The FA core complex does not move together, and that is the strongest argument against reading this as a Fanconi-pathway dependency at all
also called FA core complex spread · FANCB paradox · subunit heterogeneity
The Fanconi repair crew works as one machine: eight proteins clip together to do a single job, adding a chemical tag to a ninth protein. If a cancer cell needed that machine, losing any one of the eight should hurt it about equally. In these leukaemia cells four of the eight matter a lot, three barely matter, and the protein the machine actually acts on barely matters either — which means whatever these cells need, it is probably not the machine doing its known job.
The same thing, for a scientist
The FA core complex (FANCA, B, C, E, F, G, L, M plus FAAP20/24/100) is an obligate multi-subunit RING ubiquitin ligase whose established function is a single reaction: monoubiquitination of the FANCD2–FANCI heterodimer. A dependency on that reaction predicts that obligate subunits behave alike. In the DepMap data behind this project's FA finding they do not, by a factor of five: FANCF -0.437, FANCL -0.395, FANCA -0.357, FANCG -0.314, against FANCC -0.169, FANCE -0.123, FANCM -0.095 and FANCB -0.079 (Chronos difference, KMT2A-rearranged vs other leukaemia) [E1]. The split is not marginal — line-label permutation gives p = 0.0004 for FANCF and p = 0.136 for FANCB. Worse for the pathway reading, **the substrate is weak**: FANCD2 is -0.150 (p = 0.071, rank 411 of 8,269 by raw effect) while its obligate heterodimer partner FANCI is -0.361 (p = 0.0026, rank 15). The canonical pathway cannot be selectively required through a step whose product is dispensable. Three readings survive: the signal is in a **sub-module** rather than the complex (FANCA-FANCG and FANCF-FANCL-FAAP100, the parts that carry it); the signal is in **non-canonical functions** of those particular subunits; or the spread is **reagent quality** in the CRISPR library, which is the boring explanation and has not been excluded.
The numbers, laid out so the problem is visible
Chronos difference between 11 KMT2A-rearranged leukaemia lines and 43 other leukaemia lines, offset-corrected; p from a 5,000-draw stratified line-label permutation that holds the gene fixed and redraws the index arm 1.
| protein | role in the complex | d | line-permutation p | rank of 8,269 by raw effect |
|---|---|---|---|---|
| FANCF | core, adaptor | -0.437 | 0.0004 | 11 |
| FANCL | core, the catalytic RING E3 | -0.395 | 0.0004 | 26 |
| FANCA | core, largest subunit | -0.357 | 0.0002 | 20 |
| FANCG | core, TPR subunit | -0.314 | 0.0008 | 27 |
| FAAP100 | core, scaffolds FANCB–FANCL | -0.270 | 0.0002 | 36 |
| FANCC | core | -0.169 | 0.015 | 386 |
| FANCE | core | -0.123 | 0.018 | 498 |
| FANCM | core, translocase | -0.095 | 0.142 | 3,406 |
| FANCB | core, FANCL's own partner | -0.079 | 0.136 | 1,683 |
| — | — | |||
| FANCI | substrate, ID2 | -0.361 | 0.0026 | 15 |
| FANCD2 | substrate, ID2 | -0.150 | 0.071 | 411 |
The FANCB–FANCL–FAAP100 subcomplex is the sharpest case. Those three proteins form one module, and FANCB and FAAP100 are required for FANCL's ligase activity. In this data FANCL is -0.395, FAAP100 is -0.270 and FANCB is -0.079, which is essentially nothing. A five-fold spread inside a three-protein obligate module is not what a complex-level dependency looks like.
Why the weak FANCD2 is the hardest fact to explain away
The FA core complex has exactly one well-established catalytic output: it monoubiquitinates FANCD2 on K561 and FANCI on K523. Everything downstream — nuclease recruitment, translesion synthesis, the whole ICL repair programme — runs through ubiquitinated ID2 23.
So if KMT2A-rearranged cells are selectively dependent on the core complex because of that reaction, they must be at least as dependent on the thing the reaction acts on. They are not. FANCD2 sits at rank 411 with d -0.150 and a line-permutation p of 0.071, which does not clear any reasonable threshold, while four core subunits sit in the top 30 1.
The asymmetry within ID2 itself is stranger still: FANCI (-0.361, p = 0.0026, rank 15) against FANCD2 (-0.150, p = 0.071, rank 411), for an obligate heterodimer. FANCI does have a reported function outside ID2 — it homodimerises, and it binds UBL5 — and the FANCI–FANCD2 complex has a DNA-scanning activity that may not require ubiquitination 2. That is a real out, but it is an out that abandons the canonical pathway as the explanation.
The boring explanation, which has not been excluded
CRISPR reagent quality varies by gene. Chronos corrects for copy number and screen quality but not for how well a particular gene's guides cut, how much of the protein survives an indel in the targeted exon, or how tolerant the protein is of a truncated product. If the FANCB and FANCM guides in the Avana library happen to be poor, or if the targeted exons sit downstream of a functional core, the observed spread is an artefact of the library and not biology.
This is a testable, cheap objection and it is unresolved here, because per-guide data are not exposed through the API route this project uses. The honest position is that the core-complex spread is evidence against a pathway-level reading and is not proof of one, and anyone who wants to use the spread as an argument has to pull the per-guide DepMap data first.
Two further caveats point the same way. FANCB is X-linked, so its dosage behaves differently from the autosomal subunits. And FANCM is the subunit most often described as dissociable from the core complex, with independent functions at replication forks and in R-loop metabolism, so its low rank is the least surprising of the four.
What follows for the rest of this beat
Two things, and they pull in opposite directions.
Against. Every entry in this beat that speaks of 'the FA pathway' as the dependency is over-claiming. The defensible unit is the eight genes that actually carry the signal, and they are not a pathway — they are FANCA, FANCF, FANCG, FANCL, FAAP100 (core), UBE2T (the E2), FANCI (half of ID2) and BRIP1 (a late helicase). That set has no single published function.
For. The set that carries the signal is not random within the complex. It is FANCA-FANCG plus FANCF plus the FANCL/FAAP100 catalytic module plus the E2 that feeds it — that is, the ubiquitin-transfer machinery and its assembly scaffold, with the non-catalytic FANCC/FANCE and the semi-detached FANCM left out. A dependency on ubiquitin transfer by FANCL, licensed by UBE2T and assembled by FANCA/F/G/FAAP100, acting on a substrate other than FANCD2, would produce exactly this pattern. That is a specific, falsifiable model and it is the one this graph should carry forward.
The independent CRISPR screen that found UBE2T and FANCL as AML dependencies is consistent with that reading: it nominated the catalytic pair, not the complex 4.
| spread ratio core | 5.5 |
|---|
core complex spread — the numbers behind the chart
[
{
"gene": "FANCF",
"d": -0.437,
"line_perm_p": 0.0004,
"rank_of_8269_by_d": 11
},
{
"gene": "FANCL",
"d": -0.395,
"line_perm_p": 0.0004,
"rank_of_8269_by_d": 26
},
{
"gene": "FANCA",
"d": -0.357,
"line_perm_p": 0.0002,
"rank_of_8269_by_d": 20
},
{
"gene": "FANCG",
"d": -0.314,
"line_perm_p": 0.0008,
"rank_of_8269_by_d": 27
},
{
"gene": "FAAP100",
"d": -0.27,
"line_perm_p": 0.0002,
"rank_of_8269_by_d": 36
},
{
"gene": "FANCC",
"d": -0.169,
"line_perm_p": 0.015,
"rank_of_8269_by_d": 386
},
{
"gene": "FANCE",
"d": -0.123,
"line_perm_p": 0.018,
"rank_of_8269_by_d": 498
},
{
"gene": "FANCM",
"d": -0.095,
"line_perm_p": 0.14217,
"rank_of_8269_by_d": 3406
},
{
"gene": "FANCB",
"d": -0.079,
"line_perm_p": 0.13597,
"rank_of_8269_by_d": 1683
}
]substrate
[
{
"gene": "FANCI",
"d": -0.361,
"line_perm_p": 0.0026,
"rank_of_8269_by_d": 15
},
{
"gene": "FANCD2",
"d": -0.15,
"line_perm_p": 0.07139,
"rank_of_8269_by_d": 411
}
]surviving readings
[ "module-level dependency on the ubiquitin-transfer machinery (UBE2T-FANCL-FAAP100 plus the FANCA-FANCF-FANCG assembly arm), acting on a substrate other than FANCD2", "non-canonical functions of the individual strong subunits", "CRISPR reagent quality differences across subunits, not excluded" ]
- If the dependency is on a sub-module and not the pathway, then an inhibitor designed against the pathway concept — say, one that blocks FANCD2 ubiquitination readout — may miss the actual requirement entirely.
- Conversely, a UBE2T or FANCL inhibitor that works would disable interstrand cross-link repair across the whole patient, with the Fanconi-anaemia phenotype as the predicted on-target toxicity.
- Do the per-guide DepMap data show comparable cutting efficiency for FANCB and FANCM as for FANCF and FANCL? Until that is checked the reagent explanation stands.
- Is FANCD2 K561R knock-in tolerated in a KMT2A-rearranged line while FANCL knockout is not? That would separate the ubiquitin-transfer requirement from the canonical substrate cleanly.
- What else does the FANCL–UBE2T pair ubiquitinate in a leukaemia cell?
- Does FANCI rescue FANCD2 loss in these lines, or are the two genuinely separable here?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Across the eight obligate FA core subunits the KMT2A-selective depletion spans a factor of five (FANCF -0.437 to FANCB -0.079), the substrate FANCD2 is weak (-0.150, line-permutation p = 0.071) while its heterodimer partner FANCI is strong (-0.361, p = 0.0026), and the split does not follow the complex's architecture.
Is the Fanconi signal lineage-wide, fusion-specific, and separable from pan-essentiality? Lineage-split, subtype-control and line-label permutation re-test of the FA/BRCA enrichment · Oncofusion knowledge graph, analyses/fa-pathway-lineage-and-fusion-specificity; DepMap Chronos via Open Targets 2026What this evidence cannot show
Eleven index lines. Chronos corrects for copy number and screen quality but not for per-guide cutting efficiency, exon position or tolerance of truncated protein, and per-guide data were not pulled. FANCB is X-linked. FANCM is the subunit most often reported to act independently of the core complex.
Evidence that argues against this (1)
- The set that DOES carry the signal is coherent rather than random: FANCA-FANCG-FANCF assemble the complex, FANCL is the catalytic RING with FAAP100 as its scaffold, and UBE2T is the E2 that feeds it. Heterogeneity across a complex can mean the dependency is on a module rather than that it is noise. UniProtKB/Swiss-Prot 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The established output of the FA core complex is monoubiquitination of the FANCD2–FANCI heterodimer; FANCI promotes FANCD2 monoubiquitination by FANCL, and the FANCI–FANCD2 complex binds and scans double-stranded DNA for damage.
UniProtKB entry Q9NVI1 (FANCI) · UniProtKB/Swiss-Prot 2026What this evidence cannot show
A curated database summary of many primary papers, not a single experiment.
Evidence that argues against this (1)
- UniProt also records FANCI activities outside the canonical reaction — homodimerisation, interaction with UBL5 promoting that homodimerisation, interaction with FAN1 and POLN — so a FANCI requirement need not run through FANCD2 ubiquitination. UniProtKB/Swiss-Prot 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UBE2T is the specific E2 for the FA complex, associating with FANCL and catalysing FANCD2 monoubiquitination — so the E2 and the E3 form the catalytic pair that a module-level dependency would isolate.
UniProtKB entry Q9NPD8 (UBE2T) · UniProtKB/Swiss-Prot 2026What this evidence cannot show
Database annotation. UBE2T also has reported activity toward substrates outside the FA pathway and interacts with BRCA1.
Evidence that argues against this (1)
- If UBE2T acts on substrates beyond FANCD2, a UBE2T dependency is not evidence for an FA-pathway dependency, and the same reasoning weakens the inference from FANCL. UniProtKB/Swiss-Prot 2026
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
An independent domain-focused CRISPR screen across cancer cell lines nominated UBE2T and FANCL — the catalytic pair, not the whole complex — as the FA dependencies in acute myeloid leukaemia.
Transcriptional Silencing of ALDH2 Confers a Dependency on Fanconi Anemia Proteins in Acute Myeloid Leukemia. · Yang Z, Wu XS, et al. · Cancer Discov 2021What this evidence cannot show
The screen was domain-focused on the ubiquitination machinery, so it could only have found E2s and E3s — the fact that it nominated the catalytic pair is partly a property of the library, not only of the biology.
Evidence that argues against this (1)
- That library design point is a genuine confound: a ubiquitination-focused screen cannot nominate FANCC or FANCE, so its agreement with the module reading here is weaker than it first looks. Yang Z et al. 2021
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Humans biallelically null for an FA gene nonetheless develop KMT2A-rearranged leukaemia, so no FA gene can be an absolute requirement for a KMT2A-rearranged clone.
MLL-CBP fusion transcript in a therapy-related acute myeloid leukemia with the t(11;16)(q23;p13) which developed in an acute lymphoblastic leukemia patient with Fanconi anemia. · Sugita K, Taki T, et al. · Genes Chromosomes Cancer 2000What this evidence cannot show
A single case report; the complementation group is not given in the abstract, and this was a therapy-related leukaemia after topoisomerase-II-directed chemotherapy, which is a special context.
Evidence that argues against this (1)
- It does not refute a quantitative dependency, only an absolute one, and the DepMap effects are quantitative. A clone that forms in an FA-null host may still grow worse than it would with the pathway intact. Quentin S et al. 2011
obligate core subunits span 5x and the substrate FANCD2 is weak, so the canonical pathway is probably not the unit of dependency
a mechanism running through FANCD2 at large fragile genes is hard to reconcile with FANCD2 ranking 411 of 8,269
FANCL is the catalytic RING of the module the signal concentrates on
- part of →UBE2T — E2; rank 33 of 8,269 for KMT2A-selective dependency, and no KMT2A literature at all
UBE2T is the E2 that feeds FANCL; the published AML screen nominated this pair
FAAP100 scaffolds FANCB and FANCL — and FANCB, its own partner, is the weakest core subunit
- ← supportsFANCD2 — the FA effector menin binds, and the weakest FA gene in this graph's own selectivity scan
the substrate-versus-module discrepancy this entry documents