Cancer as Evolution: The Pressure

Every effective cancer treatment is a selective pressure. This is a consequence of what treatment does: kill some cells and spare others. The cells that survive are the ones whose genetic or transcriptional makeup happened to make them less vulnerable to the selective pressure of the chosen treatment. When those survivors divide, they pass the makeup that allowed them to survive treatment to their offspring. As a result, the population shifts; the cancer that exists after treatment differs from the cancer that existed before, and the treatment itself drove the difference.

Our previous post established that tumors are evolving populations. This post asks what happens to those populations when we intervene.

The clearest illustration of therapeutic evolution comes from EGFR-mutant non-small cell lung cancer, where three generations of drugs and three generations of resistance have played out over two decades in full molecular detail.

In 2004, researchers identified that a subset of lung adenocarcinomas carry activating mutations in the epidermal growth factor receptor (EGFR) gene, most commonly small deletions in exon 19 or a point mutation in exon 21 (L858R). These mutations lock the receptor in an active state, driving continuous cell proliferation. First-generation EGFR inhibitors, erlotinib and gefitinib, block this receptor and produce dramatic responses. Tumors shrink and patients improve. Some respond for a year or more, but the responses do not last.

In roughly 60% of patients who progress on first-generation EGFR inhibitors, sequencing of the resistant tumor reveals a single additional mutation: T790M, a point mutation in the EGFR gene that changes the shape of the drug-binding pocket so the drug no longer fits. This mutation rarely appears in pre-treatment biopsies at standard sequencing depth. Under the selective pressure of erlotinib or gefitinib, the rare cells carrying T790M gain an enormous fitness advantage, survive, expand, and subsequently become the dominant population.

Our pharmaceutical response to this was osimertinib, a third-generation EGFR inhibitor designed specifically to overcome T790M resistance. Osimertinib binds the receptor through a different mechanism and inhibits both the original activating mutation and the T790M gatekeeper. Patients who progressed on first-generation drugs respond again, and osimertinib has since moved to the front line as the standard first-line treatment for EGFR-mutant lung cancer.

Resistance continued, however. Tumors treated with osimertinib develop escape mechanisms of their own. Some cancers develop a C797S mutation that alter the cysteine residue that osimertinib bonds to, blocking the drug’s binding site entirely. Other resistance paths bypass EGFR altogether: amplification of MET, activation of alternative receptor tyrosine kinases, histological transformation from adenocarcinoma to small cell lung cancer (a wholesale change in cell identity that renders EGFR-directed therapy irrelevant).

Each drug creates a new selective pressure, and each pressure selects for a new escape. The tumor population explores its available evolutionary options and finds one that works. What looks like drug failure is evolution performing exactly as predicted.

The EGFR story illustrates a principle with broad implications: resistance often pre-exists treatment. The drug does not cause the resistant mutation. The drug changes the fitness landscape so that a mutation already present in the population, previously neutral or even slightly costly, suddenly confers a large survival advantage.

Researchers have tested this directly. Using ultra-sensitive sequencing methods capable of detecting mutations present in fewer than one in a thousand cells, several groups have identified T790M mutations in EGFR-mutant tumors before any treatment with erlotinib or gefitinib. The resistant clone was already there, a tiny minority within the larger population, and treatment selected for it rather than creating it.

This reframes the clinical problem. If resistance is pre-existing, then the question is not whether a tumor can evolve resistance to a given drug. Given sufficient genetic diversity, it almost certainly already harbors a resistant clone. The question is how large that clone is, how fast it will expand under treatment, and whether the therapeutic strategy accounts for its existence.

Checkpoint immunotherapy introduced a fundamentally different kind of selective pressure. Rather than targeting cancer cells directly, anti-PD-1 and anti-CTLA-4 antibodies release the brakes on the patient’s own immune system, enabling T cells to recognize and kill tumor cells. The selective pressure comes from the immune system itself, and the evolutionary response reflects that.

Tumors that progress on checkpoint immunotherapy frequently lose the molecular machinery required for immune recognition. The most common route is loss of beta-2-microglobulin (B2M), a protein essential for assembling MHC class I molecules on the cell surface. Without MHC class I, the cell cannot present tumor antigens to cytotoxic T cells, and the cancer cell becomes immunologically invisible.

Other tumors lose or downregulate specific HLA alleles, narrowing the range of antigens they present. Some acquire mutations in the JAK1/JAK2 signaling pathway, which mediates the cell’s response to interferon gamma, a key immune signaling molecule. Without functional JAK signaling, the cancer cell fails to upregulate MHC expression even when immune cells are actively signaling it to do so.

The pattern is consistent with the evolutionary framework. The immune system creates a selective pressure favoring cancer cells that evade recognition. Cells that lose antigen presentation gain a fitness advantage in an immune-active environment, survive, expand, and eventually dominate.

What makes immune evasion particularly challenging is that it eliminates the mechanism of action, not just the drug target. In targeted therapy resistance, the drug loses its effect on the cancer cell, but the cancer cell itself remains potentially targetable by other means. In immune evasion, the tumor has dismantled the interface between itself and the immune system. The T cells may still be present, active, and capable, but the target has vanished.

The evolutionary perspective reframes treatment decisions in a specific and uncomfortable way. Every drug administered to a cancer patient reshapes the tumor’s fitness landscape. The cells that survive are, by definition, the ones best adapted to the selective pressure imposed. The cancer that emerges on the other side has been shaped by the clinician’s choices.

Treatment is not futile. Many patients are cured. Many more gain years of life. The drugs work. But they work within an evolutionary system, and that system has consequences.

The order in which drugs are given carries evolutionary consequences. Giving drug A first and drug B second may produce a different evolutionary trajectory than B-then-A. A resistance mutation selected by the first drug may confer cross-resistance to the second, foreclosing an option that would have remained open under a different sequence. Or it may sensitize the tumor to the second drug, creating a vulnerability that the reverse order would not have produced. The optimal sequence depends on the evolutionary paths available to the tumor, and those paths depend on the genetic diversity present at the time of treatment.

As our next post will explore in detail, treatment intensity also shapes the evolutionary outcome. Maximum-dose strategies aim to kill as many cancer cells as possible, as fast as possible, following the logic that fewer surviving cells mean fewer chances for resistance. The evolutionary counterargument is that aggressive eradication can remove the competitive constraints that keep resistant subclones in check.

Timing compounds these dynamics. The same drug given when a resistant clone is vanishingly rare may succeed where it would fail if given after that clone has expanded to a significant fraction of the population.

None of these considerations emerge from thinking of cancer as a static target. They become visible only when we recognize cancer as an evolving population responding to selective pressures in real time. Treatment shapes cancer’s future. The question is whether we can learn to shape it deliberately.

Next in the series: The Ecosystem

The Cancer as Evolution series is published by Galen Health, where we are building cancer superintelligence: autonomous, self-directed AI systems for cancer research and discovery.