Bringing SwissFEL light to industrial users

John, let’s start by getting some terminology out of the way. Cristallina is the third experimental station on the hard X-ray branch of the SwissFEL. You lead a project called Cristallina-MX. What is this?

John Beale: At Cristallina-MX, our goal is to provide macromolecular crystallography using an X-ray free electron laser at an increased throughput. This will make it easier for industry to make use of this source. There’s significant potential for X-ray free electron lasers to be used in drug discovery, but currently inefficiency in data-collection and throughput inhibits this application.

Why is that?

An X-ray free electron laser such as the SwissFEL is a very special kind of facility – there are currently only five such user facilities in the world. For each of these, given their linear structure, there are a limited number of experimental stations. At SwissFEL, we have just two beamlines, compared to sixteen at the Swiss Light Source SLS. With experiments typically lasting between three and five days, it is understandably highly competitive to get beamtime at an X-ray free electron laser. These beamtimes are also highly complex, requiring lengthy optimisation by specialist users.

For industry, time is money. Industrial users need higher throughput and automation. They need to do their experiments in a few hours, not a week. If we can facilitate this, we can enable a wider base of users – including also new academic users - to benefit from the increased brilliance that an X-ray free electron laser provides.

What do pharmaceutical companies stand to gain by doing macromolecular crystallography at an X-ray free electron laser?

First, let me explain a little background.

At an X-ray free electron laser, protein crystals are destroyed almost the instant the light touches them. We have about ten femtoseconds before this happens to collect a diffraction pattern – we call this ‘diffraction before destruction’. This isn’t enough time to spin a crystal and collect multiple diffraction patterns as we would at a synchrotron. Instead, we collect diffraction patterns from thousands of different protein crystals in random orientations and put these together to calculate a structure. This is known as serial femtosecond crystallography.

For pharmaceutical companies, there are three main value propositions of serial femtosecond crystallography.

The first is to study proteins that typically only grow as small, weakly diffracting crystals and require higher brilliance than is achievable at synchrotrons. The most obvious group here are membrane proteins, which are notoriously difficult to crystallise but are a vast source of drug targets. Apparently, as many as 60 percent of current drug targets are membrane proteins.

And the other reasons? Let me guess, does it have to do with the ‘femto’ in serial femtosecond crystallography?

You’ve got it. Time-resolved measurements. Synchrotrons just can’t go to the same ultrafast timescales that we can do at an X-ray free electron laser. This means we can study the dynamics of protein structures and their molecular mechanisms at near atomic resolution.

And the third reason?

At an X-ray free electron laser, we collect genuinely radiation damage free structures. This means we can study proteins that are not cryogenically cooled and are in a more physiological state. This can enable, for example, proteins that have metal ions to be studied. These are notoriously difficult to capture correctly at a synchrotron, but at an X-ray free electron laser we can study them in their unreduced state. The metal co-factor can make these proteins quite heterogeneous in their binding partners. This makes them challenging to specifically inhibit, and as a result they are not normally targets for drug discovery. If we can capture these proteins in their non-reduced form, it may better enable their suitability as drug targets.

So, how are you increasing throughput at Cristallina-MX?

It really comes down to sample delivery. I’ll explain why. For serial femtosecond crystallography, crystals need to be delivered in a continuous manner. The typical way of doing this is to use liquid jets. A solution containing lots of little protein crystals is squirted out in front of the beam. A lot of fantastic science is done this way, but it is very sample intensive and difficult to automate.

We don’t do this. We use fixed targets. For non-membrane protein samples, we have fabricated, together with the Laboratory for X-ray Nanosciences and Technologies at PSI, a solid support that contains twenty-six thousand tiny apertures. A slurry of protein crystals can be loaded on. We can then raster scan these very quickly, taking a diffraction pattern from each well. For membrane proteins in viscous media, we use a target developed by the Max Planck Institute, Heidelberg.

These give the same effect as using a liquid jet but require less sample and give a higher hit-rate. They also lend themselves better to automation. So, all in all, we can be a lot quicker.

How much quicker are we talking?

Well, raster scanning a target can take just a few minutes. The key thing is to have a high hit-rate and collect as much data as possible in this time. After the scan the targets need to be exchanged, and this limits throughput. That’s why we’ve installed a robot to do this. Eventually, we will make the whole process completely automated.

With all this, we would hope to offer very high throughput analyses with serial femtosecond crystallography to at least two-three industrial clients per day during beamtime in addition to our academic users. We hope this will become a major resource for pharmaceutical companies.