Why has immune “checkpoint” therapy failed in most clinical trials?

Abstract

Cancer immunotherapy targeting the PD-1/PD-L1 pathway has demonstrated efficacy across a range of common solid tumors and some hematopoietic malignancies. Despite these groundbreaking successes, the clinical development of other ‘checkpoint inhibitors’ targeting molecules like TIM-3, TIGIT, ICOS and others, has largely fallen short, often showing minimal clinical benefit even in combination with anti-PD therapy. This article explores three key hypotheses that help explain the disparity in therapeutic success: (1) the absence of tumor- specific immunosuppressive logic in many checkpoint targets, (2) the dominance—but not redundancy—of immune evasion mechanisms within the tumor microenvironment (TME), and (3) the emergence of therapy-induced resistance. This is not intended as a comprehensive review of the literature. Instead, it highlights select evidence to explain past failures and to illuminate a more strategic, biologically informed path forward.

Introduction

The success of immunotherapy targeting the PD-1/PD-L1 axis (hereafter anti-PD therapy) has fundamentally transformed oncology, cementing immunotherapy as a pillar of cancer treatment. While anti-PD therapy has demonstrated exceptional performance in managing various types of solid tumors and hematologic malignancies, it has been recognized that its clinical activity is limited to a subset of cancer patients. Inspired by the concept of immune ‘checkpoint’, tremendous efforts have been made to identify and interrogate additional ‘checkpoints’, aiming to provide novel therapeutic strategies to patients who are refractory to current immunotherapies. However, most clinical trials for other “checkpoint inhibitors” have shown disappointing response rates, leading to significant doubts about the path forward for this field. These setbacks are not merely the result of flawed trial designs or inadequate biomarker-based patient selection. Rather, they reflect deeper biological principles rooted in tumor immunology and evolutionary dynamics. Hence, here we analyze the lessons we have learned from the past success and failures, with the hope to provoke therapeutic innovations that can help the immuno-oncology field overcome the current bottleneck.

The importance of functioning location: TME-specific versus systemic immunoregulation

The key to the success of anti-PD therapy lies in its specificity for a tumor-adaptive immune evasion mechanism that is both spatially and functionally confined to the TME. On infiltration into tumors, activated cytotoxic tumor-infiltrating lymphocytes recognizing tumor antigens secrete interferon-gamma.^{1 2} This cytokine induces the upregulation of PD-L1 on tumor cells and surrounding host stromal cells, thereby establishing a negative feedback loop that compromises T cell-mediated antitumor immunity at the effector phase. In this localized system, the tumor directly suppresses the same immune cells that threaten its survival. Crucially, this suppressive interaction is largely restricted to the tumor site, minimizing systemic immune dysregulation and toxicity.

In contrast, most alternative “checkpoint” targets lack such TME-specific or TME-selective logic. Cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) on T cells binds with high affinity to B7-1 and B7-2, co-stimulatory ligands broadly present on myeloid cells throughout the body. Mainly acting at the priming phase of T-cell activation, CTLA-4 blockade not only impacts tumor-reactive T cells but also disrupts the regulatory T cell-mediated maintenance of peripheral tolerance. The result is a systemic unleashing of immune responses, leading to significant risk of immune-related adverse events (irAEs) such as colitis, dermatitis, hypophysitis, and other autoimmune-like toxicities. However, CTLA-4 blockade can synergize with anti-PD therapy and salvage response in patients who failed anti-PD therapy, outweighing the risk of irAEs.^{3 4} Enormous efforts have been made to identify other actionable immune-regulatory targets, such as TIM-3 and lymphocyte activation gene 3 (LAG-3), which are similarly problematic. While they are upregulated on T cells within the TME and in peripherals, their ligands, galectin-9 (for TIM-3), major histocompatibility complex (MHC) class II, and fibrinogen-like protein 1 (FGL-1, for LAG-3), are expressed mainly in non-malignant tissues.⁵ This expression pattern leads to off-TME immune stimulation and limits therapeutic specificity. Transforming growth factor beta (TGF-β) represents another instructive example. Although it plays a pivotal role in immune suppression, TGF-β orchestrates numerous physiological processes such as wound healing and epithelial-to-mesenchymal transition. Systemic inhibition of TGF-β can thus trigger catastrophic inflammation and multi-organ fibrosis, undermining its therapeutic utility despite its relevance to immune evasion.

These observations underscore a critical principle: for checkpoint blockade to be effective and safe, the target may need to be deeply embedded in the unique immune architecture of the TME. Without such specificity, systemic immune activation may become indiscriminate and even toxic. This principle should be carefully taken into consideration during both discovery and development stages of new therapeutics.

Evolutionary dominance and the myth of redundant immune evasion

Tumors evolve under continuous immunologic pressure. In this evolutionary arms race, only the most effective immune evasion strategies are retained. The notion that tumors rely on multiple, redundant immune evasion mechanisms, such as multiple functionally similar immune checkpoints, is increasingly seen as flawed. Instead, once a dominant mechanism, such as tumorous PD-L1 upregulation, is established, there may be little to no selective pressure for additional suppressive pathways that dampen the same perspective of anti-tumor immunity.⁶

This principle explains why anti-PD therapy can work in a selected group of patients. While multiple checkpoints may be detectable in the TME, only those under evolutionary selection are likely functionally relevant. Others may be vestigial or context-dependent. For instance, TIM-3 or TIGIT expression may be a byproduct of T-cell activation and subsequent dysfunction rather than direct suppressors of immune activity—unless their ligands are also present and dominant in that specific tumor context. It is also important to recognize that tumor-derived immune evasion strategies are not limited to immune “checkpoints”, which will be elaborated below.

Therapy-induced resistance

Therapy-induced resistance is conceptually related to evolutionary dominance; however, the selective pressure in this case is artificial, driven by cancer therapies themselves. Before receiving new immune checkpoint therapies, most patients undergo extensive treatment with standard-of-care modalities, including chemotherapy, radiotherapy, and targeted therapies. These interventions exert complex and multifaceted effects on the immune system, including direct cytotoxicity to immune and host cells, modulation of immune responses, induction of cellular and molecular adaptations, and the selection of therapy-resistant tumor variants. As a result, the TME becomes highly heterogeneous. Furthermore, with the widespread adoption of anti-PD therapy as a new standard of care, tumors may evolve additional mechanisms of immune evasion. Consequently, both the TME and the host immune system display substantial diversity and heterogeneity. It is likely that, alongside pre-existing resistance mechanisms, new adaptive responses may emerge under continued therapeutic pressure.

Recent advances in single-cell analysis and spatial profiling of tumor tissues have revealed that tumors previously exposed to standard therapies, including anti-PD therapy, exhibit profound genetic, epigenetic, transcriptomic, and proteomic heterogeneity.⁷ These prior treatments impose additional selective pressures on the TME and exert systemic effects on the immune system, further shaping the course of tumor evolution. In this context, previously subdominant immune checkpoints or novel immune evasion pathways may become functionally significant. These alternative mechanisms can compensate for the disruption of primary immunosuppressive pathways targeted by therapy, allowing tumors to adapt, survive, and ultimately resist new immune checkpoint interventions.

Recognizing the role of therapy-induced resistance is essential for designing effective immunotherapy strategies in the future. It highlights the importance of considering a patient’s treatment history and the dynamic, evolving nature of the TME. The sequencing, timing, and combination of therapies should be informed by a deep understanding of how prior interventions reprogram tumor–immune interactions and foster resistance.

The path forward

Given these insights, the future of “checkpoint” therapy cannot rest on empirical combinations of immune modulators. Instead, it ought to be guided by a rigorous understanding of the tumor’s immunologic architecture, evolutionary dynamics, and context-specific vulnerabilities (figure 1). Several principles emerge:

Figure 1. The three key hypotheses that help explain the disparity in success of cancer immunotherapy. The central principle behind the success of anti-PD therapies for the subset of patients positive for both tumorous PD-L1 expression and tumor-infiltrating lymphocytes (TILs) is the tumor microenvironment (TME)-specific induction of an evolutionarily dominant immunosuppressive mechanism. From a Darwinian perspective, survival challenges imposed by the anti-PD therapies might trigger the tumor cells to select for acquired resistance mechanisms to evade antitumor immunity. Hence, instead of the “additional checkpoints” that lack specificity to the TME, presume redundant functional mechanisms and have limited relevance to therapy-induced resistance, mechanism-based discovery and development of novel immuno-therapeutic strategies against functionally orthogonal targets, while incorporating other important lessons we have learned from the past success and failures, such as biomarker-based patient stratification and dynamic target identification, will be the backbone for the next breakthrough of immuno-oncology. B7-H1, B7 homolog 1; IFN-γ, interferon-gamma; LAG-3, lymphocyte activation gene 3; MHC-I, major histocompatibility complex class I; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; TCR, T-cell receptor; TIGIT, T-cell immunoreceptor with Ig and ITIM domains; TIM-3, T cell immunoglobulin and mucin domain-containing protein 3.

Mechanistic stratification: Therapeutic strategies must be tailored to the tumor’s dominant immune evasion mechanism, which may vary across cancer types, stages, and treatment histories. This necessitates comprehensive profiling of tumor cells, tumor-infiltrating immune cells, and molecular expression patterns. Most failed “checkpoint” trials lacked rigorous biomarker stratification, leading to false negatives and misinterpretation of efficacy. Moving forward, trials should be designed around clearly defined immune-resistance phenotypes and informed by high-dimensional biomarkers such as transcriptomic profiling, spatial proteomics, and multiplex immunohistochemistry.

Dynamic target selection: Static treatment algorithms fail to account for tumor evolution. Longitudinal monitoring of the TME through serial biopsies, circulating tumor DNA, or advanced imaging techniques, ideally pretreatment, on-treatment and post-treatment, may enlighten the discoveries of evolving resistance mechanisms that enable dynamic adjustment of therapeutic targets in response to emerging resistance.

Beyond immune “checkpoints”: In many cases, combination therapy may need to move beyond additional immune checkpoints altogether. Modulation of the TME through orthogonal strategies, including metabolic reprogramming, stromal remodeling, vascular normalization, or even neural input, may prove more effective in overcoming resistance than layering more “checkpoints” with overlapping mechanisms. An example is recent promising trials with anti-PD therapy combined with Vascular endothelial growth factor (VEGF) and TGF-β inhibitors, especially those in the format of bispecific antibodies for TME-selective delivery.^{8 9} PD-L1/PD-1-mediated immune evasion mechanisms account only for a fraction of advanced patients. Many orthogonal evasion mechanisms that abrogate different core elements of antitumor immunity are present in the TME and yet to be discovered. For example, the lack of T-cell infiltration into the TME has been recognized as a pivotal resistance mechanism, potentially sharing the same level of evolutionary dominance with PD-1/PD-L1-mediated immune regulation.⁶ While one heavily exploited strategy is to boost inflammation for enhanced T-cell activation and recruitment, such as leveraging toll-like receptor agonists and oncolytic virus, we recently proposed that tumors could secrete soluble factors to exclude T cells from infiltrating into the TME.^{10 11} An example is phospholipase A2 group 10 (PLA2G10), which is aberrantly produced by various human solid tumors. PLA2G10 inhibitor could increase T cell infiltration, leading to inflammatory microenvironment in mouse tumor models.¹² Growth differentiation factor 15 (GDF-15), a cytokine in the TGF-β superfamily, is highly upregulated in the TME and plays a role in regulating inflammation. GDF inhibitor shows promising clinical responses in non-small cell lung cancer (NSCLC) and urothelial cancer.¹³ Tumor-associated endothelial cells present dysregulated (cluster of differentiation 93) CD93/insulin-like growth factor-binding protein 7 (IGFBP7) axis, resulting in vascular dysfunction in the TME. CD93 blockade has demonstrated robust in vivo single-agent activity for vascular normalization that synergized with anti-PD therapy to unleash antitumor immunity.¹⁴ Sialic acid-binding Ig-like lectin 15 (Siglec-15) is widely upregulated in cancer tissue with a mutually exclusive expression pattern with PD-L1, and therapeutic perturbation of Siglec-15-mediated immune suppression has shown encouraging clinical response in patients with NSCLC resistant to anti-PD therapy.¹⁵ These findings highlight the importance of continued discovery and validation of dominant immune evasion mechanisms that dictate the TME-specific “abnormality” in each tumor type of the Tumor Immunity in the MicroEnvironment (TIME) classification, rather than assuming uniform “checkpoint” redundancy.⁶

The technological challenge: Recent technological advancements, such as bispecific antibodies, could facilitate TME-selective delivery of immune modulatory agents, and preliminary data have shown promising activity in clinical trials.⁹ However, one of the greatest barriers to progress is the current limitation in biomarker technologies. While advances in single-cell sequencing and spatial transcriptomics have yielded valuable insights, these tools are not yet widely available or scalable for routine clinical use. Bridging this gap will require concerted efforts to develop and validate predictive, on-treatment, and post-treatment biomarkers that can stratify patients by resistance phenotype and guide combination strategies in real time. Ultimate solutions will rely on the development of real-time imaging technology to trace evolutionary dynamics of immune regulatory molecules. Moreover, computational models incorporating systems biology, machine learning, and evolutionary dynamics may help predict how tumors will respond to immune perturbations and when adaptive resistance is likely to emerge. Integration of these models with clinical and molecular data could transform how we design and interpret immunotherapy trials.

Conclusion

The failure of most ‘checkpoint’ blockade therapies is not mysterious, and they are the predictable result of ignoring the evolutionary rules that govern tumor immunology. The success of PD-1/PD-L1 blockade stems from its targeting of a dominant, tumor-adapted immune evasion mechanism embedded in the TME. In contrast, other checkpoints often lack both tumor specificity and evolutionary relevance. Moving forward, checkpoint therapy must shift from empiricism to precision, from redundancy to orthogonal dominance, and from systemic suppression to TME-specific targeting. The era of ‘checkpoint fishing expeditions’ is over. The next wave of breakthroughs will come from Darwinian-informed, mechanism-based interventions that strike at the heart of resistance. Only then can we realize the full potential of immunotherapy.