Twenty years ago, the availability of several plant genetic maps allowed the development of the first comparative genetic mapping studies [1]. Results indicated that the marker content and order were globally conserved within plant families, between species that sometimes display huge differences in genome size and organization. Thus, macrosynteny was established within the Brassicaceae, the Fabaceae, the Solanaceae and the Poaceae families [2]. In Poaceae, the spectacular conservation of macrosynteny gave the opportunity to draw a consensus map for seven different grass genomes that diverged more than 50 million years ago [3]. In contrast to within family comparative mapping, macrosynteny appears less conserved between distantly related species belonging to distant families or different clades [4, 5]. With such a comparison, the overall conservation of the genome organization appears scrambled by mechanisms of genome evolution such as chromosome rearrangements and genome duplications, to such an extent that the synteny was frequently limited over small genetic intervals. Recently, the increase of complete or partial genomic sequence of reference plant species from different families and the development of the BAC clones sequencing in experimental species gave the opportunity to perform direct comparisons at the genomic sequence level. Beside intra-family comparisons, microcollinearity was investigated between Arabidopsis and tomato [6, 7], Arabidopsis, tomato and Capsella [8], Arabidopsis, M. truncatula, P. trichocarpa and Lotus japonicus [5] and Arabidopsis, Medicago, Populus and Cucumis melo [9]. Despite a lack of macrosynteny, a significant microcollinearity was established between these phylogenetically distant species. In addition, direct sequences comparisons allowed the detection of a complex network of microcollinearity between Arabidospis and tomato [7] and between Arabidospsis, Medicago and Populus [5] suggesting the presence of ancient segmental duplication in the Arabidopsis genome. Altogether, these analyses provided new insight into the plant genome structure and evolution.

The Rubiaceae family (Euasterids I clade) is one of the largest angiosperm families, comprising about 600 genus and more than 13,000 species and including the Gardenia, the Cinchona and the Coffea genus. The genus Coffea comprises two economically important crop species: C. arabica L. and C. canephora Pierre that account respectively for 65% and 35% of the worldwide coffee production (International Coffee Organization, http://www.ico.org). C. canephora represents a model genome in the Rubiaceae family as it has a diploid genome (2n = 2x = 22), a relatively small genome size (2C = 1.43 pg) [10] compared to the allotetraploid genome of C. arabica and an extensive genetic diversity [11]. Moreover, genomic resources have been recently established for C. canephora, such as a BAC library [12] and the availability of 55,840 C. canephora ESTs from different tissues and stages of seed development [13] to facilitate the isolation and the characterization of genes with agronomic interest. However, despite its agronomical importance, very little information concerning the composition, the structure and the evolution of the C. canephora genome is now available. Particularly, it remains unclear so far how the distantly related sequenced dicotyledonous reference genomes such as Arabidopsis, Medicago, Populus, grapevine and tomato will have a predictive value to study the genome structure and to isolate genes of interest in C. canephora. Recently, in an effort to identify factors implicated in ethylene perception during the ripening of the C. canephora berries, several ethylene receptors and transcription factors genes were isolated [14–16]. One of them called CcEIN4 was identified on an isolated BAC clone [15].

In this study, we present the complete sequencing of the C. canephora BAC clone 46C02, carrying the CcEIN4 gene. Sequence analysis of the 160 kb of the BAC indicates a high density of active genes and few transposable elements. Sequence comparisons reveal a high microcollinearity between the C. canephora BAC clone and most of the current dicotyledonous sequenced genomes that diverged 114 to 125 Million Years Ago (MYA).

The high gene density in the CcEIN4 region represents a gene-rich segment of the C. canephora genome. The organization and composition of the trancriptionally active gene part of the C. canephora genome remain unexplored so far. At the CcEIN4 region, the gene density reaches one gene per 7.8 kb. Most of these genes (75%) were shown to be active or potentially active. This value is similar to the gene density observed in different euchromatic regions in tomato (genome size 950 Mb), one gene per 6.7 kb [23] and in the tomato ovate and JOINTLESS loci, one gene per 6.2 and per 8 kb, respectively [7, 24]. Similarly, the gene density in the C. canephora 46C02 BAC clone is comparable to euchromatic regions in M. truncatula (genome size ~500 Mb), one gene per 6.7 kb [25] and in rice, one gene per 6.7 kb [26]. However this density is relatively low compared to the Arabidopsis genome (125 Mb) by a factor ~2, one gene per 4 kb, [22, 27], but still significantly high compared to the Populus genome (genome size 550 Mb), one gene per 11.1 kb, [28] and the euchromatic regions in sorghum (genome size 735 Mb), one gene per 12.3 kb, [29]. The C. canephora nuclei have been estimated to contain ~1.43 pg of DNA corresponding to a diploid genome, giving an estimated haploid genome size of ~690 Mb [10]. An extrapolation of the gene number identified in the 46C02 BAC clone to the whole C. canephora genome would lead to a predicted gene content of approximately 88,000 genes. This obvious over estimation is significantly higher than the average number of predicted genes in diploid reference plant genomes such as in tomato, ~38,000 [30], Arabidopsis, ~25,000 [22], grapevine ~30,000, Populus, 45,000 [28] and rice, ~37,000 [31]. Then, our data suggest that the high density observed at the CcEIN4 region may be not representative of the overall gene density of the C. canephora genome but to a high gene density zone. Gene enriched regions were previously identified in dicotyledonous plant genomes such as tomato and Medicago [23, 32]. The tomato genome may cluster most of the active genes within 25% of the 950 Mb genome representing the euchromatin [23]. In Medicago, euchromatin, identified by FISH hybridization with gene-rich BAC clones, represents 20% of the 500 Mb genome [32]. These gene rich regions in the genomes of tomato and Medicago are currently targets of large-scale genome sequencing projects. Because genes are probably not uniformly distributed along C. canephora chromosomes, we hypothesize that the high gene density in the CcEIN4 region represents a gene-rich segment in the C. canephora genome.

The presence of a high number of identified MITEs in the CcEIN4 region is correlated with the presence of a high gene density. The transposable element composition of the BAC indicates a strong bias in the presence of Class II elements compared to Class I. With the exception of a degenerated part of a non-LTR retrotransposon, all identified elements have short lengths and are considered as non-autonomous because they lack coding capacities for a transposase involved in their mobility [33]. Due to their structural characteristics such as their small size (usually less than 500 bp), the presence of short Terminal Inverted Repeat (TIR) at both ends and a high A/T content, most of these elements (73.9%) were identified in C. canephora as Miniature Inverted-repeat Transposable Elements (MITE), a shorter derivative of non-autonomous class II DNA transposons [34, 35]. MITE families were shown to be the most abundant TEs in plant genomes, with probably more than 90,000 copies in the rice genome [36]. Particularly, MITEs were frequently found associated with coding regions in plant genomic sequence with insertion into introns as well as into untranslated and promoter regions [24, 37, 38], which was confirmed with the complete sequencing of the Arabidopsis and rice genomes. Here, the identified MITEs were closely associated with coding regions as demonstrated by their distribution on the C. canephora BAC and their redundancies in C. canephora EST sequences.

The CcEIN4 region is located in the euchromatic part of the C. canephora. Although the organization of gene-rich regions along C. canephora chromosomes remains unknown, previous cytological observations in C. canephora and C. arabica chromosomes have clearly shown a pattern of deeply and lightly staining regions indicating an overall chromosome organization in respectively condensed heterochromatin and decondensed euchromatin regions [39, 40]. Similarly to Medicago and tomato chromosome architecture, C. canephora heterochromatin regions are mainly located around centromeric regions while euchromatin constitutes the distal parts of chromosomes [13, 32].

Based on these cytological observations and the similar composition observed between euchromatic regions in tomato, in Medicago and the sequenced BAC clone 46C02, we suggest that the CcEIN4 region is located in the euchromatic part of the C. canephora chromosome corresponding to the LG H of C. heterocalyx.

Comparison of the gene order between a C. canephora BAC clone and reference dicotyledonous genome sequences shows extensive conservation. Comparative genomic studies are essential approaches to understand the evolution of genome structure and to investigate the conservation of gene order between closely and distantly related plant species. The evaluation of the genome conservation allows the transfer of information from model species as references to "orphan" genomes lacking the availability of resources and may enhance the identification of gene of interest through map-based cloning strategies [41]. The sequencing of a gene rich region in C. canephora and the availability of several complete or forthcoming sequenced model dicotyledonous genomes such as Arabidopsis, black cottonwood, barrel medic, tomato and grape allowed us to investigate multiple microcollinearity relationships over distantly related species covering two different clades: Asterids and Rosids.

The CcEIN4 region is clearly microcollinear with one homologous region in tomato, grapevine and Medicago and two homologous regions in Arabidopsis and in Populus genomes. Phylogenetically, C. canephora is closely related to tomato since they belong to the Asterid I clade and diverged from a common ancestor about 83–89 Million Years Ago (MYA) (Figure 4)[42]. Arabidopsis, Medicago, Populus and grapevine are more distant than tomato since they are all members of the Rosid clade, which diverged from the Asterid one 114–125 MYA (Figure 4). In the region studied here the number of collinear segments per genome and the degree of microcollinearity appeared heterogeneous. The CcEIN4 region is microcollinear with one homologous region both in tomato, grapevine and Medicago chromosomes, but scattered on two regions in Arabidopsis and in Populus genomes with a different degree of collinearity for each pair of homologous segments. These extra microcollinear segments lead to a complex network of conservation similar to that observed between tomato and Arabidopsis [6, 7], Medicago and Arabidopsis [4, 5], Medicago, soybean and Arabidopsis [43], Populus and Cucumis [9] and Arabidopsis and legumes genomes [5]. Such a network of conservation was recently suggested at synteny level between markers of the C. arabica S_H3 region and Arabidopsis chromosomes 1, 3, 4 and 5 [44]. The analysis of networks of collinearity contributed to the discovery of recent large-scale chromosome duplications that shaped the genomes of Arabidopsis, Populus and Medicago, as well as the understanding of the evolutionary mechanisms that reorganized the duplicated blocks [20, 21, 28, 45]. Moreover, recent comparative analysis between grapevine and the different reference genomes such as Arabidopsis, Populus and rice suggested no recent duplication, but an ancestral contribution of three genomes in grapevine common to all Eurosid plants [19], The data reported here confirm the paleopolyploid structure of the Arabidopsis and the Populus genomes. However, few collinear genes were found conserved between paralogous segments in Arabidopis and Populus, corresponding to the C. canephora CcEIN4 region. In Arabidopsis an extensive process of selective gene loss has shaped the paralogous segments while in Populus the duplicated block appeared very limited in length. Similar observations were described in other regions of paleopolyploid plant genomes where duplicated blocks were found subjected to intense mechanisms of gene movements and large rearrangements [20, 28]. Although the presence of genome duplications and ancient triplication were respectively established in Medicago and suggested in grapevine, no network of collinearity was found between the C. canephora CcEIN4 region and these model species. Similar observations were found between the melon linkage group 11 and Medicago sequences [9] and between chickpea and Medicago sequences [46] reflecting the incomplete sequence of the Medicago genome or a region not covered by ancient polyploidization events in the Medicago and the grapevine genomes.

Considering the divergence between C. canephora and the model dicotyledonous plant genomes studied in this work, we expected a higher degree of microcollinearity between the Rubiaceae and the Solanaceae families than between Asterid and the distantly related Rosid clades. Synteny was observed between tomato and C. canephora in preliminary comparative mapping using COSII markers and microcollinearity relationships were previously demonstrated between tomato and the distant species Arabidopsis [6, 7, 24]. However, in the present work, only four genes were conserved in the same order and orientation over a similar distance and homologs of other Coffea genes were not present in all currently released tomato BAC sequences. Unfortunately, to date no tomato BAC sequence overlapping the entire homologous C. canephora BAC sequence is publicly available, suggesting that the homologous CcEIN4 region was not completely sequenced as part of the tomato genome project. Progress in the tomato sequencing project and future large scale genomic sequencing in C. canephora will allow us to re-evaluate the microcollinearity over several regions and to determine the limits of tomato genome as a tool to identify candidate genes in C. canephora. Nine genes were found strictly collinear and only two coding regions, located in the grapevine interval, were absent in the C. canephora counterpart. In contrast, an unexpected degree of microcollinearity was found between C. canephora and all Rosid plant genomes used in this study. The higher degree of microcollinearity (56%) corresponds to the comparisons between C. canephora and grapevine regions. In this later case, nine genes were found strictly collinear and only two coding regions, located in the grapevine interval, were absent in the C. canephora counterpart. Despite the high degree of microcollinearity, an expansion of distance of a factor 4.9 in grapevine was observed due to differential accumulations of transposable elements in intergenic spaces. This expansion of distance in grapevine is not correlated with the genome sizes in grapevine (~504 Mb, [47]) and in C. canephora (~690 Mb, [10]).

As it was previously observed in comparative analysis in grass genomes, massive insertions and deletions of TEs may promote a rapid genome evolution and may participate in the disruption of the microcollinearity [48]. Our analyses suggest that beside the alteration of distance, insertions of TE in grapevine show no impact on microcollinearity. Our observations indicate that a significant microcollinearity may be expected between C. canephora and grapevine. Phylogenetically, grapevine belongs to the Vitaceae family that is the earliest diverging lineage of the Rosid clade [49]. To date, it remains unclear whether the phylogenetic relationships of Vitaceae are at the origin of the local genome conservation with C. canephora.

The level of microcollinearity is still fairly high with Populus (LG XIII, 49%), Arabidopsis (segment A, 40%) and Medicago segments (40%) but lower than with grapevine. Several disruptions of the microcollinearity were observed between C. canephora and Populus, Arabidopsis and Medicago giving a mosaic pattern of conservation. In Populus and Medicago, the disruptions of the microcollinearity are due to the presence of numerous nonconserved genes with the C. canephora counterpart. Most of the extra genes in Populus and Medicago are not collinear across the Rosid model species suggesting that these gene insertions may arise from independent events of gene translocation mechanisms. Five genes located downstream the collinear region between Medicago and C. canephora were found to be collinear between Medicago and tomato. The extension of the microcollinearity pinpoints the complete absence in Medicago of a cluster of six genes present in C. canephora (g16 to g21), reinforcing the mosaic pattern of the gene conservation.

Despite the observation of the microcollinearity disruption, and considering the large number of collinear genes and distantly related species involved in this analysis, we conclude that the C. canephora CcEIN4 region is significantly conserved with all reference dicotyledonous species used in our analysis. Such observations raise the question whether the microcollinearity observed at the CcEIN4 region can be representative of the conservation between distantly related species or indicative of a larger conservation of the genome organization. So far, no other comparative analysis has been performed between C. canephora or any other Rubiaceae genomic sequences, and model dicotyledonous plant species. Further comparative mapping across dicotyledonous species will provide useful information about the conservation of the plant genome structure and also to determine a set of model genomes to infer positional information of candidate genes in C. canephora.

In conclusion, our exploration of the first sequenced BAC clone in the Rubiaceaae family represents the first step in the understanding the C. canephora genome composition, structure, and evolution. Particularly C. canephora shows a remarkable level of microcollinearity with the distantly related Rosid species. In the absence of the complete tomato genome sequence, our data suggest the grapevine genome may be useful to obtain information to identify candidate genes in C. canephora.