1 Introduction

The application of monoclonal antibodies (mAbs) to cancer patients originates from the invention of hybridoma technology in 1975 [1]. Rituximab was the first mAb approved by the US FDA for cancer treatment in 1997. It binds to the B-cell-specific antigen CD20 [2] and was shown to improve the survival of B-cell non-Hodgkin lymphoma patients when combined with chemotherapy [3]. This success has catalyzed the development of the next generations of anti-CD20 mAbs as well as mAbs against other B- and T-cell antigens. Major limitations of mouse antibodies as therapeutic agents—immunogenicity, lack of effector functions, and short half-life—were subsequently identified and largely overcome by the advent of antibody chimerization and, later, humanization technologies. Today, antibody-based drugs represent the largest and fastest growing class of protein therapeutics, with 24 marketed agents in the US (28 FDA-approved and four later withdrawn from the market) [4] and at least 240 more are in clinical development [5]. The recruitment of effector cells by these antibodies has been shown to strongly affect their clinical efficacy (Fig. 1a). Despite some clinical efficacy of mAbs, it is becoming very clear that they need further improvements for optimizing their interactions with the human immune system. Towards this goal, one strategy that is at the forefront of mAb-improving technologies is optimization of the constant fragment (Fc) domain for enhanced interactions with the Fcγ receptors (FcγR) expressed on immune cells [6, 7].

Fig. 1
figure 1

Tumor targeting of accessory and effector cells by mAbs and bsAbs. a mAbs target tumor cells via their binding specificity and recruit accessory cells via their constant fragment (Fc) domain. This leads to crosslinking and to activation of the immune accessory cells against tumor cells. b If a functional Fc fragment is present, bsAbs can target accessory cells to tumor cells in the same way as mAbs (b1). Via their dual specificity, bsAbs can target T cells to tumor cells leading to the expected induction of a strong T-cell anti-tumor immunity (b2). bsAbs bispecific antibodies, CkR cytokine receptor, FcγR Fcγ receptors, mAbs monoclonal antibodies

Full size image

Another strategy is the development of so-called bispecific antibodies (bsAbs). These molecules refer to a class of artificial constructs in which two antibody-derived antigen-specific binding sites are aligned within one fusion molecule. For tumor therapy, bsAbs typically recognize simultaneously a target antigen on tumor cells, while the second site recognizes an antigen expressed on effector cells. The latter target could be CD3, CD16, or CD64, which are expressed respectively on T lymphocytes, natural killer (NK) cells or other mononuclear cells (Table 1). They bring the two cells together (effector cell recruitment), activate the effector cells, and thus facilitate tumor cell killing [8–12] (Fig. 1b). They are very attractive as they allow redirection of T lymphocytes, which lack FcγR, to tumor cells [13, 14].

Table 1 T-cell and accessory cell populations as effector cells with the target surface molecules which can be used for their triggering
Full size table

This concept of redirection and activation of T cells against tumor cells by bsAbs was first described over 20 years ago [11, 15]. Among the first formats pursued were constructs designed to bind the monomorphic T-cell receptor—cluster of differentiation 3 (TCR-CD3) complex, a strategy that offers a number of advantages. Triggering lysis by this approach allows bsAbs to directly interact with the T-cell compartment. The repertoire of T-cell specificities, which can be recruited, is much broader in case of activation via CD3 than in case of tumor antigen (TA)-specific activation. In addition, because the target-binding arm is derived, at least in part, from the variable portion of a tumor-specific antibody, the targeting of the T cells is tumor-specific while the cytotoxic activity of the T cells is largely a non-specific tumor-localized bystander effect. Such bsAbs possess the ability to bind tumor epitopes beyond the major histocompatibility (MHC) peptide complexes, which are classically recognized by TA-specific T-cell receptors (TCRs). As a result, they overcome mechanisms of tumor immune escape such as loss or downregulation of TA and/or MHC. The possibility given by such molecules to target T cells to tumor cells has raised much interest for cancer therapy [16, 17]. A plethora of bsAb formats have been developed for recruiting T cells against cancer and promising clinical results were recently obtained.

The focus of this review is the development of bsAbs over the last 20 years and their interaction with tumor cells and human effector cells. In particular, we will discuss the inclusion of further binding sites such as the Fc domain of antibodies or domains of cytokines, thereby creating trispecific reagents. Such trispecific molecules provide two types of actions:

  1. (i)

    Crosslinking CD3, CD28, or CD25 molecules on T cells to tumor cells, thereby providing respectively stimulatory signal 1 or costimulatory signals 2a and 2b (Table 1). Engagement of these receptors can trigger cytotoxicity by FcR-negative T cells.

  2. (ii)

    Recruiting accessory cells to tumor cells and activating them via FcγR or cytokine receptor (CkR) mediated signals (Fig. 1b).

We will also describe our own strategy and successful experience in the development of cancer vaccine, bsAbs, and trispecific immunocytokines. The latter represent a combination of bsAbs with cytokines. Such multivalent reagents can cross-link tumor cells, T cells, and CkR-positive accessory cells.

2 Fc and FcγRs Link Antibodies to Accessory Cells

Most of the mAbs developed for tumor therapy belong to the class G of immunoglobulins (IgG). These antibody molecules are globular glycoproteins composed of two identical light chains (L) and two identical heavy chains (H). Each IgG is thus bivalent and able to attach to two identical antigenic sites. This allows the increase of antibody functional affinity and confers high retention times via its variable domain (fragment variable, Fv). In contrast, the constant domain (fragment crystallizable, Fc) mediates the recruitment of FcR-bearing accessory cells. One of the regions of the Fc domain binds to the neonatal Fc receptor (FcRn)—an MHC class I-related molecule known to protect IgG and albumin from intracellular catabolic degradation following pinocytic capture [18]. This receptor regulates serum IgG levels by binding pinocytosed IgG in the endosomes and recycling it to the cell surface.

The Fc portion also interacts with components of the complement and of the immune system to recruit several cytotoxic effector mechanisms: (i) complement-dependent cytotoxicity (CDC) [19]; (ii) antibody-dependent cell-mediated cytotoxicity (ADCC), which corresponds to the antibody-induced lysis of target cells by activated NK cells; and (iii) antibody-dependent cell-mediated phagocytosis (ADCP) (Fig. 2a). The latter assists in the recruitment of monocytes/macrophages or dendritic cells (DCs) by binding to their Fcγ receptors, which results in engulfment of antibody-coated tumor cells. Such uptake of antibody-coated tumor cells by antigen-presenting cells may result in the induction of adaptive immune responses and systemic T-cell immunity [20, 21].

Fig. 2
figure 2

IgG, Fc receptors and their effector functions. a Four different mechanisms of effector functions derived from the Fc domain: CDC, ADCC, ADCP and immune activation. 1 Classical pathway activation of the complement (CDC). CDC can be triggered by binding of C1q to the Fc domain of antibody molecules bound on the target cells. This leads to cell cytolysis. 2 Lysis (ADCC). NK cells, which express the activating FcγRIIIa, can cause direct killing through release of cytotoxic granules and/or secretion of pro-apoptotic TNF family members. 3 Phagocytosis (ADCP), antigen presentation and T-cell activation. Phagocytic cells such as neutrophils, monocyte/macrophage, and dendritic cells (DC) can engulf target cells when FcγRIIa and FcγRIIIa are triggered, but are subject to regulation by the inhibitory receptor FcγRIIb. Phagocytosis by antigen-presenting cells (APC) such as macrophage and DC can also lead to MHC presentation of tumor antigen and activation of tumor-specific T cells. 4 Immune activation (cytokine/chemokine release). FcγR activation can also lead to release of cytokines and chemokines that could further amplify these other effector functions. For example, IFN-γ can modulate FcγR expression in favor of activating FcγR and increase MHC expression on target and APC. Chemokines can recruit additional effector cells to the site of activation. ADCC antibody-dependent cell-mediated cytotoxicity, ADCP antibody-dependent cell-mediated phagocytosis, CDC complement-dependent cytotoxicity, Fc constant fragment, FcγR Fcγ receptors, IFN interferon, NK natural killer, TNF tumor necrosis factor, ↑ indicates increase. b Structure, cellular distribution and IgG isotype binding affinity of human activating and inhibitory human constant fragment (Fc) receptors for IgG (FcγRs). FcγRs differ in function and affinity for the Fc fragment of antibody and in cellular distribution. There are five activating FcγRs: the high-affinity receptor FcγRI, which can bind monomeric IgG, and four low-affinity receptors (FcγRIIA, FcγRIIC, FcγRIIIA, and FcγRIIIB), which bind only immune-complexed IgG. Cross-linking of activating FcγRs by immune complexes results in the phosphorylation of immunoreceptor tyrosine-based activation motifs (ITAM) that are present either in the cytoplasmic domain of the receptor (FcγRIIA and FcγRIIC), or in the associated FcR common γ-chain (FcγRI and FcγRIIIA), resulting in an activating signaling cascade. FcγRIIIB is a glycosylphosphatidylinositol (GPI)-linked receptor that has no cytoplasmic domain. FcγRIIB is the only inhibitory FcγR. It is a low affinity receptor that binds immune-complexed IgG and contains an inhibitory motif (ITIM) domain in its cytoplasmic domain. FcγRIIB cross-linking by immune complexes results in ITIM phosphorylation and inhibition of the activating signaling cascade. FcγRs differ in their cellular expression; myeloid cells express FcγRI, FcγRIIA and FcγRIIIA, whereas granulocytes express FcγRI, FcγRIIA, and FcγRIIIB. In such cells, immune complex-mediated activation of these receptors is negatively regulated by FcγRIIB. FcγRIIB is the only FcγR expressed by B cells and negatively regulates B-cell receptor activation by immune-complexed antigen

aFcγR expression is dependent on activation state, typically modulated by cytokines. FcγRs are also expressed on other cell types that are not discussed in this review

bexpressed also on B cells

cmacrophages, DC, neutrophil

dFcγRIIc expression on the indicated cell types varies per individual. FcγRIIc expression on DCs is not known. Identity between FcγRIIc and FcγRIIb may vary slightly with FcγRIIc isoform

Full size image

The constant region of human IgG Abs contains a single site for asparagine (Asn)-linked glycosylation at position 297 on the two identical heavy chains, which maps to the Fc domain. These Fc glycans consist predominantly of the complex biantennary structures, which display a high degree of heterogeneity due to the presence or absence of various terminal sugars [22]. The presence of this Asn-linked glycan is essential for FcγR-mediated activity. Multiple noncovalent contact sites between the glycan and the Fc backbone, along with interactions between the glycan on one heavy chain and the glycan on the other heavy chain, serve to define Fc conformation. Consequently, the absence of the Fc glycan, whether in Abs deglycosylated by enzymatic removal of the glycan or in genetically engineered aglycosylated Abs, results in misconformations that perturb the binding regions for C1q complement protein and the FcγR. Although all IgG molecules have a common core oligosaccharide, they have marked heterogeneity due to differences in terminal glycosylation [23]. This implies that the structural analysis of sugar side chains is an important issue in the manufacturing process of antibody molecules with a functional Fc domain.

The human FcγR family contains six known members in three subgroups, including FcγRI (CD64), FcγRIIa,b,c (CD32a,b,c), and FcγRIIIa,b (CD16a,b) [24–26] (Fig. 2b). There are a number of subtle differences between the receptors that determine their biological function. FcγRI binds with high affinity (10−9 M) to monomeric IgG and is not capable of distinguishing between unbound IgG and immune complexes. By contrast, FcγRII and FcγRIII bind with low affinity (10−5–10−7) to monomeric IgG. The dynamic interaction of these receptors with immunocomplexes is thought to be an important trigger for immune activation. Central to FcγR biology is the balance between activating and inhibitory signals. Four of the receptors are activating because of a cytoplasmic immunoreceptor tyrosine-based activation motif (ITAM), which is either encoded directly (FcγRIIa/c) or gained by association with a common ITAM γ-chain (FcγRI and FcγRIIIa). By contrast, FcγRIIb possesses an inhibitory motif (ITIM) in its cytoplasmic domain, and signaling through this receptor negatively regulates effector functions. FcγRIIIb is unique in that it is linked to the membrane with a glycosyl phosphatidyl inositol (GPI) anchor and therefore does not signal. IgG1 is a preferred format for implementation of the ADCC and CDC effector functions, whereas IgG3 are potent CDC activators. IgG2 and IgG4 are poor CDC activators [27]. Slight variations in FcγRs exist in different individuals. Such allotypic variants are collectively known as FcγR polymorphism. Its association with clinical response to mAb therapy has now been increasingly appreciated [28]. This genetic polymorphism, in addition to the different expression of Fc receptor classes on various effector cell populations (Table 1) and of the effect of the antibody subtype, introduces further complexity into antibody-based molecules.

3 The First Generation of bsAbs are Large Molecules Containing Fc Domains

bsAbs were first generated by chemical conjugation or somatic hybridization (Fig. 3a). The latter mode of bsAb preparation took advantage of hybrid hybridoma or quadroma technology [29]. This was based on the somatic fusion of two hybridoma cell lines expressing murine mAbs of desired specificities. Because of the random pairing between two different Ig heavy and light chains, the quality of the desired bsAbs was poor, resulting in 80–90% of mispaired byproducts, and their purification was difficult. A certain number of such bsAb preparations were nevertheless tested in clinical trials. For some of the bsAbs, Fc fragments were eliminated by enzymatic digestion with pepsin (Fig. 3a). However, the obtained results with such reagents were difficult to reproduce because of their poor purity (Table 2).

Fig. 3
figure 3

Generation of bispecific antibodies (bsAbs) by chemical conjugation of two antibodies or somatic hybridization. a Generation of bsAbs by chemical conjugation or somatic hybridization. Chemical conjugation allows formation of bsAbs through elaboration of covalent binding between two different IgGs to generate bsAb IgG2 [or bsAb F(ab’)2 in case the constant fragments (Fc) were beforehand removed by pepsin digestion]. Somatic hybridization generates through the fusion of the two hybridomas producing, respectively, the IgGs whose binding activities have to be associated together. In this case, the immunoglobulin genes of both parental hybridomas are codominantly expressed in the resulting hybrid hybridoma and this leads to random association of heavy and light chains. Only one among the ten different antibody species is the desired bsAb IgG from which a bsAb F(ab’)2 can be derived via pepsin digestion. b Structure of a Triomab. The Triomab family are trifunctional bsAbs that maintain an IgG-like shape. These chimeras consist of two half antibodies, each with one light and one heavy chain, that originate from parental mouse IgG2a and rat IgG2b isotypes. Whereas every Triomab family member is composed of an anti-CD3 rat IgG2b half antibody for T-cell recognition, the antigen binding site presented by the mouse IgG2a isotype is exchangeable. Several Triomab® antibodies have been generated that bind to tumor-associated antigens, e.g., EpCAM (catumaxomab), HER2/neu (ertumaxomab), CD20 (FBTA05), gangliosides GD2/GD3 (Ektomun®), on appropriate tumor target cells associated with carcinomas, lymphomas or melanomas. c Activation and targeting of T cells and accessory immune cells to tumor cells by Triomab-mediated crosslinking of tumor antigen with CD3 of T cells and Fcγ receptors (FcγR) of accessory immune cells. A process of mutual cross-talk between T cells and accessory immune cells is initiated through addition of the trifunctional bsAb after triggering of the T cells via CD3 in an MHC-independent manner and stimulation of accessory immune cells via FcγR crosslinking leading to the release of proinflammatory cytokines which procure a strong costimulatory signal (signal 2) to T cells. ADCC antibody-dependent cell-mediated cytotoxicity, ADCP antibody-dependent cell-mediated phagocytosis, HER2 human epidermal growth factor receptor 2, mAbs monoclonal anti-tumor antibodies

Full size image
Table 2 Clinical trials involving the use of chemically cross-linked and somatically hybridized bispecific antibodies (bsAbs) for the engagement of effector cells
Full size table

A possible solution to the obstacles to an effective interaction of therapeutic antibodies with Fc receptors could be provided by bsAbs with one specificity for a target molecule on tumor cells and another specificity for FcγR. FcγRI (CD64)-directed bsAbs bind to monocyte/macrophages and some types of DCs. FcRII (CD32) has the disadvantage that it is present on a variety of non-effector cells, and was therefore not chosen as a trigger molecule for bsAbs. FcRIII (CD16)-directed bsAbs activate NK cells and macrophages, but will also bind to FcRIIIb which is not a cytotoxic trigger molecule. In addition to the IgG receptors, the myeloid receptor for IgA (FcαRI, CD89) was also demonstrated to function as a trigger molecule for bsAbs since this receptor is constitutively expressed on monocytes/macrophages, granulocytes, and subgroups of DCs. Such bsAbs, which were produced from quadroma and digested by pepsin to generate F(ab’)2, were also evaluated in clinical trials as Fab’2 (Fig. 3a and Table 2). However, the obtained results were not very convincing and such trials were quickly abandoned.

Table 3 Clinical trials involving the use of genetically generated bispecific antibodies (bsAbs) for the engagement of effector cells
Full size table

The formation of bsAbs via the quadroma technology was then further optimized, taking into account preferential species-restricted heavy/light chain pairing and employing hybridoma of different species (mouse IgG2a/rat IgG2b) [30]. TRION Pharma together with Fresenius Biotech develop hybrid mouse/rat IgG molecules with double specificity for a TA (epithelial cell adhesion molecule [EpCAM], human epidermal growth factor receptor 2 [HER2], CD20) and CD3 [31–33]. It is noteworthy that the Fc region composed of the two subclasses effectively binds to human FcγRI and FcγRIII on accessory cells but not to the inhibitory receptor FcγRIIb. Together with its two antigen-binding sites, this bsAb type is able to bind to tumor cells and to T cells, and simultaneously via its Fc portion, to FcγR+ accessory cells [34, 35]. Since these antibodies have three binding sites, it has been described as trifunctional and named Triomab. Catumaxomab (EpCAM × CD3) is the first of this antibody type (Fig. 3b) and so far the only bsAb on the market. It was approved in 2009 in the European Union for the intraperitoneal treatment of malignant ascites in patients with EpCAM-positive carcinomas when standard therapy is not available or no longer feasible (Table 2). Other Triomabs, targeting HER2 (ertumaxomab) and CD20 (FBTA05) on tumor cells have also reached clinical trials (phase I/II) and are being tested in metastatic breast cancer and B-cell lymphoma patients, respectively (Table 2).

The following diverse effector mechanisms are described as being induced by these antibodies composed of IgG1 and IgG2a immunoglobulin heavy chains: T-cell mediated lysis, ADCC and phagocytosis [34, 36]. These lead to tumor destruction, concomitant release of cytokines (such as tumor necrosis factor alpha [TNF-α] and interferon gamma [IFN-γ]) and induction of anti-tumor immunity [30, 37]. The formation of this postulated tri-cell complex (Fig. 3c) results in a physiological costimulation of the T cell and very efficient tumor cell destruction by various immunologic mechanisms: (i) activation of T cells which are then able to kill tumor cells by, for example, release of cytokines and lytic enzymes like perforin [35, 38]; (ii) prevention of anergy from T cells by the release of costimulatory cytokines and crosstalk between costimulatory molecules expressed on T cells and accessory cells; (iii) phagocytosis of tumor cells by FcγR+ cells such as macrophages and dendritic cells. As a consequence, TA-specific humoral as well as cellular responses are induced, leading to a long-lasting anti-tumor immunity via the involvement of antigen-presenting accessory cells [36].

A major drawback of chemically and somatically generated bsAbs is their heterogeneity. This hampers their development in the clinic because active pharmaceutical components must be homogenous and the production processes must be uniform and consistent. Recombinant DNA technology overcomes this problem.

4 Design of New Ig-Like bsAbs Containing Fc Domains by Recombinant DNA Technology

The advent of recombinant DNA technology has greatly facilitated the genetic manipulation of antibody fragments [39, 40], and the genetic manipulation of recombinant antibodies has led to the development of a large variety of engineered antibody molecules with specificities out of reach for conventional antibodies. A variety of recombinant IgG-like bsAbs, which contain a complete Fc region and thus are capable of exerting Fc-mediated activities such as ADCC, ADCP and CDC and enabling recycling by FcRn [41], has been generated. To ensure their bifunctionality, different strategies were used to optimize a correct assembly of heavy and light chains (Fig. 4a [top] and Fig. 4b).

Fig. 4
figure 4

Generation of IgG-like bispecific antibodies (bsAbs) by genetic engineering. a Recombinant bsAbs containing a constant fragment (Fc) domain or not. The recombinant bsAbs, which are obtained with this recombinant antibody technology, are a more promising tool in comparison with chemically or somatically generated bsAbs and can be categorized into two classes: (i) the Fc-containing IgG-like bispecific formats and (ii) the small bispecific formats that do not have an IgG Fc domain. DVD-Ig Dual variable-variable domain immunoglobulin

*Single-chain variable fragments (scFv) result from the fusion of the variable regions of the heavy (V H ) and light chains (V L ) of immunoglobulins, connected with a short linker peptide of 10 to about 25 amino acids. The linker is usually rich in glycine for flexibility, as well as serine or threonine for solubility, and can either connect the N-terminus of the VH with the C-terminus of the VL, or vice versa. scFv can be created directly from subcloned heavy and light chains derived from a hybridoma or from an immunized animal. Once cloned, it is then possible to increase the affinity and specificity of antigen binding by mimicking somatic hypermutations which take place during an immune response. It may even be possible to replace the existing practices of animal immunization and hybridoma development through a bacterial system (phage display) capable of synthesizing and expressing practically unlimited quantities of antibodies to almost any antigen. Such binding units can be replaced by others such as camel variable domain of heavy chains (VhHs; size about 15 kDa), VNARs which are the variable domain of antibodies from sharks and are candidates for the smallest antibody-based immune recognition units (approximately 13 kDa), or aptamers which are oligonucleic acid or peptide molecules that bind to a specific target molecule

aThe various formats of bsAbs containing an Fc fragment are represented in Fig. 4b and c

bsc-Kih: single chain knobs-into-holes (KiH)

cscFv-IgG is a C terminal fusion of an IgG to scFv

dIgG-scFv is a fusion protein having an IgG at the N terminus

eA tandem scFv molecule (TaFv) format constitutes the basis for bispecific T-cell engager (BiTE) molecules which are developed by Micromet. It consists of two scFvs connected by a peptide linker

fDiabodies (Db) comprise a heavy (VH) chain variable domain connected to a light chain variable domain (VL) on the same polypeptide chain (VH-VL) connected by a peptide linker that is too short to allow pairing between the two domains on the same chain. This forces pairing with the complementary domains of another chain and promotes the assembly of a dimeric molecule with two functional antigen-binding sites

gDerivates of single Db are generated by introducing cysteine residues at the VH-VL interface to obtain disulfide-stabilized Dbs (dsDb) or at the C termini to generate dual affinity retargeting (DART) molecules or alternatively by applying a KiH strategy to get a kiHDb

hSingle chain diabodies (scDb) recombinant molecules composed of the VH and VL chain domains of two antibodies connected by three linkers in the order VHA-VLB-VHB-VLA. Derivates of scDb are TandAbs molecules. They correspond to tetravalent bispecific antibody formats that have two binding sites for each antigen and possess the same avidity and affinity for each target as an IgG. They are generated by reducing the length of the middle linker

iSuperantigens (SAgs) are a family of bacterial and viral proteins that bind to a high frequency of T cells and activate them to cytokine release and cell-mediated cytotoxicity. Their genetic fusion with another antibody fragment allows the generation of bsAbs

jTCR-based bsAbs use the capacity of T-cell receptors (TCRs) to target HLA-peptides. A platform of TCR-based bispecifics called ImmTACs (immune mobilizing mTCR against cancer) that combine a soluble, affinity-enhanced TCR-based targeting system with an anti-CD3 scFv-based T-cell redirection effector function are developed by Immunocore. Sag fusion and TCR-based constructs are not under the scope of this review

kA multivalent immunocytokine is a fusion of a bsAb molecule with a cytokine molecule

b Scheme of some recombinant IgG-like bsAb formats. Details of each construct are described in text. CH constant domains of the heavy chain; CL constant domain of the light chain, VH variable domains of the heavy chain, VL variable domain of the light chain, green and orange knobs-into-holes (KiH), red coming from antibody with specificity for A, blue coming from antibody with specificity for B

c Schematic diagram of dual variable-variable domain immunoglobulin (DVD-Ig) protein design, with two variable domains linked in tandem in each heavy and light chain. A dual-specific Ig-like molecule can be generated from two different monospecific monoclonal antibodies (mAbs). Both the light chain and heavy chain of a DVD-Ig have two variable domains fused in tandem through a short peptide linkage. The resulting DVD-Ig molecule retains the activities of both parental mAbs

Full size image

The first strategy involves the reengineering of the CH3 domain of the Fc (i) to induce an effective heterodimerization of the two heavy chains and (ii) to differentiate between the two light-chain/heavy-chain interactions. To this end, a large amino acid side chain was introduced into the CH3 domain of one heavy chain that fits into an appropriately designed cavity in the CH3 domain of the other heavy chain (the ‘knobs-into-holes’ [KiH] methodology) [42, 43]. This hinders homodimerization of two identical heavy chains (Fig. 4b2), leading to IgG molecules composed of two different heavy chains. The second objective, the correct pairing of heavy and light chains, was more difficult to achieve. To this end, the light chains have been selected to be identical or not to contribute significantly to antigen binding (the ‘common light chain’ approach) [44] (Fig. 4b3).

A diverse set of bispecific IgG-like antibodies such as the single-chain KiH (Fig. 4b4) were generated by fusion of small bsAb fragments. Similarly, single-chain variable fragment (scFv) moieties have been fused to the N- or C-terminus of the heavy or light chain generating also tetravalent and bispecific molecules such as IgG-scFv (Fig. 4b5) and scFv-IgG (Fig. 4b6) [45, 46].

A further simplification of this strategy was realized by Trubion/Emergent BioSolutions. They fused scFv moieties to the N- and C-terminus of an Fc fragment with the advantage that only a single polypeptide chain needed to be produced (Fig. 4b6). The conservation of the classical IgG architecture, as it was selected during evolution, has many advantages for the therapeutic application of bsAbs. The Fc part is identical to that of a conventional IgG antibody, resulting in IgG-like pharmacokinetic properties and retained effector functions such as ADCC being mediated through FcγRIIIa binding. IgG-like size and molecular weight are expected to result in IgG-like diffusion, tumor penetration, and accumulation. Triple specific IgG-like antibodies were also obtained by fusion of two scFvs to one CH3 domain of an IgG (Fig. 4b8). It was observed to bind to FcRn and FcγR with the affinity of a classic IgG [47].

In the dual variable-domain (DVD) approach (Fig. 4c) developed by Abbott, a second VH domain is fused to the N-terminus of the heavy chain and a complementary second VL domain to the N-terminus of the light chain. Interestingly, it could be shown that this leads to molecules capable of simultaneously binding to two different antigens while maintaining the functions of IgG molecules [48]. The affinity of the two binding sites was, however, influenced by the order of the two specificities and by the linkers connecting the variable domains within each chain [48]. Their clinical evaluation is ongoing (Table 3).

More recently, similar approaches leading to Fc heterodimer formation were established based on an electrostatic steering effect [49] or creating complementary CH3 domains through strand-exchanged engineered domains (SEED) derived from human IgG and IgA CH3 domains [50]. These Fc heterodimers can also be used as fusion partners for scFv fragments, thus completely avoiding the use of light chains.

All these IgG-like bsAb reagents represent excellent candidates for the development of bsAbs. But the successful design and production of such molecules has been limited. Very few of them are being evaluated in clinical trials. The route to manufacturing and downstream process development is still long before these molecules could become mainstream therapeutics.

5 Design of New bsAbs Without an Fc Domain

Advances in genetic engineering also enabled the generation of defined combinations of variable regions of the heavy and light chain domains of antibodies, thus creating a great variety of smaller bsAbs without an Fc domain (Fig. 4a, bottom). Such molecules, in comparison with classical bsAbs, are expected to have low immunogenicity, high tumor penetration ability, and high efficiency because of the physical proximity between lymphocytes and cross-linked targeted tumor cells [51]. The main representative of the first category is the tandem scFv (TaFv) which is obtained by connecting together two scFvs, each scFv being generated by associating the VH and VL regions of a defined mAb by a peptide linker of approximately 15 amino acids. TaFv production is preferentially performed in mammalian cell systems, although expression in bacteria was successful in some cases [52]. Such constructs do not contain an Fc domain, therefore recruitment of accessory cells and associated cytokine release is avoided.

The BiTE (Bispecific T-cell engager) molecules, which are developed by Micromet, correspond to TaFv with two binding specificities, one for CD3, the other one for TA (Fig. 5a). They are based on the concept that simultaneous binding to a tumor-associated antigen on the tumor cell and a trigger molecule on the effector cell leads to site-directed effector-cell activation and consecutive tumor cell killing. Bypassing MHC restriction, T cells are activated by such bsAbs to selectively kill tumor cells (Fig. 5b). What is interesting to note is that a single bsAb delivering only the binding to CD3 on the T cell can induce the two signals 1 and 2 (costimulation) for physiological activation of T cells, and that no other costimulation signals are necessary. In vitro, target-dependent activation of T cells in a costimulatory-independent manner was achieved with high efficiency at low concentrations (picomolar range) and at low effector-to-target cell ratios [53]. CD8+ as well as CD4+ T cells could be activated to participate in tumor cell killing, inducing apoptosis via the perforin/granzyme B mechanism [54]. BiTE bsAbs efficiently induce formation of lytic synapses between CD8+ T cells and human tumor cells that are virtually indistinguishable in composition and subdomain arrangement from synapses induced by regular components of cytotoxic T cell recognition [55]. Epitope distance to the target cell membrane and antigen size were identified as factors that determine the BiTE efficiency [56]. Furthermore, it was shown that BiTEs can induce serial killing by T cells [57]. In vivo, promising anti-tumor effects were shown in xenograft and syngeneic mouse models and even in a non-human primate model [58–60]. Currently a growing set of BiTE molecules retargeting T cells (CD3) to diverse solid (epidermal growth factor receptor [EGFR], HER2, carcinoembryonic antigen [CEA], EpCAM, Eph receptor tyrosine kinase A2 [EphA2], melanoma-associated chondroitin sulfate proteoglycan [MCSP]) positive and leukemic (CD19, CD33) tumors are being tested [61]. Two of them, MT103 (blinatumomab) targeting CD19 and CD3 and MT110 targeting EpCAM and CD3 (Fig. 5b), entered clinical trials (Table 3).

Fig. 5
figure 5

Bispecific T-cell engager (BiTE): structure and mode of action. a Construction of a BiTE from two different monoclonal antibodies (mAbs). The variable domains of the two mAbs recognizing either tumor or T cells are genetically linked to form single-chain variable fragment (scFv). As a result, a single polypeptide chain is produced in which two scFvs are flexibly linked. b Model of a BiTE-induced cytolytic synapse. The BiTE molecule force T cells and epithelial cell adhesion molecule (EpCAM) + tumor cells to come in close contact. The resulting synapse shows all the hallmarks of a synapse formed by T cell receptor–MHC-I peptide-induced synapses. An adhesion ring involving interaction between lymphocyte function-associated antigen 1 (LFA-1) on T cells and intercellular adhesion molecule 1 (ICAM-1) on tumor cells is formed. A formed signaling domain induces a potent activation of T cells leading to expression of adhesion molecules and cytokines and to proliferation. A secretory domain is formed, releasing cytotoxic granule content into the target cell. As a consequence, the tumor cell undergoes programmed cell death and lysis

Full size image

Such bsAbs appear very efficient at triggering T cells. The anti-CD3 scFv of BiTE was also used by us to generate virus antigen-targeting bsAbs.

6 A New Strategy: Targeting bsAbs to Viral Antigens of a Tumor Vaccine

We developed in our laboratory a new type of tumor vaccine (ATV–NDV) for the treatment of cancer patients. It consists of three components: (i) autologous tumor cells (ATV), (ii) Newcastle Disease Virus (NDV) to be used for tumor cell infection, and (iii) bsAb which attach to the viral proteins—hemagglutinin-neuraminidase (HN) and fusion (F) molecules—on the infected tumor cells [62]. The avian virus NDV, when applied to humans, has the interesting ability to replicate selectively in tumor cells [63, 64]. This allows the specific labeling of the tumor cells with viral antigens. These can then be used as anchor molecules on tumor cells for the binding of bispecific or multivalent antibodies during the elaboration of the vaccine (Fig. 6a, [65, 66]).

Fig. 6
figure 6

Combination of a tumor vaccine with bispecific antibodies (bsAbs) binding to the vaccine and triggering immune T cells. a Autologous tumor cell vaccine modified by infection with Newcastle Disease Virus (ATV–NDV) and by attachment of bsAbs (left) and its effect on the T-cell anti-tumor immunity (right). This corresponds to a new strategy of tumor vaccine design. (Left) In the first step, NDV is added to the tumor cells. Virus replication leads to expression of the two viral proteins HN and F at the surface of the tumor cells. In the second step, the bsAbs are added to NDV-infected tumor cells. The constructs bsHN-CD3 and bsF-CD28 bind to the tumor vaccine via the HN and F protein, respectively, expressed at the tumor cell surface. They introduce agonistic antibody activities for CD3 (for bsHN-CD3; signal 1) and for CD28 (for bsF-CD28; signal 2) activation to the tumor vaccine. (Right) Signal 1 is important to activate non-tumor antigen (TA)-specific anti-tumor activity in naive T cells (which are not primed) whereas signal 2 is essential to re-stimulate TA-specific anti-tumor activity in memory T cells (which may be anergized). b Multivalent immunocytokines attached to the ATV–NDV tumor vaccine. Multivalent immunocytokines can bind to the tumor vaccine and to T cells as described in (a). In addition, they can recruit accessory cells of the innate immunity via binding of their cytokine part (IL-2 or GM-CSF) to the corresponding CkR (respectively CD25 [on T or NK T cells] and CD116 [essentially on DCs]). The recruitment of these cells is associated with cytokine release which amplifies the costimulatory signal and leads to an increased anti-tumor activity. ADCC antibody-dependent cell-mediated cytotoxicity, ADCP antibody-dependent cell-mediated phagocytosis, DC dendritic cells, GM-CSF granulocyte-macrophage colony-stimulating factor, HN hemagglutinin-neuraminidase, IL interleukin, NK natural killer

Full size image

Infection of tumor cells by NDV has been observed to lead to an increase in tumor cell immunogenicity [67]. A prospective, randomized, controlled clinical study of post-operative active-specific immunization (ASI) with the tumor vaccine ATV–NDV revealed evidence for long-term survival benefit for colon cancer patients [68]. Further augmentation of T-cell stimulatory capacity of the ATV–NDV vaccine was achieved by attachment of specially designed bsAbs binding to viral HN or F on the infected tumor cells and to CD3 or CD28 on T cells. Such bsAbs were designated as bsHNxCD3 and bsFxCD28. They deliver two signals to T cells: signal 1 through the CD3-associated T-cell receptor complex; and signal 2 as a costimulatory signal via CD28 [65] (Fig. 6a). These signals lead to strong and durable anti-tumor activity in human T lymphocytes [66]. Such T cell signalling appears to be required in particular in cancer patients whose T cells do not properly respond to stimulation via TAs [67, 68]. The optimized vaccine ATV–NDV/bsCD3/bsCD28 appears able to revert unresponsiveness of partially anergized TA-specific T cells. It was also demonstrated to be capable of de novo activation of anti-tumor activity from naive T cells independent of TA recognition [69–71]. Thus, this new type of vaccine has new properties particularly suited to circumvent tumor immune escape mechanisms. This is particularly due to the use of NDV. This virus introduces immunological danger signals into the vaccine such as HN at the tumor cell surface and viral RNA in the cytoplasm [62]. NDV has been reported recently to induce proinflammatory conditions and type I interferon and appears able to counteract Treg activity [72].

In recent years, we have extended our strategy to design multivalent antibodies containing additional binding sites in order to improve the recruitment of accessory cells against cancer. We avoided the use of an Fc domain because of the complexity of the Fc/FcR system. Instead, we cloned the genes coding for the human cytokines interleukin (IL)-2 and granulocyte-macrophage colony-stimulating factor (GM-CSF) between the two antibody binding sites (Fig. 6b). A strong potentiation of the T-cell stimulatory activity of the ATV–NDV vaccine was observed upon attachment of trispecific fusion proteins anti-NDV/cytokine/anti-CD28 compared with control bsAbs without a cytokine. These trispecific immunocytokines bind with one arm to viral HN molecules of the vaccine, and with the other arm to CD28 (costimulatory signal 2a) and CD25 (IL-2 receptor; costimulatory signal 2b) on T cells [73]. For example, the incorporation of the GM-CSF into the bsAb was shown to activate monocytes and to confer to them anti-tumor activity in vitro [74].

The use of multivalent immunocytokines to redirect T cells and accessory cells (i.e. NK cells and DCs) against tumor cells is an alternative to the use of an Fc domain. Their analysis revealed significant increase in functional activity. Their use should reduce risks of systemic toxicity linked to uncontrolled cytokine release.

7 Conclusions and Future Prospects

Compared with standard therapies, the targeted activation of the patient’s immune system offers many advantages. One important aspect is the induction of systemic protective anti-tumor immunity associated with long-term immunological specific memory [75].

Despite recent success in the commercial development of antibody therapeutics, current single target mAbs have demonstrated limited efficacy in the clinic. This is because T cells (which are FcR-negative) are not recruited as effector cells and also because tumor cells are heterogeneous and endowed with capacities to escape from immune monotherapy. As an example, in many primary tumor cells, and more frequently in metastatic cells, the expression of MHC-I molecules is either lost or suppressed [76]. The use of NDV and bsAbs targeting several pathways simultaneously may help to avoid such resistance mechanisms. The combination of tumor cell-targeted, T-cell-dependent and non-MHC-restricted killing activity exerted by accessory cells appears an interesting strategy to circumvent tumor immune escape mechanisms and to establish polyclonal immune memory.

In this review, we focused on bsAbs that can engage immune effector T cells by bridging them to tumor cells. We summarized how such bsAbs may be equiped with further binding domains from either immunoglobulin or from cytokines to recruit further effector cells to the tumor.

The first bsAbs, which were obtained by somatic hybridization, contained an Fc domain. They demonstrated potent tumor cell killing in vitro and in animal studies. However, despite some reported responses, early clinical trials faced several problems which were mainly attributed to the presence of the Fc region in the antibody construct. One such problem was a severe toxicity due to the release of inflammatory cytokines (cytokine storm). After initial years of disappointment, improvements were made by modifications of the quadroma strategy and by the use of the recombinant antibody technology. Today we see clinical trials with promising results, the most impressive ones being delivered by Triomab and BiTE. bsAbs appear as very powerful anti-tumor reagents. They can directly and efficiently retarget T cells to fight cancer. This is of primary importance since cytotoxic abilities of immune effector cells like cytotoxic T lymphocytes (CTLs) are often impaired in patients with advanced cancer as a result of suppressive factors present within the tumor environment [77].

bsAbs may be further improved since they can be modulated and adjusted through genetic engineering. The experience gained with the modification of the Fc domain of mAbs using protein as well as glycosylation strategies ([78], Table 4), was shown to be difficult but could be used theorically for the design of more functional bsAbs for cancer therapy. The approval of the trispecific Catumaxomab (EpCAM × CD3) in 2009 in the European Union for the intraperitoneal treatment of malignant ascities in patients with EpCAM-positive carcinomas, in circumstances where standard therapy is not available or no longer feasible, constitutes a strong signal for the inclusion of an Fc fragment in the bsAb molecule.

Table 4 Strategies for the modulation of constant fragment (Fc) activity via protein engineering and glycosylation
Full size table

However, the development of bsAbs containing an Fc domain face many problems, such as glycosylation issues, FcR polymorphism, and inhibitory Fc receptor-mediated signals. The potential devastating effects associated with non-specific lymphocyte activation have to be kept in mind. These have been clearly illustrated in the clinical trial studying TGN1412, a humanized anti-CD28 superantagonistic mAb uniquely designed to activate T cells without the need for concomitant TCR engagement [79]. Immediately after intravenous infusion with TGN1412, all six study participants experienced severe cytokine storm and were admitted to intensive care for multisystem organ failure. While multiple factors may have ultimately contributed to these outcomes, the magnitude of systemic toxicities observed in this setting reaffirms the critical importance of maintaining targeted specificity [80]. The overall challenge, then, is to develop safe and more effective anti-tumor therapies.

From the data obtained in our laboratory, bsAbs appear very appealing in combination with the ATV–NDV vaccine. These reagents showed a high safety profile [68, 81, 82]. The attachment of bsAbs allows the delivery of stronger activation signals for T cells, and the further incorporation of cytokines such as IL-2 and GM-CSF into the bsAb constructs increases in vitro anti-tumor activity [74]. It is crucial to note that glycosylation of both cytokines is not important for functional activity since both of them are not glycosylated. This is of additional advantage in comparison with Fc domains, where glycosylation plays an important role for function.

In conclusion, trispecific immunocytokines appear as very promising molecules for efficiently targeting the immune system against cancer by avoiding the biological complexity and potential risks of the Fc/FcR system.