CXXC-DNA interface of the KMT2A fusion: a validated but hard target
also called MLL CXXC domain · CpG-binding module of MLL · ZF-CxxC
Every cancer-causing KMT2A fusion keeps a small hand-like module that grips unmethylated CpG DNA, and if you break that grip the leukemia cannot form. The problem is that grabbing DNA is a shallow, charge-driven contact, which is exactly the kind of surface small-molecule drugs are worst at.
The same thing, for a scientist
The KMT2A CXXC (ZF-CxxC) domain binds non-methylated CpG dinucleotides and is retained in all leukemogenic KMT2A fusions. The solution structure of the CXXC-DNA complex (PDB 2KKF) defined the contacts and enabled structure-guided point mutants; introducing those mutants into MLL-AF9 increased CpG methylation at Hoxa9, decreased Hoxa9 transcription, abolished immortalization, and abolished leukemia in mice - genetic validation as strong as menin's [E1]. Recruitment also involves CXXC-adjacent contacts with the PAF1 complex, and has been reported to require pre-binding of wild-type MLL1 [E2] - but that requirement is contradicted by later work showing endogenous MLL1 is dispensable for MLL-rearranged leukaemia, with MLL2 the larger dependency, so the wild-type-versus-fusion window should not be argued from it. Domain-swap work further shows the DNMT1 CXXC domain can substitute for MLL CXXC in MLL-AF9 with comparable leukemogenicity, so what is required is the CpG-protection function rather than the MLL surface as such. Druggability is nonetheless poor: the interface is an extended electropositive DNA-reading surface rather than a hydrophobic pocket, the domain is small (~80 aa) and shared with paralogous CXXC proteins (KMT2B, DNMT1, TET1/3, CFP1/CXXC1, KDM2A/B, IDAX), and no inhibitor has been reported. Expect selectivity and polypharmacology problems before potency problems.
Tell me more about this picture
NMR structure of the MLL1 CXXC domain bound to a palindromic non-methylated CpG duplex.
Tell me more about this picture
The DNA-contacting surface is extended and electropositive - good for reading DNA, bad for binding a drug-like small molecule.
The genetics are excellent1 deeper
Cierpicki et al. solved the CXXC-DNA complex by NMR, used it to design DNA-binding-dead point mutants, and showed that MLL-AF9 carrying those mutants loses immortalization potential and fails to cause leukemia in mice 1. This is one of very few fusion-intrinsic surfaces with in vivo loss-of-function validation. The authors themselves flagged it as a candidate therapeutic interface.
Why loss of DNA binding kills the fusion
CXXC binding physically excludes DNA methyltransferase activity at target CpGs. Mutants that cannot bind allow CpG methylation and H3K9 methylation to spread into the Hoxa9 locus, silencing the oncogenic program 1. So the mechanism is protection of an unmethylated state, not simple tethering - which means a partial inhibitor might be enough if it lets methylation creep in over several cell divisions.
The chemistry is bad
DNA-reading interfaces bury large, flat, charged surfaces. ChEMBL contains no bioactivity series for the KMT2A CXXC domain (the only KMT2A-directed chemical matter targets the menin interface and the WDR5 WIN site). Any campaign would be starting from near zero with fragment screening or DNA-competitive peptidomimetics, against a paralog family (KMT2B, DNMT1, TET1/3, CXXC1, KDM2A/B, MBD1, CXXC4, CXXC5, FBXL19) whose members include essential genes.
CORRECTED 2026-09-21. This section previously read 'There is no reported small molecule against any ZF-CxxC domain.' There is one. Liang et al. report that dimethyl fumarate - an approved oral therapy for multiple sclerosis and psoriasis - inhibits the MLL1 CXXC-DNA complex with a relative IC50 of 0.91 +/- 0.09 uM in a single-molecule hairpin-unfolding assay, with the active metabolite monomethyl fumarate inactive at 10 uM 4.
This makes the chemistry worse, not better. DMF covalently modifies more than 2,400 cysteines and the CXXC fold coordinates two zinc ions through eight of them, so sub-micromolar inhibition by a promiscuous electrophile is the zinc-finger polypharmacology liability already listed in this entry's bite-back, now demonstrated rather than predicted. Every other member of the twelve-protein family would be expected to behave the same way; none was counter-screened. There is still no series, and 'no molecule' is the claim that has to go.
And the potency bar is higher than this entry assumed. The affinity of the natural ligand is not one number: ten published measurements span 44 uM to 33 nM, the nearest-to-physiological tagless value is 5.7 uM at 150 mM NaCl, and the most multivalent geometry gives 33 nM 5. A DNA-competitive agent may need nanomolar potency, not micromolar.
Wild-type versus fusion: is there a window?
Milne et al. showed that wild-type MLL1 needs CXXC plus PHD fingers plus H3K4me2/3 recognition for stable recruitment, whereas the fusion - which has lost the PHD fingers - depends on CXXC-DNA binding, and reported that it also depends on pre-bound wild-type MLL1 complexes 2.
Correction (2026-09-19): the second half of that claim is withdrawn as a basis for the window argument. Using independent conditional animal models, Chen et al. found endogenous MLL1 to be dispensable for MLL-rearranged leukaemia, with MLL2 - not MLL1 - the major sustaining dependency, and the two paralogues acting on distinct pathways rather than redundantly [E2 counter, Chen 2017]. A therapeutic window argued from 'the fusion needs wild-type MLL1 and we can spare it' therefore does not stand.
What survives is the first half: the fusion has lost the PHD fingers and reads chromatin with less machinery than wild-type KMT2A does, so it is plausibly more dependent on the bare CXXC-DNA contact. That remains a real, if partial, window argument.
A second correction cuts the other way on specificity. Swapping the MLL CXXC domain in MLL-AF9 for the CXXC of DNMT1 gives robust colony formation and in vivo leukemogenesis comparable to MLL-AF9, while CGBP/CFP1 and MBD1 substitutions are dead; all of them still target Hoxa9, and the discriminating property is whether specific Hoxa9 CpGs stay unmethylated [E1 counter, Risner 2013]. So the required thing is a function (protecting particular CpGs from methylation) that at least one paralogous domain performs as well as MLL's own. A ligand that merely occupies the MLL CXXC fold without abolishing that function would be inactive, and a leukaemia that can recruit an alternative CpG-protecting module has an escape route that is not usually counted in the paralog-toxicity column.
A third correction, 2026-09-21, and it removes an argument rather than adding one. Nothing in this entry quoted the CXXC-DNA affinity, which is fortunate, because the figure other entries used - 44 uM - is the extreme low-affinity point of a ten-measurement series running to 33 nM 5. The reason it matters here is that Risner's swap series, already cited in E1's counter-evidence, contains the decisive datum: the CGBP/CXXC1 domain binds unmethylated CpG 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 that bind no DNA still targeted the fusion to Hoxa9 by ChIP. DNA-binding affinity predicts neither function nor localisation in this system, so a ligand's biochemical potency against the isolated domain is a poor predictor of anything it will do in a cell. That should be the headline of any target-product profile written here.
| target class | fusion-intrinsic |
|---|---|
| modality assessed | small molecule / peptidomimetic |
| feasibility verdict | low-to-moderate |
| verdict rationale | Genetically validated as a required FUNCTION (protection of specific Hoxa9 CpGs from methylation) rather than as a uniquely required protein surface: the DNMT1 CXXC domain substitutes for MLL CXXC in MLL-AF9 with comparable leukemogenicity (Risner 2013). Flat charged protein-DNA interface, zero chemical matter, paralog family that includes essential genes. Affinity rank does not predict function in the same swap series - the tighter-binding CGBP/CXXC1 domain fails and the 34-fold weaker DNMT1 domain succeeds - and every swapped domain including two non-binders still reached Hoxa9 by ChIP, so biochemical potency against the isolated domain is a poor predictor of cellular effect. Flat charged protein-DNA interface, essentially no chemical matter (one promiscuous covalent electrophile, no series), paralog family of twelve that includes essential genes. |
| uniprot | Q03164 |
| domain boundaries | KMT2A CXXC: UniProt Q03164 CXXC-type zinc-finger feature 1147-1195 (fetched live 2026-09-21); structural construct 1147-1203 (PDB 2KKF); MT/BCB domain of the functional literature 1147-1244. CORRECTED 2026-09-21: this field previously read 'approx. aa 1147-1242 (UniProt Q03164 ZF-CXXC)', which attributed the MT-domain boundary to the UniProt feature. They are different objects. |
| chemical matter | No series in ChEMBL for the KMT2A CXXC domain. CORRECTED 2026-09-21: ONE compound is reported in the primary literature - dimethyl fumarate, an approved drug, relative IC50 0.91 +/- 0.09 uM against the MLL1 CXXC-DNA complex in a single-molecule assay (PMID 34280370). It is a promiscuous covalent electrophile with no counter-screen against the other eleven CXXC domains and no cellular data. See ana-the-cxxc-interface-has-exactly-one-published-inhibitor-and-it-is-an-approved-drug. |
| paralog risk note | DNMT1 CXXC is not only an off-target liability - it is functionally sufficient in place of MLL CXXC in MLL-AF9 (Risner 2013, PMID 23990460), so functional substitution is a resistance hypothesis as well as a selectivity problem. |
| affinity note | Ten published dissociation constants span 44 uM to 33 nM (ana-cxxc-dna-affinity-ladder). Nearest-to-physiological tagless value 5.7 uM at 150 mM NaCl; most multivalent geometry 33 nM. No measurement uses full-length KMT2A, a fusion protein or a nucleosome, so the potency a competitive ligand would need is unknown. |
structures
[ "2KKF" ]
paralog risk
[ "KMT2B", "DNMT1", "TET1", "TET3", "CXXC1/CFP1", "KDM2A", "KDM2B" ]
withdrawn claims
[ "Fusion recruitment requires pre-binding of wild-type MLL1 complexes (Milne 2010) - contradicted by Chen 2017 (PMID 28609655), which found endogenous MLL1 dispensable for MLL-rearranged leukaemia and MLL2 the larger dependency." ]
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"
}- Any CXXC ligand would likely hit KMT2B, DNMT1, TET1/3 and CXXC1 - a polypharmacology profile with real hematopoietic and developmental toxicity risk.
- Zinc-coordinating chemotypes carry generic off-target liability across the zinc-finger proteome.
- If the fusion can be recruited by partner-driven mechanisms in some cell states, CXXC blockade alone may be escapable.
- Functional substitution, not just polypharmacology: the DNMT1 CXXC domain can replace MLL CXXC in MLL-AF9 and still cause leukaemia (Risner 2013). A perfectly selective MLL-CXXC ligand may therefore be escapable by a module the cell already expresses.
- The one reported inhibitor of this interface is a promiscuous cysteine-reactive electrophile hitting a zinc-coordinating domain, which is the polypharmacology liability this entry already lists - now demonstrated. A covalent CXXC agent is closer to a domain denaturant than a competitive inhibitor, since every cysteine mutant in the structural literature unfolds the domain.
- Can a fragment screen find any pocket on the KMT2A CXXC domain, or is a DNA-competitive covalent approach (targeting a cysteine of the CxxC zinc-binding cluster) the only route?
- Does partial, reversible CXXC inhibition let CpG methylation creep into HOXA9 within a clinically useful number of cell divisions, or does the locus rebound?
- If the requirement is CpG protection rather than the MLL CXXC surface itself (Risner 2013), can a KMT2A-r leukaemia escape a CXXC ligand by recruiting DNMT1's own CXXC-containing machinery, or another ZF-CxxC protein, to the same loci? This is a testable resistance hypothesis and would be cheap to model with the existing domain-swap constructs.
- Does dimethyl fumarate inhibit the other eleven human CXXC domains at similar concentrations? That counter-screen would almost certainly close the selectivity question negatively, and it has never been run.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The MLL CXXC-DNA interface is structurally defined (PDB 2KKF) and structure-guided DNA-binding mutations abolish MLL-AF9 immortalization and leukemogenesis 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. · Nature Structural & Molecular Biology 2010What this evidence cannot show
Mouse retroviral MLL-AF9 model; point mutants may perturb folding as well as DNA binding, and the ALL-relevant KMT2A::AFF1 fusion was not tested.
Evidence that argues against this (1)
- The MLL CXXC domain is functionally replaceable, which weakens the case that THIS surface is the target. Risner et al. swapped the MLL CXXC domain out of MLL-AF9 and put in CXXC domains from DNMT1, CGBP/CFP1 and MBD1, or the MBD domain of MBD1. MLL(DNMT1 CXXC)-AF9 showed robust in vitro colony formation and in vivo leukemogenesis SIMILAR TO MLL-AF9; the CGBP and MBD1 versions were dead. All substituted domains still targeted the fusion to the Hoxa9 locus, and the discriminating property was not DNA-binding affinity per se but whether specific CpG residues at Hoxa9 were protected from methylation. Two consequences for this entry. First, the genetic validation in E1 shows that CpG-protection function is required, not that the MLL CXXC protein surface specifically is - a ligand would have to block that function, not merely bind that fold. Second, DNMT1's CXXC is not an off-target liability in the ordinary sense; it is a domain that can DO THE JOB, which raises the possibility of functional substitution as an escape route rather than only a toxicity risk. Risner LE et al. 2013
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL1 fusion proteins lack the PHD fingers and require CXXC-domain DNA recognition plus pre-binding of a wild-type MLL1 complex for stable HOXA9 association, creating a recruitment difference between wild-type and fusion.
Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis · Milne TA, Kim J, Wang GG, et al. · Molecular Cell 2010What this evidence cannot show
Domain-swap and ChIP experiments in cell lines; the requirement for wild-type MLL1 pre-binding has been debated and depends on locus and cell context.
Evidence that argues against this (1)
- The wild-type-MLL1 premise of the therapeutic-window argument has been contradicted. Chen et al. used independent conditional animal models and found that endogenous MLL1 is DISPENSABLE for MLL-rearranged leukaemia. Deleting the closest paralogue Mll2 alone significantly impacted survival of MLL-AF9-transformed cells, with additional Mll1 loss further reducing viability - but the collaboration was shown to be through regulation of distinct pathways rather than redundancy, and MLL2 rather than MLL1 was the major dependency. The recruitment model in E2, in which the fusion depends on pre-bound wild-type MLL1 complexes, cannot be assumed as the basis of a wild-type-versus-fusion window. Chen Y et al. 2017
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PDB entry 2KKF is a solution NMR structure of the MLL CXXC domain bound to palindromic CpG DNA.
PDB 2KKF: Solution structure of MLL CXXC domain in complex with palindromic CpG DNA · RCSB Protein Data Bank 2009What this evidence cannot show
NMR ensemble; no resolution figure and side-chain placement at the periphery is loosely restrained. Verified to resolve 2026-09-17.
Evidence that argues against this (1)
- A solution NMR structure of an ISOLATED domain bound to a PALINDROMIC CpG oligonucleotide is a designed substrate, not genomic context. Two limits follow for anything built on these coordinates: NMR of a small domain does not report how the flanking basic region contributes in the full-length protein, and a palindromic duplex is symmetric in a way real CpG islands are not, so specificity determinants seen here may not be the ones operating at a target promoter. RCSB Protein Data Bank 2009
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Dimethyl fumarate, an approved oral therapy for relapsing multiple sclerosis and psoriasis, inhibits the MLL1 CXXC-DNA complex with a relative IC50 of 0.91 +/- 0.09 uM measured as the dose-dependent loss of single-molecule hairpin unfolding time, while its active metabolite monomethyl fumarate is inactive at 10 uM.
Dynamics and inhibition of MLL1 CXXC domain on DNA revealed by single-molecule quantification. · Liang L, Ma K, Wang Z, et al. · Biophys J 2021What this evidence cannot show
A relative IC50 from a four-parameter logistic fit to single-molecule unfolding times (N greater than 20 traces from more than 3 molecules per concentration), not a conventional biochemical IC50; isolated synthesised domain on naked DNA, no cell, no fusion protein, no counter-screen against any other CXXC domain, and the covalent mechanism is inferred from DMF's known chemistry rather than demonstrated on this domain. The authors present it as a proof of concept for the assay. Full text read 2026-09-21 via the PMC HTML route (PMC8392069).
Evidence that argues against this (1)
- The authors note that chemical proteomics has never identified CXXC-family proteins as DMF targets, and DMF is rapidly converted after oral dosing to MMF, which is inactive here - so the clinical relevance of this activity is unsupported in both directions. It is a molecule, not a lead, and it does not change the feasibility verdict of this entry. Risner LE et al. 2013
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The KMT2A CXXC-DNA affinity is not a single number: ten published measurements span 44 uM to 33 nM across construct span (1147-1203 to 1147-1244), tag state, DNA ligand (12-bp one-CpG to 52-bp six-CpG) and ionic strength (150 to 300 mM), with the nearest-to-physiological tagless value 5.7 uM at 150 mM NaCl.
MLL protects CpG clusters from methylation within the Hoxa9 gene, maintaining transcript expression. · Erfurth FE, Popovic R, Grembecka J, et al. · Proc Natl Acad Sci U S A 2008What this evidence cannot show
The series is assembled in ana-cxxc-dna-affinity-ladder from six primary papers read in full on 2026-09-21; this evidence item cites the one that both contributes five of the ten points and states the discrepancy in print. Every measurement is of an isolated 57- to 98-residue domain on naked DNA; none uses full-length KMT2A, a fusion protein or a nucleosome, so none of them is the affinity of the species a drug would have to compete with.
Evidence that argues against this (1)
- Because no measurement uses the multivalent species that exists in a cell, the potency bar cannot actually be set from this series either. The honest position for a target-product profile is that the required potency is unknown and could be anywhere from micromolar to sub-nanomolar depending on how much avidity the assembled fusion contributes - which is an argument for measuring it before starting, not for assuming either end. Liang L et al. 2021
- related to →Step 1 — How the fusion finds its genes: AT-hooks and the CXXC domain reading unmethylated CpG
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
Counter-evidence (Chen 2017, PMID 28609655): endogenous MLL1 is dispensable for MLL-rearranged leukaemia and MLL2 is the larger dependency, so the wild-type-versus-fusion window for CXXC cannot be argued from a requirement for pre-bound wild-type MLL1
Corrected basis: the CXXC dependency is a CpG-protection function that a paralogous domain can supply (Risner 2013, PMID 23990460), which makes functional substitution a resistance hypothesis rather than only a selectivity risk
The CGBP CXXC domain cannot substitute for the KMT2A CXXC domain in MLL-AF9 despite binding unmethylated DNA and reaching Hoxa9 - so DNA binding is not the functional discriminator.
supplies the affinity and nine contact residues that dossier lacked
- part of →KMT2A / MLL1 — the gene that breaks
CXXC domain, retained in every fusion
KMT2A and KMT2B CXXC domains bind DNA indistinguishably in vitro yet go to non-overlapping genes in cells
2KKF is the structural starting point for interface ligand design
readiness layer on top of the CXXC interface entry
The druggability dossier's potency-bar reasoning depends on the ladder rather than on a single number.
Sharpens the dossier's existing functional-substitution warning with the affinity numbers that make it counterintuitive.
Corrects 'there is no reported small molecule against any ZF-CxxC domain'. One is reported; there is still no series.