Functional Anthology of Intrinsic Disorder. III. Ligands, Postranslational Modifications and Diseases Associated with Intrinsically Disordered Proteins

Abstract

Currently, the understanding of the relationships between function, amino acid sequence and protein structure continues to represent one of the major challenges of the modern protein science. As much as 50% of eukaryotic proteins are likely to contain functionally important long disordered regions. Many proteins are wholly disordered but still possess numerous biologically important functions. However, the number of experimentally confirmed disordered proteins with known biological functions is substantially smaller than their actual number in nature. Therefore, there is a crucial need for novel bioinformatics approaches that allow projection of the current knowledge from a few experimentally verified examples to much larger groups of known and potential proteins. The elaboration of a bioinformatics tool for the analysis of functional diversity of intrinsically disordered proteins and application of this data mining tool to >200,000 proteins from Swiss-Prot database, each annotated with at least one of the 875 functional keywords was described in the first paper of this series (Xie H., Vucetic S., Iakoucheva L.M., Oldfield C.J., Dunker A.K., Obradovic Z., Uversky V.N. (2006) Functional anthology of intrinsic disorder. I. Biological processes and functions of proteins with long disordered regions. J. Proteome Res.). Using this tool, we have found that out of the 711 Swiss-Prot functional keywords associated with at least 20 proteins, 262 were strongly positively correlated with long intrinsically disordered regions, and 302 were strongly negatively correlated. Illustrative examples of functional disorder or order were found for the vast majority of keywords showing strongest positive or negative correlation with intrinsic disorder, respectively. Some 80 Swiss-Prot keywords associated with disorder- and order-driven biological processes and protein functions were described in the first paper (Xie H., Vucetic S., Iakoucheva L.M., Oldfield C.J., Dunker A.K., Obradovic Z., Uversky V.N. (2006) Functional anthology of intrinsic disorder. I. Biological processes and functions of proteins with long disordered regions. J. Proteome Res.). The second paper of the series was devoted to the presentation of 87 Swiss-Prot keywords attributed to the cellular components, domains, technical terms, developmental processes and coding sequence diversities possessing strong positive and negative correlation with long disordered regions (Vucetic S., Xie H., Iakoucheva L.M., Oldfield C.J., Dunker A.K., Obradovic Z., Uversky V.N. (2006) Functional anthology of intrinsic disorder. II. Cellular components, domains, technical terms, developmental processes and coding sequence diversities correlated with long disordered regions. J. Proteome Res.). Protein structure and functionality can be modulated by various posttranslational modifications or/and as a result of binding of specific ligands. Numerous human diseases are associated with protein misfolding/misassembly/ misfunctioning. This work concludes the series of papers dedicated to the functional anthology of intrinsic disorder and describes ~80 Swiss-Prot functional keywords that are related to ligands, posttranslational modifications and diseases possessing strong positive or negative correlation with the predicted long disordered regions in proteins.

Introduction

The functionalities of many proteins are modulated by various types of posttranslational modifications (PTMs). Chemical modifications of a polypeptide chain after its biosynthesis extends the range of functions of the protein. Typically, PTMs are classified according to the involved mechanisms: addition of functional groups (e.g., acylation, alkylation, phosphorylation, glucosylation, etc.); attachement of other proteins and peptides (e.g., ubiquitination, SUMOylation, etc.); changing of the chemical nature of amino acids (deamidation, deimidation, oxidation, etc.); and dissection of the backbone by proteolytic cleavage. Some proteins require different types of posttranslational modifications for their function. One dramatic example of such proteins is provided by histones, which require methylation, acetylation, phosphorylation, ubiquitylation, ADP-ribosyation, and SUMOylation at different stages, with different modifications affecting histone–DNA interactions and also the histone–histone interfaces, thus providing the capacity to disrupt intra-nucleosomal interactions and to alter nucleosome stability.¹ Although the N-terminal domains of the core histones are known to contain an extraordinary number of sites that can be subjected to PTM,^2–4 over 30 histone modifications have been recently identified in the core domains too.^5–7

We are showing here that, in addition to the known classification based on the molecular mechanisms of PTMs, the conformational state of the site where the modification would take place can also be used to group PTMs in two major classes: the first group involves modifications that are associated primarily with structured proteins and regions, whereas the second group combines modifications that are associated primarily with intrinsically disordered proteins and regions. The first group of modifications is crucial for providing moieties for catalytic functions, for modifying enzyme activities or for stabilizing protein structure. They include formylation, protein splicing, oxidation and covalent attachment of quinones and organic radicals. The second group of modifications relies on the low affinity, high specificity binding interactions between a specific enzyme and a substrate (a protein that has to be modified); i.e., they clearly represent typical signaling interactions. Among the second group of PTMs are phosphorylation, acetylation, acylation, adenylylation, ADP ribosylation, amidation, carboxylation, formylation, glycosylation, methylation, sulfation, prenylation, ubiquitination, and Ubl-conjugation (i.e., covalent attachment of ubiquitin-like proteins, including SUMO, ISG15, Nedd8, and Atg8).

Similarly, natural ligands play a number of roles in the stabilization of proteins and in the modulation of their structures. In fact, during the course of their biological function, proteins undergo different types of structural rearrangements ranging from local to large-scale conformational changes. These changes are often triggered by protein interactions with low-molecular-weight ligands or with larger macromolecules. The interactions with natural ligands can significantly affect protein structure, a phenomenon widely used by nature. Particularly, this phenomenon represents a basis for the functioning of several cofactors. These interactions often result not only in evident local changes in the vicinity of the binding site, but also in global conformational changes. The hemoglobin-oxygen complex is a textbook example of such systems.^8–11 As a result of O₂ binding, the Fe²⁺ is shifted into the plane of the porphirine ring approximately by 0.6 Å. This slightly moves the hystidine residue, which is involved in the coordination of the iron ion. In its turn, the consequence of such local changes is an essential conformational reorganization of the whole hemoglobin molecule, accompanied by a large-scale movement (10–15° rotation of the subunits relative to each other).^{12, 13}

The structure forming and stabilizing effects of such natural ligands such as electrons, oxygen, water, metal cations, hem, fatty acids and other organic compounds were recently reviewed for different proteins.¹⁴ The range of possible structural transformations induced in a protein by ligand release is very wide, from a negligible decrease in the conformational stability to a complete protein unfolding. Furthermore, these structural alterations virtually do not depend on the nature of ligand. Comparing the structural properties and the conformational stabilities of several proteins, ligand-free forms were divided on seven major classes:¹⁴

Class I. The structural characteristics of apo- and holo-forms coincide; the release of ligands does not lead to the noticeable changes of the unique protein structure; only the conformational stability and the function are changed.

Class II. A detectable change in the 3D-structure takes place in the ligand-free form. However, the protein is still characterized by a pronounced tertiary structure while its secondary structure virtually does not change.

Class III. The apo-form of the protein possesses all the properties of the molten globule. In other words, tertiary structure completely disappears upon release of the ligands, the secondary structure is retained, and the protein molecule remains compact.

Class IV. The ligand-free form of the protein has specific features of the pre-molten globule; it has no rigid tertiary structure, its secondary structure is noticeably diminished, and its dimensions distinctly differ both from the random coil and also from the holo-protein dimensions.

Class V. The apo-form adopts a virtually completely unfolded polypeptide chain, similar to so-called random coil.

Class VI. The interaction with ligands induces large-scale movements of large parts (domains or subunits) of the protein molecule. These effects are frequently observed for the allosteric enzymes.

Class VII. Ligand binding destabilizes native protein conformation.

In agreement with this classification, we show here that many of the Swiss-Prot functional keywords related to the ligand binding are positively and negatively correlated with long protein regions predicted to be intrinsically disordered. In other words, using the bioinformatics tool developed in the first paper of this series¹⁵ some natural ligands are shown to interact preferentially with ordered proteins, whereas others prefer intrinsically disordered proteins.

Finally our analysis reveals that many diseases are strongly correlated with proteins predicted to be disordered. Contrary to this, we did not find disease associated proteins to be strongly correlated with absence of disorder. Below we describe some illustrative examples obtained from a literature survey. As previously, we were able to find at least one illustrative, experimentally validated example of functional disorder or order for the vast majority of functional keywords related to PTMs, ligand binding and diseases.

Materials and methods

The first paper of this series described the assembly of the dataset for the bioinformatics analysis to find correlation between long disordered regions and Swiss-Prot keywords.¹⁵ A bioinformatics approach was also presented that determines which functional keywords are over- or under-represented by proteins predicted to contain long (>40 consecutive amino acids) disordered regions. Using this approach, 196,326 Swiss-Prot¹⁶ proteins longer than 40 amino acid residues have been analyzed. The redundancy of Swiss-Prot database¹⁷ was reduced by the Markov Cluster Algorithm,¹⁸ which was used to group Swiss-Prot proteins into 27,217 families according to sequence similarity. Each of the analyzed proteins was annotated with at least one of the 875 functional or structural Swiss-Prot keywords.

Long disordered regions in Swiss-Prot proteins were predicted using the VL3E predictor¹⁹. Each of the 196,326 Swiss-Prot proteins was labeled as putatively disordered if it contained a region with more than 40 consecutive amino acids predicted by VL3E to be intrinsically disordered; proteins predicted not to contain such long disordered regions were labeled as putatively ordered.

The probability, P_L, that VL3E predicts a disordered region longer than 40 consecutive amino acids in a SwissProt protein sequence of length L was estimated as the fraction of putatively disordered SwissProt proteins with lengths between 0.9L and 1.1L. TribeMCL clustering was used to reduce effects of sequence redundancy in estimation of P_L, as described previously.¹⁵ Swiss-Prot keywords associated with disorder-(or order-) correlated functions were determined as those that contain a significantly larger (or smaller) fraction of putatively disordered proteins than what would be expected by a random selection of SwissProt sequences with the same length distribution.¹⁵

Results

Ligands interacting with intrinsically disordered proteins

DNA- and RNA-binding proteins

A list of keywords associated with ligands interacting with intrinsically disordered proteins is present in Table 1. One can see that the majority of ligands listed in Table 1 are related to the functionality of proteins involved in signaling, regulation and recognition. In fact, we have already pointed out that intrinsic disorder is important for functionality of different DNA- and RNA-binding proteins. The previously discussed examples include transcription factors, histones, and other proteins involved in the DNA condensation and chromosome partition.¹⁵ Similarly, ribonucleoproteins and proteins involved in rRNA-binding such as ribosomal proteins were shown to possess significant intrinsic disorder.¹⁵ Recently it has been established that serine/arginine-rich (SR) splicing factors belong to a class of intrinsically disordered proteins.²⁰ Several translation factors are involved in multiple interactions including rRNA-, mRNA-, and tRNA-binding. Analysis of the solution structure of the Methanobacterium thermoautotrophicum aIF2 beta, which is the archaeal homolog of eIF2 beta, a member of the initiation factor eIF2 heterotrimeric complex, implicated in the delivery of Met-tRNA(i)(Met) to the 40S ribosomal subunit, revealed that this translation factor subunit is composed of an unfolded N terminus, a mixed alpha/beta core domain and a C-terminal zinc finger.²¹ Also, intrinsic disorder is crucial for many viral nucleoproteins, which are responsible for the encapsulation of genomic RNA within a helical nucleocapsid. Indeed, the C-terminal domain of the nucleoprotein (N_tail) of measles virus, which is involved in a number important functions, is intrinsically unstructured.^22–25

Table 1.

All (17) ligand keywords strongly correlated with predicted disorder

Keywords	Number of proteins	Number of families	Average sequence length	Z-score	P-value
DNA-binding	13224	1751	417.25	18.22	1
RNA-binding	9781	552	261.95	17.7	1
Zinc	10661	1135	521.74	13.49	1
rRNA-binding	5852	109	153.48	11.46	1
Metal-thiolate cluster	228	11	64.67	6.99	1
Actin-binding	817	98	785.55	6.41	1
Calmodulin-binding	503	58	986.49	6.35	1
Viral nucleoprotein	437	39	573.29	6.14	1
Heparin-binding	257	51	441.11	5.53	1
Growth factor binding	53	7	562.76	3.33	1
cGMP-binding	52	4	736.37	3.07	1
cGMP	107	7	661.02	2.96	1
Sialic acid	43	17	479.95	2.17	0.97
cAMP-binding	104	6	573.42	2.06	0.97
cAMP	180	16	556.52	2.03	0.99
IgG-binding protein	30	6	349.87	1.9	0.99
tRNA-binding	1046	22	313.94	1.78	0.96

Metal-thiolate clusters

Several proteins known as metallothioneins rely on metal-thiolate clusters in their functions. There are at least ten known closely related metallothionein proteins expressed in humans. Metallothioneins are small proteins (< 7 kDa), that are able to coordinate a diverse range of metals. These proteins are characterized by a lack of definable secondary structure, a high cysteine content (~30%), and a degeneracy in the remaining residues (e.g. predominance of cysteine, serine, lysine and no aromatic residues).²⁶ The folding of metallothioneins into the functional conformation is completely determined by the coordination of the corresponding metal ions.²⁷ The mammalian metallothioneins make economic use of the protein synthesis machinery, as the protein chain is just long enough to wrap around the mineral cores, and as indicated above there is hardly any secondary structure.²⁸ Finally, the role of zinc binding in the folding of the intrinsically disordered zinc finger domains, which are the members of the metallothionein family, and which are primarily implicated in DNA binding, was already emphasized.^{15, 29}

Actin-binding proteins

Thymosin β4 is a small 5-kDa actin-binding protein with a diverse range of activities, including its function as an actin monomer sequestering protein, an antiinflammatory agent, and an inhibitor of bone marrow stem cell proliferation. Thymosin β4 is a typical natively unfolded protein with extremely low level of ordered structure in solution.³⁰ It is believed that the flexible structure of thymosin β4 may facilitate the recognition of a variety of molecular targets, thus explaining numerous functions attributed to this interesting protein. Furthermore, it has been hypothesized that thymosin β4 has a unique integrative function that links the actin cytoskeleton to important immune and cell growth signaling cascades.³¹ Troponin, another actin-binding protein, is a part of the native tropomyosin, a complex located on actin filaments and involved in the contraction of striated muscle. Troponin consists of three components, each performing specific functions: troponin C binds Ca²⁺, troponin I inhibits the ATPase activity of actomyosin, and troponin T provides for the binding of troponin to tropomyosin.³² Intrinsic disorder is abundant in troponins. In fact, troponin C possesses a relatively flexible linker providing high relative mobility between the two globular domains.³³ Troponin I, a polar protein with a high excess of positively charged residues (calculated pI is ~9.9), is highly extended in its functional state.³⁴ Finally, troponin T is also a highly polar protein but its charged residues are unevenly distributed within the amino acid sequence: the N-terminal part including residues 1–59 is enriched in negatively charged residues, whereas the C-terminal part is enriched in positively charged residues.³² Similarly to troponin I, troponoin T has an extended, rod-like structure even being involved in the tertiary complex with troponins C and I.³⁴

Calmodulin-binding

Table 1 shows the Swiss-Prot keyword calmodulin-binding to be strongly correlated with predicted disorder. Calmodulin (CaM) binding is involved in various important signaling pathways.^35–37 Numerous structural studies revealed that CaM target proteins interact with CaM in a common manner.^38–40 In fact, the interactions between CaM and its binding targets involve disorder-to-order transitions for the CaM molecule: a flexible linker, which becomes structured upon complex formation, enables the two globular domains to wrap around the CaM binding target.^38–40 The helix-helix interactions within the two globular domains of CaM are not completely rigid, so the helix-helix packing interfaces in these domains vary in a manner that depends on the detailed interactions with the different CaM binding targets.^{37, 41} Finally, the CaM target-binding surface is rich in methionines, which adopt different configurations when CaM associates with CaM binding target having different sequences. The end result of this structural plasticity is that CaM binds to a very large number of different sequences with high affinity. Furthermore, recent studies revealed that intrinsic disorder is a crucial feature of the CaM-binding targets and can be used to improve the accuracy of the calmodulin binding target predictions.³⁵

Heparin-binding proteins

Many proteins with widely diverse structures and functions are capable of binding heparin.⁴² Analysis of heparin-binding sites (HBSs) revealed that they are often disordered. For example, there are two HBSs in the anticoagulant annexin V: HBS-1 is formed by two of the calcium-binding loops, and HBS-2 includes the N terminus and nearby loop and helix regions.⁴³ The conformational behavior and fibrillation of a Parkinson’s disease related natively unfolded protein α-synuclein⁴⁴ and Alzheimer’s disease-related natively denatured protein tau^{45, 46} are both strongly modulated by heparin binding.

Growth factor binding

Insulin-like growth factor binding proteins (IGFBPs) are carriers and regulators of the insulin-like growth factors. Analysis of the solution structure of the C-terminal domain of IGFBP-6 (residues 161–240) revealed that it has substantial flexible regions (e.g., three hightly disordered loops), which include 33 of 79 residues.⁴⁷

cGMP and cGMP-binding

Cyclic GMP (cGMP) is a second messenger that regulates various metabolic processes at the cellular level and that mediates the action of certain hormones. cGMP is synthesized from GTP by guanylate cyclase (GC) and is degraded to 5'-GMP by cyclic nucleotide phosphodiesterases (PDEs 1-6). cGMP-binding PDE from retinal rods plays a key role in the process of visual signal transduction. The PDE holoenzyme is a functionally inactive heterotetramer of αβγ₂ composition. The catalytically active dimer of two large homologous α- and β-subunits is reversibly inhibited by the small γ-subunit, PDEγ.⁴⁸ It has been established that PDEγ has little or no ordered secondary structure under physiological conditions in vitro, being characterized by hydrodynamic dimensions typical for the completely unfolded polypeptide of corresponding molecular mass. This suggests that PDEγ belongs to the family of natively unfolded proteins.⁴⁹

Sialic acids

The members of this family of acidic amino carbohydrates are derived from a nine-carbon monosaccharide and are components of mucoproteins and glycoproteins. Naturally occurring sialic acids possess large structural diversity, which explains their involvement in a variety of biologically important processes. The analysis of solution structure of ovine submaxillary mucin (OSM) by NMR spectroscopy revealed that the mucin is characterized by high internal segmental flexibility and it exists in solution as a random coil.⁵⁰ Similarly, bovine submaxillary mucin (BSM) was shown to be significantly disordered, as its far-UV CD spectrum is typical of a highly unfolded polypeptide chain.⁵¹

cAMP and cAMP-binding

Cyclic AMP (cAMP) is another second messenger used for intracellular signal transduction. cAMP is produced from ATP by adenylate cyclase and is decomposed into AMP by the enzyme phosphodiesterase. cAMP is known to act through cAMP-dependent protein kinase A (also known as cAPK or PKA), and also able to directly regulate ion channels and the Rap GTPase guanine exchange factors Epac 1 and 2. The cAMP-protein kinase A (PKA) pathway is an important intracellular signal transduction cascade activated by various stimuli. Activation/inhibition of this pathway is known to affect the transcriptional regulation of various genes through distinct responsive sites. In vertebrates, the best-characterized nuclear targets of PKA are the cAMP response element-binding (CREB) proteins. One of the CREB proteins is the activating transcription factor 2 (ATF-2).⁵² Solution structure analysis of the N-terminal transactivation domain of ATF-2 revealed that the in the absence of zinc this domain is completely disordered.⁵³ However, addition of zinc induced partial folding of the ATF-2 transactivation domain.⁵³ Furthermore, the N-subdomain (residues 19–54) is well-folded, whereas the C-subdomain (residues 55–106) is highly flexible and disordered.⁵³ Other illustrative examples of cAMP-binding proteins are the Epac1 and 2 (exchange protein directly activated by cAMP) protein, which are the guanine nucleotide-exchange factors (GEFs) that activate Rap GTPase upon binding to cAMP.⁵⁴ Rap1 is a small GTPase which serves as a hub protein involved in numerous aspects of cell adhesion, including integrin-mediated cell adhesion and cadherin-mediated cell junction formation.⁵⁵ The roles of intrinsic disorder in the functions of hub proteins have been recently recognized and supported by the additional bioinformatics analysis.^{56, 57}

IgG-binding protein

Some strains of the anaerobic bacteria Peptostreptococcus magnus express a multidomain IgG-binding protein L at their surface. Protein L interacts with variable IgG light chain domain through the five homologous repeats (B1-B5) located in the N-terminal part of the protein. The 3-D solution structure of the 76 amino acid residue long B1 domain was analyzed using NMR spectroscopy. The domain was shown to contain a 15 amino acid residue long disordered N-terminus followed by a folded portion consisting of an α-helix packed against a four-stranded β-sheet.⁵⁸

IgE-binding protein

Galectins are members of the lectin family of carbohydrate-binding proteins. So far 14 mammalian galectins have been identified, all containing a conserved carbohydrate-recognition-binding domain (CRD) of about 130 amino acids. The members of galectin family have been classified into three subtypes: the prototype group (galectins-1, -2, -5, -7, -10, -11, -13, and -14), which contains one CRD; the chimera group (galectin-3), that contains one CRD and a long N-terminal proline- and glycine-rich domain; and the tandem repeat group (galectins-4, -6, -8, -9, and -12), the members of which have two CRDs separated by an non-conserved linker sequence of up to 70 amino acids.⁵⁹ The vertebrate galectins were found in the cytoplasm and the nucleus, on the cell surface, and in the extracellular space. They are present in numerous cell and tissue types, and possess various functions.⁶⁰ Galectin-3 is a 29-to 35-kDa beta-galactoside binding protein containing a C-terminal CRD and a flexible N-terminal domain which is composed of 110–130 amino acids and contains multiple homologues repeats (7–14) with a consensus sequence Pro–Gly–Ala–Tyr–Pro–Gly, followed by three additional amino acids.⁶⁰ This protein is involved in mRNA splicing activity, control of the cell cycle, regulation of cell adhesion, the modulation of allergic reactions, and the binding of the advanced glycosylation end products (AGE). These multiple functions are exerted intra- and extracellularly due to the ability of galectin-3 to bind the β-galactoside residues of cell surface and matrix glycoproteins via the CRD domain and to also interact with intracellular proteins via peptide–peptide associations mediated by its N-terminus domain.⁶¹ Among numerous functions of the extracellular galectin-3 is its IgE binding and interaction with the IgE receptor, which provide the molecular means to accomplish the modulation of inflammation.⁶² Importantly, NMR analysis of hamster galectin-3 (residues 1–245) revealed that its N-terminus (residues 1–125) is mostly unfolded in aqueous media.⁶³

tRNA-binding

Ribosomal protein L16 is an essential component of the bacterial ribosome organizing the architecture of the aminoacyl tRNA-binding site in the ribosome 50 S subunit. Although in solution L16 forms an α+β sandwich structure combined with two additional β-sheets located at the loop regions connecting the two layers, its terminal regions, Met1-Asp25 and Asp138-Gln141, and the internal loop region Thr75-Glu91 do not possess any ordered structure.⁶⁴

Ligands interacting with ordered proteins

Iron homeostasis, heme and iron-sulfur or 4Fe-4S clusters

Table 2 lists keywords related to ligands interacting with ordered proteins. Strikingly, most of the ligands and cofactors listed in the Table 2 are related to enzyme function or to the oxygen/electron transport. These and other cofactors and ligands are recruited by protein molecules to extend the chemistry available within the active sites. For example, besides the well-known heme-containing proteins (cytochromes, myoglobin and hemoglobin), transferrins, lactoferrin, ceruloplasmin and ferritins, numerous other proteins are involved in iron homeostasis.⁶⁵ The vast majority of these iron-binding proteins are known to possess well-organized rigid structure. Similarly, iron-sulfur or 4Fe-4S cluster containing proteins are almost always highly organized, as these clusters are perhaps the most abundant and the most diversely employed cofactor.⁶⁶ The simplest iron-sulfur center is comprised of a single iron atom liganded within a polypeptide by four cysteine residues, whereas more common iron-sulfur clusters have two, three or four iron atoms coordinated to polypeptide residues and bridged by inorganic sulfide. Iron-sulfur clusters serve most prominently in redox centers involved in the electron transfer reactions, as well as in several dehydratases, biotin synthase and lipoate synthase.⁶⁶

Table 2.

Top 20 of ligand keywords strongly correlated with predicted order

Keywords	Number of proteins	Number of families	Average sequence length	Z-score	P-value
Iron	9417	641	345.12	−22.78	0
NADP	3878	231	403.78	−21.72	0
Flavoprotein	2926	228	461.23	−19.39	0
NAD	5864	333	365.4	−16.48	0
FAD	2128	184	496.26	−16.47	0
Iron-sulfur	2874	227	394.78	−13.99	0
Pyridoxal phosphate	2362	81	432.38	−13.93	0
Magnesium	6186	407	441.77	−12.92	0
Plastoquinone	260	11	372.09	−12.59	0
4Fe-4S	2097	165	416.69	−12.48	0
Heme	4617	212	322.96	−10.59	0
FMN	791	74	389.7	−9.85	0
Thiamine pyrophosphate	396	16	594.21	−9.84	0
ATP-binding	19639	687	568.44	−8.95	0
Nucleotide-binding	23736	780	545.67	−8.62	0
Molybdenum	232	32	626.79	−8.51	0
Nickel	446	72	320.5	−7.34	0
Manganese	1913	194	416.88	−7.19	0
Sodium	611	81	569.15	−6.69	0
Cobalt	560	82	363.38	−6.51	0

Flavoproteins

More than 25 flavoproteins (mostly enzymes, including different oxidases, oxidocyclases, dehydrogenases, methylhydroxylases and methylenehydroxylases) have been identified, in which the flavin adenine dinucleotide (FAD) or the flavin mononucleotide (FMN) is employed as a covalently bound cofactor.⁶⁷ Furthermore, many enzymes (e.g., different dehydrogenases, reductases, oxidases, etc.) use nicotinamide adenine dinucleotide (NAD) and nicotinamide adenine dinucleotide phosphate (NADP) as their cofactors.

Pyridoxal phosphate

Like the iron-sulfur clusters, the pyridoxal phosphate cofactor is very widely used. Indeed, the pyridoxal phosphate (PLP)-dependent enzymes catalyze a wider variety of reactions than those containing any other cofactor. These PLP-dependent enzymes include different amino-acid decarboxylases and aminotransferases, serine hydroxymethyltransferase, tryptophan synthase, glycogen phosphorylase etc.⁶⁸

Magnesium

Enzymes belonging to the enolase superfamily are characterized by the highly conserved active site carboxylate residues that bind an essential magnesium ion and mediate proton transfer reactions from the carbon acid substrate to the resulting enolate ion intermediate.^{69, 70} Structures were determined for members of the superfamily that catalyze eight different reactions, including enolase, mandelate racemase, muconate lactonizing enzymes I and II, D-glucarate dehydratase, D-galactonate dehydratase, o-succinylbenzoate synthases, L-Ala-D/L-Glu epimerases, and 3-methylaspartate ammonia lyase.^{69, 70}

Plastoquinone

Finally, plastoquinone is frequently found in proteinaceous complexes associated with the electron transport chain in the light-dependent reactions of photosynthesis. For example, cytochrome b₆f complex from the thermophilic cyanobacterium Mastigocladus laminosus and the green alga Chlamydomonas reinhardtii is a highly organized machine, which also includes such prosthetic groups as hemes (c-type, b-type, and a new x-type heme), [2Fe-2S] cluster, chlorophyll a, β-carotene and plastoquinone.⁷¹

Thiamine pyrophosphate (TPP)

TPP is the derivative of vitamin B1, which is a cofactor of different enzymes performing catalysis in pathways of energy production. High resolution crystal structures have been determined for several TPP-dependent enzymes, including 2-oxoisovalerate dehydrogenase,⁷² branched-chain α-ketoacid dehydrogenase,⁷³ bacterial⁷⁴ and human⁷⁵ pyruvate dehydrogenases, transketolase,⁷⁶ pyruvate decarboxylase,⁷⁷ benzoylformate decarboxylase,⁷⁸ acetohydroxyacid synthase,⁷⁹ pyruvate oxidase,⁸⁰ and pyruvate:ferredoxin oxidoreductase.⁸¹

Nucleotide-binding and ATP-binding

Nucleotide-binding in general, and ATP-binding in particular, are crucial events for the function of numerous well-folded and highly ordered nucleotide-binding enzymes. For example, different ATP- and NAD(+)-dependent DNA ligases, ATP-dependent RNA ligases and GTP-dependent mRNA capping enzymes catalyze nucleotidyl transfer to polynucleotide 5' ends via covalent enzyme-(lysyl-N)-NMP intermediates. These enzymes share a core tertiary structure, which is composed minimally of a nucleotidyltransferase domain and an OB-fold domain.⁸²

Molybdenum

While only a minor constituent of the earth's crust, molybdenum is easily available to biological systems due to the water solubility of its high-valent oxides. Molybdenium is an integral component of the multinuclear M center of nitrogenases⁸³ and is also found in the mononuclear active sites of a diverse group of enzymes that catalyze transfer of an oxygen atom either to or from a physiological acceptor or donor molecule, respectively.⁸⁴ Crystal structures of several molybdenium-containing enzymes, e.g., aldehyde oxidoreductase, xanthine oxidase, xanthine dehydrogenase, aldehyde oxidase, sulfite oxidase, nitrate reductase, DMSO reductase, etc., have been determined.^{84, 85}

Nickel

Nickel-iron hydrogenases ([NiFe] hydrogenases, or H₂ases) are the best studied members of the hydrogenase (hydrogen acceptor oxido-reductase) family of enzymes that metabolize the most simple of chemical compounds, molecular hydrogen. The X-ray structure of a dimeric [NiFe] hydrogenase revealed that the large subunit contains the bimetallic [Ni-Fe] active site, with biologically uncommon CO and CN ligands bound to the iron, whereas the small subunit contains three iron-sulfur clusters. During catalysis, the nickel atom is most likely responsible for a base-assisted heterolytic cleavage of the hydrogen molecule, whereas the iron atom could be redox active.⁸⁶

Manganese

Catalases, also known as hydroperoxidases, are one of the most studied classes of enzymes. These proteins catalyze the degradation of two molecules of hydrogen peroxide to water and oxygen. Although the most widespread in nature and the most extensively characterized class of catalases is the monofunctional heme-containing enzymes, several hydroperoxidases are the non-heme, or manganese-containing enzymes (also known as dimanganese catalases). High resolution crystal structures of two manganese-containing catalases revealed that the catalytic center of these homohexameric enzymes is a dimanganese group.⁸⁷

Sodium

The major ion of the extracellular fluid is sodium. Transport of sodium ions across different membranes is crucial for life. In fact, production of electrical impulses in living organisms or tissues requires synchronized opening of transmembrane Na⁺ channels possessing a sodium selectivity-filter, a high-throughput ion-conductance pathway, and a voltage-dependent gating function.⁸⁸ For example, voltage-gated sodium channels (VGSCs), which are highly ordered transmembrane proteins, are important for the generation and propagation of rapid electrical signals in electrically excitable tissues such as muscle, heart, and nerve.⁸⁹ The epithelial sodium channels (ENaCs), transmembrane proteins, which are composed of three partly homologous subunits, α, β and γ, inserted into the membrane with a proposed stoichiometry of 2α:1β:1γ, are crucial for the control of sodium fluxes in epithelial cells.⁹⁰ Mutations in genes encoding voltage-gated sodium channels have been correlated with numerous inherited human disorders affecting skeletal muscle contraction, heart rhythm, and nervous system function.⁸⁹ Similarly, the role of ENaC in the overall control of sodium balance, blood volume and thereby of blood pressure is demonstrated by genetic disorders of sodium-channel activity (such as Liddle's syndrome), a rare inheritable form of hypertension associated with gain-of-function mutations in the β and γ subunits of the ENaC.⁹¹

Cobalt

The functional activities of several crucial enzymes depend on cobalt. For example, the catalytic activity of adenosylcobalamin-dependent isomerases proceeds through the formation of free radical intermediates generated by homolysis of the cobalt-carbon bond of the coenzyme.⁹² Members of the one of the three classes of ribonucleotide reductases (RNR), the enzymes responsible for the conversion of the four standard ribonucleotides, to their 2'-deoxyribonucleotide counterparts, thus providing the precursors needed for both synthesis and repair of DNA, are cobalt-containing enzymes. These proteins utilize a cobaltous cofactor, adenosylcobalamin, a vitamin B₁₂ derivative, that interacts directly with an active site cysteine to form the reactive cysteine radical needed for ribonucleotide reduction.⁹³

Posttranslational modification and intrinsically disordered proteins

Posttranslational modification keywords that are strongly correlated with intrinsic disorder are listed in Table 3. Many sites of posttranslational modifications of the kinds given in Table 3 have been experimentally associated with regions of intrinsic disorder.⁹⁴

Table 3.

All (17) posttranslational modifications keywords strongly correlated with predicted disorder

Keywords	Number of proteins	Number of families	Average sequence length	Z-score	P-value
Phosphorylation	10895	1663	592.18	27.12	1
Cleavage on pair of basic residues	867	192	194.23	14.07	1
Amidation	836	457	81.19	10.83	1
Ubl conjugation	806	154	466	8.65	1
Myristate	681	71	559.39	6.97	1
Glycoprotein	16202	1993	540.4	6.84	1
Sulfation	245	72	483.69	6.42	1
Proteoglycan	189	27	748.35	6.37	1
Lipoprotein	4337	633	393.84	5.99	1
Prenylation	723	47	265.25	5.37	1
Heparan sulfate	48	10	673	4.7	1
Gamma-carboxyglutamic acid	106	25	254.63	3.86	1
Covalent protein-DNA linkage	26	8	474.74	3.37	1
Methylation	1417	101	328.37	2.85	1
GPI-anchor	590	146	460.29	2.73	1
ADP-ribosylation	150	11	437.05	1.98	0.98
Pyrrolidone carboxylic acid	795	272	223.71	1.74	0.96

Phosphorylation

Protein phosphorylation is known to represent an important regulatory mechanism in eukaryotic cells. At least one-third of all eukaryotic proteins are estimated to undergo reversible phosphorylation.⁹⁵ Phosphorylation modulates the activity of numerous proteins involved in signal transduction, and regulates the binding affinity of transcription factors to their coactivators and DNA, thereby altering gene expression, cell growth and differentiation.⁹⁶ Recently it has been reported that amino acid compositions, sequence complexity, hydrophobicity, charge and other sequence attributes of regions adjacent to phosphorylation sites are very similar to those of intrinsically disordered protein regions.⁹⁷ These observations were employed to develop a new web-based tool for the prediction of protein phosphorylation sites, DISPHOS (DISorder-enhanced PHOSphorylation predictor, http://www.ist.temple.edu/DISPHOS).⁹⁷ Using this predictor in association with kinase substrate preference-based predictors such as ScanSite⁹⁸ appears to be a useful combination (work in progress).

Cytoplasmic domains of several immune receptors, which are members of the family of multichain immune recognition receptors (MIRRs) (e.g., T-cell receptors (TCRs), B-cell receptors (BCRs), and the high-affinity IgE receptor) represent an illustrative example of the functional importance of intrinsic disorder in protein phosphorylation.^99–101 MIRRS were shown to be intrinsically disordered.^{102, 103} They contain signaling subunits with the immunoreceptor tyrosine-based activation motif (ITAM). Phosphorylation of Tyr residues in these ITAMs takes place upon antigen binding and represents an early and obligatory event in the signaling cascade.^99–101

Protein phosphorylation sites are the substrates for specific enzymes, kinases.⁹⁴ Typically, binding of these substrates to their cognate enzymes is characterized by low affinity, and yet phosphorylation by each kinase is a highly specific process.¹⁰⁴ Combination of high specificity with low affinity, being ideal for signaling, can be achieved via the coupled binding and folding.¹⁰⁵ The low net affinity arises because the positive free energy associated with the disorder-to-order transition must be deducted from the magnitude of the negative free energy arising from the interactions within the contact interface. The usefulness of protein disorder for such high specificity/low affinity signaling interactions was pointed out almost 25 years ago.¹⁰⁵

Similar to phosphorylation, many other types of posttranslational modification including acetylation, acylation, adenylylation, ADP ribosylation, amidation, carboxylation, formylation, glycosylation, methylation, sulfation, and ubiquitination are controlled by specific enzymes.¹⁰⁶ Each of these examples is discussed below.

Lipidation

is a covalent attachement of fatty acids to the protein moiety. This type of posttranslational modification includes myristoylation, S-prenylation, and S-palmitoylation. Myristylation is the N-terminal attachment of a myristoyl lipid anchor to a glycine residue. This posttranslational modification regulates protein–membrane and protein–protein interactions. The modification is catalyzed by the enzyme N-myristoyltransferase.¹⁰⁷

GPI-anchor

Several membrane proteins contain a covalently linked lipid, glycosylphosphatidylinositol anchor (GPI-anchor), which tethers the proteins to the extracellular face of eukaryotic plasma membranes. GPI-anchored proteins are involved in a number of functions ranging from enzymatic catalysis to adhesion.¹⁰⁸ Both biosynthesis of GPI precursors and posttranslational protein modification with GPI occur in the endoplasmic reticulum. Upon GPI modification, the carboxyl-terminal signal peptide is split off from the protein and the resulting new carboxyl-terminal is then combined with the amino group of ethanolamine residue in the GPI precursors. The whole process of cleavage and GPI attachment is catalyzed by the GPI-transamidase complex.¹⁰⁹

Proteoglycans

These represent a special class of glycoproteins whose protein cores are heavily glycosylated. Proteoglycans consist of a core protein with one or more covalently attached glycosaminoglycan chain(s) (e.g., heparin and heparan sulfate). These glycosaminoglycan chains are long, linear carbohydrate polymers that are negatively charged under physiological conditions. Glycosylation of the proteoglycan occurs in the Golgi apparatus in multiple enzymatic steps, where one sugar unit is added at each step.¹¹⁰ In a more general form, a glycoprotein composed of a polypeptide and a carbohydrate (or oligosaccharide, which could be glucose, glucosamine, galactose, galactosamine, mannose, fucose and sialic acid). There are two types of glycosylation: N-glycosylation, where protein is modified at asparagines; and O-glycosylation, originating from the glycosylation at hydroxylysine, hydroxyproline, serine or threonine. It has been established that the glycosylation is a site-specific enzymatic process that involve several specific enzymes. Glycosylated proteins are often disordered, which may help to explain the observation that although it is estimated that >50% of all eukaryotic proteins are glycosylated,^{111, 112} only ~5% of all PDB entries have attached glycan chains.¹¹³

Gamma-carboxyglutamic acid (Gla)

This amino acid contains a dicarboxylic acid side chain, which occurs in a number of calcium-binding proteins. Gla has been discovered in blood coagulation proteins (prothrombin, Factor X, Factor IX, and Factor VII), plasma proteins of unknown function (Protein C, Protein S, and Protein Z), and proteins from calcified tissue (osteocalcin and bone-Gla protein).¹¹⁴ Gla is synthesized by the post-translational modification of glutamic acid residues. This reaction is catalyzed by a specific enzyme, the vitamin K-dependent carboxylase.¹¹⁵ Osteocalcin possesses a highly flexible structure which may be essential for the function of the protein.¹¹⁶

Pyrrolidone carboxylic acid (PCA, pGlu)

This acid is formed either during the later stages of protein biosynthesis at the terminal phases of translation or as a post-translational event, just prior to cellular secretion of protein with amino-terminal pGlu.¹¹⁷ The pGlu moiety results from the cyclization of an amino terminal glutamyl or glutaminyl residue by a specific enzyme, glutaminyl cyclase.¹¹⁸ Many proteins and bioactive peptides exhibit an amino terminal pGlu residue, which subsequently minimizes their susceptibility to degradation by aminopeptidases.¹¹⁹

The binding of the above-mentioned enzymes to corresponding substrates represents high specificity/low affinity signaling interactions. The molecular mechanism of kinase interaction with the modified protein might serve as aprototype for these enzyme-driven posttranslational modifications that are specific for signaling and that involve the high specificity/low affinity interactions characteristic for their function. In fact, the validity of this assumption has been recently supported for protein methylation¹²⁰ and ubiquitination (Radivojak and Iakoucheva, work in progress). In addition to ubiqutin, a number of distinct ubiquitin-like proteins (Ubls) such as SUMO, ISG15, Nedd8, and Atg8, function as protein modifiers. Post-translational covalent attachment of Ubls, Ubl-conjugation, is critical for many cellular processes, including transcription, DNA repair, signal transduction, autophagy, and cell-cycle control.¹²¹ Ubl-conjugation cascades are initiated by activating enzymes, which also coordinate the ubls with their downstream pathways.¹²²

Posttranslational modification keywords associated with ordered proteins

Table 4 lists keywords related to posttranslational modifications that occur in structured or ordered proteins.

Table 4.

All (11) posttranslational modifications keywords strongly correlated with predicted order

Keywords	Number of proteins	Number of families	Average sequence length	Z-score	P-value
Quinone	449	17	380.29	−16.17	0
Organic radical	54	3	484.74	−8.76	0
Covalent protein-RNA linkage	108	4	2365.56	−5.63	0
TPQ	22	3	645.09	−5.35	0
PQQ	26	7	663.31	−4.21	0
Formylation	57	30	168.74	−3.66	0
Zymogen	1680	128	436.22	−3.22	0
Autocatalytic cleavage	325	51	502.94	−2.42	0
Protein splicing	84	30	1027.94	−2.08	0.03
Oxidation	32	14	295.63	−1.98	0.04
Hypusine	58	1	147.19	−1.94	0.02

Quinone

Several classes of enzymes that contain post-translationally modified amino acid residues have been recently discovered.^{123, 124} One of the best characterized examples of such enzymes is the copper-containing amine oxidases (CuAOs), which contain a covalently bound cofactor, 2,4,5-trihydroxyphenylalanine quinone (TPQ), which is derived from the modification of an endogenous tyrosine residue¹²⁵ and a single copper ion located in the active site. CuAOs have been found in different organisms ranging from bacteria to mammals. Crystal structures solved for several CuAOs from different sources revealed dimeric tightly folded complexes.¹²⁶

Organic radicals

The activities of several crucial enzymes depend on organic radicals covalently attached to the protein moiety. Three illustrative examples include the following: 1) ribonucleotide reductase which contains a stable organic free radical located on a tyrosine residue in the small subunit of the enzyme;¹²⁷ 2) S-adenosylmethionine radical enzymes (e.g., coproporphyrinogen-III oxidase and biotin synthase) in which S-adenosylmethionine (SAM) serves as a precursor to organic radicals, generated by one-electron reduction of SAM and subsequent fission to form 5'-deoxyadenosyl radical and methionine;¹²⁸ and 3) galactose oxidase in which a cofactor is derived from active-site amino acid residues via the autocatalytic formation of a thioether bond between Cys-228 and Tyr-272.¹²⁹

Covalent protein-RNA linkage

The linkage between RNA and protein is found in some viral proteins that are attached to the end of a replicating viral RNA and that are necessary for RNA replication. 52% similarity between the distribution of hydrophobic and hydrophilic residues in 188 residues of the genome-linked viral protein (VPg) cistron located in the central part of potato virus Y genome and the fragment of cytoplasmic malate dehydrogenase of known crystal structure, has been used to propose a 3-D structural model of VPg.¹³⁰

PQQ

Several quinoproteins are involved in the long-range interprotein electron transfer. Quinoproteins that possess pyrroloquinoline quinone (PQQ), tryptophan tryptophylquinone (TTQ), and cysteine tryptophylquinone (CTQ) are dehydrogenases.¹³¹ PQQ is tightly but non-covalently bound to the enzyme, whereas TTQ and CTQ are derived from amino acid residues of the polypeptide chain. Illustrative examples of a PQQ-dependent enzyme are methanol dehydrogenase (MEDH), which catalyzes the oxidation of methanol to formaldehyde,¹³² and soluble quinoprotein alcohol dehydrogenase, which is a monomeric enzyme with one PQQ and one c-type heme cofactor.¹³¹

Formylation

Another posttranslational modification of enzymes is provided by formylation. For example, according to the RESID database of protein modifications,¹³³ N-formyl-L-methionine is present in several cytochrome oxidases and phosphatidylserine decarboxylase. Crystal structures of several cytochrome c oxidases, containing N- formyl-L-methionine, have been solved.¹³⁴ N-formylglycine is found in N_α-formyl melittin and N-formyl-L-lysine is a part of some peptides from bee venom.^{133, 135}

Zymogen

Many important proteolytic enzymes are synthesized as zymogens, i.e., inactive precursors or proenzymes decorated with a prosequence-inhibitor that has to be cleaved to make the enzyme functionally active. This is done to ensure precise regulation of the proteolytic enzyme activity, which is an essential requirement for cells and tissues because proteolysis at the wrong time or place may be lethal.¹³⁶ The analysis of crystal structures of several zymogens revealed that the inhibition is achieved via a specific mode of propeptide interaction with the proenzyme whereby the prosequence covers the active site cleft in a non-productive orientation.¹³⁶ An absolutely different mechanism of inhibition has been described for the malarial aspartic proteinases (plasmepsins), which are produced from inactive zymogens, proplasmepsins, having unusually long N-terminal prosegments of more than 120 amino acids.¹³⁷ Comparison of the crystal structures of plasmepsin and proplasmepsin from Plasmodium vivax revealed that a dramatic refolding of the mature N-terminus and a large (18°) reorientation of the N-domain between P. vivax proplasmepsin and plasmepsin produce a severe distortion of the active site region of the zymogen relative to that of the mature enzyme.¹³⁷

Autocatalytic cleavage

This is an important step in activation of different enzymes, including proteolytic enzymes synthesized as zymogens and in the production of inteins via protein splicing (see above). Furthermore, autocatalytic cleavage is known to produce individual proteins from polyproteins synthesized by picornaviruses.¹³⁸ Finally, γ-glutamyltranspeptidase (GGT), a heterodimic enzyme that catalyzes the hydrolysis of γ-glutamyl bonds in γ-glutamyl compounds such as glutathione and/or the transfer of the γ-glutamyl group to other amino acids and peptidesis, is generated from the precursor protein through the posttranslational autocatalytic cleavage.¹³⁹ The crystal structure of GGT from E. coli has been recently determined at 1.95 Å resolution.¹⁴⁰

Protein splicing

This is an intriguing posttranslational modification, where an intervening protein sequence (intein, internal proteins) is self-catalytically excised out from a protein precursor and the two flanking sequences (N- and C-exteins) are ligated to produce two mature enzymes.^{141, 142} This phenomenon was first described in 1990 for the TFP1 gene product of Saccharomyces cerevisiae,¹⁴³ and several years later for the recA protein of Mycobacterium tuberculosis,¹⁴⁴ an archaeal DNA polymerase,¹⁴⁵ and for the production of Vma1p (a catalytic 70-kDa subunit of the vacuolar H⁺-ATPase) and VDE (a 50-kDa DNA endonuclease or VMA1 intein) from the nascent 120-kDa translational product of the VMA1 gene in Saccharomyces cerevisiae.¹⁴⁶ The crystal structure of VDE has been determined at 2.1 Å resolution.¹⁴⁶ Protein splicing is not unique for the VMA1 precursor. Most inteins consist of two domains: one is involved in autocatalytic splicing, and the other is an endonuclease that is important in the spread of inteins.¹⁴⁷ The mechanism of this process has been elucidated^{148, 149} based on the well-established chemistry for the non-enzymatic cleavage of asparagine residues in proteins resulting in the formation of succinimide. Recently, a database of protein splicing, InBase, has been compiled¹⁵⁰ and is currently maintained by the New England BioLabs (http://www.neb.com/neb/inteins.html).

Oxidation

Proteins can undergo different types of oxidation. These include carbonylation (an irreversible process that targets different amino acids including lysine, arginine, proline and threonine^{151, 152}), nitration of tyrosine¹⁵² and oxidation of methionine to methionine sulfoxide.¹⁵³ Methionine is easily oxidized by H₂O₂, hypochlorite, chloramines, and peroxynitrite; all these oxidants are produced in biological systems.¹⁵³ However, this modification can be repaired by methionine sulfoxide reductase, which catalyzes the thioredoxin-dependent reduction of methionine sulfoxide back to methionine.¹⁵⁴ It has been pointed out that the reversible oxidation of methionine has the potential to modulate intracellular signaling under conditions involving oxidative stress in a manner analogous to other regulatory posttranslational modifications such as those involving disulfide bond formation or phosphorylation.¹⁵³

Protein cysteines can undergo various forms of oxidation, some of them reversible (disulphide formation, glutathionylation, cysteinylation, S-nitrosylation and formation of sulphenic and sulphinic acids).¹⁵⁵ Intraprotein disulphide bonds are viewed, in classical textbooks, as part of the well ordered tertiary structure of the protein and their formation is an important step in protein folding. Similarly, interprotein disulphide bonds are important in the quaternary structure of proteins and in the formation of homo- or hetero-multimers.

Hypusine

Some proteins are posttranslationally modified on lysine residues to form hypusine (N-epsilon-(4-aminobutyl)lysine). Hypusine (a molecule comprises moieties of HYdroxyPUtrescine and lySINE) was first isolated from bovine brain in 1971.¹⁵⁶ eIF5A is the only protein in eukaryotes and archaebacteria known to contain hypusine. Hypusine is formed in eIF5A by a novel post-translational spermidine-dependent modification reaction that involves two enzymatic steps. In the first step, deoxyhypusine synthase catalyzes the cleavage of the polyamine spermidine and transfer of its 4-aminobutyl moiety to the epsilon-amino group of one specific lysine residue of the eIF5A precursor to form a deoxyhypusine intermediate. In the second step, deoxyhypusine hydroxylase converts the deoxyhypusine-containing intermediate to the hypusine-containing mature eIF5A.¹⁵⁷ The structure and mechanisms of deoxyhypusine synthase and deoxyhypusine hydroxylase have been extensively characterized.¹⁵⁸ Deoxyhypusine hydroxylase is a HEAT-repeat protein with a symmetrical superhelical structure consisting of 8 helical hairpins (HEAT motifs). It is a novel metalloenzyme containing tightly bound iron at the active sites.¹⁵⁸

Intrinsic disorder in disease associated proteins

In the Swiss-Prot disease category, 10 disease-related keywords have strong correlation with intrinsically disorder proteins and no disease-associated keywords were found to be related to intrinsic order (see Table 5).

Table 5.

All (11) disease keywords strongly correlated with predicted disorder

Keywords	Number of proteins	Number of families	Average sequence length	Z-score	p-value
Proto-oncogene	405	135	567.12	8.86	1
Malaria	110	35	615.59	5.46	1
AIDS	405	17	371.9	4.95	1
Trypanosomiasis	26	19	431.85	4.06	1
Deafness	94	65	900.21	3.75	1
Cardiomyopathy	31	28	941.26	2.75	1
Albinism	25	14	638.28	2.71	1
Diabetes mellitus	36	28	566.14	2.33	0.99
Obesity	48	21	376.83	2.22	0.99
Prion	74	5	253.41	1.72	0.97
Disease mutation	1269	763	722.24	1.59	0.96

Oncoproteins

Important to this study, disorder is very common in cancer-associated proteins (or oncoproteins).¹⁵⁹ In fact, in previous studies we found that 79% of cancer-associated and 66% of cell-signaling proteins contain predicted regions of disorder of 30 residues or longer.¹⁵⁹ In contrast, in a control set of proteins with well-defined ordered structures, which was extracted from protein data bank (PDB), the content of such long disordered regions was much smaller, as only 13% of these proteins contained long regions of predicted disorder. In experimental studies, the presence of disorder has been directly observed in many cancer-associated proteins, a few examples of which include p53,¹⁶⁰ p57^kip2,¹⁶¹ Bcl-X_L and Bcl-2,¹⁶² c-Fos,¹⁶³ proto-oncogene securin,¹⁶⁴ a breast cancer associated protein BRCA1.¹⁶⁵ We have recently established that the E6 and E7 oncoproteins from the high-risk types of human papillomaviruses (HPVs) possess more predicted intrinsic disorder than proteins from the low-risk HPVs.¹⁶⁶

Malaria

Biophysical analysis of the merozoite surface protein 3 (MSP3) of Plasmodium falciparum (the parasitic agent that causes most cases of fatal malaria) revealed that MSP3 polypeptides contain a large amount of α-helix and random coil secondary structure and form highly elongated dimers and tetramers.¹⁶⁷ The MSP3 dimer was assumed to be formed via a parallel coiled-coil interaction between the leucine zipper-like regions of two monomers. Importantly, MSP3 contains numerous features associated with intrinsic disorder: its central domain includes three blocks of imperfect Ala heptad repeats, a second central domain is a very long Glu-rich region (242–294 fragment), whereas the C-terminal domain contains a leucine zipper motif.¹⁶⁷ Apical membrane antigen 1 (AMA1) of the malarial parasite Plasmodium falciparum is a merozoite antigen that is considered a strong candidate for inclusion in a malaria vaccine. The solution structure of AMA1 domain III, a 14kDa protein, has been determined using NMR spectroscopy.¹⁶⁸ It was shown that the structure has a well-defined disulfide-stabilized core region separated by a disordered loop, and both the N and C-terminal regions of the molecule are unstructured.¹⁶⁸ In another study the solution structure of a 16kDa construct corresponding to the putative second domain of AMA1 have been reported.¹⁶⁹ Interestingly, while CD and hydrodynamic data were consistent with a folded structure for domain II, its NMR spectra were characterized by broad lines and significant peak overlap, more typical of a molten globule.¹⁶⁹

Trypanosomiasis

Several surface antigens of trypanosome are glycoproteins. For example, the glutamic acid and alanine-rich procyclin from Trypanosoma congolense is a glycoprotein containing a galactosylated GPI-anchor and carrying two large mannose- and galactose-containing oligosaccharides linked to threonine residues via phosphodiester linkages.¹⁷⁰ Furthermore, in Trypanosoma brucei, the procyclic and epimastigote stage glycoproteins belong to the family of procyclic acidic repetitive proteins (procyclins) containing numerous tandem repeats of EP or GPEET.^171–173 The procyclins provide a highly acidic coat that is proposed to help the parasite survive in the harsh environment of the tsetse fly gut.¹⁷⁰ The isolated GPEET procyclin from the T. brucei is highly susceptible to proteolytic treatment,¹⁷² suggesting one possible need for the protective coat.

HIV and AIDS

The human immunodeficiency virus (HIV) is the causative agent for acquired immunodeficiency syndrome (AIDS). The HIV genome encodes a total of three structural proteins, two envelope proteins, three enzymes, and six accessory proteins.¹⁷⁴ One of the HIV structural proteins, nucleocapsid (NC) protein, is involved in interactions with nucleic acids. The solution behavior of NC may be best considered as a rapid equilibrium between conformations with weakly interacting and non-interacting knuckle (zinc finger-like) domains, and this inherent conformational flexibility plays a crucial role in the adaptive binding of NC to different nucleic acid targets.¹⁷⁵ A basic domain (Arg35 to Arg50) of the accessory protein Rev was shown to be largely unstructured in solution, but forms an α-helix upon binding to the SLIIB RNA stem.^{176, 177} NMR structural analysis of another accessory protein, Tat (trans-activator) protein from HIV-1, revealed that it possesses two highly flexible domains corresponding to a cysteine-rich and a basic sequence region.¹⁷⁸ Furthermore, recent multinuclear NMR analysis of uniformly ¹⁵N and ¹⁵N/¹³C-labelled Tat_1–72 region under reduced conditions (pH 4.1) revealed that that it exists in a random coil conformation.¹⁷⁹ The absence of a fixed conformation and the fast dynamics are consistent with the ability of Tat protein to interact with a wide variety of proteins and nucleic acid and support the concept of a natively-unfolded protein.¹⁷⁹ The envelope glycoprotein of HIV forms trimers on the virion surface, with each monomer consisting of two subunits, gp120 and gp41.¹⁸⁰ During the initial step of HIV infection, the gp41/gp120 complex associates with the CD4 receptor located on the surface of a human cell. The mature gp120 consists of 5 conserved regions (C1–C5) and five variable regions (V1–V5). The C1 and C5 domains of gp120 are involved in direct interaction with gp41 but are largely missing from the available X-ray structure,¹⁸¹ suggesting that these regions can be disordered. In agreement with this disordered region hypothesis, the HIV gp120 C5 domain (residues 489–511 of HIV-1 strain HXB2), which corresponds to the carboxy terminal region of gp120, has been recently reported to be unstructured in aqueous solutions.¹⁸²

Deafness

Connexins are integral membrane proteins that oligomerize to form intercellular channels called gap junctions between adjacent cells, and these channels promote intercellular communication. Connexin proteins are involved in pathological conditions in humans, mainly in hearing loss, neurodegenerative disorders and skin diseases. There are over 100 mutations in genes encoding connexins that are associated with deafness. Most prominent is the remarkable involvement of connexin 26 in hearing loss. Mutations in the gene GJB2, encoding connexin 26, are responsible for around 50% of genetic cases of severe to profound non-syndromic hearing loss in some parts of the world.¹⁸³ Conformational analysis of two intracellular domains of connexin43 (Cx43), cytoplasmic loop 95–144 and C-terminal domain (amino acids 254–382), revealed that they are mostly disordered and possess short transient α-helices that are connected by long, highly flexible loops of random coil.^184–186 Numerous binding partners to the carboxyl-terminal domain of Cx43 have been identified; these include tubulin, v-Src, c-Src, ZO-1, casein kinase 1 (CK1), mitogen-activated protein kinase (MAPK), cGMP-dependent protein kinase, cAMP-dependent protein kinase, and protein kinase C.¹⁸⁷ Interestingly, other than the possible binding of CK1 to Ser-325 within one of the α-helices, all these proteins interact with Cx43 in disordered regions outside the two α-helical domains.¹⁸⁶ This firther emphasizes that the disordered structure of the carboxy-terminal domain of Cx43 is advantageous for signaling between different binding partners.

Obesity, cardiovascular disease, and diabetes mellitus

Several human diseases including cardiovascular disease, diabetes mellitus, hyperlipidemia and hypertension and so-called metabolic syndrome are obesity-related disorders that are associated with the visceral fat accumulation.¹⁸⁸ Our recent bioinformatics analysis of a set of 487 cardiovascular disease (CVD)-related proteins revealed that these proteins are enriched in intrinsic disorder.¹⁸⁹ The percentage of proteins with long disordered regions was shown to be 61(±5)% for CVD-associated proteins, which is less than the value described earlier for human cancer-associated and signaling proteins (79±5% and 66±6%, respectively),¹⁵⁹ but which is significantly larger than that in eukaryotic proteins from SWISS-PROT (47±4%) and in non-homologous protein segments with well-defined 3-D structure (13±4%). Furthermore, 120 out of 487 proteins in the CVD dataset (~25%) were predicted to be wholly disordered.¹⁸⁹ This high level of intrinsic disorder could be important for function of CVD-related protein and for the control and regulation of processes associated with cardiovascular disease. In agreement with this hypothesis, ¹⁹⁸ α-helical molecular recognition fragments α-MoRFs¹⁹⁰ were predicted in 101 proteins from CVD dataset.¹⁸⁹

Albinism

is a complex genetic disease, which is a result of the melanin pigment deficiency in the skin, hair, and eye [oculocutaneous albinism (OCA)], or primarily in the eye [ocular albinism (OA)]. Mutations in six genes have been reported to be responsible for different types of oculocutaneous and ocular albinism, including the tyrosinase gene (TYR), the OCA2 gene, the gene (TYRP1), the Hermansky-Pudlak syndrome (HPS) gene, the Chediak-Higashi syndrome (CHS) gene, and the X-linked ocular albinism gene.¹⁹¹ Tyrosinase is a glycosylated transmembrane copper-containing enzyme that is responsible for conversion of tyrosine to dopaquinone, which is the rate-limiting step in the melanin pathway.¹⁹² The enzyme contains 529 amino acids, including an 18-amino acid signal peptide, two putative copper-binding sites, and a hydrophobic transmembrane region at the C-terminal end.¹⁹³ Tyrosinases are activated in vivo by limited proteolytic cleavage.¹⁹⁴ All these proteins are predicted to have long regions of intrinsic disorder.

Prion

The central event in the pathogenesis of prion diseases is a major conformational change of the prion protein (PrP) from an alpha-helical (PrP^C) to a beta-sheet-rich isoform (PrP^Sc). The mature PrP^C species consists of an N-terminal region of about 100 amino acids, which is unstructured in the isolated molecule in solution, and a folded C-terminal domain, also approximately 100 amino acids in length. It has been emphasized that the most striking feature of the full-length PrP is the random-coil nature of chemical shifts for its residues 30–124.¹⁹⁵ The C-terminal domain is folded into a largely α-helical conformation (three α-helices and a short antiparallel β-sheet) and stabilized by a single disulfide bond linking helices 2 and 3.¹⁹⁶ Although unstructured in the isolated molecule, the N-terminal region contains tight binding sites for Cu²⁺ ions and therefore may acquire structure following copper binding.^{197, 198}

Disease-associated mutations

Functional protein represents a tightly balanced machine, structure and performance of which can be easily distorted by point mutations. For example, p53 was shown to play crucial role in the development of most cancers by condemning damaged cells to death or quarantining them for repair.¹⁹⁹ p53 activity relies on its intact native conformation, which can be lost following mutation of a single nucleotide. In fact, over 10,000 somatic tumorigenic mutations in p53 gene were found,²⁰⁰ with 95% of these lie in the core DNA-binding domain.¹⁹⁹

Furthermore, even single changes of amino acid in protein sequences can change the rates at which they aggregate by an order of magnitude or more,^{201, 202} thus dramatically accelerating development of protein depositions and related diseases. In fact, the changes in aggregation rates cause by such mutations were shown to correlate with changes in simple properties that result from such substitutions, such as charge, secondary structure propensities and hydrophobicity.²⁰² Mutations modulate the aggregation propensities of both, well-folded and intrinsically disordered proteins. Numerous neurodegenerative diseases originate from misfolding and neurotoxic aggregation of specific proteins. It is established that Aβ, α-synuclein and prion protein, the major players involved in the pathogenesis of such famous diseases as Alzheimer’s and Parkinson’s diseases and prion diseases, respectively, are either completely disordered (Aβ, α-synuclein) or contain long disordered regions (prion protein).²⁰³ For example, detailed structural analysis revealed that α-synuclein, a conservative presynaptic protein, the aggregation and fibrillation of which is assumed to be involved into the pathogenesis of Parkinson's disease and several other neurodegenerative disorders, synucleinopathies, is characterized by the lack of rigid well-defined structure under the physiological conditions in vitro. However, this protein is characterized by a remarkable conformational plasticity as it adopts a series of different conformations depending on the environment, being able to either stay substantially unfolded, or adopt an amyloidogenic partially folded conformation, or fold into α-helical or β-structural species, both monomeric and oligomeric. Its aggregated forms possess astonishing morphological diversity, ranging from oligomers (spheres or doughnuts) to amorphous aggregates or amyloid-like fibrils. This unusual conformational behavior and exceptional structural plasticity of α-synuclein have led to the protein-chameleon concept, according to which a polypeptide chain can gain different structures depending on its environment.²⁰⁴ There are three point mutations in α-synuclein, A30P,²⁰⁵ E49K,²⁰⁶ and A53T,²⁰⁷ which are associated with the early onset of the Parkinson’s disease and were shown to accelerate the α-synuclein aggregation (but not necessarily fibrillation) in vitro.²⁰⁸

Conclusions

We have established that many natural ligands, PTMs and diseases are associated with proteins predicted to possess long disordered regions. Strong positive or negative correlations of different PTMs with intrinsically disordered regions is of special interest. Our data are consistent with the existence of two major protein groups subjected to PTMs. The first group involves modifications that are associated primarily with structured proteins and regions. These include the following modifications: formylation, protein splicing, oxidation and covalent attachment of quinones and organic radicals. These modifications are important for providing moieties for catalytic functions, for modifying enzyme activities or for stabilizing protein structure.

The second group involves modifications that are associated primarily with intrinsically unstructured or disordered proteins and regions. These include the following modifications: phosphorylation, acetylation, acylation, adenylylation, ADP ribosylation, amidation, carboxylation, formylation, glycosylation, methylation, sulfation, prenylation, ubiquitination, Ubl-conjugation (i.e., covalent attachment of ubiquitin-like proteins, including SUMO, ISG15, Nedd8, and Atg8). These modifications involve low affinity, high specificity binding interactions between a specific enzyme and a substrate (the protein that is modified). Combination of high specificity with low affinity, being ideal for signaling, can be achieved via coupled binding and folding. Importantly, posttranslational modifications associated with intrinsically disordered proteins and regions are especially important for signaling and regulation. For example, protein phosphorylation is known to represent a crucial regulatory mechanism in eukaryotic cells. Another illustrative example is Ubl-conjugation, which is critical for many cellular processes, including transcription, DNA repair, signal transduction, autophagy, and cell-cycle control. Ubl-conjugation cascades are initiated by activating enzymes, which also coordinate the ubls with their downstream pathways. In fact, conjugation of ubiquitin-like proteins (the Ubl conjugation pathway) to components of the transcriptional machinery is an important regulatory mechanism allowing switching between different activity states. While ubiquitination of transcription factors is associated with transcriptional activation, their SUMOylation is most often connected with transcriptional repression.

Overall, comparison of the posttranslational modifications in structured and intrinsically disordered regions revealed that the latter are more diverse and more often tend to be reversible. Thus, the irreversible posttranslational modifications are mostly used to increase stability and allow catalytic functions of the ordered proteins, whereas the reversible modifications are more frequently used for the signaling activities of intrinsically disordered proteins.