In this study, by using a combined strategy (Figure 1) that the three well-established gene finders GLIMMER (http://cbcb.umd.edu/software/glimmer) 8 , GeneMark 21 , and ZCURVE 22 were respectively applied to predict putative protein coding sequences (CDSs) within the two genomes of Xcc strains 8004 and ATCC33913 19 20 , a total of 7164 CDSs were identified after further sequence analaysis (Figure 2). Among them, 4,314 CDSs, including 146 Xcc strain 8004-specific CDSs, 60 ATCC33913-specific CDSs and 4108 shared CDSs between the two genomes, have been previously annotated (Figure 2A). The remaining 2850 predicted CDSs have not been identified in the published two genomes and were defined as putative new CDSs (Figure 2A), including 1181 CDSs by GLIMMER, 957 CDSs by GeneMark, and 612 CDSs by ZCURVE (Figure 2B). Intriguingly, there were only 126 overlapping CDSs predicted by all the three gene finders (Figure 2B). The size of these putative CDSs ranged from 90 to 4545 bps, and most of them (2202 of 2850 CDSs) were less than 1 kb long (Figure 2A). In particular, 797 CDSs were less than 180 bp in length. BLASTN analysis revealed that 2410 of the 2850 putative new CDSs were located at intergenic regions of both strands (Figure 2A, indicated by "I" and "III") in the chromosome of Xcc strain 8004 (Figure 2A). The remaining 440 CDSs were partially or fully overlapped with the annotated genes, but within different reading frames (Figure 2A, indicated by "II" and "IV"). All of 648 putative new CDSs >1000 bp in length were either antisense or overlapping to the annotated genes in two Xcc genomes (Additional file 1).

The set of 2850 putative new CDSs was probably contaminated by pseudogene fragments and false-prediction artifacts because all the 3 gene finders are entirely based on intrinsic evidence. To find true CDS, the next strategy used in this study was to get support by extrinsic evidence. All the putative new CDSs were blasted for similar entries within the NCBI non-redundent database by means of BLASTP. Based on the three criteria described in Materials and Methods, a total of 220 putative new CDSs were found to be significantly similar to other protein sequences in the database (Additional file 1).

More recently, the genome sequence of Xcc strain B100 has been published and the genome contained 496 additional CDSs 14 . About half of the these CDSs that were identified by the combined use of the gene finders GISMO 23 and REGANOR 24 were also present in the genomes of Xcc strains 8004 and ATCC33913, but have not been annotated 14 . Comparing the 2850 putative new CDSs identified in this study with the 496 additional CDSs in Xcc strain B100, we found an overlapping 72 CDSs (Additional file 1). Among them, 14 CDSs had more than one homologs in non-redundant database and have been included in the 220 putative new CDSs identified by similarity searching; the remaining 58 CDSs had no homologs in non-redundant database except in Xcc strain B100 and were also regarded as new CDSs in this study (Additional file 1). Taken together, a total of 278 CDSs were screened out of 2850 putative new CDSs by extrinsic evidence.

The majority of these CDSs (240 of 278) encodes conserved hypothetical proteins or hypothetical proteins (Figure 3). Eleven CDSs (Xcc_CDS105, Xcc_CDS107, Xcc_CDS411, Xcc_CDS1381, Xcc_CDS1831, Xcc_CDS2249, Xcc_CDS2324, Xcc_CDS2391, Xcc_CDS2668, Xcc_CDS2723, Xcc_CDS2777) encode putative secreted or exported proteins and three CDSs encode regulatory protein or transcription factors (Figure 3). Xcc_CDS002 encodes a Sir2-like transcriptional silencer protein; Xcc_CDS1553 encodes an ArsR family transcriptional regulator; Xcc_CDS1633 bears similarity to the Homeodomain of POU domain proteins or HTH_XRE domain proteins (Additional file 1). Two CDSs (Xcc_CDS2171 and Xcc_CDS2691) encode putative phenol hydroxylases and another 2 CDSs (Xcc_CDS2201 and Xcc_CDS2211) encode putative 50S ribosomal proteins. The remaining 20 CDSs respectively encode peptidase-like protein (Xcc_CDS073), hemolysin III (Xcc_CDS095), IS480b transposase (Xcc_CDS177), putative cell wall surface anchor family protein (Xcc_CDS346), endonuclease (Xcc_CDS528), chloramphenicol O-acetyltransferase (Xcc_CDS639), outer protein D (Xcc_CDS900), ABC transporter heme permease (Xcc_CDS1309), putative GTPase (Xcc_CDS1342), putative DNA methylase (Xcc_CDS1416), transmembrane protein (Xcc_CDS1446), putative tryptophan-rich sensory protein (Xcc_CDS1617), dihydroxydipicolinate synthase (Xcc_CDS1689), thermoresistant gluconokinase (Xcc_CDS1836), thiopurine methyltransferase (Xcc_CDS1899), putative tryptophan 2,3-dioxygenase oxidoreductase (Xcc_CDS2015), putative Atu protein (Xcc_CDS2546), WD40-like beta propeller (Xcc_CDS2674), ferric pseudobactins receptor protein (Xcc_CDS2714), and restriction endonuclease (Xcc_CDS2849) (Figure 3; Additional file 1).

An alternative approach to validate a CDS is to detect the transcribed mRNA. An oligonucleotide microarray chip, which contains 50-mer oligos specific for 4080 annotated CDSs and 8 negative controls, has been successfully used to analyze the DSF regulon, Clp regulon and RavR regulon in Xcc 25 26 27 . In this study, a new microarray chip with the above-mentioned oligos and additional oligos specific for the 1724 putative new CDSs was constructed. This microarray chip was used to detect transcripts of the putative new CDSs. To detect transcripts under different conditions, total RNA was extracted from cell culture grown under the following conditions: (i) different cell density: OD₆₀₀ = 1.0, 1.6 and 2.0; (ii) different genetic backgrounds: ΔrpfF strain, ΔrpfC, Δclp and ΔravR 25 26 27 ; (iii) different media: rich YEB medium and poor NYG medium. By using the screening procedures described in Materials and Methods, 147 putative new CDSs were found with detectable transcripts (Figure 4A; Additional file 1). Further analysis revealed that 75 CDSs were constitutively expressed during the growth and the remaining 72 CDSs were only expressed at high cell density (OD₆₀₀ = 2.0) (Figure 4A). Comparing the global gene expression profiles of Xcc wild type, rpfF, rpfC, and clp deletion mutants, we found that the transcription of 15 high cell density-dependent CDSs was also positively regulated by the quorum sensing signal DSF 25 . The expression levels of these CDSs in an rpfF deletion mutant were respectively 2.7 to 5.2 times lower than those in the wild type XC1 strain (Figure 4B). The transcription of Xcc_CDS2497 was only induced in poor NYG medium at higher density (OD₆₀₀ = 2.0) (Figure 4A).

In order to go further in the validation of our microarray-based method for selecting true CDS, and as we are more interested in DSF signal-regulated CDSs, we chose the 15 DSF signal-regulated CDSs for further transcriptional analysis by reverse transcription PCR. The products of 11 CDSs could be amplified by using total RNAs extracted from cell culture at OD₆₀₀ = 2.0 (Figure 4B). The resultant RT-PCR products were further verified by sequencing analysis (data not shown). RT-PCR analysis also verified the transcriptional difference of the 11 new CDSs between wild type and rpfF deletion mutant (Figure 4B).

While extrinsic evidence supported the presence of 278 new CDSs, and while transcriptional analysis indicated 147 new CDSs with detectable transcripts, a comparison of the two sets of new CDSs revealed a total of 119 overlapping CDSs that were identified by both approaches (Figure 5). Thus, a total of 306 (278+147-119) CDSs got support by extrinsic evidence or/and experimentally transcriptional analysis, suggesting that they are probably true CDSs. The remaining 2544 putative new CDSs failed to get support by extrinsic evidence or transcriptional analysis (Figure 5). Two of the overlapping 119 CDSs, Xcc_CDS002 and Xcc_CDS1553, which both encoded putative transcription factors, were chosen for further experimental characterization. The results are presented in the following sections.

Xcc_CDS002 is a new CDS of 855 bps in length. It encodes a protein with a conserved silent information regulator 2 (SIR-2) or SIR2-like domain (Figure 6A), which has been found to confer NAD-dependent protein deacetylase activity in eukaryotes 28 29 . For the convenience of discussion, Xcc_CDS002 was renamed as sir2x for

SIR2

Xcc_CDS1553 encodes a 122-aa protein with a conserved HTH_ARSR domain (Figure 7A), which occurs in arsenical resistance operon repressors and similar prokaryotic, metal-regulated homodimeric repressors that belong to the ArsR superfamily of bacterial transcription-regulatory proteins 30 31 . For the convenience of discussion, this CDS was renamed as arsR. Interestingly, arsR was only found in the genome of Xcc strain 8004, not in Xcc strains ATCC33913 and B100. In the Xcc 8004 genome, arsR is located upstream of XC_2295 and XC_2294, which respectively encode a putative high-affinity Fe²⁺/Pb²⁺ permease and an arsenite efflux pump AcR3 (Figure 7B). arsR and XC_2295 were separated by 64 bps and the gap between XC_2294 and XC_2295 was 83 bps (Figure 7B; 20). Further RT-PCR analysis showed that arsR, XC_2295, and XC_2294 belong to the same operon (Figure 7C), suggesting that ArsR, XC_2294 and XC_2295 might be functionally related. To further confirm this hypothesis, an arsR in frame deletion mutant termed ΔarsR was generated in Xcc strain 8004. The results showed that the ΔarsR strain was much more sensitive to arsenate than the wild type strain (Figure 7D). On LB plates with 0.5 mM arsenate, the wild type strain Xcc 8004 grew well, while in contrast, the deletion mutant did not grow at all on this medium (Figure 7E). The mutant phenotype could be reverted by complementation with a plasmid carrying the coding region of arsR, demonstrating that the observed phenotype was due to arsR.

Xcc strains XC1 and 8004 were grown at 30°C with shaking (250 rpm/min) in YEB, LB or NYG medium as described by He et al. 25 . E. coli strains were grown at 37°C in LB medium. Antibiotics were added at the following concentrations when required: kanamycin, 100 μg/ml, rifampicin, 25 μg/ml, and tetracycline, 10 μg/ml.

Complete genome records of the Xcc strains ATCC33913 and 8004 19 20 were downloaded from the NCBI Microbial genome database (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi?view=1). Gene prediction was conducted by the gene finders GLIMMER 2.03 8 , GeneMark 21 and ZCURVE 22 . For the prediction, the minimum length of CDS was set as 90 bp. BLASTN (http://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to find the locations of all the putative new CDSs in the genomes of Xcc strain 8004 and ATCC33913. Multiple sequence alignment analysis was performed using CLUSTAL W (1.83) (http://sbcr.bii.a-star.edu.sg/clustalw/). Domain architecture analysis was performed using the SMART database application (http://smart.embl-heidelberg.de/). The nucleic acid sequences of two well-studied regulator sir2x and arsR have been deposited in the NCBI GeneBank database and the accession numbers are JF966390 and JF966391.

The amino acid sequences of all 2850 putative new CDSs were submitted for BLASTP analysis. Homologs in the nr database were selected on the basis of the following three criteria. Firstly, only the subjects with E-values lower than 10^-4 were considered hits. Secondly, the subjects should have similar sizes as the queries. Thirdly, for each query there should be more than one matched subject unless the E-value is very low (less than 10^-30).

Based on the annotated genome sequences of the Xcc strains ATCC33913 and 8004 19 20 , we used a CDS-specific oligonucleotide selection algorithm 41 to successfully design unique 50-mer oligonucleotides for 1724 putative new CDSs. The majority of these CDSs were more than 300 bps in length. As specificity controls, 50-mer oligonucleotides were also designed based on the sinat5 (NCBI No.: AF480944) and nac1 (NCBI No.: AF198054) genes of Arabidopsis thaliana, and the genes rag1 (NCBI No.: NM_131389) of zebrafish and the olf1 (NCBI No.:U56420) of Homo sapiens 25 . Thus, a total of 5770 CDS-specific oligonucleotides representing 4042 annotated CDS 25 , plus 1724 putative new CDSs, and 4 specificity controls were used for the oligonucleotide microarray chip preparation. Oligonucleotides were synthesized at a 50 nmol scale by Operon Technologies (Alameda, CA, USA). The protocol employed for constructing the oligo-chip has been previously described 25 . Briefly, all oligos were dissolved in saline sodium citrate buffer (3 × SSC) to a final concentration of 40 μM. Oligo samples were arrayed with Pixsys 5500XL Arrayer (Cartesian) to poly-L-Lysine-coated microscope slides. DNA samples were fixed by rehydration, snap-drying and UV cross-linking. The remaining poly-L-Lysine on the slides was rendered non-reactive by treatment with blocking solution (150 mM succinic anhydride in 1-methy-2-pyrrolidinone, buffered with 85 mM sodium borate, pH 8.0) for 30 min. After washing with water, the array plates were rinsed with 95% ethanol and dried.

Bacterial cells were collected by centrifugation at 4°C for 5 min at 10,000 rpm. Total RNA samples were prepared by using RNeasy midi columns following the manufacturer's instructions (Qiagen). RNA integrity was confirmed by electrophoresis using a 1.3% formaldehyde agrose gel. The quality of DNA-free RNA was monitored by PCR and RT-PCR analysis of at least two known genes. Cy3- or Cy5-labeled cDNA was generated by using random hexamers as primers for reverse transcription (Invitrogen). cDNA labeling, purification and hybridization against the microarray were conducted as previously described 25 . Slides were scanned for the fluorescent intensity using a ScanArray 5000 laser scanner. The signal intensities were quantified by using the software ImaGene 5 (BioDiscovery). Hybridization signals were normalized using the scale normalization procedure previously described 25 . Each treatment was repeated three times and the data presented were the means of two representative replicates. The fold changes were then calculated from the normalized log ratios.

In this study, oligonucleotide microarray analysis was used to detect transcription, so as to confirm the functionality of the putative new CDSs. The putative CDSs with detectable transcript was identified using the normalized signal median of the corresponding probe. To calculate the normalized signal median, firstly the average signal median S₀ of 8 negative control probes representing 4 Arabidopsis and zebrafish genes 25 was determined by using the following formula: S₀ = ∑(S_AZ-B_AZ)/8, where S_AZ indicates the signal median of the negative control probe and B_AZ indicates the corresponding background signal median. Secondly, the normalized signal median (S) of the putative new CDSs was calculated following the formula: S = S_CDS - B_CDS -S₀, where S_CDS indicates the signal median of the putative new CDS and B_CDS indicates the background median of the putative new CDSs. Finally, if S >0, it is regarded as CDS with detectable transcript.

RT-PCR analysis was conducted using a QIAGEN^®OneStep RT-PCR Kit following the manufacturer's instructions. The primers used for RT-PCR analysis are listed in Additional file1. Total RNAs were extracted from bacterial culture grown in YEB medium at OD₆₀₀ = 2.0 and a total of 200 ng of total RNA was used for reverse transcription. The cycle number differed in the amplification of different CDS products.

Spontaneous rifampicin-resistant derivatives of strain XC1 or 8004 were used as parental strains for generation of deletion mutants. In-frame deletion of Xcc_CDS002 (sir2x) and Xcc_CDS1553 (arsR) was conducted using the primers listed in Additional file 1 following the methods described previously 25 . For complementation analysis, the coding regions of sir2x and arsR respectively were amplified by PCR using the primers listed in Additional file 1 and cloned under the control of lac promoter in expression vector pLAFR3. The resultant constructs were transferred into Xcc strains through triparental mating.

The extracellular cellulase and protease activity and EPS production in the culture supernatants of Xcc strains at OD600 = 2.3 were measured according to the methods described previously 25 . The virulence of Xcc to Chinese cabbage was determined following the scissors-clipping method described previously 26 . Fifteen plants were inoculated for each bacterial strain and the experiment was repeated three times.

Sodium arsenate (SIGMA) was added in the following final concentrations (mM): 0.10, 0.25, 0.50, 0.75 and 1.00. Fifty microliters of fresh culture of Xcc strain 8004 were inoculated into 5 ml of NYG liquid media with rifampicin (25 μg/ml) and sodium arsenate at different concentrations and grown at 28°C with shaking (250 rpm/min) for overnight. Bacterial growth was indicated by measuring the optical density at 600 nm.