LRRK2 - the AFF4 row is one of 78 hits on an in-vitro protein array, but the protein has a real AML literature that arrived through USP7 and interferon, not through the interactome
also called LRRK2 · PARK8 · dardarin · Leucine-rich repeat serine/threonine-protein kinase 2 · Q5S007
LRRK2 is the large kinase best known from Parkinson's disease. Its listing as an AFF4 partner comes from one glass-slide protein array, not from a cell - but separately, and much more interestingly, two papers put LRRK2 inside acute myeloid leukaemia biology.
The same thing, for a scientist
LRRK2 (UniProt **Q5S007, 2,527 aa**, synonym PARK8) is a dual kinase/GTPase with 2,339 distinct interaction partners across 88 publications in IntAct - by far the most-studied protein in this beat. **The row that put it in the interactome gap list is the weakest kind available**: a single IntAct row from Tomkins et al. 2018, method `protein array`, host **`taxid:-1 (in vitro)`**, one of **78** array hits for LRRK2 in that publication, in a study about the ROCO protein family rather than about AFF4 [E1]. It is annotated `direct interaction`, which reads stronger than a co-purification and is not: a purified-protein microarray reports binding on a surface, without cellular concentration, competition, compartment or post-translational state. UniProt gives LRRK2 twelve subcellular locations and **not one of them is the nucleus**, while AFF4 is nuclear and chromatin-associated [E4]. `LRRK2[tiab] AND (MLL[tiab] OR KMT2A[tiab])` returns **0** [E5]. **So the interaction is written `speculative` and should not be built on. What should be built on is the other half of this entry.** Park and Chung 2025 report that **USP7 deubiquitinates and stabilises LRRK2, that LRRK2 and USP7 expression is highest in AML among cancers, and that USP7 inhibition or knockdown suppresses AML cell growth by downregulating LRRK2, partially rescued by LRRK2 overexpression** [E2]. This graph already carries `gene-usp7` as *the only pharmacological handle on PRC1.1 that exists today*, so **there is now a published rival explanation for the anti-AML effect of USP7 inhibition that has nothing to do with PRC1.1**. Separately, Kuss et al. 2014 show IFN-gamma induces LRRK2 via **ERK5/MAPK7** in monocytes and in THP-1 acute monocytic leukaemia cells [E3] - and MAPK7 is itself one of the unwritten interactome-gap partners named in the same recount, so two members of that list sit in one published pathway.
This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.
I have never read one of these pictures before
A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.
Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.
Anything in yellow sticks is a bound small molecule, which is usually the drug.
Why a protein-array row is the weakest evidence class in this beat
IntAct's host organism field for this row is taxid:-1(in vitro). There is no cell. Purified or in-vitro-translated proteins are spotted on a surface and probed; binding is scored against the array background.
That design cannot see: physiological concentration, competition from the real binding partners of either protein, subcellular compartment, post-translational modification state, or whether the two proteins are ever expressed in the same cell at the same time. It is a useful discovery tool and a poor basis for asserting an interaction.
The interaction type label makes this worse rather than better. The row is psi-mi:"MI:0407"(direct interaction), which outranks association in the PSI-MI hierarchy and will sort higher in any ranking built on interaction type. It is a correct label - an array does test direct binding - and it is a misleading signal of strength.
And the source study is about ROCO-family proteins: LRRK2 is its subject, AFF4 is one of 78 things that stuck. Tomkins et al.'s own framing is comparative network analysis across the ROCO family, and their conclusion is about shared stress-response and cell-projection functions, not about any individual novel pair.
The USP7 finding, which is a correction owed to an entry this beat does not own
gene-usp7 in this graph exists because USP7 holds PRC1.1 together and USP7 inhibition is the only chemical route to turning PRC1.1 down. That entry's argument is that the anti-leukaemic effect of USP7 inhibitors is mediated by PRC1.1 disassembly.
Park and Chung 2025 report a different mediator in AML: USP7 directly binds LRRK2, removes K48-linked polyubiquitin and prevents its proteasomal degradation; USP7 and LRRK2 are co-highest in AML across cancer types; and USP7 inhibition or knockdown suppresses AML cell growth through LRRK2 downregulation, an effect partially reversed by LRRK2 overexpression 2.
A partial rescue by re-expressing a single substrate is a real mediation claim, and it is a rival to the PRC1.1 account of the same drug effect. The two are not mutually exclusive - a deubiquitinase has many substrates and can act through several at once - but gene-usp7 currently reads as though the PRC1.1 route is the explanation.
The correction is not mine to make. It is recorded here and in the memo with the citation, per the standing brief's rule on errors in files you do not own. What I am asserting is that a rival mediator has been published, not that the PRC1.1 account is wrong.
The system caveat is heavy: Park and Chung is AML cell lines, not B-lineage KMT2A::AFF1, and the abstract names no KMT2A-rearranged line.
Two gap-list partners in one pathway
The 2026-09-21 recount named thirteen partners not mentioned anywhere in this corpus. Two of them, LRRK2 and MAPK7, turn out to sit in a single published signalling relationship: IFN-gamma induces LRRK2 through ERK5 (the protein MAPK7 encodes) in human monocytes and in THP-1 acute monocytic leukaemia cells 3. A medicinal-chemistry paper on the same pair reports one benzo[e]pyrimido-diazepinone chemotype with activity against both ERK5 and LRRK2, so they are also pharmacologically entangled.
This is a fact about the two proteins, not about AFF4 or DOT1L, and neither the LRRK2-AFF4 array row nor whatever supports MAPK7 is strengthened by it. It is recorded because the recount treats gap-list members as independent items and they are not, and because an interferon-inducible axis is not neutral territory in this graph - cell-state-interferon-herv-mhc2 records a menin-inhibitor-induced interferon programme in this disease.
MAPK7 belongs to the other half of this cycle's interactome beat. The connection is flagged rather than developed.
| uniprot | Q5S007 |
|---|---|
| gene symbol | LRRK2 |
| length AA | 2527 |
| alphafold ID | AF-Q5S007-F1 |
| protein class | ROCO-family dual serine/threonine kinase and Roc GTPase |
| uniprot resolution | gene_exact:LRRK2 AND organism_id:9606 AND reviewed:true returned ONE reviewed row on 2026-09-21; primary gene name read back as LRRK2; entry LRRK2_HUMAN; 2,527 aa; synonym PARK8. |
| localization | Twelve curated locations, all cytoplasmic or membrane-associated: cytoplasmic vesicle, perikaryon, Golgi apparatus membrane, axon, dendrite, endoplasmic reticulum membrane, synaptic vesicle membrane, endosome, lysosome, mitochondrion outer membrane, cytoskeleton, phagosome (UniProt Q5S007, retrieved 2026-09-21). No nuclear annotation. |
| localization evidence code | ECO:0000269 for ten of twelve; ECO:0000250 for endosome and phagosome. Annotation is dominated by neuronal studies. |
| domains note | No domain boundaries asserted; the ARM, ANK, LRR, Roc, COR, kinase and WD40 ranges were not read back from UniProt with residue numbers this cycle. |
| shown for fusion | Nobody has looked in any fusion. The array pair is wild-type AFF4 against LRRK2 on a surface; there is no LRRK2 experiment in AFF1, in any KMT2A fusion, or in any KMT2A-rearranged cell. The AML evidence in this entry (USP7, IFN-gamma/ERK5) is in unspecified AML lines and in THP-1 monocytic leukaemia, neither of which is KMT2A-rearranged. |
domains
[]
interaction with the core complex
{
"partner": "AFF4",
"partner_accession": "Q9UHB7",
"our_protein_role": "prey",
"bait": "LRRK2 side annotated bait in the MITAB row; the assay is a protein microarray, so bait/prey is a spotting convention",
"method": "psi-mi:\"MI:0089\"(protein array)",
"interaction_type": "psi-mi:\"MI:0407\"(direct interaction)",
"expansion": "none",
"host_system": "taxid:-1 (in vitro) - there is no cell",
"source_publications": [
"29513927"
],
"distinct_experiments": 1,
"source_class": "comparative network analysis of the ROCO protein family; novel pairs from a protein microarray screen",
"lrrk2_hits_in_that_publication": 78,
"intact_acs": [
"intact:EBI-20587364"
],
"intact_miscore": 0.44,
"region_mapped": false,
"affinity_measured": false,
"structures": [],
"compartment_mismatch": "LRRK2 has twelve curated locations and none is nuclear; AFF4 is nuclear and chromatin-associated. The array assay has no compartment and cannot detect this."
}promiscuity controls 2026 09 21
{
"distinct_partners_in_pmid_29513927": 78,
"distinct_partners_all_of_intact": 2339,
"total_rows_all_of_intact": 4961,
"distinct_publications_all_of_intact": 88,
"note": "LRRK2 is the most heavily assayed protein in this beat by an order of magnitude, which makes any single row about it correspondingly less informative"
}counted absences
{
"kmt2a": {
"query": "LRRK2[tiab] AND (MLL[tiab] OR KMT2A[tiab])",
"records": 0
},
"leukaemia": {
"query": "LRRK2[tiab] AND (leukemia[tiab] OR leukaemia[tiab])",
"records_raw": 10,
"records_on_topic": 3,
"on_topic_pmids": [
"40907896",
"24479685",
"20697350"
]
},
"aliases_searched": [
"LRRK2"
],
"aliases_not_searched": [
"PARK8",
"dardarin"
],
"date": "2026-09-21",
"bound": "zero in titles and abstracts for the KMT2A intersection over one alias; supporting figures not searched"
}real leukaemia biology that did not come from the interactome
{
"usp7_axis": {
"claim": "USP7 deubiquitinates and stabilises LRRK2; both are highest-expressed in AML; USP7 inhibition suppresses AML growth via LRRK2 downregulation, partially rescued by LRRK2 re-expression",
"pmid": "40907896",
"graph_entry_affected": "gene-usp7",
"why_it_matters_here": "gene-usp7 attributes the anti-leukaemic effect of USP7 inhibition to PRC1.1 disassembly. A published rival mediator now exists for the same drug effect in AML."
},
"interferon_axis": {
"claim": "IFN-gamma induces LRRK2 via ERK5/MAPK7 in monocytes and THP-1 acute monocytic leukaemia cells",
"pmid": "24479685",
"cross_beat_note": "MAPK7 is itself one of the thirteen unwritten interactome-gap partners named in the 2026-09-21 recount, in the DOT1L half of this cycle's beat. Two gap-list members sit in one published pathway, and the recount treats them as independent."
}
}- If LRRK2 mediates part of the USP7 anti-AML effect, then a USP7 inhibitor developed on a PRC1.1 rationale has an unmeasured second mechanism, and its therapeutic index, resistance routes and biomarkers are all being estimated from the wrong pathway.
- LRRK2 kinase inhibitors are clinically advanced for Parkinson's disease, which makes repurposing look easy. Their documented on-target toxicity is lung type-II pneumocyte lamellar-body change, and LRRK2 is required in normal innate immunity - so an anti-leukaemic LRRK2 strategy carries an infection-risk cost on top of the pulmonary one, in a patient group already profoundly immunosuppressed.
- The ERK5 arm of the interferon result is confounded: the same benzo[e]pyrimido-diazepinone chemotype inhibits both ERK5 and LRRK2, so any 'ERK5 inhibitor lowers LRRK2' result needs its genetic arm to carry it.
- Is the USP7 anti-AML effect PRC1.1-mediated, LRRK2-mediated, or both, and does either route operate in KMT2A-rearranged B-lineage disease? A USP7 inhibitor in a KMT2A::AFF1 line with LRRK2 rescue and PRC1.1 readouts in the same experiment separates them, and this graph already argues that USP7 is the only chemical handle on PRC1.1 - so the answer changes how that handle should be interpreted.
- Does menin-inhibitor-induced interferon signalling raise LRRK2 in KMT2A-rearranged cells? This graph carries cell-state-interferon-herv-mhc2 for the menin-inhibitor interferon programme, and LRRK2 is an IFN-gamma-inducible gene in myeloid cells. Nobody has measured LRRK2 in that state.
- Is there any nuclear pool of LRRK2 in a haematopoietic cell? The twelve curated locations are largely neuronal, and the AFF4 array row is uninterpretable without knowing whether the two proteins can ever meet.
- E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.database
IntAct records a single AFF4-LRRK2 row, from Tomkins et al. 2018 (PMID 29513927), method protein array, interaction type direct interaction, host taxid:-1 (in vitro), MI-score 0.44; LRRK2 has 78 distinct partners from that publication.
IntAct molecular interaction database, PSICQUIC MITAB 2.7, query id:Q5S007 AND id:Q9UHB7, 1 row retrieved 2026-09-21 · IntAct (EMBL-EBI) 2026What this evidence cannot show
One row, one publication, one assay. The host organism field is taxid:-1 (in vitro) - there is no cell, no physiological concentration, no competition and no compartment. AFF4 is one of 78 array hits in a study whose subject is the ROCO protein family. The 'direct interaction' type outranks 'association' in the PSI-MI hierarchy and is a correct label for an array assay, which makes it a misleading signal of strength rather than a wrong one.
Evidence that argues against this (1)
- Tomkins et al.'s own design is to intersect a literature-derived weighted network with novel array hits and report the common core; the individual novel array hits are explicitly the unvalidated half of that design, and the paper's conclusions are drawn from the intersection, not from single pairs. Tomkins JE et al. 2018
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
USP7 directly binds LRRK2 and stabilises it by removing K48-linked polyubiquitin chains; USP7 and LRRK2 expression is highest in acute myeloid leukaemia among cancer types examined; and USP7 inhibition or knockdown suppresses AML cell growth via LRRK2 downregulation, an effect partially reversed by LRRK2 overexpression.
Ubiquitin-specific protease 7 promotes the growth and oncogenic potential of acute myeloid leukemia cells through the deubiquitination and upregulation of LRRK2. · Park JH, Chung KC · J Biol Chem 2025What this evidence cannot show
Read as abstract, not full text. AML cell lines; the abstract names no KMT2A-rearranged line, no B-lineage system and no primary material, so this is myeloid and in vitro. A partial rescue is a partial mediation claim - the magnitude of the LRRK2-dependent component is not stated in the abstract. Two authors, one group, no independent replication. New to this graph.
Evidence that argues against this (1)
- This graph attributes the anti-leukaemic effect of USP7 inhibition to disassembly of PRC1.1, and gene-usp7 is written on that basis. A deubiquitinase has many substrates and both routes can be real, but a partial rescue by a single substrate is a competing mediation claim for the same drug effect and gene-usp7 does not currently record it. Oncofusion knowledge graph, gene-usp7 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In human monocytes and in THP-1 acute monocytic leukaemia cells differentiated to macrophages, interferon-gamma induces LRRK2 through ERK5 (encoded by MAPK7), and pharmacological or RNAi inhibition of ERK5 suppresses that induction.
Interferon-gamma induces leucine-rich repeat kinase LRRK2 via extracellular signal-regulated kinase ERK5 in macrophages. · Kuss M, Adamopoulou E, Kahle PJ · J Neurochem 2014What this evidence cannot show
Read as abstract, not full text. Myeloid and monocytic: THP-1 is an acute monocytic leukaemia line, not a KMT2A-rearranged B-lineage model, and the paper's framing is Parkinson's-disease immunology. The ERK5 inhibitors used in that era are of the chemotype later shown to hit LRRK2 directly, so the pharmacological arm is exposed to an off-target objection that the RNAi arm is not. This relationship says nothing about AFF4.
Evidence that argues against this (1)
- One benzo[e]pyrimido-diazepinone chemotype has structure-activity relationships reported for both ERK5/MAPK7 and LRRK2, so an ERK5 inhibitor result in an LRRK2 readout is confounded by construction unless the genetic arm carries it. Deng X et al. 2013
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
LRRK2 is UniProt Q5S007, 2,527 aa, primary gene name LRRK2, synonym PARK8, a serine/threonine kinase and GTPase regulating RAB GTPases, autophagy, vesicle trafficking and innate immunity; UniProt lists twelve subcellular locations - cytoplasmic vesicle, perikaryon, Golgi membrane, axon, dendrite, ER membrane, synaptic vesicle membrane, endosome, lysosome, mitochondrial outer membrane, cytoskeleton and phagosome - and none of them is the nucleus.
UniProtKB entry Q5S007 (LRRK2_HUMAN, Leucine-rich repeat serine/threonine-protein kinase 2), subcellular location and function comments with evidence codes · UniProtKB/Swiss-Prot 2026What this evidence cannot show
Retrieved 2026-09-21 with evidence codes read, per brief rule 91: ten of the twelve locations carry ECO:0000269, two (endosome, phagosome) carry ECO:0000250. Absence of a nuclear annotation across twelve curated locations is a stronger negative than absence across one, but it is still an annotation's silence rather than a measured absence. A protein array cannot detect a compartment mismatch because there is no compartment in the assay.
Evidence that argues against this (1)
- LRRK2's annotation is dominated by neuronal studies, and its immune-cell biology - where it is interferon-inducible and highly expressed in monocytes - is more recent and less completely curated. A localisation profile built mostly in neurons may not describe LRRK2 in a myeloid or lymphoid cell. Kuss M et al. 2014
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
No literature connects LRRK2 to KMT2A or MLL: LRRK2[tiab] AND (MLL[tiab] OR KMT2A[tiab]) returns 0 records, while LRRK2[tiab] AND (leukemia[tiab] OR leukaemia[tiab]) returns 10, of which three are on-topic for LRRK2 in leukaemia biology and none mentions a KMT2A rearrangement.
PubMed counted-absence search for LRRK2 against MLL, KMT2A and leukaemia, run 2026-09-21, all ten leukaemia hits title-checked · PubMed (NCBI) 2026What this evidence cannot show
All ten leukaemia records were title-checked. On-topic: PMID 40907896 (USP7/LRRK2 in AML), 24479685 (IFN-gamma/ERK5/LRRK2 in THP-1), 20697350 (BCR-ABL1 proximal signalling network). Off-beat: three Parkinson's/autophagy papers, two cancer-outcome epidemiology papers in Parkinson's cohorts, one IBD review, one chemical-proteomics methods paper. Aliases searched: LRRK2 only; PARK8 and dardarin were not searched, and dardarin in particular carries the older literature. Zero in titles and abstracts for the KMT2A intersection; supporting figures not searched.
Evidence that argues against this (1)
- The retired-symbol rule bites here: PARK8 and dardarin were the working names in the 2004-2008 literature and were not searched, and this project has measured a tenfold undercount from exactly that omission in another family. The KMT2A zero is bounded by a one-alias list. Park JH et al. 2025
A SECOND MEDIATOR FOR THE SAME DRUG EFFECT, and it arrived from an unrelated direction. USP7 is reported as the deubiquitinase stabilising LRRK2 against proteasomal degradation, with both highest in AML and USP7 inhibition's growth suppression PARTIALLY RESCUED by LRRK2 overexpression (PMID 40907896). gene-usp7 attributes that anti-leukaemic effect to PRC1.1 disassembly. A partial rescue says some of it runs through a substrate with nothing to do with Polycomb. Note the two entries reached this graph independently - LRRK2 entered as a single protein-array row naming it an AFF4 partner - so they do not corroborate each other.
one in-vitro protein-array row (host taxid:-1), one of 78 array hits for LRRK2 in a ROCO-family study
LRRK2 is IFN-gamma-inducible via ERK5/MAPK7 in monocytes and THP-1; never measured in the menin-inhibitor interferon state
twelve curated locations and not one is nuclear