Sequence-Dependent Correlated Segments in the Intrinsically Disordered Region of ChiZ

Abstract

How sequences of intrinsically disordered proteins (IDPs) code for their conformational dynamics is poorly understood. Here, we combined NMR spectroscopy, small-angle X-ray scattering (SAXS), and molecular dynamics (MD) simulations to characterize the conformations and dynamics of ChiZ1-64. MD simulations, first validated by SAXS and secondary chemical shift data, found scant α-helices or β-strands but a considerable propensity for polyproline II (PPII) torsion angles. Importantly, several blocks of residues (e.g., 11–29) emerge as “correlated segments”, identified by their frequent formation of PPII stretches, salt bridges, cation-π interactions, and sidechain-backbone hydrogen bonds. NMR relaxation experiments showed non-uniform transverse relaxation rates (R2s) and nuclear Overhauser enhancements (NOEs) along the sequence (e.g., high R2s and NOEs for residues 11–14 and 23–28). MD simulations further revealed that the extent of segmental correlation is sequence-dependent; segments where internal interactions are more prevalent manifest elevated “collective” motions on the 5–10 ns timescale and suppressed local motions on the sub-ns timescale. Amide proton exchange rates provides corroboration, with residues in the most correlated segment exhibiting the highest protection factors. We propose the correlated segment as a defining feature for the conformations and dynamics of IDPs.

Motif-pattern dependence of biomolecular phase separation driven by specific interactions

“…These interaction motifs can include folded domains, such as in the nephrin-Nck-N-WASP system for actin regulation [7], or individual amino acids in proteins with large intrinsically disordered regions (IDRs), such as the germ granule protein Ddx4 [8]. While the precise sequences of these motifs are believed to play a major role in determining condensates' phase diagrams and material properties, the nature of this relation has only begun to be explored [9][10][11]. As a result, it remains difficult to predict the formation, properties, and composition of these diverse functional compartments.…”

Artificial intelligence guided conformational mining of intrinsically disordered proteins

“…Conformations for training and testing the autoencoders came from multiple μs-long MD simulations 14 , 23 . We collected 95,000, 140,000, and 145,000 frames, respectively, at 10 ps intervals for Q15 and 20 ps intervals for Aβ40 and ChiZ from each replicate run; the numbers of replicate runs were 2, 4, and 12, respectively.…”

“…The conformational ensembles of IDPs are broad and difficult to model 14 . A possible crucial benefit of representing the conformations in the latent space is that, due to the reduced dimensionality, the distribution of the latent vectors would be more compact and therefore easier to model.…”

“…Amyloid-beta peptides, including Aβ40, are linked to Alzheimer’s disease 13 . The cell division machinery of Mycobacterium tuberculosis , the causative agent of tuberculosis, contains a number of membrane proteins, including ChiZ, with disordered cytoplasmic regions 14 , 15 . The functional and disease mechanisms of these and other IDPs remain unclear, in large part because we lack knowledge of their conformational ensembles in various states (e.g., in isolation, in aggregation, and bound with interaction partners).…”

“…Molecular dynamics (MD) simulations offer an attractive approach for IDPs, with an atomic representation for each conformation, but the simulation time that can be presently achieved, which directly determines the extent of conformation sampling, is largely limited to 10 s of μs. The conformational ensembles of the 64-residue cytoplasmic disordered region of ChiZ (referred to simply as ChiZ hereafter) sampled by multiple replicate simulations, totaling 36 μs in solution and 38 μs at membrane, have been validated by SAXS and NMR data 14 , 15 . While we cannot answer whether 10 s of μs of simulations are really long enough, we do know that shorter simulations are insufficient.…”