Regulatory sequences constitute a small fraction of eukaryotic genomes that determine the level, location and chronology of gene expression. In parallel to functional studies, computational analysis provides different approaches for scanning genomic sequence to identify those regions predicted to participate in gene regulation [1, 2]: (i) sequence analysis of co-regulated genes within a given species, (ii) inter-species sequence comparison of orthologous genes and (iii), database construction and analysis of known transcription-factor binding sites.

Functional studies conducted to identify trans and cis-acting elements controlling the expression of translation factors and ribosomal proteins (rp) in Arabidopsis allowed us to characterize several cis-acting elements. One of them, the telo box (AAACCCTA), was first observed within the promoter of the four Arabidopsis genes encoding the translation elongation factor EF1αpromoters [3, 4] and subsequently within a few plant rp promoters [5]. This short motif is identical to the repeat (AAACCCT)n of plant telomeres [6] but differs from long interstitial telomere repeats (ITRs) which are found at discrete intrachromosomal sites in many eukaryotic species [7, 8] and probably result from chromosomal rearrangements such as end-fusions and segmental duplications. In contrast to the limited number of ITRs observed in pericentromeric and subtelomeric regions in Arabidopsis [8], a preliminary computational analysis suggested that short telomere repeats (telo boxes) were over-represented at the 5' end of Arabidopsis ESTs [9]. More recently, with the achievement of the Arabidopsis sequencing project, we showed that the occurrence of telo boxes within rp promoters is the rule rather than the exception [10, 11]. Telo boxes were also observed in promoters of several protein-encoding genes which, as is the case for rp, are expected to be over-expressed in cycling cells, suggesting that it could be involved in the coordinated expression of this class of genes. Experimental data indicated that the telo box was indeed involved in the expression in cycling cells [11–13]. However, by itself this motif is not able to activate the transcription by RNA pol II but acts in synergy with various cis-acting elements to increase the expression. These cis-acting elements include the TEF1 box identified in promoters of the translation elongation factor EF1α[14], the Trap1 box in the promoter of a rp gene [15] and redundant site II motifs initially characterized in the promoter of the proliferating cellular nuclear antigen gene (PCNA) [16] and subsequently in most Arabidopsis rp genes [11].

In this study, we analysed the distribution of telo boxes within A. thaliana and O. sativa genomes and their association with genes involved in the biogenesis of the translational apparatus. In addition, this analysis revealed a striking analogy with the genomic distribution of telo boxes and plant microsatellites.

One remarkable item of data resulting from this study is the striking similarity observed in the genomic distribution of telo boxes and microsatellites. Their preferential location in 5' flanking regions can be assigned to their role in gene expression as has been reported for both telo boxes [11, 12] and microsatellites [28, 29]. However, we think that this preferential distribution in 5' regions could also reflect a common process involved in the acquisition of these motifs. We previously proposed a model involving the telomerase and recombination events to explain the spreading of telo boxes within Arabidopsis genome [9]. A schematic representation of this model and of its possible analogy with the acquisition process of microsatellites is shown in Figure 8. It can be summarized as follows: (i) Promoter regions are hot spots for recombination and it is well established that there is a relationship between recombination and chromatin accessibility to nucleases occurring during transcription initiation and elongation processes [30–32], (Figure 8A). (ii) Free 3'OH recombinogenic ssDNA is thus generated, (Figure 8B). (iii) These free 3'OH ends are potential substrates for telomerase which, in the absence of telomere repeats interacting with the telomerase anchor site, could act in a non-processive manner by adding only one telomere motif at the 3' end [33], (Figure 8C). It must be emphasized that, as for rp genes, there is also a strong correlation between cell cycle progression and telomerase expression in Arabidopsis [34]. (iv): The 3' end invasion at homologous open sites (Figure 8D) followed by error-prone DNA repair leads to the acquisition of a telomere repeat unit (Figure 8E). A related process has been suggested for the spreading of microsatellites in the human genome by 3'OH-extension of retrotranscripts [35]. As we suggested for the putative generation of telo boxes driven by the telomerase RNA template, the authors speculate that RNA guides could give rise to specific microsatellite sequences. In a similar manner, the spreading of simple repeated sequences such as Y patches could be achieved by addition of nucleotides to free 3' ends by a terminal transferase (TdT), (Figure 8D and 8E). The occurrence in angiosperms of a TdT activity has been reported in germinating wheat embryos [36]. During V(D)J recombination in mammals, the TdT contribute greatly to the generation of diversity in the immune repertoire and the addition of template-independent nucleotides frequently consists of purine or pyrimidine tracts [37]. The common feature in the hypothetical transcription-associated recombination processes mentioned above is the availability of a free 3' end for TdT, telomerase or other related hypothetical specific RNA-guided reverse transcriptase followed by error-prone DNA repair. In the context discussed here it is interesting to mention that similarly to our data showing a high frequency of telo boxes within 5' UTRs of genes encoding components involved in the biogenesis of ribosomes, 46.5% of translation-related genes in rice contain some microsatellites in their predicted 5' UTRs, (GCC/GGC)n contributing for about half of them [19 and our unpublished data].

Biogenesis of ribosomes is a crucial process requiring the coordinate expression of hundreds of genes. In the yeast Saccharomyces cerevisiae this synchronized expression is primarily accomplished at the transcriptional level and mediated through common upstream activating sequences including in most cases Rap1p binding sites (rpg boxes) and, in a small subset of rp genes, Abf1p binding sites [38, 39]. In higher eukaryotes little is known about the transcriptional network controlling this regulon [40]. Studies conducted in our group over the last two decades have led to the identification of several transcriptional trans and cis-acting elements which participate in the over-expression of translational factor and rp genes in dividing plant cells [3, 11, 12, 14, 41]. The data reported in the present work suggest that the occurrence of telo boxes in the 5' flanking regions of rp genes is the rule not only in Arabidopsis but in angiosperms in general and therefore extend this observation to genes involved in the maturation of pre-rRNA. In agreement with data coming from a genome-wide analysis suggesting that the sequences AAACCCTA and TAGGGTTT are Arabidopsis core promoter elements [22], the majority of telo boxes observed in 5' flanking regions of plant translation-related genes are located within a narrow window located -50 to +50 relative to the transcription start site (TSS). The conservation of a topological association between telo boxes and site II motifs or TEF box cis-acting elements provides insights into the transcriptional regulation process required for the coordinate expression of plant genes involved in ribosome biogenesis. For several aspects, a parallel can be drawn between the putative role of telo boxes in plants and those achieved by the rpg cis-acting element in the yeast S. cerevisiae: (i) the rpg boxes (ACACCCAYACAY) show an homology with yeast telomere repeats (C_(1-3)A)n and are both targets for the Rap1p pleiotropic protein involved in telomere metabolism and gene expression [42]; (ii) a common characteristic of yeast genes under the control of rpg boxes is their very high transcription rate during exponential growth. Up to now, the effect of telo boxes on expression was only observed in exponentially-growing cell cultures or in cycling cells of root primordia and young leaves [11–13]; (iii) among the yeast genes up-regulated in an rpg-dependent manner during exponential growth, genes involved in the biogenesis of ribosomes constitute a major class [38, 43, 44]; (iv) the interaction of Rap1p with the rpg box does not directly act as transcriptional activator but instead as a synergistic element that allows the activation by other regulatory proteins in participating in their recruitment in protein-protein interactions or in destabilizing the DNA duplex [38, 45, 46]. Similarly, in gain-of-function experiments, the telo box is not able by itself to activate gene expression in transgenic plants but acts in synergy with other cis-acting elements like site II motifs or TEF boxes [11, 12]. Taken together, these observations support the hypothesis that there are functional similarities between the roles played by interstitial telomere motifs in plant promoters and those of the rpg box in yeast. We have estimated at about 10% the number of Arabidopsis genes harbouring a telo box within their 5' flanking regions suggesting that this element plays a much more general role than solely in the ribosome biogenesis. An intriguing question which might consequently be addressed concerns the meaning of the involvement in both yeast and angiosperms of interstitial telomere motifs in the expression of a set of genes whose expression is, at least for translation-related genes, correlated to cellular proliferation.

In contrast to that observed in vertebrates, many plant snoRNA genes are found in polycistronic clusters composed of homologous or heterologous snoRNAs [47]. Intronic snoRNA genes are frequently found in the genome of rice [26, 27] whereas they are the exception in Arabidopsis [48]. There is currently little information on how the expression of plant snoRNA genes is coordinated with the expression of other components involved in the biogenesis of the translational apparatus. When nested within introns of genes involved in ribosome biogenesis such as fibrillarin SnRNP genes in Arabidopsis or several rp genes in O. sativa the co-expression process appears to be obvious. This co-expression process is much less clear when snoRNAs are expressed from independent promoters in non-intronic genes. Some plant non-intronic snoRNAs are RNA polymerase III products as suggested in Arabidopsis and rice by the characterization of dicistronic tRNA-snoRNA genes [47, 49]. However, it remains to assess the proportion of non-intronic snoRNAs that are transcribed by pol III in plants. Our data suggest that, at least in Arabidopsis, this is probably the exception rather than the rule. The remarkable conservation of the topological association of telo boxes with site II motifs or TEF boxes observed in promoters of genes encoding ribosomal proteins or proteins required for pre-rRNA processing as well as within sequences found upstream of non-intronic snoRNA genes, strongly suggests that the association of these cis-acting elements and their interaction with related trans-acting factors might play a fundamental role in their coordinated transcription by RNA pol II. Moreover, we took advantage of the availability of TIGR-CERES data on the sequencing of full length Arabidopsis cDNAs to map the 5' end of several snoRNA precursors (Additional Files 3 and 4). These full-length cognate cDNAs were obtained by the "cap-trapping" method indicating that the identified RNA precursor molecules harbouring snoRNAs are indeed capped and polyadenylated RNA pol II transcripts. Once again, and as for rp genes, a parallel can be drawn between the putative role played by the telo box in plants and those achieved by the yeast rpg box in snoRNA gene expression. In S. cerevisiae the promoters of non-intronic snoRNA genes contain rpg boxes which are required for their full expression [50]. Thus, the analysis of conserved associations of telo boxes with site II motifs or TEF boxes allowed us to characterize new RNA pol II promoters involved in the biosynthesis of snoRNA precursors. A first analysis suggest that such an approach could be generalized to identify unexpected cryptic RNA pol II promoters within plant genomes (Figure 7). It would be of interest to investigate to what extent such promoters participate in the activation of expression in meristematic cycling cells, as is the case for plant rp or pre-rRNA processing genes showing a similar promoter configuration.

The data reported in this work support the model previously proposed for the way telo boxes spread within plant genomes and provide new insights into a putative process for the acquisition of microsatellites in plants. The conserved topological association of telo boxes with site II or TEF1 cis-acting elements appears to be an essential feature of plant genes involved in the biogenesis of ribosomes and clearly indicates that most plant snoRNAs are RNA pol II products. This conserved association could provide a powerful tool to improve genome annotation in characterizing new cryptic RNA pol II promoters.