Monday, 29 June 2009

Fragment-Based Screening and Drug Design

The last Monday seminar of the Summer term at Birkbeck was given by Dr Rob van Montfort of the Institute of Cancer Research. Van Montfort, a former postdoc in the School of Crystallography here, spent six years in industry, at the biotech company Astex Therapeutics before joining the ICR two years ago. There, he is developing high throughput, structure-based drug screening techniques and using them to design novel compounds as candidate anti-cancer drugs.

Van Montfort first described the problem that techniques such as his have been designed to solve: that of attrition in drug discovery. Not only does it often take well over 10 years to move a potential drug compound "from the bench to the bedside", but the attrition rate is immense and not sustainable. A typical drug discovery programme will involve the testing of millions of compounds, and result in, say, half a dozen candidates for Phase I clinical trials and, if all goes well, a single registered drug. And even if a compound reaches the clinic, it may well not recoup the millions of dollars that have been spent on its development.

Drug companies and adacemic groups have been turning to novel technologies to help address this problem and reduce the time and cost of drug development. The technique de Montfort described, fragment-based screening, is one of these. It was originally developed by Wim Hol's group at the University of Groningen, the Netherlands, and developed for commercial use by companies including Astex and Plexxikon.

Historically, X-ray crystallography has been used mainly in the mid stages of drug development, in modify promising lead compounds into drug candidates. Now - thanks largely to high-throughput crystallography or structural genomics programmes - X-ray technology has improved to the point where it can be used at a much earlier stage. In fragment-based screening, groups of small molecules taken from a "fragment library" are soaked into protein crystals and the resulting structures examined by X-ray crystallography to see which fragments have bound to which parts of the protein's ligand-binding sites. These fragment hits generally bind very weakly but may be "joined together", if they bind into different pockets within the binding site, and modified further to generate tight-binding inhibitors and, eventually, candidate drugs.

De Montfort went on to describe a few published case studies of Astex' protein drug targets, including the P38a MAP kinase. This kinase is involved in cellular responses to stress, and its inhibitors may be therapeutically useful in a variety of inflammatory and auto-immune diseases. Fragment-based screening identified a lipophilic fragment that bound into the selectivity pocket of this kinase (Hartshorn et al. (2005), J. Med. Chem. 48, 403-413; PDB 1W7H) and modified it to produce a larger, tight-binding inhibitor (Gill et al. (2005), J. Med. Chem. 48, 414-426; PDB 1W82 & 7 others).

Thrombin, a serine protease involved in blood coagulation, is an important therapeutic target for stroke and deep vein thrombosis (DVT). Van Montfort was involved in the development of thrombin inhibitors, a process that was particularly hard because the thrombin active site is charged, and the charged compounds that would be expected to bind there are unlikely to work as oral drugs. Howard et al. (2006) published a fragment-based screen against thrombin using a library of uncharged compounds, finding small fragments that bound to one pocket within the substrate-binding site and larger ones that bound to a different one (J. Med. Chem. 49, 1346-55; PDB 1WBG). Combining the chemistry of these fragments into a single, larger molecule produced a potent series of uncharged, non-peptide inhibitors with structures that had not previously been seen in protease inhibitors.

Kinases are signalling proteins that control many biochemical and physiological processes, including the cell cycle, and cell-cycle kinases are very important as targets for anti-cancer drugs. Both Astex and the ICR have extensive programmes developing kinase inhibitors, and thehe two organisations have collaborated on the discovery of compounds that inhibit protein kinase B (see e.g. Saxty et al. (2007), J. Med. Chem. 50, 2293-6; PDB 2UW3).

Van Montfort has now set up a fragment-based screening lab at the ICR and developed a library of over 1800 fragments. His group is investigating potential protein targets for drugs against cancer, including kinases.

All the journal references in this blog post are in the Journal of Medicinal Chemistry, which should be available via the Birkbeck e-library.

Thursday, 18 June 2009

Analytical Ultracentrifugation: Structures of Unstructured Proteins

Dr David Scott of the National Centre for Macromolecular Hydrodynamics, based at the University of Nottingham, has been visiting Birkbeck for a few days to give a course. He also gave a seminar in which he explained one of the biophysical techniques used in the Centre, analytical centrifugation, and how it is used to help determine something of the "structure" inherent in unstructured, or partially structured, proteins.

The analytical ultracentrifuge was invented by Theodor Svedberg in 1923; three years later, he won the Nobel Prize in Chemistry for research using it. It is simply a centrifuge that spins very fast (from about 1,000 - 60,000 rpm); the sample being spun is monitored optically over a period of time. The normal settling of particles in solution under gravity (sedimentation) is speeded up by spinning in a centrifuge, which essentially replaces the gravitational force by a centrifugal force. The speed of sedimentation depends on the masses and shapes of the particles involved. The maths is far too complex to be described here.

One of the first uses of ultracentrifugation was in the determination of molecular mass. Sedimentation times are measured in Svedberg units (S); 1S is exactly equivalent to 10-13 seconds. These times are related rather inexactly to molecular mass and often used to characterise large proteins and protein complexes; you have come across these in the PPS course in our discussion of ribosomal subunits. The small subunit of the Thermus thermophilus ribosome, illustrated there, is described as "30S".

Ultracentrifugation is now used routinely to determine whether a sample is homogeneous; if it is, all particles will have the same mass and shape, and therefore the same sedimentation time. A plot of sedimentation velocity for a sample can show whether the solution is homogenous or heterogenous, and whether protein is aggregated (in which case, aggregates will consist of different numbers of molecules and have different masses). Sedimentation equilibration experiments, which investigate the final steady state where sedimentation is balanced by diffusion, can be used to determine ligand binding and chemical reactions.

Scott has used ultracentrifugation and other biophysical techniques to investigate the structures of unstructured regions of a bacterial DNA-binding protein, KorB. This protein has a DNA-binding domain and a dimerisation domain each with a known structure, and it is known to interact with RNA polymerase as well as with DNA. Other parts of the protein, however, are only known to be "intrinsically unstructured". A combination of ultracentrifugation with other techniques useful for studying unstructured proteins, including circular dichroism and small angle X-ray scattering, were used to investigate the range of structures adopted by these unstructured regions. Results so far indicate that KorB forms a range of relatively compact structures when isolated in solution but that the unstructured regions extend when it binds to a related protein, KorA.

Analytical ultracentrifugation, CD spectroscopy and other biophysical techniques will be described in much more detail in the second year MSc module, Techniques in Structural Molecular Biology (TSMB).

Targeting Mycolic Acid Biosynthesis - Towards New TB Drugs

The Department of Biology and Chemistry at Birkbeck recently hosted as a seminar speaker Dr Geoff Coxon, from the University of Strathclyde, Glasgow. Dr Coxon described a series of compounds that his group had synthesised, some of which are promising candidate drugs against tuberculosis. These are inhibitors of the enzyme beta-ketoacyl-ACP synthase (otherwise known as FabH) with a novel chemical scaffold.

Mycobacterium tuberculosis now infects over a third of the world's population, and tuberculosis kills two million people a year. The threat of tuberculosis is growing largely because of two reasons: a synergy between HIV and TB infections, and the growth of drug resistant strains of the bacterium. Multi-drug resistant (MDR) TB is defined as TB that is resistant to the two most commonly used anti-TB drugs, and extensively drug resistant (XDR) TB as TB that is resistant to four drugs including a fluoroquinolone. The latter is particularly hard to treat. More information on this is available from the TB Alliance.

In order to combat drug resistance it is essential to keep introducing new drugs, and preferably drugs with novel targets and mechanisms. Ideally, a drug should be active against active, replicating TB and persistent TB, which can resist treatment by remaining dormant in macrophages. Much work in anti-TB drug development has focused on enzymes involved in synthesising the very complex cell wall of the bacteria.

The genome sequence of M. tuberculosis (Camus et al., 2002, and material in PPS section 6) revealed a number of enzymes involved in cell wall synthesis. Coxon's group is focusing on the synthesis of one cell wall component, mycolic acids. These are long chain, 2-alkyl 3-hydroxyl fatty acids with between 60 and 80 atoms in their hydrocarbon chains. Two enzyme systems are involved in their synthesis: FAS-1 synthesises the main chain and FAS-2 adds the alpha branch. FAS-2 is found only in plants and mycobacteria.

The FAS-2 system includes a large number of enzymes, all involved in the complex, cyclic process of synthesising a long lipid chain. Inhibiting any of these enzymes will prevent the synthesis of the mature long chain mycolic acids. Coxon's group has been targeting one key enzyme in this process, known as FabH, which connects the FAS-1 and FAS-2 systems. The reaction it catalyses is an extension of the lipid chain by two carbon units.

This enzyme is a homodimer with a tunnel-shaped active site; the chain extension takes place after the substrate has moved into the tunnel. Its structure was first solved in 2001 by Scarsdale and co-workers (PDB code 1HZP). Coxon and his co-workers started their search for a specific inhibitor of this enzyme from the natural antibiotic, thiolactomycin (TLM). This, however, is a relatively weak inhibitor and a chiral compound that is extremely challenging to synthesise. He used fragment-based chemical libraries to develop a range of likely inhibitors with similar skeletons but that would be simpler to synthesise.

Some of the first compounds investigated were better inhibitors of the enzyme but not active against the whole FAS-2 system or M. tuberculosis itself. It appeared that another enzyme, known as KasA, may sometimes take the place of FabH if that is inhibited. They have now produced a series of compounds based on a 2-aminothiazole-4-carboxylate scaffold that includes some very active ones. However, the compounds in this series that are the most potent inhibitors of the enzyme are not the most active against the intact bacterium, and vice versa - there are complex interactions going on that are not yet entirely understood. It is likely, even, that FabH is not their most important protein target.

A good medicine against tuberculosis must be active against both resistant and dormant forms of the bacterium, with few side effects and few interactions with other drugs, and preferably orally available, and it must be cheap, and therefore easy, to synthesise. These 2-aminothiazole-4-carboxylates are easily synthesised and some are effective against M. tuberculosis in vivo, but much more work is needed to determine whether they will fulfil the other criteria.

See Al-Balas, Q. et al. (2009) PLoS ONE 4(5) (open access) for more information about this work.

Dr Coxon also works in TB Drug Discovery UK, an alliance of scientists involved in developing new treatments for tuberculosis.

Monday, 9 March 2009

Structure of the trypanosome microtubule cytoskeleton

Twice a term, the School of Crystallography hosts a seminar for the whole of the Institute of Structural and Molecular Biology, which consists of research departments in related disciplines from both Birkbeck and neighbouring University College London. Tom Blundell's (see previous post) was an ISMB seminar; today, we heard one from Professor Keith Gull of the Sir William Gunn School of Pathology, University of Oxford.

Much of Keith Gull's work concerns the single-celled parasite Trypanosoma brucei, which is endemic in large parts of Africa and causes African trypanosomiasis, otherwise known as sleeping sickness. Diseases like this one attract relatively little research funding even though they are important causes of morbidity and mortality in many poor countries; they are classed as neglected tropical diseases. Keith and his colleagues in Oxford are studying many aspects of the molecular biology and genetics of this parasite.

The trypanosome cell surface is covered by a dense coat made up of very many copies of a single protein, called variable surface glycoprotein (VSG). A glycoprotein is a protein with carbohydrate (sugar) groups attached to one or more amino acid sidechains; the word "variable" is used because there are about 1000 variants of this protein. Each of these variants is encoded by a different gene, so the VSG genes account for about 10% of the trypanosome's genome. However, only one of these is expressed at any time, so the trypanosome coat (almost) always consists of multiple copies of a single protein. (There will be times when two proteins are present because the expressed variant is in the process of changing.) Although the sequences of VSG proteins differ considerably, their structures are very similar: they are anchored to the cell membrane and the large extracellular domain consists of an antiparallel coiled coil of alpha helices (see PDB file 2vsg).

Most of Keith's talk, however, concerned the structure of the trypanosome cytoskeleton, which consists of microtubules. Very basically, these are composed of polymers of a protein called tubulin and they are involved in maintaining the structure of components of many types of cells. Keith and his group have used the technique of electron tomography - a type of transmission electron microscopy - to obtain exquisite pictures - if at much lower than atomic resolution - of the structure of this cytoskeleton and begin to understand its function and role in trypanosomal cell division. Structures of these multi-protein complexes are not - yet - available in the PDB.

Trypanosomes have flagella - whip-like structures attached to the cells that can propel them through the host bloodstream. Flagella are found in many cell types, and if they have receptors bound they can also be used to sense cells' surroundings. The trypanosome flagellum is attached to the cell body via a filament and this point of attachment is within a pocket that, interestingly, is the end point for all vesicular traffic within the cell. This means that this pocket needs to contain a large number of proteins vital to the parasite's survival, including its transferrin receptor and haemoglobin receptor. Imaging the structure of the trypanosome cytoskeleton, particularly in the pocket where the flagellum attaches, has also given insight into the rather unusual process through which this cell conducts mitosis (cell division).

Friday, 13 February 2009

Genomes, Structural Biology and Drug Design

Professor Sir Tom Blundell, head of the School of Crystallography at Birkbeck from 1977 to 1995, returned last Monday to give an extremely well attended seminar - a tour de force of the relationship between structural biology and drug design Tom's distinguished career has also included a time as the head of the research council that funds non-medical biological research in the UK, BBSRC; he moved from Birkbeck to become head of Biological Sciences at Cambridge; and he is a director of a biotech company, Astex Therapeutics, which he founded in 1999.

Tom started his talk with a brief history of structural biology and its role in drug discovery. His personal involvement in the discipline goes back to the 1960s, when he, as a Ph.D. student, visited companies such as Eli Lilly, which manufactured insulin, with his supervisor, Dorothy Hodgkin - who solved the insulin structure, but won her Nobel prize (Chemistry, 1964) for structures of penicillin and vitamin B12.

The intervening decades have seen trends in drug discovery come and go. In the 1990s, it seemed that increases in the speed of synthesising and screening large numbers of small molecules against drug targets had made the more targeted approach of the academic structural biologists redundant. However, even the millions of compounds that can now be screened represent the tiniest fraction of "chemical space": the number of potential molecules of a size to bind to a drug target is larger than the estimated number of atoms in the universe.

And now, in the so-called age of the genome, structural biology has become an integral part of drug discovery, involved in all steps: target identification and validation, screening, and lead compound identification and optimisation. The organisation of information about sequences and structures in databases - some of which were mentioned in the PPS Bioinformatics section - began when Tom was at Birkbeck, and was spearheaded by his co-workers and collaborators, particularly Janet Thornton (now head of the EBI). The databases set up and curated by members of his Cambridge group - too many to describe properly here - are available from this page.

Some particularly useful insights arise from the relationship between the single changes in nucleic acid sequence (known as Single Nucleotide Polymorphisms, or SNPs) that are collected into databases and the structural biology of drug targets. Sometimes, with simple Mendelian diseases, one such change is sufficient to cause disease; more complex diseases arise (anything from hypertension to breast cancer to bipolar disorder) arise from interactions between many such changes that increase the chance of disease. Mapping changes in protein coding brought about by an SNP to a protein's structure can give insight into disease-causing changes in protein mechanism and lead to the identification of novel drug targets. See Burke et al. (2007), BMC Bioinformatics 8, 301 (this is an open access journal with full text available free of charge).

The drug discovery programme at Astex Therapeutics is based on experimental structural biology, on a technique known as fragment screening. In this, small chemical fragments that bind to drug targets are identified by fast X-ray crystallography. Knowing both the structures of these small compounds - which are too weak as binders to be identified by chemical means - and where they sit in a drug target's binding site enables chemists to build them out to form larger tight-binding molecules that fit into the whole site. The company's pipeline focuses on kinase inhibitors as anti-cancer drugs, and some of its lead compounds have reached early clinical trials.

Tuesday, 20 January 2009

Designing Magic Bullets

This week's Monday seminar at Birkbeck was given by Jose Saldanha, who works in Willie Taylor's group at the National Institute of Medical Research at Mill Hill, London. Both Jose and Willie are Birkbeck Ph.D. graduates. Jose described his work in using molecular modelling and bioinformatics to design specific antibodies for therapeutic applications.

These antibodies are the "magic bullets" of the title of this blog post. Interestingly, the term was invented by Paul Ehrlich, who won of one of the first Nobel prizes for Medicine - in 1908 - for his work on immunity. The magic bullet, he suggested, would be a molecule that could be targeted directly to a part of the body needing drug treatment, and which a drug molecule could be attached to.

Antibodies (or immunoglobulins) are secreted into the bloodstream in response to the presence of "foreign" molecules (e.g. from bacteria or viruses) and target those molecules in order to destroy the "invader". They do this by binding very specifically to the targeted molecule. Therefore, molecules with essentially the same structure must be able to bind an almost infinite variety of targets.

The mechanism that has evolved for this is a very elegant one. The commonest type of immunoglobulin, Immunoglobulin G (IgG) is a Y shaped molecule made up of four chains, two heavy and two light. Heavy chains contain four copies of the same domain fold (a type of beta barrel) and light chains contain two. The binding site is at the "top" of the "arms" of the Y; the domain of each chain that is closest to the binding site is much more variable in sequence than the others, and is termed the variable domain (the others are constant domains). Each variable domain has three regions that are particularly variable in sequence, termed complementarity determining regions or CDRs, and it is these that bind the target molecule (the antigen). There is a more about immunoglobulin structure on this page from the Arizona Biology Project, and it will be covered in depth in section 11 of PPS.

The only antibody that can be successful as a "magic bullet" is a completely homogenous sample, where every molecule has the same sequence and structure and binds the same antigen. Such identical antibodies, or monoclonal antibodies, can be produced by fusing myeloma (cancer) cells with spleen cells produced by an infected mouse, and selecting and cloning those resulting cells that secrete the required antibody. However, these are mouse antibodies that do not work in humans; they can also cause an immune response because they themselves are seen as "foreign".

In the 1980s, Greg Winter at NIMR developed a process called "CDR grafting" to overcome this, in which the CDR regions from the mouse antibodies are grafted on to a human antibody framework. An alternative technique involves the fusion of a few residues from the mouse antibody with a human one to form a "humanised" antibody. Jose's work at NIMR involves sequence and structure analysis to work out exactly which mouse antibody residues to fuse with which human antibodies for different indications. A "successful" antibody must express well, bind specifically to its antigen and cause no immune or other adverse reaction in human patients. Since starting the project, he has designed more than 30 different humanised antibodies, some of which are in clinical trials.

There are many structures of immunoglobulins and immunoglobulin fragments in the PDB. The most commonly crystallised fragment is the four domains that make up one of the arms of the Y, which is known as Fab (standing for "Fragment, antibody binding"). Look at this one (1BM3), which binds a peptide antigen. And remember - we will be coming back to this topic later in the course.