Application of Mass Spectrometry Profiling to Establish Brusatol as an Inhibitor of Global Protein Synthesis
2015; Elsevier BV; Volume: 15; Issue: 4 Linguagem: Inglês
10.1074/mcp.m115.055509
ISSN1535-9484
AutoresSteffan Vartanian, P. Taylur, James Lee, Peter M. Haverty, Donald S. Kirkpatrick, Kebing Yu, David Stokoe,
Tópico(s)Coenzyme Q10 studies and effects
ResumoThe KEAP1/Nrf2 pathway senses and responds to changes in intracellular oxidative stress. Mutations that result in constitutive activation of Nrf2 are present in several human tumors, especially non-small cell lung cancer. Therefore, compounds that inhibit Nrf2 activity might be beneficial in treating patients whose tumors show activation of this pathway. Recent reports suggest that the natural product brusatol can potently and selectively inhibit Nrf2 activity, resulting in cell cytotoxicity, and can be effectively combined with chemotherapeutic agents. Here, we analyzed the effects of brusatol on the cellular proteome in the KEAP1 mutant non-small cell lung cancer cell line A549. Brusatol was found to rapidly and potently decrease the expression of the majority of detected proteins, including Nrf2. The most dramatically decreased proteins are those that display a short half-life, like Nrf2. This effect was confirmed by restricting the analysis to newly synthesized proteins using a labeled methionine analogue. Moreover, brusatol increased the expression of multiple components of the ribosome, suggesting that it regulates the function of this macromolecular complex. Finally, we show that brusatol induces its potent cellular cytotoxicity effects on multiple cancer cell lines in a manner independent of KEAP1/Nrf2 activity and with a profile similar to the protein translation inhibitor silvestrol. In conclusion, our data show that the activity of brusatol is not restricted to Nrf2 but, rather, functions as a global protein synthesis inhibitor. The KEAP1/Nrf2 pathway senses and responds to changes in intracellular oxidative stress. Mutations that result in constitutive activation of Nrf2 are present in several human tumors, especially non-small cell lung cancer. Therefore, compounds that inhibit Nrf2 activity might be beneficial in treating patients whose tumors show activation of this pathway. Recent reports suggest that the natural product brusatol can potently and selectively inhibit Nrf2 activity, resulting in cell cytotoxicity, and can be effectively combined with chemotherapeutic agents. Here, we analyzed the effects of brusatol on the cellular proteome in the KEAP1 mutant non-small cell lung cancer cell line A549. Brusatol was found to rapidly and potently decrease the expression of the majority of detected proteins, including Nrf2. The most dramatically decreased proteins are those that display a short half-life, like Nrf2. This effect was confirmed by restricting the analysis to newly synthesized proteins using a labeled methionine analogue. Moreover, brusatol increased the expression of multiple components of the ribosome, suggesting that it regulates the function of this macromolecular complex. Finally, we show that brusatol induces its potent cellular cytotoxicity effects on multiple cancer cell lines in a manner independent of KEAP1/Nrf2 activity and with a profile similar to the protein translation inhibitor silvestrol. In conclusion, our data show that the activity of brusatol is not restricted to Nrf2 but, rather, functions as a global protein synthesis inhibitor. The KEAP1 1The abbreviations used are:KEAP1kelch-like ECH-associated protein 1DCMdichloromethaneNMRnuclear magnetic resonanceRTroom temperatureACNacetonitrileIDinner diameterSPSsynchronous precursor selectionGAPDHglyceraldehyde-3-phosphate dehydrogenaseVSVGvescicular stomatitis virus glycoproteinCMVcytomegalovirusRABL5Rab-like protein 5HPCL1Hippocalcin-like protein 1PYRG2pyrimidine reductaseG2AREanti-oxidant response elementCETSAcellular thermal shift assayNADPHnicotinamide adenine dinucleotide phosphateNSCLCnon-small cell lung cancerTMTtandem mass tagsUTRuntranslated region./Nrf2 pathway is one of the most important cellular mechanisms to react and respond to oxidative stresses. Under normal physiological conditions, Nrf2 protein is maintained at low levels due to constitutive ubiquitination and degradation. This is achieved through recruitment to the Cul3 ubiquitin ligase complex by the substrate adaptor protein KEAP1. KEAP1 exists as a dimer, which binds a single Nrf2 molecule through two peptide motifs at its N terminus, a high affinity GluThrGlyGLU motif and a lower affinity AspLeuGly motif. This positions Nrf2 for effective ubiquitin transfer. Polyubiquitinated Nrf2 is subsequently degraded by the proteasome. In response to oxidative stress or electrophiles, key cysteine residues in KEAP1 become oxidized, thereby impairing its ability to recruit and ubiquitinate Nrf2 (1.Rachakonda G. Xiong Y. Sekhar K.R. Stamer S.L. Liebler D.C. Freeman M.L. Covalent modification at Cys151 dissociates the electrophile sensor Keap1 from the ubiquitin ligase CUL3.Chem. Res. Toxicol. 2008; 21: 705-710Crossref PubMed Scopus (160) Google Scholar). This results in Nrf2 stabilization, allowing it to translocate to and accumulate in the nucleus (reviewed in (2.Tong K.I. Kobayashi A. Katsuoka F. Yamamoto M. Two-site substrate recognition model for the Keap1-Nrf2 system: A hinge and latch mechanism.Biol. Chem. 2006; 387: 1311-1320Crossref PubMed Scopus (387) Google Scholar)). Nrf2 binds specific DNA sequences in complex with the small Maf proteins, which were originally identified as viral oncogenes containing leucine zipper motifs (3.Nishizawa M. Kataoka K. Goto N. Fujiwara K.T. Kawai S. v-maf, a viral oncogene that encodes a "leucine zipper" motif.Proc. Natl. Acad. Sci. U.S.A. 1989; 86: 7711-7715Crossref PubMed Scopus (220) Google Scholar). Hundreds of target genes have been identified using ChIP-seq and RNAseq experiments (4.Malhotra D. Portales-Casamar E. Singh A. Srivastava S. Arenillas D. Happel C. Shyr C. Wakabayashi N. Kensler T.W. Wasserman W.W. Biswal S. Global mapping of binding sites for Nrf2 identifies novel targets in cell survival response through ChIP-Seq profiling and network analysis.Nucleic Acids Res. 2010; 38: 5718-5734Crossref PubMed Scopus (569) Google Scholar, 5.Chorley B.N. Campbell M.R. Wang X. Karaca M. Sambandan D. Bangura F. Xue P. Pi J. Kleeberger S.R. Bell D.A. Identification of novel NRF2-regulated genes by ChIP-Seq: Influence on retinoid X receptor alpha.Nucleic Acids Res. 2012; 40: 7416-7429Crossref PubMed Scopus (393) Google Scholar, 6.Hirotsu Y. Katsuoka F. Funayama R. Nagashima T. Nishida Y. Nakayama K. Engel J.D. Yamamoto M. Nrf2-MafG heterodimers contribute globally to antioxidant and metabolic networks.Nucleic Acids Res. 2012; 40: 10228-10239Crossref PubMed Scopus (268) Google Scholar), highlighting the importance of Nrf2 in reducing levels of oxidative stress by increasing glutathione and NADPH synthesis, and up-regulating drug efflux pumps. Therefore, this pathway senses and responds to oxidative stress to maintain cellular redox homeostasis. kelch-like ECH-associated protein 1 dichloromethane nuclear magnetic resonance room temperature acetonitrile inner diameter synchronous precursor selection glyceraldehyde-3-phosphate dehydrogenase vescicular stomatitis virus glycoprotein cytomegalovirus Rab-like protein 5 Hippocalcin-like protein 1 pyrimidine reductaseG2 anti-oxidant response element cellular thermal shift assay nicotinamide adenine dinucleotide phosphate non-small cell lung cancer tandem mass tags untranslated region. When dysregulated, this pathway also contributes to many human pathologies including cardiovascular and neurodegenerative diseases, inflammation and cancer. Large-scale genome sequencing efforts have identified mutations in both KEAP1 and Nrf2 in several human tumor indications, but especially in non-small cell lung cancers (NSCLC). Mutations in KEAP1 are distributed across the gene and occur in ∼20 and ∼12% of the adenomatous tumors and squamous NSCLC, respectively (7.Cancer Genome Atlas Research, N. Comprehensive molecular profiling of lung adenocarcinoma.Nature. 2014; 511: 543-550Crossref PubMed Scopus (3557) Google Scholar). In addition to KEAP1, ∼15% of squamous NSCLC cases present with mutations in Nrf2 (8.Cancer Genome Atlas Research, N. Comprehensive genomic characterization of squamous cell lung cancers.Nature. 2012; 489: 519-525Crossref PubMed Scopus (2925) Google Scholar), supporting the hypothesis that Nrf2 degradation is frequently dysregulated in this indication (9.Hast B.E. Cloer E.W. Goldfarb D. Li H. Siesser P.F. Yan F. Walter V. Zheng N. Hayes D.N. Major M.B. Cancer-derived mutations in KEAP1 impair NRF2 degradation but not ubiquitination.Cancer Res. 2014; 74: 808-817Crossref PubMed Scopus (92) Google Scholar). Nrf2 mutations cluster in two hotspot regions, ∼10 amino acids each, near the DLG and ETGE KEAP1 interacting regions, resulting in constitutive Nrf2 activation. A reasonable expectation is that these tumors become addicted to the high levels of Nrf2 activity, and some data have been presented that supports this hypothesis (10.Singh A. Boldin-Adamsky S. Thimmulappa R.K. Rath S.K. Ashush H. Coulter J. Blackford A. Goodman S.N. Bunz F. Watson W.H. Gabrielson E. Feinstein E. Biswal S. RNAi-mediated silencing of nuclear factor erythroid-2-related factor 2 gene expression in non-small cell lung cancer inhibits tumor growth and increases efficacy of chemotherapy.Cancer Res. 2008; 68: 7975-7984Crossref PubMed Scopus (304) Google Scholar). Identification of a selective inhibitor of this pathway would be of great interest for potential therapeutic applications, which led us to explore the use of brusatol as an Nrf2 inhibitor. Brusatol is a representative compound from a family of natural product quassinoids. Recent publications have suggested that it can decrease the levels of Nrf2 and sensitize cells to various chemotherapy agents (11.Ren D. Villeneuve N.F. Jiang T. Wu T. Lau A. Toppin H.A. Zhang D.D. Brusatol enhances the efficacy of chemotherapy by inhibiting the Nrf2-mediated defense mechanism.Proc. Natl. Acad. Sci. U.S.A. 2011; 108: 1433-1438Crossref PubMed Scopus (474) Google Scholar, 12.Olayanju A. Copple I.M. Bryan H.K. Edge G.T. Sison R.L. Wong M.W. Lai Z.Q. Lin Z.X. Dunn K. Sanderson C.M. Alghanem A.F. Cross M.J. Ellis E.C. Ingelman-Sundberg M. Malik H.Z. Kitteringham N.R. Goldring C.E. Kevin Park B. Brusatol provokes a rapid and transient inhibition of Nrf2 signaling and sensitizes mammalian cells to chemical toxicity-implications for therapeutic targeting of Nrf2.Free Radical Biol. Med. 2014; 78: 202-212Crossref PubMed Scopus (143) Google Scholar). To identify and characterize the cellular targets of brusatol, we chose to employ a mass-spectrometry-based strategy. The binding of small molecules to their target proteins can induce conformational changes and alter their thermal stability compared with the apo form. This fact has been recently exploited to demonstrate engagement of known protein targets by small molecules in cells using Western blotting in an approach termed the cellular thermal shift assay (CETSA) (13.Martinez Molina D. Jafari R. Ignatushchenko M. Seki T. Larsson E.A. Dan C. Sreekumar L. Cao Y. Nordlund P. Monitoring drug target engagement in cells and tissues using the cellular thermal shift assay.Science. 2013; 341: 84-87Crossref PubMed Scopus (1047) Google Scholar). More recently, this assay was coupled with multiplexed mass spectrometry to yield an assay capable of surveying small molecule–protein interactions across a large fraction of the proteome (14.Savitski M.M. Reinhard F.B. Franken H. Werner T. Savitski M.F. Eberhard D. Martinez Molina D. Jafari R. Dovega R.B. Klaeger S. Kuster B. Nordlund P. Bantscheff M. Drewes G. Proteomics. Tracking cancer drugs in living cells by thermal profiling of the proteome.Science. 2014; 346: 1255784Crossref PubMed Scopus (589) Google Scholar). Our work using this approach, and follow up studies, suggest that brusatol regulates Nrf2 through an indirect mechanism involving global inhibition of protein synthesis. A549 cells (obtained from ATCC) were maintained in RPMI 1640 medium supplemented with 10% (v/v) fetal bovine serum (FBS, Sigma Aldrich) and 10 mm l-glutamine. Cells were maintained in a humidified incubator at 37 °C/5% CO2. The ground, air-dried seeds of Brucea sumatrana (200 g) were exhaustively extracted with n-hexane (1 l). The residue, after the removal of hexane, was further extracted with methanol (1 l). The total methanol extract evaporated to a viscous residue that was dark green in color and extremely bitter. The syrupy material was mixed with an equal volume of water and the filtered solution extracted with hexane until the extracts were colorless, then with methylene chloride. The methylene chloride extract was dried and evaporated to yield a dark yellow resinous material (500 mg). TLC (DCM/MeOH = 50:1) of the crude material showed two main components that gave green colors with a ferric chloride spray. This crude material was purified via column chromatography on silica gel (DCM/MeOH = 100:1∼40:1) to give crude brusatol, which was recrystallized from DCM/ether to give pure brusatol (30 mg) as a white solid (20 mg was used for analytical studies). Analytical nmr spectra match literature reports (15.Yang Z. Xie H. Wang J. Chemical studies of the active antitumor components from the fruits of Brucea javanica.Natural Product Research and Development. 1996; 1996: 35-39Google Scholar, 16.Sim K.Y. Sims J.J. Geissman T.A. Constituents of Brucea sumatrana J.Organic Chem. 1968; 33: 429-431Crossref Scopus (40) Google Scholar). For lysate assays, A549 cells were seeded in 10 cm plates and grown to 80% confluence. The cells were trypsinized, pelleted, and washed two times with PBS. The cell pellet was resuspended in 4.4 ml KB buffer (25 mm Tris, pH 7.5, 1 mm DTT, 10 mm MgCl2, complete protease inhibitor (Roche) and phosphatase inhibitor mixture II (Sigma)). The cell suspension was subjected to three cycles of freezing in liquid nitrogen and thawing at 37 °C, vortexed, and centrifuged at 20,000 × g for 20 min at 4 °C. The supernatant (lysate) was divided into two aliquots and treated with 500 nm brusatol or DMSO for 30 min at 25 °C. Following treatment, the lysates were aliquoted into PCR tubes (50 μl per tube) and heated in a thermal cycler at 38 °C, 42 °C, 46 °C, 50 °C, 54 °C, 58 °C, 62 °C, 66 °C, or 70 °C for 3 min, followed by a cool-down at room temperature for an additional 3 min. Lysates were pooled by treatment and temperature into 1.5 ml Eppendorf tubes and centrifuged at 20,000 × g for 20 min at 4 °C. The supernatants were isolated and submitted for MS analysis. For intact cell assays, A549 cells were seeded in 10 cm plates and grown to 80% confluence. Cells were treated with either 500 nm brusatol or DMSO for 4 h (five plates per treatment). Following treatment, the cells were trypsinized, pelleted, and washed 2× with PBS. Cells from each treatment arm resuspended in 2.2 ml PBS supplemented with protease and phosphatase inhibitors. Cells were aliquoted into PCR tubes (50 μl per tube) and pelleted, removing all but ∼10 μl of remaining PBS. The PCR tubes were then heated as previously described and subjected to three cycles of freeze/thaw as before. Cells were then resuspended in KB buffer, pooled by treatment and temperature, vortexed, and centrifuged at 20,000 × g for 20 min at 4 °C. The supernatants were isolated and submitted for MS analysis. These datasets were acquired once for potential target discovery, and hits were subsequently validated by Western blots. Lysates were prepared as described above. For dose/time-course experiments, cells were treated with brusatol or cyclohexamide (EMD Millipore) at the dose and time points indicated, followed by lysate harvest in KB buffer using the snap-freeze method described above. Proteins were separated on 4–20% Tris-glycine gels and transferred to nitrocellulose membranes via iBlot (Invitrogen). Membranes were blotted with the following primary antibodies: Nrf2 (Abcam ab62352), cystatin-C (Abcam ab133495), IGFBP4 (Thermo PA5–25925), PAF1 (Bethyl A304–374A), NRCAM (ProteinTech 21608–1-AP), MCL-1 (Cell Signaling 5453), p21 (Cell Signaling 2947), Gemin-4 (Abcam ab31581), FAM96A (Acris AP51504PU-N), RL22 (Santa Cruz sc-136413), GCN1L1 (Abcam ab86139), RPL11 (Abcam ab79352), RL9 (Santa Cruz sc-292593), survivin (sc-17779), cyclin A (sc-751), and β-actin (Cell Signaling 12620). Western blotting was performed a minimum of two times using biological replicates, and a representative example is shown. For each sample, 8 m urea was added in 1:1 ratio by volume to aid in denaturation of proteins before digestion. Proteins were reduced with 5 mm dithiothreitol (DTT) at 65 °C for 25 min and then alkylated with 11 mm iodoacetamide at room temperature, in the dark for 20min. Samples were further diluted to 2 m urea with 20 mm HEPES (pH 8). Trypsin (Promega) was then added at a ratio of 1:100 (enzyme:substrate by weight)(17.Beausoleil S.A. Villén J. Gerber S.A. Rush J. Gygi S.P. A probability-based approach for high-throughput protein phosphorylation analysis and site localization.Nat. Biotechnol. 2006; 24: 1285-1292Crossref PubMed Scopus (1212) Google Scholar) and digestion carried out overnight at 37 °C. The reaction was quenched with 1% trifluoroacetic acid, and samples were desalted using C-18 Sep-Pak (Waters). Eluents were dried to completion, resuspended with 100 μl of 100 mm HEPES, pH 8.0, and labeled with TMT10plex labeling reagent (Thermo) according to the manufacturer's protocol. TMT-labeling reactions were quenched with 5% hydroxylamine for 15 min before combining into the two sets (Set 1: DMSO and brusatol at 38 °C, 42 °C, 46 °C, 50 °C, and 54 °C; Set 2: DMSO and brusatol at 38 °C, 58 °C, 62 °C, 66 °C, and 70 °C). Both sets were acidified with trifluoroacetic acid and dried to completion using a speedvac. Each set was fractionated by high pH reverse phase using an Agilent 1100 series offline HPLC system. Peptides were loaded onto an Agilent Zorbax 300 Extend C-18 analytical column with 2.1 × 150 mm dimensions and 3.5 μm particle size. Solvent A consisted of 25 mm ammonium formate (pH 9.7) while solvent B was 100% ACN. Peptide fractions were collected in 45 s intervals for a total of 63 min with a linear gradient of 15% to 60% B. Ninety-six fractions were collected, and every 16th fraction was combined to form a final set of 16 distinct groups: fraction 1, 17, 33, 49, 65, 81 into one fraction, etc. Samples were lyophilized, subjected to STAGEtip for desalting, and injected for LC-MS/MS analysis. TMT-labeled samples were analyzed on an Orbitrap-Fusion mass spectrometer (Thermo Fisher Scientific, San Jose, CA USA) connected to a NanoAcquity UPLC system (Waters, Milford, MA) via an EASY-Spray ion source (Thermo Scientific). Peptide separation was achieved over a PepMap RSLC EASY-Spray column (75 μm inner diameter, 15 cm length, 2 μm C18 beads from Thermo Scientific) flowing at 800 nl/min. A linear gradient of 2% Solvent B to 25% Solvent B was performed over 85 min with a total method run time of 120 min. Solvent B consisted of 98% ACN/2% water/0.1% formic acid, and Solvent A was composed of 98% water/2% ACN/0.1% formic acid. The mass spectrometer was operated with a spray voltage of 1.9kV. Full MS scans were collected in the Orbitrap at 120,000 resolution across a mass range of 375 to 1600 m/z. The MS1 automatic gain control target was set to of 2 × 105 with a maximum injection time of 50 ms. MS2 ions (automatic gain control target 1 × 104, maximum injection time of 50 ms) were selected using a top speed data-dependent mode and fragmented with normalized collision energy of 30 and analyzed in the linear ion trap. Utilizing the Orbitrap Fusion's SPS mode, the top 10 precursor ions previously analyzed in MS2 were selected again for MS3 analysis (isolation width of 2 Th, automatic gain control target 1 × 105, maximum injection time 120 ms). Fragment ions were generated via higher collision dissociation (normalized collision energy 65) and analyzed in the Orbitrap at a resolution of 60,000. Raw files were converted to mzXML using ReAdW (version 4.3.1 with minor modification to enable parsing rapid scan and SPS scan headers). MS/MS spectra were searched using Mascot (v.2.3.02) against target sequences from the human proteome database (UniProt Dec. 2011) and known contaminants concatenated with a decoy database containing reverse sequences of each entry (total entries are 238,494, including both forward and decoy hits). Search parameters included tryptic specificity with allowance of up to two missed cleavage events, a precursor ion tolerance of 50 ppm, and a fragment ion tolerance of 0.8 Da. Searches permitted variable modifications of methionine oxidation (+15.9949 Da) and static modifications of cysteine (+57.0215) for carbamidomethylation as well as TMT10plex tags on lysine and peptide N termini (+229.1629). Reporter ions produced by the TMT tags were quantified with an in-house software package known as Mojave. In brief, TMT reporter ions were extracted by isolating peaks within 20 ppm of theoretical TMT reporter mass. The intensity of the highest peak within the window is reported. Peptide spectra matches were filtered with a false discovery rate of 5% on the peptide level and subsequently at 2% on the protein level using linear discrimination (17.Beausoleil S.A. Villén J. Gerber S.A. Rush J. Gygi S.P. A probability-based approach for high-throughput protein phosphorylation analysis and site localization.Nat. Biotechnol. 2006; 24: 1285-1292Crossref PubMed Scopus (1212) Google Scholar). Additional filtering of quantified peptide spectra matches incorporated an isolation specificity of 0.7 or greater and a combined TMT intensity greater than 3000. The resulting dataset was then subjected to an R script for melting curve analysis. Briefly, two TMT sets were combined based on the shared 38 °C temperature point samples. A sigmoid curve was fitted to the temperature series for each treatment group using R nonlinear least-squares estimates function nls() based on logistic model "SSlogis" (package "stats," version 3.2.2). Parameters for these functions were illustrated below, where "data" is an R data frame containing temperature series and the corresponding protein intensity and "Asym," "xmid," "scal" are all computed by SSlogis function internally. nls(intensity~SSlogis(temperature,Asym,xmid,scal),data=data)(Eq. 1) The melting point was calculated at 50% of initial protein abundance level. All mass spectrometry raw datasets have been deposited to the MASSIVE database (http://massive.ucsd.edu/) and can be downloaded by the identifier MSV000079314. A complete list of peptide-spectrum-matches passing abovementioned filters (except for TMT intensity threshold) can be found in Supplemental Table 1 for the cell lysate CETSA and intact cell CETSA experiments. A complete list of proteins with normalized abundance in each condition and melting points can be found in the Supplemental Table 2. Annotated MS/MS for all one hit wonders can be found in the Supplemental Fig. 2. RNA was harvested via RNEasy Mini kits (Qiagen), cDNA synthesized with High Capacity cDNA reverse transcription kit (Life Technologies, Foster City, CA), and Taqman assays run using the following primer/probe sets (Life Technologies): Nrf2 (Hs00975961_g1), cystatin C (Hs00264679_m1), PAF1 (Hs00219496_m1), NRCAM (Hs01031598_m1), GCN1L (Hs00412445_m1), RPL11 (Hs00831112_s1), and RL22 (Hs01865331_s1). Endogenous control assays multiplexed with the target assays using GAPDH-Vic primers/probe (Hs03929097_g1). Taqman assays were performed twice, and technical replicates from one experiment are shown in the figure. Cells were seeded in six-well plates and grown to 80% confluence, then treated with brusatol or cyclohexamide for 15 min. The medium was removed and cells washed three times with PBS. Methionine-free medium was then added (with brusatol/cyclohexamide) for 30 min. Labeling compound l-azidohomoalanine (Invitrogen, Grand Island, NY) was added at a final concentration of 50 μm for 1 h. Cells were then washed three times with PBS, and lysates harvested using 200 μl lysis buffer per well (1% SDS in 50 mm Tris-HCl, pH 8.0, protease/phosphatase inhibitors). Click reactions were carried out using 90 μg of input lysate, biotinylated alkyne reagent (B10185, Invitrogen), and protein reaction buffer kit (C10276, Invitrogen) as per manufacturer's protocol. Biotinylated proteins were precipitated/cleaned up via methanol/chloroform extraction, and analyzed by Western blotting. Membranes were blotted with anti-streptavidin-HRP antibody (Thermo N504). Unprocessed lysates were also run on Western blot and probed for Nrf2 and β-actin. These experiments were performed twice, Western blots from one experiment are shown. Cells were grown in RPMI 1640 containing 2 mm glutamine and 10% fetal bovine serum (heat-inactivated) for suspension lines and 5% fetal bovine serum for adherent lines and maintained in a humidified incubator (37 °C, 5% CO2). A Multidrop Combi Reagent Dispenser (Thermo Scientific; Waltham, MA) was used for plating cells into barcoded Falcon 384-well, black, clear-bottom plates (Corning; Tewksbury, MA; Catalogue No. 353962). On the following day, the compound was added using an Oasis integrated Liquid Handling Robot Platform (Dynamic Devices; Wilmington, DE) with LiCONiC incubators (LiCONiC Instruments; Woburn, MA) with automated barcode tracking. After 72 h, 25 μl CellTiter-Glo reagent was added using a MultiFlo Microplate Dispenser (BioTek; Winooski, VT). Cell lysis was induced by mixing for 30 min on an orbital shaker; plates were then incubated at room temperature for 10 min to stabilize the luminescent signal. Luminescence readout was done via a 2104 EnVision Multilabel Plate Reader (PerkinElmer; Waltham, MA). Experiments were performed as biological triplicates, and average IC50 and viability values used for analysis. Average doubling times were calculated based on 3–5 seeding densities. Cells were harvested at 80% confluency and doubling times calculated using the following formula, where Δ h = time of harvest – time of seeding in hours, C1 = initial cell seeding, C2 = total cell number at harvest = LOG10(2)/(LOG10(C2/C1)/Δ h). Nrf2-targeting and nontargeting control shRNA lentiviral vectors were constructed in the pINDUCER10 vector (18.Meerbrey K.L. Hu G. Kessler J.D. Roarty K. Li M.Z. Fang J.E. Herschkowitz J.I. Burrows A.E. Ciccia A. Sun T. Schmitt E.M. Bernardi R.J. Fu X. Bland C.S. Cooper T.A. Schiff R. Rosen J.M. Westbrook T.F. Elledge S.J. The pINDUCER lentiviral toolkit for inducible RNA interference in vitro and in vivo.Proc. Natl. Acad. Sci. U.S.A. 2011; 108: 3665-3670Crossref PubMed Scopus (443) Google Scholar) by annealing and ligation of DNA oligos into the XhoI-EcoRI sites of the vector. shRNA lentiviral vectors were cotransfected with pCMV-VSVG and pCMV-dR8.9 plasmids into 293T cells using the Lipofectamine 2000 reagent (Life Technologies). Forty-eight hours after transfection, viral supernatants were collected and used to infect A549 H1048 H441 and H460 cells (ATCC, Manassas, VA) in the presence of 8 μg/ml polybrene (EMD Millipore Corp, Billerica, MA). Cells with stable integrants were selected with puromycin for 10 days. Stably selected pools were maintained in puromycin selective media. In preparation for predictive modeling, several cell line features were gathered: tissue of origin, growth rate, and doubling time. All sample descriptors were combined into one matrix for predictive modeling. Expression, absolute DNA copy number and loss-of-heterozygosity, seeding density and growth rate were expressed as continuous values. Mutation status for individual hotspots, for gene hotspot and pathway metafeatures, and for the three categories of function-altering mutation were each individually coded as binary variables. Categorical variables, such as tissue type, were coded with one binary dummy variable per factor level. All features, including the binary features, were standardized prior to imputation and predictive modeling. Missing data, due to a lack of genomic data or insufficient sequencing coverage, were imputed using glmnet (19.Friedman J. Hastie T. Tibshirani R. Regularization paths for generalized linear models via coordinate descent.J. Stat. Softw. 2010; 33: 1-22Crossref PubMed Scopus (8846) Google Scholar). Missing values were initially replaced with the average value for their column (feature). Missing data were then imputed, one feature at a time, in a serial fashion, by treating one column as the response variable and all other feature columns as dependent variables. Training was performed using all samples with data for the response variable, and values were then predicted for the samples previously missing data. Classification and feature selection were performed using glmnet with binarized or continuous valued mean viability as the response variable. We selected an alpha value of 0.5 to increase the richness of the selected features and to increase interpretability. We selected the minimum lambda value within one standard deviation of the absolute minimum value under cross validation. This cross validation was performed with 25 randomly selected hold-out groups consisting of 20% of the samples. Hold-out groups were selected using the caret (20.Kuhn M. Wing J. Weston S. Williams A. Keefer K. Engelhardt A. Cooper T. Mayer Z. Kenkel B. Benesty M. Lescarbeau R. Ziem A. Scrucca L. Tang Y. Candan C. caret: Classification and regression training.R package version 6.0–58. 2015; Google Scholar) tool to select sample groups with sensitive/resistant proportions or continuous response variable distributions similar to that of the full set of samples. Following cross validation and lambda selection, a glmnet run with all samples was performed to select potential biomarkers. All features with a nonzero coefficient under the selected lambda were used for further analyses. Ranked absolute coefficient values were used to assess the relative importance of features. Recent publications have suggested that
Referência(s)