Oncofusion

start / graph / ana-ell-s309-dominant-phosphosite

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q25-partner-phospho

ELL S309 is the most frequently observed phosphosite across public datasets - and it is a proline-directed site with no kinase, not an ATM site

also called ELL S309 · ELL S308 · ELL phosphorylation landscape

Someone pooled all the public data on where phosphates land on ELL and found one spot that shows up far more than any other. It is not one of the spots the DNA-damage kinase uses, nobody knows which enzyme makes it, and the technique cannot cleanly tell it apart from the serine right next door.

The same thing, for a scientist

A 2026 systematic curation of public global phosphoproteomic datasets identifies ELL **S309** as the most frequently detected and most consistently regulated ELL phosphosite, and reports co-regulation of S309 with PBRM1 S648, ATM S1885 and TP53 S392 - all DNA-damage-response-associated. Two independent resources corroborate the residue: the PeptideAtlas/PTMeXchange re-analysis of P55199 lists both **S308 and S309**, and UniProt's curated table lists **S309**. The KMT2A-rearranged AML patient phosphoproteome (74 patients) reports the adjacent **S308**. Three things follow. First, S309 sits in the sequence context `...A-A-S(308)-S(309)-P-P...`: it is a **proline-directed** site, the signature of CDK/MAPK-family kinases, **not** the S/T-Q motif ATM uses, so the dominant ELL phosphosite is not part of the ATM mechanism this beat's other entries describe. Second, S308 and S309 are adjacent serines and phosphosite localisation between adjacent residues is the classic ambiguity of the method, so the S308/S309 distinction across resources should be read as one site with uncertain assignment rather than two. Third, S309 is at residue 309, far C-terminal of the aa 1-45 discarded segment, so it is **retained in MLL-ELL** - and no kinase has ever been assigned to it.

Go deeper
The convergence, and the ambiguity inside it
resourceELL site reportedroute
systematic curation of public phosphoproteomics, 2026S309 (most frequently detected, most consistently regulated)1
PeptideAtlas / PTMeXchange re-analysis of P55199S308 and S309 (both)2
UniProt curated Modified residueS3092
KMT2A-rearranged AML patient phosphoproteome, 74 patientsS3083

Four resources, two adjacent serines. The sequence around them, read from the P55199 FASTA in this session, is S(300)LLGDPAA-S(308)-S(309)-PPGERGRSASPPQKRL 2. Two serines separated by one residue, followed by proline, is precisely the configuration in which phosphosite localisation scoring is least reliable, because both candidate assignments produce fragment-ion series that differ at only one position.

Treat this as one site with an uncertain assignment. Reporting S308 and S309 as two sites would double-count it, and reporting only one of them would make the other three resources look like they disagree when they do not.

Proline-directed, therefore not ATM - and therefore unexplained

The residue after S309 is proline. That makes it a canonical substrate position for the proline-directed kinase families: CDKs, MAPKs, GSK3 in its primed form. It is not an S/T-Q motif, and P55199's seven S/T-Q motifs are at 59, 203, 235, 345, 551, 561 and 589 2 - none of them at 308/309.

So the most-observed phosphosite on ELL is mechanistically disconnected from everything else in this beat. That is not a defect in the beat; it is the finding. The best-measured site on ELL has no kinase, no function and no paper about it, while the three sites with a kinase and a mechanism (T551/S561/S589) are barely observed in proteomics at all.

CDK9 is the obvious candidate given that ELL is a SEC subunit and CDK9 is the SEC kinase, and this project's standing rule makes that exactly the sort of suggestion that must be labelled a hypothesis for a kinase assay rather than an assignment. The PubMed query (MLLT1 OR MLLT3 OR AF9 OR ENL) AND CDK9 AND phosphorylat* returns 1 record, and it is not about this.

Retained, and a candidate for the first real measurement

S308/S309 is far C-terminal of ELL aa 45, so it is retained in MLL-ELL. It is also one of the two ELL sites observed in actual KMT2A-rearranged AML patient material rather than in cell lines 3.

That combination - most-observed site, retained in the fusion, seen in patients, no kinase - makes it the single best candidate in this beat for an in vitro kinase screen. The experiment is a synthetic phosphopeptide and a kinase panel.

Facts and figures
statusrun
genegene-ell
uniprotP55199
siteS308/S309 (one site, ambiguous assignment between adjacent serines)
motifproline-directed (S-P), not S/T-Q
retained in fusiontrue
retained basisMLL-ELL contains ELL aa 46-621
result summaryThe best-observed ELL phosphosite is retained in the fusion, seen in patient material, proline-directed, and has no kinase and no function assigned. It is the strongest candidate in this beat for an in vitro kinase screen.
gene symbolELL
candidate kinase families
[
  "CDK",
  "MAPK",
  "primed GSK3"
]
observed in
[
  "public phosphoproteomic meta-analysis (S309)",
  "PeptideAtlas/PTMeXchange (S308 and S309)",
  "UniProt curated (S309)",
  "74-patient KMT2A-rearranged AML cohort (S308)"
]
How this could bite us back
  • If a CDK inhibitor were pursued on the strength of a proline-directed motif, that would be assigning a kinase from a motif - the exact reasoning this project withdrew once already when a 14-3-3 motif candidate on KMT2A turned out not to be phosphorylated at all.
  • This site is retained in the fusion and also present on wild-type ELL, which every cell expresses. Any phospho-directed intervention here has no fusion selectivity by construction.
Still open
  1. Which kinase phosphorylates ELL S309? A synthetic phosphopeptide against a commercial kinase panel would answer it in a week.
  2. Is the site S308 or S309? Targeted MS with a synthetic standard for each would settle an ambiguity that currently makes four resources look like they disagree.
  3. Does S309 occupancy differ between KMT2A-rearranged AML and other AML in the existing 74-patient matrix? The data are deposited; nobody has asked.
  4. Is S309 phosphorylated on the MLL-ELL fusion protein itself, or only on wild-type ELL? The fusion and wild-type ELL share this peptide, so distinguishing them requires an immunoprecipitation of the fusion first.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A systematic curation and analysis of publicly available global phosphoproteomic datasets identifies ELL Serine 309 as the most frequently detected and most consistently regulated ELL phosphosite, with co-regulation across datasets between S309 and PBRM1 S648, ATM S1885 and TP53 S392.

    PMID 41843233DOI
    What this evidence cannot show

    ABSTRACT ONLY. Full text could not be obtained 2026-09-19: the Europe PMC core record for EXT_ID:41843233 returns no pmcid and the NCBI PMC ID converter returns 'Identifier not found in PMC', so the remaining routes cannot be attempted. It is a re-analysis of other people's data, not new measurement; which datasets, how many, and what 'consistently regulated' means quantitatively are unknown to this entry. Co-regulation across datasets is a correlational statistic and is not evidence of a shared kinase.

    Evidence that argues against this (1)
    • The functional ELL phosphorylation literature from a different laboratory places the causal sites at T551, S561 and S589 and does not identify S309 as functionally important; S561 rather than S309 is the site with a named kinase and a mutant phenotype. Frequency of detection in shotgun proteomics ranks sites by peptide behaviour and abundance, not by importance. Pal S et al. 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The PeptideAtlas/PTMeXchange re-analysis of P55199 reports phosphorylation at both S308 and S309, UniProt's curated table reports S309, the local sequence is S(300)LLGDPAASSPPGERGRSASPPQKRL making S309 proline-directed, and the seven S/T-Q motifs of ELL are at 59, 203, 235, 345, 551, 561 and 589 - none at 308 or 309.

    What this evidence cannot show

    Position arithmetic and motif scanning were done in this session; residue identity at every reported position was confirmed against the FASTA. Adjacent-serine localisation ambiguity means the S308-versus-S309 assignment in any individual dataset may be wrong even where the peptide is correctly identified.

    Evidence that argues against this (1)
    • A proline-directed context is a necessary but not sufficient condition for CDK/MAPK action, and this project has already had to withdraw one inference built on a short linear motif when the phosphosite atlas contradicted it. The motif here constrains the candidate kinase family and does nothing more. Bacon CW et al. 2019
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ELL S308 is observed as a phosphosite in the supplementary phosphosite matrix of an integrative phosphoproteomic study of 74 KMT2A-rearranged AML patients, alongside ELL S437.

    PMID 36843114DOI
    What this evidence cannot show

    Taken from this graph's ana-phosphosites-observed-in-kmt2a-r-aml-patients entry, which parsed the supplementary matrix; this session did not re-parse the file. Those are presence/absence row labels, not per-patient values: the number of patients carrying the site, the stoichiometry and the localisation probability are all unread. Residue 308 of P55199 was confirmed to be a serine in this session. The cohort is AML, and no kinase is assigned to any site in that table.

    Evidence that argues against this (1)
    • The same source entry records that numbering in that supplementary table is isoform-dependent and that at least two reported sites do not survive a canonical-sequence check, one of which this beat resolved to a non-canonical isoform. A site that passes the residue-type check can still be at the wrong position, because roughly one residue in eight is a serine or threonine. UniProt Consortium 2026
Connected to