CHEMMEDCHEM COMMUNICATIONS DOI: 10.1002/cmdc.201300455
Structural Basis for the Inhibition of AKR1B10 by Caffeic Acid Phenethyl Ester (CAPE) Liping Zhang, Hong Zhang, Xuehua Zheng, Yining Zhao, Shangke Chen, Yunyun Chen, Renwei Zhang, Qing Li,* and Xiaopeng Hu*[a] Caffeic acid phenethyl ester (CAPE), the major bioactive component of honeybee propolis, is a potent selective inhibitor of aldo-keto reductase family member 1B10 (AKR1B10), and a number of derivatives hold promise as potential anticancer agents. However, sequence homology between AKR1B10 and other members of the superfamily, including critical phase I metabolizing enzymes, has resulted in a concern over the selectivity of any potential therapeutic agent. To elucidate the binding mode of CAPE with AKR1B10 and to provide a tool for future in silico efforts towards identifying selective inhibitors, the crystal structure of AKR1B10 in complex with CAPE was determined. The observed interactions provide an explanation for the selectivity exhibited by CAPE for AKR1B10, and could be used to guide further derivative design.
Aldo-keto reductase family member 1B10 (AKR1B10), also called small-intestine aldose reductase,[1] is a promising antineoplastic target. This enzyme is abnormally overexpressed in many malignant tumors,[2] and has been found to interrupt cell differentiation induced by retinoid acids,[3] impair cell apoptosis mediated by intracellular toxic carbonyls and isoprenyl aldehydes,[4] and metabolize anticancer drugs such as daunorubicin and idarubicin.[5] A major difficulty in developing AKR1B10 inhibitors is the lack of selectivity between AKR1B10 and the homologous enzymes of the NADPH-dependent aldo-keto reductase (AKR) superfamily, which consists of around 150 critical phase I metabolizing enzymes that share high sequence similarities. A well-known case is the adverse effects caused by cross-inhibition with aldehyde reductase AKR1A1 (65 % sequence identity with AKR1B1), which was believed to be partially responsible for the clinical failures of aldose reductase (AKR1B1) inhibitors as type 2 diabetic complication drugs in the 1990s.[6] Because of the high sequence identity (71 %) shared between AKR1B10 and AKR1B1, an understanding of the factors controlling selectivity between these two enzymes is critical for studies of AKR1B10 inhibitors. In 2011, Soda et al. identified caffeic acid phenethyl ester (CAPE), the major bioactive component of honeybee propolis, [a] L. Zhang, H. Zhang, Dr. X. Zheng, Y. Zhao, Dr. S. Chen, Y. Chen, R. Zhang, Prof. Q. Li, Prof. X. Hu Centre for Cellular & Structural Biology School of Pharmaceutical Sciences, Sun Yat-sen University 132 East Circle, University City, Guangzhou 510006 (China) E-mail: [email protected] Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.201300455.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
as an efficient in vitro AKR1B10 inhibitor. Based on docking studies, they designed and synthesized several CAPE derivatives that exhibit much higher activity towards AKR1B10 and selectivity towards AKR1B1 (Table 1).[7] However, the proposed mechanism of inhibition of AKR1B10 by CAPE and its derivatives is not fully supported by the reported experimental results.[7, 8] Hydrogen bonds between the known AKR1B1/ AKR1B10 inhibitors and the active site residues (Tyr 49, His 111, etc.) are essential for inhibition, but the predicted hydrogen bonds of CAPE to AKR1B10 (C4 OH···Tyr 49 OH, 3.6 ; C4 OH···His 111 Ne2, 3.7 ) are too weak to efficiently anchor CAPE to the active site. Moreover, we recently identified significant inhibitor-induced conformation changes at the active site of AKR1B10,[10] which implies that the crystal structure (AKR1B10/ NADP + /tolrestat complex, PDB ID: 1ZUA) used by Soda et al.[7] in the docking studies might not be suitable to predict the interactions of CAPE derivatives with AKR1B10. To provide a solid structural basis for developing CAPEbased AKR1B10 inhibitors, here we report the 2.1 resolution crystal structure of the AKR1B10/NADP + /CAPE ternary complex (PDB ID: 4GQ0, see Table S1 in the Supporting Information for data collection and refinement statistics). This crystal structure not only shows that CAPE indeed forms strong classical hydrogen bonds via its catechol moiety with the active site residues of AKR1B10, but also clearly explains the higher activity towards AKR1B10 and excellent selectivity towards AKR1B1 of CAPE derivatives.[7] Thus, this structure might be a more suitable model for studies of interactions of CAPE derivatives with AKR1B10. Superimposing structures of the AKR1B10/NADP + /CAPE ternary complex and the AKR1B10 holoenzyme indicates that CAPE does not induce meaningful conformational changes at the active site, except for small movements of Phe 123 and Leu 302 that favor interactions between the phenethyl tail with the so-called specificity pocket (Figure S1 in the Supporting Information). This result is significantly different from the AKR1B10/NADP + /tolrestat crystal structure, in which tolrestat induces a Trp 112 side chain flip and movement of the Leu 302–Ser 304 segment, which closes the specificity pocket.[10] A schematic presentation of the AKR1B10/NADP + /CAPE ternary complex active site is shown in Figure 1 a. The electron density map clearly defines the planar conjugated system of the catechol ring and the acrylic linker (Figure S1 in the Supporting Information), as observed in the crystal structure of CAPE.[8] It is noteworthy that the catechol ring and the acrylic linker of CAPE are not coplanar in the docking model reported ChemMedChem 2014, 9, 706 – 709
706
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
Table 1. Reported IC50 values towards recombinant human AKR1B10 and AKR1B1 for caffeic acid phenethyl ester (CAPE) and other compounds. Entry
Compd R1
IC50 [mm][a]
Structure R2 R3
R4
AKR1B10
1
CAPE
(CH2)2Ph
H
OH
OH
0.5 (0.08[c])
2
Ethyl caffeate
CH2CH3
H
OH
OH
6.5
3
Caffeic acid
H
H
OH
OH
4
2
(CH2)2Ph
H
OH
H
5
3
(CH2)2Ph
H
H
OH
> 50 19[c]
Ratio[b] AKR1B1 4.1 (0.57[c])
3
> 50
–
73[c]
0.069
[c] [c]
6
9a
(CH2)2Ph
OMe
H
OH
0.013
7
Tolrestat
–
–
–
–
0.054[d]
8
18
4
2.2
[c]
31
7.2
[c]
554
0.014[d]
0.3
[a] IC50 values represent the mean of three determinations, where the standard error was less than 15 %. [b] Ratio of AKR1B1/AKR1B10 IC50 values. [c] Value determined by Soda et al. and reported in Ref. [7]. [d] Tolrestat values are taken from Ref. [[9]].
Figure 1. a) The crystal structure of AKR1B10 in complex with CAPE. Protein residues at 4 resolution are shown in grey, while the ligand is depicted in yellow. Hydrogen bonds are shown as red dashed lines. b) Model of 9 a bound to AKR1B10. The residue Trp 111 of AKR1B1 (purple) is superimposed, showing an unfavorable distance (2.0 , dashed yellow line) from the 2-methoxy group of 9 a. The hydrophobic contact between the CAPE 2-methoxy group and Leu 302 is also shown (dashed blue line, 3.2 ).
by Soda et al.[7] The catechol ring occupies the catalytic pocket (consisting of Tyr 49, His 111, and NADP + ) of AKR1B10, while the phenethyl tail is rotated about 608 to the conjugated plane, and is located in the hydrophobic specificity pocket (primarily consisting of Phe 116, Phe 123, Lys 125, Ala 131, Val 301 and Leu 302). The C4 hydroxy group of the catechol ring forms 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
two strong hydrogen bonds to the catalytic residues Tyr 49 and His 111 (C4 OH···Tyr 49 OH, 2.83 ; C4 OH···His 111 Ne2, 2.76 ). The short distance between the catechol ring and the nicotinamide moiety of NADP + (C4 O···NADP + C4N, 3.03 ) indicates a classical electrostatic interaction that is observed in many AKR1B1–inhibitor complex structures; this interaction increases the Lewis acidity of C4 OH and enhances the strengths of the two hydrogen bonds.[11] In contrast, the C3 hydroxy group cannot form hydrogen bonds with nearby residues. The catechol ring also forms parallel-displaced p–p stacking interactions with the side chain of Trp 21 (3.41 between the Trp21 Ne1 and catechol C6). The conjugated acrylic group of CAPE stacks with the phenyl ring of Phe 123 (3.36 between the Phe 123 Ce1 and acrylic group C), and forms van der Waals contacts with the side chains of Val 301 (3.75 between the Val 301 CG1 and CAPE O2) and Leu 302 (3.24 between Leu 302 CD2 and CAPE O1). The phenethyl tail of CAPE interacts with Phe 116, Phe 123, Ala 131 and Leu 302 through hydrophobic contacts in the specificity pocket. The hydrogen bonds between the inhibitors and the active site residues are critical for anchoring inhibitors of AKR1B1/ AKR1B10 in the catalytic pocket.[12] Superimposition of the AKR1B10/NADP + /CAPE structure and other AKR1B10 structures in complex with carboxylic acid type inhibitors reveals an overlap of the CAPE C4 OH oxygen atom and that of the carboxylic group (Figure S2 in the Supporting Information). Clearly, the C4 OH group of CAPE functions in a similar fashion to the carboxyl group of carboxylic-acid-type inhibitor. Thus, removal of the C4 OH eliminates most of the CAPE derivatives’ inhibition toward AKR1B10 and AKR1B1, but removal of the C3 OH has little effect.[7] Moreover, the p–p and hydrophobic interactions of the conjugated system and the phenethyl tail with the surrounding residues are synergistic with these hydrogen bonds. Replacement of the phenethyl tail with an ethyl group leads to a 14-fold loss in activity (ethyl caffeate: IC50 = 6.5 mm), while the much lower activity of caffeic acid (IC50 > 100 mm) apparently results from a total loss of hydrophobic interactions. This crystal structure provides a straightforward explanation for the dramatically improved AKR1B10/AKR1B1 selectivity of CAPE derivatives by the introduction of a 2-methoxy group on ChemMedChem 2014, 9, 706 – 709
707
CHEMMEDCHEM COMMUNICATIONS the catechol moiety,[7] represented by compound 9 a in Table 1. Our recent work revealed the critical role of Trp 112 (Trp 111) in the determination of inhibitor selectivity between AKR1B10 and AKR1B1, that is to say, an “AKR1B1-like” active site in AKR1B10 caused by inhibitor-induced Trp 112 flip and a broader active site of AKR1B10 provided by the native Trp 112 side chain orientation.[10] Because almost all AKR1B10 and AKR1B1 inhibitors involve essential hydrogen bonds formed between the inhibitors and the catalytic residues Tyr 49 and His 111 (corresponding to Tyr 48 and His 110 of AKR1B1), it is reasonable to assume that 9 a forms the same hydrogen bonds with AKR1B10. Thus, as shown in Figure 1 b, its 2-methoxy group would form an additional hydrophobic contact with the side chain of Leu 302 (3.2 ), explaining the increased activity exhibited by 9 a toward AKR1B10. On the other hand, a similar binding mode of 9 a with AKR1B1 would force the 2-methoxy group too close (2.0 ) to the side chain of Trp 111. Therefore, 9 a could likely rotate to avoid this unfavorable interaction, causing loss of the interactions between the phenethyl tail and the specificity pocket. These modeling data helps to explain why 9 a shows significant selectivity (554-fold) toward AKR1B10 over AKR1B1. This result strongly supports the critical role of Trp 112 (Trp 111) in the determination of inhibitor selectivity between these two enzymes.[10] When Soda et al. performed their docking study, the AKR1B10/NADP + /tolrestat crystal structure was the only available AKR1B10 structure. Since the protein is usually considered as a rigid body in molecular docking methodology and the conformation of its active site is critical to the docking results,[13, 14] it not surprising to see the disparity between their docking results and the crystal structure determined here. Our docking studies utilizing the structures of the AKR1B10–CAPE and AKR1B10–tolrestat complexes confirm that the former produces a similar global orientation of compound 9 a to the assumed one with a small RMSD value (< 1.3 ; Figure S3 in the Supporting Information), while the latter only produces chaotic orientations of the ligand. Thus, the AKR1B10–CAPE structure can serve as a more suitable receptor model for in silico analysis of CAPE derivatives as AKR1B10 inhibitors. CAPE efficiently inhibits the growth and metastasis of melanomas, colon, gastric, lung, prostate and breast cancers in both in vitro studies and in vivo animal models, although its stability in the body is still in dispute.[15] Its ability to regulate expression and activities of important physiological proteins was proposed to be responsible for its polypharmacology.[16] However, clear and direct analysis of CAPE interactions with its targets is rare. On the other hand, AKR1B10 inhibitors, such as known AKR1B1 inhibitors, flavonoids, nonsteroidal anti-inflammatory drugs, curcuminoids, xanthones, pentacyclic triterpenes, cholanic acids and C21-steroids, have been identified so far.[17] The binding models of these inhibitors with AKR1B10 were mostly developed by molecular docking methods using the AKR1B10/NADP + /tolrestat crystal structure.[17b, 18, 19] Because of the flexibility of the AKR1B10 active site and the structural diversity of its inhibitors, one should be very careful in employing docking methods for designing or screening new compounds. The exact inhibitor–AKR1B10 interactions might need 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org to be determined on a case-by-case basis using crystallographic methods.
Experimental Section Enzyme activity, crystallization, data collection, model building and docking: The expression, purification, enzyme activity assay and crystallization of aldo-keto reductase family member 1B10 (AKR1B10)[10] and aldehyde reductase (AKR1B1)[20] have previously been described elsewhere in detail. Caffeic acid phenethyl ester (CAPE) was purchased from the Guangdong Institute for Food and Drug Control (Guangzhou, P. R. China). All other compounds and reagents were purchased from Sigma–Aldrich. To obtain the AKR1B10/NADP + /CAPE complex structure, AKR1B10 was incubated with NADP + and CAPE before hanging drop setting. Molecular restriction of CAPE for structure refinement was prepared by using the Grade Web Server (http://grade.globalphasing.org). Coordinates and structure factors have been deposited in the Protein Data Bank with the entry code 4GQ0. Data collection and refinement statistics for all structures are shown in Table S1 of the Supporting Information. Docking of AKR1B10 with CAPE and with compound 9 a was performed using Discovery Studio 2.5 (Accelrys, Inc., San Diego, USA). The protocol was generated using the CDOCKER protocol without the crystal waters, and the parameters were set as default. Default protonation states were set at the neutral state (pH 7), and hydrogen atoms were added to the unoccupied valences of heavy atoms for the protein.
Acknowledgements The authors gratefully thank the Innovative R&D Team Leadership of Guangdong Province (P. R. China) (grant no.: 2011Y038). Keywords: anticancer agents · binding models · caffeic acid phenethyl ester derivatives · crystal structures · selectivity · aldose reductases [1] J. M. Jez, T. M. Penning, Chem.-Biol. Interact. 2001, 130, 499 – 525. [2] G. K. Balendiran, H.-J. Martin, Y. El-Hawari, E. Maser, Chem.-Biol. Interact. 2009, 178, 134 – 137. [3] B. Crosas, D. J. Hyndman, O. Gallego, S. Martras, X. Pars, T. G. Flynn, J. Farrs, Biochem. J. 2003, 373, 973 – 979. [4] C. Wang, R. Yan, D. Luo, K. Watabe, D.-F. Liao, D. Cao, J. Biol. Chem. 2009, 284, 26742 – 26748. [5] L. Zhong, H. Shen, C. Huang, H. Jing, D. Cao, Toxicol. Appl. Pharmacol. 2011, 255, 40 – 47. [6] a) J. M. Petrash, CMLS Cell. Mol. Life Sci. 2004, 61, 737 – 749; b) E. Pastel, J.-C. Pointud, F. Volat, A. Martinez, A.-M. LefranÅois-Martinez, Front. Pharmacol. 2012, 3, 148. [7] M. Soda, D. Hu, S. Endo, M. Takemura, J. Li, R. Wada, S. Ifuku, H.-T. Zhao, O. El-Kabbani, S. Ohta, K. Yamamura, N. Toyooka, A. Hara, T. Matsunaga, Eur. J. Med. Chem. 2012, 48, 321 – 329. [8] S. Son, E. B. Lobkowsky, B. A. Lewis, Chem. Pharm. Bull. 2001, 49, 236 – 238. [9] S. Endo, T. Matsunaga, M. Soda, K. Tajima, H. Zhao, O. El-Kabbani, A. Hara, Biol. Pharm. Bull. 2010, 33, 886 – 890. [10] L. Zhang, H. Zhang, Y. Zhao, Z. Li, S. Chen, J. Zhai, Y. Chen, W. Xie, Z. Wang, Q. Li, X. Zheng, X. Hu, FEBS Lett. 2013, 587, 3681 – 3686. [11] a) K. M. Bohren, J. M. Brownlee, A. C. Milne, K. H. Gabbay, D. H. T. Harrison, Biochim. Biophys. Acta Proteins Proteomics 2005, 1748, 201 – 212;
ChemMedChem 2014, 9, 706 – 709
708
CHEMMEDCHEM COMMUNICATIONS
[12] [13] [14] [15] [16] [17]
b) C. Giacomelli, F da Silva Miranda, N. S. GonÅalves, A. Spinelli, Redox Rep. 2004, 9, 263 – 269. O. El-Kabbani, A. Podjarny, Cell. Mol. Life Sci. 2007, 64, 1970 – 1978. X.-Y. Meng, H.-X. Zhang, M. Mezei, M. Cui, Curr. Comput.-Aided Drug Des. 2011, 7, 146 – 157. D. Plewczynski, M. Łaz´niewski, R. Augustyniak, K. Ginalski, J. Comput. Chem. 2011, 32, 742 – 755. N. Celli, L. K. Dragani, S. Murzilli, T. Pagliani, A. Poggi, J. Agric. Food Chem. 2007, 55, 3398 – 3407. S. Akyol, G. Ozturk, Z. Ginis, F. Armutcu, M. R. Yigitoglu, O. Akyol, Nutr. Cancer 2013, 65, 515 – 526. a) T. Matsunaga, Y. Wada, S. Endo, M. Soda, O. El-kabbani, A. Hara, Front. Pharmacol. 2012, 3, 5; b) M. Takemura, S. Endo, T. Matsunaga, M. Soda, H.-T. Zhao, O. El-Kabbani, K. Tajima, M. Iinuma, A. Hara, J. Nat. Prod. 2011, 74, 1201 – 1206; c) S. Endo, T. Matsunaga, H. Mamiya, C. Ohta, M.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org Soda, Y. Kitade, K. Tajima, H.-T. Zhao, O. El-Kabbani, A. Hara, Arch. Biochem. Biophys. 2009, 487, 1 – 9; d) M. Soda, S. Endo, T. Matsunaga, H.-T. Zhao, O. El-Kabbani, M. Iinuma, K. Yamamura, A. Hara, Biol. Pharm. Bull. 2012, 35, 2075 – 2080. [18] S. Endo, T. Matsunaga, M. Soda, K. Tajima, H.-T. Zhao, O. El-Kabbani, A. Hara, Biol. Pharm. Bull. 2010, 33, 886 – 890. [19] S. Endo, D. Hu, M. Suyama, T. Matsunaga, K. Sugimoto, Y. Matsuya, O. El-Kabbani, K. Kuwata, A. Hara, Y. Kitade, N. Toyooka, Bioorg. Med. Chem. 2013, 21, 6378 – 6384. [20] X. Zheng, L. Zhang, J. Zhai, Y. Chen, H. Luo, X. Hu, FEBS Lett. 2012, 586, 55 – 59.
Received: November 8, 2013 Published online on January 16, 2014
ChemMedChem 2014, 9, 706 – 709
709
Structural Basis for the Inhibition of AKR1B10 by Caffeic Acid Phenethyl Ester (CAPE) Liping Zhang, Hong Zhang, Xuehua Zheng, Yining Zhao, Shangke Chen, Yunyun Chen, Renwei Zhang, Qing Li,* and Xiaopeng Hu*[a] Caffeic acid phenethyl ester (CAPE), the major bioactive component of honeybee propolis, is a potent selective inhibitor of aldo-keto reductase family member 1B10 (AKR1B10), and a number of derivatives hold promise as potential anticancer agents. However, sequence homology between AKR1B10 and other members of the superfamily, including critical phase I metabolizing enzymes, has resulted in a concern over the selectivity of any potential therapeutic agent. To elucidate the binding mode of CAPE with AKR1B10 and to provide a tool for future in silico efforts towards identifying selective inhibitors, the crystal structure of AKR1B10 in complex with CAPE was determined. The observed interactions provide an explanation for the selectivity exhibited by CAPE for AKR1B10, and could be used to guide further derivative design.
Aldo-keto reductase family member 1B10 (AKR1B10), also called small-intestine aldose reductase,[1] is a promising antineoplastic target. This enzyme is abnormally overexpressed in many malignant tumors,[2] and has been found to interrupt cell differentiation induced by retinoid acids,[3] impair cell apoptosis mediated by intracellular toxic carbonyls and isoprenyl aldehydes,[4] and metabolize anticancer drugs such as daunorubicin and idarubicin.[5] A major difficulty in developing AKR1B10 inhibitors is the lack of selectivity between AKR1B10 and the homologous enzymes of the NADPH-dependent aldo-keto reductase (AKR) superfamily, which consists of around 150 critical phase I metabolizing enzymes that share high sequence similarities. A well-known case is the adverse effects caused by cross-inhibition with aldehyde reductase AKR1A1 (65 % sequence identity with AKR1B1), which was believed to be partially responsible for the clinical failures of aldose reductase (AKR1B1) inhibitors as type 2 diabetic complication drugs in the 1990s.[6] Because of the high sequence identity (71 %) shared between AKR1B10 and AKR1B1, an understanding of the factors controlling selectivity between these two enzymes is critical for studies of AKR1B10 inhibitors. In 2011, Soda et al. identified caffeic acid phenethyl ester (CAPE), the major bioactive component of honeybee propolis, [a] L. Zhang, H. Zhang, Dr. X. Zheng, Y. Zhao, Dr. S. Chen, Y. Chen, R. Zhang, Prof. Q. Li, Prof. X. Hu Centre for Cellular & Structural Biology School of Pharmaceutical Sciences, Sun Yat-sen University 132 East Circle, University City, Guangzhou 510006 (China) E-mail: [email protected] Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.201300455.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
as an efficient in vitro AKR1B10 inhibitor. Based on docking studies, they designed and synthesized several CAPE derivatives that exhibit much higher activity towards AKR1B10 and selectivity towards AKR1B1 (Table 1).[7] However, the proposed mechanism of inhibition of AKR1B10 by CAPE and its derivatives is not fully supported by the reported experimental results.[7, 8] Hydrogen bonds between the known AKR1B1/ AKR1B10 inhibitors and the active site residues (Tyr 49, His 111, etc.) are essential for inhibition, but the predicted hydrogen bonds of CAPE to AKR1B10 (C4 OH···Tyr 49 OH, 3.6 ; C4 OH···His 111 Ne2, 3.7 ) are too weak to efficiently anchor CAPE to the active site. Moreover, we recently identified significant inhibitor-induced conformation changes at the active site of AKR1B10,[10] which implies that the crystal structure (AKR1B10/ NADP + /tolrestat complex, PDB ID: 1ZUA) used by Soda et al.[7] in the docking studies might not be suitable to predict the interactions of CAPE derivatives with AKR1B10. To provide a solid structural basis for developing CAPEbased AKR1B10 inhibitors, here we report the 2.1 resolution crystal structure of the AKR1B10/NADP + /CAPE ternary complex (PDB ID: 4GQ0, see Table S1 in the Supporting Information for data collection and refinement statistics). This crystal structure not only shows that CAPE indeed forms strong classical hydrogen bonds via its catechol moiety with the active site residues of AKR1B10, but also clearly explains the higher activity towards AKR1B10 and excellent selectivity towards AKR1B1 of CAPE derivatives.[7] Thus, this structure might be a more suitable model for studies of interactions of CAPE derivatives with AKR1B10. Superimposing structures of the AKR1B10/NADP + /CAPE ternary complex and the AKR1B10 holoenzyme indicates that CAPE does not induce meaningful conformational changes at the active site, except for small movements of Phe 123 and Leu 302 that favor interactions between the phenethyl tail with the so-called specificity pocket (Figure S1 in the Supporting Information). This result is significantly different from the AKR1B10/NADP + /tolrestat crystal structure, in which tolrestat induces a Trp 112 side chain flip and movement of the Leu 302–Ser 304 segment, which closes the specificity pocket.[10] A schematic presentation of the AKR1B10/NADP + /CAPE ternary complex active site is shown in Figure 1 a. The electron density map clearly defines the planar conjugated system of the catechol ring and the acrylic linker (Figure S1 in the Supporting Information), as observed in the crystal structure of CAPE.[8] It is noteworthy that the catechol ring and the acrylic linker of CAPE are not coplanar in the docking model reported ChemMedChem 2014, 9, 706 – 709
706
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
Table 1. Reported IC50 values towards recombinant human AKR1B10 and AKR1B1 for caffeic acid phenethyl ester (CAPE) and other compounds. Entry
Compd R1
IC50 [mm][a]
Structure R2 R3
R4
AKR1B10
1
CAPE
(CH2)2Ph
H
OH
OH
0.5 (0.08[c])
2
Ethyl caffeate
CH2CH3
H
OH
OH
6.5
3
Caffeic acid
H
H
OH
OH
4
2
(CH2)2Ph
H
OH
H
5
3
(CH2)2Ph
H
H
OH
> 50 19[c]
Ratio[b] AKR1B1 4.1 (0.57[c])
3
> 50
–
73[c]
0.069
[c] [c]
6
9a
(CH2)2Ph
OMe
H
OH
0.013
7
Tolrestat
–
–
–
–
0.054[d]
8
18
4
2.2
[c]
31
7.2
[c]
554
0.014[d]
0.3
[a] IC50 values represent the mean of three determinations, where the standard error was less than 15 %. [b] Ratio of AKR1B1/AKR1B10 IC50 values. [c] Value determined by Soda et al. and reported in Ref. [7]. [d] Tolrestat values are taken from Ref. [[9]].
Figure 1. a) The crystal structure of AKR1B10 in complex with CAPE. Protein residues at 4 resolution are shown in grey, while the ligand is depicted in yellow. Hydrogen bonds are shown as red dashed lines. b) Model of 9 a bound to AKR1B10. The residue Trp 111 of AKR1B1 (purple) is superimposed, showing an unfavorable distance (2.0 , dashed yellow line) from the 2-methoxy group of 9 a. The hydrophobic contact between the CAPE 2-methoxy group and Leu 302 is also shown (dashed blue line, 3.2 ).
by Soda et al.[7] The catechol ring occupies the catalytic pocket (consisting of Tyr 49, His 111, and NADP + ) of AKR1B10, while the phenethyl tail is rotated about 608 to the conjugated plane, and is located in the hydrophobic specificity pocket (primarily consisting of Phe 116, Phe 123, Lys 125, Ala 131, Val 301 and Leu 302). The C4 hydroxy group of the catechol ring forms 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
two strong hydrogen bonds to the catalytic residues Tyr 49 and His 111 (C4 OH···Tyr 49 OH, 2.83 ; C4 OH···His 111 Ne2, 2.76 ). The short distance between the catechol ring and the nicotinamide moiety of NADP + (C4 O···NADP + C4N, 3.03 ) indicates a classical electrostatic interaction that is observed in many AKR1B1–inhibitor complex structures; this interaction increases the Lewis acidity of C4 OH and enhances the strengths of the two hydrogen bonds.[11] In contrast, the C3 hydroxy group cannot form hydrogen bonds with nearby residues. The catechol ring also forms parallel-displaced p–p stacking interactions with the side chain of Trp 21 (3.41 between the Trp21 Ne1 and catechol C6). The conjugated acrylic group of CAPE stacks with the phenyl ring of Phe 123 (3.36 between the Phe 123 Ce1 and acrylic group C), and forms van der Waals contacts with the side chains of Val 301 (3.75 between the Val 301 CG1 and CAPE O2) and Leu 302 (3.24 between Leu 302 CD2 and CAPE O1). The phenethyl tail of CAPE interacts with Phe 116, Phe 123, Ala 131 and Leu 302 through hydrophobic contacts in the specificity pocket. The hydrogen bonds between the inhibitors and the active site residues are critical for anchoring inhibitors of AKR1B1/ AKR1B10 in the catalytic pocket.[12] Superimposition of the AKR1B10/NADP + /CAPE structure and other AKR1B10 structures in complex with carboxylic acid type inhibitors reveals an overlap of the CAPE C4 OH oxygen atom and that of the carboxylic group (Figure S2 in the Supporting Information). Clearly, the C4 OH group of CAPE functions in a similar fashion to the carboxyl group of carboxylic-acid-type inhibitor. Thus, removal of the C4 OH eliminates most of the CAPE derivatives’ inhibition toward AKR1B10 and AKR1B1, but removal of the C3 OH has little effect.[7] Moreover, the p–p and hydrophobic interactions of the conjugated system and the phenethyl tail with the surrounding residues are synergistic with these hydrogen bonds. Replacement of the phenethyl tail with an ethyl group leads to a 14-fold loss in activity (ethyl caffeate: IC50 = 6.5 mm), while the much lower activity of caffeic acid (IC50 > 100 mm) apparently results from a total loss of hydrophobic interactions. This crystal structure provides a straightforward explanation for the dramatically improved AKR1B10/AKR1B1 selectivity of CAPE derivatives by the introduction of a 2-methoxy group on ChemMedChem 2014, 9, 706 – 709
707
CHEMMEDCHEM COMMUNICATIONS the catechol moiety,[7] represented by compound 9 a in Table 1. Our recent work revealed the critical role of Trp 112 (Trp 111) in the determination of inhibitor selectivity between AKR1B10 and AKR1B1, that is to say, an “AKR1B1-like” active site in AKR1B10 caused by inhibitor-induced Trp 112 flip and a broader active site of AKR1B10 provided by the native Trp 112 side chain orientation.[10] Because almost all AKR1B10 and AKR1B1 inhibitors involve essential hydrogen bonds formed between the inhibitors and the catalytic residues Tyr 49 and His 111 (corresponding to Tyr 48 and His 110 of AKR1B1), it is reasonable to assume that 9 a forms the same hydrogen bonds with AKR1B10. Thus, as shown in Figure 1 b, its 2-methoxy group would form an additional hydrophobic contact with the side chain of Leu 302 (3.2 ), explaining the increased activity exhibited by 9 a toward AKR1B10. On the other hand, a similar binding mode of 9 a with AKR1B1 would force the 2-methoxy group too close (2.0 ) to the side chain of Trp 111. Therefore, 9 a could likely rotate to avoid this unfavorable interaction, causing loss of the interactions between the phenethyl tail and the specificity pocket. These modeling data helps to explain why 9 a shows significant selectivity (554-fold) toward AKR1B10 over AKR1B1. This result strongly supports the critical role of Trp 112 (Trp 111) in the determination of inhibitor selectivity between these two enzymes.[10] When Soda et al. performed their docking study, the AKR1B10/NADP + /tolrestat crystal structure was the only available AKR1B10 structure. Since the protein is usually considered as a rigid body in molecular docking methodology and the conformation of its active site is critical to the docking results,[13, 14] it not surprising to see the disparity between their docking results and the crystal structure determined here. Our docking studies utilizing the structures of the AKR1B10–CAPE and AKR1B10–tolrestat complexes confirm that the former produces a similar global orientation of compound 9 a to the assumed one with a small RMSD value (< 1.3 ; Figure S3 in the Supporting Information), while the latter only produces chaotic orientations of the ligand. Thus, the AKR1B10–CAPE structure can serve as a more suitable receptor model for in silico analysis of CAPE derivatives as AKR1B10 inhibitors. CAPE efficiently inhibits the growth and metastasis of melanomas, colon, gastric, lung, prostate and breast cancers in both in vitro studies and in vivo animal models, although its stability in the body is still in dispute.[15] Its ability to regulate expression and activities of important physiological proteins was proposed to be responsible for its polypharmacology.[16] However, clear and direct analysis of CAPE interactions with its targets is rare. On the other hand, AKR1B10 inhibitors, such as known AKR1B1 inhibitors, flavonoids, nonsteroidal anti-inflammatory drugs, curcuminoids, xanthones, pentacyclic triterpenes, cholanic acids and C21-steroids, have been identified so far.[17] The binding models of these inhibitors with AKR1B10 were mostly developed by molecular docking methods using the AKR1B10/NADP + /tolrestat crystal structure.[17b, 18, 19] Because of the flexibility of the AKR1B10 active site and the structural diversity of its inhibitors, one should be very careful in employing docking methods for designing or screening new compounds. The exact inhibitor–AKR1B10 interactions might need 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org to be determined on a case-by-case basis using crystallographic methods.
Experimental Section Enzyme activity, crystallization, data collection, model building and docking: The expression, purification, enzyme activity assay and crystallization of aldo-keto reductase family member 1B10 (AKR1B10)[10] and aldehyde reductase (AKR1B1)[20] have previously been described elsewhere in detail. Caffeic acid phenethyl ester (CAPE) was purchased from the Guangdong Institute for Food and Drug Control (Guangzhou, P. R. China). All other compounds and reagents were purchased from Sigma–Aldrich. To obtain the AKR1B10/NADP + /CAPE complex structure, AKR1B10 was incubated with NADP + and CAPE before hanging drop setting. Molecular restriction of CAPE for structure refinement was prepared by using the Grade Web Server (http://grade.globalphasing.org). Coordinates and structure factors have been deposited in the Protein Data Bank with the entry code 4GQ0. Data collection and refinement statistics for all structures are shown in Table S1 of the Supporting Information. Docking of AKR1B10 with CAPE and with compound 9 a was performed using Discovery Studio 2.5 (Accelrys, Inc., San Diego, USA). The protocol was generated using the CDOCKER protocol without the crystal waters, and the parameters were set as default. Default protonation states were set at the neutral state (pH 7), and hydrogen atoms were added to the unoccupied valences of heavy atoms for the protein.
Acknowledgements The authors gratefully thank the Innovative R&D Team Leadership of Guangdong Province (P. R. China) (grant no.: 2011Y038). Keywords: anticancer agents · binding models · caffeic acid phenethyl ester derivatives · crystal structures · selectivity · aldose reductases [1] J. M. Jez, T. M. Penning, Chem.-Biol. Interact. 2001, 130, 499 – 525. [2] G. K. Balendiran, H.-J. Martin, Y. El-Hawari, E. Maser, Chem.-Biol. Interact. 2009, 178, 134 – 137. [3] B. Crosas, D. J. Hyndman, O. Gallego, S. Martras, X. Pars, T. G. Flynn, J. Farrs, Biochem. J. 2003, 373, 973 – 979. [4] C. Wang, R. Yan, D. Luo, K. Watabe, D.-F. Liao, D. Cao, J. Biol. Chem. 2009, 284, 26742 – 26748. [5] L. Zhong, H. Shen, C. Huang, H. Jing, D. Cao, Toxicol. Appl. Pharmacol. 2011, 255, 40 – 47. [6] a) J. M. Petrash, CMLS Cell. Mol. Life Sci. 2004, 61, 737 – 749; b) E. Pastel, J.-C. Pointud, F. Volat, A. Martinez, A.-M. LefranÅois-Martinez, Front. Pharmacol. 2012, 3, 148. [7] M. Soda, D. Hu, S. Endo, M. Takemura, J. Li, R. Wada, S. Ifuku, H.-T. Zhao, O. El-Kabbani, S. Ohta, K. Yamamura, N. Toyooka, A. Hara, T. Matsunaga, Eur. J. Med. Chem. 2012, 48, 321 – 329. [8] S. Son, E. B. Lobkowsky, B. A. Lewis, Chem. Pharm. Bull. 2001, 49, 236 – 238. [9] S. Endo, T. Matsunaga, M. Soda, K. Tajima, H. Zhao, O. El-Kabbani, A. Hara, Biol. Pharm. Bull. 2010, 33, 886 – 890. [10] L. Zhang, H. Zhang, Y. Zhao, Z. Li, S. Chen, J. Zhai, Y. Chen, W. Xie, Z. Wang, Q. Li, X. Zheng, X. Hu, FEBS Lett. 2013, 587, 3681 – 3686. [11] a) K. M. Bohren, J. M. Brownlee, A. C. Milne, K. H. Gabbay, D. H. T. Harrison, Biochim. Biophys. Acta Proteins Proteomics 2005, 1748, 201 – 212;
ChemMedChem 2014, 9, 706 – 709
708
CHEMMEDCHEM COMMUNICATIONS
[12] [13] [14] [15] [16] [17]
b) C. Giacomelli, F da Silva Miranda, N. S. GonÅalves, A. Spinelli, Redox Rep. 2004, 9, 263 – 269. O. El-Kabbani, A. Podjarny, Cell. Mol. Life Sci. 2007, 64, 1970 – 1978. X.-Y. Meng, H.-X. Zhang, M. Mezei, M. Cui, Curr. Comput.-Aided Drug Des. 2011, 7, 146 – 157. D. Plewczynski, M. Łaz´niewski, R. Augustyniak, K. Ginalski, J. Comput. Chem. 2011, 32, 742 – 755. N. Celli, L. K. Dragani, S. Murzilli, T. Pagliani, A. Poggi, J. Agric. Food Chem. 2007, 55, 3398 – 3407. S. Akyol, G. Ozturk, Z. Ginis, F. Armutcu, M. R. Yigitoglu, O. Akyol, Nutr. Cancer 2013, 65, 515 – 526. a) T. Matsunaga, Y. Wada, S. Endo, M. Soda, O. El-kabbani, A. Hara, Front. Pharmacol. 2012, 3, 5; b) M. Takemura, S. Endo, T. Matsunaga, M. Soda, H.-T. Zhao, O. El-Kabbani, K. Tajima, M. Iinuma, A. Hara, J. Nat. Prod. 2011, 74, 1201 – 1206; c) S. Endo, T. Matsunaga, H. Mamiya, C. Ohta, M.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org Soda, Y. Kitade, K. Tajima, H.-T. Zhao, O. El-Kabbani, A. Hara, Arch. Biochem. Biophys. 2009, 487, 1 – 9; d) M. Soda, S. Endo, T. Matsunaga, H.-T. Zhao, O. El-Kabbani, M. Iinuma, K. Yamamura, A. Hara, Biol. Pharm. Bull. 2012, 35, 2075 – 2080. [18] S. Endo, T. Matsunaga, M. Soda, K. Tajima, H.-T. Zhao, O. El-Kabbani, A. Hara, Biol. Pharm. Bull. 2010, 33, 886 – 890. [19] S. Endo, D. Hu, M. Suyama, T. Matsunaga, K. Sugimoto, Y. Matsuya, O. El-Kabbani, K. Kuwata, A. Hara, Y. Kitade, N. Toyooka, Bioorg. Med. Chem. 2013, 21, 6378 – 6384. [20] X. Zheng, L. Zhang, J. Zhai, Y. Chen, H. Luo, X. Hu, FEBS Lett. 2012, 586, 55 – 59.
Received: November 8, 2013 Published online on January 16, 2014
ChemMedChem 2014, 9, 706 – 709
709
