Crystal structure of the Saccharomyces cerevisiae monoglyceride lipase Yju3p

Monoglyceride lipases (MGLs) are a group of α / β -hydrolases that catalyze the hydrolysis of monoglycerides (MGs) into free fatty acids and glycerol. This reaction serves different physiological functions, namely in the laststepofphospholipidandtriglyceridedegradation,inmammalianendocannabinoidandarachidonicacidme- tabolism, and indetoxi ﬁ cation processes inmicrobes. Previous crystalstructuresof MGLsfromhumans and bac-teriarevealedconformationalplasticityinthecapregionofthisproteinandgaveinsightintosubstratebinding.In this study, we present the structure of a MGL from Saccharomyces cerevisiae called Yju3p in its free form and in complex with a covalently bound substrate analog mimicking the tetrahedral intermediate of MG hydrolysis. Thesestructures reveala high conservation of theoverall shapeofthe MGL capregion and alsoprovideevidence for conformational changes in the cap of Yju3p. The complex structure reveals that, despite the high structural similarity,Yju3pseemstohaveanadditionalopeningtothesubstratebindingpocketatadifferentpositioncom-pared to human and bacterial MGL. Substrate speci ﬁ cities towards MGs with saturated and unsaturated alkyl chains of different lengths were tested and revealed highest activity towards MG containing a C18:1 fatty acid.


Introduction
Lipases are hydrolases that cleave ester bonds in lipids and therefore act at a water-lipid interface. Many lipases contain a variable lid or cap region which covers the active site and forms the substrate binding pocket in combination with an α/β-hydrolase core domain. In some lipases, this cap region is considered to play a role in interfacial activation [1,2]. Monoglyceride lipases (MGLs) are a subclass of lipases and catalyze the breakdown of monoglycerides (MGs) resulting in free fatty acids and glycerol molecules. MGLs are predominantly specific towards MGs, yet are not reported to be regiospecific or stereospecific (1-, 2-or 3-MGs) [3][4][5][6][7][8][9]. MGLs are conserved across all species and serve different biological roles in different species and tissues. In mammals, MGLs play an essential role in energy homeostasis, where they catalyze the last step of phospholipid and triglyceride breakdown [10]. Furthermore, they play a role in lipid signaling by regulating the levels of the endocannabinoid 2-arachidonoyl-glycerol (2-AG), the most abundant endogenous agonist of cannabinoid receptors [11][12][13]. By hydrolyzing 2-AG, MGL also determines the availability of arachidonic acid for prostaglandin synthesis in the brain [14]. In microbes, MGs have been shown to be toxic rendering MGL important for detoxification processes [15][16][17]. Pathogenic bacteria, including Mycobacterium tuberculosis, accumulate lipids in the dormant phase. Enzymes involved in degradation of these lipid depots may be essential in providing energy and precursors for cell wall synthesis during the reactivation step and the chronic phase [18]. Due to these different biological roles, MGLs are interesting subjects for drug targeting [14,[18][19][20].
Despite the fact that MGLs have been studied for decades, very little three-dimensional (3D) structural data is available. The only experimentally determined 3D structures are those of human MGL (hMGL) and MGL from Bacillus sp. H-257 (bMGL) [12,[21][22][23][24][25]. hMGL was determined in its free form and in complex with inhibitors, bMGL in its free form and in complex with inhibitors and the natural substrate. The structures of the two enzymes are very similar despite the large evolutionary distance between bacteria and humans. Apart from the conserved α/β-hydrolase core, two additional features are especially remarkable, namely the similarity of the overall shape of the cap regions and its conformational flexibility [22,23]. Based on the conserved overall shape of the cap region in hMGL and bMGL, we hypothesized that this general cap architecture might also be present in other species. In order to test this, we herein focus on Yju3p, the MGL ortholog from Saccharomyces cerevisiae [26,27]. Low sequence identity to bMGL and hMGL precludes reliable homology modeling without introducing a heavy model bias, rendering an experimentally determined structure a key requirement. A solubility enhancing mutation was introduced into the coding sequence of Yju3p (hereafter named s-Yju3p) which allowed concentrations of the purified protein in the mg/ml range without the addition of detergents. Furthermore, we investigated whether the modes of substrate binding are comparable to MGLs with known structures by analyzing complex structures mimicking the tetrahedral reaction intermediate. Thus, we present crystal structures of s-Yju3p in its free form and in complex with an inhibitor mimicking the tetrahedral intermediate of a C20:0 MG during hydrolysis. s-Yju3p harbors an α/β-hydrolase core and a cap region similar to those of hMGL and bMGL. Interestingly, the structure of the inhibitor complex revealed differences in the mode of substrate binding in s-Yju3p compared to the other two MGLs. Analysis of the MG hydrolase activity of s-Yju3p unveiled differences in substrate preferences with respect to the saturation state of the MG substrate. These data can be rationalized by differences in the shapes of the active site cavities in the three MGLs.

Synthesis of substrate analogs
p-Nitrophenyl esters of alkyl phosphonic acids were used to mimic the natural MG substrate and employed for co-crystallization studies. The C20:0 MG mimicking substrate analog was synthesized as described previously [22]. The identity of the compound was confirmed using mass spectrometry and NMR spectroscopy. As outlined before, the substrate analog contained a carbon chain consisting of only 18 carbons, yet it mimics a C20:0 substrate. This can be rationalized by its chemical structure which has a phosphor-atom at the position where the carbon of the carbonyl group would be placed and an oxygenatom is at the position of the first methylene group of the carbon chain (Fig. 1E).

Cloning and mutagenesis of s-Yju3p
A solubility variant Leu175Ser was produced by applying QuikChange™ (Agilent Technologies) site directed mutagenesis to the Fig. 1. Structures of s-Yju3p in free form and in complex with a substrate analog mimicking a C20:0 MG. A, MG hydrolase activity assay for WT-Yju3p and s-Yju3p. B, overall structure of free s-Yju3p with the α/β-hydrolase core colored in dark orange and the cap region in yellow (PDB code 4ZWN). Surface mutations Leu175Ser and Gln264Arg are marked with black arrows; the catalytic triad is highlighted with a black rectangle. This orientation of the molecule is termed "standard orientation" throughout the manuscript. The "top-view" orientation is the result of a rotation of 75°around the x axis and −90°around the z axis from this standard orientation. The "side-view" orientation is the result of a rotation of 45°around the x axis and of −90°around the y axis from the standard orientation. C, cartoon of the isolated cap region of s-Yju3p in the top-view orientation. D, close-up view of the catalytic triad. E, substrate analog with a C18:0 carbon chain mimicking a C20:0 MG substrate [48]. The red oval indicates the p-nitrophenol leaving group that is cleaved off upon binding to the nucleophilic serine and is not present anymore in the complex structure. F, C20:0 MG analog (blue sticks) with the observed electron density; top: 2Fo-Fc map as blue mesh in the crystal structure of the s-Yju3p substrate analog complex (cutoff = 1 sigma) (PDB code 4ZXF), bottom: 2Fo-Fc map calculated from the s-Yju3p substrate analog complex were the substrate analog was omitted (cutoff = 1 sigma) as purple mesh and positive density in corresponding Fo-Fc map in green (cutoff =3 sigma). G, close up view of the active site of s-Yju3p in complex with the C20:0 MG analog (blue sticks) (α/β-hydrolase core in green, cap in orange, PDB code 4ZWN). H, s-Yju3p bound to the C20:0 substrate analog with the side chains surrounding the carbon chain of the substrate analog shown as sticks, green: s-Yju3p α/β-hydrolase core, orange: s-Yju3p cap region, blue: C20:0 substrate analog. I, opening of the s-Yju3p C20:0 MG analog complex embracing the substrate; the C20:0 substrate analog is shown as gray sticks; the protein is represented as surface (the colors indicate the electrostatic potential, blue = positive, red = negative). The position of Ile166 is indicated by an arrow and the position of the opening by a black rectangle. The position of the hole through which the carbon chain of the C20:0 MG analog chain leaves the enzyme is indicated by a black rectangle. In this orientation, Met159 is positioned above the paper plane and not visible due to clipping. The image was created using UCSF chimera [46]. yju3p gene according to the manufactures instructions using the following primers: forward primer: 5′-GGC GAA ATT TTC ACC AAG GGT AAG GAT CGA CAC TGG-3′; reverse: 5′-CCA GTG TCG ATC CTT ACC CTT GGT GAA AAT TTC GCC-3′. It should be noted here, that the variant used here also differs at amino acid position 264 (Gln264Arg) compared to the canonical sequence in data base entry UniProt-ID P28321. We proceeded with crystallization because i) this variant was described previously as active in hydrolyzing MGs [26], ii) the amino acid position is not conserved when comparing the sequence with hMGL or mouse MGL and, iii) the amino acid is placed in a surface exposed position of the core domain and thus is very unlikely to interfere with the fold of the protein. The resulting variant is called s-Yju3p throughout the remainder of the text.

Expression and purification
s-Yju3p was expressed and purified as described previously [27]. Expression and affinity chromatography purification for hMGL and Yju3p were done using the same protocol and buffers as for s-Yju3p with the exception that the lysis buffer contained 0.5% IGEPAL CA-630 and the wash and elution buffers 0.05% IGEPAL CA-630, respectively. bMGL was purified as described previously [22]. To compare the activities of the lipases, a buffer containing 0.05% IGEPAL was used for all proteins.

Monoglyceride hydrolase activity assay
Since MGLs are not regioselective (Fig. 1A) and the 1(3)-racemic monoglyceride (1(3)-rac MG) substrates are commercially available with different chain lengths and saturations, we used 1 (3)-rac MGs as substrates. The monoglyceride hydrolase activity assays for Yju3p, s-Yju3p, bMGL and hMGL [23] were done as previously described [26] with the exception that bovine serum albumin was exchanged for 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) with a final concentration of 5 mM and that substrate concentrations were 2 mM. The substrates used were all 1 (3)-rac MGs with the exceptions of 3-palmitoyl-sn-glycerol (which was not available as racemic mixture from the same commercial supplier) for the assays displayed in Fig 2.5. Crystallization and data collection: s-Yju3p in free form and in complex with a MG C20:0 analog Crystals of s-Yju3p in its free form were obtained from a protein solution at 14 mg/mL and a reservoir solution containing 0.1 M Bicine/Trizma base pH 8.7, 10% w/v PEG 20 000, 20% v/v PEG MME 550 and 0.03 M sodium nitrate, 0.03 M disodium hydrogen phosphate, and 0.03 M ammonium sulfate using microseeding [28] as described [27]. Diffraction data to a resolution of 2.4 Å were collected at the SLS (PSI, Villigen, Switzerland). Unfortunately, attempts to solve the structure with molecular replacement (using the program Phaser (29)) failed. Therefore structure determination was pursued with heavy atom methods.
In order to obtain a co-crystal structure of s-Yju3p bound to a C20:0 MG mimicking substrate analog, 50 μl protein solution (14.9 mg/ml) were mixed with 1 μl C20:0 solution (50 mg/ml in ethanol) and incubated for 1 h at 4°C. Excess C20:0 mimicking substrate analog and the resulting nitrophenol were then washed off by diluting the complex solution to 500 μl with dialysis buffer and re-concentrating it to 50 μl twice. This solution was then set up for crystallization using the same method and condition as for untreated s-Yju3p. Diffraction data were measured from a crystal resulting from these setups to a resolution of 2.5 Å at ID-29 at the ESRF (Grenoble, France).

Heavy metal soaking and MAD data collection
s-Yju3p crystals were harvested and transferred to an extra drop consisting of 1.5 μl of a 10 mM K 2 Pt(NO 2 ) 4 solution and 2.5 μl of the reservoir solution. The crystals were soaked for 6 days and then flash cooled and stored at liquid nitrogen temperatures. No back soaking of the crystals was performed. The crystals were taken to ID-23 at the ESRF (Grenoble, France). After an X-ray fluorescence scan, diffraction data were collected using three different wavelengths around the platinum absorption edge: 1.07155 Å (peak), 1.07182 Å (inflection) and 1.06878 Å (remote).

Data processing and structure refinement: s-Yju3p
The three data sets collected for crystals soaked with K 2 Pt(NO 2 ) 4 were indexed and integrated using XDS [30] and scaled using Scala [31]. These data sets were then uploaded to the Auto-Rickshaw server [32]. A native data set without heavy atom derivatives with a resolution of 2.5 Å measured at SLS (Villigen, Switzerland) was indexed and integrated using XDS [30], scaling and data reduction were performed using Aimless [33]. The resolution was cut to achieve reasonable CC 1/2 values. The R merge in the higher resolution shell was 140% which could be considered as very high but this was accepted since the CC 1/2 value was still 49%. The biggest fragment with the best electron density fit from the initial model produced by Auto-Rickshaw was used for molecular replacement in the native data set using PHASER [29]. The best MR solution was used to complete the structure with phenix.autobuild [34]. Structure refinement was carried out by alternating manual steps in Coot [35] and automated real-space, ADP, atomic xyz coordinate and occupancy refinement in phenix.refine. R free values were calculated from 5% of the reflections separated from the data set. Waters were added using the "update water" option in phenix.refine and questionable waters were identified via weak 2Fo-Fc map intensities and high B-factors and were manually deleted in Coot. The geometry of the structure was validated using MolProbity [36]. The structure could be refined to residual factors of 22.4% (R free ) and 18.8% (R work ). A more complete list of the statistics of data processing and refinement is shown in Table 1.
2.8. Data processing and structure refinement: s-Yju3p substrate analog complex Diffraction data from a crystal of the s-Yju3p C20:0 MG analog complex were indexed and integrated using iMosflm [37] and scaled using Scala [31]. 5% of the reflections were held back after data reduction in order to calculate R free values for evaluation of the refinement quality. The resolution was cut at 2.5 Å in order to achieve acceptable R merge values. The refinement was carried out using Coot [35] and phenix.refine [34]. The processing and refinement statistics are shown in Table 1. The model for the C20:0 MG analog was generated using MAESTRO (Maestro, version 9.3, Schrödinger, LLC, New York). A cif library containing the restraints for this model was prepared using phenix.elbow [38]. Placement of the ligand was done manually in Coot. The geometry around the phosphor atom was restrained to tetrahedral angles and the distance to the gamma oxygen of Ser123 was restrained to a maximum of 1.9 Å. The occupancy of the ligand was refined to 0.8. Waters were added to the structure using the "update waters" option in phenix.refine. Questionable waters were evaluated based on the 2Fo-Fc map and B-factors and were deleted manually in Coot. The geometry of the structure was evaluated using MolProbity [36]. The structure could be refined to an R free of 25.68%. 0.33% Ramachandran outlier remained. The average B-factors for the final structure were 44.8 ± 14.3 and 41.9 ± 8.5 for the protein and the ligand atoms, respectively. The atomic coordinates and structure factors (codes 4ZWN and 4ZXF) have been deposited in the Protein Data Bank (http://wwpdb.org/).

Molecular dynamics simulation
pKa values and protonation states of the titratable amino acids at pH 7 were calculated using TITRA employing the Tanford-Kirkwood sphere model [39]. The GROMACS 5.0.4 software package was used to perform single precision molecular dynamics simulations and equilibrations [40]. The structure was solvated with water inside a cubic box, minimized and stepwise (free water, free hydrogens, free side chains, free protein) equilibrated for 1 ns (NPT and NVT, respectively) followed by four unrestrained simulations of 100 ns with explicit solvent employing the OPLS all-atom force field and the SPC water model [41,42]. The backbone RMSD was monitored to ensure complete equilibration of the protein model. All simulations were carried out with a 2-fs time step and long-range electrostatic interactions were computed using the particle mesh Ewald (PME) method. All bonds in the system were constrained using the LINCS algorithm. The neighbor list search was updated every 5 steps within a 1.0 nm cut-off. Van der Waals interactions were computed with a cut-off of 1.4 nm. The isotropic Parrinello-Rahman protocol was used for pressure (1 bar), and the velocity-rescaling thermostat was used for temperature coupling. The components of the system were coupled at 300 K with a coupling constant of 0.1 ps. Periodic boundary conditions were applied in all three dimensions. Calculations were performed on 64-bit Linux 64core nodes of an AMD Istanbul cluster. Data analysis and image rendering were carried out with standard tools provided within GROMACS and VMD, respectively [40,43].

Engineering of s-Yju3p, a solubility enhanced variant of Yju3p
Previous reports identified Yju3p as a membrane-bound or lipid droplet associated protein [44]. Accordingly, purification and concentration of recombinant Yju3p required the presence of detergents in order to prevent aggregation (e.g. NP-40, Mega8) [26]. Therefore, we used a solubility enhancement strategy as described for hMGL to identify potential hydrophobic residues in the cap region which might be involved in aggregation [24]. Employing this strategy, hMGL carrying the variations Leu169Ser, Leu176Ser, and Lys36Ala had yielded crystals diffracting to 1.35 Å. Based on a sequence alignment, we assumed that amino acid Leu175 of Yju3p corresponds to Leu169 in hMGL. Indeed, introducing a Leu175Ser mutation proved successful and allowed purification, concentration and crystallization of the protein even in the absence of detergents. MG hydrolase assays verified that the soluble s-Yju3p variant is still active against MGs in the mmol/(hour*mg protein) range (Fig. 1A).

The 3D structure of s-Yju3p
s-Yju3p crystallized in space group P2 1 2 1 2 1 . The structure was determined using a combination of multi-wavelength anomalous dispersion (MAD) and molecular replacement using the fragments obtained during model building after MAD. We observe four molecules in the asymmetric unit. The PISA server [45] suggests a tetramer to be stable, yet we observed a peak corresponding to the size of monomeric protein in the size exclusion chromatogram. The overall structure of s-Yju3p harbors an α/β-hydrolase core and a cap containing four helices of different lengths (Fig. 1B, C). The core harbors the active site with the catalytic triad residues Ser123, Asp251 and His281 as proposed earlier based on sequence homology [26]. The spatial arrangement of these residues in our 3D structure confirms that they form a catalytic triad. His281 and Asp251 are spatially arranged so they can form the charge transfer relay involved in deprotonating Ser123 (Fig. 1D). Ser123 is in a position to act as the nucleophile attacking the partially positively charged carbonyl carbon of the substrate. Comparison with bMGL and hMGL structures suggests that the oxyanion hole is formed by the main chain amide groups of Met124 and Phe49. The cap is located around the catalytic center in a position that could facilitate substrate binding (Fig. 1B-D). In the N-terminal region of the cap, the short Helix 1 (Pro156-Asn161) and the longer Helix 2 (Thr164-Phe174) are just separated by Lys162 and Pro163, which induce a bent in the orientation of the helix axes. The cap then crosses the molecule back and forth with To calculate R free , 5% of the reflections were excluded from the refinement. R merge is defined as R merge = Σ hkl Σ i |I hkl,j − 〈I hkl 〉| / Σ hkl Σ j I hkl,j . R meas is defined as R meas = Σ hkl (n/n − 1) 1/2 Σ n j = 1 |I hkl,j − 〈I hkl 〉| / Σ hkl Σ j I hkl,j . R pim is defined as R pim Σ hkl (1/n − 1) 1/2 Σ n j = 1 |I hkl,j − 〈I hkl 〉| / Σ hkl Σ j I hkl,j . Data in parentheses correspond to the highest resolution shell. r.m.s.d., root mean square deviation. The CC 1/2 is the Pearson correlation coefficient between average intensities of two half data sets [49]. the short Helices 3 (Leu186-Ile189) and 4 (Lys193-Ser201) at the opposite end of the cap. Residues Arg179-Asp181 and Tyr209-Ser211 are arranged in an antiparallel manner and bring the forth and back leg of the cap in spatial proximity (Fig. 1C). Arg264, which is a glutamine in the native sequence, is positioned on a helix in the α/β-hydrolase core with its side chain exposed to the aqueous environment and does not show any side chain interaction with other residues. Residue 175, which is a serine in s-Yju3p as opposed to a leucine in the native enzyme, is located in the cap region in a loop near the end of Helix 2 (Fig. 1B). This spatial position corresponds nicely to the position of Leu169Ser near the end of the first helix in hMGL which served as template for our solubility enhancement design. Three small holes were observed in the surface of chain D, yet none of them seems big enough to fit the glycerol head group of an MG substrate or our MG-analog in the presence of a p-nitrophenyl group.

s-Yju3p tetrahedral intermediate analog complex
Crystals diffracting to 2.2 Å were obtained from the cocrystallization setup with a p-nitrophenyl ester of alkyl phosphonic acid mimicking a C20:0 MG analog (Fig. 1E). The crystals were isomorphous to the crystals of the free form of s-Yju3p. Additional electron density could be observed near the active site in chains A, B and D; though it was not unambiguous enough to be interpreted as the C20:0 MG analog. In chain C however, we could observe electron density resembling the shape of our ligand (Fig. 1F). In some regions of the ligand carbon chain, the electron density was poorly defined. This might be caused by a low occupancy of the ligand (e.g. due to low water solubility) or by the flexibility of the carbon chain. The substrate analog oxygen that is equivalent to the carbonyl oxygen of the substrate is stabilized by hydrogen bonds from the backbone nitrogen of the oxyanion-hole residues Met124 and Phe49 (Fig. 1E, G). The substrate entrance channel is mostly outlined by hydrophobic side chains (Leu151, Leu154, Thr158, Met159, Lys162, Gln165 and Thr182 from the cap and Phe49, Ile215, Phe218, Met219, and Ile253 from the core) (Fig. 1H, I). This hydrophobicity might be necessary to accommodate the hydrophobic carbon chain of the MG substrate in the binding pocket. Met159 and Ile166 of the cap line the outside of the entrance channel and form hydrophobic interactions with the carbon chain of the substrate analog partially sticking out into the aqueous environment (Figs. 1H, I, 2) [46]. Unexpectedly, carbon-atoms C12 to C18 (corresponding to C14 to C20 in an actual MG substrate) stick out of the hydrophobic substrate binding channel between the cap and the core regions of the lipase (Fig. 2B). This opening is quite distinct from the bMGL structures in the open conformation and from the substrate entrance tunnels observed in complex structures of bMGL with different ligands [22,23,25]. The physiological relevance of this wedge-like opening observed in presence of the covalently bound ligand still needs to be established. Crystals of the free form and the tetrahedral intermediate mimicking complex were isomorphous and the packing was identical. Therefore, we conclude that the observed conformational changes are not caused by crystal contacts in the cap region but represent an intrinsic conformational rearrangement.

Conformational flexibility in the cap of s-Yju3p
Crystal structures of hMGL and bMGL revealed conformational flexibility of the cap region. The experimental structures of free s-Yju3p and of the lipase in complex with the covalently bound inhibitor (Fig. 2A,  B) also pointed towards conformational flexibility of the s-Yju3p cap. Three of the four chains in the free form structure showed a closed surface with only one small hole in chain D ( Fig. 2A). If this observed closed cap conformation were to be rigid, such an enzyme could not be active since it does not provide a possibility for the substrate to reach the active site. Thus we hypothesized that additional cap conformations of s-Yju3p have to exist. Indeed, slight conformational changes indicating breathing movements of the lipase could be observed in one chain of free s-Yju3p. The structure of s-Yju3p mimicking the tetrahedral reaction intermediate in complex with the C20:0 MG substrate analog provided further proof for flexibility of the cap (Fig. 2B, C). In chains A, B and D, the cap region adopted the same conformation as in the free form structure. In chain C conformational flexibility could be observed especially in Helix 2 of the cap. Residues Thr164-Ile167 show marked displacement of the Cα atoms compared to the position observed in the free form (Cα-Cα distances of 2.8, 3.0, 2.4, and 2.9 Å, respectively) while the rmsd value of the cap region encompassing residues Pro156-Phe174 is 1.3 Å (Fig. 2C). In this region, a completely connected electron-density for the protein chain of free s-Yju3p can be observed which is clearly different in the s-Yju3p complex structure (Fig. S1). Upon this conformational movement, the distance between the side chains of Ile166 (located in Helix 2 of the cap) and Met159 (located in Helix 1 of the cap) is increased from 3.9 Å to 7.4 Å (Fig. 2C-E). In the complex structure, this hole is occupied by the carbon chain of the C20:0 MG substrate analog (Fig. 2B, F). The published ligand bound structures of hMGL do not contain a natural MG substrate or a substrate analog. Nevertheless, the structure of hMGL in complex with a molecule termed 'compound 1' also showed conformational rearrangements of the cap (PDB 3PE6, (24)). Different cap conformations were observed for bMGL which appeared to arise from stochastic fluctuations or enzyme breathing rather than being ligand induced [22,23,25].
In analogy, we assume stochastic fluctuations in the cap of s-Yju3p could facilitate entry of the substrate to the active site deeply buried within the protein. In order to get first clues on such a predicted open conformation of s-Yju3p, four 100 ns molecular dynamics simulations of the protein solvated in explicit water starting from the closed conformation were carried out. In all four simulations, the cap region (from residue Ile153 to Phe212) showed a significantly higher flexibility than the protein core visible in the RMSF per residue plot (Fig. 3A). In two simulation runs the protein stayed in a rather closed state, whereas a significant opening leading to an open conformation could be observed in two different simulations. The overall back bone rmsd reached a plateau around 2.0 Å, the rmsd for the α/β-hydrolase core only stayed below 1.8 Å. The cap region however, reached a maximum rmsd of 5.1 Å after approximately 22 ns, before the protein was again morphing to a closed conformation (Fig. 3B). In contrast to an unfolding of the cap region during the simulation, the observed increase and subsequent decrease of the cap rmsd (relative to the equilibrated closed conformation) is a strong indication for a directed cap movement from the closed to the open state (and back). In this open state, the active site entrance was visible and the active site Ser123 was accessible from the solvent (Fig. 3C, D). It should be noted here, that the position of the opening observed in the MD simulations is different to the one observed in our crystal structure with the C20:0 MG substrate analog exiting the substrate binding pocket between the cap and the core domain (see also Fig. 5A, B). In conclusion, the MD simulations strongly suggest that Yju3p is capable of showing an open conformation similar to the ones in hMGL and bMGL.

Cavity shape and substrate binding versus substrate selectivity
Comparison of the cap regions of s-Yju3p and the other structurally characterized MGLs (hMGL and bMGL) shows high conservation of the overall S-shape of the cap region across different species. Despite the differences in secondary structure elements, the overall course of the protein backbones in these three different MGLs is very similar (Fig. 4A-C) [12,[21][22][23][24]. It could even be argued, that the interrupted cap helix in Yju3p represents an evolutionary necessary functional intermediate when comparing MGLs from Bacillus, S. cerevisae and Homo sapiens. The introduction of the amphipathic cap helix apparently changes the intracellular localization of the MGLs from a water-soluble, cytosolic to a membrane-and LD-associated lipase. Consequently, the   4ZXF). B, open conformation of s-Yju3p resulting from the MD simulation (cap region in red and the α/β-hydrolase core in orange). C, bMGL (PDB code 4KE8, cap region in pink, α/β-hydrolase core in purple). D, hMGL (PDB code 3HJU, cap region in light blue and α/β-hydrolase core in blue). E-G, selectivity of different MGLs for saturated and unsaturated MG substrates. Saturated substrates are to the left of the dashed line and unsaturated to the right. MG hydrolase activity assay with MG substrates containing different fatty acids for Yju3p (E) bMGL (F), and hMGL (G). H, substrate binding pocket of s-Yju3p (C20:0 analog complex) (blue surface) with the bound C20:0 ligand as blue sticks. I, substrate binding pocket (yellow surface) of bMGL (PDB code 4KE8) and the bound C16:0 analog as yellow sticks. J, substrate binding pocket of hMGL (PDB code 4UUQ) as green surface, the bound inhibitor SAR127303 is displayed in stick representation. recruitment of different substrates from different cellular compartments became also possible. s-Yju3p seems to have a substrate binding mode distinct from bMGL and hMGL. In s-Yju3p, the carbon chain of the substrate analog appears to be able to act as a wedge between the α/β-hydrolase core and the cap region embedded by Leu166 and Met159 (Fig. 5A, red  circle). Thus, the end of the alkyl-chain in the complex structure of s-Yju3p is observed below cap Helices 1 and 2. The path from the active site serine to the opening of the binding pocket of hMGL and bMGL follow a straight line and suggest positioning of the substrate above the corresponding helices (Fig. 5A-D). The transition-state intermediate of s-Yju3p was crystallized with a ligand mimicking a C20:0 alkyl chain. One might assume that this conformation is linked to the preference of this long aliphatic chain to be placed in a bent form in the hydrophobic substrate binding pocket. A more direct, straight path would place larger parts of the alkyl chain into the hydrophilic environment outside the binding pocket. Interestingly, MD simulations also suggest a possible open conformation with an opening between the first stretch (Helices 1 and 2) and the second stretch (Helix 3) in the cap region. This is similar to the open conformations experimentally observed in bMGL and hMGL with the substrate entrance channel within the first turn of the S-shape within the cap region ( Fig. 5C-D, red circles).
In this study, we also determined specific activities of Yju3p, hMGL and bMGL with saturated and mono-unsaturated MGs of different chain lengths. MG hydrolase activity assays showed that Yju3p activity gradually decreased with increasing chain length using saturated MGs as substrate. Actually, we could not detect activity using 18:0 MG as substrate (Fig. 5E). This trend was also observed using saturated MGs for hMGL and bMGL (Fig. 5F, G), although these enzymes exhibited detectable activity against 18:0 MG. One might speculate that the ability to fit and enclose the alkyl chain within the binding pocket plays a role [22,23]. Notably, Yju3p showed highest activity in the presence of monounsaturated MGs C16:1 and C18:1. The different shapes of the cavities might also influence substrate specificities in different organisms ( Fig. 5H-J). The substrate preference of Yju3p nicely correlates to the most abundant fatty acids within yeast [47]. This preference of Yju3p for unsaturated long-chain fatty acids as substrates might be facilitated by the bend observed in the Yju3p cavity (Fig. 5H). A saturated fatty acid should also be able to form a bend in the carbon chain, yet the preformed bend in an unsaturated fatty acid might preferably select for accommodation of a long aliphatic chain inside the protein. hMGL and bMGL exhibit highest activity using C12:0 as substrate. MGs with short alkyl chain length are toxic to bacteria which can explain the necessity of bacterial MGL to be extremely efficient in the turn-over of these substrates. bMGL does not differentiate between C16:0 and C16:1 but shows different activity for C18:0 and C18:1 MG substrates. This might be explained by the fact that longer carbon chains need to bend to be accommodated in the substrate binding pockets while a unsaturated substrate is already pre-bent and easier to accommodate [22] ( Fig. 5H-J). hMGL shows less selectivity between C18:0 and C18:1 than Yju3p and bMGL. This could be caused by the broad substrate entrance channel of hMGL which applies less restriction on the rotamers of the fatty acid moiety that could be fitted inside the protein. It cannot be ruled out completely, that the physico-chemical properties of the substrates and concomitant different modes of substrate presentation in the used aqueous micelle systems also influence the measured absolute activities in the assay.
Although the overall shape of the cap in MGLs seems to be conserved across species, we could show that there are differences in the details of the cap structures, substrate binding pockets, and in the substrate preference of Yju3p compared to hMGL and bMGL. While the structural differences observed fit the differences in substrate preference and provide a basis for their explanation, further biochemical data including enzyme kinetics, on-and off-rates, etc. are needed in order to get a clear understanding of the mechanisms underlying the substrate selectivity in MGLs.