Paul Ehrlich gestured at immunological tumor control in 1909. Lewis Thomas and Macfarlane Burnet formalized it as immunosurveillance in the 1950s and 1970s. The claim: the immune system continuously inspects the body's own cells, identifies transformed ones, and destroys them before they become detectable tumors. The easy objection was that cancer is common — if surveillance worked, why did so many people get it? What shifted the field was mice engineered without functional T and B cells. Remove the watchers, and the cancers appear — more frequent, more aggressive. Each component of the immune killing machinery, when disabled, makes tumors worse.

The modern refinement is cancer immunoediting, described by Dunn and Schreiber in 2002. It reframes the immune system's relationship to a developing cancer not as binary success or failure but as a dynamic three-phase process. Elimination is classical surveillance: most transformed cells are destroyed before forming a tumor. Equilibrium is what the transplant case made visible: survivors held in check, dormant, not expanding — the donor's melanoma cells persisted this way for fifteen years. Escape is what we observe when we diagnose cancer: a variant acquires changes that evade immune pressure entirely, and the tumor becomes clinically visible.

The word editing is the insight. The immune system is not just failing when cancer escapes — it is, in the process of eliminating the most visible cells, selecting for the less visible ones. Every cancer cell the immune system successfully kills is one that expressed something recognizable. The cells that survive are, by selection, the ones that were harder to see, harder to kill, or better at suppressing the response. A cancer diagnosis is the endpoint of an evolutionary contest. You are seeing the winner. And here is the disquieting corollary: a more effective immune response applies more selection pressure — and may produce, in the survivors, a more aggressively evasive tumor.

For the immune system to recognize a cancer cell, the cell must display something that marks it as different. A tumor antigen is any molecule a cancer cell presents — usually as a peptide fragment on MHC — that an immune cell can recognize. The distinction that matters practically is whether that molecule exists anywhere in normal tissue. Tumor-specific antigens exist only on cancer cells — viral antigens like HPV E6 and E7, and neoantigens from somatic mutations. Tumor-associated antigens are expressed on cancer cells but also on normal tissue. The practical question to ask of any antigen is simple: what normal tissue shares this target, and what happens when the immune response hits it? That answer is your toxicity prediction.

During development in the thymus, T cells whose receptors bind strongly to the body's own peptides are deleted. This central tolerance is how a healthy immune system avoids attacking normal tissue. Tumor-associated antigens are self-proteins — tolerance to them is partial, but the activation threshold is raised. Neoantigens are different. A peptide arising from a somatic mutation that occurred after thymic education is a peptide that was never evaluated during T-cell selection. T cells capable of recognizing it survived thymic selection intact, with no tolerance imposed. This is why neoantigens are so attractive: they can be targeted without first overcoming the brakes the body deliberately placed. The number of neoantigens scales with tumor mutational burden — more mutations, more chances for novel peptides, more potential T-cell recognition.

Chen and Mellman drew the cancer-immunity cycle in 2013, and it is the most useful framework for understanding both how the immune response works and where it breaks. The cycle opens when cancer cells die and release antigens. Dendritic cells take up the debris, load peptides onto MHC, migrate to a lymph node, and present to naive T cells. Activation requires two signals in parallel: the T-cell receptor binding the antigen-MHC complex, and co-stimulation through CD28. A T cell that receives signal one without signal two becomes anergic — functionally silenced. Activated T cells expand enormously, follow chemokine gradients to the tumor, and kill through perforin-granzyme and death-receptor ligation. The chain is robust when intact, but each link is a place a tumor can sever it.

Each link in the cancer-immunity cycle is a place a tumor can sever the chain. Defective antigen processing breaks step two — no priming signal. Failure to mature dendritic cells breaks step three — co-stimulation absent, T cells become anergic. Absent chemokine expression and dense matrix block trafficking in steps four and five. Checkpoint ligands suppress the cytotoxic T cell in step six even after successful priming. The immunotherapy revolution of the last decade — checkpoint blockade — worked by releasing step six, removing an inhibitory signal on a T cell that was already present and primed but blocked from firing. The tumors that respond best are ones where the only broken link was the checkpoint brake.

So here is the chapter's claim: immunosurveillance is real — the transplant case, the knockout mice, the epidemiology of immunosuppression all confirm it. But the simple surveillance hypothesis was too simple. The modern answer is immunoediting: elimination, equilibrium, escape. The tumor that finally breaks out has been sculpted by immune pressure toward invisibility. A cancer diagnosis is the endpoint of an evolutionary contest, and you are seeing the winner. Understanding what makes cells recognizable — and how the cancer-immunity cycle fails — is what immunotherapy is built on. The questions still open are why some excluded tumors resist even combination strategies, and what drives T-cell exhaustion beyond checkpoint ligands alone.