The focus of the present study was to characterise the transcriptional organisation that is directly involved in mangotoxin production. We had previously identified the mgo operon (Mangotoxin-Generating Operon) . We determined which genes are involved in mangotoxin production by disrupting each chromosomal gene that was previously identified in pCG2-6 by mutagenesis. The disruption of ORF0 and ORF1 did not affect mangotoxin production. These two genes may belong to another independent gene cluster located close to the mgo operon that is not involved in mangotoxin production. ORF2 transcription was independent of the mgo operon, and ORF2 is homologous to the GntR family of transcriptional regulators. This family of regulatory proteins consists of the N-terminal HTH region of GntR-like bacterial transcription factors. An effector-binding/oligomerisation domain is usually located at the C-terminus . In the deposited genomes of other P. syringae pathovars, the genes in this family are often located close to gene clusters that are homologous to the mgo operon. The relationship between ORF2 and the regulation of the mgo operon remains unclear. In the present study, we observed promoter P
expression in the ORF2 mutant (UMAF0158::ORF2) when it was grown in minimal medium at 22°C but not at 28°C, in agreement with the production of mangotoxin by the ORF2 insertional mutant. These data suggest that ORF2 is not involved in mangotoxin production but provide no direct information on the possible influence of ORF2 on the mgo operon with respect to variations in temperature.
Our results demonstrate that the DNA sequence downstream of ORF2 constitutes an operon. Ma et al.  first established the correlation between the presence of a Shine-Dalgarno sequence, also known as a ribosomal binding site (RBS), and translational initiation, the expression levels of the predicted genes and operon structure . We found putative RBSs in almost all of the genes in the putative mgo operon. Only the mgoA gene, in which the start codon overlaps with the stop codon of mgoC, does contain a potential RBS sequence. mgoC and mgoA may share the same RBS, and post-translational changes may separate the two proteins; this situation could explain the absence of a putative RBS for the mgoA gene. The mutagenesis and bioinformatics analysis of each gene in the mgo operon provided insight into their relationship to mangotoxin production. The disruption of mgoB did not abrogate mangotoxin production; however, the production decreased noticeably compared with the wild-type strain. Protein domain searches indicated that mgoB is similar to haem oxygenase. This enzyme is a member of a superfamily represented by a multi-helical structural domain consisting of two structural repeats that is found in both eukaryotic and prokaryotic haem oxygenases and in proteins that enhance the expression of extracellular enzymes . The disruption mutants of the next three genes, mgoC, mgoA and mgoD, were unable to produce mangotoxin, indicating that these genes are essential for mangotoxin production. A similar conclusion was reached by Aguilera et al. , who obtained Tox- phenotypes when 11 different genes in the pht cluster, a region involved in phaseolotoxin production, were mutated by insertion, indicating that all of the genes located within this region encode proteins that are required at different stages of phaseolotoxin production, including synthesis, transport and regulation .
The sequence analysis of mgoC prompted us to search the superfamily protein domains, revealing a similarity to the N-oxygenase domain. This domain was identified in the protein PrnD, which is derived from the pyrrolnitrin biosynthesis gene cluster of Pseudomonas fluorescens. MgoC is also similar to AurF from Streptomyces thioluteus, which produces the starter unit p-nitrobenzoic acid (PNBA) for the polyketide synthase of the aureothin biosynthesis pathway . The gene mgoA, which is homologous to non-ribosomal peptide synthetases, is the largest gene in the mgo operon, and its disruption produces a mutant that is defective in mangotoxin production. Its structure, participation in mangotoxin production and influence on the virulence of the wild-type bacterium has been discussed previously . The final gene studied was mgoD; a domain localisation analysis indicated that mgoD could be a Polyketide_cyc2 belonging to the star-related lipid-transfer (START) domain superfamily. The START superfamily includes bacterial polyketide cyclase/aromatases and two families of previously uncharacterised proteins that are present only in plants and the cyanobacterium Prochlorococcus .
After analysing the elements that composed the putative mgo operon, we evaluated whether the four genes were transcribed together in a single transcript. RT-PCR experiments using the wild-type RNA showed that the four genes were connected in the single transcript (Figure 2). Moreover, the transcript size was analysed by hybridisation, which confirmed the presence of a single transcript with a sufficient size (about 6 kb) to contain the genes mgoBCAD; however, the exact size of the transcript could not be determined.
Following the identification of the mgo operon, the promoter and transcription terminator were identified and studied. The in silico analysis of the sequence identified two putative promoters. Promoter activity was detected only in a minimal medium, the same culture medium that is traditionally used for antimetabolite toxin assays [2, 13]. Promoter activity occurred in the wild-type strain at both temperatures and in the ORF2 insertion mutant at 22°C only. The other Pseudomonas spp. experimental strains, which do not produce mangotoxin, did not exhibit any β-Gal activity. The promoter activity in the wild-type strain was more intense at 28°C than 22°C. When the promoter activity was assayed at 22°C, the activity of the mutant UMAF0158::ORF2 was statistically comparable with that of the wild-type strain. These results suggest a possible influence of ORF2 on the mgo operon during its regulation in response to temperature variations. The promoter inactivity in the other two strains of Pseudomonas spp. may be due to the absence of genes homologous to the mgo operon in P. fluorescens Pf-5, but this explanation is not applicable to Pss B728a. The sequence in B728a that is homologous to the mgo operon is composed of genes that are orthologous to the mgo genes; theoretically, the promoter activity should have been similar to that of the wild-type strain, but it was not. This result suggests that there are additional genes that are necessary for mangotoxin production that are not present in B728a. In support of this explanation, additional genes involved in mangotoxin production have been identified in UMAF0158 and cloned into a different vector than pCG2-6 . The initial sequence analysis did not show any identity with the genome of B728a, and thus these additional genes may influence mgo promoter activity.
Finally, the functional promoter of the mgo operon was established by locating the start of the mgo transcript (Figure 4), which is located 18 nucleotides after the putative -10 box of the second promoter analysed in silico. Thus, the first putative promoter was eliminated as a functional promoter of the mgo operon. Once the +1 site was established, it was possible to locate additional -35 and -10 boxes, which were typical of sigma70 dependent promoters of Pseudomonas spp [19, 20] and were more closely related than the predicted -35 and -10 boxes by BPROM software developed for Escherichia coli, which are less accurate in the search for promoters of Pseudomonas spp. These results allowed us to determine the functional promoter of the mgo operon. The mgo operon terminator was found in a similar manner. The in silico analysis of the sequence identified two possible terminator sequences between the 3'-end of mgoD and the 5'-end of the 5S rRNA, both of which exhibited secondary structures typical of transcription terminators. We considered that the ribosomal transcript terminator is also likely present in the analysed sequence. RT-PCR was used to clarify which was the operon terminator, establishing T1 as the functional terminator of the mgo operon. This is a typical terminator with a stable hairpin having many GC pairs followed by a string of T's. So, it seems that the T1 terminator is a bifunctional terminator, serving this DNA region to terminate transcription of mgo operon in the sense strand and of the ribosomal operon in the antisense strand (Figure 5).
The results described above are sufficient to suggest that mgoBCAD is a transcriptional unit and therefore propose that mgo is an operon. If this argument is correct, mutations in each mgo gene should lead to the absence of a transcript for the downstream genes. A polar effect was demonstrated for UMAF0158::mgoC but not UMAF0158::mgoB. The mutation in mgoB did not prevent the transcription of the downstream genes, although the hybridisation experiments revealed that the transcription appeared to be less efficient. This reduction in transcription corresponds to the reduced production of mangotoxin by UMAF0158::mgoB relative to the wild-type strain. Therefore, the results obtained with wild-type UMAF0158 and the insertional mutants of mgoC, mgoA and mgoD support the hypothesis that the mgo genes form an operon in contrast, the results with the mutant UMAF0158::mgoB do not. We also evaluated the possible existence of an alternative promoter after the mgoB gene, which would explain the production of mangotoxin by the mutant UMAF0158::mgoB. However, during 5'RACE experiment (Figure 3) only a single transcription start site was located, eliminating the possibility of another promoter downstream of mgoB. Therefore there must be something different between the mutant and wild-type strain, which is probably the plasmid integration. In reviewing the process by which the mgo mutants were obtained, we observed that UMAF0158::mgoB was not easy to obtain. The size of mgoB is 777 bp, and the cloned sequence in pCR2.1 was 360 bp of mgoB. The integration of pCR::mgoB into mgoB occurred by single-crossover homologous recombination as it was confirmed. During this process, the plasmid could be integrated into mgoB sequence maintaining an important part of the gene. In this circumstances mgoB or sufficient fragment of it, and the remarkably other three genes of the mgo operon, could be under the influence of a promoter located in plasmid polylinker, lacZ promoter, allowing a reduced transcript expression (Figure 2) and mangotoxin production (Tables 1 and 2). To determine the insert position, a PCR was performed in which the forward primer annealed to the lacZ gene (M13F primer) and the reverse primer annealed to the 5'-end of the mgoC gene, with wild-type UMAF0158 used as the negative control. The amplicon obtained from the mutant UMAF0158::mgoB had a size of 1000 bp, confirming that the plasmid pCR::mgoB was integrated and the lacZ promoter is close to mgoB fragment (Additional file 1: Figure S1).
Because the chemical structure of mangotoxin is unknown , it is difficult to establish a hypothesis concerning the role of the mgo genes in mangotoxin biosynthesis or to determine whether they are related to the regulation of mangotoxin production. Recent studies in P. entomophila have focussed on the pvf gene cluster, which is homologous to the mgo operon, and suggest that the gene cluster serves as a regulator of certain virulence factors in pathogenic strains of Pseudomonas spp. The pvf gene cluster may be a new regulatory system that is specific to certain Pseudomonas species . In the present study, extract complementation restored mangotoxin production in the UMAF0158ΔmgoA mutant only when the culture medium was supplemented with an extract from wild-type UMAF0158. Polar effects of the deleted mgoA on mgoD expression were excluded because the construction of the deletion mutant preserved the reading phase of protein translation. Mangotoxin production was restored in the miniTn5 mutants, which contain disrupted regulatory genes, when their cultures were complemented with a wild-type extract. These results are in agreement with the results obtained by Vallet-Gely et al. , in which pvf and gac mutants were complemented by a wild-type extract. These results allow us to propose a putative regulatory role for the mgo operon in secondary metabolite production by P. syringae pv. syringae, in accordance with Vallet-Gely et al. .
To fully characterise the functions of the mgo operon, more data concerning the chemical structure of mangotoxin and a characterisation of the other genetic traits that regulate mangotoxin biosynthesis by P. syringae pv. syringae UMAF0158 are required.