How Immune Checkpoint Inhibitors Work and Why Response Is Uneven

Immune checkpoint inhibitors treat cancer by taking a foot off the brake rather than pressing the accelerator. The immune system carries built-in "off" switches, called checkpoints, that keep activated T cells from turning on healthy tissue. Many tumors survive by reaching for those same switches and holding them down. According to the National Cancer Institute, drugs that block the PD-1, PD-L1, or CTLA-4 interactions stop the "off" signal from reaching the T cell, which lets the T cell see the tumor and attack it. That one idea explains the whole clinical picture: the occasional dramatic, long-lasting response these drugs produce, and the large fraction of patients who get nothing. Releasing a brake only helps if there was a capable immune response waiting to move in the first place.

Key points#

• Checkpoint inhibitors remove an inhibitory signal on T cells; they do not create new immunity.
• CTLA-4 blockade and PD-1/PD-L1 blockade act at different times and places, which is why they can be combined and why combinations carry more side effects.
• A tumor must be visible to T cells, reachable by T cells, and not already armored with resistance for the drug to work.
• Biomarkers such as tumor mutational burden, microsatellite instability, and PD-L1 staining enrich for responders but do not guarantee response.

The brake, not the accelerator#

The most common misconception about these drugs is that they rev the immune system up. They do the opposite of what that image suggests. A healthy immune response is designed to stand down once a threat is cleared, and checkpoints are the molecular signals that enforce that. Think of a T cell as a car idling with the brake pressed: the engine is already running, and the checkpoint is the pedal holding it in place. A tumor that displays checkpoint ligands is, in effect, keeping its foot on that pedal. The inhibitor does not add horsepower. It simply lifts the foot.

That framing matters clinically, because it predicts the failure modes. If a car has no engine running, lifting the brake changes nothing. If the road is blocked, lifting the brake changes nothing. The drug can only convert potential energy that already exists into an attack, so every part of this article is really about whether that potential was present.

Two different brakes, in two different places#

Most approved checkpoint inhibitors act on one of two pathways, and the two operate at separate stages of the immune response.

CTLA-4 works early, during T cell priming in the lymph node. As a new T cell is being activated by an antigen-presenting cell, CTLA-4 competes for the B7 molecules on that presenting cell and delivers an inhibitory message that dampens the whole activation step. Blocking CTLA-4 lowers the bar for T cells to switch on at all, so it widens the pool of T cells that enter the fight.

PD-1 and its ligand PD-L1 act later, inside the tumor itself. PD-1 sits on the T cell surface; PD-L1 is the ligand that plugs into it. The NCI notes that some tumors make large amounts of PD-L1, which turns the local T cell response down right where the cancer lives. A peer-reviewed review of PD-1/PD-L1 signaling frames the drug strategy as pharmacologically preventing that PD-1/PD-L1 handshake so anti-tumor immunity can resume. A simple way to keep the two straight in your head: CTLA-4 blockade decides how many T cells show up, while PD-1/PD-L1 blockade decides whether the ones that arrive stay switched on. Because they act at different points, combining them can be synergistic, though it also stacks the risk of immune-related side effects.

Why a released brake does not always win#

If the story ended at "block the brake, win the fight," almost everyone would respond. They do not. Benefit depends on a chain of conditions, and a tumor can break the chain at any link.

First, the tumor has to look foreign#

T cells find cancer by spotting abnormal proteins, called neoantigens, that arise from mutations. A tumor carrying many mutations tends to display more of these flags, giving T cells more to lock onto. That is the reasoning behind tumor mutational burden as a biomarker. In the KEYNOTE-158 analysis, which supported the 2020 approval of pembrolizumab for tumors with high mutational burden (at least 10 mutations per megabase), the objective response rate in that selected group was roughly 29 percent, and most responses lasted a year or longer. A related marker is mismatch-repair deficiency, or high microsatellite instability, which lets mutations pile up. The FDA approval summary for pembrolizumab in mismatch-repair-deficient tumors describes the 2017 accelerated approval as the agency's first tissue-agnostic cancer indication, granted on a 39.6 percent response rate across 15 tumor types. Read those numbers in both directions: far better than unselected tumors would predict, and yet, even in these favorable groups, most patients still did not respond.

Second, T cells have to reach it#

A tumor can display all the right flags and still resist treatment if T cells cannot physically get inside. The review literature calls these "non-inflamed" or "cold" tumors, with sparse CD8-positive T cell infiltration, as opposed to "hot," inflamed tumors full of immune cells. It gets more complicated within a single mass: T cell density can differ from the rim to the core, shaped by low oxygen and by which mutations sit where. Lifting the brake in a region no T cell has entered accomplishes nothing there.

Third, the tumor fights back#

Even a visible, infiltrated tumor can neutralize the drug through resistance, and the literature separates two patterns. In innate resistance, internal signaling pathways such as PI3K-Akt and STAT3 push PD-L1 up on their own, with no immune prompting. In adaptive resistance, the tumor raises PD-L1 specifically in reply to interferon-gamma released by attacking T cells, so the immune assault itself triggers the countermeasure. Resistance can also appear after an early win, as the surviving tumor evolves to slip past a functioning attack.

Where PD-L1 testing fits, and where it fails#

It is natural to assume PD-L1 staining tells you who benefits, but the data are messy. The review is explicit that not every PD-L1-positive tumor responds, and some PD-L1-negative tumors do. Part of the trouble is sampling: PD-L1 expression varies across a tumor, so a single biopsy can misjudge the whole. This is why PD-L1 status informs some treatment decisions without being treated as a clean on-or-off switch. If you are handed one, read it as a probability nudge, not a verdict.

Reading a checkpoint result honestly#

The fair summary is that checkpoint inhibition is a conditional therapy, not a universal one. Its benefit rides on a tumor being mutated enough to be seen, infiltrated enough to be reached, and not already armored with resistance machinery. Tumor mutational burden, microsatellite instability, and PD-L1 expression each capture a slice of that picture, which is why they concentrate responders without promising response, and why no single test yet flags everyone who will benefit. Interpreting a result well means asking which of those conditions a given tumor actually met, rather than filing the whole drug class in your head as reliably effective or reliably useless.