Unsaturated Fatty Acid Synthesis in Bacteria: Mechanisms and Regulation of Canonical and Remarkably Noncanonical Pathways

Abstract

Unsaturated phospholipid acyl chains are required for membrane function in most bacteria. The double bonds of the cis monoenoic chains arise by two distinct pathways depending on whether oxygen is required. The oxygen-independent pathway (traditionally called the anaerobic pathway) introduces the cis double bond by isomerization of the trans double bond intermediate of the fatty acid elongation cycle. Double bond isomerization occurs at an intermediate chain length (e.g., C10) and the isomerization product is elongated to the C16–C18 chains that become phospholipid monoenoic acyl chains. This pathway was first delineated in Escherichia coli and became the paradigm pathway. However, studies of other bacteria show deviations from this paradigm, the most exceptional being reversal of the fatty acid elongation cycle by a reaction paralleling the initial step in the β-oxidative degradation of fatty acids. In the oxygen-dependent pathway diiron enzymes called desaturases introduce a double bond into a saturated acyl chain by regioselective cis dehydrogenation through activation of molecular oxygen with an active-site diiron cluster. This difficult hydrogen abstraction from a methylene group often occurs at the midpoint of a saturated fatty acyl chain. In bacteria the acyl chain is a phospholipid acyl chain, and the desaturase is membrane bound. Both the oxygen-independent oxygen-dependent pathways are transcriptionally regulated by repressor and activator proteins that respond to small molecule ligands such as acyl-CoAs. However, in Bacillus subtilis the desaturase is synthesized only at low growth temperatures, a process controlled by a signal transduction regulatory pathway dependent on membrane lipid properties.

1. Why do bacteria make unsaturated fatty acids (UFAs)

Note that for simplicity the unsaturated acyl chains of bacterial membrane lipids (which are generally phospholipids) will be referred to as unsaturated fatty acids (UFA), although the acyl chains are in ester linkages, so hydrolysis is required to obtain acids.

Phospholipid bilayers composed of only saturated acyl chains of chain length C14 to C18 are rigid (non-fluid) at physiological temperatures [1]. The linear acyl chains pack together tightly to produce a bilayer that has a high phase transition temperature, high viscosity and low permeability. (The phase transition is a melting of the chains at higher temperatures where the bilayer acyl chains become fluid). Physiological function requires that the membranes be permeable (but not too permeable) to allow small molecules to enter and exit the cell. A rigid membrane hinders diffusion of proteins and other molecules (e.g., enzyme substrates, quinones) in the plane of the membrane and slows protein conformational changes. Indeed, high bilayer viscosity has been shown to inhibit cellular respiration by slowing diffusion between quinone electron carriers and the transmembrane oxidoreductases [2].

Cells cope with rigid membranes by modulating the packing of acyl chains [3, 4]. In bacteria the usual modulation of packing is done by cis unsaturated chains (UFAs) that have a single cis double bond generally located in the central portion of the chain. The double bond results in a kink in the chain of about 70° which disrupts packing. This results in lower phase transition temperatures, decreased bilayer rigidity and higher permeability. In a sense UFA containing phospholipid membranes are partially melted relative to phospholipid membranes of saturated fatty acids. There is nothing specific about the double bond in that a “bump” in the middle of the chain has a similar effect. For example, the double bond can be converted to a cyclopropane ring [5] or even replaced by a bromine atom [6].

Bacteria have two mechanisms of UFA synthesis. These are introduction of double bonds into growing acyl chains for example by the classical oxygen-independent FabA-FabB pathway or by direct conversion of saturated chains to UFAs by oxygen-dependent desaturation. Both pathways are discussed below. Synthesis of UFA or branched acyl chains is essential for bacteria. Escherichia coli strains lacking either FabA or FabB were first isolated in the presence of oleic acid. When the UFA supplement is removed these mutant strains cease growth and eventually lyse [7, 8]. As discussed below Bacillus subtilis can use either a specific branched acyl chain or a UFA to form membranes functional at low temperature. The absence of both acyl chain species blocks growth at low temperatures and the cells lyse and die [9]. Note that some bacteria have both these UFA synthesis mechanisms. Pseudomonas aeruginosa has the oxygen-independent FabA-FabB pathway plus two oxygen-dependent desaturates [10] and either synthesis mechanism supports growth. In contrast, E. coli has only the FabA-FabB pathway whereas the cyanobacteria have only desaturases.

The ecological niches occupied by the bacterium dictate the UFA synthesis pathway utilized. E. coli is found in both aerobic and anerobic environments: ground water and animal digestive systems, respectively, and uses the oxygen-independent FabA-FabB pathway. In contrast cyanobacteria grow by photosynthetic conversion of carbon dioxide to oxygen so they must grow in air and have no need for an oxygen-independent pathway. Hence cyanobacteria, algae and plants essentially utilize the oxygen produced by photosynthesis for desaturation of saturated chains to UFA chains. Cyanobacteria lacking all desaturates cannot be constructed consistent with the essential nature of UFA synthesis in these bacteria [11].

2. The Canonical Oxygen-Independent Pathway of UFA synthesis.

The first definitive report of UFAs in bacteria was the discovery of cis-11,12-octadecenoic acid (cis-vaccenic acid) in the Firmicute Lactobacillus plantarum (then called L. arabinosus) [12]. The Lactobacilli cannot synthesize the biotin cofactor required for synthesis of the malonyl-CoA building block of fatty acid synthesis However, biotin can be replaced by various UFAs [13] demonstrating that these bacteria can construct their membrane lipids entirely from exogeneous UFAs. The mechanism of the exogenous fatty acid incorporation process was unknown until recent years when Rock and coworkers [14, 15] demonstrated that Firmicute bacteria activate exogenous fatty acids by conversion to acyl-phosphates, one of the acyl donors in phospholipid synthesis. The other acyl donor is acyl-ACPs which are formed from acyl-phosphates by transfer of the acyl chain [14, 15]. The presence of this pathway explains the results of Hofmann, O’Leary and coworkers [16] who had the insight that the double bond of cis-vaccenic acid could be inserted into a shorter chain acid that was elongated to cis-vaccenic acid. To test this hypothesis, they synthesized putative precursors based on the Δ11 double bond position: cis-3-decenoic acid, cis-5-dodecenoic acid and cis-7-tetradecenoic acid. Palmitoleic acid (cis-9-hexadecenoic acid) was also tested. Palmitoleic acid replaced cis-vaccenic acid in the absence of biotin whereas growth with the other acids required a trace level of biotin in the medium [16]. Each of the synthesized fatty acids except cis-3-decenoic acid restored growth and synthesis of cis-vaccenic acid (or its cyclopropane derivative) [16] arguing that the acids were first converted to their acyl-phosphates and then to acyl-ACPs that entered the fatty acid synthesis pathway. The trace biotin requirement indicated that growth on cis-5-dodecenoic acid and cis-7-tetradecenoic acid required elongation. The inability of cis-3-decenoic acid to restore growth was probably due to chain length specificity of the incorporation system. These studies also ruled out desaturation of saturated acyl chains in these bacteria [16], although, as discussed below, bacteria including Firmicutes have desaturase enzymes. These data indicated that the double bond was inserted in a short chain molecule that was elongated in two-carbon increments to cis-vaccenic acid (Fig. 1). The position of the double bond in cis-vaccenic acid dictated its insertion into either position 3 of a C10 acyl chain or position 5 of a C12 acyl chain (Fig. 1).

Fig. 1.

Oxygen-independent pathways to UFA species in bacteria postulated by Scheuerbrandt and Bloch [18]. All the UFAs shown in the enclosed area have been isolated from bacterial lipids. Clostridium butyricum (now C. beijerinckii) [17] synthesizes palmitoleate and cis-vaccenate by cis double bond introduction at the C10 level as in E. coli and also synthesizes C16Δ7 and C18Δ9 by cis double bond introduction at the C12 level as in Aerococcu viridans. A. viridans performs an additional elongation cycle to give C20Δ11. Streptococcus pneumoniae probably performs cis double bond introduction at the C6 level because they contain C16Δ11and C20Δ13 acyl chains in addition to palmitoleate and cis-vaccenate [66, 67]. The C14Δ7 precursor to palmitoleate and cis-vaccenate has been reported in E. coli [124].

A seminal finding was that of Goldfine and Bloch [17] who demonstrated the presence of UFA in Clostridia which are strictly anaerobic bacteria. This finding proved the existence of an oxygen-independent UFA synthesis pathway. Scheuerbrandt and Bloch [18] subsequently showed that anaerobically grown E. coli cells synthesized palmitoleate and cis-vaccenate. These reports caused the pathway to be called the anaerobic pathway, although the oxygen-independent pathway is more accurate because the pathway also functions under aerobic conditions. The discovery of a pathway that synthesized UFA under anerobic conditions contrasted with what was known in animals and plants where oxygen-requiring desaturases were responsible for UFA synthesis [19].

3. Discovery of the FabA Enzyme and its Encoding Gene

Studies of the mechanism of double bond insertion were greatly aided by the oxygen-independent pathway of E. coli because cells of this bacterium were commercially available in the kg quantities then needed for enzymology. Bloch and coworkers [19] hypothesized that the double bond was inserted at the C10 level, and the key enzyme (now called FabA) was found using radioactive 3-hydroxydecanoyl-NAC [20, 21]. (NAC is N-acetylcysteamine and represents the thiol-containing β-mercatoethylamine moiety of both CoA and the ACP prosthetic group.) It has long been known that 3-hydroxyacid thioesters are readily dehydrated to trans-2 acyl-thioesters so 3-hydroxydecanoyl-NAC seemed a suitable substrate. The most parsimonious reaction would be direct dehydration of 3-hydroxydecanoyl-NAC to cis-3-decenoyl-NAC. However, partially purified FabA was found to dehydrate 3-hydroxydecanoyl-NAC to trans-2-decenoyl-NAC and isomerize the trans product to cis-3-decenoyl-NAC [20, 21] (Fig. 2). Highly purified FabA was found to interconvert the 3-hydroxy, trans-2 and cis-3-decenoyl-NAC species [22, 23]. Each species was readily converted to the other two. For technical reasons the ratios of the three species varied among reports from the Bloch lab but since NAC thioesters are artificial substrates, this seemed of minor consequence. The variation in the Bloch lab reports of the FabA product ratios were due to several factors over the years. In early work [21] the thioester bonds were cleaved by base hydrolysis, which can interconvert trans-2-decenoic acid and 3-hydroxydecanoic acid. In later work a contaminating thioesterase activity that copurified with FabA was found to specifically cleave cis-3-decenoyl-NAC while leaving the other products untouched [24]. Much later Guerra and Browse [25] repeated the equilibrium experiments with ACP thioesters, the in vivo substrates, and found the three species were almost evenly distributed. They also showed that 3-hydroxydecanoyl-ACP was a much better substrate than 3-hydroxydecanoyl-NAC with a K_M >300 fold lower than that of the NAC thioester [25]. Experiments with NAC-thioesters showed that FabA was most active with C10 substrates and inactive with C8 and C10 substrates [23]. C9 and C11 substrates had some activity. Consistent with the NAC-thioester results, later experiments using an acyl-ACP synthetase to allow entry of C7 heptanoic and the C9 nonanoic acids into the E. coli fatty acid synthesis cycle gave only traces of unsaturated phospholipid chains of odd chain length [26]. These findings are not of physiological importance because E. coli does not contain odd chain length acyl chains.

Fig. 2.

In E. coli the branch point between SFA/UFA synthesis occurs at the level of the tencarbon intermediate. Both FabA and FabZ catalyze the dehydration of 3-hydroxydecanoyl-ACP to trans-2-decenoyl-ACP (Heath and Rock, 1996) whereas only FabA can isomerize the double bond to cis-3-decenoyl-ACP [ 8, 22, 27). SFA biosynthesis proceeds by the action of FabI on the trans-2 intermediate followed by further elongation cycles initiated by a long-chain 3-oxoacyl-ACP synthase (either FabB or FabF). UFA synthesis requires FabB to elongate cis-3-decenoyl-ACP [39] and thereby initiate the elongation cycles that form the major long-chain UFAs, C16:1Δ9 and C18:1Δ11. (B) In Aerococcus viridans the UFA synthesis pathway branches at the 3-hydroxydodecanoyl-ACP stage [70]. As described elsewhere in this review FabQ dehydrates 3-hydroxydodecanoyl-ACP to trans-2-dodecenoyl-ACP and isomerizes a portion of this intermediate to cis-3-dodecenoyl-ACP. FabF is required for the subsequent elongation of cis-3dodecenoyl-ACP to produce the long-chain unsaturated fatty acids. SFA are formed by the action of the FabK enoyl-ACP reductase followed by further elongation cycles initiated by FabF. Reproduced from [70] with permission.

4. Discovery of mechanism-based inhibition and its application to FabA

An unexpected result of studies of FabA of major importance to mechanistic enzymology was the discovery of mechanism-based inhibition (also called enzyme suicide or suicide inhibition). In the process of synthesizing cis-3-decenoyl-NAC, one substrate batch was not only inactive, but rapidly and irreversibly inactivated FabA [27]. That batch contained a contaminant that turned out to be the 3-decynol thioester [22]. This arose because a portion of the 3-decynoic acid used to synthesize this batch of cis-3 decenoic acid had escaped reduction and subsequently was converted into the 3-decynol thioester. When incubated with pure 3-decynol-NAC, FabA was found to catalyze isomerization of the acetylenic thioester to the corresponding 2,3-allenic thioester which attacked an active site histidine residue thereby forming an inactivating covalent adduct [28]. The specificity of 3-decynol-NAC was demonstrated by addition of the compound to E. coli cultures. This blocked growth by specific inhibition of UFA synthesis as shown by restoration of growth by addition of oleic acid while palmitic acid failed to rescue growth [29]. Although 3-decynol-NAC seemed likely to be a useful antibacterial, this was not the case. In mammalian liver 3-decynol-NAC is converted to the 2,3-allenic thioester which is inactivated by hydration to 3-hydroxy-decanoyl-NAC [30].

Mechanism-based inhibition provided a means to cope with the fleeting nature of acyl-ACP-enzyme interactions which has long hindered study of these interactions. Nguyen and coworkers [31] extended the mechanism-based inhibition work of the Bloch laboratory to crosslink FabA to acyl-ACP. These investigators synthesized a non-hydrolysable compound in which a sulfonyl group linked a 3-alkyene moiety to pantetheine. This was converted to an acyl-ACP analog via a CoA analog and the resulting acyl-ACP analog was reacted with FabA. The allene generated from this mechanism-based probe allowed site-selective covalent crosslinking of the modified acyl-AcpP to FabA that gave a 1.9 Å crystal structure of crosslinked dimers of the complex. This pioneering report showed that the 4′-phosphopantetheine (PPant) group of AcpP first binds an arginine-rich groove of FabA, followed by an AcpP helix III conformational change that secures the AcpP-FabA complex. Interaction of the acyl-AcpP with FabA flips acyl chain from the AcpP into the FabA active site. This was one of the first examples of the chain flipping mechanism whereby the acyl chain of an acyl-ACP leaves the hydrophobic ACP lumen and enters the hydrophobic active site of a fatty acid synthetic enzyme [32]. In later work [33] C6 and C8 acyl-AcpP probes were synthesized and their ability to crosslink to FabA was compared to the C10 mimic of the natural substrate used in the pioneering work. Chain length played an important role in crosslinking efficiency. Relative to the C10 species the C8 species was about half as efficient in crosslinking whereas the C6 species was only one-tenth as efficient. Extending the PPant arm by three methylene groups only slightly decreased the crosslinking efficiency of the C10 species. The effects of FabA site directed mutations on crosslinking was used to dissect the roles played by residue sidechains in the interactions of FabA with acyl-AcpP species [33]. First, the role of the FabA basic residues thought to bind AcpP was examined. Conversion of any of four basic residues to acidic residues decreased crosslinking of FabA with all four acyl-AcpP species. Crosslinking of FabA to the C6-species was abolished and that of the C8 species was significantly weaker than C10 crosslinking. These data defined the FabA “positive patch” that binds the acidic AcpP moiety. Mutants designed to restrict or enlarge the FabA acyl chain binding pocket were tested by crosslinking. In the pocket restriction approach residues having small sidechains were converted to phenylalanine residues. This resulted in decreased crosslinking of the C8 and C10 species whereas the C6 species profited by this binding pocket restriction. Mutant proteins designed to enlarge the acyl chain binding pocket involved conversion of large hydrophobic side chains to alanine or glycine. These mutant FabA proteins retained significant crosslinking activity with all four probes. Finally, these workers focused on two phenylalanine residues thought to provide gates that distinguish chain lengths. Conversion of these residue to alanine increased crosslinking of the C6 species by 3 to 4-fold but decreased the crosslinking efficiency of the C10 species (the C8 species was unchanged).

The acyl chain length specificity of FabA was subsequently probed by NMR titration and in silico docking studies [34]. Acyl-AcpP chain lengths of 6, 8 and 10 were examined. These data agreed with those obtained by crosslinking but suffered from the use of saturated acyl chains which are neither FabA substrates nor products. Indeed, the authors concede that “catalytically inactive enzymes and the native substrate will be necessary to further study specificity”. Indeed, this point has become more germane since other data from the same group argue that the acyl-AcpP surface presented to enzyme proteins is altered by the acyl chain [35]. This aspect could impart allosteric specificity to these interactions and the FabA acyl-AcpP substrates 3-hydroxydecanoyl-AcpP, trans-2-decenoyl-AcpP and cis-3 decenoyl-AcpP may differ markedly from the saturated chains tested.

5. Isolation of E. coli strains defective in UFA synthesis delineated the oxygen-independent pathway

At roughly the same time as the early FabA enzymology, Silbert and Vagelos [36] isolated an E. coli mutant strain that required UFA for growth and lacked FabA activity. The UFA requirement was satisfied by several UFAs with oleate as the usual supplement. The surprise was that saturated fatty acid synthesis continued in the mutant strain [36] indicating that E. coli must encode a second dehydrase (these enzymes are now called dehydratases). Saturated fatty acid synthesis also continued in 3-decynol-NAC inhibited cells indicating that the inhibitor was inactive on the second dehydrase [29]. The presence of a second dehydratase activity was demonstrated in the Silbert-Vagelos strain [37, 38] and the gene was later identified (see below). By this time other laboratories had isolated UFA-requiring strains, the most useful being those from the D. L. Wulff laboratory [39, 40]. The Wulff strains were missense temperature-sensitive strains that required UFA only when grown at 42°C. At 30°C the strains grew well without UFA supplementation. These strains were an advantage to genetic analysis because the presence of UFA in growth media was problematic because several calcium-dependent gene transfer methods failed because calcium ions form insoluble salts with fatty acids [40]. Since the Wulff strains did not require UFA at 30°C, this permitted in vivo and in vitro complementation analyses which showed that UFA requiring strains were of two classes called fabA and fabB [39]. The fabA temperature sensitive strains encoded a dehydrase that had decreased temperature stability indicating that fabA was the structural gene encoding the dehydrase rather than a gene that regulated FabA synthesis or activity [7]. Genetic analysis [40, 41] showed that the fabA and fabB mutants mapped at two well-separated locations on the E. coli chromosome.

The dehydratase remaining in the Slbert-Vagelos strain was shown to be encoded by the fabZ gene [42]. FabZ and FabA are related proteins (30% identical residues). E. coli FabZ was purified and shown to dehydrate 3-hydroxy-ACPs but was unable to isomerize trans-2-decenoyl-ACP to cis-3-decenoyl-ACP [43]. The question of why FabZ lacks the FabA isomerase activity was long thought to be differing architectures of the acyl chain binding tunnels of the two proteins [44]. Dodge and coworkers [45] report that this is the case by use of a new FabZ crystal structure plus a prior FabA structure and molecular dynamics simulations. The shape of the FabA tunnel provides the proper angle for the C3–C4 dihedral shift required for spontaneous trans-2 to cis-3 isomerization of the decenoyl-ACP acyl chain (Fig. 3). Indeed, the shape of the FabA tunnel may force isomerization [45]. The FabZ tunnel lacks this geometry and some FabZ tunnels have a tortuous “U” shape [46] (Fig. 3).

Fig. 3.

The active site tunnels of FabA (upper panel) and FabZ (lower panel). FabA/Z and ACP are shown in green and magenta. E. coli FabA (PDB code: 4KEH) and Helicobacter pylori FabZ (PDB code: 4ZJB) dimers are shown in green/cyan cartoons. The enzyme active sites are at the dimer interphases (FabA is a dimer whereas FabZ is a trimer of FabA-like dimers. The active site tunnels are in white and are magnified to the right. The lengths of sections of the tunnels are labeled.

6. Discovery of the FabB Enzyme and its FabF Homologue

The role of the protein encoded by the fabB class of mutants was a puzzle. Some possibilities were an enzyme that catalyzed a discrete step in UFA synthesis, or a protein required for activation of a putative FabA proenzyme. Two findings coalesced to provide an explanation. One was the serendipitous isolation of an E. coli strain blocked in cis-vaccenate synthesis [47]. This strain accumulated palmitoleate but very little cis-vaccenate (Fig. 4) arguing that there are at least two elongation enzymes in E. coli, one of which was specific for the last elongation step of UFA synthesis. However, at that time all E. coli chain elongation (β-ketoacyl-ACP synthase) activity was thought to be catalyzed by a single well-studied enzyme preparation [48] [49]. However, later those preparations were found to contain two separable activities, one of which was absent in a fabB mutant strain [50]. The fabB temperature-sensitive strains were shown to encode a β-ketoacyl-ACP synthase with decreased temperature stability indicating that fabB encoded that enzyme [51]. Hence, the remaining β-ketoacyl-ACP synthase activity seemed likely to be the enzyme missing in the strain blocked in cis-vaccenate synthesis and this was the case [51] (Fig. 4) . These enzymes are now called β-ketoacyl-ACP synthase I (FabB) and β-ketoacyl-ACP synthase II (FabF). FabB and FabF were shown to be different proteins, first by a FabB specific antibody plus peptide mapping [52] and subsequently by the sequences of the encoding genes which showed the proteins to be 37% identical [53, 54]. FabB is required for all UFA synthesis whereas FabF is required for elongation of palmitoleyl-ACP to cis-vaccenoylACP [33,34]. Both enzymes catalyze all the elongation steps of saturated fatty acid synthesis. The FabB and FabF elongation reactions proceed with the same active site residues and by the same mechanism [44] and are the only E. coli long chain β-ketoacyl-ACP synthases. Double fabB-fabF mutant strains are nonviable and cannot be rescued by UFA supplementation [51]. Hence, each β-ketoacyl-ACP synthase catalyzes a reaction in UFA synthesis that the other cannot perform or performs very poorly and loss of both enzymes blocks saturated fatty acid synthesis. It was later shown that only FabB can elongate the FabA product, cis-3-decenoyl-ACP to cis-5-dodecenoyl-ACP [55]. Apparently, the double bond located early in the acyl chain blocks elongation of this substrate by E. coli FabF. FabB elongates palmitoleyl-ACP but only very weakly [56].

Fig. 4.

Detection of two long chain 3-ketoacyl-ACP synthases in E. coli.

The left panel shows the autoradiogram of a silver nitrate impregnated silica gel thin layer chromatography plate separation of [¹⁴C]acetate-labeled fatty acid methyl esters derived from in vivo labeling of membrane phospholipids [47]. SAT denotes the saturated acyl chains, Cvc denotes cis-vaccenate and Pol denotes palmitoleate. WT denotes the wild type strain. The right panel shows the separation of two long chain 3-ketoacyl-ACP synthases by elution from a hydroxyapatite chromatography column with a potassium phosphate gradient [51]. Peak I is FabB and peak II is FabF. This separation was first demonstrated by D’Agnolo and coworkers [50]. The ratios of the two peaks vary with the assay temperature and acyl-ACP substrate [52].

In 1962 Marr and Ingraham [57] reported that E. coli alters the fatty acid composition of its phospholipid acyl chains depending on the growth temperature. At low growth temperatures cis-vaccenoyl chains replace saturated chains. This response is a property of FabF, fabF mutant strains lack temperature regulation [47, 51, 52]. FabF is expressed at all temperatures but becomes more active at low temperatures as shown by assays of the purified enzyme [52, 58]. Recently the temperature regulation seen in E. coli has been shown to occur in an unrelated bacterium, the Firmicute, Enterococcus faecalis, and to require the FabF protein [59]. Moreover, the E. faecalis FabF restored temperature regulation to an E. coli strain lacking FabF [59]. Temperature regulation is not essential but rather acts to optimize the membrane phospholipid bilayer for improved function at the new temperature.

Note that FabA releases its trans-2-decenoyl-ACP intermediate to solution [25]. Hence, a portion of the FabA dehydration product could dissociate, be captured by an enoyl-ACP reductase, and shunted off to the saturated fatty acid synthetic branch. Therefore, there is a competition between isomerization and reduction. Overproduction of an enoyl-ACP reductase can block UFA synthesis by reduction of trans-2-decenoyl-ACP before it can be isomerized [60–62]. There are several different bacterial enoyl-ACP reductases called FabI, FabL, FabV and FabK and some bacteria encode two such enzymes [63, 64].

The FabA-FabB pathway is found in many Gram-negative bacteria and in some Gram-positive Firmicute bacteria. A Firmicute example is in the E. faecalis V583 genome which was annotated as encoding two FabZs and two FabFs. However, genetic complementation of E. coli mutant strains showed that one of the putative FabZs had FabA activity and one of the putative FabFs had FabB activity [62] (the other two enzymes had the annotated FabZ and FabF activities). The E. faecalis enzymes having FabA and FabB activities were named FabN and FabO, respectively. Further work showed that swapping a FabN domain for a FabZ domain gave FabA activity [65]. E. coli FabA is a dimer whereas E. coli FabZ is a hexamer composed of three FabA-like dimers (a trimer of dimers). However, recent work has shown that FabN is a hexamer rather than the expected dimer [60].

7. Variations of the FabA-FabB pathway

The first variation the FabA-FabB pathway reported was the FabM protein of Streptococci which has isomerase activity but lacks dehydratase activity [61]. Hence FabM must use trans-2-decenoyl-ACP generated by the FabZ dehydratase as its isomerization substrate. FabM can functionally replace FabA in E. coli provided that the host enoyl-ACP reductase is partially inactivated [61]. FabM mutant strains, which are UFA auxotrophs C8 [66, 67], can be complemented by expression of E. faecalis FabN [67]. FabM is a tetrameric enzyme having no similarity to FabA or FabN and is a member of the hydratase/isomerase superfamily [61]. Thus, in Streptococci the branch point in the biosynthesis of unsaturated fatty acids in S. pneumoniae follows formation of trans-2-decenoyl-ACP rather than using the preceding 3-hydroxydecanoyl-ACP step as in the classical FabAB pathway. FabM is less specific than FabA and Streptococci contain a series of UFA species (C16 species having a double bond at C7, C9 or C11) indicating that FabM isomerizes trans-2-acyl-ACPs of chain lengths C12, C10 or C8 [66, 67] (Fig. 1).

Some bacteria (e.g, Lactococcus lactis and Clostridium acetobutylicum) make palmitoleyl and cis-vaccenoyl acyl chains but encode only one long chain β-ketoacyl-ACP synthase generally annotated as FabF. Expression of L. lactis FabF functionally replaces both FabB and FabF in E. coli, although it does not restore thermal regulation of phospholipid acyl chain composition to E. coli fabF mutant strains (L. lactis lacks thermal regulation) [68]. The C. acetobutylicum FabF also functionally replaces both E. coli FabB and FabF, although it differed from the L. lactis enzyme in that it restored thermal regulation to an E. coli fabF strain [69]. Purified C. acetobutylicum FabF was shown to elongate cis-3-decenoyl-ACP, albeit more poorly than does E. coli FabB. Hence, unlike E. coli FabF, these FabF proteins accept cis-3-decenoyl-ACP as a substrate [69]. Recent work has shown that E. faecalis strains lacking FabO grow without UFA supplementation because E. faecalis FabF has a modest ability to elongate cis-3-decenoyl-ACP [60]. Growth of strains that lack both FabO and FabF requires UFA supplementation [60].

8. Major Departures from the FabA-FabB pathway

Two bacteria, Aerococcus viridans and Helicobacter pylori synthesize C16 and C18 UFAs. However, the genome of the Gram-positive A. viridans encodes only one FabA/FabZ protein called FabQ whereas the genome of the Gram-negative H. pylori lacks a FabA homologue and uses FabX, a very different enzyme, to introduce the double bond.

FabQ, the single hexameric FabA/FabZ homolog encoded by A. viridans suggested that this enzyme could carry out the functions of both E. coli FabZ and FabA in fatty acid synthesis [70]. Indeed, FabQ not only replaces the function of E. coli FabZ in vivo, but also catalyzes the isomerization required for unsaturated fatty acid biosynthesis [70]. When FabQ was expressed in E. coli detection of the isomerization reaction required partial inactivation of the competing FabI enoyl-ACP reductase. Under these conditions FabQ synthesized UFAs having the double bond positions characteristic of A. viridans, cis-7 and cis-9 UFAs in place of the cis-9 and cis-11 species commonly found [54]. The double bond positions indicates that FabQ dehydrates the C12 substrate 3-hydroxydodecanoyl-ACP to trans-2-dodecenoyl-ACP and then isomerizes the double bond to the cis-3-isomer [70] (Fig. 2). Thus, in A. viridans the UFA biosynthetic pathway branches from the classic fatty acid biosynthesis pathway by ‘tapping off” the 3-hydroxydodecanoyl-ACP intermediate rather than the shorter 3-hydroxydecanoyl-ACP used by dual-function dehydratase/isomerase- containing bacteria Fig. 2. FabQ can act strictly as a dehydratase like E. coli FabZ or as a dual dehydratase/isomerase like FabA and the native bacterium must partition trans-2-dodecenoyl-ACP such that both saturated fatty acids and UFA are made [70]. Strikingly, FabQ in combination with E. coli FabB imparts the surprising ability to bypass reduction of the trans-2-acyl-ACP intermediates of classical fatty acid synthesis, although only trace amounts of the products are formed. This combination of enzymes allows elongation by progressive isomerization reactions to form the polyunsaturated fatty acid, 3-hydroxy-cis-5, 7-hexadecadienoic acid, both in vitro and in vivo [70].

The first report of a new oxygen-independent UFA synthesis pathway was that of Isabella and Clark [71] in Neiserria gonorrhoeae, a bacterium that lacks genes encoding homologs of FabA, FabM and FabB. These investigators did transcriptional profiling following shift of N. gonorrhoeae cultures from aerobic to anaerobic conditions and found increased transcription of a gene that encoded a FabK-like enoyl-ACP reductase. They disrupted this gene (called ufaA) and showed the mutant strain could no longer grow anaerobically, although aerobic growth was normal [71]. Given the FabK-like sequence they tested if fatty acids allowed anaerobic growth of the mutant strain and found that UFA, but not saturated fatty acids, supported growth. Isabella and Clark reported that expression of FabA, FabM or FabB failed to complement anaerobic growth of the ufaA strain indicating that the gene denoted a new anaerobic UFA synthesis pathway that seemed widely distributed in bacteria [71].

A close Ufa homolog (called FabX) is encoded by the gastric bacterium, Helicobacter pylori. The sequences of Ufa and FabX contain sequence clusters at their carboxy ends that are not present in known FabK proteins arguing that these proteins have a different function. Although the Ufa protein had no activity in E. coli [71], the H. pylori FabX was active in E. coli and synthesized the cis-vaccenyl acyl chains found in H. pylori [72]. To test if FabX expressed in E. coli utilized the FabA 3-hydroxydecanoyl-ACP substrate, the patterns of label in long chain acyl chains formed by elongation of ¹⁴C-labeled short chain acids were determined. In the classical pathway the incorporated C8 label is found in both unsaturated and saturated acyl chains whereas the incorporated C10 label is found only in saturated acyl chains [19] (Fig. 2). This is because a finished C10 chain is past the branch point where the double bond is inserted, and the enoyl-ACP reductase reaction is essentially irreversible [72]. However, an E. coli fabA strain that expressed FabX was found to shunt labeled C10 into both unsaturated and saturated acyl chains whereas the C12 acid labeled only saturated acyl chains [72]. The finding that the labeled C10 acid was incorporated into both unsaturated and saturated acyl chains indicated that FabX somehow reversed fatty acid synthesis because synthesis of cis-vaccenic acid indicated that the cis double bond was inserted between carbons 3 and 4 of a C10 acid [72]. The only way this backtracking could occur was for FabX to catalyze a 2,3-dehydrogenation reaction analogous to the acyl-CoA dehydrogenase of the β-oxidation pathway of fatty acid degradation. Dehydrogenation would produce trans-2-decenoyl-ACP which would be isomerized to cis-3-decenoyl-ACP [72]. FabX was purified and was found to be a monomeric flavin (FMN) enzyme as expected from its homology to FabK. Incubation of the purified FabX with decanoyl-ACP produced a mixture of trans-2-decenoyl-ACP and cis-3-decenoyl-ACP. In the backtracking dehydrogenation reaction, the FMN cofactor would be reduced to FMNH₂ and enzyme turnover (catalysis) would require reoxidation of the flavin. [72]. FabX was catalytic in vitro, although no oxidant was intentionally added to the reactions suggesting that the oxidant was oxygen. This was the case. No reaction occurred under anaerobic reactions but addition of 3–10% O₂ gave activity [72]. Higher O₂ concentrations were somewhat inhibitory (air is 21% O₂). Isomerization to produce cis-3-decenoyl-ACP was thought to be due to the shape of the acyl chain binding tunnel as in FabA and this was confirmed by the crystal structure described below. Deletion of the fabX gene from the H. pylori genome was not straightforward probably because H. pylori cannot incorporate UFA from the medium which precluded UFA supplementation of a mutant strain. The solution was to express E. coli FabA in H. pylori and construct the fabX deletion in this strain [73]. The fabX gene was readily deleted and FabA fully complemented growth. Hence, FabA could access the H. pylori 3-hydroxydecenoyl-ACP intermediate and produce the cis-3-decenoyl-ACP needed for UFA synthesis [73].

The 2.3Å crystal structure of FabX in a complex with octanoyl-ACP showed an overall structure like that of FabK [73]. This was expected but there also was a surprise, a small [4Fe–4S] cluster-binding subdomain was present at the C-terminus (Fig. 5). This was the first case of a 4Fe–4S] cluster in a fatty acid synthetic enzyme. Indeed, the four cysteine residues that scaffold the cluster allow differentiation of FabX proteins from FabK proteins. The [4Fe–4S] cluster is essential for FabX activity and its loss results in loss of the FMN cofactor [73]. The [4Fe-4S] cluster potentiates FMN oxidation during dehydrogenase catalysis, generating superoxide from an oxygen molecule that is locked in an oxyanion hole between the FMN and the His182 active site residue. The H. pylori [4Fe–4S] cluster is largely protected from solvent which may explain why it is more stable to oxygen than other [4Fe–4S] clusters. Indeed, the structural interactions of the [4Fe–4S] and FMN of H. pylori FabX provides an explanation of puzzling results of Srinivas and Cronan (unpublished) with the FabX of C. acetobutylicum. When first isolated the protein had the intense yellow color of a flavin protein. However, when handled in the laboratory the yellow color faded and finally disappeared. Addition of FMN to the buffer failed to restore the FabX flavin. The probable reason for the flavin loss is that the C. acetobutylicum [4Fe–4S] cluster was degraded by oxygen which caused loss of the flavin as seen in H. pylori FabX [73] and collapse of the FMN binding site prevented restoration. The C. acetobutylicum FabX may be less protected from oxygen than H. pylori FabX because C. acetobutylicum is an obligate anaerobe whereas H. pylori is a microaerophile.

Fig. 5.

The structure of FabX in magenta and octanoyl-ACP (with the ACP moiety colored blue) (PDB 7E1S). The 4’-phosphopantheine with attached octanoyl group in CPK colors has flipped from the ACP lumen into the FabX active site [32]. The FMN cofactor and the [4Fe-4S] cluster are labeled. The structure was made using iCN3D. FabX belongs to the nitronate monooxygenase (flavoprotein family and is predominately a TIM barrel protein [73].

FabX, has dual dehydrogenase and isomerase activities [72, 73]. This unprecedented combination of reactions reverses the standard fatty acid synthesis cycle at the C10 stage. The properties of FabX explains how many bacterial species lacking the classical pathway enzymes are able to produce the UFA-containing phospholipids required for functional cell membranes. However, the fact that E. coli FabA can functionally replace FabX in H. pylori [73] raises the question of why FabX arose. Isabella and Clark [71] proposed that FabX replaced the FabAB pathway in Neiserria gonorrhoeae. Bioinformatic studies suggest that FabX homologs (identity higher than 50%) are encoded widely in bacteria [73]. At present the litmus of FabX versus FabK and other nitronate monooxygenase flavoprotein family members is the cluster of cysteine residues at the C-terminus. A complication is that many of the putative FabX encoding genes are in anaerobic bacteria and H. pylori FabX requires oxygen for activity at least in vitro. Electron bifurcation provides a possible substitute for oxygen [72]. This process is utilized by the mitochondrial β-oxidation acyl-CoA dehydrogenase which catalyzes a reaction that parallels the FabX reaction. However, electron bifurcation is a complex and diverse process, and its components are challenging to identify by bioinformatics [74, 75]. The possibility of H. pylori FabX electron bifurcation cannot be tested because H. pylori requires oxygen for growth.

9. Synthesis of UFAs by the Oxygen-Dependent Desaturation Pathway

Desaturases synthesize UFA by oxygen-dependent dehydrogenations initiated at the unactivated C-H groups of saturated chains by use of a diiron active site (Fig. 6) [76–78]. These enzymes contain three clusters rich in histidine residues which together with acidic residues coordinate two iron atoms. Most of the work on bacterial desaturases has focused on three groups, the Bacilli, the pseudomonads and the cyanobacteria. Thus far, bacterial desaturases are mainly membrane-bound enzymes that modify the saturated acyl chains of membrane lipids. Most bacterial desaturases insert only a single double bond although some cyanobacteria have multiple desaturases that insert multiple double bonds [78]. The monounsaturated acyl chain becomes the substrate for the desaturase that inserts the second double bond. This product becomes the substrate for the desaturase that inserts the third double bond and so on until a chain having as many as four double bonds is formed [78]. The cyanobacteria are a very diverse set of photosynthetic bacteria and show marked diversity in their desaturases.

Fig. 6.

The fatty acid desaturase reaction.

These enzymes all contain a diiron active site and the reaction proceeds by a common mechanism. Most studies of desaturase mechanisms have been performed with plant desaturases. Some of these are soluble and use acyl-CoA or acyl-ACP substrates whereas others are membrane bound like the bacterial enzymes. Shanklin and coworkers [77] have provided an excellent review of the plant enzymes whereas Sastre and coworkers [76] have reviewed bacterial desaturases.

10. Bacillus desaturases

Fulco began studies of bacterial desaturases while in the Bloch laboratory [79] and continued this work as an independent investigator. The bacterium studied by Fulco was mainly Bacillus megaterium which desaturates palmitate to Δ5-hexadecenoate only at low growth temperatures [80]. Fulco showed that inhibitors of protein or RNA synthesis blocked desaturation after shift to low temperatures indicating synthesis of a new protein at low temperature was required [81]. This clearly differentiates this type of thermal control from that exerted by FabF where such inhibitors have no effect because FabF is a previously existing protein [58, 59]. Fulco and coworkers also reported results similar to those seen in B. megaterium in several other Bacillius species, one of which was the genetically tractable B. subtilis [80] studied by de Mendoza and coworkers [82]. In the de Mendoza laboratory the inhibitor studies were performed in B. subtilis and gave same results reported in B. megaterium [83]. They also showed that phospholipids labeled with [¹⁴C]acetate at 37°C were desaturated upon shift to 20°C. This indicated that the acyl chains of membrane phospholipids rather than acyl-CoAs or acyl-ACPs were the desaturase substrate [83]. When the des gene promoter (see below) was replaced with a synthetic promoter, synthesis of Δ5 acyl chains proceeded at 37°C. This indicated that the electron transfer components needed for desaturation were present in cells grown at high temperature [84]. After shift of cultures from 37°C to 20°C the rate of desaturation increased for several hours and then declined to a maintenance rate [83]. This behavior had previously been observed by Fulco and coworkers in B. megaterium who termed the desaturation burst seen soon after shift as hyper-induction [81, 85]. Hyper-induction of desaturation implied that the role of desaturation was to quickly adapt the membrane phospholipids to the new temperature and then maintain the adaption. That is, the pulse of desaturation des activity upon shift was somehow modulated [81]. When des transcription was driven by a synthetic promoter desaturation at 20°C did not decrease as with the native gene: the burst continued. It might be expected that temperature adaption and hence the desaturase would be essential for survival at the low temperature. To test this premise required a B. subtilis strain lacking the desaturase. Prior to the B. subtilis genomic sequence the des gene, encoding the sole desaturase from B. subtilis, was found by complementation of E. coli strains having mutations in either fabA or fabB [86]. Complementation was done using a clone bank of B. subtilis chromosome fragments inserted into a plasmid vector that replicated in both B. subtilis and E. coli. First an E. coli fabB mutant strain was complemented which could be explained by clones expressing a 3-ketoacyl-ACP synthase. However, when these same clones also complemented a fabA mutant strain, this argued that the clones must encode the desaturase [86]. Indeed, expression of the B. subtilis gene in E. coli resulted in Δ5 desaturation of [¹⁴C]palmitic acid previously incorporated into membrane phospholipids [77, 86]. The complementing gene was sequenced and named des for desaturase. The protein sequence encoded by des contained the appropriately spaced three histidine-rich clusters and putative membrane sequences characteristic of known membrane desaturates {Aguilar, 1998 #45]. The B. subtilis des gene was disrupted, and the des null strain failed to synthesize detectable UFA following temperature downshift from 37°C to 20°C. Introduction of the original clone into the des null strain restored synthesis of Δ5 acyl chains. To assay induction of desaturase expression upon a temperature shift, a construct in which the des promoter was used to drive expression of E. coli β-galactosidase was inserted into the B. subtilis chromosome. When grown at 37°C only traces of β-galactosidase were made. However, upon shift to 20°C β-galactosidase levels increased 10- to 15-fold indicating that the des gene itself was induced rather than a protein that activated a latent desaturase activity [86].

Does des play a role in adaption of the B. subtilis membrane to low growth temperatures? B. subtilis phospholipids contain few of the straight chain acyl chains found in the membrane lipids of E. coli and most bacteria. Instead, B. subtilis synthesizes acyl chains that are a mixture of terminally branched species [87]. The branched structures originate from acyl-CoA intermediates synthesized by a branched-chain ketoacid dehydrogenase which act as primers in fatty acid synthesis in place of the usual acetyl-CoA or acetyl-ACP [82, 87]. The acyl-CoAs are also intermediates in synthesis of the branched chain amino acids. The valine pathway intermediate produces C13 and C15 chains with an iso branch whereas the leucine pathway intermediate produces C14 and C16 iso branched chains. In contrast, the isoleucine pathway intermediate produces C15 and C17 chains having an anteiso branch. This last species, particularly the C15 chain length, increases fluidity of the membrane lipids [87] and its synthesis at low temperatures is increased by an unknown mechanism.

By historical accident it was found that B. subtilis adapts to low growth temperatures by either increasing anteiso branched acyl chain synthesis or by desaturation. The accident was that the 168 strain of B. subtilis used in most genetic manipulations is deficient in isoleucine synthesis at low growth temperatures whereas other B. subtilis strains synthesize sufficient isoleucine at low growth temperatures [88]. In strains derived from B. subtilis 168 des becomes essential at low temperatures in the absence of isoleucine supplementation but not when isoleucine is supplied [9, 89]. This is because isoleucine is deaminated and the resulting ketoacid is converted to the acyl-CoA intermediate required for synthesis of anteiso branched acyl chains [87].

11. Pseudomonas desaturases

The bacteria discussed above, and most other bacteria have only a single UFA synthesis pathway. However, Pseudomonas aeruginosa PAO1 has not one, but three pathways to synthesize UFAs [10]. This bacterium has the classical FabAB UFA oxygen-independent pathway plus two oxygen-requiring membrane bound desaturases. The two desaturates are an acyl-lipid desaturase (DesA) and an acyl-CoA desaturase (DesB). DesA introduces a double bond in the Δ9 position of 16:0 FA chains esterified to the sn2 position of existing glycerophospholipids [90]. In contrast DesB desaturates exogenously supplied saturated acyl chains probably following their conversion to 16:0-CoA and 18:0-CoA to give 16:1Δ9-CoA and 18:1Δ9-CoA, respectively, which are incorporated into phospholipids [10]. The presence of the DesA desaturase was detected by the ability of a ΔfabA strain to grow aerobically despite lacking the oxygen-independent pathway [10]. Bypass of the ΔfabA mutation required oxygen: anaerobic cultures failed to grow unless supplemented with a UFA. Search of the proteins encoded by P. aeruginosa PAO1 genome using the B. subtilis Des sequence as a probe showed two proteins having the hallmarks of membrane bound desaturates. These are the three appropriately spaced histidine-rich clusters and putative transmembrane helices [10]. The genes encoding the two proteins were inactivated and the resulting strains grew under both aerobic and anaerobic conditions indicating that the FabAB pathway is the major source of UFAs [10]. Radioactive precursor labeling studies analogous to those done in B. subtills showed that DesA directly inserted double bonds into phospholipid acyl chains. Strains lacking both FabA and DesA failed to grow unless supplemented with a UFA. However, supplementation with a C16 or C18 saturated fatty acid also supported growth [10]. The C16 and C18 saturated fatty acids were converted to CoA esters, desaturated by DesB desaturase and incorporated into phospholipids. Since DesB desaturates only exogenously supplied fatty acids, its substrate cannot be the acyl chains of membrane phospholipids otherwise DesB would functionally replace DesA. Therefore, the DesB substrate is either the free fatty acids or much more likely, their acyl-CoA esters. A striking finding is that exogenous stearate is not desaturated in a ΔfabA ΔdesA strain grown on oleate. Stearate is incorporated into phospholipids but somehow oleate supplementation inhibits desaturation [10]. A triple mutant strain lacking FabA, DesA and DesB could no longer grow on C16 or C18 saturated fatty acids but grew with oleate supplementation [10].

This raises the question of what advantages P. aeruginosa PAO1 derives from its three pathways of in UFA synthesis. The wild type stain grew faster than the strain lacking FabA hence the DesA desaturase gives a growth advantage [10]. However increased FabAB expression should have the same effect. DesA gives a rapid alteration of membrane lipid properties by converting phospholipid saturated acyl chains to unsaturated acyl chains whereas increased flux through the FabAB pathway would be much slower. However, transcription of the desA gene is not significantly regulated following a shift in temperature from 37°C to 16°C [10]. Hence, DesA action must impart some other physiological advantage.

DesB does not confer a growth advantage to a strain lacking FabA and DesA and functions only in the presence of long chain saturated fatty acids [10]. Such conditions seem most likely to occur in infections of animals where P. aeruginosa is an opportunistic pathogen. A scenario that proposed a role for DesB in P. aeruginosa infections of the lungs of cystic fibrous patients has been described elsewhere [90]. Note that although desA genes are found in most pseudomonads, desB seems unique to P. aeruginosa [91]. However, the presence of a desA gene does not ensure DesA activity. For example, P. putida F1 encodes a DesA that fully replaces that of P. aeruginosa PAO1 but is only very weakly active in its native bacterium [91] such that P. putida F1 strains lacking FabAB fail to grow aerobically [91]. The probable cause of this failure is the lack of an efficient electron transport system to regenerate the diiron catalyst of P. putida F1 DesA required for catalytic turnover of the enzyme.

12. REGULATION of UFA SYNTHESIS

12.1. Regulation of the FabA-FabB pathway

The first transcriptional regulator of UFA synthesis discovered was E. coli FadR. FadR was discovered by Overath and coworkers as the repressor that controlled the E. coli pathway required for transport and β-oxidative degradation of environmental fatty acids [92]. The dependence of FabA expression on FadR was discovered by genetic interaction (synthetic lethality). Unexpectedly. repeated attempts to introduce a fadR mutation into a fabA(Ts) strain at 30°C indicated failed [93]. This strain construction should have been straightforward but failed until the genetic crosses were performed with UFA supplementation. Hence, in a wild type background at 30°C, the fabA(Ts) strain did not require UFA supplementation but introduction of a FadR mutation resulted in a UFA requirement at all growth temperatures [93]. Phospholipid fatty acid analysis of the fabA(Ts) and fadR strains showed that both strains had low UFA contents although they grew normally without UFA supplementation. However, the UFA synthesis deficiencies of the fabA(Ts) and fadR mutations were additive. That is, E. coli could tolerate normal expression of the weak FabA enzyme encoded by the fabA(Ts) gene or the decreased expression of a wild type FabA enzyme in the fadR strain. However, decreased expression of a weak enzyme failed to support growth resulting in UFA auxotrophy at all growth temperatures [93]. Subsequent studies showed that FadR binds the -35 region of the fabA and fabB promoters where it acts as a transcriptional activator [55, 94, 95]. FadR also binds the β-oxidation pathway promoters where it functions as a repressor [95, 96]. The FadR regulatory ligands are acyl-CoAs synthesized from exogenous fatty acids [95, 97]. The picture that emerged was one of a simple but sophisticated expediency [98]. In the presence of UFA supplementation, the cell can use the fatty acid as a carbon source via β-oxidation and by incorporation of oleic acid into phospholipids which spares endogenous UFA synthesis.

Recent work has shown that high level overexpression of FadR transcriptionally activates many E. coli fatty acid synthetic genes [99, 100]. FadR binds to sites upstream of each of these genes, albeit very weakly, consistent with the need for high level overexpression of FadR [99]. This interesting phenomenon is peripheral to UFA synthesis and has been reviewed elsewhere [101].

Regulation of UFA synthesis in E. coli became more complex when McCue and coworkers [102] discovered the FabR protein. Following extensive bioinformatic analysis these workers identified a palindromic sequence immediately upstream of the FadR binding sites of the fabA and fabB genes. They synthesized a double stranded oligonucleotide having this sequence, immobilized it on a column and used the column to purify a binding protein they called FabR from cell extracts [102]. FabR is a weak repressor. Deletion of the fabR gene in a wild type background gave a strain having a modestly increased (<2-fold) expression of both fabA and fabB [103]. Addition of oleate to a strain lacking FadR gave no change in fabA expression and a modest (<2-fold) repression of fabB expression [103]. Deletion of both fabR and fadR resulted in a strain with low fabA expression and essentially wild type expression of fabB [103]. Addition of oleate to this strain had no effect on expression of either gene. These data argued that the physiological role of FabR is to regulate fabB transcription.

In vitro studies of E. coli FabR are hampered by insolubility of the protein upon overexpression. Refolding of FabR from inclusion bodies gave a protein that failed to bind the fabA operator sequence [104]. This was a perplexing result because that binding was the interaction used by McCue and coworkers [102] to discover FabR. To avoid FabR insolubility a coupled transcription-translation system composed of purified components was used to produce FabR [103]. Electrophoretic mobility shift assays showed that binding of such FabR preparations to the fabA and fabB operator DNAs was increased by addition of either oleoyl-CoA or palmitoyl-CoA [103]. Palmitoloyl- and oleoyl-ACPs also gave increased binding but these thioesters may act as CoA mimics because blocking UFA synthesis (hence synthesis of acyl-ACP species) failed to derepress transcription of fabA or fabB [103].

Why does E. coli have both FabR and FadR? The function of FabR is to compensate for the weak FadR regulation of fabB expression in the presence of exogenous UFA (the converse argument that FadR functions to compensate for weak FabR regulation of fabA expression by UFA is also valid). The picture that emerges is that E. coli has yet to streamline co-regulation of fabA and fabB transcription and achieving coordinate regulation may be an act in progress. Indeed, coordinated transcription of fabA and fabB is hardwired in other bacteria such as the pseudomonads and related bacteria where the genes are in a fabA-fabB operon [91]. However even those bacteria show diversity. P. putida F1 fabB is transcribed by two promoters, one that transcribes both fabA and fabB and another that transcribes only fabB whereas P. aeruginosa PAO1 has only the fabA-fabB promoter [91].

12.2. Regulation of the P. aeruginosa DesB desaturase

A relative of E. coli FabR is P. aeruginosa PAO1 DesT (37% identical to E. coli FabR). DesT regulates transcription of desB encoding the desaturase that introduces double bonds into the CoA esters of saturated fatty acids obtained from the growth medium. The desT gene is located closely upstream of a putative oxidoreductase called desC cotranscribed with desB and is transcribed from the opposite DNA strand [10]. Deletion of desT results in a 14- to 20-fold increase in desB transcription indicating that DesT is a repressor [10, 105]. A similar increase was seen in a strain that lacked both FabA and DesA. Unlike FabR, DesT is readily expressed as a soluble protein in E. coli. The purified protein binds at two sites in the desCB promoter [105]. Electrophoretic mobility shift assays showed that palmitoyl-CoA released DesT from the promoter to give derepression whereas palmitoleoyl-CoA increased DesT binding 2- to 2.5-fold over the level seen in the absence of acyl-CoA [105]. In vivo the desB mRNA levels in cultures of the wild type strain supplemented with either oleate or stearate showed opposite behaviors. Addition of oleate repressed transcription about 10-fold whereas addition of stearate gave increased desB mRNA levels (~ 9-fold). Therefore, DesT can very effectively distinguish between acyl-CoAs formed from unsaturated (oleate-fed) or saturated (stearate-fed) exogenous acids [105]. Transcription of desT appears to be autoregulated. A DesT binding site is located within the desT promoter and fatty acid supplementation represses (oleate) and derepresses (stearate) desT transcription, but the effects are only ~2-fold.

The X-ray structures of DesT alone and with either oleoyl-CoA or palmitoyl-CoA demonstrate how the unsaturated and saturated acyl chains are distinguished [106]. DesT in the absence of acyl-CoA binds DNA but the oleoyl-CoA-DesT complex is about 2.5-fold more tightly bound to the desCB promoter. The oleoyl chain is bound in an L-shaped hydrophobic pocket where the kink of the cis double bond directs the methyl end into a phenylalanine-rich region. The palmitoyl-CoA structure binds DNA poorly [106]. The first seven carbons of the acyl chain of palmitoyl-CoA form interactions like those of the first seven carbons of oleoyl-CoA but the remaining nine carbons continue on a straight path to delve deeply into the hydrophobic core of DesT. Hence, it is the differential binding of the distal carbons to different DesT pockets that gives acyl chain specificity [106]. Binding of palmitoyl-CoA alters the DesT structure such that the two DNA- binding domains of the dimeric DesT move away from one another thereby disrupting cooperative binding of DesT to the desCB promoter. Acyl-CoA thioesters with chains shorter than C14 fail to dissociate DesT from the operator DNA [106] and do not derepress DesB transcription.

In contrast to the elegant regulation of desCB transcription, the only environmental change known to give increased transcription of desA is anaerobic conditions which give an ~10-fold increase in desA mRNA [10]. This increase in DesA seems likely be a device to scavenge limiting oxygen in the environment. Further investigation of DesA regulation is needed because it seems the more physiologically useful desaturase. This because it acts on acyl chains already present in the membrane phospholipids and does not require exogenous saturated fatty acids like DesB [10].

12.3. Regulation of B. subtilis Des.

The regulatory mechanisms described above function by classical repressor/activator proteins that directly respond to small molecule ligands. Repressors generally bind within the promoter and interfere with RNA polymerase binding or action. Activators generally bind at the upstream edge of what are otherwise poor promoters and help RNA polymerase to bind or function. Regulation of the B. subtilis Des desaturase proceeds by a very different mechanism called a two-component regulatory system. Such systems sense and respond to discrete environmental changes and perform signal transduction to cope with the environmental change [107]. The two components of the system are a histidine kinase, generally a homodimeric multi-span membrane protein, having a histidine phosphotransfer domain, an ATP binding site and a phosphatase domain. The histidine kinases are commonly called sensor kinases. The second component, called a response regulator, is often a transcription factor [107].

In response to a specific change in the environment, the sensor kinase catalyzes an ATP-dependent autophosphorylation reaction to phosphorylate a specific histidine residue in a C-terminal cytoplasmic domain of the kinase. The phosphorylated sensor kinase then transfers the phosphoryl group from the histidine to an aspartate residue on the cognate response regulator protein. When the response regulator is a transcription activator, phosphorylation typically results in the multimerization of the protein to allow efficient binding of promoter DNA. Promoter binding activates transcription to express the gene that encodes the protein that copes with the environmental change. When the environmental change has stabilized to shut off or dampen expression, the sensor kinase activates a phosphatase that removes the phosphoryl group from the response regulator protein resulting in decreased dimerization and hence less transcription.

In B. subtilis two genes located immediately downstream of the des gene on the genome encoded proteins having the conserved sequences of a histidine kinase (DesK) and a response regulator (DesR). Given the many precedents in the literature these proteins seemed likely to control induction of the des desaturase. Indeed, deletion of either desK or desR abolished temperature control [108]. In these deletion strains no activation of des transcription or synthesis of UFA was seen upon shift to low temperature. Hence, the desKR genes encode a two-component regulatory system [108] that controls des expression.

In vitro experiments showed that the DesKR is a paradigm two-component regulatory system. DesK autophosphorylates its conserved histidine residue and transfers the phosphoryl group to DesR (Fig. 7). Phosphorylated DesR dimerizes and binds two sequences in the des promoter activating des transcription [108]. These activities account for induction of Des desaturase expression at low temperatures. Immediately following temperature downshift to 20°C there is a burst (hyperinduction) of Δ5 UFA synthesis that decreases to a maintenance rate [83]. Transcriptional analyses demonstrate that the des gene is tightly regulated, only traces of the des transcript are present at 37°C [84, 108].

Fig. 7.

Model of the B. subtilis des regulatory pathway [125]. A. DesK is a dimeric protein with each monomer organized into five transmembrane helices. The two N-termini are exposed to solvent whereas the C-termini are in the cytosol where they may dimerize. The C-termini catalyze the histidine autokinase, the phosphoryl transferase and the phosphatase reactions. At high temperatures where the lipids are more ordered, DesK is predominantly in the autokinase state with the histidine residue phosphorylated. A shift-down in temperature increases the order of the membrane lipids. To cope with increased lipid ordering, DesK transfers a phosphoryl group from the histidine residue to a DesR aspartate residue. Two DesR-P dimers interact with the des promoter and with RNA polymerase (large ovals) resulting in activation of des transcription. B. The Δ5-Des desaturase is synthesized and converts saturated acyl chains of membrane phospholipids to unsaturated acyl chains. DesK senses the resulting decrease in ordered membrane lipids and now favors the phosphatase state that may undergo a conformational change of the C-termini. DesR is dephosphorylated resulting in dissociation of DesR from the des promoter and decreased transcription of des. Note that the dotted arrow in B implies only that the Δ5-UFA acyl chains somehow modulate the DesK activities. There is no evidence for a direct interaction of the Δ5-Des desaturase with DesK.

DesK kinase regulates both induction of the des desaturase upon shift to low temperature and the subsequent modulation of induction (Fig. 7). Upon temperature downshift DesK autophosphorylates its conserved histidine residue and transfers the phosphoryl group to the conserved DesR aspartate residue. DesR-phosphate binds the des promoter and activates transcription. The des mRNA is translated to give the desaturase which inserts Δ5 double bonds into saturated phospholipid acyl chains. When sufficient Δ5 unsaturated chain accumulation has adapted the cells to low temperature, DesK dephosphorylates most of the DesR and transcription decreases to a maintenance level. The hyperinduction and modulation phases occur within one doubling of cell number at 20°C [83]. Hence, the DesK kinase somehow senses the levels of Δ5 unsaturated phospholipid acyl chains in the membrane where it resides. One clue is that addition of exogeneous UFA after shift from 37°C to 25°C in a Δdes strain specifically inhibited the transcriptional activity of the des promoter. Although all UFAs tested gave inhibition, the C16Δ5 UFA was the most efficient [108] The saturated acid, palmitate, had no inhibitory effect. Whether these fatty acids were incorporated into membrane phospholipids or acted by partitioning into the membrane is unclear. Partition seems more likely since addition of UFA to anaerobic B. subtills cultures which grow and make phospholipids very slowly inhibited des gene expression [109]. These data suggested that free UFAs might somehow interact with the sensor kinase and activate its phosphatase.

The current model is that the decreased membrane thickness that accompanies increased fluid lipid may trigger changes in DesK activity. The laboratory of de Mendoza found that both the first (the N-terminal helix) and fifth (that linked to the cytosolic domain) DesK transmembrane helices were both required for temperature control and made a chimeric TM1−5 construct fused to the DesK autokinase domain that largely behaved like the intact DesK in temperature control [110]. Using this simplified system, these workers found that when reconstituted into liposomes the chimeric TM1−5 fusion protein was active. The protein had autokinase activity in liposomes and activated des transcription in vivo. These activities were greatly stimulated by shift to low temperature [110]. When the chimeric TM1−5 fusion protein was reconstituted in phosphatidylcholine liposomes having UFA chains of different lengths, DesK autokinase activity increased with chain length suggesting that a bilayer thickness ruler-like mechanism regulates the activation/deactivation of DesK [110]. Similar results were observed for the full-length DesK [111]. Phosphatidylcholine liposomes composed of a C16 saturated acyl chain gave very low DesK autokinase activity. A caveat is that membrane properties other than thickness are affected by temperature and acyl chain length, including permeability to water, membrane curvature and fluidity/rigidity. Currently biochemical and genetic experiments, plus X ray structures and molecular simulations argue that stabilization/destabilization of a two-helix coiled coil, which connects the TM sensory domain to its catalytic region is crucial to control the DesK signaling state [112, 113].

Nanodiscs, small (~10-nm) plugs of lipid bilayer rendered water-soluble by an annulus of “membrane scaffold proteins” [114] have been used to answer key questions in two component systems [115, 116]. The size of the membrane scaffold protein determines the size of the nanodiscs which can vary from 7 nm to 17 nm. Investigators can vary both the number of proteins inserted in the nanodisc membrane plug and the lipids used for the lipid bilayer. This system can be used to ask questions that liposome studies cannot. For example, how many copies of the chimeric TM1−5 protein are required for autokinase activity? Nanodiscs containing different amounts of inserted proteins can be separated by chromatography [114]. The lipid contents and degree of acyl chain unsaturation can be varied over a range of temperatures [114] and nanodisc components are commercially available. Two component systems that respond to bulk properties of membranes are difficult to approach experimentally and nanodisc technology offers the most defined system for investigation. This technology coupled with the remarkably increased resolution of cryo-EM which now approaches atomic resolution [117] may allow the mechanisms of histidine kinases to be deciphered in detail [118]. Note that the osmotically regulated EnvZ-OmpR system, one of the paradigm two component systems, has been studied for over 30 years and remains under active investigation [115].

12.4. Regulation of desaturase synthesis in cyanobacteria

Temperature regulation of desaturase synthesis has been extensively studied in cyanobacteria. Most of the investigations were done in Synechocystis sp. PCC 6803 where three of the four desaturases are induced by cold shock and are required for cellular function at low temperatures[78]. As seen in B. subtilis induction requires new mRNA and protein synthesis upon shift down in temperature. DesC, the only desaturase that is not induced by cold shock introduces a Δ9 double bond into an C18 saturated phospholipid acyl chain. At low temperatures the other three desaturases sequentially introduce double bonds to give C18:4 chains with double bonds at C6, C9, C12 and C15. To explore the desaturase regulation mechanism Murata and coworkers [119] inactivated 41 histidine kinase genes and found one, HK33, that is involved in the cold-inducible expression of the desB and desD genes, which encode the Δ3 and the Δ6 desaturases, but not the desA gene. Screening gave Rer, a response regulator that regulated desB but not desD. HK33 is general regulator of cold shock responses as well as salt stress and osmotic stress [78, 119]. HK33 autokinase activity and phosphate transfer have yet to be demonstrated. Note that cyanobacterial desaturase contents are very diverse. For example, Synechococcus elongatus PCC 7942 synthesizes only a Δ9 unsaturated acyl chain [120]. There are bacteria mainly from the deep oceans that use a polyketide pathway to synthesize polyunsaturated acids [121] . However, the literature on these pathways is limited probably because these bacteria often require low growth temperatures and high pressures to produce the polyunsaturated acids.

13. Conclusions

Recent discoveries of new mechanisms of UFA synthesis and regulation suggests that studies of diverse bacteria will give rise to additional mechanisms that may involve new chemistry, some of which may be of biotechnological importance. Although not discussed in this review, the cis double bonds of bacterial phospholipids are not sacrosanct; the bonds can be modified and altered. A very common modification is conversion of cis double bonds of membrane phospholipids to cis cyclopropanes. The methylene donor is S-adenosylmethionine and the conversion is catalyzed by cyclopropane fatty acid synthase, a soluble protein. The cyclopropane-modified phospholipids have similar physical properties to the UFA phospholipids from which they are derived, although the cyclopropane-modified membranes are less permeable and play a key role in resistance to acid stress. A recent review of cyclopropane fatty acid synthesis has appeared [5].

A much less common modification is conversion of cis-UFA into their trans isomers (trans-UFA) without changing the position of the double bond. Phospholipid acyl chains are the substrates of the cis-trans isomerization reaction which is catalyzed by a large soluble protein containing a covalent heme. This reaction has been found in diverse Gram-negative bacteria where it is reported to play roles in coping with temperature increase, presence of organic solvents and heavy metals (reviewed in [122]). However, a recent exception to Gram-negative bacteria is E. faecalis, a gram positive Firmicute, which has been shown to contain low levels of trans-UFA made by isomerization of cis-unsaturated phospholipid acyl chains [123]. E. faecalis lacks a homologue of the Gram-negative cis-trans isomerase.

Highlights.

• Function of unsaturated fatty acyl chains in phospholipid bilayers

• Discovery of unsaturated fatty acids and their modes of synthesis.

• Interactions of key enzymes with acylated acyl carrier proteins (ACPs).

• Regulation of the unsaturated fatty acid synthesis pathways The author declares no conflicts of interest.