Oncofusion

start / graph / mech-men1-and-fancd2-act-in-distinct-crosslink-repair-processes

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q36-menin-interactome

The menin-FANCD2 bridge is dead: it was tested in 2008, the mutation spectra differ, and there is no genetic interaction

also called MEN1 FANCD2 distinct pathways · Marek 2008 · menin Fanconi epistasis

Because cells lacking menin and cells lacking Fanconi proteins are both killed easily by the same DNA-damaging drugs, people assumed the two work together. In 2008 a group tested that directly in fruit flies and found the two genes leave completely different kinds of mutation behind, and that changing menin up or down does nothing to how a Fanconi-deficient fly responds to the damage.

The same thing, for a scientist

This is the entry the menin-FANCD2 story needs and the graph did not have. **PMID 18258493 (Marek, Kottemann, Glazer, Bale; DNA Repair 2008)** set out explicitly to determine whether MEN1 and the FA genes function cooperatively in one interstrand-crosslink repair process or in distinct ones, and concluded distinct [E1]. Two independent lines. **Mutation spectrum:** in a new in vivo supF reporter, MEN1 mutant flies were extremely prone to single-base deletions within a homopolymeric tract while FANCD2 knockdown flies were indistinguishable from wild type; in the complementary lats assay, which detects large events, FANCD2 mutants showed a high frequency of large deletions and loss of heterozygosity while large deletions were rare in MEN1 mutants [E1][E2]. The two genes fail in opposite directions. **Modifier test:** FANCD2-RNAi flies develop a rough-eye phenotype after cisplatin; neither over-expression of MEN1 nor homozygous MEN1 loss changed that phenotype at any cisplatin dose [E3]. The authors' own limitation is stated plainly and must be carried: because MEN1 loss alone produces no eye phenotype after damage, *"additive or epistatic relationships cannot be evaluated based solely on these genetic interaction studies"* [E3]. So this is a **modifier screen returning a clean negative**, not a formal epistasis experiment. It is nonetheless the only designed test of the hypothesis that exists, in any organism, and it points away from it. The paper also names the two escape routes it cannot close, and both are written into the surviving hypothesis rather than quietly dropped.

Go deeper
The two results, in the authors' own terms

1. Mutation spectra diverge. The supF reporter is small enough to sequence and therefore resolves mutation type, but cannot see large rearrangements. The lats reporter is comprehensive for frequency but gives no spectrum. Running both is what makes the result interpretable 12.

supF (point/small events)lats (all events)
MEN1 mutantstrongly hypermutable; single-base deletions in a homopolymeric runsimilar to supF, i.e. most MEN1 mutations are the small deletions
FANCD2-RNAiindistinguishable from wild type~4x wild type after 0.004% nitrogen mustard; high LOH, i.e. large deletions

The authors' reading: "the fact that FANCD2-RNAi mutants exhibited a mutation frequency similar to wild type at the supF locus but had an elevated mutation frequency at the lats locus suggests that deficiency of FANCD2 causes an increase in large-scale rearrangements" 2. And for MEN1: "These findings indicate that MEN1 normally functions in a set of repair processes different from the Fanconi genes" 1.

2. No modifier interaction. FANCD2-RNAi driven by ey-GAL4 gives a rough eye after crosslinker treatment. Crossing in UAS-MEN1 over-expression, or making the flies homozygous MEN1 mutant, "[n]either ... changed the proportion of FANCD2 mutants with a rough eye phenotype" across a cisplatin dose range 3.

What the authors explicitly could not exclude — and it is where the fifth explanation has to live

The discussion does not overclaim, and the escape routes are theirs, not mine.

a. A mammal-versus-fly difference. "Because these immunoprecipitation assays were performed in MEFs, these findings may reflect differences between the mammalian and Drosophila repair systems." 4 This is real: Drosophila has no FANCA, FANCC, FANCE, FANCF or FANCG orthologue, so the fly FA pathway is a reduced one, and the fly menin orthologue is not human menin.

b. The S-phase checkpoint, not repair. "[I]t is possible that these genes function together in the S-phase checkpoint, but not in the repair of interstrand crosslinks." 4 Both proteins are documented IR-inducible S-phase checkpoint components, and for FANCD2 the checkpoint function runs through ATM phosphorylation of Ser-222, which is a modification independent of the K561 monoubiquitination that defines the canonical FA repair pathway (gene-fancd2). So this route is not hand-waving: it names a specific, separable arm of FANCD2 biology, and it is the only version of the menin-FANCD2 bridge that this paper leaves standing.

c. A quantitative residue. The paper concedes that "[t]he data from the lats assay do not exclude the possibility that menin also participates in a pathway that prevents large deletions, possibly the Fanconi pathway", on the grounds that MEN1 mutants' absolute rate of large deletions is elevated even though the proportion is not 2.

This kills the bridge at the FANCD2 level. What the complex-level signal is then caused by is unknown

Stated plainly, because hedging here would be the worse error: the menin-FANCD2 bridge, as an explanation for this graph's Fanconi selectivity signal, is dead. The physical association was reported once in 2003 in mouse fibroblasts; the functional prediction that follows from it was tested in 2008 and failed on two independent read-outs; and the project's own dependency data put the selectivity in the FA ubiquitin-transfer module rather than in the ID2 effector that menin actually binds.

The one number that might have argued the other way has now been withdrawn by the analysis that produced it. A dependency triage reached FANCD2 independently through the MEN1 bait at a difference of -0.153, more required in the lymphoid half (-0.70) than the myeloid (-0.48). The permutation null it had skipped was then computed - 50,000 two-sided label permutations per gene, same 11 against 43 lines - and FANCD2 gives p = 0.1241: not significant uncorrected, second weakest of the five genes in its own shortlist, and nothing in that shortlist survives Bonferroni over the 254 genes scored 6. The convergence originally described for this bridge had three legs. It now has two, and the gene-level dependency leg is the one that went.

So the honest shape is: closed at the FANCD2 level, open at the complex level. The complex-level Fanconi result - the FA ubiquitin ligase complex ranking 2nd of 719 Complex Portal complexes, median member rank 125 against a uniform expectation of 8,864 - was measured on a different unit by a different analysis and is untouched by any of this. A set of genes can shift together in a way no single member does alone; that is the argument for complex-level testing in the first place.

What, then, causes the complex-level selectivity? Honest answer: unknown. The candidates, with nothing yet distinguishing them:

  1. A sub-module function unrelated to crosslink repair. The signal sits on FANCF, FANCL, UBE2T, FANCA, FANCG and FAAP100 - the assembly and ubiquitin-transfer parts - which may be doing something in KMT2A-rearranged cells that is not monoubiquitinating FANCD2. This is mech-fa-core-complex-heterogeneity-kmt2a's second surviving reading and it has never been pursued.
  2. A B-lymphoid property rather than a fusion property. Both analyses find the effect stronger in the lymphoid half, and non-KMT2A B-ALL lines already showed roughly half the same enrichment. The lymphoid half is four lines in both.
  3. CRISPR reagent quality across the FA gene set - never excluded, and it would explain the subunit spread and FANCD2's shallowness at once.
  4. A shared upstream state: deep menin dependence and FA-module dependence as two read-outs of one continuous 'how fusion-driven is this line' variable that a binary KMT2A label does not capture. This is the null for the r = 0.670 correlation and nothing has excluded it.
  5. A genuine collective effect with no single responsible member - which is what a complex-level signal with no significant member literally is, and which may be the answer rather than a failure to find one.

The correct status for the Fanconi thread after this cycle is a real complex-level measurement with no surviving explanation, now at five tested and five failed, and a FANCD2 sub-question that is closed.

Why this belongs in the Fanconi thread and not only in the menin one

This graph's Fanconi thread has four dead explanations and, before this cycle, no surviving one (hyp-kmt2a-fusion-creates-an-fa-pathway-dependency, ana-fa-module-coherence, ana-fa-aldehyde-vs-replication, risk-fa-signal-lineage-confound). A menin-FANCD2 bridge was the natural fifth candidate, and the strongest reason to take it seriously is quantitative: in this project's own co-dependency analysis, MEN1 is the single best correlate of Fanconi-module dependency across 54 leukaemia lines, r = 0.670, surviving removal of every KMT2A-rearranged line at r = 0.486 and surviving lineage-and-fusion residualisation at partial r = 0.506 5.

That number is why the bridge deserved a real test rather than a dismissal. This entry is the test, and it is negative on the mechanism the number implies. The correlation is unexplained; a physical menin-FANCD2 bridge that operates through crosslink repair is not the explanation, because that specific mechanism was assayed and did not appear.

The entry the coordinator's brief asked for — a hypothesis written so it can be killed — is hyp-menin-fancd2-bridge-explains-the-fa-selectivity-signal, and it is written at a prior that reflects this result rather than ignoring it.

Facts and figures
organismDrosophila melanogaster
verdict plainthe menin-FANCD2 bridge is refuted as an explanation for the Fanconi selectivity signal; FANCD2's own gene-level selectivity does not clear a permutation null (p = 0.1241); the COMPLEX-level Fanconi result is untouched and remains real and unexplained
fa explanations tested and failed5
verdictthe physical association reported in 2003 does not translate into a shared crosslink-repair function in the one organism where it was tested
authors stated limitationbecause MEN1 mutation alone causes no eye phenotype after damage, additive or epistatic relationships cannot be evaluated from the genetic interaction studies alone
what the depmap selectivity would have to be instead
[
  "a non-canonical function of the FA assembly and ubiquitin-transfer sub-module, unrelated to FANCD2 monoubiquitination",
  "a B-lymphoid lineage property rather than a KMT2A-rearrangement property; both analyses find the lymphoid half stronger and both rest on four lymphoid lines",
  "CRISPR reagent quality across the FA gene set, which the source analysis states it could not exclude",
  "a shared upstream state: deep menin dependence and FA-module dependence as two read-outs of one continuous fusion-dependence variable that a binary KMT2A label does not capture",
  "a genuine collective effect with no single significant member, which is what a complex-level signal without a significant gene literally is",
  "unknown - this is the honest answer and the one the ledger should carry"
]
assays
[
  {
    "name": "supF in vivo mutagenesis reporter",
    "detects": "point mutations and small indels; blind to large rearrangements",
    "men1_result": "strongly hypermutable, single-base deletions in a homopolymeric tract",
    "fancd2_result": "frequency and spectrum similar to wild type"
  },
  {
    "name": "lats tumour-suppressor mutation assay",
    "detects": "all inactivating events including large deletions and LOH",
    "men1_result": "similar to supF, i.e. dominated by small deletions; large deletions rare",
    "fancd2_result": "about 4x wild type after 0.004% HN2; high LOH"
  },
  {
    "name": "ey-GAL4 FANCD2-RNAi rough-eye modifier test with cisplatin",
    "detects": "genetic modification of FANCD2 crosslinker sensitivity",
    "men1_result": "neither UAS-MEN1 over-expression nor MEN1 homozygous loss changed the phenotype",
    "fancd2_result": "rough eye after crosslinker, as previously reported"
  }
]
surviving alternatives
[
  "a mammal-specific interaction not reconstructed in Drosophila",
  "a shared role in the IR-induced S-phase checkpoint rather than in ICL repair",
  "a quantitative menin contribution to large-deletion avoidance that the proportional LOH read-out cannot resolve"
]
level split
{
  "gene_level_fancd2": "closed; permutation p = 0.1241, and p = 0.071 on the graph's earlier analysis",
  "complex_level_fanconi": "open; FA ubiquitin ligase complex 2nd of 719, median member rank 125 against a uniform expectation of 8,864, unaffected by the gene-level null"
}
How this could bite us back
  • A negative from a modifier screen in flies is the weakest kind of strong evidence. Treating it as closing the question would be the mirror of the error it corrects.
  • If the surviving S-phase-checkpoint reading is right, a menin inhibitor could perturb checkpoint function in normal dividing tissue without any FA-pathway phenotype to warn of it — a liability no current assay in this programme would detect.
Still open
  1. Has anyone tested the S-phase checkpoint reading directly — does menin loss alter FANCD2 S222 phosphorylation or the IR-induced S-phase arrest in a mammalian cell? Nothing found.
  2. Would a mammalian epistasis test (MEN1 knockout in a FANCD2-null background, mitomycin C survival) give the same negative the fly gave?
  3. If menin and FANCD2 do not share a repair pathway, what does explain the r = 0.670 co-dependency between MEN1 and the FA module across leukaemia lines?
  4. Does MEN1 loss produce the same homopolymeric single-base-deletion signature in a human cell, and is that signature detectable in MEN1-mutant endocrine tumour sequencing?
Every source behind this entry (6)
Connected to