Chimeric Fatty Acyl-Acyl Carrier Protein Thioesterases Provide Mechanistic Insight into Enzyme Specificity and Expression

ABSTRACT

Medium-chain fatty acids are commodity chemicals. Increasing and modifying the activity of thioesterases (TEs) on medium-chain fatty acyl-acyl carrier protein (acyl-ACP) esters may enable a high-yield microbial production of these molecules. The plant Cuphea palustris harbors two distinct TEs: C. palustris FatB1 (CpFatB1) (C₈ specificity, lower activity) and CpFatB2 (C₁₄ specificity, higher activity) with 78% sequence identity. We combined structural features from these two enzymes to create several chimeric TEs, some of which showed nonnatural fatty acid production as measured by an enzymatic assay and gas chromatography-mass spectrometry (GC-MS). Notably, chimera 4 exhibited an increased C₈ fatty acid production in correlation with improved microbial expression. This chimera led us to identify CpFatB2-specific amino acids between positions 219 and 272 that lead to higher protein levels. Chimera 7 produced a broad range of fatty acids and appeared to combine a fatty acid binding pocket with long-chain specificity and an ACP interaction site that may activate fatty acid extrusion. Using homology modeling and in silico docking with ACP, we identified a “positive patch” within amino acids 162 to 218, which may direct the ACP interaction and regulate access to short-chain fatty acids. On the basis of this modeling, we transplanted putative ACP interaction sequences from CpFatB1 into CpFatB2 and created a chimeric thioesterase that produced medium-chain as well as long-chain fatty acids. Thus, the engineering of chimeric enzymes and characterizing their microbial activity and chain-length specificity suggested mechanistic insights into TE functions and also generated thioesterases with potentially useful properties. These observations may inform a rational engineering of TEs to allow alkyl chain length control.

IMPORTANCE Medium-chain fatty acids are important commodity chemicals. These molecules are used as plastic precursors and in shampoos and other detergents and could be used as biofuel precursors if production economics were favorable. Hydrocarbon-based liquid fuels must be optimized to have a desired boiling point, low freezing point, low viscosity, and other physical characteristics. Similarly, the solubility and harshness of detergents and the flexibility of plastic polymers can be modulated. The length and distribution of the carbon chains in the hydrophobic tails determine these properties. The biological synthesis of cell membranes and fatty acids produces chains of primarily 16 to 18 carbons, which give rise to current biofuels. The ultimate goal of the work presented here is to engineer metabolic pathways to produce designer molecules with the correct number of carbons in a chain, so that such molecules could be used directly as specialty commodity chemicals or as fuels after minimal processing.

INTRODUCTION

Fatty acids are commodity chemicals with applications as detergents, plastic precursors, and biofuel precursors (1–6). Normally, bacterial fatty acids reach lengths of 16 or 18 carbons and are incorporated as membrane components. Fatty acids of these chain lengths are typical in hard-water-insoluble soap; improved lathering can be achieved through the addition of shorter-chain fatty acids. Similarly, the pliability of plastic polymers can be modulated by the incorporation of fatty acids with various chain lengths. Finally, medium-chain alkanes have desirable fuel properties, such as volatility and low viscosity, and C₈ to C₁₂ fatty acids would not require “cracking” to serve as cost-effective fuel precursors, for example, by the conversion to alkanes by a fatty acid reductase and an aldehyde decarbonylase (6, 7). Medium-chain fatty acids can be produced by the microbial expression of plant-derived thioesterases (TEs) alone or in combination with other heterologous engineering strategies (8–10).

TEs generate free fatty acids (FFAs) by hydrolyzing the thioester bond that connects the acyl chain to the acyl carrier protein (ACP) (Fig. 1a). TEs of the FatB class are proteins from plant seeds that are ∼46 kDa in size and form homodimers (11, 12). Despite the differences in activity and specificity, they all share the highly conserved hotdog motif (11, 13). It consists of a β-sheet holding two parallel α-helices (Fig. 1b). Several hypotheses about active residues in TEs have been proposed (14–17). Structures of microbial homologs have been determined. (During the preparation of the manuscript, the first structure of a plant-derived FatB enzyme with C₁₂ specificity was determined and found to be quite similar to bacterial homologues [18].) These structures suggest a potential acyl-ACP cleaving site that includes the conserved positions Asp173, Asn175, His177, and Glu211 (19) (PDB accession no. 2OWN and 2ESS, Joint Center for Structural Genomics; accession no. 4GAK, Midwest Center for Genomics) (numbering according to Fig. S1 in the supplemental material).

FIG 1.

FatB-type thioesterase (TE) biochemistry. (a) Synthetic biochemical pathway integrating plant-derived TEs that have different chain length specificities with bacterial metabolism. During E. coli fatty acid biosynthesis, the hydrocarbon chains are tethered to an acyl carrier protein (ACP). Fab enzymes elongate the fatty acid chain, through sequential malonyl-ACP condensation, reduction, dehydration, and further reduction. Heterologous TEs (CpFatB1 and CpFatB2) interrupt this elongation cycle by hydrolyzing fatty acids of different chain lengths from the ACP. (b) Ribbon model of TE dimer (PDB accession no. 2OWN). Numbering of residues corresponds to that in Fig. 2, based on an alignment of 2OWN, CpFatB1, and CpFatB2 (see Fig. S1 in the supplemental material). Each monomer (light and dark gray) exhibits the characteristic hotdog fold of two α-helices partially wrapped in a β-sheet. The proposed fatty acid pocket is indicated by an electron density that was identified in the crystal structure (space filling, orange). Side chains are shown for the putative catalytic residues Asp219, Asn221, His223, and Glu257 (orange side chains) (13), as are the key positive patch residues Arg165, Arg166, Arg192, Arg195, and Arg253 (green), which may be involved in ACP binding (18). The CpFatB2-specific amino acid deletion (red) is also highlighted.

The determinants of chain length specificity of FatB TEs are largely undefined. Previous work has demonstrated that mutations of single amino acids in the proposed N-terminal substrate binding domain changed the specificity of Umbellularia californica FatB1 from C₁₂ to C₁₄ (18, 20). In another study, the exchange of the N terminus of Arabidopsis thaliana FatA and FatB indicated that this part of the protein is largely responsible for specificity despite not containing the active site residues (21). However, these studies focused on changing a short-chain-specific TE into a long-chain-specific enzyme. To our knowledge, no attempt has been made to achieve the reverse.

The interaction dynamic between ACP and fatty acid biosynthesis enzymes may be an important factor in chain length determination (19–22). ACP itself may provide specificity in differentiating between shorter and longer fatty acid tails. Bacterial and plant chloroplast ACPs have pockets that shelter part or all of the fatty acid and the pantetheine group from the aqueous phase (23). In molecular dynamics simulations of acyl-ACP starting with an “out” conformation, longer fatty acid tails return inside the ACP less frequently, suggesting that the ratio of the in to the out states decreases with increasing fatty acid length (22). The interaction between the ACP and fatty acid-metabolizing enzymes is fleeting, but Nguyen et al. (24) solved the structure of an acyl-ACP/FabA-trapped intermediate and found an interaction with arginine side chains in a “positive patch” on FabA that might change the conformation of the ACP and induce fatty acid extrusion from within the ACP. Thus, the in versus out state of the fatty acid on the ACP is dynamic and may be regulated by the interaction with an enzymatic partner (25).

Fatty acid synthesis has many steps, and several enzymes in this process have acyl chain length specificities designed to produce chains of approximately 16 carbons, which constitute the bulk of fatty acids in membrane lipids. Several groups have engineered proteins of fatty acid synthesis to have a shortened chain length specificity (8–10, 26–28). Ultimately, it may be possible to tune the entire fatty acid synthesis pathway to produce designer molecules of practical interest. TEs will be a key enzyme in any such artificial pathway, and so here we sought to understand the features in these enzymes that control chain length specificity to enable metabolic engineers to efficiently create fatty acid products.

RESULTS

Homology modeling: homodimer structure prediction of CpFatB1 and CpFatB2.

To understand structure-function relationships within the Cuphea palustris FatB TEs, we constructed homology models of C. palustris FatB1 (CpFatB1) and FatB2 and the chimeric enzymes described below (Fig. 2a). We based the models on the structures of known hotdog fold TEs and then resampled and refined the backbone and side chain conformations using Rosetta (see Materials and Methods). The Bacteroides thetaiotaomicron acyl-ACP TE (Protein Data Bank [PDB] accession no. 2ESS) produced the best predictions, and the structures of the Lactobacillus plantarum putative oleoyl TE (accession no. 2OWN) and Spirosoma linguale acyl-ACP TE (accession no. 4GAK) were also closely related. The general validity of the gross predicted structures is supported by the fact that the highly conserved residues that are not part of the putative active site play key roles in the formation of the hydrophobic core and dimer interface in the models and by the similarity of our models with the C₁₂ thioesterase structure reported by Feng et al. (PDB identifier [ID] 5X04 [18]), which was published while this article was in preparation.

FIG 2.

Homology model and sequence alignment of TEs CpFatB1 and CpFatB2, used as parent sequences in this study to generate chimeric enzymes. (a) Models of monomers were generated with PDB accession no. 2ESS as the template using SWISS-MODEL and Rosetta (see Materials and Methods). Fragments A to F, which were used to generate chimeras, are labeled and color-highlighted in CpFatB1 (left). The electron density from structure 2OWN representing a putative alkyl chain was inserted by structure alignment and is shown as an orange space-filling model. Catalytic residue and key positive patch residue side chains are colored as in Fig. 1b. (b) Alignment of the amino acid sequences of CpFatB1 and CpFatB2. The fragments A to F, which were used to generate chimeric enzymes, are indicated above the sequences and color-highlighted. Nonconserved amino acids are highlighted in bold.

Design and construction of chimeric TEs.

The fatty acid specificity and sequences of many FatB-type TEs are known, including CpFatB1 (C₈) and CpFatB2 (C₁₄) of C. palustris (Fig. 2) (11, 29). Even though these two enzymes show distinct specificity and microbial activity levels, they share 78% sequence identity. Therefore, we hypothesized that only a small number of amino acids determine the activity and specificity. We generated chimeric proteins constructed from six 50- to 60-amino-acid-long fragments of CpFatB1 and CpFatB2 (Fig. 2, fragments A to F). The breakpoints for the fragments were placed at sequence regions with high identity between the two enzymes. We created the chimeric genes by systematically assembling and cloning segments into the arabinose-inducible expression vector pBAD-HisA (Thermo Fisher Scientific). The expression of hybrid TEs gave a variety of fatty acid synthesis phenotypes (Fig. 3). To initially analyze the free fatty acid (FFA) production of chimeric TEs when expressed in Escherichia coli, we used the free fatty acids half micro test (Roche). The final densities of the tested cultures were similar (Fig. 3a), indicating that variations in fatty acid production are most likely due to differences in enzyme expression and/or activity.

FIG 3.

Typical experiment showing fatty acid production by E. coli expressing chimeric TEs. (a) Titers of total free fatty acids in the pellet and supernatant of E. coli (DE3) ΔfadD transformed with pBAD-HisA expressing TEs obtained by performing an enzyme-based colorimetric assay (see Materials and Methods). Densities of the corresponding cultures are given as OD₆₀₀ values. Fragments A to F of chimeras and wild-type (WT) enzymes are shown as colored bars. Measurements were obtained from three biological replicates for each TE, with error bars indicating standard deviations. (b) The same samples were subsequently analyzed for their free fatty acid species using GC-MS (chimera 5 was omitted here but generated a fatty acid profile similar to that of CpFatB1 in other experiments) (see Table S2). (c) From these data, we inferred specific functions for distinct segments of the protein.

The shorter hydrocarbon chains are amphiphilic and therefore cross membranes more easily than long-chain fatty acids. C₈ fatty acids are expected to occur in the supernatant, whereas C₁₄ fatty acids should be retained in the cell pellet and membranes. We detected free fatty acids in both the supernatant and the pellet and then quantified their abundance using standard curves with known concentrations of octanoate and myristate, respectively. As expected, cells expressing CpFatB1 yielded fatty acids only in the supernatant and cells expressing CpFatB2 yielded compounds only in the pellet (Fig. 3a). The chimeras showed a variety of fatty acid production characteristics. Enzymes that resulted in fatty acid production exhibited CpFatB1-like (chimeras 4 to 6), CpFatB2-like (chimeras 1 and 8 to 10), or intermediate (chimera 7) specificities. Some chimeras did not show any fatty acid production (chimeras 2 and 3). These initial results were validated by the quantitation of specific free fatty acids in whole cultures as determined by gas chromatography-mass spectrometry (GC-MS) (Fig. 3b).

Increased microbial activity of chimera 4.

Chimera 4 maintained the specificity from CpFatB1 but had a much higher microbial activity (Fig. 3), which may make this construct useful in the engineered synthesis of octanoic acid. Conversely, chimera 5 showed the specificity and activity of CpFatB1. The only difference between the two chimeras is fragment E (amino acids 220 to 272), which also contains the proposed active site loop and adjacent helices and loops (Fig. 2). We inferred that this fragment contains the polymorphisms responsible for the microbial activity difference between CpFatB1 and chimera 4 (Chim4). To quantify the composition of the fatty acids produced, we used GC-MS analysis. The octanoate production of chimera 4 exceeded CpFatB1's production by approximately 6- to 13-fold (Fig. 3).

The protein expression levels were dramatically enhanced by CpFatB2 genetic polymorphisms between amino acids 219 to 272, which likely explains the enhanced fatty acid production by chimera 4 (Fig. 4). This region has only six FatB1/FatB2 polymorphisms: S225N, E235Q, M237V, L242F, S249G, and S269A. To understand how these polymorphisms contribute to enhanced fatty acid production, we introduced the six CpFatB2 (favorable) residues into CpFatB1 and the six CpFatB1 (unfavorable) residues into chimera 4 and measured free fatty acid production and protein expression levels via Western blotting (Fig. 4). Three of the polymorphisms had detectable effects within CpFatB1 on fatty acid production, namely, S225N, E235Q, and S249G (Fig. 4b), while the other three did not (Fig. 4b; see also Fig. S3 in the supplemental material). For example, the introduction of the change S225 to N increased fatty acid production approximately 4-fold, while the reciprocal change of N225 to S in chimera 4 reduced fatty acid production approximately 2-fold. Exchanges of E235Q and S249G had analogous but less dramatic effects. Combining these three mutations within a CpFatB1 protein background had an additive effect, with microbial activity approaching that of chimera 4. A Western blot analysis indicated that the most active proteins, chimera 4 and chimera 4 withV237M, were clearly produced at higher levels than the less active proteins (Fig. 4c). These results suggest that the folding of these plant-derived TEs may not be optimal in E. coli, with unfolded proteins being degraded, and that chimera 4 fortuitously has superior folding properties.

FIG 4.

Dissection of fragment E's effect on microbial FFA production. (a) Chimera 4 contains fragments E and F from CpFatB2. Differences in fatty acid production levels are due to fragment E (Fig. 3a). There are six amino acid substitutions between CpFatB1 and CpFatB2 in fragment E. (b) Amino acid substitutions were introduced into either CpFatB1 (converting to the CpFatB2 sequence) or into chimera 4 (converting to the CpFatB1 sequence). Constructs were expressed, and FFAs in supernatants were measured using an enzymatic assay. Shown are the averages from three biological replicates; error bars indicate standard deviations. (Substitutions of L/F242 and S/A269 had no effect [see Fig. S3].) (c) From the same cultures as in panel b, TEs were visualized by Western blotting using an anti-His6 antibody (for full blot, see Fig. S4). As a loading reference, the blot was also probed with an anti-RNA polymerase beta antibody (RPβ).

Seeking structural explanations for the phenotypes of each chimera, we constructed and examined three-dimensional homology models of CpFatB1, CpFatB2, and chimera 4. We generated the initial structures from SWISS-MODEL and refined them using Rosetta. An inspection of the structure models of CpFatB1, CpFatB2, and chimera 4 suggests that the most helpful mutation, S225N, may enhance the formation or stability of an α-helix that extends from Asn226 to Met237. In addition, Ser/Asn225 is at the +2 position relative to the putative active site histidine, and this substitution may cause a slight conformational change in the active site and could influence enzyme activity. The reasons why the polymorphisms at positions E235Q and S249G affect fatty acid production are not apparent from an inspection of the structure; these variations may affect the formation of incorrectly folded states that are degraded.

Permissive and restrictive determinants of chain length specificity.

Chimera 7 showed a novel fatty acid production phenotype that was distinct from those of both CpFatB1 and CpFatB2. This chimeric protein produced a range of fatty acids from C₈ to C₁₆, with C₁₂ being the most abundant species. Chimera 7 differs by 10 amino acids in fragment B (amino acids 63 to 116) from chimera 6, which produced primarily C₈ fatty acids. These results suggest that the CpFatB1 polymorphisms in this region together prevent longer chain fatty acids from entering the proposed binding pocket, while the CpFatB2 polymorphisms are more permissive to longer chains. (We also constructed numerous variant proteins with single amino acid substitutions in this region but were unable to find single point mutations that conferred a change in specificity [see Table S1].) The polymorphic amino acids are in the proposed fatty acid pocket, suggesting that these polymorphisms may collectively influence the binding pocket size.

Molecular modeling and an examination of homologous structures support the notion that this region of the protein defines a fatty acid binding pocket. The related structures of PDB accession no. 2ESS and 2OWN show electron densities in the putative binding pockets, consistent with this identification. Specifically, the structure 2ESS has two molecules of the crystallization agent 2-methyl-2,4-pentanediol in the pocket, and 2OWN has an unidentified electron density, modeled as a 9-carbon alkyl chain, in the same position (Fig. 1b). All together, these results suggest that a pocket defined by amino acids Gln72, Cys79, Asn86, Val104, Arg137, Trp102, and Met135, as well as Tyr/Trp157, Val 219, and Asn224 (in CpFatB2), binds to the alkyl chain of the fatty acid. This interpretation is supported by mutational and structural analyses of other related structures, including the recently solved C₁₂ TE (16, 18).

Fragment C (amino acids 117 to 161) appears to define a segment of CpFatB2 that functions to restrict the substrate specificity of this protein to longer chain fatty acyl-ACPs. Chimera 7 has a relaxed specificity and generated a range of fatty acid products, while the distribution of products from chimera 8 is much more skewed toward longer products (Fig. 3b), with C₁₄ the main product and very little C₈. Chimera 9, in which fragment D (amino acids 162 to 218) was also derived from FatB2, produced a distribution of fatty acids even more skewed toward C₁₄ and with no C₈, essentially identical to that of CpFatB2 itself. Figure 3c summarizes these observations. We note that according to the scheme in Fig. 3c, chimera 2, which is the converse of chimera 7, should contain segments that restrict both long- and short-chain fatty acyl substrates and was predicted to be inactive, which it is. However, there are other reasons why this chimera might be inactive, such as improper folding. We hypothesize that features of fragments C and D may promote ACP binding in a way that, in CpFatB1, promotes the extrusion of the fatty acid group, particularly for shorter chains.

Docking of ACP with CpFatB1 and CpFatB2 models.

We generated 10,000 docking conformations between E. coli ACPs and the models of CpFatB1, CpFatB1, and chimera 4 (see Materials and Methods). Of the most energetically favorable conformations, we clustered similar structures to identify the dominant binding modes (Fig. 5). An inspection revealed that the top sampled binding modes primarily occurred between the ACP and the “positive patch,” including Arg255 and Arg256 as well as Arg165, Arg166, and Lys195. The binding modes that correctly oriented the pantetheine-carrying Ser (modeled without the pantetheine modification) were found exclusively in the sampling for apo-ACP with CpFatB1 and chimera 4. Because sampling is not exhaustive, we also aligned this binding mode against the CpFatB2 model to determine if it was substantiated by interactions specific to CpFatB1 or CpFatB2. The residues that directly interfaced with the ACP are conserved (Lys195, Arg255, and Arg256) (Fig. 5). The two arginines directly preceding the active site (Arg255 and Arg256) intercalate between two helices of the ACP, similar to in structures of FabA-docked ACP (PDB ID 4KEH) (24). A third supporting polar interaction in the apo form of the ACP-CpFatB1/2 docking was provided by Lys195, a residue on the loop of fragment D. In CpFatB2, this loop is significantly altered relative to that in CpFatB1 (Fig. 2b), including the sole insertion/deletion, and the models place the FatB2 Lys195 side chain further from the putative ACP docking site than for CpFatB1 (Fig. 5, inset).

FIG 5.

ACP docking to CpFatB1 and CpFatB2. In silico docking of ACP (green) with modeled CpFatB1 (blue) was done using Rosetta's docking protocol (RosettaDock). We applied several solved structures of ACP (PDB IDs 1T8K, 2FAD, 2FAE, and 3EFB) and sampling was unbiased. Ten thousand arrangements were generated and scored, and the best 200 of each pair were clustered to identify 10 representative binding orientations. The proposed fatty acid pocket and active site residues (orange) are in close proximity to the ACP binding site. The upper right expanded view (rotated relative to the depicted dimer) shows direct interactions between “positive patch” residues K195, R255, and R256 and ACP residues. The predicted position of K195 in CpFatB2 is superimposed in gray.

The transplantation of amino acids 162 to 218 (fragment D) from FatB1 into FatB2 causes a relaxation of specificity, allowing medium-chain fatty acids to be produced (Fig. 6). These amino acids are not part of the alkyl chain binding pocket but do interact with the ACP in the docking models. Specifically, the production of C₈ and C₁₀ fatty acids from CpFatB2 was undetectable, while the production of these fatty acids from the FatB2 that contained fragment D from FatB1 [FatB2(FatB1-fragment D)] was approximately 10% of the total, and the C₁₂ fatty acid was only 6% of the product from CpFatB2 but was 20% of the product from FatB2(FatB1-fragment D) (Fig. 6b). These results indicated that it is possible to modulate the chain length substrate specificity without changing the amino acids around the likely alkyl chain binding pocket. We hypothesize that amino acids 162 to 218 and also 117 to 161 contribute to an ACP contact site, and the polymorphisms found in CpFatB1 enhance the extrusion of medium-chain fatty acids from the ACP. (We tested a few subsets of the 162-to-218 fragment but were unable to introduce short-chain specificity [see Fig. S5].)

FIG 6.

Region distinct from the fatty acid binding pocket, fragment D, modulates the FFA length specificity of CpFatB1/B2. (a) Chimera 7 organization. Chimera 7 includes sequences from CpFatB2 that allow binding of a long-chain fatty acid (fragment B) and sequences from CpFatB1 corresponding to the ACP binding site inferred from modeling (fragments C and D). (b) Transplantation of fragment D from CpFatB1 into CpFatB2 widens the distribution of fatty acids produced. The entire fragment D from CpFatB1 was introduced into CpFatB2, proteins were expressed in E. coli, and fatty acid production profiles were analyzed via GC-MS. Shown are the averages from three biological replicates and their standard deviations.

DISCUSSION

Plant-derived fatty acyl TEs are important tools in the microbial production of fatty acids by metabolic engineering. However, the factors that determine the microbial activity levels and the alkyl chain length specificity of these enzymes are poorly understood. Here, we used a combination of hybrid enzyme construction, site-directed mutagenesis, and in silico modeling to find determinants of the expression and specificity of the C₈ and C₁₄ fatty acyl TEs of C. palustris. These two enzymes are 78% identical at the amino acid level yet have strikingly different microbial activities and substrate specificities.

One hybrid enzyme (chimera 4) had the C₈ specificity of CpFatB1 with the higher in vivo activity of CpFatB2 (Fig. 3). This enzyme consists of amino acids 1 to 218 of CpFatB1 and amino acids 219 to 316 of CpFatB2. The enhanced microbial activity is likely due to the improved expression of the hybrid protein (Fig. 4), which appears to be expressed in E. coli at higher levels than either CpFatB1 or CpFatB2 under identical conditions. The polymorphisms Ser/Asn225, Glu/Gln235, and Ser/Gly249 in this segment appear to additively contribute to an improved in vivo activity. This hybrid enzyme may be particularly useful in the microbial production of C₈ fatty acids.

The binding site for the alkyl chain of the fatty acid is likely to be a deep pocket flanked by amino acids, including Cys79, Trp102, Met135, Arg137, and Tyr/Trp157. This conclusion is based on the observation that active chimeric enzymes with amino acids in this segment from CpFatB1 produced only a C₈ fatty acid, while those with the amino acids from CpFatB2 produced a C₁₄ fatty acid (Fig. 3). Further evidence that this region encodes the fatty acid binding pocket comes from the structure of a bacterial TE, 2ESS, which has two solvent molecules in this pocket, and the putative TE structure 2OWN, which has an electron density consistent with an alkyl chain in this area. The site-directed mutagenesis studies that converted a plant C₁₂ TE to accommodate C₁₄ substrates support this interpretation (18). CpFatB1/2 polymorphisms in this region appear to restrict or allow access by long alkyl chains in the fatty acid (Fig. 3c).

The length specificity of CpFatB1, the C₈ TE, may be explained in terms of the steric exclusion of longer alkyl chains, but the mechanism by which the CpFatB2 C₁₄ TE excludes shorter chains cannot be so simple. We hypothesize that CpFatB2 achieves a specificity for longer chains by using the equilibrium of an ACP with longer fatty acids, which favors the protruded state. ACPs in bacteria and plants have deep pockets in which attached fatty acids preferentially reside (23). This can influence the length specificity of enzymes involved in fatty acid synthesis (8). Chan et al. (22) simulated the in-to-out and out-to-in transition rates of ACP bound to fatty acids of various lengths; they found that the longer fatty acids remain outside the ACP for a longer period of time and had fewer stabilizing hydrogen bonds between pantetheine and the protein when the fatty acid was in the “in” state. We propose that the C₈ TE may have a mechanism for encouraging the protrusion of the fatty acid from the ACP, while the C₁₄ TE may rely on the natural equilibrium for the C₁₄ acyl-ACP to be in the “out” state.

ACPs are negatively charged proteins, and fatty acyl-ACPs generally interact via a “positive patch” on enzymes that process fatty acids. Nguyen et al. (24) determined the structure of the E. coli fatty acid dehydratase (FabA) with an ACP and an attached fatty acid in the extruded state. In this structure, the ACP interacts directly with two arginines within a larger positive patch, and this interaction appears to modulate the structure of the ACP to favor fatty acid release. The C₈ and C₁₄ TEs also have positive patches near the active sites: amino acids Lys97, Arg98, Arg/Lys106, Lys132, Lys163, Lys165, Arg166, Arg169, Arg175, Arg/Gly194, Lys195, Arg255, and Arg256. Most of these amino acids lie in fragments C and D, which appear to allow or block access to shorter chain substrates (Fig. 3 and 6). The entire “positive patch” is rather large, so the ACP may initially bind nonspecifically through charge-charge interactions before docking at the site that stimulates fatty acid extrusion.

Polymorphisms between CpFatB1 and CpFatB2 may change the ACP docking site to differentially promote fatty acid protrusion. Specifically, we hypothesize that the CpFatB2-ACP interaction is less favorable than the CpFatB1-ACP interaction, but this effect would be balanced by the greater intrinsic tendency of the C₁₄ acyl-ACP to remain in the “out” configuration. Figure 7 summarizes this model. To address this hypothesis, we transplanted amino acids 162 to 218 (fragment D) from CpFatB1 into CpFatB2; this segment does not contain any amino acids around the putative alkyl chain binding pocket but does include amino acids predicted to interact with the ACP. The resulting hybrid protein retains specificity for C₁₄ substrates but has the added ability to cleave C₈-, C₁₀-, and C₁₂-length substrates (Fig. 6b). This observation suggests that CpFatB TEs contain features that are permissive/restrictive for medium-chain substrates and that these features are distinct from the fatty acid binding pocket and its steric restriction of long-chain substrates.

FIG 7.

Model for structural mechanisms that determine chain length specificity in FatB-type TEs. According to this model, C₈ fatty acids can be cleaved by CpFatB1 (blue shapes), because the positive patch (green “+” signs) on this enzyme stabilizes the required ACP state (light green shapes). CpFatB2 (gray shapes) is unable to cleave C₈ fatty acids because it lacks the right conformation of the positive patch, which is mostly guided by fragment D. CpFatB1 cannot cleave C₁₄ fatty acids because its pocket is too short, causing the pantetheine group (orange) to stick out instead of residing near the catalytic residues (orange arrow). C₁₄ fatty acids are released from the ACP more readily without the requirement of stabilization through the positive patch. CpFatB2 contains a pocket that accepts larger fatty acids, which is governed by fragment B (dark gray). Hence, it is able to cleave C₁₄ fatty acids.

In summary, we analyzed the structure-function relationships that control microbial activity and the length specificity of Cuphea palustris fatty acyl-ACP TEs. We identified sequence features of these proteins that can be modified to increase protein expression in bacteria, as well as features that may control length specificity. These observations may be useful in the production of designer chemicals by metabolic engineering.

MATERIALS AND METHODS

Strain construction.

Double-stranded DNA-encoding fragments A to F of CpFatB1 and CpFatB2 (Fig. 2) were synthesized as gBlocks (Integrated DNA Technologies) and cloned into the pBAD-HisA (Thermo Fisher Scientific) vector by isothermal assembly (30) (all chimera sequences are listed in Fig. S2 in the supplemental material). Amino acid mutations of the CpFatB1 and CpFatB2 genes were incorporated by PCR amplification with mutagenesis oligonucleotides. An empty pBAD-HisA vector without a FatB gene was included as a control. The plasmids were sequence verified and transformed into E. coli BL21(DE3) ΔfadD.

Growth and induction.

Fatty acid production experiments were performed in glass tubes. Individual colonies of fresh transformants were grown overnight in 1.5 ml LB with 1% glucose and 100 μg/ml carbenicillin in 15-ml culture tubes at 37°C and then diluted 1:20 into 2 ml or 5 ml M9 with 1% glycerol, grown at 30°C in a roller drum to an optical density at 600 nm (OD₆₀₀) of 0.4 to 0.6, and induced with 1% arabinose. The cells were harvested after 24 h for enzymatic fatty acid analysis, GC-MS analysis, and Western blotting.

OD and enzymatic FFA measurements.

The OD₅₉₅ of 100-μl cultures was measured in a 96-well flat-bottom, nontreated, sterile polystyrene plate (Corning) on a Victor 3V 1420 multilabel counter (Perkin-Elmer) using a 595/60-nm filter. Enzymatic FFA measurements were performed according to Torella et al. (10) with modifications as follows. Two hundred microliters of each culture was separated into a supernatant and a pellet by centrifugation. FFAs in 5-μl supernatants were measured directly using the enzymatic free fatty acid half micro test (Roche) according to the manufacturer's instructions, scaled down to 1/10 volume in a 96-well plate with octanoate as the standard for quantification. FFAs in the pellet were extracted into ethyl acetate as follows. To a 100-μl redissolved pellet in M9 glycerol with 1% NP-40, we added 10 μl of 10% NaCl, 10 μl glacial acetic acid, and 40 μl ethyl acetate. The solution was mixed by vortexing and centrifuged for 10 min at 16,000 rpm. Seventeen microliters of the ethyl acetate phase was then evaporated using a SpeedVac and redissolved in M9 glycerol, 5 μl of which was subjected to the enzymatic free fatty acid half micro test as described above, using a myristate standard for quantification.

Fatty acid identification and quantification using GC-MS.

The identification and quantification of FFAs were essentially performed according to Torella et al. (10). In brief, we separated FFAs from the culture by acidifying 400 μl culture with 50 μl 10% (wt/vol) NaCl and 50 μl glacial acetic acid and extracting the FFAs into 200 μl ethyl acetate. FFAs in 100 μl organic phase were then esterified in 900 μl of a 30:1 mixture of ethyl alcohol (EtOH) and 37% (vol/vol) HCl by incubating them at 55°C and extracted into 500 μl hexane. Fatty ethyl esters were analyzed on an Agilent GC-MS 5975/7890 (Agilent Technologies) using an HP-5MS column (length, 30 m; diameter, either 0.25 or 0.5 mm; film, 0.25 μm). We identified chain lengths by GC retention times and mass spectra and quantified them using an internal standard (800 mg/liter pentadecanoate added to the culture before the extraction procedure). We used known concentrations of C₆, C₈, C₁₀, C₁₂, C₁₄, and C₁₆ fatty acids to generate a standard curve and to quantify the production of single fatty acid species.

Computational fold prediction.

The lack of structures for these TEs or chimeric proteins required that the structural analysis rely on representative homology models. To generate these models, we first assembled the initial structures using SWISS-MODEL to thread the CpFatB1 and CpFatB2 sequences to aligned coordinates in all structures of dimeric TEs within 25% sequence homology (31–34). We then used Rosetta to apply several consecutive rounds of random perturbations to each starting point provided by SWISS-MODEL, extending the sampling to thousands of neighboring conformations (35–37). The lowest energy perturbations of each homology model were optimized by iterative Monte Carlo adjustments to the protein backbone (38). The best scoring models were relaxed with spheroid side chain approximations, finely backbone minimized, and then side chain minimized with full-atom refinement. The final models were then clustered and selected by the Rosetta score.

In silico docking of ACP with modeled TEs.

We used Rosetta's docking protocol (RosettaDock) to identify the potential binding orientations of the ACP to CpFatB1, CpFatB2, and chimera 4 (39). To determine whether the binding modes were particular to the loading state of the ACP, we docked the solved structures of the E. coli ACPs, both empty and carrying 7-, 10-, and 14-carbon fatty acids (PDB IDs 1T8K, 2FAD, and 3EJB, respectively) (23, 40, 41). The sampling was unbiased, obtained by randomizing the rotational orientation of the ACP and testing random positions of the protein placed within a 45-Å radius of the TE center of mass. Ten thousand arrangements of ACPs and TE dimers were generated this way for each pairing, and side chains at the ACP-TE contacting surfaces were optimized. The overall structures were scored, and we took the best scoring of each pair. Clustering those structures to identify the most successful binding modes, we inspected the largest clusters as representative binding orientations.

Western blot for His-tagged thioesterases.

We directly measured thioesterase levels in expressing E. coli strains by Western blotting. Cells were grown and induced as described in “Growth and induction.” Five-milliliter cultures were spun down and resuspended in 50 μl Tris-HCl 50 mM (pH 7.4), the protein was quantified using a Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Scientific), and 25 μg was loaded in a Novex WedgeWell 4 to 20% Tris-glycine gel (Invitrogen by Thermo Fisher Scientific) according to the manufacturer's instructions. The gel was run for 2.5 h at 120 V. The proteins were transferred to a nitrocellulose membrane (iBlot gel transfer stacks nitrocellulose, regular; Invitrogen by Thermo Fisher Scientific). The membrane was blocked with 4% milk in Tris-buffered saline–Tween 20 buffer (TBS-T) at 4°C overnight. The anti-RNA polymerase beta antibody (EPR18704; Abcam) was added at a 1:2,000 dilution and incubated for 30 min, followed by four washes for 5 min with TBS-T. The horseradish peroxidase (HRP)-conjugated antibodies anti-rabbit IgG (H+L) (XG-6160HRP; ProSci Incorporated) and anti-6× His tag (ab1187; Abcam) were added at 1:5,000 dilutions, incubated, and washed as with the primary antibody. The blot was imaged using SuperSignal West Dura extended duration substrate (Thermo Scientific). Western blotting was performed on whole-cell lysates rather than supernatants, because our thioesterase constructs appear to associate with membranes (see Fig. S6).