A newly synthesised glycoprotein enters the ER carrying the precursor N-glycan Glc₃Man₉GlcNAc₂. Glucosidase I and II trim two glucoses, leaving a monoglucosylated glycan that is the binding signal for the lectin chaperones calnexin and calreticulin. While bound, the protein gets one folding attempt.
Per cycle: fold succeeds with probability p
fold fails with probability (1 − p)
On success → glucosidase II removes the last glucose
→ chaperone releases the protein → ER exit (Golgi)
On failure → UGGT (UDP-glucose:glycoprotein
glucosyltransferase) senses the exposed
hydrophobic patches and re-adds ONE glucose
→ protein rebinds calnexin/calreticulin
→ cycle repeats (cycle count += 1)
If cycle count reaches N_max → ER mannosidase I trims a
mannose residue instead → glycan no longer resembles a
"foldable" substrate → routed to ERAD (retrotranslocation
+ proteasomal degradation)
Because each cycle is an independent trial, the number of cycles to a successful fold follows a geometric distribution: P(fold exactly at cycle k) = p·(1−p)^(k−1), with expected value 1/p. The probability a given protein survives to N_max cycles without folding — and is therefore degraded — is (1−p)^(N_max). Raise p and export dominates; raise the ERAD threshold and more slow-folding-but-eventually-correct proteins get a chance to finish; shrink chaperone slots below the influx rate and new proteins queue at the ER membrane, exactly like a limited-server queue.
This checkpoint is why mutations that only mildly destabilise a protein's fold (lowering p) can still cause severe congenital disorders of glycosylation or ER-storage disease — the protein is never "broken", it is simply degraded before ever finishing.