PERK (EIF2AK3)

PERK (PKR-like endoplasmic reticulum kinase; gene EIF2AK3) is a 1,116 aa single-pass ER transmembrane kinase and the ISR arm of the unfolded-protein-response. It senses misfolded protein accumulation in the ER lumen and responds by phosphorylating eif2alpha at Ser51, globally dampening cap-dependent translation while selectively amplifying stress-adaptive transcription via atf4. In aged tissues — particularly post-mitotic neurons — chronic ER stress drives sustained PERK activity, feeding forward through ATF4 → CHOP into senescence and apoptosis programs.

Identity

UniProt: Q9NZJ5 (E2AK3_HUMAN)
NCBI Gene: 9451
HGNC: 3255
Ensembl: ENSG00000172071
Gene symbol: EIF2AK3 (alias: PERK, PEK, HsPEK)
Mouse ortholog: Eif2ak3 (one-to-one; highly conserved)
Length: 1,116 amino acids (canonical isoform)
Topology: Single-pass type I ER membrane protein; N-terminus lumenal, C-terminus cytoplasmic

Domain structure

Domain	Residues (approx.)	Function
Lumenal sensor domain	30–514	Detects misfolded proteins; regulated by BiP (GRP78) binding and release
Transmembrane helix	515–535	ER membrane anchor; enables oligomerization
Kinase domain	536–1116	Catalytic core; autophosphorylates and phosphorylates eIF2α-Ser51
Insert loop	647–888	Substrate-recognition insertion within kinase domain; absent in IRE1α

Key PTMs:

Thr982 — activation-loop autophosphorylation; required for full catalytic activity ¹
Tyr619 — autophosphorylation; dephosphorylated by PTP1B (negative feedback)
Thr802 — phosphorylated by AKT; inactivating signal coupling insulin/IGF-1 signaling to PERK suppression

UPR sensor context

PERK is one of three ER-resident stress sensors that together constitute the unfolded-protein-response:

Sensor	Branch	Immediate effector	Main outputs
PERK	ISR / translational	eif2alpha Ser51 phosphorylation	Global translation attenuation; ATF4 selective translation
IRE1α	Transcriptional / splicing	XBP1 mRNA splicing	ER biogenesis genes
ATF6	Transcriptional	ATF6 proteolytic cleavage	Chaperone induction

Under basal conditions, the ER chaperone BiP (GRP78) binds the PERK lumenal domain and keeps PERK monomeric and inactive. When misfolded proteins accumulate, BiP is titrated away to client proteins, releasing PERK to oligomerize, trans-autophosphorylate, and become catalytically active ¹.

Lumenal domain mechanism

The BiP-release model is the dominant but not sole mechanism. Structural studies indicate that the PERK lumenal domain can also directly engage misfolded protein clients, analogously to the yeast IRE1 lumenal domain. The relative contributions of indirect BiP sensing vs direct client engagement to PERK activation in vivo remain incompletely resolved. no-mechanism

Signaling axis: PERK → eIF2α → ATF4

Active PERK phosphorylates eif2alpha at Ser51 → GDP-bound eIF2α inhibits the guanine nucleotide exchange factor eIF2B → global reduction of ternary complex assembly → suppression of cap-dependent translation ¹.
Paradoxically, phospho-eIF2α selectively promotes translation of mRNAs with upstream ORFs (uORFs), most notably atf4 (ATF4). ATF4 is the master transcriptional activator of the integrated-stress-response (ISR).
ATF4 targets include: amino acid biosynthesis genes, antioxidant programs (NRF2 targets), and — under sustained or severe stress — CHOP (DDIT3/GADD153), a pro-apoptotic / pro-senescence transcription factor.

The PERK → eIF2α → ATF4 arm is the ISR module within the UPR; the ISR can also be activated by three other eIF2α kinases (HRI, PKR, GCN2) in response to heme deficiency, dsRNA, and amino acid deprivation, respectively. These four kinases converge on the same Ser51 site — hence the “integrated” nomenclature.

Aging relevance

Chronic ER stress in aged neurons

Proteostasis capacity declines with age across tissues, but the effect is particularly pronounced in post-mitotic neurons that cannot dilute misfolded proteins by cell division ². In aged brains:

Basal eIF2α phosphorylation is elevated relative to young controls.
PERK activity is chronically, rather than transiently, elevated.
Sustained PERK → ATF4 → CHOP signaling tips cells from adaptive toward apoptotic outcomes.
Reduced cognitive function in aged mice tracks with elevated ISR tone; pharmacological ISR suppression with ISRIB (see Pharmacology) restores spatial and working memory ².

Dimension	Status
Pathway conserved in humans?	yes
Phenotype conserved in humans?	partial — age-associated cognitive decline is universal; direct evidence that elevated PERK/eIF2α-P drives human neurodegeneration is correlative
Replicated in humans?	no — mouse model only for the ISRIB rescue; human ISR-aging data are observational needs-human-replication

Connection to neurodegenerative disease

Elevated phospho-eIF2α and PERK activation are observed in post-mortem brain tissue from Alzheimer’s, Parkinson’s, and prion disease patients. Whether PERK activity is causally upstream of neuronal loss or a downstream consequence of proteotoxic stress remains debated. contradictory-evidence

The ER stress / UPR activation in tauopathy models correlates with granulovacuolar degeneration, a hallmark of neurodegenerative progression ³. needs-human-replication

Disease: Wolcott-Rallison syndrome

Loss-of-function mutations in EIF2AK3 cause Wolcott-Rallison syndrome (WRS), a rare autosomal recessive disorder ⁴:

Core features: Permanent neonatal or early-infancy insulin-dependent diabetes; multiple epiphyseal dysplasia; growth retardation
Variable features: Intellectual disability, hepatic dysfunction, exocrine pancreatic insufficiency, renal abnormalities
Genetic basis: Two distinct EIF2AK3 mutations identified in the original mapping study — one producing a truncated protein lacking the catalytic domain, one causing a conserved missense in the kinase domain ⁴
Mechanism: Beta-cell ER stress is uniquely high because of insulin biosynthesis load; PERK is the obligate sensor in this cell type. Without PERK, beta cells cannot mount an adaptive UPR and undergo apoptosis.

WRS establishes that PERK is non-redundant in human pancreatic beta-cell proteostasis. It also provides a natural precedent for the pancreatic toxicity seen with pharmacological PERK inhibition (see Pharmacology).

Pharmacology

PERK inhibitors: GSK2606414 and GSK2656157

GSK2606414 (the first selective PERK inhibitor) was identified in a medicinal chemistry campaign at GlaxoSmithKline ⁵:

Mechanism: ATP-competitive; IC50 ~0.4 nM against PERK in biochemical assay; >600-fold selectivity over related kinases
In vivo: Reduced tumor growth in xenograft models; produced neurological benefit in prion-infected mice (reduced spongiosis, improved motor performance) in proof-of-concept studies
On-target toxicity: Pancreatic acinar cell atrophy and weight loss observed at therapeutically relevant doses — directly mechanistic, recapitulating the WRS phenotype. This has blocked clinical advancement.

GSK2656157 — a second-generation analog with improved pharmacokinetic profile; similar efficacy and similar pancreatic toxicity profile.

Neither compound has entered clinical trials. The pancreatic liability is considered a mechanism-based class effect, not a compound-specific off-target issue. long-term-unknown

ISRIB: downstream ISR reversal as a more tractable approach

ISRIB (Integrated Stress Response InhiBitor) acts downstream of PERK by stabilizing the eIF2B decameric complex, maintaining eIF2B GEF activity even when eIF2α is phosphorylated. This makes cells less sensitive to the translational consequences of PERK activation without blocking PERK kinase activity directly.

Advantages of ISRIB over direct PERK inhibition:

Does not block the kinase domain → upstream adaptive signaling (e.g., ATF6/IRE1 branches) is preserved
Pancreatic toxicity not observed at cognitive-rescue doses in mouse models ²
Reverses age-related cognitive decline in mice with a favorable safety window

ISRIB is an active investigational target but has not yet entered clinical trials for aging indications. needs-human-replication

See compound page isrib (implicit stub) for PubChem and dosing data.

Pathway membership

integrated-stress-response — PERK is the UPR-coupled ISR kinase; convergent with GCN2, HRI, PKR at eIF2α-Ser51
unfolded-protein-response — one of three ER stress sensors alongside IRE1α and ATF6

Key interactors

eif2alpha (EIF2S1) — direct substrate; Ser51 phosphorylation is the defining catalytic output
atf4 — translationally upregulated downstream of eIF2α phosphorylation; master ISR transcription factor
chop (DDIT3) — ATF4 target; pro-apoptotic under sustained PERK activation
BiP/GRP78 (HSPA5) — lumenal negative regulator; release licenses PERK oligomerization
p58ipk (DNAJC3) — cytoplasmic co-chaperone that inhibits PERK during the late/recovery phase of UPR

Limitations and gaps

#gap/needs-human-replication — Evidence that chronic PERK activity drives neuronal aging phenotypes in humans is correlative; no interventional human data.
#gap/contradictory-evidence — Whether PERK activity in neurodegeneration is causal or reactive to existing proteotoxic load is unresolved.
#gap/no-mechanism — The relative contributions of indirect BiP-titration vs direct misfolded-protein engagement to lumenal domain activation are not fully resolved in vivo.
#gap/long-term-unknown — Chronic PERK inhibition safety data beyond rodent preclinical models are absent; no clinical trials.
#gap/needs-canonical-id — GenAge ID for EIF2AK3 not confirmed; field left null pending verification against HAGR database.
ISRIB compound page [[isrib]] is an implicit stub; needs seeding.
[[eif2alpha]], [[atf4]], [[chop]], [[p58ipk]] are implicit stubs; need seeding.
[[unfolded-protein-response]] is an implicit stub pathway page.

Footnotes

doi:10.1038/16729 · Harding HP et al. · Nature 1999 · in-vitro + in-vivo · model: mammalian cells + yeast · PERK discovery; demonstrated ER kinase phosphorylates eIF2α at Ser51; archive status: not downloaded (not OA) ↩ ↩² ↩³
doi:10.7554/eLife.62048 · Krukowski K et al. · eLife 2020 · in-vivo · n=not extracted · randomized · model: aged C57BL/6 mice · ISRIB reverses age-related cognitive decline; reduced ATF4; improved hippocampal electrophysiology; archive status: OA pending download ↩ ↩² ↩³
doi:10.1016/j.nbd.2014.07.006 · Köhler C et al. · Neurobiol Dis 2014 · in-vivo · model: mouse tauopathy + amyloidosis · phospho-PERK co-localizes with early tangle formation; archive status: OA pending download ↩
doi:10.1038/78085 · Délépine M et al. · Nature Genetics 2000 · observational (human genetics) · n=2 consanguineous families · EIF2AK3 identified as Wolcott-Rallison syndrome gene; local PDF available in a local paper archive ↩ ↩²
doi:10.1021/jm300713s · Axten JM et al. · J Med Chem 2012 · in-vitro + in-vivo · model: biochemical + xenograft · GSK2606414 discovery; IC50 ~0.4 nM; archive status: OA pending download ↩

🧬 Aging Wiki

Explorer

perk