The ultimate goal, to systematically test every single cell surface protein against all of its potential matches, on any type of cell, is no longer squirreled away.
“With the original method, the technology limited us to testing hundreds of proteins; humans, however, have several thousand. So we needed to modify our approach and upscale things. And when you add to the number of proteins to be tested linearly, you’re actually increasing the number of potential pairings by the square of that number because we test them in a matrix of all-by-all. So we needed a way of doing this at much higher throughput.”
One of the most challenging aspects of increasing capacity was processing the proteins after they had been produced. “Purifying one protein is easy, but hundreds at a time is not. There aren’t off-the-shelf devices out there that you can buy that can do this.” says Gavin. They were assisted by Colin Barker in the engineering workshop at Sanger, who designed and made bespoke equipment.
The new version of AVEXIS, nicknamed “SAVEXIS”, makes it possible to screen tens of thousands of interactions per day while only using minute amounts of protein.
SAVEXIS goes contrary to the conventional wisdom for binding assays. Instead of using two different chemistries to create bait and prey proteins, the team found a way to base the technique off a single ‘universal’ design. Not only did that cut in half the work of making proteins, but with further tinkering it also enabled them to test surface proteins not compatible with traditional designs. It also reduced by 10 to 100-fold the amount of protein needed to perform each experiment.
Jarrod Shilts, a PhD student at the Sanger Institute, now working in Cambridge and the University of York, undertook the work of developing the new technique. He says; “it was a really cool idea. When we started I thought it probably wouldn’t work. It was new territory for us – new robotics, for example. We needed things to scale by an order or magnitude.”
But it did work, and the technique made it possible to test more proteins than ever before.
“To test our scaled up method, we chose the human immune system. It seemed like the best case because the cells are well characterised in terms of which proteins are present on their cell surface. And it's known that targeting these interactions is important for medicines. For example, a lot of cancer checkpoint inhibitors target white blood cell surface proteins, ” says Gavin.
Jarrod tested 630 proteins against one another in a matrix of 396,900 experiments. The result is a huge ‘interactome’ of the immune system. They combined their data with the vast datasets from the Human Cell Atlas, which detail the proteins that are active in individual cells, and cell types, across the whole body. From this integrated view, Jarrod could even produce a mathematical model that predicts cellular connectivity from basic physical principles.