Re-founding the address argument: the CXXC domain cannot specify a location at any affinity, and a tighter one would be worse
also called the too-weak-to-hold argument, corrected · specificity not affinity
This project used to argue that the fusion's DNA grip is too weak to hold it in place. That was based on the weakest of nine measurements. The better argument is that the grip reads a two-letter pattern that appears tens of millions of times in the genome, so it cannot pick a location however strong it is - and if it were stronger it would be worse at picking.
The same thing, for a scientist
The claim that the CXXC-DNA contact is too weak to be the fusion's address rested on Kd 44 uM, which ana-cxxc-dna-affinity-ladder shows is the extreme low-affinity point of a nine-measurement series spanning to 33 nM, measured at twice physiological ionic strength against the shortest construct and a single-CpG 12-mer. The conclusion is nonetheless correct and is re-founded here on four supports that do not depend on the magnitude: in the MLL-AF9 CXXC swap series affinity rank is inverted against functional rank; in the same series every swapped domain including two with no detectable DNA binding still targeted the fusion to Hoxa9 by ChIP; the domain has no measured base readout outside the CpG dinucleotide and gains only 2-4 fold from additional CpG motifs; and genome-wide in human SEM cells the fusion occupies 20% of unmethylated CpG sites while CFP1 occupies 50% and KDM2B 89%, with fusion occupancy tracking menin at r2 = 0.96. The reformulated claim is that the CXXC domain protects particular CpGs from methylation and does not determine where the fusion goes - which is stronger, directly measured, and immune to the affinity objection.
The old argument, why it fails, and the argument that replaces it
The old form. The CXXC-DNA contact is 44 uM, three orders of magnitude weaker than the menin contact, therefore it cannot hold the fusion on chromatin by itself, therefore menin, LEDGF and the AT-hooks supply the grip.
Why it fails. The premise is the weakest of nine measurements, taken at twice physiological ionic strength against the shortest construct and the poorest DNA ligand, and compared with a menin number taken at one-sixth the ionic strength [see ana-cxxc-dna-affinity-ladder and ana-the-menin-affinity-comparison-was-made-at-one-sixth-the-salt]. On the nearest-to-physiological pair the ratio is twelve-fold; on Birke's geometry the CXXC contact is the tighter of the two. An argument whose conclusion flips with a buffer choice is not an argument.
The new form, which is what the evidence actually supports.
The CXXC domain reads a two-base motif with no measured readout of flanking bases, so it cannot specify an address at any affinity - and higher affinity makes it worse, not better. What it does is protect particular CpGs from methylation once the fusion is there. Where the fusion goes is set by menin, and this is measured genome-wide in the index fusion.
The four supports, in order of how directly they bear:
- In the fusion, in vivo, affinity does not predict function. CGBP CXXC binds DNA more tightly than DNMT1 CXXC; DNMT1 CXXC supports MLL-AF9 leukaemogenesis and CGBP CXXC does not 1. The authors state that high DNA-binding affinity alone is not sufficient.
- In the fusion, localisation does not require DNA binding at all. Every swapped domain in that series - including MBD1 CXXC and the MBD1 MBD, which bind no unmethylated DNA detectably - still targeted MLL-AF9 to the Hoxa9 locus by ChIP 1.
- Specificity, not strength, is the authors' own conclusion. The CXXC domain "does not recognize DNA bases beyond the CpG dinucleotide and therefore is not solely responsible for specific localization to the Hoxa9 locus"; sites with multiple CpGs gain only 2-4 fold 2. A motif this short is present tens of millions of times in the genome.
- Genome-wide, in a human t(4;11) line, the fusion under-occupies the address. MLL-AF4 sits on 20% of unmethylated CpG sites, CFP1 on 50% and KDM2B on 89%; fusion occupancy tracks menin at r2 = 0.96 3.
The arithmetic argument that survives, stated correctly. A tight non-specific binder is a worse address-reader than a weak one, because the address is defined by a motif with an enormous number of copies. If the CXXC domain bound uCpG at 33 nM, then at any plausible nuclear concentration it would be saturated across bulk CpG and would carry no positional information whatever; at 44 uM it is a sampling device. Either way the positional information has to come from somewhere else. This is the one place where the higher number strengthens the project's hypothesis rather than weakening it - and it was not the reasoning the project was using.
What this costs the hypothesis, honestly. The old argument was quantitative and looked decisive. The new one is qualitative about specificity and quantitative only in Kerry's occupancy percentages, which come from one cell line and depend on a peak-calling threshold. The menin-interface hypothesis is now supported by a better argument that is somewhat less crisp, and anyone presenting it should say so rather than reach back for the number.
Where this leaves the entries that made the old argument
Six entries carried the 44 uM figure and three of them built the too-weak argument on it. This cycle:
interface-kmt2a-cxxc-cpg-dna- the title,summary_lay,summary_expert, the 'how weak this actually is' section, theaffinity_contextfield and two bite-back items all asserted the three-orders-of-magnitude comparison. Rewritten to carry the ladder and this argument; the entry now states the conditions of every number it quotes.exp-cxxc-reader-depletion-versus-kmt2a-fusion-occupancyandgene-kdm2aquoted 44 uM as the affinity. Amended to quote the range with its conditions.ana-the-fusion-sits-on-one-fifth-of-the-unmethylated-cpg-sites-and-menin-picks-which,gene-idh2andhyp-a-hypermethylating-comutation-should-antagonise-the-unmethylated-cpg-addresseach carried a caveat, written last cycle, correctly flagging the discrepancy as unreconciled and attributing it to GST dimerisation. Those caveats were right to flag it and the GST attribution is now bounded at a few-fold; each has been pointed at the resolution.target-kmt2a-cxxc-dna-interface, the druggability dossier, andtarget-phase2-cxxc-dna-dossier, the phase-2 go/no-go, both reason about the potency bar. Amended - see the bite-back below.
Brief instruction observed: do not leave an entry arguing with itself. Where the ladder shows a sentence is wrong rather than merely contested, the body prose has been changed, not only the counter-evidence.
| claim withdrawn | The CXXC-DNA contact is 44 uM, roughly three orders of magnitude weaker than the menin contact, and is therefore too weak to hold the fusion on chromatin by itself. |
|---|---|
| reason withdrawn | 44 uM is the extreme low-affinity point of a nine-measurement series spanning to 33 nM; it was measured at 300 mM NaCl against a 57-residue construct and a single-CpG 12-mer, and was compared with a menin figure measured at 50 mM NaCl. The nearest-to-physiological pair gives 12-fold, and Birke's geometry gives the CXXC contact as the tighter of the two. |
| claim substituted | The CXXC domain reads a two-base motif with no measured readout of flanking bases and so cannot specify a genomic address at any affinity; higher affinity would make it a worse address-reader, not a better one. Its function is protection of particular CpGs from methylation once the fusion is present. Where the fusion goes is set by menin, measured genome-wide in the index fusion. |
| shown for fusion | Human KMT2A::AFF1 in SEM for the occupancy arm (PMID 28076791); murine MLL-AF9 for the swap and mutant arms (PMIDs 23990460, 20010842). No affinity measurement in any fusion. |
| honest cost | The withdrawn argument was quantitative and read as decisive. The replacement is qualitative about specificity, plus occupancy percentages from one cell line at one peak-calling threshold. The hypothesis is better supported and less crisply stated, and presentations should say so rather than reach back for the number. |
supports
[ "affinity rank inverted against functional rank in the MLL-AF9 CXXC swap series (PMID 23990460)", "all swapped domains including non-binders still targeted Hoxa9 by ChIP (PMID 23990460)", "no base readout beyond the CpG dinucleotide; 2-4 fold from extra CpGs (PMID 20010842)", "fusion occupies 20% of uCpG against CFP1 50% and KDM2B 89%; fusion-menin r2 = 0.96 (PMID 28076791)" ]
entries amended this cycle
[ "interface-kmt2a-cxxc-cpg-dna", "exp-cxxc-reader-depletion-versus-kmt2a-fusion-occupancy", "gene-kdm2a", "ana-the-fusion-sits-on-one-fifth-of-the-unmethylated-cpg-sites-and-menin-picks-which", "gene-idh2", "hyp-a-hypermethylating-comutation-should-antagonise-the-unmethylated-cpg-address", "target-kmt2a-cxxc-dna-interface", "target-phase2-cxxc-dna-dossier" ]
- If the CXXC contact is nearer 33 nM than 44 uM in the cell, a DNA-competitive ligand must be nanomolar to compete rather than micromolar. The potency bar rises by up to three orders of magnitude, and the constructive angle previously recorded in interface-kmt2a-cxxc-cpg-dna - that a competitor need only be selective, not potent, because the natural ligand is weak - is withdrawn.
- A tighter contact also means a slower off-rate, so a competitive ligand faces a kinetic barrier as well as a thermodynamic one, and a soak-off experiment could show apparent inactivity at equilibrium-adequate concentrations.
- If the operative requirement is protection of specific CpGs from methylation rather than binding, then a CXXC ligand that blocks binding without changing methylation would do nothing, and one that does change methylation could be phenocopied more cheaply by a hypomethylating agent acting in the opposite direction - which is exactly the enasidenib liability already recorded in this graph.
- How many unmethylated CpG islands exist in an infant KMT2A::AFF1 blast, and what fraction does the fusion occupy? Kerry gives the fraction of uCpG SITES in SEM; the island-level number in primary infant cells is the quantity the address argument really wants.
- Under a menin inhibitor, does the fusion vacate or relocate onto the 80% of the address it currently ignores? Kerry makes menin sufficient but not necessary, so both are live. Q83 is working this question in this cycle.
- Does the reformulated argument survive in B-lineage? Every functional support here is murine myeloid except the occupancy arm.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In MLL-AF9, the DNMT1 CXXC domain (about 34-fold weaker DNA binding than MLL CXXC) supports colony formation and in vivo leukaemogenesis while the tighter-binding CGBP/CXXC1 domain does not, and every swapped domain including two with no detectable DNA binding still targeted the fusion to the Hoxa9 locus by ChIP.
Functional specificity of CpG DNA-binding CXXC domains in mixed lineage leukemia. · Risner LE, Kuntimaddi A, Lokken AA, et al. · J Biol Chem 2013What this evidence cannot show
Murine MLL-AF9 in c-Kit+ bone marrow progenitors, not human KMT2A::AFF1 and not B-lineage (rule 73). The ChIP result is a binary enrichment at one locus in a transduced population and does not quantify occupancy, so 'still targeted' means detectable, not equivalent. Full text read this cycle via the PMC HTML route (PMC3795288).
Evidence that argues against this (1)
- Cierpicki's C1188D mutant, which abolishes DNA binding in the native MLL CXXC fold rather than swapping the whole domain, DOES abolish immortalisation and leukaemia - so DNA binding by this domain is required. Requirement and address-setting are different claims and this entry only denies the second; an argument that slid from one to the other would be wrong in the opposite direction. Cierpicki T et al. 2010
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The KMT2A CXXC domain makes all base-specific contacts through a short 1182-1188 loop that reads the CpG dinucleotide only, does not recognise bases beyond it, and gains only 2- to 4-fold on sites with multiple CpG motifs, from which the authors conclude the domain is not solely responsible for specific localisation to the Hoxa9 locus.
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 2010What this evidence cannot show
The 2- to 4-fold figure and the oligonucleotide panel behind the specificity claim are in Supplementary Table 1, which is behind the Europe PMC open-access gate and was NOT read this cycle; the authors' statement of both in the main text was verified. The structural claim about the 1182-1188 loop is from a solution NMR ensemble of an isolated domain on a designed palindrome.
Evidence that argues against this (1)
- Erfurth 2008, same laboratory, measured a 9-fold affinity spread across five CpG-containing 12- to 16-mers at fixed construct, salt and method, with the MLL-protected Hoxa9 sites binding tighter than matched unprotected ones. 'No base readout beyond the CpG' and 'a flat affinity landscape across CpG-containing DNA' are different claims and only the first is supported. Erfurth FE et al. 2008
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In human SEM t(4;11) KMT2A::AFF1 cells, MLL-AF4 occupies 20% of unmethylated CpG sites while CFP1/CXXC1 occupies 50% and KDM2B 89%, and MLL(N) occupancy correlates with menin at r2 = 0.96 at fusion peaks against 0.27 for menin against CFP1 - so the fusion is placed by menin rather than by availability of its own DNA-binding motif.
MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia. · Kerry J, Godfrey L, Repapi E, et al. · Cell Rep 2017What this evidence cannot show
One human cell line; the unmethylated-CpG denominator is defined by the study's own peak calling and a weaker binder is systematically under-called at a fixed threshold. This finding is in the figures rather than the abstract and was recovered at the coordinator in cycle 45. Q83 is working on fusion occupancy under menin inhibition concurrently and owns any deeper reading of this paper; this entry cites the published occupancy contrast only and does not extend it.
Evidence that argues against this (1)
- Kerry's own tethering experiment makes the menin interaction SUFFICIENT but not NECESSARY for fusion recruitment, so 'menin picks the sites' is an inference from correlation plus sufficiency. A menin inhibitor might relocate the fusion onto the 80% of the address it currently ignores rather than evicting it, and nobody has measured where it goes. Risner LE et al. 2013
Having removed the affinity premise, the ladder hands the address argument over to the specificity argument that replaces it.
A measured 9-fold affinity spread across CpG-containing 12-mers sits in tension with, and does not overturn, the no-readout-beyond-the-CpG conclusion.
The direct in-fusion dissociation of affinity from function is the strongest support for the replacement argument.
Re-founds the menin-interface hypothesis's address argument on measured specificity and occupancy rather than on a Kd magnitude.
- regulates →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
States what the CXXC domain does for the index fusion - protect particular CpGs from methylation - and what it does not do, which is choose where the fusion goes.
- refutes →Step 1 — How the fusion finds its genes: AT-hooks and the CXXC domain reading unmethylated CpG
Withdraws the affinity-magnitude form of the too-weak-to-hold argument that this mechanism entry states, and supplies the replacement.
Withdraws the entry's claim that a 44 uM module cannot be the sole determinant of targeting because it is weak, replacing it with a specificity argument that does not depend on the magnitude.
- related to →Is the CpG address a contested resource? On current evidence, no - and the beat that went looking says so
The contested-address hypothesis is about which CXXC reader occupies a site; this entry argues that occupancy is not decided by CXXC affinity at all.
Gives the CpG-address entry the argument that survives the affinity correction.
Kerry's occupancy contrast, which refuted that entry's zero last cycle, is also the strongest support for this replacement argument.
The phase-2 dossier's selectivity-before-potency framing survives, but its implicit potency assumption changes with the ladder.