SLX4–XPF mediates DNA damage responses to replication stress induced by DNA–protein interactions

Dysfunctional DNA replication ultimately causes genomic instability and development of many human diseases, including cancers (Lecona and Fernández-Capetillo, 2014; Zhang and Walter, 2014; Gaillard et al., 2015). Replication forks traveling along chromosomal DNA are often impaired by endogenous and exogenous sources of replication stress, such as collision with transcription complexes, proteins that are tightly bound to DNA, DNA lesions, repetitive DNA sequences, starvation of deoxynucleoside triphosphates, and oncogene activation (Kotsantis et al., 2018; Lambert and Carr, 2013; García-Muse and Aguilera, 2016; Magdalou et al., 2014). In eukaryotic cells, harmful consequences such as deleterious chromosomal rearrangements are prevented by elaborate integrated pathways of DNA damage response (DDR) to diverse replication stresses (Muñoz and Méndez, 2017; Nickoloff et al., 2017; Hills and Diffley, 2014).

Among the DDR pathways, ATR (ataxia telangiectasia mutated and Rad3 related) is a central factor for maintenance of genomic integrity during DNA replication (Saldivar et al., 2017; Toledo et al., 2017). When DNA polymerases are halted, a functional uncoupling of the polymerase and replicative helicase occurs, leading to the accumulation of RPA (replication protein A)–bound single-stranded DNA (ssDNA), which serves as a platform for recruitment and activation of ATR by activator proteins TOPBP1 (DNA topoisomerase II binding protein 1) or ETAA1 (Ewing’s tumor-associated antigen 1;Saldivar et al., 2017; Blackford and Jackson, 2017). ATR is also involved in the Fanconi anemia (FA) pathway, which is mainly activated by DNA interstrand crosslinks (ICLs; Ishiai et al., 2017; Michl et al., 2016). A key step of the FA pathway is monoubiquitination of FANCI (FA group I)–FANCD2 (FA group D2) heterodimeric (ID2) complexes by the FA core E3-ligase complex, followed by incision-dependent repair steps, including ICL unhooking, lesion bypass, homologous recombination, and nucleotide-excision repair (Duda et al., 2016; Niraj et al., 2019). The scaffold protein SLX4 (synthetic lethal of unknown function 4) interacts with structure-specific endonucleases such as XPF (xeroderma pigmentosum group F)–ERCC1 (excision repair cross-complementation group 1), MUS81 (methyl methanesulfonate and UV sensitive 81)–EME1 (essential meiotic endonuclease 1), and SLX1, and has a crucial role in the ICL-repair processes following ID2 monoubiquitination (Niraj et al., 2019; Zhang and Walter, 2014; Datta and Brosh, 2019). SLX4 also functions in the resolution of Holliday junctions and the maintenance of common fragile sites (Kim and Forsburg, 2018; Dehé and Gaillard, 2017; Minocherhomji and Hickson, 2014). In addition, ATM (ataxia telangiectasia mutated) is activated at sites of DNA double-strand breaks (DSBs) and triggers DNA repair pathways and damage checkpoint signaling (Mladenov et al., 2016; Blackford and Jackson, 2017). A replication fork is converted into a DSB by a severe impediment to DNA replication or an encounter with a single-strand break on the template DNA (Syeda et al., 2014). Breakage of a replication fork may generate a one-ended DNA break, and homology-directed repair pathways and/or a converging replication fork would recover the broken fork (Ait Saada et al., 2018; Magdalou et al., 2014; Kramara et al., 2018; Helleday, 2003).

Although the DDR pathways that act in response to replication stresses induced experimentally via exogenous sources have been investigated, the mechanistic details of DDRs associated with replication interference at endogenous “difficult-to-replicate” regions remain to be clarified. In addition to replication–transcription conflict and abnormal DNA structures, large nucleoprotein structures can be features of such regions (Gadaleta and Noguchi, 2017; Mirkin and Mirkin, 2007). Protein binding to chromatin can act as a natural replication fork barrier in eukaryotic cells, as found at the ribosomal DNA array (Dalgaard et al., 2011). Tight interactions between repetitive DNA sequences and specific binding proteins would perturb fork progression at difficult-to-replicate regions, such as centromeres and telomeres, which are linked to chromosome fragility and recombinational hotspots (Sarlós et al., 2017; Salim and Gerton, 2019; Higa et al., 2017a; Black and Giunta, 2018).

To assess DDR induced by replication forks stalling at tight DNA–protein complexes, we employed an experimental model system using bacterial lacO–LacI interactions to recapitulate a site-specific replication fork barrier at lacO arrays stably integrated into the mammalian chromosome (Jacome and Fernandez-Capetillo, 2011; Beuzer et al., 2014; Kim et al., 2018). We investigated the spatiotemporal dynamics of accumulation of DDR proteins and identified novel SLX4-dependent DDR signaling at the difficult-to-replicate region. Our results also demonstrated that tight protein binding causes defective mitotic segregation of the lacO site, probably as a result of incomplete replication, and that the SLX4–ATR axis alleviates the anaphase abnormality. Although some findings for DDRs induced by lacO–LacI interactions have previously been reported (Jacome and Fernandez-Capetillo, 2011; Beuzer et al., 2014; Kim et al., 2018), our study has now provided a more comprehensive picture of a process by which human cells manage nucleoprotein obstacles ahead of replication forks to prevent chromosomal instability.

To induce LacI binding to lacO, we first transiently transfected a HA epitope–tagged LacI expression vector into asynchronous human U2OS 2–6-3 cells containing a 200-copy array of 256 lacO repeats in chromosome 1 (Janicki et al., 2004). At 24 h after transfection, fluorescence imaging demonstrated that various DDR proteins (including ATR, RPA, and FANCD2) had been recruited to lacO sites (Fig. S1 A). ChIP analysis further showed that binding of FANCD2 and TOPBP1 to lacO is specifically enhanced by LacI expression (Fig. S1 B). Furthermore, expression of LacI-induced RPA phosphorylation and FANCD2 monoubiquitination, as demonstrated by mobility shifts on SDS-PAGE (Anantha et al., 2007; Ishiai et al., 2017), as well as phosphorylation of Chk1 (checkpoint kinase 1; Fig. S1 C), indicating that LacI binding to the lacO arrays activates the ATR and FA pathways. Notably, these protein modifications were suppressed by the ATR/ATM-kinase inhibitor VE-821, which inhibited ATR, but not ATM, under the condition used here (Fig. S1, C and D).

We next established U2OS 40–2-6 cells expressing the estrogen receptor (ER)–based construct ER^T2-HA-LacI (ΔNLS), in which rapid LacI binding to lacO is induced by treatment with 4-OH-tamoxifen (4-OHT; Fig. 1 A). ChIP experiments using anti-LacI antibody further showed that in the absence of 4-OHT treatment, only a background level of LacI binds to lacO sequences, and 4-OHT treatment induces remarkable LacI binding (Fig. S2 C). In addition, the results of bromodeoxyuridine immunoprecipitation (BrdU-IP) assays showed that incorporation of BrdU at the lacO arrays is inhibited by treatment with 4-OHT, suggesting that LacI binding is an obstacle for progression of the replication fork (Fig. 1 B). As in the transiently transfected cells (Fig. S1), various DDR proteins were recruited to the LacI-bound locus 2 h after 4-OHT treatment (Fig. 1, C and D; and Fig. S2 A). We also found that components of the DSB signaling and repair pathway, including 53BP1 (p53-binding protein 1), phospho-ATM (p-ATM; Ser1981), and RAD52 (radiation sensitive 52), but not RAD51 (radiation sensitive 51), are associated with the lacO sites. ChIP analysis showed that LacI binding promotes accumulation of 53BP1 at lacO (Fig. S2 C). By immunofluorescence-based detection of BrdU-labeled DNA under nondenaturing conditions (Buisson et al., 2015), we found that ssDNA is extensively exposed at the LacI-bound lacO sites (Fig. 1 D and Fig. S2 A). Furthermore, by the use of distinct labeling methods to selectively detect nascent-strand or parental-strand ssDNA (Couch et al., 2013), we found that parental-strand ssDNA is preferentially exposed at the lacO sites (Fig. 1 D and Fig. S2 A).

Next, we examined the impacts of cell-cycle synchronization on the accumulation of DDR proteins. In G1-arrested or G0-arrested cells, the frequencies of colocalization of these proteins with the LacI focus were generally lower than in S-phase cells, although some accumulation of the DSB signaling and repair proteins p-ATM and 53BP1 was still observed in G1 cells (Fig. 1 E and Fig. S3), possibly as a result of underreplication of the lacO array (see below). Accumulation of ATR, RPA34, and FANCD2 was substantially enhanced upon release from hydroxyurea (HU)–mediated arrest in early S phase (Fig. S3 I). Taken together, these data indicate that the DDRs associated with the lacO array are induced upon blockage of fork progression by LacI binding.

We investigated the temporal dynamics of the accumulation of DDR factors to the LacI-bound lacO array in cells that had been released from HU arrest for 1–8 h and found that RPA accumulation at the array peaks ∼4–6 h after release, with accumulation of ATR and FANCD2 peaking 6–8 h after release (Fig. 2). Colocalization of p-ATM and 53BP1 with lacO peaked at 4 h, and the peak of RAD52 occurred at 6 h. In the accumulation of scaffold protein SLX4 and its associated nucleases (Kim and Forsburg, 2018), SLX4 and XPF persisted at high levels throughout the examined period of S phase, whereas colocalization of MUS81 decreased during the first 4 h, then increased to peak at 8 h. Colocalization of SLX1 gradually increased to peak at 8 h. SLX4 and XPF accumulated even in G1 phase (Fig. S3 E) and HU-arrested cells (Fig. S3 I), suggesting the possibility that part of SLX4–XPF is associated with the lacO as a difficult-to-replicate region independently of replication fork inhibition (for more details, see below).

To test the role of SLX4 in the replication stress response induced by LacI binding to lacO, expression of SLX4 was knocked down by specific siRNAs (Fig. S4 A). We confirmed that knockdown of SLX4 does not cause a reduction in the levels of XPF and MUS81 proteins (Fig. S4 A), in agreement with previous findings (Panier et al., 2019; Sarbajna et al., 2014). As shown in Fig. 3 A and Fig. S4 B, SLX4 knockdown reduced the accumulation of DDR proteins at the lacO array. In addition, knockdown experiments for SLX4-associated nucleases demonstrated that XPF, but not MUS81, is involved in the accumulation of RPA and ATR (Fig. 3, B and C; and Fig. S4, C–F). XPF knockdown neither reduced SLX4 and MUS81 protein levels (Fig. S4 C) nor impaired the accumulation of SLX1 (Fig. S4 D). In addition, MUS81 knockdown did not reduce the SLX4 and XPF protein levels (Fig. S4 E). These results suggest that SLX4 and XPF act at an early stage of the response to generate ssDNA.

We also investigated the interactions between the ATR and FA pathways. FANCD2 knockdown by siRNAs led to a significant reduction in the accumulation of ATR at lacO sites (Fig. 3 D; and Fig. S4, G and H). Inhibition of ATR-kinase activity by VE-821 significantly abrogated the accumulation of FANCD2, but not RPA (Fig. 3 E and Fig. S4 I). These results reveal that following SLX4 localization, ATR and FANCD2 are recruited interdependently to stalled replication forks.

Upon replication stress, difficult-to-replicate regions such as common fragile sites occasionally cannot be fully duplicated during S phase (Debatisse and Rosselli, 2019). Such unreplicated regions undergo mitotic DNA repair synthesis (MiDAS) in early mitosis, and some regions remain underreplicated until the next cell cycle (Minocherhomji et al., 2015; Moreno et al., 2016; Spies et al., 2019; Özer and Hickson, 2018). Here, we investigated whether MiDAS is induced on the LacI-bound lacO array. As previously described (Minocherhomji et al., 2015), we used a Cdk1 (cyclin-dependent kinase 1) inhibitor (RO-3306) to synchronize U2OS 40–2-6 cell cycles at G2 phase, with or without 4-OHT, to test the influence of LacI binding during the preceding S phase (Fig. 4, A and B). It has been proposed that FANCD2 and MUS81 accumulate at underreplicated regions in G2/M cells and protect genome stability (Chan et al., 2009; Naim et al., 2013; Ying et al., 2013; Minocherhomji et al., 2015; Duda et al., 2016). As shown in Fig. 4 C, colocalization of FANCD2 with LacI on prophase or prometaphase chromosomes was significantly greater in cells with than in those without LacI induction in S phase, suggesting that LacI binding contributes to underreplication of lacO repeats. However, 62.4% of prophase/prometaphase LacI foci showed colocalization with FANCD2 even in the absence of S-phase LacI, raising the possibility that the lacO repeat is intrinsically difficult to replicate and that it remains unreplicated regardless of S-phase LacI binding. MUS81 also accumulated on the mitotic lacO array, and S-phase LacI binding only slightly enhanced prophase/prometaphase colocalization (Fig. 4 D). In addition, we found that MiDAS, which occurs specifically in mitotic cells (not in G2-arrested cells), is activated at lacO even in the absence of S-phase LacI (Fig. 4, E–G). These data suggest that even in the absence of LacI binding, most early mitotic cells (∼60%) contain underreplicated lacO arrays, and these lesions are repaired by MiDAS in only ∼20% of the cells, leaving ∼40% of cells with underreplicated lacO arrays (see Discussion and Fig. 8). With LacI binding in the preceding S phase, ∼90% of early mitotic cells contain underreplicated lacO arrays, with a similar level of MiDAS repair, leaving ∼70% of cells with underreplicated lacO arrays.

Although MUS81-mediated cleavage of underreplicated chromosomes can induce MiDAS to complete the replication (Ying et al., 2013; Minocherhomji et al., 2015; Duda et al., 2016; Lai et al., 2017; Di Marco et al., 2017), cleavage can also result in deletions from one of the sister chromatids via end joining of the DNA breaks (Naim et al., 2013). As the capacity for MiDAS is limited, it is possible that most underreplicated LacI-bound lacO arrays in our model would not be repaired by that pathway. We therefore investigated the influence of S-phase LacI binding on mitotic segregation of the lacO array. To monitor the deletion of the lacO array, we counted the numbers of LacI foci in paired sister cells defined by symmetrical daughter nuclei in G1 phase (Fig. 5 A). We found that, even in the absence of LacI in S phase, ∼30% of G1-cell pairs contain a lacO-deleted sister (Fig. 5 B), which seems to be consistent with our evidence that (post-MiDAS) ∼40% of cells may contain underreplicated lacO arrays. Induction of LacI binding in S phase reduced the proportion of cells containing equally segregated lacO arrays and increased the frequency of lacO deletion (Fig. 5 B). These results suggest that S-phase LacI binding causes lacO deletion mediated by the end joining of MUS81-cleaved underreplicated intermediates. As our results have already demonstrated that MUS81 localizes to lacO 6–8 h after release from HU arrest (Fig. 2), deletions of the lacO arrays might occur during the late-S, G2, and M phases.

Replication stress is known to lead to the generation of ultrafine DNA anaphase bridges (UFBs), which may represent underreplication, at difficult-to-replicate regions (Sarlós et al., 2017; Fragkos and Naim, 2017; Bizard and Hickson, 2018). Therefore, we analyzed mitotic-chromosome abnormalities resulting from replication stress at the lacO array. As shown in Fig. 6, A and B, we found that UFB-like DAPI-negative abnormal structures (here referred to as anaphase abnormal lacO) are induced by LacI binding in S phase. These structures were located between the two sets of daughter chromosomes and were associated with RPA and PICH (Plk1-interacting checkpoint helicase), a key marker of UFBs (Fig. 6 A; Bizard and Hickson, 2018).

It has been proposed that underreplicated regions are converted to DNA lesions during mitosis, which are shielded by 53BP1 nuclear bodies (NBs) in the daughter cells (Moreno et al., 2016; Lukas et al., 2011; Spies et al., 2019). We examined whether LacI binding induces such 53BP1 NBs by counting the sister cells that contain symmetrical LacI foci associated with 53BP1, and calculating the frequency of such 53BP1-NB-positive cells among the cells in which lacO is equally segregated (Fig. 6, C and D). We found that LacI binding in S phase significantly enhances the generation of twin 53BP1-NB–positive cells (Fig. 6 E), indicating that LacI-induced replication stress leads to formation of anaphase abnormal lacO structures and DNA lesions that are protected in 53BP1 NBs, through the chromosomal segregation of incomplete replication intermediates.

The ATR pathway ensures sufficient DNA replication before the onset of mitosis, and regulates genome stability (Eykelenboom et al., 2013; Casper et al., 2002). In addition, as we demonstrated above, SLX4 contributes to the accumulation of the DDR factors at the lacO sites (Fig. 3 A; and Fig. S4, A and B). We therefore inhibited ATR and SLX4 with VE-821 and siRNA-mediated silencing, respectively, and examined whether the inhibition affects the chromosomal stability of the lacO arrays (Fig. 7). The results showed that inhibition of ATR and SLX4 significantly increases the frequency of anaphase abnormal lacO structures that are induced by the S-phase LacI binding (Fig. 7), demonstrating that the SLX4–ATR axis contributes to the complete replication and stability of the replication-stressed LacI-bound lacO array.

Tight DNA–protein interactions impede progression of replication machinery and cause replication stress (Jacome and Fernandez-Capetillo, 2011; Beuzer et al., 2014; Hizume et al., 2018). By taking advantage of the inducible LacI–lacO interaction, we have characterized in detail the spatiotemporal regulation of the recruitment of various DDR factors in response to a site-specific replication fork barrier on the human chromosome. DNA replication in S phase enabled accumulation of components of the ATR and FA pathways, among which SLX4–XPF functioned as an upstream factor to promote recruitment of DDR proteins, including ATR and FANCD2. S-phase LacI binding caused abnormal mitotic segregation and DNA lesions that remained until the next cell cycle, and the SLX4–ATR axis facilitated complete replication of the LacI-bound lacO arrays. Notably, even in the absence of LacI, the lacO array was underreplicated, probably because of its nature as a long repetitive sequence, in which it resembles microsatellites, telomeres, and several common fragile sites (Gadaleta and Noguchi, 2017; Kaushal and Freudenreich, 2019; Mirkin and Mirkin, 2007; Tsao and Eckert, 2018). In our ER-LacI induction system, levels of LacI binding to lacO sequences in the absence of 4-OHT are ∼25-fold lower than those in the presence of 4-OHT, and the signals are similar extent to those obtained with control IgG (Fig. S2 C). Although we cannot exclude a possibility of very small leaks, we consider that leakiness is not a major reason for the lacO instability in the absence of 4-OHT. Our data also clearly showed that LacI binding further exacerbates the replication fork block at the lacO array. We also found accumulation of DSB markers 53BP1 and p-ATM in G1 phase. Underreplicated regions can form heritable DNA lesions that are shielded by 53BP1 in G1 (Spies et al., 2019; Moreno et al., 2016; Lukas et al., 2011), so these results may reflect the persistence of the underreplicated lacO array.

Our results suggest that the collision between the DNA replication fork and LacI–lacO complex generates an excess of parental-strand ssDNA. LacI–lacO interactions have been suggested to inhibit the progression of the CDC45 (cell division cycle 45)–MCM (minichromosome maintenance)–GINS (go-ichi-ni-san) helicase ahead of the polymerase complex (Dewar et al., 2015; Hizume et al., 2018), which is unlikely to cause the accumulation of excess ssDNA at the stalled fork. Instead, unidentified exonucleases may be responsible for nucleolysis of nascent DNA (Pasero and Vindigni, 2017; Liao et al., 2018). Although MRE11 endo/exonuclease has been implicated in the degradation of replication forks stalled by HU (Ray Chaudhuri et al., 2016; Mijic et al., 2017; Schlacher et al., 2011), our results indicate that inhibition of MRE11 does not affect colocalization of RPA with LacI foci (Fig. S4 J), so other exonucleases or helicases may promote the generation of ssDNA at the fork stalled by the DNA–protein complex. For appearance of a small amount of nascent-strand ssDNA, fork reversal could potentially be involved (Pasero and Vindigni, 2017; Liao et al., 2018). We propose that SLX4 and XPF promote DNA cleavage at the stalled fork and trigger generation of ssDNA, probably through the mechanisms discussed above to recruit the effector factors such as ATR and FANCD2, which contribute to genome stability during S phase (Fig. 8). This response seems similar to the fork breakage induced by XPF in the early stage of HU-induced replication stress, which contributes to chromosome stability (Bétous et al., 2018).

FANCD2 has been suggested to function upstream of SLX4 to repair ICL-blocked replication forks (Niraj et al., 2019; Datta and Brosh, 2019; Zhang and Walter, 2014). However, we found that silencing of FANCD2 results in increased colocalization of SLX4 with LacI foci (Fig. S4 K), possibly in response to an increase in replication fork stress. Branched DNA structures may contribute to the recruitment of SLX4 to LacI-stalled replication forks (Muñoz et al., 2009; Svendsen et al., 2009; Fekairi et al., 2009). We also found that part of SLX4–XPF accumulates on the lacO array even in G1-phase or HU-arrested cells (Fig. S3, E and I). It is possible that certain DNA structures and/or DNA lesions derived from underreplication of the lacO array may promote the recruitment of SLX4–XPF. Further study is needed to elucidate the mechanisms underlying SLX4–XPF recruitment.

Our results demonstrate that RAD52, but not RAD51, is recruited to LacI foci. The lack of accumulation of RAD51 at the lacO site is consistent with previous findings (Luijsterburg et al., 2017). However, RAD51-mediated homology-directed repair pathways have been suggested to operate in response to fork stalling induced by 6×Ter–Tus complexes in mouse embryonic stem cells (Willis et al., 2014, 2018). Because our experimental system used much longer repetitive DNA than 6×Ter, it is possible that the amount of repeated sequence or DNA–protein complex affects the mechanism underlying the response to replication fork stalling. RAD52 has been implicated both in RAD51-dependent homology-directed repair and RAD51-independent subpathways such as single-strand annealing and break-induced replication–like processes (Bhowmick et al., 2016; Sotiriou et al., 2016; Kramara et al., 2018; Costantino et al., 2014; Verma et al., 2019; Zhang et al., 2019). At the lacO array, RAD52 may facilitate annealing of complementary ssDNA to repair the cleaved replication fork. Rebuilding the fork at the same genomic position would be important to preserve the copy number of the repeat sequences. However, it is also possible that RAD52-mediated single-strand annealing at a different genomic position causes chromosomal rearrangement. It will be important in the future to establish the significance of RAD52 for the replication and repeat stability of the lacO locus.

Our data demonstrate that FANCD2 and MUS81 accumulate at the lacO array in early mitosis, suggesting that, in spite of the activation of the DDR in S phase, a portion of the lacO array remains underreplicated (Fig. 8). MUS81 associates with SLX4 in a Cdk1-dependent manner and promotes DNA cleavage and subsequent MiDAS at underreplicated regions (Naim et al., 2013; Bhowmick et al., 2016; Duda et al., 2016). Although a small proportion (∼20%) of the lacO arrays undergo MiDAS, our data also suggest that underreplicated lacO arrays that are cleaved by MUS81, but not repaired by MiDAS, may result in deletion of the locus. In addition, some underreplicated lacO arrays escape from cleavage and result in anaphase abnormal lacO structures. Like single-stranded UFBs, the structures are DAPI negative and coated with RPA and PICH. In addition, a central discrete LacI-positive focus (probably consisting of double-stranded lacO) is observed in the structure. It has been suggested that during chromosome segregation, underreplicated dsDNA is converted into ssDNA lesions (Moreno et al., 2016). We therefore speculate that both single-stranded and LacI-bound double-stranded lacO derived from underreplicated lacO sequences is retained between the separating anaphase chromosomes and observed as the anaphase abnormal structures. For detailed mechanism, further study is needed. On the other hand, Jacome et al. reported that in murine 3T3^lacO cells, transfection of LacI (for 48 h) induces a DAPI-positive, PICH-coated anaphase bridge (Jacome and Fernandez-Capetillo, 2011). The difference could be due to difference in the experimental systems such as the difference in time range for LacI induction. Chromosome segregation converts underreplicated regions to DNA lesions, which attract 53BP1 in G1 daughter cells. S-phase LacI binding exacerbates replication stress and increases the frequencies of lacO deletion, anaphase abnormality, and 53BP1 NBs. Furthermore, our results support a role for SLX4 and ATR in the completion of replication of LacI-bound lacO arrays and prevention of anaphase abnormalities.

In summary, the results of the present study suggest that SLX4–XPF has a key role in generation of excess ssDNA at replication forks arrested by protein–DNA complexes. This mechanism could enable activation of DDR at the blocked replication fork without uncoupling of replicative helicase and polymerase complexes. SLX4–XPF–mediated cleavage may also be a mechanism to resolve topological stress at blocked replication forks. The human genome contains many repetitive elements, which are generally embedded in heterochromatic regions. Unscheduled fork stalling and alteration of repeat length are associated with genomic instability and human disease (Gaillard et al., 2015; Lecona and Fernández-Capetillo, 2014; Padeken et al., 2015). Although further studies are required to determine whether defects in SLX4-dependent DDR signaling are relevant to repeat instability, our findings will provide a novel framework for understanding how such repeat sequences are stably duplicated.

U2OS 2–6-3 (Janicki et al., 2004; Higa et al., 2017b) and U2OS 40–2-6 ER-LacI cells were cultured in DMEM supplemented with 8% FCS. A U2OS 40–2-6 clone overexpressing ER^T2-HA-LacI was established as follows. U2OS 2–6-3 cells were cotransfected with pAAVS-CMV-HA-ERT2-LacI (ΔNLS) and AAVS1 T2 CRIPR in pX330 (from Dr. Masato Kanemaki, National Institute of Genetics, Mishima, Japan; Addgene; #72833; Natsume et al., 2016) using Lipofectamine 2000 reagent (Invitrogen), and selected with puromycin (0.5 µg/ml). Expression and 4-OHT-inducible nuclear accumulation of ER^T2-HA-LacI protein were confirmed by immunofluorescence analysis (Fig. 1 A). Positive clones were subjected to secondary cloning, and isolated positive clone 40–2-6 was used for further experiments. Cells were cryopreserved in small aliquots and were passaged in the medium for less than 4 mo after resuscitation. We confirmed that the copy number of lacO sequences was not significantly changed during cell culture for one month or during siRNA-mediated knockdown of SLX4 for 48 h (data not shown).

pSV40-HA-LacI has previously been described (Higa et al., 2017b). pET28a-His-T7-LacI was generated by inserting LacI cDNA from pSV40-GFP-LacI (Higa et al., 2017b) into pET28a (Novagen).

To prepare pSV40-HA-ER^T2-LacI (ΔNLS), a DNA fragment encoding ER^T2 was subcloned from pCAG-CreERT2 (from Dr. Connie Cepko, Harvard Medical School, Boston, MA; Addgene #14797; Matsuda and Cepko, 2007) into pSV40-HA-LacI using the In-Fusion HD cloning system (Clontech). The following primers were used: 5′-GAT​TAC​GCC​CTG​TAC​ATA​TCT​GCT​GGA​GAC​ATG​AGA​GC-3′ (HA-ER^T2 forward) and 5′-GGA​ATT​CCT​TGT​ACA​AAG​CTG​TGG​CAG​GGA​AAC-3′ (ER^T2-LacI reverse). Subsequently, pSV40-HA-ER^T2-LacI was digested with NotI (New England BioLabs) and Eco81I (Takara Bio), treated with Klenow Fragment (Takara Bio), and self-ligated using the DNA Ligation Kit <Mighty Mix> (Takara Bio) to generate pSV40-HA-ER^T2-LacI (ΔNLS).

pAAVS CMV-HA-ER^T2-LacI (ΔNLS) was generated by inserting HA-ER^T2-LacI (ΔNLS) DNA into pMK231 (AAVS1 CMV-MCS-PURO; from Dr. Masato Kanemaki; Addgene; #105924; Okumura et al., 2018) using the In-Fusion HD cloning system. The following primers were used: 5′-GGA​CTC​AGA​TCT​GCC​ACC​ATG​TAC​CCC-3′ (AAVS1 CMV HA Forward Primer) and 5′-AGC​TCG​AGA​TCT​TCA​TCG​GGA​AAC​CTG​TCG-3′ (LacI rev pMK231 Reverse Primer).

To identify colocalization of RPA34, RPA70, ATR, FANCD2, XPF, PICH, and p-ATM with LacI, cells were fixed with 3.7% formaldehyde (Nacalai Tesque) in phosphate-buffered saline (PBS) for 15 min and permeabilized with 0.1% Triton X-100 in PBS for 10 min. For 53BP1, cells were fixed with 3.7% formaldehyde in PBS for 15 min and permeabilized with 0.1% Triton X-100 in mCSK buffer (10 mM Pipes-NaOH, pH 6.8, 100 mM NaCl, 300 mM sucrose, 1 mM EGTA, and 1 mM MgCl₂) for 10 min. For SLX1, RAD52, and RAD51, cells were preextracted with 0.1% Triton X-100 in PBS for 10 min before fixation with formaldehyde. For TOPBP1, cells were preextracted with 0.1% Triton X-100 in mCSK buffer for 3 min before fixation with formaldehyde. For MUS81, cells were preextracted with 0.5% Triton X-100 in mCSK buffer for 5 min before fixation with formaldehyde. For staining of SLX4 and BrdU, cells were fixed with chilled 100% methanol for 10 min and then permeabilized with 0.1% Triton X-100 in PBS for 10 min.

After washing with PBS, cells were incubated with appropriate primary antibodies in PBS supplemented with 1% fetal calf serum, followed by incubation with secondary antibodies, and then counterstained with DAPI. Cells were mounted in Fluoro-KEEPER Antifade Reagent (Nacalai Tesque) and analyzed using the BZ-X700 all-in-one fluorescence microscope (Keyence) with CFI Plan Apo λ 40×/0.95 NA objective (Nikon). Microscopic images were acquired by single image captures (5–20 fields for each sample) using the BZ-X viewer software (Keyence). For calculation of colocalization frequencies, at least 60 nuclei with a single prominent LacI focus were scored for each experiment. Frequencies of nuclei in which DDR proteins form a focus on the LacI focus were then manually scored.

To detect DNA synthesis in mitosis, cells were released from RO-3306–induced G2 synchronization and pulse-labeled with 10 µM BrdU for 20 min. Cells were then fixed and permeabilized simultaneously in PTEMF buffer (20 mM Pipes, pH 6.8, 10 mM EGTA, 0.2% Triton X-100, 1 mM MgCl₂, and 4% formaldehyde; Bhowmick et al., 2016) for 20 min. Samples were incubated with rabbit anti-LacI antibody and then with secondary antibody. Subsequently, cells were refixed with 100% methanol for 10 min and then incubated with 1 M HCl for 15 min to denature DNA. Samples were incubated with mouse anti-BrdU antibody (clone 3D4) and then with secondary antibody, followed by counterstaining with DAPI.

ChIP assays were performed as described previously (Higa et al., 2017b). Briefly, cells were fixed with 1% formaldehyde and lysed in SDS lysis buffer (1% SDS, 10 mM EDTA, and 50 mM Tris-HCl, pH 8.0). Chromatin was sonicated into ∼0.5-kb fragments. 5 μg antibodies or control IgG was added to aliquots of each sample, and the samples were incubated for overnight at 4°C. Immunoprecipitates were then collected with 30 µl Dynabeads (Invitrogen). The beads were washed once with RIPA buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, and 0.1% SDS), twice with RIPA buffer containing 500 mM NaCl, twice with LiCl wash buffer (10 mM Tris-HCl, pH 8.0, 250 mM LiCl, 1 mM EDTA, 1% NP-40, and 0.5% sodium deoxycholate), and once with Tris-EDTA buffer (10 mM Tris, pH 8.0, and 1 mM EDTA). The immunocomplexes were eluted with ChIP elution buffer (100 mM NaHCO₃ and 1% SDS) at RT for 15 min. After cross-linking reversal, phenol/chloroform extraction, and ethanol precipitation, DNA was dissolved in Tris-EDTA buffer.

BrdU immunoprecipitation assays were performed as described previously, with some modifications (Verma et al., 2018). Briefly, U2OS 40–2-6 ER-LacI cells were treated with 1 µM 4-OHT for 80 min. Cells were labeled with BrdU for the last 20 min. Genomic DNA was purified and sonicated to give ∼0.5-kb fragments. Anti-BrdU antibody (clone B44, 3 µg) was added to an aliquot of genomic DNA (4 µg) and incubated overnight at 4°C. Immunoprecipitates were collected with 20 µl Dynabeads. The beads were washed three times with 0.0625% Triton X-100 in PBS, and once with Tris-EDTA buffer. The immunocomplexes were eluted with ChIP elution buffer at RT for 15 min. After phenol/chloroform extraction and ethanol precipitation, DNA was dissolved in Tris-EDTA buffer.

PCR reactions were performed as described previously (Sugimoto et al., 2011; Higa et al., 2017b) using the iCycler iQ real-time PCR detection system (Bio-Rad) or the CFX96 Touch system (Bio-Rad). TB Green Premix Ex Taq II (Takara Bio) was used according to the manufacturer’s instructions. The following primers were used to detect the lacO sequence: 5′-TAG​AGG​CGC​CGA​ATT​GCA​CA-3′ and 5′-GCC​ACA​AAT​TGT​TAT​CCG​CTC​A-3′. The cycling parameters were as follows: 1 min at 95°C and then 50 cycles at 95°C for 30 s, at 64.2°C for 30 s, and at 72°C for 30 s. For the LMNB2 origin, following primers were used: 5′-GGC​TGG​CAT​GGA​CTT​TCA​TTT​CAG-3′ and 5′-GTG​GAG​GGA​TCT​TTC​TTA​GAC​ATC-3′. The cycling parameters were as follows: 1 min at 95°C; five cycles at 95°C for 30 s, 66.9°C for 30 s, and 72°C for 30 s; five cycles at 95°C for 30 s, 64.9°C for 30 s, and 72°C for 30 s; and 50 cycles at 95°C for 30 s, 62.9°C for 30 s, and 72°C for 30 s.

The following drugs were used: 4-OHT (Abcam), RO-3306 (Sigma), HU (Sigma), lovastatin (LKT Laboratories), VE-821 (Haoyuan Chemexpress), KU-55933 (Sigma), and mirin (Sigma).

For G1-phase synchronization, cells were synchronized with 40 µM lovastatin for 40 h and analyzed. For S-phase synchronization, cells were synchronized with 2.5 mM HU for 18 h and released into fresh medium. For the experiment shown in Fig. 1 E, at 4 h after release from HU, cells were further treated with 1 µM 4-OHT for 2 h and then analyzed. For G2/M-phase synchronization, cells were synchronized in the late G2 phase by incubation with 7 µM RO-3306 for 20 h (Figs. 4, 5, 6, and,7 A) or 24 h (Fig. 7 B) and then released into fresh medium.

Cell-cycle analysis was performed essentially as described previously (Yokoyama et al., 2019; Sugimoto et al., 2011). For the experiments shown in Fig. S3, A and F, cells were treated with a CycleTEST PLUS DNA reagent kit (Becton Dickinson). For the experiments shown in Fig. 4 B and Fig. S3 C, cells were suspended in PBS containing 0.1% TritonX-100 and RNase (10 µg/ml) and then stained with propidium iodide (40 µg/ml). Cells were then analyzed with a FACSCalibur flow cytometer (BD Bioscience).

U2OS 40–2-6 ER-LacI cells were transfected with 12 nM siRNA duplexes using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instructions. siRNA oligonucleotides with the following sequences (sense strand) were synthesized (IDT or Dharmacon): siGL2 (5′-CGUACGCGGAAUACUUCGAdTdT-3′), siGFP (5′-ACCCUGAAGUUCAUCUGCACCACdCdG-3′),

siLuci (5′-GGUUCCUGGAACAAUUGCUUUUAdCdA-3′), siSLX4-3 (5′-AAA​CGU​GAA​UGA​AGC​AGA​AUU-3′; Duda et al., 2016), siSLX4-4 (5′-CGG​CAU​UUG​AGU​CUG​CAG​GUG​UU-3′; Duda et al., 2016), siXPF1-1 (5′-CUUUUCUAAAAGCUAGAUCAGCAdTdT-3′), siXPF1-2 (5′-CUUGACUGAUAGAAUACCUUCAGdAdT-3′), siMUS81-1 (5′-GGGGCAUUGCAGCUUGGAAUCUAdTdT-3′), siMUS81-2 (5′-AGAGCAUGGUUCCGUCCACAACCdTdC-3′), siFANCD2-1 (5′-GAUGUUCUGAGCUUACUGGAAACdCdT-3′), and siFANCD2-2 (5′-GGAUUUACCUGUGAUAAUAAAGUdTdC-3′).

Rabbit anti-LacI antibody was obtained by immunizing rabbits with a bacterially produced His-T7-LacI protein. Mouse antibodies against human RPA34 and RPA70 were provided by Dr. Murakami (Hokkaido University, Sapporo, Japan).

Other antibodies used in this study were as follows: ATR (Santa Cruz; Goat, sc1887), ATRIP (Cell Signaling; Rabbit, #2737), RPA34 (Calbiochem; Mouse, NA19L), TOPBP1 (Abcam; Rabbit, ab2402), FANCD2 (Novus; Rabbit, NB100-182), LacI (clone 9A5, Merck, Mouse, #05-503), HA (clone 3F10; Roche; Rat, 11867423001), Chk1 (Santa Cruz; Mouse, sc-8408), phosphorylated Chk1 (Ser345; Cell Signaling; Rabbit, #2348S), phosphorylated ATM (Ser1981; Rockland; Mouse, 200–301-400; Abcam; Rabbit, ab81292), 53BP1 (Novus; Rabbit, NB100-904), RAD51 (BioAcademia; Rabbit, 70–012), RAD52 (BioAcademia; Rabbit, 70–015), BrdU (clone 3D4 and B44; BD PharMingen; Mouse, 555627 and 347580, respectively), Cyclin A (Santa Cruz; Rabbit, sc-596), Cyclin E (Santa Cruz; Mouse, sc-247), SLX4 (Novus; Rabbit, NBP1-28680), XPF (Bethyl; Rabbit, A301-315A), MUS81 (Abcam; Mouse, ab14387), SLX1 (Atlas Antibodies; Rabbit, HPA047038), PICH (clone 142–26-3, Merck, Mouse, 04–1540), goat normal IgG (Chemicon; PP40), mouse normal IgG (Southern Biotech; 0107–01), and rabbit normal IgG (Dako; X0903).

Specificity of the immunofluorescence signal derived from the anti-SLX1 antibody we used was confirmed by the finding that the siRNA-mediated knockdown significantly reduces the accumulation of SLX1 at the lacO array (data not shown). However, this antibody was not applicable to detect SLX1 protein in our immunoblotting experiments.

The following secondary antibodies were used: HRP-conjugated rabbit anti-mouse IgG (H + L; Invitrogen; 61–6520), HRP-conjugated goat anti-rabbit IgG (H + L; Invitrogen; 65–6120), Alexa488-conjugated goat anti-rabbit IgG (Invitrogen; A11034), Alexa488-conjugated goat anti-mouse IgG (Invitrogen; A11029 and A11001); Alexa594-conjugated donkey anti-rabbit IgG (Invitrogen; A21207), CF488A-conjugated goat anti-rabbit IgG (Biotium; 20019), CF488A-conjugated donkey anti-goat IgG (Biotium; 20016), CF594-conjugated donkey anti-rat IgG (Biotium; 20159), and CF594 goat anti-mouse IgG (Biotium; 20111).

Immunoblotting was performed as previously described (Fujiyama et al., 2020). Antibody binding was visualized with HRP-conjugated secondary antibodies using EzWestLumi One (ATTO) or Chemi-Lumi One Super (Nacalai Tesque). Chemiluminescent signals were captured via a cooled-CCD (charge-coupled device) camera direct detection system (Aisin; LumiVision Imager).

Unless otherwise stated, quantitative data are represented as the mean ± SD of three or more independent experiments. For Fig. 1, D and E; Fig. 3, C–E; Fig. 4, C, D, and G; Fig. 5 B; Fig. 6, B and E; Fig. S3, E and I; and Fig. S4, B, D, F, H, J, and K; data were calculated from the sum scores of two independent experiments. As described in the figures, statistical significance was assessed by using an unpaired two-tailed Student’s t test or a χ² test. P values < 0.05 were considered statistically significant. For the t tests, data distribution was assumed to be normal but this was not formally tested. For qualitative data and semiquantitative data, a representative image from multiple independent experiments is shown; for all such figures, essentially the same results were obtained in the multiple independent experiments.

Fig. S1 shows the recruitment of DDR proteins to lacO arrays and the activation of DDR after transient expression of HA-LacI in U2OS 2–6-3 cells. Fig. S2 shows the recruitment of DDR proteins to lacO arrays in U2OS 40–2-6 ER-LacI cells. Fig. S3 shows cell-cycle profiles of U2OS 40–2-6 ER-LacI cells and additional experiments to investigate the effect of cell-cycle synchronization. Fig. S4 shows additional data for silencing of SLX4, XPF, MUS81, and FANCD2, and the SLX4–XPF-dependent accumulation of DDR factors.

We thank Dr. Masato Kanemaki (National Institute of Genetics, Mishima, Japan) for plasmids and Dr. Yota Murakami (Hokkaido University, Sapporo, Japan) for antibodies. We appreciate the technical support received from the Research Support Center, Research Center for Human Disease Modeling, Kyushu University Graduate School of Medical Sciences.

This investigation was supported in part by the Japan Society for the Promotion of Science (KAKENHI grant JP15K18478 to K. Yoshida). K. Yoshida is also supported by research grants from the Fukuoka Foundation for Sound Health Cancer Research Fund and the Mochida Memorial Foundation for Medical and Pharmaceutical Research.

The authors declare no competing financial interests.

Author contributions: K. Yoshida and M. Fujita designed and managed the project. R. Ishimoto, Y. Tsuzuki, T. Matsumura, S. Kurashige, K. Enokitani, K. Narimatsu, M. Higa, and K. Yoshida conducted the experiments and data analysis. N. Sugimoto and M. Fujita supervised the experiments and data analysis. R. Ishimoto, K. Yoshida, and M. Fujita wrote the manuscript.