The phylogenetic positioning of Buchnera is still ambiguous because of the high evolutionary rate and strong AT bias of its genome (that affects most of its genes). Nevertheless, several evolutionary studies have been performed, and they have tended to converge toward positioning Buchnera within the group of the γ3-Proteobacteria [2,3,11,12]. Hence, Buchnera is closely related to the Enterobacteriaceae including Escherichia coli. The name Buchnera aphidicola has been attributed to all the lineages of aphid endosymbionts using BSg, the endosymbionts of the wheat aphid Schizaphis graminum, as the type species [3]. Since Buchnera is strictly vertically transmitted through aphid generations, speciation of host lineages has been accompanied by the divergence of the associated endosymbiont lineages, resulting in perfect congruence between the phylogenetic trees of the two groups of species. Divergence dates have been estimated on the basis of aphid fossil records and gene evolution rates [13] in order to trace speciation events within the Buchnera clade, and compare the evolutionary rates of different lineages [10,14]. The association has been estimated to have begun around 150 Myr ago [13,15]. During this co-evolution process both partners have become completely interdependent: when aphids are deprived of their primary symbiont, they are no longer able to reproduce, and so far it has been impossible to culture Buchnera without its host in vitro.

Buchnera is located within specialized eukaryotic host cells, known as bacteriocytes, which are located inside the aphid body cavity, adjacent to the ovariole cluster. The bacteriocytes (about 60 to 80 in each aphid) are giant polyploid cells within which Buchnera resides in the cytoplasm surrounded by a host-derived membrane, the symbiosome (Fig. 1). The bacteriocytes are assembled in a bilobed organ-like structure known as the bacteriome. In embryos, a thin layer of syncytial cells surrounds the bacteriome, whereas in old aphids the bacteriome structure degenerates in parallel with the nutritional and reproductive decay of the insect [5].

The most active state of development and reproduction of the aphids is during the cycles of parthenogenesis when the wingless aphid mothers produce living larvae. During this period of telescoping generations (i.e. the larvae include embryos in their own body), two distinct populations of Buchnera coexist inside the aphids: the maternal and embryonic populations. The embryonic Buchnera population corresponds to about 75% of the total Buchnera population of the mother [16]. The growth rate of Buchnera seems to reach its peak during embryo development, just after the young embryos have been colonized [17], whereas the number of Buchnera remains stable in adult aphids, and declines as the aphids get older [5], suggesting a slow growth rate of the bacterium within maternal bacteriocytes.

Like most eukaryotes, aphids can be considered to be complex micro-ecosystems involving consortia of viruses and bacteria, interacting in communities. Part of this diversity occurs in the flora of the digestive tract [18], but also inherited intracellular symbionts are also highly diversified in aphids [19]. In addition to Buchnera, most natural aphid populations also harbor other intracellular symbiotic bacteria, known as secondary endosymbionts. In general, these non-obligatory, secondary endosymbionts are not present in all aphid species nor in all the individuals of the same population. They are transmitted vertically, although horizontal transfers may occur between different aphid populations and species. These bacteria are fairly diverse, and include species of the genera Spiroplasma, Wolbachia, and Rickettsia, as well as several Enterobacteria (Regiella insecticola, Hamiltonella defensa, Serratia symbiotica). Secondary endosymbionts have been shown to have the following functional roles: thermal resistance [20], adaptation to the host plant [21], resistance to parasitism [22] and complementation of the nutritional role of Buchnera [9]. Hamiltonella defensa can invade some bacteriocytes usually reserved to Buchnera [23], and has been shown to harbor active bacteriophages, known as APSE-1 and 2, which encode toxins that may be involved in invading host cells or, on the contrary; protecting them against parasites [24]. To reach a better understanding of symbiosis and of the evolution of Buchnera, these secondary endosymbionts have to be taken into account; some recent experimental results suggest that in some laboratory aphid strains other bacteria may replace Buchnera [25].

Buchnera is transmitted vertically through the host generations, and each host generation has to undergo an infection phase. The infection of embryos with bacteria from the mother occurs during the blastoderm-stage via an opening in the posterior pole of the embryo in the viviparous morphs, whereas in the oviparous morphs, it is the eggs that are contaminated [1,17,26]. One direct consequence of this infection phase is that the symbiont population passes through successive ‘bottlenecks’, i.e. only a small proportion of the maternal symbiont population is transmitted to the offspring. Wilkinson et al. have shown that a single maternal bacteriocyte is the source of all the Buchnera cells in a given embryo, and that Buchnera enters the embryo via a specific membranous duct [17].

Hence, Buchnera populations are of small size, suffer drastic bottlenecks and are isolated within the bacteriocytes, thus preventing any possibility of any exchange of genetic material occurring between bacterial lineages. This peculiar pattern of population dynamics increases the probability that mildly deleterious mutations will become fixed, and also makes it impossible for them to be corrected (this is known as Muller's ratchet effect). Indeed, Buchnera is subjected to increased genetic drift that imposes a bias in genome composition and accelerates evolutionary rates even at non-synonymous sites [14]. This acceleration of the evolutionary rate also reflects the effect of higher mutation rates (due to mutator genes and/or the absence of a repair system), as well as changes in selection pressure due to the intracellular environment [10,14,27–30].

Aphids have developed specialized mouthparts featuring long flexible stylets that they insert into the plant tissues until they reach the phloem cells, from which they exude the sap. This phloem sap constitutes the sole food source in the aphid diet. It is rich in carbohydrates, but deficient in nitrogenous compounds, and so constitutes a very unbalanced diet, and one with a composition that varies depending on the plant, and also on the biotic and abiotic environmental conditions for the same plant. Like those of other insects, the nutritional needs of aphids also vary with their stage of development.

The provision of aphids with nutrients by Buchnera has been extensively discussed, and is the subject of considerable speculation [5]. Experimental studies using a combination of controlled artificial diets and antibiotic-treated aphids (aposymbiotic aphids), as well as genomic information, have focused on nitrogenous compounds that are not present in the phloem sap, and especially on essential amino acids. Douglas et al. have provided evidence for the synthesis of methionine by Buchnera in the aphid Myzus persicae [31], and that of tryptophan in BAp [32]. The involvement of Buchnera in the transfer of nitrogen from glutamine or glutamate has been demonstrated by Sasaki and Ishikawa [33]. Febvay et al. have provided direct evidence for the biosynthesis of threonine, isoleucine, leucine, valine and phenylalanine by BAp [34,35], highlighting the ability of Buchnera to adapt its production of amino acids to the aphid's nutritional needs [36].

If at present we know the qualitative metabolic contribution of Buchnera to the symbiotic interaction [37], the quantitative contribution of Buchnera to the adaptability of the variation of the aphid nutrition remains to be deciphered. Buchnera is often described in the literature as a degenerating bacterium, with low sensitivity to external conditions, and an inability to regulate its transcriptional and post-transcriptional responses to the variable demands of its host [38]. New insights into the symbiotic function of Buchnera provided by systemic analyses have been made possible by the fact that several complete genomes of Buchnera and other endosymbionts are now available, as well as specific tools developed for network analyses.

The genomes of the different strains of Buchnera, ranging in size from 420 to 650 kb, are among the smallest of the prokaryotic genomes to be sequenced so far [39]. The process of genome shrinkage in Buchnera has been extensively studied. Initial studies determined the gene content of the last common ancestor (LCA) Buchnera with its close relative E. coli [6,40,41]. Van Ham et al. [8] and Silva et al. [42] subsequently estimated the genome content of the last common symbiotic ancestor (LCSA) of the four sequenced strains in order to reconstruct the process of gene loss in each lineage (reviewed in [43]). Two models of gene loss have been proposed in order to explain the extreme genome shrinkage of Buchnera. The first model, proposed by Moran and Mira [41], distinguishes between the early massive gene losses caused by the fixation of large deletions and the relatively minor gene losses that have followed as a result of the pseudogenization process. According to this model, the process of genome shrinkage was highly dynamic during the early stages of symbiosis (during the transition from the LCA to the LCSA), resulting in the removal of large genomic areas containing Buchnera genes that had become redundant or unused in the context of the stable intracellular environment [40,41]. The second model, proposed by Silva et al. [40], attributes the genome shrinkage to multiple gene disintegration events dispersed over the whole genome. Recent comparative analyses of the sequenced endosymbiont genomes support this second model [44–46].

It has been proposed that in some Buchnera lineages this dynamic step was followed by a long period of stasis, lasting for more than 50 Myr, due to the loss of insertion sequences and of the genetic elements that mediate an efficient homologous recombination, as well as the reaching of an equilibrium between the number of genes and the symbiosis function [7]. However, the extreme reduction found in other lineages, such as in BCc [9], as well as those revealed by the recent work of Moran et al. [10] on BAp, have revealed that deletions and inactivations eroding the genome of Buchnera are ongoing and do not appear to have reached a limit.

The genome of BAp is composed of a 640 kb circular chromosome plus two plasmids, pLeu and pTrp, carrying all the genes specifically required for the leucine biosynthesis, and two genes encoding the enzymes for the tryptophan biosynthesis pathway [47,48]. The sequence analysis of the two plasmids suggests that they have different replication mechanisms and independent origins [43,49]. However, more accurate analyses are needed to confirm this later hypothesis. Plasmid-borne copies were acquired from an ancestral Buchnera genome, and not by horizontal transfer from an external source [49]. The evolutionary history of the two plasmids reveals highly dynamic recombination process, including an original transfer from the chromosome to the plasmids, and four independent back-transfers and gene-order reorganizations [43].

It is important to note that the genome of Buchnera is highly polyploid [50]. Buchnera has large cells about 3 to 5 μm in diameter (about 15× the volume of one E. coli cell), and contains 10× more DNA than E. coli [50]. Indeed, incomplete division due to an ineffective replication and segregation system may be the cause of the polyploidy in Buchnera [7]. The chromosome copy number per Buchnera cell varies from 50 to 200, depending on the age of the aphids [50,51]. Variable relative numbers of plasmid and chromosome copies within a Buchnera cell were probably responsible for the leucine and tryptophan gene transfers and back-transfers that have occurred between plasmids and chromosome during the evolution of Buchnera [43]. This dynamics of copy number could also be a mechanism of regulating gene expression; however, to date no experimental evidence has been found to support this hypothesis.

The genome of Buchnera consists of 70% of AT bases (90% in the intergenic regions). Important composition biases between the two DNA-strands are observed, and they make it possible to localize both the origin and the terminus of replication of the bacterial chromosome (Fig. 2). The genome of Buchnera is tightly packed with genes (some with overlapping coding sequences), indicating that superfluous DNA was rapidly eliminated during evolution [52,53].

Gene organization along the Buchnera chromosome is shown in Fig. 3. A gene distribution bias between the leading and lagging DNA strands can be seen: 56% of genes are encoded by the leading strand. The bias becomes even more marked if we consider only the essential genes: 60% of the essential genes are located on the leading strand. We investigated the effect of gene DNA strand position, operon organization and gene essentiality on the variability of the basal mRNA abundances in BAp: putative transcription units (ANOVA, F=43.204, p-value<10−4) and gene essentiality (ANOVA, F=18.722, p-value<10−4) are both significant factors in this model. No significant effect was observed between strand position and gene expressiveness, indicating the absence (or the low) selection for gene expression in Buchnera when the genome is replicating [54]. This finding is consistent with the slow growth rate of Buchnera.

The genes in Buchnera are organized into putative transcription units (PTUs) that have been inferred from their orthology and synteny with E. coli (Fig. 3). No specific regulators corresponding to these operon structures have been preserved in Buchnera (see below). However, experimental data using dedicated microarrays reveal a significant correlation between basal and differential gene expression (following nutritional stress applied to the host aphids) and PTU organization, suggesting that these operon structures play a functional role in regulating gene expression in BAp [54,55].

Genomic analysis has revealed that there are virtually no genes for regulatory systems in the BAp genome [6]. Indeed, two-component regulatory systems are also absent. None of the E. coli orthologous regulators of the operons encoding the enzymes of the essential amino acids pathways are present in Buchnera. The genes of Buchnera have no leader sequences, and the bacterium does not use attenuation systems [6]. The genes encoding adenylate cyclase (cyaA) and the AMPc receptor (crp) are absent in BAp, indicating the absence of an ability to respond to a change in carbon source, a system that is usually highly conserved in prokaryotes. Only two sigma factors are present in BAp: rpoD and rpoH. Buchnera does not display any clearly defined transcriptional terminators, and the binding sites for the generalized sigma-70 promoter, required to initiate the transcription of most genes, are not readily distinguishable. However, BAp has retained genome ancestral regulators that may control global shifts in gene expression, corresponding to different stages of the host's life cycle. This global change may be sensed by the accessibility to certain precursors (i.e. ATP), and the effectors may be Nucleoid Associated Proteins (NAPs), such as IHF and FIS. Regulation via cell division and/or genome ploidy (which alters the likelihood of interactions between transcriptional regulators and gene promoters) may also be involved in these gene expression shifts.

Several experiments have been performed to analyze gene expression in Buchnera. PCR fragment-based microarrays [38,56,57] and oligonucleotide microarrays have been produced [54,55,58,59], physiological experiments were conducted, and evidence has accumulated revealing weak gene transcriptional regulation. Gene expression normalization using genomic DNA (gDNA) has been used to compare the gene expression level between genes under the same experimental conditions [54]. Using this approach, Viñuelas et al. [54] experimentally defined a set of highly expressed genes for BAp on the basis of the gene GC rate, which were divided into two groups of slowly and quickly evolving genes respectively (some of them are shown in Fig. 3).

The group of conserved and highly expressed genes contains most of the ribosomal (not shown in Fig. 3) and chaperone (dnaK, dnaJ and mopA, ipbA) protein genes. Moran et al. [38,60,61] have provided evidence that most of the 20 heat shock genes retained in the Buchnera genome are constitutively over-expressed compared to other bacteria. Several studies have been performed on the GroEL protein, originally called symbionin, as it is the most abundant protein of Buchnera [62] representing almost 10% of the total amount of aphid proteins found in BSg [63]. It is supposed that GroEL and the other chaperonins buffer the effects of translation errors so that the proteins that have accumulated mildly deleterious mutations in Buchnera can still fold correctly [14,64].

Most of the flagellar genes (11/16), and the two orphan genes of Buchnera (yba3, yba4) have been identified among the fast-evolving highly expressed genes [54]. This observation is particularly intriguing, since Buchnera is a non-motile bacterium which has reduced its genome by eliminating energetically expensive, unnecessary or redundant genes [27]. A consensus hypothesis is that flagellar genes are in fact involved in an essential symbiotic function, the export of protein to the host [6,54,65], as is known to be the case in Salmonella typhimurium and Yersinia enterocolitica [66,67]. The high level of expression of most of these genes, their amino acid sequence divergence and the particular subset of flagellar genes conserved among the 4 sequenced Buchnera strains preserving only a minimal type-III virulence secretion system, support this hypothesis [54].

In most bacteria, selection for gene expressiveness is reflected in codon usage. Codon usage in Buchnera is dominated by an AT bias. Indeed most preferred codons end with A and T in Buchnera [68]. However, a significant correlation between tRNA-isoacceptor abundances and codon usage has been shown in Buchnera, highlighting selection for gene expressiveness [59]. Codon usage was observed thanks to the true set (measured experimentally) of highly expressed genes in Buchnera. Genes that are highly expressed in Buchnera use more C-ending rare codons, and fewer G-ending codons than more weakly expressed genes (Fig. 4). This tends to favor optimal codons for some specific amino acids in the highly expressed genes of Buchnera, suggesting the existence of a qualitative selection of the translation of certain sites (amino acids) that are functionally important in some of the essential proteins of Buchnera [59]. Recent work by Toft and Fares (2009) [69] has confirmed this hypothesis.

A global shift of gene expression in bacteria is usually governed by a change in DNA topology [70]. In Buchnera, the rates of gene expression along the chromosome were analyzed using autocorrelation models and spectral analysis [54]. Significant periodic components in the gene transcription levels along the chromosome of the BAp genome have been identified: small- (2 to 20 genes), medium- (20 to 70 genes) and long-range (80 to 152 genes), and such periodic signals have been observed in E. coli and Bacillus subtilis [71,72]. In order to determine the role of operons in such periodic structures, gene permutations were introduced in silico into the Buchnera genome. Four simulated permutations were defined as follows: (1) disrupting the gene order between the operons; (2) disrupting the gene order between the operons as well as the number of singletons between them; (3) disrupting the gene order between the operons as well as the operon ranking; and (4) disrupting the order of all the genes on the chromosome. The results obtained after these permutations suggest that operon organization in BAp is partly responsible for small periodic signals and, in addition, operon spacing and ranking contribute to the medium and long-range periods, suggesting that the BAp chromosome has a periodic topological structure [54].

Finally, a global analysis of the putative transcriptional regulators conserved in Buchnera was carried out in order to reconstruct the genetic network of the bacterium. Eight proteins, orthologous with the transcription factors of E. coli, were identified in Buchnera. This demonstrates that the genetic network of Buchnera has undergone extreme shrinkage (Fig. 5) conserving generic regulators that are probably only sensitive to major stress within the cell, where specific tuning is no longer necessary. In fact, five of the conserved transcription regulators belong to the NAP family (Nucleoid Associated Protein): (1) H-NS, histone-like nucleoid structuring protein; (2) HU, heat-unstable nucleoid protein; (3) IHF, integration host factor; (4) DnaA, DNA-binding protein A; and (5) FIS, factor for inversion stimulation. Bioinformatic structural analysis (unpublished data) reveals that these five NAPs display high levels of conservation of important amino acids and functional domains, indicating that this function has undergone severe purifying selection.

The conservation of a significant proportion of the NAPs set commonly found in bacteria (including both specific and non-specific-DNA binding NAPs) offers an especially intriguing potential explanation for gene expression shifts in BAp via DNA topology variations. This is that the combined activities of NAPs and topoisomerases shape DNA topology, which is known to act like a transcription factor in regulating gene expression [70]. NAPs are also involved in the organization of chromosome macro-domains via their involvement in the formation of a dynamic topological barrier [73], and so may contribute to the medium- and long-range periodic gene expression profiles observed in BAp.