Dynamic nuclear polarization: how a technique from particle physics is transforming medical imaging

30 May 2023

An experimental technique that started life in nuclear and particle physics is now being used to measure chemical reactions inside the human body and to help diagnose cancer and heart disease in almost 50 clinical trials. Jack Miller charts the unexpected rise of dynamic nuclear polarization, which is vastly improving the quality of magnetic resonance imaging

Powerful probe A brain scan using an experimental technique known as “dissolution dynamic nuclear polarization” (d-DNP). This process can greatly improve the images produced using magnetic-resonance imaging (MRI). (Courtesy: James Grist et al 10.1016/j.neuroimage.2019.01.027 )

Life, for physicists, is an odd thing, seeming to create order in a universe that mostly tends towards disorder. At a biochemical level, life is even stranger – controlled and thermodynamically powered by a myriad of different molecules that most of us have probably never heard of. In fact, there’s one molecule – pyruvic acid – that’s crucial in keeping us alive.

When burned, pyruvic acid releases carbon dioxide and water. If you’re exercising hard and your muscles are running low on oxygen, it’s converted anaerobically into lactic acid, which can give you a painful stitch. Later, your liver recycles the lactic acid back into sugars and the process starts anew.Magical molecule Pyruvic acid is central to the technique of dissolution dynamic nuclear polarization (d-DNP).

But pyruvic acid – known chemically as 2-oxypropanoic acid (CH3CO-COOH) – is also a marker for what’s going on inside your body. Run up a flight of stairs, skip a meal or get anaesthetised, and the rate at which pyruvic acid is metabolized (and what it’s converted into) will change. The speed with which it’s made or consumed will also vary enormously if you’re unfortunate enough to have a heart attack or develop cancer.

As it turns out, we can track this molecule by exploiting the intrinsic angular momentum, or “spin”, of the nuclei in pyruvic acid. Spin is a fundamental physical property that comes in either integer or (in the case of protons and carbon-13 nuclei for example) half-integer multiples of ħ (Planck’s constant divided by 2π). Using an experimental technique known as “dissolution dynamic nuclear polarization” (d-DNP), it’s possible to create a version of the acid where many more of its carbon-13 nuclei exist in one spin state than another.

By injecting this “hyperpolarized” pyruvic acid into a biological system, we can improve the notoriously poor signal-to-noise ratio of magnetic-resonance imaging (MRI) by a staggering five orders of magnitude. MRI, which has been of huge benefit in medicine, uses a mix of strong magnetic fields and radio waves to yield detailed images of human anatomy and physiological process inside the body. Its downside is, though, that patients often have sit for over an hour in an MRI machine for clinicians to get images that have a good enough resolution for their needs.

With d-DNP, however, we can gain spectacular MRI images that reveal in detail what happens to pyruvic acid in biological systems. Over the last 20 years, the technique has been used to image bacteria, yeast and mammalian cells. It has looked at animals such as rats, mice, snakes, pigs, axolotls – and even dogs being treated for cancer. Most importantly, about 1000 people at 20 or so research labs around the world have been imaged using d-DNP with almost 50 clinical trials under way.

So how does this technique work and what can it reveal to us about the human body?

Ups and downs of magnetic resonance

Giving clinicians valuable images of the location of water and fat in the body, the beauty of MRI is that it’s non-invasive and won’t harm a patient – even if sitting inside the bore of a magnet is not particularly pleasant. But magnetic resonance can yield far more than just pretty pictures because the behaviour of a nucleus in an applied magnetic field depends on where the nucleus is in a molecule and its precise location in the human body. In fact, we can use radio waves to measure the quantity and location of those nuclei in biological systems, turning MRI into a spectroscopic technique.

MRI spectroscopy is able to reveal the precise the distribution of molecules, such as lactic acid and adenosine triphosphate (ATP – the source of energy for use and storage at the cellular level) in almost any biological tissue. Unfortunately, these molecules are usually present at such low concentration that MRI images of them have a much lower resolution that equivalent images of water or fat. Worse, most MRI spectroscopy experiments require a patient to sit still for hours to get enough decent data, which is difficult especially if they’ve got an itchy nose or need the toilet.

In the late 1990s, however, Jan Henrik Ardenkjær-Larsen – a physicist at the Technical University of Denmark (TUD) in Copenhagen – realized that d-DNP could make MRI spectroscopy much more sensitive. Developed with his TUD colleague Klaes Golman and others, the technique of d-DNP involves some beautiful basic physics that emerged from nuclear and particle labs back in the 1950s (see box). At the heart of d-DNP is the concept of “nuclear polarization”, which comes from the energy levels of a nucleus with spin being splilt into two (or more) components when exposed to a magnetic field. The difference in energy, which is proportional to the strength of the field, provides useful information about the location of the nucleus.

To get an easily measurable signal, however, you need far more nuclei in the higher-energy state (n↑) than in the lower-energy state (n↓). The key figure of merit is the “absolute nuclear polarization”, P, which is the difference between the number of nuclei in the two states divided by their total number i.e. (n↑­ – n↓) / (n↑­ + n↓). For protons or carbon-13 nuclei, which have a half-integer spin, P depends only on temperature, magnetic field and their “gyromagnetic ratio” (magnetic moment divided by angular momentum).

1 In search of improved polarization

Magnetic resonance imaging (MRI) relies on there being many more nuclei with one kind of spin than another because the resulting polarization (P) creates a large and measurable magnetization. Unfortunately, MRI is very insensitive because the value of P for carbon atoms (blue), protons (green) and electrons (orange) is tiny at room temperature – only reaching its maximum value of 1 at very low temperatures and very strong fields (the data here are for a 3.35 T superconducting magnet). While it is hard to reach a value of 1 for nuclei, it is easy to get there for electrons using liquid helium at 1.4 K. Trouble is, most materials don’t contain unpaired electrons. That’s where dissolution dynamic nuclear polarization (d-DNP) comes to the rescue (see main text).

The value of P can range from a minimum of 0 to a maximum of 1 at absolute zero (figure 1). At room temperature and in magnetic fields that we can reasonably achieve in the lab, P is annoyingly small – typically 10–6 or less. In other words, if there are exactly a million spins in the lower state, there are only a million and one in the upper state. However, in a macroscopic biological material there will be enough spin-half nuclei for it to become magnetized – albeit still relatively weakly – when placed in a magnetic field.

Precessing around the applied field several million times per second, this weak magnetization can be measured by applying a pulse of radio waves. They generate a time-varying magnetic field, which induces a voltage in a nearby electrical circuit. To obtain an MRI image, all you need to do is vary the applied magnetic field across a sample and bathe it in radio waves. The result of such experiments is a map of the frequency and phase of the magnetic-resonance signal.

But because P is so small, the magnetization is frustratingly weak, the recorded voltages are small, and the image resolution is poor. Patients requiring, say, a high-resolution brain scan often have sit for over an hour in an MRI machine for clinicians to get a big enough signal-to-noise ratio on the images they need. So even though modern hospital MRI scanners use superconductors that generate some of the strongest and most homogeneous magnetic fields on the planet, MRI – both for imaging and spectroscopy – is still a hugely time-consuming technique. What d-DNP can do is make MRI spectroscopy much more sensitive.

2 Delicately does it

Dynamic dissolution nuclear polarization, or d-DNP, boosts the signal-to-noise ratio of MRI by a factor of 100,000, creating incredible images. It involves transferring the polarization from a substance with unpaired electrons to the protons in a material such as pyruvic acid, shown here in a vial, by zapping it with microwaves at liquid-helium temperatures. The frozen acid is then lifted from the cryostat by hand and melted by dripping superheated water onto it. In this prototype system made in about 2006, Jack Miller is shown holding a large metal stick containing superheated buffer that he is ready to drop onto the sample in the cryostat. This liquid can now be injected into a biological sample and an MRI scan carried out. But you have to be quick as the polarization decays fast. Commercial, automated d-DNP systems designed for human use are now built by firms such as GE Healthcare, which use sterile plastics to transfer the pyruvic acid into a patient. (Courtesy: Jack Miller)

The technique involves mixing pyruvic acid with a stable chemical source of unpaired electrons, typically a carbonyl radical trapped in a tiny molecular cage known as a “trityl radical”. The mixture is put in a vial, which is lowered into a bath of liquid helium, cooling it down to a temperature of 1.4K (figure 2). Microwaves are then fired at the sample, transferring the polarization from the carbonyl’s electrons to the nuclei in the pyruvic acid, which now has a polarization about five orders of magnitude higher than at room temperature.

The acid is then transferred into a patient or other biological system in a nearby MRI scanner. This is done by squirting superheated water at a temperature of about 200ºC through a pipe on to the frozen acid so it rapidly melts. Another pipe is used to suck the acid up through a sterilized filter, which removes the trityl radical. The acid is then mixed with a base (to make sure it’s pH neutral), collected in a syringe and injected into the sample or patient. As the temperature of the pyruvic acid has changed almost instantaneously, the spins in the warm liquid are completely out of thermodynamic equilibrium.

Any enterprising experimentalist has got no more than five minutes to take advantage of the humungous increase in magnetization that dynanmic dissolution nuclear polarization affords

This is not an experiment for the faint hearted as pouring boiling water on to a cryostat is not usually a great idea. It’s also a race against time. From the moment that spin-polarized pyruvic acid is created, its signal starts dropping, returning to equilibrium with a characteristic decay time of about 60 seconds. Any enterprising experimentalist has therefore got no more than five minutes or so to take advantage of the humungous increase in magnetization – and hence signal – that d-DNP affords.

Get in quick

And that’s the big drawback of pyruvic acid. Only processes that occur faster than about 60s can be studied. Researchers literally have to run from their cryostat with their syringe full of pyruvic acid to the scanner. But once injected into a living system, advanced spectroscopic imaging techniques can follow the acid as it moves through the body, monitoring where it is, how quickly it moves and – most importantly – what it changes into (figures 3 and 4).

3 Heart of the matter

These MRI images show how d-DNP was used to study a patient who had suffered a heart attack five days previously. On the far left is the “ordinary” MRI image, with the heart marked by an arrow. The second image was taken after pyruvic acid was injected into one of the patient’s veins, showing how it gets rapidly transported to every tissue in the body, including the heart and spleen (pink regions). The heart uses pyruvic acid in a complex series of biochemical reactions, with the resulting energy used to contract the heart muscle, which you can usually tell from a ring of bicarbonate produced – soon to be breathed out as carbon dioxide. Heart-attack patients do not, however, process the pyruvic acid in the normal way attack, leading to a conspicuous “hole” in the ring (see arrow in the third image). These images reveal a corresponding increase in lactic acid in both organs, especially in this patient’s spleen, which indicates they were producing more immune cells in this areas – a known and imperfectly understood reaction to a heart attack. (Courtesy: Apps et al. 2021 J Am Coll Cardiol Img 14(6) 1285)

The first people to be imaged with the technique were a group of men who had previously been diagnosed with prostate cancer. In a study led by Sarah Nelson from the University of California, San Francisco, in 2013, highly skilled pharmacists created the hyperpolarized pyruvic acid using a repurposed magnet from an Oxford Instruments nuclear magnetic resonance (NMR) machine operating at a field of 3.35T (Sci. Transl. Med. 5 198). After injecting the substance into patients, the researchers were able to detect the cancer in each person examined from the increased amount of lactic acid they subsequently produced.

Lactic acid is one of the hallmarks of cancer because tumours produce a lot of it, acidifying the local environment, disturbing nearby cells and helping the tumour to spread. In one patient, the team in San Francisco even spotted an additional tumour deposit that conventional imaging missed. Confirmed by another biopsy, the detection ultimately led to the doctors changing the treatment that the patient underwent.

4 Brain impact

Conventional MRI is great at revealing the structure of tissues, such as this healthy human brain shown at three different depths (left column, top to bottom). The images were taken by a team led by the radiologist Ferdia Gallagher at the University of Cambridge, UK (NeuroImage 189 171). When pyruvic acid was injected into a patient via a vein, d-DNP images reveal precisely how much energy the brain is using, as indicated by the bright yellow in the vein at the back of the head (second-left column). The acid is converted into lactic acid (middle column) and bicarbonate (right column). It is also possible to determine the rate, kPL, at which pyruvic acid transforms into lactic acid (second-right column). Some patients with tumours would see a dramatic increase in the rate of conversion. (Courtesy: James Grist et al 10.1016/j.neuroimage.2019.01.027 )

One difficulty with d-DNP is that the liquid helium, which is essential for the technique, cannot easily be sterilized. Spores remain visible in it, which – if they got inside a sick patient – could be deadly. It is therefore difficult to ensure that the technique is sterile, safe, repeatable and nowhere near as dangerous as it sounds. Our current solution is to transport the pyruvic acid from the cryostat to the syringe by via a single-use sterilized coaxial plastic tube.

These devices are outrageously expensive to make as they involve various sterile filters, flowing chemicals and computer-driven syringes to handle the pyruvic acid. The tube also has to be sturdy enough to withstand a temperature difference of almost 500ºC (i.e. from liquid-helium temperatures to the boiling-hot solvent) without cracking and spraying fluid around. Each scan on a human participant can therefore cost several thousand pounds.The power of polarization Jack Miller is currently running a clinical trial see how d-DNP can help women suffering from ovarian cancer. (Courtesy: Jack Miller)

But when you think how much it costs to treat cancer patients with surgery or drugs, it’s very much a price worth paying. And the results are breath-taking. What you get is a series of images that, roughly speaking, show the concentration of the pyruvic acid as it moves through the body and the concentration of what it turns into. These images are an invaluable insight into the human condition because the amount of acid depends on the specific biochemical reactions that occur in different parts of the body.

We know, for example, that anti-cancer chemotherapy drugs are successful if they slow the rate at which pyruvic acid is converted into lactic acid. So by imaging a cancer patient with d-DNP after they’ve taken the drugs, clinicians might be able to tell within days or hours if the drug is likely to work. Without d-DNP, patients often require another set of scans weeks later to see if they have worked and if the tumours have shrunk.

There are nearly 50 registered clinical trials using d-DNP around the world, including one I am setting up myself in Denmark. It will aim to help women suffering from locally advanced ovarian cancer, who in about 30% of cases currently need to undergo difficult operations that do not successfully remove the tumour. Surgeons are currently unable to accurately predict if they will be able to cut out a tumour before they start, and may – in hindsight – wish they had tried chemotherapy for longer beforehand.

The technique is the outcome of more than six decades of supposedly arcane basic physics that many would have dismissed as being irrelevant and of no use to the “real world”

Being able to quickly and objectively measure and quantify an individual’s disease – and how it is responding to therapy – is a holy grail of much medical research. Dissolution DNP could be a way to let us do this on a routine basis and is, I argue, a great example of interdisciplinary research and applied physics. The technique is the outcome of more than six decades of supposedly arcane basic physics that many would have dismissed at the time as being irrelevant and of no use to the “real world”.

I take great comfort in knowing that this wonderful mix of quantum physics, chemistry and clinical medicine is literally saving lives.

Stealing polarization from electrons

The principle behind dissolution dynamic nuclear polarization (d-DNP) can be traced back to the US theoretical physicist Albert Overhauser, who realized way back in 1953 that the gyromagnetic ratio of electrons is about 500 times bigger than for nuclei. Given that P is proportional to this ratio, the electrons’ polarization will therefore be far larger too. Overhauser predicted that by firing microwaves of just the right energy at a metal such as lithium-7, which has unpaired electrons, you ought to be able to transfer the large polarization of the electrons to its nuclei.

Three years later Thomas Carver and Charles Slichter showed that the polarization could indeed be “loaned” from electrons in this way (Phys. Rev. 102 975). Using battery-powered solenoid magnets, they increased their polarization by two orders of magnitude from about 10–9 to 10–7. Other physicists joined in the quest for higher nuclear polarizations, with much progress made by the Latvian-born physicist Anatole Abragam. Rather than using lithium-7, he cooled a particular paramagnetic salt in a strong magnetic field until almost all its electrons at thermal equilibrium were in the ground state, achieving a polarization of nearly 1.

By firing microwaves at the sample, he was able to transfer a big chunk of the electrons’ huge polarization to the nuclei. The nuclei’s polarization rose over half an hour to about 0.8, which is many orders of magnitude bigger than it would be otherwise. The nuclei are said to be “dynamically polarized” because as soon as the microwaves are switched off, the electrons and nuclei both relax back to equilibrium. The value of P falls exponentially away with a half life ranging from seconds (for electrons) to days (for nuclei at very low temperatures).

The technique, which was then known simply as dynamic nuclear polarization (i.e. without the “dissolution” term), also became of interest to high-energy physicists at labs like CERN, who realized that metre-sized blocks of cryogenically cooled paramagnetic salts could be used as targets for experiments. These materials can be given a known spin, so by firing beams of particles into them, it became possible to study how hadrons interact under controlled conditions. By the 1970s, the technique had gone from being an obscure solid-state physics “trick” to a routine and useful feature of particle physics.

But there’s a big difference between measuring hadrons at low temperature and probing living biological materials. To do so, we need a molecule that polarizes easily, decays slowly and does something biologically interesting once injected into a living organism. Pyruvic acid fits the bill perfectly. Apart from being at the heart of all chemical reactions that power life, it’s miscible with commonly used chemical electronic free radicals, readily dissolves in hot solutions and is safe when injected into humans.

Jack Miller is an associate professor of medical physics and metabolic imaging at Aarhus University, Denmark, and a Fellow by Resolution at St Hugh’s College, University of Oxford, UK, e-mail jack.miller@physics.org

FROM PHYSICSWORLD.COM     4/6/2023