An anti-MLL-N antibody sees the wild-type allele and the direct fusion, and nothing else - and cannot rank them
also called MLL-N antibody specificity · anti-KMT2A N-terminal antibody · MLL (456)
Antibodies raised against the front end of the KMT2A protein pick up two different proteins in a leukemia cell: the normal one and the fusion. On a gel you can tell them apart because they are different sizes; in any experiment without a size readout - pulling proteins down, mapping where they sit on DNA - you cannot.
The same thing, for a scientist
In a KMT2A-rearranged cell the N-terminal ~1,400 residues of KMT2A are present in two proteins: the wild-type allele (full length, and the p320 MLL-N fragment after taspase1 cleavage) and the direct fusion. An anti-MLL-N antibody binds both. It does **not** bind the reciprocal product (no KMT2A N-terminus) and does **not** bind KMT2A\* (5'-truncated from the internal promoter, so no N-terminus either) - which makes it the exact mirror of the anti-MLL-C reagent this graph already audits. The critical asymmetry is between assays that preserve a size dimension and assays that do not. **Western blot resolves them** - the wild-type species and the ~240 kDa KMT2A::AFF1 run apart. **Immunoprecipitation, ChIP, CUT&RUN and CUT&Tag do not**, and every occupancy or complex-composition result obtained with such an antibody in a KMT2A-r cell is a measurement of the sum. The field states this plainly when pressed: a 2024 study writes that *"the major obstacle of comparing chromatin occupancy of MLL1 and MLL-AF9 in MLL cells is that there is no MLL-AF9-specific antibody available. Antibodies recognizing its MLL fragment would also recognize WT MLL1, and antibodies recognizing its AF9 fragment would also recognize WT AF9"* [E1]. The problem is compounded because the wild-type allele is not a bystander: MLL-AF9 leukaemogenesis **requires** coexpression of wild-type Mll [E3], so the wild-type protein is present, functional and on chromatin in exactly the cells where its signal is being pooled with the fusion's.
The species table
| protein present in a reciprocal-positive t(4;11) cell | has KMT2A N-terminus? | seen by anti-MLL-N? | seen by anti-MLL-C? |
|---|---|---|---|
| wild-type KMT2A (full length / MLL-N p320) | yes | yes | yes (as MLL-C p180) |
| KMT2A::AFF1 direct fusion | yes | yes | no |
| AFF1::KMT2A reciprocal fusion | no | no | yes (indistinguishable from wild-type MLL-C) |
| KMT2A* (internal-promoter product) | no | no | yes (indistinguishable from wild-type MLL-C) |
Read across the rows and the two reagent classes are complementary in coverage and equally blind in attribution. Neither alone, and not both together, assigns a signal to one protein.
Where the size dimension saves the result and where it does not
This is the operational distinction that decides which published claims survive.
Survives. A western blot with an anti-MLL-N antibody showing a ~240 kDa band absent from KMT2A-germline cells is attributable: molecular weight is the discriminator, not the epitope. Quantitative comparison of fusion and wild-type band intensity is weaker, because the antibody's affinity for the two species need not be equal and nobody has measured it.
Does not survive. Any conclusion of the form "MLL occupies gene X", "MLL interacts with protein Y", or "MLL-containing complexes contain Z" obtained by IP or ChIP with an anti-MLL-N antibody in a KMT2A-r cell. The pull-down contains both proteins and the experiment has no way to say which contributed. This is not a hypothetical concern: the discrepancy between two published ChIP-seq studies of MLL1-versus-MLL-AF9 occupancy - one reporting distinct sites 2, one reporting identical ones 1 - is explicitly discussed by both sets of authors in terms of antibody choice.
Fixes that work, in increasing order of effort: (i) use a C-terminal antibody in a fusion family where the fusion has no KMT2A C-terminus, which is what 2 did and which is valid for MLL-AF9 but not for a reciprocal-positive t(4;11) cell; (ii) delete the wild-type partner allele so the partner antibody sees only the fusion, which is what 1 did with an AF9 knockout in MOLM-13; (iii) tag the fusion allele endogenously, which nobody has done.
The antibody that actually appears in the literature
The reagents are not generic. The 2024 occupancy study used two antibodies from the Canaani laboratory: MLL (456), raised against an MLL-N fragment downstream of the breakpoints, and MLL (468), against MLL-C 1. That the widely used N-terminal reagent is raised downstream of the breakpoints is the whole problem stated as a reagent property - it is, by design, an epitope both the wild-type protein and the fusion retain.
Anyone re-reading an MLL ChIP in this disease should first ask which of these two antibodies was used, because MLL (468) in an MLL-AF9 cell is attributable to the wild-type allele and MLL (456) in the same cell is not attributable at all.
| reagent class | antibodies raised against the KMT2A N-terminus (e.g. MLL (456), raised downstream of the breakpoints) |
|---|---|
| mirror entry | concept-an-mll-c-antibody-cannot-separate-wild-type-from-reciprocal-fusion |
| worked fix in the literature | AF9 knockout in MOLM-13 followed by AF9-C CUT&Tag (PMID 39002676) |
species recognised
[ "wild-type KMT2A full length", "wild-type MLL-N p320 after taspase1 cleavage", "the direct KMT2A::partner fusion" ]
species not recognised
[ "AFF1::KMT2A reciprocal fusion", "KMT2A* (5'-truncated internal-promoter product)" ]
attributable in
[ "western blot (species resolved by apparent molecular weight)" ]
not attributable in
[ "immunoprecipitation", "ChIP", "CUT&RUN", "CUT&Tag", "immunofluorescence", "flow cytometry" ]
- If chromatin-bound 'MLL' in a KMT2A-r cell is substantially wild-type protein, a therapeutic aimed at that signal is aimed partly at normal KMT2A function in an infant - the least favourable framing available, and one this graph already records for the WRAD case.
- A pharmacodynamic assay reading 'MLL occupancy' by an N-terminal antibody would show partial loss under a drug that removed only the fusion, and would be scored as partial efficacy when the effect on the intended target was complete.
- What is the ratio of wild-type KMT2A to fusion protein in an anti-MLL-N immunoprecipitate from a t(4;11) line? Nobody has measured it, and every pooled ChIP result would be interpretable if they had. A junction-peptide PRM on the IP eluate answers it in one run.
- Does the MLL (456) antibody bind the fusion and the wild-type protein with equal affinity? If not, even the western-blot ratio that this entry calls attributable is quantitatively wrong in an unknown direction.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The authors of a 2024 chromatin-occupancy study state that no MLL-AF9-specific antibody exists, that an MLL-fragment antibody also recognises wild-type MLL1 and an AF9-fragment antibody also recognises wild-type AF9, and they solve it by generating an AF9-knockout MOLM-13 line so that an AF9-C antibody recovers only the fusion.
MLL-AF9 regulates transcriptional initiation in mixed lineage leukemic cells. · Yang Z, Zhang G, Zhao R, et al. · J Biol Chem 2024What this evidence cannot show
Read in full via PMC11345648 (Europe PMC fullTextXML). The quoted statement is the authors' own framing in their Discussion. Their AF9-knockout solution is justified on the grounds that AF9 expression is low in most haematopoietic lines and its loss may be of little consequence - an assumption, not a measurement, and one now complicated by the discovery of a highly expressed YEATS-devoid MLLT3 isoform in HSCs and AML.
Evidence that argues against this (1)
- An earlier study using an MLL-C antibody that provably cannot detect MLL-AF9 reported that MLL1 and MLL-AF9 occupy distinct chromatin regions, the opposite conclusion. Its authors note they cannot rule out that the discrepancy is due to antibody differences, or to fusion genes residing at the endogenous locus versus being retrovirally integrated. Xu J et al. 2016
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
An MLL1 antibody that specifically recognises the 180 kDa MLL1 C-terminal fragment cannot detect MLL-AF9, and that property was used to attribute a genome-wide ChIP-seq signal to wild-type MLL1 rather than to the fusion.
MLL1 and MLL1 fusion proteins have distinct functions in regulating leukemic transcription program. · Xu J, Li L, Xiong J, et al. · Cell Discov 2016What this evidence cannot show
Read in full via PMC4869169. The design works because MLL-AF9 has no KMT2A C-terminus at all. It fails completely in a reciprocal-positive t(4;11) cell, where a second protein carrying the KMT2A C-terminus is present - a case this study did not face and which is the subject of `concept-an-mll-c-antibody-cannot-separate-wild-type-from-reciprocal-fusion`.
Evidence that argues against this (1)
- A later study using an AF9-knockout genetic subtraction rather than an antibody-specificity argument found MLL1 and MLL-AF9 occupancy patterns to be identical, with 87% of MLL-AF9 peaks co-occupied by MLL1 - directly contradicting the distinct-sites conclusion. Yang Z et al. 2024
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The wild-type KMT2A allele is required for MLL-AF9 leukaemogenesis, so the protein whose signal an anti-MLL-N antibody pools with the fusion's is itself functionally necessary and present on chromatin in the same cells.
MLL-AF9-induced leukemogenesis requires coexpression of the wild-type Mll allele. · Thiel AT, Blessington P, Zou T, et al. · Cancer Cell 2010What this evidence cannot show
A murine MLL-AF9 model; whether the same wild-type requirement holds for human KMT2A::AFF1 B-ALL has not been shown with the same rigour. The relevance here is not the requirement itself but that it forecloses the escape route of arguing that wild-type KMT2A is a negligible contributor to an MLL-N pull-down.
Evidence that argues against this (1)
- If the wild-type allele is required for leukaemogenesis, then a shared-epitope pull-down is not purely a nuisance: some of what is attributed to the fusion may be a genuine wild-type contribution to the same biology, which makes the pooled measurement wrong about mechanism rather than merely imprecise. Thiel AT et al. 2010
the mirror reagent: N-terminal antibodies pool wild-type with the direct fusion, C-terminal antibodies pool wild-type with the reciprocal product and KMT2A*
- related to →KMT2A / MLL1 — the gene that breaks
the epitope both the wild-type protein and the direct fusion retain
- related to →KMT2A*: a menin-blind, LEDGF-blind, SET-competent KMT2A protein made from out-of-frame reciprocal alleles
KMT2A* has no N-terminus, so this reagent class cannot see it either
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
the MLL-AF9 case, where a C-terminal antibody is attributable and an N-terminal one is not
the two reagent classes have complementary coverage and neither attributes
the two shared-epitope reagents used in series are fusion-specific even though neither is alone
the reagent property behind the MLL-N minus MLL-C subtraction is established there
the attribution failures of every retained-sequence reagent are what a junction antibody would remove
the anti-MLL-N blot lane is interpreted under that entry's rules