Oncofusion

start / graph / hyp-the-post-cxxc-basic-patch-carries-part-of-the-address

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q82-cxxc-affinity

The 41 residues after the CXXC fold (1204-1244) carry a substantial part of the fusion's DNA grip, and removing them is why the structural literature's affinity is so low

also called the BCB tail hypothesis · post-CXXC basic patch

The piece of the fusion used for all the published structures stops just after the zinc-finger fold. The original 2002 work used a longer piece with a positively charged tail. If that tail does much of the DNA gripping, then the structures and the low affinity number are both describing a truncated version of the real thing.

The same thing, for a scientist

Every structure and every tagless solution affinity for the KMT2A CpG-binding region uses a construct ending at 1203, 1208 or 1214. Birke's GST-MT, the only measurement of 1147-1244, is 19-fold tighter than the closest comparator and confounds four variables. The hypothesis is that KMT2A 1204-1244 - 41 residues, 13 lysines, net charge about +8, inside a UniProt-annotated disordered region - contributes an appreciable fraction of the DNA affinity as a sequence-independent electrostatic extension, and that its omission is a large part of why the structural literature's numbers cluster in the micromolar. If true, three consequences follow: the operative DNA-binding module of the fusion is the MT/BCB domain rather than the minimal CXXC; every structure-based design effort has been working on the minority of the interaction; and the addition strengthens rather than weakens the project's address hypothesis, because a sequence-independent affinity extension raises grip without raising specificity, which is precisely the split the hypothesis predicts. Prior probability is set at 0.5 - the architecture is the textbook one for an electrostatic enhancer, but Allen's 1146-1214 construct reaches 11 residues into the patch and is only 10-fold tighter than Cierpicki's, which is a weak prior against a very large effect.

Go deeper
The reasoning, and what argues against it

For.

  • The architecture is standard: a folded DNA-binding module followed by a disordered, strongly basic extension is the canonical form of an electrostatic affinity enhancer, and it is why the 2002 literature named the domain BCB rather than CXXC.
  • The N-terminal basic patch of the same BCB pair, KKGRRSRR at 1147-1154, is inside the minimal construct, and its residues R1150 and R1154 are among the contacts mapped in 2KKF with R1154A costing 9- to 15-fold. So the other basic flank is already demonstrated to matter. There is no principled reason the C-terminal one would not.
  • The interaction's thermodynamics are entropically driven and counterion-release dominated, exactly the regime in which extra positive charge buys affinity cheaply.
  • Birke's construct is 19-fold tighter than the closest comparator, and of the four confounded variables this is the only one that changes the protein.

Against, and this is why the prior is only 0.5.

  • Allen's construct reaches K1214, eleven residues into the patch, adding roughly three lysines - and it is only 10-fold tighter than Cierpicki's 1147-1203, against a comparison that also differs in salt (250 against 300 mM NaCl), pH (6.5 against 7.1) and method. If the patch were worth two orders of magnitude, eleven residues of it should have shown more.
  • The other three confounded variables in the Birke comparison are individually plausible: six CpGs against two, a 52-bp duplex presenting three separate footprints, and a surface geometry read kinetically. They could account for 19-fold between them with no contribution from the patch at all.
  • A disordered basic tail contributes to non-specific DNA binding, and the SPR ligand was a specific CpG duplex; the effect might therefore be smaller in that assay than in a genomic context, which would make the hypothesis true in cells and invisible in Birke's data.
Why it matters even at 0.5

The dispatch brief for this beat framed this as the branch that strengthens the project's hypothesis. It is right, and the reason is worth stating precisely.

The menin-interface hypothesis says the fusion is placed by protein-protein tethers and not by its own DNA reader. A common objection is that the DNA reader is right there, is retained in every fusion, and is genetically required - so why is it not the address?

If a sequence-independent basic tail carries much of the DNA affinity, the objection dissolves cleanly: the fusion's DNA-binding module would be split into a weak, specific part (the CXXC fold, reading CpG) and a strong, non-specific part (the basic tail, reading phosphate backbone). Affinity and specificity would live in different pieces of the same domain. A module built that way grips chromatin firmly and tells you nothing about where - which is exactly what the address hypothesis needs to be true, and it would be measured rather than argued.

If instead the patch contributes little, the Birke figure is explained by ligand valency and surface geometry, the operative module is the CXXC fold as drawn, and the hypothesis loses a support it never had.

Facts and figures
statementKMT2A 1204-1244, a 41-residue disordered basic patch omitted from every structure and every tagless affinity measurement, contributes an appreciable fraction (hypothesised: 5-fold or more) of the DNA-binding affinity of the fusion's CpG-binding module, in a sequence-independent electrostatic manner.
originliterature-gap
prior probability0.5
would strengthen the menin hypothesis if trueYes, and this is the reason to run it. A module whose affinity lives in a sequence-independent basic tail and whose specificity lives in a separate two-base reader grips chromatin firmly while carrying no positional information - which is exactly the split the menin-interface hypothesis requires, measured rather than argued.
cost tierlow
shown for fusionNO - nobody has looked in any fusion. No construct containing 1204-1244 has been tested in any functional or occupancy assay.
predictions
[
  "A tagless 1147-1244 construct titrated against the same oligonucleotide as 1147-1203, at the same salt and by the same method, will be at least 5-fold tighter.",
  "The affinity difference between the two constructs will be larger at 150 mM than at 300 mM salt, because a lysine-rich extension works by counterion release.",
  "The affinity difference will be sequence-independent: the same fold-gain on an MLL-protected Hoxa9 site and on a non-protected one.",
  "The longer construct will show a shallower dependence of affinity on CpG count, because part of its binding energy does not come from the CpG.",
  "A KMT2A fusion carrying a 1204-1244 deletion or a lysine-to-alanine block substitution will show reduced chromatin occupancy by CUT and RUN even though its CXXC fold and its C1188 are intact."
]
experiments
[
  "exp-tagless-side-by-side-cxxc-versus-mt-domain",
  "exp-gst-cleavage-and-orientation-controls-on-the-birke-spr-geometry"
]
analyses
[
  "ana-cxxc-dna-affinity-ladder",
  "ana-cxxc-occupancy-is-underdetermined-by-every-published-kd"
]
falsifiers
[
  "A tagless 1147-1244 construct binds the same oligonucleotide with a Kd within 2-fold of tagless 1147-1203, at matched salt and by the same method. The patch contributes nothing appreciable and the hypothesis is dead.",
  "The Birke 19-fold is fully accounted for by ligand valency alone - i.e. a 1147-1203 construct on a 52-bp six-CpG duplex recovers 33 nM without the extra residues.",
  "Deleting 1204-1244 from a KMT2A fusion leaves chromatin occupancy unchanged by CUT and RUN in a human KMT2A::AFF1 line.",
  "The patch is shown to be occluded in the fusion - for instance by an AFF1, menin or LEDGF contact - so that whatever it does in isolation it cannot do in the oncoprotein."
]
How this could bite us back
  • If the patch carries the affinity, a structure-based ligand designed against 2KKF or 4NW3 is designed against the minority of the interaction and would be expected to fail in cells for reasons no assay on the short construct would reveal.
  • A disordered polycationic affinity element is the least druggable object available: no pocket, no fixed conformation, and a competitor would have to be polyanionic and therefore cell-impermeant.
  • If true, the fusion has a chromatin grip that survives loss of CpG recognition, which means a CXXC-directed agent could show full target engagement and no displacement - the same false-negative endpoint problem recorded in ana-affinity-rank-does-not-predict-cxxc-function.
Still open
  1. Is 1244 the right C-terminal boundary to test, or should the construct run to 1300 to cover more of the disordered region? Testing both costs one extra construct.
  2. Does the patch bind DNA on its own as a synthetic peptide? A 41-mer peptide titration is the cleanest possible version of the question and is cheap.
  3. Is the patch phosphorylated? Thirteen lysines in a disordered region adjacent to serines and threonines is a plausible phosphoswitch, and this graph already records CK2-dependent regulation in the KMT2A N-terminus.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Birke's GST-MT construct is Q03164 1147-1244 and is 19-fold tighter (33 nM) than the closest comparable measurement (0.64 uM, GST-tagged 1147-1203 at similar ionic strength), with the 41 extra C-terminal residues being one of four confounded differences.

    The MT domain of the proto-oncoprotein MLL binds to CpG-containing DNA and discriminates against methylation. · Birke M, Schreiner S, García-Cuéllar MP, et al. · Nucleic Acids Res 2002
    PMID 11842107DOI
    What this evidence cannot show

    The 1147-1244 boundary is this cycle's translation of the primers in the paper's Methods against a live UniProt Q03164 fetch, independently corroborated by Erfurth 2008 stating 1147-1244 in print. The 19-fold is a cross-paper comparison between an SPR kinetic constant and a fluorescence-polarisation equilibrium constant, which is the weakest kind of comparison and is the reason this is a hypothesis rather than a finding.

    Evidence that argues against this (1)
    • Allen's 1146-1214 construct includes eleven residues of the patch and about three of its lysines and is only 10-fold tighter than the 1147-1203 construct, across a comparison that also differs in salt, pH and method. That is a real prior against a very large patch contribution, and it is the single strongest argument against this hypothesis. Allen MD et al. 2006
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    KMT2A 1204-1244 comprises 41 residues with 13 lysines, no arginines and 5 acidic residues (net charge about +8), lies inside a UniProt-annotated disordered region (1200-1375), and is the C-terminal basic flank that gives the domain its BCB - basic-cysteine-basic - name.

    What this evidence cannot show

    Sequence and features fetched live from rest.uniprot.org on 2026-09-21 (length 3,969, ZF-CXXC 1147-1195, disordered 1200-1375); the residue composition is counted from that sequence. A UniProt disorder region is a prediction, not a measurement (brief rule 91), and no experiment has shown 1204-1244 to be disordered in isolation or in the full-length protein. Charge is computed at neutral pH ignoring the local electrostatic environment.

    Evidence that argues against this (1)
    • A high net positive charge in a disordered region is also the signature of a nuclear localisation signal, a nucleolar targeting sequence, or a phosphoregulated interaction surface. Attributing it to DNA affinity is the most obvious reading and not the only one, and this graph already records CK2-dependent phosphoregulation elsewhere in the KMT2A N-terminus. Birke M et al. 2002
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The N-terminal basic flank of the same BCB pair, KMT2A 1147-1154 (KKGRRSRR), is inside the minimal construct and contains contacts mapped in the DNA-bound structure, with R1154A costing 9- to 15-fold and R1150A costing 5-fold, establishing that a basic flank of this domain does carry measurable DNA affinity.

    Structure of the MLL CXXC domain-DNA complex and its functional role in MLL-AF9 leukemia. · Cierpicki T, Risner LE, Grembecka J, et al. · Nat Struct Mol Biol 2010
    PMID 20010842DOI
    What this evidence cannot show

    This is an argument from the symmetry of the BCB architecture, not evidence about 1204-1244. The N-terminal flank is inside the folded module and its arginines make defined contacts in the structure; the C-terminal flank is disordered and would work by a different, non-specific mechanism, so the parallel is suggestive rather than probative.

    Evidence that argues against this (1)
    • The two flanks differ chemically as well as structurally: the N-terminal patch is arginine-rich (4 R, 2 K) and arginines make bidentate guanine contacts, while the C-terminal patch is lysine-only (13 K, 0 R) and lysines contribute mostly undirected backbone electrostatics. So if the C-terminal flank contributes at all it should contribute affinity without specificity, which is the hypothesis - but it also means the N-terminal precedent does not predict the magnitude. Allen MD et al. 2006
Connected to