There are many examples of point-of-care diagnostic tests, such as blood glucose, pregnancy and Covid-19 tests, which use protein-sensing systems to provide immediate results. But the healthcare industry is only scratching the surface of the potential of proteinbased diagnostics, according to Professor Kirill Alexandrov of the QUT School of Biology and Environmental Science, a researcher with the ARC Centre of Excellence in Synthetic Biology. He is one of the researchers behind a new approach to diagnostic testing that he and his colleagues hope could eventually lead to a scalable modular test for – well, the sky’s the limit.
Imagine a comprehensive hospital lab, he urges. Almost all the clinical diagnostic tests that are performed there – with the exception of mass spectrometry, which works using ions – are carried out with some involvement of proteins, whether that’s antibodies or various kinds of enzymes. It makes sense, given that proteins are involved in countless important functions that happen in the body, serving as structural components in cells, catalysing biochemical reactions, and regulating gene expression. They are also essential for processes like muscle contraction, cell signalling, and immune responses. Even this conversation we’re having, Alexandrov expands, is possible because proteins are turning on and off in our brains and performing sophisticated information processing at the millisecond scale.
For Alexandrov, this is the beauty of biology. Unlike an industry such as electronics, where engineers developed relays, switches and vacuum tubes without knowing that an iPhone would come out of it one day, in biology, we know what the product looks like. We live in it every day. “What we don’t know is how much time and money it will take us to replicate or to build the technologies and the industries that can deliver the performance on the same scale,” Alexandrov says. “But ultimately, biotechnology is an exercise in reverse engineering.”
He believes that he and his QUT colleagues have made an important step towards building artificial protein machines with diagnostic utility, a breakthrough that was published in the journal Nature Nanotechnology. The paper describes a new approach for designing molecular ON-OFF switches based on proteins. “What we’re doing is trying to create protein-based machines that we can train to recognise something that we would like to detect and then give us a signal that we can understand – basically we’re building biological transistors,” Alexandrov explains.
“The challenge is using our extremely basic protein engineering skills to build these things to specifications. Historically, biotechnology has been an act of trolling nature, finding things that are doing kind of the right thing and then building the hardware, chemistry and software around them to get what we need.”
Improving the existing standard of care
The example detailed in the research to demonstrate the technology focuses on a cancer chemotherapy drug methotrexate that is toxic and requires regular measurement to ensure patient safety. If too little of the drug is delivered, the cancer will survive, but too much of it could kill the patient. The sensor Alexandrov and his colleagues designed for the drug uses a colour change to identify and quantify the drug.
It’s been a long road to reach this point, with one lesson standing out as particularly painful. Ten years ago, when Alexandrov first discovered the potential of the research he and his colleagues were carrying out to develop protein switches, he got a little overexcited. “We thought, brilliant, this could solve the problem of personal diagnostics, and we rushed to commercialise the technology,” he recalls. This was about the time Theranos was hitting the headlines with its lab-on-a-chip technology, which founder Elizabeth Holmes promised could run hundreds of tests in a doctor’s office on a single drop of blood. She was ultimately proven to be a fraud and sentenced to more than 11 years in prison.
“For our part, we realised that, while our technology was real, we had very little idea about how the diagnostic industry works and couldn’t answer very basic questions, such as: Who would our customers be? What are the market forces? What do regulatory bodies want? How much does it cost to build an integrated diagnostic company?” Alexandrov admits.
“What we’re doing is trying to create proteinbased machines that we can train to recognise something that we would like to detect and then give us a signal that we can understand – basically we’re building biological transistors.”
Professor Kirill Alexandrov
He decided to take a different tack the second time round. Rather than attempting to put a catchall handheld diagnostic device on the market, they identified a real issue healthcare providers needed to solve. Sophisticated and expensive lab equipment for therapeutic drug monitoring is often unavailable in remote areas in Australia for economic and logistical reasons. “If we could provide a smart kit that could provide an answer for those patients on how their drug is performing without the need for expensive equipment, so clinical practitioners can decide where to go with the therapy, we could contribute to improving the existing standard of care,” Alexandrov explains.
The next step is for the sensor to be tested in Queensland Health laboratories for use in a clinical setting. If it is successful, it could be the first time an artificially designed protein biosensor is suitable for a real-life diagnostic application.
Potential to scale
But it’s not the specific application that Alexandrov finds most exciting. It’s the platform itself, which is modular, similar to building with Lego bricks, meaning that – in theory – you can replace parts easily to target something else, such as another drug or medical biomarker. Although the truth is that there are many scientific questions that need to be answered before the researchers can achieve anything close to the universal platform they are ultimately aiming to build.
“The platform not only needs to be able to measure a wide range of entities – from small molecules to DNA ions – but it needs to be able to do that across a huge range of concentrations,” he says. “The answer also needs to come through in a timeframe that is acceptable for the healthcare provider, which is generally a few minutes.”
Here, Alexandrov reminds me again that because this conversation is happening, we know it is possible for information processing to occur at this scale. But the big question he keeps coming back to is how do we build that real-time biological information processing system? What can we assemble it from? What can we compromise on? And what can we not compromise on?
To begin to answer some of these questions, he plans to bring in experts with additional skill sets to the next phase of the research. “We need modellers, mathematicians, people who think about systems and information processing in biology,” he says. “We are also really at the beginning of building our expertise in building protein switches, so we need to get better at that too, especially building them to technical specifications.”
Going out with a bang
Machine learning is making some aspects of this work easier. “While it hasn’t so far solved a problem we couldn’t have solved using traditional methods, it is reducing costs and speeding things up through sheer brute force,” Alexandrov says. “For example, it could be used to help us prototype a thousand switches, extract every useful parameter and feed them into a model, which would hopefully give us predictive guidance on where we should be taking the design.”
Even AI can’t speed up the regulatory process, however. “Things move slowly in the medical device field, particularly on the translational end,” he notes. “So in the short term, given that it takes on average two years to approve a medical device, we can only hope to have one product on the market that uses a synthetic protein switch in the next five years,” Alexandrov notes. “That in itself, though, would be incredibly exciting.”
Looking further ahead, this form of proteinbased sensing has potential applications beyond the diagnostic testing industry, too. “It’s actually a preface to a much more sophisticated form of bioengineering,” Alexandrov grins. “I’m talking about smart drugs – protein nano machines that go into the patient, recognise a biomarker and turn on a therapeutic function. It would use the same switch; it would just have a different chemistry. The opportunities in this area are huge.”
In the meantime, Alexandrov and his team are focusing on getting their first synthetic switch to patients, before pushing out other tests that can meet healthcare providers’ immediate needs. “It’s a long road and this concept will be the focus of my entire career,” he concludes. “But even if by the time I retire I launch three products, I’d be going out with a bang.”
