Targeting a non-oncogene addiction to the ATR/CHK1 axis for the treatment of small cell lung cancer

Small cell lung cancer (SCLC) is a difficult to treat subtype of lung cancer. One of the hallmarks of SCLC is its almost uniform chemotherapy sensitivity. However, chemotherapy response is typically transient and patients frequently succumb to SCLC within a year following diagnosis. We performed a transcriptome analysis of the major human lung cancer entities. We show a significant overexpression of genes involved in the DNA damage response, specifically in SCLC. Particularly CHEK1, which encodes for the cell cycle checkpoint kinase CHK1, is significantly overexpressed in SCLC, compared to lung adenocarcinoma. In line with uncontrolled cell cycle progression in SCLC, we find that CDC25A, B and C mRNAs are expressed at significantly higher levels in SCLC, compared to lung adenocarcinoma. We next profiled the efficacy of compounds targeting CHK1 and ATR. Both, ATR- and CHK1 inhibitors induce genotoxic damage and apoptosis in human and murine SCLC cell lines, but not in lung adenocarcinoma cells. We further demonstrate that murine SCLC tumors were highly sensitive to ATR- and CHK1 inhibitors, while Kras G12D-driven murine lung adenocarcinomas were resistant against these compounds and displayed continued growth under therapy. Altogether, our data indicate that SCLC displays an actionable dependence on ATR/CHK1-mediated cell cycle checkpoints.


Results
Transcriptome analysis in human lung cancer specimens reveals a significantly increased expression of DNA damage response genes in SCLC, compared to NSCLC. In light of the high proliferation rate of SCLC tumors and the almost universal bi-allelic loss of both tumor suppressor genes, TP53 and RB1 2,3 , we sought to further characterize the transcriptional profile of SCLC in comparison to other major lung cancer subtypes. We therefore performed a comparative analysis with the transcriptome sequencing data of 214 patient-derived lung tumors 2,3,[25][26][27][28] . In comparison to lung adenocarcinomas (ADC) and squamous cell carcinomas (SqCC), where co-occurring somatic alterations of TP53 and RB1 are less frequent and rather rare 25,26 , SCLC tumors exhibited significantly higher expression levels of genes controlling cell cycle regulation and DNA replication, as well as pathways that emphasize the neuroendocrine features of this lung cancer subtype (Fig. 1A). We furthermore observed a massive up-regulation of mRNAs encoding for different DNA damage response (DDR) and DNA repair pathways (Figs 1A,B, S1), which was similarly observed through previous proteomic studies in SCLC, as well as in a recent transcriptome analysis 23,24 . The detailed analysis of the genes involved in these cellular mechanisms pointed, among others, to CHEK1 (Fig. 1B). CHEK1 transcripts were significantly up-regulated in SCLC tumors with a median increase of 2-fold (1.7-fold) and 5-fold (4.6-fold), compared to adenocarcinomas and squamous cell carcinomas, respectively (p < 0.0001, Fig. 1C).
CHEK1 encodes for one of the three major cell cycle checkpoint effector kinases (CHK1, CHK2, MK2), which in the absence of p53 and RB1 may initiate cell cycle arrest and subsequent DNA repair mechanisms 14,15 . Intriguingly, and in line with poorly controlled cell cycle progression in SCLC, we find that the mRNAs encoding the phosphatases CDC25A, B and C are expressed at significantly higher levels in SCLC, compared to SqCC and ADC samples (Fig. 1D). Together, our observations therefore support the notion that in response to endogenous and exogenous genotoxic stress, SCLC tumors may exploit alternative pathways for DNA repair (Figs 1E, S1) and thus tumor maintenance. The elevated expression levels of CHEK1 point to a dependence on ATR/CHK1 and may suggest a particularly high vulnerability of SCLC tumors to inhibitors targeting the ATR/CHK1 signaling pathway (Fig. 1E). Murine SCLC cell lines display an actionable dependence on the ATR/CHK1 cell cycle checkpoint signaling axis. With regard to the high expression levels of CHEK1 in SCLC tumors, we next sought to address the vulnerability of these tumors to inhibitors targeting the CHK1 pathway. In order to directly address this hypothesis, we generated Rb1 fl/fl ;Tp53 fl/fl mice ( Fig. 2A, hereafter referred to as RP) and induced SCLC in these animals through intratracheal application of an adenovirus driving Cre recombinase expression (Ad-CMV-Cre), specifically in pulmonary epithelial cells 29 . In addition, we generated Kras LSL.G12D/wt ;Tp53 fl/fI mice ( Fig. 2A, hereafter referred to as KP), in which we induced lung adenocarcinomas, through intratracheal instillation of Ad-CMV-Cre 30 . Once mice reached a terminal stage of their disease, animals were sacrificed and we isolated individual tumors to derive stable tumor cell cultures, as previously described 31,32 . We specifically established five SCLC clones and five NSCLC clones. Each individual clone was established from an individual tumor isolated from different mice. As an initial characterization step, we recorded the proliferation rate of the individual cancer cell lines. As shown in Fig. 2B, all five NSCLC clones displayed rapid proliferation with only minimal variation between the individual clones. In contrast, SCLC clones displayed a more divergent proliferation pattern. Clones RP1, 3 and 4 displayed comparable proliferation rates, which were substantially slower than those observed in the NSCLC clones. Clone RP2 displayed substantially slower growth kinetics than RP1, 3 and 4, while clone RP5 showed a proliferation rate that was similar to that observed in the NSCLC clones.
To further characterize the biological properties of the individual cell lines, we next profiled their sensitivity to the frontline chemotherapeutic agents cisplatin (Fig. 2C) and etoposide (Fig. 2D), using intracellular ATP measurements as surrogate markers for cell viability. NSCLC cells were slightly more sensitive to cisplatin (GI 50 = 6.7 µMol), than SCLC cell lines (GI 50 = 12.3 µMol). In contrast, NSCLC cells were more resistant to etoposide (GI 50 = 29.1 µMol), compared to SCLC clones (GI 50 = 14.3 µMol).
To test the hypothesis that ATR-or CHK1 inhibitors display particular efficacy in SCLC cells, compared to NSCLC cells, we next profiled the sensitivity of our cell line panel to the ATR inhibitor VE-822 and the CHK1 inhibitor PF-477736. Of note, PF-477736 has been extensively profiled in a kinase screening panel representing >100 protein kinases. Among the profiled kinases, only 7 enzymes were inhibited by PF-477736 with <100-fold selectivity, namely VEGFR2, Aurora-A, FGFR3, FLT3, CSF1R, RET and YES 33 . PF-477736 displays a ~100-fold selectivity for CHK1, compared to CHK2 33 . Similarly, VE-822 is a selective ATR inhibitor with IC 50 values of 0.019 µM, 2.6 µM, and 18.1 µM for ATR, ATM and DNA-PK, respectively 34 . VE-822 displays negligible activity against mTOR and PI3Kγ 34 . Both PF-477736 and VE-822 are currently being investigated in clinical trials (e.g. NCT00437203, NCT02723864). As shown in Fig. 2E, SCLC cells were significantly (P < 0.0001) more sensitive to ATR inhibition through VE-822, than NSCLC cells (GI 50 = 1.9 µMol vs. GI 50 = 7.2 µMol, respectively). Similarly, SCLC cells were significantly (P < 0.0001) more sensitive to PF-477736, compared to NSCLC cell lines (GI 50 = 1.7 µMol vs. GI 50 = 6.9 µMol, respectively) (Fig. 2F). As ATR and CHK1 operate in a linear signaling pathway, the observation that VE-822 and PF-477736 both display cytotoxic activity in SCLC cells, cross-validates the biological relevance of the ATR/CHK1 pathway for the survival of SCLC cells. To further validate ATR and CHK1 as the relevant drug targets and to exclude structure-specific off-target effects, we re-screened our cell line panel by replacing VE-822 and PF-477736 with alternative ATR-and CHK1 inhibitors that were based on entirely distinct chemical motifs. VE-822 was exchanged against AZD-6738, and PF-477736 was replaced with AZD-7762. As shown in Fig. S2, Both AZD-6738 and AZD-7762 displayed similar cytotoxicity as VE-822 and PF-477736 in SCLC cell lines, while NSCLC cell lines remained largely unaffected, even at a concentration of 3 µM.
As cell viability assessments through intracellular ATP measurements do not distinguish between growth retardation and active cell death, we next performed flow cytometry experiments to directly test the hypothesis that ATR-and CHK1 inhibitors induce apoptotic cell death in SCLC cell lines ( Fig. 2G-J). We specifically assessed the percentage of Annexin-V/PI double positive cells in response to increasing doses of cisplatin and etoposide, which served as positive controls for the induction of apoptosis in our cell line panel. While both chemotherapeutic agents induced substantial levels of apoptosis in a dose-dependent manner, we did not detect any substantial differences in the sensitivity of SCLC and NSCLC cell lines to these agents at lower concentrations ( Fig. 2G,H). Of note, at higher doses of cisplatin (2.5 µM and 10 µM) and etoposide (5 µM), NSCLC clones displayed a significantly higher degree of apoptosis, than their SCLC counterparts. Furthermore, there were no significant differences in the chemotherapy sensitivity between the different SCLC clones that differed in their individual proliferation rates (cf. Fig. 2B). In full agreement with our cell viability assays (cf. Fig. 2E,F), we observed highly significant differences between SCLC and NSCLC cell lines in response to VE-822 and PF-477736 ( Fig. 2I,J, respectively). While NSCLC cells displayed only a mild, dose-dependent increase in the percentage of apoptotic cells, SCLC cells were significantly more sensitive to both agents. Of note, the slow growing cell line RP2 was less sensitive to VE-822 and PF-477736, compared to cell lines RP1, 3 and 4. The rapidly proliferating SCLC cell line RP5 displayed the most pronounced sensitivity to the cell cycle checkpoint abrogating agents VE-822 and PF-477736. Thus, overall, we provide data from two distinct experimental assays to show that murine SCLC cell lines are highly sensitive to ATR-and CHK1 inhibition, compared to murine NSCLC cells, which were largely resistant. Intriguingly, the NSCLC cell lines in our experiments displayed higher proliferation rates, than the SCLC cell lines included in our panel (Fig. 2B). Thus, we observed an entity-specific sensitivity to cell cycle checkpoint abrogation. Furthermore, within the SCLC entity, we could show a proliferation rate-dependent VE-822 and PF-477736 sensitivity.
Inhibition of ATR and CHK1 induces genotoxic stress in murine SCLC cell lines. We next aimed to further characterize the biological effects of ATR-and CHK1 inhibition in our murine lung cancer cell line panel. We specifically investigated the occurrence of genotoxic lesions in a longitudinal fashion, as ATR and CHK1 kinase activity is critical for a proper progression through S-phase 15,35 . Corroborating this role of the ATR/CHK1 pathway for genome maintenance in S-phase, Atr and Chek1 constitutive knockout mice were shown to be embryonically lethal 36,37 . Furthermore, Chek1 conditional knockouts display signs of genotoxic damage, upon Chek1 deletion 38 .  Lastly, ATR-and CHK1 inhibitors have previously been shown to cause genotoxic damage 39,40 . To interrogate the effects of ATR-and CHK1 inhibition on genome integrity, we assessed the induction of genotoxic damage, using indirect immunofluorescence. We specifically employed an antibody to detect phospho-Ser-139 in the histone variant H2AX (γH2AX), an established marker for DNA damage 16,32,41,42 . These experiments revealed that VE-822 (0.75 µM) and PF-477736 (1.0 µM) inflicted genotoxic damage in both, NSCLC and SCLC cell lines (Figs 3A,B, S3A-C). In line with a significantly enhanced toxicity, specifically in SCLC cells (cf. Figs 2, S2), VE-822 and PF-477736 induced significantly more DNA damage in SCLC cells, than in NSCLC cell lines at 12, 24 and 48hrs of drug exposure (Figs 3A,B, S3A-C). Of note, even RP2 cells, which displayed the lowest proliferation rate (cf. Fig. 2B), showed VE-822-and PF-477736-induced γH2AX indices that were comparable to those observed in the remaining SCLC cell lines ( Fig. S3A-C). To further substantiate these immunofluorescence data, we next performed immunoblot experiments to reproduce our observations in a distinct assay (Figs 3C and D, S3D,E). For this purpose, we treated our cell line panel with VE-822 (0.75 µM, 48 hours) or PF-477736 (1.0 µM, 48 hours). Vehicle-treated cells served as a control. Upon completion of drug exposure, cells were lysed, separated on SDS-PAGE and γH2AX was visualized using a specific antibody. β-actin staining served as a loading control. The intensity of the resulting enhanced chemiluminescence signal was quantified using densitometry. Fully in line with our immunofluorescence data, SCLC cell lines displayed a significantly stronger γH2AX signal following VE-822 and PF-477736 treatment, than their NSCLC counterparts (Figs 3C and D, S3D,E). Thus, both ATR-and CHK1 inhibition induced genotoxic damage that was significantly more pronounced in SCLC, compared to NSCLC cell lines.
Of note, we performed hematoxylin/eosin- (Fig. 4B) and Ki-67 stainings ( Fig. S4A) of lung tumor sections isolated from transplant recipients, to verify that the transplanted SCLC-and NSCLC-derived cell lines retain their histological properties and proliferation characteristics upon serial passaging in vitro and subsequent transplantation.
Once CT-morphologically verified tumors had formed in the recipient animals, mice were treated with four cycles of VE-822 (30 mg/kg, orally, on days 1-3 of a 7-day cycle), PF-477736 (20 mg/kg, i.p., on days 1-5 of a 7-day cycle) or treated with vehicle control. Tumor volume assessments were performed, using µCT imaging on the day prior to cycle 1, as well as after completion of cycle 2 and 4 ( Fig. 4A,C-G). When treated with vehicle solution, both SCLC and NSCLC tumors displayed continuous growth throughout the four cycle observation period (Figs 4C-F, S4B,C). Of note, NSCLC tumors grew more rapidly, compared to their SCLC counterparts. Due to this more rapid proliferation pattern, one vehicle-treated animal harboring KP1-derived tumors and two vehicle-treated animals carrying KP2-derived tumors did not survive until the end of cycle four (Figs 4F, S4B). Similarly, NSCLC tumors displayed continued growth when treated with VE-822 or PF-477736 (Figs 4D-F, S4B), essentially mimicking the resistance phenotype that we had observed in vitro ( Fig. 2C-J). We note that one VE-822-treated animal carrying KP1-derived tumors did not survive until completion of the fourth cycle, due to massive tumor progression (Figs 4F, S4B). In marked contrast and in line with our in vitro data, all SCLC tumors displayed significant volume shrinkage in response to VE-822 and PF-47736 compared to vehicle-treated animals, both following cycle 2 (VE-822: p = 0.0004; PF-477736: p < 0.0001) and 4 (VE-822: p = 0.0001; PF-477736: p < 0.0001) (Figs 4C,E-G, S4C). These µCT-morphological tumor response data also translated into highly significant survival gains of VE-822-and PF-477736-treated animals carrying SCLC tumors, compared to their vehicle-treated counterparts (p < 0.0001 and p = 0.0008, respectively) (Fig. 4H). While animals harboring vehicle-treated SCLC tumors displayed a median survival of 66 days following initiation of treatment, VE-822-teated mice showed a median survival of 133 days (Fig. 4H). PF-477736-treated SCLC-bearing animals showed a median survival of 141 days (Fig. 4H). Contrary to these substantial therapeutic effects inflicted by VE-822 and PF-477736 in SCLC-bearing animals, we only observed a minute survival extension in VE-822-and PF-477736-treated NSCLC-bearing mice (Fig. 4H). Median survival in vehicle-treated NSCLC-bearing mice was 40 days, compared to 55 and 54 days following VE-822-and PF-477736-treatment, respectively (Fig. 4H).
To further substantiate these observations, we next employed an autochthonous mouse model of SCLC. For that purpose, we induced SCLC in Tp53 fl/fl ;Rb1 fl/fl mice through intratracheal Ad-CMV-Cre instillation, as previously described (Fig. 5) 29 . Tumor onset was monitored by serial MRT imaging. Once MRT-morphologically-verified tumors had formed, we initiated treatment with two cycles of PF-477736 (20 mg/ kg, i.p., on days 1-5 of a 7-day cycle), or vehicle control (Fig. 5A). While tumors displayed continuous growth in vehicle-treated animals, all tumors in the PF-477736-exposed cohort displayed a substantial volume reduction after completion of the second cycle (Fig. 5B). This morphological treatment response also translated into a significant survival benefit in the PF-477736-treated cohort, compared to the vehicle-treated animals (median survival inhibition (PF-477736), than KP cell lines. RP GI 50 = 1.7 µMol; KP GI 50 = 6.9 µMol. (G-J) Flow cytometrybased apoptosis measurements. Cells were treated for 48 hours with the indicated drugs at the indicated doses. Apoptosis was assessed by quantification of the Annexin-V/PI double-positive population. Average values of three independent experiments are shown. (G-H) RP and KP cell lines displayed a similar degree of apoptotic cell death in response to cisplatin and etoposide. (I-J) Highly significant differences in apoptotic cell death were observed between RP and KP cell lines in response to VE-822 and PF-477736. RP2 was less sensitive to both treatments, compared to RP1, 3 and 4. RP5 displayed the most pronounced sensitivity. Significance was determined using an unpaired t test. Levels of significance were *p < 0.05; **p < 0.01; ***p < 0.001. 79 days, vs 19 days, respectively, p = 0.0286) (Fig. 5C,D). Collectively, our data derived from transplant models, as well as autochthonous SCLC models, indicate that cell cycle checkpoint inhibition through ATR-and CHK1 inhibition displays selective therapeutic efficacy in murine SCLC tumors, compared to NSCLC tumors, in vivo. We note that the adverse effects of VE-822 and PF-477736 were generally mild, in vivo. No animal reached the predefined termination criteria (weight loss >10%, signs of infection, bleeding, diarrhea, adynamia or affected social behavior), due to compound toxicity.  In line with our data derived from primary tumor specimens (Fig. 1), we found that SCLC cell lines display significantly higher levels of CHEK1 expression, than NSCLC and non-lung cancer cell lines (Fig. 6A). To assess whether the profound sensitivity to CHK1 inhibition that we had observed in murine cancer cell lines, was preserved in human SCLC cells, we next performed cell  Fig. S5E for H526, H82, H69, A549, H1975 and HCC44 cells 44 . H526, H69 and H82 cells are reported to carry bi-allelic TP53 and RB1 mutations. Intriguingly, a re-analysis of human lung cancer cell lines curated in the Catalogue Of Somatic Mutations in Cancer (COSMIC) database revealed that bi-allelic TP53 and RB1 aberrations can also be detected in a subset NSCLC lines (Fig. S5F).
To further probe the sensitivity of human SCLC cells against ATR-and CHK1 inhibitors, we next conducted therapy studies in xenograft tumors (Fig. 6D-F). For that purpose, we transplanted 1.5 × 10 6 H-82 (Fig. 6D), H-526 (Fig. 6E) and H-69 (Fig. 6F) cells into the flanks of 8-week-old athymic NMRI-Foxn1 nu/nu mice and allowed xenograft tumors to form. Tumor volumes were assessed by serial caliper measurements (Fig. S5G). Once, tumors reached a volume of 100mm 3 , treatment with VE-822 (30 mg/kg, orally, on days 1-3 of a 7-day cycle), PF-477736 (20 mg/kg, i.p., on days 1-5 of a 7-day cycle) or vehicle solution was initiated. While vehicle-treated xenograft tumors derived from H-82 (Fig. 6D), H-526 (Fig. 6E) and H-69 (Fig. 6F) cells displayed a rapid growth, both VE-822 and PF-477736 treatment led to a massively delayed tumor growth in all three models, which was mimicked by a substantial reduction in Ki67 staining on histological examination in tumors that were excised following completion of treatment (Fig. 6G,H). In line with this sensitivity of human SCLC xenografts, we found a substantially increased expression of CHK1 in SCLC cell lines, compared to NSCLC cells (Fig. 6I). Moreover, basal CHK1 activity, as evidenced by Ser-345 phosphorylation, was also increased in SCLC cells, compared to NSCLC controls (Fig. 6I). Thus, in summary, we demonstrate that murine and human SCLC cells and tumors display an actionable molecular dependence on the cell cycle checkpoint kinases ATR and CHK1, which could not be detected in murine and human NSCLC cells and tumors.

Discussion
SCLC is a highly aggressive disease, which accounts for ~14% of all lung cancers 45 . Despite the conduction of a series clinical trials, including more than 40 phase III trials throughout the last 45 years, the principles of systemic treatment of SCLC patients with extensive stage disease have not substantially evolved over the past several decades 45 . This lack of therapeutic progress has left the 5-year overall survival rate below 7% 45 . This is in marked contrast to NSCLC, particularly adenocarcinoma, where genome sequencing efforts have revealed the identity of numerous recurrent, actionable somatic alterations, including oncogenic EGFR mutations, as well as ALK-and ROS1 rearrangements 46 .
In the Western world, the first line treatment of SCLC patients presenting with extensive stage disease consists of a platinum-based combination chemotherapy, typically cis-or carboplatin in combination with etoposide 4,45 .
In the relapsed setting, the topoisomerase-I poison topotecan is the only US Food and Drug Administration (FDA)-approved second line regimen for SCLC patients [45][46][47] . Further second line options beyond topotecan include taxanes, irinotecan, vinorelbine, and gemcitabine. In addition, temozolomide might be an attractive regimen for the second line situation, as it can be applied orally and displays activity in central nervous system lesions 48 . There is no internationally accepted consensus on systemic treatment regimens beyond firstand second-line therapy 45,49 . Furthermore, recently published exome and genome sequencing efforts have not revealed any recurrent actionable oncogenic genome aberrations 2,3 . after two and four cycles. The yellow area indicates individual tumor lesions. The heart is indicated (H). (E-F) Tumor volume changes in response to the indicated treatments were normalized to baseline tumor volumes, assessed prior to therapy initiation. NSCLC tumors grew faster, than SCLC tumors. Vehicle-treated mice showed continuous tumor growth throughout the observation period. SCLC tumors displayed significant tumor shrinkage after cycle 2 and 4 in response to VE-822 and PF-47736 treatment. NSCLC tumors did not respond to either of the treatments. Three mice from the NSCLC control cohort and one animal from the VE-822-treated NSCLC cohort died before completion of cycle four. Here, we analyzed the transcriptomes of 214 human lung cancer patient samples, including 53 SCLC samples, as well as 970 human cancer cell lines, including 61 SCLC and 109 NSCLC cell lines. This analysis surprisingly revealed that SCLC tumors display a significantly higher expression of genes involved in cell cycle regulation, DNA damage signaling and DNA repair, than NSCLC tumors (Figs 1 and 6). One of the most strikingly differentially expressed genes was CHEK1. This CHEK1 expression in human SCLC samples has recently also been reported in an independent study 24 . We further demonstrate that ATR-and CHK1 inhibitors display a selective toxicity in SCLC cells, as well as allograft and xenograft SCLC tumors, while NSCLC cell lines isolated from Kras-driven murine lung adenocarcinoma and allograft NSCLC tumors remained largely unaffected, when treated with these compounds (Figs 2, 3, 4 and 5). We note that it has not been demonstrated that either the re-expression of TP53 and/or RB1 in SCLC or the downregulation of these tumor suppressor genes in NSCLC may be specifically correlated with the CHK1-and ATR inhibitor response. Mechanistically, we show that exposure of SCLC cells leads to the accumulation of genotoxic damage and the induction of apoptotic cell death (Figs 2, 3 and 4). A possible explanation for the increased expression of cell cycle-regulating genes may lie in the uniform bi-allelic inactivation of TP53 and RB1 in SCLC 2,3 . This genomic constellation, which is a hallmark feature of SCLC 2,3 , likely promotes a substantial destabilization of cell cycle checkpoints in SCLC. While p53 transactivates a myriad of genes that repress cell cycle progression in response to genotoxic stress (e.g. CDKN1A, 14-3-3 s, GADD45A, RPRM 9 ), RB1 governs the G 1 restriction point through sequestration of E2F1 10-12 . Thus, combined loss of TP53 and RB1 is likely to promote unrestricted proliferation and a substantially impaired ability to halt the cell cycle in response to DNA damage. In line with a destabilized G 1 /S border as a common denominator for CHK1 inhibitor sensitivity, other groups have reported that MYC-driven diffuse large B cell lymphoma and CCND1-driven mantle cell lymphoma display a marked sensitivity against the CHK1 inhibitor PF-477736 50,51 . Altogether, it is tempting to speculate that SCLC cells display a compensatory upregulation of p53-and RB1-indpendent regulators of cell cycle progression, such as the checkpoint kinase CHK1. We note that a re-analysis of cell lines curated in the Catalogue Of Somatic Mutations In Cancer (COSMIC) database revealed  that bi-allelic loss of TP53 and RB1 can also be detected in a subset of human NSCLC cancer cell lines (Fig. S5F). It would be interesting to assess the sensitivity of these cell lines against ATR-and CHK1 inhibitors.
Intriguingly, numerous checkpoint kinase inhibitors, including CHK1-and combined CHK1/CHK2 inhibitors have been developed and are currently evaluated in phase I and II clinical trials, either as single agents or in combination with chemotherapeutic compounds 52,53 . No clinically-relevant genotype-mapped drug response predictors have been reported for CHK1 inhibitors, to date. However, the CHK1 inhibitor prexasertib (LY-2606368) is currently undergoing evaluation as a single agent in a phase II trial with extensive stage SCLC. Given the data reported in our study, it will be interesting to see whether the results reported in the prexasertib trial will lend further support to the clinical development of CHK1 inhibitors for the treatment of SCLC patients. It has also been reported that ATR-and CHK1 inhibitors display synergistic activity in cancer cells 54 . It will thus be interesting to study whether combined inhibition of ATR and CHK1 may display selective synergistic activity in SCLC.

Materials and Methods
Gene expression analysis. Transcriptome sequencing data of human lung tumors was analyzed by referring to the RNA-seq data of 101 lung adenocarcinomas 25,27,28 , 60 lung squamous cell carcinoma samples 26 and 53 SCLC tumor samples 2,3 . Gene-specific transcripts were quantified with RSEM 55 . Differentially expressed genes were determined with the SAM R package at a Q-value < 0.05 56 . Differentially expressed gene expression profiles were analyzed with DAVID 57 .
Autochthonous models of small cell lung cancer and lung adenocarcinoma. To induce murine SCLC and lung adenocarcinoma, we applied 2.5 × 10 7 PFU Adeno-CMV-Cre intratracheally to Rb1 fl/fl ;Tp53 fl/fl (RP) and Kras LSL.G12D/wt ;Tp53 fl/fl (KP) mice, respectively 29,30 . 10-week-old animals were anesthetized with Ketavet (100 mg/kg) and Rompun (20 mg/kg) by intraperitoneal injection, as previously described 31,32,58 . RP mice were scanned by MRI under isoflurane (2.5%) anesthesia to confirm tumor formation approximately 30 weeks after Adeno-Cre application. KP mice were scanned by µCT imaging (Aloka, Latheta LCT-100) under isoflurane (2.5%) anesthesia to confirm tumor formation 5 weeks after Adeno-Cre application. Mice were kept on a C57Bl6 background. The local authorities and the local animal protection committee approved all animal procedures. All mouse experiments were conducted in accordance with our Institutional Animal Care and Use Committee (IACUC). All animal experiments were approved by the responsible local government authorities, namely the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein Westfalen. All animal experiments were carried out in accordance with the relevant guidelines and regulations.
Allograft mouse model. For the allograft experiments, we anesthetized (2.5% isoflurane) and injected 8-12-week-old wildtype C57Bl6mice with 2 × 10 6 murine SCLC tumor cells (RP1, 3 and 4) and 1.5 × 10 6 murine NSCLC tumor cells (KP1 and 2) into the right lung. Tumor formation was verified by µCT imaging 9 days after injection. SCLC-and NSCLC tumor-bearing mice were randomly allocated to receive four cycles of VE-822, PF-477736 or vehicle, as indicated. Tumor volume changes were monitored by µCT imaging under 2.5% isoflurane at the end of two treatment cycles and assessed by OsiriX and DICOM viewer software packages (OsiriX v.8.2, Pixmeo, Switzerland). We normalized the tumor volumes under treatment to baseline tumor volumes, assessed prior to therapy initiation. Survival curves were calculated up to 175 days after the injection of tumor cells. The local authorities and the local animal protection committee approved all animal procedures. All mouse experiments were conducted in accordance with our Institutional Animal Care and Use Committee (IACUC). All animal experiments were approved by the responsible local government authorities, namely the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein Westfalen. All animal experiments were carried out in accordance with the relevant guidelines and regulations.
Xenograft mouse model. We generated 120 xenograft tumors from three different human SCLC cell lines (H-69, H-82, H-526). For each xenograft tumor, we collected 1.5 × 10 6 tumor cells, re-suspended them in serum-free media and subcutaneously injected them into the flanks of 8-week-old male athymic nude mice (NMRI-Foxn1 nu/nu , Janvier). Mice were monitored for the formation of palpable subcutaneous tumors. Upon manifestation of tumors (≥100 mm³), VE-822 or PF-477736 were administered for two cycles following the protocol described above. Diameters of subcutaneous tumors were measured by a digital caliper every second day following initiation of therapy. For tumor volume measurements, mice were anesthetized with 2.5% isoflurane. Tumor volumes were calculated, based on the modified ellipsoid formula V = 1/2 (width × length²). We normalized the tumor volumes under treatment to baseline tumor volumes, assessed prior to therapy initiation. In accordance with our Institutional Animal Care and Use Committee (IACUC), mice with a single tumor volume larger than 1.500 mm³ were sacrificed. The local authorities and the local animal protection committee approved all animal procedures. All mouse experiments were conducted in accordance with our Institutional Animal Care and Use Committee (IACUC). All animal experiments were approved by the responsible local government authorities, namely the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein Westfalen. All animal experiments were carried out in accordance with the relevant guidelines and regulations.
Histological analysis. Mice harboring SCLC or NSCLC tumors were sacrificed at the indicated time points 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) or Ki67 the samples were assessed by two independent board-certified pathologists (M.W., R.B.).
Cell lines and cell culture. Individual murine SCLC and NSCLC tumor nodules derived from individual RP and KP mice, respectively, were isolated and cultured in RPMI medium supplemented with 1% penicillin/ streptomycin and 10% fetal calf serum, to derive stable tumor cell lines, as previously described 31,32,58 . Human cell lines were obtained from the American Type Culture Collection (ATCC), the German Resource Center for Biological Material (DSMZ), or were a kind gift from Roman K. Thomas (Department of Translational Genomics, University of Cologne). All our cell lines were cultured at 37 °C in a humidified incubator supplemented with 5% CO 2 . RPMI medium supplemented with 10% fetal calf serum (FCS) and 1% penicillin/streptomycin was the culture medium used for all cell lines. We passaged all adherent cells lines by washing with PBS and subsequent incubation in trypsin/EDTA. By addition of culture medium, we inactivated trypsin and plated or diluted the cells afterwards. Passaging of our suspension cell lines was executed by suitable dilution of the cell suspension and adding fresh culture medium. All cell lines were routinely tested for infection with mycoplasma (MycoAlert, Lonza).

Population doubling.
To assess population doubling times, triplicates of each of the five murine SCLC cell lines and the five murine NSCLC cell lines were seeded in 10 mm dishes at a density of 1 × 10 6 cells per dish for SCLC and 750,000 cells per dish for NSCLC. After 48 hours, cells were trypsinized and counted on a Beckman Coulter cell counter. Subsequently, cells were seeded at the above-mentioned density. The experiment was performed over 12 days, resulting in a total of 7 passages. Doubling times were calculated using the following equation: Cell viability assay. To assess cell viability, we employed CellTiterGlo ® (Promega) assays, according to the instructions provided by the manufacturer. In brief, 100 µl of each cell line were plated into 96-well plates at a density of 5.000 cells per well. 24 hours after seeding, triplets of each cell line were treated, as indicated. Each drug concentration was pre-prepared in deep-well plates. 25 µl of this stock solution was transferred from the deep-well plate to the 96-well plates. Cells were exposed to treatment for 48 hours in a humidified incubator at 37 °C and 5% CO 2 . After incubation, 125 µl of the CellTiterGlo ® reagent were added to each well of the 96-well plate. The contents were mixed for 2 minutes in an orbital shaker to induce cell lysis. After a 10-minute incubation at room temperature, relative cell viability was determined by measuring the ATP content in each well on a luminescence plate reader, equipped with an automated stacking module (Tecan). We normalized luminescence values to luminescence intensities of control wells on the same plate, treated with vehicle solution.
Flow cytometry. In the flow cytometry experiments reported here, we used Annexin-V/PI staining to quantify the induction of apoptosis, exactly as previously described 59,60 . All murine SCLC and NSCLC cell lines were seeded into 6-well plates at approximately 50% density and were incubated for 24 hours in a humidified incubator at 37 °C and 5% CO 2 . Subsequently, triplets of each cell line were either left untreated or were exposed to one of the following inhibitors and dosages: cisplatin (0. Immunofluorescence. Murine SCLC and NSCLC cells were seeded at a confluency of approximately 50% on 6-well plates with a 22-mm glass coverslip on the bottom. The cells were treated with VE-822 (0.75 µM) or PF-477736 (1.0 µM) or were left untreated for 12, 24 or 48 hours. Cells were fixed in 4% PFA (15 minutes on ice) and the coverslips were subsequently washed in PBS. Cells were subsequently permeabilized on ice for 10 minutes in cytoskeleton buffer (100 mM piperazine-N,N′-bis[2-ethanesulfonic acid], PIPES [pH 6.8], 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 1 mM EDTA, 0.5% Triton-X100) and incubated with cytoskeleton stripping buffer (10 mM Tris-HCl [pH7.4], 10 mM NaCl, 3 mM MgCl 2 , 2% Tween 20, 0.5% sodium deoxycholate) under the same conditions. Samples were then blocked in 5% bovine serum albumin (BSA); 2% normal goat serum (NGS), 0.01% Triton X100 in PBS and incubated overnight at 4 °C with a primary antibody detecting γH2AX (Abcam ab22551, 1:250). After 18 hours of incubation, we executed a 1 hour incubation with a secondary antibody goat-anti-mouse Alexa488 (Invitrogen, A21121, 1:600) followed by several washes. Finally, we mounted cells with ProLong Gold antifade reagent with DAPI (Life Technology, P36935), analyzed them through microscopy, using an AxioImager M1 (Carl Zeiss) and counted the γH2AX-positive cells.