Oncofusion

start / graph / hyp-fetal-state-is-permissive-for-lymphoid-and-protective-for-myeloid-kmt2a-transformation

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q133-lin28b-fetal-ontogeny

The fetal state is PERMISSIVE for lymphoid KMT2A-fusion transformation and PROTECTIVE against myeloid - which is why LIN28B has opposite signs in the two literatures

also called the lineage-splits-the-sign hypothesis · LIN28B sign reversal · fetal permissiveness is lineage-conditional

Two sets of careful experiments say opposite things about the same fetal gene: in human fetal B-cell precursors it is needed for leukaemia to start, and in mouse fetal blood cells destined to become myeloid leukaemia it actually holds the cancer back. This entry proposed that being a fetal cell helps one lineage become leukaemic and protects the other - and then found a paper showing the fetal state helping a myeloid leukaemia too, which cuts the idea down to a claim about that one gene.

The same thing, for a scientist

Three well-executed studies disagree about the sign of LIN28B in KMT2A-rearranged leukaemia, and they differ systematically by lineage. **Lymphoid arm:** in human fetal liver CD34+ cells, KMT2A::AFF1 fails to transform without LIN28B, and LIN28B knockdown removes the proliferative advantage of established CRISPR-KMT2A::AFF1 leukaemias [E1]. **Myeloid arm:** in mouse MLL-ENL, AML initiation peaks NEONATALLY rather than fetally, and enforced LIN28B - tested explicitly as a candidate accelerator - *suppressed* initiation [E2]; LIN28B acts there by stabilising MYBBP1A mRNA to repress MYB and compromise leukaemia-initiating-cell activity [E3]; and in fully transformed MLL::AF9 AML, enforced LIN28B extended survival with 80% alive beyond four months, induced monocytic differentiation and destroyed clonogenicity [E4]. **The hypothesis is that lineage, not species or artefact, is the variable that reverses the sign** - the fetal programme supplies self-renewal and proliferative competence that a B-lineage fusion needs and cannot generate alone, while in the myeloid lineage the same programme imposes a differentiation-promoting, MYB-restraining brake that a myeloid fusion must escape. **Prior probability 0.15, LOWERED FROM 0.3 BY THE BEAT THAT WROTE THIS ENTRY, in the same session.** Species, partner, perturbation direction and lesion type are all confounded with lineage across these four studies, and no single experiment has varied lineage alone. Worse, a fifth study found while writing this entry contradicts the broad claim outright: a MYELOID KMT2A fusion (MLL-AF9) in mouse FETAL LIVER cells has a pro-leukaemic Id1 dependency, and ID1 is high in human fetal-origin infant MLL-FP AML [E5]. **The fetal state is therefore NOT generally protective against myeloid KMT2A fusions.** What survives is the narrow claim that LIN28B specifically reverses sign between lineages. The title is left unchanged so the correction is legible.

Go deeper
The four studies laid out so the confounds are visible
Ling 2026 [E1]Okeyo-Owuor 2019 [E2]Eldeeb 2023 [E3]Li 2024 [E4]
specieshumanmousemousemouse
lineageB (and NK)myeloidmyeloidmyeloid
partnerAFF1MLLT1 (ENL)MLLT1 (ENL)MLLT1 + MLLT3 (AF9)
lesionendogenous CRISPR translocationtransgenetransgenetransgene
LIN28B perturbationlossgaingaingain, and fetal loss
developmental windowfetalfetal vs neonatal vs adultfetalfetal and established disease
sign of LIN28Brequired for transformationsuppresses initiationsuppresses progressionpromotes differentiation, extends survival

Five variables move together with lineage. That is the honest reason this is a hypothesis with prior 0.3 rather than a conclusion. Any of species, partner, lesion type or perturbation direction could be the real discriminator, and the lineage reading is simply the one that also makes biological sense.

But one internal control inside the myeloid arm is worth its weight. Li et al. deleted Lin28b conditionally in fetal progenitors of MLL::ENL mice and found no acceleration of AML initiation 4. So within the myeloid arm, the gain-of-function suppression is strong and the loss-of-function necessity is absent - LIN28B can protect but is not the thing that does protect. That asymmetry is itself a constraint on any reconciliation, and it means the myeloid arm is a weaker counterweight to Ling et al. than its three papers suggest.

What this predicts, and why the predictions are cheap
  1. A myeloid KMT2A fusion in human fetal liver CD34+ cells should NOT require LIN28B, and might transform better without it. Same cells, same system, same laboratory, different partner. This is the discriminating experiment and it varies lineage while holding species, lesion type and cell source fixed.
  2. A lymphoid KMT2A fusion in a mouse fetal model should require Lin28b, mirroring Ling et al. This varies species while holding lineage.
  3. MYBBP1A should be raised by LIN28B in the human fetal B system too - because the molecular link is not obviously lineage-specific even if its consequence is. If MYBBP1A rises in a cell where LIN28B is permitting transformation, then the link is conserved and the interpretation of MYB repression is what differs between lineages.
  4. Infant KMT2A-rearranged AML should behave more like the mouse myeloid arm than like infant ALL with respect to LIN28B expression. Li et al. already report that most human paediatric AML including MLLr AML do NOT express LIN28B, which is a pass for this prediction on existing data and the strongest current support for the hypothesis.
  5. A within-species, within-cell lineage comparison should be possible in the existing CRISPR fetal-liver system by reading out the myeloid and lymphoid progeny of the same transduced culture.
The alternative explanations, stated so they can win

  • Species. Human and mouse LIN28B differ in expression window and in target repertoire; Li et al. themselves call for human PDX work on the grounds of 'potential species differences in LIN28B expression and function'. If species is the variable, prediction 2 fails and prediction 1 succeeds for the wrong reason.

  • Dose. Every myeloid study used a transgene, and Li et al. state that LIN28B levels in their models exceeded physiological levels. Supraphysiological LIN28B being differentiation-promoting while endogenous fetal LIN28B is permissive is a dose explanation, not a lineage one, and it would reconcile all four papers without invoking lineage at all. This is the alternative this beat rates highest after the lineage one.

  • Perturbation direction. Ling et al. is a loss-of-function study; three of the four myeloid results are gain-of-function. Necessity and sufficiency need not be mirror images, and the one myeloid loss-of-function result (Li 2024's fetal deletion) found nothing, which is compatible with LIN28B being neither required nor protective in myeloid.

  • Partner. AFF1 versus MLLT1/MLLT3 is a difference in elongation-complex composition that this project spends most of its time on. It cannot be dismissed, and prediction 1 confounds partner with lineage because the myeloid KMT2A fusions are the non-AFF1 ones.

  • A fetal compartment result in the WRONG direction, found by this beat and recorded against itself. Man et al. put MLL-AF9 - a MYELOID KMT2A fusion - into mouse fetal liver cells and found Id1 to be pro-leukaemic there and restraining in bone marrow, with ID1 high in human fetal-origin infant MLL-FP AML 5. If the fetal state were generally protective against myeloid KMT2A fusions, that result should not exist. So the broad title of this hypothesis is already contradicted, and only the narrow form survives: LIN28B specifically reverses sign between lineages. The entry is kept under its original title with this correction in place rather than renamed, so that a reader sees what was proposed as well as what happened.
Facts and figures
statementThe fetal haematopoietic state is permissive for KMT2A-fusion transformation in the B-lymphoid lineage and protective against it in the myeloid lineage, and LIN28B is the factor through which this is most visible - which is why the human lymphoid literature reports it as required for transformation and the mouse myeloid literature reports it as tumour-suppressive.
originliterature-gap
prior probability0.15
what would make this boringIf dose explains everything - supraphysiological LIN28B being differentiation-promoting and endogenous fetal LIN28B being permissive - then there is no lineage story, and the second falsifier is the one to run first because it is also the cheapest to interpret.
shown for fusionKMT2A::AFF1 (human, lymphoid, permissive) versus KMT2A::MLLT1 and KMT2A::MLLT3 (mouse, myeloid, protective). No study has compared two partners in the same cells.
predictions
[
  "A myeloid KMT2A fusion (KMT2A::MLLT3 or KMT2A::MLLT1) engineered into human fetal liver CD34+ cells will NOT require LIN28B for transformation, and may transform more readily in its absence.",
  "MYBBP1A protein will rise with LIN28B in the human fetal B-lineage CRISPR-KMT2A::AFF1 system, showing the molecular link is conserved while its phenotypic consequence is not.",
  "Infant KMT2A-rearranged AML will show lower LIN28B expression than infant KMT2A-rearranged ALL in matched diagnostic material. PARTIALLY SUPPORTED ALREADY: Li et al. report most human paediatric AML including MLLr AML do not express LIN28B.",
  "In the existing CRISPR fetal-liver system, the myeloid and lymphoid progeny of one LIN28B-null culture will show divergent transformation outcomes with the same fusion.",
  "Restoring LIN28B in an established human infant KMT2A::AFF1 leukaemia will NOT induce differentiation, unlike the monocytic differentiation it induces in mouse MLL::AF9 AML."
]
experiments
[
  "exp-kmt2a-mllt3-versus-aff1-in-lin28b-null-human-fetal-liver",
  "exp-lin28b-null-fetal-progenitor-second-oncogene-specificity-control",
  "exp-let-7-restoration-versus-lin28b-loss-fetal-liver-kmt2a-aff1"
]
analyses
[]
falsifiers
[
  "KMT2A::MLLT3 engineered into human fetal liver CD34+ cells FAILS to transform in LIN28B-null cells at the same rate as KMT2A::AFF1 does - i.e. the transformation frequency in LIN28B-null cells is within 2-fold of the KMT2A::AFF1 LIN28B-null frequency, with both at least 5-fold below their respective LIN28B-intact controls, in at least 3 independent fetal donors. That would make the requirement lineage-INDEPENDENT and kill the hypothesis, leaving a general fetal-competence explanation.",
  "Enforced LIN28B at ENDOGENOUS FETAL LEVELS (single-copy knock-in, not a strong transgene) in mouse fetal MLL-ENL progenitors fails to suppress AML initiation - median survival within 20% of controls and no significant difference by log-rank at n>=15 per arm. That would relocate the myeloid result to a dose artefact rather than a lineage property, and the hypothesis would be answering a question that does not exist.",
  "LIN28B is found to be expressed at comparable levels in infant KMT2A-rearranged AML and infant KMT2A-rearranged ALL in matched human diagnostic material (>=20 cases per group, medians within 2-fold, no significant difference). That removes the only currently available human observation supporting a lineage split.",
  "Restoring LIN28B in an established human KMT2A::AFF1 infant ALL PDX induces differentiation and extends survival, as it does in mouse MLL::AF9 AML (>=30% survival extension, log-rank p<0.05). That would show the differentiation-promoting activity is lineage-general and the lymphoid requirement is about initiation only, not about the sign of LIN28B.",
  "ALREADY PARTLY RESOLVED IN THE WRONG DIRECTION FOR THIS HYPOTHESIS, 2026-09-23, by the beat that wrote it: a MYELOID KMT2A fusion (MLL-AF9) in mouse FETAL LIVER cells shows a PRO-leukaemic Id1 dependency, and ID1 is high in human fetal-origin infant MLL-FP AML (PMID 26944543). The broad form - 'the fetal state is protective for myeloid KMT2A fusions' - is therefore already contradicted by a 2016 Blood paper. What survives is the narrow form: LIN28B SPECIFICALLY has opposite signs in the two lineages. The falsifier as originally intended still stands for the narrow form and reads: enforced LIN28B at endogenous fetal levels in mouse fetal MLL-ENL progenitors fails to suppress AML initiation (median survival within 20% of controls, no log-rank difference at n>=15 per arm)."
]
How this could bite us back
  • If the hypothesis is right, a LIN28B inhibitor is a lymphoid-leukaemia drug and a myeloid-leukaemia hazard. KMT2A-rearranged disease switches lineage under pressure - documented under blinatumomab and CAR-T - so the same patient could move from the arm where the drug helps to the arm where it harms, during treatment.
  • If the hypothesis is wrong and dose is the answer, then the protective myeloid effect is an artefact of overexpression and the only real biology is the lymphoid requirement - which would make LIN28B a cleaner target and this entry an over-cautious one. Getting this wrong in the cautious direction costs a therapeutic opportunity.
  • Either way, the lineage-conditional framing means no single biomarker of LIN28B status suffices; a clinical programme would need lineage assignment at the moment of dosing.
Still open
  1. Has anyone engineered a MYELOID KMT2A fusion into human fetal liver CD34+ cells with the endogenous-translocation approach? If that system exists, the discriminating experiment is a partner swap away.
  2. Is there human infant KMT2A-rearranged AML material with LIN28B expression measured alongside matched infant ALL? The prediction is already partly supported by re-analysed data and could be closed properly with existing cohorts.
  3. Does the lineage split appear in the normal biology too - is the fetal programme growth-promoting in B progenitors and differentiation-promoting in myeloid progenitors, independent of any fusion? If so the fusion is a passenger in the argument and the hypothesis simplifies considerably.
  4. Why does Zhou et al. 2017 give a third sign in AML lines? Either the lines are unrepresentative, as Li et al.'s patient-expression data suggest, or the sign depends on whether the cell is transformed and on what.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Human fetal liver CD34+ cells fail to transform upon induction of the KMT2A::AFF1 translocation in the absence of LIN28B, and LIN28B-expressing CRISPR-KMT2A::AFF1 leukaemias were more proliferative in vitro and in vivo with the advantage lost on knockdown.

    The fetal specific gene LIN28B is essential for human fetal B-lymphopoiesis and initiation of KMT2A::AFF1 infant leukemia. · Ling RE, Welsh AM, Hamer L, Jackson T, Neil E, Elliott N, et al. · Blood 2026
    PMID 42213620DOI
    What this evidence cannot show

    ABSTRACT ONLY; paywalled and not in PMC or Europe PMC, verified 2026-09-23. No effect size, no n, no controls read. See exp-lin28b-loss-blocks-kmt2a-aff1-transformation-human-fetal-liver for the full read-depth accounting.

    Evidence that argues against this (1)
    • Three mouse myeloid studies give LIN28B the opposite sign. Leukemia 2024
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In inducible transgenic MLL-ENL mice, AML initiation efficiency changed across development and peaked NEONATALLY rather than fetally; LIN28B, tested as a candidate accelerator precisely because fetal/neonatal progenitors were so competent, suppressed AML initiation rather than accelerating it, which the authors propose explains why MLL rearrangements arise before birth while transformation usually follows it.

    The efficiency of murine MLL-ENL-driven leukemia initiation changes with age and peaks during neonatal development. · Okeyo-Owuor T, Li Y, Patel RM, Yang W, Casey EB, Cluster AS, Porter SN, Bryder D, Magee JA · Blood Adv 2019
    PMID 31405949DOI
    What this evidence cannot show

    Abstract read; full text not opened. Mouse, transgene, myeloid. The abstract also notes that transplantation stress mitigated the developmental barrier, i.e. the age effect is not robust to handling - which weakens the claim that the fetal window is intrinsically protective.

    Evidence that argues against this (1)
    • Transplantation stress mitigated the developmental barrier in this very study, so the fetal-versus-neonatal difference is not robust to handling. A hypothesis built on the fetal window being intrinsically protective in myeloid cells is building on an effect its own source reports as fragile. Leukemia 2024
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a dual-transgenic MLL-ENL/LIN28B mouse, LIN28B impedes MLL-rearranged AML progression through compromised leukaemia-initiating-cell activity and suppression of MYB signalling, mechanistically by binding MYBBP1A mRNA to elevate the MYB co-repressor MYBBP1A; the authors propose developmentally restricted LIN28B as a layer of protection against MYB-dependent AML, and explicitly describe the influence of prenatal signals on leukaemogenesis as controversial.

    A fetal tumor suppressor axis abrogates MLL-fusion-driven acute myeloid leukemia. · Eldeeb M, Yuan O, Guzzi N, Thi Ngoc PC, Konturek-Ciesla A, Kristiansen TA, Muthukumar S, Magee J, Bellodi C, Yuan J, Bryder D · Cell Rep 2023
    PMID 36763502DOI
    What this evidence cannot show

    Abstract read; full text not opened. Mouse, transgene, myeloid, supraphysiological LIN28B per the companion paper.

    Evidence that argues against this (1)
    • Supraphysiological transgenic LIN28B is the dose explanation this hypothesis names as its strongest rival, and this study is one of the ones exposed to it. If dose rather than lineage explains the myeloid result, this evidence item supports the rival hypothesis, not this one. Leukemia 2024
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Conditional Lin28b deletion in fetal progenitors of Tet-OFF MLL::ENL mice did not accelerate AML initiation or shorten survival, so LIN28B is not the effector of fetal leukaemia suppression; separately, enforced LIN28B in established MLL::AF9 AML extended survival with 80% of secondary recipients alive beyond 4 months, caused partial monocytic differentiation by day 6 and severely depleted colony-forming potential in 5 independent AMLs without increasing apoptosis; and most human paediatric AML including MLLr AML do not express LIN28B while expressing let-7 family microRNAs highly.

    LIN28B promotes differentiation of fully transformed AML cells but is dispensable for fetal leukemia suppression. · Li Y, Mendoza-Castrejon J, Patel RM, Casey EB, Denby E, Bryder D, Magee JA · Leukemia 2024
    PMID 38321106DOI
    What this evidence cannot show

    FULL TEXT READ (PMC10912017). A Letter, so compact. Mouse, transgenes, myeloid. The authors flag supraphysiological LIN28B and a ~6-month experimental endpoint. The human AML expression statement is a re-analysis of external data. The fetal-deletion null result is the internal control that makes the myeloid arm weaker than its paper count suggests.

    Evidence that argues against this (1)
    • A fourth paper gives LIN28B a third sign in myeloid disease: knockdown in AML lines arrests the cell cycle and blocks colony formation via let-7a de-repression, and LIN28B overexpression increases tumorigenicity - LIN28B as myeloid oncogene. The lineage hypothesis does not explain this one, and it has to be accounted for or set aside on design grounds. J Hematol Oncol 2017
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF9 expressed in mouse FETAL LIVER cells versus bone marrow cells gives opposite Id1 dependencies: Id1 loss prolongs survival in fetal-liver recipients but accelerates leukaemogenesis in bone-marrow recipients, both rescued by Cdkn1a deletion; and ID1 is elevated in human fetal-origin (infant) MLL-fusion-protein AML and reduced in postnatal MLL-FP AML.

    Differential role of Id1 in MLL-AF9-driven leukemia based on cell of origin. · Man N, Sun XJ, Tan Y, García-Cao M, Liu F, Cheng G, et al. · Blood 2016
    PMID 26944543DOI
    What this evidence cannot show

    Abstract read; full text not opened. Mouse, retroviral MLL-AF9, myeloid. THIS IS CITED AGAINST THE HYPOTHESIS, NOT FOR IT - see the counter_evidence note.

    Evidence that argues against this (1)
    • THIS IS A PARTIAL REFUTATION OF THIS HYPOTHESIS AND IS RECORDED AS SUCH. The hypothesis says the fetal state is PROTECTIVE for myeloid KMT2A fusions. Here a MYELOID KMT2A fusion (MLL-AF9 / KMT2A::MLLT3) in FETAL liver cells has a pro-leukaemic Id1 dependency - the fetal compartment is supplying something the fusion needs, exactly as in the lymphoid AFF1 case. So 'the fetal state is protective in myeloid' is too strong: it holds for LIN28B specifically and not for the fetal compartment in general. The hypothesis survives only in the narrower form 'LIN28B specifically is protective in myeloid and permissive in lymphoid', which is a much weaker claim than the one its title makes. Blood 2016
Connected to