An EGFR-Induced Drosophila Lung Tumor Model Identifies Alternative Combination Treatments Free

Lung cancer is the leading cause of cancer-related deaths, with more than 1 million newly diagnosed cases worldwide each year (1). Non–small cell lung cancers (NSCLC), and in particular adenocarcinomas, account for the majority of all lung cancers. A relatively small number of oncogenic driver mutations are responsible for both the initiation and maintenance of NSCLCs (2–4). One of the most important driver oncogenes in lung cancer is the EGFR. Somatic mutations in the EGFR gene occur in 10% of all NSCLCs (5). Although, targeted therapy for patients that carry a mutation in the EGFR gene has been successful, the risk to develop drug resistance during this treatment is still around 50% (6) and the occurrence of novel mutations in resistant tumors demanded for the development of third and fourth-generation EGFR inhibitors. Therefore, novel compounds or, more importantly, synergistically acting combinations of compounds are urgently needed as alternative therapeutic strategies. They hold the potential to interfere with simultaneously activated pathways and thereby eventually improve the effectivity of a treatment. To identify such compounds, whole-animal screens, especially those based on simple model organisms that are amenable to high-throughput screening such as the fruit fly Drosophila melanogaster could develop into invaluable tools to complement the drug discovery pipeline. The human lung and the fly's tracheal system share a large number of structural and physiologic similarities (7, 8). Moreover, homologous molecular factors shape the lung and the tracheal system during development (9). These similarities also apply for airway epithelial cells. In addition, the organization of the epithelial innate immune systems of Drosophila and mammals is surprisingly similar (10, 11). These commonalities, in combination with the unrivaled freedom of genetic manipulation possible in Drosophila (12), gave rise to highly useful Drosophila models for chronic inflammatory diseases of the lung, including asthma, COPD, and lung cancer (7, 13–15).

Here, we developed and used a Drosophila lung tumor model by targeted ectopic expression of a constitutively active isoform of the EGFR in the airway epithelium of flies. The excessive tissue growth that was seen upon this treatment resulted in early larval death, which was used as read-out for a high-throughput screening approach based on rescuing this lethality. We identified the small compounds afatinib, gefitinib, and ibrutinib that rescued this lethal phenotype. A screening designed to find compounds that act synergistically with afatinib, identified bazedoxifene, a compound that specifically interferes with JAK/STAT signaling. This finding implies that the combined targeting of tyrosine kinases and JAK/STAT signaling could be a potent strategy for future cancer therapies.

For ectopic expression specifically in the airway epithelium, the Gal4/UAS system was used (16). Experiments were performed on concentrated medium (CM) [5% (w/v) yeast extract (Becton Dickinson), 5% (w/v) sucrose, 8.6% cornmeal, 0.5% (w/v) agar, 0.03% (v/v) propionic acid, 0.3% (v/v) methyl-4-hydroxybenzoate], drug experiments were performed on low-melt CM medium (CM with low-melt agarose). Transgenic fly strains used for crosses: control line w¹¹¹⁸ (BDSC_5905), tracheal driver lines for trachea ppk4-Gal4 (10, 11) and btl-Gal4, UAS-gfp; Gal80ts (Maria Leptin Lab), UAS-EGFR^CA (BDSC_9533). This mutation is characterized by the A887T substitution leading to constitutive activation of the EGFR (17, 18). The here described EGFR-induced tumor model was generated by the crossing: ppk4-Gal4>UAS-EGFR^CA (EGFR^CA), the corresponding crossing with the wild-type (ppk4-Gal4>w¹¹¹⁸) served as control. Embryos from these crossings were used for pharmacologic tests, rescue experiments, RT-PCR, and RNA-seq. All animal studies were performed according to local animal protection rules of the state Schleswig-Holstein.

Compounds from a FDA-approved library (Selleck Chemicals, L1300), dissolved in DMSO were used for drug screening. The stock solutions were diluted with ethanol and used at a final concentration of 50 μmol/L in three replicates. DMSO (0.5%)/ethanol (0.5%) served as a negative control. Constitutively active EGFR (EGFR^CA) embryos were placed on drug containing low-melt CM medium in a 96-well format and sealed with an air-permeable membrane. The number of pupae was counted after 7 days at 25°C. Positive hits were replicated. In a second screening approach, the compounds (50 μmol/L) from the library were supplemented with a low concentration of afatinib (2 μmol/L). For rescue experiments, different afatinib and gefitinib concentrations were used and the pupae and eclosed adults were counted every day.

The tracheae of early 3rd instar larvae were dissected and the dorsal trunk of Tr8 was selected for microscopy and quantification using Z-stack analysis (see above). Z-stack images were documented in DIC and DAPI channels with the Axio Imager.Z1 with Apo Tome (Zeiss). Measurements were performed using the AxioVision SE64 Rel. 4.9 Software (Zeiss). Statistical analyses were performed with GraphPad Prism7 (GraphPad).

Tracheae of early 3rd instar were dissected in HL3 buffer, fixed with 4% PFA, and blocked with 10% normal goat serum (Sigma-Aldrich). Incubation of antibodies was carried out overnight at 4°C. Staining was performed with anti-Drosophila Coracle C2 (DSHB, 1:500) antibody in combination with Alexa Fluor 488-conjugated anti-mouse IgG (Jackson ImmunoResearch, 1:300) and with anti–phospho-histone H3 (Ser10; Cell Signaling Technology, 1:100) antibodies. Chitin staining was carried out with 505 Star-conjugated Chitin-binding Probe (New England BioLabs, 1:200).

Embryos from a 10X STAT-GFP reporter line [25] were treated with 100 μmol/L bazedoxifene as described above; 0.1% DMSO served as control. Hypoxia was applied as described earlier (13).

Trachea from control and EGFR^CA 3rd instar larvae were manually dissected and the qRT-PCR experiment was performed as described previously (19). rpl32 served as control.

RNA isolation was performed as described above (section qRT-PCR). RNA-seq libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (New England BioLabs) as per the manufacturer's recommendations. Size and quality of the libraries were visualized on a Bioanalyzer High Sensitivity DNA Chip (Agilent Technologies). Diluted libraries were multiplex-sequenced on the Illumina HiSeq 2500 Instrument (Illumina) by single-read sequencing (1 × 50 bp). Gene ontology analyses were performed with the FlyMine program package (20) and analysis of transcription factor–binding sites were done with pscan (21). The RNAseq data had been deposited at the European Nucleotide Archive with the accession number PRJEB19963.

We focused on EGFR as an oncogenic driver to initiate a tumor phenotype and induced ectopic expression of a constitutively active EGFR isoform (17) exclusively in the airway epithelium of Drosophila larvae. To address the larval tracheal system, we used two different airway-specific Gal4 driver lines, ppk4-Gal4 and btl-Gal4 (11, 13, 14). Ectopic expression of EGFR^CA by the ppk4-Gal4 driver line for a moderate EGFR^CA expression induced death not before the early 3rd instar stage (Fig. 1A and B). It is reasonable to assume that the lethality was caused by severe undersupply of oxygen. This interpretation is supported by the observation that, prior to death, these larvae left the lawn (day 4), a typical sign of oxygen deprivation (Fig. 1B; ref. 13). The control larvae left the lawn later to develop into pupae, a time point at which EGFR^CA larvae were already dead (day 6; Fig. 1A). Use of the strong tracheal driver btl-Gal4 led to death in the embryonic stage. By using the Gal4 inhibitor, Gal80, we are able to tightly regulate target gene expression to a later time point in larval development (22). By this, early death was prevented but a significant increase in tracheal mass was observed in these larvae (Fig. 1C, c′).

Dissection of the tissue from ppk4-Gal4>UAS-EGFR^CA (EGFR^CA) animals revealed a strong malformation of the whole airway structure, including thickened and shortened dorsal trunks with partially blocked airways depending on the progression of the phenotype (Fig. 1D, d′). It can be assumed that the larval death occurred because of these substantial structural changes of the airway system that prevented normal gas exchange (Fig. 1).

The chitinous inner tracheal lining, which is important for maintaining the tubular structure, was highly compressed in EGFR^CA animals (Fig. 1E, e′). Staining of the septate junction protein, Coracle, illustrated that the cell–cell borders became irregular and frayed, implying that the tightness of the septate junctions was reduced and the tissue organization was impaired because of EGFR^CA expression (Fig. 1F, f′). The epithelial layer of the airway trunks was thickened and almost completely destroyed in EGFR^CA-expressing animals (Fig. 1H), compared with the thin epithelial layer in the control animals (Fig. 1G). Moreover, the air-conducting tubes were compressed and no longer air-filled, compared with the control trachea (Fig. 1I). Expression of a chimer of the extracellular human EGFR and the cytoplasmic domain of the Drosophila EGFR led to a similar set of phenotypes with the substantial epithelial thickening as the most distinctive one (Fig. 1J).

To quantify this hyperplasia of targeted EGFR^CA expression in the tracheal system, we measured and quantified two parameters–the epithelial thickness and the number of nuclei per section of the dorsal trunk as a value that shows if cell divisions occurred (Fig. 2A–E). Trachea from larvae that experienced EGFR^CA-expression had a significantly increased epithelial thickness that was seen throughout the entire tracheal system (8 ± 2.7 μm; Fig. 2A, a′, D), and an increased number of enlarged nuclei (22.5 ± 6.6; Fig. 2B, b′, E) observed in well-defined sections of the dorsal trunk, compared with matching control (5.9 ± 2.6 μm and 18.2 ± 2.5). In addition to these phenotypes, staining of the trachea with anti–phospho-histone H3 revealed substantial DNA synthesis in almost all airway epithelial cells (Fig. 2C, c′). To understand the molecular alterations that underlie these structural changes, we performed a RNA-seq analysis of manually isolated control trachea and those that express the EGFR^CA. We identified 965 genes with a statistically significant increased abundance in transcript levels and 474 genes with a reduced abundance in EGFR^CA-expressing trachea (Supplementary Table S1). An analysis of the differentially regulated genes revealed enriched pathways that include various metabolic ones including general metabolic pathways (P = 5.4E-6), sphingolipid metabolism (P = 1.6E-4), lysosome (P = 3.5E-3), or metabolism of xenobiotics (P = 3.8E-3) indicative for a hyperplastic transformation of these epithelial cells (Supplementary Table S2). Furthermore, we performed a Gene Ontology (GO) analysis with these differentially transcribed genes and found that molecular functions of epithelial cells, especially those devoted to chitin synthesis are enriched in the group of differentially transcribed genes (Fig. 2F). This cohort of differentially transcribed genes was also analyzed regarding their transcription factor–binding sites present in their presumptive promoter sequences. With pscan (focused on −450 to +50 bp relative to the transcription start site), we found a small set of significantly enriched transcription factor–binding sites (Fig. 2G). Most interestingly, Mad, a transcription factor of the TGFβ signaling pathway was the statistically best-supported, enriched binding site (Fig. 2G).

Larvae of this specific Drosophila lung tumor model only develop to early larval stages and do not complete the development into pupal stage and adulthood (Fig. 3A). We took advantage of this lethal phenotype to screen large numbers of small compounds. Positive hits were identified by the occurrence of later developmental stages and survival of EGFR^CA-expressing animals (Fig. 3A). The pharmacologic tests were performed in 96-well format (Fig. 3B).

We screened a library containing 978 FDA-approved drugs at 50 μmol/L (Supplementary Table S3). The number of formed pupae was used as read out after 7 days of incubation (Fig. 3B). Three compounds out of the 978 (afatinib, gefitinib, and ibrutinib) were able to rescue larval lethality. Especially afatinib and gefitinib yielded impressive results in survival rate and formation of pupae and adult flies (Fig. 3E–G). The lethal phenotype induced by ectopic EGFR^CA expression in the tracheal system was completely rescued (Fig. 3C–F). The majority of the animals not only survived to the last larval stage (3rd instar), but also completed their larval life, started pupation and developed into adult flies (Fig. 3C and D). Although gefitinib treatment rescued larval lethality, the percentage of adult offspring was significantly reduced in healthy control flies, implying that the compound itself had a negative effect on development during the pupal stage (Fig. 3E and F). Afatinib, on the other hand, induced an almost complete rescue, with normal percentages of adult offspring. Higher gefitinib concentrations led to a decreased number of adult flies not only in EGFR^CA (Fig. 3F) but also in controls (Fig. 3E). The rescue effect became apparent at lower concentrations of 10 and 25 μmol/L (Fig. 3F). We used different doses of both compounds to identify the concentration ideal for rescuing the EGFR-induced larval lethality. Concentrations as low as 10 μmol/L were as effective as 100 μmol/L in rescuing development until the pupal stage (gefitinib), whereas the efficacy was reduced substantially if the concentration was reduced to 1 μmol/L or below (Fig. 3G).

Both afatinib and gefitinib could also completely reverse the structural phenotype observed in the trachea of the animals (Fig. 4), including malformation of the chitinous layer (Fig. 4A–D), cellular structure (Fig. 4a′–d′), epithelial thickness (Fig. 4a″–d″), and the nucleus number (Fig. 4a″′–d″′). With this, both meta- and hyperplasia were rescued in the treated animals, reflected by reduced epithelial thickness (gefitinib) and decreased numbers of nuclei (afatinib and gefitinib) down to a control level (Fig. 4E and F).

To identify synergistically acting compounds that can rescue the lethal phenotype together with a tyrosine kinase inhibitor (TKI) such as afatinib, a secondary pharmacologic screening of the FDA-approved library compounds was performed. A concentration of 2 μmol/L afatinib, that is too low to show a rescue by itself, was combined with each of the 978 compounds from the library. Interestingly, seven compounds, which did not show any rescue effects in the primary screening, were newly identified as positive hits. A replication of the experiment with these seven compounds revealed bazedoxifene as being significantly potent in this combination (Fig. 5A). Although the effectiveness of bazedoxifene is not as high as from the specific EGFR inhibitors, it shows an unequivocal rescue effect in combination with low dosages of the specific EGFR inhibitor afatinib (25.6 ± 17.9%) compared with the control EGFR^CA animals (2 μmol/L: 7.8 ± 3.9%; Fig. 5A).

In the fly's airways, ectopic expression of EGFR^CA was able to induce expression of the JAK/STAT ligands upd2 and upd3 (Fig. 5B), which seems to be part of a positive feedback loop (Fig. 5B). Thus, we evaluated whether bazedoxifene is able to inhibit hypoxia-induced activation of JAK/STAT signaling in the trachea of Drosophila larvae, using the 10X STAT reporter line (23). Indeed, treatment with bazedoxifene reduced this hypoxia-induced signal substantially (Fig. 5C).

In this study, we designed and evaluated an EGFR^CA-induced Drosophila model for human lung cancer that is amenable to high throughput, whole-animal compound screening approaches. This strategy is in line with other successful high-throughput-screening (HTS) assays for specific cancer subtypes based on in vivo analysis of tailored Drosophila strains (24–27). A recent study by Levine and Cagan (15) demonstrated the previously unrealized potential of personalized Drosophila models to serve as whole-animal compound screening platforms to dramatically accelerate the identification of novel lead compounds for treatment of various types of lung cancer.

We could show ectopic activation of EGFR can induce lung tumor–like phenotypes in Drosophila. The observed phenotypes, primarily hyper- and metaplasia of airway epithelial cells, are not exact copies of the histologic features of human lung cancer. The tumor-like phenotype observed in the Drosophila airway system represents a very early stage of tumor development; therefore, a direct assignment to specific human lung tumors is not possible. As the induction of this phenotype is organ-wide, lethality in larval stages results from these early steps toward tumor formation. This highly appreciated phenotype that allows to set up straight-forward screening approaches thus permits at the same time a more sophisticated characterization of the tumor type. Upon oncogene induction, the normally differentiated epithelial cells of the dorsal trunks of the airway system started to proliferate, which resulted in the death of the larvae. Furthermore, the nuclei were positive for the mitosis marker pH3 and showed striking morphologic changes. The RNA-seq analysis of airway epithelia experiencing constitutive EGFR activation supported this interpretation. The gene ontologies that were enriched in differentially regulated genes imply that cells lost their epithelial cell identity. Furthermore, Mad binding sites were enriched in promoters of differentially regulated genes, which implies that TGFβ signaling was operative. TGFβ signaling is often causally related with a hyperplastic transformation from epithelial cells into other cell identities (28, 29). The observation that EGFR pathway activation induces structural changes in the fly's airway epithelium affirms that our model could also be adapted to other oncogenes downstream of the receptor, raising the possibility of generating panels of personalized fly models for human lung cancers with different driver mutations. EGFR signaling has previously been reported to be involved in maintaining organ homeostasis during development. This highly important signaling system contributes to epithelial integrity (30) in the tracheal system and it takes a central role in controlling tube elongation during development (31).

The fast life cycle and development of the fly is beneficial and provides a fast and feasible opportunity to test drugs in a whole-animal approach. Because we started treatment with the compounds of choice during the first larval stages, full rescue is seen only if the compounds are nontoxic and do not interfere with the complex development of a multicellular organism. Moreover, animals must pass through the pupal stage, which is highly susceptible to chemically induced impairment, to give rise to normal, fully functional adults. This requirement excludes compounds that nonspecifically interfere with major developmental processes. This scenario demonstrates the superiority of whole-animal screens over cell-based assays. Moreover, the design of the screen focuses only on those compounds rescuing lethality while being well tolerated even at high concentrations. Thus it allows false negative results to occur, which is taken into account, especially as these false positive hits would be due to undesirable properties of these compounds (e.g., high toxicity).

Mutations leading to elevated activity of EGFR signaling even in the absence of cognate ligands have been detected in the corresponding cancer subtypes (5, 32, 33). These mutations induce chronically activated signaling via the EGFR pathway, leading to hyper- and metaplasia of the corresponding cell, and ultimately to cancer development. In the screen described here, we tested the compounds of a comprehensive library containing FDA-approved drugs. The mutation A887T of the Drosophila EGFR is located in the N-lobe of the tyrosine kinase domain and thus corresponds to activating mutations of the human EGFR in exons 18 and 19, such as the deletions in the 747-752 region or the G719S substitution (34). Three small compounds, which are already commonly used, were effective in rescuing the lethal phenotype and therefore validated our screening system. These lead hits belong to the class of specific TKIs. Besides the specific EGFR inhibitors, the Brutons tyrosine kinase (Btk) inhibitor, ibrutinib, turned out as positive hit. This finding is in line with recent studies showing that ibrutinib is able to interact with EGFR at higher compound concentrations (35). Despite their shared ability to rescue EGFR^CA-induced lethality and structural epithelial changes, detailed characterization of afatinib and gefitinib exhibited striking differences. Treatment with the second-generation TKI, afatinib, induced complete rescue of the phenotype, because seemingly normal adults emerged after treatment. In contrast, treatment with the first-generation TKI, gefitinib, rescued the larval lethality, but very few adults emerged. The monitored phenotype was consistent with earlier described adverse effects on adult structures (36). Differing activities of afatinib and gefitinib are also observed in human patients with lung cancer, with the former conferring longer median survival relative to the latter (37). However, the improved progression-free survival in patients, treated with afatinib, is accompanied with side effects in a higher extent.

In the second screening approach, we combined a subtherapeutic afatinib concentration with the compounds from the library and could identify additional positive hits. Bazedoxifene was confirmed to rescue the lethal phenotype in our EGFR-induced tumor model in combination with this low afatinib concentration. Bazedoxifene is approved as a selective estrogen receptor modulator and is commonly used in preventing osteoporosis. Recently, it was discovered as a novel promising GP130/STAT3 pathway signaling inhibitor (38) and considered for treatment in pancreatic cancer (39). Although the combined inhibition of EGFR- and JAK/STAT signaling has already been shown to represent a valuable strategy for a number of different cancer entities (40, 41), its application for therapies is still limited. Such a cross-talk was found to be operative in an epithelial tumor model (42) and studies employing Drosophila found a robust interaction between both signaling systems (43). The interaction between EGFR and STAT signaling often employs cytokines of the JAK/STAT signaling pathway that are released upon EGFR pathway activation (44). A similar situation is found in our model, where we observed a massive upregulation of upd2 and upd3, which both code for ligands of the JAK/STAT signaling pathways. Synergistically targeting EGFR and JAK/STAT pathway for cancer therapy is intriguing in two ways: (i) because the IL6, or more generally the gp130 signaling pathway correlated with metastasis and epithelial–mesenchymal transition in lung cancer (45), this metastatic behavior might be dampened, and (ii) IL6-mediated signaling plays an important role in the development of treatment resistance, which can be specifically targeted by a combined approach (44, 46, 47). Thus, the combination of a TKI and a JAK/STAT inhibitor could be a promising and potential drug in treating lung cancer (48). The observation that EGFR pathway activation exerts some of its effects via the JAK/STAT signaling pathway is in line with the positive effects of the afatinib/bazedoxifene combination observed in this study. Thus, targeting both, the EGFR and the JAK/STAT pathway at the same time would be a valuable strategy for the treatment of NSCLC with an EGFR mutation. With the here established EGFR^CA-induced tumor model, we offer another possibility to use Drosophila in a HTS approach to identify unknown compounds with inhibiting effect on EGFR. Furthermore, it holds the promising potential to newly identify FDA-approved compounds that act synergistically with TKIs. Thereby, TKIs could be used in low concentrations with fewer side effects while exerting the full inhibitory potential by intervening with other pathways to eventually improve clinical outcome and success for patients with lung cancer.

No potential conflicts of interest were disclosed.

Conception and design: J. Bossen, R. Pfefferkorn, C. Fink, H. Heine, T. Roeder

Development of methodology: J. Bossen, K. Uliczka, R. Pfefferkorn, Mandy M.-Q. Mai, L. Burkhardt, H. Heine, T. Roeder

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Bossen, K. Uliczka, L. Steen, L. Burkhardt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Bossen, K. Uliczka, Mandy M.-Q. Mai, M. Spohn, H. Heine

Writing, review, and/or revision of the manuscript: J. Bossen, K. Uliczka, R. Pfefferkorn, I. Bruchhaus, H. Heine, T. Roeder

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Bossen, R. Pfefferkorn

Study supervision: T. Roeder

The work was funded by grants of the Bundesministerium für Forschung und Bildung as part of program “Alternativen zum Tierversuch” (031L0110A - DroLuCa) and the Deutsche Forschungsgemeinschaft (DFG) as part of the Exc Inflammation at Interfaces to T. Roeder. We would like to thank Britta Laubenstein and Christiane Sandberg for excellent technical assistance as well as the Bloomington Stock Center, Mike Welsh (Iowa), and Maria Leptin (Heidelberg, Germany) for flies.

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.