Mechanisms for Negative Regulation by Iron of the futA Outer Membrane Protein Gene Expression in Vibrio anguillurum 775”

Synthesis of the 86-kDa FatA outer membrane protein is repressed under iron-rich conditions. Comple- mentation of transposition mutants derived from clones containing the pJMl iron uptake region re- vealed the existence of an antisense RNA, RNAa. This RNA is only expressed under iron-rich conditions and acts as a negative regulator of FatA synthesis, with slight but discernible decrease in the steady-state level of fatA mRNA determined by RNase protection and by Northern blot analysis. Primer extension experiments revealed that the level of several possible fatA tran- scripts was reduced in the presence of RNAa. In addition, we found that fatA mRNA expression is slightly reduced in the presence of Escherichia coli Fur. We have identified and cloned a chromosomally encoded fur-like gene in Vibrio anguillarum. medium 1-5 ~ L M EDDHA) conditions, as described (11). The membrane proteins were subjected to electrophoresis in 15% dodecyl sulfate-polyacrylamide gels (11) and transferred to nitrocel- lulose membranes. The presence of FatA protein was detected using absorbed polyclonal anti-FatA antibody and horseradish peroxidase- conjugated protein A as described previously (12).

proteins that are TonB-dependent (14). Expression of this pJM1 iron transport protein is negatively regulated by iron (11). In this paper we report the finding that an antisense RNA (RNAs) and a Fur-like repressor may play a role in the iron regulation of the fatA gene expression.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The wild type strain V. anguillarum 775 was used as a source of plasmid pJMl (8). V. anguillarurn 775::Tnl-6 carried plasmid pJHC9-8 which was generated by Tnl insertion of the wild type plasmid pJM1, resulting in deletion of the iron uptake region (15). The nalidixic acid resistant plasmidless V.
RNA Isolation and Northern Blot Hybridization-For iron-rich conditions, V. anguillarum strains were grown in modified M9 minimal medium (8) containing 100 pg/ml ferric ammonium citrate until the culture reached an AsW of 0.5. In iron-limiting conditions, the bacteria were grown in modified M9 medium to an Am of 0.25, then EDDHA was added to a final concentration of 10 p~, and the bacteria were allowed to grow to an A600 of 0.5. Total RNA was prepared according to the hot phenol method (33). RNAs were electrophoresed in formaldehyde-agarose gel (34) and transferred to nylon membranes. The membranes were stained with methylene blue to check Iron Regulation of fatA Expression for even loading and proper transfer of the RNAs. The Northern blots were analyzed with riboprobes (34-36). Prehybridization and hybridization were at 63 "C.
RNase Protection-RNase protection studies were performed as described by Krieg and Melton (35). Total RNA and labeled riboprobes were allowed to hybridize at 45 "C overnight. RNase A (Boehringer Mannheim) and RNase T1 (Sigma) treatment was at 34 "C for 30 min. Proteinase K (Boehringer Mannheim) treatment was at 37 "C for 15 min. Treated samples were extracted with phenol, chloroformisoamyl alcohol, and ethanol precipitated. Samples were analyzed in 6% urea-acrylamide gel. Labeled riboprobes were made by in vitro transcription of linearized plasmid templates as described by Melton et al. (36), using T3 or T7 RNA polymerases (Bethesda Research Laboratories) and [cY-~*P]UTP. Labeled HinfI -digested pBR322 fragments were used as molecular weight markers.
Primer Extension-A synthetic oligonucleotide, 5"TACCTGAT-GAGTCAT-3' complementary to the 5' terminus of fatA was end labeled using T4 polynucleotide kinase (New England Biolabs) and [ Y -~* P ]~A T P (Du Pont-New England Nuclear), as described by Sambrook et al. (34). The end-labeled oligonucleotide was annealed with 30 pg of total RNA at 42 'C, and primer extension was performed using avian myeloblastosis virus reverse transcriptase (Life Sciences Inc., St. Petersburg, FL) as described by Ghosh et al. (37). The same oligonucleotide was used as primer to generate sequencing ladders of the region immediately 5' of the fatA translation site by the dideoxy chain termination method of Sanger et al. (38).
Immunoblot Analysis-Total cell membranes were prepared from V. anguillarum grown under iron-rich or iron-limiting (M9 medium with 1-5 ~L M EDDHA) conditions, as described previously (11). The membrane proteins were subjected to electrophoresis in 15% sodium dodecyl sulfate-polyacrylamide gels (11) and transferred to nitrocellulose membranes. The presence of FatA protein was detected using absorbed polyclonal anti-FatA antibody and horseradish peroxidaseconjugated protein A as described previously (12). p J M l (20). These derivatives were then mobilized by conjugation into V. anguillarum H775-3a which is a nalidixicresistant derivative of V. anguillarum 775 which has been cured of the plasmid pJM1. V. anguillarum strains harboring all three mutants were transport-deficient and expressed a reduced amount of FatA (21). Since transposition mutations were located upstream of fatA, the data suggest that fatA may be within a polycistronic message. In the process of carrying out complementation studies for iron transport we found that clones pJHC-A122 and pJHC-A179 ( Fig. la) had an inhibitory effect on fatA expression in mutant 17 (Fig. 2a, compare lanes D and E with F ) . We have also found that the recombinant clone pJHC-LWl15 ( Fig. la) had no inhibitory effect on mutant 17 (data not shown). The data suggest that pJHC-A122 and pJHC-A179 encode a negative regulator of fatA, and the expression of this regulator requires the region 3' of the SalI site.

Inhibition of FatA Expression by Insertion
Identification of the Negative Regulator-An attractive possibility is that the negative regulator is an antisense RNA that is transcribed from the fatB region, 3' of the SalI site. Alternatively, the negative regulator could be a protein encoded within the fatB region. The negative regulator could function directly as a repressor or it could inhibit the expression of fatA by titrating an activator of transcription or translation. Another possibility is that a DNA sequence that may be the binding site for a negative regulator of fatA may occupy the fatB region that is upstream of the HindIII site in fatB but encompasses the SalI site. The lack of inhibition by the clone pJHC-LW115 may be caused by the truncation of the recognition sequence for the negative regulator or the actual truncation of the negative regulatory genetic determinants.
T o investigate these possibilities, constructs pJHC-LW95, pJHC-LW87, and pJHC-LW96 were made by insertion of the Q fragment, containing transcription and translation stop signals, in pJHC-A122 either at the HindIII site in fatD (generating pJHC-LW95) or at the PuuI site at the 5' terminus of fatB (generating pJHC-LW87) or at the HindIII site at the 3' end of fatB (generating pJHC-LW96) (Fig. 16). The recombinant clones pJHC-LW95, pJHC-LW87, and pJHC-LW96 were then mobilized by conjugation into V. anguillarum H775-3a harboring the mutant 17 plasmid. FatA synthesis was determined by immunoblot of total membrane proteins from strains grown under iron limitation, since FatA is only synthesized in iron-limiting but not in iron-rich conditions (11.12). We observed that the clones pJHC-LW87 and pJHC-LW95 still showed an ability to cause inhibition of FatA synthesis ( Fig. 2a, lanes A and C). However, the inhibitory effect was abolished when the R fragment was inserted at the fatB HindIII site as in pJHC-LW96 (Fig. 2a, lane B). The abrogation of inhibition by the R insertion at thefatB HindIII site (in pJHC-LW96) ruled out the possibility of the existence of a cis binding sequence for a negative regulator. The other possibility, that the inhibitor could be a small peptide was ruled out since not only we were unable to identify any open reading frame encoded within fatB in either strand which would include the Sal1 site, but also no new peptides in addition to those already known were detected in in vitro transcription-translation experiments using as template a plasmid carrying the fatB region (39). Therefore, the 12 insertion studies strongly suggested that an antisense RNA may be transcribed in the the fatB region, from a promoter downstream of the Sal1 site, and that this antisense RNA was likely to affect the expression of the fatA gene negatively.
RNase protection studies were also performed. A riboprobe designed to detect transcripts from the noncoding strand was constructed from the SalI-Hind111 fragment at the 3' region of fatB (Fig. la). Where there was a decrease of the FatA protein synthesis, antisense transcripts were detected in the strains H775-3a (mutant 17 together with either pJHC-A179, pJHC-A122, pJHC-LW95, or pJHC-LW87) grown in ironlimiting medium (Fig. 26, lanes E, D, C, and A). In strain H775-3a (mutant 17 and pJHC-LW96) where inhibition of FatA synthesis was abrogated (Fig. 2a, lane B ) , the antisense transcript was not detected (Fig. 26, lane B). Under  Comparison with lane 23 may not be as straightforward be-cause the construct with the R fragment, although impaired in the production of RNAn, still makes sense RNA which can interact and take out of circulation the basal level of RNAa made by mutant 17 leading to a relief of repression and an increase in the FatA level. These data confirmed that the negative regulator was indeed an antisense RNA, which was designated RNAn. Under the control of its own promoter, as in mutant 17, RNAn is observed to be expressed only in ironrich condition but not in iron limitation (29). Since RNAn was detected in strains H775-3a (mutant 17 together with either pJHC-A179, pJHC-A122, pJHC-LW95, or pJHC-LW87) under iron-limiting conditione, the RNAn transcripts were probably constitutively transcribed via a vector promoter.
Next we analyzed the effect of the antisense RNAn in the wild type context, where the FatA protein is abundantly expressed when compared with mutant 17. We found that in the presence of plasmid pJHC-A122, which encoded RNAn, the level of FatA was dramatically decreased in V. anguillarum 775, carrying the wild type plasmid pJM1, (Fig. 3, compare lanes C and D), and in the strain H775-3a harboring the recombinant clone pJHC-T7, which contained the iron uptake region of pJMl (Fig. 3, compare lanes A and E ) . RNase protection studies were performed using RNA harvested from these strains. The antisense RNA was observed in RNA harvested from strains 775 (pJM1) and H775-3a (pJHC-T7), but only when the cells were grown in iron-rich media (Fig.  4a, lanes D and F ) . RNAn was found in both iron-rich and iron-limiting conditions in the strain carrying both pJHC-T7 and pJHC-A122, which encoded determinants for RNAn cloned at the HindIII site of pACYC184, within the tetracycline resistance gene (Fig. 4a, lanes H a n d I ) . In this case the antisense transcript must be driven by promoters of the tet gene, resulting in the constitutive synthesis of RNAn. In this clone, approximately 350 nucleotides of RNAn were deleted from the 5' end, and the remainder of RNAn started at the HindIII site. Since wild type RNAn is 650 nucleotides long (29), our results demonstrate that less than 50% of the RNA sequence is sufficient for the inhibitory activity.
Regulation of fatA Gene Expression by RNAa-We used RNase protection assays to determine the levels of fatA mRNA. The riboprobe designed to detect the fatA message was constructed using the HindIII-StuI fragment from the 3' region of fatA (Fig. la). The fatA mRNA was observed only in RNA harvested from cells grown under iron-limiting conditions ( Fig. 46, lanes F, H, and J). The level of fatA mRNA from the strain H775-3a (pJHC-T7, pJHC-A122), in which RNAn was synthesized constitutively was slightly but discernibly decreased in comparison with that from H775-3a (pJHC-T7) (Fig. 46, compare lanes J and H). These observations were corroborated by Northern blot analysis of the same RNA preparations (Fig. 4c, compare lanes F and D). However account for the tremendous decrease of FatA protein synthesis.
Mapping of the 5 ' End of the fatA Message-Since the HindIII site where RNAa begins in the inhibitory clone pJHC-A122 is 484 base pairs upstream of the translation start codon of fatA (Fig. la), the fatA transcript must commence a t a location 5' of the fatR HindIII site. Primer extension was therefore performed to map the 5' end of the fatA message. Using an oligonucleotide that encompassed the fatA start codon, we detected numerous RNA species, but we have only labeled the more abundant species (MaI, MaII, and Mb) and the higher molecular weight species (Mc, MHd, and MHe). There were two major RNA species with only one nucleotide difference in size, with their 5' termini located at Ma (Ma1 and MaII), which were 103 and 102 nucleotides upstream of the fatA start codon (Figs. 5 and 6). These species of RNA were decreased in strain H775-3a harboring both the clone pJHC-T7 and the recombinant plasmid pJHC-A122, which encoded extra copies of a truncated RNAa (Fig. 5, lane 3 ) . We also noticed that new RNA species appeared in this strain ( Fig. 5, lane 3, indicated by arrows). The Ma sites are 22 and 23 nucleotides downstream of the RNAa start site, Ra, which was mapped previously (29). We also detected two minor fatA mRNA species, with 5' termini a t Mb and Mc, 243 and 362 nucleotides respectively, 5' of the fatA start codon, and therefore Mb is 118 nucleotides upstream (in the sense direction) of the RNAn start site Ra, whereas Mc coincides with the RNAa start site Rh. The results in this section suggest that the secondary fatA mRNA species, with 5' ends at sites Mb and Mc, but not the major species from site Ma, overlap and thus can interact with the antisense RNA species initiated from start site Ra.
We also detected fatA mRNA molecules, with 5' termini mapped to sites at MHd and MHe, which were distinctly present in strain H775-3a (pJHC-T7, pJHC-A122) (Fig. 5 , lane 3), and were barely discernible in the wild type strain 775. Sites MHd and MHe were approximately 593 and 713 nucleotides, respectively, from the fatA start codon. These higher molecular weight fatA mRNA species were also present in the strain harboring pJHC-T7 alone but in such low abundance that we were unable to reproduce them in photographs.
Fur Regulation of fatA Expression-The presence of RNAa is concomitant with a slight reduction of fatA mRNA levels. However, under iron-rich conditions there is a complete shutoff in the synthesis of fatA mRNA. It is therefore possible that in addition to RNAa inhibition, fatA expression may be affected by another regulatory mechanism that acts at the transcription level, that is responsive to the concentration of iron, for example a Fur-like repressor protein, as already described in other bacteria (40)(41)(42)(43)(44)(45). T o investigate this possibility, pJHC-T7 was introduced into E. coli BN4020 (fur::Tn5) (16) and into the isogenic strain which also harbored pMH15 (25), a plasmid carrying the E. coli fur gene. We observed that, although poorly expressed in E. coli, the fatA mRNA levels were found to be the same in iron-rich and iron-limiting conditions in the Fur-strain E. coli BN4020 (Fig. 7, lanes H and I). However, when the fur clone pMH15 was also present in this strain, the level of fatA mRNA was lower in RNA harvested from cells grown in iron-rich medium as compared with those grown under iron-limiting condition (Fig. 7, lanes J and K ) . The data suggest that the Is. coli Fur repressor could function, although very inefficiently, as a repressor in the regulation of the V. anguillarurn fatA gene.
Genetic and Functional Detection of fur in V. anguillarurn-Because under iron-rich conditions no fatA mRNA was detected, it is possible that a fur-like gene that encodes a more efficient product exists in V. anguillarurn. T o detect the presence of a Fur-like repressor in V. anguillarurn H775-.7a, we introduced into this bacterium the plasmid pSC27.1 which has a Fur binding sequence between the ornpF promoter and the lac2 gene (27). In this construct S-galactosidase production will be repressed in iron-rich conditions if a Fur-like protein exists. The control was plasmid pRT240 (27,45), which is identical to pSC27.1 without the Fur binding sequence. V. anguillarurn H775-3a (pRT240) produced /j-galactosidase constitutively, whereas V. anguillarurn H77.5-3a (pSC27.1) showed a 0.24 inhibition ratio of @-galactosidase activity when grown under iron-rich as compared with ironlimiting conditions (Table I). These results indicate that V. anguillarurn must have a chromosomally encoded Fur-like element.
Cloning of the fur-like Gene from V. anguillarurn-Physical detection of a fur-like gene in V. anguillarurn was performed by Southern blot hybridization of HindIII-digested chromosomal DNA using as a probe the E . coli fur gene obtained from plasmid pMH15 (25). Under low stringency conditions the probe hybridized with two bands of 6.6 and 3.5 kb. Hybridization was not observed under high stringency conditions, indicating a degree of divergence at the nucleotide level between the E. coli and the V. anguillarurn fur genes.
Isolation of a recombinant clone containing the V. anguiflarum fur-like gene was performed using a gene library (26) of the plasmidless V. anguillarurn H775-8 in the vector pVK102. Fig. 8, a and b, shows an example of hybridization   (Miller, 1972). experiments carried out to detect a recombinant clone harboring the fur-like gene. We obtained the recombinant plasmid pMET67, which included the 6.6-and 3.5-kh HindIII fragments that hybridized with the E. coli fur probe. The Fur activity of this recombinant clone was tested as described in the previous section. E. coli BN4020 (Fur-) already carrying pRT240 or pSC27.1 was transformed with either pMET67 or pMH15, and &galactosidase activity was determined in cultures grown under iron-rich conditions. Table I1 shows that the presence of the V. anguillarum fur-like gene (pMET67) resulted in an inhibition of the /3-galactosidase activity produced by plasmid pSC27.1, similar to that observed for a clone carrying the E. coli fur gene (pMH15). Neither pMET67 nor pMH15 modified significantly the @-galactosidase activity mediated by plasmid pRT240. The results in this section suggest that pMET67 carries a functional fur-like gene from the V. anguillarum chromosome.

DISCUSSION
In this work we demonstrated that at least two mechanisms may play an important role in the repression of the expression of the fatA gene when V. anguillarum 775 is grown under iron-rich conditions. One of these mechanisms is mediated by an antisense RNA (RNAa) that is induced under iron-rich conditions, and the other may be via a Fur-like protein that functions as a repressor in the presence of iron. In the first case, the presence of the antisense RNA leads to a dramatic reduction of FatA biosynthesis, whereas Northern blot and primer extension experiments showed that the level of the steady-state fatA transcripts was only slightly decreased.
Primer extension experiments also revealed the existence of various fatA mRNA with different 5' termini mapped to sites MaI, MaII, Mb, Mc, MHd, and MHe. Therefore the 2. 35 Our previous in oitro transcription-translation analysis indicated that a promoter may exist 3' of the Sun site, since insertion of an R fragment at the SalI site of an EcoRI-PstI fragment still produced FatA protein, albeit at a barely detectahle level (39). The fatA mRNA from this construct must be either at a very low level or is functionally impaired for translation and therefore plays a very small part in the overall expression of fatA. Our data indicate that fatA is most likely to be predominantly transcribed as part of a polycistronic message.
In the analysis of the E. coli lac mRNAs, it was found that the lac messages were transcribed as a polycistronic lacZYA message, which was subsequently cleaved to form six individual lac mRNAs (46,47). It was demonstrated that for the lacZYA transcript, cleavage at the start of the upstream message (lacy) inactivated the ability of the distal message (lacA) to participate in the formation of an initiation complex with ribosomes (48). In the V. anguillarum fat operon, we may have a similar phenomenon. The level of the 2.35-kb fatA mRNA was only very slightly decreased in the presence of extra RNAa, encoded by pJHC-A122. However, in this strain, little or no FatA protein was synthesized. A plausible explanation for this phenomenon is that this population of fatA mRNA may be functionally inactivated for translation in the same manner as that reported for the lacA message in the lac operon.
One of the mechanisms for functional inactivation of messages is a change in the secondary structure of the leader sequence 5' of the ribosome binding site of the mRNAs. It has been shown that changes in the translation initiation region influenced the translation efficiency of the fnr gene and of genes of the gal, trp, and mal operons (49)(50)(51)(52). In this study the decreased FatA synthesis, as assessed hy immuno-Iron Regulation of fatA Expression blot analysis, may be a result of secondary structure modifications of the fatA mRNA, generated by interaction between fatA mRNA and RNAa, prior to or after RNase I11 cleavage of the hybrid structure. Changes in secondary structure could lead to a decrease in translational efficiency, for example, by leading to inaccessibility of the ribosome binding site of f a t A mRNAs that start at sites upstream of Ma. Since pJHC-A122 encoding the truncated RNAa was able to exert such a profound effect on the translation of fatA, the sequences involved in generating the changes in the secondary structure must be located 5' of Mc and upstream of the Hind111 site in fatB.
It was of interest that primer extension of fatA mRNA obtained from V. anguillarum strains harboring the iron uptake clone (pJHC-T7) together with the RNAa clone (pJHC-A122) produced novel RNA species, in addition to those detected in the strain harboring only pJHC-T7. It is conceivable that the interaction between the large fatA mRNA molecules (initiated from start sites upstream of Ma) and the antisense RNA resulted in conformational changes in the f a t A mRNA molecules and exposed various RNA endonuclease sites. This possibility may account for the appearance of the novel fatA mRNA molecules and a slight decrease of the Ma1 and Ma11 fatA mRNA species from the strain H775-3a (pJHC-T7, pJHC-A122). From all of the above, we concluded that RNAa may regulate fatA expression primarily at the level of translation, although the data also indicate that the rate of processing of the full-length fatA transcript may also be affected. Had the translational control not been so prominent, the destabilizing effect of RNAa on the fatA message would be more readily observable.
If fatA mRNA is not part of a fatDCBA polycistronic message, a possible mechanism is that RNAa may affect the expression of fatA by interacting with a positive regulator that may be essential for the translation of the fatA message. However, a positive regulator of translation is very rare. The sole example reported was that of RNase 111-induced secondary structure changes in the region upstream of the ribosome binding site of the CIII mRNA of bacteriophage X (53, 54).
In this study we have also demonstrated that the antisense RNA is not the only regulatory mechanism. Since fatA mRNA is not detected under iron-rich conditions and RNAa does not appear to greatly change the levels of this mRNA, we therefore hypothesize that another regulator, such as a Furlike product, may exist in V. anguillarum which could act a t the level of transcription when the concentration of iron increases in the growth media. In this vein we were able to demonstrate that V. anguillarum indeed has a chromosomally mediated Fur-like protein that could recognize the E. coli Fur binding site. The V. anguillarum Fur-like product must be related to the E. coli Fur protein because it could be expressed and was functional in E. coli as well as in V. anguillarum. However, the V. anguillarum fur-like gene hybridized with that from E. coli only under low stringency conditions, indicating that there must be considerable divergence at the nucleotide level between the E. coli and the V. anguillarum f u r genes. The difference may account for the inefficiency of E. coli Fur in the regulation of the V. anguillarum fatA gene. It remains to be demonstrated whether the V. anguillarum f u r gene product is responsible for the dramatic reduction of fatA mRNA under iron-rich conditions. Our present evidence suggests that the complete shutoff of f a t A expression when the iron concentration of the medium shifts from a low to a high level is mainly controlled at the transcriptional level.
The mechanism is most likely to be mediated by a Fur-like repressor protein. The antisense RNA then functions as a fine-tuning mechanism in the iron regulation of the pJMl fatA gene, by inactivating the transcripts that have already been initiated. Inactivation is primarily at the translational level, although evidence suggests that RNAa may also contribute to the destabilization of the fatA messages.