EWS::FLI1 builds its own enhancers at GGAA microsatellites, and a single inherited base change that lengthens one repeat raises the fusion's binding and the patient's risk. KMT2A fusions occupy pre-existing, invariant CpG islands - which is why this disease has no germline modifier of its target set, and why its mouse models work at all.
also called GGAA microsatellite address · address creation versus address occupation · rs79965208 · EGR2 susceptibility repeat · the invariant CpG address
In Ewing sarcoma the fusion protein makes new gene switches out of repetitive DNA that varies between people, and one inherited single-letter change that makes a repeat longer both raises how tightly the fusion binds and raises the person's risk of the disease. The leukaemia fusion this project studies lands on ordinary, fixed features of the genome instead, so there is no equivalent inherited dial - which is a loss for finding risk genes and a gain for making the biology reproducible.
The same thing, for a scientist
EWS::FLI1's defining property is **neomorphic address creation**: it binds GGAA microsatellites that function as response elements only for the fusion and not for wild-type FLI1, and converts non-functional repeat DNA into de novo active enhancers. Gangwal and Lessnick, reviewing their own primary result, explicitly speculated that **microsatellite polymorphisms may confer differences in susceptibility to Ewing sarcoma both between individuals and between populations** [E2]. **Grunewald et al. then showed it.** Targeted germline deep sequencing of the EGR2 locus identified 291 Ewing-associated SNPs, and at **rs79965208 the A risk allele converts an interspaced GGAT motif into a GGAA motif**, joining adjacent GGAA repeats, lengthening the consecutive GGAA run and thereby **increasing the EWSR1-FLI1-dependent enhancer activity of the sequence, with the epigenetic characteristics of an active regulatory element; EWSR1-FLI1 preferentially bound the A risk allele and global and allele-specific EGR2 expression rose**, and EGR2 knockdown inhibited proliferation and induced regression of Ewing xenografts [E1]. **KMT2A fusions have the opposite address biology.** The fusion's genomic targeting runs through the retained CXXC domain's affinity for **non-methylated CpG**, i.e. CpG islands - features that are ancient, conserved, present in mouse, and essentially invariant between individuals. **This graph's own `localization-cxxc-cpg-address` is the entry that states it.** **Three consequences follow, and they are not obvious until the two families are put side by side.** (1) **Germline modifiers of the target set exist in one family and cannot exist in the other** - nobody will find a KMT2A::AFF1 analogue of rs79965208 because there is no variable element to vary. (2) **Model organisms.** GGAA microsatellites are human-specific in their length distribution and the core response elements cannot be reproduced in mouse; CpG islands are conserved, which is a large and rarely stated reason why murine KMT2A-fusion models recapitulate target-gene biology and murine Ewing models do not. (3) **Therapeutic geometry.** An address the fusion CREATES is, in principle, removable by editing or by blocking creation; an address the fusion OCCUPIES is shared with the wild-type protein and with normal CpG-island biology, so anything that blocks it blocks normal function too - which is the reason this field targets menin rather than the CXXC-CpG contact.
The mapping
| EWS::FLI1 | KMT2A::AFF1 |
|---|---|
| binds GGAA microsatellites, which are response elements only for the fusion 2 | binds non-methylated CpG through the retained CXXC domain - the same address as wild-type KMT2A |
| the address is created: repeat DNA becomes a de novo active enhancer | the address is occupied: CpG islands are already there and already active |
| addresses are polymorphic between individuals and populations 2, demonstrated at rs79965208 1 | addresses are invariant - CpG islands do not vary in a way that would tune fusion binding |
| a germline allele raises fusion binding, raises target expression, raises disease risk 1 | no germline modifier of the fusion's target set has been described |
| the elements are not conserved in mouse, so the core response elements cannot be modelled | CpG islands are conserved, which is part of why murine KMT2A-fusion models work |
| blocking address creation would spare wild-type FLI1 | blocking the CXXC-CpG contact would hit wild-type KMT2A and normal CpG-island biology |
The last row is a defence of this project's own strategy that it has not written down. The reason the field attacks menin rather than the fusion's DNA address is usually given as 'menin is druggable and DNA is not'. The comparison supplies a better reason: the KMT2A address is shared with the wild-type protein, so it has no therapeutic window at all, whereas the Ewing address is fusion-exclusive and in that sense a better target - and it is still undrugged, which says something about how hard DNA-element targeting is even when the window is perfect.
What a germline modifier would have bought, and why this disease cannot have one
rs79965208 is a remarkable object: one inherited base, changing one motif, lengthening one repeat, increasing fusion occupancy at one enhancer, raising expression of one driver gene, and raising disease risk 1. It ties a germline susceptibility variant to a somatic oncogene through a shared physical mechanism, which is a class of result cancer genetics rarely achieves.
Nothing of that shape is available for KMT2A::AFF1, and the reason is structural rather than a gap in effort. The fusion's address is a chemical feature of the genome - unmethylated CpG - not a length-variable sequence. A germline modifier would have to act on something else entirely: on the methylation state of a target island, on the abundance of a cofactor, or on the breakpoint machinery, none of which changes the fusion's address as such.
That is worth stating positively rather than as a deficit. The invariance is why a KMT2A fusion's target set is reproducible between patients, between cell lines and between species. This project's ability to write a stable interactome and a stable target list at all rests on it, and the Ewing field's inability to do the same is not a failure of diligence.
A live instrument note found while bounding this entry
While checking whether a KMT2A germline address modifier had been reported, the two gene symbols behaved very differently through an identical query template on 2026-09-23:
| query | count |
|---|---|
KMT2A[tiab] AND "CpG island"[tiab] | 0 |
MLL[tiab] AND "CpG island"[tiab] | 13 |
KMT2A[tiab] AND microsatellite[tiab] | 6 |
MLL[tiab] AND microsatellite[tiab] | 22 |
MLL[tiab] AND germline[tiab] AND enhancer[tiab] | 0 |
KMT2A[tiab] AND germline[tiab] AND enhancer[tiab] | 2 |
CXXC[tiab] AND KMT2A[tiab] (positive control) | 8 |
A beat that had queried only KMT2A would have concluded that the CpG-island literature for this gene is empty. It is not; it is thirteen papers filed under the pre-2018 symbol. This is the project's standing MLL/KMT2A rule firing live on the exact question this entry is about, and it is recorded here as a worked instance rather than as a general warning.
The absence this entry actually asserts is narrow: no germline variant modifying a KMT2A fusion's genomic target set has been described, bounded by these queries, by titles and abstracts only, and by the vocabulary 'germline', 'enhancer', 'microsatellite' and 'CpG island'. A modifier described as a 'QTL', an 'eQTL' or a 'risk allele' without those words would be missed, and the correct reading is 'not reported in this vocabulary', not 'shown not to exist'.
| domain | engineering |
|---|---|
| where the analogy breaks | It breaks on strength of evidence and on the reach of the structural argument. rs79965208 is one locus, not a demonstrated genome-wide tuning mechanism, so the Ewing column is thinner than it reads. And the claim that KMT2A cannot have a germline address modifier is an argument from the nature of a CpG island, not a measurement: a variant that created, destroyed or altered the methylation state of a target island would modify the address by a different route, and no search was run for that shape of variant. The honest statement is that the two families' addresses differ in KIND, and that only one of them has a demonstrated inherited modifier. |
| what it predicts here | No germline variant will be found that tunes KMT2A::AFF1 occupancy at its target CpG islands by altering the element itself. If one is ever reported, it should act through methylation state or through island creation, not through element length, and that prediction is specific enough to be wrong. |
| localization axis | genomic address - which sequences the fusion occupies and whether they pre-exist it |
| epigenetic axis | non-methylated CpG recognition by CXXC versus de novo enhancer creation with active-element chromatin characteristics at a repeat |
| relationship to existing entries | fusion-ewsr1-fli1 records the GGAA-microsatellite mechanism and cites the primary paper (PMID 18626011), and raises the mouse-modelling problem in an open question. localization-cxxc-cpg-address holds the KMT2A side. Neither contains the germline-modifier result: PMIDs 26214589 and 18927503 are cited in zero node files as of 2026-09-23. The contrast between address CREATION and address OCCUPATION, and its three consequences, is not stated anywhere in the corpus. |
mapping
[
{
"bio": "EWS::FLI1 CREATES enhancers at GGAA microsatellites, which are response elements for the fusion and not for wild-type FLI1",
"analog": "the KMT2A fusion OCCUPIES pre-existing non-methylated CpG islands through the retained CXXC domain - the same address as wild-type KMT2A"
},
{
"bio": "rs79965208 converts GGAT to GGAA, lengthens the repeat, raises fusion-dependent enhancer activity, raises EGR2 expression and raises disease risk",
"analog": "no germline modifier of a KMT2A fusion's target set has been reported, and the address has no length-variable element for one to act on"
},
{
"bio": "GGAA microsatellite length is not conserved in mouse, so the core response elements cannot be modelled in standard mice",
"analog": "CpG islands are conserved, which is a large and rarely stated reason murine KMT2A-fusion models reproduce target-gene biology"
},
{
"bio": "the Ewing address is fusion-EXCLUSIVE and therefore has a perfect therapeutic window - and is still undrugged",
"analog": "the KMT2A address is SHARED with wild-type KMT2A and has no window at all, which is a better reason to target menin than 'DNA is undruggable'"
},
{
"bio": "address polymorphism gives Ewing a germline-somatic mechanistic link of a kind cancer genetics rarely achieves",
"analog": "address invariance gives KMT2A-rearranged leukaemia a reproducible target set across patients, lines and species - the thing this graph is built on"
}
]generated hypotheses
[]
bounded absence queries — the numbers behind the chart
{
"KMT2A[tiab] AND germline[tiab] AND enhancer[tiab]": 2,
"MLL[tiab] AND germline[tiab] AND enhancer[tiab]": 0,
"KMT2A[tiab] AND microsatellite[tiab]": 6,
"MLL[tiab] AND microsatellite[tiab]": 22,
"KMT2A[tiab] AND \"CpG island\"[tiab]": 0,
"MLL[tiab] AND \"CpG island\"[tiab]": 13,
"CXXC[tiab] AND KMT2A[tiab] (positive control)": 8
}- The therapeutic implication is negative for the obvious idea. The KMT2A fusion's DNA address is shared with wild-type KMT2A and with normal CpG-island biology, so an agent that blocks the CXXC-CpG contact has no therapeutic window by construction - which is why menin, a contact the wild-type protein also makes but which is chemically tractable at a defined pocket, remains the better target despite sharing the same wild-type problem.
- The Ewing address is fusion-exclusive, which is the perfect window, and it is still undrugged after three decades. A project should not assume that a perfect window makes a DNA-element or enhancer-creation mechanism druggable.
- Address invariance means there is no inherited risk dial to find, so genetic-epidemiology approaches that worked in Ewing will not transfer to KMT2A-rearranged leukaemia, and effort spent looking for one is effort lost.
- The conserved-address argument is also a warning about over-trusting mouse models in the opposite direction: conservation of the ADDRESS does not imply conservation of the COFACTORS, the chromatin state or the cell of origin, and this project's standing finding is that murine MLL-AF9 results travel badly into human infant B-lineage disease.
- Is there any germline variant that alters the methylation state, presence or size of a CpG island at a KMT2A-fusion target locus, and does it change fusion occupancy? This is the shape a KMT2A address modifier would have to take, it is not the shape anyone has searched for, and methylation QTL datasets in haematopoietic cells already exist to look in.
- How much of the Ewing field's failure to build a faithful mouse model is attributable to the non-conservation of GGAA microsatellites rather than to cell-of-origin problems? fusion-ewsr1-fli1 poses this and it remains open; the KMT2A comparison sharpens it by showing what a conserved address buys.
- Does address invariance predict low inter-patient variability in fusion occupancy? The prediction is testable on existing primary-sample CUT&RUN data across several KMT2A::AFF1 patients, and a surprising amount of variability would point at something other than the CXXC-CpG contact setting the address.
- Do any of the other 52 fusion families in this graph create their addresses rather than occupy them? NUP98 fusions locking pre-existing H3K4me3 loci and CBFA2T3::GLIS2 binding Gli-family sites both look like occupation; if EWS::FLI1 is the only creator, the two-class scheme this entry proposes is really one exception and a rule.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Targeted germline deep sequencing of the EGR2 locus in Ewing sarcoma cases and controls identified 291 Ewing-associated SNPs; at rs79965208 the A risk allele connects adjacent GGAA repeats by converting an interspaced GGAT motif into a GGAA motif, increasing the number of consecutive GGAA motifs and thereby the EWSR1-FLI1-dependent enhancer activity of the sequence, which carries the epigenetic characteristics of an active regulatory element; EWSR1-FLI1 preferentially bound the A risk allele and this increased global and allele-specific EGR2 expression; and EGR2 knockdown inhibited proliferation, clonogenicity and spheroidal growth in vitro and induced regression of Ewing sarcoma xenografts.
Chimeric EWSR1-FLI1 regulates the Ewing sarcoma susceptibility gene EGR2 via a GGAA microsatellite. · Grunewald TGP, Bernard V, Gilardi-Hebenstreit P, Raynal V, Surdez D, Aynaud MM, Mirabeau O, Cidre-Aranaz F, Tirode F, Zaidi S, Perot G, Jonker AH, Lucchesi C, Le Deley MC, Oberlin O, Marec-Berard P, Veron AS, Reynaud S, Lapouble E, Boeva V, Rio Frio T, Alonso J, Bhatia S, Pierron G, Cancel-Tassin G, Cussenot O, Cox DG, Morton LM, Machiela MJ, Chanock SJ, Charnay P, Delattre O · Nat Genet 2015What this evidence cannot show
Abstract read in full from the PubMed record on 2026-09-23. Ewing sarcoma, a non-haematopoietic solid tumour with an unresolved cell of origin; the germline association derives from a genome-wide association signal near EGR2 followed by targeted resequencing, so the causal attribution to rs79965208 rests on the functional work rather than on genetic fine-mapping alone. The effect is at one locus; the paper does not establish that GGAA-repeat polymorphism tunes the fusion's target set genome-wide.
Evidence that argues against this (1)
- This entry uses the result to argue that address polymorphism is a property of fusions that CREATE their addresses. A single demonstrated locus supports the existence claim and not the generality claim; if rs79965208 turned out to be the only such variant with a measurable effect, the contrast with KMT2A would be a contrast between one locus and zero rather than between two classes of address biology. Gangwal K et al. 2008
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
EWS/FLI interacts with GGAA microsatellites to regulate target genes including NR0B1, which is required for the oncogenic phenotype of Ewing sarcoma; microsatellites typically have no ascribed function and are sometimes considered junk DNA, so this assigns them a role as transcriptional response elements; and the authors speculate that microsatellite polymorphisms may confer differences in susceptibility to Ewing sarcoma both between individuals and between populations.
Microsatellites are EWS/FLI response elements: genomic 'junk' is EWS/FLI's treasure. · Gangwal K, Lessnick SL · Cell Cycle 2008What this evidence cannot show
This is a REVIEW/perspective by the authors of the primary study and is cited as a map plus an explicitly labelled speculation, which is how the abstract itself frames the susceptibility idea ('it is tempting to speculate'). The primary GGAA-microsatellite result is PMID 18626011, which this graph already cites in fusion-ewsr1-fli1; that paper was not re-read here. Nothing in this item is a measurement.
Evidence that argues against this (1)
- The speculation in this review was published in 2008 and took seven years to be demonstrated, and then at one locus. A project reading a between-population susceptibility claim from it should note that population-level differences in Ewing incidence have several candidate explanations and that this paper demonstrates none of them. Grunewald TGP et al. 2015
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
Counted absence and instrument note, 2026-09-23: no germline variant modifying the genomic target set of a KMT2A fusion was found under the queries KMT2A[tiab] AND germline[tiab] AND enhancer[tiab] (2 records), MLL[tiab] AND germline[tiab] AND enhancer[tiab] (0), KMT2A[tiab] AND microsatellite[tiab] (6) and MLL[tiab] AND microsatellite[tiab] (22); and the two gene symbols diverge sharply on the address vocabulary itself, with KMT2A[tiab] AND "CpG island"[tiab] returning 0 while MLL[tiab] AND "CpG island"[tiab] returns 13, against a positive control of CXXC[tiab] AND KMT2A[tiab] at 8.
PubMed esearch run through scripts/pubmed_safe.py, which retries an empty querytranslation and raises rather than returning a false zero; all counts and both gene symbols recorded in the entry body · NCBI Entrez Programming Utilities (esearch, db=pubmed) · PubMed / NCBI E-utilities 2026What this evidence cannot show
Bounded by titles and abstracts and by the vocabulary 'germline', 'enhancer', 'microsatellite' and 'CpG island'. A modifier described as a QTL, eQTL or risk allele without those words would be invisible, and a functional result appearing only in a supporting figure would be invisible regardless of vocabulary. The counts were not title-screened one by one, so the small non-zero counts are upper bounds on relevance, not confirmed hits. The correct reading is 'not reported in this vocabulary', not 'shown not to exist'.
Evidence that argues against this (1)
- The structural argument that a KMT2A germline address modifier CANNOT exist is stronger than the bounded search that accompanies it, and it is also the weaker kind of claim, because it is an argument rather than a measurement. A germline variant that altered CpG-island methylation state, or that created or destroyed a CpG island, would modify the fusion's address without being a repeat-length polymorphism - so the correct statement is that the mechanism differs, not that modification is impossible. Grunewald TGP et al. 2015
- related to →EWSR1::FLI1 (EWS-FLI1), t(11;22)(q24;q12)
the family that creates its addresses at polymorphic GGAA microsatellites
the KMT2A side of the contrast - an invariant, pre-existing, wild-type-shared address
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the index fusion, whose target set is invariant between patients and conserved in mouse
- related to →KMT2A / MLL1 — the gene that breaks
the CXXC domain whose non-methylated-CpG affinity sets the shared address
sibling entry on the same family - that one on its interface programme, this one on its addresses
if the EWS module is portable the two fusions should differ only in address; if address-dependent, in kind