Step 1 — How the fusion finds its genes: AT-hooks and the CXXC domain reading unmethylated CpG
also called MLL CXXC targeting · AT-hook DNA binding · MT domain · CpG island recognition by MLL fusions
The fusion protein has two gripping parts kept from KMT2A: sticky AT-hooks that hold onto DNA generally, and a CXXC finger that only sticks to CpG letters that have no chemical tag on them. Because the genes a stem cell keeps switched on are exactly the ones with untagged CpGs, the fusion is automatically parked on the wrong set of genes.
The same thing, for a scientist
Every leukemogenic KMT2A fusion retains roughly the N-terminal 1,400 residues, which carry three AT-hooks, a subnuclear-localisation/SNL region and the CXXC (methyltransferase-homology, 'MT') domain. The CXXC domain binds non-methylated CpG dinucleotides and discriminates against 5-methylcytosine; point mutations that abolish CpG binding also abolish transactivation and myeloid transformation, and AT-hook deletion substantially impairs transformation. Targeting is therefore not sequence-specific in the classical transcription-factor sense: it is a read-out of the pre-existing DNA-methylation landscape of the cell, which is why the fusion's target set is largely the CpG-island promoter set that was already active or poised in the cell of origin rather than a novel set of loci.
The domains that survive the translocation are the targeting module2 deeper
All leukemogenic KMT2A fusions keep the amino-terminal ~1,400 residues of KMT2A and lose everything C-terminal of the breakpoint. High-resolution mutagenesis of that retained region identified three domains that are individually essential for in-vitro myeloid transformation without disturbing subnuclear localisation, and the CXXC/basic domain was one of them. Point mutations in CXXC that eliminated transformation by an MLL fusion also abolished (i) recognition and binding of non-methylated CpG DNA in vitro and (ii) transactivation in vivo — the cleanest available demonstration that DNA reading and oncogenicity are the same function 1.
The AT-hooks are a second, partly redundant DNA-contacting element. Deleting either the AT-hook motifs or the CXXC ('methyltransferase-homology') domain of KMT2A substantially impaired transformation by KMT2A::MLLT1 (HRX-ENL), and the CXXC domain was shown to bind DNA in a non-sequence-specific manner — i.e. full transforming activity requires multiple DNA-binding structures acting together rather than one high-affinity site 2.
Why this is a read-out of methylation, not a sequence code1 deeper
CXXC is a 5-methylcytosine-discriminating reader, not a sequence-specific DNA-binding domain. Its consequence is that KMT2A and its fusions are recruited to the unmethylated CpG-rich promoter set — and that is exactly what genome-scale work shows: MLL recruits RNA polymerase II to unmethylated CpG-rich promoters through its CXXC domain, and the same circuitry is co-opted by leukemic fusion proteins to drive sustained self-renewal transcription 3.
The practical consequence for this project is uncomfortable: the target list is written by the cell, not by the fusion. The same fusion in a different epigenome should pick different genes, which is the mechanistic reason cell-of-origin matters so much (see cell-state-differentiation-block and the Q4 beat) and a plausible root of the HOXA-high versus HOXA-low split.
The reader is also a writer-blocker — and that is the part that matters
The neatest experiment in this step is structural. The solution structure of the KMT2A CXXC-DNA complex explained the methylation discrimination, and structure-guided point mutants placed into MLL-AF9 did something more specific than 'lose binding': they allowed specific CpGs at Hoxa9 to become DNA-methylated, which was followed by H3K9 methylation, loss of Hoxa9 transcription, loss of immortalisation, and failure to cause leukemia in mice 5.
So CXXC is not merely an address label. By occupying those CpGs it actively protects them from DNMT activity and from the H3K9me3 silencing that follows. Domain-swap experiments make the same point from the other direction: the DNMT1 CXXC domain can substitute for MLL's inside MLL-AF9 and still cause leukemia, whereas the CGBP/CFP1 CXXC domain cannot — even though both bind unmethylated DNA, and even though every swapped fusion still reached the Hoxa9 locus. What separated them was which CpGs got protected from methylation [E1 counter-evidence].
This reframes step 1. Recruitment is cheap and promiscuous; methylation protection at the right CpGs is the selective event. It also predicts that a hypomethylating agent and a KMT2A fusion should be at least partly redundant in their effect on these loci, which is a testable and clinically relevant prediction.
Spreading: targeting does not stop at the promoter
At a subset of target genes, KMT2A::AFF1 binding spreads from the promoter into the gene body, carrying menin binding with it, raising H3K79me2/3, disrupting the normal H3K36me3 pattern — and these spreading regions coincide with unmethylated CpG stretches inside the gene body 4. Two things follow. First, the CXXC read-out operates across whole loci, not just at promoters, so intragenic unmethylated CpG islands are latent fusion-binding sites (this is how the PROM1 intragenic enhancer is reached; see mechanism-enhancer-usage-and-hijacking-kmt2a-r). Second, spreading genes are the DOT1L-inhibitor-sensitive genes, which makes spreading a candidate biomarker rather than a curiosity 4.
The bad news for a fusion-selective drug
The AT-hooks and CXXC are wild-type KMT2A sequence. There is no fusion-specific epitope here — a CXXC-directed ligand would hit wild-type KMT2A, KMT2B and (depending on chemotype) the other CXXC-domain proteins DNMT1, TET1/3, CXXC1/CFP1 and MBD1. The existing node target-kmt2a-cxxc-dna-interface covers the druggability question; the point to carry forward is that step 1 of the mechanism is the least fusion-specific step in the whole chain, and therefore probably the worst place to intervene despite being genetically the most clearly required.
| step | 1 |
|---|---|
| chain | mechanism-kmt2a-fusion-causal-chain |
| retained kmt2a region AA | approximately 1-1400 |
| reader specificity | non-methylated CpG dinucleotide; discriminates against 5-methylcytosine |
| fusion specific surface | false |
| rate limiting | genetically required but not fusion-selective; poor intervention point |
domains responsible
[ "AT-hook 1-3", "CXXC (MT) domain", "basic region" ]
related targets
[ "target-kmt2a-cxxc-dna-interface" ]
cxxc dna kd conditions
{
"_why": "Added by king sweep 2026-09-21 (cycle 46). This entry quotes the KMT2A CXXC-DNA 44 uM. That number is a property of five choices, not of the molecule, and quoting it bare is how this graph manufactured a 1,300-fold 'discrepancy' with the menin contact.",
"value": "44 uM",
"construct": "isolated CXXC domain, KMT2A 1147-1203, untagged, 15N-labelled",
"ligand": "DNA pal oligonucleotide, one CpG element",
"method": "NMR titration",
"buffer": "50 mM phosphate, pH 7.1, 300 mM NaCl, 2 mM DTT, 25 C",
"nacl_mM": 300,
"source_pmid": "20010842",
"source_quote": "\"Determination of the dissociation constants by NMR for the CXXC domain (MLL 1147-1203) and mutants was performed by titration of 15N-labeled protein with concentrated solutions of DNA pal oligonucleotide. Experiments were carried out in 50 mM phosphate buffer (pH 7.1), 300 mM NaCl, 2 mM DTT at 25 C.\" -- Cierpicki 2010 NSMB, Methods, fetched and read from PMC2908503 on 2026-09-21. The result sentence reads \"The wild type protein yielded a Kd of 44 uM\".",
"this_is_the_high_salt_end": "300 mM NaCl is roughly twice physiological ionic strength and SIX TIMES the 50 mM the menin-MBM1 53 nM was measured in. The same domain measured tagless at 150 mM NaCl gives 5.7 uM -- 7.7-fold tighter from salt alone.",
"do_not_compare_naively": {
"against": "menin-MBM1 53 nM (PMID 20961854), measured at 50 mM NaCl by FP",
"best_matched_pair": "108-fold, not ~1,000-fold",
"see": "ana-the-menin-affinity-comparison-was-made-at-one-sixth-the-salt"
},
"added": "2026-09-21"
}- The AT-hooks and CXXC are wild-type KMT2A sequence, so any direct ligand hits wild-type KMT2A/KMT2B and the wider CXXC family — an on-target toxicity problem, not a chemistry problem.
- Redundancy: AT-hook deletion only partially impairs transformation, so blocking one DNA-contact element leaves the other functional; a CXXC-selective inhibitor may be buffered by AT-hook-mediated retention.
- Because targeting reads the methylome, a cell that shifts its methylation landscape under therapy can redirect the fusion to a different, possibly still-oncogenic target set rather than losing the program.
- Is the fusion's target set fully predictable from the cell-of-origin methylome plus CpG-island annotation? If yes, a purely computational model should predict HOXA-high versus HOXA-low from the pre-leukemic methylome, which is directly testable on existing infant-ALL methylation arrays.
- Do intragenic unmethylated CpG islands predict where spreading occurs, and therefore which genes are DOT1L-inhibitor-sensitive, well enough to be a clinical biomarker?
- Can a CXXC-directed ligand ever be selective, given DNMT1, TET1/3, CXXC1 and MBD1 all carry related domains? If not, step 1 should be formally deprioritised rather than left on the target list.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Point mutations in the KMT2A CXXC domain that abolish binding to non-methylated CpG DNA also abolish transactivation in vivo and myeloid transformation by an MLL fusion protein, without disturbing subnuclear localisation.
Binding to nonmethylated CpG DNA is essential for target recognition, transactivation, and myeloid transformation by an MLL oncoprotein · Ayton PM, Chen EH, Cleary ML · Molecular and Cellular Biology 24(23):10470-8 2004What this evidence cannot show
Retroviral overexpression and in-vitro myeloid immortalisation (serial replating) in mouse bone marrow, not an in-vivo leukemia endpoint; the fusion studied is not KMT2A::AFF1. Structure-function mutagenesis cannot exclude that the CXXC point mutants also disrupt an unmeasured protein-protein contact.
Evidence that argues against this (1)
- Binding unmethylated CpG is necessary but NOT sufficient. Swapping the MLL CXXC domain for those of DNMT1, CGBP/CFP1 or MBD1 inside MLL-AF9 showed that DNMT1 CXXC rescues leukemogenesis but CGBP CXXC does not, even though both bind unmethylated DNA — and all swapped fusions still reached the Hoxa9 locus. The discriminating property was which specific CpG residues were protected from methylation, so 'CpG reader' understates the mechanism. Risner LE et al. 2013
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The solution structure of the KMT2A CXXC domain bound to DNA shows how it discriminates non-methylated from methylated CpG; structure-guided point mutations introduced into MLL-AF9 increased DNA methylation of specific CpGs at Hoxa9, increased H3K9 methylation, decreased Hoxa9-locus transcription, and abolished both immortalisation and the ability to induce leukemia in mice.
Structure of the MLL CXXC domain-DNA complex and its functional role in MLL-AF9 leukemia · Cierpicki T, Risner LE, Grembecka J, et al. (Zeleznik-Le NJ, Bushweller JH, senior authors) · Nature Structural & Molecular Biology 17(1):62-8 2010What this evidence cannot show
KMT2A::MLLT3 in mouse bone-marrow transplant, not KMT2A::AFF1 in a lymphoid context; the NMR structure is of the isolated domain with a short duplex, so the in-vivo footprint and any contribution of the flanking basic region are inferred. A single locus (Hoxa9) carries most of the mechanistic weight. LADDER POINTER added 2026-09-21 by Q82: this contact has TEN published dissociation constants spanning 44 uM to 33 nM across construct (1147-1203 to 1147-1244), tag (none / GST), DNA (12-bp one-CpG to 52-bp six-CpG) and ionic strength (150 to 300 mM). The nearest-to-physiological tagless value is 5.7 uM (PMID 34280370, 150 mM NaCl). Do not quote any single figure as the affinity - see ana-cxxc-dna-affinity-ladder. The same caveat already notes that any contribution of the flanking basic region is inferred; that region is Q03164 1204-1244, it is present in the construct behind the 33 nM figure and absent from the one behind 44 uM, and its contribution has never been measured - see interface-kmt2a-mt-bcb-domain-1147-1244.
Evidence that argues against this (2)
- A solution structure showing CXXC discriminates non-methylated from methylated CpG establishes the reading chemistry. It does not establish that methylation state is what SELECTS targets in a cell, where unmethylated CpG islands number in the tens of thousands and the fusion occupies a small subset — so the domain explains exclusion rather than selection. Nature Structural & Molecular Biology 2010
- The affinity-magnitude form of the too-weak-to-hold argument does not survive. The 44 uM figure is the extreme of a ten-measurement series and was compared with a menin figure taken at one sixth the ionic strength. What survives, and is stronger, is that affinity rank does not predict function in this system at all: the CGBP/CXXC1 domain binds DNA more tightly than the DNMT1 domain and cannot support MLL-AF9 leukaemogenesis while the weaker DNMT1 domain can, and every swapped domain including two non-binders still targeted the fusion to Hoxa9 by ChIP. Risner LE et al. 2013
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Deletion of either the AT-hook motifs or the CXXC/methyltransferase-homology domain of KMT2A substantially impairs myeloid immortalisation by KMT2A::MLLT1, and the CXXC domain binds DNA non-sequence-specifically, indicating that transformation requires multiple cooperating DNA-binding structures in the retained KMT2A portion.
The oncogenic capacity of HRX-ENL requires the transcriptional transactivation activity of ENL and the DNA binding motifs of HRX · Slany RK, Lavau C, Cleary ML · Molecular and Cellular Biology 18(1):122-9 1998What this evidence cannot show
Deletion (not point) mutants, in-vitro myeloid immortalisation only, KMT2A::MLLT1 rather than KMT2A::AFF1; 'substantially impaired' rather than abolished, so the AT-hooks are contributory rather than strictly essential in this assay.
Evidence that argues against this (1)
- The paper's headline finding is about the PARTNER, not the KMT2A side: the oncogenic capacity of HRX-ENL requires the transactivation activity of ENL as well as the DNA-binding motifs of HRX. Citing it for AT-hook and CXXC deletion effects alone takes half of a two-part result, and the ENL half is the part this graph's elongation model depends on. Molecular and Cellular Biology 1998
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
KMT2A recruits RNA polymerase II to unmethylated CpG-rich promoters via its CXXC domain and activates transcription through AEP/SEC, DOT1L and p300/CBP; leukemic fusion proteins constitutively activate the same CpG-rich promoter set, making unmethylated-CpG promoter activation the central circuit of oncogenic self-renewal.
Activation of CpG-Rich Promoters Mediated by MLL Drives MOZ-Rearranged Leukemia · Miyamoto R, Okuda H, Kanai A, et al. (Yokoyama A, senior author) · Cell Reports 32(13):108200 2020What this evidence cannot show
The paper's primary subject is MOZ/KAT6A-rearranged leukemia; the KMT2A-CXXC claims are the shared mechanistic backbone rather than the headline result. Mouse progenitor transformation assays and cell lines.
Evidence that argues against this (1)
- This result is in MOZ-REARRANGED leukaemia and concerns wild-type MLL driving CpG-rich promoter activation there. Importing it to KMT2A-fusion targeting crosses fusion family and swaps the role of MLL from fusion partner to cooperating wild-type protein, which the entry does not mark. Cell Reports 2020
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
At a subset of targets KMT2A::AFF1 binding spreads from the promoter into the gene body together with menin, raising H3K79me2/3 and disrupting H3K36me3, and these spreading regions coincide with unmethylated CpG stretches in the gene body; spreading-target genes are preferentially downregulated by DOT1L inhibition.
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. (Milne TA, senior author) · Cell Reports 18(2):482-495 2017What this evidence cannot show
ChIP-seq in a small number of t(4;11) cell lines (SEM/RS4;11-class) plus limited primary material; 'spreading' is a threshold-dependent computational call, and the causal direction between spreading and expression is correlative in this dataset.
Evidence that argues against this (1)
- Spreading from promoter into gene body with menin, raising H3K79me2/3, is a description of where the fusion goes once bound rather than of how it is addressed. If targeting is the entry's subject, spreading is downstream of it, and the same source shows these sites are distinct from super-enhancers — which removes the obvious alternative addressing mechanism without supplying one. Cell Reports 2017
CXXC-DNA binding is the second candidate menin-independent tether
the retained CXXC domain reads the unmethylated CpG regions where spreading occurs
Counter-evidence (Risner 2013, PMID 23990460): the DNMT1 CXXC domain substituted into MLL-AF9 gives colony formation and in vivo leukemogenesis similar to MLL-AF9, and all substituted domains still target Hoxa9 - what is required is protection of specific CpGs from methylation, not the MLL CXXC surface as such
Constrains the CXXC-targeting mechanism: locus targeting was preserved in every swap, functional transformation was not.
quantifies how weak the CXXC contribution is on its own
step 2 follows step 1 in the causal chain
Counter-evidence (Risner 2013, PMID 23990460): heterologous CXXC and even non-DNA-binding MBD domains swapped into MLL-AF9 still target the fusion to Hoxa9 in primary progenitors, so CXXC is a local methylation guard, not the addressing module
MLL AT-hooks were measured binding cruciform DNA, structure not sequence, not AT-rich minor groove
The fusion's CpG-island target set is a read-out of the methylation landscape; TET1 is one of the enzymes that sets it.
A fusion whose address survives an overexpressed rival reader is a fusion held by more than its CXXC domain - which is what this mechanism entry already argues.
The CXXC-plus-AT-hook targeting mechanism entry reasons from the 44 uM figure; the ladder shows that figure is the extreme of a series and the argument must be re-founded on specificity.
Withdraws the affinity-magnitude form of the too-weak-to-hold argument that this mechanism entry states, and supplies the replacement.
step 1 - unmethylated-CpG address; step measured, arrow to step 2 inferred