Crystal Structure of Pseudomonas aeruginosa Lipase in the Open Conformation

The x-ray structure of the lipase fromPseudomonas aeruginosa PAO1 has been determined at 2.54 Å resolution. It is the first structure of a member of homology family I.1 of bacterial lipases. The structure shows a variant of the α/β hydrolase fold, with Ser82, Asp229, and His251 as the catalytic triad residues. Compared with the “canonical” α/β hydrolase fold, the first two β-strands and one α-helix (αE) are not present. The absence of helix αE allows the formation of a stabilizing intramolecular disulfide bridge. The loop containing His251 is stabilized by an octahedrally coordinated calcium ion. On top of the active site a lid subdomain is in an open conformation, making the catalytic cleft accessible from the solvent region. A triacylglycerol analogue is covalently bound to Ser82 in the active site, demonstrating the position of the oxyanion hole and of the three pockets that accommodate thesn-1, sn-2, and sn-3 fatty acid chains. The inhibited enzyme can be thought to mimic the structure of the tetrahedral intermediate that occurs during the acylation step of the reaction. Analysis of the binding mode of the inhibitor suggests that the size of the acyl pocket and the size and interactions of the sn-2 binding pocket are the predominant determinants of the regio- and enantio-preference of the enzyme.

Lipases are triacylglycerol ester hydrolases (EC 3.1.1.3) that catalyze the hydrolysis of long chain acylglycerols. Since the beginning of the 1990s, the three-dimensional structures of an impressive number of lipases from mammalian, yeast, and microbial origin have been determined (1)(2)(3). In particular, structures of fungal and bacterial lipases have attracted inter-est because these enzymes are applied on a large scale in detergents and for stereoselective biotransformations. Their widespread biotechnological application is related to their cofactor independence, their broad substrate specificity, their high enantio-selectivity, and their stability in organic solvents (4).
Bacterial lipases have recently been classified into eight different families, with family I being the largest and consisting of six subfamilies (5). Families I.1 and I.2 contain lipases from the genus Pseudomonas. These lipases usually show pronounced differences in regio-and enantio-selectivity despite a high degree of amino acid sequence homology (more than 40%) (6,7). They are widely used in industry, especially for the production of chiral chemicals that serve as basic building blocks in the synthesis of pharmaceuticals, pesticides, and insecticides (8).
The prototype enzyme of family I.1 is the 29-kDa extracellular lipase from Pseudomonas aeruginosa, which was the first Pseudomonas lipase to be purified to electrophoretic homogeneity (9). This enzyme has since been studied in great detail mainly because of the two following reasons. (i) It plays an important role as a virulence factor of P. aeruginosa, which is an opportunistic pathogen, causing a variety of infectious diseases in immuno-compromised patients. This lipase can act synergistically with other lipolytic enzymes, leading to damage of host cellular membrane lipids including lung surfactants (10). (ii) P. aeruginosa lipase is widely used for biotechnological applications because it is a versatile enzyme that can catalyze the hydrolysis and synthesis of a wide variety of industrially important substrates with broad specificity (4,(11)(12)(13).
Among Pseudomonas lipases only those belonging to family I.2 are known in atomic detail (5). These include the lipases from Chromobacterium viscosum, Pseudomonas glumae, and Pseudomonas cepacia (the latter two renamed Burkholderia glumae and Burkholderia cepacia) (14 -17). The C. viscosum and B. glumae lipases are 100% identical in sequence, and their x-ray structures show the enzymes in a closed conformation, with an ␣-helical lid covering the active site and protecting it from the solvent phase (14,15). In contrast, the B. cepacia lipase was crystallized in an open conformation, both in the absence and in the presence of a bound substrate-like inhibitor (16 -19).
Here we report the x-ray structure of P. aeruginosa PAO1 lipase (PAL) 1  logue 1,2-dioctylcarbamoyl-glycero-3-O-p-nitrophenyl octylphosphonate, which binds covalently to the enzyme. The structure provides a structural description of the enzyme in its activated state and allows a detailed explanation of the regioand stereoselectivity of the enzyme toward triglyceride substrates. These results can be generalized to other members of the bacterial lipase family I.1.

EXPERIMENTAL PROCEDURES
Crystallization and Data Collection-The lipase from P. aeruginosa was cloned, overexpressed, and purified as described previously (20). Crystallization of the inhibited enzyme was achieved by the sitting drop vapor diffusion method at room temperature. The protein solution, containing 5 mg/ml lipase, 10 mM Tris/HCl (pH 8.0), and 1% ␤-octylpyranoside, was incubated for 90 h at 12°C in a 1:10 ratio with R C -(R P ,S P )-1,2-dioctylcarbamoyl-glycero-3-O-p-nitrophenyl octylphosphonate (R C -trioctyl) (21). After incubation, the lipase-inhibitor complex was equilibrated against a precipitant solution containing 25% 2-methyl-2,4-pentanediol, 20 mM CaCl 2 in 100 mM citrate buffer (pH 5.6). Rod-shaped crystals, suitable for x-ray analysis, grew within months. They were highly x-ray-sensitive, and therefore the data collection was performed at cryo temperature (100 K), using the mother liquor as the cryoprotectant.
The crystals diffracted up to 2.54 Å resolution using synchrotron radiation. They belong to space group P2 1 2 1 2 1 , with unit cell parameters of a ϭ 45.47 Å, b ϭ 50.96 Å, c ϭ 110.02 Å. This unit cell gives a V M value of 2.2 Å 3 Da Ϫ1 assuming 1 molecule of 29 kDa per asymmetric unit. A native data set was collected with a MAR Research image plate system at beamline BW7B of the EMBL Outstation at DESY, Hamburg. Data were integrated and merged using the DENZO/SCALEPACK package (22) and CCP4 software (23). Data processing statistics are given in Table I.
Structure Determination and Refinement-The structure was determined by molecular replacement using the program AMoRe (24) as implemented in the CCP4 program package (23). A homology model of P. aeruginosa lipase (12), built on the basis of its 42% sequence identity to B. cepacia lipase, of which a structure in the open conformation is known (Ref. 17; Protein Data Bank code 3LIP), was successfully used for a rotational search. The best result was obtained by including in the search model all the side chains. The rotational search clearly indicated a solution at ␣ ϭ 23.2°, ␤ ϭ 30.5°, ␥ ϭ 348.9°, with a correlation coefficient of 0.185. A translational search performed using data in the 4 -8-Å resolution range resulted in a solution at x ϭ 0.130, y ϭ 0.003, z ϭ 0.277, with a correlation coefficient of 0.277 and an R factor of 0.511. After rigid-body refinement of this solution, the final correlation coefficient and R factor were 0.304 and 0.501, respectively (Table I). Despite the poor quality of the initial electron density map, the covalent bond present between the nucleophile (Ser 82 ) and the phosphonate inhibitor was perfectly visible as well as the sn-3 moiety of the triglyceride analogue. This was interpreted as a clear indication of the correctness of the molecular replacement solution. Many cycles of manual rebuilding, using the program O (25), and refinement, using the program CNS (26) (rigid body and simulated annealing at the beginning, conventional positional refinement and restrained individual B-factor refinement in the final stages), were necessary to improve the electron density map and to make the loops visible that in the initial maps were not visible at all. The best refinement results were obtained using a flat bulk solvent correction. Water molecules and a calcium ion were placed according to strict density and distance criteria.
Model Quality and Accuracy-The final model consists of 285 amino acid residues, one calcium ion, 112 water molecules, and one molecule of R C -trioctyl (with the phosphorus atom in the S P configuration) covalently bound to the catalytic Ser 82 . All protein residues are visible in the electron density map as well as the sn-3 moiety of the inhibitor. For the sn-1 and sn-2 moieties, only discontinuous density is present, probably due to high mobility of these triglyceride chains, as reflected by their B-factors (55-68 Å 2 ). The crystallographic R factor and R free of the final model are 0.193 and 0.237, respectively. The stereochemical quality of the structure was assessed with the programs PROCHECK (27) and WHATCHECK (28). Only Ser 82 is in a disallowed region of the Ramachandran plot (29), as expected for the nucleophile of a member of the ␣/␤ hydrolase fold family. Val 258 is in a cis-peptide conformation. The structure was further analyzed using the program DALI (30). Refinement statistics are given in Table II. The atomic coordinates and the structure factors have been deposited to the Protein Data Bank with the entry code 1EX9.

RESULTS AND DISCUSSION
Overall Structure-P. aeruginosa lipase has a nearly globular shape with approximate dimensions of 35 ϫ 40 ϫ 50 Å 3 . Its structure consists of a "core" domain (residues 1-108 and 164 -285), showing the typical features of the ␣/␤ hydrolase fold topology (31)(32)(33), and a "cap" domain (residues 109 -163), with four ␣-helices (␣4, ␣5, ␣6, and ␣8) that shape the active site cleft (Fig. 1). Helix ␣5 and its neighboring loops can be considered as a "lid" in the open conformation, making the active site cavity accessible to solvent and substrate (Fig. 1A). Compared with the canonical ␣/␤ hydrolase fold (Fig. 1B), the first two ␤-strands are absent, and therefore, to be consistent with the numbering of the consensus ␣/␤ hydrolase fold, the first strand in the PAL structure is named ␤3. The helix corresponding to the canonical helix ␣E is absent too. Furthermore, an additional small antiparallel ␤-sheet (␤-strands b1 and b2) is inserted between strand ␤3 and helix ␣1, and an additional ␤-strand (␤Ј8) lines up with the sixth strand (␤8) of the central ␤-sheet, but in the opposite direction (Fig. 1B).
P. aeruginosa lipase is the first member of family I.1 of bacterial lipases of which a crystal structure has been elucidated. Its structure is similar to the family I.2 lipase structures from B. glumae (BGL) (15), B. cepacia (BCL) (16 -19), and C. viscosum (CVL) (14), which show 42% amino acid sequence identity to PAL. The structural similarity is mainly localized in the core domain, where the secondary structure elements correspond well (Figs. 2 and 3). Nevertheless, major differences occur in the region following strand ␤7, where in PAL a deletion of more than 20 residues results in the absence of the antiparallel ␤-sheet formed by strands b3 and b4 (Fig. 3). As a consequence, helix ␣8 of PAL superimposes on helix ␣9 of homology family I.2 lipases. Most importantly, helix ␣10 in where F obs and F calc are the observed and calculated structure factor amplitudes, respectively. b R free is calculated with 5% of the diffraction data, which were not used during the refinement. homology family I.2 structures (equivalent to helix ␣E in the canonical ␣/␤ hydrolase fold) is absent in PAL (Fig. 1B), with the corresponding residues 234 -237 classified as a "hydrogenbonded turn" (27,30). Residue Cys 235 in this turn forms a disulfide bond with Cys 183 , cross-linking the loops at the Nterminal sides of strands ␤8 and ␤Ј8 (Fig. 1). Although this disulfide bridge is conserved in BCL (Cys 190 -Cys 270 ) and BGL/ CVL (Cys 190 -Cys 269 ) (Fig. 3), the PAL Cys 183 does not superimpose well on the family I.2 Cys 190 due to a three-residue deletion just after helix ␣7 in PAL. This deletion brings the PAL Cys 183 C␣ atom about 1.5-2.0 Å removed from the position of the Cys 190 C␣ atom in BCL and BGL/CVL. As a consequence, the position of Cys 235 has shifted as well. The absence of an ␣-helical structure for residues 234 -237 allows PAL to compensate for the three-residue deletion and to bring Cys 235 at the correct distance to Cys 183 for disulfide bridge formation. Mutation of both cysteine residues to serines rendered PAL more sensitive to heat denaturation and proteolytic degradation, suggesting a role of the disulfide bond in stabilizing the molecule (13).
It is interesting to notice that PAL Cys 183 and Cys 235 are highly but not absolutely conserved among the members of bacterial lipase family I.1 (Fig. 3). A disulfide bond interaction similar to PAL is likely to be present in Vibrio cholerae, Acinetobacter calcoaceticus, Pseudomonas wisconsinensis, and Pseudomonas luteola lipases. In contrast, the Pseudomonas fragi, Phaseolus vulgaris, and Pseudomonas fluorescens lipases lack this interaction. The first two enzymes have only one cysteine residue in their sequence, Cys 75 and Cys 88 , respectively, whereas P. fluorescens has two cysteines, Cys 219 and Cys 233 , but none of them matches the positions of the cysteine residues in PAL (Fig. 3).
Since the PAL structure is in the open conformation similar to BCL, whereas BGL and CVL are in a closed conformation (Fig. 2), a direct comparison of PAL and BCL can show the structural analogies and differences of the activated form of lipases from bacterial homology families I.1 and I.2. Fig. 2 shows that the three-dimensional similarity of PAL and BCL extends to the cap region. The two open structures superimpose with an root mean square difference of 1.12 Å for 264 structural equivalent C␣ atoms out of 285 common C␣ atoms. However, the ␣-helices ␣4, ␣5, and ␣6 of the cap subdomain do not superimpose perfectly in the two enzymes. Helix ␣5 is one turn longer in PAL than in BCL because of a three-residue insertion in PAL at the C terminus of the helix. The loop connecting ␣4 and ␣5 in PAL is three residues shorter than in BCL. Furthermore, helix ␣4 in PAL is shorter too and ends in a 3 10 helical conformation (Fig. 3). Differences are also present in the location of helix ␣6, which in PAL immediately precedes helix ␣7, so much that helices ␣6 and ␣7 can be considered as one single ␣-helix with a bend of about 90°at the first residue of ␣7 (Ser 164 ).
Catalytic Triad-The catalytic triad residues Ser 82 (nucleophile), Asp 229 (acid), and His 251 are located at their canonical positions in the ␣/␤ hydrolase fold (31). Their positions and orientations are similar to the catalytic triad residues in the family I.2 lipases. Ser 82 is situated at the very sharp "nucleophile elbow" between strand ␤5 and helix ␣3, identified by the consensus sequence Gly-X-Ser-X-Gly (in PAL, X ϭ His at both positions). As in all other ␣/␤ hydrolase-fold enzymes of which and a comparison with the canonical ␣/␤ hydrolase fold is given. ␣-Helices and ␤-strands are represented by rectangles and arrows, respectively. G1 and G2 are 3 10   The phosphonate inhibitor covalently bound to the nucleophile Ser 82 , the calcium ion, and the disulfide bridge are in ball and stick representation in cyan, black, and yellow, respectively. B, secondary structure topology diagram of PAL. The catalytic triad residues (Ser 82 , Asp 229 , and His 251 ) and the position of the disulfide bridge are indicated, the three-dimensional structure has been elucidated (33), the nucleophile adopts unfavorable torsion angles ( ϭ 57°, ϭ Ϫ111°) and imposes steric restrictions on the residues located in its proximity. An intricate hydrogen-bonding pattern similar to that found in family I.2 lipases stabilizes the active site residues. The main chain nitrogen atom of Ser 82 is hydrogenbonded to the carbonyl oxygen atom of Val 105 . The Ser 82 side chain O␥ atom, covalently bound to the trioctyl inhibitor, is 3.0 Å from the catalytic histidine N⑀2 atom, but the geometry is not suitable for a hydrogen bond between these two atoms. Instead, the His 251 N⑀2 makes a 2.8-Å hydrogen bond to the O1 of the phosphonate inhibitor (Fig. 4B). The His 251 carbonyl oxygen and amide nitrogen atoms are hydrogen-bonded to the His 81 N⑀2 atom and Asp 229 carbonyl oxygen atom, respectively. Most importantly, the His 251 N␦1 is 3.1 Å from the O␦2 of the Asp 229 side chain that is responsible for stabilizing the positively charged His 251 side chain during catalysis. The Asp 229 O␦2 atom is not only at hydrogen-bonding distance from the catalytic histidine but is also very close (2.7 Å) to the Glu 254 O⑀2 atom. The Glu 254 O⑀1 atom is 3.6 Å away from the His 251 N␦1 atom and, therefore, does not interact directly with the triad histidine. However, it has been suggested for B. glumae lipase (15) that an acid residue at that position may assist the catalytic histidine during hydrolysis, either directly or via a water molecule, if the triad acid (Asp/Glu) is mutated. This could also be the case for the P. aeruginosa lipase.
Ca 2ϩ -binding Site-Approximately 15 Å from the nucleophile Ser 82 , at the same side of the enzyme where the catalytic His 251 is located, a calcium binding pocket is present. The calcium ion is octahedrally coordinated, and its ligands are the two carboxylate groups of Asp 209 and Asp 253 (both absolutely conserved in all family I.1 and I.2 lipases), the two carbonyl oxygen atoms of Gln 257 and Leu 261 , and two water molecules (Fig. 5). All the distances to calcium are within ϳ2.2-2.4 Å, and the low B factor values of the ion and its ligands (17-30 Å 2 ) reflect a relatively rigid conformation of the calcium binding region. Despite the medium resolution (2.54 Å) of the PAL crystal structure, Val 258 is unambiguously in a cis-peptide conformation. The cis-peptide bond between Gln 257 and Val 258 is stabilized by a conserved hydrogen-bonding network that extends to the catalytic His 251 . The Ca 2ϩ ion bridges helix ␣8, which forms part of the wall of the active site cleft, to the loop containing the catalytic histidine and, therefore, contributes to keep His 251 at the correct position in the active site. The binding mode of the Ca 2ϩ ion is conserved in BCL, BGL, and CVL, including the presence of a cis-peptide bond corresponding to Val 258 in PAL. Furthermore, a structure-based sequence alignment of the bacterial lipases from families I.1 and I.2 shows the conservation of an Asn/Asp residue at the position preceding the catalytic histidine (Fig. 3). This Asn/Asp residue hydrogen bonds one of the two water molecules that coordinate the calcium ion. The other residues in PAL that interact with these water molecules (Thr 205 , Ser 211 , and Asp 212 ) are also highly conserved, with some variation that is not likely to influence the calcium binding. This can be deduced, for instance, from the fact that in the BCL, BGL, and CVL structures an alanine replaces PAL Thr 205 , with the calcium-coordinating water molecule hydrogen bonded to another solvent molecule located at a position equivalent to the Thr 205 O␥1 atom in PAL.
Active Site Cleft and Inhibitor Binding-The PAL structure provides a rare example of a lipase complexed with a triglyceride-like inhibitor. The only other example is that of the family I. In both PAL-and BCL-inhibited structures the R C -trioctyl molecule is covalently linked at its phosphorus atom to the nucleophilic O␥ atom of the catalytic serine residue and perfectly fits into similar hydrophobic active site clefts. In PAL this cleft is 15 Å deep and has an ovoid shape with approximate dimensions of 10 ϫ 25 Å. It is bordered by helices ␣4, ␣5, ␣8 and by the loops of residues 16 to 29 and 255 to 259. Its walls are lined mostly with hydrophobic side chains, making it perfectly suited for lipid binding. At the base of the cleft the Met 16 side chain divides it in two parts, thus giving a characteristic boomerang-like shape to the active site, with four binding pockets, an oxyanion hole and three pockets for the different branches of the triacylglycerol substrate (Fig. 4B).
In PAL the Met 16 and His 83 main chain amide groups are 2.8 Å and 2.9 Å from one of the phosphorous oxygen atoms (O4) (Fig. 4B). A similar interaction is present in BCL involving Leu 17 and Gln 88 residues. Since the R C -trioctyl-complexed structure represents the putative tetrahedral intermediate conformation during the acylation step of the reaction, the Met 16 and His 83 backbone nitrogen atoms localize the oxyanion binding pocket in a position very well conserved within the ␣/␤ hydrolase-fold enzymes (31,33,34). Met 16 is the second residue of the tetrapeptide motif Gly-Hyd-X-Gly (Hyd ϭ Met, Leu, Val) located between strand ␤3 and helix ␣4 and highly conserved in families I.1 and I.2 of bacterial lipases (Fig. 3). In PAL this tetrapeptide motif is stabilized by a hydrogen bond between the His 14 carbonyl oxygen atom and the Arg 56 NH 2 atom. Arg 56 is a totally buried residue that lies at the beginning of helix ␣2, close to the nucleophile elbow. It connects the oxyanion hole tetrapeptide motif to the loop between strand ␤3 and helix ␣2 via the hydrogen bond Arg 56 NH1-Ser 48 O. Interestingly, Arg 56 is totally conserved in homology family I.1 as well as in homology family I.2 lipases (Fig. 3). In the open conformation of BCL the interaction pattern of this Arg (Arg 61 ) strongly resembles that of PAL Arg 56 , with the Arg 61 side chain hydrogen-bonded to the carbonyl oxygen atom of His 15 and Ser 50 . In contrast, in the closed conformations of CVL and BGL, the Arg 61 side chain is hydrogen-bonded to the carbonyl oxygen atoms of Leu 17 and Gly 51 and of Gly 16 and Phe 52 , respectively. Since the move-ments of the loop between strand ␤4 and helix ␣2 (PAL residues 45-52), of the oxyanion loop (PAL residues 14 -20), and of the lid region are correlated (17), the variation of the hydrogenbonding pattern of this conserved Arg residue might play an important stabilizing role during the opening of the lid of these lipases.
In PAL the sn-3 octyl chain of the inhibitor (atoms C4-C11) is accommodated in a large groove (about 8.5 Å ϫ 9.0 Å) formed by the side chains of Met 16 2 (B. cepacia, B. glumae, C. viscosum, and P. luteola) has been performed using the program CLUSTALW (36) and manually inspected and corrected on the basis of the three-dimensional structure comparison of PAL, BCL, BGL, and CVL. Shaded regions indicate the catalytic triad residues (blue) and identical (black) or similar (yellow) amino acids in all the aligned sequences. The two Asp residues that coordinate the calcium ion and the two Cys residues that make a disulfide bridge are highlighted by black and green triangles, respectively. ␣-Helices and ␤-strands are represented by rectangles and arrows, respectively. G1 and G2 are 3 10 helices. The color code of secondary structure elements is in accordance to Fig. 1. The figure has been prepared with ALSCRIPT (37). moiety in the acyl-binding pocket (HA) is almost identical in PAL and BCL.
The alcohol binding pocket is separated from the acyl pocket by the side chain of Met 16 and contains the bound lipid analogue in a bent "tuning fork" conformation (Fig. 4B). The sn-2 moiety (atoms N2-C30) fits in a hydrophobic pocket lined by Leu 17 , Phe 19 , Ile 22 , Val 25 , Tyr 27 , Leu 252 , Val 255 , Gln 257 , Val 258 , and Phe 259 (HH pocket). The latter residues are part of the calcium binding loop (residues 252 to 261). The HH pocket is slightly larger in PAL compared with that of BCL, and consequently, the enzyme-inhibitor interactions are less tight than in BCL. There is space for about 8 -10 fatty acid carbon atoms to be accommodated in this pocket. Longer chains would partly stick into the solvent or micelle. In the HH pocket all the enzyme-inhibitor interactions are hydrophobic, and the NH group of the carbamoyl function of this chain does not have any specific hydrogen-bonding interactions with the protein. In particular, the hydrogen bond which in BCL connects the Thr 18 O␥ atom to the carbonyl oxygen of the carbamoyl function of the sn-2 chain is absent in PAL due to the presence of a hydrophobic residue, Leu 17 , instead of the polar Thr 18 in BCL.
The sn-1 inhibitor moiety (atoms N1-C20) interacts less tightly with the enzyme than the sn-2 chain. It is located in a pocket more exposed to the solvent (HB pocket). The only specific interactions with the protein are van der Waals contacts with the side chains of Leu 17 , Leu 138 , and Phe 214 . The location of the sn-1 chain is mainly determined by a hydrophobic clamp formed by Phe 214 and Leu 138 . The sn-1 chain diverges considerably with respect to the position found in the BCL complex since in the latter lipase the bulky side chain of Phe 214 is substituted for an Ala 247 (Fig. 3). As for the sn-2 moiety, the sn-1 NH group of the carbamoyl function does not specifically interact with the protein. Pro 210 and Ser 211 are located between the two sn-1 and sn-2 moieties but too far away to make any direct interaction. Since the HH pocket hosting the sn-2 chain provides a more intimate interaction with the substrate analogue than the HB pocket occupied by the sn-1 chain, the HH pocket is most likely the one that determines the regio-preference of the enzyme toward the primary position of triglyceride substrates together with the size of the acyl chain pocket HA.
In view of the differences in the inhibitor binding modes in the alcohol binding pocket, it is not surprising that PAL and BCL differ somewhat in their stereospecificity. BCL prefers the R C -trioctyl inhibitor over the S C -trioctyl compound by a factor of 7. It has been suggested that Leu 287 and Ile 290 play an important role in conferring this preference by providing unfavorable interactions with the carbonyl oxygen (O7) of the S Ctrioctyl compound (18). Furthermore, the hydrophilic interaction of the sn-2 carbonyl oxygen (O6) with the O␥1 atom of Thr 18 favors this preference (18). In contrast, in PAL the residue corresponding to BCL Ile 290 is Val 255 , a less bulky residue, and BCL Thr 18 corresponds to Leu 17 in PAL, a non-polar residue (Fig.  3). These differences explain why PAL prefers the R C -trioctyl inhibitor only by a factor of 1.5 over the S C -trioctyl compound. 2 Conclusions-The x-ray structure of P. aeruginosa lipase provides the first three-dimensional structure of a member of homology family I.1 of bacterial lipases. The molecule is in the open conformation and the active site contains a triacylglycerol analogue covalently bound to the catalytic nucleophile, Ser 82 . The other catalytic residues are Asp 229 and His 251 . An octahedrally coordinated calcium ion stabilizes the loop containing the catalytic histidine. Sequence alignment with other lipases of homology family I.1 shows that the residues important for catalysis and fold stabilization are conserved.
The bound inhibitor unambigously identifies the position of the oxyanion hole and the three pockets that accommodate the sn-1, sn-2, and sn-3 fatty acid chains, thus providing a picture of the tetrahedral intermediate that occurs during the acylation step of the triacylglycerol hydrolysis reaction. At the base of the active site cleft the Met 16 side chain divides the cleft into two branches, thus creating the acyl pocket for the binding of the sn-3 moiety of the inhibitor on one side and the alcohol pocket for the binding 2 K. Liebeton and K.-E. Jaeger, manuscript in preparation. of the sn-1 and sn-2 chains on the other side. The interactions between inhibitor and enzyme are mostly of a hydrophobic nature. Since the pocket hosting the sn-2 chain provides a more intimate interaction with the substrate analogue than the one occupied by the sn-1 chain, the sn-2 binding pocket and the acyl chain pocket are the ones that predominantly determine the regio-preference of the enzyme. Furthermore, comparison of PAL and B. cepacia lipase, both in complex with R C -trioctyl, suggests Val 255 and Leu 17 to play a role in the differences in enantiospecificity of the enzymes toward the triacylglycerol analogue. These results pave the way for site-directed mutagenesis experiments to improve the enzyme for application in bioconversion reactions.