The Different Ways through Which Specificity Works in Orthosteric and Allosteric Drugs

Abstract

Currently, there are two types of drugs on the market: orthosteric, which bind at the active site; and allosteric, which bind elsewhere on the protein surface, and allosterically change the conformation of the protein binding site. In this perspective we argue that the different mechanisms through which the two drug types affect protein activity and their potential pitfalls call for different considerations in drug design. The key problem facing orthosteric drugs is side effects which can occur by drug binding to homologous proteins sharing a similar binding site. Hence, orthosteric drugs should have very high affinity to the target; this would allow a low dosage to selectively achieve the goal of target-only binding. By contrast, allosteric drugs work by shifting the free energy landscape. Their binding to the protein surface perturbs the protein surface atoms, and the perturbation propagates like waves, finally reaching the binding site. Effective drugs should have atoms in good contact with the ‘right’ protein atoms; that is, the contacts should elicit propagation waves optimally reaching the protein binding site target. While affinity is important, the design should consider the protein conformational ensemble and the preferred propagation states. We provide examples from functional in vivo scenarios for both types of cases, and suggest how high potency can be achieved in allosteric drug development.

INTRODUCTION

Specificity is crucial in drug design [1]. Toxic effects of drugs are sometimes linked to their mode of action, such as unanticipated metabolic side effects of blocking (or stimulating) a particular protein or receptor; but more often they are the result of interactions between the drug and non-target molecules. Successful drugs are specific for a targeted receptor which is either over-expressed or malfunctions due to mutations. Specificity can be defined in a number of ways, for example, the degree to which the effects of a drug are due to the pharmacological action; whether it affects only certain tissues, whether it achieves selectivity in targeting only specific target sites, or the degree of a ligand’s affinity for a receptor. The specificity of drugs relates to their affinity. The affinity is a measure of how tightly a drug binds to the receptor. If the drug does not bind well, then the action of the drug will be shorter and the chance of binding will also be lower. The vast majority of drugs which produce pharmacological effects show a high correlation between structural matching and specificity. However, while specificity is a key factor in drug action, it does not necessarily translate into mechanistically-similar modes of action for different drug types. Below, we look at orthosteric and allosteric drugs. We posit that while specificity is of paramount importance in both drug types, the mechanisms through which it works are different. Accounting for the mechanisms in drug design can help in achieving more selective drugs with fewer side effects. Below, we outline the rationale, and provide examples. Because the majority of marketed orthosteric drugs show good potency and selectivity, we illustrate these principles by using real in vivo functional scenarios involving eukaryotic transcription initiation, and by allosteric drugs presenting high potency which has been achieved in drug development.

ORTHOSTERIC AND ALLOSTERIC DRUGS

Orthosteric drugs bind at the active site, competing with the natural substrate or ligand. If their affinity to the active site surface is high, they will out-compete the native substrate and block the site. Traditionally, most drugs on the market are orthosteric. On the other hand, allosteric drugs bind elsewhere on the protein surface [2–7]. Their action is indirect: they affect protein activity by changing the conformation at the second binding site. A major difference between the two types of drugs is in their potential effects: while orthosteric drugs stop protein activity entirely, this is not the case for allosteric drugs which can modulate the activity. Allosteric modulators do not compete with endogenous ligands and therefore can exert their influence even if an endogenous ligand is bound to another site on the same target at the same time. Allosteric drugs are advantageous because of two major reasons: first, they can offer a less disruptive way to influence the functioning of a pathway; and second, they are likely to have fewer side effects [3, 8–12]. Active sites are largely conserved across a protein family. Thus, a drug which binds to the active site of one protein may also bind to the active sites of the homologous family members. By contrast, allosteric drugs bind elsewhere on the protein surface. Surface regions are less conserved across families. Thus, for targets where it has been difficult to design selective orthosteric drugs, in some cases highly selective allosteric modulators can be made. Because of the problem of drug resistance, which results from mutations in the protein target that overcome the inhibition in both orthosteric and allosteric drugs, strategies that combine allosteric modulators with orthosteric drugs can be advantageous [4, 13–15].

HOW DO THE ALLOSTERIC DRUGS WORK? A VIEW FROM THE STANDPOINT OF THE FREE ENERGY LANDSCAPE

While the mode of action of orthosteric drugs is immediately obvious, the mechanism of action of allosteric drugs is less clear. To explain how allosteric drugs work, the most important principle, based on statistical mechanics, observed by nuclear magnetic resonance (NMR) and described in terms of the free energy landscape, is that dynamic proteins exist in conformational ensembles around their native states [16–23].The states have similar energies and they are separated by low barriers. Binding of a drug perturbs the protein structure, and creates strain energy Fig. (1). To understand the origin of the strain, consider the protein surface. In solution, surface atoms interact with water molecules. The water molecules are displaced at the drug interaction site: now, instead of interacting with water, some protein surface atoms interact with drug atoms. Optimizing these interactions causes surface atoms to move and rotate. This creates strain on the neighboring atoms which in turn need to get optimized as well. In this way the strain energy propagates like waves from the binding site across the molecule toward another surface region Fig. (1). If that other surface region falls at the ligand binding site, it may change the binding site conformation and (or) its dynamics, and in this way affect ligand binding, making it less (or more) favorable. Drug binding shifts the free energy landscape: conformations that were sparsely populated before can become more populated, and vice versa [16–17 19–21, 24]. In our case, the conformation that was favorable for the ligand binding may have very low population. This could be because of mutations, etc. However, the binding of the allosteric drug can stabilize it, making its population higher, which could lead to higher protein activity via interactions with the incoming ligand.

Fig. (1).

This simplified diagram illustrates the fundamental mechanisms of how allosteric drugs and orthosteric drugs work. A functional protein (blue) is represented by two binding sites. An allosteric drug (pink) binds to the allosteric site and an orthosteric drug (green) binds to the orthosteric site. The mechanism of inhibition depicted here illustrates that a drug binding event can prevent either an agonist or substrate (orange) from binding to the orthosteric site. To achieve inhibition, the orthosteric drug has to compete with an agonist or substrate for binding at the same site. On the other hand, the allosteric drug achieves inhibition via propagation of the strain energy (depicted by the red ‘bump’ and the change in the ellipsoid contacts) which is created at the binding interface, to the orthosteric site. This causes a conformational change at the orthosteric site and alters the binding affinity.

One scenario of a malfunctioning receptor which relates to allosteric pathways and could be restored is if a mutation occurred which is on a major allosteric pathway from the active site to a second site. In such a case, substrate binding at the active site works by allosterically affecting the conformation (or dynamics) of the second site, rather than by its blocking action. An example of such a case is the glycine receptor (GlyR) 1 subunit. Four mutations lead to the hereditary hyperekplexia neuromotor disorder. Among these, are the dominant mutations of residue 271, R271Q and R271L (αR271Q/L) [25]. Residue 271 is located on the channel pore. The mutations reduce the receptor’s sensitivity to the glycine agonist [26–27]. R271 and residues in its vicinity have been observed to undergo a conformational change which is coupled with binding at the pore and channel opening in the intracellular domain [28], which suggests that they are located on a major allosteric signaling pathway which links the agonist binding site to the channel gate. Residues lying along the allosteric communication trajectory have also been identified using high-order thermodynamic coupling analysis of voltage-dependent gating in the voltage-activated K⁺ channel [29]. These combine with a local network to constitute the allosteric trajectory [30]. An allosteric drug which interacts with residues in the vicinity of mutations of these on-pathway residues, and which elicits a pathway that converges with the debilitated ones, can rescue the receptor function.

SPECIFICITY

Here we address the question of whether the different modes of drug action in orthosteric versus allosteric drugs can imply differences in the mechanisms through which specificity works. This question is important because of implications to drug design. We argue that specificity can work differently for these two cases: high affinity orthosteric drugs can be expected to lead to fewer side effects; by contrast, in allosteric drugs the key consideration is the way in which they alter the propagation pathways. Fig. (2) illustrates these differences.

Fig. (2).

This simplified diagram illustrates how allosteric drugs and orthosteric drugs achieve specificity to avoid side effects. As in Fig. (1), the functional homolog (blue) is also represented by two binding sites with the allosteric drug (pink) binding to an allosteric site and the orthosteric drug (green) to the orthosteric site. Because binding sites in homologous proteins are typically more similar than other locations on the protein surfaces, in the drawing the shapes of the orthosteric sites are more similar than those of the allosteric sites. Thus, to achieve specificity for a specific binding site, orthosteric drugs generally need to undergo fine-tuning. This is illustrated by dark green shape complementarity between the homologous sites and their respective orthosteric drugs. By contrast, the allosteric drug can achieve specificity more easily, because the allosteric sites are more different among the protein homologs.

1. SPECIFICITY IN ORTHOSTERIC DRUGS Fig. (2)

Drug discovery aims to design drugs which successfully bind the target protein and modulate the pathway without side effects. If the concentration of the drug is high, it will bind to the target protein as well as to other similar binding sites in homologous members of the protein family. Because the functions of these proteins can vary broadly, this can be expected to lead to unwanted side effects. On the other hand, if the concentration of the drug is low, it will bind only to those proteins whose binding sites display high affinity to the drug. This is expected to lead to highly specific binding. The binding of transcription factors (TFs) to DNA regulatory elements (REs) present a good example illustrating such scenario [31]. A transcription factor recognizes a large number of redundant REs. Each of the REs is associated with a specific gene; thus specific recognition of a given RE at a certain time is critical to turning genes on/off. However, there are ~3000 TFs in the human nucleus and a large number of DNA sequences that match the TF binding site from which the TF has to choose. On the human genome scale, with only four types of DNA bases, a 6 base pair RE would occur >700 000 times, and considering RE degeneracy this might be an underestimation. This argues for an immense number of possible REs for the same and different TFs. Key questions are how does a TF recognize its REs among the many similar ones, and how is this selectivity achieved? Several factors are thought to play a role including the extent of the similarity between the RE sequence and the consensus sequence of the TF, and the concentration of the TF in the cell [31–32]. If the concentration of the TF is high, it will bind all REs; however, if its concentration is low, it will bind only those with the highest affinity. The principles followed by these in vivo scenarios are similar to the orthosteric drug case: here a given drug (equivalent to a TF) can bind many homologous proteins (equivalent to REs fitting a consensus sequence); however, if the drug concentration is low, only those binding sites which have the highest affinity will selectively bind.

2. SPECIFICITY IN ALLOSTERIC DRUGS Fig. (2)

Affinity matters in orthosteric and allosteric drug binding. However, there are additional considerations in the allosteric drugs case. Allosteric drugs work through perturbation which is initiated by the physical interactions between drug and protein atoms at the binding site [33]. The perturbation propagates in the structure through multiple pathways. If the propagation pathways reach the ligand binding site and affect its shape or dynamics in a sufficiently high population of protein conformers, the drug is a candidate for consideration for additional tests and development. Two key factors determine which pathways are the major and which are the minor: (i) the protein atoms which are in contact with the drug atoms and the contact types and extents; and (ii) the protein conformational ensemble. In our case the proteins belong to the same family, and thus their global structures are expected to be similar; however, because at the same time there is also some sequence variability, the distributions of the conformational states are likely to be different. If (i) different protein atoms are in contact with the drug among the homologous proteins, the pathways can initiate from (slightly) different locations, or the perturbations could be to (slightly) different extents, which could nonetheless lead to altered pathway distributions. Similarly, (ii) sequence changes, involving residue substitutions, deletions, or insertions can shift the free energy landscape, which could favor some pathways over others [16–21]. These changes in the pathways lead to a re-distribution of the ensemble. The outcome is different preferred conformations of the ligand binding site and consequently ligand binding and protein activity. Thus, in allosteric drugs specificity works through the specific contacts of drug and receptor atoms. Two drugs may have similar (high) affinities for the protein surface; however, their specific allosteric effect on the ligand binding and protein activity may be different, because they elicit different propagation pathways Fig. (2).

To illustrate this point, we again resort to an in vivo example, taken from RNA polymerase II (Pol II) transcription activation. Mediator is a multi-protein complex that functions as a transcriptional coactivator. The Mediator complex is required for the transcription of large numbers genes in yeast and mammals. The human Mediator complex is gigantic: it has 26 subunits and is 1.2 MDa in size. Mediator has three modules: Head, Middle and Tail. It binds the activation domain of transcription factors, and the RNA Pol II complex, which includes the general transcription factors (GTFs), RNA Pol II and the TATA binding protein (TBP). It is essential for TF activation-dependent transcription. Recent compelling data illustrated how the p53 TF can activate transcription via Mediator’s ‘structural shifts’ [34]. The data showed that Mediator undergoes a dramatic conformational change upon TF binding [35–37]. The p53 activator domain (p53AD) and p53 C-terminal domain (p53CTD) interact with Mediator subunits MED17 and MED1, in the Head and Middle modules, respectively. The interactions affect Mediator structure and Pol II activity in different ways: p53AD induces conformational change and this correlates with activation of stalled Pol II; however, this is not the outcome of the p53CTD-MED1 interaction. Of particular relevance to our drug case here, mutations of p53AD residues (L22Q and W23S) which prevent expression of most p53 target genes [21, 38–39] disrupted the MED17 interaction. The authors concluded that Mediator undergoes p53AD-induced shape-shifting, which is expressed in Mediator-Pol II binding. Thus, p53 binds an RE far away from the promoter. The RE allosterically elicits a pathway through the TF DNA-binding domain, then via its activation domain, finally reaching the Mediator, and changing the Mediator conformation at the Mediator binding site with Pol II/GTFs, to activate transcription. However, the mutations in the p53AD shift the free energy landscape, and the preferred pathways; under these circumstances, rather than the p53AD binding MED17 in Mediator Head, now it binds MED1 in the Middle module. Coming back to our case here, the p53 can be considered as the drug, and the Mediator as the protein receptor. p53 mutants can be considered as different drugs. Thus, the different conformations of p53 mutants (different drugs) presented altered interactions with Mediator (different interactions with the receptor). These led to different pathways in Mediator, which in turn affected differently Mediator binding site with the Pol II machinery. In our case, this difference in drug interaction leads to different pathways in the receptor, with altered outcome at the ligand binding site.

EFFICIENCY AND POTENCY IN ALLOSTERIC DRUGS

The mechanism of orthosteric drugs is well studied, and there are many examples including the majority of the marketed drugs, which show good potency and selectivity. Allosteric drugs are more complicated as compared to orthosteric ones. Further, because the allostric sites are often shallower and narrower than orthosteric sites, the binding affinities of allosteric drugs present a major challenge for this type of drug design. Among the examples, in kinases, allosteric drugs can work by conformational changes that block productive ATP binding; these act in an ‘ATP competitive’-like manner; or by blocking kinase activation through conformational changes that are ‘ATP non-competitive’ [40]. For GPCRs, allosteric mechanisms can lead to multiple modes of target modulation: positive allosteric modulation (PAM), negative allosteric modulation (NAM), neutral cooperativity, partial antagonism (PA), allosteric agonism and allosteric antagonism [41–44]. Specific examples also include Gleevec (allosteric inhibitor of Abl) [40], Cinacalcet (allosteric activator of calcium sensing receptor) [11] and Maraviroc (allosteric inhibitor of chemokine receptor 5) [11]. Modulators of cAMP signaling include small-molecule allosteric drugs targeting phosphodiesterase 4 (PDE4), the primary cAMP-hydrolyzing enzyme in cells. Because they do not completely inhibit enzymatic activity, they have reduced potential to cause emesis, while maintaining biological activity. As such they can help central nervous system therapeutics where penetration to the brain is desired [45].

Anti-hypertensive agent ifenprodil has neuroprotective activity through its effects on NMDA (N-methyl-D-aspartate) receptors. NMDA receptors’ ion-channel activity is allosterically regulated by binding of small compounds to the amino-terminal domain (ATD) in a subtype-specific manner. Ifenprodil and related phenylethanolamine compounds specifically inhibit NMDA receptors. The phenylethanolamine binds at the interface between the GluN1 and GluN2B receptors which form heteromeric ion channels, rather than in the GluN2B cleft [46]. Of particular interest, restraining domain movement of the GluN2B ATD by inter-subunit disulphide bond, decreases the sensitivity to ifenprodil, which suggests that this biases the subunits orientation with respect to each other [47] in a way which is non-productive for ifenprodil-mediated allosteric inhibition of NMDA receptors. In another example, a series of thiazolone-acylsulfonamides act as hepatitis C virus (HCV) NS5B polymerase allosteric inhibitors. The structure-based drug designs were guided by computational docking that discovered an additional pocket in the allosteric site. The molecules contain moieties of thiazolone and a newly designed acylsulfonamide linker connected to a substituted aromatic ring. The compounds demonstrated low μM activity [48]. The positive allosteric modulator morantel binds at non-canonical subunit interfaces of neuronal nicotinic acetylcholine receptors [49] and allosterically enhances channel gating [50]. p38 MAPK inhibitory BIRB-796 [51] provides an additional example.

As we noted above, allosteric modulators are often much more selective than active site inhibitors against enzyme super families with highly conserved active sites; in addition, because they do not have to compete with the substrate for binding, they can be much more potent than competitive inhibitors when the substrate concentrations are high. Allosteric ligands could be particularly useful in controlling receptors for which the design of selective orthosteric agonists or antagonists has been elusive, such as muscarinic acetylcholine receptors [52].

CONCLUSIONS AND OUTLOOK

Specificity is a key factor in drug design. Here we highlight an aspect which has been largely overlooked: that is, specificity works differently in orthosteric and allosteric drugs. In orthosteric drugs, high specificity implies high affinity and selectivity, to avoid unwanted side effects. This points to drugs whose shape and chemistry provides a perfect match to the receptor active site. By contrast, in allosteric drugs, high specificity implies eliciting allosteric pathways which efficiently propagate to the ligand binding site. This suggests that figuring out the appropriate contact points with the receptor’s atoms is a key factor in allosteric drug design. Predicting which contact points with receptor atoms are advantageous in allosteric drug modulation effects is a challenging task. Nonetheless, methods addressing this challenge are increasingly being developed; for example, computationally this could be done via correlations of the fluctuations of residues [53], and experimentally assays using non-competitive ligands have been proposed [54]. Small molecules can also be used as probes [55].

In an engaging piece in 1999, Marc Van Regenmortel has pointed out that analysis of the interactions between biomolecules cannot be reduced to descriptions of their static molecular structures [56]. He further noted that “All binding sites are relational entities defined by their partners and not by intrinsic structural features identifiable independently of this relationship”. This description holds for orthosteric and allosteric drug binding. In both cases, ensembles need to be considered, and the drug will select a complementary favorable conformer-partner, and a population shift of the ensemble will follow [57]. The difference is in the target sites and consequently in the mode of action. Here we propose that a mechanistic view of how specificity works in different drug types may help in achieving potent drugs, with fewer side effects. We further posit that mechanistic classifications would help not only in the understanding of fundamental features of protein function [20] but also assist in forecasting drug action. While the native selection mechanism across evolution is blind to structure and selects for functional effects [56], successful drug design needs to account for the conformational distributions and the targeted mechanisms of drug action. Successful drug design needs to account for the conformational distributions and the targeted mechanisms of drug action [58].