FEBS Letters 584 (2010) 4599–4605

journal homepage: www.FEBSLetters .org

Molecular dynamics simulations of human and dog gastric lipases: Insightsinto domain movements

Anitha Selvan, Chandrabhan Seniya, Srinivas Niranj Chandrasekaran, Nithyanand Siddharth,Sharmila Anishetty ⇑, Gautam Pennathur ⇑Centre for Biotechnology, Anna University Chennai, Chennai 600 025, India

a r t i c l e i n f o a b s t r a c t

Article history:Received 18 August 2010Revised 8 October 2010Accepted 8 October 2010Available online 20 October 2010

Edited by Robert B. Russell

Keywords:Gastric lipaseAcidic pHMolecular dynamicsLidMobile domainGROMACS

0014-5793/$36.00 � 2010 Federation of European Biodoi:10.1016/j.febslet.2010.10.021

⇑ Corresponding authors. Fax: +91 44 22350299.E-mail addresses: s_anishetty@yahoo.co.uk (S. An

edu (G. Pennathur).

Mammalian gastric lipases are stable and active under acidic conditions and also in the duodenallumen. There has been considerable interest in acid stable lipases owing to their potential applica-tion in the treatment of pancreatic exocrine insufficiency. In order to gain insights into the domainmovements of these enzymes, molecular dynamics simulations of human gastric lipase was per-formed at an acidic pH and under neutral conditions. For comparative studies, simulation of doggastric lipase was also performed at an acidic pH. Analyses show, that in addition to the lid region,there is another region of high mobility in these lipases. The potential role of this novel region isdiscussed.� 2010 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.

1. Introduction

Gastric lipase, pancreatic lipase and lysosomal acid lipases aremammalian fat digesting enzymes. In humans, lipolysis is cata-lyzed by pre-duodenal and pancreatic lipases. The process beginsin the stomach with the hydrolysis of short and long chain triacyl-glycerols catalyzed by the pre-duodenal gastric lipase [1]. Partialhydrolysis by gastric lipase promotes subsequent hydrolysis bypancreatic lipase in the duodenum. Gastric lipase is active and sta-ble at an acidic pH whereas pancreatic lipase loses its activity irre-versibly at an acidic pH [2]. Human gastric lipase remains activeeven in the duodenal lumen and functions in synergy with pancre-atic lipase. Several pathological conditions like cystic fibrosis, acutepancreatitis, alcoholism, and premature birth have been associatedwith exocrine pancreatic lipase insufficiency [3]. Enzyme replace-ment therapy in the form of porcine pancreatic lipase preparationshas been used in the treatment of some of these conditions [4].However, enzymatic degradation and acid inactivation is a causeof concern in such therapies. Because of its acid stability, gastriclipase is a better choice for replacement. Insights into factorsgoverning mechanism of action and acid stability of human gastric

chemical Societies. Published by E

ishetty), pgautam@annauniv.

lipase may help in the design of better enzyme replacementtherapies.

The crystal structures of closed form of human gastric lipase [5]and open form of dog gastric lipase complexed with undecyl butyl(C11Y4) phosphonate inhibitor [6] have been elucidated. Humangastric lipase is secreted by the chief cells in the fundic region ofthe stomach [7]. It has an apparent molecular mass of 50 kDa withfour glycosylation sites (Asn 15, Asn 80, Asn 252 and Asn 308) [8].The 379 residues enzyme has a core domain [5] spanning residues9–183 and 309–379 and a cap domain spanning residues 184–308.The catalytic triad Ser 153, Asp 324 and His 353 form the activesite. The NH groups of Leu 67 and Gln 154 contribute to the oxyan-ion hole. Residues 215–244 form the putative lid [5]. Dog gastriclipase is a glycoprotein with 377 residues and a mass of 49 kDa.Its core domain is located between residues 1–183 and 309–377.The lid region spans residues 212–251. The lid along with residues184–211 and 252–308 is similar to the cap domain described in theclosed form of human gastric lipase. Ser 153, His 353 and Asp 324form the active site while Leu 67 and Gln 154 contribute to theoxyanion hole [6].

Dog and human gastric lipases share 85.7% sequence similarity.A characteristic feature of most lipases is the interfacial activationat the lipid–water interface. It involves a conformational change inthe lid from the closed form to open form allowing substrate accessto the active site [9]. Literature reports suggest that the opening of

lsevier B.V. All rights reserved.

4600 A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605

the lid in gastric lipases is not a simple rigid body movement butmay involve a more complex mechanism [3].

Molecular dynamics simulations have proved useful in under-standing the physical basis of structure and function of biomole-cules [10]. In this study molecular dynamics simulations ofhuman gastric lipase (at an acidic pH and under neutral conditions)and dog gastric lipase (at an acidic pH) were performed with anobjective of gaining insights into the domain movements of theseenzymes. Sequence analysis of acid lipases and pancreatic lipasewas also done to detect sequence specific clues if any for survivalin an acidic medium.

2. Materials and methods

2.1. Molecular dynamics simulations

The crystal structures of human gastric lipase bearing PDB ID1HLG and dog gastric lipase PDB ID 1K8Q were obtained fromthe Protein Data Bank [11]. The inhibitor in 1K8Q was removed.GROMACS [12,13] software was used for molecular dynamics sim-ulation. Chain A coordinates were used in the simulation. In humangastric lipase, the first eight residues and a short span 54–56 weremissing in the crystal structure. All the Asp, Glu and His residueswere protonated to mimic an acidic environment. The proteinwas solvated with an explicit solvent: SPC water, in a cubic boxwhich left 0.2 nm space around the solute. In the acidic environ-ment, for human gastric lipase, counter ions in the form of 19Cl� ions were added to neutralize the system while 25 Cl� ionswere added in the case of dog gastric lipase. Simulation withoutthe protonation was performed for human gastric lipase to mimicneutral conditions. In this case, 1 Na+ counter ion was added.Steepest descents method was used for energy minimization. Thesystem was weakly coupled to an external bath using Berendsen’smethod. The reference temperature was fixed at 300 K. All bondswere constrained with LINCS [14,15]. G43a1 force field in GRO-MACS was used for energy calculations. During the simulation, gridtype neighbour searching was done and long range electrostaticswas handled using PME [16]. In all the three simulations, the en-ergy minimized structures were subjected to a position restrainedmolecular dynamics simulation for 50 ps. This was followed by afull molecular dynamics simulation. The time step used was 2 fs.Coordinates were written every 0.5 ps to the trajectory. GROMACSbuilt in tools were used to compute Root Mean Square Deviation(RMSD), residue wise fluctuations: Root Mean Square Fluctuations(RMSF) and radius of gyration. All graphs were visualized usingXmgrace (http://www.plasma-gate.weizmann.ac.il/Grace/). Theaccessible molecular surface was computed using WHAT IF [17]software. The images of the superimposed structures wereobtained using VMD [18] software. Residues with RMSF of0.2 nm and above have been taken into consideration to map fluc-tuations to specific regions/domains of the enzyme. B factor anal-ysis was done by extracting B factor values for Ca atoms fromthe crystal structure of human gastric lipase.

2.2. Sequence analysis

Fasta sequences of human gastric lipase, human lysosomal acidlipase, dog gastric lipase, rat lingual lipase and human pancreaticlipase with UniProt IDS P07098, P38571, P80035, P04634,P16233, respectively, were retrieved from UniprotKB [19]. Signalpeptide sequences were removed and multiple sequence align-ment was performed using Clustal_X [20]. Kyte and Doolittlehydropathy plots were generated for human and dog gastriclipases using Protscale tool at the ExPASy proteomics server[21].

3. Results

3.1. Molecular dynamics simulation

Molecular dynamics simulation of human gastric lipase wasperformed at an acidic pH. For comparative studies a simulationof human gastric lipase was also performed without the proton-ation mimicking the acidic pH conditions. The latter simulationwill hereafter be referred to as simulation under neutral condi-tions. The RMSD of the simulated structure obtained after back-bone least squares fit with the reference structure is shown inFig. 1a (acidic pH) and Fig. 1c (neutral conditions). At an acidicpH, the RMSD steadily increased to about 0.34 nm. The enzymestabilized around 3.5 ns and remained stable till the end of thesimulation time frame 6.5 ns as is evident from the leveling off inthe RMSD graph. Under neutral conditions, the RMSD increasedto about 0.45 nm. The enzyme stabilized around 8 ns and remainedstable till the end of the simulation 11 ns. Fig. 1e shows the RMSDfrom the dog gastric lipase simulation at an acidic pH. The RMSDprogressively increased to 0.42 nm. The enzyme stabilized around5 ns and remained stable till the end of the simulation which is9 ns. In all the simulations described here, the structure obtainedat the end of the simulation time frame (6.5 and 11 ns for humangastric lipase at acidic pH and neutral conditions, respectively,9 ns for dog gastric lipase at an acidic pH) was taken as the stabi-lized form and further analyzed.

3.2. Lid mobility of human gastric lipase

Molecular dynamics simulations have been successfully em-ployed in the past to find domain movements in proteins. Analysisof residues showing fluctuations during a molecular simulationhelps in detecting such motions in a protein [22]. Residue wisefluctuations are shown as the RMSF graphs in Fig. 1b (human gas-tric lipase acidic pH) and Fig. 1d (human gastric lipase neutral con-ditions). Taking near contiguous residues into consideration,fluctuations were predominantly observed in two regions amongstothers. The crystal structure of recombinant human gastric lipasedescribes residues 215–244 to be a putative lid. Our analysis atan acidic pH revealed that a shorter stretch spanning residues215–232 shows maximum mobility in the lid region. Residues230, 219 and 229 showed maximum fluctuations in this region fol-lowed by residues 220, 231, 226 and 225. A part of this stretch221–231 corresponds to helix 12 in the crystal structure. Theremaining part of the putative lid residues 233–244 (barring resi-due 234 which had close to 0.2 nm fluctuations) showed low fluc-tuations. In an earlier study, Lohse et al. [23] had deleted the regionspanning residues 228–235 in the putative lid and had demon-strated that this region was important for catalysis. Our study indi-cates that the region of maximum mobility in the lid region isconfined to residues 215–234.

Studies from the simulation under neutral conditions showedsimilar results. Residues 215–233 showed maximum mobility inthe lid region. Residue 216 showed maximum fluctuations in thisregion followed by residues 229, 230 and 217.

3.3. Lid mobility of dog gastric lipase

The residue wise fluctuation of dog gastric lipase is shown inFig. 1f. In the lid region (residues 212–251), maximum mobilitywas found in the region 216–235. However, five residues in thisspan had fluctuations below 0.2 nm. The terminal portion of thelid, 248–251 also showed good mobility. On comparison of thelid regions of dog and human gastric lipases, we find that the re-gion with maximum mobility overlaps.

Fig. 1. RMSD and RMSF graphs. (a) and (b) RMSD and RMSF, respectively, of human gastric lipase at an acidic pH. (c) and (d) RMSD and RMSF, respectively, of human gastriclipase under neutral conditions. (e) and (f) RMSD and RMSF, respectively, of dog gastric lipase at an acidic pH.

A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605 4601

3.4. Novel region of high mobility in human gastric lipase

Interestingly, in human gastric lipase the maximum fluctua-tions were not confined to the lid region. Residues 291–302 span-ning helix 18 in the crystal structure showed the highest mobilityat an acidic pH. This region is also part of the cap domain describedin the crystal structure. Amongst the top 13 residues showing veryhigh fluctuations (0.29–0.48 nm), six were from this region. Themaximum fluctuation of 0.48 nm was shown by Val 293. The totalaccessibility of this residue becomes low. At the end of the simula-tion, the helix looked disordered and appears to have moved in-wards towards the active site. Similarly, even in the moleculardynamics simulation under neutral conditions we observe a move-ment of this helix and amongst the residues showing very highfluctuations (greater than 0.30 nm) six were from this region. The

fact that this novel region is more mobile than the lid region couldnot have been ascertained from the B factor analysis of the PDBstructure, since the B factors seen in the novel region is less thanin the lid region in the crystal structure. Superimposed VMDimages showing the movement of the region with respect to thecrystal structure are shown in Fig. 2d (acidic pH) and Fig. 2e (neu-tral conditions).

3.5. Novel region of high mobility in dog gastric lipase

Intrigued by the presence of this highly mobile region, weperformed molecular dynamics simulation of dog gastric lipaseat an acidic pH to see if we would observe a similar phenome-non. Dog gastric lipase shares 85.7% similarity with humangastric lipase and of the 12 residues in the region in question:

Fig. 2. Radius of gyration and superimposed structures. Radius of gyration profiles: (a) human gastric lipase at an acidic pH; (b) human gastric lipase under neutralconditions; (c) dog gastric lipase at an acidic pH. Superimposed structures: structures obtained at the end of the simulations superimposed with their corresponding crystalstructures: (d) human gastric lipase at an acidic pH, (e) human gastric lipase under neutral conditions; (f) dog gastric lipase at an acidic pH. The lid region is marked as 1 andthe novel region of high mobility is marked as 2 in the superimposed structures.

4602 A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605

291–302 of human gastric lipase, 10 residues are identical in doggastric lipase. Analysis of RMSF graph (Fig. 1f) of dog gastric li-pase shows that this region is indeed more mobile than the lidregion.

3.6. Other regions of mobility in human gastric lipase

In human gastric lipase, apart from the cap domain, certainother regions from the core domain also showed fluctuations of

Fig. 3. Hydropathy plots. (a) and (b) Hydropathy plots of human and dog gastric lipases, respectively.

A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605 4603

0.2 nm and above. One notable movement is the turn between he-lix 4 and 5 (residues 110–115) and part of helix 15 and 16 (253–261). Some of the residues in these regions are polar and this couldhave contributed to the movement. Under neutral conditions partof helix 11 (201–206), and part of helix 15 and 16 (248–258 bar-ring residues 251 and 252) showed mobility. At an acidic pH,amongst the active site residues, His 353 becomes more accessible,Ser 153 and Asp 324 become marginally more accessible, while un-der neutral conditions, Ser 153 and Asp 324 become less accessibleand His 353 becomes more accessible.

3.7. Other regions of mobility in dog gastric lipase

Apart from the lid region and the novel region, one other regionshowed maximum mobility in dog gastric lipase simulation at anacidic pH. This corresponded to residues 48–58. This region is richin basic residues.

3.8. Sequence alignment

Multiple sequence alignment of acid lipases: human gastric li-pase, human lysosomal acid lipase, dog gastric lipase and rat lin-

gual lipase with pancreatic lipase revealed a striking deletion(residue range 232–269 pancreatic lipase numbering) as shownin Fig. 4. This deletion overlaps with the entire lid region of pancre-atic lipase and was prominent only when all the acid lipases wereincluded in the alignment. Other deletions were of a shorter length.

4. Discussion

A characteristic feature of most lipases is the interfacial activa-tion at a lipid–water interface. It involves a conformational changein the lid from the closed form to open form allowing substrate ac-cess to the active site. Human gastric lipase used in this simulationwas in a closed form. The simulation was performed in an aqueousmedium with Asp, His and Glu residues protonated to mimic anacidic environment. Lipases are known to adopt a closed form inan aqueous medium. Although the structure used in the simulationis in a closed form, the radius of gyration graphs [Fig. 2a (acidic pH)and Fig. 2b (neutral condition)] show a decrease implying a morecompact structure after the simulation.

The starting structure in the simulation of dog gastric lipase wasthat of an open form of the enzyme. The lipase was protonated tomimic an acidic environment. On analysis, we find that the radius

Fig. 4. Multiple sequence alignment. Multiple sequence alignment of human gastric lipase, human lysosomal acid lipase, dog gastric lipase, rat lingual lipase and humanpancreatic lipase.

4604 A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605

of gyration reduced (Fig. 2c) confirming a compact structure. Sincethe simulation was in an aqueous medium, the open form wouldhave reverted to a more compact form and hence the reductionin radius of gyration. The solvent accessibility of the lid region de-creased in all the three simulations. Though both the enzymesmoved towards a more compact form, as expected some of themovements seen in the open form of dog gastric lipase were differ-ent from that of human gastric lipase.

The domain movements observed during the course of thesesimulations have not only mapped to the putative lid region pro-posed earlier but has also presented a novel region. We hypothe-size that this region could have an impact one or more of thefollowing: substrate access, water accessibility and acid stability.Given that some lipases have been shown to have a second lid

[24], the role of this helix in providing substrate accessibility tothe active site remains to be ascertained experimentally. Anotherpoint worth noting is water accessibility to the active site. Thecomparison of open form of dog gastric lipase with the closed formof human gastric lipase has provided invaluable insights [25]. Theopening of the lid in gastric lipases is not a simple rigid body move-ment but involves a more complex mechanism where the lid reor-ganizes upon opening [6]. It has also been said that when thesubstrate gains access to the active site, it completely blocks it.For hydrolysis the active site requires water and water accessmay be through lateral channels. There are a large number of cav-ities on the water accessible surface of gastric lipases [25]. Onesuch cavity with the largest volume is lined by residues 274–278and 284–287 in human gastric lipase. The authors postulate that

A. Selvan et al. / FEBS Letters 584 (2010) 4599–4605 4605

this cavity may make it possible for water to reach the active siteSerine, provided side chain displacements occur in the residuesnearby [3]. The novel region of high mobility in our study is closeto this channel and spans residues 291–302 in human and dog gas-tric lipases. We speculate that movement in this region may have arole in creating a channel for water entry. This region also happensto be the most hydrophilic region as seen in the hydropathy plotsfor human gastric and dog gastric lipases (Fig. 3a and b), respec-tively. It is also rich in charged and basic residues. This could havea potential role in acid stability. While simulations may not be con-clusive, they definitely provide good experimental leads. The roleof this region therefore remains to be validated experimentally.Gastric lipases are active not only under acidic conditions but alsoin the duodenum where they work in synergy with pancreatic li-pase. In our simulation studies also gastric lipases were found tobe stable both at an acidic pH (similar to the conditions in thestomach) and under neutral conditions (as seen in the duodenum).However, the time taken for the enzyme to stabilize was muchhigher under neutral conditions (8 ns) as opposed to time takenunder acidic pH conditions (3.5 ns). This is in concurrence withthe fact that the optimal activity of gastric lipases is seen at anacidic pH.

Analysis at the sequence level also revealed some interestingfeatures. Although pancreatic lipases do not show high similarityto acid lipases, we chose to align them since gastric lipase is knownto exhibit activity even in the duodenum. The deletion from acidlipases of a region corresponding to the lid region of pancreatic li-pase is intriguing. Given that this deletion is present in all the acidlipases, it is tempting to speculate that it confers some advantagefor survival in an acidic medium. To our knowledge this study isthe first attempt that addresses domain movements of gastric li-pases using molecular dynamics simulations. We hope that the in-sights gained in this study will help provide leads for design ofbetter and acid stable formulations for treatment of conditions inexocrine pancreatic insufficiency.

Acknowledgements

The authors thank BTIS, Department of Biotechnology for com-putational facilities. S.A. and G.P. thank DBT, Government of Indiafor support (BT/01/COE/07/01).

References

[1] Gargouri, Y., Pieroni, G., Rivière, C., Sauniere, J.F., Lowe, P.A., Sarda, L. andVerger, R. (1986) Kinetic assay of human gastric lipase on short- and long-chain triacylglycerol emulsions. Gastroenterology 91, 919–925.

[2] Gargouri, Y., Moreau, H., Pieroni, G. and Verger, R. (1988) Human gastriclipase: a sulfhydryl enzyme. J. Biol. Chem. 263, 2159–2162.

[3] Miled, N., Canaan, S., Dupuis, L., Roussel, A., Rivière, M., Carrière, F., de Car, A.,Cambillau, C. and Verger, R. (2000) Digestive lipases: from three-dimensionalstructure to physiology. Biochimie 82, 973–986.

[4] Lankisch, P.G. (1993) Enzyme treatment of exocrine pancreatic insufficiency inchronic pancreatitis. Digestion 54, 21–29.

[5] Roussel, A., Canaan, S., Egloff, M.P., Riviere, M., Dupuis, L., Verger, R. andCambillau, C. (1999) Crystal structure of human gastric lipase and model oflysosomal acid lipase, two lipolytic enzymes of medical interest. J. Biol. Chem.274, 16995–17002.

[6] Roussel, A., Miled, N., Berti-Dupuis, L., Riviere, M., Spinelli, S., Berna, P., Gruber,V., Verger, R. and Cambillau, C. (2002) Crystal structure of the open form of doggastric lipase in complex with a phosphonate inhibitor. J. Biol. Chem. 277,2266–2274.

[7] Moreau, H., Bernadac, A., Gargouri, Y., Benkouka, F., Laugier, R. and Verger, R.(1989) Immunocytolocalization of human gastric lipase in chief cells of thefundic mucosa. Histochemistry 91, 419–423.

[8] Bodmer, M.W., Angal, S., Yarranton, G.T., Harris, T.J., Lyons, A., King, D.J.,Pieroni, G., Riviere, C., Verger, R. and Lowe, P.A. (1987) Molecular cloning ofhuman gastric lipase and expression of the enzyme in yeast. Biochim. Biophys.Acta 909, 237–244.

[9] Verger, Robert (1997) Interfacial activation’ of lipases: facts and artifacts.Trends Biotechnol. 15, 32–38.

[10] Karplus, M. and McCammon Andrew, J. (2002) Molecular dynamicssimulations of biomolecules. Nat. Struct. Biol. 9, 646–652.

[11] Berman, H.M., Westbrook, J., Feng, Z., Gilland, G., Bhat, T.N., Weissig, H.,Shindyalov, I.N. and Bourne, P.E. (2000) The Protein Data Bank. Nucl. Acids Res.28, 235–242.

[12] Lindahl, E., Hess, B. and van der Spoel, D. (2001) GROMACS 3.0: a package formolecular simulation and trajectory analysis. J. Mol. Model. 7, 306–317.

[13] Berendsen, H.J.C., van der Spoel, D. and van Drunen, R. (1995) GROMACS: amessage-passing parallel molecular dynamics implementation. Comput. Phys.Commun. 91, 43–56.

[14] Miyamoto, S. and Kollman, P.A. (1992) SETTLE: an analytical version of theSHAKE and RATTLE algorithms for rigid water models. J. Comput. Chem. 13,952–962.

[15] Hess, B., Bekker, H., Berendsen, H.J.C. and Fraaije, J.G.E.M. (1997) LINCS: alinear constraint solver for molecular simulations. J. Comput. Chem. 16, 273–284.

[16] Essman, U., Perela, L., Berkowitz, M.L., Darden, T., Lee, H. and Pedersen, L.G.(1995) A smooth particle mesh eswald method. J. Chem. Phys. 103, 8577–8592.

[17] Vriend, G. (1990) WHAT IF: a molecular modeling and drug design program. J.Mol. Graph. 8, 52–56.

[18] Humphrey, W., Dalke, A. and Schulten, K. (1996) VMD: visual moleculardynamics. J. Mol. Graph. 14, 33–38.

[19] The Universal Protein Resource (UniProt) consortium (2008) Nucl. Acids Res.36, D190–D195.

[20] Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F. and Higgins, D.J.(1997) The Clustal_X windows interface: flexible strategies for multiplesequence alignment aided by quality analysis tools. Nucl. Acids Res. 25, 4876–4882.

[21] Gasteiger, E., Hoogland, C., Gattiker, A., Duvaud, S., Wilkins, M.R., Appel, R.D.and Bairoch, A. (2005) Protein identification and analysis tools on the ExPASyserver in: The Proteomics Protocols Handbook (Walker, John M., Ed.), pp. 571–607, Humana Press.

[22] Novak, William, Wang, Hongming and Krilov, Goran (2009) Role of proteinflexibility in the design of Bcl-XL targeting agents: insight from moleculardynamics. J. Comput. Aided Mol. Des. 23, 49–61.

[23] Lohse, P., Lohse, P., Chahrokh-Zadeh, S. and Seidel, D. (1997) Human lysosomalacid lipase/cholesteryl ester hydrolase and human gastric lipase: site-directedmutagenesis of Cys227 and Cys236 results in substrate-dependent reductionof enzymatic activity. J. Lipid Res. 38, 1896–1905.

[24] Cherukuvada, S.L., Seshasayee, A.S.N., Raghunathan, K., Anishetty, S. andPennathur, G. (2005) Evidence of a double-lid movement in Pseudomonasaeruginosa lipase: insights from molecular dynamics simulations. PloSComput. Biol. 1 (3), e28.

[25] Miled, Nabil, Roussel, Alain, Bussetta, Cécile, Berti-Dupuis, Liliane, Rivière,Mireille, Buono, Gérard, Verger, Robert, Cambillau, Christian and Canaan,Stéphane (2003) Inhibition of dog and human gastric lipases by enantiomericphosphonate inhibitors: a structure–activity study. Biochemistry 42, 11587–11593.