Single molecule science – at the frontier of drug discovery


Stefan Geschwindner and Chris Phillips

When Anton van Leeuwenhoek was grinding lenses for the first microscopes, he cannot possibly have imagined that his ingenuity would one day lead to advances in drug discovery based not on examining slices of tissue or clumps of cells but on high resolution images of single molecules.


As early adopters of cryo-electron microscopy (Cryo-EM) and total internal reflection fluorescence (TIRF) microscopy, researchers at AstraZeneca are at the forefront of single molecule science – using it to define the structure and function of large complex proteins and better understand interactions between potential new medicines and target receptors.


Cryo-EM – revolutionising structural biology
Cryo-EM is changing the world of structural biology by enabling us to achieve high resolution images of the structures of large, complex proteins that are difficult or impossible to study using traditional X-ray crystallography. Through a Cryo-EM consortium, involving the Medical Research Council’s (MRC) Laboratory for Molecular Biology (LMB), the University of Cambridge, and four other pharmaceutical companies, our Discovery Science team is using the cutting-edge technique to study many target proteins for the first time and gain new mechanistic insights to help guide future drug design.

For Cryo-EM, highly homogenous protein solutions are frozen in liquid ethane capturing the protein in a single molecule layer of vitreous ice. We can then observe hundreds of thousands of individual molecules with the microscope and, thanks to a series of image analysis algorithms, derive a 3D reconstruction of the molecular structure of the target protein.

Recent advances in fast-read out cameras allow us to correct for moving particles and software developments for processing images mean that we are rapidly heading towards complete atomic resolution models of large, hard to handle proteins, and membrane bound proteins.  These atomic resolution models provides huge advantages for drug design. Not only does it help us discover the biological mechanisms of disease processes, it enables us to see how our novel molecules, our potential new medicines, bind to their targets at an atomic level. By finding out where each atom goes, we can improve the structures of subsequent compounds to optimise the way they interact with the targets in which we are most interested.


ATM – as it’s never been seen before
Using high resolution Cryo-EM, scientists at AstraZeneca and the LMB recently defined the structure of human ataxia telangiectasia mutated (ATM) protein in different functional states. This is a ‘world first’ for this important player in the cell’s DNA damage response (DDR) system and is a prime therapeutic target for cancer therapies.

ATM is a phosphatidylinositol 3-kinase–related protein kinase (PIKK) which is activated by double-strand DNA breaks in conjunction with the damage-sensing Mre11, Rad50, and Nbs1 (MRN) protein complex. ATM mutations are associated with the neurodegenerative disease, ataxia telangiectasia, which can also cause genome instability and cancer.

Due to its large size – some 350 kDa – obtaining any high-resolution structure for ATM had proved impossible with other techniques. Using Cryo-EM, we achieved a resolution of up to 4.4Å for ATM, and this enabled us to show that the ATM dimer appears to exist in a dynamic equilibrium between open and closed states. In the closed state, a regulatory domain blocks the peptide substrate binding site, suggesting that, in this conformation, the enzyme is inactive. However, in its open state, the site is accessible and may therefore be more active.

Our findings, recently published in Science Advances, are an important step towards a better understanding of the science behind the regulation of ATM and its role in repairing double-strand DNA breaks in cancer cells. Using inhibitors of ATM, we will explore our ability to alter the equilibrium between the open and closed dimer populations of ATM, opening the way to novel treatments for cancers where ATM plays a role.

In addition, we are focusing our Cryo-EM research on other large proteins involved in DDR mechanisms, such as ataxia telangiectasia and Rad3 related (ATR) and DNA-PK which, like ATM, are part of the PIKK family of large complex proteins.


TIRF – watching molecules as they sparkle
TIRF is an innovative technique for imaging fluorescently tagged molecules in real time as they bind to and separate from receptors or other targets within our drug discovery assay systems, and it looks set to play an important role in second line lead characterisation. It may prove particularly useful in investigating activity of compounds on membrane bound protein targets, such as G-protein coupled receptors (GPCR) and ion channels, which are currently too unstable for traditional assay platforms or require laborious protein engineering.

While conventional fluorescence microscopy allows us to measure this type of activity by recording signals from 10-30µl volumes in the wells of our assay plates, TIRF is much more precise. It enables us to watch individual molecules less than 100nm above the surface of each well.

The other key differentiator is dynamics. With traditional techniques, we mix reagents in our wells and use fluorescence to watch ligands bind to receptors during a single event. We then have to repeat the process many times to confirm our results.

With TIRF, our system starts at equilibrium so, as long as the assay remains stable, we can watch the reaction and take measurements whenever we want – now, in an hour or tomorrow – with the same result. Not only that, we can easily see how quickly a ligand binds to our target and how long it stays on the target. We see thousands of binding and releasing events and, like a sky full of stars, we watch the lights going on and off.


Dynamic clues to drug discovery
TIRF allows differentiation of populations of receptors or ligands according to their dynamics. One population may only bind for a second while another binds for far longer, reflecting variations in either the receptor or the ligand. Such differences can have important consequences for drug discovery as one receptor sub population may be more reactive to a new compound than another.

This isn’t to say that we get instant answers to our questions. We generally need 2000-3000 observations of events to have the statistical power behind our data to start drawing conclusions. For interactions where the affinity between receptor and ligand is low and binding and release events occur as flashes of light, we may be able to collect hundreds of observations in just a few milliseconds. However, when there is high affinity and ligands remain bound to their receptors for longer periods, it can take longer to collect sufficient data.

Once we have seen how long it takes for the single molecule we’re interested in to bind and release, we can calculate how long it will take to get a complete set of observations. It may be a day, but that doesn’t matter because our system will still be working in equilibrium. Any number of technical parameters can affect the sensitivity of other fluorescence microscopy techniques, but the beauty of TIRF is that the only factor affecting sensitivity is observation time, ie. how long it takes for a binding/releasing event to occur.


Potential for antibody-free systems
TIRF has the potential for antibody-independent detection of diagnostic biomarkers. Proof-of-concept experiments have already shown that it is possible detect biomarker binding to receptors and ligands or DNA-DNA interactions without the need for an antibody as intermediary. Using ELISA, we only know that an antibody is recognising part of a biomarker and it doesn’t tell us whether that fragment is part of a larger biomarker which has already degraded or is itself active. With TIRF, we can get a functional read-out of a biomarker with which we can assess its activity state, and we hope this will lead to a new dimension of understanding of how we use biomarkers to monitor disease progression.

TIRF’s potential for imaging low affinity interactions comes into its own in antibody-free systems. Antibody-based assays work best for high affinity interactions and this can be time consuming. Once again, in an antibody-free system, TIRF dramatically shortens our measurement time.


Future directions for Cryo-EM and TIRF
Neither Cryo-EM nor TIRF is likely to replace existing platforms used in drug discovery – or indeed compete with each other. While Cryo-EM has advantages for gaining new insights into large complex proteins, X-ray crystallography has a well established role in modelling of smaller molecules.

The full potential of TIRF has yet to be confirmed but it could well prove most useful after high throughput screening in second line hit characterisation. Essential to its successful wider use within the industry and academia will be the quality of the image analysis, and that is dependent on development of state-of-the-art software to analyse the vast quantities of data that are generated and, ultimately, to guide automated systems to change and progress assays in response to results.

In just a few years, we have seen rapid and very exciting advances in single molecule science but, for many of us, the best is yet to come!