Conservation of the links between gene transcription and chromosomal organization in the highly reduced genome of Buchnera aphidicola

Past genomic studies have comprehensively described the organization of the bacterial chromosome, for example in terms of gene localization, order and orientation. The degree of organization has been shown to increase with genome size, overall GC composition and the presence of nucleoid-binding proteins [1]. This organization of the chromosome can be described as being an adaptive and functional tool, essential for the survival of the bacterial cell. More precisely, several studies have identified strand asymmetries in the distribution of genes between the leading and the lagging strand of DNA (for a review see Rocha [2]). Indeed, bacterial genomes carry, on average, from 78% (for genomes containing the polymerase PolC) to 58% (for the other genomes) of their genes on the leading strand [3]. This bias is even more important when the essentiality of genes is taken into account, and essential genes distribution bias reaches 76% and 94% in Escherichia coli and Bacillus subtilis respectively [4]. The asymmetry of the distribution of the genes between the two strands of DNA is explained as being a means of minimizing interruptions in gene transcription due to collisions between the DNA and RNA polymerases. The most generally accepted hypothesis is that co-directional collisions on the leading strand have a weaker effect on RNA polymerase processivity than the head-on collisions occurring on the lagging strand. This "replicational selection" should guarantee that genes on the leading strand, particularly the essential ones, are efficiently transcribed [5–8].

Gene organization in the bacterial chromosome has very well known effects on transcription and a standard example is the organization of genes into operons, which allows for a sophisticated regulation of gene expression [9]. It has also been shown that neighbouring genes in a bacterial chromosome tend to be co-expressed, even if they are not in the same operon [10], suggesting the existence of a "supra-operonic" organization [11]. This is a perfect illustration of the mutual influence between gene expression levels and chromosomal organization in bacteria. Another example is the significant effect of DNA supercoiling on transcription. Recent studies have suggested that the dynamical structure of the nucleoid acts as a "transcription factor" in E. coli [12], and is probably responsible for the presence of spatial transcriptional patterns in free-living bacteria [13–15].

While these observations were made on free-living bacteria, very few data exist on symbiotic bacteria characterized by reduced genomes, shaped by their adaptation to the host metabolic requirements and by their high evolution rate. Buchnera aphidicola, the endosymbiont of the aphids, is one of the best studied examples of an intracellular bacterium with a reduced genome [16, 17]. The genomes of four Buchnera from different aphid species have been sequenced so far, with sizes ranging from 0.42 to 0.65 Mb [18–21]. In all the Buchnera genomes, most of the genes involved in the biosynthesis of essential amino acids (EAAs) that the insect cannot synthesize, or find in sufficient amounts in the phloem sap, were retained, whereas almost all genes regulating their expression were lost [22]. In Buchnera from the pea aphid Acyrthosiphon pisum (BAp), among the 608 chromosomal genes, 56% are located on the leading strand DNA. This observation reveals a small gene-strand bias, although equivalent to that of its closely related free-living bacteria:E. coli (55%) [23].

Gene essentiality has been defined by experimental techniques in cultivable bacteria, such as E. coli for which a large repertoire of null mutants is available. For a small number of genes, even in the absence of mutants, essentiality is defined on the basis of their function. For example, genes encoding ribosomal proteins are usually described as essential, whereas genes involved in the flagellar apparatus are classified as non-essential in bacteria [24]. Due to the specific lifestyle and symbiotic functions of Buchnera, the definition of essentials genes for this bacterium is particularly difficult. For this reason, we considered the minimal gene set for supporting bacterial life proposed by Gil et al. [25] as the most appropriate for our study. Indeed, the determination of this set is based on comparative analyses of five bacterial insect endosymbionts with small genomes (among them three Buchnera genomes), one of the smallest bacterial genome sequenced (Mycoplasma genitalium), and the essential gene lists available for B. subtilis and E. coli in the literature. Based on this minimal set of essential genes, BAp harbour 190 essential encoding-protein genes and 60% of them are located on the leading strand. This indicates a strand bias for essentiality, albeit lower than that observed in free-living bacteria. Recent transcriptomic analyses have shown that, despite a low response to different physiological conditions imposed on the aphid host, Buchnera retained the capability to express genes at different levels in basal conditions. This expression seems to be correlated to genome organization in the putative transcription units (pTU), even in the absence of many of the known transcriptional regulatory proteins [26].

This work focuses on studying the relationships between gene transcription and chromosome organization in the highly reduced genome of Buchnera, by coupling genomic information with transcriptomic data. For that, mRNA abundances were measured with a full-genome oligo-array dedicated to BAp. The basal transcriptome of Buchnera was then analysed regarding (i) the location of the genes on the leading/lagging strand, the gene essentiality and operon organization, (ii) the gene evolution rate (as measured by the GC rate), and (iii) the spatial location of genes along the chromosome. One crucial point in this study was the choice of the data normalization procedure. Since 2002, authors have described the possible advantages of genomic DNA (gDNA) standard for microarray data normalization in the investigation of microbial genomes [27–29]. This useful attractive standard is readily available, inexpensive, and invariant over time and from laboratory to laboratory. Moreover, the main advantage of the normalization by gDNA is that it takes into account the probe/target affinity in order to allow for inter-gene calibration and expression level comparisons. For these reasons, we have developed and validated in this study a "spot by spot" microarray data normalization method based on gDNA.

To our knowledge, we have presented here the first analysis showing i) a representation of whole-genome transcriptional data in Buchnera and ii) the links between mRNA abundances and chromosomal organization in a highly reduced bacterial genome, on the basis of experimental data. The large number of prokaryotic genome sequences available in databases has made it possible to study these links in many bacterial chromosomes. Using codon adaptation index (CAI) values computed from ribosomal proteins, Rocha and Danchin have shown that, in B. subtilis and in E. coli, the importance of the expressiveness in determining the localization of the genes on the leading strand is negligible, or even absent, when essentiality is taken into account [36]. They have confirmed these results for other sequenced genomes, with the exception of some non-free-living bacteria among which Buchnera was one of the most important [4]. However, the authors underlined the difficulty of obtaining correct CAI values in the genomes of intracellular bacteria, which generally do not show sufficient codon usage bias. Moreover, another explanation of the exceptions they found for Buchnera can be attributed to authors' assignment of gene essentiality, which was exclusively based on the homology with E. coli. Indeed, while Rocha et al. [4] found no essential genes strand distribution bias in Buchnera, we found a bias of 60% using the minimal gene set proposed by Gil et al. [25]. More recently, Price et al. [37] have revisited the question of gene-strand bias in bacteria, using gene expression microarray data, and they have shown that, in B. subtilis and E. coli, the genes in operons on the leading strand DNA are more highly expressed than genes in operons on the lagging strand. This observation was true for both essential and non-essential genes.

In Buchnera, we found that highly expressed genes are generally the essential genes within pTU. Independently of the essentiality factor, the genes within pTU are more highly expressed than singletons. We also found that for genes within pTU the essential genes are more highly expressed than non essential ones whereas this effect was not observed for singletons. These results underline the conservation of a coherent relationship between mRNA abundances and gene essential functions in the reduced genome of Buchnera.

In this work we also analysed the relationship between mRNA abundances and genes GC ratio in Buchnera (taken as an indicator of gene evolution rate). By combining transcription analysis, evolution rates and comparative genomics, we were able to define new candidates for the essential gene set of Buchnera. In bacteria, genes encoding ribosomal, cell division and chaperone/protease proteins are considered as essential and they are also known to be constitutively highly expressed. As we expected, our data showed that almost all genes encoding ribosomal proteins, and genes encoding chaperonins, are relatively well conserved and also highly expressed in Buchnera. Interestingly, among the most highly conserved and expressed genes, we found genes involved in the biosynthesis of EAAs and, in particular, all the genes of the isoleucine and valine pathway. Also remarkable is that the genes encoding the aminoacyl-tRNA synthetases for 8 out of the 10 aphid EAAs are weakly expressed, whereas genes encoding aminoacyl-tRNA synthetases for non-EAAs are either moderately or highly expressed. A possible explanation for this observation is that Buchnera, which is a relatively slow-growing bacterium, does not necessitate high rates of protein production and constitutively synthesizes EAAs in order to furnish them to the aphid. Reducing the abundance of specific aminoacyl-tRNA synthetases might increase the concentration of free EAAs in Buchnera cells facilitating the transport of these amino acids to the aphid host cell. This observation is reminiscent of a similar putative adaptive response of Buchnera, which selectively underexpresses pheT under aromatic EEAs shortage [26].

In our study, we also found that the orphan genes yba3 and yba4 seem to evolve rapidly and are highly expressed, yba4 being one of the ten highest expressed genes in Buchnera. The conservation of these genes in BAp and in the Buchnera harboured by the aphid Schizaphis graminum (BSg), coupled with their high expression level, suggest that they could be of particular relevance in the symbiosis of Buchnera with the aphids S. graminum and A. pisum, as it has also been proposed by Shimomura et al. [38] for A. pisum.

Among the genes rapidly evolving (this work, Tamas et al. [19] and Reymond et al. [26]) and highly expressed in Buchnera, we found most of the flagellar genes. These data, taken together with recent experimental evidence that the Buchnera incomplete flagellar apparatus can function as a "protein transporter" [39], support the idea that flagellar genes are taking on new important functions in the symbiotic context. The flagellar gene set of BAp is composed of 26 genes of which fliEFGHIJKMNPQR and flhAB genes are located on the leading strand (except fliE), and flgNABCDEFGHIJK genes are located on the lagging strand DNA (except flgN and flgA). These flagellar genes are also conserved in BSg. However, Buchnera from a distant aphid lineage, Baizongia pistaciae (BBp), has lost five flagellar genes (flgA, flgD, flgE, flgK, flgN) but not any of the fli ones. Finally, Buchnera with the most dramatically reduced genome, from the aphid Cinara cedri (BCc), has lost the same genes as BBp but also four other flg genes (flgB,flgC, flgG, flgJ) and four fli genes (fliE, fliJ, fliK, fliM), hence preserving only a minimal type III virulence secretion system [21]. The evolutionary selection of the majority of the fli genes also suggests the possible importance of the new putative transport function of these genes in Buchnera. It is to note, that the most conserved fli genes among the Buchnera lineages, probably involved in the new function, are located on the leading strand, whereas most of the gene losses occurred on the lagging strand. Previous comparative genomics analyses [40, 41] had tempted to dissect the evolutionary forces driving the genome organization in several Buchnera lineages (i.e., gene strand bias, gene and protein composition, gene expressivity, gene evolution rate and gene loss). Our results are partly consistent with these previous analyses as we found (1) highly conserved genes are highly expressed, (2) essential genes in pTU are highly expressed and probably preserved from mutations by purifying selection, and (3) positive selection may shape new "symbiotic" functions for some genes highly expressed and highly evolving. However, we reject the former idea, based on CAI analyses, that expressiveness is a factor driving gene strand bias in Buchnera.

Finally, an important result of this study was the discovery of spatial patterns of transcriptional activity in the chromosome of Buchnera: i.e., the transcription of the genes along the chromosome is determined according to spatial constraints. From autocorrelation and spectral analysis, four groups of spatial patterns can be defined: (i) autocorrelated short-range (between 2 and 8 genes), (ii) periodic short-range (up to 17 genes), (iii) periodic medium-range (between 23 and 61 genes) and (iv) periodic long-range (over 87 genes) structural components.

Autocorrelated short-range patterns, determined by the autocorrelation function of gene transcription levels, showed that genes spaced by a gene-to-gene distance of less than 8 have highly correlated expressions. As has been suggested for E. coli and B. subtilis, we propose for Buchnera that these correlations reflect the co-ordinated transcription of genes within operons [14, 15]. This observation reinforces the result mentioned above concerning the conservation of functional transcription units in Buchnera. However, by permuting gene position along the chromosome, we observed that the organization of genes into putative operons is not sufficient to fully explain the observed periodic spatial patterns of transcription. Indeed, if the presence of spatial periodic components in Buchnera gene expression was only due to the conservation of operon structures, the modification of the order and/or the number of the singletons located between the pTU should not affect the periodicities and the autocorrelation values. We have shown, however, that these modifications reduced the spatial patterns, alleviating the importance of a high-order arrangement of all the genes along the chromosome of the endosymbiont.

Jeong et al. [14] have classified the transcriptional periods that they have found in E. coli into three categories: short-range (up to 16 genes); medium-range (100–125 genes); and long-range (600–800 genes). The existence of short periods, up to 17 genes in Buchnera, allows us to corroborate the hypothesis, proposed in a previous study on B. subtilis and on E. coli, that this short range element could be a property of the structural nucleoid common to other bacteria, corresponding to large DNA spirals on the nucleoid surface [15]. The medium and the long-range periods are shorter in Buchnera than those identified for free-living bacteria, which is probably due to the greatly reduced size of its genome. However, these two kinds of periods are not yet understood and do not correspond to the domains identified so far in the nucleoid [15]. The effect of the second and third simulated gene permutations (Perm. 2 and Perm. 3 on Figure 6) on the medium and long-range periods respectively, could be explained by the importance of the spatial location of the operons along the chromosome and by the neighbourhood of singletons that form "supra-operonic" structures in Buchnera. Moreover, the decrease of the maximum size of the autocorrelated groups of genes (Table 2), for the different permutations of gene positions, corroborates the hypothesis of "supra-operonic" structures in Buchnera. But this speculation needs to be more studied and experimentally confirmed. Finally, the observation of transcriptional periodic patterns, coupled with the conservation in its genome of some Nucleoid Associated Proteins (NAP) such as H-NS, IHF and Fis, suggest that Buchnera has maintained a nucleoid structure responsible for the differences in gene transcription levels in basal conditions. These three NAPs were previously found to be differentially expressed in Buchnera [26] facing nutritional constraints. However, their role in transcriptional regulation remains presently speculative in the aphid endosymbiont.

In conclusion, the analysis of mRNA abundances with regard to chromosomal organization in Buchnera has shown, despite an important reduction in its genome size and an apparent decay of its transcriptional regulatory capacity, a conservation of the relationship between these two parameters in the aphid symbiont. Our work shows that the organization of the genes into operons and their essential functions influence transcription in Buchnera, whereas no significant leading/lagging strand bias was observed. This work also underlines the difficulties in defining gene essentiality in intracellular symbiotic bacteria and the importance of an additional specific gene set for BAp. Finally, we showed the existence of a transcriptional periodicity along the chromosome of Buchnera. From these results, comparative analyses performed on Buchnera taken from other species of aphids, but also on other endosymbiotic bacteria, such as Blochmannia spp. from the carpenter ant or Wigglesworthia spp. from the tsetse fly would establish whether the results presented here are a common trait of the insect endosymbionts.