Although halogen substituents are omnipresent in fragment screening libraries and drugs, their contribution to the inhibitor’s affinity is poorly understood to date. In rational drug design, only the impact of halogens’ steric and lipophilic influence has often been considered. In recent years, however, the topic of halogen interactions has received a lot of attention. This article explores the range of such interactions, which halogens form with biological macromolecules, mainly halogen bonding and multipolar interactions, with the obvious omission of the halides’ reactivity.
The observation of halogen bonding is not new, but it is not yet widely utilised in structure-based drug design. Compounds containing chlorine, bromine, or iodine can form contacts of the type R−X···Y−R′, where the halogen X acts as a Lewis acid and Y can be any electron donor moiety. This interaction, referred to as “halogen bonding”, is analogous to hydrogen bonding and is driven by a positively charged region, called the σ-hole, on the hind side of X along the R−X bond axis that is caused by anisotropy of electron density on the halogen.
Organic fluorines usually do not have the ability to act as halogen bond donors due to fluorine’s strong electronegativity. However, interactions between carbofluorines and the backbone oxygens of proteins are known. One group applied their analysis of the Protein Data Bank (PDB) and the evidence of short C−F···O contacts (3.0−3.7 Å) in protein−ligand complexes to their own work on the thienopyrimidine class of menin−MLL inhibitors. They found that their inhibitor formed multipolar C−F···O interactions between two CF3 groups and the carbonyls of the two structurally diﬀerent backbone conformations. This finding could be useful in the design of small molecule inhibitors targeting protein−protein interactions (PPIs). Moreover, these interactions had unique orthogonal geometry relative to the protein backbone. This observation could be utilised in drug design cases where the formation of hydrogen bonds might not be feasible.
A group of scientists has produced a number of publications on the role of halogens and their interactions in drug discovery. In their most recent paper in order to access the importance of the halogen-containing compounds in drug discovery, Zhijian Xu et al. analysed three databases, Thomson Reuters Pharma, ZINC (ZINC Is Not Commercial), and PDB (Protein Data Bank, April 2013 release), and came to some interesting results. The first observation was that the percentage of halogenated drugs is higher in clinical trials (average 34%) than in launched drugs (25%), indicating that organohalogens are more favoured nowadays. Secondly, the percentage of heavy organohalogens has increased from clinical trials to the launched phase, highlighting the attrition of organofluorides during the drug discovery process (Figure 1).
Figure 1. Composition of the organohalogens in diﬀerent stages during drug discovery and development.
From the analysis of 2,462 PDB structures that have heavy halogenated ligands, a quarter of structures were found to possess 778 halogen bonds. Of those, 82.4% halogen bonds are formed between heavy organohalogens and biomacromolecules and 16.6% are formed between heavy organohalogens and water molecules. The percentage of the latter should be even higher, since not all water molecules could be resolved in a structure with poor resolution. Among the 778 halogen bonds, C−X···Y (O, N, S) halogen bonds were found to be more prevalent (567) in biological systems, the others being C−X···π interactions. C−X···O halogen bonds account for 84.1% of C−X···Y halogen bonds with C−Cl···O halogen bonds contributing 43.6%, two thirds of which are formed to the protein backbone. C−X···Y can also be formed between a halogenated ligand and side chain groups, such as hydroxyls in serine, threonine, and tyrosine, carboxylate groups in aspartate and glutamate, sulfurs in cysteine and methionine, nitrogens in histidine. This plethora of diﬀerent possibilities in ligand−protein interactions makes halogen bonding a very useful tool to enhance ligand aﬃnities.
In heavy organohalogens, the size of the σ-hole increases with halogen size from chlorine to iodine, resulting in a stronger halogen bonding going from chlorine to iodine. Despite the increasing strength of halogen interactions, an assessment of the possible drawbacks of using bromine and iodine in drug discovery is important. Bromine and iodine atoms considerably increase the overall molecular weight. Yet, other parameters such as size, volume, or molecular surface area, do not change dramatically. The increase in lipophilicity with these atoms is also typically moderate. It is therefore unreasonable to assume that molecular weight-based models can be applied to bromine and iodine in order to characterize them correctly. Bromo and iodo groups are also perceived as problematic in drug design, particularly with respect to oxidative dehalogenation by cytochrome P450 enzymes. This fact, perhaps, explains the prevalence of fluoro- and chloro-containing drugs on the market. However, the choice of the core scaffold, that the halogen is attached to, also plays a role on the strength of halogen bonding, as Rainer Wilcken et al. discussed in their paper. The introduction of electron-withdrawing substituents on the halogen-bearing scaffold typically increases the strength of the halogen bonding and may also have an effect on the metabolic stability of the molecule, thus allowing the possibility to tune both properties.
Currently, there are very few computational molecular design packages that recognize halogen bonding as a favourable interaction, and they have not yet become part of the regular drug discovery workﬂow, a fact that impedes more widespread use and recognition of the phenomenon. Clearly, more attention should be given to heavy organohalogens than organoﬂuorine during the drug discovery stage.
R. Wilken et al. further investigated both computationally and experimentally whether an ethynyl moiety is a suitable bioisostere to replace labile iodine in ligands that form halogen bonds with the protein backbone. They found that the molecular electrostatic potentials for halobenzenes (Cl, Br, I) and phenylacetylene are remarkably similar in distribution of positive and negative charges of the C-X/H bond (where X=Cl, Br, I or ethynyl) (Figure 2a). Such bioisosteric replacement was successfully utilised in the EGFR inhibitors erlotinib (Figure 2b,c), as well as in a series of 1,4-benzodiazepine-2,5-dione inhibitors of the HDM2-p53 interaction.
Figure 2. Potential for bioisosterism between ethynyl and halogen substituents. (a) Electrostatic potentials plotted onto the isodensity surfaces at 0.003 au for chlorobenzene, bromobenzene, iodobenzene and phenylacetylene. Color ranges of energies in atomic units are also shown. Calculations were done at the MP2/TZVPP level of theory. (b) Structural formulae for gefitinib (left) and erlotinib (right), with the chlorine-to-ethynyl substitution highlighted. (c) Co-crystal structure of gefitinib bound to EGFR (PDB: 2ITY) in an overlay with the binding mode of erlotinib from PDB 4HJO. The geometry of the Cl···O halogen bond is highlighted in yellow.
However, in their own work the authors found that similar transformation led to an approximately 13-fold loss in affinity for the p53 cancer mutant Y220C inhibitor. The computational calculations suggested that this loss in affinity could be explained by the larger extent and the reduced directionality of the ethynyl group, and therefore is specific to the particular binding site. Nevertheless, halogen to acetylene bioisosteric transformation should be explored where applicable, thanks to other successful examples of drug development, e.g. Tarceva.
Blog written by Irina Chuckowree