The Arf-inducing Transcription Factor Dmp1 Encodes a Transcriptional Activator of Amphiregulin, Thrombospondin-1, JunB and Egr1

Abstract

Dmp1 (Dmtf1) encodes a Myb-like transcription factor implicated in tumor suppression through direct activation of the Arf-p53 pathway. The human DMP1 gene is frequently deleted in non-small cell lung cancers, especially those that retain wild-type INK4a/ARF and/or p53. To identify novel genes that are regulated by Dmp1, transcriptional profiles of lung tissue from Dmp1-null and wild-type mice were generated using the GeneChip Microarray. Comparative analysis of gene expression changes between the two groups resulted in identification of numerous genes that may be regulated by Dmp1. Notably, amphiregulin (Areg), thrombospondin-1 (Tsp-1), JunB, Egr1, adrenomedullin (Adm), Bcl-3 and methyl-CpG binding domain protein 1 (Mbd1) were downregulated in the lungs from Dmp1-null mice while Gas1 and Ect2 genes were upregulated. These target genes were chosen for further analyses since they are involved in cell proliferation, transcription, angiogenesis/metastasis, apoptosis, or DNA methylation, and thus could account for the tumor suppressor phenotype of Dmp1. Dmp1 directly bound to the genomic loci of Areg, Tsp-1, JunB and Egr1. Significant upregulation or downregulation of the novel Dmp1 target genes was observed upon transient expression of Dmp1 in alveolar epithelial cells, an effect which was nullified by the inhibition of de novo mRNA synthesis. Interestingly, these genes and their protein products were significantly downregulated or upregulated in the lungs from Dmp1-heterozygous mice as well. Identification of novel Dmp1 target genes not only provides insights into the effects of Dmp1 on global gene expression, but also sheds light on the mechanism of haploid insufficiency of Dmp1 in tumor suppression.

Introduction

Dmp1 (cyclin D binding myb-like protein 1; also called Dmtf1) is a transcription factor that was originally isolated in a yeast two-hybrid screen of a murine T-lymphocyte library with cyclin D2 as bait.^1–3 Dmp1 binds to the DNA consensus sequences CCCG(G/T)ATG(T/C), a subset of which are also bound by proteins of the Ets family.^4–6 Generally, Ets proteins are known to act as positive or negative regulators of genes expressed in various biological processes, including those that control cellular proliferation, differentiation, apoptosis, metastasis, tissue remodeling, angiogenesis and transformation.⁷ It follows that Dmp1 may potentially play roles in these biological processes as well. Dmp1 mRNA is ubiquitously expressed in normal mouse tissue; its protein levels are highest in the testis and thymus, followed by the lungs and brain.^{8, 9}

We have accumulating evidence for transcriptional regulation of Dmp1. The Dmp1 promoter receives oncogenic signals from activated Ras via the Raf-MEK-ERK-Jun pathway which, in turn, causes activation of the p19^Arf promoter and induces cell cycle arrest to quench oncogenic signaling.¹⁰ Conversely, the Dmp1 promoter is repressed by physiological mitogenic stimuli and overexpression of E2F proteins.⁹ The Dmp1 promoter is also repressed by genotoxic stimuli mediated by NF-κB¹¹ and by the Wilms tumor gene product, WT1.¹² Both Dmp1-knockout and heterozygous mice are prone to tumor development. Tumors isolated from Dmp1^+/− mice seldom lose the wild-type Dmp1 allele, indicating that Dmp1 is haplo-insufficient for tumor suppression.^13–14 Deletion or mutation of Arf or p53 is rare in tumors from Dmp1-knockout mice. Thus, Dmp1 has been considered a physiological regulator of the Arf-p53 pathway in vivo.^2,3,13,14

When Dmp1 expression was studied in vivo using specific antibodies, the protein was expressed in alveolar macrophages and type II cells in the lung.⁹ The Dmp1 gene was deleted in 40 % of K-ras^LA lung cancers, and the gene deletion was mutually exclusive with mutations of the p53 gene.¹⁴ Likewise, loss of heterozygosity (LOH) of the human DMP1 gene (hDMP1) is found in ~35 % of human non-small-cell lung carcinomas (NSCLC), suggesting the primary involvement of hDMP1 in lung carcinogenesis.¹⁴ The LOH of hDMP1 was found in a mutually exclusive fashion with those of p14^ARF and/or p53 in 80–90 % of the cases, indicating that hDMP1 is a physiological regulator of the ARF-p53 pathway in human tissues as well.^2,3,14

In this study, high-density oligonucleotide microarrays were utilized to identify previously unknown target genes for the Dmp1 transcription factor that are involved in tumor suppression, tumorigenesis, and cell cycle regulation.^15–18 Real-time reverse transcription PCR (RT-PCR) of selected genes validated the microarray results and demonstrated a wide range of differentially expressed genes. Here we report that several genes that are involved in cell cycle/proliferation/transcription (amphiregulin, adrenomedullin, JunB and Egr1), angiogenesis and metastasis (thrombospondin-1), apoptosis (Bcl-3), and DNA methylation (methyl DNA-binding protein 1) were found to be downregulated in tissues of Dmp1-deficient mice, and thus, they could represent novel transcriptional targets for Dmp1. We also identified genes that were upregulated in the lungs (Gas1 and Ect2, cell cycle/proliferation) from Dmp1^−/− mice. The genes identified in this study constitute promising candidates for further studies to clarify the unrevealed functions and novel signaling pathways regulated by Dmp1.

Materials and Methods

Dmp1-knockout mice

The mice used in this experiment contained a targeted disruption of Dmp1 as described previously.⁸ Two Dmp1^+/− females were backcrossed with a single C57BL/6 male for more than 8 generations to minimize the effects of strain-dependent difference of gene expression. The mice were overall >99 % pure for C57BL/6 at G8. All 7 markers on mouse chromosome 5 where Dmp1 is located had been replaced with C57BL/6 markers at G8. Mice were observed daily and maintained in accordance with the Guide for the Care and Use of Laboratory Animals.

Gene expression profiling with Affymetrix 430A microarrays

For each phenotype, RNA was pooled from three separate mice (4 to 6 weeks of age) of the same genotype. Total RNA was isolated from the homogenized lung tissues of Dmp1 wild-type and knock-out mice utilizing Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. RNA pellets were resuspended in nuclease-free water (Sigma, St. Louis, MO). The quality of RNA was determined spectrophotometrically using an Agilent 2100 Bioanalyzer. First-strand cDNA synthesis was accomplished utilizing the Oligo dT primer method and Superscript II enzyme provided by a One-Cycle-cDNA Affymetrix Synthesis Kit # 900431 (Affymetrix, Santa Clara, CA). From cDNA, cRNA was synthesized using the Affymetrix IVT labeling kit # 900449. Synthesized cRNA was labeled during transcription (MEGAscript® system; Ambion, Inc., Austin, TX) with biotin-11-cytidine triphosphate and biotin-16-uridine triphosphate (Enzo Diagnostics, Farmingdale, NY).

Biotinylated cRNA products were hybridized to Affymetrix GeneChip arrays (murine GeneChip 430A) for 16 hours at 45°C using the manufacturer’s hybridization buffer. After hybridization, GeneChip arrays were washed and stained by a fluidics station 450 using the ProkGE-WS2 fluidics script (Affymetrix). Once the probe arrays were hybridized and washed, they were scanned using the Affymetrix GeneChip 7G plus scanner (Hewlett-Packard Gene array scanner G2500A; Affymetrix) according to the manufacturer’s procedures. The known Dmp1 target gene, Cdkn2a (p19^Arf/p16^Ink4a, exons 2–3) was expressed at very low levels in wild-type lungs, but was not detectably expressed in Dmp1^−/− lungs by our murine GeneChip 430A array, confirming the reliability of our GeneChip microarray.

Data Analysis

To process the primary quantitative data from each hybridizationexperiment, probe signals in raw CEL files were summarized using MicroArray Suite 4.0 (Affymetrix) and the expression data was then analyzed by Genesis software (Institute for Genomics and Bioinformatics - Graz University of Technology, Graz, Austria). To determine if the RNA abundance corresponding to each gene encoded on the array was detectable, several factors were evaluated, including the number of parameters representing each gene in which the intensity of the perfect match hybridization reaction exceeded that of the mismatch hybridization signal and the perfect match/mismatch ratios for each set of corresponding probe pairs. The raw data were then normalized on a per-chip basis for both Dmp1^+/+ and Dmp1^−/− lung samples. In this way, one is able to see differences between Dmp1^+/+ and Dmp1^−/− animals. Gene induction was considered significant if the change in this normalized intensity was above 2-fold.

Identification of gene expression patterns

To identify the genes that were differentially expressed in lung tissues of wild-type and Dmp1-null mice, Genesis software was utilized. A total of 300 probe sets, corresponding to 250 unique genes and few expressed sequence tags, were identified based onany expression that differed by 2-fold or more between Dmp1^+/+ and Dmp1^−/− tissue samples. The differentially expressed transcripts were clustered according to their expressionprofiles utilizing a K-means algorithm in the Genesis software. K-means analysis places gene expression profiles into a predefinednumber of clusters based on relative expression across the sample series.

Real-time PCR

To validate the gene expression levels determined by microarray analysis, the expression of several candidate genes including Areg, Tsp-1, JunB, Egr1, Bcl-3, Adm, Mbd1, Gas1 and Ect2 were checked by real-time quantitative PCR with an ABI 7500 system.^9,14 Taqman assays specific to cDNAs were used in the study (Applied Biosystems [ABI], Foster City, CA). The ID number of TaqMan primers and probes obtained from ABI were as follows: Areg: Mm00437583_m1, Tsp-1: Mm00449022_m1, JunB: Mm00492781_s1, Egr1: Mm00656724_m1; Bcl-3: Mm00504306_m1, Adm: Mm00437438_g1, Mbd1: Mm00522100_m1, Gas1: Mm01700206_g1, Ect2: Mm00432964_m1, p27^kip1: Mm00438168_m1, and internal control β-actin: Mm02619580_g1. Specific primers for Dmp1 and p19^Arf and p16^Ink4a Taqman assays were custom-designed at ABI. Values were normalized to the expression of the β-actin internal reference. Average gene expression levels from three mice of each genotype were shown in bar graphs with standard errors. Statistical analyses were conducted by unpaired Student t-tests. The significance levels were shown in the number of stars, i.e., * was used for 0.01< P <0.05, ** for 0.001< P <0.01 and *** for P <0.001.

Electrophoretic Mobility Shift Assay (EMSA) and chromatin immunoprecipitation

To confirm the binding of Dmp1 protein to the Areg promoter, EMSA was performed as described previously with recombinant Dmp1 protein prepared in Sf9 cells.⁶ In brief, the recombinant Dmp1 protein was incubated with an [α³²P]-labeled oligonucleotide probe covering the consensus sequence of the Areg promoter obtained by annealing the sense oligonucleotide 5′-TTTACCCCAGAACGGGATGGGGAGCAT-3′ and its reverse complementary sequence (Areg#1) (Figure 3a, top panel). The putative Dmp1-binding site is underlined. To demonstrate the direct binding of Dmp1 to the sequence located at the intron 1 of the Areg gene, EMSA was conducted with the oligonucleotide 5′-TTTGCTCAGGTGAGCATCCGCCACCTCTCAGG-3′ and its reverse complementary strand (Areg#2). To examine whether Dmp1 binds to the 3′ non-coding sequence of mouse thrombospondin-1 (Tsp-1) genomic DNA, oligonucleotide 5′-TTTCAGAGCCTGCGTATGTTATAACG-3′ and its reverse complementary strand (Tsp-1#1) were used (Figure 3a, the 2^nd panel). To study if Dmp1 binds to the 3′ non-coding sequence of the mouse JunB genomic DNA, the oligonucleotide 5′-TTTTTTTCTTCCAAGCTGGGCGGATGGAGGGAGAAGATT G-3′ (JunB#1) and its reverse complementary strand was used (Figure 3a, the 3^rd panel). To study whether Dmp1 binds to the mouse Egr1 promoter, oligonucleotide 5′-TTTAAAACGAAGGTGGAGGCGGATGGAGAGAGAGAGAGAGAG-3′ and its reverse complementary strand (Egr1#1) were used (Figure 3a, the 4^th panel). For control EMSA, the oligonucleotide 5′-TTTGAGCCCTGGGAGCCATCCGAGCGGCTTTCAACAGACC-3′ (JunB#2; Figure 3a, Supplementary figure 1) and its complementary sequence were used. For competition assays, a 100-fold excess of unlabeled oligonucleotide was added to the reaction mixtures before probe incubation. To verify the identityof the proteins in shifted complexes, reaction mixtures were incubated with control non-immune rabbit serum or with antibodies specific to Dmp1 (RAX: [rabbit antibody project X] to Dmp1 amino acids 136–150, RAF [rabbit antibody project F] to Dmp1 amino acids 752–760, RAJ [rabbit antibody project J] to the DNA-binding domain of Dmp1).^1,9 Chromatin immunoprecipitation for Areg, Tsp1, JunB and Egr1 genomic loci was conducted with RAX, RAZ (rabbit antibody project Z to the Dmp1 amino acids 740–756),³ and RAD (rabbit antibody project D to the full-length His-tagged Dmp1) antibodies as described previously.^10,11 For Areg #1, sense 5′-AGCAGCAAGGACAGAATGGAC-3′ and antisense 5′-GCTGGGAAAGGAGGCTGTAGT-3′; for Areg #2, sense 5′-GCTGTTGCTGCTGGTCTTAGG-3′ and antisense 5′-GTTTCCATCAGCTGGTCCTTG-3′; for Tsp-1#1, sense 5′-GAAAGCCCTACTGGTCCATCC-3′ and antisense 5′-TGCACCATCACCACATTTCTC-3′; for JunB#1, sense 5′-TGTGGAGAAGAAACACGCACT-3′ and antisense 5′-CCTGAAAGGAGCGAGCTACAA-3′; for Egr1#1, sense 5′-GGTGAGACACAGGTTTGGTCA-3′ and antisense 5′-GGGAGGTTACTCTTCGGCTCT-3′ were used. As a control, we amplified the sequence 1.8 kb downstream from the JunB#1 site with the sense primer 5′-GGACCCCAGAATAAGCAGAGG-3′ and antisense primer 5′-CTGAGAACGGGTGCTGGTTAG-3′.

Figure 3.

Dmp1 directly binds to Areg, Tsp-1, JunB and Egr1 loci. (a) Genomic structure of the murine Areg, Tsp-1, JunB and Egr1 loci. The non-coding regions are shown in white and the coding regions are shown in blue. Top panel. There are two possible Dmp1-sites in the Areg promoter and in the first intron. The second panel. The murine Tsp-1 locus. The Dmp1 consensus sequence is located at the 3′ of the Tsp-1 gene. The third panel. The murine JunB locus. JunB is a single exon gene. The Dmp1 consensus sequence (JunB#1) is located in the 3′ region of the JunB gene. JunB#2 shows the position of the oligonucleotide sequence used for the control EMSA. The fourth panel. The murine Egr1 locus. The Egr1 gene consists of 2 exons and the Dmp1 motif is located 2 kb from the transcription initiation site. (b–f) EMSA was conducted with a [α³²P]-labeled probe covering possible Dmp1-consensus sequences, using recombinant Dmp1 protein prepared in Sf9 cells. The probes were: (b) Areg#1 within the Areg promoter, (c) Areg #2 within the first intron, (d) Tsp-1#1 at the 3′ of the locus, (e) JunB#1 at the 3′ of the locus and (f) Egr1#1 in the Egr1 promoter. Competition assays were conducted with 100-fold excess scrambled oligos (n.s.), Ets-specific oligos (M3),¹ or the cold probe itself. Supershift assays were conducted with control normal rabbit serum (NRS) or with three different antibodies to Dmp1 (RAX [to the peptide 136–150 at the DNA-binding domain], RAJ [to the DNA-binding domain] or RAF [to the peptide 752–760 at the C-terminus]).¹,⁹ (g) Chromatin immunoprecipitation demonstrating the binding of endogenous Dmp1 to the Areg, Tsp-1, JunB and Egr1 genomic loci. RAZ antibody recognizes the Dmp1 peptide 740–756,⁵ RAD antibody recognizes the full-length His-tagged Dmp1 protein. Total: total input chromatin corresponding to 4 % of the precipitated samples.

Transfection of C10 cells and Actinomycin D treatment

Non-transformed mouse alveolar epithelial cell line C10¹⁹ was received from Dr. LM Anderson, NCI/NIH. C10 cells (5 ×10⁵ cells per 60 mm dish/4 mL) were transfected with 4μg of pFLEX1-Dmp1 or empty vector and were harvested for RNA extraction at 48 hours after transfection (Figure 4a). For inhibition of de novo mRNA synthesis, cells were treated with 2μg/mL of Actinomycin D (Sigma) for 6 hours before harvest. The transfection efficiency was 30 – 70 % with Genejuice (Novagen, EMD Chemicals, Gibbstown, NJ) in C10 cells.

Figure 4.

Dmp1 regulates target gene expression in a non-transformed murine alveolar epithelial cell line, C10. C10 cells were transfected with pFLEX1-Dmp1 vector or control vector and mRNA levels were quantitated 48 hours after transfection by real-time RT-PCR. (a) Induction of Areg, Tsp-1, JunB, Egr1, Bcl-3, Mbd1 and Adm genes by Dmp1. These gene inductions were nullified by Actinomycin D treatment, suggesting that Dmp1 increased the mRNA levels by transcriptional activation. (b) Repression of the Gas1 and Ect2 genes by Dmp1. This gene repression was also neutralized by Actinomycin D treatment, suggesting that the repression occurred at the transcriptional level. The p27^kip1 gene, which did not show any significant change in the GeneChip Microarray, was not affected by Dmp1 overexpression in C10 cells. (c) Primary quantitative primary transcript real-time PCR (qPT-PCR) analysis on Areg, Tsp1 and Egr1 genes. The relative abundance of immature (pre) and mature Areg, Tsp-1 and Egr1 transcripts were similarly elevated at 24 and 48 hours in C10 cells transfected with the Dmp1 expression vector.

Quantitative primary transcript real-time PCR (qPT-PCR)

C10 cells were transiently transfected with pFLEX1-Dmp1 or empty vector and were harvested for RNA extraction at 24 and 48 hours (Figure 4c). qPT-PCR was conducted as previously described²⁰. Specific primers for mouse Areg exon 4 - intron 4, Tsp-1 exon 7- intron 7, Egr1 exon 1 - intron 1 real-time PCR were custom-designed at ABI. The primer sequences will be available upon request.

Western blotting

Murine Areg, Tsp-1, JunB, Egr1, Ect2 and Actin proteins were detected by direct Western blotting with 80μg of the lung lysates from Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice. For detection of Dmp1 in murine lung tissues, immunoprecipitation (IP) - Western blotting was conducted by lysing the lungs in EBC buffer,¹ immunopprecipitation of 250μg of lysate with 5 μg of RAZ antibody,³ followed by Western blotting with the RAX antibody⁹. IP-Western blotting was necessary since the Dmp1 protein levels are relatively low in the lungs, especially in Dmp1^+/− mice. The Areg antibody (AF989) was purchased from R&D Systems (Minneapolis, MN). The antibodies for Tsp-1 (sc-12312), JunB (sc-8051), Egr1 (sc-101033), Ect2 (sc-1005), p27^kip1 (sc-1641) and Actin (sc-47778, sc-1616) were obtained from Santa Cruz Biotechnology (Santa Cruz Biotechnology, Santa Cruz, CA).

Immunohistochemistry

Sections of paraffin-embedded lung tissues from Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice in C57BL/6 background were mounted on immunohistochemistry slides. The paraffin was removed by melting at 70°C for 30 min, followed by soaking in xylene substitute (Thermo Electron Corp., Pittsburgh, PA) and rehydrating through decreasing ethanol concentrations. Antigen retrieval was performed by boiling slides in DakoCytomation Antigen Retrieval solution (Dako, Carpinteria, CA).⁹ Slides were blocked in 1 % BSA/PBS, containing 25μg/ml serum from the same species as the secondary antibody. After incubating with primary and secondary antibodies, staining was visualized with an alkaline phosphatase or peroxidase-based detection kit (ABC kit; Vector Laboratories, Burlingame, CA), followed by hematoxylin counterstaining. Negative controls (incubation with secondary antibody only) for the immunohistochemical reactions were performed on serial sections.

Results

Expression profile from microarray analysis

To identify novel transcriptional target genes of Dmp1 that could account for tumor-prone phenotypes of Dmp1-deficient mice, GeneChip Microarray was performed using murine gene chip 430A. Out of 3,771 full-length known genes available in the gene chip, 127 genes (3.7 %) in the lung were upregulated at least 2-fold in Dmp1^−/− mice in comparison to Dmp1^+/+ mice. In contrast, 104 genes in the lung (2.8 %) were downregulated a minimum of 2-fold in Dmp1^−/− mice in comparison to Dmp1^+/+ mice. Among the total of 4,371 expressed sequence tags (ESTs), 0.5 % were upregulated at least 2-fold (data not shown). According to the gene functional analysis, the selected genes were further classified into different corresponding groups: cluster I, apoptosis and cytolysis; cluster II, cell surface receptor and signal transduction; cluster III, immune and inflammatory response; cluster IV, cell cycle, proliferation and transcription; and cluster V, metabolism and enzymes (Figure 1, Table 1).

Figure 1.

Hierarchical clustering analysis of the 150 genes (displayed vertically) in the lung of Dmp1^+/+ and Dmp1^−/− tissue samples. (a) The total 150 genes were clustered based on their expression patterns among the selected tissue samples. Cluster I, apoptosis and cytolysis genes; cluster II, cell surface receptor and signal transduction genes; cluster III, immune and inflammatory response genes; cluster IV, cell cycle/proliferation/transcription factor genes; and cluster V, genes for metabolism and enzymatic reactions. (b) Expression patterns of selected genes in five groups based on their functional classification (see Table 1 for their corresponding classification). Colored bars indicate relative expression levels. Red indicates upregulated genes in Dmp1^−/−; downregulated genes in Dmp1^−/− mice are shown in green. Genes that are expressed at higher levels are assigned progressively brighter shades of red, whereas genes expressed at low levels are assigned shades of green. (c) Correlation of DNA microarray data and TaqMan RT-PCR assays. Total RNA was isolated from the lungs of Dmp1^+/+ and Dmp1^−/− mice. A representative of three separate TaqMan RT-PCR experiments, all with similar results, is shown. Real-time PCR data were normalized for levels of β-actin. The fold changes obtained by both methods were log-transformed and the values were plotted against each other to evaluate their correlation. Selected genes are shown for illustrative purposes.

Table 1.

Gene Name	Genebank	Dmp1+/+ vs. Dmp1−/− (Lung)
Apoptosis and Cytolysis
Tumor necrosis factor receptor superfamily, member 12a (Tnfrsf12a)	NM_013749	−2.54
B-cell leukemia/lymphoma 3 (Bcl3)	NM_033601	−3.18
Cell Surface Receptor and Signal Transduction
Cholinergic receptor, nicotinic, alpha polypeptide 6	AW048864	7.16
Chloride channel calcium activated 3 (Clca3)	NM_017474	6.94
Cystic fibrosis transmembrane conductance regulator (cftr11MT1)	X72693	6.36
Potassium voltage-gated channel, shaker-related subfamily, member 5 (Kcna5)	NM_008419	4.50
Glucose transporter 4 (Glut4)	AB008453	3.80
Frizzled homolog 2 (Fzd2)	BB371406	3.58
Septin 6 (Sept 6)	BG920446	2.48
Breast cancer anti-estrogen resistance 3 (Bcar3)	NM_013867	−2.06
Tumor-associated antigen 1 (Taa1)	NM_009310	−2.14
Plasminogen activator urokinase receptor (Plaur)	X62701	−2.34
Loricrin (Lor)	NM_008508	−2.80
Cytokine inducible SH2-containing protein 3 (Cish3)	BB241535	−3.58
Repetin (Rptn)	NM_009100	−4.64
Immune and Inflammatory Response
Hepcidin antimicrobial peptide (Hamp)	NM_032541	3.12
CD14 antigen (CD14)	NM_009841	−2.06
Immunoglobulin heavy chain C region (Igh-4)	M60430	−2.42
Lymphocyte antigen 6 complex, locus G6C (Ly6g6c)	NM_023463	−2.70
Adenosine deaminase (Ada)	NM_007398	−4.04
Defensin beta 4 (Defb4)	NM_019728	−4.48
Cell Cycle, Proliferation, and Transcription
Epithelial cell transforming sequence 2 (Ect2)	NM_007900	6.22
Growth arrest-specific gene 1 (Gas1)	NM_008086	5.40
Zinc finger protein 318 (Zfp318)	NM_021346	3.58
Growth differentiation factor 10 (Gdf10)	L42114	2.82
Snail homolog 1 (Snai1)	NM_011427	−2.16
Heparin binding EGF-like growth factor (Hegfl)	NM_010415	−2.24
FBJ osteosarcoma oncogene (c-Fos)	AV026617	−2.32
Pituitary tumor-transforming 1 (Pttg1)	AV105428	−2.46
NFIL3E4BP4-like transcription factor (Nfil3)	AY061760	−2.50
v-Maf musculoaponeurotic fibrosarcoma (Maf)	NM_010755	−2.52
Methyl-CpG binding domain protein 1 (Mbd1)	AK007371	−2.56
Early growth response 2 (Egr2)	NM_009704	−3.12
JunB	NM_008416	−3.20
Adrenomedullin (Adm)	NM_009627	−3.36
Amphiregulin (Areg)	NM_009704	−4.68
Early growth response 1 (Egr1)	NM_007913	−5.20
Metabolism and Enzymes
Transmembrane protease, serine 3 (Tmprss3)	NM_080727	4.80
Metallothionein 1 (Mt1)	NM_013602	−2.32
Nocturnin	AF199491	−2.58
Metallothionein 2 (Mt2)	AA796766	−2.82
Cholesterol 25-hydroxylase (Ch25h)	NM_009890	−3.24
Calmodulin 4 (Calm4)	NM_020036	−3.52
Thrombospondin 1 (Tsp1)	M87276	−3.84
Metallothionein 4 (Mt4)	NM_008631	−5.29

Using high-throughput methodology, detection of multiple genes with their related cellular functions that might be involved in tumorigenesis in Dmp1-deficient mice is possible. Amphiregulin (Areg), adrenomedullin (Adm), JunB, Egr1, thrombospondin-1 (Tsp-1), Bcl-3, methyl DNA binding protein-1 (Mbd1), Gas1 and Ect2 were selected as candidate transcriptional target genes that contribute to the tumor-prone phenotype since these genes are involved in cell growth/differentiation (Areg, Adm, Gas1 and Ect2), signal transduction/transcription (JunB and Egr1), angiogenesis/metastasis (Tsp-1), apoptosis (Bcl-3) or DNA methylation (Mbd1). These nine genes (hereafter called ‘novel Dmp1 targets’ or ‘novel Dmp1 target genes’) will be validated in later experiments.

Validation of microarray data by quantitative RT PCR

DNA microarray analysis measures expressional changes in thousands of transcripts and, therefore, the statistical evaluation of this data is prone to some errors. For this reason, we validated our initial microarray data by performing real-time RT-PCR (Figure 2). Results of the microarray analyses, real-time PCR and linear regression analyses demonstrated a strong correlation with Pearson’s correlation coefficient (R) of 0.931 for the lung (Figure 1c). In our study, the selected genes showed good correlation by real-time RT-PCR.

Figure 2.

Real-time RT-PCR analyses of target gene expression in the lungs from Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice. RNA was extracted from three mice of each Dmp1 genotype (C57BL/6, 8–12 weeks old males) and gene expression was examined by real-time Taqman assay. The Dmp1 gene was downregulated to 35 % of wild-type levels in Dmp1^+/− lungs. It should be noted that novel Dmp1 target gene expression levels were not significantly different between Dmp1^+/− and Dmp1^−/− lungs, suggesting the mechanism of haploid insufficiency of Dmp1 for tumor suppression.

Expression of novel Dmp1 target genes in Dmp1-hetrozygous lungs

Published studies have shown that Dmp1 is haplo-insufficient for tumor suppression.^13,14 Therefore, we examined relative expression levels of Dmp1 target genes in lungs with three different Dmp1 genotypes (Figure 2). The statistically significant differences are shown in stars (* for P < 0.05, **for 0.001< P <0.05 and *** for P < 0.001). The average Dmp1 gene expression level in Dmp1^+/− lungs was ~35 % of those in Dmp1^+/+ lungs, consistent with the result of our previous study (Figure 2a).¹⁴ The average Arf expression was downregulated to 14 % of wild-type levels in Dmp1^+/− mice and 5 % in Dmp1^−/− mice (Figure 2b). The expression levels of the possible Dmp1 transcriptional target genes, Areg, Tsp-1, JunB, Egr1 and Adm were downregulated to 10–25 % of wild-type levels in both Dmp1^+/− and Dmp1^−/− lungs, while Bcl-3 and Mbd1 were moderately downregulated to 30–50 % levels of those in wild-type lungs (Figure 2c–i).

We then studied the gene expression levels for Gas1 and Ect2, the two genes showed to be upregulated in lung tissue from Dmp1-deficient mice, using GeneChip Microarray (Table 1). Both of these genes were upregulated three to four times in the lungs of Dmp1^+/− and Dmp1^−/− mice as shown Figure 2j and k. In contrast, p27^kip1 mRNA levels were no different among wild-type, Dmp1^+/− and Dmp1^−/− lungs (Figure 2l). In short, cancer-related, novel Dmp1 target gene expression levels were very similar between Dmp1^+/− and Dmp1^−/− lungs.

Dmp1 directly binds to the mouse Areg, Tsp-1, JunB and Egr1 genomic DNA

We then investigated whether Dmp1 directly binds to the promoters/enhancers of Areg, Tsp-1, JunB and Egr1 since these genes were downregulated more than 80 % in Dmp1-deficient lungs. The mouse Areg promoter and the first intron have unique Dmp1-consensus sequences (#1: 5′-AACGGATGG and #2: 5′-GGCGGATGC [antisense]) (Figure 3a). EMSA conducted with recombinant Dmp1 proteins prepared in Sf9 cells show that the Dmp1 protein directly binds to the sequence #1 (Figure 3b, complex B). This complex was not antagonized by a 200-fold excess of non-specific oligos or with Ets-specific oligos (M3), but was antagonized by the cold probe. Complex B disappeared (RAJ) or was supershifted when the protein-probe mixture was incubated with specific antibody to the carboxyl terminus of Dmp1 (complex S1 with RAF) (Figure 3b). Thus, Dmp1 directly binds to the Areg promoter. Likewise, binding of Dmp1 to the intron 1 of Areg was confirmed (Figure 3c, complex C, supershifted at S2). For Tsp-1, EMSA was conducted with a [α³²P]-labeled probe covering the Dmp1-consensus sequence (5′-TGCGTATGT) located at the 3′ of exon 21 (Figures 3a and 3d). The complex D was specifically antagonized by cold probe and was interfered (RAX and RAJ) or supershifted (S3 with RAF) when incubated with antibodies to Dmp1 (Figure 3d). In JunB, Dmp1 bound to the genomic sequence 5′-GGCGGATGG (JunB#1) (Figures 3a, 3e), but not to the sequence 5′-CTCGGATGG (JunB#2, antisense) located at the 3′ region of the locus (Supplementary figure 1), showing specificity of Dmp1 binding to JunB#1. Direct binding of Dmp1 to the mouse Egr1 promoter was also demonstrated by EMSA with supershift assays (complex F was shifted to S5; Figures 3a, 3f).

Binding of Dmp1 to the endogenous Areg, Tsp-1, JunB and Egr1 genomic loci was verified by chromatin immunoprecipitation (ChIP) with three different antibodies to Dmp1 (Figure 3g). The specificity of Dmp1 binding to the target genes was confirmed by control PCR amplification of the sequence 1.8 kb downstream from the JunB#1 site (Figure 3g, JunB control). Although specific Dmp1 motifs used in EMSA may not account for the ChIP results, our data indicate that both recombinant and endogenous Dmp1 proteins bind to the genomic sequence on the Areg, Tsp-1, JunB and Egr1 promoter/enhancer regions.

To demonstrate that these genes identified by GeneChip Microarray are regulated by Dmp1 at transcriptional levels, we transfected the murine alveolar Type II epithelial cell line C10¹⁹ with a Dmp1 expression vector and studied the expression levels by real-time PCR. Areg, Tsp-1, JunB, Egr1, Bcl-3, Mbd1 and Adm transcripts were significantly upregulated (2–5 fold) (Figure 4a) in Dmp1-expressing cells while Gas1 and Ect2 gene expression was significantly repressed (30–40 %) (Figure 4b). On the other hand, the p27^kip1 gene expression did not change significantly on transient expression of Dmp1 (Figure 4b). These effects were completely abolished by simultaneous treatment of transfected cells with 2μg/mL Actinomycin D, a potent inhibitor of de novo mRNA synthesis (Figure 4a, b). In order to confirm these results, we conducted quantitative primary transcript real-time PCR (qPT-PCR) analysis on Areg, Tsp1 and Egr1 genes that consists of 2 or more exons (Figure 4c). This assay measures the relative abundance of primary, un-spliced transcripts relative to those of mature RNA. Using Areg exon 4 - intron 4 and exon 4 - exon 5, Tsp-1 exon 7- intron 7 and exon 7 - exon 8, Egr1 exon 1 - intron 1 and exon 1- exon 2 primer pairs for qPT-PCR, the relative abundance of immature (pre) and mature Areg, Tsp-1 and Egr1 transcripts were similarly elevated at 24 and 48 hours in C10 cells transfected with the Dmp1 expression vector (Figure 4c). These results indicate that regulation of these target genes by Dmp1 occurs at transcriptional levels.

Areg, Tsp-1, JunB and Egr1 proteins are barely detectable in the lungs from Dmp1^+/− and Dmp1^−/− mice

To verify the results of the GeneChip Microarray and real-time PCR at protein levels, Western blotting was conducted using the lung lysates from the three different Dmp1 genotypes. The levels of Areg, Tsp-1, JunB and Egr1 proteins were significantly downregulated in the lungs from Dmp1^−/− and Dmp1^+/− mice in comparison to wild-type tissues (Figure 5a). Conversely, the Ect2 protein was upregulated in the lungs from Dmp1^−/− and Dmp1^+/− mice (Figure 5a). Since detection of the Dmp1 protein in normal mouse lung was difficult by direct Western blotting, we performed immunoprecipitation - Western blotting using 250μg of cell lysates (Figure 5b). Interestingly, the expression of the 130 and 160 kDa bands for Dmp1 was significantly downregulated in the lungs of Dmp1^+/− mice (10–20 % of wild-type levels) (Figure 5b, lanes 3–6). No specific bands were detectable with lysates from Dmp1^−/− mice (Figure 5b, lanes 1, 2) or wild-type lung lysates immunoprecipitated with control IgG (lane 7). All these results are consistent with the findings of mRNA analysis by real-time PCR and explain the mechanism of haploid insufficiency of Dmp1.

Figure 5.

Western blotting analysis of Dmp1, Areg, Tsp-1, JunB, Egr1, Ect2 and Actin proteins in the lungs from Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice. (a) Areg, Tsp-1, JunB and Egr1 proteins were downregulated in the lungs from Dmp1^+/− and Dmp1^−/− mice while the Ect2 protein was upregulated. (b) Immunoprecipitation (IP) - Western blotting analysis of Dmp1 expression in the lungs from Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice. The Dmp1 protein was not detectable in the lungs from Dmp1^−/− mice and the expression of both 160 and 130 kDa Dmp1 proteins in the lungs from Dmp1^+/− mice was decreased to 10–20 % of levels of wild-type tissue. C: Control immunoprecipitation of the wild-type lung lysate with rabbit IgG.

To confirm the results of Western blotting and demonstrate the changes of Dmp1 and its target protein expression at a cellular level, immunohistochemical studies were conducted in the lungs from Dmp1-knockout mice with specific antibodies. We previously reported that murine lung tissue showed positive staining with antibody to Dmp1 (RAX) in the nuclei of alveolar macrophages and type II cells.⁹ We found significant downregulation of bronchial and alveolar Dmp1 signals in the lungs from Dmp1^+/− mice in comparison to the wild-type lung, reflecting its mRNA levels (Supplementary figure 2a–c). Next, the lungs from mice of three different Dmp1 genotypes were stained with antibodies to Areg, Tsp-1, JunB and Egr1 to confirm the downregulation of the target gene products in the lungs of Dmp1^+/− and Dmp1^−/− mice (Figure 6a–i; Supplemental figure 2d–f). Areg was expressed in alveolar macrophages, type II cells, and bronchial epithelial cells of a wild-type mouse (Figure 6a). The Areg signals were barely detectable in the lungs from mice with a Dmp1^+/− or Dmp1^−/− background (Figure 6b, c). This outcome is consistent with the results of Western blotting (Figures 5) and recapitulates the downregulation of Areg mRNA by real-time PCR (Figures 1, 2). The Tsp-1 staining was expressed in the bronchial epithelial cells and alveolar type II cells (Figure 6d). The cytoplasmic signals were significantly downregulated in the lungs from Dmp1^+/− and Dmp1^−/− mice (Figure 6e, f). Likewise, JunB was expressed in the nuclei of alveolar macrophages and alveolar type II cells (Figure 6g). These signals were barely detectable in the lung from Dmp1^+/− or Dmp1^−/− mice (Figure 6h, i).

Figure 6.

Detection of Areg, Tsp-1, JunB, Ect2 and Actin proteins in wild-type, Dmp1^+/− and Dmp1^−/− lungs by immunohistochemistry. Magnification, ×40 in (a)–(i), ×20 in (j)–(o). (a) Areg is mainly stained in the cytoplasm of alveolar type II cells. The inset shows the expression of Areg in bronchial epithelial cells (×100). (b), (c) The Areg signal was significantly downregulated in both Dmp1^+/− and Dmp1^−/− lungs. (d) Tsp-1 staining is localized to the cytoplasm of normal bronchial epithelial cells and alveolar cells. The insets show the expression of Tsp-1 in bronchial epithelial cells (×100). (e), (f) The Tsp-1 signal was significantly downregulated in the lungs of Dmp1^+/− and Dmp1^−/− mice. (g), (h), (i) The JunB signal was found in the nuclei of alveolar type I and type II cells and macrophages in wild-type mice, but not in the lungs of Dmp1^+/−, Dmp1^−/− mice. (j), (k), (l) Upregulation of the Ect2 protein from Dmp1^+/− and Dmp1^−/− mice. The insets show the signals from bronchial epithelial cells (×100). (m), (n), (o) Staining with the β-Actin antibody in the lungs of Dmp1^+/+, Dmp1^+/− and Dmp1^−/− mice showing no significant difference among the three genotypes. The lung tissues were stained with alkaline phosphatase for panels (a)–(f) and (j)–(l); peroxidase for panels (g)–(i) and (m)–(o).

We also confirmed upregulation of the Ect2 protein in the bronchial epithelium from Dmp1^+/−and Dmp1^−/− mice in comparison to Dmp1^+/+ mice (Figure 6j–l). Upregulation of the Ect2 protein was also found in K-ras^LA lung tumors from Dmp1^−/− and Dmp1^+/− mice¹⁴ (Supplementary figure 3), suggesting that Ect2 overexpression may play critical roles in lung cancer development in Dmp1-deficient animals. In contrast, all of Dmp1^+/+, Dmp1^+/− and Dmp1^−/− lungs showed equal signals with p27^kip1 antibody (Supplementary figure 2g–i) and cytoplasmic staining with Actin antibody (Figure 6m–o). Thus, our Western blotting and immunohistochemical studies are consistent with the results of GeneChip Microarray and real-time PCR for novel Dmp1 target genes.

Discussion

Dmp1 demonstrates its activity as a tumor suppressor through the activation of the Arf-Mdm2-p53 pathway. In this study, we have identified a number of novel target genes for Dmp1 that are involved in cell proliferation/differentiation, signal transduction/transcription, angiogenesis/metastasis, or apoptosis through GeneChip Microarray of mouse lung tissues. The results were verified by real-time PCR, Western blotting, immunohistochemistry and by expressing Dmp1 in non-transformed alveolar epithelial cells. Inhibition of novel mRNA synthesis nullified the effects of Dmp1 in lung epithelial cells, demonstrating that the gene regulation by Dmp1 occurred at transcriptional levels. Direct binding of Dmp1 to the genomic loci of Areg, Tsp-1, JunB and Egr1 indicates that these are direct transcriptional targets for Dmp1.

Among the identified genes, we chose to focus our attention on a small subset due to their involvement in cell proliferation and differentiation. For example, Areg, one of the ligands of the EGF family,^21–23 promotes the growth of carcinoma cells, keratinocytes and normal lung epithelial cells in the presence of extracellular matrix,²² but inhibits the growth of some transformed cell lines and lung epithelial cells in the absence of extracellular matrix.^22,23 Thus, Areg can either stimulate or inhibit the growth of lung cancer cells, dependent on the biological settings. It has recently been reported that brain metastasis of NSCLC is associated with higher expression of Areg protein in tumor cells.²⁴ It was also shown that high circulating Areg is linked to progressive disease of NSCLC.²⁵ Therefore, Areg is considered to be a growth promoting factor in human NSCLC. If NSCLC that shows LOH for hDMP1 is associated with low Areg expression, the patients may have a less aggressive disease and thus will have a favorable prognosis. This prediction is consistent with our recent study where we showed mutually exclusive inactivation of hDMP1 and INK4a/ARF or p53 in NSCLC since the latter two groups are associated with poor prognosis of patients.³ Further studies with clinical samples will be required to demonstrate the correlation between deletion of hDMP1, expression of Areg, and survival of patients in human NSCLC.

Our microarray data indicate that Tsp-1 is significantly downregulated in the Dmp1^−/− lung (to 10–15 %). Tsp-1 in the alveolar epithelial cells and macrophages were barely detectable in the lungs from Dmp1^−/− mice in immunohistochemical studies, and our EMSA studies demonstrate the direct binding of Dmp1 to the genomic locus of Tsp-1, indicating that Dmp1 is a critical regulator of Tsp-1. Tsp-1 is an endogenous inhibitor of angiogenesis, which limits blood vessel density in normal tissues and restrains tumor growth.^26–28 Recent studies demonstrated that in the absence of Tsp-1, proangiogenic signaling is enhanced, possibly through up-regulation of fibronectin expression. The enhanced signaling further promotes endothelial cell proliferation, migration and survival.^27,28 Since Tsp-1 levels are very low in both Dmp1^+/− and Dmp1^−/− lungs, it is expected that lung cancers from Dmp1-knockout mice will exhibit a more invasive and metastatic phenotypes than those from Dmp1^+/+ mice. Indeed, we recently found that K-ras^LA-induced lung tumors from Dmp1^+/− or Dmp1^−/− mice often showed signs of intrabronchial and intravascular invasion in comparison to the Dmp1^+/+ lung tumors of the same age.¹⁴ Thus, Dmp1 prevents tumor formation by simultaneously activating the Arf-p53 pathway and by preventing angiogenesis by upregulating Tsp-1.

Our results show significant downregulation of JunB in the lung of Dmp1^−/− mice versus wild-type. These results were confirmed by real-time PCR and immunohistochemistry. We previously reported that the Dmp1 promoter was activated by oncogenic Ras and activation of the Dmp1 promoter was strikingly impaired in c-Jun or JunB-knockdown cells.¹⁰ Thus, Jun proteins play critical roles in Dmp1 promoter activation by oncogenic Ras. It is generally considered that JunB is a negative regulator of proliferation.^29,30 JunB prevents G1-S phase transition by simultaneously activating p16^INK4a and repressing cyclin D1 transcription.³⁰ Although Dmp1 has been shown to be a physiological regulator of p19^Arf, it may regulate the transcription of p16^Ink4a as well indirectly through the upregulation of JunB. Indeed, we noticed that the p16^Ink4a mRNA was significantly downregulated (to 18 %) in the lungs from Dmp1^−/− mice in comparison to wild-type littermates (data not shown). Thus, JunB is located both upstream and downstream of the Dmp1 signaling, collectively contributing to tumor suppression.

Another intriguing Dmp1 target gene identified in this study is Egr1. Egr1 is a zinc-finger transcription factor and activates transcription as a monomer. It has been reported that Egr1 shows tumor suppressor activities through induction of TGFβ1 and direct induction of p53 and PTEN³¹. In human NSCLC, low levels of Egr1 expressions are associated with poor outcome and resistance to therapy³². Since Egr1 is a direct transcriptional target of Dmp1, it is very likely that Dmp1 deletion leads to lung cancer development by lowering Egr1 levels. Of note, Tschan et al. recently showed that the WT1, which has significant structural homology with Egr1, repressed the hDMP1 promoter through direct binding to the Egr1/Sp1 site.¹² Thus, it is possible that Egr1, like JunB, is located both upstream and downstream of the Dmp1 signaling, both contributing to tumor suppression.

The current DNA microarray and quantitative RT-PCR analyses not only revealed the down-regulation of Areg, Tsp-1, JunB, Egr1, Adm^33–35, Bcl-3^36–38 and Mbd1^39–41, but also showed significant upregulation of growth arrest-specific 1 (Gas1, 5.40 in the lung)⁴², Ect2^43–45 and other transcripts in Dmp1^−/− mice. Chalamalasetty et al. showed that nucleotide-exchange factor Ect2 localized to the central spindle and to the cell cortex in mitotic cells.⁴⁴ Depletion of Ect2 impaired cleavage furrow formation and RhoA and Citron kinase failed to accumulate at the cleavage furrow, suggesting that Ect 2 is essential for cytokinesis.⁴⁴ Hirata et al. conducted tumor tissue microarray and showed that a high level of ECT2 expression was associated with poor prognosis for patients with NSCLC.⁴⁵ They also showed that knocking down of ECT2 expression by small interfering RNAs effectively suppressed lung cancer cell growth, suggesting specific role of ECT2 in lung cancer development.⁴⁵ Interestingly, Ect2 protein was significantly upregulated in the lungs from Dmp1^−/−; K-ras^LA mice, suggesting that increased Ect2 expression caused by Dmp1 loss contributes lung tumor development. Since LOH of hDMP1 is found in ~35 % of human NSCLC¹⁴ and Dmp1 is a negative regulator for Ect2, it will be imperative to study the correlation of hDMP1 and ECT2 expression in human NSCLC samples and determine the prognostic value of hDMP1-loss in human NSCLC.³

Dmp1 has been shown to be haplo-insufficient for tumor suppression.^2,3,13,14 It has been reported that loss of one p53 allele results in a four-fold reduction of p53 mRNA and protein levels, which could explain the mechanisms of haplo-insufficiency of p53 in tumor suppression.⁴⁶ Interestingly, our study also showed that the Dmp1 mRNA levels in Dmp1^+/− lungs were ~35 % of those in Dmp1^+/+ lungs; the Dmp1 protein levels were even lower (10–20 % of wild-type) in the lungs from Dmp1^+/− mice. Consistently, all the nine cancer-related Dmp1 target genes (including p19^Arf) that were downregulated in Dmp1^−/− lungs were also repressed in Dmp1^+/− lungs. Likewise, both Gas1 and Ect2 genes that were upregulated in Dmp1^−/− lungs were highly expressed in Dmp1^+/− tissues, indicating that cancer-related gene expression is very close between Dmp1^+/− and Dmp1^−/− lungs. These data were further confirmed by immunohistochemistry with specific antibodies. EMSA showed direct binding of Dmp1 to Areg, Tsp-1, JunB and Egr1, and overexpression of Dmp1 induced either upregulation or downregulation of novel Dmp1 target genes in non-transformed alveolar epithelial cells. Although the detailed molecular mechanisms for unexpectedly low Dmp1 protein expression in Dmp1^+/− lungs remain to be determined, this study not only identified novel genes that are regulated by Dmp1, but also gave insights into the mechanisms of haploid insufficiency of Dmp1 in tumor suppression.

In summary, we performed a rigorous genome-wide examination of genes with expressional changes specifically associated with Dmp1. We found numerous genes that regulate cell growth and are downstream targets of the Dmp1 transcription factor. Interestingly, many of these are involved in the signaling pathways that regulate Dmp1. We also identified Tsp-1 as a novel target for Dmp1, which plays critical roles in the prevention of tumor angiogenesis and metastasis. Although transcriptional repressor function of Dmp1 has never been reported, significant subsets of genes were upregulated in Dmp1^−/− lung in this study. The current findings are highly worthy of further investigation to elucidate the function of Dmp1, which is located at the cusp of oncogenic and tumor suppressor signal transduction.