Background
DNA replication does not proceed in a continuous manner, but stalls and restarts at sites of DNA damage [
1‐
3]. DNA damage occurs continuously from both endogenous and exogenous sources [
1‐
3]. Replication stress occurs when the rate of proliferation overtakes the clearance of the DNA damage ahead of progressing replication forks [
1‐
3]. Cancer cells experience high levels of replication stress. Thus, efficient restart of stalled or collapsed replication forks is critical to their survival, particularly in response to common cancer therapeutics [
4,
5]. Radiation repair protein 51 (RAD51)-dependent homologous recombination (HR) is the canonical repair and restart pathway for stalled replications forks [
6‐
9]. HR is best characterized for the repair of DNA double-strand breaks (DSBs). HR is mediated by a litany of components that are regulated by breast cancer susceptibility protein-1 (BRCA1), which promotes the initial step in HR, 5′ end resection to create 3′ single-stranded (SS) DNA. BRCA2 then loads RAD51 onto this SS DNA to catalyze strand invasion into homologous sequences (typically the sister chromatid) creating heteroduplex DNA intermediates [
2,
7,
8,
10]. After the invading strand re-initiates DNA replication, HR intermediates such as Holliday junctions are resolved by Holliday junction 5′ flap endonuclease (GEN1) or MUS81 structure-specific endonuclease subunit (MUS81), with SLX4 structure-specific endonuclease subunit (SLX4) serving as a scaffold [
11‐
13].
HR repair of stressed replication forks also requires 5′ end resection as an initial step. This 5′ end resection needs a free DNA double strand (DS) end structure for the 5′ exonuclease activity in end resection. There are two ways to create this DS end: fork reversal to a chicken foot structure, or fork cleavage by a structure-specific nuclease [
14‐
16]. We previously reported that the 5′ endonuclease EEPD1 could cleave stalled replication forks, initiate EXO1-mediated 5′ end resection, and promote repair of HR replication forks independent of BRCA1 [
17‐
19]. However, BRCA1/2-mutant cancer cells lack HR, and how these cells repair stalled replication forks has been an unresolved issue. Several reports point to RAD52 in fulfilling this function. In yeast, Rad52 plays an essential role in HR, including HR-mediated restart of collapsed replication forks [
20‐
22]. However, early studies suggested that in mammals the essential roles for RAD52 in HR have been supplanted [
23], perhaps by BRCA2. We reported that human RAD52 foci appear 4–6 h after exposure to ionizing radiation, long after RAD51 foci appear, and we proposed that these late-appearing foci reflected a conserved role for human RAD52 in HR-mediated repair of collapsed replication forks [
24]. This model was supported by a subsequent study showing that DSBs arising many hours after exposure to ionizing radiation were replication-dependent and repaired by HR [
25]. In a separate line of investigation, RAD52 was identified as essential for viability of cancer cells with defects in various HR proteins including BRCA1, BRCA2, and partner and localizer of BRCA2 (PALB2) [
23,
26,
27]. Together, these results suggest that RAD52 functions in a backup HR pathway independent of BRCA1/2, in which RAD52 loads RAD51 onto SS DNA for HR repair at stalled forks [
26‐
29]. RAD51 then promotes the strand invasion required to complete HR repair and replication fork restart [
26‐
28].
However, RAD52 still needs an end-resected 3′ SS upon which to load RAD51. Thus, 5′ end resection is still required for the backup pathway, yet this is problematic if BRCA1 is not functional, because BRCA1 promotes resection [
30‐
32]. Since EEPD1 can operate independently of BRCA1 to initiate EXO1-mediated 5′ end resection after replication fork stalling [
18,
19], this suggests that in the absence of functional BRCA1, EEPD1 can initiate the 5′ end resection needed for generation of the 3′ SS DNA required for RAD51 loading.
Repairing stressed replication forks is a high priority for the cell. If stressed replication forks are not repaired in timely manner, they may convert into toxic structures that make fork restart difficult [
3,
9,
12,
13,
33], leading to mitotic catastrophe as demonstrated by nuclear abnormalities, including nuclear bridges and micronuclei. These nuclear abnormalities can arise from non-homologous end joining (NHEJ)-mediated fusion of free DS ends at unrepaired replication forks [
3,
34‐
36]. Unbalanced chromosomal fusions can result in chromosomes without centromeres, which are retained as micronuclei after mitosis, and chromosomes with two centromeres, which form chromosomal bridges between daughter cells during mitosis [
36,
37]. Since BRCA1/2 mutant cancer cells use RAD52 as an escape pathway for HR-mediated replication fork repair and restart, depleting RAD52 causes mitotic catastrophe and synthetic lethality in these cells. Thus, RAD52 has emerged as a target of interest for pharmaceutical intervention for novel synthetic lethal treatment strategies for BRCA1/2 mutant cancers [
38‐
40]. However, the molecular mechanism by which RAD52 deficiency causes synthetic lethality of BRCA1/2 mutant cancer cells has not been identified [
26,
27,
29].
In this study we show that EEPD1 is required for death of BRCA1 mutant breast cancer cells that have been depleted of RAD52. Specifically we show that depletion of EEPD1 rescues the synthetic lethality of RAD52-depleted, BRCA1 mutant breast cancer cells. Co-depletion of EEPD1 with RAD52 promotes restart of stalled replication forks, and suppresses chromosome aberrations and mitotic catastrophe compared to RAD52 depletion alone. These results suggest that EEPD1 may play a role in generating a toxic replication fork intermediate that leads to mitotic catastrophe.
Methods
Cell culture, transfection, and survival assays
EEPD1 and/or RAD52 were selectively depleted using small interfering RNA (siRNA) transient transfection (Lipofectamine RNAiMAX transfection Reagent, Life Technologies). SMARTpool ON-TARGETplus Non-target pool (SiRNA control) (D-001810-10-20), EEPD1 SiRNA (L-014641-01-0020), RAD52 SiRNA (L-011760-00-0010), X-ray repair cross-complementing protein (XRCC4) siRNA (L-004494-00-0005), DNA ligase IV (LIG4) (L-004254-00-0005), polymerase theta (POLQ) siRNA (L-015180-01-0010) and BRCA1 siRNA (L-003461-00-0005) were purchased from Dhamarcon RNAi Technologies (Pittsburgh, PA, USA). Briefly, the day prior to transfection, cancer cells (MDA-MB-436, SUM149PT and MCF7) were plated at a density of 1.4 × 10
5 per well. Transfection reagents were prepared by mixing 6 μl of RNAiMAX/250 μl Opti-MEM (Life Technologies) to 50 nM of siRNA/250 μl Opti-MEM at room temperature (RT) for 20 min before adding to cells. Between 4 h and 6 h after transfection, 0.5 ml of fresh medium was added to each well for all cell types except MCF7. Instead of 50 nM siRNA, 10 nM of siRNA was used in all MCF7 transfections [
27]. Cells were harvested at 2 days post-transfection for clonal colony formation (survival), western analysis, immunofluorescence or other assays. All experiments were performed at least three times in triplicate (n
> 9).
Clonal survival was determined by seeding transfected cells (800 MDA-MB-436 (+ or -/-), 1000 MCF7 or 1000 SUM149PT) per well of a 6-well plate and cells were allowed to expand for 10–12 days (MDA-MB-436 BRCA1+, SUM149PT) or 14–18 days (MDA-MB-436 BRCA1-/-, MCF7). Cells were then rinsed with 1 × PBS, fixed with 1% formaldehyde for 10 min and stained with 0.1% crystal violet before counting. Colonies with > 50 cells were counted as a surviving clone. For hydroxyurea (HU) treatment (H8627) (Sigma, St. Louis, MO, USA), cells were treated with 10 mM HU overnight at the 48-h post-transfection time point and subsequently harvested for colony formation assay as described. The unpaired Student t test was used for all statistical analysis, unless otherwise indicated.
MDA-MB-436 breast cancer cells (BRCA1 mutant (-/-) or replete (+)) (ATCC, Manassas, VA, USA) and MCF7 (ATCC) were cultured in D-MEM (Life Technologies, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin and streptomycin (Life Technologies).
SUM149PT BRCA1-/- breast cancer cells (Asterand Bioscience, Detroit, MI, USA) were cultured in Ham’s F-12 medium (Invitrogen) supplemented with 5% heat inactivated FBS (Hyclone), 10 mM HEPES (Invitrogen), 1 μg/ml hydrocortisone (Sigma) and 5 μg/ml insulin (Sigma).
Western blot analysis
Protein expression of EEPD1, RAD52, DNA ligase 4 (LIG4), XRCC4, POLQ, BRCA1, BRCA2, and the constitutively expressed cyclophilin B was monitored by standard western blotting. EEPD1 expression was detected by a custom-produced mouse polyclonal antibody to EEPD1 protein (Interdisciplinary Center for Biotechnology Research Core Facility, UF, Gainesville, FL, USA) [
19,
41]. RAD52, LIG4, BRCA1, and BRCA2 antibodies were purchased from Santa Cruz Biotech (sc-8350, sc-365341, sc-271299, sc-6954 and sc-1818). POLQ and XRCC4 antibodies were purchased from ThermoFisher Scientific (PA5-39885 and PA5-27104). Cyclophilin B antibodies were purchased from Abcam (ab178397) (Cambridge, MA, USA). Secondary antibodies used for enhanced chemiluminescence (EC) detection were ECL Rabbit IgG, HRP-linked Whole Ab (NA934-1ML), HRP-conjugated mouse secondary antibody (NA931-1ML) (Thermo Fisher Scientific, Waltham, MA, USA) and HRP-conjugated goat IgG (sc-2020, Santa Cruz Biotec). SuperSignal West Pico Chemiluminescent Substrate (ECL) (34078) and High Performance Chemiluminescence film; Amersham Hyperfilm ECL (45001508) were purchased from Thermo Fisher Scientific.
Expression levels of proteins involved in the ATM/ATR DNA damage signaling pathway were examined using ataxia-telangiectasia mutated kinase (ATM) (2873), p-ATM (5883), ATM-related and Rad3-related kinase (ATR) (2790), Checkpoint kinase 1 (Chk1) (2341), p-Chk1 (2348), Chk2 (2662) and p-Chk2 (2662) antibodies from Cell Signaling Technology (Danvers, MA, USA), p-ATR (GTX128145) antibodies from GeneTex (Irvine, CA, USA), replication protein A 32 (RPA32) (A300-244A) and p-RPA32 (A300-245A) antibodies from Bethyl Laboratories (Montgomery, TX, USA).
Immunofluorescence
Immunofluorescence foci assays were performed as we previously described with minor modifications [
19]. In brief, MDA-MB-436 BRCA1
-/- cells were cultured on coverslips followed by siRNA transfection. At the predetermined time points (1, 2, 3, or 4 days post transfection), cells were fixed with 1% formaldehyde for 10 min at ambient temperature, rinsed with 1 × PBS, incubated with methanol for
> 5 min at − 20 °C, rinsed with 1 × PBS and permeabalized with 0.1% Triton-X for 3 min before incubation with phosphorylated histone 2A family member X (γH2AX) antibodies (05-636) (1:200) (Millipore, Temecula, CA, USA) at 4 °C overnight. The cells were then rinsed with 1 × PBS multiple times. Secondary antibodies (Goat anti-Mouse IgG, Alexa Fluor® 568 conjugate, A11004) (1:400) (Thermo Fisher) were added to the cells at ambient temperature and protected from light for 1 h. After washing thrice with 1 × PBS, coverslips were mounted in an anti-fade solution containing 4',6-diamidino-2-phenylindole (DAPI).
All samples were analyzed using either a Zeiss fluorescence microscope (Axiovert 200 M) (Carl Zeiss Microscopy, LLC, Thornwood, NY, USA) or a Leica TCS SP5 confocal scanning microscope (Leica Microsystems, Exton, PA, USA). Immunofluorescence images were taken using a Hamamatsu ORCA-ER digital camera (Hamamatsu Photonics K.K, Bridgewater, NJ, USA) and processed by Zeiss Axiovision Release 4.6 software. Images from confocal microscopy were processed by Leica LAS AF imaging software. Cells with ≥ 5 foci were scored as positive. Photomicrographs of distinct cell populations were taken at equal magnifications and equal fluorescence intensities. To assess nuclear structural abnormalities (micronuclei and post-mitotic bridging), MDA-MB-436 BRCA1-/- cells, with or without EEPD1 and/or RAD52 depletion, were fixed as described above, and stained with 300 nM DAPI (Beckman) in PBS for 5 min. After washing with PBS, coverslips were mounted in anti-fade solution and analyzed using confocal microscopy. Each immunofluorescence assay was performed at least three times.
Analysis of chromosome breaks
Cytogenetic analysis of chromosome breaks was performed as we described with minor modifications [
19]. Briefly, 24 h prior to transfection, 1.6 × 10
5 cells were plated into wells containing 1.5- mm coverslips. Cells were transfected with control, EEPD1, and/or RAD52 siRNA using Lipofectamine RNAiMAX reagent as described. After 48 h, cells were treated with Colcemid (final concentration of 0.1 ug/ml) (Sigma) for 2 h at 37 °C, 5% CO2. After 2 h, cells were treated with 75 mM KCl at 37 °C, 5% CO2. After 15 min, 0.1 volume of fixative (methanol/acetic acid 3:1) was added to the KCL for a few seconds and the supernatant was aspirated and replaced by 1 ml fresh fixative for 5 min at room temperature. The step was repeated two more times. Coverslips were then left for air drying and cells were rehydrated with PBS for 5 minutes and stained for the Giemsa Stain (Karyomax Giemsa Stain, 10092) (Gibco, Carlsbad, CA, USA) for 5 min at RT. The coverslips were rinsed three times with deionized water, mounted and examined with a × 63 objective connected to a Zeiss Axiovert 200 M microscope. At least 50 metaphases were analyzed for chromatid breaks in each cell preparation. Data were collected from three separate experiments.
DNA fiber analysis of replication fork repair and restart
DNA fiber analysis for measuring stalled replication fork repair and restart was performed as we previously described [
30,
31]. Briefly, 600,000 of MDA MB 436 BRCA1
-/- cells were incubated overnight at 37 °C in 6-well plates, then 20 mM iodo-deoxyuridine (IdU) was added to the growth medium and incubated for 20 min at 37 °C. The IdU medium was removed and cells washed in fresh medium. Cells were then treated with 5 mM HU for 120 min or mock-treated. The HU-containing medium was replaced with fresh medium containing 100 mM chloro-deoxyuridine (CldU). Cells were then incubated for varying times at 37 °C. The CidU medium was removed, cells harvested, resuspended in PBS, and ∼ 1000 cells were transferred to a positively charged microscope slide (Superfrost/Plus, Daigger), and processed for DNA fiber analysis as we described previously [
32]. Slides were mounted in PermaFluor aqueous, self-sealing mounting medium (Thermo Scientific), and DNA fibers were visualized using an Olympus FV1000D confocal scanning microscope (Olympus America Inc., Center Valley, PA, USA). Images were analyzed using ImageJ software.
DNA resection at stalled replication forks
Single-label DNA fiber end-resection analysis was carried out as previously described with some modifications [
16,
42]. Briefly, MDA-MB-436 BRCA1 mutant cells transfected with the indicated siRNA were grown in 6-well dishes (2 × 10
5 cells/well), and then 20 μM IdU was added to the growth medium and incubated for 45 min at 37 °C. After washing with fresh medium, cells were treated with 5 mM hydroxyurea for 0 or 10 h at 37 °C. Cells were harvested and suspended in PBS, and 1000 cells were transferred to a positively charged microscope slide and processed for DNA fiber analysis as we described previously [
43]. Slides were mounted in PermaFluor aqueous, self-sealing mounting medium (Thermo Scientific), and DNA fibers were visualized using a confocal microscope (Olympus, FV1000D, × 63 oil immersion objective). Images were analyzed using ImageJ software.
Discussion
BRCA1 mutant or BRCA2 mutant malignancies rely on RAD52 to repair stressed replication forks [
26,
27,
29]. Depletion of RAD52 results in synthetic lethality of these homologous recombination (HR)-deficient cancers [
26‐
29]. This study demonstrates that the synthetic lethality seen when RAD52 is depleted in BRCA1 mutant breast cancer cells depends on the HR endonuclease EEPD1. Previously, we have shown that EEPD1 nicks stressed replication forks to initiate 5′ end resection, which creates 3′ SS DNA for RAD51 loading to affect HR repair of stalled forks [
18,
19]. The fact that synthetic lethality of RAD52-depleted BRCA1-deficient cells can be suppressed by downregulating EEPD1 implies that EEPD1 cleavage of stalled forks may create a toxic fork repair intermediate that is lethal if repair is not completed. In cells that lack BRCA1/2 and RAD52, EEPD1 would cleave stressed replication forks to permit 5′ end resection, but those cells would not progress past this repair intermediate. Repair would be arrested before RAD51-dependent homology-mediated single-strand invasion could occur. These cleaved replication forks would have free DS ends that could produce chromosomal fusions mediated by tumor protein P53 binding protein (53BP1)-dependent NHEJ [
34,
35]. In BRCA1 mutant cancer cells, unrepaired replication forks can generate chromosomal fusions, leading to chromosomal instability and mitotic catastrophe [
34‐
37]. Depleting 53BP1 rescues replication stress-induced chromosomal instability in these cells [
19,
34,
35]. Thus, one mechanism of cell death in the RAD52-depleted BRCA1 mutant cells could be chromosomal fusion of cleaved, but unrepaired forks, resulting in mitotic catastrophe.
Interestingly, our study finds that depletion of XRCC4 or DNA ligase IV, key components in the cNHEJ repair pathway, has little impact on cell viability in the MDA-MB-436 BRCA1
-/- cells (Fig.
5, Additional file
5: Figure S5). In contrast, depletion of POLQ, a mediator of aNHEJ, induces severe synthetic lethality in the BRCA1
-/- cells (Fig.
5, Additional file
5: Figure S5). These observations are consistent with a previous report showing that BRCA1-deficient cancer cells are dependent on the aNHEJ pathway for replication [
46]. Our study shows that MDA-MB-436 BRCA1
-/- cells rely heavily for survival upon the RAD52 recombination repair pathway, or if that fails, on the aNHEJ pathway (Fig.
1, Additional file
5: Figure S5). This implies that the RAD52 alternative HR repair pathway is effectively arrested in the co-depleted BRCA1 mutant cells, whereas the aNHEJ repair pathway remains active. In addition, unlike their ability to restore chromosomal integrity and replication efficiency (Figs.
2,
3 and
4), co-depletion of EEPD1 and RAD52 in BRCA1
-/- cells only has a moderate effect on DNA end-resection (Additional file
5: Figure S5). This is not surprising, since EEPD1 depletion would arrest replication fork repair before 5′ end resection, and repair of the fork might default to aNHEJ. POLQ-mediated aNHEJ typically does not compete with HR repair of stressed replication forks, probably because it is non-conservative, and the cell would designate it as a backup for conservative HR repair to protect its genome [
55‐
57].
Repairing and restarting stressed replication forks is one of the highest priorities for any cell. Indeed, there is a great deal of evidence that cells will bypass DNA lesions to maintain replication rates [
41,
47]. We and others have found that even delaying fork repair by 15–30 min can have lethal consequences [
17,
19,
35,
47]. The data presented here suggest a model in which an EEPD1-cleaved replication fork repair intermediate is rapidly recognized by the cell as toxic, and in the absence of BRCA1 and RAD52, the free DS ends are shunted toward the aNHEJ repair pathway. Without the toxic EEPD1-cleaved fork repair intermediate, aNHEJ can repair the stressed replication fork and permit the RAD51-depleted BRCA1 mutant cell to survive [
55‐
57]. This implies a hierarchy of replication fork repair pathways, with classical HR as favored, but in the absence of BRCA1/2, the cell chooses the RAD52 backup HR pathway [
26‐
29]. When cells lack both classical HR and the RAD52-mediated backup pathway, they choose aNHEJ to repair stalled replication forks [
13,
14,
55‐
57]. This may also explain why depleting both EEPD1 and RAD52 improves the survival of BRCA1 mutant cells after HU replication stress (Fig.
1c, d); this would force cells away from the RAD52 pathway to the more efficient but less conservative aNHEJ pathway. We had previously shown that EEPD1 actively represses aNHEJ while promoting HR, consistent with this hypothesis [
41].
There are several other lines of evidence that endonucleases mediate stressed replication fork collapse and cell death if fork repair is not completed in a timely manner. For example, BRCA2 protects stressed replication forks from degradation by the nuclease Mre11 [
41], perhaps by promoting timely loading of RAD51 onto end resected 3′ SS DNA [
55,
56]. In BRCA2 mutant cells, the histone H3 lysine 4 methylase (MLL3/4) complex component Pax transcription activation domain interaction protein 1-like protein (PTIP) was found to recruit double-strand break repair nuclease (Mre11) to stalled replication forks, causing degradation of nascent DNA strands [
57]. The Werner’s syndrome helicase (WRN) also stabilizes stressed replication forks and prevents their destruction by Mre11 [
48]. WRN interacting protein1 (WRNIP1) stabilizes RAD51 on 3′ end-resected SS DNA, and prevents prolonged and excessive end resection by Mre11, thereby protecting stressed replication forks from degradation [
49]. Finally, bi-orientation defect 1-like (BOD1L), a large protein with N-terminal homology to the mitotic regulator BOD1, is recruited to stressed replication forks where it stabilizes RAD51 on the 3′ SS end-resected DNA, which blocks further Bloom syndrome recQ-like helicase (BLM) unwinding and DNA2-mediated end resection [
50]. In each of these examples, excessive nuclease degradation of stressed replication forks is prevented by proteins recruited to stressed forks to regulate end resection. These reports demonstrating that sophisticated mechanisms have evolved to protect excessive degradation of stressed forks provides further evidence that such repair intermediates are toxic to the cell.
There is a significant effort to create small molecular inhibitors of RAD52 in order to clinically treat BRCA1 and BRCA2 mutant breast and ovarian cancers [
26,
38‐
40,
51]. Inhibition of RAD52 might be epistatic with PARP1 inhibition, since both strategies rely on the failure of replication fork repair and restart for their lethal effects, albeit at distinct steps in that pathway. Thus, combining PARP1 and RAD52 inhibitors to treat BRCA1 or BRCA2 mutant cancers might provide little or no therapeutic gain, and might increase normal tissue toxicity. Synthetic lethality from either PARP1 or RAD52 inhibition involves toxic replication fork repair intermediates that generate mitotic catastrophe and cell death [
26,
29,
52,
53]. Given that depleting EEPD1 prevents synthetic lethality of RAD52 repression in BRCA1 mutant cancer cells, potential mechanisms by which cancer cells could become resistant to clinically useful RAD52 inhibitors are by repressing EEPD1 expression or acquiring EEPD1 loss-of-function mutations [
52]. Either of these mechanisms would shunt stressed replication fork repair to aNHEJ [
55‐
57]. Thus, EEPD1 could be used as a biomarker for treatments that target RAD52 in BRCA1/2 mutant cancers.