BCOR, not CBX8, is the AHD contact MLL-AF9 cannot lose - and the same two partners bind ENL far more tightly than AF9
also called BCOR MLL-AF9 · AF9 E531R · CBX8 MLL-AF9 · 2N4Q · 6B7G
Four different proteins compete for the one small patch of AF9 that the cancer protein keeps, and they cannot bind at the same time. Break just one of those contacts - the one with BCOR - and the leukaemia never forms in mice, even though the cells grow faster in a dish.
The same thing, for a scientist
The retained AF9 AHD binds **AF4, DOT1L, BCOR and CBX8 at the same site**, so all four contacts are mutually exclusive; the four complexes superpose on one another with backbone RMSDs of 1.4-1.9 A over the structured AF9 residues [E1]. Two of the structures were determined in this work: **CBX8 (aa 326-349) with the AF9 AHD, PDB 2N4Q**, and **BCOR (aa 1176-1207) with the AF9 AHD, PDB 6B7G** [E1]. The functional result is the sharpest separation-of-function yet reported on this surface. **AF9 E531R** was designed from chemical-shift mapping to disrupt an electrostatic contact with BCOR's C-terminal residues; it drastically reduces BCOR affinity with little effect on AF4, DOT1L or CBX8 binding, and it disrupts BCOR co-immunoprecipitation in cells. Introduced into MLL-AF9 it **does not impair colony formation, and actually increases in vitro proliferation, yet completely abrogates the ability of MLL-AF9 to cause leukaemia in vivo** [E1]. Mechanistically, EYA1 is the most downregulated direct target on loss of BCOR binding, SIX1 falls with it, c-Myc protein falls, and forced EYA1 expression partially rescues the leukaemia-initiating cell population [E1]. Two things follow that matter beyond this contact. **The in-vitro readout is wrong here**: a mutant that grows better in a dish cannot make leukaemia in a mouse. And **the AHD is not paralogue-equivalent**: BCOR binds ENL 9-fold more tightly than AF9, and CBX8 - too weak to saturate against AF9 - binds ENL with a Kd of 112 nM, while AF4 and DOT1L affinities are comparable between the two [E1].
One site, four partners, mutually exclusive
All four characterised AHD partners bind the same surface on AF9, so the fusion cannot hold two of them at once 1:
| partner | motif on the partner | affinity for AF9 AHD |
|---|---|---|
| AF4/AFF1 | LXLXXL-type | comparable to ENL |
| DOT1L | three sites; the strongest is aa 877-900 | comparable to ENL |
| BCOR | aa 1176-1225 | Kd 18 +/- 2 nM |
| CBX8 | aa 326-349 | too weak to saturate; A335V variant Kd 12 +/- 1 nM |
Backbone RMSD of the BCOR-AF9 complex against the AF4-, DOT1L- and CBX8-AF9 complexes: 1.9, 1.4 and 1.6 A over structured AF9 residues 1. All four are mixed alpha-beta complexes in which a partner beta strand of alternating hydrophobic residues completes an antiparallel beta sheet with an AF9 beta hairpin, with three AF9 helices folding around it.
This is what the fusion's retained half actually is: a single competitively shared docking site. Any statement of the form 'the fusion recruits DOT1L and AF4 and BCOR and CBX8' is a statement about a population of fusion molecules, not about one molecule.
AF9 E531R: the cleanest separation-of-function on this surface, and what it shows
Designing a mutation that hits one of four partners at a shared site is hard, and the authors say so. Their route was to compare NMR spectra of AF9 bound to a short BCOR peptide (1176-1207) against a longer one (1176-1228); the largest chemical-shift difference was AF9 E531, and because BCOR's C-terminal extension carries several positive charges they built the charge-reversal E531R 1.
Results 1:
- E531R greatly weakens binding of the long BCOR peptide and disrupts BCOR co-IP with full-length AF9 in cells.
- Binding of AF4, DOT1L sites 2 and 3, and CBX8 A335V changes little to none - and the residual changes are smaller than those of the AF9 D544R mutant that retains colony-forming ability, which is the internal benchmark for 'not biologically relevant'.
- MLL-AF9 E531R: colony formation unaffected, in vitro proliferation increased, partial differentiation - and no leukaemia in vivo.
The last line is the finding. A mutant that looks better on the standard in vitro assay is dead in the animal. Every entry in this graph that scores an AF9 AHD mutant by serial replating alone should be read against it.
Downstream: RNA-seq puts EYA1 as the most downregulated direct target on loss of BCOR binding, with SIX1 following and c-Myc protein reduced; forced EYA1 expression partially rescues the leukaemia-initiating cell population 1. EYA1 is also the gene whose expression falls when the ENL YEATS domain is deleted from MLL-ENL 4 - so two structurally unrelated lesions, on two different partners, converge on the same target gene.
The correction this paper makes to the CBX8 story, and it is sourced
The 2011 report that CBX8 is essential for MLL-AF9-induced leukemogenesis showed that CBX8 ablation and point mutations in MLL-AF9 that disrupt the CBX8 interaction both abolish HOX gene upregulation and leukaemic transformation 2.
The structural work says those point mutations were not specific 1:
- the AF9 mutations used in that study 'also affect AF4 and DOT1L binding';
- AF9 T542A causes a significant reduction in BCOR binding, which the authors state makes the results difficult to interpret and 'strongly suggest[s] that the observed effects derived from reduced recruitment of binding partners other than CBX8'.
And tested with mutants that do separate the contacts: CBX8 R336D, which weakens CBX8-AF9 binding, has no effect on colony formation, while CBX8 A335V, which strengthens it, inhibits MLL-AF9 colony formation 1.
The honest position is therefore: CBX8 the protein may still be required for MLL-AF9 leukaemia - the knockout data stand - but the direct CBX8-AF9 contact is not what that requirement runs through. This graph should not carry 'CBX8 binding to MLL-AF9 is essential' as an unqualified claim.
The paralogue asymmetry nobody has used
Measured side by side in one assay 1:
| partner | AF9 versus ENL |
|---|---|
| AF4 | comparable |
| DOT1L | comparable |
| BCOR | 9-fold tighter to ENL |
| CBX8 | too weak to saturate on AF9; Kd 112 +/- 13 nM on ENL |
So the AHD is not a paralogue-equivalent module. For the two partners the field treats as secondary, ENL is much the better binder; for the two it treats as primary, the two are the same.
That cuts directly against a convenient assumption in this graph's transfer rule. It is not only that MLL-ENL has a YEATS domain and MLL-AF9 does not; their shared docking module also has a different partner preference, and the difference is largest for the contact that MLL-AF9 cannot survive losing. A BCOR-directed or CBX8-directed strategy would therefore be expected to behave differently in t(9;11) and t(11;19) for reasons that have nothing to do with the reader.
| fusion | fusion-kmt2a-mllt3 |
|---|---|
| surface | AF9 ANC1 homology domain (AHD), the retained partner module |
| mutually exclusive | true |
| rescue | forced EYA1 expression partially restores the leukaemia-initiating cell population |
| in vitro in vivo discordance | MLL-AF9 E531R proliferates better in vitro and cannot make leukaemia in vivo |
| correction to the literature | AF9 point mutants previously used as CBX8-selective are not selective; the direct CBX8-AF9 contact does not behave as a requirement |
partners
[ "AFF1/AF4", "DOT1L", "BCOR", "CBX8" ]
structures
{
"2N4Q": "CBX8 (aa 326-349) with the AF9 AHD, solution NMR",
"6B7G": "BCOR (aa 1176-1207) with the AF9 AHD, solution NMR",
"2MV7": "DOT1L site 3 (aa 877-900) with the AF9 AHD",
"2LM0": "AF4 peptide with the AF9 AHD"
}rmsd to bcor complex angstrom — the numbers behind the chart
{
"AF4": 1.9,
"DOT1L": 1.4,
"CBX8": 1.6
}affinities NM
{
"BCOR_1176_1225_to_AF9": 18,
"CBX8_A335V_to_AF9": 12,
"CBX8_WT_to_AF9": "too weak to saturate",
"CBX8_to_ENL": 112
}paralogue preference
{
"AF4": "comparable AF9 vs ENL",
"DOT1L": "comparable AF9 vs ENL",
"BCOR": "9-fold tighter to ENL",
"CBX8": "unmeasurable on AF9, 112 nM on ENL"
}separating mutants
[
{
"mutant": "AF9 E531R",
"target_contact": "BCOR",
"in_vitro": "colony formation unaffected; proliferation increased; partial differentiation",
"in_vivo": "completely abrogates MLL-AF9 leukaemia"
},
{
"mutant": "CBX8 R336D",
"target_contact": "CBX8-AF9, weakened",
"in_vitro": "no effect on MLL-AF9 colony formation"
},
{
"mutant": "CBX8 A335V",
"target_contact": "CBX8-AF9, strengthened",
"in_vitro": "inhibits MLL-AF9 colony formation"
},
{
"mutant": "AF9 T542A",
"target_contact": "used historically as CBX8-selective",
"note": "significantly reduces BCOR binding, so earlier CBX8-attributed phenotypes are confounded"
}
]downstream
[ "EYA1 (most downregulated direct target)", "SIX1", "c-Myc protein", "SIX-regulated gene programme" ]
- A drug that blocks the AF9 AHD blocks all four contacts at once, including the BCOR contact whose loss increases proliferation in vitro - so a compound could plausibly accelerate blast proliferation while impairing leukaemia initiation, and an in vitro screen would score that as failure or as toxicity.
- BCOR and CBX8 bind ENL far more tightly than AF9, so a ligand optimised against the AF9 AHD will engage wild-type ENL preferentially, and ENL is required for AML maintenance.
- The separation-of-function result means the field's standard colony-formation assay has at least one documented false negative on this surface; any AHD-directed programme that screens on replating alone is using an assay with a known blind spot.
- Does AF9 E531R abolish leukaemia because BCOR is lost, or because the other three partners now occupy the site more? The competitive equilibrium after the mutation has not been measured.
- Is the BCOR requirement partner-specific? BCOR binds ENL 9-fold more tightly, and no equivalent E531R-style experiment has been done on MLL-ENL.
- Does the internal-promoter YEATS-devoid AF9 isoform contribute a third AHD-bearing competitor in myeloid blasts, and is it present at levels that matter?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AF4, DOT1L, BCOR and CBX8 all bind the same site on the AF9 AHD and their binding is mutually exclusive, with the CBX8(326-349)-AF9 structure deposited as PDB 2N4Q and the BCOR(1176-1207)-AF9 structure as PDB 6B7G and backbone RMSDs of 1.9, 1.4 and 1.6 A between the BCOR complex and the AF4, DOT1L and CBX8 complexes; BCOR(1176-1225) binds AF9 with a Kd of 18 +/- 2 nM while CBX8 binds too weakly to saturate and its A335V variant binds at 12 +/- 1 nM; BCOR binds ENL 9-fold more tightly than AF9 and CBX8 binds ENL with a Kd of 112 +/- 13 nM while AF4 and DOT1L affinities are comparable between the paralogues; the structure-guided AF9 E531R mutation selectively disrupts BCOR binding and, in MLL-AF9, leaves colony formation intact and increases in vitro proliferation while completely abrogating leukaemia in vivo, with EYA1 the most downregulated direct target, SIX1 and c-Myc protein reduced, and forced EYA1 partially rescuing the leukaemia-initiating cell population; CBX8 R336D, which weakens CBX8-AF9 binding, has no effect on colony formation.
BCOR Binding to MLL-AF9 Is Essential for Leukemia via Altered EYA1, SIX, and MYC Activity. · Schmidt CR, Achille NJ, Kuntimaddi A, et al. · Blood Cancer Discov 2020What this evidence cannot show
Read in full text from the PMC HTML page (PMC7497807, 74,855 characters of stripped text) on 2026-09-20. All structures are isolated partner peptides with an isolated AF9 AHD construct - no KMT2A sequence, no chromatin, no fusion protein. The functional work is retroviral MLL-AF9 in murine bone marrow progenitors with transplantation, so species, lineage and expression-level imports all apply. E531R is selective but not perfectly so; the authors argue the residual effects on AF4, DOT1L and CBX8 are smaller than those of a mutant known to retain function, which is an internal benchmark rather than an independent one.
Evidence that argues against this (1)
- E531R increases in vitro proliferation while abolishing leukaemogenesis, so the mutant is not simply a loss-of-function; an alternative reading is that losing BCOR pushes the cells out of the self-renewing state into a more proliferative, more differentiated one that cannot initiate disease - which would make BCOR a state-maintenance factor rather than a required recruitment partner, and the two readings predict different things about a BCOR-directed drug. Kuntimaddi A et al. 2015
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
CBX8 interacts with MLL-AF9 and TIP60 and is required for MLL-AF9-induced transcriptional activation and leukemogenesis; CBX8 ablation and point mutations in MLL-AF9 that disrupt the CBX8 interaction both abolish HOX gene upregulation and abrogate MLL-AF9 leukaemic transformation, while Cbx8-deficient mice are viable with no apparent haematopoietic defects.
CBX8, a polycomb group protein, is essential for MLL-AF9-induced leukemogenesis. · Tan J, Jones M, Koseki H, et al. · Cancer Cell 2011What this evidence cannot show
The MLL-AF9 point mutations used here were later shown not to be specific for the CBX8 contact: they also affect AF4 and DOT1L binding, and AF9 T542A significantly reduces BCOR binding. The CBX8 knockout result is unaffected by that criticism; the interaction-specific conclusion is.
Evidence that argues against this (1)
- With mutants that do separate the contacts, weakening the CBX8-AF9 interaction (CBX8 R336D) has no effect on MLL-AF9 colony formation and strengthening it (CBX8 A335V) inhibits colony formation, so the direct contact does not behave as a requirement in either direction. Schmidt CR et al. 2020
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The AF9 AHD is intrinsically disordered and recruits multiple transcription factors through coupled folding and binding, retaining substantial dynamic behaviour in complex that may facilitate exchange between disordered partners.
Leukemia fusion target AF9 is an intrinsically disordered transcriptional regulator that recruits multiple partners via coupled folding and binding. · Leach BI, Kuntimaddi A, Schmidt CR, et al. · Structure 2013What this evidence cannot show
Isolated peptides in vitro; the exchange interpretation is a proposal rather than a measurement.
Evidence that argues against this (1)
- If the AHD really does exchange partners rapidly, then a point mutation that abolishes one contact reshapes the competitive equilibrium among the other three rather than simply removing one - so the E531R phenotype may be as much about what now binds more as about what no longer binds. Schmidt CR et al. 2020
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Deleting the ENL YEATS domain from MLL-ENL decreases expression of pro-leukemic genes including Eya1 and Meis1, so EYA1 falls both when the ENL reader is removed and when the BCOR contact on the AF9 AHD is broken.
The ENL YEATS epigenetic reader domain critically links MLL-ENL to leukemic stem cell frequency in t(11;19) Leukemia. · Hu H, Saha N, Yang Y, et al. · Leukemia 2023What this evidence cannot show
Two different fusions, two different laboratories, two different perturbations and two different transcript panels; the convergence on EYA1 is an observation across papers, not a controlled comparison, and EYA1 is a prominent MLL-fusion target that many perturbations move.
Evidence that argues against this (1)
- EYA1 is downstream of many MLL-fusion perturbations, so its shared appearance in two unrelated loss-of-function experiments is weak evidence of a shared mechanism and may simply mark a sensitive node in the programme. Schmidt CR et al. 2020
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
An internal promoter in MLLT3 drives a YEATS-domain-devoid AF9 isoform expressed in haematopoietic stem cells and in poor-prognosis AML, and a similar internal promoter exists in MLLT1 that is inactive in the haematopoietic lineage and AML but active in other tissues.
Tissue- and tumor-type-specific expression of internal-promoter-driven YEATS-domain-devoid isoforms of MLLT1 and MLLT3. · Dailhau S, Zamani A, Pfeifer R, et al. · Biomark Res 2026What this evidence cannot show
Abstract read; expression is transcript-level, and the paper does not establish that the truncated AF9 protein is made at meaningful abundance in leukaemic blasts.
Evidence that argues against this (1)
- If a YEATS-devoid AF9 isoform is abundantly expressed in AML blasts, it carries an intact AHD and therefore competes with the fusion for BCOR, DOT1L, AF4 and CBX8 - so the competitive equilibrium measured in this beat may involve a third AHD-bearing species that no experiment has counted. The effect would be specific to the myeloid lineage, because the equivalent MLLT1 promoter is inactive there. Dailhau S et al. 2026
the essential contact on the retained partner module
same site, competing partner
CBX8 binds the AF9 AHD weakly; the direct contact is not a requirement
the AHD itself is not paralogue-equivalent: BCOR and CBX8 prefer ENL
the YEATS-devoid isoform is a third AHD-bearing competitor
what the retained AHD actually does
the query that found it, and the simulation that disagrees with it
the AHD ligands are mutually exclusive, so links 3 and 4 cannot be serial
the mutually exclusive AHD is the structural reason to expect partner-dependent ordering
the mutually exclusive AHD is the structural premise