IDD388 Polyhalogenated Derivatives as Probes for an Improved

Subscriber access provided by La Trobe University Library

Article

IDD388 polyhalogenated derivatives as probes for an improved structure-based selectivity of AKR1B10 inhibitors Alexandra Cousido-Siah, Francesc X. Ruiz, Jindrich Fanfrlik, Joan Giménez-Dejoz, Andre Mitschler, Martin Kamlar, Jan Vesely, Haresh Ajani, Xavier Pares, Jaume Farrés, Pavel Hobza, and Alberto D. Podjarny ACS Chem. Biol., Just Accepted Manuscript • DOI: 10.1021/acschembio.6b00382 • Publication Date (Web): 30 Jun 2016 Downloaded from http://pubs.acs.org on July 7, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Chemical Biology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Chemical Biology

IDD388 polyhalogenated derivatives as probes for an improved structure-based selectivity of AKR1B10 inhibitors Alexandra Cousido-Siah †,‡, Francesc X. Ruiz †,‡,#,*, Jindřich Fanfrlík &,*, Joan Giménez-Dejoz #, André Mitschler †, Martin Kamlar ||, Jan Veselý ||, Haresh Ajani &,£, Xavier Parés #, Jaume Farrés #, Pavel Hobza &,£ and Alberto D. Podjarny †. †

Department of Integrated structural Biology, Institut de Génétique et de Biologie Moléculaire et Cellulaire, CNRS, INSERM, UdS, 1 rue Laurent Fries 67404 Illkirch CEDEX, France &

Institute of Organic Chemistry and Biochemistry (IOCB) and Gilead Science and IOCB Research Center, Academy of Sciences of the Czech Republic, Flemingovo nám. 2, 166 10 Prague 6, Czech Republic #

Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, E-08193 Bellaterra, Barcelona, Spain ||

Department of Organic Chemistry, Charles University in Prague, Hlavova 2030, 128 43 Prague 2, Czech Republic

£

Regional Centre of Advanced Technologies and Materials, Department of Physical Chemistry, Palacký University, Olomouc, 771 46 Olomouc, Czech Republic

ABSTRACT: Human enzyme aldo-keto reductase family member 1B10 (AKR1B10) has evolved as a tumor marker and promising antineoplastic target. It shares high structural similarity with the diabetes target enzyme aldose reductase (AR). Starting from the potent AR inhibitor IDD388, we have synthesized a series of derivatives bearing the same halophenoxyacetic acid moiety with increasing number of bromine (Br) atoms on its aryl moiety. Next, by means of IC50 measurements, X-ray crystallography, WaterMap analysis and advanced binding free energy calculations with a quantummechanical (QM) approach, we have studied their Structure-Activity Relationship (SAR) against both enzymes. The introduction of Br substituents decreases AR inhibition potency but improves it in the case of AKR1B10. Indeed, the Br atoms in ortho position may impede these drugs to fit into the AR prototypical specificity pocket. For AKR1B10, the smaller aryl moieties of MK181 and IDD388 can bind into the external loop A subpocket. Instead, the bulkier MK184, MK319 and MK204 open an inner specificity pocket in AKR1B10 characterized by a π-π stacking interaction of their aryl moieties and Trp112 side chain in the native conformation (not possible in AR). Among the three compounds, only MK204 can make a strong halogen bond with the protein (−4.4 kcal/mol, using QM calculations), while presenting the lowest desolvation cost among all the series, translated into the most selective and inhibitory potency AKR1B10 (IC50 = 80 nM). Overall, SAR of these IDD388 polyhalogenated derivatives have unveiled several distinctive AKR1B10 features (shape, flexibility, hydration), that can be exploited to design novel types of AKR1B10 selective drugs.

1. INTRODUCTION Cytosolic AKRs are mostly monomeric enzymes of approximately 37 kDa which fold into a typical and conserved (α/β)8 barrel. They belong to a rapidly growing enzyme superfamily and metabolize reactive aldehydes of lipid peroxidation, prostaglandins, retinoids, steroids, chemical carcinogens and drugs. Because of their broad substrate specificity and potential contribution in the cancer aetiology and chemoresistance, research in the structural and functional properties of mammalian AKRs has recently attracted considerable interest. Human enzymes include aldose reductase (AR, ALR2 or AKR1B1) and AKR1B10, sharing 71% sequence identity, but with

very different kinetic properties versus some relevant substrates such as retinaldehyde or glucose. In addition, a third human enzyme, AKR1B15 (92% sequence identity with AKR1B10), has been recently discovered. It is lowly expressed in placenta, testes and adipose tissues and, distinctly from most AKRs, is located in the mitochondrial fraction. AR has been extensively related to secondary diabetic complications and recently to cancer and some inflammatory diseases, while AKR1B10 is induced in several cancer types, especially hepatocellular carcinoma and non-small cell lung cancer correlated with smoking (1-3).

1 ACS Paragon Plus Environment


Page 2 of 19

Table 1. IC50 values of IDD388 halogenated derivatives

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

substituent position IC50 (µM)* Selectivity# # Inhibitor R2 R3 R4 R5 R6 AR AKR1B10 AR AKR1B10 MK181 H H Br H H 0.71 ± 0.07 4.5 ± 0.7 (3.7 ± 0.8) 6 (5) − IDD388 F H Br H H 0.40 ± 0.02 4.4 ± 0.5 (2.7 ± 0.2) 11 (7) − MK319 F F Br F F 0.30 ± 0.02 7.0 ± 0.8 (3.4 ± 0.4) 23 (11) − MK184 Br H Br H Br 21.4 ± 1.9 1.0 ± 0.1 (0.43 ± 0.04) − 21 (50) MK204 Br Br Br Br Br 21.7 ± 2.3 0.08 ± 0.01 (0.03 ± 0.005) − 271 (720) *IC50 values: mean ± SD of n = 2 independent experiments; # IC50 values vs AKR1B10 with 60mM glyceraldehyde and with 0.2mM pyridine-3-aldehyde (in brackets); idem for selectivity

Although over 100 crystal structures of AR and its complexes with inhibitors have been determined, only a dozen of structures of AKR1B10 are yet available (4-6). In this regard, Dr. Podjarny lab has contributed with the generation of a crystallization engineering form of AKR1B10, the methylated K125R/V301L AKR1B10 (i.e. AKME2MU), which improves the success in solving AKR1B10 holoenzyme-drug complexes (4, 7, 8). Halogens and halogen bonding have been recognized as being important in many areas of chemistry, biochemistry, and material science (9-13). Recently, we have joined efforts in combining the theoretical QM/MM approach with ultrahighresolution X-ray crystallography to the case study of AR and IDD388 derivatives to investigate the halogen bond (X-bond) tuning and the effect of X-bond to-hydrogen bond (H-bond) substitution (14, 15). In this work, we have taken advantage of the IDD388 inhibitor scaffold (also selective against another offtarget, aldehyde reductase or AKR1A1 (16)) and have synthesized a series of derivatives with different composition in Br atoms in the aryl moiety. The joint use of in vitro inhibition assays with recombinant protein, structural biology and computational techniques has allowed us to identify an unexpected induced-fit binding to AKR1B10 by some of the designed drugs, and to assert the importance of the shape complementarity, holoenzyme hydration and proper X-bonding interactions in the inhibitor binding strength. Besides, AR and AKR1B10 present some differences in these aspects, which might help into the development of new programs of structure-based drug design of potent selective AKR1B10 inhibitors targeting cancer. 2. RESULTS AND DISCUSSION 2.1. Synthesis and IC50 determination Based on the previous structural knowledge on AR and AKR1B10, one of the notable differences between these very similar enzymes is that the active site of the latter is broader (5, 17, 18). Hence, we tested the previously described compounds IDD388, MK181 and MK319 against AKR1B10 and additionally synthesized the bulkier MK184 and MK204 (see Methods section) and tested them with both enzymes to study the effect of the increased aryl moiety volume on binding and selectivity to both proteins.

Figure 1. Structural formula of the IDD388 halogenated derivatives.

Our next step was to determine their IC50 values (Figure 1 and Table 1). The IC50 values vs AKR1B10 in Table 1 were performed using 60 mM glyceradehyde and 0.2mM pyridine-3-aldehyde as substrates, giving qualitatively similar values (Table 1). We will use as a reference the former from now on. IDD388, MK181 and MK319 values against AR were below 1 µM, as previously reported (14), and the addition of more Br atoms in MK184 and MK204 provoked a dramatic increase in their IC50 against AR, both over 20 µM. Conversely, the values obtained against AKR1B10 followed an opposite trend. IDD388, MK181 and MK319 values were over 1 µM, while the bulkier MK184 and MK204 inhibited AKR1B10 with 1.0 and 0.08 µM values, respectively. 2.2. Structure-activity relationships (SAR) with AR and the IDD388 derivatives Originally, the IDD388 derivatives were designed and synthesized to analyse the effect of different aryl substitution patterns in the binding of the well-characterized AR specificity pocket. IDD388, MK181 and MK319 (potent ARIs) differ in their number of fluorine atoms (F), which tune the X-bond potency between their para Br and AR Thr113 Oγ1 atom. However, both MK184 and MK204 have been shown here as low affinity ARIs (Table 1) and thus we were not able to observe them in X-ray structures. Hence, we have superimposed both compounds with MK319 complexed with AR holoenzyme (Figure 2A-C). In both cases, the Br atoms in ortho position facing loop C would clash with Ala299 Cα and C ( 0.4 Å per thousand atoms) (19) and van der Waals radii of Br and C atoms (20), a clash will probably occur below the indicated distance). It has to be noted that three main conformations have been described for ARI binding, represented by the sorbinil-, tolrestat-, and IDD594-bound states (21). The last one is the corresponding to all the IDD388 derivatives (14, 15) and it has the π-π stacking interaction of the aryl moiety and Trp111 side chain as common denominator. Leu300 is also locked by a hydrophobic interaction with the aryl moiety, which in turn keeps Ala299 fixed in the position observed in Figure 2. Thus, the illustrated steric hindrance (as anticipated) might be the likely cause for the low potency of MK184 and MK204 against AR, as we have reported recently for JF0049, another polybrominated selective AKR1B10 inhibitor (i.e. poor AR binder) (8).


Page 3 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


2.3. X-ray structures of AKME2MU holoenzyme complexed with the IDD388 derivatives: observation of a novel AKR1B10 binding site conformer Regarding AKR1B10, the situation was quite reversed, as the bulkier aryl moieties were preferred for strong AKR1B10 inhibition (Table 1), fitting with its broader active site. In order to understand the rationale behind, we attempted to co-crystallize the IDD388 derivatives with AKME2MU, whose kinetic properties were indistinguishable from those of the unmodified wild-type AKR1B10, as also reported before (4, 22). We succeeded in obtaining the AKME2MU holoenzyme complexed with all the IDD388-derived compounds. Table 2 shows the data collection and refinement statistics and Fo-Fc omit maps allow to unambiguously observe them (Figures 3A-3E). To

note that for AKME2MU holoenzyme complexes with MK319 and MK204, two molecules of the inhibitor are bound at the same time in the active site (Figures 3C and 3E). The first molecule is positioned with the carboxylic acid binding to the anion-binding site, while the second one is inverted i.e. the chlorophenoxyacetic acid moiety of the first molecule is performing a π–π stacking interaction with the halogen-substituted benzyl moiety of the second molecule (about 3.4 Å distance between planes, data not shown). This had already been observed for the ARIs alrestatin and tolrestat with AR (23, 24); however, mass spectrometry studies indicated (at least for tolrestat) that the most abundant conformation in solution is the case of only one bound ligand, at 10-fold excess (25) (in the obtained complexes, there was a 10-fold and 12-fold excess, respectively).

Table 2. Data collection and refinement statistics AKME2MU-MK181

AKME2MU-IDD388

AKME2MU-MK319

AKME2MU-MK184

AKME2MU-MK204

PDB ID

4XLJ

4XLK

4XM9

4XLM

4XMJ

Inhibitor concentration (mM)

10

10

8

5

5

Wavelength (Å)

1.54178

1.54178

1.54178

1.54178

1.54178

Resolution range (Å)

50 - 2.05 (2.12 - 2.05)

50 - 1.75 (1.81 - 1.75)

50 - 1.75 (1.81 - 1.75)

50 - 1.95 (2.02 - 1.95)

50 - 2.05 (2.12 - 2.05)

Space group

P 31

P 31

P 31

P 31

P 31

78.66 78.66 48.22

79.39 79.39 50.09

79.59 79.59 49.85

79.56 79.56 50.06

79.46 79.46 49.42

Unit cell (Å) and (°)

90 90 120

90 90 120

90 90 120

90 90 120

90 90 120

Total reflections

104630

161256

168955

185172

91769

Unique reflections

20884

35928

33155

24794

21542

Multiplicity

5.0 (3.9)

2.5 (2.1)

2.6 (1.7)

7.5 (4.2)

2.3 (2.2)

Completeness (%)

99.7 (98.5)

88.1 (61.5)

92.4 (81.1)

96.0 (85.4)

90.6 (83.9)

Mean I/sigma(I)

24.58 (3.17)

15.36 (2.78)

19.03 (3.32)

19.53 (3.21)

9.95 (1.92)

Wilson B-factor

34.63

23.69

21.77

26.78

34.78

R-sym^

0.058 (0.422)

0.058* (0.331)

0.05* (0.203)

0.089 (0.384)

0.079* (0.413)

R-factor ^^

0.2002 (0.2179)

0.2066 (0.2501)

0.1778 (0.1222)

0.1948 (0.2201)

0.1813 (0.1921)

0.2471 (0.2714)

0.2536 (0.2215)

0.2086 (0.1626)

0.2433 (0.2946)

0.2216 (0.2552)

Number of atoms

2771

2853

2851

2813

2801

macromolecules

2598

2598

2573

2676

2646

ligands

79

80

110

81

114

water

93

170

160

53

40

Protein residues

316

316

316

316

316

RMS(bonds)

0.012

0.011

0.011

0.010

0.012

RMS(angles)

1.38

1.31

1.41

1.48

1.53

Ramachandran favored (%)

96

98

98

96

97

Ramachandran outliers (%)

0.31

0

0

0.31

0

Clashscore

11.04

7.03

7.95

10.05

9.52

Average B-factor

R-free

#

31.30

24.10

32.60

33.20

38.90

macromolecules

31.50

24.00

32.30

33.30

39.00

solvent

31.00

26.30

37.00

29.50

35.70

Statistics for the highest-resolution shell are shown in parentheses. *Anomalous pairs were treated separately. ^Rsym=Σ|I‘I’|/ΣI, for which I=observed intensity and ‘I’=statistically weighted average intensity of multiple observations of symmetry-related reflections. ^^Rfactor=Σ||Fo| − |Fc||/Σ|Fo|, for which Fo=observed structure factor amplitudes and Fc=calculated structure factor amplitudes. #Rfree: same definition as that for Rfactor for a cross-validation set of ~5^% of the reflections.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 19

Figure 2. Structural analysis of AR-NADP+-IDD388 derivative complexes. A) Atomic representation of the X-ray structure of ARMK319 active site (PDB ID 4LB4, MK319 in violet sticks), B) AR-MK184 (cyan sticks) and C) AR-MK204 (yellow sticks) superimposed to MK319 coordinates in PDB ID 4LB4. The protein residues and cofactor are in sticks (white and orange respectively). The short distances noting steric hindrance are displayed in red dashed lines and red labels, while allowed fluorine-protein distances are in yellow dashed lines.

Figure 3. X-ray structures of the complexes of AKME2MU-IDD388 derivatives. Atomic representations including the inhibitor electron densities, shown as σA-weighted Fo–Fc omit maps contoured at 2σ level in a cyan mesh: A) AKME2MU-MK181, B) AKME2MUIDD388, C) AKME2MU-MK319, D) AKME2MU-MK184 and E) AKME2MU-MK204. The protein is displayed in white cartoon, with the active site residues shown with white lines and the cofactor with orange sticks. In F), the X-ray structures are superimposed, focusing in the inhibitors (same color code as from A to E) and the active site residues Trp112, Gln114 and Ser304 and NADP+ (in lines and in orange for the cofactor and in the same color as the corresponding inhibitor for the amino acids). Trp112 and Gln114 native and flipped conformations are noted.


Page 5 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The comparison of all AKME2MU-ligand structures herein reported shows that MK181 and IDD388 bind in an extended form (Figure 4A), while MK319, MK184 and MK204, instead, bend over and perform a face-to-face oriented π–π stacking with the Trp112 indole ring (Figure 3F). In these complexes, Trp112 side chain adopts the “native conformation” of the holoenzyme structure (5), perpendicular to the “flipped conformation” present in the AKR1B10 complex with tolrestat (17). Trp112 native conformation in AKR1B10 is stabilized by a specific Hbond network between the former, Gln114 and Ser304

(Figure 3F). Indeed, this was already described for other AKR1B10-inhibitor complexes and shown as an important selectivity feature in regard to AR (5, 6). However, the complexes with MK319, MK184 and MK204 constitute an unprecedented conformation of the AKR1B10 binding site, as the stacking with Trp112 is happening with the native conformation. Besides, as shown, MK184 and MK204 can only bind deep in the active site of AKR1B10, but not of AR, which make them selective for the first.

Figure 4. Structural analysis of the AKME2MU-IDD388 derivatives complexes. Atomic representations of the AKME2MU holoenzymes complexed with A) MK181, B) MK319, C) MK184 and D) MK204. The active site residues and the cofactor are displayed in white and orange sticks, respectively. The inhibitors are displayed as indicated in Figure 3. Water molecules (W, red spheres) and active site residues are labeled, with mutant residues in italics. Short halogen...O contacts and H-bonds (green dashed lines) distances are indicated in Å. Trp112 conformation is specified.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 19

Figure 5. Superimposition of the AKME2MU-MK204 X-ray structure with the three main AKR1B10 active site conformers. Atomic representation of the AKME2MU-MK204 X-ray structure following the same guidelines than in Figure 4 superimposed to the complex A) AKR1B10-tolrestat (inhibitor and protein residues in orange and pale yellow sticks, respectively), B) AKR1B10-zopolrestat (inhibitor and protein residues in yellow and light brown sticks, respectively), and C) AKR1B10-CAPE (inhibitor and protein residues in brown and chocolate sticks, respectively). The cofactor is displayed in orange sticks in all cases. Residues that have the same conformation in both complexes are labeled in black, while residues colored as the superimposed protein residues are those presenting a different conformation than the AKME2MU-MK204 X-ray structure.

The interactions with the halophenoxyacetic acid moiety are similar in all the complexes showed herein (Figure 4), including H-bonds with residues Tyr49 and His111. In the case of MK181 (Figure 4A), there is also an interstitial water molecule mediating a H-bond network with Trp112 in flipped conformation (not possible for the remaining ligands). To note that in the complexes with MK181 and IDD388, Trp112 and Gln114 are observed in both native and flipped conformations (Figure 3F). Considering that the occupancy of these two inhibitors is 70% and 78% respectively, we might be observing coexistence in those crystals between the holoenzyme (with Trp112 and Gln114 native conformations) and the ternary complexes with the inhibitors (with Trp112 and Gln114 flipped conformations). The interactions with the aryl moiety clearly differ between MK181 and IDD388 and the rest of derivatives (Figure 4B-4D). The two first display hydrophobic interactions with the mutated Leu301 (Val301 in the wild-type) and Phe123 (Figure 4A, with equivalent binding geometry for IDD388, data not shown), while their Br atom has a short Br…O contact with the O of the main chain of Pro124 (3.5 Å). Regarding the remaining compounds, all have two short halogen…O contacts (possible X-bonds, see also 2.4.4.). While MK319 and MK204 have short halogen…O contacts with Cys299 main chain O and Ser304 side chain O (Figures 4B and 4D), MK184 aryl moiety is slightly rotated in regard to them and makes short Br…O contacts with Tyr210 and Asn300 (Figure 4C). This is due to the different conformation of the halophenoxyacetic acid moiety in MK319 and MK204, which present a second inhibitor molecule bound in the crystal (the halophenoxyacetic acid moiety of MK184 is superimposing to the ones of MK181 and IDD388, Figure 3F). In any case, the aryl moiety stacking with Trp112 and the two short halo-

gen…O contacts are conserved in both conformations, allowing us to extract SAR of these X-ray structures. Up to date, three main AKR1B10 binding-site conformers have been described: tolrestat-, zopolrestat- and caffeic acid phenethyl ester (CAPE)-bound (5, 6, 8, 17). The superimposition of this unprecedented AKR1B10 binding site conformer (observed in the complexes with MK319, MK184 and MK204) with the previously reported shows its novelty. The AKR1B10-tolrestat conformer (Figure 5A) differs with it because of a more closed loop A subpocket around the inhibitor, between Val301, locked by the Trp112 flipped conformation (along with Gln114), and Phe123 (more open in the novel conformer). To note that the polybrominated compound JF0049 is binding to this conformer (8) and not to the novel one described here. The AKR1B10-zopolrestat conformer (Figure 5B) shows a very similar conformation to its AR cognate (data not shown), with the benzothiazole ring involved in a stacking interaction with the Trp112 flipped side chain (opposite to the native conformation of the novel AKR1B10 conformer), with the single difference that Thr113 and its interaction with the trifluoromethyl moiety of zopolrestat (in AR) is absent in AKR1B10, with Gln114 moving away from the active site. This is the main difference with the novel AKR1B10 conformer, but not the only one: Phe123 is in a more closed conformation and Val301 side chain is oriented into an opposite direction. Finally, the AKR1B10CAPE conformer (Figure 5C) is the most similar one to the novel conformer, albeit with some differences. The most intuitive difference is that MK204, MK184 and MK319 address a deep region of the active site and CAPE targets the more external loop A subpocket. Indeed, the CAPE conformer has a slightly more closed conformation (see residues Trp220, Lys125 and Phe123 in Figure 5C). But


Page 7 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


most importantly, MK204 and congeners provoke the movement of Leu302 away from the active site and do not need to induce the shift of residue 301 side chain, as the loop A subpocket remains unoccupied in the novel AKR1B10 conformer. 2.4. SAR with AKR1B10 and the IDD388 derivatives In this series of IDD388 derivatives, a general SAR trend can be observed with AKR1B10. The binding geometry of MK181 and IDD388 provides a lower affinity than the binding geometry of MK204 and MK184. Nevertheless, MK319 presents a nearly equivalent binding geometry to that of MK204 and is presenting the worst affinity for AKR1B10 to that of the former. Therefore, the current compound series is an interesting example of the complexity to discern the key interactions for molecular recognition and the contribution to the affinity of the observed interactions in the crystal structures. Even the same interaction, given different contexts, is able to provide a different energetic contribution (9). In order to tackle this issue, we have performed a deep analysis of the

SAR data, taking advantage of several novel computational approaches that allowed us to deconvolute the attractive and repulsive interactions involved in the affinity of this series of ligands versus AKR1B10. 2.4.1. Shape complementarity of the aryl moiety with the loop A subpocket. The shape complementarity of a ligand filling a pocket is critical for a proper binding (9, 26, 27). In order to analyse the distinct active site pocket architectures obtained with the five AKME2MU-inhibitor complexes determined, we ran them in the DoGSiteScorer server (http://dogsite.zbh.uni-hamburg.de/, available in Figure 6 and in Figure S1, Table S1) (28). Figure 6 shows the inhibitors and a surface representation of the calculated pockets along with the protein fold in cartoon. It can be seen that, regarding the IDD388 and its derivatives, only MK181 and IDD388 can fill the loop A subpocket characteristic of AKR1B10 (4, 18) (and not present in AR with the exception of AR-tolrestat, see Table S1 in SI). In fact, their aryl moieties have volumes lower than 130 Å3 and the volume of this subpocket for this series of ligands is ~190 Å3 (Table 2). MK319, MK184 and MK204 aryl moie-

Figure 6. Pocket conformation of the X-ray determined AKME2MU-inhibitor complexes. The five different AKME2MUinhibitor complexes are displayed in the same exact orientation (top view with loop B in the left and loop A in the right), with the protein in white cartoon and the inhibitor in sticks following the coloring of Figure 3. The surfaces (displayed in aquamarine color) are the resultant from each one of the DogSiteScorer automatically detected pockets, which graphically depict the active-site pocket surface.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ties are too bulky (> 70% of the subpocket volume) and hence their binding deep into the active site stacking to Trp112 native conformation may be favoured (Figures 4B, 4C and 4E). While MK181 and IDD388 fit properly into the active site of AKR1B10 (Figures 6A and 6B), their IC50 values remain around 5 µM, which indicates that the observed interactions might be cancelled by other factors a priori not evident. One factor that might cause that is the suboptimal filling of the loop A subpocket by these compounds. Indeed, a hydrophobic pocket optimal filling is often achieved long before the van der Waals limit is reached (9). For synthetic host-guest complexes, it has been empirically established that optimal binding is observed when the guest occupies about 55 ± 9% of the volume of the host and this has been observed once for a series of ligands binding to plasmepsin II (29). DogSiteScorer computes it through the poc_cov parameter (percentage of the pocket covered by the co-crystallized ligand, see SI). MK181 and IDD388 present poc_cov of 68% and 65% respectively and hence will not follow the so-called “55% rule”. Of the described AKR1B10 binding site conformers, only AKR1B10-CAPE is suitable to compare with the current series. The AKR1B10-zopolrestat conformer does not present this subpocket separated (see Table S1 in SI) and AKR1B10-tolrestat, though having it, shows the ligand barely occupying the subpocket (Figures S1 and S2 in SI). CAPE bound to AKR1B10 fulfills the “55% rule”, with a poc_cov of 49% for the loop A subpocket. With the bulkier ligands in the series, given that are placed in a much broader pocket shared with the cofactor, the comparison is also difficult to make. In any way, Figure S2 is graphically depicting the importance of shape complementarity for a proper binding in the case of loop A subpocket. 2.4.2. Solvation of AKR1B10 holoenzyme and of AKME2MU ternary complexes: WaterMap analysis. Displacement of long-residence water molecules from the protein active site plays an important role in ligand binding and is an important source of binding free energy. Long-residence water molecules in protein active sites are usually energetically unfavourable. This is due to the orientation and positional constraints imposed by the protein surface and also by their inability to form a full complement of hydrogen bonds when solvating the protein surface (30). In this regard, AKR1B10 presents the aforementioned loosely packed loop A subpocket, able to trap a water molecule, described in the AKR1B10 V301L-NADP+-fidarestat complex (and also observed in other ternary complexes, PDB IDs 1ZUA, 4JII, 4WEV), differing in AR, which presents a well-packed hydrophobic subpocket unable to accommodate any water molecule (18). Even the AR-tolrestat complex, with the most open conformer of Phe122 for AR and presenting the only appreciable loop A subpocket, would have short contacts in the presence of the water molecule observed in the AKR1B10 cognate (data not shown). Besides, the AKR1B10 holoenzyme (PDB ID 4GQG), and the herein described AKME2MU-MK319 and AKME2MUMK204 complexes, present this buried water molecule.

Page 8 of 19

The non-observation of other surrounding water molecules might be due to the fact that they might be highly mobile and disordered, pointing to an imperfectly hydrated hydrophobic pocket, which can lead to large enthalpic gains when a ligand displaces such water molecules (31), as recently shown in JF0049 binding to AKR1B10 (8). Apart from the cited water molecule, a comparison of the binding sites of MK319, MK184 and MK204 shows the decrease of the number of water molecules with the increasing number of Br atoms, probably related to the stronger hydrophobicity of Br over F atoms (Figure S3). Table 3. Thermodynamic signature (in kcal/mol) and average number of hydrogen bonds (#HB) of active-site water molecules Water molecules (PDB numbering)

∆GSolv

∆HSolv −T∆SSolv

#HB (PW)

#HB Occupancy (WW)

W*

1.79

−1.81

3.60

0.8

1.6

0.92

W626

4.24

0.99

3.25

0.9

1.3

0.90

W673

6.58

3.53

3.05

0.9

1.1

0.87

W707

8.24

4.8

3.44

0.7

1.0

0.87

W596

4.63

0.01

4.62

0.7

1.0

1.00

W767

4.93

3.61

1.32

0.8

1.5

0.47

W779

2.22

0.33

2.55

0.8

1.9

0.51

Data for the AKR1B10 holoenzyme structure (4GQG) determined by WaterMap. P and W stand for protein and water. The aforementioned buried water molecule (W707) is highlighted in bold. * Water molecule not found in the X-ray structure.

With the obtained experimental data, however, it was difficult to properly weight the importance of the solvation in the inhibitor binding. In order to have further insight, we analyzed all the long-residence water molecules of the AKR1B10 holoenzyme structure and the herein described AKME2MU-ligand structures using WaterMap (30). WaterMap estimates the free energy of water molecules expelled into bulk solvent when a ligand is binding to a protein. The free energy can be decomposed into enthalpic and entropic contribution. The solvation free energy of unfavourable water molecules is positive and is released upon the ligand binding. The information of the AKR1B10 holoenzyme (PDB ID 4GQG) is summarized in Table 3 and Figure 7. In the holoenzyme structure, WaterMap found the aforementioned buried water molecule (W707) close to Phe123, which was very unfavourable in this position (ΔG of about +8 kcal/mol). Another slightly unfavourable water molecule (W*) was found close to Lys125. Finaly, four long-residence water molecules were found by WaterMap in the vicinity of Tpr112 (W596, W626, W673 and W767). These four water molecules were unfavourable (ΔG of about +4 to +7 kcal/mol). WaterMap calculations on protein-ligand (P-L) complexes showed that the W707, W*, W596, W626 and W673 water molecules were replaced when either MK181 or IDD388 was bound. W596, W626 and W673 were re-


Page 9 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


leased upon MK181 or IDD388 binding even though the inhibitor aryl moieties were bound to a different binding subpocket. This was caused by the changed conformation of Tpr112 and few neighbouring residues.

release W626, W673 and W767 but not W596 due to the smaller size of the inhibitor (see also Figure 6B), even though W596 became more unfavourable (ΔGSolv of +7.49 kcal/mol and −TΔSSolv of +2.11 kcal/mol). Additionally, WaterMap found a water molecule with long-residence

Figure 7. WaterMap analysis. A) WaterMap was used to infer positions of the long-residence waters of the holoenzyme structure of AKR1B10 (PDB ID 4GQG) (larger spheres, colored from red to green (unfavourable to favourable free energy, as indicated in B)). Smaller red spheres show the X-ray water positions. The protein is in green cartoon and the active-site residues and NADP+ are displayed in grey sticks (C atoms); B) Superimposition of the AKR1B10 holoenzyme unfavourable long-residence water molecules (spheres coded as in A)) with the AKME2MU ternary complexes with MK319 (flesh sticks), MK184 (violet sticks) and MK204 (pink sticks). Green dashed lines represent contacts ≥ 2.5 Å, while red dashed lines correspond to contacts ≤2.0 Å. The distance between W596 and Br of MK204 is about 1.8 Å, between W596 and Br of MK184 is 1.6 Å, while with W596 and F of MK319 is 2.5 Å.

Moreover, W767 was not released when MK181 or IDD388 was bound according to the WaterMap simulations: specifically, WaterMap found single water molecules buried in the area between the enzyme and the ligand in the complexes of MK181 (ΔGSolv of +1.96 kcal/mol, −TΔSSolv of +4.60 kcal/mol and occupancy of 0.96) and IDD388 (ΔGSolv of +2.77 kcal/mol, −TΔSSolv of +4.90 kcal/mol and occupancy of 0.98). The water molecule between MK181 and AKME2MU corresponds to W42 in the AKME2MU-MK181 crystal structure (PDB ID 4XLJ). However, there is not such a bridging water molecule in the AKME2MUIDD388 (PDB ID 4XLK) crystal structure. W5 found in this crystal structure is in conformation B, pairing with Trp112 native conformation, presumably before IDD388 binding. The polyhalogenated compounds (MK319, MK184 and MK204)) are fitting into the novel binding site conformer. WaterMap calculations on the P-L complexes showed that the four aforementioned unfavourable longresidence water molecules (W596, W626, W673 and W767) bound in this area were released when MK184 or MK204 were bound. Additionally, there was a crystallographic water molecule (W23) close to MK184 in the structure of the AKME2MU-MK184 complex. W23 was, however, not found by WaterMap. Finally, WaterMap analysis was performed on the AKME2MU-MK319 complex. The result showed that the binding of MK319 only

time in the area between the enzyme and MK319 in the complex: specifically W36 (numbering from PDB ID 4XM9 structure; ΔGSolv of +5.32 kcal/mol and −TΔSSolv of +4.06 kcal/mol). 2.4.3. Scoring of the inhibitor binding to AKR1B10. As similarly done with AR (14, 15), a scoring was performed using advanced binding free energy calculations involving a QM/SQM approach, this time in combination with the WaterMap results aforementioned, i.e. the entropic term of WaterMap was added to the −TΔSSolv scoring term (the enthalpic part is reliably described by the QM/SQM approach used). The entropy of the released water molecules was subtracted (W*, W707, W596, W673 and W626 for MK181 and IDD388; W596, W673, W626 and W767 for MK184 and MK204; W673, W626 and W767 for MK319), while the entropy of water molecules interacting with the ligand was added (change of entropy of W596 and entropy of W36 was added for MK319; change of entropy of W767 was added for MK181 and IDD388). The wild-type (WT) enzyme was used by means of the in silico mutation of R125 and L301 in the crystal structures to the WT K125 and V301 residues. These mutations and the subsequent QM/SQM optimization did not, however, change the binding mode significantly. The largest difference between the X-ray crystal structures and their cognate QM/SQM optimized WT structures were found for the MK181 and IDD388 inhibitor complexes (Table S2 in SI).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

However, even here, the binding mode was conserved (RMSD below 2 Å; see also Figure S4 in SI). The calculated scores are summarized and compared with the experimental IC50 values in Graph 1 and Table 4. The computational results are overall in good agreement with the experimental values (coefficient of determination (R2) of 0.92 and predictive index (PI) of 0.94). MK204 and MK184 have a more negative score than MK181, IDD388 and MK319 (−92.5 and −83.8 vs. −81.2, −76.4 and −78.3 kcal/mol, respectively). The scores of MK204 and MK184 are more negative than scores of MK181 and IDD388 due to the ΔΔGsolv term. It can be caused by properties of the MK184 and MK204 compounds (the higher number of halogen atoms) and also because they bind into a more hydrophobic part of the enzyme than MK181 and IDD388. Effectively, MK184 and MK204 aryl moieties bind in a deeply buried area of the active site performing a strong hydrophobic interaction with Trp112 (Figure 3F, 4C and 4D). On the other hand, MK181 and IDD388 also displace W707 and W*. Moreover there is another long-residence water molecule interacting with MK181 (W42, figure 4A). Table 4. Scoring values *Inh.

ΔEint

ΔΔGsolv ΔGconfw(L) −TΔSSol Score ΔGbo v

IC50 (µM)

MK181

−127.1

60.5

0.0

−14.7

−81.2

-7.7

4.5

IDD388

−126.9

64.4

0.5

−14.4

−76.4

−7.7

4.4

MK319

−111.4

34.8

4.5

−6.2

−78.3

−7.4

7.0

MK184

−113.8

40.8

1.4

−12.2

−83.8

−8.6

1.0

MK204

−117.6

34.9

2.5

−12.2

−92.5

−10.1

0.08

*Inh.=Inhibitor. Calculated Gas-Phase Interaction Energy (ΔEint), Interaction Desolvation Free Energy (ΔΔGsolv), Change of Ligand Conformational Energy (ΔGconfw(L)), Entropy (−TΔSSolv) of expelled buried water molecules determined by WaterMap and ΔGbo (calculated by ΔGbo = RT ln(IC50/2)) and Experimental IC50 values (in grey). ΔGbo refers to free energy for the reaction under standard conditions of all reactants and products at a concentration if 1.0 M and it was calculated from the Experimental IC50 values. All values, except IC50 values, are in kcal/mol. The score and the IC50 values are highlighted in bold and with grey filling of the column.

Graph 1. Calculated scores plotted vs. experimental ΔGbo (all in kcal/mol)

Page 10 of 19

The comparison of MK184 and MK204 with MK319 is more complicated. Given the pharmacological concentrations expected in solution, we can assume that one molecule of MK319 will bind to the AKR1B10 holoenzyme in this state (the same can be hypothesized for MK204), as already discussed here. MK319 binds to the same part of the MK184 and MK204 and it does not displace W707 and W* either. However, it also does not displace W596, which is bound in the novel active site (Figure 6B). Finally, another unfavourable water molecule (W36) was found in the complex. Thus, a 6 kcal/mol entropy penalty might apply to it. Moreover, it presents the lowest ΔEint and the highest ΔGconfw(L) in the series. Finally, MK204 is the best binder (lowest IC50 value) and it also has the most negative score. Its score is more negative than the score of MK184 due to the more negative ΔEint and smaller ΔΔGsolv terms. To note that, in the above calculations, we used W4 in the IDD388-AKME2MU complex (described in 2.4.2.). If W4 had not been considered, the RMSD between heavy atoms of the ligand, in the X-ray crystal structure and the QM/SQM optimized structure, increased (from 0.99 to 1.12 Å) and correlation between the calculated scores and experimental binding data decreased (R2 decreased from 0.92 to 0.77). 2.4.4. Contribution of X-bonds to inhibitor binding. To better understand the binding of the studied inhibitors to the enzyme and see the importance of every single Xbond, we calculated interaction energies (ΔEint) of the inhibitors with the selected amino acid side chains and peptide bonds of the enzyme. These results are shown in Tables S3 and S4. The calculated ΔEint values showed one strong and one weak X-bonds between Br atoms of the inhibitor and an electron donor of the enzyme. The strong X-bond was between Br of MK204 and O atom of Cys299 (Br...O distance of 2.89 Å, C-Br...O angle of 169.3 degrees and calculated ΔEint between the peptide bond and the inhibitor of -−4.4 kcal/mol, Figure 7). Indeed, this is not possible for MK319, because F does not have a positive σ-hole (14), while MK184 displays a shifted conformation with no halogen bond in this position. The weak X-bond was found between the Br of IDD388 and the O atom of Pro124 (distance 3.43 Å, angle 167.7 degrees, ΔEint −0.4 kcal/mol). A similar Br...O contact can be found between MK181 and Pro124 (distance 3.53 Å, angle 172.8 desgrees). However, the calculated ΔEint value was positive in this case, which shows that there is not an attractive interaction between MK181 and Pro124. This might be explained by the smaller (less positive) σ-hole on the Br atom of MK181 due to the missing electron withdrawing F atom (14). There are additional short Br…O contacts (3.1 – 3.3 Å) between MK184 and Asn300 and Tyr210, MK319 and Ser304, and MK204 and Ser304. However, these Br…O contacts were far from the optimum linear arrangement of the X-bonds. The C-Br…O angle ranges between 123.8 and 150.4 degrees here.


Page 11 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


3. CONCLUSIONS Being AKR1B10 a promising antineoplastic target, the development of potent and specific AKR1B10 inhibitors has been growing in the last years, and given the high similarity with AR, structural studies providing insights into inhibitor binding are essential for designing potent and selective AKR1B10/AR inhibitors (the latest known as ARIs). However, there is not much clinical data available for AKR1B10 as yet (as it can be visualized in http://www.ncbi.nlm.nih.gov/pubmed/clinical/), which anticipates that more inhibitor scaffolds might be useful for AKR1B10 drug development, given the difficult translation of cancer research into clinical success (32). Using a known ARI scaffold and through SAR in its aryl moiety, here we have shown that AR and AKR1B10 active sites present small but significant differences that enable distinct conformations. We have been able to unveil the underlying reasons, uncovering in the process differences between AR and AKR1B10 that can be exploited to narrow the binding of drugs to AKR1B10 (27). For AR, the bulkier substituents (MK184 and MK204) are not suitable because of steric hindrance within the previously described IDD388 AR specificity pocket. Indeed, the conformation of IDD388 (and other potent binders derivatives like MK319) corresponds to the most frequent conformer in AR (21). On the other hand, the two polybrominated compounds are the most potent and selective inhibitors for AKR1B10. Thus, an intriguing possibility would be to add Br substituents in AR inhibitors like ponalrestat, zopolrestat or lidorestat in their ortho positions, which would allow them to interact with AKR1B10 but not AR, as observed here for the two cited IDD388 derivatives. This strategy has been shown to be successful with the compound JF0049, a poor AR but good AKR1B10 binder, also with a polybrominated moiety (but a shorter link between aryl moieties) unable to fit into the AR specificity pocket (8). For AKR1B10, the bulkier substituents do not fit into the characteristic AKR1B10 loop A subpocket, and the lighter MK181 and IDD388, in spite of occupying this subpocket, do not provide strong enough interactions to bind in the vicinity of residue Trp112. This is further confirmed by the QM/SQM calculations that show that their X-bonds between Br in position R4 and the protein are weak. The three bulkier ligands are able to fit nicely into a novel AKR1B10 binding site conformer, mainly through a stacking interaction to the Trp112 native conformation. To achieve strong AKR1B10 inhibition, the ligand binding in this novel pocket might be hydrophobic enough to displace the unfavourable water molecules residing in the AKR1B10 broader active site pocket without paying a high desolvation penalty, which makes MK319 a poor binder. Besides, MK204 has an additional feature over the rest of inhibitors: it can establish a strong halogen bond while MK184 and MK319 cannot. Thus, MK204 is a lead compound targeting the distinct AKR1B10 specificity pocket

and with a strong X-bond. The possibility to modify MK204 in order to modulate the strength of the X-bond and improve its predicted bioavailability (it violates the logP

IDD388 Polyhalogenated Derivatives as Probes for an Improved

Recommend Documents