We showed previously that CCAAT/enhancer binding protein α (C/EBPα), a tissue-specific transcription factor, is a candidate tumor suppressor in lung cancer. In the present study, we have performed a transcriptional profiling study of C/EBPα target genes using an inducible cell line system. This study led to the identification of hepatocyte nuclear factor 3β (HNF3β), a transcription factor known to play a role in airway differentiation, as a downstream target of C/EBPα. We found down-regulation of HNF3β expression in a large proportion of lung cancer cell lines examined and identified two novel mutants of HNF3β, as well as hypermethylation of the HNF3β promoter. We also developed a tetracycline-inducible cell line model to study the cellular consequences of HNF3β expression. Conditional expression of HNF3β led to significant growth reduction, proliferation arrest, apoptosis, and loss of clonogenic ability, suggesting additionally that HNF3β is a novel tumor suppressor in lung cancer. This is the first study to show genetic abnormalities of lung-specific differentiation pathways in the development of lung cancer.
Lung cancer remains a public health problem with ∼170,000 cases in the United States per year (1) . It is the leading cause of cancer deaths in both men and women with a 5-year survival rate of only 15%. A recent analysis of trials performed over the last 30 years demonstrated clearly that only minimal progress has been made in the treatment of this disease (2) . The disappointing results of recent studies have led to the realization that we have reached a “chemotherapy efficacy plateau” (3) . Additional progress in the treatment of lung cancer will depend critically on a better understanding of the molecular events leading to the development of epithelial neoplasias as well as the critical pathways sustaining the neoplastic, invasive phenotype. Our understanding of the genetic abnormalities underlying the development of lung cancer remains quite limited (4 , 5) . Both a number of tumor suppressors, such as p53, p16, and retinoblastoma, as well as several proto-oncogenes, such as k-ras and the epidermal growth factor receptor, are known to play a role, but no abnormalities of lung-specific tumor suppressors or proto-oncogenes have been identified yet. A recent high-frequency allelotyping study demonstrated that in individual lung cancers, as many as 22 areas of loss of heterozygosity can be detected, suggesting that many tumor suppressor genes remain unidentified (6) .
Aberrant differentiation is one of the hallmarks of cancers, and the contribution of differentiation arrest to the multistep carcinogenesis process has been accepted widely recently (7) . In particular, it is increasingly clear that aberrant regulation of transcriptional control pathways of normal differentiation is one of the most common abnormalities in hematological malignancies, such as acute myeloid leukemia (8) . The transcriptional control of differentiation pathways in airway epithelium is poorly understood, and its abnormalities in the aberrant differentiation of lung cancers are largely unknown. A number of key transcription factors, such as thyroid transcription factor-1, helix-loop-helix transcription factors, forkhead transcription factors, such as hepatocyte nuclear factor 3β (HNF3β), and CCAAT/enhancer binding protein α (C/EBPα) are implicated in the complex developmental genetic instruction of lung morphogenesis and cell lineage determination (9 , 10) .
C/EBPα is a leucine zipper transcription factor that serves as a tissue-specific differentiation factor in a number of tissues, such as hepatocytes, myeloid cells, and adipocytes (11) . C/EBPα was also identified recently as a novel tumor suppressor in acute leukemia (12) . Recurrent mutations of C/EBPα have been identified by a number of groups in subtypes of acute leukemia (13 , 14) . C/EBPα is expressed strongly in the lung, more specifically in both type II pneumocytes as well as cells of the bronchial epithelium (15 , 16) . It also regulates the expression of several genes, directly or indirectly, during lung differentiation, including surfactant B and uteroglobin (17 , 18) . Specific lung abnormalities, such as an abnormal proliferation of type II pneumocytes, have been described in C/EBPα−/− knockout mice (19) , suggesting that C/EBPα is important for normal lung development and the maintenance of normal alveolar structure. It is postulated that this hyperproliferation is because in the absence of C/EBPα, the alveolar type II cells can continue to proliferate. In previous studies, we have demonstrated that C/EBPα is down-regulated in a large proportion of lung cancers (20) . We also developed a tightly regulated, highly inducible cell line model system using a zinc-inducible metallothionein promoter-based system. With the use of this stably transfected cell line system, we have shown that the induction of C/EBPα expression leads to growth arrest, apoptosis, and cellular changes suggestive of differentiation, all supporting its role as a candidate tumor suppressor gene.
To gain additional insight into the downstream effects of C/EBPα expression, we have performed transcriptional profiling studies on this inducible cell line system. From the many C/EBPα-regulated genes identified, we have focused our additional studies on one, HNF3β (also known as Foxa2), given its known important role in airway epithelial differentiation (21, 22, 23) . Our studies demonstrate down-regulation and novel mutations of HNF3β in lung cancer cell lines. We also identified hypermethylation of the promoter region of the HNF3B gene as a novel mechanism for epigenetic silencing of HNF3β expression. To assess the functional consequences of HNF3β expression, we also generated a tetracycline-regulated inducible cell line system for the conditional expression of HNF3β. With the use of this system, we demonstrate that induced expression of HNF3β in the H358 lung cancer cell line leads to growth reduction, proliferation arrest, apoptosis, and loss of clonogenic ability, supporting its identification as a novel tumor suppressor in lung cancer.
MATERIALS AND METHODS
Cell Lines and Cell Culture.
The following lung cancer cell lines were used in our study: squamous cell cancer, Calu-1, SK-MES-1, H157, H520, SW900, U1752, and EPLC103H; adenocarcinoma, A427, SK-LU-1, Calu-3, H23, and H441; adenocarcinoma, bronchoalveolar type, H358, A549, and H322; adenosquamous cancer, H125, H292, and H596; large cell cancer, H460 and H661; anaplastic, Calu-6; and small cell lung cancer, H526, H187, H69, H345, H211, H60, H82, N417, H128, and UMC19. All of the non-small cell lung cancer cell lines were grown in RPMI 1640 supplemented with 10% fetal bovine serum, whereas all of the small cell lung cancer cell lines were grown in RPMI 1640 supplemented by HITES medium (final concentration of 2.5% fetal bovine serum, 2.8 mm glutamine, 10−8 m hydrocortisone, 10−8 m estradiol, and 1% insulin/transferrin/Na-selenite (Sigma Chemical Co., St. Louis, MO).
Oligonucleotide Array Analysis.
Ppc22-transfected H358 cells were grown to 60% confluence in RPMI containing 10% fetal bovine serum. Triplicate plates were induced by the addition of 100 μm of ZnSO4 for 6 and 12 h. Control cells (0 h of induction) were grown without the addition of ZnSO4 to the medium. Cells were collected at the same time, and total cellular RNA was isolated using the TRIzol method. RNA specimens were then processed and hybridized to Affymetrix Hu95 microarrays and scanned. The expression value for each gene was calculated using Affymetrix GeneChip software.
Preprocessing, Rescaling, and Filtering.
The raw expression data consisted of the scanner “signal” units as obtained from the GeneChip MAS5 software of Affymetrix. These raw data were rescaled to account for different chip intensities. Each chip in the data set was multiplied by the factor constant/chip_intensity, where chip_intensitydenotes the average intensity of the chip (i.e., the expression level of the sample averaged across all of the probe sets in the chip), and constant is the same quantity for all of the chips (chosen to be the average intensity of the median chip). From the initial set of 12,626 genes, a final set of 4,984 genes was obtained as follows: (a) setting the minimum signal to 10 and the maximum signal to 20,000; (b) excluding genes for which the fold change (i.e., the maximum:minimum threshold value) was <3; and (c) excluding those genes for which the δ change (i.e., the difference between the maximum and minimum threshold values) was <100. For the identification of differentially expressed genes, a paired t-score was used. When analyzing the pooled data, the “0 versus 6 h” pairwise differences and the “0 versus 12 h” pairwise differences were pooled together for the computation of the score. For the analysis, GeneCluster software and scripts written in R (an open-source statistical package) were used.
Total cellular RNA from cell lines was isolated using TRIzol reagent. RNA (20 μg/lane) was separated on 1% agarose/4-morpholinepropanesulfonic acid/formaldehyde gels and transferred to MagnaGraph membranes (Osmonics, Westborough, MA). Hybridization was performed with [32P]dCTP-labeled probes using Church-Gilbert hybridization solution. The following probes were used: rat C/EBPα-350 bp fragment flanking the rat C/EBPα/SV40 polyadenylic acid junction from plasmid ppc22; HNF3β, 1.6 kb full-length cDNA of rat HNF3β (kind gift of Dr. Robert Costa, University of Illinois, Chicago, IL); cyclooxygenase-2 (COX-2), 700-bp fragment of the 3′ untranslated region of the human COX-2 mRNA subcloned into pGEM-T, cut with NcoI-SacI; and interleukin 8, 500-bp EcoRI insert from PCDNA3 (kind gift of Dr. Isaiah Fidler, M. D. Anderson Cancer Center, Houston, TX).
Whole cell lysates were isolated using radioimmunoprecipitation assay lysis buffer and protease inhibitors (phenylmethylsulfonyl fluoride, pepstatin, and leupeptin), and 20 μg of protein/lane were electrophoresed in 10 or 12% polyacrylamide minigels. A 1:2000 dilution of a polyclonal rabbit anti-C/EBPα antibody and a 1:2000 dilution of a polyclonal goat anti-HNF3β antibody (both from Santa Cruz Biotechnology, Santa Cruz, CA), and a 1:1000 dilution of a monoclonal mouse anti-β-actin and a 1:100 dilution of a monoclonal mouse anti-β-tubulin (both from Sigma Chemicals) were used, respectively. Detection was performed using enhanced chemiluminescence (Amersham Life Science, Piscataway, NJ).
Electrophoretic Mobility Shift Assay.
Nuclear extracts were prepared from untreated or 25 μm CdSO4-treated H358 ppc18-transfected or ppc22-transfected cells after 16 h of induction as described previously (24) . The gel shift assay consisted of a binding reaction allowing the formation of DNA-protein complexes, which were then separated from unbound probe by native gel electrophoresis. 50 ng of double-stranded DNA probe containing the C/EBP binding site of the HNF3β promoter was end-labeled with [γ-32P]ATP and T4 polynucleotide kinase (New England Biolabs). The sequences of the oligonucleotides used to generate the probes were as described previously (25) : sense, 5′-AATTCCCTGTTTGTTTTAGTTACGAAATGCGTTG-3′; and antisense, 5′-AATTCAACGCATTTCGTAACTAAAACAAACAGGG-3′. Binding reactions were performed by incubating 5 μg of nuclear extracts with 50,000 cpm of the double-stranded probe in 20 μl of reaction mixtures consisting of ×1 binding buffer [2 mm HEPES-KOH (pH 7.9), 10 mm KCl, 0.5 mm MgCl2, 0.2 mm DTT, and 2% glycerol] in the presence of 50 ng/μl BSA and 25 ng/μl poly(deoxyinosinic-deoxycytidylic acid) for 20 min at room temperature. For the supershift assays, 2 μg of a rabbit polyclonal anti-C/EBPα antibody (sc-61X; Santa Cruz Biotechnology) was added to the reaction mixture. The binding reactions were separated on 4% acrylamide gel at 150 V. Subsequently, the gel was dried under vacuum at 80°C for 1 h and submitted to autoradiography with an intensifying screen for 6 h at −80°C.
Chromatin Immunoprecipitation Assays.
Mock-transfected and ppc22-transfected cells were treated with 25 μm CdSO4 for 48 h. Chromatin immunoprecipitation assay was performed as described previously (26) . Briefly, 1 × 108 cells were cross-linked by addition of formaldehyde to the medium at a final concentration of 0.37%. Nuclear lysates were prepared, and before immunoprecipitation, 20% of each lysate was removed for analysis of input chromatin DNA. Immunoprecipitation was performed with 5 μg of normal rabbit IgG (Santa Cruz Biotechnologies) or 5 μg of rabbit polyclonal C/EBPα antibody (sc-61; Santa Cruz Biotechnology) at 4°C for 6 h with rotation. Immune complexes were collected by incubating with protein A-agarose beads overnight at 4°C with rotation. Cross-links of the immunoprecipitated samples and input samples were reversed by heating at 67°C in the presence of NaCl and RNase A (Sigma) for 5 h followed by proteinase K (Roche Diagnostics) digestion. The DNA was recovered by phenol-chloroform extraction followed by ethanol precipitation and resuspended in distilled sterile water. Binding of C/EBPα to the HNF3β promoter was assessed by PCR with the following primers: sense, 5′-GCCTCCACATCCAAACACC-3′; and antisense, 5′-CTCTCCGACTCCTCAGACACC-3′, which amplify a region that spans bps −75 to −104 (containing the C/EBP binding site) of the HNF3β promoter.
The promoter region and three exons of HNF3β were sequenced by the use of seven primer sets as described previously (27) . PCR products were sequenced by the use of both the sense and antisense primers. Both the actual sequences and the traces were compared with that of wild-type HNF3β using DNAStar software. All of the abnormal sequences were reamplified twice and resequenced in both directions. In cases where the abnormal sequence was confirmed, the PCR product was subcloned into pGEM-T vector, and at least five subclones were sequenced.
The 5′-aza, 2′ deoxycytidine was a kind gift of Dr. Stephen Baylin (Johns Hopkins Oncology Center, Baltimore, MD). Cells were grown to 20% confluence, and then deoxyazacytidine was added at 0.2 μm or 1 μm concentration to duplicate specimens. Medium was changed, and fresh drug was added every day. RNA was collected using the TRIzol method after 96 h of treatment. Real-time reverse transcription-PCR was performed as described below.
Promoter Methylation Studies.
Bisulfite sequencing was performed according to established methods (28) . In brief, 2 μg of genomic DNA was bisulfite treated and purified (Promega Wizard DNA Clean-UP System; Promega). The resultant bisulfite-modified DNA was amplified using a primer set amplifying a 369-bp fragment of the HNF3β promoter and part of exon 1 (positions 192–540). The following primers were used: sense, 5′-TTGGAAGATAGAGAGGATAGA-3′; and antisense, 5′-CCCCTCCCTATTACCAATTCAA-3′. The amplified PCR product was subcloned into pGEM-T vector, and clones were sequenced with the use of Sp6 antisense primer. Peripheral blood mononuclear cell DNA served as negative control, whereas universally methylated DNA (CpGenome Universally Methylated DNA; Intergen) was used as positive control.
The full-length cDNA of rat HNF3β was a kind gift of Robert Costa (University of Illinois, Chicago, IL). The 1.6-kb rat HNF3β cDNA was released from the pGEM-T vector backbone by digestion with EcoRI. The fragment was blunt ended and subsequently subcloned into the dephosphorylated PvuII site of the multiple cloning site of vector pTRE2puro (Clontech, Palo Alto, CA) to generate plasmid pTRE2HNF3B.
Generation of Inducible Cell Lines.
H358 cells were transfected with the transactivator Tet-off plasmid using lipofectamine transfection (LipofectAMINE PLUS; Invitrogen Life Technologies, Inc., Carlsbad, CA) according to the manufacturer’s instructions. Clones were selected on the basis of G418 resistance (G418 concentration of 500 μg/ml). Highly repressible clones were identified by transient transfection with a TRE-luciferase reporter plasmid (pTRE2pur-luc; Clontech). Clones were screened by performing luciferase assays after the cells were grown for 24 h with or without addition of doxycycline (1 μg/ml). Clone 6 showed the highest level of repressibility after doxycycline withdrawal and was selected for additional studies. The second round of transfection with the pTRE2HNF3B plasmid was performed using identical methods. Cells were grown continuously in the presence of doxycycline to suppress transresponder gene induction, and clones were selected on the basis of dual resistance to G418 and puromycin (1 μg/ml). Clones were screened for inducibility of HNF3β expression after 24–48 h of doxycycline withdrawal.
Real-Time PCR Assay.
TRIzol-extracted RNA was DNase treated, reverse transcribed, and subsequently amplified using an ABIPrism 7700 Sequence Detector (Applied Biosystems) by the following parameters: 50°C (30 min); 95°C (15 min) followed by 40 cycles of 94°C (15 s); and 60°C (60 s). Primers and probe (FAM-labeled) were as follows: human HNF3β forward primer, 5′-AAGATGGAAGGGCACGAGC-3′; reverse primer, 5′-TGTACGTGTTCATGCCGTTCA-3′; and probe, 5′-TCCGACTGGAGCAGCTACTATGCAGAGC-3′. Rat HNF3β: forward primer, 5′-CTGAAGCCCGAGCACCAT-3′; reverse primer, 5′-GCTGCTCGGAGGGACATGA-3′; and probe, 5′-TCCGACTGGAGCAGCTAC-3′. Primers and probe (VIC-labeled) for 18 S rRNA were from Applied Biosystems.
Bromodeoxyuridine Proliferation Assay.
Proliferation assays were performed with the use of BrdU Flow kit (Becton Dickinson PharMingen, San Diego, CA). In brief, bromodeoxyuridine was added to medium to achieve a final concentration of 10 μm for 45 min, then cells were trypsinized and treated according to the manufacturer’s instructions. Flow cytometry was performed on a fluorescence-activated cell scan cytometer (Becton Dickinson).
Annexin/Propidium Iodide Apoptosis Assay.
Cells were collected after trypsinization, washed with PBS, and stained with annexin/propidium iodide according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany). Samples were analyzed on a fluorescence-activated cell scan cytometer (Becton Dickinson).
One thousand each of H358 pTRE2HNF3B/4 and pTRE2HNF3B/31 cells were mixed with 1.5 ml of 1.25% methylcellulose/Iscove’s modified Dulbecco’s medium/tetracycline-free fetal bovine serum/puromycin with or without 1 μg/ml of doxycycline and plated onto 20-mm cell culture plates. Doxycycline was added every 2–3 days to the appropriate plates. Six plates/condition were seeded. The number of colonies was counted on day 14.
Transcriptional Profiling of C/EBPα-Inducible H358 Cells.
As described previously (20) , we have generated a stably transfected cell line from H358 adenocarcinoma cells using a mammalian expression vector construct (ppc22; Ref. 29 ) harboring the rat C/EBPα gene under the control of the zinc-inducible metallothionein prom (MT-C/EBPα). We also generated control cell lines by stable transfection of H358 cells with a control vector (ppc18). H358 cells have practically no detectable native C/EBPα expression either on the RNA or protein level. In this cell line system, marked increase in C/EBPα mRNA can be detected as early as 3 h after induction by addition of zinc or cadmium sulfate to the medium. The tight regulation and very strong inducibility of this cell line system appeared to be ideally suited for transcriptional profiling studies to identify downstream targets of C/EBPα and gain additional functional insights into the mechanisms leading to C/EBPα-induced growth arrest and differentiation.
To perform oligonucleotide array analysis, ppc22-transfected H358 cells were induced to express C/EBPα by the addition of zinc for 6 and 12 h, respectively. Control cells (0 h induction) were grown without the addition of zinc to the medium. Conditional expression of C/EBPα was confirmed by Northern as well as Western blotting (Fig. 1, A and C) ⇓ . Transcriptional profiling was then performed on total cellular RNA using Affymetrix Hu-95 chips. The raw expression data were preprocessed, rescaled, and filtered as described in “Materials and Methods.” Analysis of the results was performed by GeneCluster software (data on the entire gene set is available as Supplementary Data). Experiments were carried out in triplicates, with three time points for each experiment (0, 6, and 12 h). We were interested in identifying those genes manifesting a differential expression between 0 h and 6 h to identify the earliest wave of transcriptional changes most likely highly enriched in direct targets of C/EBPα. We also performed a comparison between the uninduced control and induced “pooled” specimens (6 samples, 3 at 6 h and 3 at 12 h). This comparison yields a set of genes that is up- or down-regulated consistently between 6 and 12 h, therefore is less likely to harbor false-positive candidate genes and might also identify important up- or down-regulated genes that might not be significantly changed at 6 h but become evident by 12 h. We felt the two comparisons would provide complementary information and would also serve as internal controls for the validity of our findings. We used paired t-statistic to rank genes. From the ranked list of genes, we excluded genes with a t-statistic < 1 (representing no association). To arrive at a biologically relevant set of genes, we used two additional sets of criteria. We selected only those genes where the level of induction or repression was a minimum 3-fold (mean difference using means of triplicate experiments). We also excluded genes where the absolute difference between induced/uninduced was <500 fluorescence intensity signal units, thereby removing genes where the changes might appear significant based on fold-differences, but this is likely to be biologically irrelevant or spurious secondary to low levels of expression. The top genes obtained this way are shown in Tables 1 ⇓ and 2 ⇓ . Of note is that a very strong overlap was noted between the 0 versus 6 h and the 0 versus pooled comparisons (34 of the top 45 up-regulated and 23 of the top 35 down-regulated genes from the 0 versus 6 h comparison appeared in the 0 versus pooled most highly induced/repressed gene list); therefore, the pooled analysis contributed only a few genes to the lists. Such strong overlap does suggest that the gene set arrived at this way represents an enriched set of genes regulated directly or indirectly by C/EBPα in H358 cells.
To confirm the findings of the oligonucleotide array analysis, the induction/repression of several genes was confirmed by Northern blot analysis (Fig. 1A) ⇓ . As control, RNA was collected from H358 cells stably transfected with empty vector (ppc18) and treated in an identical fashion to cells collected for the oligonucleotide array analysis. Fig. 1 ⇓ shows representative results for interleukin 8, COX-2, and HNF3β genes. These studies confirmed the gene chip results in that all of the examined genes showed consistent findings with our gene chip data, and none of the examined genes appeared regulated by zinc itself, as demonstrated by the absence of induction in the control (ppc18) cell line.
An analysis of the highly up-regulated genes demonstrated three major clusters of genes in this set. First, many of the up-regulated genes are acute phase reactants, such as interleukin 8, COX-2, tumor necrosis factor-inducible gene 14, and exodus-1 among others. A second group of genes comprises genes of metabolic pathways, such as hepatic dihydrodiol dehydrogenase, folate receptor, glutaredoxin, monocarboxylate transporter 2, and quinone oxidoreductase. The third group of genes includes genes known to be involved in fat metabolism or adipocyte differentiation, such as adipophilin, ADP-ribosylation factor-like protein-4, and ceramide glucosyltransferase. Lastly, one of the most highly induced genes is HNF3β. This finding is of particular importance because HNF3β is known to play a major role in the transcriptional control of cellular differentiation, including airway epithelial differentiation (21 , 30) , thereby suggesting that C/EBPα might indeed act as master regulator of airway epithelial differentiation.
Our analysis identified also a set of highly repressed genes by C/EBPα. Many of the repressed genes are growth factors, growth factor receptors, or proangiogenic molecules, such as connective tissue growth factor, Cyr61 (also called CCN1), fibroblast growth factor-9, epidermal growth factor receptor, and urokinase-plasminogen activator. This is in line with the established role of C/EBPα in proliferation arrest (29 , 31 , 32) . C/EBPα is identified as one of the down-regulated genes. This is not surprising, because the ppc22 plasmid construct RNA does not hybridize with the oligonucleotide on the Affymetrix chip.
HNF3β Is Regulated by C/EBPα.
Our transcriptional profiling studies showed an ∼6-fold induction of HNF3β mRNA as early as 6 h after C/EBPα induction. This finding is particularly interesting given the established role of HNF3β in lung development as well as cellular differentiation (10 , 21, 22, 23) . Therefore, we decided to establish whether this regulation was direct or indirect and also focused in our additional studies on dissecting the role of HNF3β in lung cancer as a possible critical target of the changes induced by C/EBPα. We confirmed the up-regulation of HNF3β as a result of the conditional expression of C/EBPα by Northern blotting, Real-time reverse transcription-PCR assay, as well as Western blotting and showed that zinc itself does not induce the expression of HNF3β (Fig. 1, B and C) ⇓ . The degree of induction on the protein level is ∼3-fold. Although the peak induction on the RNA level occurs by 6 h, on the protein level the peak expression is somewhat delayed and occurs between 24 and 48 h.
The HNF3β promoter does have a putative C/EBP-binding site, and one report did show previously indirect evidence of C/EBP proteins regulating the HNF3β promoter (25) . In electrophoretic mobility shift assays, we demonstrated a strong increase in specific binding activity upon induction of C/EBPα (Fig. 2A) ⇓ . The specificity of binding was confirmed by supershift assays using a C/EBPα antibody. We confirmed the same finding by chromatin immunoprecipitation assays using primers flanking the putative C/EBP-binding site of the HNF3β promoter (Fig. 2B) ⇓ . These assays demonstrated direct C/EBPα binding to this promoter element of the HNF3β promoter in vivo, suggesting that the induction identified through our transcriptional profiling studies is indeed due to direct regulation by C/EBPα.
Abnormalities of HNF3β in Lung Cancer Cells.
Given our finding of C/EBPα-mediated induction of HNF3β and the established role of HNF3β in airway development and differentiation, a process also regulated by C/EBPα, we decided to focus our additional studies on dissecting the role of HNF3β in lung cancer. We have determined the expression of HNF3β by Northern blotting in 25 lung cancer cell lines representing all of the histological subtypes of lung cancer. Although HNF3β is strongly expressed in normal lung, its expression is undetectable or very weak in 15 of 25 cell lines examined (Fig. 3A) ⇓ . Western blotting demonstrated strong correlation of HNF3β expression between mRNA and protein levels (Fig. 3B) ⇓ . Of the 10 cell lines that expressed HNF3β mRNA at substantial levels, 5 had transcripts of sizes different from that observed in normal lung, most likely representing products of alternative splicing. We have performed genomic sequencing of the HNF3β gene to exclude the possibility that these cell lines carry mutant forms of HNF3β. The HNF3β gene is located on chromosome 20p11 and has 3 exons. A sequencing strategy was adopted (27) using a set of seven primers to sequence a portion of the HNF3β promoter, exons 1 and 2, and all of the coding regions of exon 3, as well as the first 68 bps of the 3′-untranslated region in 31 lung cancer cell lines. Besides a number of single-base polymorphisms, two mutant forms of HNF3β were found (Fig. 3C) ⇓ . One cell line, H60 (small cell lung cancer), harbors a heterozygous G-A mutation at position 2916 (GenBank accession no. AF176110) resulting in a G-D amino acid change at codon 92 inside the NH2-terminal activation domain (TAD II). A homozygous C deletion at codon 194 (position 3220 in AF176110) in the middle of the forkhead domain was found in the SKLU-1 cell line (adenocarcinoma) leading to a frameshift and a truncated 218-amino acid protein. Interestingly, in neither cell line is HNF3β detectable at either the mRNA or at the protein level, suggesting that the resulting mRNAs are unstable and/or that the expression of HNF3β is silenced by another mechanism (e.g., promoter methylation). None of the cell lines with aberrant-sized transcripts carried any mutations.
Promoter Methylation of the HNF3β Promoter.
To establish whether promoter methylation could play a role in silencing HNF3β expression, 4 HNF3β nonexpressor cell lines, A427, H596, SKLU-1, and H322, were treated for 96 h with two different concentrations (200 nm and 1 μm) of deoxyazacytidine, a demethylating agent. Significant up-regulation of HNF3β mRNA was observed in 3 of the 4 cell lines examined (H322, A427, and H596; Fig. 4 ⇓ ). SKLU-1 cells had no detectable HNF3β mRNA expression regardless of treatment. The promoter of HNF3β is very rich in CpG dinucleotides and meets the criteria of a CpG island. We performed bisulfite sequencing of a 369-bp segment of the promoter as well as the 5′ untranslated region of the first exon of the HNF3β gene (positions +192 to 560). This region contains 25 CpG dinucleotides that are putative targets for promoter methylation. After bisulfite treatment, products were subcloned into pGEM-T vector, and multiple clones were sequenced. All 4 of the above cell lines showed evidence of methylation with the densest methylation observed in the case of SKLU-1 (9 of 9 clones sequenced had methylated Cs; 32–96% of all of the Cs methylated/clone). Two of 6 clones of H596 (28–52% Cs methylated), 5 of 8 clones of H322 (32–40%), and 2 of 3 clones of A427 (16–76% Cs methylated) had evidence of methylation. These results strongly suggest that promoter hypermethylation could be a putative mechanism for silencing of HNF3β expression.
Conditional Expression of HNF3β Leads to Growth Arrest.
To analyze additionally the role of HNF3β in airway epithelium, we set out to establish an inducible cell line system. We selected a tetracycline-inducible system requiring the establishment of doubly stably transfected cell lines (33) to develop lines with tight control of expression and to avoid potential toxicity from the heavy metal inducers. The H358 cell line was initially transfected with the Tet-Off transactivator plasmid, and stably transfected inducible transactivator clones were selected. For the second transfection, a plasmid construct was generated consisting of the HNF3β cDNA under the control of the tTA-regulated TRE-driven promoter (pTRE2HNF3B). Double-transfectant transresponder clones were selected based on G418 and puromycin resistance. From 50 clones screened, 2 clones were selected for additional analysis. These clones (clones 4 and 31) demonstrated a 6–8-fold induction of HNF3β RNA on withdrawal of doxycycline from the medium, which translated into an ∼3–4-fold increase in HNF3β protein. It was noted that HNF3β expression could be fully suppressed even using a 1000-fold lower concentration of doxycycline (1 ng/ml as opposed to 1 μg/ml) than recommended (Clontech). When these lower concentrations were used, the induction of HNF3β occurred significantly earlier, most likely because even several washes of the cells might not reduce the doxycycline concentration sufficiently to allow transresponder gene activation when the higher suppressive concentrations are used. At the lower doxycycline concentrations used, the induction of HNF3β mRNA occurred as early as 2 h after the withdrawal of doxycycline as assessed by real-time PCR assay (Fig. 5A) ⇓ , whereas on the protein level, the induction is clearly detectable by Western blotting as early as 16 h after the withdrawal of doxycyline from the medium (Fig. 5B) ⇓ . Of note is that this degree of induction is very similar to the level of HNF3β induction obtained after C/EBPα expression in the ppc22-transfected H358 cell line.
In both of these clones, induction of HNF3β protein led to very substantial growth reduction noticeable as early as within 7 days (Fig. 6A) ⇓ . In fact, in 1 of the clones (clone 31), after day 4 of HNF3β induction, no additional increase in cell numbers was noted, and by day 14, no viable cells could be seen. In clone 4, the growth reduction was very substantial, but slow cell proliferation did continue despite induction of HNF3β. As expected, no change in cell proliferation was noted on the withdrawal of doxycycline from the medium in the parental cell line.
Changes in cell proliferation and cell cycle profile were also assessed in a bromodeoxyuridine proliferation assay (Fig. 6B) ⇓ . Although no change in cell cycle profile was noted on days 2 or 4 (data not shown), by day 7, a very significant increase in the fraction of apoptotic cells was noted. This was accompanied by a depletion of cells in G0/G1 and S phase and an increase in the number of cells in G2-M phase. These findings are suggestive of G2-M arrest and apoptosis as a result of HNF3β induction. The changes were slightly more prominent in clone 31 than in clone 4. No change in the cell cycle profile or in the rate of apoptosis was noted in the control cell line. Changes in the rate of apoptosis were also determined by annexin/propidium iodide flow cytometry. These studies have shown that by day 7 of HNF3β induction, there was a highly significant increase in the rate of cell death in both clones (from 3 to 20% apoptotic cells in clone 31; Fig. 6C ⇓ ). No change in the rate of apoptosis was seen on doxycyline withdrawal in the parental cell line.
We have also performed methylcellulose-based clonogenic assays to assess the ability of H358 cells to form colonies with and without the induction of HNF3β expression (Fig. 6D) ⇓ . In clone 4, we observed a significant reduction in the colony-forming ability of the cells when grown in the absence of doxycycline. Interestingly, clone 31 cells were not able to form colonies even in the presence of doxycycline, suggesting that possibly even a slight “leakiness” of HNF3β expression might lead to very substantial reduction of colony-forming ability.
We previously showed that C/EBPα is a tumor suppressor in lung cancer and demonstrated strong growth-inhibitory activity of C/EBPα in lung cancer cell lines (20) . In the present study, we have identified transcriptional changes secondary to conditional expression of C/EBPα in a C/EBPα nonexpressor lung adenocarcinoma cell line, H358. The tight regulation and strong inducibility of C/EBPα expression proved to be optimal for this study. As shown in Tables 1 ⇓ and 2 ⇓ , the changes in gene expression levels were marked and highly consistent in-between triplicate samples but only for a very limited set of genes, suggesting that we were able to identify the initial wave of transcriptional changes as a result of C/EBPα expression.
The genes identified this way also correlate very well with prior knowledge about the function of C/EBPα. Three groups of genes stand out from the list of up-regulated genes. The first is genes involved in acute phase reaction, such as interleukin 8, COX-2, and numerous tumor necrosis factor-inducible genes. It is well known that C/EBP family members regulate the acute phase response (34, 35, 36) . In fact, many acute phase genes have both C/EBP as well as nuclear factor κB binding sites in their promoters, suggesting that C/EBP proteins and nuclear factor κB cooperatively regulate the acute phase response (37, 38, 39, 40) . In addition, prior studies have also shown that C/EBPα can directly regulate a number of acute phase genes in hepatocytes (41) . Secondly, many genes up-regulated by C/EBPα play a role in terminal metabolism, such as enzymes of metabolic pathways (hepatic dihydrodiol dehydrogenase, folate receptor, and quinone oxidoreductase) and are suggestive of the transcriptional profile of a more differentiated cell. The third group of genes (such as adipophilin and ceramide glucosyltransferase) is involved in lipid metabolism. For instance, adipophilin, a gene up-regulated by C/EBPα, is a prominent protein component of lipid storage droplets and is thought to be necessary for the formation and cellular function of these structures (42 , 43) . It is very interesting to note that the induction of C/EBPα in the H358 cell line did indeed lead to the appearance of lipid droplets in the cytoplasm as determined by electron microscopy (20) , a feature of more mature pneumocytes. C/EBPα plays a major role in the development of preadipocytes to adipocytes and is known to regulate the expression of many genes involved in lipid metabolism in adipose tissues (44, 45, 46, 47, 48, 49, 50) . It might not be surprising that C/EBPα plays a similar role in alveolar cells, where the production of lipids in the form of surfactant is critical to the proper functioning of the airway epithelium. Also, both of these sets of genes appear to be markers of a more differentiated cellular state. C/EBPα is a critical differentiation factor in a number of cell types, such as hepatocytes, adipocytes, and myeloid cells (44 , 51, 52, 53, 54) . On the basis of our findings, it is likely to play a similar role in airway epithelial cells as well. A hyperproliferation of type II pneumocytes has been observed in C/EBPα knockout mice supporting such a role (19) . In fact, in previous studies, we have described intracellular changes detected by electron microscopy suggestive of airway epithelial differentiation in the identical cell line system, where these transcriptional profiling studies were done (20) . Also, a strikingly large number of C/EBPα-repressed genes are proangiogenic factors or growth factors. These findings are consistent with the role of C/EBPα in growth arrest (29 , 31 , 32 , 55) .
A validation of our hypothesis that using such an approach would enable us to identify direct C/EBPα targets is that a number of genes identified are known to be regulated directly by C/EBP members (such as interleukin 8 and COX-2; Refs. 56 and 57 ). Additional validation comes from a study (58) in which primary human CD34+ cells were transduced with a retroviral construct that expresses the C/EBPα cDNA fused in-frame with the estrogen receptor ligand-binding domain. In these cells, the addition of estradiol leads to granulocytic differentiation. This system was used to identify target genes of C/EBPα in primary human hematopoietic cells by the use of Affymetrix oligonucleotide arrays. Quite strikingly, many of the regulated genes (e.g., tumor necrosis factor-inducible gene 14, COX-2, HM74, glutaredoxin, ARF-like protein, adipophilin, and p63) identified were common to the C/EBPα target genes identified in our study.
Our transcriptional profiling study identified also HNF3β (also known as Foxa2) as one of the most highly induced downstream targets of C/EBPα in lung cancer cells. This finding is particularly intriguing given the known function of HNF3β in foregut development and transcriptional regulation in the mature airway epithelium (10 , 59 , 60) . HNF3β is a member of the forkhead family of transcription factors. The amino acid sequence of HNF3β is highly conserved (97.8% homology with between the human and rat orthologue; Ref. 27 ). HNF3β binds DNA through a homologous winged helix motif common to a number of proteins known to be critical for determination of events in embryogenesis, the forkhead box (61) . Targeted disruption of HNF3B results in embryonic lethality with defective development of the foregut endoderm (62) . In the adult, HNF3β regulates the transcription of numerous liver-enriched genes, and the HNF3 proteins play a pivotal role in the regulation of metabolism and in the differentiation of metabolic tissues such as the pancreas and liver (63 , 64) . HNF3β is known to be a key regulator of airway epithelial differentiation (21 , 23 , 30 , 65) . It influences the expression of a number of target genes in the respiratory epithelium, such as thyroid transcription factor-1, another master gene of airway epithelial cell differentiation, surfactant protein-B, and Clara-cell secretory protein (21 , 22 , 66) . HNF3β is expressed at the onset of lung morphogenesis (day 10 gestation) and throughout lung development (65 , 67) . It is expressed at highest levels in proximal bronchial and bronchiolar epithelial cells, but it is also expressed in type II pneumocytes. It has been shown previously that members of the C/EBP family, including C/EBPα, bind and activate the LF-H3β site of the HNF3β promoter, which might mediate cell-specific expression of HNF3β (25) . Our studies using both in vitro electrophoretic mobility as well as in vivo chromatin immunoprecipitation assays clearly show direct binding of this promoter element by C/EBPα. This finding further suggests that C/EBPα might act as a master regulator of airway epithelial differentiation not only by controlling the expression of sets of genes characteristic of the differentiated alveolar pneumocytes but also by turning on a secondary wave of differentiation events by inducing the expression of HNF3β itself.
Although several other members of the forkhead family of proteins, such as FoxO family members, have been implicated in carcinogenesis (68 , 69) , to our knowledge, no studies have ever investigated the role of HNF3β in any type of cancer. Our results show that whereas HNF3β, as expected, is expressed strongly in normal lung, its expression is lost in more than half of all of the lung cancer cell lines. Genomic sequencing of 31 lung cancer cell lines also identified two mutant forms of HNF3β. To our knowledge, these are the first mutant forms of HNF3β ever described in cancer. Treatment of HNF3β nonexpressor cell lines with the demethylating agent deoxyazacytidine also leads to up-regulation of HNF3β expression in 3 of 4 cell lines examined, suggesting that promoter methylation might be a mechanism of silencing. Indeed, bisulfite sequencing of these HNF3β nonexpressing cell lines revealed evidence of promoter hypermethylation in a CpG-rich segment of the promoter/exon 1 junction in all of the 4 cell lines examined. These results do suggest both point mutations and promoter hypermethylation as genetic/epigenetic mechanisms leading to aberrations of the HNF3β transcriptional program of airway epithelial differentiation. It is interesting to note that a recent high-frequency allelotyping study did find an ∼40% rate of loss of heterozygosity of the chromosomal region of HNF3β, 20p11 in non-small cell lung cancers suggestive of the presence of an as yet unidentified tumor suppressor gene in this region (6) . Our results suggest that HNF3β is a strong candidate as the 20p11 tumor suppressor. We are conducting additional studies to define the full spectrum of mutations, loss of heterozygosity, and promoter hypermethylation in primary non-small cell lung cancer specimens to fully establish the role of HNF3β as a novel tumor suppressor.
We have also generated a doubly stably transfected, tetracycline-regulatable cell line system for the conditional expression of HNF3β to study its cellular effects. This cell line system allowed us to show that conditional expression of HNF3β leads to proliferation arrest, apoptosis, and loss of clonogenic ability, further corroborating our hypothesis that HNF3β might act as a tissue-specific tumor suppressor in the development of lung cancer. Although HNF3β has a well-established role in driving maturation of certain cell types, its role in growth control or apoptosis has not been demonstrated as of yet. On the other hand, other forkhead family members are known mediators of cell signaling pathways in control of cell cycle progression (70 , 71) . This inducible model system should also allow us to define the transcriptional program of differentiation of airway epithelial cells as well as the pathways involved in cell cycle arrest and apoptosis as result of HNF3β induction. We are in the process of performing transcriptional profiling studies on the above-described cell line system to fully delineate the transcriptional changes after HNF3β induction.
In summary, our studies have identified the downstream targets of C/EBPα, a candidate tumor suppressor in lung cancer, in neoplastic airway epithelial cells. These studies led to the identification of HNF3β, a known differentiation factor in airway epithelium as a downstream target of C/EBPα. Additional studies of HNF3β show down-regulation of its expression in lung cancer cell lines, identify novel mutant forms of HNF3β in lung cancer, and demonstrate that promoter methylation is a putative mechanism for epigenetic silencing of the HNF3β gene. Our studies also show strong growth-inhibitory and proapoptotic properties of HNF3β and suggest that HNF3β can indeed act as a tissue-specific tumor suppressor in lung cancer.
Grant support: NIH Grant Specialized Programs of Research Excellence in Lung Cancer PA20-CA090578-01A1 (D. Tenen), American Association for Cancer Research/Cancer Research Foundation of America/Astra-Zeneca Young Investigator Award for Translational Lung Cancer Research (B. Halmos), and the Clinical Investigator Training Program of Harvard/MIT (B. Halmos).
Note: Supplementary data for this article can be found at Cancer Research Online (http://cancerres.aacrjournals.org).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Daniel G. Tenen, Harvard Institutes of Medicine, Harvard Medical School, 77 Avenue Louis Pasteur, Boston, MA 02115. Phone: (617) 667-5561; Fax: (617) 667-3299; E-mail:
- Received December 30, 2003.
- Revision received March 19, 2004.
- Accepted April 9, 2004.
- ©2004 American Association for Cancer Research.