Oncofusion

start / graph / interface-psip1-gdf5-declined-a-secreted-cartilage-growth-factor-crosslinked-in-nuclei

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q129-the-never-discussed-partners-of-kmt2a-and-psip1

PSIP1-GDF5 is a single DSSO crosslink in isolated U2OS nuclei with no bait on either side; GDF5 is a secreted cartilage growth factor, and the source paper itself attributes 30% of its cross-compartment links to contamination during nuclei isolation - declined

also called PSIP1-GDF5 · LEDGF-GDF5 · BMP-14 · CDMP1 · EBI-20905152

A database lists GDF5, a growth factor cells secrete to build cartilage and bone, as a partner of the nuclear protein LEDGF/PSIP1. The single supporting measurement is a chemical crosslink found inside isolated cell nuclei, and the paper that produced it says outright that some of its cross-compartment links come from contamination while the nuclei were being purified.

The same thing, for a scientist

**Verdict: DECLINED. Detection method is chemical crosslinking mass spectrometry (XL-MS), not affinity purification, and there is no bait: both participants are recorded as 'neutral component'** [E1]. One IntAct record, EBI-20905152, from PMID 30021884 (Fasci et al., MCP 2018), host U2OS osteosarcoma, intact-miscore 0.4, annotated *'interaction detected in: TX100 soluble fraction (fractionated nuclei), unfractionated nuclei'* [E1]. GDF5 is UniProt P43026, 501 aa, subcellular location **Secreted [ECO:0000269] and Cell membrane [ECO:0000269]**, a BMP-family growth factor that signals through BMPR1B/BMPR1A in cartilage and bone development, predominantly expressed in long bones during embryonic development [E3]. **The decline does not rest on my judgement of plausibility - it rests on the source paper's own control.** Fasci et al. report that **20% of their interactions are nucleus-cytosol/plasma-membrane, 6% nucleus-mitochondria and 4% nucleus-endoplasmic-reticulum**, and write that this *"could also be partly influenced by contamination with other subcellular compartments during our nuclei isolation"* [E2]. **Thirty per cent of the dataset is cross-compartment and the authors flag it.** This row is one of them. The dataset is otherwise strong - 8,710 unique crosslinks at 1% FDR, validated against high-resolution structures - and the paper is about histones and nucleosome interaction hot spots; **neither PSIP1, LEDGF, KMT2A nor GDF5 appears anywhere in its text** [E2].

Go deeper
Why an XL-MS row is a different kind of evidence from an AP-MS row, and why that cuts both ways

Eight of the thirteen partners in this beat are affinity-purification rows in which a partner was tagged and KMT2A or PSIP1 fell out as prey. The five Fasci rows are not that. A DSSO crosslink is a covalent distance restraint formed in situ, before lysis, between two lysines within roughly 30 angstroms. There is no bait, no tag, no over-expression, and IntAct records both sides as neutral component with no spoke expansion 1.

That is, in principle, better evidence than a spoke-expanded pull-down. A crosslink says the two proteins were physically close in an intact nucleus. It does not depend on a lysate surviving a wash.

Three things pull it back down.

  1. The identification is a peptide-spectrum match at 1% FDR over ~8,700 crosslinks 2. At that scale, a 1% error rate is roughly 87 wrong crosslinks, and a wrong crosslink names two real proteins in a plausible-looking pair. One record is one PSM-level call.
  2. Replicate reproducibility is 58% between two independent biological replicates 2. This record is annotated to two preparations - the Triton X-100-soluble fraction and unfractionated nuclei - which is better than most of the Fasci rows in this beat, and is the strongest single point in its favour.
  3. The compartment problem is the paper's own 2. Thirty per cent of the network crosses compartments, and the authors name nuclei-isolation contamination as a partial cause. A secreted BMP-family growth factor in an isolated nucleus is the clearest possible instance of the class they flagged.

The honest position: this row is not evidence of nothing. It is evidence that a GDF5 peptide and a PSIP1 peptide were crosslinked in a preparation that the authors say contains material from other compartments. That is not an interaction claim.

The Fasci dataset contributes five of the thirteen partners in this beat, and two of the five are membrane proteins

PMID 30021884 supplies five of the thirteen never-discussed partners: GDF5, IGHMBP2, PPP1R13B and TMEM214 on PSIP1, and TMCO1 on KMT2A 1.

IntAct serves 4,856 rows from this publication, covering 3,660 distinct molecules, every one of them crosslink and every one neutral component on both sides - confirming there is no bait anywhere in the dataset 4.

Within the two proteins this beat cares about, the cross-compartment fraction matches the paper's global figure almost exactly:

bait proteincrosslink partners in this datasetnon-nuclear membrane or secreted
PSIP1GDF5, HSPA8, IGHMBP2, PPP1R13B, SCAF11, TMEM214GDF5 (secreted), TMEM214 (ER membrane) - 2 of 6
KMT2ACCT6A, ITGB1, NPM1, RPL13, TMCO1ITGB1 (plasma membrane), TMCO1 (ER/Golgi membrane) - 2 of 5

Four of eleven, 36%, against the paper's own global 30% 24. The local neighbourhoods of KMT2A and PSIP1 in this dataset are not cleaner than the dataset as a whole - they are, if anything, slightly worse, which is what one expects when the proteins in question are themselves low-abundance and near the stated detection floor of ~1.5e4 copies per cell 2.

Facts and figures
partner agene-psip1
partner a uniprotO75475
partner a length AA530
partner bGDF5
partner b uniprotP43026
partner b length AA501
verdictDECLINED
adjudicated byQ129, cycle 92, 2026-09-23
source publication pmid30021884
source datasetFasci 2018 nuclear XL-MS (DSSO, U2OS intact nuclei)
detection methodchemical crosslinking mass spectrometry (DSSO XL-MS)
experimental role partnerneutral component
experimental role psip1neutral component
bait exists in datasetfalse
intact miscore0.4
n publications1
n methods1
orthogonal method existsfalse
dataset fdr1% crosslink FDR over ~8,710 unique crosslinks (XlinkX Validator / Percolator)
dataset replicate overlap PCT58
dataset detection depth~1.5e4 protein copies per cell (first quartile of detected copy numbers)
dataset cross compartment PCT30
dataset cross compartment authors note20% nucleus-cytosol/plasma membrane, 6% nucleus-mitochondria, 4% nucleus-endoplasmic reticulum; the authors write this 'could also be partly influenced by contamination with other subcellular compartments during our nuclei isolation'
druggabilityNot assessable. The crosslinked residue positions are in supplemental Table S1/S2 and were not retrieved, so not even the crosslinked lysines are known here - which is the one thing an XL-MS row could in principle contribute to an interface map.
absence classmeasured-not-searched
absence class noteThe decisive missing datum is NOT a new experiment - it is the crosslinked lysine positions and PSM score for EBI-20905152, which sit in supplemental Tables S1 and S2 of PMID 30021884 and in the PRIDE deposition for that paper, neither of which was opened. Opening them is an afternoon and would convert this entry from an argument about compartments into an argument about a spectrum. The separate, genuinely never-measured absence is that nobody has tested PSIP1 against GDF5 in any haematopoietic cell or with PSIP1 as a bait.
shown for fusionNO - nobody has looked in any KMT2A fusion. U2OS osteosarcoma nuclei, no fusion protein present, and PSIP1's relevance to this project is as the KMT2A-menin complex's chromatin tether.
epigenetic readoutnone - no histone mark, occupancy or chromatin measurement exists for this pair, and the source paper is about histone crosslinking without ever naming PSIP1.
host cell lines
[
  "U2OS osteosarcoma"
]
preparations detected in
[
  "TX100 soluble fraction (fractionated nuclei)",
  "unfractionated nuclei"
]
records
[
  "EBI-20905152"
]
compartment mismatch
{
  "partner_compartments": [
    "Secreted [ECO:0000269]",
    "Cell membrane [ECO:0000269]"
  ],
  "partner_nuclear_annotation": "none at any evidence level",
  "partner_expressed_in_host_line": "plausible - U2OS is a bone-lineage line and GDF5 is a bone morphogenetic factor"
}
interface residues a
[]
structures
[]
How this could bite us back
  • PSIP1/LEDGF is the tether that holds the menin-KMT2A complex on chromatin and is the target of the LEDGIN class of allosteric inhibitors developed for HIV integrase. Any claimed PSIP1 partner has intervention consequences, which is why a floor-level row deserves a written decline rather than silence.
  • BMP signalling is active in the bone-marrow niche, and a genuine GDF5-PSIP1 axis would connect a niche growth factor to a chromatin tether. That would be interesting enough that it should not be believed on one crosslink in an osteosarcoma line - the attractiveness of the story is a reason for more scepticism, not less.
Still open
  1. What are the crosslinked lysine positions and the PSM score for EBI-20905152? They are in supplemental Table S1/S2 of PMID 30021884, they were not opened here, and they are the only thing that could turn this row into an interface constraint rather than a compartment argument.
  2. How many of this graph's PSIP1 and KMT2A partners come from the Fasci dataset, and would a compartment filter over all 4,856 rows of it give a usable subset? The dataset is genuinely good for nuclear complexes and is being used here only in its weakest corner.
  3. Does PSIP1 crosslink to anything in a haematopoietic nucleus? The whole Fasci dataset is U2OS, and no equivalent nuclear XL-MS dataset from a leukaemia line was searched for in this beat.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IntAct curates one record of a PSIP1-GDF5 physical association, EBI-20905152, from PMID 30021884. Detection method 'crosslink' (XL-MS), interaction type 'physical association', NO expansion method, host Homo sapiens U2OS osteosarcoma cell. Both GDF5 (P43026) and PSIP1 (O75475) are recorded as 'neutral component' - there is no bait and no prey. intact-miscore 0.4. Annotations: 'figure legend (Table S2-pink, Table S1)' and 'comment (interaction detected in: TX100 soluble fraction (fractionated nuclei), unfractionated nuclei)'. The same publication supplies four further never-discussed partners: IGHMBP2, PPP1R13B and TMEM214 on PSIP1, and TMCO1 on KMT2A.

    What this evidence cannot show

    Read live from IntAct REST on 2026-09-23, not PSICQUIC (rule 93: PSICQUIC returns about 44% of IntAct's rows for PSIP1). Supplemental Tables S1 and S2 of the source paper were NOT opened, so the crosslinked residue positions, the peptide-spectrum-match score and the meaning of the 'pink' colour coding in Table S2 are unknown here. Those are the fields that would let a reader judge the identification rather than the annotation.

    Evidence that argues against this (1)
    • The record is annotated to TWO preparations - TX100-soluble fractionated nuclei and unfractionated nuclei - which is more internal support than most rows in this beat have, and the paper reports only 58% overlap between independent replicates, so appearing in two preparations is not trivial. That is the strongest available argument for the row. Fasci D et al. 2018
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In isolated intact human U2OS nuclei, Fasci et al. crosslinked with the MS-cleavable reagent DSSO (250 micromolar, 4 million nuclei/ml) and identified ~8,710 unique crosslinks at 1% FDR (XlinkX Validator/Percolator), of which about two thirds connect distinct proteins, with ~58% overlap between two independent biological replicates and a reliable detection depth of ~1.5e4 protein copies per cell. The paper states: 'Notably, 20% of the interactions were nucleus-cytosol/plasma membrane, 6% nucleus-mitochondria and 4% nucleus-endoplasmic reticulum. This result is in line with the multiple subcellular localization reported for many proteins but could also be partly influenced by contamination with other subcellular compartments during our nuclei isolation.'

    Histone Interaction Landscapes Visualized by Crosslinking Mass Spectrometry in Intact Cell Nuclei. · Fasci D, van Ingen H, Scheltema RA, Heck AJR · Molecular & cellular proteomics : MCP 2018
    PMID 30021884DOI
    What this evidence cannot show

    Full text read 2026-09-23 through the PMC HTML route (PMC6166682, 92,127 stripped characters) after Europe PMC fullTextXML returned HTTP 500 and efetch db=pmc returned a document with no body element. The strings 'PSIP1', 'LEDGF', 'KMT2A', 'MLL', 'GDF5', 'TMCO1', 'TMEM214', 'IGHMBP2' and 'PPP1R13B' each appear ZERO times in the paper's text - every row in this beat from this publication exists only in a supplementary table that was not opened. The paper therefore does not report these interactions; it deposited tables from which IntAct curated them.

    Evidence that argues against this (1)
    • The authors offer the contamination reading as one of two, and state the other first: multiple subcellular localisation is genuinely reported for many proteins. Their sentence does not assign a proportion to either reading, so quoting it as though it settles a specific row overreads it - it establishes that the class exists and is acknowledged, not that this member belongs to it. Fasci D et al. 2018
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB P43026 is 501 aa, primary gene name GDF5, recommended name 'Growth/differentiation factor 5'. Its only subcellular-location comments are Secreted [ECO:0000269] and Cell membrane [ECO:0000269]; there is no nuclear comment at any evidence level. Function: growth factor in bone and cartilage formation, regulating chondrogenic differentiation through high-affinity binding to BMPR1B and lower-affinity binding to BMPR1A. Tissue specificity: predominantly expressed in long bones during embryonic development; expressed in monocytes at protein level. PSIP1 (O75475) is 530 aa with a single subcellular-location comment, Nucleus [ECO:0000269].

    What this evidence cannot show

    UniProt location comments are curated pointers, not measurements (rule 91); the cited papers were not opened. Note the one qualification that argues the other way and is recorded here rather than buried: GDF5's tissue-specificity comment includes 'Expressed in monocytes (at protein level)', so it is not absent from the haematopoietic compartment - it is simply secreted from it rather than nuclear in it. U2OS is an osteosarcoma line, which is the one cell type in which GDF5 expression would not be surprising.

    Evidence that argues against this (1)
    • U2OS is a bone-lineage tumour line, and GDF5 is a bone morphogenetic factor, so unlike the BioPlex secreted partners this pair does at least occur in a cell that plausibly expresses the protein. The objection here is compartment, not expression - which makes it a weaker objection than the one used against ANOS1 and FGFBP1 in this beat, where neither expression nor compartment fits. UniProt Consortium 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IntAct serves 4,856 interaction rows from PMID 30021884, resolving to 3,660 distinct molecules. Every one of the 4,856 rows has detection method 'crosslink' and BOTH participants recorded as 'neutral component' - there is no bait and no prey anywhere in the dataset, and no spoke expansion. Within it, PSIP1 has six crosslink partners (GDF5, HSPA8, IGHMBP2, PPP1R13B, SCAF11, TMEM214) and KMT2A has five (CCT6A, ITGB1, NPM1, RPL13, TMCO1); four of those eleven - GDF5, TMEM214, ITGB1 and TMCO1 - are UniProt-annotated secreted or membrane proteins.

    What this evidence cannot show

    Instrument note: the IntAct REST endpoint returns 0 for the field-scoped form `pubid:30021884` and 4,856 for the bare PMID as free text. A beat using the field-scoped form would record a clean, well-formed false zero - the same shape as rule 85 and rule 95. The bare form may also match the PMID in other fields, so 4,856 is an upper bound; rows were filtered on publicationPubmedIdentifier before counting, and all 4,856 passed. The 4 of 11 cross-compartment figure is this entry's classification from UniProt comments, not the paper's, and 11 is a small denominator.

    Evidence that argues against this (1)
    • Comparing a local 4 of 11 against the paper's global 30% is not a statistical test: n=11 gives a 95% Wilson interval of roughly 15-65%, which contains 30%. So the statement that these neighbourhoods are 'no cleaner than the dataset' is supported, but the statement that they are 'slightly worse' is not - that difference is noise and should not be repeated as a finding. Fasci D et al. 2018
Connected to