Taming the beast: engineering strategies and biomedical potential of antibody-based cytokine mimetics

KEYWORDS:

• Antibody
• antibody engineering
• bispecific antibody
• cytokine
• cytokine mimetic
• single domain antibody
• surrogate agonist
• VHH

1. Introduction

Cytokines [1–3] are potent immunoregulatory proteins that have proven to be of relevance as therapeutic modalities. In this regard, cytokines have been approved for the treatment of a wide range of diseases such as anemia, multiple sclerosis, sarcoma, melanoma, or chemotherapy-induced disorders [4]. However, since decades, no regulatory approvals were made for new cytokine therapeutics. This can be at least partially explained by the complex biology involving cellular and functional pleiotropy, as elegantly reviewed by Garcia and colleagues [5]. Moreover, dose-limiting toxicities when administered systemically due, in large part, to downstream cytokine expression and release (i.e. cytokine cascades) clearly restrict the general applicability of cytokines for disease treatment [6]. In addition, the short half-life and unfavorable biodistribution [7] combined with an often poor manufacturability of wild-type cytokines contributes to their ‘non-drug-like’ properties.

To address the main limitations of cytokines, most importantly the narrow therapeutic window, different strategies were developed to spatiotemporally regulate cytokine functions [8]. In this respect, Philogen S.p.A. recently reported that their investigational drug Nidlegy™ (Daromun) met the study`s primary objective in a Phase III trial (PIVOTAL, NCT02938299) in improving the Recurrence-Free Survival in patients with locally advanced fully resectable melanoma as neoadjuvant treatment. Daromun is a combination of two intralesionally administered immunocytokines (IL-2 and TNF-α each fused to an antibody-derived targeting moiety addressing the alternatively spliced extra-domain B of fibronectin) [9,10]. Besides intralesional administration, systemic administration of antibody-cytokine fusions is also feasible [11,12]. Other interesting examples of next-generation-cytokines exploit mutant versions of a particular cytokine in order to tailor a specific function [5,7,13–16].

An alternative to engineering the cytokine itself toward an intended biology relies on harnessing multifunctional antibody derivatives (or scaffold proteins) that mimic the function of a given cytokine by agonizing the cognate cytokine receptor. Those (typically) bispecific or multispecific antibodies, referred to as cytokine mimetics or surrogate agonists emerged as very promising tools to tailor cytokine functionalities in manifold ways. In 2015, the group of Garcia exploited erythropoietin receptor (EpoR) targeting diabodies to re-orient the receptor dimer geometry and consequently to tune cytokine receptor signaling (Table 1). Diabodies are bivalent antibody fragments composed of VH and VL domains of the same or different antibodies that assemble in a head-to-tail orientation [22,23]. Intriguingly, different EpoR-specific paratopes elicited differential receptor agonism profiles [17]. In 2019, the same group generated a high affinity designed ankyrin repeat protein (DARPin [24]) targeting the EpoR (Table 1) [18]. By further applying protein design and engineering involving a homodimeric DARPin structure, as well as ankyrin repeat spacers, the group was able to really fine-tune EpoR downstream signaling ranging from full over biased to partial agonism, both, in terms of maximal signaling as well as potencies. Importantly, different design architectures for instance with respect to paratope distance provoked unique ramifications regarding differentiation and proliferation of hematopoietic stem and progenitor cells.

Table 1. Selected examples of cytokines that were mimicked by exploiting multifunctional antibodies or scaffold proteins.

Cytokine	Class of mimetic
EPO	diabody [17] DARPin [18]
IL-2	sdAb [19] VHH [20]
IL-18	VHH [21]
IL-2/IL-10 chimera	VHH [20]
Type I IFN	VHH [20]

Due to their high flexibility concerning reformatting options for the construction of bispecific and multispecific antibody designs [25], ease of generation [26–28] and humanization [29–32], single domain antibodies (sdAbs) such as camelid-derived variable domains of the heavy chain of heavy chain-only antibodies (VHHs) are extraordinary building blocks for the engineering of cytokine mimetics. Harris et al. engineered a biased bispecific sdAb-based surrogate agonist for the IL-2 receptor (IL-2 R) [19]. IL-2 is an immunomodulatory protein with pleiotropic functions, driving proliferation and survival of NK cells, B cells and T cells as well as differentiation of T cell subsets either in a positive or negative manner [33,34]. The functional form of the IL-2 R comes in two different flavors, either as intermediate affinity receptor for IL-2 composed of the IL-2 Rβ (CD122) and IL-2 Rγ (CD132) that is primarily expressed on resting NK and CD8⁺ T cells or as trimeric high affinity variant containing IL-2 Rβ, IL-2 Rγ as well as IL-2 Rα (CD25). The high affinity trimeric receptor is constitutively expressed on regulatory T cells (Tregs) but also on other lymphocytes following activation. Harris et al. engineered IL-2 mimetics by generating sdAbs targeting the IL-2 Rβ and IL-2 Rγ subunits but sparing IL-2 Rα. Hence, despite robust T cell and NK cell activation, this Fc comprising bispecific antibody (bsAb) circumvented the natural IL-2 bias for preferential Treg activation (Table 1) [19].

This strategy was recently further refined by Garcia and coworkers [20]. The group generated sdAbs against IL-2 Rβ and IL-2 Rγ and combinatorically assembled multiple different VHH-based paratopes in a beads-on-string manner (Table 1). Thereby, surrogate agonists with different degrees of IL-2 R agonism were identified, ranging from minimal agonism over partial and full agonism to super agonism, as measured by STAT5 phosphorylation as proximal activation marker. Even more importantly, distinct surrogate agonists were biased in their ability to elicit receptor downstream signaling. In this matter, the authors identified bsAbs triggering differential phosphorylation of STAT5 and AKT in direct comparison to the wild-type cytokine resulting in significant functional differences in a CD8+ T cell differentiation assay i.e. an effector memory bias, an effector/central memory bias or a central memory bias.

In the same work, the group broadened the applicability of the cytokine mimetic approach by generating surrogate agonists for type I interferons (IFN) by engineering bsAbs targeting the type I IFN receptor composed of the IFNAR1 and IFNAR2 subunits (Table 1) [20]. Identified bsAbs were partial agonists displaying a signaling bias in terms of pSTAT activation compared with IFN-ω. Fascinatingly, those partial agonists were similarly potent as natural IFN-ω in inhibiting Sendai virus replication and also potently inhibited SARS-CoV-2 replication. Decisively, constructed cytokine mimetics functionally decoupled antiviral functions from inflammation showing a reduced ability to trigger a proinflammatory gene expression. Essentially, this can be envisioned to translate into an improved safety profile in terms of inflammation when administered into patients [35,36].

Intriguingly, Garcia and colleagues also engineered novel cytokine specificities, i.e. bispecific cytokine mimetics that trigger a functional response by targeting a heterodimeric receptor which does not exist in nature [20]. This was achieved by means of combining IL-2 Rβ-specific VHHs with sdAbs targeting IL-10 Rβ (CDw210b). These artificial surrogate agonists were able to activate STAT5 but not STAT3 and elicited CD8+ T cell proliferation as well as NK cell cytotoxicity in functional assays (Table 1).

Finally, our group was able to generate sdAb-based cytokine mimetics for IL-18, another proinflammatory cytokine with therapeutic relevance (Table 1) [21]. Initially, surrogate agonists were constructed in a monovalent fashion for IL-18 Rα (CD218a) as well as for IL-18 Rβ (CD218b) by employing a Fc heterodimerization technique and fusing the respective sdAb to the hinge region of each chain. Generated bsAbs triggered a dose-dependent activation of IL-18 R signaling, as well as IFN-γ release (in combination with low-dose IL-12), but were moderate agonists in direct comparison to wild-type IL-18. By modifying paratope valencies as well as the spatial orientation of individual paratopes within the overall design architecture we showed that potencies as well as the magnitude of receptor activation, and hence, IFN-γ release could be significantly augmented, resulting in different flavors of IL-18 mimetics. Most importantly, those surrogate agonists were resistant to IL-18 binding protein (IL-18BP) inhibition, which is a natural decoy receptor for IL-18 and is overexpressed in certain types of tumors [37].

2. Expert opinion

Cytokine mimetics have emerged as versatile entities for immunomodulation. The modularity of such building blocks i.e. the fact that surrogate agonists are highly engineerable in terms of both, antibody architecture, as well as paratope valency enables a tailor-made exploitation of the cognate cytokine receptor biology. Although it is known since several decades that antibodies can agonize receptors [38,39], most of what we know today about antibody-based (or scaffold-based) surrogate agonists has been unveiled by the group of Christopher Garcia within recent years. Garcia and coworkers elegantly revealed by harnessing diabodies as well as DARPins that different receptor geometries can be induced having substantial ramifications with regards to receptor agonism (in terms of magnitude of agonism and potency) and equally important also with respect to receptor downstream signaling bias [17,18,20]. Essentially, this highly tunable receptor biology can be achieved in different ways such as employing epitopic diversity of incorporated paratopes, paratope valencies as well as the overall design architecture involving paratope orientation and linker length.

It is obvious that this might open up new avenues regarding biomedical applications. For instance, biased cytokine mimetics have been described for type I IFNs [20]. These agonists elicited very potent virus neutralization but triggered significantly less proinflammatory effects compared with wild-type IFN-ω. As such, those biased surrogate agonists might be promising antiviral agents for therapy harboring a beneficial safety profile [35,36,40].

In another example, our group generated cytokine mimetics for IL-18 [21]. Since from a sequence as well as structural perspective, those surrogate agonists share very little to no sequence similarity with wild-type IL-18, engineered mimetics were resistant against the natural inhibitor, IL-18BP. While decoy receptor resistant mutant versions of IL-18 have also been engineered (and in fact are actually scrutinized in clinical trials), the feature of being completely different from the wild-type cytokine might pose a substantial benefit for diseases that are associated with autoantibodies against cytokines [41,42]. Here, the anti-cytokine autoantibody response is most likely poly-clonal and the epitopic diversity might be patient-individual (at least to a certain extent). Hence, for selected cytokines, if possible at all, huge engineering efforts would be needed in order to generate a mutant version of the natural cytokine which is resistant to this antibody response. On the other hand, although it was shown that it is possible to humanize camelid-derived sdAbs in a sufficient manner leading to marketing approval [43–45], the risk of immunogenicity or preexisting anti-drug antibodies binding to neoepitopes [46] cannot be fully excluded for surrogate agonists.

Apart from this, the high modularity by utilizing bispecific or multispecific antibody derivatives allows for the engineering of cytokine-like functions with novel specificities. This was again shown by the group of Garcia [20]. They generated an IL-2-like mimetic by targeting IL-2 Rβ and IL-10 Rβ and consequently without the need for IL-2 Rγ. Albeit their molecules did not signal via STAT3 activation (IL-10 R response) but STAT5 (IL-2 R signaling), these findings are giving clear evidence that cytokine-like entities can be tailor-made in a way that allows for targeting a specific cell population in a more selective manner. Beyond, this might pave the way for combining functional features of different cytokines in a single entity.

Another to this date theoretical advantage is the epitopic diversity of surrogate agonists. This might facilitate to target different epitopes of the receptor (subunits) compared to the corresponding cytokine. Accordingly, the natural cytokine:receptor axis would be kept unaffected, a feature that might become relevant for specific cytokines given their pleiotropic modes of action.

Ultimately, although the authors of this perspective are not aware of any ongoing clinical trial involving antibody-based cytokine mimetics, we believe that it is most likely a matter of time until this novel class of engineered proteins will migrate into clinical development, given the evidence of their huge potential as therapeutics.

Declaration of interest

SK, LP and SZ are affiliated with Merck Healthcare KGaA, a company developing and commercializing drugs. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosure

One peer reviewer received honoraria for advisory board work from Philogen S.p.A. Peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Additional information

Funding

This paper was not funded.