If you have an ACS member number, please enter it here so we can link this account to your membership. (optional)

ACS values your privacy. By submitting your information, you are gaining access to C&EN and subscribing to our weekly newsletter. We use the information you provide to make your reading experience better, and we will never sell your data to third party members.



Piece By Piece

More and more companies are using fragment-based lead design as a drug discovery strategy

by Sarah Everts
July 21, 2008 | A version of this story appeared in Volume 86, Issue 29

shaking hands
Credit: Maurizio Pellechia
Small fragments that bind in nearby pockets can be connected to make a lead.
Credit: Maurizio Pellechia
Small fragments that bind in nearby pockets can be connected to make a lead.

IF YOU WERE TO amble up to a typical medicinal chemist with a compound that you hoped they could optimize into a drug lead, their first question would probably be, "How well does it bind?"

In the pursuit of new pharmaceuticals, many medicinal chemists want to start their leg of the drug discovery race with a drug-sized molecule that binds with a tenacious grip—we're talking nanomolar potency—to its biological target. After all, there are so many molecular traits to optimize, such as reducing a drug lead's toxicity and increasing its solubility in the body, that beginning with high-binding affinity seems like starting on the right foot. That's why high-affinity hits are the primary aim of high-throughput screening (HTS), a bread-and-butter starting point for drug lead discovery.

But a growing number of medicinal chemists are leaving the high-affinity paradigm behind. These researchers are sidestepping some of the cherished tenets of HTS in favor of an emerging drug discovery strategy called fragment-based lead discovery (FBLD).

Medicinal chemists who embrace FBLD try to build blockbusters by identifying small chemical fragments—often mere functional groups, which may bind only weakly to biological targets. Then they combine or expand the fragments to make a lead with stronger affinity. This approach enables researchers to take a second glance at hits that bind with millimolar concentrations, an affinity that's several orders of magnitude lower than what is considered a promising HTS hit.

Without closer examination, the FBLD strategy may sound a bit absurd: It's as if a jockey was given one hoof, a skinny thigh, and a sparse mane and told to build a horse that could win the Kentucky Derby.

But in its relatively short lifespan—it only got going as a drug discovery strategy in the late-1990s—FBLD has inspired 10 or so leads that are now percolating through clinical trials. "Currently, people who do drug discovery are intrigued" by FBLD, says President and Chief Scientific Officer Vicki Nienaber of Zenobia Therapeutics, La Jolla, Calif., who is a pioneer of FBLD. "They notice that FBLD strategies are getting leads into the clinic fast."

At least 20 companies have firmly established FBLD units. They are a mix of start-up and established biotech firms and a sprinkling of big pharma. The technique's appeal was also evident earlier this year at a conference in San Diego, Calif., which attracted 350 participants and included representatives from all major pharma. Some biotech companies are solely focused on FBLD, while many more companies are starting to use FBLD in combination with other established techniques such as HTS, combinatorial chemistry, and in silico drug design.

The FBLD strategy is fundamentally minimalist. Unlike users of HTS, who sift through millions of drug-sized molecules in compound libraries, FBLD proponents set their sights far more simply.

They opt to screen small libraries that each contain only a few thousand tiny chemical fragments, such as amines or biaryl carboxylates. These fragments are 150–300 daltons in size, about half the size of most HTS library compounds. The fragments are "minimal" molecular architectures—structures having usually fewer than 20 heavy (nonhydrogen) atoms—that are potentially capable of binding choice locations on biological targets.

THE LOGIC of sampling small chemical fragments is that it enables a greater fraction of chemical space to be examined for a given target, says James A. Wells, a pharmaceutical chemist at the University of California, San Francisco.

"Chemical space is vast," Wells explains. There are an estimated 1062 different molecular scaffolds possible for a 500-Da compound—the typical size of a molecule in an HTS library. "In a typical high-throughput screen you may test millions, or 106, compounds. Even though that's a lot of molecules, you are not actually sampling a huge fraction of that chemical space," Wells continues.

"But in fragment land, where the molecules are around 200 Da, the total number of possible molecular scaffolds is approximately 1010," Wells says. So even if you screen just a few thousand fragments, or 103 compounds, you have accessed a greater total fraction of that available chemical space, Wells notes.

While they do deliver drug leads for some targets, large corporate HTS libraries "barely scratch the surface of available chemical space," said Philip J. Hajduk, a medicinal chemist at Abbott Laboratories, at the FBLD conference.

Fragment libraries are designed with several parameters besides just chemical minimalism in mind. For example, library developers exclude compounds with known reactive or toxic groups. They use clustering techniques to select fragments that are chemically diverse. And then chemists do further filtering to only include fragments that are soluble and seem suitable as starting points for chemical optimization, says Roderick E. Hubbard, professor of chemistry at the University of York and a senior fellow at Winnersh, England-based Vernalis.

After screening a fragment library, chemists expand the weakly binding fragment hits into a drug lead either by adding chemical groups—"growing," as the FBLD lingo goes—or by linking a sequence of fragments together.

Growing a fragment into a high-affinity lead is most effective when the target has a well-defined pocket, such as the adenosine triphosphate binding site of kinases, allowing "the impact of each new functional group or atom to be carefully assessed and maximized," notes Glyn Williams, director of biophysics at Cambridge, England-based Astex Therapeutics. On the other hand, linking together several weakly binding fragments is useful when the target has several proximal, shallow pockets that need to be plugged.

"FBLD relies on 'avidity,' the idea that two things bind better than one," Wells says. "It's about discovering a drug one piece at a time."

FBLD proponents often cite examples of how weakly binding hits can be made into a lead with nanomolar affinity. At Astex, "we expect a 1 millimolar fragment hit could be transformed to a 1 micromolar lead in six months, and to a 100 nanomolar lead in nine to 12 months," Williams says.

Proponents also like to point out that a millimolar hit may not be as daunting a starting point when you consider a concept called ligand efficiency, which is a ligand's binding energy for a specific protein averaged over all the ligand's heavy atoms. A large hit with nanomolar affinity may have the same ligand efficiency as a small fragment binding with millimolar affinity. "It's a different perspective," says David Rees, vice president of medicinal chemistry at Astex.

Inspiration for fragment expansion can come directly from other fragment screen hits, from in silico modeling, or sometimes from HTS hits—any tool that works to make a potent clinical candidate. Medicinal chemists often make use of the opportunity to grow an FBLD compound by starting with an initial fragment hit and adding chemistry that avoids intellectual property issues, increases solubility, or reduces toxicity—all without messing with the target-binding moiety. But the initial fragment hit "is the anchor—it binds in the hotspot of your target," Rees says.

Many proponents say the ideal biological target for FBLD drug development is one with one or more well-defined pockets, where a fragment can be grown to fill the space. One of the favorite targets of FBLD researchers is HSP90, a protein that has just such a pocket, which it uses to help its protein clients fold into their three-dimensional shapes. "You haven't done FBLD unless you've done HSP90," Nienaber jokes.

"HSP90 has a list of clients that reads like a who's who of oncogenic proteins," says Martin Drysdale, deputy research director at Vernalis. It also plays a role in central nervous system disease and malaria.

Another instance where FBLD shines is when a company wants to develop leads for a new biological target, Hubbard says.

Most corporate collections contain molecules that have been optimized for historical targets, Hajduk explains. So HTS will bring up hits for targets that are structurally similar to targets it has found hits for in the past. But the technique sometimes fails to bring up hits that interfere with new targets, such as protein-protein interaction surfaces. FBLD allows researchers to be therapeutic-area agnostic. "It allows us to look for naÏve binders," Wells notes. Adds Hubbard, "There's only a small chance that within even a large corporate collection you'd find compounds with the perfect constellation of shapes to fit a new target."

In fact, proponents like to point out that not only can you get hits from FBLD for a variety of new targets, but the approach can also generate a lot of them.

For example, a group of FBLD chemists at Novartis led by Angsar Schuffenhauer reported in Current Topics in Medicinal Chemistry that fragment screens had 10- to 1,000-fold higher hit rates than conventional HTS screens (2005, 8, 751). And Berkeley, Calif.-based Plexxikon has filed five FBLD-based Investigational New Drug Applications with the Food & Drug Administration in six years of active chemistry, a relatively high number for a start-up.

ALTHOUGH getting a selection of hits from fragment screens is common, the ability to detect these hits was one of the initial technical challenges to overcome before fragment-based screening could be applied widely, says Daniel A. Erlanson, associate director of medicinal chemistry at San Francisco-based Sunesis Pharmaceuticals. The biochemical screens common in HTS were simply not sensitive enough to detect the low-affinity hits from FBLD screens.

The key was to take biophysical methods, which are sensitive to millimolar binding, and to develop them for medium-throughput screening, which happened in the 1990s. "That's when the FBLD technique could take off," Erlanson notes.

Abbott Laboratories was the first company to seriously cultivate the FBLD technique. In 1996, Abbott scientists led by Stephen Fesik, now divisional vice president of cancer research, published a seminal paper in Science (1996, 274, 1531) that "really demonstrated that FBLD was not just a theoretical possibility but a practical approach," Erlanson notes.

Abbott initially focused on using nuclear magnetic resonance chemical shifts of protein targets to detect fragment binding. Not only is NMR sensitive enough to detect weak binders, but because chemical shifts are correlated with the local environment of specific atoms, the technology also allows researchers to pinpoint the precise amino acid residues near to which a fragment binds. Thus it is possible to distinguish binding of a fragment in a desired location—for example, in an active site or within a protein–protein interaction pocket—from binding at an undesirable site.

Credit: Joe Patel/Astex Therapeutics
A fragment for a heat shock protein target grows into a lead.
Credit: Joe Patel/Astex Therapeutics
A fragment for a heat shock protein target grows into a lead.

OTHERS DETECT binding by observing the NMR signal of the fragment, instead of that of the protein target. The logic is the same: An atom's local environment will change upon binding, so the NMR spectrum of its molecule will also change.

Following closely on the heels of NMR-based screening came the development of X-ray crystallography for fragment-based screening at Abbott and elsewhere. Typically, crystallographers soak the crystals of a biological target with fragments, or cocktails of fragments, and solve the structures of the bound targets. Again, the analyses provide structural information about where hits bind.

Although NMR is still widely used for screening, some FBLD researchers say that X-ray crystallography screening is starting to become more predominant than NMR in industry for FBLD screening because more companies are better set up for medium-throughput X-ray crystallography than for medium-throughput NMR. Others maintain that NMR is still as predominant as ever.

The weakness of both NMR and X-ray crystallography screening is that the techniques rely on high concentrations of both fragments and target. This means fragment libraries must be soluble at high concentrations. It also often requires that the target's structure be solvable and, in the case of X-ray crystallography, that the crystal have a suitably accessible binding site to allow ligand binding.

More recently, surface plasmon resonance (SPR) has emerged as a tool for sensitively detecting millimolar binding. SPR works by measuring changes in the refractive index at a surface. If a protein target sits on that surface and a fragment binds to the protein, the index of refraction of the entire surface will change, and the hit will be registered.

One of the perks of using SPR is that it provides quantitative dynamics data on the binding interaction, such as binding constants, which are complementary to the structural information from X-ray and NMR screens, says David G. Myszka, head of Center for Biomolecular Interaction Analysis at the University of Utah and a pioneer in developing SPR for fragment studies.


Another advantage is that SPR makes it extremely easy to distinguish nonspecifically binding hits from good hits, and "you don't need a garden hose worth of protein to do the screens," Myszka adds.

"We do SPR much more often and much earlier than we used to," says Vernalis' Drysdale. Indeed, most companies actively involved in FBLD do some mix of NMR, X-ray, SPR, and virtual screening to obtain hits.

RESEARCHERS ARE also using other techniques such as mass spectrometry and calorimetry, albeit to a lesser extent, to detect fragment binding. And some biotech firms use their own patented screening technology to detect fragment binding to targets. For example, Sunesis has developed a technique called "tethering" to find fragment hits. First, through protein engineering, they place a cysteine residue near the target protein's active site. Then they attach a disulfide group to every component of a fragment library and screen the modified library. Fragments that bind near the target's cysteine can participate in a disulfide exchange, which chemically tethers the fragment to the protein. Mass spectrometry then detects fragment hits.

Other innovative chemistry can also be used to detect hits. For example, click chemistry has been used to screen for inhibitors of a zinc-containing metalloenzyme. Knowing that an aromatic sulfonamide could associate with the enzyme's zinc ion, researchers led by chemists K. Barry Sharpless and Chi-Huey Wong at Scripps Research Institute, added an acetylene to the sulfonamide and then used a library of azide moieties and click chemistry to capture inhibitory fragments that bound near the enzyme's zinc-containing active site.

In addition, some groups are building upon existing FBLD protocols to create new sophisticated screening strategies. For example, Maurizio Pellecchia, of the Burnham Institute for Medical Research, in La Jolla, attaches paramagnetic moieties near the active site of his biological target. Then he screens fragments and looks at the NMR signals of the fragment molecules, taking note of those whose signals have been quenched by binding near the paramagnetic compounds in the active site.

FBLD is certainly not as prevalent in universities as in industry. But "it's also ideal for an academic setting," Pellecchia says. Because fragment-based screening requires access to a library of only a few thousand compounds, "it's not as expensive as HTS." Pellecchia primarily uses FBLD to look for new compounds for biological target validation, which might then be developed into leads in collaboration with industry. Other academic labs, such as that of Gabriele Varani at the University of Washington, use FBLD to develop drugs for neglected diseases, such as malaria and HIV.

And because the function and inhibition of the targets from different fragments can be explored in a "hypothesis driven" manner, FBLD can be more stimulating than HTS as a basis for academic research, Pellecchia adds. In fact, academic researchers have substantially contributed to FBLD technology development. For example, Wim G. J. Hol, who is now at the University of Washington, and his colleagues performed some of the groundbreaking work to apply X-ray crystallography to FBLD.

Nevertheless, industrial interest in FBLD is especially strong. Some 20 companies are heavily involved in the approach, and many more are testing the waters. Every major pharmaceutical company was represented at the FBLD conference in San Diego earlier this year. And several contract research organizations are adding FBLD to their roster of services.

However, because FBLD is still primarily the purview of start-ups, practitioners have been vulnerable to job loss. For example, an FBLD company called ActiveSight was closed last month by its parent company, the Japanese instrument company Rigaku. Some of the former ActiveSight employees have re-formed into Zenobia, under the direction of Nienaber, who earlier established the X-ray FBLD screening technology at Abbott Laboratories.

In addition, all of the preclinical research staff at Sunesis, including medicinal chemists doing FBLD research, learned in June that they would be laid off in August so the company could focus on getting several of its clinical candidates to market. Plexxikon also recently cut a third of its staff "to tighten up in the area of discovery research in order to focus on developing our existing products," which include two FBLD-inspired compounds in clinical trials, says the company's media spokeswoman, Jennifer Cook Williams.

THESE CHANGES don't indicate that FBLD has failed, comments Wells, who was involved in founding Sunesis, "but that these start-up companies can only survive if they have late-stage clinical programs," which absorb a huge budget.

"This trend ironically comes when the FBLD technology is really in log-phase growth," Wells adds.

Despite the turmoil among some of the FBLD start-ups, the FBLD technique has been more easily implemented in smaller biotech companies than in big pharma—although this is changing, Hubbard says. "FBLD requires researchers with different skill sets—such as NMR spectroscopy, medicinal chemistry, and virtual screening—to come together at an early stage in drug development.

"The big pharma companies have always had the relevant skill sets; they just haven't always brought them together," Hubbard says. "So you are seeing a lot of larger companies partnering with smaller biotechs to learn how to bring all the skill sets together early on," he explains.

For example, in early July, Eli Lilly & Co. announced that it had signed a merger agreement with the FBLD biotech SGX Pharmaceuticals, based in San Diego (C&EN, July 14, page 23). According to its press announcement, Lilly acquired the company to have access to "SGX's fragment-based, protein structure, guided drug discovery technology, and to a portfolio of preclinical oncology compounds focused on a number of high-value kinase targets."

When asked what the main challenges of FBLD are, many point to chemists who cringe at the thought of being handed over a millimolar hit to optimize. Michael Hennig, the head of discovery technologies at F. Hoffmann-La Roche, says it was initially hard to change the mindset of chemists who are not used to starting with weak binders.

"When I would mention that I had a millimolar lead, people would think it was nonspecific and certainly too weak to be considered a lead," Abbott's Fesik adds.

"There's a cult of affinity. Many medicinal chemists were taught that if a hit is not nanomolar, it's nothing," Sunesis' Erlanson says. "FBLD is a revolution in the way chemists think about drug development."

FBLD-inspired drug candidates such as these are currently percolating through clinical trials.
FBLD-inspired drug candidates such as these are currently percolating through clinical trials.

WHILE FBLD scientists are quick to describe the benefits of FBLD relative to HTS, even the heartiest of FBLD proponents don't claim HTS will be ousted by fragment methods. But the frequent comparisons beg the question, "Which technique provides more drug leads?" Abbott, the big pharma company that has been longest involved in FBLD research but is also committed to HTS, is probably most informed to answer.

In a Nature Reviews Drug Discovery article (2007, 6, 211), Hajduk and coworkers at Abbott compared and contrasted the relative successes of HTS and FBLD techniques that were concurrently used to find drug leads for 45 different biological targets.

Fragment-based screens were successful in identifying solid hits for three-quarters of these targets, compared with just over half the targets for HTS screens, "validating the premise that fragment screening can deliver more hits against a larger number of protein targets," they write. In lead optimization for hits from those FBLD and HTS studies, nanomolar leads were obtained for 31% of targets using FBLD hits but only for 26% of targets using HTS hits. The Abbott chemists conclude that the two lead generation methods are "highly complementary."

"In some cases the fragment leads were optimized in parallel to yield completely novel, alternative chemotypes, whereas in other cases the fragment leads were incorporated directly into the optimization of the HTS leads," they note. FBLD has indeed sometimes been used in concert with other techniques such as HTS, combinatorial chemistry, or in silico screening to develop the leads currently percolating through clinical trials.

Furthermore, reports of HTS-FBLD crossbreed screening strategies—with names like "high-concentration screening" or "reduced complexity screening"—are beginning to pepper the literature.

Amid the confidence many proponents of FBLD have about their drug discovery strategy, "there is still skepticism within the pharmaceutical industry that the optimization process to give nanomolar potency leads from very low-affinity hits can be more broadly and routinely achieved," notes Astex's Miles Congreve in a recent review of FBLD (J. Med. Chem. 2008, 51, 3661).

So many FBLD proponents look to the year 2011 for vindication. It would be the earliest that a drug developed from FBLD techniques could be marketed, given the typical 15-year drug discovery pipeline.

An FBLD drug that reaches the market by 2011 would be a "psychological" victory for the whole FBLD community, Erlanson says.

"Consider the trajectory of any new technology," he explains. "First there's great excitement, followed by a crash, and then some steady state. You saw it happen with high-throughput screening and combinatorial chemistry.


"For FBLD, we are still on the upswing. But there's a huge push to get FBLD leads to the clinic, first, because of the basic drug discovery pipeline pressures, second, because many of the companies doing FBLD are small start-ups, and venture capital money will eventually run out. But third, we are still on the upswing of excitement about FBLD, and we want to get a drug before FBLD interest peaks, then falls."

Others consider the field more established. "An FBLD drug on the market would be a nice milestone," Wells says. But FBLD-based compounds "have reached the late-stage clinic. It's already obvious that the strategy works."

[Top of Page]


This article has been sent to the following recipient:

Chemistry matters. Join us to get the news you need.