CCDC6 protein levels contribute to the differential sensitivity to PARG and PARP inhibitors in high-grade serous ovarian carcinoma cell lines
The identification of tumour biomarkers that can predict sensitivity and/or resistance to PARPi as well as further enhance the selective cytotoxicity of such therapeutic drugs is highly envisaged in order to establish an accurate personalised medicine of HGSOC. According to preclinical research, attenuation of CCDC6 in several cancer cell cultures enhances sensitivity to PARPi, which have the ability to work in concert with cisplatin [36,37,38,39,40,41].
Here, we investigated if CCDC6 expression would influence how sensitive a panel of ovarian cancer cell lines was to PARPi and PARGi.
To do so, in three ovarian cancer cell lines, namely Kuramochi, OVCAR3 and OV-90 cells, harbouring different genetic backgrounds (Additional File 1: Table S1) [42], we assessed relative sensitivity to the PARPi olaparib in colony formation assays: for these set of experiments we used a single representative drug concentration, as indicated [shCTRL] (Fig. 1A-C). The Kuramochi cells appeared minimally sensitive to PARPi (Fig. 1A) in comparison to the OVCAR3 and OV-90 cells (Fig. 1B, C), as supported by statistical analysis (Fig. 1A-C). The silencing of CCDC6 (shCCDC6) or pharmacological inhibition of the de-ubiquitinase USP7 (P5091), which results in proteasome-dependent degradation of CCDC6 [30] (Additional File 5: Figure S2A-F), determined a positive modulation of the PARPi sensitivity in all three cell lines (Fig. 1A-C).
Conversely, the Kuramochi cells resulted very responsive to the PARGi (PDD00017273) treatment, while OVCAR3 and OV-90 cells scarcely responded to PARGi (Fig. 1A-C) [24]. The low level of PARG and the high levels of ADP-ribosylated proteins, which are revealed by a pan-ADP-ribose antibody, supported the lower PARPi sensitivity and the higher PARGi response of Kuramochi cells, in comparison to OVCAR3 and OV-90 cells [6] (Additional File 5: Figure S2 G, H). In a 2-D colony-forming assay with one concentration drug, CCDC6 silencing diminished sensitivity to PDD00017273 in Kuramochi cells (Fig. 1A). By contrast, the CCDC6 attenuation minimally affected the PARGi sensitivity in OVCAR3 and OV-90 cells (Fig. 1B-C). Nevertheless, when these ovarian cancer cells were challenged with different concentration of PARGi and their vitality assessed by MTT assays, the change of PARGi sensitivity upon CCDC6 downregulation appeared evident, as described below.
The relative CCDC6 and USP7 levels in all the analysed ovarian cancer cells are shown in Additional File 5: Figure S2G.
Together these results demonstrate that the genetic or pharmacological modulation of the CCDC6 protein levels in ovarian cancer cells affects sensitivity to the anti-cancer PARPi as well as to the recently developed PARGi.
CCDC6 depletion affects γH2AX foci formation in response to PARP and PARG inhibitors and impairs double strand breaks repair by Homologous Recombination
We evaluated the formation of γH2AX foci, a marker for double strand breaks (DSBs), in order to thoroughly study the molecular mechanisms of CCDC6-dependent effects on the susceptibility of HGSOC cells to PARPi and/or PARGi [43, 44]. In contrast to the untreated cells, foci formation was significantly induced by both PARPi and PARGi in the Kuramochi and OV-90 cell lines, and CCDC6 silencing significantly reduced these effects. In comparison, PARPi and PARGi had minimal effect on γH2AX foci in control OVCAR3 (shCTRL) cells. However, CCDC6 silencing significantly reduced PARPi-induced foci formation in these cells (Fig. 2A, B). Kuramochi cells responded dramatically to the addition of the PARGi therapy, as previously described [24].
Strikingly, silencing of CCDC6 resulted in a considerable reduction in γH2AX foci in all three HGSOC cell lines as seen in the Figure and by relative intensity quantification (Fig. 2A, B). Notably, over-expression of myc-CCDC6 plasmid rescued γH2AX foci formation in CCDC6-downregulated cells, thus excluding off-target issues (Additional File 6: Figure S3, Additional File 7: Figure S4, Additional File 8: Figure S5).
The drugs’ effect observed upon CCDC6 silencing suggested an impairment of DSBs repair by HR in the analysed ovarian cancer cells. In order to assess the efficacy of the HR repair, we employed the DR-GFP reporter system [32]. A schematic representation of the DR-GFP assay is shown in Additional File 2: Figure S1. In these assays, we decided to restrain endogenous CCDC6 activity by using the pharmacologic inhibitor P5091. The ovarian cancer cells, left untreated or after pre-treatment with P5091, were transfected with the DR-GFP reporter plasmid alone, as a control, or together with the I-SceI plasmid able to induce DSBs. The ability to repair the DSBs by HR was measured by flow cytometry and the frequency of HR was reported as a percentage of GFP positive cells. Treatment with USP7 inhibitor determined a significant decrease of the GFP positive cells, compared to non-treated cells, suggesting that the reduction of CCDC6 levels affected the DNA repair by HR in all the ovarian cancer cells analysed, even if the effects were most evident in OVCAR3 and OV-90 cells, with respect to Kuramochi cells (Fig. 2C). The phenotype induced by the P5091 treatment was mainly dependent on the CCDC6 increased turnover, as it was almost completely (Kuramochi) or partially (OVCAR3 and OV-90) reverted following the myc-CCDC6 transient transfection (Fig. 2C). In ovarian cancer cells the transfection efficiency of HA-ISceI, in presence or absence of the myc-CCDC6 plasmid, was assessed by western blot (Fig. 2D-F).
These data suggest that CCDC6 loss of function confers an HR-deficiency phenotype in HGSOC cancer cells.
CCDC6 loss sensitises the high-grade serous ovarian carcinoma cell lines to combined treatment with PARP inhibitors and cisplatinum
Since the HR deficiency is also accompanied by sensitivity to PARPi, we used cellular metabolic activity as a marker of cell survival to characterize and quantify the effects of CCDC6 downregulation on olaparib sensitivity in HGSOC. To do this, we reduced CCDC6 protein levels in ovarian cancer cells using the USP7 inhibitor P5091 and evaluated sensitivity to various olaparib concentrations. Similarly, PARGi and cisplatin sensitivity were also assessed (Fig. 3A-I). Generally, CCDC6 downregulation by P5091 improved sensitivity to olaparib in all cell lines herein analysed. Specifically, the sensitivity to olaparib in Kuramochi cells (IC50 7.76 μM ± 0.74) was positively modulated to 3.45 μM ± 0.39 by the CCDC6 chemical attenuation (Fig. 3A). Stronger cytotoxic effects were attained in P5091-pretreated OVCAR3 and OV-90 cells when exposed to olaparib compared to Kuramochi cells; these cells' sensitivity increased by about three and five times, respectively (Fig. 3D, G). Similarly, genetic downregulation of CCDC6 by shRNA leads to overlapping improved sensitivity to olaparib in all cellular models herein analysed (Additional File 9: Figure S6A, D, G).
Platinum containing drugs are widely used for the treatment of ovarian cancer. However, ovarian cancer patients initially responsive to platinum containing drugs invariably relapse becoming resistant. Here we asked whether CCDC6 downregulation may affect the vulnerability of HGSOC to cisplatin alone or in combination with olaparib. While sensitivity to cisplatin was only slightly affected by P5091 (IC50 5.89 μM ± 0.07) in Kuramochi cell lines compared to vehicle (IC50: 6.63 μM ± 0.02) (Fig. 3A), CCDC6 chemical downregulation significantly modifies cisplatin sensitivity in OVCAR3 (IC50 2.42 μM ± 0.03 vs IC50 1.20 μM ± 0.04) and OV-90 cells (IC50 4.49 μM ± 0.05 vs IC50 0.48 μM ± 0.03) (Fig. 3D, G ). Strikingly, in these cells the combined treatment of olaparib with cisplatin showed a synergistic effect, magnified by the CCDC6 accelerated turnover, upon P5091 addition (Fig. 3E, H).
Conversely, sensitivity to PARGi in CCDC6 downregulated cells behaved differently compared to olaparib treatment, which is explained by their different mechanism of action (as discussed later in the text). As reported, Kuramochi cells resulted highly responsive to the PARGi treatment (IC50 0.49 μM ± 0.35). Notably, the CCDC6 downregulation by P5091 treatment affected PARGi-induced cytotoxicity (in the presence of P5091 at 2.5 μM, the IC50 raised to 1.33 μM ± 0.32) (Fig. 3A). Similar results were obtained by silencing CCDC6 by shRNA (Additional File 9: Figure S6A). The CCDC6-dependent weak response to PARGi was also visible in intrinsically PARGi-resistant OVCAR3 and OV-90 cell lines, where the low sensitivity to PARGi was enhanced by P5091 (Fig. 3D, G). In detail, following the PARGi addition, the CCDC6-depleted OV-90 cells showed an IC50 of 17.05 μM ± 0.72 compared with an IC50 of 11.02 μM ± 0.43 of control cells. The CCDC6-depleted OVCAR3 cells also exhibited a mild modulation showing an IC50 of 18.94 μM ± 0.59 compared with an IC50 of 17.60 μM ± 0.10 of control cells (Fig. 3D, G ). Importantly, all these phenotypes were rescued by CCDC6 exogenous expression in CCDC6 chemically or genetically downregulated cell lines (Additional File 9: Figure S6 A-B, D-E and G-H), thus supporting the specificity of our observations. In detail, in Kuramochi-CCDC6-depleted cells the re-expression of CCDC6 (Additional File 9: Figure S6B) led to an IC50 of 0.76 μM ± 0.10 vs an IC50 of 1.11 μM ± 0.08 detected in control cells (transfected with the empty vector, EV); in OVCAR3 CCDC6 silenced cells the re-expression of CCDC6 (Additional File 9: Figure S6E) determined an IC50 of 18.52 μM ± 0.08 vs an IC50 of 19.36 μM ± 0.05 relieved in control cells; and in OV-90 cells CCDC6-depleted, the re-expression of CCDC6 (Additional File 9: Figure S6H) led to an IC50 of 15.08 μM ± 0.07 vs an IC50 of 17.80 μM ± 0.02 detected in control cells (transfected with the empty vector, EV). By contrast, the overexpression of CCDC6 in wild type Kuramochi, OV-90 and OVCAR3 cells did not determine any variation in the IC50 values upon treatment with different doses of PARGi (Additional File 9: Figure S6B, E, H).
By subjecting HGSOC cells to various PARPi with varied trapping and inhibitory potencies, we further tested whether the increased sensitivity to olaparib could be replicated [45]. Both talazoparib and veliparib, PARP inhibitors with greater PARP activity inhibition and trapping action, have recently received clinical approval for breast cancer treatment, enabling dosage reduction and improving clinical efficacy [17]. Therefore, we examined their effectiveness alone or in conjunction with P5091 in our HGSOC preclinical models. Intriguingly, pharmacological downregulation of CCDC6 enhances the cytotoxic effects of both PARPi in all three cellular models, with P5091-treated OVCAR3 cells showing the most impressive 18.6-fold increase in talazoparib sensitivity (Fig. 3C, F, I).
These findings imply that CCDC6 downregulation enhanced HGSOC cell line sensitivity to various PARPi, regardless of their trapping and inhibitory efficacy.
CCDC6 downregulation rescues the sensitivity to olaparib in newly generated PARP inhibitor resistant ovarian cancer cells
Our results suggest the potential of predicting PARPi sensitivity in HGSOCs harbouring CCDC6 loss of functions as well as of downregulating CCDC6 by USP7 inhibitor, even in PARPi-resistant, BRCA wild type (WT), HR competent HGSOC cells. In order to address the intriguing properties of CCDC6 downregulation in re-sensitising PARPi-resistant HGSOC, we generated olaparib-resistant OVCAR3 and OV-90 cell lines, thereafter named OVCAR3olaR and OV-90olaR, respectively, by exposing cell lines to PARPi at IC50 doses and up to 10 μM for two weeks, as described in material and methods section. Compared to the parental OVCAR3 and OV-90 cells, the cytotoxic effects of the PARPi treatment were quantified in the newly generated OVCAR3olaR and the OV-90olaR by cell viability assays (Fig. 4A). Of note, compared to parental cells, CCDC6 protein levels were slightly increased in OV-90olaR at IHC and western blot, while resulted comparable in PARPi-resistant OVCAR3 cells (Additional File 10: Figure S7 D-G). However, the CCDC6 attenuation upon the P5091 treatment, rescued the sensitivity to PARPi in OVCAR3olaR and OV-90olaR HGSOC cells (Fig. 4C, D).
The generation of the olaparib resistance phenotype left nearly unaffected the cisplatin sensitivity in these cells (Fig. 4A). However, while in OV-90olaR cells the combined treatment of cisplatin and olaparib resulted in an antagonistic effect (CI > 1), in OV-90 parental cells it determined a synergistic effect (CI < 1) (Fig. 4B). Interestingly, in the presence of P5091, which increases the CCDC6 degradation, the combined treatment of cisplatin and olaparib turned into an additive effect in OV-90olaR cells (CI = 1) (Fig. 4E). No variations in the combination index between OVCAR3 parental and resistant cells upon cisplatin and olaparib treatment were registered (Fig. 4B).
Then, by hypothesizing CCDC6 as a possible biomarker of cisplatin response, we investigated the CCDC6 expression levels in three ovarian patient-derived xenografts (PDXs) couples, sensitive and with acquired resistance to cisplatin (MNHOC 266, MNHOC 124 and MNHOC 239), obtained after multiple in vivo drug treatment [46,47,48]. In particular, the PDX MNHOC 266, carries mutations in BRCA1, whereas the PDXs MNHOC 124 and MNHOC 239 are BRCA1/2 WT. The models recapitulate the clinical setting of cisplatin resistance. Following an immunohistochemical staining to assess the levels of CCDC6, a strong expression was found in all PDXs (Fig. 5A, B), with a pattern of CCDC6 staining that was comparable in sensitive and the matching cisplatin resistant PDXs.
Although the CCDC6 staining data on PDXs do not discriminate a primary from a resistant ovarian tumour model, we may hypothesise, based on in vitro evidence, that the CCDC6 positive staining could otherwise predict the response to PARPi, as described in other tumour models [27].
All the PDXs were non-responders to the olaparib treatment. They reproduced the phenotype of the newly generated olaparib-resistant ovarian cancer cells, where CCDC6 showed a high level of expression, mostly mimicking the PDX phenotype (Fig. 5, Additional File 10: Figure S7 D-E).
We used the well-known PEO1 and PEO4 cells [49], which represent naturally occurring CCDC6 null and CCDC6 competent ovarian cancer cell types, respectively, that display differing protein levels despite having equivalent amounts of CCDC6 transcripts, to test this hypothesis (Fig. 5C, D). These cells were from the same patient who had a poorly differentiated serous adenocarcinoma and who after chemotherapy had a status change from platinum sensitivity (PEO1) to platinum resistance (PEO4) [49].
Notably, the PEO1 cells are the first and only human BRCA2 defective ovarian cancer cell line identified thus far and, like the original patient, possess a BRCA2 hemizygous nonsense mutation 5193C > G (Y1655X). PEO4 cells, derived from ascites at the time of relapse with cisplatin resistance, have the secondary mutation (Y1655 mutation spontaneously reverted) and are BRCA2 proficient.
Following CCDC6 knockdown by shRNA in the cisplatin-resistant PEO4 primary cells, the cytotoxic effects of olaparib and the cisplatin sensitivity were measured. The very low sensitivity to olaparib in the PEO4 cells (IC50, 56.15 μM ± 0.18) was positively modulated by the CCDC6 lowering (IC50 dropped to 36.45 μM ± 0.06) (Fig. 5E); strikingly, in these cells, the cisplatin response (IC50 35.12 μM ± 0.15) was also positively modulated by the CCDC6 attenuation (IC50 23.23 μM ± 0.06) (Fig. 5E, Additional File 10: Figure S7A). Most importantly, upon CCDC6 depletion, the combined treatment of cisplatin and olaparib determined a synergistic effect (CI < 1), while an antagonistic effect (CI > 1) was observed in the PEO4 parental cells (Fig. 5F). Remarkably, the dose response index (DRI) was impressively modulated by the drug combination in PEO4 cells, silenced for CCDC6 (DRI > 1), since the IC50 concentration dropped by more than 30% for both the drugs concentration (Fig. 5F).
Notably, the PEO1 cells resulted naturally null for the CCDC6 protein expression, although CCDC6 mRNA is well expressed in these cells and in PEO4 cells. This suggests that post-translational processes are responsible for CCDC6 deregulation in PEO1 cells. According to preliminary results (Morra F et al., in preparation), the low amount of CCDC6 protein detected in PEO1 cells might be ascribed to GSK3β gain of activity, which is common in BRCA2-mutant cells [50], in turn able to sustain the FBXW7-dependent CCDC6 degradation [30]. Importantly, CCDC6 protein levels spontaneously re-established in BRCA2 reverted PEO4 cells (Fig. 5C). The CCDC6 null, cisplatin sensitive, PEO1 cells showed a high PARPi sensitivity (IC50: 2.3 μM ± 0.15). This phenotype was abolished by replenishing the expression of CCDC6, as observed by a 70% increase of the IC50 drug concentration (IC50: 4.11 μM ± 0.17) (Fig. 5E). Moreover, cisplatin sensitivity remained nearly unaffected in the presence of CCDC6 ectopic expression (in PEO1 CCDC6 + : IC50 0.78 μM ± 0.11 vs. IC50 1.01 μM ± 0.06 in PEO1 EV cells). However, upon cisplatin and olaparib combined treatment, no significant variations in the combination index between the PEO1 cisplatin sensitive cells, overexpressing either the Myc empty vector (EV) or the Myc CCDC6 plasmid (CCDC6 +), were registered. Thus, CCDC6 expression affects olaparib, but not cisplatin sensitivity of these cells. Furthermore, the ability to repair the DNA DSBs by HR was determined in the PEO1 cells by GFP reporter assays, which revealed a slightly increase in the GFP positive cells following the CCDC6 transient transfection [Myc CCDC6 plasmid (CCDC6 +)], in comparison to the control cells [Myc empty vector (EV)] where a very low percentage of GFP positive cells was detected, mostly ascribed to the presence of BRCA2 inactivating mutation (Additional File 10: Figure S7A-C).
This observation indicates that, in the context of BRCA2 deficiency, CCDC6 activity in HR repair is limited and is epistatic to BRCA2. Nevertheless, in the PEO4 cells, in which BRCA2 functional activity and CCDC6 levels are restored, the CCDC6 depletion determined a significant decrease of GFP positive cells, compared to the control HR proficient cells (Additional File 10: Figure S7B-C).
Overall, our results imply that assessing CCDC6 levels in tumours may provide critical information for therapy choices in HGSOC.
The CCDC6-PP4c interaction depends on phosphorylation and determines PARP inhibitors sensitivity by modulating γH2AX levels
To further investigate the molecular mechanisms of PARPi improved sensitivity in HGSOC where CCDC6 is downregulated, we dissected CCDC6 roles in HR response to DNA damage. Indeed, by regulating histone H2AX phosphorylation status, CCDC6 contributes to efficient DDR through HR. Upon DNA damage exposure and in an ATM-dependent manner, CCDC6 moves from cytosol to the nucleus where it binds the main phosphatase responsible for the maintenance of histone H2AX phosphorylation status, PP4c [28, 51]. The PP4 holoenzyme, consisting of the catalytic subunit (PP4c) and the two major isoforms of PP4R3 (PP4R3α/β, also known as SMEK1/2) binds to the F423xxP426 motif in CCDC6, providing specificity [29]. Moreover, the residue Threonine 427, close to the motif, can modulate the CCDC6 binding to PP4 and determine the CCDC6 intracellular localization through its phosphorylation status [29]. In order to exploit the functional outcome of the Threonine 427, flanking the FxxP interaction motif and preventing, when phosphorylated, the binding to PP4, we performed site-directed mutagenesis of this residue in alanine (T427A) or aspartate (T427D). We investigated whether the disruption of CCDC6-PP4c interaction might affect the histone H2AX phosphorylation leading to HR deficiency and to PARPi sensitivity.
In HGSOC cells, pharmacologically or genetically silenced for CCDC6, and upon treatment with different concentrations of olaparib, as indicated, the cytotoxic drug effects were quantified by a cell viability assay upon overexpression of myc-CCDC6WT (wild type), CCDC6T427A, CCDC6 T427D or empty vector (EV), as control (Fig. 6A). The olaparib sensitivity observed in the OVCAR3 stably silenced for CCDC6, and transiently transfected with the EV, (IC50: 1.90 μM ± 0.45), decreased following the transient transfection of myc-CCDC6WT (IC50: 3.37 μM ± 0.23) and myc-CCDC6T427A (IC50: 3.08 μM ± 0.20) plasmids; this data suggests that the T427A mutant, as well as the wild type protein, maintains the ability to interact with PP4c and to inhibit its phosphatase activity toward histone H2AX [29]. However, even after forced expression of the CCDC6T427D mutant, OVCAR3 remains sensitive to olaparib whether treated with the USP7i P5091 or stably silenced for CCDC6, indicating that this mutant is unable to engage with PP4c phosphatase (Fig. 6A, B). As seen in the western blot analysis, this causes an increase in PP4c activity and a corresponding decrease in histone H2AX phosphorylation, which reproduces the CCDC6 deletion phenotype (Fig. 6C). γH2AX levels were assessed in olaparib-treated (1 μM) or untreated CCDC6-silenced ovarian cancer cells by western blot. This analysis revealed lower γH2AX levels in cells overexpressing the T427D mutant with respect to those expressing the T427A mutant or the wild type CCDC6 protein (Fig. 6C, Additional File 11: Figure S8).
In addition, the suppression of the regulatory component PP4R3α hampered the rescue of the CCDC6 wild type or CCDC6T427A mutant plasmids overexpression, as measured by cell survival experiments, upon olaparib treatment, by blocking the CCDC6-PP4 interaction through the FxxP motif (Fig. 6D, E).
According to our findings, the PP4 complex activity, which is made up of the catalytic subunit PP4c and the regulatory subunit 3α, (PP4R3α,), is necessary for the CCDC6-dependent BRCA-ness status.