The RIF1-long splice variant promotes G1 phase 53BP1 nuclear bodies to protect against replication stress

Human cells lacking RIF1 are highly sensitive to replication inhibitors, but the reasons for this sensitivity have been enigmatic. Here, we show that RIF1 must be present both during replication stress and in the ensuing recovery period to promote cell survival. Of two isoforms produced by alternative splicing, we find that RIF1-Long alone can protect cells against replication inhibition, but RIF1-Short is incapable of mediating protection. Consistent with this isoform-specific role, RIF1-Long is required to promote the formation of the 53BP1 nuclear bodies that protect unrepaired damage sites in the G1 phase following replication stress. Overall, our observations show that RIF1 is needed at several cell cycle stages after replication insult, with the RIF1-Long isoform playing a specific role during the ensuing G1 phase in damage site protection.

RIF1 is also implicated in protecting cells from replication stress (Buonomo et al., 2009;Kumar and Cheok, 2014;Mazouzi et al., 2016). Replication stress can be induced by various conditions including drugs such as Aphidicolin, which interrupts replication fork progression by inhibiting the replicative DNA polymerases a, d and e (Syvaoja et al., 1990). Replication stress leads to genomic instability, mutation and eventually disease (Burrell et al., 2013;Kerzendorfer et al., 2013;Ogi et al., 2012), so understanding the cellular response is central for understanding accurate genome duplication and the action of replication inhibitors as anti-cancer drugs (Feng et al., 2003;Imai et al., 2016). RIF1-deficient cells are acutely sensitive to replication stress, in fact appearing to be more sensitive to replication inhibitors than to double-strand break-inducing agents (Buonomo et al., 2009), suggesting that protection from stress is a critical RIF1 function. Several roles have been described for RIF1 in replication control. RIF1 acts as a Protein Phosphatase 1 (PP1) 'substrate-targeting subunit' (Peti et al., 2013) that suppresses replication origin initiation by directing PP1 to dephosphorylate the MCM replicative helicase complex (Alver et al., 2017;Davé et al., 2014;Gnan et al., 2020;Hiraga et al., 2017;Hiraga et al., 2014;Mattarocci et al., 2014). RIF1 moreover stimulates origin licensing during G1 phase, and protects replication forks from unscheduled degradation (Garzó n et al., 2019;Hiraga et al., 2017;Mukherjee et al., 2019). However, whether deficiency in these functions accounts for the replication stress sensitivity of cells lacking RIF1 has remained unclear (Buonomo et al., 2009;Feng et al., 2003).
RIF1 also acts in mitosis to maintain genomic stability. During anaphase, RIF1 is recruited to ultrafine anaphase bridges (UFBs), along with the BLM and PICH proteins that ensure proper chromosome segregation (Hengeveld et al., 2015). UFBs are believed to correspond to unresolved topological entanglements that escape checkpoint surveillance and persist into mitosis (Bergoglio et al., 2013;Bhowmick et al., 2016). Unresolved DNA damage that passes to daughter cells leads to the formation of 53BP1 nuclear bodies during G1 phase, which protect the damaged DNA (Bruhn et al., 2014;Lukas et al., 2011;Moreno et al., 2016). RIF1 has also recently been described as functioning at the midbody during cytokinesis (Bhowmick et al., 2019).
The human RIF1 transcript undergoes alternative splicing producing two protein isoforms: a long variant of 2472 amino acids ('RIF1-L'), and a short variant ('RIF1-S') which lacks 26 amino acids close to the C-terminus of the protein (Xu and Blackburn, 2004). RIF1-S was reported to be more abundant in various cancer cell lines (Xu and Blackburn, 2004), hinting at distinct effects of the isoforms. Although RIF1-L was designated the canonical form, studies using cloned RIF1 have predominantly used RIF1-S (Batenburg et al., 2017;Escribano-Díaz et al., 2013;Xu et al., 2010;Xu and Blackburn, 2004), without testing for distinct functions of the isoforms.

Results
Analysing fluorescent degron-tagged RIF1 reveals highly dynamic cell cycle localisation We aimed to understand how RIF1 guards against replication stress. First, we confirmed in a colony formation assay (CFA) that HEK293 cells depleted for RIF1 are sensitive to the DNA polymerase inhibitor Aphidicolin ( Figure 1A,B). HCT116 cells deleted for both copies of RIF1 were also sensitive ( Figure 1C). Doses of Aphidicolin were designed to slow replication fork progression and induce replication stress, as opposed to blocking the cell cycle (Buonomo et al., 2009). These results imply a specific role for RIF1 in protecting cells under replication stress conditions.
Since RIF1 functions at various cell cycle stages, we explored when RIF1 is needed to maintain cell proliferation following replication stress. Specifically, we tested if RIF1 function is required during DNA replication stress, after its occurrence, or both during and after stress. Using auxin-inducible degron (AID) technology we constructed a cell line allowing rapid depletion and re-expression of RIF1 at different phases of the cell cycle (Natsume et al., 2016;Nishimura et al., 2009). In an HCT116-based cell line carrying the auxin-responsive degron recognition protein OsTIR1 under doxycycline (DOX) control, we N-terminally tagged both RIF1 genomic copies with a degron-Clover construct, termed 'mAC', consisting of a mini-auxin-inducible degron and monomer Clover (a derivative of GFP) ( Figure 1D; Yesbolatova et al., 2019). The expressed construct remains under control of the endogenous RIF1 promoter. Western blot analysis indicated that expression levels of mAC-RIF1 in the absence of Auxin were similar to those of endogenous untagged RIF1 ( Figure 1E). Treatment for 24 hr with DOX and Auxin led to near-complete degradation of RIF1, as visualised by western blotting ( Figure 1E) and microscopy ( Figure 1F). Using flow cytometry analysis of the RIF1-fused mClover tag, we established suitable concentrations of DOX and Auxin to allow effective depletion ( Figure 1-figure supplement 1A,B). Degradation was largely complete after 3 hr of Auxin treatment ( Figure 2-figure supplement 1A), whilst expression was restored to almost normal levels 5 hr after Auxin removal (Figure 2-figure supplement 1B). Together, these results confirm the construction of a cell line allowing rapid depletion and re-expression of RIF1.
We visualised mAC-RIF1 based on Clover fluorescence, in the first study of the dynamic behaviour of tagged endogenous RIF1 in living cells. In live-cell imaging experiments we observed a pattern of numerous mAC-RIF1 foci throughout S phase nuclei ( Figure 1G, top row), often with 3-6 prominent foci superimposed on a pattern of more numerous smaller foci, consistent with previous Colony Formation Assay (CFA) confirming Aphidicolin sensitivity of HEK293 cells treated with siRIF1. Plot shows mean and range of technical triplicates. ***p<0.001. (C) CFA testing Aphidicolin sensitivity of HCT116 RIF1-KO cells. 'RIF1' cell line is HCT116 mAC-RIF1. Plot shows mean and range from two biological replicates (each performed in technical triplicate). (D) Structure of auxin-inducible degron (AID)-tagged RIF1 construct (mAC-RIF1), located at both endogenous RIF1 loci in HCT116 cells carrying the auxin-responsive F-box protein Oryza sativa TIR1 (OsTIR1) under DOX control. The RIF1 gene is fused to a tag containing a self-cleaving P2A peptide, hygromycin resistance marker, mini-auxin-inducible degron (mAID) and monomer Clover (mClover) protein. After self-cleavage at the P2A, RIF1 protein is expressed as N-terminal in-frame fusion with mAID and Clover. (E) Confirmation of mAC-RIF1 protein degradation. Cells were incubated with 2 mg/ml DOX and 500 mM Auxin for 24 hr, then protein extracts analysed by western blotting with antibody against RIF1. Tubulin is shown as a loading control. Testing of drug concentrations is shown in Figure 1-figure supplement 1. (F) mAC-RIF1 degradation assessed by microscopy. mAC-RIF1 cells were treated with 2 mg/ml DOX and 500 mM Auxin for 24 hr. DNA was stained with SiR-DNA (magenta). Scale bar = 10 mm. (G) Examples of mAC-RIF1 localisation at different cell cycle stages. DNA stained with SiR-DNA. Scale bar = 10 mm. (H) mAC-RIF1 co-localises with BLM at UFBs but not with 53BP1 or FANCD2 repair proteins. Fixed cells were stained with the above-mentioned antibodies. Scale bar = 10 mm. The online version of this article includes the following figure supplement(s) for figure 1: immunofluorescence studies (Buonomo et al., 2009;Xu and Blackburn, 2004;Yamazaki et al., 2012). In prometaphase mAC-RIF1 exhibits a novel, circumpolar localisation pattern similar to that described for kinetochores ( Figure 1G, second row, showing polar view of condensed chromosomes) (Magidson et al., 2015). At anaphase, even in unperturbed cells, mAC-RIF1 was sometimes observed at bridge structures ( Figure 1G, third row) where it colocalised with the bridge marker protein BLM, but not always with conventional DNA dyes ( Figure 1H, bottom panel) (Barefield and Karlseder, 2012;Chan et al., 2007;Hengeveld et al., 2015). At telophase, mAC-RIF1 formed multiple small foci associated with separated chromosomes ( Figure 1G, bottom row), as described previously (Xu and Blackburn, 2004;Yamazaki et al., 2012). Time-lapse imaging of HCT116 mAC-RIF1 cells containing an mCherry-tagged PCNA revealed intense RIF1 foci that accumulated through S phase and G2 (Video 1), but disappeared by metaphase (Video 2). RIF1 signal was absent for a short period at metaphase, quickly followed by reappearance of numerous smaller foci on telophase chromosomes.
The mAC-RIF1 fusion therefore retains the localisation and functional characteristics of the endogenous RIF1 protein. Its highly dynamic behaviour as revealed by this live-cell study indicates that RIF1 functions in several cell cycle phases to maintain chromosome stability.

RIF1 is needed during and after replication stress to promote cell proliferation
Prolonged auxin-induced degradation of RIF1 caused sensitivity to Aphidicolin, comparable to an HCT116 RIF1-knockout cell line (RIF1 KO) ( Figure 3B lane 4). To investigate when RIF1 function is required to protect against the effects of Aphidicolin, we performed a 'staged depletion' experiment ( Figure 2B). We synchronised cells first in G1 phase with Lovastatin (Javanmoghadam- Kamrani and Keyomarsi, 2008;Rao et al., 1999), and 8 hr after release from Lovastatin added Aphidicolin to induce replication stress ( Figure 2B, upper timeline). The mitosis-specific CDK inhibitor RO-3306 (Vassilev et al., 2006) was simultaneously added, to induce a temporary G2 arrest and prevent cells from proceeding into mitosis. After 28 hr, Aphidicolin and RO-3306 were removed to allow release, and then 4 hr later cells were plated for CFA measurement of cell viability (Vassilev et al., 2006). Synchronisation timings were optimised using flow cytometry analysis of cell cycle progression ( Figure 2C Within the above synchronisation procedure, we either depleted RIF1 during the S phase Aphidicolin treatment period and re-expressed it for the recovery period (condition I, Figure 2B), or expressed RIF1 during the S phase treatment period and depleted it for the recovery period (condition II, Figure 2B). We included control samples where RIF1 was either expressed or depleted throughout the entire experiment ( Figure 2D, RIF1+ and RIF1-). Cells depleted of RIF1 only during the replication stress treatment period (condition I) showed sensitivity similar to the RIF1-depleted (RIF1-) condition, displaying a mean surviving fraction of 33% (data from two biological replicates shown in Figure 2-figure supplement 2C). Cells depleted of RIF1 only during the recovery period (condition II) also showed high sensitivity, again similar to RIF1-with a mean surviving fraction of 26% ( Figure 2D). These results indicate that molecular functions of RIF1 occurring both during and after a stressed S phase are important for survival. The cell death observed did not reflect an Video 1. mClover-tagged RIF1 from early to late S phase. Time-lapse imaging video of unsynchronised HCT116 mAC-RIF1 mCherry-PCNA cells transitioning from early to late S phase. Blue represents DNA (DAPI stain), red represents PCNA (mCherry-PCNA) and green represents mAC-RIF1 (mClover).
We performed similar conditional depletion experiments in asynchronous cell populations (as outlined in Figure 2-figure supplement 2D), and observed comparable results (Figure 2-figure supplement 2E), in that depletion of RIF1 either during or after Aphidicolin treatment led to sensitivity.
To summarise, these results imply that RIF1 must be present during both treatment and recovery to protect cells from the effects of replication stress induced by Aphidicolin. The observation that RIF1 function remains important after a stressed replication period to promote cell survival is consistent with its highly dynamic pattern of localisation through late cell cycle stages ( Figure 1G,H), suggesting that RIF1 operates in chromosome maintenance processes occurring outside of S phase.
Only the RIF1-Long splice isoform protects cells from replication stress The RIF1 messenger RNA undergoes alternative splicing resulting in expression of 'Long' and 'Short' protein isoforms that we name RIF1-L and RIF1-S. RIF1-S lacks 26 amino acids corresponding to exon 31 ( Figure 3A, exon 31 shown in red; see Materials and methods for Exon designation). We confirmed that transcripts encoding both the isoforms are widely expressed in various tissues in vivo ( Figure 3-figure supplement 1), and are present in U2OS and HEK293 cell lines (data not shown). Although they were mentioned as showing differential expression in cancer cells (Xu and Blackburn, 2004), distinct functions of the two isoforms have not previously been examined. We constructed cell lines expressing only mAC-RIF1-L or only mAC-RIF1-S, by inserting at the 3' end of Exon 29 a 'pre-spliced' cDNA construct consisting of Exons 30-35 including or excluding Exon 31 ( Figure 3A). Clones were selected where both copies of the mAC-RIF1 gene contained the insertion. Western analysis confirmed that these mAC-RIF1-L and mAC-RIF1-S constructs encoded proteins expressed at levels similar to parental mAC-RIF1 ( Figure 3B, lanes 2 and 3).
We found that whilst the mAC-RIF1-L cell line showed resistance to Aphidicolin very similar to that of the parent mAC-RIF1 cells ( Figure 3C, Figure 3-figure supplement 2, black and red lines), the mAC-RIF1-S isoform in contrast conferred little protection against drug, producing sensitivity similar to that of cells lacking RIF1 altogether ( Figure 3C, Figure 3-figure supplement 2, grey and dotted lines). This result suggests that only RIF1-L can protect cells from replication stress caused by Aphidicolin, and that RIF1-S is ineffective in this role.
To confirm this finding in a different cell line, we used HEK293-derived stable cell lines, containing siRIF1-resistant cDNA constructs encoding either GFP-RIF1-L or GFP-RIF1-S expressed under DOX control. Treatment of these cell lines with siRIF1, followed by DOX induction, allows replacement of endogenous RIF1 with either its Long or Short isoform (Figure 3-figure supplement 3A,B and C; Hiraga et al., 2017). We found that also in this cell line RIF1-L was able to protect against Aphidicolin treatment, whilst RIF1-S could not ( Figure 3D We considered molecular roles through which RIF1-L and RIF1-S might confer differing replication stress sensitivity. We found that both isoforms however appear equally functional in the previously described mechanisms through which RIF1 affects DNA replication. The isoforms are equally effective in preventing hyperphosphorylation of the MCM complex ( Figure 3E, Figure 3-figure supplement 4A) and protecting nascent DNA at blocked replication forks ( Figure 3F). Nor did we observe significant differences between cells expressing RIF1-S or RIF1-L in replication fork speed (as assessed by nascent DNA tract length; . RIF1-S therefore can apparently fulfil the known functions of RIF1 during DNA replication, and yet cannot safeguard cells from Aphidicolin treatment. These observations imply that responding to replication stress demands a further RIF1-mediated mechanism to promote cell survival, probably one that operates after the period of stress ( Figure 2) and that specifically requires RIF1-L ( Figure 3C,D).
In its functions in origin repression and nascent DNA protection during S phase, RIF1 acts as a PP1 substrate-targeting subunit (Garzó n et al., 2019;Hiraga et al., 2017;Hiraga et al., 2014;Mukherjee et al., 2019). To further test for separability of the effect of RIF1 in replication stress survival from its known roles in replication control, we investigated whether PP1 interaction is essential for RIF1-L to protect against Aphidicolin treatment. We used the HEK293 cell line expressing a ) .+%(# +!*&%-%)( ,-+!,, version of RIF1-L mutated at the PP1 interaction motifs to prevent PP1 interaction ( Figure 3G,H; Hiraga et al., 2017). This 'RIF1-L-pp1bs' protein was almost as effective as wild-type RIF1-L in conferring resistance to Aphidicolin ( Figure 3I, Figure 3-figure supplement 5A), indicating that RIF1-L acts largely independent of PP1 function in protecting cells from the effects of Aphidicolin. This independence from PP1 reinforces the evidence that the function of RIF1 in protecting from replication stress is distinct from its previously known roles in replication control, which do require PP1.

RIF1-Long and -Short isoforms do not appear to differ in mitotic functions
We considered other routes through which RIF1 might promote survival after Aphidicolin treatment, focusing on events occurring after the replication stress period itself. Chromosomal bridge structures have been reported to form after replication stress-in particular UFBs, believed to correspond to sites of under-replicated DNA that become single-stranded when chromosomes separate (Chan et al., 2009). UFBs do not stain with DAPI (which recognises double-stranded DNA). RIF1 acts at UFBs (Hengeveld et al., 2015), so we tested whether the RIF1-L and RIF1-S isoforms show differential recruitment to these sites or affect their rate of formation. UFB formation is induced by treatment with the topoisomerase II inhibitor ICRF-193 ( Figure 4Ai,ii). We found no clear difference in rates of UFB formation in the HCT116 RIF1-L or RIF1-S cells after ICRF-193 treatment (Figure 4Aii, left), and both RIF1-L and RIF1-S were localised at the UFBs formed ( Figure 4Ai, and Figure 4Aii, right). Treatment of these HCT116-based cell lines with Aphidicolin surprisingly did not induce UFBs, but rather increased the incidence of anaphase chromosome bridges (defined as bridge structures that stain with DAPI) (Figure 4Ai, Bi). These anaphase bridges were formed at similar rates in the RIF1-L and RIF1-S-expressing cells (Figure 4Bii, left) and both RIF1-L and RIF1-S were localised to these anaphase bridges ( Figure 4Bi, Bii, right). Overall therefore, both RIF1 isoforms can localise effectively to both anaphase bridge and UFB structures; and the loss of RIF1 does not greatly affect the rate at which such bridge structures occur. The results of this analysis were consistent with observations made by Imagestream flow cytometry, where analysis of untreated cell lines revealed Figure 2. RIF1 is essential during and after drug treatment to protect against effects of replication stress. (A) CFA testing Aphidicolin sensitivity of HCT116 mAC-RIF1 cells depleted for RIF1. Cells were treated with DOX and Auxin for 48 hr before seeding at low density and treatment with Aphidicolin at the concentrations indicated. 'No Auxin' cells were treated with DOX and Auxinole. After Aphidicolin removal, medium was replaced with DOX + Auxin medium for '+ Auxin' cells or DOX + Auxinole medium for 'No Auxin' cells. Plot shows mean and range from three biological replicates, each plated in technical triplicate. (B) Procedure for testing effect of depleting RIF1 during or after Aphidicolin treatment in synchronised cultures. HCT116 mAC-RIF1 cells were incubated with Lovastatin to arrest cells in G1. After 24 hr, cells were released from G1 arrest with Mevalonic Acid. 4 hr into release (28 hr from the initial addition of Lovastatin), RIF1 depletion was induced in 'condition I' cells by addition of DOX and Auxin. 8 hr into release (32 hr), 1 mM Aphidicolin was added (to both cultures) to induce replication stress and simultaneously, RO-3306 was added to arrest cells in G2 phase. 24 hr later (at 56 hr), in 'condition I' cells RIF1 was re-expressed by removal of DOX and Auxin and addition of Auxinole. Also at 56 hr, in 'condition II' cells RIF1 depletion was induced by addition of DOX and Auxin. 4 hr later (60 hr), both cell cultures were released from RO-3306. After 4 hr (64 hr), cells were seeded in triplicate at 250/well in 6-well plates (condition I in medium containing DOX + Auxinole and condition II in medium containing DOX + Auxin) then incubated for 7 days after which colonies were counted. mAC-RIF1 depletion and re-expression data is shown in Figure 2 Overall, these investigations provided no evidence that the deficiency of the RIF1-S isoform in protecting cells against Aphidicolin is caused by its failure to function in mechanisms that manage under-replicated DNA during mitosis.

RIF1-long promotes 53BP1 nuclear body formation in G1 phase
A further consequence of replication stress is the formation of large 53BP1 nuclear bodies (NBs) in the subsequent G1 phase, which protect unreplicated DNA damaged by chromosome breakage at mitosis (Lukas et al., 2011;Moreno et al., 2016). We examined the formation of 53BP1 NBs 12 hr after removal of Aphidicolin in our cells expressing only mAC-RIF1-L, only mAC-RIF1-S, or neither isoform. We limited our analysis to G1 phase cells by counting only those that were cyclin A2-negative (  (E) Both RIF1-L and RIF1-S can counteract MCM4 hyperphosphorylation caused by depletion of endogenous RIF1. For HEK293 cells, 24 hr after transfection with either siCon or siRIF1, DOX was added to the culture medium. 24 hr later, chromatin-enriched protein fractions were prepared and analysed by western blotting with anti-MCM4 antibody. Hyperphosphorylated MCM4 protein (MCM4-P) is detected based on its retarded mobility as indicated by bracket. (F) DNA fibre assay to assess nascent DNA degradation. HEK293-derived cell lines were transfected with siCon or siRIF1 to deplete endogenous RIF1. The following day, expression of the stably integrated GFP, GFP-RIF1-L and GFP-RIF1-S was induced by addition of DOX. 2 days later cells were labelled with CldU then IdU, followed by treatment with 2 mM hydroxyurea for 4 hr, after which the IdU:CldU ratio was measured in fibre analysis. Black bars represent median value. 75 forks were analysed per sample. NS: not significant; ****p 0.0001. (G) Illustration of RIF1 PP1-interaction motifs. To prevent PP1 interaction, critical residues in all three potential PP1 interaction motifs were replaced with alanine, creating a RIF1-pp1bs allele as described (Hiraga et al., 2017).     panel), demonstrating that RIF1 contributes to the formation of 53BP1 bodies after replication stress. Examining the cell lines expressing only Long or Short RIF1 isoforms, we found that mAC-RIF1-L localised normally with 53BP1 bodies, and supported their formation at a near normal rate ( Figure 5A second row, 5B,C red bars and Figure 5-figure supplement 1C top right panel). In contrast, the number of 53BP1 NBs formed in mAC-RIF1-S cells was similar to that in RIF1-KO cells, and mAC-RIF1-S showed reduced co-localisation with the 53BP1 bodies ( Figure 5A third row, 5B,C grey bars and Figure 5-figure supplement 1C bottom left panel). The reduced rate of 53BP1 NB formation in mAC-RIF1-S cells was not due to reduced abundance of 53BP1 in mAC-RIF1-S when compared to mAC-RIF1-L cells ( Figure 5-figure supplement 1A). The two RIF1 isoforms therefore differ in their effectiveness in promoting 53BP1 nuclear body formation following replication stress, with RIF1-L but not RIF1-S functional in this role. Since 53BP1 nuclear bodies are known to protect DNA damaged as a consequence of Aphidicolin treatment (Lukas et al., 2011), the defect in 53BP1 body formation in the absence of RIF1-L is likely to be an important factor in the replication stress sensitivity of RIF1-deficient cells, and may explain the isoform specificity of the replication stress protection function of RIF1.
To test in a different cell line whether RIF1 contributes to the number of 53BP1 nuclear bodies formed, we examined U2OS cells depleted of RIF1, 12 hr after release from a 24 hr Aphidicolin treatment ( Figure 5D, Figure 5-figure supplement 1D). We found that RIF1 depletion caused a reproducible reduction in the number of NBs, similar to that observed in our HCT116-derived cell lines.
TopBP1 has been reported to be involved in the transition of under-replicated DNA sites damaged during mitosis into 53BP1 NBs in the following G1 phase (Pedersen et al., 2015). We therefore tested for involvement of TopBP1 in the stimulation of 53BP1 NB formation by RIF1-L. We found that depleting TopBP1 substantially reduced the rate of 53BP1 NB colocalisation with RIF1-L ( Figure 5-figure supplement 2A,B). We conclude that TopBP1 is needed for effective localisation of RIF1-L with 53BP1 NBs, potentially enabling the specific function of RIF1-L in recovery from Aphidicolin treatment.

Discussion
In investigating mechanisms through which RIF1 protects against interruption to replication, we tested RIF1 isoforms and found that RIF1-L is able to protect against replication stress whilst RIF1-S cannot. This deficiency of RIF1-S function was initially surprising, since RIF1-S appears competent to fulfil the known functions of RIF1 in DNA replication management-in particular RIF1-S is able to support replication licensing and control MCM phosphorylation (Hiraga et al., 2017 and Figure 3E), and to protect against nascent DNA degradation ( Figure 3F). Consistently however, all of these known functions of RIF1 in replication control (promotion of replication licensing, control of MCM phosphorylation, and nascent DNA protection) depend on PP1 recruitment by RIF1 (Garzó n et al., Hiraga et al., 2017;Hiraga et al., 2014); whilst we find that protection against replication stress does not require PP1 interaction ( Figure 3I, Figure 3-figure supplement 5), again suggesting that the role of RIF1 in protecting from replication stress might involve additional mechanisms. Testing the effects of conditional depletion in synchronised cultures revealed that RIF1 is still needed after a period of replication stress to guard against toxicity, implying that protection from Aphidicolin involves a further function of RIF1. We therefore investigated whether RIF1 operates in post-S phase replication stress response pathways. RIF1 was reported to act at chromosome bridge  structures formed during mitosis after replication stress, so we tested whether the RIF1-L and RIF1-S isoforms function differently in this role. However, our investigations provided no evidence for differential roles of RIF1-L and RIF1-S at UFBs or anaphase bridges (Figure 4). Nor did we find any role for RIF1 in determining rates of MiDAS or chromatin breaks (Figure 4-figure supplement 1A,B). Our finding that RIF1 does not affect chromatin breaks contrasts with the results of Debatisse and Rosselli, 2019, who observed that RIF1 depletion suppresses fragile site breakage in human lymphoblastoid cells. Debatisse and Rosselli interpreted this suppression as due to earlier replication of fragile site DNA in lymphoblastoid cells in the absence of RIF1. The lack of any effect in our HCT116-based line may reflect cell line-specificity of RIF1 effects on replication timing (Gnan et al., 2020;Klein et al., 2019). Rather than a differential role of the RIF1 isoforms during mitosis, our investigation revealed a role for RIF1-L in promoting the assembly of G1 phase 53BP1 nuclear bodies ( Figure 5E). Although in DNA repair RIF1 is generally thought to act downstream of 53BP1, RIF1 was recently shown to contribute to the correct substructure of 53BP1 ionizing radiation-induced foci (IRIF) that form at radiation-induced double-strand breaks (Ochs et al., 2019). Ochs et al did not report the number of 53BP1 IRIF foci formed in the absence of RIF1, or test effects of the RIF1 isoforms. The factors important for the larger, replication stress-induced G1 phase 53BP1 nuclear body (NB) structures are less well understood, and there has been no investigation of how RIF1 affects NB formation. The role we have discovered for RIF1-L could be related to the effect described by Ochs et al for RIF1 in 53BP1 IRIF organisation, but in RIF1-S cells we found no evident reduction in IRIF formation after DSB-inducing treatment, in contrast to the clear defect in replication stress-induced G1 phase 53BP1 NB formation ( Figure 5). This difference suggests that the pathway through which RIF1 organises IRIF substructure (Ochs et al) may differ from the pathway through which RIF1-L promotes 53BP1 NB formation (this study). Factors that have been implicated upstream of 53BP1 NB formation include phospho-ATM and RNF168 (Lukas et al., 2011), but an initial investigation of these factors in RIF1-L and RIF1-S cells revealed no striking differences.
Our finding that RIF1-L but not RIF1-S can function in promoting 53BP1 body formation potentially explains the specific requirement for RIF1-L in protecting against replication stress, since 53BP1 NB assembly represents an important step in correct handling of stress-associated damage to enable ongoing proliferation. RIF1-L may act to directly promote 53BP1 NB assembly, or possibly assist with the transit of damaged sites through to G1 phase to allow such protective bodies to form. Presently it is unclear how the apparently small difference of RIF1-L from the RIF1-S isoform (the inclusion of just 26 amino acids) specifically supports 53BP1 nuclear body formation. Interestingly however, this region contains the motif 'SPKF' at amino acids 2265-2268, with the S-2265 residue identified as phosphorylated in high-throughput proteomic studies (Schweppe et al., 2014;Sharma et al., 2014) and in our own proteomics data. This sequence conforms to a 'phosphoSPxF' BRCT interaction motif, raising the possibility that RIF1-L interacts with one of the several BRCT domain-containing proteins involved in chromosome maintenance (which include BRCA1, TopBP1 and 53BP1 itself) to effect its role in promoting 53BP1 NB formation and protecting from Aphidicolin. However, the physical interaction of RIF1 with 53BP1 does not appear to be mediated through a BRCT domain-mediated interaction (Callen et al., 2013). We found that TopBP1 is important to enable the observed association of RIF1-L with 53BP1 NBs. However, co-immunoprecipitation analysis of RIF1-L following Aphidicolin treatment did not identify any interaction with TopBP1 (data not shown), so the molecular mechanism through which RIF1-L promotes 53BP1 NB formation remains an open question.
The findings described here suggest that RIF1 contributes to recognising under-replicated and unrepaired sites for special protection and handling later in the chromosome cycle-in particular for delayed replication that guards against unscheduled recombinational repair to prevent the formation of pathological intermediates, as recently described (Spies et al., 2019). Overall, this study highlights the multifunctional role of RIF1 in ensuring chromosome maintenance to promote the survival and proliferation of cells after replication stress, emphasising the importance of RIF1 for determining response to replication-inhibiting chemotherapeutic drugs.

Cell lines
Stable HEK293 Flp-In T-REx GFP and GFP-RIF1-S cell lines were as described (Escribano-Díaz et al., 2013;Hiraga et al., 2017). Constructed using the same procedure, described briefly under 'HEK293 cell lines and culture conditions', were cell lines: HEK293 Flp-In T-REx GFP-RIF1-L and HEK293 Flp-In T-REx GFP-RIF1-L-pp1bs. Construction of the necessary plasmids is described in the next section. The HCT116 mAC-RIF1 cell line was constructed as outlined below in 'HCT116 cell lines and culture conditions', using the system described (Natsume et al., 2016;Yesbolatova et al., 2019). HCT116 mAC-RIF1-L, HCT116 mAC-RIF1-S and HCT116 RIF1 KO cells were constructed as described below. HCT116 mAC-RIF1 mCherry-PCNA was constructed by introducing the mCherry-PCNA construct under control of the EF1 alpha promoter using the piggyBac system (Yusa et al., 2011).
For the HEK293-derived cell lines, presence of the Tet-repressor modification provides confirmation of cell line identity. For HCT116 OsTIR1-derived cells and HCT116 RIF1 KO cell lines, presence of the OsTIR1 modification confirms cell line identity. U2OS cells are from ATCC. Cell lines tested negative for mycoplasma.
The GFP-RIF1 constructs used in this study are based on pcDNA5/FRT/TO-GFP-RIF1 (Escribano-Díaz et al., 2013), which carries human RIF1-S cDNA with GFP fused at its N-terminus. To construct pcDNA5/FRT/TO-GFP-RIF1-L, a PCR fragment containing RIF1-S cDNA was amplified from pcDNA5/FRT/TO-GFP-RIF1 using primers SH593 and SH594, and cloned into pIRESpuro3 vector (linearised by EcoRV and EcoRI) using the In-Fusion HD cloning system, to create plasmid pSH1009. The NheI-NotI fragment of the plasmid pSH1009 was replaced using the In-Fusion HD system by two PCR fragments (amplified by SH572 and SH595 and SH596 and SH597 respectively), to construct pSH1011 which has RIF1-L cDNA. The NheI-PspOMI fragment of pcDNA5/FRT/TO-GFP-RIF1 was replaced by NheI-NotI fragment of the pSH1011 plasmid to construct pcDNA5/FRT/TO-GFP-RIF1-L. Construction of a GFP-RIF1-S-pp1bs plasmid was previously described (Hiraga et al., 2017). The GFP-RIF1-L-pp1bs construct was made following a similar strategy.
The plasmid pX330-U6-chimeric_BB-CBh-hSPCas9 from Feng Zhang (Addgene, 42230) (Cong et al., 2013) was used to construct the CRISPR/Cas9 vector for the guide RNA according to the protocol described . Donor plasmids were based on pBluescript and constructed as described (Natsume et al., 2016;Yesbolatova et al., 2019). Primers for amplification of the homology arms and cDNA from RIF1-L and RIF1-S are listed in the key resource table.
HEK293 cell lines and culture conditions HEK293-derived cell lines were cultivated in Dulbecco's Modified Eagle's Minimal medium supplemented with 10% foetal bovine serum (tetracycline-free), 100 U/ml penicillin, and 100 mg/ml streptomycin at 37˚C with 5% CO 2.
To construct cell lines, pOG44 (O'Gorman et al., 1991) and pcDNA5/FRT/TO-based plasmids carrying the GFP-RIF1-L or GFP-RIF1-L-pp1bs gene were mixed in 9:1 molar ratio and used for transfection of Flp-In T-REx 293 cells (Invitrogen) with Lipofectamine 3000 reagents (Invitrogen). Transfections and hygromycin B selection of stably transfected cells were performed as described by the manufacturer. Clones were tested for doxycycline-dependent induction of GFP fusion proteins by western blot and microscopy.
To assess the effect of ectopically expressing RIF1, cells were transfected with either Control siRNA or siRNA against human RIF1 (see key resource table for sequence). 2 days later, cells were split with addition of 1 mg/ml DOX then incubated for 24 hr to induce expression of GFP-RIF1 variant proteins. siRNA transfection was carried out using Lipofectamine RNAiMAX (Invitrogen). Synonymous base mutations in the ectopically expressed GFP-RIF1 constructs make them resistant to siRNA targeted against endogenous RIF1 (Escribano-Díaz et al., 2013). RIF1 expression was assessed by western blot using RIF1 and GFP antibodies.

HCT116 mAC-RIF1
To construct miniAID-mClover-fused RIF1 stable cell lines, HCT116 cells expressing the auxinresponsive F-box protein Oryza sativa TIR1 (OsTIR1) under the control of a Tet promoter were transfected using FuGENE HD (Promega) with a CRISPR/Cas9 plasmid (pX330-RIF1-N) targeting nearby the 1 st ATG codon of the RIF1 gene (5'-TCTCCAACAGCGGCGCGAGGggg-3') together with a donor plasmid (pmAC-RIF1) based on pMK345 (Yesbolatova et al., 2019), that contains a cassette (hygromycin resistance marker, self-cleaving peptide P2A, and mAID-mClover; Yesbolatova et al., 2019) flanked by 500 bp of homology arms on each end. Two days after transfection cells were diluted in 10 cm dishes, to which 100 mg/mL of Hygromycin B Gold (Invivogen) was added for selection. After 10-12 days, colonies were picked for further selection in a 96-well plate. Bi-allelic insertion of the donor sequence was checked by genomic PCR. Clones taken forward for analysis of RIF1 function were selected based on near endogenous RIF1 levels of RIF1 and efficient AID-mediated degradation of RIF1.
To induce degradation of miniAID-mClover-fused RIF1, OsTIR1 expression was first induced by 0.2 mg/ml DOX added to the culture medium, to produce a functional SCF (Skp1-Cullin-F-box) ubiquitin ligase that directs degradation of an AID-tagged protein (Natsume et al., 2016;Nishimura et al., 2009). After 24 hr, 10 mM Auxin (indole-3-acetic acid) was added to the culture medium to promote the interaction of mAC-RIF1 with SCF-OsTIR1, driving ubiquitination and subsequent degradation of mAC-RIF1. To suppress premature degradation of RIF1 in the presence of DOX, 100 mM of the auxin antagonist Auxinole, which interferes with TIR1 activity, was added (Hayashi et al., 2012;Yesbolatova et al., 2019). In subsequent depletion experiments DOX was first added in the presence of Auxinole, and then Auxinole was replaced with Auxin. Validation of DOX and Auxin concentrations is shown in Figure 1-figure supplement 1A and B. The abovementioned drug concentrations were used except where otherwise stated.

HCT116 RIF1 KO
HCT116 cells were transfected with the same CRISPR/Cas9 plasmid that was used to construct HCT116 mAC-RIF1 cells (pX330-RIF1-N), targeting near the 1 st ATG codon of the RIF1 gene. Simultaneously transfected was a donor plasmid (pRIF1-KO) containing a hygromycin resistance marker flanked by 500 bp homology arms. Transfection and clonal selection was carried out as described above. Clones were tested for loss of RIF1 expression by western blot.

Protein extraction and western blotting
To prepare whole cell protein extracts, cells were trypsinised and washed with Dulbecco's Phosphate-Buffered Saline (PBS) before treating with lysis buffer (10 mM Tris pH 7.5, 2 mM EDTA) containing a protease and phosphatase inhibitor cocktail (Roche). Chromatin-enriched protein fractions were prepared essentially as described (Mailand and Diffley, 2005). Protein concentrations were determined using the Bio-Rad RC-DC protein assay kit and equal amounts of total proteins were loaded in each lane. Proteins were transferred to PVDF membrane using the Trans-Blot Turbo Blotting System (Bio-Rad). See key resource table for antibodies used. Proteins were detected by Clarity Western ECL blotting substrate (Bio-Rad) and the Bio-Rad Chemidoc Touch Imaging System.

Colony formation assay (CFA)
In the case of HEK293-based cell lines, cells were seeded and transfected with 50 nM RIF1 siRNA using Lipofectamine RNAimax (Invitrogen) together with Optimem (Gibco). For HCT116-based cell lines, cells were seeded with 0.2 mg/ml DOX. After two days, cells were counted using the Invitrogen Countess II FL Automated Cell Counter and 250 cells added to each well of a 6-well plate (samples always plated in technical triplicate). For HEK293-based cell lines, 1 mg/ml DOX was added to the culture medium to induce ectopic RIF1 expression whilst for HCT116-derived cell lines, 10 mM Auxin was added to the culture medium to degrade RIF1. Cells were incubated for 24 hr after which Aphidicolin (or Hydroxyurea) was added and cells incubated for a further 24 hr, before washing twice with PBS and replacement with the appropriate Aphidicolin-free medium. For HEK293-derived cell lines, 1 mg/ml supplementary DOX was re-added 72 hr after Aphidicolin removal and cells were then incubated for a further 4 days. In the case of HCT116-derived cell lines, after Aphidicolin removal, cells were incubated for 7 days in medium containing appropriate drug (DOX, Auxin, Auxinole). At the end of the incubation period, colonies consisting of more than 20 cells were counted directly using a Nikon Eclipse TS100 microscope, as in Buonomo et al., 2009. Values were normalised to the 0 mM APH (DMSO) control. Statistical significance was calculated using a student's T-test in Prism (Graphpad).

Flow cytometry
To assess DNA content, cells were recovered by trypsinisation, then fixed with 70% ethanol. Cells were spun down and resuspended in 0.5 ml FxCyclePI/RNase staining solution (Molecular Probes, F10797) and incubated for 30 min at room temperature, protected from light. Manual gating on DNA content was used to distinguish cell cycle phases.
To measure the mClover fluorescence of the mAC-RIF1 fusion protein, cells were recovered by trypsinsation and fixed with a 10% neutral buffered formalin solution (Sigma, HT-5012) for 30 min at 4˚C, protected from light.
For EdU pulse labelling, 10 mM EdU was added to cells for a 15 min pulse before harvesting. The Click-iT reaction (Invitrogen) using Alexa Fluor 647 was carried out to fluorescently label EdU.
Flow cytometry was performed on a Becton Dickinson Fortessa analytical flow cytometer and FlowJo software was used for analysis. Doublet discrimination was performed by gating FSC-A against FSC-H. An equal number of events are shown in each set of histogram plots.

Cell cycle synchronisation
HCT116 cells were seeded in 12-well dishes and treated with 20 mM Lovastatin for 24 hr to induce G1 arrest (Rao et al., 1999). Cells were washed and medium containing 2 mM Mevalonic acid (MVA) added to release the cells from the arrest. 8 hr after release, 9 mM RO-3306 was added to hold cells at the G2/M boundary (Javanmoghadam-Kamrani and Keyomarsi, 2008;Vassilev et al., 2006). After 28 hr in RO-3306, cells were washed and drug-free medium added allowing cells to enter mitosis. Flow cytometry was used to analyse synchronisation efficiency, and to establish and optimise the procedure for the experiment in Figure 2 based on assessment of cell cycle progression kinetics.

DNA fibre assay
Cells were pulse-labelled with 50 mM CldU (20 min), followed by another pulse with 250 mM IdU (20 min). After treatment with 2 mM Hydroxyurea for 4 hr, cells were harvested and lysed on a microscope slide with spreading buffer (200 mM Tris pH 7.4, 50 mM EDTA, 0.5% SDS). Slides were tilted to allow the DNA suspension to run slowly and spread the fibres down the slide. Slides were fixed in cold (À20˚C) methanol-acetic acid (3:1) and DNA denatured in 2.5 M HCl at RT for 30 min. Slides were blocked and incubated with primary antibodies for 1 hr at RT in a humidity chamber (anti-CldU, anti-IdU, anti-ssDNA). After washes with PBS, the slides were incubated with secondary antibodies (anti-rat IgG Alexa Fluor 594, anti-mouse IgG1 Alexa Fluor 488, anti-mouse IgG2a Alexa Fluor 350). Slides were air-dried and mounted with Prolong (Invitrogen). Samples were imaged under a Zeiss Axio Imager and analysed using ImageJ. CldU and IdU tract lengths were measured in doublelabelled forks and the IdU/CldU ratio was used as an indicator of nascent DNA degradation. Statistical significance was calculated using a Mann-Whitney test.
Wide-field immunofluorescence microscopy HCT116 cells were cultured in a glass-bottomed dish (MatTek). Cells were fixed with 4% paraformaldehyde/PBS, for 15 min at room temperature. Cells were antibody stained with primary antibodies (anti-53BP1, anti-FANCD2, anti-BLM) and secondary antibodies (anti-rabbit Alexa Fluor 594, antigoat Alexa Fluor 594, GFP-Booster ATTO 488). Microscopy was performed using a DeltaVision microscope (GE) with a x60 objective. 41 Z-stacks of 0.3 mM were taken and deconvolution was performed using the SoftWoRx software (DeltaVision). Maximum intensity projections (MIP) were created using Volocity software (PerkinElmer). Unless stated otherwise, images shown are projections.

Live-cell imaging
HCT116 cells were cultured in a glass-bottomed dish (MatTek) containing medium without phenol red. To visualise nuclei in live cells, 0.5 mM SiR-DNA (Spirochrome) was added to the medium before observation. Live-cell imaging was performed using a DeltaVision microscope equipped with an incubation chamber and a CO 2 supply (GE). Image analysis and quantification was performed using the Volocity software (PerkinElmer).

Confocal immunofluorescence microscopy
HCT116 cells and U2OS cells were cultured in ibiTreat m-slide eight well dishes (Ibidi). Cells were treated with 1 mM Aphidicolin for 24 hr after which Aphidicolin was removed and cells were incubated for a further 12 hr. Where indicated siTopBP1 (Ambion s21823, see key resource table for sequence) transfection was performed 48 hr before Aphidicolin treatment. Cells were fixed with either a 10% neutral buffered formalin solution (Sigma, HT-5012) for 10 min at room temperature or with 100% methanol for 15 min at À20˚C. Blocking was for 30 min with PBST/1%BSA at 4˚C after which antibody staining was performed (anti-cyclinA2-Alexa Fluor 555, anti-53BP1 self-conjugated to Alexa Fluor 647). Slides were mounted with Ibidi mounting medium containing DAPI. Confocal microscopy was performed using a LSM880 + Airyscan (Zeiss) with a x63 objective. Z-stacks were imaged (40 slices). Images were processed first to an Airyscan image and then to a MIP using ZEN Black (Zeiss). Cyclin A2 staining was used to identify G1 cells (Cyclin A2 negative = G1 cells). Both image analysis (colocalisation) and quantification were performed using CellProfiler (Broad Institute) with a custom-made pipeline. The 'IdentifyPrimaryObjects' module was used to identify individual cells (DAPI positive), the G1 cell population (CyclinA2 negative), mAC-RIF1 foci (mClover positive) and 53BP1 foci (Alexa Fluor 647 positive). The 'RelateObjects' module was used to count the number of mAC-RIF1 and 53BP1 foci in a cell and to assess colocalisation of 53BP1 foci with mAC-RIF1 foci. Statistical significance was calculated using a one-way ANOVA using Prism (Graphpad).

cDNA analysis of RIF1 splice variants
A human cDNA panel covering 48 major tissues was obtained from Insight Biotechnology (Origene HMRT104). The two RIF1 splicing variants were amplified by competitive PCR using a single primer pair (see key resource table). The PCR products were analysed on a 2% agarose gel (3:1 mix of lowmelting temperature agarose and normal agarose). Bands were visualised by staining with SYBR Green I dye, and fluorescent images were captured using Bio-Rad ChemiDoc Touch system. Intensities of bands specific to RIF1-L, RIF1-S and an additional minor band caused by heteroduplex formation, were measured using ImageLab software (Bio-Rad). The relative abundance of RIF-L and RIF1-S were calculated based on the band intensities, with 50% of the heteroduplex band intensity assigned to RIF1-L and 50% assigned to RIF1-S. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Additional files

Supplementary files
. Source data 1. Underlying data for graphs.
. Transparent reporting form

Data availability
All data generated or analysed during this study are included in the manuscript and supporting files.