A clinically relevant heterozygous ATR mutation sensitizes colorectal cancer cells to replication stress

Colorectal cancer (CRC) ranks third among the most frequent malignancies and represents the second most common cause of cancer-related deaths worldwide. By interfering with the DNA replication process of cancer cells, several chemotherapeutic molecules used in CRC therapy induce replication stress (RS). At the cellular level, this stress is managed by the ATR-CHK1 pathway, which activates the replication checkpoint. In recent years, the therapeutic value of targeting this pathway has been demonstrated. Moreover, MSI + (microsatellite instability) tumors frequently harbor a nonsense, heterozygous mutation in the ATR gene. Using isogenic HCT116 clones, we showed that this mutation of ATR sensitizes the cells to several drugs, including SN-38 (topoisomerase I inhibitor) and VE-822 (ATR inhibitor) and exacerbates their synergistic effects. We showed that this mutation bottlenecks the replication checkpoint leading to extensive DNA damage. The combination of VE-822 and SN-38 induces an exhaustion of RPA and a subsequent replication catastrophe. Surviving cells complete replication and accumulate in G2 in a DNA-PK-dependent manner, protecting them from cell death. Together, our results suggest that RPA and DNA-PK represent promising therapeutic targets to optimize the inhibition of the ATR-CHK1 pathway in oncology. Ultimately, ATR frameshift mutations found in patients may also represent important prognostic factors.

. The clinically relevant ATR mutation is associated with a dysregulation of the ATR/M activities, endogenous replication stress and DNA damage. (A) Schematic representation of the ATR peptide with its functional domains. The "9A" frameshift mutation induces a stop codon in ATR's ORF (Stop sign). Hence, this mutant allele is unable to code for the catalytic C-terminal domains of ATR. (B) Wild-type (WT, blue) and mutant (MUT, red) ATR clones were isolated from HCT116 by limit dilution. The poly-A region located in ATR's exon 10 was sequenced. WT harbors the homozygous wild-type ATR sequence, whereas MUT harbors the clinically significant heterozygous frameshift mutation, leading to a downstream premature stop codon (red box). A revertant "REV" (green) clone has been obtained by genetic engineering of MUT using ZFN nucleases to restore a homozygous wild-type ATR genotype while conserving the genetic background of MUT. (C) Western Blot analysis of the ATR-CHK1 and ATM-CHK2-KAP1 pathways.GAPDH served as a loading control for each individual blot (see [ Fig. S5]). A representative GAPDH panel is shown. The quantifications (relative to WT) were calculated from 3 to 5 biological replicates. Mann-Whitney tests were applied. (Full blots: [Fig. S5]) (D) DNA fiber assay performed in ATR mutant cells. IdU and CldU were sequentially added to culture media. DNA was manually spread, and fibers were observed (× 400). Representative images are shown. IdU-positive CldU tracks were measured using ImageJ (exclusion of origin firing and termination events). (n = 600, 200 fibers × 3 independent experiments). Means are displayed on the graphs. Mann-Whitney significance tests were applied. (E) γ-H2AX (red) and 53BP1 (green) immunofluorescent detection. DNA was counterstained with DAPI (blue). Cells were observed with a Zeiss Axio Imager with Apotome (× 63 objective). 53BP1 and γH2AX foci were scored in > 200 cells per condition. The experiment was repeated 3 times (total > 600 cells) and the percentage of cells harboring > 10 γH2AX or > 5 53BP1 foci were plotted on graphs. Mann-Whitney significance tests were applied. www.nature.com/scientificreports/ Altogether, these results demonstrate that the ATR mutation correlates with altered ATR/M pathway activities, endogenous replication stress and increased endogenous DNA damage levels (notably DSBs) but no change in the cell cycle profile.

The ATR heterozygous mutation found in MSI + tumors sensitizes cells to topoisomerase I and ATR inhibitors.
Since it is known that the ATR-CHK1 replication checkpoint has a major role in drug response, we checked if this endogenous ATR deficiency has any impact on cell survival upon a panel of drugs. To do so, sulforhodamine survival assays (SRB) were performed. We showed that the ATR mutation increases cell sensitivity to several molecules. Amongst them, SN-38 (active metabolite of Irinotecan, topoisomerase I inhibitor) and VE-822 (ATR inhibitor) were those that showed the most dramatic sensitization of ATR mutant cells [ Fig. 2A]. A twofold decrease of the IC 50 of both SN-38 and VE-822 was observed in ATR mutant as compared with WT and REV cells. To discriminate between cytotoxic and cytostatic effects, we completed our analysis of SN-38 and VE-822 toxicity using Celigo Imaging Cytometer (Nexcelom Bioscience) on living cells. The percentage of dead cells was assessed using the CellTox kit (Promega), at doses of SN-38 and VE-822 matching their respective IC 50 [Fig. 2B]. A significant increase in cell death was highlighted in ATR mutant cells for both molecules, although mutant cells also presented a slightly higher spontaneous death rate. As SN-38 is known to induce replication stress and notably ATR dependent responses, we combined both molecules in 72 h survival assays. The combination of both molecules resulted in highly synergistic effects, reaching a maximum of 50% [ Fig. 2C] in WT cells. The maximum synergy was achieved at doses roughly matching the IC 50 of each molecule in single drug treatments. The cytotoxic effects of the SN-38 + VE-822 combination was strongly enhanced in all clones as compared to single drug treatments, reflecting the synergistic effects observed with SRB assays. The ATR mutation was here linked to a twofold increase in cell death [ Fig. 2D]. Similar but slightly weaker synergistic effects were highlighted between VE-822 and other widely-used replication stress inducers such as hydroxyurea (HU) and aphidicolin (APH), but not with the microtubule depolymerization inhibitor taxol, used here as a negative control since it does not target replicating cells [ Fig. S2A]. Altogether, these results highlight the cytotoxic sensitization of ATR mutant cells to topoisomerase I and ATR inhibitors, either used separately or combined.
The SN-38 + VE-822 combination triggers a caspase-3 dependent apoptosis as cells undergo extensive DNA damage due to a failure of the replication checkpoint. To better characterize the cytotoxic effects of the drugs, we assessed for caspase 3 activation using the Caspase 3-7 live kit (Nexcelom) on the Celigo, as a measure of apoptotic induction [ Fig. 3A]. The basal levels of caspase activity in untreated cells were enhanced by both SN-38 and VE-822 treatments, in an ATR mutation-dependent manner. The combined (SN-38 + VE-822) treatment greatly enhanced apoptosis induction, especially in ATR mutant cells. Furthermore, assessing for the DNA content of the cells showed a dramatic increase in the sub-G1 population in response to SN-38 ± VE-822, especially in ATR mutant cells [ Fig. S2B]. Moreover, the levels of cleaved-PARP-a product of caspases activation-followed the same pattern, thereby reinforcing the previous results [ Fig. S2C]. Together, these data demonstrate that the ATR mutation is linked to an increased apoptosis induction by SN-38 ± VE-822. Because SN-38 induces RS, we suspected this apoptosis induction to be linked to DNA damage. To answer this question, we performed further analyses at the 48 h timepoint (↓B-E). Upon SN-38 + VE-822 treatment, increased levels of both cleaved caspase-3 and the DNA damage marker γH2AX were observed in the ATR mutant cells by WB [ Fig. 3B Fig. 3D], strongly correlating the ATR mutation with extensive apoptosis induction and DNA damage. We hypothesized that such phenotypes could be caused by a failure of the replication checkpoint in ATR mutant cells. We therefore quantified ATR-CHK1 and ATM-CHK2 activities in cells challenged by the drugs [Fig. 3E]. SN-38 alone induces a massive ATR-CHK1 pathway activation, which is significantly downregulated in ATR mutant cells (upper panel). Therefore, in presence of SN-38, ATR mutant cells fail to achieve a full activation of the ATR-CHK1 pathway. Furthermore, we demonstrated that this checkpoint failure was linked to a permissive replication progression despite the presence of SN-38 in ATR mutant cells challenged with SN-38 for 24 h [Fig. 3F]. Indeed, SN-38 tends to reduce the percentage of replicating cells while increasing in the G2/M, this being significantly exacerbated in the ATR mutant (red arrows). Moreover, while VE-822 totally abrogated the SN-38-induced ATR-CHK1 activation, ATM targets CHK2 and KAP1 were found hyper phosphorylated in response to SN-38 + VE-822, especially in ATR mutant cells [ Fig. 3E] (lower panel). Altogether, these data suggest that the ATR mutation weakens the replication checkpoint of cells, leading to the accumulation of DNA damage and ultimately, to an enhanced apoptosis induction.
The ATR signaling is the critical barrier protecting cells from ssDNA accumulation, RPA exhaustion and replication catastrophe during SN-38 treatment. As mentioned in the introduction, whether SN-38 could induce exhaustion of RPA and replication catastrophe (RC) is not known yet. Hence, we postulated that in response to SN-38, the ATR signaling serves as a critical barrier to replication progression and, potentially, RPA exhaustion and RC. In an attempt to answer these questions, we used higher and isotoxic (~ 20 IC 50 ) concentrations of SN-38 (40 nM) and VE-822 (2 µM) to achieve optimal replication checkpoint activation and repression,as assessed by western blotting [ Fig. S3A]. At these doses, a strong SN-38 (40 nM)induced ATR-CHK1 activation is significantly impaired in ATR mutant cells, harboring more DNA damage. VE-822 (2 µM) completely suppressed the SN-38 -induced CHK1 phosphorylation. We then monitored cell cycle using complementary BrdU/PI labelling strategies during the acute response To complete this analysis, we designed experiments to study the co-labellings of γH2AX and RPA32 coupling immunofluorescent and flow cytometry approaches in order to cover both sub-nuclear and pan-nuclear levels, respectively [ Fig. 4D]. Representative cells pointed by white arrows are magnified below, with labellingintensity cross sections. In normally growing cells, immunofluorescence highlighted few γH2AX foci spread throughout the nuclei. VE-822 did not affect these parameters. SN-38 led to an intense and co-occurring foci formation of γH2AX and RPA32. However, in a fraction of cells, SN-38 + VE-822 induced a strong, homogenous and pan-nuclear H2AX phosphorylation covering the RPA32 foci distribution, a classical phenotype of replication catastrophe (RC) 24,33 . Flow cytometry showed that SN-38 leads to the induction of γH2AX/RPA32 double positive cells. VE-822 by itself induced no phenotype as compared to the untreated control. However, when coupled to SN-38, VE-822 increased the percentage of double positive cells by enhancing the γH2AX intensity of RPA32 positive cells. This suggests that the RPA "protection threshold" was reached, inducing the accumulation of extensive DNA damage. Furthermore, IF and FACS experiments highlighted similar 53BP1pS1778 patterns, another DSB marker 34 [ Fig. S3C]. Together, these results demonstrate that SN-38 induces DNA damage (depleting RPA pools) which in turn lead to an ATR-dependent replication arrest. By abrogating this replication checkpoint, VE-822 contributes to the exhaustion of RPA as cells are forced to progress through replication, eventually leading to RC, characterized by extensive DNA damage. As we anticipated, these phenotypes were significantly enhanced in ATR mutant cells [ Fig. 4E] hence giving a molecular explanation of their hypersensitivity to these drugs. A higher proportion of unchallenged ATR mutant cells showed increased levels of chromatin recruitment of RPA and H2AX phosphorylation [ Fig. S3B], it is therefore likely that MUT cells are predisposed to the RPA exhaustion and RC induced by the SN-38 + VE-822 combination. Interestingly, roscovitine, a CDK inhibitor that prevents origin firing, partially protected cells from RPA exhaustion and DNA damage accumulation induced by the SN-38 + VE-822 combination [ Fig. 4F]. This observation suggests that the negative regulation of origin firing by the ATR signaling is a critical mechanism targeted by VE-822, leading to the protection of cells from RPA exhaustion and RC in presence of SN-38.
As SN-38 + VE-822 leads to a 4 N-post replicative arrest of the cell cycle, we then wanted to identify the protein(s) involved in this post replication checkpoint. Indeed, having a complete understanding of compensatory checkpoint pathways that protect cells in the absence of ATR kinase activity is of high interest for future drug association in order to optimize SN-38/VE-822 efficiency, especially in ATR mutant cells.
DNA-PK-but not the ATM-CHK2 axis-is primarily involved in the post replicative cell cycle arrest in response to SN-38 + VE-822. Amongst the DNA Damage response (DDR) major pathways, the three PI3K (ATR, ATM and DNA-PK) are known to hold concerted and non-redundant function in response to DNA lesions occurring during replication. Previous studies highlighted the relative roles played by these 3 branches of the DDR network in checkpoint responses to several DNA damaging agents, using combined siRNA and inhibitors of ATM, ATR (VE-821) and DNA-PK 32 . Here, we studied these processes in the context of the clinically relevant SN-38, using the most recent and specific ATR inhibitor VE-822. Therefore, in a replication

Discussion
The ATR mutation found in MSI + CRCs is critical during the cellular response to SN-38 and VE-822. RS is a hallmark of most cancers 4,35 By triggering the replication checkpoint, the ATR-CHK1 axis holds a pivotal role in the management of RS. Our results show that cells harboring the clinically relevant heterozygous mutation in ATR display a dysregulation of this checkpoint response, associated with higher endogenous replication stress and DNA damage. However, ATR mutant cells seem to somehow compensate for these phenotypes, as cell cycle and S-phase progression remain unaffected. Notably, the slowdown of replication forks observed in ATR mutant cells could likely be compensated by dormant origin firing, a phenotype observed during ATR inhibition 36 . Nevertheless, we rationally hypothesized that this ATR mutation, found in MSI + cancers, could be critical during drug response, especially for molecules known to be ATR-CHK1 inducers such as SN-38 37 . Indeed, our results highlighted a twofold-increased sensitivity of ATR mutant cells to SN-38. Moreover, a similar phenotype is observed when treating the cells with VE-822, a highly specific ATR inhibitor 38 . Furthermore, in agreement with recent publications, strong synergistic effects were highlighted between the two molecules 39 . In addition, we showed that the inhibition of the ATR pathway by VE-822 greatly potentiates the Caspase-3 dependent apoptosis induced by SN-38. Here again, the ATR mutation was linked to a twofold increased apoptotic induction in response to the drug combination SN-38 + VE-822. As VE-822 is currently in clinical trial (notably in combination with DNA damaging agents), we could expect significant benefits of these drug combinations especially in ATR mutated tumors. Interestingly, mutations in CHK1 exon7 have been reported as well and may induce similar phenotypes 40 . This could give the possibility to adapt posology depending on the tumor genotype, to reduce side effects.

The ATR mutation predisposes cells to the exhaustion of RPA and replication catastrophe induced by the SN-38 + VE-822 combination.
The RPA protein complex plays various critical roles to preserve genome integrity 41,42 . It has essential functions for the physical protection of single-stranded DNA and checkpoint activation 43 . It is notably involved in Nucleotide Excision Repair (NER) DNA repair pathway and   26 , we show that it creates a propitious background to the exhaustion of RPA. Indeed, the generation of DSBs (as marked by γH2AX and 53BP1pS1778) likely promotes the generation of ssDNA via DNA resection by various exonucleases such as MRE11, EXO1, DNA2 or CtIP [44][45][46] . ATR-CHK1 activation is crucial to set-up the early-S replication arrest. As VE-822 allows cells to bypass this checkpoint and resume replication, most probably via unscheduled origin firing 36 , more DSBs are generated during the collisions between replication forks and Top-1CCs, inducing higher levels of ssDNA as cells progress into replication. This exhausts the soluble pool of RPA and cells eventually reach a RC state, most likely in late-S/G2. This is consistent with our finding that the CDK inhibitor roscovitine protects cells from RPA exhaustion and with the known protective role of the ATR-CHK1 pathways in other replication stress contexts 24,25 . Moreover, we show that ATR mutant cells, which show higher basal levels of ssDNA, RPA chromatin recruitment and DNA damage, are predisposed to these detrimental outcomes, which likely explains their hypersensitivity to SN-38 and VE-822. RPA is abundantly expressed in the nucleus of eukaryotic cells and is overexpressed in numerous tumor types including lung, ovarian, and colorectal cancers 47 . Many efforts have been made to find RPA inhibitors in order to potentiate the anti-tumor activity effects of several molecules 48 . Indeed, the use of RPA inhibitors such as HAMNO 49 or TDRL-505/551 50 , could lower the "RPA protection threshold", potentially increasing the efficiency of the SN-38 + VE-822 combination.
We observed that cells surviving the SN-38 + VE-822 combination complete replication and accumulate in a post-replicative state characterized by RPA hyperphosphorylation at serines 4/8 and high levels of DNA damage, notably DSBs marked by 53BP1pS1778 34 .
Master regulator of the post-replication cell cycle arrest, DNA-PK is a potential co-target to achieve synthetic lethality with ATR inhibitors. Double-strand breaks (DSBs) are among the most toxic DNA lesions. At the cellular level, DSBs are mainly managed by the mutually exclusive HR (Homologous Recombination) and NHEJ (Non-Homologous End-Joining) pathways 51 . The accurate HR repair notably depends on ATM, the MRN complex and BRCA1/2 factors, while the repair by NHEJ (a far less reliable repair process) is mediated by the DNA-PK complex. During the canonical NHEJ, the catalytic subunit DNA-PKcs is recruited to the DSBs and auto-phosphorylates (Ser2056) 52 . We showed that the SN-38 + VE-822 combination triggers DNA-PK phosphorylation at Ser2056, which is correlated with the high levels of DNA damage resulting from the exhaustion of RPA. We showed that cells eventually get stopped in a post replicative state by DNA-PK, keeping them from entering mitosis. A contrario, DNA-PK -/cells fail to arrest in G2, progress through mitosis and are strongly sensitized to SN-38 + VE-822. Also, these processes were independent of the ATM-CHK2 pathway, as CHK2 deficient cells (as well as cells treated with the ATMi KU-55933) display a functional postreplication arrest.
Therefore, in response to SN-38 and in the absence of ATR kinase activity, DNA-PK seems to be preferentially involved in the post-replication checkpoint, preventing cells from progressing into a mitotic catastrophe. Indeed, DNA-PK deficient cells fail to phosphorylate serines 4/8 of RPA, which have been shown to protect cells from mitotic catastrophe following DNA damage 32 . Because of their non-redundant and coupled functions, ATR and ATM kinases were for a long time seen as putative co-targets for drug combination in cancer therapy 53 . Even so, our results suggest that targeting DNA-PK rather than ATM-CHK2 could better potentiate the ATR inhibition strategies, especially when coupled to a DNA damaging agent such as SN-38. www.nature.com/scientificreports/

Conclusions
In summary, we showed that the heterozygous ATR mutation is associated with high levels of endogenous replication stress, DNA damage and ATM-CHK2-KAP1 activities. Furthermore, the ATR mutation sensitizes cells to SN-38 and VE-822, both alone and combined. The increased apoptosis induction observed in mutant cells correlated with an impaired ATR-CHK1 induction and cell cycle arrest following SN-38 treatment. The highly synergistic combination SN-38 + VE-822 induces a massive chromatin recruitment of RPA in replicating cells, especially these harboring the ATR mutation. This exhaustion of soluble RPA leads a significant fraction of cells to a replication catastrophe state, while DNA-PK is responsible for their post-replicative arrest before the onset of mitosis. Thus, RPA and DNA-PK represent promising targets to potentialize the effects of ATR inhibitors coupled with DNA damaging agents. Furthermore, the significant impact of the ATR mutation in the response to SN-38 and VE-822 could open new doors in the management of ATR-mutated MSI + cancers.

Material and methods
Cell culture. HCT116 cells (NCI-DTP Cat# HCT-116, RRID:CVCL_0291) were cultured in McCoy's 5A medium supplemented with 10 mM HEPES, 10 mM Sodium Pyruvate, 1% antibiotics and 10% FBS at 37 °C, 5% CO 2 . Each batch of cells used throughout the study was cultured for max. 3 months and passaged 2-3 times a week. Cells were tested negative for mycoplasma at every thawing and once a month during their culture periods. Cells were maintained in exponential growth phase. Before any drug treatments, cells were seeded at an equivalent density and allowed to attach overnight.  s.e.m from at least 3 independent biological replicates, unless stated otherwise. One-tailed Mann-Whitney tests significances are: *: p < 0.05; **: p < 0.01; ***: p < 0.001.

ATR
Sulforhodamine B survival assays (SRB assay). 1500 cells were seeded in 96-well plates and treated in triplicate. After 72 h, cells were fixed overnight in 10% Tri-Chloroacetic Acid. Cells were washed 3 times with Milli-Q water and 50µL of 1% acetic acid-0.4% SRB were added for 30 min. Cells were washed 3 times in 1% acetic acid and SRB was resuspended in 100µL of 10 mM Tris-Base. OD was measured at 560 nm. The survival fraction (SF) (%) at each point was calculated as follows: Survival curves were plotted using GraphPad Prism. The experiments were replicated three times, standardized curves and IC 50 were generated using non-linear regression in prism 5. Western blotting. 10 6 cells were seeded in 6-well plates. After drug treatments, cells were trypsinized and washed with ice-cold PBS. Proteins were extracted in M-PER extraction buffer containing 1 mM EDTA and a proteases/phosphatases inhibitor cocktail (Thermo Fisher) (1000 rpm agitation, RT, 15 min). Debris were spun down (20 min, 13,000 rpm, 4 °C) and Protein extracts were dosed (Pierce BCA Kit, Thermo Fisher). Equivalent amounts of proteins were then diluted into loading buffer, sample reducing agent (Thermo Fisher) and Milli-Q water. Migration was performed using 3-8% (Tris-acetate buffer) and 4-12% (Bis-Tris buffer) acrylamide gradient pre-cast gels (Thermo Fisher), according to manufacturer's instruction. Transfer was performed on nitrocellulose membranes (iBlot, Thermo Fisher). Membranes were saturated in TBS-Tween 0.1%-5% non-fat milk and incubated (overnight, 4 °C) with primary antibodies diluted in TBS-0.1%Tween-5%BSA (Sigma-Aldrich). Membranes were washed 3 × 10 min in TSB-Tween 0.1% and incubated (1 h, RT) with peroxidase-conjugated secondary antibodies (Southern-Biotech), diluted at 1:3000 in TBS-0.1%Tween-5% milk. Membranes were washed 3 × 10 min in TSB-0.1%-Tween. ECL RevelBlot Plus was used and membranes were revealed in a G:BOX (Ozyme) using the Syngene software. Non-saturating blots were displayed. Full length blots are available [ Fig. S5]. www.nature.com/scientificreports/ overnight at 4 °C. Nuclei were analyzed on a Gallios Flow cytometer. A FSC/SSC graph was used to gate out debris. PI Height/PI Area graph was used to gate out doublets. 20,000 cells were analyzed for each sample on Kaluza.

Apoptosis and cell death quantification on living cells (celigo imaging cytometer).
All experiments were performed in 96-well plates (Thermo, 165,305). 1250 cells (100 µL) were seeded per well and left overnight. Cells were treated with a mixture of drugs and live stains (Live Caspase-3/7 or CellTox green reagent, at recommended dilutions). The combined use of these two kits allowed us to study the early (Caspase-3/7 activation) and later (CellTox Green) cell death events. An untreated proliferation control with no live stain was assayed for each clone in every experiment. The green fluorescence was analyzed on the Celigo in living cells at the indicated timepoints. Each experiment was repeated 5 times, and curves were plotted using GraphPad prism (errors bars: s.e.m). Acquisition and analysis settings were the following. Acquisition and analysis modes of the Celigo were set to "dead/total". Exposure time was arbitrarily set to 100,000. General settings used for the analysis were: Use well mask: 1; Well Mask: 95%; Well shape override:0. Dead frame settings:

Data availability
Raw data were generated at the IRCM -INSERM U1194. Derived data supporting the findings of this study are available from the corresponding authors A.C and T.E. on request. www.nature.com/scientificreports/