FOXM1 activates AGR2 and causes progression of lung adenomas into invasive mucinous adenocarcinomas

Lung cancer remains one of the most prominent public health challenges, accounting for the highest incidence and mortality among all human cancers. While pulmonary invasive mucinous adenocarcinoma (PIMA) is one of the most aggressive types of non-small cell lung cancer, transcriptional drivers of PIMA remain poorly understood. In the present study, we found that Forkhead box M1 transcription factor (FOXM1) is highly expressed in human PIMAs and associated with increased extracellular mucin deposition and the loss of NKX2.1. To examine consequences of FOXM1 expression in tumor cells in vivo, we employed an inducible, transgenic mouse model to express an activated FOXM1 transcript in urethane-induced benign lung adenomas. FOXM1 accelerated tumor growth, induced progression from benign adenomas to invasive, metastatic adenocarcinomas, and induced SOX2, a marker of poorly differentiated tumor cells. Adenocarcinomas in FOXM1 transgenic mice expressed increased MUC5B and MUC5AC, and reduced NKX2.1, which are characteristics of mucinous adenocarcinomas. Expression of FOXM1 in KrasG12D transgenic mice increased the mucinous phenotype in KrasG12D-driven lung tumors. Anterior Gradient 2 (AGR2), an oncogene critical for intracellular processing and packaging of mucins, was increased in mouse and human PIMAs and was associated with FOXM1. FOXM1 directly bound to and transcriptionally activated human AGR2 gene promoter via the -257/-247 bp region. Finally, using orthotopic xenografts we demonstrated that inhibition of either FOXM1 or AGR2 in human PIMAs inhibited mucinous characteristics, and reduced tumor growth and invasion. Altogether, FOXM1 is necessary and sufficient to induce mucinous phenotypes in lung tumor cells in vivo.

Introduction Lung cancers are classified into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC). Adenocarcinoma, the most common subtype of NSCLC, is a complex disease harboring activating mutations in KRAS (30%), EGFR (15%), or ALK (5%) genes [1]. Recently, specific and effective receptor tyrosine kinase inhibitors have been generated to treat patients with EGFR and ALK mutations. Efforts to pharmacologically inhibit oncogenic KRAS, however, have been largely unsuccessful and developing targeted therapies for KRAS-driven lung cancer remains a significant challenge. While the majority of human NSCLCs have robust expression of NKX2.1, a homeobox transcription factor which is routinely used as a marker of NSCLC [2], a subset of the KRAS-driven tumors was identified with reduced expression of NKX2.1 [3,4] associated with mucinous characteristics and poor prognosis in NSCLC patients [5]. NKX2.1 functions as a tumor suppressor in KRAS-driven mucinous adenocarcinomas, but has an oncogenic role in EGFR mutated lung tumors [3]. Haploinsufficiency of Nkx2.1 in mice induced Kras G12D -mediated lung tumorigenesis and increased production of mucins in Kras G12D -driven lung tumors [3]. While NKX2.1 represses mucinous differentiation in NSCLCs, transcriptional activators of mucinous phenotype remain unknown.
Forkhead Box M1 (FOXM1) is a transcription factor activated by the RAS/ERK signaling pathway [6][7][8]. Activation of RAS-ERK drives cell cycle progression by regulating the temporal expression of cyclin regulatory subunits that bind to and activate their corresponding cyclindependent kinases (CDK). CDK2/Cyclin E and CDK1/Cyclin B complexes phosphorylate a variety of cell cycle regulatory proteins, including FOXM1, promoting G 1 /S and G 2 /M transitions [9,10]. Activated ERK1/2, PLK1, CDK4 and CDK6 also phosphorylate FOXM1 and are required for full transcriptional competency of the FOXM1 protein [11,12]. FOXM1 transcriptionally activates cell cycle regulatory genes critical for DNA replication and progression into mitosis, including Cyclin B1, PLK1, JNK1, CDC25B, TOPO2 and Aurora B [13][14][15][16]. FOXM1 is required for KRAS/ERK signaling during lung morphogenesis, since deletion of Foxm1 prevented defects in branching lung morphogenesis caused by Kras G12D [6]. Deletion of Foxm1 from respiratory epithelial cells blocked tumorigenesis by oncogenic Kras G12D [8], and decreased the tumor initiation and growth of Kras-mutations-associated chemically-induced mouse lung tumors [17] [18]. In humans, increased FOXM1 was correlated with higher grades of lung cancers and poor patient survival [19]. While the role of FOXM1 in lung tumor initiation and growth is well established, it is unclear whether FOXM1 enhances progression from adenomas to adenocarcinomas and regulates metastasis of lung tumors in vivo.
In the present study, we used a doxycycline-inducible transgenic mouse model to express an activated FOXM1 transcript in pre-existing, benign lung adenomas. FOXM1 caused progression of lung adenomas into invasive, metastatic adenocarcinomas with a mucinous phenotype. Inhibition of FOXM1 in human mucinous adenocarcinoma cells inhibited mucinous characteristics and reduced tumor invasion in an orthotopic xenograft mouse model. We have demonstrated that FOXM1 is sufficient to drive progression of adenomas to adenocarcinomas and is required for maintainance of the mucinous phenotype.

FOXM1 causes progression to invasive adenocarcinomas
To examine the role of FOXM1 in pulmonary carcinogenesis, we used a urethane-induced lung cancer model. Mice containing Spc-rtTA and tetO-GFP-FoxM1-ΔN transgenes were generated. Treatment with doxycycline (Dox) expresses an activated form of FOXM1, the FoxM1-ΔN, tagged with GFP [20]. To induce lung adenomas, the Spc-rtTA/tetO-GFP-FoxM1-ΔN double transgenic mice (epiFoxM1-ΔN) were treated with urethane ( Fig 1A). Low-grade lung adenomas were present 14 weeks after initiation of urethane [16,21], at which time epiFoxM1-ΔN mice were treated with Dox to express the FoxM1-ΔN transgene in SP-C-expressing cells and lung epithelial cells. Lungs were harvested 10 weeks later for tumor assessment. Expression of activated FOXM1 increased the number and size of tumors ( Fig 1B) and led to histologically less differentiated tumors shown with H&E ( Fig 1C). qRT-PCR analysis of micro-dissected tumors demonstrated human (transgenic) FOXM1 mRNA in epiFoxM1-ΔN lungs ( Fig 1D). FoxM1-ΔN protein was detected in epithelial cells in the tumors by immunostaining (Fig 1E).
Consistent with published studies [21][22][23], the majority of tumors in control urethanetreated mice were classified as low grades 1 and 2 non-small lung tumors (72%). Only 14% of tumors were high grades 3 and 4 lung adenocarcinomas (Fig 1F and S1 Fig). Expression of FoxM1-ΔN in low-grade adenomas increased the frequency of highgrade 4 tumors (23% vs. 7%), and caused progression into poorly differentiated adenocarcinomas (7%) which were not detected in the controls (Fig 1F and S1 Fig). All tumors in control and epiFoxM1-ΔN mice were positive for proSP-C (Fig 2A), which is consistent with previous studies and confirms the type II alveolar epithelial origin of tumor cells [16,24]. EpiFoxM1-ΔN tumors often invaded the airways as shown by proSP-C stained tumor cells in the bronchial lumen stained with CCSP (CC10) (Fig 2A). Peritoneal lymph node metastasis were found in epiFoxM1-ΔN mice which expressed lung-specific NKX2.1 (TTF-1) protein ( Fig 2B, right panel), indicating that the metastases were derived from FoxM1-ΔN-positive tumors that expressed stabilized human FOXM1 transgene [20]. Metastases were not present in control mice. Altogether, expression of activated FoxM1-ΔN in benign lung adenomas caused tumor progression into poorly differentiated, metastatic adenocarcinomas.

FOXM1 increases cellular proliferation in epiFoxM1-ΔN tumors
Since FOXM1 is often expressed in proliferating cells and activates cell cycle regulatory genes, we used immunostaining for Ki-67 to visualize proliferating cells. Tumors from epiFoxM1-ΔN mice had increased numbers of Ki-67-positive cells (Fig 3A). No differences were observed in the frequency of apoptosis, as assessed by cleaved-caspase 3 immunostaining (Fig 3B). SOX2, a marker of stem-like and less differentiated NSCLC tumor cells [25,26] and a known transcriptional target of FOXM1 [27], was increased in the epiFoxM1-ΔN tumors (Fig 3C, left panels). Sox2 mRNA was increased in micro-dissected lung tumors from epiFoxM1-ΔN mice ( Fig 3C, right panel). Thus, expression of activated FoxM1-ΔN in lung adenomas increased tumor cell proliferation and induced SOX2.

Mucinous characteristics of epiFoxm1-ΔN tumors
NKX2.1 is often used as a diagnostic marker of human NSCLCs [28], and its expression is decreased in mucinous tumor subtypes [3,4]. While lung tumors from control mice maintained robust expression of NKX2.1, the NKX2.1 staining was decreased in epiFoxM1-ΔN tumors ( Fig 4A, brown nuclei) and associated with decreased Nkx2.1 mRNA (Fig 4B). NKX2.1 was absent in a majority of FoxM1-ΔN-expressing tumor cells ( Fig 4C) and associated with following urethane treatment in control (n = 6 mice) and epiFoxM1-ΔN (n = 9 mice) mice. (C) Representative H&E staining of lung tumors from control and epiFoxM1-ΔN mice demonstrates invasive and less differentiated phenotype of epiFoxM1-ΔN tumors compared to controls. (D) Efficient expression of FOXM1-ΔN mRNA in microdissected epiFoxM1-ΔN tumors (n = 5) compared to control tumors (n = 5) shown by qRT-PCR. mRNA levels were normalized to β-actin mRNA. (E) Efficient expression of transgenic FOXM1 protein in epiFoxM1-ΔN tumors shown by immunochistochemictry (IHC) using an anti-GFP antibody (top panels) and anti-FoxM1 antibody (bottom panels). (F) Schematic diagrams of the tumor grades distribution in control and epiFoxM1-ΔN mouse lungs. 25 tumors were analyzed from 9 epiFoxM1-ΔN mice, and 23 tumors from 6 control lungs. 10 images from each mouse lung were used for analysis. A p-value <0.05 is marked with a single asterik (*) and a pvalue <0.01 is marked with a double asterik (**).

FOXM1 regulates AGR2 in mouse and human lung adenocarcinomas
To determine whether FOXM1 regulates AGR2, we used shRNA lentiviral vectors to inhibit FOXM1 expression in A549 cells, a human mucinous lung cancer cell line with a Kras G12S mutation. Inhibition of FOXM1 reduced expression of AGR2 (Fig 6D and 6E) and was associated with decreased expression of proliferation-specific Cyclins B1, E1, A1 and D1 (S2A Fig). FOXM1 depletion reduced mRNAs of goblet cell associated genes MUC5AC, MUC5B, MUC1 and SPDEF (S2B Fig), the latter is a critical transcriptional regulator of mucinous phenotype [31]. To demonstrate that the FOXM1 regulation of the mucinous phenotype is not limited to one cell line, we also inhibited FOXM1 in H2122 cells, another human mucinous lung adenocarcinoma with an activating mutation in Kras (Kras G12C ). Similar to findings in A549 cells, inhibition of FOXM1 reduced AGR2 mRNA and protein in H2122 cells (Fig 6D and 6E). Thus, FOXM1 is essential for AGR2 expression and the mucinous phenotype in human mucinous lung adenocarcinoma cell lines.

FOXM1 induces transcription of AGR2
Since AGR2 was increased in FoxM1-ΔN expressing mouse lung tumors and human PIMAs (Fig 6), we tested whether FOXM1 directly activates the transcription of the AGR2 gene. A potential FOXM1-binding site was found within the -257/-247 bp region of the human AGR2 gene promoter (Fig 7A). Chromatin immunoprecipation (ChIP) in A549 cells demonstrated that FOXM1 directly bound to the -257/-247 bp region of the AGR2 promoter ( Fig 7A). Next we cloned a -2.0 kb fragment of the human AGR2 promoter into luciferase (LUC)-expressing vector and co-transfected the -2.0 kb AGR2-LUC plasmid with a CMV-FOXM1 expression vector into A549 cells. FOXM1 increased the transcriptional activity of the AGR2 promoter FoxM1 causes progression of benign lung adenoma to mucinous lung adenocarcinoma Representative images of lung tissue sections from patients with PIMAs (n = 12 patients, S2 Table). Images include adjacent normal lung tissue (left panels) and tumor lesions (right panels) in matched patients. Tissue sections were stained with antibodies against FOXM1, AGR2, NKX2.1 or stained for mucus using Alcian blue (left panels). Image J software was used to quantify intensity of staining (right graphs). A minimum of 5 random 20x field images per patient were quantified. Increased FOXM1 staining in PIMAs was associated with abundant Alcian blue staining, loss of NKX2.1 and increased expression of AGR2 in tumor cells. https://doi.org/10.1371/journal.pgen.1007097.g005 FoxM1 causes progression of benign lung adenoma to mucinous lung adenocarcinoma In mouse urethane-induced lung cancer model, AGR2 was induced in epiFoxM1-ΔN tumors. Immunofluorescence staining for AGR2 in control and epiFoxM1-ΔN lung tumors (left panels, n = 6 control mice and n = 9 epiFoxM1-ΔN mice, ten images per each mouse lung) and qRT-PCR of Agr2 mRNA using RNA from microdissected control (n = 5) and epiFoxM1-ΔN (n = 5) lung tumors (right panel). β-actin mRNA was used for (Fig 7B). Site-directed mutagenesis of the FOXM1 binding site blocked the ability of CMV-FOXM1 to activate the AGR2 promoter, indicating that the -257/-247 bp AGR2 region was required for FOXM1-mediated activation of AGR2 (Fig 7B). Thus, AGR2 is a direct transcriptional target of FOXM1.
Interestingly, overexpression of AGR2 in FOXM1-defecient A459 cells rescued mRNA and protein levels of Cyclin D1, but was not able to rescue expression of other cyclins and mucinoius genes (Supplemental S3 Fig). Depletion of AGR2 alone was sufficient to decrease expression of proliferation-specific and mucin-specific genes (S3 Fig) and reduce the growth of A549 cell in vitro (S4A Fig). These data demonstrate that both FOXM1 and AGR2 are critical for cell growth and maintenance of the mucinous phenotype in PIMA cells in vitro.

FOXM1 and AGR2 maintain mucinous characteristics of lung adenocarcinomas in vivo
We used lentiviral shRNA to stably inhibit FOXM1 in A549 cells in an orthotopic xenograft tumor model. Single cell suspensions of control and FOXM1-deficient A549 cells were delivered into the tracheas of immunocompromised mice. Eight weeks after inoculation, control A549 cells formed numerous tumors while FOXM1-deficient cells formed smaller and fewer tumors (Fig 7C). FOXM1-deficient xenografts were also less mucinous, with substantially reduced cytoplasmic swelling and reduced staining with Alcian blue (Fig 7C and 7D) and associated with decreased staining of AGR2 and MUC5AC (Fig 7E). Interestingly, parental A549 cells frequently metastasized into the mediastinal lymph nodes (6 out of n = 6 mice, 100% incidence of lymph node metastasis) and into the liver (2 out of n = 6 mice, 33% incidence of liver metastasis) (Fig 7F). In contrast, FOXM1-depleted tumors did not develop liver metastasis (0 out of n = 6 mice, 0%) and had decreased metastatic potential for mediastinal lymph nodes (1 out of n = 6 mice, 17%) ( Fig 7F). In addition, control A549 tumors were locally invasive, infiltrating the pulmonary bronchioles and alveolar regions, whereas FOXM1-deficient A549 tumors were non-invasive ( Fig 7G). Similar to intra-tracheal administration, a direct inoculation of FOXM1-deficient A549 cells into the left lung lobe inhibited the tumor growth and metastasis into the liver and mediastinal lymph nodes (S4B and S4C Fig). Similar effect was found after inhibition of AGR2 (S4B and S4C Fig). Thus, knockdown of FOXM1 or AGR2 reduced tumor invasion and inhibited mucinous phenotype of human lung A549 adenocarcinomas in vivo.

Discussion
Activating KRAS mutations are associated with poor prognosis in patients with non-small cell lung cancers (NSCLCs), and KRAS mutant NSCLC tumors are often resistant to common anti-cancer drugs. Several KRAS-regulated kinases, including ERK1/2, PLK1, CDK1, CDK2, CDK4 and CDK6 phosphorylate FOXM1, contributing to its transcriptional activation. While FOXM1 alone was insufficient to induce lung tumors in transgenic mice [32], genetic deletion of Foxm1 from mouse respiratory epithelium inhibited the initiation of lung tumorigenesis by Kras G12D [8]. These results indicate that FOXM1 functions downstream of oncogenic KRAS to induce lung tumorgenesis. Deletion of Foxm1 from fetal mouse lungs prevented the effects of normalization. A p-value <0.01 is marked with a double asterik (**). (C) In mouse Kras G12D -induced lung cancer models, AGR2 was induced in lung tumors of Kras G12D / epiFoxM1-ΔN mice (n = 3) compared to control Kras G12D mice (n = 3). (D) FoxM1 is required for AGR2 expression in PIMAs. Human PIMA cell lines A549 and H2122 were stable transduced with control (scramble) or shRNAs against human FoxM1 (shFoxM1 #1 or shFoxM1 #2). Efficient inhibition of FoxM1 decreased Agr2 expression in both A549 (top) and H2122 (bottom) PIMA cells, shown with qRT-PCR. β-actin mRNA was used for normalization. A p-value <0.01 is marked with a double asterik (**). (E) Western blot shows the correlation of the loss of FOXM1 and AGR2 in the A549 (top) and H2122 (bottom) cell lines. https://doi.org/10.1371/journal.pgen.1007097.g006

Fig 7. FOXM1 increases transcription of AGR2 and is required to maintain mucinous phenotype in PIMAs. (A)
Schematic of a potential FOXM1 binding site in the AGR2 promoter (left panel). ChIP shows binding of FOXM1 to the AGR2 promoter in A549 cells (right panel). A549 cells were fixed, lysed, sonicated, and used for immunoprecipitation with an antibody against FOXM1 or rabbit control IgG. PCR was performed encompassing a predicted FOXM1 binding site from -257bp to -247bp of the human AGR2 promoter. DNA region located at -6.0 kb of Agr2 promoter was used as a negative control. (B) FOXM1 transcriptionally activates AGR2 promoter. Sequence denotes the predicted binding motif and the sequencing results following activated Kras G12D during lung development [6], demonstrating an important role for FOXM1 downstream of the KRAS/ ERK signaling cascade. In the present study, we found that FOXM1 is necessary and sufficient to induce mucinous characteristics in lung adenocarcinomas induced by either urethane or Kras G12D . The FoxM1-ΔN-expressing mouse tumors resembled human pulmonary invasive mucinous adenocarcinomas (PIMAs), an aggressive subtype of NSCLC associated with activating mutations in KRAS [1]. Consistent with these findings, we detected robust expression of FOXM1 in human PIMAs. Since FOXM1 is required for KRAS/ERK signaling in mouse tumor models, our studies provide a rationale for pharmacological targeting FOXM1 in PIMA tumors with activating KRAS mutations.
We found that transgenic expression of activated FoxM1-ΔN increased cell proliferation and tumor growth, findings consistent with the known role of FOXM1 in activation of cell cycle regulatory genes [13,14]. Expression of FOXM1 in Rosa26-FoxM1 transgenic mice accelerated proliferation of tumor cells and increased the number and size of lung adenomas after tumor initiation/ promotion with 3-methylcholanthrene (MCA)/butylated hydroxytoluene (BHT) [15,32]. Likewise, genetic deletion of the Foxm1 gene from adult mouse respiratory epithelium inhibited tumorigenesis caused by MCA/BHT or urethane [17]. In the present study, we observed abundant extracellular mucin and reduced NKX2.1 expression in FoxM1-ΔNexpressing tumors, features characteristic of human PIMAs [5]. This unexpected finding was further investigated using PIMA cell lines in which FOXM1 expression was required for expression of MUC5AC, MUC5B, MUC1, and the mucin-associated disulfide isomerase, AGR2 in vitro and in an orthotopic xenograft mouse model. Our data are consistent with published studies demonstrating that deletion of Foxm1 from respiratory epithelial cells or pharmacological inhibition of FOXM1 by RCM-1 compound reduces MUC5AC and prevents differentiation of mucin-producing goblet cells in mouse asthma models [33,34].
AGR2 is an ER chaperone which functions as a proto-oncogene in breast, colon, and esophageal adenocarcinoma [30,[35][36][37]. Since AGR2 was highly expressed in the Kras-LSL G12D ; Nkx2.1-/+ mucinous lung carcinomas [3,38], we investigated AGR2 expression in a carcinogen-induced FoxM1-ΔN transgenic mouse model and human PIMAs. We identified AGR2 was highly expressed in both PIMAs and FoxM1-ΔN-expressing mouse lung tumors, as a result of direct transcriptional activation by FOXM1. Although the oncogenic functions of AGR2 remain largely unknown, AGR2 expression was demonstrated to be prognostic in human lung cancer [39]. In normal goblet cells, AGR2 is required for folding and processing of secreted (MUC5AC) and membranous mucins (MUC1, MUC2, MUC5B;) which have also been implicated in various tumor-promoting processes [30,40]. AGR2 was also shown to interact with reptin an inhibitor of p53, which may contribute to tumor cell proliferation and survival [41]. These molecular interactions may promote the mucinous and aggressive nature of PIMA tumors. Consistent with important role of FOXM1 and AGR2 in PIMAs, inactivation of either FoxM1 causes progression of benign lung adenoma to mucinous lung adenocarcinoma AGR2 or FOXM1 was sufficient to inhibit the tumor growth, invasion and metastasis in orthotopic mouse model. Altogether, our studies support a model in which FOXM1 functions downstream of KRAS and stimulates expression of AGR2 and other proliferation-specific and mucinous genes (Fig 8). Both FOXM1 and AGR2 induce tumor growth, progression, invasiveness and maintain mucinous characteristics in PIMAs. Since PIMA is a particularly aggressive cancer subtype without specific therapeutic options, our findings suggest that pharmacological inhibition of either AGR2 or its upstream regulator, FOXM1, may be beneficial for PIMA patients.
Modeling mucinous lung adenocarcinomas in mice requires both activation of oncogenic KRAS and downregulation of Nkx2.1 [3]. Neither KRAS activation nor reduction of NKX2.1 alone is sufficient for carcinoma development or mucin production, indicating context specific crosstalk between these two genes. In the present study, expression of FoxM1-ΔN in urethaneinduced lung adenomas decreased NKX2.1 and increased tumor progression and mucin production. Since urethane generates activating mutations in Kras, FOXM1 may regulate genetic interactions between Kras and Nkx2.1 during development of mucinous lung adenocarcinomas. Interestingly, FoxM1-ΔN-expressing lung tumors contained many SOX2 expressing cells [27]. Since SOX2 is a marker of stem-like and dedifferentiated tumor cells [42], activation of SOX2 can contribute to increased invasiveness of lung tumors in FoxM1-ΔN transgenic mice.
In summary, activation of FOXM1 in pre-existed lung adenomas caused progression to invasive, metastatic adenocarcinomas. FOXM1 enhanced mucinous characteristics in lung tumors and transcriptionally activated Agr2, a mucin-associated oncogene. We have developed a mouse model of mucinous, metastatic lung adenocarcinomas and demonstrated that FOXM1 is necessary and sufficient to induce mucinous phenotype in lung tumor cells.

Ethics statement
The cohort consisted of an independent lung adenocarcinoma patients who underwent lung cancer resection at the Department of Thoracic Surgery of the Thoraxklinik at the University Animal models SPC-rtTA mice harboring a 3.7kb fragment of the human SPC promoter was used to drive expression of an rtTA cassette as previously described [43]. (TetO) 7 -CMV-GFP-FoxM1-ΔN mice possessed a stabilized human FOXM1 cDNA cassette fused to GFP as previously described [20]. SPC-rtTA mice were bred to (TetO) 7 -CMV-GFP-FoxM1-ΔN to generate SPC-rtTA;TetO-FoxM1-ΔN mice as previously described [20]. Mice with single or no transgenes were used as controls. SPC-rtTA;TetO-FoxM1-ΔN mice without doxycycline were also used as controls. 6-8 week old male mice received six intraperitoneal (I.P.) injections of urethane at a dose of 1 gram per kilogram of body weight. 14 weeks after the first injection, mice were maintained on chow containing doxycycline, except for SPC-rtTA;TetO-FoxM1-ΔN control mice. The SPC-rtTA/ TetO-Kras G12D / TetO-FoxM1-ΔN mouse tumor model was used as previously described [20]. For A549 orthotopic xenograft experiments, 2.5x10 5 A549 cells were inoculated into the tracheas or left lung lobes of 8 week old male NOD.Cg-Prkdc scid ;Il2rg tm1Wjl /SzJ (NSG) immunocompromised mice.

RNA isolation and real-time qRT-PCR
Total mRNA was isolated from microdissected lung tumors or cell lines using RNA Stat60 according to the manufacturer's instructions. Isolated total mRNA was treated with DNASE I (Promega) according to the manufacturer's instructions. DNASE treated RNA was reverse transcribed into cDNA using the High Capacity cDNA Reverse Transcription Kit (Invitrogen). Real time PCR was performed using the TaqMan Gene Expression Assays and a StepOnePlus Real-Time PCR System (Invitrogen) as described previously [44]. Inventoried TaqMan mouse gene assays are summarized in S1 Table. All reactions were performed in triplicate and normalized to β-actin mRNA.

Lung histology, immunohistochemistry, and immunofluorescence
Mouse and human lung tissue paraffin sections (5μm) were stained with the following antibodies: anti-FOXM1 (Santa Cruz), anti-GFP (Abcam), anti-SPC (Seven Hills Bioreagents), anti-CC10 (Santa Cruz), anti-NKX2.1 (Seven Hills Bioreagents), anti-AGR2 (Cell Signaling), anti-MUC5Ac, anti-MUC5B (Abcam), anti-Ki67 (ThermoFisher Scientific), anti-Cleaved Caspase 3 (R&D Systems), anti-SOX2 (Santa Cruz). H&E Staining and Alcian Blue staining were performed according to the manufacturer's protocol and as previously described [33,45,46]. Human lung tissue samples were provided by the tissue bank of the National Center of Tumor Diseases (NCT, Heidelberg, Germany) in accordance with the regulations of the tissue bank and the approval of the Ethics Committee of the Heidelberg University. Image J software was used for quantification of staining intensity in human lung tissue samples. Mean positive signal was quantified on 8-bit images with the following thresholds (low/high): FOXM1 (22/176), AGR2 (41/165), and NKX2.1 (58/160). For quantification of Alcian blue staining, the blue hue was quantified on color images using the following threshold settings: hue 136/143 pass, saturation 0/255 pass, and brightness 200/255 pass. A minimum of 5 random 20x field images per patient were quantified. Adjacent normal lung tissue was included only if clearly discernable in the patient sample.

Western blot
Protein lysates were prepared from A549 and H2122 cell lines 48 hours after puromycin selection using RIPA buffer (Abcam) supplemented with 1mM PMSF (Sigma) and protease inhibitor cocktail (Roche). Western blot analysis was done as described previously [48,49]. Primary antibodies were incubated overnight at 4˚C in 1% nonfat, dry milk. β-actin was used as loading control. The following antibodies were used for Western blot: FOXM1 (Santa Cruz), AGR2 (Cell Signaling), or β-actin (Santa Cruz).

Cloning of the human AGR2 promoter and dual luciferase assay
The human AGR2 promoter region -1827bp to +197bp was amplified from A549 human genomic DNA using the following PCR primers: Forward-5' CCTTGCCATCCGTCAGCCACTA 3'; Reverse-5'TGCTGTCAGGAGCCTTACCTGG 3'. The PCR product was cloned into pCR2.1 TOPO and verified by DNA sequencing. The promoter fragment was subcloned into pGL2-Basic luciferase vector (LUC) by double digestion with Acc651 and XhoI restriction enzymes (New England Biolabs). For site-directed mutagenesis of the putative FOXM1 binding site, primers containing a 4 base pair change were used with the Site-Directed Mutagenesis kit (Invitrogen): Forward-5' TGTGTGTCTTCAAGTGACTGACGGCAATCTGCCCACGGA 3'; Reverse-5' TCCGTGGGCAGATTGCCGTCAGTCACTTGAAGACACACA 3'. Hek293T cells were transfected with a CMV-Empty or CMV-human FOXM1b plasmids, as well as with LUC reporter driven by -2.0 kb AGR2 promoter region (AGR2-LUC) WT or mutated AGR2-LUC promoter region. CMV-Renilla was used as an internal control to normalize transfection efficiency (relative luciferase activity). A dual luciferase assay (Promega) was performed 24 hours after transfection, as described previously [51,52]. The standard error of the mean was calculated using the mean relative luciferase activity for each of the three wells in each group. All values were normalized to cells transfected with AGR2-LUC + CMV-Empty plasmids.

Cloning of the human AGR2 cDNA
The human AGR2 cDNA was cloned from total RNA from A549 cells. RNA was isolated and reverse-transcribed into cDNA using random primers, as described above. For amplification of the AGR2 cDNA, we used Q5 High-Fidelity Polymerase (New England Biolabs) and the following primers: Forward-5' ATGGAGAAAATTCCAGTGTCAG 3'; Reverse-5' CTTTACAAT TCAGTCTTCAGCAAC 3'. The 530bp band corresponding to the AGR2 cDNA was gel digested, cloned into PCR2.1 TOPO (Invitrogen) and sequence verified. The AGR2 cDNA was subcloned into BamHI and XhoI restriction sites in the retroviral plasmid MIEG3.

Statistical analysis
Statistical analysis was performed using Prism software. ANOVA and Student's T-test were used to determine statistical significance. Right skewed measurements were log-transformed to meet normality assumption prior to analysis. P values less than 0.05 were considered significant. Values for all measurements were expressed as the mean ± standard deviation (SD). Data were graphically displayed using GraphPad Prism v.5.0 for Windows (GraphPad Software, Inc., La Jolla, CA, USA).
Supporting information S1 Fig. FOXM1 induces lung tumor progression. Representative H&E images depicting tumors of each grade observed in the lungs of control (n = 6) and epiFoxM1-ΔN (n = 9) mice. Scale bar = 500μM. (TIF) S2 Fig. FOXM1 is essential for the expression of cell cycle regulatory genes and markers of mucinous tumors. mRNA isolated from control A549 (scramble) or shFoxM1 A549 stable cell lines were used for qRT-PCR. Knockdown of FOXM1 decreased mRNAs of cell cycle regulators (A) and mucinous characteristics (B) in A549 human pulmonary invasive mucinous adenocarcinoma. mRNA expression is normalized to β-actin mRNA. A p-value <0.05 is marked with a single asterik ( Ã ) and a p-value <0.01 is marked with a double asterik ( ÃÃ ).