Fused dimerization increases expression, solubility, and activity of bacterial dehydratase enzymes

Abstract

FabA and FabZ are the two dehydratase enzymes in Escherichia coli that catalyze the dehydration of acyl intermediates in the biosynthesis of fatty acids. Both enzymes form obligate dimers in which the active site contains key amino acids from both subunits. While FabA is a soluble protein that has been relatively straightforward to express and to purify from cultured E. coli, FabZ has shown to be mostly insoluble and only partially active. In an effort to increase the solubility and activity of both dehydratases, we made constructs consisting of two identical subunits of FabA or FabZ fused with a naturally occurring peptide linker, so as to force their dimerization. The fused dimer of FabZ (FabZ‐FabZ) was expressed as a soluble enzyme with an ninefold higher activity in vitro than the unfused FabZ. This construct exemplifies a strategy for the improvement of enzymes from the fatty acid biosynthesis pathways, many of which function as dimers, catalyzing critical steps for the production of fatty acids.

Abbreviations

ACP

acyl carrier protein

Introduction

The production of fatty acids in bacterial cultures remains an important objective toward the sustainable generation of biofuels and biomaterials (see Ref. 1 for a review). Different approaches have been adopted as a means to increase commercial fatty acid yields in bacterial cultures, such as the modulation of the fatty acid biosynthetic pathway, the deletion of fatty acid degradation genes, the regulation of gene transcription, and the overexpression of thioestererases and other enzymes.2, 3, 4, 5, 6, 7, 8 In Escherichia coli, as in most bacteria, the biosynthesis of fatty acids in most bacteria involves the use of independent enzymes to catalyze reactions in a stepwise manner involving specific biocatalysts for acyltransfer, condensation, ketoreduction, dehydration, and enoyl reduction.9, 10, 11, 12

The dehydration of acyl intermediates in the biosynthesis route of E. coli is catalyzed by two dehydratrase enzymes: FabA and FabZ.10 Both dehydratases act on β‐hydroxy intermediates to form the conjugated double bond in the acyl intermediate. While FabZ will dehydrate intermediates of varying lengths, FabA exhibits specificity for acyl chains that are 10‐carbons long. This conjugated double bond that results from FabA/Z activity can either be further reduced by an enoyl reductase (FabI) or isomerized from the trans 2,3 position to a cis 3,4 position (Fig. 1). The reaction catalyzed by FabA and FabZ has been shown to be a bottleneck in the production of fatty acids in numerous microbial hosts.13, 14 Thus, an improvement in FabA/Z activity or availability could result in an enhancement in the production of fatty acids in bacterial cultures.

Figure 1.

Both FabA and FabZ catalyze the dehydration of fatty acyl‐ACP intermediates. FabA catalyzes an additional double bond isomerization step to form the 3,4 cis double bond.

Both FabA and FabZ form a double hotdog dimeric structure in which the active site consists of amino acids from both subunits. The conserved histidine in position 70 in FabA is complemented with an aspartate in position 84 of the partnering subunit.15 While the expression in E. coli and purification of both FabA and FabZ have been reported as part of a greater effort to reconstitute the full fatty acid synthase pathway of E. coli, our experience with FabZ is that ∼90% of the protein produced, sediments together with the insoluble cellular debris.14

In an effort to increase the production and activity of FabA and FabZ, we designed, built and produced a construct encoding two fused copies of the same dehydratase linked by a peptide stretch from a polyunsaturated fatty acid synthase multienzyme from Photobacterium profundum by overlap PCR, overexpressed in E. coli, purified and assayed for hydratase activity. While no differences in yield or activity were observed between FabA and FabA‐FabA, the fused dimer FabZ‐FabZ was found to be more soluble and more active in vitro than the native FabZ enzyme, raising the possibility of enhancing fatty acid production through the forced dimerization of individual enzymes.

Results

Protein solubility and purification

Proteins were expressed in E. coli and purified by Ni‐Sepharose affinity chromatography, which resulted in proteins that were highly pure. The overall protein yields for FabA and the fused FabA‐FabA were 40 mg and 52 mg/L of culture, respectively. The yield for FabZ was substantially lower since the protein was highly (although not entirely) insoluble [Fig. 2(A)]. From the soluble lysis supernatant, a total of 0.5 mg FabZ was obtained per liter of culture. Interestingly, the fused dimer FabZ‐FabZ was recovered in the soluble lysis supernatant in higher yields than the parent FabZ molecule at 2.8 mg/L of culture, indicating that dimer fusion promotes enzyme solubility [Fig. 2(B)].

Figure 2.

Soluble versus insoluble yields of FabZ and FabZ‐FabZ. Both the lysis supernatant and the lysis pellet were loaded into a Ni²⁺‐sepharose column both for (A) FabZ and (B) FabZ‐FabZ. The lysis pellet was solubilized using Bugbuster detergent (Novagen) prior to loading. Elution fractions from the Ni²⁺‐sepharose were analyzed by SDS–PAGE (Lane 1: Protein Ladder; Lanes 2–8 contain the elution fractions from the resolubilized lysis pellet; Lanes 9–15 contain the elution fractions from the soluble supernatant) The FabZ was mostly found in the insoluble lysis pellet (0.30 vs. 0.16 mg/L soluble), whereas the FabZ‐FabZ, was mostly recovered from the soluble supernatant at 2.75 mg/L.

Enzyme activity assays

All enzymes in this study were assayed for their ability to hydrate crotonyl‐CoA since it has been shown for this family of dehydratases that equilibrium favors the reverse reaction10 (Fig. 3). Not surprisingly, FabA and the fused dimer FabA‐FabA showed very little activity toward crotonyl‐CoA, a substrate analog, which is substantially shorter than the 10‐carbon fatty acid that is the natural substrate for FabA [Fig. 3(A)]. Similarly, the activity of FabZ toward crotonyl‐CoA was modest (k _cat = 144 s⁻¹; K _m = 64 µM). The FabZ‐FabZ fused dimer showed a significant increase in activity as compared to the FabZ native dimer with an ninefold enhancement in catalytic efficiency with a k _cat = 3814 s⁻¹ and K _m = 179.8 μM [Fig. 3(B)].

Figure 3.

Enzyme kinetic assays. Initial velocities of crotonyl‐CoA hydration were expressed in enzyme units per μg of enzyme. The measurements were performed in triplicate and the error bars represent the standard deviation. (A) The enzyme activity for FabA and FabA‐FabA was minimal. (B) FabZ‐FabZ shows a significant ninefold increase in V _max compared to wild‐type FabZ.

Quaternary structure of fused dimers

Using size exclusion chromatography on a Superdex 200 column, we estimated the quaternary structure for the different fused dimers (Fig. 4). The elution volume of FabA‐FabA corresponded to that of a protein weighing 45 kDa, which is the theoretical molecular weight for the fused dimer and in agreement with the dimeric nature of wild‐type FabA. Eluting at an apparent weight of ∼116 kD, FabZ‐FabZ displays a elution volume indicative of a trimer of fused dimers. This is consistent with the published three‐dimensional structure of FabZ (PDB ID: 1U1Z), which shows a hexameric arrangement in the unit cell possibly due to the formation of a trimer of dimers.16

Figure 4.

Determination of oligomeric states by size exclusion chromatography. FabA‐FabA elutes at a volume indicative of a 46 kDa dimer, whereas FabZ‐FabZ elutes as a 116 kDa hexamer, or the expected trimer of dimers that is shown in most crystal structures of FabZ.

Discussion

Enhancing the production of fatty acids in bacterial or other microbial cultures is of paramount importance in the fields of biofuels, agriculture and human nutrition. Toward this end, several groups have devised strategies for increasing the microbial production of fatty acids by the targeted deletion of beta‐oxidation enzymes to cause fatty acid accumulation, the overexpression of thioesterases to increase fatty acid release and the overexpression of transcriptional regulators of fatty acid biosynthesis.2, 3, 4, 5, 6, 7, 8 Here we show a complementary strategy that involves the optimization of individual enzymes within a biosynthesis network through covalent linkage. In this instance, the catalytic efficiency of dehydratase FabZ was dramatically enhanced by the introduction of a peptide connector that presumably encouraged favorable interactions between identical subunits. While the efficiency of FabZ was enhanced ninefold, that of FabA was unaffected by homodimerization. This lack of activity of FabA could be explained by the fact that the surrogate substrate in this study, crotonyl‐CoA, is much shorter than the 10‐carbon intermediate that is the true natural substrate recognized by FabA. Thus, it is not surprising that neither FabA nor FabA‐FabA had much activity toward this short substrate.

The covalent linkage of enzyme domains also affects their expression, solubility and stability. We first attempted the production of wild‐type FabZ and found it to be highly insoluble, mostly in inclusion bodies with some protein present in soluble form. By contrast, FabZ‐FabZ was highly soluble in common buffers and needed no extra treatment for solubilization. In the case of FabA and FabA‐FabA, both were highly soluble but the expression and purification of FabA‐FabA dimer yielded more protein than that of FabA alone. These results indicate that the presence of a linker peptide not only does not impair protein solubility, but can significantly enhance protein production and recovery, an observation with implications in the field of biotechnology.

The idea of fusing protein dimers to improve upon their properties has been explored.17, 18 Protein hormones have been fused together either to increase stability, to increase expression or to enhance biological activity.19, 20, 21 Also, the idea of fusing enzymes that catalyze consecutive reactions along a metabolic pathway has been extensively explored.22, 23, 24 For instance, the fusion of enzymes that catalyze consecutive steps in the degradation of dextrins or plant biomass have been shown to be more efficient than the parents enzymes working separately.25, 26, 27 Critical in the design of fused proteins is the choice of peptide linker, which can be designed either to be rigid through the presence of repeated Ala residues or to be flexible through the presence of repeated Gly residues.28 In this report, we have fused a dimeric enzyme by introducing a naturally occurring peptide linker from a multi‐domain PUFA synthase of sequence STQNVAIQTAAPVASASNGLDAAQVQGT. The presence of multiple alanine residues in this linker sequence suggests a rigid structure, which is consistent with earlier observations that multiple acyl carrier protein (ACP) domains naturally fused by this linker and its homologs, resemble beads on a string with little or no interactions between domains.29 In this case, however, the FabA‐FabA homodimer clearly forms an intramolecular dimeric structure similar to the intermolecular dimer that FabA naturally forms, as evidenced by size exclusion chromatography. Similarly, FabZ‐FabZ forms the expected trimer of dimers that the parent FabZ also forms. Thus, although the chosen linker is predicted to have an extended conformation, it is flexible enough to allow the formation of a functionally competent fused dimer. The three‐dimensional model for FabZ‐FabZ shown in Figure 5, shows how an elongated linker (magenta) could accommodate a native‐like structure in the fused dimer.

Figure 5.

Three‐dimensional model of FabZ‐FabZ. Panel A shows a three‐dimensional model was built for FabZ‐FabZ using the I‐Tasser platform and visualized using PyMol.30, 31, 32, 33 The model contains the FabZ repeats (green and cyan respectively), the linker (magenta) and the active site His and Glu residues (red spheres). According to this model, an elonogated linker could accomodate a fused dimeric arrangement that is similar to that of the native dimer of FabZ from Pseudomonas aeruginosa (shown in Panel B; PDB Id 1U1Z 34).

Taken together, our results open the possibility of employing fused dimerization to enhance the activity of naturally dimeric enzymes along the fatty acid biosynthesis pathway. The fused enzymes could become a new orthogonal tool in current efforts to increase fatty acid biosynthesis in bacterial cultures. Future work should be aimed at assessing how these fused enzymes work within the metabolic network and whether they increase overall production or accumulation of fatty acids in microbial cultures.

Materials and Methods

Primer design

DNA sequences for E. coli fabA and fabZ genes were obtained from NCBI (Accession no. ACB02154.1 and BAA77855.1, respectively). The pfaA gene from P. profundum had been previously cloned as described in Trujillo et al.29 The linker region corresponded to amino acids 1511–1538 of the pfaA gene from P. profundum (Accession No. CAG19871.1). The primers for the amplification of fabA and fabZ contained overhangs of additional sequence that are complementary to either the 5′ or 3′ terminus of the linker sequence (Table SI, Supporting Information). The reverse primers for the amplification of fabA/Z contained overhangs complementary the 5′ terminus of the linker sequence, whereas forward primers contained overhangs complementary to the 3′ terminus of the linker sequence. All primers were purchased from the RCMI Core Lab at UPR Medical Sciences Campus after verification of secondary structure and primer dimer formation using the online Thermo Scientific Multiple Primer Analyzer tool (http://www.thermoscientificbio.com/webtools/multipleprimer).

Construction of fused dimers and molecular cloning

A general scheme for our generation of fused dimers by overlap PCR is presented in Figure S1, Supporting Information. All PCRs were done using PfuUltra II Fusion HS DNA Polymerase (Agilent). All PCR reaction were preceded by a denaturation step at 95°C for 2 min and finished with a final extension step at 68°C for 3 min. For the amplification of the DNA corresponding to the linker region of pfaA, 100 ng of plasmid DNA containing the ACP domains from P. profundum 29 were mixed with primers 9 and 10. DNA was denatured at 95°C for 1 min, followed by annealing at 45.8°C for 1 min and finished with an extension 68°C for 1 min. For the amplification of FabA, ∼2 ng of genomic DNA from DH10B E. coli cells were mixed with primers 1 and 4. For the amplification of FabZ, ∼2 ng of genomic DNA were mixed with primers 5 and 8. The PCR parameters were the same as for their respective homodimers. For the construction of fused dimers, a number of sequential PCR reactions were performed as outlined in Tables S2 and S3, Supporting Information. For the construction of homodimers, PCR reactions were run for 30 cycles without primers, and then 15 more cycles with primers. The final PCR reactions for the construction of heterodimers were completed for 30 cycles with primers. Individual PCR reaction products were separated by electrophoresis on an agarose gel (2%) and purified using the QIAQuick Gel Extraction Kit (Qiagen) (Fig. S2, Supporting Information). The pfaA linker adds 84 base pairs to each construct. These 84 base pairs correspond to a 28 amino acid sequence of STQNVAIQTAAPVASASNGLDAAQVQGT. All constructs were cloned into pET‐200/TOPO (Invitrogen), transformed into E. coli strain TOP10 (Invitrogen), and their sequences were fully verified using the Sanger method.

Expression and purification

E. coli BL21(DE3)‐Codon Plus RIL cells (Invitrogen) were transformed with each plasmid and grown in liquid Luria‐Bertani medium supplemented with 0.4% glycerol, 1% glucose, and contained kanamycin (100 mg/L) and chloramphenicol (25 mg/L). Cultures were grown at 37°C, 250 rpm until OD₆₀₀ = 0.2–0.3, at which point the temperature was lowered to 22°C. Protein expression was induced with 1 mM IPTG once the OD₆₀₀ reached 0.5–0.6. After 5 h, the cells were harvested by centrifugation at 4°C and 11,000g on a Sorvall Lynx 4000 Centrifuge using a Fiberlite™ F14‐14 × 50cy Fixed‐Angle Rotor (Thermo). Samples were stored at −20°C overnight. Pellets were resuspended in lysis buffer [(50 mM Tris, 150 mM NaCl, 1 mM DTT, 20% glycerol, pH 7.8 for FabA and FabA‐FabA), (20 mM Tris, 500 mM NaCl, 1 mM DTT, 20% glycerol, pH 7.8 for FabZ, FabZ‐FabZ, FabA‐FabZ, and FabZ‐FabA)] in the presence of lysozyme (10 mg/mL), DNase (1 mg/mL), 2× protease inhibitor cocktail (Pierce) and sonicated. The lysates were collected by centrifugation (11,000g, 4°C, 30 min). The insoluble lysis pellet containing FabZ was solubilized using the inclusion body protocol from BugBuster (Novagen). The soluble lysates (and separately the resolubilized lysis pellet) were poured through a column filled with Ni‐Sepharose (Sigma) that had been equilibrated with the corresponding buffer + 5 mM imidazole and washed twice with the same buffer + 10 mM imidazole. His‐tagged proteins were eluted in the in the corresponding buffer containing 200 mM imidazole. Purities of elution fractions were analyzed by SDS‐PAGE. Protein yields were calculated using Nanodrop A280 quantification and dividing total milligrams of protein per volume of culture. Due to the low abundance of aromatic residues in FabZ and FabZ‐FabZ, they were quantified at 205 nm using the Scopes method as recommended by the Nanodrop instrument manufacturer (Thermo).35

Enzyme activity assays

The activity for all enzymes was measured in the reverse direction by incubation with crotonyl‐CoA (Sigma) as substrate. This reverse reaction format has been employed previously by others because the equilibrium point of the reaction favors enoyl hydration.10, 36 Enzymatic reactions were followed spectrophotometrically by monitoring the absorbance at 270 nm as a function of time in a 96‐well plate format on a Spectramax 190 instrument (Molecular Devices). Reactions were carried out in a buffer containing 25 mM Tris, 150 mM NaCl, 10% glycerol, pH 7.8, with varying concentrations of crotonyl‐CoA in the range of 1–2500 μM. Enzyme concentrations were fixed to 11 μM FabA‐FabA, 22 μM FabA, 8 μM FabZ‐FabZ, 16 μM FabZ. All reactions were measured at 37°C for 10 min, with a 1 s shake every 5 s. Pre‐enzyme data was also collected for 10 min. The values for the absorbance slopes were converted to units of μmole of product per minute by using the equation μmoles/min = (slope × L)/(b × ɛ), where the slope is given by the instrument (mAU/s), b is the path length (0.61cm), ɛ is the molar extinction coefficient resulting from the loss of a double bond as defined by the difference in absorbance between crotonyl‐CoA and β‐hydroxybutyryl‐CoA at a particular wavelength (971.4 M⁻¹ cm⁻¹), and L is the reaction volume (150 μL). The data was normalized by dividing initial velocity values by total amount (μg) of enzyme. These values were plotted as a function of substrate concentration to generate Michaelis–Menten saturation curves. Dividing by the K _m values yielded k _cat/K _m. Data analysis was performed with GraphPad Prism 6 and using the first 5 min of the reaction.

Size exclusion chromatography

The purified proteins were infused into a Superdex 200 Increase 10/300 GL column (GE Healthcare) equilibrated in the NiNTA lysis buffers minus glycerol. Each run with the enzymes was preceded by a run with a mixture of standard proteins (GE Healthcare) [aprotinin (6,500 Da), ribonuclease (13,700 Da), ovalbumin (44,000 Da), conalbumin (75,000 Da), aldolase (158,000 Da), ferritin (440,000 Da)] to generate a Kav versus logMW curve, where Kav = (Ve – Vo)/(Vc – Vo). Enzymes were eluted at a flowrate of 0.8 mL/min, and the elution volumes were determined using the Unicorn software integration function. The resulting standard curve was used to estimate the molecular weight for our proteins (Fig. S3, Supporting Information).

Three‐dimensional protein models

Models for FabA‐FabA and FabZ‐FabZ were built using the I‐TASSER (Iterative Threading ASSEmbly Refinement) server and visualized using PyMol.30, 31, 32, 33 The I‐Tasser server selected the crystal structure of the FabA dehydratase from E. coli (PDB ID: 1MKA) as the best template for FabA‐FabA and the FabZ dehydratase from Pseudomonas aeruginosa (PDB ID: 1U1Z) as the best template for FabZ‐FabZ (shown in Fig. 5).

Conflict of Interest

The authors declare that they have no conflict of interest with the content of this article.

Supporting information

Supporting Information