If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Enzyme activation remains a largely under-represented and poorly exploited area of drug discovery despite some key literature examples of the successful application of enzyme activators by various mechanisms and their importance in a wide range of therapeutic interventions. Here we describe the background nomenclature, present the current position of this field of drug discovery and discuss the challenges of hit identification for enzyme activation, as well as our perspectives on the approaches needed to overcome these challenges in early drug discovery.
Enzymes have evolved over millions of years to be optimal for specific biological functions therefore identifying small molecule activators to further improve them can be challenging. An opportunity arises when enzymes lose function through mutation whereby increasing activation of mutant enzymes can return enzymatic function to wildtype levels. Enzyme inhibitors have long been a focus for modern drug discovery with the origins perhaps dating back as far as the 1850s with the use of acetylsalicylic acid for the treatment of pain and fever [
]. However, the concept of using enzyme activators as drugs has been slower to gain traction and focus on this mechanism as a potentially attractive therapeutic approach has sometimes only arisen during the quest for inhibitors [
]. Enzyme activation, therefore, represents an under exploited opportunity to identify novel chemical equity that may provide an ability to modulate enzyme function. Indeed, the identification of enzyme activators will enhance our understanding of the effects of small molecule ligands on target pathways and will supplement our ability to utilise both activators and inhibitors [
Enzyme activators have been categorised previously with the main distinction often being between non-essential (sometimes termed allosteric regulation) and essential (or obligatory) activation. The activating ligand is considered non-essential if it increases activity beyond an existing basal level, indicating that the reaction can occur in the absence of the activator, as well as in its presence. Non-essential activation may be treated similarly to general nonlinear inhibition with the changes in rate being in the opposite direction. In essential activation there is an absolute requirement for the activator to be present for observed enzyme activity [
], indicating that the reaction will not take place in the absence of the activator. Below we provide a description of the different mechanisms of activation.
Mechanisms for enzyme activation
Essential activation
In essential activation a small molecule or cation is absolutely required in addition to the substrate for catalysis. The enzyme will be in an inactive state and will not respond to the substrate unless it has been activated by the small molecule or ligand. The binding of the activator will likely cause a conformational change resulting in reorientation of the active site residues switching the enzyme from the inactive to the active form. This normally involves a large entropic penalty during rearrangement of the site [
]. There are two different classifications for essential activation, which, in a similar manner to linear inhibition, are characterised by their α (the reciprocal allosteric coupling value) [
]. Essential activators may bind only to free enzyme (E, where αKx = ∞, where Kx is the equilibrium dissociation constant for activator binding to free enzyme) or have mixed behaviour, binding before the substrate to E or after substrate binding to the ES (enzyme-substrate) complex. There are three subsets of mixed behaviour: preferential binding to free enzyme over ES complex (1< α< ∞); preferential binding to the ES complex (0 < α< 1) and equal affinity for binding to either enzyme form (α = 1). Discovery of small molecule essential activators is not usually an approach taken in drug discovery, as restoration or improvement in the rate of enzyme activity is often required rather than the ab initio generation of substrate turn over.
Essential cationic activation
A large majority of essential activators are cations. In essential cationic activation a monovalent or divalent metal is necessary for catalysis and the enzyme cannot even turnover substrate in the presence of excess substrate [
]. Type Ia typically involves using the metal ion to anchor the substrate into the active site this can be considered analogous to a ‘competitive’ activation mechanism. The metal binds to free enzyme coordinates and rearranges the active site for catalysis and enhances the substrate binding. In this type of activation there is no dependence on kcat (turnover number) as a function of metal concentration and a hyperbolic dependence on kcat/Km (where Km is the Michaelis constant) this is due to the change in Km9[9]. Enzymes like diol and glycerol dehydratases can be activated by monovalent cations through a type Ia mechanism [
The mechanism of type Ib is allosteric, where the metal binding site is separate to the substrate recognition site. In this case the metal ion does not discriminate between free enzyme and ES complex and can bind to either form. This is similar to a ‘non-competitive’ activation mechanism. The dependency on kcat and kcat/Km (as a function of metal concentration) is hyperbolic so type Ib and type Ia cationic activation can be distinguished from each other easily [
]. Pyruvate kinase is activated by potassium ions (K+) and kcat is zero when K+ is not present. K+ can bind to different forms of pyruvate kinase; active (closed) or inactive (open). This behaviour indicates that the mechanism is typical of type Ib activation [
Non-essential activation is characterised by the enzyme demonstrating an ability to turn over the substrate in the absence of the activator, albeit at a lower rate than in its presence. As with simple inhibitor mechanisms the activator can bind either before substrate, to the free enzyme or after the substrate, to the ES complex (Fig. 1). Clearly, binding of the activator must occur without detriment to the ability of the enzyme to bind and process the substrate, with either or both the binding and/or the chemical step being enhanced compared to the absence of activator. The ability for the modifier to bind preferentially to one enzyme form or the other gives rise to different classes of activation, which are analogous to the mixed, uncompetitive and non-competitive classes of inhibition and are determined by their α and β (the factor determining the overall rate enhancement of the catalytic step induced by the activator) values [
], Fig. 2. The general modifier scheme[7] (Fig. 1) covers all cases of non-essential activation as well as linear and hyperbolic inhibition. The associated general modifier equation (Equation 1) can be used to calculate α and β values and therefore define mechanism of action [
Fig. 1General modifier scheme. This scheme shows the enzyme species involved in catalysis following a general modifier scheme for essential and non-essential activation where X is an activating compound. Kx is the equilibrium dissociation constant for activator binding to free enzyme, Ks is the equilibrium dissociation constant for substrate binding to free enzyme, kcat is the rate for the chemical step generating product, α is the reciprocal allosteric coupling value which determines the relative affinity for X binding to free enzyme (E) vs Enzyme-Substrate (ES) complex and β is the factor determining the overall rate enhancement induced by the activator. The rate of the enzyme catalysed reaction in the presence of activator is given by the equation, where [X] and [S] are the free activator and substrate concentrations respectively and [E]t is the total enzyme concentration.
Fig. 2Characteristics of Enzyme Inhibition and Activation. (a) different cases of linear inhibition and broad classing of mechanism of inhibition (MoI) determined by their α value. (b) different cases of non-essential activation and broad classing of mechanism of activation (MoAct) analogous to those in linear inhibition. Desirable activators in drug discovery are those with low α (tends towards ‘uncompetitive’, α<1) and high β values (β ≥1).
'Competitive’ activation can be considered analogous to competitive inhibition, where the activator would compete with the substrate for binding to the free enzyme. However, it clearly cannot occur when α = ∞, as this would lead to a dead-end EA complex (Fig. 1), which would be catalytically inert.
‘Mixed’ activation is similar to mixed inhibition, where the binding of activator may occur with varying affinity before or after substrate has bound. Mixed activation, where the value of α (the reciprocal allosteric coupling value) is very large, will tend to disfavour the binding of substrate (analogous to mixed inhibition, which tends towards competitive in nature) and so to demonstrate significant activation the value of β would need to also be large. Hence, modifiers conforming to a ‘mixed’ mechanism may achieve activation by various combinations of α and β. For example, compounds may increase the binding affinity of the substrate (α < 1, tends to uncompetitive in nature) or may even decrease the substrate binding affinity (α >1) but will always enhance the rate of the chemical step (β >1), see Fig. 2. A compound that induces a conformational change or lowers the activation energy barrier for substrate binding (α < 1) will alter the apparent binding affinity of the substrate and may also change the rate of the chemical step. In this case of α < 1, the substrate binds more tightly to the activator-bound form of the enzyme leading to higher levels of activation being observed at low substrate concentrations [
]. An example is aspartate transcarbamylase (ATCase) which has endogenous positive and negative modulators. Positive modulators bind to the active state, which has higher affinity for its substrate and small molecules can be designed to mimic these modulating ligands [
‘Non-competitive’ activation occurs when the activator may bind before and after the substrate binds with equal affinity to the free enzyme and ES forms. Activators which bind in this way, where α = 1, do not change the apparent Michaelis constant (Km) but will increase the maximal rate (β > 1). Modifiers of this type are expected to enhance the catalytic rate, potentially by realigning the catalytic machinery, increasing the rate of the chemical step without effecting the substrate binding [
‘Uncompetitive’ activation arises when the activator binds only after the substrate has already bound (0 < α < 1), or in effect only to the ES complex. This activator mechanism stimulates a concurrent decrease in apparent Km and increase in kcat (β > 1). As an example activators stabilising the enzyme-substrate complex are exemplified for glucokinase (GK), below [
]. Examples of compounds that mimic the behaviour of a protein binding partner, which subsequently leads to activation of the enzyme, have also been demonstrated for PDK1 [
A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase Czeta (PKCzeta ) and PKC-related kinase 2 by PDK1.
Outside of the terminology generally applied to inhibitors there are also combinations of α and β that can lead to either activation or inhibition depending upon the concentrations of activating molecule and substrate. These two remaining types of non-essential activation may be less common. Not only do these types of mechanism depend upon the individual values of α and β but the behaviour also is dependent upon the relative values of α and β. Where 1 < α < ∞, 1 < β <∞ and α > β, then the compound is inhibitory at low substrate concentrations, but shows activation at high substrate concentrations. The converse occurs where 0 < α < 1, 0 < β <1 and α < β [
Glyphosate sensitivity of 5-enol-pyruvylshikimate-3-phosphate synthase from Bacillus subtilis depends upon state of activation induced by monovalent cations.
]. The different mechanisms of inhibition and activation determined by their α and β values and described above are summarised in Fig. 2.
Non-essential cationic activation
Non-essential activation can be induced with cations. Non-essential cationic activation, also known as type II cationic activation, is where the cation is capable of promoting a higher level of enzymatic activation but the enzyme can turnover substrate in the absence of the cation [
]. Type II activation typically involves allosteric coordination of a metal ion leading to an improvement in activity likely through conformational changes enhancing enzyme activity from a basal level. The increase in activity typically results in a hyperbolic dependence on kcat and kcat/Km. These dependencies are equivalent to those observed in type Ib cationic activation, so they cannot be distinguished using these parameters. However, in type II cationic activation the enzyme will have activity in the absence of the specified cation. In branched chain α-ketoacid dehydrogenase (BCKD), K+ binds to two separate allosteric sites which stabilise the enzyme and allow maximal activity [
Crystal structure of human branched-chain alpha-ketoacid dehydrogenase and the molecular basis of multienzyme complex deficiency in maple syrup urine disease.
Examples of small molecule non-essential enzyme activation in drug discovery
Glucokinase
Glucokinase (GK) is an essential enzyme in blood glucose homeostasis and is responsible for phosphorylating glucose to produce glucose-6-phosphate. Loss of function mutations in GK cause a decrease in Vmax (reaction rate when the enzyme is fully saturated by substrate) and affinity for its substrates (glucose and ATP (adenosine triphosphate)) which results in reduced glucose sensitivity in β cells and decreased glycogen accumulation [
Mutants of glucokinase cause hypoglycaemia- and hyperglycaemia syndromes and their analysis illuminates fundamental quantitative concepts of glucose homeostasis.
]. Modulating GK is, therefore, an attractive therapeutic target for regulating type II diabetes. There has been particular interest in enhancing GK activity using small molecule activators that increase apparent affinity (Km’) for glucose and substrate turnover (Vmax) [
]. A library of 120,000 compounds was screened using a biochemical assay which identified a compound capable of increasing GK activity. Binding of the compound has been shown to concomitantly enhance the basal activity of GK with a 4-fold decrease in Km’ and 1.5-fold increase in Vmax. No effect on the Km’ for ATP (the second substrate) was observed [
Piragliatin (RO4389620), a novel glucokinase activator, lowers plasma glucose both in the postabsorptive state and after a glucose challenge in patients with type 2 diabetes mellitus: a mechanistic study.
]. Precedented activators for GK typically bind to an allosteric site on the enzyme and stabilise the active conformation. Upon binding glucose, GK undergoes a slow conformational change from ‘super-open’ to ‘open’ which reveals an allosteric site on the enzyme only present once the enzyme-substrate complex has formed [
Phosphoinositide-dependent protein kinase 1 (PDK1) is responsible for activating and/or stabilising the kinases in the AGC kinase subfamily (e.g.P90 ribosomal s6 kinase (RSK), Protein kinase C (PKC), P70 ribosomal s6 kinase (S6K), Protein kinase B (PKB or AKT), serum/glucocorticoid regulated kinase (SGK) by phosphorylation of the conserved threonine residue present in the activation loop (T308 in PKB). PDK1 contains a regulatory allosteric site, termed the PDK1 interacting fragment (PIF) pocket, which interacts with the hydrophobic motif of selected downstream phosphorylated substrate enzymes such as RSK, PKCζ, PRK2 (protein kinase C-related kinase), S6K and SGK allowing docking of the substrate kinases prior to phosphorylation [
A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase Czeta (PKCzeta ) and PKC-related kinase 2 by PDK1.
Interestingly, PKB is phosphorylated by PDK1 through a different mechanism compared to the other kinase substrates where PKB does not need to dock in the PIF pocket for its phosphorylation and activation [
]. However, occupation of the PIF (PDK1 interacting fragment) pocket by PIFtide has shown a four-fold increase in phosphorylation of PKB peptide (T308tide) and a decrease in Km of T308tide for PDK1 from >10 mM to 0.14 µM [
]. The efficiency of phosphorylation of PKB is also heavily dependent on phosphoinositide, which binds to the PH domains on both enzymes and co-localises them to the membrane [
In assays targeting the PIF pocket the PIF peptide (PIFtide) is regularly used instead of the full-length protein, as the PIFtide shares similarity with the hydrophobic motif conserved in the substrate kinases. Upon docking of the peptide substrate a conformational change in PDK1 occurs promoting stabilisation of the active conformation, which, in conjunction with phosphorylation on the activation loop, allows binding of the second substrate, ATP. Prior to binding of the first substrate the PIF and ATP pockets are both disrupted [
Hit finding campaigns have been undertaken to identify small molecules that induce the same conformational change, leading to stabilisation of the active conformation and enhancement of PKB phosphorylation. An in-silico screen of 60,000 compounds identified two small molecule activators [
] which were demonstrated by surface plasmon resonance (SPR) to compete with PIFtide binding. An in vitro kinase assay confirmed that this compound increases activity by 5-fold and decreases Km’ from >10 mM to 35 µM [
A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase Czeta (PKCzeta ) and PKC-related kinase 2 by PDK1.
]. This activator compound and others found more recently mimic substrate binding at the allosteric PIF pocket, binding preferentially to free enzyme and working cooperatively to enhance a stronger interaction between the T308 peptide and PDK1 leading to an increase in overall activation [
However, whilst phosphorylation of the T308 peptide was enhanced by the compound, phosphorylation of substrate enzymes that bind to the PIF pocket were inhibited as the compound blocked these from docking in the PIF pocket [
]. There is an obvious advantage in identifying small molecule inhibitors of PDK1 as this will inhibit the downstream cascade, including PKB. However, finding small activator molecules with the ability to modulate conformational states has provided a good rationale for alternative ways to target phosphorylation-dependent transitions. This could lead to the development of more potent activators [
2-(3-Oxo-1,3-diphenylpropyl)malonic acids as potent allosteric ligands of the PIF pocket of phosphoinositide-dependent kinase-1: development and prodrug concept.
3,5-diphenylpent-2-enoic acids as allosteric activators of the protein kinase PDK1: structure−activity relationships and thermodynamic characterization of binding as paradigms for PIF-binding pocket-targeting compounds†PDB code of 2Z with PDK1: 3HRF.
], using different combinations of technologies such as deuterium exchange experiments, fluorescence based binding assays, disulphide trapping (covalent), virtual screening and fragment screening [
A comment on nomenclature: K type and V type classification
A simple way to classify both inhibitor and activator modifiers is by their ability to affect the Km or Vmax of the enzymatic assay. Compounds decreasing Km are classified as K-type activators, whilst compounds increasing Vmax of the enzymatic reaction in a concentration dependent manner are classified as V-type activators. If the compound increases Km and decreases Vmax of the reaction, then this is classified as K-type or V-type inhibition, Table 1. However, caution should be applied as there are cases where a modifier compound may demonstrate K-type activation and V-type inhibition or K-type inhibition and V-type activation and so does not distinguish whether the modifier is an activator, inhibitor or has dual behaviour, depending upon the conditions of substrate concentration, Fig. 2. Despite this problem the classification has been widely used throughout the literature, but is usually accompanied by supporting information to help discern modifier behaviour [
Kinases, such as adenosine monophosphate-activated protein kinase (AMPK), frequently contain regulatory domains or autoinhibition domains (AID) that control access to the active site on the catalytic domain. In inactive AMPK protein-protein interactions between the regulatory domain (βγ complex) and the catalytic domain (α subunit) hold the kinase in an open conformation. AMP binds to the regulatory domain and disrupts the interactions between the α subunit and βγ complex allowing the α subunit to transition into its active state. This can be treated as de-inhibition. Modulation of AMPK is seen as an attractive therapeutic target and could treat metabolic syndromes and type II diabetes [
]. This compound activates AMPK with an improved EC50 (concentration of ligand giving 50% effect) value of 38 µM, compared to AMP, which has an EC50 value of 56 µM, but no improvement in Vmax was observed. Whilst there is no overall increase in Vmax, the presence of the activating compound triggers increased levels of enzyme activation at lower non-saturating concentrations of substrate. The A-592107 was optimised and A-769662 a more potent activator was discovered (EC50 value of 0.8 µM) [
Challenges for identifying non-essential enzyme activators
Drug discovery approaches may be required to identify non-essential activators that restore or enhance the basal level of enzyme activity to modulate a disease state. The identification of putative activators may be achieved by using many of the existing traditional methods applied for inhibitors, but there are additional considerations that may need to be addressed.
Firstly, enzymes are biological catalysts that are present in all living organisms. Their role is to increase the rate of chemical reactions in biological systems. They have evolved over time specifically to efficiently increase the rate of the specific chemical reactions. Thus, the likelihood of identifying small molecules that can increase this activity further may possibly be expected to be lower than that for identifying ligands that decrease the activity – disrupting the catalytic machinery may be easier than improving upon nature's solution for rate enhancement.
Lack of tool compounds is often an issue when developing enzyme assays. Having a reagent that can replicate the desired outcome expected for a suitable hit is useful in defining the response of the assay and the degree of activity of identified hits. This dearth of tool compounds is likely to be greater for activators than for inhibitors, partly due to the reason above, but also because substitutes for eliminating activity are more readily conceived than for increasing activity. Measures of control blank rates may be obtained in various ways: omitting enzyme, substrate or another essential component, using denatured (boiled) enzyme, adding stop solution (e.g. low or high pH) before the reaction is started, using chelators to remove essential metal ions or specific or non-specific inhibitors. These approaches may allow estimation of the expected decrease in rate when the enzyme is fully inhibited. Unfortunately, determining the degree of enzyme activation expected is not straightforward, as this is, in its simplest form determined by both the degree of saturation as well as the magnitude of the effect on kcat, which contrasts with the effect of simple linear inhibitors. In other words, the bottom of the plot of rate versus ligand concentration is well defined for an inhibitor, whereas the top of the curve for an activator is effectively unknown and potentially different for each activating ligand. The effect of activation may be mimicked by increasing the functional enzyme concentration but the magnitude of the effect may bear no relationship to that expected or observed for different activators.
By definition, an enzyme activator will increase the rate of the catalysed reaction compared to the control rate in the absence of activating ligand. This will change the progress curve for the reaction and may reduce or remove the linear portion indicative of steady state kinetics [
]. Usually, steady state conditions are identified by monitoring the reaction over time at a fixed enzyme concentration with inhibitors not associated with changing the concentration of enzyme-bound intermediates and affecting the steady state period. Conversely, activators may reduce the length of the approximate steady state period as they increase the rate of turnover and with substrate concentration changing more quickly. Additional effort may be required during assay development to ensure that the expected degree of activation allows for remaining in the steady state and initial rate measurements to be reliable for detecting activators. Practically, this may mean that assays must run for shorter times, impacting on time-critical processes such as the application of automated liquid handling and reading during a high-throughput screen. Slow-binding kinetics are common during drug discovery [
] where low concentrations of potent compounds are often employed. Although slow binding does not usually impact upon primary screens as the high concentrations, ability to pre-incubate and relatively long assay times usually lead to full equilibration, or at least enough compound binding to lead to a detectable decrease in rate for an enzyme inhibitor. For slow-binding activators, the incubation period may need to be extended to allow full equilibration, as the relative magnitude in the change in rate is expected to be much lower – for example it may not be unexpected for an inhibitor to show 90 % inhibition (10-fold reduction in rate), whereas similar magnitude increases in enzyme activity are expected to be highly unlikely. Often compounds providing around 2-fold activation are considered respectable hits [
The mechanism of activation may also be a key factor in the ability to detect activators. Like inhibitors, activators may interact with the enzyme before or after the substrate has bound, although for activators, the ternary complex between the enzyme, activator and substrate must form for substrate turnover to occur. Thus, this limits the utility of the term competitive activator [
]. In general, the a and b terms from the scheme above (Fig. 1) will often be below 1 and above 1 respectively. However, mixed-type schemes will be possible where the values of both a and b could both be above 1 or both be below 1. For situations where both a and b are less than 1 it is possible for a compound to activate at low substrate concentrations but to demonstrate inhibition at high concentrations. Thus, the choice of substrate concentration will affect the ability to detect compounds with the desired mode of action [
] in a manner that is, perhaps, even more stark than for inhibitors.
Identifying small molecule enzyme activators generally requires overcoming more challenges than identifying enzyme inhibitors. Although most biochemical assay technologies that are employed for high throughput screens are amenable for identifying activators, assay quality parameters such as Signal Window, and Z’ (Z-prime, a statistical measure of assay performance) [
] may be estimated with less certainty and may appear lower because of high background and more variable top signal. The assay window may be more difficult to estimate for activator assays compared to inhibitor assays due to the uncertainty over the degree of activation that may be observed and the maximal signal achieved. In cases where the standard deviation increases with raw signal increase, variability around the minimum signal controls would be larger for activation assays since the minimum signal in this instance represents basal activity rather than fully inhibited signal. In inhibitor assays the two controls correspond to 0 % and 100 % inhibition, where 0 % inhibition is the uninhibited rate and 100 % inhibition represents fully inhibited enzyme, often accessed by using a tool compound at high concentration. These controls provide a defined signal window with easily quantifiable and often lower variability. In contrast, for activator assays the minimum control corresponds to non-activated rate with basal activity. The high signal control requires a tool compound, if available, at a concentration where saturation is reached, but it may be possible to reach higher signal levels if more potent activators are identified during the screen. Assuming the variance increases proportionally with signal increase (i.e. constant coefficient of variation), these higher signal controls will have a higher standard deviation compared to the lower signal controls experienced in inhibition assays. Hence, there may be a tendency for the assay statistics to be poorer for activation compared to inhibition assays and for the probability for identifying compounds, outside of the screening noise to be lower for activation assays.
Where tool compounds are not available, it may be necessary to use other controls to define the signal window. This is presumably simpler for inhibition assays, where the fully inhibited rate may be mimicked by various approaches as described above. For activation assays, the only option to mimic the activated enzyme rate, without affecting the likelihood of identifying activators with different mechanisms in the absence of a tool activating compound, is to increase the functional enzyme concentration. Assessing the extent to which this may be a relevant approach is also difficult – what degree of activation may be expected from an active compound – so that an appropriate increase in enzyme concentration may be used to provide a reasonable signal window estimate? Even if a tool compound can be used, there remains the possibility for screened compounds to have higher degrees of activation and therefore to display >100 % activation with respect to the stimulator control. In either case the assay window is variable and the maximal activation potential unknown making assay characterisation and subsequent data analysis more difficult than in more frequently used inhibition assays.
Concentration response curve analysis also presents further challenges for activator compounds. The most-commonly used 4-parameter logistic fit model requires well-defined curves to reliably estimate the top, bottom, midpoint and slope at the mid-point. In contrast to inhibition, where the bottom of the inhibitor curve might be expected to drop to zero, the top of an activator curve is often poorly defined by the data and more difficult to estimate. Consequently, this results in more uncertain estimates for compound potency for activators, requiring alternative methods to assess the effectiveness of active compounds in mechanistic studies (see below). Interestingly, potency may not be the only nor the most important factor for identifying activators, with the degree of activation induced by the compound being an important consideration. Understanding the target biology and the required level of activation to observe a phenotypic effect will be important when considering which compounds to progress. For example, it may be that during any subsequent chemical optimisation, potency, achieved from the molecular recognition component, may be easier to improve compared to being able to further improve the rate of the chemical step involved in catalysis. Hence, it may be advantageous to progress compounds with a high degree of activation even if they may not be the most potent start points.
Finally, identifying ligands that may subsequently be observed as inhibitory is often achieved using affinity binding methods as a primary screen. These involve detecting binding of ligands to proteins (directly or indirectly), rather than measuring biological activity. Often, these binding assays prove to be beneficial for unprecedented targets where no tool compounds are available. It may be reasonable to expect that ligands identified in this manner may be more likely to have an inhibitory effect on the target activity, from the rationale that filling a binding pocket (such as the active site) is potentially more likely to reduce activity than to stimulate activity. This consideration may be extended from orthosteric sites to allosteric disruption of substrate binding or disruption of the catalytic machinery, having a higher probability than having a stimulatory effect. Activity assays, of course, will therefore be ultimately required to discern true activator MoA behaviour. Whilst there is no reason these affinity screening methods cannot be applied to activators, the probability that they will identify potential activators may be reasonably expected to be significantly lower and so may tend to limit their application compared to inhibitor screening.
Addressing the challenges of activator discovery
Above, we discussed some of the difficulties with drug discovery for activators. To further illustrate some of these challenges and potential strategies for overcoming them, we simulated the changes in rate of product formation at different compound and substrate concentrations using the general modifier equation (Fig. 3 and Fig. 4).
Fig. 3Simulated substrate dependence curves for enzyme activators. Rate of product formation plotted against substrate concentration for four simulated activator compounds. The general modifier equation was used to simulate for compound for an enzyme with kcat = 1 µM/min, Ks = 10 µM and enzyme concentration set to a constant of 1 to yield a maximum product formation rate of 1 µM/min (Vmax = 100%). The “starting activator compound” profile (Panel A) was simulated with Kx = 30 µM, α = 0.2, β = 2. Increasing compound concentration [µM] is indicated, using the viridis colour palette, with compound concentration increasing from blue (0 µM) to yellow (100 µM).
Fig. 4Simulated concentration response curves for enzyme activators. Rate of product formation plotted against compound concentration using the same simulation parameters as Fig. 3. The logarithm of compound concentration is plotted on the x-axis. Increasing substrate concentration [µM] is indicated, using the viridis colour palette, with substrate concentration increasing from blue (0 µM) to yellow (140 µM).
Fig. 3 shows the changes in the rate of product formation plotted against substrate concentration for different concentrations of four exemplar compounds. The panels illustrate how changing the different compound-associated parameters affect the global profile of product formation. The starting point of the simulation (Panel A) is an activator compound with Kx = 30, a = 0.2 and β = 2. As seen in the “Starting Compound” panel of Fig. 3, in the absence of activator compound the maximum rate of product formation is 1 µM/min. Concentrations of compound above 3 µM result in an observable increase in product formation compared to enzyme alone. Expectedly, a compound with 10-fold lower Kx (Panel B) shows a similar profile but with visible activation at lower concentrations, here discernible above 0.3 µM. Whereas, lowering a by ten-fold (Panel D) results in achieving maximal rate at both lower concentrations of substrate as well as compound. This is demonstrated by the steeper increase in rate with substrate concentration demonstrated in the substrate dependent rectangular hyperbola. This contrasts with the change in Kx which has a less pronounced effect on the shape of the substrate dependence plot but lowers the threshold for compound-dependent activation. Additionally, and unsurprisingly, changes in β have the most profound effect on the rate of product formation as shown in Panel C of Fig. 3. This investigation of the effects of changes in the parameters associated with the binding and catalytic steps suggests that a useful way to rank the activation effects for these four simulated compounds is to compare the sum of the area under all the curves in each individual panel in Fig. 3. In this example, demonstrated in the plots of Fig. 5, examining the profiles in this manner reveals that changing β has the highest impact for overall activation, followed by lowering a and then changing Kx. This empirical approach is likely to have value when rapidly comparing activators where a combination of changes in these parameters may be present. Although this will potentially identify the most effective compounds, it will not fully substitute for a detailed multi-variate experimental approach allowing the estimation of values for a, β and Kx. However, we believe that this is certainly preferable to an approach which aims to measure EC50 which may be termed, AC50 (activator concentration giving 50 % activation) and the top of the curve, like the determination of inhibitor EC50, termed IC50 (inhibitor concentration giving 50 % inhibition), often used in the lead optimisation process.
Fig. 5Simulated Area under Curve (AUC) concentration response curves for enzyme activators. Relative Area under Curve (AUC) values have been plotted against compound concentration using the same simulation parameters as Fig. 3. The logarithm of compound concentration is plotted on the x-axis versus AUC for the changes in α, β or Kx indicated.
To illustrate the difficulties in using EC50 and top of the curve for ranking activators, Fig. 4 plots the same profiles but with compound concentration on the x-axis. This shows how the analysis of activating compounds might be hindered by incomplete curves. Both the “starting compound” (Panel A) and the compound with “double β” (Panel C) have incomplete curves, such that the top concentration of compound does not reach a plateau for the increase in rate of product formation. Often, the top concentration of compound and substrate tested are limited by compound solubility or the DMSO tolerance of the assay. In addition, full-length protein substrates may require prohibitively high quantities of protein to be made limiting the top substrate concentration further. The non-linear 4-parameter concentration-response fits for the starting compound and the compound with the doubled value for β would necessarily be less certain, with wider confidence intervals for the EC50 and an uncertain estimate for the top of the curve and hence the degree of activation delivered. Such uncertain fits could be penalized when ranking compounds on concentration-response curve quality and the generated EC50 value. To prevent useful compounds from being discarded may require a degree of manual intervention in viewing the curves, assessing the variability of EC50 values and the position of the estimation of the degree of activation [
]. In contrast, compounds with high affinity (lower values of Kx) and lower values of a would have more complete curves (see Fig. 4 panels B and D), smaller confidence intervals for the EC50 and well-characterised tops. Therefore, compounds with lower values of a and Kx may be prioritised over compounds which may, in fact, be better activators such as those with higher values of β. To maintain diversity, offering a range of options for medicinal chemistry optimisation early in the discovery process, it would be beneficial to progress compounds with a variety of profiles and the bias for high affinity compounds may “trap” chemists in local optimisation minima with little scope for Structure Activity Relationship (SAR) work in lead generation. In contrast, it may be prudent to prioritise compounds with genuinely high values for β and to focus medicinal chemistry efforts on the usual practice of improving affinity as part of the lead generation process. This yields a familiar approach of maintaining efficacy, whilst increasing potency, often by improving interactions within the binding site and perhaps minimises the difficulty of understanding how to improve catalytic rate further. For example, by attempting to realign substrate binding pockets, enhancing interactions between important catalytic residues, stabilising the transition state or improving leaving group ability. This approach may yield improved results overall as it utilises the strengths of current medicinal chemistry.
In addressing the challenges with activator drug discovery, we recommend the following approaches are considered.
Firstly, understanding the behaviour of the assay system under conditions of activation is important to address early in a project. Approaches to mimic the effects of activation may be useful to assess the ability of the assay to perform under the conditions of increased reaction rate. At this stage, where the availability of tool compounds may be lacking, characterising the behaviour of the assay with increased enzyme concentration, increased substrate concentration or under other conditions that may deliver rate enhancement (temperature, pH) can be valuable. This ensures that the system will be capable of identifying hits that show the expected level of activation.
Since activators may reduce the time-period during which initial rate conditions prevail, kinetic assays are preferred. However, this may be difficult during primary screening and so understanding the behaviour of compounds identified during the screen may require additional, detailed follow up in lower throughput experiments before further progression.
As with inhibitor primary screening, we recommend that balanced assay conditions are utilised for hit finding. This condition, where substrates are used at concentrations close to their Km values, should provide the best chance for identifying activators with different values of a.
Once an assay system is in place that delivers an ability to successfully identify activators with a desired or expected level of activation, it is valuable that attempts are made to identify tool compounds that might be useful in establishing initial screening parameters and that can be used to act as positive controls. We advocate rapid, smaller screens during the transition of assays into high-throughput screens, using a diverse range of compounds supplemented with compounds that may have been identified that have a higher probability of success, for example, compounds identified via virtual screening. At AstraZeneca, we have found that screens of this nature, comprising just under 10,000 compounds, have been successful in identifying both useful tools and compounds that are chemically attractive for further optimisation.
To reduce the effects of a reduced window of effect for activators and the potential for increased variability for the fitted parameters, we recommend that replicates are used accordingly to reduce sampling error and reduce or allow the detection of human and technical error.
To fully characterise compounds, full matrix experiments, where substrate concentration and activator concentration are varied should be used to obtain full concentration response curves, which may be analysed appropriately. We recommend generating AUC plots to visually exemplify the behaviour of compounds having different values of α and β. These analysis approaches may be employed on representative examples of hit series or on all compounds once the numbers have been reduced during the lead optimisation process. This will allow focus on compounds with the desired profile of α and β values.
In conclusion, the challenges around developing activator compounds may be overcome by maintaining a mechanistic awareness of the pitfalls of working with activators and by performing experiments which systematically assess the effects of activation. Initially, activation profiles may be quantified in a non-parametric approach by summing the area under the rate vs substrate concentration curve for each compound concentration. The summed area under the curves can be used as a first-pass filter to separate inactive compounds and rank activators. Detailed characterisation requires varying both compound and substrate concentrations in matrix experiments and the estimation of the values of α and β. During lead optimisation fitting a global model using the general modifier equation and estimating a, β and Kx for each compound would enable the progression of more diverse compounds and may potentially result in the design of superior activators. In contrast to the classical approach for inhibitors, where compounds are ranked by IC50 values, the analogous approach for activators, utilising AC50 and maximal activation, has additional disadvantages and our recommendation is that taking the time to undertake the additional mechanistic characterisation described above may lead to compounds with better activation potential (β). Having information on the magnitudes of α and β at very least provides a greater degree of flexibility in terms of the diversity of compound behaviour that can be progressed.
Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests. Geoff Holdgate is a member of the Editorial Board for SLAS Discovery
References
Miner J.
Hoffhines A.
The discovery of aspirin's antithrombotic effects.
A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase Czeta (PKCzeta ) and PKC-related kinase 2 by PDK1.
Glyphosate sensitivity of 5-enol-pyruvylshikimate-3-phosphate synthase from Bacillus subtilis depends upon state of activation induced by monovalent cations.
Crystal structure of human branched-chain alpha-ketoacid dehydrogenase and the molecular basis of multienzyme complex deficiency in maple syrup urine disease.
Mutants of glucokinase cause hypoglycaemia- and hyperglycaemia syndromes and their analysis illuminates fundamental quantitative concepts of glucose homeostasis.
Piragliatin (RO4389620), a novel glucokinase activator, lowers plasma glucose both in the postabsorptive state and after a glucose challenge in patients with type 2 diabetes mellitus: a mechanistic study.
2-(3-Oxo-1,3-diphenylpropyl)malonic acids as potent allosteric ligands of the PIF pocket of phosphoinositide-dependent kinase-1: development and prodrug concept.
3,5-diphenylpent-2-enoic acids as allosteric activators of the protein kinase PDK1: structure−activity relationships and thermodynamic characterization of binding as paradigms for PIF-binding pocket-targeting compounds†PDB code of 2Z with PDK1: 3HRF.
13693-2&id=gr1.jpg){kind=link}
13693-2&id=gr2.jpg){kind=link}
13693-2&id=gr3.jpg){kind=link}
13693-2&id=gr4.jpg){kind=link}
13693-2&id=gr5.jpg){kind=link}