ZNF384 fusions make lineage-ambiguous leukaemia by acting in a haematopoietic stem cell - and the founding lesion is in the HSC itself, unlike Ph-like B-ALL
also called ZNF384 cell of origin · ZNF384 lineage aberrancy · EP300::ZNF384 knockin
The ZNF384 fusion is not made in an already-committed B cell. It is made much earlier, in a blood stem cell, and from there it pushes the cell to make a bit of everything - which is why the resulting leukaemia cannot decide what lineage it is.
The same thing, for a scientist
Two independent lines of evidence place the ZNF384 lesion upstream of lineage commitment. **Genomically**, Alexander et al. purified progenitor, blast and mature populations from a ZNF384-rearranged B/myeloid MPAL patient: TCF3-ZNF384 and the case's other somatic alterations were present in the blasts, in multiple myeloid- and lymphoid-potent progenitor populations, **and in a subset of HSCs**, but not in non-leukaemic T or NK cells - and the paper explicitly contrasts this with Ph-like B-ALL, where founding lesions reach a primitive myelo-lymphoid progenitor **but not HSCs** [E1]. **Experimentally**, Dickerson et al. expressed ZNF384 fusion oncoproteins in mouse and human HSPCs and built an Ep300::Znf384 knockin mouse: the fusions promote haematopoietic expansion, myeloid lineage skewing and self-renewal; in human CD34+ cells TCF3::ZNF384 produced lethal disease in NSG-SGM3 mice at a median latency of 74 days with CD33, CD19 and MPO co-expression **consistent with human B/myeloid MPAL**; in mouse HSPCs a cooperating lesion such as NRAS G12D was required for fully penetrant leukaemia [E2]. The chromatin correlate is occupancy of **predominantly intragenic and enhancer regions with increased histone H3 lysine acetylation**, assayed by ChIP-seq for H3K27ac, H3K18ac and H3K4me1, with deregulation of haematopoietic stem-cell transcription factors [E2]. **The resulting model - lesion in the stem cell, lineage decided downstream - is the same model this graph already records for KMT2A in `mechanism-kmt2a-mpal-mixed-phenotype`, reached independently.**
The cell-of-origin experiment, and the comparison that gives it teeth
Verbatim 1:
"Alterations identified in the unfractionated samples (for example, TCF3-ZNF384 and mutations in MYCN, NTSD2 and DNAH17 in the ZNF384r sample) were identified in the purified blast populations but not in non-leukaemic T or natural killer (NK) cells. Each alteration was also present in multiple haematopoietic progenitor populations with myeloid and lymphoid potential, and a subset of HSCs... these contrast with Ph-like B-ALL, in which founding lesions are detectable in a primitive progenitor with the capacity for myelo-lymphoid differentiation, but not in HSCs. These data support the notion that mutations are acquired in a haematopoietic stem cell that is primed for lineage aberrancy."
The Ph-like contrast is what makes this informative. Without it, 'the lesion is in a progenitor' is nearly vacuous for a leukaemia. With it, there is a documented B-ALL subtype whose founding lesion stops short of the HSC compartment, and ZNF384-rearranged disease does not. The cost is that n = 1 patient was fractionated for the ZNF384 arm.
What the fusion does once it is there - and the partner-dependence nobody predicted
Dickerson et al. 2:
"expression of ZNF384 FO transforms mouse HSPCs, with variable immunophenotype dependent upon fusion partner"
"only EP300::ZNF384 was able to promote formation of serially transplantable leukemia without concomitant genomic alterations"
"Cells expressing TCF3::ZNF384, EP300::ZNF384, or TAF15::ZNF384 and NRAS G12D produced an aggressive leukemia with a median latency of 82 days"
And in human cells:
"NSG-SGM3 mice transplanted with cells expressing TCF3::ZNF384, but not wild-type ZNF384 or the other ZNF384 FO developed lethal disease with a median latency of 74 days" with "expansion of hematopoietic cells expressing CD33, CD19, and MPO, consistent with the immunophenotype of human B/myeloid MPAL"
These two results disagree about which partner is strongest, and the disagreement is the finding. In mouse HSPCs, EP300::ZNF384 alone made serially transplantable leukaemia and the others did not. In human CD34+ cells, TCF3::ZNF384 alone made lethal disease and the others did not, including EP300::ZNF384. Either the species differs, or the readout differs, or the partner effect is not robust. Any claim of the form 'the partner sets the phenotype' has to survive that inversion, and the standard version of the ZNF384 story does not address it. It is recorded as an open question below, and it is one of the tests in ana-q173-the-inverse-architecture-framing-survived-one-of-four-tests.
The chromatin picture, stated at the resolution the source supports
The published summary 2: ZNF384 fusion oncoproteins "occupy a subset of predominantly intragenic/enhancer regions with increased histone 3 lysine acetylation and deregulate expression of hematopoietic stem cell transcription factors", with ChIP-seq for H3K27Ac, H3K18Ac and H3K4me1.
Independently, Alexander et al. ran ChIP-seq of fusion versus wild-type ZNF384 and found 2,298 peaks with new or increased binding of the fusion proteins compared to wild-type ZNF384, and 495 peaks with reduced binding 1 - one of the few places in this literature where the fusion and the wild-type protein were actually separated experimentally, which matters because the FLT3-enhancer work in target-flt3-in-znf384-rearranged-leukaemia could not do that with the reagents available.
H3K18ac is worth flagging for this corpus specifically: data/symbol-collisions.json already carries a row about the compact H3K18ac mark spelling as a search trap, and CREBBP/EP300 are the acetyltransferases responsible for it. A ZNF384 fusion whose 5' partner is EP300 or CREBBP, occupying regions of elevated H3K18ac, is the most mechanistically legible part of this family - and it applies to at most 51% of cases by the Ponte di Legno partner distribution.
| cell of origin | haematopoietic stem cell / multipotent progenitor - the lesion is detectable in a subset of HSCs and in multiple myeloid- and lymphoid-potent progenitors |
|---|---|
| occupancy class | predominantly intragenic and enhancer regions with increased histone H3 lysine acetylation |
models
[ "mouse HSPC retroviral expression of TCF3::ZNF384, EP300::ZNF384, TAF15::ZNF384", "Ep300-exon6::Znf384-V5-HA floxed inverted cDNA minigene knockin into Ep300 intron 5 (mouse)", "human CD34+ HSPC expression, transplanted into NSG-SGM3", "Arf-null mouse pre-B cell cultures" ]
chromatin marks assayed
[ "H3K27ac", "H3K18ac", "H3K4me1" ]
measurements — the numbers behind the chart
[
{
"quantity": "latency_median",
"subject": "lethal disease in NSG-SGM3 from human CD34+ HSPCs expressing TCF3::ZNF384",
"value": 74,
"units": "days",
"n": null,
"universe": "human CD34+ HSPCs transduced with TCF3::ZNF384 and transplanted into NSG-SGM3 mice; wild-type ZNF384 and the other ZNF384 fusions did NOT produce lethal disease in this system",
"null": null,
"note": "no mouse n was extracted; the comparator arms are the other fusions and wild-type ZNF384, which are the right comparators and all negative"
},
{
"quantity": "latency_median",
"subject": "aggressive leukaemia from mouse HSPCs expressing a ZNF384 fusion plus NRAS G12D",
"value": 82,
"units": "days",
"n": null,
"universe": "mouse HSPCs co-expressing TCF3::ZNF384, EP300::ZNF384 or TAF15::ZNF384 with NRAS G12D; NRAS G12D alone gave malignancy with longer latency",
"null": null,
"note": "the three fusions are pooled in this figure, so the number does not discriminate between partners"
},
{
"quantity": "latency_median",
"subject": "ARF-null-derived tumours from TCF3::ZNF384 plus NRAS G12D",
"value": 29,
"units": "days",
"n": null,
"universe": "ARF-null mouse pre-B cell cultures; tumours were of MYELOID lineage despite the pre-B starting material",
"null": 82,
"note": "the null is the ARF-wild-type arm at 82 days. Myeloid tumours from a pre-B culture is itself the lineage-aberrancy phenotype."
},
{
"quantity": "count",
"subject": "ChIP-seq peaks with new or increased binding of ZNF384 fusion proteins relative to wild-type ZNF384",
"value": 2298,
"units": "peaks",
"n": null,
"universe": "ChIP-seq comparison of fusion versus wild-type ZNF384 in Alexander et al.; 495 peaks showed reduced binding",
"null": 495,
"note": "the 495 reduced peaks are the natural null here - the fusion is not simply a hyperactive wild-type protein, it gains about five sites for every one it loses"
},
{
"quantity": "count",
"subject": "proteins interacting with ZNF384 fusion oncoproteins but not wild-type ZNF384",
"value": 74,
"units": "proteins",
"n": null,
"universe": "RIME/mass spectrometry in Dickerson et al.; FDR and replicate number not extracted",
"null": null,
"note": "one of the 74, MED23, was confirmed by co-immunoprecipitation. The rest are unvalidated."
}
]- A lesion resident in the HSC compartment is a lesion that survives lymphoid-directed therapy. If TCF3-ZNF384 is present in a subset of HSCs at diagnosis, then any therapy that clears the blast population - chemotherapy, blinatumomab, CD19 CAR-T - leaves a reservoir in a compartment with myeloid potential, which is the mechanistic version of the lineage-switch risk this graph already records for KMT2A.
- It also constrains what 'MRD-negative' can mean. MRD assays for this subtype are flow-based or fusion-transcript-based on blasts; neither interrogates a rare HSC carrying the same lesion. The reservoir would be invisible by construction.
- And it undermines any intervention aimed at a lineage-specific surface antigen. The target population is defined by lineage, and the lesion sits above lineage.
- Why does EP300::ZNF384 transform mouse HSPCs alone while TCF3::ZNF384 is the only fusion that makes lethal disease from human CD34+ cells? Both results are in the same paper and they invert. Until this is resolved, no statement about which ZNF384 partner is the potent one is safe, and the field's habit of ranking partners by clinical outcome and then assuming a matching potency ranking is unsupported.
- Is the myeloid skewing an artefact of the cytokine-humanised host? NSG-SGM3 mice express human SCF, GM-CSF and IL-3 and are known to bias toward myeloid output. The lineage-aberrancy readout is the paper's headline, and the control that would settle it - the same constructs in a non-cytokine-humanised NSG host - was not reported in what this beat read.
- Does a ZNF384 fusion introduced into a committed B-cell progenitor, rather than an HSPC, produce anything at all? The whole cell-of-origin argument predicts no, or a different disease, and this graph's KMT2A material contains the same untested prediction. A single side-by-side experiment - one fusion, two starting cells - would speak to both families.
- Is the KMT2A protein really in the ZNF384 fusion interactome? It appears in an unvalidated RIME list of 74. If a real contact exists, it is a direct physical bridge between this graph's index fusion family and ZNF384, and it would be the most consequential single fact in this beat. A co-immunoprecipitation would settle it in a week.
- The Ep300::Znf384 knockin places the fusion cDNA in Ep300 intron 5, so the allele produces the fusion but not the reciprocal product and not the native regulatory context. Do ZNF384-rearranged patients express a reciprocal transcript, and does anything depend on it? Alexander et al. report that patients translate BOTH wild-type ZNF384 and the chimaeric protein, which the knockin does model, but the reciprocal is a separate question nobody has asked here.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In a ZNF384-rearranged B/myeloid MPAL patient, TCF3-ZNF384 and co-occurring somatic alterations were detected in purified blasts, in multiple myeloid- and lymphoid-potent progenitor populations and in a subset of HSCs, but not in non-leukaemic T or NK cells; the authors contrast this with Ph-like B-ALL, where founding lesions reach a primitive myelo-lymphoid progenitor but not HSCs. ChIP-seq identified 2,298 peaks with new or increased binding of the fusion proteins relative to wild-type ZNF384 and 495 peaks with reduced binding.
The genetic basis and cell of origin of mixed phenotype acute leukaemia. · Alexander TB, Gu Z, Iacobucci I, Dickerson K, Choi JK, Xu B, et al., Mullighan CG · Nature 2018What this evidence cannot show
FULL TEXT READ via the PMC author-manuscript rendering of PMC6195459 (Europe PMC fullTextXML returns HTTP 500 for this PMCID). **THE FRACTIONATION WAS DONE ON ONE ZNF384-REARRANGED PATIENT.** Two further patients were fractionated but both were WT1-altered T/myeloid MPAL, a different entity. Detection in 'a subset of HSCs' is a sequencing call on a sorted population and is sensitive to sort purity; the paper does not report a contaminating-blast estimate that this beat read. The ChIP-seq fusion-versus-wild-type comparison is a separate experiment from the fractionation and its system is not specified in the text quoted here.
Evidence that argues against this (1)
- One patient cannot establish a cell of origin for a subtype, and the Ph-like contrast is between two studies rather than two arms of one experiment, so a technical difference in sort strategy or sequencing depth could produce it. The more specific tension is inside the same paper: GSEA found ZNF384r B/myeloid MPAL arrested at a MORE MATURE stage of development than other B/myeloid MPAL, with HSC gene sets NEGATIVELY enriched in that comparison - while the same subtype shows enrichment of stem-cell and ETP-ALL programmes relative to B-ALL. A stem-cell origin and a relatively mature arrest are both supported, against different comparators, and neither reading is the whole picture. Alexander TB et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Viral expression of ZNF384 fusion oncoproteins in mouse and human HSPCs, plus an Ep300::Znf384 knockin mouse, showed that the fusions promote haematopoietic expansion, myeloid lineage skewing and self-renewal; in mouse HSPCs a concomitant lesion such as NRAS G12D was required for fully penetrant leukaemia (median latency 82 days for TCF3::ZNF384, EP300::ZNF384 or TAF15::ZNF384 with NRAS G12D), while only EP300::ZNF384 promoted serially transplantable leukaemia without concomitant genomic alterations; in human CD34+ HSPCs, TCF3::ZNF384 but not wild-type ZNF384 or the other fusions produced lethal disease in NSG-SGM3 mice at a median latency of 74 days with CD33, CD19 and MPO expression consistent with human B/myeloid MPAL. Mechanistically the fusions occupy predominantly intragenic and enhancer regions with increased histone H3 lysine acetylation (ChIP-seq for H3K27Ac, H3K18Ac and H3K4me1) and deregulate haematopoietic stem-cell transcription factors.
ZNF384 Fusion Oncoproteins Drive Lineage Aberrancy in Acute Leukemia. · Dickerson KM, Qu C, Gao Q, Iacobucci I, Gu Z, Yoshihara H, Backhaus EA, Chang Y, Janke LJ, Xu B, Wu G, Papachristou EK, D'Santos CS, Roberts KG, Mullighan CG · Blood Cancer Discov 2022What this evidence cannot show
Read from the PMC rendering of PMC9414741 in this session; the quoted sentences are from the abstract, results text and figure text, and the extraction was performed by a page-fetch summariser rather than by reading the whole PDF, so per-experiment n values, mouse numbers and statistics are NOT recorded here and should not be inferred. The knockin is a floxed, inverted Ep300-exon6::Znf384-V5-HA cDNA minigene placed in Ep300 intron 5 - a minigene knockin, not a chromosomal translocation, so the fusion is expressed from the Ep300 locus but the reciprocal product and any genomic-context effects are absent. NSG-SGM3 mice are cytokine-humanised, which is known to favour myeloid output and is therefore a confounder for any myeloid-skewing readout.
Evidence that argues against this (1)
- The two systems in this paper disagree about which partner is the strongest oncoprotein, and the paper does not reconcile them. In mouse HSPCs, 'only EP300::ZNF384 was able to promote formation of serially transplantable leukemia without concomitant genomic alterations'. In human CD34+ cells, lethal disease developed with 'TCF3::ZNF384, but not wild-type ZNF384 or the other ZNF384 FO' - i.e. EP300::ZNF384 did not. Any statement that a particular partner is the potent one is therefore true only within one species and one assay, and the widely repeated claim that the partner determines the phenotype rests on data that invert between systems. Dickerson KM et al. 2022
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ZNF384 fusion oncoproteins, but not wild-type ZNF384, interacted directly or indirectly with 74 proteins including components of the preinitiation complex, the Mediator and cohesin complexes, chromatin-modifying enzymes including ARID1A and KMT2A, RNA polymerase subunits, E3 ubiquitin ligases and transcription factors including TRPS1 and FOXP1; MED23 was a top unique interactor confirmed by co-immunoprecipitation.
ZNF384 Fusion Oncoproteins Drive Lineage Aberrancy in Acute Leukemia. · Dickerson KM, Qu C, Gao Q, Iacobucci I, Gu Z, Yoshihara H, Backhaus EA, Chang Y, Janke LJ, Xu B, Wu G, Papachristou EK, D'Santos CS, Roberts KG, Mullighan CG · Blood Cancer Discov 2022What this evidence cannot show
Read from the PMC rendering; the interactome is from a RIME/mass-spectrometry experiment whose bait, cell system, replicate number and statistical threshold were NOT extracted and are not recorded here. '74 proteins' is therefore a headline count with an unstated FDR. Only MED23 was validated by co-immunoprecipitation. 'Directly or indirectly' is the paper's own wording and means the list contains complex members as well as contacts. KMT2A appearing in this list is notable for this graph but is an unvalidated mass-spectrometry hit, not a demonstrated ZNF384-fusion/KMT2A contact.
Evidence that argues against this (1)
- A RIME list is a proximity list, and the two entries in it that matter most to this graph - KMT2A and ARID1A - are exactly the kind of abundant chromatin protein that appears in affinity proteomics of many baits. This project's own rule 93 records that the more famous the hit, the more likely the instrument is at fault, and the converse applies here: a famous hit in an unvalidated list should be read as a lead. Only MED23 was co-immunoprecipitated. No orthogonal confirmation of a ZNF384-fusion/KMT2A interaction was found by this beat, and a targeted search for one was not run. Dickerson KM et al. 2022
ZNF384 fusion oncoproteins drive haematopoietic expansion, myeloid skewing and self-renewal from HSPCs
TCF3-ZNF384 detectable in a subset of HSCs and in multiple myeloid- and lymphoid-potent progenitors, unlike Ph-like B-ALL
- analogous to →Lineage switch in KMT2A-rearranged leukemia — the same clone changes what kind of blood cell it is
same model in a second fusion family: a stem-cell-resident lesion with lineage decided downstream
- related to →KMT2A / MLL1 — the gene that breaks
KMT2A appears in the ZNF384-fusion RIME interactome of 74 proteins - an UNVALIDATED mass-spectrometry hit, flagged as a lead not a contact
MED23 was the top unique ZNF384-fusion interactor confirmed by co-immunoprecipitation; wild-type ZNF384 did not interact
- part of →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
occupancy of intragenic and enhancer regions with increased H3K27ac, H3K18ac and H3K4me1
uses the validated human CD34+/NSG-SGM3 and mouse HSPC systems and their published positive and negative controls