The making of multispecific immunoglobulins – a clinical perspective

ABSTRACT

Over the past two decades, bi- and multispecific antibodies have emerged as a rapidly advancing class of therapeutic biologics, transforming oncology and immunotherapy. By simultaneously binding two or more distinct antigens or epitopes, these molecules achieve mechanisms of action beyond those of conventional monoclonal antibodies, including immune cell redirection, dual pathway modulation, and enhanced tissue selectivity. Bispecific and multispecific antibodies exhibit considerable structural diversity, encompassing a wide range of molecular architectures covering a steady growing ‘zoo’ of formats. The therapeutic success and diversity of molecules and formats is reflected in the 2021 revision of the international nonproprietary name system, which introduced the suffix – mig to denote multispecific immunoglobulins. In this review, we provide an overview of multispecific antibodies in clinical development, focusing on format, molecular design, and clinical status. In total, data for 501 multispecific antibodies were compiled and analyzed, identifying 112 different formats. Overall, this analysis highlights the rapid growth, enormous format diversity, and translational potential of multispecific antibodies. It underscores their emerging role as versatile therapeutics not only in oncology, but also in non-cancer indications, reflecting a field that continues to evolve rapidly in response to both scientific innovation and clinical needs.

Introduction

Over the past several decades, bi- and multispecific immunoglobulins (migs) have emerged as one of the most transformative classes of therapeutic biologics in clinical oncology and immunotherapy. Defined by their ability to simultaneously or consecutively bind two or more distinct antigens or epitopes, bi- and multispecific antibodies enable mechanisms of action that are unattainable with conventional monoclonal antibodies.¹ Through this multiple targeting capability, they can bridge immune effector and target cells, co-modulate signaling pathways, or improve tissue selectivity, offering new solutions to long-standing therapeutic challenges.

The concept of dual-specific antibodies was first explored in the late 1960 s, with early experimental constructs appearing in the 1980 s through chemical crosslinking and hybrid-hybridoma (“quadroma”) technologies.² While these initial prototypes demonstrated proof-of-principle for immune redirection, clinical translation was challenging due to manufacturing issues, immunogenicity, toxicities, and limited therapeutic efficacy.^3–6 The advent of recombinant antibody engineering in the 1990 s catalyzed a resurgence of interest in the field, enabling rational design of bispecific architectures.^7,8 These innovations improved chain pairing fidelity and manufacturability, paving the way for clinical development. A pivotal milestone occurred with the approval of blinatumomab (Blincyto) in 2014, the first recombinant bispecific T-cell engager (BiTE) targeting CD19 and CD3. This achievement demonstrated that bispecific antibodies could be produced at clinical scale and deliver meaningful therapeutic benefit. Since then, multiple next-generation bispecific antibodies have reached the market, with 17 of 20 molecules approved within the past 4 years, and the clinical development pipeline is rapidly expanding.⁹

In contrast to most of the approved antibodies, with 91% being conventional IgGs, a diverse set of formats is used to make bi- and multispecific antibodies. The complexity of the design space and format diversity of bi- and multispecific antibodies is already reflected by the approved bispecific antibodies, but even more by the molecules in clinical development. In 2017, we published a review of the zoo of bispecific antibody formats populated by many different species.¹⁰ Since then, this zoo has expanded to include new species and many of the formats were applied to generate molecules that entered clinical development.

Until 2021, the international nonproprietary names (INN) of antibodies, including bispecifics, ended with the suffix -mab. In 2021 the system was revised and it was decided to substitute the INN stem -mab by four new stems (-tug, -bart, -mig, and -ment) to reflect the diversity of therapeutic antibody molecules and to ease their distinguishability.¹¹ Since then, bispecific antibodies have their own INN stem, -mig, which stands for “multispecific immunoglobulin” and covers all antibodies with more than one specificity. Although all approved multispecific antibodies and the majority in clinical development are bispecific antibodies, we will use the terms multispecific immunoglobulin and mig throughout the review.

This review provides an overview of multispecific immunoglobulins in clinical development, with an emphasis on format and molecular properties. Our analysis covers migs listed in the U.S. ClinicalTrials.gov database identified using publicly available information from sources such as the global substance registration system (GSRS), Antibody Society, TheraSabDab, and NCI EVS Explore databases, which were searched using terms such as “bispecific”, “trispecific”, and “multispecific”. Data were compiled up to December 2025. Antibody mixtures and antibody fusion proteins, for example with peptides and ligands such as SIRPα, TGF-β or receptor fragments (traps) such as VEGFR, DNA (aptamers), and multimers of peptides and scaffold proteins, were excluded. However, antibody fusions with antibody-mimetic scaffold proteins with at least one antibody moiety were included.

We identified more than 150 companies and sponsors with at least 1 mig in clinical development. Most of the companies have between 1–3 migs in development, 12 companies have 5–7 migs in development, and 4 companies (Innovent, Janssen, Regeneron, Roche) have 10 or more migs in clinical trials.

In summary, we compiled 501 migs, 358 migs currently listed with at least one study having the status not yet recruiting, recruiting, active or which was completed in 2023 or later (category 1). Furthermore, 143 migs that are listed as completed before 2023, terminated, withdrawn, suspended or having an unknown status (category 2) were identified and considered separately. For 322 migs of category 1 and 128 migs of category 2 we were able to identify the format. For 118 of the 332 migs of category 1 there is a sequence entry in the GSRS database (37 for migs of category 2).

It is well established that the structural format of migs can influence their functional properties, and that even minor changes in geometry or linker length/composition may alter biological activity. It is therefore noteworthy that for several clinically evaluated migs we were unable to unambiguously determine the exact molecular format; for even more candidates, format information was entirely unavailable, and their architectures had to be inferred from schematic representations in the public domain rather than from peer-reviewed sources. Although sequences are sometimes disclosed in patents or patent applications, the identifiers used in these documents seldom correspond to the names of clinical candidates, complicating the definitive assignment of molecular compositions in ongoing trials. While recognizing that the engineering of therapeutic antibody derivatives is a highly competitive field, it would nevertheless be advisable to ensure that detailed format information and precise molecular compositions of migs become readily accessible at the latest when these molecules are first administered to humans.

Our review summarizes the diversity of mig formats, their mode-of-actions, targets and target combinations, as well as engineering approaches used to enable Fc heterodimerization, modulate Fc-mediated effector functions, and extend serum half-life. A complete list of all migs analyzed in this study is provided in the supplementary Tables S1 and S2. Obviously, the field of multispecific antibodies is rapidly moving forward, with new molecules constantly entering clinical trials, and others being withdrawn or terminated. We tried to be as comprehensive as possible, but we are aware that we might have missed some molecules, especially those not listed in the ClinicalTrials.gov database. Our intention is not to provide a full list of migs in clinical development, but to provide an overview of multispecific antibody formats, especially those commonly used to generate migs.

Overview of migs in clinical development and mode-of-actions

Of the compiled 358 migs of category 1, 182 (51%) were listed as in Phase 1, 117 (32%) in Phase 1/2 or 2, 39 (11%) in Phase 2/3 or 3, and 20 (6%) are approved (Figure 1A). In total we identified more than 1500 clinical studies with migs. The first approved mig was catumaxomab (Removab), which obtained approval in Europe in 2009 for treatment of malignant ascites, but was voluntarily withdrawn in 2017 for commercial reasons. Of note, catumaxomab, was relaunched as Korjuny in 2025. In 2014 and 2017 two further migs were approved, blinatumomab for the treatment of B-cell precursor acute lymphoblastic leukemia (B-ALL) and emicizumab for treatment of hemophilia A. Since 2021, we see an increase in approval of migs, with one approved in 2021, four in 2022, six in 2023, and six in 2024.

Figure 1.

Properties of multispecific immunoglobulins in clinical use or development.

The majority (86%) of migs are for oncology indications (Figure 1B). Here, 76% of migs are used to treat solid tumors and 24% to treat hematologic malignancies. In non-cancer therapeutic areas, inflammatory and autoimmune diseases comprise the majority (55%), followed by hematologic diseases (11%), metabolic disorders (11%), infections (9%), vascular diseases, and neurologic diseases (Figure 1B). Thus, migs have found broad applications with new indications being explored.

This is also reflected by the modes-of-action that migs utilize. A large proportion of migs of category 1 are in the field of engaging and modulating immune responses, mainly for cancer treatment. Thus, 44% of the migs are designed for effector cell engagement and 31% provide immune checkpoint inhibition (CPI) and/or costimulation of immune cells. Further modes-of-action include blockade of growth factors or their receptors and down-stream signaling, inhibiting cytokines or their receptors, payload delivery, delivery through the blood-brain barrier (BBB), extending the half-life of the molecule, inhibiting viral infections, and mimicking the activity of natural ligands or cofactors (Figure 1C). Please note that especially tri- and tetraspecific migs can combine different modes-of action within one molecule.

Overview of migs in clinical development in cancer indications

For cancer therapy, effector cell-engaging migs are dominated by T-cell engaging molecules, which account for 89% of the effector cell engaging migs, 57% to treat solid tumors, and 43% to treat hematologic malignancies. These T-cell engagers (TCEs) are designed to bind simultaneously to a tumor-associated antigen (TAA) expressed on tumor cells or within the tumor microenvironment and to an activating receptor on T cells, most commonly CD3, a key component of the T-cell receptor complex. This dual engagement induces target-dependent T-cell activation and subsequent tumor cell killing.¹² Most TCEs use antibody-derived antigen-binding domains to recognize targets on tumor cells, whereas a smaller subset uses T-cell receptors (TCRs) that recognize tumor-associated peptide – MHC complexes. In addition, TCR-mimetic antibodies (TCRm) are being explored for target cell recognition, which emulate the binding characteristics of TCRs while retaining the structural advantages of antibody scaffolds. A few TCEs are further combined with binding sites to mediate CPI and/or provide costimulatory signals. Several others (11%) engage natural killer (NK) cells or macrophages, for example, through inhibition of “don’t eat me” signals, which can also be considered as innate immune CPI (Figure 1C).¹³

Most of the immuno-modulatory activities of migs utilize targeting of two different CPIs, either in cis or trans (35%), or target costimulatory signals to tumor cells (34%). Furthermore, migs within this group combine CPI with costimulation (14%), dual costimulation, or CPI with ligand blockade, for example, by interfering with tumor angiogenesis (Figure 1C).

Finally, in cancer therapy migs are applied for dual or biparatopic blockade of growth factor receptors or their ligands, or for dual or biparatopic drug delivery including chemotherapeutic drugs and radioactive isotopes (Figure 1C).

Overview of migs in clinical development in non-cancer indications

In contrast to oncology, the portfolio of migs developed for non-cancer indications remains comparatively limited. In our analysis, we identified a total of 65 migs investigated outside of oncology, of which 14 are also being explored for cancer-related applications. Non-cancer indications are predominantly focused on inflammatory and autoimmune diseases, which together account for approximately 55% of all identified programs. This distribution likely reflects the well-established role of targeted immune modulation and cytokine signaling in these pathologies, as well as the potential of multispecific formats to enhance selectivity or functional control. Nonetheless, mig development is not restricted to these areas, and ongoing studies also encompass a broader range of therapeutic indications, including hematologic, metabolic, vascular, and neurologic disorders, as well as infectious diseases (Figure 1A).

Targets and target combinations

In total, we identified 146 different targets used by the 358 migs of category 1, combined into 212 different target combinations. Including category 2, 173 different targets have been used so far to generate migs with a total of 264 target combinations. An overview of targets and target combinations (category 1 migs) is shown in Figure 2. Because a large proportion of migs is developed for T-cell engagement, CD3 is the most widely used target structure, utilized by 41% of the migs. The CD3-targeting TCEs are frequently directed toward BCMA (16 migs), CD20 (12 migs) and CD19 (12 migs), all targets on malignant B-cells and used in already approved migs (Figure 2). Another 50 different targets are combined with CD3, both from hematologic and solid tumors.

Figure 2.

Specificities and target combinations of multispecific immunoglobulins. A) Overview of use of targets, B) Selected combinations commonly used in migs.

PD-1, a receptor for PD-L1 and PD-L2, is used in 39 different migs. The majority are combined with VEGF targeting (28%) or other immunoregulatory targets (e.g., CTLA-4, LAG-3, TIGIT, PD-L1, and TIM3) for dual CPI in cis or trans. A few migs are used for targeted delivery of PD-1 inhibition by binding to TAAs such as HER2. 4-1BB is a costimulatory receptor of the TNF-receptor superfamily that is used in 36 different migs and used mainly in combination with CPI by targeting PD-L1 (36%), followed by TAAs for targeted delivery of costimulation into the tumor. PD-L1 targeting is used in 35 different migs and combined in addition to 4-1BB (37%) with other immunomodulatory targets such as CD47, PD-1, TIGIT and LAG-3, but also with targeting of various TAAs as well as soluble factors (VEGF, TGF-β) for neutralization.

Commonly used targets of migs further comprise EGFR (HER1) (32) and HER2 (21), both members of the EGF receptor family and validated targets in cancer therapy. EGFR is often combined with c-MET targeting (40%) and one mig targeting EGFR and c-MET (amivantamab) is already approved (Table 3). Additional combinations for EGFR include CD3 for T-cell engagement, and HER3 for dual receptor inhibition. In contrast, HER2-targeting migs are often directed against two different epitopes of HER2, i.e., being biparatopic (48%), followed by CD3 targeting or combination with immuno-modulatory targets.

Table 3.

Approved multispecific immunoglobulins.

INN/ brand name	targets	format	MoA	indication	approval	company
Catumaxomab Korjuny	EpCAM x CD3	Triomab (Quadroma) (1 + 1)	TCE	malignant ascites	2009– 172,025	Lindis/Pharmanovia
Blinatumomab Blincyto	CD19 × CD3	Tandem scFv (1 + 1)	TCE	ALL	2014	Amgen
Emicizumab Hemlibra	FXIa × FX	IgG (1 + 1)	CA	hemophilia A	2017	Roche/Chugai
Amivantamab Rybrevant	EGFR x c-MET	IgG1 (1 + 1)	GRI, ADCC	NSCLC with Exon 20 insertion mutation	2021	J&J/Janssen
Tebentafusp KIMMTRAK	gp100 × CD3	scFv-TCR (1 + 1)	TCE	uveal melanoma	2022	Immunocore
Faricimab Vabysmo	VEGF x Ang-2	IgG-hetFc CH1-CL CrossFab (1 + 1)	AA	DME, AMD, RVO	2022	Roche
Mosunetuzumab Lunsumio	CD20 × CD3	IgG-hetFc (1 + 1)	TCE	R/R fNHL	2022	Roche/ Genentech
Cadonilimab Anniko	PD-1 × CTLA-4	IgG-scFv2 (2 + 2)	Dual CPI	R/M CC, GC/GEJ cancer	2022	AkesoBio
Ozoralizumab Nanozora	TNF x HSA	sdAb3 (2 + 1)	AI + HLE	RA	2022	Taisho/Pfizer/Sanofi/Ablynx
Teclistamab Tecvayli	BCMA x CD3	IgG4 Duobody) (1 + 1)	TCE	R/R MM	2023	J&J/Janssen
Glofitamab Columvi	CD20 × CD3	Fab-CrossMab-hetFc (2 + 1)	TCE	R/R DLBCL	2023	Roche
Epcoritamab Epkinly	CD20 × CD3	IgG4 (Duobody) (1 + 1)	TCE	R/R DLBCL	2023	GenMab/Abbvie
Talquetamab Talvey	GPRC5D x CD3	IgG4 (Duobody) (1 + 1)	TCE	R/R MM	2023	J&J/Janssen
Elranatamab Elrexfio	BCMA x CD3	IgG-hetFc (1 + 1)	TCE	R/R MM	2023	Pfizer
Tarlatamab Imdelltra	DLL3 × CD3	Tandem-scFv-scFc (1 + 1)	TCE	R/R SCLC	2024	Amgen
Zanidatamab Ziihera	HER2-D2 × HER2-D4	Fab-/scFv-hetFc (1 + 1)	GFRI, ADCC, CDC	HER2+ BTC	2024	JAZZ Pharm./Zymeworks
Zenocutuzumab Bizengri	HER2 × HER3	IgG-hetFc-cLC (1 + 1)	GFRI, ADCC, CDC	NSCLC, PDAC with NRG1 gene fusion	2024	Merus
Ivonescimab AK112	PD-1 × VEGF	IgG-scFv2 (2 + 2)	CPI + AA	NSCLC TKI resistant	2024	AkesoBio
Odronextamab Orspono	CD20 × CD3	IgG4-hetFc-cLC (1 + 1)	TCE	R/R DLBCL, R/R fNHL	2024	Regeneron
Linvoseltamab Lynozyfic	BCMA x CD3	IgG4-hetFc-cLC (1 + 1)	TCE	R/R MM	2024	Regeneron

Diseases: ALL – acute lymphoblastic leukemia, AMD age-related macular degeneration, BTC – biliary tract cancer, CC – cervical cancer, DLBCL – diffuse large B-cell lymphoma, DME – diabetic macular edema, fNHL – follicular non-Hodgkin lymphoma, GC – gastric cancer, GEJ – gastroesophageal junction adenocarcinoma, NSCLC – non-small cell lung cancer, PDAC – pancreatic ductal adenocarcinoma, RA- rheumatoid arthritis, R/M – relapsed/mtastatic, R/R – relapsed/refractory, MM – multiple myeloma, RVO – retinal vein occlusion, SCLC – small cell lung cancer.

MoA: AA – anti-angiogenesis, ADCC – antibody-dependent cellular cytotoxicity, CA – coagulation activation, CDC – complement-dependent cytolysis, CoStim – Co-stimulation, CPI – checkpoint inhibition, GRI – growth factor receptor inhibition, RT – receptor targeting, TCE – T-cell engagement.

Formats

The design of migs can be reduced to the combination and use of a limited set of building blocks. Variable fragments (Fvs), single-chain Fvs (scFvs), single domain antibodies (sdAbs; VHH, nanobodies), antigen-binding fragments (Fabs) and Fab-derivatives modified to allow cognate VH-VL pairing (modFab), and TCR Vα/Vβ domains are the building blocks commonly used as antigen-binding moieties. These antigen-binding moieties can further be combined with a homodimerizing wild-type Fc, a single-chain derivative thereof (scFc) or a Fc modified to allow for the formation of Fc-heterodimers (hetFc). These building blocks are either directly connected or fused together by linkers of varying length and composition, very often comprising natural sequences or arrangements of glycine and serine residues (GS-linkers) providing a flexible connection.¹⁴

The molecular formats used to generate migs are selected to align with the intended modality and application. Fc-less, and in many cases small-sized formats, are preferred when rapid systemic distribution, efficient tumor or tissue penetration, and rapid clearance are desired. In contrast, molecules designed for extended pharmacokinetic profiles incorporate an Fc region or other half-life – extension modules. Modified Fc domains may also be used to modulate or eliminate interactions with Fc receptors. Engineered valencies, including bi-targeted and avidity-driven architectures, can enhance specificity or promote receptor clustering.¹⁵ In such designs, binder orientation, epitope selection, and the nature of linkers or connectors, collectively referred to as the molecular “geometry,” are critical determinants of function. Geometric considerations are particularly important for migs that crosslink or engage distinct cell populations, such as T-cell or NK-cell engagers (see below).

We identified 89 formats used by 332 migs of category 1 and 23 further formats used by migs of category 2, in total 112 formats used so far (Figure 3). Historically, the first bispecific antibodies were either generated by chemical coupling of two different monoclonal antibodies or fragments thereof, or by the hybrid-hybridoma technology, including catumaxomab, thus being of non-human origin. Many of these early bispecific antibodies were applied for effector cell engagement (T-cells, NK-cells). Clinical studies revealed limitations of these formats, especially complications through severe cytokine release mediated by recruitment and activation of other immune cells and immunogenicity.⁶ Nevertheless, the first mig, approved in Europe in 2009, was a mouse/rat hybrid-hybridoma antibody, catumaxomab, directed against EpCAM and CD3 and used as a TCE to treat malignant ascites.¹⁶

Figure 3.

Multispecific immunoglobulin formats in clinical development. Migs are grouped into Fc-less and Fc-comprising formats and are arranged according to increasing valency and specificity.

Many of the further developments focused on the generation of Fc-less bispecific molecules, for example, by conjugation of Fab’ fragments or by genetic engineering.^2,4 This led to the approval of blinatumomab, a tandem-scFv molecule (scFv2), in 2014. Since then, various other Fc-less migs have entered clinical trials. However, one of the drawbacks of these formats, as mentioned above, is the rather short plasma half-life, often requiring continuous infusions. A breakthrough was the establishment of strategies to silence the Fcγ receptor binding functionality of the Fc region or to use IgG isotypes, such as IgG2 and IgG4 and further silenced derivatives thereof, with reduced or abolished effector functions. Furthermore, for many other indications the presence of an Fc region is less critical or even advantageous. Consequently, the majority of migs in clinical development for which we could identify the format include an Fc region (Figure 1E).

Most of the migs for which we could identify specificities are bispecific molecules (89%), followed by 10% being trispecific and only a few that are tetraspecific. Furthermore, migs combine different numbers of binding sites for each antigen. Bispecific migs are dominated by 1 + 1 (51%) and 2 + 2 formats (30%), with an emerging number of migs having a 2 + 1 stoichiometry (9%). For trispecific migs we see a similar diversity, with many having a 1 + 1 + 1 stoichiometry (77%), following by 2 + 1 + 1, 2 + 2 + 1 and 2 + 2 + 2 formats. Tetraspecific migs utilize a 2 + 2 + 2 + 2 or a 1 + 1 + 1 + 1 stoichiometry.

Regarding molecular composition, migs can further be divided into Fc-less and Fc-comprising molecules. An important effect of the Fc region is an extended half-life due to FcRn-mediated recycling of certain IgG subclasses (IgG1, IgG2, IgG4).¹⁷ Consequently, most include an Fc region (79%) while only 11% of migs were identified to be Fc-less (for 37 migs we did not find information) (Figure 1E). Fc-less migs are characterized by a rather short half-life. To circumvent this drawback, approximately half of the Fc-less migs further utilize half-life extension strategies, mainly through binding to HSA or fusion to PEG-mimetic polypeptides. Fc-comprising migs are mainly of the γ1 isotype (67%), followed by γ4 (18%) and γ2, one mig derived from a hybrid hybridoma has a hybrid mouse/rat Fc. For 42 molecules the isotype was not identified. One reason to use γ4 and γ2 Fc regions in migs is their reduced effector functions. Approximately half of the γ4 migs incorporate further mutations in their Fc region to increase silencing of effector functions. Similarly, 74% of the Fcγ1-comprising migs bear silencing mutations, a few others carry mutations to further extend half-life or to enhance effector functions, and combinations thereof.

Fc-less formats

The 38 Fc-less migs utilize 20 different formats, 6 additional formats were used in migs of category 2 (Figure 3). Two-thirds of the Fc-less migs are used for effector cell engagement. The most widely used Fc-less formats are tandem-scFv (scFv2) (6) and sdAb3 (5), followed by scFv-sdAb2 or sdAb2-scFv, and scFv-TCR (3). All of the Fc-less migs have a molecular mass of around or below 100 kDa, and all would, thereby, suffer from rapid renal clearance (renal exclusion ~50–60 kDa). Thus, many of the Fc-less migs (53%) utilize half-life extension strategies, dominated by HSA binding. Of note and outlined in the following section, combining the Fc-less building-blocks with an Fc region is another strategy for half-life extension.

The prototype molecule is a tandem arrangement of two scFv molecules fused by a peptide linker (tandem scFv, scFv2) with a size of approximately 55 kDa. This format, which was first described in 1994,^18,19 is currently used in 6 different migs, with blinatumomab as the most advanced mig, being approved in 2014 for the treatment of B-ALL. Blinatumomab (MT-103, AMG-103) is a TCE produced in CHO cells and directed against CD19 and CD3. It has the following arrangement of domains and linkers: VL(CD19)-linkerA-VH(CD19)-linkerB-VH(CD3)-linkerC-VL(CD3)-His6. Linker A is a (G₄S)₃ linker, linker C has the sequence VE(GGS)₄GGVD, linker B connecting the two scFv moieties is 5 residues long G₄S. The same arrangement is used in another tandem-scFvs, hEGFRvIII-CD3 bi-scFv.²⁰ However, here both linkers A and C have the sequence (G₄S)₃. In other tandem scFvs, including several Amgen molecules, both scFvs have the arrangement VH-VL with (G₄S)₃ linkers within each scFv and G₄S as connecting middle linker. Of note, 11 tandem scFv migs that entered clinical development were terminated or are no longer pursued due to business decisions, including etevritamab (AMG-596, EGFRvIII), eluvixtamab (AMG-330, CD33), pasotuximab (AMG-212/BAY-2010112, PSMA), and pacanalotamab (AMG-420, BCMA).

The same strategy of fusing individual binding sites in tandem is also applied to sdAb, e.g., nanobodies. Here, up to 5 sdAbs have been fused together. Three tandem sdAb (sdAb2) are currently in clinical trials, the most advanced (gefurulimab/ALXN-1720, C5xHSA) being in a Phase 3 study. In gefurulimab, the two sdAb, one targeting C5 and one binding to HSA for half-life extension, are connected by a flexible 15 aa linker ((GGGGA)₂GGGGS).²¹ Most advanced are triple sdAbs (sdAb3), with one bispecific sdAb3, ozoralizumab, directed against TNF and HSA being approved for the treatment of rheumatoid arthritis,²² four others being in clinical development. In ozoralizumab (ATN-103), a central HSA-binding nanobody is flanked by two identical TNF-binding nanobodies connected by G₄SG₃S linker.²³ Triple sdAbs can further be used to make trispecific molecules, e.g., sonelokimab, which targets IL-17A, IL-17F and HSA and is currently in Phase 3 studies.²⁴ The approach was further extended to generate tetra- and pentavalent molecules by fusing 4 or 5 sdAbs in a row. SAR444200 is a tetravalent sdAb4 comprising two binding sites for GPC3, one for the TCRβ, and one for HSA.²⁵ Brivekimig (SAR442970) is a trispecific mig (Phase 2) comprising 5 nanobodies (sdAb₅), two directed against TNF, two against OX40L and one against HSA.²⁶ Similarly, lunsekimig (SAR443765) is a pentavalent mig that binds biparatopically to TSLP and IL-13 and monovalently to HSA.²⁷ Lunsekimig is in Phase 3 trials for the treatment of chronic obstructive pulmonary disease.

SdAb and scFvs were further combined to make trispecific migs by fusing either 2 sdAb to one scFv (sdAb2-scFv; scFv-sdAb2) or one sdAb to 2 scFv (sdAb-tandem scFv). SdAb2-scFv comprise two migs currently in clinical development, one further being terminated. Gocatamig (HPN-328) is a TCE composed of one scFv directed against CD3 fused to an HSA-binding sdAb and a DLL3-binding sdAb, further comprising a C-terminal His6-tag.²⁸ Flexible (G₄S)₂ linkers are used to connect the two sdAb to the scFv moiety. This so-called TriTAC format is designed to overcome challenges of TCEs, having a small size to facilitate tumor penetration combined with a long serum half-life due to binding to serum albumin.²⁹ Gocatamig is currently in a Phase 2 trial for treatment of patients with relapsed/refractory extensive-stage small cell lung cancer (SCLC). The same format is used in podentamig (HPN-217) directed against BCMA, HSA and CD3, and HPN-526 targeting mesothelin, HSA and CD3, both in Phase 1. In CLN-978, an HSA-binding sdAb is fused to a low-picomolar affinity anti-CD19 scFv followed by an anti-CD3 scFv (sdAb-scFv2).³⁰ HPN-536 is a further TriTAC targeting mesothelin, HSA and CD3, but here the domains are arranged in the order sdAb-sdAb-scFv, with the mesothelin binding sdAb placed at the N-terminus and the CD3-binding scFv at the C-terminus.³¹ HPN-536 completed a Phase 1 study in 2023.

The variable domains of two different antibodies can further be arranged in a way that leads to assembly of a bispecific molecule. This was originally shown for the diabody format where two polypeptide chains were expressed to assemble into a heterodimeric bispecific molecule.³² The two polypeptide chains have the composition VH(A)-linker-VL(B) and VH(B)-linker-VL(A), with the variable domains within one polypeptide chain connected by a short glycine-serine linker that prevents pairing of VH and VL of one polypeptide chain, thus forcing dimerization. However, here homo- and heterodimers can form, with only the heterodimers being bispecific and the homodimers being nonfunctional. This drawback was solved for instance by converting the diabody format (Db) into the DART (dual affinity retargeting) format.³³ In this format, the VH and VL domains within one polypeptide chain are connected by a 9 residues long GGGSGGGG-linker. Originally, dimers were stabilized by a C-terminal extension to allow disulfide bond formation. In a comparative study, a DART directed against CD19 and CD3 was shown to be more potent in redirecting B-cell lysis then a corresponding tandem scFv. In flotetuzumab (MGD006),³⁴ a DART targeting CD123 × CD3 that is currently in Phase 1, the two polypeptide chains are C-terminally extended by heterodimerizing coiled-coil peptide sequences (K-coil, E-coil) forming charge-charge interactions.³⁵

Alternatively, the two chains of a diabody can be connected by a further 15 to 20 residue long middle linker, generating single-chain Diabodies.³⁶ Here, the middle variable domains assemble in a scFv-like structure. Shortening this middle linker leads to head-to-tail homodimerization of two polypeptide chains, resulting in a ~ 100 kDa tetravalent bispecific molecule, the TandAb format.³⁷ This format is used in acimtamig (AFM-13), an NK-cell engager directed against CD30 and CD16. In acimtamig, the four variable domains within one polypeptide chain are all connected by 9 residue long (G₂S)₃ linker. Acimtamig is currently in several Phase 2 trials for treatment of various forms of CD30⁺ lymphomas.^38,39

Fc-less bispecific antibodies can, furthermore, be designed using Fab and scFv as building blocks. This was originally shown by fusing scFv to the C-terminal ends of the light chain and/or the Fd-fragment of a Fab (Fab-scFv/-scFv).⁴⁰ BNT142, currently in a Phase 1/2 study, is a lipid-nanoparticle encapsulated mRNA formulation encoding RiboMab02.1, a trivalent bispecific Fab-scFv/-scFv TCE comprising two scFv directed against CLDN6 fused to the C-termini of an anti-CD3 Fab.⁴¹ The same format is used in CDR404 (Phase 1), a Fab-scFv/-scFv TCE (M-gager) comprising two TCR-mimetic scFv targeting an MHC-presented MAGE-A4 peptide on solid cancer cells fused to a CD3-targeting Fab.⁴² Since a Fab is a heterodimer, the Fab-scFv/-scFv format allows furthermore the generation of trispecific, trivalent molecules without further modifications by fusing scFvs with different specificity to the Fab chains. This was applied to, for example, galvokimig (UCB9741), which has recently completed a Phase 1/2 trial. Here, a Fab directed against IL-17A/F is C-terminally fused with scFvs directed against HSA and IL-13 using 11 residue long linkers (SG₄TG₄S or S(G₄S)₂). The same format was used to generate donzakimig (UCB1381) directed against IL-13, IL-22 and HSA, with the Fab and scFvs connected by the S(G₄S)₂) linker.

The strategy of Fab and scFv as building blocks was further extended by fusing one or two scFv furthermore to the N-termini of a Fab. A-319 is a bispecific bivalent scFv-Fab in which an anti-CD3 scFv fused to the Fd chain of an anti-CD19 Fab.⁴³ A-319 is currently in a Phase 1 study in patients with active or refractory systemic lupus erythematosus. A-337 is a bispecific trivalent molecule composed of two EpCAM-targeting scFv fused to the N-termini of an CD3-binding Fab (scFv2-Fab), a format belonging to the iTab platform.

The Fab moiety can be substituted by soluble TCRs, which are composed of Vα-Cα and Vβ-Cβ chains, adopting a Fab-like structure. Although not antibodies in the strict sense, the Vα/Vβ domains form a binding site capable of recognizing peptides displayed by MHC molecules, with applications to recognize tumor cells or infected cells.⁴⁴ Fusion proteins of TCR with antibody moieties end with -fusp. Tebentafusp (IMCgp100) is such a fusion protein (ImmTAC) designed as TCE by fusing a CD3-binding scFv to an affinity-improved anti-gp100 TCR.⁴⁵ Here, the scFv is fused to the N-terminus of the Vβ domain by a short G₄S linker. Since removal of the transmembrane domains from the TCR reduces the stability due to increased dissociation, the soluble TCR moiety is further stabilized by introducing an artificial disulfide bond, mutating T48 in Cα and S57 in Cβ to cysteines.⁴⁶ Tebentafusp was approved in 2022 for the treatment of uveal melanoma.⁴⁷ Two additional ImmTACs are currently in clinical testing, brenetafusp (IMC-F106C) (Phase 3) directed against PRAME,⁴⁸ and IMC-R117C (Phase 1/2) directed against PIWIL1.

The development of the DutaFab technology has enabled generation of dual-targeting Fab molecules without the need to fuse different binding moieties.⁴⁹ Here, the complementarity-determining regions (CDRs) of the VH domain are directed against a first target, and the CDRs of the VL against a second target. Both paratopes can be independently selected and it was shown that DutaFabs are capable of binding two antigens simultaneously since the two paratopes are facing away from each other. Zifibancimig (RO-7250284) is a first DuaFab in clinical development (Phase 1/2). The molecule is directed against VEGF-A and angiopoetin-2 (Ang-2) and studied in patients with neovascular age-related macular degeneration (nAMD).

Natural serum proteins can also be applied to combine different antibody moieties with a further biological function. An example is fusion of antibody fragments to IL-15, which has immuno-stimulatory activity that enhances anti-tumor immune responses, for example, of NK cells. This approach has been applied to generate trispecific killer engagers (TriKEs).⁵⁰ GBT-3560 is a TriKE composed of a CD16-targeting sdAb, IL-15, and an anti-CD33 scFv, currently in Phase 1.⁵¹

Conditionally active Fc-less formats

T-cell engaging molecules represent a major class of migs. Almost all of these migs include a CD3-binding site for T-cell engagement and activation. Binding to CD3 must be well balanced to avoid systemic T-cell activation and excessive cytokine release, which can lead to on-target, off-tumor activities and severe side effects, e.g., cytokine release syndrome (CRS). Several strategies have been used to cope with these obstacles, for example the conversion of the TCEs into molecules which are conditionally active at the tumor site only.¹² One approach is to mask the CD3 binding site while in circulation and release the mask at the tumor site, for example, through proteolytic cleavage. Several Fc-less formats were modified and generated to be conditionally active. This includes AMX-818, AMX-500, and VIR-5525 that include PEG-mimetic polypeptide sequences (XTEN) fused to the N- and C-terminus of a tandem scFv directed against CD3 and either HER2, PSMA, or EGFR. The XTEN sequences have a length of 200 or more amino acids and further comprise a cleavage site of a protease dysregulated in tumors. It sterically masks the two binding sites, reducing the affinity for HER2 about 10-fold and for CD3 about 6-fold.⁵² The XTEN sequences can be removed at the tumor site, liberating the binding sites. Furthermore, they mediate a prolonged half-life.⁵³ These XPAT molecules are produced in E. coli. AMX-818 directed against HER2 and CD3 is currently in a Phase 1/2 trial. In a non-human primate study, AMX-818 demonstrated a 200-fold increase in tolerated C_max in comparison to unmasked variants, thus providing a strong safety margin.⁵²

In another approach, masking of binding sites was achieved by fusing short peptides that interact with the paratope, further comprising proteolytic cleavage sites. This approach is utilized in the TRACTr format. JANX007 is composed of an anti-HSA sdAb for half-life extension, followed by CD3-masking peptide, a cleavage site for proteases of the tumor microenvironment, and an anti-CD3 scFv fused to an anti-PSMA Fab.⁵⁴ Masking of the CD3-binding site reduces CD3 binding approximately 600-fold. JANX007 is currently in a Phase 1 study in patients with castration-resistant prostate cancer. A similar approach was used to generate JANX008 (Phase 1), a TRACTr targeting HSA, CD3 and EGFR. Here, the EGFR-binding Fab is in addition masked by a cleavable peptide to further reduce on-target, off-tumor activities.⁵⁵

An entirely different approach is utilized in the COBRA (COnditional Bispecific Redirected Activation) format. This format is composed of one long polypeptide chain, consisting of a tumor-targeting sdAb fused to CD3-targeting VH-VL domains, followed by another tumor-targeting sdAb, a proteolytic cleavage site, dummy VH-VL domains, and an anti-HSA sdAb for half-life extension.⁵⁶ The linkers in the two VH-VL units have lengths that prevent intra-chain pairing of the VH and VL domains, like that used to generate diabodies. Hence, two polypeptide chains assemble into a homodimeric molecule in which the CD3-binding domains pair with the dummy variable domains. Thus, due to the presence of dummy VH and VL domains, these homodimers do not form functional CD3-binding sites. However, after cleavage the C-terminal part comprising the dummy variable domains and the HSA-binding sdAb is removed and the N-terminal part, comprising two TAA-targeting sdAbs and the CD3-binding variable domains, assemble into a diabody-like homodimer with flanking sdAbs, thereby restoring CD3-binding activity. One COBRA mig, TAK-186 (MVC-101) targeting EGFR was in a Phase 1/2 study. TAK-186 utilizes a matrix metalloproteinase-9 (MMP9) cleavable linker (SGGPGPAGMKGLPGS) to connect the first sdAb-VH-VL-sdAb part targeting EGFR and CD3 with the inactive VL-VH-anti-HSA sdAb part.⁵⁷ Further development of TAK-186 was, however, terminated, due to limited anti-cancer activity. A recent study revealed the occurrence of anti-drug antibodies (ADAs) in many of the treated patients, with a preference for the HSA-binding VHH.⁵⁸

Fc-less formats no longer in clinical development

It should be noted that, in addition to TAK-186, a few Fc-less formats have entered clinical developed but are no longer pursued, including F(ab‘)₂ molecules generated by chemical conjugation and mainly used in early studies for effector cell engagement.⁵⁹ NM21-1480 is a trispecific scDb-scFv fusion protein directed against 4-1BB, PD-L1, and HSA, thus combining CPI, costimulation and half-life extension in one molecule. NM21-1480 was in a Phase 1/2 study terminated in 2024 due to a business decision. MM-111 is a scFv-HSA-scFv fusion protein for dual inhibition of HER2 and HER3 using the HSA moiety for half-life extension.⁶⁰ MM-111 was in a Phase 2 study terminated in 2015. TF2 is a bispecific, trivalent Fab2-DNL-Fab molecule directed against CEA and a hapten-peptide bearing a single DOTA moiety developed for pre-targeted radioimmunotherapy, but it is no longer in clinical studies.⁶¹ DT2219ARL is an immunotoxin generated by fusing a diphtheria toxin moiety to a scFv2 part directed against CD19 and CD22, with the variable domains of each scFv moieties connected by aggregation reducing/stabilizing linkers (ARL).⁶²

Fc-comprising formats

The majority (88%) of the 322 migs of category 1 for which we could identify the format include an Fc region. The Fc region is a versatile building block to fuse antigen-binding moieties to the N- and/or C-termini. Furthermore, it mediates an extended half-life due to FcRn-mediated recycling.⁶³ In migs of category 1, we identified 69 different formats using an Fc region, 18 additional formats were used in migs no longer listed as in clinical development (category 2). Of the Fc-comprising migs of category 1, 34% are IgGs or have an IgG-like structure, i.e., are composed of two heavy chains and two light chains. The remaining migs are extended IgGs (38%) or Fc-fusion proteins comprising two or more antigen-binding building blocks fused to the N- and/or C-terminus of the Fc part (28%).

IgG and IgG-like migs

The production of bispecific IgGs composed of two different heavy and light chains faces two challenges: 1) the heavy chain problem, which requires heterodimeric assembly of two different heavy chains, and 2) the light chain problem, which requires cognate pairing of the light chains with their corresponding heavy chains. Expression of all four chains within one cell can result in formation of 16 different combinations, with only 2 of them being bispecific. Therefore, several strategies have been developed and applied to solve these problems.^64,65

A straight-forward approach is the chemical conjugation of two different monospecific antibodies, e.g., IgGs or Fab’ fragments. Such bispecific conjugates were especially used in early clinical trials, but most of them have not been further pursued.^4,6 Nevertheless, several migs based on chemical IgG-IgG conjugates composed of a tumor-targeting antibody and a CD3-binding antibody are currently in clinical trials. These TCEs are preincubated in vitro with patient-derived, activated T-cells, resulting in bispecific antibody-armed activated T-cells (BAT) which are injected back into the patient. Thus, T-cells are loaded ex vivo with the bispecific antibodies. Several of these migs, targeting EGFR, HER2, or CD30, are in Phase 1 or 2 clinical studies.^66–68 They resemble chimeric antigen receptor T cells (CAR-Ts), with the differences being that the T cells have not been genetically engineered and the bispecific antibodies are not inserted into the membrane of the T-cells but non-covalently bound.

Catumaxomab, a bispecific IgG targeting EpCAM and CD3, was generated by a technology already developed in 1983 by fusing two different hybridomas into a hybrid-hybridoma, or quadroma.⁶⁹ Catumaxomab comprises heavy and light chains from a mouse IgG2a hybridoma antibody directed against EpCAM and a rat IgG2b hybridoma antibody directed against CD3.⁷⁰ A preferential species-restricted heavy/light chain pairing was described for these mouse/rat hybridoma. Furthermore, it was found that the heterodimeric rat/mouse heavy chains show an intermediate binding to protein A, which allows a single-step purification.⁷¹ Catumaxomab is described as a trifunctional antibody (Triomab) capable of recruiting T cells to tumor cells and to activate accessory immune cells through its Fc region.⁷² Catumaxomab (Removab) received first approval in Europe in 2009 for the intraperitoneal treatment of malignant ascites, but it was withdrawn from the market in 2017 for commercial reasons. It was relaunched in Europe in 2025 as Korjuny.⁷³ Korjuny prolongs the time patients live without the need for paracentesis, indicating that fluid builds up slower after treatment, and has an acceptable safety profile, although side effects occur more frequently than with paracentesis. Due to the rat/mouse origin catumaxomab is immunogenic and ADAs have been described to occur in most patients after 11 to 16 days. Hence patients are treated on days 0, 3, 7 and 10, i.e., treatment is stopped before ADA development.⁷⁴

Fab-arm exchange (FAE) is a naturally occurring process found for IgG4 molecules that leads to exchange of half-antibodies (HC/LC) between two antibodies.⁷⁵ Key determinants of this exchange are residue S228 in the core hinge region and R409 in the CH3 domain of IgG4.⁷⁶ This process was adapted to generate bispecific IgG1 and IgG4 antibodies using controlled Fab-arm exchange (cFAE). Here, the two antibodies are produced separately and subsequently half-antibodies are exchanged between the two IgGs by a process involving mild reduction conditions. To facilitate heterodimeric assembly and a directional exchange, matching mutations were introduced into the CH3 domain and Ser228 in the hinge region of IgG4 was mutated to proline to prevent in vivo exchange.^76,77 IgG4 Duobodies carry S228P in both chains and F405K and R409K in one of the chains. IgG1 Duobodies carry a F405L mutation in one chain and a K409R mutation in the other chain. Fifteen migs generated by cFAE, so-called Duobodies, are in clinical development, four are already approved (teclistamab, talquetamab, epcoritamab, and amivantamab). Teclistamab and talquetamab are IgG4-based TCEs targeting BCMA and CD3 or GPRC5D and CD3, respectively, epcoritamab is an IgG1 TCE targeting CD20 and CD3, and amivantamab is an IgG1 Duobody targeting EGFR and c-MET.⁷⁸

The heavy chain problem was further solved by introducing mutations into the CH3 domains that force heterodimeric assembly and avoid formation of homodimers (Table 1). The knobs-into-holes (kih) strategy was the first approach demonstrating feasibility. Here, a tryptophan possessing a large side chain is introduced into a first CH3 (T366W) and three positions in the second CH3 are mutated to amino acids with smaller side chains (T366S, L368A, Y407V).⁷⁹ These Fc heterodimers were further stabilized by introducing an additional disulfide bond (S354C, Y349C) (C-kih). Subsequently, further strategies were established based on steric and/or electrostatic complementarity introducing either a few mutations into the CH3 domain or regions from other immunoglobulin domains, for example, from IgA (SEED) or the T-cell receptor (BEAT).⁸⁰ Furthermore, an approach (Fc*) utilizing a differential purification strategy was developed by introducing mutations affecting protein A binding into one of the heavy chains (H435R, Y436F).⁸¹ Thus, heterodimeric IgGs elute at an intermediate pH value and can be separated from homodimers. The Fc* approach was further combined with kih and C-kih mutations to facilitate generation of pure heterodimer preparations.

Table 1.

Heterodimerizing fc variants.

Name	isotype	chain 1	chain 2	#	example
C-kih	γ1	S354C, T366W	Y349C, T366S, L368A, Y407V	48	glofitamab
kih	γ1	T366W	T366S, L368A, Y407V	33	mosunetuzumab
Fc*	γ1	H435R, Y436F	wt	12	odronextamab
kih + Fc*	γ1	T366W	T366S, L368A, Y407V, H435R, Y436F	6	obrixtamig
C-kih + Fc*	γ1	S354C, T366W	Y349C, T366S, L368A, Y407V, H435R, Y436F	2	emzotamig
XmAb	γ1	S364K	Q295E, L386D, K370S, N384D, Q418E, N421D	10	plamotamab
ZW1	γ1	T350V, T366L, K392L, T394W	T350V, L351Y, F405A, Y407V	6	zanidatamab
kih-KA	γ1	T366W, K409A	T366S, L368A, F405K, Y407V	2	anbenitamab
Kih-ETYY	γ1	S354Y, T366Y	Q347E, Y407T	1	GB261
DEKK	γ1	L351D, L368E	L351K, T366K	7	zenocutuzumab
RCVT/CEW	γ1	Y349C, K360E, K409W	Q347R, S354C, D399V, F405T	3	azerutamig
ELV/LR	γ1	L351E, Y407L, K409V	T366L, D399R	2	fidasimtamab
KE/KKR	γ4	R409K, K439E	E356K, R409K, H435R	3	emicizumab
EE/KR	γ1	D221E, L368E	D221K, K409R	1	tilrekimig
EEE/RRRR	γ2	C223E, P228E, L368E	C223R, E225R, P228R, K409R	1	elranatamab
cFAE γ1	γ1	F405L	K409R	7	amivantamab
cFAE γ4	γ4	F405L, R409K	wt	6	teclistamab
BEAT	γ1	TCR Cγ insertion	TCR Cβ insertion	2	ISB1342
SEED	γ1	IgG/A hybrid	IgA/G hybrid	1	M1231 ADC

These strategies for solving the heavy chain problem have been combined with strategies solving the light-chain problem. One approach is the use of a common light chain (cLC), which circumvents any further modifications in the Fab arms.⁸² Antigen-binding sites comprising a common VL can be selected from combinatorial antibody libraries or transgenic animals comprising a single VL.^81,83 In total, 39 bispecific IgGs incorporate a cLC, 28 using a heterodimerizing Fc and 11 using an Fc*. This includes emicizumab, an agonistic antibody targeting coagulation Factor IXa and X approved for treatment of hemophilia A, which uses a heterodimerizing Fc (electrostatic steering), and odronextamab, a CD20 × CD3 targeting TCE approved for treatment of relapsed/refractory follicular lymphoma or diffuse large B-cell lymphoma, which uses the Fc* technology.⁸¹

Another approach successfully applied to generate bispecific IgGs utilizes post-production assembly of two different antibodies comprising a heterodimerizing Fc. This process has the advantage that any Fab arms can be used without further engineering. This strategy was applied to generate several approved migs, including mosunetuzumab, a TCE directed against CD20 and CD3 and approved for the treatment of relapsed or refractory follicular lymphoma, and elranatamab, a TCE directed against BCMA and CD3 and approved for the treatment of multiple myeloma.⁸⁴ Mosunetuzumab is an IgG1 antibody that includes kih mutations. For generation of this so-called TDB (T cell-dependent bispecific antibody) and as described for cevostamig, a FcRH5 × CD3 TCE, half-antibodies are produced separately purified, and thereafter combined to form an IgG-shaped bispecific antibody.⁸⁵ Elranatamab is an IgG2 antibody using electrostatic steering for heavy chain assembly. For elranatamab, in the first heavy chain 3 positions are mutated to arginine in the hinge region and one further in the CH3 domain, and in the second heavy chain 2 positions in the hinge and one in the CH3 domain are mutated to glutamic acid (EEE/RRRR).⁸⁶

Alternatively, heterodimerizing heavy chains have been combined with modified Fab arms (modFab) engineered in a way to allow cognate pairing with their corresponding heavy chain in cells expressing two different heavy and light chains.⁶⁴ We identified 13 migs using this approach, which comprise mutated Fab arms combined with different Fc modifications for heavy chain heterodimerization. These Fabs include engineered CH1/CL domains with a newly introduced disulfide bond, mutations introducing charge-charge pairs, or combinations thereof, as well as mutated VH-VL interfaces. Five migs use the DuetMab format, which comprises in one Fab arm a light chain carrying S121C and C214V mutations and one heavy chain with F126C and C233V mutations, thus removing the natural disulfide bond and introducing one at a new position between the CL/CH1 interface. This forces pairing of the mutated light chain with the mutated heavy chain due to the engineered disulfide bond.⁸⁷ The most advanced DuetMabs are in Phase 3, including rilvegostomig (AZD2936), a dual checkpoint-inhibitor targeting PD-1 and TIGIT, and volrustomig (MEDI5752), a dual checkpoint-inhibitor targeting PD-1 and CTLA-4. The Fc region of rilvegostomig includes C-kih mutations for heterodimerization. Reozalimab (LY3434172, IBI318) is a dual checkpoint inhibitor targeting PD-1 and PD-L1 currently in Phase 2.⁸⁸ Here, electrostatic steering is used for cognate HC/LC pairing. In heavy chain 1 two residues in CH1 were mutated to glutamic acid (L124E, Q179E) and a lysine at position 145 was mutated to threonine. Two adjacent positions in the cognate kappa CL were mutated to arginines (S131R, S176R) plus mutating V133 to glycine. Furthermore, serine 188 in heavy chain 2 was mutated to lysine and three mutations were introduced in the cognate lambda CL (S176E, Y178E, T212A).

The IgG format with modified Fab arms was further applied to generate trispecific molecules by combining a Fab with dual-Fab that competitively binds either antigen A or B. This Fab-/dual-Fab-hetFc format was used to generate SAIL66, a trispecific mig targeting CLDN6 with its Fab arm and CD3/CD137 with its dual-Fab arm, with both arms fused to heterodimerizing Fc.⁸⁹ Charge-charge pairs were introduced into the constant domains of the Fab/dualFab arms to force cognate light and heavy chain pairing. Thus, SAIL66 redirects T cells to CLDN6-expressing tumor cells, but can only bind CD3 or CD137, which reduces the risk of systemic T-cell activation by simultaneous binding to CD3 and CD137 (4-1BB), a costimulatory receptor of the TNF superfamily.

Instead of introducing mutations into variable and/or constant domains of a Fab fragment, other approaches completely substitute the CH1/CL domains of one Fab by alternative heterodimerization domains (IgG-hetFc Cα/Cβ Fab). For example, in D3L-001, a bispecific IgG (Wuxibody) targeting HER2 and CD47, these domains were substituted by the Cα and Cβ domains of the human TCR.⁹⁰

The CrossMab (CrossFab) approach combines a heterodimerizing Fc with Fabs having an altered arrangement of the variable and/or constant domains.^91,92 In IgG-hetFc-CH1/CL CrossFabs, the CH1 and CL domains of one Fab arm are exchanged between heavy and light chain. Thus, in these molecules the VH domain is fused to the CL domain followed by the Fc region, and the cognate light chain is composed of the VL domain fused to the CH1 domain. This prevents pairing of the CrossFab arm with a natural heavy chain. The format is, for example, used in faricimab for dual blockade of VEGF and Ang-2, which is approved for the treatment of nAMD or diabetic macular edema (DME).⁹³ Similarly, in an IgG-hetFc-VH/VL CrossFab, the VH and VL domains of one Fab arm are exchanged between heavy and light chain. This arrangement was applied to generate tobemstomig (RO7247669), a dual checkpoint inhibitor targeting PD-1 and LAG-3 currently in Phase 2.⁹⁴ It should be noted that further mutations were introduced in the CH1 and CL domains of the CrossFab arm (R108A, E123R, Q124K in the CL domain and K145E, K221E in the CH1 domain. Furthermore, the entire light chain can be swapped with the Fd fragment of the heavy chain. To our knowledge, this format was used to generate PF-07826390 targeting LILRP1 and LILRP2, which is, however, no longer in clinical development.

The light chain problem was further solved by converting the heterodimeric Fab arm into a single-chain moiety (scFab).⁹⁵ Two such scFab are fused to a heterodimerizing Fc in the ZweiMab format (scFab-/scFab-hetFc).⁹⁶ In these scFab the C-terminus of the light chain is fused to the N-terminus of the heavy chain by a 38 residues long linker (G₄S-(EGKSSGSGSESKST)₂-G₄S). This format is used in obrixtamig (Phase 2), a TCE targeting DLL3 and CD3, further comprising a kih-Fc*⁹⁷ and in BI-765049 (Phase 1), a TCE targeting B7-H3 and CD3.⁹⁸

A strategy that relies on a three-step purification process is used in κ/λ-bodies. Here, a common heavy chain is combined with a first light chain of the kappa isotype and a second light chain of the lambda isotype. By using appropriate columns, bispecific antibodies comprising kappa and lambda light chains are purified.⁹⁹ Corresponding antibodies are isolated from combinatorial libraries using a common VH combined with kappa or lambda VL sequences. Three of these κ/λ-bodies are in clinical Phase 1, NI-1801 which is directed against mesothelin and CD47 blocking a “don-t-eat-me” signal on tumor cells by interfering with the CD47-SIRPa interaction on myeloid cells,¹⁰⁰ NILK-2301, a TCE directed against CEACAM5 and CD3, and zeripatamig (TG-1801) directed against CD19 and CD47.¹⁰¹

Finally, an entirely different approach is used by mAb² antibodies. These antibodies also have an IgG-like structure and are bispecific. However, here the second binding site is introduced at the end of the Fc region by mutating structural loops of the CH3 domain into antigen-binding sites (Fcab).¹⁰² Since mAb² are symmetric molecules comprising two identical heavy and light chains, they are tetravalent with two binding sites for each antigen. The format is utilized in FS120, directed against OX40 and 4-1BB for dual co-stimulation, and FS222, targeting PD-L1 and 4-1BB, for combined CPI and co-stimulation, both in Phase 1.^103,104

Extended IgG

Migs with an extended IgG structure are composed of an IgG, either mono- or bispecific, and additional binding moieties (sdAb, scFv, Fab) fused to the N- and/or C-terminus of the IgG. These extended IgGs can be symmetric by using a monospecific IgG or asymmetric by using an IgG with a heterodimerizing heavy chain (mono- or bispecific). In total, we identified 106 migs with an extended IgG applying 23 different formats. Two of these extended IgG formats (Fab-IgG-hetFc with a VH-VL CrossFab, IgG-scFv2 (HC)) are used in three approved migs (glofitamab, cadonilimab, ivonescimab).

Extended IgGs with C-terminal fusions

Various bispecific formats were generated by fusing a scFv or a sdAb to the C-terminus of the light or heavy chain of an IgG. These tetravalent bispecific formats of often referred to as Morrison-type antibodies.¹⁰⁵ The advantage is that the IgG itself does not need to be modified and can be used directly as a building block, at least for symmetric molecules.

The most commonly used format is IgG-scFv2 (HC), with an scFv moiety fused to the C-terminus of the IgG heavy chain. We identified 47 different migs using this 2 + 2 format, two of them (cadonilimab, ivonescimab) approved, three further in Phase 3, including 1 bispecific antibody-drug conjugate (ADC) (BL-B01D1). Seven others are no longer in clinical development. For several of these IgG-scFv2 (HC) there is a sequence entry in the GSRS database. Analysis of these sequences showed that linkers of different length are used to connect the scFv ((VH-VL or VL-VH orientation) to the heavy chain ranging from short G₄S (e.g., cabotamig), intermediate ((G₄S)₂, (G₄S)₃), to long (G₄S)₄ linkers (e.g., cadonilimab). Most of these IgG-scFv2 (HC) use linkers of 15 or 20 amino acids to provide sufficient flexibility and accessibility of the scFv binding site. The majority of these migs are applied for dual targeting strategies, only a few (6) are TCEs. In these TCEs, the anti-CD3 binding site is encoded by the scFv. Its C-terminal location presumably prevents simultaneous binding of two CD3 molecules and, thus are not T-cell activating on their own.¹⁰⁶ Cadonilimab (AK104) is a dual checkpoint inhibitor, targeting PD-1 with its IgG moiety and CTLA-4 through the scFv, approved for the treatment of gastric and cervical cancer.¹⁰⁷ Ivonescimab (AK112), which targets PD-1 and VEGF and thus combines CPI with anti-angiogenesis, is approved for treatment of EGFR-mutated locally advanced or metastatic non-squamous non-small cell lung cancer (NSCLC).¹⁰⁸

Alternatively, scFvs were fused to the C-terminus of the light chain (IgG-scFv2 (LC)). Three such migs are currently in clinical testing, all TCEs in Phase 1 with the CD3-binding site fused as scFv to the light chain. This includes CT-95 targeting mesothelin and CD3, BA1202 targeting CEA and CD3, which was described to bind bivalently to the CEA, but functionally monovalent to CD3,¹⁰⁹ and BA3182, a conditionally active TCE (Conditionally Active Biologic, CAB) targeting EpCAM and CD3.¹¹⁰

C-terminal fusion to an IgG heavy chain is also applied using sdAb (IgG-sdAb2 (HC). Of 12 migs identified that use this format, five are no longer in clinical development or have an unknown status. In this format, flexible glycine-serine linkers are also used to connect the sdAb to the heavy chain, like those used in IgG-scFv2(HC) molecules. The most advanced molecule is BNT327 (pumitamig, PM8002) directed against PD-L1 and VEGF and currently in Phase 3 studies to treat SCLC patients.¹¹¹

Instead of sdAb, scaffold proteins with antigen-binding activity have been used to generate extended IgGs (IgG-scaffold2). Three such molecules are in clinical development (Phase 1 or 2), all composed of an anti-4-1BB anticalin¹¹² fused to the C-terminus of the heavy chain of a tumor-targeting antibody. Because anticalins are not antibodies their INN end with -fusp. For example, cinrebafusp alfa (PRS-343) (Phase 2) targets HER2 and provides tumor-localized T-cell costimulation.¹¹³

Common light chain (cLC) Fabs were furthermore used to generate tetravalent bispecific molecules, fusing the Fd fragment of an additional Fab to the C-terminus of an IgG heavy chain. This solves the light chain problem in the context of IgG-Fab2 fusion proteins. One such mig, CTX-8371, targets PD-1 and PD-L1 and is in Phase 1. Of note, it was found that co-engagement of PD-1 and PD-L1 induces proteolytic cleavage and loss of cell surface of PD-1, further enhancing blockade of the PD-1/PD-L1 interaction.¹¹⁴

Extended IgGs with C-terminal fusions have further been applied to generate asymmetric migs with a 2 + 1 stoichiometry. This includes fusion of an scFv to monospecific IgG comprising a heterodimerizing Fc (IgG-hetFc-scFv). This format is used in ABL301, a so-called Grabody-B designed to deliver an antibody across the BBB.¹¹⁵ ABL301 (SAR446159) has two binding sites targeting aggregated forms of α-synuclein associated with different forms of neurodegeneration and one binding site for IGF1R mediating BBB penetration.¹¹⁶ The approach further includes IgG-hetFc-CrossFab molecules where a CrossFab is fused to the C-terminus of an IgG with an heterodimerizing Fc. This format is used in trontinemab, targeting with its IgG amyloid-β and with its CrossFab the transferrin receptor (TfR) used as BBB shuttle.¹¹⁷ Heavy chain heterodimerization is achieved by C-kih. The VH-CL chain of the CrossFab is fused by an 18 residues long glycine-serine linker to a knob-comprising heavy chain and the corresponding VL-CH1 is co-expressed together with the other heavy (hole) and light chains. To further force correct pairing of the LC and HC, mutations were introduced into the CH1 and CL domains (K145E and K221E in the HC; E123R, Q124K in the LC).

Extended IgGs with N-terminal fusions

Different antigen-binding moieties have been fused to the N-terminus of the heavy or light chain of an IgG to generate tri- or tetravalent bispecific or trispecific molecules with an extended IgG structure. Most of the migs in clinical development utilize Fabs fused to the N-terminus of the heavy chain. Thus, 6 migs are symmetric tetravalent (2 + 2) Fab2-IgG fusion proteins and 11 are trivalent bi- or trispecific (2 + 1, 1 + 1 + 1). As for bispecific IgGs, the formats face the light chain problem. Thus, the same or similar approaches were adapted for 2 + 2 Fab2-IgG and 2 + 1 Fab-IgG fusion proteins. Four 2 + 2 migs with N-terminal Fab fusions, one 2 + 1 bispecific antibody and one 1 + 1 + 1 trispecific antibody use cLCs. SSGJ-706, SSGJ-707, and KK2269 are 2 + 2 Fab2-IgG cLC molecules. KK2269 (Phase 1) is directed against EpCAM and CD40 for targeted co-stimulation and derived from the REGULGENT platform.¹¹⁸ SSGJ-706 targets PD-1 and PD-L1 and SSGJ-707 targets PD-1 and VEGF, with the Fd fragment of the outer PD-1-binding Fab fused by a (G₄S)₃ to the heavy chain of the IgG.¹¹⁹ SSGJ-707 is the most advanced mig with a Fab2-IgG cLC format, currently in Phase 3 to treat patients with advanced NSCLC.

One further Fab2-IgG (bafisontamab) uses the FIT platform (Fab-in-tandem immunoglobulins).¹²⁰ Here, the light chain of the outer Fab is fused to the IgG heavy chain, combined with expression of the corresponding Fd fragment and the second light chain. Although the Fd fragment can bind the second light chain to form a Fab, purification by, for example, protein-A results in bispecific tetravalent molecules. In principle, this resembles a CrossFab approach. Bafisontamab (EMB-01) is in Phase 2 (with a Phase 3 study having an unknown status) and targets EGFR and c-MET.¹²¹

OT-A201 is a 2 + 2 Fab2-IgG targeting two different immune checkpoints in Phase 1. It uses different sets of complementary pairing mutations in the CH1/CL interface to force correct pairing of cognate light chains and Fc regions (modFab2-IgG).¹²²

In addition to these 2 + 2 Fab2-IgG migs, Fab fusion was adapted to generate 2 + 1 molecules by fusing one Fab to the heavy chain of an IgG with a heterodimerizing Fc (Fab-IgG-hetFc). Here, different arrangements are possible in these asymmetric molecules, for example, the first specificity on the two Fabs on the same heavy chain and the second specificity on the second heavy chain. Alternatively, the Fab-Fab arm can carry both specificities, with the second specificity either on the inner or outer Fab. As for the 2 + 2 Fab2-IgG fusion proteins, the light chain problem has been addressed by different approaches.

In the Fab-CrossMab format, the IgG building block is a CrossMab. Here, the Fd chain of the outer Fab is fused to the heavy chain of the CrossFab arm by a 10 residues long glycine-serine linker. The format is used in glofitamab, a TCE directed against CD20 and CD3, approved for the treatment of relapsed or refractory large B-cell lymphomas.¹²³ In the CrossFab of glofitamab, the VH and VL domains are swapped.¹²⁴ The same arrangement is used in three more 2 + 1 Fab-IgG-hetFc molecules, including various TCEs (e.g., forimtamig, alnuctamab) and one dual checkpoint inhibitor targeting PD-1 and TIM-3 (lomvastomig)¹²⁵ that has completed a Phase 2. These migs carry additional charge-charge mutations in the CH1-CL to further force correct pairing. Alternatively, the Fab-CrossMab format was used by swapping the CH1-CL domain in the CrossMab, which was applied to generate cibisatamab, a TCE targeting CEACAM5 and CD3.¹²⁶

Alternatively, cLC can be applied to generate 2 + 1 Fab-IgG. The approach is used in XmAb808, combining bivalent binding to B7-H3 with monovalent binding to CD28 for targeted costimulation. Here, the two B7-H3 binding sites are in the Fab-Fab arm, and Fc heterodimerization is achieved using the XmAb Fc platform further modified with Xtend for half-life extension. XmAb808 completed a Phase 1 in 2025.¹²⁷

TQB2825 is a CD20 × CD3-targeting TCE derived from the WuxiBody format, which generated a 2 + 1 Fab-IgG. Here, in the single anti-CD3 Fab the CH1 and CL domains are substituted by TCR Cα/Cβ (Fab-IgG-Cα/CβFab-hetFc).⁹⁰

The 2 + 1 Fab-IgG format was further modified to generate trispecific, trivalent molecules. Tilrekimig targets IL-13, IL-4, and TSLP and combines a bispecific IgG with kappa/lambda light chains with an outer CrossFab, i.e., with the VL-CL fused to the heavy chain and expressing the corresponding VH-CH1 as separate chain. Fc heterodimerization is achieved by charge-charge interactions introduced into the CH3 and upper hinge. Tilrekimig is currently in Phase 2 studies for treatment of moderate to severe asthma and atopic dermatitis. ISB 2001 is also a trispecific, trivalent Fab-IgG TCE targeting BCMA, CD38, and CD3 applying cLCs and the BEAT (Bispecific Engagement by Antibodies based on the T-cell receptor) platform for heavy chain heterodimerization.¹²⁸

Instead of fusing a Fab to an IgG, an scFv moiety can be used as building block to generate bispecific, trivalent 2 + 1 scFv-IgGs. In MBS303, a stability-enhanced anti-CD20 scFv is fused to the CD3 binding site of a bispecific CD20 × CD3 IgG with a kih hetFc.¹²⁹ Similarly, a trispecific, trivalent scFv-IgG was generated by fusing an scFv moiety to a bispecific IgG comprising a hetFc and using cLCs (BG-T187).¹³⁰

DVD-Igs use an approach where a second VH is fused to the N-terminus of the heavy chain, and a second VL to the N-terminus of the light chain. The linkers connecting the outer and inner variable domains have a length that allows their arrangement in a way that does not interfere with the inner binding site, i.e., both binding sites are accessible.¹³¹ Using a symmetric arrangement, bispecific, tetravalent molecules can be generated. This format is used in dilpacimab (ABT-165) targeting DLL4 and VEGF, which completed a Phase 2 study in 2022,¹³² and in lutikizumab (ABT-981) targeting IL-1α and IL-1β,¹³³ currently in a Phase 3 study to treat moderate to severe hidradenitis suppurativa. An 8 and 13 amino acid linker (ASTKGPEV, RTVAAPSVFIFPP) was used in dilpacimab to connect VH-VH and VL-VL, respectively, and 8-amino acid linkers (ASTKGPQV, QITRTVAAP) are used in lutikizumab, resembling sequences found in VH-CH1 and VL-CL connections. The same arrangement can be combined with a heterodimerizing IgG. CS-2009 (Phase 1) is such a trispecific DVD-Ig (DVD2-Ig). Here, two different outer Fv directed against PD-1 and CTLA-4, respectively, are fused to an anti-VEGF-A IgG with a heterodimerizing Fc.

Furthermore, sdAbs are used as building blocks fused to the N-terminus of an IgG. A sdAb has been fused to the N-terminus of an IgG heavy chain (sdAb2-IgG (HC)) or light chain (sdAb2-IgG (LC)). Four migs using the sdAb2-IgG (HC) and two migs using the sdAb-IgG (LC) format are currently in clinical trials. SIM0348 and SHR-2002 are sdAb2-IgG (HC) currently in Phase 1. Both migs target PVRIG and TIGIT for dual CPI.¹³⁴ BC008-1A, currently in Phase 1 trials, is directed against PD1 and TIGIT and derived from the so-called SMABody platform (single-domain antibodies fused to monoclonal antibody). However, we did not find information on the exact fusion site.

Hexavalent, trispecific sdAb2-/sdAb2-IgG were generated by using a monospecific IgG with a heterodimerizing Fc (IgG-hetFc) and fusing tandem arrangements of two different sdAb (sdAb-sdAb)2-IgG (HC) to the heavy chain N-terminus. This arrangement is used in GB268 (Phase 1), which has a symmetric 2 + 2 + 2 stoichiometry targeting with the tandem sdAbs PD-1 and CTLA-4 fused to an anti-VEGF IgG.¹³⁵

Extended IgGs with N- and C-terminal fusions

Currently, only one format combines fusion of antigen-binding moieties to the N- and C-terminus of an IgG. Fusion of a first scFv to the N-terminus of the heavy chain and two scFvs in tandem with the C-terminus of the heavy chain results in a tetraspecific, octavalent IgG fusion protein with a 2 + 2 + 2+ stoichiometry. This format is used in three different migs. Umizortamig (GNC-039, Phase 1) targets CD3 (first scFv), EGFvIII (IgG), PD-L1 and 4-1BB (second and third scFv), emfizatamab (GNC-038, Phase 2) targets CD3, CD19, PD-L1 and 4-1BB, and nebratamig (GNC-035, Phase 2) is directed against CD3, ROR1, PD-L1 and 4-1BB.^136–138 Thus, these guidance and navigation control (GNC) migs combine T-cell engagement with CPI and immune cell co-stimulation. Ten amino acid long glycine-serine linkers ((G₄S)₂) are used to connect the different scFvs.

Fc fusion proteins

A third diverse group of migs comprises fusion of various antigen-binding building blocks to homo- or heterodimerizing Fc regions. We identified 29 different formats used in 75 migs. These Fc fusion proteins cover bispecific and trispecific molecules of varying valencies. The majority uses Fab, scFv or sdAb as building blocks with antigen-binding activity. All have these building blocks fused to the N-terminus of the Fc chains, some in addition have building blocks fused to the C-terminus.

Bivalent bispecific Fc fusion proteins

The dominant format within this group is the bivalent, bispecific Fab-/scFv-hetFc, which is used in 30 different migs, six further are no longer in clinical development. Here, a Fab is fused to the first heterodimerizing Fc-chain and an scFv to the second heterodimerizing Fc-chain. Various hetFc are used in these migs, including kih, C-kih, the ZW kih, XmAb charge-charge, SEED (Strand-Exchange Engineered Domain), BEAT, and kih in combination with Fc*. A first Fab-/scFv-hetFc, zanidatamab (ZW25), was approved in 2024 for the treatment of previously treated unresectable or metastatic HER2-positive biliary tract cancer.¹³⁹ Zanidatamab is a biparatopic mig directed against domain 2 and 4 of HER2 with the binding sites derived from trastuzumab and pertuzumab. The scFv has the arrangement VL-VH with a 16 amino acid glycine/serine-rich linker ((GGSG)₄), connected by a 7 residues long linker (AAEPKSS) to the hinge of the Fc region. It uses the Asymmetric platform for Fc heterodimerization. Zanidatamab was described to induce a unique HER2 clustering and complement-mediated cytotoxicity, not observed for the parental antibodies, in addition to antibody-dependent cellular phagocytosis (ADCP) and antibody-dependent cellular cytotoxicity (ADCC).¹⁴⁰ Five further Fab-/scFv-hetFc migs are in Phase 3, including silevimig targeting two different surface antigens of rabies virus,¹⁴¹ pasritamig, a TCE targeting KLK2 and CD3,¹⁴² TQB2930 targeting HER2 domain 2 and 4,¹⁴³ TQB2102, an ADC derived from TQB2930,¹⁴⁴ and M701, a TCE targeting EpCAM and CD3.¹⁴⁵ Of note, the Fab-/scFv-hetFc format is used in three additional ADCs, ZW49 derived from zanidatamab, IMGN151, and M1231 (see also the section on bispecific ADCs below). Most of the Fab-/scFv-hetFc use a rather short linker of 5 to 8 residues to connect the scFv to the Fc (analyzing 11 molecules with GSRS entries), often comprising an EPKS sequence proceeding the upper hinge in the γ1 heavy chain flanked by 1–3 additional residues. The Fab-/scFv-hetFc format is furthermore used to generate TriNKETs, which bind to a TAA, NKG2D and co-engaging CD16 through their Fc-region, together synergizing in activating NK cells.¹⁴⁶

Instead of an scFv, a sdAb moiety is used to generate Fab-/sdAb-hetFc. This format is used in three migs currently in clinical development, one additional was recently terminated. The most advanced one is surovatamig, a CD19 × CD3 targeting TCE currently in Phase 3 in previously untreated follicular lymphoma patients.¹⁴⁷ Here, the Fab is directed against CD3 and derived from F2B¹⁴⁸ and the sdAb against CD19. A modification of the Fab-/sdAb-hetFc format is used in TNB-738. Here, the Fab arm is composed of a sdAb (as VH domain) completed by a dummy VL.¹⁴⁹

Two different scFv or two different sdAb can further be used as fusion partner, with the first moiety fused to a first Fc-chain and the second moiety fused to the second Fc-chain. The scFv-/scFv-hetFc format is used in TGI-6, a TCE targeting B7-H6 and CD3, and the sdAb-/scAb-hetFc format is used in PF-08046052 (SGN-EGFRd2, LAVA-1223), a TCE targeting EGFR and the Vδ2 chain of the γ9δ2-TCR,¹⁵⁰ both currently in Phase 1. Another sdAb-/scAb-hetFc (LAVA-1207) was recently terminated.

Fusing of one scFv and one sdAb to the N-termini of a hetFc results in the scFv-/sdAb-hetFc format. This format is used in CM-350, a TCE targeting GPC3 and CD3.¹⁵¹ Here, the scFv binds to CD3 and the sdAb to GPC3.

Blinatumomab is an Fc-less tandem-scFv (scFv2) approved for cancer therapy, applied due to a rather short half-life as infusion with a portable pump.¹⁵² Second-generation molecules have addressed this limitation by fusing the tandem-scFv to an Fc-region for half-life extension. To maintain the 1 + 1 stoichiometry of such molecules, the tandem scFv was fused to a single-chain version of the Fc-region (scFc). In this scFc, the first Fc-chain is connected to the second Fc chain by a 30 residue long glycine/serine-rich linker. This scFv2-scFc format was used to generate tarlatamab (AMG-757), a TCE targeting DLL3 and CD3, approved for the therapy of patients with extensive-stage SCLC.^153,154 Here, the tandem-scFv moiety (both scFvs in the VH-VL orientation) is linked to the scFv by a short GGGG linker to the scFc (GSRS 74X82ST8Q1). Of note, the first scFv directed against DLL3 is further stabilized by an interdomain disulfide bond introduced between VH-C44 and VL-C100. Population pharmacokinetics determined a terminal half-life of 11.2 h, which is significantly longer than that of tandem-scFvs.¹⁵⁵ It should be noted that several migs using this format have recently been terminated for strategic or business reasons, including vepsitamig (AMG-199), gresonitamab (AMG-910), emirodatamab (AMG-427), acapatamab (AMG-160), pavurutumab (AMG-701), emerfetamab (AMG-673), and AMG-562.

Fusion of two binding sites to one Fc-chain was furthermore applied to generate a Fab2-hetFc molecule, EMB-07, a TCE targeting ROR1 and CD3. In this MAT-Fab format, the light chain of the outer Fab is fused to the N-terminus of a first heavy chain, combined with a second Fc-chain comprising kih mutations and co-expression of a soluble Fd chain for the outer Fab and a light chain for the inner Fab.¹⁵⁶

A similar approach is used to generate Fab-scFv-hetFc molecules. Here, the Fd-chain of a Fab is fused to a scFv, which is then fused to a first heterodimerizing Fc-chain, coexpressed with a second Fc chain (Fab-scFv-hetFc). This format is used in QLS32015 targeting GPRC5D with its Fab arm and CD3 with its scFv.¹⁵⁷ QLS32015 is currently in Phase 3 in patients with relapsed or refractory multiple myeloma.

Like blinatumomab, a tandem-scFv, DART molecules are composed of the variable domains of two different antibodies, which assemble into a bivalent diabody-like structure further stabilized by heterodimerizing coiled-coiled peptides (see above). Similar in size than tandem-scFvs, DARTs also have a short half-life. Consequently, half-life extension strategies have been applied to the DART format. One approach was fusion of one of the DART chains to a first heterodimerizing Fc-chain, co-expressed with a second Fc-chain (DART-hetFc). This format was applied to several migs, three of them are currently in clinical development or have completed a Phase 1 (four others have been terminated). Three of these DART-hetFc are 1 + 1 TCEs used either for cancer therapy (MGD-024 targeting CD123 and CD3)¹⁵⁸ or depletion of HIV-infected cells (MGD-014, MGD-020 targeting gp120 or gp41, respectively).¹⁵⁹

The TCER format is an Fc fusion protein comprising a diabody-like structure composed of VH-Vβ and Vα-VL fused to a heterodimerizing Fc. The two chains assemble into a heterodimer with cognate cross-over pairing of VH/VL, directed against CD3, and Vα/Vβ, derived from a TCR recognizing a tumor-associated peptide in the context of MHC. The TCER format is used in IMA401 (Phase 1) directed against MAGE4/8 and in IMA402 (Phase 1/2) directed against PRAME.¹⁶⁰

Fc fusion was furthermore applied to extend the half-life of ImmTACs. Here, the TCR α-chain of ImmTAC is fused to a first Fc-knob chain and the second scFv-TCR β-chain is fused to a second Fc-hole chain, resulting in a scFv-TCR-hetFc molecule. This format was applied to generate IMC-P115C, a half-life extended ImmTAC targeting PRAME and CD3.

Bivalent Fc-fusion proteins further comprise ANKET molecules, which engage CD16 through an Fc region, combined with a Fab and an scFv targeting NKp46 and a TAA. The arrangement of the building blocks is Fab-Fc-Fab or Fab-Fc-scFv. An example of such an ANKET format is zemlikafusp alfa (IPH6501) which is a Fab-Fc-scFv targeting CD20 and NKp46 further fused with IL-2 v.¹⁶¹ The Fd and LC of the CD20-binding Fab are fused to the two chains of a hinge-comprising Fc, and Fd chain of the anti-NKp46 scFv is fused to the C-terminus of VH-CH1-hinge-CH2-CH3 chain. Zemlikafusp alfa is currently in a Phase 1/2 trial.

Finally, in the Fv-Fab-monoFc format the variable domains of a Fv fragment are fused to the N-termini of the two chains of a Fab, which is further fused to a monomeric Fc, resulting in a 1 + 1 mig. The two Fc chains of this H-BiTE format are modified in the CH3 domain to avoid Fc dimerization, thus the two Fc chains are present as individual, non-interacting chains. The format was used to generate SMET12, a mig targeting EGFR and CD3.¹⁶²

Trivalent, bispecific Fc fusion proteins

Several formats have extended the 1 + 1 Fc fusion proteins by adding a further binding moiety to generate trivalent bispecific migs with an asymmetric geometry and increased avidity for one of the target antigens. Thus, a tandem-sdAb was fused to a first hetFc chain and combined with a Fab-hetFc generating Fab-/sdAb2-hetFc with a 2 + 1 stoichiometry. The format is used in four different migs. The most advanced molecule is etentamig (ABBV-383, TNB-383B) directed against BCMA and CD3 and currently in Phase 3. Here, the two sdAb are directed against BCMA and the Fab against CD3.¹⁶³ A short G₄S linker is used to connect the two sdAb, which are directly fused to the upper hinge of a γ4 hetFc. Similarly, AZD5863 is a Fab-/sdAb2-hetFc TCE targeting CLDN18.2 and CD3.¹⁶⁴ Simridarlimab (IBI322) (Phase 2) targets with its two sdAb PD-L1 and with its Fab CD47.¹⁶⁵ The two sdAb are connected by a (G₄S)₄ linker. BGB-B3227 is a further sdAb2-/Fab-hetFc comprising one binding site for MUC1 (Fab) and two binding sites for CD16 (sdAbs) for NK-cell engagement¹⁶⁶

The Fab-/scFv-hetFc was extended to generate trivalent, bispecific molecules by fusing either a second Fab to the scFv moiety (Fab-scFv-/Fab-hetFc) or a second scFv to the Fab moiety (scFv-Fab-/scFv-hetFc). The Fab-scFv-/Fab-hetFc format is used in seven migs, all TCEs, benefiting from increased target cell binding due to avidity effects. The most advanced one is xaluritamig (AMG-509), a TCE targeting STEAP1 and CD3, which is in a Phase 3 study to treat participants with progressive metastatic castration-resistant prostate cancer.¹⁶⁷ Both, the Fd chain of the fused Fab and the scFv moiety are connected by a (G₄S)₂ linker and the CD3 binding site is encoded by the scFv. The scFv-Fab-/scFv-hetFc format is used in ZW171 (Phase 1), a TCE targeting mesothelin and CD3. Here the CD3-binding site is formed by the Fab.¹⁶⁸

Tetravalent bispecific Fc fusion proteins

Tetravalent bispecific fusion proteins are symmetric molecules with a homodimerizing Fc and fusion of two different binding sites to each Fc-chain. This includes DART moieties that were fused to an Fc (DART2-Fc). Two such molecules are currently in clinical development. Lorigerlimab (MGD-019) is dual checkpoint inhibitor targeting PD-1 and CTLA-4 and is currently in Phase 2.¹⁶⁹ Tebotelimab (MGD-013), is a dual checkpoint inhibitor targeting PD-1 and LAG-3 that completed a Phase 2/3 trial in combination with margetuximab, retifanlimab, and chemotherapy in HER2-positive gastric cancer/ gastroesophageal junction adenocarcinoma patients.¹⁷⁰ In these molecules, one chain of each DART moiety is fused to the upper hinge of γ4 Fc chains. Furthermore, the K-coil and E-coil at the C-terminus of each DART chain comprise 2 cysteine residues capable of covalently linking the two chains of one DART moiety.

Alternatively, tandem sdAbs were fused to the Fc region ((sdAb2)2-Fc). This format is utilized in erfonrilimab (KN046), a dual check-point inhibitor targeting PD-L1 and CTLA-4 currently in a Phase 3 study in subjects with advanced pancreatic ductal adenocarcinoma,¹⁷¹ and in lofacimig (KN060), a biparatopic mig directed against the apple 2 and 3 domains of Factor XI, which is in a Phase 2 study for the prevention of venous thromboembolism in patients undergoing total knee arthroplasty.¹⁷² In these molecules the two sdAb are connected by a short GAP linker.

Instead of an outer sdAb, a Fab was used to generate a 2 + 2 (Fab-sdAb)2-Fc molecules. The format is used in JS207 directed against PD-1 and VEGF-A currently in several Phase 2 studies.¹⁷³ Here, the Fd chain of the anti-PD-1 Fab is fused to the anti-VEGF-A sdAb, a nanobody (DotBody) linked by a flexible glycine-serine linker to the Fc. In a related format ((Fab-scFv)2-Fc), a Fab moiety is fused at the C-terminus of the Fd chain with an scFv, further fused to a homodimerizing Fc. This format is used by AZD0292, a half-life extended version of gremumamab (MEDI13902), directed against two Pseudomonas aeruginosa antigens for the treatment of bronchiectasis.¹⁷⁴

A Fab was further fused to an Fv moiety to generate 2 + 2 (Fab-Fv)2-Fc molecules. Here, the Fd-chain of the outer Fab is fused to the VL domain of the Fv, which is further fused to a Fc-chain. The LC of the Fab is fused to the VH domain. The format is used in Y101D (YM101) targeting PD-L1 and TGF-β, currently in Phase 2.¹⁷⁵

Two formats use N- and C-terminal fusion of binding modules to generate tetravalent, bispecific molecules. In the scFv2-Fc-scFv2 format, scFvs are fused to the N- and C-terminus of a homodimerizing Fc chain, resulting an homodimeric molecule in a 2 + 2 format with two scFv for the first specificity at the N-terminus and two scFv for the second specificity at the C-terminus. This format is used in mipletamig (APVO436), a TCE targeting CD123 and CD3 currently in Phase 1. The first scFv (VL-VH) is fused to the N-terminus of the upper hinge and the second scFv (VH-VL) connect by a S(G₄S)₃ linker to the C-terminus of the Fc-chain (substituting the C-terminal lysine by the first serine).¹⁷⁶ The same format is used in ALG.APV-527 (Phase 1/2) providing targeted costimulation by binding to 5T4 and 4-1BB.¹⁷⁷ Similarly, sdAbs were fused to the N- and C-terminus of an Fc fragment (sdAb2-Fc-sdAb2). The format is used in PM1003 (Phase 1/2) for combined checkpoint inhibitor and costimulation by targeting PD-L1 and 4-1BB.¹⁷⁸

Multivalent, trispecific Fc fusion proteins

A few Fc fusion proteins were developed into multivalent, trispecific formats. In one format, a bispecific tandem scFv is fused to a first Fc-chain, while the other Fc-chain of a hetFc carries a Fab (scFv2-/Fab-hetFc). This format is used in alveltamig (ZG006), a TCE that binds with the outer scFv and the Fab arm to two different epitopes of DLL3 and with its inner scFv to CD3. Alveltamig is in Phase 3 studies for treatment of patients with SCLC.¹⁷⁹

In a related format, one scFv and one Fab are fused to the N-termini of a hetFc and another scFv to the C-terminus of one of the hetFc chains (Fab-/scFv-hetFc-scFv). The format is used in JNJ-79635322 and JNJ-80948543, TCEs targeting CD3 and either GPRC5D and BCMA or CD20 and CD79b, respectively.^180,181 The C-terminal scFv is fused by a (G₄S)₄ linker to Fc-chain. Both compounds are in Phase 1.

AMG-305 (Phase 1) uses an extended form of the scFv2-scFc format by fusing another scFv2 to the C-terminus of the scFc moiety. Here, the first, N-terminal scFv2 binds CDH3 and CD3, the second, C-terminal scFv2 binds mesothelin and CD3, thus AMG-305 has a 1 + 2 + 1 stoichiometry. Through affinity tuning, AMG-305 can target double-positive tumor cells, while sparing single-positive cells.¹⁸²

Yet another format utilizes the CODV (cross-over dual variable) building block¹⁸³ Here, a first chain has the composition VHA-VHB-CH1 and the second chain has the composition VLB-VLA-CL. The domains are connected by different linkers allowing pairing of the cognate VH-VL pairs, further stabilized by CH1-CL. SAR-442257 combines a bispecific CODV building block with a Fab, both fused to the N-terminus of a hetFc (CODV-/Fab-hetFc). SAR-442257 is a TCE with costimulatory activity by targeting with its Fab CD38, and with its CODV CD3 and CD28, currently in Phase 1.¹⁸⁴

One format (sdAb3-Fc) has extended the sdAb-Fc fusion approach to generate a trispecific, hexavalent molecule by fusing a sdAb3 moiety directed against three different targets to a homodimerizing Fc. This format is used in SOA101, a TCE targeting HLA-G, CD3, and PD-L1, thus further including CPI.

TAVO412 is a trispecific tetravalent sdAb2-/Fab-hetFc-scFv comprising two sdAb for dual targeting of EGFR, a Fab-arm directed against c-MET (Fab arm) and an anti-VEGF scFv fused to the C-terminus of the sdAb2-comprising Fc chain.¹⁸⁵

Finally, remigromig (MK-3000, EYE-103) comprises a biparatopic diabody (Db) moiety at the N-terminus of a heterodimerizing Fc and two identical Fab moieties fused to the C-terminus, resulting in a trispecific, tetravalent mig. Remigromig binds with the Db moiety to two epitopes of the low-density lipoprotein receptor-related protein 5 (LRP5) and with the Fab arms to Frizzled-4 (Fz-4), thus acting as agonist of Wnt signaling pathway.¹⁸⁶ Remigromig is currently in Phase 2/3 for treatment of DME.

Conditionally active, Fc-comprising formats

Several approaches were applied to generate conditionally active Fc-comprising migs. Trevotamig (BA3182) is a conditionally active IgG-scFv2 (LC) TCE (Conditionally Active Biologic, CAB) targeting EpCAM and CD3. Here, the CD3-binding scFv exhibits a pH-dependent binding activity. Thus, the scFv shows little binding at the neutral pH of healthy tissue, but is capable of binding CD3 on T-cells in the acidic tumor microenvironment.¹¹⁰ Trevotamig is currently in a Phase 1 study in patients with advanced adenocarcinoma.

CX-904 is a conditionally active 2 + 2 TCE targeting EGFR and CD3 where an scFv is fused to the heavy chain N-terminus, thus placing both antigen-binding sites in close proximity. CX-904 carries further modifications to allow for conditional activation by fusing a first masking peptide to the N-terminus of the CD3-binding scFv and a second masking peptide to the N-terminus of the light chain, thus blocking the EGFR-binding site (Probody).¹⁸⁷ These 21 amino acid long masking peptides are fused by a glycine-serine linker and further comprise an 8 amino acid protease cleavage site. Recently the clinical study was, however, discontinued due to a sponsor decision.

In another approach, the Fab2-IgG format was adapted to generate conditionally active molecules. In the LockBody format the two light chains are connected by a 11 residues long linker (GPAPLGLGGGS) and the Fd fragment of the outer Fab is connected by an IgG-derived 11 residue long linker (GPAPELLGGGS) to the heavy chain, which introduces a steric blockade of the inner binding site by the outer Fab. This lock is opened by proteolytic cleavage of the heavy chain linker, for example, by MMPs and cathepsin, making the inner binding site accessible. The format is utilized in itanistomig (LB101) targeting PD-L1 (outer Fab) to drive tumor enrichment, and CD47 (inner Fab) being conditionally active at the tumor site.¹⁸⁸ Itanistomig was in a Phase 1/2 trial, which was recently terminated due to a business decision.

Fc-comprising migs no longer in clinical development

Fourteen Fc-comprising formats were identified which are no longer in clinical development (Figure 3). PF-06671008 is a diabody-Fc fusion protein (LP-DART-hetFc).¹⁸⁹ PF-06671008 was terminated due to limited anti-tumor activity. The same format was used in PF-07062119 targeting GUCY2c and CD, which was terminated due to strategic considerations.¹⁹⁰ Duligotuzumab, is a symmetric bispecific IgG comprising two-in-one Fab arms (2in1 IgG). These two-in-one Fab arms bind competitively to targets as shown for an EGFR x HER3 antibody (MEHD7945A, duligotuzumab).¹⁹¹ Duligotuzumab completed Phase 2 studies, but is no longer pursued due to lack of efficiency.¹⁹² ABBV-184 is a 1 + 1 TCE comprising a soluble scTCR (Va-Vb) fused to an anti-CD3 Fab that is further fused to one chain of a hetFc.¹⁹³ Here, the scTCR recognized a peptide derived from the oncogene survivin in the context of HLA presentation. Clinical development of ABBV-184 has been terminated. The CODV building block was used to generate a bivalent, bispecific CODV-hetFc fusion protein targeting NKp46 and BCMA (SAR445514). Here, the VHA-VHB-CH1 chain is fused to the hinge region of a heterodimerizing Fc-chain, co-expressed with a non-fused Fc-chain and the VLB-VLA-CL chain. The Fc-part is capable of further engager NK-cells through CD16 binding with its ADCC-enhanced Fc (ADE: G236A, S239D, I332E), thus was assigned as being trifunctional.¹⁹⁴ SAR-440234 is a CODV-Fc fusion protein, where one chain of the CODV moiety is fused to the first Fc-chain and the other chain of the CODV is fused to the second Fc-chain, thus forming a bivalent, bispecific CODV-Fc molecule.¹⁹⁵ SAR-440234 is a TCE targeting CD123 and CD3, whose clinical development was terminated. RO6874813 is a 2 + 2 IgG-CrossFab2 directed against FAP and DR5 developed for targeted activation of death receptor 5 (TRAILR2) by binding to fibroblast activation protein (FAP) in the tumor microenvironment.¹⁹⁶ RO6874813 completed a Phase 1 study in 2017. ATOR1015 is a tetravalent bispecific IgG-sdAb2 fusion protein with the sdAb moieties fused to the C-terminus of the IgG light chain. ATOR1015 is directed against CTLA-4 and OX40, and thus combines CPI with costimulation.¹⁹⁷ ATOR1015 completed a Phase 1 in 20,121. MEDI4276 is a biparatopic HER2-binding molecule generated by fusing an scFv derived from trastuzumab to the N-terminus of the heavy chain of an HER2 domain 2-binding antibody.¹⁹⁸ MEDI1476 has completed a Phase 1 study in 2018. GB-263T was in a Phase 1/2 (unknown status) and is composed of two different sdAb fused to the N-terminus of a monospecific IgG with a het Fc. It targets with its sdAbs two different epitopes of c-Met fused to an anti-EGFR IgG. Thus, this format has an asymmetric 1 + 1 + 2 stoichiometry (or 2 + 2 regarding antigens).¹⁹⁹ Ivicentamab is a biparatopic DuoHexabody targeting two epitopes of CD37. The format comprises six bispecific Duobody building blocks with a hexadimerization-enhancing mutation (E430G) in the CH3 domains, together with F405L and K409R for controlled Fab-arm exchange.²⁰⁰ Clinical development was terminated due to strategic evaluation. An IgM-scFv format was used to generate TCEs with 10 binding sites for a target antigen with monovalent binding of CD3. Here, an anti-CD3 scFv is fused to the J-chain, which is integrated into a pentameric IgM molecule with 10 binding sites for the target antigen and one binding site for CD3. The format was used to generate imvotamab (IGM-2323) targeting CD20²⁰¹ and CD3 and IGM-2644 targeting CD38 and CD3. Clinical development of both migs was terminated.

Based on the available data, it is difficult to define generalizable format- or binder-associated determinants of failure. Most publicly disclosed terminations provide only high-level rationales, such as strategic reprioritization or insufficient clinical benefit, rather than mechanistic insight into liabilities related to molecular architecture, valency, geometry, or epitope selection. As more multispecific formats advance clinically and as post-hoc analyses become available, these datasets will be essential for identifying recurrent design pitfalls and establishing evidence-based guidelines for format selection and optimization.

Fc heterodimerization

Of 284 Fc-comprising migs, 65% utilize a heterodimeric Fc region required to generate asymmetric molecules.⁸⁰ Twenty different strategies were identified by sequence analysis, a few more could not be assigned (Table 1).

The most frequent Fc heterodimerization strategy is the C-kih (disulfide-stabilized knobs-into-holes) used in 48 migs, followed by kih used in 33 migs, 8 further migs combine kih or C-kih with Fc* modifications (see below). Thus, 55% of migs with a heterodimeric Fc comprise kih and its disulfide-stabilized version C-kih.

The majority of the kih and C-kih modification was applied to γ1 Fc chains (87%), but also to a few γ4 Fc chains. In both cases, a tryptophan is introduced into a first CH3 domain (T366W) and three positions in the second CH3 are mutated to amino acids with smaller side chains (T366S, L368A, Y407V) (kih) further stabilized by introducing an artificial disulfide bond between position 354 and 349 of the first and second CH3 domain (S354C, Y349C) (C-kih).⁶⁵ One approved mig (mosunetuzumab) uses the kih approach and post-production assembly of the two heavy chains, two approved migs (faricimab, glofitamab) use the C-kih approach and co-expression of both heavy chains, together with the two light chains in one cell (both combined with the CrossMab technology to solve the light chain problem).

The kih platform was further improved by mutating the bulky F205 of one CH3 domain to a lysine, and changing the charged K409 to an alanine in the other CH3 domain (kih-KA), which favors the formation of heterodimers and further preventing homodimer formation.²⁰² The technology was applied to generate anbenitamab (KN026) and its ADC derivative (JSKN003), a 1 + 1 IgG1 targeting domain 2 and 4 of HER2 by combining the binding sites of trastuzumab and pertuzumab, both currently in Phase 3 studies.^203,204

A kih-like approach (kih-ETYY) was applied to generate GB261. Here, a kih was generated by introducing T366Y in one CH3 and Y407T in the other CH3, further combined with an engineered ionic bond and hydrophobic interaction (Q347E, S354Y). Of note, the isoelectric point (pI) of GB261 was differentiated by introducing mutations into the framework of the VH domains.²⁰⁵

The Fc* technology⁸¹ is used in 12 IgG4 molecules all comprising a cLC. Here, two mutations are introduced into one of the CH3 domains (H435R, Y436F), which affects binding to protein A. Hence, heterodimers bind with intermediate pH sensitivity to protein A, allowing a differential elution and separation of homo- and heterodimers. The Fc* modification is used in odronextamab, a 1 + 1 TCE targeting CD20 and CD3.²⁰⁶ However, these mutations alone do not force heavy chain heterodimerization. Therefore, the Fc* approach was further combined with kih or C-kih mutations to facilitate assembly of heterodimers and further separation of homodimeric impurities. Six γ1 migs use Fc* together with kih. This includes obrixtamig (BI-764532), a 1 + 1 scFab-Fc targeting DLL3 and CD3, currently in Phase 2 to treat patients with SCLC and neuroendocrine cancers.^207,208

The XmAb platform combines engineered pI differences in the Fc region to allow for selective purification of heterodimeric species with Fc substitutions that use electrostatic steering to force Fc-heterodimerization.²⁰⁹ Here, a serine is mutated to a lysine in the first CH3 domain, and five residues in the second CH3 are mutated to negatively charged residues and one to serine (Table 1). The XmAb modifications are used in 11 migs, three of them in Phase 2 (plamotamab, vibecotamab, vudalimab). All three Phase 2 XmAbs are 1 + 1 Fab-/scFv-hetFc molecules, plamotamab targeting CD20 and CD3, vibecotamab targeting CD123 and CD3, and vudalimab targeting OD-1 and CTLA-4.^210–212

Like the kih technology, the ZW1 platform uses mutations in the CH3 domain to generate a complementary interface.²¹³ Here, four mutations are introduced in each CH3 domain (Table 1). We found 6 migs using this technology, five of them being Fab-/scFv-hetFc molecules. One, zanidatamab is a biparatopic Fab-/scFv-hetFc directed against domain 2 and 4 of HER2 approved for treatment of HER2-positive biliary tract cancer. Another one, pasritamig (JNJ-78278343) is a TCE targeting KLK2 and CD3, currently in Phase 3 to treat patients with metastatic castration-resistant prostate cancer.¹⁴²

Electrostatic interactions are used in various migs to steer Fc-heterodimerizations. This includes DEKK. Here, negatively charged residues are introduced in one CH3 (L351D, L368E) and positively charged residues in the other CH3 (L351K, T366K) at opposing positions.²¹⁴ DEKK is used in seven migs, including zenocutuzumab, a bispecific IgG-hetFc cLC, targeting HER2 and HER3, approved for the treatment of advanced, unresectable or metastatic pancreatic adenocarcinoma or NSCLC harboring an NRG1 gene fusion.²¹⁵ It uses a “dock & block” approach by guided targeting of HER2-expressing tumor cells and blocking of adjacent HER3.²¹⁶ Other electrostatic steering variants are KE/KKR, EE/KR, EEE/RRRR (Table 1). KE/KKR is used in emicizumab, a bispecific IgG1 directed against FIXa and FX approved for the treatment of hemophilia A.²¹⁷ EE/KR combines a charge pair in the upper hinge with a charge pair in CH3 and is used in tilrekimig (PF-0725215), a 2 + 1 Fab-IgG CrossMab directed against IL-4, IL-13, and TSLP currently in Phase 2 to treat moderate-to-severe asthma. Similarly, elranatamab is an approved 1 + 1 IgG2 TCE targeting BCMA and CD3 carrying charge-charge pairs in the hinge and CH3 domain (EEE/RRRR).

The RCVT/CEW approach involves a combination of kih and electrostatic interactions. Here, charged residues at the rim of hydrophobic core of the CH3 domains are substituted by larger or smaller side chains of hydrophobic residues together with a substitution of weakly involved interactions with charged, long side chain residues (Q347R/D399V/F405T – K360E/K409W).²¹⁸ These RVT/EW modifications were extended by introducing a disulfide bond at the interface (Y349C – S354C). These mutations are used in three migs, including azerutamig (DF1001), a Fab-/scFv-Fc TriNKET molecule targeting HER2 and NKG2D, currently in Phase 1/2 studies in patients with advanced solid tumors.²¹⁹

A combination of hydrophobic and electrostatic mutations is furthermore utilized by ELV/LR, carrying three mutations in the first CH3 domain (L351E, Y407L, K409V) and two mutations in the second CH3 domain (T366L, D399R). This approach is used, for example, in fidasimtamab (IBI315), a bispecific IgG1 directed against HER2 and PD-1 for targeted CPI.²²⁰

Heterodimerization of Fc parts can further be achieved by the BEAT technology. Here, residues from the TCR Cα and Cβ interface are grafted into the CH3-CH3 interface.²²¹ We identified two migs using this technology. ISB 1342 is a bispecific Fab-/scFv-hetFc TCE with specificity for CD38 and CD3²²² and ISB 2001 is a trispecific Fab-IgG-hetFc cLC TCE directed against BCMA, CD38 and CD3 currently in Phase 2.¹²⁸

A related approach is the SEED technology. Here, the CH3 domains comprise alternating regions derived from IgG and IgA that create a complementary interface.²²³ This approach was applied to generate M1231, an ADC targeting EGFR and c-MET, which recently completed a Phase 1 study.²²⁴

An entirely different approach is used by the cFAE, described already above, which was applied to generate bispecific IgG1 and IgG4 antibodies (Duobodies). We identified seven IgG1 Duobodies and six IgG4 Duobodies currently in clinical development, two of them in Phase 3 (acasunlimab, denecimig), and four of them being approved: teclistamab, an IgG4 antibody directed against BCMA and CD3 approved to treat multiple myeloma²²⁵; epcoritamab, an IgG1 TCE targeting CD20 and CD3 approved to treat relapsed/refractory diffuse large B-cell lymphoma (DLBCL) and high-grade B-cell lymphoma (HGBCL); talquetamab, an IgG4 TCE targeting GPRC5D and CD3 for the treatment of relapsed/refractory multiple myeloma; and amivantamab, an IgG1 DuoBody targeting EGFR and c-MET approved for the treatment of NSCLC with exon 20 insertion mutations.²²⁶

Fc silencing

Fc-comprising antibodies are capable of exerting Fc-mediated effector functions such as ADCC, ADCP, and complement-dependent cytolysis (CDC), which can be beneficial for therapeutic interventions, for example, by killing cancer cells.²²⁷ However, engagement of immune cells and complement by antibodies can lead to an inappropriate immune system activation and cytokine release, and thus unwanted side effects²²⁸ This was especially recognized for TCEs and it was postulated that TCEs should not contain Fc portions.⁴ Consequently, TCEs that lack an Fc region, such as blinatumomab, were developed. However, the presence of an Fc region can be beneficial by, for example, mediating a prolonged serum half-life, increasing stability, and facilitating manufacturing. This has led to the development of antibodies with reduced or abolished effector functions.²²⁹

Of the 284 Fc-comprising migs we identified, 178 implement an approach to silence Fc effector functions (the term “silenced” is used in its widest sense, including Fc with reduced effector functions) (Figure 1E, Table 2). This includes the use of unmodified IgG4 (24), as well as 164 engineered IgG1, IgG2, and IgG4 derivatives. Of these engineered migs, 126 migs have an IgG1 background, 2 an IgG2 and 26 an IgG4 backbone. The majority (71%) of the Fc-silenced migs comprise an engineered γ1 heavy chain. In total, we identified 30 different silencing approaches (Table 2). For a few migs, the isotype or exact modifications are unknown. Half of the silenced Fc-comprising migs are TCEs, with the majority targeting CD3 on T cells as trigger molecule, approximately 41% are involved in modulation immune reactions, for example, by CPI or by blocking or providing immuno-stimulatory signals, 3 silenced migs are ADCs.

Table 2.

Silenced Fc variants.

iso- type	name	mutations	#	example
IgG1	LALA	L234A, L235A	37	Ivonescimab
IgG1	LALA-PG	L234A, L235A, P329G	11	glofitamab
IgG1	LALA-GA	L234A, L235A, G237A	7	cadonilimab
IgG1	LALA-PA	L234A, L235A, P329A	4	BA-1202
IgG1	LALA-DS	L234A, L235A, D265S	8	pasritamig
IgG1	LALA-GA-KA	L234A, L235A, G237A, K322A	4	danvilostomig
IgG1	LALA-aglyco	L234A, L235A, N297A	1	MBS303
IgG1	aglyco	N297G/A/Q	13	mosunetuzumab
IgG1	aglyco-RC-VC	N297G, R292C, V302C	2	inezetamab
IgG1	aglyco-KA	N297A, K322A	1	AP203
IgG1	PVAΔG-SK	E233P, L234V, L235A, ΔG236, S267K	10	vibecotamab
IgG1	PAAΔG-PA	E233P, L234A, L235A, ΔG236, P329A	2	BGB-B455
IgG1	PVAΔG-DG-AQ-AS	E233P, L234V, L235A, ΔG236, D265G, A327Q, A330S	2	CC-1
IgG1	PVAΔG-DG-NQ-AQ-AS	E233P, L234V, L235A, ΔG236, D265G, N297Q, A327Q, A330S	1	CLN-049
IgG1	PVAΔG-PS aglyco	E233P, L234V, L235A, ΔG236, N297Q, P331S	2	IMA-401
IgG1	PVAΔG-SK aglyco-RC-VC	E233P, L234V, L235A, ΔG236, S267K, N297G, R292C, V302C	1	xaluritamig
IgG1	LFLE	L234F, L235E	4	SGN-EGFRd2
IgG1	LFLE-DA	L234F, L235E, D265A	6	acasunlimab
IgG1	LFLE-PS	L234F, L235E, P331S	5	rilvegostomig
IgG1	LALE-GA-AS-PS	L234A, L235E, G237A, A330S, P331S	2	MTX101
IgG1	LALQ-KQ (AQQ)	L234A, L235Q, K322Q	1	CDX-585
IgG1	LG-GR	L235G, G236R	2	besufetamig
IgG1	LE-PG	L235E, P329G	1	CT-95
IgG1	γ2CH2-DA	γ2 CH2, D270A	1	YM101
IgG2	IgG2 DA-AS-PS	D270A, A330P, S331S	1	elranatamab
IgG4	IgG4	–	23	emicizumab
IgG4	FALA	F234A, L235A	12	teclistamab
IgG4	PVAΔG	E233P, F234A, L235A, ΔG236	10	odronextamab
IgG4	LR-SK-aglyco	L235R, S239K, N297A	1	ERY974

The LALA mutations (L234A, L235A) within the lower hinge region of IgG1 can be considered the prototype silencing strategy.²³⁰ We identified the LALA mutations in 33 of the migs from category 1, 35 additional migs carry LALA derivatives comprising further mutations. LALA is, for example, used in ivonescimab, an IgG-scFv2 directed against PD-1 and VEGF-A approved to treat patients with EGFR-mutated locally advanced or metastatic non-squamous NSCLC who have progressed after TKI therapy.¹⁰⁸ Furthermore, 3 migs with LALA mutations are in Phase 3, pumitamig (BNT327; PM8002), an IgG-sdAb2 fusion protein targeting PD-L1 and VEGF-A,¹¹¹ lutikizumab (ABT-981), a DVD-Ig directed against IL-1α and IL-1β to treat adult and adolescent participants with moderate to severe hidradenitis suppurativa,²³¹ and QLS32015 targeting GPRC5D and CD3.¹⁵⁷

Detailed analysis of Fc-silenced antibodies demonstrated that the LALA mutations are not able to completely abolish effector functions, especially regarding FcγRI-mediated activities.^229,232 This led to the development of LALA derivatives. LALA-PG is a derivative further carrying a P329G mutations, which was described to exhibit completely abolished immune effector functions.²³² LALA-PG is used in 11 migs, the most advanced one being glofitamab, a TCE targeting CD20 and CD3.¹²³ Alternatively, position 329 was mutated to alanine (LALA-PA), which is used in 4 migs, most of them in early-stage development. LALA-GA, with an additional G237A mutation close to the LALA mutations, is used in 7 migs, including cadonilimab (AK104), an approved IgG-scFv2 fusion protein targeting PD-1 and CTLA-4.²³³ LALA-GA was further combined with K322A (LALA-GA-KA), a mutation originally described by Hezareh and coworkers.²³⁰ This LALA variant is used in 4 migs, the most advanced of them in Phase 2 (danvilostomig, emfizatamab, nebratamig). LALA-DS combines LALA with a D265S mutation used in 8 different migs, with pasritamig (JNJ-78278343) directed against KLK2 and CD3 in Phase 3.¹⁴² Finally, LALA was combined with a N297A mutation that results in an aglycosylated LALA variant (LALA-aglyco), which is used, to our knowledge, in MBS303, a CD20 × CD3 targeting TCE.¹²⁹

An aglycosylated CH2 domain (aglyco) without further mutations is present in 13 migs. Here, either N297 is mutated to glutamate, or T299 to an alanine or a glycine, which are part of the N-glycosylation sequon -NST-. Mosunetuzumab, a CD20 × CD3 targeting 1 + 1 TCE, is such an aglycosylated variant with an N297G mutation.²³⁴ However, it was noted that aglycosylated IgGs exhibit reduced thermal stability.²³⁵ This was addressed by further stabilizing the CH2 domain by introducing an additional disulfide bond (R292C, V302C), resulting in aglyco-RC-VC and aglycosylated IgG molecules with similar stability than the parental antibody.²³⁶ This variant is used in 2 migs, including tarlatamab, a sdAb2-scFc fusion protein directed against DLL3 and CD3 and approved for the treatment of patients with extensive-stage SCLC.¹⁵⁴ Another aglycosylated variant further comprises a K322A mutation is used in AP203, an IgG-scFv2 targeting PD-L1 and 4-1BB currently in Phase 2.²³⁷

Instead of the LALA motif, part of the lower hinge from IgG2 was used to silence IgG1 and IgG4 antibodies. For IgG1 the following mutations are introduced (E233P, L234V, L235A, ΔG236 - PVAΔG), thus the glycine at position 236 is deleted. This motif was further combined with additional mutations in the CH2 domain. In PVAΔG-SK the motif is combined with S267K used in 10 migs. Examples include two Phase 2 migs: vudalimab (XmAb207171) targeting PD-1 and CTLA-4 and vibecotamab (XmAb14045), a TCE targeting CD123 and CD3.²¹¹ Alternatively, PVAΔG was combined with 3 mutations (D265G, A327Q, A330S) to generate PVAΔG-DG-AQ-AS. This silencing strategy is used, for example, in CC-1, an IgG-scFv2 TCE targeting PSMA and CD3.¹⁰⁶ A related variant, PVAΔG-DG-NG-AQ-AS with an additional N297G mutation is used in CLN-049, an IgG-scFv2 TCE targeting FLT3 and CD3.^238,239 Xaluritamig (AMG-509), a 2 + 1 Fab-/Fab-scFv-hetFc TCE targeting STEAP1 and CD3, combines PVAΔG with S267K and N297G, which is further stabilized by the disulfide bond-forming R292C/V302C mutations (PVAΔG-SK-aglyco-RC-VC).¹⁶⁷ Based on patent information, IMA401 and IMA402, TCER molecules targeting PRAME or MAGEA4/8, respectively, comprise PVAΔG-PS-aglyco, with N297Q and P331S mutations. Finally, PAAΔG-PA carries a L234A instead of L234V and the P329A mutation used also in LALA-PA. Based on patent information, this is used in BGB-B455 and BGB-B2033.

Another set of mutations converts L234 and L235 to phenylalanine and glutamate (LFLE). This variant is used in 4 migs, including PF-08046052 (SGN-EGFRd2), a biparatopic TCE targeting DLL3 and CD3.²⁴⁰ A derivative thereof, LFLE-PS carries in addition a P331S mutation used in 4 migs. This includes two migs in Phase 3, rilvegostomig (AZD2936) for dual blockade of PD-1 and TIGIT,²⁴¹ and volrustomig (MEDI5752), a DuetMab for dual CPI by targeting PD-1 and CTLA-4.²⁴² A recent study showed that LFLE-PS still shows substantial binding to various Fcγ receptors leading to ADCP and ADCC.²²⁹ This was improved by a further variant, LFLE-DA with an additional D265A mutation, although this variant still showed some complement activation.²²⁹ The LFLE-DA variant is used in 4 migs, including epcoritamab, an IgG1 Duobody directed against CD20 and CD3 approved for the treatment of B-NHL (Frampton, 2023)²⁴³ and acasunlimab (GEN1046, BNT311), an IgG1 Duobody for dual CPI by targeting PD-L1 and CTLA-4 currently in Phase 3.²⁴⁴

Two further silencing variants, LALE-GA-AS-PS and LALQ-KQ, carry the L234A mutation but use alternative mutations for L235 (either LE or LQ), combined with additional mutations. LALQ-KQ is used in CDX585, an IgG-scFv fusion protein directed against PD-1 and ILT4, which has completed a Phase 1.²⁴⁵ The LG-GR variant combines L235G with G236R, thus focus as LALA on the lower hinge, but introducing a bulky and charged residue at position 236. This variant is used in 2 migs, including besufetamig (ONO-4685), a IgG1-hetFc with a cLC directed against PD-1 and CD3 currently in Phase 1 to treat patients with T-cell lymphoma.²⁴⁶ LE-PG combines L235E with P329G for silencing. To our knowledge it is used in CT-95, a TCE targeting mesothelin and CD3.²⁴⁷

Furthermore, mutations were applied to further silence IgG2 antibodies. Elranatamab is an approved IgG2 TCE carrying D270A, A330P, and P331S in its CH2 domain and targeting BCMA and CD3 for the treatment of relapsed or refractory multiple myeloma.⁸⁴

Instead of mutating individual residues, entire domains can be exchanged between antibody isotypes. Thus, the CH2 domain of an IgG1 antibody was substituted by that of IgG2 to generate Y101D (YM101) a Fab-Fv-Fc Checkbody targeting PD-L1 and TGF-β, further comprising a D270A mutation in the CH2.¹⁷⁵

Twenty-three migs have an unmodified γ4 heavy chain or Fc part and thus have reduced Fc-mediated effector functions. Emicizumab is an approved IgG4-hetFc cLC antibody designed to mimic the activity of coagulation Factor VIII by binding to FIXa and FX.²¹⁷

IgG4 was further silenced by introducing mutations into the CH2 domain. Like the LALA mutations introduced into the γ1 lower hinge, the same positions were mutated in γ4 (F234A, L235A; FALA). We identified 12 migs carrying these mutations. This includes two approved TCE, teclistamab targeting BCMA and CD3 and talquetamab targeting GPRC5D and CD3. Both antibodies are IgG4 Duobodies approved for the treatment of relapsed or refractory multiple myeloma.^248,249 In another IgG4 antibody, ERY974, three mutations were introduced (L235R, S239K, N297A) resulting in a mutated, aglycosylated heavy chain.²⁵⁰

Fc with enhanced effector functions

Strategies to enhance Fc-mediated effector functions, such as ADCC, ADCP or CDC, are utilized by a few migs, most of them designed for dual or biparatopic receptor blockade and target cell killing. Several strategies have been established in the past, including protein engineering by introducing mutations affecting binding to Fc receptors and complement, and glycol-engineering, affecting the composition of the N-glycans in the Fc region.^251,252 Thus, it was shown that afucosylated IgGs exhibit increased ADCC.²⁵³

Amivantamab, an approved IgG1 Duobody targeting EGFR and c-Met, is produced in a low fucose production cell line. Thus, amivantamab preparations include less than 10% of antibodies containing a core fucose attached to the N-glycan at N297 of the CH2 domain, resulting in increased FcγRIII binding and enhanced ADCC.²⁵⁴ Migs completely lacking fucosylation were further obtained by production in cell lines deficient in fucosylation. MBS301, KM257 and its ADC derivative KM501, are glyco-engineered, biparatopic IgG-hetFc cLC molecules binding to domains 2 and 4 of HER2. MBS301 is generated using dual-cell expression and in vitro post-assembly after protein A purification. The half antibodies comprising either a hole or a knob were derived from trastuzumab and pertuzumab and are produced separately in glycol-engineered CHOK1-AF cells in which the GDP-fucose transporter (GFT) gene SLC35c1 was removed.²⁵⁵ MBS2301 has a 2.7 to 7-fold increased affinity for FcγRIIIa, resulting in enhanced ADCC. MBS201 is currently in Phase 1 to treat HER2⁺ recurrent and metastatic malignant solid tumors. Pamvatamig (MCLA-129) is an EGFR x c-MET targeting IgG-hetFc-cLC currently in Phase 2 expressed using the GlymaxX technology resulting in afucosylated antibodies.²⁵⁶

We identified further migs with enhanced effector functions, BMS-986442 (AGEN1777) directed against CD96 and TIGIT, and GB-263T, a tetravalent, trispecific sdAb-/sdAb-IgG fusion protein targeting EGFR and c-MET.²⁵⁷ However, we were not able to identify the exact approach.

Mutations affecting half-life of Fc-comprising migs

Immunoglobulins of the IgG subclasses IgG1, IgG2, and IgG4 have an exceptionally long half-life due to FcRn-mediated recycling. Various studies have shown that the half-life can be further extended by introducing mutations into the Fc region affecting binding to FcRn without interfering with recycling.²⁵⁸ Several of these mutations were also applied to extend the half-life of Fc-comprising migs.

The YTE approach, which comprises three mutations (M252Y, S254T, T256E) in the CH2 domain, increases binding to FcRn at pH 6 while maintaining release at neutral pH.²⁵⁹ Using this approach, Dall’Acqua et al. showed that the terminal half-life of an anti-RSV IgG (MEDI-524) was extended in cynomolgus monkeys from approximately 6 h for the parental antibody to approx. 21 h for the YTE variant. We identified four migs, two IgG1 and two IgG4 utilizing YTE for half-life extension. CDX585, an IgG1-scFv2 targeting PD1 and ILT4, EMB-09, a γ1 FIT-Ig targeting PD-L1 and OX40, and two 2 + 2 DART-Fc (MGD-013, MGD-019) having a γ4 Fc region.

The Xtend modifications (M428L + N434S; ML-NS) are found in 4 different migs. These mutations were able to extend the terminal half-life of cetuximab in hFcRn-bearing mice approximately five-fold.²⁶⁰ ML-NS-comprising migs include vudalimab (XmAb20717), a Fab-/scFv-hetFc directed against PD-1 and CTLA-4,²¹² and tilrekimig (PF-07275315), a trispecific Fab-IgG-hetFc CrossMab targeting IL-4, IL-13 and TSLP, both in Phase 2 studies. Of note, a single mutation derived from ML-NS (M428L; ML) is most likely used in KM257, a biparatopic IgG-hetFc cLC targeting domain 2 and 4 of HER2.

NXT007 is a FVIII-mimetic bispecific antibody targeting FIXa and FX.²⁶¹ It is described as a half-life extended IgG-hetFc with non-common LCs. Although the exact modifications are not disclosed, it might include combinations of N434A, M428L, Y436T, Q438R, and S440E (ATC series).^262,263

The triple mutation I253A/H310A/H435A is used in faricimab, which targets Ang-2 and VEGF-A and is approved for the treatment of nAMD or DME.⁹³ These mutations were described to reduce binding to FcRn at both neutral and acidic pH.²⁶⁴ Studies in mice indicate that these mutations result in reduced subcutaneous bioavailability. Furthermore, the H435A mutation was shown to result in reduced phagocytosis by polymorphonuclear neutrophils.²⁶⁵

T-cell engagers

TCEs represent the major class of migs, with 139 molecules (38% of all migs) approved or in clinical development (Suppl. Table S1). For solid tumors, 73 migs were identified, 64 migs are used or developed to treat hematologic malignancies. Two additional ones are used in non-cancer indications, in addition to 14 TCEs, including 4 approved ones, studied in cancer and non-cancer indications, mainly to treat autoimmune diseases or light chain amyloidosis. Twelve TCEs are already approved, 12 further TCEs are in Phase 3 (Tables 3, 4; Figures 4, 5). The majority (75%) of TCEs are bispecific, 14% are trispecific and 4 TCEs are tetraspecific. For 12 TCEs we could not identify the format or valency. In total, 48 different formats were used to generate TCEs, 36 comprising an Fc-part, 12 being Fc-less (see suppl. Table S1 for details). Approximately half of the TCEs have a 1 + 1 stoichiometry, followed by 2 + 1 (14%), 1 + 1 + 1 (14%) and 2 + 2 (9%) valencies. A few are trispecific with a 1 + 2 + 1 or 2 + 2 + 2 stoichiometry, or tetraspecific with a 2 + 2 + 2 + 2 or 1 + 1 + 1 + 1 stoichiometry (Figure 6). Of the 17 trispecific migs, 8 comprise an HSA-binding site for half-life extension, 5 have a binding site to mediate costimulatory signals through binding to 4-1BB or CD28, respectively, and the remaining target two different TAAs. Three of the identified tetraspecific TCE target a TAA, CD3, 4–1-BB as costimulatory signal and PD-L1 for CPI. All three are scFv-IgG-scFv-scFv fusion proteins. A further tetraspecific 1 + 1 + 1 + 1 TCE targets two TAAs (TROP2, c-MET), CD3 and CD28 for costimulation. Most likely, this TCE (MDX2001) uses a CODV2-hetFc format.²⁶⁶

Table 4.

Late-stage (phase 2/3 or 3) multispecific immunoglobulins.

INN/Brand name	Targets	format	MoA	Indications	company
Xaluritamig AMG-509	STEAP1 × CD3	Fab-/Fab-scFv-hetFcγ1 (2 + 1)	TCE	PC	Amgen
Brenetafusp IMC-F106C	PRAME x CD3	scFv-TCR (1 + 1)	TCE	advanced melanoma	Immunocore
Surovatamig AZD0486	CD19 × CD3	Fab-/sdAb-hetFcγ4 (1 + 1)	TCE	follicular lymphoma	AstraZeneca
TQB2825	CD20 × CD3	Fab-IgG4-hetFc-modFab	TCE	R/R FL	Chia Tai Tianqing
Etentamig ABBV-383	BCMA x CD3	Fab/sdAb2-hetFcγ4	TCE	R/R MM	Abbvie
CM-336	BCMA X CD3	IgG4-hetFc modFab	TCE	R/R MM	Keymed
M701	EpCAM x CD3	Fab-/scFv-hetFcγ1	TCE	malignant ascites	YZY Biopharma
Alveltamig ZG006	DLL3 × CD3 (biparatopic)	Fab-/scFv2-hetFcγ1	TCE	SCLC	Zelgen
Pastritamig JNJ-78278343	KLK2 × CD3	Fab-/scFv-hetFcγ1	TCE	mCRPC	Janssen
Ramantamig JNJ-79635322	GPRC5D x BCMA x CD3	Fab-/scFv-hetFcγ1-scFv	TCE	R/R MM	Janssen
QLS-31905	GPRC5D x CD3	Fab-scFv-hetFcγ1	TCE	R/R MM	Qilu
QLS-32015	CLDN18.2 × CD3	IgG1-hetFc-scFv	TCE	PC	Qilu
Tebotelimab MGD-013	PD-1 × LAG-3	DART2-Fc	dual CPI	HER2+ GC/GEJ	Macrogenics
Volrustomig MEDI5752	PD-1 × CTLA-4	IgG1-hetFc modFab (DuetMab)	dual CPI	NSCLC, H&N, Cervix, Mesothelioma	AstraZeneca
Rilvegostomig AZD2936	PD-1 × TIGIT	IgG1-hetFc modFab (DuetMab)	dual CPI	NSCLC, GC, HCC, BTC	AstraZeneca
Erfonrilimab KN046	PD-L1 × CTLA-4	sdAb-sdAb-Fcγ1	dual CPI	NSCLC	Alphamab
LB1410	PD-1 × TIM-3	scFv2(HC)-IgG?	dual CPI	CC	L&L Bio
Acasunlimab GEN-1046	PD-L1 × 4-1BB	IgG1 Duobody	CPI + CoStim	NSCLC	Genmab
Pumitamig BNT327	PD-L1 ×VEGF	IgG1-sdAb2	CPI + AA	NSCLC, TNBC	BioNTech
SSGJ-707	PD-1 × VEGF	Fab-IgG-hetFcγ4 cLC	CPI + AA	NSCLC	Sunshine
HS-20117, PM1080	EGFR x c-MET	IgG1-modFab-hetFc	dual GRI	NSCLC	Hansoh
Petosemtamab MCLA-158	EGFr × LGR5	IgG1-hetFc-cLC	dual GRI	H&N	Merus
Izalontamab SI-B001	EGFR x HER3	IgG1-scFv2	dual GRI	NSCLC	Baili
izalontamab brengitecan BL-B01D1	EGFR x HER3	IgG1-scFv2 (HC)	dual RT ADC	SCLC, NSCLC, OC, BC, UC	Baili
Anbenitamab KN026	HER2 × HER2	IgG1-hetFc-cLC	biparatopic GRI	BC, GC	JMT-Bio
Anbenitamab ADC JSKN003	HER2 × HER2	IgG1-hetFc-cLC	bipara. RT ADC	BC, OC	JMT-Bio
TQB2930	HER2 × HER2	Fab-/scFv-hetFcγ1	bipara. RT ADC	BC	Chia Tai Tianqing
TQB2102	HER2 × HER2 ADC	Fab-/scFv-hetFcγ1	bipara. RT ADC	BC	Chia Tai Tianqing
Tovecimig CTX009	VEGF x DLL4	IgG1-scFv2	AA	BTC	Compass
Silevimig GR1801	RABV-GI x GIII	Fab-/scFv-hetFcγ1	VN	rabies	Genrix
Sonelokimab M1095	IL-17A/F x HSA	sdAb3	CKI + HLE	HS, PS	MoonLake
Gefurulimab ALXN1720	C5 × C5	sdAb2	bipara. CI	MG	Alexion
Lunsekimig SAR443765	IL-1a × IL-1b	sdAb5	CKI	COPD	Sanofi
Arumakimig MAS-825	IL-1b × IL-18	IgG1-hetFc	CKI	XIAP deficiency	Hospital for Sick Childen (Novartis)
Denecimig Mim-8	FIXa × FX	IgG4 Duobody	CA	Hemophilia A	Novo Nordisk
Trontinemab RO-7126209	Amyloid-β x TfR	IgG1-CrossFab 2 + 1	BBB delivery + Aβ clearance	AD	Roche
Lutikizumab ABT-981	IL-1α x IL-1β	DVD-Igγ1	CKI	HS	AbbVie
Remigromig MK-3000, EYE-103	LRP5 × Fz-4	Db-hetFcγ1-Fab2	RA	DME	Merck, Eyebiotech

Diseases: AD – Alzheimer’s disease, BC – breast cancer, BTC- biliary tract cancer, CC – cervix cancer, COPD – chronic obstructive pulmonary disease, DME – diabetic macular edema, EC – esophageal cancer, FL – follicular lymphoma, GC – gastric cancer, GEJ – gastroesophageal junction cancer, HCC – hepatocellular carcinoma, H&N – head and neck cancer, HS- hidradenitis suppurativa, mCRPC – metastatic castration-resistant prostate cancer, MG – myasthenia gravis, NSCLC – non-small cell lung cancer, OC – ovarian cancer, PA – psoriatic arthritis, SCLC – small cell lung cancer, TNBC – triple negative breast cancer, UC – urothelial cancer.

MoA: CA – coagulation activation, CI – complement inhibition, CKI – cytokine inhibition, GRI – growth factor receptor inhibition, AA – anti-angiogenesis, TCE – T-cell engagement, CPI – checkpoint inhibition, CoStim – Co-stimulation, RA- receptor activation, RT – receptor targeting, VN – virus neutralization.

Figure 4.

Formats of approved multispecific immunoglobulins. A) Cancer indications, B) Non-cancer indications.

Figure 5.

Formats of multispecific immunoglobulins in phase 2/3 or 3. A) Cancer indications. B) Non-cancer indications.

Figure 6.

Properties of T-cell engagers.

Except for three migs, all TCEs use CD3 as a trigger molecule on T cells, with at least 33% of the CD3 antigen-binding site derived from SP34, which is an antibody reactive with human and cynomolgus CD3.²⁶⁷ Two TCEs were found to bind to Vδ2 of γδ TCRs, one TCE is described to reactive with TCRβ. Furthermore, 6 migs utilize a strategy for conditionally activation/liberation of the CD3 binding site.²⁶⁸ Two of these migs, JANX007 and JANX008, have the CD3 binding site masked by a peptide sequence that can be removed by proteolytic cleavage within the tumor, thus liberating CD3 binding.⁵⁴ Others such as AMX-818 and AMX-500 exhibit PEG-mimetic polypeptides (XTEN) that sterically interfere with target binding.²⁶⁹ Proteolytic removal of these sequences makes these binding sites accessible. BA-3182, an IgG-scFv2(LC) TCE targeting EpCAM and CD3, is based on the CAB technology, in which both binding sites have little or no binding to the target antigens at neutral pH, but strong binding in the acidic tumor microenvironment.¹¹⁰

With 29 molecules, bivalent, bispecific IgGs are the leading class of TCEs. Different strategies are used to generate these IgGs. The IgG-hetFc cLC format is used in 11 TCEs using different Fc heterodimerization strategies. With linvoseltamab and odronextamab, two such TCEs are approved (Table 3). Both are IgG4 molecules with a silenced Fc, further comprising a S228P mutation in the hinge region, and using the Fc* strategy for purification of heterodimeric molecules. Seven TCEs are IgG1 or IgG4 Duobodies. Three such Duobodies are approved: teclistamab and talquetamab, both IgG4 Duobody with a silenced Fc targeting CD3 and either BCMA or GPRC5D, respectively, and epcoritamab, an IgG1 Duobody with a silenced Fc targeting CD20 and CD3.^{225,243,248,249} Five IgG migs comprise a heterodimerizing Fc and utilize post-purification assembly processes to generate bispecific molecules. This includes two approved migs, mosunetuzumab with a kih Fc targeting CD20 and CD3, and elranatamab, using charge-charge interactions for heavy chain heterodimerization, directed against BCMA and CD3.^84,233 One further bispecific IgG TCE (NILK-2301) is an unmodified molecule comprising a common heavy chain and κ and λ light chains (κλ-bodies).¹⁰¹ Finally, the group includes catumaxomab, a hybrid hybridoma generated by fusing a mouse and rat hybridoma directed against EpCAM and CD3, which was approved from 2009 to 2017 and just recently was approved again.⁷³

The Fab-/scFv-hetFc format is another frequently used format, found in 12 different TCEs. Two such TCEs, pasritamig and M701 are currently in Phase 3, both comprising a silenced γ1 Fc.^142,270 Several other 1 + 1 formats, 14 in total, have been applied to generate TCEs. Here, the scFv2 (tandem scFv) format should be mentioned, which is used in blinatumomab, the second approved TCE targeting CD19 and CD3, a format applied to generate 5 additional TCEs, two of them expressed by viral vectors. Tarlatamab, a TCE targeting DLL3 and CD3, is a next-generation half-life extended scFv2 derivative approved for therapy.¹⁵⁴ Here, the scFv2 moiety is fused to a silenced γ1 scFc. Finally, ImmTacs represent another class of 1 + 1 TCEs, combining an anti-CD3 scFv with a soluble TCR for targeting of MHC-displayed intracellular targets. One such ImmTac, tebentafusp, is approved for therapy of uveal melanoma, two further are in clinical development.⁴⁵

Migs with a 2 + 1 stoichiometry are an emerging class of TCEs. They combine a single CD3 binding site with two binding sites for a target antigen, thus utilizing avidity effects for target cell binding. In total, 19 TCE have a 2 + 1 valency comprising 7 different formats. This class of migs is dominated by two formats, Fab-/Fab-scFv-hetFc (7 migs) and Fab-IgG-hetFc (7 migs). A first Fab-IgG-hetFc, glofitamab targeting CD20 and CD3 and generated by the CrossMab technology, was approved in 2023,¹²³ one further Fab-IgG-hetFc, TQB2825, is in Phase 3, targeting the same antigens. The Fab-/Fab-scFv-hetFc are in Phase 1 or 2 trials. In this format the light chain problem, which applies to the Fab-IgG-hetFc molecules, is solved by having the Fab moieties with the same specificity, thus using the same light chain, and the CD3 binding site embedded as scFv, fused between an outer Fab and the Fc region.¹⁶⁷ In a related approach, a CD3 Fab fused to a first Fc-chain is combined with two sdAb moieties fused to the second Fc-chain to create sdAb2-/Fab-hetFc molecules. One such mig, etentamig directed against BCMA and CD3 is currently in a Phase 3.²⁷¹

These and other formats were furthermore applied to generate trispecific 1 + 1 + 1 migs. We identified 18 different 1 + 1 + 1 formats, five of them being Fc-less. In these Fc-less TCEs, the third binding site is directed against HSA for half-life extension. Examples of Fc-less formats include TriTACs with either the arrangement sdAb2-scFv or scFv-sdAb2, used by three migs, e.g., gocatamig and podentamig.²⁹ Similarly, the TRACTr format has the arrangement sdAb-scFv-Fab used by JANX007 and JANX008. Both molecules are conditionally active TCEs, by masking either one or two binding sites with a cleavable peptide. Several other 1 + 1 + 1 TCEs include an Fc region. For example, in Fab-/scFv-hetFc-scFv molecules an additional scFv is fused to the C-terminus of one Fc-chain.¹⁸⁰ Alveltamig (ZG006) is a biparatopic TCE targeting DLL3 and CD3 currently in Phase 3. It is composed of two scFvs fused to a first Fc-chain and a Fab fused to the second Fc-chain (scFv2-/Fab-hetFc).¹⁷⁹

All the above-described TCEs exhibit one CD3 binding site to avoid systemic T-cell activation by bivalent binding. Nevertheless, several formats are in clinical development that comprise two CD3 binding sites. This includes a few trispecific TCEs with a 2 + 1 + 1 stoichiometry, e.g., AMG-305,¹⁸² which consists of a central scFc flanked by scFv2 moieties. It thus represents an extension of the scFv2-scFv format used in tarlatamab by an additional tandem scFv fused to the C-terminus of the scFc moiety. AMG-305 binds monovalently to mesothelin and CDH3, and bivalently to CD3. TCEs with two CD3 binding sites furthermore comprise symmetric IgG fusion proteins with a 2 + 2 or 2 + 2 + 2 + 2 stoichiometry. A 2 + 2 valency is achieved, for example, by fusion an scFv to the C-terminus of the light or heavy chain, which is utilized by nine different TCEs, all in early phase development. In one case, trevotamig, the CD3 binding site is conditionally active. Alternatively, two different scFvs have been fused to the N- and C-terminus of an Fc region (scFv2-Fc-scFv2), a format used by mipletamig directed against CD123 and CD3.²⁷² Yet another approach uses chemical conjugates of two IgG antibodies. These conjugates are then used to arm patient-derived T cells with the bispecific antibodies (BAT) before reinjecting them back into the patient.²⁷³ This approach resembles the CAR-T approach, without the need for genetic manipulations.

Finally, it should be noted that basically all TCEs (94%) use strategies to avoid or at least reduce Fc-mediated effector functions. This includes the use of Fc-less formats, γ2 and especially γ4 heavy chains, which exhibit already reduced effector functions, although not completely silent, and as outlined above silencing strategies applied to γ1 and γ4 heavy chains (Figure 6). Of the 12 approved TCEs (Table 3, Figure 5), all Fc-comprising TCEs, except catumaxomab, comprise a silent Fc. The same applies for the Fc-comprising TCEs in Phase 3 (Table 4, Figure 6).

Migs as dual checkpoint-inhibitors

CPI with antibodies has become a central strategy in cancer immunotherapy. The interactions between tumor cells and immune cells, as well as between different immune cell types, play a key role in this process and are mediated by various pairs of cell surface receptors and ligands.²⁷⁴ Among these, the interaction between PD-1 on immune cells and PD-L1 on tumor cells is one of the best characterized, leading to the development of several approved monoclonal antibodies. Recent approaches have extended this concept to the use of migs targeting either immune-regulatory receptors and ligands expressed on the same cell (cis) or on different cells (trans). A total of 34 migs act as CPIs in cis. This includes 5 migs targeting PD-1 and CTLA-4, three further being trispecific for PD-1, CTLA-4 and VEGF to include inhibition of tumor angiogenesis. Furthermore, 5 migs target PD-1 and TIGIT, 5 migs targeting PD-1 and LAG-3, and 34 migs target PD-1 and TIM-3. Additional cis-acting multispecific CPIs include targeting of LAG-3 × CTLA-4, PRVRIG x TIGIT (4 migs), PD-1 × ILT4, CD96 × TIGIT, ILT2 × ILT4, and PD-1 with TGFβ-R2.

Trans-acting multispecific CPIs are combinations of PD-L1 targeting with either PD-1 (3 migs), LAG3, CTLA-4, TIGIT, ILT4, or TGF-β. One mig combines LAG-3 with CD73. In addition, 2 migs use targeted CPI by combining PD-1 with HER2 targeting (Suppl. Table S1). In total, we identified 44 migs using one of these approaches for CPI. The diversity of these molecules is reflected in the variety of formats, which are mainly dominated by 1 + 1 IgG molecules (33%) and 2 + 2 IgG – scFv2 fusion proteins (24%) (Figure 7). Except for two trispecific migs targeting PD-1, CTLA4, and VEGF, all migs are either 1 + 1 formats (42%) or 2 + 2 formats (59%). One multispecific CPI, cadonilimab (AK104), was approved in 2022 for the treatment of various forms of solid tumors. Cadonilimab is an IgG-scFv2 molecule that blocks PD-1 and CTLA-4 (Figure 4).²³³ Four further molecules are in Phase 2/3 or 3 trials: 1) volrustomig (MEDI5751), a IgG1-hetFc with a modified Fab (DuetMab) also directed against PD-1 and CTLA-4; 2) rilvegostomig (AZD2936), another DuetMab which is also a IgG1-hetFc blocking PD-1 and TIGIT; LB1410 with an unclear structure (presumably a 2 + 2 IgG fusion protein) which is directed against PD-1 and TIM-3; and 4) tebotelimab (MGD013), a DART2-Fc directed against PD-1 and LAG-3, which has recently completed a Phase 2/3 study (Figure 5). Notably, most migs employed as CPIs feature a silenced Fc region, minimizing effector function.

Figure 7.

Formats used for various mode-of-actions in cancer indications.

Migs as checkpoint-inhibitors and angiogenesis inhibitors

The combination of checkpoint inhibitors with anti-angiogenesis agents has emerged as a therapeutic option, for example, in lung cancer.²⁷⁵ Consequently, this approach was applied to generate migs. We found the combination of CPI by targeting either PD-1 or PD-L1 with inhibition of tumor angiogenesis by targeting VEGF in 13 migs. One mig, ivonescimab (AK112), an IgG-scFv2 fusion protein targeting PD-1 and VEGF is already approved for the treatment of lung cancer. Two additional migs are in Phase 3: pumitamig (BNT327), a 2 + 2 IgG1-sdAb2 fusion protein directed against PD-L1 and VEGF, and SSGJ-707, a 2 + 2 Fab2-IgG4 cLC blocking PD-1 and VEGF (Figure 5). All migs of this group with a known structure use a 2 + 2 stoichiometry, with 5 different formats being identified. Three migs of this group use an IgG-sdAb2 format, 2 further are IgG-scFv2 fusion proteins. The group includes three trispecific migs combining dual CPI with VEGF blockade (GB268, CS2009, and HC010), blocking PD-1, CTLA-4, and VEGF.

Migs as checkpoint-inhibitors and immune co-stimulators

CPI was further combined with immune cell costimulation.²⁷⁶ This approach is used by 13 migs, 9 of them directed against PD-L1 and 4-1BB. The other migs combine PD-L1 with CD27 or OX40, 4-1BB with CD73, or PD-1 with ICOS. Nine of the migs combining CPI with costimulation have a 2 + 2 stoichiometry, seven of them using the IgG-scFv2 format, and four other migs have a 1 + 1 stoichiometry. The most advanced mig is acasunlimab (BNT-311), a PD-L1 × 4-1BB targeting IgG1 Duobody, which is currently in Phase 3.²⁷⁷

Migs as targeted immune co-stimulators

A large group of migs comprising 35 molecules are applied for targeted immune cell costimulation combining a TAA with an immune stimulatory receptor, especially 4-1BB targeted by 20 migs, followed by CD28 used by 7 migs, and CD40 used by 4 migs. As found for the other migs with CPI or costimulatory activity, the majority (21 migs) has a 2 + 2 stoichiometry, many of them using the IgG-scFv2 or the IgG-sdAb2 format. Most of the other migs have a 1 + 1 stoichiometry using, for example, the IgG-hetFc-cLC format. One mig, CB-307, targeting PSMA and 4-1BB, is Fc-less but includes an HSA binding site. The group furthermore includes four tri- or tetraspecific migs, combining T-cell engagement with costimulation, and in three cases with CPI. All of these migs are in early-stage clinical development (Phase 1, 1/2 or 2) (Suppl. Table S1).

Growth receptor-blocking migs

Blocking of growth receptor-mediated signaling can be achieved by either blocking the receptors, their ligands, or a combination of it. Currently 12 migs are in clinical development targeting two different growth receptors (Suppl. Table S1). Eight of them are directed against EGFR and c-Met, with amivantamab approved in 2021⁷⁸ and HS-20117 in Phase 3. One of these migs is a trispecific molecule (BG-T187) which is biparatopic for c-Met. A further mig, TAVO412, is a trispecific antibody targeting EGFR, c-Met, and VEGF, thus combining receptor inhibition with anti-angiogenesis. Another mig (izalontamab) is directed against EGFR and HER3 and one (petosemtamab) against EGFR and LGR5.^278,279 Both antibodies are in Phase 3. Another mig, zenocutuzumab (MCLA-128) approved in 2022, is directed against HER2 and HER3 and was described to use HER2 to guide HER3 blockade to HER2-expressing tumor cells.²¹⁶ Inhibition of a single growth receptor is further achieved using biparatopic migs recognizing two different epitopes on the same receptor.²⁰³ Seven of such biparatopic migs are in clinical development, six of them directed against domain 2 and 4 of HER2, one against c-MET. The majority of these migs (11) are IgG molecules, i.e., binding monovalently to each receptor or epitope. Eight of them use a cLC in combination with a heterodimerizing Fc. Three migs are 2 + 2 IgG-scFv2 fusion proteins and two are 1 + 1 are Fab-/scFv-hetFc molecules (Figure 7). One further mig, tovecimig, is directed against two ligands, VEGF and DLL4 (a membrane-bound delta-like ligand 4), both involved in tumor angiogenesis. Tovecimig (CTX-009) is a IgG1-scFv2 fusion protein currently in Phase 2/3 trials.

Multispecific ADCs

Dual binding to tumor cell antigens can be used for drug delivery, for example, through ADCs. Consequently, migs were developed into ADCs.^280,281 We identified 26 different multispecific ADCs, 19 targeting two different receptors, and 7 being biparatopic (Suppl. Table S1). Like receptor-blocking migs, these ADCs commonly include EGFR as a target. Thus, 14 dual-targeting receptor-blockers comprise an EGFR binding site, combined with HER3 (5), c-MET (5), MUC1 (1), B7-H3 (1), or TROP2 (1) targeting. Biparatopic ADCs are dominated by HER2 targeting. In addition, c-MET, folate receptor (FRα), and 5T4 are employed. Several of these ADCs use as parental antibody a mig developed for receptor inhibition. Three ADCs are currently in Phase 3, including izalontamab brengitecan (BL-B01D1), anbenitamab repodatecan (JSKN003), and TQB2102 (Figure 5).^144,204,282 The majority of these ADCs use a 1 + 1 IgG-hetFc as format, followed by 4 Fab-/scFv-hetFc and one 2 + 2 IgG-scFv2 (Figure 7).

Migs and non-cancer indications

Compared to monospecific antibodies, where approximately an equal number of molecules are used for treatment of cancer and non-cancer indications, cancer is the dominating field for migs. Nevertheless, we identified 47 migs approved or in clinical studies for various non-cancer indications, using a diverse set of formats (Suppl. Table S1).

With 36 migs, the field of inflammatory/autoimmune diseases is the major application in non-cancer indications. Combined blockade of ligands and/or receptors involved in mediating proinflammatory signals is the major mode of action of migs used for treatment of inflammatory/autoimmune diseases. One mig, ozoralizumab, a sdAb3 with two binding sites for TNF and one for HSA is approved,²⁸³ and four others are in Phase 3, for example, targeting proinflammatory interleukins such as IL-17A x IL-17F (sonelokimab), IL1α x IL1β (lutikizumab) or IL1β x IL-18 (arumakimig).^133,284,285 Of note, several TCEs studied as cancer therapy (hematologic malignancies) by targeting CD3 in combination with other antigens such as CD19 or CD20 have found an application also in this indication.

The field of hematologic diseases (7 migs) is dominated by migs being pro-coagulants, for example, by targeting FIXa × FX or FVIIa × TLT-1. The prototype mig is emicizumab, a FVIII-mimetic FIXa × FX mig approved in 2017.²¹⁷ All the procoagulant migs are IgGs with a 1 + 1 stoichiometry.

The 6 migs for treatment of infectious diseases are designed to block relevant pathogen host cell interactions, and thus interfere with the infection process. A main indication is HIV infection, with migs directed against, for example, gp120 × CD3, gp41 × CD3, or gp41 × CD4.²⁸⁶ The most advanced molecule, however, is silevimig, which is directed against two surface proteins (RABV-GI and RABV-GIII) of the rabies virus, and which has completed a Phase 3 study.¹⁴¹

Four migs used in vascular diseases target Ang-2 and VEGF, thus interfering with angiogenesis with applications mainly to treat AMD and related diseases. The prototype mig is faricimab, a 1 + 1 IgG-hetFc, which was approved in 2022 for treatment of AMD, DME, and retinal vein occlusion.⁹³ One further mig, remigromig, is directed against LRP5 and Fz-4, antagonizing the Wnt signaling pathway, and thus restoring the blood-retinal barrier.¹⁸⁶

In neurologic diseases, the 4 migs currently in clinical development are used to deliver a therapeutic antigen-binding site through the BBB into the brain. Trontinemab (RO-7126209), which is in Phase 3, targets amyloid-β and uses binding to transferrin receptor (TfR) for brain shuttling. Here, a 2 + 1 IgG-Fab CrossMab format is used to achieve monovalent binding to TfR through the C-terminal CrossFab moiety, and bivalent binding to amyloid-β.¹¹⁷ Similarly, ABL-301, which recently completed a Phase 1, is a 2 + 1 IgG-hetFc-scFv directed against synuclein and IGF-1 R as brain shuttling target.¹¹⁶

Finally, two migs (AGA-2115; AGA-2118) are used to treat a metabolic disease (osteogenesis imperfecta) by blocking sclerostin and dickkopf-1 (DKK1), in addition to five TCEs (three approved migs), all directed against BCMA and CD3, studied for treatment of light chain amyloidosis (Suppl. Table S1).

Approved migs

The composition of the approved 20 migs is summarized in detail in Figure 4 and further information is provided in Table 3. Twelve of the approved migs are TCEs, most of them approved to treat hematologic malignancies. The majority of the approved migs are IgG derivatives, with different strategies utilized to generate them. Some of them are produced by a post-assembly strategy from two separately produced IgGs employing a heterodimerizing Fc (mosunetuzumab, elranatamab) or cFAE (Duobodies, 4 migs). Several others (3 migs) combine a heterodimeric Fc with a cLC (3 migs) or combine a hetFc with a CrossFab (e.g., faricimab). To date, only one Fab-/scFv-hetFc mig (zanidatamab) is approved, although the format is extensively used in other migs currently in clinical development. The same applies to the IgG-scFv2(HC) format used by more than 40 migs in clinical development. Two migs with such a format are approved (cadonilimab, ivonescimab). Three migs have an Fc-less format, blinatumomab (scFv2), ozoralizumab (sdAb3) and tebentafusp (scFv-TCR). One of them, ozoralizumab, comprises an HSA-binding sdAb for half-life extension. A half-life extended version of the scFv2 format is used by tarlatamab, with the scFv2 moiety fused to an scFc. Finally, catumaxomab, the first approved mig, is derived from a hybrid-hybridoma generated from mouse and rat parental hybridomas.

Migs in late-stage clinical development

Currently, 39 migs are listed on ClinicalTrials.gov in late-stage (Phase 2/3 or 3) development (Table 4). Thirty are studied for cancer therapy, nine for non-cancer indications. For 37 we could identify the format (Figure 5), one further, LB1410, is presumably a 2 + 2 IgG fusion protein. Four of them are Fc-less, including brenetafusp using the same scFv-TCR format as in tebentafusp. The other three Fc-less migs are multimers of sdAbs, all comprising an HSA-binding sdAb. A highly diverse set of formats are used by the 35 Fc-comprising migs in Phase 3. Ten are IgGs (including 3 IgG-hetFc-cLC, 2 Duobodies and 2 DuetMabs), 8 migs are extended IgGs with a 2 + 1 or 2 + 2 stoichiometry, 3 of them IgG-scFv2(HC), and the remaining 14 migs are Fc-fusion proteins, with 5 of them being Fab-/scFv-hetFc. Of note, a first trispecific mig, ramantamig, targeting GPRC5D, BCMA, and CD3, is in late-stage clinical development. In total, 23 different formats (counting the different IgGs as one group of formats) are used.

Conclusions and outlook

The clinical development of bispecific and multispecific antibodies has revealed an extraordinary diversity of molecular formats, reflecting the creativity of antibody engineering over the past two decades. Formats vary widely in valency, specificity, geometry, and affinity, and these structural features are carefully designed to influence functional outcomes such as immune cell recruitment, receptor crosslinking, or simultaneous pathway blockade. The rapid evolution of formats raises important questions: has the zoo of bispecific formats transformed into a jungle where formats compete with each other for the best performance and do we need so many designs? In practice, format selection is largely dictated by functional requirements, target biology, and desired pharmacokinetic and pharmacodynamic profiles.²⁸⁷ The principle that “format follows function” is often valid,^288,289 but the choice of a particular architecture also impacts manufacturability, stability, and immunogenicity^290–292

Formats that have failed during development have frequently suffered from poor stability, aggregation, immunogenic responses, or suboptimal drug-like properties, providing important lessons for future molecular design. Because clinical applications vary widely, the optimal pairing of binders and molecular formats must be tailored to the intended mechanism of action and cannot be generalized. For instance, bridging of coagulation factors (for example, to replace FVIIIa), forming cell – cell synapses in T- or NK-cell engagers, enabling avidity-driven simultaneous binding to multiple cell-surface antigens, or neutralizing soluble targets each require distinct architectural solutions rather than a one-size-fits-all format. Although different functionalities will continue to necessitate diverse molecular designs, it is increasingly evident that successful formats must exhibit high stability and minimal immunogenicity while avoiding major developability liabilities. Consequently, among clinical multispecific antibodies, certain formats have clearly emerged as preferred choices, likely due to their favorable developability and safety profiles (Figure 8). Of the 322 migs of category 1 for which the format was identified, at least 95 adopt an IgG-based structure, spanning 12 distinct engineering approaches. Within these, molecules using a cLC are particularly prevalent, 28 employing a hetFc and 11 using Fc*. Another frequently applied IgG architecture is IgG-hetFc with a modified Fab, including CrossFabs, Fabs with engineered disulfide bonds, e.g., DuetMabs, or charge-pair-stabilized Fabs (19 migs), followed by Duobodies (15 migs). The Fab-/scFv-hetFc format, also characterized by a 1 + 1 stoichiometry, is represented in 30 migs. Among formats with a 2 + 2 stoichiometry, IgG-scFv2(HC) is used in 47 migs, IgG-sdAb2(HC) in 12 migs, and Fab2-IgG formats in 6 migs. Beyond the 1 + 1 and 2 + 2 architectures, migs with variable valencies have also emerged. Notably, several 2 + 1 formats have been developed, particularly for TCEs, which feature one binding site for CD3 and two identical or distinct binding sites for target antigens. Key examples include various Fab-IgG-hetFc designs (8 migs) and the Fab-/Fab-scFv-hetFc format (7 migs), which collectively account for 15 different migs. Furthermore, the Fc-less migs are predominantly represented by the scFv2 (6 migs) and sdAb3 (5 migs) formats. Notably, comparison of category 1 and category 2 migs reveals a marked shift toward Fc-containing architectures, with Fc-less designs comprising 26% of category 2, but only 12% of category 1.

Figure 8.

Overview of usage of multispecific immunoglobulin formats. A) Summary of frequency of formats in category 2 migs (with have been completed before 2023, terminated, withdrawn, suspended or having an unknown status. B) Summary of frequency of formats in category 1 migs with the status not-yet-recruiting, recruiting, active or which has been completed in 2023 or later (IgG-like migs, 2 + 1 fab-IgGs, and 2 + 2 Fab2-IgG have been grouped). The most frequently used formats are shown in colors. C) Top 9 mig formats frequently used by category 1 migs.

Collectively, the above described formats account for 67% of all category 1 migs (216 migs), while the remaining 65 formats – each appearing between one and four times – represent the remaining 33% of molecules (Figure 8). This distribution highlights that while a subset of IgG-based and related formats dominates clinical development, the broader landscape continues to explore diverse architectures to optimize function, valency, and target engagement.

Conclusions from clinical studies are that a balance between functional innovation and drug-like attributes is critical.²⁹⁰ Formats with optimized geometry, valency, and affinity can enhance potency and selectivity, but excessive complexity can complicate production and regulatory approval. Consequently, when selecting a format, developers must weigh functional requirements against manufacturability, stability, solubility, immunogenicity risk, and the intended clinical application.

Looking forward, the field may see a gradual consolidation of formats around those architectures that consistently demonstrate favorable safety, efficacy, and manufacturability, particularly for indications with high clinical need.²⁹³ Nevertheless, continued exploration of novel designs is likely to remain a hallmark of bispecific antibody research, as antibody engineering enables solutions for increasingly complex therapeutic challenges. Ultimately, the diversity of formats is both a reflection of the field’s ingenuity and a reminder that thoughtful, rational design remains essential to translating multispecific immunoglobulins into safe, effective, and clinically viable therapeutics.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Disclosure statement

UB is an employee of Roche. REK is a consultant to BioCopy, BioNtech, Immatics, Oncomatryx, Plectonic, Roche, and SunRock.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2026.2613548