Agonistic bispecific antibodies: an emerging frontier in antibody therapeutics

Abstract

Antibody-based cancer therapies have long been valued for their high target specificity and favorable safety profiles compared with conventional chemotherapy and radiotherapy. The field has expanded rapidly, driven in part by advances in antibody engineering and the emergence of antibody-derived modalities. Among these, bispecific antibodies (bsAbs) and agonists have independently reshaped the therapeutic landscape by enabling novel mechanisms of action. However, the clinical development of agonists has been hampered by challenges such as systemic toxicity arising from excessive receptor activation and limited efficacy associated with inadequate receptor clustering or context-dependent signaling. To address these limitations, a distinct class of engineered molecules—agonistic bsAbs—has recently emerged as a promising therapeutic strategy. By spatially and contextually restricting receptor activation, these antibodies aim to enhance agonistic potency while improving safety and therapeutic index. Notably, in addition to their roles in cancer therapy, agonistic bsAbs also exhibit therapeutic potential in metabolic diseases. In this review, we classify agonistic bsAbs according to the nature of the receptors requiring activation, including immune co-stimulatory receptors, death receptors, and growth factor receptors. For each category, we summarize representative agonistic bsAbs, with a focus on their molecular design principles, mechanisms of action, preclinical and clinical progress, and current limitations. Together, this overview highlights agonistic bispecific antibodies as an evolving and versatile platform with the potential to redefine antibody-based therapeutics.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-08684-z.

Introduction

Over the past century, several major therapeutic modalities, including cytotoxic chemotherapy and radiation therapy, have been established for the treatment of cancer. Despite their substantial clinical impact, these approaches generally lack sufficient target specificity and often cause unacceptable systemic toxicity. In contrast, monoclonal antibodies (mAbs) offer a high degree of molecular precision, enabling selective targeting of disease-associated antigens [1, 2]. The therapeutic potential of antibodies was unlocked by several foundational technologies: the hybridoma technique allowed the generation of murine mAbs with high specificity; subsequent chimeric antibodies improved pharmacokinetic properties and reduced immunogenicity; fully human antibodies were made possible through transgenic mice and phage display [3]. Nevertheless, the clinical utility of mAbs can be limited by their single-target nature, which may be insufficient to overcome pathway redundancy or compensatory mechanisms in complex diseases (Table 1).

Table 1.

The key differences between mAbs, bsAbs, agonists and agonistic bsAbs

Modality	Design	MoA	Advantages	Limitations
Monoclonal antibodies (mAbs)	Single-antigen targeting	Target blockade or effector-mediated cytotoxicity	High target specificity; Established development platforms; Favorable safety profile	Single-target activity; Limited ability to overcome pathway redundancy; Restricted biological functionality
Bispecific antibodies (bsAbs)	Dual-antigen targeting	Simultaneous engagement of two distinct targets	Dual-pathway modulation; Enhanced therapeutic efficacy; Cell-bridging capability	Increased molecular complexity; Manufacturing challenges; Potential for unexpected toxicities
Agonists	Receptor-targeted agonism	Receptor activation and downstream signaling	Potent biological activity; Direct pathway modulation; Broad therapeutic applicability	Limited spatial selectivity; On-target off-tissue toxicity; Risk of excessive pathway activation
Agonistic bsAbs	Dual-target agonistic architecture	Conditional receptor activation and spatially restricted signaling	Enhanced selectivity; Context-dependent agonism; Improved therapeutic index	Complex molecular engineering; Manufacturing challenges; Safety optimization requirements

To overcome the limitations associated with single-target therapies, bispecific antibodies (bsAbs) have been developed to simultaneously bind two distinct antigens to coordinate or bridge two biological mechanisms in a single molecule. By dual pathway modulation, bsAbs can improve efficacy, enable cell-bridging and even reduce need for combination therapies. Although bsAbs have demonstrated potent antitumor activity in preclinical models, these advantages come at the cost of increased molecular complexity, more challenging manufacturing, and unexpected biological effects and toxicities may emerge [4] (Table 1). Nevertheless, continued advances in bsAbs platforms highlight their considerable potential to further expand the therapeutic landscape of cancer treatment. Meanwhile, researchers have focused on agonists, molecules that activate a receptor or signaling pathway, enhancing physiological signaling to stimulate a desired biological response. By direct pathway activation, agonists can drive strong biological effects and be useful for immune stimulation, tissue regeneration, metabolic regulation, and cell expansion. While agonists can potently stimulate signaling pathways, their efficacy often requires receptor clustering or crosslinking, and uncontrolled activation may lead to systemic toxicity [5] (Table 1). These intrinsic limitations have prompted the development of engineered agonistic modalities that aim to achieve more controlled and context-dependent receptor activation.

Building on the concept of agonists, researchers have sought to combine receptor activation with the dual-targeting capabilities of bsAbs to enhance both potency and specificity. Agonistic bispecific antibodies (agonistic bsAbs) represent a specialized subclass of bsAbs that are designed to actively induce receptor signaling by simultaneously engaging two distinct antigens, typically one expressed on immune cells and the other on tumor cells or accessory cells within the tumor microenvironment, yet both antigens can also be co-expressed by a single cell population. In contrast to conventional bsAbs, which primarily function by blocking ligand–receptor interactions or enhancing effector mechanisms such as antibody-dependent cellular cytotoxicity (ADCC), agonistic bsAbs are engineered to promote productive receptor clustering or conformational activation, thereby directly triggering immune signaling pathways. A prominent example involves the conditional co-stimulation of immune receptors, such as CD28, leading to enhanced T-cell activation, proliferation, and effector function, ultimately resulting in robust antitumor responses [6]. Importantly, compared with monospecific agonistic antibodies—which have often been limited by narrow therapeutic windows and dose-limiting toxicities—agonistic bsAbs offer the potential to spatially and contextually restrict receptor activation, thereby improving efficacy while reducing systemic toxicity. Nonetheless, clinical development of agonistic bsAbs carries potential for novel safety risks, including excessive immune activation and cytokine release syndrome. Translating preclinical findings into the clinic remains challenging due to unpredictable efficacy and toxicity. Furthermore, their highest molecular design complexity and manufacturing difficulties add additional hurdles to development (Table 1). Therefore, substantial room remains for further exploration and development in this field.

Notably, beyond their essential roles in the treatment of cancer, agonistic bsAbs targeting growth factor receptors also exhibit prominent therapeutic potential in metabolic diseases. In this review, we classify agonistic bsAbs according to the nature of the receptors requiring activation, including immune co-stimulatory receptors, death receptors, and growth factor receptors. For each category, we summarize representative agonistic bsAbs with respect to their molecular design principles, mechanisms of action, preclinical and clinical progress, and current limitations. Representative agonistic bispecific antibodies were selected based on published preclinical evidence, clinical investigation, or publicly disclosed development programs. Literature was primarily retrieved from PubMed (2010–2025) using keywords related to agonistic antibodies and bispecific antibodies.

Agonistic bispecific antibodies targeting immune co-stimulatory receptors

During adaptive immune responses, full T cell activation requires the integration of two distinct signals. The first signal is delivered through engagement of the T cell receptor (TCR) upon recognition of antigen-derived peptides presented by antigen-presenting cells (APCs) in the context of major histocompatibility complex (MHC) molecules. The second signal, referred to as co-stimulation, is provided by the interaction between immune co-stimulatory receptors expressed on immune cells—including T cells, B cells, dendritic cells, and natural killer cells—and their cognate ligands on APCs [7]. Engagement of co-stimulatory receptors initiates downstream signaling cascades, most notably the NF-κB and MAPK pathways, thereby promoting T cell proliferation, cytokine production, survival, and cytotoxic effector function [8]. In the absence of appropriate co-stimulatory signaling, T cells may enter a state of anergy or tolerance despite antigen recognition, underscoring the critical role of co-stimulation in productive immune activation. Immune co-stimulatory receptors are broadly categorized into two major families: the co-stimulatory members of tumor necrosis factor receptor superfamily (TNFRSF), comprising receptors such as CD40, CD137 (4-1BB), OX40 (CD134), and CD27, and the CD28 superfamily, which includes CD28 and inducible T cell co-stimulator (ICOS) [7]. Activation of these receptors enhances T cell survival, metabolic fitness, and effector differentiation, thereby supporting both the initiation of adaptive immune responses and the formation of durable immunological memory. Collectively, these processes are essential not only for effective antitumor immunity but also for the regulation of immune responses in chronic infections and autoimmune diseases [9].

CD40

CD40 is a member of the TNFRSF and is expressed on a broad range of immune cells, including dendritic cells, B cells, macrophages, and monocytes. Therapeutic activation of CD40 has the potential to enhance antigen-presenting cell (APC) function, stimulate robust antitumor T cell responses, and reprogram tumor-associated macrophages toward a tumoricidal phenotype capable of remodeling the tumor stroma [10]. However, the clinical development of CD40 agonist immunotherapies has historically been limited by significant systemic toxicities associated with widespread receptor activation. To overcome these challenges while preserving therapeutic efficacy, substantial efforts have focused on strategies that spatially restrict CD40 agonism, among which bsAbs represent a particularly promising approach. Accordingly, several CD40-targeting agonistic bsAbs have been developed and are discussed below [11].

ABBV-428 is a bsAb designed to simultaneously bind CD40 and the tumor antigen mesothelin (MSLN) (Fig. 1A and Supplementary Table 1), thereby confining CD40 activation to the tumor microenvironment and minimizing systemic immune activation. Preclinical studies demonstrated that ABBV-428 effectively activates APCs and enhances downstream T cell responses in a mesothelin-dependent manner, resulting in significant tumor growth inhibition in xenograft models. Structurally, ABBV-428 is based on an Fc-engineered IgG scaffold with appended single-chain antibody fragments, a design intended to restrict productive CD40 cross-linking to mesothelin-expressing sites and thereby promote tumor-localized immune activation with reduced peripheral exposure. These findings provided a strong rationale for subsequent clinical evaluation [12–14]. In a first-in-human phase 1 study (NCT02955251), 59 patients with advanced solid tumors were treated at doses between 0.01 and 3.6 mg/kg. The maximum tolerated dose (MTD) was not reached, and 3.6 mg/kg was selected as the recommended phase 2 dose (RP2D). The most common treatment-related adverse events (TRAEs) were fatigue (41%), decreased appetite (36%), asthenia (29%), constipation (25%), and dyspnea (25%). A total of 36 patients (61%) reported grade ≥ 3 TEAEs (Table 2) and no dose-limiting toxicities (DLTs) were observed. The pharmacokinetic (PK) data revealed that drug exposure increased in an approximately dose-proportional manner across the 0.4 to 3.6 mg/kg dose range. 9 (36%) of 25 subjects with ovarian cancer and mesothelioma, at the RP2D, achieved stable disease (SD). Over 90% peripheral CD40 receptor occupancy on B cells was attained starting at 0.8 mg/kg. However, in both the dose escalation phase and a small expansion cohort of patients with advanced mesothelioma or ovarian cancer, ABBV-428 monotherapy demonstrated minimal clinical activity. It suggests that higher dosing, optimized scheduling, or the selection of tumors with higher mesothelin expression may be required to enhance therapeutic efficacy, either as monotherapy or in combination regimens [15].

Fig. 1.

Overview of representative agonistic bsAbs targeting key receptor classes in cancer therapy. (A) Immune co-stimulatory receptors: BsAbs engage immune checkpoint ligands or tumor-associated antigens with co-stimulatory receptors (e.g., CD40, CD137, OX40, CD27, CD28, ICOS) to enhance T-cell activation and antitumor immunity. Representative examples include ABBV-428 (CD40×MSLN), CEA-CD40 (CD40×CEA), RO7300490 (CD40×FAP), GEN1042 (CD40×CD137), Cinrebafusp alfa (HER2×CD137), ATG-101 (PD-L1×CD137), MCLA-145 (PD-L1×CD137), AP203 (PD-L1×CD137), PD-L1/OX40 bsAb (OX40×PD-L1), FS120 (CD137×OX40), CDX-527 (PD-L1×CD27), RO7443904 (CD19×CD28), REGN7075 (EGFR×CD28), XmAb23104 (PD-1×ICOS), and KY1055 (ICOS×PD-L1). (B) Death receptors: MCSP×DR5, BI 905,711 (CDH17×DR5) and TR2/CDH3 BAB (CDH3×DR5) selectively engage death receptors on tumor cells to induce apoptosis, exploiting tumor-restricted surface antigens to minimize off-tumor toxicity. (C) Growth factor receptors: BFKB8488A (KLB×FGFR1) mimics FGF21 activity to modulate metabolic pathways, including lipid and glucose metabolism, with potential implications for metabolic diseases. Orange bars indicate clinical-stage agents, blue bars indicate preclinical-stage agents, and gray bars indicate agents that have not yet entered clinical development

Table 2.

Agonistic bsAbs in clinical development for tumor therapy

Name	Institute/ Company	Target	Phase/ clinical trial ID	Trial design/ Patient population	Efficacy	Safety	Reference
ABBV-428	AbbVie Inc., AbbVie	CD40, MSLN	Phase 1/ NCT02955251	ABBV-428 or ABBV-428 + nivolumab in advanced solid tumors refractory to all standard therapies	DCR 36% (9 SD)	Most common TRAEs: fatigue (41%), decreased appetite (36%), asthenia (29%), constipation (25%), and dyspnea (25%). Grade ≥ 3 TRAEs in 36 (61%) patients.	[15]
RO7300490	Hoffmann-La Roche	CD40, FAP	Phase 1/ NCT04857138	RO7300490 or RO7300490 + atezolizumab in locally advanced or metastatic solid tumors refractory to all standard therapies	DCR 53.8% (14 SD)	Most common TRAEs: low-grade arthralgia (31%). Grade ≥ 3 TRAEs in 2 (7.7%) patients.	[16]
GEN1042	Genmab	CD40, CD137	Phase 1/2 NCT04083599	GEN1042 or GEN1042 + pembrolizumab or GEN1042 + pembrolizumab + chemotherapy in metastatic or locally advanced solid tumors	DCR 4%, ORR 4% (2 PR)	DLT in 1 patient.	[17]
Cinrebafusp alfa	Pieris Pharmaceuticals, Inc	HER2, CD137	Phase 1/ NCT03330561; Phase 2/ NCT05190445	Cinrebafusp alfa in HER2 + advanced or metastatic solid tumors (NCT03330561); Cinrebafusp alfa + ramucirumab + paclitaxel in HER2-low gastric or gastroesophageal junction (GEJ) adenocarcinoma (NCT05190445)	NCT03330561 DCR 52.5%, ORR 12.5% (16 SD, 5 PR)	NCT03330561 Most common TRAEs: IRRs (21.8%), fatigue (14.1%), and nausea (11.5%). Grade ≥ 3 TRAEs in 12 (15.4%) patients. DLT in 1 patient.	[18]
ATG-101	Antengene Corporation	PD-L1, CD137	Phase 1/ NCT04986865; Phase 1/ NCT05490043	ATG-101 in metastatic/advanced solid tumors and mature B-cell non-Hodgkin lymphomas	N/A	N/A	N/A
MCLA-145	Merus N.V.	PD-L1, CD137	Phase 1/ NCT03922204	MCLA-145 or MCLA-145 + pembrolizumab in advanced or recurrent/metastatic solid tumors or B-cell lymphomas refractory to all standard therapies	Monotherapy: DCR 37%, ORR 9.6% (14 SD, 5 PR) Combination: DCR 68%, ORR 5.3% (12 SD, 1 PR)	Monotherapy Most common TRAEs: fatigue (51%), decreased appetite (34%), dyspnea (32%), anemia (30%), ALT/AST increase (25%), and pyrexia (25%). DLTs in 6 patients. Combination Most common TRAEs: fatigue (58%), cough (32%), constipation (32%), and ALT/AST increase (21%).	[19]
AP203	AP Biosciences Inc.	PD-L1, CD137	Phase 1/2/ NCT05473156	AP203 in locally advanced or metastatic solid tumors	N/A	N/A	N/A
FS120	invoX Pharma Limited	OX40, CD137	Phase 1/ NCT04648202	FS120 in locally advanced, unresectable or metastatic solid tumors	N/A	N/A	N/A
CDX-527	Celldex Therapeutics	PD-L1, CD27	Phase 1/ NCT04440943	CDX-527 in recurrent, locally advanced or metastatic solid tumor cancer excluding the following: MSI-low colorectal cancer, glioblastoma multiforme, prostate cancer, pancreatic cancer, mucosal and ocular melanoma	N/A	N/A	N/A
RO7443904	Hoffmann-La Roche	CD28, CD19	Phase 1/ NCT05219513	RO7443904 + glofitamab in relapsed/refractory B-cell non-Hodgkin’s lymphoma	(aNHL) ORR 64%, CRR 39% (FL) ORR 90%, CRR 90%	Most common TRAEs: CRS (54%), anemia (27%), ALT increase (19%), COVID-19 (19%), diarrhea (19%) and thrombocytopenia (19%).	[20]
REGN7075	Regeneron Pharmaceuticals	CD28, EGFR	Phase 1/2 NCT04626635; Phase 2 NCT06465329	REGN7073 + cemiplimab in advanced solid tumors (NCT04626635); REGN7073 + cemiplimab + chemotherapy in non-small cell lung cancer (NSCLC) (NCT06465329)	NCT04626635 (MSS CRC without liver metastases) DCR 80%, ORR 20%, CRR 6.6% (9 SD, 2 PR, 1 CR) (MSS CRC with liver metastases) DCR 8.3% (3 SD)	NCT04626635 Most common TRAEs: fatigue, increases in AST, diarrhea, hypothyroidism, pyrexia, and rash. Grade ≥ 3 TRAEs in 6 (7%) patients.	[21]
XmAb23104	Xencor, Inc; Adil Daud, University of California, San Francisco; Memorial Sloan Kettering Cancer Center	PD-1, ICOS	Phase 1/ NCT03752398; Phase 1b/2 NCT05695898; Phase 2 NCT05879185	XmAb23104 in selected advanced solid tumors (NCT03752398); XmAb23104 + XmAb22841 in advanced/metastatic melanoma (NCT05695898); XmAb23104 in advanced sarcoma (NCT05879185)	NCT03752398 DCR 8.1%, ORR 3.2% (2 SD, 3 PR)	NCT03752398 Most common TRAEs: diarrhea (9.7%), decreased appetite (9.7%), and fatigue (9.7%). Grade ≥ 3 TRAEs in 6 (9.7%) patients.	[22]
BI 905,711	Boehringer Ingelheim	DR5, CDH17	Phase 1/ NCT04137289; Phase 1/ NCT05087992	BI 905,711 in advanced unresectable or metastatic gastrointestinal cancers or colorectal adenocarcinoma (NCT04137289); BI 905,711 + chemotherapy in advanced unresectable or metastatic colorectal adenocarcinoma (NCT05087992)	NCT04137289 DCR 22.1% (23 SD) NCT05087992 DCR 91.7% (11 SD)	NCT04137289 Most common TRAE: nausea (16.4%). NCT05087992 Most common TRAEs: decreased appetite (33.3%), increased ALT (25%), increased AST (25%), and diarrhea (25%). DLTs in 2 patients.	[23]

MSLN, mesothelin; FAP, fibroblast activation protein alpha; HER2, human epidermal growth factor receptor 2; PD-L1, programmed death-ligand 1; EGFR, epidermal growth factor receptor; PD-1, programmed cell death protein 1; ICOS, inducible T-cell co-stimulator; CDH, cadherin; DCR, disease control rate; ORR, objective response rate; CRR, complete response rate; SD, stable disease; PR, partial response; CR, complete response; aNHL, aggressive non-Hodgkin lymphoma; FL, follicular lymphoma; MSS CRC, microsatellite stable colorectal cancer; TRAEs, treatment-related adverse events; DLT, dose-limiting toxicity; IRRs, infusion-­related reactions; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CRS, cytokine release syndrome; N/A, not applicable. For ongoing studies without published efficacy or safety results, corresponding entries are indicated as N/A

CEA-CD40 is a bispecific CD40 agonist engineered to bind CD40 and the tumor-associated antigen carcinoembryonic antigen (CEA) (Fig. 1A and Supplementary Table 1), thereby confining CD40 activation to CEA-expressing tumor cells and minimizing systemic CD40 engagement. In preclinical in vitro assays, CEA-CD40 was shown to induce potent dendritic cell (DC) activation and subsequent CD8⁺ T cell cross-priming only in the presence of CEA, while exhibiting little to no activation in its absence. Mechanistically, CEA-CD40 promotes the co-localization of CEA⁺ tumor-derived extracellular vesicles (EVs) with DCs, thereby facilitating the acquisition of EV-associated MHC class I/peptide complexes and enhancing antigen cross-presentation to naïve CD8⁺ T cells. These properties contribute to its favorable safety profile, as FcγR binding is silenced through Fc mutations to prevent systemic cross-linking, and receptor activation is driven by clustering on CEA⁺ cells [24]. To date, CEA-CD40 has not entered clinical testing, and all available data remain preclinical.

Fibroblast activation protein α (FAP)–targeted CD40 bsAb, RO7300490, was developed to confine CD40 agonism to the tumor stroma and to widen the therapeutic window of CD40 activation (Fig. 1A and Supplementary Table 1). Preclinical studies demonstrated that RO7300490 induces robust intratumoral dendritic cell maturation and T-cell inflammation, resulting in markedly enhanced tumor growth inhibition in multiple syngeneic and humanized tumor models, while substantially attenuating the systemic toxicities that have limited non-targeted CD40 agonists. Mechanistically, anchoring CD40 agonism to FAP-expressing stromal elements concentrates receptor cross-linking within the tumor microenvironment, thereby enabling the administration of higher biologically active doses that overcome peripheral CD40 sink effects [25]. These findings supported first-in-human evaluation, and RO7300490 has entered phase 1 trial (NCT04857138). In this study, 29 patients with advanced solid tumors were treated with RO7300490 at doses ranging from 16 mg to 1100 mg. The majority of adverse events (AEs) were mild to moderate and non-serious. The most common TRAE was low-grade arthralgia (31%). While Grade 3 TRAEs occurred in two patients, no Grade 4 or 5 TRAEs were reported. A single case of Grade 1 CRS was recorded. Three AEs resulted in treatment withdrawal, two of which were deemed related to RO7300490. No DLTs were observed. Notably, there was no evidence of a correlation between dose levels and the incidence or severity of AEs. The best overall response observed was SD, achieved in 14 of 26 evaluable patients (Table 2). PK analysis indicated non-linear pharmacokinetics at lower doses, with a trend toward linear pharmacokinetics at higher doses [16]. Collectively, FAP-directed CD40 agonism exemplifies a stromal-targeting strategy for conditional immune activation within the tumor microenvironment and remains of strong translational interest, particularly for combination regimens aimed at converting immunologically cold tumors [12].

BiA92_HF represents a distinct class of CD40 agonistic bsAbs engineered to enable the rapid formation of antibody–peptide conjugates, thereby coupling antigen delivery with concurrent CD40-mediated activation of APCs. In vitro and in vivo studies demonstrated that, when loaded with peptide cargo, BiA92_HF markedly enhances peptide-specific CD8⁺ and CD4⁺ T cell proliferation in draining lymph nodes and significantly improves antigen cross-priming compared with peptide alone or control antibodies. In neoantigen- and virus antigen–driven mouse tumor models, BiA92_HF combined with peptide cargo produced robust antitumor efficacy, including complete tumor regressions in select cohorts and measurable abscopal effects, while maintaining a favorable safety profile relative to non-targeted CD40 agonists. Mechanistically, this format is designed to concentrate both antigen payloads and CD40 agonism on CD40-expressing myeloid cells, thereby enhancing dendritic cell activation and antigen presentation without inducing systemic CD40 cross-linking [26]. To date, BiA92_HF remains at the preclinical stage.

GEN1042 (DuoBody-CD40×CD137/BNT312) is a novel agonistic bsAb that combines targeting of costimulatory molecules CD40 and CD137 (Fig. 1A). Constructed via controlled Fab‑arm exchange with Fc‑silencing mutations, it eliminates FcγR and C1q binding to avoid off‑target activation (Supplementary Table 1). Its conditional agonism depends on simultaneous dual‑target crosslinking, strengthening the DC‑T cell immunological synapse, inducing dendritic cell maturation, and enhancing CD8⁺ T‑cell proliferation, cytokine secretion, and cytotoxicity. Preclinically, it promotes ex vivo expansion of tumor‑infiltrating lymphocytes and synergizes with PD‑1 blockade [27]. In an ongoing phase 1/2 trial (NCT04083599), GEN1042 displayed controllable safety across 0.1–400 mg Q3W without identified MTD, with only one DLT (grade 4 transaminase elevation) observed at 200 mg. Among 50 evaluable patients, two partial responses (PRs) cases occurred at 3 mg and 30 mg respectively (Table 2). On the basis of Pharmacodynamics (PD) data, doses ≥ 30 mg triggered obvious elevation of peripheral IFN‑γ and TARC alongside robust T cell activation, and 100 mg Q3W was finally selected as the recommended expansion dose based on modeling prediction and clinical PD/efficacy data [17] (Table 2). However, GEN1042 is limited by dose-dependent hepatotoxicity, narrow effective dose window from bell-shaped pharmacodynamic curve and poor efficacy against cold tumors [27].

In addition to the strategies described above, several other CD40-targeting bsAbs have been reported. The anti–PD-L1/CD40 bsAb BA4415 was designed to activate CD40 signaling in a PD-L1–dependent manner while simultaneously blocking PD-1/PD-L1 immune checkpoint interactions [28]. Other CD40×tumor-associated antigen (TAA) bsAbs leverage TAA-binding domains to capture tumor-derived EVs, while the CD40-binding domain directs these vesicles to dendritic cells, thereby facilitating EV uptake and neoantigen cross-presentation that promotes tumor-specific T cell expansion [29]. Collectively, these CD40 bsAbs achieve localized CD40 agonism through divergent dual-targeting designs to synergistically boost anti-tumor immune responses.

CD137(4-1BB)

CD137 (4-1BB, TNFRSF9) is a co-stimulatory receptor belonging to the TNFRSF and plays a pivotal role in antitumor immunity by promoting the activation and survival of cytotoxic T cells and natural killer (NK) cells, as well as by inducing type I interferon (IFN) production. CD137 expression on CD8⁺ and CD4⁺ T cells is inducible and markedly upregulated following T-cell receptor (TCR) stimulation [30]. Two first-generation agonistic CD137 monoclonal antibodies, utomilumab (PF-05082566) and urelumab (BMS-663513), were developed to therapeutically exploit this pathway. However, their clinical development was limited by modest antitumor efficacy in the case of utomilumab and dose-limiting hepatotoxicity associated with urelumab [31, 32]. These divergent clinical outcomes have been attributed, at least in part, to differences in CD137 epitope engagement and reliance on Fcγ receptor (FcγR)–mediated receptor clustering for agonistic activity [33]. To overcome these limitations, several next-generation CD137 agonistic bsAbs have been developed with the aim of enhancing efficacy while minimizing systemic toxicity.

Cinrebafusp alfa (PRS‑343) is arguably the first agent to reach the clinic with a tumor-localized CD137 clustering strategy. Distinct from conventional agonistic bsAbs, Cinrebafusp alfa is a bsAb‑Anticalin fusion protein targeting HER2 and CD137 (Fig. 1A). It comprises an engineered anti‑HER2 IgG4 fused to a 4‑1BB‑specific Anticalin, with a silenced Fc domain that abrogates FcγR binding and effector functions to avoid nonspecific T-cell activation (Supplementary Table 1). It acts by bridging tumor cells and T cells to induce tumor‑localized 4‑1BB clustering and T‑cell costimulation. Preclinically, it induced HER2‑dependent T‑cell activation and tumor growth inhibition with minimal systemic toxicity [34]. In a phase 1 trial (NCT03330561), for safety analysis, 78 patients across all dose-escalation cohorts (0.0005–18 mg/kg) who received at least one dose of the study drug were included. Overall, single-agent Cinrebafusp alfa demonstrated a favorable safety and tolerability profile. Only one DLT was observed: a patient with breast cancer in the 12 mg/kg Q2W cohort developed grade 3 pleural effusion on day 28, leading to treatment discontinuation. Furthermore, no MTD was established at doses up to 18 mg/kg. TRAEs occurred in 44 patients (56.4%), including grade 3 events in 11 patients (14.1%) and grade 4 events in 1 patient (1.3%). No grade 5 TRAEs were reported. The most common TRAEs were infusion-­related reactions (IRRs) (21.8%), fatigue (14.1%), and nausea (11.5%). Among 40 evaluable patients in active dose cohorts (2.5–18 mg/kg), the overall response rate reached 12.5% and disease-control rate (DCR) was 52.5%, with cohort-specific objective response rates (ORR) of 28.6% at 8 mg/kg Q2W and 25.0% at 18 mg/kg Q2W [18] (Table 2). A phase 2 trial (NCT05190445) of Cinrebafusp alfa in combination with ramucirumab and paclitaxel has been initiated for HER2-positive gastric or gastroesophageal junction cancer, yet no efficacy or safety data have been published (Table 2). Limitations of Cinrebafusp alfa include efficacy restricted to HER2‑positive tumors, low response rates as monotherapy, and immunogenicity at lower doses [34].

ATG-101 is a tetravalent agonistic bsAb targeting programmed death-ligand 1 (PD-L1) and CD137 (Fig. 1A and Supplementary Table 1), designed to simultaneously block PD-L1/PD-1 interactions and conditionally agonize CD137 within PD-L1–enriched tumor microenvironments. In preclinical models, ATG-101 demonstrated robust antitumor activity in tumors resistant to immune checkpoint inhibitors, accompanied by increased infiltration of effector CD8⁺ T cells, expansion of effector memory T-cell populations, and favorable CD8⁺/regulatory T-cell ratios within the tumor microenvironment, while exhibiting minimal hepatotoxicity in non-human primates [35]. Biodistribution studies, including immuno-PET imaging with ⁸⁹Zr-labeled ATG-101, confirmed preferential tumor accumulation and PD-L1–dependent tissue uptake in murine xenograft models [36]. Consistent with these findings, GLP toxicology studies indicated a favorable safety profile. Clinically, ATG-101 has entered first-in-human phase 1 trials (NCT04986865/ NCT05490043) in patients with advanced or metastatic solid tumors and mature B-cell non-Hodgkin lymphomas (Table 2), with primary endpoints including safety, dose-limiting toxicity, pharmacodynamics, and preliminary efficacy.

MCLA-145 is a bivalent agonistic bsAb that co-engages PD-L1 on tumor or antigen-presenting cells and CD137 on T cells (Fig. 1A and Supplementary Table 1), thereby licensing CD137 agonism in a PD-L1–dependent, tumor-proximal manner to limit systemic CD137 activation. Mechanistically, MCLA-145 promotes the re-localization of PD-L1 and CD137 to the immune synapse and induces clustering of CD137 receptors on the T-cell membrane, resulting in conditional co-stimulatory signaling that correlates with PD-L1 expression levels (Fig. 2). Preclinical in vitro and in vivo studies demonstrated that MCLA-145 selectively induces PD-L1–restricted CD137 signaling, enhances intratumoral CD8⁺ T-cell proliferation and memory formation, and overcomes immunosuppressive mechanisms mediated by regulatory T cells and macrophages, leading to significant antitumor activity in both xenograft and syngeneic models [37, 38]. In an ongoing phase 1 clinical study (NCT03922204), 72 patients, 53 in monotherapy and 19 in combination(MCLA-145 combines with pembrolizumab), across 26 cancer types had been treated. In the monotherapy arm, 6 patients had DLTs in the 25-75 mg dose range. The most common TRAEs were fatigue (51%), decreased appetite (34%), dyspnea (32%), anemia (30%), ALT/AST increase (25%), and pyrexia (25%). In the combination arm, no DLTs were reported. The most common TRAEs were fatigue (58%), cough (32%), constipation (32%), and ALT/AST increase (21%). Consequently, the recommended dose for expansion (RDE) was established as 40 mg Q3W for both monotherapy and combination regimens. For the monotherapy with 52 evaluable patients, 5 PRs were observed and DCR was 37%, simultaneously, for the combination with 19 evaluable patients, 1 PR was observed and DCR was 68% (Table 2). Drug exposure was dose-dependent (terminal half-life: 69 h at 75 mg), and peripheral blood Ki67 + CD8+ T cell data supported maximal PD activity at 40 mg [19].

Fig. 2.

Mechanism of action of MCLA-145. MCLA-145 is a bsAb that simultaneously engages PD-L1, expressed on tumor cells or antigen-presenting cells (APCs; shown in pink), and CD137, expressed on effector T cells (shown in green). This dual engagement induces clustering of CD137 receptors on the T-cell membrane, thereby activating downstream NF-κB signaling. In parallel, PD-L1 binding by MCLA-145 blocks its interaction with PD-1, relieving PD-1–mediated suppression of T-cell receptor (TCR) signaling that activates downstream NF-κB and NFAT signaling. Trans-binding of MCLA-145 between tumor cells/APCs and effector T cells orchestrates a coordinated immunomodulatory response, including upregulation of CD137 on T cells and PD-L1 on tumor cells/APCs, secretion of pro-inflammatory cytokines, promotion of T-cell differentiation and clonal expansion, and enhanced effector T-cell activation. Collectively, these processes facilitate the cytotoxic elimination of tumor cells

AP203 is a tetravalent agonistic bsAb targeting PD-L1 and CD137 (Fig. 1A and Supplementary Table 1), designed to integrate immune checkpoint blockade with conditional co-stimulatory signaling to enhance antitumor immune responses. In vitro, AP203 exhibited high-affinity binding to both PD-L1 and CD137 at sub-nanomolar levels and induced potent T-cell activation, including IL-2 and IFN-γ secretion, in mixed lymphocyte reactions, memory recall assays, and co-culture systems with PD-L1–expressing tumor cells, outperforming parental monospecific antibodies administered either alone or in combination. In vivo, AP203 elicited dose-dependent antitumor activity in both humanized and immunocompetent mouse tumor models, accompanied by increased CD8⁺ T-cell infiltration, reduced regulatory T-cell frequencies, and elevated CD8⁺/CD4⁺ ratios, without evidence of excessive inflammatory cytokine release in human PBMC assays. Mechanistic analyses confirmed that CD137 agonism by AP203 is strictly dependent on PD-L1–mediated cross-linking, thereby limiting off-tumor receptor activation [39]. The U.S. FDA cleared the investigational new drug (IND) application for AP203, enabling initiation of a phase 1/2 clinical trial (NCT05473156) to evaluate the safety, tolerability, pharmacokinetics, immunogenicity, pharmacodynamics, and antitumor activities of AP203 in adult patients with locally advanced or metastatic solid tumors (Table 2).

Although ATG-101, MCLA-145, and AP203 share a common design principle—dual targeting of PD-L1 and CD137 to integrate immune checkpoint blockade with conditional co-stimulation—they differ substantially in molecular architecture, Fc engineering, and activation dependency. ATG-101 is a Fc-silenced tetravalent bsAb that achieves balanced PD-L1 blockade and CD137 engagement, resulting in robust CD8⁺ T-cell activation and enhanced antitumor memory formation while maintaining favorable tolerability [35]. MCLA-145 is a bivalent bsAb, developed using the Biclonics^® platform, having a fully silenced Fc effector function and integrating dual checkpoint inhibition with co-stimulatory signaling, yielding potent immune activation in preclinical models and encouraging early clinical signals [37]. AP203, a tetravalent bsAb, employs a silenced Fc and a strictly PD-L1–dependent activation mechanism that confines potent T-cell co-stimulation to PD-L1⁺ tumor sites, thereby effectively minimizing systemic cytokine release [39]. Despite both being tetravalent agonistic bsAbs, AP203 and ATG-101 adopt distinct structural engineering designs in Fc region mutation profiles and interdomain linker configurations. AP203 incorporates a dual Fc-silencing mutation combination of N297A and K322A and utilizes a fixed short flexible linker of (GGGGS)₂ between the heavy chain and scFv domain [39]. In contrast, ATG-101 only carries a single N297A Fc mutation [35], and its relevant information of the scFv linker has not been disclosed. Collectively, PD-L1×CD137 bsAbs represent a new generation of tumor microenvironment–restricted immunomodulators designed to overcome the safety and efficacy limitations associated with conventional CD137 agonists.

OX40(CD134)

OX40 (CD134) is a co-stimulatory receptor belonging to the TNFRSF and is predominantly expressed on activated CD4⁺ and CD8⁺ T cells following antigen stimulation. Engagement of OX40 by its cognate ligand, OX40L (CD252), which is primarily expressed on antigen-presenting cells (APCs) such as dendritic cells, B cells, and macrophages, delivers a potent secondary co-stimulatory signal that promotes T-cell proliferation, survival, effector differentiation, and memory formation. Structurally, OX40–OX40L interactions involve trimeric receptor–ligand assemblies characteristic of TNFR family signaling, and efficient downstream activation requires higher-order receptor clustering, typically mediated by ligand multimerization or Fcγ receptor (FcγR)–dependent crosslinking. However, clinical trials evaluating agonistic anti-OX40 monoclonal antibodies as monotherapies have demonstrated limited efficacy, likely due to insufficient receptor clustering and suboptimal agonistic potency [40].

To overcome these limitations, a tetravalent PD-L1/OX40 bsAb was developed (Fig. 1A and Supplementary Table 1). This molecule is designed to exert antitumor activity through multiple complementary mechanisms. First, PD-L1 binding disrupts the PD-1/PD-L1 axis, thereby alleviating T-cell exhaustion. Second, agonistic engagement of OX40 enhances co-stimulatory signaling in activated T cells. Third, simultaneous binding to PD-L1 and OX40 promotes physical bridging between tumor cells and T cells, facilitating immune synapse formation. Finally, OX40 clustering and activation are conditionally enhanced within the PD-L1–enriched tumor microenvironment, thereby improving spatial selectivity. Importantly, the OX40-binding moiety does not interfere with endogenous OX40L–OX40 interactions, a feature that may contribute to an improved safety profile. In murine tumor models, the PD-L1/OX40 bsAb exhibited superior antitumor efficacy compared with combination therapy, with activity observed across a wide range of PD-L1 expression levels. These findings support further clinical development of PD-L1/OX40 bsAb for the treatment of advanced solid tumors [41].

In addition to tumor-targeted OX40 agonism, dual co-stimulatory strategies have also been explored. FS120 is a bsAb that simultaneously targets CD137 and OX40 (Fig. 1A and Supplementary Table 1). This molecule activates both CD4⁺ and CD8⁺ T cells through an FcγR-independent mechanism that requires concurrent engagement of both targets. A murine surrogate of FS120 demonstrated robust antitumor activity in syngeneic tumor models that was independent of regulatory T-cell (Treg) depletion and FcγR engagement, but correlated with enhanced peripheral T-cell activation and proliferation. Notably, compared with crosslink-independent CD137 agonists, FS120 induced lower levels of hepatic T-cell infiltration, supporting its favorable safety profile and clinical development potential [42]. Mechanistic studies further revealed that the antitumor efficacy of dual OX40/CD137 agonism was partially dependent on IFN-γ production by Tregs induced following FS120 treatment, a phenomenon not observed with anti–PD-1 therapy. These findings suggest a distinct immunomodulatory mechanism and provide a rationale for evaluating FS120 in combination with PD-1 blockade, such as pembrolizumab [43]. FS120 is currently undergoing a phase 1 clinical trial (NCT04648202) to evaluate its safety and antitumor activity, alone and in combination with pembrolizumab, in subjects with advanced malignancies (Table 2).

CD27

CD27 is a TNFRSF co-stimulatory receptor expressed on resting and activated αβ CD4⁺ T cells, including both conventional and regulatory subsets, CD8⁺ T cells, and a subset of innate-like Vγ9Vδ2 T cells with documented antitumor activity [44]. Engagement of CD27 by its ligand CD70 delivers potent co-stimulatory signals that enhance T-cell activation, proliferation, survival, and effector differentiation. Agonistic CD27 antibodies have demonstrated robust T-cell–activating and antitumor effects in both murine tumor models and human primary immune cell assays [45, 46].

Notably, CD27 agonism has shown pronounced synergy with PD-1/PD-L1 blockade. Transcriptomic analyses of CD8⁺ T cells revealed that CD27 agonists and PD-1 blockade induce partially distinct transcriptional programs associated with proliferation and cytotoxicity, respectively, which converge under combination therapy to amplify T-cell activation–related gene signatures [47]. These observations provided a strong rationale for integrating CD27 co-stimulation with immune checkpoint inhibition within a single molecular entity.

Accordingly, CDX-527, a tetravalent PD-L1×CD27 bsAb, was developed (Fig. 1A and Supplementary Table 1). CDX-527 is designed to simultaneously block PD-1 signaling while inducing CD27-mediated T-cell co-stimulation through PD-L1-dependent and Fc receptor–mediated crosslinking. In vitro studies demonstrated that CDX-527 exhibited greater functional potency than the combination of the corresponding parental monoclonal antibodies, consistent with enhanced CD27 agonism. In human CD27 transgenic mouse models, a PD-L1×CD27 surrogate bsAb significantly augmented vaccine-induced T-cell responses and achieved superior antitumor efficacy in syngeneic tumor models. Furthermore, a pilot toxicology study in cynomolgus macaques revealed favorable pharmacokinetics and no overt safety concerns, supporting ongoing clinical evaluation of CDX-527 [48]. Clinically, a phase 1 trial (NCT04440943) indicated that CDX-527 up to and including the 1 mg/kg dose level has been well tolerated (Table 2). Subsequent data readouts will further elucidate the safety profile at escalated dose levels, clinical activity, and comprehensive pharmacokinetic and pharmacodynamic parameters.

CD28

CD28 is a central co-stimulatory receptor required for full T-cell activation, survival, and differentiation. Interaction of CD28 with its ligands CD80 and CD86 on APCs provides the canonical second signal necessary for effective T-cell activation in conjunction with TCR engagement of peptide–MHC complexes. CD28 signaling promotes T-cell proliferation, cytokine production, and resistance to anergy, while also supporting long-term survival and memory formation. Given its pivotal role in immune activation, CD28 has long been considered an attractive target for cancer immunotherapy. However, excessive or systemic CD28 activation can lead to severe immune-related toxicities, underscoring the need for precise spatial and contextual control of CD28 agonism [49].

To address these challenges, tumor-targeted CD28 bispecific strategies have been developed. RO7443904 (RG6333, CD19-CD28) is a bsAb that simultaneously targets CD28 and CD19 (Fig. 1A and Supplementary Table 1), enabling localized delivery of CD28 co-stimulatory signals to T cells in proximity to CD19⁺ B-cell malignancies. By restricting CD28 activation to the tumor microenvironment, RO7443904 aims to enhance antitumor immunity while minimizing systemic immune activation and associated toxicities. Subsequent studies demonstrated that RO7443904 significantly potentiated the efficacy of glofitamab and related T-cell bsAbs (TCBs) by providing the required CD28-mediated co-stimulatory signal (signal 2) to tumor-infiltrating T cells. In combination with glofitamab, RO7443904 enhanced T-cell effector functions in ex vivo assays using lymphoma patient-derived PBMCs and splenocytes, and improved tumor regression in humanized mouse models. Moreover, incorporation of CD19-4-1BBL into a triple-combination regimen further improved durable tumor control compared with glofitamab monotherapy or dual combinations. Collectively, these findings support RO7443904 as a promising and controllable co-stimulatory platform [50]. Clinically, the phase 1 trial (NCT05219513) data showed that RO7443904 in combination with glofitamab demonstrated controllable safety and favorable clinical efficacy in relapsed/refractory B-cell non-Hodgkin’s lymphoma (B-NHL). A total of 53 patients were enrolled, including 33 aggressive NHL (aNHL) and 19 follicular lymphoma (FL). No MTD was identified across the tested dose range of 0.15 to 28 mg. The combination of RO7443904 and glofitamab had a manageable safety profile with no additive or new safety signals, and no DLTs were noted. The most common TRAEs were CRS (54%), anemia (27%), ALT increase (19%), COVID-19 (19%), diarrhea (19%) and thrombocytopenia (19%). All grade 5 events occurring in 4 (8%) patients were unrelated to the study treatment. The ORR and complete response rate (CRR) were 64% and 39% for aNHL, and 90% and 90% for FL [20] (Table 2). However, limited by CD19 expression, RO7443904 acts solely on CD19⁺ B-cell malignancies and fails against CD19⁻ lymphoma and solid tumors.

REGN7075, one of the first immunotherapeutics to show activity in microsatellite-stable colorectal cancer (MSS CRC), is an agonistic bsAb targeting epidermal growth factor receptor (EGFR) and CD28 (Fig. 1A), which is built on a human IgG4 backbone with Fc-region engineering to abrogate FcγR-dependent effector functions (Supplementary Table 1). It triggers T-cell costimulation via crosslinking EGFR-positive tumor cells with CD28-expressing T cells. The phase 1/2 study (NCT04626635) demonstrated that the combination of REGN7075 and cemiplimab yielded manageable safety and modest antitumor efficacy in patients with MSS CRC. For 84 patients with diverse solid tumors, REGN7075 combined with cemiplimab showed an acceptable safety profile without reaching the MTD. TRAEs were reported in 90% of patients (mostly Grade 1–2 at 83%, with 7% Grade 3/4). The most common TRAEs were fatigue, increases in AST, diarrhea, hypothyroidism, pyrexia, and rash. As of data cutoff, no DLTs, CRS, or treatment-related deaths were reported. For 51 patients with MSS CRC treated at an active dose level, including 1 complete response (CR), 2 PR, and 12 SD, the overall response rate and DCR were 6% and 29%. Among the subset of 15 patients without liver metastases, the ORR and DCR were 20% and 80%. Among the 36 patients with liver metastases, there were 3 SD as of data cutoff (Table 2), and 1 patient achieved a PR following data cutoff [21]. A phase 2 trial for NSCLC (NCT06465329) is still recruiting (Table 2). Key limitations are confined applicability to EGFR-positive malignancies, limited monotherapy efficacy and mandatory co-administration of anti-PD-1 agent.

ICOS(CD278)

Inducible T-cell co-stimulator (ICOS, CD278), a member of the CD28 superfamily, is dynamically upregulated on activated T lymphocytes predominantly within the immunosuppressive tumor microenvironment after PD-1 engagement. Physiologically, ligation of ICOS with its cognate ligand ICOSL delivers essential co-stimulatory signals to drive effector T-cell proliferation, cytokine secretion and memory T-cell formation, while ICOS overexpression on tumor-infiltrating regulatory T cells often facilitates local immune suppression. Conventional ICOS agonist monotherapies display inconsistent clinical efficacy, largely limited by off-target peripheral T-cell activation and insufficient selective enrichment in tumor sites [51]. To overcome these drawbacks, bsAb engineering has emerged as a promising strategy; accordingly, a growing portfolio of ICOS-agonistic bsAbs has been developed for tumor-selective immune activation, which will be discussed in the following section.

XmAb23104, an agonistic bsAb targeting PD-1 and ICOS (Fig. 1A), eliminates FcγR binding to avoid off-target immune depletion (Supplementary Table 1). XmAb23104 exerts dual functions via simultaneous PD-1 inhibition and ICOS co-stimulation to promote T-cell proliferation. It preferentially targets tumor microenvironment-residing PD-1/ICOS double-positive T cells for superior antitumor efficacy and improved safety compared with single-agent anti-PD-1. In a phase 1 study (NCT03752398), 62 patients with advanced solid tumors were treated in escalation at doses up to 15 mg/kg. No DLTs were observed, and MTD was not reached. TRAEs occurred in 37 subjects (59.7%), and the majority of TRAEs were Grades 1 or 2, with 6 subjects (9.7%) having a Grade 3 or higher TRAE. The most common TRAEs were diarrhea (9.7%), decreased appetite (9.7%), and fatigue (9.7%). Regarding efficacy, 3 subjects achieved PR, and 2 subjects achieved SD lasting longer than 12 months (Table 2). Based on PK, safety, and clinical activity data in consultation with investigators, the 10 mg/kg dose was selected and continued to be evaluated in the expansion part of the study [22]. Its Phase 1b/2 study that XmAb23104 combine with XmAb22841 (CTLA-4 × LAG3) (NCT05695898) proceeds in checkpoint-refractory melanoma (Table 2). Its single-agent phase 2 trial in people with sarcoma (NCT05879185) was terminated (Table 2). Limitations of XmAb23104 include modest single-agent efficacy, absent predictive biomarkers and potential aggravated immune-related toxicities from ICOS overstimulation.

KY1055 is an agonistic bsAb targeting ICOS and PD‑L1 (Fig. 1A). Structurally, it is a human IgG1 with fully human Fab arms for ICOS and a modified Fcab domain for PD‑L1 binding (Supplementary Table 1). Mechanistically, it bridges ICOS‑ and PD‑L1‑expressing cells, activates ICOS signaling, blocks PD‑1/PD‑L1 inhibition, and mediates ADCC to deplete ICOS‑high cells. Preclinically, KY1055 shows potent anti‑tumor efficacy in syngeneic models, depletes Tregs, and elevates the TEff/Treg ratio [52].

Agonistic bispecific antibodies targeting death receptors

Death receptors, which are members of the death domain-containing TNFRSF, mediate apoptotic cell death through the initiation of intracellular signaling pathways upon ligand binding. These receptors contain a cytoplasmic death domain (DD) that is essential for the initiation of apoptosis. Key death receptors include Fas (CD95), TRAIL receptors (DR4 and DR5), and TNFR1, each of which activates apoptosis through the extrinsic pathway upon binding to their respective ligands, such as FasL, TRAIL, or TNF. Ligand binding induces receptor oligomerization, resulting in the recruitment of adaptor proteins such as FADD and TRADD, activation of caspase-8, and initiation of a downstream proteolytic cascade leading to apoptosis. Beyond apoptosis, death receptors are also involved in the regulation of immune responses, inflammation, and tumor suppression through modulation of immune cell survival and activation [53]. In cancer therapy, agonistic antibodies targeting death receptors are being developed to induce apoptosis in tumor cells or to enhance immune cell–mediated cytotoxicity. Although these therapies hold considerable promise, achieving selective receptor activation remains challenging, as excessive signaling may result in systemic toxicity. Next, several DR5 (TRAILR2) agonistic bsAbs that may address these associated challenges will be discussed.

The MCSP×DR5 is designed to selectively target melanoma cells through melanoma cell surface proteoglycan (MCSP) binding while concomitantly activating DR5-mediated apoptotic signaling (Fig. 1B and Supplementary Table 1). Preclinical studies have demonstrated that this bispecific construct exerts potent and selective DR5-dependent cytotoxicity against MCSP-expressing melanoma cells; moreover, Fcγ receptor–mediated crosslinking further enhances its cytotoxic activity without compromising antigen selectivity. This therapeutic approach effectively integrates three cooperative mechanisms, including targeted delivery of DR5 agonism, robust activation of apoptotic signaling in tumor cells, and recruitment of Fcγ receptor–bearing immune effector cells to enhance antitumor immunity. In vitro and in vivo evaluations support its capacity to induce tumor-selective cell death with reduced systemic toxicity, indicating potential utility in overcoming therapeutic resistance in advanced melanoma [54, 55]. Although clinical studies remain limited, this bsAb represents a promising next-generation immunotherapeutic strategy.

BI 905,711 is a tetravalent bsAb that cross-links DR5 with the tumor-associated membrane protein cadherin-17 (CDH17) (Fig. 1B and Supplementary Table 1), thereby inducing selective apoptosis in CDH17-expressing tumor cells. In preclinical models, BI 905,711 promoted CDH17-dependent DR5 oligomerization, induced tumor-selective cell death and tumor regression in CDH17-positive colorectal cancer and other gastrointestinal tumor models, and exhibited a favorable liver-sparing safety profile compared with untargeted DR5 agonists. Mechanistically, the tumor selectivity of BI 905,711 is attributed to antigen-driven receptor clustering that enhances DR5-mediated apoptotic signaling exclusively on CDH17-positive cells; additionally, FcγR engagement further augments cytotoxicity without compromising selectivity [56]. On the basis of these findings, two phase 1a/1b studies tested BI 905,711 monotherapy (NCT04137289) or combination therapy (NCT05087992) in advanced, refractory gastrointestinal (GI) cancers. In NCT04137289, 110 patients (80% with colorectal cancer) received monotherapy. No DLTs occurred, MTD and RP2D were undetermined. The most common TRAE was nausea (16.4%) and affected 48.2% of patients. And 22.1% of 104 evaluable patients achieved SD. In NCT05087992, 12 colorectal cancer patients received combination therapy. Two DLTs were reported, while MTD was not reached. The RDE was 0.6 mg/kg plus biweekly FOLFIRI and bevacizumab. The most common TRAEs were decreased appetite (33.3%), increased ALT (25%), increased AST (25%), and diarrhea (25%). TRAEs occurred in 91.7% of patients and 91.7% attained SD (Table 2). Linear dose exposure and ≥ 2-fold activation of plasma caspase 3/7 were revealed by PK/PD analyses [23] (Table 2). Taken together, BI 905,711 exemplifies a targeted death receptor–agonist strategy that integrates tumor antigen recognition with DR5 activation to maximize tumor cell killing while minimizing off-tumor toxicity.

TR2/CDH3 BAB is a bsAb designed to simultaneously engage cadherin-3 (CDH3) and DR5 (TRAILR2) on the cell surface (Fig. 1B and Supplementary Table 1), thereby promoting higher-order TRAILR2 clustering and potent apoptotic signaling selectively in CDH3-positive tumor cells. Its antitumor activity was confirmed to depend on concurrent recognition of both targets, as CRISPR/Cas9-mediated knockout of either TRAILR2 or CDH3 completely abrogated therapeutic efficacy. High-throughput screening using the PRISM cancer cell panel revealed that multiple tumor types—including pancreatic, gastric, colorectal, and triple-negative breast cancers—exhibited marked sensitivity to TR2/CDH3 BAB, consistent with robust caspase activation observed both in vitro and in xenograft tissues. In pancreatic ductal adenocarcinoma (PDAC), where existing cytotoxic regimens provide limited clinical benefit, tumor cells frequently exhibit uniform surface expression of both CDH3 and TRAILR2, suggesting that a substantial proportion of patients may be candidates for TR2/CDH3 BAB–directed therapy. This bsAb demonstrated potent antitumor efficacy in multiple PDAC patient-derived xenograft models, achieving tumor regression in several cases. Moreover, when administered in combination with chemotherapeutic agents, TR2/CDH3 BAB produced more profound and durable tumor control than single-agent treatment. Taken together, these findings, combined with its promising preclinical activity and potential for a favorable safety profile, justify further clinical evaluation of TR2/CDH3 BAB in PDAC [57].

MCSP×DR5, BI 905,711, and TR2/CDH3 BAB represent a new generation of death receptor–engaging bsAbs that integrate tumor antigen targeting with death receptor agonism to induce selective apoptosis in malignant cells. All three exploit tumor-restricted surface antigens (MCSP, CDH17, and CDH3) to spatially confine TRAILR2/DR5 activation to tumor cells, thereby enhancing therapeutic specificity and safety. Key differences lie in the targeted antigen pairings, molecular architectures, and cancer indications under investigation: MCSP×DR5 is intended for melanoma, BI 905,711 for CDH17-positive colorectal and gastrointestinal cancers, and TR2/CDH3 BAB for CDH3-positive pancreatic ductal adenocarcinoma and other solid tumors. Although preclinical studies demonstrate tumor regression and enhanced efficacy in combination with chemotherapy for all three agents, their distinct antigen targets and expression patterns define different patient populations and may entail divergent safety or dosing considerations. Collectively, these agents exemplify a shared mechanistic paradigm of tumor antigen–directed death receptor activation; however, variations in antigen selectivity, receptor biology, and clinical development trajectories are expected to influence their translational potential and therapeutic niches.

Agonistic bispecific antibodies targeting growth factor receptors

Growth factor receptors constitute a distinct class of signaling modules that couple ligand-dependent receptor tyrosine kinase activation with systemic metabolic regulation. Prominent among these are the fibroblast growth factor receptors (FGFRs), which operate in concert with tissue-restricted co-receptors such as Klothoβ (KLB) [58]. Binding of FGFs (notably FGF19 and FGF21) to the FGFR–KLB complex confers ligand and tissue specificity, as KLB expression restricts responsiveness principally to adipose tissue, the liver, and selected central nervous system sites, thereby enabling FGFs to elicit context-dependent effects on energy homeostasis, glucose and lipid metabolism, and bile acid regulation [59, 60]. Mechanistically, FGFR activation triggers canonical receptor tyrosine kinase cascades (RAS–MAPK, PI3K–AKT, and PLCγ pathways), resulting in transcriptional reprogramming of metabolic gene networks, increased energy expenditure, and modulation of insulin sensitivity. From a translational perspective, the modular nature of FGFR–KLB signaling has motivated therapeutic strategies aimed at recapitulating or modulating FGF activity with high tissue selectivity. Because receptor–coreceptor pairing determines both the pharmacodynamic profile and adverse-event risk, interventions targeting the FGFR–KLB axis must balance potency with spatial selectivity to maximize favorable metabolic outcomes while limiting off-target mitogenic or hepatobiliary effects [60]. Building upon these mechanistic insights, recent studies have explored the potential of bispecific agonist antibodies to mimic or fine-tune receptor–coreceptor activation in a controlled manner.

BFKB8488A represents a prototype humanized bsAb that concurrently binds FGFR1 and KLB (Fig. 1C and Supplementary Table 1), thereby mimicking the biological activity of FGF21. By enforcing receptor–coreceptor dimerization, this antibody activates canonical MAPK and PI3K–AKT signaling cascades, resulting in increased adiponectin secretion, enhanced thermogenic gene expression, and improved lipid and glucose metabolism in preclinical models [58]. Clinically, BFKB8488A was tested in two early-phase trials. The phase 1a trial (NCT02593331) in overweight and obese participants with likely insulin resistance showed good safety/tolerability with single dosing, elevated adiponectin and improved insulin sensitivity markers. The phase 1b trial (NCT03060538) in participants with Type 2 diabetes mellitus (T2DM) and participants with non-alcoholic fatty liver disease (NAFLD) confirmed safety of repeated administration, significantly improved blood lipids, liver enzymes and hepatic fat content, but had no expected glucose-lowering effect and no serious adverse events [61]. The phase 2 trial (NCT04171765) will evaluate the efficacy, safety, and pharmacokinetics of BFKB8488A compared to placebo in participants with non-alcoholic steatohepatitis (NASH). Mechanistically, bispecific engagement of FGFR1 and KLB produces a spatially restricted receptor activation pattern that overcomes the pharmacokinetic limitations of native FGF21, including rapid clearance and off-target exposure. This design illustrates the potential of antibody engineering to harness receptor–coreceptor agonism and produce durable and selective metabolic effects [58]. Collectively, these findings establish FGFR1/KLB-targeted bispecific agonists as a promising therapeutic modality for metabolic diseases and underscore a conceptual bridge between endocrine growth factor biology and antibody-based agonist design, thereby setting the stage for comparison with immune co-stimulatory bispecific agonists.

Challenges and future perspectives

Agonistic bsAbs have emerged as a versatile class of therapeutics that deliver co-stimulatory signals selectively within the tumor microenvironment. By simultaneously engaging a tumor-associated antigen and an immune receptor, agonistic bsAbs aim to enhance T-cell activation while minimizing systemic immune toxicity. Despite promising preclinical and early clinical data, several interrelated challenges continue to limit their broad translation.

Safety and clinical translation

Safety remains the foremost challenge for agonistic bsAbs. Systemic immune overactivation can result in hepatotoxicity, as observed with the potent CD137 agonist urelumab, which caused dose-limiting hepatotoxicity in clinical trials due to activation of liver-resident Kupffer cells [32]. To address the shortcoming, an agonistic bsAb, MCLA-145, was identified through an unbiased functional screening approach. It incorporates conditional activation mechanisms that confine receptor agonism to the tumor microenvironment and has not elicited significant liver enzyme elevations in preclinical studies (Fig. 3A). This dual-dependent activation mode avoids robust CD137 signaling in PD-L1-low normal organs including the liver [37]. Nevertheless, as clinical trials mature, careful monitoring for hepatotoxicity and systemic immune effects remains essential.

Fig. 3.

Challenges and future perspectives for agonistic bsAbs. (A) Safety: Engineering strategies, exemplified by MCLA-145, aim to mitigate hepatotoxicity observed with first-generation CD137 agonists, such as urelumab, by conditional activation mechanism. (B) Clinical translation: Optimizing dosing regimens and interpatient variability remains a critical challenge to achieve clinical translation of agonistic bsAbs. (C) Molecular design: Rational engineering approaches, such as binding affinity, epitope location, valency, and Fc engineering, are essential to enhance the potency, stability, and functional performance of agonistic bsAbs. In addition, immunocytokines, a novel molecular design, that couple checkpoint antibody scaffolds with receptor-selective cytokines constitute a highly promising therapeutic modality to decouple antitumor efficacy from systemic cytokine toxicity. ANV600, a novel PD-1 targeted, IL-2Rβγ-biased IL-2 molecule, serves as a representative example of this strategy. (D) Functional selectivity: Agonistic bsAbs enable receptor agonism to occur predominantly within the tumor microenvironment rather than systemically. For instance, CD28 agonist mAb TGN1412 causes severe systemic toxicity, while tumor-targeted CD28 bsAb RO7443904 restricts agonism to the tumor microenvironment. (E) Novel target discovery: It is a highly promising strategy to discover novel targets for agonistic bsAbs, aiming to expand the spectrum of treatable malignancies and enhance the precision of tumor targeting

Optimizing clinical regimens further complicates translation. Strategies such as step-up or split dosing may mitigate cytokine release, but the optimal dosing regimen remains undefined. Interpatient variability further complicates the establishment of uniform dosing guidelines [62] (Fig. 3B). Additionally, the heterogeneous tumor microenvironment strongly influences responses to co-stimulatory agonists, and combination therapies with checkpoint inhibitors, T-cell engagers, or adoptive cell transfers may improve efficacy while simultaneously raising the risk of toxicity [63]. Collectively, these issues represent key challenges to the clinical safety of agonistic bsAbs and to their successful translation from preclinical studies into clinical application.

Molecular design and functional selectivity

Molecular design and functional selectivity are critical to achieving an appropriate balance between efficacy and safety. Potency and tissue distribution of co-stimulatory signals are highly sensitive to parameters such as binding affinity, epitope location, valency, and Fc engineering (Fig. 3C). Excessive receptor clustering may provoke uncontrolled signaling, whereas insufficient clustering reduces biological activity [64]. Therefore, precise molecular design that is based on these parameters above is indispensable to generate optimal receptor crosslinking and controlled co-stimulatory signaling. In addition, beyond conventional agonistic bsAbs, immunocytokines that couple checkpoint antibody scaffolds with receptor-selective cytokines constitute a highly promising therapeutic modality to decouple antitumor efficacy from systemic cytokine toxicity. ANV600 serves as a representative example of this strategy: its rationally engineered structure combines a non-blocking PD-1-binding Fab and an IL-2Rα-masked IL-2 agonist (Fig. 3C), enabling localized IL-2Rβγ co-stimulation solely on intratumoral PD-1⁺ CD8⁺ T cells. This refined molecular tuning minimizes off-tumor signal activation and Treg proliferation, matching the core design principles for balanced potency described above [65]. Taken together, refined molecular engineering serves as a significant strategy to balance antitumor potency and systemic safety for agonistic bsAbs.

Functional selectivity—defined as receptor activation confined to the intended cellular or spatial context—has emerged as a key design objective. Approaches such as affinity tuning, conditional activation, and tumor antigen–restricted co-engagement enhance selectivity, ensuring that agonism occurs predominantly within the tumor microenvironment rather than systemically [66]. This strategy was largely motivated by the catastrophic clinical failure of unrestricted CD28 agonist mAb TGN1412, which triggered life-threatening systemic cytokine storm via unconstrained peripheral T cell crosslinking independent of tumor antigen or TCR priming. Consequently, tumor-targeted CD28 bsAbs represented by RO7443904 were developed and they exemplified this principle by delivering localized co-stimulatory signals to tumor-infiltrating T cells while minimizing systemic toxicity. For instance, RO7443904 adopts low-affinity monovalent CD28 arm paired with silenced Fc domain, rendering the molecule functionally inert in circulating peripheral blood and only capable of stable CD28 crosslinking upon simultaneous binding to CD19-expressing malignant cells and TCR signal 1 at tumor immune synapses [50] (Fig. 3D). Such constructs illustrate the broader principle that functional selectivity, rather than mere receptor occupancy, is paramount in the design of agonistic bsAbs.

Novel target discovery

Currently, several translationally promising strategies for agonistic bsAbs have advanced through preclinical and early-phase clinical research. Among currently investigated approaches, TAA-restricted conditional TNFR co-stimulation bsAbs and dual immunomodulatory bsAbs have demonstrated encouraging preclinical and early clinical activity [67]. Among currently investigated target combinations, PD-L1×CD137 is one of the most clinically advanced and extensively studied formats. This pairing delivers synergistic anti-tumor immunity via concurrent checkpoint inhibition and conditional co-stimulation, while restricting CD137 agonism strictly to PD-L1-rich tumor lesions to mitigate the systemic hepatotoxicity seen with untargeted CD137 monoclonal agonists [68]. Multiple candidates including MCLA-145 have progressed to phase 1 clinical trials, yielding durable tumor regressions and favorable safety profiles across a broad spectrum of solid tumors. However, the clinical utility of PD-L1×CD137 bsAbs is constrained by variable PD-L1 positivity across cancer types and limited efficacy in checkpoint-refractory lesions [67]. Consequently, it drives urgent demands to discover novel targets for agonistic bsAbs, aiming to expand the spectrum of treatable malignancies and enhance the precision of tumor targeting (Fig. 3E).

Outlook

Future advances in agonistic bsAb development will rely on the integration of molecular engineering, systems immunology, and translational medicine. Innovations in structure-guided design, Fc engineering, and switchable antibody platforms enable fine-tuned control of agonist potency and spatial restriction. Concurrently, improved humanized preclinical models and biomarker-driven clinical monitoring are expected to enhance safety prediction and dosing precision. The emerging paradigm emphasizes contextual co-stimulation—transient, tumor-restricted, and tunable—rather than indiscriminate immune activation. Achieving this balance promises to unlock the full therapeutic potential of agonistic bsAbs across cancer immunotherapy and potentially beyond.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Jinyan Yang: Conceptualization, Writing – original draft, Writing – review & editing. Junwen Deng: Conceptualization, Writing – original draft, Writing – review & editing. Zhaowei Sun: Conceptualization, Writing – original draft, Writing – review & editing. Yumeng Liu: Writing – review & editing. Yuhui Wang: Funding acquisition, Writing – review & editing. Xiangzheng Meng: Funding acquisition, Writing – review & editing. Xinlin Liu: Resources, Funding acquisition, Writing review & editing. Xiang Ma: Resources, Writing review & editing. Peng Sun: Conceptualization, Project administration, Funding acquisition, Writing – review & editing.

Data availability

Not applicable. All data cited in this review are available from the original published articles referenced in the manuscript.

Declarations

Ethics approval and consent to participate

Not applicable. This is a review article and no original studies with human participants or animals were performed by the authors.

Consent for publication

Not applicable. This is a review article and no individual patient data or identifiable personal information is included.

Competing interests

The authors declare no competing interests.