What checkpoint inhibitors change
Ask about this lessonA checkpoint inhibitor removes one brake from T cells that already recognize the cancer. In early triple-negative breast cancer (TNBC), adding pembrolizumab to chemotherapy raised the complete-response rate and improved survival, whether or not the tumor stained for PD-L1.
Before you start:
- Immune checkpoint: a normal inhibitory signal that limits T-cell activity.
- Immune checkpoint inhibitor: a drug that blocks one such signal.
- T-cell exhaustion: the restrained program T cells enter under persistent antigen.
- Prognostic vs predictive: tracking outcome versus tracking benefit from one treatment.
Where this step sits
The earlier lessons followed a response from display through priming, cell states, location, measurement and escape. This step asks what happens when a drug changes one of those steps.
The drug acts at the last contact. It cannot create the earlier steps: a displayed target, a matching T cell and access to the tumor.
The brake that the tumor borrows
Programmed cell death protein 1 (PD-1) is a receptor on activated and exhausted T cells. When it binds programmed death-ligand 1 (PD-L1), the T cell's activating signals are damped. In healthy tissue this protects organs from prolonged inflammation.
Tumors often display PD-L1 because T cells are attacking them. T cells that recognize a tumor release interferon gamma, and interferon gamma induces PD-L1 on nearby tumor and immune cells. This is called adaptive immune resistance: the brake is pulled in response to the attack (Taube et al., 2012). That is why PD-L1 staining and a lymphocyte-rich "hot" pattern often travel together.
Pembrolizumab is an antibody that binds PD-1 and blocks its contact with PD-L1. Cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) is a different checkpoint whose role is emphasized during priming in lymph nodes; the checkpoint concept compares them.
What blockade can and cannot do
| Step in the route | Can PD-1 blockade change it? | Why |
|---|---|---|
| A tumor peptide is displayed on human leukocyte antigen (HLA) | No | Lost antigen, HLA or beta-2 microglobulin (B2M) is an escape route the drug does not repair |
| A matching T cell exists and is primed | Indirectly | PD-1 blockade may let more primed cells expand; it does not create a new receptor |
| T cells reach cancer-cell nests | Not directly | Exclusion by stroma or missing chemokines acts upstream of PD-1 |
| T cells in the tumor keep working | Yes | This is the main target: restrained, PD-1-positive cells can act again |
| The response lasts | Partly | Exhausted cells keep a stable chromatin pattern and can relapse into exhaustion (Pauken et al., 2016) |
Where do the responding cells come from? In mice, mainly from progenitor exhausted cells that divide after blockade. In paired human skin-cancer biopsies, many expanded clones were new arrivals (Yost et al., 2019). Both may contribute; see T-cell exhaustion.
What the trials showed in TNBC
| Trial | Population | Comparison | Result |
|---|---|---|---|
| KEYNOTE-522 | 1,174 patients with untreated stage II or III TNBC, enrolled regardless of PD-L1 | Pembrolizumab or placebo with chemotherapy before surgery, then pembrolizumab or placebo for up to nine cycles after | Pathological complete response (pCR) 64.8% vs 51.2% in the first 602 patients (Schmid et al., 2020); 36-month event-free survival 84.5% vs 76.8%, hazard ratio (HR) 0.63 (Schmid et al., 2022); 60-month overall survival 86.6% vs 81.7% (Schmid et al., 2024) |
| KEYNOTE-355 | 847 patients with untreated advanced (metastatic or inoperable) TNBC | Pembrolizumab or placebo with chemotherapy | Median overall survival 23.0 vs 16.1 months (HR 0.73) when the combined positive score (CPS) was ≥10; not significant when CPS was ≥1 (HR 0.86) (Cortés et al., 2022) |
Two details are worth carrying. First, the 64.8% pCR figure comes from the first interim analysis; across all 1,174 patients the US label reports 63.0% vs 55.6% (FDA label, section 14.20). Second, KEYNOTE-522 was not designed to separate the effect of pembrolizumab before surgery from its effect after. The KEYNOTE-522 regimen page gives the full schedule.
Which markers select patients
- PD-L1 is predictive in advanced TNBC. In KEYNOTE-355, the benefit was concentrated in tumors with CPS ≥10 by the 22C3 assay, and the US label requires that cutoff for advanced disease. See PD-L1 scoring for why a different assay's "positive" is not the same.
- PD-L1 does not select patients in early TNBC. KEYNOTE-522 enrolled regardless of PD-L1, and the early-disease label has no PD-L1 requirement.
- sTILs are prognostic. In 2,148 chemotherapy-treated patients with early TNBC, each 10% increase in stromal tumor-infiltrating lymphocytes (sTILs) was associated with an overall survival HR of 0.84 (Loi et al., 2019). That does not show that sTILs pick out who gains from pembrolizumab; see stromal TIL scoring.
The cost: immune-related adverse events
Releasing a brake that protects healthy tissue can cause inflammation in any organ. These immune-related adverse events (irAEs) usually appear during treatment but can start after it ends. Across 2,799 patients given pembrolizumab alone in trials, hypothyroidism occurred in 8% and adrenal insufficiency in 0.8%. In KEYNOTE-522, pembrolizumab was stopped for adverse reactions in 20% of patients, and the label asks for blood cortisol checks before surgery in this setting (FDA label, sections 5.1 and 6.1).
What can go wrong at this step
- Treating PD-L1 as a universal gate. Its role depends on the setting, assay and cutoff.
- Reading "hot" as "will respond". A lymphocyte-rich tumor is prognostically favorable, but escape by HLA or B2M loss can still block the drug's route.
- Reading a pCR gain as a cure rate. pCR is a surgical finding; survival is a separate endpoint, which KEYNOTE-522 also measured.
Try it
A friend says: "My tumor was PD-L1 negative, so pembrolizumab won't help me in early TNBC." What does the evidence say?
Answer: In early TNBC, KEYNOTE-522 enrolled patients regardless of PD-L1, and the approved early-disease use has no PD-L1 requirement. The CPS ≥10 cutoff applies to advanced TNBC, where KEYNOTE-355 found the benefit concentrated in that group. Individual benefit still cannot be read from one marker.
Explain it back
"Anti-PD-1 drugs act on ___; they cannot supply ___. PD-L1 selects patients in ___ TNBC but not in ___ TNBC." One answer: "T cells that already recognize the tumor but are restrained; a missing target, HLA or access; advanced; early."
Takeaway
Checkpoint blockade lifts one late brake; in early TNBC it added to chemotherapy's effect regardless of PD-L1, and it can inflame healthy organs.
Next: step 8, compare immune-cell recognition systems, or return to the guide.
Sources and scope
Source-checked November 7, 2026, against PubMed abstracts and the FDA pembrolizumab label (supplement 194). Population-level trial results; no individual treatment recommendation. Expert and learner review pending.
- Schmid et al., 2020: KEYNOTE-522 pCR; 2022: event-free survival; 2024: overall survival.
- Cortés et al., 2022: KEYNOTE-355 overall survival.
- FDA: pembrolizumab label, supplement 194 (2026), sections 1.20, 5.1, 6.1 and 14.20.
- Taube et al., 2012: adaptive immune resistance.
- Loi et al., 2019: sTIL prognosis in early TNBC.
- Yost et al., 2019 and Pauken et al., 2016: where responding cells come from, and why reinvigoration fades.