Drugbaron Blog

March 6, 2018 no comments

Cult of Crystallography

Biological macromolecules, such as proteins and DNA, are very small and very complex. It is something of a technological miracle, therefore, that we can visualise their three-dimensional structures at atomic resolution. The first, and still by far the dominant, method of solving these structures is crystallography.

The principal is simple: by packing many millions of molecules into a repeating crystal lattice, the diffraction pattern of X-rays passing through the crystal captures information about the structure of the individual molecules. In effect, the structural information from many identical molecules amplifies the signal to the point where it can be read and decoded.

This ability to see the structures of individual protein molecules was a huge breakthrough in biology. It may not be as famous as the DNA double helix, but the first protein structure, earned a Nobel prize for Max Perutz in 1962, the same year that Watson, Crick and Wilkins were honoured for solving the DNA structure by the same crystallographic technique.

In the intervening six decades, many thousands of protein structures have been solved in this way, and delivered mechanistic insight into the many functions of proteins, as structural components, molecular motors, enzymes, synthetic factories and signalling networks. They have underpinned the development of numerous drugs, designed to bind in highly specific ways to particular proteins. Today, variations on a theme even allow us to visualise structures of membrane-bound proteins that traditionally eluded crystallisation.

So great has been the positive impact of having a window on these ultra-microscopic structures that its hard to imagine that the ubiquitous application of crystallography has had a unintended negative consequence for our understanding of biology.

Structures obtained using these methods that average across millions of molecules are so embedded in our thinking that we accept them as representative of the protein in vivo, almost without thinking, and certainly without constantly bearing in mind the single biggest assumption of crystallography: that all the molecules are the same.

Together with the cult of DNA centricity, this cult of crystallography has led to a very distorted view of what we think of as a protein. When we name a protein, such as apolipoprotein E (one of DrugBaron’s favorite proteins), the implicit assumption is that we refer to a homogeneous population of molecules, all with the same primary sequence (that is, the same amino acid building blocks in the same order, exactly as encoded by the DNA sequence of the apoE gene) and all with the same three-dimensional structure and conformation (the one revealed to us by crystallography).

But work at Methuselah Health, a Medicxi portfolio company founded in 2014, to exploit the insights of Professor Miroslav Radman into the role of protein damage in ageing and age-related diseases such as neurodegeneration and autoimmunity, has demonstrated just how misleading this pervasive view really is.

In order to investigate protein damage in a non-hypothesis driven way (by that we mean without assuming ahead of time in what ways certain proteins might be damaged), Dr David Mosedale, the Chief Technology Officer at Methuselah Health, had to refine another ground-breaking analytical technique: liquid chromatography coupled to mass spectrometry (or, more simply, LC-MS-MS). LC-MS-MS allows us to look at the chemical composition of individual molecules, and with refinement the Methuselah Health team have been able to quantify the different types of damage occurring in complex mixtures of proteins.

And the results are staggering. It was immediately obvious that precisely none of the proteins in a real biological sample fit the assumption of crystallography. What we might classically have called “apolipoprotein E” in a blood sample is breathtakingly diverse. There are literally thousands, perhaps millions, of subtle variations between the individual molecules of apoE, many of which are so prevalent in the population that what might once have been thought of as apoE (that is a perfect expression of the DNA-encoded gene sequence) is very much in the minority, if indeed it can be found at all.

“We have called this diversity within the population of molecules of a single protein the ‘quantum zoo’” says David Mosedale. “There is a clear parallel between this new level of detail in studying proteins and the shift from classical Newtonian physics to quantum mechanics. At some level, assuming a protein to exist in only one or two states provides an acceptable approximation, and allows real biological insight. But it is becoming clear that there are also processes – critical processes involved in disease – that can only be understood at this new quantum resolution, viewing each protein as a population of subtle variants.”



There are essentially two sources of variants: errors in converting the gene sequence into a protein and post-translational modifications. The Methuselah Health platform readily quantifies both.

“We tend to forget that ribosomes, like every other biological process, are imperfect. Every now and then, a wrong amino acid gets permanently incorporated into the growing polypeptide chain. This might be because the wrong tRNA engages with a codon, or because the correct tRNA has been loaded with the wrong amino acid. But either way, a copy of the protein gets made that doesn’t match the gene sequence” explains David Mosedale.

“Then, even if a perfect copy is made, some amino acid side-chains are subject to chemical changes” he continues. “Some of these are deliberate, controlled by enzymes, such as glycosylation, phosphorylation and acetylation, but many others are accidental, such as oxidation or glycation. Around half of all amino acids are susceptible to such modifcations, which means the universe of combinations of changes to even a small protein is almost infinite.”

When Methuselah Health turns its analytic tools on a biological sample, however, it sees a biased population of these “errors”. The equilibrium population of variants that make up the ‘quantum zoo’ depends on both the frequency of producing each variant and on its lifetime. The vast majority of changes destabilise the protein, and so are cleared quickly and even if they are made reasonably frequently are hardly represented in the equilibrium population.

But just occasionally, a variant (whether an error in translation or a post-translational modification) has a longer half-life than the gene-encoded version of the protein. These variants will accumulate over time, and even if they are made only rarely, given enough time they can constitute a significant fraction of the ‘quantum zoo’. These are variants that Methuselah Health identify.



Because these variants, invisible to classical biology, blinded by a cult of crystallography, can cause diseases. Imagine a variant of apoE that does not bind to the LDL receptor – lets call it apoE*. ApoE* will selectively accumulate in tissue, because it will be cleared more slowly than the ‘wild-type’ apoE. Accumulation of LDL in the blood vessel wall in atherosclerosis, therefore, may depend on the propensity to form apoE* – a process that has nothing to do with genetics, and is invisible to analytical tools that assume every copy of apoE is the same.

Professor Radman believes these quantum processes underlie the majority of degenerative diseases of old age. Accumulation of rare variants that have increased stability (perhaps because they have been mis-trafficked into a compartment where the normal degradation pathway for that protein is not operative) gradually destroys the structural, and hence functional, integrity of many tissues. In the blood vessel wall, it leads to atherosclerosis. In the brain, it leads to neurodegeneration.

This quantum complexity of protein structure may even explain, at least in part, why some proteins are so challenging to crystallise. Forming a regular lattice requires that all the protein molecules in the sample have very nearly identical structures – if the sample actually contains a diverse ‘quantum zoo’ of variants, some with quite different structures, it may be difficult to form a crystal at all. And even when one does form, it will contain only the molecules that conform to the lattice – structurally-distinct variants will be expelled into the supernatant. The very act of crystallisation simplifies the true complexity.

Unsurprisingly, therefore, techniques like crystallography have shaped our view of proteins as homogeneous populations of identical structures. As Methuselah Health are now showing us, nothing could be further from the truth. And hidden in this new quantum complexity may lie the answer to why we age, why we develop diseases like Alzheimer’s disease – and why genetics can miss all of this important biology.

Yearly Archive