Crosstalk of Escherichia coli FadR with Global Regulators in Expression of Fatty Acid Transport Genes

Escherichia coli FadR plays two regulatory roles in fatty acid metabolism. FadR represses the fatty acid degradation (fad) system and activates the unsaturated fatty acid synthetic pathway. Cross-talk between E. coli FadR and the ArcA-ArcB oxygen-responsive two-component system was observed that resulted in diverse regulation of certain fad regulon β-oxidation genes. We have extended such analyses to the fadL and fadD genes, the protein products of which are required for long chain fatty acid transport and have also studied the role of a third global regulator, the CRP-cAMP complex. The promoters of both the fadL and fadD genes contain two experimentally validated FadR-binding sites plus binding sites for ArcA and CRP-cAMP. Despite the presence of dual binding sites FadR only modestly regulates expression of these genes, indicating that the number of binding sites does not determine regulatory strength. We report complementary in vitro and in vivo studies indicating that the CRP-cAMP complex directly activates expression of fadL and fadD as well as the β-oxidation gene, fadH. The physiological relevance of the fadL and fadD transcription data was validated by direct assays of long chain fatty acid transport.


Introduction
Much of our current knowledge of bacterial fatty acid metabolism comes from studies with Escherichia coli [1]. The fatty acid degradation (fad) pathway is primarily responsible for the transport, activation and b-oxidation of fatty acids [2]. The known long-chain fatty acid (LCFA) transport system components are the outer-membrane transport protein, FadL [3,4] and the innermembrane associated acyl-CoA synthetase, FadD [5,6]. In this system, the FadL transporter delivers exogenous long chain fatty acids across the cell membrane into the periplasmic space [3,4]. From there by an unknown mechanism the LCFA enter the cytosol where they become activated to their acyl-CoA thioesters by FadD which allows entry of the acyl chains into the b-oxidation cycle [5,6]. E. coli FadR coordinates the catabolic and anabolic fatty acid pathways (Fig. 1). In this dual role FadR acts as a repressor for the entire set of fad regulon genes and also functions as an activator for unsaturated fatty acid biosynthesis pathway by increasing transcription of both fabA [7,8] and fabB [9] (Fig. 1). The physiological ligands that antagonize FadR binding to its cognate operators are LCFA CoA thioesters synthesized by the cytosolic acyl-CoA synthetase, FadD [7,10]. In strains lacking either FadD or FadL addition of LCFA fails to induce fad regulon expression because the regulatory ligand cannot be synthesized. Therefore, extremely stringent regulation of fadD and fadL genes is precluded because this would block derepression.
Similar to most of the other fad regulon genes, fadL [4,11,12] and fadD [12,13] are controlled at the transcriptional level by two different regulatory systems in addition to FadR, the oxygen-sensitive ArcA-ArcB two-component system and the cyclic AMP (cAMP) receptor protein-cyclic AMP (CRP-cAMP) complex. Our interest in fadL and fadD arose from the presence of two FadRbinding sites in the promoters of both genes whereas the other FadR regulon fad genes each have only a single site. This raised the question of regulatory interactions among the multiple regulators, FadR, ArcA and CRP-cAMP when bound to the respective promoter regions. Cho and coworkers [12] initially used quantitative RT-PCR to assay regulation of fadL and fadD by FadR and ArcA under anaerobic conditions and reported that deletion of either arcA, fadR alone, or both arcA and fadR resulted in increased fadL expression by 34-, 4-or 69-fold, respectively, whereas fadD transcription increased 69-, 4.5-, and 84-fold. These workers did not investigate the effects of CRP-cAMP. Classical catabolite repression of the fad pathway has long been known. Pauli and coworkers [14] reported that Fad enzyme levels were very low in wild type cells grown in glucose even in the presence of inducing levels of fatty acid (or in constitutive fadR mutant strains) and that glucose inhibition could be partially relieved by addition of cAMP. CRP mutant strains also had low fad enzyme activities. In this laboratory Clark [15] using early lacZ fusion technology showed that glucose acted at the transcriptional level in fadBA and fadE expression. Although as discussed below CRP-cAMP binding sites for several fad genes have been proposed, none had been directly tested for the ability to bind the complex. Therefore, it remained possible that CRP-cAMP regulation of the fad regulon was an indirect effect.
We report experiments defining the interactions of these regulatory proteins and their effects on fatty acid transport.

Bacterial strains and growth conditions
All the strains used here are E. coli K-12 derivatives ( Table 1). The bacterial media were used for bacterial growth and analyses of b-galactosidase activity. These included LB medium (Luria-Bertani medium containing 10 g of tryptone, 5 g of yeast extract and 10 g of NaCl per liter), Rich broth (RB medium containing 10 g of tryptone, 1 g of yeast extract, and 5 g of NaCl per liter) and minimal medium M9 [16] supplemented with 0.4% glucose or other carbon source, 0.1% Vitamin-Free Casamino Acids, 0.1 mM CaCl 2 , and 0.001% thiamine. When necessary, antibiotics were used as follows (in mg/liter): sodium ampicillin, 100; kanamycin sulfate, 50; chloramphenicol, 20 and tetracycline HCl, 15. To monitor transcriptional regulation by ArcA-ArcB, an anaerobic environment (5% H 2 , 75% N 2 , and 20% CO 2 ) was generated in an anaerobic environmental system (Bio-Bag environmental chamber type A; Becton Dickinson). The E. coli strains were grown on modified LB media in which potassium nitrate (5 mM) was added as an electron acceptor [17]. Fatty acids such as oleic acid (Sigma) were neutralized with KOH, solubilized with Tergitol NP-40, and used as an inducer at 5 mM final concentration.

Plasmids and genetic manipulations
The pCR2.1-TOPO vector (Invitrogen) was used for PCR cloning and sequencing whereas expression vector pET28a (Novagen) was used for protein preparation ( Table 2). All plasmids constructed were validated by PCR analyses plus direct DNA sequencing.
Both FYJ159 and FYJ295 are, respectively, derivatives of E. coli strain MC4100, in which either a fadD-lacZ transcriptional fusion [18] or a fadL-lacZ transcriptional fusion, was integrated into the chromosomal attl site (Table 1). Strain FYJ104 which carries a chromosomal fadL-lacZ transcription fusion (Table 1) was constructed using the FLP-mediated site-specific recombination method [19]. The kanamycin resistance cassette was removed from the fadL::km strain JW2341-1 ( Table 1) by expression of the FLP recombinase encoded by plasmid pCP20 (Table 1) to give strain FYJ103, which retained a single FLP recombinase target (FRT) site. The FRT site was used for site-specific integration of the lacZ fusion plasmid, pKG37 (an improved version of pCE71) containing a FRT site upstream of a promoterless lacZY genes, a kanamycin resistance gene, and the R6K origin of replication (Table 1) [19]. The transformants were screened on LB agar plates containing kanamycin and 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-Gal) at 37uC to obtain the chromosomal fadL-lacZ fusion strain FYJ104. The fusion plasmid was stably integrated due to its R6K origin and loss of the temperature-sensitive pCP20 plasmid [19]. The fadL promoter-lacZ junction was validated by a PCR using primers, fadL-P plus lacZ-R (

b-Galactosidase assays
Mid-log phase cultures in LB, RB or minimal media (with or without supplementation with various carbon sources), were collected by centrifugation, washed twice with Z Buffer [21] and assayed for b-galactosidase activity after lysis with sodium dodecyl sulfate-chloroform [21]. The data were recorded in triplicate with no less than three independent experiments.

Protein expression and purification
Hexahistidine-tagged E. coli FadR (and/or CRP) proteins were produced in E. coli BL21 (DE3) carrying the expression plasmid pET28-fadRec (and/or pET28-crp) ( Table 1) by induction of bacterial cultures at an OD 600 nm of 0.8-1.0 with 0.3 mM IPTG at 30uC for 3 h [17,22]. The cells were pelleted washed twice with ice cold PBS buffer (101.4 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, 8% glycerol, pH 7.4), dissolved in the same buffer and lysed using a French pressure cell. The extracts were centrifuged to remove bacterial debris and the supernatants loaded onto a nickel chelate column (Qiagen). Following washing with ten column volumes of with PBS buffer containing 50 mM imidazole, the FadR proteins were eluted with 150 mM imidazole. Appropriate eluted protein fractions were pooled and dialyzed against PBS buffer then concentrated by ultrafiltration (30 kDa cut-off, Amicon Ultra) [17]. The protein purity was judged by 12% SDS-PAGE, followed by staining with Coomassie brilliant blue R250 (Sigma, St. Louis, MO). Both FadR and CRP proteins were characterized by liquid chromatography quadrupole time-of-flight mass spectrometry of tryptic peptides and chemical cross-linking as described previously [18]. In the absence of a long chain acyl-CoA E. coli FadR represses the fad regulon genes [8,23] whereas it activates transcription of fabA and fabB, the two genes of the unsaturated fatty acid synthetic pathway [7][8][9]. B.     Electrophoretic mobility shift assays These assays of the interaction between the fadD and fadL promoters, FadR and the cAMP-CRP complex were done essentially as previously reported [18,23]. All of the FadR (and/ or CRP) probes were prepared by annealing two complementary primers ( Table 2) by incubation in TEN buffer (10 mM Tris-HCl, 1 mM EDTA, 100 mM NaCl, pH 8.0) at 95uC for 5 min followed by slow cooling to 25uC and then DIG labeling by terminal transferase with DIG-ddUTP (Roche). DNA probes ( Table 2) for assay of DNA binding by the CRP-cAMP complex were similarly synthesized. The digoxigenin-labeled DNA probes (either 0.1 or 0.2 pmol) were incubated with either DNA binding protein in binding buffer (Roche) for 15 min at room temperature and then analyzed by native PAGE (6.5% PAGE for the CRP probes and 7% PAGE for all other probes). The separations were then visualized as previously described [18,23].

Fatty acid transport assays
Fatty acid transport was assayed as described by Klein et al. [24] with minor modifications. To avoid complications by b-oxidation strains that carried a fadBA disruption (strains FYJ82, FYJ83 and SI203) were used (Table 1). To test ArcA-P regulation of fatty acid transport, FYJ82 strain (DarcA) was compared with the wild type strain SI203 (Table 1). Overnight cultures were inoculated into 10 ml of RB liquid media supplemented with potassium nitrate (5 mM) as electron acceptor and kept in a fully anaerobic chamber at 37uC for ,10 hrs [17]. The anaerobic environment (5% H 2 , 75% N 2 , and 20% CO 2 ) was generated by an anaerobic environmental chamber (Bio-Bag type A, Becton Dickinson) [17]. Cultures in exponential phase were treated with 100 mM chloramphenicol for 10 min prior to assay. 1-14 C-Oleic acid (American Radio-labeled Chemicals) was injected into the anaerobic bacterial cultures using a syringe fitted with a fine The cultures were mixed well by vigorous vortex mixing and incubated at room temperature for about 15 min with anaerobiosis monitored by an anaerobic indicator (0.001% resazurin). Finally, the bacteria were collected by centrifugation (4.2006 g, 16 min), and washed for five times with iced RB medium. One ml of the cultures were subjected to membrane phospholipid extraction [18,25] after measuring culture absorbance (A 600 ) and adjusted to an absorbance of 1.5. The phospholipids were then separated from any residual fatty acid by thin layer chromatography [18,25]. Assay of the effects of the CRP-cAMP system on logarithmic phase cultures aerobically grown in LB liquid proceeded in a similar manner.

Bioinformatic analyses
The known or predicted DNA binding sites recognized by either FadR or ArcA (or CRP) were all from the E. coli literature. Multiple alignments were done using ClustalW2 (http://www.ebi. ac.uk/Tools/clustalw2/index.html), and the resultant output was processed by program ESPript 2.2 (http://espript.ibcp.fr/ ESPript/cgi-bin/ESPript.cgi), generating the final BLAST version.

Results and Discussion
Regulatory complexity in the LCFA transport promoters, fadL and fadD The fadL and fadD genes encode the proteins known to be required for LCFA transport (Fig. 2A). The fadL and fadD transcriptional start sites are located 95 bp and 60 bp upstream of the translation start sites, respectively [6,26] (Fig. 2B). Despite a seemingly straightforward role in metabolism, both promoters contain demonstrated or annotated binding sites for three different transcription factors, FadR, ArcA and CRP (http://www.ecocyc. org). Moreover, fadL and fadD promoters are the only fad regulon promoters that contain two distinct FadR binding sites [6,26]. In the fadL promoter the FadR sites are separated by only 8 bp whereas in the fadD promoter the two FadR sites are separated by 68 bp (Fig. 2B and C). The locations of the ArcA binding sites determined by Cho and coworkers [12] also differ. The fadL ArcA site is 5 bp from FadR site 1 whereas the fadD ArcA site overlaps FadR site 2 by one bp, raising the possibility of crosstalk between the two repressor proteins (Fig. 2). The fadL CRP site was reported to lie downstream of the transcription start site [4] (Fig. 2B), a position incompatible with the usual activator function of CRP. Searches using the CRP consensus sequence of Zheng et al. [27] produced another fadL candidate binding site upstream of the transcription start site. Although this seemed a much more plausible position for activation of transcription, experimental verification of the CRP binding site was required. Indeed, although CRP-cAMP regulation of the fad regulon genes at the physiological level has been known for many years [14], none of the proposed sites have been experimentally validated.

The two FadR-binding sites of fadL and fadD result in only modest repression
To obtain a parallel comparison of the FadR binding sites of the two promoters we used electrophoretic mobiltity shift analyses (EMSA) to assay the binding abilities of the individual sites over a range of FadR concentrations (Fig. 3). The two fadL FadR sites bound FadR with equivalent affinities (essentially complete binding of the probes by 50 nM FadR, Fig. 3B and D). In contrast, although fadD site 1 showed a binding affinity comparable to that of the fadL sites (Fig. 3F), FadR binding by fadD site 2 (Fig. 3H) was .10-fold weaker than the other three sites (Fig. 3B, D and F). Addition of 25-50 mM oleoyl-CoA resulted in loss of FadR binding by all four sites (Fig. 3C, E, G and I). Therefore, our data are in good agreement with the reported DNase I footprinting results [6,11] and provide data on the effects of the acyl-CoA regulatory ligand that is lacking in the foot-printing experiments. To monitor expression of fadL and fadD, each promoter was fused to a LacZ reporter gene to allow expression to be assayed by b-galactosidase activity (Fig. 4). Deletion of FadR resulted in only modest derepression of the two LCFA transport system genes (2 to 2.5 fold for fadL and 2 to 3-fold for fadD) in  medium with acetate as sole carbon source ( Fig. 4A and B) whereas expression of fadBA and fadH increased by 5 to10-fold ( Fig. 4C and D). Increased expression of genes fadL and fadD in the DfadR strain was also seen when the carbon source was either glucose or glycerol ( Fig. 4A and B), These observations are generally consistent with those obtained upon oleate induction of strains carrying a functional FadR (Fig. 4E).

Expression of fadL and fadD is directly activated by the CRP-cAMP complex
The global regulator cAMP-CRP complex [28] can act as either an activator [29] or a repressor [30] in expression of genes involved in many E. coli metabolic pathways. Although putative class I cAMP-CRP binding sites had been proposed upstream of some fad regulon genes and activation of fadH transcription was observed in vivo [17], direct physical evidence for DNA binding of these promoters by the cAMP-CRP complex was lacking. We therefore used EMSAs with purified CRP in the presence or absence of cAMP to test the proposed sites. SDS-PAGE analyses indicated that the apo-form of CRP of molecular weight ,23 kDa was homogenous (Fig. 5A). Chemical cross-linking showed the protein was dimeric as previously reported [28] (Fig. 5B). Liquid chromatography mass spectrometry analysis of tryptic peptides demonstrated that our recombinant protein matched E. coli CRP with 66% coverage (Fig. 5C).
Higashitani and coworkers [4] predicted a fadL CRP binding site (Fig. 2B) centered at 57 bp downstream of the fadL transcription start site (termed fadL2) that is covered by our fadL probe 2, Fig. 6C). In contrast, we favored a site centered 157 bp upstream ( Fig. 2C and 6A) covered by fadL probe 1 (Fig. 6B). In the fadH promoter, we previously proposed a CRP binding site (called fadH site 1 covered by fadH probe 1, Fig. 6A and E) and subsequently have predicted a second site (called fadH2 and covered by fadH probe 2, Fig. 6A and F). Given that sequence alignments of these sites showed only six conserved bp (Fig. 6A), the function of these sites required direct testing by EMSAs. As expected from numerous prior investigations (e.g., Lawson et al. [31]), CRP lacked DNA binding activity in the absence of its cAMP ligand (Fig. 6). Gel shift assays confirmed that the CRP-cAMP complex efficiently bound fadL probe 1 (Fig. 6B), but failed to bind fadL probe 2 (Fig. 6C). Therefore, fadL site 1 is a functional CRP-cAMP complex binding site whereas the site predicted by Higashitani and coworkers [4] is nonfunctional. Unlike Class II sites which overlap the 235 hexamer and Class III sites which have tandem CRP binding sites, the position of fadL site 1 identifies it as an atypical site that is located too far upstream to be reached the C-terminal domain (CTD) of the a-subunit of RNA polymerase which seems to have maximum stretch of 90-120 bp upstream of the promoter. In contrast the demonstrated fadD CRP-cAMP complex binding site [27] can be clearly identified as a Class I site (Fig. 2C and 6D). Both proposed fadH promoter sites bound the CRP-cAMP complex and therefore comprise a bona fide Class III site of the type that is composed of two Class I sites (some Class III sites have a Class II site coupled to a Class I site) ( Fig. 6E and F).
To evaluate the effects of CRP on expression of the fatty acid transport genes, in vivo assays of b-galactosidase activities of the fadD and fadL lacZ fusions in Dcrp mutant strains were compared to strains expressing wild type CRP (Fig. 7). As a control we also assayed fadH expression and found that expression decreased in the absence of CRP (Fig. 6E and F) (Fig. 7D). As expected, the activities of both transport promoters were modestly decreased upon the loss of crp (Fig. 7A and B) verifying that CRP activates fadL and fadD transcription. Although in agreement with others (http://www.ecocyc.org), we failed to identify any CRP-binding site candidates in the fadBA promoter, expression of fadBA in the crp deletion strain is decreased relative to the wild type strain (Fig. 7C). This may indicate an indirect activation of the fadBA promoter by the CRP-cAMP complex. It seems possible that the indirect activation is due to action of the FIS protein, expression of which is activated by CRP binding to three discontinuous sites [27]. Indeed, a putative FIS binding site (TTGCA-TATTTTTAACACAA, 236 to 18) lies within the fadBA promoter and FIS is a known activator of fadBA transcription [32].

Expression of fadL and fadD are additively repressed by FadR and ArcA under anaerobic conditions
The weak effects of FadR inactivation on expression of fadL and fadD suggested that other regulatory proteins may be involved and thus we tested the effects of deletion of the genes encoding the other proteins proposed to bind these promoters. The oxygensensitive two-component system ArcA-ArcB was reported to negatively regulate transcription of several fad regulon genes, including fadL, fadD, fadBA and fadH [12]. That report [12] also contained in vitro and in vivo evidence that the phosphorylated ArcA regulator (ArcA-P) bound the promoters of the target genes (Fig. 8A). We recently reported cross-talk between FadR and ArcA in expression of other fad regulon genes and found that the interaction could be either additive (fadH) (Fig. 8E) or synergistic (fadBA) (Fig. 8D) [17]. The fadD promoter FadR site 2 and the ArcA site overlap by a single bp (Fig. 2C) as also seen in fadH [17] whereas in the fadL promoter the ArcA and FadR site 1 sequences are separated by 5 bp (Fig. 2B). Thus, cross-talk between FadR and ArcA in fadL and/or fadD expression seemed likely. To test for cross-talk strains carrying arcA and/or fadR null mutations plus the chromosomal fadL-lacZ (or fadD-lacZ) transcriptional fusions were constructed. In general agreement with the report of Cho et al. [12], we found that the absence of arcA under anaerobic conditions resulted in increased transcription of fadL and fadD by 2 to 3-fold and .20-fold, respectively, ( Fig. 8B and C). Our levels of derepression are somewhat less than those reported by Cho and coworkers [12] which we attribute to the gene used by these workers as their internal reference [17,33]. However, our data differed from those of Cho and coworkers in the effects reported on FadR regulation of fadL transcription under anaerobic conditions when ArcA is functional (we see no effect versus their reported 4-fold increase). Although our statistical analysis argued that these two regulatory proteins control transcription of fadL and fadD in an additive manner, the wide variations in the data and the overlapping error bars provide caveats to this interpretation.
In the absence of the cAMP receptor protein (CRP) complex (cAMP-CRP), FadR and ArcA-ArcB repression of fadL and fadD expression was relatively weak (Fig. 8) and similar to that seen with fadBA and fadH [17]. Thus it seems that regulation by both FadR and ArcA-ArcB rely on the cAMP-CRP complex to activate transcription such that it can be further modulated.
Physiological relevance of ArcA and the CRP-cAMP complex to fatty acid transport Uptake of [1-14 C]oleic acid was assayed to evaluate the physiological consequences of ArcA-ArcB and the CRP-cAMP complex on fatty acid transport (Fig. 9). To rule out potential interference by b-oxidation, the strains studied lacked fadBA ( Table 1). Incorporation of [1-14 C]oleic acid into the membrane phospholipids of the DarcA strain (FYJ82) was consistently .2fold greater that that of its parental strain under anaerobic conditions (Fig. 9A). Further analyses by thin layer chromatography showed that incorporation into the three major membrane phospholipids phosphatidylethanolamine (PE), phosphatidylglycerol (PG) and cardiolipin (CL) (Fig. 9B), consistent with anaerobic repression of fadL (and/or fadD) by ArcA-P ( Fig. 8B and C). In agreement with the observed CRP-cAMP activation of fadL and fadD transcription ( Fig. 7A and B), quantitative determination of total 14 C-labeled bacterial membrane phospholipids showed that the level of oleic acid incorporation in the strain lacking crp is only about one third that of the parental strain ( Fig. 9C and D). To our knowledge these are the first direct physiological data that directly show that these two global regulators modulate LCFA transport pathway in E. coli.

Conclusions
Transcriptional regulation of fatty acid transport in E. coli involves three distinct regulatory systems, the specialized FadR system and the two global systems, ArcA and CRP-cAMP. FadR action is straightforward, it acts as a classical LacI-type repressor and only weakly represses fadD and fadL expression. CRP-cAMP regulation is also straightforward, fatty acids are a low status carbon source and E. coli prefers to use the highest status carbon source, glucose. In the presence of glucose (or in the absence of CRP), the other two regulators have little transcription to regulate. The apparent surprise is the stringent repression by ArcA seen under anaerobic conditions even in the absence of FadR because E. coli has a pathway to degrade fatty acids under anaerobic conditions [34] and low expression of FadD and FadL would seem likely to compromise function of the anaerobic pathway. However, under anaerobic conditions a new acyl-CoA synthetase, FadK, is induced that replaces FadD [35]. Unlike FadD which is inactive with short chain fatty acids [24,35], FadK strongly prefers short chain length acids and such acids are the preferred growth substrates of the anaerobic b-oxidation pathway [34] although LCFA can also be utilized. Short chain acids readily enter fadL strains [36] and thus decreased expression of FadL is irrelevant for these growth substrates. However, degradation of LCFA such as oleate via the anaerobic pathway should require FadL. These apparent contradictions can be reconciled by our finding that ArcA represses fadL expression less than fadD expression (only about 4-fold) ( Fig. 8B and C) and that FadK activates oleate poorly [35]. Hence, the significant level of FadL expressed under ArcA repression probably provides sufficient oleate transport to satisfy the poor catalytic activity of FadK with this substrate and allow the observed slow anaerobic growth on oleate [35].   DfadR), respectively. The strains were grown on LB agar plates supplemented with 5 mM nitrate (KNO 3 ) as the electron receptor. Anaerobic environments were generated using Bio-Bag environmental chamber type A as described [17]. beta-Galactosidase activities were recorded from at least six independent assays and are expressed as means 6 standard deviations. *, P,0.005; **, P,0.001. doi:10.1371/journal.pone.0046275.g008