Background
The long lag time between initiation of cigarette smoking and cancer induction (estimated at 25 to 50 pack-years) [
1,
2] raises several fundamental questions concerning the eventual induction of tobacco-induced diseases for which there is little information: e.g., how does the lung adapt to the chronic assault of many decades of cigarette smoke (CS) exposure, what are the biological sequelae that occur in response to this adaptation and the continuous disruption of normal cellular homeostasis in the lung, and is this adaption a help or hindrance to lung cancer development? Our working hypothesis is that a) tobacco-induced lung cancer is a complex process in which numerous pro-survival cellular systems have important contributory functions that both augment and modify the central role played by tobacco carcinogens and reactive oxygen/nitrogen species, and b) CS temporally shapes the course of lung carcinogenesis through chronic activation, and eventual dysregulation, of normal cellular defense mechanisms. In our published [
3‐
6] and unpublished studies using high-density oligonucleotide arrays and other techniques to define relevant CS-induced alterations in gene/protein expression and function in lung cells, we have attempted to place the impacted genes into biological context by developing a plausible mechanistic model relating disruption of specific cellular circuits to pulmonary disease. Thus, in addition to revealing that CS affects the functioning of several important molecular pathways (e.g., redox homeostasis, detoxification of xenobiotics and cell cycle control), these data highlighted a potential role for the unfolded protein response (UPR) program.
Successful maturation of secretory and membrane proteins in the endoplasmic reticulum (ER) involves proper folding, assembly, and post-translational modification [
7]. A wide range of stressful situations (e.g., hypoxia, viral infection, alterations in glycosylation status, disruption of calcium homeostasis, and oxidative stress), can disrupt this maturation process, resulting in the accumulation of unfolded or misfolded proteins and causing ER stress [
8]. The ER attempts to attenuate this stress by activating an adaptive set of stress response signaling pathways termed the Unfolded Protein Response (UPR) [
8,
9]. The primary function of the UPR is to reduce the accumulation of aberrantly folded proteins in the ER and promote cell survival through a transient decrease in protein translation coupled with increases in the ER's capacity to refold and degrade these proteins[
10,
11]. If this pro-survival response fails to restore homeostatic equilibrium in the ER, a secondary response, triggered in part by the same ER stress sensors that activate the UPR program, promotes apoptosis and cell death. The importance of a properly functioning ER in maintaining cellular and tissue health is clear from the mounting evidence that a chronic increase in defective protein structures coupled with dysregulation within the ER can play a pathogenic role in diabetes, cardiovascular disease, Alzheimer's and Parkinson's syndromes, and cancer [
12‐
14].
An accumulation of misfolded proteins induces the dissociation of the ER-resident master chaperone regulator, BiP/GRP78 (Binding Immunoglobulin Protein/Glucose Response Protein 78), from three ER transmembrane sensor proteins: ATF6 (Activation of Transcription Factor 6), Ire1 (Inositol Requiring Enzyme 1α), and PERK (Protein Kinase R-like ER Kinase) resulting in activation of their respective molecular functions [
15,
16]. A second mechanism driving activation of these sensor proteins may also involve binding of unfolded protein domains to a peptide-binding groove in both IRE1 and PERK, and possibly ATF6 [
17]. Upon experiencing stress the 90 kDa ATF6 protein translocates from the ER to the Golgi where it is proteolytically processed to a functional 50 kDa transcription factor that binds to specific ER stress elements and directs the synthesis of chaperone proteins that mitigate protein misfolding through various mechanisms [
18,
19]. IRE1 has, in addition to a kinase domain, an endoribonuclease domain that splices an intron from the XBP1 (X-box Binding Protein) mRNA resulting in the synthesis of a transcriptional activator that modulates expression of a number of genes involved in ER homeostasis, DNA damage repair, and redox homeostasis [
20‐
22]. As well as being a transcriptional activator, IRE1 can also mediate the rapid degradation of a specific subset of mRNAs that would interfere with the coordinated reestablishment of normal ER function [
23]. In contrast to these pro-survival functions, IRE-1 can also directly regulate both pro- and anti-apoptotic circuits via activation of the stress kinase JNK-1 and mitogen-activated protein kinase ERK-1 [
24‐
26]. Unlike the transcriptionally active ATF6 and XBP1 proteins, PERK, upon release from BiP/GRP78, undergoes autophosphorylation and activation of a kinase function that phosphorylates the alpha subunit of eIF2 (eukaryotic translation initiation factor 2) and a transient repression of global protein synthesis. This temporary decrease in newly synthesized proteins entering the ER provides time to reestablish homeostatic equilibrium and resume normal protein maturation [
27,
28].
Since habitual U.S. smokers consume an average of 16.6 cigarettes/day [
29], and probably inhale somewhere between 120–180 puffs/day of a complex mixture of reactive gases and particulate matter composed of a wide range of entities (e.g., carcinogens and reactive oxygen/nitrogen radicals) that cause both immediate and delayed damage to proteins, lipids, and nucleic acids, it is likely that the lung cell is protected by multiple pro-survival mechanisms, including the UPR. Two recent studies have begun to amass data showing that CS induces elements of the UPR program in lung cells [
30,
31]. These reports speculate that elements of the UPR signaling pathway are relevant to human smoking-related diseases. However, there is minimal data detailing the impact of CS on all three UPR effector arms in both normal and malignant lung cells. Moreover, there is inconclusive evidence of UPR activation in human lung cancers. Thus, we were interested in determining the impact of CS on the UPR pathway in NHBE lung cells in vitro, and assessing the status of UPR activation in vivo in human lung cancers. We report here that CS exposure induces ER stress and triggers the UPR in both normal and malignant human lung cells. Of greater clinical relevance, however, is that an immunohistochemical analysis of a human lung tumor tissue array containing 110 lung cancer specimens and 10 normal lung tissue controls from a total of 120 patients showed a statistically significant increase in the total levels of phospho-eIF2α, BiP, and eIF2α in lung cancers compared to non-malignant lung cells.
These data implicate dysregulation of the UPR pathway in the pathogenesis of human lung cancers and indicate that phospho-eIF2α and BiP (and possibly eIF2α), may have diagnostic and/or therapeutic potential [
32,
33]. Furthermore, the activation of the UPR program via eIF2α phosphorylation indicates a previously unknown pathogenic effect of CS and suggests that chronic induction of one or more protein effectors of the UPR pathway may play an etiological role in lung cancer. Finally, the upregulation of several UPR regulators (in particular) in lung cancers may provide a pro-survival advantage by increasing cellular resistance to various cytotoxic stresses such as hypoxia, chemotherapeutic drugs, and immune attack [
34‐
38], especially in tumor cells that have one or more pro-apoptotic pathways disabled, which is a common feature of lung neoplasms [
39].
Methods
Cell Culture and Smoke Treatment
A549 cells were purchased from American Type Culture Collection (ATCC no. CCL-185, Manassas, VA) and were cultured in Ham's F12K medium with 2 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate (Gibco/Invitrogen, Carlsbad, CA) and supplemented with 10% fetal bovine serum (ATCC). NHBE cells from nonsmoking, nondiabetic donors were purchased from Cambrex Corporation (Walkersville, MD). Cells were cultured in complete Bronchial Epithelial Cell Growth Medium, prepared by supplementing Bronchial Epithelial Basal Medium with retinoic acid, epidermal growth factor, epinephrine, transferrin, T3, insulin, hydrocortisone, antimicrobial agents and bovine pituitary extract by addition of SingleQuots,™ (Cambrex Corporation). All cell cultures were treated before their sixth passage. All incubations were at 37°C in a humidified atmosphere of 5% CO
2 in air. CS treatment was performed as follows: cells were seeded into 35 mm Petri dishes (Fisher Scientific, Falcon #35-3001, Pittsburg, PA) at a density of 10
5 cells/dish and were typically at 70% confluency at the time of exposure to CS. At least three replicate dishes were treated for each condition, and each replicate was analyzed using a separate microarray (i.e., the RNA from the dishes was not pooled). The cell culture medium was replaced with 37°C Dulbecco's PBS (D-PBS) containing calcium and magnesium (Gibco/Invitrogen) for the smoke exposure. The covers were removed from the Petri dishes and they were placed in a smoke exposure chamber designed to deliver a consistent dose of diluted CS. CS was generated under Federal Trade Commission (FTC)[
40] smoking conditions (35 ± 0.3 cc puff, one puff every 60 seconds, 2-second puff duration with none of the ventilation holes blocked) using a KC 5 Port Smoker (KC Automation, Richmond, VA), from 2R4F reference research cigarette (designed to represent the average 'lights' cigarette marketed in the U.S. with FTC values of 9.7 mg 'tar' and 0.85 mg nicotine; University of Kentucky, Louisville, KY) using FTC machine smoking conditions, or two leading commercially available U.S. cigarette brands representative of either the 'lights' (FTC values of 11 mg 'tar' and 0.8 mg nicotine) or 'full-flavor' (FTC values of 15 mg 'tar' and 1.1 mg nicotine) styles of cigarettes. Cigarettes were smoked to within 3 mm of the filter tip. All cigarettes had been equilibrated at 23.9°C ± 1.1°C and 60% ± 2% relative humidity for a minimum of 24 hours and a maximum of 14 days. The smoke exposure chamber was designed to deliver smoke uniformly diluted with 5% CO
2 in air and passed through the cell exposure chamber at a constant flow rate of 500 cc/min. Briefly, each 35 cc puff was first drawn into a 250 cc round chamber containing 5% CO
2 in air and mixed via a stir bar. The standard smoke dilution used in most of our experiments was 35 cc delivered over 1 min in a 250 cc or 500 cc volume, and the intensity of exposure was varied by varying the length of time the cells spent in the exposure chamber (typically 15 min or 20 min). The time and distance that the smoke traveled from the end of the cigarette to the exposure chamber was minimized by using the shortest lengths of tubing possible between the parts of the apparatus. Mock-exposed cells were treated under identical conditions as the exposed cells except for the absence of a cigarette in the smoking port. Following treatment or mock treatment, the D-PBS covering the cells was aspirated and replaced with 1 ml per chamber of fresh culture medium at 37°C. The cells were placed in the 37°C, 5% CO
2 incubator for the times indicated.
Thapsigargin, tunicamycin, and dithiothreitol treatment
UPR-stimulating reagents thapsigargin, tunicamycin, and dithiothreitol (DTT) were obtained from Sigma, St. Louis, MO. Treatment conditions used in our experiments were 1 uM thapsigargin, 10 μg/mL tunicamycin, or 2 mM DTT added to the cell culture medium in DMSO (thapsigargin and tunicamycin) or water (DTT), and the cells were placed in the 37°C, 5% CO2 incubator for the times indicated. In experiments where thapsigargin or tunicamycin treatment followed smoke exposure, the PBS was removed from the cell cultures immediately after smoke exposure and replaced with fresh medium containing the thapsigargin or tunicamycin, and placed in the 37°C, 5% CO2 incubator for the times indicated.
Microarray analysis
Cells were harvested for total RNA extraction after either mock or smoke treatment. The medium was aspirated and the dishes were rinsed twice with 1 mL prewarmed PBS per dish. After the second rinse, 350 ul of Buffer RLT (Qiagen Inc, Valencia, CA) was added per dish. NHBE cell lysates were homogenized using a QIAshredder spin column and RNA extracted using Qiagen RNeasy spin columns according to the manufacturer's protocol (Qiagen). The eluted RNA was frozen and stored at -80°C. Microarray data was generated at Expression Analysis (Durham, NC). RNA integrity was assessed using capillary gel electrophoresis (Agilent BioAnalyzer, Agilent Technologies, Palo Alto, CA) to determine the ratio of 28s:18s rRNA in each sample. The RNA quality of all samples was extremely high with no RNA Integrity Number (RIN) being less than 9.5 out of 10.0 RIN units (a detailed explanation of RIN units can be found at
http://www.chem.agilent.com/scripts/LiteraturePDF.asp?iWHID=37507). Two micrograms of total RNA was reverse transcribed into double stranded cDNA using an oligo(dT)/T7 promotor chimeric primer, and in vitro transcribed using reagents provided by Affymetrix (Santa Clara, CA). Fragmented biotin-labeled cRNA at a concentration of 50 ng/ul was hybridized to Affymetrix Human Genome U133 Plus 2.0 GeneChip
® expression arrays according to the manufacturer's recommendations for a minimum of sixteen hours. Post-hybridization washing and staining was performed on the GeneChip
® Fluidics Station 450 and arrays were scanned with the GeneChip
® Scanner 3000 7G, under the control of the Affymetrix GeneChip
® Operating Software (GCOS). Preliminary expression data analysis was performed by GCOS. Two-Group Comparisons with Permutation Analysis for Differential Expression (PADE) was used to determine differential gene expression between CS-exposed and mock-exposed groups at each time point, each group consisting of multiple arrays. PADE analysis provides statistically valid summary measures including false discovery rates along with transcript sets that are typically much more useful and indicative of differential expression between groups than using techniques such as p-values alone, corrected p-values, or p-values with fold change estimates. A complete description of this type of analysis can be found at the Expression Analysis website:
http://www.expressionanalysis.com/pdf/PADE_TechNote_2005.pdf. A false discovery rate of 1% and a fold-change level of 1.5 were used as cutoff values. The full microarray data sets have been deposited according to MIAME standards on the NCBI GEO (Gene Expression Omnibus) website:
http://www.ncbi.nlm.nih.gov/geo/ under the GEO accession numbers GSE10700 and GSE10718.
PCR analysis of XBP1 splicing
RNA was harvested using the Qiagen RNeasy mini kit according to manufacturer's instructions immediately. First-strand cDNA synthesis was performed with the High Capacity cDNA Reverse Transcription kit for RT-PCR (Applied Biosystems, Foster City, CA). To amplify XBP1 mRNA (NM_005080), PCR was performed for 35 cycles (95°C for 30s; 58°C for 30s; 72°C for 1 min) using the PCR primers 5'-CTG GAA AGC AAG TGG TAG A-3' and 5'-CTG GGT CCT TCT GGG TAG AC-3' with AmpliTag Cold DNA polymerase (#N808-0241; Applied Biosystems). Fragments representing spliced and unspliced XBP1 (398 bp and 424 bp, respectively) were visualized on 2% agarose gels with ethidium bromide staining. NIH ImageJ software (
http://rsb.info.nih.gov/ij/, [
41]) was used to quantify gel band intensities.
Transfection of A549 Cells with ATF6 plasmid
A549 cells were seeded in 35 mm culture dishes at 1.2 × 105 cells per dish. At 24 hours after the seeding, cells were transfected with ATF6 expression plasmid ATF6/pCMV6-XL5 (OriGene Technologies, Inc., Rockville MD, Catalog # SC115551). Transfection was performed with Lipofectamine LTX (Invitrogen catalog # 15338-100) according to the manufacture's recommendation. Briefly, 2 μg of the plasmid DNA was diluted in 400 ul serum-free medium and mixed with PLUS Reagent (Invitrogen catalog #11514-015) at 1:1 ratio (DNA μg: PLUS vol. in μl). The mixture was incubated at room temperature for 5 min. Lipofectamine LTX was then added to the DNA/PLUS mixture and incubated at room temperature for an additional 25 min. At the end of the incubation, the original growth medium was removed and 2 ml of fresh pre-warmed growth medium was added to each dish. The DNA/PLUS/Lipofectamine complex was immediately added to the appropriate culture dishes, gently mixed and returned to the incubator for another 6 hours after which the medium was replaced again with fresh pre-warmed complete growth medium and the cells were returned to the incubator until smoke exposure or DTT treatment 24 hours post-transfection. Control cells were transfected with pCMV6-XL5 plasmid using the same procedure as above.
Transfection of A549 cells with PERK siRNA
A549 cells were seeded in 100 mm2 culture dishes at 9 × 105 cells per dish. At 20 hours after seeding, cells were transfected with PERK siRNA oligos (Dharmacon, Lafayette, CO, catalog # L-004883-00) or non-target siRNA oligos (Dharmacon, catalog # D-001810-10) using DharmaFECT1 transfection reagent (Dharmacon catalog # T-2001-02). Briefly, 21 μl of DharmaFECT1 was diluted in 679 μl of serum-free medium and was incubated at room temperature for 5 minutes. In a separate sterile tube, 70 μl of siRNA oligos (20 μM stock) was mixed with 630 μl of serum-free medium and incubated at room temperature for 5 minutes. The diluted DharmaFECT1 and diluted siRNA oligos were then mixed together and incubated at room temperature for another 20 minutes. At the end of the incubation period, 12.6 ml of complete growth medium was added to the mixture and 14 ml of this final mixture was dispensed to each of the 100 mm2 dish after the original growth medium was removed by aspiration. The dishes were returned to the incubator for another 6 hours before the oligo-DharmaFECT1 containing medium was replaced with fresh prewarmed complete medium. The culture was then allowed to grow in the incubator for another 20 hours when the cells were trypsinized and reseeded in 35 mm2 dishes at 1 × 105 cells per dish. The 35 mm2 dishes were returned to the incubator and the cells were allowed to grow for another 48 hours before smoke exposure or thapsigargin treatment.
Cell Lysis for Western Blots
Following treatment, the culture medium was aspirated, cell monolayers were washed twice with cold D-PBS, 2 ml/dish, and 1 ml (per 106 cells) of RIPA cell lysis buffer (Pierce, Rockford, IL) containing protease and phosphatase inhibitors was added to the monolayer. Cells were dislodged using a cell scraper, transferred to an eppendorf tube, vigorously pipetted, and left on ice for 25 minutes to allow complete lysis. The cell lysates were centrifuged at 10,000 × g for 25 minutes and the supernatants transferred to fresh eppendorf tubes. All manipulations were done at 4°C. Protein concentration of the lysates were determined against a commercially available protein standard using the BioRad protein assay kit (both from Bio-Rad Laboratories, Hercules, CA)
NHBE cell nuclear fractionation for ATF4 detection
NHBE cells (3rd passage) were seeded at 6 × 105 cells per 100 mm2 dish two days prior to the smoke exposure. On the day of smoke exposure cells were trypsinized and counted at 4 hours and 7 hours post-CS exposure or after 6 hours of incubation in media containing 1 μM thapsigargin. The total cell numbers were calculated and used for adjusting buffer volumes during cell compartment fractionation. The cellular compartment fractionation was carried out using the Qproteome Cell Compartment Kit (Qiagen catalog # 37502) according to manufacturer's instructions.
Western Blotting
Protein lysates were run on Criterion Tris-precast polyacrylamide gels (Bio-Rad Laboratories) with SeeBlue Plus2 and Magic Mark protein standards (both from Invitrogen, Carlsbad, CA) and transferred to a PDVF membrane using a Criterion Blotter apparatus (Bio-Rad Laboratories) for ~40 min at room temperature using a transfer buffer containing 10% methanol (Bio-Rad Laboratories). Primary antibodies were from the following sources: Cell Signaling Technology, Danvers, MA (eIF2α cat#9722, phospho-eIF2α cat#3597, BiP cat#3183, Lamin A/C cat#2032 & α-tubulin cat#2144), GeneTex, Inc., San Antonio, TX (eIF2A (phospho S52)Cat # GTX24837), Santa Cruz Biotechnology Inc., Santa Cruz, CA (ATF3 cat#sc-188, ATF4 cat#sc-200), Abcam Inc., Cambridge, MA (GAPDH cat#ab8245, α-tubulin cat#24246), Imgenex, San Diego, CA (ATF6 cat#IMG-273). Membranes were blocked with 5% BSA (for eIF2α, phospho-eIF2α, ATF6, BiP and α-tubulin), 5% BLOTTO (Pierce) for ATF3 and ATF4, and T-20 (Pierce) for GAPDH for 1 hour at room temperature, and then washed with TBST (Bio-Rad) for 3 × 10 min. Membranes were placed in 5% BSA, 5% Blotto or T-20 buffer solution containing the primary antibody (1:1000 dilution for anti-eIF2α, anti-phospho-eIF2α, BiP, and Lamin A/C, 1:200 dilution for ATF3 and ATF4, 1:500 for ATF6, 1:2000 for α-tubulin, and 1:40,000 for GAPDH) and incubated at 4°C overnight with gentle shaking, followed by washes in TBST for 3 × 10 min. Membranes were then placed in 5% BSA, 5% Blotto or T-20 solution containing anti-rabbit-HRP, or anti-mouse-HRP secondary antibody (Cell Signaling Technology) at a 1:2000 dilution and incubated for 1 hour at room temperature with gentle shaking, after which they were washed with TBST for 3 × 10 min. Blots were developed using Western Lightning chemiluminescence reagent (Perkin-Elmer, Boston, MA) according to manufacturer's instructions. NIH ImageJ software (
http://rsb.info.nih.gov/ij/, [
41]) was used to quantify gel band intensities.
Tissue Microarrays (TMAs)
Commercial Cancer/Normal Lung Tissue arrays were acquired from Asterand (Detroit, MI). Arrays were constructed from formalin-fixed paraffin-embedded tissues representing 120 normal and lung cancer cases. Each tissue microarray contained 10 normal cases and 110 tumor cases from a broad range of lung cancer types. Every tissue sample was reviewed by a board certified pathologist to confirm tissue type, diagnosis and optimal region for coring before use. Cores of 0.6 diameter were removed from each selected sample and placed in a paraffin block using a Beecher Manual Array instrument. Each case was represented in triplicate for a total of 360 cores. An H&E slide of each array post construction was completed as a quality control measure. Table
1 lists the pathological characteristics of these 120 formalin-fixed, paraffin-embedded archival tissues from patients.
Table 1
Clinicopathological characteristics of the assessed lung cancers.
Age at excision: 32–76 (mean = 60.2 yrs) | | |
Small Cell Carcinoma | Small cell carcinoma | 5 |
Non-Small Cell Carcinoma | | 93 |
| Adenocarcinoma | 39 |
| Adenosquamous carcinoma | 12 |
| Squamous cell carcinoma | 37 |
| Large cell carcinoma | 4 |
| Bronchioloalveolar carcinoma | 1 |
Mixed Carcinoma Types | | 12 |
| Carcinoid tumor | 2 |
| Carcinoma | 3 |
| Clear cell carcinoma | 1 |
| Undifferentiated carcinoma | 2 |
| Malignant mesothelioma | 4 |
Antibodies and immunohistochemical analysis
Identical TMAs were immunostained with one of the following antibodies: a) a rabbit monoclonal anti-BiP antibody (Cell Signaling Technology, Danvers MA, cat. # 3177) recognizing BiP (C50B12) from rabbits immunized with a synthetic peptide derived from the sequence around Gly584 of human BiP (used at 1:200 dilution and a concentration of 1 μg/ml); b) a mouse monoclonal anti-EIF2 antibody (Cell Signaling Technology cat. #2103) recognizing eIF2α (L57A5) from mice immunized with purified recombinant human eIF2α (used at 1:50 dilution and a concentration of 0.1 μg/ml); c) a rabbit monoclonal anti-phospho-EIF2 antibody (Cell Signaling Technology cat.# 3597) recognizing Ser51 (used at 1:50 dilution and a concentration of 1 μg/ml); d) an irrelevant rabbit IgG antibody control (Jackson ImmunoResearch, West Grove, PA) (used at a concentration of 1 μg/ml). Prior to antibody staining, slides were deparaffinized in 3 changes of xylene and rehydrated in graded ethanol. All slides were subjected to antigen unmasking in 10 mM citrate buffer, pH 6.0, quenching in 3% H2O2 and blocking prior to incubation with primary antibodies overnight at 4°C. Following incubation with the primary antibodies, detection was performed using the ABC Elite Kit (Vector Laboratories, Burlingame, CA) with either a biotinylated goat anti-rabbit secondary antibody or a biotinylated goat anti-mouse secondary antibody (Jackson ImmunoResearch). Both eIF2α mouse monoclonal and BiP rabbit monoclonal antibodies from Cell Signaling Technology had previously been validated at the company using human breast carcinoma samples and multi-tissue arrays. The sequence specificity of the BiP antibody was verified using peptide blocking. The EIF2 ?antibody was raised against a fusion protein and thus was not validated with peptide blocking. The antibody did give the expected expression and subcellular localization. Human breast carcinoma slides used as a control were obtained from Newcomer Supply (Madison, WI). The anti-phospho-eIF2α antibody has been validated at Cell Signaling Technology using 3T3 -/+ thapsigargin cell pellets, human breast carcinoma samples and a multi-tissue array. Phospho-specificity of the antibody was confirmed using lambda phosphatase pre-treatment of tissue sections, and sequence specificity was verified using peptide blocking.
Clinicopathological variables
The following variables were correlated with protein expression: a) tumor stage, b) age at tumor excision, c) gender, d) ethnicity, e) smoking status (i.e., never user, occasional user, former user, current user); f) number of cigarettes per day, and g) smoking duration. Staging adhered to the AJCC pTNM (tumor-node-metastasis) staging system and included both clinical and histological data.
Staining criteria and statistical analyses
Tumor cells and non-tumor cells were scored separately for percent immunohistochemical reactivity and staining intensity for BiP, eIF2α, and phospho-eIF2α expression. The percentage of cells staining positive was assigned one of the following grades: (0) < 5%; (1) 5–25%; (2) 25–50%; (3) 50–75%; (4) > 75%. Staining intensity was independently graded on the following scale: (0) none; (1) weak; (2) moderate; (3) strong; (4) intense. In order to more accurately reflect the synergistic contribution of both criteria to overall protein expression, the product of the percentage and intensity scores was used to generate an immunohistochemical staining index (ISI) for each cellular component (i.e., tumor and non-tumor) of an individual tissue specimen [
42,
43]. Staining was independently assessed by two pathologists. Measurements were made in triplicate. For purposes of statistical analysis, triplicate measurements were represented by their mean. To compare mean ISI levels between each carcinoma diagnostic group with the normal group, one-way analyses of variance followed by Dunnett's tests were used. To compare the ISI levels in the tumor and normal cell compartments of each specimen, repeated measures analyses of variance followed by Tukey's tests were used. A result was considered statistically significant if the resulting
P value was less than or equal to 0.05.
Discussion
The pyrolysis process in cigarettes generates a highly complex mixture of reactive gases and suspended particulate matter containing a wide range of carcinogens, tumor promoters, toxins, and free radicals which cause direct and indirect damage to the genome [
77,
78], transcriptome [
79,
80], and proteome [
81] of cells in the aerodigestive tract of smokers on a daily basis [
82,
83]. This incessant cycle of tissue injury and repair is presumed to be a major contributor to the development of lung cancer [
75,
84‐
86]. However, a precise biological understanding of the nature and temporal sequence of specific damaging events that drive the formation and progression of this disease remains elusive [
87‐
91]. Compelling data support the conclusion that one prominent form of CS-induced airway damage and a key etiological factor in lung cancer is chronic oxidative stress and resulting inflammation [
72‐
75,
84‐
86,
92,
93]. For example, 1) lung cancer is increased in patients with chronic inflammation [
84,
94‐
96]; 2) polymorphisms and mutations in genes regulating the inflammatory process are linked to lung cancer risk [
97,
98]; 3) gene and proteomic studies of smoke-exposed airway epithelium consistently shows the upregulation of genes that respond to oxidants [
30,
31,
99]; 4) the cyclooxygenase-2 (COX-2) isoenzyme, a key player in diverse pro-inflammatory conditions, is frequently up-regulated in lung neoplasms and correlates with a poor prognosis [
100,
101]; 5) non-steroidal anti-inflammatory drugs (e.g., selective COX-2 inhibitors such as Celecoxib) can reduce the relative risk of lung cancer [
102‐
105]; 6) CS can depress levels of endogenous antioxidants such as glutathione [
74,
106‐
108]; 7) synergy between pro-oxidant beta-carotene cleavage products and resulting oxidative stress is believed responsible for the increased rates of lung cancer in long-term smokers enrolled in both the ATBC (Alpha-Tocopherol Beta-Carotene) and CARET (beta-Carotene and Retinol Efficacy) chemoprevention trials [
109]; 8) a mouse model system suggests that oxidative stress induced in the lung by CS vapor phase components is a key player in lung oncogenesis [
85,
110]; and 9) we have previously shown that an early genomic defect caused by the copious amounts of free radical generators in CS is the induction of DNA double-strand DNA breaks (DSBs), which are probable tumorigenic lesions in multiple cancers including those of the lung [
111‐
113], and that free radical scavenging antioxidants can prevent these DSBs [
3,
5].
Although a detailed understanding of the molecular mechanisms that directly link oxidative stress, inflammation, and CS-induced pathologies is still lacking, a wealth of data suggest that the large amounts of free radicals and reactive species (such as
•O
2
-,
•NO,
•OH, etc) in the gas phase, and more stable organic reactive species/oxidants (such as phenols, hydroquinone, epoxides, etc.) in the particulate phase of CS, overwhelm the respiratory tract's steady-state antioxidant capacity causing a marked imbalance in its redox status [
114‐
116]. This persistent exogenous source of oxidative stress is amplified and augmented by an endogenous chronic host inflammatory response at the sites of tissue damage that provides additional quantities of reactive oxygen and nitrogen compounds (e.g., H
2O
2 and
•NO), and reactive intermediates such as peroxynitrite (ONOO
- [
86,
117,
118]. Combined, these sources of reactive species can cause significant cellular and tissue damage that the cell responds to in several ways. One mechanism is to alter the expression of genes that attenuate the effects of oxidative stress. For example, as shown in additional file
1 –
Supplemental Table S1: Cigarette smoke signature genes, CS exposure induces the expression of: a) thioredoxin reductase 1, a component of a ubiquitous thiol oxidoreductase system that protects the cell from oxidative stress; b) heme oxygenase I (HO-1), an enzyme that catabolizes heme containing proteins with the subsequent production of free iron, CO, and bilirubin; c) ferritin, which sequesters reactive iron molecules and d) NAD(P)H:quinone acceptor oxidoreductase 1 (NQO1), a cytosolic flavoenzyme that catalyzes reduction of quinones to hydroquinones and is part of the oxidative stress response [
119]. Each of these CS-responsive genes is frequently upregulated in a range of cancers including lung carcinomas [
120‐
123]. Our data further show that CS causes an increase in the expression of genes pivotal to redox homeostasis as well as the synthesis of glutathione: gamma glutamylcysteine ligase, catalytic and modifier subunits. Chronic depletion in glutathione related antioxidant enzymes by CS can lead to lung damage and disease [
124,
125]. Since proteins are major targets of oxidative damage, another predictable biological result would be an accumulation of misfolded, aggregated, or cross-linked proteins in the ER of the respiratory tract which responds by activating a wide array of stress responsive genes in order to restore homeostasis [
126]. Support for this hypothesis comes, in part, from the data presented in this study showing that CS causes ER stress and activates the UPR pathway via phosphorylation of eIF2α in a PERK-dependent manner in human lung cells, which supports a similar conclusion reached in two recent papers [
30,
31]
Our data further show that reactive radical species in both whole smoke and the vapor phase are a dominant CS-component causing ER stress [
127]. A recent publication by Hengstermann and Muller has concluded that the CS gas phase component acrolein is a major inducer of ER stress [
31]. The ability of both GSH (the major intracellular antioxidant) and NAC, a synthetic acetylated form of the amino acid L-cysteine and a thiol-containing antioxidant, to scavenge CS-generated reactive oxygen species (ROS) (e.g.,
•O
2
-, H
2O
2, and
•OH) [
128,
129] or sequester reactive CS compounds that spawn free radicals intracellularly (
e.g., aldehydes, epoxides, quinones, etc.) further supports the mechanistic relationship between CS-induced reactive species and ER stress [
130]. In these in vitro experiments it is perhaps not surprising that the particulate phase of CS did not induce the UPR. This is presumably because many of the compounds that have reactive potential in the particulate phase require either enzymatic activation or generation via quinone/hydroquinone redox coupling in the presence of oxygen over time [
127,
131]. Thus, the particulate phase does not result in a bolus of free radicals at short time periods but rather chronic low-level radical generation over extended time periods [
131,
132]. Vapor phase components are inherently smaller/more reactive than particulate phase components and therefore have a higher likelihood of reacting with biological components immediately upon exposure as well as crossing phospholipid membranes more quickly than larger molecules. Consequently, it is probable that since we exposed cells to CS for only 15–20 minutes, there is insufficient time for reactive species to form within the cell and induce the UPR pathway. Presumably, this is not the case in vivo where chronic cigarette smoking would provide sufficient time for the reactive species in the particulate phase to impact protein structures and induce molecular circuits like the UPR.
Another interesting observation to emerge from this study is that CS significantly suppressed the splicing of the XBP1 mRNA to its active truncated form. The ability of CS to modulate the expression of XBP1, if mirrored in vivo in smokers, may have significant ramifications for the development and progression of lung neoplasms since XBP1 not only has a primary role in maintaining ER homeostasis [
22,
133], but also important regulatory functions in DNA damage and repair pathways, redox homeostasis and oxidative stress responses [
22]. Thus, impairment of this key effector arm of the UPR pathway could have considerable detrimental short and long-term effects in smokers. The mechanism by which CS suppresses XBP1 splicing is not obvious. It is possible that one or more components of CS can attack the structure of the IRE1 transmembrane protein thereby disabling its ribonuclease function and subsequent splicing of the XBP1 mRNA to its transcriptionally active isoform. Another possibility is that CS activates a protein that inhibits elements of the ER-stress induced UPR pathway. For example, the Bax inhibitor-1 (BI-1) protein has been shown to suppress apoptosis [
134], ER-stress related protein expression, and XBP1 splicing [
135]. Moreover, it has been shown that, depending upon the subtype, some 43% to 82% of lung adenocarcinomas overexpress BI-1 [
136]. Eliminating or down-regulating proapoptotic signals are a major step in the evolution of most, if not all, human malignancies. We are currently assessing if CS suppresses XBP1 splicing by inducing or augmenting BI-1 expression in lung cells. A recent paper has presented evidence showing that mouse 3T3 cells exposed to aqueous extracts of CS displayed only minute amounts of spliced XBP1 mRNA [
31]. The conclusion of this paper was that CS did not activate the IRE1 sensor and therefore resulted in negligible XBP1 splicing. In contrast, we provide evidence that CS actively suppresses the splicing of XBP1. If this phenomenon occurs in vivo in smokers, it could have physiological relevance in terms of altering the balance between cell survival and cell death. Lin et al have recently shown that IRE1 and ATF6 activities were attenuated by persistent ER stress in human cells resulting in increased cell death [
61]. Clearly, further research to determine if IRE1 activation and/or XBP1 splicing is corrupted in actual smokers would be valuable. Finally, we showed that the third effector arm of the UPR signaling pathway, ATF6, is also activated by CS as evidenced by cleavage of the ATF6 protein and the appearance of downstream transcriptional targets of ATF6 such as XBP1 and BiP.
If activation of the UPR program occurs in smokers on a daily basis due to oxidative stress, it could evolve into a long-term problem either because of persistent damage directly related to its activation, or because as a functional barrier to transformation and tumor progression, it provides an attractive target for disablement. Either scenario would allow the respiratory cell to sustain additional damage to mechanisms that maintain lung homeostasis which could eventually lead to neoplastic transformation [
137,
138]. A number of studies have suggested that habitual activation of UPR plays a major role in etiology of cancer as well as many other diseases [
35,
36,
139,
140]. Tumor cells are subjected to considerable internal stress due to genetic instability, hypoxia, signaling distortion, immune attack, and disorganized cross-talk with the surrounding normal microenvironment [
141]. Many of these stressful situations can cause ER dysfunction from which the cell must defend itself. Consequently, a prevailing hypothesis is that upregulation of UPR components in malignant cells provides an anti-apoptotic, pro-survival advantage by increasing resistance to ER stress induced by endogenous sources (e.g., hypoxia, genetic instability, etc.) [
10,
36,
64,
142] or exogenous sources (such as chemotherapeutic drugs) [
32,
143]. Evidence to support this idea comes from numerous studies showing the increased expression of one or more UPR-relevant proteins in multiple tumors including those of the breast (i.e., BiP [
144]), liver (ATF6, XBP1 and BiP [
145]), stomach (BiP [
146]), brain (BiP [
143]), esophagus (Grp94 [
147]), and lung (BiP and Grp94 [
148,
149]). Our analysis of lung cancer tissue specimens is also consistent with this hypothesis. For example, we found that compared to non-malignant lung tissue, there is a significant increase in expression of phospho-eIF2α in a majority of cases of NSCLCs (55.9%;
p = 0.0025) but not in either SCLCs or MCs. We further found a significant increase in expression of BiP in a majority of cases of NSCLCs (87.5%;
p = < 0.0001) and MCs (100%;
p = < 0.0001), but not in SCLCs. The eIF2α protein was also overexpressed in a majority of cases of both NSCLCs (60.6%;
p = 0.0002) and MCs (87.5%;
p = 0.0017), but not in SCLCs. Rosenwald et al. have previously shown that eIF2α is frequently increased in bronchioloalveolar carcinomas (BAs) but only rarely in squamous cell carcinomas (SCCs) [
150]. Since both BAs and SCCs are subgroups of NSCLCs that share the histological feature of being derived from lung epithelium [
151], these data differ somewhat from ours which showed a statistically significant increase in eIF2α expression in the tumor cell compartments of SCCs as well as in the other major NSCLC histological subtypes (i.e., adenocarcinoma, adenosquamous carcinoma, and large cell carcinoma) for which we had a sufficient number of cases. Resolution of the differences between our results and that of Rosenwald et al. awaits further research. However, an important caveat to our data is that the small number of SCLC cases available for assessment (i.e., 5) could potentially obscure any reliable protein expression differences.
Although our data support the conclusion that modulators of the UPR pathway are chronically impacted in this tumor type, aside from a modestly significant increase in BiP in the NSCLC diagnostic group with increasing age (p = 0.0187), we found no obvious correlation with any histopathological criteria such as gender, pathologic stage, histological type, or TMN-status and the increased expression of eIF2α, phospho-eIF2α, and BiP. Uramoto et al. [
149] similarly found no significant difference between BiP expression and any clinicopathological parameter. A recent lung cancer study by Wang et al. [
148], though finding no correlation of increased expression of BiP with pathologic tumor type, did relate augmented expression in less differentiated tumors and in more advanced stage III tumors, both aspects predicting a poorer prognosis. However, Uramoto et al. [
149] found increased expression of BiP in a majority of lung cancers but these patients had a better prognosis than those with BiP-negative cancers. In our current study, we could not determine the relationship between expression and prognosis due to a lack of detailed clinical outcome data on the patients from whom the tumor specimens we assessed originated.
Paradoxically, although our study showed that increased expression of eIF2α, phospho-eIF2α, and BiP are pathogenic features of lung cancers, none of these proteins identified lung cancers as having arisen in a smoker or nonsmoker. Does this mean that the induced expression of these UPR-related proteins is not directly related to CS exposure or, alternatively, that their induction is a general convergent feature of lung cancers regardless of the provoking stimulus? Since oxidative stress resulting from passive cigarette smoking is certainly one etiological factor in the development of lung cancers in nonsmokers [
75,
84,
85,
92,
152,
153] the latter interpretation is more likely. It is also possible that the UPR is induced in CS-exposed lung cells prior to malignancy but also subsequent to the development of a cancer that has regions of hypoxia. For example, it is well documented that hypoxia is present in a majority of human solid tumors, including lung cancers, and that hypoxic regions can have selective resistance to various therapeutic modalities [
53,
154]. Thus, one prosurvival mechanism of hypoxic tumor cells is to diminish protein translation and energy utilization, which can occur as a direct result of activation of the UPR pathway. Accordingly, induction of UPR via eIF2α phosphorylation is required for hypoxic cell survival and tumor growth [
53]. In addition, hypoxia can also induce the expression of BiP [
155]. Thus, it remains to be determined if the expression of peIF2α and BiP that we observed in a majority of human lung cancers occurs relatively late in their evolution as the result of hypoxic conditions, or reflects an early activated prosurvival mechanism in asymptomatic lung cells undergoing chronic ER stress due to CS or some other environmental contaminant. While our data to date do not strongly support either possibility, a recent study using a proteomic approach showed that lung samples from chronic smokers demonstrated a number of differentially expressed proteins compared to nonsmokers [
30]. For example, several UPR proteins, including BiP and calreticulin, were found to be up-regulated in smokers. The conclusion of these data was that activation of UPR by CS may protect the lung from oxidant injury and the development of chronic obstructive pulmonary disease, a strong risk factor for the development of lung cancer.
In summary, while more studies are needed to clarify how chronic activation, expression, or dysregulation of key UPR-regulated proteins impact the trajectory of CS-induced lung disease, it seems highly likely that the UPR pathway is a one of several molecular mechanisms promoting tumor cell evolution that could be attenuated or reversed resulting in a more efficacious treatment strategy for lung cancers [
87‐
90,
156‐
158]. Targeting one or more UPR effectors, either unilaterally or in combination with conventional cytotoxic drugs, may be a particularly important treatment opportunity since the current standard of care for patients with advanced lung cancer remains disappointing [
33,
35,
36,
137,
156]. Direct support for this proposition has recently come from a study showing that bortezomib (PS-341, Velcade™), a potent proteasome inhibitor currently approved for the treatment of multiple myeloma, sensitizes pancreatic cancer cells to ER stress-induced apoptosis and strongly enhances the anticancer activity of cisplatin [
159].
Authors' contributions
EJ performed the microarray analysis, participated in the design of the study, and the writing of the manuscript; AS, LS, JY, and DG contributed to the data generation and participated in study design; and APA participated in the design, coordination, and performance of the study, assisted in the writing of the manuscript, and funded the study. All authors have read and approved the final manuscript.