当前位置: 首页 > 期刊 > 《核酸研究》 > 2004年第21期 > 正文
编号:11370087
Computational method for discovery of estrogen responsive genes
http://www.100md.com 《核酸研究医学期刊》
     Knowledge Extraction Lab, Institute for Infocomm Research, 21 Heng Mui Keng Terrace, Singapore 119613

    * To whom correspondence should be addressed. Tel: +65 6874 8800; Fax: +65 6774 8056; E-mail: bajicv@i2r.a-star.edu.sg

    ABSTRACT

    Estrogen has a profound impact on human physiology and affects numerous genes. The classical estrogen reaction is mediated by its receptors (ERs), which bind to the estrogen response elements (EREs) in target gene's promoter region. Due to tedious and expensive experiments, a limited number of human genes are functionally well characterized. It is still unclear how many and which human genes respond to estrogen treatment. We propose a simple, economic, yet effective computational method to predict a subclass of estrogen responsive genes. Our method relies on the similarity of ERE frames across different promoters in the human genome. Matching ERE frames of a test set of 60 known estrogen responsive genes to the collection of over 18 000 human promoters, we obtained 604 candidate genes. Evaluating our result by comparison with the published microarray data and literature, we found that more than half (53.6%, 324/604) of predicted candidate genes are responsive to estrogen. We believe this method can significantly reduce the number of testing potential estrogen target genes and provide functional clues for annotating part of genes that lack functional information.

    INTRODUCTION

    Estrogen is a steroid hormone that has a profound impact in maintaining human physiological function. The unbalance of its production and metabolism can cause many diseases, such as breast cancer (1), obesity (2), osteoporosis (3) and Alzheimer's disease (4). Estrogen exerts its influence through multiple modes of action. Although the mechanisms vary, the ultimate effects of estrogen are manifested by the function of its responsive genes. Thus, identification of estrogen responsive genes becomes a crucial step for overall understanding of estrogen effects and diseases it influences.

    The classical mode of action of estrogen is mediated by its receptors (5), known as estrogen receptor (ER) alpha and beta (ERalpha and ERbeta). These receptors act as ligand-dependent transcription factors (TFs) that bind to specific DNA sequences known as estrogen response elements (EREs) in the target gene's promoter region. Both ERalpha and ERbeta recognize the same ERE fragment, though they have different affinities. Importantly, ERs are not the only factors involved in estrogen-controlled transcription regulation. Other co-activators and co-repressors are essential in the process. The distribution of ERs and their co-regulators are tissue- and cell-type specific (6). Relations between ERs, co-regulators and sets of TFs are established partly within the context of promoter region.

    Traditionally, hormone research is based on studies of individual genes. Recently, large-scale gene expression and protein–protein interaction technologies significantly expanded the coverage of genes included in the experiments. Now, high-throughput technologies make it possible to screen a large number of estrogen responsive genes under the same experimental condition. While large number of data is accumulated this way, researchers still face at least three problems: (i) A significant number of genes found in microarray experiments are lacking well-documented functional information, though they respond to estrogen stimulation (7); (ii) While large-scale gene expression approach can provide a systematic overview of the biological response, it is not able to directly link such response to specific regulation pathways; the expression outcome can be direct or indirect reaction to estrogen treatment; (iii) In-depth studies of individual estrogen responsive genes are still very tedious and expensive. Thus, there is a need to introduce a fast, economic, yet efficient method to identify human estrogen responsive genes and provide rational clues for those that lack functional information but may be responsive to estrogen. In this study we propose a simple, economic and effective computational method to predict a subclass of genes responsive to estrogen and thus provide a partial solution to the above-mentioned problems.

    MATERIALS AND METHODS

    Before giving details of methods, we need to introduce several definitions:

    Detectable ERE site—We define a detectable ERE site as a 17 bp ERE pattern that can be detected by Dragon ERE Finder (8) system at its default setting (http://research.i2r.a-star.edu.sg/promoter/ERE-V2b/).

    ERE frame—It is a 117 bp fragment that contains a detectable ERE in its center flanked by 50 bp upstream and downstream (Figure 1).

    Master genes—These are human genes for which responsiveness to estrogen has been experimentally proved and which contain detectable ERE sites in their promoter regions.

    Figure 1. An illustration of the ERE frame that contains a 117 bp segment centered by a core 17 bp ERE.

    Set of human promoter sequences

    Since our goal is to hunt human estrogen responsive genes, this study focuses on human genes. We collected human promoter sequences using two bioinformatic tools: PromoSer (9) (http://biowulf.bu.edu/zlab/PromoSer/) and FIE2 (10) (http://research.i2r.a-star.edu.sg/FIE2.0/). We used both programs with their default settings to identify transcription start site (TSS) and consequently extract promoter sequences covering segment relative to TSS determined by these tools. In total, >18 000 human promoters are collected in this way. Both of these tools have been developed to cater for the accurate extraction of promoter sequences based on the expressed sequence data, thus ensuring a great accuracy of TSS estimates. Details of their functioning and algorithms are provided in (9,10), and they have implemented the best current practices for determination of TSS location using EST/cDNA/mRNA available data.

    Selection of master genes

    As a source of master genes, we used ERGDB (11) (http://research.i2r.a-star.edu.sg/promoter/Ergdb-v11/index.htm), which is a database of manually curated data on genes reported to be responsive to estrogen in different wet-lab experiments. We are aware that not all of these genes are direct estrogen targets, since estrogen responsive genes could be activated in the ER-independent manner (12,13). Out of all (251) human genes in ERGDB that have detectable ERE sites in their promoters, we made a random selection of 60 of these genes as our master genes. We expect that a part of these 60 genes are direct estrogen targets.

    Hunting estrogen responsive genes

    Using each of the frame sequences from 60 master genes and the BLAST program (14), we matched the ERE frames with our internal human promoter sequence collection. To insure high-quality matches, we used the following criteria to filter the blast results:

    The identity rate between the ERE frame of the master gene and the matched target segment is 90%.

    The minimum length of the matched sequence has to be 94 bp (80% of 117).

    A detectable ERE site in the matched sequence should align to the ERE site of the master sequence.

    The E-value of the blast match has to be 10–6, which indicates that the similarity of two sequences is very unlikely by chance.

    Only the top 50 genes in the blast match list are considered in further analysis.

    As a result, 26 ERE frames from 60 master genes have shown at least one significant match with other human gene promoters. In total, 604 non-redundant candidate genes are found in this way.

    Classifications of candidate genes

    We searched for functional information of 604 candidate genes in public databases, including PubMed search, and found that many of them do not have any associated functional information or they produce hypothetical proteins. We denote such genes as genes with ‘function unknown’. We classified the genes that had some information about their function into two categories which we denote as confirmed (where the reaction to estrogen is documented in PubMed) and non-confirmed (where there was no published evidence in PubMed that the gene responds to estrogen). The category of confirmed genes must show direct reaction of the gene to estrogen treatment. As an example, genes found in ERGDB are either up- or down-regulated by estrogen and thus they have been considered to be in this category.

    Evaluation process of candidate genes

    Candidate genes had been subjected to a double evaluation process. We searched for candidate genes through several published microarray studies related to estrogen treatment. Also, we used literature search to find out if a candidate gene is directly responsive to estrogen. The details of these evaluation steps are given below.

    Searching microarray expression data

    We consider the high similarity between the ERE frame of the master gene and the found high-scoring frames in the candidate genes as a valuable sequence-based indicator that candidate genes are potentially responsive to estrogen. To evaluate our prediction of genes reactive to estrogen, we examined 10 microarray gene expression datasets most of which are from human breast cancer cell lines (15–23). We considered genes as responsive to estrogen when changes of their expression levels appear to be at least 1.5-fold (either up or down) after estrogen treatment in at least one microarray dataset. The threshold of 1.5-fold is considered in many microarray studies as sufficient to indicate reaction of the gene (24).

    Searching published literature

    Additionally, we performed a literature search for evidence of gene's responsiveness to estrogen. Consequently, for all 604 non-redundant candidate genes, we searched documented evidence through published literature in PubMed database.

    RESULTS

    Out of sixty master human genes from ERGDB (11) database, we found that 26 genes have high quality hits in 604 non-redundant human promoters. The 604 genes (see Materials and Methods) are considered as candidate genes that have increased likelihood to react to estrogen treatment. Details of matching information are provided in Supplementary Tables 1 and 2.

    Comparing our list of 604 candidate genes with the microarray data, we found that 294 of them (48.7%) are responsive to estrogen based on 1.5-fold threshold. Details are given in Table 1. Additionally, we examined the existence of functional annotation for the respective 604 candidate genes and found that 207 of them do not have any functional information in public databases. For the remaining 397 genes, we found that 19.14% (76/397) of them have documented evidence in PubMed of direct reaction to estrogen treatment. Combining this literature search with the information from microarray studies where 1.5-fold threshold is used, we found that 53.6% (324/604) of all 604 candidate genes react to estrogen. Moreover, if we exclude genes classified by ‘function unknown’, the accuracy of prediction for the remaining genes is significantly improved to 65.0% (258/397). Supplementary Table 2 gives details of the evaluation results from gene expression data and literature search in per gene basis.

    Table 1. Results of prediction evaluation based on starting set of 60 master genes

    The results we have obtained confirm by hard evidence that candidate genes predicted by our computational method are responsive to estrogen in 53.6% of cases. If we combine the computational method with the search of public databases for gene functionality and exclude from the candidate gene list those genes that can be classified by ‘function unknown’, then our method is successful in 65.0% of cases for the remaining genes, i.e. for genes which have some of their functions already known. The real obstacle in assessment whether the predicted candidate genes are responsive to estrogen is that there are very few published results available of relevant microarray experiments with all data. Although, currently, for less than half (46.4%) of our candidate genes, we could not find supporting evidence for their reaction to estrogen, further experiments may prove that some of these genes possess such functions.

    DISCUSSION

    This study emphasizes how to computationally discover a subclass of estrogen responsive genes whose regulation is likely via ERE involvement. In the most well-studied estrogen reaction pathway, the hormone influence gene transcription by binding its receptors to functional ERE. Therefore, we propose that similarity of ERE frames in promoter regions is a good indicator to suggest similar gene reaction to estrogen. Through searching gene expression data, public databases and biomedical literature, we evaluated our predictions for genes responsive to estrogen. The result shows that our computational approach is feasible and that the predictions are successful in significant proportions.

    Several important features in our approach are worth mentioning. The consensus ERE is a 13 bp palindromic sequence, but many functional EREs identified to date have even several bases deviations from the consensus (25). The site mutation and transient transfection assays found that one or more bases different from the consensus can either preserve ERE binding of ER and induce transcription activity, or may completely abolish such functions (26). Due to the cost, experimental validation of ER binding to mutated EREs has its limit in covering each possible variation of the consensus. In addition, the immediate flanking dinucleotide sequences of ERE are also important for ER binding and hormone induction (26). Hence, using the 13 bp palindromic consensus (or 15 bp when consensus is extended by 1 bp on each side) is neither accurate nor appropriate for identification of most functional EREs. A more efficient method is based on a computational approach which uses position weight matrices (PWMs) that can cater for great variability in motif sequences. To detect ERE motifs, we used Dragon ERE Finder (8) tool built on a set of such PWMs, which is designed to detect functional ERE on genome scale with relatively small number of false predictions. Based on evaluation on the known EREs and human chromosome 21, this tool achieves 83% sensitivity while making one prediction in 13 300 nt on average (8). This tool uses core ERE motif of 13 bp flanked by two dinucleotides on each side and searches for the target ERE motif of 17 bp. This tool is publicly available (website in Materials and Methods).

    Promoter motif analysis including TF binding sites (TFBSs) in yeast has revealed motif preservation in functionally related or co-expressed genes (27). Several tools (28,29) exist for detection of such sites. Our approach of assigning a part of the functions (in this case the responsiveness to estrogen treatment) to genes is based on the frequently observed conservation of functionally important cis elements in genes with similar functions. We propose that in humans, the ERE site with its flanking regions is a key promoter segment associated with estrogen-induced transcription for a subclass of estrogen inducible genes, i.e. for those genes whose transcription initiation is characterized by direct binding of ER to ERE sites.

    Our proposal is based on the following assumptions: (i) sequences around ERE sites are frequently highly conserved because of their functional importance, and (ii) the combined effect of ERE and several neighboring TFBSs in promoters are characteristic for transcription regulation of these genes; however, the sets of TFBSs may vary according to the biological characteristics of the gene.

    We base our hypothesis on the fact that a relatively large promoter segment (117 bp) of a gene, which is known to be responsive to estrogen has very similar matches in other promoters. The preservation of such large promoter segments, which include detectable EREs and can harbor sets of TFBSs, suggests possible functional association among the respective genes. This is strengthened by the fact that genes regulated through common mechanisms or co-expressed under specific conditions often share common cis-regulatory elements in their promoters (30).

    Due to the nature of our algorithm, it is important to make reference to EREs embedded in Alu repeats. Our algorithm relies on the similarity of a large segment of promoter region between human promoters. We examined whether this similarity is the consequence of Alu repeat segments in promoters, as well as the presence of EREs within Alu repeats. The results are summarized in Table 2. The proportion of EREs within Alu repeats in promoters of master genes is 33% (20/60), suggesting that it is likely that estrogen responsive genes possess such EREs. The similarity of promoter segments required by our method will favor selection of candidate genes that contain Alu-embedded EREs. This is exactly what one can observe, although a small portion of the candidate genes (7/604) does not contain Alu-embeded EREs. However, one should note that the functional Alu-embeded EREs are known for a long time (31). Moreover, the gene involved in breast cancer (32) is characterized exactly by the ERE within Alu repeat. Finally, many repeats are functional and may contain functional TFBSs. Thus, the presence of EREs within Alu repeats is not unusual and does not contradict our assumptions. Our method predominantly suggests genes characterized in this way.

    Table 2. EREs and Alu repeats

    In evaluation of what is the proportion of the estrogen responsive genes based on published microarray experiments, we faced several problems. First, very few microarray experiments provided the full datasets. Second, the naming convention for the genes included in microarrays is not uniform, making sometimes impossible to unambiguously list all genes that were included in the microarray. Third, some microarrays were custom-made for special purposes, making proportion of estrogen responsive genes from such microarray experiments highly non-representative. For these reasons, it is impossible to make any accurate estimate about the proportion of genes that respond to estrogen, except that the spectrum of such genes is broad. In different microarray experiments that proportion varies from 0.29% up to 65.31%. In Table 3, we provide the summary of these proportions for genes responsive to estrogen based on the criteria that we used. From the six microarray datasets (R1, R4, R5, R6, R7 and R10, see Table 3) for which we had full access to data, we were able to identify 21 259 unique genes. The additional 5246 entries from these data, we were not able to associate to any gene. These make a total of 26 505 microarray entries. We found 9514 estrogen responsive entries based on our criteria.

    Table 3. Genes included in microarray datasets

    On the other hand, note that the microarray data we used come from only four cell lines (MCF-7, ZR-75, T-47D and GH3), first three of which have been used from breast cancer and the last one from pituitary gland. Due to a broad impact that estrogen has on the human body, it is also likely that with other cell lines different sets of genes may appear responsive, thus changing our current perception of what should be the expected proportion of human genes responsive to estrogen.

    Another important feature of this study is the increased accuracy of identification of TSSs and subsequent extraction of appropriate promoter sequence for further analysis. Identification of TSSs and extraction of promoter sequences as accurately as possible are essential for our promoter analysis. In principle, TSS can be identified by mapping the full-length mRNA to the genome. However, mRNA sequences are frequently not full length and many mRNAs stored in public databases miss the most 5' part sequences (35). To cater to this problem, in this study we used both PromoSer (9) and FIE2 (10) tools to extract promoter sequences with high accuracy. Both tools are publicly available (websites in Materials and Methods).

    Obviously, bioinformatics methodology has its strength to detect functional motifs on genome scale as suggested by a recent review (36). Computational algorithms have been applied in a recent genome-wide screen for high-affinity ERE in human and mouse (36). However, our study differs in at least four important aspects from Bourdeau et al. (37). Contrary to Bourdeau's, where, in principle, equal importance is given to nucleotides of the core ERE motif, we used PWM-based model to detect the presence of ERE sequences. Even in the above study, binding ability of ER to ERE shows big difference ranging from <20% to >80% in an array of single replacement in the core ERE. Thus, simply allowing single or two bases difference in ERE motifs in screening for functional EREs is not appropriate. Second, in prediction of ERE motifs, we used Dragon ERE Finder which utilized two groups of PWM models: one which describes EREs which maintain ER binding, and the other which abolish ER binding abilities, because even a single nucleotide replacement in ERE can completely eliminate ER binding (26). Third, we used 117 bp ERE frame (see Materials and Methods) which contains a 17 bp core ERE pattern and similarity between such frames in different human promoters to predict functional relationship among estrogen responsive genes. Large segments of promoter sequences used by ERE frames reduce dramatically the number of false positive prediction caused by random similarity. Fourth, we used much more accurate promoter regions.

    In conclusion, we propose a simple and efficient computational approach to identify human genes which are responsive to estrogen and are probably characterized by direct interaction between their ERE regulatory units and ERs. Our prediction has been extensively evaluated by both examining results from published microarray studies and manual literature checking. By this, we demonstrate that we can make partial function assignment to genes suggesting its responsiveness to estrogen. This method can significantly cut down the number of candidates to be tested as estrogen target genes and provide functional clues for a group of human genes that lack any functional information. However, only a subclass of estrogen responsive genes is covered by our method. The limitation of this approach is that genes reacting to estrogen through membrane receptors (12) or other TFs (13) are ERE-independent and their function cannot be suggested by this method.

    SUPPLEMENTARY MATERIAL

    Supplementary Material is available at NAR Online.

    REFERENCES

    Frasor,J., Danes,J.M., Komm,B., Chang,K.C., Lyttle,C.R. and Katzenellenbogen,B.S. ( (2003) ) Profiling of estrogen up- and down-regulated gene expression in human breast cancer cells: insights into gene networks and pathways underlying estrogenic control of proliferation and cell phenotype. Endocrinology, , 144, , 4562–4574.

    Geisler,J.G., Zawalich,W., Zawalich,K., Lakey,J.R., Stukenbrok,H., Milici,A.J. and Soeller,W.C. ( (2002) ) Estrogen can prevent or reverse obesity and diabetes in mice expressing human islet amyloid polypeptide. Diabetes, , 51, , 2158–2169.

    Bonnelye,E., Kung,V., Laplace,C., Galson,D.L. and Aubin,J.E. ( (2002) ) Estrogen receptor-related receptor alpha impinges on the estrogen axis in bone: potential function in osteoporosis. Endocrinology, , 143, , 3658–3670.

    Xu,H., Gouras,G.K., Greenfield,J.P., Vincent,B., Naslund,J., Mazzarelli,L., Fried,G., Jovanovic,J.N., Seeger,M., Relkin,N.R., Liao,F., Checler,F., Buxbaum,J.D., Chait,B.T., Thinakaran,G., Sisodia,S.S., Wang,R., Greengard,P. and Gandy,S. ( (1998) ) Estrogen reduces neuronal generation of Alzheimer beta-amyloid peptides. Nature Med., , 4, , 447–451.

    Warner,M., Nilsson,S. and Gustafsson,J.A. ( (1999) ) The estrogen receptor family. Curr Opin. Obstet. Gynecol., , 11, , 249–254.

    McDonnell,D.P. and Norris,J.D. ( (2002) ) Connections and regulation of the human estrogen receptor. Science, , 296, , 1642–1644.

    Charpentier,A.H., Bednarek,A.K., Daniel,R.L., Hawkins,K.A., Laflin,K.J., Gaddis,S., MacLeod,M.C. and Aldaz,C.M. ( (2000) ) Effects of estrogen on global gene expression: identification of novel targets of estrogen action. Cancer Res., , 60, , 5977–5983.

    Bajic,V.B., Tan,S.L., Chong,A., Tang,S., Strom,A., Gustafsson,J.A., Lin,C.Y. and Liu,E.T. ( (2003) ) Dragon ERE Finder version 2: a tool for accurate detection and analysis of estrogen response elements in vertebrate genomes. Nucleic Acids Res., , 31, , 3605–3607.

    Halees,A.S. and Weng,Z. ( (2004) ) PromoSer: improvements to the algorithm, visualization and accessibility. Nucleic Acids Res., , 32, : W191–W194.

    Chong,A., Zhang,G. and Bajic,V.B. ( (2003) ) FIE2: a program for the extraction of genomic DNA sequences around the start and translation initiation site of human genes. Nucleic Acids Res., , 31, , 3546–3553.

    Tang,S., Han,H. and Bajic,V.B. ( (2004) ) ERGDB: Estrogen Responsive Genes Database. Nucleic Acids Res., , 32, (Database issue), D533–D536.

    Nadal,A., Ropero,A.B., Laribi,O., Maillet,M., Fuentes,E. and Soria,B. ( (2000) ) Nongenomic actions of estrogens and xenoestrogens by binding at a plasma membrane receptor unrelated to estrogen receptor alpha and estrogen receptor beta. Proc. Natl Acad. Sci. USA, , 97, , 11603–11608.

    Driggers,P.H. and Segars,J.H. ( (2002) ) Estrogen action and cytoplasmic signaling pathways. Part II: the role of growth factors and phosphorylation in estrogen signaling. Trends Endocrinol. Metab., , 13, , 422–427.

    Altschul,S.F., Gish,W., Miller,W., Myers,E.W. and Lipman,D.J. ( (1990) ) Basic local alignment search tool. J. Mol. Biol., , 215, , 403–410.

    Finlin,B.S., Gau,C.L., Murphy,G.A., Shao,H., Kimel,T., Seitz,R.S., Chiu,Y.F., Botstein,D., Brown,P.O., Der,C.J., Tamanoi,F., Andres,D.A. and Perou,C.M. ( (2001) ) RERG is a novel ras-related, estrogen-regulated and growth-inhibitory gene in breast cancer. J. Biol. Chem., , 276, , 42259–42267.

    Cicatiello,L., Scafoglio,C., Altucci,L., Cancemi,M., Natoli,G., Facchiano,A., Iazzetti,G., Calogero,R., Biglia,N., De Bortoli,M., Sfiligoi,C., Sismondi,P., Bresciani,F. and Weisz,A. ( (2004) ) A genomic view of estrogen actions in human breast cancer cells by expression profiling of the hormone-responsive transcriptome. J. Mol. Endocrinol., , 32, , 719–775.

    Terasaka,S., Aita,Y., Inoue,A., Hayashi,S., Nishigaki,M., Aoyagi,K., Sasaki,H., Wada-Kiyama,Y., Sakuma,Y., Akaba,S., Tanaka,J., Sone,H., Yonemoto,J., Tanji,M. and Kiyama,R. ( (2004) ) Using a customized DNA microarray for expression profiling of the estrogen-responsive genes to evaluate estrogen activity among natural estrogens and industrial chemicals. Environ. Health Perspect., , 112, , 773–781.

    Lobenhofer,E.K., Bennett,L., Cable,P.L., Li,L., Bushel,PR and Afshari,C.A. ( (2002) ) Regulation of DNA replication fork genes by 17beta-estradiol. Mol. Endocrinol., , 16, , 1215–1229.

    Lin,C.-Y., Str?m,A., Vega,V.B., Kong,S.L., Yeo,A.L., Thomsen,J.S., Chan,W.C., Doray,B., Bangarusamy,D.K., Ramasamy,A., Vergara,L.A., Tang,S., Chong,A., Bajic,V.B., Miller,L.D., Gustafsson,J-?. and Liu,E.T. ( (2004) ) Estrogen receptor a target gene and response element discovery in breast tumor cells. Genome Biol., , 5, , R66.

    Cunliffe,H.E., Ringner,M., Bilke,S., Walker,R.L., Cheung,J.M., Chen,Y. and Meltzer,P.S. ( (2003) ).The gene expression response of breast cancer to growth regulators: patterns and correlation with tumor expression profiles. Cancer Res., , 63, , 7158–7166.

    Hodges,L.C., Cook,J.D., Lobenhofer,E.K., Li,L., Bennett,L., Bushel,P.R., Aldaz,C.M., Afshari,C.A. and Walker,C.L. ( (2003) ) Tamoxifen functions as a molecular agonist inducing cell cycle-associated genes in breast cancer cells. Mol. Cancer Res., , 1, , 300–311.

    Hayashi,S.I., Eguchi,H., Tanimoto,K., Yoshida,T., Omoto,Y., Inoue,A., Yoshida,N. and Yamaguchi,Y. ( (2003) ) The expression and function of estrogen receptor alpha and beta in human breast cancer and its clinical application. Endocr. Relat. Cancer., , 10, , 193–202.

    Fujimoto,N., Igarashi,K., Kanno,J., Honda,H. and Inoue,T. ( (2004) ) Identification of estrogen-responsive genes in the GH3 cell line by cDNA microarray analysis. J. Steroid Biochem. Mol. Biol., , 91, , 121–129.

    Zhu,G., Reynolds,L., Crnogorac-Jurcevic,T., Gillett,C.E., Dublin,E.A., Marshall,J.F., Barnes,D., D'Arrigo,C., Van Trappen.P.O., Lemoine,N.R. and Hart,I.R. ( (2003) ) Combination of microdissection and microarray analysis to identify gene expression changes between differentially located tumour. Oncogene., , 22, , 3742–3748.

    Gruber,C.J., Gruber,D.M., Gruber,I.M., Wieser,F. and Huber,J.C. ( (2004) ) Anatomy of the estrogen response element. Trends Endocrinol Metab., , 15, , 73–78.

    Klinge,C.M. ( (2001) ) Estrogen receptor interaction with estrogen response elements. Nucleic Acids Res., , 29, , 2905–2919.

    Cliften,P., Sudarsanam,P., Desikan,A., Fulton,L., Fulton,B., Majors,J., Waterston,R., Cohen,B.A. and Johnston,M. ( (2003) ) Finding functional features in Saccharomyces genomes by phylogenetic footprinting. Science, , 301, , 71–76.

    Pavesi,G., Mereghetti,P., Mauri,G. and Pesole,G. ( (2004) ) Weeder Web: discovery of transcription factor binding sites in a set of sequences from co-regulated genes. Nucleic Acids Res., , 32, , W199–W203.

    Liu,Y., Wei,L., Batzoglou,S., Brutlag,D.L., Liu,J.S. and Liu,X.S. ( (2004) ) A suite of web-based programs to search for transcriptional regulatory motifs. Nucleic Acids Res., , 32, , W204–W207.

    Pilpel,Y., Sudarsanam,P. and Church,G.M. ( (2001) ) Identifying regulatory networks by combinatorial analysis of promoter elements. Nature Genet., , 29, , 153–159.

    Norris,J., Fan,D., Aleman,C., Marks,J.R., Futreal,P.A., Wiseman,R.W., Iglehart,J.D., Deininger,P.L. and McDonnell,D.P. ( (1995) ) Identification of a new subclass of Alu DNA repeats which can function as estrogen receptor-dependent transcriptional enhancers. J. Biol. Chem., , 270, , 22777–22782.

    Singleton,D.W., Feng,Y., Chen,Y., Busch,S.J., Lee,A.V., Puga,A. and Khan,S.A. ( (2004) ) Bisphenol-A and estradiol exert novel gene regulation in human MCF-7 derived breast cancer cells. Mol. Cell. Endocrinol., , 221, , 47–55.

    Nagashima,T., Matsuda,H., Silva,D.G., Petrovsky,N., Konagaya,A., Schonbach,C., Kasukawa,T., Arakawa,T., Carninci,P., Kawai,J. and Hayashizaki,Y. ( (2004) ) FREP: a database of functional repeats in mouse cDNAs. Nucleic Acids Res., , 32, , D471–D475.

    Schonbach,C. ( (2004) ) From masking repeats to identifying functional repeats in the mouse transcriptome. Brief Bioinform., , 5, , 107–117.

    Suzuki,Y., Yamashita,R., Sugano,S. and Nakai,K. ( (2004) ) DBTSS, DataBase of Transcriptional Start Sites: progress report 2004. Nucleic Acids Res., , 32, , D78–D81.

    O'Lone,R., Frith,M.C., Karlsson,E.K. and Hansen,U. ( (2004) ) Genomic Targets of Nuclear Estrogen Receptors. Mol. Endocrinol., , First published on March 18, 2004, 18, , 1859–1875.

    Bourdeau,V., Deschenes,J., Metivier,R., Nagai,Y., Nguyen,D., Bretschneider,N., Gannon,F., White,J.H. and Mader,S. ( (2004) ) Genome-wide identification of high-affinity estrogen response elements in human and mouse. Mol. Endocrinol., , 18, , 1411–1427.(Suisheng Tang, Sin Lam Tan, Suresh Kumar)