PAF1C couples transcription to H2B ubiquitination - but the RTF1 half of that coupling is yeast-specific and contested in human cells, and PAF1C is one of the few complexes that genuinely binds the retained KMT2A fragment
also called PAF1 complex and H2Bub · RTF1-dependent H2B ubiquitination · transcription-coupled H2Bub
The ubiquitin tag only lands on genes that are actively being read, and the reason is a complex called PAF1C that rides along with the reading machinery and brings the tagging enzyme with it. That matters here because PAF1C is one of the few things that binds directly to the part of KMT2A kept in the fusion.
The same thing, for a scientist
The specificity of H2BK120ub for transcribed gene bodies is not intrinsic to RNF20/RNF40; it is imposed by the polymerase-associated factor complex. In yeast, Rtf1 is essential for global H3K4 and H3K79 methylation but not H3K36 methylation, and the reason is that **Rtf1 is required for H2B ubiquitination** — its effects on the two methylation marks are an indirect consequence [E1]. **That half does not transfer cleanly.** Human Rtf1 is not stably associated with PAF1C, appears to act independently of it, and PAF1C recruitment in HeLa cells was found to be Rtf1-*independent*, the opposite of the yeast result (PMID 26217014, counter-evidence on [E1]); the human arm of the coupling therefore rests on hPAF-bound hBRE1 [E2] and on the UbcH6-RNF20/40-hPAF trimer [E3], not on RTF1. In human cells, hRAD6 is recruited to transcribed genes through direct interaction with **hPAF-bound hBRE1**, and hPAF-mediated transcription is required for efficient H2B ubiquitination [E2]; independently, UbcH6 and RNF20/40 form a stimulatory trimeric complex with hPAF [E3]. This is the mechanistic joint between two layers the graph already holds separately. It matters for this disease because PAF1C is, in this project's own assessment (`gene-rtf1`), one of the few interactions that is genuinely **on the fusion**: MLL interacts directly with PAF1C through sequences flanking the CxxC domain, a region retained in every KMT2A fusion, and MLL deletions that abolish the PAF1C interaction eliminate MLL-AF9-mediated immortalisation [E4]. The chain therefore closes: the fusion binds PAF1C; PAF1C recruits the H2B ubiquitination machinery; H2Bub licenses DOT1L. **Nobody has tested whether the second link is required for the first to produce the third in a KMT2A-fusion cell**, and that single epistasis experiment is the most valuable thing this beat can point at.
The yeast result that established the coupling, and exactly what it says
Ng, Dole and Struhl showed that Rtf1 is essential for global methylation of H3-Lys4 and H3-Lys79 but not H3-Lys36, that this role resembles Rad6's, and that Rtf1 is required for H2B ubiquitination — from which they inferred the methylation effects are indirect 1.
Two details of that paper are often lost and both are load-bearing here.
Dot1 recruitment is not the mechanism. Dot1 associates with transcriptionally active genes, and unlike Set1 and Set2, that association is largely independent of Rtf1 1. So PAF1C does not bring the methyltransferase; it brings the licence. The enzyme is already there and cannot work.
The H2Bub function is separable from elongation. The authors explicitly suggest Rtf1 affects genome-wide H2B ubiquitination by a mechanism distinct from its function as a transcriptional elongation factor 1. That is a separation-of-function claim, and it is the reason a RTF1-directed intervention is conceptually different from a general elongation-blocking one.
And a third detail that cuts against reading this into human cells at all. Cao et al. found human Rtf1 is not a stable PAF1C subunit, that it and PAF1C regulate distinct subsets of genes in HeLa, and that PAF1C recruitment to the genes examined was Rtf1-independent - the reverse of the yeast finding (counter-evidence on 1). Every RTF1-specific statement above is a yeast statement, and the experiment this entry points at, exp-rtf1-h2bub-separation-of-function-kmt2a, has to survive that.
Why this is the highest-value untested experiment in this beat
Lay the three established facts side by side:
| Fact | Source | Where it lives in this graph |
|---|---|---|
| KMT2A binds PAF1C directly through sequences flanking CxxC; deleting that contact abolishes MLL-AF9 immortalisation | 4 | gene-rtf1, gene-paf1 |
| PAF1C/RTF1 is required for H2B ubiquitination | 123 | this entry |
| H2Bub is required to maintain Dot1l-dependent H3K79me at Hoxa9 and Meis1 | gene-rnf20 E1 | mark-h2bk120ub |
Each link is established. The composite has never been tested. The field's explanation for why the KMT2A–PAF1C contact matters is that PAF1C helps recruit the fusion to target loci and supports elongation. The alternative explanation the chain above permits is that the contact matters because it brings the ubiquitination machinery, and that H3K79 methylation is downstream of that.
The discriminating experiment is written up as exp-rtf1-h2bub-separation-of-function-kmt2a: a separation-of-function RTF1 allele that supports elongation but not H2B ubiquitination, in a KMT2A-fusion cell, reading out H2Bub, H3K79me2 and fusion target-gene expression. If the allele leaves H3K79me2 intact, the ubiquitin arm is dispensable in this disease and this whole beat is a chromatin footnote. If it collapses H3K79me2, PAF1C's role in KMT2A leukaemia is substantially a ubiquitination role, and the graph's account of mechanism-chromatin-elongation-hijack needs amending.
What the DepMap numbers say about PAF1C as a target, which is not encouraging
From the project's genome-wide scan 5:
| Subunit | KMT2A-r | other leuk | difference | rank / 17,728 | pan-essentiality |
|---|---|---|---|---|---|
| RTF1 | -0.703 | -0.620 | -0.083 | 1,645 | -0.666 |
| LEO1 | -0.420 | -0.252 | -0.168 | 367 | -0.213 |
| PAF1 | -0.818 | -0.780 | -0.038 | 4,529 | -0.832 |
| CDC73 | -1.177 | -1.196 | +0.020 | 11,645 | -1.122 |
| CTR9 | -0.854 | -0.999 | +0.146 | 17,413 | -0.873 |
| SKIC8 | -0.831 | -0.815 | -0.016 | 7,012 | -0.746 |
| MEN1 (benchmark) | -1.263 | -0.397 | -0.866 | 1 | -0.363 |
Only LEO1 reaches the top 2.1% of the genome on selectivity, and it does so with a small absolute requirement. The complex as a whole is pan-essential and unselective, which is the same verdict this beat reached for the H2Bub writers. The interesting number is that the two PAF1C subunits with any selectivity signal (LEO1, RTF1) are the two with the lowest pan-essentiality, which is what a real window looks like when there is one — but -0.168 is not -0.866.
The elongation half has the wrong sign, too. Chen et al. showed that reducing PAF1 increases release of paused Pol II at thousands of genes, via recruitment of the super elongation complex (counter-evidence on 4). SEC is the complex KMT2A::AFF1 recruits. A PAF1C perturbation in a KMT2A-fusion cell therefore removes the ubiquitination route and de-represses SEC-driven elongation at the same time, and no published experiment separates the two.
| coupling | elongating Pol II -> PAF1C (RTF1) -> RNF20/RNF40 + E2 -> H2BK120ub -> DOT1L -> H3K79me |
|---|---|
| key subunit | RTF1 |
| separability claim | Rtf1's effect on H2B ubiquitination is suggested to be distinct from its elongation-factor function |
| fusion relevance | PAF1C binds sequences flanking the CxxC domain of KMT2A, a region retained in every KMT2A fusion |
| untested composite | whether the KMT2A-PAF1C contact matters because it delivers the H2B ubiquitination machinery rather than because it supports elongation |
| human caveat | human RTF1 is not a stable PAF1C subunit and PAF1C recruitment was Rtf1-independent in HeLa (PMID 26217014); the RTF1-specific coupling is a yeast result |
| opposite sign effect | PAF1 depletion increases paused-Pol-II release via SEC recruitment (PMID 26279188) - SEC is the fusion's own elongation machinery |
experiments
[ "exp-rtf1-h2bub-separation-of-function-kmt2a" ]
depmap paf1c — the numbers behind the chart
{
"RTF1": {
"difference": -0.0825,
"rank": 1645,
"pan": -0.6659
},
"LEO1": {
"difference": -0.1678,
"rank": 367,
"pan": -0.2126
},
"PAF1": {
"difference": -0.0381,
"rank": 4529,
"pan": -0.8316
},
"CDC73": {
"difference": 0.0195,
"rank": 11645,
"pan": -1.1215
},
"CTR9": {
"difference": 0.1456,
"rank": 17413,
"pan": -0.8729
},
"SKIC8": {
"difference": -0.0163,
"rank": 7012,
"pan": -0.7458
}
}- PAF1C is required for transcription of essentially everything; blocking the coupling blocks H2Bub genome-wide, not at fusion targets, and the pan-essentiality numbers (PAF1 -0.832, CDC73 -1.122) say what that costs.
- If the KMT2A-PAF1C contact matters for elongation rather than for ubiquitination, then intervening at the ubiquitination step leaves the oncogenic mechanism running and only adds toxicity.
- LEO1's selectivity signal is the one encouraging number in this section and it comes from 11 cell lines; the project has previously found that small DepMap differences in small line panels do not survive better annotation.
- Depleting PAF1C in a KMT2A-fusion cell does two opposing things at once - it removes the H2B ubiquitination route and it increases SEC-driven pause release (PMID 26279188) - so a null result would be uninterpretable and a positive one would be hard to attribute.
- Does human RTF1 do the job at all? Cao et al. (PMID 26217014) report human Rtf1 acting outside PAF1C and PAF1C recruitment being Rtf1-independent, which would make the yeast coupling the wrong model for this disease and would redirect the experiment onto a catalytically dead RNF20 knock-in.
- Does the KMT2A-PAF1C contact support H2B ubiquitination at fusion target genes? The contact is established and the PAF1C-H2Bub coupling is established, but the composite has never been measured in a KMT2A-fusion cell.
- Is RTF1's H2Bub function separable from its elongation function in human cells? The suggestion is twenty-three years old and no separation-of-function human allele is described in the literature this beat found.
- Does the graph's account of PAF1C need amending? `mechanism-chromatin-elongation-hijack` explains PAF1C's contribution in elongation terms; the ubiquitination route is an alternative that has not been excluded.
- Why is LEO1 the only PAF1C subunit with meaningful KMT2A selectivity (-0.168, rank 367) and a low pan-essentiality (-0.213)? That combination is the profile of a window, and LEO1's specific role in H2Bub is not established.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Rtf1 is essential for global methylation of H3-Lys4 and H3-Lys79 but not H3-Lys36; Rtf1 is required for H2B ubiquitination, suggesting its effects on H3-Lys4 and H3-Lys79 methylation are an indirect consequence; Dot1 associates with transcriptionally active genes but, unlike Set1 and Set2, this association is largely independent of Rtf1; and the authors suggest Rtf1 affects genome-wide H2B ubiquitination by a mechanism distinct from its function as a transcriptional elongation factor.
The Rtf1 component of the Paf1 transcriptional elongation complex is required for ubiquitination of histone H2B. · Ng HH, Dole S, Struhl K · J Biol Chem 2003What this evidence cannot show
Abstract read in full; full text not read. Saccharomyces cerevisiae. 'Essential for global methylation' is a bulk measurement; locus-level behaviour may differ. The separation-of-function claim is offered by the authors as a suggestion, not demonstrated with a separation-of-function allele, and no such allele is asserted to exist by this entry.
Evidence that argues against this (1)
- The coupling this entry is built on may be yeast-specific. Cao et al. showed that human Rtf1 is NOT stably associated with PAF1C - unlike in S. cerevisiae - that it is a transcription elongation factor that may function independently of PAF1C, that RNA-seq and ChIP in HeLa cells give Rtf1 and PAF1C distinct roles over a subset of genes, and, directly contrary to the yeast result, that PAF1C was recruited to the genes examined in an Rtf1-INDEPENDENT manner. If human RTF1 is not the PAF1C handle on the ubiquitination machinery, the chain 'fusion binds PAF1C -> PAF1C recruits RNF20/40 via RTF1' does not transfer from yeast. A published correction to this article exists (PMID 26518638); its abstract is empty in the PubMed record and this repair did not establish what was corrected. Cao QF et al. 2015
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
hRAD6, through direct interaction with hPAF-bound hBRE1, is recruited to transcribed genes and ubiquitylates chromatinised H2B at lysine 120, and hPAF-mediated transcription is required for efficient H2B ubiquitylation as a result of hPAF-dependent recruitment of hBRE1-hRAD6 to the Pol II transcription machinery.
RAD6-Mediated transcription-coupled H2B ubiquitylation directly stimulates H3K4 methylation in human cells. · Kim J, Guermah M, McGinty RK, Lee JS, Tang Z, Milne TA, Shilatifard A, Muir TW, Roeder RG · Cell 2009What this evidence cannot show
Abstract read in full; full text not read. Reconstituted human systems; not leukaemia, not a KMT2A fusion. The E2 assignment in this paper is contested - see `gene-ube2a`.
Evidence that argues against this (1)
- The recruitment model was established in human cells for hRAD6 via hPAF-bound hBRE1, and 'transcription-coupled' is inferred from co-occupancy with transcribed genes. Co-occupancy does not establish that transcription recruits the ligase rather than the ligase being required for transcription. Cell 2009
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
UbcH6 physically interacts with RNF20/40 and with the hPAF complex, formation of a trimeric complex with hPAF stimulates H2B monoubiquitination activity in vitro, and UbcH6, RNF20/40 and the hPAF complex are all recruited to transcriptionally active genes in vivo; RNAi against the hPAF complex reduces H2B monoubiquitination and lowers H3K4 and H3K79 methylation.
Monoubiquitination of human histone H2B: the factors involved and their roles in HOX gene regulation. · Zhu B, Zheng Y, Pham AD, Mandal SS, Erdjument-Bromage H, Tempst P, Reinberg D · Mol Cell 2005What this evidence cannot show
Abstract read in full; full text not read. RNAi against a multi-subunit complex perturbs complex stoichiometry as well as function.
Evidence that argues against this (1)
- A trimeric complex with hPAF stimulating monoubiquitination in vitro demonstrates that the components can act together. Stimulation in a defined system with purified parts routinely exceeds what the same interaction contributes in chromatin, where PAF1C has several other functions competing for it. Molecular Cell 2005
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
MLL interacts directly with the polymerase-associated factor complex through sequences flanking the CxxC domain; PAFc augments MLL- and MLL-AF9-mediated transcriptional activation of Hoxa9; PAFc knockdown disrupts MLL-fusion-mediated transcriptional activation and MLL recruitment to target loci; and MLL deletions that abolish PAFc interaction also eliminate MLL-AF9-mediated immortalisation.
The PAF complex synergizes with MLL fusion proteins at HOX loci to promote leukemogenesis. · Muntean AG, Tan J, Sitwala K, Huang Y, Bronstein J, Connelly JA, Basrur V, Elenitoba-Johnson KS, Hess JL · Cancer Cell 2010What this evidence cannot show
PMID re-fetched from NCBI eutils by this beat on 2026-09-19 and the title read back; abstract read in full, full text not read. The abstract's phrase is that PAFc 'stimulates posttranslational histone modifications' - it does not name H2B ubiquitination, so the link from this contact to the ubiquitination machinery is supplied by [E1]-[E3] and not by this paper. MLL-AF9 mouse immortalisation, not a KMT2A::AFF1 B-ALL model. The same PMID is cited in this graph's `gene-rtf1`.
Evidence that argues against this (2)
- PAF1C's effect on elongation runs in the opposite direction from the one this entry's title assumes, and it runs through the fusion's own machinery. Chen et al. found that reducing PAF1 levels INCREASES release of paused Pol II into gene bodies at thousands of genes, raising nascent and mature transcripts and Ser2 phosphorylation, and that this is explained by recruitment of the super elongation complex. SEC is precisely what KMT2A::AFF1 recruits through AFF1/AFF4. So 'PAF1C couples elongation to H2Bub' is not a simple positive coupling: losing PAF1C removes the ubiquitination route while simultaneously de-repressing SEC-driven elongation, and in a KMT2A-fusion cell those two effects push the target programme in opposite directions. Chen FX et al. 2015
- This entry's own scan is the counter to reading the KMT2A-PAF1C contact as a therapeutic entry point: no PAF1C subunit is meaningfully KMT2A-selective (PAF1 -0.038 rank 4,529; RTF1 -0.083 rank 1,645; CTR9 +0.146 rank 17,413) on a strongly pan-essential background, against MEN1's -0.866 at rank 1. Immortalisation-abolishing in a mouse MLL-AF9 assay and selectively required in human KMT2A-rearranged lines are different claims and only the first is supported. Broad Institute DepMap; Cellosaurus; Open Targets Platform; scan run by this project et al. 2026
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In the project's genome-wide DepMap KMT2A-selectivity scan, PAF1C subunits show little selectivity on a pan-essential background: RTF1 difference -0.083 (rank 1,645; pan -0.666), LEO1 -0.168 (rank 367; pan -0.213), PAF1 -0.038 (rank 4,529; pan -0.832), CDC73 +0.020 (rank 11,645; pan -1.122), CTR9 +0.146 (rank 17,413; pan -0.873), SKIC8 -0.016 (rank 7,012; pan -0.746), against MEN1 at -0.866 (rank 1).
DepMap CRISPR (Chronos) genome-wide KMT2A-selectivity scan of 17,728 protein-coding genes, 11 Cellosaurus-annotated KMT2A-rearranged leukaemia lines vs 43 other leukaemia lines · Broad Institute DepMap; Cellosaurus; Open Targets Platform; analysis run by this project (analyses/depmap-kmt2a-selective-scan-full) · Database and derived analysis 2026What this evidence cannot show
Eleven KMT2A-rearranged lines, AML-weighted, none infant-derived. Selectivity differences are compressed for pan-essential genes because both arms die. CTR9's large positive difference on a strongly essential background is more likely noise than biology.
Evidence that argues against this (1)
- SCAN CONFOUND: this number comes from the project's genome-wide KMT2A-selectivity scan, whose index arm is MYELOID-ENRICHED (8 AML lines against 5 B-ALL). A gene whose requirement differs by lineage is mis-scored in BOTH directions by that design — CEBPA ranks 25 genome-wide with no dependency at all in KMT2A-rearranged B-ALL, and conversely a gene required in one lineage and not the other can average to apparent non-selectivity. See `ana-lineage-confound-myeloid-tf-dependencies`. For PAF1C the pan-essential background is the dominant problem and it is independent of lineage: a complex required by most cells has little dynamic range left in which to show KMT2A selectivity, so 'little selectivity on a pan-essential background' is close to uninformative about the fusion. Oncofusion knowledge graph (internal analysis) 2026
Counter-evidence (Cao 2015, PMID 26217014): human Rtf1 is not stably associated with PAF1C, regulates a distinct subset of genes from PAF1C in HeLa, and - opposite to the yeast result - PAF1C recruitment to the genes examined was Rtf1-independent, so the RTF1 half of this coupling may not transfer to human cells
Counter-evidence (Chen 2015, PMID 26279188): reducing PAF1 increases release of paused Pol II at thousands of genes by recruiting the super elongation complex - the very complex KMT2A::AFF1 recruits - so PAF1C loss de-represses elongation at the same time as it removes the ubiquitination route
Counter-evidence (DepMap scan, this project): no PAF1C subunit is meaningfully KMT2A-selective (PAF1 -0.038 rank 4,529; RTF1 -0.083 rank 1,645; CTR9 +0.146 rank 17,413) against MEN1 at -0.866 rank 1, so an immortalisation result in a mouse MLL-AF9 assay is not evidence of selective requirement in human KMT2A-rearranged lines
PAF1C/RTF1 is required for H2B ubiquitination and imposes its transcription-coupled specificity
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
PAF1C binds sequences flanking CxxC, retained in every KMT2A fusion; whether that contact delivers the ubiquitination machinery is untested
separates PAF1C's elongation role from its H2B ubiquitination role in a KMT2A-fusion cell