ED-1 and ED-2 murine lung cancer lines, C-10 murine immortalized lung epithelial cells, BEAS-2B human immortalized bronchial epithelial cells, and human lung cancer cell lines (A549, HOP-62, H-522, U-1752, NCI-H1730, and NCI-H2122) were each cultured in RPMI-1640 medium with 10% fetal bovine serum (FBS) and 1% antibiotic and antimycotic solution at 37°C in 5% CO₂ in a humidified incubator, as before (15,16,19–21). Cell lines were obtained from and authenticated (using genotypic and phenotypic assays) by ATCC except for murine ED-1 and ED-2 lung cancer cell lines that were previously described and authenticated (19,21).

Cyclopamine (LC Laboratories, Wobrun, MA) and tomatidine (Sigma-Aldrich, St. Louis, MO) were purchased as were recombinant mouse sHH (R&D Systems, Minneapolis, MN) and FBS (Gemini Bioproducts, Inc, Calabasas, CA). The Smo inhibitor MK-4101 (22) was provided by Merck. The SANT-1 Smo inhibitor (15) was purchased (Tocris Bioscience, Ellisville, MO) as was the SAG Smo agonist (EMD Millipore, Billerica, MA).

Cells were independently treated with the Smo inhibitors: cyclopamine, SANT-1 and MK-4101. In vivo Smo inhibition was achieved in mouse lung cancer models with cyclopamine (intraperitoneal injections, 40 mg/kg) treatments or with short hairpin RNA (shRNA)-mediated Smo knock-down in ED-1 cells. Individual small interfering RNA (siRNA)-mediated or shRNA-mediated repression of Gli1, Gli2, or Gli3 was achieved.

Cyclopamine growth effects were investigated in 705 human cancer cell lines using a high-throughput screen (19,23,24). Cells were treated with cyclopamine at 10 μM (and lower dosages) in media with 5% FBS and were assayed at 72 h with quantification by the SpectraMax M5 plate reader (Molecular Devices, Sunnyvale, CA). Means of triplicate cyclopamine treatment experiments were compared to vehicle controls, using optimized methods (19,23,24).

The HH pathway affects expression of Ptch1, cyclin D1, cyclin E, IGF2, IGFBP6, GILZ, Gli family members, and other species (2–5). The cor.test function (25) of R (26) compared cyclopamine-dependent growth responses to expressed species. Expression values were from U133 Plus 2.0 Affymetrix arrays and are publically available (27). The data set consisted of 490 samples corresponding to 164 unique cell lines that were typically examined in triplicate. Correlations were done: a) using all samples and b) using samples with a cyclopamine growth inhibitory response ≤0.75 at the 10 μM dosage, corresponding to 110 samples (in triplicate). Cyclopamine-mediated growth response was compared with mutation data available from the Sanger database (28).

Sequencing of the coding regions of Ptch1 was performed as before (29). For murine cell lines, murine-specific primers were used to sequence homologous domains of Ptch1. To assess for Smo mutations in murine and human cell lines, prior mutations in human Smo were searched for (A324T, V404M, D473H, E518K, W535L, and T640A) and corresponding regions of murine Smo were sequenced.

Logarithmically growing ED-1 (3×10⁴ and 4.5×10³), ED-2 (8×10⁴), C-10 (1×10⁵), BEAS-2B (3×10⁵), A549 (5×10⁴), HOP-62 (5×10⁴), H-522 (3×10⁵), U1752 (5×10⁴), NCI-H1703 (4.5×10⁵), and NCI-H2122 (6×10⁵) cells were individually plated onto each well of 6-well tissue culture plates (BD Bioscience, San Jose, CA) 24 h before transfection or drug treatments. Three replicate experiments were performed in triplicate. Logarithmically growing cells were assayed using the CellTiter-Glo assay (Promega, Madison, WI) (21). Apoptosis was scored with the Caspase-Glo 3/7 Assay System (Promega).

SiRNA transfections were with Oligofectamine reagent (Invitrogen, Carlsbad, CA). SiRNAs targeting Gli1, Gli2, Gli3, Smo, IGFBP6 or a RISC-free control were each purchased (Dharmacon, Lafayette, CO). Two different siRNAs independently targeted Gli1, Gli2, Gli3, Smo, or IGFBP6 sequences: 5′-GGUUGGAACUUCUGUGAUG-3′ (Gli1.1), 5′-GAGCAGGCCUCCGUUGUA-3′ (Gli1.2), 5′-GGGAGAAGAAGGAGUUCGU-3′ (Gli2.1), 5′-GGUUUGUGGUUGAGCGGAA-3′ (Gli2.2), 5′-CCAUCGGUGGAAAAGCGU-3′ (Gli3.1), 5′-GAAACGCAAUCACUAUGCA-3′ (Gli3.2), 5′-AGAACCCGCUGUUCACCGA-3′ (Smo1.1), 5′-GCAUCUGUUUUGUAGGCUA-3′ (Smo1.2), 5′-CAUCGAGGCUUCUACCGGA-3′ (IGFBP6-1), and 5′-CAACAGAGGAAUCCAGGCA-3′ (IGFBP6-2). Transient transfections were with FuGENE 6 reagent (Roche Applied Science, Indianapolis, IN). Reporter assays were performed after transfection of a GLIBS-Luciferase reporter construct or TK-Luciferase control plasmid (19). The Dual-Luciferase reporter assay system (Promega) was used in three replicate experiments, each performed in triplicate.

ED-1 cells stably expressing green fluorescent protein (GFP) were individually infected with lentivirus expressing shRNAs targeting Smo: TRCN0000026245 (Smo1) and TRCN0000026295 (Smo2); IGFBP6: TRCN0000114766 (IGFBP6.1) and TRCN0000114768 (IGFBP6.2); GILZ: TRCN0000085743 (GILZ.1) and TRCN0000085746 (GILZ.2), or a scrambled control shRNA (RHS4080) (Open Biosystems, Huntsville, AL). Selection was with puromycin (5 μg/ml). Knock-downs were determined using semi-quantitative real-time reverse transcription (RT) polymerase chain reaction (PCR) assays. Cells with the most robust Smo knock-down versus controls were selected for tail vein injection experiments into syngeneic mice (19,20,30).

Total RNA was isolated using the RNeasy kit (Qiagen, Valencia, CA). RT was with the High Capacity cDNA RT kit (Applied Biosystems, Foster City, CA) and a Peltier Thermal Cycler (MJ Research, Waltham, MA). Semi-quantitative real-time RT-PCR assays were with SYBR Green PCR Mastermix (Applied Biosystems) and the 7500 Fast real-time PCR system (Applied Biosystems). Human IGFBP6 and β-actin qPCR assays were performed with TaqMan Universal PCR Master Mix and the manufacturer’s described assays (Applied Biosystems). Three replicate experiments were performed. Primers were: murine Gli1 forward: 5′-CCAAGCCAACTTTATGTCAGGG-3′, and reverse: 5′-AGCCCGCTTCTTTGTTAATTTGA-3′; murine Gli2 forward: 5′-CAACGCCTACTCTCCCAGAC-3′, and reverse: 5′-GAGCCTTGATGTACTGTACCAC-3′; murine Gli3 forward: 5′-CACAGCTCTACGGCGACTG-3′, and reverse: 5′-CTGCATAGTGATTGCGTTTCTTC-3′; murine Smo forward: 5′-GAGCGTAGCTTCCGGGACTA-3′, and reverse: 5′-CTGGGCCGATTCTTGATCTCA-3′; murine IGFBP6 forward: 5′-TGCTAATGCTGTTGTTCGCTG-3′, and reverse: 5′-CACGGTTGTCCCTCTCTCCT-3′; murine GILZ forward: 5′-ACCACCTGATGTACGCTGTG-3′, and reverse: 5′-TCTGCTCCTTTAGGACCTCCA-3′; murine glyceraldehyde 3-phosphate dehydrogenase (GAPDH) forward: 5′-AGGTCGGTGTGAACGGATTTG-3′, and reverse: 5′-TGTAGACCATGTAGTTGAGGTCA-3′; human Gli1 forward: 5′-GGCACCATGAGCCCATCTC-3′, and reverse: 5′-ATCACCTTCCAAGGGTTCCTC-3′; human Gli2 forward: 5′-GGTGAAGCCTCCACCCTTTC-3′, and reverse: 5′-TGCATGTAGTTTACCCTGGGG-3′; human Gli3 forward: 5′-CTCCACGACCACTGAAAAGAAA-3′, and reverse: 5′-TCTCTGTGATAAGTCTGTCCAGG-3′; human Smo forward: 5′-GGCAACAGCATTGCAGTGAAG-3′, and reverse: 5′-GAGGAGAGACACACGAGCCT-3′; and human GAPDH forward: 5′-ATGGGGAAGGTGAAGGTCG-3′, and reverse: 5′-GGGGTCATTGATGGCAACAATA-3′.

Two hundred logarithmically growing ED-1 cells were plated for triplicate replicate clonal growth assays, as before (20). Colonies were stained with Diff Quick (IMEB Inc, San Marcos, CA) and were counted with the Oxford Optronix Col Count counter (Oxford Optronix, Oxford, UK) (31).

The New Hampshire State Cancer Registry and the Dartmouth-Hitchcock Medical Center Tumor Registry have been described (20). Analyses of normal versus malignant lung tissue arrays were by a pathologist (Vincent A. Memoli) unaware of findings from the associated clinical database. Signed consents were obtained; studies were reviewed and approved by Dartmouth’s Institutional Review Board (IRB) for human subjects. A paired normal-malignant lung tissue microarray was constructed (20).

Results were expressed as means ± standard deviations. Results of all independent experiments were pooled to assess for statistical significance. Z-test and two-sided t-tests were used for statistical analyses with significance considered for values of P<0.05.

Three 9-month-old female mice expressing transgenic wild-type human cyclin E were each treated daily (intraperitoneal) for 5 consecutive days with cyclopamine (40 mg/kg) or vehicle (45% hydroxypropyl-ß-cyclodextrin, HBS) with a total of six mice studied. Mice were examined using an Institutional Animal Care and Use Committee (IACUC)-approved protocol. Tissues were formalin-fixed, paraffin-embedded and sectioned for histopathology (20). Hematoxylin and eosin staining as well as Ki-67, cyclin D1, and cyclin E immunostaining were performed as before (16,19). Analyses were by a pathologist (Vincent A. Memoli) unaware whether harvested tissues were from cyclopamine or vehicle-treated mice.

ED-1 cells were transduced with a GFP expression vector and transductants were sorted (15,16) with 8×10⁵ cells injected into tail veins of each female FVB mouse (8-week old). Ten mice per arm were used and replicate experiments were performed. To investigate cyclopamine antineoplastic effects, 10 mice were each intraperitoneally treated daily for 14 days with cyclopamine (40 mg/kg); 10 additional mice were each treated with vehicle (45% HBS). Treatments began 2 weeks after tail-vein injections because lung tumors formed by then (data not shown). Ten syngeneic FVB mice were each injected with ED-1 cells that were treated with either MK-4101 (10 μM) or vehicle as part of an IACUC-approved protocol. Mice were sacrificed 28 days post-tail vein injections. Replicate experiments were performed. Lung tissues were formalin-fixed, paraffin-embedded and sectioned for histopathology (16,19,20). A rabbit polyclonal anti-GFP antibody (product ab290) (Abcam, Cambridge, MA) was used to identify lung tumors. Hematoxylin counterstaining was used.

A total of 705 different human cancer lines were examined for responses to the Smo inhibitor cyclopamine (10 μM). Proliferation was reduced in a subset of cancer cell lines after 72 h of treatment (Fig. 1A, left panel). Of these lines, 353 were displayed in the Sanger database. No Ptch1 or Smo mutations were detected (3 lines had Smo amplifications that did not confer cyclopamine hypersensitivity). Data from the Sanger database showed the distribution of somatic mutations in Ptch1 for 860 unique cancer cell lines. Six had variant sequences of uncertain functional consequences. For the Smo gene, only 6 out of 866 different cancer cells had sequence variants of uncertain functional consequence. There were 152 lung cancer cell lines analyzed. None of these lines had Ptch1 or Smo mutations.

The HH pathway can affect expression of the Gli family members, Ptch1, cyclin D1, cyclin E, IGF2, IGFBP6, or GILZ (1–5). Cyclopamine-dependent growth responses were compared to these expressed species using the described database. Most highly significant associations occurred for high cyclin E (P=0.000009) and low IGFBP6 (P=0.000004) levels (Fig. 1A). Less significant associations were detected for IGF2 (P=0.005), cyclin D1 (P=0.024), Gli1 (P=0.04), and Gli2 (P=0.05). High GILZ levels were linked to reduced cyclopamine response (P= 0.002).

Twenty lines were of SCLC and 107 were of NSCLC origins. Subsets of SCLC (Fig. 1B) and NSCLC (Fig. 1C) lines responded to cyclopamine with substantial growth inhibition. NSCLC responses occurred independently of mutations for Ptch1 or Smo (Table IA). These responses were also independent of ras, epidermal growth factor receptor (EGFR), or p53 mutations (Table IB). K-RAS mutations in NSCLC are associated with resistance to EGFR-tyrosine kinase inhibitors (TKIs) (32–35). Growth inhibition in the most cyclopamine-responsive NSCLC cells lines (NCI-H1703 and NCI-H2122) were independently confirmed in Fig. 2.

Studies were next conducted in lung cancer because an autocrine HH pathway is reported in lung cancer (36). HH signaling is important for growth of NSCLC and SCLC (6,15). Other reasons for studying lung cancer included the finding that increased cyclin E expression was associated with cyclopamine response (Fig. 1A) and cyclin E transgenic mice recapitulated key features of human lung cancer biology (16,30).

Gain or loss of expression was achieved for the most significantly associated candidate regulators of cyclopamine response (cyclin E and IGFBP6) or resistance (GILZ). Gli1 is an established HH pathway regulator in lung cancer (15). To determine whether cyclin E affected cyclopamine response, cyclin E was overexpressed in murine immortalized C-10 lung epithelial cells (Fig. 2A, left panel shows a 2.26-fold increase in cyclin E protein versus actin expression) and this significantly (P<0.01) increased cyclopamine-mediated growth inhibition versus the inactive isomer tomatidine (Fig. 2A, right panel). Similar results were obtained using a second Smo inhibitor, SANT-1 (data not shown), which was consistent with prior studies (15).

Two shRNAs (Fig. 2B, left panel) were selected to reduce IGFBP6 expression independently in ED-1 lung cancer cells derived from a transgenic mouse that expressed wild-type cyclin E (30). Individual knock-down with these shRNAs significantly (P<0.01) increased Smo-mediated growth inhibition in this murine lung cancer cell line versus an inactive shRNA control (Fig. 2B, right panel). Thus, gain of cyclin E or loss of IGFBP6 expression regulated response to a Smo inhibitor. In contrast, two different shRNAs independently repressed GILZ expression (Fig. 2C, left panel), but these did not appreciably affect cyclopamine-mediated growth suppression at 72 h (Fig. 2C, right panel). Findings were confirmed in human NSCLC cells engineered with IGFBP6 knock-down (Fig. 2D and E).

Lung cancers arising in transgenic mice from increased cyclin E expression in the lung exhibit high levels of Gli1 (16,30). These transgenic lung cancers are also shown to express low levels of IGFBP6 and GILZ versus the adjacent normal lung (Fig. 3A). Whether cell lines derived from these murine lung cancers responded to a Smo inhibitor was independently studied in ED-1 and ED-2 lung cancer cells derived respectively from transgenic mice whose lung tumors expressed wild-type (ED-1) or degradation-resistant (ED-2) cyclin E species (30).

Effects of Smo inhibition were explored by targeting Smo in ED-1 cells. Gli1, Gli2, Gli3, and Smo were independently targeted using two different siRNAs engineered to repress each of these species. Independent knock-down of Gli1, Gli2, and Smo each significantly inhibited ED-1 cell growth (as did cyclopamine treatment), but Gli3 knock-down did not appreciably affect growth (Fig. 3B, left and middle panels). As expected, the Smo agonist SAG augmented Gli1 expression; cyclopamine co-treatment prevented this (Fig. 3B, right panel). Compensatory changes in expressed HH pathway members were found after these different knock-downs in Fig. 3C and D. Smo knock-down substantially decreased lung cancer cell growth.

Cyclopamine treatment also significantly (P<0.01) decreased ED-1 clonal growth while increasing apoptosis (Fig. 4A). ED-1 and ED-2 cells expressed multiple HH pathway components as did the examined murine and human immortalized pulmonary epithelial or cancer cell lines (Fig. 4B). Genomic DNA from each of these cell lines was sequenced for the entire coding region of Ptch1 and for Smo mutations in regions that conferred Smo inhibitor resistance (A324T, V404M, D473H, E518K, W535L, and T640A). No mutations were detected (data not shown). Human lung cancer cell lines studied were those in which responses to Smo inhibition were already reported (15,36). To establish whether the HH pathway was functional in ED-1 cells, these cells were treated with recombinant sHH at different dosages in low serum (0.5%) containing medium. HH pathway activity in these cells was confirmed; sHH treatment significantly (P<0.01) increased mRNA expression for Gli1 and Ptch (Fig. 4C). Although ED-1 cells arose from transgenic lung cancers that activated the HH pathway (16), these cells had relatively low basal Gli1 and Gli2 mRNA levels (Fig. 4B) and low basal Gli1-reporter activity (Fig. 4D). This is consistent with differences reported between ex vivo versus in vivo HH pathway dependence of lung cancer cells (37). It was therefore not unexpected to detect prominent Gli1 immunostaining of lung tumors arising from ED-1 cells after their tail-vein injections into syngeneic FVB mice (data not shown).

To explore independently Smo inhibitor effects on lung cancer cells, ED-1 cells were engineered to have Smo knockdown. This was done using two independent shRNAs versus an inactive control shRNA (Fig. 4E). This conferred significant repression (Smo1 shRNA P=0.002 and Smo2 shRNA P=0.001) of transplanted lung cancers (Fig. 4F). Ex vivo treatment of ED-1 cells with another Smo inhibitor MK-4101 significantly reduced ED-1 colony formation (P<0.01, data not shown) and tumorigenicity (P=0.009) after tail-vein injections into syngeneic mice of these versus control cells (Fig. 4G).

Cyclopamine responses were examined in cyclin E transgenic mice because their lung cancers expressed a gene profile that was indicative of dependence on HH signaling. Cyclopamine-treatment decreased immunohistochemical expression of the proliferation markers Ki-67, cyclin D1 and cyclin E in the neoplastic versus normal lungs of treated versus age and sex-matched control mice (Fig. 5A, left panel). This result suggested that Smo inhibition would also reduce lung tumor formation in the transplantation model. ED-1 cells were injected into the tail veins of syngeneic mice. Cyclopamine treatment of mice began after lung tumors were histologically present (data not shown). This significantly (P=0.00002) reduced lung cancers in cyclopamine versus vehicle-treated mice (Fig. 5A, right panel).

To learn whether a profile indicative of HH pathway dependence was expressed in human lung cancers, a normal-malignant lung tissue array (20) was studied. Immunohistochemical expression profiles revealed significant differences in the malignant versus normal lung for cyclin E (P<0.01), IGFBP6 (P<0.01), and GILZ (P<0.01) in Fig. 5B. Lung cancers with high cyclin E levels expressed significantly (P<0.01) higher levels of Gli1 than did cases with reduced cyclin E levels (Fig. 5B). Cases with high cyclin E and low IGFBP6 levels had increased (P<0.0001) Gli1 immunohistochemical expression. Representative results are presented (Fig. 5C).