Two patients. The same cancer. The same anti-PD-1 antibody.
One develops a deep response that lasts for years.
The other progresses within months.
This striking difference has followed immune checkpoint inhibitors since their introduction and remains one of the central unanswered questions in immuno-oncology.
The explanation is often reduced to PD-L1 expression, tumor mutational burden (TMB), or whether a tumor is immunologically “hot” or “cold.” But none of these variables alone explains why checkpoint blockade can be transformative for one patient and ineffective for another.
The biology is considerably more interesting.
Checkpoint inhibitors do not create antitumor immunity from nothing. They remove specific inhibitory signals from an immune response whose other components must already exist or be capable of developing.
That means the real question is not simply whether PD-1 is blocked.
It is whether the entire biological chain required for tumor immunity is sufficiently intact for releasing PD-1 to matter.
Can 3D Ex Vivo Lymphoma Spheroids Predict Immunotherapy Response?
The First Misconception: Checkpoint Inhibitors Do Not Simply “Boost” Immunity
PD-1 and CTLA-4 are not abnormal cancer pathways. They are physiological regulators of immunity.
Without inhibitory checkpoints, activated lymphocytes could damage normal tissues and produce uncontrolled inflammation and autoimmunity. Cancer exploits these protective mechanisms.
PD-1 engagement by PD-L1 or PD-L2 can attenuate signaling in activated T cells, reducing proliferation, cytokine production and cytotoxic function. CTLA-4 regulates T-cell activation through a different biological context, particularly during priming and competition for costimulatory signals.
Blocking these pathways can restore antitumor activity, but only when checkpoint signaling represents an important barrier to an immune response that has something meaningful to attack.
A useful way to think about checkpoint blockade is therefore not as turning the immune system on, but as removing one constraint from a complex biological system.
This distinction is becoming even more important as checkpoint biology itself expands. The 2026 Nature Reviews Cancerarticle “Regulation of Immune Checkpoint Molecules in Cancer Immune Evasion and Therapy” by Cansu Eris, Cheng Zu, Michael Platten, Chong Sun and colleagues describes a much broader network of checkpoint regulation involving not only receptor expression but transcriptional, epigenetic, post-transcriptional and metabolic mechanisms. Modern checkpoint biology is therefore considerably more complex than a simple PD-1/PD-L1 interaction.
Before You Release the Brake, There Has to Be an Engine
For an anti-PD-1 antibody to produce tumor regression, several biological events must align.
A cancer must generate antigens that can distinguish malignant cells from normal tissue. Those antigens must be processed and presented. Dendritic cells must capture antigen and support effective T-cell priming. Appropriate tumor-reactive lymphocytes must expand, enter the circulation, traffic into tumor tissue, recognize malignant cells and retain enough functional capacity to kill them.
Only then does releasing an inhibitory checkpoint become meaningful.
This framework was elegantly described by Daniel Chen and Ira Mellman in their landmark 2013 Immunity article “Oncology Meets Immunology: The Cancer-Immunity Cycle.” The cancer-immunity cycle connected antigen release, presentation, T-cell priming, trafficking, infiltration, recognition and tumor-cell killing into a continuous biological process.
Checkpoint blockade addresses only part of that cycle.
A patient can therefore fail anti-PD-1 therapy even when PD-1 is successfully blocked because the dominant defect may lie somewhere else entirely.
More Mutations, More Targets? The Genomic Clue
One of the earliest major clues came from tumor genomics.
Cancer cells accumulate somatic mutations, some of which produce altered proteins. If mutation-derived peptides are expressed, processed, presented by HLA molecules and recognized by T cells, they can function as neoantigens.
This led to an intuitive hypothesis: perhaps tumors carrying more mutations provide the immune system with more opportunities for recognition.
In the landmark New England Journal of Medicine study “Genetic Basis for Clinical Response to CTLA-4 Blockade in Melanoma,” Alexandra Snyder, Timothy Chan, Jedd Wolchok and colleagues used whole-exome sequencing to examine melanomas treated with CTLA-4 blockade. Higher mutational burden was associated with clinical benefit, although mutation number alone was not sufficient to determine response.
Soon afterward, Naiyer Rizvi, Matthew Hellmann, Timothy Chan and colleagues reported “Mutational Landscape Determines Sensitivity to PD-1 Blockade in Non-Small Cell Lung Cancer” in Science. Higher nonsynonymous mutational burden was associated with improved response, durable clinical benefit and progression-free survival with pembrolizumab.
These studies helped establish the biological foundation for TMB as an immunotherapy biomarker.
But they also exposed one of its major limitations.
A mutation is not automatically an immune target.
A mutation must be expressed. The resulting protein must generate an appropriate peptide. That peptide must bind an HLA molecule. The peptide–HLA complex must reach the tumor-cell surface. A compatible T-cell receptor must exist. And the corresponding T cell must remain capable of attacking the cancer.
TMB therefore describes the potential antigenic opportunity of a tumor much better than it describes the actual quality of its antitumor immune response.
ZNF683+ CD8+ T Cells and SPP1 Signaling May Define Immunotherapy Response in NSCLC
MSI/dMMR: When Genomic Instability Becomes Immunologically Visible
Mismatch repair-deficient cancers provided one of the strongest demonstrations that genomic biology can predict sensitivity to immunotherapy.
In “Mismatch Repair Deficiency Predicts Response of Solid Tumors to PD-1 Blockade,” Dung T. Le, Luis Diaz, Drew Pardoll and colleagues demonstrated substantial activity of PD-1 blockade across mismatch repair-deficient tumors originating in different organs.
The importance of this work went far beyond one biomarker.
It showed that cancers traditionally classified as different diseases could share a common immunological vulnerability created by their genomic biology.
But even MSI-high/dMMR tumors do not universally respond.
Why?
Because generating abnormal proteins is still only the beginning of the immune response.
A Tumor Can Carry the Perfect Antigen—and Still Hide It
Imagine that a tumor contains an exceptionally immunogenic mutation.
That should make it vulnerable.
Unless the cancer stops showing that mutation to the immune system.
CD8+ T cells generally recognize intracellular antigens as peptides presented by HLA class I molecules. Defects involving HLA expression, β2-microglobulin or other components of antigen processing and presentation can therefore allow malignant cells to become functionally invisible to otherwise capable T cells.
This creates one of the most important distinctions in cancer immunology:
Having an antigen is not the same as presenting an antigen.
Checkpoint blockade cannot restore recognition if the malignant cell is no longer adequately displaying what the T cell needs to see.
This is also one mechanism through which initially sensitive cancers can later escape immune pressure.
The Number of Immune Cells Is Not the Whole Story Either
For years, much of immuno-oncology focused on the presence or absence of tumor-infiltrating lymphocytes.
That remains important—but the field is now asking a much more sophisticated question:
Which immune cells are present, where are they located, what state are they in, and who are their neighbors?
A particularly interesting 2026 example comes from Nature Cancer.
In “Temporal and Spatial Composition of the Tumor Microenvironment Predicts Response to Immune Checkpoint Inhibition in Metastatic TNBC,” Noah Greenwald and colleagues analyzed longitudinal specimens from 103 patientsin the phase II TONIC trial using highly multiplexed imaging.
The study evaluated 37 proteins across 270 tumors and extracted more than 800 spatial features per sample.
One of the most striking findings was that features from metastatic lesions, rather than primary tumors, were informative for outcome. Spatial characteristics including immune diversity and T-cell infiltration at tumor borders, together with T-cell-to-cancer-cell ratios and PD-L1 expression on myeloid cells, were associated with response. Multivariable models reached an AUC of 0.90 in stratifying treatment outcome.
This adds an important dimension to the checkpoint-response problem.
It may not be enough to ask: “How many CD8+ T cells are in the tumor?”
We may need to ask: “Where exactly are those T cells, which cells are they interacting with, and what functional state are they in?”
The Tumor Is Not Just Cancer Cells Plus T Cells
Another major change in modern immuno-oncology is the recognition that checkpoint response cannot be understood by studying malignant cells and CD8+ T cells alone.
The tumor microenvironment contains dendritic cells, macrophages, monocytes, neutrophils, B cells, regulatory T cells, fibroblasts, endothelial cells and multiple other populations communicating through cytokines, metabolites and direct cell–cell interactions.
The 2025 Nature Cancer review “The Tumor Microenvironment’s Role in the Response to Immune Checkpoint Blockade” by Konstantinos Aliazis, Anthos Christofides, Sizun Jiang, Vassiliki Boussiotis and colleagues emphasizes exactly this broader framework.
Their analysis highlights how myeloid and non-immune stromal populations can regulate adaptive immunity and influence both sensitivity and resistance to checkpoint blockade. It also reflects an important evolution in the field: resistance can no longer be explained only through T-cell exhaustion.
A tumor can contain potentially effective T cells yet surround them with suppressive macrophages.
It can recruit lymphocytes but maintain abnormal vasculature that restricts their access to malignant-cell nests.
Cancer-associated fibroblasts can shape immune exclusion.
Hypoxia, lactate, adenosine and nutrient competition can alter immune-cell metabolism.
TGF-β and other suppressive pathways can reshape the entire immune architecture.
In such a tumor, PD-1 may be only one layer of suppression among many.
And Then There Are the Immune Neighborhoods
One of the most fascinating developments in recent cancer immunology is the increasing importance of tertiary lymphoid structures, or TLS.
These are organized lymphoid aggregates that develop within or around tumors and can contain B-cell follicles, T cells, dendritic cells and other components capable of supporting local adaptive immune responses.
Their importance has challenged the older T-cell-centric view of checkpoint response.
In the landmark Nature paper “Tertiary Lymphoid Structures Improve Immunotherapy and Survival in Melanoma,”Rita Cabrita and colleagues found that the co-occurrence of CD8+ T cells and CD20+ B cells—and the presence of TLS-related biology—was associated with improved clinical outcomes in melanoma treated with checkpoint blockade.
Subsequent work has strengthened the association between mature TLS and immunotherapy outcomes. A Nature Cancerstudy, “Mature Tertiary Lymphoid Structures Predict Immune Checkpoint Inhibitor Efficacy in Solid Tumors Independently of PD-L1 Expression,” found that mature TLS were associated with improved objective response, progression-free survival and overall survival independently of PD-L1 status and CD8+ T-cell density.
And the biology continues to evolve.
The 2026 Nature Cancer perspective “The Future of Tertiary Lymphoid Structures in Cancer Immunotherapy as Biomarkers and Therapeutic Targets” by Hye Mi Kim, Tejashree Joglekar, Tina Cascone, Tullia Bruno and colleagues goes a step further: TLS are being investigated not merely as biomarkers of an immune-responsive tumor, but as structures that might potentially be therapeutically induced or manipulated to convert immune-excluded tumors into more immune-infiltrated states.
That is a remarkable evolution.
A few years ago, the question was whether lymphocytes were present.
Now we are asking whether tumors contain organized immune communities capable of coordinating local antitumor immunity.
Why “Exhausted T Cell” Is Also Too Simple
Even when two tumors contain similar numbers of CD8+ T cells, their responses to checkpoint blockade can be dramatically different.
One reason is that tumor-infiltrating T cells are not biologically equivalent.
Chronic antigen stimulation can drive T cells into dysfunctional or exhausted states, but exhaustion is not one irreversible endpoint. Single-cell and lineage-tracing studies have identified different populations along the exhaustion continuum, including progenitor or stem-like populations with greater proliferative potential and more terminally exhausted populations with limited capacity for reinvigoration.
This matters for PD-1 blockade.
The therapeutic effect may depend not simply on whether PD-1-positive T cells exist, but on whether the tumor contains T-cell populations still capable of proliferating and generating effective descendants after checkpoint release.
Recent TLS research adds another layer to this story. The 2025 Nature Communications paper “Mature Tertiary Lymphoid Structures Evoke Intra-Tumoral T and B Cell Responses via Progenitor Exhausted CD4+ T Cells in Head and Neck Cancer” used single-cell RNA sequencing, antigen-receptor sequencing and spatial transcriptomics to characterize immune organization within TLS. Mature TLS were enriched for stem-like T-cell populations and diverse B-cell maturation states, illustrating how immune-cell state and spatial organization can intersect within the same tumor.
The question is therefore no longer merely:
Are T cells present?
It is: Are the right T cells present, in the right state, in the right place, surrounded by the right immune partners?
Immunophenotypic Signatures Predict Immunotherapy Response and Survival in NSCLC
The Tumor Is Moving While We Are Trying to Measure It
There is another reason checkpoint response is difficult to predict.
Cancer is not static.
Neither is immunity.
One biopsy represents one location at one moment in an evolving disease.
Treatment itself changes that disease.
Immune-sensitive tumor clones can be eliminated, leaving resistant populations behind. Antigens can disappear. HLA expression can change. Interferon signaling can be altered. Alternative inhibitory pathways can emerge. Stromal and myeloid populations can remodel under therapeutic pressure.
This evolutionary framework is increasingly central to contemporary resistance research. The 2026 review “Adaptive Resistance in Cancer Immunotherapy” by Ke Yang, Chunqian Yang, Kai Xiong, Jiangtao Hao and colleagues distinguishes different clinical patterns of immune resistance, including progression during treatment following an initial response and relapse after remission when durable antitumor immune memory fails to persist.
This helps explain why a biomarker measured before the first dose may not describe what is happening months later.
A 2026 Nature Reviews Clinical Oncology review, “Towards Liquid Biopsy-Based Analysis of Antitumour Immunity,” highlights the growing interest in using circulating tumor DNA, circulating immune cells, plasma proteins and other blood-based measurements to follow tumor–immune interactions longitudinally and potentially identify response or emerging resistance before conventional clinical progression becomes apparent.
The future may therefore require us to stop treating immunotherapy response as a fixed pretreatment characteristic.
It may be something we need to measure repeatedly while it is happening.
So What Separates an Exceptional Responder From a Non-Responder?
There is no single answer.
A favorable checkpoint response is more likely when several layers of biology align:
- the tumor generates sufficiently immunogenic antigens;
- those antigens are efficiently processed and presented;
- dendritic-cell priming is functional;
- tumor-reactive T-cell clones exist;
- those cells can traffic into the tumor;
- the immune architecture supports productive cellular interactions;
- T cells remain in a state capable of responding to checkpoint release;
- suppressive myeloid, stromal and metabolic mechanisms do not dominate;
- and the blocked checkpoint is actually an important limiting pathway in that particular tumor.
This explains why PD-L1, TMB, MSI/dMMR, T-cell infiltration, gene-expression signatures, TLS and other biomarkers can all contain useful information without any one of them perfectly predicting response.
Each measures a different piece of the same biological system.
