Thirty-one previously characterized strains of X. arboricola (Young et al., 2008; Palacio-Bielsa et al., 2011; Pothier et al., 2011b; Garita-Cambronero et al., 2016a; López-Soriano et al., 2016) from the pathovars pruni, corylina, juglandis, and populi were utilized. Besides, 40 strains showing Xanthomonas-like colonies were collected during the Spanish outbreaks of bacterial spot disease of stone fruit trees and almond, as well as from routine screenings performed on Spanish nurseries (Table S1). These strains were screened for identification as Xanthomonas arboricola pv. pruni (Xap). All the bacterial strains listed in Table S1 are available in the collections from the Instituto Valenciano de Investigaciones Agrarias (IVIA, Valencia, Spain) and the Centro de Investigación y Tecnología Agroalimentaria de Aragón (CITA, Aragón, Spain).

Bacterial strains were cultured on Luria Bertani (LB) 1.5% agar plates or in LB broth at 27°C for 48 h. The commensal bacterial strains, isolated from Prunus and used in this study (Table S1), were identified to genus level based on the partial sequence of the 16S rDNA gene according to a method described previously (Lagacé et al., 2004).

For an initial Xap classification, a real-time PCR reaction in the gene ftsX of an ABC transporter (Palacio-Bielsa et al., 2011, 2015), and a multiplex PCR for plasmid pXap41 (Pothier et al., 2011b) were performed. Those strains that showed a positive result only for the real-time assay were considered as Xap-look-a-like strains, and were further identified according to a multilocus sequence typing scheme (MLSA) based in the partial sequences of the housekeeping genes dnaK, fyuA, gyrB and rpoD (Young et al., 2008).

Additionally, sequences of these housekeeping genes from the Prunus-non-virulent X. arboricola strain CITA 44 and sequences from X. arboricola pathovars celebensis (CFBP 3523 = ICMP 1488 = NCPPB 1832), corylina (CFBP 1159 = ICMP 5726 and CFBP 1846), juglandis (CFBP 2528 = ICMP 35 and IVIA 2113), populi (CFBP 3123) and pruni (CFBP 2535 = ICMP 51, CFBP 5530, Xap 33 = CITA 33 and IVIA 2626.1), as well as X. citri subsp. citri strain CFBP 2525 = ICMP 24, included as outgroup, were obtained from the National Center for Biotechnology Information database (NCBI).

Purified PCR products were sequenced at STAB VIDA (Lisbon, Portugal), and edited using Geneious (Kearse et al., 2012). Obtained nucleotide sequences were aligned with ClustalW version 1.83 (Hall, 2011) using default parameters. Both ends of each alignment were trimmed to the following sizes: dnaK, 842 positions; fyuA, 601 positions; gyrB, 631 positions and rpoD, 759 positions. Then, they were aligned and concatenated to give a total length of 2,833 nucleotide positions. For the analysis of the concatenated gene dataset, Tamura-Nei (TN93) model was selected and maximum likelihood trees, using 1,000 bootstrap re-samplings, were generated using MEGA 6.0 software (Tamura et al., 2013).

Nucleotide sequences were deposited in GenBank. Accession numbers for the partial sequences of the genes used in this study are: KX357115 to KX357120 for dnaK; KX357133 to 357138 for fyuA; KX357127 to KX357132 for gyrB and KX357121 to KX357126 for rpoD.

Six strains isolated from Prunus spp. and classified as Xap-look-a-like, as well as the pathogenic Xap strains CITA 33 and CFBP 5530, and the Prunus-non-virulent strain CITA 44, were typed by PCR for 11 genes related to structural and regulatory components of the T3SS, 19 genes for the T3Es and two genes that encoded the type III secreted proteins (T3SPs) hpaA and hrpW predicted in Xap (Hajri et al., 2012; Garita-Cambronero et al., 2016a). PCR reactions were performed according to the conditions proposed previously (Hajri et al., 2012) with the exception of the T3SS genes, hrpD5 and hrpF, as well as the T3SP hpaA and the T3Es genes xopAQ and xopZ, for which new sets of primers were designed based in orthologues available in databases for X. arboricola (Table 1). PCR amplifications with the primers for hrpD5, hrpF, hpaA, xopAQ, and xopZ were performed in 20 μl of PCR reaction containing 1X PCR buffer (10 mM Tris-HCl, 50 mM KCl, 0.1% Triton X-100 [pH 9.0]); 0.5 μM of each primer; 0.25 U Taq DNA polymerase (Biotools, Madrid, Spain); 0.2 mM each dNTP (Biotools Madrid, Spain); 1.5 mM MgCl₂ and 1.0 μg/μl of DNA template. PCR conditions consisted in an initial denaturation step of 94°C for 2 min, 30 cycles of denaturation at 94°C for 1 min, annealing at 60°C for 1 min, extension at 72°C for 2 min and a final extension step at 72°C for 10 min. PCR products were visualized in 2% agarose gel containing Midori Green nucleic acid gel staining solution (Nippon Genetics Europe, Dueren, Germany).

Pathogenicity tests on barley (H. vulgare), Nicotiana benthamiana, Nicotiana tabacum, Prunus persica (rootstock GF-305) and tomato (Solanum lycopersicum) were carried out for the six Xap-look-a-like strains as well as for the pathogenic strain of Xap CITA 33 and the non-virulent strain of X. arboricola CITA 44. Bacterial suspensions in sterile phosphate buffered saline (PBS pH = 7.5), adjusted to a final concentration of 1 × 10⁶ colony forming units (CFU)/ml, were infiltrated in three leaves per plant using a syringe without needle and sterile PBS was utilized as blank control. All the infiltrated plants were kept in a growth chamber with high humidity and 16 h of light and 8 h of darkness at 26 and 22°C, respectively. Infiltration results were graphically recorded at 0, 7, 14, and 21 days post inoculation (dpi). After 21 dpi, the infiltrated leaves were macerated on sterile distilled water and tenfold dilution of the macerated were plated on YPGA (0.5% yeast extract, 0.5% bactopeptone, 1.0% glucose, 2.0% agar) supplemented with 250 mg/l of cycloheximide. Colonies showing a Xanthomonas-like phenotype were confirmed as Xap-look-a-like using a real time PCR protocol indicated above (Palacio-Bielsa et al., 2011, 2015). Additionally, for P. persica rootstock GF-305, colonies were counted in order to determine the bacterial concentration in the inoculated tissue at the end of the assay (Ah-You et al., 2007).

From the six Xap-look-a-like strains analyzed previously, one representative of each cluster, according to the MLSA analysis (Table S1; Figure 1), was selected for genome sequencing. Genome sequencing conditions and features for CITA 44 have been discussed in a previous paper (Garita-Cambronero et al., 2016b). In addition, in this study the genome features of the Xap-look-a-like strains CITA 14 and CITA 124 are described. For these two strains, the genome sequencing and assembly conditions have been previously announced and deposited at DDBJ, EMBL, GenBank databases under the accession numbers LXIB00000000 for CITA 14 and LXKK00000000 for CITA 124 (Garita-Cambronero et al., 2016c).

The assembled draft genome sequence of strains CITA 14 and CITA 124 were automatically annotated using the NCBI's prokaryotic annotation pipeline (Tatusova et al., 2013). Signal peptides and transmembrane domains were predicted using the signalP 4.1 (Petersen et al., 2011) and the TMHMM 2.0 (Krogh et al., 2001) servers. The assignment of genes to each cluster orthologous group (COG) and its Pfam domain was performed with the NCBI's conserved domain database using an expected value threshold of 0.001 (Marchler-Bauer et al., 2014). The circular genome maps of the draft genome sequences of X. arboricola strains CITA 14 and CITA 124, representing the COG categories of the genes, were constructed using CGView (Stothard and Wishart, 2005). The contigs of both strains were arranged by Mauve (Darling et al., 2004, 2010) using the complete genome sequence of X. arboricola pv. juglandis Xaj417 as the reference (Pereira et al., 2015).

The genome sequence variation among CITA 14, CITA 124 and the publicly available genomes of X. arboricola (strains 3004, CFBP 7634, CFBP 7651, and CITA 44; Cesbron et al., 2015; Ignatov et al., 2015; Garita-Cambronero et al., 2016b), X. arboricola pv. celebensis (NCPPB 1630 and NCPPB 1832; Harrison et al., 2016), X. arboricola pv. corylina (NCCB 100457; Ibarra Caballero et al., 2013), X. arboricola pv. juglandis (CFBP 2528, CFBP 7179, Xaj2, and Xaj417; Cesbron et al., 2015; Higuera et al., 2015; Pereira et al., 2015) and X. arboricola pv. pruni (CITA 33, IVIA 2626.1, MAFF 301420, and MAFF 301427; Garita-Cambronero et al., 2014; Garita-Cambronero et al., 2016b) (Table S2), were determined by using sequence-based and sequence content approaches (Snipen and Ussery, 2010).

Comparison among genome sequences based on sequence alignment and evolutionary analysis, of the shared protein coding sequences (CDS), were determined using Roary (Page et al., 2015). Contigs of the 17 genome sequences of X. arboricola were ordered by Mauve. Afterwards, all the genome sequences were automatically annotated using PROKKA (Seemann, 2014) and used as the input for the search of shared homologous genes among the studied strains.

Those CDS shared by all the analyzed genomes of X. arboricola, with an identity and a coverage percentage over 80%, were considered as homologous genes. The concatenated sequences of the genes that composed the core genome sequence of X. arboricola were aligned using the PRANK (Löytynoja and Goldman, 2008) and subsequently, a maximum likelihood tree (1,000 bootstrap resamplings) was constructed to determine the phylogenetic position of the strains CITA 14 and CITA 124 within X. arboricola. Maximum likelihood tree was performed with RaxML (Stamatakis, 2014) and visualized using Dendroscope (Huson et al., 2007).

Additionally, the gene-content comparison was performed using the R implemented package for microbial-pangenomics micropan (Snipen and Liland, 2015). The CDS were obtained from nucleotide genome sequences of the 17 strains mentioned above using Prodigal v2.6.1 (Hyatt et al., 2010). To determine the similarity of the proteins within and across the genomes, a reciprocal all-against-all BLAST search was performed using the blastAll package. BLAST distance between sequences was obtained using the bDist package. A hierarchical clustering was performed using bClust, and a complete linkage function was selected with a liberal threshold of 0.80. A similarity matrix, with the number of protein sequences contained in each cluster for each genome, was constructed using the panMatrix function and Jaccard distance was calculated. The matrix was used to perform a principal component analysis for showing how the genomes were distributed in the space according to the two first principal components which revealed the dominant differences between them, and this was computed using the panpca and plotScores functions. Additionally, the similarity of the analyzed genomes was represented in a pan-genome tree using the panTree function (Snipen and Ussery, 2010). Tree construction was based in the distance between genomes according to the Manhattan distance. Bootstrap values were calculated by re-sampling the columns of the similarity matrix and the re-clustering of these data, therefore, the bootstrap value represented was the percentage of the re-sampled trees that showed a similar node.

In order to determine potential groups of genes with a putative function related to pathogenesis, genes with an identity and a coverage percentage over 80% associated with tonB-dependent transporters (TBDTs), sensors of the two-component regulatory system (STCRs), methyl accepting chemotaxis proteins (MCPs), flagella, type IV pilus, non-fimbrilar adhesins, production of xanthan, quorum-sensing regulation, repertoire of cell-wall degrading enzymes, type II, III and IV secretion systems, as well as T3Es and T3SPs, previously reported in Xanthomonas spp. (da Silva et al., 2002; Filloux, 2004; Chevance and Hughes, 2008; He and Zhang, 2008; Vorhölter et al., 2008; Wang et al., 2008; White et al., 2009; Subramoni et al., 2010; Szczesny et al., 2010; Guo et al., 2011; Mhedbi-Hajri et al., 2011; Potnis et al., 2011; Ryan et al., 2011; Hajri et al., 2012; Vandroemme et al., 2013; Guglielmini et al., 2014; Li et al., 2014; Cesbron et al., 2015; Dunger et al., 2016; Nascimento et al., 2016), were searched in the genomes.

The presence of the putatively virulence-associated plasmid pXap41 (Pothier et al., 2011b) was also evaluated in the analyzed genomes based in the nucleotide sequence similarity, graphically represented using the BLAST Ring Image Generator (BRIG) tool (Alikhan et al., 2011) and blastn was used for the sequence comparative analysis with an expected value threshold of 0.001.

In order to discriminate Xap from other atypical X. arboricola strains present in Prunus spp., a partial sequence of the xopE3 gene that is encoded on the pXap41 plasmid, described as specific for Xap (Pothier et al., 2011b), was used. Sequences of the xopE3 gene available in GenBank database from strains CITA 33 (GenBank locus tag DK27_00095), IVIA 2626.1 (AN652_04270), MAFF 301420 (XPR_2580), MAFF 301427 (XPN_1257) and CFBP 5530 (XAP_pXAP410005) were aligned with ClustalW and the consensus sequence used as template for xopE3 primers and probe design using the ABI PRISM Primer Express software v. 2 (Applied Biosystems, Foster City, CA). Specificity of the primers was firstly evaluated in silico using the Primer-BLAST tool available at NCBI. Graphical representation of the primers and the probe hybridization was performed in a set of X. arboricola genome sequences using the BRIG software.

Real-time PCR was conducted in a total volume of 25 μl containing, 12.5 μl of GoTaq probe qPCR MasterMix (Promega, Madison, WI, USA), 0.4 μM of each primer, 150 nM of TaqMan probe, and 5 μl of sample. Real-time PCR amplifications were performed in an ABI 7,500 Fast thermocycler (Applied Biosystems, Foster city, CA) and consisted of an initial denaturation step of 95°C for 5 min followed by 45 cycles, each one of 1 min at 95°C and 1 min at 59°C.

Specificity of the real-time PCR test was assessed in 99 bacterial strains, which comprised 54 strains of Xap, seven strains of Xap-look-a-like, ten strains from other pathovars of X. arboricola, 11 strains from other species of Xanthomonas, eight strains from other genera of phytopathogenic bacteria, and nine strains from the natural microbiota of Prunus spp. (Table S1). Bacterial suspensions of 10⁸ CFU/ml were treated at 95°C during 10 min and used for real-time PCR reactions. Sterile distilled water was used as negative control. Additionally, a real-time PCR protocol previously described (Palacio-Bielsa et al., 2011, 2015) was also applied on all the 99 bacterial strains tested for xopE3 gene.

Serial dilutions from 10 to 10⁸ CFU/ml of a 48 h LB broth culture of strain CITA 33 were prepared in sterile distilled water and heat-treated (95°C for 10 min) for real-time PCR reactions. Additionally, serial dilutions of pure bacterial DNA (QIamp DNA miniKit, Qiagen, Hilden, Germany), ranging from 0.001 pg/μl to 10⁸ pg/μl were also prepared in sterile distilled water. A volume of 5 μl of each dilution was used as template for amplification. Seven replicates of each sample were evaluated in each experiment, and the experiment was repeated in three independent assays. Appropriate negative controls containing no bacteria or no DNA were subjected to the same procedure. The limit of detection of the test was defined as the lowest target amount giving positive results in at least 15 of the 21 total reactions tested in the three independent assays (Caraguel et al., 2011). Analysis of variance was used to test for differences in the threshold cycles (C_ts) at each bacterial concentration in the three independent assays. Statistical analyses were performed by using Statgraphics Plus v.5.1 software. The amplification efficiency of the protocol for each kind of sample was calculated as described previously (Palacio-Bielsa et al., 2011, 2015). Linear regression curves representing the C_ts of each reaction were plotted against the logarithmic values of bacterial or DNA concentration. The slope of the curves (k) was used to determine the amplification efficiency (E) according to the equation E = 10^[−1/k], where E = 2 corresponded to 100% efficiency.

A total of 40 strains isolated from Prunus spp. and phenotypically similar to X. arboricola were initially identified as Xap by using a real-time PCR protocol (Palacio-Bielsa et al., 2011, 2015). Additionally, the plasmid pXap41, which is considered a specific feature of Xap, was not detected by PCR amplification of the genes repA1, repA2 and mobC (Pothier et al., 2011b) from strains CITA 14, CITA 42, CITA 49, CITA 51, CITA 124, and CITA 149 and, therefore, they were not considered as Xap but Xap-look-a-like strains.

In order to further characterize, six Xap-look-a-like strains mentioned above, and the Prunus-non-virulent strain CITA 44 (Garita-Cambronero et al., 2016b), were analyzed using a MLSA scheme based in partial sequences of the genes dnaK, fyuA, gyrB and rpoD (Young et al., 2008). The Maximum likelihood analysis of the concatenated sequences revealed that none of the Xap-look-a-like strains could be consistently clustered with any of the reference strains that belong to the pathovars described within X. arboricola. On the contrary, these strains were distributed in three separated clusters, one composed by strains CITA 14 and CITA 149, another composed by strains CITA 44, CITA 49 and CITA 51 and a third one composed by strains CITA 42 and CITA 124. These clusters were located in a basal phylogenetic position with respect to most of the strains used as reference. Sequence analysis of the concatenated sequence (2,833 nucleotide positions) revealed a mean similarity of 98.30 ± 0.2% between the Xap-look-a-like strains and the remaining ten strains of X. arboricola (Figure 1). As expected, according to other studies (Essakhi et al., 2015; Fischer-Le Saux et al., 2015), the phylogenetic clustering deduced from individual genes did not result in the same phylogenetic arrangement observed, which reinforced the need of a compendium of genetic characters as used in the MLSA (Figure S1).

Conventional PCR typing of 32 genetic determinants of the T3SS, its related T3SPs and T3Es, brought out a variable repertoire in the six Xap-look-a-like strains. The structural and regulatory components of the T3SS were only detected in strains CITA 14 and CITA 149, which harbored some of the 11 components tested. Similarly, only strains CITA 14 and CITA 149 harbored five and two T3SPs and T3Es, respectively. On the other hand, CITA 42, CITA 49, CITA 51 and CITA 124 did not harbor any of these genes. As expected, Xap strains CITA 33 and CFBP 5530, showed positive amplification from all the analyzed genes, while the non-virulent strain CITA 44 resulted negative (Table 2).

In addition to the variability found in the T3SS components, its related effectors and other secreted proteins, the ability of the Xap-look-a-like strains to cause disease symptoms after bacterial infiltration on leaves of barley, N. benthamiana, N. tabacum, tomato and the susceptible peach rootstock GF-305 was evaluated. None of the assayed strains were able to cause disease symptoms on barley, with the exception of strain CITA 14, which caused necrosis and chlorosis in the infiltrated zone. Strains CITA 14, CITA 149, and Xap strain CITA 33 were able to cause necrosis and chlorosis in N. benthamiana and N. tabacum. On the contrary, strains CITA 42, CITA 49, CITA 51, CITA 124 and the Prunus-non-virulent strain CITA 44 only showed a chlorosis effect after 21 dpi. In tomato at 21 dpi, all the assayed strains, with the exception of CITA 44, were able to cause necrosis and, in most of the cases, the necrotic area was surrounded by a chlorotic halo. Infiltration on peach rootstock GF-305 showed that only CITA 44 did not cause damage on the leaves; however, the remaining strains CITA 14, CITA 42, CITA 49, CITA 51, CITA 124 and CITA 149 caused necrosis in the infiltrated area after 7 dpi, but these necrotic spots did not expand beyond this zone and were different to typical bacterial spot symptoms.

On the other hand, the Xap strain CITA 33, caused necrosis in the infiltrated area and these necrotic zones were surrounded by a chlorotic halo (Figure S2). In addition, in GF-305, variation in bacterial populations on the infiltrated leaves was determined after 21 dpi, and all the Xap-look-a-like strains, as well as strain CITA 44, showed a reduction in their bacterial populations leading to concentrations equal or lower than 10⁵ CFU/ml. For strain CITA 33, the bacterial concentration increased from 10⁶ to 10¹⁰ CFU/ml by the end of the assay (Table 2). Positive results in real-time PCR analysis of the isolated colonies after 21 dpi, using the standardized protocol for Xap detection (Palacio-Bielsa et al., 2011, 2015), corroborated that the re-isolated strains corresponded to the same inoculated at the beginning of the study.

According to the results obtained in the MLSA analysis, one member of each one of the three clusters (CITA 14, CITA 44 and CITA124) observed for the Xap-look-a-like strains was selected for whole genome sequencing (Figure 1). The draft genome sequence of the Prunus-non-virulent strain CITA 44 has been reported and analyzed previously (Garita-Cambronero et al., 2016a). Moreover, main sequencing and structural features of CITA 14 and CITA 124 genome sequences have been previously announced (Garita-Cambronero et al., 2016c).

Draft genome sequence of CITA 14 was 4,864,444 bp in length with an average GC content of 65.60%; 3,870 over 3,974 genes predicted were identified as protein coding genes and a putative function was assigned to 2,991 of them (Table 3). This strain presented four rRNAs and 53 tRNAs. In the case of CITA 124, its draft genome sequence was 4,752,241 bp in length with an average GC content of 65.80%. A total of 4,004 genes were predicted and, among them, 3,798 were identified as protein coding genes with a putative function assigned to 2,838 of them. In addition, three rRNA and 50 tRNA genes were predicted (Table 3). From the total of genes with a predicted function, a COG functional category was assigned to 3,090 and 2,668 genes in CITA 14 and CITA 124, respectively. Those genes related to the amino acid and carbohydrate transport and metabolism, as well as those associated with translation and ribosomal structure and biogenesis, were predominant for both strains (Figure 2). Moreover, CITA 14 and CITA 124 presented 3,320 and 2,668 genes, respectively, with a protein domain in the Pfam database. Finally, a total of 625 and 942 genes of CITA 14 presented peptide cleavage signals and transmembrane helices, respectively. While, in CITA 124, 569 genes with peptide cleavage sites and 954 genes with transmembrane helices were predicted (Table 3, Table S3).

A comparative gene content analysis was performed among the genome sequences of the two Prunus-non-virulent Xap-look-a-like strains, CITA 14 and CITA 124, and 15 genome sequences of other strains of X. arboricola. As result, a total of 7,074 potential groups of homologous genes were found in the 17 analyzed genomes, from which 2,714 were shared by all the X. arboricola strains and comprised the core group of orthologus genes. CITA 14 and CITA 124 presented 76 and 124 unique cluster genes, respectively, which distinguished them from the other 15 strains (Figure 3A; Table S3). Strains CITA 14 and CITA 124, the pathogenic Xap strains CITA 33 and IVIA 2626.1 and the Prunus-non-virulent CITA 44, all isolated from Prunus spp. in Spain, shared 3,103 groups of homologous genes. A total of 889 cluster genes were found only in the non-virulent strains, while 708 cluster genes were found in the two Spanish strains of Xap (Figure 3B). Additionally, 236 CDS, 17 CDS, and 131 CDS, were unique for the pathovar corylina, juglandis and pruni, respectively (Table S3).

The mean of the gene content similarity among the analyzed genomes was 0.23 according to the Jaccard distance distribution, which means that the analyzed genomes shared a mean of 77% of their gene content, while the remaining 23% was unique for each one (Figure S3A).

Based in the gene cluster content of each genome, a principal component analysis showed that only 41.0% of the total difference was due to the variation by the two first principal components (Figure S3B). Three distinct clusters were elucidated, one of them was comprised by the Prunus-non-virulent or walnut strains (CITA 14, CITA 44, CITA 124, CFBP 7634 and CFPB 7651) and the strains with low-pathogenic-activity (3004, NCPPB 1630 and NCPPB 1832) which cause disease on barley and banana. The pathogenic strains of the pathovar pruni formed another cluster that comprised two subgroups, one formed by the strains isolated in Spain (CITA 33 and IVIA 2626.1) and another by the strains MAFF 301420 and MAFF 301427 isolated in Japan. Strains from the pathovar juglandis (CFBP 2528, CFBP 7179, Xaj2 and Xaj417) formed the third group. Finally, the strain NCCB 100457 from the pathovar corylina tended to group together with the strains isolated from walnut (Figure S3B).

The difference in the gene content cluster was illustrated using a pan-genome tree (Snipen and Liland, 2015) after computing the distance among the genomes using the Manhattan distance algorithm. The pan-genome tree for the 17 analyzed genomes (Figure 4) showed the same clustering organization that was visualized previously with the principal component analysis. Besides this, a division of the cluster comprised by the low-virulent and non-virulent strains was shown. A first group was composed by those strains that harbored components of the T3SS and T3Es, isolated from banana (NCPPB 1630 and NCPPB 1832), walnut (CFBP 7634 and CFBP 7651) and peach (CITA 14). A total of ten CDS differentiated this group from all the remaining clusters observed. The second group was comprised by the strain 3004 isolated from barley, and the strains CITA 44 and CITA 124 isolated from Prunus, and 27 CDS differentiated this cluster from all the other analyzed strains (Table S3). The same strains grouping and distribution was obtained using a Maximum likelihood phylogenetic analysis based in the concatenated sequence of the genes that comprised the core genome sequence of the 17 analyzed strains according to a sequence based methodology recently described (Page et al., 2015; Figure S4).

In addition to the gene content comparison, the profiles of genetic components associated with pathogenesis were determined for CITA 14 and CITA 124 and compared to those in other strains of X. arboricola isolated from Prunus spp. in Spain, especially in the two pathogenic strains of Xap, CITA 33, and IVIA 2626.1, and in the Prunus-non-virulent strain of X. arboricola CITA 44.

Regarding the profiles of cell wall degrading enzymes, a total of ten genes that encoded for pectolytic enzymes were found in the five strains. CITA 14 and CITA 124 showed seven and eight of these genes, respectively. For these enzymes, two orthologs, NP_635517.1 and NP_635516.1, were shared by the Prunus-non-virulent strains, meanwhile the degenerated pectate lyase AAM37225.1 was only found in Xap strains CITA 33 and IVIA 2626.1 (Table S4). Regarding the profile of cellulolytic enzymes, nine of them were shared over 11 genes found in all the strains. In this case, only the presence of the cellulase AAM38359, described in X. citri subsp. citri 306, differentiated non-virulent strains from Xap (Table S4). In the case of the hemicellulolytic enzymes, a total of 11 orthologs were found in the Prunus-associated strains; pathogenic strains of Xap were differentiated from the non-virulent strains due to the presence of the genes that encoded the xylanase NP_638385.1 and the xylosidase/arabinosidase NP_637752.1, both described previously in X. campestris pv. campestris strain ATCC 33913. Finally, orthologous genes for the virulence associated lipases NP_638797.1 and AAO29541.1, from X. campestris pv. campestris ATCC 33913 and Xylella fastidiosa strain Temecula, were found in the five genomes (Table S4).

The profiles of genes related to sensing and chemotaxis varied among the analyzed strains. Four of the five Spanish Prunus-isolated strains presented the same gene profile for those genes associated with chemotaxis. However, CITA 124 did not harbor homologous genes to cheD (AAM36751.1), cheZ (AAM36793.1), and cheA (AAM36792.1), described in X. citri subsp. citri 306. Variants in other sensing mechanisms, such as TBDTs, were found in the Prunus-associated strains of X. arboricola. From the 17 TBDTs encoding genes found, those homologs of the proteins NP_635515.1, NP_635700.1, NP_635699.1 and NP_639391.1, initially described in X. campestris pv. campestris ATCC 33913, differentiated the atypical strains from Xap. Additionally, a large repertoire of 60 genes associated with STCRs was found in the analyzed genomes and, from them, 55 were shared for all the strains isolated from Prunus. In addition, the STCRs AAM36681.1, AAM35218.1, AAM37649.1 and NP_637535.1 were only present in the Prunus-non-virulent strains CITA 14, CITA 44, and CITA 124. Finally, from a total of 26 MCPs genes, 11 were found in all Prunus-associated strains, but the absence of an ortholog to CAJ23610.1, described in X. campestris pv. vesicatoria 85–10, in strains CITA 14 and CITA 124 differentiated them from the remaining strains (Table S4).

Besides to those genes related to environmental sensing, variations in some other genes associated with the initial steps of the pathogenesis process, such as motility, attachment, biopolymerization of the xanthan gum, and the inter-cellular cross-talk process controlled by the quorum-sensing system, were also found (Table S4). Pathogenic and non-virulent strains of stone fruit trees and almond shared 35 orthologs associated with molecular components of the flagellar system. The exception to this was the strain CITA 124, which did not have homologous genes to the flagellar components of X. citri subsp. citri 306, flhF (AAM36797.1), fliH (AAM36814.1), fliJ (AAM36812.1) and motB (AAM38537.1). In addition, an interesting polymorphism was observed in the flagellin protein, encoded by fliC, of non-pathogenic strains. In CITA 14 and CITA 124, this protein was identical to protein WP_024939608.1, which has been previously associated with all the non-virulent strains of X. arboricola or with low virulent strains of the pathovar celebensis. Pathogenic strains of Xap harbored a flagellin protein identical to protein WP_039814449.1, which present a substitution of aspartic acid for valine in the amino acid 43 of the N-terminal region that has been associated to pathogenic strains in other species of Xanthomonas (Sun et al., 2006; Cesbron et al., 2015).

Another bacterial structure related to motility, as well as to attachment, is the type IV pilus. The pathogenic Xap strains CITA 33 and IVIA 26262.1 harbored 25 orthologs to the 31 genes described in X. citri subsp. citri, while the atypical strains CITA 14, CITA 44, and CITA 124 were differentiated for the absence of orthologs to the genes fimA (AAM38084.1), fimT (AAM37516.1), pilV (AAM37515.1), pilW (AAM37514.1), pilX (AAM37513.1) and pilY1 (AAM37512.1). With regard to the attachment function carried out by the non-fimbrial adhesins, most of the Prunus-non-virulent strains shared five of the six genes found, with the exception of CITA 14, which did not harbor a homolog to fhaB1 of X. campestris pv. vesicatoria 85–10 (CAJ23537.1). Presence of a homologous gene to fhaB2 (CAJ23538.1) of X. campestris pv. vesicatoria differentiated Xap from the atypical strains isolated from Prunus spp.

Pathogenic and non-virulent strains of X. arboricola isolated from Prunus spp. shared the same profile of xanthan-associated genes, which are involved in bacterial attachment and biofilm formation. None of the strains had homologous sequences to gumG (NP_637802.1) which was found in other xanthomonads (Lee et al., 2005). Regarding quorum sensing system, which is associated with the regulation of the pathogenic activity, all the analyzed strains, with the exception of CITA 14, harbored the same gene pattern conformed by 11 of the 12 genes associated to this process in Xanthomonas (He and Zhang, 2008). In addition, CITA 14 harbored an ortholog to the transcriptional regulator NP_636589.1 described in X. campestris pv. campestris ATCC 33913 (Table S4).

Bacterial type II, III, and IV secretory systems (T2SS, T3SS, and T4SS), which are related to the secretion of proteins and DNA, also play a crucial role in pathogenesis (Ryan et al., 2011). Regarding to T2SS, CITA 14 and CITA 124 presented 19 and 18 orthologs, respectively, of the 23 genes associated with xcs and xps T2SS gene clusters described in Xanthomonas (Filloux, 2004; Szczesny et al., 2010). The only difference among atypical strains and Xap was the presence in the latter of an ortholog to the gen xcsK (NP_638764.1; Table S4).

Pronounced differences were observed among pathogenic and non-virulent to Prunus strains regarding the gene profile associated with T3SS and its related effectors, as well as with T4SS. T3SS-related gene profile in CITA 14 was comprised by 24 orthologs of the 28 T3SS described in Xanthomonas, and this profile was the same observed in pathogenic strains of X. arboricola (Cesbron et al., 2015; Garita-Cambronero et al., 2016a). On the other hand, CITA 124 only harbored four T3SS-related genes (hpaS, hpaR2, hrpG and hrpX) which correspond to the regulators of this secretory system (Jacobs et al., 2015). For this strain, as for the Prunus-non-virulent strain CITA 44, none of the genes that composed the macromolecular structure of the T3SS were found (Table S3). Regarding to the T3SS related effectors, from the 61 T3Es and other T3SPs described in Xanthomonas, the genome sequence of CITA 14 presented a total of six orthologs to the genes avrBs2, hpaA, hrpW, xopA, xopF1 and xopR, while genome sequence of CITA 124 did not present any of these effectors. Moreover, X. arboricola strains isolated from Prunus showed variants in the number of T4SSs. Most of the strains, regardless of their pathogenic activity, contained ten of the 12 components associated with the VirB/VirD4 T4SS of Agrobacterium tumefaciens (Christie, 2004). In addition, the absence of orthologs to the core components associated with the type IV conjugation cluster tfc, described in Haemophilus influenzae, in the strains CITA 14 and CITA 124, differentiated them from the Xap strains.

Finally, comparative sequence analysis among the nucleotide sequence of the plasmid pXap41 and the draft genome sequence corroborated the absence of this plasmid in both non-virulent strains (Figure S5).

In silico analysis of the primers XopE3F (5′-TCAGCGATCACGCATCCA-3′), XopE3R (5′-CGCACCAGATCGACAAACAC-3′) and the probe XopE3p (5′-6-carboxyfluorescein [FAM]-CATGCGCAGGCCGCACAT-[TAMRA]-3′), indicated that they were able to amplify the gene xopE3 in X. arboricola only in those sequences from the pathovar pruni. Sequence analysis on the available complete genome sequences of X. arboricola showed that xopE3 was only present in those strains of the pathovar pruni (Figure S6). In addition, this set of primers and the designed probe were also able to amplify the xopE3 gene in other species of Xanthomonas such as one strain of X. campestris (IVIA 2734.1), one strain of X. citri subsp. citri (306) and X. fuscans subsp. fuscans strains NCPPB 381 and IVIA 151835DA (Table S1).

Besides the nucleotide sequence-based analysis, the specificity of the real-time PCR assay was conducted by testing the protocol on the bacterial strains listed in Table S1. Among the X. arboricola strains, only those identified previously as Xap, by the presence of the plasmid pXap41, and with a positive result for the standardized real-time protocol based in the gene ftsX of the ABC transporter in Xap (Palacio-Bielsa et al., 2011, 2015), presented consistent positive results. None of the seven Xap-look-a-like strains (CITA 14, CITA 42, CITA 44, CITA 49, CITA 51, CITA 124 and CITA 149) showed positive results from this PCR. Undesired specific PCR results for xopE3 were observed from one strain of X. campestris (IVIA 2734-1), three strains of X. citri subsp. citri (306, IVIA 2889-1 and IVIA 3026-1), one strain of X. hortorum pv. pelargonii (CITA Xp-2), two strains of X. fuscans subsp. fuscans (NCPPB 381 and IVIA 151835DA), and the strain IVIA 3619-1 of X. vesicatoria (Table S1). When the analyzed strains were amplified using the real-time PCR protocol for the ABC transporter-associated gene ftsX (Palacio-Bielsa et al., 2011, 2015), positive results were obtained with all the strains of Xap, the seven strains of Xap-look-a-like, two strains of X. arboricola pv. corylina (CFBP 1846 and IVIA 3978) and the strain of X. citri subsp. citri 306. Double positive PCR results, using xopE3 or ABC primers, were only observed for all the Xap strains but also for the strain 306 of X. citri subsp. citri which is unlikely to be found in Prunus spp. (Table S1).

No significant differences were found among the three independent assays conducted to determine the sensitivity of the real-time PCR protocol to amplify xopE3 using heat-treated cells or purified DNA as samples. Calibration curves, obtained from serial dilutions of heat-treated cells of Xap strain CITA 33, demonstrated that the real-time PCR assay showed a sensitivity of 10 CFU/ml or 100 pg/μl of DNA, with a PCR efficiency of 2.2 ± 0.22 or 1.8 ± 0.03 for bacterial cells or purified DNA, respectively (Figure 5).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.