Oncofusion

start / graph / concept-graft-versus-leukaemia-in-kmt2a-rearranged-infant-leukaemia

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q53-transplant

Graft-versus-leukaemia in KMT2A-rearranged infant leukaemia - the mechanism the whole transplant rests on, demonstrated in ALL generally, assumed in this disease, and with a published mechanistic reason to doubt it

also called GVL · graft-versus-leukemia effect · graft versus leukaemia · donor immune effect · alloreactivity against leukaemia

A transplant is supposed to cure leukaemia not by the chemotherapy but by the donor's immune cells recognising and killing any leukaemia left behind. In childhood leukaemia as a whole this has been measured and it is real. In this particular infant leukaemia it has never been demonstrated, and a laboratory study found that these cells are unusually resistant to exactly the killing signals the donor cells use.

The same thing, for a scientist

Graft-versus-leukaemia is the mechanism that justifies allogeneic transplant over intensification of chemotherapy, and it is the one antileukaemic mechanism in this disease that shares no escape route with any pharmacological agent. In 1,619 paediatric ALL patients transplanted in CR1/CR2 and reported to CIBMTR, acute or chronic GVHD was associated with a lower relapse hazard (HR 0.49-0.69), with a net overall survival benefit only for grade I-II acute GVHD without chronic GVHD [E1]. That is a demonstration in paediatric ALL as a whole, not in this subtype. Restricted to infants with KMT2A rearrangement, a bounded PubMed search on 2026-09-20 returns FOUR records at the intersection of GVL and (KMT2A OR MLL OR 11q23) and (infant OR infants), of which the only case report of a documented GVL effect in an infant is explicitly in an MLL-NEGATIVE leukaemia [E2], and the only mechanistic study concludes the opposite: KMT2A-rearranged infant ALL and AML cells are highly resistant to TRAIL and poorly sensitive to FasL, which the authors propose as a mechanism for poor sensitivity to GVL [E3]. A 2026 study now shows that menin inhibition restores MHC class II expression and enhances GVL in KMT2A-rearranged AML [E4] - which is an argument that GVL is impaired in this biology and is correctable, not that it was working all along.

Go deeper
What is demonstrated, and in whom

The CIBMTR analysis of 5,215 ALL transplants is the reference demonstration 1:

cohortfinding
2,593 adults CR1/CR2acute or chronic GVHD -> lower relapse, HR 0.49-0.69
1,619 paediatric CR1/CR2same direction, same hazard range
1,003 advanced ALLgrade III-IV aGVHD or cGVHD -> lower relapse, HR 0.52-0.67

And the net effect on survival is narrow: 'adult and children in CR1/CR2 with grades I and II aGVHD without cGVHD experienced the best OS compared with no GVHD (reduction of mortality with HR, 0.83-0.76)', while grade III-IV acute GVHD raised non-relapse mortality 2.69- to 3.91-fold and abrogated any protection from relapse 1.

So GVL in paediatric ALL is real, and it is beneficial only in a narrow band of GVHD severity. Too little and there is no antileukaemic effect; too much and the child dies of the graft rather than the leukaemia. The therapeutic window is the GVHD grade.

None of this is stratified by fusion, and the infant fraction of 1,619 paediatric patients is not reported.

The bounded absence in this disease1 deeper

Measured on PubMed on 2026-09-20, with controls through the identical query shape:

queryrecords
"graft-versus-leukemia"[tiab] OR "graft versus leukemia"[tiab] OR "graft-versus-leukaemia"[tiab] OR GVL[tiab]3,381
the same GVL clause AND (KMT2A[tiab] OR MLL[tiab] OR "11q23"[tiab] OR "t(4;11)"[tiab])19
the GVL clause AND (infant[tiab] OR infants[tiab])9
all three clauses together4

The instrument composes and returns non-zero at every level, so the four is a measurement and not an artefact. Title-checking those four:

  • PMID 17066095 - the TRAIL-resistance study, which argues GVL does not work here 3
  • PMID 17960638 - the only case report of a demonstrated GVL effect in an infant, and it is MLL-negative (t(7;12) HLXB9/ETV6) 2
  • PMID 27777140 - the EBMT cord blood series, which asserts the GVL rationale in its opening sentence without measuring it 5
  • PMID 28646023 - a TNF-alpha sensitisation study in KMT2A::AFF1 ALL, not a transplant study

Zero of the four demonstrate a graft-versus-leukaemia effect in KMT2A-rearranged infant leukaemia. Name list used: graft-versus-leukemia, graft versus leukemia, graft-versus-leukaemia, GVL; KMT2A, MLL, 11q23, t(4;11); infant, infants. A search on 'donor lymphocyte infusion' or 'alloreactivity' was not run and could return more.

A note on one of those tokens: "t(4;11)"[tiab] quoted inside a larger parenthesised query does not trigger this project's der(n) trap - PubMed's querytranslation returns "t 4 11"[Title/Abstract] and the term returns 3 records when paired with the GVL phrase. The quoting is what saves it.

The mechanistic reason to doubt it

Inukai and colleagues tested MLL-rearranged ALL and AML cells against the two cytotoxic pathways that cytotoxic T lymphocytes and NK cells use 3:

  • Most cell lines and primary cells were highly resistant to TRAIL - in ALL because of low cell-surface death-receptor expression, in AML because of simultaneous decoy-receptor expression
  • Nearly half of cell lines and the majority of primary cells showed low sensitivity to FasL

And the sentence that frames it: 'Infant leukemias with mixed-lineage leukemia (MLL) rearrangement have a remarkably short latency, and GVL effect after allo-SCT has not been clearly evidenced in these leukemias.' That was written in 2006 and, on the search above, is still accurate in 2026.

This is in vitro, it is a proposed mechanism rather than a measured clinical failure, and the alloreactive repertoire includes perforin/granzyme killing that these assays do not test. But it is a specific, published, mechanistically coherent reason why the one antileukaemic mechanism transplant offers might be weaker in this disease than in the paediatric ALL population where it was demonstrated.

The 2026 result that changes the shape of the question

Fetsch and colleagues report that menin inhibition 'induced class II transactivator and major histocompatibility complex II (MHC-II) expression in KMT2A-rearranged and NPM1-mutated AML cells in vitro and in vivo', that increased MHC-II 'sensitized AML cells to T-cell-mediated elimination after allo-HCT in mice', and that menin inhibition 'enhanced the graft-versus-leukemia (GVL) effect in human xenograft models' 4. The mechanism runs through derepression of human endogenous retroviruses, interferon-stimulated gene upregulation and consequent MHC-II induction, plus a direct effect on donor T-cell effector function.

Read what that implies about the baseline. If pharmacologically restoring MHC-II enhances GVL, then MHC-II was limiting GVL in untreated KMT2A-rearranged cells - which is an antigen-presentation argument for exactly the conclusion Inukai reached from a death-receptor argument twenty years earlier. Two independent mechanisms, two decades apart, both saying the KMT2A-rearranged blast is a poor target for donor T cells.

It is AML, in mice and xenografts, not infant ALL in children. But it converts 'is there a GVL effect in this disease?' from an unanswered question into a tractable one with a candidate intervention, and it supplies the mechanistic rationale that the five registered post-transplant menin-maintenance studies in concept-hsct-in-kmt2a-r-leukemia were launched without.

Assumed, and the assumption is visible in print

The largest infant transplant dataset opens by stating the mechanism as established: umbilical cord blood 'is an attractive cell source for this population because of the low risk of chronic graft-versus-host disease (GVHD), the strong graft-versus-leukemia effect, and prompt donor availability' 5.

That sentence is in a paper reporting 252 children with acute leukaemia diagnosed before one year of age, and the study measures relapse (27% at 4 years), acute GVHD (40% grades II-IV) and chronic GVHD (13% at 4 years) without relating them to each other. The GVL claim is a premise of the paper, not a result of it.

This is the shape this graph should record: in infant leukaemia, graft-versus-leukaemia is a rationale, appearing in the introduction of papers rather than in their results.

Facts and figures
status of the mechanismDEMONSTRATED in paediatric ALL as a population (CIBMTR, n=1,619 paediatric CR1/CR2); NOT demonstrated in KMT2A-rearranged infant leukaemia; ASSUMED in the introductions of infant transplant papers
cibmtr relapse hazard with gvhdHR 0.49-0.69 for acute or chronic GVHD versus none, CR1/CR2
net survival benefit windowgrade I-II acute GVHD without chronic GVHD only (mortality HR 0.83-0.76); grade III-IV acute GVHD raises NRM 2.69-3.91-fold and abrogates the relapse protection
largest infant transplant datasetEBMT cord blood, 252 children diagnosed before 1 year: aGVHD II-IV 40%, cGVHD 13%, TRM 23% at 1 year, relapse 27% at 4 years, LFS 50% (ALL 40%, AML 66%); age at diagnosis under 3 months independently worse
bounded absence
{
  "date": "2026-09-20",
  "tool": "PubMed E-utilities esearch, querytranslation read back",
  "gvl_clause_alone": 3381,
  "gvl_and_kmt2a_clause": 19,
  "gvl_and_infant_clause": 9,
  "all_three_clauses": 4,
  "on_topic_after_title_check": 0,
  "name_list_gvl": [
    "graft-versus-leukemia",
    "graft versus leukemia",
    "graft-versus-leukaemia",
    "GVL"
  ],
  "name_list_disease": [
    "KMT2A",
    "MLL",
    "11q23",
    "t(4;11)"
  ],
  "name_list_age": [
    "infant",
    "infants"
  ],
  "not_searched": "donor lymphocyte infusion, alloreactivity, minor histocompatibility antigen, DLI - any of which could return additional records",
  "parenthesis_note": "\"t(4;11)\"[tiab] inside a larger parenthesised query translated to \"t 4 11\"[Title/Abstract] and returned 3 records with the GVL phrase; QUOTING the token avoids this project's der(n) parenthesis trap"
}
mechanistic arguments against
[
  "TRAIL resistance through low surface death-receptor expression in MLL-rearranged ALL (Inukai 2006)",
  "TRAIL resistance through simultaneous decoy-receptor expression in MLL-rearranged AML (Inukai 2006)",
  "low FasL sensitivity in nearly half of cell lines and most primary cells (Inukai 2006)",
  "MHC class II being limiting, inferred from the fact that restoring it pharmacologically ENHANCES GVL (Fetsch 2026)"
]
mechanistic arguments for correctability
[
  "menin inhibition induces CIITA and MHC-II via HERV derepression and interferon-stimulated genes, enhancing GVL in KMT2A-r AML xenografts",
  "menin inhibition directly raises donor T-cell TNF-alpha, IFN-gamma, perforin and granzyme A/B and reduces T-cell exhaustion",
  "these are the effector routes the 2006 death-ligand assays did not test"
]
How this could bite us back
  • The mechanism that justifies the intervention is the one thing in this disease that has never been measured. Every argument for transplant over chemotherapy intensification runs through graft-versus-leukaemia, and in KMT2A-rearranged infant leukaemia the literature contains a resistance mechanism and an assumption, and no demonstration.
  • The therapeutic window is a GVHD grade, and it is narrow. Grade I-II acute GVHD helps; grade III-IV kills more children than it saves. Nothing in transplant practice lets a clinician aim for grade I-II, and in an infant the consequences of overshooting are worst.
  • Restoring GVL pharmacologically implies GVL was broken. The most encouraging result in this beat - menin inhibition enhancing graft-versus-leukaemia - is simultaneously the strongest new evidence that the baseline effect in KMT2A-rearranged disease is impaired.
  • A post-transplant menin inhibitor is being given for an immunological reason that was discovered after the trials were designed, in a different lineage and a different species. If the MHC-II axis does not operate in B-precursor disease, five registered maintenance studies are testing a drug in a setting whose mechanism does not hold.
  • Fludarabine-containing conditioning depletes exactly the effector cells GVL requires, and it has become the international standard. The regimen that best protects the graft may be the one that most blunts its antileukaemic action, and in a disease where GVL may already be weak that trade-off has never been quantified.
  • If GVL is absent, the transplant is a delivery vehicle for alkylator intensity and the late effects are being accepted for a purely cytotoxic benefit - one that a better-tolerated chemotherapy intensification might match.
Still open
  1. Does GVHD reduce relapse in KMT2A-rearranged infant ALL specifically? The CIBMTR dataset that demonstrated GVL in paediatric ALL holds the patients needed to answer this; the analysis has simply never been stratified by fusion and age, and it is a query rather than a new study.
  2. Does the menin-inhibitor MHC-II mechanism operate in B-precursor KMT2A::AFF1 leukaemia, where MHC class II is already constitutively expressed? If the antigen-presentation half of the mechanism has no headroom in B lineage, the rationale for post-transplant menin maintenance in infant ALL rests only on the direct T-cell effects.
  3. Are the TRAIL and FasL resistance phenotypes of 2006 still present in contemporary KMT2A-rearranged primary samples, and do they predict post-transplant relapse? Nobody has re-measured them in twenty years, and the original study reported only in vitro sensitivity.
  4. If GVL is weak in this disease, does that change what transplant is FOR? A transplant whose immunological benefit is unproven is a very intensive way to deliver alkylator dose intensification, and that is a different intervention with a different comparator.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 5,215 allogeneic transplants for ALL reported to CIBMTR, including 1,619 paediatric patients in CR1/CR2, development of acute or chronic graft-versus-host disease was associated with a lower risk of relapse than no GVHD (HR 0.49-0.69); grade I-II acute GVHD without chronic GVHD gave the best overall survival (mortality HR 0.83-0.76), while grade III-IV acute GVHD raised non-relapse mortality 2.69- to 3.91-fold and abrogated any protection from relapse.

    The impact of the graft-versus-leukemia effect on survival in acute lymphoblastic leukemia. · Yeshurun M, Weisdorf D, Rowe JM, Tallman MS, Zhang MJ, Wang HL, Saber W, de Lima M, et al. · Blood Advances 2019
    PMID 30808685DOI
    What this evidence cannot show

    Registry data, observational, and structurally vulnerable to the same guarantee-time problem as the transplant-benefit literature: GVHD is a time-dependent exposure, so patients who relapse early have less opportunity to develop it. No fusion stratification is reported, the infant fraction of the 1,619 paediatric patients is not stated, and the conditioning regimens, donor types and GVHD prophylaxis span decades. It establishes that GVL exists in paediatric ALL as a population; it establishes nothing about any subtype.

    Evidence that argues against this (1)
    • The only mechanistic study of GVL susceptibility in this specific disease reaches the opposite conclusion. MLL-rearranged infant ALL and AML cells were highly resistant to TRAIL - in ALL through low surface death-receptor expression, in AML through simultaneous decoy-receptor expression - and nearly half of cell lines and the majority of primary cells were also insensitive to FasL, which the authors propose as a mechanism for poor sensitivity to the GVL effect. A registry-wide GVL demonstration does not transfer to a subtype with a published resistance mechanism against the effector pathways involved. Inukai T et al. 2006
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    The only published case report of a demonstrated graft-versus-leukaemia effect in an infant leukaemia recovered by a bounded search is in an MLL-NEGATIVE infant AML: an 8-month-old with t(7;12)(q36;p13) HLXB9/ETV6 rearrangement who received haploidentical HSCT in relapse without prior re-induction and became MRD-negative over 53 days.

    Graft versus leukemia effect after haploidentical HSCT in a MLL-negative infant AML with HLXB9/ETV6 rearrangement. · Hauer J, Tosi S, Schuster FR, Harbott J, Kolb HJ, Borkhardt A · Pediatric Blood & Cancer 2008
    PMID 17960638DOI
    What this evidence cannot show

    A single patient. Attribution of MRD clearance to graft-versus-leukaemia rather than to the conditioning regimen rests on the absence of re-induction chemotherapy and on the 53-day timescale, which is suggestive and not conclusive. The leukaemia is AML with a different fusion, in the haploidentical setting, and the point being made here is about what the LITERATURE contains, not about this patient's biology. The claim that this is the 'only' such report is bounded by the exact query and name list recorded in this entry's data payload; a search on donor lymphocyte infusion or alloreactivity was not run.

    Evidence that argues against this (1)
    • This graph already holds a patient story in which a KMT2A-USP2 infant received donor lymphocyte infusion after transplant alongside blinatumomab, which is a clinical use of the graft-versus-leukaemia principle in a KMT2A-rearranged patient. A case in which DLI is given is not the same as a case in which a GVL effect is demonstrated, but it means the practice exists in this fusion and the literature search above does not capture every instance. Inukai T et al. 2006
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Most MLL-rearranged ALL and AML cell lines and primary leukaemia cells were highly resistant to TRAIL-mediated cytotoxicity - in ALL primarily through low cell-surface death-receptor expression and in AML through simultaneous decoy-receptor expression - and nearly half of cell lines and the majority of primary cells showed low sensitivity to FasL; the authors state that a graft-versus-leukaemia effect after allogeneic transplant 'has not been clearly evidenced' in these leukaemias and propose death-ligand resistance as a mechanism.

    PMID 17066095DOI
    What this evidence cannot show

    Entirely in vitro. Recombinant TRAIL and FasL sensitivity assays test two of several effector pathways: perforin/granzyme-mediated killing, which is a dominant CTL and NK mechanism in vivo and which the 2026 menin-inhibition study measures directly, is not assessed here. No clinical correlation is reported, so 'poor sensitivity to the GVL effect' is a hypothesis the authors offer, not an outcome they measured. Published 2006, before the modern understanding of MHC class II regulation in KMT2A-rearranged leukaemia. The cell-line panel composition and the number of primary samples are not given in the abstract.

    Evidence that argues against this (1)
    • A 2026 study measuring the effector functions this one omits finds that the blocking step is correctable rather than absolute: menin inhibition increased donor T-cell production of TNF-alpha, interferon-gamma, perforin and granzyme A/B, reduced T-cell exhaustion, and enhanced the graft-versus-leukaemia effect in human xenograft models of KMT2A-rearranged AML. Death-ligand resistance in 2006 assays is compatible with an intact and pharmacologically restorable cytolytic route. Fetsch V et al. 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Menin inhibition induced CIITA and MHC class II expression in KMT2A-rearranged and NPM1-mutated AML cells in vitro and in vivo, sensitised them to T-cell-mediated elimination after allogeneic HCT in mice, and enhanced the graft-versus-leukaemia effect in human xenograft models; the mechanism runs through increased expression of multiple human endogenous retroviruses, consequent interferon-stimulated gene upregulation and MHC-II induction, together with direct promotion of donor T-cell effector function (TNF-alpha, interferon-gamma, perforin, granzyme A/B) and reduced T-cell exhaustion.

    Menin inhibition enhances graft-versus-leukemia effects by T-cell activation and endogenous retrovirus induction in AML. · Fetsch V, Schwoebel L, Ozyerli-Goknar E, Stell AV, Punta M, Plenge T, Klaus T, Gupta MK, et al. · Blood 2026
    PMID 41144759DOI
    What this evidence cannot show

    AML, in mouse models and human xenografts, not infant ALL in patients. Xenograft GVL models use human T cells in immunodeficient mice, where the alloreactive repertoire and the MHC context are not those of a matched paediatric transplant. The MHC class II argument is also lineage-loaded: MHC-II is constitutively expressed on B-lineage cells, so an MHC-II-restoration mechanism discovered in myeloid leukaemia may not transfer to a B-precursor ALL where the molecule is already present. The paper's own conclusion is a rationale for a clinical trial, not a clinical result, and no PMC record was available at the time of retrieval.

    Evidence that argues against this (1)
    • The mechanism as described may not apply in the graph's index disease for a lineage reason. KMT2A::AFF1 infant leukaemia is B-precursor and MHC class II is normally expressed in that lineage, so the CIITA/MHC-II derepression axis that makes myeloid blasts visible to donor T cells may have no headroom in a B-lineage blast. The direct donor T-cell effects reported in the same paper would still apply; the antigen-presentation half might not. Inukai T et al. 2006
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 252 children with ALL (n=157) or AML (n=95) diagnosed before 1 year of age receiving single-unit unrelated cord blood transplantation after myeloablative conditioning in EBMT centres between 1996 and 2012, day-100 grade II-IV acute GVHD was 40% +/- 3%, 4-year chronic GVHD 13% +/- 2%, 1-year transplant-related mortality 23% +/- 3%, 4-year relapse 27% +/- 3% and 4-year leukaemia-free survival 50% +/- 3% (40% ALL versus 66% AML, p=.001); the paper states the graft-versus-leukaemia rationale for cord blood in its opening sentence without measuring it.

    Unrelated Cord Blood Transplantation for Acute Leukemia Diagnosed in the First Year of Life: Outcomes and Risk Factor Analysis. · Ruggeri A, Volt F, Locatelli F, Michel G, Diaz de Heredia C, Abecasis M, Zecca M, Vora A, et al. · Biology of Blood and Marrow Transplantation 2017
    PMID 27777140DOI
    What this evidence cannot show

    Registry, retrospective, 1996-2012, and confined to single-unit cord blood grafts with one HLA disparity in 57% - not the Interfant-06 donor hierarchy, which prefers unmanipulated marrow from a matched sibling or 10/10 donor. Fusion status is not reported, so the KMT2A-rearranged fraction of the 157 ALL patients is unknown; 'diagnosed before 1 year of age' is not the same as 'transplanted as an infant' and median age at transplant was 1.1 years. Transplant before 2004 was independently associated with worse leukaemia-free survival, so the pooled estimates understate contemporary results. GVHD and relapse are reported separately and never related, so nothing here measures GVL.

    Evidence that argues against this (1)
    • The same study's multivariable model identifies age at diagnosis younger than 3 months as independently associated with worse leukaemia-free survival (p=.012), which is the age band most likely to be KMT2A-rearranged and is exactly the group Interfant-06 selects for transplant. A cord blood series reporting 50% leukaemia-free survival overall is reporting a substantially worse figure in the subgroup the transplant indication targets, and a 40% ALL figure against 66% for AML in the same cohort. Ruggeri A et al. 2017
Connected to