Crystal structure of linoleate 13R-manganese lipoxygenase in complex with an adhesion protein1

The crystal structure of 13R-manganese lipoxygenase (MnLOX) of Gaeumannomyces graminis (Gg) in complex with zonadhesin of Pichia pastoris was solved by molecular replacement. Zonadhesin contains β-strands in two subdomains. A comparison of Gg-MnLOX with the 9S-MnLOX of Magnaporthe oryzae (Mo) shows that the protein fold and the geometry of the metal ligands are conserved. The U-shaped active sites differ mainly due to hydrophobic residues of the substrate channel. The volumes and two hydrophobic side pockets near the catalytic base may sanction oxygenation at C-13 and C-9, respectively. Gly-332 of Gg-MnLOX is positioned in the substrate channel between the entrance and the metal center. Replacements with larger residues could restrict oxygen and substrate to reach the active site. C18 fatty acids are likely positioned with C-11 between Mn2+OH2 and Leu-336 for hydrogen abstraction and with one side of the 12Z double bond shielded by Phe-337 to prevent antarafacial oxygenation at C-13 and C-11. Phe-347 is positioned at the end of the substrate channel and replacement with smaller residues can position C18 fatty acids for oxygenation at C-9. Gg-MnLOX does not catalyze the sequential lipoxygenation of n-3 fatty acids in contrast to Mo-MnLOX, which illustrates the different configurations of their substrate channels.

The present study had three major goals. The first goal was to solve the crystal structure of Gg-MnLOX from the data obtained during the first crystallographic analysis (21). The second goal was to compare the active sites of Gg-and Mo-MnLOX and a model of the active site of Fo-MnLOX to identify amino acids of catalytic importance. The third goal was to study hydrogen abstraction at the n-5 and n-8 positions of long chain fatty acids and hydroperoxides at the n-10 position by Gg-, Mo-, and Fo-MnLOX to find catalytic differences and, if possible, their relation to structural elements at the active sites. We also report the first crystal structure of a fungal adhesion protein with the same fold as collagen adhesins of Staphylococcus aureus and other Gram-positive bacteria.

Enzyme expression and purification
The Gg-MnLOX precursor consists of 618 amino acids, including a predicted secretion signal of 16 amino acids (GenBank identification number (ID) AAK81882). Gg-MnLOX without the secretion signal was expressed (602 amino acids) and secreted by P. pastoris using pPICZA with the yeast -secretion signal (pPICZA_MnLOX_602) (16). Fermentation was carried out in a 10 l bioreactor with 6 l of buffered minimal methanol medium for 4-5 days, as described (21). The medium was harvested, 136 g (NH 4 ) 2 SO 4 was added per liter, the medium was adjusted with in addition to M. oryzae (9)(10)(11)(12). They are named by their catalytic properties with 18:2n-6 as a substrate or by their origin: Gaeumannomyces graminis (13R-or Gg-MnLOX), Magnaporthe salvinii (9S-or Ms-MnLOX), Aspergillus fumigatus (13-or Af-MnLOX), Fusarium oxysporum (11R/13S-or Fo-MnLOX), and Colletotrichum gloesporioides (9S/11S-or Cg-MnLOX) (2,(9)(10)(11)(12). Based on sequence homology to Gg-MnLOX, a number of LOX sequences have been denoted as putative MnLOXs at NCBI. We will denote the enzymes by their origin in this report. A phylogenetic tree of the six known and four tentative MnLOXs is shown in Fig. 1A. The amino acids at their C-terminal end are partly conserved and this motif appears to be characteristic of MnLOXs (Fig. 1B), whereas the C terminal of plant and mammalian LOXs ends by the hexamers Pro-AsnSerIleSerIle and GluAsnSerValAlaIle, respectively.

Structure analysis
Sequences were aligned with the ClustalW program. Structure comparison and superpositioning were performed with PyMOL (PyMOL molecular graphics system, version 1.7.4, Schrödinger, LLC) and Coot (28), and the former was used to prepare figures. The phylogenetic tree was constructed with MEGA6 with bootstrap tests of the nodes as described (30). Mascot analysis was used to construct peptide maps as described (http://www. matrixscience.com). Swiss-Model was used to generate two homology models of Fo-MnLOX with Mo-MnLOX and Gg-MnLOX as templates, respectively (31). The 18:2n-6 was docked to Gg-MnLOX and Mo-MnLOX with the SwissDock program (32), and the mol2 file for docking of 18:2n-6 was generated from the ligand library in SwissDock.

LC-MS analysis
Reversed-phase (RP)-HPLC with MS/MS analysis was performed with a Surveyor MS pump (ThermoFisher) and an octadecyl silica column (5 m; 2.0 × 150 mm; Phenomenex), which was eluted at 0.3 ml/min with methanol:water:acetic acid, 750:250:0.05 for separation of hydroxy fatty acids and 650:350:0.05 for separation of dihydroxy fatty acids. The effluent was subject 10 M KOH to pH 6.8, which was followed by centrifugation or filtration. The medium was then used immediately for enzyme purification or stored at 80°C.

MALDI-TOF/TOF analysis
SDS-PAGE analysis revealed that the purified sample after gel filtration contained two dominating bands, an unknown protein of 23 kDa and a protein of 70 kDa (Gg-MnLOX after deglycosylation). They were identified by tryptic digestion and peptide mapping by MALDI-TOF/TOF analysis (Bruker Ultraflex). The Mascot search of the peptide map identified Gg-MnLOX in the 70 kDa band, but did not identify the unknown protein. MS/MS analysis of the two dominant peaks in the MS spectrum in the peptide map of the unknown protein identified two peptides, which contained residues 20-32 and 216-233 of a hypothetical protein, zonadhesin, of Komagataella phaffii (GenBank ID CCA40153). K. phaffii is identical to the strain of P. pastoris commonly used in gene expression studies (22). The peptide with residues 20-32 was identified as formed by tryptic cleavage at one position and thus contained the N-terminal end.

Enzyme assay
LOX activity was measured on a dual beam spectrophotometer (Shimadzu UV-2101PC). Enzyme was mixed with 100 M of fatty acids in 0.1 M NaBO 3 (pH 9.0; at 22°C) and the UV-absorbance was followed at 235 or 237 nm or by repetitive UV spectra between 200 and 300 nm. The 9-HPOTrE (40-50 M) (or the methyl ester) was incubated in the same way. K m for Mo-MnLOX with 2.7-64 M 9S-HPOTrE was estimated in triplicates (270 nm). Products were extracted on a cartridge of octadecyl silica (SepPak/C 18 ) and hydroperoxides were reduced to alcohols with triphenylphosphine before analysis (11).

Crystallization and X-ray diffraction data collection
Crystallization of Gg-MnLOX was set up using the sitting-drop vapor diffusion method at 20°C. The drops were prepared by mixing protein solution containing 20 mg/ml of protein with an equal amount of reservoir solution of the Morpheus screen in a 96-well plate (21). Data were collected with a Pilatus detector at wavelength 0.976 Å at 100 K on beam line ID29 at the European Synchrotron Radiation Facility, Grenoble, France. The data were processed with XDS (23) and iMosflm (24). The integrated data (iMosflm) were then scaled using SCALA in the CCP4 package (25). A set of 5% of the reflections was set aside and used to calculate the quality factor R free (25). Statistics of data collection and processing are presented in Table 1.

Structure determination of Gg-MnLOX and zonadhesin
The structure of Gg-MnLOX was solved by molecular replacement with Phaser (26,27) using molecule A of the Mo-MnLOX structure as a search model [Protein Data Bank (PDB) entry: to Val-618 with gaps between Ala-424 and Asp-427 and between Ala-510 and Leu-519. Zonadhesin was built from Ala-20 to Val-339 with gaps between Ala-233 and Ser-241 and between Ala-301 and Asp-307. The atomic coordinates and structure factors have been deposited in the PDB (PDB entry: 5FX8).

Overall structure of zonadhesin and its interaction with Gg-MnLOX
The 3D structure of zonadhesin 20-339 is shown as a cartoon in the asymmetric unit in Fig. 2A. The molecule has two domains (N 1 and N 2 ), which are dominated with strands. Each domain contains two anti-parallel -sheets forming one -sandwich and the N-terminal domain has a short -helix (Glu-135 to Thr-138) connecting two strands. The fold of the two domains with -sandwiches can be described as IgG-like (see below). The sequence of the missing C-terminal domain with 248 residues lists 64 Thr/Ser residues and 68 Gly/Ala/Val residues in characteristic repeat of adhesins (34,35).
to electrospray ionization in a linear ion trap mass spectrometer (LTQ, ThermoFisher). The heated transfer capillary was set at 315°C, the ion isolation width at 1.5 atomic mass units, the collision energy at 35 (arbitrary scale), and the tube lens at about 110 V. Prostaglandin F 1 was infused for tuning. Samples were injected by an auto-injector (Surveyor autosampler plus; Thermo).

Miscellaneous
Protein concentration was estimated by UV absorption at 280 nm. SDS-PAGE was performed as described (11). Diazomethane was used for methylation.

Crystallization and structure determination
Commercial crystallization screens were used to identify suitable crystallization conditions for Gg-MnLOX. The best crystals appeared as described (21)  A data set with more than 1,100 consecutive images with an oscillation range of 0.15 degrees and exposure time of 0.043 s was collected to a resolution of 2.6 Å from a single crystal, as described (21). The crystal belongs to the space group C2 with unit-cell parameters: a = 226.6 Å, b = 50.6 Å, c = 177.92 Å, and  = 91.70° (21).
There are two molecules of Gg-MnLOX and one molecule with residues 20-339 of zonadhesin (zonadhesin 20-339 ) (GenBank ID CCA40153) in the asymmetric unit with a Matthews coefficient of 3.2 Å 3 Da 1 and a solvent content of 62%. We previously reported that there could be either two or three molecules of Gg-MnLOX in the asymmetric unit, as judged from the self-rotation function (26), and this has now been clarified. The third molecule, zonadhesin, consists of 587 residues (33), but the amino acids 1-19 and 340-587 could not be detected. Loss of the former as a secretion signal of 19 amino acids was confirmed by MALDI-TOF/TOF analysis of the tryptic peptide, which was formed by single cleavage, and contained residues 20-32 as described above. This crystallized protein was designated zonadhesin  . The sequence identity between Mo-MnLOX and Gg-MnLOX is 56%, and this allowed us to solve the structure by molecular replacement. The electron density was poor on a few locations. Gg-MnLOX could be built from Gln-46  His-290, His-294, His-478, Asn-482, the carboxyl oxygen of Val-618, and a water molecule coordinate the catalytic manganese (Fig. 3C). Four of the five manganese ligands are essential for catalysis (16). His-290 and His-294 are separated by three residues in Gg-MnLOX, 4 and the corresponding His-284 and His-289 residues of Mo-MnLOX 5 by an additional residue, Pro-288. The electron density over this area, from Tyr-289 to Glu-298, of Gg-MnLOX is shown in Fig. 4A. This region of the structure is superimposed with the corresponding region of Mo-MnLOX (Phe-283 to Glu-298) in Fig. 4B. The two pairs of His residues are well aligned, likely due to the compactness of the inserted Pro residue. Interestingly, the Pro-288 residue is not conserved in other MnLOX. A Thr residue is found at this position of Fo-MnLOX and often a Gly residue in FeLOX, but insertion of a Thr or Gly residue between His-290 and His-294 of Gg-MnLOX at the corresponding position of Pro-288 inactivated the enzyme (19). A summary of the effects of replacement of residues in the active site of Gg-MnLOX is shown in Table 2.
A small loop connecting -helix 16 (16) and 17 close to the active site harbors the metal ligand Asn-482 . This The interaction at the interface of zonadhesin  and Gg-MnLOX appeared to be partly electrostatic, as judged from the surface charge of the molecules at the contact area. The PDBePISA program (http://www.ebi.ac.uk) identified four salt bridges and fourteen hydrogen bonds between Gg-MnLOX (chain A) and zonadhesin  (chain U) at an interface area of 686 Å 2 . The predicted pI of zonadhesin 20-339 is 3.6, whereas the isoelectric point of Gg-MnLOX is estimated to be 9.7 (12). The crystal packing of two A and B chains of Gg-MnLOX with two molecules of zonadhesin 20-339 is shown in Fig. 2B. The latter did not bind near the entrances of the substrate channels. Zonadhesin 20-239 can be removed by cation exchange chromatography (16), and we found that the catalytic activity of Gg-MnLOX with and without zonadhesin 20-239 appeared to be similar.

Overall structure of Gg-MnLOX and a comparison with Mo-MnLOX
An illustration of the overall structure of Gg-MnLOX is presented in Fig. 3A. Gg-MnLOX lacks the -barrel domain with homology to the polycystin-1-lipoxygenase--toxin domain of FeLOX. Gg-MnLOX and Mo-MnLOX share the same fold as illustrated by superimposing their structures (Fig. 3B). The C  carbons of the metal ligands of Gg-and Mo-MnLOX align with a root-mean-square (rms) deviation of 0.34 Å. The geometry of the metal ligands is also Asn-482 may form hydrogen bonds with the catalytic water (2.6, 3.0, and 3.2 Å, respectively).
The 2 helix extends along one side of the protein from residue Glu-87 to Ser-124, and four of its residues are found at the entrance to the active site (Trp-100, Ala-104, Thr-108, and Tyr-112). The sequence from 9 to the end of 10 is described as the arched helix (1), and it contains a series of important residues, e.g., Ile-328, which clamps the fatty acid against the catalytic base for hydrogen abstraction (1), the Gly-Ala switch/Coffa-Brash determinant Gly-332 in the wall of the active site (37), Leu-336, Phe-337, and the Sloane determinant Phe-347 in the bottom of the active site (38). Trp-343, Leu-535 and Met-288 also build this part of the channel. Arg-538 at the end of 18 tethers the carboxyl group of fatty acids, and the side chain of the next residue, Phe-539, is positioned close to Leu-336 and Phe-337 in the active site.
The "clamp" residue Leu-336, the catalytic water, manganese, and His-474 align close to a straight line and C-11 of 18:2n-6 is positioned here for hydrogen abstraction. Ggand Mo-MnLOX catalyze suprafacial hydrogen abstraction and oxygenation, which implies that O 2 can access the pentadienyl radical from the same side as the catalytic metal. The other side is likely shielded by Phe-337 of Gg-MnLOX during oxygenation of C-13 and C-11. The aligned sequences of Gg-, Mo-, and Fo-MnLOX with marked residues from known positions of the 3D structures are shown in Fig. 5 to facilitate a comparison of the enzymes and the numbering of amino acids in critical positions. Four residues of Mo-and Fo-MnLOX have also been investigated by site-directed mutagenesis ( Table 2).
The substrate channel of Gg-MnLOX contains two conspicuous side pockets near the catalytic center (Fig. 6A, B). The side pockets are delineated by a series of residues, e.g., Leu-77, Met-288, Val-291, Thr-295, Ile-328, Phe-342, and Trp-343 (Figs. 5, 6). Some of them are replaced with larger or smaller hydrophobic residues in Mo-MnLOX (Ile-282, Ala-290, Val-323, Leu-337, and Phe-338) (Fig. 6C, D). The channel entrance and the proximal pocket appear to be slightly larger in Mo-than in Gg-MnLOX and the distal pocket larger in Gg-MnLOX. The proximal and distal side pockets near the catalytic center might be important for oxygenation at C-9 and C-13 of C 18 fatty acids, respectively. Oxygen likely enters by the same route as the fatty acid, as we could not identify a tentative oxygen channel of Gg-MnLOX.
We used docking of 18:2n-6 to estimate the substrate binding position in the active site. A model of 18:2n-6 in the active site of Mo-MnLOX yielded the best docking score based on energy minimization. The docked molecule fitted in the substrate channel with the 9Z double bond facing the proximal pocket, C-11 near the metal center, the 12Z double bond facing the distal pocket, and the carboxyl group at hydrogen bond distance to Arg-525. To compare the active site with the docked substrate, Gg-MnLOX and a model of Fo-MnLOX (see below) were superimposed onto Mo-MnLOX.
Four residues of 2 (Trp-100, Ala-104, Thr-108, and Tyr-112), two residues of 9 (Val-331 and Val-335), and Phe-342 delineate the entrance to the active site of Gg-MnLOX (Fig. 7A). Arg-538 is positioned in the substrate channel loop may bring Asn-482 close to the catalytic metal ( Table 3). This loop is also present in Mo-MnLOX (Fig. 4C), but corresponds to an -helix in 8R-LOX (7,36). The oxygen of Gln-287 forms tentative hydrogen bonds to the nitrogen of Asn-482 with a distance of 2.9 Å. His-290, Val-618, and and substrates to the active site (Fig. 8B, C). The effects of the Gly332Ala and Gly332Val replacements are summarized in Table 2.
We conclude that the different catalytic properties of Ggand Mo-MnLOX could be related to the volumes and the width of the substrate channel and the two side pockets.

Homology models of Fo-MnLOX
Fo-MnLOX shares about 55% sequence identity with Gg-and Mo-MnLOX (Fig. 1). Two models of Fo-MnLOX were generated using the PDB files of Gg- and Mo-MnLOX as templates (PDB entries: 5FX8, 5FNO). As expected, the two homology models were similar and showed an rms deviation of 0.5 Å for all atoms. Unexpectedly, both models showed a tentative oxygen channel to the active site and the side pockets of the two templates were hardly detected. The entrance to the substrate channel was relatively narrow as illustrated in Fig. 7C.

Lipoxygenation of 9S-and 9R-HPOTrE
The Arg525Leu mutant of Mo-MnLOX oxygenated 18:2n-6 insignificantly, but 9S-HPOTrE was oxygenated rapidly at C-16 (7). Whether 9S-HPOTrE is tethered by another Arg residue or metabolized by Gg-or Fo-MnLOX are unknown. Based on these considerations, we examined the mechanism of oxygenation of 9S-HPOTrE, the methyl ester, and the 9R stereoisomer. and could tether the carboxyl group of fatty acids in the same way as Arg-525 of Mo-MnLOX (7). The entrance of the substrate channel could be less accommodating in Gg-MnLOX than in Mo-MnLOX due to the Phe-342 residue, which is replaced by a Leu residue in Mo-MnLOX (Fig.  7B). The entrance appears to be even smaller in the model of Fo-MnLOX (Fig. 7C) described below.
Gly-332, Leu-336, Phe-539, Val-618, and Ile-328 of Gg-MnLOX are located near the catalytic metal below the entrance to the substrate channel, as shown in Fig. 8A. Replacement of Gly-332 (Gly-Ala switch/Coffa-Brash determinant) with larger hydrophobic residue (Ala, Val) likely narrows the channel and could reduce the access of oxygen  (16). b This mutation increases the biosynthesis of epoxyalcohols, but has little effect on the regiospecificity (17). c These mutations increase the relative oxygenation at C-9 of 18:2n-6 (19). d This mutation shifts the direction of oxygen insertion but retains the steric abstraction of the hydrogen (19). e This residue aligns with the Sloane determinant at the bottom of the substrate channel (10). f Data from (7). g Data from (9). h Oxygenation of fatty acids was almost abolished, but 9-HPOTrE was oxygenated at C-16 (7).  (7). b Ile-668 is the C-terminal residue of CspLOX2 (6). c Ile-839 is the C-terminal residue of soybean lipoxygenase-1 (sLOX-1) (8).
Mo-MnLOX oxygenated 9S-and 9R-HPOTrE to 9S,16S-DiHPOTrE and 9R,16S-DiHPOTrE, respectively, which occurs by hydrogen abstraction at the n-5 position. The two diastereoisomers were separated by RP-HPLC (Fig. 9A). Important residues, which were deduced from the 3D structure, are marked under the alignment as follows: Residues at the entrance of the substrate channel are marked by a blue ring, the Coffa-Brash and the Sloane determinants with blue hash tags (#). Important residues, which are discussed in the text and figures, are denoted in the same way with green (conserved residues) or red (variable residues) angles (^). The metal ligands are marked by an asterisk (*).

Oxygenation of long chain fatty acids by Mo-, Fo-, and Gg-MnLOX
Methyl esters of fatty acids are not oxygenated by Gg-MnLOX, and the carboxyl group of C 18 fatty acids are likely tethered to the Arg residues at the entrance of the substrate channel (7). The number of double bonds reduces the flexibility of fatty acids, and a comparison of the oxygenation of 20:4n-6 and 20:5n-3 may therefore illustrate the conformation of the active sites.
The 20:5n-3 was oxygenated by Gg-MnLOX to 13-and 15-hydroperoxyeicosapentaenoic acid (HPEPE) in almost equal amounts, and to small amounts of 10-HPEPE (Fig.  9B). The 13-HPEPE is formed by hydrogen abstraction and oxygenation of the bis-allylic C-13, whereas 15-HPEPE is formed by oxygenation of C-15 with the typical shift in the position of the double bond (14Z→13E). The n-3 double UV analysis showed that the methyl ester of 9S-HPOTrE was transformed by Mo-MnLOX to 9,16-DiHPOTrE with the development of the typical triene UV spectrum with  max at 270 nm and shoulders at 260 and 280 nm (11). K m for the oxygenation of 9S-HPOTrE was estimated by UV analysis (270 nm) in triplicate to be 3 M.
UV analysis showed that Fo-MnLOX transformed 9S-HPODE to a triene, which was identified by LC-MS as 9S,16S-DiHPOTrE (>95%). Gg-MnLOX did not oxygenate 9S-HPOTrE to a triene according to UV analysis, and this was confirmed by LC-MS analysis of products after prolonged incubation. The oxygenation of 9-HPOTrE by Mo-and Fo-MnLOX, and not by Gg-MnLOX, could be due to binding of 9-HPOTrE to Phe-342 or other hydrophobic residues in unproductive configurations. 11S-HPETE as judged from NP-and CP-HPLC-MS/MS analysis (39).

DISCUSSION
We report the three dimensional structures of Gg-Mn-LOX and zonadhesin 20-339 at a resolution of 2.6 Å. The structure of Gg-MnLOX revealed that the geometry of the metal ligands was conserved in comparison with Mo-MnLOX, but demonstrated volume differences in the substrate channels. Many assumptions of the positions and functions of amino acid residues of Gg-MnLOX, which have been subject to site-directed mutagenesis ( Table 2), can now be analyzed with greater confidence.
Gg-MnLOX preferentially oxygenates C-13/C-11, Mo-MnLOX C-9/C-11, and Fo-MnLOX C-11/C-13 of 18:2n-6. The U-shaped substrate channels differ by the size of two pockets on both sides of the catalytic center (Fig. 10). The n-7 position is also catalyzed by FeLOX of the cyanobacterium Acaryochloris marina (40), but oxygenation at the n-5 and n-3 positions by Fo-MnLOX appears to be unprecedented.
We conclude that there are productive configurations in the active sites, which allow hydrogen abstraction at the chirality of the products (Table 2), presumably by repositioning of the substrate (9).
Leu-336 and Phe-337 of Gg-MnLOX appear to position the substrate near Mn 2+ OH 2 for suprafacial hydrogen abstraction and oxygenation (Table 2). Phe residues are conserved in this position of MnLOXs, whereas Ile or Leu is present in FeLOXs (19). Phe337Ile retained abstraction of the proS hydrogen at C-11 of 18:2n-6, but altered the oxygenation by Gg-MnLOX from biosynthesis of 13R-to 13S-HPODE (19). This result and the three dimensional structure suggest that Phe-337 could shield one side the n-6 double bond for oxygenation, which could be essential for the suprafacial oxygenation mechanism. The crystal structure of Gg-MnLOX with 18:2n-6 in the active site will be needed to confirm this mechanism.
Gly-332 is located at the position known as the Gly-Ala switch (1,37). Gly or Ala at this position of many FeLOX can shift the oxygenation of 18:2n-6 between C-9 and C-13 (37). Gly-332 is located in the wall of the active site of Gg-MnLOX. The fact that Gly332Ala increased the hydroperoxide isomerase activities and Gly332Val inhibited catalysis can now be explained by space restrictions in the active site (17) (Fig. 8). Replacement of the corresponding Ala-451 of eLOX-3 with Gly reduced the prominent hydroperoxide isomerase activities, presumably by increasing oxygen access to the active site (41). Ala-451 of eLOX3 and Gly-332 of Gg-MnLOX illustrate a second effect of the Gly-Ala switch.
The 9S-HPOTrE is not oxygenated at C-16 by Gg-MnLOX. This is the most striking catalytic difference to Mo-and Fo-MnLOX. This lipoxygenation is unique in several aspects. The Arg-525 residue of Mo-MnLOX is not required (7), the methyl ester of 9S-HPOTrE is oxygenated, and the chain length is not critical (Fig. 9). Tethering of the carboxyl group of 9S-HPOTrE by other positively charged residues can therefore be excluded. Both 9S-and 9R-HPOTrE are oxygenated with S stereo configuration at C-16 (Fig. 9A). This suggests that 9-HPOTrE is positioned with C-14 (n-5) of the 12Z,15Z-pentadiene at the catalytic base for hydrogen abstraction. The lack of oxygenation of 9-HPOTrE by Gg-MnLOX could be due to binding of 9-HPOTrE in an unproductive configuration, apparently caused by differences in the substrate channels of Gg-and Mo-MnLOX (Figs. 6, 7, 10). The homology model of Fo-MnLOX suggests a relatively narrow substrate channel. This enzyme nevertheless oxygenated 9S-HPOTrE by hydrogen abstraction at the n-5 position. Fo-MnLOX lacks an Arg residue at the entrance of the substrate channel for tethering of carboxyl groups of fatty acids (Figs. 5, 7C). The oxygenation of fatty acids by hydrogen abstraction at the n-8 or n-5 positions and 9-HPOTrE suggest that the substrates enter the active site of Fo-MnLOX "tail first" as indicated in Figs. 7 and 10.
The metal ligands, His-290 and His-294, of Gg-MnLOX are located along a standard helix. Mo-MnLOX harbors an additional Pro-288 residue between the corresponding His-284 and His-289 residues (Figs. 4, 5). This nevertheless results in an almost identical orientation of the two pairs of His residues. volume of the proximal pocket (marked I in Fig. 10) is the largest in Mo-MnLOX, slightly smaller in Gg-MnLOX, and hardly noticeable in the model of Fo-MnLOX. The distal side pocket (marked II) decreases in volume from Gg-to Fo-MnLOX. The model suggests that the distal pocket of the latter could be connected to the surface by a tentative oxygen channel. Overall, the interior of the substrate channel of Gg-MnLOX is the widest and the deduced channel of Fo-MnLOX is presumably the most restricted of the three enzymes.
Phe-347 is positioned at the far end of the active site, known as the position of the Sloane determinant (38) (Fig. 6). The importance of Phe-347 for the regiospecificity of Gg-MnLOX is well-documented. First, Phe347Leu, Phe347Val, and Phe347Ala increased sequentially the oxygenation of 18:2n-6 at C-9 (10) ( Table 2). Second, Ms-MnLOX with Leu-350 in this position mainly oxygenated C-9 of 18:2n-6 (9). Ser-346 is found in this position of Fo-MnLOX (Fig. 5) and the Ser346Phe mutation changed the Bacterial adhesins, which are formed by Gram-positive bacteria, have been crystallized and the folds contain strands in analogy with IgG molecules [see (49) for review]. We used the Dali server (http://ekhidna.biocenter. helsinki.fi) to find similar structures to zonadhesin in the PDB and bacterial collagen adhesins yielded high scores with CNA of Staphylococcus aureus as the top candidate (50). Both zonadhesin and CNA contain two subdomains (N 1 and N 2 ; Fig. 11A). Each domain can be superimposed separately (Fig. 11B, C), but the orientations of the N 1 and N 2 domains differ in the two proteins.
We conclude that we have accidentally crystallized the first fungal adhesion protein with structural similarities to bacterial adhesins. The physiological function of zonadhesin is unknown, but it may adhere to proteins other than Gg-MnLOX by salt bridges and hydrogen bonds.

CONCLUSIONS
We report the crystal structure of Gg-MnLOX in complex with an adhesion protein (zonadhesin) at a resolution of 2.6 Å. The structure confirms that the metal ligands of Gg-and Mo-MnLOX are conserved, but the configuration and side pockets of the U-shaped active sites differ. The secondary coordinating sphere and hydrogen bonds likely tune the Mn 2+ and Fe 2+ metal centers to support catalysis. Zonadhesin appears to be the first crystallized fungal adhesin with structural similarities to collagen adhesins of Gram-positive bacterial pathogens.
The ligand, Asn-482, of Gg-MnLOX is situated on a short loop 2.8 Å away from manganese and not on a helix in analogy with Asn-473 of Mo-MnLOX (7), and these ligands could therefore be relatively mobile. The mutants, Asn482Leu and Asn482Gln, of Gg-MnLOX retained catalytic activities, whereas the other metal ligands are required (Table 2) (16).
The 11-HPODE is subject to rapid enzymatic fragmentation by Gg-and Mo-MnLOX (11,13). This also occurs nonenzymatically by Mn 3+ in methanol, but not by Fe 3+ (42,43). FeLOX catalyze -fragmentation at an insignificant rate with one exception, the 9R-LOX of Cyanothece (CspLOX2) (18,44). Enzymatic -fragmentation likely requires positioning of the hydroperoxide group near the catalytic center and steric factors are therefore of obvious importance (42,43). In addition, the metal ligands may adjust the redox potentials of protein-bound iron and manganese to support -fragmentation and lipoxygenation. The redox potentials of Mn 2+ /Mn 3+ and Fe 2+ /Fe 3+ differ by a factor of two, and Mn-substituted FeLOX lose their catalytic activities (44). How do Mnand FeLOX adjust redox potentials to support lipoxygenation and -fragmentation?
The first coordinating spheres of Mn-and FeLOX appear to be conserved ( Table 3). The ionic radii of the oxidized metal ions are almost identical (45). In contrast to iron, manganese is always in high spin state in biological systems, and there is a strong Jahn-Teller distortion of the octahedral coordination during oxidation to Mn 3+ (46). The second coordinating sphere and coordinated solvent molecules might play an important role in tuning the redox potentials (47,48). This has been observed with superoxide dismutases of Escherichia coli with catalytic iron or manganese. Hydrogen bonds and steric factors adjust the redox potentials for Mn-and Fe-superoxide dismutases to similar values (45). The redox potentials of Fe-and Mn-LOX may be tuned by related mechanisms, but to resolve these issues will be a challenging task and was beyond the scope of this investigation.
Zonadhesin is known as a predicted protein with characteristic repeats of Thr, Ser, Val, and Ala residues at the C-terminal domain in analogy with adhesion molecules of Candida albicans and Saccharomyces cerevisiae (33,35). Amino acid sequences with homology to zonadhesin  occur in peptidases and few other proteins of P. pastoris. 6 Adhesion molecules of bacteria and fungi are crucial for cell-cell interaction, adherence to surrounding tissues, and the virulence of pathogens (34). Zonadhesin and several other fungal adhesins contain a ligand binding N-terminal domain of 300 residues and a C-terminal domain with repeats of Thr, Ser, and hydrophobic residues (34,35). Zonadhesin 21-339 did not interact near the substrate channel of Gg-MnLOX. This may explain why we did not observe any apparent catalytic difference between preparations of recombinant Gg-MnLOX with and without zonadhesin present.