Resistance mechanism atlas
Every cancer drug eventually meets resistance. For 10 major classes, the 38 known escape routes, sorted into eight kinds, how often they occur where that is known, and the countermeasures, linked to the products, targets, and ideas in the map.
10 drug classes · 38 documented escape routes · darker cells mean more routes of that kind. Click a row to jump to it, a column to follow one kind of escape across the page.
- – On-target EGFR C797S
- – MET amplification / bypass
- – Histologic transformation
- – Pre-existing minor resistant clones
- – Solvent-front G1202R and compound mutations
- – Bypass signalling (MET, EGFR, KRAS)
- – Payload resistance: TOP1 mutation or loss
- – SLFN11 loss
- – Efflux pump upregulation (ABCG2, ABCB1)
- – Antigen loss or downregulation
- – Impaired internalisation / lysosomal processing
- – Alternative checkpoints (LAG-3, TIM-3, TIGIT)
- – Loss of neoantigens / low TMB
- – Loss of antigen presentation (B2M, HLA, JAK1/2)
- – Immune-desert / excluded tumours
- – Immunosuppressive myeloid cells and VEGF
- – ESR1 ligand-binding-domain mutations
- – RB1 loss
- – Cyclin E / CDK2 activation
- – PI3K/AKT/mTOR activation
- – Replication fork protection and PARP1 loss
- – BRCA1/2 reversion mutations
- – Restoration of HR via 53BP1/Shieldin loss
- – Drug efflux (ABCB1)
- – Secondary KRAS mutations (Y96D, R68S, H95) and amplification
- – Adaptive RTK feedback (EGFR, others)
- – Bypass alterations (MET amplification, NRAS/BRAF mutations, RTK fusions)
- – BCMA antigen loss (biallelic TNFRSF17 deletion, extracellular mutations)
- – T-cell exhaustion and low fitness
- – Soluble BCMA decoy
- – CD19-negative relapse
- – Immunosuppressive microenvironment and T-cell exhaustion
- – CD19-positive relapse from poor CAR-T persistence
- – AR amplification and ligand-binding-domain mutations
- – AR splice variants (AR-V7)
- – PI3K/AKT activation via PTEN loss
- – Glucocorticoid receptor substitution
- – Lineage plasticity to neuroendocrine prostate cancer
EGFR tyrosine kinase inhibitors (osimertinib)
Lung cancers on osimertinib escape by mutating the drug's binding site, switching on a bypass receptor (MET), or changing cell type entirely.
On-target × 1
The drug's binding site mutates, so the drug no longer fits.Mutation of the cysteine that osimertinib binds covalently; abolishes drug binding while EGFR stays active.
- Fourth-generation allosteric EGFR inhibitors (in trials); amivantamab-based regimens
- ADCs that bypass genotype: Dato-DXd, HER3-DXd, iza-bren
Bypass × 1
Another pathway takes over the job the blocked one was doing.Amplified MET signals to PI3K/MAPK independently of EGFR.
- EGFR×MET bispecific amivantamab; MET TKI + osimertinib combinations
- c-MET-directed and EGFR×c-MET bispecific ADCs
Lineage switch × 1
The cell changes type and no longer depends on the target.Conversion to small-cell lung cancer (RB1/TP53 co-loss) or squamous histology; EGFR mutation persists but the cell no longer depends on it.
- Re-biopsy at progression; platinum-etoposide for SCLC transformation
Other × 1
Escape routes that do not fit the classes above, such as pre-existing minor clones.Small subclones that already carry a resistance route expand under single-agent TKI; adding chemotherapy up front kills them before they take over.
ALK tyrosine kinase inhibitors
Each ALK drug generation was beaten by a new mutation in the kinase; lorlatinib covers nearly all of them, so resistance now runs through other pathways.
On-target × 1
The drug's binding site mutates, so the drug no longer fits.Steric clash blocks first- and second-generation inhibitors; compound mutations (G1202R + L1196M etc.) emerge after lorlatinib.
- Lorlatinib covers G1202R; fourth-generation neladalkib for compound mutations
Bypass × 1
Another pathway takes over the job the blocked one was doing.Alternative receptors or downstream mutations re-activate MAPK/PI3K.
- Combination with MET or MEK inhibitors (trials); chemotherapy; ADCs
Topoisomerase-I payload ADCs (T-DXd, sacituzumab govitecan, Dato-DXd)
Resistance can be to the address (antigen) or to the poison (payload). Payload resistance is shared across every TOP1 ADC regardless of target, which is why a second one often fails.
Payload × 3
The poison stops working: TOP1 loss, SLFN11 silencing, efflux pumps.TOP1 mutations (e.g., E418K) or reduced expression prevent trapping of the cleavage complex.
- Switch payload class (tubulin, DNA-crosslinking, degrader) rather than antigen
- Radioconjugates against the same antigen
Schlafen-11 is required for replication-stress-induced death; its epigenetic silencing confers resistance to TOP1 (and platinum) agents.
SN-38 is an ABCG2 substrate; DXd and MMAE are ABCB1 substrates; mesenchymal states upregulate both.
- Payloads reported to be weaker efflux substrates (sac-TMT's belotecan derivative)
- Efflux-agnostic modalities: radiation, T-cell engagers
Antigen loss × 1
The target disappears from the cell surface.Reduced HER2 or TROP2 surface expression after treatment; less frequent than payload resistance for HER2-low disease.
- Antigen PET to detect loss and pick the next target
- Bispecific ADCs hitting two antigens
Pharmacology × 1
The drug does not reach its site or is cleared: decoys, sanctuary sites, poor persistence, dosing.Defective endocytosis or lysosomal cathepsin activity limits payload release.
- Biparatopic antibodies that force receptor clustering (zanidatamab-type)
PD-1 / PD-L1 checkpoint inhibitors
Most patients never respond (primary resistance) and some responders relapse (acquired). The routes are loss of antigen presentation, no T cells in the tumour, and a suppressive microenvironment.
Bypass × 1
Another pathway takes over the job the blocked one was doing.Exhausted T cells co-express other inhibitory receptors.
- Relatlimab + nivolumab (LAG-3) works; TIGIT combinations failed
Antigen loss × 1
The target disappears from the cell surface.Immunoediting removes the clones that carried immunogenic mutations.
- Vaccines against shared antigens (KRAS)
Immune evasion × 3
Antigen presentation is lost, the tumour is cold, or the microenvironment suppresses T cells.Mutations in B2M or HLA class I stop tumour cells displaying antigen; JAK1/2 loss removes interferon responsiveness (and PD-L1 induction).
- NK-cell and CAR-based approaches that do not need MHC; T-cell engagers
No pre-existing T-cell infiltrate (cold tumour) or T cells held at the margin by TGF-β and stroma.
- Radiation, oncolytic viruses, ADC + IO to prime
- Personalised neoantigen vaccines to supply T cells
- TIL therapy after PD-1 failure
MDSCs, M2 macrophages, and VEGF suppress T-cell function and dendritic-cell maturation.
- PD-1 + VEGF blockade; PD-1×VEGF bispecifics
CDK4/6 inhibitor + endocrine therapy
Hormone-positive breast cancer escapes either by mutating the oestrogen receptor so it no longer needs oestrogen, or by rewiring the cell-cycle engine (RB loss, cyclin E) so CDK4/6 no longer matters.
On-target × 1
The drug's binding site mutates, so the drug no longer fits.Y537S/D538G render ER constitutively active; arise under aromatase-inhibitor pressure, detectable in ctDNA.
- Oral SERDs (elacestrant, imlunestrant, camizestrant) and PROTAC vepdegestrant; ctDNA-guided early switch (SERENA-6)
Bypass × 3
Another pathway takes over the job the blocked one was doing.Without RB, CDK4/6 inhibition cannot arrest the cell cycle.
- Switch to chemotherapy or ADCs (T-DXd for HER2-low, sacituzumab, Dato-DXd)
- CDK2 inhibitors (AVZO-021 and others) and CDK4-selective inhibitors in trials
PIK3CA mutation, PTEN loss, or AKT1 E17K sustain growth independent of ER.
- Capivasertib, inavolisib, alpelisib, everolimus, gedatolisib by genotype
PARP inhibitors
Tumours that lost BRCA can regain repair by re-mutating BRCA back into working order, or by finding another way to protect their DNA.
On-target × 1
The drug's binding site mutates, so the drug no longer fits.Bypass × 2
Another pathway takes over the job the blocked one was doing.Secondary mutations restore the open reading frame and homologous recombination; also confers platinum resistance.
- ctDNA detection of reversions to avoid futile re-challenge; switch to non-DDR agents (ADCs such as mirvetuximab)
Loss of end-protection factors lets BRCA1-deficient cells resect DNA ends and repair by HR.
- ATR inhibitors; POLQ inhibitors (trials)
Pharmacology × 1
The drug does not reach its site or is cleared: decoys, sanctuary sites, poor persistence, dosing.Olaparib and rucaparib are P-gp substrates.
- Talazoparib and niraparib are weaker substrates
KRAS G12C inhibitors
Blocking one RAS mutant makes the cell turn up every upstream receptor and often mutate KRAS again; that is why responses are short and why combinations and pan-RAS drugs followed.
On-target × 1
The drug's binding site mutates, so the drug no longer fits.Alter the switch-II pocket or overwhelm the drug.
- Pan-RAS(ON) tri-complex inhibitors (daraxonrasib) bind a different site
Bypass × 2
Another pathway takes over the job the blocked one was doing.Relief of ERK-mediated negative feedback re-activates receptors within hours, producing new wild-type KRAS-GTP the drug cannot bind.
- Add anti-EGFR antibody in colorectal cancer (CodeBreaK 300, KRYSTAL-1)
- SHP2 or SOS1 inhibitor combinations (trials)
Alternative MAPK activation.
- Combination with MEK/ERK inhibitors; re-biopsy-guided therapy
BCMA-directed therapy (CAR-T, bispecifics, ADC)
Myeloma escapes BCMA drugs by deleting or mutating the target, or by exhausting the T cells that were supposed to do the killing.
Antigen loss × 1
The target disappears from the cell surface.Deletion or mutation removes or alters the epitope; more common after bispecifics than CAR-T.
- Switch to GPRC5D-directed therapy (talquetamab, GPRC5D CAR-T) or FcRH5
Immune evasion × 1
Antigen presentation is lost, the tumour is cold, or the microenvironment suppresses T cells.Prior lines, high tumour burden, and continuous bispecific dosing exhaust T cells; CAR-T products from heavily pretreated patients expand poorly.
- Earlier-line use (CARTITUDE-4, MajesTEC-3); fixed-duration or less frequent bispecific dosing
CD19 CAR-T
Leukaemia and lymphoma relapse after CD19 CAR-T either without CD19 (the target is gone) or with it (the CAR-T cells are gone or exhausted).
Antigen loss × 1
The target disappears from the cell surface.Alternative splicing, mutation, or lineage switch (to myeloid) removes the CD19 epitope.
- CD22 CAR-T, CD20 bispecifics, dual-target CARs
Immune evasion × 1
Antigen presentation is lost, the tumour is cold, or the microenvironment suppresses T cells.PD-1 upregulation, TGF-β, and myeloid suppression in lymphoma.
- PD-1 knockout or blockade with CAR-T (trials)
Pharmacology × 1
The drug does not reach its site or is cleared: decoys, sanctuary sites, poor persistence, dosing.Limited expansion or early loss of CAR-T cells; 4-1BB products persist longer than CD28.
- Allogeneic or in vivo re-dosing; armoured CARs
Androgen receptor pathway inhibitors (abiraterone, enzalutamide)
Prostate cancer keeps the androgen receptor working without hormones (amplification, splice variants), or abandons it and becomes a neuroendocrine cancer.
On-target × 2
The drug's binding site mutates, so the drug no longer fits.More receptor, or mutations (F877L, T878A) that turn antagonists into agonists.
- AR degraders and N-terminal-domain inhibitors (trials); PSMA radioligand therapy
Truncated receptor lacking the ligand-binding domain is constitutively active and invisible to enzalutamide.
- Taxanes retain activity; N-terminal-domain inhibitors
Bypass × 2
Another pathway takes over the job the blocked one was doing.Reciprocal feedback between AR and PI3K pathways.
- Capivasertib + abiraterone (approved 2026 for PTEN-deficient disease)
GR drives an AR-like transcriptional programme under enzalutamide.
- GR antagonists (relacorilant tested in prostate cancer)
Lineage switch × 1
The cell changes type and no longer depends on the target.RB1/TP53 loss enables transdifferentiation; AR-indifferent, DLL3-positive, PSMA-negative.
- Platinum-etoposide; DLL3 engagers (tarlatamab) and B7-H3 ADCs in trials