Fig 1.
Illustration and assignment of binding modes.
(A) Binding modes considered in this work. Binding modes are shown for the interferon-induced, double-stranded RNA-activated protein kinase (RPK). The activation segment (residues 440–450) is not visible in the crystal structure of the monomeric form (PDB: 6d3l [52]), and remains disordered in the dimeric form (PDB: 3uiu, 6d3k). This binding mode represents a disorder-to-disorder transition. Interactions with eukaryotic initiation factor 2 (eIF2), however, triggers folding of the activation segment to mediate inter-molecular contacts (PDB:2a1a [53]), which process is coupled to auto-phosphorylation of Thr446. The RPK binding to eIF2 is classified as a disorder-to-order transition. (B) Assignment of context-dependent binding. Structures corresponding to the same sequence (P19525, residues 440–460) were collected in the monomeric and complex forms. Residues were observed (O) in the crystal structures were assigned as 'ordered', missing residues (M) were assigned as disordered. 'Context-dependent' residues (blue bar) were disordered in the monomeric form, but were represented both in ordered and disordered forms in different complexes. 'Disorder-to-order' residues were disordered in the monomeric structure and ordered (O) in all complexes; whereas 'disorder-to-disorder' residues also remained to be disordered (M) in all bound state structures.
Fig 2.
Determination of binding mode diversity.
(A) Assignment of possible binding sites. The sequence of the N-terminal 20-residue region of glycogen synthase kinase 3 (GSK3, UniProt P49841) is shown. The possible 5 to 9 residue binding regions of Ser9 are displayed together with their probabilities for disorder-to-order transition (pDORi). (B) Frequencies of binding modes. The distribution of the pDO(R) values for Ser9 are shown. The bimodal distribution of the pDO(R) values indicates that Ser9 can populate both disorder-to-order and disorder-to-disorder binding modes. The interactions with low-density lipoprotein receptor-related protein 6 (LRP6) peptides (wheat) and axin (violet) exemplifies the disorder-to-disorder binding modes (PDB: 4nm5), where the N-terminal region (dashed, cyan) does not adopt a well-defined structure in the complex. Phosphorylation of Ser9 induces folding of the N-terminal peptide (lime), which mediates an auto-inhibitory interaction (PDB: 4nm3)[18]. (C) Shannon entropy for binding modes. The Shannon entropy (Eq 5) is evaluated for the binding mode distribution of each residue. The values predict increased number of possible binding modes for residues 7–11, which is consistent with their conditional folding.
Fig 3.
Predicted binding modes of disorder-to-order (DOR), context-dependent (CDR), disorder-to-disorder (DDR) and fuzzy regions.
(A) Binding mode probabilities. The probabilities of disorder-to-disorder transitions are shown for DOR (blue), CDR (lime), DDR (salmon) and fuzzy (red) regions. The pDD values indicate significantly elevated disorder for interactions of DDRs and fuzzy regions as compared to DORs and CDRs. (B) Shannon entropy of binding modes. values for DOR (blue), CDR (lime), and DDR (salmon) regions significantly differ between these binding modes. Context-dependent regions exhibit the highest binding mode diversity as compared to DORs and DDRs. Fuzzy, disordered binding regions (from the Fuzzy Complexes Database [25]) also have elevated
values indicating their context dependence. Statistical significances were determined by Mann-Whitney tests as implemented in the R program. p values as compared to CDRs are shown (** p < 10−2; *** p < 10−5).
Fig 4.
Prediction of context-dependent regions by the FuzPred method.
(A) Prediction of binding mode profiles. Comparable probabilities for disorder-to-order transition (pDO, dark gray) and disorder-to-disorder transition (pDD, light gray) indicate a disordered binding mode for the region of residues 45–55 (grey box), which involves both the docking and the KIS motif, consistently with the experimental data [33] (top panel). Based on the binding profile, this region can fluctuate between ordered and disordered interactions (bottom panel), which will depend on the signaling pathway. The values indicate that both the docking motifs and the N-terminal part of the KIS domain are capable to establish different binding modes, consistent with their involvement in disordered interactions. Selected MKK4 conformers docked onto p38α structure (PDB:1lew). The docking motif (marine) and the KIS domain (light blue) are shown (coordinates as a courtesy of Dr. Malene Ringkjobing-Jensen). (B) Prediction of phosphorylation-induced folding. Trans-autophosphorylation induces folding of the activation loop in the dual-activity enzyme Ire1, which promotes its oligomerisation [19]. Packing of four monomers (wheat, light blue, pale green and light pink surfaces) (PDB: 3fbv) are stabilised by the ordered activation loop (cartoon, the phosphorylated Ser841 is shown by spheres). FuzPred predicts slightly higher probabilities for disorder-to-order transition (pDO, dark gray, top panel) for the activation loop (grey box) than for disorder-to-disorder transition (pDD, light gray, top panel), indicating that it can fold upon binding. The high
values (bottom panel) corroborate that the activation loop can sample both disordered and ordered states in the bound form, which could be shifted towards the folded form by phosphorylation.
Fig 5.
A binding mode landscape for disordered protein interactions.
Residues are characterised by their binding modes to increase or decrease order upon interactions and the context-dependence of such binding modes. (A) The binding modes, reflecting the level of disorder in the bound state, are represented on the x axis; ranging from structured, well-defined to disordered, heterogeneous interactions, as quantified by the pDD values. Context-dependence, reflecting the level of fuzziness, is displayed on the y axis, ranging from stable, uniform to diverse, inducible binding modes, as quantified by the values. The pDD and
values are predicted from the sequence by the FuzPred program. A disorder-to-order binding with low context-dependence is exemplified by a disordered loop (504–512 aa, blue squares) in Taq polymerase, which adopts a stable structure upon interacting with DNA (PDB: 3lwl [54]). A disorder-to-disorder binding with low context-dependence is represented by the heterogeneous interactions between the elongation factor AF4 (residues 747–754, orange diamonds) with leukemia fusion protein AF9 (PDB:2lm0 [55]). Fuzzy, context-depedent interactions sample a wide variety of binding modes ranging from disorder-to-order to disorder-to-disorder transitions. Context-dependent disorder-to-order binding is exemplified by the polymorphic interactions of the ribosomal S6 kinase 1 (RSK1, residues 697–703, light blue dots), which adopts different secondary structures upon binding to S100B, corresponding to autoinhibited and active forms (PDB:5csf, 5csi, 5csj [23]). Conditional folding upon binding is represented by the N-terminal region (residues 15–25, lime dots) of the large chain of ribonucleoside-diphosphate reductase, which can be structured or disordered in different oligomers (PDB: 1zyz, 1zzd [56]). Context-dependent disordered binding is exemplified by the p150 subunit of the eukaryotic initiation factor 4F (residues 225–235, light orange dots). eIF4 wraps around the translation initiation factor 4E, but the flanking region remains to be highly dynamic in the assembly (PDB: 1rf8 [57]). The interaction sites are shown by the same colours as interaction modes, and partner proteins are displayed by grey surfaces. (B) The characteristics of the different binding modes, which are represented in panel A. The binding mode landscape comprises a continuum of interaction behaviours, the major trends of which are illustrated by the distinct modes.
Fig 6.
Binding mode landscape for p53 interactions.
The oligomerisation domain (residues 325–356, blue squares) exhibits a strong preference for disorder-to-order transitions and forms stable tetramers (PDB:1c26) [36] and higher-order structures. Short linear peptides (residues 378–386, orange diamonds) at the disordered C-terminal regulatory region interact with sirtuin (PDB: 4zzj [37]) and the cyclin dependent kinase cyclin A (PDB:1h26 [38]) exhibit heterogeneous binding modes. On the top of the binding mode landscape two context-sensitive regions are shown. The disordered N-terminal transactivation region interacts with Mdm2 (PDB:1ycr [58]) via a short helical segment (19–25 aa, lime dots). The beginning of the disordered C-terminal region folds into an α-helical conformation (residues 278–285, green dots) to recognise DNA via a variety of dynamic binding modes (PDB: 2ady, [36]). The high values for both regions indicate fuzzy interactions, which are strongly influenced by the cellular context. The interaction sites are shown by the same colours as interaction modes, and partner proteins are displayed by grey surfaces.