Fatty acyl-CoA binding domain of the transcription factor FadR. Characterization by deletion, affinity labeling, and isothermal titration calorimetry.

The Escherichia coli transcription factor FadR regulates genes required for fatty acid biosynthesis and degradation in an opposing manner. It is acting as an activator of biosynthetic genes and a repressor of degradative genes. The DNA binding of FadR to regions within the promoters of responsive genes and operons is inhibited by long chain acyl-CoA thioesters but not free fatty acids or coenzyme A. The acyl-CoA binding domain of FadR was localized by affinity labeling of the full-length protein and an amino-terminal deletion derivative, FadRDelta1-167, with a palmitoyl-CoA analogue, 9-p-azidophenoxy[9-3H]nonanoic acid-CoA ester. Analysis of labeled peptides generated by tryptic digestion of the affinity-labeled proteins identified one peptide common to both the full-length protein and the deletion derivative. The amino-terminal sequence of the labeled peptide was SLALGFYHK, which corresponds to amino acids 187-195 in FadR. Isothermal titration calorimetry was used to estimate affinity of the wild-type full-length FadR, a His-tagged derivative, and FadRDelta1-167 for acyl-CoA. The binding was characterized by a large negative DeltaH0, -16 to -20 kcal mol-1. No binding was detected for the medium chain ligand C8-CoA. Full-length wild-type FadR and His6-FadR bound oleoyl-CoA and myristoyl-CoA with similar affinities, Kd of 45 and 63 nM and 68 and 59 nM, respectively. The Kd for palmitoyl-CoA binding was about 5-fold higher despite the fact that palmitoyl-CoA is 50-fold more efficient in inhibiting FadR binding to DNA than myristoyl-CoA. The results indicate that both acyl-CoA chain length and the presence of double bonds in the acyl chain affect FadR ligand binding.

been implicated as effectors of vesicular transport and fusion. A number of studies have been directed at elucidating a role for acyl-CoA compounds in the regulation of the activity of transcription factors. However, there has been only indirect evidence that acyl-CoA compounds regulate gene activity of eucaryotes at the level of transcription (2,3).
To date, the Escherichia coli FadR protein is the only transcription factor for which there is substantial and convincing evidence that direct binding of LCACoA to the protein prevents DNA binding, transcription activation, and repression (4 -8).
FadR is a 239-amino acid protein that regulates the transcription of many unlinked genes and operons encoding proteins required for fatty acid synthesis and degradation. Among the genes directly regulated by FadR are those encoding a specific membrane-associated fatty acid transport protein (FadL), acyl-CoA synthetase, all of the enzymes required for the ␤-oxidation of fatty acids, two enzymes essential for unsaturated fatty acid biosynthesis, and the repressor of the genes encoding the glyoxylate bypass genes, IclR.
The effect of FadR on the level of transcription is caused by its direct binding to DNA in the promoter regions of FadRresponsive genes (5,6). This binding in vitro is specifically prevented by long chain fatty acyl-CoA esters and not medium chain acyl-CoA esters or fatty acids (5). The protein binds DNA as a dimer (9). Our interests lie in understanding the mechanism by which acyl-CoA controls FadR activity. Previous genetic and biochemical analyses have identified amino acid residues in the carboxyl terminus of FadR that are likely to constitute in part the acyl-CoA (CoA presumably) ligand-binding pocket (8,9). In the present work, we have further localized the acyl-CoA binding domain by deletion and affinity labeling. Additionally, we have used isothermal titration microcalorimetry to assess acyl-CoA binding to the full-length protein, a His-tagged derivative and an amino-terminal deletion protein.
Together these studies contribute to our understanding of FadR-DNA and FadR-acyl-CoA interactions and help to further define the region of FadR involved in forming the acyl-CoA binding pocket.
Plasmid pCD129 was used for the production of full-length wild-type FadR as previously detailed (5). Plasmid pCD307-6 encoding His 6 -FadR was constructed as follows. An NdeI site was generated at the initiating methionine codon of the wild-type FadR gene by site directed mutagenesis of pCD152 to generate pCD306-6 using the Altered Sites System of Promega as described previously (8). An NdeI-BamHI fragment from pCD306-6 containing the complete coding sequence of FadR was cloned into the T7 RNA polymerase-responsive expression plasmid pET15b (Novagen) such that the coding sequence of FadR was fused in frame at the amino terminus to the His tag in the vector to generate pCD307-6 encoding His 6 -FadR. Plasmid pCD307-3 encoding FadR⌬1-167 was * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. constructed in a similar manner, but in this case an NcoI site was constructed in pCD152 at Met 168 in wild-type FadR to generate pCD306 -3. The NcoI-BamHI fragment encoding amino acid residues 168 -239 was subcloned from pCD306 -3 into the expression plasmid pET3a to generate pCD307-3. All constructions were verified by restriction enzyme analysis and DNA sequencing using promoter T7-and FadR-specific oligonucleotides as primers.
Protein Overexpression and Purification-Wild type FadR, its truncated mutant FadR⌬1-167, and full-length soluble His 6 -FadR were overexpressed in E. coli BL21(DE3)/pLysS harboring the appropriate plasmid using T7 polymerase expression system. Cultures (4 liters) were grown in an Amersham Pharmacia Biotech fermenter with automatic pH control at 37°C in the previously described medium (11) containing ampicillin (100 g/ml) and chloramphenicol (15 g/ml) until A 600 was about 4. T7 RNA polymerase was induced by the addition of isopropyl thio-␤-D-galactopyranoside to 0.4 mM. Growth continued for 3-4 h, and cells were harvested by centrifugation.
For isolation of full-length, native FadR encoded within pCD129, bacterial cells were disrupted in a French press in 300 ml of 20 mM Tris-HCl buffer (pH 8.0), 1 mM dithiothreitol, 1 mM EDTA (TDE buffer), and the homogenate was centrifuged at 70,000 ϫ g for 20 min. The FadR protein was subsequently recovered from inclusion bodies as follows. The pellet was dissolved in the same buffer containing 6 M urea, which was then removed by dialysis against TDE buffer containing 20% (v/v) glycerol. A part of the protein that precipitated during urea removal was redissolved and redialyzed. Soluble protein in TDE buffer containing 20% (v/v) glycerol was applied to a Q-Sepharose Fast Flow column preequilibrated with the same buffer, and the protein was eluted with a linear gradient of KCl (to 0.5 M). Fractions containing the protein of interest were pooled and dialyzed against 20 mM ammonium acetate buffer (pH 5.5) containing 1 mM dithiothreitol, 20% (v/v) glycerol. This was applied to an S-Sepharose high performance column, and the protein was eluted with a linear gradient of 1 M KCl in the same buffer at pH 6.5. FadR-containing fractions were pooled and dialyzed against 20 mM Tris-HCl buffer (pH 8.0), 20% (v/v) glycerol. Protein was concentrated by binding to a 5-ml Q-Sepharose High Trap column and eluting with a linear gradient of 0.4 M KCl. Final purification was achieved by gel filtration chromatography on a Sephacryl S-200 HR column developed with 20 mM Tris-HCl buffer (pH 8.0), 200 mM KCl containing 20% (v/v) glycerol. Protein was concentrated by ultrafiltration using a Centriprep 10 concentrator and stored at Ϫ20°C. Protein purity and identity was confirmed by electrospray mass spectrometry (ES-MS) and N-terminal sequencing.
For isolation of FadR⌬1-167 encoded in pCD307-3, bacterial cells were disrupted in a French press in 300 ml of TDE buffer, and the homogenate was separated by centrifugation as above. FadR⌬1-167 protein was recovered from inclusion bodies as follows. The pellet was dissolved in the same buffer containing 6 M urea and purified to apparent homogeneity by single step reverse-phase HPLC on a Nucleosil RP18 (230 ϫ 150-mm) column. Prior to injection, the sample was supplemented with trifluoroacetic acid to 0.1%. The column was equilibrated with 0.1% trifluoroacetic acid in water, and protein was eluted with a linear gradient of 96% ethanol containing 0.08% trifluoroacetic acid (0 -100% in 50 min). Fractions containing pure protein, as evidenced by the appearance of a single protein band upon 15% SDSpolyacrylamide gel electrophoresis, were freeze-dried. Lyophilized protein was redissolved in 20 mM ammonium acetate buffer (pH 6.0), 1 mM dithiothreitol and stored at Ϫ20°C. Protein purity and identity were further confirmed by ES-MS and N-terminal sequencing.
For isolation of His 6 -FadR encoded within pCD307-6, bacterial cells were disrupted in 200 ml of the binding buffer (50 mM potassium phosphate buffer (pH 8.0), 300 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol). Most of the target protein (ϳ90%) was in the soluble fraction after centrifugation. The crude protein preparation was incubated at 4°C for 2 h with 15 ml of a 50% slurry of nickel-nitrilotriacetic acid-agarose (Qiagen), which had been prewashed several times with the binding buffer. The slurry was then packed into a column and subsequently washed with 20 column volumes of the binding buffer to elute unbound proteins. The target protein was then eluted with 250 mM imidazole in the binding buffer. Due to a high binding capacity of nickel-nitrilotriacetic acid-agarose, eluted protein was so concentrated that some precipitation occurred but disappeared upon dilution. Imidazole was removed by dialysis against the binding buffer. Protein was essentially pure (initiating Met lost) as confirmed by HPLC, ES-MS, and N-terminal sequencing. The preparation was stored at Ϫ20 degrees C. Buffer exchange for isothermal titration calorimetry (ITC) experiments was later done using a 5-ml fast desalting High Trap column.
Functional Characterization of His 6 -FadR-The binding of purified His 6 -FadR to the fadB promoter in vitro was tested using protein-DNA gel shift assays. A 159-base pair fragment containing the fadB promoter was amplified by thermocycling using the oligonucleotides BFW (5Ј-TGATTTCTGCCGAGCGTG-3Ј) and BRE (5Ј-AGTCAAGGTACAGGGT-GTC-3Ј) as primers and pCD154 (5) as template. Reactions were cycled 35 times at 94°C for 1 min; 36°C for 1 min; and 72°C for 1 min. The fragment of interest was gel-purified and end-labeled using polynucleotide kinase and [␥-32 P]ATP (3000 Ci/mmol). Conditions for gel shifts were as described previously (5).
In vivo activity of His 6 -FadR was assessed by measuring ␤-galactosidase activities in transformants of strain LS1155 (fadR fadB-lacZ) to test repression and LS1348 (fadR fabA-lacZ) to test activation as described (6). The plasmids used to transform LS1155 and LS1348 to test FadR activity included pCD129, which encodes wild-type FadR under its native promoter as a positive control; pET15b, the His 6 vector, a negative control; pGP1-2, which encodes T7-RNA polymerase under the control of an isopropyl thio-␤-D-galactopyranoside-inducible promoter; and pCD307-6, which encodes His 6 -FadR under the control of a T7-RNA polymerase-responsive promoter.
Synthesis of 9-p-azidophenoxy [9-3 H]nonanoyl-CoA ([9-3 H]APNA-CoA)-Oleic acid (4 mmol) was dissolved in 40 ml of dioxane, and then 4 ml of 1 M NaOH and 3.2 ml of 2% OsO 4 were added. To this mixture, NaIO 4 (8 g) was added slowly over 5-10 min with stirring, which was continued for 18 h. The solvent was removed by evaporation, and the residue dissolved in water, made alkaline with 2 M Na 2 CO 3 , and extracted three times with petroleum spirits (40 -60°C). The aqueous phase was acidified with 4 M HCl and extracted three times with petroleum spirits (40 -60°C)/diethyl ether (1:1). The solvent was evaporated with a stream of nitrogen, and the product 9-oxononanoic acid was dissolved in petroleum spirits/diethyl ether (4:1) and purified on a silica gel column. The yield was 121 mg (21%). The 9-oxononanoic acid was dissolved in 8 ml of absolute ethanol titrated to pH 12 with 2 M NaOH and reduced with NaBH 4 in two steps. NaBH 4 (1.9 g) was dissolved in 1 ml of absolute ethanol and mixed with 750 mCi of [ 3 H]NaBH 4 (specific activity 20 -40 Ci/mmol, Amersham, UK) dissolved in 1 ml of ethanol. This solution was added slowly over 30 min to the 9-oxononanoic acid solution under continuous stirring. The stirring was continued for 90 min, an additional 5.9 mg of NaBH 4 was added, and stirring continued for another 3 h. The solvent was removed with a stream of nitrogen; 10 ml of water was added; the solution was acidified with HCl; and the 9-hydroxy[9-3 H]nonanoic acid was extracted with diethyl ether. The ether phase was dried over MgSO 4 for 6 h. The yield was 131 mg. The 9-hydroxy[9-3 H]nonanoic acid was methylated with diazomethane (12). The methylated product was dried carefully with benzene and dissolved in 1 ml of dry pyridine and placed on ice; 315 mg of dry tosylchloride was added, and the reaction was allowed to continue overnight with stirring. The reaction mixture was acidified with 2 ml of 1 M HCl and extracted three times with 10 ml of diethyl ether. The combined extracts were washed twice with 5 ml of 1 M HCl and dried with MgSO 4 ; the solvent was evaporated with nitrogen; and the product was dissolved in petroleum spirit and purified on a silica gel column. The yield was 132 mg of 9-tosyl[9-3 H]nonanoic acid methyl ester. All the following synthetic steps were carried out under dim light. p-Azidophenol (111 mg) was dissolved in 1 ml of ethanol, 44 mg of sodium methoxide was added, and then the solvent was evaporated in a vacuum centrifuge over night. The dry residue was dissolved in 1.5 ml of dry hexamethyl formamide and added to the dried 9-tosyl[9-3 H]nonanoic acid methyl ester. The reaction was allowed to run for 5 h at room temperature with stirring. The product 9-p-azidophenoxy[9-was incubated for 3-5 min on ice with [9-3 H]APNA-CoA (873 Ci/M) at a molar ratio of 1:1.25 in a final volume of 100 l in TBE buffer (pH 8.0) followed by UV illumination (300 nm) for 30 s. For competition experiments, a 5-fold excess of palmitoyl-CoA over photoaffinity ligand was included in the reaction. Ten microliters of the reaction mixture were subjected to 15% SDS-polyacrylamide gel electrophoresis. The gel was stained with Coomassie Blue, soaked in Amplify reagent, dried, and fluorographed.
For peptide mapping, radiolabeled protein was HPLC-purified from nonreacted [9-3 H]APNA-CoA on a Dynasphere column (8 ϫ 60 mm) with a 2-propanol linear gradient. This column material was found to provide superior recovery of both protein and very hydrophobic peptides. Label incorporation efficiency was estimated to be 15%. Photoaffinity-labeled protein was dissolved in 50 mM Tris (pH 7.4), 6 M urea and digested with trypsin (2%, w/w) after dilution to bring urea concentration to ϳ0.5 M. Peptides were resolved on a Dynasphere column preequilibrated with 0.1% trifluoroacetic acid and developed with a linear gradient of acetonitrile containing 0.08% trifluoroacetic acid (0 -70% in 40 min). Poorly resolved peaks containing most of the radioactivity were rechromatographed on the same column using a shallower gradient, and individual peptides were collected and sequenced.
ITC-Calorimetric measurements were carried out using an OMEGA titration microcalorimeter from MicroCal, Inc. (Northampton, MA). This instrument has been described in detail by Wiseman et al. (14). The reference cell was filled with water containing 0.02% sodium azide. The calorimeter was electrically calibrated at each temperature. All solutions used for the experiments were thoroughly degassed by stirring under vacuum. If necessary, protein solutions were spun for several minutes in a bench top centrifuge to remove any visible particles. The concentration of the protein was estimated spectrophotometrically at 280 nm using ⑀ ϭ 33060 M Ϫ1 cm Ϫ1 for FadR and His 6 -FadR and 12,210 M Ϫ1 cm Ϫ1 for FadR⌬1-167 and was approximately 0.03 mM. Protein under study in appropriate buffer was placed in the sample cell, and a ligand (dissolved in the same buffer as the protein) was drawn into the injection syringe, which was then mounted into a stepper motor for delivery into the sample cell. The syringe with stirrer paddle was rotated at 400 rpm during the experiment to assure immediate mixing. Experiments were performed at a temperature of 27 or 31°C. The concentration of the ligand, about 0.5 mM, was chosen to ensure full saturation well before final injection. Appropriate blank runs were conducted and subtracted from the corresponding data. The peaks of the thermograms obtained in this manner were integrated using the ORIGIN software supplied with the instrument. A nonlinear regression fitting to the isotherm was done using the CALREG (version 3.0) program (15). The fitting procedure yields the binding constant of the ligand K a , the heat of binding H, and the concentration of the binding sites (stoichiometry) N.
Mass Spectroscopy-Electrospray mass spectra of the proteins were recorded on a Vestec instrument (Vestec Corp., Houston, TX). Buffers and salts were removed from the purified proteins by HPLC. The sample in aqueous buffer was loaded onto a 60 ϫ 8-mm Dynasphere column and then stepwise eluted with a mixture of 90% acetonitrile, 10% water, and 0.08% trifluoroacetic acid. The sample was dried and brought to a final concentration of ϳ20 pmol/l with methanol/water (1:1, v/v) containing 1% acetic acid. The sample was introduced into the mass spectrometer by infusion with a syringe pump with a flow rate of 0.3 l/min. Spectra were acquired in the positive ion mode at 10 s/scan and mass window of m/z 600 -1500 using Teknivent Vector 2 data system. The molecular mass of the protein was calculated by weighted averaging as described by Mann et al. (16). The spectrometer was independently calibrated using myoglobin.
Determination of Amino Acid Sequences-Purified proteins and radiolabeled peptides were sequenced on a Knauer 910 pulsed liquid sequencer with chemicals and program as recommended by the manufacturer. Samples of 20 l of the amino acid phenylthiohydantoin derivatives were used for amino acid identification on a Knauer on-line HPLC 64 using a 250 ϫ 4-mm Lichrosphere 100 C-18 (5-mm particle size) column and a gradient of acetonitrile in 50 mM sodium acetate buffer, pH 5.2, as described by the manufacturer.
Materials-All protein purification was done on an FPLC system (Amersham Pharmacia Biotech) with columns supplied by the same manufacturer. HPLC separations were performed using the Kontron system with a gradient former equipped with a double wavelength detector (used at 216 and 280 nm). A Knauer 8 ϫ 60-mm column was packed with Dynasphere PD-102-RE monosized particles (from Dyna Particles, Lillestrøm, Norway). Nucleosil ODS, 10-m particle size, 30-m pore size, was from Machery Nagel (Duren, Germany). Propan-2-ol, acetonitrile (both HPLC grade), and trifluoroacetic acid (trifluoro-acetic acid, gas phase sequenator grade) were from Rathburn (Walkerburn, Scotland). L-1-tosylamido-2-phenylethyl chloromethyl ketonetreated trypsin (EC 3.4.21.4) was from Sigma. Fluorographic reagent Amplify was from Amersham. Other reagents were obtained from commercial sources and were of reagent grade. Water was of Milli-Q quality. All acyl-CoA esters were synthesized from corresponding fatty acids as described by Sanchez et al.(17)

RESULTS
Affinity Labeling of FadR and Fad⌬1-167-In previous work, we identified by random and site-directed mutagenesis a region in FadR including amino acids 216 -228 of the 239amino acid protein that was specifically required for acyl-CoA binding (8). This led us to hypothesize that the acyl-CoA binding domain was located toward the carboxyl terminus of the protein and that the acyl-CoA binding domain might be structurally and functionally separable from the amino-terminal DNA binding domain of FadR (9,18). Therefore, to further localize and analyze the acyl-CoA binding domain, we have constructed pCD307-3, which overexpresses a protein made up of amino acids 168 -239. Amino acid 168 was chosen, since it is an internal methionine that would make a convenient site for translational initiation (18). The protein encoded within pCD307-3, called FadR⌬1-167, has been purified to apparent homogeneity as evidenced by SDS-polyacrylamide gel electrophoresis and ES-MS (data not shown). Dimerization of a portion of the purified FadR⌬1-167 protein was observed when the protein was analyzed by ES-MS, and dimerization could be prevented by dithiothreitol treatment. Partial N-acetylation was also suggested by ES-MS.
Both full-length FadR and FadR⌬1-167 were photoaffinity labeled under identical conditions with [9-3 H]APNA-CoA. [9-3 H]APNA-CoA was previously shown to mimic palmitoyl-CoA in binding to acyl-CoA-binding protein, and the labeled peptides identified in that study were later shown by NMR to indeed be involved in acyl-CoA binding (19). In the present work, we were able to show that [9-3 H]APNA-CoA can photoaffinity-label FadR, and this cross-linking was prevented by an excess of palmitoyl-CoA (Fig. 1). No labeling of the protein was detected when it was incubated with [9-3 H]APNA-CoA without subsequent illumination. Similar labeling experiments in the presence of delipidated bovine serum albumin (which binds acyl-CoA with a K d of ϳ0.5 M) showed that FadR could be labeled in the presence of a 2-fold excess of bovine serum albumin and therefore that FadR competes effectively with bovine serum albumin for the [9-3 H]APNA-CoA (data not shown). The radiolabeled proteins were digested with trypsin in order to identify peptides cross-linked to [9-3 H]APNA-CoA. Since the affinity label is at the -end of the acyl chain, the amino acid residues within a labeled tryptic peptide are expected to be part of a hydrophobic pocket within FadR that binds long chain acyl-CoA. Additionally, some of the amino acid residues within the labeled peptide may make specific contacts with the acyl chain. A tryptic map of full-length FadR photo-cross-linked to [9-3 H]APNA-CoA is shown in Fig. 2. The peptides generated by digestion with trypsin were separated by HPLC on a Dynasphere column preequilibrated with 0.1% trifluoroacetic acid and developed with a linear gradient of acetonitrile with 0.08% trifluoroacetic acid (0 -70% in 40 min) (Fig. 3A). The major amount of radioactivity was eluted in five fractions. The first two (containing 68% of the radioactivity) were combined into pool 1, and the three following fractions (containing 20%) were combined into pool 2. The two pools were each rechromatographed using a shallower gradient, and individual peptides were collected. When pool 1 was rechromatographed, the radioactivity coeluted with a single absorbance peak (Fig. 3B). Mass spectrometry and sequence analysis of this peak showed that it contained three peptides, namely T19 (amino acids 187-195), T3 (amino acids 25-35), and T4 (amino acids 36 -45). When pool 2 was rechromatographed, the majority of the radioactivity was eluted in two fractions associated with a single peak (Fig. 3C). Mass spectrometry and sequence analysis on the peak containing radioactivity in Fig. 3C identified peptides T8 (amino acids 58 -70) and T23 (amino acids 229 -239). Total radioactivity recovered from the Dynasphere column was about 40%. FadR⌬1-167, like the wild-type protein, was photoaffinity-cross-linked to [9-3 H]APNA-CoA and digested with trypsin. The generated peptides were separated by HPLC (Fig. 4A). Upon rechromatography of the two fractions containing the major amount of radioactivity as appears in Fig. 4A, the radioactivity coeluted with a single peak (indicated by the arrow in Fig. 4B). The peptide in the fractions containing this peak was determined to be identical to T19 peptide from full-length FadR by mass spectrometry and amino acid sequencing.
Construction and Characterization of His 6 -FadR-Construction of plasmid pCD129, which allowed production, purification, and characterization of native, full-length FadR protein was initially developed to yield enough material to allow DNA binding studies (5). However, the amount of protein typically obtained from a 4-liter fermenter after a lengthy purification was about 5-6 mg despite good overexpression in E. coli. Mass spectroscopy data showed a loss of initiating methionine and partial removal of two or four C-terminal amino acids. This was not considered to be satisfactory for isothermal titration microcalorimetry, mainly because each ITC measurement required about 1.7 mg of FadR. This was also compounded by the fact that purified FadR partly precipitated during storage. Never-theless, the purified soluble FadR was sufficient for photoaffinity cross-linking to [9-3 H]APNA-CoA to determine a possible ligand-binding pocket as detailed above and a limited number of ITC experiments (see below).
To obviate the problems associated with full-length native FadR and to facilitate purification of large quantities of protein, we constructed an amino-terminal His-tagged derivative of the full-length protein called His 6 -FadR. The His 6 -FadR construct encoded in pCD307-6 was determined to be functional by comparison with native FadR in vivo with regard to (i) repression of the fadB gene by assaying ␤-galactosidase activities of a fadR fadB-lacZ strain, LS1155, and (ii) activation of the fabA gene by assaying ␤-galactosidase of a fadR fabA-lacZ strain, LS1348 (Table I). When T7RNA polymerase encoded within pGP1-2 was coexpressed with pCD307-6 encoding His 6 -FadR, ␤-galactosidase activities of LS1155 were very low and comparable with those seen for cells carrying pCD129 encoding wild type FadR. In contrast, LS1155 cells carrying pGP1-2 alone or together with the His tag expression vector pET15b had high levels of ␤-galactosidase, indicating that expression of fadB-lacZ was constitutive. Like the native protein, the His 6 -FadR also was inactivated by inclusion of oleate in the growth media resulting in elevated ␤-galactosidase activities. When ␤-galactosidase activities were evaluated in LS1348 transformed with the same set of plasmids, the opposite pattern was observed. ␤-Galactosidase activities in cells transformed with pGP1-2 and pCD307-6 were higher than the activities observed for cells carrying pGP1-2 alone. This indicated that His 6 -FadR was functional with regards to activation of expression of the fabA promoter. FadR-DNA binding in vitro of His 6 -FadR was assessed using electrophoretic mobility shift assays and was equivalent to purified native FadR as previously reported (Fig.  5) (5). Therefore, we concluded that His 6 -FadR was equivalent to native FadR in its ability to bind to DNA to repress or activate target promoters.
Affinity of Full-length and FadR⌬1-167 for Acyl-CoA Compounds Estimated by Isothermal Titration Microcalorimetry-We compared the binding affinities of full-length native FadR, His 6 -FadR, and FadR⌬1-167 for acyl-CoA using isothermal titration microcalorimetry, which allows simultaneous measurement of K d , enthalpy, and free energy of any heatevolving reaction and the stoichiometry of binding ( Fig. 6 and Tables II and III). The limited availability of the amount of native, full-length FadR precluded detailed microcalorimetry studies of its binding to a large number of different acyl-CoA esters. However, the quantity of native FadR was sufficient to perform a limited number of ITC binding studies (Table II). These experiments indicated that FadR has high affinity for myristoyl-CoA and oleoyl-CoA. The binding affinity of S-ether palmitoyl-CoA was not significantly different from that of natural palmitoyl-CoA ligand estimated for His 6 -FadR or FIG. 2. Tryptic peptide map of photo-cross-linked [9-3 H]APNA-CoA-FadR complex. Tryptic peptides were HPLC-separated, and peaks containing radioactive label were sequenced. Peptide sequences identified in radioactivity-containing peaks are underlined to show their position in FadR primary sequence. Methionine 168 is identified by an arrow. Amino acid substitutions in FadR that eliminated or reduced regulation of transcription in response to long chain fatty acids in vivo (also called superrepressors) are in boldface type (9,22). FadR⌬1-167. For each of the three FadR proteins studied, the stoichiometry of ligand binding for palmitoyl-CoA and oleoyl-CoA was determined to be approximately one acyl-CoA per FadR monomer.
The availability of large quantities of the His 6 -tagged FadR and FadR⌬1-167 made determination of binding and thermodynamic parameters to acyl-CoA esters by ITC possible (Tables  II and III). The results of acyl-CoA ester binding to either of these FadR derivatives showed clear preference toward long chain length compounds (C14-to C18-CoA) and no binding to the medium chain ligand (C8-CoA) as expected (i) from studies examining the chain length dependence of fad gene induction (20), (ii) from in vitro studies using protein-DNA electrophoretic mobility shift assays (5), and (iii) fluorescence analysis of FadR-acyl-CoA binding (8).
FadR⌬1-167 had about a 4-fold reduced affinity for myristoyl-CoA and oleoyl-CoA by comparison with the full-length protein (Table II). Estimates of FadR⌬1-167-palmitoyl-CoA binding were approximately equivalent to His 6 -FadR. However, the apparent affinity of FadR⌬1-167 followed the same pattern as the full-length native FadR and His 6 -FadR: no binding of C8-CoA and high affinity for long chain acyl-CoA. Importantly, all long chain ligands studied bound with high affinity to both His 6 fusion and truncated FadR, supporting the view that the acyl-CoA binding site of this protein is largely localized in its C-terminal part.

DISCUSSION
The Acyl-CoA Binding Domain of FadR-Two lines of evidence led us to suggest that FadR has a separable acyl-CoA binding domain located in the C-terminal portion of the protein. In the first set of experiments, we selected and characterized noninducible mutations in the FadR gene, which were required for binding of long chain acyl-CoA esters (8). Each of these mutations was localized to a single amino acid change. One altered protein carrying the change Ser 219 to Asn (FadRS219N) was purified and shown by DNA gel retardation assay to have a reduced affinity for oleoyl-CoA. FadRS219N retained the ability to bind DNA and to repress or activate transcription in the presence of oleoyl-CoA. Alanine substitution of amino acid residues 215-230 identified Gly 216 and Trp 223 as also required specifically for induction. This region of FadR shares amino acid identities and similarities with the Bars represent relative amount of radioactivity (in dpm) in 1-ml fractions collected by an automated fraction collector. Elution of radioactivity is slightly delayed due to the dead volume between the UV recorder and the outlet. The first two fractions and three following fractions containing most of the radioactivity (ϳ68 and 20%) were pooled separately and rechromatographed, and individual peaks were collected (B and C, respectively). The indicated peak (B) was sequenced and contained peptides T3, T4, and T19. Peptides identified in C were T8 and T23 (see also Fig. 1).

FIG. 4. Peptides identified in a tryptic digest of photoaffinitylabeled FadR⌬1-167.
A, the tryptic digest of FadR⌬1-167 photocross-linked with [9-3 H]APNA-CoA was separated on a Dynasphere column. Bars show the amount of radioactivity in 1-ml fractions collected as described in the legend to Fig. 3. B, the radioactivity-containing fractions from A were pooled and rechromatographed on the same column, and individual peaks were collected. The only radioactive peak (arrow) was sequenced and identified as T19.
coenzyme A binding site of Clostridium thermoaceticum CO dehydrogenase/acetyl-coenzyme A synthase (CODH) (21). The position of Trp 223 in the putative binding site of FadR is identical to the localization of Trp 418 in the CODH binding pocket. Trp 418 was protected from chemical modification by coenzyme A, thus identifying this Trp residue as part of the CoA binding site within the enzyme. Due to the alteration in acyl-CoA binding affinity of the purified FadRS219N protein, the noninducible phenotype of several FadR mutants carrying alanine substitutions in residues 215-230 and the sequence similarities of this region to the sequence of the CODH binding pocket, we proposed that this region of FadR forms part of the acyl-CoA binding domain and in particular is required for binding the CoA moiety. In a second series of experiments, we generated and analyzed protein fusions between FadR and the DNA binding domain of the E. coli lexA gene (9). One LexA-FadR fusion, LexA 1-87 -FadR 102-239 , retained the LexA DNA binding activity and was inducible by long chain fatty acids, indicating that the fusion contained the long chain acyl-CoA binding domain of FadR.
In an effort to further delineate the acyl-CoA binding domain to facilitate structural analyses, we constructed a deletion derivative of FadR, FadR⌬1-167, which removes the amino-terminal DNA binding domain but retains the ability to specifically bind long chain acyl-CoA (9,18). In the present study, we have demonstrated that purified FadR⌬1-167 has similar specificity for acyl-CoA as full-length native or His 6 -tagged FadR and can be affinity-labeled with [9-3 H]APNA-CoA. These data support our hypothesis that the determinants for acyl-CoA binding reside in the C terminus of the protein included within amino acids 168 -239.
Amino Acid Residues 187-195 Are Likely to Be in Contact with the Acyl-Chain of the Ligand-Both full-length FadR and FadR⌬1-167 could be photo-cross-linked to [9-3 H]APNA-CoA. The labeling reaction was blocked by an excess of palmitoyl-CoA (C16:0-CoA), indicating that FadR binds photoreactive ligand in a specific manner (Fig. 1). Proteolytic digestion of the radiolabeled full-length native protein followed by HPLC separation of the resulting peptides yielded a poorly resolved peak containing most of the eluted radioactivity (Fig. 3A). Rechromatography (Fig. 3B) of the pooled fractions from the major radioactivity peak in Fig. 3A identified three peptides, which were T19 (amino acids 187-195), T3 (amino acids 25-35), and T4 (amino acids 36 -45). Each of these peptides could conceivably be involved in formation of the fatty acid binding part of the ligand-binding pocket. However, we suggest that only T19 was specifically labeled for two reasons. First, when the deletion derivative FadR⌬1-167 was photolabeled, only one peptide identical to T19 was isolated from tryptic digestion of the cross-linked protein mutant (Fig. 4B). Second, in recent work, we have demonstrated that fusions between the DNA binding domain of the bacterial repressor LexA and amino acid resi- The host strains for these experiments were LS1155 or LS1348 transformed with the plasmids indicated. These included pGP1-2, encoding T7 RNA polymerase; pCD129, which encodes wild-type FadR under the control of the native FadR promoter; and pET15b, which is the vector used in the construction of pCD307-6. The last encodes His 6 -FadR under the control of the T7 RNA polymerase-responsive promoter.
b The growth medium was tryptone broth (TB) or tryptone broth containing 1 mM oleate (TBO).  dues 102-239 of FadR result in a protein that has the DNA binding specificity of LexA but is derepressed by long chain acyl-CoA, a function associated only with FadR (9). Rechromatography (Fig. 3C) of the second pool of fractions from Fig.  3A identified peptides T8 and T23. T8 includes amino acid residues within the DNA binding domain of FadR, and T23 consists of the C-terminal sequence of FadR. These peptides were labeled to a lesser extent than T19, and their contribution to acyl-CoA binding is unclear. We suggest that the T8 peptide may have been a coeluting contaminant in the poorly resolved peak after HPLC. The T23 peptide, in contrast, was retarded in the chromatogram from its theoretical calculated elution time and is therefore likely to be the labeled peptide in this mixture. These data suggest that in the native protein structure, this C-terminal peptide may be adjacent to the acyl-CoA binding pocket. Future detailed structural analysis will determine whether or not amino acids within T8 and T23 peptides actually contact acyl-CoA. One additional line of evidence that also suggests the acyl-CoA-binding domain that is localized in the C-terminal part of FadR has been obtained by treatment of the protein with vinylpiridine, which blocks free cysteine SH groups. Binding of palmitoyl-CoA to vinylpiridine-treated FadR⌬1-167, which contains single a cysteine residue corresponding to amino acid 200 in FadR (Cys 200 ) was abolished, as measured by ITC. One plausible explanation of this phenomenon is that either the single free thiol group is important for the ligand binding or simple steric hindrance around Cys 200 prevented the ligand from interacting with the protein. Using site-directed mutagenesis techniques, Cys 200 was substituted with either alanine or aspartate with no apparent affect on FadR activity or inducibility by long chain acyl-CoA as evaluated in vivo using a fadB-lacZ reporter (22). Therefore, the free thiol in Cys 200 does not appear to be making a critical contact with the ligand, and steric hindrance is the most probable cause of inhibition of palmitoyl-CoA binding by vinylpiridine.
Acyl-CoA Binding Affinity and Specificity Determined by Isothermal Titration Microcalorimetry-Titration microcalorimetry is a unique method for simultaneous determination of enthalpy of reaction, K d , and stoichiometry, from which free energy and entropy can then be calculated. The specific interactions that are involved in protein-ligand interactions include hydrogen bonding, electrostatic interactions, hydrophobic bonding, and proton ionization. Therefore, the computed parameters reflect a sum of those processes when more than one is contributing to the protein-ligand interaction. We observed in the experiments reported here that titration of either His 6 -FadR protein or FadR⌬1-167 with different acyl-CoA esters caused some precipitation, suggesting protein aggregation. Thus, calculated values should be taken as apparent that, despite all limitations, are nevertheless useful for comparing data obtained for related proteins.
Using ITS, we determined, as anticipated, that FadR, His 6 -FadR, and FadR⌬1-167 all exhibited high binding affinity for long chain fatty acyl-CoA esters, while no binding to medium chain ligand C8-CoA was detected (Tables II and III). FadR bound myristoyl-CoA and oleoyl-CoA with the same affinity as His 6 -FadR, indicating that the two proteins have similar binding affinity and that His 6 -FadR is suitable for binding studies. However, we obtained universally higher K d values for the truncated protein FadR⌬1-167 as compared with the fulllength proteins (Table I), 2 supporting the idea either that the native conformation of the truncated derivative differs somewhat from the full-length protein or that regions of the protein required for affinity are missing in the truncated protein. An exception from this universal observation is palmitoyl-CoA binding. The K d values obtained for palmitoyl-CoA binding were generally 5-6-fold higher than the K d values obtained from binding of myristoyl-CoA and oleoyl-CoA and 1.7-fold higher than binding of palmitoyl-CoA to FadR⌬1-167. There are some striking differences in specificity noted by comparison with previous estimates of FadR-acyl-CoA interaction (5,8). Specifically, the fact that C14-CoA binding to His 6 FadR is stronger than that of C16-CoA. This result was unexpected, since, when potencies of these two ligands to dissociate DNA-FadR complex are compared, palmitoyl-CoA is 50 times more effective than myristoyl-CoA (5). One explanation might be the fact that binding of the shorter acyl-CoA to FadR does not result in the allosteric transition required to prevent DNA binding. Alternatively, it is possible that affinity of FadR alone for any given acyl-CoA may differ from that of the FadR-DNA complex. The large increase in K d from myristoyl-CoA to palmitoyl-CoA followed by a similar drop in K d from palmitoyl-CoA to oleoyl-CoA indicates that the acyl chain length and/or double bonds may play a significant role in inducing conformational changes in FadR tertiary structure upon ligand binding. Our present model is that acyl-CoA binding prevents DNA binding; however, we cannot yet rule out the possibility that binding of acyl-CoA releases FadR from the DNA. The present ITC data may therefore hint that the latter mechanism is more probable. Future experiments will be required to clarify these interpretations.