|Home | About | Journals | Submit | Contact Us | Français|
The newly discovered FabV enoyl-ACP reductase, which catalyzes the last step of the bacterial fatty acid biosynthesis (FAS-II) pathway, is a promising but unexploited drug target against the re-emerging pathogen Yersinia pestis. The structure of the Y. pestis FabV in complex with its cofactor reveals that the enzyme features the common architecture of the short chain dehydrogenase reductase superfamily, but contains additional structural elements which are mostly folded around the usually flexible substrate binding loop, thereby stabilizing it in a very tight conformation that seals the active site. The structures of FabV in complex with NADH and two newly developed 2-pyridone inhibitors provide first insights for the development of new lead compounds, and suggest a mechanism by which the substrate binding-loop opens to admit the inhibitor, a motion that could also be coupled to the interaction of FabV with the acyl-carrier-protein substrate.
Plague is amongst the most infamous ancient diseases, depopulating medieval Europe in disastrous epidemics. The causative agent is the Gram negative bacterium Yersinia pestis, which was first described by Alexandre Yersin during an outbreak of plague in Hong Kong in 1894 (Alexandre Yersin, 2010). In the 20th century, plague caused extensive deaths in India and Vietnam, and even nowadays it remains a life-threatening disease in Africa, Asia and America (Perry and Fetherston, 1997; Titball and Leary, 1998). Between 1992 and 2001, 28956 cases were reported with 2064 deaths, of which about 84% occurred in Africa alone (2004). If the disease is diagnosed soon after infection, treatment with conventional antimicrobial agents such as gentamycin, streptomycin or doxycycline is usually effective (Butler, 2009). However, in 1995 two highly drug-resistant strains of Y. pestis were reported in Madagascar (Galimand et al., 2006; Guiyoule et al., 2001; Guiyoule et al., 1997). The continuous reports of cases in endemic areas, as well as the evolution of resistant strains, led to the classification of plague as a re-emerging disease by the WHO (Galimand et al., 2006). The emergence of drug-resistant strains in Madagascar (1995) and the possibility that this pathogen could be used as a biological weapon (Richard and Grimes, 2008) underline the danger arising from this neglected organism, and hence the need to identify new drug targets in the struggle against this disease.
The bacterial fatty acid biosynthesis (FAS-II) pathway carries an untapped potential for drug discovery (Lu and Tonge, 2008; Wright and Reynolds, 2007). This pathway differs significantly from the mammalian FAS-I pathway, which utilizes a multi-enzyme complex, in contrast to the discrete enzymes that separately accomplish every step of the bacterial elongation cycle. In the FAS-II pathway the acyl-intermediates are shuttled between the enzymes by the small, negatively charged acyl-carrier-protein (ACP). Some of the currently used antibacterial agents have been shown to target the FAS-II pathway. Isoniazid, one of the front-line drugs against tuberculosis, and triclosan, a broad-spectrum antiseptic, both inhibit the last and rate limiting step of the FAS-II pathway (Heath et al., 1998; Quemard et al., 1995). In this last step the double bond of the enoyl-ACP is reduced in an NAD(P)H dependent step to acyl-ACP by an enoyl-ACP reductase. Currently, four enoyl-ACP reductase isoenzymes are known: FabI, FabL and FabV, which belong to the short-chain dehydrogenase/reductase superfamily (SDR), and FabK, which is a flavoprotein, that uses the flavin-cofactor FMNH2 to reduce the enoyl double bond (Marrakchi et al., 2003; Massengo-Tiasse and Cronan, 2009). The recently identified isoform FabV (Massengo-Tiasse and Cronan, 2008) is the only known isoenzyme in Vibrio cholerae and Y. pestis, and is present along with FabI in Pseudomonas aeruginosa and Burkholderia mallei(Lu and Tonge, 2010; Massengo-Tiasse and Cronan, 2009; Zhu et al., 2010). FabV differs from the previously characterized SDR enoyl-ACP reductases in various aspects. With ~400 residues, it is about 60% larger than the established SDR-family members FabI and FabL, and the overall sequence identity with these isoenzymes is limited to ~15%. In contrast to the other SDR enoyl-ACP reductases, which contain the Y-(Xaa)6-K active site consensus sequence, the FabV active site residues are separated by two additional amino acids in a Y-(Xaa)8-K motif (Massengo-Tiasse and Cronan, 2008). However, a detailed kinetic analysis of B. mallei FabV (bmFabV) revealed that it follows a sequential bi-bi mechanism, with NADH binding first and NAD+ dissociating last, which is similar to the mechanism observed for other FabIs (Lu and Tonge, 2010). Nevertheless, inhibition of the enoyl-ACP reductases reveals differences among these enzymes. While triclosan is a slow-onset nM to pM inhibitor of all studied FabIs apart from that from Mycobacteriumtuberculosis, the flavoprotein FabK is triclosan resistant, and FabL does not form a stable ternary complex with NAD+ and triclosan (Heath et al., 2000). Likewise, triclosan is a rapid reversible inhibitor of bmFabV with a Ki of 0.4 μM (Lu and Tonge, 2010) and for P. aeruginosa, a pathogen that contains both FabI and FabV, it was shown that FabV mediates triclosan resistance (Zhu et al., 2010). Consequently, in order to interfere with the bacterial fatty acid biosynthesis pathway in severe Gram-negative pathogens that contain the FabV enoyl-ACP reductase, such as Y. pestis, P. aeruginosa or V. cholerae, new potent inhibitors of FabV are needed. Here, we present the structure of FabV from Y. pestis in the presence of its cofactor NADH and in complex with two novel 2-pyridone inhibitors, that advance our understanding of this very distinct enoyl-ACP reductase isoform. These structures provide intriguing first insights into the mechanism of substrate recognition and will aid in the development of novel inhibitors against this reemerging pathogen.
The structure of the Y. pestis FabV (ypFabV) was solved by X-ray crystallography using single isomorphous replacement with anomalous scattering (SIRAS) from a crystal soaked in 300 mM NaI. The structure revealed that the His-tag and the following linker region lead to the formation of a dimer in the crystal (Supplementary Figure S1). To exclude the influence of these additional residues on the overall protein structure, part of the linker and His-tag were removed by thrombin cleavage. The cleaved protein crystallized in the space group P3121, with one molecule in the asymmetric unit. The structure was solved by molecular replacement using the model of the His-tagged structure as a search model and refined to R-factors of 16.8% and 21.1% (Rfree). The resulting model contains all 405 residues.
Recently it was shown, that FabV from Vibrio cholerae forms a monomer in solution (Massengo-Tiasse and Cronan, 2008). The crystal form of ypFabV analyzed in this study also indicates a monomer as the biological assembly. This result is quite surprising, since the well characterized FabI and FabL enoyl-ACP-reductases of this familyform tetramers, and the SDR superfamily in general consists of dimers or tetramers.
In spite of its significantly increased size, the core of FabV shows the typical structure of an SDR family member (Figure 1 A/C). Most of the N-terminal part of the protein forms a Rossmann fold, with an eight-stranded parallel β-sheet surrounded by three and four α-helices, respectively, on either side. The Rossmann fold (Buehner et al., 1973) provides the cofactor binding site, whereas the C-terminal part of the protein, which is usually smaller in the SDR family, forms the catalytic site and generates substrate specificity. Additional insights into the properties of the protein arise from the comparison of the FabV structure with the structure of its related isoenzyme FabI. In the structural comparison search against the Protein Data Bank using secondary structure matching (SSM) (Krissinel and Henrick, 2004), the protein structures with the highest structural similarity were FabI from E. coli (ecFabI) in a ternary complex with NAD+ and different inhibitors such as triclosan (ecFabITCL, PDB:1C14/1QG6/1QSG (Qiu et al., 1999; Stewart et al., 1999; Ward et al., 1999)) (Figures 1 B–D). Although ypFabV shares only 15% sequence identity with ecFabI (Figure 2), the three-dimensional structures are significantly more similar and 50% of the secondary structure elements of ypFabV overlap with the ecFabITCL structure according to SSM analysis (using PDBeFold). A detailed comparison between the FabI and FabV structures is presented in the supplementary material (S2). YpFabV is ~140 residues larger than ecFabI, and the majority of these additional residues are located in and around the ypFabV active site. An extended β-hairpin (β9/β10) lies on top of the active site and covers together with an N-terminal β-hairpin (β1/β2) the substrate-binding loop and the following helix (α9), which interacts with ACP in FabI (α8). Four additional helices (α11, α12, α13, α14) surround this central part of the enzyme, thereby stabilizing the substrate-binding pocket of FabV. These additional secondary structure elements may stabilize the protein sufficiently, and thus further stabilization by oligomerization into dimers or tetramers, as observed for the usually smaller SDR superfamily members, is not required.
The major difference between apo-FabI and its inhibitor-bound structure is the substrate-binding loop. In the apo or cofactor-bound FabI structures the substrate-binding loop is flexible and therefore missing in the structure. However, in the inhibitor bound complex the substrate-binding loop is defined and stabilized by the inhibitor. In contrast to FabI the conformation of the substrate-binding loop in FabV is already well defined in the presence of the cofactor, and does not require further stabilization by either a substrate or an inhibitor. Superposition of FabV with the recently published FabL structures from Bacillus subtilis in the presence (PDB:3OID) and absence (PDB:3OIC) of triclosan (Kim et al., 2011) reveal even more pronounced differences. The ternary FabL complex displays a structure comparable to that of the cofactor bound FabV complex (Figure 1 B); in the FabL apo structure, however, not only the substrate-binding loop (η5 in ecFabITCL) but also the helix carrying the catalytic residues is disordered.
The Rossmann fold is the conserved feature of all members of the diverse SDR superfamily, which share low sequence identity and can be classified into five subfamilies (classical, extended, intermediate, divergent and complex SDRs) according to their cofactor binding and active site motifs (Kallberg et al., 2002; Kavanagh et al., 2008; Oppermann et al., 2003). FabI and FabL are members of the “divergent” subfamily with a cofactor binding motif consisting of “GxxxxxSxA”. This motif is partially conserved in ypFabV, although the conserved serine is replaced by a glycine (marked by a box in Figure 2). The NADH cofactor is bound in a tight network of hydrogen bonds involving fixed solvent molecules (Figure 3). Most residues of the glycine-rich conserved motif (G48–G58) are involved in NADH binding and form several hydrogen bonds to the cofactor. The active site residue K244 forms hydrogen bonds to both hydroxyl residues of the nicotinamide ribose, while Y235 interacts via a water molecule with one of the hydroxyl groups. A detailed description of all the interactions formed between the cofactor and the protein can be found in the supplemental material (S3). Since FabV shares only parts of the sequence motif for cofactor binding, this subfamily-defining motif has to be revised if FabV should belong to the subfamily of “divergent” SDRs.
The second, most critical motif for subfamily assignment is the active site motif. For the divergent subfamily, a consensus sequence of “YxxMxxxK” was described in which the tyrosine and the lysine residues play a direct role in catalysis (Kavanagh et al., 2008). In FabV the active site residues are separated by two additional residues and are missing the conserved methionine, which is replaced by isoleucine leading to an active site motive of Y(x)8K. Despite this insertion, the position of the two active site residues Y235 and K244 in the catalytic center of the enzyme is structurally conserved in comparison to Y156 and K163 of ecFabITCL (Figure 4 A). Y235 is thought to bind the thioester carbonyl oxygen, thereby stabilizing the transition state for substrate reduction. In addition, the second active site tyrosine residue in FabI (Y146) is also structurally conserved in ypFabV (Y225). Adjacent to the known active site lysine (K244), which is thought to interact with the cofactor as well as with the acyl substrate, an additional lysine (K245) is present. Interestingly, this second lysine is conserved in FabV from different organisms. It was shown for bmFabV, that the mutation of this lysine to methionine (K245M) did not impair NADH binding, but increased the Km value for trans-2-dodecenoyl-CoA about 10-fold (Lu and Tonge, 2010). However, in the cofactor bound ypFabV structure the side chain of K245 points away from the binding pocket. In order for K245 to interact directly with the substrate, the helix that carries both lysine residues has to rotate. This is highly unlikely since the active site lysine K244 would then move out of its cofactor-binding position. Instead of binding the substrate directly, K245 forms hydrogen bonds to the backbone carbonyl of W236 and the side chain oxygen of N237, and in this way connects the loop that carries the active site tyrosine (Y235) to helix α7, which carries the catalytic lysine (K244) (Figure 4 B). Thus, it is highly probable that K245 holds the catalytic residues in the correct orientation, and thereby has an indirect effect on substrate binding.
A surprising finding of the cofactor-bound structure is the well-defined substrate-binding loop (helix α8, residues S276-M284), which is ordered despite the absence of either a bound inhibitor or substrate. The substrate-binding loop displays the same helical fold as in the ecFabITCL structure, and seems to close the binding pocket, constricting both entrance and volume of the active site pocket relative to FabI. In the ternary FabI complexes, the helical substrate-binding loop has to be stabilized by hydrophobic interactions with a bound inhibitor or substrate. In contrast, the substrate-binding loop in FabV is stabilized by interactions with the cofactor and some of the additional structural elements (Figure 4 C). The N-terminal portion of the substrate-binding loop is embedded in a network of hydrogen bonds, arising mainly from the Rossmann fold and the cofactor. Q277 is involved in a hydrogen bonding network via four water molecules, connecting it to the NADH adenosine phosphate moiety and residues T51, E75, S86, G87, N90 and L373. On the N-terminal β-hairpin (β1/β2), F10 forms hydrogen bonds to S276 and S280, thereby stabilizing the central part of the substrate-binding loop again via a water bridge. These polar interactions are supported by hydrophobic interactions, mostly located at the other end of the substrate-binding loop involving A140/L157 (before/after the β8/9 hairpin), L181 (β10), I230/I234 (η1, next to the active site), W349 (α12) and F370/F374 (α14). The hydrophobic pocket, that surrounds the C-terminal portion of the substrate-binding loop, is complemented by additional structural elements such as a β-hairpin (β9/β10), which closes the binding pocket. Thus, the binding pocket of FabV is quite tightly closed and rigid, in contrast to the flexible, open binding site observed in binary FabI complexes. Although our binary complex of ypFabV contains NADH, in contrast to the binary complex structures of bacterial FabIs that all contain NAD+, there is no distinct interaction between NADH and the substrate-binding loop, which could not also be mediated by NAD+. The closed conformation of the substrate-binding loop in the binary complex can thus not be explained by the fact, that it is a substrate-complex instead of a product-complex as in the binary FabI structures. Despite these differences between FabI and FabV, the architecture of the active site is remarkably conserved.
The low affinity of the FabI inhibitor triclosan for FabV is another interesting aspect of this newly discovered enoyl-ACP reductase. Triclosan (Figure 5 A) is a slow onset inhibitor of ecFabI (Ki=7 pM) (Lu and Tonge, 2008; Sivaraman et al., 2004) and it has been hypothesized that during the slow step of the binding process the flexible substrate-binding loop forms hydrophobic interactions with the inhibitor, resulting in a defined helical conformation (Qiu et al., 1999). Since the substrate-binding loop is already ordered in ypFabV and its conformation is very similar compared to that of the triclosan-bound ecFabI structure, one might expect that triclosan would also bind with high affinity to ypFabVof FabV to triclosan. However, in previous studies (Lu and Tonge, 2010), it was shown that triclosan is a rapid-reversible inhibitor of bmFabV with a Ki value of only 0.4 μM. Here we show that triclosan binds with even lower affinity to ypFabV with a Ki' value of 71±4 μM, which is ~100× less than the corresponding value for bmFabV (Figure 5 B).
SAR studies based on the diphenyl ether scaffold of triclosan have resulted in a number of potent inhibitors of the FabI enzyme from pathogens such as Staphylococcus aureus (saFabI), M. tuberculosis, and Plasmodium falciparum (am Ende et al., 2008; Chhibber et al., 2006; Lu and Tonge, 2008; Sullivan et al., 2006). One such compound is PT70 (Figure 5 A), which is a slow-onset inhibitor of the M. tuberculosis FabI (InhA) (Luckner et al., 2010). However, despite their ability to potently inhibit FabI, the diphenyl ethers are susceptible to Phase II conjugation reactions, and we are currently exploring the utility of replacing the diphenyl ether A-ring with a 2-pyridone, which has been identified as a promising scaffold for the inhibition of saFabI (Tipparaju et al., 2008; Tipparaju et al., 2010; Yum et al., 2007). In the present study, the A ring of our novel 2-pyridone inhibitors is substituted with a C6 alkyl chain to mimic the ypFabV substrate, while the B-ring carries either an ortho chlorine substituent (PT172) or an ortho methyl group together with a meta amine (PT173) (Figure 5 A). Interestingly, PT172 (Figure 5 C) and PT173 (Figure 5 D) are competitive inhibitors of ypFabV with Ki values of 2.1±0.4 μM and 1.5±0.3 μM, respectively.
To characterize the interactions of these promising 2-pyridone inhibitors with ypFabV and to gain more insight into their molecular effect on ypFabV, both inhibitors were co-crystallized with the protein in the presence of the cofactor. The protein portion of the two ternary complexes is almost identical to each other with an rmsd of 0.09 Å (all atoms). Both structures can be superimposed with the cofactor-bound ypFabV structure (rmsd= 0.2 Å, all atoms) with one notable exception (Figure 6). The substrate-binding loop, together with the β-turn of the N-terminal β-hairpin (β1/β2), is shifted by 3 Å out of the active site pocket. The binding pocket is thereby broadened, which allows the inhibitor entrance to the pocket and provides sufficient space for binding. This shift of the substrate-binding loop indicates a completely reversed inhibitor binding behaviour of ypFabV compared to FabI. FabV has to loosen its tight conformation to provide space for the inhibitor. Furthermore, in contrast to FabI, the substrate-binding loop does not contribute to the hydrophobic interactions with the inhibitor. The distance between the substrate binding loop and the inhibitor is roughly 6 Å, which is too far for direct interactions. Instead, water molecules between polar residues such as S276 and S279 and the inhibitor mediate interactions between the cofactor and the substrate-binding loop (Figure 7 A).
Apart from this intriguing difference, the enzyme-inhibitor interactions in the ypFabV structures closely resemble those observed in FabI-triclosan structures (Figure 7 A) (Qiu et al., 1999; Sivaraman et al., 2004; Sivaraman et al., 2003; Stewart et al., 1999). The carbonyl oxygen in the pyridone ring forms a hydrogen bond to the hydroxyl group of Y235 as well as to the nicotinamide ribose. In addition, the parallel-displacedπ-π–stacking interaction between the A-ring and the nicotinamide ring of the cofactor is also present. The position of the A-ring is identical in both inhibitor structures, whereas the carbon chain and the B-ring show a slightly different conformation (Figure 7 B). The acyl chain is flexible, which is reflected in a more diffuse electron density. The B-ring, on the other hand, is well defined in the density but is tilted by 24°. The B-ring and the acyl-chain of the PT172 inhibitor form hydrophobic interactions to residues A140, M196, Y225, T231, A273 and M285 (Figure 7 C). In PT173, the amine group on the B-ring forms hydrogen bonds via a water bridge to the side chain of S155 and the carboxyl oxygen of S141 (Figure 7 D). This causes the B-ring to move towards the water molecule, which leads to the loss of the hydrophobic interaction with M196.
To determine the reason for the decreased affinity of triclosan compared to the new 2-pyridone inhibitors, docking trials where performed using the ternary complex form of ypFabV. The docking results suggest that triclosan is unlikely to adopt the binding mode in ypFabV that is observed in other enoyl-ACP reductases. Instead appropriate positioning of the diphenyl ether skeleton in the active site was only obtained when docking was carried out for a triclosan derivative in which the chloro-substituent of the A ring was replaced by an n-hexyl chain, as in PT172 and PT173. This finding underlines the importance of the acyl-chain for ypFabV inhibition. A detailed description of the triclosan docking approach can be found in the supplemental material (S4 and S5). Although the structural data demonstrate how the 2-pyridone inhibitors interact productively with ypFabV, further studies are required to determine why the 2-pyridones are preferred over the diphenyl ether scaffold.
Even though the binding pocket and the active site of FabV share similarities to FabI, the additional structural elements occlude the active site pocket from the surrounding solvent, and thus pose the question of how substrates reach their destination. An additional criterion for a possible substrate channel is the availability of a potential ACP interaction site. In general, negatively charged residues located on helix αII of ACP interact with hydrophobic and electropositive surfaces on the target proteins which include lysine and arginine residues. ACP is thereby capable to interact with a variety of different enzymes, yet it binds only with low affinities (low μM range) to the target proteins (Chan and Vogel, 2010; Parris et al., 2000; Rafi et al., 2006; Zhang et al., 2003a; Zhang et al., 2003b). In FabV, the basic residues, which presumably interact with ACP, are not conserved and the helix, which seems to be responsible for ACP binding in FabI, is covered by two β-hairpin structures in ypFabV.
For FabI, two main entry sites to the active site pocket have been termed the minor and the major portals, with the latter being much wider (Rozwarski et al., 1999). Nevertheless, in the ecFabI-ACP complex structure (Rafi et al., 2006) the acyl-pantetheine adduct was modeled by molecular dynamic simulations through the minor portal. For the cofactor-bound structure of ypFabV, molecular channels leading from both entry sites could be verified in MOLE and Caver (Petrek et al., 2007; Petrek et al., 2006) (Figure 8 A). However, the channel through the minor portal in FabV is very narrow due to the tightly closed conformation of the substrate-binding loop and the long β-hairpin (β9/β10) covering the binding pocket. The other possibility is an entry of the substrate through the major portal. In a substrate analogue complex structure of InhA the N-acetyl cysteamine portion of the C16 fatty acid substrate, which would be physiologically coupled to the ACP, faces the major portal (Rozwarski et al., 1999). Next to the major portal in FabV, the N-terminal β-hairpin (β1/β2) spreads out with three very prominent basic residues K4, R6 and R8 on β1 and a stretch of hydrophobic residues (G9, F10, I11, V13, A15) on β-turn (β1/β2) and β2 (Figure 8 B). ACP could bind to these basic residues and present the substrate to the hydrophobic patch that leads into the binding pocket. This scenario is supported by the inhibitor structures, if it is assumed that the alkyl substituents mirror the interaction of substrates with the enzyme. Thus, a comparable outward movement of the substrate-binding loop, as observed in the inhibitor bound structures, will probably also occur upon substrate binding. Apart from the substrate-binding loop, only the N-terminal β-hairpin (β1/β2) moves upon inhibitor binding, underlining its strong connection to the substrate-binding loop. Furthermore, this movement only broadens the major portal, whereas the minor portal remains narrow. One could thus speculate that the β1/β2-hairpin, containing the three basic residues, is pulled away from the active site upon ACP binding (Figure 8 C). The substrate-binding loop, to which the the β1/β2-hairpin is connected, follows this outward movement and opens the active site for the enoyl substrate, which can then easily enter the active site via the hydrophobic patch, located adjacent to the putative ACP binding site.
The structure of ypFabV confirms the unique character of this SDR enoyl-ACP reductase. YpFabV is ~140 residues larger than the typical FabI isoforms, and these additional residues are mainly located around the substrate-binding loop. This usually flexible loop is central for the function of FabI, since it enables the enzyme to adapt to substrates of various sizes and is only closed when ligands are bound. In contrast, FabV seems to follow an inverse mechanism. Without a substrate bound it adopts a closed conformation which has to open to allow access of the enoyl substrate or an inhibitor, as observed in the PT172 and PT173 complex structures. These mechanistic differences might lead to different requirements for the inhibition of each enzyme. In addition, the novel pyridone inhibitors, characterized in this study, reveal that FabV can be targeted by a structure-based drug design approach, and holds the promise that inhibitors can be identified which target both FabI and FabV simultaneously.
The fabV gene was amplified from the Y. pestis CO92 genome (NCBI Reference Sequence YP_002348955.1) using the following primers (Integrated DNA Technologies): 5'-CCGCTCGAGATGATTATAAAACCAC G TGATA- 3' (forward) and 5'-CCGGAATTCTTAACCCTGAATCAAGTTAGG- 3' (reverse). The PCR product was digested with XhoI and EcoRI and inserted into the pET15b vector (Novagen) adding a His-tag at the N-terminus. The target sequence was confirmed by sequencing.
Protein expression and purification were performed as described previously using E. coli BL21(DE3) pLysS cells (Xu et al., 2008).
To remove the N-terminal His tag, 1 μL of biotinylated thrombin was added to 1 mL of ypFabV (20 μM) in PIPES buffer. The reaction mixture was incubated at room temperature for 16 h. Streptavidin agarose (20 μL) was added, and the solution was incubated for an additional hour, after which the agarose was removed by centrifugation at 13000 rpm for 10 min. The supernatant was further purified by chromatography on a Sephadex G-25 column using PIPES buffer as the eluent, and fractions containing ypFabV lacking the His tag were pooled and analyzed via 12% SDS-PAGE.
Trans-2-dodecenoyl-CoA (DD-CoA) was synthesized from trans-2-decenoic acid using the mixed anhydride method (Parikh et al., 2000). ESI-MS ([M-H]−) gave a mass of 946.2 Da for DD-CoA in perfect agreement with the mass of 946.2 Da calculated for [C33H53N7O17 P3S].
All kinetic experiments were performed on a Cary 300 Bio (Varian) spectrometer at 25 °C in 30 mM PIPES buffer pH 6.8 containing 150 mM NaCl and 1 mM EDTA. Initial velocities were measured by monitoring the oxidation of NADH to NAD+ at 340 nm (ε = 6,300 M−1 cm−1).
The inhibition mechanism was characterized by measuring initial velocities at a fixed concentration of NADH (250 μM) and at various concentrations of DD-CoA and inhibitor. Data were analyzed in Lineweaver-Burk plots. The dissociation constant was calculated by fitting the data to eqns 1 and 2 which describe uncompetitive and competitive inhibition, respectively.
In eqns 1 and 2, [S] is the concentration of DDCoA, [I] is the concentration of inhibitor, Km is the Michaelis-Menten constant for DDCoA, Vmax is the maximum velocity, Ki and Ki' are the inhibition constants.
For purification of ypFabV the cells were thawed, resuspended and lysed in 50 mM Na3PO4 pH 7.4, 300 mM NaCl, 5 mM imidazole. The soluble fraction of the cell lysate was loaded on a Ni-TED (Protino) affinity column, washed with 10 mM imidazole in the same buffer and eluted with 250 mM imidazole. The protein was further purified by size exclusion chromatography (Superdex 200 26/60, GE Healthcare/ ÄKTA) using 20 mM PIPES pH 7.6, 300 mM KCl and 1 mM EDTA as the eluent. The protein sample was concentrated to 68 mg/mL, flash-frozen in liquid nitrogen and stored at −80°C.
Thrombin cleavage was integrated into the purification protocol after affinity chromatography. The elution of the Ni-TED column was concentrated up to 65 mg/mL and cleaved with 1 U thrombin (GE healthcare)/mg protein at room temperature over night. The protease was removed via a benzamidine column, and cleaved protein was purified by size exclusion chromatography.
The protein sample (68 mg/mL) was preincubated with the cofactor NADH (fivefold molar excess) for 30 minutes. 1 μL drops were equilibrated in hanging drop vapor diffusion experiments at 20° C against 1 mL of the reservoir solution, containing 1.25–1.5 M sodium malonate and 100 mM buffer (citrate pH 5.6/imidazole pH 8.25). Crystals grew in space group P6522 with cell parameters of a=b=104 Å, c=220 Å. For the SIRAS approach, crystals were soaked with 300 mM NaI in the reservoir solution including 30% glycerol as cryo-protectant for 10–30 seconds. A dataset of the iodide-soaked crystals was collected at beamline ID29 (ESRF, Grenoble) at a wavelength of 1.25 Å, and a native dataset was obtained at beamline MX 14.1 (Bessy, Berlin). The cleaved FabV was treated likewise andcrystallized in 150 mM (NH4)2SO4, 25% polyethyleneglycol 4000 and 100 mM MES pH 5.5. The diffraction dataset was collected at beamline MX 14.1 (Bessy, Berlin).
For co-crystallization experiments thrombin-cleaved FabV was preincubated with the cofactor NADH (fivefold excess) and the inhibitor (5mg/mL) for 2 hours. The inhibitor complexes crystallized in 150 mM (NH4)2SO4, 35–37% polyethyleneglycol 4000 and 100 mM MES pH 5.5. Data were collected at a Rigaku MicroMaxTM-007HF x-ray generator with a Raxis HTC detector from cryopreserved crystals.
The datasets were indexed and integrated with imosflm (Leslie, 1992) and scaled in Scala (1994). Phasing of the tagged FabV crystal was achieved by using the program autoSHARP (Vonrhein et al., 2007) in a SIRAS approach. An initial model covering 88% of the ypFabV sequence was automatically built by ARP/wARP (Langer et al., 2008), and the model was subsequently revised and extended manually in Coot (Emsley and Cowtan, 2004). For refinement in Refmac and Phenix (Adams et al., 2010) TLS parameters were created using the TLSMD server (Painter and Merritt, 2006) and a library file supplying restrains for the NADH cofactor was created with the Prodrg server (Schuttelkopf and van Aalten, 2004). The tagged ypFabV structure was refined and validated using Phenix (Adams et al., 2010) and Procheck (in PDBsum (Laskowski, 2009)). The structure of the cleaved ypFabV was solved by molecular replacement using Phaser (McCoy et al., 2007) with the structure of the tagged FabV as a search model. Refinement and validation was performed as in the tagged protein. Interactions were analyzed with the help of Ligplot (Wallace et al., 1995). Images were created using the programs Pymol (Schrodinger, 2010), ESPript (Gouet et al., 2003), Topdraw (Bond, 2003), Caver and MOLE (Petrek et al., 2007; Petrek et al., 2006; Schrodinger, 2010).
This work was supported by grant AI065357 from the National Institutes of Health to PJT and through the Deutsche Forschungsgemeinschaft (SFB 630 to C.K. and C.S. and Forschungszentrum FZ82 to C.K.). We thank the staff of BL 14.1 at BESSYII, Berlin and ID29 of the ESRF, Grenoble for technical support. In addition, CN was supported by the Chemistry-Biology Interface Training Program grant T32GM092714 from the National Institutes of Health.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.