Xylella Genomics and Bacterial Pathogenicity to Plants

Xylella fastidiosa, a pathogen of citrus, is the first plant pathogenic bacterium for which the complete genome sequence has been published. Inspection of the sequence reveals high relatedness to many genes of other pathogens, notably Xanthomonas campestris. Based on this, we suggest that Xylella possesses certain easily testable properties that contribute to pathogenicity. We also present some general considerations for deriving information on pathogenicity from bacterial genomics.


Introduction
Since the publication of the complete genome sequence of Haemophilus in¯uenzae in 1987, rapid progress has been made in bacterial genomics and today approximately 100 bacterial genomes have been or are about to be completed. Many of the organisms chosen are human or animal pathogens. It is not surprising that funding agencies have supported work on human pathogens because functional genomics provides new strategies for the discovery of urgently needed antibacterial therapeutic agents. Despite the obvious importance of human pathogens, it should not be forgotten that plant pathogens can also be indirect agents of human misery. It is generally believed that, globally, up to 20% of potential crop yields are lost because of disease. Bacterial plant diseases are most severe in the tropics, where their effects can be locally catastrophic. Apart from exploitation of genetically resistant crops (if they exist), few satisfactory control measures are available. The need to devise new control strategies based on understanding pathogenicity is a compelling argument for supporting sequencing projects. The publication of the ®rst complete genome sequence of a bacterial phytopathogen, Xylella fastidiosa (Xf), by a consortium of researchers in Brazil [39] is therefore to be welcomed. The choice of organism is interesting. It appears to have been driven by economic considerations: Xf is a major pathogen of citrus trees in Brazil. As its name suggests, Xf is a dif®cult microorganism to handle. It has hitherto been largely ignored by molecular geneticists, who have preferred to work with more tractable genera, such as Erwinia, Pseudomonas, Ralstonia and Xanthomonas. The genomes of members of these genera are typically twice the size of the Xf genome.
Analysis of the Xf genomic sequence has revealed a number of genes whose products have very high amino acid sequence relatedness to gene products in Xanthomonas spp., in particular Xanthomonas campestris pv. campestris (hereafter called X. c. campestris). Many of these Xanthomonas genes have established roles either in virulence or in the biosynthesis of the extracellular polysaccharide xanthan, which as well as being an important industrial product is believed to be a virulence determinant. To some extent amino acid sequence relatedness between Xf and Xanthomonas is not particularly surprising, since these bacterial species are closely related [46]. However, the high degree of relatedness strongly suggests that the Xanthomonas and Xf genes are orthologous. In contrast, the relative size of the genomes of Xf (y2.7 Mb) and Xanthomonas (y5.5 Mb) indicates that there are distinct differences in the complement of genes carried by the two bacteria. Xf may represent a `minimal' plant pathogen. In addition to its plant hosts, Xf can colonize tissues of its sharpshooter insect vector. However, it is not pathogenic to the insects and thus differs from Pseudomonas aeruginosa, certain strains of which are polyphyletic pathogens [12,32] able to cause disease in mammals, plants and invertebrates. Although this paper only discusses plant pathogenesis, studies of Xf may also reveal determinants of bacteria±insect interaction speci®city.
The differences in genome size between Xf and X. c. campestris probably re¯ect the different disease strategies and growth capabilities of the two organisms. Xf depends on leafhoppers for its transmission and in plants is limited to the xylem. X. c. campestris can infect intact plants and is capable of growth and survival on plant leaf surfaces as well as in soil. In the following sections we highlight some of the similarities (and differences) between gene products and gene organization in Xanthomonas and Xf that may inform studies of Xf pathogenesis. We also consider general strategies for the functional genomics of plant pathogens.
Xylella has no type III secretion system A principal difference between Xf and Xanthomonas spp. is the absence in Xf of a type III secretion system, which in other plant pathogenic bacteria is encoded by genes within hrp gene clusters. Type III secretion systems are widely distributed in phytopathogenic bacteria, where they function to deliver effectors of disease into plant cells (for recent reviews, see [16,23]). Some of these virulence determinants may act to suppress plant defence responses and hence promote disease, whereas others may act to promote the release of nutrients from the plant cell. In some cases, however, secreted effectors can be recognized by the plant to trigger plant defence. In this case they are considered to be avirulence determinants. Examination of the Xf genome revealed no gene product with amino acid sequence relatedness to the products of known avirulence genes in other plant pathogenic bacteria. The lack of dependence of Xf pathogenesis on a type III secretion system may re¯ect the insectmediated transmission, vascular restriction and slow growth of the bacterium. For example, pathogenesis by Xf may not require the suppression of speci®c processes within living plant cells. It should be noted, however, that the bacterium may be able to kill plant cells through the action of plant cell wall degrading enzymes (see below) and haemolysin-like toxins. In contrast to the situation in Xf, type III secretion systems are required for the pathogenesis of X. c. campestris and X. oryzae pv. oryzae [1,49], vascular pathogens that colonize the xylem elements in crucifers and rice, respectively. These bacteria enter intact plants through hydathodes, structures located at the leaf margin at the ends of the veins. Having colonized the hydathodes, the bacteria gain access to the xylem elements, where they proliferate. At later stages of disease, bacteria can be released from the vascular system to invade the surrounding mesophyll tissue. The type III-secreted effectors could clearly have a role in each of these phases of the disease cycle.
Type II protein secretion in Xylella and Xanthomonas X. c. campestris synthesizes a number of extracellular enzymes that are capable of degrading components of the plant extracellular matrix (cell wall), including pectin, cellulose, proteins and glycoproteins [8]. The export of these extracellular enzymes (and perhaps other proteins) from the bacterial cell is achieved via a type II secretion system encoded by the cluster of 11 genes, xpsEF-GHIJKLMND [11,25]. Type II secretion is signal sequence-dependent, the pathway is the main terminal branch of the general secretion pathway [31,34]. The type II secretion pathway is required for virulence in both X. c. campestris and X. o. oryzae [9,33]. The same gene organization as found in X. c. campestris occurs in Xf and most of the gene products (XF1517±1527) have very high amino acid sequence relatedness to their Xanthomonas counterparts. The Xf type II secretion system is probably involved in the export of cell wall degrading enzymes, including endoglucanase (XF0810, XF0818, XF2708), polygalacturonase (XF2466) and perhaps several proteases (XF2330, XF0267, XF1851, XF0816). The occurrence of several isoforms of endoglucanase and protease is consistent with observations of many plant pathogenic microorganisms. One possible function of these enzymes in Xf pathogenesis is to degrade the pit membranes, allowing bacteria to move into previously uncolonized xylem vessels. Other roles may be to mobilize cell walls for nutritional purposes or to overcome plant defences. The putative polygalacturonase precursor from Xf is highly related to the enzyme from Ralstonia solanaceraum, the causal agent of bacterial wilt. The polygalacturonase from R. solanacearum is required for the full virulence of this vascular pathogen to tomato [37].

Extracellular polysaccharides
Many plant pathogenic bacteria produce extracellular polysaccharides that contribute to the ability of these organisms to cause disease on plants [7]. In X. c. campestris, the extracellular polysaccharide is xanthan, which is also an important commercial product. Xanthan is a polymer of repeating pentasaccharide units with the structure mannose-(b-1,4)glucuronic acid-(b-1,2)-mannose-(a-1,3)-cellobiose [42]. The pentasaccharide units are derivatized with acetyl and pyruvyl moieties. Biosynthesis of xanthan is believed to occur in at least two stages. In the ®rst, the repeating unit is sequentially assembled linked to a polyprenol through a diphosphate bridge. In the second stage, the repeating units are polymerized and the polymer liberated to the growth medium. The genes that encode the enzymes involved in the transfer of the sugars and the non-glycosidic constituents are located in a cluster that comprises 12 predicted open reading frames, gumB to gumM. The function of some of the gene products has been established, whereas others have only been suggested [18,21,26,44]. GumD, GumH, GumI, GumK and GumM are known to be involved in the assembly of the pentasaccharide lipid intermediate. Gum L is the pyruvate ketal transferase. GumB, GumC, GumE and GumJ are speculated to have a role in the polymerization and translocation of xanthan. The Xf genome contains genes with highly related products with the gene order gumBCDEFHJKM (XF2370±XF2360). Two short ORFs (XF2363 and XF2368) are also present in this region. However, gumI (encoding the transferase which incorporates the terminal mannose), gumL and gumG (encoding one of two acetyl transferases) were missing. BLAST searches of the Xf genome revealed no other gene products with amino acid sequence relatedness to GumL or GumI. These results suggest that Xf may make a modi®ed polysaccharide that lacks the terminal mannose residues and is not pyruvylated (Figure 1), although this has not been experimentally determined. Mutants of X. c. campestris carrying non-polar insertions in gumI are also able to produce a polytetrasaccharide, although only to 10% of the amount of polymer produced by the wild-type [21]. Signi®cantly, gumI Figure 1. The structure of the repeating pentasaccharide unit of xanthan of Xanthomonas campestris together with the proposed functions of the gum gene products involved its assembly, decoration with acetyl and pyruvyl residues, polymerization and export from the bacterial cell. Dashed lines indicate that the repeating unit is variably decorated; the terminal mannose residue carries either an acetyl or pyruvyl moiety. Abbreviations: GT, glycosyl transferase; AT, acetyl transferase; KPT, ketal pyruvate transferase. Those gum gene products that are common to Xylella and Xanthomonas are indicated in red; those that are only found in Xanthomonas are shown in blue mutants of X. c. campestris are not altered in their virulence when inoculated into the petioles of cabbage plants [21]. Mutants of X. c. campestris with non-polar insertions into gumK can synthesize a lipid-linked trisaccharide in vitro but produce only very low amounts of a polytrisaccharide polymer in vivo [21]. These modi®ed polysaccharides may have particular properties that make them useful in biotechnological or other industrial applications [42]. One limitation of the study and exploitation of some of these modi®ed polymers is the low level of production, which may be in part a consequence of the substrate speci®city of the polymerization/ export apparatus for the different lipid-linked oligosaccharides. The apparent natural variation in the polysaccharides produced by Xf and X. c. campestris may re¯ect differences in the substrate speci®cities for lipid±oligosaccharide precursors that could be exploited to elevate the production of modi®ed xanthans in X. c. campestris.

The regulation of the synthesis of virulence determinants
The production of extracellular enzymes and extracellular polysaccharide by X. c. campestris is strictly regulated both during growth in liquid media and during disease. Regulation probably occurs as an adaptation to particular environmental changes. Work in our own laboratory has identi®ed a cluster of genes that act to regulate the synthesis of these virulence factors. This cluster, which we have called rpf (for regulation of pathogenicity factors), comprises nine genes (rpfA-I) and is located within a 21.9 kb region of the X. c. campestris chromosome [2,10,43,48]. The left-hand part of this region of the chromosome comprises six contiguous rpf genes with the gene order rpfABFCHG. Transposon insertion in any of these genes leads to coordinate downregulation of synthesis of all extracellular enzymes and EPS. The remaining rpf genes (rpfD, rpfE, rpfI), which are located to the right of rpfA-G, are grouped with genes of diverse function. Transposon insertions in these latter rpf genes lead to only minor effects (rpfD) or to complex effects on only certain enzymes (rpfE, rpfI).
In some cases, sequence analysis of the rpf genes and/or biochemical studies of the rpf mutants have given a guide to possible function of the gene products. rpfA encodes the major aconitase of X. c.
campestris and is implicated in iron homeostasis [48]. rpfB encodes a long chain fatty acyl CoA ligase and, together with rpfF, is involved in regulation mediated by a small diffusible molecule, which we have called DSF (for diffusible signal factor) [2]. Neither rpfF nor rpfB mutants are able to make DSF, although rpfF mutants can be phenotypically corrected for the production of extracellular enzymes and EPS by the exogenous addition of DSF or by growth on plates in proximity to wildtype strains. RpfF has limited amino acid sequence relatedness to enoyl CoA hydratase and we have speculated that RpfF and RpfB are involved in diverting intermediates of lipid metabolism to DSF production [2].
rpfC encodes a hybrid two-component regulator, containing both sensor kinase and response regulator domains, and contains an additional Cterminal phosphorelay (HPt) domain [40,43]. rpfC is in an operon with rpfH and rpfG [40]. The predicted protein RpfG has a regulatory input domain attached to a specialized version of an HD domain, suggested previously to function in signal transduction [17]. The predicted protein RpfH is structurally related to the sensory input domain of RpfC. RpfG, RpfH and RpfC are proposed to participate in a signal transduction system linking perception of environmental signals, including perhaps DSF itself, to the activation of pathogenicity gene expression and to the negative regulation of DSF synthesis [40]. The predicted proteins, RpfD, RpfE and RpfI, show the highest amino acid sequence relatedness to hypothetical proteins from Caulobacter crescentus, Bordetella pertussis and Klebsiella pneumoniae, respectively [10].
Remarkable synteny exists between the region of the chromosome of X. c. campestris that contains the rpf gene cluster and that region of the Xf genome encoding products XF1108±XF1115 ( Figure 2). Furthermore, the amino acid sequence similarities between all of the genes which are represented in both chromosomes is very high; in BLASTP comparisons E values were all less than 10 x99 and homologies exist throughout the length of the proteins. Genes encoding RpfA (XF0290) and RpfB (XF0287), which are immediately to left of rpfF in X. c. campestris, are located in a different part of the Xf genome although they are closely linked. In contrast, BLASTP searches revealed no gene products in Xf with signi®cant amino acid sequence similarity to RpfD, Orf1, Orf2, Orf3 or RpfI.

Intercellular signalling
The extremely high degree of sequence relatedness between gene products implicated in the synthesis and perception of DSF in X. c. campestris and predicted gene products from Xf provides strong circumstantial evidence for the existence of a DSF regulatory system in Xf. It is increasingly evident that cell±cell signalling mediated by diffusible molecules plays a large role in regulating diverse physiological processes, including virulence to plants and animals, across distant genera of bacteria [20,22,35]. Notably, cell±cell signals enable bacteria to behave in a group fashion, coordinating changes in gene expression in response to the changing environment and growing cell population, a phenomenon which has been called`quorum sensing'. The most common signalling molecules found among Gram-negative bacteria, including plant pathogens, are N-acyl derivatives of homoserine lactone (N-AHLs) but other molecules such as lipids, peptides and c-butyrolactones are also found [14,15,35]. X. c. campestris has apparently evolved a novel system for cell±cell signalling. DSF is not an N-AHL and a recent survey has indicated that production of compounds with N-AHL activity is very rare in Xanthomonas species [4]. A second low molecular weight signalling molecule in X. c. campestris that has overlapping functions with DSF has been tentatively characterized as a butyrolactone [5,28,29,30].
Proteins of the LuxI and LuxR families are involved in the synthesis of N-AHLs and in quorum sensing-mediated gene regulation, respectively, in many bacteria [15]. Homology searches (BLASTP) with LuxI family members from a number of bacterial species, including R. solanacearum and Erwinia spp., did not reveal any signi®cant hits in Xf. There were also no gene products with signi®cant sequence relatedness to a number of proteins unrelated to LuxI implicated in the synthesis of autoinducers or diffusible signal molecules in several other bacteria. These included LuxM and LuxL, which are implicated in the synthesis of N-AHLs in Vibrio harveyi [3], LuxS, which is responsible for autoinducer synthesis in Escherichia coli and Salmonella typhimurium [41], and PhcB, which is responsible for the synthesis of hydroxypalmitic acid methyl ester, a regulator of synthesis of extracellular enzymes and extracellular polysaccharide in R. solanacearum [13,36].
What is the physiological state of xyleminfecting bacteria?
Electron micrographs of infected plants often show high local concentrations of bacteria in the tissues. Micrographs of Xf-infected citrus can be seen on the Brazilian group's web site (http://onsona.lbi.ic. unicamp.br/xf/). There are many similar examples in the literature, including a study of cabbage infected with X. c. campestris [45]. The implications of such pictures are often overlooked. We calculate that the concentration of X. c. campestris in xylem vessels is at least 10 13 cells/ml; for comparison, a dense stationary phase laboratory broth culture contains about 10 9 cells/ml. It is instructive to consider the bacteria as growing in a continuous-¯ow system such as a chemostat, fed by the¯ow of xylem sap. The¯ow rate of nutrients through xylem vessels can be estimated [19,38] and, making the most optimistic assumptions about the ability of bacteria to sequester nutrients from sap, it is clear that the rate of increase of bacterial biomass must be very small, much less than one doubling in 24 h. The conclu- Figure 2. Synteny between the rpf gene clusters of Xylella fastidiosa and Xanthomonas campestris. The rpf genes of X. c. campestris, which regulate the synthesis of pathogenicity factors, are clustered in a region of the chromosome that also contains genes with broader regulatory functions and others of diverse function. Rec J is a single-stranded DNA-speci®c exonuclease, GreA is a transcriptional elongation factor, PrfB is peptide release factor 2, LysU is lysyl tRNA synthetase. The dashed line indicates that recJ and prfB are contiguous in the Xf genome sion is that bacteria in infected plants are subject to severe nutritional stress and hence mechanisms employed by bacteria to respond to and protect against physiological stress will be important in pathogenicity. The processes by which bacteria accumulate to such high levels merit further research.

Strategies for functional genomics of plant pathogens
Although Xf is the only published phytopathogen genome sequence, several others are in progress. These include Xanthomonas axonopodis pv. citri (Brazil, by the consortium responsible for Xf), X. c. campestris (China), X. o. oryzae (Japan), R. solanacearum (France), P. syringae pv. tomato (USA), E. carotovora (Finland) and Agrobacterium tumefaciens (USA). This set includes examples of the major taxonomic, physiological and economically important pathogens. It is timely to consider how functional studies could be implemented to yield information on pathogenicity mechanisms. The Brazilian consortium has already established a programme for functional analysis of Xf (http:// www.lbm.fcav.unesp/br/fun/). The portfolio of projects highlights some of the disadvantages of using an`unpopular' organism such as Xf. Many of the necessary techniques, such as gene disruption, that have been used routinely for other pathogens for many years, will have to be developed from scratch. Methodology used for studying bacterial pathogenicity is described in two books [24,47]; here we brie¯y consider approaches to evaluating the role of sequenced genes in bacteria±plant interactions.

Gene expression
The concept that some`pathogenicity' genes might show differential expression in X. c. campestris inhabiting host tissues was proposed in 1987 [27] and the approach used was the forerunner of`in vivo expression technology' (IVET). Comparisons of bacterial gene expression levels have mainly used gene fusion technology. However, the availability of complete sequences of organisms invites the use of microarrays for genome-wide surveys. It is not yet clear what technical obstacles will have to be surmounted in order to measure bacterial transcript levels directly in infected plants. Experimental manipulation of expression levels can yield useful information on gene function. Apart from relatively crude methods, such as multicopy cloning, no general methods are available for in¯uencing bacterial gene expression in infected plants.

Gene disruption
Procedures for gene disruption based on either direct transposon mutagenesis or recombinational exchange between the genome and suitably altered cloned sequences have been developed for all the major genera of plant pathogens. The use of marked and unmarked and in-frame deletions is particularly informative.

Phenotype assessment
The`pre-genomic era' of molecular phytobacteriology relied heavily on isolation of mutants that produced altered disease symptoms. The need to screen large numbers of individuals dictated the use of simple, rapid and unsophisticated techniques for testing pathogenicity. The disadvantage of simple methods is that the results of pathogenicity tests may depend on how the inoculation of plants is carried out [6]. As an example, rpfC mutants of X. c. campestris produce symptoms indistinguishable from the wild-type if in®ltrated directly into leaf panels, but inoculation into vein endings (which more closely resembles the natural infection route) shows them to have much reduced virulence [40]. The feasibility constraints imposed by mass screening programmes no longer apply when evaluating the phenotype of speci®c mutants and more re®ned and time-consuming methods become acceptable. Based on experience with many classes of X. c. campestris mutant, we suggest the following set of phenotypic tests of mutants as a starting point for evaluating the role of a gene in plant±bacterial interactions: (i) Cultural characteristics (auxotrophy, quantitative growth properties, sensitivity to physicochemical factors, substrate utilization, motility and chemotaxis). (ii) Production and/or secretion of factors considered to be candidate pathogenicity factors (extracellular enzymes, polysaccharides, toxins From genomics to disease control? In many discussions on the use of genomic information for the development of therapeutic agents against human pathogens, a ®ve-step strategy has been proposed: In principle, an analogous strategy could be adopted for plant pathogens. Chemical control of fungal diseases has been highly effective but, for a variety of reasons, bacterial diseases of plants have, historically, not been satisfactorily controlled by chemicals (an exception is those pathogens, such as Xf, which need insect vectors and for which insecticides may disrupt the disease cycle). For other bacteria, the most effective control measure, apart from use of good agronomic practices, is the deployment of genetic resistance to pathogens. At present the search for new sources of genetic resistance to incorporate into crop breeding programmes requires large-scale glasshouse and ®eld inoculation with bacteria and measurement of disease development. It is to be expected that functional genomics will lead to the identi®cation of previously unsuspected bacterial virulence factors and mechanisms. This should in turn permit the development of biochemical probes and assays to reveal natural genetic variation in susceptibility of targets in plants or to guide the manipulation of targets in the laboratory.