Amid the complexity and expense of the small-molecule drugdiscovery process, from identification and validation of a "drugable" target tothe development of an understanding of the impact of pharmacogenomicdifferences in patient populations on drug action, lies the "hit-to-lead"process in which compounds that show activity in an assay system areiteratively improved upon through medicinal chemistry that is guided by moredetailed assays and filtering criteria.
The assays used involve a careful balance between simplicity(which inversely correlates with time and expense) and biological relevancy,and should ensure (or at least not contradict) a correlation between assayresponse and ultimate in vivo drugaction. While it is easy to build and perform robust assays to measure theaction of a soluble, unmetabolized compound on a single, well-characterized,recombinantly expressed protein target, doing so in a manner that mimicsbiological relevancy remains a challenge.
At a minimum, for the hit-to-lead process to be successful,the assay systems employed to evaluate compounds should provide a clearunderstanding of compound affinity for the intended target, as well as a clear"profile" of compound specificity, which is typically thought of as ameasurement of the affinity of the compound for its intended target relative tothe affinity of its interactions with other, often related, targets. Both ofthese properties can contribute to the ultimate success of a compound: Thosethat bind tightly (have high affinity for their intended target) can in theorybe used at lower concentrations, and those that bind selectively should bydefinition have fewer off-target effects, which can contribute to side effectsand toxicity.
While compound affinity for the target of interest can oftenbe assessed using the same assay system used to initially identify and/orcharacterize the compound, specificity is often assessed by performing assaysagainst a larger number of targets. Because of the scale and complexity of thisprofiling and a requirement for standardization, nuances may be lost during theprofiling process.
For example, in the case of kinase-directed compounds,profiling is often performed against panels containing the active forms ofkinases. While this simplifies the assays being performed (it is easier tomeasure the activity of an enzymatic reaction when the enzyme is active), itremoves the nuance of compounds that may bind preferentially to thenon-activated form of a kinase, and thereby stabilize the non-active state,which may prevent further activation.
Additionally, although more than 500 protein kinases existin the human genome, methods to express and/or assay all of these kinases donot exist, and even the broadest panels lack full kinase coverage. The lack offull target family coverage and lack of easy control over kinase activationstate is compounded by the fact that for both technical and economic reasons,full-length targets are often not used (catalytic activity may be assessedusing only the catalytic domain of the kinase), and the targets are oftenexpressed in non-mammalian systems or as domains expressed on the surface of phageparticles.
Given the fact that the majority of small-molecule drugsdirected toward kinases presently target the ATP binding site, and that aplethora of proteins use ATP as a substrate (and therefore contain ATP bindingsites), the ability of any profiling process that is limited to determiningspecificity against only kinase targets is clearly incomplete. Although elegantmethods have been described that can identify the target of a small-moleculedrug from within a cellular lysate (thereby exposing a compound to all possiblebinding partners present in a particular cell type), such methods still containshortcomings and are difficult to implement in a cost-effective manner.
Despite the recognized challenges of developing appropriateassays for moving compounds forward during the hit-to-lead process, and theshortcomings inherent in any of the available methods, this process remainscrucial to the development of selective and efficacious drugs, with theimportant caveat being that the information that can be gleaned by any onemethod used in the process should be evaluated with a clear understanding ofthe limitations associated with that method.
In addition to target affinity and specificity, a third (andless commonly appreciated) property of a compound that can correlate with both invivo efficacy as well as safety is oftenreferred to as drug-target residence time. This property is related to theaverage time that the compound remains associated with its intended targetbefore dissociation.
There are several reasons why this property can correlatewith compound success. The first is intuitive: For a drug to be active againsta target, it needs to be physically associated with that target. Whileassociation of a compound with a target is dependant on the concentration ofthe compound, the rate of dissociation is independent of drug concentration andis a property of the drug-target complex. Since any drug that is not associatedwith a target is available for metabolism, degradation or excretion, if acompound has a slower off-rate than these competing processes, it can remainefficacious for a longer period of time than if it had a shorter residencetime.
This leads to a second corollary that can be associated withresidence time, that being an increase in the "effective" selectivity of acompound. For example, a drug may bind to multiple targets, but if the"off-target" events have short residence times such that the drug may beeliminated before these off-target events are detrimental, then side effectsand toxicity may be lessened.
The importance of drug-target residence time has beenrecognized for many decades, and at least via retroactive analysis, there arenumerous examples where the in vivoproperties of a compound or set of compounds can be explained or rationalizedbased upon residence times. Despite the recognized importance of drug-targetresidence time, measurement of this property is not commonly performed whenprioritizing compounds early in the drug discovery process. While affinity canoften be measured using fairly standard methods, and first-pass specificity canbe determined by profiling compounds against an appropriate panel of relatedtargets, compound residence time is often measured using complex kineticexperiments or in systems using immobilized targets.
For example, a traditional enzymatic method for measuringthe rate at which a compound dissociates from a target is to first incubate thetarget and the drug at a concentration above the Kd value for that interaction,to rapidly dilute the sample into assay buffer such that the totalconcentration of drug is now below the Kd value, and then to measure theinitial rate of a catalytic reaction at various time points after the dilutionhas been performed. As the compound dissociates from the target the enzymaticactivity is restored, and the rate at which this restoration is seen can thenbe correlated with residence time.
Although conceptually simple, the dissociation processcannot be monitored in real-time, and multiple catalytic reactions(experiments) are required to determine residence time for a single compound.Alternative non-activity based assays can be performed in a similar formatusing radioactive probes that bind to the active site as compound dissociates,and bound radioactivity can be measured after a separation step is performed(to remove unbound radioactivity), but these methods come at the regulatory andsafety expense associated with the use of radiation.
An alternative approach to measuring compound dissociationis by surface plasmon resonance (SPR) or similar optical methods in which thetarget is immobilized (attached) to a surface, and compound binding (anddissociation) can be measured in real-time as solutions containing compound arepassed over the immobilized target. Limitations on this approach can include anappropriate level of sensitivity, as the signal is dependant in part of opticalchanges induced by a small molecule (ca. 500 Daltons in molecular weight)binding to a much larger protein (ca. 50-100 kilo Daltons), which can beextremely small. However, due to the flexibility of the technique (whichincludes no requirement for labeling either the target or the receptor with anysort of "tag" or handle) and the rapid pace at which technological improvementsare being made, it is expected that these types of measurements will push intoroutine use at earlier stages in the drug discovery process in the future.
Recently, several techniques were described at the 2010Society for Biomolecular Sciences (SBS) conference in April in Phoenix, fromboth my own group as well as scientists at GlaxoSmithKline, that combinecertain elements of existing approaches in an attempt to develop a homogenous,real-time assay system for measuring drug-target residence time. As with thetraditional radiometric approach to measuring compound off-rates, labeledcompounds that can bind to a target active site as a drug dissociates are used,but in these cases the labels used are fluorescent rather than radiometric.
In either case, binding of the fluorescent probe to thetarget active site can be monitored in real time as the drug dissociates fromthe target by a change in fluorescence signal. Although in each approach thereis a requirement to generate a fluorescent probe that will bind to the targetof interest, for many target classes (such as kinases), there arewell-described tool compounds that bind broadly across a target class, and cantherefore be used to develop a small set of probes that could be used todetermine residency time across a large number of related targets. Time willtell if these methods can be developed and validated to the extent that theymay become a routine part of the lead optimization process to help developbetter drugs faster.
Dr. Kurt Vogel is director of R&D within the discovery and ADMETsystems segment of Life Technologies in Madison, Wis. In this role, he has leadteams focused on the development and commercialization of products and servicesaimed at early-stage drug discovery, particularly those involvingfluorescence-based readouts and formatted for high-throughput screeningapplications.