INTRODUCTION

In all organisms lipid metabolism is tightly regulated to coordinate the cellular requirements for metabolic energy and to maintain the fluidity of cellular membranes. In Escherichia coli the transcription factor FadR plays a pivotal role in regulating the anabolic and catabolic pathways of fatty acid metabolism. FadR specifically represses the transcription of each of the genes that are essential for fatty acid transport and -oxidation including fadL, fadD, fadE, fadF, fadG, fadBA, and fadH (1-3). This transcription factor is required for long term stasis survival, partly through the regulation of expression of the uspA gene, which encodes a universal shock protein (4). FadR is also an activator of transcription of the fabA and fabB genes, which are required for unsaturated fatty acid biosynthesis (1, 3) and of the gene encoding the regulatory protein IclR, which in turn represses the expression of the aceBAK operon encoding the glyoxylate shunt enzymes (5). The growth of E. coli in minimal medium containing long chain fatty acids (C₁₄-C₁₈) results in derepression of the genes negatively controlled by FadR and in a decrease in the expression of the genes activated by FadR (2, 3). It has been demonstrated in vitro that FadR binds to DNA and that this interaction is specifically prevented by long chain fatty acyl-CoA (2, 6, 7).

Previously, we estimated the native molecular mass of FadR to be 29 kDa using gel filtration chromatography, suggesting that this protein is a monomer in solution, since the molecular mass predicted from the DNA sequence is 26,972 Da (2, 8). However, two lines of evidence argue that FadR interacts with its DNA binding sites as a multimer: (i) dyad symmetry in the operator regions of target genes, a feature commonly observed in transcriptional regulators that bind to their respective operators as homomultimers, (9) and (ii) the identification of a potential helix-turn-helix (H-T-H)¹ motif (between amino acid residues 34 and 54) in the amino-terminal region of FadR, a motif most commonly observed in homodimeric or homotetrameric prokaryotic DNA-binding proteins (8, 9). While many transcriptional regulators may be stable homomultimers in solution, for others dimerization in solution may be required for DNA binding or, in other cases, dimerization may occur on the DNA (10, 11). In the latter two situations, dimerization may be limiting for DNA binding.

In this work, we reassessed the native molecular weight in solution using purified FadR by nondenaturing polyacrylamide gel electrophoresis and glycerol gradient ultracentrifugation to determine whether dimerization may be limiting for DNA binding. The results of these experiments support that dimerization is not limiting for DNA binding. We also identified fadR alleles which exhibited a dominant negative phenotype (fadR^d). These altered alleles were expected to have mutations within the putative DNA binding domain that would decrease DNA binding but not multimerization (12-14). Four such alleles were identified after hydroxylamine mutagenesis of fadR, and the changes were localized to amino acid residues Ala⁹, Arg³⁵, His⁶⁵, and Gly⁶⁶. Furthermore, alanine scanning mutagenesis within and adjacent to the putative H-T-H motif identified five additional fadR^d mutations. The amino acid substitutions in these mutations were localized to Arg³⁵, Arg⁴⁹, His⁶⁵, Gly⁶⁶, and Lys⁶⁷, thus confirming the importance of the three amino acid residues Arg³⁵, His⁶⁵, and Gly⁶⁶ identified by hydroxylamine mutagenesis. To further distinguish functional domains of FadR, chimeric fusions between the amino-terminal DNA binding domain of LexA and different portions of FadR were constructed and analyzed. Our results indicate that a fragment of FadR including amino acid residues 102-239 contains determinants for dimerization and ligand binding.

EXPERIMENTAL PROCEDURES

The E. coli strains used were: JM109 for propagation of pSELECT-1 derived phagemids (15), LS1085 for plasmid isolations (3), LS1155 (fadR fadB-lacZ) to assess -galactosidase activity in response to plasmid encoded wild-type or mutant FadR (3), LS1154 (fadB-lacZ) to test negative transdominance (3), BL21(lDE3)/pLysS for controlled expression of FadR (16), and JL1436 to assess -galactosidase activity in response to plasmid encoded wild-type LexA or chimeric LexA-FadR fusions (17). For biochemical assays cells were routinely grown in Tryptone broth (18). Where necessary to maintain plasmids, antibiotics were added to a final concentration of 100 µg/ml ampicillin, 40 µg/ml kanamycin, or 12.5 µg/ml chloramphenicol. The minimal medium used was medium E (18). Fatty acids were provided at 5 mM in 0.5% Brij-58. Liquid cultures were grown at 37 °C with shaking in a New Brunswick gyratory incubator. Bacterial growth was monitored in a Klett Summerson colorimeter equipped with a blue filter.

Chemical mutagenesis of wild-type FadR encoded within pCD126 using hydroxylamine was described previously (6). The mutagenized DNA was transformed into LS1155, and transformants were tested for -galactosidase activity on Tryptone broth plates containing ampicillin and X-gal (40 µg/ml). Plasmids which conferred high levels of -galactosidase were isolated and transformed into LS1154 to test for negative transdominance.

Alanine substitutions in FadR were constructed using the Altered Sites System® (Promega) as described previously (6). The wild-type template was single-stranded pCD152. Mutagenic oligonucleotides (21-27 nucleotides in length) carried an alanine codon in place of the targeted native codon as the central residues. Plasmids encoding the mutation of interest were identified by sequencing. For phenotypic analysis of the mutants, a HinDIII-BamHI fragment was removed from the pCD152 derivative and subcloned to pACYC177 (19). The resulting plasmids were tested for FadR function in LS1155 and for the dominant negative phenotype in LS1154.

For the overexpression of wild-type and altered fadR proteins, the T7 system of Studier was used as previously detailed (2). Final cell extracts were maintained in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol containing 50% glycerol. Samples were stored in 100-µl aliquots at 80 °C (crude preparations). Purified FadR was prepared as described elsewhere (2). DNA binding affinity of wild-type or mutant FadR was assessed using the standard gel shift assay (2). For the characterization of binding to O_B, the 377-base pair HinDIII-EcoRI fragment from pCD154 containing the fadB promoter was used (2).

Antibodies were produced against purified FadR in an egg-laying hen by East Acre Biologicals (Southbridge, MA). Antisera was screened for FadR-specific antibody production using immunoblots (20). When specific, high titer antigen recognition was achieved, eggs were collected, and IgY was prepared. Partial purification of the antibody preparation from the egg yolks was achieved by a modification of Bade and Stegemann (21) as follows. Six yolks were separated from the egg white and rinsed with glass-distilled water, and the yolks were combined and scrambled in a 1-liter beaker using a Teflon-coated stir bar and magnetic stir plate on medium speed. Cold isopropyl alcohol (20 °C, 600 ml) was slowly added and the mixture was stirred on ice for 5 min. The precipitated protein was collected by centrifugation at 10,000 × g for 5 min at 4 °C. The isopropanol extraction was repeated two additional times followed by three extractions with 600 ml of cold acetone (20 °C). The final acetone precipitate was dried under a gentle stream of air and stored at 80 °C as a pellet or at 4 °C after resuspension in phosphate-buffered saline containing 0.01% sodium azide (10 ml/yolk).

Immunoblotting of cell extracts prepared as described above was performed essentially as described in Burnette (20). Following resolution on a 15% sodium dodecyl sulfate-polyacrylamide gel, proteins were electroeluted to Zetaprob (Bio-Rad). The blots were pretreated for 1 h with 5% calf serum in Tris-buffered saline (pH 7.5) and then for 1 h with anti-FadR or preimmune sera. The blots were washed and then treated with goat anti-chicken IgG conjugated to horseradish peroxidase (Capel Laboratories). Following a series of washes with Tris-buffered saline, the blots were developed using 3,3-diaminobenzidine and H₂O₂ as specified by the supplier.

The native molecular mass of FadR was evaluated using nondenaturing electrophoresis (23, 24). Standards used were carbonic anhydrase (29 kDa), chicken egg albumin (45 kDa), and bovine serum albumin (monomer, 66 kDa, and dimer, 132 kDa). Protein standards (10 µg) and purified FadR (10 µg) in 1 mM sodium phosphate, pH 7.0, and 50 mM NaCl were electrophoresed on a set of nondenaturing gels with polyacrylamide concentrations of 7.5, 10, or 12.5%. The distance of protein migration relative to the distance of migration of the tracking dye bromphenol blue (R_F value) was calculated from a scan made using an NIH image analysis system. A value of 100(log(R_F × 100)) for each protein was plotted against the percent concentration of acrylamide and the slope of the line (i.e. the retardation coefficient, K_R) was determined. The negative slopes obtained for each protein were then plotted against the native molecular masses of the standards to produce a linear log/log plot from which the molecular mass of FadR was extrapolated.

Glycerol density gradient sedimentation assay developed by Martin and Ames (25) was used to define the native molecular mass of FadR. Protein standards (100 µg) and pure FadR (25 µg) were layered on the top of a 5-25% glycerol gradient in 20 mM HEPES, pH 7.9, 100 mM KCl, 4 mM dithiothreitol, and 0.2 mM EDTA, which was spun at 32,000 rpm at 4 °C for 32 h in a SW41 Ti rotor in a Beckman L8-M ultracentrifuge. Following centrifugation fractions were collected under gravity flow. Quantification of the protein in each fraction was determined by Bradford (26) assay.

A 1-kilobase pair EcoRI-HinDIII fragment was subcloned from pJWL184 (22) to pSELECT-1 to generate pCD170. Using pCD170 as a template, lexA-specific oligonucleotides (19-24 bases in length) were synthesized to engineer either NdeI or NcoI sites at amino acid 88 to generate pCD171 and pCD172, respectively.

A 1.3-kilobase pair HinDIII-EcoRI fragment encoding FadR was subcloned from CD1901 (8) to pSELECT-1 to generate pCD152R. Using pCD152R as a template fadR-specific oligonucleotides (19-24 bases in length) were synthesized to engineer either NdeI sites at amino acid positions 81, 100, 120, or 167 to generate pCDR307-1, -4, -5, or -3, respectively, or an NcoI site at amino acid 140 to generate pCDR307-2. An EcoRI-NdeI fragment from pCD171 was then fused in frame with NdeI-BamHI fragments from pCDR307-1, -4, -5, -3, and ligated into pUC18 vector to generate pCD205 (LexA_1-87 HM FadR_83-239), pCD208 (LexA_1-87 HM FadR_102-239), pCD209 (LexA_1-87 HM FadR_122-239), and pCD207 (LexA_1-87 HM FadR_167-239), respectively. An EcoRI-NcoI fragment from pCD172 was fused in frame with NcoI-BamHI fragment from pCDR 307-2 and ligated into pUC18 vector to generate pCD206 (LexA_1-87 HM FadR_142-239). The resulting chimeric plasmids were sequenced across the fusion juncture to confirm the in-frame fusions. The expression of all the fusion proteins was driven off the isopropyl-1-thio--D-galactopyranoside-inducible promoter (lacUV5) (15).

-Galactosidase activities in strains LS1154, LS1155, or JL1436 harboring the plasmid of interest were assayed as described by Miller (18).

DNA was sequenced using Sequenase V 2.0 and a reagent kit purchased from U. S. Biochemical Corp. (Cleveland, OH).

Restriction enzymes, T4 DNA ligase, T4 DNA polymerase, and Sequenase V 2.0 were purchased from U.S. Biochemical Corp., -³⁵S-dATP was purchased from DuPont NEN. Fatty acids and molecular mass markers were obtained from Sigma.

RESULTS

In previous work the native molecular mass of FadR was estimated to be 29 kDa using gel filtration chromatography which is close to the calculated monomeric molecular weight 26,972 from the DNA sequence (8). However, we predicted the native form of FadR is a multimer because (i) the binding sites for FadR have dyad character (1), (ii) the protein is predicted to bind DNA through an H-T-H motif (8, 27), and (iii) as reported here, specific mutations confer a dominant negative phenotype. Since gel filtration chromatography can sometimes lead to aberrant estimates of native molecular (e.g. due to nonspecific interactions between the protein and the gel matrix) (28), we have reexamined the native molecular weight of FadR using two independent methods: nondenaturing polyacrylamide gel electrophoresis and glycerol gradient ultracentrifugation.

Analysis of purified FadR using nondenaturing polyacrylamide gel electrophoresis estimated the native molecular mass of this protein was 53.5 kDa (Fig. 1A). Using glycerol gradient ultracentrifugation, the native molecular mass was estimated to be 57.8 kDa (Fig. 1B). These data support the conclusion that FadR exists as a stable homodimer in solution. On the basis of these data and because of the dyad character of the binding sites, we propose FadR interacts with DNA as a homodimer.

[View Larger Version of this Image (14K GIF file)]

We performed hydroxylamine mutagenesis of wild-type fadR encoded within the moderate copy number vector, pACYC177 (6). A pool of mutagenized DNA was transformed into LS1155 (fadR fadB-lacZ) and selected on plates containing ampicillin and X-gal. Colonies that appeared dark blue had high levels of -galactosidase. Plasmid DNA was isolated from those colonies, and the fadR alleles encoded within the plasmids were then tested for the ability to confer a dominant negative phenotype over wild-type FadR in strain LS1154 (fadB-lacZ). Those mutant alleles induced by treatment of plasmid DNA with hydroxylamine, which resulted in at least a 3-fold increase in -galactosidase activities in LS1154, were designated fadR^d (Table I). Alleles that fulfilled this criterion had one of the following amino acid changes: A9V, R35C, H65Y, G66D, or G66S. Each of the substitutions were localized to the amino terminus of FadR; however, only one residue, Arg³⁵, resides within the putative H-T-H motif (8). To further test the importance of other amino acid residues within and adjacent to the predicted H-T-H motif, we constructed a series of fadR alleles carrying substitutions of alanine.

Table I. Transdominance of altered FadR proteins generated by hydroxylamine mutagenesis of wild-type FadR Plasmid Amino acid Codon  -Galactosidase activitya Negative dominanceb Native Substituted Native Substituted nmol min1 mg1 pACYC177 NRc 110  (24)d 1.0 pCD126 Wild type 101  (36) 0.9 pRW5 Ala9 Val GCG GTG 508  (83) 4.6 pRW9 Gly66 Asp GGC GAC 481  (50) 4.4 pRW12 Gly66 Ser GGC AGC 646  (65) 5.9 pRW17 His65 Tyr CAT TAT 410  (40) 3.7 pRW20 Arg35 Cys CGT TGT 607  (47) 5.5 a The results are the averages of three separate experiments each assayed in triplicate. b Dominance determined by dividing the value obtained for the strain carrying the plasmid encoded mutant allele by same strain carrying the vector pACYC177. c Not relevant, vector control. d Number in parentheses indicates standard error.

As above, each of these alleles substituted with alanine were tested for repressor function in strain LS1155 (fadR fadB-lacZ) (Fig. 2). The fadR alleles that showed constitutive levels of -galactosidase were subsequently tested for negative transdominance in strain LS1154 (Table II). Substitution of alanine for residue Arg³⁵, like the substitution with cysteine induced by hydroxylamine mutagenesis, resulted in a 5-fold dominance. Therefore we suggest that this amino acid is critical for DNA binding of FadR to O_B, the FadR binding site within the fadB promoter. Additional residues that appeared to be important for regulation of fadB included Arg⁴⁹, His⁶⁵, Gly⁶⁶, and Lys⁶⁷, each of which resulted in levels approximately 3-fold higher when substituted with alanine. The fadR alleles with alanine substitutions at Glu³⁴, Leu³⁷, Arg⁴⁵, Leu⁴⁸, Arg⁵⁴, Trp⁶⁰, Ile⁶³, and Thr⁶⁹ showed slightly elevated levels (1.4-2-fold) with respect to the control.

-Galactosidase activities in transformants of LS1155 (fadR fadB-lacZ).

[View Larger Version of this Image (25K GIF file)]

Table II. Negative dominance of FadR alleles carrying alanine substitutions to wild-type FadR in strain LS1154, fadB-lacZ Plasmid encoded FadR allelea  -Galactosidase activityb Negative dominancec hmol min-1 mg-1 None 190  (17)d 1.0 Wild type 135  (12) 0.7 E34A 327  (21) 1.7 R35A 962  (14) 5.1 L37A 135  (37) 0.7 V43A 238  (28) 1.2 R45A 270  (51) 1.4 L48A 395  (46) 2.1 R49A 575  (46) 3.0 R54A 300  (45) 1.6 W60A 312  (20) 1.6 I63A 273  (18) 1.4 H65A 589  (86) 3.1 G66A 506  (105) 2.7 K67A 524  (37) 2.8 T69A 256  (33) 1.3 a None indicates values obtained for transformants carrying the vector pACYC177. Wild type indicates transformants carrying pCD126. All other plasmids were pACYC177 derivatives carrying the designated amino acid substitution in fadR. b The results are the averages of three separate experiments each assayed in triplicate. Within each experiment a single colony after transformation of LS1155 was used as an innoculum to 10 ml of Tryptone broth. c Dominance determined by dividing the value obtained for transformants of plasmids encoding the relevant fadR allele by that obtained for cells transformed with the vector pACYC177 (None, above). d Number in parentheses indicates standard error.

To test whether or not the fadR^d isolates described above, including A9V, R35C, H65Y, G66D, and G66S, were able to bind to DNA, we subcloned the mutant genes into the T7 RNA polymerase expression plasmid, pT7-5, so that FadR was produced as the major protein after induction of T7 RNA polymerase. Fig. 3A represents a Coomassie Blue-stained 15% SDS-polyacrylamide gel of extracts of the cells after FadR induction. There was no discernible difference in the amount of protein produced in the mutants compared with the wild type. To verify that the major protein band was FadR, antibodies generated against purified FadR were used to probe an immunoblot of a duplicate gel (Fig. 3B). The specificity of the antibody preparation was demonstrated by reacting preimmune antisera or immune antisera against extracts from strain BL21(DE3) carrying pT7-5 (vector) or pCD129 (fadR⁺) (data not shown).

[View Larger Version of this Image (107K GIF file)]

The concentration of FadR in the crude extracts was estimated to be to be 3 × 10⁶ M by comparing the FadR band in the crude extract with a standard concentration of purified FadR on a Coomassie Blue-stained polyacrylamide gel (protein concentration was compared as band intensity using a Millipore BioImage analysis system). Each of the extracts was used to test binding to fadB DNA using the gel shift assay and a DNA fragment carrying O_B (2). A FadR-specific shift was observed for the crude extract enriched in wild-type FadR over a dilution range of 0.05-0.0005-fold (data not shown). This was estimated to be 2 × 10⁷ to 6 × 10¹¹ M FadR. The apparent K_eq for wild-type FadR in extracts of BL21(DE3) carrying pCD129 was 3 × 10¹⁰ M or essentially that which was determined for the purified protein (2). Fig. 4 is a representative gel shift experiment using extracts of cells harboring the vector pT7-5, pCD129 (wild type), or pCD20 (R35C). Under the same conditions we were not able to detect FadR-specific DNA binding activity to O_B for the fadR^d mutants A9V, R35C, H65Y, G66D, or G66S (data not shown).

[View Larger Version of this Image (39K GIF file)]

The bacterial repressor protein LexA binds to a site within the sulA promoter to repress transcription. Binding to DNA and subsequent repression of sulA is dependent upon homodimerization of the LexA protein (10). Chimeras between the LexA DNA-binding domain (amino acids 1-87) and segments of proteins such as GAL4, Jun, glucocorticoid hormone receptor, and AraC have been used successfully to identify and analyze functional heterologous dimerization domains (29-33). For the present study, amino acids 1-87 of LexA were fused in frame to FadR. None of the LexA-FadR chimeras were able to repress fadB-lacZ expression, since they were missing the predicted DNA binding domain of FadR (data not shown). In contrast, two chimeras (LexA_1-87FadR_83-239 and LexA_1-87FadR_102-239) were able to repress the LexA-responsive sulA-lacZ reporter (Table III). The chimera LexA_1-87FadR_83-239 repressed at levels comparable to wild-type LexA, while LexA_1-87FadR_102-239 had an intermediate effect. Three additional LexA-FadR chimeras which included amino acid residues 122-239, 142-239, and 167-239 of FadR did not repress sulA-lacZ, suggesting that either determinants critical for dimerization were missing or that the protein structure had been altered so that it was not capable of binding DNA.

Table III. Repression of sulA-lacZ by LexA-FadR chimeras in the presence and absence of fatty acid a Filled rectangles represent LexA and hatched rectangles represent FadR. The numbers above the filled rectangles indicate amino acid residues derived from LexA and the numbers below the hatched rectangles indicate amino acid residues from FadR. Transformants of JL1436 (lexA sulA-lacZ) carrying the plasmid encoded the chimeras were assayed for -galactosidase activity as described under "Experimental Procedures." b The values given are the averages of three experiments each assayed in triplicate. (SE) indicates standard error of the mean between experiments. c TB (Tryptone broth) indicates repressing media and TBO (Tryptone broth containing 5 mM oleate) derepressing media. d ND, not determined.

Previous studies found that amino acids in the carboxyl-terminal region of FadR were specifically required for acyl-CoA binding (6). Therefore we tested whether LexA_1-87FadR_83-239 and LexA_1-87FadR_102-239 were inducible by long-chain fatty acids (Table III). LexA_1-87FadR_102-239 was 4-fold inducible, indicating that it retained the ability to dimerize and bind long chain fatty acyl-CoA. LexA_1-87FadR_83-239 was not inducible, but acted like a super-repressor (6). This phenotype could result if the protein was unable to undergo the allosteric transition associated with long chain fatty acyl-CoA binding or if it was impaired in inducer binding.

DISCUSSION

Most prokaryotic transcriptional regulators which bind DNA through a helix-turn-helix motif do so as homodimers (e.g. CI and Cro) or homotetramers (e.g. TrpR) (9). The proteins may exist as stable multimers in solution at physiological concentrations. Alternatively, dimerization may be limiting for DNA binding or may occur on the DNA (10, 11). Previous work on purified FadR suggested that the protein was monomeric (2). To understand the function and mechanism of FadR-dependent gene regulation, it was important to investigate the possibility that dimerization was limiting for DNA binding. The results of native gel electrophoresis and glycerol gradient ultracentrifugation studies presented here clearly demonstrate that FadR is a stable dimer in solution. These data support the conclusion that dimerization is therefore not a limiting parameter for DNA binding. Our previous results suggesting FadR was a monomer in solution was based upon the behavior of this protein a gel filtration column. We suspect that the contradictory data result from anomalies of FadR migration when compared with molecular weight standards. Alternatively, FadR may interact with the gel matrix resulting in an anomalous elution pattern and an incorrect molecular weight determination.

FadR regulates a large number of genes required for fatty acid biosynthesis and degradation. FadR DNA binding is inhibited by long chain fatty acyl-CoA thioesters (2, 6, 7). In previous work we used random and site-directed mutagenesis to identify mutations in fadR that led to decreased ligand binding (6). The fadR alleles which showed decreased affinities toward long chain fatty acyl-CoA contained mutations that were mapped to the carboxyl terminus to amino acid residues including Gly²¹⁶, Glu²¹⁸, Ser²¹⁹, Trp²²³, and Lys²²⁸ (6). In an analogous manner, we used random and site-directed mutagenesis in the present studies to probe for mutations that affected DNA binding but not multimerization. FadR was predicted to bind DNA using an H-T-H motif formed by amino acid residues 34-53 (8). By comparison to the Cro family, amino acids in this region of FadR that fulfill the criterion to form this structure include: 1) Gly at position 42, which is critical to form the turn; 2) hydrophobic residues Leu⁷, Ile⁴¹, Val⁴³, and Leu⁴⁸; 3) hydrophilic residues Glu³⁴, Arg³⁵, Glu³⁶, Glu³⁹, Thr⁴, Arg⁴⁵, Thr⁴⁶, Thr⁴⁷, Arg⁴⁹, and Glu⁵⁰, which might provide DNA contacts; and 4) the absence of proline, which would disrupt domain architecture. FadR shares amino acid identities in this region with the GntR family subclass of H-T-H proteins (27). To test for amino acid residues that disrupt DNA binding but retain the ability to multimerize, we used the standard dominant negative test (34, 35). In general, those residues which are expected to be intolerant to substitution by alanine would be those making direct contact with the DNA and those which are required for maintaining the architecture of the protein exclusive of the determinants for multimerization (34, 35). Residues that are not critical for contacting DNA or to maintain the structure of the protein would result in a wild-type phenotype upon substitution of the native amino acid with alanine. Our results show that amino acid residues critical for DNA binding include two residues in the putative H-T-H domain, Arg³⁵ and Arg⁴⁹, as well as Ala⁹, His⁶⁵, Gly⁶⁶, and Lys⁶⁷. Yoshida et al. (36) reported the identification of four amino acid residues in Bacillus subtilis GntR that were required for maximal DNA binding and exhibited a dominant negative phenotype including Ser⁴³, Ala⁶⁶, Glu⁷⁴, and Arg⁷⁵. Two of these residues in GntR, Glu⁷⁴ and Arg⁷⁵, correspond exactly by position to His⁶⁵ and Gly⁶⁶ identified here for FadR based on the alignment of Haydon and Guest (27). Since mutations in both FadR and GntR were localized to homologous residues outside of the H-T-H motif, this extended DNA binding domain may constitute a DNA-binding structural motif characterisic of this family of proteins. Residues Ile⁴¹ through Thr⁴⁴ constitute the predicted turn of the H-T-H motif and therefore are expected to be critical to maintain the architecture of this portion of the DNA binding domain. This region of FadR proved to be relatively tolerant to substitution with alanine. Substitution of alanine for Leu³⁷ and Leu⁴⁸, in contrast, eliminated repression and were only weakly dominant negative. These residues, therefore, appear to be structurally important, and we further speculate that they may be necessary for maintaining the geometry of the two helices relative to one another as has been demonstrated to be the case for Ala³⁷ and Val⁴⁷ in the H-T-H region of CI (9).

Based on the in vivo -galactosidase activities in strains carrying the sulA-LacZ reporter fusions, it was concluded that two of five chimeras generated between the DNA binding domain of LexA and segments of FadR, LexA_1-87FadR_83-239 and LexA_1-87FadR_102-239, were functional and able to repress sulA-LacZ. Previous studies with FadR have demonstrated that amino acids 215-230 in the carboxyl-terminal region of FadR are specifically required for acyl-CoA binding (6). Therefore we expected that the chimeras LexA_1-87FadR_83-239 and LexA_1-87FadR_102-239 would be inducible by long chain fatty acids (Table III). LexA_1-87FadR_102-239 was inducible by long chain fatty acids indicating that it retains both the ability to dimerize and to bind acyl-CoA. However, LexA_1-87FadR_83-239 behaved like a super-repressor. The phenotype of this chimera may have resulted from: 1) a decreased affinity for inducer molecule, 2) an increased affinity for operator, or 3) an impaired ability to undergo the appropriate allosteric change required for induction. Since the determinants for DNA binding are attributable to the LexA segment of the chimera, increased affinity for DNA is not likely to be the cause for non-inducibility. Additionally, LexA_1-87FadR_83-239 is expected to retain the determinants for inducer binding present in LexA_1-87FadR_102-239. We therefore favor the view that the chimera cannot undergo the allosteric transition that must take place upon ligand binding which results in either the inhibition of, or release of, the protein from DNA binding.

FadR is a global regulator of lipid metabolism in E. coli, regulating more than 10 genes and operons. The ligands which modulate the activity of FadR are essential intermediates in fatty acid metabolism, long chain fatty acyl-CoAs. Therefore, the molecular mechanisms by which FadR controls gene expression and by which long chain fatty acyl-CoAs control FadR activity are of general interest. In the present work, we have determined that FadR functions as a homodimer and we have provided supportive evidence that the DNA binding domain lies in the amino terminus and that the determinants for dimerization lie within amino acid residues 102-122.