Structure of a Calcium-dependent 11R-Lipoxygenase Suggests a Mechanism for Ca2+ Regulation*

Background: Lipoxygenases vary in their catalytic specificity and regulation. Results: 11R-LOX, strictly Ca2+-dependent, displays novel structural features in the membrane-binding domain. Conclusion: A model for how access to an enclosed active site is linked to Ca2+-dependent membrane binding is proposed. Significance: The 11R-LOX model provides structural insights into the allosteric regulation of lipoxygenases. Lipoxygenases (LOXs) are a key part of several signaling pathways that lead to inflammation and cancer. Yet, the mechanisms of substrate binding and allosteric regulation by the various LOX isoforms remain speculative. Here we report the 2.47-Å resolution crystal structure of the arachidonate 11R-LOX from Gersemia fruticosa, which sheds new light on the mechanism of LOX catalysis. Our crystallographic and mutational studies suggest that the aliphatic tail of the fatty acid is bound in a hydrophobic pocket with two potential entrances. We speculate that LOXs share a common T-shaped substrate channel architecture that gives rise to the varying positional specificities. A general allosteric mechanism is proposed for transmitting the activity-inducing effect of calcium binding from the membrane-targeting PLAT (polycystin-1/lipoxygenase/α-toxin) domain to the active site via a conserved π-cation bridge.

Lipoxygenases (LOXs) 2 are non-heme iron dioxygenases that catalyze the stereo-and regiospecific hydroperoxidation of polyunsaturated fatty acids (1). LOX catalysis products of arachidonic acid (AA), which is the main substrate in animals, are hydroperoxyeicosatetraenoic acids (HpETEs), these lipid mediators and their metabolites have been implicated in cancer (2), atherosclerosis (3), and allergic inflammation (4). Consequently, LOXs are targets for drug design. A complicating factor in the development of LOX inhibitors is that there are several LOX isoforms in an organism, all with equivalent catalytic machinery and chemical mechanism (5). Thus, differences in regiospecificity or regulation must be exploited to design iso-form-specific inhibitors. Given the limited amount of structural information of arachidonate-metabolizing LOXs, each new structure provides crucial details related to LOX catalysis mechanism.
LOX catalysis begins with a stereoselective hydrogen abstraction by the catalytic non-heme iron from the methylene carbon (CH 2 ) of the selected 1,4-cis,cis-pentadiene unit on the fatty acid substrate, and is followed by regioselective dioxygen addition on the opposite face of the substrate either at Ϫ2 or ϩ2 carbon ( Fig. 1) (6). For such specific reactions to take place, a very distinct substrate channel that goes past the non-heme iron must position the fatty acid. This binding site must also vary among LOX isoforms to facilitate the different catalytic properties. The substrate-binding cavity has been described as "boot-shaped"; it is directly accessible from the surface of the protein and ends with a hydrophobic pocket (7). The pocket residues of several 12/15-LOXs have been mutated to demonstrate that bulkier side chains favor 15-lipoxygenation, whereas less space-filling residues, which would allow the fatty acid tail to penetrate deeper into the cavity, confer 12-LOX activity. These results are consistent with aliphatic tail-first entry (8 -12). Also, a cationic arginine near the entrance of the cavity has been shown to stabilize the carboxylate head of the fatty acid (13). Computational docking studies based on x-ray crystallography data further support the boot-shaped substrate channel (14,15). For some LOXs, carboxylate head-first binding has been suggested to explain differing specificity or double dioxygenation of AA (16,17). In the light of the coral 8R-LOX crystal structure, however, a novel binding model was proposed with an alternative U-shaped channel that neglects the hydrophobic pocket (18). According to this hypothesis, the substrate is bound in a culvert that runs under a conserved arched helix; distinct lipoxygenases allow access to the catalytic iron from one of two possible directions. Although several lipoxygenase crystal structure models have been published, including rabbit 12/15-LOX (19,20), coral 8R-LOX (18,21), and recently a modified human 5-LOX (22) representing Animalia, the lack of experimental evidence on substrate binding, such as a crystallized enzyme-substrate complex, has precluded the emergence of a uniform theory.
The activity of various LOXs depends more or less on the presence of Ca 2ϩ that promotes interactions with the lipid membrane, from where the enzyme obtains its fatty acid substrate (21,23). The human 5-LOX is effectively translocated to the nuclear envelope upon Ca 2ϩ release, the C2-like PLAT domain being the selective membrane-targeting module (24). The calcium-binding sites of the PLAT domain appear to be conserved among human 5-LOX, coral 8R-LOX, gangrene ␣-toxin, as well as coral 11R-LOX, which are all induced by Ca 2ϩ , but not in rabbit 12/15-LOX, which is only mildly affected (21,25). The molecular mechanism of Ca 2ϩ -and membrane-induced allosteric regulation is not clear. Although many mammalian lipoxygenases retain their reaction specificity after PLAT domain truncation, this is accompanied by reduced turnover rates (26). Moreover, tight association of PLAT and catalytic domains has been shown to be important for protein stability and catalytic activity (27). A possible structural element that may be under allosteric control is the ␣2 region that forms a "lid" over the putative substrate channel entrance, which can adopt different conformations, thereby either opening or closing the orifice (20,28,29). Another plausible allosteric mechanism could involve oligomerization, which has been noticed in case of human platelet 12-LOX and rabbit 12/15-LOX (30,31), but no definite assembly has been described to date.
The arachidonate 11R-LOX from the white sea coral Gersemia fruticosa is the first described lipoxygenase with 11Rspecificity (25). Based on primary structure comparison it is most closely related to the 8R-LOX from the allene oxide synthase-lipoxygenase fusion protein of the Caribbean sea whip coral Plexaura homomalla (42% identity) (32) and the analogous enzyme from G. fruticosa (43%) (33). The closest mammalian counterparts are 5-LOXs (about 33%). Experiments with alternative substrates suggest that the fatty acid enters the active site tail-first, as the catalysis specificity depends on the distance of the bisallylic methylene from the methyl end of the aliphatic tail (25). The conserved Gly/Ala sequence determinant, which acts like a switch that directs oxygen to either Ϫ2 or ϩ2 carbon (R-or S-stereospecificity, respectively) (34), has been identified as Gly 416 (25), which agrees with previous findings linking glycine to R-stereoconfiguration. An important feature of the 11R-LOX is its complete dependence on both Ca 2ϩ and lipid membranes, the presence of both components is necessary for catalytic activity (25). This, together with the remarkable stability and relative similarity of the enzyme to human 5-LOX, makes it an exceptional subject to study the mechanism of lipoxygenase catalysis specificity and regulation.
Hereby we report the 2.47-Å crystal structure model of the coral 11R-LOX that suggests a potential allosteric mechanism involving the PLAT domain and the ␣2 lid region. A highly conserved -cation bridge was found that could mediate the regulatory effect of the PLAT domain to the active site. Additionally, a mutation analysis of the hydrophobic pocket in the boot-shaped cavity was conducted to address questions regarding substrate binding. A general hypothesis of possible substrate orientations in the active site is described.

EXPERIMENTAL PROCEDURES
Expression and Purification-Recombinant G. fruticosa 11R-LOX with an N-terminal His 4 tag in pET-11a vector was transformed into Escherichia coli BL21(DE3) cells (Novagen). Colonies were grown overnight in 25 ml of LB containing 100 g/ml of ampicillin at 37°C. A 500-ml volume of autoinducing medium ZYM-5052 (35) (with 100 g/ml ampicillin) was inoculated with 5 ml of overnight culture. The culture was incubated at 37°C for 3-4 h, followed by growth to saturation at 20°C. Cells were harvested by centrifugation and frozen at Ϫ80°C when the absorbance at 600 nm had remained stable for 4 h (usually 27-30 h after the inoculation).
Cell pellets were resuspended in Bugbuster (Novagen) with added DNase I, pepstatin, and PMSF. The suspension was stirred and incubated on ice for 30 min, lysed in a French pressure cell, and centrifuged at 39,000 ϫ g for 40 min at 4°C. The supernatant was applied onto a HisTrap Ni-Sepharose column (GE Healthcare) and washed with binding buffer (50 mM Tris-HCl, 500 mM NaCl, 20 mM imidazole, pH 8.0) on an ÄKTA FPLC system (GE Healthcare). The protein was eluted with an imidazole gradient from 20 to 200 mM. Protein fractions were dialyzed overnight against 20 mM Tris-HCl, pH 8.0, or desalted in a Sephadex G-25 Fine column (Amersham Biosciences). The sample was then applied onto a Mono Q anion exchanger (GE Healthcare), washed with 20 mM Tris-HCl, pH 8.0, and eluted with a NaCl gradient from 0 to 500 mM. Concentrated protein fractions were run on a Superdex 200 size exclusion column (GE Healthcare) with 20 mM Tris-HCl, pH 8.0, and 150 mM NaCl. For long term storage, the protein was concentrated to ϳ15 mg/ml, flash frozen in liquid nitrogen, and stored at Ϫ80°C.
Protein Crystallization-Showers of plate-like crystals formed in 1-2-day-old hanging-drop vapor diffusion experiments that contained a 1:1 mixture of 5 mg/ml of protein solution and reservoir solution (0.1 M bis-Tris, pH 7.2, 11-12% (w/v) PEG 3350, 15% (w/v) sucrose) at 22°C. To obtain fewer and bigger crystals a microseed stock was prepared using the Seed Bead (Hampton Research) in stabilizing solution with 0.1 M bis-Tris, pH 7.2, 9% (w/v) PEG 3350, and 15% (w/v) sucrose. Seeding experiments were conducted with drops containing 2:1:1 mixture of protein and reservoir solution as described The catalysis begins with stereoselective hydrogen abstraction from C7, C10, or C13 (labeled proS/R), followed by antarafacial dioxygen addition either at Ϫ2 or ϩ2 carbon (grouped by colors). Based on data from Ref. 6. above and serial dilutions of seed stock as described in the Seed Bead user guide. Larger single crystals grew in 2-3 days. For cryoprotection, crystals were transferred into 0.1 M bis-Tris, pH 7.2, 12% (w/v) PEG 3350, 20 -25% (w/v) sucrose in two consecutive steps and then frozen in liquid nitrogen or a 100 K cryostream.
Data Collection and Structure Determination-Preliminary screens for crystal diffraction were conducted at the Gulf Coast Protein Crystallography Consortium beamline at the Center for Advanced Microstructures and Devices (CAMD, Louisiana State University). A full dataset was collected at the NE-CAT beamline 24-ID-E at the Advanced Photon Source (Argonne, IL) using 0.98-Å radiation at 100 K. Data were processed to a resolution of 2.47 Å ( Table 1) using xia2 (36). The structure was determined by molecular replacement with MrBUMP (37, 38) using 3.2-Å P. homomalla 8R-LOX model (PDB code 2fnq). The initial refinement cycles were performed with REFMAC5 (39). Both MrBUMP and REFMAC5 are part of the CCP4 suite (40). Manual model building was done with COOT (41) and further refinement in PHENIX (42) using the program phenix.refine with non-crystallographic symmetry and Ramachandran restraints, individual isotropic atomic displacement factors, and automatic water picking. For the final refinement, hydrogens were added to the model, Ramachandran restraints were released and both stereochemistry and atomic displacement weights were optimized. Illustrations were prepared with UCSF Chimera (43), surfaces were obtained with MS-MS (44). The dimerization interface was analyzed using PISA (Protein Interfaces, Surfaces and Assemblies) (European Bioinformatics Institute) (45). Sequences were aligned with ClustalW2 (46) and rendered with ESPript (47).
Site-directed Mutagenesis-The V430A, L431A, V609A, and V609W mutations were introduced using whole plasmid PCR primed with complementary primers that additionally contained silent mutations for restriction analysis. The M606A mutant was obtained by separately cloning the upstream and downstream fragments of the recombinant cDNA using mutation-containing primers with respective cDNA upstream or downstream primers. Purified fragments were merged using PCR and the cDNA was ligated back into pET-11a vector (Stratagene) into the BamHI site. The desired mutations were confirmed by sequencing. For whole plasmid PCR protocol, the following DNA primers were used with their complementary primers for mutagenesis: 5Ј-GGT GCG GCT GAC AAA GCG CTG AGC ATT GGT GGA GG-3Ј for V430A, 5Ј-GGT GCG GCT GAC AAG GTG GCT AGC ATT GGT GGA GG-3Ј for L431A, 5Ј-GTT ACA ATG GTT TCA GCT GTG AAT GCG C-3Ј for V609A, 5Ј-ACA ATG GTT TCT TGG GTT AAC GCG CTA ACC ACG A-3Ј for V609W, and the following primers for cloning fragments with M606A mutation: 5Ј-AA GGA TCC ATG CAT CAC CAT CAC ATG AAG TAC AAG-3Ј (11R-LOX cDNA upstream) and 5Ј-G CGC ATT CAC AAC AGA AAC AGC TGT AAC AGC TTG-3Ј (M606A downstream) for the upstream fragment, 5Ј-CAA GCT GTT ACA GCT GTT TCT GTT GTG AAT GCG C-3Ј (M606A upstream) and 5Ј-GAT GGA TCC TTA GAT GGC AAT ACT GTT CGG-3Ј (11R-LOX cDNA downstream) for the downstream fragment.
For product analysis, aliquots of 2.5 ml of culture were resuspended in 500 l of 50 mM Tris-HCl, pH 8.0, with 1 mM PMSF and sonicated 3 ϫ 5 s using a Torbéo 36810-Series cell disruptor (Cole Parmer) at a setting of 5. The suspension was centrifuged at 13,000 ϫ g for 20 min at 4°C and the supernatant was harvested. CaCl 2 was added to the enzyme solution in final concentration of 10 mM. Incubations of 1-20 ml were carried out in 50 mM Tris-HCl, 250 mM NaCl, pH 9.0 buffer with 50 M arachidonic acid at room temperature for 5 min with constant stirring. [1-14 C]-Labeled arachidonic acid (GE Healthcare) was used in 1-ml incubations. HpETEs were reduced to corresponding hydroxy acids (HETEs) with 10 mM SnCl 2 , the mixture was acidified with KH 2 PO 4 /HCl (1:1) to pH 6 and the products were extracted using ethyl acetate. Incubation products (10 -20 ml volumes) were purified prior to HPLC analysis using thin layer chromatography: the extract was applied on a silica gel plate, eluted with a hexane/ethyl acetate/acetic acid (3:4: 0.05) mixture, and the product band was determined using UV light (254 nm) and extracted with methanol.
For kinetic studies, wild-type enzyme and selected mutants were purified on an ÄKTA FPLC system as described above. Conjugated diene formation was monitored on a UV-1601 spectrophotometer (Shimadzu) at 236 nm. Reactions of 1 ml (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM CaCl 2 , ϳ60 M liposome) were performed in a thermostatted (20°C) cuvette with continuous stirring. The concentration of arachidonic acid (Cayman) was varied from 2 to 200 M. The reaction was initiated by adding 6 nM wild-type enzyme, or up to 32 nM for the less active mutants. The reaction velocity was determined from the slope of the linear portion of the curve. K m and k cat values were obtained by nonlinear regression analysis with the Michaelis-Menten equation. As 11R-LOX exhibits very strong substrate inhibition and, thus far, no suitable kinetic model has been derived, only the ascending part of the curve (2-30 M AA) was used for fitting.
HPLC-MS Analysis-Catalysis products were analyzed by reverse phase HPLC using an Agilent Eclipse 3.5 m 150 ϫ 2.1-mm ODS column thermostatted at 35°C. The sample was eluted isocratically with methanol/water/acetic acid (75:25: 0.01, v/v/v) at 0.25 ml/min on an Agilent 2200 Series HPLC system. Products were detected using a diode array detector monitoring wavelengths 210 -280 nm, followed by an Agilent LC/MSD Trap XCT mass spectrometer. The MS/MS spectra of arachidonic acid derivates were obtained in negative mode using an APCI interface.
[1-14 C]-Labeled products were additionally analyzed using a Radiomatic 500TR Flow Scintillation Analyzer (Packard Bioscience) preceded by an Agilent Eclipse 5 m 150 ϫ 4.6-mm ODS column thermostatted at 35°C and a diode array detector. The same eluent was used at a flow rate of 1 ml/min.
Liposome Preparation-Small unilamellar vesicles were prepared from L-␣-phosphatidylcholine, L-␣-phosphatidylethanolamine, and L-␣-phosphatidylserine (Avanti Polar Lipids). A mixture of phosphatidylcholine/phosphatidylethanolamine/ phosphatidylserine (40:30:30 mol %) in chloroform was dried in a round-bottom flask using a nitrogen stream to form a thin film and was incubated in vacuum for 1 h at room temperature. Buffer (50 mM Tris-HCl, 100 mM NaCl, pH 8.0) was added to the dried lipids and the film was soaked for 1 h. The suspension was shaken vigorously, sonicated 15 ϫ 5 s on a Torbéo 36810-Series cell disruptor (Cole Parmer) with 1-min intervals at the setting of 5 in a water bath at room temperature. The sonicate was centrifuged at 13,000 ϫ g for 20 min at 4°C. The supernatant was saved and stored at 4°C.

RESULTS
Overall Structure-The crystal structure of 11R-LOX in a calcium-free environment was refined at 2.47 Å with R work and R free values of 0.20 and 0.23, respectively ( Table 1). The protein crystallized in space group C2 with two molecules in the asymmetric unit. Phasing was done with molecular replacement using P. homomalla 8R-LOX (PBD code 2fnq) as a template. A total of 96.1% of the residues are in the favored region of the Ramachandran plot and there are no outliers according to Mol-Probity validation (48).
Like the majority of lipoxygenases, 11R-LOX consists of two distinct domains: the N-terminal C2-like PLAT domain (residues 1-115) and a larger mainly ␣-helical catalytic domain (residues 129 -679), which are connected by a small linker region (residues 116 -128). The catalytic domain contains the nonheme iron, which is coordinated by three highly conserved histidines, His 373 , His 378 , and His 556 , and the carboxylate group of the C-terminal Ile 679 . In general, the structure is very similar to that of the 8R-LOX domain of P. homomalla allene oxide synthase-LOX fusion protein, a root mean square deviation of 644 C ␣ pairs being 1.37 Å. The greatest differences are found in the putative entrance to the active site, which more closely resembles the recently published human Stable-5-LOX (22). Similarly to 5-LOX, the ␣2 helix of 11R-LOX that covers the putative entrance is only about 2 turns long and is flanked by loops and small 3 10 -helices. It should also be mentioned that the overall similarity between the 11R-and 5-LOX is very high (root mean square deviation of 641 C ␣ pairs is 1.49 Å).
Another heterogeneous region includes the putative Ca 2ϩbinding sites in the PLAT domain. The PLAT domain is a ␤-sandwich consisting of two antiparallel 4-strand ␤-sheets. At the sheets' ends proximal to the catalytic domain (opposite to the N terminus) there are four loops that connect the two sheets, three of the loops being rather extensive, this is the region that contains the Ca 2ϩ -binding sites. The electron density map is less clear in the region of the PLAT domain, which is characterized by higher mean B-factors, 37.3 Å 2 for the catalytic domain versus 70.5 Å 2 in the PLAT domain. The least well defined densities are found in the putative calcium-binding loops, especially residues His 43 -Glu 47 and Gly 73 -Lys 77 ; whereas main chain density is not ambiguous, several side chains were modeled primarily according to optimal geometry (side chains present at Ն0.5 ). For refinement, all occupancies were set to 1. The apparent mobility of these residues is described by their elevated B-factors.
Dimerization-Size exclusion chromatography indicated that 11R-LOX appears as a dimer in a calcium-free buffer. The purified enzyme eluted with aldolase (158 kDa), which is double the molecular mass of the recombinant protein (79 kDa) (supplemental Fig. S1). Based on ultrafiltration assays, it was previously concluded that 11R-LOX is in a monomeric state in solution: in calcium-free conditions the enzyme passed a filter with a 100-kDa cutoff without major losses (25). The findings presented here clearly dispute those claims. In the presence of 10 mM CaCl 2 , however, size exclusion chromatography analysis indicated the formation of large aggregates, as the protein eluted with the void volume (data not shown). Similar results were obtained using ultrafiltration (25). Potential dimerization interfaces were searched among the crystal contacts using PISA (45), but according to the criteria established by the algorithm, no significant assemblies were found.
Substrate Channel and Its Entrances-There is an array of consecutive, mostly hydrophobic cavities concealed in the catalytic domain alongside the coordinated non-heme iron (Fig.  2). These cavities are covered by a conserved arched helix (␣10 -␣11 in G. fruticosa), and the potential entrances on either side are blocked by short helices and loops. The arched helix harbors the R/S-stereospecificity determinant Gly 416 (34). There is a small confined chamber (about 34 Å 3 ) next to the catalytic iron that is surrounded by several conserved aliphatic residues, which have been thoroughly discussed in P. homomalla 8R-LOX by Neau et al. (18). Of those residues, Leu 374 , Leu 420 , and Leu 613 form an orifice that leads to the largest cavity (188 Å 3 ), which is located under the arched helix toward the "rear" end of the enzyme (away from the PLAT domain). The bottom of this cavity is composed of residues Thr 365 , Val 430 , and Val 609 , which coincide with the regiospecificity determinants described for the boot-shaped channel in 12/15-LOX (7); Leu 431 , also a regiospecificity determinant of some LOXs (8,12); and Met 606 , a position claimed to be relevant in the binding orientation of the fatty acid substrate (49). On one side of the arched helix, the entrance to the active site chamber is blocked by Phe 185 that sits on a loop between helices 4 (3 10 -helix) and ␣2. On the other side, a gap between Leu 374 , Ile 412 , and Leu 420 leads to a pair of cavities (95 and 113 Å 3 ) that also reach toward the protein surface. These cavities are lined perpendicularly to the largest one, and are separated from each other by a constriction formed by Leu 379 , Ile 421 , and a conserved salt bridge Glu 382 -Arg 417 , that participates in lodging the arched helix. Access on this side is obstructed by Tyr 154 and the loop where that residue is situated. Let us call the orifices blocked by Phe 185 and Tyr 154 entrances A and B, respectively. Both of these entrances have a positively charged residue in the vicinity, which could neutralize the carboxylate group of a fatty acid, and therefore, aid to position the substrate for catalysis: these are Arg 186 for entrance A, and Arg 153 for B. It is interesting to note that the chain fragments that constitute the lids over both entrances interact with the PLAT domain, there is a -cation interaction between Trp 107 and Lys 172 , as well as an H-bond connecting the main chain N-H of Trp 107 with Asp 173 (Fig. 2A). The lids of entrances A and B are on the C-and N-terminal sides of the interface, respectively. Therefore, either one of the entrances could be regulated via this bridge. The -cation bond appears Additionally, there are hydrophobic residues in the C-terminal part of the lid that could bind the lipid membrane. B, in the active site, the non-heme iron is coordinated by highly conserved residues (orange). The volume of the hydrophobic pocket (light green) has been found to be important in catalysis specificity (7,12). Residues marked in green were mutated in this study to investigate that hypothesis. Another set of cavities (cyan) could give access to the active site via the alternative entrance B. C, schematic depictions of hypothesized substrate-binding channels and substrate orientations (viewing angle is analogous to panels A and B). The boot-shaped cavity is defined as a passage from entrance A to the bottom of the hydrophobic pocket (7,12). The U-shaped channel would stretch between the two entrances (18). Yet another possibility is that the fatty acid tail is always bound into the hydrophobic pocket, but either one of the entrances is used depending on the enzyme isoform, this yields the T-shaped channel.
to be conserved as it is present in all published lipoxygenase crystal structures (18, 20, 22, 50 -53) (Fig. 3A). The PLAT domain Trp 107 is invariant among studied LOXs and is a part of the conserved sequence FPCYRW on the ␤7 strand of animal LOX (28). The cationic residue of the catalytic domain (Lys 172 in 11R-LOX) is more variable, but can still be found in either of the two positions shown on the alignment (Fig. 3B).
Ca 2ϩ -binding Sites-The 11R-LOX crystals were obtained in calcium-free conditions, but for catalytic activity, the presence of Ca 2ϩ is a must. When compared with available structures of Ca 2ϩ -PLAT complexes, the Ca 2ϩ -binding loops of the apo-domain in 11R-LOX differ significantly. The PLAT domains of coral 8R-LOX (21) and gangrene ␣-toxin (54) both contain three occupied binding sites, these are formed by three adjacent loops and are well conserved both in sequence and structure (21). All three sites are also preserved in the 11R-LOX sequence, yet are absent in the tertiary structure (Fig. 4). The first Ca 2ϩ complex is formed by a turn in the ␤3-␤4 loop; in the apo-PLAT domain the turn is missing, and instead, there is an extra ␤-strand-like segment. The second site is situated between another turn in the ␤3-␤4 loop and ␤1-␤2 loop; again, the loops in 11R-LOX are arranged differently making a  smoother curve and reaching further away from the ␤-sandwich core. The same goes for the third site that lies between a turn in the ␤1-␤2 loop and the ␤5-␤6 loop. Site III is also the hardest to detect by sequence comparison as it is mostly defined by main chain atoms. Fig. 4 illustrates that the Ca 2ϩ -binding sites are intertwined; therefore, a cascade of conformational changes could occur upon metal chelation. The aforementioned Trp 107 is right next to site III and makes contacts with residues that define the site in both apo-and holo-PLAT domains. This suggests it could be communicating the allosteric effect that occurs upon calcium and membrane binding in the PLAT domain to the lid segment.
Site-directed Mutagenesis of Substrate Channel-Several residues were mutated in the largest of the internal cavities to study its intrinsic role in the specificity of catalysis. Bulky, hydrophobic Val 430 , Leu 431 , Met 606 , and Val 609 were substituted with a compact alanine to deepen the boot-shaped channel, and potentially alter the regiospecificity of the enzyme, as has been previously shown by Kühn et al. (7). In the case of 11R-LOX, the additional space was expected to cause a frameshift of substrate binding, resulting in a novel 8-LOX activity, unseen in the wild-type. The V430A and V609A substitutions failed to alter the position of hydrogen abstraction, although the corresponding residues have been described as regiospecificity determinants for rabbit 12/15-LOX (11) and many others (12) (full sequence alignment is provided in supplemental Fig.  S2). Rather, the V609A mutant suffered from general loss of positional control, exhibiting increased 15-HpETE production ( Table 2). On the other hand, modest 8-LOX activity was observed with L431A and M606A mutants, reaching up to 10% of 8-HpETE production. Further kinetic studies of those enzymes showed that there was no remarkable change in substrate affinity (K m ) (Table 3). However, the catalysis efficiency (k cat /K m ) of the L431A mutant was only one-fourth of the wildtype as the turnover rate (k cat ) had dropped. One possible interpretation of this data is that the substrate cannot align in an orientation appropriate for hydrogen abstraction as the hydrophobic pocket has a role in substrate positioning. Granted, the data do not rule out an effect on the catalytic machinery itself, but the fact that enzyme retains catalytic activity might suggest that the iron coordination sphere remains intact. Moreover, the residue at this position varies significantly among LOX isoforms so it is in all likelihood not an essential element of the core LOX fold.
Although the data display a trend, these studies do not rule out the use of the alternative U-shaped channel (Fig. 2C). To find more substantial evidence to differentiate between the channels, Val 609 , which lies near the proximal side of the hydrophobic pocket, was substituted with a large tryptophan to block the distal end of the cavity, and propagate the usage of the U-shaped channel. As a consequence, the catalytic efficiency of the enzyme plunged 50-fold, but surprisingly, the K m remained unaffected despite the dramatic reduction in binding space that such a mutation should have caused. Apparently, the loss of activity was entirely due to the diminished turnover rate (Table  3). Reaction specificity suffered greatly, as well, as the share of 11-HpETE dropped down to 69% and a multitude of by-products (15/5/8/12-HpETE in descending order by proportion) was formed ( Table 2). The chirality of the three major products of V609W mutant, 11/15/5-HpETE, was determined using chiral phase HPLC, to confirm the substrate orientation in the active site. Practically pure 11R and 5S products were detected, oxygenation at C15 created an R/S (35:65) mixture (data not shown). The formation of 5S-HpETE intimates a head-first binding if entrance A is considered as the point of entry, whereas for 11R-HpETE, tail-first orientation has been suggested (25).
The interpretation of these data is not straightforward. If one invokes the use of the boot-shaped cavity, the residual activity of the V609W mutant may be a result of incomplete closure of the cavity and a motional flexibility to allow room for substrate entry despite the bulky tryptophan (Fig. 2C). On the other hand, if the fatty acid binds into the U-shaped channel, the role of the hydrophobic pocket may be to provide the flexibility necessary for the substrate to product transition. This would easily explain the unchanged K m . Nevertheless, the various regiospecificities of distinct LOXs must somehow be reflected by their binding sites (e.g. different cavity volumes). The U-shaped channel is highly conserved, as emphasized by Neau et al. (18), but the invariant amino acids alone cannot explain the distinct products among lipoxygenases. Those side chains that impart specificity would be expected to lie outside the cluster of conserved amino acids. The hydrophobic pocket fulfills that criterion, and binding of AA in that cavity is supported by product shifts in L431A and M606A mutants.

DISCUSSION
Allosteric Lid Segment-The substrate channel entrance A of 11R-LOX is blocked by Phe 185 . Interestingly, a similar element, a Phe-Tyr "cork," has been described in human 5-LOX (22). This cork is situated where the other LOX crystal structure models are open to allow access to the catalytic site. In these  two structures, however, the ␣2 helix is considerably shorter and is flanked by loops and 3 10 -helices forming a lid that covers entrance A (Fig. 5). In the N-terminal end of this motif is the conserved -cation bridge (Trp 107 -Lys 172 ) that connects the PLAT and the catalytic domain. The -cation bridge is preceded in turn by the lid of entrance B, with Tyr 154 blocking the access. The corresponding cork in human 5-LOX is Trp 147 , and again, the similarity between the coral and human enzyme lids is remarkable. Notably, this is the entrance Gilbert et al. (22) have suggested to be utilized in 5-LOX, as opening it requires only a rotamer flip, and this way the substrate can enter the channel "tail first." The stereochemistry of the products of 5and 11R-LOX suggests that AA is bound in inverse orientations in these enzymes: may it be a result of the substrate entering the active site in inverse orientation using the same orifice or the utilization of different entrances. In either case, access to the catalytic site in both enzymes requires a conformational change, an opening of the lid. The structure of 11R-LOX can provide a stable framework for understanding the relationship between Ca 2ϩ -dependent membrane binding and lid opening. It has been speculated that the interface of PLAT and catalytic domains, including the conserved FPCYRW fragment on the ␤7 strand, might be involved in an allosteric regulatory mechanism, transmitting a conformational change in the PLAT domain induced by calcium and membrane binding to the lid component in the catalytic domain (28). We propose, that the conserved -cation bond could be mediating this interaction, as it stands right between the Ca 2ϩ -binding sites and the allosteric lid. The different regulatory properties of various lipoxygenases (e.g. the necessity of calcium and membranes for 11R-LOX catalysis) can be explained by differing conditions that are needed to set off the cascade of conformational changes, which among other contributors depend strongly on the structure of the Ca 2ϩ -binding loops (55). The activity that is present without any inducing factors in most LOXs can be attributed to a semi-open orifice, or possibly allosteric binding of the substrate itself. Additionally, in human 5-LOX the FPCYRW fragment has been shown to be involved in the binding of the coactosinlike protein, which promotes leukotriene formation (56).
Although the exact mechanism of the latter is unknown, it further substantiates the regulatory role of the bridge.
In 11R-LOX, entrance A seems the more plausible access route for several reasons. First, the same orifice is used in rabbit 15S-LOX according to docking studies (14). For 11R-and 15Sspecificity, the same hydrogen must be abstracted in the initial step of catalysis; thus, substrate binding should also be identical. Second, coral 8R-LOX, the binding of which should differ from 11R-and 15S-LOX only by "frameshift," has been suggested to employ entrance A as well (18). Another detail in favor of entrance A is that there are several bulky, hydrophobic residues like Phe 192 , Phe 201 , and Trp 204 on the C-terminal end of its lid fragment, distal to the PLAT domain interface (Fig. 5). With slight conformational changes, these residues could readily anchor the catalytic domain to the lipid membrane, and facilitate an additional mechanism of lid removal. Analogous residues are present in other lipoxygenases, too, including human 5-LOX. Furthermore, the entrance A lid is followed by the putative PDZ domain, which might also contribute to allosteric regulation, making this opening the more likely candidate not only for 11R-LOX, but for other LOXs as well.
Additional experimental data are essential to elucidate the possible access portals in this family of enzymes. In a recent experiment, removing the entrance A cork of olive LOX1 by site-directed mutagenesis augmented the activity of the enzyme remarkably (57). A similar approach could be used in further study to determine the true substrate entrance of 11R-LOX, but also of other LOXs.
T-shaped Substrate Channel-In general, the 11R-LOX model contains a closed, roughly T-shaped system of cavities, wherein the active site iron is located at the junction of perpendicular channels, and the potential entrances for the substrate are situated at both ends of the "T-bar" (Fig. 2C). Although the channel system seems to be segmented in the model, minimal side chain movements are necessary to connect the neighboring pockets. In this system of cavities, the so called boot-shaped channel described by Kühn et al. (7,12) would constitute the passage from entrance A to the bottom of the hydrophobic pocket. The alternative U-shaped channel proposed by Neau et al. (18), on the other hand, would consist of a culvert stretching below the arched helix and connecting both entrances, and thus, disregarding the pocket altogether.
The results obtained by site-directed mutagenesis of 11R-LOX suggest that the integrity of the T-shaped channel is required for proper positioning of the substrate. The fact that L431A and M606A substitutions resulted in an 8/11-LOX, albeit with modest amounts of the 8-product, suggests that AA enters the hydrophobic pocket tail-first. This model is also supported by the dramatic reduction of catalytic activity and specificity when the pocket was blocked by the V609W substitution, even though the kinetic parameters for that mutant have left room for alternative interpretations. It is likely that in regard to substrate binding, 11R-LOX is analogous to the enzymes described to have a boot-shaped channel (e.g. rabbit 12/15-LOX).
The presence of cavities that connect the hypothetical entrance B with the active site still makes one question their potential role. The U-shaped channel is lined with highly con- served Leu and Ile residues that imply a structure-functional importance. However, the highly conserved amino acids alone cannot define the different catalytic properties of lipoxygenases. And whereas the cavity that forms the B side of the T-site may provide an entry way for molecular oxygen access as suggested for soybean LOX-1 (58,59), it is not clear whether leucines, as opposed to any hydrophobic amino acids, are necessary for an O 2 channel. It just might be that distinct lipoxygenases each utilize the central core of the binding site, but regiospecificity is defined by the access to that core. One could imagine a theory that merges the boot-and U-shaped passages, yielding a T-shaped substrate channel. Depending on the catalytic specificity of a particular LOX, the substrate could enter tail-first utilizing either one T-bar entrances. Additional mechanisms like positively charged residues could further induce and stabilize the substrate binding. Yet, for specificity, the aliphatic tail requires the internal hydrophobic pocket. Further studies, especially co-crystallization of the enzyme with the substrate could bring more definitive answers to these matters.