It is inevitable that any article about G protein–coupled receptors (GPCRs) will begin with a statistic reflecting the importance of these proteins—which transmit essential signals from a wide range of hormones and neurotransmitters—as drug targets. In fact, the numbers are striking: 40–50% of all marketed drugs are thought to modulate GPCR activity. Equally striking, however, is the fact that so many of these drugs were formulated with a limited understanding of the deep complexity underlying GPCR biology. “There are so many challenges in identifying high-affinity, selective GPCR drugs with the desired efficacy that it's amazing we have anything at all,” says Brian Kobilka of Stanford University.

The traditional model of GPCR function is complicated enough. Ligand binding at the receptor extracellular domain induces intracellular domain rearrangements that allow a GTP-bound heterotrimeric G-protein αβγ complex, which can comprise hundreds of individual subunit combinations, to bind. The G-protein complex then hydrolyzes GTP to GDP and dissociates into α and βγ subunits, each of which then activates distinct downstream signal cascades.

This, however, pales in comparison to the byzantine complexity that scientists have subsequently uncovered regarding the remarkable flexibility of GPCR structure and signaling activity, which is, in turn, closely dependent on cellular context and interactions with activity-modulating binding partners. Even their name is misleading, as evidenced by the recent determination that some GPCRs do not even use G proteins, leading some to renounce the family moniker altogether. “It's a horrible name,” says Terry Kenakin of GlaxoSmithKline's Department of Assay Development and Compound Profiling. “They're seven-transmembrane domain receptors that bind a lot of other signaling proteins, and to just think of G proteins is really limiting.”

Intercepting signals

On a positive note, this maturing understanding of GPCRs has enabled the development of increasingly sophisticated and high-throughput methods for receptor characterization and drug discovery. With more than 100 functionally opaque 'orphan' GPCRs and many more whose function is at best partially understood, researchers have their work cut out for them.

For initial assessment of potential ligands, binding assays are still a typical first step, and the use of radiolabeled ligand molecules remains the standard approach for quantifying specificity. “You can get some very precise measurements, not just of the receptor expression density but also on-off rates of ligand association and dissociation,” says Richard Eglen, president of Bio-discovery at PerkinElmer. As a less hazardous alternative for high-throughput screening, nonradioactive fluorescence-based platforms are also available, such as PerkinElmer's DELFIA and LANCE assays and the recently launched Tag-lite system from Cisbio.

Arthur Christopoulos and colleagues are investigating and designing molecules that modulate GPCR activity allosterically.

However, even a highly informative binding assay needs to be followed up by more detailed mechanistic studies, points out Bryan Roth, whose team at the University of North Carolina is dedicated to characterizing the specificity and functional properties of both pharmaceutical compounds and drugs of abuse, many of which also target GPCRs. “Our screening is biased toward functional screens just because they're higher-throughput, although the false positive rate is high,” says Roth. “Radioligand binding assays tend to be a lot more specific, but obviously there's information that can't be gained from them.”

The most mature functional assays are those that quantify second-messenger activity induced via G-protein activation. Activation is often accompanied by an influx of calcium ions, and a variety of sensitive assays are available for high-throughput screening of changes in Ca2+ in GPCR-expressing cells upon treatment with a compound or compounds of interest. PerkinElmer offers a pair of photoproteins that serve as effective indicators:aequorin and Photina, both of which are jellyfish-derived enzymes that convert the chemical substrate coelenterazine into luminescent coelenteramine in the presence of calcium. “They're engineered into the cell line, and they give a very bright luminescent response as the intracellular calcium is elevated,” explains Eglen. The FLIPRTETRA instrument from MDS Analytical Technologies is suitable for high-throughput analysis of such aequorin and Photina assays, but it also enables fluorescence-based analysis via the widely-used FLIPR Calcium Flux Assay.

Aequorin-based assays rely on calcium to facilitate the enzymatic production of coelentarimide, a bioluminescent compound. Credit: PerkinElmer

However, this only covers a subset of receptor activities. “Calcium mobilization assays are a well-established method for Gq-coupled GPCRs, but Gi- and Gs-coupled GPCRs have been traditionally assayed via cyclic AMP (cAMP) activity,” explains David Yamane, senior director of drug discovery marketing for MDS Analytical, which offers the CatchPoint immunoassay for quantifying cAMP production. Promega's GloSensor assay also represents a sensitive and straightforward tool for detecting GPCR-induced increases in adenylate cyclase activity using an engineered luciferase enzyme whose capacity to produce luminescence is directly activated by cAMP binding. “It [has] a fantastic dynamic range, and we can theoretically monitor the cAMP signaling in real-time,” says Arthur Christopoulos of Monash University.

The FLIPRTETRA instrument from MDS Analytical Instruments enables high-throughput screening of calcium flux for GPCR studies. Credit: MDS Analytical Instruments

Alternative routes

As mentioned above, GPCR activity does not end with G protein–mediated signaling, and researchers who stop their investigations at the second-messenger level are likely to miss the full story. For example, most activated GPCRs bind the protein β-arrestin, and although this was once thought to be primarily a mechanism for restricting receptor activation, it is now clear that arrestins trigger signal cascades of their own.

“You can inhibit G proteins altogether, and you'll still get arrestin activation,” says Tom Wehrman, director of cell biology at DiscoveRx. “And there are actually two GPCRs, CXCR7 and C5L2, that don't actually couple to any G protein but do activate arrestin.” DiscoveRx offers the PathHunter assay, which monitors receptor-arrestin interaction via enzyme fragment complementation1. In this approach, one segment of the β-galactosidase enzyme is linked to the GPCR of interest and the other is attached to β-arrestin, and only when the two interact is full enzymatic activity restored, enabling processing of a bioluminescent substrate. “The main advantage is that you're actually monitoring the binding event: it's a stoichiometric readout on receptor activation,” says Wehrman. Invitrogen also uses a fusion protein system in their Tango arrestin assay, with a protease-tagged arrestin that specifically cleaves a transcription factor fused to the target receptor upon binding, releasing it to enable activation of a downstream reporter.

Other assays are also emerging that gauge more pathway-independent responses to receptor activation, such as MAP kinase activity or receptor endocytosis, but ultimately, experts agree that the key to success lies in effectively combining readouts (for example, cAMP and Ca2+ or G protein–dependent and –independent activation) to overcome the biases inherent to any assay-based screening strategy and derive a more complete overall picture of receptor activity. “You must use multiple functional endpoints,” asserts Christopoulos. “In my lab, nobody can ask any questions without doing at least three functional assays and one binding assay where possible.”

The plot thickens

Even with many approaches available, uncovering the true nature of receptor activity is a challenge. The same receptor may not only activate different downstream pathways in response to different ligands—a phenomenon known as 'functional selectivity' or 'biased agonism'—but might even respond differently to the same ligand depending on the cellular context. This complexity is exacerbated by the existence of allosteric modulators, which bind outside the active site but can profoundly alter receptor ligand specificity or kinetics. “If you put an allosteric modulator on the receptor, you know you're going to make a new conformation, and so you could engender functional selectivity even with agonists that don't normally have it,” says Christopoulos. “We have examples of pathway-biased allosteric modulation.”

Such subtleties can potentially be lost in an overly engineered and artificial assay system. “In many cases, you identify a GPCR as a target based on physiological data, and the receptor might be expressed in the brain in a particular neuron, but then you perform a high-throughput screen in over-expressing, immortalized cell lines that are nothing like the cell in which the receptor normally resides,” says Kobilka.

Corning's EPIC system tracks changes in the distribution of molecules at the cell surface as an indicator of GPCR-triggered cellular activity. Courtesy of Y. Fang, Corning.

One potential solution might be the directed differentiation of embryonic stem or induced pluripotent stem cells to yield mature cells that are appropriate for the study of a given GPCR under normal or disease-specific conditions. In the end, however, animal models remain the ultimate goal. “If you keep introducing complexities, you're going to get a lot of interesting observations but you'll probably go down mechanistic paths that aren't relevant,” says Christopoulos. “You're better off taking a small, judicious sampling of compounds in vivo sooner and then coming back to deconvolute.”

Roth, whose team has developed cell lines for screening purposes that express nearly 250 different GPCRs, concurs. “There are so many ways to be misled by using an overexpressed receptor in a non-native cell line,” he says. “If we make a discovery, we try as best we can to validate it in a native tissue and then preferably with a wild-type and a knockout mouse.”

Without a mark

A newer alternative that has given hope to some are the so-called 'label-free' screening technologies, which are designed to enable screening of receptor activity in virtually any cell line or primary cell culture, without the need for receptor engineering or other genetic modification. “That's going to be the coming thing: we'll re-emerge into the old way of doing things, in real time with human tissues and primary cells,” says Kenakin. “Tissues that are not healthy but actually model the illness we're trying to cure.”

In the CellKey system from MDS Analytical a constant voltage is applied across cell monolayers cultured atop electrodes and changes in the ratio of voltage to current across that monolayer—a property known as impedance—resulting from drug-induced rearrangements in cellular organization are tracked. “Changes in cell adherence, shape, volume, and cell-cell interactions that occur upon exposure of cells to a stimulus, such as a ligand, contribute to the impedance signal,” explains Yamane. “These factors, individually or collectively, affect the flow of current and thereby influence the magnitude and characteristics of the impedance signal.”

Corning's EPIC system implements an alternative, optical approach using a resonant waveguide biosensor, which quantifies alterations in the refraction index at the membrane-sensor interface of cultured cells. In the context of GPCR screening, these changes result from molecular rearrangements induced by receptor activation. “Cell signaling, particularly mediated through receptors, often involves protein trafficking, microfilament remodeling, cell adhesion alterations and morphological changes of cells, all of which can lead to [substantial] dynamic mass redistribution,” explains Ye Fang, senior research manager at Corning. “Such redistribution is not random; instead, it is tightly regulated and is often dynamic both spatially and temporally, and the biosensor simply acts as a noninvasive monitor to record [this] in real time.”

On one hand, label-free platforms offer multiple advantages, enabling screening of virtually any cell type that can be successfully cultured in real time and in a pathway-independent manner. “[Label-free systems] can act as a single platform for all classes of GPCRs,” says Fang. However, some screeners are skeptical about how effectively these data can be interpreted. “They're really what most of us would consider to be 'black box' readouts,” says Roth. “All you can basically say is that you added a drug and something happened to the cell.”

Recent successes in GPCR crystallization, including this structure of the β2-adrenergic receptor from Brian Kobilka and colleagues, are helping propel the field forward. Courtesy of B. Kobilka.

In some cases, this may save time and resources. “When you do a cyclic AMP assay, you know what's happening there but you don't know what else is happening; I don't know if that's any better than getting a generic response and not understanding exactly what's happening,” says Kenakin. “There are tools you can use to get at that, like [small interfering] RNA or enzyme inhibitors.” For some scientists, though, this technology is simply still too new and for now, too expensive for any categorical assessment. “I don't know how far it's going to go in telling us things, ... but I don't want to be too negative,” says Christopoulos. “Once things settle down, we'll see where this ends up.”

Making a match

The floppy, complex structure of GPCRs has made them exceedingly difficult to crystallize; for many years, the structure of rhodopsin was the only high-resolution structure available, making direct chemical screening the only game in town for drug discovery. “For a virtual library screen, you have to get the contacts exactly right, and so you're looking at less than two-angstrom resolution in the binding pocket,” says Christopoulos. “Most virtual screens that aren't based on real crystal structures fail.”

This is starting to change, with recent structures for the β1-adrenergic2, β2-adrenergic3 and A2A adenosine4 receptors potentially heralding accelerated progress in GPCR crystallography and thus new opportunities for in silico drug discovery. Brian Shoichet's team at the University of California at San Francisco is at the forefront in this regard, having recently applied their molecular docking algorithm, DOCK, to the Kobilka group's β2-adrenergic receptor structure to identify potential ligands from among one million commercially available compounds5. “By our standards, it was an unbelievable success,” says Shoichet. “We found that about 25% of the molecules we predicted to bind, did bind, which is about five- to tenfold better than we do with non-GPCR sites. The best hit we got was a nine nanomolar inverse agonist, ... and it turns out that this is the best inverse agonist that's been characterized to date.” Members of the Christopoulos lab are likewise gaining momentum in their ab initio drug design efforts, using the steady increase in available structural data to facilitate the design of allosteric modulators that specifically target peptide-binding GPCRs. “Transmembrane regions are where the crystal structures are getting quite good, ... and we believe those will be novel allosteric sites ripe for targeting,” he says.

Shoichet and Roth have also been working together to explore the potential of applying computational methods to a more 'classical' pharmacological analysis, finding targets that match a given drug rather than the other way around. In a recently published study6, they analyzed thousands of pharmaceutical compounds, comparing them chemically to known receptor ligands and using those similarities to reveal new information about drug specificity and potential off-target effects. According to Shoichet, the results revealed insights that would have been overlooked by conventional genomic or proteomic analysis. “Things that look unrelated by sequence look highly related by the ligands they bind,” he says. “For instance, serotonergic receptors have seven major subtypes; serotonin 3 is an ion channel, and all the others are GPCRs; ... their sequences are completely different, but they're recognizing similar molecules.” Their findings also enabled them to identify the physiological basis for known drug side-effects, such as the capacity of fluoxetine (Prozac) to act as a beta-blocker and unexpected affinity of the antiviral drug delavirdine (Rescriptor) to the histamine H4 GPCR.

Shoichet believes this is only the beginning of what promises to be a fruitful, parallel avenue for drug research. “I think that going back to the classical pharmacology view of typing receptors by the ligands that bind to them as an overarching idea is a game-changer,” he says.

Tracking a moving target

With so many technological resources at hand for functional analysis and ligand discovery, most scientists working with GPCRs agree that aggressive pursuit of structural data is now a top priority, although it would be a mistake to assume that even 'solved' structures can be considered a closed book. “I think the binding pocket is going to differ for different active states of the receptor, so you have to have other structures,” says Kobilka, adding that ongoing investigations with nuclear magnetic resonance (NMR) spectroscopy in his and other labs may complement crystallography and help capture some of the more dynamic details of receptor behavior. “The methodology and instrumentation for high-resolution structure determination of GPCRs by NMR [spectroscopy] aren't quite there yet, but NMR [spectroscopy] can provide a lot of interesting information about the ligand-induced movement of specific domains and the dynamics of different parts of the protein,” he says.

In the meantime, every new structure helps. “There are still so few that each one is really going to affect a lot of people,” says Shoichet. “It really brings the field a big bolt of energy.”

Table 1 Suppliers guide: companies offering reagents and hardware for GPCR-based research
Full size table