Identification of Mammalian E2F Regulatory Networks Using DNA Microarray Hybridization Analyses

The E2F family of transcription factors has been intensely studied since the discovery that E2F1 can interact with the retinoblastoma tumor suppressor protein. Candidate gene approaches have led to the identification of E2F targets that play a role in cell cycle regulation, DNA replication, and apoptosis. The advent of new, microarray-based approaches has greatly expanded the base of knowledge about E2F and has revealed that E2F plays a role in far more cellular processes than originally envisioned. Here, we review the recent literature in which microarray technology has been used to gain insight into E2F activity. Introduction The E2F family of transcription factors plays a key role in cell cycle progression and the deregulation of E2F activity is linked to the development of human cancer. For example, certain E2Fs bind to and are repressed by the retinoblastoma (Rb) tumor suppressor protein, which is frequently mutated in several types of cancer. The E2F family (which to date consists of E2Fs 1-7 and the heterodimerization partners DP1 and DP2) has been intensively studied since the identification in 1987 of E2F as a factor that recognizes the viral E1A promoter; for a recent review see Trimarchi et al.3 Initial discoveries included the characterization of a consensus E2F binding site (TTTSSCGC) and the finding that E2Fs bind to and regulate promoters of genes that control processes such as cell cycle progression, DNA replication, and apoptosis. The pace of discovery has accelerated with the recent sequencing of the human genome, the increased use of bioinformatics, and the development of microarray technologies, the combination of which has allowed the identification and analysis of E2F targets in vivo in a high-throughput manner. To date, several E2F microarray analyses have been performed, identifying sets of genes regulated by individual E2Fs. Although one might criticize such studies as “fishing expeditions” that simply provide a large list of genes, carefully designed microarray experiments provide the opportunity to move beyond reductionist science by employing massively parallel analyses to identify metapatterns that can give novel insights into whether a factor regulates a specific process. In particular, such techniques can ultimately lead to the delineation of transcriptional regulatory networks in various normal tissues, assist in the charMicroarrays and Transcription Networks, edited by M. Frances Shannon and Sudha Rao. ©2005 Eurekah.com. Microarrays and Transcription Networks 2 acterization of the deregulation of these networks in pathologies such as human cancers, and spur the development of therapeutics that can revert the regulatory network back to the normal state. Obviously, this is not a trivial goal given that there are estimated to be more than 2000 transcription factors in the human genome. However, large steps have been taken in the laboratory of R. Young who has begun to identify most of the transcriptional regulatory networks in yeast using a combination of gene expression analysis, chromatin immunoprecipitation coupled to microarray analysis (ChIP-chip), and bioinformatic algorithms. Although such comprehensive studies have not yet been performed in mammalian cells, investigators have taken two types of microarray-based approaches to high-throughput experimental identification of E2F targets: mRNA expression analysis following modulation of E2F activity and direct analysis of E2F binding by coupling ChIP with genomic microarray analysis. In this review, we summarize how these studies have provided novel insights concerning biological processes in which E2F participates and propose future experiments designed to more fully understand the complex regulatory networks in which E2F functions. Identification of E2F Target Genes Using Oligonucleotide-Based Microarrays E2F was initially identified as a cellular factor that regulates transcription from the adenoviral E2 promoter.6 The adenoviral system provided some knowledge about the function of E2F and allowed the sequence of a high affinity E2F binding site to be identified. However, E2F’s role in regulating transcription of cellular genes continued to be unclear for several years. The first cellular gene determined to be regulated by the E2F family was dihydrofolate reductase (dhfr), a gene required for de novo purine biosynthesis (and thus important for the production of nucleotides required for DNA replication). The dhfr gene displays cell cycle-specific transcriptional regulation, with the highest promoter activity occurring at the G1/S phase boundary. Deletional analysis of the dhfr promoter revealed that an E2F site is critical for the increase in promoter activity at the G1/S phase boundary. Although this was the first indication that the E2F family is a critical regulator of S phase events, other genes involved in cell cycle progression or DNA replication were soon shown to be regulated by E2F. These early studies of the E2F family relied upon candidate gene approaches to identify E2F-regulated genes. Such studies were time-consuming and biased towards identifying E2F target genes that controlled processes already known to be regulated by E2F. Therefore, it was not clear if the E2F family would be limited to controlling the G1/S phase transition or if it was more broadly involved in other cellular functions. Fortunately, the development of DNA-based microarrays soon allowed a less biased examination of the global cellular role of E2F. DNA microarrays have revolutionized the study of gene expression and are produced by the utilization of robotics that allows precise spatial control over the synthesis (or deposition) of DNA-based oligonucleotide sequences or PCR fragments onto solid platforms (usually glass slides) to use as a massively parallel probe of unknown samples of DNA or RNA. To date, five independent groups have used high-density oligonucleotide microarrays to study the effects of overexpression of E2F1 protein levels on gene expression. Although each group employed a different system to introduce exogenous E2F1 and also used different mammalian cell lines, the studies resulted in complementary results that have contributed fundamentally to our knowledge of E2F activity in vivo. Interestingly, each group used an inducible system to express E2F1, which may have been required given that it has been demonstrated that E2F1 overexpression causes growth inhibition, both in normal and transformed cells. In one of the first studies to emerge in this crowded field, Kalma et al studied the effects of E2F1 overexpression in a Rat1a cell line that had been stably transfected with a zinc-inducible E2F1 cassette.12 They used this system to prepare RNA from cells in which expression of E2F1 3 Mammalian E2F Regulatory Networks Using DNA Microarray Hybridization Analyses had been induced for 12 or 16 hours. The RNA was then used to probe both an Atlas membrane containing several hundred rat cDNAs, as well as an Affymetrix rat-specific oligonucleotide array that represented ~8700 individual mRNAs. They identified >35 genes from the Affymetrix screen and three genes on the Atlas membrane screen that were reproducibly upregulated by E2F1 overexpression. Their results were similar to what was already known about the E2F family in that at least nine putative targets identified in their screens had previously been shown to be involved in DNA replication. One of the genes they found, RPA2, had not previously been identified as an E2F target, and they used several different types of follow-up analyses to conclude that RPA2 was indeed a bona fide E2F1 target in vivo. Thus, this study was one of the first to demonstrate that high-throughput microarray expression analysis could allow the identification of novel targets of E2F. However, these initial studies did not reveal any new processes in which E2F participates. The Ginsberg group next expanded their analysis and found 58 upand 28 downregulated genes upon induction of E2F3 in Rat-1a fibroblasts. Clustering analysis showed that many of the putative E2F targets fell into three categories: DNA replication, DNA repair, and mitosis. Because of the already known link between certain E2F target genes and cell cycle regulation, the laboratory of J. Nevins used Affymetrix arrays to compare global gene expression changes as mouse embryo fibroblasts (MEFs) move through the cell cycle to changes caused by introduction of exogenous E2Fs. Using Affymetrix microarrays containing ~6200 sequences, they first identified changes in gene expression that occurred as cells progressed from serum starvation-induced quiescence into S phase and secondly, as cells were released from a block at the G1/S-phase boundary that was induced by treatment with hydroxyurea (HU). The results obtained were clustered based on expression patterns. Interestingly, examination of the gene expression clusters between cells brought out of quiescence and cells released from the HU block demonstrated that differences between the two different types of cycles exist, and these differences were most evident at the G1/S-phase boundary and during the G2 phase of the cell cycle. To compare these two types of cell cycle-induced changes in gene expression with E2Finduced gene expression, they infected quiescent MEFs with adenoviruses containing either E2F1 or E2F2 and collected RNA after 18 hours. They identified genes involved in transcriptional regulation and DNA repair as likely E2F target genes and, not surprisingly, found many common genes between those regulated by introduction of E2F and those that changed expression at the G1/S phase of the cell cycle. However, an additional set of putative E2F target genes were identified that changed expression in the G2 phase. Many genes involved in mitotic functions are activated in G2, and this study was the first to suggest a role for E2F in regulating mitosis. A fairly comprehensive study of the effects of overexpression of human E2F1, -2, and -3 has also been performed. The Helin group used human U2OS cells to create stable cell lines that expressed estrogen receptor ligand binding domain (ER)-E2F fusion genes. They showed that upon addition of an estrogen receptor agonist, the ER-E2F protein moved into the nucleus and activated known E2F-regulated genes, such as cyclin E1. For novel target gene identification, they induced expression of ER-E2F1, ER-E2F2, or ER-E2F3 for eight hours. RNA from these cells was used for hybridization to Affymetrix oligonucleotide arrays that represented ~35,000 different mRNAs. Using the McNemar test to identify genes that were significantly altered in expression following E2F activation, they found 1240 mRNAs were regulated by at least one of the E2Fs. Although it was not practical to confirm each of these putative targets, independent validations of a subset of the mRNAs (via Northern analysis) revealed a very low false positive rate of 2%, suggesting that most of the identified genes are in fact responsive to changes in the levels of the E2Fs. Interestingly, they found that overexpression of E2Fs could lead to both increases and decreases in mRNA levels. They found that activation of genes by Microarrays and Transcription Networks 4 ER-E2F fusion proteins did not require de novo protein synthesis, suggesting a direct involvement of E2F in the transcription of those genes. However, repression by ER-E2Fs did require de novo protein synthesis, suggesting that the repression required production of another transcription factor and that E2F indirectly regulates the repressed genes. To categorize large sets of E2F target genes into different functional groups, they performed a target gene bias analysis. The basis behind such an analysis is that if a disproportionate number of genes that are involved in specific cellular functions are regulated by E2F activation, then that cellular function must be regulated by E2F. Not surprisingly, they too found that E2F targets significantly overlapped mRNAs that are regulated by serum stimulation of quiescent fibroblasts. However, their analysis also showed a correlation between E2F target genes and genes involved in the TGFβ signaling pathway as well as targets of other transcription factors such as homeobox family members, p53, and c-Myc. These data suggested that E2F plays a much broader physiological role than the previously characterized roles in DNA synthesis, apoptosis, and cell-cycle regulation.4 The laboratory of D. Cress used a mouse model system to examine the effects of expression of exogenous E2F1 in quiescent 3T3 cells. Thirty hours after infection with an adenovirus expressing E2F1, total RNA was isolated and hybridized to an Affymetrix array containing ~6500 sequences. This study produced several novel findings when compared to the other E2F1 overexpression studies. For example, Ma et al found that infection of quiescent fibroblasts with E2F1 affected the regulation of genes involved in signal transduction and cell membrane biology. Similar to the other studies, they found numerous genes that were repressed by exogenous E2F