Meeting Report

The year-end meeting of the Chemical Biology Discussion Group offered a variety of perspectives on the theme of biomolecular recognition; namely, finding and synthesizing molecules that can bind to specific regions of proteins or DNA in predictable and useful ways. Featuring a keynote talk and presentations by students and postdocs in New York-area chemical biology labs, all linked back to the central issue: harnessing the chemistry of biology for new applications.

In the keynote talk, Adrian Whitty of Boston University described an effort to find small molecules that disrupt protein–protein interactions and that can be orally administered—a holy grail for drug discovery. Although some drugs exist that target these interactions, almost all are peptides or proteins that must be administered by injection.

Traditional small molecule targets typically have pronounced clefts, which small molecules can exploit to achieve strong binding. However, surfaces where two proteins interact are often relatively flat, Whitty said, with no pre-existing pockets suitable for binding a small molecule. As a result, it is difficult to find drug-like molecules that bind strongly to a protein–protein interface using traditional small molecule screens.

However, binding surfaces between proteins typically have energetic "hotspots"; that is, smaller regions within the interface that generate the bulk of the binding energy. Targeting small molecules to those hotspot regions has the potential to achieve stronger binding. In addition, it has been shown that some protein–protein interaction surfaces can conformationally adapt upon small molecule binding to produce a cleft or pocket. Screening for these potential adaptive binders cannot yet be done reliably with modeling or docking studies: identifying such binders requires experimental screens.

In an attempt to exploit the binding energy of hotspots and potential adaptive binding, researchers sometimes screen for drug leads using the method of fragment-based lead identification. Screening a drug target against libraries of low molecular weight "fragments" allows researchers to efficiently examine how ligands could productively bind to a target in different arrangements. Growing these initial fragment hits, by adding functionality to pick up additional binding interactions with nearby regions of the protein target, can generate stronger binders and eventually drug leads and drugs. However, fragments typically bind only weakly to their targets, so identifying fragment hits and quantifying their affinity can be challenging.

Sunesis has developed an approach to fragment-based lead identification called Tethering®, which involves engineering a cysteine residue on the protein's surface near a region of interest and screening with fragments that contain a disulfide linkage. Through disulfide exchange any fragment can form a covalent disulfide bond with the engineered cysteine. However, fragments that also engage in favorable noncovalent binding to the target will produce more stable adducts. Reaction is carried out under equilibrium conditions, so the extent of disulfide adduct formation with the protein is governed by binding affinity, allowing stronger binders to be identified.

Screening disulfide-linked drug fragments against proteins with added cysteine residues provides a way to find new targets.

Whitty and colleagues targeted a set of TNF family cytokines that bind to related receptors from the TNF receptor superfamily. These trimeric proteins can contribute to pathophysiology in autoimmune diseases such as rheumatoid arthritis (RA), Crohn's disease, and systemic lupus erythematosus. Current drugs available to treat the inflammation of RA and Crohn's disease include etanercept (Enbrel, Amgen/Wyeth), infliximab (Remicade, Centocor Ortho Biotech), and adalimumab (Humira, Abbott), all of which are protein drugs that must be administered by injection. No selective small molecule TNFα antagonists had previously been reported.

To allow the extensive receptor binding interface of TNFα to be screened for fragment binding using Tethering®, the researchers engineered 15 different TNFα cysteine mutants. Initial screens generated several hits containing a pyridyl pyrimidine core structure as well as a number of other structures. Molly He of Sunesis solved the crystal structures of the disulfide adducts of the most promising hit fragments bound to the mutant TNFα proteins, The structures showed that the fragments bound to two primary sites within the receptor binding region of TNFα. Whitty and his colleagues then worked to optimize the most promising hits that bound in that region, adding functional groups that could contribute binding affinity based on their interactions with adjacent sites on the protein, and eventually removing the disulfide linker to create an extensive set of soluble candidate ligands.

To validate these soluble molecules, the team used an array of biochemical and biophysical assays, including NMR analysis, to evaluate binding. X-ray crystallography was used to confirm the binding mode of the most promising hits. Of the 772 soluble compounds synthesized, only two made it to the structural evaluation stage. When the pyridyl pyrimidine portion of these small molecule ligands was resolved, the group found that the untethered molecules interacted with the same general region of the protein but did not recapitulate the binding mode seen in the tethered probes. These hits were determined to be unsuitable for further advancement as drug leads.

The team carried out similar tethered fragment-based screens for two other TNFα family members, CD40L and BAFF. CD40L gave only weak tethered hits. BAFF gave multiple strong hits and crystallographic analysis showed that all bound in the same small region of the protein surface. However, there were no nearby secondary binding regions that might be exploited to grow the fragment hits into stronger binders, and so work on this target was also discontinued.

Although the effort to produce small molecule inhibitors of these proteins failed, the researchers learned important lessons for future studies. It's possible that cysteine mutations necessary to enable fragment screening by Tethering® altered the protein dynamics, producing a more flexible structure than the wild-type protein and allowing formation of a spurious binding site for the tethered pyridyl pyrimidines. In addition, compound design was somewhat biased toward conventional drug-like compounds, which might not be optimal for protein–protein interactions. It's possible that these TNF family proteins are simply too difficult to target with conventional small molecules, Whitty said. According to Lipinski's Rule of Five—guidelines that describe properties that tend to correlate with drug-like behavior—most drugs are of only moderate lipophilicity and have molecular weights less than 500 Daltons. To solve this so far intractable problem, it might be necessary to use molecules that don't fit these conventional criteria, Whitty said.

Efficient access to natural-product-like molecules for screening

Renato Bauer continued the discussion of generating diverse lead molecules for drug discovery. In Derek Tan's laboratory at Memorial Sloan-Kettering Cancer Center, Bauer and his colleagues are using diversity-oriented synthesis to generate large libraries of potential drug leads, modeled after common structural characteristics in natural products. Bauer and postdoctoral researcher Christine DiBlasi generated a library of compounds with dense mono- and polycyclic ring systems modeled after alkaloid and terpenoid natural products. The syntheses were developed from cyclization reactions of t-butylsulfinamide enynes or diynes. Collaborating researchers at the Broad Institute have taken that initial library and added other commercially available building blocks through solid-phase synthesis.

Statistical analysis reveals a library that shares features of drug candidates and natural products.

With the resulting library, Bauer and his colleagues have used statistical analysis to compare 17 molecular descriptors for each molecule—attributes such as the number of stereocenters and the molecular complexity—with the parameters for existing drug libraries and known natural products. This analysis of chemical space indicates that the new library bridges characteristics of both existing drugs and natural products, which suggests that molecules in this library could encompass the best characteristics of each group.

Understanding the activity of nitrogen mustards

Nitrogen mustards are well-established chemotherapy agents that act by chemically cross-linking two strands of DNA via nitrogen atoms on the bases. Angelo Guainazzi of Stony Brook University described his work in Orlando Schärer's laboratory that aims to better understand the structure of these cross links and the biochemistry of how DNA repair enzymes excise these lesions. The cross-linking structures are typically difficult to isolate and are chemically unstable. Therefore, Guainazzi and his colleagues have synthesized analogs that replace the guanosine nitrogens with carbon atoms and produce aldehydes that can then be cross-linked by reaction with hydrazine.

Guainazzi and his colleagues have successfully synthesized these analogs and they have used DNA denaturing gels to demonstrate that the analogs can form cross links. Molecular dynamics studies carried out in collaboration with Arthur Campbell showed that nitrogen mustards form cross links of 7.5 Å. The group was able to vary the length of the analog cross link to match this length. They also have made cross links of 8.9 Å (similar to B-form DNA) and longer cross links to probe how cross-link length might affect cellular processing of these lesions.

A bipedal DNA Brownian motor with coordinated legs

Working in Ned Seeman's laboratory at New York University, Tosan Omabegho and his colleagues have developed a bipedal DNA-based motor with coordinated legs. The motor's motion is modeled after the movement of biological motors such as kinesin that move processively along a track. Developing synthetic motors with similar properties requires a system that anchors a foot to the track at all times. DNA provides a useful template for walker, track, and fuel to propel the walker forward.

Previous work in this area has included the synthesis of nonautonomous walkers that required the addition of fuel to drive them and autonomous walkers that were not directional. In this system, the walker feet are complementary to stepping points on the track. The addition of a hairpin of DNA "fuel" releases the foot on the left side, leading it to make a forward step to the next available complementary strand. Coordination between the two feet is essential in order to achieve forward motion. Spent fuel occupies the previous points on the track, preventing backward motion. Steps in the forward motion were identified by cross-linking the DNA strands with psoralen at individual steps.

Bacterial isoprenoid biosynthesis as an antibiotic target

According to research by Scott Lefurgy and his colleagues in Tom Leyh's laboratory at Albert Einstein College of Medicine, targeting the isoprenoid biosynthesis pathway in Streptococcus pneumoniae could lead to the development of new antibiotics. This pathway is required for survival of the pathogen and a downstream product, diphosphomevalonate (DPM) serves as a non-competitive inhibitor, binding at an allosteric site on the enzyme mevalonate kinase. The human homolog of this enzyme is not inhibited by DPM.

The inhibition in S. pneumoniae by DPM might extend to other gram-positive bacteria, Lefurgy said. He has tested homologous enzymes from related bacteria and will publish the results shortly. In future work he and his colleagues will target another enzyme in the pathway, diphosphomevalonate decarboxylate (DPM-DC), using mechanism-based inhibitors.

Scaffolds for engineering allosterically controlled biomolecular recognition

Mark Blenner of Scott Banta's laboratory at Columbia University described the development of biomolecular switches based on the RTX motif of the adenylate cyclase of Bordetella pertussis. Calcium binding controls the folding of this motif from a disordered strand into a beta roll, a parallel β-helix structure. The peptide that the researchers isolated from the protein demonstrates the desired conformational change and that conformation is reversible.

Blenner determined that the flanking residues outside the beta roll structure are required for folding. The C-terminal cap is required for the calcium-induced conformational change, but the N-terminal cap is not. Entropic capping may play a role in the conformational change: when a non-folding peptide sequence had fluorescent proteins added as N- and C-terminal caps, those new caps induced a calcium response. Blenner and his colleagues are now randomizing the surface-exposed residues on these RTX sequences and doing directed evolution studies to evolve binding function for other targets. These peptide regions are highly expressed within fusion proteins in cells and translocate to the cell membrane.

Visualization of fatty-acylated proteins in mammalian cells during Salmonella infection

Salmonella infection changes the membrane trafficking in host cells, and protein lipidation appears to be an important part of this process. To better understand this process, Guillaume Charron in Howard C. Hang's laboratory is using chemical reporters to track fatty acylation of proteins, without the need for radioactively labeled lipids. The assay is based on click chemistry, where the desired moiety—the fatty acid—and a fluorescent tag are labeled with uniquely reactive functional groups. Upon lysis of the cell, the reactivity of the desired tagged fragments with the corresponding fluorescent molecules allows the researchers to identify the proteins of interest.

Fluorescent labeling allows tracking of fatty-acylated proteins in cells.

In this system, the most efficient labeling method was to produce fatty acids with an alkyne that could react with an azide group on a fluorescent label. Charron and his colleagues could distinguish among different fatty acid groups on proteins and in a variety of cell types. They are now using this system to look at fatty acid acylation in Salmonella infection. A cleavable detection tag allows them to efficiently retrieve labeled proteins for proteomic analysis, and they have also used the fluorescent label to visualize the fatty-acylated proteins in cells.

Open Questions

Will it be possible to find small molecule drugs that target the TNF family of cytokines?

Will targeting protein–protein interactions require researchers to rethink Lipinski's rules that have guided drug discovery over the last decade?

Will libraries of small molecules with characteristics more like natural products lead to new or better pharmaceuticals?

What will be involved in the design of an autonomous, bi-directional DNA walker?

Which proteins are tagged by Salmonella bacteria during infection?

Will inhibitors of the mevalonate kinase pathway in S. pneumoniae lead to new antibiotics?