Bringing molecular models to life
Anatomical models have been used for centuries in medical teaching. They play a vital role in helping students understand the relative size and position of bodily structures. For much of that history there was little concept of what physically existed below the range of the naked eye. And where that did emerge, flattened on microscopy slides, the 3D nature of cells and their components was difficult to appreciate. Technology marched on and eventually we could ‘see’ proteins and molecules, things smaller than the wavelengths of light itself. These objects have positions, shapes and volumes in 3D space but our ingrained expectations of what these things ‘look like’ break down more and more, the smaller we go. Various abstracted representations attempt to convey relevant details – the ball and stick molecular models found in school chemistry labs are a familiar example. Ribbon models of proteins are similarly ubiquitous to biochemistry students. Valuable as these approaches undoubtedly are, the simple sense of spatial presence is gradually eroded.
In biology and medicine, conception of the 3D arrangement of molecular structures is crucial to a full understanding of cellular signalling pathways, their modulation by drugs and clinical applications. Many students find it hard to appreciate the relative scale, shape and orientation of molecular structure, particularly how to integrate this with other concepts such as adverse drug reactions. Traditionally, 2D printed figures have been used to explain molecular interactions but these are static, highly stylised and do not adequately capture all the spatial rearrangement involved in biological signalling. An example of pushing this to its metaphorical limits is shown here from my own teaching, in this representation of a receptor as a hand catching a ligand and being activated. A receptor protein does, in a certain sense, catch a ligand and this does frequently lead to a change in receptor structure which enables a functional signal to be activated. But this representation, albeit intentionally, says nothing about the spatial relationship between receptor, ligand, membrane and effector.
A more familiar representation from science teaching is the whiteboard sketch shown below which since its arrival in 1975 has pretty much completely ousted chalk from the classroom. This type of diagram is found in every biochemistry or pharmacology textbook: receptors become rounded rectangles interacting with ligand circles via a flurry of arrows. The figure below shows a neuronal synapse – typically two mushrooms headbutting – and the release of 5-HT (serotonin, ornage dots) binding to its receptor on the postsynpatic cell. Also shown as a blue, rounded rectangle, is the protein of interest the serotonin tranporter (SERT) the target of many antidepressant medications such as fluoxetine (Prozac). As students explore molecules in more depth they meet the ribbons and cylinders, alpha helices and beta sheets of protein structure representation. These convey more specialist information necessary for experts communicating, understanding and exploring macromolecular dynamics yet concomitantly become more impenetrable to the non-specialist. Software such as PyMOL allows the ready translation of protein structure data into less abstracted space-filling models which can be manipulated and rendered in three dimensions.
Modern computer graphics allow sophisticated three-, and even four-, dimensional rendering of molecular dynamics but typically come with specialised interfaces, vertiginous learning curves and limited scope for manipulation by learners.
This is the story of how I’ve tried to address this, starting with custom 3D-printed molecular models and moving to digital versions and augmented reality (AR) tools. I’m no technical expert in these areas, at the end I provide links to resources I’ve found helpful. I have two main aims in writing this. One is to give insight as an artist-scientist hybrid, where maybe many people would consider themselves one or the other. The second is to show that these approaches are accessible. No-one thinks twice about embedding clip art or images into their presentations and posters. The tools are there for everyone to do this with 3D objects. And yes, even in posters, read on to find out how…
The value of custom 3D-printed models as medical education tools has been published for insight into tissue structures and surgical interventions. At a molecular level, atomic models enzymes and their substrates, receptors and their ligands, are used in teaching pharmacy and chemistry. The mid-point of protein complexes, relevant to cellular biochemistry and physiology remains relatively unexplored. 3D printing is a rapidly advancing technology, offering a cost-effective and bespoke means to develop custom teaching resources. For this aspect, I am immensely grateful for the expertise of Dr Darren Gowers from Molecular Models. I had ideas of what I wanted my protein models to look like but he was the one who turned them into high-quality, visually appealing and durable (they’re to be used in a classroom!) resources. In addition, he’s been a wealth of inspiration in using cut-aways and embedded magnets to reveal the inner workings of proteins and their interactions with other molecules.
Using colour to show function
3D printed models are more than a physical replica of a molecular structure. Careful choice of design can illustrate function and even kinetics. You need to think through exactly what you are trying to convey, what do your learners or audience actually need to know. Most of my teaching is for medical students and for the majority of them and in the majority of contexts, the technical detail is a fraction of what would be taught in a pure undergraduate science degree (I realise views on this differ widely and quite vehemently but this is completely irrelevan to the design point here). It is tempting to include every protein domain, colour every key residue, have magnets attaching all sorts of post-translational modifications but if this detracts from the important learning points they will be counter-productive. And I say this from the point of view of a molecular biologist who wants to include every tiny molecular detail! This is where I have to step back and think like an educator, even like an artist.
Haemoglobin – the protein that carries oxygen round the body, packed into red blood cells – is a good example of all these considerations. Haemoglobin occurs in two major states, either bound to oxygen or not. How it switches between the states is critical to how it delivers oxygen efficiently to tissues and under different conditions such as exercise, altitude or inflammation. In each state the protein adopts a specific 3D conformation, understanding this is essential to appreciate physiological function and pathology. Adequate oxygen is associated with the bright red colour of blood, the opposite results in the bluish fingertips and extremities seen in oxygen starvation. Simply using redder shades for the oxygenated structure and bluer ones for the deoxygenated one immediately help students to associate the model with the relevant state without even to see if a tiny oxygen molecule is bound. Haemoglobin also binds to carbon monoxide, this toxic interaction disrupts oxygen binding and results in a ruddier complexion in cases of severe poisoning, hence the third model is coloured with more saturated reds.
Haemoglobin protein is made of four polypeptide chains of amino acids, called globins, two alpha-globins and two beta-globins. The correct pairing of globins is essential for normal haemoglobin function. Each globin has a haem chemical group embedded in it, these are what actually bind the oxygen molecules. In these models, we coloured the alpha and beta globins different shades so in addition to the overall colouration described above, they can be readily distinguished. The haem groups are 3D printed as separate units and contain mini magnets so they can be placed in and out of the globin. Many factors can affect the assembly of these globin and haem domains, resulting a range of blood disorders, notably anaemia.
Haemoglobin is one of the most intensively studied proteins in the human body. Hundreds of genetic variants have been identified, some have no detectable effect on function, many are life-threatening. They could all be coloured differently on the printed model but how to choose the most relevant ones? This is where carefully prioritising learning objectives is so important. There is a further consideration, when amino acids fold up into a protein, most of them end up buried inside the protein so how to see these on a solid model? To answer the first question, I grouped the variations by their functional outcome. Some reduce haemoglobin’s ability to bind oxygen; others disrupt the pairing and assembly of the globins; still others make the haem groups wobble around and come loose. Some even make haemoglobin bind too tightly to oxygen, this is actually bad news because then it doesn’t release it to the body tissues! Each of these groups tends to have a specific clinical presentation my students need to know so I coloured one or two of each. The most famous example is a single change, shown here as E6 and coloured purple. This sole alteration has such an effect on haemoglobin that it changes the entire shape of the red blood cell, resulting in sickle cell anaemia. The second question was answered by the skill of Darren Gowers, he cut out curved chunks of the model which are then reattached by magnets. You can just about see this in the bottom right panel of the figure above where one of these units has been removed to show the insides of the protein and in this case the A138 variant.
Using form to show function
One of the drawbacks of the rounded rectangle sticking to a coloured ball 2D depiction is that ligands often actually bind deep inside the protein structure. Some drugs compete with endogenous molecules for binding to these sites, others bind at completely distinct sites on, or in, the protein. An example of this is the receptor for a neurotransmitter called GABA, one of the major signalling systems for the brain and the target of many drugs, including recreational ones.
Like haemoglobin, the GABA receptor protein is a cluster of separate subunits. Rather than a ball like haemoglobin, the GABA receptor is a channel which sits in the cell membrane. A pore runs through the middle connecting the outside and inside of the neuron and regulates the flow of ions which underlie the electrial signals of nerves. Textbooks show variations on a standard theme of circles of GABA sticking to a pair of rectangles spanning a line of membrane. The left side of the figure above shows the 3D printed equivalent. The model is transected horizontally and held together magnetically. Looking from above, as in the righthand side, the top half of the receptor can be lifted off, showing that GABA actually has to diffuse through the outer extent of the receptor to sit in a binding pocket further inside the protein. Benzodiazepines exert their effect by binding to the GABA receptor but at a different site from the GABA itself. This causes a structural change which increases the likelihood of GABA binding. GABA inhibits nerve firing, hence the depressant effect of benzodiazepines. Flumazenil is used to treat benzodiazepine overdose by competing with it for its binding site. The model enables students to grasp more fully the spatial aspects of these complex interactions, called allosteric modulation. This model is another example of how colour can be used – the domains are coloured to match the ones used in the research paper that reported the protein structure so students who want to go deeper can directly align the model with the original data.
Using models to show kinetics
Many molecular interactions rely on dynamic changes in protein structure and it might appear hard to represent this in a static, printed object. In some cases, these changes are transient, lasting a tiny fraction of a millisecond and are deduced rather than directly observed. Increasingly, the technology is available to capture proteins in different orientations and represent them as models.
A cell’s preferred energy source is glucose however glucose cannot diffuse across cell membranes so transporter proteins must carry it from the blood into the cell. Most glucose transporters (called GLUT) carry glucose passively, it moves down a concentration gradient from a higher concentration in blood to a lower concentration in the cell. GLUT is shaped like a bucket, open to the extracellular side, which the blood glucose drops into. Unlike a bucket, the GLUT protein can’t be simply tipped up into the cell. Binding of glucose causes a conformational change as if the bottom of the bucket falls upon, exposing the glucose to the intracellular space which it enters, diffusing away from the GLUT. We have mounted the structure of outward and inward facing GLUT side by sidet o show these different conformations. The models (like haemoglobin) have curved chunks cut out which can be removed to reveal the binding site of the glucose deep inside the GLUT. Glucose is such a key physiological molecule that understanding GLUT function is important in many conditions. As listed in the figure above, these include metabolism, exercise, neuron and red blood cell physiology, tumour biology. The wider relevance of each model is an important practical consideration. Robust, high quality models are not cheap so a widely used one is a better use of limited faculty funds.
Along came a virus
Shortly after I started using these models, the COVID-19 pandemic hit. A tactile, shared resource that encouraged learners to peer closely, touching and breathing all over it could hardly have been less appropriate! Necessity being the mother of invention and all that, I needed to find a different way to utilise the potential of molecular models. I took the first steps on my journey into Augmented Reality…