J. Mol. Biol. (1992) 227, 818-839
The Crystal and Molecular Structure of the Rhizmmaw Triacylglyceride Lipase at 1*9 A Resolution
miehei
Zygmunt S. Derewendat, Urszula Derewenda MRC Group in Protein Structure and Function Department of Biochemistry, University of Alberta 4-74 Medical Sciences Building, Edmonton, Alberta, Canada T6G 2H7
and Guy G. Dodson University
Department of Chemistry of York, Heslington, York YOl
(Received 23 March
5DD,
U.K.
1992; accepted 6 June 1992)
The crystal and molecular structure of a triacylglyceride lipase (EC 3.1.1.3) from the fungus Rhizomucor miehei was analyzed using X-ray single crystal diffraction data to 1.9 A resolution. The structure was refined to an R-factor of 0169 for all available data. The details of the molecular architecture and the crystal structure of the enzyme are described. A single polypeptide chain of 269 residues is folded into a rather unusual singly wound j-sheet domain with predominantly parallel strands, connected by a variety of hairpins. loops and helical segments. All the loops are right-handed, creating an uncommon situation in which the central sheet is asymmetric in that all the connecting fragments are located on one side of the sheet. A single N-terminal a-helix provides the support for the other. distal. side of the sheet. Three disulfide bonds (residues 29-268, 40-43, 235-244) stabilize the molecule. There are four cis peptide bonds, all of which precede proline residues. In all, 230 ordered water molecules have been identified; 12 of them have a distinct internal character. The catalytic center of the enzyme is made up of a constellation of three residues (His25T. Asp203 and Ser144) similar in structure and funct’ion to the analogous (but not homologous) triad found in both of the known families of serine proteinases. The fourth residue in this system equivalent to Thr/Ser in proteinases), hydrogen bonded to Asp. is Tyr260. The catalytic site is concealed under a short amphipatic helix (residues 85 to 91), which acts as “lid”, opening the active site when the enzyme is adsorbed at the oil-water int,erface. In the native enzyme the “lid” is held in place by hydrophobic interactions. Keywords:
lipase; protein crystallography;
targets for chemotherapy.
1. Introduction
interest
3.1.1.3), Triacylglyceride hydrolases (EC commonly known as neutral lipases, constitute a ubiquitous and diverse family. In mammals they play a key role in lipid digestion (e.g. lingual, gastric and pancreatic lipases), reconstitution and catabolism (e.g. intestinal lipases, hepatic, lipoprotein and lysosomal lipases). The recent advances in lipoprotein biochemistry and the impact these studies have on various medical problems, such as obesity interest in atherosclerosis, stimulated and structure-function relationships in these enzymes. Some mammalian enzymes are thought to be good
$08.00/O
in
the
industrial
There is also considerable potential
of
lipases,
although in this case bacterial and fungal enzymes appear to be more useful. In spite of this interest, relatively little is known so far about the molecular architecture of lipases. To date, preliminary accounts of crystal structure determinations for three neutral triglyceride lipases have been reported: an extracellular enzyme purified from a fungus Rhizomucor miehei (RmL$: Brady et al., 1990), human pancreatic lipase $ Abbreviations used: RmL, Rhizomucor miehei hPL, human pancreatic lipase; GcL, (?eotrychum candidurn lipase; m.i.r., multiple isomorphous replacement: AChE, Torpedo culifornica acetylcholinesterase.
t Author to whom all correspondence should be addressed. 0022-2836/92/190818-22
serine hydrolase
818
lipase;
$;) 1992 Academic Press Limited
R. miehei Lipase at 1.9 d Resolution (Winkler et al., 1990) and, most recently, another fungal enzyme from Geotrychum candidurn (Schrag et al., 1991). All three enzymes contain mixed (though predominantly parallel) /?-pleated sheets, although the connectivities between strands vary. In all cases the active centers were found to be buried. In RmL a short helix, referred to as the “lid”, covers the active site and its displacement in response to the enzyme’s adsorption to an oil-water interface was postulated as the structural basis for the phenomenon of interfacial activation (Brady et al., 1990). Similarly, a conformational change was also postulated by Winkler et aE. (1990) for the human pancreatic lipase (hPL) and by Schrag et al. (1991) for the G. candidum lipase (GcL). Recently, single crystals of RmL complexed with a covalent inhibitor (n-hexylphosphonate-ethylester) were obtained and the subsequent crystallographic resolution analysis carried out at 3A (1 A = 0.1 nm) demonstrated that the helical lid of RmL indeed undergoes a dramatic hinged body type movement (Brzozowski et al., 1991). Another crystallographic study at 2.6 A resolution of RmL by diethyl p-nitrophenylphosphate inhibited (Derewenda et al., 1992) confirmed these results and provided a more detailed molecular description of the activated enzyme. No other structures of lipases in their “active” state have been reported to date. Here, we present a detailed analysis of the threedimensional structure of the Rhizomucor miehei enzyme
in its native
experimental
model described
(Brzozowski
form.
A full
account
of the
work which resulted in the molecular below has been published
elsewhere
et al., 1992).
2. Materials and Methods Crystallization, data collection, phasing and refinement have been described in detail elsewhere (Brzozowski et aZ., 1992). A short summary is given below. Crystals were grown by the hanging drop method. Purified recombinant protein, expressed in Aspergillus
oryzae was dissolved in 20 mn-Tris. HCI (pH &05) to a concentration of 15 to 16 mg/ml; 55 to 75% (v/v of saturated solution) phosphate buffer was used as a precipitant to grow single crystals. The crystals were b = 75.0 A, orthorhombic a = 71.6 A, u3,212,, c = 55-O A) with a specific volume of 2.5 A3 per dalton indicating solvent content of -45%, well within the range expected for globular proteins. Native data to l-9 A resolution and all heavy atom derivative data were collected using the Xentronics (Siemens) area detector and processed using the XENGEN suite of programs (Howard et al., 1987). Three derivatives (Hg, Pt and I) were used. Partial interpretation of the multiple isomorphous replacement (m.i.r.) electron density map was made possible with the assistance of a molecular replacement solution using the atomic model of the Humicola lanuginoso lipase (for a full account of the chronology of this study, see Brozozowski et al., 1992). Many of the loops, however, were not visible, and routine least-squares refinement did not result in any significant improvment. Poly-Ala segments were introduced in place of these loops and simulated annealing (Brunger, 1988) was used to find their conformations. Two rounds of this procedure finally
819
yielded a fully interpretable map. The atomic model was refined further by restrained crystallographic refinement and the hydrogen atoms were generated using XPLOR (Brunger, 1988). The co-ordinates have been deposited in the Protein Data Bank (accession number TGL3). Unless otherwise specified, all the programs referred to in this paper have been used in their Vax VMS or Silicon Graphics Unix versions of the CCP4 suite of crystallographic programs (SERC Daresbury Laboratory, U.K.). A CCP4 version of FRODO (Jones, 1978) modified by Dr P. Evans (MRC Laboratory of Molecular Biology, Cambridge, U.K.) for a PS330 Evans & Sutherland computer graphics system was used for all structural comparisons.
3. Results and Discussion (a) Quality of the rejkement and of the structure The final structure consists of 2289 non-hydrogen atoms of which 230 are water oxygens. Table 1 shows some of the stereochemical parameters of this model; full details are given by Brzozowski et al. (1992). The mean error in co-ordinates calculated according to the method of Luzzati (1952) was 613 A. The final crystallographic R-factor was 9129 for data with a low-resolution cutoff of 7.5 A and an intensity cutoff of 2o(F). In addition, those reflections for which P&FC,,C or F~,,/F,,,, exceeded 2.5 were also excluded. The latter criterion was introduced to eliminate the bad reflections not screened out during the merging step. The probability of such bad errors in area detector data sets is relatively high and we find that the elimination of even a handful of such bad terms improves the quality of refinement. Consequently, out of 19,404 unique reflections, 18,960 were used. The variation of the R-factor with resolution (Table 2) shows the expected increase at low and high-resolution ends; the former is associated with inaccurate description of the solvent continuum and the lack of hydrogen atoms, while the latter arises from the decrease in the value of F/a(F) and the quality of data. The average temperature (B) factor, including solvent molecules, is 22.3 A’. This value is within the range normally observed in protein crystals. An analysis of the final difference electron density map and of the temperature factors for the water molecules indicates no ions or molecules heavier than water.
Table 1 Stereochemical parameters for the refined atomic model
(l-2) Bond distances (d) Planarity (A) Chiral volumes (A3) Non-bonded single-torsion contacts ((A) Conformation planar torsion angles (“) Isotropic temperature factor (mainchain) (A*)
Standard deviation
0
0015 0010 0.216 0.210 2.30
O-010 0.010 0.100 0500 300
523
190
820
Z. S. Derewenda et al.
n II
The Crystal and Molecular Structure of the Rhizmmaw Triacylglyceride Lipase at 1*9 A Resolution
miehei
Zygmunt S. Derewendat, Urszula Derewenda MRC Group in Protein Structure and Function Department of Biochemistry, University of Alberta 4-74 Medical Sciences Building, Edmonton, Alberta, Canada T6G 2H7
and Guy G. Dodson University
Department of Chemistry of York, Heslington, York YOl
(Received 23 March
5DD,
U.K.
1992; accepted 6 June 1992)
The crystal and molecular structure of a triacylglyceride lipase (EC 3.1.1.3) from the fungus Rhizomucor miehei was analyzed using X-ray single crystal diffraction data to 1.9 A resolution. The structure was refined to an R-factor of 0169 for all available data. The details of the molecular architecture and the crystal structure of the enzyme are described. A single polypeptide chain of 269 residues is folded into a rather unusual singly wound j-sheet domain with predominantly parallel strands, connected by a variety of hairpins. loops and helical segments. All the loops are right-handed, creating an uncommon situation in which the central sheet is asymmetric in that all the connecting fragments are located on one side of the sheet. A single N-terminal a-helix provides the support for the other. distal. side of the sheet. Three disulfide bonds (residues 29-268, 40-43, 235-244) stabilize the molecule. There are four cis peptide bonds, all of which precede proline residues. In all, 230 ordered water molecules have been identified; 12 of them have a distinct internal character. The catalytic center of the enzyme is made up of a constellation of three residues (His25T. Asp203 and Ser144) similar in structure and funct’ion to the analogous (but not homologous) triad found in both of the known families of serine proteinases. The fourth residue in this system equivalent to Thr/Ser in proteinases), hydrogen bonded to Asp. is Tyr260. The catalytic site is concealed under a short amphipatic helix (residues 85 to 91), which acts as “lid”, opening the active site when the enzyme is adsorbed at the oil-water int,erface. In the native enzyme the “lid” is held in place by hydrophobic interactions. Keywords:
lipase; protein crystallography;
targets for chemotherapy.
1. Introduction
interest
3.1.1.3), Triacylglyceride hydrolases (EC commonly known as neutral lipases, constitute a ubiquitous and diverse family. In mammals they play a key role in lipid digestion (e.g. lingual, gastric and pancreatic lipases), reconstitution and catabolism (e.g. intestinal lipases, hepatic, lipoprotein and lysosomal lipases). The recent advances in lipoprotein biochemistry and the impact these studies have on various medical problems, such as obesity interest in atherosclerosis, stimulated and structure-function relationships in these enzymes. Some mammalian enzymes are thought to be good
$08.00/O
in
the
industrial
There is also considerable potential
of
lipases,
although in this case bacterial and fungal enzymes appear to be more useful. In spite of this interest, relatively little is known so far about the molecular architecture of lipases. To date, preliminary accounts of crystal structure determinations for three neutral triglyceride lipases have been reported: an extracellular enzyme purified from a fungus Rhizomucor miehei (RmL$: Brady et al., 1990), human pancreatic lipase $ Abbreviations used: RmL, Rhizomucor miehei hPL, human pancreatic lipase; GcL, (?eotrychum candidurn lipase; m.i.r., multiple isomorphous replacement: AChE, Torpedo culifornica acetylcholinesterase.
t Author to whom all correspondence should be addressed. 0022-2836/92/190818-22
serine hydrolase
818
lipase;
$;) 1992 Academic Press Limited
R. miehei Lipase at 1.9 d Resolution (Winkler et al., 1990) and, most recently, another fungal enzyme from Geotrychum candidurn (Schrag et al., 1991). All three enzymes contain mixed (though predominantly parallel) /?-pleated sheets, although the connectivities between strands vary. In all cases the active centers were found to be buried. In RmL a short helix, referred to as the “lid”, covers the active site and its displacement in response to the enzyme’s adsorption to an oil-water interface was postulated as the structural basis for the phenomenon of interfacial activation (Brady et al., 1990). Similarly, a conformational change was also postulated by Winkler et aE. (1990) for the human pancreatic lipase (hPL) and by Schrag et al. (1991) for the G. candidum lipase (GcL). Recently, single crystals of RmL complexed with a covalent inhibitor (n-hexylphosphonate-ethylester) were obtained and the subsequent crystallographic resolution analysis carried out at 3A (1 A = 0.1 nm) demonstrated that the helical lid of RmL indeed undergoes a dramatic hinged body type movement (Brzozowski et al., 1991). Another crystallographic study at 2.6 A resolution of RmL by diethyl p-nitrophenylphosphate inhibited (Derewenda et al., 1992) confirmed these results and provided a more detailed molecular description of the activated enzyme. No other structures of lipases in their “active” state have been reported to date. Here, we present a detailed analysis of the threedimensional structure of the Rhizomucor miehei enzyme
in its native
experimental
model described
(Brzozowski
form.
A full
account
of the
work which resulted in the molecular below has been published
elsewhere
et al., 1992).
2. Materials and Methods Crystallization, data collection, phasing and refinement have been described in detail elsewhere (Brzozowski et aZ., 1992). A short summary is given below. Crystals were grown by the hanging drop method. Purified recombinant protein, expressed in Aspergillus
oryzae was dissolved in 20 mn-Tris. HCI (pH &05) to a concentration of 15 to 16 mg/ml; 55 to 75% (v/v of saturated solution) phosphate buffer was used as a precipitant to grow single crystals. The crystals were b = 75.0 A, orthorhombic a = 71.6 A, u3,212,, c = 55-O A) with a specific volume of 2.5 A3 per dalton indicating solvent content of -45%, well within the range expected for globular proteins. Native data to l-9 A resolution and all heavy atom derivative data were collected using the Xentronics (Siemens) area detector and processed using the XENGEN suite of programs (Howard et al., 1987). Three derivatives (Hg, Pt and I) were used. Partial interpretation of the multiple isomorphous replacement (m.i.r.) electron density map was made possible with the assistance of a molecular replacement solution using the atomic model of the Humicola lanuginoso lipase (for a full account of the chronology of this study, see Brozozowski et al., 1992). Many of the loops, however, were not visible, and routine least-squares refinement did not result in any significant improvment. Poly-Ala segments were introduced in place of these loops and simulated annealing (Brunger, 1988) was used to find their conformations. Two rounds of this procedure finally
819
yielded a fully interpretable map. The atomic model was refined further by restrained crystallographic refinement and the hydrogen atoms were generated using XPLOR (Brunger, 1988). The co-ordinates have been deposited in the Protein Data Bank (accession number TGL3). Unless otherwise specified, all the programs referred to in this paper have been used in their Vax VMS or Silicon Graphics Unix versions of the CCP4 suite of crystallographic programs (SERC Daresbury Laboratory, U.K.). A CCP4 version of FRODO (Jones, 1978) modified by Dr P. Evans (MRC Laboratory of Molecular Biology, Cambridge, U.K.) for a PS330 Evans & Sutherland computer graphics system was used for all structural comparisons.
3. Results and Discussion (a) Quality of the rejkement and of the structure The final structure consists of 2289 non-hydrogen atoms of which 230 are water oxygens. Table 1 shows some of the stereochemical parameters of this model; full details are given by Brzozowski et al. (1992). The mean error in co-ordinates calculated according to the method of Luzzati (1952) was 613 A. The final crystallographic R-factor was 9129 for data with a low-resolution cutoff of 7.5 A and an intensity cutoff of 2o(F). In addition, those reflections for which P&FC,,C or F~,,/F,,,, exceeded 2.5 were also excluded. The latter criterion was introduced to eliminate the bad reflections not screened out during the merging step. The probability of such bad errors in area detector data sets is relatively high and we find that the elimination of even a handful of such bad terms improves the quality of refinement. Consequently, out of 19,404 unique reflections, 18,960 were used. The variation of the R-factor with resolution (Table 2) shows the expected increase at low and high-resolution ends; the former is associated with inaccurate description of the solvent continuum and the lack of hydrogen atoms, while the latter arises from the decrease in the value of F/a(F) and the quality of data. The average temperature (B) factor, including solvent molecules, is 22.3 A’. This value is within the range normally observed in protein crystals. An analysis of the final difference electron density map and of the temperature factors for the water molecules indicates no ions or molecules heavier than water.
Table 1 Stereochemical parameters for the refined atomic model
(l-2) Bond distances (d) Planarity (A) Chiral volumes (A3) Non-bonded single-torsion contacts ((A) Conformation planar torsion angles (“) Isotropic temperature factor (mainchain) (A*)
Standard deviation
0
0015 0010 0.216 0.210 2.30
O-010 0.010 0.100 0500 300
523
190
820
Z. S. Derewenda et al.
n II
