Live long and active: Polypeptide-mediated assembly of antibody variable fragments

Abstract

Antibodies possess multiple biologically relevant features that have been engineered into new therapeutic formats. Two examples include the adaptable specificity of their variable (Fv) region and the extension of plasma circulation times through their crystallizable (Fc) region. Since the invention of the single chain variable fragment (scFv) in 1988, antibody variable regions have been re-engineered into a wide variety of multifunctional nanostructures. Among these strategies, peptide-mediated self-assembly of variable regions through heterologous expression has become a powerful method to produce homogenous, functional biomaterials. This manuscript reviews recent reports of antibody fragments assembled through fusion with peptides and proteins, including elastin-like polypeptides (ELPs), collagen-like polypeptides (CLPs), albumin, transmembrane proteins, leucine zippers, silk protein, and viruses. This review further discusses the current clinical status of engineered antibody fragments and challenges to overcome.

1. Introduction

Coupled with advancements in antibody engineering, peptides adopted from human and non-human sources have become powerful platforms to generate supramolecular antibody complexes that are starting to address needs unmet by conventional antibody therapeutics. Biomaterials that have been investigated for supramolecular antibody fragment assembly include: i) bioinspired recombinant polypeptides, such as elastin-like polypeptides (ELPs), silk-like polypeptides (SLPs), or collagen-like polypeptides (CLPs); ii) hybrid block copolymers containing polypeptides that produce micellar, tubular, or vesicular structures; iii) endogenous human proteins, such as human serum albumin, transmembrane proteins or heterodimeric coiled-coil domains; and iv) exogenous sources such as viral proteins [1]. Although these biomaterials vary in shape, size, conformation, responsiveness to external stimuli, and pharmacology, they possess common features: i) high fidelity components produced by ribosomal biosynthesis; ii) biocompatible and biodegradable; iii) tertiary and quaternary structure promote supramolecular assembly; and iv) tunable structure-activity relationship by genetic or post-translational engineering [2]. Further advancements in their engineering and manufacturing are gradually allowing progression of antibody fragment-biomaterial conjugates into the clinic [3].

In line with efforts to functionalize biomaterials with engineered antibody fragments, this review intends to provide insights in three ways. First, general characteristics of antibodies, their functional derivatives, and a complete list of fragment-based therapeutic antibodies in clinical trials are discussed. Second, features of biomaterials that promote supramolecular assembly of antibody fragments are discussed with the hope that their features may improve current modalities under investigation, diversify future applications, and overcome limitations of traditional antibodies. For this, a panel of polypeptide/protein platforms are examined that range from the proof-of-concept stage to early phases of human studies. This list includes recombinant polypeptides that are inspired by consensus motifs identified in human endogenous proteins, and natural proteinsin human, yeast, silkworms, and viruses. Third, important biophysical characteristics are discussed that may enable antibody fragment-biomaterial conjugates to become clinically relevant drugs.

Our discussion begins below with the general characteristics of antibodies and their functional derivatives, how multivalency affects antibody fragment engineering, and a complete list of fragment-based antibodies in clinical trials.

2. Engineering antibody fragments for therapeutics

Antibodies are utilized by the innate immune system to identify and neutralize foreign entities. There are five naturally occurring antibody classes: monomeric IgD, IgE, IgG, dimeric IgA, and pentameric IgM, from which the IgG class is the most abundant (>80%) in the systemic circulation (Fig. 1A) [4]. Myriad fragment-based products that maintain the parent IgG's specificity have been developed and exploited as therapeutics (Table 1) or as a targeting agent for nano-formulations. These include monovalent Fab (C_H1-V_H paired with C_L-V_L) or scFv (single chain variable fragment, V_H paired with V_L via short polypeptide linker) as well as bivalent, trivalent and tetravalent derivatives, that engage multiple targets, such as Fab₂ (bispecific), bis-scFv (bispecific), diabody (bispecific), Fab₃ (trispecific), minibody (bivalent), triabody (trivalent) and tetrabody (tetravalent) [5]. Distinct antibodies discovered from cartilaginous fishes and camels, such as Ig-NAR (immunoglobulin new antigen receptor, found in sharks) and hcIgG (heavy chain IgG, found in camels, llamas, and alpacas), have also been explored owing to their simple structure (yet equivalent affinity) compared to human scFv (Fig. 1A) [6,7].

Fig. 1.

Antibody fragments subjected to engineering. A. Different antibody formats identified in human and non-human vertebrates. Conventional antibody (Conv. Ab) found in humans, heavy chain–only antibody (HCAb or hcIgG) found in camelids, and immunoglobulin new antigen receptor (IgNAR) found in sharks. B. The role of glycosylation in immune response. The composition of the glycan tree mediates phagocytosis, superoxide production, cytokine release, antibody-dependent cellular cytotoxicity (ADCC), inflammatory mediator secretion and further immune cell attraction through modified Fc-FcγR interactions. Figures reproduced from refs [19,20] with permission.

Table 1.

Fragment-based antibody therapeutics approved by the USFDA or in clinical trials.

INN (trade name)	Target	Fragment type	Primary indication	Approval year/Phase (NCT#)
Brolucizumab (Beovu®)	VEGF-A	scFv	Wet Age-related Macular Degeneration	2019
Caplacizumab-yhdp (Cablivi®)	Von-Willebrand Factor	Nanobody	Acquired Thrombotic Thrombocytopenic Purpura	2019
Moxetumomab pasudotox-tdfk	CD22	scFv-Pseudomonas Exotoxin PE38 fusion	Relapsed/Refractory Hairy Cell Leukemia	2018
Idarucizumab (Praxbind®)	Dabigatran	Fab	Reversal of Dabigatran(anticoagulant)	2015
Blinatumomab (Blincyto®)	CD19, CD3	BiTE	B Cell precursor Acute Lymphoblastic Leukemia (ALL)	2014
Certolizumab pegol (Cimiza®)	TNFα	Fab	Crohn's Disease	2008
Ranibizumab (Lucentis®)	VEGF-A	Fab	Wet Age-related Macular Degeneration	2006
Abciximab (ReoPro®)	GP IIb/IIIa	Fab	Unstable Angina	1994
Ozoralizumab	TNF	Nanobody	Rheumatoid Arthritis	III (NCT04077567)
Vicinium/Oportuzumab monatox	EpCAM	scFv-P.Aeruginosa Exotoxin A fusion	Non Muscle Invasive Bladder Cancer	III (NCT02449239)
Daromun	Extra-domain B of fibronectin (L19)	Combination of scFv-IL2 fusion (Darleukin) and scFv-TNFα fusion (Fibromun)	Stage IIIb/c Melanoma	III (NCT03567889/NCT02938299)
Lampalizumab	Complement Factor D	Fab	Geographic Atrophy secondary to Age-related Macular Degeneration	III (NCT02247531/NCT02247479)
Dapirolizumab-pegol	CD40 ligand	Fab-peg	Systemic Lupus Erythematosus	III (NCT04294667)
MGD013	PD-1, LAG-3	DART	Unresectable or Metastatic Neoplasm	III/II (NCT04082364)I (NCT03219268)
Naptumomab estafenatox	5T4 Tumor Antigen	Fab-Staphylococcal Enterotoxin A fusion	Advanced Renal Cell Carcinoma, Advanced or Metastatic Solid Tumor	III/II (NCT00420888)I (NCT03983954)
M1095/ALX-0761	IL-17a,f	Bispecific Nanobody	Psoriasis	IIb (NCT03384745)
Vobarilizumab	IL-6r	Nanobody	Rheumatoid Arthritis	IIb (NCT02287922)
V565	TNFα	Nanobody	Crohn's Disease	II (NCT02976129)
VHH batch 203027	Rotavirus	Nanobody	Rotavirus Diarrhea	II (NCT01259765)
68GaNOTA- VHH	HER2	Nanobody	Breast Carcinoma, Brain Metastasis of Breast Carcinoma	II (NCT03924466/NCT03331601))
Flotetuzumab/MGD006	CD123, CD3	DART	Acute Myeloid Leukemia (AML)	II (NCT03739606) I/II (NCT02152956)
AFM13	CD30, CD16A	Tetravalent bsAb	CD30 Positive T-cell Lymphoma, Hodgkin's Lymphoma (HL)	II (NCT04101331)
Istiratumab/MM-141	IGF-1Rb, HER3	Tetravalent bsAb	Metastatic Pancreatic Cancer	II (NCT02399137)
MM- 111	HER2, HER3	Bispecific scFv	HER2 Positive Carcinomas of Distal Esophagus, Gastroesophageal Junction and Stomach	II (NCT01774851)
Gremubamab/MEDI-3902	PcrV, Psl	Fab2-scFv-Fc	Nosocomial Pneumonia	IIb (NCT02696902)
MT-3724	CD20	scFv-Shiga-like toxin 1A fusion	Relapsed/Refractory Diffuse Large B Cell Non-Hodgkin's Lymphoma (NHL)	II (NCT03488251/NCT02361346)
IMCgp100	Gp100, CD3	ImmTAC	Advanced Uveal Melanoma	II (NCT03070392)
IMCnyeso	NY-ESO-1, CD3	ImmTAC	NY-ESO-1 and/or LARGE-1A Positive Cancer	II/I (NCT03515551)
OXS-3550/GTB-3550	CD16, CD33	TriKE, scFv-IL-15-scFv fusion	High Risk Myelodysplastic Syndrome, Refractory/Relapsed AML, Advanced Systemic Mastocytosis	II/I (NCT03214666)
OXS-1550/DT2219ARL	CD19, CD22	BiKE, scFv-scFv-Diphtheria Toxin fusion	Relapsed/Refractory B-Lineage Leukemia, Lymphoma	II/I (NCT02370160)
CD19/20 CAR-T cell	CD19, CD20	Bispecific Nanobody derived CAR-T cell	B-cell Lymphoma	I (NCT03881761)
M6495	ADAMTS-5	Nanobody	Knee Osteoarthritis	I (NCT03583346)
ALX-0651	CXCR4	Nanobody	Hematopoietic Stem Cell Transplant	I (NCT01374503)
PF-05230905	TNFα	Nanobody	Rheumatoid Arthritis	I (NCT01284036)
BI 836880	VEGF, ANG2	Bispecific Nanobody	Wet Age-related Macular Degeneration	I (NCT03861234, NCT03468426)
Solitomab/MT110/AMG 110	EpCAM, CD3	BiTE	Lung Cancer, Gastric Cancer, Adenocarcinoma of Gastro-esophageal Junction, Colorectal Cancer, Breast Cancer, Hormone-Refractory Prostate Cancer, and Ovarian Cancer	I (NCT00635596)
AMG 160	PSMA, CD3	BiTE	Metastatic Castration-resistant Prostate Cancer	I (NCT03792841)
AMG 199	MU-17 CD3	BiTE	Metastatic Gastric and Gastroesophageal Junction Cancer	I (NCT04117958)
MEDI-565/AMG 211	Human CEA, CD3	BiTE	Gastrointestinal Adenocarcinoma	I (NCT02760199)
Pasotuxizumab/BAY 2010112/AMG 212	PSMA, CD3	BiTE	Prostate Cancer	1 (NCT01723475)
AMG 330	CD 33, CD3	BiTE	Relapsed/Refractory AML	I (NCT02520427)
AMG 420 M	BCMA, CD3	BiTE	Relapsed/Refractory Multiple Myeloma	I (NCT03836053)
AMG 562	CD19, CD3	BiTE	Diffuse Large B-cell Lymphoma, Mantle Cell Lymphoma, Follicular Lymphoma	I (NCT03571828)
AMG 596	EFGRvIII, CD3	BiTE	Glioblastoma	I (NCT03296696)
AMG 673	CD33, CD3	BiTE	Relapsed/Refractory AML	I (NCT03224819)
AMG 701	BCMA, CD3	BiTE	Multiple Myeloma	I (NCT03287908)
AMG 757	DLL3, CD3	BiTE	Small Cell Lung Cancer	I (NCT03319940)
BFCR 4350A	FCRH5, CD3	BiTE	Relapsed/Refractory Multiple Myeloma	I (NCT03275103)
AFM 11	CD19, CD3	Tetravalent bsAb	Relapsed/Refractory NHL	I (NCT02106091)
MGD 007	gpA33, CD3	DART	Relapsed/Refractory Metastatic Colorectal Carcinoma	I (NCT02248805)
MGD 009	B7-H3, CD3	DART	Relapsed/Refractory B7-H3 expressing tumor	I (NCT03406949)
MGD 010/PRV 3279	CD32B, CD79B	DART	B Cell Mediated Autoimmune Disorder	Ib (NCT03955666)
Duvortuxizumab/MGD 011	CD19, CD3	DART	Relapsed/Refractory B Cell Malignancy	I (NCT02454270)
PF-06671008	P-Cadherin, CD3	DART	P Cadherin expressing TNBC, CRC, or NSCLC	I (NCT02659631)
AMV 564	CD33,CD3	TandAb	Relapsed/Refractory AML	I (NCT03144245)

Information retrieved from www.fda.gov and www.clinicaltrials.gov.

BiKE = Bi-specific Killer Engager; BiTE = Bispecific T Cell Engager; DART = Dual Affinity Re-Targeting (Di-scFv); ImmTAC = T Cell Receptor (TCR)-scFv fusion; INN = International Nonproprietary Name; Nanobody = V_HH; TandAb = Tandem Diabody; TriKE = Tri-specific Killer Engager.

Not only the variable (Fv) region, but also the crystallizable (Fc) region of an antibody has been a target of engineering. Two rationales for Fc engineering are to improve pharmacokinetics (Fc-FcRn interaction, increase/decrease systemic half-life) or to optimize pharmacodynamics (Fc-FcγR interaction, enhance/decrease recruitment of FcγR-bearing immune effector units). Two main ways to achieve these are amino acid substitutions and glycoengineering [8]. It is relatively straightforward to substitute amino acids at the genetic level; however, the choice of the host cell that produces the optimal glycoform is crucial to generate the glycosylation pattern that elicits downstream cellular and humoral immune responses at a desired level (Fig. 1B). Currently, USFDA approved antibodies or Fc-fusions (Table 2) with nonimmunogenic glycoforms are produced using human embryonic kidney (HEK), NSO hybridoma (cell line derived from non-secreting murine myeloma), Sp2/0 (hybrid cell line of BALB/c mouse spleen cell and mouse myeloma P3X63AG8), or Chinese hamster ovary (CHO) cell lines. These cell lines produce slightly different glycoforms, which result in different pharmacodynamics.

Table 2.

Fc-fusion therapeutics approved by the USFDA or in clinical development.

INN (Trade name)	Target	Therapeutic construct	Primary indication	Approval year/NCT #
Luspatercept-aamt (Rebrozyl®)	TFG-β ligand	Human activin receptor type IIB-Fc(IgG1)	Anemia with Beta Thalassemia	2019
Asfotase alfa (Strensiq®)	Factor substitute	Human Tissue-nonspecific Alkaline Phosphatase-Fc(IgG1)	Hypophospatasia	2015
Dulaglutide (Trulicity®)	GLP-1 receptor	Glucagon-like peptide-1 (GLP-1) analog-Fc (IgG4)	Type II Diabetes	2014
Antihemophilic Factor Recombinant FcFusion Protein (Eloctate®)	Coagulation	FVIII-Fc(IgG1)	Hemophilia A	2014
Alprolix (Alprolix®)	Coagulation	FIX-Fc(IgG1)	Hemophilia B	2014
Ziv-aflibercept (Zaltrap®)	VEGF-A, VEGF-B, PIGF	VEGFR1-VEGFR2-Fc(IgG1)	Colorectal cancer	2012
Aflibercept (Eylea®)	VEGF-A, VEGF-B, PIGF	VEGFR1-VEGFR2-Fc(IgG1)	Wet Age-related Macular Degeneration	2011
Belatacept (Nulojix®)	CD80, CD86	CTLA-Fc(IgG1) (Two amino acid substitutions, A29Y and L104E, on CTLA-4 from abatacept)	Transplant rejection	2011
Rilonacept (Arcalyst®)	IL-1α,β,RA	IL-1R-Fc(IgG1)	Cryopyrin-Associated Periodic Syndrome (CAPS)	2008
Romiplostim (NPlate®)	Thrombopoietin receptor (CD110)	Thrombopoietin binding peptide-Fc(IgG1)	Refractory Immune Thrombocytopenia	2008
Abatacept (Orencia®)	CD80, CD86	CTLA-4-Fc(IgG1)	Rheumatoid Arthritis	2005
Alefacept (Amevive®)	CD2	CD58 (LFA-3)-Fc(IgG1)	Plaque Psoriasis, Transplant rejection	2003 (Withdrawn from market in 2011)
Etanercept (Enbrel®)	TNFα (Both soluble and membrane bound)	TNFR2-Fc(IgG1)	Rheumatoid Arthritis, Plaque Psoriasis	1998
Trebananib/AMG 386	Ang1, Ang2	TIE2 mimetic peptide-Fc(IgG1)	Ovarian Cancer, Primary Peritoneal Cancer, Fallopian Tube Cancer	III (NCT01204749, NCT01281254, NCT01493505)
Efgartimimod/ARGX-113	FcRn	Fc(IgG1)	Myasthenia Gravis	III (NCT03770403)
Blisibimod/AMG 623	BAFF	BAFF-Fc(IgG1)	Systemic Lupus Erythematosus	III (NCT02514967, NCT01395745)II/III (NCT02062684)
Atacicept	B Lymphocyte, stimulator (BLys), A proliferation-inducing ligand (APRIL)	BLys-APRIL-Fc(IgG1)	Systemic Lupus Erythematosus	II/III (NCT00624338)
ALT-803	IL-15	IL-15-Fc(IgG1)	Acute Myeloid Leukemia	II (NCT03050216) (35 other Phase II Studies)
Asunercept/APG101	CD95 ligand	CD95-Fc(IgG1)	Glioblastoma Multiforme, Myelodysplastic Syndrome	Orphan Drug status, II (NCT01071837)
SBI-087	CD20	Small modular immunopharmaceutical (SMIP) anti-CD20 scFv-Fc(IgG1)	Rheumatoid Arthritis, Systemic Lupus Erythematosus	II/I (NCT01008852, NCT00714116)
TRU-015	CD20	SMIP anti-CD20 scFv-Fc(IgG1)	Rheumatoid Arthritis	II (NCT00634933)
Otlertuzumab/TRU-016	CD37	anti-CD37 scFv-Fc(IgG1)	Relapsed Indolent Lymphoma	II (NCT01317901)
AMG 171	GDF15	GDF15-Fc(IgG1)	Obesity	I (NCT04199351)
Efavaleukin alfa/AMG 592	IL-2	IL-2-Fc(IgG1)	Lupus Erythematosus, Rheumatoid Arthritis	I (NCT03451422, NCT03410056)
CNTO 528	Erythropoietin receptor	Erythropoietin mimetic peptide (EMP1)-Fc (IgG1)	Anemia	I [17,18]

Information retrieved from www.fda.gov.

With the maturation of technology to engineer antibody variable fragments, methods have been explored to develop multivalent antibodies that maximize their avidity, target-clustering potential, and multi-target specificity. Multivalency enables tighter binding (avidity) to the target through multiple antibody-ligand interactions. Compared to monovalent constructs, multivalent constructs may increase apparent binding affinity, from several fold to several hundred fold, which can improve therapeutic efficacy [9]. In addition to avidity, multivalency enables receptor clustering at the cell surface. As monovalent constructs may require an additional secondary molecule to cluster receptors, multivalency-mediated receptor clustering can be more effective in initiating therapeutically relevant intracellular signaling upon binding [10]. Moreover, multivalent constructs that are engineered to have multi-target specificity can result in better therapeutic outcome compared to multivalent monospecific constructs. For instance, a bispecific antibody faricimab, that targets and neutralizes both VEGF-A and ANG-2 in the vitreous and subretinal regions in the eyes, shows better suppression in subretinal neovascularization in a pre-clinical animal model compared to monospecific anti-ANG-2 antibody [11,12]. By neutralizing multiple disease factors with one compound, the therapeutic efficacy of faricimab once every 16 weeks in humans showed equivalent efficacy to aflibercept (anti-VEGF) once every 8 weeks or ranibizumab (anti-VEGF) once every 4 weeks, both of which are the standard of care for subretinal neovascularization [13].

One way to generate multivalent antibody constructs is to use peptide-based biomaterials that promote supramolecular assembly, such as oligomerization domains from naturally occurring proteins or bioinspired polypeptides that are capable of oligomerization. Use of these scaffolds to initiate supramolecular assembly has possible advantages compared to multivalent constructs generated by consecutive conjugation of antibody fragments. First, supramolecular assembly may address renal clearance by increasing the size and molecular weight of resulting protein complexes [14]. Incorporating multiple antibody fragments into a single fusion, either at the genetic level or post-translational level, can also slow renal clearance through increased molecular weight and size; however, these fusions commonly suffer the instability of antibody fragments [15]. Due to low-yield expression and protein instability upon conjugation of multiple fragments, antibody fragments advanced into the clinic have thus far been limited to bivalency (Table 1). Second, biomaterials that promote supramolecular assembly of antibody fragments can generate high-valency complexes, which for the purpose of this review are considered more than ten antibodies per complex. As high-valency constructs can be further transformed into the multi-target specific complexes [16], they have the potential to enhance therapy compared to low-valency antibody-based therapeutics. Based on the abovementioned reasons, peptide-based biomaterials possess an untapped potential to generate alternative therapeutic modalities. The following sections discuss peptide-based biomaterials that have been explored as a platform for generating multivalent antibody fragment complexes.

3. Human proteins that mediate self-assembly of antibody variable regions

3.1. Bioinspired recombinant polypeptides

3.1.1. Elastin-like polypeptides (ELPs)

Elastin-Like Polypeptides (ELPs) are an emerging class of protein polymer whose sequence is derived from tropoelastin, the precursor of elastin [21]. Tropoelastin is a soluble monomeric protein that is encoded by a human gene ELN. It makes an insoluble fibrous elastin structure once it is crosslinked via its lysine residues (Fig. 2A,B). Inspired by a hydrophobic repeat motif in tropoelastin (Fig. 2C), ELPs are recombinant polypeptides of pentapeptide repeats (Val-Pro-Gly-X-Gly)_n, where X can be any amino acid residue (guest residue) and n is number of repeats. Since the sequence of ELP polymers are similar to that of the naturally occurring tropoelastin, they appear to be biodegradable, biocompatible, and non-immunogenic [22,23]. One of the most unique characteristics of ELPs is their thermo-responsive phase separation [24]. ELPs phase separate above a transition temperature (T_t), while remaining highly soluble below T_t. The T_t can be precisely controlled by various factors, including molecular weight (n), hydrophilicity of the guest residue (X), ELP concentration, pH, and co-solutes (Supplemental Video S1). There are multiple ways to utilize their thermoresponsiveness, one of which is as a purification strategy [25]. Phase separated ELPs can be collected by centrifugation at above its T_t (hot centrifugation), which promotes protein pellet formation at the bottom of the tube [26]. After discarding soluble impurities (supernatant), pelleted coacervates are resuspended in a cold buffer and insoluble impurities are cleared by another centrifugation below the T_t (cold centrifugation). Repeating 2– 6 rounds of hot and cold centrifugation steps usually yields highly pure ELPs (>95% purity based on SDS-PAGE) [27]. This method, called inverse temperature cycling, is an advantage compared to conventional methods that involve affinity chromatography. Cost-effective expression from bacterial fermentation and ease of genetic modification are other advantages. ELP can be expressed in Escherichia coli, yeast, or plant cells. As peptides, these can be co-expressed in fusion with other biological and therapeutic proteins or peptides, which avoids the need to optimize post-purification bioconjugation strategies. Unconstrained genetic level modification has made ELPs a favorable fusion platform. Examples of functional units include adenovirus knob domain [28], antibody [29,30], peptide inhibitors [31], growth factors [32], receptor antagonist [33], non-invasive imaging tool [34], and intracellular molecules [35].

Fig. 2.

Extracellular Matrix (ECM) environment, the biogenesis of elastin and collagen, and self-assembly of antibody fragments using collagen. A. Primary components of ECM and their interactions. B. Representation of the classical elastogenesis model. The tropoelastin packages are crosslinked by lysyl oxidase upon transport to the extracellular space. Accumulation of elastin to the microfibrillar scaffolds results in elastic fibers. C. Domain structure of the mature human tropoelastin. The hydrophobic domains are shown in light brown. The consensus sequence motif identified in these hydrophobic segments is adopted for ELPs. D. Production of collagen fibrils. Procollagen chains synthesized from the endoplasmic reticulum (ER) are gathered by interactions in the C-propeptide to form a rod-like triple helical domain. After post-translational modifications in the ER, procollagens are transported across the Golgi and plasma membrane to the extracellular space, where truncation of the N- and C-propeptides occurs. The triple-helical collagen forms covalent crosslinks to generate a fibril. The triple-helical region is highly enriched with the GPO repeats. E. Self-assembly of antibody fragment followed by supramolecular assembly via collagen trimerization domain or CLPs (collagen-like polypeptides). F. Schematic showing the C-terminus of the collagen XVIII that consists of the triple helix, non-collagenous (NC1) trimerization domain, hinge region and the endostatin domain. The NC1 trimerization domain is used to trimerize self-assembled antibody fragments. Figures reproduced from refs [21,49,51,55-57] with permission.

While useful as a purification tag, the thermo-responsiveness of ELP fusions has also been explored in drug delivery. Topcic et al. developed a strategy to overcome the barriers with current antiplatelet drug class glycoprotein IIb/IIIa (GPIIb/IIIa) inhibitors that can cause bleeding problems during cardiac surgery [36]. GPIIb/IIIa inhibitors such as abciximab, eptifibatide and tirofiban act by competing with fibrinogen and von Willebrand Factor for GPIIb/IIIa binding, which in turn interferes with platelet cross-linking and induction of a thrombus [37]. During cardiac surgeries, long-duration neurosurgical procedures, or out-of-hospital cardiac arrests, therapeutic hypothermia (deep: <20 °C; moderate: 25– 30 °C; mild: 31– 34 °C) is often incorporated into surgical procedures to downregulate oxygen consumption and to protect organs from ischemia. However, therapeutic hypothermia can induce platelet activation, which leads to thrombocytopenia and postoperative neurological impairment. GPIIb/IIIa inhibitors, especially the monoclonal antibody abciximab, show nonspecific binding to both the active and inactive GPIIb/IIIa receptor. This prolongs antiplatelet activity even after rewarming, which in turn, risks major post-operative bleeding. As a solution, Topcic and coworkers engineered a thermally sensitive scFv-EMP (elastin-mimetic peptide) fusion that only binds to activated GPIIb/IIIa under surgical hypothermia. The authors propose that at 37 °C, the highly ordered β-spiral structure of the EMP prevents the association of scFv to the platelets while the disordered EMP under hyperthermia conditions permits epitope recognition by the scFv. Due to this thermal responsiveness, scFv-EMP fusions showed: i) selective affinity towards activated platelets; ii) blockage of fibrinogen binding to activated platelets; iii) inhibition of ADP-induced platelet aggregation; iv) prolonged arterial occlusion time at hypothermic conditions (below 28 °C) but not at physiological temperature (37 °C) compared to approved drugs, including eptifibatide, tirofiban, or abciximab.

Antibody-ELP fusions sometimes self-assemble into a multivalent nanoparticle. Our group reported an anti-CD20 scFv-ELP fusion in which a single-chain variable fragment (scFv) derived from rituximab was fused to 192 repeats of VPGAG. Surprisingly, these self-assembled into worm-like nanoparticles that are ~80 nm in length and ~10 nm in width. These multivalent ‘Nanoworms’ enhanced apoptosis in CD20 positive Raji and SU-DHL-7 cell lines, which was comparable to that of the rituximab crosslinked with a secondary crosslinking molecule (goat anti-human Fc antibody) [38]. Nanoworm assembly is not restricted to rituximab-derived scFvs but appears to be a common phenomenon for many scFv-ELP fusions [39]. Three scFv-A192 variants, namely anti-CD19-A192 (targeting CD19 in B cells), anti-HLA-DR10-A192 (targeting HLA-DR10 in B cells), and anti-CD3-A192 (targeting CD3 in T cells), all oligomerized into Nanoworms regardless of the origin of scFv. These three additional Nanoworms were capable of activating their respective target receptors by spontaneously clustering them on the cell surface, which induced non-apoptotic cell cycle arrest (in Raji cells), apoptosis (in Raji cells), and activation-induced cell death (in HuT78 cells), respectively. Inspired by the unique activity of these four scFv-ELPs in non-Hodgkin lymphoma models, Nanoworms were further developed for acute myeloid leukemia by targeting cell-surface receptors CD99 or FLT3 (Fms Like Tyrosine kinase 3 or CD135) [40,41]. In CD99 and FLT3 positive MOLM-13 xenografts, both CD99 and FLT3 targeting Nanoworms extended survival compared to ELP controls. When evaluated for pharmacokinetic half-life in mouse plasma, both Nanoworms exhibited surprisingly long terminal half-lives of 15.8 and 14.7 h respectively, which strengthens the potential for scFv-ELP fusions as an alternative therapeutic modality.

As opposed to self-assembly into worm-like or extended chain-like structures using monoblock ELPs, diblock ELPs that have both hydrophobic and hydrophilic domains can also stabilize scFvs into micellar nanoparticles. Using diblock ELP NH₂-(GAGVPG)₇₀-(GVLPGVG)₅₆– (GC)₄, Zhao et al. reported micellar scFv-ELP fusion nanoparticles that block the PD-1 immune checkpoint [29]. The diblock ELP, named amphiphilic immune-tolerant elastin-like polypeptides (iTEP), is composed of a hydrophilic domain (GAGVPG)₇₀, a hydrophobic domain (GVLPGVG)₅₆, and multiple cysteine residues at the C-terminus to stabilize the resulting micellar nanoparticles through disulfide bonds. Anti-PD-1 treatment is known to worsen diabetes by ablating the immune checkpoint [42]. Using this endpoint, this group tested the efficacy of anti-PD-1-iTEP fusion in non-obese diabetic mice (NOD/ShiLtj) and showed equivalent efficacy compared to an anti-PD-1 igG control in exacerbating diabetes.

While E. coli is a commonly exploited expression system for ELP or ELP fusions, Conrad and coworkers have pioneered the use of plant expression system, transgenic tobacco plant Nicotiana tabacum, to express functional antibody-ELP fusions [43]. They fused an anti-human tumor necrosis factor (TNF) nanobody to 100 repeats of VPGXG (X: Gly, Val, and Ala with different ratios) to extend the systemic circulation half-life. The serum half-life of anti-TNF-ELP fusion was extended by 24-fold, from 0.5 h to 11.4 h compared to a non-fusion control. This fusion protein functionally blocked the lethality of LPS/D-gal induced sepsis in an animal model, while maintaining comparable survival to anti-TNF nanobody treated control animals.

Although antibody-ELP fusions developed in pre-clinical studies require significant optimization prior to clinical evaluation, ELPs do seem poised to allow new therapeutic opportunities in the future. Currently, a long-acting vasoactive intestinal peptide (ViP) analogue in fusion with ELP (PB1046) is in a Phase 2b trial in pulmonary arterial hypertension (PAH) patients (NCT03556020, www.phasebio.com). Although it is not an antibody-ELP fusion, clinical approval of the first ELP fusion could transform the potential applications of ELP-based therapeutic modalities, including antibody-ELP fusions.

3.1.2. Collagen-like polypeptides (CLPs) and collagen trimerization domain

Collagen is the most abundant fibrous protein found in the extracellular matrix (Fig. 2A,D). Although there are many types of collagens with different roles in different tissues, a distinctive triple helix conformation characterizes the secondary structure of most collagens [44]. Due to their mechanical properties, collagens are marketed predominantly for tissue engineering [45]; furthermore, their derivatives are explored in fusion with other recombinant polypeptides [46]. The triple helical regions contain repeats of the amino acid motif (Gly-X-Y)_n, where X is mostly proline (Pro, P) and Y is mostly hydroxyproline (Hyp, O) or to a lesser extent, hydroxylysine (Hyl). About 50% of the GXY motif is comprised of GPO, largely owing to its high thermal stability [47]. The high imino acid content (structure that contains an amide group bonded to the alpha carbon on the peptide backbone) and the presence of hydroxyproline stabilizes the triple helix and collagen fibrils as a whole via π-π stacking and hydrogen bonding [48]. Although there are ways to process them to a medical grade material, collagen produced from various biological sources can sometimes cause heterogeneity as well as immunogenicity. As an alternative, collagen-like polypeptides (CLPs) with a signature motif (GPO)_n have gained traction due to their well-defined hierarchical structure, ease of genetic engineering through recombinant technology, and superior resistance to enzymatic degradation.

Several teams have explored the triple helical nature of collagen as a platform to produce multivalent multi-specific antibodies. Fan et al. expressed anti-EGFR scFv-CLP fusion proteins in mammalian cells under the presence of 4-hydroxylase, which hydroxylates prolines (Fig. 2E) [49]. The incorporation efficiency of hydroxyproline in GXY motif was about 60% (maximum is 10 residues in (GPO)₁₀). Expressed scFv-CLP fusions self-assembled into trivalent ‘collabody’ via CLP-mediated trimerization. Attributed to multivalency, the collabody showed a remarkable improvement in equilibrium dissociation constant (K_D). The K_D for the trivalent scFv-CLP was 4.8 nM, which was 1000-fold and 22-fold lower than monovalent scFv (4960 nM) and bivalent scFv-Fc fusion (104 nM), respectively. The serum stability of scFv-CLP was better than an scFv-Fc fusion or scFv alone in that 60% of its binding activity was retained after 72 h incubation at physiological temperature. In the same study, trivalent anti-CD3 scFv-CLP fusion was also generated, which showed a 3-fold reduction in IC₅₀ and better target binding compared to its parent monoclonal antibody OKT3.

The trimerization domain found in non-collagenous (NC) segments in several collagen subtypes was also used to assemble functional antibody fragments. Chain selection and initiation of collagen trimerization rely on the NC1 domain present at the C-terminus of the core triple helix domain [50]. This particular role of the NC1 segment is found in all collagen types except two transmembrane collagens, XIII and XVII [51]. Having no conserved or repeat sequences such as (GPO)_n, the NC1 is stabilized mainly by hydrophobic interactions and hydrogen bonds. Among the identified NC1 domains, the one in collagen XVIII was proposed as an antibody assembly platform for its relatively short sequence (54 amino acids) compared to that of others (Fig. 2F). Álvarez-Vallina and co-workers pioneered the use of this NC1 segment to display antibody fragments. The resultant antibody fusion, named ‘trimerbody,’ formed a non-covalently oligomerized trivalent antibody. In one study, they constructed two types of monospecific trivalent trimerbodies and one bispecific hexavalent trimerbody, whereby the latter is a composite of two trimerbodies [52]. For the monospecific trimerbody, either anti-laminin scFv or anti-CD3 scFv was fused to N- or C-terminus of NC1 segment, generating a trivalent construct that displays three scFvs at the N- or C-terminus of the triple helix. To generate a bispecific hexavalent trimerbody, two different scFvs were fused to the N- and C-terminus of the NC1 segment, respectively. Anti-laminin scFv and anti-CD3 scFv were fused to the N- and C-terminus of NC1, generating three anti-laminin scFvs displayed at the N-terminus of triple helix and three anti-CD3 scFv assembled at its C-terminus. Purified trimerized molecules showed high stability in serum. To test specificity and function, they incubated bispecific hexavalent trimerbodies on laminin-coated plates and then attempted to target and activate T cells. By showing that laminin-immobilized bispecific hexavalent trimerbodies were better at T cell activation compared to an immobilized OKT3 monoclonal antibody, they demonstrated the feasibility of the collagen NC1 segment as a platform to assemble multi-specific antibody fusions.

Following this success, the same group suggested two different constructs for cancer immunotherapy. The first construct that they proposed was named IMTXTRICEAαS, which contained three anti-carcinoembryonic antigen (CEA) scFv and three immunotoxin α-sarcin at the N- and C- terminus of the trimerization domain. Functional studies in CEA positive cells and colorectal cancer xenograft model suggested that the IMTXTRICEAαS had a more promising effect on cell survival and tumor suppression than its monomeric version [53]. The second construct was the anti-EGFR + anti-CD3 scFv bispecific construct called Asymmetric Tandem Trimerbody for T cell Activation and Cancer Killing (ATTACK). The difference between ATTACK and other modalities introduced in this section is the trimerization strategy. While other constructs were generated upon homodimerization via trimerization domains, ATTACK was generated via intra-trimerization. In one DNA vector, three identical cDNA cassettes that encode anti-EGFR nanobodies and the trimerization domain were consecutively cloned followed by the cDNA encoding an anti-CD3 scFv. This resulted in an intra-trimerized tetravalent molecule that is trivalent to EGFR and monovalent to CD3. In a flow cytometry experiment, ATTACK showed better binding to the EGFR receptor and similar binding to CD3 compared to the control molecules, which consists of a single anti-EGFR nanobody linked to a single anti-CD3 scFv [54]. ATTACK was able to activate T cells and redirect them to exhibit cytotoxicity towards EGFR positive HeLa cells, which was 10– 20-fold more potent than the control molecules. Moreover, ATTACK was more potent in inhibiting the growth of EGFR positive A431 cells than the USFDA approved anti-EGFR antibody, cetuximab. EGFR phosphorylation was decreased by 75% for ATTACK, while it was 36% upon cetuximab treatment. These two constructs, IMTXTRICEAαS and ATTACK, show both improved pharmacokinetic properties and encouraging in vivo data, which may promote further development of these multivalent antibodies as therapeutics.

3.2. Human serum albumin (HSA)

Human serum albumin (HSA, 66.5 kDa) is the most abundant protein in the human blood, which comprises approximately half of serum protein. Produced from the liver, it has a simple yet highly stable structure, regulates osmotic pressure of the plasma, and transports various metabolites such as long chain fatty acids, bilirubin, steroid hormones, tryptophan, calcium. Due to its long half-life (20 days) [58] and tendency to accumulate in tumor cells [59], several therapeutic studies have explored albumin as a drug delivery vehicle to modulate the pharmacokinetics of short-lived pharmaceuticals. Interaction between albumin and neonatal Fc receptor (FcRn) on the surface of endothelial cells is the mechanism by which albumin maintains prolonged serum half-life. Its binding occurs in a pH dependent manner. Albumin binds tightly to the FcRn at a slightly acidic condition (pH 6.0) but dissociates from it at pH 7.4. This characteristic allows pinocytosed albumin to tightly associate with FcRn and then to recycle back to the extracellular space, thereby escaping from the late endosome-lysosomal degradation pathway. Albumin or other molecules that contain the Fc region, such as IgG, maintain their half-life up to 2–3 weeks through this mechanism (Fc-FcRn interaction) [60].

Blood flow is irregular in the tumor microenvironment, which is a major barrier for systemic delivery and homogenous distribution of therapeutic compounds inside the tumor. While therapeutic entities that rely on passive diffusion to the tumor through the leaky vasculature may suffer from irregular blood flow, albumin has the ability to overcome this limitation. Albumin, albumin fusions, or albumin-drug conjugates traverse blood vessels (capillary lumen) through transcytosis by binding to albondin (gp60), which is selectively present on the surface of plasma membrane of continuous endothelium (Fig. 3). It has been reported that up to 50% of albumin binds to albondin and transcytose, while the remainder traverse via intercellular junctions or fluid-phase endocytosis. Binding to albondin and subsequent internalization of the whole complex through caveolin-dependent endocytosis promotes evasion of lysosomal degradation and successful translocation to tumor interstitium [61,62]. At the tumor interstitium, albumins or albumin fusions travel further by diffusion and convection with the aid of SPARC (Secreted Protein Acidic and Rich in Cysteine), which is overexpressed in tumor cells to facilitate extracellular matrix proliferation and tumor cell migration. SPARC captures the albumin-gp60 complex through a high-affinity interaction with albumin and facilitates tumor accumulation [63]. SPARC-albumin interaction is thought to enable homogenous intratumoral distribution of albumin or its fusions, which is often limited in untargeted therapeutic formulations or conventional monoclonal antibodies.

Fig. 3.

Transcytosis of the albumin or albumin-drug complex. Transcytosis of albumin or albumin-drug complex upon binding to gp60 allows delivery of albumin-drug complex to tumor interstitium where its binds to SPARC (secreted protein, acidic and rich in ysteine). Albumin-drug complex reaches tumor cells through association and dissociation at the tumor interstitium. Figure reproduced from ref. [61] with permission.

The transportation mechanism of albumin was extensively utilized to improve the in vivo residence of short-lived antibody fragments. McDonagh et al. focused on HER2-HER3 dimerization in HER2-overexpressing tumors, such as breast and gastric carcinoma [64]. In tumor cells, cell surface receptor HER2 and HER3 preferentially dimerize to activate the AKT pathway and maintain a strong oncogenic signal. Heregulin binding to HER3 with high affinity induces tumor growth without inducing overexpression of HER3, mediates the resistance to anticancer agents targeting HER receptor family, and leads to poor prognosis. Due to the lack of an active kinase domain and low expression of HER3, HER2 has been the primary therapeutic target for interrupting the HER2-HER3 heterodimeric complex. Currently available HER2 targeting drugs, trastuzumab and lapatinib, have greatly improved treatment outcome of the HER2 overexpressing patients (~30%); however, a significant proportion of patients (~70%) do not benefit from these drugs due to the lack of HER3 inhibition. To address these limitations, they developed a novel bispecific single polypeptide albumin fusion (MM-111) protein in which one arm binds to HER2, and the other arm binds to HER3. MM-111 showed better inhibition of ligand-induced oncogenic signaling when compared to trastuzumab or lapatinib treatment alone. Furthermore, co-treatment of MM-11 with trastuzumab or lapatinib significantly inhibited tumor growth by suppressing p-AKT signaling. MM-111 also showed extended in vivo half-life. Serum half-life of the MM-111 in the murine model was 16.2– 17.5 h, a significant improvement from ~5 h observed for scFv alone control. MM-111 even achieved a circulating serum half-life of 99 h in cynomolgus monkeys, making it a promising agent as a second-line treatment for HER2-overexpressing tumors. However, despite successful completion of a Phase I clinical trial, the further development of MM-111 as a second-line treatment for metastatic HER2-positive gastric and gastroesophageal junction (gastric/GEJ) cancers was terminated due to inferior progression-free survival in the subjects.

Yazaki et al. engineered an scFv-albumin fusion as an imaging agent. They made genetic fusion of a murine anti-CEA monoclonal antibody T84.66 scFv to a truncated HSA that referred to as ‘immunobumin’ [65]. Two different radiolabeled immunobumins, [¹²⁵I]-T84.66 immunobumin and [¹¹¹I]-DOTA-T84.66 immunobumin, were used to study biodistribution in human colon carcinoma LS-174 T xenografts. [¹²⁵I]-T84.66 immunobumin showed significant increase in tumor uptake of 22.7% ID/g at 18 h, compared to the 4.9% ID/g for scFv alone. Tumor uptake was even more prominent for [¹¹¹I]-DOTA-T84.66 immunobumin, which was marked at 37.2% ID/g at 18 h. Remaining at the tumor site after 72 h was 27.3% ID/g with a tumor-to-blood ratio of almost 19:1. By showing a higher tumor uptake and blood half-life for immunobumins compared to scFv alone control, they demonstrated the potential of immunobumins as an alternative imaging tool.

One of the challenges in producing antibody-albumin fusions is the deterioration of albumin's binding capability to its cognate receptor FcRn. Anderson et al. recently investigated FcRn binding capacities of various albumin fusion proteins by fusing a short peptide or scFv to the N- or C-terminus of albumin [66]. These fusions at the N-terminus of albumin did not result in a significant reduction in binding (although there was a trend in reduction) whereas fusion to its C-terminus resulted in a 2-fold increase in K_D (2-fold reduction in affinity). While a small difference, it is possible that this difference will be amplified in vivo, where the albumin fusions compete with 40 mg/mL of intact albumins in the blood for FcRn binding. To circumvent the issue, current efforts in utilizing albumin for antibody delivery include incorporation of albumin binding domains within a therapeutic antibody. The three most successful platforms are Albumod™ (MedImmune, now licensed to Affibody), AlbudAb™ (GSK) and Albumin binding Nanobodies® (Ablynx). Albumod™ uses an albumin binding domain (ABD) identified in protein G of Streptococcus strain G148. ABD is a 46 amino acids polypeptide that forms three-helix bundle with affinity (K_D) of ~4 nM to HSA [67]. For example, ABY-035, an 18 kDa fusion of Affibody® and Albumod™, exhibits antibody-like half-life due to Albumod™ binding to HSA. The AlbudAb™ platform uses an anti-albumin single domain antibody (variable heavy chain or variable kappa chain) selected from a phage display [68]. GSK2374697, a fusion of Exendin-4 and anti-albumin variable kappa chain (V_k) showed an increased half-life of Exendin-4 from ~2.5 h to 6 days in humans [69]. GSK3128349, a fusion of ⁸⁹Zr and AlbudAb™, was also tested in a clinical trial using positron emission tomographic imaging (NCT02829307). Ablynx uses albumin binding Nanobodies® for half-life extension. Ozorulizumab (trivalent, bispecific antibody), M1095/ALX-0761 (trivalent, trispecific antibody), and vobarilizumab (bivalent, monospecific antibody) are examples of albumin binding nanobodies that have entered clinical trials (Table 1).

3.3. Transmembrane proteins for exosome retargeting

Exosomes are naturally occurring lipid-bilayer vesicles with hydrodynamic diameters of 30–150 nm that are secreted by most cell types. Inspired by their biological role in delivering nucleic acids, amino acids, and proteins for intercellular communications [70], therapeutic exosomes are being proposed with an expanded list of cargo including ribonucleic acids (such as siRNA and miRNA), chemotherapeutic compounds (such as paclitaxel and doxorubicin), cell growth inhibitors, viral vectors, and bioimaging agents [71]. To improve efficacy, exosomes have been engineered to have active targeting properties. To target exosomes requires an appreciation for the biological complexity of the exosome lipid bilayer composition [72]. As chemical conjugation of antibody fragments to purified exosomes may render the exosomes non-functional, endogenous proteins that are enriched in exosomes were explored as a fusion platform. Proteins were chosen that are highly expressed in the donor cells, highly enriched in the secreted exosomes, and capable of presenting targeting moieties at the corona of the exosome bilayer [73]. Incorporation of targeting moieties, such as RVG (targets the acetylcholine receptor) [74], folate receptor α (FRα, promotes crossing the blood brain barrier) [75] or GE11 (targets EGFR) [76] have all been investigated to target exosomes.

Kooijimans et al. produced nanobody-displaying exosomes from Neuro2A cells transfected with the vector that encodes anti-EGFR nanobody followed by 37 amino acids of human decay-accelerating factor (DAF). DAF (or CD55) is a 70 kDa protein that interacts with cell membrane bilayer via glycosylphosphatidylinositol (GPI) anchor and is selectively sorted into secreted exosomes during reticulocyte maturation [77]. The expressed nanobody-DAF fusion gets inserted into a lipid bilayer after the DAF peptide is replaced with a GPI anchor through post-translational modification. This anti-EGFR nanobody incorporated in exosomes was enriched with conventional exosome markers, showed identical size distribution to the unmodified exosomes, and retained a ‘cup-shaped’ morphology that is a hallmark of exosomes. Nanobody-decorated exosomes successfully targeted EGFR-expressing tumor cells in both static cell cultures as well as in flow conditions (shear rate: 82.5 s-¹); however, they had suboptimal cellular uptake, which was attributed to low expression and heterogenous enrichment of the anti-EGFR nanobody within purified exosomes.

Limoni et al. achieved siRNA delivery to HER2 positive breast cancer cells by engineering exosomes with an antibody-mimetic designed ankyrin repeat protein DARPin [78]. By fusing an anti-HER2 DARPin between the signal peptide and the mature gene segment of a membrane protein Lamp2b in a lentiviral vector, they produced DARPin-decorated exosomes that contain 75 kDa Lamp-DARPin chimeric proteins. These anti-HER2 DARPin exosomes, loaded with siRNA, successfully targeted HER2/neu positive cells, internalized, and suppressed gene expression of tumor protein D52 (TPD52) by 70%. These DARPin-decorated exosomes were used for tumor imaging in another study [79]. In a SKOV03 ovarian cancer xenograft model, they showed significantly higher accumulation of ^99mTc-DARPin-exosomes compared to a trastuzumab pre-treated control group, which saturates HER2 receptors.

Shi et al. developed an exosome platform, termed synthetic multivalent antibodies retargeted exosome (SMART-Exo), that displayed both anti-human epidermal growth factor 2 (HER2) and anti-CD3 scFv on the exosome surface [80]. Their αCD3 + αHER2 bispecific SMART-Exo utilized the human platelet-derived growth factor receptor (PDGF-R) transmembrane domain for self-assembly and display of scFvs on the exosome surface. SMART-Exo showed a dose dependent T cell activation in the presence of HER2 positive HCC 1954 cells, which confirmed its ability to recruit CD3 positive T cells to HER2 positive cells. Under co-incubation of non-activated peripheral blood mononuclear cells (PBMC) + SMART-Exo mixture and HER2 positive cells (SK-BR-3 and HCC 1954) at a 10:1 ratio, SMART-Exo exhibited cytotoxicity with an EC₅₀ of 0.85 ± 0.23 ng/mL and 50.2 ± 7.67 ng/mL (concentration of total exosome protein) in SK-BR-3 and HCC 1954 cells, respectively. Under the same condition, SMART-Exo had no effect on HER2 negative MDA-MB-468 cells. In addition, PBMCs depleted of CD3 positive cells had no effect towards cell viability of HER2 positive cells. Intravenous SMART-Exo showed significant tumor suppression in immunodeficient NSG (NOD scid gamma) mice that were subcutaneously implanted with HCC1954 cells. No toxicities were observed as the body weight, creatine level (kidney damage marker), and alanine aminotransferase (liver damage marker) was similar to that of control animals. Their data demonstrates the encouraging potential for bispecific SMART-Exo in the immunomodulation of breast cancer.

4. Non-human proteins for self-assembly of antibody variable regions

4.1. Leucine zippers and coiled coils

The leucine zipper is a coiled-coil structure of two α-helices. Its homodimerization is promoted through a hydrophobic interaction between leucine residues that are present at every 7th residue within an α-helix forming peptide, such as RMKQLEDKVEELLSKNYHLENEVARLKKLVGER (leucine residues are bold and underlined) [81]. This homodimerization can be utilized to generate oligomeric multivalent antibody fusions. For example, two α-helix leucine zipper forming peptides, each peptide fused with an antibody fragment either at its C- or N-terminus, assemble a bivalent antibody complex upon expression and homodimerization [82].

Since the successful expression and validation of binding activity of scFv-leucine zipper (scFv-ZIP) in E. coli by Pack et al. [83], leucine zippers and variants were further investigated as an antibody fusion platform. One of the strategies selected for further development was amino acid substitutions within the α-helix forming peptide that produced multimeric leucine zippers. It was identified that the hydrophobic residue (valine, leucine, or isoleucine) substitutions within the heptad repeat (a and d residues in abcdefg heptad repeat within α-helix) affect the oligomeric state of the leucine zipper, promoting dimeric, trimeric, or tetrameric assembly [84]. Incorporation of isoleucine at the d position in every heptad repeat relaxed the tight ‘knobs-into-holes’ configuration formed by interactions between a and d amino acids, and results in trimeric leucine zippers. Trimeric leucine zippers were further modified to become a tetrameric by incorporating leucine at the a position (which now have leucine and isoleucine at every a and d position). Based on these molecular characteristics, Klement et al. generated scFv-ZIP (dimeric), scFv-TriZIP (trimeric), and scFv-TetraZIP (tetrameric) (Fig. 4A) that target cell-surface podocalaxin like protein-1 [85]. The rationale for developing multivalent scFv-ZIPs would be to achieve better tissue penetration as well as improved retention compared to the parent monoclonal antibody IgM84. IgM84 shows cytotoxicity towards undifferentiated human embryonic stem cells (hESC) by clustering podocalaxin like protein-1, which is known to induce cell death similar to oncosis (cell death with swelling); however, it shows inferior tissue penetration due to its size, thus limiting its use in removing undifferentiated hESC from differentiated cells in vitro prior to transplant into the patient. To address this issue, scFv-ZIPs were tested for their cytotoxicity and tissue penetrability compared to IgM84. In terms of cytotoxicity, all scFv-ZIPs were less potent than mAb84, while scFv-TetraZIP was better tolerated than both scFv-ZIP and scFv-TriZIP due to its higher valency. On the contrary, scFv-ZIPs were better for tissue penetration than mAb84. More than 80% of the given dose was found in the hESC–derived embryoid bodies (EB) for all scFv-ZIPs, which was remarkably higher than that of the mAb84 (25%). The ability to penetrate to the core of the tissue was slightly different among scFv-ZIPs. A significant fraction of scFv-ZIP or scFv-TriZIP was found in the core of the EB, while scFv-TetraZIP was scarce at the center of the tissue, possibly due to its high MW (127 kDa) and valency (4). Based on cytotoxicity and tissue penetrability, it was concluded that the scFv-TriZIP may be a better candidate for further study than scFv-ZIP or scFv-TetraZIP.

Fig. 4.

Non-human peptide biomaterials used to self-assemble antibody fragments. A. Leucine zipper (ZIP)-mediated-assembly of antibody fragments. Engineered leucine zippers can form trimers (TriZIP) or tetramers (TetraZIP). scFv84: single-chain variable fragment of monoclonal antibody 84; mIgG3UH: mouse IgG3 upper hinge region that is used as a linker between scFv and α-helix forming peptide. B. Schematic of the Zipbody. Heavy chain (dark blue) and light chain (cyan) are genetically fused to an individual α-helix forming peptide, respectively. Dimerization of α-helix forming peptides to leucine zipper enhances correct folding of the Fab. C. Protein structure of SpyCatcher and SpyTag. SpyCatcher and SpyTag is depicted in grey and red, respectively. Lys31(yellow) of the SpyCatcher forms an isopeptide bond with Asp117 (orange) of the SpyTag. D. Structure of silks spun by Bombyx mori. Fibrils make up the fibroin protein, which is covered by sericin. Heavy (~325–350 kDa) and light (~25 kDa) chain are linked by a single disulfide bond to form the fibroin fibril. The heavy chain is made up of 12 hydrophobic chains (red bar) with hydrophilic linkers (blue line). Silk 1 is formed by an α helix while silk II consists of antiparallel β-sheet nanocrystal conformation. Figures reproduced from refs [85,95,113,114] with permission.

While the leucine zipper from GCN4 remains the most extensively studied platform for α-helical peptide assembly, other human homologs have been explored for antibody assembly. Ojima-Kato et al. tested two kinds of leucine zippers: i) a c-Fos/c-Jun leucine zipper pair identified in humans; and ii) an artificially designed leucine zipper pair LZA/LZB. Both approaches were used to pair the Fab that targets O157, one of the O-antigens found on the surface of E. coli. [86]. It should be noted that the purpose of the study presented by Ojima-Kato et al. is to use leucine zippers to correctly fold the heavy and light chain of Fab (Fig. 4B), whereas Klement et al. utilized leucine zippers to produce multivalent antibody formulations (Fig. 4A). Fab domains contain multiple disulfide bonds that require correct folding. Fab fragments expressed under oxidizing conditions, i.e., cytoplasm of E. coli, are usually recovered from inclusion bodies as these conditions result in incorrectly or partially folded, non-functional material. To determine if leucine zippers enhance the correct folding and solubility of Fab, fusion of the heavy chain and light chain of Fab to Fos and Jun (or LZA and LZB), respectively, was paired through leucine zipper heterodimerization. This generated two types of Fab fusions, Fab-Fos/Jun and Fab-LZA/LZB (named Zipbody), which bound specifically to an E. coli strain that expresses O157 and not to other strains. Fab domains expressed without a leucine zipper failed to bind their targets. This Zipbody platform was further optimized for the antibody screening process to rapidly develop monoclonal antibodies via cell-free protein synthesis [87].

Besides leucine zippers, Zhang et al. incorporated a coiled-coil system named CCE/CCK to their drug-free macromolecular modality [88]. CCE and CCK are oppositely charged penta-heptad molecules that bind to each other to form an antiparallel heterodimer. Using the heterodimerization of CCE and CCK, they showed that cell surface CD20 was bound by anti-CD20 Fab-CCE. This was further engaged by multiple CCK (HSA-(CCK)₇), which was significantly better in inducing apoptosis than the rituximab in CD20 positive Raji cells, a model of non-Hodgkin lymphoma. In-situ assembly between CCE and CCK on the cell surface clustered CD20 bound Fab-CCEs and induced apoptosis. Since it is a two-component system, they compared a therapeutic efficacy between consecutive (administer Fab-CCE followed by HSA-(CCK)₇) and premixed (pre-assemble Fab-CCE with HSA-(CCK)₇ before administration) administration. For both consecutive and premixed at three different ratios of CCE and CCK (1:1, 1:5, and 1:25), the caspase 3 activity was 17%, 26% and 31%, respectively. Although there was no significant difference between the consecutive and premixed administration, the possible advantage of the consecutive over premixed would be the high tumor-to-normal tissue ratio as administration of the HSA-(CCK)₇ after the nonspecifically bound Fab-CCE gets cleared, may minimize off target side effects. [89].

4.2. SpyCatcher/SpyTag, an adaptor system

Streptococcus pyogenes (also called group A Streptococci, GAS) is an exclusive human pathogen that causes significant diseases from a mild, self-healing mucosal membrane, and skin infections such as pharyngitis, impetigo, and pyoderma (a pus forming skin infection) to potentially life-threatening diseases such as necrotizing fasciitis and toxic shock syndrome [90]. Streptococcus pyogenes (as well as many other gram-positive bacteria) has many virulence factors to hijack the host cells. One of the factors is called MSCRAMM (microbial surface components recognizing adhesive matrix molecules), which contains adhesins that can attach to the host cells. MSCRAMM adhesins have two different domains: CnaA (collagen adhesin domain) domain and CnaB domain in which the latter forms a spontaneous intramolecular isopeptide bond within the same domain. The formation of this intramolecular bond significantly stabilizes CnaB domain (adhesin) as a whole, thereby providing a stronger binding to the host cell [91]. Mark Howarth and coworkers at the Oxford University developed the SpyCatcher/SpyTag system by engineering the CnaB2 domain of the FbaB (fibronectin binding protein F2), a MSCRAMM found in Streptococcus pyogenes (Spy) [92,93]. By knowing that the intra-isopeptide bond is spontaneously formed between the lysine (K31) and the aspartic acid (D117) within the CnaB2 domain, the 116 amino acid residues (15 kDa) within the CnaB2 that contains lysine became the SpyCatcher and the 13 amino acid residues that flanks aspartic acid were named SpyTag (Fig. 4C) [94]. Functional moieties are recombinantly fused to either N- or C-terminus for both SpyCatcher and SpyTag. Especially, due to its small size, SpyTag can be fused even in the middle of the protein. The SpyCatcher/SpyTag system is compatible with both bacterial and mammalian expression systems [95] and various purification tags, such as histidine, GST (Glutathione S-transferase), or FLAG are used to purify SpyCatcher and SpyTag through column-based purification [96].

Since its invention, SpyCatcher/SpyTag system has been applied to various biological research, including vaccine production and targeted drug delivery. To address several problems observed in conventional vaccine development, such as sub-optimal expression for large antigenic fusion proteins and excessive cost and time for the de novo generation [97], Liu et al. used a SpyCatcher/SpyTag system to target and activate dendritic cells to evoke immune responses [98,99]. Association of anti-DEC205 scFv fused SpyTag with the SpyCatcher that carries both CD8 T cell epitope (ovalalbumin_257-264 or OVA₈) and tick-borne encephalitis virus envelope protein domain 3 (TBEV ED3), termed αDEC205-SpyTag/SpyCatcher OVA₈-ED3, elicited a 15-fold higher antibody titer compared to the non-targeted construct, SpyCatcher/SpyTag OVA₈-ED3. The percentage of cytotoxic T cells (IFNγ producing CD8+ cells) in splenocytes in C57BL/6 mice vaccinated with αDEC205-SpyCatcher/SpyTag OVA₈-ED3 was 3.5-fold higher than that in mice vaccinated with non-targeted construct. This result shows the feasibility of SpyCatcher/SpyTag system as an antibody fusion platform and its utility in vaccine production. Compared to the conventional de novo production of the vaccine, generation of the library by dividing the antibody fragment and antigens to SpyCatcher and SpyTag, could speed the vaccine development process, especially in the current COVID-19 pandemic and other outbreaks where discovery, screening and mass production is pivotal.

Park et al. used SpyCatcher/SpyTag system to deliver an immunotoxin derived from Corynebacterium diptheria, which induces apoptosis by inhibiting protein synthesis [100]. By using two different antibody mimetic-SpyTag fusions, this group showed how simple it is to pair one immunotoxin-SpyCatcher fusion with different antibody mimetic-SpyTag fusions. To target HER2 positive tissues, the SpyCatcher fused with an A fragment and T domain (dtA-T-SC) was covalently paired with the HER2 targeting affibody fused SpyTag (ST-Afb). The A fragment in the immunotoxin irreversibly inactivates eukaryotic elongation factor 2 upon endocytosis into a host cell, whereas the T domain enables endosomal escape of A fragment by forming a pore on the endosome. The in vivo anti-tumor activity of the targeted immunotoxin, named dtA-T-SC/ST-HER2Afb, in HER2 positive allograft animal model was significantly higher compared to animals treated with immunotoxin (dtA-T-SC) alone. By simply swapping ST-HER2Afb to ST-EGFRAfb, the anti-tumor activity observed in HER2 positive tumors was replicated in EGFR positive tumors. Intravenously administered dtA-T-SC/ST-EGFRAfb significantly inhibited tumor growth in an EGFR positive xenograft animal model compared to dtA-T-SC alone. By showing an anti-tumor activity in two different animal models, this study reconfirms ease of target variation at the post-translational stage.

Geyer and co-workers used SpyCatcher/SpyTag system as a platform to construct multivalent antibody construct [101]. Three independent anti-HER3 scFv fused SpyCatchers (anti-HER3 scFv-SpyCatcher) were combined into a single trivalent construct by pairing with the tri-SpyTag backbone. Biolayer interferometry kinetic analysis showed that trivalency (anti-HER3 Tri-scFv-SpyCatcher) lowered K_d by 12-fold compared to monovalent control (anti-HER3 scFv-SpyCatcher) by significantly lowering the dissociation rate constant, k_off. The flow cytometry analysis confirmed the improvement in K_d, by showing an increase in mean fluorescence intensity (MFI) in HER3 positive FaDu cells with increase in valency: 3834, 4636, and 6919 for monovalent, bivalent, and trivalent construct, respectively. These results suggest the SpyCatcher/SpyTag system can assemble antibody variable regions and serve as a platform for multivalency.

4.3. Fibroin – a silk protein

About 150,000 species including spiders and insects can produce silks [102]. Among all different kinds of silks, mulberry silk produced by silkworm Bombyx mori is mostly use in textile industries [102]. A silk fiber produced by B. mori is composed of fibroin and sericin (Fig. 4D). Sericin is a glycoprotein that surrounds the core protein named fibroin. Fibroin is composed of three parts: heavy chain, light chain, and glycoprotein P25 [103]. The 12 hydrophobic anti-parallel β-sheets stabilized by hydrogen bonds and van der Waals forces make the core of the heavy chain. Within this core, an extended segments of amino acids (GAGA)_nG(Y/S) were found to be responsible for the physico-chemical properties and crystallinity of silk films and silk fibers. The hexapeptide repeats (GAGAGX)_n derived from this motif has been recombinantly expressed and named as silk-like polypeptides (SLPs) [104,105]. Due to their assembly of fibrous structures, both fibroin and SLPs are widely explored in tissue engineering [106].

Even though the heavy chain of fibroin or its recombinant derivative SLPs are widely used, it may not be a suitable platform for antibody assembly due to their tendency to form crystalline domains. Instead, a more hydrophilic segment of fibroin was explored as a fusion site [107]. The hydrophilic nature of the fibroin light chain produces elastic semi-crystalline proteins instead of insoluble fibers [108]. The light chain of the fibroin is different in its amino acid composition and solubility. Certain amino acids (aspartic acid, alanine, glycine, and serine) that are relatively hydrophilic comprise approximately 50% of the light chain; however, there is no consensus motif. One group in Japan pioneered fibroin light chain in antibody assembly using transgenic silkworms [109]. The scFv-fibroin light chain fusions (scFv-FibL) that were purified from the cocoons of the transgenic silkworms retained their binding activity towards Wiskott-Aldrich syndrome protein (WASP), even after they were freeze-dried to powder. Compared to its parent molecule anti-WASP mAb, scFv-FibL showed equivalent immunoprecipitation capability towards recombinant WASP and native WASP (extracted from cell lysates). After this initial success, they further targeted CEA and validated its function using ELISA [110,111]. An interesting feature across their reports is that scFv-FibL maintained binding activity regardless of its formulation: powder, thin film, and aqueous solution. Although they showed a novel approach to self-assemble antibody fragments in silk proteins, low expression was reported as a limitation. The expression level of scFv-FibL was about 10– 25% of total light chain proteins or 1– 2% of total silk proteins extracted from cocoons [112].

4.4. Antibody-virus fusions

Despite their potential immunogenicity, the superb specificity and infectivity of viruses have prompted viral surface engineering to redirect them to tumor tissues. Surface engineering can be performed in several ways [115]. First, pseudotyping is a method to transplant viral glycoproteins from other viruses, either from the same family or between families, to manipulate the original tropism (affinity to bind a specific target) and expand the host range of the virion [116]. In the wild, human immunodeficiency virus type 1 (HIV-1) incorporates glycoproteins from other viruses through phenotypic mixing (pseudotyping) to expand its host range. In the lab, this can be applied to both enveloped viruses (retrovirus or lentivirus) and non-enveloped viruses (adenovirus or adeno-associated virus). Compared to adaptor systems, genetic engineering, or polymer coating, pseudotyping is less technologically challenging; however, the limited availability of novel pseudotypes have restricted their applications [117,118]. Second, adaptor systems can be employed to modify intact viral particles. These include: i) utilizing a peptide that binds to cognate virus receptor (e.g CAR: coxsackievirus and adenovirus receptor) to graft molecules on the surface of the virus [119]; ii) bispecific antibodies where one fragment binds the virus and the other end targets a receptor of interest [120]; iii) chemical linkages that covalently link a targeting peptide to the viral surface [121]; iv) avidin-biotin interactions whereby a biotinylated virus interacts with an avidin-bound targeting moiety [122]; and v) an antibody that functions as a targeting moiety via an immunoglobulin (Ig) binding domain expressed on the viral surface [123]. Adaptor systems have little effect towards the original viral structure, allowing exploration of different targeting molecules using similar procedures. A third approach is genetic modification. Several examples include: i) serotype exchange that produces chimeric virus decorated with a donor serotype [124,125]; ii) fusion of a targeting peptide [126]; iii) scFv expressed on the outer surface of the virus particle [127]; iv) two or more distinct viral attachment proteins expressed on one virus particle (mosaicism) [128]; and v) gene mutations that ablate the original viral tropism to reduce off-target viral transduction [129]. Genetic fusion enables homogenous production; however, technical barriers can result from changing the original structure and function of the virus, such as replication and encapsulation. A fourth strategy is the incorporation of a polymer coating to evade opsonization and neutralization. Polyethylene glycol (PEG) [130] and poly-N-(2-hydroxypropyl) methacrylamide (poly-HPMA) [131] are widely used polymers to coat viral particles, which helps them escape clearance by the mononuclear phagocyte system (MPS).

Among the strategies discussed, genetic modifications of the genes that encode viral capsid proteins or envelop proteins have been extensively investigated for antibody fusion after a landmark report by Russell et al. in 1993 [127]. Technological advancements have enabled assembly of antibody fragments on the viruses with high titer and exceptional specificity. Chowdhury and coworkers fused anti-CEA scFv to the N-terminus of the murine leukemia virus surface subunit A, allowing scFvs to assemble on the surface of retrovirus envelop [132]. After scFv-retrovirus binds to CEA, viral particles transduce into tumor cells via interaction between its surface unit gp70 and Pit-2 on the target cell. Attributed to specificity and transduction efficiency, CEA positive tumor xenografts (HT29 and Mawi) were highly transduced, while CEA negative tumors were not (A375 and HT1080). They claimed that it was 10⁴ times less likely for scFv-retrovirus to infect host organs, as proviral DNA was not detected in spleen, liver, and kidney, in CEA positive tumor xenografts. Achievement of high titer and maintenance of in vivo transduction efficiency demonstrated the feasibility of scFv fused virus as targeted therapeutics.

Nakamura and coworkers tested two oncolytic measles viruses (MV) retargeted by scFv fusion at the C-terminus of the H (hemagglutinin) gene. To precisely evaluate the scFv-mediated targeting ability, the H gene was mutated to blind MV from its natural receptors CD46 and SLAM [133]. scFv-MVs that target either CD38 or EGFR had specificity against an EGFR positive cell line (SKOV3ip.l) or two CD38 positive cell lines (Ramos and Raji), respectively. No residual infectivity was reported on target-negative control cells. When these were injected intratumorally, both anti-EGFR-MV and anti-CD38-MV showed a remarkable anti-tumor activity in their respective subcutaneous xenograft models. Upon intravenous administration, anti-CD38-MV treated metastatic Raji xenografts showed the best survival rate compared to animals treated with unmodified MVs or anti-EGFR-MVs.

A scFv-MV-H fusion was further characterized by Ayala-Breton et al. [134]. They pseudotyped VSV (an oncolytic vesicular stomatitis virus) with both scFv-MV-H and scFv-MV-F (hemagglutinin (H) and fusion (F) envelope glycoproteins) to target either EGFR, alpha-folate receptor (αFR), or prostate-specific membrane antigen (PSMA). In these fusions, they ablated the original tropism of VSV by deleting the G gene from its genome (VSVΔG). Therefore, reconstituted scFv-MV-H/F pseudotyped VSVΔGs were neutralized only by anti-measles serum, but not by anti-VSV serum. Cells expressing either EGFR, αFR, or PSMA were efficiently targeted by pseudotyped VSVΔGs, while cells expressing CD46 or SLAM were not. This confirmed infectivity was dependent on the pseudotyped scFv-MV-H/F proteins onto VSVΔG. The in vitro specificity was also maintained in vivo. The gene expression of green fluorescent protein reporters was restricted to the receptor-positive tumors in SCID or athymic xenografts engrafted with various cell lines: KAS 6/1 (EGFR−, αFR−, PSMA−), SKOV3ip.1 (EGFR+, αFR+), PC3 (PSMA−), and PC3-PSMA (PSMA+).

Antibody-decorated ‘onco-lytic’ viruses, that lyse the oncogenic tissues, have been explored for the treatment of cancer. Oncolytic viruses have proven challenging for multiple reasons. First, viral producer cells tend to die before substantial viral particles are expressed. Moreover, the spread of oncolytic virus inside the tumor is insufficient, which seems to lead to suboptimal therapy. However, due to their exceptional infectivity, they may still possess therapeutic potential. To address these challenges, Fernández-Ulibarri and coworkers genetically linked a peptide that is a fusion of anti-EGFR scFv and onconase (ONC_EGFR) to the capsid of oncolytic adenovirus [135]. Onconase (sometimes referred to as immunoRNase) is an RNase from the Northern leopard frog Rana pipiens that resists the human cytosolic RNase inhibitor. Its expression in the cytosol degrades endogenous RNAs and blocks proliferation of human cancer cells [136,137]. On top of ONC_EGFR fusion, additional modifications were explored to eliminate EGFR positive cells that were not initially infected: i) viral expression was optimized to maintain full replication by shifting Onconase expression until the late phase of replication; ii) a deletion mutation in pRb interacting domain of E1A (E1AΔ24) promoted cancer-selective viral replication; iii) supplemented with adenovirus serotype 3 knob, thus making chimeric serotype 5/3, for better infection. These ONC_EGFR-adenovirus showed strong binding and specificity to EGFR and induced cytotoxicity in all EGFR positive cells A431, Cal27, Panc-1, and primary HNO210 cells. Impressively, they reported more than an 100-fold improvement in IC₅₀ on A431 cells compared to control virus-treated A431 cells. ONC_EGFR-adenovirus had no effect in the receptor negative Mel624 melanoma cell line. The ONC_EGFR-adenovirus was especially effective in overcoming the antibody resistance. A non-small cell lung cancer (NSCLC) cell line H460, in which the cytotoxic effect of an anti-EGFR monoclonal antibody cetuximab was minimal, showed a significant decrease in its viability upon ONC_EGFR-adenovirus treatment. This superior cytotoxicity was successfully replicated in subcutaneous A431 xenografts. Single intratumoral injection of ONC_EGFR-adenovirus virotherapy showed a significant tumor regression as well as exceptionally high onconase expression in tumor cells compared to a non-targeted virotherapy control.

5. Concluding discussion

Engineering conventional human antibodies and identification of single domain antibodies (V_HH or V_NAR) from non-human sources have accelerated innovation in biologics. Biomaterials that mediate assembly of these antibodies have generated a wealth of strategies to modify and strengthen their applications. Having described various antibody-biomaterial fusions, this concluding section discusses three fundamental aspects important to moving these pre-clinical modalities into the clinic: scale up, affinity-efficacy relationship, and biophysical characteristics.

First, with regard to scale up, having more than two components necessitates separate optimization for each in terms of expression, purification, and quality control. If one component's manufacture is sub-optimal compared to another's, the mass production of the whole modality may have a high cost or complexity, which may be unfavorable for investors, despite the novelty of intellectual property and promise of pre-clinical efficacy. In addition, the USFDA requires added documents, such as Module 2 and parts of Module 3 of the Common Technical Document (which are the parts of the New Drug Application), which requires additional time and capital [138,139]. As shown in Tables 1 and 2, all antibody fragment-based therapeutics currently in the clinic are single component constructs that are devoid of these issues. Among the discussed domain-based oligomerization systems, CLPs, leucine zipper, coiled coil, or SpyCatcher/SpyTag system (or other adaptor systems) could experience this issue if each sub-component has different biophysical and therapeutic properties (i.e., each sub-component is fused to a different scFv). While sub-optimal expression and formulation stability may hinder their scale up, their innovations in biological research (i.e., imaging application, assay development, vaccine screening) cannot be neglected. Applications such as leucine zippers for expedited antibody candidate identification by Ojima-Kato et al. [86] or the SpyCatcher/SpyTag system for vaccine screening by Liu et al. [97] demonstrate how these modalities could directly contribute to advancements of clinically relevant targets even without a clear pathway to mass production.

Second, higher affinity does not always correlate to an improved therapeutic effect. Advancements in antibody engineering towards generating multivalent antibodies introduced avidity into consideration. However, the ‘binding-site barrier’ effect, proposed by Weinstein et al., explains why high antibody affinity can lead to poor tumor penetrance and mitigate therapeutic efficacy [140]. In the context of solid tumors, once antibodies escape from vasculature, high tumor interstitial pressure is the next barrier they encounter. To reach target antigens buried deep inside the tumor, antibodies may go through a series of association and dissociation events against targets to gradually move forward. Under this condition, high affinity with a low dissociation constant decreases the availability of free antibodies to diffuse farther into the tumor core, which leads to significant accumulation at the tumor margins. Heterogeneous distribution of high affinity antibodies has been confirmed using bioimaging techniques, which often reveal a dense localization on outer layers of the tumor (or in close proximity to functional blood vessels). Regarding this matter, Cuesta et al. postulated the ‘tumor target zone’, which suggests optimal biophysical conditions for the multivalent antibody constructs in tumor targeting [141]. The monovalent antibodies, such as scFv, have higher tumor penetration, but they experience fast off-rates and rapid systemic clearance due to their smaller size and weight. In contrast, the multivalent antibodies overcome the abovementioned setbacks leading to a tighter engagement to the targets (longer tumor retention) and longer half-life in the blood through their avidity. However, they suffer from inferior tumor penetrability. The molecular weights between 70 and 120 kDa with bi- or trivalency was suggested to achieve optimal tumor targeting, tumor uptake, and low systemic clearance, which may include diabodies (bivalent), minibodies (bivalent), and trimerbodies (trivalent) [141]. If the delivery of anti-tumor agents deep into the tumors is crucial, use of comparably lower affinity antibody should be considered. On the other hand, if binding itself is the goal, such as induction of apoptosis upon binding to tumor cells, then the higher affinity antibody may be favorable [142]. Multiple factors, such as tumor microenvironments, density of target antigen, vascular permeability, and antibody's affinity, should be precisely assessed in conjunction with one another to achieve optimal therapeutic effect. CLP (or collagen trimerization domain), HSA, or leucine zipper fusions may conform with the characteristics described in the tumor target zone. These fusion platforms are more confined than others in terms of valency and structure. ELPs, transmembrane proteins for exosomes, and antibody-virus fusions may be superior in tumor targeting and residence due to their high valency but inferior in tumor penetration compared to CLP, HSA, or leucine zipper fusions. Antibody-virus fusions might be comparably better in tumor penetration among the three, owing to their ability to infect the host tissue; however, a binding-site barrier may remain a factor that retains them at the outer layers of the tumor. High valency modalities may be more suitable when tissue penetration is not necessary, such as targeting vascular endothelial cells [143] or circulating tumor cells [144]. Since those targets do not require tissue penetration, high valency fusions may provide better therapeutic outcomes based on enhanced engagement with receptors at the target tissue. Another point to mention in regard to the valency would be the determination of the effective degree of valency in high valency modalities (ELPs, transmembrane fusions of exosomes, antibody fused viruses). Most research articles report their valency in a form of a range or simply describe as a ‘multivalent construct’ without further detail. Given that such information would allow better understanding of thermodynamic properties, capacity of oligomerization of a given biomaterial, and valency-therapeutic efficacy relationship, it is important to always quantify the effective antibody valency (or asa form of ‘degree of oligomerization’) as a standard part of their biophysical characterization. One way to determine this is to measure absolute molar mass of a fusion construct using a multi angle light scattering (MALS). Since MALS determines the exact molar mass of a given molecule, the degree of valency (or the degree of oligomerization) can be estimated. Our research group has been performing MALS analysis on antibody-ELP fusions to provide readership with this particular information [38-41].

Thirdly, biophysical characteristics should be precisely determined alongside therapeutic efficacy. Despite the promising theoretical concept and preliminary in vitro studies, many compounds are not successful in vivo. A major limitation of many formats is sub-optimal pharmacokinetics. The key factors governing pharmacokinetics of antibody-conjugated biomaterials are size, charge, morphology, and serum stability. Under pathologic conditions, therapeutic molecules can readily cross or escape the abnormal vascular endothelium and reach the tumor microenvironment [145]. However, unlike the tumor periphery, blood flow is scarce in the tumor core, where many prominent target antigens reside on populations of dormant, resistant tumor cells. Therefore, whether they are molecularly targeted or non-targeted, it takes substantial time for therapeutic molecules to accumulate deep inside the tumor. Although smaller proteins penetrate better to the tumor core in accordance with their higher diffusion constants, they suffer from rapid clearance from the blood. Therefore, the size of the antibody-complex has to be precisely determined to guarantee both the elongated residence in blood and the tumor penetration [146]. Molecules that are less than 10 nm in hydrodynamic radius are eliminated by glomerular filtration (renal clearance), whereas those ranging from 150 to 300 nm are cleared by the liver and spleen (reticuloendothelial system) [147]. Therefore, antibody complexes with size between 70 and 200 nm may have a higher chance to reach the target site with prolonged circulation in the bloodstream with minimal glomerular filtration or capture by the reticuloendothelial system [148,149]. The net surface charge of the molecule also determines their longevity. In vivo toxicity study revealed that positively charged nanoparticles increased liver enzyme release, body weight loss, and interferon type I response. They also induced 15– 25-fold higher mRNAs of interferon responsive genes, and 10– 75-fold higher expression of pro-inflammatory cytokines compared to negatively-charged or neutral nanoparticles [150]. Due to this reason, positively charged nanoparticles are rapidly removed by macrophages through opsonization. Negatively charged or charge neutral nanoparticles are better than positively charged nanoparticles in terms of systemic half-life. Several reports suggest that net negative charge may be slightly better over charge neutral in preventing non-specific cellular uptake [151,152]. The morphological shape of antibody complexes may also govern their pharmacokinetics and efficacy. Single-walled carbon nanotubes that are 200– 300 nm in length and 350– 500 kDa in molecular weight experienced rapid renal clearance with half-life of only 6 min. Although they were 10– 20 times higher in mass than the glomerular filtration cut-off (~60 kDa), their narrow and elongated tubular structure negatively affected their residence time in systemic circulation [153]. Despite their importance, these parameters are often overlooked. The majority of research articles exploring antibody fragments, including some introduced in this manuscript, focus on pre-clinical efficacy without providing sufficient information on their biophysical characteristics. Having adequate biophysical characteristics (size, shape, colloidal stability, charge, etc.) alongside functional measurements of pharmacokinetics and preclinical efficacy will be necessary to advance antibody-based biomaterials into the clinic.

Given that antibody-biopolymer fusion modalities are scarce in clinics, the platforms discussed above likely experienced obstacles to move beyond the pre-clinical stage. To overcome these hurdles, research into their biophysical and pharmacological properties must continue until these modalities can complement the limitations of conventional antibody therapy and ultimately become better therapeutic alternatives for patients.