Functional Genomic Responses to Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) and CFTRΔ508 in the Lung*

Cystic fibrosis (CF), a common lethal pulmonary disorder in Caucasians, is caused by mutations in the cystic fibrosis transmembrane conductance regulator gene (CFTR) that disturbs fluid homeostasis and host defense in target organs. The effects of CFTR and Δ508-CFTR were assessed in transgenic mice that 1) lack CFTR expression (Cftr-/-); 2) express the human Δ508 CFTR (CFTRΔ508); 3) overexpress the normal human CFTR (CFTRtg) in respiratory epithelial cells. Genes were selected from Affymetrix Murine Gene-Chips analysis and subjected to functional classification, k-means clustering, promoter cis-elements/modules searching, literature mining, and pathway exploring. Genomic responses to Cftr-/- were not corrected by expression of CFTR Δ508. Genes regulating host defense, inflammation, fluid and electrolyte transport were similarly altered in Cftr-/- and CFTR Δ508 mice.CFTRΔ508 induced a primary disturbance in expression of genes regulating redox and antioxidant systems. Genomic responses to CFTRtg were modest and were not associated with lung pathology. CFTRtg and CFTRΔ508 induced genes encoding heat shock proteins and other chaperones but did not activate the endoplasmic reticulum-associated degradation pathway. RNAs encoding proteins that directly interact with CFTR were identified in each of the CFTR mouse models, supporting the hypothesis that CFTR functions within a multiprotein complex whose members interact at the level of protein-protein interactions and gene expression. Promoters of genes influenced by CFTR shared common regulatory elements, suggesting that their co-expression may be mediated by shared regulatory mechanisms. Genes and pathways involved in the response to CFTR may be of interest as modifiers of CF.

Cystic fibrosis (CF) 2 is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) that regulates chloride and fluid transport in various epithelia, including the lung (1). The CFTR gene encodes a large transmembrane-spanning protein that is distributed primarily in the apical regions of secretory epithelia (2). CFTR mediates cAMP-dependent chloride transport that plays a critical role in the regulation of fluid homeostasis in the airway and in other organs. Although CF affects the function of various organs, pulmonary disease represents the most common source of morbidity and mortality associated with the disorder. CFTR mediates fluid and chloride transport in epithelial cells, and interacts directly with the cytoskeleton and with various cellular proteins, including receptors, chaperones, and other transport proteins (3). Numerous mutations in the CFTR have been identified and associated with CF in humans (4). The most common mutations associated with CF include both missense and deletion mutations that result in the synthesis of CFTR proteins that are mistrafficked and degraded. CFTR ⌬508 is the most common mutation causing CF in Caucasians and is caused by the deletion of phenylalanine at position 508 in the polypeptide, resulting in its misfolding and degradation and failure of its association with the apical membrane (5). The severity of CF-related lung disease is highly variable, even among individuals bearing the same mutations, indicating that the disorder is influenced by complex, hereditary, and environmental factors (6,7). Similarly, marked variability in the severity of gastrointestinal and pulmonary disorders is observed among various strains of mice bearing CFTR mutations (8). Taken together, these findings support the concept that the cellular dysfunction associated with CFTR mutations may be influenced by the expression of modifying genes that compensate for or exacerbate the cellular functions disrupted by the lack of CFTR or the presence of misfolded CFTR. Previous in vivo and in vitro studies indicated that the lack of CFTR influences complex transcriptional responses that may influence the pathogenesis of CF-related disorders (9,10). Lack of CFTR in the lung is associated with changes in the expression of genes regulating fluid and electrolyte transport, intracellular trafficking, inflammation, as well as of proteins that directly bind and interact with CFTR, indicating that CFTR functions within a complex, transcriptional, and functional milieu within the cell (10). In the present study, functional genomics were used to identify cellular responses caused by lack of CFTR expression, CFTR ⌬508 expression, and increased expression of wild-type human CFTR in respiratory epithelial cells of the adult mouse lung.

EXPERIMENTAL PROCEDURES
Transgenic Mice and Animal Husbandry-The genotypes and characteristics of the mice used in this study are listed in Table 1. Briefly, wild-type (FVB-N) mice (Cftr ϩ/ϩ ) were compared with FABP-hCFTR tg , Cftr Ϫ/Ϫ mice (Cftr Ϫ/Ϫ ) (11) and mice expressing CFTR ⌬508 in the respiratory epithelium in Cftr Ϫ/Ϫ background, and mice in which the normal human CFTR was selectively expressed in bronchiolar and alveolar respiratory epithelial cells (CFTR tg ) in Cftr ϩ/ϩ background, previously identified as J4 transgenic mice (12). Mice have been backcrossed for a number of years in the FVB-N strain. CFTR ⌬508 mice were produced by nuclear injection of oocytes using standard procedures in FVB-N mice (13). The construct consisting of the 3.7hSP-C promoter element controlling the human CFTR ⌬508 in which a hemagglutinin tag was inserted at the C-terminal end of the CFTR, was kindly provided by Dr. Ray Frizzell, University of Pittsburgh).
RT-PCR Analyses-hCFTR, mCftr, CFTR ⌬508 /Cftr ϩ/ϩ , and CFTR ⌬508 mRNAs were assessed using regular PCR. Whole lung tissue was isolated from freshly sacrificed adult mouse from each genotype. Total RNA was isolated from homogenized lung tissue using TRIzol reagent (Invitrogen) based on the manufacturer's instruction. Lung RNA was isolated from sexand age-matched littermates to minimize the potential influence of strain differences. Individual mouse lung was carefully dissected, and the conducting airways and mediastinal structures were removed. PCR was performed on cDNAs prepared from these samples at the conditions of 95°C ϫ 30 s, 56°C ϫ 30 s, and 72°C ϫ 1 min for total of 35 cycles. The primers for mCftr and hCFTR were designed to recognize sequences unique to the mouse endogenous Cftr or the human CFTR sequence expressed as transgene. The primers for mCftr were forward: 5Ј-ACCT-CATTGCCTCACGGAGACATC-3Ј and reverse: 5Ј-AACAAGTCCCA-GAGAAGCCCCATC-3Ј. The primers for hCFTR are forward: 5Ј-TAAACCTACCAAGTCAACCA-3Ј and reverse: 5Ј-AATTCCATG-AGCAAATGTC-3Ј. The primers for ␤-actin are forward: 5Ј-TGGAAT-CCTGTGGCATCCATGAAC-3Ј and reverse: 5Ј-TAAAACGCAGC-TCAGTAACAGTCCG-3Ј. The mRNA levels of selected genes were confirmed by LightCycler reverse transcription-PCR using the following probes for mouse Hsp40, Hsp70, Foxf1a, and Thsd1. The primers for Hsp40 were forward: 5Ј-TTC CAA TGG GTA TGG GTG GC-3Ј and  reverse: 5Ј-CCG CTT GTG GGA GAT TTT CATC-3Ј. The primers for  Hsp70 were forward: 5Ј-GCC AAA AAT GCG GTG GAG TC-3Ј and  reverse: 5Ј-CTT TCT CTG CCA TCT GGT TTCG-3Ј. The primers for  Foxfla were forward: 5Ј-CCT TCA CCA AAA CAG TCA CAA CGG-3Ј  and reverse: 5Ј-TCA CCT CAC ATC ACA CAC GGC TTG-3Ј. The primers for Thsd1 were forward: 5Ј-CCA GAA AGT CAG TCA GCT TAG  CC-3Ј and reverse: 5Ј-TTG GAG CCG AAG TAT CCT GG-3Ј. Lung Histology and Immunofluorescent Staining of CFTR and CFTR ⌬508 -MLE15 cells were seeded into 6-well plates containing 3 coverslips in each well. 1 g of 3.7hSP-C/CFTR-FLAG or 1 g of 3.7hSP-C/CFTR ⌬508 -FLAG plasmids were transfected to each well using Fugene6 transfection reagent (Roche Applied Science). After 48 h, coverslips holding transfected cells were rinsed with phosphate-buffered saline and fixed with 4% paraformaldehyde for 20 min. After incubation with 0.1% Triton in phosphate-buffered saline for 20 min, the cells were blocked with 5% goat serum for 1 h and incubated with anti-FLAG M2 antibody for 3 h and Texas red-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 h. Immunofluorescent staining on mouse lung was performed with M3A7 antibody (Chemicon International, Temecula, CA), as previously described (14). Immunostaining for CFTR and RNA analysis of J4 mice expressing the normal hCFTR in the lung were reported previously (12). Lung histology was performed on mice of all genotypes after inflation fixation at 25 cm in 4% paraformaldehyde. Tissue was embedded in paraffin, sectioned, and stained with hematoxylin-eosin using standard methods.
RNA Microarray and Data Analysis-Microarray experiments were conducted to identify transcriptional responses caused by 1) expression of CFTR ⌬508 in the Cftr Ϫ/Ϫ background (herein termed CFTR ⌬508 ) and 2) increased expression of normal human CFTR in the lungs of wildtype mice (CFTR tg ) under control of the human Sftpc promoter (J4 mice), as previously described (12). RNA microarray findings were compared with previous data generated from the lungs of Cftr Ϫ/Ϫ mice (10). Table 1 defines the genotypes of the mice used in the study. RNAs were isolated from lungs of adult littermates as previously described (10) and RNAs were hybridized to murine genome U74AV2 or murine genome MOE430A GeneChip (Affymetrix Inc.). Three biological replicates were used for each comparison. Total lung RNA was subjected to reverse transcription using oligo(dT) with T7 promoter sequences attached, followed by second strand cDNA synthesis. Antisense cRNA was amplified and biotinylated using T7 RNA polymerase using protocols recommended by Affymetrix. Affymetrix MicroArray Suite version 5.0 was used to scan and quantitate intensity using default scan settings. Data were first normalized using RMA (Robust Multichip Average model) (15). Normalized data were then further analyzed using Gene Spring 6.0 (Silicon Genetics, Inc.) and SAM (Significance Analysis of Microarrays) (16). Differentially expressed genes were identified by Student's t-test at p value Յ 0.05 with estimates of false discovery rate Յ 10% (SAM). In addition, Affymetrix "Present Call" in at least two of three replicates and coefficient of variation among replicates Յ 50% were set as a requirement for gene selection.
K-Means Clustering-Differentially expressed genes identified in lungs from CFTR ⌬508 and Cftr Ϫ/Ϫ mice were mapped to the MOE430 genome base on the U74V2-MOE430 best match (www.affymetrix.com/support/) and compared with genes selected from RNAs from the SP-C-CFTR tg mice using modified Best K-Means clustering (Silicon Genetics BioScript 2.1) with Pearson's correlation as similarity measures. The number of clusters was chosen empirically based on the combination of percent variance explained (select k-Means results with higher percentage explained variability) and visual inspection. To control for age, Cftr Ϫ/Ϫ mice were compared with other genotypes at 6 weeks of age. Genes with no match between U74V2 and MOE430 were excluded from the cluster analysis. To reduce the background and ensure the clustering quality, only genes with detectable hybridization signals in the arrays from all three genotypes were used for clustering. The results are shown in supplemental Fig. S3 and supplemental Table SIII.
Principal Component Analysis-Principal component analysis (PCA) was performed using Spotfire DecisionSite 7.2 software (Spotfire, Inc.). Data were standardized using z-score calculation before PCA. The PCA scores corresponding to the first two components were plotted in a scatter plot in which the position along the axes shows the PCA score of the sample (see Fig. 3A).
Functional Classification-Differentially expressed genes were classified into functional categories based on the gene ontology definition using publicly available web-based tools Onto-express and David (data FABP-hCFTR ϩ/Ϫ /mCftr Ϫ/Ϫ /SP-C-⌬508CFTR ϩ/ϩ CFTR ⌬508Flag Mixture of FVB/N and C57BL/6 CFTR ⌬508 /mCftr ϩ/ϩ FABP-hCFTR ϩ/Ϫ /mCftr ϩ/ϩ /SP-C-⌬508CFTR ϩ/ϩ CFTR ⌬508Flag and mCFTR Mixture of FVB/N CFTR tg SP-C-hCFTR ϩ/ϩ /mCftr ϩ/ϩ hCFTR FVB/N base of annotation, visualization, and integrated discovery) (17,18). Overly represented functional categories were determined by Fisher Exact Test using corresponding murine genome as a reference dataset. Potential protein-protein or protein-DNA interactions were identified using Pathway Assist (Ariadne Genomics), a recently developed software application for analysis of biological pathways, regulatory networks, and protein interaction maps provided in a comprehensive data base termed ResNet. The current version of the data base contains over 500,000 biological interactions built by applying the MedScan text-mining algorithms to all PubMed abstracts. Promoter Analysis-Original genomic sequences and gene mapping information were downloaded from Ensemble (February, 2004 release). 2 kb upstream of exon 1 were identified and retrieved from human and mouse genome assemblies and used to form our promoter data base. Upstream sequences (2 kb) of differentially expressed genes were searched for over-represented cis-elements using MatInspector (Genomatix) vertebrate matrix library (supplemental Table SIV). Human and mouse promoters (Ϫ3 to 1 kb relative to exon 1) were compared using CisMols Analyzer, an internally developed program that identifies compositionally similar and phylogenetically conserved cis-element modules from a list of co-expressed genes (Results are shown in Figs. 6 and 7) (19).

Comparison of Lung Gene Expression Profiles: CFTR ⌬508 , Cftr
Ϫ/Ϫ , and Cftr ϩ/ϩ -CFTR ⌬508 cDNA was expressed under control of the SP-C promoter in bronchiolar and alveolar epithelial cells in Cftr Ϫ/Ϫ mice to determine whether the mutant protein might complement the abnormalities in gene expression caused by the lack of CFTR. Both human CFTR and CFTR ⌬508 cDNAs were available and well characterized. Because both are expressed in the early fetal lung, they are not likely to be seen as nonself proteins and therefore, are not likely to elicit autoimmune responses. There is no a prior expectation that the proteins (mouse versus human) will have distinct functions. Re-cloning mCFTR and mutations might introduce cryptic mutations that might complicate these studies further, therefore, we choose to use hCFTR and hCFTR ⌬508 for this study. CFTR ⌬508 mRNA was detected in RNA isolated from the lungs of CFTR ⌬508 mice at levels similar to that of the endogenous murine CFTR, while the expression of hCFTR in the CFTR tg mice was increased. mCftr expression was not changed in the CFTR tg and CFTR ⌬508 /Cftr ϩ/ϩ mice as compared with the Cftr ϩ/ϩ mice ( Fig. 1). When transfected into MLE15 cells in vitro, immunostaining for CFTR ⌬508 was detected in a diffuse, non-membrane associated pattern. In contrast, staining of normal hCFTR was primarily membrane-associated (Fig. 2). CFTR ⌬508 staining was detected in bronchial and alveolar epithelial cells of the transgenic SP-C-CFTR ⌬508 mice, consistent with the abnormal trafficking of the CFTR ⌬508 protein (Fig. 2). Lung histology and survival were not perturbed by the expression of CFTR ⌬508 in the lungs of Cftr Ϫ/Ϫ or Cftr ϩ/ϩ mice (supplemental Fig.  S1). Both CFTR ⌬508 and Cftr Ϫ/Ϫ mice were maintained in FABP-CFTR tg/tg backgrounds in which gastrointestinal function was corrected by expression of the normal CFTR in the gastrointestinal tract (11). We sought to address two separate questions: 1) does CFTR ⌬508 correct changes in mRNA expression caused by CFTR deficiency? and 2) does expression of CFTR ⌬508 have unique effects on lung RNA expression in CFTR-deficient mice?
PCA was used as an exploratory, multivariant statistical tool to reduce data dimensions, filter noise, and extract the most significant patterns in the dataset. PCA was applied to triplicate biological samples, CFTR ⌬508 , Cftr Ϫ/Ϫ , and wild-type mice (Cftr ϩ/ϩ ) at three different ages to determine the similarity/dissimilarity of each genotype on global mRNA expression profiles in the lung. Global gene expression patterns were most similar in Cftr Ϫ/Ϫ and CFTR ⌬508 mice and distinct from Cftr ϩ/ϩ mice (Fig. 3A). The expression profiles of genes differentially influenced by the lack of CFTR correlated closely with their corresponding changes in response to CFTR ⌬508 expression (positive correlation of r ϭ 0.85). Thus, the expression of CFTR ⌬508 did not correct alterations in gene expression caused by the absence of CFTR (Fig. 3B).  . Immunofluorescent staining of CFTR and CFTR ⌬508 . Immunofluorescent staining for CFTR on lung sections of Cftr Ϫ/Ϫ mice (A) and CFTR ⌬508 mice (B) indicated that CFTR ⌬508 protein was present in bronchial epithelial cells. Immunostaining on MLE15 cells transfected with wild type Cftr (C) and SP-C-CFTR ⌬508 (D) showed the lack of CFTR ⌬508 staining on the apical surface of epithelial cells and diffuse cytoplasmic staining, indicating the lack of trafficking to the cell membrane.
Transcriptional Response to CFTR ⌬508 -To identify lung RNAs influenced by expression of CFTR ⌬508 in the respiratory epithelium and distinct from those related to lack of CFTR, lung RNAs from CFTR ⌬508 mice were compared with those from Cftr Ϫ/Ϫ littermates. The mRNAs encoded by 65 distinct genes were significantly altered by expression of CFTR ⌬508 . Of these RNAs, 33 were increased while 32 were decreased. 36 genes were uniquely altered by CFTR ⌬508 (those that were not different in Cftr Ϫ/Ϫ and Cftr ϩ/ϩ mice). This set of RNAs was considered to be a reflection of the response to the expression of CFTR ⌬508 in the lung (supplemental Table SI). Among the remaining 29 genes, 14 changed in the same direction in Cftr Ϫ/Ϫ and Cftr ϩ/ϩ mice, indicating that CFTR ⌬508 has more profound effect than the lack of CFTR (Cftr Ϫ/Ϫ ) alone; 10 could not be determined due to a low hybridization signal or inconsistent changes among replicates; only 5 genes changed in opposite directions (up in Cftr Ϫ/Ϫ but down in the CFTR ⌬508 or vice versa), and none of the 5 mRNAs changed by Ͼ1.5-fold (the largest change is seen in Cyp4v3, a cytochrome P450 family member, induced 2.5-fold by CFTR ⌬508 , but down-regulated Ϫ1.4-fold in response to Cftr Ϫ/Ϫ ), indicating that there is no significant compensation by CFTR ⌬508 .
Functional Categories of Genes Influenced by CFTR ⌬508 -The genes that were influenced by CFTR ⌬508 were further assessed according to the Gene Ontology classification of biological process, and the Fisher Exact Test was used to calculate the probability that each category was over-represented in differentially expressed genes by comparing the data with the U74Av2 dataset as reference. The mRNAs defined within a group of genes responsive to "Heat/external Stimulus" (including Hspa8, Hspa2, Hspa1a, Dnaja1, Dnajb1, Ddit3, and Arnt) were the most overly represented (p ϭ 0.001). RNAs included in "Cell Growth and Development" (Nup50, Igfbp5, Shc1, Ndr2, Xpo1, Ktn1, and Krt2-8) were decreased in response to CFTR ⌬508 (p ϭ 0.005). LightCycler analysis of lung mRNAs confirmed the changes in Hspa8 and Dnaja1, as well as the observed differences in Foxfla (supplemental Fig. S2).
Genes and proteins that may interact with CFTR or with each other were identified by Pathway Assist (Ariadne Genomics), a recently developed software program application for analysis of biological pathways, gene regulation networks, and protein interactions (ResNet) (20,21). Interactions among the candidate genes and proteins influenced by CFTR ⌬508 are shown in Fig. 4A. Genes encoding heat shock proteins/ chaperones, ion transporters, and carriers and genes involved in regulation of intracellular redox state and phospholipid metabolism were significantly altered.
Enhanced Expression of Genes Encoding Heat Shock Proteins/ Chaperones-The "Heat shock proteins/chaperones" was the most significantly over-represented functional category induced by CFTR ⌬508 that was linked directly or indirectly by known physical associations with CFTR. These findings are consistent with the concept that CFTR ⌬508 is misfolded, misrouted, and/or degraded (5,22). Molecular chaperones Hsc70, Hspa8, and Hsp40 (Dnaja1 and Dnajb1) are known to facilitate CFTR biogenesis. Hsc70 directly interacts with the NBD1 domain of CFTR to increase folding efficiency, retain misfolded CFTR in endoplasmic reticulum, and enhance its ubiquitination and degradation (23,24). Hsp40 proteins (Dnaja1 and Dnajb1) are important co-chaperones with Hsc70 that facilitate the folding of CFTR. The expression of cytosolic chaperonin, Tcp1-ring complex (also called Tric or CCT), and peptidogylcan in the recognition protein (Pglyrp or Tag7) were decreased in response to CFTR ⌬508 . Both Hsc70 and Tcp1 FIGURE 3. A, principal components analysis of Cftr Ϫ/Ϫ and Cftr ϩ/ϩ (wt) at three different ages. The PCA scores corresponding to the first two components were plotted in scatter plot in which the position along the axes shows the PCA score of each sample. All probe sets on the chip were used to include all observable variation. PCA captured 75% of the variation observed in the experiment in the first three principal components. Sample distribution indicating that Cftr Ϫ/Ϫ and CFTR ⌬508 clustered together and were distinct from Cftr ϩ/ϩ (wt). Correlation of transcript changes in Cftr Ϫ/Ϫ to the corresponding changes in CFTR ⌬508 (B) and in CFTR tg (C). Genes used in this analysis were selected from a previous study (10). Correlation results were plotted in scatterplots. The y axis is the ratio of Cftr Ϫ/Ϫ to wt. The x axis represents the ratio of CFTR ⌬508 to wt (B) and the ratio of CFTR tg to wt (C). Linear regression was calculated to fit a regression line through the data points. The regression equations were included in the corresponding scatter plots. The correlation table was made using Pearson Product-Moment Correlation.  A and B). Regulatory relationships are denoted by arrows and signs within the box (ϩ, positive regulation; Ϫ, negative regulation; and no sign, regulation is unknown). mediate de novo protein folding and distinct and overlapping substrate specificity (25,26). Tag7 forms a stable 1:1 complex with Hsp70. Neither Hsp70 nor Tag7 are cytotoxic, but the complex induces apoptosis (27). Thus, reduced expression of Tag7 may represent a protection mechanism against oxidative stress-induced apoptosis. All of these chaperones (Hsc70, Dnaja1, Dnaja2, and Tcp-1) have been coprecipitated with nascent CFTR previously (28). Complex formation between CFTR ⌬508 and Dnaja1/Hsc70 was greater than with wild-type CFTR, supporting the concept that Dnaja1/Hsc70 catalyzes CFTR folding, which may be defective in CFTR ⌬508 (28). The observed increase in Hsc70 and Hsp40 mRNAs likely indicates that expression of CFTR ⌬508 caused compensatory responses in respiratory epithelial cells that might promote the folding and/or degradation of the misfolded protein. It is possible that the decrease in Tcp1 mRNA caused by CFTR ⌬508 may reflect a response to increased Hsc70 or other chaperones, in turn maintaining a balance among chaperone activities.
Altered Expression of Genes Sensitive to Cellular Redox State-Glutathione (GSH) is an important intracellular and extracellular protective antioxidant. GSH deficiency is associated with lung inflammation, increased mucus viscoelasticity, and oxidative stress-induced cell death (29 -32). Recent findings suggest that CFTR may serve as a modulator of GSH efflux. Decreased GSH efflux and extracellular GSH deficiency were observed in cells with various CFTR mutations (33)(34)(35). Consistent with a link between CFTR and GSH redox systems, decreased expression of RNAs encoding sensors of cellular redox status and those regulating GSH metabolism, including vanin-1 (Vnn1), peroxiredoxin 2 (Prdx2), and glutathione S-transferase (GST) mu 2 (Gstm2) (Fig. 5) were observed. Gstm2 encodes a GST that belongs to the mu class that conjugates and detoxifies electrophilic compounds, including products of oxidative stress. Vnn1, an enzyme with pantetheinase activity, functions as a sensor of oxidative stress, regulating endogenous GSH concentrations. The Vnn1 gene is regulated under stress and generates cystamine, which in turn inhibits ␥-glutamylcysteine synthetase, the rate-limiting enzyme for GSH synthesis. A recent study by Berruyer et al. (36) demonstrated that Vnn Ϫ/Ϫ mice are resistant to oxidative injury and that protection is reversed by treating mutant mice with cystamine. Prdx2, an antioxidant enzyme, catalyzes the reduction of hydrogen peroxide and alkyl hydroperoxides to the corresponding alcohol (or water) with the simultaneous oxidation of its active site cysteine. In yeast, Prdx2 functions both as a peroxidase and molecular chaperone, functions that can be reversibly switched by oxidative stress (29). Thus decreased expression of this group of genes in the CFTR ⌬508 mice may indicate a cellular response serving to maintain the GSH homeostasis. A group of mRNAs involved in phospholipid metabolism and intracellular redox state regulation, including phospholipase A2, group V (Pla2g5), arachidonate 12-lipoxygenase (Alox12b), arachidonate 15-lipoxygenase (Alox15b), and platelet-activating factor acetylhydrolase (Pafah2) were significantly decreased. Pafah2 functions as an anti-oxidant phospholipase, translocating between cytoxol and membranes in response to oxidative stress and protecting cells from oxidative stress-induced apoptosis by hydrolyzing oxidized phospholipids in membrane (37). Oxidant stress or GSH depletion may activate arachidonic acid mobilization and signaling by activation of Pla2 and releasing arachidonic acid from membrane phospholipids, which can be utilized by 12/15-lipooxygenase to generate a variety of bioactive eicosanoids that may modulate inflammation and contribute to oxidative stress-induced cell death (38 -40).
Lack of Classic Endoplasmic Reticulum Response to CFTR ⌬508 -Expression of misfolded proteins may induce ER stress that can cause apoptosis. A group of genes involved in regulation of apoptosis were induced by the expression of CFTR ⌬508 (Fig. 5), including DNA-dam-age-inducible transcript 3 (Ddit3), Kruppel-like factor 10 (also known as transforming growth factor ␤-inducible early growth response (Klf10/ Tieg1)), aryl hydrocarbon receptor nuclear translocator (Arnt), and myeloid leukemia factor 1 (Mlf1) (positive regulation) (41)(42)(43)(44)(45)(46)(47) and Hspa1a, Hspa8, Dnaja1, and Dnajb1 (negative regulation) (48 -50). Among these, Ddit3 (also known as GADD153 or Chop) is a member of the CCAAT/enhancer-binding protein (C/EBP) family of transcriptional factors (51). The expression of Ddit3 is low during normal cell growth but is markedly induced in response to a variety of cellular stresses, including oxidative and ER overload stress, the induction of Ddit3 will, in turn, lead to apoptosis and programmed cell death (52,53). Tieg1 is a member of the Kruppel-like factor family of zinc finger transcription factors that play critical roles in regulation of cell proliferation, differentiation, and apoptosis by a mechanism that involves formation of reactive oxidant species and GSH depletion (45,54). Carboxylesterases and GSTs are drug-metabolizing enzymes that mediate hydrolysis and conjugation of xenobiotics. Ces1 plays an important role in lipid metabolism and lung detoxification. The increase in Ces1 RNA, including CFTR ⌬508 , may represent compensation for decreased GST. Although a number of genes encoding ER chaperone proteins and apoptotic pathways were up-regulated by CFTR ⌬508 , genes directly involved in the ER-associated degradation pathway (ERAD), including Xbp1, Bip, Erdj4, Erdj5, and EDEM, were not altered.
Oxidative Stress and Anti-oxidant Effect-The present findings suggest that CFTR ⌬508 may cause a primary dysfunction in the redox system by altering GSH metabolism that may deplete GSH in lung epithelium, creating an oxidative cellular environment that may induce inflammation and lipid preoxidation. The lack of functional or morphological abnormalities in CFTR ⌬508 expressing mice may indicate the presence of strong adaptive mechanisms that maintain a critical balance between the oxidative stress and antioxidants; pro-inflammatory and anti-inflammatory mediators; apoptosis and anti-apoptosis that maintain lung function. As shown in Fig. 5, CFTR ⌬508 enhanced expression of antioxidant proteins (primarily various heat shock proteins, including Hsa1a, Hsa1b, Hspa2, Dnaja1, Dnajb1, and Hspa8) and decreased expression of RNAs encoding enzymes that should act to decrease the utilization of reduced GSH (Prdx2 and Gstm2), increase GSH biosynthesis (Vnn1), and diminish the damage caused by phospholipid oxidation (Pafah2, Alox12b, Alox15b, and Pla2g5).
CFTR ⌬508 Influences Expression of Ion Transporters and Carriers-CFTR encodes a cAMP-dependent chloride conductance that is proposed to regulate airway surface liquid and the activities of other ion transporters and channels. A number of RNAs, including Slc4a1ap, Scn7a, Slc16a7, and Nsccn1 were induced by the expression of CFTR ⌬508 in the respiratory epithelium (Fig. 4A). Slc4a1 (AE1), expressed on both apical and basolateral membranes, functions as a chloride/bicarbonate exchanger (55,56) that may provide an alternative pathway for chloride secretion. Scn7a encodes a sodium sensing channel that is expressed in alveolar type II cells known to facilitate sodium and water absorption (57,58). Nsccn1 encodes a nonselective cation channel expressed abundantly in brain, heart, and lung. Nsccn1 mRNA was increased 2.7-fold in the lungs of CFTR ⌬508 mice. Nsccn1 is known to be induced by cAMP, calcium, insulin, and enhances intracellular sodium and calcium (59 -61). The observed increase in expression of Scn7a and Nscc1 may represent compensatory responses to the CFTR ⌬508 and accompanying ion transport abnormalities associated with CFTR dysfunction. It is of interest that CFTR is known to directly interact with and regulate Scnn1b, the amelioride-sensitive sodium channel, which has been implicated in the enhancement of sodium and fluid uptake across the airway epithelium in cystic fibrosis. Expression of Scnn1b FIGURE 5. Schematic representation of a model in which oxidative stress and anti-oxidant cell responses are induced by the expression of CFTR ⌬508 . CFTR ⌬508 may cause a primary dysfunction in the redox system by directly inhibiting GSH efflux. This may cause GSH depletion in lung epithelium and create an oxidative cellular environment that induces inflammation, lipid peroxidation, and apoptosis (33)(34)(35). Cell responses to the stress with strong adaptive defense mechanism, including enhanced expression of antioxidant proteins (primarily various forms of heat shock proteins, including Hsa1a, Hsa1b, Hspa2, Dnaja1, Dnajb1, and Hspa8) and decreased expression of RNAs encoding multiples enzymes to decrease the utilization reduced GSH (Prdx2 and Gstm2) (29), increase GSH biosynthesis (Vnn1) (36), and altered phospholipid oxidization (Pafah2, Alox12b, Alox15b, and Pla2g5) (37)(38)(39)(40). mRNA in the respiratory epithelium increased airway Na ϩ absorption and initiated cystic fibrosis-like lung disease in vivo (62).
Identification of Genes Responsive to the Expression of Normal Human CFTR-To discern the effects of increased expression of normal human CFTR, lung mRNA levels were assessed in mice expressing the normal human CFTR cDNA under control of the Sftpc gene promoter in respiratory epithelial cells of adult mice. Human CFTR cDNA and protein were increased in lung tissue and epithelial cells of bronchioles as previously reported (12). Homozygous SP-C-hCFTR tg/tg , SP-C-hCFTR tg/Ϫ , and wild-type mice were compared. Unlike CFTR ⌬508 , the expression profile of mice with increased CFTR in respiratory epithelial cells (in wild-type background mice) was most similar to the gene expression profile in lungs of Cftr ϩ/ϩ (wild type) mice but not similar to that in Cftr Ϫ/Ϫ mice as assessed by K-Means cluster analysis and correlation study (supplemental Fig. S3 and Fig. 3C). Significant differences were detected in the expression of 57 genes (p Ͻ 0.05 in analysis of variance test with -fold change ϾϮ1.5-fold). Expression profiles correlated well with gene dosage when comparing wild type, heterozygous, and homozygous SP-C-CFTR transgenic mice (data not shown). Expression of human wild-type CFTR increased in 22 and decreased in 35 lung mRNAs (supplemental Table SII). In general, changes induced by expressing CFTR in the lung were subtle, and fewer RNAs were changed by Ͼ2-fold. mRNAs that were induced Ͼ2-fold included thrombospondin (Thsd1), insulin-like growth factor binding protein 5 (Igfbp5), and an expressed sequence tag that is similar to promyelocytic leukemia zinc finger protein. Thsd1 encodes a transmembrane molecule with thrombospondin module whose biological function is unknown. It is an integral membrane protein with highest expression in lung and next in salivary gland based on electronic Northern blot analysis (symatlas.gnf.org/SymAtlas/). Igfbp5 is a family member of insulinlike growth factor-binding proteins. Although there is no known direct evidence to support a relationship between these candidates and changes in CFTR expression, Igfbp5 forms a complex with IGF-1 and IGFBP-3, both of which are known to regulate CFTR expression by activating CREB1 (63)(64)(65). Levels of IGF-1 and IGFBP-3 are decreased in patients with CF and may influence growth (66,67). IGFBP5 and promyelocytic leukemia zinc finger protein interact with FHL2, a Lim domain protein that is also a co-activator for CREB-mediated transcription (68 -70). Thsd1 binds to IGFBP-5 with very high affinity (71). We hypothesize that these induced genes might form a functional complex in response to the expression of excess CFTR and regulate CFTR expression through a CREB1/cAMP-PKA pathway (Fig. 4B) (72).
RNAs decreased Ͼ2-fold by expression of hCFTR include syntaxinbinding protein 3 (Stxbp3), protein phosphatase 1A (Ppm1a), gelsolin (Gsn), calbindin 1 (Calb1), and a RIKEN cDNA C030006K11 gene, with carboxypeptidase activity. Multiple lines of evidence demonstrate that CFTR trafficking and channel activity are influenced by an interaction between syntaxin and Munc18 isoforms (73,74). Munc18c (i.e. Stxbp3), a family member of Munc18 associates with synaptosome-associated protein of 23 kDa (SNAP-23) and SNAP-23 physically associates with CFTR by binding to its N-terminal tail, a region that modulates channel gating (74,75). Gelsolin is an important intracellular and extracellular actin-severing protein. Gelsolin co-immunoprecipitates with c-Src (76). CFTR modulates c-Src expression and activity, because c-Src modulates the expression of multiple mucin genes (77), c-Src may serve as a bridge between CFTR dysfunction and mucin overproduction in CF (77). Protein phosphatase 1A, magnesium-dependent, ␣ isoform (Ppm1a) is a member of the PP2C family of Ser/Thr protein phosphatases. Ppm1a (formerly Pp2c␣) dephosphorylates and inactivates wildtype CFTR (78,79). Recent studies using various approaches, including patch clamp studies of channel rundown, co-immunoprecipitation, chemical cross-linking studies, and pull-down assays all indicate that CFTR and PP2C are closely associated and form a stable regulatory complex (80). Remarkably, the mRNAs that changed most significantly in response to increased expression of hCFTR form direct or indirect physical interactions with CFTR (Fig. 4B), supporting the concept that CFTR exists within a multiprotein complex whose activities are modulated by their interactions.
Gene ontology analysis revealed that the most significantly overrepresented molecular functions responding to increased wild-type CFTR were "chaperones" ( p ϭ 0.023). Those overly represented in the group decreased by CFTR were grouped with "electron transport": ( p ϭ 0.018). The most represented biological process influenced by increased CFTR was "cell growth and maintenance" (33%). The most represented cellular compartment was "membrane" (43%). Among the genes whose expression was enhanced by CFTR were several molecular chaperones (Hspa1a, Hsp105, and Fkbp5). Fkbp5 (also known as FK506) directly interacts with Hsp90 and Hsp70, contributing to protein folding and trafficking. Hspa1a and Hsp105 belong to the Hsp70 protein family and Hsp105/110 protein family, respectively. Their transcription is known to be stimulated in response to heat shock, oxidation, and other stresses (gene summary from NCBI). Hsp105 forms complexes with Hsp70, and recent studies suggest that it serves as a negative regulator of the Hsp70/ Hsc70 chaperone system (81,82). The enhanced expression of these molecular chaperones perhaps indicates compensatory responses to the stress caused by excessive production of CFTR protein. Expression of several electron transport proteins, including Alox12, Sdhd, Txn2, and Txn12, were significantly decreased by expression of hCFTR. Arachidonate 12-oxidoreductase (Alox12), a key enzyme in arachidonic acid metabolism, was decreased by expression of both wild-type hCFTR and CFTR ⌬508 . Alox12 catalyzes the transformation of arachidonic acid into active lipids (12-hydroxyeicosatetraenoic acid). Alox12 products influence ion channel conductivity, including apical membrane chloride channels, in airway epithelia (83)(84)(85). Freedman et al. (86 -89) demonstrated an altered ratio of arachidonic to docosahexaenoic acid in CFTR-deficient tissues. A recent study by Karp et al. (40) indicates that patients with CF produce less lipoxin, enhancing their susceptibility to neutrophil-mediated inflammation. Our data provide further support for the concept that CFTR may influence fatty acid metabolism by influencing Alox12 expression.
Gene Clustering and Promoter Analysis-Groups of genes may be co-regulated by the actions of transcription factors that are active or cis-acting elements shared among the markers of the clusters. To identify the genes co-regulated with CFTR, we first identified differentially expressed genes in lungs from CFTR ⌬508 , CFTR tg , and Cftr Ϫ/Ϫ mice. Gene probes from Affymetrix genome U74Av2 were translated to the corresponding probe set in MOE430 based on the "Best Match" provided by Affymetrix for cross-chip comparisons (www.affymetrix.com). Genes that shared no matching probes or do not have detectable hybridization signals on any of the analyses were excluded from further analysis. 135 mRNAs from the 3 data sets qualified for this combined cluster analysis. K-Means clustering was used to regroup these genes sharing similar expression patterns in the CFTR mouse models (supplemental Fig. S3 and supplemental Table SIII). The largest cluster (cluster 1) included mRNAs that were decreased in the absence of CFTR (with or without CFTR ⌬508 ) but did not change after expression of hCFTR (i.e. Cftr Ϫ/Ϫ ϭ CFTR ⌬508 Ͻ hCFTR tg ϭ Cftr ϩ/ϩ ). The second largest cluster (cluster 8) had the opposite profile (Cftr Ϫ/Ϫ ϭ Cftr Ϫ/Ϫ and CFTR ⌬508 Ͼ hCFTR tg ϭ Cftr ϩ/ϩ ). Cluster analysis again supports the concept that the CFTR ⌬508 does not correct CFTR-dependent changes in gene

cis-element modules shared by Cftr and its co-expressed genes
CisMols Analyzer (cismols.cchmc.org) was used to search for conserved modules within phylogenetically conserved regions (Ն70% sequence similarity by BlastZ alignment) (19). MatInspector was used to identify the potential cis-elements in the conserved regions for each individual gene. Phylogenetically conserved modules were compared across all genes in cluster 1 to identify common cis-element modules. Each colored symbol represents a module that contains at least two individual cis-elements. The lower panel summarizes the detail constituent of each cluster and the total frequency of each individual cis-element; the upper panel depicts the frequency of a given module occurring in cluster 1 genes and the total number of modules located in the promoter region of a given ortholog gene pair. expression. Because large clusters reflect predominant expression patterns that are likely to be functionally relevant, we chose cluster 1 for further analysis.
Promoter sequences (2 kb) of genes listed in cluster 1 were extracted from our data base (see "Experimental Procedures"). The promoter sequences were then searched for common cis-elements using MatInspector (Genomatix) seeking a core similarity of 0.95 and a matrix similarity of 0.85 in the complete vertebrate matrix library. The cis-elements significantly enriched in this cluster as compared with the sequence of randomly chosen gene promoters were selected based on binomial probability calculation and their percentage frequency in the given cluster ( p Ͻ 0.05, 80%). Several cis-elements located within 2-kb promoter sequences that were shared in cluster 1 were EGRF, EBOX, EKLF, CMYB, CEBP, and HNF1 (supplemental Table SIV). Elements that were highly represented in this cluster, but also frequently appear in random promoter sequence, include CREB and forkhead domain factors (occurrences in cluster 1 are 100 and 98%, respectively) were excluded by the probability test ( p Ͼ 0.05). CREB and Foxa1/a2 are known to regulate CFTR expression in vitro and in vivo (72,90), sug-gesting that recruitment of other transcription factors may be needed to form biologically relevant promoter modules.
Identification of a Shared Transcriptional Module-Given the nature of transcription factor binding sites (their short length and degenerate nature), single site prediction is error-prone, having information about the combination of transcription factors (modules) and phylogenetic conservation can significantly reduce false-positive rate and lead to more accurate and valuable predictions (91,92). We therefore sought to identify transcriptional modules containing two or more cis-elements. Common modules shared by co-expressed genes that were responsive to the changes in CFTR were identified via comparing the human and mouse promoter synteny region (Ϫ3 to 1 kb) using CisMols Analyzer, an internally developed program, which is designed to analyze cis-element modules that are shared across a list of co-expressed or functional related genes (cismols.cchmc.org) (19). Briefly, phylogenetically conserved regions of all ortholog gene pairs in cluster 1 were identified by BlastZ alignment (sequence similarity Ն70%). MatInspector was used to identify the potential cis-elements in the conserved regions for each individual gene, and then those phylogenetically conserved modules FIGURE 6. CFTR shares a conserved region with ATP2A2, PRKCN, and ITGA3 that contains CREB/OCT1/HOXF modules located in their promoter regions (indicated by light blue arrows). Other genes contain two or three elements of this module, including PRKCE, SOX7, KIF3A, IGFBP7, HDAC2, GJA4R, ASA1, and FIGF. HNF1/NKXH forms another transcriptional module (indicated by dark blue arrows) located in the promoter regions of several cluster 1 genes, including CFTR, ATP2A2, ITGA3, NR2F1, PRKVN, UPS33, OSBPN11, and PTGFR (see Table 2  were compared across all genes in cluster 1 to identify common ciselement modules. The common modules were determined by an algorithm that uses a 200-bp window to look through the cis-elements present within the conserved sequence region. The system has been successfully tested in several microarray profile-based data sets searching for tissue-specific gene expression (93,94). In this study, 34 ortholog-matched promoter pairs in cluster 1 were found (UCSC Golden Path). Overly represented cis-elements in cluster 1 (supplemental Table SIV) plus known transcription regulators of CFTR (CREB, forkhead domain factors, and GC-box factors SP1/GC) were searched to identify common modules (two or more individual elements) within the 200-bp window that were shared across species (human and mouse) and across multiple promoter regions of cluster 1 genes. To select for the most common transcriptional modules that may regulate CFTR and co-regulated genes in cluster 1, the following filters were used: 1) the modules must be shared in syntenic regions in the CFTR gene, 2) for a two-element modules to be chosen, a minimum of 10 genes must contain this module (for a 3-element module to be chosen, at least 6 genes need to share this module), and 3) genes must share a minimum of 5 common modules with CFTR in their syntenic regions. Genes and modules passing these criteria are listed in Table 2. Genes sharing Ͼ50% of the common modules with CFTR include PRKCE (protein kinase C, epsilon), ITGA3 (integrin alpha 3), NR2F1 (nuclear receptor subfamily 2, group 4, member 1), RASA1 (RAS p21 protein activator 1), EDN1 (endothelin 1), and ATP2A2 (sarco/endoplasmic reticulum Ca2 ϩ -ATPase 2) (highlighted in green in Table 2). The module with highest frequency of occurrence in cluster 1 contained three potential cis-acting elements MAZF/SP1F/ZBPF (80%). This module generally was located within 500 bp of the predicted transcriptional start site of genes in cluster 1 ( Table 2 and Fig. 7). Other promoter modules occur with high frequency (Ͼ40%), including ETSF/HOXF, EGRF/ETSF, ETSF/GATA, EKLF/ZBPF, MAZF/MZF1/ZBPF, CREB/ETSF, CREB/HOXF, ETSF/ MAZF, EGRF/MAZF, EGRF/ZBPF, and HNF1/NKXH. CFTR expression is subject to temporal, spatial, hormonal, and cell type-specific controls (95)(96)(97). The present analysis identified several known and potential regulatory elements that may serve as common regulators for CFTR and co-regulated genes. Fig. 6 shows that CFTR shares a conserved region with ATP2A2, PRKCN, and ITGA3 that contains CREB/ OCT1/HOXF modules located Ϫ2 to Ϫ1.5 kb relative to their start sites. Other genes contain two or three elements of this module, including PRKCE, SOX7, KIF3A, IGFBP7, HDAC2, GJA4R, ASA1, and FIGF. CREB is known to regulate CFTR expression in vitro and in vivo (72).
However, CREB was not specific for cluster 1 genes when compared with its frequency in random gene promoters. CREB interacts with other transcriptional elements (such as OCT1 and/or HOXF) that may be more specific to the transcriptional regulation of CFTR and its related genes. HNF1 (hepatocyte nuclear factor-1) sites were significantly enriched in cluster 1. An HNF1/NKXH transcriptional module was located in the promoter regions of several genes, including CFTR, ATP2A2, ITGA3, NR2F1, PRKVN, UPS33, OSBPN11, and PTGFR (Table 2 and Fig. 6). Consistent with our prediction, recent studies demonstrate that HNF1␣ binds to the CFTR gene and regulates CFTR gene expression (98).
Predicted binding sites for C/EBP were also enriched in cluster 1. CEBP sites were located in close proximity to NKXH, GATA, and HOXF within multiple cluster 1 genes. Deletion and mutational analysis of the CFTR promoter has identified C/EBP binding sites (CCAAT-like element) that are required for basal and cAMP-mediated regulation of CFTR expression (99 -102). In additional to identification of regulatory modules containing elements that were known to regulate CFTR, we identified new elements that might be important for regulation of CFTR expression. EGR1 is a transcriptional regulator of the ATP2A2 gene (103,104). In the present study, we found that ATP2A2 shares Ͼ50% of these common regulatory modules with CFTR, and the EGRF/MAZF/ ZBPF module that was identified in multiple cluster 1 genes within 500 bp upstream of exon 1, including ATP2A2 and CFTR, suggesting that EGRF module may be involved in the transcriptional regulation of CFTR (Fig. 7).

CONCLUSIONS
The effects of CFTR, CFTR ⌬508 , and deletion of CFTR were compared in lung tissue from adult transgenic mice. Mice of all genotypes do not suffer from discernible lung pathology, indicating remarkable functional adaptation to changes in CFTR status. In the present transgenic mouse models, genomic responses to CFTR deficiency were not corrected by expression of the CFTR ⌬508 mutant protein in respiratory epithelial cells. Genes regulating host defense, inflammation, fluid, and electrolyte transport were commonly altered in Cftr Ϫ/Ϫ and CFTR ⌬508 , Cftr Ϫ/Ϫ mice. Our data demonstrate that CFTR ⌬508 may cause a primary disturbance in the redox system by inhibiting GSH homeostasis. The presence of a strong antioxidant defense response in our mouse models, however, may be critical for protection from oxidative stress and maintenance of normal lung function. Expression of hCFTR and CFTR ⌬508 in respiratory epithelial cells of the lung induced expression of genes classified as heat shock proteins and chaperones but did not induce classic ERAD transcriptional responses. Expression of a number of genes known to interact with CFTR via protein-protein interactions was influenced by CFTR ⌬508 , hCFTR, and Cftr Ϫ/Ϫ , supporting the hypothesis that CFTR exists within a multiprotein complex whose constituents are co-regulated in response to presence or absence of CFTR or CFTR ⌬508 . Cis-elements and modules consisting of shared complex elements were located in promoter regions of the CFTR and co-regulated genes, suggesting common regulatory mechanisms governing a number of the proteins interacting with CFTR. These findings may have important implications toward the identification of modifier genes, novel therapeutic targets, and new pathways that may influence the pathogenesis of CF.