"Dispensable for haematopoiesis" is a summary of a real phenotype: halved white counts, perturbed HSC and progenitor numbers, and failing serial replating
also called Psip1 conditional knockout haematopoiesis · LEDGF therapeutic index
Mice that lose LEDGF in their blood system live normally, which is the main reason people think a LEDGF drug would be safe. They also have about half the normal white blood cell count and blood stem cells that stop making colonies after repeated challenge.
The same thing, for a scientist
The therapeutic-index argument for the LEDGF axis rests on one sentence - LEDGF/p75 is dispensable for haematopoiesis but essential for MLL-rearranged leukaemogenesis - and the sentence is a fair summary of an unfair reading. In the conditional Vav-iCre Psip1 knockout the mice were viable with no evident phenotypic abnormalities, and the authors conclude Psip1 is not essential for HSC maintenance or survival in adult mice. In the same paper they report an **approximately twofold decrease in peripheral white cell counts**, **perturbed numbers of HSCs and progenitor cells**, **impaired colony-forming capacity on serial replating**, **decreased Hox gene expression in haematopoietic stem cells** and **decreased MLL occupancy at MLL target genes**, and note that targeted disruption of HoxA9 produces a similar set of defects. None of this is lethal and none of it was tested under stress: no competitive repopulation, no serial transplantation, no genotoxic or infectious challenge. For a drug given to infants alongside myelosuppressive chemotherapy, the untested half is the half that matters.
What was measured, and what was not
Measured 1: viability; peripheral blood counts; HSC and progenitor immunophenotyping; serial CFU; Hox expression in HSC; MLL ChIP at Hoxa9/Hoxa7; RNA-seq of lineage-negative cells; HoxA9 rescue of the CFU and differentiation defect.
Not measured: competitive repopulation against wild-type marrow, serial transplantation, recovery after myeloablation, response to infection or to a genotoxic agent, ageing, or any human primary cell.
The omission is not a criticism of the paper, which was asking whether steady-state haematopoiesis needs LEDGF and answered it. It is a criticism of how the answer is used. This graph already records the analogous case for menin, where a competitive-repopulation experiment exposed a defect that steady-state analysis had missed.
The HoxA9 axis runs through both sides
The knockout's haematopoietic defects were rescued by HoxA9 overexpression, and the authors note that targeted HoxA9 disruption gives similar defects - lower peripheral lymphocytes, reduced spleen and thymus cellularity, defects in B-cell production, impaired CFU capacity 1.
That is uncomfortable for the therapeutic-index argument, because HOXA9 is also the output the leukaemia depends on. The normal and the malignant requirement are not two separable pathways that happen to share a protein; they are the same axis at different set points. A drug that lowers HOXA9 output enough to matter in the leukaemia is operating on the same lever that produced the knockout's B-cell and progenitor defects - and infant KMT2A-rearranged disease is a B-lineage disease treated in patients whose B-cell compartment is still developing.
Two other normal-tissue findings worth carrying
The germline phenotype is severe. Gene-trap disruption of Psip1 caused perinatal mortality in the majority of homozygotes, with survivors showing low fertility, absent epididymal fat pads, blepharitis, motor and behavioural defects, brachycephaly, small rib cages and homeotic skeletal transformations with incomplete penetrance - a Hox phenotype 2. The lens epithelium, despite the protein's name, was normal.
The brain phenotype is recent and specific. A 2026 conditional Nestin-Cre Psip1 knockout showed early motor deficits, altered MeCP2 condensates in cerebellar neurons, downregulation of synaptic transmission genes and a differentially expressed gene set dominated by PRC2 regulation; the same paper reports brain-region-dependent isoform ratios, with p75 over p52 in cerebellum and p52 over p75 in hippocampus 3. That isoform gradient is the first quantitative evidence this graph holds that the two isoforms are differentially deployed across tissues, which bears directly on concept-psip1-p52-is-not-p75.
| model | Psip1 fl/fl x Vav-iCre |
|---|---|
| summary claim | dispensable for steady-state haematopoiesis |
| rescued by | HoxA9 overexpression |
| germline phenotype | perinatal mortality, homeotic skeletal transformations, motor and behavioural defects |
| isoforms removed | both p75 and p52 |
measured defects
[ "~2-fold decrease in peripheral WBC", "perturbed HSC and progenitor numbers", "impaired serial CFU", "decreased Hox expression in HSC", "decreased MLL occupancy at MLL target genes" ]
not tested
[ "competitive repopulation", "serial transplantation", "myeloablation recovery", "genotoxic challenge", "infection", "ageing", "human cells" ]
- The normal and malignant requirements run through the same HOXA9 axis, so the therapeutic window is a dose window on one lever rather than a selectivity window between two pathways.
- Every published tolerability datum is steady-state; none is under the myelosuppressive and genotoxic stress a paediatric leukaemia patient is actually under.
- The germline phenotype includes homeotic skeletal transformation, and the patients are infants whose skeletons are still patterning - a class of toxicity this graph has already had to take seriously for menin inhibitors and growth plates.
- Loss of LEDGF in the CNS produces motor deficits in mice, and CNS penetration is a design goal for leukaemia drugs because of CNS disease.
- Does Psip1-deficient marrow lose a competitive repopulation assay? The mouse exists, the assay is standard, and the analogous menin experiment changed that target's safety picture.
- Does a Psip1-deficient haematopoietic system recover from myeloablation or tolerate etoposide? A LEDGF agent would be given alongside exactly those insults.
- Is the haematopoietic requirement p75-specific? The allele removes both isoforms, so nothing in the therapeutic-index argument is isoform-resolved on either side.
- Do human HSPCs behave like mouse HSPCs here? The one human datum is that LEDGF-depleted CD34+ cells engraft NSG mice, which is a much weaker statement.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Conditional Vav-iCre Psip1 knockout mice were viable with no evident phenotypic abnormalities and the authors conclude Psip1 is not essential for HSC maintenance or survival in adult mice, while the same animals showed an approximately twofold decrease in peripheral white cell counts, perturbed numbers of HSCs and progenitor cells, impaired colony-forming capacity on serial plating and decreased Hox gene expression in haematopoietic stem cells; the CFU and differentiation defects were rescued by HoxA9 overexpression, and the authors note targeted HoxA9 disruption produces similar haematopoietic defects.
LEDGF/p75 is dispensable for hematopoiesis but essential for MLL-rearranged leukemogenesis. · El Ashkar S, Schwaller J, et al. · Blood 2018What this evidence cannot show
Mouse, steady state only. No competitive repopulation, serial transplantation, myeloablation recovery, infectious or genotoxic challenge. The allele removes both p75 and p52. Read in full via the PMC HTML route after fullTextXML returned HTTP 500 and efetch db=pmc returned no body element.
Evidence that argues against this (1)
- The same paper's leukaemia arm is unambiguous - no disease at 285 days in LEDGF-deficient recipients of MLL-ENL-transduced cells - so the therapeutic-index argument has real content even if the normal-tissue half is understated. El Ashkar S et al. 2018
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Gene-trap disruption of Psip1 caused perinatal mortality in the majority of homozygous mice; survivors showed low fertility, absent epididymal fat pads, a tendency to blepharitis, motor and behavioural defects including hind-limb clenching and reduced grip strength, and craniofacial and skeletal abnormalities including brachycephaly, small rib cages and homeotic skeletal transformations with incomplete penetrance, while the lens epithelium was normal.
Disruption of Ledgf/Psip1 results in perinatal mortality and homeotic skeletal transformations. · Sutherland HG, Newton K, et al. · Mol Cell Biol 2006What this evidence cannot show
The allele is a Psip1-betageo gene-trap fusion that retains chromatin binding and the PWWP and AT-hook domains while lacking the conserved C terminus, so it is a C-terminal truncation rather than a clean null. Incomplete penetrance and mixed background. In Europe PMC (PMC1592893).
Evidence that argues against this (1)
- A whole-animal developmental phenotype from a germline C-terminal truncation is a weak predictor of the toxicity of partial, reversible pharmacological occupancy in an adult - the standing database-and-model scope caveat this project applies to gnomAD constraint applies here too. El Ashkar S et al. 2018
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A conditional Nestin-Cre Psip1 knockout depleting both isoforms in the central nervous system produced early motor deficits, more and smaller MeCP2 condensates in cerebellar neurons and downregulation of synaptic transmission genes whose differentially expressed set was dominated by PRC2 regulation; wild-type brain showed region-dependent isoform ratios, with more p75 than p52 in cerebellum and more p52 than p75 in hippocampus.
Absence of Ledgf in mouse brain affects the Kmt2a/b and polycomb balance, synaptic transmission and motor function. · Debusschere L, Bentea E, et al. · Acta Neuropathol Commun 2026What this evidence cannot show
Mouse brain, not haematopoiesis; both isoforms removed. Cited here for the normal-tissue liability and for the isoform-ratio observation. Open access in Europe PMC (PMC12910792).
Evidence that argues against this (1)
- Neuronal depletion from a Nestin-Cre allele acts during development, so this speaks to a developmental requirement rather than to the consequences of inhibiting LEDGF in an adult brain behind an intact blood-brain barrier. Debusschere L et al. 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The requirement is not absolute even within the haematopoietic system: E2A-HLF, which is independent of HoxA9 for transformation, transformed Psip1-deficient cells with comparable capacity to wild-type, so the LEDGF requirement tracks the HoxA9 axis rather than transformation in general.
LEDGF/p75 is dispensable for hematopoiesis but essential for MLL-rearranged leukemogenesis. · El Ashkar S, Schwaller J, et al. · Blood 2018What this evidence cannot show
A single comparator oncogene in serial CFU assay. It establishes specificity of the LEDGF requirement, not its magnitude.
Evidence that argues against this (1)
- If the LEDGF requirement is really a HoxA9 requirement, then LEDGF is a less specific target than the dossier implies: anything that lowers HOXA9 output would substitute, and this graph already holds several such agents. El Ashkar S et al. 2018
the normal-tissue phenotype behind the therapeutic-index claim
qualifies that entry's therapeutic-index argument
- depends on →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it
the knockout's haematopoietic defects are rescued by HoxA9
the allele removes both isoforms, so the safety argument is not isoform-resolved
opposite sign of the Hox effect in a different cell type