Tapping the Potential of Protein Drug Discovery

July 1, 2005 (Vol. 25, No. 13)

Next-Generation Therapeutics Likely to Emerge from this Promising Field

Although protein drug discovery is still a new field, many companies are developing sophisticated technologies to enhance its potential. These companies will be presenting information on their most recent efforts at the upcoming “Drug Discovery Technology World Congress” in Boston this August.

Focusing on Protein-Protein Interaction Targets

Protein-protein interactions play a large role in cellular processes and represent therapeutically active points of intervention. However, these interaction sites tend to be wide and flat, making screening difficult.

Utilizing its proprietary NMR-based screening technology, Combinature Biopharm (www. combinature.com) has successfully identified small molecule ligands to a challenging protein-protein interaction target. This is the PDZ domain of the human gene af6, which modulates the activity of the pivotal oncogene ras.

“These PDZ domains are present in more than 400 human genes and represent a large class of underexplored protein-protein interaction targets binding to the C-termini of transmembrane receptors and ion channels through a barely druggable shallow grove,” says Markus Schade, Ph.D., vp, NMR drug discovery.

He adds that NMR screening provides structural resolution, which is important for the chemical optimization of fragment hits toward high affinity and selectivity.

Additional advantages of this technology include site-selective screening, acceleration of lead optimization by structure-aided connecting of building blocks, and generation of lead structures with desirable pharmacokinetic profiles. The company has its own NMR-fragment library of 20,000 synthetic compounds for screening.

Customer services currently include NMR-based screening using its in-house library or customer libraries, determination of binding sites and affinity of hits by NMR, and determination of 3-D protein-ligand complexes by NMR. Although Dr. Schade says NMR screening is a new technology, “the potential for applications are wide because there are thousands of protein-protein interactions and many enzyme targets requiring the design of highly selective inhibitors.”

Protein-Epitope Mimetics

Polyphor (www.polyphor.com) is also focusing on protein-protein interactions, but is zeroing in on the most relevant secondary-structure motifs occurring in proteinsbeta-hairpin loops. These are involved in approximately 90% of all protein-protein interactions.

The company has developed a technology process, in collaboration with Professor John A. Robinson at the University of Zurich, to make mimetics of beta-hairpin loops called Protein Epitope Mimetics (PEM).

“These are fully synthetic structural mimetics of proteinogenic beta-hairpin loops,” says Daniel Obrecht, Ph.D., CSO. “They are synthesized on an automated, parallel platform that allows for rapid iterative cycles for optimization of biological activity, and in vitro and in vivo ADMET properties.”

The molecules are approximately 12 kD, and can be assessed via NMR. Its 3-D information can be integrated into a small molecule design process.

The company presented information on two applications. One involves inhibitors of serine proteases, cathepsin G, elastase, and tryptase, which are all involved in chronic obstructive pulmonary disease and asthma. Most small molecule inhibitors lack good selectivity and have side effects. The company developed fully reversible and selective inhibitors of cathepsin G and elastase exhibiting good ADMET properties.

The second application is the discovery of antagonists of the chemokine receptor, CXCR4, a protein-ligand CPCR initially described as a co-receptor for T-tropic HIV, but recently shown to be an important receptor in angiogenesis, metastasis, and stem cell release.

“Antagonists of CXCR4 could be useful for the treatment of HIV, as well as for cancer, stem cell transplants, and inflammation,” states Dr. Obrecht. The company expects to bring these compounds into preclinical trials soon, and will also be looking at intracellular protein-protein interaction targets.

Generating Novel Proteases

As an alternative to therapeutic antibodies, Direvo Biotech (www.direvo.com) has developed two technology platforms. One is the directed evolution of natural proteins, which optimizes existing therapeutic proteins. The second, called NBE (New Biological Entities), creates new proteases with custom-made specificities and targets.

The directed evolution platform consists of an ultra high throughput screening process that can handle 100,0001,000,000 individual protein variants per day. Optimization includes efficacy, selectivity, substrate specificity, stability, and yields. Applications include in vitro assays (cell-based and pure chemical), assays for solvents, and assays for antibodies and enzymes.

NBE is relatively new, but has already been used to develop a new anti-inflammatory protease to inactivate TNF-alpha. “We can generate proteases with this platform which do not exist in nature with tailor-made specificities against nearly any protein target,” says Andre Koltermann, Ph.D., CEO. “This is comparable to the therapeutic antibody approach, but our method has several advantages.”

These advantages include: cleaving their target, causing irreversible inactivation; inactivating hundreds to thousands of target proteins with one NBE protease; cost-efficient production via bacteria expression systems that work in vitro; and a production time of only three to six months. Antibodies can only neutralize one target protein, require mammalian expression systems and immunogenicity, and take much longer than six months to create.

The NBE process begins with a library based on the human protease backbone, and then random loops in the protease gene are inserted into specific regions called SDRs (Selectivity Determining Regions). These are stretches of a few amino acids at certain regions of the protease that can determine and alter its specificity. Then the NBE library is screened for desired specificity, and if necessary, the protease is further optimized for process conditions.

Preclinical studies are currently under way for the company’s novel anti-TNF-alpha protease. “This is an efficient way to treat diseases like rheumatoid arthritis, Crohn’s disease, and psoriasis,” adds Dr. Koltermann. “Our future plans are to broaden our protein platform to generate more proteases for specific targets.”

Antibacterial Agents

Polymedix (www.polymedix. com) creates small molecule mimetics of host defense proteins, eliminating the potential of drug-resistant bacteria as seen with current antibiotics. Exclusively licensed from researchers at the University of Pennsylvania, this process involves several computational design tools.

“The first thing we did was to design a novel antibiotic,” says Nicholas Landekic, Ph.D., president and CEO. “Instead of targeting a biochemical molecule, we’re disrupting the bacterial cell membrane from the outside. It’s difficult to envision how resistance would occur.”

The tools include a special force-field, GOLDYN (Global Optimization of Long-Time Dynamics), which takes into consideration solvent effects (i.e., water) on molecular shape and interaction. “Molecular dynamics is a key tool if one is trying to create membrane-active compounds, which is what we’re doing with our antibiotic compounds,” says Dr. Landekic.

“With the COSMOS (Coarse Grain Molecular Dynamics Simulations) approach, we’re able to do molecular dynamic simulations of over several hundred microseconds, even using modest amounts of computer power. This is real-time for the interaction of many molecules with their membrane targets.”

PACE (Proteomic-Assisted Computational Engine) is a set of algorithms that build nonpeptide backbones from common organic building blocks. These are built in a variety of shapes to mimic the backbone of the active portion of a protein the company is trying to mimic. Later, functional groups are added to give it biological activity.

The company has tested 300 novel small molecule antibiotic compounds to date and approximately 65% are considered hits. Approximately 50 of the 300 compounds are potent, selective, non-toxic, and considered leads.

“By developing small molecules that mimic host defense proteins, we’ve achieved compounds that are broad-spectrum, work on a wide range of gram positive and gram negative strains, as well as fungus, and are fast actingbactericidal within seconds to minutes. Most current antibiotics take one to three days to work and only arrest bacterial growth,” states Dr. Landekic.

FivePrime Therapeutics (www.fiveprime.com) has developed a platform for the production and screening of all secreted proteins and receptors. The principles of high throughput screening of small molecule compounds are applied to the screening of entire sets of human secreted proteins on primary cells.

The company has the most comprehensive human cDNA collection, with more than 300,000 cDNA, of which more than 90% are full length, with an average of 12 clones per gene.

“Full-length cDNAs give a much better chance to make functional proteins rather than using ESTs, which most companies use,” says Ge Wu, Ph.D., director of assay development and screening. These are derived from a tissue bank containing over 2,400 samples from 135 different normal and disease tissues.

High throughput protein production allows for the production of more than 2,000 proteins per week for screening. “Then we put these 2,000 proteins in different assays that are medically relevant. We decide on a disease first, then decide what type of cells are a good target, so we match the cell to the screening. We use primary cells, not cell lines,” explains Dr. Wu.

This screening process is automated and has shown results to be reproducible. A database captures activity profiles of all proteins across all screens, which is a key element for lead selection.

The proprietary Espresso in vivo testing system generates information on secreted protein function in animals by mimicking intravenous injection of the protein without the need for protein production.

FivePrime is currently applying this technology to four disease areas: oncology, type 2 diabetes, immune disorders, and regenerative medicine. Dr. Wu says they currently have a “promising target called FPT025, which targets monocytes and T-cells.”

A new area is cardiology, where Dr. Wu says the company currently has two projects under way. One is screening for factors that protect the cardiomyocyte after ischemia, and the other is screening for factors that stimulate the regeneration of heart cells from stem cells.

Combining In Vitro Assays with In Silico ADME Prediction

Nimbus Biotechnology (www. nimbus-biotechnology.net) combined in vitro assays with in silico approaches to optimize drug profiling and enable early profiling of new chemical entities or whole compound libraries. There are two early ADME assaysone for lipophilicity/membrane affinity (characterizes intestinal absorption of the compounds) and one for serum/HAS binding (mostly responsible for drug distribution).

These high throughput TRANSIL bead-based assays are in ready-to-go plates, adaptable to all major screening platforms, and state of the art read-out (UV, HPLC, LC/MS) is supported. It is capable of measuring more than 1,000 compounds/day, and processing time is less than one minute per drug in the 384-well format, according to Nimbus.

Pk-Map (originally developed by Bayer Technology Services and exclusively distributed by Nimbus) is a software tool for assessing ADME properties of compounds in different stages of the drug R&D process via the TRANSIL assays and uses this as major input parameters (lipophilicity, protein binding, buffer solubility, molecular weight of compound).

“The prediction models are optimized by using close to nature’ assays like the TRANSIL approaches for the measurement of lipid-water partition coefficients and binding to HAS (human serum albumin),” explains Thorsten Hartmann, Ph.D., group leader, analytics. “This combination allows a fast and precise ranking of drug-like compounds as soon as possible in drug discovery.”

A user-friendly graphical interface allows selection of compounds with suited properties without looking through complex databases.

Dr. Hartmann says that the biggest concern of most companies is to provide the right information base for selection decisions during a drug discovery project. “Therefore researchers want to have a maximum amount of information about their compounds right from the beginning, which is provided through the combination of in vitro and in silico tools.”