Domain Swapping between Enterococcus faecalis FabN and FabZ Proteins Localizes the Structural Determinants for Isomerase Activity*

Anaerobic unsaturated fatty acid synthesis in bacteria occurs through the introduction of a double bond into the growing acyl chain. In the Escherichia coli model system, FabA catalyzes both the dehydration of β-hydroxydecanoyl-ACP and the isomerization of trans-2-decenoyl-ACP to cis-3-decenoyl-ACP as the essential step. A second dehydratase, FabZ, functions in acyl chain elongation but cannot carry out the isomerization reaction. Enterococcus faecalis has two highly related FabZ homologs. One of these, termed EfFabN, carries out the isomerization reaction in vivo, whereas the other, EfFabZ, does not (Wang, H., and Cronan, J. E. (2004) J. Biol. Chem. 279, 34489–34495). We carried out a series of domain swapping and mutagenesis experiments coupled with in vitro biochemical analyses to define the structural feature(s) that specify the catalytic properties of these two enzymes. Substitution of the β3 and β4 strands of EfFabZ with the corresponding strands from EfFabN was necessary and sufficient to convert EfFabZ into an isomerase. These data are consistent with the hypothesis that the isomerase potential of β-hydroxyacyl-ACP dehydratases is determined by the properties of the β-sheets that dictate the orientation of the central α-helix and thus the shape of the substrate binding tunnel rather than the catalytic machinery at the active site.

Anaerobic unsaturated fatty acid (UFA) 1 biosynthesis occurs through the insertion of a double bond into the growing acyl chain (Fig. 1), and the proportion of UFA produced is a critical determinant of the biophysical properties of biological membranes. In bacteria, fatty acids are synthesized using the dissociated, type II fatty acid synthase system in which each of the steps is catalyzed by distinct enzymes that are each encoded by separate genes (1,2). The key players in UFA synthesis in Escherichia coli were first defined by the isolation and characterization of UFA auxotrophs (3). The double bond is introduced at the 10-carbon intermediate by ␤-hydroxydecanoyl-ACP dehydratase, FabA (4), which is capable of both the removal of water to generate trans-2-decenoyl-ACP and the isomerization of this intermediate to the cis-3-decenoyl-ACP (1, 5) (Fig. 1). A second UFA auxotroph was isolated which corresponds to the fabB gene, which encodes ␤-ketoacyl-ACP synthase I (6). In fabA and fabB mutants, saturated fatty acid synthesis persists because of the presence of another dehydratase, FabZ (7), and another elongation-condensing enzyme, FabF (8,9). The FabZ isozyme (␤-hydroxyacyl-ACP dehydratase) is expressed ubiquitously in type II systems, cannot carry out the isomerase reaction, and is the only type of dehydratase that exists in most bacteria. Although FabA and FabZ have many primary sequence characteristics in common, bioinformatic analysis clearly divide the two subtypes. Specifically, there are distinct differences in the active site residues, an Asp in FabA and a Glu in FabZ, and FabA is a dimer (10) whereas FabZ is a hexamer (11). FabZ does play a role in UFA synthesis by functioning in the elongation of both saturated and unsaturated long chain acyl-ACP, whereas FabA most efficiently processes saturated chain lengths 10 carbons and shorter (12). Likewise, the FabF condensing readily elongates 16:1 to 18:1 UFA (9); however, the inability to support UFA synthesis in fabB mutants leads to the conclusion that FabF cannot elongate a key intermediate in UFA biosynthesis in vivo, most likely cis-3-decenoyl-ACP (1,2).
The availability of numerous bacterial genome sequences allows the reconstruction of type II fatty acid synthesis in these organisms using bioinformatics analysis tools. It is notable that fabA and fabB genes occur together in Gram-negative bacteria that produce UFA (13). However, many anaerobes that synthesize UFA do not have a recognizable fabA homolog in their genomes and also have a FabF rather than a FabB subtype of elongation-condensing enzyme. In these organisms UFA is synthesized by a different mechanism. Streptococcus pneumoniae produces straight chain saturated and monounsaturated fatty acids predominantly of 16-and 18-carbon chain lengths (14). This organism does not utilize a FabA-like mechanism for introducing a double bond into the growing acyl chain, but rather accomplishes this task using FabM, a trans-2, cis-3decenoyl-ACP isomerase (15). Enterococcus faecalis also has a fatty acid composition similar to E. coli, but also lacks a FabA, FabB, and FabM. However, E. faecalis has two FabZ homologs, and Wang and Cronan (16) show that one of these genes, now called FabN, functions as a dehydratase/isomerase analogous to FabA, whereas the other possessed only dehydratase activity (Fig. 1). Thus, one cannot predict the biochemical activity of this class of proteins based on bioinformatics.
The comparison of the x-ray structures of FabA (10) and FabZ (11), coupled with the analysis of site-directed mutants shows that the differences in the catalytic activities of the enzymes are not the result of distinct catalytic residues in the active site or the side chains of the residues that compose the substrate binding pocket (11). Both proteins adopt a "hot dog" fold with a long central ␣-helix (the hot dog) surrounded by several ␤-sheets (the bun). A detailed comparison of these closely related structures led to the hypothesis that the different biochemistry of these enzymes is caused by differences in the ␤-sheet structures that alter the orientation of the central ␣-helix and thus the shape of the active site tunnel preventing the substrate from adopting a cis conformation in FabZ. The existence of two even more closely related enzymes, EfFabN and EfFabZ, with different catalytic properties but identical active site residues, allows us the opportunity to test the working hypothesis using domain swapping experiments to determine whether the isomerase activity is caused by the structure of the ␤-strands that control the shape of the active site tunnel.
Cloning and Construction of Chimeric Enzymes-The EffabN and EffabZ genes were amplified from genomic DNA of E. faecalis by primer sets EfFabNstart/EfFabNend and EfFabZstart/EfFabZend. The PCR products were ligated into plasmid pCR2.1 and sequenced. The plasmids were isolated and digested with NdeI and BamHI, and the gene fragments were isolated and ligated into plasmid pET15b digested with the same enzymes to generate the EfFabN and EfFabZ expression vectors pYL3 and pYL4. Chimeras were generated by overlapping PCR method. Sequences of internal primers are listed in Table I. T7 promoter primer with reverse primers and T7 terminator primer with forward primers were used to generate PCR fragments corresponding to the specific region of the two enzymes. After purification from agarose gels, these fragments were mixed with the right combination to serve as template, and T7 promoter and T7 terminator primers were used to amplify the recombinant molecule. PCR products from this step were ligated to PCR 2.1 vector and sequenced to ensure the chimeric constructs were correct. The chimeric genes were cut out with NdeI and BamHI and transferred into pET15b to generate the expression vectors.
Protein Purification-The pET15b plasmids were used to transform E. coli Rosetta-competent cells for protein expression. The selected transformants were cultured in LB medium with antibiotic (50 g/ml carbenicillin and 34 g/ml chloramphenicol) at 37°C until A 600 reached 0.6. Isopropyl 1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM, and incubation continued overnight at 20°C. Cells were collected by centrifugation (6,000 rpm, 4°C, 15 min), and cell pellets were lysed using a French press. Soluble proteins were applied to a Ni 2ϩ -agarose column and washed with 40 mM imidazole-containing metal chelation affinity chromatography buffer (20 mM Tris-HCl, pH 7.4, 0.5 M NaCl). His-tagged proteins were eluted with 500 mM imidazole in the same buffer. Proteins were quantitated by the Bradford method (17). The purified proteins were stored at Ϫ20°C. Relevant characteristics of the plasmids used in this study are shown in Table II.
Purified EfFabN or EfFabZ proteins were applied to a Superdex 200 HR 16/60 column (Amersham Biosciences) and eluted with 20 mM Tris-HCl, pH 7.4, 1 mM EDTA, 100 mM NaCl, 1 mM dithiothreitol). The molecular masses of EfFabN and EfFabZ were estimated using globular protein standards and a calibration curve.
Enzymatic Assays-The ability of the individual His 6 -tagged EfFabN and EfFabZ to carry out the dehydratase and/or isomerase activities was measured using a reconstituted system essentially as described previously (12,15,22)  The key intermediate in anaerobic olefin formation is ␤-hydroxydecanoyl-ACP. This molecule is dehydrated to trans-2-decenoyl-ACP and then utilized by enoyl-ACP reductase (FabI) and the elongationcondensing enzymes to form saturated fatty acids. All of the dehydratases in E. coli and E. faecalis carry out this reaction, and all of the elongationcondensing enzymes function in the production of saturated fatty acids. UFAs are formed by the isomerization reaction of the EcFabA and Ef-FabN dehydratases. These specialized enzymes exhibit both dehydratase and isomerase activity against ␤-hydroxydecanoyl-ACP and form cis-3decenoyl-ACP, which is elongated by the special activity of either the EcFabB or EfFabF1 elongation-condensing enzymes.
GTCGCTTTGCCGATGCCagcagaagcgcggactt Chr7for aagtccgcgcttctgctGGCATCGGCAAAGCGAC mationally sensitive gel electrophoresis in 13% polyacrylamide gels containing 2.5 M urea. Electrophoresis was performed at 25°C and 32 mA/gel. The gels were dried, and the bands were quantitated using a PhosphorImager screen. In the experiments determining FabN isomerase activity, EcFabZ was added to the reaction mixture, and the reaction was incubated at 37°C for 30 min before the addition of EfFabN. Specific activities were calculated from the slopes of the plot of product formation versus protein concentration in the assay.

RESULTS AND DISCUSSION
Characteristics of E. faecalis fab Genes-An analysis of the type II fatty acid biosynthetic genes in E. faecalis shows that they are located in two clusters in the genome ( Fig. 2A). S. pneumoniae, a closely related bacteria that also synthesizes UFAs, has only one fab gene cluster in its genome corresponding to the large gene cluster in E. faecalis. A comparison of the predicted protein sequences of E. faecalis fab genes with that of S. pneumoniae showed that the 12-gene cluster contains all of the gene homologs of the S. pneumoniae fab gene cluster except fabM, which is an essential isomerase to make UFAs in S. pneumoniae (15,23). A separate 3-gene cluster contains homologs of E. coli fabI and fabF1, in addition to fabN, a fabZ homolog that possess the ability to introduce double bonds into growing acyl chains in vivo (16). FabT, the predicted transcriptional regulator in several Gram-positive bacterial fatty acid biosynthetic pathways (24), is located at the beginning of the 12-gene cluster. FabT belongs to the MarR superfamily, which are typically dimers that utilize a winged helix motif to bind a DNA palindrome (25,26). Often bacterial transcription factors are autoregulated, and their DNA binding motifs are located within their own promoter regions. One DNA palindrome was found in the promoter region of FabT ( Fig. 2A). Significantly, this same palindrome is found in the promoters of the fabI, fabF1, and fabK genes ( Fig. 2A). Unraveling the transcriptional regulation in this large cluster is beyond the scope of this study. The significance of the bioinformatics analysis is that it ties the 3-and 12-gene clusters for fatty acid biosynthesis and suggests that the fabI, fabF1-fabN, and fabK genes may be coordinately regulated.
Purification of EfFabN and EfFabZ-EfFabN and EfFabZ were cloned into pET15b vector and purified as described under "Experimental Procedures." Purified His-tagged EfFabZ and EfFabN had monomer molecular masses of Ϸ18 kDa based on SDS-gel electrophoresis, consistent with their primary sequence (Fig. 3). Recently, the crystal structure of Pseudomonas aeruginosa FabZ was solved, and it forms a classic "trimer of dimers" structure (11). Accordingly, PaFabZ behaves as a hexamer exhibiting a Stokes radius on gel filtration chromatography corresponding to a 112-kDa protein (11), whereas EcFabA is a dimer both in its x-ray structure (10) and in solution. Both EfFabN and EfFabZ are hexamers in solution as determined by gel filtration chromatography (Fig. 3), which illustrates that they have structures similar to that of PaFabZ.
Enzyme Activity of EfFabN and EfFabZ-We used a complex, multicomponent reconstituted fatty acid biosynthetic system described under "Experimental Procedures" to detect the dehydratase and isomerase activities of purified EfFabN and EfFabZ in vitro (Fig. 4). This assay used enzymes from different sources to synthesize substrate for the dehydratases and followed the same principals as used previously in the characterization of the FabA and FabM isomerases (12,15). The identify of the band indicated as ␤-hydroxy-cis-5-dodecenoyl-ACP was established by mass spectrometry (15

FIG. 4. EfFabN catalyzes the formation of cis unsaturated acyl-ACP intermediates in vitro.
The ability of the His 6 -tagged EfFabN and EfFabZ proteins to carry out the dehydratase and isomerase activities was measured using a reconstituted gel assay system. The enzymes used in each lane are indicated on the top. The assays contained the indicated Fab enzymes and were initiated with octanoyl-CoA and [2-14 C]malonyl-CoA. The appearance of ␤-hydroxy-C12: 1(⌬5c)-ACP indicated the ability of the dehydratase to isomerize C10:1(⌬2t)-ACP to C10:1(⌬3c)-ACP, which was elongated by SpFabF.
This in vitro gel assay system was used to evaluate the relative activities of these two enzymes. The EcFabD/MtFabH/ EcFabG system was used to present EfFabN and EfFabZ with ␤-hydroxydecanoyl-ACP as a dehydratase substrate. The formation of enoyl-ACP was measured as a function of EfFabN or EfFabZ concentration, and a specific activity calculated by linear regression curves (Fig. 5A). The EcFabD/MtFabH/Ec-FabG/EcFabZ system was used to provide the trans-2-enoyl-ACP substrate for measuring isomerase activity. The formation of an elongation product with an excess concentration of Sp-FabF arising from EfFabN isomerase activity as a function of EfFabN in the assay was used to determine the isomerasespecific activity of EfFabN (Fig. 5B). The dehydratase activity of EfFabN, 0.07 Ϯ 0.002 pmol/min/ng, was much less than that of EfFabZ, which was 2.2 Ϯ 0.07 pmol/min/ng, whereas the dehydratase activities of the two controls, EcFabA and EcFabZ, were 0.34 Ϯ 0.03 and 0.36 Ϯ 0.003 pmol/min/ng, respectively. Thus, EfFabN had the lowest dehydratase activity of the four enzymes in the coupled assay. The isomerase activity of Ef-FabN for the 10-carbon substrate was 0.6 Ϯ 0.03 pmol/min/g, whereas the specific activity of EcFabA under these same assay conditions was 52 Ϯ 1 pmol/min/g (Fig. 5B). Thus, FabN was 86-fold less efficient than EcFabA in the formation of cis-double bonds in the in vitro fatty acid synthase assay reconstituted with the indicated constellation of E. coli enzymes and ACP cofactor. Also, FabN was about 2-fold less active than the FabM isomerase required for unsaturated fatty acid synthesis in S. pneumoniae (15). Although this assay is capable of clearly distinguishing dehydratases from isomerases, the activities of the FabM and FabN isomerases are low compared with FabA. The reasons for the low activities of the Gram-positive isomerases in this reconstituted assay are attributed in part to the use of heterologous ACP and elongation-condensing enzymes in the experiment. These data are consistent with the inability of fabN expression to complement the growth defect in fabA(Ts) mutants (16).
Chimera Construction and Enzyme Activity-The working hypothesis developed from our structural analysis of EcFabA and PaFabZ (11) was that the isomerase activity depends on the shape of the substrate binding tunnel, which is controlled by the positioning of the ␤-strands (the bun) surrounding the long central helix (the hot dog). The existence of these two FIG. 5. Specific activities of EfFabN and EfFabZ. The specific dehydratase activities (A) assay was performed using ␤-hydroxy-C10-ACP as substrate, and the isomerase activities (B) were analyzed using trans-2-C10:1-ACP as substrate. The coupled assay using a series of Fab enzymes to generate the substrates was performed in the presence of different concentrations of either EfFabN or EfFabZ as described under "Experimental Procedures." Radiolabeled product formation was quantitated using a PhosphorImager. A standard curve of [ 14 C]malonyl-CoA was used to calibrate the instrument. closely related FabZ enzymes with different catalytic properties provided an opportunity to test this hypothesis concerning the basis for their distinct catalytic properties. The goal was to determine the minimal structural features required to convert a dehydratase (EfFabZ) into a dehydratase/isomerase (Ef-FabN). Structural models of EfFabN and EfFabZ were generated using the PaFabZ x-ray structure (11) as the template. There are several amino acid differences between EfFabN and EfFabZ which were predicted to reside in the substrate binding tunnel and which might be responsible for the isomerase activity of EfFabN. These include Asn-16, Ile-20, Thr-57, Ile-88, and Asn-92. We introduced each of these point mutations into EcFabZ and measured their activities. All of the mutants retained dehydratase activity, and none acquired the isomerase function (data not shown). The next approach was to construct a series of chimeric EfFabZ/N proteins to determine the minimum structural elements required to transform EfFabZ into an isomerase. The transition points for chimera construction were selected by based upon the structure of PaFabZ (11) and are mapped onto the primary sequences in Fig. 2B. EfFabN and EfFabZ both have high similarity to each other and PaFabZ (EfFabN has 55% similarity and 41.6% identity to PaFabZ; EfFabZ has 56.5% similarity and 43.5% identity to PaFabZ), and their structural folding patterns are clearly predicted to be similar to those determined for PaFabZ. The divisions between chimeras were made at the boundary of loops between the ␤-sheet structures with at least one identical residue in both proteins occurring after the joint (Fig.  2B). Fig. 2C depicts schematically the chimeras made from introducing different EfFabN domains into EfFabZ. All chimeras were hexamers as judged by gel filtration chromatography and retained catalytic activity. First, EfFabN and Ef-FabZ were divided to two parts, and a swap of the corresponding regions of EfFabN and EfFabZ was made to generate Chimera1 and Chimera2. The analysis of these two proteins showed that both of them retained the ability to dehydrate ␤-hydroxydecanoyl-ACP to enoyl ACP (Fig. 6,  lanes 7 and 8). On the other hand, only Chimera2 containing the C-terminal part of EfFabN possessed 43% of the isomerase activity of EfFabN as indicated by the formation of the ␤-hydroxy-cis-5-dodecenoyl-ACP (Fig. 6, lane 8). We then divided the C-terminal part of EfFabN into three pieces to narrow down the region. Chimera3 and Chimera4 had only dehydratase activity (Fig. 6, lanes 9 and 10), leading to the conclusion that ␤3 and the following loop region of EfFabN were essential requirements for isomerase activity. However, the fact that Chimera5 lacked isomerase activity (Fig. 6, lane  11) illustrated that ␤3 and following loop were not sufficient to support isomerase activity. The combination of ␤3 and ␤4 (Chimera6) possessed isomerase activity at 38% of the Ef-FabN level (Fig. 6, lane 12), whereas the combination of ␤3 with ␤5 and ␤6 (Chimera7) did not (Fig. 6, lane 13). These data show that the ␤3-␤4 region of EfFabN was necessary and sufficient to transform EfFabZ from a dehydratase to a dehydratase/isomerase.
The isomerase activity of Chimera6 showed the important role of the combination of the ␤3/␤4 strands in specifying isomerase activity. Two residues, Phe-86 from ␤3 and Val-108 from ␤4 in EfFabZ, were candidates for mediating this effect based on the modeling of EfFabN and EfFabZ using the struc- The crystal structure of FabZ (PDB code 1U1Z) from P. aeruginosa shows the location of helix ␣3 (hot dog) relative to ␤-strands ␤3 and ␤4 within the curved ␤-sheet (the bun), which is characteristic of isomerase-negative dehydratases. Helix ␣3 is shown in orange, the ␤-sheets are yellow, and ␤3 and ␤4 sheets are magenta. The active site residues that span the interface with the associated monomer (blue/ green) are indicated. Helix ␣1 would obscure ␣3 in this orientation and has been removed for clarity. Superimposed on the FabZ structure are helix ␣3 from the E. coli FabA structure (transparent) and the bound inhibitor (gray) that was captured in the complex (PDB code 1MKA). FabZ and FabA were structurally aligned based on their ␤-sheets, and this alignment reveals the subtle but significant differences in the placement of helices ␣3 in their respective buns. The acyl chain of the inhibitor indicates the location of the active site tunnel whose shape is modulated by the position of helix ␣3. The figure was produced using MOLSCRIPT (27) and rendered with RASTER3D (28). ture of PaFabZ as a template. The corresponding residues in EfFabN are Ile-88 and Phe-110. In both structures, these two residues sit face to face on the outside of the ␣3-helix in a position to modulate the orientation of the helix, therefore, we prepared single and double mutants to replace these residues and assessed the isomerase activity of the constructs. EfFabZ(F86I), EfFabZ(V108F), and EfFabZ(F86I,V108F) were all found to retain dehydratase activity, but all three mutants lacked the ability to isomerize the substrate (not shown). These data suggest that the conversion of EfFabZ to EfFabN requires the sum of the differences between the two ␤3/␤4 strands, rather than a simple single amino acid substitution.
Conclusions-These data are consistent with the hypothesis that the shape of the substrate binding tunnel is the major determinant of the isomerase activity of the FabZ class of ␤-hydroxyacyl-ACP dehydratases. The replacement of the ␤3/␤4 strands in EfFabZ with the corresponding strands from EfFabN were the minimal requirements to convert EfFabZ from a dehydratase to a dehydratase/isomerase. Molecular modeling of the EfFabZ and EfFabN clearly indicate that their structures are both very similar to the PaFabZ crystal structure. The ␤3/␤4 strands that control the catalytic activity of the enzymes interact with both ends of the central helix ␣3, and the conformation of these strands along this interaction interface determines the orientation of helix ␣3, and hence the shape of the active site tunnel (Fig. 7). Single and double mutants within these strands failed to transform EfFabZ into an isomerase, only when the ␤3/␤4 strands were inserted as a unit was isomerase activity reconstituted. The structural analysis of FabA and FabZ (11) led to the hypothesis that differences in the shapes of the active site tunnels caused by the structures of these ␤ strands, rather than active site chemistry, explain why the two enzymes differ in their ability to carry out the isomerization reaction. EcFabA (10) and PaFabZ (11) possess long, narrow, hydrophobic tunnels that span both monomers. The substrate tunnels both begin at the surface with narrow openings that can be completely occluded by a tyrosine, snake under the N terminus of the long, central ␣3-helix of the first monomer, and extend along the side of the corresponding helix ␣3 from the second monomer on its way back to the surface (Fig.  7). The tunnels are roughly 20 Å long, and the catalytic residues are located about halfway down their length. Both proteins have an active site histidine that acts as a general base with the N␦ atom hydrogen-bonded to a backbone carbonyl oxygen. Both proteins also have catalytic water molecules that are held in place by hydrogen bonds to a backbone amide at the N-terminal end of the central helix and to an acidic residue, Asp-84 in EcFabA and Glu-68 in EcFabZ. Although the acidic residues are different, they hold the water molecules in essentially the same spatial position within the tunnels. Site-directed mutagenesis swapping the aspartate in FabA for the glutamate in FabZ, and vice versa, do not change the catalytic properties of the proteins, making it clear that the catalytic machinery is equivalent and capable of carrying out both the dehydratase and isomerase activities (11). Rather, the ability of FabA, but not FabZ, to place the trans-2 substrate in an appropriate conformation to allow isomerization was proposed to arise from the differences in the protein backbones in the distal part of the substrate binding tunnels, encompassing helix ␣3 and strands ␤3/␤4. Differences in the orientation of helix ␣3 are the primary structural differences leading to differently shaped active site tunnels (Fig. 7), which in turn position on the substrate in a conformation that in FabA/N, but not FabZ, is appropriate for the isomerization reaction. Strand ␤4 pushes the ␣3-helix upward, and strand ␤3 moves in the same direction to allow this movement and form one side of the active site tunnel (Fig. 7). Thus, the catalytic properties of the FabA/Z class of dehydratases is specified by the ␤3/␤4 strands, not the catalytic residues, which alter the shapes of the active site tunnels.