GPHN (gephyrin) — synaptic scaffold and molybdenum-cofactor enzyme, KMT2A fusion partner in t(11;14)(q23;q24)
also called gephyrin · GPH · KIAA1385 · GEPH_HUMAN
Gephyrin is a brain protein that acts like a peg board: it clusters the receptors that calm nerve cells, and in the rest of the body it helps build a chemical the liver needs. It has nothing to do with switching genes on, which is why it matters so much that a tiny 15-letter piece of it is enough to turn a leukaemia gene into a cancer driver.
The same thing, for a scientist
GPHN (UniProt Q9NQX3, GEPH_HUMAN, 736 aa) is a bifunctional protein: the terminal two steps of molybdenum-cofactor biosynthesis are carried out by its N-terminal MPT Mo-transferase (G) domain at aa 14-166 and its C-terminal MPT adenylyltransferase (E) domain at aa 326-736, while in the CNS it scaffolds glycine and GABA-A receptors at the inhibitory postsynapse [E1][E2][E5]. Between the two enzymatic domains lies a poorly ordered central linker (the C domain, disordered at aa 181-232 and 260-290) encoded in part by alternatively spliced exons 8, 13 and 14, which Ramming et al. predicted encode motifs mediating oligomerisation and tubulin binding [E2]. Gephyrin binds polymerised tubulin with high affinity and its transcripts are found in all rat tissues examined, so it is not a neuron-specific protein [E3]. Its self-assembly is the best-characterised part of its biology: G-domain trimerisation and E-domain dimerisation generate higher-order scaffolds, and cryo-EM has now resolved gephyrin filaments built from E-domain dimers linked by Z-shaped subdomain-II interfaces [E5]. **The leukaemia-relevant point is that gephyrin carries no transcriptional activity, no chromatin-reading or -writing domain, and no elongation-machinery interaction** — yet a 15-amino-acid tubulin-binding segment of it is necessary and sufficient to convert MLL into a transforming oncogene, and the same segment confers oligomerisation capacity on MLL [E4]. GPHN is not a dependency anywhere: DepMap Chronos mean +0.071 across 1208 lines and rank 13,812 of 17,728 for KMT2A selectivity, and it survives a lineage split as a non-dependency in both halves of the KMT2A-rearranged arm [E6].
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.
Two jobs, neither of them gene regulation
Gephyrin is one of the few proteins in the human genome that does two unrelated things well. In peripheral organs it catalyses the last two steps of molybdenum-cofactor (Moco) synthesis, using domains homologous to the bacterial MogA and MoeA proteins 12. In the central nervous system it clusters glycine receptors and GABA-A receptors beneath the postsynaptic membrane; gephyrin-knockout mice lose synaptic GABA-A receptor clusters containing the gamma2 or alpha2 subunit entirely, while glutamate receptors localise normally and total GABA-A receptor levels are unchanged [E5a].
Neither job involves DNA. There is no bromodomain, no YEATS domain, no PHD finger, no transactivation domain and no annotated interaction with the super elongation complex or with DOT1L in the UniProt record 1. This is the whole reason gephyrin is interesting to this project: it is a clean negative control that nevertheless transforms.
The oligomerisation machinery, and where the 15 residues sit
Gephyrin self-assembles through both enzymatic domains: the G domain trimerises and the E domain dimerises, and the combination is the classical explanation for the hexagonal postsynaptic lattice. Single-particle cryo-EM has now shown that E-domain dimerisation produces filaments, in which two E-domain dimers are linked by Z-shaped interfaces formed between the subdomain II (SDII) regions of adjacent dimers; deleting SDII, introducing either of two epilepsy-causing pathogenic variants, or neutralising an opposing charge in that interface abolishes filament formation, in vitro phase separation and synaptic receptor clustering in hippocampal neurons 5.
The 15 residues that matter for leukaemia are not in either enzymatic domain. Eguchi et al. localised the transforming activity to a small tubulin-binding domain 4, and the tubulin-binding and oligomerisation motifs of gephyrin map to the central linker encoded by exons 8, 13 and 14, between the G and E domains 2. Two consequences follow. First, the transforming module of MLL-GPHN is structurally unrelated to the coiled coils of EPS15 and GAS7, so "forced oligomerisation" in KMT2A fusions is not a statement about coiled coils. Second, the enormous structural literature on gephyrin self-assembly — trimers, dimers, filaments, phase separation — is about the G and E domains and is therefore not direct evidence about how the MLL-GPHN oncoprotein oligomerises.
There is a sharp irony in the linker's status. In the phase-separation account of the inhibitory postsynaptic density, gephyrin alone does not condense at all, and the G-to-E linker autoinhibits condensate formation, with phosphorylation of linker residues or binding of dynein light chain to the linker relieving that inhibition [E5b]. So the same central region that the leukaemia literature treats as gephyrin's oligomerisation module is treated by the synaptic literature as a brake on assembly. The reconciliation is presumably that removing the G domain (as the fusion does) removes the thing the linker was inhibiting — but nobody has tested that, and it is a concrete, cheap experiment.
Honest limit: the exact residue numbers of the 15-amino-acid domain were not obtained. The Eguchi 2004 full text is subscription-only (Europe PMC isOpenAccess: N, no PMC record) and only the abstract was read. Per the graph's residue rule, no domain entry with guessed or null boundaries has been written for it.
Expression, localization and cell state — what is known and what is not
Tissue. Gephyrin transcripts are present in all rat tissues examined, with at least five alternatively spliced brain mRNAs, and further splice variants are differentially expressed across non-neuronal tissues and brain regions 23. It is not a neuron-restricted protein, which removes the obvious objection that a neuronal protein could not be expressed in a haematopoietic cell.
Subcellular. Submembranous, at the cytoplasmic face of the postsynaptic membrane in neurons, and associated with polymerised tubulin — gephyrin coprecipitates with polymerised tubulin when expressed in 293 kidney cells, i.e. the tubulin association is intrinsic and not neuron-specific 3.
Subnuclear. No nuclear localisation for wild-type gephyrin is established, and no subnuclear distribution has been reported for the MLL-GPHN fusion protein. This is stated explicitly because the analogous experiment was done for KMT2A::SEPTIN6 (nuclear speckles) and for several other cytoplasmic-partner fusions; for GPHN it simply has not been done.
Cell type and cell state in the leukaemia context. MLL-GPHN transduced into haematopoietic progenitor cells caused both myeloid and lymphoid lineage leukaemias in mice, which the authors read as evidence that MLL-GPHN targets multipotent progenitors 4. The specific progenitor compartment was not resolved by marker sorting in that report.
Epigenetics. Nothing. No H3K79 methylation, no H3K4 methylation, no HOXA-locus ChIP and no transcriptome has ever been published for MLL-GPHN or for any GPHN-rearranged patient sample. The entry says so rather than importing a figure from MLL-AF9.
Essentiality: not a dependency, in either lineage half
GPHN is not required by cancer cells. Across 1208 DepMap lines the mean Chronos gene effect is +0.071 (i.e. very slightly growth-promoting when cut, which is noise around zero), and in the project's genome-wide KMT2A-selectivity scan GPHN ranks 13,812 of 17,728 with a KMT2A-rearranged mean of +0.091 against +0.052 in other leukaemia lines — marginally less required in KMT2A-rearranged lines 6.
Because the coordinator established that the KMT2A-rearranged DepMap panel is myeloid-enriched and that a myeloid-identity gene therefore scores as KMT2A-selective for free, the index arm was split by lineage for this entry rather than the combined number being quoted. GPHN scores -0.41 SD in the myeloid half and +1.47 SD in the B-lineage half of the KMT2A-rearranged arm, relative to the other-leukaemia arm 6. Both halves are consistent with no dependency, and the large gap between them in the absence of any actual effect is the signature of noise in a gene nobody needs, not of lineage specificity.
The decisive caveat is the same one that applies to every partner in this beat: no GPHN-rearranged cell line exists in DepMap. This measures wild-type gephyrin in leukaemias driven by other partners.
Human loss-of-function phenotype
Complete gephyrin loss is not tolerated in mouse: geph-/- animals lose synaptic clustering of glycine and GABA-A receptors [E5a]. In humans, GPHN is the gene of molybdenum cofactor deficiency type C, a severe neonatal neurological disease — relevant here only as a bound on how far a hypothetical gephyrin-directed agent could be pushed, which is a reason NOT to target gephyrin itself and to target the fusion's assembled interface instead. This entry does not cite a primary Moco-deficiency paper because none was fetched for it; the Moco-biosynthetic function itself is cited from 12.
| uniprot | Q9NQX3 |
|---|---|
| uniprot verified | Q9NQX3 = GEPH_HUMAN, PRIMARY gene name GPHN, synonyms GPH and KIAA1385, 736 aa, Homo sapiens. Fetched from the UniProt REST API on 2026-09-19 and the primary gene name read back rather than trusting a gene_exact synonym match; sequence length agrees with the value asserted here. |
| length AA | 736 |
| ensembl | ENSG00000171723 |
| ensembl verified | ENSG00000171723 confirmed to resolve to approvedSymbol GPHN by the Open Targets Platform target query used in the lineage-split run; the run aborts any gene whose Ensembl id resolves to a different symbol, and two ids in the same panel were caught that way (see data.identifier_failures). |
| PDB note | 1JLJ = '1.6 Angstrom crystal structure of the human neuroreceptor anchoring and molybdenum cofactor biosynthesis protein gephyrin', X-ray diffraction, 1.6 A. Title, method and resolution verified against the RCSB entry API on 2026-09-19. It is the only PDB entry cross-referenced from UniProt Q9NQX3 and it covers an enzymatic domain, NOT the central tubulin-binding segment that matters for leukaemia. No structure of MLL-GPHN or of its oligomerisation interface exists. |
| alphafold ID | AF-Q9NQX3-F1 |
| domains omitted | The 15-amino-acid tubulin-binding / oligomerisation domain that is necessary and sufficient to transform MLL is deliberately ABSENT from the domains array. Its residue boundaries were not verified from the primary text (Eguchi 2004 is subscription-only and both publisher mirrors returned HTTP 403), and the graph's rule forbids a domain entry with guessed or null coordinates. It lies in the central linker between the G and E domains, encoded in the region Ramming et al. assign to exons 8/13/14. |
| localization | submembranous at the inhibitory postsynapse in neurons; microtubule-associated (binds polymerised tubulin) and cytoplasmic in non-neuronal cells |
| subnuclear localization | NOT ESTABLISHED. No nuclear localisation is reported for wild-type gephyrin and no subnuclear distribution has ever been reported for the MLL-GPHN fusion protein. |
| epigenetic readout | NONE PUBLISHED. No H3K79me2, H3K4me3, HOXA-locus ChIP or transcriptome has been reported for MLL-GPHN or for a GPHN-rearranged patient sample. |
| family | MoaB/Mog and MoeA fusion protein; synaptic scaffold |
| second function | terminal two steps of molybdenum-cofactor biosynthesis; GPHN is the gene of molybdenum cofactor deficiency type C |
PDB IDS
[ "1JLJ" ]
domains — the numbers behind the chart
[
{
"name": "MPT Mo-transferase domain (G domain)",
"start": 14,
"end": 166
},
{
"name": "GABARAP-interaction region",
"start": 140,
"end": 316
},
{
"name": "MPT adenylyltransferase domain (E domain)",
"start": 326,
"end": 736
}
]essentiality
{
"depmap_common_essential": false,
"chronos_mean": 0.0711,
"chronos_n_lines": 1208,
"kmt2a_selectivity_rank": 13812,
"kmt2a_selectivity_of": 17728,
"kmt2a_r_mean": 0.091,
"other_leukaemia_mean": 0.0519,
"lineage_split_z_kmt2a_myeloid": -0.41,
"lineage_split_z_kmt2a_b_lineage": 1.47,
"lineage_split_verdict": "no dependency in either half; the gap between halves is noise around a zero effect",
"mouse_ko": "geph-/- mice lose synaptic glycine and GABA-A receptor clustering; haematopoiesis was not examined"
}identifier failures
[ "ENSG00000213445, used in the first draft of the lineage-split panel for MLLT11, resolves to SIPA1 in the Open Targets target record. The correct MLLT11 id is ENSG00000213190 (Ensembl REST xrefs/symbol lookup, 2026-09-19).", "ENSG00000079462, used for EPS15L1, resolves to PAFAH1B3. No corrected id was needed because EPS15L1 was a reference gene, not a target of this beat." ]
depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": 0.091,
"chronos_mean_other_leukaemia": 0.0519,
"difference": 0.0391,
"pan_essentiality_all_lines": 0.0711,
"selectivity_rank": 13812,
"rank_denominator": 17728,
"provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
"how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}- Gephyrin itself is a terrible drug target: it performs an essential metabolic function (molybdenum-cofactor biosynthesis, whose human loss causes a lethal neonatal disease) and an essential synaptic one, while being a dependency in no cancer line anywhere. Any tractable intervention must target the fusion's assembled interface, not the protein.
- If the 15-residue module works purely by forcing proximity, then almost any oligomerisation module substitutes for it — which is exactly what the synthetic-dimeriser experiments showed. An inhibitor that blocks one partner's oligomerisation interface would be expected to have no effect on any other KMT2A fusion, making the addressable population a handful of patients worldwide.
- The same logic cuts against the whole partner-directed strategy: a mechanism that any of dozens of unrelated modules can supply is a mechanism with dozens of escape routes.
- Which 15 residues? The single most valuable missing fact in this entry is the residue range of the gephyrin minimal transforming domain, which is stated in a subscription-only 2004 paper and is needed before anyone can model, mutate or target the MLL-GPHN oligomerisation interface. A library copy of Blood 103(10):3876-82 would settle it in five minutes.
- Does MLL-GPHN require menin? No KMT2A-GPHN fusion has ever been tested against a menin inhibitor, and because the minimal transforming module is only 15 residues with no elongation or chromatin function, MLL-GPHN is the cleanest available test of whether menin dependence is a property of the KMT2A N-terminus alone rather than of the partner.
- Is the MLL-GPHN oligomer a dimer, a trimer or a filament? Gephyrin's own assembly is trimeric through the G domain and filamentous through the E domain, and the fusion retains neither intact — so the stoichiometry of the oncoprotein is unknown and is not predictable from gephyrin structural biology.
- Does the gephyrin segment in the fusion still bind tubulin in a haematopoietic cell, and does the fusion therefore have a cytoplasmic pool? Nobody has imaged the protein.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Human gephyrin is UniProt Q9NQX3 (GEPH_HUMAN), primary gene name GPHN, 736 aa, Homo sapiens, with an MPT Mo-transferase region at aa 14-166, a GABARAP-interaction region at aa 140-316, disordered segments at aa 181-232 and 260-290, and an MPT adenylyltransferase region at aa 326-736; the record annotates no chromatin-binding, transcription-activation or elongation-complex domain.
UniProtKB Q9NQX3 (GEPH_HUMAN) Gephyrin, Homo sapiens, 736 aa · UniProt Consortium · UniProtKB 2026What this evidence cannot show
Canonical isoform only. Gephyrin is heavily alternatively spliced in a tissue-specific way, so residue numbers are isoform-dependent and the isoform present in a haematopoietic cell is not established. Resolved by fetching the accession and reading back the PRIMARY gene name (GPHN) and the sequence length, not by trusting a gene_exact synonym match.
Evidence that argues against this (1)
- An absent annotation is not an absent function. UniProt annotates no nuclear role for gephyrin, but gephyrin has documented partners in signalling and cytoskeletal pathways (GABARAP, dynein light chains, RAFT1/mTOR) that were unknown when the entry was first built, and the founding fusion report explicitly speculated that gephyrin domains 'could contribute novel signal sequences'. The negative should be read as 'not annotated', not as 'shown not to exist'. Eguchi M et al. 2001
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Gephyrin is essential both for postsynaptic localisation of inhibitory neurotransmitter receptors in the CNS and for molybdenum-cofactor biosynthesis in peripheral organs; the murine gene has 29 exons of which eight are alternatively spliced, the N- and C-terminal domains encoded by exons 3-7 and 16-29 resemble bacterial, invertebrate and plant Moco-biosynthetic proteins, and the central exons 8, 13 and 14 encode motifs that may mediate oligomerisation and tubulin binding.
Diversity and phylogeny of gephyrin: tissue-specific splice variants, gene structure, and sequence similarities to molybdenum cofactor-synthesizing and cytoskeleton-associated proteins. · Ramming M, Kins S, Werner N, Hermann A, Betz H, Kirsch J · Proc Natl Acad Sci U S A 2000What this evidence cannot show
Mouse gene structure and mouse/rat splice variants; the exon-to-motif assignment for oligomerisation and tubulin binding is stated by the authors as a sequence-based inference ('may mediate'), not a mapped biochemical result. Exon numbering is murine and does not transfer to the human GPHN transcript without checking.
Evidence that argues against this (1)
- The structural work that followed places the dominant oligomerisation determinants in the G and E domains, not in the central linker: E-domain dimerisation drives filament formation through subdomain-II interfaces, and disrupting those abolishes assembly. So 'the central exons encode oligomerisation motifs' is at best incomplete as an account of gephyrin self-assembly, even though the central region is where the leukaemogenic 15 residues lie. Macha A et al. 2025
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Gephyrin is a 93 kDa polypeptide associated with the inhibitory glycine receptor that binds with high affinity to polymerised tubulin; its transcripts are found in all rat tissues examined, at least five different gephyrin mRNAs are generated by alternative splicing in brain, and gephyrin expressed in 293 kidney cells coprecipitates with polymerised tubulin, defining it as a novel type of microtubule-associated protein.
Primary structure and alternative splice variants of gephyrin, a putative glycine receptor-tubulin linker protein. · Prior P, Schmitt B, Grenningloh G, Pribilla I, Multhaup G, Beyreuther K, et al. · Neuron 1992What this evidence cannot show
Rat protein and rat tissue panel; 'all tissues examined' is a Northern-blot statement from 1992 and is not a quantitative expression atlas, and it says nothing about protein level in haematopoietic progenitors specifically. The tubulin coprecipitation was done in transfected 293 cells with polymerised tubulin, not in situ.
Evidence that argues against this (1)
- Ubiquitous transcript presence conceals strongly tissue-specific isoform usage: splice variants of gephyrin are differentially expressed between non-neuronal tissues and between brain regions, so a statement that gephyrin is 'present everywhere' does not establish that the tubulin-binding, oligomerisation-competent isoform is present in a haematopoietic progenitor. Ramming M et al. 2000
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Gephyrin, a neuronal receptor assembly protein and rare MLL fusion partner, has the capacity as an MLL-GPHN chimera to transform haematopoietic progenitors despite lacking transcriptional activity; a small 15-amino-acid tubulin-binding domain of GPHN is necessary and sufficient for this activity in vitro and in vivo, the same domain confers oligomerisation capacity on MLL protein, and transduction of MLL-GPHN into haematopoietic progenitor cells caused myeloid and lymphoid lineage leukaemias in mice.
The small oligomerization domain of gephyrin converts MLL to an oncogene. · Eguchi M, Eguchi-Ishimae M, Greaves M · Blood 2004What this evidence cannot show
READ FROM THE ABSTRACT ONLY. Europe PMC reports this article as not open access with no PMC record, and both the publisher and ScienceDirect mirrors returned HTTP 403, so the full text was not obtained. Consequently: the residue numbers of the 15-aa domain, the assay used to demonstrate oligomerisation, the mouse strain and disease latencies, the transcriptional-activity assay, and any Hox data are NOT verified here and are not asserted anywhere in this entry. This is a single laboratory's result and has not, as far as the searches in ana-kmt2a-deferred-partner-enumeration found, been independently replicated in 22 years.
Evidence that argues against this (1)
- The strongest published challenge to reading any of this as a general oligomerisation mechanism is that the paradigm MLL fusion protein is NOT dimerised: Martin et al. showed that induced dimerisation of truncated MLL immortalises bone marrow and upregulates Hoxa7, Hoxa9 and Meis1, but stated explicitly that MLL-AF9 is not dimerised in vivo. Forced oligomerisation is therefore a sufficient route to transformation, not the route the common fusions take. Martin ME et al. 2003
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Single-particle cryo-EM shows that gephyrin E-domain dimerisation promotes formation of filaments in which two E-domain dimers are linked by Z-shaped interfaces between subdomain II (SDII) of adjacent dimers; deletion of SDII, introduction of either of two epilepsy-causing pathogenic variants, or neutralisation of an opposing charge in the interface abolishes filament formation, in vitro phase separation, and synaptic receptor clustering in hippocampal neurons.
Gephyrin filaments represent the molecular basis of inhibitory postsynaptic densities. · Macha A, Liebsch F, Bruckisch EHW, Burdina N, von Stülpnagel I, Benting K, et al. · Nat Commun 2025What this evidence cannot show
Purified protein plus cultured hippocampal neurons. This is the biology of full-length gephyrin at the inhibitory postsynapse and is included here as background on how gephyrin oligomerises; the MLL-GPHN oncoprotein retains only the C-terminal portion of gephyrin and its assembly has never been structurally characterised, so this evidence must not be transferred to the fusion.
Evidence that argues against this (1)
- A competing model for the same system holds that the inhibitory postsynaptic density forms by phase separation of a gephyrin/receptor complex rather than as a defined filament; in that work gephyrin ALONE does not form condensates, multivalent gephyrin-E-domain/receptor interactions are essential, and — directly relevant here — the linker between the G and E domains AUTOINHIBITS condensate formation. Since the leukaemogenic 15-residue module lies in that same linker, the two literatures assign opposite signs to the same region, and the structural basis of gephyrin higher-order assembly is not settled. Bai G et al. 2021
- E5aestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In gephyrin knockout mice, synaptic GABA-A receptor clusters containing the gamma2 or alpha2 subunit are absent from brain sections and cultured hippocampal neurons while glutamate receptor subunits localise normally, normal levels of functional GABA-A receptors are still expressed, and the intracellular receptor pool is enlarged — so gephyrin is required for synaptic localisation of glycine and GABA-A receptors but not for their delivery to the plasma membrane.
Loss of postsynaptic GABA(A) receptor clustering in gephyrin-deficient mice. · Kneussel M, Brandstätter JH, Laube B, Stahl S, Müller U, Betz H · J Neurosci 1999What this evidence cannot show
A germline mouse knockout scored for a neuronal phenotype. It bears on the toxicity ceiling for any gephyrin-directed agent and on nothing else in this entry; it is not evidence about haematopoiesis, which was not examined.
Evidence that argues against this (1)
- The knockout phenotype is dominated by loss of the Moco-biosynthetic function, which is lethal in its own right, so the neuronal clustering defect cannot be cleanly separated from systemic metabolic failure in this model; gephyrin's dual function makes its null mouse a poor instrument for asking what the synaptic scaffold alone does. Ramming M et al. 2000
- E5bstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Gephyrin in complex with glycine or GABA-A receptors spontaneously forms highly condensed assemblies by phase separation in solution and on supported bilayers; multivalent interactions between the dimeric gephyrin E domain and the receptor multimer are essential; gephyrin ALONE does not form condensates; and the linker between the G and E domains inhibits condensate formation by autoinhibition, with phosphorylation of specific linker residues or binding of dynein light chain to the linker regulating gephyrin-mediated receptor clustering.
Gephyrin-mediated formation of inhibitory postsynaptic density sheet via phase separation. · Bai G, Wang Y, Zhang M · Cell Res 2021What this evidence cannot show
Reconstituted purified proteins and supported lipid bilayers, not neurons or leukaemic cells. The autoinhibition result is about full-length gephyrin containing the G domain; the MLL-GPHN oncoprotein retains the linker and the E-domain region but not the G domain, so whether the autoinhibition applies to the fusion is untested and is asserted nowhere in this entry.
Evidence that argues against this (1)
- The cryo-EM account reaches a different structural conclusion for the same system, finding that E-domain dimers polymerise into defined filaments through subdomain-II interfaces and that abolishing those interfaces abolishes both filaments and phase separation — which makes the condensate a consequence of the filament rather than an alternative to it. Macha A et al. 2025
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
GPHN is not a dependency in any leukaemia context measured: DepMap Chronos mean gene effect +0.0711 across 1208 lines, KMT2A-rearranged mean +0.091 versus +0.052 in other leukaemia lines, rank 13,812 of 17,728 (77.9th percentile) for KMT2A selectivity; and when the KMT2A-rearranged index arm is split by lineage as the myeloid-confound analysis requires, GPHN scores -0.41 SD in the 8-line myeloid half and +1.47 SD in the 5-line B-lineage half of the other-leukaemia distribution, i.e. no dependency in either half.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API: genome-wide KMT2A-selectivity scan (analyses/depmap-kmt2a-selective-scan-full/results.json) plus a lineage-split re-analysis of the deferred KMT2A partners run for this beat (analyses/depmap-deferred-partners-lineage-split/) · Broad Institute DepMap; Open Targets Platform; lineage split run by this project · Database and derived analysis 2026What this evidence cannot show
Decisive limitation: no GPHN-rearranged cell line exists in DepMap, so this is wild-type gephyrin measured in leukaemias driven by other KMT2A partners. The lineage split rests on 8 myeloid and 5 B-lineage KMT2A-rearranged lines, so one line moves a mean; the +1.47 SD in the B-lineage half is noise around a gene effect of essentially zero, not a protective effect. Chronos scores near zero are where noise dominates. DepMap release version is not stated by the API.
Evidence that argues against this (1)
- The lineage split can itself mislead in the direction of false reassurance: the coordinator's own analysis records that the split is by lineage LABEL rather than by expression signature, that KMT2A-rearranged disease includes genuinely ambiguous-lineage cases forced into one bin, and that the procedure bounds lineage specificity without testing therapeutic relevance. Oncofusion knowledge graph et al. 2026
C-terminal half of gephyrin including the tubulin-binding site and the MoeA/E domain
- ← fuses withKMT2A / MLL1 — the gene that breaks
t(11;14)(q23;q24), two published cases
the partner protein of the t(11;14)(q23;q24) fusion
deferred by that enumeration at 3 PubMed records; written here, and its 'single therapy-related patient' description corrected to two independent index cases
the minimal transforming module is named a tubulin-binding domain by its discoverers
a 15-residue tubulin-binding oligomerisation module is what the fusion retains