Systematic Computational Design and Identification of Low Picomolar

Subscriber access provided by UNIVERSITY OF TOLEDO LIBRARIES

Article

Systematic Computational Design and Identification of Low Picomolar Inhibitors of Aurora Kinase A Hwangseo Park, Hoi-Yun Jung, Shinmee Mah, and Sungwoo Hong J. Chem. Inf. Model., Just Accepted Manuscript • DOI: 10.1021/acs.jcim.7b00671 • Publication Date (Web): 05 Feb 2018 Downloaded from http://pubs.acs.org on February 6, 2018

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 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 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.

Journal of Chemical Information and Modeling 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 33 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

Journal of Chemical Information and Modeling

Systematic Computational Design and Identification of Low Picomolar Inhibitors of Aurora Kinase A Hwangseo Park,*,† Hoi-Yun Jung, ‡,§ Shinmee Mah, ‡,§ and Sungwoo Hong*,‡,§

†

Department of Bioscience and Biotechnology & Institute of Anticancer Medicine

Development, Sejong University, 209 Neungdong-ro, Kwangjin-gu, Seoul 05006, Korea b

Center for Catalytic Hydrocarbon Functionalizations, Institute for Basic Science (IBS),

34141, Korea §

Department of Chemistry, Korea Advanced Institute of Science and Technology (KAIST)

Daejeon 34141, Korea

Author ORCID ID Hwangseo Park: 0000-0001-5806-2472 Hoi-Yun Jung: 0000-0003-4111-850X Shinmee Mah: 0000-0001-9459-5073 Sungwoo Hong: 0000-0001-9371-1730

*Correspondence may be addressed to either author.

Telephone: +82-2-3408-3766; +82-42-350-2811 Fax: +82-2-3408-4334; +82-42-350-2812 E-mail: [email protected]; [email protected] 1

ACS Paragon Plus Environment

Journal of Chemical Information and Modeling 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

Abstract Aurora kinase A (AKA) has served as an effective molecular target for the development of cancer therapeutics. A series of potent AKA inhibitors with the (4-methoxy-pyrimidin-2-yl)phenyl-amine (MPPA) scaffold are identified using a systematic computer-aided drug design protocol involving structure-based virtual screening, de novo design, and free energy perturbation (FEP) simulations. To enhance the accuracy of the virtual screening to find a proper molecular core and de novo design to optimize biochemical potency, we preliminarily improved the scoring function by implementing a reliable hydration energy term. The overall design strategy proves successful to the extent that some inhibitors reveal exceptionally high potency at low picomolar levels; this was achieved by substituting phenyl, chlorine, and tetrazole moieties on the MPPA scaffold. The establishment of bidentate hydrogen bonds with backbone groups in the hinge region appears to be necessary for the high biochemical potency, consistent with the literature X-ray crystallographic data. The picomolar inhibitory activity also stems from the simultaneous formation of additional hydrogen bonds with the side chains of the hinge region and P-loop residues. The FEP simulation results show that the inhibitory activity surges to the low picomolar level because the interactions in the ATP-binding site of AKA become strong by structural modifications enough to overbalance the increase in dehydration cost. Due to the exceptionally high biochemical potency, the AKA inhibitors reported in this study are anticipated to serve as a new starting point for the discovery of anticancer medicine.

2


Page 2 of 33

Page 3 of 33 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


Introduction The Aurora kinase family is a class of serine/threonine kinases that plays a pivotal role in regulating cell cycle and division. Although three isoforms (Aurora A, B, and C) share a highly conserved C-terminal catalytic domain containing the auto-phosphorylation sites, they exhibit different subcellular distributions in mitotic events and perform physiological functions independently with distinct mechanisms.1 Aurora kinase A (AKA) tends to be upregulated in the G2/M phase of the cell cycle,2 the enzymatic activity of which depends on the phosphorylation state of Thr288 in the activation loop (A-loop, residues 274-299). Residing at the centrosomes and spindle poles, AKA is involved in centrosome maturation, spindle assembly, and metaphase I spindle orientation during the cell division.3 Aurora kinase B (AKB) influences chromosome arms and centromeres in prometaphase to form a chromosomal passenger protein complex, which facilitates chromosome-microtubule alignment, chromosome segregation, and cytokinesis.4 In contrast to the relative abundance of AKA and AKB, the expression of Aurora kinase C is limited to the testis and placenta to be involved in meiosis.5 Aberrant Aurora kinase activity has been implicated in a variety of human cancers together with polyploidy and chromosome instability.6 For example, the overexpression of AKA and AKB has been observed frequently in colorectal, glioma, breast, ovarian, and pancreatic cancers. The inhibition of auto-phosphorylation at the activation sites of Aurora kinases leads to transient G2/M arrest and cell apoptosis.7 Much experimental evidence has thus accumulated for the suitability of Aurora kinases as a potential therapeutic target for the development of anticancer medicines. Over the past decade, a great deal of effort has been devoted to the discovery of smallmolecule inhibitors of Aurora kinases, as reviewed recently in a comprehensive fashion.8 3



Page 4 of 33

These scientific endeavors have enabled the discovery of potent Aurora kinase inhibitors involving

4-aminopyrimidine-2-thiol

phthalazin-1-amine

(AMG900),11

(tozasertib),9

3-aminopyrazole

5H-benzo[c]pyrimido[4,5-e]azepine

(danusertib),10 (alisertib),12

quinazolin-4-amine (barasertib),13 and pyrimidine-2,4-diamine (HLM008598)14 as the key structural elements. Although a number of Aurora kinase inhibitors were assessed in clinical studies, none has yet been approved as a new cancer drug due to limited efficacy in solid tumors. In addition to the anticancer effect, the impairment of Aurora kinase activity may also enhance the response rate against hematologic malignancies,1 further motivating the discovery of new potent Aurora kinase inhibitors.

Chart 1. Chemical Structures of Aurora A Inhibitors.

This study was undertaken to identify highly potent AKA inhibitors using rational computer-aided design in combination with chemical synthesis and biochemical validation. The computational protocol involves virtual screening with molecular docking to find a suitable molecular core, de novo design to generate putative AKA inhibitors, and free energy 4


Page 5 of 33 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


perturbation (FEP) calculations to screen the final candidates to be synthesized. Although virtual screening and de novo design proved useful for structure-based drug discovery,15-17 their applicability has been limited due to the roughness of the scoring function used to estimate the protein-ligand binding affinity.18-21 This stems in large part from the underestimation of ligand dehydration energy in protein-ligand association, which inevitably culminates in overestimating the binding affinity of a ligand with hydrophilic groups.22 To alleviate this dehydration problem, the scoring function is modified by substituting a suitable hydration free energy term. Virtual hits are further screened with the FEP method based on molecular dynamics (MD) simulations to enhance the possibility of discovering potent AKA inhibitors. This was motivated by recently reported successful applications of the FEP method in structure-based drug design.23-27 It will be exemplified that even low picomolar AKA inhibitors can be identified when the ligand hydration effects and dynamic aspects of proteinligand interactions are reflected properly in the individual design steps.

5



Computational Methods Preparation of an all-atom receptor model for AKA. Three-dimensional atomic coordinates required for virtual screening and de novo design were prepared using the X-ray structure of AKA in complex with a potent inhibitor (PDB entry: 4UYN).28 The formal charges of titratable residues of AKA were assigned according to patterns for hydrogen bond formation from the original X-ray crystal structure. Upon determination of protonation states, hydrogen atoms were attached to all heavy atoms of AKA after removing the crystallographic solvent molecules to complete the all-atom receptor model for AKA. The energy minimization was conducted last to correct physically unreasonable steric contacts produced due to the addition of a vast number of hydrogen atoms.

Virtual screening for a molecular core to bind tightly in the hinge region. Because the establishment of bidentate hydrogen bonds with backbone groups in the hinge region proved to be necessary for high inhibitory activity,28-31 the design of AKA inhibitors began with finding the most suitable molecular substructure that would exhibit this peculiar interaction pattern. This could be accomplished through structure-based virtual screening of a chemical library comprising approximately 230,000 building block molecules, which were distributed by several compound vendors including ChemDiv, InterBioScreen, ASINEX, SPECS, ENAMINE. More specifically, docking simulations of small molecules with molecular weight below 210 amu were performed near the hinge region of AKA with the automated AutoDock program32 to estimate the binding modes and binding free energies. To augment the accuracy of docking simulations, the original scoring function was modified to include an accurate ligand dehydration free energy term. The introduction of this new energy term would enhance predictive capability by reflecting ligand hydration effects in protein-ligand binding. The 6


Page 6 of 33

Page 7 of 33 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


modified scoring function to estimate AKA-inhibitor binding free energy (∆Gbind) can be expressed using the following mathematical formula.

 Aij Bij   C ij Dij  ∆Gbind = WvdW ∑∑  12 − 6  + Whbond ∑∑ E (t ) 12 − 10   r rij  rij  i =1 j =1  rij i =1 j =1  ij rij  − 2 max  2σ + Wsol ∑ S i Occi − ∑ V j e  i =1 j ≠i  2

+ Welec ∑∑ i =1 j =1

qi q j

ε (rij )rij

+ Wtor N tor

   

(1)

The weighting parameters for van der Waals contacts (WvdW), hydrogen bonds (Whbond), electrostatic interactions (Welec), entropic penalty (Wtor), and ligand dehydration free energy (Wsol) were set to 0.1485, 0.0656, 0.1146, 0.3113, and 0.1711, respectively, as in the original AutoDock program.33 All the energy parameters in eq 1 were also extracted from the original AutoDock program except for those in the dehydration term that had been derived with the extended solvent-contact model.34,35 Further details on the empirical parameters in eq 1 are provided in Supporting Information. Using this scoring function including the modified dehydration term, docking simulations were conducted to find the molecular core that could bind tightly in the hinge region with multiple hydrogen bonds.

De novo design of AKA inhibitors adequate for binding to the ATP-binding site. Beginning from the molecular core selected in the precedent virtual screening, structure-based de novo design was performed with the scoring function in eq 1 to generate the derivatives suitable for binding tightly in the ATP-binding site of AKA. To make the design of potent AKA inhibitors more possible, the fragment library for the de novo design was prepared including the chemical moieties present in the actual aurora kinase inhibitors as retrieved in PubChem (https://pubchem.ncbi.nlm.nih.gov). This fragment library contained some unusual 7



moieties such as tetrazole, morpholine, and benzo[d]imidazole as well as popular chemical groups including alkyl chains and substituted phenyl rings. A variety of putative AKA inhibitors were then constructed by operating the genetic algorithm to introduce the chemical moieties at the substitution positions of the molecular core.36 Tight binding in the ATPbinding site served as the criterion for selecting putative AKA inhibitors. To reduce the computational burden for the entire design procedure, only the derivatives that satisfied the bioavailability conditions were filtered as virtual hits for further analysis.37

Final virtual screening with FEP simulations. To select the candidate AKA inhibitors to be prepared by chemical synthesis, FEP calculations were performed via MD simulations using the all-atom model of AKA in complex with the putative inhibitors produced in the earlier de novo design. All MD-FEP simulations of the AKA-inhibitor complexes were based on the thermodynamic cycle, as depicted in Figure 1. Within this theoretical framework, the difference (∆∆Gbind) between the binding free energy of inhibitor I1 (∆G1) and that of inhibitor I2 (∆G2) was estimated through the nonphysical path connecting I1 and I2. Since the sum of all energy components of the thermodynamic cycle in Figure 1 is zero, the ∆∆Gbind value between I1 and I2 can be calculated by subtracting the free energy change during the transformation of the former into the latter in water (∆Gu) from that in the ATP-binding pocket of AKA (∆Gb). It is therefore possible to calculate the relative binding affinities among AKA inhibitors using the relevant structural transformations, providing one of the most rigorous methods for virtual screening.

8


Page 8 of 33

Page 9 of 33 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


Figure 1. Thermodynamic cycle adopted to compute the free energy change in water (∆Gu), that in the ATP-binding pocket of AKA (∆Gb), and the difference between binding free energies (∆∆Gbind) of two AKA inhibitors (I1 and I2).

The transformation of I1 (state 0) to I2 (state 1), used to calculate ∆Gu and ∆Gb, was performed by gradually perturbing the Hamiltonian of the total system by changing the variable (λ) to describe the initial and final state with 0 and 1, respectively. Upon completion of the structural perturbation, ∆Gu and ∆Gb were obtained by summing the partial changes of free energy over discrete and equally spaced λ values.38

 Vλ (i +1) − Vλ (i )   ∆G (0 → 1) = − RT ∑ ln exp − RT i  

(2) λ (i )

In this equation, Vk denotes the potential energy for a representative state k, and 〈⋅⋅⋅〉 k means the ensemble average of the enclosed physical quantity at that state. Standard AMBER force field parameters were used to calculate ∆Gu and ∆Gb for all types of AKA and the candidate inhibitors. The procedure for computing ∆Gu and ∆Gb values via MD simulations are described in detail in Supporting Information.

9



Page 10 of 33

Chemical Synthesis and enzyme inhibition assays of AKA inhibitors. The general synthetic route for the preparation of 2-aminopyrimidine derivatives is depicted in Scheme 1. In all cases, the C4 position of pyrimidine core was functionalized prior to performing nucleophilic aromatic substitution (SNAr) reaction at the C2 position. Thus, the isopropoxide and various phenol groups were installed at the C4 position by base-promoted SNAr displacement of 2,4,5-trichloropyrimidines (I in Scheme 1) at room temperature. Aniline groups were then introduced at the 2-position through another substitution reaction with requisite anilines by heating to 100 °C, which afforded the final products. Biochemical potencies of the synthesized putative AKA inhibitors were measured with enzyme inhibition assays as described in Supporting Information.

Scheme 1. Synthetic route of 2-aminopyrimidine derivativesa

O

O R

c

R

N

N

R'

a N

Cl R

I

Cl

IIa

N N

N

1-2, 4-14

Cl b

1, R = H 2, 4-11, 13, 15-17, R = Cl 3, 12, 14, R = Me

O

Ar

O c

R

N H

R

N

N N

Cl

Ar

N

R' N H

3, 15-17

IIb 3, Ar = Ph 15, Ar = 4-OMe-Ph 16, Ar = 4-F-Ph 17, Ar = 2-F-Ph a

Reagents and conditions: (a) NaOiPr, anh. Et2O, 0 ℃ to rt, 12 h; (b) ArOH, KOtBu, DMF, rt,

16 h; (c) aniline, TFA, 2-BuOH, 100 °C, 3-12 h. 10


Page 11 of 33 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


Results and Discussion Prior to designing the AKA inhibitors, we examined the accuracy of the modified scoring function in comparison with the original one. The test set for this assessment comprised 20 known AKA inhibitors including those in Chart 1 and 980 decoy molecules selected at random in the docking library. The percentages of true hits found by the two scoring functions are compared in Figure 2 in varying top-scoring fractions for producing a hit list. The modified scoring function outperforms the original one in achieving the highest enrichment irrespective of fraction cutoffs. For example, the former selects four actual inhibitors as hits among top 1% of the screened molecules in contrast to only one with the latter. When top 10% candidate molecules in the data set are regarded as virtual hits, the new scoring function recognizes a total of thirteen actives whereas only five are picked with the original one. These results indicate the outperformance of the modified scoring function over the original one in predicting the biochemical potencies of AKA inhibitors.

Figure 2. Cumulative percentage of the actual AKA inhibitors selected as hits in virtual screening with respect to the fraction of selection in the test set. 11



To discover the new and potent AKA inhibitors, we adopted a computational protocol comprising the two steps illustrated in Figure 3. The first step was the virtual screening of building block molecules to find the best substructure for the hinge region in the ATP-binding site of AKA. Among 230,000 molecules screened with docking simulations, (4-methoxypyrimidin-2-yl)-phenyl-amine (MPPA) moiety was the top-scoring virtual hit with the lowest binding free energy. We note that MPPA includes the 2-aminopyrimidine group, as in some known kinase inhibitors. Nonetheless, MPPA deserves a new molecular core because it possesses the etheric oxygen as a linking group instead of the amino moiety observed in known AKA inhibitors.14

Figure 3. Flowchart for a systematic computer-aided design approach to discover potent AKA inhibitors.

Figure 4 shows the binding mode of MPPA in the ATP-binding site of AKA predicted with the scoring function in eq 1. MPPA appears to be stabilized through the interactions with amino acid residues in the hinge region, the glycine-rich phosphate loop (P-loop), and the Aloop (Figure 4a), which are the key functional elements of AKA. Judging from the proximity to hot spot residues, MPPA is anticipated to serve well as a molecular core from which new potent AKA inhibitors can be derived using de novo design and chemical synthesis.

12


Page 12 of 33

Page 13 of 33 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


Figure 4. (a) Calculated binding of MPPA in the ATP-binding pocket of AKA (b) Detailed interaction patterns of MPPA in the ATP-binding pocket of AKA. Carbon atoms of AKA and MPPA are displayed in green and cyan, respectively. Two hydrogen bonds are indicated with dotted lines.

In the calculated AKA-MPPA complex, we note that the bridging amino moiety and one of aromatic nitrogens on the pyrimidine ring of MPPA donates and receives a hydrogen bond to the backbone aminocarbonyl oxygen and from the amidic nitrogen of Ala213 in the hinge region (Figure 4b), respectively. The capability to form these bidentate hydrogen bonds proved necessary for the high biochemical potency of AKA inhibitors.28-31 MPPA appears to be further stabilized in the ATP-binding site through van der Waals contacts of the phenyl and pyrimidinyl rings to the nonpolar side chains of Leu139, Val147, Ala160, Leu194, and Leu263. These hydrophobic interactions would have the effect of protecting the bidentate hydrogen bonds by restricting the approach of solvent molecules. Some synergistic effects are thus expected in strengthening the AKA-MPPA interactions by positioning the hydrophobic interactions and the bidentate hydrogen bonds in proximity to each other. Indeed, it proved to 13



be a facile strategy in structure-based drug design to reinforce the hydrogen-bonding and hydrophobic interactions in a cooperative fashion.39 Notably, the two sub-binding pockets (Pocket 1 and 2 in Figure 4b) in the ATP-binding site remain largely vacant in the AKA-MPPA complex. Pocket 1 comprising the side chains of Lys141, Val147, Leu194, Leu210, and Asp274 points to the pyrimidinyl ring of MPPA, while Pocket 2 includes the backbone groups of residues 215-217 in the hinge and the side chains of Arg137, Leu139, Thr217, and Arg220 to accommodate the phenyl ring of MPPA. Therefore, biochemical potency would be enhanced significantly by introducing the proper chemical groups on the two aromatic rings of MPPA, due to the strengthening of binding in the ATPbinding site. Beginning from the calculated AKA-MPPA complex shown in Figure 4, a variety of MPPA derivatives were generated in the de novo design to find the putative AKA inhibitors that would have maximal binding affinity in the ATP-binding site. The position of MPPA was fixed during the entire course of de novo design to retain its binding affinity for the hinge region. This design strategy has an advantage in the context that the biochemical potency of a putative inhibitor can be maximized by keeping the molecular core in the optimal position. To maximize the strength of binding in Pocket 1 and 2, the substitution positions of MPPA included the terminal methyl carbon (R1) and C5 atom on the pyrimidine ring (R2) as well as the meta (R3) and para (R4) positions of the terminal phenyl ring with respect to the bridging −NH group. The introduction of a large chemical moiety at the R1 and R2 positions may induce tighter binding in Pocket 1 flanking P-loop and A-loop, while the substitutions at the R3 and R4 positions would have the effect of strengthening the interactions in Pocket 2 at the end of the ATP-binding site (Figure 4). Among the MPPA derivatives generated in the de novo design, those with a calculated binding free energy lower than MPPA by more than 5 14


Page 14 of 33

Page 15 of 33 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


kcal/mol were selected for further analysis. A total of 35 candidates were finally screened with MD-FEP simulations to select 17 MPPA derivatives with the same criterion in the precedent de novo design. These were prepared via chemical synthesis and evaluated with enzyme inhibition assays. Table 1 lists the structures and inhibitory activities of the newly identified AKA inhibitors with the MPPA scaffold. Notably, most MPPA derivatives reveal high biochemical potencies at low-nanomolar and picomolar levels. This exemplifies the efficiency of the design strategy adopted in this study.

Table 1. Structures and inhibitory activities (IC50 values) of MPPA derivatives against AKA compared to staurosporine as a reference.

O R2

*

1

*

2

3

4

5

*

*

*

R4

N N

R1a

R1

R3

N H

R2

R3a

R4a

IC50 (nM)

H

H

H

512

Cl

H

H

42.0

CH3

H

H

8.10

Cl

CH3

H

12.7

H

13.4

*O

Cl

15



*

6

*

7

*

8

Cl

NO2

H

20.2

Cl

NH2

H

1.40

Cl

H

Cl

H

279

O

*

9

Page 16 of 33

21.4

* O

*

10

Cl

H

18.9

*

HN O

*

11

*

13

CH3

H

*

Cl

H

*N

*

CH3

H

O

*

O

* H N

Cl

H

* N

16

F

*

Cl

H

H N

* N H N

F

Cl

17

H

*

* N

16


3.30

N

N

15

37.3

N

H N

*

14

H

O

*

12

O S

Cl

N

7.58

0.438

N

N

0.043

N

N

0.068

N

N N

0.012

Page 17 of 33 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


staurosporine

a

0.252

Asterisk indicates the atom attached to the position of substitution.

Remarkably, a submicromolar AKA inhibitor (1) was found simply by the replacement of methyl group at the R1 position of MPPA with an isopropyl moiety. The inhibitory activity increased further to the nanomolar level in 2 with chlorination at the R2 position of 1. Only chloride atom and methyl group appear to be allowed at the R2 position of all MPPA derivatives generated in this work; large substituents such as isopropyl and substituted phenyl groups are favored at the R1 position to occupy a large vacant volume in Pocket 1. The preference for a small substituent at the R1 position can be understood because the C-5 atom of the pyrimidine ring resides in close proximity to the side chains of Val147, Ala160, and Leu210 (Figure 4). The respective introductions of phenyl and methyl groups at the R1 and R2 positions lead to low-nanomolar inhibitory activity in 3. The high biochemical potencies of 2 and 3 with only two substituents imply the usefulness of MPPA as a molecular core to design new AKA inhibitors. Additional substitutions of methyl (4), methoxy (5), nitro (6), and amino (7) groups at the R3 position of 2 lead to a more than two-fold increase in biochemical potency. Like R2, only small substituents appear to be favored at the R3 position to retain the nanomolar inhibitory activity. This can be attributed to the proximity of the meta carbon of the phenyl ring to the backbone groups of the hinge region (Figure 4). The R4 position may be more effective than R3 in terms of enhancing the biochemical potency because, as can be seen in Figure 4, the former points toward a vacant space bordered by the amino acid residues in the P-loop (Arg137 and Leu139) and the hinge region (Leu215Thr217 and Arg220). Although the phenyl ring in 8 is a poor substituent at the R4 position, 17



the introduction of acetone (9), acetamide (10), and sulfonamide (11) restores nanomolar inhibitory activity. Even low-nanomolar inhibitors can be obtained by substituting dimethylamide (12) and morpholine (13) groups at the R4 position. The inhibitory activity surges to the picomolar level with the introduction of a tetrazole moiety at the R4 position in 14-17 (Table 1). In particular, IC50 values of 15-17 become even lower than that of the reference compound (staurosporine) due to the substitutions of phenyl ring, chloride atom, and tetrazole moiety at the R1, R2, and R4 positions, respectively. Although the tetrazole ring was suggested as an effective chemical moiety for AKA inhibitors,28 until this work it has been unprecedented to enhance biochemical potency to the low-picomolar level using a tetrazole substituent. This result confirms the efficiency of the computational protocol (Figure 3) devised to discover new potent AKA inhibitors. To seek a rationale for the exceptionally high inhibitory activity, we calculated the binding mode of the most potent inhibitor (17) with docking simulations in the ATP-binding site of AKA. As shown in Figure 5, the chloride atom and fluorobenzene group of 17 are bound with moderate strength through van der Waals contacts with nonpolar residues including Lys141, Ala160, Val147, Leu194, Leu210, and Leu263. Major contribution to the low-picomolar inhibitory activity comes from the terminal tetrazole moiety, which establishes the two strong hydrogen bonds with the side-chain guanidinium ions of Arg137 and Arg220 in Pocket 2. It can also form a stable hydrogen bond with the backbone aminocarbonyl oxygen of Leu139. The capability to form hydrogen bonds with Arg137 and Leu139 is important for AKA inhibition because they reside in the P-loop, which serves as a receptor for the phosphate group during ATP binding. The formation of multiple hydrogen bonds with Ploop residues is therefore likely to preventing ATP from binding to AKA. This can be

18


Page 18 of 33

Page 19 of 33 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


invoked to explain the picomolar inhibitory activity of the MPPA derivatives with a tetrazole moiety.

Figure 5. Calculated binding mode of 17 in the ATP-binding pocket of AKA. Carbon atoms of AKA and 17 are displayed in green and cyan, respectively. Hydrogen bonds are indicated with dotted lines.

It is worth noting that the MPPA core of 1-17 are structurally similar to pyrimidine-2,4diamine moiety in HLM008598 (Chart 1). However, the former can be discriminated from the latter in that the etheric oxygen (instead of -NH- moiety) serves as the linker group to connect the central pyrimidine and the terminal phenyl ring. This substitution may have a significant effect on the biochemical potency by perturbing the positions of the flanking chemical moieties to target the hinge region and the P-loop. It is apparent that the etheric oxygen is the better linker than amino moiety because the biochemical potency increases from low19



nanomolar level in HLM008598 and its structural analogues14 to picomolar level in MPPA derivatives (Table 1). To provide experimental evidence for the tight binding of 17 in the ATP-binding site of AKA, we investigated its steady-state kinetic behavior in enzyme inhibition assays at varying ATP concentrations. The results for these kinetic analyses are summarized in Figure 6 as Lineweaver–Burk plots. Four linear plots obtained at 0.41, 1.23, 3.70, and 11.1 nM of 17 appear to cross the intercept at approximately -0.5 on the abscissa. This indicates that 17 would impair the enzymatic activity of AKA in the ATP-competitive fashion, consistent with the binding mode estimated with docking simulations (Figure 5). The tight binding of 17 in the ATP-binding site is further supported by a low Ki value (0.54 nM) with respect to unliganded AKA, in contrast to an almost infinite value (1.65×1016 nM) for the AKA-ATP complex. The differential inhibitory activities among the MPPA derivatives against AKA may therefore be elucidated by comparing their binding modes in the ATP-binding site.

Figure 6. Lineweaver-Burk plot associated with inhibition of AKA by 17. Enzymatic reaction velocities (V) were measured at 1, 2.5, 5, 25, and 75 µM of ATP concentrations. 20


Page 20 of 33

Page 21 of 33 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


We next address the energetic features associated with AKA inhibition by MPPA derivatives. Summarized in Table 2 are the ∆∆Gbind values of AKA inhibitors calculated with FEP simulations (∆∆Gcalc) in comparison with the corresponding experimental data (∆∆Gexp). MPPA served as the initial (λ=0) state only for 1, which was in turn perturbed to 2-17 to estimate the differential biochemical potencies among the AKA inhibitors. Changing the initial state from MPPA to 1 was inevitable because of the large statistical errors in FEP calculations of 2-17 using MPPA as the template. Although the root mean square error (RMSE) is as high as 1.30 kcal/mol for the comparison of ∆∆Gcalc and ∆∆Gexp, the squared Pearson coefficient (R2) amounts to 0.75, implying a good correlation. The predictive capability of FEP simulations for AKA inhibitors is comparable in terms of R2 to those for beta secretase inhibitors and G-protein-coupled receptor ligands.24-26

Table 2. Perturbation-induced free energy change in water (∆Gu) and in the ATP-binding site (∆Gb), and the difference in calculated binding free energies (∆∆Gcalc) between the two AKA inhibitors compared to the corresponding experimental data (∆∆Gexp). All energy values are given in kcal/mol. perturbation

∆Gu

∆ Gb

MPPA → 1

-1.1 ± 0.2

-3.4 ± 0.0

-2.3 ± 0.2

NA

1→2

-1.5 ± 0.1

-4.3 ± 0.2

-2.8 ± 0.2

-1.5

1→3

-2.2 ± 0.4

-5.3 ± 0.3

-3.2 ± 0.5

-2.5

1→4

-3.8 ± 0.1

-7.2 ± 0.4

-3.4 ± 0.4

-2.2

1→5

-4.6 ± 0.2

-7.9 ± 0.3

-3.3 ± 0.4

-2.2

1→6

-4.2 ± 0.4

-7.6 ± 0.7

-3.4 ± 0.8

-1.9

1→7

-4.1 ± 0.4

-7.0 ± 0.2

-2.9 ± 0.4

-3.5

1→8

-4.2 ± 0.6

-7.3 ± 0.3

-3.1 ± 0.7

-0.4

1→9

-5.5 ± 0.6

-9.1 ± 0.5

-3.6 ± 0.8

-1.9

21


a

∆∆Gcalc

b

∆∆Gexp


a

Page 22 of 33

1 → 10

-7.3 ± 0.7

-10.7 ± 0.9

-3.4 ± 1.1

-2.0

1→ → 11

-8.7 ± 0.5

-11.9 ± 0.5

-3.2 ± 0.7

-1.6

1→ → 12

-7.7 ± 0.8

-11.7 ± 0.9

-4.0 ± 1.2

-3.0

1→ → 13

-8.9 ± 0.9

-12.6 ± 1.3

-3.7 ± 1.6

-2.5

1→ → 14

-8.3 ± 0.7

-13.4 ± 1.3

-3.9 ± 1.5

-4.2

1→ → 15

-9.6 ± 1.6

-14.3 ± 1.1

-4.7 ± 1.9

-5.5

1→ → 16

-9.3 ± 1.2

-13.8 ± 1.4

-4.5 ± 1.8

-5.3

1→ → 17

-9.1 ± 0.8

-14.3 ± 1.2

-5.2 ± 1.4

-6.3

RMSE

1.30

R2

0.75

Error in ∆∆Gcalc was approximated as square root of the sum of the squares of those in ∆Gu

and ∆Gb values. b∆∆Gexp values were calculated from Ki values of 1-17 obtained with Cheng−Prusoff equation [Ki = IC50/(1 + [ATP]/Km)].

As can be seen in Table 2, the statistical errors of the ∆Gu and ∆Gb terms contribute similarly to those of the ∆∆Gcalc values, which tend to increase with the increase of molecular size. For example, the errors accumulate to almost 2 kcal/mol during the structural transformations from 1 to 15 and 16 but were lower than 0.5 kcal/mol in 1 → 2, 1 → 4, 1 → 5, and 1 → 7 perturbations. The small errors in the latter can be understood because only small substituents (chloride, methyl, methoxy, and amino groups) were introduced at the R2 and R3 positions of 1. The candidate AKA inhibitors possessing a substituent larger than the six-membered ring at the R4 position were excluded in chemical synthesis because the relative errors in ∆∆Gcalc exceeded 50%. This exemplifies a drawback of FEP simulations in which structural changes should not be significant for accuracy. Consistent with potency enhancement with the introduction of various substituents into the MPPA scaffold (Table 1), ∆∆Gcalc values appear to become more favorable in the 22


Page 23 of 33 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


structural perturbations of 1 to 2-17 (Table 2). Because ∆Gb is more negative than ∆Gu in all cases (Table 2), the improved inhibitory activity can be attributed to higher stabilization of MPPA derivatives in the ATP-binding site of AKA than in water. It can thus be argued that the increase in dehydration cost should be overcome by strengthening the interactions with AKA to enhance the biochemical potency of MPPA derivatives. Despite reasonably good agreement between ∆∆Gcalc and ∆∆Gexp data, a large discrepancy is observed for 8 with the error of 2.7 kcal/mol. We note that 8 possesses two large nonpolar substituents (isopropyl and phenyl groups at the R1 and R4 positions, respectively), indicating that van der Waals interactions would play a significant role in stabilizing the AKA-8 complex. In this case, a relatively large error accumulates during FEP simulations because of the singularity problem in the Lennard-Jones potential, which may arise as the potential function is scaled from the initial to final state.40 The large deviation of the ∆∆Gcalc value of 8 from the experimental counterpart can thus be attributed to the imperfection of the van der Waals interaction term. In agreement with the experimental data, the ∆∆Gcalc values of 4 and 5 are similar although the latter has more negative ∆Gb than the former (Table 2). Therefore, the extra stabilization of 5 in the ATP-binding site of AKA is likely to be negated by the increased stabilization in water, which can be inferred from the more negative ∆Gu value in the 1→5 perturbation than in the 1→4 counterpart. This confirms that to improve the biochemical potency of AKA inhibitors by chemical modifications, the interactions in the ATP-binding site should be made strong enough to surmount the increased dehydration cost. When ∆Gu and ∆Gb in 1→11 and 1→12 perturbations are compared, it follows immediately that the increase in binding affinity to AKA in going from 11 to 12 stems from

23



the reduced dehydration cost rather than the increased stabilization in the ATP-binding site. This indicates that inhibitory activity can be enhanced by reducing the dehydration cost as well as by strengthening the interactions with the amino acid residues in the ATP-binding site. However, ∆Gu and ∆Gb increase and decrease in going from 1→13 to 1→14 perturbations, respectively, implying that the increase in inhibitory activity from the low nanomolar (13) to subnanomolar (14) level stems from the combined effects of reduced dehydration cost and strengthening of the interactions with AKA. A common energetic feature of the MPPA derivatives involving phenyl and tetrazole moieties at the R1 and R4 positions (15-17), respectively, is that ∆Gb is even more negative than ∆Gu by 4.5–5.2 kcal/mol. Therefore, their picomolar-level inhibitory activity (Table 1) can be elucidated in terms of strengthening the interactions in the ATP-binding site of AKA to the extent that it can negate the increased stabilization in water. With respect to high biochemical potency, molecular substructures of the MPPA derivatives identified in this study appear to serve cooperatively to optimize the interactions with hot spot residues in the ATP-binding pocket. For example, ∆∆Gexp becomes more favorable by 1.5, 2.7, and 1.3 kcal/mol (Table 2) in going from 1 to 2 by introducing a chloride atom at the R2 position, from 2 to 14 with the substitution of tetrazole ring at the R4 position, and from 14 to 15 due to the replacement of isopropyl with anisole at the R1 position, respectively. Remarkably, the sum of the three ∆∆Gexp values is very similar to the single

∆∆Gexp value between 1 and 15. Hence, the substituents at the R1, R2, and R4 positions contribute to the picomolar inhibitory activity of 15 in the additive fashion, indicating that the MPPA scaffold serves as an effective linking group to retain key interactions between the substituents and hot spot residues of AKA. The discovery of low-picomolar AKA inhibitors was thus enabled by the identification of an optimal linker group with virtual screening and 24


Page 24 of 33

Page 25 of 33 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


subsequently by introducing the proper substituents to optimize biochemical potency using de novo design and FEP simulations.

Conclusions We identified a series of potent AKA inhibitors using a systematic computer-aided drug design protocol involving virtual screening to find the optimal molecular core for the hinge region, de novo design to optimize the binding affinity, and FEP simulations to select the candidates to be synthesized. Preliminary to virtual screening and de novo design, the scoring function was modified by implementing a proper hydration energy term to enhance the accuracy of the affinity prediction. This design strategy was very efficient and led to the identification of several picomolar AKA inhibitors. This exceptionally high inhibitory activity was achieved by introducing the substituted phenyl, chlorine, and tetrazole groups onto the MPPA scaffold. Consistent with the X-ray crystallographic data, the formation of bidentate hydrogen bonds with backbone groups in the hinge region appeared to be necessary for the high biochemical potency. The simultaneous hydrogen-bond interactions with the side chains of the hinge region and the P-loop residues were also found to be required for the picomolarlevel inhibitory activity. The FEP calculation results indicated that the biochemical potency of MPPA derivatives could surge to low-picomolar levels because the interactions in the ATPbinding site were strong enough to overbalance the increase in dehydration cost for binding to AKA. Due to their high activity, the AKA inhibitors found in this study may serve as a new starting point for the discovery of anticancer medicines.

25



Page 26 of 33

Supporting Information General methods and materials, procedures of synthesis, characterization data, and copies of 1H and 13C NMR spectra for 1-17. This material is available free of charge via the Internet at http://pubs.acs.org.

Notes The authors declare no competing financial interest.

Acknowledgments This research was supported by Basic Science Research Program through the National Research

Foundation

of

Korea

funded

by

the

Ministry

of

Education

(NRF-

2016R1D1A1B01014187) and the Institute for Basic Science (IBS-R010-G1).

References (1) Bavetsias, V.; Linardopoulos, S. Aurora Kinase Inhibitors: Current Status and Outlook. Front. Oncol. 2015, 5, 278. (2) Gavriilidis, P.; Giakoustidis, A.; Giakoustidis, D. Aurora Kinases and Potential Medical Applications of Aurora Kinase Inhibitors: A Review. J. Clin. Med. Res. 2015, 7, 742–751. (3) Keen, N.; Taylor, S. Aurora Kinase Inhibitors as Anticancer Agent. Nat. Rev. Cancer 2004, 4, 927–936. 26


Page 27 of 33 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


(4) Goto, H.; Yasui, Y.; Nigg, E. A.; Inagaki, M. Aurora-B Phosphorylates Histone H3 at Serine 28 with Regard to the Mitotic Chromosome Condensation. Genes Cells 2002, 7, 11–17. (5) Khan, J.; Ezan, F. Cre´met, J.-Y.; Fautrel, A.; Gilot, D.; Lambert, M.; Benaud, C.; Troadec, M.-B.; Prigent, C. Overexpression of Active Aurora-C Kinase Results in Cell Transformation and Tumour Formation. PLoS One 2011, 6, e26512. (6) Green, M. R.; Woolery, J. E.; Mahadevan, D. Update on Aurora Kinase Targeted Therapeutics in Oncology. Expert Opin. Drug Discovery 2011, 6, 291–307. (7) Kaestner, P.; Stolz, A.; Bastians, H. Determinants for the Efficacy of Anticancer Drugs Targeting Either Aurora-A or Aurora-B Kinases in Human Colon Carcinoma Cells. Mol. Cancer Ther. 2009, 8, 2046–2056. (8) Yan, M.; Wang, C.; He, B.; Yang, M.; Tong, M.; Long, Z.; Liu, B.; Peng, F.; Xu, L.; Zhang, Y.; Liang, D.; Lei, H.; Subrata, S.; Kelley, K. W.; Lam, E. W.-F.; Jin, B.; Liu, Q. Aurora-A Kinase: A Potent Oncogene and Target for Cancer Therapy. Med. Res. Rev. 2016, 36, 1036– 1079. (9) Harrington, E. A.; Bebbington, D.; Moore, J.; Rasmussen, R. K.; Ajose-Adeogun, A. O.; Nakayama, T.; Graham, J. A.; Demur, C.; Hercend, T.; Diu-Hercend, A.; Su, M.; Golec, J. M. C.; Miller, K. M. VX-680, a Potent and Selective Small-Molecule Inhibitor of the Aurora Kinases, Suppresses Tumor Growth in Vivo. Nat. Med. 2004, 10, 262–267. (10) Fancelli, D.; Moll, J.; Varasi, M.; Bravo, R.; Artico, R.; Berta, D.; Bindi, S.; Cameron, A.; Candiani, I.; Cappella, P.; Carpinelli, P.; Croci, W.; Forte, B.; Giorgini, M. L.; Klapwijk, J.; Marsiglio, A.; Pesenti, E.; Rocchetti, M.; Roletto, F.; Severino, D.; Soncini, C.; Storici, P.; Tonani, R.; Zugnoni, P.; Vianello, P. 1,4,5,6-Tetrahydropyrrolo[3,4-c]pyrazoles: Identification 27



of a Potent Aurora Kinase Inhibitor with a Favorable Antitumor Kinase Inhibition Profile. J. Med. Chem. 2006, 49, 7247–7251. (11) Payton, M.; Bush, T. L. Chung, G.; Ziegler, B.; Eden, P.; McElroy, P.; Ross, S.; Cee, V. J.; Deak, H. L.; Hodous, B. L.; Nguyen, H. N.; Olivieri, P. R.; Romero, K.; Schenkel, L. B.; Bak, A.; Stanton, M.; Dussault, I.; Patel, V. F.; Geuns-Meyer, S.; Radinsky, R.; Kendall, R. L. Preclinical Evaluation of AMG 900, a Novel Potent and Highly Selective Pan-Aurora Kinase Inhibitor with Activity in Taxane-Resistant Tumor Cell Lines. Cancer Res. 2010, 70, 9846– 9854. (12) Manfredi, M. G.; Ecsedy, J. A.; Chakravarty, A.; Silverman, L.; Zhang, M.; Hoar, K. M.; Stroud, S. G.; Chen, W.; Shinde, V.; Huck, J. J.; Wysong, D. R.; Janowick, D. A.; Hyer, M. L.; LeRoy, P. J.; Gershman, R. E.; Silva, M. D.; Germanos, M. S.; Bolen, J. B.; Claiborne, C. F.; Sells, T. B. Characterization of Alisertib (MLN8237), an Investigational Small-Molecule Inhibitor of Aurora A Kinase Using Novel in Vivo Pharmacodynamic Assays. Clin. Cancer Res. 2011, 17, 7614–7624. (13) Mortlock, A. A.; Foote, K. M.; Heron, N. M.; Jung, F. H.; Pasquet, G.; Lohmann, J. J.; Warin, N.; Renaud, F.; Savi, C. D.; Roberts, N. J.; Johnson, T.; Dousson, C. B.; Hill, G. B.; Perkins, D.; Hatter, G.; Wilkinson, R. W.; Wedge, S. R.; Heaton, S. P.; Odedra, R.; Keen, N. J.; Crafter, C.; Brown, E.; Thompson, K.; Brightwell, S.; Khatri, L.; Brady, M. C.; Kearney, S.; McKillop, D.; Rhead, S.; Parry, T.; Green, S. Discovery, Synthesis, and in Vivo Activity of a New Class of Pyrazoloquinazolines as Selective Inhibitors of Aurora B Kinase. J. Med. Chem. 2007, 50, 2213–2224. (14) Lawrence, H. R.; Martin, M. P.; Luo, Y.; Pireddu, R.; Yang, H.; Gevariya, H.; Ozcan, S.; Zhu, J.-Y.; Kendig, R.; Rodriguez, M.; Elias, R.; Cheng, J. Q.; Sebti, S. M.; Schonbrunn, E.; 28


Page 28 of 33

Page 29 of 33 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


Lawrence, N. J. Development of O-Chlorophenyl Substituted Pyrimidines as Exceptionally Potent Aurora Kinase Inhibitors. J. Med. Chem. 2012, 55, 7392–7416. (15) Murray, C. W.; Verdonk, M. L.; Rees, D. C. Experiences in Fragment-Based Drug Discovery. Trends Pharmacol. Sci. 2012, 33, 224−232. (16) Baker, M. Fragment-Based Lead Discovery Grows up. Nat. Rev. Drug Discov. 2013, 12, 5–10. (17) Daniel, A. E. Introduction to Fragment-Based Drug Discovery. Top. Curr. Chem. 2012, 317, 1–32. (18) Adeniyi, A. A.; Soliman, M. E. S. Implementing QM in Docking Calculations: Is it a Waste of Computational Time? Drug Discov. Today 2017, 22, 1216−1223. (19) Yan, Y.; Wang, W.; Sun, Z.; Zhang, J. Z. H.; Ji, C. Protein-Ligand Empirical Interaction Components for Virtual Screening. J. Chem. Inf. Model. 2017, 57, 1793–1806. (20) Ragoza, M.; Hochuli, J.; Idrobo, E.; Sunseri, J.; Koes, D. R. Protein-Ligand Scoring with Convolutional Neural Networks. J. Chem. Inf. Model. 2017, 57, 942–957. (21) Liu, Z.; Su, M.; Han, L.; Liu, J.; Yang, Q.; Li, Y.; Wang, R. Forging the Basis for Developing Protein–Ligand Interaction Scoring Functions. Acc. Chem. Res. 2017, 50, 302– 309. (22) Shoichet, B. K.; Leach, A. R.; Kuntz, I. D. Ligand Solvation in Molecular Docking. Proteins 1999, 34, 4–16. (23) Abel, R.; Wang, L.; Harder, E. D.; Berne, B. J.; Friesner, R. A. Advancing Drug Discovery through Enhanced Free Energy Calculations. Acc. Chem. Res. 2017, 50, 29



1625−1632. (24) Keränen, H.; Pérez-Benito, L.; Ciordia, M.; Delgado, F.; Steinbrecher, T. B.; Oehlrich, D.; van Vlijmen, H. W. T.; Trabanco, A. A.; Tresadern, G. Acylguanidine Beta Secretase 1 Inhibitors: A Combined Experimental and Free Energy Perturbation Study. J. Chem. Theory Comput. 2017, 13, 1439–1453. (25) Lenselink, E. B.; Louvel, J.; Forti, A. F.; van Veldhoven, J. P. D.; de Vries, H.; MulderKrieger, T.; McRobb, F. M.; Negri, A.; Goose, J.; Abel, R.; van Vlijmen, H. W. T.; Wang, L.; Harder, E.; Sherman, W.; IJzerman, A. P.; Beuming, T. Predicting Binding Affinities for GPCR Ligands Using Free Energy Perturbation. ACS Omega 2016, 1, 293–304. (26) Ciordia, M.; Pérez-Benito, L.; Delgado, F.; Trabanco, A. A.; Tresadern, G. Application of Free Energy Perturbation for the Design of BACE1 Inhibitors. J. Chem. Inf. Model. 2016, 56, 1856–1871. (27) Free Energy Calculations in Rational Drug Design; Reddy, M. R., Erion, M. D., Eds.; Kluwer Academic Publishers: New York, 2001. (28) Carry, J. C.; Clerc, F.; Minoux, H.; Schio, L.; Mauger, J.; Nair, A.; Parmantier, E.; Le Moigne, R.; Delorme, C.; Nicolas, J. P.; Krick, A.; Abecassis, P. Y.; Crocq-Stuerga, V.; Pouzieux, S.; Delarbre, L.; Maignan, S.; Bertrand, T.; Bjergarde, K.; Ma, N.; Lachaud, S.; Guizani, H.; Lebel, R.; Doerflinger, G.; Monget, S.; Perron, S.; Gasse, F.; AngouillantBoniface, O.; Filoche-Romme, B.; Murer, M.; Gontier, S.; Prevost, C.; Monteiro, M. L.; Combeau, C. SAR156497, an Exquisitely Selective Inhibitor of Aurora Kinases. J. Med. Chem. 2015, 58, 362–375. (29) Jonathan M. E.; Stefano S.; Andrea M.; Stefan K. Crystal Structure of Human Aurora B 30


Page 30 of 33

Page 31 of 33 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


in Complex with INCENP and VX-680. J. Med. Chem. 2012, 55, 7841–7848. (30) Martin, M. P.; Zhu, J.-Y.; Lawrence, H. R.; Pireddu, R.; Luo, Y.; Alam, R.; Ozcan, S.; Sebti, S. M.; Lawrence, N. J.; Schönbrunn, E. A Novel Mechanism by Which Small Molecule Inhibitors Induce the DFG Flip in Aurora A. ACS Chem. Biol. 2012, 7, 698–706. (31) Badrinarayan, P.; Sastry, G. N. Specificity Rendering ‘Hot-Spots’ for Aurora Kinase Inhibitor Design: The Role of Non-Covalent Interactions and Conformational Transitions. PLoS ONE 2014, 9, e113773. (32) Park, H.; Lee, J.; Lee, S. Critical Assessment of the Automated Autodock as a New Docking Tool for Virtual Screening. Proteins 2006, 65, 549–554. (33) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; Belew, R. K.; Olson, A. J. Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J. Comput. Chem. 1998, 19, 1639–1662. (34) Park H., Extended Solvent-Contact Model Approach to SAMPL4 Blind Prediction Challenge for Hydration Free Energies. J. Comput. Aided Mol. Des. 2014, 28, 175–186. (35) Chung, K.-C.; Park, H. Accuracy Enhancement in the Estimation of Molecular Hydration Free Energies by Implementing the Intramolecular Hydrogen Bond Effects. J. Cheminform. 2015, 7, 57. (36) Wang, R.; Gao, Y.; Lai, L. Ligbuilder: A Multi-Purpose Program for Structure-Based Drug Design. J. Mol. Model. 2000, 6, 498–516. (37) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Experimental and Computational Approaches to Estimate Solubility and Permeability in Drug Discovery and 31



Development Settings. Adv. Drug Deliv. Rev. 1997, 23, 3–25. (38) Zwanzig, R. W. High-Temperature Equation of State by a Perturbation Method. J. Chem. Phys. 1954, 22, 1420–1426. (39) Muley, L.; Baum, B.; Smolinski, M.; Freindorf, M.; Heine, A.; Klebe, G.; Hangauer, D. G. Enhancement of Hydrophobic Interactions and Hydrogen Bond Strength by Cooperativity: Synthesis, Modeling, and Molecular Dynamics Simulations of a Congeneric Series of Thrombin Inhibitors. J. Med. Chem. 2010, 53, 2126−2135. (40) Genheden, S.; Kongsted, J.; Söderhjelm, P.; Ryde, U. Nonpolar Solvation Free Energies of Protein−Ligand Complexes. J. Chem. Theory Comput. 2010, 6, 3558−3568.

32


Page 32 of 33

Page 33 of 33 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


TOC graphic

33


Systematic Computational Design and Identification of Low Picomolar

Recommend Documents