2. Major Classes of Multispecific Biotherapeutic Drugs

By definition, multispecific biotherapeutic drugs are protein-based therapeutic molecules that can engage multiple drug-target binding interfaces concurrently. The modalities of these biomolecules can be antibodies, different formats of antigen-binding fragments, small scaffold protein domains, peptides, enzymatic domains, protein receptors, chemical molecules, or oligonucleotides like anti-sense RNA or small interference RNA (siRNA). These biological entities are covalently linked together to form a single molecule to achieve multiple therapeutic modes of action. For the protein components of multispecific biotherapeutics, they are typically generated through recombinant DNA technology. For entities like chemical molecules, oligonucleotides, and some peptides, they are covalently conjugated to the protein moieties through in vitro manipulation.

Antigen-binding moieties are the dominant components for multispecific biotherapeutics as they are the main building blocks for further genetic fusion or chemical conjugation. The molecular designs for the antigen-binding moieties have been extensively reviewed [9,10], with around 100 different formats developed over the past two decades. These formats consist of various binding modules arranged in different domain repeats and fusion orders. As summarized in Table 1, the binding modules typically include fragments of antibody (Fabs), single-chain variable fragment (scFv), diabodies, single-domain antibodies (VHH from llama or camel, VNAR from shark), synthetic peptides, receptor domains, enzyme domains, CH2-based nanoantibodies or non-antibody scaffolds (Affibody, Fynomers, Monobodies, DARPins, Knottins, VLRs) [9,10,21,34,35]. The connections between these modules are either amino acid linkers encoded by DNA or chemical linkers synthesized in vitro.

One important criterion for multispecific polypeptides is the presence or absence of the antibody’s “fragment crystallizable region” (Fc) portion which plays a critical role in effector functions and FcRn-mediated recycling [36,37]. Based on the heterodimerization of Fc, multispecific molecules can be further classified into symmetric [9,10,38,39] or asymmetric architectures [4,30,40]. Symmetric multispecific biotherapeutics are homodimers with identical binding modules in each Fc fusion monomer, whereas asymmetric multispecific biotherapeutics are heterodimers with different binding modules in either of the Fc fusion monomers.

One unique problem posed for Fc-containing antibody-like multispecific biotherapeutics is the “chain-pairing” issue (detailed discussions are seen in several excellent reviews [4,9,10,30]), due to their need for the correct assembly of two antibody-like heavy chain (HC) polypeptides, and two or more antibody-like light chain (LC) polypeptides into the final multispecific molecules. A number of technologies have been developed to overcome this “chain-pairing” problem. As summarized in Table 2, various genetic alternations in the polypeptides, which introduce specific protein-protein interaction fits through shapes, charges or domain swaps between protein chains, have been proven successful in producing correctly assembled multispecific antibodies.

Table 1. Building blocks of multispecific biotherapeutics.

Binding Modules	Binding Modules	Molecular Weight	Molecular Properties	References
Fragments of antibody (Fabs)		~50 kDa	Medium half-life/less aggregated	[9,10]
Single-chain variable fragment (scFv)		~25 kDa	Short half-life/less aggregated	[9,10]
Diabodies (Db)		~25 kDa	Short half-life/less aggregated	[9,10]
Single-domain antibodies (VHH from llama or camel, VNAR from shark)		~15 kDa	Short half-life/less aggregated	[9,10]
Synthetic peptides		~3 kDa	Extremely short half-life/less aggregated	[41,42,43,44]
TCR domains		~50 kDa	Medium half-life/less aggregated	[45,46,47,48]
Enzyme domains		~50 kDa	Medium half-life/less aggregated	[49]
Small scaffolds (Affibody, Fynomers, Monobodies, DARPins, Knottins, VLRs, Nanoantibodies)		~5–12 kDa	Short half-life/less aggregated	[21,50,51,52,53,54,55,56]
Chemical molecules		~3 kDa	Short half-life/aggregated	[6,17]
Oligonucleotides		~10 kDa	Short half-life/less aggregated	[19,57]

Non-protein entities utilize small molecules and oligonucleotides to greatly broaden the mechanism of action and design versatilities of the multispecific biotherapeutics [6,16,17,18,19]. They enable specific binding to intracellular drug targets such as those involved in replication and division of cells and DNAs. These non-protein moieties are covalently linked to protein backbones (typically antibodies) that target specific surface antigens in cells and tissues. Upon the arrival to either the extracellular space or inside cells through endocytosis, the conjugated chemical molecules or oligonucleotides are released to exert therapeutic effects.

The current generation of multispecific innovative drugs have ridden a wave of promising clinical and regulatory successes. As of January 2021, thirteen multispecific molecules have been commercialized. Among them, there are three approved bispecific antibodies (catumaxomab (Removab^TM), blinatumomab (Blincyto^®), emicizumab (Hemlibra^®)), one approved antibody binding fragment fusion with immunotoxin (Lumoxiti^TM), and nine approved antibody-drug conjugates (gemtuzumab ozogamicin (Mylotarg^®), brentuximab vedotin (Adcetris^®), ado-trastuzumab emtansine (Kadcyla^®), inotuzumab ozogamicin (Besponsa^®), enfortumab vedotin (Padcev^TM), fam-trastuzumab deruxtecan-nxki (Enhertu^®), polatuzumab vedotin-piiq (Polivy^TM), belantamab mafodotin-blmf (Blenrep^TM), sacituzumab govitecan (Trodelvy^TM)). These successes, along with nearly 200 multispecific biotherapeutic candidates in the clinical pipelines [3,4,5,6,10,16,17,18,19,20,21,22,23,24,25,26,27,28], indicate that multispecific biotherapeutics are promising molecules. However, discovering and developing them into an approved product is challenging and requires extensive efforts in engineering and development. The following sections summarize the recent advances of the major classes of multispecific biotherapeutics drugs (Figure 1) in designs and applications.

2.1. Immune-Cell Engagers

Directing immune cells to target tumor cells represents one major group of mechanism of action for the multispecific biotherapeutics. These so-called immune-cell engagers are the quintessential example of the obligate multispecific biotherapeutics which take advantage of the presence of different binding specificities within a single polypeptide molecule to achieve a new function. They are designed to form an artificial cytolytic synapse by bringing immune cells, such as cytotoxic T-cells and natural killer (NK) cells, into close proximity with tumor cells [4,5,11,23]. Subsequent release of granzymes and perforin from activated T or NK cells into tumor cells results in tumor cell death. In the meantime, activated NK cells can also evoke target cell caspase activation via TNF-related apoptosis-inducing ligand and Fas ligand. Versatile factors can be secreted by NK cells for modulating functions of other immune cells.

The formats for the immune-cell engagers are generally classified into IgG-like molecules and non-IgG-like molecules [4,5,9,10], each with advantages and disadvantages. IgG-like molecules retain the Fc component that may endow effector functions and FcRn-mediated recycling for longer serum half-lives, whereas large sizes of these molecules could hinder their tissue penetration. To generate multispecific IgG-like engagers that interact with both immune cells and tumor cells, a number of antibody formats have been reported. Quadroma relies on the fusion of two distinct hybridomas [85], through which catumaxomab is produced in a rat/mouse quadroma cell line. Engineering chain-pairing scaffolds with “Knobs-into-holes” [58,59] have been widely used [86,87,88], whereas electrostatic steering strategies [60,61] have been shown with successful applications [89,90]. Dual-variable domains Ig (DVD-Ig) [91,92] (formed by the fusion of the variable domains of two antibodies in tandem), IgG-scFv [9,10] (produced by fusing scFv or Fv to the termini of antibody chains), κλ bodies [77] (sharing same heavy chain with two different light chains (κ or λ) with highly selective affinity resins for purification), trivalent format [93], and various tetravalent formats [80,81,94] have been utilized for the design of new immune-cell engagers.

Figure 1. Examples of multispecific biotherapeutics drugs. (A). T-cell and NK-cell engagers (TCE/KCE) [4,5,11]. (B). Antibody-drug conjugates (ADC) [3,6,15,17,18,31]. (C). Tetherbody cytokine [95]. (D). Piggybacking tetherbody [96,97,98]. (E). Emicizumab matchmaker [99,100]. (F). Sweeping antibody matchmaker [3,101]. (G). Agonist matchmaker [102,103]. (H). Linker matchmaker [104]. (I). Receptor attenuating matchmaker [105].

Non-IgG-like molecules do not contain Fc, but rather have different antigen binding fragments. They can become dimers, trimers, or tetramers depending on linker length and fragment sequences. A number of non-IgG-like formats have been reported: scFv-based molecules (exemplified by blinatumomab with the fusion protein containing light-chain variable domain (VL) and heavy-chain variable domain (VH), or tandem scFvs), diabody format (single-chain diabody, tandem diabodies, exemplified by dual-affinity retargeting molecules (DART)), and nanobodies from llamas, camels, or shark. The smaller size of the non-IgG-like formats offers advantages like enhancing tissue penetration, ease of manufacturing, and accessing sterically hindered epitopes. Their drawbacks are short half-lives and poor target-site retention. Besides the improvement with constant administration in the case of blinatumomab, a number of strategies have been put in place for serum half-life extension [106]. Attachment of highly flexible and hydrophilic molecules such as polyethylene glycol, carbohydrates, N-(2-hydroxypropyl) methacrylamide, dextran, and amino acid polymer XTEN [107] can increase the hydrodynamic volume. Fusions with human serum albumin (HSA) or albumin-binding domain like scFv to HSA can increase half-life significantly [108,109]. Additionally, multimerization can increase sizes and binding valency of the molecules [10,29].

2.2. Antibody Drug Conjugates

Antibody drug conjugates (ADCs) (Table 4) are an important class of multispecific biotherapeutics that combine two or more binding activities into one entity [3,6,15,17,18,31]. They are composed of chemical compounds (cytotoxic drugs or nucleotide acids) covalently linked to an antibody component that binds to a specific target. The selective binding to the target brings the conjugated chemical components into specific cells or tissues where they can exert their therapeutic effects. This should significantly limit the undesirable side effects of cytotoxic drugs or oligonucleotides. ADCs are often considered as prodrugs that are supposed to be nontoxic in circulation. A key attribute for an ADC is the average number of payloads (conjugated drugs) per antibody or the so-called drug-to-antibody ratio (DAR). The DAR can affect an ADCs’ potency and toxicity as well as plasma stability, biophysical properties, and pharmacokinetics.

Similar to that of immune-cell engagers, the successful development of an ADC depends on the selection of an appropriate target antigen for the antibody component. Besides high tumor-association to minimize on-target toxicity of ADCs, the antigens engaged by ADCs need to be endocytosed upon antibody binding. High copy numbers of antigens’ surface expression in tumor cells (>10⁵/cell, [154]) is crucial for the efficacy of ADCs because as more surface antigens bind to the ADC, then more chemical drugs can be delivered into the tumor cells [6,17]. Since most TAAs also express in normal tissues at lower levels, bispecific ADCs targeting two TAAs should increase tumor cell specificity. Such molecules have been recently shown to increase ADCs’ internalization and reduces tumor resistance in preclinical models [155].

For designing an effective ADC, selection criteria for cytotoxic payloads should include strong cell toxicity, preservation of potency after conjugation release, acceptable aqueous solubility, stability in aqueous formulation and physiological conditions, as well as synthetic feasibility under good manufacturing practice [6,15,17,18,31]. Most cytotoxic drugs are hydrophobic and tend to induce antibody aggregation which cause fast clearance and acute host responses [17,156]. The current major categories for cytotoxic payloads are microtubule inhibitors and DNA-damaging agents. Antimitotic agents cause cell death by interfering with the ability of mitotic spindles to segregate chromosomes and alter cellular cytoskeletal architecture [154,157]. The two most widely used antimitotic agents for ADCs are based on auristatins and maytansinoids [154,157]. For DNA-damaging agents, they are DNA binding agents that attach to the double-helix and promote DNA strand alkylation, scission or crosslinking [154,157]. The representative examples for this agent class are calicheamicins, duocarmycins, camptothecins, anthracyclines, and pyrrolobenzodiazepines. Other mechanisms of action for ADCs’ cytotoxic payloads include Topoisomerase I inhibitor-based payload, RNA splicing inhibition using natural product-derived thailanstatin, and RNA polymerase I inhibition using α-amanitin derivatives [18]. Besides small molecule payloads, synthetic oligonucleotide therapeutics (antisense oligonucleotides and siRNA) [19,57,158] have been conjugated into the antibody backbone showing efficacy in vivo and in vitro [159,160,161]. These payloads can target the RNAs encoding those intracellular proteins that are inaccessible and un-druggable by antibodies alone.

Linkers that connect payloads to antibodies are another critical feature of ADCs [6,15,17,18,31]. They must be stable in circulation and able to release the drugs in active form in the targeted location. The intrinsic features of the linkers and the payloads determine the physicochemical properties of ADC metabolites in nearby cells. As a result, the linkers induce a “bystander effect” to kill neighboring cells [6,15,17,18,31]. They can be classified into non-cleavable (like the one in trastuzumab emtansine) and cleavable (like the one in brentuximab vedotin). Non-cleavable linkers remain attached to the drugs after proteolytic degradation of the antibody in lysosome, which with linkers and an amino acid or short peptide are released into the cytosol to exert their activities [162]. The most common example of a non-cleavable linker is the thioether linker. For cleavable linkers, they exploit the differences between bloodstream and cytoplasm of targeted tissues. They can be cleaved by certain specific proteases found in lysosomes, or by low pH (acidic environment), or by a reducing environment (high glutathione concentrations in cytoplasm) to release free drugs [163]. Peptide linkers, hydrazone, and disulfide are a typical type of cleavable linkers.

Conjugation strategy is another important aspect for ADCs. Linker-payload moieties are typically conjugated to antibodies via lysine or cysteine residues. Ideally, the conjugation process should maintain the integrity of antibodies and the activity of drugs, as well as produce homogeneous batches that remain stable during the storage and lyophilization process [17]. Because the IgG scaffold has over 80 lysine residues, about 20 of which are highly solvent-accessible [164], the conjugation process results in highly heterogenous mixtures. For cysteine conjugation, four interchain disulfides in one IgG1 also lead to a heterogeneous ADC [165]. More homogeneous ADCs can be produced through site-specific conjugation via genetic engineering [6,15,17,18,31,166,167]. Engineered cysteine substitution at light chain and heavy chain positions provides reactive thiols and allows site-specific drug conjugation [168,169,170,171]. Unnatural amino acids capable of bio-orthogonal chemical conjugation can be introduced to antibodies through amber codon suppression [172]. Transglutaminase couple glutamine residues side-chain with linker-payloads through amide bonds [173,174]. Sortase catalyzed the ligation of the LPXTG tag at the C-terminus of a polypeptide with the N-terminal oligoglycine [175]. Glycotransferases can be engineered to transfer modified monosaccharides that enable bio-orthogonal chemical conjugation to a linker-payload [176]. Formylglycine-generating enzyme can covert an engineered cysteine residue in a specific peptide sequences such as CXPXR to produce an aldehyde tag for linker conjugation [177]. Most recently, engineered methionine residues have been utilized for efficient site-specific drug conjugation [178,179], and ADP-ribosyl cyclases enable a facile approach for generating site-specific ADXs CD38 fusion proteins [49].

2.3. Tetherbodies

Tetherbodies are a class of multispecific biotherapeutics drugs that utilize one binding interface to engage a docking protein for action enrichment to a particular location, with another binding interface to engage an effector protein. Regarding the mechanism of action, tetherbodies are similar to ADC, but its modalities are all protein-based and produced by recombinant DNAs [3].

The typical examples for tetherbodies are antibody-immunotoxin fusion (Lumoxiti^TM) or antibody-cytokine fusion [95]. Immuotoxin or Cytokines (such as IL-2) have adverse effects when administered systemically. When fused to an antibody that binds to a TAA, immunotoxin or cytokines like IL-2, tetherbodies can be enriched and interact with receptors at those tumor sites. This interaction can greatly limit the toxicities of the resulting tetherbodies. Another example for tetherbodies is the “piggyback” or “hijacking” approach [3,4]. Because blood-brain barrier (BBB) restricts access of large molecules to the brain parenchyma, multispecific biotherapeutics with one antibody arm against plasma membrane receptors, e.g., transferrin receptor (TfR) that transcytose through brain endothelial cells, have been explored for efficient brain entry [96,97,98,180,181]. The efficacies of these tetherbodies have been demonstrated in animal models [97,181], yet clinical successes remain to be seen. Identifying novel receptors that are brain endothelial cell-specific represents a new opportunity. A similar piggyback mechanism for tetherbodies has been used for the indication of infectious diseases. By targeting persistent factor Psl and the needle tip protein PcrV of Pseudomonas aeruginosa [182], the resulting bsAb can increase internalization and localization in acidic vacuoles for neutrophil killing. In addition, by utilizing one arm targeting an extracellular exposed epitope in the filovirus glycoprotein (GP) of Ebola virus to gain access to endosome during viral uptake [183], a “Trojan horse” bispecific antibody used the other arm against a cryptic receptor binding site of GP or the corresponding Niemann-Pick C1 intracellular receptor to block virus entry to cytoplasm.

One novel subclass of tetherbody-like multispecific biotherapeutics drugs is the condition-activated tetherbody that can enrich the active drugs in a particular location through the design of conditional activation. The conditional activation mechanisms take advantage of special conditions in tumor microenvironment such as increased levels of specific proteases [184], ATP [185,186], and others [20]. Because the presence of tumor antigen in healthy tissues can cause on-target toxicity by ADCs or bsTCEs, Probodies^TM (CytomX) with a protease-cleavable peptide linker and a masking peptide can presumably remain inert until proteolytically activated locally in disease tissues [184]. A novel class of ADC called Probody-drug conjugates has shown the potential to protect normal tissues expressing target antigens [187]. The probody for EGFRxCD3 bsTCE has reported a 60-fold increase in dose tolerance compared to the unmasked construct [20,25]. A similar conditional activated TCE called ProTIA (protease-triggered immune activator) has been developed by Amunix and its AMX-168 is an anti-EpCAM BiTE-TCE fused to a masking XTEN polymer by a protease-cleavable linker [25].

2.4. Biologic Matchmaker Drugs

Biologic matchmaker drugs are a relatively new class of multispecific biotherapeutics that utilize one binding interface to engage an action partner, with another binding interface to engage a target [3]. They bring the drug targets in close proximity to an action partner for efficient biological activation or inactivation.

The best known example of a biologic matchmaker drug is emicizumab that replaces the scaffold function of factor VIII for generating activated FX [99,100]. Through hetero-binding arms, this bispecific antibody brings the substrate FX to the protease enzyme FIXa. A similar biologic matchmaker is a molecular linker that connects laminin-211 and the dystroglycan beta-subunit [104]. A novel bispecific antibody targeting both laminin and dystroglycan proteins has been developed to ameliorate sarcolemmal fragility for improving muscle function [104].

Biomimetics for signaling molecules are another type of biologic matchmaker. Bispecific agonistic antibodies that engage with FGFR1 with one arm and Klothoβ with another arm [102,103] have been developed to mimic fibroblast growth factor 21 for its signaling action. Most recently, bispecific biologic matchmakers have been constructed to recruit promiscuous cell-surface phosphatase CD45 to attenuate signaling of kinase-activated cell surface receptors such as PD-1 and SIRPα through dephosphorylation [105].

There are also other novel formats for biologic matchmaker drugs in early stage preclinical research. One example is the multispecific biotherapeutics that utilized the mechanism similar to heterobifunctional proteolysis-targeting chimeric molecules (PROTACs) [188,189,190]. Through the PROTACs’ first ligand to the target and the second ligand to ubiquitin ligase, PROTACs induce ubiquitylation of the target followed by proteolytic degradation in the proteasome [3]. A similar PROTACs-like multispecific biotherapeutics has been produced by Kanner et al. [191]. This PROTACs-like matchmaker consisted of a llama single-chain antibody fused to the catalytic domain of E3 ligase which can increase targeted ubiquitination of ion channels and consequentially diminish the surface expression of these membrane proteins. Interestingly, an opposite application has also been recently reported by the same research team. A multispecific deubiquitinase, in which a deubiquitinase was fused to a nanobody, deubiquitinated trafficking-deficient ion channels and consequentially rescued their functional surface expression [192]. Similarly, a multispecific matchmaker, named sweeping antibody, utilizes a lysosomal degradation system without taking advantage of proteasome activity [3]. This multispecific matchmaker binds to either FcRn or a surface antigen (i.e., TfR) in one arm at both neutral and acidic pH, with another arm interacting with the target only at neutral pH [3,101]. When the multispecific is endocytosed to endosome, the target is then released and proceeds to lysosome for degradation.