The result that runs the wrong way: in Psip1-null cells Hoxa and Hoxd expression goes UP, because Psip1 also recruits Bmi1 and Ctbp1
also called Psip1 restrains Hox · LEDGF polycomb recruitment
LEDGF is usually described as helping switch Hox genes on. In one careful study, cells without LEDGF switched Hox genes on more, not less, because LEDGF also brings in the machinery that keeps them off.
The same thing, for a scientist
Psip1/Ledgf p75 regulates homeotic genes by recruiting **both** the trithorax-group proteins Mll1/Mll2 and the polycomb-group protein **Bmi1**, together with the co-repressor **Ctbp1**. In Psip1-null cells the binding of Mll1/2, Bmi1 and Ctbp1 at Hox loci is all abrogated and **Hoxa and Hoxd mRNA expression is increased**. That is the opposite sign to the leukaemia field's summary - LEDGF tethers MLL to HOXA9, therefore removing LEDGF lowers HOXA9 - and it is not a contradiction that can be waved away as a different cell type, because the Psip1 conditional knockout in haematopoietic cells reports **decreased** Hox expression in HSC in the same year-range. Both results are real; they were obtained in different cells with different Hox baselines, and neither has been reconciled with the other. The reconciliation matters for a drug, because a LEDGF-directed agent is being proposed on the assumption that it lowers HOXA9. In a cell where LEDGF's polycomb-recruiting arm dominates, it would raise it.
The two results side by side
| Study | Cells | Psip1 status | Hox expression |
|---|---|---|---|
| Pradeepa 2014 1 | Psip1-/- (mouse embryonic fibroblast-type) | null | Hoxa and Hoxd mRNA increased |
| El Ashkar 2018 2 | haematopoietic stem cells, Vav-iCre conditional | null | Hox genes decreased |
Both papers report loss of Mll binding at Hox loci on Psip1 loss. They differ on what happens to transcription, and the difference is explained by what else leaves: in the first, Bmi1 and Ctbp1 leave too, so repression is lifted along with activation and the net is derepression.
A third data point points the same direction as the first: a 2026 Nestin-Cre Psip1 knockout found that the differentially expressed genes in cerebellum were dominated by PRC2 regulation, and the authors frame the result as LEDGF operating on a Trithorax-Polycomb balance rather than on Trithorax alone 3.
Why a balance model changes the drug prediction
If LEDGF is a tether for activation, removing it lowers HOXA9 and kills the leukaemia. If LEDGF is a fulcrum for a trithorax-polycomb balance, removing it changes the set point in a direction that depends on which arm dominates at that locus in that cell, and the answer is not knowable from the model system used.
This is not an abstract worry. The direction of the effect on HOXA9 is the entire mechanistic rationale for the axis. The one KMT2A-rearranged measurement - HoxA9 target genes down in Psip1-depleted cells after a week in culture, phenotype rescued by HoxA9 overexpression 2 - supports the activation model in the cell type that matters. But it is one experiment, in mouse, and it sits against two independent results in other tissues that show the opposite or show polycomb dominance.
A LEDGF-directed agent should therefore carry a HOXA9 transcript readout in every model it is tested in, and an increase should be treated as a live possible outcome rather than as an assay failure.
The bridge to menin
There is an obvious structural link nobody in this graph has drawn: menin's own pocket binds both MLL and JunD, with opposite effects on transcription. Menin is therefore already established as a bivalent switch rather than a pure activator cofactor, and the graph holds JunD liberation under menin inhibition as a resistance mechanism. A LEDGF that recruits both Mll and Bmi1 is the same shape of object one step further out in the complex. If both members of the clamp are bivalent, then 'disrupting the clamp' has two possible transcriptional outcomes at every locus, and the field's one-directional framing of this module is an oversimplification at both ends.
| direction in non haematopoietic cells | Hoxa and Hoxd mRNA UP on Psip1 loss |
|---|---|
| direction in HSC | Hox expression DOWN on Psip1 loss |
| unreconciled | true |
| implication | a LEDGF-directed agent must carry a HOXA9 transcript readout, and an increase is a possible outcome |
factors lost at hox loci in both
[ "Mll1/Mll2" ]
additional factors lost in psip1 null fibroblast type
[ "Bmi1", "Ctbp1" ]
- If LEDGF restrains as well as activates Hox genes, a LEDGF-directed agent could raise HOXA9 in some cellular contexts, which is the precise opposite of the intended effect.
- Loss of polycomb recruitment at developmental loci is a derepression event with no obvious ceiling, and the graph already records PRC1 loss as a menin-inhibitor resistance route.
- The one KMT2A-rearranged measurement supporting the intended direction did not actually show HoxA9 itself falling in the relevant fraction, only its targets.
- Does LEDGF recruit polycomb at HOXA9 in a KMT2A-rearranged blast? Nobody has done a BMI1 or CTBP1 ChIP in a LEDGF-depleted leukaemic cell, and the answer decides the sign of the drug effect.
- Why do the two published directions differ - cell type, Hox baseline, or isoform? All three are confounded in the existing comparison.
- If LEDGF is a trithorax-polycomb fulcrum, is there a locus class in KMT2A-rearranged cells where LEDGF loss DERE-presses a tumour suppressor? That would be a resistance mechanism created by the drug.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Psip1/Ledgf p75 regulates homeotic genes by recruiting both Mll complexes and the polycomb group protein Bmi1; in Psip1-/- cells the binding of Mll1/2, Bmi1 and the co-repressor Ctbp1 at Hox loci is abrogated and Hoxa and Hoxd mRNA expression is increased.
Psip1/Ledgf p75 restrains Hox gene expression by recruiting both trithorax and polycomb group proteins. · Pradeepa MM, Grimes GR, et al. · Nucleic Acids Res 2014What this evidence cannot show
Non-haematopoietic cells with a different Hox baseline from a leukaemic blast; the Psip1-null allele removes both isoforms. Open access in Europe PMC (PMC4132756).
Evidence that argues against this (1)
- In haematopoietic stem cells from a conditional Psip1 knockout, Hox gene expression is DECREASED, and the CFU and differentiation defects are rescued by HoxA9 overexpression - the opposite direction in the cell type that matters for this disease. El Ashkar S et al. 2018
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In the haematopoietic conditional Psip1 knockout, Hox gene expression in haematopoietic stem cells is decreased, MLL occupancy at Hoxa9 and Hoxa7 promoter regions is reduced, HoxA9 targets are downregulated in Psip1-depleted cells after a week in culture, and the observed phenotypes are rescued by HoxA9 overexpression.
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 haematopoietic cells; the authors note they did not observe decreased HoxA9 expression in the lineage-negative fraction itself, in contrast to other HoxA cluster members, so even within this paper the HoxA9 result is indirect. Read via the PMC HTML route.
Evidence that argues against this (1)
- That HoxA9 itself was not decreased in the lineage-negative fraction, while its targets were, means the simple 'LEDGF loss lowers HOXA9' statement is not what the data show even in the haematopoietic system. El Ashkar S et al. 2018
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In a conditional Nestin-Cre Psip1 knockout, transcription factor network analysis showed the differentially expressed genes were mainly regulated by PRC2, and the authors frame LEDGF/p75 as acting on the balance between Trithorax and Polycomb rather than on Trithorax alone.
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; network inference rather than direct PRC2 occupancy measurement; both isoforms removed. Open access (PMC12910792).
Evidence that argues against this (1)
- A transcription-factor enrichment on a differential expression list is an inference about upstream regulators, not a measurement of polycomb occupancy, and PRC2 targets are enriched in any list of developmentally regulated genes. Pradeepa MM et al. 2014
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Bivalence is already established for the other member of the clamp: the same pocket in menin binds both MLL and JunD, with opposite effects on transcription.
The same pocket in menin binds both MLL and JUND but has opposite effects on transcription. · Huang J, Gurung B, et al. · Nature 2012What this evidence cannot show
Cited for the precedent of a bivalent cofactor within this module, not as evidence about LEDGF.
Evidence that argues against this (1)
- Menin's bivalence is between two different partners at one pocket; LEDGF's proposed bivalence is between two complexes at one locus. The analogy is suggestive rather than mechanistically parallel. Pradeepa MM et al. 2014
direction of the effect is cell-type dependent and unresolved
this protein recruits polycomb as well as trithorax at Hox loci
opposite sign of the Hox effect in a different cell type
both clamp members are bivalent transcriptional cofactors
third system implicating a polycomb arm of LEDGF function