Due to the specific peculiarities of their amino acid sequences (high net charge and low hydrophobicity), conformationally, extended IDPs do not follow the rules established for ordered globular proteins and domains that are known to lose specific structure and biological function at various denaturing conditions (e.g., extreme temperatures or pH). In fact, extended IDPs, such as α-synuclein [75], caldesmon 636-771 fragment [76], phosphodiesterase γ-subunit [77], the receptor extracellular domain of nerve growth factor [78], α_s-casein [79], and several other IDPs are known to gain more ordered structure based on temperature, and are usually more structured at higher temperatures (due to the enhancement of the hydrophobic interaction), while being more disordered at lower temperatures [35, 80].

Similarly, many extended IDPs, such as pig calpastatin domain I [81], naturally occurring human peptide LL-37 [82], prothymosin α [83], histidine rich protein II [84], human α-synuclein [75], and several others, undergo partial folding at extremely acidic and/or alkaline conditions as a result of neutral pH minimizing their large net charge, leading to the minimization of the intramolecular charge/charge repulsion and thereby permitting hydrophobicity-driven folding to a more ordered/collapsed conformation [80].

Obviously, these considerations should be protracted from IDPs to the extended IDPRs commonly present in various hybrid proteins. Furthermore, it is likely that the peculiarities of the amino acid composition of a given IDP/IDPR or even the particularity of distribution of charges in its sequence could define the degree of IDP/IDPR responsiveness to changes in its environment. In other words, differently disordered parts of protein molecules can differently respond to a similar environmental cue, providing another level of complexity for these highly heterogeneous entities.

A very specific feature of IDPs/IDRs is the fact that these biologically active proteins/regions fail to form specific 3D structures and exist as highly dynamic structural ensembles, either at the secondary or at the tertiary level [7, 27, 32, 40, 56, 85, 86]. It is recognized now that IDPs/IDRs may possess extended disorder (where intrinsic disorder is present in a form of random coil or pre-molten globule), and collapsed disorder (where the intrinsic disorder is present in a molten globular form) [7, 86, 87]. Furthermore, proteins may also have regions of semi-disorder; i.e., fragments that have the same predicted probability to be ordered or disordered, and that play vital roles in protein aggregation and participate in protein-protein interactions, being able to undergo binding-induced folding [88]. More generally, IDPs/IDPRs are characterized by a very complex and heterogeneous spatiotemporal structural organization, possessing foldons (independent foldable units of a protein), unfoldons (ordered regions that have to undergo an order-to-disorder transition to become functional), semi-foldons (regions that are always in a semi-folded form), inducible foldons (disordered regions that can fold at least in part due to the interaction with binding partners), and non-foldons (non-foldable protein regions), [35, 89–91]. This intricate structural, mosaic-like “anatomy” of IDPs/IDPRs defines their unique molecular “physiology,” where differently (dis)ordered structural elements might have well-defined and specific functions [92], thereby defining a possibility of a protein molecule to be multifunctional and be involved in regulation of, interaction with, and be controlled by multiple structurally unrelated partners.

Overall, IDPs/IDPRs are complex systems with sophisticated structurally and functionally heterogeneous organization. They are uniquely placed at the core of the structure-function continuum concept, where instead of the classical (but heavily oversimplified) “one gene–one protein–one structure–one function” view, the actual protein structure-function relationship is described by the more convoluted “one-gene–many-proteins–many-functions” model [92, 93]. Therefore, intrinsic disorder represents a cornerstone of the proteoform concept [94] introduced to explain an important observation that the complexity of a biological system is primarily determined not by the genome size, but by its proteome size [95]. In fact, the number of functionally different proteins is known to dramatically exceed the number of protein-encoding genes (e.g., the human genome is approaching 20,700 genes [96], but the actual number of functionally different proteins is in the range of a few million). The increased size of a functional proteome over a corresponding genome is determined by multiple factors, ranging from the allelic variations and various pre-translational mechanisms affecting genes (e.g., production of numerous mRNA variants by the alternative splicing and mRNA editing) to numerous chemical changes induced in proteins by various post-translational modifications (PTMs) [97–101]. As a result, a single gene can efficiently encode for a set of distinct protein molecules, giving rise to the aforementioned proteoform concept [94]. Furthermore, in addition various genetic and non-genetic factors increasing chemical variability of a polypeptide chain, the protein structural diversity can be further increased by intrinsic disorder and functioning [93].

It is clear that internal dynamics is crucial for the biological activity of many (if not all) proteins, indicating that protein functionality requires at least some degree of conformational flexibility and structural dynamics [102–106]. In fact, even in ordered proteins, functional dynamics involve movements of not only individual amino acid residues or groups of amino acids in an active site relative to each, but also displacements of entire domains, and these function-related movements that happen in a wide spread of time-scales, ranging from femtoseconds to seconds, are required for the facilitation of catalytic activity [103, 104]. Furthermore, enzymes abundantly (in fact, as amply as non-enzymes) contain lengthy IDPRs with functions similar to those of IDPRs in non-enzymes (i.e., recognition, regulation, signaling, and control) [46]. Finally, it was pointed out that enzymes might possess evolutionary conserved dynamic traits, where, for example, RNase homologs with different biological functions, such as angiogenesis, anti-pathogenicity, immunosuppressivity, and highly efficient RNA cleavage, are not only grouped into evolutionarily distinct functional sub-families, but also show remarkable intra-subfamily conservation of dynamical properties evaluated by the quantitative characterization of the slowest modes of corresponding molecular dynamics (MD) simulations [107] that are not directly related to the conserved or semi-conserved features in the corresponding disorder profiles showing distribution of the per-residue disorder predisposition of these proteins [108]. These observations suggest that even for ordered globular proteins, in addition to the well-recognized structural conservation, which is a functionally important feature, functionality may also be linked to the evolutionarily conserved dynamical traits [107], which are only partially related to the conserved intrinsic disorder predisposition [108]. Therefore, such conservation of structural dynamics represents an important constituent of the evolutionary conservation of a functional protein fold [107, 108].

However, IDPs/IDPRs are different from ordered proteins and domains in terms of the amplitudes, time scales, and spatio-temporal coordination of their functional and structural dynamics. As a matter of fact, it is recognized now that the full structural description of IDPs/IDPRs is only feasible via the utilization of experimental [e.g., NMR-based chemical shifts (CSs)], residual dipolar couplings (RDCs), ¹⁵N R₂ relaxation rates, paramagnetic relaxation enhancement (PRE) distance restraints, J-couplings, pulsed field gradient (PFG)-derived R_h values, and ¹H-¹⁵N heteronuclear nuclear Overhauser effects (hetNOEs), complemented by SAXS [109, 110] and computational ensemble approaches (such as molecular dynamics simulations, or generation of a large number of conformations and selection of a subset that fits the experimental constrains [60, 111–113] that describe the structure of an IDP/IDPR as an ensemble of conformations and that can generate detailed information on short- and long-range contacts and backbone flexibility [114]. Because of their heteropolymeric nature, IDPs/IDPRs are rarely completely unstructured (i.e., behave as polymeric random coils) but often contain regions of transiently populated elements of secondary structure (so-called pre-structured motifs or PreSMos), which commonly serve as targets for their interaction partners to facilitate binding in a kinetically efficient manner [91, 92, 115–122]. Because of the recognition of the crucial importance of structural dynamics and flexibility for the functionality of IDPs and IDPRs, this field of protein science is currently undergoing a phase of explosive development, which precludes comprehensive consideration of this subject in a short overview article. Therefore, interested readers should look for dedicated review articles, of which there are many.

Over the years, understanding protein interactability has undergone a dramatic shift from a highly static to a very dynamic view. Everything started with the “lock-and-key” model of protein functionality that implied the crucial need of a spatially and temporally fixed structure of an active site [1, 123]. In this model, functional protein interaction with its partners resembles fitting a specifically shaped key into a unique lock, where only the substrate with the correct shape/size (key) is able to fit into the specifically shaped active site (key-hole) of the particular protein (lock), and where a high degree of geometrical precision in the substrate-protein interface complementarity defines their specificity of interaction [124]. Later, to account for changes in protein structure during the interaction process, the “induced fit” model was proposed [102], where a protein possesses some partial flexibility, and substrate binding induces some changes in the shape of the active site [125]. The next step was recognition of the presence of the “preexisting population, conformational selection, and population shift” mechanism [126–130], implying that, due to some flexibility, the native protein is better described as a conformational ensemble rather than as a unique rigid conformation [131], and further suggesting that interaction with a binding partner is based on the selection of the most binding-compatible conformation from this ensemble, thereby leading to shifting conformational equilibrium toward such an interaction-prone member of the conformational ensemble [131]. Finally, consideration of intrinsic disorder has opened seemingly unlimited opportunities of how a protein can feel and react to its environment, and how it can interact with its binding partners [91, 118].

In fact, despite their lack of stable structure, many IDPs/IDPRs are promiscuous binders that are never-nude, being always-complexed, invariably interacting with various partners via multiple binding scenarios [4, 7, 35, 117], and forming static, semi-static, and fuzzy or dynamic complexes [91, 118]. IDPs/IDPRs are capable of semi-static and dynamic polyvalent interactions [119], where one protein's multiple binding sites are simultaneously bound to another protein's receptors [132]. Some examples of such behaviors include bivalent and polybivalent scaffolds formed by two monovalent IDPs via interaction with a dimeric protein with two symmetrical binding sites or by self-association [133], IDP/IDPR wrapping around the binding partner [118], or highly dynamic shuffle complexes, where several binding sites of one protein are involved in concomitant, non-cooperative, tri-partite interaction with binding sites of another protein, thereby functioning similarly to holding a hot potato [134].

Many IDPs/IDPRs can fold to their partners at binding [29, 30, 91, 92, 115–135], with the degree of such binding-induced folding being different in various systems, thereby forming complexes with broad structural and functional heterogeneity [91, 118]. IDPs/IDPRs can act as molecular glue or mortar [29] by filling the gaps and cracks between the structural elements of a binding partner [136]. They can also serve as molecular epoxy, where partners undergo at least partial binding-induced folding at the complex formation [137–139]. Intrinsic disorder forms the basis of the dynamic “on-off” switch-type interactions commonly found in signaling networks, where IDPs/IDPRs bind partners with low affinity and high specificity [4, 140, 141]. Many IDPs/IDPRs serve as morphing shape-changers that are able to differently fold as a result of binding to different partners [56, 87, 135, 142–144], with the binding region of such a morphing IDP adopting completely different structures in the rigidified assemblies formed by the binding to divergent partners [28, 116, 117, 145, 146].

Some IDPs/IDPRs form fuzzy complexes, where significant disorder remains at least outside the binding interface [76, 147–153]. In such fuzzy complexes, under-folded bound regions are used as alternative contacts for specific partners and can also be fine-tuned by PTMs [147]. Furthermore, preservation of intrinsic disorder within dynamic signaling complexes provides disorder carriers with the capability to be engaged in entropic chain activities, which rely entirely on an extended random-coil conformation of a polypeptide that is able to carry out its function while staying flexible [4, 154]. Such dynamic signaling complexes with entropic chain activities represent stochastic machines [155] that contain structured domains connected by long flexible linkers [146, 155–157], and which work by stochastic, uncoordinated movements of the long disordered linkers (rather than by coordinated conformational changes) that move constantly and chaotically, but eventually enable productive functionality [155].

Although in the multitude of the aforementioned binding scenarios protein functionality seems to commonly originate from disorder-to-order transitions, many functions of ordered proteins were shown to rely on local or even global functional unfolding [89]. This regulated unfolding [158] or conditional disorder has an induced nature [159] and transient character [160], and may be initiated by several environmental factors, ranging from changes in temperature, pH, mechanical force, redox potential, or light exposure, to specific interactions with ligands, membranes, nucleic acids, or other proteins, to the release of autoinhibition, to various posttranslational modifications [89]. It was pointed out that such ordered proteins with dormant disorder serve as an important illustration of the global importance of intrinsic disorder for protein function.

Recently, another important property of IDPs/IDPRs pertaining to their foldability-interactability relationship was uncovered, namely the ability of IDPs/IDPRs to form tight complexes (with the affinity approaching picomolar levels [161]) without gaining any ordered structure and retaining long-range flexibility and highly dynamic character [161–163]. Formation of such extremely disordered complexes with ultrahigh affinity was described for oppositely charged proteins, and was proposed to rely on long-range electrostatic attraction between the charge-rich regions of the polypeptide chains, instead of being dependent on some structurally defined binding sites or some specific interactions between individual residues [161].

Therefore, IDPs/IDPRs clearly act as interaction specialists with an astonishing arsenal of binding modes, utilizing which can lead to the formation of a broad variety of complexes with remarkable structural and functional diversification.

Being a complex, “edge of the chaos” system, IDPs/IDPRs are characterized by emergent behavior based on the intricate self-organization processes, leading to the appearance of unanticipated novel structures, patterns, and properties [35]. One of the well-studied examples of such self-organized emergent behavior is given by the spatiotemporal oscillations of the Min protein system (MinD, MinC, and MinE) that moves from pole to pole of the rod-shaped Escherichia coli cells, and which is related to the spatial regulation of the positioning of the cytokinetic Z ring that determines the division plane at the middle of the cell [164, 165]. Min system oscillates in vivo with the intrinsic wavelength comparable to the size of the E. coli cell [166]. This is illustrated by Figure 4A representing a model of the pole-to-pole oscillations of these proteins in a rod-shaped system with a length of 3.8 μm [167]. Oscillations in the Min system originates from the ATP-dependent interactions of Min proteins with each other and with the cytoplasmic membrane [164], where the MinC/MinD oscillations between the ends of the cell are regulated by MinE acting as the “Tarzan of the jungle” while interacting with MinD [169]. A molecular mechanism of action of Min system is illustrated by Figure 4B, showing that the major regulator of oscillations, the MinE homodimer, is present in two functional forms, either a latent 6β-stranded conformation that diffuses in the cytoplasm or an active 4β-stranded form bound to the membrane and MinD [168]. The latent 6β-stranded form itself exists in two rapidly interconverting states, where the cryptic membrane-targeting sequences (MTSs) present at the N-tails of each protomer are dynamically associated with the β-sheet leading to the closed and open forms, with the open form being able to reversibly interact with the membrane via the released MTSs [168]. If the loop region connecting MTS with the remaining part of MinE in the membrane-bound open form of the 6β-stranded conformation encounters the membrane-bound MinD dimer, this sensor region undergoes a disorder-to-order transition, and a nascent α-helix is formed, which then propagates into the β₁-strand leading to the dissociation of this region from the MinE dimer interface. In the resulting 4β-stranded form, this new long α-helix protrudes from the surface of MinE dimer and is further stabilized by binding to the dimeric interface of MinD (see Figure 4C). Formation of the MinE-MinD complex leads to the stimulation of the MinD ATPase activity, ATP hydrolysis, MinD dimer dissociation, and the release of the MinE dimer. These released 4β-stranded active forms of MinE are either handed off to another membrane-bound MinD dimer, or dissociate from the membrane and fold back to the 6β-stranded structure, where the six-stranded β-sheet is sandwiched between the MTS on one side, and a coiled coil formed by two long antiparallel α-helices on the opposite side [170]. The probability of MinE to be handed off to the neighboring MinD dimer is rather high due to the high MinD density on the membrane [168, 169]. Figure 4D represents the PONDR® FIT-generated disorder profile of the E. coli MinE and shows that this protein is predicted to possess a noticeable amount of intrinsic disorder, especially in its N-tail containing the aforementioned MTS and conformational switch. This observation indicates that the intrinsically disordered nature of this region might be related to its capability to undergo structural transitions required for MinD binding [35].

The aforementioned oscillating behavior of Min system on bacterial and artificial membranes can be reproduced by MinD and MinE via MinD-driven recruitment of MinE to the membrane [171–173]. In fact, on the supported lipid bilayers in vitro, these Min proteins self-organize to form mesoscale patterns of traveling protein surface waves emerging from the repetitive binding-detaching cycles of proteins to the membrane [174, 175]. Besides the Min system, prokaryotes contain several other members of the WAKA protein family (Walker A cytomotive ATPase; also knows as ParA), which also possess oscillatory behavior and are related to regulation of development, spatial regulation of cell division, and segregation of chromosomes and plasmids [176–179]. For example, the intracellular localization of E.coli plasmids carrying par2 locus is determined by the interplay of the DNA binding protein ParB, the oscillating ATPase ParA, and the specific cis-acting DNA regions to which ParB binds [179]. Here, in the presence of ParB and aforementioned cis-acting DNA regions, ParA is able to form oscillating spiral-shaped structures, whereas in the absence of ParB and DNA, ParA was engaged in the formation of stationary helices that were extended from one end of the nucleoid to the other [179]. Furthermore, liquid-liquid phase separation-driven formation of various proteinaceous membrane-less organelles (PMLOs), which are commonly found in both prokaryotic and eukaryotic cells, formation of large protein clusters on the membrane surface, and liquid-gel phase transitions might represent other examples of the emergent behavior of intrinsically disordered proteins. These three scenarios are briefly considered below.

On the other side of the spectrum of disorder-based protein complexes described above are proteinaceous membrane-less organelles (PMLOs), also known as non-membranous cytoplasmic/nucleoplasmic granules, or cellular/nuclear micro-domains, or intracellular/ intranuclear bodies, that are commonly found in the nucleus and cytoplasm of various cells [180–184]. The dimensions of PMLOs are cell size–dependent, and they contain proteins and RNA (and/or DNA) [185]. These very large (detectable by light microscope), highly dynamic (but stable), and liquid-like assemblages are formed via the intracellular liquid-liquid demixing phase separations [186–189] and originate due to the colocalization of molecules at high concentrations within a small cellular or nuclear micro-domain [190, 191], leading to the liquid-liquid phase transitions (LLPTs) or the intracellular liquid-liquid demixing phase separation [185, 186, 188]. Since PMLOs are not covered by the membranes, their components are involved in direct contact and exchange with the nucleoplasm or cytoplasm [190, 191]. Furthermore, PMLOs are characterized by a liquid-like behavior, being capable of dripping, wetting, and relaxing into spherical structures upon fusion [192–195]. These liquid-droplet phases of the nucleoplasm/cytoplasm [192–197] possess relatively low intrinsic density and viscosity that are comparable to those of the cytoplasm or nucleoplasm [198, 199].

Formation of PMLOs may be initiated by a number of factors, including changes in the concentrations of specific small molecules or salts, changes in the concentration of proteins undergoing phase separation, changes in the pH and/or temperature of the solution, changes in osmolarity, by the binding of these proteins to some definite partners, by various PTMs and alternative splicing of the phase-forming proteins, or by changes in other environmental conditions that affect the protein-protein or protein-nucleic acid interactions [185, 188, 200–202]. PMLOs, of which are many [181, 182] (see Figure 5), represent an intricate form of the disorder-based protein complexes [92, 182, 183, 187, 204, 205], which serve as important illustrations of emerging behavior, and which are highly dynamic in nature and are typically formed without any noticeable structural changes in the proteins undergoing LLPTs [206]. The structural integrity and biogenesis of PMLOs are both exclusively determined by protein–protein, protein–RNA, and/or protein–DNA interactions [207, 208], and the process of PMLO formation is highly controlled, completely reversible, and strongly condition-dependent [181, 182].

Proteins capable of biological phase separation are characterized by structural/sequence modularity, invariably contain high levels of intrinsic disorder, and as a result of their highly disordered nature, are able to participate in weak multivalent interactions [92, 182, 183, 204, 205]. These weak multivalent interactions range from the heterologous electrostatic attraction between the oppositely charged biological polymers, such as oppositely charged proteins or positively charged nucleic acids and proteins, to homologous interactions of the same protein molecules containing repetitive donor and acceptor domains/regions (e.g., multiple stretches of positively- and negatively-charged residues) needed for the multivalent binding [181, 182]. In addition to specific environmental cues, the PMLO biogenesis is regulated via various PTMs and alternative splicing of the phase-forming proteins [188, 209]. Since many of the PMLO resident proteins are IDPs, and since the formation of all the PMLOs analyzed so far relies on IDPs/IDPRs, it is clear that intrinsic disorder is needed for PMLO biogenesis [188, 210].

In addition to liquid-liquid phase transitions, some proteins can (at least in vitro) undergo liquid-gel phase separation leading to the formation of hydrogels [211]. Here, instead of liquid-like PMLOs, the resulting hydrogels cannot flow under steady-state conditions [212–214], likely because they are composed of uniformly polymerized amyloid-like fibers [213]. Importantly, fibers found in these hydrogels are different from the pathological fibrils associated with numerous human diseases in at least two ways. In fact, hydrogels are highly dynamic systems that can be reversibly formed and disassembled in the response to some environmental cue (e.g., addition of specific small molecule or phosphorylation of a protein undergoing phase separation), and they are formed via heterotypic polymerization [213]. This is in stark contrast to pathological amyloids and amyloid-like fibrils, which are characterized by an extremely high conformational stability (once formed, they almost never disintegrate), and formation of which typically represents a homotypic polymerization reaction that specifically engages molecules of the same protein. Importantly, similar to the LLPT, the dynamic liquid-gel phase separation depends on multivalent interactions between proteins with low complexity (LC) domains, many of which are known to be intrinsically disordered. Examples of systems undergoing dynamic liquid-hydrogel phase separation include heterotypic polymerization of the LC domain of the fused in sarcoma (FUS) RNA-binding protein with RNA [213]; polymerization of mutant FUS forms associated with amyotrophic lateral sclerosis [215]; RNA-dependent hydrogel formation of the LC domains of RBM3, hnRNPA1, hnRNPA2, CIRBP, and the yeast Sup35 protein [213, 216]; RNA-dependent hydrogel formation of the LC domains of Ewings sarcoma (EWS) and TAF15 proteins [212]; and Phe and Gly (FG)-rich repeats of nucleoporins, e.g., yeast nucleoporin Nsp1p [214].

Although the aforementioned liquid-liquid and liquid-gel phase transitions were described in three-dimensional solutions, it was pointed out that the dynamic interactions between the multivalent cytoplasmic tails of transmembrane proteins and their multivalent binding partners can trigger the formation of large (at least micron-sized) two-dimensional protein clusters on the membrane surface [217]. This possibility was illustrated by the system that included the phosphorylated cytoplasmic domain of Nephrin and its intracellular targets, Nck and N-WASP [217]. Although in a three-dimensional solution, these three proteins phase separated into dynamic micron sized liquid droplets when critical protein concentration (that depended on the valency and affinity of interacting species) in solution was achieved [197], attachment of phosphorylated Nephrin to supported lipid bilayers of DOPC in the presence of Nck and N-WASP resulted in the formation at membranes of the micron-sized concentrated puncta containing all three proteins [217]. Furthermore, these phase-separated two-dimensional protein clusters were able to successfully promote actin filament assembly via the Arp2/3 complex recruited to the membrane through binding N-WASP, and were themselves remodeled by the resultant filament network [217]. These observations suggest that the multivalent protein interactions leading to phase separation can be responsible for regulation and control of some signaling pathways via generation of spatially organized micron-scale protein clusters [217].

These observations indicated that multivalency-induced polymerization and phase separation can occur in three-dimensional solutions and in two-dimensional systems. Importantly, computational analysis revealed that all three members of this system contain high levels of intrinsic disorder. In fact, more than 60% of residues in the C-terminal cytoplasmic tail of human Nephrin (PMID: O60500, amino acids 1077–1241) are predicted to be disordered. In one study, a rat neural Wiskott-Aldrich syndrome protein (N-WASP, PMID: O08816) construct containing residues 183–193 fused to 273–501 was predicted to be completely disordered. Finally, more than 35% of human cytoplasmic protein NCK1 (PMID: P16333) with C139S, C232A, C266S, and C340S mutations are predicted to be disordered as well.