The drinking water contaminant dibromoacetonitrile delays G1-S transition and suppresses Chk1 activation at broken replication forks

Chlorination of drinking water protects humans from water-born pathogens, but it also produces low concentrations of dibromoacetonitrile (DBAN), a common disinfectant by-product found in many water supply systems. DBAN is not mutagenic but causes DNA breaks and elevates sister chromatid exchange in mammalian cells. The WHO issued guidelines for DBAN after it was linked with cancer of the liver and stomach in rodents. How this haloacetonitrile promotes malignant cell transformation is unknown. Using fission yeast as a model, we report here that DBAN delays G1-S transition. DBAN does not hinder ongoing DNA replication, but specifically blocks the serine 345 phosphorylation of the DNA damage checkpoint kinase Chk1 by Rad3 (ATR) at broken replication forks. DBAN is particularly damaging for cells with defects in the lagging-strand DNA polymerase delta. This sensitivity can be explained by the dependency of pol delta mutants on Chk1 activation for survival. We conclude that DBAN targets a process or protein that acts at the start of S phase and is required for Chk1 phosphorylation. Taken together, DBAN may precipitate cancer by perturbing S phase and by blocking the Chk1-dependent response to replication fork damage.


Results
DBAN interferes with S phase. Informed by the ability of bromoacetonitrile (12 μM) to block mitosis in Chinese hamster ovary (CHO) cells 18 , we first tested whether haloacetonitriles (HANs) impact on cell cycle progression. Wild type cells were synchronised in G2 by lactose gradient centrifugation 29 and released into medium with 10 μM bromo-, dibromo-, chloro-, dichloro-or trichloroactonitile (BAN, DBAN, CAN, DCAN, TCAN). Samples were withdrawn every 20 min to score septated G1/S cells. While the monohalogen compounds, BAN and CAN, allowed cells to complete two cell cycle rounds, their dihalogen forms, DCAN and DBAN, delayed entry into the second cycle by 40 min and 60 min, respectively (Fig. 1C,D). Unlike CHO cells 18 , DBAN-treated yeast cells arrested in G2 before the onset of mitosis (Fig. S1A,B). Since a second cycle G2 arrest is typical for drugs like hydroxyurea (HU) or camptothecin (CPT) that interfere with DNA replication, we concluded that DBAN and DCAN perturb S phase thereby triggering the G2 delay (Fig. 1G). HU stalls DNA replication forks by depleting the dNTP pool, whereas CPT breaks forks by immobilising topoisomerase 1 in front of the advancing replication complex. Both events delay onset of mitosis through the Rad3-dependent activation of the checkpoint kinases Cds1 and Chk1, respectively 27,30 . To investigate whether the DBAN-induced S phase perturbations activate the checkpoint, we synchronised a checkpoint-defective Δrad3 Δtel1 double mutant (Tel1/ATM is the second checkpoint kinase besides Rad3/ATR) in G2 and released cells into medium with or without 10 μM DBAN. Interestingly, the checkpoint deficient strain arrested like wild type for 80 min (Fig. S1A,C) showing that the G2 delay is independent of the DNA damage response. An unexpected observation was made when we analysed the cell cycle impact of trichloroacetonitrile. Unlike DBAN or DCAN, TCAN (10 μM) blocked entry into the first cycle for around 280 min (Fig. 1E). Since such a first cycle arrest is typical for agents that break DNA 30 and because HANs were linked with DNA breaks in mammalian cells 10 , we tested whether they increase the phosphorylation of histone 2AX at S129 by Rad3 and Tel1, an established marker of chromosomal breaks 31 . Intriguingly, all HANs (BAN, CAN, DBAN, TCAN), with the exception of DCAN, reduced the phosphorylation of H2AX showing that DNA breaks are not the cause of the arrest (Fig. S1D). Since H2AX is also phosphorylated during unperturbed S phase 31 , this drop may be caused by depleting the pool of S phase cells due to the G2 arrest. We can however not exclude the possibility that the HANs affect H2AX phosphorylation more directly since BAN and CAN do not show a strong cell cycle arrest but reduce H2AX modification (Fig. 1A,B; Fig. S1D). The decline in H2AX phosphorylation was concentration dependent starting at 8 μM DBAN (Fig. S1E). We next tested whether the first cycle arrest is unique to TCAN and found that a higher concentration of DBAN (20 μM) also delayed cells in the first G2 phase (Fig. 1F). This implies that TCAN is more effective than DBAN in eliciting this response. As in the case of the second cycle arrest, the DNA damage checkpoint was not required (Fig. S1F). A checkpoint-defective Δcds1 Δchk1 double mutant delayed mitosis for 220 min, although this arrest was 40 min shorter compared to wild type cells (Fig. S1F). We did not further investigate this first cycle arrest as we wanted to learn more about how DBAN perturbs S phase given the importance of DNA replication stress in malignant transformation 28 .

DBAN delays G1-S progression.
To map the execution point of DBAN in S phase, wild type cells were enriched in G1 by nitrogen starvation and released back into the cell cycle by replenishing the medium with a nitrogen source 29 . The DNA content was measured by flow cytometry over 8 hours to monitor progression from G1 (1 copy of the chromosomes, 1 C) into G2 (two copies, 2 C) (Fig. 2). While untreated cells reached G2 4 h post-release ( Fig. 2A), DBAN (10 μM) delayed exit from G1 whereas TCAN (10 μM) blocked cells in G1 (Fig. 2B,C). We also arrested cells in early S phase with 15 mM HU to have an internal marker for unreplicated DNA (Fig. 2B). Eight hours after the release from G1, only HU and TCAN treated cells remained arrested with a 1 C DNA content, while DBAN permitted the completion of DNA replication (Fig. 2B,C). This shows that DBAN only delays G1-S transition, whereas TCAN blocks this step more effectively. None of the other HANs (BAN, CAN, DCAN) had a similar effect on G1-S transition (Fig. 2B,C).
To exclude the possibility that the G1 arrest protocol impacts on this interesting finding, we synchronised nda3-KM311 cells in mitotic prophase. This cold sensitive beta-tubulin mutant stops with condensed chromosomes without a mitotic spindle at 20 °C and returns to the cell cycle within minutes upon a temperature up-shift to 30 °C 32 . In line with the first experiment, TCAN prevented the accumulation of G2 cells whereas DBAN only delayed it (Fig. S1G).
We next arrested cells in early S phase by incubating a wild type strain in 15 mM HU for 3.5 h 29 to test whether addition of DBAN or TCAN (10 μM) after the G1-S transition would still delay the cell cycle. While the latter was not the case for DBAN (Fig. S2B), TCAN-treated cells delayed progression through S phase by 20 min compared Wild type cells (ade6-M210 leu1-32 ura4-D18) were synchronised in G2 by lactose gradient centrifugation 29 and released into rich medium (3% glucose, 0.5% yeast extract, 100 mg/L adenine) without (UT = untreated) or with 10 μM of the indicated haloacetonitriles (HANs). All HANs were diluted from a 12 mM stock solution in DMSO to a final concentration of 10 μM.  to untreated cells (Fig. S2C). Since DNA replication was complete within 60 min in untreated cells (Fig. S2A), a 20 min difference (1/3 of S phase) may well be significant.
Taken together, these data show that DBAN delays G1-S transition while allowing cells to complete DNA replication. In contrast, TCAN blocks cells effectively before the G1-S transition and slightly delays DNA replication. This conclusion is in line with the higher potency of TCAN as a G2 blocker (Fig. 1E). DBAN affects DNA polymerase delta. Since DBAN allows DNA replication to proceed after an initial G1-S delay (Figs 2C, S1G), we tested whether this affects the three replicative DNA polymerases, alpha (Pol1, swi7), epsilon (Pol2, cdc20) or delta (Pol3, cdc6). Since these essential genes cannot be deleted, we used temperature-sensitive mutants at the semi-restrictive temperature of 30 °C. Serial dilutions of the strains were applied to rich medium plates containing no HAN or 10 μM DBAN. We also incubated one plate at 37 °C to confirm the temperature sensitivity. While mutations in the three essential subunits of Pol delta (cdc6.23 [catalytic], cdc27.P11 [non-catalytic], cdc1.P13 [non-catalytic]) impaired cell viability, mutations in the catalytic subunits of Pol epsilon (cdc20.M10) or Pol alpha (swi7.H4) did not (Fig. 3B). Interestingly, deletion of the fourth, non-essential Pol delta subunit, cdm1, had also no effect (Fig. 3C). We next tested mutations in the MCM [2][3][4][5][6][7] helicase that unwinds the DNA template (MCM2 [cdc19.P1], MCM4 [cdc21-M68] or MCM5 [nda4-108]), but failed to detect loss of viability (Fig. 3C). Also no impact on cell growth was found for mutations in DDK (Cdc7/ Hsk1) kinase (hsk1-1312) and in the replication factor Rad4 (TopBP1) (rad4.116) (Fig. 3C). Whether a mutation in Ctf4 (mcl1-1), which binds Pol alpha to the replication complex, impairs cell viability was difficult to judge since the strain grew very poorly even in the absence of DBAN (Fig. 3B). Interestingly, none of the DNA pol delta mutants showed a growth defect on TCAN plates even at 20 μM (Fig. 3D). This was unexpected given the high impact of TCAN on cell cycle progression. A possible explanation is provided by the replication delay caused by TCAN (Fig. S2C) that may prevent loss of viability of pol delta mutants. The latter mutants were also not sensitive to CAN, BAN or DCAN at 10 μM (Fig. 3E).

DBAN overcomes the intra-S arrest of a pol delta mutant.
To find out why pol delta mutants are DBAN sensitive, we synchronised wild type, pol alpha (swi7.H4), pol delta (cdc27.P11) and pol epsilon (cdc20. M10) strains in early S phase using the HU arrest protocol 29 . After HU was washed out, cells were released into medium with or without 10 μM DBAN. Flow cytometry showed that untreated wild type cells completed S phase within 60 min (Fig. 4A). As previously reported 33 , the untreated mutant strains delayed S phase progression at the semi-permissive temperature of 30 °C (Fig. 4B-D). While DBAN had no impact on S phase within the first 60 min in the case of the pol alpha and pol epsilon mutants, it did significantly advance DNA replication of the pol delta (cdc27.P11) strain (Fig. 4C). This advancement was clearly detectable at the 60 min and 90 min time points. After 2 h, the DBAN-treated cdc27.P11 cells had initiated already the next cell cycle round compared to the untreated sample (Fig. 4C). Cdc27 connects the catalytic (Cdc6) and non-catalytic (Cdc1) subunits, and binds Pol delta to the DNA sliding clamp PCNA 34 . DBAN also advanced DNA replication in the other two mutant strains but later and to a lesser extent (Fig. 4B,D).
These results imply that DBAN abolishes the intra-S phase arrest of the pol delta (cdc27.P11) strain, which may be linked with its loss of viability on DBAN plates (Fig. 3B). Since the viability of pol delta mutants depends on Chk1 kinase 33,35 we next tested whether DBAN interferes with the activation of the checkpoint kinases Cds1 and Chk1.
DBAN suppresses the activation of Chk1. To find out whether DBAN perturbs activation of the intra-S checkpoint kinase Cds1 at stalled forks, asynchronous cds1-His 6 HA 2 cells 36 were incubated with 10 μM DBAN, 12 mM HU or with both chemicals simultaneously for 4 h. Total protein extracts were loaded onto a 6% phos-tag SDS gel to assay the phosphorylation status of Cds1. Phostag electrophoresis reveals the phosphorylation pattern of proteins as their mobility is inversely related to the extent of their modification 37 . While DBAN did not promote the modification of Cds1, the kinase was intensively phosphorylated when DNA replication forks stalled in the presence of HU. Although DBAN did not impact on this hyperphosphorylation, it induced a faster migrating band (Fig. 5A). The latter band could be a hypophosphorylated form of Cds1.
We then repeated this experiment with a chk1-HA 3 strain 38 but replaced HU with 12 μM camptothecin (CPT) to break DNA replication forks. In contrast to Cds1, DBAN effectively suppressed Chk1 phosphorylation at serine 345 (Fig. 5B). On normal SDS page, S345 phosphorylation was detected as a band shift, as previously reported, which disappeared upon DBAN exposure (Fig. 5B) 38 . DBAN also induced very slowly migrating, hyperphosphorylated bands of Chk1 independently of CPT. The suppression of Chk1 phosphorylation at damaged replication forks could explain why DBAN impairs the viability of DNA polymerase delta mutants which rely on this kinase for survival (Fig. 3B) 35 .
To test whether DBAN is an inhibitor of Rad3 kinase, we replaced CPT with MMS (methyl-methanesulfonate) that damages DNA by alkylation 39 . Since DBAN did not to block the MMS-induced phosphorylation of Chk1 (Fig. S1H), it is unlikely that the HAN impairs Rad3 kinase directly. We then exposed wild type cells (chk1-HA 3 ) to CPT or to the combination of DBAN and CPT on plates to gauge whether replication forks still break. The latter seems to be the case as DBAN rendered wild type cells (chk1-HA 3 ) CPT sensitive (Fig. 5D). This increase in sensitivity, which was similar to the sensitivity of a kinase-dead chk1 mutant (chk1-D155E-HA 3 ), indicates that forks still break while DBAN prevents activation of Chk1. We can however not exclude the possibility that CPT generates a block to DNA replication like a mutation in DNA polymerase delta that renders cells sensitive to DBAN.
To find out when DBAN acts on Chk1 in the cell cycle, we synchronised chk1-HA 3 cells in early S phase with 15 mM HU and released them into medium with CPT (12 μM) or with CPT and DBAN or TCAN (10 μM). In line with the idea that Chk1 is activated once bulk DNA synthesis had been completed 40 , the shift band of Chk1, indicative of serine 345 phosphorylation, appeared 60 min post-release after the levels of the DNA replication marker Mrc1 had declined (Fig. 5E) 24 . Since Chk1 is weakly phosphorylated during the HU arrest, all samples displayed a weak shift band that increased strongly when forks were damaged by CPT (Fig. 5E, bottom panel). The presence of DBAN or TCAN effectively suppressed activation of Chk1 (Fig. 5F). Since Rad3 modifies also the Rad9 subunit of the 9-1-1 checkpoint ring at broken forks (Fig. 5C) 41 , we repeated this experiment with a HU-synchronised rad9-HA 3 strain 42 . As in the case of Chk1, Rad9 phosphorylation is detectable as a band shift. This shift was not affected by DBAN strongly suggesting that Chk1 is specifically targeted by the haloacetonitrile (Fig. 5G). We noticed however that the Rad9 phosphorylation peaked 30 min earlier in the presence of DBAN (Fig. 5G, panels 3 + 4). The phos-tag assay did not reveal any changes in the Rad9 phosphorylation pattern which were DBAN specific (Fig. S1I).
Collectively, these results suggest that DBAN kills pol delta mutants (Fig. 3B) and renders wild type cells sensitive to camptothecin (Fig. 5D) by preventing the phosphorylation of Chk1 at damaged DNA replication forks.

Discussion
The evidence presented here reveals novel activities of DBAN at two stages during the cell cycle, at the G1-S transition and later at damaged replication forks (Fig. 6). DBAN affects both processes in a negative way as it delays entry into S phase (Fig. 2) and suppresses phosphorylation of Chk1 (Fig. 5B,F) without affecting the activation of Rad9 (Fig. 5G) or Cds1 (Fig. 5A). Replication fork damage can originate from the inhibition of topoisomerase 1 by camptothecin or from mutations in the lagging strand DNA polymerase delta 26,33 . Collectively, these findings imply that DBAN blocks an event or protein that is required for both entry into S phase and activation of Chk1 Scientific REPORTs | 7: 12730 | DOI:10.1038/s41598-017-13033-8 at damaged forks (Fig. 6). Early in S phase, human and yeast cells phosphorylate the histone H2AX in a cell cycle specific manner independently of DNA breaks 31,43 . The same chromatin modification is later required for the recruitment of Crb2 (53BP1) to a broken fork where the scaffold protein associates with Chk1 44,45 . DBAN and TCAN may therefore compromise both processes by reducing H2AX phosphorylation (Fig. S1D,E). For example,  29 . HU was washed out and cells were released into pre-warmed rich medium. The DNA content was measured at the indicated times. The dotted lines indicate 1 copy of the chromosomes (1 C, G1) and two copies (2 C, G2), respectively. The DNA content of nitrogen starved cells (no N) was measured in a parallel experiment to have an internal standard. (B-D). The indicated mutant strains were HU-synchronised and released into rich medium at 30 °C with or without 10 μM DBAN. The green histogram is the DNA content of untreated cells, the red histogram is the DNA content in the presence of DBAN. The brown colour indicates that both histograms overlap.
the HANs may up-regulate a phosphatase, like human PP2A-B56ϵ, that dephosphorylates H2AX at damaged forks upon CPT treatment 46 . What argues against this model is the survival of the DNA pol delta mutants in the presence of BAN, CAN or TCAN (Fig. 3) which all reduce H2AX phosphorylation (Fig. S1D). An alternative explanation is provided by the dual function of Rad4/Cut5 (TopBP1) during G1-S transition and in the activation of Chk1 47,48 . Rad4 associates with Sld3 and Sld2/Drc1 at start and with Crb2 at broken DNA 45,47 . What argues however against Rad4 is its requirement for the activation of Cds1 at stalled forks 48 which is not affected by DBAN (Fig. 5A). A third candidate is DDK kinase as it initiates the assembly of the replication complex at the end of G1 (reviewed in 19 ) and terminates Chk1 activation by phosphorylating the Rad9 subunit of the 9-1-1 complex 49 . We have however not found any evidence that the Rad9 phosphorylation pattern changes upon DBAN exposure (Fig. S1I). A fourth possibility is provided by Mrc1 (Claspin) that associates with early replication origins at start 25 and recruits human Chk1 to broken forks 50 . A similar interaction between S. pombe Chk1 and Mrc1 has not yet been reported. Finally, DBAN may act directly on Chk1 as indicated by its hyper-phosphorylation (Fig. 5B). Chk1 is required in S. pombe and human cells at the start of the G1-S transition 51,52 and at broken forks 26 . Further work will however be required to dissect the different possibilities that are of great interest as it is still enigmatic why human Chk1 occupies a dominant role in S phase whereas yeast Chk1 appears to act mainly in G2 (Fig. 5E) 40 .
DBAN is known to block a large number of enzymes in vitro 7 but it is as yet unclear how it interacts with proteins. Halogen atoms like bromine and chlorine have a positive charge, known as the alpha-hole, that can make electrostatic contacts with the protein backbone or amino acid side chains 53 . While bromine interacts preferentially with the side chain of arginine, chlorine prefers leucine 54 . Whether this explains why TCAN is a stronger cell cycle inhibitor than DBAN (Fig. 1) is difficult to tell as it is unclear to which protein they bind. It is even not entirely safe to conclude that they bind to the same protein. DBAN and TCAN behave in a similar way regarding the G1-S delay (Fig. 2) and the inhibition of Chk1 phosphorylation (Fig. 5F). Both HANs differ however in their lethality when DNA polymerase delta is mutated (Fig. 3D). Although this appears to contradict the earlier notion that DBAN kills these mutant strains by blocking Chk1 phosphorylation at damaged forks, it could be explained by the ability of TCAN to delay DNA replication (Fig. 4C). If this delay were to prevent fork damage in pol delta mutant cells, the inhibition of Chk1 would not affect cell viability.
Although DNA replication stress is a good explanation for why DBAN (10 μM) triggers a second cycle delay (Fig. 1D), it would not provide an answer to why 20 μM DBAN block cells in the first G2 (Fig. 1F). The concentration dependency implies that DBAN has either more than one target in cells with different affinities for the HAN or that DBAN upregulates the expression or activity of a protein in a concentration dependent manner. From the different options discussed earlier, the up-regulation of a phosphatase like PPA2 would connect the diverse cell cycle activities of DBAN and TCAN. Interestingly, induction of PPA2 arrests cells in G2 independently of the DNA damage checkpoint when the accessory protein Vpr of human immunodeficiency virus type 1 (HIV-1) is over-expressed in S. pombe 55 . Since this arrest shares a similar independence from the Rad3-Tel1 checkpoint as the G2 arrest produced by DBAN and TCAN (Figs 1E,F; S1F), it is quite possible that both HANs delay cell cycle progression in G2 through a mechanism that involves the activation of a phosphatase acting on the cell cycle machinery.
The final point to consider is whether DBAN, which is frequently found in water supply systems 2 , posses a serious cancer risk. The concentration of 10 μM used in this study is approximately 30 times higher than the WHO guideline of 0.35 μM (70 μg/L). It is therefore unlikely that DBAN levels reach such high concentrations in tap water. The peak concentration found in water supply systems in Western Australia, for example, was 0.13 μM (26.6 μg/L) 56 , whereas the peak value in the United Kingdom was with 0.04 μM (8 μg/L) much lower 2 . It is however not yet clear whether haloacetonitriles accumulate over a longer consumption period in the liver, gastrointestinal tract or the kidneys 7 . The latter may explain why the consumption of chlorinated drinking water was linked with cancer 3 .

Materials and Methods
Yeast Strains. The genotype of the strains used in this study is ade6-M210 leu1-32 ura4-D18. Wild type cells contained no additional mutation and the different mutant alleles are mentioned in the text. The cds1-HA 2 His 6 [URA4 + ] ura4-D18 strain is described in 36 and the chk1-HA 3 strain in 38 . Before cells were synchronised in G1 by nitrogen starvation, all auxotrophic markers were crossed out.
Cell synchronisation. Cells were synchronised as described in 29 . HU was used at a final concentration of 15 mM for 3.5 h at 30 °C in rich medium. For the G1 arrest, cells without auxotrophic markers were first grown in minimal medium with nitrogen before being transferred to minimal medium without nitrogen for 16 h at 30 °C. Lactose gradients were centrifuged for 8 min at 800 rpm. The nda3-KM311 mitotic arrest was performed in rich medium as reported in 57 . Flow cytometry. The DNA content was measured using a CUBE 8 (Sysmex) instrument as described in 29 .