Unexpected Functional Diversity among FadR Fatty Acid Transcriptional Regulatory Proteins *

The FadR protein of Escherichia coli has been shown to play a dual role in transcription of the genes of bacterial fatty acid metabolism. The protein acts as a repressor of (cid:1) -oxidation and an activator of unsaturated fatty acid synthesis. FadR DNA binding is antag-onized by long chain acyl-CoAs, and thus FadR acts as a sensor of fatty acid availability in the environment. When viewed from a genomic viewpoint, FadR proteins are unusual in that the DNA binding domain is very highly conserved among FadR-containing bacteria, whereas the C-terminal acyl-CoA binding domain shows only weak conservation. To further our understanding of the role of FadRinbacteriallipidmetabolismwehaveexaminedthe invivo and in vitro properties of a diverse set of FadR proteins expressed in E. coli . In addition to E. coli FadR the proteins examined were those of Salmonella enterica , Vibrio cholerae , Pasteurella multocida , and Haemophilus influenzae . These FadR proteins were found to differ markedly in their effects on repression and induction of (cid:1) -oxida-tion in E. coli and in their acyl-CoA binding abilities as measured by isothermaltitrationcalorimetry.The E. coli and S.enterica proteins were the most similar, although they differed in their effects on utilization of oleic acid and acyl-CoA binding affinities,whereas the P. multocida and H. influenzae proteins showed only weak repres-sionandpooracyl-CoAbindingaffinities.The V.cholerae FadRwas strikingly superior to the other proteins in the amplitude of its regulatory

Escherichia coli FadR protein is a transcription factor that plays a central role in the regulation of fatty acid metabolism. FadR specifically represses the transcription of each of the genes essential for fatty acid transport, activation and ␤-oxidation, including fadL, fadD, fadE, fadBA, fadF, fadG, fadH, and fadIJ (1)(2)(3)(4)(5)(6). FadR also acts as a transcriptional activator of the fabA and fabB genes of the unsaturated fatty acid biosynthetic pathway (7,8) and of the iclR gene that encodes the glyoxylate operon repressor (9). The function of FadR as a repressor or an activator is readily explained by the location of its binding sites within the promoter regions. When the binding site either overlaps the Ϫ10 or Ϫ35 regions or lies just downstream of the Ϫ10 sequence (6,8,10) FadR represses transcription. However, when the binding sites are located immediately upstream of the Ϫ35 region of the promoter, FadR binding activates transcription presumably by promoting the binding or action of RNA polymerase. The apparent role of FadR is to optimize fatty acid metabolism in response to exogenously supplied fatty acids (6).
Addition of long chain (ϾC 12 ) fatty acids induces the fad enzymes, whereas medium (C 7 -C 11 ) or short chain-length (C 4 -C 6 ) fatty acids cannot (11,12). Therefore, wild-type E. coli strains utilize long-chain fatty acids such as oleate (C18:1), but not medium-chain-length fatty acids such as decanoate (C10:0) as sole carbon and energy sources. However, mutants that grow on decanoate are readily isolated by plating wild-type cells onto minimal media containing this acid as sole carbon source (11,12). These stains have constitutive levels of the fad enzymes and have mutations that map in fadR (2,11,13). The utilization of fatty acids shorter than C6 requires enzymes other than those encoded by the fad regulon (3,14).
E. coli FadR is a 239-residue protein that functions as a dimer (15). The N-terminal DNA binding domain has the winged helix structure often found in prokaryotic regulator proteins (16,17), whereas the C-terminal half of FadR contains the acyl-CoA binding domain (18,19). In recent years four FadR crystal structures have been reported from two research groups and show a gratifying agreement and complementarity. One structure is of the unliganded protein (20), another is an FadR-myristoyl-CoA complex (19), and two others are complexes of FadR with the fadBA operator DNA (19,21). Given these crystal structures it is instructive to examine alignments of FadR homologues (see Fig. 1). Note that in the completed bacterial genomes FadR is found only in those ␥-proteobacteria that colonize vertebrates (or in one case plants). This distribution suggests that FadR regulation evolved to utilize fatty acids available in these ecological niches. The alignments are unusual in that the N-terminal DNA binding halves show extremely strong sequence conservation (Ͼ95% identical residues), whereas the C-terminal acyl-CoA binding halves are much less conserved (ϳ25% identical residues). Moreover, all of the residues critical for DNA binding by E. coli FadR are conserved in all known FadR proteins, which strongly implies that the operator sequences bound by FadRs of these disparate organisms are the same or very similar to those of E. coli. Indeed, in these genome sequences one can find plausible operator sequences upstream of the relevant genes (22). In contrast, the C-terminal acyl-CoA binding regions are much more diverse. For example, in the E. coli FadR helices that contact myristoyl-CoA in the co-crystals (19,21) only 20 -30% of the residues are conserved in the other FadR proteins. These findings suggest that different FadR proteins may be tuned to bind different acyl-CoA ligands with differing affinities.
In the present study, we analyzed the FadR proteins of E. coli, Salmonella enterica, Vibrio cholerae, Pasteurella multocida, and Haemophilus influenzae. The E. coli and S. enterica FadR proteins differ in only 7 of 239 residues, but interestingly all but one of the mismatches are in the C-terminal halves (Fig. 1). The other FadR homologues studied were chosen because their sequences seemed the most different from E. coli and from one another (Fig. 1). This includes P. multocida and H. influ-enzae, which despite both being Pasteurellaceae, have FadRs that are only 54% identical. V. cholerae FadR is unusual in that relative to the E. coli protein 40 residues are inserted into the midst of the protein (Fig. 1).

EXPERIMENTAL PROCEDURES
Media and Growth Conditions-Rich broth contained 10 g of Tryptone, 5 g of NaCl, and 1 g of yeast extract per liter. 2ϫ YT medium contained 16 g of Tryptone, 5 g of NaCl, and 10 g of yeast extract per liter. Minimal medium was M9 medium (23) supplemented with 1 mM MgSO 4 , 0.1 mM CaCl 2 , and 0.5 mg/liter thiamine. Acetate was used at a final concentration of 0.4% as sole carbon source. Fatty acids were neutralized with KOH and solubilized with Tergitol NP-40 and were used at final concentrations of 1 g/liter (w/v) as carbon source and at 5 mM for induction. Solid media contained 1.5% (w/v) Bacto agar (Difco, Milwaukee, WI). Antibiotics were used at the following concentrations (in mg/liter): ampicillin, 100; kanamycin, 50; tetracycline, 12; and chloramphenicol, 25.
Bacterial Strains and Plasmids-All bacterial strains are derivatives of E. coli K-12. Phage transduction and other basic genetic techniques were generally carried out as described by Miller (24). Strains MFH9 (fadR::Tn5) and MFH13 (fadR::Tn5, fabA ts ) (8) were used for fadR complementation and characterization of growth phenotype on mini-mal media containing fatty acids. The transcriptional fusion strain SI203 (fadBA-lacZ) was used to assay induction of the fad regulon genes. Strain SI207 (fadR fadBA-lacZ) derived from SI203 was used to assay the repressor activity of plasmid encoded FadR homologues. Strain BL21(DE3)/pLysS was used as host for overproduction of FadR proteins (25). Other strains used in this study were: UB1005 (F Ϫ metB1 relA1 spoT1 gyrA216), MG1655 (rph-1 fnr?), and MC1061 (F Ϫ araD139 ⌬(ara-leu)7696 galE15 galK16 ⌬(lac)X74 rpsL hsdR2 (r K Ϫ m K ϩ ) mrcA mcrB1. Strain SI203 was constructed using Red and flippase (FLP) 2mediated site-specific recombination (26). The aminoglycoside 3Ј-phosphotransferase (Kan) gene of plasmid pKD4 (27) was amplified via PCR using primers 1 plus 2 (see TABLE ONE). The resulting PCR product was used to replace the entire coding sequence of fadBA in strain MG1655 by homologous recombination catalyzed by the red genes encoded by plasmid pKD46 (27). The chromosomal structures of several kanamycin-resistant transformants were verified by colony PCR using primers three plus four, and one of these was called strain SI186. The temperature-sensitive plasmid pCP20 (28)  cassette leaving behind a single FRT site. The FRT site in the resulting strain was then used for FLP recombinase site-specific integration of a lac fusion plasmid pCE70 (26) containing a FRT site upstream of the promoterless lacZY genes, a kanamycin resistance gene, and the R6K origin of replication. Transformants were plated on RB plates containing kanamycin and 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside at 37°C. This resulted in the stable integration of the fusion plasmid due to the loss of the temperature-sensitive replication of pCP20 to produce fadBA-lacZ transcriptional fusion strain SI198. Strain SI203 was obtained by P1 transduction of strain MC1061 with a lysate grown on SI198 and selection for kanamycin resistance. Strain SI207 was generated by P1 transduction of strain SI203 with a lysate grown on MFH8 (8) followed by selection for tetracycline resistance. Plasmid pSH11 was made by PCR amplification of the fadR gene from the chromosomal DNA of strain UB1005 using primers 5 plus 6 (TABLE ONE). The resultant 983-bp PCR product was purified via Qiagen spin columns and cloned into pCR2.1 (Invitrogen). The insert of the resultant plasmid was sequenced to verify the fidelity of the cloned PCR product. The sequenced plasmid was digested with HindIII and ClaI (sites introduced by the primers) and ligated to pACYC177 cut with the same enzymes. Note that this manipulation inserted the fadR gene into the kanamycin resistance gene of the vector in the orientation opposite to that of the resistance gene and its promoter. Similarly, plasmids pSH12, pSH13, and pSH14 were constructed using the PCR products of S. enterica, V. cholerae, and P. multocida genomic DNA obtained with primer sets 7 plus 8, 9 plus 10, and 11 plus 12, respectively. The resultant PCR products were cloned into pCR2.1 and then subcloned into pACYC177 using HindIII-ClaI double digest for pSH12 and pSH13, whereas BamHI-ClaI double digest was used in case of pSH14. Plasmid pSH44 was constructed by subcloning a 1850-bp fragment carrying H. influenzae fadR gene from pGHIDM89 into pACYC177 using a HindIII-XhoI double digest. Plasmids pSH56 and pSH57 contain the fadR coding sequence of P. multocida and H. influenzae and the native promoters of both genes are replaced by E. coli fadR promoter sequence. Plasmids pSH56 and pSH57 were made by PCR amplification using pSH14 and pSH44 as templates with primer sets 13 plus 14 and 15 plus 16, respectively. Both forward primers 13 and 15 contained a 68-bp E. coli fadR promoter sequence that included the fadR ribosome binding site and 20 bp of sequence homologous to the first codons of either the P. multocida or H. influenzae fadR coding sequences. The resulting PCR products were cloned into pCR2.1 to give pSH54 and pSH55. A BamHI-XhoI fragment from pSH54 and a HindIII-XhoI fragment from pSH55 (sites from pCR2.1) were subcloned into pACYC177 cut with the same enzymes to produce pSH56 and pSH57, respectively. Plasmid pSH89 encoding V. cholerae FadR⌬132-171 was constructed by overlap PCR (29). Briefly, two separate PCR amplifications were done from template plasmid pSH13 using primer sets 9 plus 27 and 28 plus 10. Primers 27 and 28 were designed such that their 3Ј ends hybridized to a template sequence on one side of the deletion, and the 5Ј ends are complementary to a template sequence on the opposite side of the deletion. The resulting PCR products of expected size were purified via Qiagen spin columns and used together as templates in a subsequent fusion PCR reaction with primers 9 plus 10. The PCR product of expected size was again purified and cloned into pCR2.1 (Invitrogen). The insert of the resultant plasmid was sequenced and subcloned into pACYC177 using HindIII-ClaI double digest (sites within the primers) to give plasmid pSH89.

Construction of Genes Encoding His 6 -tagged FadR Homologues-
The fadR gene of strain UB1005 was amplified via PCR using Easy-A High-Fidelity Polymerase (Stratagene) under standard PCR conditions using primer set 17 plus 18. The resultant 822-bp PCR product was purified via Qiagen spin column and cloned into pCR2.1 (Invitrogen). The insert of the resultant plasmid was sequenced to verify the fidelity of the cloned PCR product. The sequenced plasmid was digested with NdeI and BamHI (sites within the primers) and ligated to pET16b cut with the same enzymes to give in-frame fusion of the coding sequence of the FadR N terminus to that of the His tag of the vector to give pSH77. Plasmids pSH78, pSH79, pSH80, and pSH81 were constructed in the same manner using primer sets 19 plus 20, 21 plus 22, 23 plus 24, and 25 plus 26, respectively, with the appropriate chromosomal DNA preparations.
Western Blot Analyses-Western blotting of cell extracts from E. coli fadR strains transformed with the plasmid constructs harboring the different fadR homologues or the empty vector control was carried out as follows. Cultures were grown overnight in RB medium plus ampicillin, then sub-cultured 1:100 in the same media and grown until mid-log phase. Crude extracts were loaded on an equal cell number basis and separated on 12% resolving SDS-polyacrylamide gels using a Mini-Protean II apparatus (Bio-Rad). The separated proteins were then electrophoretically transferred to Immobilon-P membranes (Millipore) for 60 min at 90 V. The membrane was incubated first in TBS buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl) containing 0.05% Tween 20 and 5% nonfat milk powder for 2 h at room temperature with gentle shaking. Following this blocking step the membrane was probed with rabbit polyclonal antibody (raised against E. coli His 6 -FadR protein) diluted 1:1000 in the antibody buffer (0.25% Triton X-100 and 2% nonfat milk powder in TBS buffer) for an additional hour. After rinsing four times with wash buffer (0.05% Tween 20 in TBS buffer), the membrane was incubated with a secondary antibody conjugated with horseradish peroxidase (Amersham Biosciences) diluted 1:20,000 in antibody buffer for 1 h at room temperature. The membranes were washed as above and the FadR proteins were visualized by incubation of the membrane in ECL Plus chemiluminescent detection reagents (Amersham Biosciences) and exposure to ECL Hyperfilm (Amersham Biosciences).
␤-Galactosidase Assays-Overnight cultures were grown in either RB medium or RB medium supplemented with oleate. The overnight cultures were subcultured into the same medium and shaken at 37°C. When the cultures reached mid-log phase the cells were pelleted and then washed twice with RB medium containing 0.5% Tergitol NP-40 and three times with RB medium to remove fatty acid and detergent. After the final rinse the cells were resuspended to 4 ϫ 10 8 cells/ml and assayed for ␤-galactosidase activity after chloroform/sodium dodecyl sulfate lysis as described by Miller (24). The cell debris in the assay mix was removed by centrifugation prior to reading the absorbance at 420 nm. All cultures contained Tergitol NP-40 and received the washing treatment regardless of oleate supplementation.
Expression and Purification of His 6 -FadR Proteins-The His 6 -FadR proteins of E. coli, S. enterica, V. cholerae, H. influenzae, and P. multocida were overexpressed in E. coli BL21(DE3)/pLysS harboring a FadR coding plasmid using the T7 polymerase expression system (25). Cultures (1 liter) were grown in LB broth supplemented with ampicillin and chloramphenicol at 37°C until an optical density of ϳ0.5 was reached, and then 0.4 mM isopropyl-␤-D-thiogalactopyranoside was added followed by incubation for 3 h. The cells were harvested by centrifugation at 4,000 ϫ g for 20 min. The pellets were resuspended (4-fold concentration) in ice-cold 20 mM Tris-HCl (pH 8.0) and centrifuged for 5 min at 4,000 ϫ g. The pellets were stored at Ϫ80°C until further use. All further purification steps were performed at 5°C. For purification of the His-tagged proteins frozen cell pellets were thawed on ice for 15 min prior to the addition of 20 ml of buffer A (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 10 mM imidazole). Lysozyme (1 mg/ml) was added, and after 30 min of incubation on ice the cells were disrupted via sonication. The lysates were centrifuged at 10,000 ϫ g for 30 min, and 3 ml of Ni 2ϩnitrilotriacetic acid resin (Qiagen) was added to the resultant supernatant. The slurry was rotated slowly at 5°C for 60 min, and the resin was loaded into a 5-ml column and washed with 5 volumes of Buffer B (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 20 mM imidazole). The His 6tagged protein was eluted by 1 ml of buffer C (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 300 mM imidazole). The elution was repeated 4 times. The purity of the samples was monitored via SDS-PAGE. Fractions containing the protein of interest were pooled and dialyzed against two changes of calorimetry buffer (20 mM Tris-HCl (pH 7.0) 50 mM NaCl, 1 mM EDTA, 0.1 mM Tris(2-carboxyethyl)phosphine hydrochloride) for 12 h at 5°C. Following dialysis 10 mM 2-mercaptoethanol and 10% glycerol were added prior to storage at Ϫ80°C.
Isothermal Titration Calorimetry-Isothermal titration calorimetry experiments were performed using a VP-isothermal titration calorimeter from MicroCal Inc. (Northampton, MA). This instrument has been described in detail by Wiseman et al. (30). To improve baseline stability, the temperature of the system was kept about 5°C below the working temperature. All solutions were degassed under vacuum with stirring immediately before use. All samples were prepared in, and thoroughly dialyzed against, the same batch of buffer to minimize artifacts due to minor differences in buffer composition. The ligand concentration in the titration syringe was usually 10-fold higher than the protein concentration in the cell. The protein concentrations were ϳ0.01-0.03 mM and were measured spectrophotometrically at 280 nm using calculated (31) molar extinction coefficients (M Ϫ1 cm Ϫ1) ) of 33,120, 33,120, 37,200, 34,520, and 35,800 for the E. coli, S. enterica, V. cholerae, H. influenzae, and P. multocida His 6 -FadRs, respectively. The experiments were performed at 30°C, and injections were begun after equilibration to baseline stability. Each injection consisted of a volume of 10 l and a 10-s duration, with a 240-s interval between injections. The syringe was rotated at 300 rpm during the experiment to assure immediate mixing. Nonspecific heat effects were estimated from the magnitude of the peaks appearing after complete saturation. Raw data were integrated, corrected for nonspecific heat, and analyzed using the ORIGIN software supplied with the instrument. The total heat of binding (⌬H), the association constant (K a ), and the binding stoichiometry (n) were obtained by nonlinear regression fitting to the isotherm.

RESULTS
Cloning and Expression of fadR Gene Homologues-The fadR gene homologues of E. coli, S. enterica, V. cholerae, P. multocida, and H. influenzae were cloned into the moderate copy number vector pACYC177 under control of their native promoters to give plasmids pSH11, pSH12, pSH13, pSH14, and pSH44, respectively (TABLE TWO). We used the pACYC177 vector due to prior work (32), and because it seemed possible that some of the FadR homologue proteins might be weak repressors, the function of which could be overlooked at lower expression levels. These plasmids were introduced into the E. coli fadR null mutant strain MFH9, and the FadR expression was assayed in extracts of these strains by Western blotting using a polyclonal antibody raised against E. coli FadR (Fig. 2). These analyses showed that the plasmids encoding the E. coli, S. enterica, and V. cholerae FadRs expressed similar levels of the FadR proteins. However, the P. multocida and H. influenzae FadR proteins were only very weakly expressed. This could be due to either a lack of cross-reactivity with the antibody raised against E. coli FadR and/or poor recognition of the Pasteurellaceae fadR promoters by the E. coli transcriptional machinery. Because the latter hypothesis seemed more likely, we replaced the native promoters and ribosome binding sites of the P. multocida and H. influenzae fadR genes in plasmids pSH14 and pSH44 with the E. coli fadR promoter to obtain plasmids pSH56 and pSH57, respectively ("Experimental Procedures"). Western blotting indicated that the levels of expression of the P. multocida and H. influenzae FadRs from these plasmids were comparable to those of FadRs from the other organisms. Note that extracts of the fadR strain carrying the parental vector plasmid pACYC177 contained no detectable FadR.
Complementation Analyses-As described earlier, wild-type E. coli strains utilize long chain fatty acids such as oleate but not medium chain fatty acids such as decanoate as sole carbon and energy sources (2,11,12). However, strains encoding defective FadR proteins grow on decanoate as a sole carbon and energy source due to the constitutive fad enzyme expression (2,(11)(12)(13). We therefore tested the ability of the foreign FadRs to block growth of the fadR::Tn5 null strain MFH9 on minimal medium plates containing decanoate as the sole carbon and energy source (  (TABLE  THREE). These results showed that all FadR homologues are functional in E. coli and that H. influenzae FadR appears to have weaker operator binding affinity, because repression was seen only at elevated FadR levels. E. coli FadR also activates the transcription of fabA and fabB genes required for unsaturated fatty acid biosynthesis (7,8). Strains carrying a fabA ts mutation grow well at 30°C but fail to grow at 42°C without unsaturated fatty acid supplementation, whereas strain MFH13 (fadR, fabA ts ) requires unsaturated fatty acid supplementation at all growth temperatures (33). To test if the FadR homologues were able to activate fabA expression, plasmids encoding each of the proteins were introduced into MFH13 and transformants were assayed for growth at 30°C on RB medium lacking unsaturated fatty acid supplementation. All FadR homologues permitted growth of the double mutant strain (  (Fig. 2) were tested on oleate minimal medium plates (Fig. 3). The increased levels of E. coli FadR produced from the multicopy plasmid resulted in strong inhibition of growth on oleate in agreement with previous findings (32). Expression of S. enterica FadR at similar levels blocked growth on oleate suggesting that this FadR protein was a more potent (or less easily neutralized) repressor than E. coli FadR. In contrast, to the E. coli and S. enterica proteins expression of the V. cholerae or P. multocida FadRs at similar intracellular levels (Fig. 2) had no effect on growth with oleate; the strains grew as well as the strain that carried an empty vector (Fig. 3). Expression of H. influenzae FadR from the high copy plasmid  also had no effect on growth on oleate (Fig. 3). All strains grew well on minimal medium plates containing acetate as sole carbon and energy source showing that the defective growth on oleate was due to inhibition of induction of the ␤-oxidation pathway. Due to the blockage of growth on oleate by the S. enterica FadR we measured the relative levels of E. coli and S. enterica FadR proteins in these strains using the purified FadR proteins as standards (Fig. 4). The strain expressing S. enterica FadR contained 1.7 g of FadR per 10 9 cells, a value very similar to that of the strain expressing E. coli FadR (2.1 g of FadR per 10 9 cells). Thus the greater growth inhibition of the S. enterica FadR was not due to higher levels of expression.
The Effects of FadR Homologues on fad Regulon Transcription in Vivo-To quantitate the efficiencies of FadR repression by the various homologues as well as the efficiencies of induction by oleate, we analyzed fad regulon transcription using lacZY transcriptional fusions. For this purpose we constructed transcriptional fusions of both the fadBA operon and the fadE gene in the chromosome of E. coli using homologous recombination catalyzed by the Red proteins followed by removal of the antibiotic cassette and insertion of the lac genes by FLP-mediated site-specific recombination ("Experimental Procedures"). The two fusion constructs gave very similar results, and hence we will present only those data obtained using the fadBA-lacZY fusion. This fusion was transduced into the ⌬lac strain MC1061 by selection for kanamycin resistance to give strain SI203. A fadR null mutation derivative of strain SI203 was then constructed by P1 transduction with a lysate grown on MFH8 and selection for tetracycline resistance to give strain SI207 (fadR fadBA-lacZY). As expected ␤-galactosidase activities in strain SI203 were low in RB medium lacking inducer and high in the RB medium supplemented with oleate, whereas the ␤-galactosidase activities of its fadR derivative strain, SI207, were high in cells grown in either medium (TABLE FOUR). ␤-Galactosidase activities were also determined for the derivatives of the fadR strain SI207 that carried plasmids encoding the various FadR homologues. It should be noted that the presence of E. coli fadR on pACYC177 (the vector we used) gave about 2-fold stronger repression of the fad regulon than that of a wild-type cell (32) presumably due to increased occupancy of the fad operators by the increased number of FadR molecules. Similar ␤-galactosidase levels were seen in derivatives of strain SI207 that expressed the E. coli, S. enterica, or V. cholerae FadR proteins under non-inducing conditions and were about ϳ20-fold than those seen in the strain carrying the empty vector, pACYC177 (TABLE FOUR). In contrast we observed only a 5-fold repression of ␤-galactosidase activity in the strain that expressed P. multocida FadR from same vector. Because the H. influenzae FadR expressed from the high copy number vector gave only a 7-fold repression of ␤-galactosidase activity, it seems that both of the Pasteurellaceae FadRs bound the fadBA operator less strongly than did the other proteins. Significant differences were also observed when ␤-galactosidase activities were determined in the fadR plasmid-containing strains after induction with oleate. The highest levels of induction were observed in the strain that expressed V. cholerae FadR where the ␤-galactosidase activity was elevated 14-fold relative to non-induced levels, whereas only 6-and 2.5-fold increases were observed in the same strain expressing the E. coli or S. enterica FadRs, respectively. Both the P. multocida and H. influenzae FadR proteins, which showed poor repression of fadBA promoter activity, also showed weak responses to oleate sup-    plementation. These results suggest that the P. multocida and H. influenzae FadR proteins possess significantly lower affinities for oleoyl-CoA as well as the fadBA operator than the other FadR proteins studied. Although the FadR proteins of E. coli, S. enterica, and V. cholerae had similar affinities for the fadBA promoter region in vivo, the transcriptional reporter activity in oleate supplemented medium indicated sub-stantial differences in the interactions of these FadR proteins with oleoyl-CoA (TABLE FOUR), which provides an explanation for the growth differences observed on oleate plates (see below).
Isothermal Titration Calorimetry-Given the interdependency between acyl-CoA binding and DNA binding, we needed a method to directly measure acyl-CoA binding of the FadR proteins. The binding affinities of different His 6 -FadR homologues for acyl-CoA were compared using isothermal titration calorimetry, which allows simultaneous determination of the dissociation constant (K d ), enthalpy (⌬H°), entropy (⌬S°), as well as the stoichiometry (n) of binding (Figs. 5 and 6 and TABLES FIVE and SIX). Typical isothermal titration calorimetry profiles of the binding of oleoyl-CoA to S. enterica FadR with the raw and integrated data are shown in Fig. 5. The negative displacement of the peaks is typical of an exothermic reaction. The calorimetric data were analyzed using the ORIGIN software supplied with the instrument, and the binding isotherms were fitted using a model based on a single set of binding sites. The predictions from the in vivo work were confirmed by the isothermal titration calorimetry results where V. cholerae FadR showed a 2.5-fold higher affinity compared with E. coli FadR, whereas S. enterica FadR had about 2-fold reduced affinity for oleoyl-CoA (TABLE FIVE). In contrast both P. multocida and H. influenzae FadR homologues demonstrated very poor binding of oleoyl-CoA, ϳ30fold lower affinity than that of E. coli FadR. The thermodynamic parameters of oleoyl-CoA binding to the different FadR proteins are given in TABLE FIVE. The specificities of the different FadR proteins for various chain length fatty acyl-CoAs were also determined (TABLE SIX). No significant binding of the short chain fatty acyl-CoA, hexanoyl-CoA, was detected for any of the FadR proteins, and decanoyl-CoA binding was weak (or absent in the case of the Pasteurellaceae proteins). S. enterica FadR bound palmitoyl-CoA with slightly reduced affinity compared with E. coli FadR, whereas it demonstrated slightly higher affinity for decanoyl-CoA. Plots of the integrated heat signals from the calorimetric titration of S. enterica FadR with different chain length acyl-CoAs are presented in Fig. 6. V. cholerae FadR exhibited the highest binding affinities for both long and medium chain length acyl-CoAs having K d values of 25, 64, and 1041 nM for oleoyl-CoA, palmitoyl-CoA, and decanoyl-CoA, respectively (TABLE SIX). The dissociation con-  stants for the interactions of long chain fatty acyl-CoAs with E. coli FadR we measured agree well with those reported by DiRusso and coworkers (34).

DISCUSSION
In E. coli FadR plays a dual role in transcriptional regulation of the levels of the enzymes of fatty acid biosynthesis and degradation. The role of FadR in fatty acid synthesis is to decrease the activity of the unsaturated fatty acid synthetic pathway when exogenous fatty acids are available for membrane lipid synthesis (6), whereas in the fad regulon it functions as a classic transcriptional repressor (2,4,10,13,18,35). E. coli FadR has been extensively characterized, but the function of FadR in other bacteria had been addressed only by computational analysis of putative operator sequences (22). These investigators concluded that FadR regulation would be defective in V. cholerae and H. influenzae, whereas the pattern of regulation in Yesinia pestis would closely approximate that of E. coli. However, this analysis was based only on the predicted strengths of FadR binding sites and assumed that the various FadR proteins had identical operator binding and inducer binding properties (22). However, we show that this assumption is unwarranted; the FadR proteins we examined have strikingly different properties.
The two most similar proteins were the FadRs of E. coli and S. enterica, which was not surprising given the very close relationship between these organisms. However, despite the fact that these two proteins differed at only 7 of 239 residues (with several of the mismatches being conservative substitutions) expression of S. enterica FadR blocked growth of an E. coli fadR strain on oleate, whereas E. coli FadR when expressed at slightly higher protein levels, allowed growth, although growth was greatly slowed (Fig. 3). Because S. enterica FadR was no more efficient than E. coli FadR in repression of the fadBA promoter (TABLE SIX), it seemed likely that the difference in growth on oleate was due to weaker oleoyl-CoA binding by S. enterica FadR. That is, binding of fad operator sequences by S. enterica FadR was less efficiently neutralized by oleoyl-CoA than was E. coli FadR. Direct analysis by iso-thermal titration calorimetry confirmed this hypothesis; the K D of the S. enterica FadR for oleoyl-CoA was twice that of E. coli FadR (TABLE  FIVE). The residue changes that cause the less efficient binding of oleoyl-CoA by S. enterica FadR relative to E. coli FadR are not obvious because none of the residues that differ between the two proteins have been reported to contact myristoyl-CoA in the co-crystals with E. coli FadR (19). However, because oleoyl-CoA is four carbon atoms longer than myristoyl-CoA, it could contact residues that myristoyl-CoA cannot reach. This scenario seems unlikely to explain the different oleoyl-CoA binding, because although E. coli FadR-myristoyl-CoA structure has room to accommodate additional methylene groups (19), the residues that line this cavity are strictly conserved between the E. coli and S. enterica FadRs. It would be of interest to examine the response of the two FadRs to myristic acid supplementation in vivo to take better advantage of the FadR-myristoyl-CoA structure. However, the extremely poor solubility of long chain saturated fatty acids in bacterial media makes such experiments problematical (8). It should be noted that relative to the E. coli sequence there are almost as many nucleotide changes in the 5Ј-half of the S. enterica fadR as in the 3Ј-half (50 versus 58). However, only one base change in the 5Ј-half gives rise to a change of amino acid residue, whereas six such changes are found in the 3Ј-half indicating that markedly different selection processes have acted on the two halves of the genes. This should be viewed in light of the fact that the degrees of homology between the E. coli and S. enterica FadR proteins (97%) and the fadR coding sequences (83%) are typical of the values found for other homologous genes of the two organisms.
In contrast, the FadR proteins of P. multocida and H. influenzae were rather poor regulators of E. coli ␤-oxidation. They repressed the fadBA operator weakly and had low affinities for oleoyl-CoA. The poor operator binding might result from the N-terminal five-residue extensions that are not found in the proteins that give stronger repression. Residues 7-9 of E. coli FadR have been reported to make several stabilizing nonspecific contacts with the phosphate backbone of the operator (19), and the N-terminal extensions might hinder these interactions.
The "best" FadR from the viewpoint of the amplitude of the regulatory response was that of V. cholerae. The V. cholerae protein repressed as efficiently as the E. coli and S. enterica proteins, but failed to show the inhibition of growth on oleate imparted by the other strongly repressing proteins (Fig. 3). Moreover, oleate addition almost completely reversed repression of fadBA expression by V. cholerae FadR, whereas the E. coli and S. enterica proteins still exerted very appreciable repression (TABLE FOUR). For a transcriptional regulator to be an efficient regulatory switch, the protein must efficiently bind both the operator sequences of the cognate proteins and the regulatory ligand (acyl-CoA in the present case). However, efficient induction (derepression) requires that, upon binding of the small molecule ligand, the regulatory protein would lose most or all of its affinity for the operator sequences. At the concentrations of oleoyl-CoA accumulated in vivo 3 and the affinities of the V. cholerae, E. coli, and S. enterica FadR proteins it seems likely that each of the FadR proteins is present as the proteinligand complex in oleate grown cells. If so, then the repression remaining after oleate addition must be due to residual binding of the protein-ligand complex to the operator sites. If this scenario is valid, the greater dynamic range of the V. cholerae FadR must be the product of both the higher affinity of this protein for oleoyl-CoA and decreased affinity of the protein-ligand complex for the operator sites. The greater dynamic range of V. cholerae FadR seems very likely to result from the insertion (relative to the other proteins) of 40 residues into the midst of the primary sequence. Relative to E. coli FadR, the protein of known structure, alignments based on various criteria suggest that the insertion(s) occurred somewhere within residues 3 S. H. Iram and J. E. Cronan, unpublished data. 129 and 139. This is consistent with the surface location of this protein segment and its distance from the dimer interface. Because there are several non-conserved residues at the ends of the insertion, we cannot predict the site of insertion with confidence. However, based on the accessibility of the residues, the most likely points seem to be at the end of E. coli FadR helix 7 or within the coil region between helices 7 and 8. Upon threading on the E. coli FadR structures the PSIPRED program (bioinf.cs.ucl.ac.uk/psipred) predicts that the inserted segment would form a helix of 19 residues and might also significantly extend both helices 7 and 8. Helix 8 of E. coli FadR is known to interact with myristoyl-CoA (19) and together with neighboring helices is proposed to play a role in the disruption of the precise DNA binding architecture of the N-terminal domain of the dimer (21). It seems possible that the structure formed by the extra 40 residues of V. cholerae FadR might more thoroughly distort the structure or dynamics of the DNA binding determinants of the liganded protein. We attempted to remove the insertion but seem to have chosen endpoints incompatible with proper protein folding, because no expression of the deleted species was seen consistent with the observed lack of regulation (Fig. 3). Given the nearly ideal regulatory properties of V. cholerae FadR, it seems ironic that prior workers concluded that the fad genes of this organism are not FadR-regulated (22). Bioinformatics analyses of putative operator sequences found upstream of the V. cholerae fad genes led these workers to conclude that these sequences were not functional FadR binding sites. However, the sequence recognition rule derived in that work included bases that are now known not to be directly involved in operator recognition. If the consensus binding site derived from the DNA-protein contacts seen in the crystal structures (19,21) and in vitro selection (9) studies is used, the operators of the V. cholerae fadBA, fadE, and fadH genes match the consensus perfectly. As also seen in E. coli there are only weak matches to the putative operators of the fadD and fadL genes. This is expected, because these gene products are required for accumulation of the acyl-CoA inducer. Hence, stringent FadR regulation of these genes would prevent or greatly slow induction.
In contrast the FadR proteins of P. multocida and H. influenzae, particularly the latter, are poor regulatory proteins in that they weakly repress fadBA and bind oleoyl-CoA poorly. However, this seems of little physiological consequence, because the genomes of these organisms lost the ␤-oxidation cycle genes during the shrinking of the ancestral genome that accompanied adoption of the Pasteurellaceae commensal lifestyle (E. coli and H. influenzae are thought to have diverged about 680 million years ago). The only fad genes present in these small genomes are those involved in fatty acid uptake, fadD and fadL. It seems likely that FadR has been retained in these genomes because of its role as a transcriptional activator of the genes of unsaturated fatty acid synthesis, fabA and fabB. These genes have putative operators that are reasonable matches to the known fabA and fabB operators of E. coli. Indeed, because these organisms lack the generic FabF protein that catalyzes the elongation reaction of long chain fatty acid synthesis and have only FabB to catalyze these elongations (36), FadR activation of the fabB promoter may be more important than in E. coli.
The regulatory properties of the FadR proteins analyzed seem likely to reflect the ecological niches of the parent organisms. Pasteurellaceae such as P. multocida and H. influenzae grow only in commensal association with warm-blooded animals and are locked into this ecological niche (the upper respiratory tract) by their inability to synthesize several essential metabolites. It, therefore, seems possible that these organisms have dispensed with fatty acid degradation, because fatty acids are not sufficiently abundant in their environment to act as carbon sources and may use FadR mainly as a constitutive transcriptional activator. In contrast, E. coli and S. enterica colonize the digestive tracts of many organisms and in this environment should often be rich in fatty acids. How-ever, these organism also are found in waterways where fatty acids would be scarce, hence the advantage of inducible versus constitutive expression of the fad genes. In contrast V. cholerae is found in estuarine and marine environments that may or may not contain fatty acids. This organism does not colonize mammals, but rather causes a disease (cholera) that quickly clears the bacterium from the host. Hence, the diversity of habitats sampled by V. cholerae may have selected for a FadR of superior regulatory properties. The type of fatty acids available in various environments might also play a role. The marked diversity in sequence homology within the C-terminal halves of the different FadR proteins versus the strong sequence conservation seen in the N-terminal halves suggests that function of these proteins might be tuned to differing acyl-CoA species. Indeed, the differences in the values for the enthalpies and entropies of oleoyl-CoA binding strongly suggest that the mode of acyl-CoA binding by V. cholerae FadR may be significantly different from those of the E. coli and S. enterica FadR proteins.