Identification of a novel type III secretion-associated outer membrane-bound protein from Xanthomonas campestris pv. campestris

Many bacterial pathogens employ the type III secretion system (T3SS) to translocate effector proteins into eukaryotic cells to overcome host defenses. To date, most of our knowledge about the T3SS molecular architecture comes from the studies on animal pathogens. In plant pathogens, nine Hrc proteins are believed to be structural components of the T3SS, of which HrcC and HrcJ form the outer and inner rings of the T3SS, respectively. Here, we demonstrated that a novel outer membrane-bound protein (HpaM) of Xanthomonas campestris pv. campestris is critical for the type III secretion and is structurally and functionally conserved in phytopathogenic Xanthomonas spp. We showed that the C-terminus of HpaM extends into the periplasm to interact physically with HrcJ and the middle part of HpaM interacts physically with HrcC. It is clear that the outer and inner rings compose the main basal body of the T3SS apparatus in animal pathogens. Therefore, we presume that HpaM may act as a T3SS structural component, or play a role in assisting assembling or affecting the stability of the T3SS apparatus. HpaM is a highly prevalent and specific protein in Xanthomonas spp., suggesting that the T3SS of Xanthomonas is distinctive in some aspects from other pathogens.

Scientific RepoRts | 7:42724 | DOI: 10.1038/srep42724 effector proteins from animal pathogens and plant pathogen effectors can be secreted by the T3SS of animal pathogens 9,10 . Based on these facts, it is presumed that the Hrc proteins are the components of the T3SS in plant pathogens and the core T3SS apparatus may be conserved among plant and animal pathogens 6,11 . According to their homology to the T3SS components of animal pathogens, the function of the nine conserved Hrc proteins is believed to be: (1) HrcC is an outer membrane ring protein; (2) HrcJ is an inner membrane ring protein; (3) HrcR, S, T and U are integral inner membrane proteins with periplasmic extensions, taking part in the rod formation of the T3SS apparatus; and (4) HrcV, Q and N are inner membrane or peripheral cytoplasmic proteins engaged in initiation of effector secretion from the cytoplasm 6,8,[11][12][13][14] .
Xanthomonas is a large genus of Gram-negative bacteria, which comprises 27 species and some of which include multiple pathovars. Many members of the genus are important plant pathogens, such as X. campestris pv. campestris (the crucifer black rot pathogen), X. citri subsp. citri (the citrus canker pathogen), X. euvesicatoria (the pepper and tomato bacterial spot pathogen), X. oryzae pv. oryzae (the rice bacterial blight pathogen), and X. oryzae pv. oryzicola (the rice bacterial leaf streak pathogen), and most of which rely on an efficient T3SS for their pathogenicity 15,16 . The T3SS-encoding hrp cluster of Xanthomonas spp. consists of six operons (hrpA to hrpF) which harbor more than 20 different genes including the nine conserved hrc genes [17][18][19] . Recently, we identified a novel outer membrane-bound protein that is involved in the HR and pathogenicity of X. campestris pv. campestris (Xcc), which was designated as HpaM (for Hrp-associated membrane-bound protein). Here, we present evidences showing that the protein is essential for type III secretion and conserved in Xanthomonas spp.

Results
HpaM is essential for the virulence and HR induction of Xcc. In our previous work, we isolated a large number of Xcc mutants from a library constructed by the transposon Tn5gusA5 insertion in the genome of Xcc wild-type strain 8004. One of the mutants, 083E12, was due to a Tn5gusA5 insertion in the ORF XC_2847 (named hpaM in this study). Plant tests showed that the mutant strain 083E12 almost completely lost virulence and hardly induced any disease or HR symptoms in the host plant Chinese radish or the non-host plant pepper (cultivar ECW-10R). The gene XC_2847 was annotated to be 1161 bp in length, locating at nucleotide (nt) positions from the 3426325 th to the 3427485 th nt, and predicted to encode a hypothetical protein 20 . Using a standard 5′ -RACE method, the transcription initiation site (TIS) of XC_2847 was mapped at 89 nucleotides downstream of the predicted translational start codon GTG (Fig. S1). There is an in-frame ATG codon 22 bp downstream of the determined TIS (Fig. S1). Based on these data, we propose that the XC_2847 ORF should start with the ATG and consist of 1050 bp instead of 1161 bp.
To facilitate further studies on the function of hpaM, a deletion mutant, named Δ hpaM, was constructed by using the suicide vector pK18mobsacB (Table S1). Simultaneously, a complemented strain was also constructed by introducing the recombinant plasmid pLChpaM, which carries an entire hpaM gene, into the mutant Δ hpaM. The resulting complemented strain was named as CΔ hpaM (Table S1). As anticipated, the mutant Δ hpaM could hardly induce visible disease or HR symptoms (Fig. 1). However, the complemented strain CΔ hpaM could produce wild-type disease and HR symptoms (Fig. 1), suggesting that the pathogenicity and HR of Δ hpaM could be restored by hpaM in trans. The growth in planta of the hpaM mutant was suppressed significantly, although its growth rate was not affected in minimal medium (Fig. S2), suggesting that mutation in hpaM decreased significantly fitness in planta. Taken together, the above data indicate that HpaM is essential for the virulence and HR induction of Xcc.
HpaM is required for T3Es secretion of Xcc. As mentioned above, the T3SS is critical for the pathogenicity and HR induction of Xcc. To gain an insight into the mechanisms by which HpaM affects the virulence and HR induction, we examined whether HpaM is involved in the T3SS. The T3SS of Xcc is encoded by six hrp operons (hrpA to hrpF) and the expression of the hrp operons is positively controlled by several key regulators including HrpG and HrpX [21][22][23] . To determine whether HpaM influences the expression of hrp genes, the plasmid-driven promoterless β -glucuronidase (gusA) transcriptional fusion reporters of hrpG and hrpX regulators as well as the six hrp operons, in which a DNA fragment containing the promoter region of each of the hrp operons (hrpA to hrpF) and hrpG and hrpX genes fused to the promoterless gusA gene with its ribosome binding site (RBS) was cloned into the vector pLAFR6 (Table S1), were introduced from E. coli JM109 by triparental conjugation into the hpaM mutant Δ hpaM and the wild-type strain 8004, and transconjugants (reporter strains) were screened on NYG medium as described previously 22 . As the expression of the hrp genes is induced in minimal media but inhibited in rich media 23 , β -glucuronidase (GUS) activities produced by the obtained reporter strains (Table S1) were measured after cultivation in MMX minimal medium. The results revealed that each of the reporters produced similar GUS activity in wild-type and hpaM deletion backgrounds (Fig. S3A), suggesting that mutation of hpaM did not affect the expression of the hrp genes. To clarify whether the expression of hpaM is subject to HrpG and HrpX regulation, the promoter-gusA transcriptional fusion reporter of hpaM was constructed. A 404-bp DNA fragment upstream of the hpaM ORF, amplified from the wild-type strain 8004, was fused with the coding region of promoterless gusA gene and cloned into pLAFR6, generating the reporter plasmid named pGUShpaM (Table S1). The GUS activities produced by the reporter plasmid in wild-type background and hrpG or hrpX mutation background were not significantly different (P = 0.05 by t test) (Fig. S3B), indicating that the expression of hpaM is not controlled by HrpG and HrpX. In addition, the reporter plasmid pGUShpaM in hpaM mutation background and wild-type background produced similar GUS activities (Fig. S3B), implying that HpaM plays no impact on its own expression.
We further investigated whether HpaM is involved in T3Es secretion. It is well known that T3Es have a modular structure and the targeting signal generally resides in the N-terminal 50 or 100 amino acids (aa) 24 . Two reporter plasmids, pGUSavrAC and pGUSxopN (Table S1), were employed to study the secretion efficiency of Xcc T3SS. The reporters were previously constructed by fusing the promoterless gusA gene with a fragment including the promoter and targeting signal-encoding region of avrAC (XC_1553) or xopN (XC_0241), which encode the T3Es AvrAC and XopN, respectively 25,26 . pGUSavrAC and pGUSxopN were introduced into the hpaM mutant strain Δ hpaM and the wild-type strain 8004, respectively. The plasmids were also introduced into the hrcV-deficient mutant strain Δ hrcV as negative controls. HrcV is a conserved inner membrane protein of the core T3SS and the mutant Δ hrcV is defective in type III secretion 13 . The recombinant plasmid pL6gus, which was constructed by cloning a 1,832-bp promoterless gusA ORF into the promoterless cloning site of the plasmid pLAFR6, was introduced into the wild-type strain 8004 and the resulting strain 8004/pL6gus, which did not produced any significant GUS activity, was used as a negative control for the GUS assay. As shown in Fig. 2A,B, both reporters produced large amount of GUS activity in the wild-type and hpaM mutation backgrounds; however, the GUS activities in the cultural supernatants of the hpaM mutation background strains were significantly lower than those in the cultural supernatants of the wild-type background strains (P = 0.01 by t test), implying that mutation of hpaM significantly diminished the secretion of the T3Es AvrAC and XopN.
To further verify the effect of hpaM on the type III secretion, western blot assay was performed to examine the secretion of the T3E AvrAC in the hpaM mutation background. For this purpose, an avrAC deletion mutant (∆ avrAC) and an avrAC/hpaM double deletion mutant (∆ avrAC-hpaM) were constructed. Another double deletion mutant (∆ avrAC-hrcV) that lacked avrAC and hrcV was also constructed and used as a negative control strain. The recombinant plasmid pRavrACH6, which was constructed by fusing 6× His-tag coding sequence to the 3′ end of the avrAC gene with its own promoter and cloning the fused fragment into the Figure 1. HpaM is essential for pathogenicity and HR induction of Xcc. The Xcc wild-type strain 8004 and its derivatives from overnight culture were washed and resuspended in 10 mM SPB or sterile distilled water (for electrolyte leakage assay) to an OD 600 of 0.1 (1 × 10 8 CFU ml −1 ). (A) Disease symptoms on Chinese radish (Raphanus sativus) leaves. Xcc strains were inoculated by cutting leaves with scissors dipped in the bacterial suspensions. (B) Lesion lengths were scored 10 days postinoculation. Values represent means and standard deviation from twenty inoculated leaves in one experiment. The experiment was repeated three times with similar results. (C) HR symptoms induced in pepper leaves (Capsicum annuum cv. ECW-10R) by Xcc strains. Approximately 5 μ l bacterial resuspension (1 × 10 8 CFU ml −1 ) was infiltrated into the leaf mesophyll tissue with a blunt-end plastic syringe. Pictures were taken at 8, 16 and 24 h after infiltration. Three replications were done in each experiment, and the experiment was repeated three times. The results presented are from a representative experiment, and similar results were obtained in all other independent experiments. hrcV and avrBs1 Xcc deletion mutants Δ hrcV and Δ avrBs1 Xcc were used as negative controls. (D) Electrolyte leakage from pepper leaves inoculated with Xcc strains. For each sample, four 0.4 cm 2 leaf disks were collected from the infiltrated area and incubated in 5 ml distilled water. Conductivity was measured with a DDS-307A conductometer. Three samples were taken for each measurement in each experiment. Results presented are from a representative experiment, and similar results were obtained in two other independent experiments. hrcV deletion mutant Δ hrcV was used as a negative control.
promoterless cloning site of the plasmid pLAFR6, was then introduced into the mutants. The resulting strains ∆ avrAC/pRavrACH6, ∆ avrAC-hpaM/pRavrACH6 and ∆ avrAC-hrcV/pRavrACH6 (Table S1) were used to test the secretion of AvrAC protein by western blot assay. As shown in Fig. 2B, AvrAC protein was present in the cells of all the strains tested and the cultural supernatant of the strain ∆ avrAC/pRavrACH6. Similar to the negative control strain ∆ avrAC-hrcV/pRavrACH6, no AvrAC protein was detected in the cultural supernatant of the strain ∆ avrAC-hpaM/pRavrACH6 under the test conditions (Fig. 2B), indicating that deletion of hpaM abolished the secretion of AvrAC. These data confirm that HpaM is indispensable for the type III secretion of Xcc.
To further estimate the effect of HpaM on T3Es translocation into plant cells, the N-terminal 102 aa of the T3E AvrAC were fused with the calmodulin-dependent reporter protein Cya 27 and the resulting reporter plasmid, named pLavrAC 102 ::CyaA (Table S1), was introduced into the hpaM mutant strain Δ hpaM, the wild-type strain 8004, and the T3SS-defective hrcV mutant Δ hrcV. The obtained recombinant strains were inoculated into radish leaves at 10 8 cfu ml −1 (OD 600 = 0.1), and the cAMP levels were measured 24 h post-inoculation. Strain Δ hrcV/ pLAFR6, which was constructed by introducing the vector pLAFR6 into the hrcV mutant strain Δ hrcV, was used as a negative control. As shown in Fig. 2C, the cAMP level in the leaves inoculated with the wild-type strain harboring the reporter plasmid was higher than that in the leaves inoculated with the mutants carrying the reporter plasmid. As the Cya protein produces a measurable cAMP level only in plant cells but not in bacterial cells or plant apoplasts 28 , the result reveals that HpaM is essential for T3Es translocation into plant cells.
Xcc secretes a series of extracellular enzymes including exoproteases by the type II secretion system (T2SS). To evaluate whether HpaM affects the T2SS, we compared the exoprotease activities produced by the hpaM mutant strain Δ hpaM and the wild type strain 8004. The result showed that the two strains produced similar enzyme activities (Fig. S4A), suggesting that HpaM is not involved in the T2SS. The extracellular polysaccharide (EPS) Figure 2. HpaM is essential for secretion of T3SS effectors in Xcc. Type III secretion signal sequence-gusA fusion reporter plasmids pGUSavrAC and pGUSxopN were introduced into Xcc strains. The resulting recombinant strains were cultured in XVM2 medium for 12 h and the β-glucuronidase (GUS) activities were determined. Values are the means ± standard deviation from three repeats. (A) GUS activities in the cultural supernatant (Secreted) and the total culture (Total) produced by pGUSavrAC and pGUSxopN in different background strains. (B) Western blot assay. The recombinant plasmid pRavrACH6, which contains the T3E AvrAC encoding sequence fused with 6× His tag in its C-terminus, was introduced into Xcc strains. The resulting recombinant strains were cultured in XVM2 medium for 12 h and proteins in cultural supernatant (secreted protein) were collected by ultra-filtration using Amicon Ultra-15 centrifugal filter (Millipore Corporation, Billerica, MA, USA) and the total proteins in Xcc cells were prepared as previously described 62 . 30 μ g of secreted or cell protein was electrophoresed in SDS-PAGE gel and transferred to a PVDF membrane. The presence of AvrAC was detected by anti-His 6 monoclonal antibody. (C) Cya protein translocation assay. The pLavrAC 102 ::CyaA fusion construct was transferred into Xcc strains and the resulting recombinant strains were then used to inoculate Chinese radish (Raphanus sativus) leaves. The cAMP level was determined 24 h postinoculation. Values given are the means ± standard deviations of triplicate measurements from a representative experiment; similar results were obtained in two other independent experiments. 8004, wild type strain; ∆ hpaM, hpaM deletion mutant; ∆ hrcV, hrcV deletion mutant.
Scientific RepoRts | 7:42724 | DOI: 10.1038/srep42724 production and the motility of the mutant Δ hpaM were also determined. No significant difference on either EPS production or motility was observed between the mutant and the wild type ( Fig. S4), indicating that HpaM does not affect EPS production and cell motility.
To validate whether HpaM is a membrane-bound protein, the cellular location of HpaM in Xcc was determined. We constructed a recombinant strain, Δ hpaM/pRhpaMH6, which expressed HpaM with a 6× His tag on its C-terminus in the hpaM deletion strain Δ hpaM. The total, periplasmic, and outer membrane protein fractions of the strain Δ hpaM/pRhpaMH6 grown at the late log phase were prepared. Western blot analysis revealed HpaM present in the total-protein and the outer membrane fractions but not in the periplasmic protein fraction (Fig. 3A). The cytoplasm protein HpaR1 29 and the outer and inner membrane protein HpaS 21 were taken as controls (Fig. 3A). To further determine whether HpaM also locates in the inner membrane, the outer and inner membrane fraction proteins were prepared using the method as described by Chen and associates 30 . The result showed that HpaM was detected only in the outer membrane fraction but not in the inner membrane fraction, while the control protein HpaS was detected in both outer and inner membrane fractions (Fig. 3B). These combined data indicate that HpaM is an outer membrane protein in Xcc.
HpaM physically interacts with HrcC and HrcJ. The above data demonstrate that HpaM locates in the bacterial outer membrane and contributes to T3Es secretion, but is not involved in the regulation of the T3SS expression. From these facts we presumed that HpaM may act as a component of T3SS apparatus or a factor affecting the assembly or stability of the T3SS apparatus. To verify these possibilities, we employed the BacterioMatch II two-hybrid system (Stratagene, La Jolla, CA, USA) to determine whether HpaM physically interacts with the T3SS apparatus outer and inner membrane ring proteins HrcC and HrcJ 31 . A truncated hpaM gene excluding the N-terminal 22-aa signal peptide coding sequence was cloned into the bait vector pBT, yielding a recombinant plasmid named pBhpaM LN22 (Table S1). DNA fragments of truncated hrcC and hrcJ (excluding the N-terminal 33-and 21-aa signal peptide encoding sequences of hrcC and hrcJ, respectively) were fused into the target vector pTRG, yielding recombinant plasmids named pThrcC LN33 and pThrcJ LN21 (Table S1). The plasmids were introduced into the reporter strain XL1-Blue MRF′ . The resulting recombinant strains, which harbor a pair of plasmids (Table S1) were tested for their growth ability on the double-selective indicator plate. In the reporter strain, if the HpaM and HrcC or HrcJ proteins interact with each other, the expression of HIS3 and addA reporter genes will be activated, leading to the growth of the bacterial cells in the presence of 3-amino-1, 2, 4-triazole (3-AT) and streptomycin; however, if no interaction between the proteins occurs, the bacteria cannot grow in the same conditions. As shown in Fig. 4A, like the positive control strain XL1-Blue MRF′ /pBThpaS LN54 /pTRGhrpG that showed an interaction between the histidine kinase HpaS and the response regulator HrpG of a two-component regulatory system 21 , the reporter strain XL1-Blue MRF′ harboring the plasmid pair pBhpaM LN22 /pThrcC LN33 or pBhpaM LN22 / pThrcJ LN21 grew well in the selective agar plate, while the negative control strains (the reporter strain harboring the plasmid pair pBT/pTRG, pBhpaM LN22 /pTRG, or pBT/pThrcC LN33 ) did not grow (Fig. 4A). These results indicate that HpaM interacts with HrcC as well as HrcJ in the reporter strain XL1-Blue MRF′ . To evaluate whether the interaction between HpaM and HrcC or HrcJ is specific, the membrane-bound protein HpaS was included in the Xcc strains were cultured to an OD 600 of 1.0 and proteins were prepared using the method described by Feilmeier and associates (2000) (A) or the method described by Chen and associates (2010) (B). 30 (for total protein) or 10 μ g of protein sample was separated by SDS-PAGE electrophoresis and transferred to a PVDF membrane. The presence of HpaM was detected by anti-His 6 monoclonal antibody. The histidine sensor kinase HpaS and the transcription regulator HpaR1 were used as controls. HpaM, protein sample was prepared from strain Δ hpaM/pRhpaMH6; HpaS, protein sample was prepared from strain ∆ hpaS/pRhpaSH6; HpaR1, protein sample was prepared from strain ∆ hpaR1/pRhpaR1H6. bacterial two-hybrid analysis. A truncated HpaS protein (lacking the N-terminal 54 aa transmembrane domain encoding sequence) was cloned into the target vector pTRG and the obtained plasmid pTRGhpaS LN54 was used in the analysis. The result showed that the reporter strain XL1-Blue MRF′ harboring the plasmid pair pBhpaM LN22 / pTRGhpaS LN54 could not grow on the selective agar plate, indicating no interaction existed between HpaM and HpaS (Fig. 4A). It has been supposed that the periplasmic domains of the HrcC and HrcJ proteins interact with each other and compose the T3SS periplasmic rod of the T3SS apparatus 32 . We therefore tested whether HpaM interacts with the periplasmic domains of HrcC and HrcJ. For this purpose, a 1011 bp DNA fragment encoding the aa from the 34 th to the 370 th of HrcC and a 555-bp fragment encoding the aa from the 22 th to the 206 th of HrcJ were amplified and cloned into the target vector pTRG, yielding recombinant plasmids named pThrcC  and pThrcJ   (Table S1). As shown in Fig. 4A, the reporter strain XL1-Blue MRF′ harboring the plasmid pair pBhpaM LN22 /pThrcC  or pBhpaM LN22 /pThrcJ  was able to grow on the selective agar plate, indicating that HpaM interacts with the periplasmic domains of HrcC and HrcJ in the reporter strain.
To confirm the interactions, pull-down biotinylated protein-protein assays were performed. For this purpose, an attempt was made to overproduce recombinant 6× His-tagged truncated HrcC and HrcJ proteins by cloning truncated hrcC and hrcJ excluding the N-terminal 33-and 21-aa signal peptide encoding sequences into the To gain a primary insight into the molecular interaction between HpaM and HrcC or HrcJ, we defined the peptides in HpaM required for the interaction. As described above, the first 22 aa in the N-terminus of HpaM was predicted to be a signal peptide. We therefore tested the N-terminal portion exclusive of the first 22 aa. 540, 609, and 678 bp DNA fragments encoding the peptides of the 23 th -202 th aa, 23 th -225 th aa, and 23 th -248 th aa, respectively, were amplified by using the corresponding primer sets listed in Table S2 and cloned into the vector pBT, respectively. 372, 441, 510, and 276 bp DNA fragments encoding the C-terminal peptides of the 226 th -349 th aa, 203 th -349 th aa, 180 th -349 th aa, and 180 th -271 th aa were also amplified and cloned into the vector pBT. The obtained recombinant plasmids (Table S1) as well as plasmid pThrcC  or pThrcJ  were introduced into the reporter strain XL1-Blue MRF′ and the growth of the resulting recombinant strains was examined. As shown in Fig. 5B, the recombinant strains that harbored the plasmid pair pBM 23-225 /pThrcC  , pBM 23-248 /pThrcC  , pBM 180-349 /pThrcC  , pBM 180-271 /pThrcC  or pBM 180-349 /pThrcJ  could grow on the selective plate but other strains could not, indicating that the peptide consisting of the 180 th to the 225 th aa of HpaM is essential for the interaction between HpaM and HrcC, and the C-terminus of HpaM from the 180 th aa is involved in the interaction between HpaM and HrcJ. The 180 th to the 225 th aa of HpaM was further tested to see whether it suffices the interaction with HrcC, and the truncated proteins consisting of the 180 th aa to the 340 th , the 320 th , or the 300 th aa of HpaM were also further tested for their interactions with HrcJ, respectively. As shown in Fig. 5, the peptide from the 180 th to the 225 th aa of HpaM is sufficient for the interaction with HrcC (the strain containing pBM 180-225 /pThrcC  could grow on the selective agar plate), and the peptide from the 180 th to the 320 th (but not to the 300 th ) aa of HpaM is sufficient for the interaction with HrcJ (the strain containing pBM 180-320 /pThrcJ  could grow on the selective agar plate). The interactions were further confirmed by pull-down assays (Fig. 4B, lanes 12  and 14).
To evaluate whether the interaction with HrcC or HrcJ is essential for HpaM function, the HpaM derivatives with deletion in 180 th -202 th aa consisting a PbH1 domain of parallel β -helix repeats and 288 th -311 th aa consisting a PbH1 domain of parallel β -helix repeats, respectively, were constructed, and the obtained hpaM partial deletion mutants were named Δ hpaM 180-202 and Δ hpaM 288-311 , respectively. Plant assays revealed that the two mutant strains, similar to the hpaM full deletion mutant Δ hpaM, scarcely caused any disease or HR symptoms in the host plant Chinese radish or the non-host plant pepper (Fig. S6). Additionally, the recombinant plasmid pLChpaM carrying a full length hpaM gene was introduced into the mutants Δ hpaM 180-202 and Δ hpaM 288-311 , respectively. The resulting complemented strains CΔ hpaM 180-202 and CΔ hpaM 288-311 showed wild-type virulence and HR phenotypes (Fig. S6).
Evidences that HrcC, HpaM and HrcJ are outer and inner membrane-bound proteins, respectively, and HrcC of Xcc interacts directly with HrcJ. In animal pathogens, the EscC/InvG/YscC family proteins compose of the outer membrane ring, and the EscJ/PrgK/YscJ family members are one of the inner membrane ring components. Periplasmic domains of EscC/InvG/YscC and EscJ/PrgK/YscJ proteins interact with each other and form the T3SS periplasmic rod [32][33][34] . HrcC and HrcJ in phytopathogens are isoforms of the EscC/ InvG/YscC and EscJ/PrgK/YscJ families, respectively 6 . Deletion of hrcC or hrcJ abolished the virulence and HR induction of Xcc (Fig. S7). The N-termini of HrcC and HrcJ were predicted to be the periplasmic domains and their C-termini were supposed to integrate into the cell membranes. To verify the HrcC and HrcJ integration in Xcc cells, recombinant strains ∆ hrcC/pRhrcCH6 and ∆ hrcJ/pRhrcJH6 were constructed, which produced HrcC and HrcJ with a 6× His tag on the C-terminus in the mutants ∆ hrcC and ∆ hrcJ, respectively. The outer and inner membrane protein fractions of the two strains grown to the late-log phase were prepared and exposed to western blot analysis. As shown in Fig. 6, HrcC and HrcJ were present in the outer and inner membrane fractions, respectively, indicating that HrcC and HrcJ in Xcc, as speculated, are outer and inner membrane-bounded proteins, respectively.
Our above data revealed that HpaM is an outer membrane-bound protein. As HrcC is believed to compose the outer membrane ring of the type III apparatus, we concerned that whether the outer membrane localization of HpaM depends on the presence of HrcC. We therefore detected the location of HpaM in the hrcC deletion mutant background. To do this, an hpaM and hrcC double deletion mutant named ∆ hpaM-hrcC (Table S1) was constructed, and the recombinant plasmid pRhpaMH6 was introduced into the mutant. The resulting recombinant strain ∆ hpaM-hrcC/pRhpaMH6 (Table S1) was used to locate HpaM protein. As shown in Fig. 6, HpaM protein was still present in the outer membrane fraction of the bacterial cells, indicating that the presence of HpaM in the outer membrane does not rely on HrcC, i.e. HpaM is in itself an outer membrane-bound protein.
To verify the Xcc HrcC and HrcJ proteins interact with each other, the truncated hrcC and hrcJ genes excluding the signal peptide coding sequence were cloned into the vector bait pBT and the prey pTRG, respectively, resulting the plasmids pBhrcC  and pThrcJ   (Table S1). The plasmids were introduced into the reporter strain XL1-Blue MRF′ . As shown in Fig. 4A, the strain harboring the plasmid pair pBhrcC 34-370 /pThrcJ  grew well on the selective indicator plate, while the strain harboring the plasmid pair pBThpaS LN54 /pThrcJ  or pBhrcC 34-370 /pTRGhpaS LN54 could not grow. These results indicate that the interaction between HrcC and HrcJ existed. Protein pull-down assay was carried out to further verify the bacterial two-hybrid assay result.  . Evidence from western blot analysis reveals that HpaM, HrcC and HrcJ are outer and inner membrane-bound proteins, respectively. The outer and inner membrane fraction proteins from strain ∆ hrcC/ pRhrcCH6 (for HrcC detection), ∆ hrcJ/pRhrcJH6 (for HrcJ detection), and Δ HpaM-HrcC/pRhpaMH6 (for HpaM detection) were prepared. 10 μ g of protein for each sample was separated by SDS-PAGE electrophoresis and transferred to a PVDF membrane. The presence of HrcC, HrcJ, and HpaM was detected by anti-His 6 monoclonal antibody. The histidine sensor kinase HpaS (from strain ∆ hpaS/pRhpaSH6) was used as a control.
HpaM is highly conserved in phytopathogenic Xanthomonads. To date, the whole genome sequences of more than one dozen Xanthomonas spp. or pathovars are available. A protein blast revealed that HpaM is conserved in all sequenced Xanthomonas spp. (Table S3). Although the rate of their amino acid sequence homology is varied among different species or pathovars, most of which share more than 90% similarity and 87% identity. Only three species, i.e., X. translucens, X. sacchari, and X. albilineans, share an HpaM homologue with lower similarity (71-74%) and identity (about 60%) to Xcc HpaM. Additionally, an HpaM homologue also exists in Pseudoxanthomonas spadix and Xylella fastidiosa, which shares ~55% identity and ~68% similarity with Xcc HpaM (Table S3). Transmembrane domain analysis using the TMPRED program (http://www.ch.embnet.org/ software/TMPRED_form.html) revealed that the N-termini of all the HpaM homologues contain a transmembrane helice (Table S3).
The Xanthomonas oryzae homologues of HpaM exhibit similar functions to Xcc HpaM. As described above, HpaM is highly conserved in Xanthomonas pathogens. To verify whether the HpaM homologues in other Xanthomonas spp. play similar roles to Xcc HpaM, we investigated the function of the HpaM homologues in the species Xanthomonas oryzae. X. oryzae consists of two pathovars, oryzae (Xoo) and oryzicola (Xoc), which are the causative agents for bacterial leaf blight and bacterial leaf streak of rice, respectively. The whole-genome sequences are available for Xoo strain PXO99 A 35 and Xoc strain GX01 (our unpublished data), therefore, we used these strains in the study. The HpaM homologues in strain PXO99 A and strain GX01 were designated as HpaM Xoo and HpaM Xoc , respectively. HpaM Xoc is completely identical to its counterpart in the Xoc strain BLS256 36 . If HpaM Xoo and HpaM Xoc are entrusted with similar functions to Xcc HpaM, they should be able to replace Xcc HpaM and restore the virulence and HR induction of the Xcc hpaM deletion mutant. Therefore, we cloned the hpaM homologues of Xoo and Xoc into the vector pLAFR3 (Table S1) and introduced the resulting recombinant plasmids pLChpaM Xoo and pLChpaM Xoc (Table S1) into the Xcc hpaM deletion mutant strain Δ hpaM, respectively. Plant tests showed that either of pLChpaM Xoo and pLChpaM Xoc could restore the ability of the mutant to induce typical black rot symptoms in the host plant Chinese radish and HR in the non-host plant pepper leaves (Fig. 1A,B, Fig. S8), indicating that Xcc HpaM and its counterparts in Xoo and Xoc probably have similar functions.
To further investigate the function of HpaM Xoo and HpaM Xoc in Xoo and Xoc, hpaM Xoo and hpaM Xoc deletion mutants were constructed from strain PXO99 A and strain GX01 by homologous recombination using the suicide vector pK18mobsacB 37 , and the resulting mutants, named Δ hpaM Xoo and Δ hpaM Xoc (Table S1), were tested for virulence in rice and HR in tobacco. As shown in Fig. 7, both mutants almost completely failed to stimulate disease symptoms in rice and HR in tobacco, while the complemented strains could induce wild-type disease symptoms and HR. As the T3SS is also essential for the pathogenicity and HR induction of both pathogens, the plant test result suggests that HpaM is probably indispensable for a functional T3SS of Xoo and Xoc. To verify this, the type III secretion efficiency of the mutants Δ hpaM Xoo and Δ hpaM Xoc was detected. To do this, the type III secretion reporter plasmid pLGUSavrAC (Table S1) was introduced into the mutants Δ hpaM Xoo and Δ hpaM Xoc as well as the wild type strains of Xoo and Xoc. The GUS activities of the resulting recombinant strains were then determined. As shown in Fig. 8A, both mutants harboring pLGUSavrAC produced significantly weaker GUS activity in cultural supernatants, compared to the wild type strains harboring pLGUSavrAC, suggesting that the type III secretion efficiency of the mutants was significantly weakened. These combined data demonstrate that the HpaM homologues of Xoo and Xoc are also critical for the type III secretion. The cellular location of HpaM Xoo and HpaM Xoc was also determined by western blot assay. The HpaM Xoo and HpaM Xoc encoding sequences fused with 6× His tag at their C-termini were cloned into pLAFR3 and the resulting recombinant plasmids named pRhpaM Xoo H6 and pRhpaM Xoc H6 (Table S1) were introduced into the mutant strains Δ hpaM Xoo and Δ hpaM Xoc , respectively. The outer and inner membrane proteins from the obtained recombinant strains Δ hpaM Xoo /pRh-paM Xoo H6 and Δ hpaM Xoc /pRhpaM Xoc H6 (Table S1) were prepared and analyzed by western blot assay. The result revealed that HpaM Xoo and HpaM Xoc were also located in the outer membrane of Xoo and Xoc (Fig. 8B). Taken together, the above combined data indicate that the Xoo and Xoc homologues of HpaM may have similar functions to Xcc HpaM.

Discussion
Here we have demonstrated that the novel outer membrane-bound protein HpaM is critical for the type III secretion of Xanthomonas spp. Mutation of hpaM did not alter the production of extracellular enzymes and polysaccharides as well as cell motility, suggesting that HpaM may specifically affect the T3SS. HpaM is not involved in the regulation of the expression of hrp genes that encode the components of the T3SS machinery, but interacts with HrcC and HrcJ, the homologues of the components that compose the outer and inner rings of the T3SS basal body of all bacterial pathogens that possess a T3SS. Mutation of hrcC or hrcJ almost completely broke the type III secretion of Xcc, resulting in loss of the ability to cause disease symptoms and HR. In animal pathogens, it has been shown that the outer and inner ring proteins are outer and inner membrane proteins, respectively, and they physically interact with each other directly [5][6][7][8] . In this work, we authenticated that Xcc HrcC and HrcJ, as expected, are located in the outer and inner membrane, respectively, and they interact with each other directly. These data provide supporting evidence to the inference that HrcC and HrcJ act as T3SS outer and inner ring proteins in Xanthomonas spp. The peptide consisting of 46 amino acids from the 180 th to 225 th aa of HpaM is sufficient for interaction with HrcC but not HrcJ, and the most portion of the C-terminus, containing the amino acids from the 180 th to 320 th aa, is indispensable for the interaction with HrcJ. Bioinformatics analysis revealed that Scientific RepoRts | 7:42724 | DOI: 10.1038/srep42724 transmembrane helices are present in the N-terminus of HpaM. Taken together, these data suggest that HpaM is integrated into the outer membrane with its N-terminal domain and extends into the periplasm, where its middle part interacts with the outer ring protein HrcC and the C-terminal portion including the middle part interacts with the inner ring protein HrcJ, forming a three protein complex. It is worth noting that HpaM is predicted to have six right-handed parallel β -helix repeats from the 120 th to 311 th residues (i.e. 120-163, 180-202, 203-225, 226-248, 249-271, and 288-311), five of which lie in the region related to its physical interaction with HrcC and HrcJ. The right-handed parallel β -helix repeats are most commonly associated with autotransporter proteins, many of which are extracellular enzymes. It is clear that the β -helix repeats are essential not only for protein folding but also for functions, such as forming an appropriate structure that recognizes the substrates 38,39 . The presence of the β -helix repeats within the region interacting with HrcC and HrcJ suggests that they may be critical for HpaM stability and the formation of the protein complex.
Comparative bioinformatics analysis revealed that HpaM is conserved in all sequenced Xanthomonas species. To expand our knowledge on the function of HpaM in other Xanthomonas spp., we also investigated the HpaM homologues in X. oryzae pathovars oryzae and oryzicola. The results demonstrated that the HpaM homologues in the two pathovars of X. oryzae also localize in the outer membrane and are critical for pathogenicity and HR as well as efficient type III secretion. Furthermore, they can replace HpaM in Xcc for the type III secretion. These results indicate that HpaM is conserved not only in structure but also in function in Xanthomonas spp. Interestingly, HpaM homologs are also present in the species Pseudoxanthomonas spadix and Xylella fastidiosa (Table S3). Like the genus Xanthomonas, Pseudoxanthomonas and Xylella genera also belong to the family Xanthomonadaceae. It is possible that HpaM homologues are also prevalent in the members of these genera. However, Pseudoxanthomonas spadix and Xylella fastidiosa do not seem to have a T3SS. To investigate the function of the HpaM homologues in these bacteria will be a valuable topic.
At this stage, the precise role of HpaM in the T3SS is not clear. However, given the facts that: 1) HpaM does not act as a regulator for hrp gene expression but is critical for type III secretion; 2) its N-terminus integrates in the outer membrane and C-terminus extends deeply into the periplasm to interact physically with the inner ring protein HrcJ; and 3) its middle part interacts physically with the outer ring protein HrcC, forming a HpaM-HrcC-HrcJ complex, we presume that HpaM is most likely to be a structural component of the T3SS in Xanthomonas spp., although it is not encoded by a gene within the hrp cluster. In general, T3SS structural components of animal and plant pathogens are encoded by chromosomal or plasmid-borne gene clusters that were probably acquired during evolution by horizontal gene transfer 6,40 . As described above, the cluster consists of more than 20 genes and nine of which are conserved among plant and animal pathogens. These conserved genes are believed to encode the core components of the T3SS machinery in both plant and animal pathogens. However, more than 50% of the genes in the clusters are varied, suggesting that the clusters have changed a lot in different pathogens during the long-term evolution. A phylogenetic tree analysis divides the T3SSs of plant and animal pathogens into at least six families including two families from plant pathogens 41 . Therefore, it is possible that although the architectures of the T3SS apparatuses in different pathogens are similar, some fittings may not be the same. As described above, the T3SS extracellular appendage (pilus-like) of plant pathogens is different from that (needle-like) of animal pathogens. In addition, the inner rod may be another case showing different fittings in the T3SSs between plant and animal pathogens. The inner rod is a part of the T3SS basal body found in animal pathogens, which is formed by a periplasmic protein that connects the outer and inner rings 33,34 . However, the inner rod homologous protein is missing in plant pathogens. A non-homologous protein, HrpB2, has been supposed to be a putative inner rod protein of X. euvesicatoria, based on the features that it contains a VxTLxK amino acid motif that is conserved in the inner rod proteins of animal pathogens, localizes to the periplasm and the outer membrane, and is essential for T3SS pilus formation 6,42 .
A periplasmic protein, named VrpA, encoded by a gene outside the hrp cluster of X. citri subsp. citri, was recently reported to contribute to the secretion efficiency of the T3SS 43 . Similar to HpaM, VrpA is conserved in Xanthomonas spp. and also physically interacts with HrcC and HrcJ but not HrpB2. It was presumed that VrpA may affect activation of secretion and assembly or stability of the T3SS apparatus via interacting with HrcC and HrcJ 43 . We cannot exclude the possibility that HpaM associates with the T3SS via assisting the apparatus assembling or affecting the apparatus stability rather than as a T3SS structural component. Nonetheless, given the fact that HpaM as well as VrpA are Xanthomonas genus-specific proteins which are absent in other bacterial pathogens that possess a T3SS, our results suggest that the T3SS of Xanthomonas is distinctive in some aspects from other pathogens. To further investigate the precise role of HpaM and VrpA will greatly facilitate our understanding of the T3SS biogenesis.

Materials and Methods
Bacterial strains, plasmids and growth conditions. The bacterial strains and plasmids used in this study are listed in Table S1. Escherichia coli strains were grown in Luria-Bertani medium 44 at 37 °C. Xcc strains were grown at 28 °C in NYG medium 45 , the minimal medium MMX 46 or XVM2 31 . Xoo and Xoc strains were grown at 28 °C in OB medium 47 , NB medium 48 , or the minimal medium XOM2 49 . Antibiotics were added at the following concentrations as required: kanamycin (Kan) 25 μ g ml −1 , rifampicin (Rif) 50 μ g ml −1 , ampicillin (Amp) 100 μ g ml −1 , spectinomycin (Spc) 50 μ g ml −1 , gentamicin (Gm) 5 μ g ml −1 , streptomycin (Sm) at 100 μ g ml −1 , and tetracycline (Tet) 5 μ g ml −1 for Xanthomonas spp. and 15 μ g ml −1 for E. coli. DNA and RNA techniques, SDS-PAGE and western blotting. DNA manipulations followed the procedures described by Sambrook and associates 50 . Plasmids were transformed into cells of E. coli and Xanthomonas spp. by electroporation or conjugation described by Turner and associates 51 . The restriction endonucleases, T4 DNA ligase, and pfu polymerase were provided by Promega (Shanghai, China). The total RNAs from Xanthomonas spp. were extracted with a total-RNA extraction kit (Promega), and reverse transcription was performed using a cDNA synthesis kit (Fermentas Co., Vilnius, Lithuania). Each kit was used according to the manufacturer's instructions.
Western blotting was carried out as previously described 21 . Briefly, bacterial proteins were separated by 12% (w/v) SDS-PAGE and transferred onto PVDF (polyvinylidene difluoride) membrane (Millipore Corporation, Billerica, MA, USA). After blocking, the 1:2500 diluted anti-His-tag mouse monoclonal antibody (Qiagen, Shanghai, China) was used as the primary antibody, and the 1:2500 diluted horseradish peroxidase conjugated goat antimouse IgG (Bio-Rad, Hercules, CA, USA) was used as secondary antibody.

Deletion mutant construction and complementation.
The hpaM in Xcc and its homologues hpaM Xoo (in Xoo) and hpaM Xoc (in Xoc) were deleted by the method described by Schäfer and associates 37 . For construction of Xcc hpaM deletion mutant, 747-bp upstream and 726-bp downstream fragments flanking hpaM (XC_2847) were amplified with the primer sets LhpaM-F/R and RhpaM-F/R (Table S2), respectively, using the total DNA of the Xcc wild type strain 8004 as a template. Primers were modified to give EcoRI-, XbaI-or HindIII-compatible ends (underlined) ( Table S2). The two fragments were cloned together into the vector pK18mobsacB 37 , and the resulting plasmid named pK18mobsacBhpaM was introduced into the Xcc strain 8004 by triparental conjugation. The transconjugants were screened on selective agar plates containing 5% sucrose. The obtained hpaM deletion mutant was further confirmed by PCR and named Δ hpaM.
The HpaM derivatives with deletion in 180 th -202 th aa or 288 th -311 th aa were constructed by using the same method. For the HpaM derivative with deletion in 180 th -202 th aa, a 767-bp fragment spanning the 230 th nt upstream to the 537 th nt downstream of the start codon ATG of hpaM ORF and a 560-bp fragment spanning the 607 th nt to the 1166 th nt downstream of the start codon ATG of hpaM ORF were amplified with the primer sets LhpaM-F 180 /R 180 and RhpaM-F 180 /R 180 . The resulting hpaM partial deletion mutant was named Δ hpaM 180-202 . For the HpaM derivative with deletion in 288 th -311 th aa, a 715-bp DNA fragment spanning the 147 th nt to the 861 th nt downstream of the start codon ATG of hpaM ORF and a 588-bp DNA fragment spanning the 934 th nt to the 1521 th nt downstream of the start codon ATG of hpaM ORF were amplified with the primer sets LhpaM-F 311 /R 311 and RhpaM-F 311 /R 311 . The resulting hpaM partial deletion mutant was named Δ hpaM 288-311 .
For deletion of hpaM Xoo (PXO_01147) or hpaM Xoc (XOC_3053 homologue), 879-bp upstream and 591-bp downstream fragments flanking hpaM Xoo or hpaM Xoc were amplified with the corresponding primer sets (Table S2) from the Xoo strain PXO99 A and the Xoc strain GX01, respectively. The resulting deletion mutants were named Δ hpaM Xoo and Δ hpaM Xoc (Table S1). For complementation of the hpaM deletion mutant, a 1432-bp DNA fragment containing the hpaM coding region and extending from 352-bp upstream of the 5′ end to 30-bp downstream of the 3′ end of the ORF was amplified by PCR from the total DNA of the Xcc strain 8004 with the primer set ChpaM-F/R (Table S2). Primers were modified to give BamHI-or HindIII-compatible ends (underlined) ( Table S2). The amplified fragment was confirmed by sequencing, and ligated into the promoterless cloning site of the plasmid pLAFR6 52 , generating the recombinant plasmid named pLChpaM (Table S1). The plasmid was introduced into the hpaM deletion mutant or partial deletion mutants by triparental conjugation, generating complemented strains named CΔ hpaM, CΔ hpaM 180-202 and CΔ hpaM 288-311 , respectively (Table S1). 1053-bp DNA fragments of the hpaM Xoo (PXO_01147) ORF and hpaM Xoc (XOC_3053 homologue) ORF were also amplified by PCR from the Xoo strain PXO99 A and the Xoc strain GX01, respectively, and cloned into the plasmid pLAFR3 53 . The resulting recombinant plasmids named pLChpaM Xoo and pLChpaM Xoc (Table S1) were used to complement the mutant strains Δ hpaM, Δ hpaM Xoo , and Δ hpaM Xoc .

Determination of transcriptional start site.
To determine the transcriptional start site of the hpaM gene, 5′ -RACE (5′ rapid amplification of cDNA ends) method was carried out with the hpaM sequence-specific primers hpaM-RTP1-4 (Table S2). The assay was performed as previously described 21 . Briefly, total cellular RNA was extracted from the Xcc wild type strain 8004 grown in NYG medium to an OD 600 of 1.0. cDNA fragments were obtained using the 5′ -RACE kit (Invitrogen Life Technologies, San Diego, CA, USA), and PCR products were cloned into the vector pMD19-T and sequenced.

Construction of promoter reporter plasmid.
A promoter reporter plasmid for hpaM was constructed by fusing a 404-bp DNA fragment upstream of hpaM ORF (including the translation start codon ATG) with the promoterless β -glucuronidase (GUS)-encoding ORF (excluding the translation start codon ATG). The hpaM promoter region was amplified from the total DNA of the Xcc wild type strain 8004 by using the primer set RP-hpaMF/R (Table S2). The gusA coding region was amplified by PCR with the primer set GusF/R (Table S2), using the transposon Tn5gusA5 DNA as template. Primers were modified to give EcoRI-, BamHI-or PstI-compatible ends (underlined) ( Table S2). The two fragments obtained were cloned into the promoterless cloning sites of the plasmid pLAFR6 to generate the reporter plasmid named pGUShpaM (Table S1).
Bacterial two-hybrid assay. The BacterioMatch II two-hybrid system (Stratagene, La Jolla, CA, USA) was used to detect protein-protein interaction in vivo. The truncated (or full length) hpaM, hrcC and hrcJ were amplified by PCR using the total DNA of the Xcc wild type strain 8004 as template and corresponding oligonucleotide set as primers (Table S2), respectively. The 981 bp truncated hpaM gene (from the 67 th to the 1047 th nt of the hpaM gene coding sequence, excluding the signal peptide coding sequence) was cloned into the BamHI/XhoI sites of pBT (bait), generating the plasmid pBhpaM LN22 (Table S1). 1716-and 1011-bp fragments of truncated hrcC, and 699-and 555-bp fragments of truncated hrcJ were cloned into the BamHI/XhoI sites of the vector pTRG (prey), respectively, generating the plasmids named pThrcC LN33 , pThrcC  , pThrcJ LN21 and pThrcJ   (Table S1). The bacterial two-hybrid assay was performed according to the manufacturer's instructions. To test the interaction of a variety length of HpaM fragments with the periplasmic domain of HrcC and HrcJ, 540, 609, 678, 372, 441, 510, 276, 138, 363, 423 and 483-bp DNA fragments containing partial hpaM gene were amplified by PCR using the corresponding primer sets (Table S2), respectively, and the obtained DNA fragments were cloned into BamHI/XhoI sites of pBT, resulting a series of pBM recombinant plasmids (Table S2). The plasmid pairs (Fig. 5) were used to co-transform the reporter strain XL1-Blue MRF′ on M9 salt agar without 3-AT. Colonies were then restreaked on M9 salt agar containing 5 mM 3-AT and incubated at 37 °C for 24 h for the first detection of interaction. For confirmation, the colonies were cultured on dual selective medium containing 5 mM 3-AT and 12.5 μ g ml −1 Sm, as described in the manual.
GUS activity assay. GUS activity was determined by measurement of the absorbance of OD 415 using ρ -nitrophenyl-β -D-glucuronide as the substrate, as described by Henderson and associates 54 , after growth of bacterial cells in medium for a period of time. To determine the GUS activity of secreted proteins, the bacterial cells of 200 μ l culture for each strain were separated by centrifugation and the cell-free supernatant was taken for GUS activity determination.
Plant assay. The virulence of Xcc to Chinese radish (Raphanus sativus) was tested by the leaf-clipping method 55 . Bacterial cells from overnight culture were collected, washed with 10 mM sodium phosphate buffer (SPB, 5.8 mM Na 2 HPO 4 and 4.2 mM NaH 2 PO 4 , pH 7.0) and resuspended in the same buffer to an OD 600 of 0.1 (1 × 10 8 CFU ml −1 ). Leaves were cut with scissors dipped in the bacterial suspensions. Lesion length was measured 10 days after inoculation, and data were analysed by t-test. The HR was tested on the pepper plant ECW-10R (Capsicumannuum cv. ECW-10R) as previously described 21 . For each Xcc strain tested, an approximately 5 μ l bacterial resuspension (1 × 10 8 CFU ml −1 ) was infiltrated into the abaxial leaf surface of pepper plant. The inoculated plants were maintained in appropriated conditions, and HR symptoms were observed and photographed at 8, 16 and 24 h after inoculation. For the electrolyte leakage assay, bacterial cells were resuspended in sterile distilled water at a concentration of OD 600 of 0.1. Four 0.4 cm 2 leaf disks for each sample were collected from the bacteria-infiltrated area and incubated in 5 ml of distilled water. Conductivity was measured with a DDS-307A conductometer.
For Xoo virulence assay, the wild-type strain PXO99 A 56 and its derivatives were tested on susceptible rice plant Oryza sativa L. ssp. Japonica cultivar Nipponbare using leaf clip inoculation method 57 . For Xoc, the wild type strain GX01 and its derivatives were infiltrated into rice leaves by needleless syringe 58 . Bacterial cells were grown for 72 h at 28 °C in NB medium with appropriate antibiotics. The cells were collected and resuspended in sterile distilled water to a concentration of OD 600 = 0.3. Inoculation was carried out on 6-week-old rice plants under relevant conditions. Symptoms were recorded by photography and the disease lesion lengths were measured 14 days after inoculation. Twenty-five leaves were inoculated for each strain in each experiment. The experiment was repeated three times.
The HR test of Xoo and Xoc was conducted as described by Guo et al. 59 and Zou et al. 60 , respectively. Briefly, Xoc or Xoo strains were cultured in NB medium to logarithmic phase, and cells were pelleted and suspended in water to a concentration of OD 600 = 0.5. The suspensions were infiltrated into leaves of glass house-grown tobacco (Nicotina benthamiana), and the results were observed at 24, 36 and 48 h after infiltration. If the strain had the ability to trigger HR, the phenomenon of programmed cell death would be observed around the inoculation sites on tobacco leaves. The detection of the electrolyte leakage in tobacco leaves inoculated with Xoo or Xoc strains was similar to the method used to detect that in pepper leaves inoculated with Xcc strains.
Cya protein translocation assay. To determine Cya enzyme activity in vivo, a modification of the procedure described by Roden and associates 61 was carried out. Briefly, a 894-bp DNA fragment spanning nucleotides 588-bp upstream to 306-bp downstream of the translation start codon ATG of avrAC (XC_1553) was fused with the ORF of cyaA (excluding the translational start codon ATG) and ligated into pLAFR6, resulting the plasmid pLavrAC 102 ::CyaA (Table S1). The plasmid was then introduced into Xcc wild type, hpaM mutant, and hrcV mutant strains. The resulting recombinant strains were cultured in NYG medium, and bacterial cells from the cultures were resuspended in 10 mM MgCl 2 to a concentration of OD 600 of 0.1 and then infiltrated into 20 plant leaves. A direct cyclic AMP (cAMP) correlation enzyme immunoassay kit (Amersham) was used to process the leaf samples and measure the cAMP concentrations following the manufacturer's instructions. The protein content of each sample was determined using the Bio-Rad protein assay (Bio-Rad). Cya enzyme activity was expressed as pmol of cAMP per mg of total protein.
Preparation of total, periplasmic, and outer membrane proteins. The bacterial total and periplasmic proteins were prepared using the method described previously 62 . The bacterial outer and inner membrane proteins were prepared as described by Chen and associates 30 . 100 ml of the bacterial culture for each strain was collected and disrupted by sonication, the unbroken cells and cell debris were removed by centrifugation at 14,000 g at 4 °C for 30 min. Supernatants were then centrifuged at 135,000 g at 4 °C for 1 h. The pellets, which contain membranes and ribosomes, were suspended in 1.0 ml cold TM buffer (10 mM Tris, pH 8.0, containing 8 mM MgSO 4 ), and followed by centrifugation at 135,000 g. Pellets were then rinsed with 1.0 ml cold TM buffer, resuspended in 3.9 ml 0.25% (w/v) Sarkosyl and loaded into 3.9 ml ultracentrifuge tubes. After incubation at room temperature for 1 h, the tubes were centrifuged at 135,000 g for 1 h. Supernatants, containing the Sarkosyl-soluble inner membranes, were retained. The Sarkosyl-insoluble pellets, containing the outer membrane fraction, were washed twice with 1.0 ml 0.25% (w/v) Sarkosyl, incubated at room temperature for 1 h and centrifuged at 135,000 g. The pellets, containing the outer membrane fraction, were resuspended in 40 μ l cold TM buffer.
The outer membrane proteins were also prepared as described by Leuzzi et al. 63 . Briefly, bacterial cells were disrupted by sonication and the supernatant containing the total membrane fraction was recovered and further centrifuged at 50,000 g for 90 min at 4 °C. The pellet containing the membranes was resuspended in 2% Sarkosyl in 20 mM Tris-HCl, pH 7.5 and 2 mM EDTA and incubated at room temperature to solubilize the inner membranes. To remove aggregates the suspension was first centrifuged at 10,000 g for 20 min at 4 °C and then centrifuged overnight at 75,000 g at 4 °C. The pellet containing the outer membranes was resuspended in SPB.