Predicting disordered regions driving phase separationof proteins under variable salt concentration

Abstract

We investigate intrinsically disordered regions (IDRs) of phase separating proteins regarding their impact on liquid-liquid phase separation (LLPS) of the full protein. Our theoretical approach uses a mean-field theory that accounts for sequence-dependent electrostatic interactions via a random-phase approximation (RPA) and in addition allows for variable salt concentration for the condensed and dilute protein phases. The numerical solution of the complete phase diagrams together with the tie lines that we derive for this model system leaves two parameters to be determined by fitting experimental data on concentrations of all species involved in the system. For our comparisons, we focus on two proteins, PGL-3 and FUS, known to undergo LLPS. For PGL-3 we predict that its long IDR near the C-terminus promotes LLPS, which we validate through direct comparison with in vitro experimental results under the same physiological conditions. For the structurally more complex protein FUS the role of the low complexity (LC) domain in LLPS has been intensively studied. Apart from the LC domain we here investigate theoretically two IDRs, one near the N-terminus and another near the C-terminus. Our theoretical analysis of these domains predict that the IDR at the N-terminus (aa 1-285) is the main driver of LLPS of FUS by comparison to in vitro experiments of the full length protein under the same physiological temperature and salt conditions. SIGNIFICANCEIntrinsically disordered proteins are drivers of cellular liquid-liquid phase separation. However, it remains a challenge to directly predict the phase behaviour of a protein based on its primary sequence, and under physiological conditions. We present a random-phase approximation that allows for variable salt concentration and thus accounts for salt partitioning. We use this to link the sequence of the disordered regions with the behaviour of the complete protein through direct comparisons to in vitro phase-separation assays. In particular, for FUS we determine the exact region responsible for LLPS, weighting in a long-standing debate.