Antibody-cytokine fusion proteins: A novel class of biopharmaceuticals for the therapy of cancer and of chronic inflammation.

Cytokines are small immunoregulatory proteins, which are secreted by leukocytes and, in some cases, by endothelial cells, fibroblasts and stromal cells. Cytokines can induce important functional changes in tissues, such as an increase in vascular permeability or alteration of body temperature. Importantly, these proteins can also have inhibitory or activating effects on immune cells, depending on their features, on their concentration and on the environment in which they act [ 1 ]. Considering the important role in many physiological and pathological conditions, cytokines have been considered as products in their own right or as targets for the development of blocking agents. The antibody-based blockade of pro-inflammatory cytokines [e.g., tumor necrosis factor (TNF), interleukin-2 (IL2), interleukin-12 (IL12), interleukin-17 (IL17)] or their cognate receptors [e.g., the interleukin-6 receptor (IL6R)] has led to the development of successful products for the treatment of chronic inflammatory conditions, such as rheumatoid arthritis, inflammatory bowel diseases and psoriasis. On the other hand, certain recombinant cytokines have received marketing authorization for the treatment of cancer (e.g., IL2, TNF), of viral infections [e.g., Interferon (IFN) α], and for few other biomedical applications. The systemic administration of pro-inflammatory cytokines can lead to severe off-target related adverse effects, which may limit the dose and prevent escalation to therapeutically active regimens. Certain cytokine products (e.g., IL2, TNF, IL12) have exhibited recommended doses in the single-digit milligram range or even below [ 2 – 4 ]. Adverse effects associated with the intravenous administration of pro-inflammatory cytokines may include hypotension, fever, nausea or flu-like symptoms, but cytokines may occasionally also cause serious hematologic, endocrine, autoimmune or neurologic events [ 5 ]. In view of these considerations, there is a clear biomedical need for the development of next-generation cytokine products, which are better tolerated and which display a preferential action at the site of disease, helping spare normal tissues. Antibodies specific to accessible markers, which are over-expressed at the site of disease, may represent ideal “vehicles” for the targeted delivery of therapeutic payloads, including cytokines. In mice, it has been clearly shown that certain tumor-homing antibody-cytokine fusion proteins (“immunocytokines”) can dramatically increase the therapeutic index of the corresponding cytokine payload [ 6 – 12 ]. Similarly, immunocytokines with anti-inflammatory properties, capable of selective accumulation at sites of tissue remodeling, have been considered for the treatment of chronic inflammatory conditions and of endometriosis [ 13 , 14 ]. In this review, we survey some basic concepts associated with the development of immunocytokine biopharmaceuticals and discuss how engineered cytokine products may look like in the future.

Ideal target antigens for the development of immunocytokine products should be abundantly expressed at the site of disease and absent in healthy tissues. In tumors and in chronic inflammatory conditions, tissue remodeling and neo-vascularization processes expose antigens, which are otherwise virtually undetectable in healthy organs [ 33 – 35 ]. One example is represented by splice isoforms of fibronectin, a glycoprotein of the extracellular matrix (ECM). The extra-domains A and B (EDA and EDB) of fibronectin are strongly expressed in tumors, at sites of tissue remodeling and during fetal development, but are otherwise not found in normal tissues, exception made for the female reproductive system [ 36 , 37 ]. Similarly, splice variants of tenascin-C are specifically found in tissues and tumors undergoing neo-angiogenesis, in a process which is regulated by intracellular pH [ 38 ]. EDA, EDB and splice variants of tenascin-C represent suitable targets for the delivery of bioactive payloads, including cytokines. In oncological malignancies molecular targets may include fibroblast activation protein (FAP) [ 39 , 40 ], cellular antigens (e.g., CEA and PSMA) [ 41 – 43 ] or proteins, which become accessible in necrotic lesions, such as histones [ 44 ]. Antibodies which have been extensively characterized in the context of cytokine fusions include F8 (targeting EDA-fibronectin) [ 45 ], L19 (targeting EDB-fibronectin) [ 46 ], F16 (targeting the A1 domain of tenascin-C) [ 47 ], scFv36 (targeting FAP) [ 40 ], hu14.18 (targeting the GD2 ganglioside) [ 48 ], chCLL-1 (targeting CD20) [ 49 ] and anti-HER2/ neu [ 50 ]. In tissues undergoing chronic inflammatory processes, specific cell subtypes that are present at higher local density than in normal tissues, may be targeted (e.g., neutrophils and macrophages) [ 51 ]. CD64 (i.e., the high-affinity Fcγ receptor 1) is expressed on the surface of macrophages, monocytes and their progenitor cells. Immunotoxins directed against CD64 were tested in animal models of inflammation [skin, rheumatoid arthritis and ischemia-induced kidney injury] and did not display adverse effects related to the presence of the antigen on other circulating leukocytes [ 52 ].

Both for cancer therapy as well as for the treatment of chronic inflammation, several cytokine payloads have been developed and tested in preclinical trials. Pro-inflammatory cytokines such as IL2, TNF, and IL12 have been investigated for tumor therapy, as they have been found to increase and activate the local infiltrate of leukocytes at the tumor site [ 53 – 59 ]. By contrast, immunosuppressive cytokines (e.g., IL10) may be considered as payloads for the treatment of chronic inflammatory conditions [ 60 , 61 ] or of other diseases (e.g., endometriosis) [ 62 , 63 ]. The first antibody-cytokine fusions were reported by the groups of Reisfeld/Gillies, Alan Epstein and Sherie Morrison [ 48 , 64 , 65 ]. The authors described the fusions of IgG antibodies with various interleukins, interferons and members of the TNF superfamily [ 48 – 50 , 64 , 66 – 69 ]. Our group has mainly focused on the development of antibody-cytokine fusion proteins, which targeted extracellular matrix components (e.g, those based on the F8 and L19 antibodies). Quantitative biodistribution studies using radiolabelled immunocytokine preparations have allowed us to learn some lessons about payloads that can be efficiently delivered to solid tumors, as well as payloads which abrogate the targeting properties of the parental antibody. Payloads can broadly be grouped into five distinct categories in the mouse: i) Cytokines, which are specifically delivered to the tumor mass, by fusion to the antibody moiety (e.g., IL2, TNF, IL4, IL6, IL10, IFNα) [ 14 , 53 , 70 – 72 ]. ii) Cytokines, which are specifically delivered to the tumor mass, by fusion to the antibody moiety in a specific format (IL12) [ 29 ]. iii) Cytokines which can be efficiently delivered to the tumor site only at higher doses, after saturation of the cognate receptor (e.g., IFN γ, GM-CSF) [ 73 , 74 ]. iv) Payloads, which are too large or too negatively or positively charged to be able to extravasate and reach the target antigen in vivo (e.g., VEGF(120) vs. VEGF(164)) [ 75 ]. v) Cytokines, which are heavily glycosylated and are captured by asialoglycoprotein receptor in the liver, which abrogates the tumor-homing performance of the antibody moiety (e.g., B7.2, IL9) [ 76 , 77 ]. Cytokines, which are specifically delivered to the tumor mass, by fusion to the antibody moiety (e.g., IL2, TNF, IL4, IL6, IL10, IFNα) [ 14 , 53 , 70 – 72 ]. Cytokines, which are specifically delivered to the tumor mass, by fusion to the antibody moiety in a specific format (IL12) [ 29 ]. Cytokines which can be efficiently delivered to the tumor site only at higher doses, after saturation of the cognate receptor (e.g., IFN γ, GM-CSF) [ 73 , 74 ]. Payloads, which are too large or too negatively or positively charged to be able to extravasate and reach the target antigen in vivo (e.g., VEGF(120) vs. VEGF(164)) [ 75 ]. Cytokines, which are heavily glycosylated and are captured by asialoglycoprotein receptor in the liver, which abrogates the tumor-homing performance of the antibody moiety (e.g., B7.2, IL9) [ 76 , 77 ]. Our group has observed that glycosylation patterns may vary depending on the protein production protocol used (e.g., stable vs. transient gene expression), leading to striking differences in disease-targeting performance [ 77 ].

A large variety of immunocytokines has been generated for multiple therapeutic applications in the past two decades. Even though only a few of those proceeded to clinical stage so far, the potential of antibody-cytokine fusion proteins is enormous. Immunocytokines may synergize with several combination partners, such as cytotoxic drugs [ 26 , 44 , 53 , 71 , 107 – 109 , 186 ], radiation [ 44 , 96 , 120 , 187 ], monoclonal antibodies [ 44 , 99 , 111 ], SMDCs [ 119 ], antibody drug conjugates [ 188 , 189 ], cancer vaccines [ 159 , 161 ], immunecheckpoint inhibitors [ 39 , 56 , 57 , 134 , 190 , 191 ], bispecific antibodies [ 192 – 194 ], and other immunocytokines[ 17 , 57 , 114 , 134 , 160 – 162 , 195 , 196 ]. Over the next few years, the outcome of on-going clinical trials will shed light on the therapeutic benefit that can be achieved, while new protein engineering approaches will contribute to the development of second-generation products.

Antibodies can be used in full immunoglobulin G (IgG) format, in order to exploit the stability and long circulatory half-life of these products. Alternatively, antibody fragments [e.g., single-chain variable fragments (scFvs), diabodies, Fab fragments] may be considered, when a faster blood clearance and a more efficient extravasation is desired [ 15 , 16 ]. Figure 1 illustrates some of the most popular antibody formats, which have been considered for immunocytokine development. The format may determine different pharmacokinetic and pharmacodynamic properties. These in vivo functional aspects may not be evident on the basis of simple in vitro assays, since the fusion of cytokine payloads with various types of antibodies often proceeds without loss of biological activity. On the other hand, the residence time of immunocytokines in blood and the ability of the product to extravasate and localize at the site of disease are crucial determinants of pharmaceutical performance [ 17 ]. Intact IgG molecules can be fused to cytokines at various sites (N- or C-terminus of heavy or light chain), giving rise to large fusion proteins (e.g., 180 kDa for IL2 fusions). The size and recycling properties mediated by the interaction with the neonatal FcRn receptor may contribute to a longer circulatory half-life in vivo [ 18 ]. As a potential drawback, a larger molecular size may impair extravasation and penetration into the tumor mass [ 19 ]. In addition to the bioactivity of the cytokine moiety, an IgG-based immunocytokine would retain the ability to bind to Fc gamma receptors on leukocytes, with the potential to impair localization at the tumor site and to mediate undesired activation of white blood cells. Changing the amino acid composition, the isotype or the glycans of the Fc portion may reduce these off target effects [ 20 , 21 ]. Immunocytokines based on small antibody fragments have a much shorter half-life in the circulation, compared to those based on the full IgG format. A rapid blood clearance may lead to a decreased uptake at the site of disease, since extravasation typically represents the rate-limiting step for ligand-based pharmacodelivery applications [ 15 , 22 ]. Short linkers (less than 11 aminoacids) between the variable heavy (VH) and variable light (VL) chains form non-covalent homodimers (diabodies) [ 23 ]. Diabodies and other bivalent formats fused to the cytokine payloads retain a high binding avidity to the cognate antigen, thus potentially leading to a long residence time on the biological target [ 16 , 24 , 25 ]. Multivalent immunocytokines based on antibody fragments have exhibited tumor:organ ratios greater than 10:1 in biodistribution studies performed in mouse models of cancer, 24 hours after intravenous administration [ 11 , 26 ]. The antibody format and the position of the cytokine payload, as well as the amino acid composition of the linker connecting various protein domains, can substantially influence the in vivo performance of fusion proteins, thus offering rich opportunities for protein engineering applications [ 16 , 27 ]. Design options further increase, when cytokines consisting of multiple subunits (e.g., components of the TNF or of the IL12 superfamily) are considered [ 7 , 17 , 26 , 28 – 32 ] ( Figure 2 ).

The in vivo activity of cytokines and of their derivatives depends on the immunological environment and on the concentration of the cytokine at the site of disease. Certain payloads may display a pro- or anti-inflammatory action depending on the immunological context [ 164 – 166 ]. For example, IL12 caused disease worsening and increased TNF production in the collagen-induced model of arthritis, when the product was given at low doses (5ng/day). By contrast, higher doses of (500ng/day) suppressed disease development inducing production of the anti-inflammatory cytokine IL10 and corticosterone [ 167 ]. More examples, reviewed by Cavaillon JM (e.g., IL6, IL4, IL10, TGFβ), indicate the dependence on the responding cell, the timing and even the experimental model [ 165 ]. Even though research has mainly focused on immunocytokines directed against oncological targets, potential applications in chronic inflammatory diseases have recently gained momentum. One product based on IL10 (F8-IL10) is currently being investigated in Phase II clinical trials in patients with rheumatoid arthritis or ulterative colitis, while other products have shown promising signs of efficacy in animal models [ 14 , 168 ]. IL10 is a homodimeric, pleiotropic cytokine, with a potential to display an immunosuppressive activity at sites of inflammation, for example by reducing antigen presentation and the release of pro-inflammatory mediators [ 169 ]. Recombinant IL10 has been tested in various mouse models of inflammation showing encouraging results [ 60 , 61 ]. Initial clinical studies have shown no significant differences in remission rates or disease improvements were reported [ 170 , 171 ]. However, a pegylated version of murine IL10 induced rejection of solid and metastasizing tumors in mice by induction of cytotoxic T cells [ 172 , 173 ]. In clinical trials the recombinant pegylated interleukin-10 (PEG-rIL-10) increased the density of activated intratumoral cytotoxic T cells in patients [ 174 ]. PEG-rIL-10 is currently being investigated in combination with immunecheckpoint inhibitors [anti-PD-1 antibodies Nivolumab ( NCT03382912 ) and Permbrolizumab ( NCT03382899 )] or with chemotherapy ( NCT02923921 ). The antibody-cytokine fusion L19-IL10 selectively localized to sites of inflammation in mouse models [ 13 ]. L19-IL10 features the L19 antibody (specific to EDB fibronectin) in diabody format, fused to human IL10. In the mouse model of collagen induced arthritis (CIA), L19-IL10 treatment induced a significant inhibition of disease progression, compared to control mice treated with saline or with an IL10 fusion based on an antibody of irrelevant specificity [ 13 ]. Similarly, fusion of IL10 with the F8 antibody (specific to EDA fibronectin) [ 175 ] showed disease targeting and an encouraging activity in the murine CIA model [ 175 , 176 ], both, alone or in combination with methotrexate or a TNF inhibitor. The product was also active in a mouse model of endometriosis [ 63 ]. The fully human F8-IL10 fusion protein (Dekavil) was tested in a Phase Ib clinical trial against rheumatoid arthritis in combination with methotrexate ( NCT02076659 ). 15 out of 23 treated patients experienced therapeutic benefit, including 2 long-lasting remissions [ 177 ] and is now being investigated in multiple Phase II clinical trials ( NCT02270632 , EudraCT number 2013-005418-37). Additionally, F8-IL10 is tested in a Phase II clinical trial as add-on therapy to infliximab in patients with ulcerative colitis (EudraCT number 2017-002108-28). In another approach, viral IL10, which is considered less immunostimulatory, was fused to an antibody fragment targeting ROS-CII (1-11E). Treatment of arthritic mice with 1-11E/IL10 inhibited disease progression and lowered pro-inflammatory cytokines serum levels [ 178 , 179 ]. Immunomodulatory products can also be produced by antibody fusion with IL4, a so-called “prototypic immunoregulatory cytokine” [ 180 ], able to interact with various target cells. IL4 production is limited to a small subset of hematopoietic cells and the effects are mediated by a high affinity heterodimeric IL4 receptor. Recombinant IL4 reduced disease progression in preclinical models of rheumatoid arthritis [ 181 ], but failed to reproduce these results in human clinical studies[ 182 ]. By contrast, recombinant IL4 switched the immune environment to a T H 2 pattern and markedly improved conditions in patients with moderate-to-severe T H 1 mediated psoriasis [ 182 ]. The antibody-based delivery of IL4 at the site of chronic inflammation is thought to generate a more specific and focused immunomodulatory action. The F8-IL4 fusion protein (Tetravil), based on the EDA targeting antibody in diabody format, selectively accumulated in the proximity of newly formed blood vessels in arthritic paws and toes of mice with CIA. Therapy studies in CIA mice induced 100% complete remission of the disease when F8-IL4 was given together with dexamethasone [ 14 ]. The combination therapy rapidly resolved the inflammatory process in the paws, by recruiting TH2 cells and Tregs and by normalizing the concentration of pro-inflammatory cytokines [ 14 , 168 , 183 ]. In addition, treatment with F8-IL4 significantly inhibited endometriotic lesion growth in a mouse model, compared to untargeted IL4, which did not exhibit in vivo activity [ 62 ]. The therapeutic effect of F8-IL4 was also observed in imiquimod-induced and contact-hypersensitivity-induced mouse models of skin inflammation [ 184 ]. Motivated by these encouraging data, F8-IL4 is currently being considered for clinical testing for the treatment of rheumatoid arthritis and endometriosis [ 185 ].