Microarray analysis of E2Fa-DPa-overexpressing plants uncovers a cross-talking genetic network between DNA replication and nitrogen assimilation

Previously we have shown that overexpression of the heterodimeric E2Fa-DPa transcription factor in Arabidopsis thaliana results in ectopic cell division, increased endoreduplication, and an early arrest in development. To gain a better insight into the phenotypic behavior of E2Fa-DPa transgenic plants and to identify E2Fa-DPa target genes, a transcriptomic microarray analysis was performed. Out of 4,390 unique genes, a total of 188 had a twofold or more up- (84) or down-regulated (104) expression level in E2Fa-DPa transgenic plants compared to wild-type lines. Detailed promoter analysis allowed the identification of novel E2Fa-DPa target genes, mainly involved in DNA replication. Secondarily induced genes encoded proteins involved in cell wall biosynthesis, transcription and signal transduction or had an unknown function. A large number of metabolic genes were modified as well, among which, surprisingly, many genes were involved in nitrate assimilation. Our data suggest that the growth arrest observed upon E2Fa-DPa overexpression results at least partly from a nitrogen drain to the nucleotide synthesis pathway, causing decreased synthesis of other nitrogen compounds, such as amino acids and storage proteins.

tobacco, carrot, Arabidopsis and rice (Ramírez-Parra et al., 1999;Sekine et al., 1999;Albani et al., 2000;Magyar et al., 2000;Ramirez-Parra and Gutierrez, 2000;Kosugi and Ohashi, 2002a). In the Arabidopsis genome there are three E2F (E2Fa, E2Fb and E2Fc) and two DP (DPa and DPb) genes (Vandepoele et al., 2002). Recently, we have analyzed the phenotypes of plants co-overexpressing the E2Fa-DPa genes . Transgenic plants were smaller than control plants, had curled leaves and cotyledons, and were arrested in growth at an early stage of development. Microscopic analysis revealed that E2Fa-DPa-overproducing cells underwent ectopic cell division or endoreduplication, depending on the cell type. Whereas extra cell divisions resulted in cells smaller than those seen in the same tissues of control plants, supplementary endoreduplication caused the formation of giant nuclei. By using reverse transcription (RT)-PCR, we demonstrated that the expression levels of genes involved in DNA replication (CDC6, ORC1, MCM and DNA pol α) were strongly up-regulated .
Physiologically important targets of the mammalian E2F-DP transcription factors have been identified by microarray hybridization experiments, chromatin immunoprecipitations and computer-assisted prediction (Ishida et al., 2001;Kel et al., 2001;Müller et al., 2001;Weinmann et al., 2001;Ren et al., 2002). E2F-DP-responsive genes can be found among genes involved in cell division, DNA repair and replication, mitotic

Previously we have shown that overexpression of the heterodimeric E2Fa-DPa transcription factor in Arabidopsis thaliana results in ectopic cell division, increased endoreduplication, and an early arrest in development. To gain a better insight into the phenotypic behavior of E2Fa-DPa transgenic plants and to identify
E2Fa-DPa target genes, a transcriptomic microarray analysis was performed. Out of 4,390 unique genes, a total of 188 had a twofold or more up-(84) or down-regulated (104) expression level in E2Fa-DPa transgenic plants compared to wild-type lines. Detailed promoter analysis allowed the identification of novel E2Fa-DPa target genes, mainly involved in DNA replication. Secondarily induced genes encoded proteins involved in cell wall biosynthesis, transcription and signal transduction or had an unknown function. A large number of metabolic genes were modified as well, among which, surprisingly, many genes were involved in nitrate assimilation. Our data suggest that the growth arrest observed upon E2Fa-DPa overexpression results at least partly from a nitrogen drain to the nucleotide synthesis pathway, causing decreased synthesis of other nitrogen compounds, such as amino acids and storage proteins.
progression, apoptosis and differentiation. Although little is known about the plant E2F-DP target genes, a database search has been published recently, in which the Arabidopsis genome was screened for genes harboring the TTTCCCGCC cis-acting element in their promoter (Ramirez-Parra et al., 2003). However, it is still unclear whether this specific cis-acting element is the only one recognized by the plant E2F-DP complexes, or whether the presence of the TTTCCCGCC element is sufficient to mark a gene as a true E2F-DP target gene. In order to identify the functional classes of genes regulated by E2Fa-DPa and to understand the nature of the phenotype of the E2Fa-DPa-overexpressing plants, we designed a microarray experiment that compared the transcript levels of 4,571 genes of wild-type and transgenic lines. We found distinct classes of genes that were up-or down-regulated in the E2Fa-DPa plants. Promoter analysis allowed us to distinguish among the downstream expressed genes, the genes that were putatively under direct control of E2Fa-DPa. Furthermore, we found that the increased expression levels of E2Fa-DPa have a large impact on the expression levels of genes involved in nitrogen assimilation and metabolism.

Construction of microarrays
The Arabidopsis thaliana (L.) Heynh. microarray consisted of 4,608 cDNA fragments spotted in duplicate, distant from each other, on Type V silane-coated slides (Amersham Biosciences, Little Chalfont, UK). The clone set included 4,571 Arabidopsis cDNAs from the unigene clone collection Arabidopsis Gem I (Incyte Genomics, Palo Alto, CA). The functional annotation of the genes related to the spotted cDNAs was retrieved by BLASTN against genomic sequences. To facilitate the analysis, a collection of genomic sequences was built each bearing only one gene. In each of these sequences, the upstream intergenic sequence was followed by the exon-intron structure of the gene and the downstream intergenic sequence, or, in other words, the whole genomic sequence between start and stop codons from neighboring protein-encoding genes. From the BLASTN output, the best hits were extracted and submitted to a BLASTX search against protein databases. From this analysis, the set of 4,571 cDNAs appeared to constitute 4,390 unique clones. To obtain more detailed information concerning the potential function of the genes, protein domains were searched using ProDom. The complete set can be found at http://www.psb.ugent.be/E2F/. The cDNA inserts were amplified by PCR with M13 primers, purified with MultiScreen-PCR plate (Millipore, Bedford, MA), and arrayed on slides using a Generation III printer (Amersham Biosciences). Slides were blocked in 3.5% SSC (1× SSC, 150 mM NaCl, 15 mM sodium citrate, pH 7.0), 0.2% sodium dodecyl sulfate (SDS), 1% bovine serum albumin for 10 minutes at 60°C.

Array hybridization and post-hybridization processes
The probes were resuspended in 30 µl hybridization solution (50% formamide, 5× SSC, 0.1% SDS, 100 µg/ml salmon sperm DNA) and prehybridized with 1 µl poly(dT) (1 mg/ml) at 42°C for 30 minutes to block hybridization on the polyA/T tails of the cDNA on the arrays. Mouse COT DNA (1 mg/ml) (Invitrogen) was added to the mixture and placed on the array under a glass coverslip. Slides were incubated for 18 hours at 42°C in a humid hybridization cabinet (Amersham Biosciences). Post-hybridization washing was performed for 10 minutes at 56°C in 1× SSC, 0.1% SDS, twice for 10 minutes at 56°C in 0.1× SSC, 0.1% SDS, and for 2 minutes at 37°C in 0.1× SSC.
Scanning and data analysis Arrays were scanned at 532 nm and 635 nm using a Generation III scanner (Amersham Biosciences). Image analysis was performed with ArrayVision (Imaging Research Inc, St. Catharines, Ontario, Canada). Spot intensities were measured as artifact-removed total intensities (ARVol) without correction for background. We first addressed withinslide normalization by plotting for each single slide a 'MA-plot' (Yang et al., 2002), where M=log2 (R/G) and A=log2√R -×G -. Dye intensity differences were corrected with the 'LOWESS' normalization. Subsequently, between-slide normalization and differentially expressed gene identification between the two genotypes were performed by sequential analysis of variances (ANOVAs), according to the method of Wolfinger et al. (Wolfinger et al., 2001). (i) The base-2 logarithm of the 'LOWESS'-transformed measurements for all 73,136 spots (yiklm) were subjected to a normalization model yiklm=µ+Ak +AkDlRm+εiklm, where µ is the sample mean, Ak the effect of the kth array (k=1-4), AkDlRm the channel effect (AD) for the mth replication (m=2; left and right) of the total collection of i (i=1, …, 4571) cDNA fragments, and εiklm the stochastic error. (ii) The residuals from this model were subjected to 4,571 gene-specific models rijkl=µ+GiAk+GiDl+GiCj+γijkl, where GiAk is the spot effect, GiDl the gene-specific dye effect, GiCj the signal intensity for genes that can specifically be attributed to the genotypes (effect of interest), and γijkl the stochastic error. All effects were assumed to be fixed, except for εiklm and γijkl. T-tests for differences between the GiCj effects were performed, all based on n1+n2-6 degrees of freedom, where n1 and n2 correspond to the number of wild-type and E2Fa-DPa hybridizations, respectively. Bonferroni adjustment for the 4,571 tests to assure an experiment-wise false positive rate of 0.05 results in a P-value cut-off of 1e -5.0 , which is certainly too conservative. Thus, no further adjustments for multiple testing were done. Therefore, we chose to set the P-value cut-off arbitrarily at the 0.05 level. We used Genstat for both the normalization and gene model fits.

Promoter analysis
The intergenic sequence corresponding to the promoter area of each gene spotted on the microarray was deduced from genomic sequences. From these intergenic sequences, up to 500 bp upstream of the ATG start codon were extracted and subjected to motif searches to retrieve potential E2F elements. Of the 4,571 expressed sequence tags (ESTs) spotted on the microarray, we could retrieve the genomic sequence of 4,390. This difference is due to the presence of duplicate genes and mitochondrial or chloroplast DNA on the microarray. Both the position and frequency of occurrence were determined with the publicly available MatInspector (version 2.2) by using matrices extracted from PlantCARE and matrices made especially for this particular analysis (Lescot et al., 2002). The relevance of each motif was evaluated against a background consisting of all the sequences from the dataset by using the Fisher exact test.

Results and Discussion
Experimental setup and statistical analysis A microarray containing in replicate 4,571 Arabidopsis expressed sequence tags (ESTs) was used to compare the transcriptome of the wild-type with that of E2Fa-DPaoverexpressing plants. cDNA was synthesized from total RNA of plants harvested 8 days after sowing. At that stage, transgenic plants can be distinguished from control plants by the appearance of curled cotyledons that display ectopic cell divisions and enhanced endoreduplication . In the first two hybridizations, including a biological repeat, fluorescently Cy3-and Cy5-labeled probe pairs of control and E2Fa-DPa cDNAs were used. Subsequently, a dyeswap replication was performed for both hybridizations, resulting in a total of four cDNA microarray hybridizations. Because each cDNA was printed in duplicate on the array, eight data points for every gene were obtained.
Fluorescence levels were analyzed to establish whether the expression level of each gene varied according to the overexpression of the E2Fa-DPa transcription factor. Two sequential ANOVA models were used, as proposed by Wolfinger et al. (Wolfinger et al., 2001). First, the model called 'normalization model' accounts for experiment-wise systematic effects, such as array and channel effects, which could bias inferences made on the data from the individual genes. The residuals from this model represent normalized values and are the input data for the second model, called the 'gene' model. The gene models are fitted separately to the normalized data from each gene (see Materials and Methods). In this procedure, normalized expression levels rather than ratios are used as units.
For each of the 4,571 genes on the arrays the genotypespecific signal intensity was determined and t-tested for significant differences (P<0.05). Fig. 1 presents the P values obtained (as the negative log10 of the P value) against the Fig. 1. Volcano plot of significance against effect. Each x represents one of the 4,571 genes, with the negative log10 of the P value from the gene model plotted against the difference between least-square means for the genotype effect. The horizontal line represents the testwise threshold of P=0.05. The two vertical reference lines indicate a twofold cut-off for either repression or induction. magnitude of the effect (log2 of estimated fold change). This so-called volcano plot illustrates the substantial difference of significance testing as opposed to cut-offs strictly based on the fold change. The two vertical reference lines indicate a twofold cut-off for either repression or induction, whereas the horizontal reference line refers to the P-value cut-off at 0.05. These reference lines divide the plot into six meaningful sectors. The 3,126 genes in the lower middle sector have low significance and low fold change, and both methods are in agreement that the corresponding changes are not significant. The 188 genes in the upper left and right sectors have high significance (P<0.05) and high fold change (≥2); 84 of these genes show a significant two-or-more-fold induction of expression, whereas the remaining 104 genes show a significant two-or-more-fold repression of expression in the E2Fa-DPa plants. The identity of these genes was confirmed by sequencing, and the induction of a random set of selected genes was confirmed by RT-PCR analysis (Fig. 2). Finally, the 1,257 genes in the upper middle sector represent significant (P<0.05) up-or down-regulated genes, but with a low (≤2) fold change. The full dataset of genes can be viewed at http://www.psb.ugent.be/E2F/.

DNA replication and cell cycle genes
Genes up-or down-regulated in the E2Fa-DPa transgenic plants can be classified into clear groups according to their function (Tables 1 and 2). Among the genes that are twofold or more up-regulated, 14 belong to the class of DNA replication and modification, correlating with the observation that E2Fa-DPa-overexpressing plants undergo extensive endoreduplication. Most of these genes have previously been   shown to be up-regulated by E2F-DP overexpression in mammalian cells, including a putative thymidine kinase, replication factor c, adenosylhomocysteinase, DNA (cytosine-5)-methyltransferase, and histone genes (Ishida et al., 2001;Müller et al., 2001;Ren et al., 2002). Other E2Fa-DPa-induced S phase genes include a linker histone protein, the topoisomerase 6 subunit A, and two subunits of the histone acetyltransferase HAT B complex, namely HAT B and Msi3. The HAT B complex is responsible for the specific diacetylation of newly synthesized histone H4 during nucleosome assembly on newly synthesized DNA (Lusser et al., 1999). In addition to the overexpressed E2Fa gene (90-fold more abundant in transgenic than in control plants), only one cell cycle gene (CDKB1;1) has a twofold or more change in expression level upon E2Fa-DPa overexpression. CDKB1;1 had already been predicted to be a candidate E2F-DP target by the presence of a consensus E2F-DP-binding site in its promoter (de Jager et al., 2001). Whereas CDKB1;1 activity is highest at the G2 to M transition, its transcript levels start to increase during S phase (Porceddu et al., 1999;. Therefore, up-regulation of CDKB1;1 might be a mechanism linking DNA replication with the following mitosis. That other cell cycle genes modulated in the E2Fa-DPa plants are not detected can be explained by the lack of many important E2F-DP target genes on the microarray and the putative difficulty in identifying changes in expression levels of lowly expressed genes in microarray hybridizations.

Cell wall biogenesis genes
Four members of the xyloglucan endotransglucosylase (XET) gene family are found to be twofold or more up-regulated in the E2Fa-DPa plants, one of them identical to the previously described Meri-5 gene (Medford et al., 1991). XETs are enzymes that modify cell wall components and are presumed to play a role in altering size, shape and physical properties of plant cells. Reversal breakage of the xyloglucan tethers by XETs has been proposed as a mechanism for allowing cell wall loosening in turgor-driven cell expansion (Campbell and Braam, 1999). However, there are several reasons for believing that E2Fa-DPainduced XETs are not required for cell expansion. First, cells divide more frequently in the E2Fa-DPa plants, but the overall cell size is smaller in transgenic than in control plants; so, no overall increase in expansion rates is needed. Second, no induction is seen of genes with a known role in cell expansion, such as expansins. Therefore, the hydrolytic activity of the XETs might rather be required to incorporate the newly synthesized cell walls formed during cytokinesis into the existing cell wall structure. Alternatively, because XET activity has been shown to be involved in the postgerminative mobilization of xyloglucan storage reserves in Nasturtium cotyledons (Farkas et al., 1992;Fanutti et al., 1993), induction of XETs in E2Fa-DPa plants might be related to polysaccharide breakdown to serve the metabolic and energy needs that are required to synthesize new nucleotides (see below). Interestingly, two XETs can be identified in the set of twofold-or-more down-regulated genes. These XETs are more related to each other than to the induced XET genes. This differential response of XETs toward the E2Fa-DPa-induced phenotypes suggests that plant XETs can be classified into at least two different functional classes.

Genes involved in metabolism and biogenesis
A relatively large number of genes involved in metabolism and biogenesis were found in both the up-regulated and downregulated gene groups. Most remarkable is the induction of genes involved in nitrogen assimilation, such as nitrate reductase (NIA2), glutamine synthetase (GS), and glutamate synthase (GOGAT) (Fig. 3). Although not present on the microarray, the nitrite reductase (NiR) gene was found to be induced as well in the transgenic lines, as demonstrated by RTmediated PCR analysis (Fig. 2). Nitrogen and nitrite reductase catalyze the first two steps in the nitrogen assimilation pathway, whereas GS and GOGAT are involved both in the primary assimilation of nitrogen and the reassimilation of free ammonium. This mechanism supplies the plant with all nitrogen needed for the biosynthesis of amino acids and other nitrogen-containing compounds.
There are other indications that the nitrogen metabolism is altered in the E2Fa-DPa plants; these include the modification of genes homologous to genes expressed during the formation of nitrogen-fixing nodules in Medicago sativa (MTN3 and a nodulin-like gene), and the down-regulation of genes involved in sulfur assimilation (two different genes encoding adenylylsulfate reductase [APR] and a putative adenine phosphosulfate kinase). Genes involved in sulfur assimilation have been shown before to be transcriptionally down-regulated during nitrogen deficiency (Koprikova et al., 2000).
The altered expression of genes involved in nitrogen assimilation and metabolism in the E2Fa-DPa transgenic plants might reflect the need for nitrogen for the nucleotide biosynthesis, because purine and pyrimidine bases are rich in nitrogen. If nitrogen assimilation were indeed stimulated by E2Fa-DPa overexpression, two requirement should be fulfilled. Firstly, there should be enough α-ketoglutarate to act as an acceptor molecule for ammonium (Lancien et al., 2000) Journal of Cell Science 116 (20) Fig. 3. Sources of α-ketoglutarate in the E2Fa-DPa-overproducing cells. Genes encoding enzymes shown in red and green are up-regulated or down-regulated in the E2Fa-DPa versus wild-type plants, respectively. Products indicated in blue act as precursors for nucleotide biosynthesis. α-KG, α-ketoglutarate; GOGAT, glutamate synthase; GS, glutamine synthetase; NIA2, nitrate reductase; NiR, nitrite reductase. and secondly, because assimilation of nitrogen is energy consuming, the rate of reductant production should be higher in the E2Fa-DPa transgenic than in the wild-type plants.
Our microarray data suggest that in the accumulation of αketoglutarate in E2Fa-DPa-overexpressing plants is stimulated in different ways. First, α-ketoglutarate production is improved by increased photosynthetic activity, as indicated by the 4.7-fold up-regulation of the large subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (Fig. 3), with accumulation of glyceraldehyde-3-phosphate as a result. Glyceraldehyde-3-phosphate can be converted into fructose-1,6-bisphosphate by fructose bisphosphate aldolase. However, a sixfold down-regulation of the fructose bisphosphate aldolase gene rather suggests the conversion of glyceraldehyde-3phosphate into pyruvate, which can be converted into αketoglutarate in the citrate cycle. The preferential conversion of glyceraldehyde-3-phosphate to pyruvate fits the increased need for amino acids rather than for sugars to drive nucleotide biosynthesis (Fig. 3).
A second source of α-ketoglutarate can be provided by the glyoxylate cycle. In E2Fa-DPa-overproducing plants we observed a 3.1-fold increase in expression of isocitrate lyase, suggesting an increased lipid turnover. Isocitrate lyase activity cleaves isocitrate into glyoxylate and succinate (Fig. 3). Whereas the produced glyoxylate can be converted into glycine, which is also required for de novo nucleotide biosynthesis, succinate can be converted into α-ketoglutarate in the citrate cycle. A 2.3-fold decrease in the expression of the fumarase gene presumably stimulates the subsequent conversion of α-ketoglutarate to glutamate by triggering an accumulation of succinate and fumarate, which are also side products formed during de novo nucleotide biosynthesis (Fig.  3).
Reductant in plants mainly originates from photosynthetic electron transport in leaves. Corresponding with the increased need for reductant, several components of the chloroplast electron transport chain and associated ATP-synthesizing apparatus, such as cytochrome B6, a photosystem II subunit, and the ATPase ε subunit, are up-regulated in the E2Fa-DPa transgenic plants. Increased expression of the protochlorophyllide reductase precursor even indicates an increase in chlorophyll biosynthesis.
E2Fa-DPa plants may suffer from nitrogen starvation that has an impact on amino acid biosynthesis. Three different amino acid aminotransferases are down-regulated in the E2Fa-DPa plants. Shortage of nitrogen-rich amino acids is also evident from the reduced expression of genes encoding vegetative storage proteins (VSP1 and VSP2) and ERD10, a protein with a compositional bias toward glutamate (Kiyosue et al., 1994). Additional evidence for amino acid shortage comes from the down-regulation of a myrosinase-binding protein and the cytochrome P450 monooxygenase CYP83A1. Both proteins are involved in the biosynthesis of glucosinolates, nitrogen-and sulfur-containing products derived from amino acids (Wittstock and Halkier, 2002).

Promoter analysis of E2Fa-DPa-regulated genes
The DNA-binding domains of the E2F and DP proteins are highly conserved between plants and mammals and, correspondingly, plant E2F-DP proteins have been shown by the technique of electrophoresis mobility shift assay to bind to the same canonical DNA-binding site as their mammalian counterparts (Albani et al., 2000;Ramirez-Parra and Gutierrez, 2000;de Jager et al., 2001). Furthermore, these E2F-binding sites regulate the expression of several plant genes involved in DNA synthesis (Kosugi and Ohashi, 2002a;Chabouté et al., 2000;Castellano et al., 2001;Egelkrout et al., 2001;Stevens et al., 2002).
To distinguish between the putatively direct target genes of E2Fa-DPa and the secondarily induced genes, the first 500 bp upstream of the ATG start codon of the genes with 2-fold or higher change in expression were scanned for the presence of an E2F-like-binding site matching the (A/T)TT(G/C)(G/C)C(G/C)(G/C) sequence, which corresponds to all the different E2F-DP-binding motifs that have been described in plants. Of all the different permutations only the TTTCCCGC and TTTGGCGG elements were enriched significantly (P<0.01) in the set of E2Fa-DPaupregulated genes, suggesting these are the preferred binding site of the E2Fa-DPa complex (Table 3). Moreover, six out of eight target genes containing one or more of these elements belong to the group of genes involved in DNA replication and modification. The observation that not all genes that enclase this DNA sequence in their promoter are induced upon E2Fa-DPa overexpression suggests that the presence of the TTTCCCGC or TTTGGCGG motif is not the only element to  TTTCCCCC  49  2  1  TTTCCCCG  31  1  1  TTTCCCGC  46  5  0  TTTCCCGG  61  1  1  TTTCGCCC  16  0  0  TTTCGCCG  76  2  0  TTTCGCGC  19  2  0  TTTCGCGG  30  2  1  TTTGCCCC  35  1  1  TTTGCCCG  13  0  0  TTTGCCGC  24  0  1  TTTGCCGG  34  0  0  TTTGGCCC  54  0  1  TTTGGCCG  38  0  3  TTTGGCGC  18  1  0  TTTGGCGG  47  4  0  ATTCCCCC  14  0  2  ATTCCCCG  21  0  0  ATTCCCGC  11  1  0  ATTCCCGG  23  0  0  ATTCGCCC  10  0  0  ATTCGCCG  42  0  0  ATTCGCGC  13  1  0  ATTCGCGG  9  0  1  ATTGCCCC  14  0  1  ATTGCCCG  6  0  0  ATTGCCGC  15  0  0  ATTGCCGG  0  0  0  ATTGGCCC  42  0  2  ATTGGCCG  13  0  1  ATTGGCGC  12  2  0  ATTGGCGG  28  0  0   Total  864  25  17 *Promoters of mitochondrial and chloroplastic genes were omitted from this analysis make a gene responsive toward E2Fa-DPa, and that E2Fa-DPa may cooperate with other factors to activate transcription. Alternatively, the promoters of non-responsive genes might be shielded with other transcription factor complexes. A putative candidate is the E2Fc protein which, in analogy with the mammalian E2F6 protein, lacks a strong transactivation domain (del Pozo et al., 2002). Alternative candidates are the recently discovered DEL proteins, proven to bind as monomers to the canonical E2F-binding site (Kosugi and Ohashi, 2002b;Mariconti et al., 2002). Because of a lack of transcriptional activation domain, the DEL proteins are postulated to act as repressors of E2F-DP-regulated genes by competing for the same binding site.
It is not excluded that genes without an E2F-like-binding site are not directly activated by E2Fa-DPa. Chromatin immunoprecipitation experiments have shown that mammalian E2F factors can bind to promoters without a clear E2F recognition motif (Kiyosue et al., 1994), suggesting that E2F-DP might recognize non-canonical binding sites, or might be recruited by promoters through the association of other factors. In this respect, the Chlorella vulgaris nitrate reductase gene, of which the Arabidopsis homologue was shown here to be induced by E2F-DPa, binds an E2F-DP complex, although a clear consensus binding site is lacking (Cannons and Shiflett, 2001).
E2Fs can activate as well as repress promoter activity (Trimarchi and Lees, 2002). In the PCNA, MCM3 and RNR2 promoters, E2F sequences have been identified that act as a negative regulatory element during development (Chabouté et al., 2000;Egelkrout et al., 2001;Stevens et al., 2002). In the set of down-regulated genes, no particular enrichment of a specific E2F sequence could be seen (Table 3). Therefore, the data suggest that the E2Fa-DPa complex works as a transcriptional activator and that other E2F-DP complexes are involved in E2F-mediated transcriptional repression.

Conclusions
Microarray analysis of E2Fa-DPa-overexpressing lines identified a cross-talking genetic network between DNA replication, nitrogen assimilation and photosynthesis. The putatively direct E2Fa-DPa target genes as identified by the presence of an E2F-DP-binding site, belong to the group of genes involved in DNA synthesis, whereas the secondarily induced genes are mainly linked to nitrogen assimilation. In a recently published microarray experiment in which the periodic expression of genes during the cell cycle was monitored, genes with a role in nitrogen assimilation (aspartate aminotransferase and a nitrate transporter) were found to be specifically expressed during the S phase . Because purine and pyrimidine bases are nitrogen rich, we postulate that induction of nitrogen assimilation genes during DNA synthesis in wild-type and E2Fa-DPa transgenic plants is required to supply enough nitrogen for nucleotide biosynthesis. However, in the EFa-DPa transgenic plants, increased nitrogen assimilation most probably does not meet all the nucleotide biosynthesis needs, as seen by the expression modulation of many genes involved in nitrogen and carbohydrate metabolism. The drain of nitrogen from essential biosynthetic pathways to the nucleotide biosynthesis pathway is expected to affect other aspects of plant metabolism, as can be seen from the reduced expression of vegetative storage protein genes and genes involved in amino acid biosynthesis. This altered metabolism might, at least in part, contribute to the growth arrest observed in E2Fa-DPa transgenic plants.
The exact regulatory pathways and factors controlling the nitrogen assimilation pathway in plants are still unknown. In addition to the genes involved in DNA replication and metabolism, our data contain a relatively large number of genes with unspecified function (Tables 1 and 2). For instance, a GATA zinc-finger-encoded gene with a still unknown function is found between the up-regulated regulatory genes. This gene might encode the ortholog of the Neurospora crassa nit-2 protein that has been shown to positively regulate expression of the nitrate reductase gene (Fu and Marzluf, 1990). Other regulatory genes modified in the E2Fa-DPa plants encode protein kinases and several putative receptor kinases. These genes might include some novel key regulatory components in the process of nitrogen assimilation or regulation of efficient nitrogen usage. It will be of great interest to analyze their role in nitrogen assimilation, metabolism and plant growth.