Review article

Current understanding of fatty acid biosynthesis and the acyl carrier protein

David I. Chan, Hans J. Vogel

This article has a correction. Please see:


FA (fatty acid) synthesis represents a central, conserved process by which acyl chains are produced for utilization in a number of end-products such as biological membranes. Central to FA synthesis, the ACP (acyl carrier protein) represents the cofactor protein that covalently binds all fatty acyl intermediates via a phosphopantetheine linker during the synthesis process. FASs (FA synthases) can be divided into two classes, type I and II, which are primarily present in eukaryotes and bacteria/plants respectively. They are characterized by being composed of either large multifunctional polypeptides in the case of type I or consisting of discretely expressed mono-functional proteins in the type II system. Owing to this difference in architecture, the FAS system has been thought to be a good target for the discovery of novel antibacterial agents, as exemplified by the antituberculosis drug isoniazid. There have been considerable advances in this field in recent years, including the first high-resolution structural insights into the type I mega-synthases and their dynamic behaviour. Furthermore, the structural and dynamic properties of an increasing number of acyl-ACPs have been described, leading to an improved comprehension of this central carrier protein. In the present review we discuss the state of the understanding of FA synthesis with a focus on ACP. In particular, developments made over the past few years are highlighted.

  • acyl carrier protein
  • antibiotic
  • fatty acid biosynthesis
  • lipid
  • protein complex


De novo FA (fatty acid) synthesis represents a crucial pathway in all living organisms. Its products are mainly utilized in membrane biosynthesis, but are also needed in a number of other processes such as bacterial quorum sensing and protein post-translational modification. Owing to its ubiquitous nature and importance, FAS (FA synthase) systems have long been a subject for investigation and significant progress has been made [1]. For example, a virtually complete catalogue of high-resolution structures is available for each of the proteins involved in bacterial FA synthesis [2]. Nevertheless, there have been many important contributions to this field in recent years. Structures of the mega-synthases from mammalian and fungal sources have become available and researchers also continue to make strides in our understanding of FA synthesis regulation [37]. Interest in the FA synthesis pathways has remained high since it is believed that its machinery represents a promising target for the development of novel antibiotic agents [8]. Compounds such as triclosan and the drug isoniazid used against tuberculosis have fuelled the search for further medicinal agents that target the FAS systems. In addition to these commercially available compounds, a large number of natural products, such as platensimycin and cerulenin, have been noted to target FA synthesis [9,10], although these have yet to be developed to the point of being actively used pharmaceuticals. In addition, a recent study suggests that Gram-positive organisms may be able to completely circumvent de novo FA synthesis by acquiring FAs from their surroundings, illustrating the need for a better understanding of FASs [11]. Inhibition of FASs has also been implicated as an avenue for treatment and prevention of cancers as well as obesity, further raising interest in the study of FA metabolism [12,13].

Aside from supplying FAs for biomembrane synthesis, the FAS system also provides acyl chains for a host of other processes. ACP (acyl carrier protein), the central cofactor protein for FA synthesis, supplies acyl chains for lipid A and lipoic acid synthesis, as well as quorum sensing, bioluminescence and toxin activation [1418]. Furthermore, ACPs or PCPs (peptidyl carrier proteins) are also utilized in polyketide and non-ribosomal peptide synthesis, which produce important secondary metabolites such as the lipopeptide antibiotic daptomycin and the iron-carrying siderophore enterobactin [19,20]. In the present review, we provide an overview of the FA synthesis processes in bacteria and higher organisms. In particular, recent discoveries in this field will be highlighted with a special focus on the universal carrier protein ACP.


De novo fatty acid synthesis

FAS systems are grouped into two major classes, type I and II. The main difference between the two is the organization of the genes and the proteins needed for FA biosynthesis. In dissociated type II FAS systems, the proteins are all expressed as individual polypeptides from a series of separate genes [2]. These systems are found mostly in bacteria but also in specialized eukaryotic organelles such as mitochondria and plastids in plants. In contrast, type I FAS systems express large multienzyme complexes that carry all the proteins necessary for FA biosynthesis on one or two large polypeptide chains [21]. Type I FAS systems are divided into two sub-groups, type Ia and Ib. Type Ia systems are found in fungi as well as a few bacterial species and produce either α6β6 or α6 multimeric proteins respectively. The second sub-group, type Ib, is found in mammals and produce α2 homodimers, which each contain the entire suite of proteins needed for FA synthesis [22].

De novo biosynthesis of FAs occurs via a conserved set of reactions and, although the architecture of the FAS system can vary greatly between various organisms, the individual enzymatic steps are essentially the same. In fact, the FA biosynthetic machinery that is present in humans evolved from the bacterial systems, and the reactions carried out during the elongation cycle are identical [23]. To this end, FA synthesis is described as it occurs in Escherichia coli, the most studied system to date, and differences to the type I systems are noted later in the present review. Saturated FA synthesis in E. coli involves more than ten genes and proteins (Table 1). Central to this process is the ACP, a cofactor protein that covalently binds all fatty acyl intermediates. Before ACP can accept acyl chains, it must be activated by ACPS (ACP synthase), which attaches a phosphopantetheine group from CoA on to a serine residue of ACP (Ser36 in E. coli) in a conserved Asp-Ser-Leu motif (Figure 1). The resulting terminal sulfhydryl group of the phosphopantetheine arm is then used to bind all the FAs in a covalent high-energy thioester bond. The counterpart to ACPS is ACPH, which is a phosphodiesterase that cleaves acylated phosphopantetheine groups off of ACP [24]. This protein is only found in Gram-negative bacteria and the physiological purpose of this enzyme has not yet been established [25].

View this table:
Table 1 E. coli proteins involved in FA biosynthesis and their abbreviations
Figure 1 Conversion of apo- into holo-ACP by ACPS

CoA donates the phosphopantetheine arm, measuring approx. 18 Å, that is attached to the hydroxy group of the conserved serine residue in a Asp-Ser-Leu motif of ACP. The sulfhydryl in the prosthetic group is used to bind acyl chains via a thioester bond.

The FA biosynthesis process begins with the conversion of acetyl-CoA into malonyl-CoA using ACC (acetyl-CoA carboxylase), which adds an HCO3 group to the substrate (Figure 2; step 1). ACC consists of four proteins encoded for by accABCD and requires a biotin cofactor as well as ATP [26]. Malonyl-CoA-ACP transacylase (FabD) next transfers malonyl-CoA to ACP (Figure 2; step 2). The malonyl-ACP then combines with an acetyl-CoA group in the first condensation reaction, which is catalysed by β-oxoacyl synthase III (FabH) and forms β-oxobutyryl-ACP (Figure 2; step 3). This intermediate represents the first compound in the FA synthesis cycle, which performs the same four reactions in a cyclic manner to extend the saturated fatty acyl chain by two carbon units during each cycle. In this first instance, the acetyl-CoA unit (C2) is expanded to a butyryl group (C4). The following cycles give rise to a hexanoyl (C6) and then an octanoyl (C8) group, which would be continued until a final length of 16 or 18 carbons is reached. In recent years, an alternative to the FabD reaction has been described. Holo-ACP has been shown to be capable of self-acylating and trans-acylating other holo-ACP proteins [27,28]. This function of ACP was first described in PKS (polyketide synthase) systems (see below) and subsequently FAS proteins have been demonstrated to possess one or both of these activities as well [2730]. It is unclear at this time what the physiological function of these reactions is; however, it has been shown that bacteria with a temperature-sensitive version of FabD may be rescued if the ACP that is present possesses both self- and trans-acylation activities [28].

Figure 2 Overview of the bacterial FA synthesis mechanism

After initiation of the acyl chain synthesis (reaction 3), the synthetic cycle is repeated multiple times until saturated C16 or C18 acyl-ACP is diverted for utilization in membrane biosynthesis. For the description of the reactions please refer to the text.

The first reaction in the FA synthesis cycle is a NADPH-dependent reduction of the β-oxoacyl-ACP into β-hydroxyacyl-ACP, which is catalysed by β-oxoacyl reductase (FabG) (Figure 2; step 4). The hydroxy group is removed by one of two β-hydroxyacyl dehydratases (FabA or FabZ), which converts the chain into a trans-2-enoyl group (Figure 2; step 5). The double bond is then reduced in an NADH-dependent reaction by an enoyl-reductase (FabI). This yields acyl-ACP bound by a saturated acyl chain extended by two carbon units compared with the start of the cycle (Figure 2; step 6). In the first round just described, acetyl-CoA represents the start of the cycle. At this point the cycle begins again through the condensation reaction of acyl-ACP with another malonyl-ACP group, generated as described above by ACC. The condensation reactions other than the first step involving acetyl-CoA are catalysed by either β-oxoacyl synthases I or II (FabB or FabF) (Figure 2; step 7). Notably, during each of the condensation reactions, the acyl chain is detached from ACP and bound in a thioester bond by a cysteine residue in the active sites of FabH, FabB or FabF (Figure 3) [2]. An extender malonyl-ACP then enters the active site and the acyl chain is added to the tip of the malonyl unit, which loses a CO2 group in the process. Therefore the acyl chain is constructed from the ‘inside out’ as the additional carbon groups are added to the base of the acyl chain rather than the tip (Figure 3). The cycle is repeated until the acyl chain reaches 16–18 carbon groups in length, at which point the vast majority of acyl-ACPs are utilized in membrane biosynthesis.

Figure 3 Extension of fatty acyl chains

Fatty acyl chain extension occurs via the condensation reaction, in which ACP transfers its acyl chain on to one of the condensing enzymes FabB or FabF (step 1). The holo-ACP from this step is lost (step 2) and a malonyl-ACP enters, carrying the acyl chain extender unit. The malonyl group first loses a carbon dioxide (blue) and the growing acyl chain (red) is subsequently added onto the remaining two carbon groups attached to ACP (step 3). The resulting β-oxoacyl-ACP represents the first intermediate of a new fatty acyl elongation cycle. ‘R’ represents a saturated acyl chain of length 3+2n carbons. FabH catalyses only the very first condensation reaction that forms β-oxobutyryl-ACP. In that case, the acetyl starter group from acetyl-CoA is bound by FabH and condensed with malonyl-ACP.

In bacteria, the process of membrane biosynthesis takes place in two different ways, the most common being the PlsX/PlsY pathway [31]. First, the peripheral membrane protein PlsX cleaves acyl-ACP to form an acyl-phosphate. This is then attached on to glycerol-3-phosphate to form 1-acyl-glycerol-3-phosphate (LPA; lysophosphatidic acid) by the acyltransferase membrane protein PlsY [31]. Another acyltransferase, PlsC, that is expressed in all bacteria, then adds a second acyl chain to the 2-position of LPA to form PA (phospatidic acid) [32]. PA represents the fundamental building block from which the phospholipids are derived, giving rise to phosphatidylserine, phosphatidylethanolamine and phosphatidylglycerol [33]. In a second, not as widely distributed pathway, the acyltransferase PlsB utilizes acyl-ACP and transfers the acyl chain to glycerol-3-phosphate directly, followed by the same reaction catalysed by PlsC to generate PA [34]. The PlsB pathway has the advantage that it can utilize both acyl-ACP and acyl-CoA, which is the form of FAs from exogenous sources [34]. In E. coli, PlsC also has the ability to function with either substrate, but this is not the case in all organisms [31,35].

Biosynthesis of unsaturated fatty acids

The pathway described above details the production of SFAs (saturated FAs). The reaction scheme is slightly different for the production of UFAs (unsaturated FAs). The first step in the production of these acyl chains is catalysed by the fabA gene product β-hydroxyacyl dehydratase (FabA), which catalyses both a dehydration as well as an isomerization reaction. The isomerase function is exclusively performed on C10 substrates. FabA converts β-hydroxydecanoyl-ACP into trans-2-decenoyl-ACP and subsequently isomerizes this fatty acyl intermediate to cis-3-decenoyl-ACP (Figure 4) [36,37]. Subsequently, the cis-3-decenoyl group undergoes a condensation reaction with another malonyl-ACP group, yielding cis-5-dodecenoyl-ACP, since the two carbon groups are introduced closest to the carbonyl carbon (Figures 3 and 4). This reaction is catalysed by the second essential enzyme for UFA synthesis, β-oxoacyl synthase I or FabB [38]. Thus in E. coli the FabB and FabI enzymes compete for cis-3- and trans-2-decenoyl-ACP respectively (Figure 4), and the relative abundance of these two enzymes determines the ratio of saturated to unsaturated FAs in biological membranes. In the following acyl chain elongation cycles, the same enzymes are used as in SFA synthesis, with the exception of FabA, as only FabZ (the other β-hydroxyacyl dehydratase; Figure 2) is capable of utilizing cis-acyl chains [36]. Owing to the nature of the FA synthetic cycle, there are only a small number of specific UFAs produced in E. coli. The inside-out carbon group additions mean the cis double bond essentially gets ‘pushed’ away from the thioester link by two carbon units per cycle (Figure 2). Therefore the vast majority of UFAs are cis-9-hexadecenoic acid (palmitoleic acid) and cis-11-octadecenoic acid (vaccenic acid) [39,40].

Figure 4 UFA synthesis pathways in bacteria

(A) Two UFA synthesis pathways are shown for E. coli (left) and S. pneumoniae (right). E. coli utilizes FabA, which has a dual dehydratase and isomerase activity, whereas S. pneumoniae has a dedicated cis/trans isomerase (FabM), both of which are specific for C10 acyl chains. (B) The structure of the resulting cis-3-decenoyl-ACP.

The E. coli system described above represents the most extensively studied bacterial system. However, variations to this UFA synthesis mechanism exist and are notably different in Gram-positive bacteria. For example, Streptococcus pneumoniae does not possess a FabA enzyme, but contains a specific cis/trans isomerase, termed FabM, which is responsible for the conversion of trans-2- into cis-3-decenoyl-ACP [41]. Furthermore, additional condensation reactions are not dependent on FabB, but rather FabF extends both the saturated and unsaturated FAs. Enterococcus faecalis possesses a system more similar to the E. coli pathway, where FabN replaces FabA and FabO replaces FabB [42]. Interestingly, FabN in terms of its amino acid sequence is more closely related to FabZ, despite its functional relationship to FabA [42]. The mechanisms in other organisms, such as Clostridium acetobutylicium, are still being investigated and it appears a number of enzymes involved in UFA synthesis have yet to emerge [43].

Some bacteria also possess desaturases that can introduce unsaturated bonds into acyl chains after FA synthesis. For example, Bacillus subtilis and Pseudomonas aeruginosa both express a membrane-bound enzyme known as DesA that introduces cis double bonds at either the 5- or 9- position respectively [44,45]. Ps. aeruginosa expresses a second desaturase known as DesB, which also introduces cis double bonds at the 9-position. However, DesB is active on acyl-CoA taken up from exogenous sources rather than phospholipids that are part of the membrane [45]. Unsaturated cis-acyl chains may be modified further in the membrane via other mechanisms, including cis/trans isomerization and the introduction of cyclopropane FAs [46,47]. These adjustments help membranes adapt to environmental factors such as acid stress in the case of cyclopropane FAs [48]. In B. subtilis, a two-component system regulates the expression of DesA via the membrane-bound thermosensor DesK and the response regulator DesR [44]. At temperatures lower than 30 °C, DesK phosphorylates DesR, which activates the transcription of DesA and leads to lipid desaturation. Recently, the structure of the catalytic portion of DesK was determined in a number of conformations, showing major rearrangements of its four helix bundle [49]. This explains how DesK alternates between its different states in response to changes in membrane fluidity.

Type I fatty acid synthase systems

The type I FASs are believed to have evolved from a dissociated system, much like the type II systems found in bacteria and plants [23]. Various evolutionary processes such as gene duplication and gene fusion probably led to the emergence of the multifunctional systems now found in modern type I FASs [23]. Despite obvious differences in the overall organization of the two FAS systems, the enzymatic reactions and mechanism of de novo FA synthesis are essentially identical [22]. For example, exactly the same intermediates and reactions are present in the elongation cycle. However, some differences do occur in the initiation and termination processes. The mammalian type Ib systems contain a malonyl-/acetyl-CoA ACP transacylase (FabD) that loads both the initiator and extender units on to ACP. This contrasts the bacterial systems, which utilize FabH and FabD for these processes (Figure 2). Fungal type Ia systems have a discrete acetyl transferase that is specific for acetyl-CoA in the transfer to ACP. The malonyl-CoA extender unit is transferred on to ACP via a malonyl/palmitoyl transferase which is also involved in the FA synthesis termination process, as it transfers the mature palmitoyl group from ACP back on to CoA for use in various cellular processes. On the other hand, type Ib systems utilize a dedicated thioesterase enzyme that cleaves acyl chains off of ACP and releases free FAs. Type II systems instead use enzymes, such as the aforementioned PlsB and PlsX, to transfer acyl chains directly from acyl-ACP for integration into the membrane [31].

Mitochondrial fatty acid synthase

In addition to the cytosolic FAS systems that are present in eukaryotic organisms, mitochondria harbour a further set for de novo FA synthesis. The system found in mitochondria differs from the cytosolic FAS as it is of the type II dissociated organization and many of the proteins are highly homologous with their bacterial counterparts [50]. Human and yeast mitochondrial genes equivalent to the bacterial type II proteins continue to be discovered and the structures of a number of mitochondrial proteins have been solved [5154]. Although both systems are present in eukaryotic cells, the vast majority of the FAs synthesized are produced by the cytosolic machinery. The mitochondrial FAS synthesizes octanoic acid, which serves as the precursor for lipoic acid [55]. There is also evidence that the mitochondrial system can produce longer FAs, however, it is presently unclear what these acyl chains are used for [55,56]. The exception is β-hydroxytetradecanoyl-ACP, which is known to be part of the respiratory complex I in mammalian mitochondria [57]. Although the role of the mitochondrial FAS system is not entirely clear, it has been shown that an impaired system leads to respiratory incompetence in yeast [58], whereas in Trypanosoma brucei malfunctioning mitochondrial FAS leads to abnormal morphology [59], and in mammalian tissues to cell death [60]. Recently, evidence has been presented that this FAS system is implicated in monitoring the metabolic condition of the cell [61]. Decreasing the amount of pyruvate available in mitochondria will limit the acetyl-CoA pool for FAS synthesis and thereby lower the rate of lipoic acid formation. Since pyruvate dehydrogenase, the committing enzyme for TCA (tricarboxylic acid) cycle activity, is strictly dependent on lipoic acid, this presents a positive-feedback loop involving FAS in mitochondria. Furthermore, it has been suggested that a mitochondrial FAS product is implicated in the RNase P maturation process, which is required for tRNA processing [61]. As such, inhibition of the FAS will lead to decreased protein synthesis in mitochondria. A link between RNase P and 3-hydroxyacyl dehydratase was also found in humans, as they are expressed in a bicistronic fashion [62]. That these two proteins, which are seemingly only distantly related in function, are transcribed together further substantiates the notion that mitochondrial FAS plays a role in sensing the metabolic state of a cell.


Regulation through product inhibition

A great deal of energy in the form of ATP is committed to de novo synthesis of fatty acyl chains and it is therefore of paramount importance to regulate this process. The primary pathway for the regulation of FA synthesis in E. coli is through feedback inhibition by long-chain acyl-ACPs. This was initially demonstrated through the overexpression of thioesterases that cleave acyl-ACPs, leading to excessive FA synthesis [63]. Long-chain acyl-ACPs affect the pathway at three discrete steps by inhibiting ACC, FabH and FabI (Table 2) [64]. Inhibition of ACC prevents the creation of malonyl-CoA, which is the extender unit needed for the elongation of the growing acyl chains (Figure 2; step 1) [65]. FabH catalyses the formation of the starter unit β-oxobutyryl-ACP and thus its inhibition prevents the initiation of further acyl chain formation cycles (Figure 2; step 3) [66]. Finally, FabI catalyses the reduction of enoyl-ACP, which is critical for the completion of the acyl chain elongation cycle due to the reversible nature of the preceding dehydration reaction (Figure 2; step 6) [67].

View this table:
Table 2 Summary of bacterial FAS regulation

Since the vast majority of the synthesized FAs are utilized in biomembranes, the acyl-ACP concentration is highly dependent on the level of acyl chain integration into membranes, which is catalysed in part by the enzyme PlsB. PlsB activity is in turn inhibited by guanosine tetraphosphate (ppGpp), which is a general stress response metabolite in E. coli and related bacteria as well as in plants [6870]. When the bacterium undergoes a ppGpp-mediated stress response, PlsB activity is decreased, which leads to a build-up of long-chain acyl-ACPs and consequently to the attenuation of FA synthesis (Table 2). On the other hand, it has been shown that ACP interacts and alters the activity of the enzyme SpoT, which is responsible for both the synthesis and breakdown of ppGpp [71,72]. This means that ACP itself can act as a sensor of disrupted FA metabolism, leading to a build-up of ppGpp and consequently to a stress response in E. coli [71].

Transcriptional level regulation of fatty acid synthesis

E. coli as well as other Gram-negative organisms also regulate FA synthesis on a transcriptional level via the FabR and FadR proteins (Table 2). Both proteins are involved in the regulation of the fabA and fabB genes, which code for the essential enzymes required for UFA synthesis. FadR is a transcriptional activator protein that is bound at position −40 to the fabA and fabB genes, whereas the protein FabR is a repressor that can bind in the promoter region. FadR binds to both saturated and unsaturated acyl-CoA groups, which causes the protein to release from the DNA and thereby inhibit fabA and fabB transcription [73]. FadR actually plays a dual role in the modulation of FA levels, as it also acts as a repressor of β-oxidation [74]. FabR, on the other hand, responds to both acyl-CoA and acyl-ACP, but unsaturated and saturated ligands elicit different responses [7]. SFAs decrease the DNA affinity of FabR, which leads to its release from the promoter and subsequent transcription of fabA and fabB, whereas unsaturated groups increase the affinity resulting in the repression of these genes [7].

Although this is the only known transcription-level regulation of FA synthesis in E. coli, there are additional mechanisms in Gram-positive species (Table 2). For example, in Staphyloccocus aureus and B. subtilis, the protein FapR is a transcription factor that represses all FAS genes [75]. FapR is regulated by malonyl-CoA, which leads to the release of the inhibitor and transcription of the FAS genes [76]. Moreover, malonyl-ACP was recently shown to play a similar role as malonyl-CoA in regulating FapR in B. subtilis [77]. Interestingly, this process does not seem to involve protein–protein interactions but requires the exposed malonyl-phosphopantetheine group.

The protein FabT was identified to regulate the entire FAS gene cluster with the exception of FabM in S. pneumoniae, which is involved in UFA synthesis (see above) [78]. It was subsequently discovered that FabT is a repressor that responds to acyl-ACP levels [6]. It was shown to be most sensitive to vaccenyl-ACP but responds to other long chain acyl-ACPs as well in the presence of DNA [6].

Additional means of regulation of FA biosynthesis may yet be uncovered. Recently, malonyl-ACP has been found to copurify and crystallize in a complex with the cytoplasmic STAS domain of YchM, an E. coli inner membrane protein that is involved in the transport of bicarbonate. This observation directly links regulation of the transport of the substrate bicarbonate to the activity of acetyl-CoA carboxylase (T. Moraes and R. Reithmeier, personal communication). Clearly, the role of this protein–protein complex in regulating FA biosynthesis requires further study.



ACP is the carrier protein that represents the central player responsible for attachment and presentation of all acyl chain intermediates to the enzymes of FA biosynthesis. In E. coli, ACP is highly abundant, comprising approx. 0.25% of all soluble proteins and it represents one of four major protein–protein interaction hubs, the others being DNA and RNA polymerases as well as ribosome-associated proteins [79,80]. Bacterial ACP is the protein encoded by the acpP gene. It is expressed as apo-ACP and requires activation to holo-ACP by ACPS through the attachment of a phosphopantetheine moiety from CoA (Figure 1). The prosthetic arm is attached at a specific serine residue in a conserved Asp-Ser-Leu motif. Bacterial ACPs are small, acidic proteins, the prototypical version being E. coli ACP, which has a molecular mass of 9 kDa and a pI of 4.1 [79]. It has a very high concentration of anionic residues and only few cationic residues, which number 20 and 6 respectively, out of a total of 77 amino acids for the E. coli protein. The high density of anionic residues is responsible for the unusual migration pattern of ACP on SDS/PAGE systems, where it appears as a 20 kDa protein. In addition, ACP only has 22 hydrophobic amino acids, giving it a high ratio of charged to hydrophobic residues. This amino acid distribution is highly reminiscent of natively unfolded proteins and ACP is known to be of rather marginal stability [81,82]. E. coli ACP displays a large hydrophobic expansion at an elevated pH, which can be visualized by a decreased electrophoretic mobility in native PAGEs run at alkaline pH [83]. Although E. coli ACP is folded at low ionic strength and pH 7, Vibrio harveyi and Helicobacter pylori ACPs are unfolded or partially folded respectively, under these conditions [84,85]. ACP is also a highly dynamic protein and E. coli, Plasmodium falciparum and spinach ACPs have even been noted to exist in a conformational equilibrium between two states [8688]. Deuterium exchange and 15N-heteronuclear NMR dynamics studies reveal that loop I is among the most flexible regions of ACP along with helices II and III [85,89,90]. Recent 15N relaxation dispersion NMR studies reveal that spinach ACP displays μs and ms time-scale motions, further complicating the nature of ACP dynamics [91]. Conformational flexibility is thought to be important for ACP's role in substrate delivery, since the acyl chain must be extracted from the protein and presented to the enzymes of FA synthesis. Although the flexibility of ACP is believed to aid in acyl chain insertion and extrusion and is also important for enzyme interactions, unfolding of the protein is not required for these processes. This was demonstrated utilizing a functional, cyclized version of ACP, which stabilizes the folded conformation and thereby prevents complete unfolding [92].

Researchers began describing the structure of ACP over 40 years ago and the high-resolution solution structure of the E. coli protein was first published in the late 1980s by the Prestegard group [1,93,94]. Since then, a number of additional ACP structures have been published from various organisms (Table 3, Figure 5). Each of these structures falls in line with a consensus fold consisting of a four α-helix bundle. Three of these helices run approximately parallel with each other, whereas helix III is much shorter and lies almost perpendicular to the axes of the major helices (Figure 5). In E. coli ACP, helices I–IV run from residues 3–15, 36–49, 56–59 and 65–74 respectively, and helix I is antiparallel with helices II and IV [95]. The helices form a scaffold around a hydrophobic pocket in the centre of ACP that has long been thought to harbour the growing acyl chain during FA biosynthesis [96,97]. This was confirmed when structures of acyl-ACP from spinach and E. coli ACP became available and it was shown that the opening to the pocket lies in the gap between helices II and III (Figure 5) [88,95,98]. The electrostatic surface of E. coli ACP is highly anionic and helix II in particular possesses a segment of conserved negatively charged amino acids (Figure 6) [99]. The structure of ACP is remarkably conserved throughout many organisms even at low sequence identity. For example, E. coli and rat ACPs only share 22.5% sequence identity, but the overall global fold is still largely the same [100,101]. Furthermore, when expressed in E. coli, rat ACP is partially capable of replacing the endogenous protein where it can be utilized by the host FAS system [102].

View this table:
Table 3 High-resolution structures of FAS ACPs solved to date
Figure 5 Sample high-resolution type II acyl-ACP structures

E. coli butyryl- and decanoyl-ACPs (A and B respectively), as well as octadecanoyl-ACP from spinach (C), show the highly conserved ACP structural motif consisting of a four α-helix bundle. The spinach ACP NMR structures are an overlay of the five lowest energy conformations and illustrate that the octadecanoyl chain is too large to fit into the hydrophobic pocket in its entirety. The acyl chains and prosthetic groups are coloured according to atom type. A three-dimensional interactive structure of (B) is available online at

Figure 6 Electrostatic properties of ACP

Anionic and cationic residues are marked on E. coli ACP in red and blue respectively in (A) and (B). (B) The recognition helix, helix II, is highlighted, which contains a number of conserved anionic residues as well as two hydrophobic residues (white). (C) Electrostatic surface representations of ACP, the left panel being orientated in the same way as (A) and the right panel rotated by 180 °, displaying the highly anionic face.

Current understanding of acyl-ACPs

High-resolution acyl-ACP structures and MD (molecular dynamics) simulations have greatly increased our understanding of how various acyl chains are bound and accommodated by ACP. E. coli acyl-ACP structures have been published with butyryl, hexanoyl, heptanoyl, octanoyl and decanoyl chains attached, and decanoyl- and octadecanoyl-ACP structures from spinach have also been reported [88,95,98]. The spinach acyl-ACP structures were determined by NMR spectroscopy and an ambiguous NOE (nuclear Overhauser effect) docking protocol was used to determine the co-ordinates of the acyl chain, whereas the E. coli structures were obtained from X-ray crystallography. The different acyl-ACP forms from E. coli are highly similar [Cα RMSD (root mean square deviation) ranging from 0.2 to 0.94 Å (1 Å=0.1 nm) for decanoyl-ACP to the other forms], whereas the spinach structures display an RMSD of 1.2 Å from the E. coli protein (Figure 5). The structures reveal that all the acyl chains enter the hydrophobic cavity of ACP between helices II and III and are positioned along helix II into the protein cavity. MD simulations of E. coli acyl-ACP captured the spontaneous transition process into the hydrophobic pocket of the protein starting from a completely solvent-exposed acyl chain conformation (see Supplementary Movie S1 at [103]. As the acyl chain enters the cavity, the phosphopantetheine linker arranges itself at the cavity entrance and usually hydrogen bonds with Asp35, Thr38 and Glu60. Inside the protein, the simulations demonstrate that the acyl chains can enter a second sub-pocket that is directed towards helix I of the protein, which was subsequently documented by NMR analysis as well [103,104]. The opening of this cavity is mediated by two leucine residues which either block or allow entry into the second sub-cavity via their χ1 dihedral angles (see Supplementary Movie S2 at [103]. Harbouring an acyl chain inside the cavity of ACP expands its volume with increasingly longer acyl chains. This has been described as an overall swelling, as helices II and IV are observed to move further apart, as well as helix III [88,104,105]. In spinach ACP, this increases the cavity volume from 157 to 228 Å3 going from decanoyl- to octadecanoyl-ACP [88]. The shorter acyl chains attached to ACP up to octanoic or decanoic acid are buried inside the pocket and the hydrocarbon groups are not readily solvent-accessible [88,95,98,103]. This is in agreement with earlier experiments such as octyl-sepharose binding studies and native-PAGE performed at pH 9.5, which suggest that acyl chains up to C8 are fully bound inside the hydrophobic pocket, whereas longer acyl groups are partially solvent-exposed [79,106]. We have also conducted MD simulations of the most prominent E. coli UFAs, cis-palmitoleic and cis-vaccenic acids, bound to ACP. Preliminary results indicate that these groups are more rigid, but nonetheless utilize the same hydrophobic pocket as all the other acyl chains (D.I. Chan, D.P. Tieleman and H.J. Vogel, unpublished work).

Although there is now substantial evidence to understand the behaviour of saturated acyl-ACP, only little is known on ACP bound to the elongation cycle intermediates. It is conceivable that the entire acyl chain does not have to be extracted from the protein during the β-carbon processing reactions, since the reactions take place at the β-carbon. However, high resolution structures of FabA and FabI bound to substrate analogues consistently show interactions between the acyl chains and the active site of the enzymes [37,107]. Furthermore, ACP in the crystal structure complex with FabI is positioned in a manner that the acyl chain would have to be extruded entirely for the β-carbon to reach the active site of FabI [108]. In that study, the authors also performed MD simulations that position the tip of the acyl chain away from ACP and into the active site of FabI [108]. As such, it is likely that in the β-carbon processing reactions, the acyl chain is completely extruded from the hydrophobic pocket of ACP, similar to what takes place during the condensation reactions. A recent MD study of β-oxoacyl-, β-hydroxyacyl- and trans-2-enoyl-ACPs (see cycle in Figure 2) conducted by our group provides additional information on ACP and the elongation cycle intermediates [105]. These intermediates were also observed to spontaneously enter the hydrophobic pocket of ACP and sample both sub-pockets within the protein (see Supplementary Movie S2). Furthermore, it was observed that the β-oxone and β-hydroxy groups affect the positioning of short-chain acyl groups, as these polar moieties prefer not to reside within the hydrophobic binding pocket inside the protein [105]. That study also managed to capture the acyl chain extraction out of ACP in the case of the β-oxo- and β-hydroxybutyryl-groups [105]. These short acyl chains possess very little hydrophobic character and so were not constrained inside the hydrophobic pocket as stringently. Similarly, a recent crystal structure of B. subtilis malonyl-ACP reveals that this moiety is not housed inside the hydrophobic pocket [77]. As the acyl chains are extracted in the MD simulations, the structure of ACP remains unchanged and it is essentially the reverse of the acyl chain entry process [105]. A steered MD study of P. falciparum β-hydroxydecanoyl-ACP suggests a similar acyl chain extrusion pathway, however, helix III in those simulations swings away from the protein, unlike what was observed in the unbiased simulations [109]. Whether this reflects a real difference in the behaviour of short- and long-chain acyl groups will have to be evaluated in further studies.

Type I ACPs

Unlike the wide range of type II ACP structures that are currently available, only the rat, human and Saccharomyces cerevisiae type I ACP structures have been characterized to date (Table 3). Rat and human apo-ACPs were cloned and expressed as discrete proteins, separated from the large multi-enzyme complexes that they are usually a part of. The rat ACP structure was determined by solution NMR, whereas the human protein was crystallized in a complex with ACPS [100,101,110]. On the other hand, the structure of yeast holo-ACP was determined together as part of the entire 2.6 MDa FAS complex using X-ray crystallography [5,111,112]. The structures display similar overall folds to the type II ACPs and also possess four α-helices (Figure 7), despite relatively low amino acid sequence identities between the two types (rat compared with E. coli ACP: 22.5% identity). Intriguingly, the yeast protein possesses a second domain in addition to the usual ACP fold, which consists of four additional helices extending from the C-terminus, approximately doubling the size of the protein (18 kDa) (Figure 7) [5,111]. NMR experiments conducted on the isolated yeast ACP domain support the conformation observed in the X-ray mega-complex [113]. The rationale for this additional domain in yeast ACP is currently unknown. In two of the yeast FAS complex structures, ACP is positioned in contact with the oxoacyl synthase domain and the phosphopantetheine group is directed away from the protein into the active site of the enzyme [5,111]. The interactions between ACP and the oxoacyl synthase unit occur both between the typical and the extended domains of yeast ACP. Notably, helix II, similar to what is seen in type II complexes, is implicated as part of the interaction interface (see below) [111,114]. Helix II is also involved in the complex between human ACP and ACPS [110]. However, Bunkoczi et al. [110] indicated that the most important protein–protein interactions in this case occur between hydrophobic regions and are based on shape complementarity. It is also noted that human ACP is far less negatively charged compared with the type II E. coli protein (overall charges of −5 and −14 respectively), which supports an increased role for hydrophobic interactions in the type I systems.

Figure 7 High-resolution structures of type I ACPs

Human (A), yeast (B) and rat (C) type I ACPs. Overall, the secondary structure is conserved with type II proteins but other differences are present. ACP from yeast possesses a second domain (blue) and the entrance to the hydrophobic pocket appears to be closed off in type I ACPs (compare red circles in B and C with Figure 5B). The opening is closed off by loop I, which is orientated more towards helix III (thick arrow, C) and by Arg2150 (spheres, C). Ser36 is highlighted in stick representation.

It has been postulated that type I ACPs do not harbour their bound acyl chain inside the hydrophobic pocket, which would represent a major difference between the type I and II proteins [101]. The C-terminal region of loop I in type I ACPs is observed to rest further towards helix III compared with type II proteins, in which this portion of the loop is directed closer to helix I (compare Figure 5 with Figure 7). This arrangement is held in place by a hydrophobic triad composed of Leu2144, Leu2149 and Val2174 (residues 29, 34 and 59 in E. coli numbering) which is not conserved in type II ACPs [101]. This hypothesis was derived from the observation that only very small NMR chemical shift perturbations are seen upon conversion of holo-ACP into hexanoyl- or hexadecanoyl-ACPs in rat [101]. Yeast type I ACP is in apparent contrast to this hypothesis, since it does not possess the same hydrophobic triad and a substantial hydrophobic pocket was detected [5]. Nevertheless, the loop I arrangement in yeast ACP is still equivalent to that seen in the rat protein, further complicating the interpretation of these data [101]. We conducted MD simulations of isolated type I rat ACP similar to the simulations performed with acyl-ACP from E. coli [103]. Unlike the E. coli protein, neither rat octanoyl-ACPs nor butyryl-ACPs displayed a transition of the acyl chain inside the hydrophobic pocket during 3×20 ns of simulations (see Supplementary Movie S3 at Interestingly, the acyl chains appear to have a propensity to associate with loop I of ACP, similar to simulations of the bacterial type II ACPs [105]. The opening to the pocket is observed to narrow further during the simulations of rat ACP, and Arg2150 (Arg60 in E. coli numbering) frequently hydrogen bonds with the phosphopantetheine linker and thereby somewhat blocks the entrance to the cavity. This arginine residue is well conserved among type I ACPs and represents a hydrogen bond donor group in the binding pocket entrance, which is absent in type II ACPs. Thus these MD simulations support the hypothesis that type I ACPs do not protect acyl chains inside a hydrophobic pocket, and it remains to be seen what role Arg2150 may play. Recently, data was also presented on type I PKS ACPs (see below), indicating that these proteins do not bind their acyl chains inside the hydrophobic pocket either, further supporting this theory for type I FAS ACPs [115].

Although additional data is needed to evaluate the mechanism of acyl chain binding in type I ACPs, it is certainly conceivable that the acyl-chain in these systems need not be shielded inside the ACP pocket. In type II systems, ACP is expressed as an individual protein, whereas in type I FAS, ACP is attached to much larger 540 kDa or 2.6 MDa protein complexes. If type II ACPs expose their bound acyl chains to the solvent, it may very well be targeted to the cell membrane, much like myristoylated or palmitoylated proteins in mammalian cells (e.g. recoverin) [116]. However, in type I FASs, because ACP is covalently bound to the other protein subunits, it is unlikely that a single acyl chain would target the whole complex to the membrane. In fact, in the fungal type I systems, ACP is located inside the scaffold of the overall barrel shape (see below) and therefore would not be accessible to the cell membrane.

ACP–protein interactions

For each of the reactions that occurs in the type II FA synthesis cycle, the substrate that is covalently bound to ACP and hidden inside its pocket must be extracted and placed into the active site of the partner enzyme. Because the substrate is not readily accessible inside the hydrophobic pocket of ACP, the partner enzyme must bind the carrier protein and then somehow access the bound cargo. Although ACP interacts with a large number of proteins during FA synthesis, no common ACP-binding sequence has been noted, unlike, for example, the protein kinases. Nevertheless, information has been gained on ACP–protein interactions from a number of structural, mutagenesis, NMR chemical shift perturbation and bioinformatic studies. Sequence alignment of ACPs from various organisms indicates that many anionic residues in helix II are conserved and are therefore believed to be important for ACP–enzyme interactions (Figure 6) [99]. Helix II was subsequently dubbed the ‘recognition’ helix [99] and it has been a recurring theme in virtually all studies of type II ACP–enzyme interactions. It has been shown that ACPs from a broad range of organisms are capable of restoring growth in an E. coli acpP temperature-sensitive strain, which is at least partially related to the similarity of helix II in ACPs from different organisms [117]. The two ACP–enzyme complex structures that are available support the important role of helix II (Figure 8) [108,118]. In both the ACP–ACPS and ACP–FabI structures, helix II represents the central interaction site. Furthermore, chemical-shift perturbation and docking studies involving FabG and FabH respectively, heavily implicate helix II of ACP [119,120]. Mutations of various portions of helix II suggest that different enzymes recognize specific regions of the helix. For example, interactions with ACPS are highly affected by mutations to the N-terminal region of helix II, whereas LpxA, which is involved in lipid A biosynthesis, and acyl-ACP synthetases were more affected by mutations to the C-terminal region of this helix [121].

Figure 8 ACP–enzyme complex structures

The structure of the B. subtilis ACPS–ACP complex shows the functionally relevant homotrimer of ACPS (blue) that binds ACP (coloured) between the interface of two dimers. (B) An expanded snapshot showing the electrostatic interactions between ACP (coloured from N- to C-terminal in red to white) and ACPS (blue). Anionic residues are shown in red. The β-strands of ACPS are transparent for clarity. (C) Structure of the docked complex between FabG and ACP (purple), which binds between helices II and III and across the dimer interface of FabG. The β-oxoacyl substrate (starred arrow) would have to be extracted from the internal hydrophobic pocket of ACP and directed into the active site of FabG next to the NADPH cofactors (spherical representation). For simplicity, only the dimer of FabG is shown instead of the tetramer.

Another important feature for ACP binding appears to be a common surface pattern on the partner enzyme, which consists of an electropositive/hydrophobic surface patch and lies in close proximity to the active site [119,122]. In one study of FabH, it was shown that the mutants R249A or A253Y in the patch region, representing cationic and hydrophobic residues respectively, largely affect the affinity for ACP [122]. These observations again stress the importance of helix II of ACP, since the docking model of FabH–ACP predicts that Arg249 of FabH interacts with Glu41, which resides in the recognition helix [122]. Arginine residues in the cationic portion of the ACP-binding enzymes appear to play a particular role in these protein–protein interactions. In each of the complexes characterized to date (ACP and ACPS, FabD, FabH, FabG and FabI), arginine residues were identified that, when mutated to uncharged residues, have highly detrimental consequences for ACP binding [108,118120,122]. Although some lysine residues have been implicated as well, it appears that arginine residues may be more suitable, potentially due to their superior hydrogen-bonding ability.

Although helix II has been heavily implicated as an interaction feature between ACP and various enzymes, other regions of ACP are involved as well. The docking study mentioned above between ACP and FabH also shows that the N-terminal region of ACP participates in binding, whereas residues close to helix III are involved in FabG and ACPS binding [118,119,122]. Type I FAS and PKS systems provide yet another situation. For example, thioesterases use carbon rulers to select for appropriate substrates and, in the case of at least one PKS thioesterase, it was shown that protein–protein interactions make only negligible energetic contributions to their association [123,124]. A recent EM (electron microscopy) study of the yeast type I FASs suggests that ACP spends different amounts of time at each enzyme site depending on the protein–protein interactions [125]. Furthermore, it was shown that the FapR of B. subtilis, involved in FAS regulation, only recognizes the malonyl-phosphopantetheine moiety of malonyl-ACP [77]. Therefore the picture is not as straightforward as it might appear. Part of the problem is the relatively low affinity (low μM range) between ACP and its partner enzymes [119,121], which has been noted as a likely reason for why only two ACP–protein complexes have been crystallized to date. The highly dynamic nature of ACP itself probably contributes to this situation as well.

NMR chemical-shift perturbations and computational docking algorithms may present one approach to help circumvent the difficulties of studying the structures of ACP–enzyme complexes. This methodology has been used to characterize the ACP–FabD complex of the FAS system in Streptomyces coelicolor [120] and also the ACP–LpxA complex [126]. LpxA is an acyl transferase involved in lipid A synthesis, which is a major component of Gram-negative bacterial cell walls [127]. In addition to the NMR chemical shift perturbations, RDCs (residual dipolar couplings) were also utilized in the information-driven docking, which provides additional orientational restraints for the placement of ACP [126]. The resulting complex of ACP–LpxA is highly consistent with the experimental data and reveals important salt bridges, again involving helix II of ACP [126]. We used a similar data-driven docking approach to determine the interactions between ACP and FabG. We utilized previously published NMR chemical-shift perturbation data as well as mutational data for Arg129 and Arg172 of FabG [119] and employed experimental information driven docking by HADDOCK [128]. HADDOCK is one of a number of programs, such as ZDOCK [129] and PatchDock [130], that are available and allow various forms of data-driven docking. Since the location of the FabG active site is known, as well as the length of the phosphopantetheine linker of ACP, we were able to significantly limit the search space for ACP and arrive at a model highly consistent with the experimental data (Figure 8). This complex once more highlights the importance of helix II of ACP in its interaction with partner enzymes, as it is in close proximity to Arg129 and Arg172 of FabG, which have been shown to be crucial for ACP binding [119]. Furthermore, the prosthetic group with the attached acyl chain is well situated for delivery of the substrate into the active site of FabG (Figure 8). With the continued development and improvement of data-driven docking protocols, this approach may prove itself as a feasible and effective alternative to obtaining crystal structures of ACP–enzyme complexes. In particular, the accuracy of the docked complex can be increased vastly if mutagenesis or similar information is available on both the target and the ligand proteins.

ACP within type I fatty acid mega-synthases

In type I FAS systems, ACP is part of large, multi-domain polypeptides that also carry the other protein domains for FA synthesis in a linear fashion (Figures 9B and 10A). These chains arrange the mammalian and fungal systems in different multimeric assemblies to build the FAS machinery (Figures 9 and 10). Structures for mammalian, fungal and yeast type I FAS systems have been determined previously by X-ray crystallography [35,111,112,114,131,132]. The type Ia mega-complexes (2.6 MDa) of the yeast S. cerevisiae and the fungus Thermomyces lanuginosus display similar overall folds, described as a barrel shape with a central wheel in the middle that splits the complex into two reaction chambers (Figure 10) [3,5]. Each chamber contains three units each possessing the enzymes necessary for performing the acyl chain elongation cycle. The mammalian FAS structure looks significantly different and is markedly smaller with each monomer of the homodimer measuring approx. 270 kDa in size. In the mammalian complex, of the entire polypeptide chain, only 9% is used in linker regions, whereas in the much larger, fungal type Ia protein approx. 50% of the polypeptide is used for scaffolding [21]. The human protein shape resembles a person sitting with their legs and arms stretched out (Figure 9) [131]. The last two proteins on the mammalian polypeptide chain are the ACP and thioesterase domains. Neither could be identified in the electron density of the mammalian FAS complex, in contrast with the yeast protein in which the structure of ACP was observed (see above). In the mammalian system, ACP is anchored by the oxoreductase domain on the N-terminal side, whereas the thioesterase domain lies on its C-terminal side. In contrast, ACP is connected via linker regions to the side wall of the fungal barrel structure N-terminally and to the central wheel C-terminally. In addition, the N-terminal linker region is alanine- and proline-rich, which might make it more rigid and further restrict the motion of ACP in fungal FAS systems [21]. It is therefore perhaps not surprising that the electron density for ACP was not observed in the mammalian complex.

Figure 9 Mammalian type I FAS structure

(A) Crystal structure of the mammalian type I FAS complex in surface representation, coloured according to the proteins' subunits (B). The yellow arrow denotes the location of ACP, for which there was no electron density in the crystal structure. The broken arrows indicate the directions of FAS movement as was identified by EM (as explained in the main text). (B) The mammalian FAS is composed of two α-chains which each contain a complete set of proteins for FA synthesis. DH, dehydratase; ER, enoyl reductase; KR, oxoacyl reductase; KS, oxoacyl synthase; MAT, malonyl-/acetyl-CoA transacylase; TE, thioesterase.

Figure 10 Structures of the fungal 2.6 MDa type I crystal structure

(A) The heterododecameric arrangement consisting of six α- and β-chains (lower diagram), resembling an overall barrel shape. (B) X-ray structures of the yeast protein reveal ACP (surface representation) locked into position at the oxoacyl synthase domains. (C) EM experiments captured ACP at the active sites of a number of enzymes, revealing the dynamic manner of substrate shuttling during FA synthesis. (C) reproduced from Gipson, P., Mills, D.J., Wouts, R., Grininger, M., Vonck, J. and Kuhlbrandt, W. (2010) Direct structural insight into the substrate-shuttling mechanism of yeast fatty acid synthase by electron cryomucroscopy. Proc. Natl. Acad. Sci. U.S.A. 107, 9164–9169; [125]. AT, acetyl transferase; DH, dehydratase; ER, enoyl reductase; KR, oxoacyl reductase; KS, oxoacyl synthase; PT, phosphopantetheine transferase; TE, thioesterase.

In the crystal structure of the yeast FAS, the ACP was locked into a fixed position with the active site of the oxoacyl synthase enzyme (Figure 10) [5,111,112]. Hence this provided a somewhat static picture, whereas ACP is believed to move dynamically from enzyme active site to active site within each chamber of the complex. Fortunately, for both the mammalian and fungal FASs, EM studies have provided further insights into the dynamic aspects of ACP and FAS action [125,133]. In the mammalian system, a continuous spectrum of FAS conformations was observed, indicating that the lower half of the X-shaped structure can rotate around and thereby access both active site cavities (Figure 9A, lower broken arrow) [133]. In addition to rotation around the vertical axis of the protein, an in-plane swinging by 25 ° is also observed, which alters the active site chambers between open and closed conformations (Figure 9A, upper broken arrow) [133]. These data consolidate the biochemical evidence that each ACP in the dimer can access both reaction chambers, which was not apparent from the crystal structures alone [134]. The yeast FAS seen in the EM studies is slightly relaxed compared with the crystal structure, probably due to the absence of crystal contacts [125]. Densities were identified that correspond to ACP being located at the acetyl transferase, oxoacyl synthase, oxoacyl reductase and the enoyl reductase active sites [125]. ACP exposes the same anionic surface to each domain and the enzymes all possess a cationic region, highly reminiscent of what has been described for type II FAS systems. It was also noted that the electron density inside the reaction chamber is highly variable, suggesting that ACP moves about stochastically in each of the six reaction chambers [125]. This means that the ACP position in one of the reaction chambers is not co-ordinated with the carrier protein position in the other chambers. These EM results clearly support the notion of ACP being able to move around and present the correct acyl chains sequentially to the desired enzymes, so that it can be synthesized in an efficient assembly line fashion. Because ACP cannot leave the reaction chambers, it does not need to hide its acyl chain inside the protein, which could also help expedite its processing in the enzymes of the type I systems.

Although the architecture and sequence identity of the type I FAS systems are drastically different from the type II dissociated enzymes, many of the functional units in these complexes are remarkably similar. For example, the oxoacyl reductase domain (FabG) of E. coli only has 21% sequence identity with the fungal protein, but the RMSD for both structures measures 2.5 Å [114]. On the other hand, other domains, such as the enoyl reductase and dehydratase enzymes, vary significantly between the type Ia, Ib and II systems. It should be noted that human mitochondrial type II ACP shares similar properties to E. coli ACP and is expected to have a similar structure [135], capable of harbouring acyl chains.


Three major processes involve carrier proteins that bind substrates via a phosphopantetheine prosthetic group. These are FA biosynthesis, PKS systems and NRPSs (non-ribosomal peptide synthetases). Although FAS and PKS systems exist in both type I and type II organizational units, NRPSs mainly exist as multi-functional type I polypeptides [136]. In contrast with FAS systems that only utilize one ACP, NRPS and PKS systems may contain a large number of carrier proteins. For example, Streptomyces avermitilis has been documented to have over 80 carrier proteins to accommodate all of its PKS and NRPS assembly lines [137]. Both NRPS and PKS systems are involved in the synthesis of small natural molecules such as the antimicrobials gramicidin S, erythromycin, vancomycin, and iron-chelating compounds such as enterobactin [19,136]. The key difference between the two systems is that the former catalyses the synthesis of peptide bonds, whereas the latter forms carbon–carbon bonds, much like the FAS system. The organization of these multidomain units can become quite complex and even PKS–NRPS hybrids exist, but each system can be decomposed into minimal essential components that are needed for function. The NRPS system minimally requires a carrier protein, an adenylation domain and a condensation domain [136]. Since this system is involved in peptide synthesis, the ACP analogue here is usually termed PCP. In the first step, the adenylation domain catalyses the initiation of the PCP by adding an amino acid via a thioester linkage. The condensation domain then forms a peptide bond by combining two PCP-bound substrates. Similarly, PKS systems minimally require ACP, oxo synthase and acyl transferase domains [136]. The acyl transferase usually loads a malonyl group from malonyl-CoA on to ACP. The oxo synthase domain then extends this substrate through a condensation reaction with another loaded ACP protein. These systems function in an assembly-line fashion and product diversity can be varied depending on the modular units that are present. Whereas in FAS systems each cycle is catalysed by four separate enzymes that reduce the β-oxoacyl chain to its saturated form, PKS and NRPS systems may omit various enzymes in this cycle and thereby create a variety of products. Modifying the assembly line of enzymes has been proposed as a convenient way to create novel bioactive compounds with high stereoselectivity [138,139]. A system of interest is the biosynthesis of daptomycin, which is an antibacterial lipopeptide that is synthesized by an NRPS system [140] and is in use as an antibacterial agent for the treatment of endocarditis and skin infections. Not only are ACP-like proteins involved in the NRPS pathway leading to biosynthesis of daptomycin, but it has recently been shown that the fatty acyl group attached to this lipopeptide is provided by DptF, which is itself also an ACP [140].

Many NRPS PCP and PKS ACP structures have been determined (Table 4), and the overall folds are largely conserved with FAS ACPs (Figure 11). They possess the same helical bundle arranged in a highly similar manner to the FAS counterparts (Figures 11B and 11D). However, unlike FAS ACPs, a distinct difference between the apo and holo forms of PCPs has been noted. For example, in the tyrocidine A synthetase system from B. subtilis, a PCP has been observed to have three distinct conformations. Two states are only accessed by either apo- or holo-PCP, whereas a third conformation is shared by both PCP forms, known as the A/H state (Figure 11) [141]. The A/H form is very similar to the canonical ACP conformation consisting of a four-helix bundle that resembles those seen in the FAS carrier proteins (Figure 11B) [141]. The conformation exclusive to apo-PCP contains shortened helices I–III and, most notably, the region corresponding to helix III is moved further towards the protein core. The structure only seen in holo-PCP is characterized by an altered helix IV that lies almost perpendicular to helix I (Figure 11C). The conformational changes of PCP allow it to alter its interaction interface, exposing different regions of the protein, which selects the appropriate binding partner in the synthetic cycle [141]. This is not observed in PKS ACPs, which behave more similarly to the FAS counterparts (Figure 11D). In PKS systems, differences between apo- and holo-ACP have been noted, but they are far more subtle [142]. Interestingly, the structure of the type I PKS ACP involved in norsolorinic acid synthesis was determined, and although this is not the first type I PKS ACP determined [143], it was noted in this case that, similarly to the rat ACP structure, the acyl chain does not enter the hydrophobic pocket of the protein [115].

View this table:
Table 4 High-resolution PKS ACP and NRPS PCP structures available
Figure 11 Carrier proteins from NRPS and PKS systems

The PCP from the tyrocidine NRPS system undergoes changes depending on whether apo- or holo-PCP is present (AC). The A form shown in (A) is only used by apo-PCP, whereas the A/H form (B) is shared by the apo- and holo-proteins. The H form (C) is exclusive to the holo state. The structures are coloured in a continuous spectrum from red (N-terminus) to blue (C-terminus). (D) The 3-oxobutyl-ACP intermediate from the S. coelicolor actinorhodin PKS, displaying the canonical ACP fold. The 3-oxobutyl group, which does not enter the ACP pocket, is shown in stick representation.


Owing to obvious architectural differences between the human type I FAS and the bacterial type II FAS and their perceived essential function, this particular system has long been considered a good target for the development of novel antibiotics. In addition, there are high-resolution structures available for each enzyme in the FAS pathway, which provides a good starting point for rational structure-based approaches to drug design and lead optimization [2]. Nevertheless, the presence of the type II FAS system in mitochondria, which closely resembles the bacterial systems, puts a high demand on the selectivity of any new compounds developed in this manner. Moreover, whether FASs are plausible targets has been under debate recently, as it was demonstrated that numerous pathogenic Gram-positive bacteria can actually survive in the absence of a functioning FAS system [11]. It was shown that these bacteria have the ability to grow on FAs taken up from external sources, such as human blood plasma [11]. Nevertheless, this does not apply to all bacteria and considering that there are already antibiotics in use that successfully target bacterial FASs, it remains likely that these systems will continue to be an important target for the development of new antibacterial compounds.

The two antibacterial compounds that are currently in use are triclosan and isoniazid, which both inhibit the enoyl reductase FabI [8]. Triclosan is widely used in household products such as hand soap, whereas isoniazid is an anti-tuberculosis drug. It has also been shown that the pantothenamide class of antibiotics is metabolized to form non-productive ACP, which contributes to the antibiotics' activity [144]. Pantothenate is a precursor for CoA, and the pantothenamides are inactive analogues thereof [145]. Through the same biosynthetic pathway that is used to make CoA, an inactive derivatized pantothenamide is made instead and added on to ACP in place of a phosphopantetheine group [144]. FA synthesis is consequently shut down because this form of ACP is not capable of binding the fatty acyl groups. There is also a broad range of natural products that inhibit various steps in the FA synthesis process. For example, cerulenin inhibits FabB and FabF by binding at their active sites [146,147]. Considerable interest has been placed on platensimycin and numerous analogues have been developed [148]. As researchers continue to improve the understanding of how to specifically target bacteria [112], it only seems a matter of time until these will become available commercially.

More recently, the connection of FA synthesis to two pervasive diseases in our society have further sparked interest in FA metabolism. Administration of C75, an analogue of the natural FAS inhibitor cerulenin, led to drastic weight loss in mice [149]. Since other areas of FA metabolism are influenced by C75 as well, it is not clear how much the inhibition of FA synthesis contributes to this effect and as such, additional investigation is needed in this area [12,13]. Furthermore, the commercially available drug Orlistat has been shown to inhibit the thioesterase domain in human FAS and thereby it halts tumour progression in mice [150]. This illuminates a very intriguing possibility of using human FAS inhibitors for targeted cancer treatment, as FA synthesis is only active in adipose and liver tissues and is largely not needed in normal tissue function [12]. Additionally, FAS inhibitors have not displayed any toxicity in chemoprevention studies, making them attractive targets for both prevention and tumour treatment options.


ACP and FAS have long been prominent proteins for study in biochemistry [1]. Their central role and importance in all cells, as well as the abundance and small size of ACP have undoubtedly contributed to their appeal for investigation. Nevertheless, there are many areas in this field that need further detailed study. Only now are we beginning to understand how type II ACP accommodates acyl chains of various lengths and excitingly, the structures of the type I mega-synthases have shed some light on the organization and mechanism of FAS function. Furthermore, with the continued improvements made to computational docking protocols, this approach has become a feasible alternative to X-ray crystallography for the determination of ACP–protein complexes [151,152]. Input data to drive the docking can stem from a wide range of techniques, such as NMR chemical-shift perturbations and cross-saturation experiments [153], mutagenesis and hydrogen/deuterium exchange data or even bioinformatic information can be included. This further enhances the applicability and accessibility of this technique. In light of the potential role FAS inhibitors may play as various therapeutic agents, it is crucial to continue to further our understanding of FASs. FA synthesis is also being investigated for use in biofuel production. Some researchers have recently engineered photosynthetic cyanobacteria to efficiently over produce FAs and secrete them. This enables ready harvesting of FAs for renewable biofuel production [154]. Improving our understanding of FASs will also contribute to our ability to utilize PKS and NRPS systems for our benefit. To date, E. coli has already been used as the host organism for the synthesis of various active natural, non-endogenous products such as macrolides [155], cyclic peptides [156] and aromatic polyketides [157], which are all made using either PKS or NRPS systems. Variations of the native protein clusters found in these systems have been made to synthesize libraries of unnatural, natural products [158]. However, many aspects are still poorly understood, such as the importance of linker regions between the subunits and the basis for subunit preference, as yields are often low in these modified systems [159]. With a better understanding of ACP and, importantly, the factors that govern effective ACP–protein interactions, we will be able to engineer more proficient molecular assembly lines [160162]. Clearly, there are many avenues still to explore in the field of FA synthesis and with the significant developments in recent years, the field will be able to continue to move forward swiftly.


Research in this laboratory on ACPs was supported by a grant from the Canadian Institutes for Health Research. H.J.V. holds a Scientist award from the Alberta Heritage Foundation for Medical Research.

Abbreviations: ACC, acetyl-CoA carboxylase; ACP, acyl carrier protein; ACPS, ACP synthase; EM, electron microscopy; FA, fatty acid; FAS, FA synthase; LPA, lysophosphatidic acid; MD, molecular dynamics; NRPS, non-ribosomal peptide synthetase; PA, phospatidic acid; PCP, peptidyl carrier protein; ppGpp, guanosine tetraphosphate; PKS, polyketide synthase; RMSD, root mean square deviation; SFA, saturated FA; UFA, unsaturated FA


View Abstract