Underestimated Halogen Bonds Forming with Protein Backbone in

Jun 23, 2017 - Department of Computer Science and Technology, East China Normal University, Shanghai 200241, China. ‡ CAS Key Laboratory of Receptor...
7 downloads 8 Views 2MB Size
Download PDF
Article pubs.acs.org/jcim

Underestimated Halogen Bonds Forming with Protein Backbone in Protein Data Bank Qian Zhang,† Zhijian Xu,*,‡ Jiye Shi,§ and Weiliang Zhu*,‡ †

Department of Computer Science and Technology, East China Normal University, Shanghai 200241, China CAS Key Laboratory of Receptor Research, Drug Discovery and Design Center, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, China § UCB Biopharma SPRL, Chemin du Foriest, B-1420 Braine-l′Alleud, Belgium ‡

S Supporting Information *

ABSTRACT: Halogen bonds (XBs) are attracting increasing attention in biological systems. Protein Data Bank (PDB) archives experimentally determined XBs in biological macromolecules. However, no software for structure refinement in X-ray crystallography takes into account XBs, which might result in the weakening or even vanishing of experimentally determined XBs in PDB. In our previous study, we showed that side-chain XBs forming with protein side chains are underestimated in PDB on the basis of the phenomenon that the proportion of side-chain XBs to overall XBs decreases as structural resolution becomes lower and lower. However, whether the dominant backbone XBs forming with protein backbone are overlooked is still a mystery. Here, with the help of the ratio (RF) of the observed XBs’ frequency of occurrence to their frequency expected at random, we demonstrated that backbone XBs are largely overlooked in PDB, too. Furthermore, three cases were discovered possessing backbone XBs in high resolution structures while losing the XBs in low resolution structures. In the last two cases, even at 1.80 Å resolution, the backbone XBs were lost, manifesting the urgent need to consider XBs in the refinement process during X-ray crystallography study.

1. INTRODUCTION The halogen bond (XB),1−3 a net attractive interaction between the σ-hole (positive electrostatic potential)4,5 associated with a halogen atom (XB donor) and a nucleophile (XB acceptor), are attracting increasing attention in biological systems.6−14 The most reliable evidence and source for XBs are presented in the Protein Data Bank (PDB), which collects experimentally determined biological macromolecules (proteins, DNA, RNA) at the atomic level.15 PDB structures have long been a cornerstone for the studies on XB, especially on the development of the XB scoring function16−20 in drug discovery and design. The PDB structures are also used for the interpretation of increase in specificity and affinity of halo-organics toward their molecular targets in hindsight.21 Therefore, the quality and reliability of XBs in the PDB structures is critical to explore the QSAR studies of halogen-containing compounds and to explore the mechanism of the halogen-containing ligand binding process. The most typical XB is in C−X···Y (X = Cl, Br, I; Y = O, N, S) form. Y is either from protein side chains (side-chain XB) or protein backbone (backbone XB). Boeckler et al. showed that nitrogen−halogen bonds with histidine side chains are underexploited22 and stated that XBs with polar protein side chains are rather underrepresented in PDB.23 As is often the case, a protein side chain is flexible and a protein backbone is rigid, resulting in more unclear electron density of the side chain compared to the © 2017 American Chemical Society

backbone when the resolution becomes lower. Using the proportion of side-chain XBs to overall XBs side − chain XB ( ) as an indicator, we showed that sideside − chain XB + backbone XB

chain XBs are overlooked in PDB with the phenomenon that a significant decrease in the proportion is observed as the resolution becomes lower.24 Here, 35.4% C−X···Y XBs are side-chain XBs and 64.6% are backbone XBs in a PDB (April 2013 release) survey.25 There are twice as many C−X···Y XBs with protein backbone carbonyls as side-chain XBs in the PDB (September 2013 release) survey.26 In addition, most of the XBs in drug discovery and design are backbone XBs.24 Refmac, CNS, and phenix.refine are the three most popular refinement programs for biological macromolecules. However, no software for structure refinement in X-ray crystallography takes into account XBs, which might result in the weakening or even vanishing of experimentally determined XBs in PDB. For example, the VDWR/VAND keyword in Refmac restrains VDW repulsions. van der Waals options in CNS nbonds ⟨parameter-nbonds-statement⟩ controls repulsive potential for VDW repulsions. Prolsq_repulsion_function_changes in phenix.refine controls repulsive force. These antibumping repulsions Received: April 28, 2017 Published: June 23, 2017 1529

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534

Article

Journal of Chemical Information and Modeling

The protein atoms should satisfy the following criterion:

might push the atoms involving XBs away from each other. Hence, whether the backbone XBs are underestimated is an important question. However, there is no study, to the best of our knowledge, on the question. Here, we aim to explore the reliability of the backbone XBs and whether they are underestimated in PDB structures. Using a modified version of the ratio (RF) adapted from Taylor,27 the observed XBs’ frequency of occurrence to their frequency expected at random, i.e., if determined solely by the solvent accessible surface areas of the surrounding protein atoms, demonstrated that backbone XBs are largely overlooked in PDB, too. Furthermore, three complexes (cyclothiazide−glutamate receptor 2 (PDB IDs: 3tkd,28 3h6t,29 1lbc30), DMAT− ribosyldihydronicotinamide dehydrogenase [quinone] (PDB IDs: 4u7h7, 4u7f7), and RXP470.1−macrophage metalloelastase (PDB IDs: 4gql,31 5l7f,32 5l7932)) were discovered with backbone XBs in high resolution structures while lost in low resolution structures. In the last two complexes, even at 1.80 Å resolution, the backbone XBs were lost. These results manifested the urgent need to consider XBs in the refinement process during X-ray crystallography study.

dX − Z ≤ vdW(X) + vdW(Z) + 1.0 Å

(1)

where X = Cl, Br, I; Z is an arbitrary atom from protein backbone or side chain. 2.4. Statistics. For a given structure containing halogen atoms (X) and nucleophile (Y), the null hypothesis is stated as

S(Y ) ×

H 0: p =

40 180

S(total)

(2)

and the alternative hypothesis as

H1: p >

S(Y) ×

40 180

S(total)

(3)

where p is the probability that a halogen atom (X) will form XB with an atom of type Y, S(total) is the total SASA of all the atoms (Z) which could satisfy eq 1. Here, S(Y) is the SASA of atoms of type Y (i. e., backbone O and N atoms) which could satisfy eq 1. Therefore, the null hypothesis states that the probability of the formation of XB depends solely on the proportion of SASA contributed by Y. Also, 40 means that only the θ in the range of 180 140°−180° could form XBs. Let N be the number of halogen atoms (X: Cl, Br, I). Under the null hypothesis, the expected number of XBs is

2. MATERIALS AND METHODS 2.1. PDB Survey for Backbone XBs. The current PDB (March 2017 release) was explored in this study. The criteria for C−X···Y XBs are set as X···Y distances (d) shorter than the sum of vdW radii, and the C−X···Y angle (θ) is larger than 140° (Figure 1).24 Similar to the previous side-chain XB study,24 only the PDB structures with resolution equal or better than 3.0 Å were considered in this study.

E(XB) =

N × S(Y) × S(total)

40 180

(4)

The following statistic (RF, ratio of frequencies) can then be defined i=1

RF =

∑n O(XB)i i=1

∑n E(XB)i

where the summations are over all crystal structures that contain at least one XB. Here, O(XB)i is the observed number of XB in the ith structure, and E(XB)i is the expected number. Also, RF = x implies that XB occurs x times more often than expected by chance. 2.5. Bootstrapping. Bootstrapping was used to quantify uncertainty in the following steps. First, resample the PDB structures 1000 times with replacements. Second, calculate the percentage or RF from each sample and yield from the distribution in 1000 samples. Third, a percentile bootstrap interval was used to estimate the 95% confidence interval (CI). The data were presented as statistics (95% CI: lower limit, upper limit).

Figure 1. Geometrical parameters for backbone XBs.

2.2. Drug Targets Survey. Primary drug targets information from DrugBank33 that are mapped to PDB structures were retrieved at http://www.rcsb.org/pdb/ligand/drugMapping.do on March 14, 2017. 2.3. Solvent Accessible Surface Area (SASA). SASA is the surface area of a protein that is accessible to water molecules and accordingly accessible to halo-organics in the binding process. SASA of the protein atoms around halogens was computed by the get_area command in PyMOL (parameters: dot_solvent, 1; dot_density, 4). Since no refinement software during X-ray crystallography takes into account XB, a bump occurs if a halogen atom (X) and a nucleophile (Y) approach closer than the sum of their vdW radii. There should be a balance between the bump and the electron density. The halogen atom and protein atoms involving in backbone XBs might be pushed x Å (tolerance values) away from the sum of their vdW radii in the refinement process (Figure S1). In agreement with the previous study,27 1.0 Å was used in this study (eq 1), and different tolerance values were also evaluated.

3. RESULTS AND DISCUSSION 3.1. Most PDB Structures and Drug Targets Are in 1.5− 3.0 Å Resolution Range. Considering structural resolution, the PDB structures with ≤1.5 Å resolution should be most reliable due to their clear electron density. There are 113,888 X-ray structures in PDB (March 2017 release), only 10.5% (95% CI: 10.4%, 10.7%) in the ≤1.5 Å resolution range (Figure 2A). In addition, there are 1391 primary drug target PDB structures from X-ray experiments, just 9.5% (95% CI: 7.9%, 10.9%) in the ≤1.5 Å resolution range (Figure 2B). There are 4013 PDB structures containing organo-halogens (X = Cl, Br, I), only 8.8% (95% CI: 7.9%, 9.6%) in the ≤1.5 Å resolution range (Figure 2C) and just 8.9% XBs (95% CI: 6.9%, 10.9%) (Figure 2D) in the ≤1.5 Å 1530

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534

Article

Journal of Chemical Information and Modeling

other during the structural refinement, leading to the weakening or even disappearance of the backbone XBs. Therefore, the apparent decreasing trend of the RF values suggested that backbone XBs are underestimated in PDB. Furthermore, the 95% confidence intervals (CIs) of the RF were calculated by bootstrapping. The CIs are (19.9, 33.7), (11.6, 17.1), (11.7, 16.1), and (8.5, 19.0) for the resolution values of ≤1.5, 1.5−2.0, 2.0−2.5, and 2.5−3.0 Å, respectively. Impressively, the upper limit of the CIs in the 1.5−3.0 Å resolution range are still smaller than the lower limit of the CI in the ≤1.5 Å resolution range, which implies the significance of the underestimated XBs in the 1.5−3.0 Å resolution range. In addition, if we change the 1.0 Å in eq 1 to different values, i.e, 0.0, 0.5, 1.5, and 2.0 Å, similar results are yielded (Table S2). To investigate the role of the time line on RF, the PDB structures containing backbone XBs were divided into two equal parts according to deposition date. Similar trends of the RF ratios and 95% CIs to the overall period (Figure 3) were observed for both the first part (Table S3) and second part (Table S4), suggesting little effect of time line on RF. With a normalization factor (vdW(X) − vdW(Z) + dX − Z + 2.4) × (vdW(X) + vdW(Z) − dX − Z + 2.4) ) applied (

Figure 2. PDB structures at different resolution ranges. (A) All X-ray structures in PDB. (B) Primary drug target structures from DrugBank mapping to PDB. (C) PDB structures with organo-halogens. (D) Backbone XBs.

resolution range of the 845 backbone XBs (Table S1). Here, 83.4% (95% CI: 83.2%, 83.7%) of the X-ray structures in PDB are in the 1.5−3.0 Å resolution range (Figure 2A), while 81.4% (95% CI: 79.4%, 83.4%) of the primary drug targets are in the 1.5−3.0 Å resolution range (Figure 2B). One macromolecule might have several structures with different resolutions in PDB. So, we keep the ones with the highest resolution and reperform the survey which yields 80.4% (95% CI: 80.0%, 80.9%) of the X-ray structures in PDB in the 1.5−3.0 Å resolution range and 78.9% (95% CI: 76.4%, 81.5%) primary drug targets in the 1.5−3.0 Å resolution range (Figure S2). Taken together, the major information in PDB resides in the 1.5−3.0 Å resolution range. Therefore, whether the backbone XBs are underestimated and whether the resolution affects the reliability are key questions in the study. 3.2. Backbone XBs Are Underestimated in 1.5−3.0 Å Resolution Range. There are 845 backbone XBs (Table S1) in the current PDB database (March 2017 release) with a resolution no worse than 3.0 Å, 75 backbone XBs in the ≤1.5 Å resolution range, 226 in the 1.5−2.0 Å resolution range, 305 in the 2.0−2.5 Å resolution range, and 239 in the 2.5−3.0 Å resolution range. Statistical analysis of backbone XBs at different resolution ranges would shed some light on the question of whether the backbone XBs are underestimated. Here, RF, the observed backbone XBs’ frequency of occurrence to their frequency expected at random, was utilized in this study. The RF values are 25.7, 14.0, 13.7, and 13.0 for the resolution values of ≤1.5, 1.5−2.0, 2.0−2.5, and 2.5− 3.0 Å, respectively (Figure 3). Figure 3 reveals that as the resolution becomes worse, the RF values of XBs become lower. Because the XBs are considered as clash in the refinement progress, the atoms of XBs with clear electron density would not move too far away from each other in the ≤1.5 Å resolution range. However, the atoms of XBs with poor electron density in the 1.5−3.0 Å resolution range would be forced away from each

4(vdW(Z) + 1.4) × dX − Z

to SASA, the highest RF ratio was also found in the ≤1.5 Å resolution range (see SASA with a normalization factor in the Supporting Information, Figure S3 and Table S5). With these results taken together, we could state with confidence that backbone XBs are underestimated in PDB. 3.3. Backbone XBs with Different Amino Acids. All of the 20 amino acids could form backbone XBs (Figure 4). Leu is the most abundant amino acid (19%) in the 845 backbone XBs, Gly the second (13%), Ala the third (12%), Val the forth (10%), Gln and Glu tied for fifth (6%), Ser, Tyr, and Ile tied for seventh (4%), Phe, Thr, and Met tied for 10th (3%), Asp, Lys, His, Arg, and Cys tied for 13th (2%), and Pro, Asn, and Trp tied for 18th (1%). The most abundant four residues (Leu, Gly, Ala, Val) account for 54% of all the backbone XBs. The 95% CIs of the percentage for backbone XBs with different amino acids are presented in Table S6. Furthermore, we calculated the propensities of backbone XBs forming with different amino acids. The propensities were calculated as (number of amino acid A forming backbone XBs)/(number of A satisfying eq 1) × 100%. For statistical confidence, only amino acids that could form more than 10 backbone XBs in the ≤1.5 Å resolution range were included for analysis. With this criteria, three amino acids, viz., Gly, Val, and Leu, will be further analyzed (Figure 5 and Table S7). It is obvious that all of the three amino acids have the highest propensities in the ≤1.5 Å resolution range, while significantly lower in the 1.5−3.0 Å resolution range (Figure 5 and Table S7), which is consistent with the hypothesis that backbone XBs in the 1.5−3.0 Å resolution rage are underestimated in PDB. 3.4. Cases of Backbone XBs Lost in Low Resolution Range. In the PDB survey, we showed that as the resolution becomes worse, the RF values of backbone XBs becomes lower, suggesting the weakening and even the disappearance of the backbone XBs in the low resolution range. Many protein−ligand complexes have been deposited in PDB at different resolution ranges. If our postulation is correct, there might be some protein−ligand complexes housing backbone XBs in the high resolution range and losing the XBs in the low resolution range. Hence, we reperformed the PDB survey and found three proteins that form complexes with the same organo-halogens but with different resolutions. In the first protein, a marketed drug

Figure 3. Values of RF of backbone XBs at different resolution ranges. 1531

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534

Article

Journal of Chemical Information and Modeling

Figure 4. Frequency of backbone XBs forming with 20 amino acids at different resolutions.

found that another complex at 1.80 Å resolution crystallized at pH 6.5 also forms XB (d(Cl···N) = 3.00 Å, θ(C−Cl···N = 158.8°, Figure 6C, PDB ID: 1lbc30). With the three PDBs taken together, we could state that the real backbone XB between cyclothiazide and glutamate receptor 2 is lost in the low resolution range. In the second protein, 2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole (DMAT) forms a backbone XB (d(Br···O) = 3.01 Å, θ(C−Br···O) = 174.6°, Figure 7A) with Gly174 of ribosyldihydronicotinamide dehydrogenase [quinone] at 1.48 Å resolution (PDB ID: 4u7h7). However, this XB is lost at 1.80 Å resolution (d(Br···O) = 6.60 Å, θ(C−Br···O) = 114.7°, Figure 7B, PDB ID: 4u7f7). Three water molecules in 4u7f replace the bromide atoms in 4u7h with the displacement distances of 0.37, 0.47, and 1.33 Å (Figure 7C). If we extract the DMAT from 4u7h to 4u7f, there would be a nice agreement between DMAT and the electron density (Figure 7D). Therefore, the three water molecules might be the misassignment of the bromide atoms, and the electron density in 4u7f would correspond to two alternative conformations of DMAT, with one conformation housing a backbone XB. Although the largest problem concerns multiple location of halogenated ligands, the refinement algorithm might force an alternative conformation in the DMAT case. In the third protein, RXP470.1 forms a backbone XB (d(Cl···O) = 2.69 Å, θ(C−Cl···O) = 166.7°, Figure S5A) with Pro232 of macrophage

Figure 5. Propensities of backbone XBs forming with Gly, Val, and Leu at different resolutions.

cyclothiazide forms a backbone XB (d(Cl···N) = 3.16 Å, θ(C− Cl···N) = 153.2°, Figure 6A) with Asp248 of glutamate receptor 2 at 1.45 Å resolution (PDB ID: 3tkd28) with clear electron density (Figure S4A and C). However, if the electron density of the chloride atom becomes slightly unclear (Figure S4B and D), this XB is lost at 2.25 Å resolution (d(Cl···N) = 3.40 Å, θ(C−Cl··· N = 153.9°, Figure 6B, PDB ID: 3h6t29). When checking the crystallization experiments details, one may argue that 3tkd was crystallized at pH 4.5, while 3h6t at pH 6.5. The different pH values might result in different conformations. Further analysis

Figure 6. Backbone XBs in glutamate receptor 2 and lost in low resolution. (A) There is a backbone XB between Asp248 and cyclothiazide at 1.45 Å resolution (PDB ID: 3tkd, chain A). (B) XB is lost at 2.25 Å resolution (PDB ID: 3h6t, chain A). (C) XB exists at 1.80 Å resolution (PDB ID: 1lbc, chain B). (D) Superimposition of the above three systems. Glutamate receptor 2 is shown in gray cartoon. Asp248 and cyclothiazide are black sticks at 1.45 Å resolution, green sticks at 2.25 Å resolution, and red sticks at 1.80 Å resolution. Distance and angles are in black dashed lines. 1532

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534

Article

Journal of Chemical Information and Modeling

Figure 7. Backbone XB in ribosyldihydronicotinamide dehydrogenase [quinone] and lost in low resolution. (A) There is a backbone XB between Gly174 and 2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole (DMAT) at 1.48 Å resolution (PDB ID: 4u7h, chain B). (B) XB is lost at 1.80 Å resolution (PDB ID: 4u7f, chain B). (C) Superimposition of panels (A) and (B). (D) PDB 4u7f with DMAT extracted from 4u7h. Ribosyldihydronicotinamide dehydrogenase [quinone] is shown in gray cartoon. 2Fo−Fc electron density map contoured at the 1.0 σ level is shown in blue mesh. Gly174 and DMAT are black sticks at 1.48 Å resolution and red sticks at 1.80 Å resolution. Water molecules are shown in spheres. Distance and angles are in black dashed lines.

metalloelastase (MMP12) at 1.15 Å resolution (PDB ID: 4gql31). However, this XB is lost at 1.80 Å resolution (d(Cl··· O) = 6.42 Å, θ(C−Cl···O) = 36.7°, Figure S5C, PDB ID: 5l7f32) and also lost at 2.07 Å resolution (d(Cl···O) = 6.55 Å, θ(C−Cl··· O) = 38.9°, Figure S5B, PDB ID: 5l7932). These three cases showed that well-defined backbone XBs might vanish in a low resolution range (1.5−3.0 Å) in an intuitive way which is consistent with the above statistical results. Specifically, in the last two cases, backbone XBs are lost even at 1.80 Å resolution, which is seen as high quality structures in most cases, suggesting the urgent need to reinvestigate all the halogen-containing PDB structures.

Diagrammatic sketch of disappearance of an XB during refinement progress. X-ray protein structures of the highest resolution at different resolution ranges. SASA with a normalization factor. SASA of Z’s spherical cap. Backbone XB in glutamate receptor 2 and lost in low resolution. Backbone XB in macrophage metalloelastase (MMP12) and lost in low resolution. Backbone XBs (784) in PDB. Values of RF at different distance values. Ratios of RF and 95% CIs of 238 structures with the deposit date between 11/21/1991 and 06/15/2011. Ratios of RF and 95% CIs of 237 structures with the deposit date between 06/25/2011 and 02/17/2017. Ratios of RF and 95% CIs with a normalization factor applied. CIs (95% ) of the percentage for backbone XBs with different amino acids. CIs (95%) of propensities of backbone XBs forming with Gly, Val, and Leu at different resolutions. (PDF)

■

CONCLUSIONS Whether the backbone XBs forming between organo-halogens and protein backbone are underestimated is still a mystery. The quality of the backbone XBs are very critical to the development of the XB scoring function in drug discovery and design and critical to the interpretation of increase in specificity and affinity of halo-organics toward their molecular targets. Here, utilizing the ratio (RF) of the observed backbone XBs’ frequency of occurrence to their frequency expected at random, i.e., if determined solely by the solvent accessible surface areas of the surrounding protein atoms, we demonstrated that backbone XBs are largely overlooked in PDB. Furthermore, three cases were discovered with the backbone XBs existing in the high resolution range and lost in low resolution range. Considering our previous study that side-chain XBs are underestimated in PDB, we could conclude that the refinement software should be improved to deal with XBs during X-ray crystallography study.

■

■

AUTHOR INFORMATION

Corresponding Authors

*Phone: +86-21-50806600-1304. E-mail: [email protected] (Z.X.). *Phone: +86-21-50805020. Fax: +86-21-50807088. E-mail: [email protected] (W.Z.). ORCID

Zhijian Xu: 0000-0002-3063-8473 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Key R&D Program of China (2017YFB0202600 and 2016YFA0502800), National Natural Science Foundation of China (81273435), and Special Program for Applied Research on Super Computation of the NSFC-Guangdong Joint Fund (second phase) under Grant No. U1501501.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jcim.7b00235. 1533

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534

Article

Journal of Chemical Information and Modeling

■

(22) Lange, A.; Zimmermann, M. O.; Wilcken, R.; Zahn, S.; Boeckler, F. M. Targeting Histidine Side Chains in Molecular Design through Nitrogen-Halogen Bonds. J. Chem. Inf. Model. 2013, 53, 3178−3189. (23) Zimmermann, M. O.; Lange, A.; Zahn, S.; Exner, T. E.; Boeckler, F. M. Using Surface Scans for the Evaluation of Halogen Bonds toward the Side Chains of Aspartate, Asparagine, Glutamate, and Glutamine. J. Chem. Inf. Model. 2016, 56, 1373−1383. (24) Zhang, Q.; Xu, Z.; Zhu, W. The Underestimated Halogen Bonds Forming with Protein Side Chains in Drug Discovery and Design. J. Chem. Inf. Model. 2017, 57, 22−26. (25) Xu, Z.; Yang, Z.; Liu, Y.; Lu, Y.; Chen, K.; Zhu, W. Halogen Bond: Its Role beyond Drug-Target Binding Affinity for Drug Discovery and Development. J. Chem. Inf. Model. 2014, 54, 69−78. (26) Zimmermann, M. O.; Lange, A.; Wilcken, R.; Cieslik, M. B.; Exner, T. E.; Joerger, A. C.; Koch, P.; Boeckler, F. M. Halogen-Enriched Fragment Libraries as Chemical Probes for Harnessing Halogen Bonding in Fragment-Based Lead Discovery. Future Med. Chem. 2014, 6, 617−639. (27) Taylor, R. Which Intermolecular Interactions Have a Significant Influence on Crystal Packing? CrystEngComm 2014, 16, 6852−6865. (28) Krintel, C.; Frydenvang, K.; Olsen, L.; Kristensen, M. T.; de Barrios, O.; Naur, P.; Francotte, P.; Pirotte, B.; Gajhede, M.; Kastrup, J. S. Thermodynamics and Structural Analysis of Positive Allosteric Modulation of the Ionotropic Glutamate Receptor GluA2. Biochem. J. 2012, 441, 173−178. (29) Hald, H.; Ahring, P. K.; Timmermann, D. B.; Liljefors, T.; Gajhede, M.; Kastrup, J. S. Distinct Structural Features of Cyclothiazide are Responsible for Effects on Peak Current Amplitude and Desensitization Kinetics at iGluR2. J. Mol. Biol. 2009, 391, 906−917. (30) Sun, Y.; Olson, R.; Horning, M.; Armstrong, N.; Mayer, M.; Gouaux, E. Mechanism of Glutamate Receptor Desensitization. Nature 2002, 417, 245−253. (31) Czarny, B.; Stura, E. A.; Devel, L.; Vera, L.; Cassar-Lajeunesse, E.; Beau, F.; Calderone, V.; Fragai, M.; Luchinat, C.; Dive, V. Molecular Determinants of a Selective Matrix Metalloprotease-12 Inhibitor: Insights from Crystallography and Thermodynamic Studies. J. Med. Chem. 2013, 56, 1149−1159. (32) Bordenave, T.; Helle, M.; Beau, F.; Georgiadis, D.; Tepshi, L.; Bernes, M.; Ye, Y.; Levenez, L.; Poquet, E.; Nozach, H.; Razavian, M.; Toczek, J.; Stura, E. A.; Dive, V.; Sadeghi, M. M.; Devel, L. Synthesis and in Vitro and in Vivo Evaluation of MMP-12 Selective Optical Probes. Bioconjugate Chem. 2016, 27, 2407−2417. (33) Law, V.; Knox, C.; Djoumbou, Y.; Jewison, T.; Guo, A. C.; Liu, Y.; Maciejewski, A.; Arndt, D.; Wilson, M.; Neveu, V.; Tang, A.; Gabriel, G.; Ly, C.; Adamjee, S.; Dame, Z. T.; Han, B.; Zhou, Y.; Wishart, D. S. DrugBank 4.0: Shedding New Light on Drug Metabolism. Nucleic Acids Res. 2014, 42, D1091−1097.

REFERENCES

(1) Kolar, M. H.; Hobza, P. Computer Modeling of Halogen Bonds and Other σ-Hole Interactions. Chem. Rev. 2016, 116, 5155−5187. (2) Cavallo, G.; Metrangolo, P.; Milani, R.; Pilati, T.; Priimagi, A.; Resnati, G.; Terraneo, G. The Halogen Bond. Chem. Rev. 2016, 116, 2478−2601. (3) Cavallo, G.; Metrangolo, P.; Pilati, T.; Resnati, G.; Terraneo, G. Halogen Bond: a Long Overlooked Interaction. Top. Curr. Chem. 2014, 358, 1−17. (4) Clark, T.; Hennemann, M.; Murray, J. S.; Politzer, P. Halogen Bonding: the σ-hole. J. Mol. Model. 2007, 13, 291−296. (5) Politzer, P.; Murray, J. S.; Clark, T. Halogen Bonding and other σhole Interactions: a Perspective. Phys. Chem. Chem. Phys. 2013, 15, 11178−11189. (6) Persch, E.; Dumele, O.; Diederich, F. Molecular Recognition in Chemical and Biological Systems. Angew. Chem., Int. Ed. 2015, 54, 3290−3327. (7) Leung, K. K.; Shilton, B. H. Quinone Reductase 2 is an Adventitious Target of Protein Kinase CK2 Inhibitors TBBz (TBI) and DMAT. Biochemistry 2015, 54, 47−59. (8) Guerra, B.; Bischoff, N.; Bdzhola, V. G.; Yarmoluk, S. M.; Issinger, O. G.; Golub, A. G.; Niefind, K. A Note of Caution on the Role of Halogen Bonds for Protein Kinase/Inhibitor Recognition Suggested by High- And Low-Salt CK2α Complex Structures. ACS Chem. Biol. 2015, 10, 1654−1660. (9) Scholfield, M. R.; Zanden, C. M.; Carter, M.; Ho, P. S. Halogen Bonding (X-bonding): a Biological Perspective. Protein Sci. 2013, 22, 139−152. (10) Poznanski, J.; Shugar, D. Halogen Bonding at the ATP Binding Site of Protein Kinases: Preferred Geometry and Topology of Ligand Binding. Biochim. Biophys. Acta, Proteins Proteomics 2013, 1834, 1381− 1386. (11) Ho, P. S. Biomolecular Halogen Bonds. Top. Curr. Chem. 2014, 358, 241−276. (12) Parisini, E.; Metrangolo, P.; Pilati, T.; Resnati, G.; Terraneo, G. Halogen Bonding in Halocarbon-Protein Complexes: a Structural Survey. Chem. Soc. Rev. 2011, 40, 2267−2278. (13) Lu, Y.; Wang, Y.; Zhu, W. Nonbonding Interactions of Organic Halogens in Biological Systems: Implications for Drug Discovery and Biomolecular Design. Phys. Chem. Chem. Phys. 2010, 12, 4543−4551. (14) Auffinger, P.; Hays, F. A.; Westhof, E.; Ho, P. S. Halogen Bonds in Biological Molecules. Proc. Natl. Acad. Sci. U. S. A. 2004, 101, 16789− 16794. (15) Rose, P. W.; Prlic, A.; Altunkaya, A.; Bi, C.; Bradley, A. R.; Christie, C. H.; Costanzo, L. D.; Duarte, J. M.; Dutta, S.; Feng, Z.; Green, R. K.; Goodsell, D. S.; Hudson, B.; Kalro, T.; Lowe, R.; Peisach, E.; Randle, C.; Rose, A. S.; Shao, C.; Tao, Y. P.; Valasatava, Y.; Voigt, M.; Westbrook, J. D.; Woo, J.; Yang, H.; Young, J. Y.; Zardecki, C.; Berman, H. M.; Burley, S. K. The RCSB Protein Data Bank: Integrative View of Protein, Gene and 3D Structural Information. Nucleic Acids Res. 2017, 45, D271−D281. (16) Zimmermann, M. O.; Lange, A.; Boeckler, F. M. Evaluating the Potential of Halogen Bonding in Molecular Design: Automated Scaffold Decoration Using the New Scoring Function XBScore. J. Chem. Inf. Model. 2015, 55, 687−699. (17) Koebel, M. R.; Schmadeke, G.; Posner, R. G.; Sirimulla, S. AutoDock VinaXB: Implementation of XBSF, New Empirical Halogen Bond Scoring Function, into AutoDock Vina. J. Cheminf. 2016, 8, 27. (18) Liu, Y.; Xu, Z.; Yang, Z.; Chen, K.; Zhu, W. A Knowledge-Based Halogen Bonding Scoring Function for Predicting Protein-Ligand Interactions. J. Mol. Model. 2013, 19, 5015−5030. (19) Yang, Z.; Liu, Y.; Chen, Z.; Xu, Z.; Shi, J.; Chen, K.; Zhu, W. A Quantum Mechanics-Based Halogen Bonding Scoring Function for Protein-Ligand Interactions. J. Mol. Model. 2015, 21, 138. (20) Kolar, M.; Hobza, P.; Bronowska, A. K. Plugging the Explicit σholes in Molecular Docking. Chem. Commun. (Cambridge, U. K.) 2013, 49, 981−983. (21) Ford, M. C.; Ho, P. S. Computational Tools To Model Halogen Bonds in Medicinal Chemistry. J. Med. Chem. 2016, 59, 1655−1670. 1534

DOI: 10.1021/acs.jcim.7b00235 J. Chem. Inf. Model. 2017, 57, 1529−1534