Regulation of Cell Division by Intrinsically Unstructured Proteins; Intrinsic Flexibility, Modularity and Signaling Conduits

Abstract

It is now widely recognized that intrinsically unstructured (or disordered) proteins (IUPs, or IDPs) are found in organisms from all kingdoms of life. In eukaryotes, IUPs are highly abundant and perform a wide range of biological functions, including regulation and signaling. Despite increased interest in understanding the structural biology of IUPs/IDPs, questions remain regarding the mechanisms through which disordered proteins perform their biological function(s). In other words, what are the relationships between disorder and function for IUPs? There are several excellent reviews that discuss the structural properties of IUPs/IDPs since 2005(1-3). Here, we briefly review general concepts pertaining to IUPs and then discuss our structural, biophysical, and biochemical studies of two IUPs, p21 and p27, which regulate the mammalian cell division cycle by inhibiting cyclin-dependent kinases (Cdks). Some segments of these two proteins are partially folded in isolation and they fold further upon binding their biological targets. Interestingly, some portions of p27 remain flexible after binding to and inhibiting Cdk2/cyclin A. This residual flexibility allows otherwise buried tyrosine residues within p27 to be phosphorylated by non-receptor tyrosine kinases (NRTKs). Tyrosine phosphorylation relieves kinase inhibition, triggering Cdk2-mediated phosphorylation of a threonine residue within the flexible C-terminus of p27. This, in turn, marks p27 for ubiquitination and proteasomal degradation, unleashing full Cdk2 activity which drives cell cycle progression. p27, thus, constitutes a conduit for transmission of proliferative signals via post-translational modifications. The term “conduit” is used here to connote a means of transmission of molecular signals which, in the case of p27, correspond to tyrosine and threonine phosphorylation, ubiquitination and, ultimately, proteolytic degradation. Transmission of these multiple signals is enabled by the inherent flexibility of p27 which persists even after tight binding to Cdk2/cyclin A. Importantly, activation of the p27 signaling conduit by oncogenic NRTKs contributes to tumorigenesis in some human cancers, including chronic myelogenous leukemia (CML) (4) and breast cancer (5). Other IUPs may participate in conceptually similar molecular signaling conduits and dysregulation of these putative conduits may contribute to other human diseases. Detailed study of these IUPs, both alone and within functional complexes, is required to test these hypotheses and to more fully understand the relationships between protein disorder and biological function.

INTRINSICALLY UNSTRUCTURED (or DISORDERED) PROTEINS

Many proteins, which play a wide range of biological roles, either entirely lack secondary and/or tertiary structure, or possess long segments that lack secondary and/or tertiary structure, under physiological conditions (3, 6-8). These are commonly termed intrinsically unstructured (or disordered) proteins, abbreviated IUPs (or IDPs). Bioinformatics analyses of whole genome sequences using disorder predictors (9, 10) indicated that 6-33% of proteins in bacteria and 35-51% of proteins in eukaryota contain disordered regions of 40 or more consecutive residues (9, 11). The greater abundance of IUPs in the latter was proposed to be due to the greater need for protein-mediated signaling, regulation and control that is required in eukaryotes (11). It is now widely recognized that IUPs play broad biological roles in all kingdoms of life (11, 12).

Functions of Disordered Proteins

IUPs are involved in many cellular functions, including regulation of cell division, transcription and translation, signal transduction, protein phosphorylation, storage of small molecules, chaperone action, transport, and regulation of the assembly or disassembly of large multi-protein complexes, amongst many others (13, 14). Indeed, the majority of transcription factors (15) and proteins involved in signal transduction (16) in eukaryotes are predicted to be disordered or contain long disordered segments. Further, 79% of human cancer-associated proteins (HCAPs) have been classified as IUPs, compared to 47% of all eukaryotic proteins in the SWISS-PROT database (16). The latter observation highlights the importance of disorder in the function of proteins that regulate processes often dysregulated in cancer such as cell proliferation, apoptosis, and DNA repair.

Although many IUPs function by folding into an ordered conformation upon binding their biological targets, for many others, disordered conformations mediate biological function. For example, disordered segments serve as linkers in many IUPs (e.g. between folded domains in multi-domain proteins such as p53 (17)). In other cases, IUPs function as entropic bristles (e.g. NF-M and NF-H which serve a repulsive spacers in neurofilaments (18)), springs (e.g. titin, which induces passive tension in muscle filaments) (19)) and semi-permeable barriers (FG-domains within nucleoporins of the nuclear pore complex (20)). In these examples, disordered polypeptide segments often occur in conjunction with other folded or partially folded domains.

Amino Acid Composition and Primary Structure of IUPs

The sequences of intrinsically unstructured proteins exhibit low complexity compared to those of globular proteins (21). Further, IUPs are depleted in hydrophobic amino acids (i.e. Val, Leu, Ile, Met, Phe, Trp and Tyr) and correspondingly enriched in polar and charged amino acids (i.e. Gln, Ser, Pro, Glu, Lys, Gly and Ala) relative to globular proteins (21). Consequently, IUPs occupy a different region of “charge-hydrophobicity space” compared to globular proteins (22) and lack tertiary structure because they possess too few hydrophobic residues to independently form a stable hydrophobic core. The term “charge-hydrophobicity space” refers to a two-dimensional graph of values of the mean net charge (y-axis) and mean hydrophobicity (x-axis) of proteins (22). Uversky and co-workers showed that the vast majority of IUPs exhibit values of these parameters which lie to the left on such a graph, and folded proteins to the right, of a boundary line (22). These sequence-based features have led to the development of algorithms that identify disordered segments within protein sequences with up to 80% accuracy (10).

While disordered segments within proteins can be reliably identified using these algorithms, it remains difficult to differentiate between disordered segments that, for example, play structural roles as linkers and those that mediate protein-protein interactions through folding upon binding. Such algorithms would need to be trained to recognize the sequence features of polypeptide segments with these distinct functions. While some progress has been made toward the goal of identifying interaction sites within IUPs (23, 24), only when a database of sequence-structure/disorder-function relationships based on detailed analysis of the conformational and functional properties of a wide variety of functional complexes of disordered proteins is established will more discriminating predictions of disordered protein function be possible.

How disordered are IUPs; Evidence for Partially Populated Secondary Structure within Certain IUPs

Structural analysis of IUPs is challenging because their polypeptide backbones exhibit a high degree of flexibility and disorder due to rapid inter-conversion amongst multiple conformers. This causes measurable parameters such as far UV circular dichroism (CD) and NMR chemical shift values to be dynamically averaged amongst the inter-converting states. Early studies of IUPs were performed using CD, NMR and/or structural mapping using limited proteolysis (25-27) and the results suggested that IUPs adopt conformations very similar to those of random coils. However, CD studies of the IUPs p21^Cip1 (p21) (25) and p27^Kip1 (p27) (28, 29) showed that these proteins possess nascent α-helical secondary structure. Detailed analysis using NMR spectroscopy revealed that several different segments of isolated p27 adopt secondary structure in solution; these were termed “intrinsically folded structural units”, or IFSUs (30). In fact, NMR studies by many laboratories have now shown that IUPs exhibit varied degrees of nascent structure and disorder. Some IUPs completely lack secondary and tertiary structure (partial list in Table 1A) while others exhibit partial secondary structure (Table 1B).

Table 1.

Structural features of IUPs which either (A) lack or (B) exhibit some degree of secondary structure as determined at individual residue resolution using NMR spectroscopy.

Protein Name	Protein length (# of amino acids)	Boundaries of unstructured region(s) (residue numbers)	Structural observations (type of secondary structure observed (residues))	Reference (s)
A. Proteins which lack secondary structure under physiological conditions
β-Dystroglycan segment derived from dystroglycan precursor	893	654-750	Highly disordered	(74)
IA3, inhibitor of yeast proteinase A	68	1-68	Highly disordered	(75)
Myelin basic protein, Golli isoform BG21	194	1-194	Highly disordered	(76)
N protein of phage λ	107	1-53	Highly disordered	(77)
p19Arf tumor suppressor	169	1-37	Highly disordered	(78)
SNARE (soluble NSF attachment protein receptor) protein, Snc1	117	Cytoplasmic domain, residues 1-94	Highly disordered	(79)
B. IUPs which exhibit nascent secondary structure under physiological conditions
α-Synuclein	140	1 - 140	α-helical propensity (∼10% α-helix:18-31), slight α-helical propensity (1-100), possible β-turn (110- 140)	(80, 81)
[β-synuclein and γ- synuclein]			[Data are also available for these alternative isoforms]	(81) and references therein; (82)
cAMP response element-binding (CREB)	341	101 - 120 (CREB kinase- inducible activation domain, pKID)	α-helix (119-130)	(83)
Cyclin-dependent kinase inhibitor p27Kip1	198	1-198, 22-105 (kinase inhibitory domain, KID), 105- 198 (C- terminal domain)	α-helix (37-59)	(29)
			α-helix (37-59), β- hairpin (65-75), single turn of helix (87-90)	(30)
			Highly disordered (105- 198)	(70)
Cyctic fibrosis transmembrane conductance regulator	1480	654-838 (R region)	α-helical propensity (>5% up to ∼30% α- helix: 654-668, 759-764, 766-776, 801-817), β- strand propensity (>5% up to ∼30% β-strand: 744-753)	(84)
Dynein intermediate chain, IC74	640	84-143, 198- 237	α-helical propensity (222-232), random coil (84-143)	(85)
Fibronectin binding protein	1018	745-874 (Fibronectin binding domains D1- D4)	α-helical propensity (773-778, 793-799, 811- 816, 831-837)	(86)
Histone mRNA Binding Protein (Stem-loop binding protein, SLBP)	276	1 - 175	α-helix (28-45, 50-57, 66-75, 91-96)	(87)
Merozoite surface protein 2 (MSP2), isoform of Plasmodium falciparum	221	1-221	α-helical propensity (14-21, 140-150)	(88)
Microtubule- associated protein tau, isoform tau-F	441	1-441	α-helix (253-267, 315- 328, 346-361), β-strand (307-312)	(89)
Negative regulator of flagellin synthesis (anti-sigma factor FlgM)	97	1 - 97	α-helix (60-73, 83-90)	(90)
Nrf2	597	1-98 (Neh2 domain)	α-helix (39-71), β- strand (74-76, 82-85)	(91)
Potassium channel shaker chain beta 1a	401	1 - 62	α-helix (2-10, 44-52, 56-61)	(92)
Retinal phosphodiesterase inhibitory γ-subunit	87	1-87	α-helical propensity (∼50% α-helix: 68-84)	(93)
Thymosin β4	44	1-44	α-helix (5-17)	(94)
Titin	834	425-452 (Elastic PEVK motifs)	Polyproline II helix (425-429, 438-442, 445- 449), highly disordered (430-437, 443-444, 450- 452)	(95)
Tumor suppressor protein p53	393	1-75 (trans- activation domain) 1-73	α-helix (18-26), nascent turn (40-44, 48-53) α-helix (18-24), mixture of α-helix, β-strand and random coil (39-59)	(96) (97)

Folding-Upon-Binding

While IUPs are disordered in isolation under physiological conditions, they often perform their biological functions by binding specifically to other biomolecules through the process of folding-upon-binding. In general, folding-upon-binding reactions are enthalpically driven to overcome the accompanying large and unfavorable entropies of binding, as shown for protein-DNA interactions (31) and protein-protein interactions (29, 32). Due to the extended nature of many IUPs which fold upon binding their targets, the magnitudes of both the favorable enthalpy change for binding (ΔH) and unfavorable entropy change for binding (ΔS) are approximately proportional to the length of the disordered polypeptide segment involved in binding (29). This allows a range of different size binding sites to be targeted by IUPs through evolutionary tuning of the binding favorability and structural complementarity of IUPs and the protein surfaces they target. While the loss of conformational freedom due to folding upon binding (ΔS_conf) is entropically unfavorable, it is partially compensated by the entropically favorable release of bound water molecules (ΔS_HE) upon binding of an IUP to a protein surface (the hydrophobic effect). While some segments of the polypeptide backbone of IUPs involved in specific protein-protein interactions may become rigid after folding upon binding, other segments may remain dynamic within complexes (33), mitigating to some extent the unfavorable ΔS_conf. Further, the methyl groups of either IUPs and/or their binding targets, that mediate inter-molecular hydrophobic interactions, may experience motional restriction to different extents upon binding, providing an additional mechanism for modulating ΔS of binding (34). These two mechanisms allow tuning of the affinity of interactions (ΔG) through evolutionary variation of the associated entropy changes. Consequently, the values of dissociation constants (K_d) observed for IUPs binding their biological targets span a wide range, from low nanomolar values (tight binding; e.g. p27 binding to Cdk2/cyclin A (29)) to high micromolar values (weak binding; e.g. WASP binding to Cdc42 (35)). As a general rule, weak interactions involving IUPs involve relatively small amounts of buried surface area and tight interactions involve the burial of very large surfaces.

Functional Advantages of Disorder

The inherent flexibility of IUPs is thought to confer certain functional advantages over more highly structured proteins. First, some IUPs bind specifically to more than one biological target and thus exhibit diverse biological functions, often with involvement in signaling and regulation (36). For instance, p21 binds and regulates the catalytic activity of several cyclin-dependent kinase (Cdk)/cyclin complexes, an early example of “binding promiscuity” (25). Further, p21 and p27 bind additional partners in both the cell nucleus and cytoplasm, extending their functions to include regulation of apoptosis, cell motility and transcription ((37) and references therein). Another example is the p53 tumor suppressor protein. While the DNA binding and tetramerization domains are folded, the N-(residues 1-97) and C-terminal domains (residues 363-393) of p53 are intrinsically unstructured in isolation and mediate interactions with numerous binding partners that modulate p53 activity in diverse ways ((38) and references therein). Promiscuous binding activities allow p53 to regulate diverse cellular processes such as cell division, apoptosis and DNA repair (38). Finally, it has been suggested that IUPs are specialized to function as hubs in protein interaction networks due to their propensity for promiscuous interactions (39); however, the generality (40) and validity (41) of this general concept has been challenged.

Second, because a large fraction of residues within IUPs are solvent exposed, even within multi-protein assemblies, these sites are accessible for post-translational modification (PTM), allowing control of protein function, localization and turnover. For example, the majority of known phosphorylation, acetylation and ubiquitination sites in p53 occur within the disordered N- and C-terminal domains and modification of these sites alters the function, localization and turnover of p53 (42). PTM sites are often clustered within disordered polypeptide segments, affording accessibility not only to modification enzymes but also to other proteins that interact specifically with the modified sites to transduce biological signals. An example of this is the phosphorylation/ubiquitination cascade that regulates p27 function (43).

Third, disordered polypeptide segments within proteins are often highly susceptible to proteolytic cleavage in vitro, and this may be a factor which influences the rate of IUP degradation in cells. However, a recent study by Tompa and co-workers of >3,000 yeast proteins showed that protein disorder was a poor predictor of the in vivo rate of protein turnover; hence, while it is intuitively obvious that polypeptide disorder is associated with proteolytic susceptibility, protein degradation in vivo is highly regulated and influenced by many other factors (44). For example, Shaul and co-workers discovered that p53 is degraded by the 20S proteasome via a “default” pathway, without the need for ubiquitination. These authors proposed that disordered segments of p53, and other proteins (45), are signals for 20S proteasome-mediated degradation and that the formation of multi-protein assemblies masks these signals and guards against degradation (46). This may represent a mechanism for sensing imbalances in the levels of subunits within multi-subunit assemblies, allowing subunits present in excess to be degraded by default (45). Thus, the physical properties of disordered polypeptide segments allow proteins to be extensively regulated by PTM and provide the opportunity for rapid turnover and possibly quality control during assembly of multi-protein complexes.

Finally, the non-compact nature of IUPs may facilitate biomolecular interactions by increasing intermolecular association rates. Wolynes and co-workers (47) postulated that disordered proteins have a greater “capture radius” than compact, folded proteins. According to their so called “fly-casting” mechanism, a segment of an extended, unfolded protein first binds relatively weakly to the surface of a target, followed by folding to reel in the target. By being extended, IUPs sample larger solution volumes, in a sense reducing the dimensionality of the search for their partners (47). For example, p27 binds via a sequential mechanism to Cdk2/cyclin A, with an extended domain at the N-terminus binding first to a compact surface on cyclin A, followed by extensive folding of p27 and remodeling of Cdk2 as the inhibited p27/Cdk2/cyclin A complex is fully assembled (29). This fly-casting-like mechanism may facilitate assembly of Cdk/cyclin complexes under the low concentration conditions found in cells. Another example of this mechanism was revealed recently by Wright and co-workers (48) in studies of the phosphorylated KID (pKID) domain of CREB binding to the KIX domain of the CREB binding protein (CBP). Intrinsically unstructured pKID was shown to initially dock non-specifically on the surface of KIX, allowing rapid searching of the KIX surface for the specific binding site, followed finally by folding into the specific, high-affinity complex (48). IUPs often exhibit many, compact recognition elements throughout their sequences; the conformational disorder of IUPs may facilitate the assembly of multi-protein complexes by bringing together their many components via the fly-casting mechanism. Not only does intrinsic disorder promote association events, but Hilser and Thompson have proposed that linking ligand binding with disordered domain folding provides a mechanism for optimizing allosteric coupling in multi-domain proteins (49).

INTRINSICALLY UNSTRUCTURED PROTEINS IN MAMMALIAN CELL CYCLE REGULATION

In eukaryotes, cyclin-dependent kinases (Cdks) are the master time keepers of cell division (50). Many proteins regulate the Cdks, both directly and indirectly, and, in turn, the catalytic activity of the Cdks regulates the activity of myriad downstream targets (51, 52). While many of these regulatory proteins are folded, many others are intrinsically unstructured. In this review, we focus on a small subset of these IUPs: the cyclin-dependent kinase regulators (CKRs) p21, p27 and p57^Kip2 (p57) (51) that regulate cell division through direct interactions with Cdk/cyclin complexes (Fig. 1A). Through binding promiscuity (25), the CKRs regulate Cdk/cyclin complexes that control 1) entry into G₁ phase (Cdk4 and Cdk6 paired with D-type cyclins) and 2) progression from G₁ to S phase (Cdk2 paired with A- and E-type cyclins) (51). Further, p21 and p27 exhibit functional diversity by having seemingly opposite effects on these different Cdk/cyclin complexes, promoting the assembly and catalytic activity of some (e.g. Cdk4 paired with D-type cyclins) and potently inhibiting others (e.g. Cdk2 paired with A- and E-type cyclins) (51). In the following sections, we discuss results from our laboratory and others on the structural and dynamics features of the CKRs and the relationship of these features to their diverse biological functions.

Figure 1. Regulation of the eukaryotic cell division cycle.

(A) Illustration of the various stages of the cell division cycle and the Cdk/cyclin complexes that play roles in regulating progression through the different stages are indicated. Initiation of cell division in G₁ phase requires the activity of Cdk4/cyclin D and Cdk6/cyclin D and progression into S phase (when DNA synthesis or replication occurs, “S”) requires Cdk2/cyclin E and Cdk2/cyclin A (and similar complexes with Cdk1). The activity of Cdk1/cyclin B and Cdk1/cyclin A are required for entry into mitosis (“M”). While initially thought to be universal inhibitors of these Cdk/cyclin complexes, p21 and p27 have been show to activate Cdk4/cyclin D and Cdk6/cyclin D under certain circumstances (indicated by arrow). (B) Alignment of sequences of the kinase inhibitory domains (KID) of p27, p21 and p57. The boundaries of sub-domains D1 (blue), LH (black), D2 (red) and 3₁₀ (green) are indicated, as is the “RxL” motif which is recognized by cyclin A. (C) Illustration of the domain structure of p21, p27, and p57. PCNA, PCNA binding domain; NLS, nuclear localization signal; QT, QT domain which contains one or more QT motifs that are either known or putative phosphorylation sites. PAPA, domain with multiple repeats of PAPA motif. The locations of known phosphorylation sites, Y74, Y88 and T187 in p27 and T310 in p57, are also indicated.

Domain structure of CKRs

The CKRs p21, p27 and p57 contain a conserved, 60 residue-long N-terminal kinase inhibitory domain (KID, residues 28-90 in p27) (Fig. 1B) and nuclear localization signals (NLSs) (53, 54) within their C-terminal domains (Fig. 1C). In addition, p21 and p57 contain a PCNA-binding domain within their C-termini that, when bound, inhibits the ability of PCNA to stimulate DNA synthesis (55, 56). Further, p27 and p57 possess a C-terminal QT domain that contains a critical threonine residue (T187 in p27 and T310 in p57) that, when phosphorylated by Cdk2, triggers ubiquitination of p27 (57, 58) and p57 (59) by SCF/Skp2. Human and mouse p57 have an additional domain comprised of multiple Pro-Ala repeats and mouse p57 contains a segment rich in acidic residues; these domains were proposed to mediate protein-protein interactions (60).

The structure of p27 bound to Cdk2/cyclin A

Pavletich and co-workers showed that the KID of p27 binds in a highly extended conformation to both subunits of the Cdk2/cyclin A complex (Fig. 2), burying over 2,000 Ų of solvent exposed surface (61). Several sub-domains within p27-KID, which possess many residues conserved among p21, p27 and p57 (Fig. 1B), adopt secondary structure in the Cdk2/cyclin A-bound state (Fig. 2): sub-domain D1, containing the conserved “RxL” motif (62), binds in an extended conformation on the surface of cyclin A; sub-domain LH forms a 22 residue-long α-helix that spans the nearly 40 Å gap between Cdk2 and cyclin A; sub-domain D2 forms a β-hairpin and an intermolecular β-sheet (with Cdk2) on the surface of the N-terminal lobe of Cdk2; and sub-domain 3₁₀ forms a 3₁₀-helix that inserts into the ATP binding pocket of Cdk2. p27 inhibits Cdk2 1) by inserting sub-domain 3₁₀ into the active site and blocking access to ATP and 2) by remodeling the catalytic cleft through displacement of a β-strand of Cdk2 by sub-domain D2 of p27 (61). Further, sub-domain D1 of p27 blocks the substrate binding site on cyclin A which recognizes Cdk2 substrates possessing a RxL motif (62). These results can be extended to generally understand how the CKRs regulate the Cdks that control the G₁/S transition during cell division (29, 63). However, due to the inherent limitations of X-ray crystallography this structural model does not provide insights into the role of p27's intrinsic flexibility in recognizing and binding to Cdk/cyclin complexes.

Figure 2. Structure of p27-KID bound to Cdk2/cyclin A.

Crystal structure of p27-KID/Cdk2/cyclin A determined in 1996 by Pavletich and co-workers. The sub-domains of p27, D1, LH, D2 and 3₁₀ are indicated using the color scheme of Fig. 1.

The CKRs are IUPs

Analysis using proteolysis, CD and NMR spectroscopy showed that p21, p27 and p57 are largely disordered, with ∼15-20% α-helical content (25, 28, 29, 64). Secondary structure (29) and disorder prediction (using FoldIndex (65), IUPred (66) and PONDR (21)) also indicated that these proteins are predominantly disordered (Fig. 3 shows results for p27). 2D ¹H-¹⁵N HSQC NMR spectra for p21, p27 and p57 exhibited limited resonance dispersion (backbone amide protons resonate between 7.8 and 8.5 ppm), a feature typical of intrinsically unstructured proteins (25, 29, 64). However, despite being disordered, these proteins were shown to be potent inhibitors of various Cdks in vitro (67-69) and in vivo (51). As early as 1996, structural data for p21 (25) and the previously identified heat-resistant nature of p27 (69) strongly suggested that polypeptide disorder was associated with the biological function of the CKRs.

Figure 3. Disorder predictions for p27.

Three disorder prediction programs were used to analyze the structure of p27, including FoldIndex (A), IUPred (B) and PONDR (C). The approximate locations of sub-domains D1, LH, D2 and 3₁₀ are indicated at the top, as is the QT domain.

The CKRs exhibit partially populated secondary structure in solution

Although p21, p27 and p57 can be categorized as IUPs based on their lack of tertiary structure, CD spectra indicated the presence of a small amount of α-helical secondary structure (25, 28, 29, 64) within the KID of each protein (25, 28, 64). Subsequently, we used NMR spectroscopy to localize this secondary structure within p27-KID. Analysis of secondary ¹³C_α chemical shift values (Δδ¹³C_α) indicated that the small amount of α-helical secondary structure observed using CD was localized to sub-domain LH (∼30% α-helical, Fig. 4A) while other sub-domains of p27-KID appeared to lack secondary structure (29). To probe the dynamics of p27-KID, we measured {¹H}-¹⁵N heteronuclear NOE (hetNOE) values; this NMR relaxation parameter is sensitive to fluctuations of amide groups on the high picosecond to low nanosecond timescale. The hetNOE values observed for p27-KID (Fig. 4B) indicated that the polypeptide backbone experiences intermediate dynamics, being less rigid than a folded protein but more rigid than a random coil. Interestingly, partially restricted motions were observed not only for sub-domain LH, but also for most residues in sub-domains D2 and 3₁₀. In contrast, residues within sub-domain D1 (residues 27-35) exhibited negative or near zero hetNOE values, consistent with a high degree of flexibility.

Figure 4. NMR results on the solution structure and dynamics of isolated p27-KID in solution.

(A) Secondary ¹³C_α chemical shift values (δΔ ¹³Cα) and (B) {¹H}-¹⁵N heteronuclear NOE (hetNOE) values for p27-KID. (C-F) Patterns of ¹H_N-¹H_N NOEs which indicated the existence of partial secondary structure (as indicated) within sub-domains LH (C), D2 (D and E), and 3₁₀ (F) of p27-KID. A and B are taken from Lacy, et al., 2004 (29) and C-F are taken from Sivakolundu, et al., 2005 (30).

More recently, we probed the structure and dynamics of p27-KID using a novel approach that utilized amide proton-amide proton (¹H_N-¹H_N) NOE data from NMR spectroscopy to provide insights into time-averaged structure and molecular dynamics (MD) computations to provide insights into how structure fluctuates with time (30). Interestingly, these studies revealed that several segments of p27-KID exhibited discreet structures which we termed intrinsically folded structural units (IFSUs) (Fig. 4C-F). The IFSUs occurred within sub-domains LH, D2 and 3₁₀, which also exhibited positive hetNOE values. Sub-domain LH adopted helical structure, as expected from chemical shift analysis, while subdomain D2 formed a β-hairpin and nascent helical structure and sub-domain 3₁₀ formed a single turn of helix. These results indicated that p27-KID is quite rich in transient, discrete structures, in contrast to the picture which first comes to mind for an IUP. Importantly, with the exception of the nascent helical segment of sub-domain D2, these structural features resemble those observed when p27-KID is bound to Cdk2/cyclin A.

Finally, we recently characterized the solution structure of the ∼100 residue-long C-terminal domain of p27 using NMR spectroscopy (70). This domain contains several phosphorylation sites, including T187 mentioned earlier and threonine 157 (T157) within the NLS which is phosphorylated by Akt in breast cancer (71), and several lysine residues that are likely sites of ubiquitination. Earlier results from CD and NMR suggested that this segment of p27 is highly disordered (29); these results were confirmed by the recent NMR studies which showed that this domain lacks tertiary and secondary structure on the basis of Δδ¹³C_α and hetNOE values (70).

The CKRs Fold Sequentially Upon Binding Specific Cdk/cyclin Complexes

In 1996, Kriwacki and Wright demonstrated that sub-domains D2 and 3₁₀ of p21 folded upon binding to Cdk2 using NMR spectroscopy and, through proteolytic mapping, deduced that sub-domain D1 bound to cyclin A within the Cdk2/cyclin A complex (25). Later in the same year, Pavletich and co-workers demonstrated that p27-KID adopted a highly extended, folded conformation when bound to Cdk2/cyclin A (61). As noted earlier, several of the IFSUs detected in p27-KID prior to binding were also observed in the Cdk2/cyclin A bound state. For example, one part of sub-domain D2 and sub-domain 3₁₀ maintained their β-hairpin and helical conformations, respectively, when bound to Cdk2. Importantly, the single turn of helix observed for apo sub-domain 3₁₀ occupied the ATP binding site of Cdk2 within the ternary complex, with tyrosine 88 (Y88) bound in place of the purine ring of ATP (61). Sub-domain LH formed a 22 residue long α-helix that linked sub-domains D2 and D1, which were bound to Cdk2 and cyclin A, respectively. Sub-domain D1 exhibited a high degree of disorder and flexibility in the free state and adopted an extended, rigid conformation upon binding to a pocket on the surface of cyclin A that is conserved in many cyclins that regulate cell division (29, 61, 63). Finally, another part of sub-domain D2, which exhibited nascent helical features prior to binding, adopted an extended conformation and formed an inter-molecular β–sheet upon binding to the N-terminal domain of Cdk2.

While the crystal structure defined the Cdk2/cyclin A-bound conformation of p27 and provided key insights into the molecular basis of specific recognition of Cdk/cyclin complexes (61), it does not explain why the CKRs have evolved to be disordered or how disorder plays a role in their biological functions. Answers to these questions came from studies that probed the mechanism through which p27 binds to Cdk2/cyclin A (29). Isothermal titration calorimetry (ITC) was used to determine values of thermodynamic parameters (ΔG, ΔH and ΔS) associated with p27 binding to Cdk2/cyclin A and to quantitatively characterize the polypeptide folding which accompanies binding (31). Further, surface plasmon resonance (SPR) was used to analyze the kinetics of p27 association with and dissociation from Cdk2/cyclin A. Results from these two methods, coupled with our knowledge of structure and dynamics, indicated that the sub-domains of p27 bind to Cdk2/cyclin A via a sequential mechanism; first the highly flexible sub-domain D1 binds cyclin A, followed by docking of helical sub-domain LH and finally by docking and folding of sub-domains D2 and 3₁₀ to Cdk2 (Fig. 5). In addition, other studies (63) suggested that sub-domain D1 rapidly scans the surfaces of protein complexes for a conserved binding pocket, as found on cyclin A and other cyclins that regulate cell division. When sub-domain D1 encounters the cyclin pocket, p27 transiently binds, providing time for other sub-domains to sequentially dock and fold into the final, inhibited ternary complex with Cdk2/cyclin A. Amino acids that comprise the cyclin A docking site for p27 are highly conserved in Cdk/cyclin complexes that directly regulate cell division and are inhibited by p21 and p27. However, these residues are not conserved in Cdk/cyclin complexes involved in other biological functions. Therefore, we proposed that the intrinsic disorder of p21 and p27 evolved to allow specific molecular recognition through sequential folding upon binding (29, 63). The extended character of p27 when bound to Cdk2/cyclin A has evolved to accommodate the large distance (40 Ǻ) between the specificity determining site on cyclin A and the site of inhibition (ATP binding pocket) on Cdk2. It is possible that simultaneously engaging these two sites through interactions with the two ends of an extended polypeptide chain provided evolutionary advantages over an alternative scheme involving interactions mediated by multiple, folded protein domains.

Figure 5. The sub-domains of p27 sequentially fold and bind Cdk2/cyclin A.

Results from NMR, ITC and SPR support a scheme in which the “RxL” motif within sub-domain D1 of p27 binds first to cyclin A (“1”), followed by folding and docking of sub-domain LH, followed by binding of sub-domain D2 to Cdk2 (with extensive remodeling of Cdk2, followed finally by binding of sub-domain 3₁₀ in the ATP binding pocket of Cdk2 (“2”). This sequential scheme provides a mechanism for specificity toward Cdk/cyclin complexes which preserve the binding site for the “RxL” motif within sub-domain D1 of p27. A movie depicting this sequential binding scheme is available at http://www.stjuderesearch.org/data/kriwackilab/p27.html. Taken from Lacy, et al., 2004 (29).

Regulation of p27 Function through Post-translational Modification; p27 as a Molecular Signaling Conduit

Post-translational modifications regulate the localization, turnover, and activity of p27 ((72) and references therein). For example, Akt-mediated phosphorylation of T157 within p27's NLS in breast cancer cells prevents its interactions with the nuclear import machinery and leads to cytoplasmic localization. p27, normally located in the nucleus, encounters new targets in the cytoplasm and exhibits a gain of oncogenic function. In a further example, phosphorylation of p27 on Ser 10 promotes its interaction with the shuttling protein, CRM1, leading to export from the nucleus. Finally, cells entering the cell division cycle contain high levels of p27 and phosphorylation-dependent ubiquitination and degradation of p27 by the 26 S proteasome is required for progression through the cell division cycle.

p27 degradation is regulated by two E3 ubiquitin ligases during cell division, KPC1 in the cytoplasm in G1 phase and SCF^Skp2 in the nucleus in S and G2 phases ((43) and references therein). While KPC1 ubiquitinates unphosphorylated and free p27, SCF^Skp2 targets p27 that is phosphorylated on T187 and which is bound to Cdk2/cyclin E or Cdk2/cyclin A. Ubiquitinated p27 is degraded by the 26S proteasome. Active cyclin E/Cdk2 can phosphorylate cyclin/Cdk-bound p27 on T187. However, while free, active Cdk2/cyclin phosphorylates Cdk-bound p27 efficiently, structural and biochemical studies have demonstrated that p27-bound Cdk2 is catalytically inactive (29). This presented an apparent paradox since elimination of p27 appeared to require the activity of the enzyme (Cdk2) that it was known to potently inhibit. However, recent studies by Hengst, Kriwacki and co-workers (43) have demonstrated that phosphorylation of p27, within the p27/Cdk2/cyclin A complex, at sites other than T187 provided a solution to this conundrum.

Preliminary in vivo studies indicated that p27 was phosphorylated on tyrosine 88 (Y88) by non-receptor tyrosine kinases, including Abl (43), Lyn (43), and Src (5). Phosphorylation of residue Y88 on p27 bound to Cdk2/cyclin A relieved inhibition of Cdk2 and promoted Cdk2-mediated phosphorylation of T187 via a pseudo uni-molecular reaction mechanism (43). NMR studies showed that phosphorylation of Y88 (pY88) within the Cdk2 binding domain of p27 by Abl kinase led to ejection of the inhibitory 3₁₀-helix of p27 (sub-domain 3₁₀) from the ATP binding pocket of Cdk2 while leaving other interactions between p27 and Cdk2/cyclin A unperturbed (Fig. 6 and Fig. 7, “Step 1”). Surprisingly, Cdk2 retained significant catalytic activity even though pY88-p27 remained tightly bound to the Cdk/cyclin complex (43). Consequently, residue T187 within the intrinsically unstructured C-terminal domain of p27 could then be phosphorylated by the partially reactivated Cdk2 within the same pY88-p27/Cdk2/cyclin A ternary complex (43) (Fig. 7, “Step 2”). Although Cdk2 within this phosphorylated ternary complex exhibited sub-maximal catalytic activity, tethering of p27 to cyclin A/Cdk2 strongly promoted the phosphorylation of T187 through the uni-molecular mechanism (Fig. 7B). p27 that has been phosphorylated on both Y88 and T187 (pY88/T187-p27) can be polyubiquinated by the SCF^Skp2 ubiquitin ligase and degraded, resulting in complete reactivation of Cdk2/cyclin A. The accumulation of free, active cyclin A/Cdk2 may further promote Thr187 phosphorylation of p27 within remaining p27/Cdk2/cyclin A complexes and accelerate progression from G₁ to S phase of the cell division cycle.

Figure 6. Phosphorylation of p27 on Y88 (pY88) ejects only the 3₁₀-helix (sub-domain 3₁₀) from the ATP binding pocket of Cdk2; other sub-domains of p27 remain bound to Cdk2/cyclin A.

(A) NMR chemical shift differences (Δδ) between (A) free p27-KID and p27-KID bound to cyclin A/Cdk2 are compared with those for (B) free pY88-p27-KID and pY88-p27-KID bound to cyclin A/Cdk2. The locations of sub-domains within p27 are shown at the top, as is the amino acid sequence of p27-KID. The asterisk indicated the location of Y88. Amino acid residues are indicated using the single letter code. Note: For technical reasons, a mutant form of p27-KID, with Y89 mutated to F, was used in these experiments. Taken from Grimmler, et al., 2007 (43).

Figure 7. The p27 molecule is a signaling conduit.

(A) A single snap-shot from a 13 ns molecular dynamics trajectory illustrating the structure of p27 bound to Cdk2/cyclin A (cyan and magenta, respectively). The 100 residue-long C-terminal domain of p27 (yellow tube), which contains T187 (orange), is intrinsically unstructured and highly dynamic in this trajectory. Also illustrated are two critical tyrosine residues, Y74 and Y88 (red and green, respectively), which are phosphorylated by non-receptor tyrosine kinases (NRTKs). Phosphorylation of Y88 (“Step 1”), and possibly Y74, ejects sub-domain 3₁₀ from the ATP binding pocket of Cdk2 (indicate by white arrow), allowing T187 within the flexible C-terminal domain to encounter the Cdk2 active site (“Step 2”, indicated by grey arrow) and be phosphorylated by Cdk2. (B) Scheme illustrating the two-step p27 phosphorylation mechanism involving Y88 and T187 which triggers p27 poly-ubiquitination and 26S proteasomal degradation in both normal and cancer cells. The pseudo-uni-molecular nature of step 2 is illustrated.

We propose that the intrinsic disorder and flexibility of p27 are evolutionarily advantageous by enabling the structural fluctuations and post-translational modifications associated with the phosphorylation/poly-ubiquitination cascade that regulates p27 turnover at the G₁/S boundary during cell division. First, segmental flexibility between the sub-domains D2 and 3₁₀ allows Y88 within sub-domain 3₁₀ to fluctuate between ATP pocket-bound and solvent exposed conformations. This allows Abl and other NRTKs to access and phosphorylate otherwise occluded Y88. Notably, in addition to targeting Y88, Src also phosphorylates p27 on tyrosine 74 (Y74). This implies that the β-hairpin IFSU harboring Y74 within p27 sub-domain D2 (Fig. 7A) also fluctuates between bound and solvent exposed conformations so as to provide Src access to Y74. Second, after Y88 (and sometimes Y74) has been phosphorylated and sub-domain 3₁₀ has been ejected from the Cdk2 active site, the extreme flexibility of the p27 C-terminus allows T187 to approach the substrate binding site of Cdk2 and be phosphorylated. This flexibility also ensures that phosphorylated T187 is accessible for recognition by SCF^Skp2 which is bound to p27/Cdk2/cyclin A (or cyclin E). Third, the p27 C-terminus contains six lysine residues that are likely sites for poly-ubiquitination by SCF^Skp2. Intrinsic flexibility within this polypeptide segment will not only make these sites accessible for poly-ubiquitination, but also ensures that covalently linked poly ubiquitin chains are accessible for processing by the 26S proteasome and its various accessory proteins (73). Thus, the intrinsic flexibility of p27 critically mediates each step of this multi-step PTM cascade. In view of this, we have proposed that p27 acts as a molecular signaling conduit which integrates proliferative signals from NRTKs (through phosphorylation on Y74 and Y88), participates in the processing of these signals (through reactivation of Cdk2 and phosphorylation of T187) and finally transduces these signals by triggering its own poly-ubiquitination and degradation (through phospho-T187-dependent interactions with SCF^Skp2). While p27 exhibits modular structure within the kinase inhibitory domain when free in solution, the segments connecting these modules are highly flexible, allowing the individual modules to function independently as part of this signaling conduit. The flexibility and modularity which enable this structural independence allow signals to flow through the p27 conduit as a consequence of these sequential phosphorylation and ubiquitination modifications.

CONCLUSIONS

It is now well recognized that IUPs are highly abundant and that they play critical functional roles in biological systems, with many mediating signaling and regulation. Bioinformatics studies have dramatically increased awareness of IUPs and their properties. However, structural, biophysical, and biochemical studies of IUPs has progressed at a slower pace, creating gaps in our knowledge of the relationship between the physical properties of these proteins and their wide ranging biological functions. Our studies of the cell cycle regulators, p21 and p27, have revealed many interesting, often unexpected, results that have established important concepts that are likely to apply to many other, uncharacterized IUPs. These concepts include the existence of both highly disordered and partially folded modules (IFSUs) within IUPs prior to interacting with their biological targets; the participation of these modules in highly specific, sequential binding events; the role of post-translational modification (PTM) of residues within these modules in regulating IUP function; and the linkage of multiple PTMs within individual IUPs into signaling conduits. While disorder prediction algorithms can accurately identify proteins, or protein segments, that are likely to exhibit extensive disorder, because relatively few IUPs have been structurally and dynamically characterized at atomic resolution, it is not yet possible to empirically predict the detailed structural features of IUPs. Several of the IFSUs identified within p27 play critical roles in interactions with Cdk/cyclin complexes; just as critical, however, is flexibility between these modules so that their functions can be regulated independently, for example, by PTMs. A challenge for the future is to characterize in detail additional IUPs, both with regard to their detailed structural features but also with regard to the relationship(s) of these features to biological function. Multidisciplinary approaches are required to fully understand the physical properties of IUPs, for example, using NMR spectroscopy complemented with computational (MD), hydrodynamic (AUC) and scattering (SAXS) methods, as illustrated here for p27. When a larger body of biophysical data on IUPs is available in the future, we will gain a deeper understanding of relationships between their structural and dynamic features and biological functions, and may be better positioned to broadly predict structure/function relationships for IUPs. As additional data is gathered, we will more clearly see that functionally relevant protein “structures” span a continuum which extends from near complete disorder (e.g. FG-domains within nucleoporins (20)), to partial order (e.g. IFSUs within p27), to partial disorder (e.g. disordered N- and C-terminal domains of p53 (38)), and finally to near complete order (e.g. many enzymes). We hope that this detailed review of the relationships between protein disorder and function for p21 and p27, and the experimental approaches used to gain these insights, will motivate others to undertake similar, detailed studies of other biologically important IUPs. Only through detailed study of the large number of intrinsically unstructured proteins evident in higher organism will we begin understand why they have evolved to be ubiquitous.

Abbreviations

AUC

analytical ultracentrifugation

IDP

intrinsically disordered protein

IFSU

intrinsically folded structural unit

IUP

intrinsically unstructured protein

KID

kinase inhibitory domain

MD

molecular dynamics

NOE

nuclear Overhauser effect

PTM

post-translational modification

SAXS

small-angle X-ray scattering