Outflanking RANK with a Designer Antagonist Cytokine

See allHide authors and affiliations

Science Signaling  19 Aug 2014:
Vol. 7, Issue 339, pp. pe20
DOI: 10.1126/scisignal.2005674


Members of the tumor necrosis factor (TNF) superfamily of cytokines are noncovalently linked trimers that play important roles in regulating the immune system and have emerged as successful therapeutic targets in various rheumatic and autoimmune conditions. Traditionally, antibodies to cytokines or receptor-Fc fusion proteins have been used to block signaling by TNF family cytokines. In this issue of Science Signaling, Warren et al. have taken a new approach to blocking the action of the TNF superfamily member RANKL [receptor activator of nuclear factor κB (RANK) ligand], which plays an important role in regulating bone turnover through stimulation of its receptor RANK on osteoclasts. Beginning with a single-chain fusion protein of three RANKL subunits, the authors used directed mutagenesis to generate a trimer consisting of a nonreceptor binding subunit fused to two “super-agonist” subunits that have increased affinity for RANK. This molecule antagonized the osteoclastogenic activity of wild-type RANKL in vitro and in vivo, thus providing insights into RANK signaling and a paradigm for the development of other antagonists of TNF family cytokines.

The tumor necrosis factor (TNF) superfamily consists of 19 structurally related cytokines that mediate important functions, such as lymphocyte costimulation, inflammation, and induction of cell death (1). Blocking TNF family cytokines is a successful therapeutic strategy in rheumatic and autoimmune diseases (2). Warren et al. have generated a designer inhibitory cytokine against the TNF family member RANKL [receptor activator of nuclear factor κB (RANK) ligand] (3). Their method outlines a new strategy for inhibiting TNF family cytokines, and their findings have interesting implications for the mechanism of signaling by RANK and related TNF family receptors. Most current therapeutics targeting TNF family cytokines are either monoclonal antibodies against cytokines or Fc-receptor fusion proteins. These agents prevent the interaction between the targeted cytokine and its receptors. Because a number of TNF family cytokines bind to more than one receptor and may mediate disparate functions through these distinct receptors, more specific therapies are needed. In addition, monoclonal antibody–mediated therapies, particularly those that deplete target-bearing cells, may have undesirably long half-lives and do not permit easy ways to reverse their actions in case serious adverse events occur during therapy.

The strategy used by Warren et al. to block RANKL relies on the particular biology of TNF family cytokines, which are composed of noncovalently linked trimers of cytokine subunits, which bring together three receptors into the minimal signaling complex (4). Two interesting concepts have been integrated by Warren et al. to make their cytokine antagonist (3). First, the authors generated a single-chain RANKL encoding three monomers of the cytokine linked by flexible linkers. Second, the authors applied the idea that mutant TNF subunits can interfere with the activity of wild-type TNF subunits, resulting in heterotrimers with antagonistic function. Single-chain variants of other members of the TNF superfamily were previously described (5) and have been used to investigate the structural basis for the binding of TNF by its receptors (6). The concept that dominant-negative TNF mutants can be used to block TNF activity as a potential therapeutic approach was first described by Steed et al., who engineered a nonreceptor-binding version of TNF and allowed this subunit to intermix with trimers of soluble, wild-type TNF (7). Dominant-negative TNF variants formed inactive heterotrimers with wild-type TNF, thus preventing the formation of active TNF homotrimers. As a result, TNF signaling was abolished, and TNF-mediated pathology was attenuated in mouse models.

Although RANKL was originally identified as a T cell–derived cytokine that promotes the functions of antigen-presenting cells (8), it was subsequently found to play diverse roles outside the immune system, regulating bone turnover, mammary gland formation, and thermoregulation in the brain through its signaling receptor RANK. The effects of RANKL are modulated by osteoprotegerin (OPG), a soluble decoy receptor encoded by a distinct gene (9). RANKL produced by osteoblasts promotes the resorption of bone by osteoclasts, promoting bony destruction during inflammatory arthritis and in malignant bone metastases. Deficiency in RANK or RANKL or blockade of RANKL with antibodies leads to hypermineralization of bone and reduced bone destruction in the setting of inflammatory arthritis. In mouse tumor models, bone destruction and the frequency of metastases is RANK-dependent (10). Studies of Mendelian genetic diseases with gain or loss of function of RANK suggest that RANK is also an essential modulator of bone turnover in humans (9). Denusomab, a blocking monoclonal antibody against RANKL, is effective in blocking bone resorption in the setting of osteoporosis and metastatic cancer, diseases for which its use is approved by the U.S. Food and Drug Administration.

The RANK antagonist generated by Warren et al. is based on a covalently linked trimer of RANKL in which key amino acid residues important for the binding of one or two subunits to RANK have been mutated [“single-block” and “double-block” single-chain RANKL (scRANKL)]. These molecules retained RANK binding commensurate with the number of wild-type RANKL subunits and did not stimulate RANK signaling, but they failed to antagonize signaling induced by wild-type RANKL, probably because the wild-type RANKL out-competed the mutant RANKL for binding to RANK. To improve the chances that a mutant heterotrimer would act as an antagonist, Warren et al. turned to random mutagenesis through a yeast surface display technique and selected for RANKL mutants that specifically improved binding to RANK, but not OPG. The authors then secondarily selected for mutants with low off-rates for RANK binding by selecting for monomeric RANKL that could bind to RANK on cells, as detected with flow cytometry. These selections generated a RANKL trimer with an affinity for RANK that was 500-fold greater than that of wild-type RANKL and that had no detectable binding to OPG. A RANKL trimer consisting of two subunits of this “super” RANKL and one of the non–RANK-binding subunits failed to activate RANK signaling, potently blocked the binding of wild-type RANKL to RANK in vitro, and modestly reduced bone turnover induced by injection of RANKL in a short-term osteoclast stimulation assay in mice.

Why does one bad subunit of three block signaling so potently? There are a number of possibilities (Fig. 1). RANK may need to form a trimer to adopt the correct conformation to initiate signaling. We know from structural studies that TRAF [TNF receptor (TNFR)–associated factor] molecules critical for signaling by RANK and other TNF family receptors must be recruited as trimers, presumably to a trimeric receptor. Having a mutant subunit in the ligand trimer may interfere with the recruitment of the third receptor subunit. The two high-affinity subunits presumably function to “lock” RANK into a nonsignaling conformation and render it impervious to competition by wild-type RANKL.

Fig. 1 Generation of an antagonist from a covalent scRANKL fusion protein.

The principal forms of RANKL used by Warren et al. are shown in sequence, with mutations in the receptor-binding sites of one chain of wild-type (WT) RANKL (left) leading to the formation of a nonfunctional RANKL (middle), and then the generation of an antagonist by increasing the affinity of the functional chains of RANKL to become “super” RANKL subunits (right). Because of the 3:3 ratio of interacting ligands and receptors, with two surfaces of ligand and receptor making contact, a single mutant subunit of RANKL likely destroys the ability of two RANK molecules to bind, thus preventing productive formation of trimers or higher-order clusters of RANK. The increased affinity of the receptor-binding chains of the fusion protein may enable the antagonist to more effectively prevent receptor subunits from signaling when wild-type RANKL is added.


Compared with TNF inhibitors that are monoclonal antibodies or Fc-fusion molecules, the cytokine antagonist engineered by Warren et al. may provide some advantages. First, because mutating the binding sites of single-chain variants can alter the affinity of ligands for their receptors, in systems in which there is more than one receptor for the cytokine (for example, in the binding of TNF to either TNFR1 or TNFR2), such an approach could produce a receptor-specific antagonist. Second, because they lack Fc domains, these fusion proteins would not be likely to become agonists after FcR-mediated cross-linking. To progress into a therapeutic pipeline, these antagonists would need to block endogenous RANKL in more physiological animal models of inflammatory arthritis, osteoporosis, and bony metastases. In addition to becoming therapeutics themselves, these engineered antagonists may help to define important interfaces and residues of cytokines that control receptor binding and signaling. This knowledge may aid in designing small-molecule antagonists against TNF, which has been a long-standing goal in the field.

References and Notes

Funding: This work was funded through the intramural research program of the National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH.
View Abstract

Stay Connected to Science Signaling

Navigate This Article