The RecQ helicase WRN is required for normal replication fork progression after DNA damage or replication fork arrest

Cells and culture

Primary human dermal fibroblasts were obtained from Clonetix and used at passages 14–18. SV40-transformed GM639 and GM847 fibroblast cell lines were obtained from the Coriell Institute Cell Repositories (Camden NJ). GM639cc1 is a pNeoA-carrying derivative of GM639²⁴.

Drugs and Dyes

Stock solutions of 5-bromodeoxyuridine (BrdU; 10 mM in water), 5-iododeoxyuridine (IdU, 2mM in water), 5-chlorodeoxyuridine (CldU, 10mM in water), caffeine (100mM in DMSO), nocodazole (0.5mg/ml in DMSO), and hydroxyurea (HU, 1M in PBS) were stored at –20° C until use. MMS was diluted in PBS or water to 1–5% prior to use. Mimosine (10mM in growth media) was stored at 4° C. Propidium iodide (10mg/ml in PBS) was stored at 4° C and 4,6′-diamidino-2-phenylindole (DAPI, 1mg/mL in water) was stored at –20° C. DAPI was obtained from Accurate Chemical and Scientific Corp (Westbury, NY), and the rest of the chemicals from Sigma (St. Louis, MO).

Cell growth, synchronization and drug treatments

All cell strains or lines were grown as adherent monolayers in Dulbecco Modified Minimal Essential Medium (DMEM) supplemented with 2 mM L-glutamine and 10% fetal bovine serum (Hyclone, Ogden, UT) in a humidified 5% CO₂, 37°C incubator.

To synchronize cells, subconfluent cultures were treated with 0.5mM mimosine for 12–14 hours²⁵. Cells were then washed in PBS and incubated with fresh media for 8–10 hours. At this time, cells released from the block reached mid to late S phase, and were treated with HU or MMS for replication assays or cell cycle recovery experiments. HU was added at 2mM, and MMS was added at 0.005% (0.55mM) for 1 hr unless stated otherwise. BrdU and other halogenated nucleotides were added to cells to a final concentration of 50μM, caffeine was used at 3mM, and nocodazole at 50ng/ml.

RNAi-mediated depletion of WRN

WRNsi is a hairpin with the stem sequence corresponding to positions 160 to 184 in the WRN ORF (accession # NM_000553; 1 is A in ATG). It was expressed from the pBABEpuro retroviral vector²⁶. WRN2-4 is a hairpin with the stem sequence corresponding to WRN ORF positions 578 to 597. This hairpin was cloned into pLKO.1 lentiviral vector²⁷ between EcoRI and AgeI sites, under the control of the human U6 promoter. Virus was generated by transient transfection of 293T cells. Human fibroblasts were transduced with pBABEpuro-WRNsi, pLKO.1-WRN2-4, or the respective empty vectors, and placed on puromycin selection (1μg/ml for primary fibroblasts and GM847 and 1.5μg/ml for GM639) at 20 hours after infection.

Western blotting

Cells were harvested for Western blot analysis and cell cycle and replication assays at 5–7 days post infection. Western blotting of WRN was done as previously described²⁴, with the rabbit α-WRN (Novus Cat. #NB100472A). For loading controls, α-Chk1 (Santa Cruz Cat. #sc-8408) antibody was used against CHK1 and α-GAPDH (Abcam Cat. # ab9482) was used against GAPDH. Proteins were visualized by ECL (Amersham) and quantified using the Storm Phosphorimager and ImageQuant software (Molecular Dynamics).

Staining for BrdU incorporation and FACS

DNA of ethanol-fixed cells was denatured with 2N HCl and 0.5% Triton X100, and neutralized in 100mM Na borate pH8.5. Cells were next washed in IFA buffer (10mM HEPES pH7.4, 150mM NaCl, 5% normal goat serum, 0.1% Na azide) supplemented with 0.5% Tween 20. Cells were resuspended in IFA containing 10μL of FITC-conjugated mouse αBrdU antibody (Beston-Dickinson) per 10⁶ cells and incubated on ice in the dark for 1 hr. Cells were washed again with IFA/Tween and resuspended for FACS analysis in PBS with 10μg/mL propidium iodide and 100μg/mL RNAse A.

DAPI staining was performed as previously described²⁸.

Data analysis and presentation were done with Summit software (Dako, Carpinteria, CA), and cell cycle phase quantitations were done with Mcycle software (Phoenix Flow Systems, San Diego, CA). To quantify BrdU incorporation, the position of the BrdU-negative population was determined using flow cytometric profiles of relevant negative control samples with no incorporation. Cells with fluorescence above the negative control level were considered positive.

Microchannel fabrication, DNA fiber stretching and replication track analysis

PDMS microchannels were fabricated using standard photolithography and soft lithography procedures²⁹. SU8-2 was spun on silicon wafers at 500 rpm for 10 sec. and 3000 rpm for additional 30 sec. The heights of microchannels were 3.25 μm as measured with a surface profilometer (KLA-Tencor, model P15, San Jose, CA). DNA stretching was performed as described³⁰, with modifications. Glass cover slips were cleaned in nitric acid:hydrochloric acid 2:1, silanized with a solution of 12.6 μl N-trimethoxysilylpropyl-N,N,N-trimethylammonium chloride (Gelest, SIT8415.0) and 6 μl vinyltrimethoxysilane (Gelest, SIV9220.0) per 50ml of water at 65°C for 17.5 hrs, and stored in ethanol at 4°C. DNAs from cells harvested by trypsinization were isolated in agarose plugs as described in manufacturer recommended protocols for the CHEF-DR II PFGE apparatus (Bio-Rad), with the exception that MMS-treated DNAs and their control counterparts were processed at temperatures at or below 37°C. To release DNA, plugs were briefly heated to 75°C and incubated with β-agarase (NEB). To stretch DNA, oxygen plasma-treated (to decrease hydrophobicity) PDMS matrices with series of microchannels 50–450 μm wide were laid over cover slips, and DNA was loaded into channel space by capillary tension. PDMS matrices were then removed and cover slips were treated with methanol:acetic acid 3:1 for 3 min and air dried. DNA was stained with YOYO-1 (Molecular Probes, 10mM in TE with 20% β-mercaptoethanol) to inspect the quality of stretching. For immunostaining, cover slips were incubated in 2.5N HCL for 40 min, neutralized in 0.1M Na borate pH 8.0 and PBS, and blocked in PBS/5% BSA/0.5% Tween 20. The following antibodies were used, in that order: rat α-CldU/BrdU (Serotec, Cat.# MCA2060), goat α-rat Alexa 594-conjugated (Molecular Probes, Cat.# A11007), mouse α-IdU/BrdU (BD Biosciences, Cat.# 347580), and goat α-mouse Alexa 488-conjugated (Molecular Probes, Cat.# A11001). Antibody dilutions were made in PBS/5% BSA/0.5% Tween 20 with 10% normal goat serum and cover slips were also blocked in this buffer between α-rat secondary antibody and mouse α-IdU/BrdU antibody. Washes between antibodies were done in PBS/1%BSA/0.1% Tween 20. Cover slips were mounted in 10% PBS/90% glycerol/10mM DTT.

Confocal microscopy of stretched DNAs was performed on the Zeiss Axiovert microscope with a 100x objective. Scanning for areas with optimal density of molecules was done with the filter set to the color corresponding to the first label, and then ten to twenty digital images per channel were generated by tracking along their lengths. Lengths of tracks were measured in digital images using the attached Zeiss AxioVision software. Chi square tests were applied to frequency distributions of lengths of tracks. The ratios of first label to second label segment lengths in double-labeled tracks were analyzed with two nonparametric tests: frequency distributions were subjected to chi square tests, and datasets of ratio values were also subjected to Kolmogorov-Smirnov (KS) test. This was done to eliminate the risk that variations between tail end values in the ratio distributions will disproportionately influence confidence.

WRN depletion delays recovery from DNA damage during S-phase

In order to determine the role of WRN in recovery from replication stress we found it necessary to use cell populations maximally enriched for S phase cells. Average tissue culture cell populations are abundant in G1 cells, which are not affected by HU and are differentially affected by camptothecin, cis-platinum, and MMS. Also, many WRN−/− cell line cultures have an increased fraction of G1 cells compared to WRN+/+ lines²⁸, which can further dilute effects produced by replication stress in WRN-deficient cells.

In order to avoid these potential pitfalls, we used assays that focused on S phase cells (see experimental design in Figure 1A). First, we used synchronization to enrich cell populations for S phase fraction. Cell cultures were arrested in late G1 (G1(I) in Figure 1A) by treatment with mimosine. Then mimosine was removed and the cells were incubated for 8–10 hrs. At this time, when the bulk of the population was in S phase, cells were pulse-treated with MMS, and followed through the time course of recovery.

To compare isogenic cell lines, we depleted WRN from human fibroblasts. We first used a previously characterized retroviral shRNA, WRNsi^{26, 28}. In addition, we developed a new, lentivirus-expressed shRNA, WRN2-4. Typically, 80–90% of WRN protein was depleted (Supplementary Figure 1A), and this level was maintained throughout the experiment.

In the absence of DNA damage, WRN-depleted and mock-depleted SV40-transformed fibroblasts finished the cell cycle and returned to G1 (G1(II)) with similar kinetics (Figure 1B). Treatment of cells with MMS during S phase induced cell cycle delays in both WRN-depleted and control cells (Figure 1B). For the first 10 hrs after MMS, cells traversed through S and G2 phases (not shown). At later time points, cells began to accumulate in G1(II), with WRN-depleted cells markedly delayed compared to controls (Figure 1B). Addition of a mitosis inhibitor nocodazole to cells recovering from MMS confirmed that not only progression from G2/M to G1 was delayed in WRN-deficient cells, but also the completion of S phase (not shown).

The same experiment was performed using a different shRNA (WRN2-4, Supplementary Figure 1B) and with the addition of BrdU labeling step prior to MMS treatment (see experimental design in Figure 1A). BrdU was present between 0 and 8–10 hrs after release from mimosine, thereby labeling every cell that entered S phase during this interval. This BrdU+ population was identified in FACS profiles by comparison with BrdU− controls (mimosine-arrested cells incubated with BrdU), and the cell cycle progression of this population was followed for about 20 hrs (see an example in Figure 1C). BrdU incorporation allowed distinguishing between BrdU− cells remaining in G1(I), and BrdU+ cells reaching G1(II) (an arrowhead in Figure 1D, right panel), as well as cells in first and second S phases. Once again, WRN-depleted SV40 fibroblasts were slower than controls in completing the cell cycle after MMS treatment in S phase (Supplementary Figure 1B). This effect of WRN depletion on recovery from MMS was even more pronounced in a different SV40 fibroblast line, GM639 (Figure 1D). For example, 13 hrs after MMS treatment, control cells were largely in G1(II), while WRN-depleted cells remained predominantly in the first G2 (time point 24 hrs in Figure 1D). A large proportion of WRN-depleted population remained in the first G2, while the bulk of control cells traversed into the second S phase (time point 32 hrs in Figure 1D).

To rule out a possibility that the difference between WRN-deficient and wild type cells was induced by mimosine, we preformed experiments in an asynchronous population (Figure 2A, B). WRN-depleted and control fibroblasts were pulse-labeled with BrdU for two hours. MMS was added during the second hour of BrdU labeling and then both the drug and the nucleotide analog were removed. Cell cycle distributions prior to (all cells) and during MMS treatment (BrdU+ only cells) in both types of cells were similar (Figure 2A). However, by 19 hrs after MMS, the progression of WRN-depleted cells that were in S phase during MMS treatment lagged behind controls (Figure 2A, B) in the generation of BrdU+ G1 cells. Addition of caffeine during recovery shortened the delay of cell division in both types of cells (Supplementary Figure 1C). Thus MMS-induced delays were at least in part mediated by the ATR/ATM-dependent checkpoint. As before, without MMS treatment, WRN-depleted cells were not slower than controls in their progression from S phase to the next G1 (Supplementary Figure 1C).

Taken together, the data obtained with both synchronized and unsynchronized SV40 fibroblasts indicate that loss of WRN causes a marked extension of the checkpoint-dependent delay of cell division induced by MMS treatment during S phase. This extension arises from prolonged pausing of cells in late S and/or G2 phases of the cell cycle.

WRN depletion impairs recovery from HU-mediated replication arrest

In order to determine if the defect in recovery from replication stress observed in WRN-depleted cells was MMS-specific or more general, we treated WRN-depleted cells with HU. The experimental scheme was as shown in Figure 1A, except that HU was added to BrdU-labeled cells instead of MMS (BrdU was not present during incubation with HU). After 8 hrs of HU arrest, we could detect an approximately 4–5 hr delay in the kinetics of progression of WRN-depleted cells out of S phase and into the next G1, compared to controls (not shown). This WRN-dependent delay became more prominent after 11–13 hrs of HU arrest in S phase (Figure 2C, D). For example, 12 hrs after HU (and 34.5 hrs after the start of the experiment), some control cells entered G1(II) while WRN-depleted cells remained in G2 (Figure 2C, panel 34.5 hr, also see graph in Figure 2D) Thus, WRN loss impairs recovery from replication arrest induced by HU.

To address whether the transformed status of SV40 fibroblast lines affects the WRN-dependent phenotype, we repeated the HU-mediated arrest and recovery experiments in primary human fibroblasts. WRN-depleted primary fibroblast cultures (see Supplementary Figure 1A for depletion data) had significantly fewer cells that entered S phase after mimosine synchronization (Figure 3B, upper panels). This phenotype is unlikely to be caused by mimosine treatment, since untreated WRN-depleted cultures also had fewer S phase cells (not shown). Also, previous work has described an accumulation of G1 cells in WRN-depleted primary fibroblasts^{28, 31}. Interestingly, WRN-depleted primary fibroblasts that entered the cell cycle after mimosine synchronization did so on the same schedule as controls (compare 9 hr panels in Figure 3C). However, the appearance of BrdU+ G1(II) cells that completed the cell cycle was delayed in WRN-depleted fibroblasts, with more cells staying in late S/G2 (Figure 3A, 17 to 22.5 hr panels, also Figure 3D). This phenotype is consistent with the previous observations made with WRN−/− primary fibroblasts¹⁹, and suggests that WRN loss caused an extension of S/G2 phases.

HU treatment during S phase delayed cell cycle progression of both mock-depleted and WRN-depleted primary fibroblasts (Figure 3B, C, D). However, 11.5 hrs after release from HU block when 10% of control population reached the subsequent G1, WRN-depleted cells remained in late S and/or G2, and no G1(II) cells could be detected (Figure 3C, panel 34hr, also Figure 3D). Thus, WRN depletion from both primary and SV40-transformed human fibroblasts impairs recovery from HU-mediated replication arrest. Taken together, the data obtained thus far suggest a general defect in recovery from replication stress caused by depletion of WRN.

WRN depletion affects fork progression on MMS-damaged DNA

The defect in recovery from replication stress in WRN-depleted cells could have at least two distinct causes. First, WRN may stabilize or restore active replication forks that are slowed or stalled due to DNA damage or nucleotide depletion. In this case, when WRN is depleted, the fraction or activity of forks that are able to elongate after replication stress should be reduced. Alternatively, WRN may be involved in the repair of DNA strand breaks that emerge after irreversible breakdown of replication forks. In this case, WRN depletion should not have impact on the fate of replication forks, but should reduce efficiency of DNA repair in S and G2 phases.

In order to determine whether WRN affects fork activity, we looked at the behavior of DNA replication in living cells by using immunofluorescent detection of replication tracks in stretched DNA fibers. DNA was stretched on silanized glass with the aid of micro-fabricated capillary channels (see Materials and Methods). The experimental design is outlined in Figure 4A: mimosine synchronized, mid S-phase SV40-transformed fibroblasts were labeled for 40 min with CldU, treated with 0.02% MMS for 20 min, and then labeled for an additional 40 min with IdU (in a separate experiment, we established that virtually no replication can be detected by fiber track analysis if the label is added during incubation with this dose of MMS, not shown). Untreated controls were consecutively labeled with CldU and IdU. DNAs were isolated in agarose plugs and all treatments were conducted at temperatures at or below 37°C to minimize potential breakdown of thermolabile alkylated bases³². Replication was detected by staining with antibodies to IdU or CldU. Since MMS treatment has been shown to reduce the rate of progression of replication forks^{33, 34}, we first focused on measuring the lengths of the IdU and CldU segments in double-labeled tracks (Figure 4A, track type a), as well as the lengths of IdU and CldU single-labeled tracks (Figure 4A, track types b and c). In addition, we determined the ratios of CldU to IdU segment lengths in double-labeled tracks (see³⁵ for an in-depth discussion of the approach).

In our experimental design (Figure 4A), CldU and IdU track length distributions of untreated cells should be the same, while MMS treatment should shorten the lengths of tracks labeled with IdU (post-damage) but not with CldU (pre-damage, Figure 4B). In addition, CldU track lengths in untreated and MMS-treated DNAs should be similar. All these expectations were met: in untreated WRN-depleted and control fibroblasts, CldU and IdU track lengths were similar, and the ratios of CldU to IdU segment lengths of double-labeled tracks were distributed around 1 (Figure 5A, C). This confirms the assumption that double-labeled tracks are in fact generated by single forks (Figure 4A). Also, single-labeled tracks were overall longer than segments of double-labeled tracks of the same color (Figure 5A). Again, this is consistent with the interpretation that a vast majority of double-labeled tracks is produced by single replication forks, while single-labeled tracks can be produced by two diverging (or converging) forks.

Treatment of control cells with MMS between CldU and IdU pulses resulted in shortening of IdU (post-damage) but not CldU (pre-damage) tracks, as seen both by track length distributions (Figure 5B) and by a shift of a bulk of CldU/IdU ratios of double-labeled tracks to values above 1.5 (Figure 5C, +MMS panels). This indicates a decrease in fork progression rates after MMS, consistent with previous reports^{33, 34}. In WRN-depleted fibroblasts, this effect of MMS on fork rates was exacerbated. All IdU track lengths, whether by newly fired forks (single-labeled, Figure 5B, top panel), or by ongoing forks (double-labeled, Figure 5B bottom panel), were significantly shorter in WRN-depleted cells after MMS than in controls, whereas CldU track lengths were virtually identical. Moreover, WRN-depleted cells had more forks in which post-damage progression was severely depressed. For example, in WRN-depleted cells 16% of ongoing forks had post-damage segments that were 10 or more times shorter than their pre-damage segments, compared to only 6% in control cells (Figure 5C, P<0.01 by chi square test and D=0.325, P<<0.001 by KS test; also see Table 1).

This prompted us to examine whether WRN-depleted cells also had a higher proportion of forks that failed to elongate after MMS damage. Results of experiments performed with the original (CldU, then IdU) and a reversed (IdU, then CldU) order of addition of labels, were averaged to derive a mean value for the recovery of ongoing, e.g. double-labeled, forks. The labeling order was varied in order to eliminate the risk of a potential bias in track type identification caused by antibody staining artifacts. The results consistently showed a more dramatic MMS-dependent reduction in the fraction of ongoing forks in WRN-depleted cells (Figure 5D).

Taken together, our data suggest that MMS damage resulted in more dramatic fork slowing in WRN-depleted cells, as compared with controls. This was evident both in ongoing forks and in forks originated after MMS treatment. Moreover, slowing of ongoing forks in WRN-depleted cells was accompanied by a higher degree of apparent fork inactivation.

WRN depletion impairs fork progression, though not fork reactivation, after HU arrest

Next, we asked whether forks stalled by HU were also inactivated more readily in cells lacking WRN. Synchronized, mid-S phase SV40 fibroblasts were pulse-labeled with the first halogenated nucleotide (CldU or IdU) for 40 minutes, and then HU was added to the media for 4 hrs (in the presence of the first label), prior to transfer to media containing the second label (respectively, IdU or CldU) and no HU (Figure 6). Untreated controls were sequentially labeled with two labels for the same periods of time.

Forks arrested with HU are thought to remain functional, at least for a period of time, unless cells lack protein factors that preserve activity of arrested forks^{36, 37}. Therefore, in order to determine whether WRN plays a role in stabilizing arrested forks, we first determined the percentages of double- and single-labeled tracks in WRN-depleted and control SV40 fibroblasts before and after HU arrest.

Treatment with HU for 4 hrs resulted in a decrease in recovery of ongoing forks, even though a substantial number of forks that were labeled prior to HU addition, was still able to resume DNA synthesis and incorporate the second label (Figure 6B). Remarkably, WRN-depleted cells were not different from controls in their ability to preserve activity of HU-arrested forks. An average 1.6-fold reduction in recovered forks was observed in WRN-depleted cells compared to a 1.4-fold reduction in controls. We next asked if the forks that resumed elongation after HU-mediated arrest progressed slower in WRN-depleted cells than in controls. Track length analysis of the data from one of the experiments analyzed in Figure 6B, showed that this was the case. For example, 29% of the forks that resumed replication after HU in WRN-depleted cells, traveled less than a fifth of the distance they traveled before HU (Table 1). By comparison, in control cells only 7.2% of forks that resumed after HU fell into this category.

We also exposed SV40 fibroblasts to a longer HU arrest (7 hr), and measured the lengths of double-labeled tracks (in this case, IdU-BrdU, Figure 6C, D). The analysis confirmed that forks resuming replication after HU-mediated arrest covered on the average less distance in WRN-depleted cells than in controls. For example, overall length distributions of post-HU segments of double-labeled tracks were significantly shorter in WRN-depleted cells (Figure 6C). Also, the ratios of pre- to post-HU segments in double-labeled tracks were on the average higher in WRN-depleted cells. As seen before (Table 1), a greater proportion of forks in WRN-depleted cells had very short post-HU segments (less than a fifth of the length of their corresponding pre-HU segments; Figure 6D, D=0.436, P<<0.001 for the track segment ratio comparison between WRN-depleted and control cells by KS test; also see Table 1). In the absence of HU treatment, the majority of ongoing forks in both WRN-depleted and control cells progressed at a uniform rate (Figure 6D), though it should be noted that in WRN-depleted cells a slightly higher proportion of forks may have traveled with an irregular rate (D=0.186, P=0.02 for pre HU track comparison between WRN-depleted and control cells by KS test).

In summary, these data indicate that WRN facilitates fork progression after HU-induced replication arrest, which is similar to the case observed for MMS-induced stress. However, WRN does not appear to be required for stabilizing forks during HU arrest, since a similar proportion of forks are able to resume DNA synthesis after HU removal in WRN-depleted and control cells.