Oncofusion

start / graph / mechanism-constitutive-kinase-activation

Mechanismestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q11-comparative

Mechanism class 3 — constitutive kinase activation (BCR::ABL1, NPM1::ALK, ETV6::NTRK3)

also called kinase fusion class · oligomerisation-activated kinase · enzyme active-site class

In these fusions an enzyme that should switch on only on demand gets stuck on, because the partner glues copies of it together. Enzymes have a real chemical slot a pill can plug, which is why this class produces drugs - but the pill has to be given early and alongside chemotherapy, because on its own, late in the disease, it does very little.

The same thing, for a scientist

The 3-prime partner contributes an intact tyrosine kinase catalytic domain (ABL1, ALK, NTRK3, ROS1, RET, PDGFRB) and the 5-prime partner contributes a constitutive oligomerisation element (BCR coiled-coil, NPM1 oligomerisation domain, ETV6 SAM/PNT polymerisation domain) that removes the normal ligand and autoinhibition requirement. The fusion signals continuously through RAS-MAPK, PI3K-AKT and STAT5. Uniquely among fusion classes, the oncogenic activity resides in a deep, evolutionarily conserved, ATP-binding cleft, so classical structure-based small-molecule design applies directly and has produced multiple approved drugs. **Two qualifiers belong in the class definition.** First, the flagship single-agent results are results about INDOLENT disease: the same drug against the same fusion gave median survival of 6.9 months in CML myeloid blast crisis and 4.9 months in relapsed Philadelphia-positive acute lymphoblastic leukaemia [E6]. The qualifier is monotherapy at relapse, not acuity: given frontline with intensive chemotherapy to newly diagnosed children with Philadelphia-positive ALL, the same drug produced a 3-year event-free survival of 80% against 35% in historical controls and removed the advantage of transplant (counter-evidence on [E6]). Line of therapy and combination partner, not disease acuity, set the ceiling. Second, what rescued the class from point-mutation resistance was not chemotype diversity at the ATP site but a second, functionally ORTHOGONAL site -- the ABL1 myristoyl autoinhibitory pocket, now registered as asciminib [E7] -- a property of one protein's regulatory architecture rather than of kinases generally.

Go deeper
Two modules, one of which is a real enzyme pocket2 deeper

Structurally the pattern is monotonous and that is the point. BCR supplies a coiled-coil oligomerisation domain to ABL1; NPM1 supplies its N-terminal oligomerisation domain to ALK; ETV6 supplies the SAM/PNT domain, which polymerises head-to-tail, to NTRK3. In every case the partner's only essential contribution is to force the kinase domains together so they trans-autophosphorylate without a ligand.

The therapeutically decisive fact is that the retained module is a catalytic domain with a buried ATP site. Crystal structures of the ABL1 kinase domain with imatinib show the drug occupying the canonical ATP-binding site and trapping a specific inactive conformation of the activation loop 1. That is a textbook druggable pocket: enclosed, hydrophobic, ~500 cubic angstroms, with a conserved gatekeeper residue to exploit for selectivity.

What the clinical record looks like

Imatinib in interferon-refractory chronic-phase CML gave complete haematologic responses in 53 of 54 patients at doses of 300 mg or more in the phase 1 study 2, and at almost 11 years of follow-up in the randomised IRIS trial the estimated overall survival for imatinib-assigned patients was 83.3% 3. In the ALK and NTRK families the same logic transfers across tissues: larotrectinib produced an overall response rate of about 75% across 17 different TRK-fusion tumour histologies in adults and children 4, which is why NTRK became the second tissue-agnostic approval in oncology.

No other fusion mechanism class has anything approaching this record.

The cost: a pocket drug selects pocket mutants

Because the drug and the substrate share the same site, a single amino-acid change can abolish drug binding while preserving catalysis. Clinical resistance to imatinib was traced to BCR-ABL kinase-domain point mutations (and amplification) within months of the drug reaching patients 5. The field answered with second- and third-generation inhibitors and finally an allosteric myristoyl-pocket drug — an option that only exists because the target is an enzyme with more than one regulatory site.

The structural parallel to the menin-inhibitor story is exact: in both cases resistance came from mutations at the drug-binding surface rather than from loss of the oncogenic mechanism 5. The difference is that for kinases there was a deep bench of alternative chemistry to switch to.

Therapeutic implication — stated plainly

This class has yielded drugs, repeatedly, and is the reason the whole field expects fusions to be druggable. It should be treated as the exception, not the template. Its tractability comes from one property that no chromatin or transcription-factor fusion has: the oncogenic function is catalysis in an enclosed active site.

KMT2A::AFF1 shares none of this. The KMT2A SET methyltransferase domain — the only catalytic module in the parent protein — is always lost in the fusion, and AFF1 has no enzymatic activity. There is therefore no active site on the fusion to inhibit, which is precisely why the field went after the partner interface instead. Any strategy that reasons 'fusions are druggable because BCR::ABL1 was' is importing an assumption that fails at the first structural question.

Correction 2026-09-19 — 'disease acuity sets the ceiling' was wrong

This entry previously asserted, in its summary and its bite-back, that "Disease acuity, not pocket depth, sets the ceiling" and that "Matched for acuity, the kinase class's precedent is 6% sustained responses and 4.9 months median survival." Both statements are withdrawn as stated.

They rested on imatinib given as a single agent to relapsed or refractory patients. Matched for acuity but given the way a targeted agent is actually deployed - frontline, with intensive chemotherapy, in newly diagnosed disease - the same drug against the same fusion in paediatric Philadelphia-positive ALL gave a 3-year event-free survival of 80% +/- 11% against 35% +/- 4% in historical controls, with no appreciable added toxicity; at five years the chemotherapy-plus-imatinib disease-free survival was 70% +/- 12%, indistinguishable from sibling-donor (65%) or unrelated-donor (59%) transplant, so transplant lost its advantage (counter-evidence on 6).

The corrected statement is: line of therapy and combination context, not disease acuity, set the ceiling for this class. That change matters for the lab's own programme, because the comparison that was being drawn - a chronic-disease drug against an acute paediatric leukaemia - was the wrong comparison. The right one is a frontline targeted-agent-plus-chemotherapy combination in newly diagnosed acute paediatric leukaemia, and in the single case where that has been run to maturity it doubled cure and replaced transplant. Menin inhibitors are at present tested almost entirely in the relapsed/refractory setting, which is the setting in which the kinase precedent looks worst.

Facts and figures
mechanism classconstitutive-kinase
invariant moduleretained tyrosine kinase catalytic domain with an intact ATP-binding cleft (ABL1, ALK, NTRK3)
variable module5-prime oligomerisation element that removes ligand dependence (BCR coiled-coil, NPM1 oligomerisation domain, ETV6 SAM/PNT polymerisation domain)
druggable nodethe enzyme active site on the fusion itself, plus allosteric regulatory pockets on the same domain
druggability classclinically-validated
why tractableThe oncogenic activity is catalysis inside a deep, enclosed, conserved ATP pocket, so classical structure-based design applies and multiple chemotypes and allosteric sites are available.
shared with kmt2a aff1none — the KMT2A SET catalytic domain is always lost in the fusion and AFF1 has no enzymatic activity, so there is no active site to inhibit
resistance signaturepoint mutation at the drug-binding surface of the retained catalytic domain
member fusions
[
  "fusion-bcr-abl1",
  "fusion-npm1-alk",
  "fusion-etv6-ntrk3"
]
drugs yielded
[
  "drug-imatinib",
  "dasatinib",
  "nilotinib",
  "ponatinib",
  "asciminib",
  "drug-crizotinib",
  "alectinib",
  "lorlatinib",
  "larotrectinib",
  "entrectinib"
]
drugs failed or stalled
[]
acuity qualifier
{
  "chronic_phase_cml_10y_os_pct": 83.3,
  "myeloid_blast_crisis_median_survival_months": 6.9,
  "relapsed_ph_positive_all_median_os_months": 4.9,
  "pmids": [
    "28273028",
    "11986204",
    "12200353"
  ],
  "implication": "the class reliably produces DRUGS; it does not reliably produce CURES in acute leukaemia, and the fusion classes this beat compares against are all acute"
}
the resistance escape was an orthogonal site
{
  "site": "ABL1 myristoyl autoinhibitory pocket (asciminib/ABL001)",
  "pmids": [
    "28329763",
    "36717654"
  ],
  "generalisability": "a property of ABL1's regulatory architecture, disabled by the fusion itself; no analogous autoinhibitory pocket is described for menin or for any non-enzymatic fusion target in this graph"
}
How this could bite us back
  • Active-site inhibition selects gatekeeper and activation-loop mutants very quickly, and each generation of inhibitor selects the next mutant, so the class trades cure for indefinite therapy in most settings.
  • The target kinases have essential normal functions (ABL1 in development, NTRK in neural tissue, ALK in the nervous system), producing on-target toxicities that matter more in growing children than in adults.
  • The success of this class is the main source of unwarranted optimism about the rest of the atlas; importing its assumptions into a chromatin-fusion programme wastes time and money.
  • Quoting chronic-phase CML numbers in a comparison with infant acute leukaemia is the single most flattering move available to this beat - and quoting the relapsed-monotherapy numbers is the mirror-image error. Matched for acuity AND for line of therapy, the kinase precedent is 6% sustained responses and 4.9 months median survival as single-agent salvage, and 80% 3-year EFS given frontline with chemotherapy. Which number this beat quotes should follow the setting being compared. (Corrected 2026-09-19 against PMID 19805687; the second sentence previously read 'Matched for acuity, the kinase class's precedent is 6% sustained responses and 4.9 months median survival', with no mention of line of therapy.)
  • An allosteric or orthogonal site is still a drug-binding surface. Asciminib, the class's answer to ATP-site resistance, is itself defeated by myristoyl-pocket mutations (A337V/T, I502L) and by mutations that break the autoinhibited conformation it requires; in the randomised setting it reached major molecular response in 37.6% at 96 weeks. Any argument that an interface drug escapes the mutational arms race needs to explain why this one did not.
Still open
  1. Is there any latent enzymatic or nucleotide-binding function in a KMT2A fusion complex component (CDK9 in the SEC is the obvious candidate) that could be attacked with kinase-grade chemistry while sparing global transcription?
  2. Does the strict dependence of these kinase fusions on partner-driven oligomerisation mean their oligomerisation interfaces are a second, untried drug target that would be mutation-resistant in a way the ATP pocket is not?
  3. In paediatric Ph-positive ALL, why does kinase inhibition plus chemotherapy fall so far short of the CML result, and does that ceiling tell us something about how much of leukaemia a single driver actually explains?
  4. How much of this class's reputation is disease biology rather than chemistry? A like-for-like comparison would be TKI monotherapy in ALK-positive ALCL or Ph-positive ALL against menin-inhibitor monotherapy in KMT2A-r, all in acute disease -- and on that comparison the gap narrows sharply.
  5. If the kinase precedent in acute paediatric leukaemia is a frontline drug-plus-chemotherapy combination rather than salvage monotherapy, what is the earliest line in which a menin inhibitor has been tested, and is the field repeating the relapsed-monotherapy design that made imatinib look ineffective in acute disease for a decade?
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Imatinib binds the canonical ATP-binding site of the ABL1 kinase domain and captures a specific inactive conformation of the activation loop, as shown by crystal structures.

    PMID 12154025
    What this evidence cannot show

    Structures are of the isolated ABL1 kinase domain, not of the full BCR::ABL1 fusion, so conformational effects of BCR-driven oligomerisation are not captured.

    Evidence that argues against this (2)
    • The conformation in the crystal does not explain the drug's behaviour, which is the property the entry is using it for. NMR and fast kinetics show that Abl and Src adopt almost identical structures when bound to imatinib yet have vastly different affinities, and that the origin of Abl's high affinity lies predominantly in a conformational change AFTER binding - an induced fit - rather than in selection of the inactive activation-loop conformation the structures display. Every conformational-selection model of imatinib specificity, including the one this claim implies, was tested against this data and failed. The structure is real; the causal story usually told from it is not. Agafonov RV et al. 2014
    • What is actually modelled is much less than 'the fusion'. The primary structure of the cited paper, PDB 1IEP, is a 2.1 A X-ray structure of a SINGLE polymer chain of 293 residues, and SIFTS maps that chain to UniProt P00520 - ABL1_MOUSE, the 1123-residue mouse Abl1 - beginning at reference residue 229. There is no BCR moiety in the crystal, no SH3-SH2 regulatory cap, no myristoylated C-lobe and no human sequence. So the claim is about how imatinib binds an isolated murine kinase domain, roughly a quarter of the parent protein, not about how it engages the oncoprotein in a patient's cell. That matters because the myristoyl pocket that later produced asciminib [E7] and the SH3/SH2 clamp whose loss causes primary asciminib resistance both lie outside the modelled region. Checked against the RCSB entry record on 2026-09-19. Nagar B et al. 2002
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In the first-in-human trial of imatinib in interferon-refractory chronic-phase CML, 53 of 54 patients treated at 300 mg per day or more achieved a complete haematologic response.

    Efficacy and safety of a specific inhibitor of the BCR-ABL tyrosine kinase in chronic myeloid leukemia · Druker BJ, Talpaz M, Resta DJ, et al. · New England Journal of Medicine 2001
    PMID 11287972DOI
    What this evidence cannot show

    Phase 1 dose-escalation, n=83, single disease phase, no randomisation and short follow-up; cytogenetic complete remissions were achieved in only 7 patients at that stage.

    Evidence that argues against this (2)
    • The class's headline claim -- 'the one fusion class that reliably produces drugs' -- is a claim about CHRONIC-PHASE CML. In CML myeloid blast crisis the same drug against the same fusion gave sustained haematologic response in 31%, major cytogenetic response in 16% and MEDIAN SURVIVAL 6.9 MONTHS. Sawyers CL et al. 2002
    • In relapsed or refractory Philadelphia-positive ACUTE LYMPHOBLASTIC leukaemia, imatinib monotherapy gave complete haematologic or marrow responses in 29% but sustained for four weeks in only 6%, with median time to progression 2.2 months and median overall survival 4.9 months. A deep ATP pocket does not make an acute leukaemia tractable; it makes a chronic myeloproliferative disease tractable. Ottmann OG et al. 2002
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    With a median follow-up of 10.9 years in the randomised IRIS trial, estimated 10-year overall survival among imatinib-assigned patients with chronic-phase CML was 83.3%, and 82.8% had a complete cytogenetic response.

    Long-term outcomes of imatinib treatment for chronic myeloid leukemia · Hochhaus A, Larson RA, Guilhot F, et al. · New England Journal of Medicine 2017
    PMID 28273028DOI
    What this evidence cannot show

    65.6% of the comparator arm crossed over, so the randomised comparison is effectively lost and the imatinib estimates are single-arm long-term follow-up; this is chronic-phase adult CML, not paediatric Ph-positive ALL.

    Evidence that argues against this (1)
    • IRIS-scale durability is also conditional. In EURO-SKI, 755 chronic-phase patients pre-selected for at least 3 years of TKI and at least 1 year of deep molecular response had molecular relapse-free survival of only 50% at 24 months after stopping -- i.e. the pocket controls rather than cures even in the most favourable subgroup. Saussele S et al. 2018
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Larotrectinib produced a high, histology-independent response rate in TRK fusion-positive cancers across many tumour types in both adults and children.

    Efficacy of larotrectinib in TRK fusion-positive cancers in adults and children · Drilon A, Laetsch TW, Kummar S, et al. · New England Journal of Medicine 2018
    PMID 29466156DOI
    What this evidence cannot show

    Pooled analysis of three single-arm studies with a small and histologically heterogeneous cohort; median duration of response was not reached at the time of reporting, and acquired NTRK kinase-domain resistance mutations were already documented.

    Evidence that argues against this (2)
    • 'Histology-independent' is an eligibility statement as much as a biological one. The pooled analysis that expanded the primary 55-patient set to 159 treated patients restricted eligibility to locally advanced or metastatic TRK-fusion tumours with a NON-CNS primary, so the most therapeutically difficult compartment was excluded by design rather than found to respond. Within that selected population the objective response rate was 79% (121/153 evaluable) but only 24 patients (16%) had a complete response, i.e. the great majority of the responses quoted as evidence that a kinase fusion is fully druggable were partial. Response rate across histologies is a weaker claim than depth of response within them. Hong DS et al. 2020
    • The tissue-agnostic response does not make the fusion a durable target. In a documented CTRC-NTRK1 pancreatic adenocarcinoma, larotrectinib gave a partial response at two months and the disease progressed at six; the next-generation TRK inhibitor selitrectinib, designed for exactly this resistance, held it for two months before further progression, and the patient died two months after that. One case cannot establish a rate, but it is the shape the class takes: a high initial response rate reported at a timepoint, and acquired resistance at the same drug-binding surface underneath it. O'Reilly EM et al. 2019
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Clinical resistance to imatinib is caused by BCR-ABL kinase-domain point mutation or gene amplification, i.e. escape at the drug-binding surface rather than loss of the oncogenic mechanism.

    PMID 11423618DOI
    What this evidence cannot show

    Early small series of relapsed patients; the relative contribution of kinase-domain mutations versus BCR-ABL-independent resistance and quiescent stem-cell persistence was established only later.

    Evidence that argues against this (1)
    • The class survived point-mutation resistance not because 'an enzyme offers more than one chemotype at the same site' but because ABL1 has a second, functionally orthogonal site: ABL001/asciminib occupies the myristoyl AUTOINHIBITORY pocket, and dual myristoyl-plus-ATP-site targeting suppresses resistance neither controls alone. That is a property of this one protein's regulatory architecture, not of kinases in general, and it is the feature no pocketless fusion class can copy. Wylie AA et al. 2017
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Against the same fusion in acute disease, imatinib's effect largely disappears: median survival 6.9 months in CML myeloid blast crisis and 4.9 months in relapsed/refractory Philadelphia-positive acute lymphoblastic leukaemia, with sustained responses in 6% of the latter.

    PMID 12200353DOI
    What this evidence cannot show

    Abstract only for both sources (PMIDs 11986204 and 12200353). Monotherapy, 2002, relapsed/refractory adults; modern Ph+ ALL outcomes with TKI plus chemotherapy, blinatumomab or transplant are far better. The point is about the pocket alone, which is what the class claim rests on.

    Evidence that argues against this (2)
    • This is the most consequential correction in this entry, because the entry built a class-level rule on it. The 6.9- and 4.9-month figures are imatinib as MONOTHERAPY in relapsed or refractory disease. Given frontline with intensive chemotherapy to newly diagnosed children and adolescents with Philadelphia-positive ALL, imatinib produced a 3-year event-free survival of 80% +/- 11% in the continuous-exposure cohort against 35% +/- 4% in historical controls treated with chemotherapy alone (p < .0001), with no significant added toxicity, and transplant conferred no advantage over chemotherapy plus imatinib. So the same drug against the same fusion in the acute paediatric disease did not 'largely disappear' - it more than doubled cure. What fails is single-agent salvage in chemotherapy-refractory blast populations, which is a statement about sequencing and disease burden, not about acuity setting a ceiling on what a targeted agent can do. Schultz KR et al. 2009
    • The frontline result held up on longer follow-up rather than decaying, which removes the obvious objection that the 3-year figure was an early-timepoint artefact. At five years the disease-free survival for the chemotherapy-plus-imatinib cohort was 70% +/- 12%, statistically indistinguishable from sibling-donor (65% +/- 11%) and unrelated-donor (59% +/- 15%) transplant, and the authors conclude that allogeneic transplant offers no advantage. A targeted agent added frontline to chemotherapy in acute paediatric leukaemia replaced transplant; that is the precedent the KMT2A programme should be reasoning from, not the relapsed-monotherapy figures. Schultz KR et al. 2014
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The escape from ATP-site resistance came from a second, orthogonal site on the same oncoprotein -- the ABL1 myristoyl autoinhibitory pocket, targeted by ABL001/asciminib, with dual targeting suppressing resistance that neither site controls alone.

    The allosteric inhibitor ABL001 enables dual targeting of BCR-ABL1. · Wylie AA, Schoepfer J, Jahnke W, et al. · Nature 2017
    PMID 28329763DOI
    What this evidence cannot show

    Abstract only. Industry discovery paper; the clinical standing rests on ASCEMBL (PMID 36717654), a molecular-response endpoint against a weak comparator. The structural point -- a second, functionally distinct pocket created by the fusion's own disabling of autoinhibition -- is not disputed.

    Evidence that argues against this (2)
    • The orthogonal site is not an escape from the problem, only a move of it. In ASCEMBL, the randomised trial of asciminib against bosutinib after two or more prior TKIs, the major molecular response rate at week 96 was 37.6% with asciminib - better than bosutinib's 15.8%, and still a minority of patients - with grade >=3 adverse events in 56.4%. A second site bought a roughly 22-percentage-point absolute improvement in a heavily pretreated population, not control of resistance. Hochhaus A et al. 2023
    • The myristoyl pocket mutates too, which is the specific prediction this entry's 'dual targeting suppresses resistance' framing needs to survive and does not. Clinical asciminib resistance runs through mutations in the myristoyl pocket itself (A337V/T, I502L) and through mutations outside it that break the autoinhibited conformation asciminib requires (F359C/I/V at the kinase-SH2 interface, M244V in the N-lobe), so an allosteric site is a drug-binding surface with the same liability as an ATP site. It is a review and is cited here as a map of the reported mutations rather than as primary evidence for any one of them. Costa A et al. 2025
Connected to