Mechanistic and Structural Insights into the Regioselectivity of an Acyl-CoA Fatty Acid Desaturase via Directed Molecular Evolution*

Membrane-bound fatty acid desaturases and related enzymes play a pivotal role in the biosynthesis of unsaturated and various unusual fatty acids. Structural insights into the remarkable catalytic diversity and wide range of substrate specificities of this class of enzymes remain limited due to the lack of a crystal structure. To investigate the structural basis of the double bond positioning (regioselectivity) of the desaturation reaction in more detail, we relied on a combination of directed evolution in vitro and a powerful yeast complementation assay to screen for Δx regioselectivity. After two selection rounds, variants of the bifunctional Δ12/Δ9-desaturase from the house cricket (Acheta domesticus) exhibited increased Δ9-desaturation activity on shorter chain fatty acids. This change in specificity was the result of as few as three mutations, some of them near the putative active site. Subsequent analysis of individual substitutions revealed an important role of residue Phe-52 in facilitating Δ9-desaturation of shorter chain acyl substrates and allowed for the redesign of the cricket Δ12/Δ9-desaturase into a 16:0-specific Δ9-desaturase. Our results demonstrate that a minimal number of mutations can have a profound impact on the regioselectivity of acyl-CoA fatty acid desaturases and include the first biochemical data supporting the acyl-CoA acyl carrier specificity of a desaturase able to carry out Δ12-desaturation.

Membrane-bound fatty acid desaturases and related enzymes play a pivotal role in the biosynthesis of unsaturated and various unusual fatty acids. Structural insights into the remarkable catalytic diversity and wide range of substrate specificities of this class of enzymes remain limited due to the lack of a crystal structure. To investigate the structural basis of the double bond positioning (regioselectivity) of the desaturation reaction in more detail, we relied on a combination of directed evolution in vitro and a powerful yeast complementation assay to screen for ⌬x regioselectivity. After two selection rounds, variants of the bifunctional ⌬12/⌬9-desaturase from the house cricket (Acheta domesticus) exhibited increased ⌬9-desaturation activity on shorter chain fatty acids. This change in specificity was the result of as few as three mutations, some of them near the putative active site. Subsequent analysis of individual substitutions revealed an important role of residue Phe-52 in facilitating ⌬9-desaturation of shorter chain acyl substrates and allowed for the redesign of the cricket ⌬12/⌬9-desaturase into a 16:0-specific ⌬9-desaturase. Our results demonstrate that a minimal number of mutations can have a profound impact on the regioselectivity of acyl-CoA fatty acid desaturases and include the first biochemical data supporting the acyl-CoA acyl carrier specificity of a desaturase able to carry out ⌬12-desaturation.
Fatty acid desaturases are nonheme diiron-containing enzymes that introduce double bonds within fatty acyl chains and belong to two main groups. The soluble desaturases are localized in the stroma of plastids where they typically introduce the first double bond in saturated fatty acids esterified to an acyl carrier protein. In contrast, the majority of the fatty acid desaturases are integral membrane proteins recognizing acyl chains esterified to either CoA or phospholipids. Although a similar diiron active site is present in membrane-bound desaturases, they are unrelated to soluble desaturases and are believed to have evolved independently (1). Although crystal structures for soluble desaturases from both castor (Ricinus communis) and the English Ivy (Hedera helix) (2,3) have paved the way for engineered variants displaying intriguing novel catalytic activities and specificities (4 -6), our understanding of the structure-function relationship of membrane-bound desaturases remains limited and scattered at best. Consequently, engineering of this class of desaturase enzymes has lagged behind that of the soluble desaturases.
Historically, three main types of regioselectivity have been distinguished (7,8). Desaturases with ⌬x regioselectivity introduce the first double bond x carbon atoms away from the carboxyl end. A second group comprises y-desaturases that insert a double bond at position y referenced from the methyl end. Finally, vϩz-desaturases require a preexisting double bond as a reference point (v) and generate the new double bond z carbon atoms further along the acyl chain. These modes of regioselectivity are not mutually exclusive, as illustrated by several bifunctional ⌬12/3 fungal desaturases (9 -12).
Thus far, membrane-bound fatty acid desaturases have been the subject of only a limited number of regioselectivity studies. Site-directed mutagenesis and the exchange of topological features between homologous fungal vϩ3-, bifunctional vϩ3/3-, and 3-desaturases have pointed to the C-terminal end and residues in the neighborhood of and in between the first two conserved histidine boxes as important structural determinants for vϩ3 versus 3 regioselectivity (9, 10, 13). Although such experiments have their merit, they rely heavily on sequence information. Typically, sets of homologous desaturases displaying different regioselectivities are required to identify potential amino acids or regions that might be involved in the regioselectivity outcome.
In an attempt to further increase our understanding of the structural principles that determine the regioselectivity of membrane-bound fatty acid desaturases, we relied on directed evolution in vitro rather than on more rational protein design principles that formed the basis of previous studies. Because directed evolution mimics Darwinian evolution in vitro, protein engineering puzzles can be tackled without the need for detailed structure-function information (14,15). The recently cloned bifunctional ⌬12/⌬9-desaturase from the house cricket (Acheta domesticus) (16) was chosen as a target because of the high amino acid sequence identity (65%) it shares with the ⌬9-desaturase from the same insect (17). Because the former enzyme probably evolved from a ⌬9-desaturase, it provides a unique opportunity to study for the first time the structure-function relationship behind ⌬x and vϩ3 types of regioselectivity. Two cycles of random mutagenesis and selection for ⌬x regioselectivity by genetic complementation of an ole1⌬::kanMX4 yeast deletion strain yielded three desaturase mutants exhibiting significant changes in their chain length specificity and regioselectivity. Analysis of individual contributions of each selected substitution allowed for the subsequent design of a highly specific ⌬9-desaturase mutant that only differed in 3 amino acids from the wild type enzyme. In addition, we provide biochemical evidence for the acyl-CoA specificity of the wild type cricket ⌬12/⌬9-desaturase and the most active of the evolved mutants, setting these enzymes apart from other ⌬12-desaturases that typically recognize membrane phospholipids.
Functional Expression in Yeast-Yeast transformations were carried out using the yeast-1 transformation kit (Sigma-Aldrich, Castle Hill, New South Wales, Australia) and 1 g of plasmid DNA according to the manufacturer's instructions. Transformants were selected on drop-out agar plates containing 2% glucose and lacking uracil (SC-Ura; 2 for selection of pYES2-derived vectors) or tryptophan (SC-Trp; for selection of pYES3-derived vectors). For transformations involving the ole1⌬ strain, plates were supplemented with 0.5 mM 16:1 ⌬9 , 0.5 mM 18:1 ⌬9 , and 1% Tergitol. Colonies that developed after 3-5 days at 30°C were tested for the presence of the appropriate plasmid DNA in a colony PCR.
Precultures of ole1⌬ transformants were set up in 5 ml of SC-Ura medium supplemented with 2% glucose, 1% Tergitol, and any exogenous fatty acid required. After incubation overnight at 28°C, yeast cells were washed with 5 ml of water and resuspended in 5 ml of SC-Ura induction medium containing 2% galactose, 1% Tergitol, and the appropriate fatty acid. When a different exogenous fatty acid was used during induction, two extra washing steps using 5 ml of 1% Tergitol and 5 ml of 0.5% Tergitol were included. Desaturation of 16:0 and 18:0 was assessed by adding 0.5 mM 17:1 ⌬10-cis to preculture and induction medium. Conversion of 18:1 ⌬9 was determined by growing cultures in the presence of 0.5 mM 18:1 ⌬9-cis during preculture and induction steps. For yeast feeding experiments involving 16:1 ⌬6-cis or 18:1 ⌬9-trans substrates, 0.5 mM of either fatty acid was added to the induction medium, whereas precultures were supplied with 0.25 mM 17:1 ⌬10-cis . Induction cultures were incubated for 3 days at 28°C and washed with equal volumes of 1% Tergitol, 0.5% Tergitol, and water to remove any unincorporated fatty acids. Final yeast pellets were freeze-dried and stored at Ϫ20°C until further analysis.
Random, Targeted, and Saturation Mutagenesis-Random mutagenesis of the cricket ⌬12/⌬9-desaturase gene was carried out as described previously (19) except for some minor modifications. The error-prone PCR mixture consisted of 100 ng of pXZP282 plasmid DNA, 0.2 mM each dNTP, 5 units of Taq DNA polymerase (Stratagene, Jolla, CA), and 0.15 mM MnCl 2 . Primers (10 pmol of each) flanking the desaturase gene were chosen to avoid mutations in the GAL1 promotor and the CYC1 transcriptional terminator. The PCR program consisted of 94°C for 2 min, 33 cycles of 94°C for 20 s, 63°C for 30 s, and 72°C for 1 min 30 s and a final extension step of 72°C for 10 min. Amplified products were purified using the QIAquik gel extraction kit (Qiagen, Doncaster, Victoria, Australia) and subsequently used as megaprimers to amplify the entire pXZP282 vector backbone with the high fidelity PfuUltra DNA polymerase (Stratagene). The PCR mixture contained 100 ng of pXZP282 template, 500 ng of the purified error-prone PCR product, 2.5 units of PfuUltra DNA polymerase, 1ϫ PfuUltra HF reaction buffer, and 0.2 mM of each dNTP in a final volume of 50 l. The thermocycling program included an initial denaturation step at 95°C for 2 min followed by 30 cycles of 95°C for 50 s, 60°C for 50 s, and 72°C for 10 min. Prior to transformation, each PCR mixture was digested with 20 units of DpnI (New England Biolabs, Ipswich, MA) for 2 h at 37°C to remove the original pXZP282 template DNA.
The library diversity was assessed by sequencing five randomly selected clones after transforming an aliquot into Escherichia coli DH5␣. On average, 5.8 missense mutations were found in each random mutant, and the mutation frequency was estimated at 0.75%.
ole1⌬ Complementation Assay-The desaturase library was divided in 10-l aliquots, transformed into ole1⌬ using the yeast-1 transformation kit (Sigma-Aldrich), and plated on SC-Ura medium containing 2% glucose, 0.5 mM 16:1 ⌬9 , and 1% Tergitol. Plates were kept for 5 days at 30°C, after which ole1⌬ transformants were collected in 20 ml of water. A 500-l aliquot of the resuspended library was diluted until an A 600 of 0.5 was reached and plated on SC-Ura selection medium containing 2% galactose (3 ml/150-mm Petri dish for optimal plating density). To confirm the increased level of complementation when compared with the parent of the respective selection round, individual colonies that developed after 6 -7 days at 30°C were tested in an ole1⌬ complementation assay as described below prior to lipid analysis. In addition, the integrity of the GAL1 promotor of each selected desaturase variant was verified by sequencing to rule out increased expression due to improved promotor efficiency.
Assaying single clones for ole1⌬ complementation was done by resuspending cells in 100 l of water. After adjusting the A 600 to 1, each sample was further diluted 3 ϫ 0.1, and 10 l of each dilution was spotted on SC-Ura agar plates containing 2% glucose or 2% galactose. Cell growth was scored after 4 -5 days at 30°C. The absence of growth on the glucose plates confirmed that complementation was not a result of adaptation of the ole1⌬ host itself.
Desaturase-Elongase Coexpression in Yeast-S288C precultures were grown overnight at 28°C in 15 ml of SC-Ura-Trp medium containing 2% glucose and resuspended the following day in 45 ml of fresh medium. After a second overnight incubation, cell pellets were washed with an equal volume of water and resuspended in 45 ml of SC-Ura-Trp induction medium containing 2% galactose. At three different time points (6, 12, and 24 h), separate cell cultures were washed with an equal volume of water. A 5-ml aliquot of each sample was used for total lipid analysis. Pellets obtained from the remaining 40 ml of cell suspensions were resuspended in 1 ml of water, centrifuged, and used for lipid fractionation.
Lipid Analysis-Fatty acid methyl esters were prepared from total yeast lipids by transesterification with 0.75 ml of 1 N methanolic-HCl (Supelco, Bellefonte, PA) for 3 h at 80°C. After adding 0.5 ml of 0.9% NaCl to each sample, methylated fatty acids were extracted with 0.5 ml of hexane and analyzed by GC as described previously (16). The proportion of each fatty acid as a weight percentage of total fatty acids was based on the integrated peak area as calculated by the Agilent Chemstation software (Revision B.02.01.SR1 (260), Agilent, Palo Alto, CA) and with 17:1 ⌬10-cis response based on the 16:1 ⌬9-cis standard. To confirm the position of the introduced double bond, fatty acid methyl esters were derivatized with 4,4-dimethyloxazoline and analyzed by GC-MS as described previously (16).
For analysis of different lipid classes, total lipids were extracted from yeast pellets using the method described by Bligh and Dyer (20). Neutral lipids (NL) and polar lipids including phosphatidylcholine (PC) were fractionated by TLC using precoated silica gel plates (Silica Gel 60; Merck) and a solvent system comprising chloroform/methanol/acetic acid/water (85/15/10/3.5, v/v). The lipid spots impregnated in silica were visualized using iodine vapor and identified according to the authentic standards run alongside. After isolating lipid spots from the TLC plates, fatty acid methyl esters were prepared and analyzed by GC.
Topology Prediction-Topology predictions for the cricket ⌬9and ⌬12/⌬9-desaturases were based on the TMPRED (21) and TMHMM version 2.0 (22) algorithms and Kyte and Doolittle (23) hydropathy profiles using a window size of 19 amino acids. Both models were in close agreement with predicted and validated topologies of the yeast and mouse ⌬9 acyl-CoA desaturases (24,25).
Statistical Analysis-One-way analysis of variance and pairwise two-tailed t tests (Holm-Bonferroni) were done with R-2.11.1 for Windows.
To obtain a better understanding of possible constraints imposed by the shape of the binding pocket of the cricket ⌬12/ ⌬9-desaturase onto the various C16 and C18 substrate molecules, we performed two additional ole1⌬ yeast feeding experiments. In the presence of the 18:1 ⌬9-trans , no further desaturation was observed (data not shown). The cricket ⌬12/ ⌬9-desaturase therefore requires a preexisting double bond in 18:1 ⌬9 in the cis-conformation in accordance to other typical vϩ3-desaturases. Because 18:1 ⌬9-cis serves as a substrate for ⌬12-desaturation, we next investigated whether 16:1 ⌬6-cis might also be accommodated into the binding pocket and undergo ⌬9-desaturation. Again no desaturation product was observed (data not shown), implying that the ⌬9-desaturation of 16:0 as catalyzed by the cricket ⌬12/⌬9-desaturase is an example of ⌬x regioselectivity.
Random Mutagenesis of the Cricket ⌬12/⌬9-desaturase-The low ⌬9-desaturation activity on 16:0 led us to investigate whether the cricket ⌬12/⌬9-desaturase could be "evolved back" to the proposed ancestral ⌬9-desaturase form. Because ole1 gene-disrupted yeast strains require unsaturated fatty acids or the heterologous expression of a ⌬9-desaturase for growth on 3 X. R. Zhou, personal communication. minimal medium (18), they are an ideal screening tool for desaturase variants that display changes in regioselectivity. Because unsaturated fatty acids other than 16:1 ⌬9 and 18:1 ⌬9 have also been reported to restore the growth of ole1 yeast cells (26), such a complementation assay would select for ⌬x regioselectivity in general rather than for double bonds exclusively at the ⌬9 position.
Random mutagenesis of the cricket ⌬12/⌬9-desaturase followed by screening for ole1⌬ complementation yielded a single desaturase mutant that allowed for minimal growth restoration of ole1⌬ yeast cells on minimal medium (Fig. 1). The isolated desaturase variant 1.1 contained two amino acid substitutions (E37G and A98V) (Fig. 2B and Table 1) and was used as a parent for a subsequent similar round of directed evolution. Three second generation mutants 2.1-2.3 were able to complement for the ole1⌬ deletion in yeast (Fig. 1). Sequencing revealed four new mutations (E22K, F52Y, I89V, and K136R) (Fig. 2B and Table 1). When mapped upon a predicted topological model of the cricket ⌬12/⌬9-desaturase, all targeted residues appeared to be either exposed into the cytosol or situated near the membrane surface (Fig. 2A). Two substitutions, A98V and K136R, are located close to the active site within the first and near the second conserved histidine box, respectively. Remarkably, the original mutation A98V selected during the first round was absent in mutant 2.2.
Analysis of fatty acid methyl esters revealed some interesting differences between the activity profiles of the selected desaturase mutants and the wild type enzyme (Figs. 3 and 4). The random mutants 2.1-2.3 all exhibited between 3.9-and 5.9-fold significantly improved ⌬9-desaturation activities on 16:0 (p Ͻ 0.01), which explains their selection based on the ole1⌬ complementation assay. In addition, expression of desaturase mutants 2.2 and 2.3 resulted in significantly increased ⌬12desaturation activities on 18:1 ⌬9 (p Ͻ 0.01). Comparison of peak retention times with those of standard isomers showed that the newly introduced ⌬9 and ⌬12 bonds in 16:1 and 18:2 were in the cis-conformation (data not shown). In addition, increased levels of 14:1 ⌬9 were also observed with the regioselectivity of the double bond confirmed by 4,4-dimethyloxazo-   APRIL 15, 2011 • VOLUME 286 • NUMBER 15

Engineering of Fatty Acid Desaturation Regioselectivity
line derivatization (Fig. 4). Interestingly, only minimal ⌬9-desaturation activity was detected on 18:0. Analysis of Individual Desaturase Mutations-To better understand the individual contributions of each of the selected random amino acid substitutions, we constructed a series of single mutants of the cricket ⌬12/⌬9-desaturase (Table 1). In addition, homologous mutations were also introduced separately into the cricket ⌬9-desaturase (Table 1). Because residues Glu-22 and Glu-37 of the cricket ⌬12/⌬9-desaturase are not conserved in the cricket ⌬9-desaturase (Ala-22 and Arg-39, respectively), we also exchanged amino acids at these two positions between the two desaturases (Table 1). For reasons of clarity, amino acid substitutions and residues of the cricket ⌬9 and ⌬12/⌬9 desaturases are referred to below as AdD9 and AdD12/9, respectively.
Results obtained with the cricket ⌬12/⌬9and ⌬9-desaturase single mutants point to an important role of AdD12/9-F52 and AdD9-F54 for specific activity on 18:1 ⌬9 and 18:0 fatty acid substrates, respectively. The functional role of this particular residue in both desaturases was therefore explored further by saturation mutagenesis. Analysis of the individual AdD12/9-F52 saturation mutants revealed that only the original tyrosine substitution and to a lesser extent asparagine resulted in an increased ⌬9/⌬12-desaturation ratio (data not shown). Similarly, saturation mutagenesis of AdD9-F54 showed that the shift in substrate specificity toward 16:0 is unique to the AdD9-F54Y substitution (data not shown).
Engineering of a Cricket ⌬12/⌬9-desaturase Variant with Increased ⌬9 Regioselectivity-Based on the effect of the single amino acid substitutions, we explored the possibility of further altering the regioselectivity of the cricket ⌬12/⌬9-desaturase by recombining some of the selected random mutations. In the first experiment, we redesigned variant 2.2 in an attempt to further improve ⌬12-desaturation activity on 18:1 to levels obtained with the T. castaneum and A. thaliana ⌬12-desaturases (Fig. 6). The AdD12/9-A98V substitution that was lost during the second selection round was reintroduced, and residue AdD12/9-E37 was restored. Desaturation activity of the triple mutant was reduced and comparable with the wild type enzyme. Secondly, we attempted to improve the ⌬9/⌬12-desaturation ratio of variant 2.1 by introducing AdD12/9-I89V and reversing AdD12/9-E37G. The %16:0/%18:1 conversion ratio of the new triple variant was 3.1-and 13.5-fold higher relative to that of the parent mutant 2.1 and the wild type cricket ⌬12/⌬9-desaturase, respectively, illustrating a further increase in specificity toward 16:0 (Fig. 6).

Engineering of Fatty Acid Desaturation Regioselectivity
Acyl Carrier Specificity of Wild Type and Mutant Desaturases-Previous in vitro biochemical studies with isolated insect microsomes have demonstrated that ⌬12-desaturation takes place on fatty acid substrates esterified to CoA rather than on phospholipid-linked acyl groups (27,28). However, such acyl-CoA specificity has not been confirmed directly for the ⌬12/ ⌬9and ⌬12-desaturases recently cloned from the house cricket and the red flour beetle (T. castaneum), respectively (16). To rule out differences in the acyl carrier specificity of the wild type cricket ⌬12/⌬9-desaturase and the high activity mutant 2.2, we performed a ⌬9-elongase coexpression experiment as described by Sayanova et al. (29). Any acyl-CoA ⌬12desaturase hereby provides the direct substrate for subsequent elongation in the same metabolic pool by the ⌬9-elongase from I. galbana, which only accepts 18:2 ⌬9,12 -CoA and 18:3 ⌬9,12,15 -CoA substrates. Elongation should therefore proceed more efficiently than in the case of a phospholipid-specific desaturase such as the A. thaliana FAD2 ⌬12-desaturase, which produces 18:2 ⌬9,12 in a different lipid fraction.
Analysis of total lipids confirmed elongation of the ⌬12-desaturation product 18:2 ⌬9,12 to 20:2 ⌬11,14 after a 12-and 24-h induction ( Table 2). Expression of the A. thaliana FAD2 ⌬12desaturase resulted in the highest 18:2 ⌬9,12 product levels. Despite the lower desaturation activities of the wild type and mutant cricket desaturases, between 40 and 50% of the available desaturation product was further elongated by the coexpressed ⌬9-elongase. Sayanova et al. (29) reported a similar elongation efficiency (57%) when 18:2 ⌬9,12 was supplied to yeast cultures. Efficient coupling between desaturation and elongation is to be expected when both reactions take place in the acyl-CoA lipid fraction. In the case of the A. thaliana FAD2 ⌬12-desaturase, on the other hand, efficiencies of the subsequent elongation step were found to be markedly lower at both time points. Additional acyltransferases that are required to shuttle the desaturation products from the phospholipids to the acyl-CoA lipid pool thereby possibly restrict the flux between acyl-lipid-specific desaturation and acyl-CoA-dependent elongation.
To obtain further evidence of the acyl-CoA specificity of the wild type cricket ⌬12/⌬9-desaturase and its mutant 2.2, NL and different phospholipids of induced yeast cells were separated by TLC. Expression of the A. thaliana FAD2 ⌬12-desaturase resulted in an enrichment of 18:2 ⌬9,12 in PC after a 24-h induction ( Table  3). Such enrichment was not observed for either the wild type or the mutant cricket desaturases. Instead similar or even higher levels of 18:2 ⌬9,12 were detected in the NL. Similar results were obtained after a 12-and 6-h induction (data not shown). These findings are in agreement with previous results obtained with ⌬5and ⌬6-acyl-CoA desaturases from microalgae (30,31). Although we cannot rule out additional acyl-lipid-specific desaturation, the biochemical characterization of the wild type cricket ⌬12/⌬9-desaturase and mutant 2.2 provides evidence for the acyl-CoA substrate specificity of both enzymes.

DISCUSSION
In nature, the nutritionally important docosahexaenoic acid 22:6 ⌬4,7,10,13,16,19 is an apt example of the wide range in regioselectivity displayed by fatty acid desaturases. However, many other fatty acid modifications such as hydroxylation that are performed by closely related enzymes are restricted to only a handful of specific positions within the acyl chain of the fatty acid substrate (32). Therefore, a better understanding of the structure-function relationship determining the regioselectivity of fatty acid-modifying enzymes not only opens up further possibilities to manipulate double bond positioning but could also lead to novel and high value designer fatty acids   ⌬12-desaturase (TcD12des), A. thaliana FAD2 ⌬12-desaturase (AtD12des), wild type cricket ⌬12/⌬9desaturase (AdD12/9des), and targeted and random mutants of the cricket ⌬12/⌬9-desaturase. 16:0 and 18:1 conversion efficiencies (%) were calculated from relative substrate and product levels (percentage of total fatty acids) as 100 ϫ product/(substrate ϩ product). 16:0/18:1 conversion ratios are shown relative to the wild type AdD12/9des desaturase. Data represent means Ϯ S.D. of three replicate cultures.
containing various functional groups at tailored positions that are of immediate importance as renewable chemical feedstocks.
The evolution of fatty acid desaturases with novel regioselectivities in insects and other organisms is thought to have occurred via gene duplication or retrotransposition on multiple independent occasions (32)(33)(34)(35)(36)(37). Most likely, subsequent accumulation of mutations first resulted in a generalist desaturase displaying a broader substrate acceptance or exhibiting some degree of relaxed catalytic activity (38). Any low level side activity would then serve as a potential starting point for the evolution of a new enzymatic specificity such as altered regioselectivity. As an example, several bifunctional vϩ3/3-desaturases have been cloned from fungi and are believed to be an intermediate form during the change from vϩ3 toward 3 regioselectivity (9 -12). Unlike vϩ3-and 3-fatty acid desaturases, however, ⌬x-desaturases cluster together in a distinct phylogenetic class (33)(34)(35). The bifunctional cricket ⌬12/⌬9desaturase forms a notable exception given its homology to the ⌬9-desaturase from the same insect. Furthermore, its broad specificity and low general activity are typical for a generalist enzyme and hint at a relatively recent ⌬9-desaturase duplication event. As such, this particular desaturase provides a unique evolutionary and structural link between ⌬x and vϩ3 types of regioselectivity.
Despite the high amino acid identity between the cricket ⌬9and ⌬12/⌬9-desaturases, both enzymes differ markedly in their specificity. Unlike the ⌬9-desaturase, the ⌬12/⌬9-desaturase displays low but significant activity on a variety of saturated and mono-unsaturated acyl chains and uniquely combines both ⌬x and vϩ3 types of regioselectivity depending on the chain length of the fatty acid substrate. Particularly intriguing is the minimal ⌬9-desaturation activity on 18:0. A reduction in size at the bottom of the substrate binding channel has been known to influence acyl chain length specificity of soluble desaturases as fewer carbon atoms can be accommodated between the terminal double bond and the methyl end of the acyl chain (4,39). Although a different geometry has been suggested for the binding cavity of membrane-bound desaturases (40), it is tempting to speculate that a similar effect occurs in the cricket ⌬12/⌬9desaturase, thereby preventing a correct orientation of the 18:0 substrate with its C9-C10 carbon pair in an eclipsed conformation in close proximity to the diiron active site.
Generalist enzymes such as the cricket ⌬12/⌬9-desaturase are ideal starting points for directed evolution experiments aimed at improving promiscuous activities (41,42). Analysis of the fatty acid products of three desaturase variants that were selected using the ole1⌬ yeast complementation assay revealed significant shifts in their chain length specificity and regioselectivity preference (Figs. 3 and 4). Mutants 2.2 and 2.3 can be regarded as improved bifunctional ⌬12/⌬9-desaturases yielding significant elevated levels of 16:1 ⌬9 and 18:2 ⌬9,12 . Mutant 2.1 displayed increased ⌬9-desaturation of 16:0, whereas ⌬12desaturation activity on 18:1 was comparable with that of the wild type enzyme. Western blot analysis of ole1⌬ yeast cell lysates expressing wild type and mutant desaturases fused to an N-terminal FLAG epitope suggested that changes in the fatty acid product levels as observed for mutants 2.1 and 2.3 were the result of altered specificity and/or increased catalytic activity rather than improved protein stability or expression (data not shown). Such a stability or expression effect could not be completely ruled out for the high activity mutant 2.2, however. Coexpression of the I. galbana ⌬9-elongase in yeast allowed us to determine whether a possible change in the acyl carrier spec- Desaturation and elongation efficiencies (%) were calculated from relative substrate and product levels (% of total fatty acids) as 100 ϫ product/(substrate ϩ product). Data represent means Ϯ S.D. of three replicate cultures after a 6-, 12-, or 24-h induction.

AtD12des
AdD12  ificity of the desaturase mutants in combination with different substrate availabilities in the acyl-CoA and phospholipid pools of induced yeast cultures accounted for some of the observed differences in their fatty acid profiles. Analysis of desaturation and elongation products in NL and PC confirmed no change in acyl carrier specificity of the wild type cricket ⌬12/⌬9-desaturase and mutant 2.2 (Tables 2 and 3). This is to our knowledge the first biochemical evidence for the acyl-CoA specificity of a cloned ⌬12-desaturase and is of great interest for the production of polyunsaturated fatty acids such as docosahexaenoic acid in oilseed crops via an exclusive acyl-CoA-dependent pathway (31). Although unconfirmed, desaturase mutants 1.1, 2.1, and 2.3 are likely to display a similar specificity for acyl-CoA substrates because they differ in only 3-4 amino acids with mutant 2.2 and all share the AdD12/9-E37G substitution. Each of the selected desaturase random mutants contained three amino acid substitutions with AdD12/9-E37G and AdD12/9-A98V being shared by more than one desaturase variant ( Fig. 2B and Table 1). When we combined the mutations AdD12/9-F52Y, AdD12/9-I89V, and AdD12/9-A98V in the wild type ⌬12/⌬9-desaturase, a highly specific 16:0 ⌬9-desaturase was obtained that exhibited only minimal ⌬12-desaturation activity on 18:1 ⌬9 (Fig. 6). Our random and targeted mutagenesis results thus demonstrate that a limited number of amino acid changes can have a significant effect on the regioselectivity and chain length specificity of an acyl-CoA fatty acid desaturase. Similarly, the substitution of as little as 1 or 2 amino acids in some cyanobacterial and fungal acyl-lipid desaturases has been shown to change the positioning of the introduced double bond (10, 43).
Although the lack of a crystal structure precludes a detailed structural interpretation of the selected mutations and the majority of fatty acid desaturase mutagenesis studies are based on acyl-lipid desaturases, mapping onto a topological model of the cricket ⌬12/⌬9-desaturase and analysis of desaturase single mutants allowed for some interesting observations. Two mutations, AdD12/9-A98V and AdD12/9-K136R, were found to be located within or close to the first two conserved histidine boxes that together with the third histidine box are believed to form the active site (44). Residues in the vicinity of the histidine boxes and particularly in the amino acid stretch following the second histidine box have been reported to influence the catalytic reaction outcome as well as the chain length specificity of FAD2-related enzymes such as the Lesquerella fendleri ⌬12hydroxylase and the Crepis alpina ⌬12-acetylenase (45)(46)(47). Among the cricket ⌬12/⌬9-desaturase single mutants, AdD12/ 9-K136R displayed the largest increase in desaturase activity. A similar substitution, AdD9-K138R, in the cricket ⌬9-desaturase also increased catalytic activity. Therefore, this residue seems to be of importance for desaturation in general rather than influencing regioselectivity or chain length specificity.
Residues AdD9-F54 and AdD12/9-F52 appear to influence the specificity of the desaturation reaction in the cricket ⌬9and ⌬12/⌬9-desaturase, respectively. Exchanging AdD9-F54 for tyrosine shifted the substrate specificity of the ⌬9-desaturase from 18:0 to 16:0. Such a change toward shorter fatty acid substrates without an immediate effect on regioselectivity is reminiscent of soluble fatty acid desaturase mutants with reduced binding pockets behind the active site (4,39). In the cricket ⌬12/⌬9-desaturase, a single AdD12/9-F52Y mutation reduced ⌬12-desaturation significantly, whereas ⌬9-desaturation of 16:0 was comparable with the wild type enzyme. The effect of this mutation is more difficult to interpret because steric hindrance at the end of the substrate binding cavity does not explain the reduction in ⌬12-desaturation activity. Possibly, direct or long distance effects in the region that is in close contact with the CoA thioester or changes in the upper portion of the substrate binding channel interfere with a correct positioning of the 18:1 ⌬9 substrate relative to the active site. A similar mechanism was suggested for changes in the regioselectivity of soluble desaturases (3,4). In a topological model of the cricket ⌬12/⌬9-desaturase, AdD12/9-F52 is located in the first predicted transmembrane region close to the membrane surface. When the first two putative transmembrane helices were exchanged between the ⌬6 fatty acid desaturase and the ⌬8 sphingolipid desaturase from Borago officinalis, the chimeric desaturase displayed a preference for 16:0 and 14:0 fatty acids instead of 18:0 (48). Although the cricket desaturase recognizes a different lipid head group, our results suggest that a similar region is also involved in a substrate binding site in acyl-CoA desaturases.
In summary, our results demonstrate for the first time that directed evolution can provide a valuable alternative to previous targeted mutagenesis and domain-swapping studies when studying the structural basis behind regioselectivity of fatty acid desaturases. Key residues in the cricket ⌬12/⌬9-desaturase including AdD12/9-K136 and AdD12/9-F52 provide promising targets for further activity or specificity improvement by means of iterative or combinatorial saturation mutagenesis (39,49). The results presented in this work are a step forward in our understanding of the structure-function relationship of membrane-bound fatty acid desaturases and will aid in fully unlocking the remarkable enzymatic plasticity displayed by this class of enzymes.