In their rather cloistered environmental niche inside xylem vessels,
X. albilineans and
X. fastidiosa may have largely avoided surveillance by general and specific plant defense systems. Their lack of a T3SS of the Hrp1 or Hrp2 families may be explained by the fact that
X. albilineans and
X. fastidiosa live and multiply essentially in a dead-cell environment. However, like other bacterial vascular pathogens, they may interact with living xylem parenchyma cells through pit membranes [
25]. If they do, they do not use a Hrp TTSS but another system that remains to be identified. The adaptation of
X. albilineans and
X. fastidiosa to a xylem-limited lifestyle is also illustrated by their enzymes adapted to the use of plant cell breakdown products as carbon sources. The low number of genes unique to both
X. albilineans and
X. fastidiosa (11, see Additional file
4) may be explained by a very early divergence of the two xylem-limited
Xanthomonadaceae lineages, possibly followed by strong selective pressure to adapt to their different biological niches and lifestyles.
X. fastidiosa is vector-transmitted by various xylem sap-feeding insects and is able to colonize many plant species (citrus, wine grape, coffee, alfalfa, peach, plum, almond, elm, maple, pear, etc) (reviewed in [
26]). On the other hand,
X. albilineans is mainly transmitted by mechanical means and is not known to be insect-transmitted, and is able to colonize only sugarcane and few other monocots in the
Poaceae family (reviewed in [
8]).
The genome of
X. albilineans encodes a T3SS that displays similarities with the
Burkholderia pseudomallei bsa T3SS which belongs to the injectisome family SPI-1 (
Salmonella Pathogenicity Island -1) and which is required for the virulence of this human pathogen. The SPI-1 injectisome family mainly includes T3SSs from human and insect bacterial pathogens or symbionts [
21]. Interestingly, the genomes of
Erwinia amylovora strain Ea273 and
Erwinia tasmaniensis strain Et1/99 both contain two copies of a SPI-1 T3SS [
27,
28]. The role of these SPI-1 T3SSs in these plant-invading
Erwinia spp. remains unknown.
E. amylovora is insect-disseminated, although the interactions between this pathogen and its insect hosts remain poorly understood. It was suggested that the presence of a SPI-1 T3SS in
Erwinia spp. indicates a common ancestry and close phylogenetic relationship between
Erwinia spp. and insect-related enteric bacteria, raising the possibility that an insect host might be serving as a mixing vessel for the exchange of genes between
Erwinia strains and other enteric bacteria [
27]. Similarly, the presence of a SPI-1 T3SS in the genome of
X. albilineans could indicate an insect-associated life style of this plant pathogen.
The MLSA performed herein resulted in a phylogenetic tree that included
X. fastidiosa into the
Xanthomonas group. This phylogenetic tree is in accordance with the presence of the unique
gum genes in both
X. fastidiosa and
Xanthomonas species of the Hrp
Xanthomonas group. The
gum genes, which are involved in the biosynthesis of extracellular polysaccharides and the formation of biofilms, play a key role in pathogenicity of these
Xanthomonadaceae. These genes, which were probably acquired by the progenitor of the
Xanthomonas genus, were most likely lost by
X. albilineans and conserved by
X. fastidiosa during their speciation. Our MLSA phylogenetic tree is also in accordance with i/the presence of 11 unique genes, including
metE, in
X. albilineans and
X. fastidiosa, and ii/the alignment of the 5' end of the 16S RNA of
Xanthomonadaceae (Additional file
5).
Additionally, based on this MLSA, the same 480 ancestral genes appeared to be lost by both
X. albilineans and
X. fastidiosa. Interestingly, 209 of the 480 ancestral genes lost by both
X. albilineans and
X. fastidiosa are also absent in
X. oryzae pv.
oryzae (a xylem invading pathogen belonging to another phylogenetic clade), indicating that independent but convergent evolution events were involved in genome erosion of
X. oryzae pv.
oryzae and the xylem-limited
Xanthomonadaceae. Some of these genes lost by three xylem-invading pathogens are orthologous of genes with assigned functions and are organized into clusters. The five following ancestral gene clusters were lost by
X. albilineans,
X. fastidiosa and
X. oryzae pv.
oryzae: i/ the ancestral genes XCV0258 to XCV265 encoding enzymes involved in the glyoxylate cycle; ii/ the ancestral genes XCV0592 to XCV0602 encoding enzymes involved in malonate metabolism; iii/ the ancestral genes XCV1316 to XCV1334 including one TonB-dependant receptor gene, a two component signal transduction system (TCSTS) and chemotaxis genes, iv/ the ancestral genes XCV2187 to XCV2196 including one TCSTS and a type I secretion system and v/ the ancestral genes XCV2796 to XCV2803 encoding enzymes involved in catabolism of polysaccharides (Additional file
1). These examples support the hypothesis of a link between the convergent erosion of three xylem-invading
Xanthomonadaceae and the adaptation to a same restricted environment (the xylem) in which these lost functions are useless. However, only 38 of the 480 ancestral genes lost by both
X. albilineans and
X. fastidiosa are also absent in another xylem invading pathogen,
X. campestris pv.
campestris, indicating that adaptation to xylem lifestyle favoured or allowed genome erosion, but did not necessarily induce it. Alternatively, the convergent genome erosion of the two xylem-limited
Xanthomonadaceae may be linked to similar insect-associated lifestyles that may have favoured genome erosion because most of the genes required for a plant-associated life style are most likely not required for an insect-associated life style.
Similar striking convergence in fundamental genomic features associated with a restricted lifestyle is very well documented for obligate animal symbionts and pathogens, especially for
Buchnera (reviewed in [
29]). In these bacteria, gene losses are non-random but can affect all functional categories. The most dramatic losses affect genes that are involved in metabolism but are not required for survival. Another general feature is the loss of most DNA repair systems and transcriptional regulatory mechanisms, indicating that there is reduced need for transcriptional regulation in a stable environment [
29]. In
X. fastidiosa, and to a lesser extent in
X. albilineans, losses also affected genes involved in metabolism and transcriptional regulatory mechanisms (Figure ). Metabolic capabilities essential for other habitats may have been lost in the genome reduction process coincidently with the adaptation of
X. fastidiosa and
X. albilineans to the nutrient-poor xylem environment. For
X. fastidiosa, genome erosion has been extreme. For example,
X. fastidiosa retained only one transcriptional sigma factor gene and one outer membrane efflux protein
tolC gene, and it lost all genes involved in synthesis of the flagellar apparatus. This extreme erosion allowed
X. fastidiosa to save energy (synthesis and operation of the flagella confer a growth disadvantage of about 2%; [
30]).
In obligate animal symbionts and pathogens, the process of genome shrinkage might have taken place in two separate stages [
29]. A massive gene loss must have occurred soon after the establishment of the obligate symbiosis, probably by means of large deletions that eliminated a series of contiguous genes. The large DNA region containing the flagellar genes was probably lost by
X. fastidiosa during a similar stage. The accumulation of mobile elements, representing a source of chromosomal rearrangements and gene inactivation, seems to have an important role in this first stage. A similar process is likely responsible for the limited genome erosion of
X. oryzae pv.
oryzae, which possesses a very high number of insertion sequences (IS) covering 20% of the genome [
31]. During the second stage of genome reduction in obligate animal symbionts and pathogens, genome shrinkage seems to have mostly occurred through a process of gradual gene loss, scattered along the genome. Such losses seem to follow a pattern that starts with the inactivation of a gene (pseudogenization) by single-nucleotide mutations, and continues with a rapid reduction in length until the original gene is completely eroded [
29,
32]. A similar process is likely responsible for the genome erosions of
X. fastidiosa and
X. albilineans (Additional files
2 and
3). Furthermore, the coding density of
X. fastidiosa strain Temecula1 is significantly smaller than that of xanthomonads probably because of the degradation of ancestral genes. In
X. fastidiosa strain 9a5c, the number of short annotated CDSs is considerably higher than in other
Xanthomonadaceae (Table ), although the functionality of these shortened CDSs, which may result from the degradation of ancestral genes, is questionable.
Obligate animal symbionts and pathogens display rapid evolution and have highly biased nucleotide base compositions with elevated frequencies of adenine and thymine (A+T) (reviewed in [
29]).
X. fastidiosa also displays rapid evolution (note that the length of the branch separating
X. fastidiosa from the ancestor common to
X. albilineans and
X. fastidiosa is much longer, Figure ) and has a high A+T content in comparison with other
Xanthomonadaceae (Table ). Furthermore, the GC skew pattern of the chromosome of
X. fastidiosa has very high amplitude and contains a high number of diagram distortions (Figure ). A similar atypical GC skew pattern was observed for the chromosome of a
Buchnera aphidicola strain [
33]. This latter atypical GC skew coincides with the loss of genes involved in the replication restart process (
recA and
priA) and may be explained by a higher frequency of cytosine deaminations [
34]. The loss of DNA repair genes
recX,
dinG and
dinP may explain, similarly, the very high GC skew of
X. fastidiosa. It may also explain the more extensive genome erosion of
X. fastidiosa, compared to
X. albilineans and
X. oryzae pv.
oryzae. Alternatively, the most important factor affecting genome erosion of
X. fastidiosa may reflect the insect-associated lifestyle specific to this
Xanthomonadaceae [
26].
The GC skew pattern of the
X. albilineans chromosome contains a lower number of distortions and has a significantly higher amplitude than the GC skew pattern of other
Xanthomonas species (Figure ), indicating that no recent events of recombination have occurred in
X. albilineans. Furthermore, the synteny between the chromosomes of
X. albilineans strain GPE PC73 and
X. axonopodis pv.
vesicatoria strain 85-10 also indicated that recombination events were limited during the speciation of
X. albilineans (Additional file
1). The limited recombination of the chromosome of
X. albilineans, its limited erosion, its high G+C content and its low number of IS elements may indicate that a distinctive process was responsible for the reductive genome evolution of this pathogen.
We propose a unique mechanism of genome erosion involving the unique toxin albicidin produced by
X. albilineans. Albicidin is a potent DNA gyrase inhibitor with 50% inhibitory concentrations (40 to 50 nM) lower than those of most quinolones [
9]. DNA gyrase inhibitors block the religation of cleaved DNA intermediate during gyrase catalysis, resulting in lethal double-stranded DNA breaks [
9,
35]. In the presence of subinhibitory doses of DNA gyrase inhibitors, the SOS response mediates survival of the bacteria by allowing DNA replication to continue past breaks that would normally block it. In exchange for this survival advantage, there is an increased mutation rate because the polymerases that perform the repair are prone to error [
36,
37]. Several studies showed that subinhibitory doses of quinolones result in an increased mutation rate in
Escherichia coli,
Staphylococcus aureus,
Pseudomonas aeruginosa and
Mycobacterium tuberculosis [
35,
38-
40].
X. albilineans has two genes conferring resistance to albicidin: an albicidin efflux pump gene that is present in the albicidin biosynthesis gene cluster XALB1 [
20,
41] and an albicidin-resistant DNA gyrase A gene elsewhere on the chromosome. This albicidin-resistant DNA gyrase A is unique to
X. albilineans [
42]. It contains a unique insertion of 43 amino-acids length close to the albicidin binding site. Production of albicidin in ancestral bacteria that possessed both the albicidin biosynthesis gene cluster and a DNA gyrase A sensitive to albicidin may have induced genome erosion. In these ancestral bacteria, most of the albicidin molecules were secreted by the albicidin efflux pump. Occasionally, molecules of albicidin that were not secreted most likely had the same effect as subinhibitory doses of quinolones: the SOS response was induced, thus resulting in DNA repair, recombination and mutagenesis. Successive and cumulative effect of albicidin at each replication cycle eventually resulted in genome erosion. The genome erosion induced by albicidin was likely arrested by evolution of the albicidin-resistant DNA gyrase A.
Acquisition of the albicidin biosynthesis genes by the ancestor of
X. albilineans conferred a selective advantage because of the potent antibiotic activity of albicidin. The DNA damage caused by albicidin may rapidly have induced the mutation of DNA gyrase A gene and thus stopped the process of genome erosion, possibly explaining the distinctive genomic characteristics of
X. albilineans. Albicidin inhibits the growth of
X. axonopodis pv.
vesicatoria (data not shown), suggesting that the DNA gyrase A of the ancestral
Xanthomonas was sensitive to albicidin. Transfer of the albicidin biosynthesis gene cluster to
X. axonopodis pv.
vesicatoria led to production of functional albicidin [
43], demonstrating that the production of albicidin per se is not lethal for a producer that possesses an albicidin-sensitive DNA gyrase A. No remains of the albicidin biosynthesis genes were found in the complete genome sequences of
X. fastidiosa. Therefore, albicidin is most likely not responsible for genome erosion of
X. fastidiosa. However, we cannot exclude the hypothesis that albicidin biosynthesis genes were lost during evolution of
X. fastidiosa. For example, cluster XALB1 could have been lost concurrently with the flagellar biosynthesis gene cluster because these two gene clusters are close on the chromosome of
X. albilineans.
1
This article has been 
