CHAPTER 1: Modulation of Biological Targets Using Allosteric Ligands: Food for Thought

The dawn of the new century found several major pharmaceutical companies facing significant challenges in keeping their pipelines populated with new drugs. For nearly two decades, robust increases in research and development investments had not produced the hoped for number of new medical entities (NMEs) approvals per year. In some therapeutic areas such as diseases of the central nervous system (CNS), and in spite of a growing patient population and major medical needs, only a handful of new drugs were approved during this period, and mostly based on mechanisms already known (“me too” drugs). As a response to this crisis of innovation, and propelled by major discoveries in molecular and cell biology research, in particular in the areas of oncology and immunology, a chemically broad range of new therapeutic modalities emerged, such as antibodies, proteins, nucleic acids, vaccines, cell, or gene therapies. These are generally grouped under the denomination “biologics”, to differentiate them from “small molecules”. Drugs in the latter group have historically been mostly thought of as acting competitively with endogenous ligands, binding at their target receptors in a model that became known as “lock and key”.¹  These two types of drugs differ in terms of a number of attributes, such as molecular weight, preparation, characterization of the active pharmaceutical ingredient (API), physicochemical properties, possible route of administration, pharmacokinetic distribution, metabolism, clearance mechanisms, drug–drug interactions, dosing regimen, safety and toxicology, antigenicity and hypersensitivity, and pharmacology – including side effects.^2,3  A comparison of these attributes highlights some of the challenges associated with the development of biologics as drugs. (Table 1.1) It also begs the question whether new strategies could be discovered to modulate drug targets with a different type of small molecules. This is, in essence, the promise of what allosteric modulators might do.

Broadly, allostery has been defined as an indirect interaction between topographically distinct binding sites in a protein, mediated by a conformational change. Monod and Jacob used the term for the first time in the printed version of the Proceedings of the 26th Cold Spring Harbor Symposium on Quantitative Biology entitled “Cellular Regulatory Mechanisms” in 1961, in the general discussion written as a conclusion of the conference. The earliest appearance of the term “allostery” in a publication listed in PubMed dates to more than half a century ago.^4,5  Since then, the concept has evolved and its use expanded notably. Today, allosterism is considered an inherent property of all dynamic proteins and other macromolecules of biological relevance.⁶  Examples of important drugs exist that are allosteric modulators of diverse targets such as ion channels, G protein-coupled receptors (GPCRs), and enzymes. Benzodiazepines, one of the earlier drug classes with major impact on society, act as negative (flumazenil, 1) or positive (diazepam, 2; flunitrazepam, 3) allosteric modulators of gamma aminobutyric acid A (GABA_A) ion channels. The first inhibitors of human immunodeficiency virus reverse transcriptase enzyme (HIVRT) useful in the treatment of AIDS were nucleoside analogs. These compounds act competitively and inhibit the binding of the appropriate deoxynucleotide triphosphate (ATP, TTP, GTP, or CTP) to the reverse transcriptase (RT) enzyme. Nevirapine (4) and efavirenz (5) are non-nucleoside RT inhibitors (NNRTIs). Both nucleoside and non-nucleoside RTIs inhibit the same target, the reverse transcriptase enzyme, an essential viral enzyme which transcribes viral RNA into DNA. Unlike nucleoside RTIs, which bind at the enzyme's active site, NNRTIs bind allosterically at a distinct site away from the active site called the NNRTI pocket. Maraviroc (6) is another antiretroviral drug used in the treatment of HIV infection, which works through a negative allosteric modulation (NAM) mechanism inhibiting the cell-surface chemokine CCR5 receptor. The chemokine receptor CCR5 is a co-receptor for most HIV strains, necessary for the entrance of the virus into the host cell. The drug binds to CCR5 and changes its conformation, thus blocking the HIV protein gp120 from associating with the receptor. HIV is then unable to enter human macrophages and T-cells. Cinacalcet (7), a calcimimetic, is a positive allosteric modulator (PAM) of the calcium-sensing receptor (CaSR), which is activated according to the natural rise and fall of endogenous calcium ion levels. This drug acts by lowering the threshold for activation of feedback on the parathyroid chief cells. Trametinib (8) is a highly selective reversible allosteric inhibitor of mitogen-activated protein kinase (MAPK)–extracellular-signal-regulated kinase enzymes MEK1 and MEK2 activity. Compound 8 is an ATP non-competitive inhibitor that binds MEK at a location adjacent to the ATP binding site. Approved in 2013, it is a first-in-class oral treatment for metastatic melanoma with BRAF V600 mutations (Figure 1.1).^7,8 

In addition to these marketed drugs, a number of compounds working through allosteric mechanisms are undergoing advanced clinical trials at the time of writing. The serine–threonine kinase AK thyoma oncoprotein homolog (Akt), also known as protein kinase B (PKB), plays a key role in the phosphoinositide 3-kinase–Akt–mammalian target of rapamycin (PI3K–Akt–mTOR) signaling cascade, tightly associated with cell proliferation, survival, migration, angiogenesis and other activities directly linked to tumor genesis and progression. The allosteric Akt inhibitor MK-2206 (9) is currently being evaluated in clinical Phase 1 and 2 studies on breast, colon/rectal and other cancers.⁹  The muscarinic M1 receptor positive allosteric modulator (PAM) 1-(4-cyano-4-(pyridine-2-yl)piperidin-1-yl)methyl-4-oxo-4H-quinolizine-3-carboxylic acid (PQCA) (10) has shown in vivo efficacy in a number of preclinical models in both rodents and non-human primates, suggesting this mechanism may have therapeutic potential for the treatment of Alzheimer's disease and cognitive deficits.¹⁰  The metabotropic glutamate receptor 4 (mGluR4) PAM Lu AF21934 (11) has been extensively studied in a number of in vitro and in vivo mechanistic tests. It differentiates from congeners by its remarkable ability to enhance in vitro agonist concentration–response curves for both, the mGluR4 homodimer as well as the mGlu2/4 heterodimer, as opposed to the earlier tool compounds, which are functionally selective for the mGluR4 homodimer (Figure 1.2).¹¹ 

In some cases, strong support was obtained using X-ray crystallography to enable the discovery of allosteric drugs. Phosphodiesterase 4 (PDE4) is an enzyme that hydrolyzes cAMP in cells, and a promising drug target for the treatment of Alzheimer's disease, Huntington's disease, schizophrenia, and depression. Drugs treating these diseases by aiming at central biological targets must cross the blood–brain barrier (BBB) and penetrate into the brain to exert therapeutic benefit, which restricts the available chemical space for potential competitive inhibitors. PDE4 inhibitors exist which are cAMP-competitive marketed drugs, but the lack of subtype selectivity is thought to lead to side effects, including nausea, depression, and weight loss. Studying crystal structures of PDE4 with bound inhibitors exposed the structural basis of this enzyme’s regulation and inspired, together with mutagenesis and kinetic studies, the design of subtype-selective negative allosteric modulators of PDE4D with reduced potential to cause the dose-limiting side effects observed with existing active-site-directed PDE4 inhibitors.¹²  BPN14770, a first in class PDE4D NAM (structure not disclosed at present time) is being developed with support from the NIH Blueprint Neurotherapeutics Network and is the first compound funded by the program to reach a Phase 1 clinical trial.¹³  Likewise, the recent disclosure of crystallographic studies with GPCR ligands confirmed their allosteric nature. Notable examples are the corticotropin-releasing factor-1 (CRF-1) functional inhibitor CP-376395 (12),¹⁴  the mGluR5 NAM mavoglurant (13),¹⁵  the mGluR1 NAM FITM (4-fluoro-N-(4-(6-(isopropylamino)pyrimidin-4yl)thiazol-2-yl)-N-methylbenzamide) (14),¹⁶  and the P2Y₁ receptor non-nucleotide antagonist BPTU (1-(2-(2-(tert-butyl)phenoxy)pyridin-3-yl)-3-(4-(trifluoromethoxy)phenyl)urea) (15). The latter has the peculiarity of being the first structurally characterized selective GPCR ligand located entirely outside of the helical bundle (Figure 1.3).¹⁷ 

These last several examples strongly suggest that the use of allosteric ligands has the potential to help solve druggability issues in areas where ligands binding at the orthosteric site present challenges, thus providing a new chemistry modality to deliver therapeutic treatments using small molecules, with which chemists are highly familiar working. Yet, the purposeful targeting of allosteric sites in proteins that are biological targets of drug discovery efforts is a relatively recent undertaking in drug discovery projects.

The pharmaceutical industry has witnessed a major change over the last decade, as it went from working on relatively well-understood biological targets with clinical validation, where efficacy in the patient population was expected, to novel biological targets with unprecedented track records in clinical settings. The reasons for this change are not only scientific, and include social, political, and financial factors, which are beyond the scope of this book. The bottom-line is that today, unmet medical needs and financial incentives dictate the discovery of drugs with enhanced efficacy over existing treatments. A way to achieve this goal is by using novel mechanisms of action. However, the novelty entails a risky task: demonstrating clinical efficacy in a broad patient population, usually requiring Phase 3 clinical studies. Late failures are very expensive, and while they might be explicated by our imperfect understanding of disease etiology and flawed replication of human (patho)physiology in the preclinical disease models used, that is no consolation after a billion-dollar drug development exercise yields negative clinical data. The current premise is that translational sciences will, at least in part, mitigate this risk and enhance success rates.

Allosteric drugs were envisioned as having the potential to provide two ways to mitigate the risk of late clinical failure:

1. By providing an alternative molecular mechanism to attack biological targets (clinically validated or not) that require a different, friendlier drug-like chemistry space than originally attempted with orthosteric ligands (e.g., GABA, glutamate or acetylcholine receptors); and

2. Giving novel ways to attack previously intractable targets (e.g., the calcium sensing receptors or sweet taste receptors; voltage gated receptors).

Therefore, it seemed reasonable that some of the same biological targets previously aimed at, and which had been proven recalcitrant to drug discovery efforts using orthosteric ligands, yet not technically invalidated,¹⁸  were given a second life and addressed using allosteric ligands.

A number of excellent publications have discussed in great detail the potential advantages of using allosteric ligands in drug discovery, including comparisons with their orthosteric counterparts.^19–23  A review of this literature consistently points to a variety of factors as drivers to encourage the exploration of allosteric ligands as drugs and chemical probes. These include factors such as: being easier to achieve selectivity, lack of intrinsic activity, ability to “dim” endogenous agonist function while preserving spatial and temporal effects, etc. (Figure 1.4). Some of these initial views have evolved as they were challenged by experimental data, suggesting a reconsideration of these idealized properties.

Target selectivity is generally thought to be easier to obtain with allosteric ligands than with orthosteric ligands. This belief would be a consequence of a highly conserved amino acid sequence in the orthosteric binding site. Remote allosteric sites tend to be less well conserved, and sufficiently different that selectivity for allosteric ligands may be expected. For example, when comparing different mGluR subtypes binding the same endogenous agonist (glutamate), the amino acids forming the binding site are indeed highly conserved.^24,25  Furthermore, these allosteric sites are thought to be an evolutionary artifact and less sensitive to mutations in terms of the receptor functional output, as there would be no endogenous ligands binding at such sites. This would make the design of selective orthosteric ligands somewhat challenging.

However, this thinking may in part be a consequence of how orthosteric analogs were conceived in the past. Historically, these ligands were often designed by constraining the conformation of the endogenous natural ligand and keeping a similar ligand size. Thus, the same amino acids are involved in the binding process of both the endogenous ligand and the putative drug. Indeed, recent work has shown that orthosteric ligand selectivity may be achieved by making analogs with added substituents, so that additional amino acids that may no longer be conserved among receptor subtypes are involved in the binding process. Examples of such selective ligands are LSP4-2022 (16,²⁶  mGluR4-selective) and LY2812223 (17,²⁷  mGluR2-selective) (Figure 1.5; glutamate backbone shown in color). Furthermore, evidence suggesting that allosteric sites might play a role in determining receptor function came from the study of the GABA_A receptor, and the discovery of endogenous ligands binding at the benzodiazepine site, including oleamides, nonpeptidic endozepines and the protein diazepam-binding inhibitor (DBI).²⁸ 

Potent and selective adenosine A₁R receptor orthosteric agonists and antagonists have also been discovered and tested in clinical settings for a number of indications.²⁹  The use of bulky N⁶-adamantyl substituents on adenosine has led to agonists previously shown to be A₁R-selective with respect to binding affinity at rat receptors.³⁰  Recently, the discovery of a series of subtype-selective N⁶-bicyclic and N⁶-(2-hydroxy)cyclopentyl derivatives of adenosine as agonists was reported and their A₁R/A₂R selectivity assessed based on a modified Saccharomyces cerevisiae strains screening platform. In addition, the activity at A₁R and A₃R of a few select analogs which were active against the A₁R in the yeast screen was studied in mammalian CHOK1 cells; compounds such as 18 and 19 (Figure 1.5) were shown to be highly selective A₁R agonists. Preferred compounds based on potency and selectivity contained N⁶-adamantyl substitution in combination with 5′-N-ethylcarboxamido or 5′-hydroxymethyl groups.³¹ 

In some cases, allosteric ligands have provided a path forward to selectively activate CNS receptors when orthosteric ligands failed to do so. One such example is the use of compounds 10, 20 and 21 (Figure 1.6), characterized as having a mixed agonist–positive allosteric modulator profile highly functionally selective for the M1 receptor. M1 agonists such as xanomeline produce beneficial cognitive effects in Alzheimer's disease and schizophrenia patients. Unfortunately, its therapeutic use is limited due to cholinergic side effects (sweating, salivation, gastrointestinal distress). These are suggested to result from nonselective activation of other muscarinic receptor subtypes such as M2 and M3. However, this theory was refuted as the M1-selective activators 10, 20 and 21 tested in rats, dogs, and cynomologous monkeys still elicited cholinergic side effects such as salivation, diarrhea, and emesis.³²  This suggests that activation of M1 receptors alone is sufficient to produce unwanted cholinergic side effects such as those seen with xanomeline.

Another example of an idealized property of allosteric modulators is that they frequently exhibit little or no intrinsic activity, since their mode of action is to enhance or inhibit the action of the endogenous agonist.³³  The observation of agonist-independent activity (allosteric agonism) by allosteric modulators is a rather complex phenomenon. Studies have shown that the specific nature of the biological system under interrogation may play a role in the functional response observed to the same PAM. For example, higher levels of receptor expression in a cell line under investigation may facilitate the observation of intrinsic activity, as was reported for the case of the glucagon-like peptide 1 (GLP-1) receptor (vide infra).³⁴  Small modifications to the chemical structure of an allosteric potentiator may provide compounds in the same chemotype with varying degrees of agonist-independent activity, including none. Thus, distinct drug target product profiles may be defined. Allosteric agonists or “ago-PAMs” are compounds with intrinsic activity independent of the endogenous ligand. On the other hand, “pure PAMs” show activity only in the presence of the orthosteric agonist and no activity in its absence. In any case, medicinal chemistry optimization to a preferred target product profile is feasible.³⁵ 

Allosteric ligands are also thought to allow fine tuning of functional responses. While this may be the case for properly designed and characterized PAMs, most notably for mGluR5 PAMs,^36,37  it is also possible to achieve super-physiological functional responses, above what the endogenous natural system may deliver.³⁸  This design attribute may be an advantage in diseases where the tone of the endogenous agonist is high but the receptor functional response needs augmentation.³⁹ 

As acknowledged by Dougall and Unitt, the “potential advantages of allosteric modulators remain largely theoretical as very few such agents have to date reached the market”.⁴⁰  About a decade into conducting research with the intent to engineer certain attributes into allosteric drugs using structure–activity relationships (SAR) as has been done with orthosteric ligands, it seems prudent to maintain some level of skepticism regarding the extent to which this potential has been realized.

Once the idea of modulating the functional activity of a biological target using allosteric drugs found support in reliable early experimental data, it also became clear than the concept was more complex than initially thought. Empirically, differences could be found between developing SAR for allosteric compounds that are inhibitors (NAMs) or enhancers (PAMs) of biological target function. Furthermore, subtler aspects further differentiate different types of allosteric enhancers. For enzymes, K-type modulators affect the affinity of the substrate, while V-type modulators affect the catalytic rate. Distinct cases (Type II, Types Ia and Ib) generate unique kinetic signatures which may be of diagnostic value.⁴¹  For membrane-bound targets (e.g., GPCRs, transporters), some compounds may act by enhancing the affinity of the orthosteric ligand (K_d), others by increasing the efficacy at the receptor (E_max), or by acting as allosteric agonists activating the receptor in the absence of orthosteric agonist (e.g., Figure 1.7). In practice, compounds with a combination of these modes of action are often encountered.⁴²  Indeed, examples of allosteric modulators exist which enhance the binding affinity of the orthosteric ligand (attribute of a PAM) while reducing the receptor efficacy (attribute of a NAM).⁴³ 

One of the models used to describe the effect of an allosteric ligand on the functional response elicited by the action of an orthosteric ligand at a receptor is the Operational Model.⁴⁴  Mathematically, this model can be represented by eqn (1.1), where pK_A and pK_B are the affinity of orthosteric and allosteric ligand for the free receptor, respectively; α is the affinity cooperativity factor, β is the efficacy cooperativity factor, τ_A and τ_B represent the agonist and modulator efficacy, and n is a scaling factor.⁴⁵  Indeed, the simulations shown in Figure 1.7 were generated using this equation. Recommendations for its proper application have been discussed.⁴⁶ 

A protocol known as “triple addition” has become broadly used in allosteric drug research.⁴⁷  This methodology enables the fast and efficient screening of compound libraries in projects seeking to identify novel modulators of receptor function by exploring in a single assay the effects of compound alone (which identifies new agonists, orthosteric or allosteric), followed by orthosteric agonist at two concentrations. A 20% maximal effective concentration (EC₂₀) identifies PAMs, and this is followed by an 80% maximal effective concentration (EC₈₀) to identify NAMs or antagonists. This screen is conducted at a single test compound concentration (usually 10 µM), and delivers different potentiation patterns, so that compounds can be qualitatively classified based on their functional response (Figure 1.8).

To qualify the potency of compounds identified in this first step, these are then studied using a concentration–response curve (CRC). PAMs are typically screened at a fixed agonist EC₂₀ concentration and NAMs at a fixed agonist EC₈₀ concentration. For example, for the case of PAMs, these data-points correspond to those inside the black boxes in Figure 1.9A–C, and are usually plotted as in Figure 1.9D, to provide half maximal effective concentration (EC₅₀) and E_max values. These two values are then used by medicinal chemists during the lead optimization stage of allosteric drug discovery projects. All three compounds have low nanomolar EC₅₀ values, but their E_max values, and therefore their functional effects at the target, are different.

Analyzing PAM EC₅₀ data in this paradigm may lead to observations that could be interpreted as non-tractable SAR. While the approach is fast and efficient, using these data mechanically may lead to comparisons of compounds of fundamentally different nature, as the simulation in Figure 1.10 shows. In this simulation, three compounds were chosen to show similar potentiation of agonist at the EC₂₀ concentration, but different response at the EC₈₀. Therefore, if the “triple addition” approach is chosen, it is recommended to confirm that the compounds tested are at least qualitatively comparable in their functional profile (e.g., they are all PAMs maintaining maximal efficacy at 100% E_max, they are all full NAMs, etc.).

In contrast to this methodology, albeit more laborious, full functional CRCs can be determined where the range of relevant concentrations are tested for both, orthosteric and allosteric ligands.⁴⁸  This protocol allows the experimental determination of a number of parameters relevant to orthosteric and allosteric ligands and the derivation of SARs for their interaction in the system under study. The exact translational nature of these in vitro-derived parameters continues to be investigated.

In summary, the characterization of a ligand based only on the effects (inhibition or potentiation of functional response) of a given receptor in an in vitro system must be conducted judiciously, avoiding oversimplification that might lead to translational disconnects with other preclinical assays or even in the clinical setting.

The extent to which effects of allosteric ligands are interrogated leads to an important, and often overlooked consequence: compounds are nowadays tested more in-depth than in the past. In so doing, compounds that were considered as having essentially the same pharmacology based on a reduced number of screens may now be differentiated by assays based on a new technology. This suggests that concepts such as “class effects” may have been misleading. This newly found differentiation also provides an opportunity to re-purpose older compounds.

Given the diverse and complex effects that an allosteric modulator may produce on the functional response of a biological target, understanding the exact nature of these effects is key in drug discovery. Ideally, an appropriate preclinical characterization of allosteric drug candidates (or chemical probes) using a number of in vitro test systems and in vivo models will enable the correct translation to human pharmacology. This is generally far from trivial, and efforts at an early stage will go a long way to avoid costly mistakes and minimize the risk of preventable clinical attrition.

Some of these precautions are common to all ligands, whether they act allosterically or not. For example, when testing compounds with receptors cloned on a certain cell line, even when receptor homology across species is high, using the correct receptor sequence for the appropriate species (e.g., rat for in vivo studies in rat, human for clinical drug candidates) is consistent with a robust translational approach.

As mentioned above, understanding the effects of receptor expression levels on functional response is particularly important when testing allosteric ligands using receptors overexpressed in a cell line. This is critical when studying compounds with allosteric agonist effects. For example, a tendency to an increase in ligand-independent functional effects has been correlated with an increased level of heterologous GPCR expression for the dopamine D_1B receptor subtype.⁴⁹ 

The GLP-1 receptor is highly conserved in mammals. An interesting report described cloning of the homologous GLP-1 receptors from mouse, rabbit, pig, cynomolgous monkey, and chimpanzee. Receptors were expressed by stable transfection of BHK cells at both high expression levels and lower, more physiologically relevant, expression levels. In the system with high expression levels of cloned GLP-1 receptors, an increased potency of GLP-1 peptide and other high-affinity ligands was observed, while K_d values stayed the same. When studying a low-affinity compound thought to be an allosteric agonist (known as “compound 2”, see Chapter 12), expression levels of the human GLP-1 receptor were found to be important for maximal efficacy and potency. The two natural peptidic metabolites of GLP-1, GLP-1(9–37) and GLP-1(9–36) amide, behaved as agonists in the cell line with high expression of human GLP-1 receptor, but they were low-potency antagonists when the receptor was expressed at lower levels. Thus, it was concluded that for the proper functional characterization of endogenous or synthetic ligands of the GLP-1 receptor, the level of receptor expression is a key parameter.³⁴ 

The early years of research into the discovery of GPCR PAMs saw a number of statements in the primary literature such as:

or

or, referring to mGluR4 PAMs,

Very quickly, allosteric modulators became notorious as research accounts were surfacing reporting “flat SAR” with a higher propensity than expected based on classical medicinal chemistry knowledge. Why is this so? Or is it?

To begin answering this question, we can consider an interesting example from the area of mGluR4 PAMs. First, 4-(1-phenyl-1H-pyrazol-4-yl)quinoline (22, Figure 1.11) was reported as a relatively potent, selective, and brain-penetrant mGluR4 PAM. This compound was a confirmed high-throughput screening (HTS) hit, quite attractive based on its low molecular weight (271.3 amu) and appropriate physicochemical properties as a CNS lead compound (e.g., Log D_7.4 = 3.9 and PSA = 31 Ų). However, after making several close analogs across the molecule, most were inactive and only a couple of them had activity at the target.⁵³  Clearly, this was an HTS hit with poorly tractable SAR. Second, an early lead compound, (±)-VU0155041 (23, peripherally restricted distribution) and the closely related amide (+)-Lu AF21934 (11, brain-penetrant) showed some SAR trends as mGluR4 PAMs. For example, replacing Cl with F was tolerated, as well as moving around the halogens in the 3- and 4- positions of the aniline ring. Still, even with this somewhat narrow SAR, these compounds became useful probe molecules to help develop the target’s preclinical validation package. Third, the N-heteroaryl-4-(1H-pyrazol-4-yl)thiazol-2-amine chemotype showed nicely tractable and additive SAR, and also led to an excellent tool compound that enabled interrogation of the biology in the field (ADX88178, 24),^54,55  as well as publications,⁵⁶  and a number of patent applications all based on the same scaffold,^57–59  indicating the highly tractable nature of this chemical lead. Thus, all three hits derived from HTSs presented different attributes in terms of SAR tractability as mGluR4 PAMs. It is inferred that SAR tractability (or the lack thereof, a.k.a. flat SAR) is an attribute of a chemotype interrogating a biological system, and not necessarily a reflection of the mechanism of action of the compounds (e.g., PAMs or NAMs).

Could SAR tractability be predicted? It is important to note that following an HTS campaign, confirmed validated hits are often times rank-ordered based on their physicochemical attributes. However, to our best knowledge, there is no established link between tractability of SAR and physicochemical descriptors or other easily calculated molecular parameters that can be estimated prior to making a few analogs. Thus, a validated way to establish whether a series has tractable SAR is to make a number of compounds purposely designed to explore SAR and/or surface plasmon resonance (SPR) trends, rather than to enhance potency or stability.

Also, tractable SAR should not be considered as an entitlement of all confirmed hits derived from an HTS campaign. As written by Robert Rydzewski,

A number of bona fide reasons may account for flat SAR. For example, a molecular fragment may be partially solvent-exposed and therefore, a significant portion is unable to interact with amino acids on the receptor and contribute to constructive binding interactions.⁶¹  More importantly for the case of allosteric modulators, when establishing SAR trends for PAMs, some researchers compare EC₅₀ and E_max values for the analogs synthesized. As explained in Section 1.5, these values are determined at a single agonist concentration. EC₅₀ values so obtained reflect a combination of compound contributions to affinity and efficacy. This may be the reason why SAR trends are hard to interpret (leading to the conclusion of flat SAR). Instead, when the same compounds are analyzed using the operational model methodology, trends are found for α, β, τ_B, and pK_B, in that specific system. pK_A and τ_A may be used as a control, since they are independent of the PAM used.^62–64 

In summary, finding novel, good quality chemical lead series with tractable SAR or SPR may not be an exclusive attribute of allosteric ligands, but rather one of the many challenges faced in drug discovery.

During SAR studies of allosteric modulators, it is not uncommon to encounter chemotypes where relatively minor structural differences have resulted in dramatically altered, even opposite functional effects—as in going from agonist to antagonist. The structural differences involved may be as small as replacing an –H for –CH₃, –OH, or –F, and may derive from medicinal chemistry design or be the consequence of in vivo metabolism.⁶⁵  Thus, when establishing pharmacokinetic–pharmacodynamic relationships during in vivo efficacy studies, ignoring the presence of circulating metabolites may lead to the wrong conclusions. For example, a metabolite with opposite functionality to the parent compound may lead to a diminished observed effect.

It is important to recognize that these functional switches are not an exclusive attribute of allosteric modulators. Indeed, interconversion of GPCR agonists and antagonists in the same chemical scaffold has been shown to be a general phenomenon, and considered to be a viable tactical approach to generate new leads.⁶⁶  Allosteric substrate-enabled switching was recently reported in a voltage-sensing lipid phosphatase.⁶⁷  Chemical switches have been exploited to enable the transformation of an mGluR2 silent allosteric modulator (SAM) into dual metabotropic glutamate receptor 2/3 negative/positive allosteric modulators.⁶⁸ 

The 20th century has been called the “century of physics”, as technologies were developed to take advantage of energy; namely combustion, electricity and nuclear power. Likewise, the following hundred-year period has been envisioned as the “century of biology”, as it was anticipated that decoding the complete genetic blueprint of the human species would define scientific advance, and play a major impact on our understanding and treatment of disease.⁶⁹ 

We began this chapter discussing the crisis of productivity of the pharmaceutical industry and the potential role that allosterism could play in helping to establish a more successful drug discovery paradigm. Towards this end, there seems to be a consensus about the importance of understanding in enhanced detail the biology of life, and the pathophysiology of disease. The key question is: what is the best strategy to move forward?

We believe that the ideal path forward to understand human disease is to engage and empower all branches of the life science ecosystem in the research. Some organizations are following this strategy. Others prefer to focus on understanding new aspects of biology and leave the discovery of new drugs to third parties, which are not fully integrated into the major research path.⁷⁰  Time will tell how this story progresses, but just one-and-a-half decades into the century, the interface between biology and chemistry, where allosterism lies, is developing at a robust pace and as a result, exciting new knowledge is emerging that we think will be vital to arrive at some of the disease therapies of the future. Allosterism is a broad attribute of molecular components of live systems, and learning to design allosteric drugs mandates that we understand the fundamental nature of these interactions. We have a lot of work ahead of us. So, shall we start?

The authors are grateful to team members and management at Lundbeck Research for supporting the scientific research conducted on allosteric modulation CNS drug discovery projects. We also thank many colleagues in academic and contract research organizations for so generously sharing their ideas and making this area of research lots of fun. In particular, the contributions of Dr A. Pilc and Dr J. Wieronska (Polish Academy of Sciences), Dr F. Acher (Université Paris Descartes, France), Dr P. Gubellini (Aix-Marseille Université, France), Dr S. Duty, Dr S. Salvage and colleagues (King’s College London, UK), Dr T. Johnstone and colleagues (Atuka Ltd, Canada), Dr J. Sprouce and colleagues (Neuroservice, France), and Dr T. Wolinsky (Porsolt & Associates, France) are warmly acknowledged.