Cancer Cells Haploinsufficient for ATM Are Sensitized to PARP Inhibitors by MET Inhibition

Abstract

The MET oncogene encodes a tyrosine kinase (TK) receptor. Its activation protects cells from death but also stimulates DNA damage response by triggering excess replicative stress. Transcriptomic classification of cancer cell lines based on MET expression showed that response to the PARP inhibitor (PARPi) olaparib is poorer in MET overexpressing cell lines. Accordingly, a high MET expressing lung carcinoma cell line was sensitized to PARPi by MET TK inhibition. This was not linked solely to MET overexpression: other MET overexpressing cell lines were biochemically but not functionally responsive to combined inhibition. Moreover, exogenously induced MET overexpression was unable to induce resistance to PARPi. The MET overexpressing cell line, responsive to the combined PARP and MET inhibition, carried a heterozygous mutation of the ATM gene and showed an attenuated response of ATM to PARPi. Among the downstream targets of ATM activation, NuMA was phosphorylated only in response to the combined PARP and MET inhibition. Given the role played by NuMA in mitosis, data show that the latter is affected by MET and PARP inhibition in cells with haploinsufficient ATM. This is important as ATM heterozygous mutation is frequently found in human cancer and in lung carcinomas in particular.

1. Introduction

The treatment of cancer has changed dramatically over the last decade, driven by increased understanding of the cancer genome, leading to the development of targeted agents. Among the most successful targeted drugs, inhibitors of the poly-ADP-polymerase (PARPi) are approved for the treatment of tumors harboring defects of genes involved in repairing DNA double-strand breaks [1]. First, seminal papers introduced the concept that PARP inhibitors have synthetic lethal interaction in homologous recombination defective tumors [2,3], such as ovarian and breast cancer carrying defects of the BRCA1/2 genes. Then, given the encouraging clinical results obtained with PARPi in BRCA1/2 deficient cancers, several studies were undertaken to identify other proteins whose defects might confer PARPi sensitivity (for a review, see [4]). Among mutated genes potentially conferring susceptibility to PARPi, proteins directly involved in homologous recombination repair [5], as well as those involved in sensing DNA damage, have been associated with PARPi response/resistance [4]. Examples are REV7, RINN1/2/3, MRE11 and EZH2 (for reviews, see [4,6]). Among the latter, ataxia telangiectasia mutated (ATM) is a DNA damage checkpoint also able to indirectly activate homologous recombination repair [7] and/or to counter toxic non-homologous end-joining (NHEJ) at broken replication forks [8]. ATM is a serine-threonine kinase at the apex of a cascade responsible for global orchestration of cellular responses (for a review, see [9]). In ATM proficient cells, irradiation results in the phosphorylation of consensus sites recognized by ATM and ATR encompassing over 700 proteins [10]. Homozygous or compound heterozygous germline ATM mutation causes ataxia-telangiectasia, an autosomal recessive multisystem disorder presenting in childhood with progressive cerebellar ataxia, oculocutaneous telangiectasia, immune deficiency, radiosensitivity, and cancer predisposition [11]. In human sporadic cancers, somatic ATM aberrations have been frequently found using next-generation sequencing [12]. COSMIC reports ATM point mutations in 1% to 10% of lung, endometrial, kidney, liver, esophageal, ovarian, salivary gland, gastric, thyroid and urinary tract cancers. Excluding variants of unknown significance, 167 distinct, possibly functional, somatic mutations in ATM have been observed in a broad range of malignancies, and are associated with loss of heterozygosity in approximately 20% of cases (for a review, see [13]).

DNA damage response (DDR) is also impacted by signaling of growth factor tyrosine kinase receptors (TKRs). Several oncogenes activated in cancer lead to DDR activation as a consequence of excess replicative stress, which may result in genomic instability and alterations in checkpoint and repair mechanisms [14]. Contrarily, activation of oncogenic TKRs has been associated with resistance to genotoxic stress (see e.g., [15,16]), as it protects cells from DNA damage via different downstream signaling pathways. Remarkable advances have been made over the last two decades in the discovery and clinical development of a wide range of molecular entities targeting TKRs (for a review, see [17]). Among them, monoclonal antibodies such as Trastuzumab targeting HER2 for the treatment of breast cancer and TK small molecule inhibitors such as Gefitinib and Erlotinib targeting the epidermal growth factor receptor (EGFR) for the treatment of lung cancer have shown their clinical efficacy.

One such TKR possibly involved in the DDR is the hepatocyte growth factor (HGF) receptor, encoded by the MET oncogene. This oncogene was named after its isolation as an oncogenic gene from an osteosarcoma cell line transformed in vitro by the chemical carcinogen methyl-nitro-nitroso-guanidine (MNNG) [18] and found deregulated and/or activated in a variety of human tumor types and serves as an oncogenic target (for a review, see [19]). Phosphoproteomic analysis of MET overexpressing cells after irradiation and treatment with METi revealed the modulation of several substrates of the DDR [20]. In line with these observations, MET inhibition was shown to sensitize glioblastoma cells to irradiation (IR) [21]. Cooperation of MET inhibitors with PARPi has been reported in breast cancer cell lines exposed to hypoxia that resulted in MET activation [22] and nuclear transport [23] where it could phosphorylate PARP1 directly.

The aim of this work was to find a way to enlarge the spectrum of applications of approved drugs in the treatment of highly aggressive human cancers. Using in silico and in vitro experimental approaches, including functional assays and protein analyses, we show PARPi effectiveness in cells with heterozygous ATM aberration in combination with low doses of MET inhibitors. This work is novel and has clinical impact as low and thus manageable doses of these inhibitors in combination treatments have not been studied. In addition, we demonstrate a new mechanism for explaining the effectiveness of this drug combination.

2. Results

2.1. MET Overexpressing Cells Are Made Susceptible to PARPi by MET Inhibition

We interrogated MET gene expression and susceptibility to the PARPi olaparib in 547 cancer cell lines reported in the Cancer Cell Line Encyclopedia (CCLE) project database [24], available through the tools for CCLE Data Visualization and Analysis DepMap portal (https://depmap.org/portal (accessed on 17 May 2022)). Data on olaparib susceptibility were available for 200/547 cell lines. We found that cancer cell lines overexpressing the MET oncogene were less sensitive to olaparib than cell lines not expressing or showing lower levels of MET expression (Figure 1).

Therefore, we aimed at verifying in preclinical experiments if MET overexpressing cells might be sensitized to olaparib by MET inhibition. We used the following cell lines that show the highest levels of MET expression [25]: EBC-1 cells established from a lung carcinoma and GTL16 cells established from a liver metastasis of gastric adenocarcinoma, both carrying MET amplification, and the Hs746T established from a muscle metastasis of gastric adenocarcinoma, harboring also MET mutation. As expected, these cells showed constitutive activation of the MET receptor (Figure 2A) and are the most sensitive to the highly selective MET kinase inhibitor JNJ-38877605 (hereafter referred to as JNJ) in the list of 923 cell lines reported in the Genomics of Drug Sensitivity in Cancer (GDSC) database (www.cancerRxgene.org (accessed on 17 May 2022)), a public resource for information on drug sensitivity in cancer cells and molecular markers of drug response [26]. In this database, the IC₅₀ values for JNJ are as follows: EBC-1, 4.96 μM; Hs746T, 0.65 μM; and MKN45, 1 μM. Figure 2B–D show that these cell lines are not intrinsically susceptible to olaparib, as they all displayed a GR₅₀ (normalized growth rate) [27] similar to that of lines described as olaparib resistant [28,29]. Our data are in line with the IC₅₀ for olaparib reported in GDSC1, as follows: EBC-1, 45.27 μM; Hs746T, 52.16 μM; and MKN45, 111.44 μM. We then treated cells with olaparib in combination with low doses of JNJ (10 nM, i.e., 100 times lower than the IC₅₀ mentioned above). The EBC-1 cells (Figure 2B), but not the GTL16 and Hs746T cells (Figure 2C,D), were more susceptible to olaparib. Notably, the JNJ dose able to sensitize EBC-1 cells to olaparib did not fully abrogate the constitutive MET TK phosphorylation at tyrosines 1234–1235, which is a proxy of the receptor constitutive activation (Figure 2A). We tested likewise the EGFR inhibitor gefitinib, as EGFR TK is also known to protect cells from genotoxic stress ([15]). The IC₅₀ values for gefitinib reported in GDSC1 are as follows: EBC-1, 66.31 μM; Hs746T, 41.93 μM; MKN45, 75.09 μM. As shown in Figure 2B–D, none of these cell lines was affected by the combination of gefitinib and olaparib.

To further assess the possible translational impact of the cooperation between PARPi and METi, we tested this combination in a panel of ovarian cancer cell lines, using PARP inhibitors approved for the treatment of human epithelial ovarian carcinomas (EOC) that represent a fraction of susceptible ovarian carcinomas [30]. Susceptibility to PARPi was tested in the following BRCA1/2 mutated EOC cell lines: UWB1.289 (homozygous p.2594delC BRCA1), IGROV1 (heterozygous frameshift deletion of BRCA1 and missense mutation of BRCA2), OVCAR8 (BRCA1 gene methylated), COV362 (splice site mutation in BRCA1) and, as a control, the BRCA1/2 fully proficient A2780 cell line. These cell lines showed differential susceptibility to PARP inhibitors (olaparib, veliparib, rucaparib and niraparib, Supplementary Figure S1), in agreement with previous reports [28,29] that showed that not all BRCA1/2 mutated lines are sensitive to PARP inhibition. Notably, in agreement with previously published data [31], the UWB1.289 cells showed the highest susceptibility to all of the PARP inhibitors tested (Supplementary Figure S1).

Expression of the MET receptor was detected in all ovarian cancer cell lines except A2780 (Figure 3A). UWB1.289 and IGROV1 also showed a basal level of MET phosphorylation at tyrosines 1234–1235 in the absence of the added specific ligand HGF and irrespective of the level of MET protein expression (Figure 3A). These phospho sites represent established markers of activation of the MET receptor. However, we found that this activation was not indicative of susceptibility to MET inhibition: cells showing a basal constitutive MET activation were neither per se affected by JNJ (Figure 3B) nor showed a significantly increased susceptibility to olaparib when administered in combination with JNJ (Figure 3D).

2.2. Exogenous MET Activation Is Not Sufficient to Convert PARPi Susceptible Cells to Be Resistant

We tested whether an increased level of MET expression, resulting in further MET activation, could impair ovarian cancer cell response to PARPi. Using lentiviral vectors, we transduced the most PARPi responsive UWB1.289 EOC cells to overexpress the wild-type MET receptor (Figure 4A). Exogenously induced MET overexpression resulted in increased constitutive activation, in the absence of the ligand, and in biochemical susceptibility to JNJ (Figure 4A). However, MET overexpression was not associated with either increased functional susceptibility to JNJ (Figure 4B) or increased resistance to PARP inhibitors olaparib and rucaparib (Figure 4C,D).

Altogether, these data show that the exogenously induced MET activation was not, on its own, sufficient to make PARPi-susceptible cells resistant to PARP inhibition.

2.3. In MET Overexpressing Cells with ATM Mutation, Combined PARP and MET Inhibition Affects a Downstream Cascade Leading to the Phosphorylation of the Nuclear Mitotic Apparatus (NuMA) Protein

As PARPi and METi cooperation in triple negative breast cancer cells has been previously described and reported as dependent on both proteins’ localization within cell nuclei [22] where the two proteins might interact, leading to PARP1 phosphorylation, we assessed PARP1 and MET localization in the above-listed MET overexpressing cell lines. As shown in Supplementary Figure S2, PARP1 and MET were respectively located in the nuclei and at the cell membrane, and were never found in the same cellular compartment.

To understand the mechanism of the distinctive response of EBC-1 cells to the combined PARPi and METi treatment, we compared their genomic profiles to those of the non-responsive cell lines and found that EBC-1 cells carry a heterozygous mutation of the ATM gene.

As reported above (Figure 2B), this genetic alteration in EBC-1 cells was not sufficient per se to make these cells responsive to olaparib. In agreement, Figure 5 and Figure 6B show that olaparib treatment was able to induce ATM phosphorylation in EBC-1 cells only at higher concentrations and at a later timepoint than those reported to be necessary for ATM activation by olaparib in non-ATM mutated cells [32,33]. The addition of JNJ to olaparib did not further increase the phosphorylation of ATM in response to olaparib alone (Figure 6B).

We then asked which mechanism might affect cell proliferation upon treatment with the combination of olaparib and JNJ. We used an unbiased approach to discover connections among molecules associated with cell response to either olaparib or JNJ. Ingenuity Pathway Analysis (IPA, Figure 6A) showed that the pathway from one inhibitor to the other involves ATM. It is known that ATM phosphorylates hundreds of substrates [10], although it is unclear which of these is functionally important. Moreover, ATM phosphorylates and activates other protein kinases that phosphorylate yet more substrates, meaning that ATM-dependent signaling events are not just restricted to factors directly phosphorylated by ATM. Interestingly, the unbiased IPA analysis resulted in connecting olaparib and JNJ cellular response through the ATM substrate CHECK2 and the nuclear mitotic apparatus (NuMA) protein. As shown in Figure 6B, we have confirmed that CHECK2 phosphorylation was associated with ATM activation after prolonged treatment with olaparib alone and also when combined with JNJ. Interestingly, NuMA was observed to be phosphorylated only in cells treated with the combination of olaparib and JNJ (Figure 6B).

3. Discussion

After the initial biological and subsequent clinical demonstration of the effectiveness of PARPi in cells with BRCA1/2 defects, other biomarkers of response to PARPi have been sought to select potentially responsive tumors (for a review, see [4]). We show here that heterozygous somatic ATM mutation alone is not sufficient to make cells susceptible to PARPi, but that heterozygous ATM mutation, leading to attenuated ATM activation by PARPi, makes cells more susceptible to the combined inhibition of PARP and MET tyrosine kinase. Indeed, ATM signaling attenuation has been reported to sensitize cells to DNA damaging agents, such as chemotherapeutics [34]. Moreover, lung cancer cells carrying ATM mutation and wild-type RAS and BRAF, such as EBC-1, display a strong dependency on MEK linked to the ATM heterozygous mutation [35].

As mentioned above, ATM is one of the most common genes that is somatically mutated in cancer, in particular in lung adenocarcinomas [13]. Germline heterozygous variants of ATM have been associated with lung cancer susceptibility and onset [36]. While ATM heterozygous knockout mice with 50% loss of Atm protein did not show a pronounced DNA repair phenotype [37], and ATM null mice were viable, mice with kinase dead ATM died before birth [38]. Although neither observation in mice can directly be transferred to the interpretation of somatic mutations in human cancer, this reinforces the concept that the absence of a protein does not equal the presence of a defective counterpart. ATM protein might be such an example: resting ATM forms a homodimer, and heterozygous mutations might exert dominant negative effects by interfering with the dissociation and activation associated with autophosphorylation [39].

ATM haploinsufficiency is the feature that make cells overexpressing MET susceptible to the combined treatment even at low doses of the small molecule METi. Data shown here confirm that MET expression and activation are not sufficient to make MET overexpressing cells, such as ovarian cancer cells, susceptible to METi or to the combined treatment, in line with the report that no synergistic effects of the combined treatment with PARPi and METi were found in cell lines with BRCA1 or BRCA2 deficiency [40]. This agrees also with the recent report that increased MET protein expression was found in primary cultures of PARPi resistant high-grade serous ovarian carcinomas [41]. Again, in line with data reported here, it has been recently shown that in cellular models of castration-resistant prostate cancer, MET inhibition enhances the efficacy of PARPi by suppressing the ATM/ATR and PI3K/AKT pathways [42].

We unveiled the mechanism of the efficacy of this combined PARPi and METi treatment, which relies on the failure of kinase activation triggered by ATM. We show here that in cells carrying heterozygous somatic ATM mutation, ATM activation by PARP1 inhibition requires higher doses and longer treatment with PARPi than those necessary for ATM phosphorylation in ATM proficient cells [32,33]. We also found that prolonged and stronger treatments were required for the substrate CHK2 phosphorylation to be observed. The mechanism by which olaparib affects the phosphorylation status of ATM and other molecules is still a conundrum. It is well known that olaparib causes DNA damage and that DDR results in ATM phosphorylation. On the other hand, it is not clear which DNA damage caused by olaparib is directly eliciting ATM phosphorylation. Indeed, PARP1 protein was initially identified for its role in the detection and repair of single-strand DNA breaks, but subsequent evidence suggested that PARP1 may also have a role in alternative DNA repair pathways, including nucleotide excision repair, nonhomologous end joining (both classical and alternative), homologous recombination, DNA mismatch repair and fork damage (for a review, see [43]).

In search of a protein involved in the specific response to PARPi and METi combination, we used IPA, which drew attention to the nuclear mitotic apparatus (NuMA) protein. This was not totally unpredictable, as the Ser395 of NuMa is one of the hundreds of consensus sequences phosphorylated by ATM in response to irradiation [10]. Moreover, the phospho-proteomic analysis of EBC-1 cells treated with irradiation and MET inhibition showed the increased phosphorylation of NuMA [20]. NuMA is a key structural nuclear protein, that binds microtubules and plays a role in the formation and maintenance of the spindle poles and the alignment and segregation of chromosomes during mitotic cell division (for a review, see [44]). NuMa localization and phosphorylation status vary throughout mitosis [45]. NuMA is phosphorylated by a number of different kinases, such as CDK1 [46,47] and ATM [48], and its phosphorylation is required for attachment to, and assembly of, the mitotic spindle. NuMA phosphorylation is also required for its efficient PARylation (by PARP3 and tankyarase 1), in turn necessary for correct spindle assembly [49]. It is noteworthy that olaparib, as well as rucaparib, inhibits PARP3 as well as PARP1 and PARP2 [50]. PARP3, which participates also in the DSB repair pathway(s), by interacting with NuMA and Tankyrase 1 [51], regulates mitotic progression. Altogether, our data suggest that a more complex interplay between MET activation/inhibition and PARP inhibitor activity might interfere with mitotic spindle organization, resulting in mitotic blockade and directing cells to mitotic catastrophe.

In conclusion, the model shown here introduces the proof of concept that ATM haploinsufficiency, which is found quite frequently in human cancers and alone is not able to limit the effectiveness of genotoxic pharmacological agents, such as PARP inhibitors, can be circumvented by interfering with other pathways (e.g., MET) leading to mitosis regulation. These results have clinical impact as low and thus manageable doses of these inhibitors in combination treatments are effective in killing cancer cells with a specific and identifiable genetic profile.

4. Materials and Methods

4.1. Cell Lines and Reagents

All but the GTL16 cell lines used are commercially available. The EBC-1 cells were derived from a metastatic tumor of a patient with lung squamous cell carcinoma and were purchased from the Japan Cancer Resources Bank. The other cell lines were purchased from ATCC and cultured as suggested by the provider. GTL16 was a clonal gastric cell line established in our laboratory [52]. JNJ-38877605 was from OCTIMET Oncology, and olaparib, rucaparib and gefitinib were purchased from Selleck Chemicals (Munich, Germany). Olaparib and rucaparib were used in functional assays in escalating doses, given that 10–40 μM are the effective doses reported in the literature. JNJ-38877605 and gefitinib were used at concentrations of 10 nM and 500 nM, respectively.

4.2. Immunofluorescence

Cells were plated on glass coverslips coated with 0.5% gelatin in PBS. After 24 h, cells were fixed in 4% formaldehyde for 15 min at room temperature and permeabilized in PBS, 1% BSA and 0.1% Triton X-100 for 15 min at room temperature. Subsequently, cells were stained with the MET mouse mAb DO24 [53] and rabbit mAb anti-PARP (Cat# 9532) purchased from Cell Signaling Technology (Denver, MA, USA). Primary antibodies were revealed by Alexa Fluor 555 and 488 conjugated secondary antibodies (Thermo Fisher Scientific; Waltham, MA, USA). Cell nuclei were stained with DAPI (Thermo Fisher Scientific; Waltham, MA, USA). Confocal analysis was performed on a Leica TCS SPE and processed using ImageJ.

4.3. Western Blot Analysis

For Western blot analysis, cells were plated and treated in complete medium after 24 h with the indicated drug or vehicle for the indicated doses and time points. Proteins were extracted in ice cold elution buffer (TrisHCl pH 7.4, containing EDTA, 1% Triton X-100, 10% glycerol), with protease inhibitor cocktail (1:1000), sodium orthovanadate (1:100) and phenylmethylsulfonyl fluoride (1:100), all purchased from Sigma-Aldrich (St. Louis, MI, USA), freshly added to lysis buffer. The extracted proteins were fractionated by SDS–PAGE and blotted onto nitrocellulose membrane using Trans-Blot Turbo Transfer System (BioRad, Hercules, CA, USA). The membranes were labeled overnight at 4 °C with the following antibodies: rabbit mAb anti-phosphoATM^Ser1981 (Cat# ab81292; Abcam, Cambridge, UK), and the following antibodies, all purchased from Cell Signaling Technology (Denver, MA, USA): rabbit mAb anti-phosphoMET^Tyr1234/1235 (Cat# 3077), rabbit mAb anti-phosphoEGFR^Tyr1068 (Cat#3777), rabbit mAb anti-phosphoCHK2^Thr68 (Cat#2661), rabbit mAb anti-phosphoCHK1^Ser345 (Cat#2341), rabbit mAb anti-phosphoNuMA^Ser395 (Cat#3429), rabbit mAb anti-NuMa (Cat#3888), rabbit mAb anti-ATM (Cat#2873), rabbit mAb anti-CHK2 (Cat#2662), mouse mAb anti-CHK1 (Cat#2360); mouse mAb anti-MET DL21 [53] and mouse mAb anti-EGFR (Cat# sc373746) purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Membranes were then incubated with HRP conjugated secondary antibodies, and the chemo-luminescent signals were reveled with ECL (Thermo Fisher Scientific; Waltham, MA, USA) using the ChemiDoc Touch Imaging System (BioRad, Hercules, CA, USA).

4.4. CellTiter-Glo^® Viability Assays

CellTiter-Glo^® assays were performed according to the manufacturer’s protocol (Promega Madison, WI, USA) and used to evaluate the viability of cell lines after 72 h long treatment with drugs. After incubation, the luminescent signal was measured using microplate reader BioTek Synergy HTX (Winooski, VT, USA). Results were analyzed with GR metrics, which normalize cell drug response on cell doubling, and data obtained by GR analyses were plotted using GraphPad Prism 7. All experiments were performed in triplicate, each being repeated at least three times, and all data are expressed as the mean ± standard deviation. T test was performed to compare between two experimental groups using GraphPad Prism 7 software. The synergistic effects were evaluated with the combination index (CI) using the Chou–Talalay method.

4.5. Crystal Violet Cytotoxicity Assay

Cells were treated with the indicated drugs for 5 days and, at the end of experiments, fixed with 2% paraformaldehyde (PAF) in PBS for 40 min. After washing, cells were stained with crystal violet (10% in 20% methanol) for 40 min. Crystal violet was dissolved, incubating with 10% acetic acid for 20 min on shaker. Crystal violet absorbance signal was measured at 595-nm with the microplate reader BioTek Synergy HTX. Results were analyzed with GR metrics, which normalize cell drug response on cell doubling, and data obtained by GR analyses were plotted using GraphPad Prism 7. All experiments were performed in triplicate, each being repeated at least three times, and all data are expressed as the mean ± standard deviation. T-test was performed to compare between two experimental groups using GraphPad Prism 7 software.