Designing disorder: Tales of the unexpected tails.

Protein tags of various sizes and shapes catalyze progress in biosciences. Well-folded tags can serve to solubilize proteins. Small, unfolded, peptide-like tags have become invaluable tools for protein purification as well as protein-protein interaction studies. Intrinsically Disordered Proteins (IDPs), which lack unique 3D structures, received exponentially increasing attention during the last decade. Recently, large ID tags have been developed to solubilize proteins and to engineer the pharmacological properties of protein and peptide pharmaceuticals. Here, we contrast the complementary benefits and applications of both folded and ID tags based on predictions of ID. Less structure often means more function in a shorter tag.

Introduction

IDPs are abundant in all eukaryotic proteomes. Short ID regions mediate some 40% of all protein-protein interactions, for instance in regulatory tails of proteins., Large ID regions occur in essential scaffolds and molecular chaperones.-

What makes IDPs special? IDPs don’t form rigid hydrophobic cores due to lower fractions of hydrophobic and aromatic residues (W, I, Y, F, L, H, V). They are often more elongated due to a predominant self-repulsive charge (E or K) or disorder-promoting hydrophilic (S) or structure-breaking residues (P, G)., Enlarged accessible surfaces of IDPs can facilitate multiple types of interactions including phosphorylation by protein kinases., The ~500 human kinases regulate crucial cellular processes including cell cycle and cell differentiation by phosphorylation at thousands of sites, which are typically part of short linear motifs.

Whereas X-ray crystallography has transformed our structural understanding of folded proteins, still relatively little is known about conformational ensembles of IDPs. Unique formidable challenges in IDP research start at the stage of soluble overproduction and subsequent protein purification. High susceptibility to proteolysis or a high aggregation-propensity of many IDPs has slowed down progress in the IDP research field., Improvements in solubilizing IDPs, facilitating their purification and efficient prevention of their aggregation, have the potential to catalyze future progress in the IDP research field, which is already growing exponentially.

Protein tags of various sizes and shapes have helped solubilizing, purifying and characterizing proteins (Table 1)., Fusing a folded tag may solubilize proteins during production and purification. Multiple small tags mediate reversible to tight or even covalent binding during protein purification and subsequent characterization and applications of the purified proteins., Here we review these commonly used folded and ID tags with an emphasis on contrasting their known structure vs known or predicted ID propensity and applications. Finally, we aim to highlight recently emerging applications of large designed ID tags. Large ID tags have the potential to transform protein production and purification, protein-based medicine and to make medicine more patient-specific.-

Table 1. Overview of protein fusion tags and their PONDR-FIT predicted ID propensity

ID [%]	Name	Application	Obeserved effect	Composition	Residues	References
6	MBP	protein solubility	65% > Histag	X	396	22 , 26
6	NusA	protein solubility	solubility	X	495	122
6	HaloTag	coupling	covalent coupling	X	297	123
7	GroEL	solubility	aggregation prevention	X (- W)	548	30
9	DsbA	peptide stabilization	high peptide yield	X	208	124
11	GST	affinity	dimer/affinity	X	200	125
11	IN	split-intein	induced folding	X(- M)	123	35
15	Fc tag	ProtA/G affinity	high affinity (10 nM)	X	232	126
16	Lipoyl	solubility	solubility	X (- RM)	80	127
18	GFP	fluorescence	tunable fluorescence	X (- CW)	238	128 , 129
18	SnapTag	coupling	in vivo labeling	X	182	39 , 130
18	GFP11	split-GFP	in vivo complementation	X (- NCQPSWY)	17	129
21	Trx	protein solubility	peptide stabilization	X (- HR)	103	18
21	DnaK	solubility	aggregation prevention	X	638	30
22	CBD	chitin affinity	high capacity		51	131
24	mtHsp70	solubility	high yields	X(- C)	625	132
40	Sortag	coupling	covalent coupling	LPXTG	5	58 , 59 , 62
41	IC	split-intein	coupled folding/binding	X (- CWMTWY)	34	35
46	ySUMO	protein solubility	peptide stabilization	X (- CW)	100	85 , 133
47	GB1	solubility	solubility	X(- PHCR)	64	134
50	Histag	affinity	specific binding	H	6	20
100	XTEN	half life	tunable increase	PESTAG	36–1008	25
100	HAP	half life	3-fold increase	SG	100, 200	23
100	PAS	half life	tunable increase	PAS	100–600	24
100	EB60A	solubility	75% > Histag	PESQ	60	22
100	EB60B	solubility	75% > Histag	PEGQ	60	22
100	EB144	solubility	95% > Histag	PESDGQ	144	22
100	EB250	solubility	100% > Histag	PESDGQMVILF	250	22
100	Antigen 13 rep.	half-life	2-fold increase	PESAK	280	135
100	SAPA repeats	half life	5-fold increase	PASTDH	156	135
100	HRM	half life	4.5–6-fold increase	PASTD	105	136
100	poly(E)	drug solubility	tumors reduced	E	277	137 , 138
100	ELP	drug efficacy	enhanced efficacy	PVGX	160- 750	139
100	ELP	half life	8.7h half-life	PVGX	450–600	140
100	Random coil	length	tunable gel	PEGQAN	100–800	141
100	NNT/NNS	drug properties	tuned glycosylation	NTS	60–750	142
100	ELP	purification	phase cycling possible	PVGX	50–900	114
100	ELP	solubility of Ig	improved solubility	PESTAGX(- HY)	22–61	143
100	ELP	cost reduction	tailored transition	PVGX	450	144
100	ELP	simple purification	effective as Histag	PVGX	?	145
100	HIS3	immunoaffinity	co-crystallization	H	3	74
100	c-MYC	Detection	highly specific	ELDQIKS	10	18
100	NSS	Detection	highly specific	FINQHMKT	9	51
100	StreptagII	streptactin affinity	highly specific	WSHPQFEK	8	43 , 44
100	V5 tag	detection	highly specific	PSTKLINDG	14	146
100	HA tag	detection	highly specific	YPDAV	9	147
100	FLAG tag	detection	highly specific	DKY	8	18
100	Avi/BAPtag	biotin binding	biotinylation by BirA	EIANDQGHLKF	15	38
100	Ybbr tag	coupling	covalent coupling	ALSDEIKF	11	70
100	Spytag	coupling	covalent coupling	KAVDIHPTYM	13	72 , 73
100	Zn hook	dimerization	fM dimer	CGKRADELPTV	14	148
100	Leucine Zipper	dimerization	nM hetero-dimer	KELQANW	30+30	149 , 150
100	Arg tag	purification	surface adhesion	R	5	75 , 151
100	tetracystein tag	FlAsH binding	FlAsH dequench	CPG	6	152
100	CBP	calmodulin affinity	high affinity	KASRNIFGLWV	26	83 , 153
n.a.	Cysteine	specific reactivity	chemical coupling	C	1	154

Composition is indicated by single-letter representation of amino acids with X denoting any of the 20 amino acids; amino acids in brackets are not occurring in the respective tag. For calculation of the indicated ID fractions in percent of the total number of residues, we counted residues with a PONDR-FIT predicted disorder propensity > 0.5 as disordered and residues with lower predicted disorder propensity as ordered.

Folded Tags Facilitate Production, Purification, and Rapid Labeling In Vivo

Folded fusion tags, ranging in size from ~5 kDa to 70 kDa, are versatile tools that serve to increase folding and solubility during recombinant protein production or to enable fluorescent tracing of single proteins.

Commercially available examples of solubility-enhancing folded protein fusion tags include MBP, NusA and GST., These tags significantly improved the soluble overproduction of a range of target proteins on a proteomic scale. Similar effects have been described more recently for the smaller ySUMO, Lipoyl and Trx fusions, which seem to be of particular use for the overproduction of peptides and small IDPs., Fusion of molecular chaperones, chiefly of the Hsp70 class, helped solubilizing aggregation-prone IDPs. For instance, the highly aggregation-prone prion protein could be solubilised during overproduction by fusing it to DnaK, the E. Coli homolog of Hsp70.

In addition to their role in protein production, well-folded proteins can mediate a large number of specific functions, including fluorescence. Green fluorescent protein (GFP) and its variants are popular genetically encoded fluorescent labels, which generally do not interfere with the function of their fusion partner. An elegant application of folded tags is the so-called split-FP assay, used in protein-protein interaction studies, where two halves of GFP are appended to putatively interacting proteins to map their interactions in space and time in living cells. Only upon close encounter of two proteins, both fused GFP halves complement each other to form the mature fluorophore. A recent application of split-GFP to trace direct molecular contacts of proteins involved in peroxisome-biogenesis demonstrated that this approach yields super-resolved insights on the low nm scale using a conventional wide-field fluorescence setup.,

Other examples of popular folded tags include enzymes that catalyze the formation of specific covalent bonds. Inteins, for instance, catalyze bond-formation and self-excision. Two halves of self-excising split-intein tags can be appended to both ends of a target peptide to cyclize peptides in vivo for rapid generation of stabilized libraries of small peptides.- Two recently established enzyme tags called SnapTag and HaloTag facilitate rapid fluorescent labeling. Fluorescent tracing of low-abundant receptors using a rapid but non-denaturing SnapTag coupling protocol was recently demonstrated., HaloTag coupling was used to site-specifically couple single protein molecules for optical tweezers studies of protein folding.,

Small ID Tags Are Popular Multi-Talents

Many of the most commonly used tags are small and predicted to lack structure (Table 1). Despite being around 20 residues at maximum, they mediate highly specific and diverse functions, from tunable affinity probes to covalent coupling points.

Scientists can now choose from a large range of small tags, which ensure that the protein fused to it can be highly selectively detected and enriched by specific binders. Frequently used examples for applications in pull-downs and chromatographic purifications include Hexahistidine (short: His)-tags and the eight amino acids short StreptagII.,,, His-tags mediate reversible and tunable affinity under a wide range of conditions from near physiological to protein denaturing conditions, have been used successfully in thousands of protein purifications, and are an integral part of most structural biology pipelines. Many small His-tagged proteins are sufficiently overproduced in E. coli to achieve greater than 95% purity in one immobilised metal affinity chromatography (IMAC) purification step. In purifications from mammalian hosts, histidine-rich metal-binding intracellular proteins might interfere with IMAC. The more selective Strep-tag II can be a valuable alternative or additional purification step along with His-tags. Its specificity was further increased by repeats allowing immobilisation of single molecules under applied force (up to 65 pN) for hours and purification of low abundant mammalian proteins.-

Popular epitope tags for immunoblots, pulldowns, or immunoprecipitations include FLAG tag, Myc tag, V5 tag and HA tag., Since their sequence composition and hydrophobicity varies significantly, screening several variants in parallel helps optimising the results. Repeats often help to increase the effective binding strength or apparent specificity of binding of the epitope tag to its cognate antibody., The recently described NSS tag was suggested to be comparable to multi-Myc tags and might be an interesting alternative for future co-immunoprecipitation and protein interaction studies.

The AviTag, or biotin acceptor peptide (BAP), allows for in vivo biotinylation. Biotin has an ultra-high affinity for Streptavidin (KD of ~10−14 M) and forms an unusually stable complex that is still 50% folded at 112°C., Many sensitive detection assays, immobilisation strategies and nano-patterning strategies exploit the strength of this interaction.- Recently, “Traptavidin,” an engineered variant of streptavidin, further improved the thermostability of the complex and decreased the off-rate by an order of magnitude.

Several small tags have been developed to allow site-specific covalent coupling, either spontaneously or upon addition of a highly specific enzyme. Sortase-mediated protein ligation (SMPL), for instance, is a method to couple polypeptide chains to a wide variety of (bio)molecules including nucleotides, sugars, lipids, organic and inorganic particles, or to circularise peptides efficiently., Polypeptides with a C-terminal Sortag (LPXTG motif) can be specifically recognized by Sortase A and transferred to either the natural nucleophile (GGGGG) or triple-Glycine-modified target molecules. Most applications use Staphylococcus aureus Sortase A., Complementary semi-orthogonal calcium-independent Sortases allow incorporation of a second coupling point using the LPXTA motif in the same polypeptide chain or enable transpeptidations in vivo in specific cellular compartments.- The small size of the Sortag, combined with good expression of the Sortase A enzyme as well as the flexibility of Sortase A as “molecular stapler” for a large range of bio-conjugation challenges made this system popular.,

A major limitation of the Sortase A coupling chemistry is the low efficiency of the reaction at nanomolar or lower concentrations of reactants. Directed evolution of rate-enhanced variants of Sortase A and immobilisation of the Sortase A to a flexible solid helped to increase yields and speed of reaction.- The six residues longer ybbR tag can be specifically recognized by the Sfp enzyme. This tag was successfully applied to attach single stalled ribosomes to beads to assess ribosome-associated folding in optical tweezer studies. The recently engineered Isopeptag and Spytag permit enzyme-free, efficient site-specific covalent coupling to the spontaneously bond-forming proteins Spy0128 and SpyCatcher and might become a valuable new addition to the protein engineering toolbox.,

The Better Tag Remains Unseen

Folded tags typically need to be removed after protein purification to prevent structural or functional interference during biophysical or structural characterization of the target protein. Crystallographic structure determination usually even requires removal of short disordered tags as their flexibility may hamper crystallization, although a few cases have been reported in which the presence of the Arg-6, His-3, His-6, FLAG, or c-Myc tag did not preclude structural studies.- Several highly site-specific enzymes cleave at short but rather rare sequence motifs, which lack 3D structure propensity, and are inserted between the tag and protein of interest: ENLYFQ*G/S for TEV protease, DDDK* for Enterokinase, IE/DGR* for Factor Xa, LEVLFQ* GP for HRV 3C, LPXT*G for Sortase A, LVPR* GS for Thrombin (cleavage occurs at *).- Sortase A uses the same motif for both specific cleavage and subsequent protein ligation. Thus, tag removal can be simultaneously combined with attachment of other tags using Sortase A or inteins, for instance to include NMR-invisible solubility tags.,

Although the above-mentioned cleavage motifs are relatively rare in proteomes, it is not excluded that they are present in some natural or engineered target proteins. Hence selecting a unique cleavage site to specifically remove a fusion protein from the protein of interest requires caution. Enhanced cleavage specificity can be attained if cleavage depends on recognition of a unique 3D structure as in the case of Ubiquitin, SUMO, or Intein proteins.- Briefly, “visible,” folded tags can be efficiently removed using unstructured motifs and “invisible,” either unlabelled or unstructured tags are often compatible with or even required for downstream applications.

ID Tags Are Not Limited by the Need to Fold During Purification

Structured tags are often structurally well-characterized by high-resolution methods like X-ray crystallography and NMR spectroscopy (Fig. 1). Because of their defined, compact structure, they are less likely to be degraded by proteases, and amenable to rational design and engineering of their thermal stability.- However, if the conditions are not ideal for folding, for instance upon exposure to physicochemical stresses or when folding intermediates of the target protein engage in non-native interactions, irreversible aggregation and loss or unpredictable alteration of tag function could occur. Even though folded tags are well-established, their relatively large size and their need to fold can thus be limiting for many applications.

Figure 1. Structural representation of tag order and disorder. (A) Cartoon representation of selected protein fusion tags as indicated in the text and Table 1. His6 tag is shown in stick representation. All tags are shown at the same scale. Numbers between brackets indicate the PDB ID from which the represented structures were derived. The fusion tags are colored by secondary structure: cyan indicates an α-helix, red indicates a β-sheet, and purple indicates a loop or turn. (B) Overlay of the different conformations of lipoyl NMR structures (PDB ID: 1QJO), showing limited flexibility.

IDP tags and short ID tags are structurally less characterized, newer, and therefore less established for some applications. Their larger solvent-exposed surface and lack of rigid structure might make them more susceptible to proteolysis. Several variants of large ID tags have, however, been described, which can be expressed at high yields for biophysical characterization., Up to 400 mg/L of large ID tag fusions have been obtained in a recent example. ID tags do not need to fold, therefore the stability of ID tags in denaturing conditions is frequently exploited in denaturing affinity purifications.

ID and Its Flavors Can be Predicted With High Accuracy

Propensity for ID is encoded in the sequence of amino acids and can be efficiently predicted for all sequenced proteomes. Most prediction algorithms combine physico-chemical properties of amino acids and empirical knowledge of structure propensities as derived from large experimental data sets. One of the most widely used algorithm is PONDR. Its accuracy is, like similar methods, around 80% for water-soluble proteins; these estimates might be biased by the limited availability of data for large disordered regions above 40 residues, which rarely crystallize.,- Thousands of sequenced genomes have been analyzed with this algorithm. These PONDR predictions indicate that some 40% of all eukaryotic protein residues are located in ID regions. More recent methods of ID prediction combine several predictors into one meta-prediction, which significantly improved the accuracy of predictions in several cases. Disorder predictions are increasingly accurate and fast. Future progress in algorithm development will benefit from novel large-scale experimental data sets on ID propensity in larger fractions of complete proteomes including yet understudied membrane proteins.,,,

A good experimental proxy for the average apparent size and extension of an IDP is the hydrodynamic radius.- Both hydrodynamic experiments (such as SEC or more recently SEC-MALLS) as well as scattering experiments (DLS, SAXS, SANS) can give insights into hydration radii., Drawbacks of these experimental approaches are that they require either substantial amounts of highly purified protein or access to expensive equipment. Computational approaches to model global features of IDPs are not limited by these restraints. An empirically optimised formula to predict hydration radii has been proposed based on a large set of experimentally solved radii and statistical correlation with sequence features of these proteins. Mainly based on proline content and net charges, this predictor achieves high accuracy for many small disordered proteins suggesting broad applicability. An alternative approach approximates IDPs as polyampholytes because of their enrichment in both positive and negative charges. While weak polyampholytes collapse to globule-like shapes, stronger polyampholytes are more extended and their shape depends on local distribution of positive and negative charges within the primary sequence. Their predicted ensemble properties can be efficiently modeled computationally using implicit salvation in the Absinth force-field. Interactions of small ID regions with folded proteins often involve coupled folding and binding.- Combining initial rigid-body docking and subsequent flexible adjustment appears to be a computationally efficient solution to this problem. Several novel computational approaches expand the scope of prediction of ensemble properties of IDPs.

Can we also predict aggregation propensity of ID regions from their amino acid sequence? Irreversible aggregation is a well-known disease-associated feature of multiple naturally occurring IDPs. Several ID regions are, however, highly soluble or aggregate reversibly. Sequence composition, particularly the relative fractions of proline and glycine, strongly affect amyloidogenicity of unfolded polypeptides. Also a large percentage of charges, especially glutamate, can help preventing aggregation in a broad range of conditions. Several net negatively charged proteins like tau or the engineered PESTAG sequences remain soluble after heating to 95°C., Precise control of reversible thermal aggregation has been demonstrated using engineered elastin-like polymers (ELPs). ELPs that contain large fractions of proline, glycine, valine enable cycles of thermally controlled reversible aggregation and solubilisation around an engineered transition temperature between 40°C and 80°C. Thus, multiple strategies help predicting and controlling the aggregation-propensity of IDPs.

Designed, Large ID Tags: Applications from Biotechnology to Next Generation Protein Pharmaceuticals

Increased understanding of naturally occurring IDPs and improved predictability of their features enabled the design of several ID tags for overcoming protein aggregation, facilitating non-chromatographic purification and tuning pharmaco-kinetic profiles of protein drugs.

Inspired by the amino acid composition of plant dehydrins, which efficiently cope with extremes of dessication stress and prevent aggregation upon rehydration, and aided by predictions of ID and solubility, the Keith Dunker lab recently developed several large, negatively charged, fully disordered “entropic bristle” (EB) tags to solubilise difficult-to-produce proteins. EB fusions successfully solubilised target proteins from several protein classes including kinases, transcription factors, proteases, and neurodegeneration-related proteins. The EB tags did not interfere with GST enzyme function and structural stability and therefore did not have to be removed after production in this particular case. If tag removal is, however, required and the target protein is stably folded and proteolytically resistant, rapid thermally accelerated proteolysis of the unstructured tags could be performed using the inexpensive, commercially available enzyme thermolysin.

Many proteins have been purified using short affinity tags and chromatography. Scaling up chromatography-based purifications can be very costly as most chromatography materials are costly. Chromatography-free purification methods that yield highly purified proteins are therefore desirable for many applications. One possible solution to this problem exploits repeated cycles of reversible aggregation and solubilisation of designed elastin-like sequences, which have a defined transition temperature point. A recent extension of this concept by simultaneous genetic fusion of both an elastin-like peptide (ELP) sequence and a Sortase A fusion tag achieved one-step purification of large amounts of soluble target proteins without the need of chromatography for several tested proteins. This method yielded, for instance, 28 mg/L pure, sortagged GFP and 35 mg/L pure, sortagged Trx without chromatography.

Protein and peptide pharmaceuticals have long been suggested as possible complementation for traditional small-molecule drugs and inspire hopes for reduced side-effects and more targeted therapies. Current limitations for their broader application include large production costs, less convenient administration by injection and a short half-life. The plasma half-lives of peptide pharmaceuticals can be as short as ten minutes. Increasing the half-life could help overcoming both cost problems and simplify drug dosage and application. Several groups have recently described complementary, effective strategies to use designed ID tags for extending the half-life of protein pharmaceuticals. The Pim Stemmer lab developed super-sized, unfolded PESTAG sequences, which increased the half-time of peptides and hormones up to 125-fold by mimicking attachments of globular molecules in the Megadalton range. Curiously, this supercharged fusion also completely prevented the thermal aggregation of human growth hormone and might be beneficial in transport and storage of protein drugs. The Arne Skerra lab more recently presented the PASylation technology, which uses an uncharged but stably unfolded and largely extended polymer consisting of proline, alanine, and serine. This technology also mediated large increases of half-lives of protein pharmaceuticals and its fusions could be expressed at very high levels (400 mg/L E.coli fermentation culture volume). Thus, these novel, large ID fusion tags might become invaluable alternatives for organic polymer extensions using polyethylene glycol that turned out to be more immunogenic than previously anticipated.

Conclusions

Folded and ID tags can have overlapping as well as different features, resulting in complementarity and synergisms of their applications (Fig. 2):

Figure 2. ID tags complement and expand the use of folded tags. Several functions are unique to folded tags or unfolded ID tags. Auto-fluorescence, enzymatic activity and rigidity in 3D are known features of folded proteins. ID tags permit tuning of size enlargement (super-sizing), metal affinity or aggregation-propensity. ID tags and ordered tags (O tags) overlap in many applications including protein purification, protein binding and enzyme-substrate interactions.

Folded tags excel in the following features: 1. Rigidity in 3D 2. Specific enzymatic functions 3. Auto-fluorescence

ID tags, on the other hand, permit tuning: 1. enlargement (“super-sizing”) compared with compact globular proteins 2. metal affinity from rapidly reversible binding (His-tag) up to femtomolar affinity (Zn hook) 3. thermal aggregation behavior

ID tags and folded/ordered tags overlap in: 1. Protein solubilisation 2. Protein binding (reversible to covalent) 3. Enzyme-substrate interactions

In a nutshell, ID tags are invaluable additions to the toolbox of protein sciences. Future applications of ID tags will be catalyzed by the rapid growth of IDP research.,, In nature, ID tails have evolutionary diverged into multifarious roles., We can learn from nature’s great example and create novel ID tails to harness their unique features for new purposes.