Abstract
Defects in maintaining genome integrity are a hallmark of cancer. The DNA damage response kinase ATM is frequently mutated in human cancer, but the significance of these events to chemotherapeutic efficacy has not been examined deeply in whole organism models. Here we demonstrate that bi-allelic Atm deletion in mouse models of Kras-mutant lung adenocarcinoma does not affect cisplatin responses. In marked contrast, Atm-deficient tumors displayed an enhanced response to the topoisomerase-II poison etoposide. Moreover, Atm-deficient cells and tumors were sensitive to the PARP inhibitor olaparib. This actionable molecular addiction to functional PARP1 signaling was preserved in models that were proficient or deficient in p53, resembling standard or high-risk genetic constellations, respectively. Atm deficiency also markedly enhanced sensitivity to the ATR inhibitor VE-822. Taken together, our results provide a functional rationale to profile human tumors for disabling ATM mutations, particularly given their impact on PARP1 and ATR inhibitors. Cancer Res; 77(11); 3040–56. ©2017 AACR.
Introduction
Lung cancer is the most common cancer entity after nonmelanocytic skin cancer, and lung cancer–related deaths surpass those from any other neoplastic disease worldwide (1). In 2012, lung cancer was the most frequently diagnosed cancer and the leading cause of cancer-related death in male populations. In female patients, lung cancer was the leading cause of cancer-related death in the developed countries, and the second leading cause of cancer-related death in developing countries (2). Through recent cancer genome sequencing efforts, we are beginning to understand the complex genomic aberrations that lead to the development of lung cancer (3). Particularly in lung adenocarcinoma, these efforts have led to the identification of numerous actionable genetic vulnerabilities, such as oncogenic EGFR mutations, as well as ALK and ROS1 rearrangements (4). However, the majority of lung adenocarcinomas appear to be driven by oncogenic alterations that are not amenable to direct therapeutic intervention (5). The most prominent example in this regard is oncogenically mutated KRAS, which occurs in approximately 32% of lung adenocarcinomas and for which no direct targeting approaches have been clinically developed, thus far (5). While KRAS is clearly a potent oncogenic driver, additional genomic aberrations may exist in KRAS-driven lung adenocarcinomas that could impact the therapeutic response. For instance, KRAS mutations were recently shown to cocluster with mutations affecting the p53 response in lung adenocarcinoma (6). These data suggest that an impaired DNA damage response maybe selected in KRAS-mutant lung adenocarcinoma. In line with this observation, it was recently shown that approximately 40% of human lung adenocarcinomas lack ATM protein expression (7).
The DNA damage response (DDR) constitutes a complex, kinase-based signaling network, which is activated in response to genotoxic stress (8). The DDR consists of two major branches, namely the ATR/CHK1 and the ATM/CHK2 pathways, in which the proximal kinases ATR and ATM directly phosphorylate and activate the effector kinases CHK1 and CHK2, respectively (8). Beyond these canonical kinase branches, additional signaling components, such as the p38MAPK/MK2 pathway, are recruited into the DDR network, depending on the type of genotoxic stress (9–12). The proximal DDR kinase ATM is a master regulator of the DDR and is involved in mediating cell-cycle arrest, DNA repair and apoptosis, following DNA damage (13–15). In keeping with the prominent role of ATM in cell cycle control, DNA repair, and genome maintenance, the ATM gene is recurrently inactivated through mutations and/or deletions in various cancer entities, ranging from hematologic malignancies to solid tumors, including lung adenocarcinoma (3, 5, 16–19). Specifically in chronic lymphocytic leukemia, it was shown that biallelic loss of ATM is associated with poor prognosis of the affected patients (20). This may be rationalized by the observation that ATM depletion impairs chemotherapy-induced p53 activation and subsequent induction of p53-dependent apoptosis (21). Beyond its role in regulating p53-mediated apoptosis, ATM is also involved in DNA double-strand break (DSB) repair, particularly through the homologous recombination (HR) pathway, with a less well defined role in the nonhomologous end joining pathway (15). ATM was specifically shown to mediate DSB resection and subsequent recruitment of the critical HR component RAD51 to the sites of damage (15). Recent reports further suggest that ATM is particularly involved in mediating DSB repair in heterochromatin regions of the genome through a slow-acting repair process (22, 23). It was shown that ATM directly phosphorylates the heterochromatin-building factor KAP-1, which in turn promotes HR-mediated DSB repair within heterochromatin regions (22, 23). In line with these observations, KAP-1 depletion was demonstrated to rescue the DSB repair defect induced by ATM deficiency (22, 23).
Intriguingly, both preclinical and clinical observations strongly suggest that a defective HR pathway is associated with distinct molecular liabilities. For instance, it was recently shown that BRCA1- or BRCA2-deficient cells and tumors, which display a manifest HR defect, are extraordinarily sensitive against PARP1 inhibition (24, 25). Furthermore, in vitro single-agent ATR inhibition was recently shown to induce selective toxicity in Chinese hamster cells that were HR-defective, due to ATM, BRCA2, or XRCC3 alterations, nucleotide excision repair–defective, due to ERCC2 mutations, or base excision repair–defective, due to XRCC1 defects (26). Specifically the selective toxicity of ATR inhibition in HR-defective cells is further corroborated by in vitro experiments that demonstrated that RAD51 depletion or inhibition led to a massively increased sensitivity against ATR or Chk1 inhibitors (27). However, the in vivo efficacy of ATR inhibitors has not been conclusively shown to date.
Here, we employ autochthonous mouse models of Kras-driven lung adenocarcinoma, as well as orthotopic transplantation models of genetically engineered lung adenocarcinoma cells in a syngeneic system to assess molecular vulnerabilities associated with Atm deficiency in lung adenocarcinoma. Our models span standard risk (KrasG12D) to high-risk (KrasG12D;Tp53−/−) genetic constellations. We show that loss of Atm is tolerated in Tp53-proficient backgrounds, while it appears to be counterselected in Tp53-deficient settings, in vivo. We further demonstrate that Atm deficiency is associated with markedly increased sensitivity against PARP1 and ATR inhibition, regardless of the Tp53 status. These data strongly suggest that even in lung adenocarcinomas displaying a high-risk mutational constellation, response to targeted small-molecule drugs is dictated by additional mutations affecting the DDR. Our data further indicate that the ATM status should routinely be assessed in lung adenocarcinomas, as defective ATM signaling is associated with an actionable dependence on PARP1 and ATR.
Materials and Methods
Autochthonous murine lung adenocarcinoma model
We employed the KrasLSL.G12D/wt (K) and KrasLSL.G12D/wt;Tp53fl/fl (KP) mouse model for Kras-driven lung adenocarcinoma, as described previously (28). In addition, we combined a conditional Atm allele (Atmfl) (29) with the K and KP models, to obtain KrasLSL.G12D/wt;Atmfl/fl (KA) and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl (KPA) mice, respectively. Mice were kept on a mixed C57Bl6/Sv129 background. To induce lung tumor formation, 8- to 12-week-old mice were anesthetized with Ketavet (100 mg/kg) and Rompun (20 mg/kg) by intraperitoneal injection followed by intratracheal instillation of replication-deficient adenovirus expressing Cre-recombinase (Adeno-Cre, 2.5 × 107 PFU). To confirm tumor formation, KP and KPA mice were scanned five weeks after Adeno-Cre application by μCT imaging (Aloka, Latheta LCT-100) under isoflurane (2.5%) anesthesia. K and KA mice were imaged 12 weeks after tumor induction in the same manner. All mouse experiments were conducted in accordance with an Institutional Animal Care and Use Committee (IACUC).
Allograft model
For the syngeneic allograft experiment, mice were anesthetized (2.5% isoflurane) and injected with 1.5 × 106 tumor cells into the right lung. Tumor formation was verified by μCT imaging one week after injection. Mice were subjected to four different treatment regimens and tumor volume changes were monitored by weekly μCT imaging for four weeks. Tumor volume was assessed by OsiriX and DICOM viewer software packages (OsiriX v.8.2, Pixmeo).
Compound solutions were prepared as follows: the PARP inhibitor olaparib (Axon Medchem, AZD2281) was dissolved in dimethyl sulfoxide to a final concentration of 50 mg/mL and then added to 10% 2-hydroxy-propyl-beta-cyclodextrin/PBS solution. Olaparib was administered daily at a dose of 50 mg/kg i.p. The ATR inhibitor VE-822 (Abmole, M3115) was dissolved in an equilibrated (v/v) mixture PBS (70%), polyethylene glycol 300 (PEG300, 29.5%) and Tween 80 (0.5%) at a concentration of 18 mg/mL. VE-822 was administered weekly for 3 consecutive days at a dose of 30 mg/kg by oral gavage. Etoposide (Hexal) was administered two times for 4 consecutive days, intraperitoneally, at a concentration of 10 mg/kg. Cisplatin (Accord) was administered once a week for 3 consecutive weeks at a concentration of 7.5 mg/kg, i.p.
Histologic analysis and tumor volume quantification
Mice harboring tumors of all four genoytpes were sacrificed 4, 8, and 12 weeks after Adeno-Cre inhalation and lungs were fixed in 4% PFA. Formalin-fixed paraffin-embedded (FFPE) murine lung samples were cut into 4 μm–thick sections and mounted on slides. After staining with haematoxylin and eosin (H&E), the tumors were assessed by a board-certified pathologist to evaluate the relative tumor burden. A tumor grading system with two tumor grades (I and II) based on the relative amount of diffuse or nodular tumor growth was applied, exactly as described previously (30). Grade I (“diffuse”) tumors were defined by a predominantly diffuse tumor architecture (>90% diffuse). Grade II (“nodular”) tumors predominantly show a nodular growth pattern (>90% nodular). In addition, FFPE lungs were stained for the proliferation marker Ki-67 (Cell Signaling Technology, #9661).
Tumor isolation and cell culture
Individual murine lung adenocarcinoma tumor nodules from KP and KPA mice were isolated and cultured in RPMI medium containing 10% FBS and 1% penicillin/streptomycin. The retained LoxP-flanked Atm allele in KPAfl/Δ cells was recombined by treatment with 2.5 × 107 PFU of Adeno-Cre and individual Atm-deficient clones (KPAΔ/Δ) were generated by clone picking.
Results
KRAS-mutant lung adenocarcinomas with cooccurring TP53 mutations constitute a high-risk collective
KRAS is one of the most frequently mutated oncogenes in human cancer and is altered in 32% of human lung adenocarcinomas. KRAS-mutant lung adenocarcinomas remain a clinically challenging subentity, as no effective targeted single-agent or combination regimens are available in the clinical setting. In fact, current first-line regimens for the treatment of stage IV KRAS-mutant lung adenocarcinomas consist of platinum-based combination chemotherapy together with a VEGF-blocking antibody (4). Only very recently, immune checkpoint–blocking antibodies are entering the first-line setting in selected patient populations. To approach these difficult to treat KRAS-mutant tumors, we initially analyzed publicly available cancer genome sequencing data to ask whether we could identify cooccurring genetic aberrations (7). Consistent with previously published data (3), we found that protein-damaging TP53 mutations, MDM2 amplifications, and CDKN2A alterations are frequently detected in KRAS-mutant lung adenocarcinomas (Fig. 1A). Moreover, patients with KRAS mutations and cooccurring alterations in TP53, MDM2, or CDKN2A displayed a reduced overall survival, compared with patients carrying KRAS-mutant tumors, in which the integrity of TP53, MDM2, or CDKN2A was preserved (Fig. 1B). These data are mimicked in a well-established murine model of Kras-driven lung adenocarcinoma. In these mice, expression of oncogenic KrasG12D is prevented through the insertion of a LoxP-flanked transcriptional and translational STOP cassette in the endogenous locus (KrasLSL.G12D allele; refs. 6, 28). Intratracheal administration of adenoviral Cre recombinase (Adeno-Cre) leads to the expression of oncogenic KrasG12D from its endogenous locus. In addition to this simple Kras-driven model, we also used KrasLSL.G12D/wt;Tp53fl/fl mice, in which both Tp53 alleles are flanked by LoxP sites (6, 28, 31). In these compound-mutant mice, Adeno-Cre drives expression of KrasG12D and simultaneous deletion of both Tp53 alleles. As shown in Fig. 1C, and consistent with previously published data (28), biallelic deletion of Tp53 in Kras-driven lung adenocarcinomas leads to a significantly reduced overall survival, compared with p53-proficient settings.
Combined genomic alterations in KRAS and the TP53 pathway are associated with adverse risk in lung adenocarcinoma patients. A, Pie chart based on 146 patients analyzed for the occurrence of mutations in KRAS and/or the TP53 pathway. TP53 mutations, MDM2 amplifications, or CDKN2A deletions were considered as p53 pathway mutations for this analysis. B, Lung adenocarcinomas with combined mutations in KRAS and the TP53 signaling pathway are associated with poor prognosis. Survival of patients that with KRAS mutations (n = 58), but lacking TP53 pathway mutations, was compared to the survival of patients with KRAS mutations and TP53 pathway mutations (n = 88) by log-rank (Mantel–Cox) test. C, KrasLSL.G12D/wt;Tp53fl/fl (KP) animals (n = 40) succumb to their disease significantly earlier than KrasLSL.G12D/wt (K) animals (n = 44). Mice from both cohorts were subjected to intratrachael instillation of Adeno-Cre with 8–12 weeks of age. Survival of the two groups was compared by log-rank (Mantel–Cox) test.
A recent report indicates that ATM is inactivated in approximately 40% of human lung adenocarcinomas (7). Furthermore, it was recently shown that, albeit rarely, inactivation of ATM can also occur in TP53-deficient human tumors (21). This observation is somewhat surprising, as the proximal DNA damage response kinase ATM directly phosphorylates and activates p53 in a linear pathway. Thus, these observations suggest that cancer cells might derive additional benefit from inactivating the ATM/p53 pathway both, on the level of ATM and p53. Given that ATM is a master regulator of the cellular DNA damage response that is not only involved in activating p53, but also regulates DNA repair and kinase-driven cell-cycle checkpoint networks, we speculated that KRAS-driven lung adenocarcinomas with combined ATM and TP53 mutations might display actionable molecular liabilities, despite the fact that simultaneous KRAS- and TP53 alterations constitute a high-risk scenario, per se.
Atm alterations lead to reduced overall survival in an autochthonous KrasLSL.G12D/wt;Tp53fl/fl mouse model of lung adenocarcinoma
To directly investigate the phenotype associated with Atm deficiency in Kras-driven tumors in vivo, we employed a conditional LoxP-flanked Atm allele (29), which was homozygously crossed into our KrasLSL.G12D/wt and KrasLSL.G12D/wt;Tp53fl/fl models, yielding KrasLSL.G12D/wt;Atmfl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice, respectively (Fig. 2A; Supplementary Fig. S1A).
Atm alterations in Tp53-deficient Kras-driven murine lung adenocarcinomas leads to reduced survival and a more aggressive tumor phenotype. A, Schematic representation of the three alleles used in the mouse model KrasLSL.G12D, Tp53fl, and Atmfl. Two experimental groups were generated, KrasLSL.G12D/wt;Tp53fl/fl (KP) and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl (KPA) mice. B, Atm alterations significantly reduce the overall survival of lung adenocarcinoma-bearing KP mice, in vivo. Kaplan–Meier curves illustrate the overall survival of KP (n = 20) and KPA (n = 16) mice 160 days following Adeno-Cre application. Log-rank test: P < 0.0001. C, Representative histology, H&E stainings of lungs isolated from mice 4, 8, and 12 weeks after tumor induction. KP and KPA mice display progressively growing lung adenocarcinomas following intratracheal Adeno-Cre application. Histologic examination revealed that KPA animals display more aggressively growing lung adenocarcinomas than KP animals. D, Quantification of H&E-stained lung tissue samples indicating tumor-free mass, tumor grade I, and tumor grade II, according to the grading system detailed in Materials and Methods. E, No differences in the number of tumor lesions were observed between KP and KPA mice. Lungs were isolated 4, 8, and 12 weeks after tumor induction (n = 2). F, Atm is still expressed in lungs of KPA mice isolated 8 weeks after tumor induction. Atm in situ hybridization on lung sections (KP and KPA) isolated 8 weeks after tumor induction. Tumor-free lungs isolated from an Atm−/− mouse were used as a negative control.
The biological effects of Atm deletion were primarily assessed through recording of overall survival of KrasLSL.G12D/wt, KrasLSL.G12D/wt;Tp53fl/fl, KrasLSL.G12D/wt;Atmfl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice that were intratracheally instilled with Adeno-Cre (2.5 × 107 PFU) at 8–12 weeks of age. As shown in Supplementary Fig. S1A, the survival of KrasLSL.G12D/wt (n = 17) and KrasLSL.G12D/wt;Atmfl/fl animals (n = 9) was not statistically different (P = 0.1596). Furthermore, neither the number of individual tumor nodules per lung, nor the relative tumor volume, related to normal lung volume, differed significantly between KrasLSL.G12D/wt and KrasLSL.G12D/wt;Atmfl/fl animals at 4, 8, and 12 weeks after Adeno-Cre–mediated recombination (Supplementary Fig. S1C and S1D). Of note, efficient deletion of Atm was verified in this model, using a RNA in situ hybridization approach (Supplementary Fig. S1B). We specifically chose RNA in situ hybridization for the assessment of Atm deletion, as no specific antibody against murine Atm is available for IHC.
In marked contrast, KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals (n = 16) displayed a significantly (P < 0.0001) reduced overall survival, compared with their Atm-proficient KrasLSL.G12D/wt;Tp53fl/fl counterparts (n = 20; Fig. 2B). Moreover, histologic examination of lungs isolated from KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals isolated 8 weeks after intratracheal Adeno-Cre instillation, displayed a significantly (P < 0.0000001) larger relative tumor volume per lung, compared with KrasLSL.G12D/wt;Tp53fl/fl controls (60% ± 0% vs. 10% ± 0% at 8 weeks; Fig. 2C and D). A similar trend was observed in lungs isolated 4 weeks after Adeno-Cre instillation (7.5% ± 2.5% vs. 5% ± 2.5%). However, this trend failed to reach statistical significance (Fig. 2C and D). We did not observe a substantial difference in tumor grade in KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals 12 weeks after intratracheal Adeno-Cre instillation, as might have been expected given the massively reduced overall survival of KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice compared with KrasLSL.G12D/wt;Tp53fl/fl controls (Fig. 2B). As observed in the Tp53-proficient setting (Supplementary Fig. S1E), the number of individual tumor nodules per lung did not significantly (P = 0.1723) differ between KrasLSL.G12D/wt;Tp53fl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals (Fig. 2E).
To further dissect these observations and to directly assess the efficacy of Atm deletion in tumors, we next performed RNA in situ hybridization of lungs isolated from tumor-bearing KrasLSL.G12D/wt, KrasLSL.G12D/wt;Tp53fl/fl, KrasLSL.G12D/wt;Atmfl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice 12 weeks after Adeno-Cre application. Surprisingly, we observed incomplete Atm deletion in KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals (Fig. 2F; Supplementary Fig. S1B). Of note, lungs isolated from Atm−/− mice (32) stained entirely negative for Atm (Fig. 2F, right). Together, these data from KrasLSL.G12D/wt, KrasLSL.G12D/wt;Tp53fl/fl, KrasLSL.G12D/wt;Atmfl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice suggest that biallelic Atm deletion is selected against in p53-deficient settings, while it is tolerated in the context of a functional p53 response. Furthermore, the significantly reduced overall survival observed in KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl mice, compared with KrasLSL.G12D/wt;Tp53fl/f animals (Fig. 2B) suggests that partial deletion of Atm in a Tp53-deficient background might enhance tumorigenesis.
Homozygous Atm deletions can be obtained in Tp53-deficient cells in vitro
As combined homozygous deletions of both Tp53 and Atm could not routinely be obtained in KrasG12D-driven lung adenocarcinomas in vivo (Fig. 2F), we next isolated individual cell lines from tumor-bearing KrasLSL.G12D/wt;Tp53fl/fl and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl animals, to assess their Tp53 and Atm status by genotyping PCR and immunoblot analysis (Fig. 3A–C). We found that KrasLSL.G12D/wt;Tp53fl/fl- and KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl-derived tumor cell lines (n = 30 independent clones for each genotype) uniformly displayed a homozygous Tp53 deletion and a concomitant lack of p53 protein expression (Fig. 3A–C). A different picture emerged, when we analyzed the recombination status of the conditional Atm allele in cell lines derived from KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl tumors (n = 30 independent clones). Only 1 of 30 independent clones displayed biallelic Atm deletion (clone 3; Fig. 3A). Consistent with these genotyping results, ATM protein expression was preserved in 29 of 30 independent clones (only clone c3 lacked ATM protein expression; Fig. 3C and D). Resequencing of the KrasLSL.G12D/wt;Tp53fl/fl;Atmfl/fl–derived tumor cell lines revealed that the LoxP sites flanking the nonrecombined Atm alleles were intact in all 29 non-recombined clones. These data indicate that acute combined biallelic loss of Tp53 and Atm might be counterselected in Kras-driven tumors. Furthermore, the uniform deletion of both Tp53 alleles might suggest that Tp53 deficiency provides the incipient Kras-mutant cancer cell with traits that are advantageous, compared with biallelic Atm deficiency. We next asked whether we could achieve biallelic Atm deletion in KrasG12D/wt;Tp53Δ/Δ;Atmfl/Δ cell lines. For this purpose, we reexposed these cell lines to Adeno-Cre (2.5 × 107 PFU/mL of culture media) and assessed the recombination status in individual clones isolated following this in vitro Cre-application. As shown in Fig. 3B and C, this in vitro Adeno-Cre exposure resulted in efficient deletion of the remaining LoxP-flanked Atm allele, and subsequent loss of ATM protein expression in all 9 KrasG12D/wt;Tp53Δ/Δ;Atmfl/Δ parental cell lines. We note that these in vitro–derived KrasG12D/wt;Tp53Δ/Δ;AtmΔ/Δ clones displayed proliferation kinetics that were not significantly different from those observed in KrasG12D/wt;Tp53Δ/Δ- and KrasG12D/wt;Tp53Δ/Δ;Atmfl/Δ clones (Supplementary Fig. S2). Thus, in summary, biallelic Tp53 and Atm deletion can be achieved in vitro, but appears to be counterselected in vivo.
Deletion of both Atm alleles is tolerated in KrasLSL.G12D/wt; Tp53fl/fl;Atmfl/fl lung adenocarcinoma cells, in vitro. A, Schematic representation of the targeted alleles and the primer-binding sites. B and C, Atm is heterozygously deleted in the majority of KPA cell lines and homozygously deleted following in vitro exposure to Adeno-Cre. B, KP and KPA cell lines were genotyped to detect Cre-mediated recombination in the Kras, Tp53, and Atm loci. All cell lines show efficient recombination of Kras (monoallelic) and Tp53 (biallelic), while only one out of eight KPA cell lines showed biallelic Atm deletion. Following in vitro administration of Adeno-Cre, biallelic Atm deletion was readily detected in all KPA cell lines. Genotyping was performed by PCR. Displayed bands for Kras: wild-type, 622 bp; LoxP-flanked allele, 500 bp; recombined allele, 650 bp. Displayed bands for Tp53: wild-type, 288 bp: LoxP-flanked allele, 370 bp; recombined allele, 612 bp; unspecific band, 400 bp. Displayed bands for Atm: wild-type, 328 bp; LoxP-flanked allele, 420 bp; recombined allele, 903 bp. c1 = tail DNA from KrasLSL.G12D/wt;Tp53fl/fl; Atmfl/fl mouse. c2, tail DNA from wild-type mouse. C, Immunoblot to detect ATM and TP53 in lysates derived from KP-, KPA-, and KPA cells that were exposed to Adeno-Cre in vitro. All cell lines displayed complete loss of TP53 in all samples. ATM was expressed in seven out of eight KPA cell lines and was not expressed following in vitro administration of Adeno-Cre. β-Actin was used as a loading control. D, Schematic representation of the ATM status of all cell lines generated from KP (n = 30) and KPA tumors (n = 30) and of KPA cell lines, which were treated with Adeno-Cre in vitro (n = 9).
Atm-deficient KrasG12D/wt;Tp53Δ/Δ lung adenocarcinomas display increased sensitivity against the topoisomerase-II inhibitor etoposide
We next asked whether KrasG12D/wt;Tp53Δ/Δ- (referred to as KP hereafter), KrasG12D/wt;Tp53Δ/Δ;Atmfl/Δ- (referred to as KPAfl/Δ hereafter), and KrasG12D/wt;Tp53Δ/Δ;AtmΔ/Δ (referred to as KPAΔ/Δ hereafter) cells displayed differential chemotherapy sensitivity in vitro (Fig. 4; Supplementary Fig. S3). For this purpose, we employed flow cytometry–based apoptosis measurements and clonogenic colony survival assays. KP-, KPAfl/Δ-, and KPAΔ/Δ cells were exposed to cisplatin (1 μmol/L), etoposide (0.1 μmol/L), or vehicle control, harvested following 24 hours of drug exposure and subsequently stained, using Annexin V and PI labeling. As shown in Fig. 4A, cisplatin readily induced apoptotic cell death in KP-, KPAfl/Δ-, and KPAΔ/Δ cells (21.88% ± 0.828%, 17.34% ± 1.57%, and 18.25% ± 2.54%, respectively). However, we did not detect a statistically significant difference in the cisplatin sensitivity of the three different genotypes (Fig. 4A). To validate the results obtained with these flow cytometry experiments, we next performed clonogenic survival assays, using KP-, KPAfl/Δ-, and KPAΔ/Δ cells. In brief, cells were pretreated with cisplatin (1 μmol/L, 72 hours), trypsinized, and seeded onto new cell culture dishes (5,000 cells per 6-well plate). Fourteen days following seeding, cells were stained with crystal violet and surviving colonies were counted (Fig. 4B and C). Reminiscent of the results obtained in our flow cytometry–based apoptosis measurements, we observed a massive and statistically significant reduction of viable colonies in all three genotypes [P < 0.0001 (KP), P < 0.0001 (KPAfl/Δ) and P < 0.0001 (KPAΔ/Δ); Fig. 4B and C]. However, this reduction in surviving colonies did not differ between KP-, KPAfl/Δ-, and KPAΔ/Δ cells, further validating our flow cytometry results (Fig. 4B and C). To verify the induction of genotoxic damage induced by cisplatin, we performed a longitudinal assessment of nuclear γH2AX foci, using indirect immunofluorescence (Fig. 4D and E; Supplementary Fig. S3A). These experiments revealed that cisplatin inflicted γH2AX-positive DNA lesions in all three genotypes. The kinetics of DNA damage induction were similar in KP-, KPAfl/Δ-, and KPAΔ/Δ cells, with a mild induction after 3 hours of drug exposure and a peak at 24 hours. This genotoxic damage remained stable for 72 hours in all three cell types. We did not detect a statistically significant difference in the kinetics and intensity of γH2AX staining between the different genotypes (Fig. 4D and E; Supplementary Fig. S3A).
Atm-defective lung adenocarcinoma cell lines display hypersensitivity to DNA double-strand break-inducing agents. A, KP-, KPAfl/Δ-, KPAΔ/Δ cell lines display a similar degree of apoptotic cell death in response to cisplatin (1 μmol/L; red). Blue, vehicle-treated cells. Cells were treated for 72 hours and apoptosis was assessed by quantification of the Annexin-V/PI double-positive population using flow cytometry. Average values of three independent experiments are shown. B and C, KP-, KPAfl/Δ-, and KPAΔ/Δ cells show reduced clonogenic survival following cisplatin treatment. B, KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines were exposed to 1 μmol/L cisplatin or vehicle control for 72 hours. Crystal violet stainings were performed after 10 days in culture. C, Quantification of crystal violet–positive cells. Scale bar, 500 μm. D and E, Cisplatin treatment (10 μm, 30 minutes) induces genotoxic damage KP-, KPAfl/Δ-, and KPAΔ/Δ cells. D, Cells were stained with an antibody detecting γH2AX, as well as a DAPI counterstain 3, 24, 48, and 72 hours after treatment. Representative immunofluorescence images are shown. Scale bar, 20 μm. E, The percentage of γH2AX positive cells from the experiment detailed in D is quantified. F, KPAΔ/Δ cells display an increased apoptotic response to etoposide treatment (red), compared with KP- and KPAfl/Δ cells. No substantial apoptotic response was observed in the vehicle-treated controls (blue). Cells were treated with 0.1 μmol/L etoposide or vehicle control for 72 hours and the apoptotic fraction was quantified as the Annexin V/PI double-positive population using flow cytometry. Average values of three independent experiments are shown. G and H, KPAΔ/Δ cells display a reduced clonogenic survival after etoposide treatment, compared with KP- and KPAfl/Δ cells. G, KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines were exposed to 0.1 μmol/L etoposide or vehicle control for 72 hours. Crystal violet stainings were performed after 10 days in culture. H, Quantification of crystal violet–positive cells. Scale bar, 500 μm. I and J, Etoposide treatment induces genotoxic damage in KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines, which is not effectively resolved in KPAΔ/Δ cells. I, KP-, KPAfl/Δ-, and KPAΔ/Δ cells were either treated with a vehicle control or with etoposide (10 μmol/L, 30 minutes). Cells were stained for γH2AX and counterstained with DAPI 3, 24, 48, and 72 hours after treatment. Representative immunofluorescence images are shown. Scale bar, 20 μm. J, The percentage of γH2AX-positive cells from the experiment detailed in I is quantified.
A strikingly different picture emerged, when we exposed KP-, KPAfl/Δ-, and KPAΔ/Δ cells to the topoisomerase-II inhibitor etoposide (0.1 μmol/L, 72 hours). Using both flow cytometry–based apoptosis measurements (Fig. 4F), as well as clonogenic survival assays (Fig. 4G), we found that KP- and KPAfl/Δ cells were essentially resistant against etoposide. In marked contrast, etoposide treatment robustly induced substantial levels of apoptosis in KPAΔ/Δ cells (Fig. 4F). This difference in apoptosis induction was statistically significant between KP- and KPAΔ/Δ cells (P = 0.0385; Fig. 4F). Similarly, etoposide treatment led to a significant reduction of colony survival only in KPAΔ/Δ cells (P < 0.0001), but not in KP- or KPAfl/Δ cells (Fig. 4G). Thus, in summary, biallelic Atm deletion leads to enhanced etoposide sensitivity in KP cells, but does not affect cisplatin sensitivity. We further assessed the severity of genotoxic stress inflicted by etoposide in KP-, KPAfl/Δ-, and KPAΔ/Δ cells, using immunofluorescence-based longitudinal quantification of nuclear γH2AX foci (Fig. 4I and J; Supplementary Fig. S3B). In all three cell types, we observed a similar peak in genotoxic damage at 24 hours of drug exposure. KP cells displayed efficient clearance of γH2AX-positive DNA lesions at 48 and 72 hours following drug treatment. KPAfl/Δ cells displayed clearance kinetics that were significantly delayed, compared with that observed in KP cells. In contrast, KPAΔ/Δ cells completely failed to remove γH2AX-positive DNA lesions at 48 and 72 hours. Thus, our cell line panel essentially constitutes an allelic series, where dependent on the Atm gene dosage, etoposide-induced DNA damage is repaired (Fig. 4I and J; Supplementary Fig. S3B).
To further substantiate these in vitro observations, we next performed orthotopic transplantation experiments to assess chemotherapy sensitivity of KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors, in vivo. For this purpose, we induced tumors in the lungs of syngeneic-recipient animals through intrathoracal injection of 1.5 × 106KP-, KPAfl/Δ-, or KPAΔ/Δ cells, respectively. Tumor onset and therapy response was longitudinally monitored through μCT imaging. Once CT-morphologically visible tumors had formed, animals received either cisplatin (7.5 mg/kg, intraperitoneally, days 1, 8, and 15), etoposide (10 mg/kg, intraperitoneally, days 1–5 and 15–19), or vehicle control (Supplementary Fig. S4). As shown in Supplementary Fig. S5A–S5C and corroborating our in vitro data, cisplatin treatment led to a reduction in tumor volume, which did not significantly differ between KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors. We note that vehicle-treated KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors displayed continuous growth throughout the entire observation period of 21 days (Supplementary Fig. S5B and S5C). The cisplatin-induced reduction in tumor volume did not translate into a significant survival gain in any of the three different tumor genotypes, although we detected a trend toward enhanced overall survival in all three genotypes, following cisplatin treatment (Supplementary Fig. S5D). The apparent resistance of the KP model against cisplatin-based chemotherapy is in line with observations that we have previously published (30). We next employed IHC to assess the proliferation index (Ki-67–positive tumor cell fraction) in KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors, following cisplatin treatment (Supplementary Figs. S5E and S5F and S6). In line with our previous results, we detected a significantly reduced Ki-67 index in all three tumor genotypes, following cisplatin treatment (Supplementary Fig. S5E and S5F and S6). Thus, overall, there appears to be no differential cisplatin sensitivity between KP-, KPAfl/Δ-, and KPAΔ/Δ cells in vitro and tumors in vivo.
In contrast to the cisplatin response, our orthotopic transplantation model revealed statistically significant differences in the etoposide response of KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors (Supplementary Fig. S7). While KP-derived tumors were entirely resistant against etoposide and even displayed mild tumor volume gains (1.476 ± 0.237 fold change in mean tumor volume) one week following application of two cycles of etoposide (10 mg/kg, intraperitoneally, days 1–5 and 15–19), KPAfl/Δ-derived tumors shrank in response to etoposide (0.876 ± 0.032 fold change in mean tumor volume). However, this effect did not reach statistical significance (P = 0.1272) at day 7 (Supplementary Fig. S7C) and only reached significance at the day 14 and 21 μCT scans (Supplementary Fig. S7B). In contrast to KP- and KPAfl/Δ-derived tumors, KPAΔ/Δ tumors displayed a robust and significant (P = 0.0053) shrinkage following etoposide treatment (0.620 ± 0.103 fold change in mean tumor volume). This difference was manifest on day 7 (Supplementary Fig. S7C), following the first etoposide dose and remained stable on days 14 and 21 following etoposide (Supplementary Fig. S7B). We note that vehicle-treated KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors displayed continuous growth throughout the entire observation period of 21 days (Supplementary Fig. S7B). In line with these CT-morphologic tumor volume–based response data, we also observed a significantly enhanced overall survival of KPAfl/Δ- and KPAΔ/Δ- tumor-bearing mice, compared with KP tumor–bearing animals in response to two cycles of etoposide treatment (Supplementary Fig. S7D). The CT-morphologic response data, as well as the overall survival data, are also reflected by results obtained from IHC analysis of tumor sections from animals that were sacrificed after the first cycle of etoposide treatment (Supplementary Figs. S7E and S7F and S6). While KP-derived tumors showed only a mild reduction in Ki-67 positivity following etoposide, KPAfl/Δ- and, even more pronounced, KPAΔ/Δ-derived tumors displayed a massive reduction in Ki-67 positivity (Supplementary Figs. S7E and S7F and S6). Overall, these data suggest that heterozygous and particularly homozygous Atm deletion in KP-derived tumors is associated with a significantly enhanced etoposide response. These data indicate that the clinical annotation of the ATM status might be a useful tool to predict the etoposide response in high-risk KRASmut and TP53mut human tumors.
Atm-deficient KrasG12D/wt;Tp53fl/fl lung adenocarcinoma cells display actionable molecular dependencies on PARP1 and ATR
ATM is a master regulator of the cellular DNA damage response, which phosphorylates a variety of proteins involved in cell-cycle arrest, DNA repair, and programmed cell death (8, 13, 14). Particularly the role of ATM in DNA DSB repair might be therapeutically relevant: human and murine cells employ two dominant pathways for DSB repair, namely HR, which partially depends on ATM (15) and nonhomologous end joining (15, 33–35). One of the first steps of HR-mediated DSB repair is the resection of the DSB leading to the generation of a single-stranded (ss) 3′-overhang, which is immediately coated by the single-strand–binding protein RPA (15, 36–39). RPA is ultimately replaced by RAD51 in an ATM/CHK2/BRCA1/BRCA2/PALB2-dependent fashion (15, 36, 37, 39, 40). Rad51 is a critical component of the HR process that mediates homology search, strand exchange, and Holliday junction formation (15). Intriguingly, mutationally encoded defects in the cellular HR machinery are frequently detected in various cancer entities. Particularly, the HR defect in BRCA1- or BRCA2-deficient settings has previously been shown to be associated with an actionable dependence on PARP1 (24, 25, 41). As ATM is also involved in the HR process and given the PARP1 dependence of BRCA1- or BRCA2-deficient cells and tumors, we speculated that our Atm-deficient KP tumors might display an actionable PARP1 dependence. To directly address this question, we assessed the sensitivity of KP-, KPAfl/Δ-, and KPAΔ/Δ cells against the PARP1 inhibitor olaparib. We initially employed flow cytometry–based apoptosis measurements to directly investigate whether olaparib (3 μmol/L, 120 hours) displayed differential cytotoxicity in KP-, KPAfl/Δ-, and KPAΔ/Δ cells. As shown in Fig. 5A, olaparib induced only a marginal increase in the fraction of apoptotic KP cells, compared with vehicle-treated controls (6.188% ± 0.855% vs. 19.770% ± 3.168%). In contrast, olaparib robustly induced apoptosis in KPAfl/Δ cells (4.431% ± 0.565% vs. 39.750% ± 4.909%), which was even more pronounced in KPAΔ/Δ cells (6.585% ± 0.750% vs. 62.780% ± 7.999; Fig. 5A). We next aimed to validate these observations using clonogenic survival assays. Reminiscent of our flow cytometry results, olaparib induced a mild reduction in surviving KP colonies (60.980% ± 3.799%; Fig. 5B and C). The same effect was observed in KPAfl/Δcells (65.480% ± 2.368%), while KPAΔ/Δ cells (11.810% ± 2.288%) showed a striking decrease in surviving colonies (Fig. 5B and C). Thus, our data suggest that heterozygous, and more importantly, homozygous Atm deficiency is associated with an actionable PARP1 dependence in KP lung adenocarcinoma cell lines.
Combined loss of Atm and Tp53 in Kras-driven lung adenocarcinoma cell lines constitutes a genetic constellation that is associated with sensitivity to ATR and PARP inhibition. A, Olaparib induces apoptosis in KPAfl/Δ cells, which was even more pronounced in KPAΔ/Δ cells. In contrast, olaparib did not induce a substantial degree of apoptosis in KP cells. Cells were treated with 3 μmol/L olaparib (red) or vehicle control (blue) for 120 hours and the apoptotic fraction was quantified as the Annexin V/PI double-positive population using flow cytometry. Average values of three independent experiments are shown. B and C, Olaparib induces a mild reduction in clonogenic survival of KP cells. This effect was enhanced in KPAfl/Δ cells and even more striking in KPAΔ/Δ cells. B,KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines were exposed to 3 μmol/L olaparib or vehicle control for 120 hours. Crystal violet stainings were performed after 10 days in culture. C, Quantification of crystal violet–positive cells. Scale bar, 500 μm. D, KPAΔ/Δ cells display an increased apoptotic response to VE-822 treatment (red), while KP- and KPAfl/Δ cells do not induce substantial apoptosis in response to VE-822 treatment. Blue, vehicle-treated cells. Cells were treated with 0.1 μmol/L VE-822 or vehicle control for 72 hours and apoptosis was quantified as the Annexin-V/PI double-positive population using flow cytometry. Average values of three independent experiments are shown. E and F, KPAΔ/Δ cells display reduced clonogenic after VE-822 treatment, which cannot be observed in KP- and KPAfl/Δ cells. E, KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines were exposed to 0.1 μmol/L VE-822 or vehicle control for 72 hours. Crystal violet stainings were performed after 10 days in culture. F, Quantification of crystal violet–positive cells. Scale bar, 500 μm. G and H, Olaparib treatment induces genotoxic damage, which is not effectively resolved in KPAΔ/Δ cells. KPAfl/Δ cells are partially competent to resolve olaparib-induced genotoxic damage. G, KP-, KPAfl/Δ-, and KPAΔ/Δ cells were either treated with a vehicle control or with olaparib (3 μmol/L, 72 hours). Cells were stained for γH2AX and counterstained with DAPI 3, 24, 48, and 72 hours after treatment initiation. Representative immunofluorescence images are shown. Scale bar, 20 μm. H, The percentage of γH2AX-positive cells from the experiment detailed in G is quantified. I and J, VE-822 treatment induces genotoxic damage in KP-, KPAfl/Δ-, and KPAΔ/Δ cell lines, which cannot be effectively resolved in KPAΔ/Δ cells. I, KP-, KPAfl/Δ-, and KPAΔ/Δ cells were either treated with vehicle control or with VE-822 (0.1 μmol/L, 72 hours). Cells were stained for γH2AX and counterstained with DAPI 3, 24, 48, and 72 hours after treatment initiation. Representative immunofluorescence images are shown. Scale bar, 20 μm. J, The percentage of γH2AX-positive cells from the experiment detailed in I is quantified.
In addition to profiling the effects of the PARP1 inhibitor olaparib, we next aimed to assess the efficacy of the ATR inhibitor VE-822 (0.1 μmol/L) on KP-, KPAfl/Δ-, and KPAΔ/Δ cells. These experiments were motivated by recent reports suggesting that ATM depletion or loss might be associated with ATR inhibitor sensitivity in chronic lymphocytic leukemia (42). In addition, the ATR inhibitor AZD6738 was recently shown to potently synergize with cisplatin in ATM-deficient non–small cell lung cancer cells (43). Furthermore, combined cisplatin and AZD6738 treatment was shown to induce robust shrinkage of xenograft tumors derived from KRAS-mutant and ATM-defective human H23 lung adenocarcinoma cells (43). To further interrogate the effects of ATR inhibition on Atm-defective lung adenocarcinoma cells, we performed flow cytometry–based apoptosis measurements in KP-, KPAfl/Δ-, and KPAΔ/Δ cells. As shown in Fig. 5D, VE-822 treatment (0.1 μmol/L, 72 hours) induced only a mild increase in the percentage of apoptotic cells in KP- and KPAfl/Δ cells, compared with vehicle-treated controls (5.548% ± 0.694% vs. 10.170% ± 0.901% and 5.896% ± 0.522% vs. 9.752 ± 0.992%, respectively). In contrast, KPAΔ/Δ cells displayed markedly increased levels of apoptosis in response to VE-822 exposure (18.440% ± 1.429%), compared with KP- and KPAfl/Δ cells. Vehicle treatment did not induce substantial levels of apoptosis in KPAΔ/Δ cells (5.978% ± 0.776%). These flow cytometry–based apoptosis measurements are further supported by the results of clonogenic survival assays that we performed in KP-, KPAfl/Δ-, and KPAΔ/Δ cells following ATR inhibition (Fig. 5E and F). Similar to the results shown in Fig. 5D, VE-822 treatment of KP- and KPAfl/Δ cells led to a mild, but statistically significant, reduction in the number of surviving colonies, compared with vehicle-treated cells (P < 0.0001 and P < 0.001, respectively; Fig. 5E and F). Furthermore, VE-822 exposure led to a substantial and significant reduction in surviving KPAΔ/Δ colonies, compared with vehicle controls (P < 0.0001; Fig. 5E and F). Of note, the cytotoxic effect of VE-822 was significantly more pronounced in KPAΔ/Δ cells, compared with KP- and KPAfl/Δ cells (P < 0.0001 and P < 0.0001, respectively; Fig. 5D). These differential effects of olaparib (3 μmol/L, 120 hours) and VE-822 (0.1 μmol/L, 72 hours) on KP-, KPAfl/Δ-, and KPAΔ/Δ cells were also mirrored in longitudinal immunofluorescence-based γH2AX measurements. As shown in Fig. 5G and H and Supplementary Fig. S8A, olaparib induced γH2AX-decorated genotoxic lesions in all three tumor cell genotypes that peaked at 24 hours. While KP- and KPAfl/Δ-derived cells displayed substantial clearance of these lesions at 72 hours following drug exposure, KPAΔ/Δ cells failed to remove olaparib-induced DNA damage at 72 hours. Similarly, VE-822 was genotoxic in all three genotypes and induced maximal damage at 24 hours following drug exposure (Fig. 5I and J; Supplementary Fig. S8B). These lesions were readily removed by KP- and KPAfl/Δ cells 72 hours following drug exposure. In marked contrast, KPAΔ/Δ cells did not display a substantial removal of genotoxic lesions even 72 hours following VE-822 treatment. Thus, in summary, our data indicate that homozygous Atm deletion in KP murine lung adenocarcinoma cell lines is associated with two distinct molecular vulnerabilities, namely an actionable dependence on PARP1 (Fig. 5A–C, G, and H), as well as an actionable ATR dependence (Fig. 5D–F, I, and J).
Olaparib and VE-822 display selective toxicity against Atm-deficient KrasG12D/wt;Tp53fl/fl lung adenocarcinomas in vivo
To solidify our results and to address a potential relevance for human patients, we next asked whether ATM is indeed recurrently altered in human lung adenocarcinoma. By reanalyzing human whole-exome sequencing data of 139 lung adenocarcinomas provided by TCGA (44), we found one case that harbored two frameshift deletions in ATM (p.R250fs and p.L2006fs). In addition, this sample has undergone a whole-genome duplication and the allelic fractions of both ATM mutations clearly indicate that they occurred before the duplication. This suggests that both hits of ATM have emerged rather early in the tumor evolution of this patient. These data are further corroborated by recently published observations, which indicate that approximately 40% of human lung adenocarcinomas lack ATM protein expression (7). Mutations are not the only mechanism by which ATM expression may be repressed. Epigenetic regulation, as well as posttranscriptional mechanisms may also be involved, suggesting that the fraction of ATM-defective lung adenocarcinoma patients is larger than 1/139 and that both genomic and IHC-based stratification algorithms may be clinically useful.
To further validate our in vitro observations, which strongly suggest that Atm-deficient KrasG12D/wt;Tp53fl/fl lung adenocarcinoma cells display actionable dependencies on PARP1 and ATR, we next assessed the therapeutic efficacy of single-agent olaparib and VE-822 in our orthotopic transplantation models of KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors (Figs. 6 and 7). While KP tumors displayed continued growth (157.5 ± 43.96 percent change in mean tumor volume) in respone to olaparib (50 mg/kg, intraperitoneally, days 1–21), KPAfl/Δ tumors remained largely stable in size (107.0 ± 21.09 fold change in mean tumor volume; Fig. 6A–C). In marked contrast, KPAΔ/Δ-derived tumors displayed a robust and statistically significant (P < 0.001) volume reduction, which was detectable as early as 7 days following initiation of olaparib treatment (Fig. 6A–C). We note that vehicle-treated KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors displayed continuous growth throughout the entire observation period of 21 days (Fig. 6A–C). In line with these CT-morphologic tumor response data, we observed that olaparib treatment led to a significantly prolonged overall survival in animals bearing KPAΔ/Δ-derived tumors, compared with vehicle-treated controls (36.5 vs. >140.0 days median survival, P < 0.0001; Fig. 6D). While olaparib treatment also induced mild survival gains in mice bearing KP- and KPAfl/Δ-derived tumors (33.0 vs. 70.5 and 28.5 vs. 82.0 days median survival, respectively), this trend was substantially less obvious and the overall survival gains were marginal, compared with the effects observed in KPAΔ/Δ tumor–bearing mice (Fig. 6D). Of note, olaparib treatment (50 mg/kg, intraperitoneally, days 1–21) of KrasLSL.G12D/wt and KrasLSL.G12D/wt;Atmfl/fl animals bearing autochthonous lung adenocarcinomas revealed similar results (Supplementary Fig. S9A and S9B). While KrasLSL.G12D/wt tumor–bearing animals did not derive a survival advantage from 21 days of continuous olaparib treatment, KrasLSL.G12D/wt;Atmfl/fl tumor–bearing mice displayed a massive and statistically highly significant survival gain following a 21-day course of intraperitoneal olaparib exposure (Supplementary Fig. S9A and S9B). Together, these data strongly suggest that it is the lack of Atm, which dictates the response to olaparib. This synthetic lethal interaction between biallelic Atm deletion and PARP1 inhibition appears to be independent of the functional Tp53 status in KrasG12D-driven lung adenocarcinoma.
Combined loss of Atm and Tp53 constitutes a genetic constellation that is associated with sensitivity to PARP-inhibition in Kras-driven lung adenocarcinomas in vivo. A, μCT-based response assessment of tumor formation revealed tumor shrinkage in mice bearing KPAΔ/Δ-derived tumors 7 days following initiation of olaparib treatment (50 mg/kg, i.p., once daily). 10 days after tumor injection, lungs injected with either KP-, KPAfl/Δ-, and KPAΔ/Δ cells were imaged (μCT) to confirm tumor formation. After 7 days of vehicle (left) or olaparib treatment (right), mice were reimaged. Representative μCT images before (top) and after treatment (bottom) are shown. Red arrows, individual tumor lesions. H, heart. B and C, Olaparib treatment (50 mg/kg, i.p., once daily) led to a tumor volume reduction in lungs injected with KPAΔ/Δ cells. B, Tumor volume was longitudinally measured for 21 days. C, Tumor volumes were quantified after 7 days of treatment and normalized to pretreatment values (x-fold change, y-axis). Volume changes in the treatment cohort (red) were compared with the vehicle-treated control group (blue). Significance is indicated by asterisk and calculated by two-tailed Student t test. D, Olaparib treatment (50 mg/kg, i.p., once daily) increases overall survival in mice bearing KPAΔ/Δ-derived tumors, compared with vehicle-treated controls. Mild survival gains were also observed in animals bearing KP- and KPAfl/Δ-derived tumors. Survival of animals was followed for 150 days. Survival curves were compared by log-rank (Mantel–Cox) test. E and F, PARP inhibition induces a significant reduction in Ki-67 positivity in KPAΔ/Δ-derived tumors, indicating reduced proliferation. E, Representative images of Ki-67 stainings. Tumors of all three genotypes (KP, KPAfl/Δ, and KPAΔ/Δ) were treated with vehicle or olaparib for five consecutive days. Scale bar, 50 μm. F, Quantification of Ki-67 scores. Significance is indicated by asterisk and calculated by two-tailed Student t test. n.s., nonsignificant.
Combined loss of Atm and Tp53 constitutes a genetic constellation that is associated with sensitivity to ATR-inhibition in Kras-driven lung adenocarcinomas in vivo. A, μCT-based response assessment of tumor formation revealed tumor shrinkage in mice bearing KPAΔ/Δ-derived tumors 7 days following initiation of VE-822 treatment (30 mg/kg, orally, days 1–3, 8–10, and 15–17). Ten days after tumor injection, lungs injected with either KP-, KPAfl/Δ-, and KPAΔ/Δ cells were imaged (μCT) to confirm tumor formation. After 7 days of vehicle (left) or VE-822 treatment (right), mice were reimaged. Representative μCT images before (top) und after treatment (bottom) are shown. Red arrows, individual tumor lesions. H, heart. B and C, VE-822 treatment (30 mg/kg, orally, days 1–3, 8–10, and 15–17) led to a tumor volume reduction in lungs injected with KPAΔ/Δ cells. B, Tumor volume was longitudinally measured for 14 days. C, Tumor volumes were quantified after 7 days of treatment and normalized to pretreatment values (x-fold change, y-axis). Volume changes in the treatment cohort (red) were compared with vehicle-treated control group (blue). Significance is indicated by asterisk and calculated by two-tailed Student t test. D, VE-822 treatment (30 mg/kg, orally, days 1–3, 8–10, and 15–17) increases overall survival in mice bearing KPAΔ/Δ-derived tumors, compared with vehicle-treated controls. Survival of animals was followed for 100 days. Survival curves were compared by log-rank (Mantel–Cox) test. E and F, ATR inhibition induces a significant reduction in Ki-67 scores in mice bearing KPAΔ/Δ-derived tumors, indicating reduced proliferation. E, Representative images of Ki-67 stainings. Tumors of all three genotypes (KP, KPAfl/Δ, and KPAΔ/Δ) were treated with vehicle or VE-822 for three consecutive days. Scale bar, 50 μm. F, Quantification of Ki-67 scores. Significance is indicated by asterisk and calculated by two-tailed Student t test. n.s., nonsignificant.
The cytotoxic effect of olaparib on KPAΔ/Δ-derived tumors was also validated in IHC analyses. We specifically stained KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors with an antibody detecting Ki-67 following in vivo olaparib treatment. As shown in Fig. 6E and F and Supplementary Fig. S9, olaparib did not have an effect on the size of the Ki-67–positive tumor cell fraction in KP-derived tumors and only induced a mild reduction in Ki-67 positivity in KPAfl/Δ-derived tumors. In marked contrast, olaparib treatment led to a massive and statistically highly significant reduction of the Ki-67 index in KPAΔ/Δ-derived tumors. (Fig. 6E and F; Supplementary Fig. S10).
We next aimed to validate the therapeutic efficacy of VE-822 in vivo, using our orthotopic transplantation models of KP-, KPAfl/Δ- and KPAΔ/Δ-derived tumors (Fig. 7). As shown in Fig. 7A–C, VE-822 treatment (30 mg/kg, orally, days 1–3, 8–10, 15–17) of KP- and KPAfl/Δ tumor–bearing mice did not result in any significant tumor volume changes, compared with vehicle-treated controls. In marked contrast, and in line with our in vitro results (Fig. 5D and F), KPAΔ/Δ-derived tumors displayed a substantial and statistically significant volume shrinkage, compared with vehicle-treated controls at day 7 (0.701 ± 0.203 fold change in mean tumor volume, P = 0.0054; Fig. 7B and C). This difference was readily detectable on day 7 following initiation of VE-822 treatment and remained detectable throughout the entire observation period of 21 days (Fig. 7B and C). In agreement with these tumor volume assessments, VE-822 treatment also led to a statistically significant (P < 0.0426) overall survival extension in KPAΔ/Δ tumor–bearing animals, compared with vehicle-treated controls (36.5 vs. 76 days median survival; Fig. 7D). In contrast, VE-822 did not lead to any significant overall survival gains in mice bearing KP- or KPAfl/Δ-derived tumors (Fig. 7D). These data are further corroborated by IHC experiments. We specifically stained sections of KP-, KPAfl/Δ-, and KPAΔ/Δ-derived tumors following in vivo VE-822 treatment with an antibody detecting Ki-67 (Fig. 7E and F; Supplementary Figs. S4 and S10). While VE-822 treatment did not induce a significant reduction in Ki-67 positivity in KP- and KPAfl/Δ-derived tumors, this compound induced a significant reduction in Ki-67–positive KPAΔ/Δ-derived tumors. Altogether, our in vivo results clearly demonstrate that the Atm status in high-risk KP tumors dictates the response to olaparib and VE-822. These data thus imply that the mutational status of ATM should be evaluated particularly in high-risk KRASmut and TP53mut tumors, as biallelic ATM alterations predict susceptibility to targeted therapeutic intervention with PARP1 or ATR inhibitors.
Discussion
Lung cancer is the leading cause of cancer-related deaths (1). Through recent cancer genome sequencing efforts, numerous oncogenic driver mutations could be identified in lung adenocarcinoma. Among these are several druggable genomic aberrations, such as oncogenic EGFR mutations, as well as ALK and ROS1 rearrangements (4). However, for the majority of lung adenocarcinomas, no direct pharmacologic approach is available to intercept oncogenically rewired signaling. A prominent example for nondruggable oncogenic driver mutations are KRAS alterations, which are detected in approximately 32% of human lung adenocarcinomas (5). There is robust experimental evidence indicating that KRAS-mutant carcinomas are oncogene addicted (45), which in turn makes KRAS an attractive drug target. Recent efforts of suppressing KRAS signaling led to the development of small-molecule inhibitors that bind to the prenyl-binding pocket of PDEδ to prevent the interaction between KRAS and PDEδ (46). Such compounds interfere with oncogenic RAS signaling by altering its intracellular localization and have shown preclinical activity in vitro and in vivo using xenograft models (46). Currently, the clinical feasibility of such an approach awaits further experimentation. In addition, irreversible allosteric inhibitors of KRASG12C have recently been developed and were shown to display in vitro activity in KRASG12C-mutant cancer cell lines (47). These compounds are thus far limited to the KRASG12C-mutant, as they rely on the mutant cysteine for binding. Hence, direct targeting of KRAS remains a difficult challenge and thus far no clinically available KRAS inhibitors have been developed.
As direct targeting of KRAS is difficult, indirect approaches have been proposed. For instance, interception of the downstream effector pathways using combinations consisting of MEK and PI3K inhibitors have been evaluated (48). These regimens have shown promising preclinical activity and multiple clinical trials are underway, but the clinical feasibility and validity has yet to be demonstrated. In addition, combined inhibition of the checkpoint kinases CHK1 and MK2 was recently shown to display preclinical activity against different Kras-driven cancer entities (6).
Here, we employed autochthonous mouse models, as well as orthotopic transplantation models of Kras-driven standard- and high-risk lung adenocarcinoma to investigate whether cooccurring Atm alterations might create a window for therapeutic intervention independent of the driving oncogenic Kras p.G12D mutation. While biallelic Atm deletions could readily be induced in Tp53-proficient Kras-driven lung adenocarcinomas (Supplementary Fig. S1), the combined acute Adeno-Cre-induced loss of Tp53 and Atm appeared to be counterselected, in vivo. In only 1 of 30 cell lines isolated from primary murine lung adenocarcinoma cell lines isolated from KPA mice, could we detect a biallelic loss of Tp53 and Atm. In the remaining 29 cases, we found both alleles of Tp53 to be effectively recombined, while only one of the Atm alleles was deleted. These observations are in line with recently published cancer genome sequencing efforts, which revealed that comutation of ATM and TP53 are significantly underrepresented in lung adenocarcinoma (3, 44, 49). A possible rationale for this observation may be that both genes are involved in the cellular response to genotoxic stress and that combined loss of these central components of the DDR may overwhelm the cellular response mechanisms against DNA damage. Two pieces of circumstantial evidence may support this hypothesis. First, it was recently shown that acute expression of oncogenic Kras causes DNA damage, likely due to the induction of replicative stress (6, 50). Second, it was also shown that murine embryonic fibroblasts that were engineered to lack both Tp53 and Atm display a substantially enhanced sensitivity against genotoxic chemotherapeutics, such as the topoisomerase I poisons camptothecin and topotecan, the topoisomerase II poisons doxorubicin, epirubicin and etoposide, as well as the antimetabolites 5-fluorouracil and gemcitabine, but not to cisplatin, carboplatin, and oxaliplatin, and the taxanes, docetaxel and paclitaxel (51). These data are further in line with a previous report, suggesting that combined inactivation of Tp53 and Atm is statistically significantly underrepresented in human tumors (21). Thus, overall, we interpret our observation as follows: the acute combined loss of Atm and Tp53 may be detrimental to the affected cell, due a severely impaired DDR. However, the sequential loss of the two potent tumor suppressor genes may be tolerable, due to the induction of escape mechanisms, such as alternative DNA repair pathways.
Our data clearly indicate that Kras-driven murine lung adenocarcinomas that lack Tp53 and Atm are hypersensitive to olaparib, but not cisplatin. This is somewhat surprising, as BRCA1- or BRCA2-mutant breast cancer patients typically respond very well to cisplatin and olaparib (52). Furthermore, ATM, BRCA1, and BRCA2 are all involved in homologous recombination–mediated DNA repair (15). However, in contrast to ATM, at least BRCA2 is also centrally involved in the Fanconi anemia pathway, which is employed to clear interstrand crosslinks—a DNA lesion typically induced by platinum salts (15). Thus, the differential sensitivity of Atm-defective lung adenocarcinoma to cisplatin and olaparib may, at least in part, be the result of an impaired HR pathway in these tumor cells. Our data are in line with phenotypes observed in ATM-defective DT40 cells, which have been shown to be hypersensitive to PARP inhibition, as well as topoisomerase-I and II poisons (53, 54).
The data presented here clearly indicate that Kras-driven lung adenocarcinomas lacking Atm develop with similar kinetics as their Atm-proficient counterparts in Tp53 wild-type settings. Furthermore, Kras-driven lung adenocarcinomas lacking both Tp53 and Atm do form in vivo, albeit at low frequency, thus resembling the human situation. We further demonstrate that biallelic Atm deletions are associated with an actionable molecular dependence on PARP1 and ATR activity in both Tp53-proficient (standard-risk) and Tp53-deficient (high-risk) Kras-driven lung adenocarcinomas. Through reanalyzing publicly available exome sequencing data of human lung adenocarcinoma patients, we could demonstrate that tumors with biallelic loss of ATM do exist in patients. However, we found only one sample that display compound heterozygous ATM mutations out of 139 tumors that were analyzed. This is somewhat contrasted by recent IHC data indicating that up to 40% of human lung adenocarcinomas display lack of ATM expression. This discrepancy between the different technologies to assess ATM status indicates that a clinical patient stratification should probably involve both, sequencing data and IHC. Overall, our observations indicate that genomic aberrations that exist in parallel to the major driving lesions encode therapeutically relevant liabilities. Furthermore, our data implicate that the functional ATM status should be evaluated in human lung adenocarcinomas for which no directly targeted therapeutic approach exists, regardless of the Tp53 status of these tumors. In fact, our data suggest that the Atm deficiency–associated sensitivity against PARP1 and ATR inhibitors is preserved even in Tp53-mutant high-risk lung adenocarcinomas.
Disclosure of Potential Conflicts of Interest
M. Peifer is a consultant/advisory board member for NEO New Oncology GmbH. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design:A. Schmitt, H.C. Reinhardt
Development of methodology:A. Schmitt, D. Welcker
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.):A. Schmitt, J. George, M. Nowak, R. Buttner, S. Perner
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis):A. Schmitt, G. Knittel, T.-P. Yang, S. Perner, M. Peifer
Writing, review, and/or revision of the manuscript:A. Schmitt, G. Knittel, D. Welcker, R. Buttner, S. Perner, H.C. Reinhardt
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):A. Schmitt, J. George, U. Leeser, R. Buttner, H.C. Reinhardt
Study supervision:H.C. Reinhardt
Grant Support
This work was supported by the Volkswagenstiftung (Lichtenberg Program; H.C. Reinhardt), the Deutsche Forschungsgemeinschaft (KFO-286; H. C. Reinhardt and M. Peifer), the Bundesministerium für Bildung und Forschung (SMOOSE; H. C. Reinhardt, M. Peifer, R. Büttner and MILES, M. Peifer), the German federal state North Rhine Westphalia (NRW) as part of the EFRE initiative (grant LS-1-1-030a to H. C. Reinhardt), the Else Kröner-Fresenius Stiftung (EKFS-2014-A06 to H. C. Reinhardt), and the Deutsche Krebshilfe (111724 to H. C. Reinhardt).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Acknowledgments
We are indebted to our patients, who provided primary material. We thank Alexandra Florin, Marion Müller, and Ursula Rommerscheidt-Fuß from the Institute of Pathology, University Hospital Cologne, for their outstanding technical support. Some analyses conducted in this work were in parts based upon data that was generated by The Cancer Genome Atlas managed by the NCI and NHGRI.
Footnotes
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received December 20, 2016.
- Revision received March 9, 2017.
- Accepted March 27, 2017.
- ©2017 American Association for Cancer Research.
References
Volume 77, Issue 11
