DNA double strand breaks (DSBs) can be generated by both endogenous and exogenous sources of damage. DSBs can be induced directly (DSB node) by, for example, ionizing radiation or can be originated as a byproduct during the repair of another types of DNA damage, such as an interstrand crosslink (ICL) (Knipscheer et al., 2009; Ceccaldi et al., 2016b). The current evidence indicates that the non-homologous end-joining (NHEJ) pathway repairs most direct-DSBs, irrespective of the cell cycle phase in which they appear; whereas the homologous recombination repair (HRR) pathway repairs ICL-derived DSBs, the repair of these lesions requires end-resection (R-DSB node) mediated by DNA exonucleases (Ceccaldi et al., 2016a).

When a DSB appears in the DNA molecule it is recognized first by the KU70-KU80 heterodimer and the 53BP1 protein (KU-53BP1 node). KU70/80 ensures the DNA ends and 53BP1 represses the function of HRR proteins, such as FANCS/BRCA1, thus funneling the repair of the DSB to the NHEJ pathway, composed by DNA-PKcs kinase, the Artemis nuclease and the XLF-Lig4 complex (NHEJ node), that directly seal the DSB and restore the DNA continuity. Although the NHEJ is considered highly mutagenic, it is the main DSB repair pathway along the cell cycle given its promptness (Lieber, 2010; Ceccaldi et al., 2016a).

An R-DSB is generated during the repair of an ICL. Briefly, an ICL encountered during S-phase stalls the DNA replication fork and activates the FA/BRCA pathway through the FANCM-FAAP24 complex (Xue et al., 2008; Schwab et al., 2010; Blackford et al., 2012). The FANCM-FAAP24 complex recruits the FA core complex (FAcore node) (Kottemann and Smogorzewska, 2013) which, through its E3 ubiquitin ligase, activates the FANCD2-FANCI heterodimer (FANCD2-I) (FANCD2I node) (Meetei et al., 2004). The monoubiquitinated FANCD2-I complex remains in the chromatin and promotes the recruitment of FANCP/SLX4, a scaffold for the nucleases associated to the FA/BRCA pathway: MUS81, FANCQ/XPF, and SLX1 (Kottemann and Smogorzewska, 2013) (NUC1 node). These nucleases unhook the ICL and generate a DNA adduct in one of the previously joined strands, and a DSB in the opposing strand that will be resected by DNA exonucleases to generate a R-DSB (Ceccaldi et al., 2016b). The adduct is bypassed by translesion synthesis polymerases and repaired by nucleotide excision repair, whereas the R-DSB is funneled to the HRR pathway (HRR node) (Rodriguez et al., 2012).

Both DSBs and R-DSBs activate the phosphorylation of the histone variant H2AX at S139, producing γH2AX (γH2AX node), the main histone modification associated to DSBs signaling that also functions as a scaffold for the recruitment of additional DNA repair factors (Celeste et al., 2003; Kolas et al., 2007). As mentioned, the R-DSB was generated by DNA exonucleases to generate a single-stranded 3'DNA overhang (ssDNA) that is rapidly covered by RPA subunits (Cejka, 2015; Ceccaldi et al., 2016b). RPA is subsequently substituted by RAD51 that in association with the recombination proteins FANCD1/BRCA2 and other RAD51 family members would compose the pre-synaptic filament that searches homology for DNA repair (Taylor et al., 2015; Ceccaldi et al., 2016a,b; Godin et al., 2016). This strand invasion will mediate HRR and is considered the process with the highest fidelity for DSB repair (Jasin and Rothstein, 2013). This entire process is included in the HRR node. As previously mentioned, the NHEJ is the main DSB repair pathway (Wu et al., 2008), however the proteins of the FA/BRCA pathway block the activity of the NHEJ proteins during ICL processing thus ensuring that the HRR will be the only mechanism repairing R-DSBs (Bunting and Nussenzweig, 2010; Pace et al., 2010).

Importantly, error-prone alternative pathways for the processing of ICLs and the repair of ICL-derived DSBs might be activated when the FA/BRCA pathway is non-functional. We included these possibilities in the FA-CHKREC BNM. First, as the generation of ICL-derived DSBs in FA cells is not completely abolished, we hypothesized that an alternative ICL-unhooking pathway might emerge to improve the tolerance of these cells to ICL-arrested replication forks. In ATR mutant cells, for example, the constant replication stress due to defective DNA damage response (DDR) activates PLK1 and RNF4 (RNF4 node) that cooperatively take the control of the stalled replication fork and coordinate the recruitment of alternative DNA nucleases, thus collapsing the replication fork for continuation of DNA synthesis (Ragland et al., 2013). Although this mechanism has not been proven in FA cells, we hypothesize that the replication stress due to unrepaired ICLs could activate the same or a similar cascade of events for collapsing the replication forks in absence of the FA pathway. The NHEJ, on the other hand, is well-known to be over-activated in FA cells as a backup DNA repair pathway that might collaborate in the repair of the DSBs accumulated in FA cells (Adamo et al., 2010; Bunting and Nussenzweig, 2010; Pace et al., 2010). Finally, when inhibited, the alternative-NHEJ initiated by PARP-1, displays synthetic lethality in FA mutants and PARP-1 has been shown to have an important role protecting the HRR process from NHEJ interference (Hochegger et al., 2006).

Whilst DNA damage is present, it is critical to avoid cell cycle progression from G2 to M, the last checkpoint before cell division (Giglia-Mari et al., 2011). As the G2/M progression is promoted by the mitosis promoting factor, also known as Cyclin B/CDK1 complex, (Cyclin B-CDK1 node), a tight regulation over this complex is necessary, above all when the cell cycle has been perturbed by DNA damage (Charrier-Savournin et al., 2004).

Cyclin B-CDK1 blockage is accomplished through several mechanisms, in the presence of a DSB, cell cycle arrest is initiated through the ATM kinase (ATM node), while in presence of an ICL, the ATR kinase through HCLK2 and RPA accumulation (ATR node) takes charge on signaling (Yazinski and Zou, 2016). These kinases spread the DNA damage signal through their downstream effectors CHK2, CHK1 and p53 (p53 node) (Collis, 2008; Schwab et al., 2010). In addition, ATM and ATR phosphorylate additional components of the FA/BRCA pathway (Traven and Heierhorst, 2005; Matsuoka et al., 2007). Once triggered, p53 activates the expression of p21, the main cell cycle blocker that inhibits the kinase activity of the Cyclin B/CDK1 complex and avoids cell cycle progression while the DNA repair transactions are performed (Charrier-Savournin et al., 2004).

Another branch for cell cycle blockage is composed by the WEE1 (WEE1 node) and MYT1 (MYT1 node) kinases that directly inhibit the Cyclin B/CDK1 complex formation by phosphorylating the Thr14 and Tyr15 residues of CDK1 (Wang et al., 2004). Removal of these inhibitory modifications is controlled by members of the CDC25 phosphatases and leads to a rapid activation of the Cyclin B/CDK1 complex. In turn, the initial activation of the Cyclin B/CDK1 complex stimulates the activity of CDC25 and inactivates WEE1, creating a feedback that increases the Cyclin B/CDK1 complex activity. Activation of the DNA damage checkpoint in G2 leads to a reduction in the activity of the CDC25 family and increases WEE1 function thus keeping the Cyclin B/CDK1 complex limited and preventing mitotic entry (Takizawa and Morgan, 2000; Mailand et al., 2002; Ray and Kiyokawa, 2008).

When the DNA repair process has been completed, the cell needs to switch off the checkpoint, which in turn would allow the division of the cell (Bartek and Lukas, 2007; Medema and Macurek, 2011). To accomplish this, several phosphatases and kinases of the checkpoint recovery (CHKREC) take charge and inactivate the checkpoint components as well as activate the promoters of cell division, such as the Cyclin B/CDK1 complex (Shaltiel et al., 2014). The control of this switch is of central importance given that failing its execution might conduct to the delivery of unrepaired DNA. One of the critical steps in this process is the inactivation of p53 by the WIP1 phosphatase (WIP1 node) and the MDM2 ubiquitin ligase (Bose and Ghosh, 2007), which themselves are direct transcriptional targets of p53. WIP1 dephosphorylates p53 at S15 and prevents its activity (Belova et al., 2005).

WIP1 dephosphorylates γH2AX, ATM, CHK2, and CHK1 thus inactivating the checkpoint (Cha et al., 2010). Other phosphatases have been shown to collaborate with WIP1 to dismantle the checkpoint including PP2A-B55, PP4, and PP6 (PP2A-B55 node) (Schmitz, 2010). Importantly, WIP1 deficiency results in a failure in checkpoint recovery and the cell remains arrested in G2 (Belova et al., 2005). Aside from the reversal of phosphorylations by phosphatases, the CHKREC is also regulated by several kinases (Peng, 2013), including PLK1 (PLK1 node) which is of the highest importance (Seki et al., 2008). PLK1 phosphorylates several DNA repair and checkpoint components in such a way that the targeted proteins are recognized by βTrCP- ubiquitin proteosome system and degraded (Mamely et al., 2006). Well-recognized PLK1 targets include WEE1, Claspin, 53BP1, CHK2, and p53 (van Vugt et al., 2010). During CHKREC, PLK1 becomes essential for re-entry into the cell cycle as well as its upstream activators Aurora A and Bora (CDK1-AurA node) (Seki et al., 2008). Finally, the Cyclin B/CDK1 complex is reactivated by the CDC25 phosphatase family (CDC25 node) (Boutros et al., 2006). Given the pro-mitotic activity of Cyclin B, PLK1 and Bora, just to name a few, they are actively targeted for degradation or remain down-regulated whereas the cell cycle arrest is active (He et al., 2005; Tsvetkov and Stern, 2005).

In conclusion, cell cycle arrest and DNA repair are two mechanisms that act in a concerted way in presence of DNA damage, the proteins that mediate both processes are inactivated by the CHKREC once the DNA damage has been eliminated. The CHKREC consists of two concerted mechanisms: (a) the phosphatase negative pathway that inactivates DDR and its effectors and (b) the kinase cascade that mediates cell cycle progression (Shaltiel et al., 2014).

We modeled the FA-CHKREC network as a Boolean network using the nodes and logical rules that appear in Table 1. The implementation of the network as a dynamical system followed standard procedures (Rodriguez et al., 2012, 2015; Ramírez and Mendoza, 2018), and the simulations were performed using BoolNet (Müssel et al., 2010) under a synchronous updating scheme. We simulated the wild type network as well as all possible gain-of-function mutants and null mutants of the model. The null and constitutive activation mutants were simulated by fixing the mutant node at 0 or 1, respectively.

Two biologically relevant conditions were simulated: namely, a short exposure to DNA damage (represented by the node ICL), which is assumed to be repaired fast and efficiently, and a persistent exposure to DNA damage, which results in a saturation of the DNA repair pathways. The response to short DNA damage exposure was simulated synchronously both in the wild type and the mutants by setting the ICL activation state to 1 only at the initial state, whereas a continuous exposure to DNA damage was simulated by fixing the DNA damage node activation state to 1. The effect of removing interactions was also evaluated when considered pertinent in combination with null/persistent activation mutants and in response to short/persistent exposures to DNA damage. The trajectories from all possible initial states were analyzed until the system reached an attractor. The model is available as the Supplementary files FA-CHKREC_for_BoolNet.txt and FA-CHKREC.smbl.

The EUFA316+EV, EUFA316+G, VU817 (kindly donated by Dr. Hans Joenje, VU University Medical Center) and NL53 (DNA repair proficient) lymphoblast cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum (GIBCO, Waltham, Massachusetts, USA). Cancer derived cell lines TOV21G+EV and TOV21G+F, as well as MCF7 cell lines were maintained in RPMI medium 1640 supplemented with 10% fetal calf serum (GIBCO, Waltham, Massachusetts, USA); HeLa cells were maintained in DMEM medium supplemented with supplemented with 10% fetal calf serum (GIBCO, Waltham, Massachusetts, USA) For survival curves, cells were seeded in 96-well plates at 2,000, or 5,000 cells per well, treated with serial dilutions of GSK2830371, BI2536, TC-S 7010, CDC25 phosphatase inhibitor II, Carboplatin, Olaparib, and/or MMC. Cell survival was analyzed after 5 days of culture using Cell Titer Glo (Promega) following manufacturer instructions.

Cells were treated with MMC and CCT007093 (WIP1 inhibitor) (both from Sigma-Aldrich Co, St. Louis MO, USA) during 96 h and harvested every 24 h. For chromosome aberrations analysis, colchicine (Sigma-Aldrich Co, St. Louis MO, USA) (final concentration 0.1 μg/ml) was added to cell cultures 1 h before harvesting. Twenty-five metaphases per experimental condition were scored and the number of chromatid breaks, chromosome breaks and radial figures was assessed. For cell cycle and DNA damage analysis, cells were fixed with the Cytofix/Cytoperm buffer (BD Biosciences), washed with the Perm-Wash buffer (BD Biosciences) and stained with anti- H3S10ph antibody (Biolegend) and anti-γH2AX antibody (BD Biosciences). For DNA content analysis cells were counterstained with Propidium Iodide (Sigma-Aldrich). A total of 20,000 events were acquired in a BD FACSCalibur or a FACScan cytometer (BD Biosciences) and analyzed in the FlowJo software version 10.1. For gene expression analysis RNA was extracted with the QIAGEN mini kit, PCR arrays were performed following the manufacturers instructions, using the RT 2 Profiler™PCR Array “Human DNA Repair” kit (QIAGEN). PCR reaction was performed in an ABI Prism 7000 Real-Time PCR System (Applied Biosystem) and data were analyzed using the QIAGEN on-line tool (https://www.qiagen.com/us/shop/genes-and-pathways/data-analysis-center-overview-page/) Experiments were run by triplicate.

The attractors obtained with the FA-CHKREC model in every DNA damage simulation were considered to be a cell fate decision. These attractors were classified, according to the activation of critical nodes, in two categories: (1) cell-cycle arrest (CCA) attractor and (2) cell-cycle progression (CCP) attractor. The attractors obtained in every mutant simulation were not necessarily identical. However, a “signature” for every attractor category can be defined based on the activity of critical nodes. The CCP attractor is defined by activation of the cell division promoter nodes (CDK1-AurA, PLK1, CDC25, and Cyclin B-CDK1) and inactivation of any other node in the network. The CCA attractor is defined by the activity of the DNA damage nodes (ICL, DSB, and R-DSB), activity of the cell cycle arrest nodes (ATM, ATR, p53, p21, WEE1, and MYT1) and inactivation of the critical node CyclinB-CDK1. It is important to mention that deviations from these two broad attractor categories exist. For example, CCP-DDA (Cell Cycle Progression with DNA Damage Adaptation Attractor), the one that is critical for this work. This attractor shows activation of cell division nodes along with activation of DNA damage nodes (see also Table 2).

As previously mentioned, we included in the FA-CHKREC BNM 15 unreported interactions. These are indicated in Figure 1 as blue arrows for inferred positive regulations, and as dashed purple arrows for inferred negative regulations (see also Table S2). Importantly, if these 15 interactions are removed the model is not able to recover all the attractors relevant for the cellular dynamics studied in this work. Without these 15 interactions all the initial conditions of the system are driven exclusively toward a CCP attractor (see Table S1B). The relevance of these 15 interactions in cellular context will be described below.

According to our simulations (Figure 3A), turning off the CHKREC nodes dampens cell division capacity and drives the simulation to CCA attractor, therefore, chemical inhibition of CHKREC components might prevent FA mutants, such as FAcore, from reaching cell division (Figure 3B). We tested this model-derived hypothesis in the EUFA316+EV cell line, a lymphoblast cell line derived from a FA patient with inherited mutations in FANCG, using chemical inhibitors against the components of the CHKREC nodes, namely GSK830371 (Sigma-Aldrich) for WIP1 inhibition, TC-S 7010 (Selleckchem) for inhibiting Aurora A, BI2536 (Selleckchem) for inhibiting PLK1 and CDC25 Phosphatase Inhibitor II (Santa Cruz Biotechnologies) for inhibiting CDC25. Figure 3C shows survival curves from the different CHKREC inhibitors tested. Of note, FA deficient cells were more sensitive than their corrected counterparts (EUFA3156+G) to chemical inhibition of WIP1, PLK1 and Aurora A, and to a lesser extent to CDC25 inhibition. We decided to keep studying WIP1 phosphatase since it remains poorly explored and WIP1 inhibitors are gaining relevance in the oncology field, WIP1 inhibition, for example, has been associated to reduced tumorigenesis and induction of cell cycle arrest in cancer cells (Belova et al., 2005; Lu et al., 2005; Shaltiel et al., 2014).

In addition to the EUFA316+EV cell line, other different FAcore mutants were found to be sensitive to WIP1 inhibition with GSK2830371, including the TOV21G+EV, an ovarian cancer cell line with epigenetic silencing of the FANCF gene (Figure 4A), and the HeLa and MCF7 cancer cell lines that have previously been shown to be resistant and sensitive to GSK2830371, respectively (Figure 4B). In addition, co-treatment with GSK2830371 sensitizes the TOV21G+EV cell line to treatment with carboplatin, an inductor of interstrand-crosslinks (Figure 4C), and Olaparib, a PARP inhibitor (Figure 4D); GSK2830371 also sensitizes the EUFA316 cell line to MMC (Figure 4E).

Seven different concentrations of MMC (4–1,000 ng/ml) were evaluated during 4 days of continuous treatment to monitor the maximum amount of DNA damage that a FAcore mutant cell line can tolerate, in this case the VU817 cell line, mutant in the FANCA gene (Figure S2). 10 ng/ml of MMC allows the division of 100% multi-aberrant cells with up to 10 chromosomal aberrations per cell, in addition, almost the entire culture is positive for γH2AX staining, a marker of unrepaired DSBs. Despite the dramatic amount of unrepaired DNA damage, an important number of cells reaches mitosis and cell death remains below 10%, suggesting that this is an amount of DNA damage in which DDA occurs (Figure S2). The effect of WIP1 inhibition was studied in this setting and we evaluated cell division parameters. We observed that, effectively WIP1 might contribute to DDA in FA-A cells, since its inhibition with the chemical inhibitor CCT007093 (Sigma-Aldrich) reduces the mitotic index, measured by cells positive for phosphorylation of the mitotic marker H3S10, and activates apoptosis, measured by PARP-cleavage positive cells (Figures 5A,B). Figure 5C shows how treatment with CCT007096 reduces the mitotic index (H3S10ph+ circled population) of FA cells with respect to untreated cultures and MMC-treated cultures. WIP1 inhibition with CCT007093 does not modify the number of chromosome aberrations (Figure 5D), suggesting that WIP1 phosphatase is not directly involved in DNA repair mechanisms, but controls cell cycle transitions.

WIP1 collaborates turning-off the checkpoint through dephosphorylation of several of its components including p53, CHK1, CHK2, γH2AX, and the ATR and ATM kinase target motif p(S/T)Q (Lindqvist et al., 2009a; Shaltiel et al., 2014), therefore in its absence the checkpoint machinery remains active and the continuation of the cell cycle is not allowed. In such a situation of long lasting unrepaired DNA damage two scenarios are feasible, a DNA damage induced senescence, which has previously been described in other contexts of unrepairable DNA damage (von Zglinicki et al., 2005; Kim et al., 2014) or activation of apoptosis. We confirmed the gene expression landscape for these scenarios using a real time PCR array that simultaneously detects 84 DNA damage related transcripts (QIAGEN). First we found that both PPM1D and CDKN1A, which codify for WIP1 and p21 proteins, respectively, become activated in FA-A cells after exposure to MMC, thus demonstrating that the CHKREC program becomes activated in presence of DNA damage (Figure 5E). Furthermore, chemical inhibition of WIP1 in MMC-treated FA-A cells activated the expression of definitive cell cycle arrest genes and apoptotic genes, namely CDKN2A (which codifies for p16) and APAF1, respectively (Figure 5F), supporting our notion that WIP1 is critical for division and survival of FA cells with unrepaired DNA damage.

In summary, our results show that FA cells with a large amount of DNA damage preserve the capacity of cell division after DNA damage induction, a result that is consistent with the FA-CHKREC BNM simulations. The finding that inhibition of WIP1 in particular, and of the CHKREC in general, prevents cell cycle progression in FA-A cells deserves further exploration so as to elucidate a potential larger impact of CHKREC studies in the FA field. CHKREC stimulation would allow survival of the hematopoietic stem cell compartment and amelioration of the total cell count numbers, however overexpression of the same components may conduct to stem cell pool attrition and potential selection of malignant cell clones. Given this, a profound knowledge of this process is necessary to understand the balance between the normal hematopoiesis and the hematopoiesis resulting from CHKREC hyper-activation, the later in a context of DNA repair deficiency might result in hematopoietic stem cells with high amounts of DNA damage and potentially contribute to cancer.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.