Protein Quality Control and Organelle Turnover

Starve a yeast cell of nitrogen and it starts to digest pieces of its own cytoplasm. A mutant that cannot do this loses viability faster than normal yeast during the starvation.

That mutant, apg1, came from a 1993 screen by Miki Tsukada and Yoshinori Ohsumi, and it points at an idea you will use throughout this lesson: a cell destroys parts of itself on purpose, and it needs to. Proteins misfold, and organelles wear out. This lesson follows a protein from the endoplasmic reticulum, where it is checked, to the two machines that destroy what fails, and then to the lysosome route that can take apart a whole organelle.

The idea: checkpoints for proteins

Think of a factory with inspectors on the line. A part that fails inspection is held back and sent round again. If it keeps failing, it is scrapped. If too many parts fail at once, the factory slows the line and hires more inspectors.

The endoplasmic reticulum (ER) runs this kind of line for secreted and membrane proteins (see protein sorting). Its lumen, the space enclosed by the ER membrane, is a folding compartment. A protein that cannot reach its native shape is kept in the ER and later destroyed instead of being sent on to the Golgi. When unfolded proteins pile up, the cell answers with a signalling program, the unfolded protein response (UPR), that slows protein synthesis and makes more folding machinery.

Two destruction routes back up the checkpoints. The proteasome, a barrel-shaped protease in the cytosol and nucleus, destroys single proteins that carry ubiquitin tags. The lysosome, an acidic compartment full of hydrolases, digests whatever is delivered to it, including membranes and whole organelles.

Folding in the ER and the unfolded protein response

A new chain enters the ER lumen through the translocon. BiP, an Hsp70-family chaperone of the ER lumen, binds its exposed hydrophobic stretches and helps it fold. Many proteins entering the ER also receive an N-linked glycan, a branched sugar tree attached to asparagine, with three glucose residues at its tip. That sugar becomes a tag that reports folding.

1. Glucosidases trim two of the glucoses, leaving one. Calnexin (membrane-bound) and calreticulin (soluble) are lectin chaperones, proteins that bind sugars, and they hold the monoglucosylated protein while it folds.
2. Glucosidase II removes the last glucose, and the protein is released.
3. UGGT, an enzyme that adds glucose back, inspects the released protein. If it is still non-native, UGGT re-glucosylates it and it can bind calnexin or calreticulin again. If it is folded, it can leave the ER.

The cycle has been traced in detail for particular clients. In HEK293-derived cells, for example, it was followed for the IGF-1 receptor, and the two human UGGT enzymes turned out to prefer different substrates. A protein that cycles too long is diverted to ER-associated degradation (ERAD). In the standard account, trimming of mannose residues from its glycan helps mark it.

ERAD moves a protein across a membrane in the wrong direction. The misfolded protein is carried from the lumen back to the cytosol (retrotranslocation), tagged with a ubiquitin chain by an E3 ligase in the ER membrane (Hrd1 in yeast), pulled out by the ATPase Cdc48 (p97, or VCP, in mammals) and handed to the proteasome. In 2016 Ryan Baldridge and Tom Rapoport put purified yeast Hrd1 and a membrane-anchored misfolded substrate into synthetic lipid vesicles. Hrd1 ubiquitinating itself triggered movement of the substrate across the membrane, and the results suggest that Hrd1 forms a ubiquitin-gated channel, with Cdc48 completing extraction. That is a reconstitution with yeast proteins; whether every ERAD substrate in every cell crosses this way is not settled.

When misfolded proteins accumulate, sensors in the ER membrane notice. Budding yeast has one, Ire1, reported in 1993 by Jeffery Cox, Caroline Shamu and Peter Walter, and in parallel by Kazutoshi Mori and colleagues. Ire1 spans the ER membrane and carries a kinase and an RNase on its cytosolic side. Activated, it cuts an intron out of the mRNA for the transcription factor Hac1, by a route outside the spliceosome, and the spliced mRNA makes Hac1. Mammals have three branches:

• IRE1 removes a 26-nucleotide intron from XBP1 mRNA (shown for mouse XBP1), which yields the active transcription factor XBP1s.
• PERK phosphorylates the translation factor eIF2α on serine 51. That slows most translation, lowering the load entering the ER, while allowing preferential translation of ATF4 (the kinase side is in stress signals).
• ATF6 travels to the Golgi, where Site-1 and Site-2 proteases cut it and release a transcription factor.

The branches overlap but can be separated. In mouse fibroblasts lacking IRE1α, ATF6 was still cleaved, and tunicamycin, a drug that blocks N-glycosylation, still induced BiP mRNA. One broken branch does not silence the response. If the stress is not relieved, the UPR can also drive cell death (see cell death).

Two shredders: proteasome and lysosome

The 26S proteasome is a hollow cylinder, the 20S core, capped by regulatory particles. Its proteolytic sites face inward, so only an unfolded chain threaded into the barrel is cut. The caps recognise ubiquitin chains, remove the ubiquitin for reuse, and use ATP to unfold the substrate and push it in. The tagging system itself is covered in ubiquitin and proteolysis, and the chemistry of protein breakdown in protein turnover. The proteasome shreds single proteins, not large aggregates or whole mitochondria.

The lysosome can. A V-ATPase pumps protons into its lumen and holds mammalian lysosomes at a strongly acidic pH (see organelles as environments), where its hydrolases work best. Material reaches it from outside by endocytosis and from inside by autophagy.

In macroautophagy, the main form, a cup-shaped membrane grows around a portion of cytoplasm and closes into an autophagosome, a vesicle bounded by two membranes. The autophagosome fuses with a lysosome, and the inner membrane and its contents are digested; the building blocks return to the cytosol. Starvation turns the process up through nutrient-sensing kinases such as mTOR (see mTOR and AMPK). Ohsumi’s yeast genes, later renamed ATG genes, encode much of the machinery, including a system that attaches the protein LC3 (Atg8 in yeast) to the lipid phosphatidylethanolamine. Lipid-linked LC3, called LC3-II, coats autophagosome membranes, and the copies on the inner membrane are destroyed in the lysosome.

Eating organelles: autophagy and mitophagy

Autophagy is not only bulk eating. In selective autophagy, receptors link a chosen cargo to the LC3 family. CCPG1 is a receptor for ER-phagy, the selective digestion of parts of the ER; it binds mammalian ATG8-family proteins and the autophagy protein FIP200 through distinct motifs. Mice with reduced Ccpg1 accumulate insoluble protein in the ER of pancreatic acinar cells.

Mitophagy removes mitochondria. In 2008 Derek Narendra, Atsushi Tanaka, Der-Fen Suen and Richard Youle showed in cultured cells that Parkin, a ubiquitin ligase, moves to mitochondria that have lost their membrane potential. In 2010 the same laboratory reported that the kinase PINK1 is kept low on healthy, polarized mitochondria by voltage-dependent proteolysis and builds up on depolarized ones, where it was necessary and sufficient for Parkin recruitment in the cells tested. Parkin then ubiquitinates outer-membrane proteins, and the mitochondrion is engulfed.

Two limits matter. These experiments used uncouplers such as CCCP, which depolarize every mitochondrion at once, and largely overexpressed Parkin. And in reporter mice lacking Pink1, basal mitophagy in the sampled tissues was mostly preserved, with pancreatic islets the exception. PINK1 and Parkin are a well-proven route for damaged mitochondria in culture, not the only route in an animal.

How do you measure autophagy? Counting LC3 dots, or LC3-II on a blot, fails on its own. More autophagosomes could mean more are made or fewer are destroyed, just as a fuller sink could mean the tap is on harder or the drain is blocked.

Worked problem: autophagic flux. In this made-up data set, cultured cells are immunoblotted for LC3-II, each band normalized to actin, with the control set to 1.0. Bafilomycin A1, which blocks the V-ATPase and so lysosomal degradation, is added for the last few hours. Assume it is saturating, so any extra LC3-II it reveals is LC3-II that would otherwise have been destroyed.

Condition	Without bafilomycin	With bafilomycin
Control	1.0	2.0
Starved	1.5	4.5
Drug D	3.0	3.2

flux = LC3-II (with bafilomycin) − LC3-II (without bafilomycin)

Step 1. Control: 2.0 − 1.0 = 1.0. Starved: 4.5 − 1.5 = 3.0. Drug D: 3.2 − 3.0 = 0.2.

Step 2. Starvation triples flux (3.0 / 1.0 = 3): more autophagosomes are made and destroyed.

Step 3. Drug D has the most LC3-II at steady state, 3.0, but a flux of only 0.2, one fifth of control. It blocks fusion or degradation, not induction. Read without bafilomycin, it would look like the strongest inducer.

Condition (made-up data set)

Control Starved Drug D Your own

add bafilomycin A1 (saturating) for the last few hours

LC3-II made in the window (= the + baf band)

2.0

relative units Share of newly made LC3-II destroyed (in this sketch)

50

percent

Keep this rule: a high LC3-II band on its own can mean more made or less destroyed, so only the LC3-II saved by bafilomycin, the flux, shows which. In this sketch the inhibitor is assumed saturating, the numbers are made up, and the ‘without’ band is simply the made amount × (1 − share destroyed); the 1.25× and 0.75× cut-offs for ‘up’ and ‘down’ are arbitrary, so the picture shows the logic, not a real blot.

Step 4. Confirm with cargo. p62, a receptor destroyed along with its cargo, should fall with starvation and rise with drug D. A tandem reporter, LC3 tagged with red and green fluorescent proteins, also helps: green fluorescence is quenched in the acidic lysosome, so normal flux gives red-only dots, while a block before fusion leaves dots that are both red and green.

Real data need the same care. In HeLa cells exposed to CCCP for 48 hours, electron micrographs showed 0.62 mitochondrial profiles per µm² of cytoplasm without YFP-Parkin and 0.074 with it.

reduction = (0.62 − 0.074) / 0.62 = 0.546 / 0.62 = 0.881, so 88.1% lower

That compares mean two-dimensional densities in two groups of cells at one time. It is not a rate. And in Atg5-null mouse fibroblasts, which cannot build autophagosomes, Parkin-marked mitochondria stayed.

Design problem

A mutant yeast strain accumulates a misfolded ER membrane protein. Is ERAD blocked, or is the UPR broken? Design experiments that tell the two apart, and write down what each model predicts before you run them.

One way to do it

If ERAD is blocked, the substrate is made normally but destroyed slowly, and the UPR should still work (it may even be switched on without added stress, because misfolded protein builds up). If the UPR is broken, the cell cannot raise its folding and degradation capacity, so the substrate may build up mainly under stress, and the alarm itself will be silent.

1. Cycloheximide chase. Stop synthesis and blot for the substrate at 0, 30, 60 and 90 minutes. A stable band in unstressed cells points to an ERAD block. Near-normal decay that slows only under stress points to the UPR.
2. Ubiquitination. Immunoprecipitate the substrate and blot for ubiquitin. No ubiquitin suggests the E3 step (Hrd1) is lost; ubiquitinated substrate stuck in membranes suggests a later step, such as extraction by Cdc48.
3. UPR reporter. Treat with tunicamycin and measure HAC1 mRNA splicing by RT-PCR, or a reporter driven by the UPR element. No splicing means the UPR is broken; normal splicing means it works.
4. Complementation and rescue. Transform the mutant with plasmids carrying candidate genes such as HRD1 or IRE1. The gene that restores normal degradation (or normal splicing) names the defect, and the rescue confirms that the mutation, not a second hit elsewhere, causes the phenotype.

If the chase, the ubiquitin blot and the rescue all point to HRD1, and splicing is normal, ERAD is blocked. If splicing fails and IRE1 rescues, the UPR is broken.