Cell Death and Clearance

In the hermaphrodite worm Caenorhabditis elegans, development makes 1,090 somatic cells and then kills 131 of them, leaving 959 in the adult body. The same cells die in every animal.

A death that happens to the same cell, at the same time, in every animal is hard to call an accident. It looks like a fate, as much as becoming a neuron or a muscle cell is a fate. This lesson asks three questions about it. How do you tell one kind of death from another? How was it shown that the cell’s own genes do the killing? And what happens to the body once the cell is dead? A lesson in Cell Signaling covers the protein cascade of apoptosis; here you will see the cell-level evidence, the assays and their traps, and the cleanup.

What a dying cell looks like, and why looks are not mechanism

In 1972 John Kerr, Andrew Wyllie and Alastair Currie described a kind of death they saw in many tissues and named it apoptosis. The cell shrinks. Its chromatin condenses against the nuclear envelope, the nucleus breaks into fragments, and the surface throws out blebs, bubbles of membrane that can pinch off as membrane-bound pieces. Through all of this the plasma membrane stays intact until late, and in a tissue a neighbouring cell usually eats the fragments first.

The older picture of death was necrosis: the cell and its organelles swell, and the plasma membrane ruptures and spills the contents. For decades the two were taught as a pair, one orderly and programmed, the other accidental.

That pairing has not survived. A cell can die with necrotic looks through dedicated machinery that genes or drugs can turn up or down, which is the working definition of regulated death. Three such deaths are now named by what they depend on:

• Necroptosis needs the kinase RIPK3 and its substrate MLKL. In the mammalian model, activated RIPK3 phosphorylates MLKL, and MLKL oligomers move to the plasma membrane.
• Pyroptosis needs gasdermins, proteins whose cut fragments form pores in the plasma membrane. It often, but not always, follows activation of inflammatory caspases.
• Ferroptosis is driven by severe lipid peroxidation that depends on iron. It is held off by GPX4, an enzyme that reduces lipid hydroperoxides, and in cultured cells it is blocked by iron chelators and lipophilic antioxidants.

Apoptosis itself has a defined engine. In the intrinsic route BAX and BAK permeabilize the mitochondrial outer membrane, cytochrome c escapes, and caspases (proteases that cut after aspartate) are activated in a cascade. The lesson on death receptors, BCL-2 and caspases teaches that cascade, and stress signals covers how p53 can push a damaged cell toward it.

The lesson for an experimenter is that appearance names an endpoint, not a cause. To say which death happened, you have to perturb the machinery and see the death change. That is the logic of experiments that distinguish, applied to a dying cell.

The worm shows death is a program

Watching was the first step. Using a light microscope on living transparent embryos, John Sulston, Einhard Schierenberg, John White and Nichol Thomson traced every division from the fertilized egg to the hatched larva. They reported in 1983 that the lineage and the fates it produced were highly invariant, and that 113 cells die during embryogenesis in the hermaphrodite. The rest of the 131 die after hatching. These numbers are for the standard hermaphrodite; males have a different lineage and a different count.

An invariant pattern says death is reproducible. It does not say a gene causes it. A cell might die because it is squeezed or starved at the same place every time. The test is genetic: find mutants in which those cells fail to die.

Hilary Ellis and H. Robert Horvitz did that. Their 1986 paper reported that wild-type function of two genes, ced-3 and ced-4 (ced for cell death abnormal), is needed for the deaths they examined. In the mutants, cells normally fated to die survived, and some differentiated into recognizable cell types. That last observation matters. Fewer corpses alone could mean corpses were being cleared too fast to count. Surviving, differentiated cells show the death itself did not happen.

A third gene acted the other way. In 1992 Michael Hengartner, Ronald Ellis and Horvitz reported that an activating mutation in ced-9 prevented the normal developmental deaths, while inactivating mutations made cells that normally live die. Genetic ordering placed a fourth gene, egl-1, upstream of ced-9, then ced-4, then ced-3. When the genes were cloned, CED-3 proved to be a caspase, CED-4 resembled human Apaf-1 and CED-9 resembled BCL-2. A worm screen had found the core of the human machinery. In a human cell-free system, Li and colleagues showed in 1997 that cytochrome c and dATP let Apaf-1 bind and activate caspase-9, which then cleaves and activates caspase-3.

One caution. The textbook cartoon has CED-9 simply holding CED-4 until EGL-1 releases it. A 2024 study of CED-9 mutants that cannot bind CED-4 found that the binding was needed for a pro-apoptotic role of CED-9 but not for its protective role, in their assays. The genetic phenotypes stand; the mechanism by which CED-9 protects is less settled than the cartoon suggests. For how the same logic shapes the nervous system, see naturally occurring death and death machinery in neurons.

Reading death assays without fooling yourself

Each common marker reports one event, and each event can occur in more than one kind of death.

• Annexin V is a calcium-dependent protein that binds phosphatidylserine (PS), a lipid normally kept on the inner leaflet of the plasma membrane (see lipids and membrane organization and topology). Annexin V binding means PS is reachable from outside. PS is also exposed in necrosis.
• Propidium iodide (PI) is a DNA dye that live cells exclude. PI entry means the plasma membrane is leaky, whatever made it so.
• TUNEL labels free DNA ends in fixed cells. In rat liver it labelled necrotic and autolytic cells as well as apoptotic ones.
• Cleaved caspase-3 on an immunoblot shows the protease was activated, not that the cell died or that caspase-3 killed it.
• LDH release into the medium reports rupture. NINJ1-null mouse macrophages died yet stayed swollen and released little LDH.

The way out is to combine markers and follow them in time. An apoptotic cell exposes PS first and stays PI-negative for a while. In culture, with no phagocyte to eat it, it then generally becomes leaky, which is called secondary necrosis.

A worked problem: what share of the death needs caspases?

Here is a made-up data set. A flow cytometer reads 10,000 Jurkat cells (a human T-cell line) per sample, after annexin V and PI staining. The quadrant counts are given in the order AnnV−PI−, AnnV+PI−, AnnV+PI+, AnnV−PI+.

• Untreated: 9,300 / 400 / 200 / 100
• Drug, 6 hours: 5,800 / 2,900 / 1,000 / 300
• Drug plus a pan-caspase inhibitor, 6 hours: 8,600 / 600 / 500 / 300

Assume the inhibitor does not affect the cells on its own, and that the untreated sample is the baseline for every condition.

Step 1. Check the totals. 9,300 + 400 + 200 + 100 = 10,000. 5,800 + 2,900 + 1,000 + 300 = 10,000. 8,600 + 600 + 500 + 300 = 10,000. Every cell is counted once.

Step 2. Total annexin-positive cells (both AnnV+ quadrants). Untreated: 400 + 200 = 600, or 6%. Drug: 2,900 + 1,000 = 3,900, or 39%. Drug plus inhibitor: 600 + 500 = 1,100, or 11%.

Step 3. Drug-induced excess. Drug alone: 39 − 6 = 33 percentage points. With the inhibitor: 11 − 6 = 5 points.

Step 4. The caspase-dependent share of the induced death.

caspase-dependent share = (33 − 5) / 33 = 28 / 33 = 0.85, about 85%

Step 5. Interpret. The AnnV+PI− cells are the clearest apoptotic signal: PS is out and the membrane still holds. The AnnV+PI+ cells could be late apoptotic cells in secondary necrosis or cells that died by necrosis from the start, and one time point cannot tell them apart. A time course would: apoptotic cells should pass through AnnV+PI− before becoming double positive. The remaining 5 points are caspase-independent death or incomplete inhibition, and the data cannot decide between them.

Find me, eat me: clearance

A tissue loses many cells, and in healthy tissue you rarely see a corpse. That is because clearance is fast and comes in steps. Clearance of apoptotic cells is often called efferocytosis.

Find me. In apoptotic human Jurkat cells, the effector caspases 3 and 7 cut the plasma membrane channel PANX1 and open it. ATP and UTP leave through it and attract monocytes. Knocking PANX1 down cut nucleotide release without slowing caspase activity, so the signal can be separated from the death.

Eat me. Caspases also cut and activate Xkr8, a scramblase, an enzyme that moves phospholipids between the two leaflets. In mouse cells lacking Xkr8, apoptotic cells failed to expose PS and were engulfed poorly. Valerie Fadok, Peter Henson and colleagues had reported in 1992 that macrophages recognize exposed PS on apoptotic lymphocytes.

Hold, then swallow. In resident mouse peritoneal macrophages, the receptor Tim4 binds PS directly and tethers the dying cell. MerTK, a receptor tyrosine kinase, binds PS through the bridging proteins Gas6 or Protein S and drives the engulfment. The engulfed cell is enclosed in a phagosome, which fuses with lysosomes and is digested, the same route described in the lesson on endocytosis.

Two cautions. Apoptosis and its clearance are not always immunologically silent; outcomes depend on context, as the lesson on TLRs and NF-κB helps explain. And phagocytes differ: microglia also clear dying cells and pruned synapses (see microglia and complement pruning).

Design problem

Mouse resident peritoneal macrophages are mixed with apoptotic cells. Macrophages lacking Tim4 bind fewer apoptotic cells. Macrophages lacking MerTK bind them normally but take few inside. Design an assay that separates binding from engulfment, and say what each receptor does.