New era of lower-cost, lower-energy cryo-EM ushered in by MRC Laboratory of Molecular Biology in Cambridge
It has already revolutionised the study of biological structures, but now a new era of electron cryo-microscopy (cryo-EM) is being ushered in thanks to astonishing work led by the MRC Laboratory of Molecular Biology in Cambridge.
A new electron cryo-microscope has been built at a fraction of the multi-million pound cost and size of current models by overturning conventional thinking about their operation.
Drawing in expertise from around the world to help, the LMB team has spent four years creating the new microscope, which promises to broaden access to one of the most sought-after technologies in the field and further unleash its potential to drive discoveries about human biology and disease.
“We are very excited about it,” the LMB’s Chris Russo told the Cambridge Independent. “We know it’s the right way to go and we hope multiple companies will get involved. There is a whole new opportunity in cryo-EM.”
The master stroke behind the new microscope is that it operates at a far lower energy than usual.
Current electron cryo-microscopes typically operate at 300 electron volts (keV) as this was thought to be the optimum level.
But using such high electron energies has sent the cost of these microscopes spiralling – a state-of-the-art machine today might set a lab back £5million, cost about £250,000 a year to run and require a sizeable facility, as they are several metres high.
Many institutes, laboratories and universities that would love to use cryo-EM are simply priced out of the market.
However, previous research published in 2019 by Chris’ group in the LMB’s Structural Studies Division, and reported by the Cambridge Independent, showed that 100 keV, not 300 keV, is in fact the optimum energy for imaging.
Since that proof of concept work, Chris’s group, along with Richard Henderson – who shared the 2017 Nobel Prize in Chemistry with two other scientists for his crucial role in the development of cryo-EM – have been working with their colleagues across the LMB and collaborators to build their new 100 keV microscope, beginning with a column from Japanese manufacturer JEOL.
“We made a list of all the things we wanted as part of this microscope and went round the commercial microscopes and settled on a JEOL1400 column as the starting point for a whole new design,” explains Chris.
“We chose it for a couple of reasons. It had many of the things we wanted for a base design. It’s a very robust, cheap, widely-used microscope. Its most common use is in pathology labs.
“We lopped the top off the column, took out the lens and did a whole new design for the objective lens.
“We have a new election emissions source developed by our colleagues in York and they developed a new power supply as well. Then we have a detector developed in Switzerland by a company well known for making detectors for X-rays for synchrotrons but which is very interested in moving into electron microscopy.”
With funding from the Wellcome Trust among others, the team compiled the components they needed for their new design.
“The general philosophy was to start with something that is cheap and build it up to what we need, rather than taking an expensive microscope and making it smaller and cheaper. We spent two years building it and finished in December last year.”
Led by Greg McMullan, and assisted by the LMB’s scientific computing facility, electronics and mechanical workshops, and Shaoxia Chen and Giuseppe Cannone from the LMB’s electron microscope facility, the new microscope was built for less than £500,000 – a tenfold reduction on the £5million cost of today’s microscopes.
“Once it’s commercialised it might creep up a bit, but under £1million is very realistic here, so it is a fraction of the price for a saleable product,” said Chris.
The costs of establishing a microscopy room have also been reduced tenfold by the new design – while the running costs will be just five per cent of current levels, meaning many more labs could afford to use it for studies, including helping to discover new drugs.
Having built the machine though, the team needed to prove its capabilities.
“We decided there are more hardware improvements to improve this even more, but they will take a few more years, so now is the time to see how this really performs on realistic specimens,” said Chris.
“We went down the hall to our collaborators and two of our students made some specimens. They downloaded the genes for several proteins from databases, expressed them and purified them from bacteria. We put them in the microscopes and had three more structures in two weeks.”
The researchers – including PhD students Katerina Naydenova, Josh Dickerson and Daniel Mihaylov, plus postdocs Mathew Peet and Hugh Wilson – used the new microscope to determine 11 atomic structures in total.
“The turnaround time is phenomenal. Part of that is having a microscope that is dedicated to it,” explained Chris. “There’s no waiting for access to a facility and sending grids to a major national centre and then waiting for the data to come back to process – we cut out all of that loop, so you can solve structures extremely fast.”
The team chose a diverse range of macromolecular specimens, of different sizes and symmetries and with subunit numbers ranging from one to 60, to show the microscope’s capabilities. Each structure was solved in a single day of data collection.
“Essentially we get the same result as you would out of a Titan Krios microscope, but with about 100 times less data,” said Chris. “The camera is only 1K by 1K – 16 times smaller than what we are used to on the latest 300 keV microscopes, which are normally 16 megapixels, so 4K by 4K.
“People collect massive datasets but we didn’t need to. We collected a few hundred micrographs, which were 16 times smaller, so the equivalent of maybe 10 or 12 micrographs on a Krios and we got structures from which we could build very nice atomic models.”
Given the results, published in a paper in PNAS this week, you might wonder why 300 keV was ever thought to be the optimum energy level.
“There were several things that people thought were plaguing images in cryo-EM of biological specimens that would all get better if you went up in energy,” explained Chris.
Specimens in cryo-EM are frozen in a thin, glass-like layer of water and an electron beam is fired through them.
“When you put an electron beam through the specimens, you generate a charge – the high energy electrons come in and kick out other electrons and you are left with positive charge. That can move around, or not, and it can cause distortion in the images.
“Glass is an insulator so the charge builds up. The thought was the potential image distortion would be worse the lower the energy goes – because one volt in 300,000 is less than one in 100,000,” explained Chris.
“So both dynamic, twinkling charge as well as static build-up of charge were both thought to be worse when you went down in energy.
“Five or six years ago, I started to question that assumption and we did some experiments and published two papers proving that charging was not a problem on these cryo-EM specimens. It did exist, we could measure it, but it was way too small to make any difference. So that wiped out that reason to go up in energy.”
Another reason that high energies were adopted related to how information is gleaned from the phase contrast images and turned into 3D structures.
“We used the ‘projection approximation’, which means when we are taking an image we assume the beam goes completely through it, like an X-ray of your hand, and you are left with the shadow pattern. Approximation gets better the higher you go up in energy. The more energy, the more it looks like a projection.
“But we published another paper in 2019 on that. We developed a method that would enable you to correct for the non-projection part of the image just by processing the data correctly in software. Now everybody uses this in all the programmes they use to process cryo-EM software. It’s just a button you push.
“Once we ruled these out, there was no reason to go up in energy. In fact, we found, if you do the maths and the physics, you find there is an optimum energy and for the vast majority of specimens it is 100 keV, not 300 keV. You get 30 per cent more signal in every image. It was a real surprise. The first motivation to go down in energy was that it would be cheaper, but the thought was it would be worse. But actually, it’s cheaper and better and that’s the real surprise of all this.
“People have been making all these microscopes and they are at the wrong energy to solve structures as effectively as possible. This paper is a monumental moment, because there is no argument any more that this is correct.”
There is a major environmental benefit to the change too.
“It means we can get rid of this gas we have to use in high voltage supplies called SF6,” said Chris. “Most very high voltage transformers and high voltage power supplies in electronics have to do something to stop the electrons arcing to ground and shorting out inside the electronics. The most common way they do this is to use sulphur hexachloride. But this gas is nasty – it’s about 2,600 times worse than carbon dioxide as a greenhouse gas. It’s becoming heavily regulated. You have to reclaim it – you can’t release it into the atmosphere.
“It would be best if you never used it at all. All the 300 keV microscopes have to use it – nobody knows how to use it without it. But the 100 keV design that we’ve done now has completely got rid of it. There’s no SF6 at all. It’s better for the environment and the costs are a lot lower.”
Getting rid of this gas was the biggest challenge when designing the field emissions source.
“But our collaborators in York are very good at this,” noted Chris. “Each individual component had its own engineering challenges.
“On the microscope side, the biggest challenge was getting the improvement in the objective lens which does the imaging, without adding cost. By reducing the energy, you have to have better lenses to get to the same resolution, because the effect of the aberrations is increased. But we managed to do that in collaboration with JEOL. We’ve got a new lens which outperforms previous lenses that had been used for cryo-EM by a significant amount – and we showed that in the paper.
“On the detector side, taking one that was essentially designed for X-rays and synchrotrons and making it compatible with the microscope was no small task. It had to go inside a vacuum chamber and be adapted for the purpose.
“Greg McMullan, the lead author, put all this together and wrote new software, new algorithms to drive the whole system and make it a usable microscope.”
While resolution-for-resolution a £5m microscope today will still beat the new machine, Chris believes that will change – and the team is already working on a second machine.
“With another round of improvements, I think we will be at the point where it even outperforms a Titan Krios pixel for pixel, which will make this unstoppable,” he said.
For now, there are already early adopters ready to try out the new design. And Chris is hopeful that microscope manufacturers will jump on the team’s work.
“There are maybe 400 of these high-end microscopes in the world, distributed among places that can afford to purchase and run them. That limits access. There are plenty of places in Cambridge that would like to do more cryo-EM but they can’t afford it. Decreasing the cost of this really ought to have a big impact.”
The LMB has been at the forefront of developing cryo-EM for decades.
This latest work is set to catapult this technology – at last – into the labs of many more researchers.
It was funded by UKRI MRC, UKRI BBSRC, UKRI Innovate UK, the Wellcome Trust, UKRI EPSRC, Astex Pharmaceuticals Sustaining Innovation postdoctoral fellowship and a Herchel Smith fellowship.
How does it work?
Cryo-EM has become pivotal in recent years to many discoveries about human biology. It has become a primary method for determining the structures of biological samples, which is key to the discovery of new drugs.
X-ray crystallography has been used for this purpose for more than a century, but it relies on turning samples into crystals before they can be studied, and some molecules or complexes of molecules do not form crystals. Many proteins - key to the study of disease - are particularly difficult to crystallise.
In cryo-EM, samples in an aqueous solution are applied to a grid mesh, then plunged into liquid ethane, which vitrifies the water around the sample.
“When we freeze our specimens, they are frozen into this thin layer of water,” explains Chris. “The water is turned into a glass - an amorphous material. There are no crystals, unlike normal ice. We do this because we want to see the specimens as if they are frozen in a pane of glass, it’s just our pane is made out of water instead of sand.”
Accelerated electrons are fired through the thin sample and, unlike photons of light in standard microscopes, these tiny charged particles have such a minute wavelength that they can interact with nanoscale components. By collecting two-dimensional images, researchers can build up a 3D shape and achieve atomic resolution.