Reflection on design and testing of pancreatic alpha-amylase inhibitors: an in silico comparison between rat and rabbit enzyme models
© Khalil-Moghaddam et al; licensee BioMed Central Ltd. 2012
Received: 25 September 2012
Accepted: 13 November 2012
Published: 20 November 2012
Inhibitors of pancreatic alpha-amylase are potential drugs to treat diabetes and obesity. In order to find compounds that would be effective amylase inhibitors, in vitro and in vivo models are usually used. The accuracy of models is limited, but these tools are nonetheless valuable. In vitro models could be used in large screenings involving thousands of chemicals that are tested to find potential lead compounds. In vivo models are still used as preliminary mean of testing compounds behavior in the whole organism. In the case of alpha-amylase inhibitors, both rats and rabbits could be chosen as in vivo models. The question was which animal could present more accuracy with regard to its pancreatic alpha-amylase.
As there is no crystal structure of these enzymes, a molecular modeling study was done in order to compare the rabbit and rat enzymes with the human one. The overall result is that rabbit enzyme could probably be a better choice in this regard, but in the case of large ligands, which could make putative interactions with the −4 subsite of pancreatic alpha-amylase, interpretation of results should be made cautiously.
Molecular modeling tools could be used to choose the most suitable model enzyme that would help to identify new enzyme inhibitors. In the case of alpha-amylase, three-dimensional structures of animal enzymes show differences with the human one which should be taken into account when testing potential new drugs.
Carbohydrate digestion has been targeted as a mean to control both postprandial increase of blood glucose and weight gain . Inhibitors of carbohydrate digesting enzymes, such as alpha-amylase and alpha-glucosidase, are now actively searched for, since they could ultimately make useful medicines against diabetes and obesity [2, 3]. There are numerous examples where inhibitors that were found to be effective on glycosidic enzymes in vitro[4–7], proved to possess hypoglycemic and weight decreasing effect in vivo[5, 7–9]. Pancreatic alpha-amylase inhibitors could be foreseen to become part of the drugs used for the so-called “diabesity” state. This enzyme inhibitors are especially of interest, as they have been reported to be devoid of the disturbing gastro-intestinal side-effect of some alpha-glucosidase inhibitors [10, 11].
Although the final target of this compounds is usually the human pancreatic alpha-amylase, it is still common place to use similar enzymes as models in the in vitro studies [5, 6]. In the next level, use of animal models is also a common method to assess the effect of these compounds, and in most of these studies, rats are privileged, as accessible, established and reproducible models of diabetes and obesity. There are also other alternatives, such as the possibility of using rabbits , it would be thus interesting to see which animal’s pancreatic enzyme is more similar to the human one, and how this similarity/ difference would affect the possibility of extrapolating the outcome of animal tests to human beings. Since there is no crystal structure of rat and rabbit enzymes, computer-generated models were used in this study in order to compare rabbit, rat, and human pancreatic enzymes with regard to their binding to a carbohydrate-based ligand as mimic of an inhibitor.
Sequences of rabbit (Oryctolagus cuniculus _protein accession number: XP_002715871) and rat (Rattus norvegicus _protein accession number: P00689) enzymes were retrieved from the NCBI (protein) (http://www.ncbi.nlm.nih.gov). The 3OLD.pdb file of the human enzyme was used to assess the length of the signal peptide, which was omitted in the alignments. ClustalW  and the BLAST tool of the NCBI site were used to perform alignments. The BLAST program compares the protein sequences and calculates the statistical significance of sequences that are matched together, and reports the percentage of “identity” (residues that are identical) and “similarity” (residues that are conserved), which have been mentioned in the Results section. Clustal W was used to make a multiple alignment in order to have an insight into areas of similarity (usually associated with conserved functional domains) as well as differences that occur as a consequence to substitutions and deletions of amino acids.
The ModWeb server version SVN.r1368M (web based version of Modeller ) was used in order to make the rabbit and rat pancreatic amylase model (https://modbase.compbio.ucsf.edu/scgi/modweb.cgi). Modeller performs comparative modeling by satisfaction of spatial restraints . The program uses the protein sequence, and a three-dimensional structure that has a high enough similarity to the sequence to make a three-dimensional protein model. In this case 1hx0.pdb corresponding to a porcine alpha-amylase with 1.38 Å resolution was used. Usually, 30% of identity between template and target is sufficient to obtain a reliable model ; in this case, percentage of identical residues of human and porcine enzymes with rat and rabbit is higher than 80%. Although multiple alpha-amylase structures have been elucidated, at the time this study has been done, the 1hx0.pdb structure had the best resolution between available mammalian alpha-amylase structures. The sequence of rat and rabbit alpha-amylase were given to the server, and the resulting models were retrieved. Model quality was assessed with the Molprobity  server, which analyses side chain rotamers and provide Ramachandran plots, which are indicative of the geometrical parameters of the model.
Docking and molecular dynamics simulation
Docking was performed with Autodock vina . The 3OLI.pdb file was first processed with the use of MOE 2009.10 (Chemical Computing Group Inc., Montreal,Canada). Additional molecules to alpha-amylase, including solvent, were deleted prior to docking. MGLtools (v. 15.4, revision 24) which is the graphical interface to Autodock was used to define the docking box and assign gasteiger charges to protein and ligand molecules. The docking box was positioned at x = −8.548, y = −21.483, z = 18.925 with a size of 64x82x58. A configuration file was prepared as the input of Autodock vina. In addition to the docking box specifications, in this configuration file, exhaustiveness (related to the docking precision and usually set at 8) was set on 20 and 100 poses were generated for the ligand. To validate the docking method that was used, RMSD between actual pose of the co-crystallized ligand and docked one was assessed with the use of the Profit server (http://www.bioinf.org.uk/software/profit) which is based on the McLachlan algorithm . The default method was used with inclusion of heteroatoms only and calculating RMSD for all atoms.
The dynamics simulation module of MOE 2009.10 was used to perform a short simulation of 5000 picoseconds length with T = 300K using the Noise-Poicare-Andersen Hamiltonian equations of motion. The protein was first minimized. Then solvation was done with 17725 water molecules. A preliminary simulation of t = 100 picoseconds, T = 300K was performed to equilibrate the system, and the output was used as the input of the actual 5000 picoseconds run. Sampling was done at each 5 picoseconds. The MMFFx94 forcefield was used for all the calculations.
Results and discussion
In the next stage, a three-dimensional model was made for RPA and RABPA. It should be mentioned that the tertiary structure of mammalian alpha-amylase has been reported only for human salivary and pancreatic enzyme, as well as porcine pancreatic enzyme. Small differences in the spatial arrangement of residues (in similar enzymes) could produce large effects on their interactions with potential ligands , and the fact that proteins belong to the same family is not enough to predict their ligand binding profile . In the absence of an experimentally obtained structure, the best way to detect these differences is to study a model. Protein models that are made by comparative modeling are considered to be highly reliable, especially when the percentage of identical residues is elevated . The template used to model RPA and RABPA was the 1hx0.pdb  structure which originally contained a carbohydrate-based ligand derivated from acarbose. As mentioned in the Methods section, this structure had the highest resolution in mammalian alpha-amylases structures of the protein database (PDB), at the time this study was performed. No missing atoms were detected in this template. Ramachandran plot of 1 h × 0.pdb indicates that 100% of its residues are situated in the allowed region. RPA and RABPA models were generated by the ModWeb server and subsequently minimized in MOE. Quality assessment of the generated models indicated them to be reliable. First, the results of evaluation methods that are implemented within the ModWeb server output were acceptable. The GA341 score was 1.0 for both models which is indicative of a reliable fold (GA341 > 0.7 is related to ≥ 95% probability of correct fold). The MPQS (ModPipe Quality Score) was 1.96 for RABPA and 2.08 for RPA which is a further approval of the reliability of the model (MPQS value higher than 1.1 is indicative of a reliable model) . Furthermore, as a mean to assess the geometrical validity of the models. Ramachandran plots  of RABPA and RPA were obtained from the MolProbity server. Ramachandran graphs of the two models indicated that for RABPA 99.8% (492/493) of all residues were in allowed regions and the only outlier was N378, where RPA had 100% (490/490) of all its residues in the allowed region, with no outlier in the structure. HPA, RPA and RABPA possess five conserved disulfide bonds, which correspond to residues pairs 28–86, 70-115, 141–160, 378–384, 450–462 in HPA and RABPA, and residues pairs 28–86, 70-115,141-157,375-381, and 447–459 in RPA, in which the numbering is altered due to the three-residue gap. It should be mentioned that the numbering here has been made according to the putative expressed protein (with omission of the signal peptide, by comparison with HPA_see Figure 1). In light of these quality assessment results, and with regard to the high similarity of the templates with the modeled targets, we believe that these models could be considered to have enough accuracy and biological plausibility for further ligand binding studies.
The three missing residues detected in the rat sequence are found to be located in a loop of domain B, which makes the loop distinctively shorter in RPA (Figure 2).
In order to get an approximation of the possible effectiveness of this ligand as a potential inhibitor of the enzyme, docking score was obtained for the HPA co-crystallized pseudo saccharide ligand acarviostatin II03 (with 7 carbohydrate-based units). This score was −14.6 kcal/mol for acarviostatin II03 while our ligand achieved a −13.4 kcal/mol. In comparison, acarbose (with four units) had a score of −11 kcal/mol. The Ki of acarviostatin II03 is reported to be of 0.0147 μM and the Ki of acarbose that is around 2.6 μM , which is suggesting that our ligand could be potentially better than acarbose, but this assumption remains to be verified.
Residues N53, Q63 and G104 are conserved in all three enzymes. However, V51 replaces I51 in RABPA, but it is conserved in RPA as I51. As observed in Figure 7b, this makes a shorter side chain in this place for RABPA, and consequently, more space exists in this location. Instead of A106, both RABPA and RPA possess G106, again with the consequence of providing more available space in that position. The most important difference occurs in V107, where it is changed to Q107 in RABBPA and N107 in RPA. This is a more radical change, which results in polar, and bulkier side chains taking the place of an aliphatic side chain. Thus, a potential designed ligand possessing a bulkier moiety in place of ring 1 of our ligand could probably fail to interact properly in this region with RABPA or RPA, while it might have in fact a good positioning in the human enzyme. Reversely, if some designed ligand find enough space in the vicinity of residues 106 or 51 in RABPA or RPA, it may be too large for HPA. A review of the other residues that are interacting with rings 2–7 of the ligands (based on the diagram shown in Figure 7a) shows their perfect conservation in the three enzymes, with the exception of T163 being replaced by a serine residue in RPA, while remaining T163 in RABPA.
Overall, the structural differences between RPA and HPA could be assessed to be more important than the ones between RABPA and the human enzyme, with regard to inhibitor design. Especially, the shortening of a loop in domain B of rat enzyme should be highlighted, as it may produce long-range effect in case of conformational changes. About the active site itself, if ligands are to be designed that would span subsites −3 to +3, RABPA would be a better choice for testing the compounds, and consequently, rabbit would be a better model in this regard. If a ligand present a structure capable of interacting with subsite −4, then results obtained from the use of both RABPA and RPA (and the animals themselves) could be different with the results obtained for the human enzyme. Since modeling techniques have now evolved to become more accessible, it would be advisable to use them in order to assess the possible positive/negative outcome of tests in animal models.
- Preuss HG: Bean amylase inhibitor and other carbohydrate absorption blockers: effects on diabesity and general health. J Am Coll Nutr. 2009, 28: 266-276.View ArticlePubMedGoogle Scholar
- Tundis R, Loizzo MR, Menichini F: Natural products as alpha-amylase and alpha-glucosidase inhibitors and their hypoglycaemic potential in the treatment of diabetes: an update. Mini Rev Med Chem. 2010, 10: 315-331. 10.2174/138955710791331007.View ArticlePubMedGoogle Scholar
- Tarling CA, Woods K, Zhang R, Brastianos HC, Brayer GD, Andersen RJ, Withers SG: The search for novel human pancreatic alpha-amylase inhibitors: high-throughput screening of terrestrial and marine natural product extracts. Chembiochem. 2008, 9: 433-438. 10.1002/cbic.200700470.View ArticlePubMedGoogle Scholar
- Le Berre-Anton V, Bompard-Gilles C, Payan F, Rouge P: Characterization and functional properties of the alpha-amylase inhibitor (alpha-AI) from kidney bean (Phaseolus vulgaris) seeds. Biochim Biophys Acta. 1997, 1343: 31-40. 10.1016/S0167-4838(97)00100-3.View ArticlePubMedGoogle Scholar
- Jo SH, Ha KS, Moon KS, Lee OH, Jang HD, Kwon YI: In vitro and in vivo anti-hyperglycemic effects of omija (Schizandra chinensis) Fruit. Int J Mol Sci. 2011, 12: 1359-1370. 10.3390/ijms12021359.PubMed CentralView ArticlePubMedGoogle Scholar
- Najafian M, Ebrahim-Habibi A, Hezareh N, Yaghmaei P, Parivar K, Larijani B: Trans-chalcone: a novel small molecule inhibitor of mammalian alpha-amylase. Mol Biol Rep. 2011, 38: 1617-1620. 10.1007/s11033-010-0271-3.View ArticlePubMedGoogle Scholar
- Tsujita T, Takaku T, Suzuki T: Chestnut astringent skin extract, an alpha-amylase inhibitor, retards carbohydrate absorption in rats and humans. J Nutr Sci Vitaminol (Tokyo). 2008, 54: 82-88. 10.3177/jnsv.54.82.View ArticleGoogle Scholar
- Layer P, Rizza RA, Zinsmeister AR, Carlson GL, DiMagno EP: Effect of a purified amylase inhibitor on carbohydrate tolerance in normal subjects and patients with diabetes mellitus. Mayo Clin Proc. 1986, 61: 442-447. 10.1016/S0025-6196(12)61978-8.View ArticlePubMedGoogle Scholar
- Najafian M, Ebrahim-Habibi A, Yaghmaei P, Parivar K, Larijani B: Core structure of flavonoids precursor as an antihyperglycemic and antihyperlipidemic agent: an in vivo study in rats. Acta Biochim Pol. 2010, 57: 553-560.PubMedGoogle Scholar
- Boivin M, Flourie B, Rizza RA, Go VL, DiMagno EP: Gastrointestinal and metabolic effects of amylase inhibition in diabetics. Gastroenterology. 1988, 94: 387-394.PubMedGoogle Scholar
- Aoki K, Muraoka T, Ito Y, Togashi Y, Terauchi Y: Comparison of adverse gastrointestinal effects of acarbose and miglitol in healthy men: a crossover study. Intern Med. 2010, 49: 1085-1087. 10.2169/internalmedicine.49.3218.View ArticlePubMedGoogle Scholar
- Chatzigeorgiou A, Halapas A, Kalafatakis K, Kamper E: The use of animal models in the study of diabetes mellitus. In Vivo. 2009, 23: 245-258.PubMedGoogle Scholar
- Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, McWilliam H, Valentin F, Wallace IM, Wilm A, Lopez R: Clustal W and Clustal X version 2.0. Bioinformatics. 2007, 23: 2947-2948. 10.1093/bioinformatics/btm404.View ArticlePubMedGoogle Scholar
- Pieper U, Webb BM, Barkan DT, Schneidman-Duhovny D, Schlessinger A, Braberg H, Yang Z, Meng EC, Pettersen EF, Huang CC: ModBase, a database of annotated comparative protein structure models, and associated resources. Nucleic Acids Res. 2011, 39: D465-D474. 10.1093/nar/gkq1091.PubMed CentralView ArticlePubMedGoogle Scholar
- Sali A, Blundell TL: Comparative protein modelling by satisfaction of spatial restraints. J Mol Biol. 1993, 234: 779-815. 10.1006/jmbi.1993.1626.View ArticlePubMedGoogle Scholar
- Ginalski K: Comparative modeling for protein structure prediction. Curr Opin Struct Biol. 2006, 16: 172-177. 10.1016/j.sbi.2006.02.003.View ArticlePubMedGoogle Scholar
- Chen VB, 3rd Arendall WB, Headd JJ, Keedy DA, Immormino RM, Kapral GJ, Murray LW, Richardson JS, Richardson DC: MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr D Biol Crystallogr. 2010, 66: 12-21.PubMed CentralView ArticlePubMedGoogle Scholar
- Trott O, Olson AJ: AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem. 2010, 31: 455-461.PubMed CentralPubMedGoogle Scholar
- McLachlan AD: Rapid comparison of protein structures. Acta Crystallogr A: Crystal Phys, Diffraction, Theor and Gen Crystallography. 1982, 38: 871-873. 10.1107/S0567739482001806.View ArticleGoogle Scholar
- Chen BY, Bandyopadhyay S: Modeling regionalized volumetric differences in protein-ligand binding cavities. Proteome Sci. 2012, 10 (Suppl 1): S6-10.1186/1477-5956-10-S1-S6.PubMed CentralView ArticlePubMedGoogle Scholar
- Najmanovich RJ, Allali-Hassani A, Morris RJ, Dombrovsky L, Pan PW, Vedadi M, Plotnikov AN, Edwards A, Arrowsmith C, Thornton JM: Analysis of binding site similarity, small-molecule similarity and experimental binding profiles in the human cytosolic sulfotransferase family. Bioinformatics. 2007, 23: e104-e109. 10.1093/bioinformatics/btl292.View ArticlePubMedGoogle Scholar
- Qian M, Nahoum V, Bonicel J, Bischoff H, Henrissat B, Payan F: Enzyme-catalyzed condensation reaction in a mammalian alpha-amylase. High-resolution structural analysis of an enzyme-inhibitor complex. Biochemistry. 2001, 40: 7700-7709. 10.1021/bi0102050.View ArticlePubMedGoogle Scholar
- Ramachandran GN, Ramakrishnan C, Sasisekharan V: Stereochemistry of polypeptide chain configurations. J Mol Biol. 1963, 7: 95-99. 10.1016/S0022-2836(63)80023-6.View ArticlePubMedGoogle Scholar
- Brayer GD, Luo Y, Withers SG: The structure of human pancreatic alpha-amylase at 1.8 A resolution and comparisons with related enzymes. Protein Sci. 1995, 4: 1730-1742. 10.1002/pro.5560040908.PubMed CentralView ArticlePubMedGoogle Scholar
- Brayer GD, Sidhu G, Maurus R, Rydberg EH, Braun C, Wang Y, Nguyen NT, Overall CM, Withers SG: Subsite mapping of the human pancreatic alpha-amylase active site through structural, kinetic, and mutagenesis techniques. Biochemistry. 2000, 39: 4778-4791. 10.1021/bi9921182.View ArticlePubMedGoogle Scholar
- Maurus R, Begum A, Kuo HH, Racaza A, Numao S, Andersen C, Tams JW, Vind J, Overall CM, Withers SG, Brayer GD: Structural and mechanistic studies of chloride induced activation of human pancreatic alpha-amylase. Protein Sci. 2005, 14: 743-755. 10.1110/ps.041079305.PubMed CentralView ArticlePubMedGoogle Scholar
- D’Amico S, Gerday C, Feller G: Structural similarities and evolutionary relationships in chloride-dependent alpha-amylases. Gene. 2000, 253: 95-105. 10.1016/S0378-1119(00)00229-8.View ArticlePubMedGoogle Scholar
- Qin X, Ren L, Yang X, Bai F, Wang L, Geng P, Bai G, Shen Y: Structures of human pancreatic alpha-amylase in complex with acarviostatins: Implications for drug design against type II diabetes. J Struct Biol. 2011, 174: 196-202. 10.1016/j.jsb.2010.11.020.View ArticlePubMedGoogle Scholar
- Kandra L, Remenyik J, Batta G, Somsak L, Gyemant G, Park KH: Enzymatic synthesis of a new inhibitor of alpha-amylases: acarviosinyl-isomaltosyl-spiro-thiohydantoin. Carbohydr Res. 2005, 340: 1311-1317. 10.1016/j.carres.2005.03.003.View ArticlePubMedGoogle Scholar
- Corbeil CR, Williams CI, Labute P: Variability in docking success rates due to dataset preparation. J Comput-aided Mol design. 2012, 26: 775-786. 10.1007/s10822-012-9570-1.View ArticleGoogle Scholar
- Malleda C, Ahalawat N, Gokara M, Subramanyam R: Molecular dynamics simulation studies of betulinic acid with human serum albumin. J Mol Model. 2011, 18: 2589-9257.View ArticlePubMedGoogle Scholar
- Ul-Haq Z, Khan W, Kalsoom S, Ansari FL: In silico modeling of the specific inhibitory potential of thiophene-2,3-dihydro-1,5-benzothiazepine against BChE in the formation of beta-amyloid plaques associated with Alzheimer's disease. Theor Biol Med Model. 2010, 7: 22-10.1186/1742-4682-7-22.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.