Children with NPM1::ALK lymphoma immunise themselves against their own fusion protein, and the antibody titre predicts relapse - while no autoantibody against any KMT2A fusion has ever been reported. The difference is not the junction; it is that ALK is an ectopic self-antigen and KMT2A and AFF1 are not.
also called anti-ALK antibodies · ALCL immune response · fusion as tumour antigen · the immunogenicity asymmetry
In one childhood cancer, patients' own immune systems spontaneously attack the fusion protein driving the disease, and the strength of that response predicts whether the cancer comes back. Nobody has ever found the equivalent response in children with KMT2A-rearranged leukaemia - and the likely reason is that the two halves of the KMT2A fusion are proteins the body already knows, so there is almost nothing foreign for the immune system to see.
The same thing, for a scientist
**More than 90% of children and adolescents with ALK-positive anaplastic large cell lymphoma have detectable anti-ALK antibodies in serum or plasma, and the antibody titre inversely correlates with the risk of relapse**; ALK-specific CD8 and CD4 T-cell responses are documented, and ALK DNA vaccination protected against lymphoma growth in a murine model [E2]. Epitope mapping across **129 ALK-positive ALCL patients and 21 controls** by overlapping peptide microarray localised the dominant epitopes to the **C-terminal ALK portion of NPM-ALK** - amino acid positions 469-496, 561-588 and 617-644 - with high-titre patients additionally recognising three epitopes at the N-terminus of the kinase domain [E1]. **This is the field's only example of a fusion oncoprotein functioning as a spontaneous, clinically prognostic tumour antigen in children**, and it is the natural comparator for every fusion-directed immunotherapy idea in this project. **The comparison for KMT2A is a clean zero.** `KMT2A[tiab] AND autoantibody[tiab]` returns **0** with querytranslation `"KMT2A"[Title/Abstract] AND "autoantibody"[Title/Abstract]`, and `MLL[tiab] AND autoantibody[tiab]` returns **0** with `"MLL"[Title/Abstract] AND "autoantibody"[Title/Abstract]`, both run through an esearch wrapper that raises rather than returning a silent zero. The identical template returns **8** for `ALK[tiab] AND autoantibody[tiab]`, so the instrument can return a one [E3]. **The break in the analogy is the most useful part, and it is structural rather than accidental.** Knorr et al.'s epitopes are in the ALK moiety, not at the fusion junction [E1] - the immune system is responding to an **ectopically expressed kinase**, not to a neo-junction. ALK's normal expression in humans is largely restricted to neural tissue; KMT2A and AFF1 are both ubiquitously expressed nuclear proteins. **Inference, labelled as such and not asserted by either source:** if central tolerance to KMT2A and AFF1 is complete, then the only genuinely foreign sequence in KMT2A::AFF1 is the junction peptide itself, which is a single short epitope that must bind an individual patient's HLA - a vastly smaller antigenic surface than a whole ectopic kinase domain. **So the ALK precedent should not be read as encouragement for fusion-junction immunotherapy in KMT2A-rearranged leukaemia. It is the opposite: it shows what a fusion needs in order to be immunogenic, and KMT2A::AFF1 does not have it.**
The mapping, with the epitope locations doing the work
| NPM1::ALK in ALCL | KMT2A::AFF1 in infant ALL |
|---|---|
| >90% of children have detectable anti-fusion antibodies 2 | no autoantibody report exists, under both KMT2A and MLL spellings 3 |
| titre inversely correlates with relapse risk 12 | no analogous biomarker |
| dominant epitopes at aa 469-496, 561-588, 617-644 of NPM-ALK - i.e. inside the ALK moiety 1 | the only certainly foreign sequence is the junction; no epitope has been mapped |
| ALK-specific CD8 and CD4 T cells documented; ALK DNA vaccine protects in mice 2 | no fusion-specific T-cell response reported |
| 3-prime partner is a receptor tyrosine kinase with restricted normal expression | 3-prime partner AFF1 is a ubiquitously expressed nuclear elongation factor |
| 5-prime partner NPM1 is ubiquitous - and is not where the epitopes are 1 | 5-prime partner KMT2A is ubiquitous |
Read the third and fifth rows together. The epitopes sit in the partner with restricted normal expression and not in the ubiquitous one. That is the pattern a tolerance argument predicts, and it is visible in the data rather than assumed. A fusion is immunogenic to the extent that it places a tissue-restricted protein where the immune system has not learned to ignore it - which is a property of which genes broke, not of the fact that a break occurred.
By that criterion KMT2A::AFF1 scores as badly as any fusion in this graph: both partners are ubiquitous, both are nuclear, and neither is tissue-restricted.
What this says about the immunotherapy that is actually working in this disease
KMT2A-rearranged B-ALL does have effective immunotherapy - blinatumomab and CD19 CAR-T - and it targets CD19, a lineage antigen, not the fusion. The ALK comparison explains why that is not a failure of imagination. In ALCL the fusion itself is the antigen because ALK is ectopic; in KMT2A-rearranged B-ALL the fusion supplies almost no antigenic surface, so the only available targets are lineage markers.
And that has a direct clinical consequence this graph already tracks: lineage switch. A lineage antigen can be abandoned by the leukaemia - KMT2A-rearranged B-ALL switching to a myeloid phenotype under CD19-directed pressure is documented and is one of the disease's defining resistance routes. A fusion antigen cannot be abandoned without abandoning the driver. ALCL therefore has the escape-proof antigen and does not need it, because chemotherapy already cures most of it; KMT2A-rearranged infant ALL needs an escape-proof antigen badly and, on this evidence, does not have one.
The honest conclusion is that the fusion-antigen route for this disease is not merely unexplored - it is unpromising for a specific, stated reason, and a project should say so rather than list it as an untried option.
How the absence was bounded, and what it does not cover
Both gene symbols were queried, because a search keyed only on KMT2A systematically excludes the pre-2018 MLL literature:
| query | count | querytranslation |
|---|---|---|
KMT2A[tiab] AND autoantibody[tiab] | 0 | "KMT2A"[Title/Abstract] AND "autoantibody"[Title/Abstract] |
MLL[tiab] AND autoantibody[tiab] | 0 | "MLL"[Title/Abstract] AND "autoantibody"[Title/Abstract] |
ALK[tiab] AND autoantibody[tiab] (positive control) | 8 | "ALK"[Title/Abstract] AND "autoantibody"[Title/Abstract] |
Context counts through the same wrapper, recorded so the absence is not mistaken for an absence of any immunological literature: KMT2A[tiab] AND "immune response"[tiab] 6, MLL[tiab] AND "immune response"[tiab] 38, "MLL-AF4"[tiab] AND "T cell"[tiab] 24, KMT2A[tiab] AND epitope[tiab] 2, MLL[tiab] AND epitope[tiab] 6, KMT2A[tiab] AND vaccine[tiab] 1, MLL[tiab] AND vaccine[tiab] 14. There is a substantial immunology literature around these fusions; what is absent is specifically a report of a patient autoantibody against the fusion protein.
What this absence does not cover. It is bounded by titles and abstracts, so a serological measurement reported only in a supporting figure would be invisible. It is bounded by the word autoantibody: a paper describing 'humoral response', 'seroreactivity' or 'patient sera recognise' without that noun would be missed. And it is a statement about published reports, not about biology - the absence of a reported autoantibody is compatible with a weak response that nobody has looked for, which is in fact the most likely state of affairs.
Related but distinct, and already in this graph: ana-q115-the-kmt2a-junction-antibody-absence-survives-in-the-catalogue-that-can-see-it bounds the absence of a junction-specific laboratory reagent antibody in the Antibody Registry. That is a question about reagents; this is a question about patients' own immune systems. The two absences are independent and both hold.
| domain | engineering |
|---|---|
| where the analogy breaks | It breaks at the antigen itself, and the break is visible in the epitope map rather than assumed: the anti-ALK response targets the ectopically expressed ALK kinase domain, not the neo-junction. KMT2A::AFF1 has no ectopic moiety - both partners are ubiquitously expressed nuclear proteins - so the ALK precedent does not transfer, and reading it as encouragement for fusion-junction vaccines or junction-directed T-cell therapy would invert its lesson. |
| what it predicts here | A serological screen of banked sera from KMT2A-rearranged patients against recombinant fusion protein should be NEGATIVE or very weakly positive, in contrast to ALCL. That is a cheap, decisive experiment on existing biobank material, and a clean negative would be publishable and would close a route this field keeps listing as open. It also predicts that fusion-directed immunotherapy effort in this disease is better spent on lineage or fusion-DRIVEN surface antigens than on the junction. |
| immune axis | tumour antigen and spontaneous host immunity; cell type is the paediatric lymphoid compartment in both diseases |
| tolerance argument status | INFERENCE, not a claim made by either cited source. Neither Knorr et al. nor Stadler et al. states that ALK's restricted normal expression explains the response, and neither discusses KMT2A. The observation that the mapped epitopes lie in the ALK moiety and not in the ubiquitous NPM1 moiety is from the source; the extension to a tolerance mechanism and to KMT2A::AFF1 is this entry's reasoning and is labelled as such. |
| relationship to existing entries | ana-q115-the-kmt2a-junction-antibody-absence-survives-in-the-catalogue-that-can-see-it bounds the absence of a junction-specific laboratory REAGENT in the Antibody Registry; this entry bounds the absence of a PATIENT autoantibody in the literature. Different objects, different instruments, both absences hold. fusion-npm1-alk describes the fusion and its kinase biology; PMIDs 30077013 and 29642597 are cited in zero node files as of 2026-09-23. |
mapping
[
{
"bio": "NPM1::ALK places an ectopic receptor tyrosine kinase in a lymphoid cell; >90% of paediatric patients make anti-ALK antibodies and the titre inversely tracks relapse risk",
"analog": "KMT2A::AFF1 joins two ubiquitously expressed nuclear proteins; no autoantibody against any KMT2A fusion has been reported under either gene symbol"
},
{
"bio": "dominant epitopes map to the ALK moiety (aa 469-496, 561-588, 617-644), NOT to the fusion junction",
"analog": "for KMT2A::AFF1 the junction is the only certainly foreign sequence, and a junction peptide is one HLA-restricted epitope rather than a whole ectopic domain"
},
{
"bio": "ALK-specific CD8 and CD4 T cells exist and ALK DNA vaccination protects mice",
"analog": "no fusion-specific T-cell response or vaccine result exists for a KMT2A fusion"
},
{
"bio": "the antigen cannot be abandoned without abandoning the driver",
"analog": "the antigen that works in KMT2A-rearranged B-ALL is CD19, a LINEAGE antigen - and lineage switch is a documented escape route precisely because a lineage antigen is abandonable"
},
{
"bio": "immunogenicity is a property of WHICH genes broke, not of the fact that a break occurred",
"analog": "on that criterion KMT2A::AFF1 is among the least promising fusions in this graph for a fusion-directed immunotherapy"
}
]generated hypotheses
[]
- The most likely misuse of this entry is as a rationale for a KMT2A::AFF1 junction vaccine. It argues the opposite: the one fusion with a documented spontaneous immune response is immunogenic because of an ectopic kinase, not because of a junction, and KMT2A::AFF1 has no ectopic moiety.
- A junction peptide is HLA-restricted, so even if it were immunogenic the approach would serve only the subset of patients whose HLA presents it - and infant ALL is a small population before that subsetting.
- The immunotherapy that currently works in this disease targets a lineage antigen that the leukaemia can and does abandon by switching lineage. This entry explains why the escape-proof alternative is unavailable rather than merely untried, which makes lineage switch a harder problem than it looks.
- Any attempt to raise immunity against a nuclear protein expressed in every cell of the body carries an autoimmunity risk that scales with how well it works. The reason KMT2A::AFF1 is a poor antigen is the same reason targeting it immunologically would be dangerous if it succeeded.
- Do banked sera from KMT2A-rearranged patients contain any reactivity against recombinant KMT2A::AFF1 or against a junction peptide? This is the experiment the ALCL field ran, the material exists in paediatric biobanks, and a clean negative is as valuable as a positive because it would close the route with data instead of with an argument.
- Is the anti-ALK response directed at the neo-junction at all, or entirely at the ectopic kinase? The published epitope map says the kinase moiety, and if a junction epitope is genuinely absent even in the most immunogenic fusion known, that is strong general evidence against junction-directed immunotherapy for every fusion.
- Across the 52 fusion families in this graph, which fusions place a TISSUE-RESTRICTED protein into a haematopoietic cell? That set is the candidate list for spontaneous fusion immunity, it is computable from GTEx or HPA plus this graph's own partner table, and nobody has produced it. CBFA2T3::GLIS2 and TCF3::HLF both import a factor normally expressed elsewhere and would be the first two rows.
- Does antibody titre predict relapse in ALCL because the response is protective, or because titre tracks antigen load? The distinction decides whether ALK vaccination is a reasonable clinical idea, and the review states the determinants are largely unknown.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Overlapping peptide microarray analysis of plasma or serum from 129 ALK-positive anaplastic large cell lymphoma patients and 21 controls showed that antibodies in patient sera bound epitopes mainly in the C-terminal ALK portion of NPM-ALK at amino acid positions 469-496, 561-588 and 617-644; patients with higher ALK antibody titres more frequently recognised the 561-588 epitope and three further epitopes at the N-terminus of the kinase domain; and the titre of these autoantibodies inversely correlates with the risk of relapse.
Epitope mapping of anti-ALK antibodies in children with anaplastic large cell lymphoma. · Knorr F, Weber S, Singh VK, Pulford K, Reiter A, Woessmann W, Damm-Welk C · Clin Immunol 2018What this evidence cannot show
Abstract read in full from the PubMed record on 2026-09-23. The inverse titre-relapse correlation is stated in this paper's opening as established background rather than derived here; E2 is the review that carries it. The epitope localisation is to the ALK moiety and the abstract does not report a junction-spanning epitope, which is the observation this entry leans on - but absence of a reported junction epitope in an abstract is not proof that none was detected, and the array's coverage of the junction region was not verified from the full text.
Evidence that argues against this (1)
- If the peptide array under-sampled the NPM-ALK junction region, the conclusion that the response is directed at the ectopic ALK moiety rather than at the neo-junction would be an artefact of array design. The entry's tolerance argument rests on this distinction and the full text was not read, so the distinction is recorded as the abstract reports it and not as verified. Stadler S et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review
Patients with ALK-positive anaplastic large cell lymphoma mount both humoral and cellular immune responses against ALK: more than 90% of children and adolescents with ALK-positive ALCL have detectable anti-ALK antibodies in serum or plasma, the antibody titre inversely correlates with the risk of relapse, ALK-specific CD8 and CD4 T-cell responses have been described, and vaccination with ALK DNA protected against lymphoma growth in a murine model - together suggesting the ALK-specific immune response is involved in control of the disease.
Immune Response against ALK in Children with ALK-Positive Anaplastic Large Cell Lymphoma. · Stadler S, Singh VK, Knorr F, Damm-Welk C, Woessmann W · Cancers (Basel) 2018What this evidence cannot show
This is a REVIEW and is cited as a map of the field rather than as a primary observation; E1 is the primary epitope-mapping study from the same group. The mouse DNA vaccination result is preclinical. The review explicitly notes that tumour and host factors determining the individual immune response are still largely unknown, so the mechanism of the titre-relapse relationship is not established.
Evidence that argues against this (1)
- An inverse correlation between antibody titre and relapse risk is also consistent with titre being a marker of tumour burden or of antigen load rather than of protective immunity, in which case the immune response would be a consequence of the disease biology rather than a control mechanism. The review's own framing - that host and tumour determinants are largely unknown - leaves this open. Knorr F et al. 2018
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
Counted absence, 2026-09-23: a PubMed title/abstract search for an autoantibody against a KMT2A fusion returns zero under both gene symbols - KMT2A[tiab] AND autoantibody[tiab] returns 0 with querytranslation "KMT2A"[Title/Abstract] AND "autoantibody"[Title/Abstract], and MLL[tiab] AND autoantibody[tiab] returns 0 with querytranslation "MLL"[Title/Abstract] AND "autoantibody"[Title/Abstract] - while the identical template applied to ALK[tiab] AND autoantibody[tiab] returns 8, establishing that the instrument can return a positive.
PubMed esearch, run through scripts/pubmed_safe.py, which retries an empty querytranslation and raises rather than returning a false zero; queries and querytranslations recorded in the entry body · NCBI Entrez Programming Utilities (esearch, db=pubmed) · PubMed / NCBI E-utilities 2026What this evidence cannot show
Bounded by titles and abstracts, so a serological measurement appearing only in a supporting figure is invisible - the standing project rule for counted absences on a MEASUREMENT rather than an entity. Bounded by the single noun 'autoantibody'; 'humoral response', 'seroreactivity' and 'patient sera' were not searched as alternative phrasings, and that is the most likely way a positive would have been missed. Context counts through the same wrapper show the surrounding literature is not empty: KMT2A AND 'immune response' 6, MLL AND 'immune response' 38, 'MLL-AF4' AND 'T cell' 24, KMT2A AND epitope 2, MLL AND epitope 6, KMT2A AND vaccine 1, MLL AND vaccine 14.
Evidence that argues against this (1)
- The absence of a published autoantibody report is not evidence that no response exists. The ALCL field found its antibodies because somebody built the assay and ran it on banked sera; nobody appears to have run the equivalent assay against a KMT2A fusion. The correct reading is 'unlooked-for', not 'shown to be absent', and the entry's tolerance argument is what makes a negative result plausible in advance rather than any measurement. Stadler S et al. 2018
- related to →NPM1::ALK (NPM-ALK), t(2;5)(p23;q35)
the family whose patients spontaneously immunise themselves against the fusion
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the index fusion, for which no patient autoantibody has been reported under either gene symbol
the kinase moiety carrying the mapped epitopes at aa 469-496, 561-588 and 617-644
- related to →Blinatumomab in infant KMT2A-rearranged ALL — the largest single outcome gain in thirty years
the lineage-antigen route that works here because the fusion-antigen route is unavailable
the disease in which a fusion-directed immunotherapy would be needed and is least likely to be available
sibling entry - that one closes the junction-neoantigen route, this one opens the fusion-driven surface-antigen route