Twenty-two different strains/isolates were collected for this study. Of these, the Xcc strain ATCC33913 is a type strain, isolated from Brussels sprout (Brassica oleracea var. gemmifera) in the UK in 1957 20, and the Xcc strain 8004 is a laboratory strain with spontaneous rifampicin-resistance, derived from Xcc NCPPB No.1145 isolated from cauliflower (B. oleracea var. botrytis) in the UK in 1958 14. The other 20 isolates were collected from different infected cruciferous plants in various geographic locations over a wide range of latitudes across China and named CN01 to CN20 (Table 1). These isolates were validated by morphological, virulent and molecular analyses. All the isolates formed typical X. campestris colonies of yellow mucoid texture on NYG agar medium 14 and caused typical black rot disease symptoms on the host plant radish (Raphanus sativus var. radicula; data not shown). To further confirm the isolates, their partial 16S-23S rDNA intergenic spacer (ITS) regions 41 were examined by PCR and sequencing. A PCR fragment 464 bp in length was obtained for every isolate except CN13 and CN19, for which no PCR product was obtained. Sequencing results showed that five isolates have identical ITS sequences to that of strain 8004, while the ITS sequences of the other 13 isolates differ from that of 8004 by only one or two nucleotides (Additional data files 1 and 2). The isolates CN13 and CN19 were not used for further study in this work as they were not confirmed to be Xcc by the 16S-23S rDNA ITS analysis. The phylogenetic analysis by the maximal parsimony method 42 showed that the 18 proven Xcc isolates were grouped into two clusters and each cluster contains previously identified Xcc strains (Additional data file 2). These two groups were significantly distinguished from other Xanthomonas species and X. campestris pathovars (Additional data file 2), further confirming the 18 isolates as Xcc at the molecular level. The word 'strain' will be used for the identified Xcc 'isolates' hereafter.

The in planta pathogenicity test of Xcc strains was carried out by the leaf-clipping inoculation method on eleven different cultivars (cv.) of four cruciferous species (see Materials and methods). The results showed that seven of the eleven cultivars were susceptible to all of the Xcc strains tested, whereas the other four plants manifested resistance to particular Xcc strains (Tables 1 and 2). Based upon these results, a gene-for-gene relationship governing the outcome of the interactions between the Xcc strains and the host plants could be postulated (Table 3). The key essentials are: first, the host plants that were susceptible to all of the Xcc strains possess no resistance genes against the Xcc strains; second, mustard cv. Guangtou possesses a resistance (R) gene, arbitrarily designated Rc1, for which the postulated interacting avirulence (avr) gene is designated avrRc1, present in strains 8004, ATCC33913, CN03, CN07, CN09, CN10, CN11, and CN20; third, cabbage cv. Jingfeng-1 and radish cv. Huaye possess an R gene named Rc2 that interacts with an avr gene named avrRc2, present in strains ATCC33913, CN03, CN14, CN15, and CN16; and fourth, Chinese cabbage cv. Zhongbai-83 possesses an R gene, Rc3, that interacts with the postulated avrRc3 in strains 8004, ATCC33913, CN02, CN03, CN06, CN07, CN08, CN12, CN14, CN15, CN16, CN18, as well as CN20 (Tables 2 and 3).

We also examined the hypersensitive response (HR) 43 of the Xcc strains on the nonhost pepper ECW10R, a plant commonly used to test the HR of Xcc. The results showed that eight hours after inoculation strains 8004, ATCC33913, CN01, CN03, CN09, CN10, CN11, and CN20 elicited a typical HR while the others did not (Table 2). According to the results, we postulated that strains 8004, ATCC33913, CN01, CN03, CN09, CN10, CN11, and CN20 possess an avirulence gene, designated avrRp1, that interacts with a cognate resistance gene, named Rp1, in the non-host plant pepper ECW10R (Tables 2 and 3).

To investigate genetic similarity and diversity among Xcc strains, a DNA microarray encompassing 4,186 CDSs was constructed, representing all CDSs (non-redundant) in the reference strain 8004 21. Primer design was based on the genomic sequence of 8004, which is composed of 4,273 CDSs 21. Of the 4,186 CDSs, gel electrophoresis revealed successful amplification of 4,043 CDSs, representing 96.58% of the non-redundant genome content. For the CDSs predicted to be less than 100 bp in length, for which optimized primers could not be designed, and those for which PCR amplification did not work, a 70-mer oligo probe for each CDS was designed. The word 'gene' will be used in reference to the CDS that each spot corresponds to unless otherwise indicated.

To determine the sensitivity of our aCGH analysis, self-to-self hybridization was performed using genomic DNA of the reference strain 8004. After removal of faint spots for which the intensity was lower than the average plus two standard deviations of the negative controls (blank spotting solution) on the array, it was found that more than 95% of all genes on the array could be detected and the intensity ratio of the detected genes lay between 0.6 and 1.6. aCGH analyses were then carried out using the reference strain 8004 and its derivative strain C1430nk, described previously 44. The strain C1430nk is derived from 8004 and harbors the cosmid pLAFR6 containing the open reading frames (ORFs) XC1429 and XC1430. The aCGH results revealed that only two genes, XC1429 and XC1430, had an intensity ratio of approximately 1.9-2.4 (C1430nk/8004), indicating that sole copy alteration at the genomic scale could be detected in this study (Figure 1). Based on the above results, it was presumed that the microarray can detect the 1.6-fold alteration when ignoring sequence diversity. After passing the initial tests, aCGH analyses were performed using the fully sequenced Xcc strains 8004 and ATCC33913. The results showed a good agreement with the complete genome sequences of 8004 and ATCC33913 (Figure 1). It was found that for the genes of strain ATCC33913, whose sequences are >90% identical to those of strain 8004, 99% of their spots on the array showed intensity ratios ≥0.5. Therefore, intensity ratios ≥0.5 were selected to be the threshold for genes detected as present/conserved within strain 8004. Furthermore, 98% of the genes previously reported to be specific to strain 8004 (that is, that are absent in the genome of strain ATCC33913) were detected as absent genes in the aCGH analysis of strain ATCC33913 (Figure 1). Our selected threshold for conserved genes here is similar to that described by Taboada et al. 30, who used a Log₂ ratio (sample/reference) threshold of -0.8 to detect conserved genes in aCGH analyses with an acceptable level of false positives.

The validity of the aCGH results was further tested by PCR examination of the presence or absence of 30 genes showing a range of ratios in the aCGH analysis. The PCR primers used and PCR results are presented in Additional data file 3. The results show that a ratio (sample/8004 strain) of <0.5 gives high confidence (98%) that the gene is absent/highly divergent (AHD) in the sample strain.

Using the parameters established above, the gene composition of 18 Chinese Xcc strains was analyzed by aCGH using the genome of strain 8004 as the reference. The results are shown in Tables 4 and 5, Figure 2 and Additional data file 4. Of the 4,186 CDSs spotted on the microarray slides, 3,405 are conserved in all of the strains tested (Table 5). These conserved CDSs may represent the common backbone ('core' genes) of the Xcc genome, which contains most of the genes encoding essential metabolic, biosynthetic, cellular, and regulatory functions (Table 5). The genes relevant to central intermediary metabolism, replication, transcription, translation, the TCA cycle, and nucleotide, fatty acid and phospholipid metabolism are largely conserved. Genes encoding the components involved in the type I (T1SS), type II (T2SS) and type III secretion systems (T1SS-T3SS) as well as extracellular polysaccharide production, and the rpf (regulation of pathogenicity factors) gene cluster 1112 are highly conserved among the Xcc strains investigated, although some predicted pathogenicity- and adaptation-related genes are AHD (Table 5).

The aCGH results showed that 730 CDSs are absent or highly divergent among all the Chinese strains tested (Tables 4 and 5). In addition, a total of 51 invalid hybridization spots (CDSs) were observed in all the aCGH analyses of the 18 Chinese strains. The 730 AHD genes, which account for 17.6% of all valid hybridized CDSs in the aCGH analyses, may constitute the Xcc flexible gene pool. The functional categories of all the AHD genes are given in Table 5. Half of the AHD genes have been predicted to encode proteins with unknown function.

The differences in the numbers of the AHD genes in different strains are significant (Table 4). Compared with the reference strain 8004, the most divergent Chinese Xcc strain is CN14, of which 475 CDSs are AHD; and the most closely related strain is CN07, of which only 137 CDSs are AHD. Fifty-seven Xcc 8004 CDSs, most of them encoding hypothetical proteins, are AHD in all eighteen Chinese strains. Of the 57 CDSs, 16 are conserved in strain ATCC33913. A hierarchical clustering program 45 was used to explore the relationship of the different Xcc strains based on the aCGH analysis (Figure 2). The result shows that the Chinese strains and the reference strain are divided into five groups (Figure 2). Some Xcc strains classified in the same phylogenetic group based on 16S-23S rDNA ITSs showed a similar grouping pattern in hierarchical clustering (Figure 2 and Additional data file 2). However, no significant relationship was observed between phylogenetic group and pathogenicity, or pathogenicity and hierarchical cluster.

No significant correlations were observed between the gross genome composition of Xcc strains and their pathogenicity, or the genome composition of the strains and their geographical origins. However, strains CN14, CN15, and CN16, which were isolated from different host plants around Guilin city, are significantly conserved in genome composition and exhibit similar pathogenicity (Tables 1 and 2; Additional data file 4). This suggests that the three strains may share a most recent common ancestor that is different from that (those) of the other Chinese strains.

The locations of the variable genes in the different strains identified by the aCGH analysis were mapped onto the chromosome of strain 8004. The results revealed that there are 27 such chromosomal regions, each of which consists of more than three contiguous CDSs in the 8004 genome (Figure 2). These regions were named XVRs for Xanthomonas variable genomic regions and numbered from 1 to 27 in accordance with the genome coordinates of strain 8004 (Table 6). The boundaries of the XVRs were determined at the CDS level, to fit in with the resolution of the array hybridization analysis in this study. The 27 XVRs contain 402 CDSs and account for 48.4% of the AHD genes, representing 9.41% of the total CDSs of Xcc strain 8004.

The size of the XVRs ranges from 1,778 bp (XVR24 with only three CDSs) to 98,358 bp (XVR13 with 81 CDSs) (Table 6). There are 15 XVRs larger than 10 kb and 4 larger than 50 kb. Within the XVRs, there are 27 genes encoding proteins for pathogenicity and adaptation, 9 for regulatory functions, 25 for cell structure and cell processes, 19 for intermediary metabolisms, 95 for mobile elements, 21 for DNA metabolism related to mobile elements, and 219 encoding hypothetical or function-unknown proteins (Table 6 and 7).

The distribution patterns of XVRs show significant diversity among the Xcc strains tested (Table 8). Five XVRs (XVR02, XVR17, XVR18, XVR20 and XVR27) are AHD from all the Chinese strains tested (Table 8). XVR17 and XVR18 are also absent from the British strain ATCC33913 as pointed out by Qian et al. 21. Most of the genes in these five XVRs encode hypothetical proteins for which there are no significantly similar sequences in GenBank.

XVR04 is a typical integron, which contains a gene for a DNA integrase (intI) catalyzing the site-specific recombination of gene cassettes at the integron-associated recombination site (attI), and a cassette array of 14 genes with unknown function 2146. Integrons are best known for assembling antibiotic resistance genes in clinical bacteria. They capture genes by integrase-mediated site-specific recombination of mobile gene cassettes. It has been postulated that the ancestral xanthomonad possessed an integron at ilvD, an acid dehydratase gene flanking the intI site-specific recombinase 46. The microarray results showed that all of the Chinese strains tested possess the ilvD gene, although whether its organization is conserved in these strains is unknown. However, significant diversity found in the integron cassette array among these Chinese strains suggests that the integron might also generate diversity within the pathovar, in addition to between pathovars 46.

XVR14 contains 21 CDSs with two copies of the phi Lf-like Xanthomonas prophage, which harbors the putative dif site of replication termination of the Xcc strains 8004 21 and Xc17 47. In strain ATCC33913, the two copies of Lf-like prophage possess the typical genetic organization of filamentous phages, that is, a symmetrical head-to-head constellation, with genes functioning in DNA replication, coat synthesis, morphogenesis and phage export 20. In strain 8004, only one copy of the Lf-like prophage is intact and the other lacks two genes (gII and gV) 2021. This phi Lf-like prophage is missing from or highly divergent in most of the Chinese strains tested and most other xanthomonads sequenced, but present in Xcv 85-10 48 (Table 9 and Figure 3). It is worth mentioning that the P2-like prophage 49, which occurs in strain ATCC33913 but is missing from strain 8004, is found to be AHD from all of the Chinese strains tested by hybridization analysis using a probe from ATCC33913 2021.

There are two clusters of the type I restriction-modification system in strain 8004, of which one is present in strain ATCC33193 and the other is unique to strain 8004 2021. XVR22 is one of these clusters. In contrast to ATCC33913, which lacks this locus, most of the Chinese strains possess it. Restriction and modification systems are responsible for cellular protection and maintenance of genetic materials against invasion of exogenous DNA. There is evidence that they have undergone extensive horizontal transfer between genomes, as inferred from their sequence homology, codon usage bias and GC content difference. In addition to often being linked with mobile genetic elements, such as plasmids, viruses, transposons and integrons, restriction-modification system genes themselves behave as mobile elements and cause genome rearrangements 50.

XVR23 consists of 14 ORFs that contains several genes for lipopolysaccharide (LPS) O-antigen synthesis, including wxcC, wxcM, wxcN, gmd and rmd 19, which is discussed below. Some predicted functions of other XVRs are shown in Table 7 based on the annotation of their component CDSs.

The detection of DNA segments in which integrase genes are associated with tRNA or tmRNA genes 515253, or regions of anomalous GC content with mobile elements 54, facilitates the identification of horizontally acquired sequences in genomes. Horizontally acquired sequences are also detectable by comparing their dinucleotide composition (genome signature) dissimilarity (δ* value) with that of the host genome. The higher δ* values of XVRs can be indicative for horizontal acquisition 55. The data presented in Tables 6 and 7 show that XVR09, XVR13, XVR18 and XVR19 are integrated adjacent to or within tRNA genes with an integrase or insertion sequence (IS) flanking the ends. XVR04, an integron 46, and XVR14, a phi Lf-like prophage 2021, are also actively transferred DNA sequences. Obviously, the five XVRs, XVR02, XVR17, XVR18, XVR20 and XVR27, which are ubiquitously AHD from all the Chinese strains tested, could be the most recently acquired DNA in strain 8004. It is possible that the donors of these five XVRs are probably absent in mainland China. In contrast, we consider that the XVRs present in the other sequenced xanthomonad strains may be a result of acquisition events during the early stage of Xanthomonas evolution and lost from certain Xcc strains at a later stage, probably due to DNA deletion events.

The identification of Xcc DNA loss events can be carried out by analysis of the sequenced xanthomonads for the presence of collinear blocks that encompass the targeted DNA segments. Whole genome comparisons among Xcc 8004 21, Xcc ATCC33913 20, X. axonopodis pv. citri 306 20, X. campestris pv. vesicatoria 85-10 48, X. oryzae pv. oryzae KACC10331 56 and X. oryzae pv. oryzae MAFF311018 57, allowed identification of a number of XVRs (XVR03, XVR05, XVR08, XVR10, XVR11 and XVR22) as DNA segments inherited from the common ancestral xanthomonad (Figure 3). In each case, large DNA segments containing each of these XVRs have a high degree of synteny in other xanthomonads (Figure 3 and Table 9).

Analysis of the structure of XVR13 and its distribution pattern in Xcc strains revealed that this region might undergo a series of multiple insertion and deletion events during the Xcc evolution (Figure 4). This region is near the terminus of chromosome replication, which is susceptible to gene acquisition and/or gene loss 20. XVR13 is the largest genomic island identified in Xcc 8004, which spans nucleotide coordinates from 2,414,668 to 2,513,025 and contains 81 CDSs. To its left flank are three tRNA genes and an integrase gene. Genome comparison showed that the central part of XVR13, named XVR13.1, is totally absent in strain ATCC33913. XVR13.1 is 58,007 bp in length. The aCGH results reveal that three Chinese strains (CN01, CN03 and CN11) contain the XVR13 locus, which is almost identical to that of Xcc 8004, and four Chinese strains (CN07, CN09, CN10 and CN20) contain an incomplete XVR13 locus without XVR13.1 that is almost identical to that in strain ATCC33913, and the rest of the Chinese strains probably have no XVR13 (Table 8 and Figure 4). To elucidate the dynamic relationship between XVR13 and XVR13.1, re-annotation was done for XVR13.1 and 63 CDSs were identified (Figure 4 and Additional data file 5). A truncated yeeA-like gene was found across the right border of XVR13.1 (Figure 4). Intriguingly, yeeB- and yeeC-like genes occur in both Xcc strains 8004 and ATCC33913 (Figure 4). This suggests that XVR13.1, or at least part of it, has been lost from the British strain ATCC33913 and most of the tested Chinese strains during their evolution.

XVR23, part of the wxc cluster, contains several genes for O-antigen synthesis of LPS 21. The aCGH results revealed that this region is highly divergent, with a mosaic structure among the Chinese strains tested. Sequence comparisons showed that wxc cluster of Xcc 8004 is significantly divergent from that of Xcc B100 19, although it is almost identical to that of Xcc ATCC33913 2021. The wxc cluster of strain 8004 is truncated by IS elements and some of the wxc genes have low similarity to the corresponding genes of strain B100. Significant differences in wxc clusters among other xanthomonad strains have also been reported 485658. The Xcc wxc cluster not only has a significantly lower GC content (56.82%) than the average genome level (64.95%), but also has a very high δ* value of 81.182. These suggest that Xcc might have acquired the wxc cluster by horizontal DNA transfer.

Bioinformatic analysis revealed that strain 8004 contains 197 CDSs that show homology to the confirmed or annotated putative pathogenicity genes of plant or animal pathogenic bacteria, in addition to 108 genes that have been proven to be involved in Xcc pathogenicity (Additional data file 6). Of these 305 proven or presumed pathogenicity genes, 304 were spotted on the microarray slides of strain 8004 in this study. The other CDS (XC3591) encoding pectate lyase was not spotted as it has a redundant DNA sequence in the genome of strain 8004. The aCGH analysis revealed that 258 of the pathogenicity genes (84.8% of the pathogenicity genes spotted) are present in all of the Xcc strains tested and 46 (15.1%) are AHD in at least one of the strains (Table 5 and Additional data file 6). The results show that the pathogenicity genes involved in the type I, II and III secretion systems (T1SS, T2SS and T3SS), host cell wall degradation, extracellular polysaccharide production, and the quorum sensing system are highly conserved in almost all of the Xcc strains tested (Table 5 and Additional data file 6). In addition, genes encoding proteins of the gluconeogenic pathway 59, Mip-like protein 60, the catabolite repressor-like protein Clp 61, and zinc uptake regulator protein Zur 44, which have been demonstrated to play important roles in Xcc virulence, are also highly conserved. However, genes relating to T4SS, T3SS-effectors and candidates, LPS synthesis, toxin as well as adhesin are highly diversified (Table 5 and Additional data file 6).

LPS is an indispensable component of the cell surface of Gram-negative bacteria and has been demonstrated to play important roles in pathogenicity of several phytopathogenic bacteria, including Xcc 626364. More than 20 genes for LPS synthesis have been characterized in Xcc. These include xanAB 65, rmlABCD 66, rfaXY 64, lpsIJ 67 and the wxc cluster consisting of 15 genes 19. The aCGH results suggest that lpsIJ, rfaXY, rmlABCD and xanAB are highly conserved while wxc genes are divergent in the Xcc strains tested. The wxc genes are involved in the biosynthesis of the LPS O-antigen, which is the most variable portion of LPS 1968. The diversity of the wxc cluster indicates that the LPSs produced by Xcc different strains may be varied.

T4SSs have been validated as having important roles in the pathogenesis of several animal and plant bacterial pathogens 36373840. The T4SS of Agrobacterium tumefaciens is essential for virulence and is assembled from the proteins encoded by the virB cluster and virD4. Many T4SSs are highly similar to the A. tumefaciens VirB/D4 T4SS 40. Burkholderia cenocepacia strain K56-2 can produce the plant tissue watersoaking phenotype (a plant disease-associated trait) and possesses two T4SSs similar to the VirB/D4 system 69. Mutational studies in B. cenocepacia strain K56-2 revealed that the plasmid-encoded T4SS is involved in eliciting the plant tissue watersoaking phenotype and responsible for the secretion of a plant cytotoxic protein(s), while the chromosome-encoded T4SS is not 69. Genome annotation revealed that the Xcc strain 8004 has an A. tumefaciens VirB/D4-like T4SS 21. Although genomic sequence comparison showed that the Xcc strain ATCC33913 possesses an almost identical virB cluster to that of strain 8004, the aCGH analyses displayed that the virB cluster of most Chinese strains tested is AHD. Since all these strains were fully virulent and their aCGH intensity ratios were extremely low (as low as 0.1-0.025; Additional data file 4), a query on the role of the T4SS in Xcc pathogenicity was raised. To answer this question, we constructed a T4SS mutant derived from strain 8004 (Figure 5). A mutant with deletions of the virB cluster as well as virD4 was confirmed by PCR and designated 8004ΔT4 (Figure 5 and Additional file 7). The virulence of the mutant was tested on host plants cabbage (B. oleracea var. capitata) cv. Jingfeng-1, Chinese cabbage (B. rapa subsp. pekinensis) cv. Zhongbai-83, Chinese kale (B. oleracea var. alboglabra) cv. Xianggangbaihua, pakchoi cabbage (B. rapa subsp. chinensis) cv. Jinchengteai, and Radish (R. sativus var. radicula) cv. Manshenghong by the leaf-clipping inoculation and spray methods. The results showed that the virulence of the mutant was as severe as on the wild type strain 8004 on all the tested plants inoculated by leaf-clipping (Figure 5) or spray (data not shown). This suggests that the T4SS is not involved in the virulence of Xcc.

Genes involved in the host specificity of Xcc are of central interest in this study. All of the Xcc strains used in this work are able to cause disease in their host plants but show specificity for a host range. Apart from four strains (CN01, CN04, CN05 and CN17) that could infect all of the host plants tested, the other 16 strains were avirulent on certain host plant(s) (Table 2). The host specificity of pathogens is determined by gene-for-gene interactions 10 involving avirulence (avr) genes of the pathogen and cognate resistance (R) genes of the host. Disease resistance occurs in a host-pathogen interaction in which an R gene in the host is matched by a cognate avr gene in the challenging pathogen. A pathogen-host interaction without such a cognate avr-R combination will lead to disease.

To elucidate the genetic determinants for host specificity of Xcc, the correlation between the virulence scale on host plants and the gene distribution pattern of the 20 Xcc strains was analyzed. The correlation between HR induction on non-host plants and gene distribution patterns of the strains was also determined. Twelve operations were performed and the correlation coefficient (CC) values of these are given in Additional data files 8 and 9. Seven of the eleven host plants are susceptible to all of the 20 Xcc strains tested (Table 2), indicating that they have no CC values. Correlation analyses for the other four host plants and one non-host plant discovered four candidate genes responsible for the virulence-deficiency (negative CC value) of Xcc strains on a particular host plant(s) and one candidate for HR induction (positive CC value) on the non-host plant pepper ECW10R. These genes are candidates of the three postulated avr genes avrRc1, avrRc3 and avrRp1 (Table 10). The candidates XC2004 and XC2084 are correlative to avrRc3 and have the same CC value. XC2084 encodes a transposase 21, suggesting that its postulated avrRc3 is much smaller than that of XC2004. Therefore, XC2084 was removed from the candidate list. The candidate genes XC2602, XC2004 as well as XC2081 have been annotated as encoding Avr-homologous proteins 21.

To identify avr genes from the candidates, we further investigated their biological functions by mutagenesis. The candidate avr genes of Xcc 8004 were disrupted by using the plasmid pK18mob 70, a conjugative suicide plasmid in Xcc (see details in Materials and methods). The obtained nonpolar mutants of XC2602, XC2004 and XC2081, named NK2602, NK2004, and NK2081, respectively, were inoculated on corresponding host or non-host plants to test their virulence or HR. The results revealed that mutation in XC2004 or XC2602 altered the reaction of the pathogen on the corresponding host plant mustard cv. Guangtou or Chinese cabbage cv. Zhongbai-83, respectively, from non-pathogenic to pathogenic (Figure 6 and Table 10). Disruption of XC2081 resulted in the loss of the ability to elicit an HR on the non-host plant pepper ECW10R (Figure 6 and Table 10). These alterations in plant response caused by mutation in XC2004, XC2602 or XC2081 could be restored to the wild-type phenotype by expression in trans of the intact corresponding CDS carried by a DNA fragment cloned into pLAFR3 or pLAFR6 (Figure 6 and Table 10). These results demonstrate that XC2004, XC2602 and XC2081 are the postulated avrRc1, avrRc3 and avrRp1, respectively. XC2004, XC2602 and XC2081 of strain 8004 have been annotated as avrXccC, avrXccE1 and avrBs1, respectively, based on their sequence homology to avr genes identified in other pathogens 21. Therefore, we renamed these postulated avr genes avrRc1, avrRc3 and avrRp1 as avrXccC, avrXccE1 and avrBs1, respectively (Table 10). Recently, Castañeda and associates 22 have shown that the avirulence of Xcc strain 528^T (Xcc ATCC33913) on Florida Mustard is attributed to avrXccFM, which shares the same locus as avrXccC but is longer than the avrXccC ORF annotated in the genome of ATCC33913 20. Our results further confirm that the avrXccC locus dominates the avirulence of Xcc on mustard plants. The avirulence function of Xcc avrBs1 is similar to that of the homologue avrBs1 of Xcv on the resistant pepper ECW10R, which contains the corresponding R gene Bs1 71.

To verify the avirulence function of Xcc avrXccE1 (XC2602), the cosmid pLAFR6 carrying a PCR-generated 1,605 bp fragment encompassing the region 514 bp upstream of the start codon to 29 bp downstream of the stop codon of XC2602 was introduced by triparental mating into the Chinese strains CN01, CN05, CN10 and CN11, which showed virulence on Chinese cabbage cv. Zhongbai-83 (Table 2). The obtained transconjugants for all the four strains lost virulence on Chinese cabbage cv. Zhongbai-83 (Figure 7). These results demonstrate that avrXccE1 (XC2602) of Xcc is endowed with an avr function determining host specificity.

Xcc isolates used in this study were collected from various geographical locations over a wide range of latitudes across mainland China (Tables 1 and 2). The bacteria were isolated from the infected leaves of cruciferous plants with typical symptoms of black rot disease. Recovered colonies were picked and re-streaked onto NYG 14 agar plates to verify the bacterial identity. Each isolate was inoculated onto radish (R. sativus var. radicula) cv. Manshenhong by the leaf clipping method 85 to evaluate its pathogenicity. The 16S-23S rDNA ITS was amplified as described by Gurtler and Stanisich 41 using primer R1 and primer R2 (Additional data file 1).

Molecular manipulations, genomic DNA preparations, restriction endonuclease digestions and PCR amplifications were performed as described by Sambrook et al. 86. Enzymes were supplied by Promega (Shanghai, China) and used in accordance with the manufacturer's instructions.

The virulence of Xcc strains was evaluated on 11 host plants: cabbage (B. oleracea var. capitata) cv. Jingfeng-1, Chinese cabbage (B. rapa subsp. pekinensis) cv. Zhongbai-4, Chinese cabbage (B. rapa subsp. pekinensis) cv. Zhongbai-83, Chinese kale (B. oleracea var. alboglabra) cv. Xianggangbaihua, kohlrabi (B. oleracea var. gongylodes) cv. Chunqiu, mustard (B. juncea var. megarrhiza Tsen et Lee) cv. Guangtou, pakchoi cabbage (B. rapa subsp. chinensis) cv. Jinchengteai, pakchoi cabbage (B. rapa subsp. chinensis) cv. Naibaicai, radish (R. sativus var. sativus) cv. Cherry Belle, radish (R. sativus var. longipinnatus) cv. Huaye, and radish (R. sativus var. radicula) cv. Manshenhong (Table 2). All of these cultivars are available from the Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences, Beijing 100081. Each Xcc strain was tested on all the 11 cultivars. The bacteria grown overnight in NYG medium 14 were washed and resuspended in water to a cell density OD of 0.01 at 600 nm. The last completely expanded leaf of the five-week old seedlings was inoculated by cutting with scissors dipped in bacterial suspensions 85 or by spraying the bacterial suspensions with a sprayer. Twenty leaves were inoculated for each strain-plant combination. The inoculated plants were kept in a culture room at a temperature of 28°C and a relative humidity of 80% under 16 h light day, after 24 h moisture preservation in a plastic chamber at a temperature of 28°C and a relative humidity of about 100%. First symptoms appeared five days post-inoculation, and the lesion lengths of 20 leaves were measured 10 days post-inoculation for each strain-plant combination. The virulence of each Xcc strain on each host plant was rated according to the disease symptoms: non-pathogenic, leaves with no visible effect, or with localized necrosis (HR) or with few small lesions (less than 3 mm) near cuts; weakly virulent, leaves with chlorosis extending from cuttings; and fully virulent, blackened leaf veins, death, and drying of tissue with V-shape lesions. This rating method was modified from Ignatov et al. 87.

For HR tests, Xcc strains were cultured as for the virulence assay, adjusted to a density of 10⁸ colony forming units per ml with distillated water and introduced, by the infiltration method with a needleless syringe 59, into the intercellular spaces of the leaves of non-host plant pepper (Capsicum annuum cv. Early Cal Wonder) ECW10R (from Laboratoire de Biologie Moleculaire des relations Plantes Microorganismes INRA-CNRS, Castanet Tolosan, France). After inoculation, the plants were kept at 28°C under continuous illumination of 6,000 lux light intensity. The ΔhrpG mutant, a Xcc deletion mutant of hrpG 88, was used as a negative control.

A high-density PCR-based DNA array was designed by using the genome sequence data of Xcc strain 8004 (GenBank accession number CP000050). The genome has 5,148,708 bp and encodes 4,273 predicted CDSs 21. An in-house high-throughput computer algorithm based on the Linux operating system and Python programming language was employed to design PCR primers for all CDSs. The fundamental rules of our computer algorithm include that all the primer annealing temperatures range from 57.5-68.7°C, and the PCR product sizes fall within 200-1,000 bp, with an optimum of 500 bp. The PCR amplicons should have a minimum sequence similarity with cut-off e-value <1 e^-3 and sequence identity <70% when using the BLAST program. There are 87 genes which were designed not to be spotted on the array because of their high sequence similarity to other genes in the genome. The PCR amplifications were performed in a 100 μl reaction volume and PCR success was confirmed by agarose gel electrophoresis. The confirmed PCR products were precipitated with isopropanol and redissolved in DNA Spotting Solution (CapitalBio Corp., Beijing, China). For ORFs that were too small or those genes for which PCR amplification failed, 70-mer sense oligonucleotides were designed; 143 such oligonucleotide probes were synthesized. PCR products and 70-mer oligonucleotides (20 μM) were printed on amino silaned glass slides (CapitalBio Corp.) using a SmartArray™ microarrayer (CapitalBio Corp.). Each CDS was printed in triplicate to facilitate subsequent data analysis. After printing, the slides were baked at 80°C for 1 h and stored dry at room temperature till use. The Xcc 8004 microarray slides are available to the public from CapitalBio Corp. 89.

Prior to hybridization, the slides were rehydrated over 65°C water for 10 s, and UV cross-linked at 250 mJ/cm². The unimmobilized DNAs were washed off with 0.5% SDS for 15 minutes at room temperature and SDS was removed by dipping the slides in anhydrous ethanol for 30 s. The slides were spin-dried at 1,000 rpm for 2 minutes.

Genomic DNA was fragmented by Dpn II endonuclease digestion, and then purified with the PCR Clean-up NucleoSpin Extract II kits (Macherey-Nagel, Düren, Germany). For each labeling reaction, 2 μg of digested DNA and 4 μg of random nonamer were heated to 95°C for 3 minutes and snap cooled on ice, then 10× buffer, dNTP and Cy5-dCTP or Cy3-dCTP (GE HealthCare Bio-Sciences AB, Björkgatan, Uppsala, Sweden) were added at final concentrations of 120 μM each dATP, dGTP, dTTP, 60 μM dCTP and 40 μM Cy-dye. Klenow enzyme (1 μl; Takara, Dalian, China) was added and the reaction was performed at 37°C for 1 h. The labeled DNA was purified with a PCR Clean-up NucleoSpin Extract II kit, resuspended in elution buffer and checked for its optical density.

For aCGH, the final products hybridized with microarrays were fluorescence-labeled DNA, so an identical hybridization strategy was employed. Labeled control and test samples were quantitatively adjusted based on the efficiency of Cy-dye incorporation and mixed into 80 μl hybridization solution (3× SSC, 0.2% SDS, 50% formamide). DNA in hybridization solution was denatured at 95°C for 3 minutes prior to loading on the microarray. Hybridization was performed under LifterSlip™ (Erie Scientific Company, Portsmouth, NH, USA), which allows for even dispersal of hybridization solutions between the microarray and coverslip. The hybridization chamber was laid on a Three-phase Tiling Agitator (CapitalBio Corp.) to prompt the microfluidic circulation under the coverslip. The array was hybridized at 42°C overnight and washed with two consecutive washing solutions (0.2% SDS, 2× SSC for 5 minutes at 42°C and 0.2% SSC for 5 minutes at room temperature).

Arrays were scanned with a confocal LuxScan™ scanner (CapitalBio Corp.) and the data of obtained images were extracted with SpotData software (CapitalBio Corp). In order that the aCGH results were also represented with the fluorescence intensity ratio, a spatial and intensity-dependent normalization based on a LOWESS program was employed, which is prevalent in microarray expression profiling 31. Since each gene was represented in triplicate on each slide and the experiments were performed in duplicate by dye swap, producing six data points, the average ratio (always sample/reference strain 8004) of each gene was input into hierarchical clustering with an average linkage algorithm for aCGH analysis.

All the aCGH data can be accessed at the National Center for Biotechnology Information Gene Expression Omnibus (GEO) database 90 with accession number GSE5087.

Putative AHD CDSs identified by aCGH were examined by PCR using the primers designed within CDSs in strain 8004. The oligonucleotide primers and the PCR results are shown in Additional data file 3.

Whole genome comparison of the CDS set of strain 8004 with that of each sequenced xanthomonad strain was carried out using the BLASTN program 91. Shared genes were defined using an e-value cutoff of e^-20. The CDS sets were obtained from GenBank with the following accession numbers (in parentheses): Xcc 8004 (CP000050), Xcc ATCC33913 (AE008922), X. axonopodis pv. citri (Xac) 306 and plasmids pXAC33 and pXAC64 (AE008923, AE008924 and AE008925, respectively), X. oryzae pv. oryzae (Xoo) KACC10331 (AE013598), Xoo MAFF311018 (NC_007705), X. campestris pv. vesicatoria (Xcv) strain 85-10 and its four plasmids pXCV2, pXCV19, pXCV38, and pXCV183 (AM039948, AM039949, AM039950, AM039951, and AM039952, respectively).

The phylogenetic relationships of all the Xcc strains tested and other xanthomonad strains used as references were constructed by the maximal parsimony method based on pairwise comparisons of partial 16S-23S rDNA ITSs, which were obtained from direct ITS rDNA sequencing of Chinese strains and from GenBank with the accession number of each xanthomonad strain: X. axonopodis pv. aurantifolii (Xaa) strain X84 (AF442739.1), Xcc strain XCC15 (AF123092.2), X. axonopodis pv. dieffenbachiae (Xad) ATCC23379 (AY576642.1), Xad X195 (AY576648.1), X. arboricola pv. pruni (Xap) (AJ936965.1), X. gardneri (Xg) strain CNPH496 (AY288083.1), X. vesicatoria (Xv) strain CNPH345 (AY288080.1), Xv XV1111 (AF123088.2), and other strains, such as Xcc 8004, Xcc ATCC33913, Xac 306, Xcv 85-10, Xoo KACC10331, and Xoo MAFF311018 with the same accession number as that for each genome.

The genomic dissimilarity δ* values (the average dinucleotide relative abundance difference) between the putative variable genomic region in Xcc (XVR) and the genome sequence of Xcc strain 8004 were determined by the δρ-WEB program 5592 and are listed in Table 5. A BLASTN search in GenBank was carried out for each XVR in order to identify the origin of potential horizontal gene transfer if the homology was high enough.

To identify the genetic determinants for host specificity of Xcc, a correlation analysis was performed using the CORREL tool in Excel (Microsoft Office 2000). Prior to statistical operation, the aCGH result of each gene in any Xcc strain was transformed from the ratio value to the numerical code: 0 = absent or highly divergent; 1 = present. The pathogenicity test results were transformed from a qualitative description to a numerical code: 0 = non-pathogenic; 1 = pathogenic. The HR results were transformed from a qualitative description to a numerical code: 0 = no HR; 1 = HR. For one round of statistical operation, a direct correlation analysis between virulence scales of the 20 Xcc strains (including 18 Chinese stains, strain ATCC33913 and strain 8004) on one given plant cultivar and the distribution pattern of each gene in 20 Xcc strains was carried out using the program CORREL. Twelve operations were performed and the CC values of each operation were listed in one column, parallel to the gene list of strain 8004 (Additional data files 8 and 9).

In each correlation analysis (for each plant assay), the Xcc genes with the maximal R absolute values were selected as the candidates responsible for host specificity of Xcc strain 8004 on a particular plant. Due to the possibility of more than one gene having the same distribution pattern among 20 Xcc strains, more than one candidate gene for each genetic determinant was able to be selected (Table 10).

The virB/D4 T4SS deletion mutant was generated by the marker exchange method. The upstream and downstream fragments flanking the virB/D4 cluster were amplified with the primer sets DT4-LF/DT4-LR (Additional data file 1) (the coordinate position of the amplified fragment in Xcc 8004 chromosome is from 1956072 to 1957097, and DT4-RF/DT4-RR (the coordinate position of the amplified fragment in Xcc 8004 chromosome is from 1965913 to 1966832, respectively). Simultaneously, the gentamicin resistant fragment was amplified with the primer sets Gm-F/Gm-R (Additional data file 1). The obtained fragments were cloned into the EcoRI-KpnI-BamHI-XbaI sites of the suicide vector pK18mobsacB 70 one by one, yielding the recombinant plasmid pKDT4. The plasmid pKDT4 was transferred into Xcc wild-type strain 8004 by triparent conjugation and kanamycin resistant transconjugant colonies were screened. Bacterial cells cultured in NYG broth without antibiotics overnight from a single transconjugant colony chosen randomly were diluted gradiently and plated on the NYG agar plats with 5% sucrose and appropriate gentamicin. The gentamicin resistant and kanamycin sensitive colonies were screened, generating the deletion mutant of virB/D4 T4SS, named 8004ΔT4 (Figure 5). The deletion mutant 8004ΔT4 was further confirmed by PCR with the primer sets DT4-LF/DT4-RR (Additional data file 1) and the primer sets of each ORF of virB/D4 T4SS (Additional data file 7). The virulence of the mutant was tested on host plants by the leaf-clipping inoculation method 85.

The candidate avr genes (XC2602, XC2004 and XC2081) of Xcc 8004 were disrupted by using the plasmid pK18mob, a conjugative suicide plasmid in Xcc 70. The internal fragment of each target gene was amplified by PCR using chromosomal DNA of Xcc strain 8004 as template and the primers designed according to certain CDSs (Additional data file 1), and cloned into the plasmid pK18mob to generate a recombinant plasmid. The identity of the cloned fragment was confirmed by sequencing. Each recombinant plasmid was transformed into Escherichia coli JM109 (Additional data file 1) and then introduced into the wild-type strain 8004 by triparental conjugation using the helper plasmid pRK2073 (Additional data file 1). Transconjugants were selected on the NYG plates containing rifampicin and kanamycin. Mutants were screened for disruption of the target gene by PCR using primer PMOB-SP (Additional data file 1), a specific primer from pK18mob, and a specific primer of the upstream gene of each target gene (Additional data file 1). The obtained mutants of XC2004, XC2602 and XC2081 were named NK2004, NK2602 and NK2081, respectively.

The complementation of the mutation of each target gene was carried out by introduction of the broad host range cosmid pLAFR3 carrying the intact target gene into the corresponding mutant strain. The intact target gene was amplified by PCR using chromosomal DNA of Xcc 8004 as template and the specific primer sets (Additional data file 1), and cloned into the plasmid pLAFR3 under the control of the P_lac promoter. The identity of the cloned target gene was confirmed by sequencing. The confirmed recombinant plasmid was transformed into E. coli JM109 and then introduced into the corresponding mutant strain by triparental conjugation. The transconjugants were screened on NYG plates with rifampicin, kanamycin and tetracycline. The created complementary strains for the mutants NK2602, NK2004 and NK2081 were named CNK2602, CNK2004 and CNK2081, respectively.

For verification of the avr function of putative avrXccE1, the plasmid containing XC2602 was transferred into the Chinese strains CN01, CN05, CN10 and CN11, which contain no homologs of XC2602. A 1,605 bp fragment that includes the region from 514 bp upstream of the star codon to 29 bp downstream of the stop codon of XC2602 was amplified with the primer set XC2602CM-F/XC2602CM-R (Additional data file 1) using the total DNA of Xcc 8004 as template. After confirmation by sequencing, the fragment was cloned into the promoterless cosmid pLAFR6 to generate the recombinant plasmid named pC2602. The recombinant plasmid pC2602 was transferred into the strains CN01, CN05, CN10 and CN11 by triparental conjugation. The transconjugants carrying pC2602 were screened on NYG plates with rifampicin and tetracycline, and named CN01/pC2602, CN05/pC2602, CN10/pC2602 and CN11/pC2602, respectively. The virulence of the obtained strains CN01/pC2602, CN05/pC2602, CN10/pC2602 and CN11/pC2602 on Chinese cabbage cv. Zhongbai-83 was tested by the leaf-clipping method described above.