Indian Journal of Chemistry Vol. 45B, July 2006, pp. 1692-1698
':I:~'-J '~_' .r.,
2,3-Diphenyltetrahydrofurans (DPTF) - A new class of stereogenic diaryl heterocycles as potential COX-2 inhibitors - Computational evaluation of
COX-2-DPTF binding behaviour
Palwinder Singh*, Pervinder Kaur, Anu &Subodh Kumar Department of Chemistry, Guru Nanak Dev University, Amritsar 143005 India
E-mail: palwinder_singh_2000@yahoo.com
Received 02 June 2005; accepted (revised) 05 September 2005
Based on the reported COX-2 selectivity of a,p-diarylheterocycles based NSAIDs' - a new class of 2,3-diphenyl- tetrahydrofurans (DPTF) possessing COOH, CH20H or CH2NH2 groups at C-5 and three chiral centers C-2, C-3 and C-5 have been designed. In total 12 title compounds have been studied for their binding in the active sites of COX-2 and COX-I enzymes. For comparison, the binding behavior of known COX-2 selective (rofecoxib, celecoxib, valdecoxib) and non- selective (aspirin, ibuprofen) drugs has also been studied. The trends of COX-2 selectivity of known NSAIDs' are parallel with their reported results. In general, DPTF derivatives show binding energies in COX-2 active site better than rofecoxib and comparable to aspirin and poor binding energies in COX-I active sites pointing towards these molecules being selective COX-2 inhibitors. The stereochemistries at C-2, C-3 and C-5 carbons of DPTF derivatives considerably affect their bindings in the active sites of COX-I and COX-2 and the COX-2 selectivity. The ligands with configuration (2S, 3R,5S) show higher selectivity for COX-2 over COX-I in comparison to other configurational isomers.
IPC: Int.CI.8C07D
Keywords: cyclooxygenase, NSAIDs, diphenyltetrahydrofurans, stereogenic molecules, doc kings
Cyclooxygenases (COXs) and lipoxygenases (LOXs) are the key enzymes in arachidonic acid metabolism'".
One isoform of cyclooxygenase i.e., COX-I impart a desirable function in the formation of thromboxane, gastric acid secretions and induction of labor pains"
while the second form, COX-2 is responsible for an undesirable role in the production of prostaglandins during its expression in inflammatory cells that leads to pains". The traditional non-steroidal anti-inflam- matory drugs (NSAIDs) like aspirin, ibuprofen, diclofenac etc. block both the COX-I and COX-2 enzymes and their non-selectivity leads to undesirable side effects'. Once the X-ray structures of COX-I and COX-2 enzymes were established, a number of selective inhibitors of COX-2 have been reported in which the basic skeleton involves the presence of aryl/substituted aryl moieties at adjacent Sp2 hybridized carbons of pyrazoles'", furanones'", cyclopentenes", imidazoles'", isoxazoles", pyrroles'j, oxazolones':', acyclic olefins", indoles'v'" etc., and a few like celecoxib ', rofecoxib", valdecoxib' are in clinical use. In a recent report", the theoretical studies have been carried out for the inhibition mechanism of
COX-2 with various inhibitors like celecoxib, rofecoxib etc. In spite of the fact that COX-2 active site being chiral and difference in activity of two enantiomers of ibuprofen" is known, most of the COX-2 selective inhibitors like celecoxib, rofecoxib are achiral in nature.
Bearing in mind the lack of flexibility of the reported COX-2 inhibitors at the two aryl rings due to their presence on Sp2 hybridized atoms, reasonable flexibility in the COX-2 recognition site which can adapt itself to subtle changes in the conformations of the inhibitor molecule and due to chirality of the COX-2 recognition site, the search of new molecules having structural rationale similar to the known inhibitors and having varied chiral centers for the effective binding is necessitated. During the programme on the development of green approaches for the synthesis of new organic molecules, a simple strategy was developed for the synthesis of 1,2- diaryltetrahydrofuran (DPTF) derivatives". Here, DPTF derivatives 1-4 (Figure 1) have been designed in which the two aryl rings are placed at C-2 and C-3 carbons on the opposite faces of the nearly planar
4
SINGH et a/: STEREOGENIC DIARYL HETEROCYCLES AS COX-2 INHIBITORS
R
la-c 2a-c
R
3a-c 4a-c
Figure 1
tetrahydrofuran ring. Either of hydroxymethyl, aminomethyl or carboxylic acid group, are present on the C-5 carbon.
Further, since the tetrahydrofuran ring has three chiral centers and it is possible to have eight stereo- isomers, the fixation of trans-relative geometries of the two aryl rings at C-2 and C-3 carbons results in four stereo isomers. These constitute two pairs of diastereorners.
All the four stereo isomers of tetrahydrofuran (Figure 1) possessing hydroxymethyl, aminomethyl or COOH group at C-5 have been considered for computational evaluation of their interactions with COX-2 and COX-I enzymes. So, in total twelve tetrahydrofuran derivatives have been studied. For comparison and validation of the procedures, the known NSAlDs, both selective (rofecoxib, celecoxib, valdecoxib) and non-selective (ibuprofen, diclofenac) towards COX-2 have also been made part of docking studies. The results clearly demonstrate the role of stereogenecity and the nature of the substituents on the selective binding of OPTF derivatives with COX-2.
Computational details Methodology
All molecular techniques used in this manuscript were performed on CAChe Worksystem Pro 6.1. The crystal structures of the enzymes COX-l and COX-2 were downloaded from Protein Data Bank (http://www.rcsb.org/)_as POB files.
COX-2: The file containing the crystal structure of CQX-? 'Jiit)l. its: selective inhibitor SC-558 in the
1693
active site (POB entry 6COX) was downloaded. It is a dimeric structure with chains A and B each consisting of 550 residues. Each chain has SC-558, 3 molecules of Nsacetyl-Dcglucosamine (NAG) and 1 heme (HEM) group. The chain A.with the residues, water and the hetero groups (HEM, NAG) within a radius of 5
A
was refined and further cleaned by ascertaining the hybridization and valence of each atom of SC-558 and introducing H-atoms to the protein residues. The cleaned structure of COX-2 carried a net charge of -3 and 9047 atoms. The active site residues of COX-2 are shown in Figure 2.COX-I: The crystal structure of COX-l cornplexed with ibuprofen (POB entry 1EQG)2o was downloaded.
It was having a dimeric structure with two chains, each having 551 residues, 7 molecules of N-acetyl-O- glucosamine (NAG), 2 molecules of Bvocrylglucoside (BOG), one heme group and a molecule of ibuprofen.
The monomer within the 5
A
radius of''chain A'was refined and cleaned by checking the: 'hYbd~!.izatjbli, valence of the ligand and introducing;;H-atoms.1(Hhe protein residues. COX-l carries a~n€t,~:cgai-ge':df"+2 and' 9680 atoms. The active site residues'or
G:0)~Ylare shown in Figure 3. ' ,:i',' :';'Jfi1i ,',)"
All the tetrahydrofuran derivatives were built in the CAChe workspace and the structures were refined by performing an optimized geometry calculation in MOPAC using PM5 parameters.·..",
Docking and binding evaluations
In the automated module of CAChe Work system
]?ro 6.1, the ligands were docked into the active site of COX-2 and COX-J using a genetic algorithm with a fast, simplified Potential of Mcan Force (PMF)2I, It is
Figure 2- Active site residues of COX-2 with SC-558.
Hydrogens are suppressed for clarity
1694 INDIAN 1.CHEM., SEC B, JULY 2006
Figure 3 - Active site residues ofCOX-I with ibuprofen.
Hydrogens are suppressed for clarity
assumed that the protein and the ligand dock non- covalently. The standard PMF implementation used Amber" van der Waal's potential for this purpose.
The docking is carried with flexible ligand intoarigid protein active site. The general procedure for docking process starts with the addition of energy minimized target ligand on the enzyme. The active site and the ligand were specified in the programme. The different starting parameters were optimized by using 15x15x15
A
box located at the centre of the target active site using aunited atom (explicit hydrogens are not considered) potential of mean force (PMF) with a docking algorithm that has a population of 50 chromosomes and runs for 6000 generations. The process of docking is repeated until a constant value of docking score is reached. This takes about 12000- 18000 generations. The final results are parameterized.in terms of docking score in KcaIlmol. The docked ligand-COX-2 complex is interpreted by looking at the H-bonding or hydrophobic interactions of the ligand with the amino acid residues in the active site.
The same procedure was followed for docking of different DPTFs' into the active site of COX-l and COX-2.
Results
Validation of PMF method: To ensure the validity of the programme, before docking the test compounds, the docking of SC-558 into the active site of COX-2 was performed. This selective inhibitor of COX-2 binds in the active site with abinding score of -85 KcaIlmol and nns deviation of 0.45 is observed.
The docked structure of SC-558 in the active site of
Figure 4 - Figure showing the close overlapping of docked structure ofSc-558 with its crystal structure (rms error is 0.45).
Native ligand islabeled, other is docked ligand.
Table I - Comparison ofICso (IlM) of various COX-2 and COX-I inhibitors inHuman Blood with their docking scores
(Kcal/mol)
Entry Ligand IC50 (IlM) Dock score
(Kcal/mol) COX-I COX-2 COX-1 COX-2
I Celecoxib 14 1.2 1713 -87.2
2 Rofecoxib 40 0.3 833 -54.4
3 Valdecoxib 25 0.89 479 -55.4
4 Etoricoxib 12 I.l 1042 -58
5 Ibuprofen Non-selective -21 -47.4
6 Diclofenac Non-selective -41 -27.4
COX-2 is shown in Figure 4. The close overlapping of the docked structure with the native ligand (X-ray crystal structure) ensures the validity of the programme.
Docking of known selective and non-selective cyclooxygenase inhibitors into the active sites of COX-l and COX-2: The non-selective behaviour of ibuprofen, diclofenac and selective behaviour of celecoxib, rofecoxib, valdecoxib towards COX-2 is quite in agreement with their binding energies (docking score) with COX-1 and COX-2 as calculated from the docking of these compounds in COX-l and COX-2 active site (Table I). In case of COX-2 selective NSAIDs viz. celecoxib, rofecoxib, valdecoxib, etoricoxib, the negative binding energies for COX-2 and positive binding energies with COX-l
---_.
__ ._
...._--._--- ---
SINGH et al: STEREOGENIC D1ARYL HETEROCYCLES AS COX-2 INHIBITORS
are in agreement with their COX-2 selectivity as reported in terms of ICsovalues (entries 1-4, Table I).
The non-selective compounds ibuprofen and diclofenac show negative binding score with both COX-l and COX-2 enzymes in agreement with their reported non-selectivity. During binding of these known NSAIDs in the binding pocket of COX-2, the conformational placement of amino acid residues in the active site is observed.
Celecoxib binds in the active site of COX-2 mainly through hydrophobic interactions. Nr-phenyl ring is surrounded by Y355 and V523 and the sulphonamide group present on this ring approaches A516, H90 and G 192 where NHcelecoxib---O=CG192distance is in the range of weak H-bonding (2.31 A). C-2 phenyl ring is enveloped by F518, W387, M522, G526, Y385, S353.
The CF3 group present at C-4, through H-bonding between CF---HNRI20 (1.73 A) has blocked R120, the prominent residue involved in arachidonic acid metabolism. Therefore, the aromatic rings present on celecoxib through 1t-1t interactions with amino acid residues keep it placed in the active site of COX-2 and CF3 group blocks R120 via H-bonding.
Docking of DPTF Derivatives 1-4 into COX-1 and COX-2 active sites: The dockings of 12 DPTF derivatives (Figure 1) in COX-2 and COX-l active sites has been studied (Table II). In these compounds, all the carbon atoms of the central core (tetrahydrofuran ring) are Sp3hybridized. The C-2 and C-3 carbons carry aryl moieties and the groups chosen to be present at C, are H-donors (CH20H) as well as H-acceptors/H-donors (COOH, CH2NH2). The pairs
1695
1a-c, 4a-c and 2a-c, 3a-c are enantiomers and la-c, 2a-c; 1a-c, 3a-c; 3a-c, 4a-c; 2a-c, 4a-c have diastereomeric relation with each other. These DPTF derivatives having two phenyl rings on the tetrahydrofuran moiety have some structural similarities with COX-2 selective drug rofecoxib-a 3,4-diphenyl-5H-furan-2-one. The decrease in steric repulsion between two phenyl rings due to their presence on the opposite faces of the tetrahydrofuran ring provides a degree of freedom for free rotation of the two aryl rings. All the 12 diphenyl tetrahydrofurans go into the active site of COX-2 and show negative docking scores. These DPTFs show positive docking score (except for 4b, 3c and 4c) for the COX-l enzyme, resulting in reasonable to high binding specificities for COX-2. Therefore these molecules have potential to act as COX-2 selective NSAIDs. The difference in binding of various DPTF molecules with COX-2 and COX-l enzymes points towards the significant role of various stereo isomers and the C-5 substituents in their binding abilities.
Discussion
All the 12 DPTF molecules show binding in the COX-2 active site with binding scores between -53.0 and -67.4 Kcal/mol. However, in case of dockings of DPTF molecules in COX-l cavity, the docking scores vary between -19.4 to 94.2 Kcal/mol. These results clearly point that the stereo chemistries at C-2, C-3 and C-5 centers of DPTF molecules and the nature of the substituent at C-5 significantly affect their binding in COX-lICOX-2 active sites.
Table II-Docking score (Kcal/mol) of various DPTF derivatives 1-4 with COX-I
(pdb 10:1EQG) and COX-2(pdb 10:6COX).
ligand Configuration at Docking score (Kcal/mol) COX-2 selectivity C2, C3, C5 COX-I (Ei) COX-2 (E2) (Ei-E2)
1a 2S,3R,5R 30.8 -61.0 91.8
2a 2S,3R,5S 47.2 -56.3 103.5
3a 2R,3S,5R 28.7 -54.6 83.3
4a 2R,3S,5S 8.6 -60.0 68.6
Ib 2S,3R,5R 52.6 -67.4 120
2b 2S,3R,5S 94.2 -62.7 156.9
3b 2R,3S,5R 68.0 -60.0 128
4b 2R,3S,5S -19.4 -65.6 85
Ie 2S,3R,5R 16.9 -60.1 77
2e 2S,3R,5S 39.4 -55.7 95.1
3e 2R,3S,5R -1.9 -53.0 54.9
4e 2R,3S,5S -11.8 -60.8 72.6
1696 INDIAN J. CHEM., SEC B, JULY 2006
Amongst the four stereo isomers, irrespective of the substituent at C-5, the (2R,3S,5S)-stereoisomers (4a, 4b and 4e) show highest order of binding with COX-I resulting in lowest binding specificities in each category. The ligands with (2S,3R,5R)- configuration (la, Ib and Ic) which are enantiomers of respective ligands 4, exhibit highest binding with COX-2 in each category with appreciable selectivity for COX-2 over COX-I. The graph plotting the binding energy of DPTF molecules with COX-2 vs
their energy difference for binding with COX-2 and COX-l (selectivity) clearly shows that in case of enantiomers la-c and 4a-c, the selectivity are directly proportional to the binding energies. As the bindings of enantiomers la-c and 4a-c in the COX-2 active site increase, their selectivity towards COX-2 also increase (Figure 5). Among the enantiomers Ia-c and 4a-e, 1b shows highest binding energy (docking score) and also the highest selectivity for COX-2 over COX-I. The compounds 2 and 3 are enantiomers with
•
1b----.- .•-2-b_ ...---- ..---- -"--160 .~
140-i!
.3b I
120-J
.-
o
I> ,
•.::: I
ul<1)1
v
VJ-70
100
l
80,
3c 60~
.
.---- --- --- 40~
-65 -60 -55 -50
Binding energy (KCallmoI) .4b
Figure 5 - The plot of binding energies of DPTF derivatives with COX-2 (E2) vs.difference inbinding energies (E1-E2)
respect to each other and have diastereomeric relationship with Iand 4. Here again, the selectivity is directly proportional to binding with COX-2 (Figure 5). 2b shows the highest selectivity among the 12 DPTFs. Based upon the substituent at C-5, in all the 12 diphenyl tetrahydrofurans, the order of binding with COX-2 as well as the selectivity for COX-2 over COX-l is COOH >NH2>OH.
Compounds Ib, 2b and 4b show the highest binding with COX-2 in comparison to their respective isomers. The enantiomers 2b and 3b with COOH functional group at C-5 show higher specificity than 2a, 2e and 3a, 3e respectively. Hence, out of the 12 compounds lb and 2b show appreciable binding as well as selectivity for COX-2 over COX-I.
The further rationalization of mode of binding of these DPTF molecules in active site of COX-2 has been based upon the amino acid residues present around the ligands in the active site pocket of COX-2.
On the basis of structural features essential for binding in the cavity, the DPTF molecules could be divided into three segments: 2-phenyl, 3-phenyl and 5-carboxylic/amino, hydroxymethyl. The residues surrounding each phenyl group and the C-5 substituent have been determined (Table III). Based upon the stereochemistry at the three chiral centers of tetrahydrofuran, three segments of DPTFs are placed into different sub-pockets of COX-2 active site. The most prominent bindings are observed in the case of ligands Ib and 2b. For ligand 2b, the C-2 phenyl ring is stabilized by hydrophobic interactions from Y385, W387, M522 and C-3 phenyl ring is surrounded by H90, S353 and V523. The C-5 substituent is having very characteristic orientation where it forms H-
ligand Configuration at C2,C), C,
Table III -Amino acid residues around each segment ofDPTF derivatives when docked into COX-2.
Ia 2a 3a 4a Ib 2b 3b 4b Ie 2e 3e 4e
2S,3R,5R 2S,3R,5S 2R,3S,5R 2R,3S,5S 2S,3R,5R 2S,3R,5S 2R,3S,5R 2R,3S,5S 2S,3R,5R 2S,3R,5S 2R,3S,5R 2R,3S,5S
2-phenyl
R 120,Y355, A527, L531, L352, W387, Y385,
V116, RI20, Y355, A527, L531 R I20,S353, Y355, A527 L352,S353, F518,V523 Y385, W387, M522 L384, W387, S530 RI20, V349, Y355, L531 R120, A527, L531 V349, L352, Y385, W387 Y355, A527, L531 R120, Y355, A527
Amino acid residues surrounding 3-phenyl
F381, Y385, W387, S530 H90,Y385, W387, F518 H90, L352, Y355, R513, V523, L384, Y385, W387, M522 Y348, Y385, W387 H90,S353,S523
R I20, S353, Y385, A527 L352, Y385, W387, F518 Y385, W387, S530 H90, S353, F518 H90,L352, Y355,V523 L352, Y385, W387,F518
5-substituent S353,L352 RI20,L352 W387
H90,S353, V523 RI20, Y355, V523, RI20
H90 H90
L352, S353, V523 R120, Y355 F381, G526, S530 S353
SINGH et al: STEREOGENIC DIARYL HETEROCYCLES AS COX-2 INHIBITORS
¥
IFigure 6 - 2b docked in the active site of COX-2
~8.
Figure 7 - I bdocked in the active site of COX-2
bonding with two NH groups of R120 (Figure 6).
This functionality of R120 is responsible for arachidonic acid metabolism and blockage of this guanidine moiety would render the COX-2 inactive. It seems as 2b could be the potential COX-2 inhibitor.
In case of lb, the C-2 phenyl ring is surrounded by L352, S353, F518, V523 and C-3 phenyl ring is enveloped by Y348, Y385, W387. Its carboxylic group at C-5 has shifted away from the guanidine part of R 120 with distances 2.21
A
and 2.85A
from two NHs of RI20 (Figure 7) and does not show H- bonding with R120.1697
Conclusions
The control of stereochemistries at C-2, C-3 and C- 5 carbons of DPTF molecules significantly affects the binding and selectivity in COX-l active site which results these molecules to be COX-2 selective with varied selectivity orders. The presence of COOH moiety at C-5 further enhances the binding capacity and selectivity and results in better discrimination between COX-l and COX-2 active sites.
Acknowledgements
The authors thank CSIR [project no.
0I(l735)/02/EMR-II] and DST (project no.
SR/FTP/CS-20/2001) for financial support. One of the authors (Anu) thanks CSIR for SRF.
References
Hamberg M, Svensson J, Wakabayashi J & Samnelsson B Proc Nat Acad Sci USA. 71,1974,345.
2 Nugteren D H & Hazelhof E, Biochim Biopliys Acta 326, 1973,448.
3 Musser J H&Kreft A F,JMed Chem 35, 1992, 250 I.
4 Khanapure S P, Garvey D S,Young D V, Ezawa M, Earl RA, Gaston R D, Fang X, Marty M, Martino A, Shamway M, Trocha M, Marek P, Tam S W, Janero D R & Letts L G, J Med Client 46, 2003, 5484 and references cited therein.
5 Shin S S, Byun Y, LirnK M, Choi J K, Lee Ki-W, Moh J H, Kim J K, Jeong Y S, Kim JY, Choi Y H, Koh H-J, Park Y-H, Oh Y I,Noh M-S &Chung SJMed Gem 47, 2004, 792.
6 Penning T D,Talley J J, Bertenshaw S R, Carter J S, Collins P W, Docter S, Graneto M J, Lee L F, Malecha JW, Miyashiro J M, Rogers R S, Rogier DJ, Yu SS, Anderson G D, Burton EG, Cogburn JN, Gregory S A, Koboldt C M, Perkins W E, Seibert K, Veenhuizen A W, Zhang Y Y & Isakson P C, J Med Gem 40,1997, 1347.
7 Singh S K, Vobbalareddy S, Shivaramakrishna S, Krishnamraju A, Rajjak S A, Casturi S R, Akhila V & Rao Y K,Bioorg Med Clietn LeI!2004, 1683.
8 Prasit P, Wang Z, Brideau C, Chan C C, Charleson S, Cromilish W, Ethier D, Evans J F, Ford-Hutchinson A W, Gauthier J Y, Gordon R, Gyay J, Gresser M, Kargman S, Kennedy B, Leblanc Y, Leger S, Mancini J, 0' eill G P, Ouellet M, Percival M D, Perrier H, Riendeau D, Rodger I, Tagari P,Therien M, Vickers P,Wong E, Xu L-J, Young R N
&Zamboni R, Bioorg Med Chem Leu 9,1999, 1773.
9 Li JJ, Anderson G D, Burton E G, Cogburn JN, Collins J T, Garland D J, Gregory SS,Huang H-C, Isakson PC, Koboldt C M, Logusch E W, Norton M B, Perkins W E, Reinhard E.I, Seibert K, Veenhuizen A W, Zhang Y & Reitz D B, JMed Chern 38, 1995,4570.
10 Khanna I K, HuffY Y R M, Weier R M, Xu X, Koszyk F J, Collins P W, Cogburn J N, Isakson PC, Koboldt C M, Masferrer J L, Perkins W E, Seibert K, Veenhuizcn A W,
Yuan J,Yang D&Zhang YY,JMed Che1ll43, 2000, 3168.
II Talley JA, Brown D L, Carter JS, Masferrer M J, Perkins W E, Rogers R S, Shaffer A F, Zhang Y Y, Zweifel B S &
Seibert K,JMed Cliem43, 2000, 775.
1698
Tetrahedron
te«
45, 2004, 3793.17 Soliva R, Almansa C, Kalko S G, Luque F J & Orozco M, J MedChern,46, 2003, 1372.
18 Cleij M, Archelas A & Furstoss R, J Org Chern64, 1999, 5029 and references cited therein ..
19 Kumar S, Kaur P, Chimni S S·&'Singh P,Synlett, 2001, 1431.
20 Selinsky B S, Gupta K, Sharkey C
T
&Loll P J,Biochemistry, 40,2001,5170.21 Muegge I&Martin Y,JMed Chern,42,1999,791.
22 Cornell W D, Cieplak P, Bayly CI, Gould I R, Merz Jr K M, Ferguson D M, Spellmeyer D C, Caldwell J W& Kollman P A,JAm Chern Soc, 117, 1995,5179 .
INDIAN J. CHEM., SEC B, JULY 2006
12 Khanna I K, Weier R M, Yu Y, Collins P W, Miyashiro J M, Koboldt C M, Veenhuizen A W, Currie J L, Seibert K &
Isakson P C,JMed Chern40, 1997, 16 I 9.
13 Puig C, Crespo M I, Godessart N, Feixas J, Ibarzo J, Jimenez J, Soca L, Cardelus I, Heredia A, Miralpeix M, Puig J, Beleta J, Huerta J M, Lopez M, Segarra V, Ryder H&Palacios J M, JMed Chern43, 2000, 214.
14 Uddin Md J, Praveen Rao P N &Knaus E E, Bioorg Med Chern Leu 14,2004, 1953.
15 Hu W, Guo Z, Yi X, Guo C, Chu F&Cheng G,Bioorg Med Chern11,2003,5539.
16 Campbell J A, Bordunov V, Broka C A, Dankwardt J, Hendricks R T, Kress J M, Walker K A M & Wang J-H,
