Introduction

Immunotherapies attempt to harness the innate and adaptive immune system to attack cancer cells.1 Since early systematic clinical applications of immunotherapy in oncology, such as the use of Coley’s bacterial toxin for sarcoma more than 100 years ago and Bacillus Calmette-Guérin vaccine for bladder cancer in the 1970s,2 there has been an exponential evolution accelerated by the epochal FDA approvals of the first immune checkpoint inhibitors (ICIs), the antibody ipilimumab against anti-cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) in 2011 and the first antibodies against anti-programmed cell death protein 1 (PD-1) pembrolizumab and nivolumab in 20143 (Fig. 1).

Fig. 1
figure 1

Timeline showing representative events of drug approval, clinical trials, and key biological discovery for immunoregulatory targets of anti-cancer therapeutics. Events involving immunoregulatory receptors and bispecific antibodies are in blue and dark gray boxes and those related to immuno-epigenetics and cytokines are in light gray and pink boxes, respectively. Boxes with solid lines indicate approvals, while other events are in dashed line boxes. Lines with rounded corners indicate the period of clinical trials. The approvals include the final approval, the accelerated approval by FDA, and the approval by EMA. Cancer stage descriptions such as advanced or metastatic, pathological subtypes, and details of combination therapies are omitted. ICI immune checkpoint inhibitors, mAb monoclonal antibody, BiTE bispecific T cell engager, ALL acute lymphoid leukemia, NSCLC non-small cell lung cancer, MSI-H microsatellite instability-high, dMMR deficient mismatch repair, TNBC triple-negative breast cancer, SCLC small cell lung cancer, bsAb bispecific antibody, HCC hepatocellular carcinoma, TCR T-cell receptor, TCE T cell engager, rhIL-2 recombinant human interleukin-2

Full size image

Despite the remarkable success achieved by ICIs and ICI-based treatment combinations in some tumor entities,4,5,6,7,8,9,10,290,291 (Fig. 2). Therefore, the efficiency of agonists is affected by unique factors. First of all, agonists with very high affinity or at excessive dose can lose their agonistic function,292 suggesting a bell-shaped affinity-agonism and dose-response relationship and an optimal affinity and dose. Secondly, agonistic antibodies can bind both natural ligand binding sites and exclusive epitopes.290,293,294 For example, different domains of CD40 are associated with agonistic or antagonistic effects of anti-CD40 antibodies.295 Characterizing the antibody binding epitope is therefore very important for agonist development.

Moreover, the interaction between the antibody Fc domain and FcγRs can induce both agonist and ADCC/CDC effects. Except for the inhibitory FcγRIIB, other FcγRs are activating and FcγRI has the highest affinity for the Fc region. Binding FcγRIIB is proposed to promote target receptor crosslinking and to maintain immune synapses, thus providing true agonism290,291 (Fig. 2). Instead, binding activating FcγRs can elicit ADCC, which can be utilized to deplete Tregs, especially using the IgG1 isotype with the strongest binding to activating FcγRs290,291 (Fig. 2). Therefore, agonists can either activate anti-cancer immune cells or deplete immunosuppressive populations. However, issues might arise from indiscriminate ADCC triggered by activating FcγRs, depleting Tregs but also effector cells. Binding activating FcγRs also contributes to toxic side effects, e.g., in case of 4-1BB agonists.296,297 For these reasons, Fc engineering is crucial and has been shown to be highly useful for the development of pure agonists by removing the Fc segment,297,298 mutation methods abating Fc-FcγR interactions299 or selectively enhancing Fc-FcγRIIB binding.300

In particular, human IgG2 agonists can activate co-stimulatory molecules including CD40, 4-1BB, and CD28 independent of FcγRs.290,301 Later studies showed that agonists with rigid conformation constrained by “tight” hinge region promote clustering of co-stimulatory molecules301,302,303 and tend not to bind excess epitopes mediating antagonism as is the case for more flexible antibodies,301,304 thus providing sufficient agonism even without FcγRs,295 and this phenomenon exists on natural IgG2 isotype mAb.301,302

IgSF co-stimulatory receptor: ICOS

ICOS (CD278) is the receptor of ICOSL (B7-H2, CD275, B7h). Upon initial activation of TCR and CD28 signaling, ICOS is upregulated on T cells and this can non-redundantly enhance T cell immunity288,290,291 while ICOS is constitutively expressed on Tregs.291 ICOSL is constitutively expressed on APCs.288 After activation, ICOS induces phosphoinositide 3-kinase (PI3K)-Akt signaling,305 mammalian target of rapamycin (mTOR),306 and nuclear factor of activated T cells (NFAT)-responsive genes290 in T cells.

Anti-ICOS agonistic antibodies currently under development include vopratelimab and alomfilimab (Table 1 and Supplementary Table 3). The widely reported IgG4 pure agonist feladilimab has been removed from the GlaxoSmithKline pipeline due to its unsatisfactory clinical activity in phase II studies. The IgG1 mAbs vopratelimab and alomfilimab are designed to deplete intratumoral Tregs. Although vopratelimab plus nivolumab only elicited a total ORR of 2.3%, patients with ICOShigh CD4+ effector T cells had longer PFS and OS than patients without these cells (6.2 vs. 1.9 and 20.7 vs. 9.0, months).306 This finding guided the patient selection for the phase II SELECT study in NSCLC, where the combination of vopratelimab at 0.03 mg/kg with pimivalimab (a PD-1 inhibitor) yielded an ORR of 40% and a 6-month PFS rate of 80%. However, the study did not reach the primary endpoint of tumor shrinkage between vopratelimab plus pimivalimab and pimivalimab monotherapy groups.307 Alomfilimab depleted ICOShigh Tregs, had monotherapy anti-tumor efficacy, and improved anti-PD-L1 efficacy in a pre-clinical study.308 According to a preliminary report there were 5 OR cases out of 103 patients in a phase I/II trial testing alomfilimab ± atezolizumab.309 In summary, ICOS drug development is still challenging.

TNFRSF co-stimulatory receptor: CD40, OX40, and 4-1BB

Upon ligand trimer ligation, TNFRs on T cells trimerize to recruit TNFR-associated factor (TRAF)1-6 in different preferences and activate distinct downstream adapters but predominantly converge at nuclear factor-κB (NF-κB) signaling.289,290 According to the chronological impact on T cell activation as discussed above, we first discuss CD40, then focus on OX40 and 4-1BB that aroused most incentives of industries. Unlike OX40 and 4-1BB, the development of GITR agonists has been largely terminated due to limited responses.310,311,312,313,314,315 Similarly, agents targeting the CD27-CD70 pathway, such as the widely reported CD27 agonist varlilumab and CD70 agonist cusatuzumab, have also been removed from the pipelines of Celldex and Argenx, respectively, due to unfavorable developmental prospects. Likewise, the development of TNFR1/2 agonists remains immature, with almost all agents still under preclinical investigation.316,317 Therefore, other TNFRSF receptors, including GITR, CD27/CD70, and TNFR1/2, are not the focus of our discussion.

CD40

CD40 (TNFRSF5) expressed mainly by APCs plays an important role in initial activation of CD4+ T cells following the CD28 signal. CD40L (CD154) mainly expressed by CD4+ T cells ligates and activates CD40, triggering the maturation of DCs which is crucial for the efficient priming of T cells including CD4+ Th cells and cross-primed CD8+ T cells288,318 (Fig. 2). Activated CD40 stimulates expression of CD80 and CD86 on DCs thus stimulating the CD28 coreceptor on T cells which in turn leads to upregulation of CD40L on T cells coordinately driving T cell stimulation and DC maturation (Fig. 2).

The agents presently developed all entered phase II clinical trials (Table 1 and Supplementary Table 3), while only the development of selicrelumab has been discontinued. SEA-CD40, mitazalimab, sotigalimab, and giloralimab are IgG1 FcγR-dependent DC activators, whereas YH003 and CDX-1140 are IgG2 pure agonists. In the phase Ib/II PRINCE study of sotigalimab plus chemotherapy ± nivolumab in pancreatic adenocarcinoma, the total ORR was 58% in the phase Ib part,319 while in phase II part, the confirmed ORR of sotigalimab plus chemotherapy was 33%.319 Mitazalimab efficiently upregulated CD80/CD86 expression and IL-12 secretion by DCs, induced antigen-specific T cell proliferation and anti-tumor activity preclinically.320,321 Efficacy evaluation is ongoing in the phase II OPTIMIZE-1 study combining mitazalimab and chemotherapy in pancreatic ductal adenocarcinoma (PDAC) patients. More studies will be needed on combinations with other agents or regarding optimizing indication selection.

OX40

OX40 (CD134, TNFRSF4) is temporarily expressed by memory T cells and activated T cells following TCR/CD3 signaling and has important roles in their survival, yet it does not participate in T cell priming.288,291 It is also constitutively expressed by Tregs.288,291 Interestingly, OX40 agonism does not impair the immunosuppressive functions of Tregs but only confers them an inflammatory phenotype.322 Expression of OX40L (CD252) is upregulated on APCs after their activation and can be promoted by activated CD40.288 After binding of OX40L,323 trimerized OX40 recruits TRAF2-3 and TRAF5 to transmit canonical and non-canonical NF-κB and other signals288,289 (Fig. 2).

Several major companies have withdrawn from the development of OX40 agonists due to unfavorable clinical efficacies,324,325,326 indicating the necessity of strategy improvement for further development. OX40 agonists under development currently include revdofilimab, HFB301001, and BGB-A445 (Table 1 and Supplementary Table 3). The IgG1 agonist INCAGN1949 is proven to FcγR-dependently stimulate OX40 and deplete OX40high Tregs.327 However, in a phase I/II study, INCAGN1949 monotherapy only elicited an ORR of 1.15%,328 hence it has been removed from the pipeline of Agenus. Trials of other agonists are all still ongoing. The development of many OX40 agonists has been discontinued. Due to the transient expression of OX40, the timing of OX40 agonist administration may be important.291 Further development of OX40 agonists may need either combining with other agents in an appropriate order or develo** msAbs.

4-1BB

4-1BB (CD137, TNFRSF9) is also transiently upregulated following TCR/CD3-mediated signaling mainly on activated T cells289 but is also detected on NK cells and APCs.291 Upon ligation of 4-1BBL (TNFSF9), 4-1BB recruits TRAF1-2 to activate downstream signaling similar to OX40289 (Fig. 2). Considering the substantial liver toxicity at doses of ≥1 mg/kg290,293,294 and modest ORR of 3.8%329 observed in trials of the first generation 4-1BB agonistic antibodies urelumab and utomilumab respectively, Bristol-Myers Squibb and Pfizer deprioritized the development of these two drugs. However, subsequent analyses have guided further design of 4-1BB agonists. As many reports indicated, utomilumab showed insufficient clinical monotherapy activity while urelumab induces strong agonism but also severe toxicity in a fraction of the patients.291 Structural analysis indicated that utomilumab blocks natural ligands and binds 4-1BB at proximal domains while urelumab binds the distal one,330,331 which is consistent with antibodies against CD40.295 This reflects the importance of the binding epitopes in the design of agonists. The toxicity of urelumab mostly stems from Fc-FcγR interaction, thus Fc engineering is relevant for toxicity management of 4-1BB agonists. Based on such considerations, next-generation 4-1BB agonists including ADG106,332 LVGN6051,333 AGEN2373,334 and ATOR1017 have been developed and are being investigated in clinical trials (Table 1 and Supplementary Table 3). In a phase I trial of ADG106, treatment appeared to be safe with a DCR of 57%.332 LVGN6051 monotherapy elicited a DCR of 70% and induced preliminary ORR of 25% combined with pembrolizumab in a phase I study.335 AGEN2373 induced a DCR of 26.3% without liver toxicity.336 Dose escalation for ATOR-1017 is still ongoing with the best response of SD observed.337

In summary, agonists targeting costimulatory receptors appear powerful candidates for future immunotherapy and a wave of new agonistic molecules has been developed many of which have entered clinical trials. However, agonist development is more difficult than the development of antagonists because more parameters have to be taken into account. Clinical trials have shown that agonist monotherapies scarcely induce favorable responses hence combination with ICIs or other agents may become particularly important. Next-generation constructs including Fc-engineered mAbs, multi-valent mAbs, and bsAbs/msAbs seem promising.

Immunoregulatory bispecific and multi-specific antibodies

The concept of bsAb targeting two different molecules was proposed in the last century.338 At that time, shortly after gaining insights into immunoglobulin biology, Alfred Nisonoff envisioned combining two distinct antigen-binding sites within a single molecule. He connected rabbit Fab fragments with different specificities using chemical methods and demonstrated bispecificity of the resulting product.339 Subsequently, other researchers advanced the field of bsAbs by introducing hybridoma methods for mAbs, phage display techniques, and strategies to direct antibody effects towards various target cells.340,341,342 However, bsAbs/msAbs with promising efficacy and acceptable safety had not been developed until the last decade, when the CD3×CD19 bispecific T-cell engager (BiTE) blinatumomab was approved by the FDA.342 Along with the advances in antibody format design, and further comprehension of cancer immunology, anti-cancer bsAbs/msAbs targeting immunoregulatory and other cancer-related molecules are under intensive development. Here we present an update of the developmental landscape of these agents (Fig. 3a–d, Table 2 and Supplementary Table 4) compared with previous summaries342,343,344,345,346 according to data from the pipeline and clinical trials. We briefly introduce the characteristics of anti-cancer immunoregulatory bsAbs/msAbs, mainly discussing their categories according to mechanism-of-action, and clinical vista of widely reported agents.

Fig. 3
figure 3

Various formats and categories of bsAbs. a Examples of common bsAb formats. Short bars indicate the antibody Fab segment, long bars indicate the Fc segment, and for ImmTAC, the Fab segment is linked to an antigen-specific TCR. In the same antibody icon, different color combinations of the Fab segment indicate the binding to different target proteins. BsAb examples of these formats are in the gray box. b According to our statistics of bsAbs that have entered clinical trials, the TCEs comprising the anti-CD3 scFv account for about half of all bsAbs under development at present. The development of other bsAb categories presents a diversified landscape. The two ends of each arc indicate two targets of bsAbs. Only bsAbs are counted for this figure, with msAbs with higher valency excluded. The data of this figure are consistent with Table 2 and Supplementary Table 4. The statistic is up to October 2022 and bsAbs with terminated development are excluded. c Prism of developmental strategies of bsAbs. The strategy of bsAb development is mainly to combine four types of targets: immune cell targets, tumor cell targets, co-stimulatory molecules/immunostimulatory cytokines, and immune checkpoints or other immunosuppressive molecules. By these designs, immune cells and immunomodulatory signals can be introduced into the TAA-expressing environment. The black lines on the edge of the prism indicate that the corresponding bsAb category targets the target types directed by the arrows: (1) TCEs and NKEs; group I general bsAbs: (2) co-stimulatory molecule × TAA or TME protein and (3) co-stimulatory molecule × co-stimulatory molecule; group II general bsAbs: (4) inhibitory checkpoint × TAA and (5) inhibitory checkpoint × inhibitory checkpoint; group III general bsAbs: (6) inhibitory checkpoint × co-stimulatory molecule. d Mechanism-of-action of bispecific cell engagers and group I-III general immunoregulatory anti-cancer bsAbs. DART dual-affinity retargeting, scFv single-chain variable fragment, TAA tumor-associated antigen, HLA human leukocyte antigen, HSA human serum albumin, TriTAC Tri-specific T cell activating construct, CAF cancer-associated fibroblast, FAP fibroblast activation protein

Full size image
Table 2 Immunoregulatory anti-cancer bispecific and multi-specific antibodies
Full size table

BsAbs/msAbs have both similar and distinct mechanisms of action compared with mAbs. Fc-FcγR interactions are thought to be mainly responsible for the toxicity of early bsAbs, as in the case of the bispecific trifunctional antibody catumaxomab (anti-EpCAM×anti-×anti-CD3).342 Thus, now T cell engagers (TCEs) are mainly constructed without Fc segment or with a functionally silenced Fc segment. Complete removal of the Fc segment as in the cases of BiTEs and DARTs has not been the main trend and the development of many BiTEs and DARTs has been discontinued due to insufficient efficacy and safety issues. With the Fc segment silenced by mutation, the leading format of TCE development at present is the 1 + 1 asymmetric IgG-like form (Fig. 3a). The affinity toward different targets of a single bsAb can be fine-tuned by adjusting the two single-chain variable fragment (scFv) arms independently, thus ameliorating safety or pharmacokinetic/pharmacodynamic (PK/PD) properties. Moreover, in terms of PK/PD characteristics, the optimal dose for bsAbs is one that results in maximum target-bsAb-target trimer formation.347

Some bsAbs/msAbs can elicit biological effects that cannot be induced by the corresponding mAb mixture, therefore they are called obligate bsAbs/msAbs.342 For immunoregulatory anti-cancer bsAbs/msAbs, this has been demonstrated by redirecting CD3+ T cells, or immune cells expressing checkpoint receptors or co-stimulatory molecules to TAA-expressing cells or the TME. Moreover, bsAbs/msAbs binding different immunoregulatory targets can, at the same cellular spatial location, target multiple immunoreceptors or simultaneously enhance the co-stimulatory signal and inhibit immune checkpoints, hence potentially causing stronger anti-cancer immunity compared with the mAb mixture. These bsAbs/msAbs can be divided into cell engagers involving CD3, CD16a, or TAA-specific TCRs and general immunoregulatory anti-cancer bsAbs/msAbs combining all other immunoregulatory molecules or TAAs (Fig. 3c, d and Table 2).

In summary, bsAbs/msAbs have several potential advantages, including (1) superior specificity, safety, and therapeutic efficacy compared with the corresponding mixture of mAbs, (2) the ability to bridge and recruit immune cells, and (3) dual or multiple signal regulation. Nevertheless, disadvantages of bsAbs/msAbs still exist including chain mispairing in production, risk of inducing cytokine release syndrome (CRS), and the potential for inducing anti-drug antibodies (ADAs). In bsAb/msAb production, diverse combinations of light and heavy chains could lead to the dilution of the target bsAb, posing challenges in its isolation and resulting in low yield.348,349 Innovative development platforms, such as CrossMab350,351 and orthogonal Fab interface,352 have emerged to mitigate the impact of this issue. CRS is a common and distinctive adverse effect in the clinical application of bsAbs,353,354,355 mainly associated with TCEs containing the anti-CD3 arm. It is a systemic inflammatory response with symptoms ranging from fever, fatigue, and headache to multiorgan failure, triggered by T cell activation, with myeloid cells and TNF-α being the main mediators of the systemic cytokine release.356,357 To advance the further application of TCEs, the management of their using and the handling of adverse events should be improved, for example, with stepwise dosing, properly using tocilizumab, corticosteroids, or TNF-α blockade, and supportive353,354,355,358 care. Regarding the induction of ADAs, increased engineering and artificial design may result in greater differences between bsAbs and endogenous immunoglobulins, and bsAbs could therefore potentially contain new epitopes that elevate antigenicity and subsequently increase the likelihood of ADA development. Therefore, early monitoring of immunogenicity is crucial for increasing clinical success rates in bsAb development.359,360

Bispecific T cell engagers

TCEs are representative obligate bsAbs combining anti-CD3 and anti-TAA scFvs to redirect any T cell to TAA-expressing tumor cells. TCEs make up nearly half of the immunoregulatory anti-cancer bsAbs/msAbs currently in clinical trials (Fig. 3b). Of note, the formats of TCEs comprise BiTE, dual-affinity re-targeting (DART), IgG-like full-length format, and others342,343 (Fig. 3a). Another type of TCE utilizing a TAA-specific TCR instead of an anti-TAA scFv is called ImmTAC. The development of TCEs surged after the approval of blinatumomab, which, as explained above, is an Fc-free BiTE. Blinatumomab yielded a CR rate of 43% in a phase II trial in Ph- relapsed or refractory (r/r) B-precursor acute lymphoblastic leukemia (ALL) patients361; it was thus approved by FDA in 2014. After blinatumomab, the CD3×CD20 IgG-like TCE mosunetuzumab was conditionally approved in the European Union,362 and also received accelerated approval by FDA in 2022 because it induced a CR rate of 60% for r/r follicular lymphoma (FL) in phase I and II trials.363,364 Likewise, teclistamab monotherapy was conditionally approved in the European Union365 and approved by FDA366 in 2022 for r/r multiple myeloma (MM) due to an ORR of 63.0%, a CR rate of 39.4% and mPFS of 11.3 months in the phase I/II MajesTEC-1 trial.367,368 Because of the reported mOS of 21.7 months in HLA-A*02:01+ uveal melanoma patients in a phase III trial,369 tebentafusp became the first approved ImmTAC in 2022.

The indications of TCEs depend on the TAA expression of the cancer type. For example, TCEs targeting CD20, CD19, and CD38 are all designed for hematological malignancies and are rivals of CAR-T cell therapies in hematology. The development of TCEs against solid tumors seems more challenging. Challenges include heterogeneity in TAA expression, on-target off-tumor toxicity for normal tissue, the immunosuppressive TME, disordered vasculature, and limited tumor penetration. These challenges might be overcome by further structure design exploration, antibody avidity fine-tuning, or therapy combinations.

At present, TCEs that have been approved or entered phase III clinical trials all target hematological TAAs. TCEs advanced into phase III trials before approval include epcoritamab (CD3×CD20), glofitamab (CD3×CD20), and elranatamab (CD3×B-cell maturation antigen) (Table 2 and Supplementary Table 4). For epcoritamab, the phase I/II EPCORE NHL-1 study showed an ORR of 68% and 90% for r/r B-cell non-Hodgkin lymphoma (B-NHL) and r/r FL patients with monotherapy,370 supporting the ongoing phase III EPCORE DLBCL-1 study. The majority of trials of glofitamab combine it with rituximab, obinutuzumab, or tocilizumab pretreatment to mitigate cytokine release.371 A phase I study combining glofitamab and obinutuzumab pretreatment in r/r B-NHL patients showed an ORR of 53.8% and a CR rate of 36.8%.372 For elranatamab, the phase I MagnetisMM-1 study has demonstrated an ORR of 75% at high doses,373 supporting two ongoing phase III trials.

Beyond conventional TCEs, other components are introduced in novel formats to refine immunostimulatory properties, PK/PD attributes, and toxicity (Fig. 3d and Table 2). By introducing a CD28 immunostimulatory arm, Sanofi designed Fc-silenced CD3 × CD38 × CD28 TCE with better stimulation of anti-tumoral T cells.374 Based on this design, SAR442257 has been developed and is being tested in a phase I trial (NCT04401020). Another category called Tri-specific T Cell-Activating Construct (TriTAC) introduced anti-human serum albumin scFv to improve PK/PD properties for solid tumors. Preclinical results showed superior T-cell killing compared with conventional BiTEs targeting EGFR or PSMA and favorable efficacy,375 supporting phase I/II trials (Table 2 and Supplementary Table 4). To improve safety, XTENylated protease-activated T cell engagers (XPATs) were created by introducing scFvs with TME-specific degradable masking, thus avoiding off-tumor T cell activation. Sanofi completed the acquisition of this technology in 2022, including the HER2 XPAT AMX-818. Moreover, as functions are being continuously discovered, innate immune cell populations are also evaluated for immuno-oncology agent development. BsAbs targeting CD16A/FcγRIIIa, an activating FcγR, to redirect NK cells to TAA-expressing cells are called NK cell engagers (NKEs) or innate cell engagers (ICEs) (Table 2 and Supplementary Table 4). Unfortunately, AFM13, a representative CD16A × CD30 NKE for r/r Hodgkin lymphoma, only induced an ORR below 25% in several trials as monotherapy.376,377 However, an ORR of 88% was induced by combining AFM13 and pembrolizumab,378 suggesting combination therapy for further development.

General immunoregulatory anti-cancer bsAbs/msAbs

Apart from CD3-engaging TCEs, many other anti-cancer bsAbs/msAbs target immunoregulatory proteins other than the CD3 complex. Based on the design, this category includes three subgroups: bsAbs/msAbs stimulating co-stimulatory molecules (group I), blocking immune checkpoints (group II), and the combination of these two tactics (group III) (Fig. 3d, Table 2 and Supplementary Table 4). These bsAbs/msAbs are currently mainly developed for the treatment of solid tumors.

Two designs are used for group I bsAbs/msAbs (Table 2). The first one is by binding co-stimulatory molecules on immune cells and TAA-expressing tumor cells or fibroblast activation protein on cancer-associated fibroblasts. The second one is to concurrently target distinct IgSF/TNFRSF co-stimulatory molecules on immune cells.

Group II bsAbs/msAbs include three subtypes (Table 2). The first one redirects PD-1/PD-L1 blockade toward TAAs or tyrosine kinase expression-enriched TME. The second one concurrently targets different immune checkpoint ligand-receptor axes. Due to thorough research on ICI combination therapies, the development of this subtype is the main trend for group II bsAbs/msAbs and is also most advanced in this category. The third one targets PD-1/PD-L1 and immunosuppressive molecules beyond IgSF checkpoints, such as CD47 and TGF-βRII. Group III includes designs mainly combining anti-PD-1/PD-L1 and co-stimulatory agonist arms, and fusion proteins combining ICI and immunostimulatory cytokines (immunocytokines) (Fig. 3d and Table 2). In a preclinical study, anti-PD1–IL-2v immunocytokine was proved to have superior ability to expand tumor-specific CD8+ effector-like T cells and therapeutic efficacy than the (agonistic) IL-2Rβγ-biased mutant IL-2 variant IL-2v in tandem with an anti-FAP scFv.379 These findings support the clinical development of RG6279, a bispecific anti-PD1–IL-2v fusion protein directing IL-2v to PD-1+ tumor-reactive T cells.

Encouraging preclinical results have been reported for various agents in these bsAb categories, including 4-1BB×HER2,380 4-1BB×CD40,381 4-1BB×PD-1/PD-L1,382,383 PD-1×GITRL,384 PD-L1×LAG-3,385 PD-1×CTLA-4,386 PD-L1×IL-15/IL-15RA387 bispecifics, and others. However, in general, except for several group II bsAbs, most others are still at early phases of development. BsAbs entered in phase III trials include cadonilimab, erfonrilimab, tebotelimab, retlirafusp alfa, and ivonescimab (Table 2 and Supplementary Table 4). Cadonilimab is an Fc-silenced symmetric IgG1 PD-1×CTLA-4 bsAb. Combined with chemotherapy, cadonilimab elicited an ORR of 65.9% in phase I/II trial for GC/GEJC.388 In the phase I/II trial for PD-L1 TPS ≥ 1% NSCLC, cadonilimab combined with anlotinib induced an ORR of 62.5%.389 Thus, cadonilimab combined with chemotherapy or targeted therapy elicited excellent ORRs (Table 2). Erfonrilimab is a symmetric full-length IgG1 PD-1×CTLA-4 bsAb. Combined with chemotherapy, erfonrilimab induced ORRs of 50.6%, 58.3%, and 55.6% in NSCLC,390 ESCC,391 and PDAC patients.392 A similar ORR of 57% was also observed combining erfonrilimab and lenvatinib in HCC patients.405 Notably, antagonizing m6A modifiers can sensitize tumors to PD-1 blockade in mice.406,407,408,409,410 However, most agents targeting m6A regulators are still in preclinical development and none has entered clinical evaluation.411 Thus, considering the volume and scope of this review, we mainly focus on DNA methylation by DNA methyltransferases (DNMTs), histone deacetylation by histone deacetylases (HDACs), recognition of acetylated histone by the mammalian bromodomain and extra-terminal (BET) proteins, and demethylation by histone methylase polycomb repressive complex 2 (PRC2) and lysine-specific histone demethylase 1 (LSD1) (Fig. 4).

Fig. 4
figure 4

Epigenetic targets and their impact on different immune cell types and tumor cells in the TME. Lines of the same color indicate the impacts on different cell types of the same epigenetic process. And the color of lines corresponds to the background color of the specific process those lines indicated. Epigenetic regulation mainly comprises transcriptional regulation via DNA methylation, histone modification, and post-transcriptional modification. Immune-related pharmacological development has mainly focused on DNA methylation by DNMTs, histone deacetylation by HDACs, recognition of acetylated histone by BET proteins, and histone demethylation by PRC2 and LSD1. DNA methylation, which is mainly mediated by DNMTs, represses gene transcription when located in a gene promoter and regulates anti-tumor immunity with the orchestration of different cell members. The aforementioned histone modifications are capable of remodeling chromatin structures and interactions with other regulating factors (e.g., recruitment of transcription factors) and affect gene transcription of various cell types in the TME. The post-transcriptional m6A methylation represents a new layer of epigenetic regulation that mainly affects the fate of RNAs via promoting or antagonizing their degradation or translation. Classification of drugs of each epigenetic target are indicated in the blue boxes. inh inhibitor, DNMT DNA methyltransferase, HDAC histone deacetylase, BET bromodomain and extraterminal domain, BRD bromodomain, BRDT bromodomain testis-specific protein, RNA pol RNA polymerase, PRC2 polycomb repressive complex 2, EZH enhancer of zeste homolog, EED embryonic ectoderm development, SUZ suppressor of zeste, LSD1 lysine-specific demethylase 1, m6A N6-methyladenosine, SASP senescence-associated secretory phenotype

Full size image

DNA methyltransferases (DNMTs)

Targeting DNA methylation has become important for the treatment of certain hematological malignancies with the intention to reactivate tumor suppressors and promote differentiation of the malignant cells. Regarding anti-tumor immunity, therapeutic DNA demethylation can enhance tumor immunogenicity by inducing expression of endogeneous retroviral elements and of neoantigens normally silenced by DNA methylation. Expression of the former induces double-strand RNA which, in turn, can induce interferon-based innate immune activation essential for adaptive antitumor immunity, and is one reason why DNMT inhibitors can cause immunogenic cell death of malignant cells.412,413 Furthermore, therapeutic DNA demethylation can alter the composition and behavior of immune cells; it can increase the expression of MHC molecules, alleviate T cell exhaustion and enhance T cell effector and memory potential, increase secretion of Th1-type cytokines, and reduce immunosuppressive myeloid and Treg cells.412,413 DNA methylation status and demethylating agents can also directly affect the expression of multiple immune checkpoints, including PD-1,414,415 PD-L1,414,416,417, LAG3,414 TIM-3,414,418,419, CTLA-4,414,420 and TIGIT,421,422 by recruiting of proteins involved in gene repression or by inhibiting the binding of them.

Because of these interesting antitumor immune effects, combinations of hypomethylating agents (HMAs), currently mainly DNMT inhibitors, with immunotherapeutics are being investigated. Decitabine plus camrelizumab caused high response rates and long-term benefits in patients with Hodgkin’s lymphoma who failed PD-1 inhibitors.423,424 The combination of decitabine and pembrolizumab induced better response in patients with relapsed AML, with transcriptional signs of immune activation.425 Other combinations of HMAs and ICIs also show good safety and preliminary anti-tumor effects in patients with hematological malignancies in clinical trials426,427,428,429 (Supplementary Table 5). Regarding solid tumors, although the preclinical and some early clinical results using the combination of PD-1 blockade and HMAs are highly promising,430,431,432 most clinical data has been disappointing. No responses were observed after guadecitabine plus atezolizumab in metastatic urothelial carcinoma which had progressed on previous immune checkpoint blockade (ICB).433 The combination of guadecitabine or azacytidine and pembrolizumab or durvalumab produced only modest anti-tumor effects in a variety of solid tumors.434,435,436 The addition of azacytidine or CC-486 (oral azacytidine) to pembrolizumab437,438 or durvalumab439 was not more effective than standalone ICI treatment. Lack of robust tumor DNA demethylation and of viral mimicry was found to be associated with a missing clinical response in one study.439

Overall, the combination of HMAs and ICIs needs further studies, especially in solid tumors. Notably, investigations of how dosing and scheduling of these drug classes affect the immunomodulatory and anti-tumor effects in the clinical setting are expected. In mouse solid tumor models, low-dose HMAs plus ICIs outperform either HMAs or ICIs alone in restricting tumor growth and prolonging survival, with significant HMA-related immune modulation.430,431 Epigenetic priming using HMAs with sequential ICIs has the potential to produce durable clinical benefit associated with immune responses in patients with solid tumors.440,441 In addition, there is considerable interest in the development of compounds targeting a selective subtype of DNMTs, which may enhance the tolerability and efficacy.442,443,482 Considering the current evidence and the potential of BET proteins in cancer and immune-related diseases,483 the exploration of their impacts on anti-tumor immunity and the development of drugs targeting BETs worth more effort.

Histone methylase polycomb repressive complex 2 (PRC2)

PRC2, which is formed when zeste homolog 2 (EZH2) associates with embryonic ectoderm development (EED) protein and SUZ12, is responsible for histone methylation mainly at histone 3 lysine 27 (H3K27). It has a broad impact on cancer immunity484 (Fig. 4). It mediates long-term transcriptional silencing of the MHC-I antigen processing pathway485 and represses CXCL9 and CXCL10 production by tumors, two critical chemokines for effector T-cell trafficking.486,487 The orchestrated immune modulation also includes higher MDSC infiltration, less NK cell-mediated killing and more Treg-mediated immune suppression.484,488 EZH2 inhibition could enhance the efficacy and overcome resistance to current immunotherapies.488 Tazemetostat, an inhibitor of EZH2, the main catalytic unit, demonstrated clinical activity in epithelioid sarcoma in a phase II trial (ORR: 15%, duration of response: not reached).489 It was approved by FDA for locally advanced or metastatic epithelioid sarcoma in 2020. EZH1, a paralog of EZH2, can also form functional PRC2 complexes as a compensatory mechanism for tumor cells to escape EZH2 inhibition.490,491 Therefore, co-inhibition of EZH2 and EZH1492,493,494 or EED inhibition492,495 could more completely inhibit the activity of PRC2, especially in the presence of innate or acquired resistance mutations in EZH2 and by addressing the potential compensatory mechanism of EZH1-driven tumor growth. SETD2, an upstream regulator of EZH2, can also be targeted to combat EZH2-high tumors.496 Future preclinical and clinical investigations may identify novel drug targets and formats, and will provide more insight into the value of PRC2 inhibition in cancer immunotherapy.

Lysine-specific histone demethylase 1 (LSD1)

LSD1 inhibitors are widely applied in myeloid hematological malignancies as they promote the differentiation of myeloid cells via regulation of myeloid transcription factors GFI1 and PU.1.497,498 Regarding anti-tumor immunity, LSD1 undermines T cell-mediated cytotoxicity via promoting terminal differentiation of T cells499,500 (Fig. 4). Accordingly, LSD1 inhibition expands progenitor exhausted T cells with stem-like properties, thereby enhancing the efficacy of immunotherapy.499,500 LSD1 inhibition also increases antigen presentation mediated by MHC I complexes on cancer cells485,501 and decreases exosomal PD-L1.502 Although tranylcypromine-based flavin adenine dinucleotide (FAD) domain-binding irreversible inhibitors exert long-lasting inhibition on LSD1 and yield encouraging clinical results both in myeloid malignancies503 and solid tumors,504 they induce significant TRAEs due to their covalent binding to FAD domains contained in critical enzymes other than LSD1 and the ensuing off-target reactivity.505 This could be ameliorated using reversible LSD1 inhibitors. For example, minimal inhibition of the cytochrome P450 enzymes containing a FAD domain was reported using the reversible LSD1 inhibitor TACH101.506,507 Another two clinical stage reversible LSD1 inhibitors, seclidemstat508 and CC-90011,509 also show immune activation and efficacy in combination with ICIs. Selectively targeting nuclear LSD1 phosphorylated at serine 111 (nLSD1p) might also be a plausible therapeutic approach to tackle the safety issue.510 Similar to most other anti-cancer agents targeting immuno-epigenetic modifiers, LSD1 inhibitors are combined with PD-1/PD-L1 blockade in phase I/II trials with promising preliminary results.511

Targeting epigenetics faces problems related to broad specificity and pleiotropic activity. Discovered immunomodulatory effects of some existing epigenetic modulators might contrast their previously known antitumor functions. For example, although HDAC activity generally seems to impair anti-tumor immunity, the intrinsic HDAC activity of Tcf1 and Lef1 is crucial for maintaining CD8+ T cell identity.555,648,649 leads to an arduous efficacy-toxicity balance (Fig. 5a). Tissue-/cell-specific therapeutics and/or conditionally activated agents might help to overcome these problems. Both cytokine biology research and protein engineering and novel delivery platforms for cytokines have greatly advanced in recent years. They will hopefully help to design better drug structures and to expand the realm of targetable cytokines, continuously promoting the development of cytokine-based therapeutics.

It is worth noting that from a clinical perspective, differences exist in clinical practices for treatment of different cancer types. In fact, due to the varying immune backgrounds and intrinsic differences between cancer types, immunoregulatory anti-cancer therapies targeting different targets indeed have different optimal indications. Melanoma is well known for its robust immune responsiveness, which made it predestined for initial evaluation of therapeutic potential of LAG-3, TIM-3, CD40, and other immunoregulatory targets. Pembrolizumab induces CR in melanoma patients, and over 90% maintain CR for 5 years,650 highlighting potent efficacy of ICB. Relatlimab and tebentafusp were also approved for melanoma as their first indication. Activating the immune system against melanoma through cytokine-based therapies such as aldesleukin, darleukin, tavokinogene telseplasmid, has also proven to be effective.529,651,652

For other solid tumors, additional checkpoints like LAG-3 and TIGIT are likely to play significant roles. In the case of NSCLC, the relatively favorable immune environment in most NSCLC cases653 suggests that targeting these additional checkpoints could potentially be advantageous. LAG-3 agents have demonstrated efficacy in solid tumor entities such as NSCLC and HNSCC,52,53,54 and TIGIT agents are currently in several clinical trials in combination with PD-1/PD-L1 agents in NSCLC (Table 1). Exploration of B7 is ongoing across various solid tumors.654 The ADC enfortumab vedotin targeting nectin-4 has shown promising results particularly in combination with pembrolizumab in bladder cancer (NCT04223856). The situation is different for SCLC. While atezolizumab combined with carboplatin/etoposide is approved as first-line treatment for extensive-stage SCLC, many SCLC subtypes still respond weakly.655,656 These non-immunogenic tumor subtypes may rely on TAAs, such as DLL3 to be targeted, for example, with CD3×DLL3 TCE tarlatamab and HPN328. Neuroendocrine features of SCLC can also be managed with LSD1 inhibitors such as ladademstat to suppress neuroendocrine transcription factors.656,657,658

In gastrointestinal tumors, excellent efficacy for GC/GEJC is primarily observed with regimens based on anti-PD-1 agents and bsAbs containing anti-PD-1 scFv, such as cadonilimab388 and tebotelimab.659 For PDAC, CAFs are the main component of its TME, forming a strong physical barrier with the ECM that hampers T cell infiltration.660 CD40 agonists like sotigalimab can enhance T cell infiltration and show efficacy in combination with chemotherapy and nivolumab.319 Inhibiting TGF-β with NIS793 in combination with anti-PD-1 agents may help remodel the CAF-rich TME of PDAC.578,579 HCC is immune-privileged, with abundant MDSCs and an abnormal vascular system.661 Non-inflammatory HCC subtypes predominate,661 requiring anti-PD-1-based immunotherapy combined with anti-angiogenic therapy or dual immunotherapy to enhance immune response. Currently, atezolizumab plus bevacizumab is the first-line treatment for advanced HCC, with nivolumab plus ipilimumab and durvalumab plus tremelimumab also demonstrating efficacy.662,663 New bsAbs such as erfonrilimab and cadonilimab plus lenvatinib have achieved very high ORR.671,672 with no immunotherapy yet proven to improve prognosis for gliomas.672 Adenovirus-encoded IL-12 INXN 2001 (Table 3 and Supplementary Table 6) may help ameliorate the suppressive microenvironment. CD155 serves as both the ligand for the inhibitory receptor TIGIT and the poliovirus receptor. The polio-rhinovirus chimera lerapolturev (Table 1 and Supplementary Table 1) offers some hope for treatment of gliomas.673 Redirecting anti-tumor immunity relying on TAAs is also an important strategy for immunotherapy of pediatric and nervous system tumors, with the mAbs or TCEs targeting B7-H3 and GD2 showing promise.674,675

For gynecologic tumors, each of the main cancer types presents distinct characteristics. Immunotherapy for endometrial cancer is mainly limited to the MSI-H/dMMR subgroup, represented by PD-1 agents such as pembrolizumab and dostarlimab,676 with the potential of other immunoregulatory therapies yet to be explored. Ovarian cancer exhibits a highly immunosuppressive TME and relatively low immunogenicity,677 resulting in poor response to immunotherapy. Combination therapies blocking multiple checkpoints, bsAb like ubamatamab or non-α IL-2 variant nemvaleukin alfa modulating the TME, can possibly enhance immune responses.677,678,679 Cervical cancer shows better responses to immunotherapy,680 with promising outcomes observed with pembrolizumab plus chemotherapy (recently approved by FDA), anti-PD-1 plus anti-CTLA-4 agents, and bsAbs.680,681,682

Beyond the drug therapies discussed in detail, other therapies, such as cancer gene therapy and cancer vaccines, are also promising anti-cancer treatments with immunoregulatory effects. While they may not be totally classified as conventional drug therapies, their rapid development and effectiveness are noteworthy. Cancer gene therapy can alter genes in vivo or ex vivo. Ex vivo gene therapy, represented by CAR T cell therapy, has achieved great clinical successes.683,684,685 Additional genetic modifications hold promise to further improve cell therapy, as manifested by the good safety and feasibility of CRISPR-edited TCR T cells and CAR T cells in patients with solid tumors.686,687 In-vivo gene therapy introduces the target gene directly into patients using a vector. Stimulating intra-tumoral cytokine gene expression (elaborated in the section on cytokines) and co-stimulatory molecules as well as inhibiting immunosuppressive molecules/cell types with anti-sense oligonucleotides (ASOs) and small interfering RNAs (siRNAs) attract high research interest. The combined local delivery of OX40L, CD80, and CD86 mRNAs cause significant local and systemic immune activation and facilitate tumor regression at both local and abscopal sites.688

Cancer vaccines amplify the signal of tumor-specific antigens (TSAs) or TAAs via encapsulated antigen-encoding DNA and RNA, peptides, or antigen-loaded APCs.689,690,691 They actively stimulate patients’ own anti-tumor immune response at the very beginning of the cancer immunity cycle. With promising potential,692,693,694,695,696 current research focuses on identifying antigens with the best quality and optimizing the delivery platform.689,690,691 High-throughput sequencing and bioinformatics tools in recent years have greatly facilitated the screening of highly immunogenic neoantigens.692,697 Improved vectors698,699 and immune adjuvants700,701,702 have been reported to enhance the efficacy of vaccine delivery and the ensuing immune activation and tumor-killing effects in preclinical studies. Combining cancer vaccines with ICIs and TME-reprogramming may help tackle the problem of immunosuppressive TME observed in classical immunotherapies, and is now being extensively tested in clinical trials with some producing promising results and advancing into phase III evaluations (NCT05141721, NCT06077760), but the optimal combination, dosage, and sequence of combination therapy still require further exploration. In the broader landscape of immunoregulatory cancer therapeutics, it is essential to recognize the contributions of these diverse approaches.

In summary, we have reviewed highly promising avenues for the development of immunoregulatory anti-cancer therapeutics by analyzing a large volume of recent published research, also including conference reports, and clinical trials. We summarized recent advances in the understanding of the mechanisms of action of classes of immunotherapy drug targets and the progress of the corresponding drug development. Despite considerable success so far, further research is necessary to boost drug development to improve treatment responses and prolong cancer patient survival. Moreover, next-generation drug development in these immunotherapy fields will continue to rely on clarification of immunological target biology and progress in drug developmental platforms, whereas the final evaluation of drug efficacy depends on rigorous high-quality clinical trials. This needs effective cooperation of academia, pharmaceutical and biotech industry, and the clinical medical community. More and more promising pharmacological immunoregulatory anti-cancer therapeutics are likely to be developed in innovative forms to the benefit of patients. This will further expand and enrich the landscape of immunoregulatory anti-cancer therapies.