# CDK5 notes

CDK5 is reviewed in the PN autophagosome maturation / lysosome fusion "specific function unknown" bucket. The PN row cites Drosophila Cdk5 phosphorylation of Acinus as basal-autophagy context, but the PN mapping is no-mapping/context-only, so that fly evidence is not used as direct human CDK5 GO evidence.

Human CDK5 is a proline-directed serine/threonine kinase activated by CDK5R1/p35 or CDK5R2/p39, with core roles in neuronal development, synaptic function, cytoskeletal regulation, and stress-linked neuronal death. UniProt states that CDK5 is a "Proline-directed serine/threonine-protein kinase" and lists many substrates, including MAPT/tau, MAP2, MAP1B, SYN1, DNM1, AMPH, SYNJ1, SH3GLB1, p35/CDK5R1, MEF2A, TP53, and HTT [file:human/CDK5/CDK5-uniprot.txt, "Proline-directed serine/threonine-protein kinase"; file:human/CDK5/CDK5-uniprot.txt, "Phosphorylates SRC, NOS3, VIM/vimentin"]. UniProt also summarizes neuronal functions including "neuronal survival, migration and differentiation, axonal and neurite growth, synaptogenesis, oligodendrocyte differentiation, synaptic plasticity and neurotransmission" [file:human/CDK5/CDK5-uniprot.txt, "neuronal survival, migration and differentiation, axonal and neurite growth, synaptogenesis, oligodendrocyte differentiation, synaptic plasticity and neurotransmission"].

The core kinase-complex context is CDK5 plus p35/p25/p39-family activators. UniProt states that CDK5 is "Activated by interaction with CDK5R1 (p35) and CDK5R2 (p39)" and that the active enzyme is a "Heterodimer composed of a catalytic subunit CDK5 and a regulatory subunit CDK5R1" [file:human/CDK5/CDK5-uniprot.txt, "Activated by interaction with CDK5R1 (p35) and CDK5R2 (p39)"; file:human/CDK5/CDK5-uniprot.txt, "Heterodimer composed of a catalytic subunit CDK5 and a regulatory subunit CDK5R1"]. The early structural/biochemical evidence supports CDK5/p25 complexes, but broad "protein binding" annotations should be replaced by the specific kinase-complex and kinase-activity terms.

CDK5 has direct human/neuron autophagy evidence through SH3GLB1/endophilin B1. The Nat Cell Biol paper reports an "unexpected role for Cdk5 in the regulation of induced autophagy in neurons", identifies EndoB1/SH3GLB1 as a substrate, and shows that CDK5-mediated EndoB1 phosphorylation is required for "autophagy induction in starved neurons" [PMID:21499257, "unexpected role for Cdk5 in the regulation of induced autophagy in neurons"; PMID:21499257, "autophagy induction in starved neurons"]. This supports retaining `regulation of macroautophagy` as a real, PN-relevant CDK5 function, even though CDK5 is not a constitutive core autophagy-machinery component.

Synaptic/NMDAR evidence supports specific receptor and synaptic-plasticity annotations, but not broad generic protein binding. Plattner et al. show that "serine 1116 of NR2B is phosphorylated by cyclin-dependent kinase 5 (Cdk5)" and that CDK5-dependent NR2B phosphorylation "controls the receptor's cell surface expression" [PMID:24607229, "serine 1116 of NR2B is phosphorylated by cyclin-dependent kinase 5 (Cdk5)"; PMID:24607229, "controls the receptor's cell surface expression"]. The same abstract says disrupting the NR2B-CDK5 interaction increases NR2B surface levels, facilitates synaptic transmission, and improves memory formation [PMID:24607229, "increases NR2B surface levels, facilitates synaptic transmission, and improves memory formation"].

Some microtubule annotations need caution. PMID:17491008 is about the p35 activator, not direct CDK5 microtubule binding: it says "p35 binds directly to alpha/beta-tubulin and microtubules" and that microtubules inhibit CDK5-p35 activity [PMID:17491008, "p35 binds directly to alpha/beta-tubulin and microtubules"; PMID:17491008, "Microtubule polymers but not the alpha/beta-tubulin heterodimer block p35 interaction with Cdk5"]. Therefore CDK5 microtubule-binding annotations should be kept non-core or marked over-annotated unless the annotation is about the CDK5R1/p35-containing complex context rather than CDK5 itself.

Curation decisions:
- Accept protein serine/threonine kinase, cyclin-dependent protein serine/threonine kinase, tau-protein kinase, ATP-binding, and protein kinase 5 complex annotations.
- Accept major neuronal/synaptic/cytoskeletal process rows as CDK5 core biology where they match UniProt/reviewed evidence; keep very broad cell-cycle/localization/process rows non-core.
- Accept `regulation of macroautophagy` based on human/neuron SH3GLB1/endophilin B1 evidence, but describe it as PN-relevant substrate-specific kinase regulation rather than a direct autophagy-machinery role.
- Mark generic `protein binding` rows as over-annotated.
- Treat receptor-binding/activator/inhibitor activity terms from similarity transfer conservatively unless direct evidence makes them more specific than kinase regulation.

## Description cleanup note

The YAML `description` field was revised to keep it as a standalone biological summary. Project-specific curation framing moved here instead.

- Moved out of the YAML description: PN autophagy context was used to interpret CDK5 mainly through substrate-specific phosphorylation of SH3GLB1/endophilin B1; Drosophila Cdk5-Acinus basal-autophagy evidence was treated as orthology context rather than direct human CDK5 evidence. Generic protein-binding rows were considered less informative than kinase, complex, substrate-specific, and neuronal process terms.

## 2026-06-20 second-pass audit

The second-pass audit added manual `reference_review` metadata for CDK5/p35 activation, CDK5/p25 structural context, induced-autophagy phosphorylation of SH3GLB1/endophilin B1, NR2B/GluN2B phosphorylation, and the p35-specific microtubule-binding caveat. No annotation action changes were needed: CDK5 remains curated as the catalytic serine/threonine kinase in CDK5R1/CDK5R2-activated complexes, with autophagy and receptor/synaptic effects treated as substrate-specific kinase outputs rather than broad direct binding functions.

## Falcon deep research integration (2026-06-21)

The Falcon (Edison) report in `CDK5-deep-research-falcon.md` strongly corroborates the existing review: it confirms CDK5 as a CMGC-family proline-directed Ser/Thr kinase activated by non-cyclin p35 (CDK5R1)/p39 (CDK5R2) rather than cyclins, with calpain-cleaved p25/p29 driving mislocalized hyperactivation, tau/cytoskeletal/synaptic substrates, and neurodegeneration roles — all already captured. The following Falcon-sourced citations are *not yet independently verified against full text*.

Already covered (no change): non-cyclin p35/p39 activation; p25 (and p39→p29) calpain generation, loss of myristoyl membrane targeting, nuclear mislocalization, and neurotoxic hyperactivation; tau phosphorylation (T181/S202/T205/T217/S235/S396/S404) and Alzheimer's relevance; neuronal migration, synaptic/NMDAR (NR2A/NR2B), endocytosis (dynamin/amphiphysin), DARPP-32, neurodevelopment; p25/CDK5 cytoskeletal degeneration/apoptosis [Pao & Tsai, J Biomed Sci 2021, doi:10.1186/s12929-021-00774-y; Tian et al., Front Mol Neurosci 2022, doi:10.3389/fnmol.2022.1030639; Nikhil & Shah, Mol Cancer 2023, doi:10.1186/s12943-023-01895-8].

Genuinely NEW or REFINED beyond current notes/review (candidate substrate-specific kinase outputs; would expand BP coverage, not the core MF):
- Mitochondrial fission: CDK5 phosphorylates Drp1/DNM1L at Ser616 promoting fission, and Parkin at Ser131 reducing its E3 ligase activity (mitochondrial quality control); also Prx1 T90 / Prx2 T89 reducing peroxidase activity / increasing ROS [Tian et al., Front Mol Neurosci 2022, doi:10.3389/fnmol.2022.1030639; Pao & Tsai 2021, doi:10.1186/s12929-021-00774-y]. Relevant BP: GO:0000266 mitochondrial fission / GO:0090140 regulation of mitochondrial fission (not in review).
- mTORC1 activation: novel CDK5-PRMT1(S307)-WDR24/GATOR2 cascade activating mTORC1 in response to amino acids [Yin et al., Cell Reports 2023, doi:10.1016/j.celrep.2023.112316]. Relevant BP: GO:0032008 positive regulation of TOR signaling (not in review).
- Circadian clock: CDK5 phosphorylates CLOCK (T451/T461, nuclear translocation) and PER2 (S394, stabilization) [Pao & Tsai 2021, doi:10.1186/s12929-021-00774-y]. Relevant BP: GO:0042752 regulation of circadian rhythm (not in review).
- SIRT2 (S331/S335) nuclear translocation driving dopaminergic neuron death in Parkinson's models [Yan et al., npj Parkinsons Dis 2022, doi:10.1038/s41531-022-00311-0] — refines the existing p25/neurodegeneration framing with a specific substrate.
- APP phosphorylation at Thr668 regulating APP localization/processing [Pao & Tsai 2021, doi:10.1186/s12929-021-00774-y] — Alzheimer's-relevant substrate beyond tau.
- Tau Thr217 as a specific AD-biomarker site impairing synaptic structure [Fu et al., Transl Psychiatry 2025, doi:10.1038/s41398-025-03551-9] — refines the tau-kinase annotation.

Notable DISCREPANCY to flag: Falcon highlights a 2024 Nature paper (Zheng et al., doi:10.1038/s41586-024-07888-x) showing CDK5 forms a complex with cyclin B1 and regulates mitotic fidelity, "challenging the traditional view that CDK5 has no cell-cycle function." The current review keeps cell-cycle terms (GO:1901987 regulation of cell cycle phase transition; GO:0051726 regulation of cell cycle) as KEEP_AS_NON_CORE on the basis that CDK5 is non-canonical/non-mitotic. This new primary evidence (if it holds up on full-text review) could argue those cell-cycle annotations have a genuine direct basis rather than being incidental — worth revisiting the NON_CORE rationale, though not necessarily changing the action. Note also the report contains a likely citation-mismatch artifact (a "synaptic remodeling in stroke (2026)" claim attributed to the Zheng cyclin-B1 reference), so treat Falcon's author-year tokens with caution.

No existing annotation appears contradicted by Falcon; the new findings are additive process candidates, not corrections.