This study was carried out in accordance to the recommendations of the Federation of European Laboratory Animal Science Association. The protocol was approved by the local Ethics Committee of Animal experiments and the Landesamt für Natur, Umwelt und Verbraucherschutz of North Rhine-Westphalia in Germany (reference number: 84-02.04.2015.A199). Six to eight weeks old, genetically engineered, male and female mice with C57BL/6J lineage background were anesthetized with Ketamin/Xylazin (100 mg/kg/(body weight) BW i.p./0.5 mg/kg/BW i.p.) and 1 × 10⁷ pfu Adeno-Cre was applied intratracheally [14]. Viral vectors were provided by the University of Iowa Viral Vector Core (http://www.medicine.uiowa.edu/vectorcore). Prior to tumor induction, the mice were genotyped. We crossed Kras ^LSL-G12V mice, commonly called RASLO, which express the Kras G12V mutant after Cre-recombinase exposure [13] (Supplementary Figure S1a) with the conditional Cre-inducible Notch1 knock-out model (Notch1 ^tm1Agt, commonly called Notch1 ^fl/fl) [15] (Supplementary Figure S1b), kindly provided by Freddy Radtke and colleagues. Eight weeks after Cre-application, mice were sacrificed by cervical dislocation and lung tissue was harvested and formalin fixed and paraffin embedded (FFPE). In total, seven mice per group have been analyzed and are indicated as follows: R+ N1w/w (n = 7) harboring mutated Kras driven tumors with wild type Notch1 alleles, R+ N1f/w (n = 7) harboring mutated Kras driven tumors with heterozygous Notch1 knock-out and R+ N1f/f (n = 7) harboring mutated Kras driven tumors with homozygous Notch1 knock-out.

Genotyping of mice was performed using PCR (Supplementary Figure S1c) in 25 µl total volume according to the manufacturer’s protocol (New England Biolabs, Frankfurt, Germany) using 2 µl of DNA, 0.2 µl Hot Start Taq polymerase (5000 U/ml) 2.5 µl Standard Taq Reaction buffer (10x), 0.5 µl dNTPs (10 mM Mix), 0.5 µl Primer fwd (10 µM), 0.5 µl Primer rev (10 µM), and 18.8 µl DNase/RNase free water. The following primers were used: Ras-fwd 5′CAG​TGC​AAT​GAG​GGA​CCA​GT3′, Ras-rev 5′CAC​CCT​GTC​TTG​TCT​TTG​CTG​ATG3′, Notch1-fwd 5′ CTG​AGG​CCT​AGA​GCC​TTG​AA3′ and Notch1-rev 5′TGT​GGG​ACC​CAG​AAG​TTA​GG3′. Eartag material was lyzed using 100 µl lysis buffer (25 mM NaOH, 0.5 mM EDTA) at 95°C for 1 h and neutralized using 100 µl 40 mM Tris-HCl. The Notch1 PCR indicates whether both Notch1 alleles were floxed (500 bp product) as prerequisite for conditional Notch1 knock-out. The wild type allele which was not floxed revealed a 445 bp product. Thus, in a Notch1 flox/wild type condition, a double band was produced (see control R+ N1f/w). The Ras PCR (280 bp) detected the Kras ^LSL-G12V expression construct enabling tumor induction. Excision of the loxP flanked stop cassette was mediated by intratracheal inhalation of adenoviral Cre-recombinase as described above.

Three micrometre sections from FFPE lung tissue, were deparaffinized and treated according to standard protocols of the routine diagnostics pipeline (Institute for Pathology, University Hospital Cologne, Germany). Staining was performed using hematoxylin & eosine (H&E) and primary antibodies against KI67 (SP6, 1:100, Cell Marque, Rocklin, United States), NICD1 (ab8925, 1:50, Abcam, Cambridge, United Kingdom), SPC (AB3786, 1:100, Millipore, Burlington, United States), CC10 (07-623, 1:250, Millipore) and TAZ (H-70, 1:100, Santa Cruz Biotechnology, Dallas, United States) and the secondary Histofine Simple Stain antibody detection kit (Medac, Wedel, Germany). The LabVision Autostainer 480S (Thermo Fisher Scientific, Waltham, United States) was used for visualization and slides were scanned by the Pannoramic 250 slide scanner (3D Histech, Budapest, Hungary). Murine tumors were analyzed according to the recommendations of mouse models of the human cancers consortium [16]. Protrusions from bronchiolar lining and clear bronchiolar lumen without protrusions have been determined in R+ N1w/w (n = 7), R+ N1f/w (n = 7) and R+ N1f/f (n = 7) mice by three independent H&E stained transverse sections of the lower lung periphery from each of the FFPE lung biopsies.

TAZ protein expression was scored by IHC in murine tumor samples and in clinical samples combined in tissue micro arrays (TMAs). Thirty-eight squamous cell carcinomas (SCC), 37 adenocarcinomas (ADC) and 32 small cell carcinomas (SCLC) of patients have been investigated. TAZ expression in the nucleus or the cytoplasm was rated as positive if more than 10% of tumor cells exhibited expression. The TAZ score was determined by two independent investigators and graded on an integer number four-grade scale (0: no expression, 1: weak expression, 2: moderate expression, 3: strong expression). The TAZ score was obtained by adding the nuclear and the cytoplasmic staining intensity, the nuclear multiplied with 0.75, the cytoplasmic multiplied with 0.25, in order to have a measure of the TAZ activation status [17].

Lung cancer cell lines were kindly provided by Roman K. Thomas (Department of Translational Genomics, University of Cologne, Germany) and Reinhard Büttner (Institute for Pathology, University Hospital Cologne, Germany). Cells were cultivated in RPMI medium (Life Technologies, Carlsbad, United States) supplemented with 10% FCS (PAA laboratories, Cölbe, Germany). WWTR1/TAZ and NOTCH1 siRNAs (100 nM, Santa Cruz Biotechnology) were transfected using Lipofectamine 3000 (Life Technologies). Plasmid transfection and vector control were performed as previously described [18]. pTight-hASCL1-N174 [19] was a gift from Jerry Crabtree (Addgene plasmid #31876; http://n2t.net/addgene:31876; RRID:Addgene_31876). Cell proliferation was determined by Cell Proliferation kit I [(MTT) assay (Hoffmann-La Roche, Basel, Switzerland)]. MTT assays were performed in 96-well plates with 3,000 cells per well. Cells have been seeded after validated TAZ knock-down in medium containing tetrazolium salt, the reagent which enables the colorimetric detection of proliferating cells. Tetrazolium salt is cleaved to the water insoluble dye formazan by succinate-tetrazolium reductase system. This enzyme is only active in viable cells. The cell proliferation was determined after 24, 48 and 72 h, by measuring the absorbances at A550/690 nm using a plate reader. The protocol was performed according to the manufacturer’s instructions.

For the generation of cell pellet blocks, 1 × 10⁷ cells were centrifuged and fixed in 4% paraformaldehyde (PFA) over night. PFA was removed by centrifugation for 15 min at 2400 rpm and cells were washed with 2 ml 96% ethanol with three drops of glycerin. Cells were again centrifuged for 15 min at 2400 rpm and transferred to tissue capsules and were further processed for paraffin embedding by standard FFPE biopsy protocol.

One microgram of total RNA were transcribed into cDNA using the Super Script™ III, H Reverse transcriptase and Oligo(dt)12-18 Primer (Life Technologies). SYBR Green (Qiagen, Hilden, Germany) was used for qRT-PCR. Used primers were directed against NOTCH1 (forward 5′-GTC​AAC​GCC​GTA​GAT​GAC​C-3′ and reverse 5′- TTG​TTA​GCC​CCG​TTC​TTC​AG-3′), ASCL1 (forward 5′-GCA​TGG​AAA​GCT​CTG​CCA​AGA-3′ and reverse 5′- TGG​CAA​AGA​AAC​AGG​CTG​CG-3′), WWTR1/TAZ (forward 5′-GAA​GTC​CAT​CCC​CTT​CTG​GT-3′ and reverse 5′-CAA​GCA​GAG​AAT​GAG​GGG​AA-3′) and GAPDH (forward 5′-ACT​GCC​AAC​GTG​TCA​GTG​GT-3′ and reverse 5′-GTC​AAA​GGT​GGA​GGA​GTG​G-3′). Primers were derived from the qPrimerDepot of Wenwu Cui and synthetized by Sigma-Aldrich. The 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, United States) and the SDS2.2 software (Applied Biosystems) were used. Expression levels were calculated using ∆∆CT method.

Clinical samples were acquired and analyzed as part of routine diagnostic procedures (Institute of Pathology, University Hospital Cologne, Germany) according to the guidelines and with approval of the local ethics committee (reference number: 13-091).

In addition, the datasets analyzed during the study are available in the TCGA repository, accessed by cBioportal [20, 21] webpage https://www.cbioportal.org/. Particularly, a data set of 230 lung ADC samples [22] was accessible under: https://www.cbioportal.org/results/oncoprint?Action=Submit&RPPA_SCORE_THRESHOLD=2.0&Z_SCORE_THRESHOLD=2.0&cancer_study_list=luad_tcga_pub&case_set_id=luad_tcga_pub_3way_complete&data_priority=0&gene_list=NOTCH1%2520KRAS&geneset_list=%20&genetic_profile_ids_PROFILE_COPY_NUMBER_ALTERATION=luad_tcga_pub_gistic&genetic_profile_ids_PROFILE_MUTATION_EXTENDED=luad_tcga_pub_mutations&profileFilter=0&tab_index=tab_visualize.

Moreover, a dataset of 178 lung SCC samples [23] was used, accessible under: https://www.cbioportal.org/results/oncoprint?Action=Submit&RPPA_SCORE_THRESHOLD=2.0&Z_SCORE_THRESHOLD=2.0&cancer_study_list=lusc_tcga_pub&case_set_id=lusc_tcga_pub_cnaseq&data_priority=0&gene_list=NOTCH1%2520KRAS&geneset_list=%20&genetic_profile_ids_PROFILE_COPY_NUMBER_ALTERATION=lusc_tcga_pub_gistic&genetic_profile_ids_PROFILE_MUTATION_EXTENDED=lusc_tcga_pub_mutations&profileFilter=0&tab_index=tab_visualize.

The datasets are available in the TCGA repository, accessed by cBioportal [20, 21] webpage https://www.cbioportal.org/.

Statistical analysis was performed with Graph Pad Prism (STATCON, Witzenhausen, Germany) and SPSS (IBM Corp., Armonk, United States). The two-sided Student’s t-test and the Pearson Correlation Coefficient were used to analyze data for significant differences and correlations. Error bars indicate standard error of the mean (SEM). p-values <0.05 were regarded as significant and indicated in the figures as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Mice, carrying the conditional Kras ^LSL-G12V expression construct [13] were crossed with conditional Notch1 knock-out mice [15]. Hence, both, the expression of the Kras mutant and the Notch1 knock-out depend on adenoviral Cre-recombinase application. Eight weeks after intratracheal inhalation of adenoviral Cre-recombinase, lung tissues were harvested (Figure 1A). The size of single tumor lesions was determined and tumors were analyzed histologically (Figure 1B). Immunohistochemical analysis of the intracellular domain of Notch1 (NICD1), shown to be the active Notch1 signaling component, confirmed that Notch1 was efficiently knocked out in the Kras mutant derived tumors (Figure 1C). There was no significant difference in tumor size observed between the mutated Kras driven tumors with wild type Notch1 alleles (R+ N1w/w), the mutated Kras driven tumors with heterozygous Notch1 knock-out (R+ N1f/w) and the mutated Kras driven tumors with homozygous Notch1 knock-out (R+ N1f/f).

Morphologically, the tumors presented as papillary ADCs and showed cuboidal cell shape in R+ N1w/w mice harboring active Notch1 signaling. Notch1 deficiency resulted in a mosaic-like pattern and complete loss of NICD1 in R+ N1f/w and R+ N1f/f mice, resulted in papillary ADCs comprised of cuboidal and columnar tumor cells, respectively (Figure 1C).

The determined tumor area upon Notch1 knock-out was slightly decreased from 24.68 to 17.90% (p = 0.1672) as well as tumor cell proliferation, indicated by the KI67 index, from 21.00 to 17.86% (p = 0.3034) in a dose-dependent manner, though not significantly, comparing R+ N1w/w and R+ N1f/f mice (Figures 2A–C). Intriguingly, tumor cell accumulation inside the bronchiolar lumen with proliferating cells which protrude from the bronchiolar lining, was significantly increased both upon the loss of one Notch1 allele [from 9.82 to 28.40% (p = 0.0390)] and the loss of two Notch1 alleles [from 9.82 to 40.34% (p = 0.0020)] (Figure 2D). The cell lineage markers, Surfactant Protein C (SPC) and Clara Cell 10 KDa Secretory Protein (CC10) were used to identify the differentiation of dysplastic cells accumulations inside the bronchiolar space. SPC+ cells representing alveolar type II cells, were located in the alveolar periphery (arrow 1) and CC10 + Clara cells lined the bronchiolar walls (arrow 2). ADCs which grew predominantly peri-bronchiolarly were composed of SPC+ tumor cells (arrow 3). In R+ N1w/w mice, bronchiolar airways were rarely invaded by SPC+ tumor cells whereas in R+ N1f/f mice, SPC+ cells which were located within the bronchiolar lining formed epithelial carcinoma in situ lesions protruding from the bronchiolar wall (arrow 4) and budding into the bronchiolar lumen (arrow 5) (Figure 2E).

Taken together, these data suggest that conditional Notch1 deficiency in this autochthonous Kras ^LSL-G12V driven lung cancer mouse model induces papillary ADCs accompanied by SPC+ tumor cell accumulation inside the bronchiolar lumen, which is not found under Notch1 wild-type conditions.

To determine whether Notch1 deficiency due to genomic aberrations can also be observed in human mutated KRAS driven ADCs, we used TCGA database and investigated the publicly available lung ADC dataset containing 230 cases [22] for genomic aberrations in KRAS and NOTCH1. We found ten genomic alterations in NOTCH1 (4.3%) (Table 1), of which five were putative loss-of-function aberrations (2.15%). One of the NOTCH1 alteration was annotated as an oncogenic likely loss-of-function frame shift insertion. The other four of the NOTCH1 alterations were missense mutations in domains coding for the extracellular EGF-like repeats, which may result in interference with Notch ligand binding and reduce NOTCH1 signaling [18]. In four out of five cases harboring putative NOTCH1 loss-of-function alteration, KRAS was also altered by genomic amplification and mutation (Table 1). In 178 cases of SCCs obtained from TCGA database [23], genomic aberrations in NOTCH1 and KRAS are mutually exclusive and do not co-occur (Supplementary Table S1).

Taken together, this data suggests that Notch1 deficiency by putative inactivating NOTCH1 alterations occurs together with KRAS aberrations particularly in human ADCs of the lung, thus giving clinical importance to our findings.

Since TAZ and Notch signaling cooperate in epidermal cell fate decisions [10], we aimed to decipher whether upon Notch1 deletion TAZ expression is altered in mutated Kras driven ADCs. ASCL1 is a master regulator in Notch signaling [18] and is used as a positive control for Notch1 deficiency in the following analyses.

We investigated TAZ expression in R+ N1w/w, R+ N1f/w and R+ N1f/f mice by IHC and determined a total protein expression score per single lesion. We found that partial and full Notch1 knock-out in mutated Kras driven ADCs increased TAZ expression significantly (t-test, two-sided, p = 0.0012). In addition, Pearson Correlation analysis revealed a weak correlation of moderate and high TAZ expression to partial and full Notch1 knock-out R = 0.27231; p = 0.00029) (Figure 3). Similar results were obtained in vitro using the ADC cell line PC9 which has NOTCH1 expression and low levels of ASCL1 (Supplementary Figure S2a). ASCL1 is negatively regulated by NOTCH1 [18]. Thus, PC9 cells showed increased ASCL1 and WWTR1/TAZ expression upon NOTCH1 knock-down (Supplementary Figure S2b). Upon plasmid derived ASCL1 expression in PC9 cell clones we detected WWTR1/TAZ up-regulation, as well (Supplementary Figure S2c). These data show that TAZ expression is regulated by Notch1.

Furthermore, we investigated TAZ expression in cell blocks of human lung carcinoma cell lines, representing the most frequent lung carcinoma entities, including four SCC cell lines (H226, HCC15, HCC95, H1703), four ADC cell lines (H1975, PC9, H441, A549) and four SCLC cell lines (GLC8, GLC1, N417, DMS114). All SCC cell lines exhibited high TAZ expression, among the ADC cell lines, A549 and H441 cells, both harboring KRAS mutations, showed the highest TAZ expression. Among the SCLC cell lines only DMS114 cells showed TAZ expression (Figures 4A–C). Knock-down of TAZ in H1703, A549 and DMS114 cells using three pooled target specific siRNAs significantly reduced tumor cell proliferation after 72 h determined by MTT assay (Figures 4D–F). In the cell lines GLC8 and GLC1, that showed only marginal TAZ expression, TAZ knock-down did not affect tumor cell proliferation (Supplementary Figure S3). Thus, we hypothesize that TAZ protein expression correlates to the inhibitory effect of TAZ knock-down on cell proliferation and that TAZ protein expression might serve as a predictive marker for treatment with TAZ signaling inhibitors.

To show whether this TAZ distribution also applies to clinical patient samples, we investigated TAZ expression in lung carcinoma samples acquired during routine diagnostics. Patients’ lung SCCs, ADCs and SCLCs were analyzed by IHC using TMAs with regard to the increased TAZ expression and nuclear TAZ localization (Figure 5; Table 2; Supplementary Figure S4) because the nuclear translocation of TAZ indicates activation of the Hippo signaling pathway [24]. Within the clinicopathologic parameters we found no association of TAZ expression to age and gender (Table 2).

Importantly, high total and nuclear TAZ protein expression was detected in ADCs, associated to the SCC entity using Pearson Correlation (R_total = 0.66464; R_nucleus = 0.5999; p < 0.0001), but less present in the SCLC cohort (Figures 5C, D; Table 2). Addressing TAZ expression and survival in publicly available TCGA data [22, 23], WWTR1/TAZ amplification served as a negative prognostic factor in ADCs (Logrank test; p = 0.0533), but had no impact on survival in SCCs (Logrank test; p = 0.988) (Supplementary Figure S5).

This data demonstrates that TAZ expression is found in SCCs, ADCs and SCLCs of the lung, serves as a negative prognostic factor in ADCs and is mainly associated to the SCC entity. Interestingly, TAZ expression which is related to a poor prognosis in ADCs, might be regulated by NOTCH1 deficiency.