How Apoptosis Works: Why Healthy Cells Choose to Die

Apoptosis is a form of cell death that the cell carries out on itself, using a built-in set of enzymes called caspases. It is how an embryo removes the tissue between developing fingers, how a tadpole loses its tail, and how the body discards worn-out or dangerous cells without setting off inflammation. A cell can be told to die by signals from outside or by damage it detects inside, and in both cases the result is a tidy dismantling rather than a burst. This article explains the machinery, the discoveries that revealed it, and what goes wrong when the process fails.

Death as a tool for building bodies

Most people think of cell death as a failure: an injury, an infection, a poison. In a growing embryo it is closer to sculpture. Material is added, and then selected parts are removed to leave the final shape, the way a stone carver works by taking away.

Fingers are the textbook case. Early in development, the hand is a paddle, with tissue filling the spaces between the future digits. A 2015 review in the International Journal of Developmental Biology, summarizing the field, describes how the tissue between the digits degenerates through several routes, among them caspase-dependent apoptosis, lysosome-mediated cell death, and a senescence process, and says that the process begins through interacting signaling pathways that also drive limb outgrowth. In 1996 studies in chicken limbs, the signaling molecules BMP-2 and BMP-4 induced apoptosis of interdigital cells in culture, and blocking BMP signaling left more webbing between the toes (later reviewers cautioned that the blocking method may have acted indirectly, by depleting BMP ligands).

Ducks and chickens

The comparison across species makes the point sharper. The same review reports that interdigital cell death is absent or reduced in the feet of swimming birds, with duck feet as an example, and in bat wings and aquatic mammal flippers. The difference in how much tissue dies is, in effect, much of the difference between a chicken's foot and a duck's. When interdigital cell death is repressed in humans, the result can be syndactyly, fused digits.

Tadpoles and turnover

Metamorphosis offers another example. As a tadpole's tail regresses, apoptosis is involved, and an apoptosis review in Toxicologic Pathology notes that inhibiting macrophages can disrupt tissue remodeling during tail regression. Throughout life, turnover continues. That same review states that around 10 billion cells are made each day just to balance those dying by apoptosis. A retrospective on the 1972 apoptosis paper adds a smaller example: human neutrophils, a type of white blood cell, live about 5.4 days, and roughly 20 percent of them die by apoptosis each day.

The takeaway from these cases is that the body is full of cells whose job includes dying on schedule. A tissue that never lost cells could not renew itself.

Giving the process a name

For many years, developmental biologists knew that cells died as a normal part of development. The phrase "programmed cell death" came into use in 1964, when Richard Lockshin and Carroll Williams used it in a paper, as recounted in H. Robert Horvitz's Nobel lecture. But there was no consensus on what it looked like in a cell.

That changed in 1972, when John Kerr, Andrew Wyllie, and Alastair Currie published "Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics" in the British Journal of Cancer. They proposed that the death they observed under the electron microscope was a distinct, controlled process, complementary to mitosis in regulating animal cell populations. A 50-year retrospective in the same journal records that they chose the word from the Greek for petals falling from a flower. It describes two stages: formation of apoptotic bodies, then their ingestion and breakdown by other cells. The features it lists include cell shrinkage, nuclear condensation, membrane blebbing, and fragmentation.

Their paper already noted links to disease. The retrospective says the authors recognized the potential to use programmed cell death for cancer therapy and observed that cell death goes on continuously inside malignant tumors. The 1972 paper has more than 20,000 citations.

Apoptosis versus necrosis

The contrast with the other major way for a cell to die, necrosis, helps define the term. Necrosis is accidental death from injury, and it is messy. Apoptosis is orderly.

Feature	Apoptosis	Necrosis
Cell size	Shrinks	Swells
Membrane	Stays intact; cell breaks into apoptotic bodies	Loses integrity; contents spill out
Inflammation	Essentially none	Usually triggers inflammation
Scope	Single cells or small clusters	Often large fields of cells
Trigger	Developmental or damage signals, tightly controlled	Injury, toxins, lack of oxygen

The first four rows follow a widely used apoptosis review by Elmore. The point that matters most for the body is the third row. A dying cell that keeps its contents wrapped up does not alarm the immune system, so millions can be removed daily with no one noticing.

The worm that cracked the code

The detailed mechanism came from a transparent roundworm about a millimeter long, Caenorhabditis elegans. In the 1970s, Sydney Brenner chose it as a model organism, and John Sulston and H. Robert Horvitz set out to follow every cell division from the fertilized egg onward. Horvitz describes in his Nobel lecture how Sulston could place a larva on a slide, watch individual cells under a microscope, and see them move, divide, and, in some cases, die.

What they found was precise. In addition to the 959 cells in the adult worm, 131 more cells are generated during development and are absent in the adult, because they undergo programmed cell death. It happens to the same cells every time. In the Horvitz lecture's words, 105 of these 131 deaths occur in the nervous system. Death, in other words, is a cell fate as predictable as becoming a muscle cell.

A genetic pathway for dying

If death is a fate, there should be genes for it, and that is what the Horvitz lab found. The first gene linked to the process, nuc-1, was discovered by Sulston and controls the breakdown of DNA in dying cells. Ed Hedgecock then found two genes, ced-1 and ced-2, needed for neighboring cells to engulf the corpses. Horvitz's student Hilary Ellis then carried out a screen that found a mutant with no cell corpses at all, and the gene was named ced-3. When ced-3 activity is reduced or eliminated, essentially all 131 cells that normally die instead survive. The finding showed that cell death requires a specific gene, which makes it an active biological process rather than just a failure.

Two more genes followed. ced-4 was also needed for the deaths. And ced-9 turned out to do the opposite: a gain-of-function mutation in ced-9 made cells that should die live, while loss of ced-9 function made cells that should live die. As the lecture puts it, ced-9 controls the decision between survival and programmed cell death.

The punchline: it is conserved

The worm genes had human counterparts. The CED-3 protein was found to be similar to a human enzyme, interleukin-1-beta-converting enzyme, a protease that was being studied in inflammation. CED-3 and that enzyme proved to be founding members of a family of cysteine proteases now called caspases. The lecture notes that studies by many laboratories had by 2003 identified at least 14 mammalian caspases, many of which function in programmed cell death. CED-4 resembled a mammalian protein called Apaf-1. And ced-9 resembled Bcl-2, a gene first known because its misexpression was tied to a human B-cell lymphoma. Horvitz recalls that the date he first learned of the similarity was February 12, 1992, from a fax sent by his graduate student Michael Hengartner.

Sydney Brenner, Horvitz, and Sulston shared the 2002 Nobel Prize in Physiology or Medicine for this work. A pathway worked out in a worm turned out to run, with modifications, in people.

Caspases: the enzymes that dismantle the cell

Caspases are proteases, enzymes that cut other proteins. The name comes from "cysteine-aspartic proteases": they use a cysteine in their active site and cut after an aspartate. They are made as inactive precursors, called procaspases, so that the cell can hold them ready without harm. Activating one means cleaving the precursor.

Elmore's review groups ten major caspases into three categories:

• Initiators: caspase-2, -8, -9, and -10, which start the process in response to signals.
• Effectors (executioners): caspase-3, -6, and -7, which do the demolition. Caspase-3 is regarded as the most important.
• Inflammatory: caspase-1, -4, and -5, which have a different role and are not part of the quiet apoptotic program.

The arrangement is a cascade. An initiator caspase, once activated, cleaves and activates executioners, which cut many cellular proteins, producing the shrinkage and fragmentation into apoptotic bodies that Kerr and colleagues described. A cascade has a useful property: it amplifies, so a small signal becomes a decisive one, and once the executioners are active the process is hard to reverse.

Think of a row of mousetraps set on a table, each loaded with a ping-pong ball. Trigger the first and the whole table erupts. The cell works this way by design, because a death decision that dithered halfway would be worse than either outcome.

Two routes in: the intrinsic and extrinsic pathways

Cells can be directed to apoptosis by two main routes, which converge on the same executioners.

The intrinsic pathway: mitochondria as the judge

The intrinsic, or mitochondrial, pathway responds to trouble inside the cell: severe DNA damage, stress, loss of survival signals. Mitochondria, which most people know as the cell's power plants, also play the role of arbiter of its death.

The decision depends on a family of proteins related to Bcl-2. A review of mitochondrial outer membrane permeabilization (often abbreviated MOMP) explains the logic. Some family members protect the cell, others promote death, and the outcome depends on how these signals are integrated. Proteins called BH3-only proteins act as sensors: they move to the outer mitochondrial membrane and either activate two effector proteins, Bax and Bak, or inhibit the protective relatives. Bax and Bak oligomerize and punch holes in the outer membrane. Cells lacking both resist this step. The authors call MOMP the point of no return.

Once the membrane is permeable, proteins leak from the space between the mitochondrial membranes into the cytosol. One of them is cytochrome c, a protein better known for its role in the cell's energy production (the wider chemistry is covered in the Krebs cycle article). In the cytosol, cytochrome c assembles with a protein called APAF1 into a caspase-activating complex called the apoptosome. The apoptosome activates caspase-9, which cleaves caspases-3 and -7.

A protein that works inside the powerhouse becomes a death signal when it escapes. That is a neat example of one molecule having two jobs, depending on where it sits.

The extrinsic pathway: orders from outside

The extrinsic pathway begins at the cell surface. When a ligand such as Fas ligand or tumor necrosis factor binds a death receptor, it leads to activation of initiator caspase-8. According to the same review, active caspase-8 propagates the signal by directly cleaving and thereby activating caspases-3 and -7.

The Fas receptor and its ligand also help keep immune cells in check, as the autoimmunity section below shows. For how immune cells identify what to eliminate, see the explainers on how the immune system recognizes threats and T-cell receptors.

Crosstalk through Bid

The two pathways are not sealed off from each other. Caspase-8 can cleave a BH3-only protein called Bid, and the truncated form, tBid, moves to mitochondria and induces MOMP. In some cell types, which the review calls type II cells, this loop is needed for death-receptor signaling to work because a caspase inhibitor, XIAP, holds back the direct route. Another mitochondrial protein, SMAC/Diablo, is released along with cytochrome c and blocks XIAP, lifting the brake.

Feature	Intrinsic pathway	Extrinsic pathway
Typical trigger	DNA damage, cell stress, loss of survival signals	Ligand binding a death receptor (for example Fas)
Where it starts	Mitochondria	Cell surface
Key regulators	Bcl-2 family (BH3-only proteins, Bax, Bak)	Death receptors, caspase-8
Initiator caspase	Caspase-9 (via the apoptosome)	Caspase-8
Converges on	Executioner caspases-3 and -7	Executioner caspases-3 and -7
Link between them	Receives the amplified signal from Bid	Caspase-8 cleaves Bid, which engages mitochondria

Cleaning up the corpse

A cell that has finished apoptosis is still there, in pieces, and a body cannot leave debris lying around. A neat feature of the process is that the cell advertises its own condition. Phosphatidylserine, a lipid that normally faces the inside of the cell membrane, moves to the outer surface during apoptosis, where it signals phagocytes to clear the cell. Exposure of this signal supports the noninflammatory uptake of apoptotic cells, as Elmore's review explains.

The worm genetics showed the same step from the other direction. Mutants in ced-1 and ced-2 start the death process, but the corpses are not engulfed and persist, which is how Hilary Ellis in Horvitz's laboratory could see them for her screen. Without cleanup the dead cells linger.

This is why apoptosis can occur in huge numbers without causing the swelling and redness of an injury. The contents never spill into the tissue, and neighbors swallow what is left. For a look at the wider system that deals with debris and infection, the immune system overview describes the cells that do the engulfing.

When the program fails

Because apoptosis removes cells that are damaged, dangerous, or no longer needed, failures run in two directions. Too little death leaves cells that should be gone. Too much removes cells the body needs. The best-characterized cases of too little are cancer and autoimmunity. (This section describes mechanisms, not treatment advice.)

Cancer: cells that will not die

A cell that has accumulated DNA damage is supposed to be stopped or removed. The protein p53 senses damage and can halt the cell cycle for repair or, if the damage cannot be fixed, trigger apoptosis. Elmore's review describes p53 as the most widely mutated gene in human tumors, mutated in over 50 percent of human cancers. A cell without functional p53 can survive damage that would otherwise have killed it.

Bcl-2 supplies another example. Its name stands for B-cell lymphoma 2, because overexpression of Bcl-2 occurs in certain B-cell lymphomas, and the review calls this one of the first strong lines of evidence that failed cell death contributes to cancer. A tumor cell can also gain protection by losing a pro-death protein such as Bax. For the wider picture of how tumors arise, see the article on what causes cancer.

The reverse logic has produced drugs. Venetoclax is a small molecule designed to mimic the action of BH3-only proteins by selectively blocking BCL-2. A review of its use in chronic lymphocytic leukemia reports that BCL-2 expression is increased in 95 percent of patients with that disease, and describes venetoclax as the first BH3 mimetic to receive FDA approval. The immune side of the story, including how treatments recruit immune cells against tumors, is covered in the immunotherapy explainer.

Autoimmunity: immune cells that survive

A different kind of failure appears in the immune system. The apoptosis review explains that too little apoptosis of auto-aggressive T cells leads to autoimmunity. Autoimmune lymphoproliferative syndrome (ALPS) is the clearest example: type 1A results from a mutation in the death domain of the Fas receptor, type 1B from a mutation in Fas ligand, and type 2 involves reduced caspase-10 activity. The shared theme is that cells that should be removed by the extrinsic pathway are not.

Dividing and dying in balance

Cell division and cell death are two halves of the same budget. A tissue holds a steady size when births match deaths, which is why the 1972 paper called apoptosis complementary to mitosis. The article on how cells divide covers the growth side of that ledger. And the way a stem cell's descendants commit to one fate or another, including the fate of dying, links to the questions in how stem cells know what to become and to classic work such as the Spemann-Mangold organizer experiment.

Common misunderstandings

"Apoptosis and cell death are the same thing." Apoptosis is one controlled form of cell death. Necrosis is another, and the table above shows how they differ. Researchers recognize additional forms, so "programmed" should not be read as "only apoptosis." In the developing digits, for example, apoptosis is one of several routes.

"Apoptosis is a sign something has gone wrong." Mostly it is the opposite. In the worm, 131 cells die as part of normal development, and in humans huge numbers die every day to keep tissues at the right size.

"Cells die passively, like a battery running flat." Apoptosis depends on active enzymes. The ced-3 discovery established that death needs a specific gene, which is why it is called a program.

"More apoptosis is always better for cancer." Cell death in tumors is a double-edged process. The 1972 retrospective notes that apoptosis can also promote tumors, for example by driving clonal selection of premalignant cells, meaning that killing most of a cell population can leave behind the survivors best at resisting death.

"Cell death only removes; it never builds." Dying cells may also send signals. The 2015 limb review raises the possibility that dying cells in the digit region signal to their neighbors, a possibility the authors flag as an open question. How some animals rebuild whole structures, and how cell turnover figures in that, is covered in the article on animal regeneration.

The takeaway

Apoptosis turns death into a regulated step, with its own genes, enzymes, and cleanup crew. It was worked out in a worm because that animal kills the same 131 cells every time, and the genes found there turned out to have counterparts in us: the caspases, the Bcl-2 family, and Apaf-1. Mitochondria hold the key to the internal route, death receptors drive the external one, and both converge on executioner caspases that cut the cell apart without alarm. The practical lesson of the cancer and autoimmunity cases is that a body must be able to remove cells as reliably as it makes them. Health is a balance of the two, not a one-way push toward more cells. For the surrounding biology, the Complete Biology Guide is the best starting map.

Frequently asked questions

Q: What is apoptosis in simple terms? A: Apoptosis is programmed cell death, a controlled process in which a cell dismantles itself using enzymes called caspases. The cell shrinks, its contents are packaged into small bodies, and neighboring cells or phagocytes clear them away. Because the contents are not released, there is essentially no inflammation.

Q: What is the difference between apoptosis and necrosis? A: Apoptosis is an orderly, regulated process in which the cell shrinks and the membrane stays intact. Necrosis is usually a response to injury, in which the cell swells and loses membrane integrity, spilling its contents and usually triggering inflammation. Apoptosis typically affects single cells; necrosis often affects fields of cells.

Q: What are caspases? A: Caspases are protease enzymes that cut specific proteins in the cell. They are made as inactive precursors and activated in a cascade. Initiator caspases such as caspase-8 and caspase-9 activate executioner caspases such as caspase-3 and caspase-7, which carry out the demolition.

Q: What do mitochondria have to do with cell death? A: In the intrinsic pathway, Bcl-2 family proteins such as Bax and Bak make the outer mitochondrial membrane permeable. Cytochrome c then leaks into the cytosol and forms the apoptosome with APAF1, which activates caspase-9. This step is considered the point of no return.

Q: Why do 131 cells die in the worm C. elegans? A: They are generated during normal development and then removed by programmed cell death, the same 131 every time, 105 of them in the nervous system. Studying mutants with no deaths led to the discovery of the ced-3 caspase gene, and later of ced-4 and ced-9, whose counterparts exist in humans.

Q: How does apoptosis shape the hand? A: Tissue between the developing digits dies. Signaling molecules such as BMP-2 and BMP-4 induced interdigital cell death in 1996 chick-limb studies. Interdigital death is reduced in webbed feet such as those of ducks. When the process is repressed in humans, fused digits can result.

Q: How is apoptosis connected to cancer? A: Tumors often escape apoptosis. The protein p53, which can trigger apoptosis in damaged cells, is mutated in over 50 percent of human cancers, and overexpression of Bcl-2 occurs in certain B-cell lymphomas. These are mechanisms researchers study; they are not a basis for personal medical decisions.

Sources

• Worms, Life, and Death (Nobel Lecture), H. Robert Horvitz, ChemBioChem 2003 (C. elegans lineage, 959 and 131 cells, ced-1, ced-2, ced-3, ced-4, ced-9, caspases, Bcl-2).
• Apoptosis: a review of programmed cell death, Elmore, Toxicologic Pathology (apoptosis versus necrosis, caspase classes, phosphatidylserine, p53, Bcl-2, ALPS).
• 50 years on and still very much alive: Apoptosis, British Journal of Cancer retrospective (Kerr, Wyllie, and Currie, 1972; the origin of the word).
• Mitochondrial outer membrane permeabilization review, PubMed Central (Bcl-2 family, cytochrome c, apoptosome, caspase-8, Bid, SMAC).
• Interdigital cell death review abstract, Lorda-Diez and colleagues, International Journal of Developmental Biology, 2015 (digit sculpting).
• Venetoclax in the Treatment of Chronic Lymphocytic Leukemia, PubMed Central (BCL-2 expression in CLL, BH3 mimetic).