Cancer Immunotherapy: 2016 vs 2026

A decade can completely change the language of oncology.

In 2016, cancer immunotherapy had already emerged as one of the most transformative areas in cancer treatment. CTLA-4 and PD-1/PD-L1 checkpoint inhibitors were demonstrating that restoring antitumor T-cell activity could produce durable responses, while CAR-T cells, engineered T-cell therapies, and increasingly sophisticated immune biomarkers were beginning to redefine what might be possible.

The scientific literature of the time reflected this transition. Reviews such as “The Future of Cancer Treatment: Immunomodulation, CARs and Combination Immunotherapy” and “Engineered T Cells: The Promise and Challenges of Cancer Immunotherapy” described a field moving beyond conventional cancer treatment toward immune modulation and cellular engineering.

Ten years later, many of those emerging concepts have developed into major therapeutic platforms. By 2026, immunotherapy extends far beyond checkpoint blockade, encompassing engineered cellular therapies, TILs, NK-cell approaches, bispecific and multispecific immune engagers, perioperative immunotherapy, increasingly sophisticated biomarker strategies, and individualized neoantigen-based approaches.

The story of 2016 to 2026 is therefore not simply one of more immunotherapy drugs. It is the evolution of immuno-oncology from broadly releasing immune inhibition toward understanding, redirecting, engineering, and increasingly personalizing the antitumor immune response.

NK-Cell Immunotherapy: The Next Frontier in Cancer Treatment

2016: When Releasing the Immune Brakes Was Transforming Oncology

To understand how dramatically immuno-oncology has changed, it is worth returning to the scientific landscape of 2016.

That year, Danny N. Khalil, Eric L. Smith, Renier J. Brentjens, Jedd D. Wolchok and colleagues published “The Future of Cancer Treatment: Immunomodulation, CARs and Combination Immunotherapy” in Nature Reviews Clinical Oncology. The review captured a field at a pivotal moment: immune-modulating antibodies were producing unprecedented clinical responses, while adoptive cellular therapies were beginning to demonstrate what genetically modified lymphocytes might accomplish. The authors also anticipated that combinations would become central to expanding the reach of immunotherapy.

At the center of this transformation were CTLA-4 and PD-1/PD-L1, physiological regulators that limit excessive T-cell activation but can also be exploited by tumors to escape immune destruction. Engagement of PD-1 by PD-L1 or PD-L2 suppresses activated T-cell function; blocking this interaction can restore activity in tumor-reactive lymphocytes that have already encountered antigen but are functionally restrained.

This was the defining therapeutic logic of the early checkpoint era: rather than giving the immune system a new target, checkpoint blockade could remove inhibitory signals preventing an existing antitumor response from functioning.

But the extraordinary responses produced in some patients immediately exposed the next problem: most patients did not experience the same benefit.

From PD-L1 to the Immune Architecture of Cancer

Why the same checkpoint inhibitor could produce durable tumor control in one patient and little activity in another became one of the central questions of immuno-oncology.

In the 2016 Nature Reviews Cancer article “Mechanism-Driven Biomarkers to Guide Immune Checkpoint Blockade in Cancer Therapy,” Suzanne L. Topalian, Janis M. Taube, Robert A. Anders, Drew M. Pardoll and colleagues argued that biomarkers should reflect the biology of the immune pathway being targeted. PD-L1 was emerging as an important marker for anti-PD-1 therapy, but its limitations soon became evident: expression can vary between tumor regions, primary and metastatic lesions, tumor and immune cells, and across different points in time.

Other work was already broadening the biological framework. In “Innate Immune Signaling and Regulation in Cancer Immunotherapy,” Leticia Corrales, Vyara Matson, Blake Flood, Stefani Spranger, Thomas F. Gajewski and colleagues highlighted the importance of innate immune sensing, type I interferon signaling, the STING pathway, and Batf3-lineage dendritic cells in generating spontaneous antitumor T-cell responses.

Together, these observations helped establish a principle that is now central to IO: successful checkpoint blockade depends on a much larger sequence of biological events. Tumor antigens must exist and be presented; tumor-reactive lymphocytes must be primed and recruited; those cells must enter the tumor and retain sufficient effector function; and local suppressive mechanisms must not completely overwhelm them.

By 2026, biomarker science increasingly reflects this complexity. PD-L1 remains clinically important, but it now sits within a broader landscape that includes:

• MSI/MMR and tumor mutational burden, reflecting aspects of tumor immunogenicity;
• genomic and antigen-presentation alterations, including pathways capable of affecting immune recognition;
• immune-cell composition and functional states within the tumor;
• spatial architecture, revealing where immune populations are located relative to malignant and stromal cells;
• ctDNA and molecular residual disease, offering the possibility of monitoring tumor burden dynamically rather than relying exclusively on pretreatment tissue.

The conceptual change is from identifying a single marker associated with response toward understanding the biological state of the tumor–immune ecosystem.

This direction is also reflected in the 2026 Nature Reviews Cancer article “Advancing AI for Multi-Omics and Clinical Data Integration in Basic and Translational Cancer Research” by Fei Liu, Stephan Beck, Lei Yang, Huiyan Luo and colleagues, which examines how genomic, transcriptomic, proteomic, imaging and clinical information can increasingly be integrated computationally. For immuno-oncology, such approaches are particularly relevant because a highly mutated tumor may still be poorly infiltrated, an infiltrated tumor may have defective antigen presentation, and a tumor containing abundant immune cells may nevertheless maintain a profoundly suppressive environment.

The future IO biomarker may therefore be less likely to resemble one molecule and more likely to represent an integrated immune state.

Immunotherapy Moves Into Earlier Disease

The therapeutic setting has changed just as substantially as the biomarker landscape. The first checkpoint-inhibitor successes were largely established in advanced and metastatic cancers, but over the following decade immunotherapy moved progressively into adjuvant, neoadjuvant and perioperative treatment.

The biology of neoadjuvant immunotherapy is particularly compelling. Before surgery, the primary tumor remains available as a broad source of antigens, while tumor-draining lymph nodes and tumor-reactive lymphocyte populations are still present. Checkpoint blockade during this period may therefore promote expansion of tumor-specific T-cell populations while the immune system continues to encounter a substantial antigenic repertoire. For a limited period, the tumor can function not only as the disease being treated but also as an in situ source of antigen for systemic immune priming.

The clinical questions are consequently evolving beyond whether immunotherapy should be administered before or after surgery. Pathological response, ctDNA clearance and molecular residual disease are increasingly being investigated as measures of treatment effect and residual risk, raising the possibility that future treatment intensity could be adapted according to an individual patient’s biological response.

This is a fundamentally different model from conventional fixed-duration therapy: baseline biomarkers may help determine how treatment begins, while dynamic biomarkers may eventually help determine how treatment continues—or when it can stop.

From CAR-T to a Broader Cellular Immunotherapy Ecosystem

While checkpoint blockade was transforming solid-tumor oncology, another immunotherapy revolution was developing in parallel.

In August 2016, Andrew D. Fesnak, Carl H. June and Bruce L. Levine published “Engineered T Cells: The Promise and Challenges of Cancer Immunotherapy” in Nature Reviews Cancer. Early CAR-T and TCR-engineered T-cell studies were already producing striking responses in relapsed and refractory hematologic malignancies, particularly B-cell acute lymphoblastic leukemia, but CAR-T therapy remained investigational. The first FDA approvals followed in 2017.

CAR-T introduced a fundamentally different therapeutic principle. Checkpoint blockade attempts to restore endogenous tumor-reactive immunity; a chimeric antigen receptor instead provides a T cell with synthetic recognition of a selected surface antigen. Cancer immunotherapy had moved from pharmacologically modifying immune regulation to genetically modifying the effector cell itself.

A decade later, CD19- and BCMA-directed CAR-T therapies have established cellular immunotherapy as a major treatment modality in hematologic malignancies. Yet the scientific frontier has moved beyond demonstrating that engineered cells can work. The more difficult question is what determines whether those cells can generate deep and durable tumor control.

The 2026 Nature Reviews Immunology article “Hallmarks and Correlates of Effective Adoptive Cell Immunotherapy for Cancer” by Sri Krishna, Paul F. Robbins, Frank J. Lowery, Steven A. Rosenberg and colleagues approaches adoptive cell therapy from this perspective. Effective cellular therapy depends not simply on the number of cells infused but on several interconnected characteristics:

• antigen specificity and target quality;
• T-cell differentiation and functional state;
• proliferative capacity and persistence after infusion;
• ability to traffic into and survive within tumor tissue;
• resistance to suppressive signals in the tumor microenvironment.

This represents a maturation of the field from receptor engineering toward cell-state engineering.

At the same time, cellular immunotherapy has become much broader than CAR-T. TCR-engineered T cells can recognize HLA-presented peptides derived from intracellular proteins, expanding the potential target space beyond surface antigens. TIL therapy takes another route entirely: rather than constructing new recognition, it isolates lymphocytes that have already entered the tumor and can contain naturally occurring tumor-reactive clones, expands them ex vivo, and reinfuses them at therapeutic scale.

The approval of lifileucel in 2024 for previously treated unresectable or metastatic melanoma marked an important milestone by establishing TIL therapy as an approved cellular treatment for a solid tumor.

The 2026 review “Cell-Based Cancer Immunotherapy: Milestones, Mechanistic Insights, and Emerging Therapeutic Directions” by Ji-zhao Cao, Wei Zhao and Xiao-jun Xia captures the increasing breadth of the field, encompassing CAR-T, TCR-T, TILs, dendritic-cell approaches, NK-cell therapies and macrophage-directed strategies.

Among these, NK cells provide particularly different biology. Rather than depending on conventional antigen-specific TCR recognition, NK cells integrate activating and inhibitory receptor signals, respond to cellular stress and altered HLA class I expression, mediate direct cytotoxicity through perforin and granzymes, and participate in antibody-dependent cellular cytotoxicity through CD16. These properties are driving development of allogeneic NK products, CAR-NK cells, cytokine-induced memory-like NK cells and antibody- or engager-based combinations.

The expansion of cellular therapy therefore reflects more than the arrival of additional platforms. It allows researchers to ask which immune-cell biology is best suited to a particular tumor, antigen and therapeutic objective.

Redirecting Immunity Without Engineering the Cell

Not every strategy requires removing immune cells from the patient and genetically modifying them.

Bispecific T-cell engagers can bind a tumor-associated antigen and CD3 on endogenous T cells, physically bringing the two cells into proximity and promoting cytotoxic activity. This makes their mechanism distinct from checkpoint blockade: rather than simply removing inhibition from an existing tumor-reactive lymphocyte, an engager can help construct the interaction between an endogenous T cell and a predefined tumor target.

By 2026, this has developed into a substantial therapeutic class. In “Multispecific T-Cell Engagers Attract Cancer Drug Developers,” Asher Mullard described the growing clinical footprint of bispecific T-cell engagers and the development of trispecific and other multispecific constructs intended to overcome limitations of earlier platforms.

The evolution toward multispecific molecules is particularly interesting because immune redirection is becoming increasingly programmable. A therapeutic construct may potentially combine tumor recognition with additional signals affecting immune-cell activation, costimulation or selectivity.

CAR-T changes the immune cell itself; immune engagers change how immune and malignant cells are connected. Both approaches demonstrate how far the field has moved beyond nonspecific immune stimulation.

The Tumor Microenvironment Becomes Part of the Therapy

The difficulty of translating many immunotherapies into solid tumors has made one problem increasingly clear: generating a tumor-reactive lymphocyte is only part of the challenge. The cell must reach the cancer, survive there, and remain functional within an environment that has often evolved specifically to suppress immune attack.

The tumor microenvironment can contain multiple interacting barriers:

• regulatory T cells, myeloid-derived suppressor cells and tumor-associated macrophages capable of suppressing effector immunity;
• cancer-associated fibroblasts and abnormal vasculature that can contribute to immune exclusion;
• TGF-β, adenosine and other suppressive signaling pathways;
• hypoxia, lactate accumulation and nutrient competition, creating metabolic conditions unfavorable to immune-cell function;
• defects in antigen processing or presentation, allowing malignant cells to become less visible to T cells.

These mechanisms help explain why the simple classification of tumors as “hot” or “cold” is increasingly inadequate. One tumor may lack sufficient antigenicity, another may generate antigens but fail to present them, another may prime T cells that cannot penetrate the tumor, and another may be heavily infiltrated yet contain predominantly dysfunctional immune populations.

Contemporary strategies are therefore attempting to intervene directly in this ecosystem. In the 2026 Nature Cancerreview “Oncolytic Viruses and Cytokine-Based Gene Therapies Reprogram the Tumor Microenvironment,” Joshua D. Bernstock, Lennard Spanehl and E. Antonio Chiocca discuss approaches designed to alter local immune infiltration and signaling through engineered oncolytic viruses and cytokine-based gene therapy.

A different strategy is explored by Mads Hald Andersen in “Immune Modulatory Vaccines Targeting Tumor Microenvironment Antigens: Recent Advances in Oncology and Beyond.” Here, vaccination is directed not primarily toward malignant-cell antigens but toward immunoregulatory components associated with the TME, including IDO-, PD-L1-, ARG1- and TGF-β-related biology.

This expands the definition of an immunotherapy target. The target does not necessarily have to be the malignant cell itself; it can also be the biological environment that allows the malignant cell to escape immunity.

Tumor Microenvironment: Why It Matters for Cancer Immunotherapy

Neoantigens and the Reinvention of Cancer Vaccines

The development of neoantigen-based therapies represents another major conceptual evolution.

Tumor mutational burden initially attracted attention because tumors carrying more mutations might generate more abnormal proteins and therefore more potential immune targets. But mutation quantity alone cannot determine immunogenicity.

For a somatic mutation to become an effective T-cell target, several biological steps must succeed:

• the mutation must generate an altered protein that is expressed by the tumor;
• an appropriate peptide must be generated through antigen processing;
• that peptide must bind the patient’s HLA molecules and reach the cell surface;
• a T-cell clone capable of recognizing the peptide–HLA complex must exist and expand;
• those lymphocytes must ultimately reach and function within the tumor.

A mutation is therefore not automatically a neoantigen, and a computationally predicted neoantigen is not automatically a therapeutically relevant immune target.

This is why contemporary neoantigen research increasingly focuses on neoantigen quality, presentation, clonality and immunogenicity, rather than mutation count alone.

In the 2026 Nature Communications review “Developing Neoantigen Cancer Vaccines: Where Are We Now?”, Nune Markosyan and Robert H. Vonderheide describe how next-generation sequencing and computational prediction of individual MHC-restricted neoantigens have made personalized vaccine strategies increasingly feasible.

Neoantigens offer a particularly attractive solution to one of the historical difficulties of cancer vaccination. Many traditional tumor-associated antigens are also expressed to some extent in normal tissues and are therefore constrained by central and peripheral immune tolerance. Mutation-derived neoantigens can instead create peptide sequences absent from normal cells, providing potentially more tumor-specific targets.

Cancer vaccines are also being reconsidered through approaches that do not depend exclusively on predefined personalized neoantigens. In “Therapeutic Cancer Vaccines: Development, Challenges, and Future Perspectives,” Di Di, Luan-feng Li, Tian-yang He and colleagues discuss persistent barriers including antigen heterogeneity, impaired antigen presentation, immune exhaustion and an immunosuppressive TME. Meanwhile, “Combinatorial In Situ Cancer Vaccines: Unlocking Broad and Enhanced Antitumor Responses” by Weihsuan Chen and colleagues explores strategies that use the tumor itself as an endogenous antigen source while simultaneously promoting local immune activation.

These approaches represent two complementary directions: one attempts to identify and deliver highly selected tumor-specific antigens, while the other attempts to expose the immune system to the broader antigenic repertoire already present within the tumor.

Individualized mRNA Immunotherapy Reaches Phase 3

One of the clearest examples of the convergence between tumor genomics and immunotherapy is INTerpath-001.

Intismeran autogene is an individualized mRNA-based neoantigen therapy designed according to the unique mutational profile of an individual patient’s tumor. Each individualized construct can encode up to 34 selected neoantigens.

In August 2026, Merck and Moderna announced that the Phase 3 INTerpath-001 trial had met its primary endpoint of recurrence-free survival and the key secondary endpoint of distant metastasis-free survival with intismeran plus pembrolizumab compared with pembrolizumab alone in patients with completely resected stage IIB–IV cutaneous melanoma.

The biological logic of the combination illustrates how several generations of immunotherapy can now intersect. Intismeran is designed to generate T-cell responses against selected patient-specific tumor neoantigens, while pembrolizumab blocks PD-1-mediated inhibitory signaling. One component provides antigenic direction, while the other attempts to prevent checkpoint-mediated suppression of the resulting immune response.

The current Phase 3 announcement remains topline. Detailed hazard ratios, confidence intervals, absolute RFS and DMFS rates, subgroup analyses and the complete safety dataset have not yet been publicly reported, while overall survival remains under evaluation.

Nevertheless, the conceptual significance is considerable. Tumor sequencing is no longer being used only to classify a cancer, identify a molecular driver or select among existing therapies. Molecular information from an individual tumor can increasingly contribute to constructing the therapeutic product itself.

Resistance Becomes a Biology of Its Own

As immunotherapy has matured, resistance has become increasingly difficult to describe as simply “response” versus “no response.”

Primary resistance can occur when the biological conditions necessary for an effective immune response are absent from the beginning. Acquired or adaptive resistance can emerge after an initially successful response as therapeutic pressure reshapes the cancer–immune interaction.

Potential mechanisms include loss of target antigens, alterations in HLA or antigen-presentation machinery, changes in interferon signaling, emergence of alternative inhibitory pathways, progressive T-cell dysfunction, immunosuppressive remodeling of the TME and failure to maintain durable antitumor immune memory.

The 2026 review “Adaptive Resistance in Cancer Immunotherapy” by Ke Yang, Chunqian Yang, Kai Xiong, Jiangtao Hao and colleagues examines clinically distinct patterns of resistance, including progression after an initial response during treatment and relapse following remission.

This distinction has therapeutic consequences. Antigen loss cannot necessarily be solved in the same way as T-cell exclusion; defective antigen presentation requires a different strategy from alternative checkpoint upregulation; and failure of immune memory may require a different intervention from primary absence of tumor-reactive lymphocytes.

The future of combination immunotherapy should therefore become increasingly mechanism-driven—identifying the biological reason for failure before deciding what additional immune intervention to add.

AI and Multi-Omics Enter the Immunotherapy Laboratory

The biological information available from an individual tumor has expanded dramatically. Modern immuno-oncology can interrogate cancer and its immune environment at several complementary levels:

• Single-cell sequencing can define immune-cell populations and functional states.
• Spatial technologies can reveal where those populations are located and how they interact with malignant and stromal cells.
• Genomic and transcriptomic profiling can identify mutations, potential neoantigens, inflammatory programs and mechanisms of immune suppression.
• Digital pathology can quantify tissue architecture and patterns of immune infiltration.
• Longitudinal ctDNA analysis can provide information about molecular disease burden and its evolution during and after treatment.

The challenge is increasingly not simply generating these datasets but integrating them into a biologically meaningful picture.

The 2026 Nature Reviews Cancer article “Advancing AI for Multi-Omics and Clinical Data Integration in Basic and Translational Cancer Research” by Fei Liu, Stephan Beck, Lei Yang, Huiyan Luo and colleagues describes how artificial intelligence is being developed to integrate genomic, transcriptomic, proteomic, imaging and clinical information for patient stratification, response prediction and investigation of treatment resistance.

For immuno-oncology, this is particularly relevant because no single measurement fully represents antitumor immunity. AI has not produced a universal predictor of immunotherapy response, but it provides tools for interrogating biological complexity that increasingly exceeds what PD-L1, TMB or any other individual biomarker can capture.

What the Decade Has Actually Changed

The difference between cancer immunotherapy in 2016 and 2026 cannot be reduced to a longer list of drugs. The deeper change is that modern IO can intervene at increasingly different stages of the cancer–immune interaction:

• Checkpoint inhibitors remove inhibitory immune signaling.
• Cancer vaccines and neoantigen therapies attempt to generate or broaden tumor-specific recognition.
• CAR-T and TCR-T therapies engineer antigen recognition directly.
• TIL therapy amplifies naturally occurring tumor-reactive lymphocyte populations.
• NK-cell platforms introduce different recognition and effector biology.
• Bispecific and multispecific engagers redirect endogenous immune cells toward defined tumor targets.
• TME-directed therapies attempt to dismantle barriers preventing effective immune attack.
• ctDNA and other dynamic biomarkers may increasingly allow treatment to be informed by the evolution of residual disease.
• Single-cell, spatial and multi-omic technologies are providing increasingly detailed maps of the tumor–immune ecosystem.

Yet increasingly sophisticated technology does not eliminate the fundamental biology of antitumor immunity. A therapeutically relevant antigen must still be recognized, an appropriate immune effector must reach the tumor, that cell must retain sufficient function in a hostile microenvironment, and the resulting immune response must be durable enough to prevent immune escape.