CUL4A/B depletion potentiates the cytotoxicity of crosslinking agents
We started out with short-term viability assays, based on the cell-mediated resazurin reduction, to establish that the neddylation inhibitor MLN4924 potentiates the cytotoxic effect of the crosslinking agents cisplatin and MMC in HeLa cells, as demonstrated before with several other cancer cell lines [21, 22]. MLN4924 at a concentration of 10 μM reduces the IC50 of cisplatin from ~ 10 to ~ 2.5 μM and the IC50 of MMC from ~ 4 to ~ 1.5 μM (Fig. 1a). MLN4924 also increases the cytotoxicity of cisplatin and MMC in SKOV3 ovarian carcinoma cells (Additional file 1: Figure S1a and S1b).
Next, we depleted different cullins by siRNA transfections to understand which of the possible cullin targets of neddylation modulates this vulnerability to DNA-crosslinking agents. Cell viability assays, carried out in the presence of 5 μM cisplatin, confirmed a potentiation of cisplatin toxicity upon down regulation of CUL3 as reported before for SKOV3 and ES2 ovarian carcinoma cells [29]. The new finding of this screen is that a sensitization to cisplatin cytotoxicity is also detected upon simultaneous down regulation of the two scaffold paralogs of CRL4, i.e., CUL4A and CUL4B (Fig. 1b). Dose dependence experiments showed that this co-depletion of CUL4A and CUL4B mimics to a considerable extent the sensitizing effect of MLN4924 when cells are treated with cisplatin or MMC for 48 h (Fig. 1c and d). Nearly the same increase of sensitivity to cisplatin was achieved upon depletion of the CRL4 adaptor protein Damaged DNA-binding 1 (DDB1) instead of the CUL4A/B scaffold. Instead, no sensitization was elicited upon individual depletion of only one of the cullins, CUL4A or CUL4B, indicating that the two interchangeable scaffolds have a redundant function. These results were confirmed using distinct combinations of siRNA sequences targeting CUL4A and CUL4B to exclude off-target effects (Additional file 1: Figure S1c and S1d). The efficiency of protein down regulation upon siRNA transfections is documented in Additional file 1: Figure S2.
Further assays measuring the release of lactate dehydrogenase as a marker of membrane disruption (Fig. 1e) confirmed that the CUL4A/B depletion enhances cisplatin-induced cell death. Finally, the increased cytotoxicity of cisplatin upon a combined CUL4A/B depletion, but not after down regulation of only one of the cullins individually, was confirmed in a long-term colony-forming assay (Fig. 1f).
CUL4A/B depletion reduces H2AX/RPA phosphorylation upon ICL induction
Considering that CUL4A and CUL4B have an impact on the cytotoxicity of crosslinking agents, we tested the role of CRL4 in modulating DNA damage signaling following ICL induction. Resolution of ICLs by the FA pathway generates transient DNA breaks and ssDNA intermediates, which activate the checkpoint kinases ATR and ATM [5, 8, 30,31,32]. Phosphorylation of downstream factors like histone H2AX and the ssDNA-binding protein RPA generates docking motifs for effectors that mediate S phase checkpoints essential for DNA repair [3, 33,34,35]. Phosphorylation of H2AX (generating γH2AX) as well as phosphorylation of RPA2, the middle subunit of RPA, on serines 4/8 (generating pS4/8) and serine 33 (generating pS33), was assessed by immunofluorescence using phospho-specific antibodies. CRL4-proficient cells respond to cisplatin treatment with a dose-dependent increase of γH2AX, pS4/8 and pS33, but this phosphorylation was markedly reduced for the direct ATR targets pS33 and γH2AX (Fig. 2a-c) in CUL4A/B-depleted cells. A significant reduction was also observed in CUL4A/B-depleted cells for the formation of pS4/8, but only at the highly cytotoxic cisplatin concentration of 20 μM (Fig. 2d).
The phosphorylation of RPA2 at serines 4, 8 and 33 was further investigated by immunoblotting (Fig. 2e). Some increase of RPA2 phosphorylation was observed in CUL4A/B-deficient cells without any genotoxic treatment. This response is expected from the loss of CRL4-dependent licensing regulation (see Discussion). In blots with the generic RPA2 antibody, there was an apparent reduction of the overall RPA2 signal in CRL4-deficient cells compared to the CRL4-proficient counterparts. However, we did not find in the literature any indication that CRL4 would positively regulate RPA stability, such that a CRL4 down regulation could result in diminished RPA2 levels. We favor the view that the higher background phosphorylation of RPA2 in CRL4-deficient cells results in a reduced signal intensity in the electrophoretic position corresponding to the unmodified protein. This interpretation is supported by the immunofluorescence analyses (with quantifications) of Fig. 3, where in the absence of any crosslinking agent there is no reduction of RPA2 in CRL4-deficient cells compared to controls. When using phospho-specific antibodies, the immunoblots of Fig. 2e confirmed the observed reduction of pS4/8 and pS33 in CUL4A/B-depleted cells compared to CRL4-proficient counterparts after 24-h cisplatin exposures This finding was validated using a second siRNA sequence for the down regulation of CUL4A and CUL4B (Additional file 1: Figure S3a and S3b). To ensure that the observed shift in RPA2 electrophoretic mobility results from phosphorylation and not from a hypothetical ubiquitination by CRL4, cell lysates were subjected to phosphatase incubation prior to immunoblotting. Such a dephosphorylation clearly diminished the pS33 signal, thus confirming that we detected truly phosphorylated RPA2 (Additional file 1: Figure S3c). These results indicate that CUL4A/B-deficient cells are impaired in DNA damage signaling following cisplatin treatment.
CRL4 deficiency impairs interstrand crosslink-dependent assembly of ssDNA-RPA complexes
In view of the observation that the cisplatin-dependent RPA phosphorylation is reduced in CRL4 deficient cells, we next tested whether CRL4 is required for the recruitment of RPA to sites of cisplatin damage. By immunofluorescence analysis, some increase of RPA2 foci in chromatin was observed in CUL4A/B-deficient cells even without any genotoxic treatment (Fig. 3a and b). This response is expected from the loss of CRL4-dependent regulation of CDT1 described in previous reports (see Discussion). The replication licensing factor CDT1 is nearly completely degraded within 24 h after genotoxic stress caused by cisplatin in CRL4-proficient cells (Additional file 1: Figure S4a and S4b). Instead, the CUL4A/B depletion results in a pronounced stabilization of CDT1, such that the cells maintain high CDT1 levels despite cisplatin exposure. This results in uncoupling of the minichromosome maintenance (MCM) helicase activity and an uncontrolled re-replication that triggers RPA recruitment and other ATR-dependent signaling reactions [36]. However, this RPA recruitment to chromatin was not or only slightly further increased by cisplatin treatment of CRL4-deficient cells. As a consequence, CUL4A/B-depleted cells exposed to cisplatin display significantly lower levels of RPA foci when compared to CRL4-proficient controls treated with the same cisplatin concentrations (Fig. 3a and b).
Next, we tested whether RPA recruitment to chromatin in response to cisplatin damage is related to the ssDNA formation. For that purpose, ssDNA induction was assessed using a well-established method based on the incorporation of 5-iodo-2′-deoxyuridine (IdU), which allows for the probing of cells with an antibody that binds to IdU only in the ssDNA conformation. Using this same approach, Huang et al. [5] did not detect ssDNA intermediates after 4- to 6-h treatments with MMC or psoralen (plus UV-A radiation). In agreement with this earlier report, we also observed that ssDNA as well as pS33 remain below the detection threshold within the first 6 h of cisplatin exposure (Additional file 1: Figure S4c-e). However, a longer incubation time of 24 h revealed the formation of clearly detectable ssDNA foci in control cells (Fig. 3a). Again, an increase of ssDNA was observed in CUL4A/B-deficient cells even without genotoxic treatment, as expected from the loss of CRL4-dependent licensing regulation and the notion that the display of ssDNA provides an initial signal for ATR-mediated S phase checkpoint activation upon uncontrolled re-replication [36]. Consistent with the differential recruitment of RPA, ssDNA foci were substantially reduced in CUL4A/B-depleted cells compared to CRL4-proficient controls exposed to the same cisplatin concentrations (Fig. 3a and c). It is important to ascertain in these experiments that IdU is equally incorporated into DNA under the different experimental conditions, as shown by immunofluorescence after DNA denaturation (Additional file 1: Figure S4f-g).
To demonstrate the general relevance of the above-described findings, we also assessed the appearance of ssDNA and consequent RPA phosphorylation in MMC-exposed HeLa and SKOV3 cells (Fig. 3d and e). Immunofluorescence quantifications confirmed that the CUL4A/B depletion counteracts partially the ICL-dependent display of ssDNA and pS33 upon exposure to the crosslinking agent (Fig. 3f and g). Decreased pS33 and pS4/8 levels upon CUL4A/B depletion were also found in immunoblots following MMC treatment of both HeLa and SKOV3 cells (Additional file 1: Figure S4 h). These results indicate that CRL4 activity is required for the generation of a ssDNA-RPA signaling platform essential for the DDR mitigating the cytotoxicity of crosslinking agents.
FANCD2-dependent ssDNA formation upon cisplatin treatment
The FA pathway is responsible for the recruitment of nucleases required for the unhooking of crosslinked bases, thus inducing double strand breaks and ssDNA intermediates at ICL sites [37,38,39,40,41,42]. To corroborate the role of CRL4 in stimulating the formation of ssDNA-RPA complexes in cisplatin-treated cells, we exploited the increased level of ssDNA foci induced by 5-μM cisplatin incubation for 24 h (Fig. 4a and b). After this incubation treatment for 24 h, 95 ± 4.3% (N = 7) of control cells remain viable, arguing against the possibility that the accumulation of ssDNA is due to replication catastrophe caused by severely damaged DNA. We then depleted, by siRNA transfection, different members of the FA pathway that have been implicated in DNA damage processing and RPA recruitment. Immunofluorescence analyses (Fig. 4a) and subsequent quantifications (Fig. 4b) revealed that a depletion of FANCD2 is sufficient to prevent ssDNA formation detected after 24 h of cisplatin treatment. This observation, although unexpected in view of previous findings focusing on 4–6 h as the time points for analyses [5], is in line with the notion that FANCD2 constitutes a central member of the FA pathway that organizes downstream effector nucleases [37,38,39,40]. This dependence on FANCD2 indicates that ssDNA formation is triggered by ICLs rather than other forms of damage resulting from cisplatin. We also used the siRNA-mediated strategy to down regulate the upstream FA pathway members FANCM and FAAP24 [43, 44] as well as the core nucleotide excision repair subunit XPA. Unlike FANCD2, depletion of these factors failed to detectably reduce ssDNA formation during the same 5-μM cisplatin treatment for 24 h (Fig. 4a and b). It is possible, however, that low residual level of these factors remaining after siRNA transfections (see Additional file 1: Figure S2 for the efficiency of the siRNA-mediated depletion) were sufficient for their action in the display of ssDNA after ICL induction.
Additionally, RPA2 phosphorylation on serine 33 was assessed (Fig. 4a and c) to prove that depletion of FANCD2 not only suppresses the formation of ssDNA but also the consequent foci of phosphorylated RPA2 in cisplatin-exposed cells. This result was confirmed by immunoblots using two distinct siRNA sequences to deplete FANCD2 (Additional file 1: Figure S5a and S5b). In all cases, the lack of FANCD2 reduced markedly the level of pS33 in cisplatin-treated cells. We concluded that the experimental conditions of our study verify the involvement of FANCD2 in the induction of ssDNA serving as a hub for the initiation of RPA signaling at ICL sites.
CUL4A/B depletion impedes recruitment of FANCD2 and XPF-ERCC1
The above results prompted us to use FANCD2 as the molecular target to test whether CRL4 might impact on the FA pathway in cisplatin-damaged cells. A central step in repair of ICLs is monoubiquitiation of FANCD2 by the FA core machinery. Subsequently, a nuclease complex that includes the structure-specific endonuclease XPF-ERCC1 is recruited to chromatin in order to unhook the ICLs [30, 38, 45]. As expected, exposure to cisplatin increases the nuclear foci of FANCD2 and ERCC1, and, interestingly, this recruitment is stimulated by CRL4 activity. The down regulation of CUL4A/B reduces the level of FANCD2 foci (Fig. 5a and b) as well as the level of ERCC1 foci in cisplatin-treated cells (Fig. 5c and d). These findings imply that XPF-ERCC1 may play a key role in the formation of ssDNA at ICLs. This view is supported by the observation that down regulation of ERCC1 with two different siRNA sequences reduces both the ssDNA foci and the consequent RPA phosphorylation in cisplatin-treated cells (Additional file 1: Figure S5c-e).
As stated above, monoubiquitination of FANCD2 is a key prerequisite for recruitment of the downstream nuclease complex including XPF-ERCC1 to ICL sites [37]. We therefore tested whether CRL4 might influence the FANCD2 ubiquitination in response to crosslinking agents. Upon analysis in immunoblots, the monoubiquitination of FANCD2 is indeed impaired in CUL4A/B-depleted cells relative to CRL4-proficient controls (Fig. 5e). Quantifications are presented in Fig. 5f as the ratio of ubiquitinated and unmodified FANCD2 for each condition over three independent experiments. After treatment with 5-μM cisplatin, ~ 50% of FANCD2 is ubiquitinated in control cells. Instead, in CUL4A/B-depleted cells, only ~ 30% of FANCD2 molecules appear in this modified form after the same cisplatin exposure. These results indicate that CRL4 activity stimulates the monoubiquitination of FANCD2 and, accordingly, the FANCD2-dependent recruitment of downstream nucleases.
CUL4A/B depletion suppresses the S phase checkpoint after exposure to crosslinking agents
Upon DNA damage, replication is inhibited by S phase checkpoints to ensure repair of the damage before cells enter mitosis, which is an important strategy to prevent cell death by replication catastrophe. As the above described results indicate that CRL4 supports damage signaling by stimulating the FANCD2-ERCC1-dependent formation of the ssDNA-RPA complexes, we next assessed whether the abrogated signaling in CRL4-deficient cells affects S-phase progression. HeLa cells depleted of CUL4A/B were incubated for 24 h with cisplatin or MMC. Thereafter, DNA content and DNA synthesis were monitored by measuring 4′,6-diamidino-2-phenylindole (DAPI) binding and 5-ethynyl-2′-deoxyuridine (EdU) incorporation, respectively, in flow cytometry analyses. When cells were transfected with non-coding control RNA, cisplatin inhibited their replicative DNA synthesis in a dose-dependent manner (Fig. 6a and b). At a cisplatin concentration of 5 μM, the EdU incorporation was decreased by nearly 90%. As expected [20], the CUL4A/B depletion on its own was sufficient to elicit intra-S phase checkpoint responses lowering the rate of DNA synthesis. However, such a reduced DNA synthesis compared to non-coding siRNA controls was observed only as long as the cells were not exposed to cisplatin. This is demonstrated by the fact that the combination of CUL4A/B depletion and 5-μM cisplatin treatment resulted in a 2-fold higher EdU incorporation compared to the same cisplatin treatment in CRL4-proficient controls (Fig. 6b).
This elevated DNA synthesis is in agreement with the dampened activation of the intra-S checkpoint transducer CHK1, a direct target of ATR. An increase of CHK1 phosphorylation (generating pCHK1) was observed in CRL4-proficient HeLa cells treated with cisplatin lasting at least 24 h after initiation of drug exposure (Fig. 6c and d). Although there was an initial increase of pCHK1 in CRL4-deficient cells, at later timepoints pCHK1 levels were significantly lower. In MMC-exposed cells, CUL4A/B depletion also impedes CHK1 phosphorylation, albeit partially (Additional file 1: Figure S6a). Although CHK1 had been identified as a possible CRL4 substrate [46], we did not observe any overall changes of CHK1 protein level. One may argue that the reduced phosphorylation of CHK1 in CUL4A/B-depleted cells exposed to crosslinking agents results from a compromised viability. However, over three independent experiments the combined CUL4A/B deficiency had no statistically significant influence on the phosphorylation of CHK2 protein in cisplatin-treated cells (Additional file 1: Figure S6b and S6c). These findings indicate that CRL4 activity stimulates mainly the ATR/CHK1 signaling pathway in cells treated with crosslinking agents.
An identical response with stimulation of DNA synthesis in CRL4-deficient relative to CRL4-proficient counterparts was detected after exposure to MMC (Additional file 1: Figure S6d and S6e). These findings confirm that the cells respond to cisplatin and MMC treatment with an effective S phase checkpoint that suppresses DNA synthesis. However, this cell cycle checkpoint is at least in part abrogated by concomitant depletion of the CRL4 scaffold proteins CUL4A and CUL4B, such that cisplatin- or MMC-exposed and CUL4A/B-depleted cells display higher rates of DNA synthesis than control cells treated with these same crosslinking drug. Cell cycle analyses established that a depletion of FANCD2 or ERCC1 results in a defective S phase checkpoint and elevated DNA synthesis in cisplatin-treated cells (Additional file 1: Figure S6f and S6 g) exactly as observed after CUL4A/B depletion (Fig. 6a). This identical checkpoint defect is consistent with the notion that CRL4 activity positively regulates the observed ICL –> FANCD2 –> XPF-ERCC1 –> ssDNA-RPA –> ATR/CHK1 pathway.
We next tested whether the weakened S phase checkpoint, translating to an accelerated S phase progression, leads to an increase of the M phase population. For that purpose, cisplatin-treated cells were stained for DNA content and histone H3 phosphorylation at position Ser10 (pH3), a well-established marker of mitosis, and subsequently analyzed by flow cytometry (Fig. 6e). The proportion of cells reaching M phase was reduced upon cisplatin exposure in a dose-dependent manner. However, the CUL4A/B depletion doubled the fraction of M phase cells relative to the non-coding siRNA controls (Fig. 6f). Because the tested cisplatin concentration of 5 μM is toxic in CUL4A/B-depleted cells, we concluded that the CUL4A/B deficiency allows for entering mitosis despite irreparable DNA damage, thereby causing mitotic catastrophe.
Inhibition of Neddylation recapitulates the effects of CUL4A/B depletion
Our results suggest that the cytotoxic effect of the neddylation inhibitor MLN4924 is mediated partially by inactivation of the CRL4 ligase (Fig. 1). For further confirmation, we examined whether MLN4924 would reiterate the effect of CUL4A/B depletion on the S-phase checkpoint regulation following cisplatin and MMC exposure (Fig. 7). Efficient CRL4 inhibition is indicated by the disappearance of the slower migrating CUL4 bands, representing the active neddylated form, upon incubation of HeLa cells with MLN4924. The inhibitor alone only weakly activates the DNA damage response, reflected by slight increases in pS33, pS4/8 and pCHK1, whereas cisplatin and MMC induce a pronounced phosphorylation of RPA2 and CHK1, accompanied by nearly complete degradation of CDT1. This ICL-induced DNA damage response was progressively suppressed in the presence of increasing MLN4924 concentrations, thus supporting the described role of CRL4 in stimulating FA pathway-induced ssDNA signaling.