Unfolded Protein Response

What the unfolded protein response actually does

Every secreted hormone, every antibody, every cell-surface receptor begins life as a floppy chain threaded into the endoplasmic reticulum (ER). Inside the ER lumen — an oxidizing, calcium-rich, chaperone-packed compartment — these chains fold, get disulfide bonds, and acquire sugar decorations before moving on. The ER is a folding factory running near capacity. When the order book exceeds the factory's throughput, partly folded and misfolded proteins accumulate. That backlog is ER stress, and the cell's response to it is the unfolded protein response.

The UPR is not a single switch but a three-pronged decision system with a clear logic. Step one: reduce the inflow by throttling translation, so fewer new clients arrive. Step two: increase the folding capacity by transcribing more chaperones, folding enzymes, lipid-synthesis genes (to physically expand the ER), and components of ER-associated degradation (ERAD), which retrotranslocates hopeless misfolds back to the cytosol for the proteasome. Step three, taken only if the first two fail: trigger apoptosis. In short, the UPR first tries to fix the problem, then — if it cannot — deletes the cell to protect the tissue.

How the cell senses a folding jam: BiP and three sensors

The central sentinel is BiP (binding immunoglobulin protein, also called GRP78), an HSP70-family chaperone that is the most abundant protein in the ER lumen. BiP does double duty. It chaperones folding clients by gripping their exposed hydrophobic stretches, and it sits on the lumenal domains of the three stress sensors, holding them inactive. This is an elegant titration sensor: when unfolded proteins accumulate, their exposed hydrophobic patches compete for BiP. As BiP is pulled away to deal with the misfolds, it lets go of the sensors, and they switch on. The folding load and the alarm threshold are linked by the same molecule.

The three sensors are transmembrane proteins, each a separate signaling branch:

• IRE1 (inositol-requiring enzyme 1) — the most ancient branch, present in budding yeast and conserved to humans. Its cytosolic side is both a kinase and an endoribonuclease (RNase). On activation, IRE1 oligomerizes and clusters into foci, trans-autophosphorylates, and its RNase excises a 26-nucleotide intron from XBP1 mRNA. This unconventional cytoplasmic splicing (by IRE1 plus the ligase RTCB, no spliceosome) frameshifts the message to produce XBP1s, a potent transcription factor for chaperones, ERAD genes, and ER/Golgi biogenesis.
• PERK (PKR-like ER kinase) — the rapid-brake branch. On activation it phosphorylates the translation initiation factor eIF2α on serine 51, which blocks the eIF2B guanine-nucleotide exchange step and shuts down roughly 80% of cap-dependent translation within minutes. Paradoxically, this lull lets a few special mRNAs with upstream open reading frames — notably ATF4 — translate better, turning on stress genes and, later, the death factor CHOP.
• ATF6 (activating transcription factor 6) — the membrane-released transcription factor. When BiP releases it, ATF6 traffics to the Golgi where two proteases, Site-1 protease (S1P) and Site-2 protease (S2P), clip it. The freed cytosolic fragment (ATF6f) enters the nucleus and drives chaperone and XBP1 transcription. This is the same regulated intramembrane proteolysis machinery that processes the cholesterol sensor SREBP.

The numbers: speed, scale, and energetics

The UPR's branches operate on strikingly different timescales, which is what makes it a graded, intelligent response rather than an on/off alarm. PERK acts within minutes — translational attenuation is the fastest, cheapest move because it does not require new transcription. The transcriptional branches (IRE1-XBP1s and ATF6f) take 30 minutes to a few hours to ramp up chaperone mRNA and protein. Adaptive recovery, if it happens, plays out over hours; the commitment to apoptosis typically arrives after 24–72 hours of unresolved stress.

The energetics are dominated by the cost of folding and the cost of failure. Protein synthesis is among the most expensive things a cell does — roughly 4 ATP-equivalents are spent per peptide bond, and a secretory cell can make tens of thousands of proteins per second. Throttling translation during stress is therefore both a load-shedding and an energy-saving move. Disulfide bond formation in the ER, catalyzed by protein disulfide isomerase (PDI) and the oxidase ERO1, consumes reducing equivalents and generates hydrogen peroxide as a byproduct, linking ER stress to oxidative stress. A professional secretory cell such as a plasma cell can devote a huge fraction of its volume to ER and secrete on the order of 1,000–10,000 antibody molecules per second, which is exactly why those cells lean so heavily on a robust UPR.

Adaptive output: throttle inflow, expand capacity, clear the jam

When the UPR succeeds, three things happen in parallel. Translation drops via PERK-eIF2α, lowering the rate at which new clients enter the ER. Transcription of chaperones surges — BiP itself, the lectin chaperones calnexin and calreticulin, PDI, GRP94 — increasing folding capacity. And the ER physically grows: XBP1s and ATF6f turn on phospholipid biosynthesis, so the membrane network expands to give clients more room and more chaperone surface. Simultaneously, ERAD ramps up: terminally misfolded clients are recognized, retrotranslocated through the Sec61/Hrd1 channel, ubiquitinated, and fed to the cytosolic proteasome. In severe cases bulk ER fragments are removed by ER-phagy, a selective form of autophagy. Once the backlog clears and BiP is freed again, BiP re-binds the sensors and the alarm resets.

Maladaptive output: how the UPR commits a cell to death

If the inflow keeps exceeding capacity, the UPR pivots from repair to elimination. The decisive node is CHOP (also called GADD153/DDIT3), induced downstream of PERK-ATF4. CHOP tips the apoptotic balance: it represses the survival factor BCL-2 and induces the pro-death BH3 proteins BIM and PUMA, pushing BAX/BAK to permeabilize the mitochondrial outer membrane and release cytochrome c. CHOP also induces GADD34, a phosphatase regulatory subunit that dephosphorylates eIF2α and restarts translation — which, under unresolved stress, refills the ER and worsens the problem, a built-in deadline. Meanwhile IRE1, when chronically active, switches from cytoprotective XBP1 splicing to RIDD (regulated IRE1-dependent decay), shredding many ER-bound mRNAs and even some microRNAs, and it recruits TRAF2 to activate the JNK and ASK1 stress kinases. The shift from "splice XBP1" to "degrade everything and signal death" is one of the clearest examples of a single enzyme encoding a life-or-death timer.

Comparing the three UPR branches

Branch	Sensor type	Immediate action	Main output	Timescale	Pro-death role
PERK	eIF2α kinase	Phosphorylate eIF2α; halt translation	ATF4 → stress genes, later CHOP	Minutes	CHOP, GADD34 (translational restart)
IRE1	Kinase + RNase	Splice XBP1 mRNA (cytoplasmic)	XBP1s → chaperones, ERAD, ER growth	30 min – hours	RIDD mRNA decay; JNK/ASK1 activation
ATF6	Membrane-bound TF	Traffic to Golgi; S1P/S2P cleavage	ATF6f → chaperones, XBP1 transcription	Hours	Mostly adaptive; minor death role

Concrete examples across organisms and tissues

Yeast. In Saccharomyces cerevisiae the entire UPR is essentially a single branch: Ire1 splices HAC1 mRNA (the fungal counterpart of XBP1). This minimal system, dissected by Peter Walter and Kazutoshi Mori in the early 1990s, revealed cytoplasmic mRNA splicing and earned the field a 2014 Lasker Award. The PERK and ATF6 branches were added later in metazoan evolution, layering rapid translational control and a second transcriptional arm onto the ancestral IRE1 core.

Plasma cells. When a B cell becomes an antibody factory, XBP1s is required for the dramatic ER expansion that lets it pump out immunoglobulin. Knock out XBP1 and you cripple antibody secretion — the UPR here is a normal developmental program, not just an emergency response.

Pancreatic beta cells. Beta cells secrete insulin on demand and run their ER hot. Mutations in PERK cause Wolcott-Rallison syndrome (early infantile diabetes), and a point mutation in proinsulin that prevents folding causes the Akita mouse and human MIDY diabetes by overwhelming the UPR and killing beta cells via CHOP. This is the textbook case where chronic UPR failure becomes metabolic disease.

Photoreceptors and neurons. Misfolded rhodopsin in retinitis pigmentosa, mutant prion protein, and aggregation-prone proteins in Alzheimer's, Parkinson's, and ALS all engage the UPR. Sustained PERK signaling, by keeping translation off, starves neurons of synaptic proteins — and PERK inhibition has reversed neurodegeneration in prion-infected mice, a striking proof of principle.

UPR versus other proteostasis stress responses

Feature	Unfolded protein response (UPR)	Cytosolic heat-shock response (HSR)	Mitochondrial UPR (UPRmt)
Compartment guarded	ER / secretory pathway	Cytosol and nucleus	Mitochondrial matrix
Master sensor	BiP releases IRE1/PERK/ATF6	HSF1 (released from HSP70/90)	ATFS-1 (worms); ATF5 (mammals)
Key chaperones induced	BiP, calnexin, PDI, GRP94	HSP70, HSP90, small HSPs	mtHSP70, HSP60, ClpP
Translational control	Yes — PERK throttles globally	No global throttle	Limited
Membrane expansion	Yes — ER biogenesis	No	No
Death program	CHOP-driven apoptosis	Not a primary output	Mitophagy on failure

Evolutionary and clinical significance

The UPR is a roughly billion-year-old solution to a universal problem: the ER's folding capacity is finite and fluctuating demand will sometimes exceed it. The ancestral IRE1 branch already coupled sensing to a transcriptional fix; vertebrates added PERK for instant load-shedding and ATF6 for a second transcriptional channel, giving a graded, redundant system. That redundancy is why losing one branch is survivable but losing the response entirely is not.

Clinically, the UPR sits at the crossroads of metabolism, immunity, and cancer. In type 2 diabetes, lipid overload and chronic insulin demand drive beta-cell ER stress and CHOP-mediated loss. In cancer, tumor cores are hypoxic and nutrient-starved, and IRE1-XBP1s lets cancer cells survive and adapt — making IRE1 RNase inhibitors and PERK inhibitors active drug-development targets. Many viruses (flaviviruses, coronaviruses) co-opt the UPR to remodel the ER into replication factories. And in neurodegeneration, tuning the PERK branch — for instance with the small molecule ISRIB, which restores eIF2B activity downstream of eIF2α — can rescue protein synthesis and memory in animal models. Understanding the unfolded protein response is, increasingly, understanding the difference between a cell that adapts and one that dies.