Scalable methods to study protein solubility: Applications for biomolecular condensates

Proteins are the machines of life, but they are inherently not very soluble. While mostly ordered and well-folded, many proteins contain highly dynamic and disordered regions with poorly understood functions. A recently proposed role for these is the formation of essential cellular structures via liquid-liquid phase separation (LLPS) to form biomolecular condensates. This behaviour is related to their solubility, and some of them addition- ally form stronger and more persistent aggregates tied to neurodegeneration in diseases such as Amyotrophic lateral sclerosis (ALS). Therefore, it is essential to understand the difference between transient interactions in liquid protein phases and those in malignant aggregation, but the limitations of current experimental methods hinder these studies.

The work in this thesis has focused on empowering such solubility studies by developing novel methods capable of generating the large-scale data needed to understand protein LLPS better. A protein’s solubility is determined by its sequence of amino acids and its surrounding environment, which requires a detailed study of both. To this end, two experimental approaches have been used; one focusing on large-scale screening of the amino acid sequence dependence of LLPS using display technology, and one centred around detailed screening of solution conditions using microfluidics.

Firstly, we show that partitioning experiments using the powerful technique mRNA-display can be used to probe phase separation. We studied the disordered N-terminal domain of human DEAD-box Helicase 4 (Ddx4) involved in forming essential germ granules, and directly show the influence of individual amino acids on liquid protein assembly. We corroborate known important roles of aromatic groups and arginine but show more broadly that partitioning correlates with the propensity for the protein to fold. The approach is unique in its ability to investigate isolated LLPS behaviour in vitro and might open further inquiry into dynamic and ”fuzzy” protein interactions.

Secondly, two microfluidics-based approaches were developed to enable rapid and highly controlled experiments to be carried out using an easy-to-use and fully automated instrument. Microfluidic studies on LLPS have superior analytical capabilities but are currently limited to expert users. The new methods described here can robustly deliver detailed equilibrium and non-equilibrium characterisation of LLPS using a minimal amount of sample in relatively high throughput. We uncover how some proteins, which undergo LLPS, have a re-entrant transition with respect to buffer ionic strength and additionally show how drug-like compounds can be efficiently screened.

While Ddx4 forms only liquid assemblies, the methods could also be used for disease-related systems undergoing more persistent aggregation. More specifically, the two approaches can act in concert to help understand how single genetically inherited mutations can tip the balance and cause malignant aggregation. Generally, both approaches are useful for studying protein solubility and might also help optimise the solubility of protein-based drugs and enzymes.