EGR1 Target Genes in Prostate Carcinoma Cells Identified by Microarray Analysis*

The EGR1 transactivator is overexpressed in prostate cancer, and its expression pattern suggests that EGR1 could potentially regulate a number of steps involved in initiation and progression of prostate cancer, such as mitogenesis, invasiveness, angiogenesis, and metastasis. To identify potential EGR1 target genes in an unbiased manner, we have utilized adenovirus-mediated expression of EGR1 in a prostate cancer cell line to identify specific genes that are induced by EGR1. Using oligonucleotide arrays, a number of EGR1-regulated genes were identified and their regulation was confirmed by quantitative reverse transcription-polymerase chain reaction analysis. One of the largest gene classes identified in this screen includes several neuroendocrine-associ-ated genes (neuron-specific enolase, neurogranin), suggesting that EGR1 overexpression may contribute to the neuroendocrine differentiation that often accompanies prostate cancer progression. This screen also identified several growth factors such as insulin-like growth fac-tor-II, platelet-derived growth factor-A, and transforming growth factor- b 1, which have previously been implicated in enhancing tumor progression. The insulin-like growth factor-II gene lies within the 11p15.5 chromosomal locus, which contains a number of other imprinted genes, and EGR1

ferentially expressed genes in cancer, as they are uniquely poised to coordinately regulate gene networks. For example, the protooncogene c-myc can be activated through gene amplification or dysregulation of tumor suppressor pathways, and is able to promote proliferation by activating the expression of many genes involved in cell cycle progression (1,2). To understand the physiological significance of this overexpression, it is imperative to identify target genes that become activated in response to such overexpression. Until recently, identifying target genes of specific transactivators has been impeded by the scarcity of promoter sequence data. However, even comprehensive genome sequence is insufficient to unambiguously identify target promoters, since many transactivators bind to cognate sequences that deviate considerably from consensus binding sites. Techniques such as DNA footprinting and reporter assays have been extremely useful in the analysis of suspected target promoters, but these techniques do not provide proof that a transactivator activates a given promoter in vivo. In addition, these methods are time-consuming and are not suited for efficient identification of novel target genes.
Recent studies have demonstrated that at least two transcription factors, ETS2 and EGR1, are overexpressed in prostate cancer (3)(4)(5). The EGR1 transactivator was originally identified as an immediate-early gene that is rapidly induced in response to a variety of stimuli. More recently, several studies have focused attention on the role of EGR1 in coordinating responses to hypoxia and vascular injury. In these systems, EGR1 activates expression of tissue factor (which eventually triggers vascular fibrin deposition) and several growth factors such as PDGF-A, 1 PDGF-B, TGF-␤1, IGF-II, and bFGF (6,(7)(8)(9). Interestingly, many of these same factors have also been implicated in various stages of prostate tumor progression (e.g. angiogenesis, metastasis), adding further evidence that at least some steps of tumor progression are mechanistically related to wound healing and hypoxic responses (10). Although increased expression of several of these genes has been implicated in development and progression of prostate cancer (11,12), it has not yet been established whether their up-regulation in this context is functionally linked to increased levels of transactivators such as EGR1.
To determine the physiological significance of EGR1 overexpression in prostate cancer, we have developed a high-throughput screen for genes that are induced by EGR1. This strategy employs a recombinant adenovirus that expresses EGR1 in the LAPC4 prostate cancer cell line. Changes in gene expression are then analyzed using microarray technology, which has made it possible to simultaneously track changes in expression levels of thousands of genes. Finally, expression of the candidate target genes in primary prostate tumor specimens is rapidly determined using quantitative RT-PCR analysis. Compared with reporter-based assays, a major advantage of this approach is that it measures the response of endogenous promoters in their native chromatin context. Using this strategy, we have identified several genes that are regulated by EGR1 overexpression. These include signaling proteins, transcription regulators, neuroendocrine proteins, and membrane-associated proteins involved in adhesion and signaling. These results not only illuminate the consequences of EGR1 overexpression in prostate cancer, but also provide a model for identifying target genes of specific transactivators in other types of cancer.

EXPERIMENTAL PROCEDURES
Adenovirus Infection-LAPC4 human prostate carcinoma cells (13) (kindly provided by C. Sawyers, UCLA, Los Angeles, CA), were maintained in Iscove's growth medium supplemented with 10% fetal bovine serum. Adenoviral recombinants were prepared essentially as described (14). EGR1 I293F (15) was subcloned into the pAC adenoviral transfer plasmid and inserted by homologous recombination into the E1 region of adenovirus Ad5PacIGFP (14,16). As a negative control, we used an adenovirus (Ad5PacIGFP) expressing the Gal4 DNA-binding domain (1-147) fused to a mutant, nonfunctional form of the EGR1 R1 domain (EGR1 residues 269 -304 with I293F mutation). LAPC4 cells were infected at a viral titer of 1 ϫ 10 8 plaque-forming units/ml for 2 h. Thereafter, cells were washed once with medium and then cultured for another 24 h. Examination of the cells for GFP expression revealed that each virus infected Ͼ90% of the cells. For the immunoblot analysis of Fig. 1, lysates from adenovirus-infected cells were harvested 24 h after infection, resolved on a 10% polyacrylamide gel, and blotted onto nitrocellulose. Culture and stimulation of PC12 cells with NGF was performed as described previously, and the blot was probed with the 6H10 anti-EGR1 monoclonal antibody (17).
Oligonucleotide Microarray Analysis-Hybridization probes for Ge-neChip analysis were prepared from poly(A) ϩ RNA prepared from cultures of LAPC4 cells that had been infected with either adenovirus expressing EGR1 (I293F) or the control adenovirus. The poly(A) ϩ RNA was converted to double-stranded cDNA using an oligo(dT) primer containing the T7 promoter, and this was used to prepare biotinylated cRNA using the Bioarray HighYield kit (Enzo) according to the manufacturer's directions. The biotinylated cRNA probes were fragmented and applied as described (18,19) to individual oligonucleotide Hu-GeneFL GeneChip arrays (Affymetrix), which contain probe sets for 5600 human genes. The signal intensities from hybridized cRNA were quantified, and the GeneChip analysis software was used to identify differentially expressed genes.
Quantitative RT-PCR (TaqMan) Analysis-Total RNA was purified, and 1 g was used to prepare cDNA (20). Quantitative RT-PCR was performed by monitoring in real time the increase in fluorescence of the SYBR Green dye as described (21,22) using the TaqMan 7700 sequence detection system (PerkinElmer Life Sciences). For comparison of transcript levels between samples, a standard curve of cycle thresholds for several serial dilutions of a cDNA sample was established and then used to calculate the relative abundance of each gene. Values were then normalized to the relative amounts of glyceraldehyde-3-phosphate dehydrogenase cDNA, which were obtained from a similar standard curve. All PCR reactions were performed in duplicate. Sequences of primers used for PCR analysis are available upon request.
Tissue Specimens-Prostate tissue specimens, derived from radical prostatectomy, were obtained from the Alvin J. Siteman Cancer Center Tissue Procurement Core Facility at Washington University. All samples had a Gleason grade of 3 ϩ 3 (23). Guided by hematoxylin and eosin-stained frozen sections, the tissue blocks were grossly dissected so that, by visual estimate, the epithelial component of the isolated tissue contained at least 75% carcinoma cells. The tissues were sectioned at 50 m on the cryostat microtome and used for RNA isolation. A serial frozen section was stained to verify that the tissue sections used for RNA preparation were predominantly carcinoma. Quantitative RT-PCR analysis was performed as described above, with the exception that 18 S rRNA was used to normalize for the amount of input cDNA.
Analysis of IGF-II Imprinting-Detection of IGF-II imprinting in human samples was performed as described (24 -26). Briefly, genomic DNA from LAPC4 cells was isolated and amplified using two primers that span an IGF-II gene segment that contains a single nucleotide polymorphism. Sequencing of both strands of the PCR fragment revealed that LAPC4 DNA is heterozygous for this polymorphism. To determine which allele is induced by EGR1 expression, cDNA from LAPC4 cells infected with adenovirus expressing EGR1 I293F was amplified with the same primers, and this product was sequenced.

Activation of Endogenous EGR1 Target Genes-Previous
work has shown that the EGR1 transactivator is overexpressed in a majority of prostate cancers (4,5). To identify genes that are regulated by EGR1, we wished to achieve overexpression of EGR1 in a prostate cell line without using stimuli (e.g. growth factors) that would activate signaling pathways and induce other transcription factors. Therefore, we utilized a recombinant adenovirus that expresses EGR1 (I293F), a mutant that is resistant to repression by endogenous NAB transcriptional corepressors (27,28), which could repress any activation by wild type EGR1. The recombinant adenovirus was used to infect the LAPC4 prostate cell line. This cell line was derived as an explant of metastatic prostate cancer, and retains many of the characteristics of normal prostate cells such as prostatespecific antigen expression and androgen dependence (13). In addition to EGR1 (I293F), the adenovirus also expresses GFP from an independent transcription unit, which allows monitoring of infection. After infection with EGR1-expressing adenovirus, visualization by fluorescence microscopy confirmed that essentially all (Ͼ95%) of the cells in culture were infected.
To determine if the level of EGR1 expression in adenovirusinfected cells is significantly higher than is ever observed physiologically, the expression level of EGR1 (I293F) created by the recombinant adenovirus (Fig. 1, lane 4 of inset) was compared The lower band in lane 2 is a previously described proteolytic product of EGR1 found in NGF-stimulated PC12 cells (17). Equal amounts of protein lysates were loaded in each lane. B, the expression levels of the indicated genes were determined by quantitative RT-PCR analysis of cDNA samples obtained from LAPC4 cells that had been infected with adenovirus expressing EGR1 I293F. Expression levels for each gene were normalized to the level of glyceraldehyde-3-phosphate dehydrogenase expression, and then normalized to the level found in LAPC4 cells infected with a control adenovirus (set as 1). All reactions were performed in duplicate, and the standard error is indicated.
with the induced level of EGR1 in NGF-stimulated PC12 cells (lane 2) (29). This immunoblot reveals that use of recombinant adenovirus is an efficient means to target EGR1 overexpression to a cell line, and that the resulting level of EGR1 expression is comparable to the level of induced EGR1 observed in the NGFtreated PC12 system. The endogenous expression level of EGR1 in cultured LAPC4 cells is relatively low, at a level that is comparable to that observed in normal prostate tissue (data not shown).
To test whether EGR1 (I293F) could activate potential target genes in this cell line, we measured expression levels of three EGR1 target genes that have been identified in other systems: IGF-II, PDGF-A, and TGF-␤1 (30 -34). After 24 h of infection, RNA was purified from these cultures and used to generate cDNA. We employed a technique for quantitative RT-PCR analysis, in which a fluorescent dye (SYBR Green) that binds double-stranded DNA is used to quantitate the amount of amplicon as it accumulates during the PCR reaction (22,(35)(36)(37). For each cDNA sample, we measured the cycle number at which PCR product accumulation reaches a defined threshold. Then, the relative levels of gene expression were determined using a standard curve obtained from assays of serial dilutions of a cDNA sample containing the gene of interest. To control for genes induced by adenovirus infection alone, expression levels of these three genes in EGR1-infected LAPC4 cells were compared with those obtained in LAPC4 cells infected with a control adenovirus. As shown in Fig. 1B, IGF-II, TGF-␤1, and PDGF-A were all induced in response to EGR1 expression, indicating for the first time that these genes, in their endogenous loci, are induced by EGR1 in a prostate cell type.
Microarray Analysis of LAPC4 Cells-The same cDNA samples used in Fig. 1 were used to prepare biotinylated cRNA targets, which were then hybridized to individual oligonucleotide HuGeneFL GeneChip arrays (Affymetrix), which contain probe sets for 5600 human genes. The signal intensities from hybridized cRNA were quantified as described (18,19). Using the default parameters of the GeneChip analysis software, 37% and 33% of the genes on the HuGeneFL array were scored as being present (or marginal) in the control-infected and EGR1 (I293F) expressing LAPC4 cells, respectively. The normaliza-tion factor used to compare the two data sets indicated that the global levels of hybridization from the two cRNA samples were roughly equivalent. Using defined copy numbers of synthetic, biotinylated cRNA transcripts that were added to the hybridization mixture, we estimate that the detection threshold in this experiment was approximately 5-10 copies/cell.
The GeneChip data was first used to identify genes that are abundantly expressed in this cell line, since this profile may provide information regarding potential diagnostic markers and therapeutic targets in prostate cancer. A list of the 50 most highly expressed genes (excluding genes for ribosomal proteins) in Table I contains several genes that have previously been associated with various types of tumors. For example, the thymosin ␤-10 protein binds and sequesters G-actin and is overexpressed in a wide range of tumor types (38). In addition, CD81, a cell-surface molecule involved in cell adhesion and integrin signaling (39), is also expressed at a high level. Neuroleukin is a multifunctional protein that is a phosphoglucose isomerase, but also functions as a tumor-secreted cytokine that regulates invasion and metastasis (40). The significance of the high expression levels of these proteins in the LAPC4 cell line remains to be established, but similar analyses of other prostate cancer models may substantiate their overexpression as a general characteristic of prostate cancer.
Identification of EGR1 Target Genes by Microarray Hybridization-Because we had observed up-regulation of several potential EGR1 target genes ( Fig. 1) after expression of EGR1 (I293F), we used the GeneChip data sets to identify other genes that become induced as a consequence of EGR1 (I293F) expression. The average hybridization intensity across all probe sets using the cRNA prepared from LAPC4 cells expressing EGR1 I293F was normalized to that obtained from LAPC4 cells infected with a control virus. Comparison of the two data sets revealed that 144 of the genes found to be "present" in both samples (2.1% of total genes represented on the array) were induced in LAPC4 cells expressing EGR1 (I293F), but only 30 of these genes (0.5%) were induced more than 3-fold.
Analysis of the results indicated several genes that were significantly altered in response to EGR1 overexpression (Table II), most of which had not previously been identified as EGR1 target genes. Many of the induced genes could be grouped into functional classes of molecules. These include transcriptional regulators, signaling molecules, as well as some neuroendocrine proteins. One signaling molecule was the Rad gene, a Ras homolog that was originally identified to be overexpressed in the muscle of patients with type II diabetes (41). More recently, Rad expression has been shown to potentiate serum-stimulated DNA synthesis in a melanoma cell line. In addition, this activity of Rad is inhibited by the nm23 gene product, a putative suppressor of tumor metastasis (42). An example of the transcription factor group is CBF-␤, the non-DNA-binding subunit of the heterodimeric transcription factor core-binding factor (CBF)/polyoma enhancer-binding protein 2. Chromosomal translocations involving the human CBF-␤ gene (CBF-␤-MYH11) are associated with a large percentage of human leukemias (43). Recently, CBF activity has been shown to be required for angiogenesis in an endothelial cell line, where expression of all CBF subunits is induced by angiogenic factors, such as bFGF and vascular endothelial growth factor (44). The neuroendocrine genes are particularly interesting since neuroendocrine differentiation is often observed during prostate cancer progression. Neuron-specific enolase is a widely used marker for determining the extent of neuroendocrine differentiation (45,46). Validation of GeneChip Results Using Quantitative RT-PCR Analysis-To independently measure the fold induction of specific EGR1 target genes, quantitative RT-PCR analysis was used to measure expression of some of the genes identified by GeneChip analysis. The induction by EGR1 (I293F) of the genes chosen for this analysis spanned a range from 2-to 50-fold. Fold induction calculated from the GeneChip data was compared with that obtained using quantitative RT-PCR analysis (Table III). Induction of specific genes was confirmed for the most part by our quantitative RT-PCR analysis. However, for some genes, the actual fold induction by quantitative RT-PCR (e.g. protease M, Rad, IGF-II, and TGF-␤1) was significantly greater than that derived from the GeneChip analysis. In these cases, the expression level in the control RNA sample appeared to be below the threshold of detection by GeneChip analysis. In such samples (marked with Ͼ), the software employed the background noise (computed by Affymetrix software) to calculate fold induction.
Since the GeneChip detection system significantly underestimated the fold induction of several genes (e.g. IGF-II and TGF-␤1), the data were sorted to identify genes that were absent in the base-line chip (control infected LAPC4 cells) and then emerged above the detection threshold in the sample obtained from EGR1-expressing LAPC4 cells. Testing several of these genes revealed others that were significantly induced beyond the level indicated by the GeneChip analysis. For example, the EF1-␣2 and telencephalin/ICAM-5 genes were induced 3.7-and 3-fold, respectively, in the GeneChip data set, but subsequent quantitative RT-PCR analysis demonstrated inductions of 9.4-and 13.3-fold, respectively.
The identification of several known EGR1 target genes suggests that most of these genes are probably activated directly by EGR1. However, this analysis cannot exclude the possibility that activation occurs indirectly through activation of one or more intermediary molecules. For some target genes, such as TGF-␤1, PDGF-A, and IGF-II, in vitro assays have identified one or more EGR1 binding sites in these promoters that appear to mediate EGR1 activation (30 -34). However, the promoters of most of the other genes have not been studied in any great detail. The promoters of these genes were screened for sequences that conform to the EGR1 consensus-binding site defined previously (47). The presence of such sites in many of the promoter regions (i.e. within 1000 base pairs upstream of the transcription start site) suggests that activation by EGR1 is direct (Table III).

TABLE III Validation of EGR1 target genes
The induction of several EGR1 target genes identified in the Gene-Chip screen was independently tested using quantitative RT-PCR analysis. The -fold induction was computed using a standard curve analysis and normalized to the level of GAPDH as described under "Experimental Procedures." All PCR reactions were performed in duplicate, and the average -fold induction is shown. The Ͼ symbol in the GeneChip column indicates that the expression level of the gene in the control sample was below the detection threshold determined by Affymetrix software. In the Taqman column, the Ͼ symbol indicates that the uninduced level falls below the standard curve derived from the level found in the induced sample, meaning that the induction is at least 125-fold. When available, the promoter sequences for the selected genes were scanned for EGR1 binding sites that lie within 1000 base pairs 5Ј of the transcription start site, and the number of these sites are indicated in the last column. NA indicates that promoter sequence is not currently available. cer-To extend our analysis to prostate tumors, we examined the expression of a several EGR1-regulated genes identified above in tumor specimens obtained after prostatectomy. Because the infiltrative nature of most prostate carcinomas makes it difficult to obtain homogeneous samples of prostate cancer, the frozen tissue blocks were examined to select samples in which the epithelial component consisted of Ͼ75% carcinoma. RNA was prepared from these samples, and EGR1 expression was determined by quantitative RT-PCR analysis. For subsequent analysis of EGR1 target genes, we chose four samples with high EGR1 levels and three samples with low EGR1 expression, similar to that found in normal prostate. The increased expression of EGR1 in these samples should be considered to be a minimum estimate since the isolated carcinoma samples also contain some stromal cells that do not express EGR1. We found that the relative expression levels of a number of EGR1 target genes (IGF-II, NSE, Rad, Id4, and EF-1␣) correlated well with that of EGR1 in many of the tumors (Fig. 2). The carcinomas with high levels of EGR1 generally expressed high levels of these target genes, whereas tumors with low EGR1 expressed low levels of the target genes. Some EGR1 target genes (neurogranin, protease M, IPL, and telencephalin) were expressed at very low levels in all of the prostate cancer samples tested (data not shown), suggesting that these genes may not be relevant to prostate tumorigenesis. Overall, these results support the idea that EGR1 overexpression in prostate tumors activates expression of downstream target genes that may influence tumor growth.
EGR1 Activation of Genes in the Imprinted 11p15 Locus-Analysis of the genes induced by EGR1 expression in LAPC4 cells surprisingly revealed that many potential EGR1 target genes lie within the 11p15 chromosomal locus. This locus contains a cluster of imprinted genes, and loss of imprinting for several of these genes has been associated with a variety of adult and childhood cancers (48,49). The genes up-regulated by EGR1 in this locus are IGF-II, IPL/TSSC3, and p57 KIP2 . Elevated IGF-II expression is observed in many types of cancer and is often associated with loss of imprinting in which the normally silent, maternal allele becomes activated (26). Some of the more recently identified imprinted genes in this domain are not represented in the human FL array that we used in our GeneChip experiment. Therefore, we used quantitative RT-PCR to assay expression of the imprinted genes in the 11p15 locus that are not represented on the array (Fig. 3). Not all of the imprinted genes in this locus respond to EGR1 overexpression, but our results do indicate a clustering of several EGR1 target genes in this domain. Of the genes activated by EGR1, IGF-II would appear to be one of the more likely candidates to enhance prostate tumor progression, and we therefore examined activation of IGF-II expression by EGR1 in greater detail.
The human IGF-II gene has four independent promoters, which are differentially regulated in a tissue-and developmental stage-specific manner (50,51). To determine which IGF-II promoter(s) are activated by EGR1 in LAPC4 prostate cells, we used promoter-specific primers (26) to analyze transcription from the four promoters following EGR1 expression. As shown in Fig. 4, the EGR1-mediated activation of IGF-II expression occurs principally through the P3 and P4 promoters, although all four promoters are activated to some extent. This pattern of promoter activation correlates well with the number of EGR1 binding sites (47) that were found in the IGF-II promoters. Within 1000 base pairs upstream of the respective transcription initiation sites, promoters P1, P2, P3, and P4 have 1, 0, 4, and 6 binding sites, respectively.
It has been previously reported that elevated IGF-II expression in prostate cancer is associated with loss of imprinting (25). To determine if the activation of IGF-II expression by EGR1 (I293F) in LAPC4 cells is also biallelic, we took advantage of a single nucleotide polymorphism in the IGF-II 3Јuntranslated region, which has previously been used to examine IGF-II imprinting status (24 -26). Genomic DNA obtained from the LAPC4 line was amplified and found to be heterozygous for this polymorphism (Fig. 5). When we examined the cDNA from EGR1-expressing LAPC4 cells, it was evident that both alleles are being expressed, consistent with results previously obtained in prostate cancer samples (25). DISCUSSION Two groups have recently identified EGR1 as a gene that is overexpressed in prostate cancer (4,5). As we have also confirmed these observations ourselves, 2 these data have prompted our investigation of two questions. 1) Does EGR1 promote prostate cancer initiation and/or progression, and 2) what genes activated by EGR1 participate in its functional role in prostate cancer? For the first question, we have recently mated a mouse model of prostate cancer developed by Garabedian et al. (52) with the targeted disruption of the mouse Egr1 2 S. A. Abdulkadir, J. Svaren, and J. Milbrandt, submitted for publication.
FIG. 2. Expression of EGR1 target genes in prostate carcinoma samples. RNA was prepared from prostate carcinoma samples, and the relative EGR1 expression levels in these samples were determined by quantitative RT-PCR analysis. Four samples with high EGR1 levels and three samples with low EGR1 were then assayed for expression of the indicated genes. For each gene, a standard curve was used to calculate relative expression levels in the samples, which were then normalized to the amount of 18 S rRNA in each sample. The expression level of each gene is indicated relative to the sample containing the highest level of that gene, which was set as 1. All reactions were performed in duplicate, and the standard error is indicated. gene that we had previously generated (53). Our results indicate that activation of Egr1 is an early event in development of prostate cancer in this model, and that the absence of the mouse Egr1 gene results in significantly delayed tumor development as measured by tumor mass and survival rates. 2 These results provide genetic evidence that Egr1 plays an important role in tumor development subsequent to formation of prostatic intraepithelial neoplasia.
Therefore, it is imperative to address the second question (i.e. identification of EGR1 target genes) in order to understand the functional role of EGR1 in prostate cancer. To explore the EGR1 transcriptional network, we have employed adenovirusmediated expression of the EGR1 transactivator followed by microarray hybridization to determine how EGR1 overexpression alone (in the absence of other stimuli) is able to change patterns of gene regulation in the LAPC4 prostate cancer cell line. One class of genes that was identified in our screen is associated with neuroendocrine cells. Neuroendocrine differentiation occurs frequently in prostate cancer, and it is thought that neuroendocrine cells may secrete factors that allow prostate carcinoma to become androgen-independent (45,46,52). Neuron-specific enolase is a common marker used to identify neuroendocrine differentiation in prostate carcinoma. Other such genes identified in this screen include neurogranin, hneuro-d4 (a zinc finger protein related to requiem), and telencephalin/ICAM5. Many of the neuroendocrine genes are also highly expressed in the central nervous system, where EGR1 and related family members (i.e. EGR2/Krox20, EGR3, and EGR4) are present at high levels.
Another class of EGR1-induced genes includes several growth factors (PDGF-A, IGF-II, and TGF-␤1) that respond to EGR1 activity in other systems (30 -32, 34). All of these factors have been implicated in accelerating one or more aspects of tumor progression, such as mitogenesis, angiogenesis, or invasiveness (12, 54 -59). Analysis of mouse models has directly demonstrated that IGF-II is an autocrine factor that can contribute to malignant hyperproliferation (60,61). Furthermore, ribozyme-mediated reduction of IGF-II levels has been shown to inhibit growth of the PC3 prostate cancer cell line (62). It has been proposed that EGR1 and WT1 have opposing effects on IGF-II expression, since the two proteins have similar DNA binding specificity (30,63). Furthermore, decreased WT1 levels are associated with elevated IGF-II levels in benign prostatic hyperplasia (64).
IGF-II overexpression is the most common molecular event observed in Beckwith-Wiedemann syndrome (BWS), which is characterized by prenatal overgrowth phenotypes and predisposition for several childhood tumors (48,49). Transgenic mouse models that overexpress IGF-II recapitulate many of the phenotypes observed in BWS (65,66). A major cause of IGF-II overexpression in BWS appears to be biallelic expression of the IGF-II gene, which is normally maternally imprinted. Loss of imprinting of the IGF-II gene is similarly observed in a wide range of cancers (including prostate) (25,26). The biallelic induction of IGF-II expression leads us to speculate that EGR1 overexpression may help bypass the imprinting mechanism that normally silences the maternal allele, and thereby be at least one factor that contributes to higher levels of IGF-II in prostate cancer.
In examining the genes induced by EGR1 activity, we found that two additional genes within the 11p15.5 locus are induced by EGR1 activity, IPL/TSSC3 and p57 KIP2 . The induction of the p57 KIP2 CDK inhibitor would not seem to be consistent with FIG. 4. Activation of IGF-II in EGR1-overexpressing LAPC4 cells. The diagram indicates the 10 exons and 4 alternate promoters (P1, P2, P3, and P4) that comprise the human IGF-II gene (70). Promoter-specific primers (26) were individually paired with a reverse primer at the junction of exons 8 and 9 (common to all IGF-II transcripts). The products of PCR reactions using cDNA prepared from control-infected (con) and AdEGR1 I293F (labeled EGR1) are shown below. Note that the transcript initiating at promoter P2 has an alternatively spliced exon (exon 5), which gives rise to the two bands in the PCR reaction (71). The promoters were scanned for EGR1 binding sites, and promoters P1, P2, P3, and P4 were found to have 1, 0, 4, and 6 binding sites, respectively.
FIG. 5. EGR1 activates both alleles of IGF-II. Genomic DNA from LAPC4 cells was amplified with primers that encompass a single nucleotide polymorphism in the 3Ј-untranslated region of the IGF-II gene. Sequencing of both strands of the PCR fragment demonstrated that LAPC4 cells were heterozygous for this polymorphism. The same primers were then used to amplify cDNA prepared from cells infected with the adenovirus expressing EGR1 I293F. The sequence traces reveal that both alleles of IGF-II are induced by expression of EGR1. Under the PCR conditions used, there was no observable band using cDNA from control-infected LAPC4 cells. accelerated tumor progression. However, both IGF-II and p57 KIP2 have very similar developmental expression patterns, and recent work has documented a genetic interaction between these genes in a mouse model of Beckwith-Wiedemann syndrome (67). From these results, it was proposed that IGF-II and p57 KIP2 play antagonistic roles in controlling cell proliferation during normal development. Therefore, it is possible that coinduction of these two genes by EGR1 might reflect a normal regulatory loop that is disrupted by increased expression of IGF-II and/or loss of function (mutation) of p57 KIP2 .
One caveat in interpreting the results of GeneChip analysis is that the detection sensitivity of the current technology results in a significant underestimate of -fold induction for certain genes. Although the detection threshold may vary between individual genes because of differential hybridization efficiency, use of labeled control transcripts in our experiment indicate that the detection threshold is approximately 5-10 copies/cell (assuming complete conversion of mRNA to cDNA), a number that is consistent with that obtained in other experiments employing the same technology. This estimate, however, does not include loss of sensitivity associated with conversion of mRNA to cDNA and subsequent in vitro transcription to generate labeled cRNA target. In a typical cell, it has been estimated that most of the expressed genes (ϳ10,000) are expressed at 5-15 copies/cell. Therefore, a large number of these genes would be expressed at (or below) the detection threshold of GeneChip technology. Genes that are expressed in this abundance class (5-15 copies/cell) include many genes that do not need to be expressed at extremely high levels in order to be biologically active, such as growth factors and transcriptional regulators. For example, the expression level of important transcriptional regulatory proteins such as Sp1 and TATA-binding protein is at or below the level of detection in our experiment, as well as similar data sets published by other groups (68,69). Therefore, using a more sensitive method such as quantitative RT-PCR analysis is a complementary strategy that can not only validate GeneChip data, but also help identify significantly induced genes whose transcripts are in the lower abundance classes.
Over the last 15 years, enormous progress has been made in identifying transcriptional regulators and understanding the molecular details of how they modulate gene expression. However, the relatively small number of promoters with which they have been tested in vitro has limited our knowledge of these factors. The recent growth of genome data bases and microarray techniques has now made it possible to explore the physiological significance of these transcriptional regulators by identifying the complement of genes that they regulate. Although these techniques are also well suited for identifying target genes in loss-of-function models (e.g. mouse knockouts), the gain-of function approach that we have described will be particularly helpful in cases where redundancy of related transactivators makes loss-of-function studies impractical.