Fusion protein linkers: property, design and functionality

Chen et al. (2012). Adv Drug Deliv Rev DOI: 10.1016/j.addr.2012.09.039 Citations: 2179

Linker choice in recombinant fusion proteins was often neglected and rarely systematized. This review sorts empirical linkers into flexible, rigid and in vivo cleavable classes and reports that a 50-residue helical linker raised G-CSF-transferrin bioactivity about 10-fold.

Key findings

  • Natural linkers between domains of multi-domain proteins are short and become more hydrophilic as they lengthen. In a database of 1280 natural linkers the mean length is 10.0 ± 5.8 residues (small, medium and large classes average 4.5 ± 0.7, 9.1 ± 2.4 and 21.0 ± 7.6 residues), whereas a smaller set of 51 linkers averages 6.5 residues. Solvent accessibility rises and hydrophobicity falls with increasing linker length, so longer linkers are more exposed to solvent.

  • Natural interdomain linkers are enriched in Pro, Thr and Gln in both of two independent surveys of natural multi-domain proteins. Pro has a cyclic side chain and no amide hydrogen, which restricts its conformation and limits hydrogen bonding with neighbouring domains, so including Pro may increase the stiffness and structural independence of a linker. Natural Pro-rich interdomain linkers include the pyruvate dehydrogenase linker between the lipoyl and E3-binding domains and the P9 linker between the central and C-terminal domains of a cysteine proteinase.

  • In a survey of 1280 natural linkers, 38.3% adopt an alpha-helical and 37.6% a coil or bend secondary structure, and 8.4% adopt turns, while in a survey of 51 natural linkers 59% adopt a coil structure. Helical linkers are rigid spacers stabilized by intra-segment hydrogen bonds, and non-helical linkers tend to be Pro-rich and relatively rigid, so natural linkers mainly adopt extended conformations that do not interact with the adjacent domains.

  • The most widely used flexible linker is the Gly-Ser repeat (GGGGS)n, built from small Gly residues that give mobility and polar Ser residues that hydrogen-bond to water and keep the linker soluble in aqueous solution. The copy number n sets the linker length, so the separation between the joined domains, or the interdomain interaction needed, can be tuned.

  • A (GGGGS)3 flexible linker, calculated to span about 5.7 nm, bridges the 3.5 nm distance between the C terminus of the antibody heavy-chain variable region (VH) and the N terminus of the light-chain variable region (VL) in their natural orientation. It allows a correctly oriented single-chain variable fragment (scFv) without interfering with domain folding, and it has since been reused in many other scFvs.

  • The flexible linkers KESGSVSSEQLAQFRSLD and EGKSSGSGSESKST, designed by computational search of peptide structures from the Protein Data Bank to span the VL C terminus to the VH N terminus, support a bioactive scFv. Their Gly and Ser residues provide flexibility and their added Glu and Lys residues improve solubility.

  • An eight-glycine (Gly)8 linker inserted between a Myc epitope tag and Est2p in Saccharomyces cerevisiae is stable against proteolytic digestion during purification and improves the in vivo function of the epitope-tagged Est2p and of several other epitope-tagged telomere-maintenance proteins. The improvement is attributed to correct folding of the tagged protein and reduced steric hindrance between the tag and the protein.

  • The Gly-Ser linker GSAGSAAGSGEF, which avoids large hydrophobic residues, gives a GFP folding-reporter fusion performance similar to a longer (GGGGS)4 linker. Its DNA sequence lacks the highly homologous repeats of (GGGGS)4, so it is less likely to be deleted by homologous recombination during DNA shuffling and cloning.

  • A flexible (GGGGS)3 linker does not always rescue a fusion. A transferrin (Tf) to granulocyte colony-stimulating factor (G-CSF) fusion in the Tf-G-CSF orientation is almost undetectable in expression even with a (GGGGS)3 linker, and a protein G-Vargula luciferase fusion does not recover immunoglobulin binding after insertion of a single GGGGS linker. The failures are attributed to inefficient separation of the domains or insufficient reduction of their interference with each other.

  • An alpha-helical A(EAAAK)nA linker (n = 2 to 5) is stabilized by Glu-/Lys+ salt bridges within its segments. Inserted between enhanced blue fluorescent protein (EBFP) and EGFP, FRET efficiency falls as the number of EAAAK repeats rises, and is much lower than with a flexible linker of the same length, indicating that helical linkers separate the domains more effectively. Synchrotron X-ray small-angle scattering of variably linked chimeric proteins suggests the chimera with the helical linker is more elongated than the one with a flexible linker.

  • Pro-rich linkers of the form (XP)n, where X is preferably Ala, Lys or Glu, are rigid and extended. 1H-NMR shows that the (Ala-Pro)7 dipeptide repeat in the N-terminal alkali light chain of skeletal muscle is extended and rigid, and 33-residue peptides of repeating Glu-Pro or Lys-Pro are elongated and stiff.

  • Inserting the 50-residue helical linker (H4)2, A(EAAAK)4ALEA(EAAAK)4A, between the domains raises expression in transiently transfected HEK293 cells. hGH-(H4)2-Tf is expressed 1.66-fold and Tf-(H4)2-hGH 2.39-fold higher than the respective fusion without a linker, G-CSF-(H4)2-Tf 1.44-fold higher than G-CSF-Tf, and Tf-(H4)2-G-CSF 11.2-fold higher than Tf-G-CSF, whose expression is almost undetectable.

  • The expression gain from the helical (H4)2 linker is sequence- and structure-specific. A linker encoded by the reversed oligonucleotide sequence of (H4)2, which gives a peptide of identical length without a helical structure, does not improve expression of the Tf fusion proteins.

  • A direct fusion of G-CSF to Tf retains less than 10% of the bioactivity of parent G-CSF. Inserting a helical linker [A(EAAAK)nA]m (n = 2 to 4, m = 1 or 2) or a flexible (GGGGS)3 linker improves in vitro G-CSF activity, and the helical (H4)2 linker, A(EAAAK)4ALEA(EAAAK)4A, gives the highest activity, about 10-fold higher than the fusion without a linker.

  • The helical (H4)2 linker, A(EAAAK)4ALEA(EAAAK)4A, between human growth hormone (hGH) and Tf improves in vitro hGH bioactivity of hGH-Tf. Orally administered hGH-(H4)2-Tf elicits significant body weight gain in hypophysectomized rats, the hGH response endpoint, whereas hGH-Tf without the linker shows no pharmacological effect.

  • In a human serum albumin (HSA) to interferon-alpha2b (IFN-alpha2b) fusion, a flexible GGGGS, a Pro-rich PAPAP or a helical AEAAAKEAAAKA linker increases antiviral activity by 39%, 68% and 115%, respectively, compared with the fusion without a linker. Without a linker, the disulfide bond between two cysteines of IFN-alpha2b cannot form because of interference from HSA, and each of the three linkers restores disulfide formation and gives a homogeneous product on non-reducing SDS-PAGE.

  • Bioactivity of an enzyme-chaperone fusion rises with the length of a flexible (GGGGS)n linker. In a penicillin amidase to rTHS chaperone chimera with n = 1, 2 or 4, activity increases with n and is almost equivalent to that of the parent enzyme with the longest linker.

  • Bioactivity of an interferon-gamma-gp120 fusion rises with the length of a rigid (Ala-Pro)n linker. A 10-residue linker gives relatively low interferon-gamma activity, activity increases gradually with length and peaks at 88% of free interferon-gamma with a 34-residue linker, whereas a (Cys-Trp)n linker of the same length does not improve activity.

  • A 15-residue (EAAAK)3 helical linker allows a 133-residue protein A fragment to be displayed on the viral coat protein of a rod-shaped plant virus with formation of functional virions, whereas otherwise only peptides shorter than 20 residues can be expressed on the coat protein of plant rod-shaped viruses without preventing assembly of functional virions. The rigid helical linker is thought to give the fusion enough distance from the coat protein to fold correctly.

  • In bifunctional beta-glucanase to xylanase fusion enzymes, helical (EAAAK)n linkers (n = 1 to 3) and a flexible linker significantly improve stability and catalytic activity. The thermal stability of beta-glucanase in the fusion improves as the helical linker lengthens (higher n), and helical linkers appear more efficient than flexible linkers at increasing thermal stability.

  • A cyclopeptide linker LEAGCKNFFPR|SFTSCGSLE, which contains a disulfide bond between its two Cys residues and a thrombin-sensitive PRS sequence between them, joins G-CSF and Tf. In vitro thrombin treatment cleaves the thrombin-sensitive sequence but leaves the two domains joined by the reversible disulfide. After intravenous injection in CF1 mice, free G-CSF is released from the thrombin-treated, disulfide-linked G-CSF-Tf fusion as early as 5 minutes and peaks at about 15 minutes, whereas no free G-CSF is detected from a G-CSF-Tf fusion with a stable peptide linker.

  • Reducing the disulfide of the cleavable cyclopeptide linker LEAGCKNFFPRSFTSCGSLE in a G-CSF-Tf fusion with dithiothreitol releases the G-CSF domain and gives a 2-fold increase in in vitro G-CSF bioactivity and a much higher maximal response than the intact covalently linked fusion. The gain may be due to reduced G-CSF receptor blockage once G-CSF is separated from Tf.

  • The dithiocyclopeptide linker CRRRRRREAEAC joins interferon-alpha2b (IFN-alpha2b) and HSA through a disulfide between its two Cys residues and is cleaved by the Pichia pastoris secretory proteases Kex2 at CRRRRRR|EAEAC and then Kex1 and Ste13. The residues between the two Cys are removed during secretion, so the disulfide-linked IFN-alpha2b-HSA fusion is expressed directly.

  • Cleavable linkers raise the in vivo antiviral efficacy of IFN-alpha2b-HSA although they lower exposure. The area under the antiviral-activity curve increases by 450% for IFN-RKRR-HSA, 25% for IFN-RR-HSA and 47% for IFN-SS-HSA (disulfide linker) relative to the fusion without a linker, even though the area under the concentration curve is generally lower, consistent with release of short-half-life free IFN-alpha2b.

  • A coagulation factor IX (FIX) to albumin fusion with a flexible Gly-Ser linker has poor FIX activity, probably because albumin blocks FIX interaction with other coagulation factors. A linker VSQTSKLT|RAETVFPDV derived from the FIX activation region is cleaved together with the FIX activation sites during clotting by tissue factor/Factor VIIa or Factor XIa, which raises clotting activity 10- to 30-fold and leaves FIX inactive with a prolonged half-life before cleavage.

  • A linker of two flexible GGGGS sequences flanking the MMP-1/9 cleavage sequence PLGLWA joins an interferon (IFN) to the latency-associated peptide (LAP) of transforming growth factor beta1. Because the linker is cleaved only by matrix metalloproteinases overexpressed at disease sites, the IFN stays masked by LAP in the circulation, and the LAP-IFN fusion has a plasma half-life of 55 hours, 37-fold longer than that of native IFN.

  • A C-terminal 41-residue fragment of the E. coli antitoxin MazE fused to the N terminus of the toxin MazF by a linker with the HIV protease site RVL|AEA, the HCV NS3 protease site EDVVCC|SMSY or the Factor Xa site GGIEGR|GS keeps MazF inactive. In vitro incubation with HIV protease, HCV NS3 protease or Factor Xa, respectively, releases MazF, which then shows mRNA interferase activity.

  • Furin, a cellular endoproteinase located mainly in the trans-Golgi network that recycles between the trans-Golgi network, early endosomes and the cell surface, cleaves the consensus Arg-X-Arg/Lys-Arg|. A furin-sensitive linker TRHRQPR|GWEQL is cleaved efficiently in vitro in a restrictocin ribotoxin-scFv immunotoxin, and the immunotoxin with this linker kills target cells 2- to 30-fold more strongly than the immunotoxin without a linker on various cell lines.

  • A synthetic polyarginine furin site and the furin sites TRHRQPR|GWE and AGNRVRR|SVG (from Pseudomonas exotoxin A and diphtheria toxin) in scFv-caspase-3 and scFv-granzyme B immunoproapoptotic proteins allow selective binding to target tumor cells, cleavage at the furin site within the endosome, translocation to the cytosol, cell death and reduced tumor size in nude mice.

  • A cathepsin B-sensitive GFLG peptide combined with the furin sequence R2KR6 links a tumor-targeting moiety (a fragment of Clostridium perfringens enterotoxin) to the toxin recombinant gelonin, so that the toxin is released in the lysosome. Cathepsin B, a lysosomal protease, also cleaves a Phe-Lys dipeptide linker in an albumin-binding doxorubicin prodrug.

  • In hGH-Tf fusion proteins, the linker sets receptor binding affinity and plasma half-life. In competitive receptor binding assays the short dipeptide LE linker gives the weakest binding (IC50 17.7 nM for the hGH receptor and 21.2 nM for the Tf receptor) compared with the cyclic disulfide linker LEAGCKNFFPRSFTSCGSLE (8.2 and 4.2 nM) and the helical (H4)2 linker (7.0 and 8.7 nM), and the weakest-binding hGH-LE-Tf has the longest plasma half-life (4.97 ± 0.34 h versus 1.76 ± 0.27 h for the cyclic linker and 1.87 ± 0.44 h for (H4)2).

  • In hGH-Tf fusion proteins, hGH receptor binding leads to endocytosis and lysosomal degradation, whereas Tf receptor binding leads to recycling. Blocking hGH receptor binding with excess hGH prolongs plasma half-life of the dipeptide-linked, cyclic disulfide-linked and (H4)2-linked fusions to 5.95, 8.66 and 6.73 h, and blocking the Tf receptor with excess Tf shortens the half-life of the dipeptide-linked fusion to 3.00 h.

  • The LINKER program generates fusion-protein linker sequences from a database of 14,734 loop sequences taken from X-ray crystal and NMR structures. Its inputs are the desired length (residues or angstroms), protease or restriction sites to avoid and amino acid composition preferences, and the IBIVU linker database at Vrije University of Amsterdam is searchable by PDB code, linker length, solvent accessibility, secondary structure or sequence.

  • Direct fusion of functional domains without a linker can cause misfolding of the fusion protein, low production yield or impaired bioactivity. Reduction of a disulfide linker occurs rapidly in the blood circulation, whereas protease-sensitive linkers are cleaved more slowly at constrained compartments, which permits site-targeted activation of a fusion protein.

  • Recombinant fusion linkers fall into three classes. Flexible linkers such as (GGGGS)n and (G)n are rich in small or hydrophilic residues and allow domain interaction or movement, rigid linkers such as (EAAAK)n and (XP)n maintain a fixed distance through a helical or Pro-rich structure, and cleavable linkers such as disulfide or protease-sensitive sequences allow in vivo separation of the domains. Flexible and rigid linkers are stable in vivo and do not release the joined proteins.

1 of 35

Abstract

As an indispensable component of recombinant fusion proteins, linkers have shown increasing importance in the construction of stable, bioactive fusion proteins. This review covers the current knowledge of fusion protein linkers and summarizes examples for their design and application. The general properties of linkers derived from naturally-occurring multi-domain proteins can be considered as the foundation in linker design. Empirical linkers designed by researchers are generally classified into 3 categories according to their structures: flexible linkers, rigid linkers, and in vivo cleavable linkers. Besides the basic role in linking the functional domains together (as in flexible and rigid linkers) or releasing free functional domain in vivo (as in in vivo cleavable linkers), linkers may offer many other advantages for the production of fusion proteins, such as improving biological activity, increasing expression yield, and achieving desirable pharmacokinetic profiles.