In the twentieth century, drug discovery largely involved the identification of small molecules that exert their therapeutic effects by interacting with the binding sites of endogenous small-molecule ligands such as neurotransmitters or hormones on target proteins. Reflecting this, drug targets fell mainly into one of four classes of proteins: receptors, enzymes, ion channels and transporters. However, while small molecules remain the predominant therapeutic modality among approved agents1,2,3,4,5, advances in molecular and cellular biology, genomics and pharmacology in the past few decades have enabled the development of several other therapeutic modalities, including full-length monoclonal antibodies (mAbs), decoy receptors, antibody fragments and mimetics, oligonucleotide therapies and gene therapies6. Antibody-based therapies in particular have dramatically changed the landscape of drug development; more than 80 mAbs have been approved since the pioneering mAb muromonab-CD3 was approved by the FDA in 1986, and three of the top ten best-selling drugs in 2018 and 2019 were ligand-targeting mAbs7,8,9.

The expansion in the types of biopharmaceutical agents has broadened the types of targets of approved drugs and drugs in clinical trials because characteristics of these agents, such as their target specificity and size, enable the effective modulation of target types that are far less tractable for small molecules. In particular, soluble protein ligands such as cytokines and growth factors can be readily targeted with mAbs, fusion proteins and other modalities such as aptamers to achieve a therapeutic goal such as blockade of the interaction between protein ligands and their receptors — a goal that is far more challenging to achieve with small-molecule drugs. Protein ligands have been successfully targeted by many drugs since the first FDA approval of the ligand-targeting agents etanercept and infliximab in 1998, including cytokines such as tumour necrosis factor (TNF) and growth factors such as vascular endothelial growth factor A (VEGFA).

Therapeutic mAbs and their targets have been investigated, often with a focus on specific indications — predominantly cancer or inflammatory diseases — and other modalities such as gene therapies and their targets have also been reviewed10,11,12. However, to our knowledge, a comprehensive study of the endogenous soluble ligands targeted by all therapeutic modalities has not been published. Here we present an analysis of soluble ligands as a class of drug targets based on approved agents and agents in clinical trials between 1992 and 2020 (see Box 1 for details of the origin of the dataset and analysis), including the types of ligands being targeted and the characteristics of the drugs and their indications. Furthermore, we discuss strategic issues such as the pros and cons of different ligand-targeting therapeutic modalities and why targeting ligands versus their receptors might be preferable in different disease contexts.

Overview of ligand-targeting drugs

The FDA has approved 34 new molecular entities (NMEs) that target soluble ligands since the first such agents — etanercept and infliximab — were approved in 1998 (Fig. 1). There is a trend towards an increasing number of ligand-targeting drugs being approved, with more than half of the NMEs gaining approval in the last 6 years and the highest number of approvals in the last 2 years, with six agents approved in 2018 and five agents approved in 2019.

Fig. 1: Overview of ligand-targeting drugs and therapeutic targets.
figure 1

a | Number of agents that target ligands approved by the FDA per year. In total, 34 ligand-targeting drugs have been approved by the FDA since 1998. b | A total of 291 ligand-targeting agents were identified through the analysis described in Box 1, of which 34 are FDA-approved agents, 119 are currently in clinical trials or apparently still in active clinical development and 138 have been discontinued. c | A total of 99 unique ligands were identified, of which 22 are targeted by FDA-approved drugs, 52 are targeted by agents in clinical development and 25 are the targets of agents for which clinical development has been discontinued. The data are to 15 June 2020.

In total, we identify 291 ligand-targeting agents that are approved or have been in clinical trials since 1992, which target a total of 99 unique ligands, with 84 ligands targeted individually and 25 different combinations of ligand targets. At present, 119 investigational ligand-targeting agents are currently in clinical trials or apparently still in active development, while the development of 138 agents that reached clinical trials has been discontinued (Fig. 1b). There are 22 clinically validated ligand targets — that is, ligands that are targeted by FDA-approved drugs — and 77 further novel ligand targets of investigational agents that have reached clinical trials. Of these, 52 ligands are the targets of agents currently in clinical development, while clinical development of agents for 25 of the ligand targets has been discontinued (Fig. 1c).

The number of ligand-targeting agents entering clinical trials has increased substantially over the past two decades, with eight agents entering trials between 1990 and 1999, 115 entering trials between 2000 and 2009 and 168 entering trials between 2010 and 2020 (Fig. 2a). Furthermore, the number of agents that bind multiple soluble ligands has increased substantially in the last decade (36 agents) in comparison with the previous two decades (10 agents). Such agents may bind multiple ligands through an epitope that is common to closely related proteins (for example, luspatercept, a decoy receptor that functions through binding growth/differentiation factor 8 (GDF8; also known as myostatin) and GDF11 of the transforming growth factor-β superfamily13) or may be designed to bind unrelated ligands (for example, faricimab, a bispecific mAb that binds to angiopoietin 2 (ANGPT2) and VEGFA14). While the number of ligand targets that have been investigated in trials has ranged from a total of 58 between 2000 and 2009 to 35 between 2010 and 2019, the number of unique combinations of ligand targets has increased 1.5-fold in the last decade to a total of 15 (Fig. 2b), and the opportunities for multitargeting are discussed further in the following sections.

Fig. 2: Therapeutic agents and targets entering clinical trials.
figure 2

a | The blue bars represent the total number of therapeutic agents targeting ligands (291 agents) that have entered clinical trials each year beginning in 1992, which is the first year in which we identify one of the agents beginning a registered clinical trial, and extends to 15 June 2020. The dark blue line indicates the subset of agents that bind multiple targets that have entered clinical trials. The ‘study start date’ field from was used to determine when an agent entered trials. Owing to the possible inconsistent registration of clinical trials in during the decade preceding and the years immediately following the turn of the century, this may be an under-representation of all the ligand-targeting drugs and trials (see Box 1 for details). b | The green bars indicate the number of unique ligand targets that entered clinical trials per year, while the darker green line represents the number of novel target combinations that entered clinical studies. The year a target ligand (or combination) entered trials is derived from when the earliest therapy targeting that ligand (or combination) entered clinical trials. The data include 99 unique ligands and 25 combinations. c | More than 5,700 clinical trials of ligand-targeting drugs have been registered since 1992. Approximately 30% involve novel investigational agents entering trials, while ~70% are additional studies involving the 34 approved agents (grey) after they had already been approved. The number of trials in each phase initiated per year is shown using the study start date from The data include clinical trials that have a designated phase, are listed as an interventional study type and have the primary purpose of treatment. d | The major therapeutic categories being treated by active agents. Inflammation and autoimmune diseases, cancer and ophthalmological diseases are three major disease groups being treated by both approved agents and agents in development. Inflammation and autoimmune diseases are the predominant areas for approved agents, while cancer is an increasingly popular area for agents in earlier stages of development. The data include all active agents and the highest phase achieved for a specified disease area is selected to prevent redundancies (agents in multiple trials for similar indications). Two agents have FDA approval in multiple disease areas and investigational agents may be in trials for more than one disease area. The data are to 15 June 2020.

More than 5,700 clinical trials for ligand-targeting agents have been initiated since 1992 (Fig. 2c), with ~30% of trials initiated involving novel agents and ~70% involving approved agents. The number of trials initiated for novel agents increased between 2010 and 2019 compared with 2000–2009, reaching 142 in 2018, including 46 phase III trials. Most FDA-approved ligand-targeting agents are for the treatment of inflammatory and autoimmune disorders (Fig. 2d), while cancer is an increasingly popular area for agents in earlier stages of clinical development.

Which ligands are being targeted?

Cytokines constitute 50% (11 cytokines) of the 22 ligands targeted by FDA-approved drugs and 40% (31 cytokines) of the 77 novel ligands for which agents are in development (Fig. 3a,b). Growth factors are the second-largest group and constitute 27% of the ligands targeted by FDA-approved drugs and 31% of the novel ligand targets. The other ~30% of ligands targeted by approved and investigational agents can be classified in a range of categories, including neuropeptides such as calcitonin gene-related peptide (CGRP), inflammatory mediators in the complement pathway15 and other signalling molecules such as sclerostin (SOST) (Fig. 3c). Selected examples from the major ligand classes based on the number of FDA-approved agents and/or the number of drugs in clinical trials are shown in Table 1, and the FDA-approved ligand-targeting agents and their approved indications are shown in Table 2.

Fig. 3: Classes of the target ligands.
figure 3

a | The 22 unique soluble ligands that have been clinically established as targets by FDA-approved agents are separated into different classes. Cytokines include interleukins, interferons and chemokines in parts a, b and c. b | The 77 unique ligands that are targeted by novel investigational agents in clinical trials are separated into different classes. c | The different classes of the 99 unique ligands that have been targeted by all agents are separated into the number of ligands that have been clinically established as targets by FDA-approved drugs, the number of ligands that are being targeted by therapeutic agents in active development and the number of ligands that are targeted by agents that are now discontinued. d | Ten selected ligands that are targeted by the greatest number of agents in the dataset are shown with the number of FDA-approved agents that target the ligand, the number of active agents that target the ligand and the number of discontinued agents that previously targeted that ligand in clinical trials. GDF8, growth/differentiation factor 8; IL, interleukin; TGFβ1, transforming growth factor-β1; TNF, tumour necrosis factor; VEGFA, vascular endothelial growth factor A.

Table 1 Selected examples of ligand targets from different ligand classes
Table 2 FDA-approved ligand-targeting drugs and their indications

TNF has been one of the most targeted ligands (Fig. 3d). Of the 32 agents that target TNF, 25 agents target it exclusively, of which 5 have been approved, and 7 drugs target two unique combinations that include it. This reflects its position at the apex of the proinflammatory cytokine cascade, which makes it a key target for inflammatory disease treatment16. Anti-TNF drugs were initially developed to treat rheumatoid arthritis and Crohn’s disease, with two of these — the decoy receptor etanercept and the full-length mAb infliximab — both gaining FDA approval in 1998. Since then, three additional anti-TNF drugs have been approved (Table 2), and this class of drugs has been investigated in many other autoimmune diseases, gaining approval for indications such as ankylosing spondylitis, ulcerative colitis, juvenile idiopathic arthritis and psoriasis.

VEGFA, a key regulator of normal and pathological angiogenesis17, has also been a heavily targeted ligand, with 29 agents, of which 5 have been approved. There are currently 18 active novel agents targeting it in clinical trials — the most for any ligand (Fig. 3d). Initially, research on VEGFA as a drug target focused on its role in tumour angiogenesis, and bevacizumab, a full-length mAb that binds to VEGFA, became the first drug against this target to reach the market when it was approved in 2004 for treatment of colorectal cancer in combination with chemotherapy. It is now approved for treatment of multiple cancers, as both a single agent and in combination with chemotherapy. VEGFA was also found to be important in eye disorders characterized by ocular neoangiogenesis, such as neovascular age-related macular degeneration (AMD) and macular oedema18. This has led to the approval of several drugs targeting VEGFA for treatment of such disorders, beginning with the aptamer pegaptanib for treatment of AMD in 2004 (discussed further later).

The interleukin-17 (IL-17) pathway has also been a popular target, because of its role in inflammatory and autoimmune diseases such as psoriatic diseases19,20 (Fig. 3d). IL-17A is part of a family of six cytokines (IL-17A–IL-17F), and can exist as a homodimer or as a heterodimer with IL-17F. Of the 20 agents that target IL-17A (Fig. 3d), 8 target just IL-17A, while 12 target 4 unique combinations, including 6 drugs that target IL-17A plus TNF. Two agents that target IL-17A have been approved in the last 5 years: secukinumab and ixekizumab in 2015 and 2016, respectively. IL-17 is produced by T helper 17 (TH17) cells that are primarily regulated by IL-23, which has also been a popular target for treatment of psoriasis and other immunoinflammatory diseases. IL-23, a heterodimeric cytokine composed of the subunits IL-23p19 and IL-12p40, is the interleukin with the greatest number of approved agents targeting it — three full-length mAbs that target the IL-23p19 subunit (guselkumab, tildrakizumab and risankizumab, approved in 2017, 2018 and 2019, respectively) and one full-length mAb that targets the p40 subunit and consequently also binds and inhibits IL-12 (ustekinumab, approved in 2009).

Although many approved drugs targeting ligands are indicated for treatment of immuno-inflammatory disorders, reflecting the role of the ligands in the immune system (Fig. 2d), there are a growing number of examples in other therapeutic areas. For example, SOST is a secreted protein that has recently been clinically validated as a target for treatment of osteoporosis, leading to the FDA approval of the full-length mAb romosozumab in 2019 (ref.21). SOST negatively regulates the formation of mineralized bone matrix and bone mass, and so inhibition of SOST results in increased spine and hip bone mineral density22. In another recent example, CGRPα, the protein product of CALCA, has been validated with the FDA approval of three CGRP-specific full-length mAbs for the treatment of migraines: fremanezumab and galcanezumab in 2018 and eptinezumab in 2020 (refs23,24,25). These mAbs also bind to CGRPβ (encoded by CALCB), which has 90% structural similarity to CGRPα26. CGRPα, the principal form of CGRP in the nervous system, contributes to the pain of a migraine attack through its roles in vasodilation, neurogenic inflammation and peripheral sensitization, and antagonism of the CGRP pathway has been shown to ameliorate migraine pain as well as to offer the potential to prevent migraine attacks27.

Classes of ligand-targeting drugs

Among the approved ligand-targeting drugs, approximately three-quarters (25 of 34) are full-length mAbs, with the remainder made up of four decoy receptors, three antibody fragments, one nanobody and one aptamer (Fig. 4a). These drugs all bind to a specific soluble ligand (or ligands) and sequester it (or them), thereby modulating the targeted pathway(s) by inhibiting the ligand–receptor interaction(s) and subsequent signalling.

Fig. 4: Classes of ligand-targeting agents.
figure 4

a,b | The 34 FDA-approved therapies (part a) and 257 investigational agents (part b) were identified and classified by the different therapeutic platforms. Antibody therapies including full-length monoclonal antibodies and antibody fragments such as nanobodies, Fabs and single-chain variable regions are shown in blue tones. Other platforms include aptamers and spiegelmers, gene therapies, antibody mimetics, therapeutic vaccines, decoy receptors, oligonucleotide therapies such as antisense oligonucleotides and anti-microRNAs, and other therapies. The number of agents and the percentage in each category are shown. c | The 291 therapeutic agents in the dataset classified by their different platforms and separated by the number of approved agents, the number of active agents and the number of discontinued agents in each platform.

There is greater diversity in the novel agents in clinical trials. Overall, nearly two-thirds of the 257 biopharmaceutical agents in clinical trials are antibody-based drugs and derivatives, including 151 full-length mAbs, 8 antibody fragments such as Fabs and single-chain variable regions (scFvs) and 6 single-domain antibodies such as nanobodies (Fig. 4b). The other classes in the dataset include decoy receptors, engineered protein scaffolds, gene therapies, therapeutic vaccines and various types of oligonucleotide therapies. Therapeutic antibodies and decoy receptors, which have become well-developed classes of agents over time, have the highest number of active agents (Fig. 4c) as well as the greatest number of drugs in phase III trials, further indicating the maturity and continued interest in these therapeutic modalities. In the following sections, we discuss the characteristics of the major therapeutic modalities in the dataset in more depth.

Monoclonal antibodies

As already noted, mAbs have become established as a key therapeutic modality in the past two decades — a success that has been based on the specificity, stability and adaptability of the antibody framework28 (Fig. 5a). Therapeutic mAbs can be distinguished by several characteristics, including the technology from which the mAb is derived, the immunoglobulin G (IgG) subtype and the engineering strategies applied to the mAb29. Here we focus briefly on aspects that are most relevant to ligand-targeting mAbs.

Fig. 5: Established therapeutic platforms for ligand-targeting drugs.
figure 5

a | Structural components and functional activities of antibody therapies. The antibody is constructed of variable (light blue) regions that contain three complementarity-determining regions (CDRs; or hypervariable regions) that together form the target-binding paratope, which binds to an antigen epitope. Modifications in the variable region can affect immunogenicity, antigen-binding affinity and specificity. The Fc region and glycosylation patterns modulate effector functions and antibody half-life. Specific amino acid alterations in this region or alteration of the glycosylation patterns can induce or suppress effector functions and alter antibody circulating times. b | Antibody engineering used in FDA-approved ligand-targeting agents. The first row illustrates the progression in strategies to address immunogenicity issues, from chimeric full-length antibodies, shown with purple variable regions and CDRs to indicate material from other species (2 approved agents) to humanized antibodies with non-human CDRs (12 approved agents) to fully human antibodies (10 approved agents). The second row shows antibody fragment constructs for approved agents: one humanized Fab (ranibizumab); one humanized Fab conjugated with polyethylene glycol (PEG) for increased half-life (certolizumab pegol); one single-chain variable region (scFv; brolucizumab); and one single-domain antibody (sdAb) — a VHH fragment derived from the heavy-chain regions from camelids and called a ‘nanobody’ (caplacizumab). c | Decoy receptors are the second most common therapeutic platform among approved ligand-targeting agents; there are four approved decoy receptors, including aflibercept (shown), which targets vascular endothelial growth factor A (VEGFA), VEGFB and PGF. One ligand-targeting aptamer has been approved: pegaptanib is initially a single RNA strand that folds into a specific 3D structure that binds and inhibits VEGFA.

Immunogenicity, leading to the presence of anti-drug antibodies that compromise drug efficacy and safety, is a major challenge for therapeutic mAbs as well as other biological therapies30. The pioneering therapeutic mAbs in the 1980s were derived from mice, and produced human anti-mouse antibody responses owing to the non-human sequences in the constant regions28. This led to the development of successive generations of technology to reduce the extent of foreign-derived sequences in the antibody: chimeric mAbs with foreign variable regions and human constant regions, humanized mAbs that are human except for foreign complementarity-determining regions and, most recently, fully human mAbs31 (Fig. 5b). The first fully human mAb, adalimumab, was approved in 2002, and since then mAbs that are fully human have become increasingly popular. Of the 25 approved full-length ligand-targeting mAbs in the dataset, 13 (52%) are humanized, 10 (40%) are fully human and 2 (8%) are chimeric. The 151 novel ligand-targeting mAbs that are in clinical trials also have a high proportion of humanized or fully human mAbs; 72 are humanized (48%), 68 are human (45%) and 3 are chimeric (1%) (data were not obtainable for six agents and an additional two older agents are full mouse constructs).

The choice of IgG subtype may be important depending on the intended therapeutic application of a mAb, as IgG subtypes can differentially interact with the five activating Fcγ receptors (FcγRs), the single inhibitory receptor FcγRIIb and the C1q component of the complement cascade28,32,33. Binding to activating FcγRs can stimulate immune responses and evoke antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis and/or complement-dependent cytotoxicity (CDC)32, whereas binding with the inhibitory FcγRIIb can dampen immune responses34. The IgG1 subclass is the most prevalent in our dataset, with 17 (68%) of the approved ligand-targeting full-length mAbs and 81 (53%) of the agents in clinical trials. IgG2 frameworks (three mAbs) and IgG4 frameworks (three mAbs) and combinations (two mAbs) have also been used in approved ligand-targeting mAbs. However, for investigational agents, IgG4 subtypes (34 mAbs; 23%) are more than twice as common as IgG2 subtypes (17 mAbs; 11%) (data were not obtainable for 19 agents).

In general, although mAbs targeting soluble ligands have low Fc effector function potential, one matter to consider in selecting the IgG subtype is whether there is a membrane-bound form of the target ligand that the antibody may interact with, as with membrane-bound TNF, or whether ligand–antibody complexes could bind to the cell surface and elicit effector function activity35. Selecting IgG2 and IgG4 subtypes, which have lower potential to elicit effector functions, and/or engineering modifications such as aglycosylation of the Fc region to minimize effector functions might be advantageous for ligand-targeting agents that are intended to dampen immune responses. For example, the development of the approved complement C5 inhibitor eculizumab involved engineering the heavy-chain constant region with components of human IgG2 and IgG4 constant regions to create a mAb that lacks the ability to bind to Fc receptors or activate the complement cascade36. Fc engineering can also be used to extend the half-life of therapeutic mAbs by enhancing binding to the neonatal Fc receptor FcRn37. An example is provided by the second-generation C5 inhibitor ravulizumab, which was developed with eculizumab as a starting point; it has two amino acid substitutions in the Fc region that increase the affinity of the antibody for human FcRn and contribute to its longer half-life38.

Decoy receptors

The second most common class of agents among the approved ligand-targeting drugs (four drugs; 12%) and the agents in clinical trials (22 drugs; 9%) are decoy receptors, also known as Fc-fusion proteins, which incorporate the binding domains from endogenous ligand receptors to trap the ligand (Fig. 5c). The fusion of the receptor binding moiety with the Fc region helps to extend the half-life of the drug, although the large size (~150 kDa) can limit tissue penetration. The first such agent to reach the market was the TNF-receptor–Fc-fusion protein etanercept, which received approval by the FDA for treatment of rheumatoid arthritis in 1998, and has since been approved for treatment of several other inflammatory diseases (Table 2).

Three other decoy receptors have been approved so far. Rilonacept, an engineered dimeric fusion protein consisting of the ligand-binding domains of the extracellular portions of the human IL-1 receptor component (IL-1RI) and IL-1 receptor accessory protein (IL-1RAcP) linked to the Fc portion of human IgG1, was approved in 2008 for the treatment of a group of rare autoinflammatory diseases known as cryopyrin-associated periodic syndromes39. Aflibercept, which contains the second binding domain of the VEGFR1 receptor and the third domain of the VEGFR2 receptor to bind isomers of the VEGF family40, was approved for treatment of wet AMD in 2011. It was also approved in a combination treatment for colorectal cancer in 2012 (ref.41), for treatment of diabetic macular oedema and macular oedema following retinal vein occlusion in 2014 and for treatment of diabetic retinopathy in 2015. Luspatercept is a fusion protein that contains the extracellular domain of human activin receptor type IIB (ActRIIB) that has been modified to reduce activin binding13. It inhibits select ligands of the transforming growth factor-β superfamily to reduce aberrant SMAD2/3 signalling and promote late-stage erythropoiesis and was approved in 2019 for treatment of adult patients with β-thalassaemia and for treatment of anaemia in adults with myelodysplastic syndromes in 2020. Fifteen decoy receptors are currently in clinical trials, and more than half (ten agents) of these began trials in the last 5 years, indicating continued industry interest in this class (Fig. 4c).

Antibody fragments and single-domain antibodies

Although full-length mAbs have successfully been developed to treat a variety of disorders, their size (~150 kDa) can impede the desired tissue penetration42. Various strategies have been pursued to develop antibody-based drugs with reduced size to address this issue, including antibody fragments and nanobodies. However, antibody fragments have short half-lives as they are preferentially cleared via lysosomal degradation in addition to renal filtration28, which could result in more frequent dosing requirements. This may be undesirable for chronic therapies, but can be addressed by modifications such as pegylation, as with the TNF-targeted antibody fragment certolizumab pegol28.

Roughly 6% of the antibody therapies (18 agents) in our total dataset consist of fragment constructs including Fab, (Fab)2, scFv and single-domain antibodies such as nanobodies (Fig. 5b). Antibody fragment therapies were clinically established in 1998 with the FDA approval of abciximab, and the first ligand-targeting Fabs ranibizumab, certolizumab pegol and brolucizumab were approved in 2006, 2008 and 2019, respectively43 (see later for further discussion of the rationale for pursuing fragments for these targets).

Nanobodies are constructs engineered from the unique antibodies produced by members of the Camelidae family (camels and llamas), which are composed solely of a VHH domain where the target recognition module contains a single variable domain44. In 2019, the FDA approved the first nanobody-based medicine, caplacizumab, which targets von Willebrand factor (VWF) for treatment of the rare disease acquired thrombotic thrombocytopenic purpura45. The drug has a short half-life, but patients undergo plasmapheresis every day, and so this is not expected to be a major barrier to its use.

Aptamers and spiegelmers

Aptamers and spiegelmers are nucleic acid polymers that fold into distinct 3D structures that bind specifically and with high affinity to a defined target, including ligands, ions or small molecules46. They are essentially nucleotide analogues of antibodies that could have several advantages, including smaller size, rapid chemical production and lack of immunogenicity47. Spiegelmers are a mirror-image configuration of aptamers composed of non-natural l-nucleotides, which makes them highly resistant to degradation by nucleases48.

The RNA aptamer pegaptanib, which binds to VEGFA, is the first and to date only aptamer approved by the FDA49. Pegaptanib was approved for treatment of wet AMD in 2004, and was the first treatment that could effectively slow disease progression; however, it has since been superseded by other VEGFA-targeted therapies, as discussed further later. There are two active aptamers and two spiegelmers that are currently in phase II trials. Recent reviews indicate that there are also many aptamers in preclinical development50,51, suggesting that this platform may attract continued therapeutic interest.

Engineered protein scaffolds

Several types of engineered protein scaffolds, also known as antibody mimetics, can bind to antigens and ligands in a similar fashion to antibodies and have been developed with the aim of addressing some of the limitations of mAbs (for recent reviews on antibody mimetics, see52,53). Their advantages could include higher affinity, greater stability and solubility, and inexpensive large-scale production. Furthermore, as with antibody fragments, the smaller sizes of mimetics could enable an increased rate of diffusion and more efficient penetration into the target tissues. However, similarly to antibody fragments, antibody mimetics can also have problems such as immunogenicity and short plasma half-life52.

The nine antibody mimetics that target ligands in our dataset represent a rich cross section of different scaffold designs, including designed ankyrin repeat proteins (DARPins), affibodies and avimers. Two of the three agents in active clinical trials are DARPins, which are artificial protein scaffolds based on ankyrin repeat domains that mediate a broad range of protein interactions54. They are highly stable and soluble, and their small size (14–18 kDa) could enable greater tissue penetration than full-length mAbs53. The anti-VEGFA DARPin abicipar pegol55 has shown positive effects in phase III trials for treatment of neovascular AMD, although recently the FDA declined to approve this agent. The DARPin agent MP0250 is currently in phase I/II trials. The other mimetic in clinical development is an affibody, which is in phase II trials. Affibodies are small (~6-kDa) constructs that consist of three α-helices that hold the variable domains that are responsible for epitope recognition and for binding to a range of proteins56.

Gene therapies

Gene therapies — which deliver a therapeutic gene to target cells using a vector such as an adeno-associated virus (AAV) — are becoming established as a treatment modality after years of setbacks57. Most such therapies involve the delivery of a functional copy of a defective gene, typically for the treatment of rare monogenic disorders. However, there are six investigational therapies in our dataset for which the mode of action involves the delivery of a gene to express a protein that targets a ligand. The two currently active gene therapies are both in phase I trials; one is an AAV vector which encodes the anti-VEGF agent aflibercept and has been optimized for intravitreal delivery and strong protein expression58, while the other is an AAV vector that carries a gene that encodes a soluble anti-VEGF protein59. Eye disorders have been one of the areas in which gene therapies in general have been successful in recent years, and in the case of VEGFA-targeted agents for treatment of AMD, such agents could circumvent the need for repeated intravitreal injections, which is a limitation of current therapies60.

Therapeutic vaccines

There are 13 therapeutic vaccines in our dataset that contain an antigen–carrier construct that, on delivery, evokes antibodies to a specific ligand target to achieve a therapeutic effect61. However, the extent of recent developments in this class of agents is limited compared with other classes, possibly due to issues with lack of efficacy in various indications62,63. Three ligand-targeting vaccines are in active trials, including an EGF-targeted immunotherapy known as EGF-PTI64 that is in two phase II trials, although three phase III trials of this agent were terminated by the sponsor because of enrolment issues. Ten agents have been discontinued, with most of the agents not progressing past phase I or phase I/II trials.

Oligonucleotide therapies

All the agent classes discussed so far directly target the ligand protein. However, it is also possible to affect the production of the ligand protein with various platforms, including antisense oligonucleotides (ASOs)65, small interfering RNAs (siRNAs)66 and microRNA-based agents67.

ASOs, which can bind with high specificity to the mRNA for a protein and promote its degradation, are the most developed of these platforms, with several ASOs having gained FDA approval, although none of these target ligands. There are 11 ASOs in our dataset that modulate the expression of endogenous ligands, with 6 currently active in trials. The most advanced agent is ISIS-HTTRx, which targets the huntingtin (HTT) gene for treatment of Huntington disease and is in phase III trials68.

There are also three small siRNA-based therapies that silence the expression of genes coding for ligands by exploiting the RNA interference pathway. One of the first siRNA drugs to begin clinical trials was the anti-VEGFA agent bevasiranib in 2004, although its development was terminated following a phase III trial owing to poor efficacy in reducing vision loss69. Two siRNAs which have targets that are not ligands have recently received FDA approval, and of the ligand-targeting siRNAs currently in clinical trials, the most advanced is the angiotensinogen targeting agent ALN-AGT70 for treatment of hypertension, which is in phase I trials.

Finally, six agents in the dataset affect the expression of ligand targets by targeting regulatory microRNAs. One of the agents that has progressed the furthest in clinical development is cobomarsen, which targets miR-155 (which subsequently targets FGF7 (ref.71)) and is in phase II trials for treatment of haematological cancers72.

Strategic issues in targeting ligands

When we noted in 2014 that soluble ligands were becoming one of the largest categories of potentially novel drug targets in clinical trials, there were 11 established ligand targets for FDA-approved drugs and 37 ligand targets being investigated in clinical trials1. The number of established ligand targets has doubled in the past 5 years to a total of 22, while potentially novel targets in clinical trials have increased twofold to 77 ligand targets in our analysis. Ligands now constitute 10% of novel targets in clinical trials and are currently the third-largest target class for therapeutic agents targeting human protein products, after enzymes and receptors2.

There are several factors that contribute to the increasing importance of soluble ligands as drug targets. First, a key factor is the growing understanding of the role of the immune system in many diseases and the tractability of cytokines as therapeutic targets for agents such as mAbs. Second, it may be easier to reach ligands than their receptors. For example, a receptor target expressed in a solid tumour can be inaccessible owing to challenges related to tissue distribution and penetration of the therapeutics73, while ligands that are involved in cancers may be more readily targetable. A third factor is that it may be easier to map epitopes on protein ligands in comparison with transmembrane receptors owing to the complications inherent in resolving tertiary membrane-bound structures. Epitope characterization helps elucidate the mechanism of action by which an antibody and its target interact and can be beneficial in selecting potential therapeutic candidates, as well as for intellectual property claims74. Fourth, some proteins have membrane-bound forms that then become soluble via cleavage; for example, TNF (discussed later), TNFSF13B (also known as BAFF)75 or FASLG76. The different forms of soluble ligand, cleaved transmembrane protein and also possibly both together in a complex may have different roles in normal physiology and disease progression, potentially presenting opportunities to more specifically intervene in disease processes by targeting a particular form. In this section, we consider such strategic issues in the development of ligand-targeting drugs.

Targeting ligands versus receptors

Inflammatory mediators such as cytokines often have complex signalling pathways, with the potential to signal through more than one type of receptor complex, and/or redundant actions if the receptor can interact with more than one cytokine. Consequently, there could be disease scenarios where it may be beneficial to target the ligand rather than the receptor (or vice versa). For example, clinical trials with the IL-17A-targeting mAb ixekizumab showed improved signs and relief of symptoms in patients with rheumatoid arthritis77, whereas trials of the IL-17R-targeting mAb brodalumab, which inhibits all signalling by all members of the IL-17 family, showed no response78,79. It has been speculated that this non-response may be a result of the additional inhibition of signalling by the cytokine IL-17E (also known as IL-25) by brodalumab80.

An interesting example beyond the immune system in which the benefits of blocking the ligand versus the receptor are currently unclear is in the treatment of migraines with the four recently approved therapeutic mAbs. Fremanezumab, galcanezumab and eptinezumab are humanized mAbs that bind CGRPα and CGRPβ, while erenumab is a fully human mAb that targets the CGRP receptor23,24,81. Erenumab blocks only the canonical CGRP receptor and does not affect actions of CGRP mediated by other CGRP-binding receptors27, which could limit the side effect profile. However, the effects mediated by the interaction of CGRP with other receptors, such as amylin 1 (AMY1), have not been fully elucidated yet and they may also be involved in migraine pathology, which might mean that targeting the CGRP receptor rather than CGRP could have lower efficacy27. In this case, the relative merits of ligand or receptor targeting have yet to be determined, and a deeper understanding of the CGRP pathways in connection with migraines and other effects, assessments of genetic factors such as possible individual polymorphisms that influence therapeutic response and analyses of the comparative efficacy of these agents are needed to establish the appropriate agent given the circumstances.

Opportunities to target multiple ligands

Approximately 16% of the dataset (46 agents) are agents that bind more than one soluble ligand, with 25 unique combinations of ligands being targeted. As mentioned previously (Fig. 2a,b), increases in the number of agents targeting more than one ligand as well as the number of unique combinations of ligands can be seen in the last decade in comparison with previous years. The agents include 26 bispecific or multispecific agents that are designed to bind multiple unrelated targets or bind different epitopes on related proteins. These constructs include 16 full-length mAbs and four antibody fragments, three antibody mimetics, two decoy receptors that contain multiple domains to target unrelated proteins and one peptide–Fc-fusion protein. Another 19 agents bind more than one protein because one epitope is common to more than one protein, such as in closely related members of a family. This includes 14 decoy receptors, four full-length mAbs and one peptide–Fc-fusion protein. Additionally, there is one combination (or mixture) that contains two separate monospecific agents (we did not include three agents considered bispecific as the second target was human serum albumin to extend the half-life).

Bispecific antibodies became validated as a therapeutic class in the USA with the 2014 FDA approval of blinatumomab for treatment of acute lymphoblastic leukaemia82, although it is a bispecific T cell engager that does not target soluble ligands. Although bispecific constructs that target cell-surface ligands and receptors primarily for treatment of cancer are a large portion of the bispecific antibody pipeline, they will not be further discussed here (see refs83,84,85 for reviews). However, nearly 30% of the 26 multispecific ligand-targeting therapies are currently in clinical trials for treating cancer. Multispecific antibodies are able to address the multifactorial nature of complex diseases, and in particular cancer, where therapeutic resistance can emerge, as cancer cells have the ability to downregulate targeted pathways, upregulate alternative pathways and generate crosstalk between pathways, which can lead to resistance and reduction of antibody efficacy83. This appears to occur primarily for antibody therapies where a single epitope of a target in one specific pathway is targeted. Indeed, as tumour angiogenesis is known to play a crucial role in tumour growth, seven bispecific therapies for oncological applications target either VEGFA plus DLL4 to enhance antitumour effects while limiting VEGFA-induced vascular sprouting86 or VEGFA plus ANGPT2 for dual inhibition of the VEGFA–VEGFR and ANGPT2–TIE2 signalling pathways.

Advances in engineering technology and the ability to effectively and safely target more than one ligand has combined with increased knowledge of disease pathways and also clinical experience from single-target agents to better identify and effectively utilize therapies. Additionally, the development of monotherapies targeting multiple targets can be less complex than the development of multiple drug products in combination, as well as contributing to reducing production costs. Furthermore, using multiple paratopes on multispecific antibodies can promote prolonged serum half-life, improved tissue distribution and local enrichment87.

Platform choices may influence therapeutic effects in different contexts

For some popular ligand targets, several strategies have been applied. For example, there are five anti-TNF agents that have been approved by the FDA: three full-length mAbs (infliximab, adalimumab and golimumab), one TNF receptor–Fc-fusion protein (etanercept) and one pegylated Fab (certolizumab pegol). All of these agents are successful anti-inflammatory agents that target TNF; however, each platform may confer different attributes, and the choice of platform may affect the efficacy for specific indications.

Infliximab, adalimumab, golimumab and certolizumab pegol have all been approved for treatment of inflammatory bowel diseases (IBDs) such as Crohn’s disease and ulcerative colitis88, whereas etanercept has been shown to not be effective for IBDs89. The primary mechanism of action of anti-TNF agents is to bind and neutralize the proinflammatory cytokine TNF, predominantly the soluble form. However, the agents that contain an Fc domain also bind to tmTNF, the transmembrane-bound precursor to soluble TNF, albeit to a lesser degree90. Fc-mediated interactions with tmTNF appear to play a key role in IBD treatment: they induce apoptosis in immune cells through reverse signalling in tmTNF-expressing cells90,91, they promote cell lysis via ADCC and CDC92, and they downregulate proinflammatory mediators93,94. TNF inhibitors also decrease intestinal permeability95, which minimizes mucosal inflammation, and induce regulatory macrophages96, which promotes mucosal healing. Certolizumab, however, does not contain the Fc portion of IgG1 and has been shown to be unable to induce ADCC, CDC, regulatory macrophages or apoptosis via reverse signalling97. Nevertheless, it has been shown to be effective for induction of clinical remission and response in patients with Crohn’s disease98, which could be because certolizumab induces a direct cytotoxic effect on tmTNF-expressing cells by a non-apoptotic mechanism99. Etanercept, which also has an Fc region, is unique in that it binds and neutralizes members of the lymphotoxin family, possibly due to the fusion protein construct retaining the ligand-binding specificity of its parent receptor TNFR2 (ref.100). Etanercept inhibits lymphotoxin-α3, which is a key cytokine in the regulation of the mucosal immune system, and it is suspected that this inhibitory action contributes to the ineffectiveness of etanercept in Crohn’s disease and it may even slightly increase the risk of developing IBD88.

All five of these anti-TNF agents have been approved for treatment of rheumatoid arthritis. TNF is important in the pathogenesis of rheumatoid arthritis owing to its ability to induce the production of other proinflammatory cytokines and chemokines and facilitate the destruction of underlying articular cartilage and subchondral bone101,102. Lymphotoxin-α3 and TNF both mediate proliferation and proinflammatory cytokine secretion of rheumatoid arthritis synovial fibroblasts, and as etanercept blocks both of these cytokines, it may allow better control of inflammation and synovial proliferation in the treatment of rheumatoid arthritis103. There have not been any direct head-to-head comparison trials between anti-TNF biologics, although systematic reviews and meta-analyses have found comparable effectiveness and safety between the five biologics104,105.

Another popular target for which several platforms have been applied is VEGFA. There are currently five FDA-approved agents that target soluble VEGFA: the pegylated RNA aptamer pegaptanib for treatment of neovascular AMD, the full-length mAb bevacizumab for treatment of various cancers and off-label use for ophthalmic indications, the Fab ranibizumab and the scFv brolucizumab for ophthalmic indications, and the recombinant fusion protein aflibercept for ophthalmic indications and colorectal cancer106.

For this target, ophthalmic indications such as wet AMD have been a focus of optimization strategies. The eye is an immunoprivileged site and the blood–retinal barrier prevents free entry from the circulation, so all of the therapeutic agents are administered locally through intravitreal injections. It was anticipated that the molecular size of the agent could be relevant in improving the delivery of the agents, which encouraged the development of antibody fragments107,108. However, despite their structural differences, bevacizumab, ranibizumab and aflibercept all appear to have comparable efficacy and safety, although aflibercept may be superior in patients with initial lower visual acuity109,110. Brolucizumab, the smallest agent so far and the most recently approved, has higher tissue penetrance than the other antibody fragment ranibizumab and has shown longer anti-VEGFA suppression111 as well as more favourable anatomical outcomes in comparison with aflibercept112, but the clinical relevance remains to be established.

However, current treatment paradigms with all of the approved agents involve regular intravitreal injections, which are associated with complications and possible side effects as well as posing a significant burden for the patient and the caregiver113. Of the nine antiangiogenic VEGFA inhibitors currently in clinical trials for treatment of AMD, two are gene therapies, which may provide sustained anti-VEGFA levels in the retina following a single injection and improve patient adherence and comfort.

Investigating alternative delivery routes

Even with the success of therapeutic antibodies, there are still difficult-to-reach sites of action and previously undruggable targets that could be pursued with continued engineering developments114. Advances in therapeutic platforms along with emerging delivery and administration options can facilitate more effective and safer disease management and possibly access privileged areas, for example, behind the blood–brain barrier (BBB). Investigational recombinant proteins have been designed that cross the BBB by binding to transferrin receptors, which transport the therapeutic agent across the BBB and enable it to bind to a target in the central nervous system115. Intranasal administration of agents, where drug absorption is through the olfactory and trigeminal routes, has also been explored to cross the BBB116. Methods to effectively deliver therapies to mucosal tissues are also being investigated. For example, there are three anti-TNF agents — a nanobody, a decoy receptor and a mAb — with an oral administration route for treatment of Crohn’s disease and ulcerative colitis in clinical trials, of which the nanobody V565 is the most advanced, having reached phase II trials117. An inhaled formulation of a high-affinity antibody fragment, abrezekimab, that targets IL-13 was well tolerated in a phase I safety trial, and is proposed to provide a more rapid onset of action at lower doses than subcutaneous systemic administration for the treatment of asthma118.

Repurposing approved drugs for new indications

It is evident that the repurposing of approved drugs for new indications is common; more than half of the 34 approved drugs that target ligands have been approved for additional indications. Furthermore, more than 70% of the initiated clinical trials in our dataset are studies of agents after they have gained FDA approval (Fig. 2c).

One factor in the high number of clinical trials of approved ligand-targeting drugs for new indications is the role of some of the most targeted ligands in immune pathways that are involved in several inflammatory disorders. For example, since the pioneering approvals for treatment of rheumatoid arthritis and Crohn’s disease, TNF-targeted agents have also been approved for other indications, including psoriasis, psoriatic arthritis, ulcerative colitis, ankylosing spondylitis and hidradenitis suppurativa. Similar trends can be seen with other popular cytokine targets, and in some cases these efforts are expanding into new therapeutic areas. For example, inflammation has been suggested to play a role in a psychiatric illnesses such as major depressive disorder119,120, and the effects of various cytokine-targeted drugs on depressive symptoms have been investigated in a number of trials121.

There may also be a substantial number of rare diseases in which a ligand is a promising target; indeed, nearly 45% of the approved drugs targeting ligands have had additional approvals for a rare disease indication, with ten of them having initial approval for treatment of a rare disease. Furthermore, nearly 15% of the novel agents in clinical trials are being investigated as treatments for rare diseases.

Finally, given emerging understanding of the involvement of the immune system in patients with severe COVID-19 (ref.122), including high levels of inflammatory cytokines such as IL-6, IL-17, CSF2, interferon-γ, TNF and IL-2 in some patients123, there has been intense recent interest in repurposing various ligand-targeting agents for treatment of COVID-19. Several approved ligand-targeting agents are in clinical trials involving patients with COVID-19, including the IL-6-targeting drug siltuximab (NCT04329650 and NCT04330638), the VEGFA-targeting drug bevacizumab (NCT04305106, NCT04344782 and NCT04275414), the C5-targeting drugs ravulizumab (NCT04369469 and NCT04390464) and eculizumab (NCT04346797), the interferon-γ-targeting drug emapalumab (NCT04324021) and the IL-1B-targeting drug canakinumab (NCT04362813 and NCT04365153). Additionally, several investigational agents that had entered clinical trials for other indications are now in COVID-19 studies, including three CSF2-targeting agents, gimsilumab (NCT04351243), lenzilumab (NCT04351152) and TJ003234 (NCT04341116); two C3-targeting agents, AMY-101 (NCT04395456) and APL-9 (NCT04402060); the C5-targeting agent IFX-1 (NCT04333420); and the IL-6-targeting agent clazakizumab (NCT04343989, NCT04348500, NCT04351724, NCT04363502 and NCT04381052).

Drugs that target ligands are economically successful

Therapeutic agents that target ligands have proved financially successful. From 2010 to 2018, the number of unique therapeutic antibodies in the top 200 performing drugs based on US retail sales rose from 2 to 29, and roughly half of these drugs have consistently targeted ligands124. Furthermore, in the last 3 years, ligand-targeting drugs have been responsible for nearly 20% of US sales among the top 200 drugs, and they account for more than 50% of all therapeutic antibody sales124. According to a recent analysis of the financial performance of agents modulating the top 20 drug targets, 24 FDA-approved drugs that target 10 cytokines plus the growth factors VEGFA and VEGFB had more than US$250 billion in cumulative sales from 2011 to 2015 (ref.125). The cytokine TNF was the most valuable target, with sales generating more than US$163 billion dollars from 2011 to 2015 (ref.125). Recently approved drugs that target ligands are also expected to generate large financial returns, such as the IL-23-targeted mAbs guselkumab and risankizumab, which are forecast to have sales in excess of US$3.4 billion by 2022 (ref.126) and US$3.2 billion by 2024 (ref.127), respectively.


Our analysis highlights the increased number of ligand-targeting drugs approved in the past decade compared with the previous one, as well as substantial interest in agents for novel ligand targets in clinical trials. Given the increasing body of evidence that inflammation is involved in many diseases beyond typical inflammatory disorders such as rheumatoid arthritis and the position of ligand-targeting drugs at the forefront of anti-inflammatory therapies, there will be continued interest in this class of agents. In this respect, there is vast territory to explore, as there are more than 580 endogenous peptides according to the IUPHAR/BPS Guide to Pharmacology, with more than 140 of these identified as specific immunopharmacology ligands that have curated immunological data associated with them128. Furthermore, the potential to target ligands also continues to be expanded by technological accomplishments such as bispecific constructs, antibody fragments and innovative engineering approaches such as sweeping antibodies129. In conclusion, while ligands as a type of drug target were previously categorized as ‘miscellaneous’ or ‘other’5, they now merit consideration as a distinct and expanding group of targets for a range of biopharmaceutical agents that can exert their effects through single targets or in selected combinations.