Death Receptors, BCL-2 and Caspases

Cytochrome c, the small protein you met carrying electrons to complex IV, has a second job: once it leaves the mitochondrion, it is the order for the cell to take itself apart. A cell that can destroy itself within hours must be certain never to do so by mistake. This lesson asks what the killing enzyme is and what holds it off, and then how a signal from outside the cell or from inside it lets it go.

The idea: a cell can die by accident or by program

In the early 1960s John Kerr, a doctoral student in London, tied off blood vessels to part of a rat’s liver and studied how the starved cells died. Over the following years he found two deaths in one tissue. In the first, sheets of cells swelled, burst and spilled their contents. That is necrosis, death by injury, and the spill often causes inflammation. In the second, single cells scattered through working liver shrank, grew dense and broke into parcels wrapped in unbroken membrane. Neighbouring cells ate the parcels, and no inflammation followed.

Kerr published the second death in 1971 as “shrinkage necrosis”. In 1972, with Andrew Wyllie and Alastair Currie in Aberdeen, he gave it its own name: apoptosis, from a Greek word for the falling of leaves. The name made a claim: this death is something a cell does, as an ordinary part of a tissue’s life.

Why would a cell carry a program for its own death? An animal has to remove cells as well as make them. Development builds cells that are later taken away, and an adult tissue keeps its size only if losses match births. An infected cell, or an immune cell aimed at the body’s own tissues, is safer gone. So there are two groups of apoptosis-involving diseases: too little cell death, as in cancers and autoimmune disease, and too much, as in neurodegeneration.

The proof that such deaths are planned came from a worm. A hermaphrodite, the self-fertilising form of the nematode Caenorhabditis elegans, makes 1,090 body cells as it develops, and the same 131 of them die every time. Robert Horvitz’s laboratory hunted for mutants in which the deaths fail. It found two genes needed for nearly all 131 deaths, ced-3 and ced-4, then a brake on them, ced-9, and a gene that releases the brake, egl-1. Putting two mutations into one animal and seeing which effect won set the genes in order: EGL-1 blocks CED-9, CED-9 holds back CED-4, and CED-4 switches on CED-3. Each has human relatives, which the rest of this lesson follows.

The killer is an enzyme kept as a precursor

CED-3 is a protease, and its relatives in your cells are the caspases, short for “cysteine-dependent aspartate-specific proteases”. A caspase uses a cysteine in its active site to cut other proteins just after an aspartate.

Every caspase is made as an inactive precursor, a zymogen, like the digestive proteases you know. There are two kinds. Initiator caspases, such as caspase-8 and caspase-9, are single inactive chains that become active when a platform gathers several copies and they pair up. Executioner caspases, which are caspases 3, 6 and 7, are already paired but stay inactive until an initiator cuts them. The first step is a gathering, and every step after it is a cut.

Active executioners cut hundreds of the cell’s proteins, in what is called the demolition phase. One target is the inhibitor of a DNA-cutting enzyme. Freed, that enzyme cuts DNA between nucleosomes, the protein spools that DNA is wound on, and leaves a ladder of fragments about 200 base pairs apart. Other targets support the nucleus and the cytoskeleton, so the chromatin condenses, the cell rounds up and its surface bubbles into blebs.

Compare this with a kinase cascade. A phosphate that a kinase adds, a phosphatase removes. No enzyme rejoins a peptide bond that a caspase has cut. Inhibitor proteins can restrain an active caspase, but nothing undoes a cut. That is why the main safeguards act earlier, on whether initiators are gathered at all.

Killed from outside: a death receptor gathers caspase-8

The outside route, called the extrinsic pathway, starts at a death receptor: a membrane protein whose tail inside the cell carries a death domain, a small fold that binds other death domains. The best studied is Fas. A killer T cell that has found an infected cell can present Fas ligand, which comes as a trimer and pulls Fas receptors into a cluster.

The clustered tails bind an adaptor, FADD. Like Grb2, FADD has no enzyme activity and two binding ends. Its death domain holds Fas, and its death effector domain holds a matching domain on the precursor of caspase-8. Receptor, adaptor and caspase together make the death-inducing signaling complex, or DISC. In it, copies of caspase-8 sit side by side, pair up and become active, and then cut the executioners. The receptor needs no second messenger and no kinase.

Mice showed that animals use this route. Shigekazu Nagata’s laboratory in Osaka cloned human Fas in 1991, starting from an antibody that killed the cells it bound. In 1992 the laboratory found that mice of the lpr strain, kept since the 1970s for their swollen lymph nodes and lupus-like disease, have a defective Fas gene. A second strain with the same disease, gld, proved in 1994 to have a mutation in Fas ligand. Lymphocytes that should have been cleared after an immune response were staying alive.

The receptor for TNF, TNFR1, has a death domain too, but its usual output is NF-κB and survival, and it kills only when that arm fails. That is “the anti-death, death pathway” of the NF-κB lesson.

Killed from inside: a series of checks and balances

Many deaths start inside the cell, with damaged DNA acting through p53, as in the previous lesson, or with the loss of a survival signal. This inside route, the intrinsic pathway, is decided at the outer membrane of the mitochondrion by a family of proteins named for its first member, BCL-2.

BCL-2 was found at a chromosome break. In follicular lymphoma, a cancer of antibody-making B cells, a piece of chromosome 18 is joined to chromosome 14 next to an antibody gene. Yoshihide Tsujimoto in Carlo Croce’s laboratory, and then Stanley Korsmeyer’s laboratory, cloned the junction in 1984 and 1985 and found a gene beside it. In 1988 David Vaux, Suzanne Cory and Jerry Adams in Melbourne showed what the gene does. It does not make cells divide. It keeps them from dying.

In 1993 Korsmeyer’s laboratory found the opposition: BAX, which resembles BCL-2, binds to it, and speeds death. The family is a series of checks and balances with three classes.

• BAX and BAK are the effectors. Once activated, they open the outer mitochondrial membrane.
• BCL-2 and relatives such as BCL-xL are the guards. They hold BAX and BAK in check.
• BH3-only proteins, such as BAD and BID, are the sensors of stress. They share with the family only one short helix, the BH3 helix, which fits into a groove on a guard and occupies it.

Each sensor has its own control: BAD is switched off when a survival kinase such as Akt phosphorylates it. Korsmeyer read the family as a rheostat: guards and effectors are both present in a healthy cell, and their proportion sets how hard the cell must be pushed.

When BAX and BAK open the membrane, cytochrome c escapes. Xiaodong Wang’s laboratory found what it does next, in 1996, by making apoptosis run in a tube. They took the soluble contents of HeLa cells, added the nucleotide dATP, and saw caspase-3 switch on. Fractionation narrowed the requirement to one small protein. An antibody that removed it stopped the reaction, and adding the pure protein back restored it. It was cytochrome c. The next year the laboratory purified its partner, APAF-1, a relative of the worm’s CED-4. Cytochrome c and dATP make seven copies of APAF-1 assemble into a wheel, the apoptosome, which gathers caspase-9 as the DISC gathers caspase-8.

The two routes meet. Active caspase-8 cuts the sensor BID, and the cut piece goes to the mitochondrion and sets off BAX and BAK, as the laboratories of Junying Yuan and Xiaodong Wang reported in 1998.

Design problem

A cell must be able to destroy itself within hours, and must never do so by accident. Design the safeguards. Say in what form the killing enzyme is kept, what holds it off, and why you would require two separate events before it is released. Then say what a tumour gains by making too much BCL-2, and how a small molecule shaped like a BH3 helix would take that gain away.

When the brakes are too strong, or the machinery is missing

The lymphoma has too strong a brake. A B cell with the translocation makes too much BCL-2 and outlives its time, and further mutations can then make it a cancer. The drug venetoclax, approved in the United States in 2016 for a form of chronic lymphocytic leukaemia, is the small molecule of the design problem. It sits in the BH3 groove of BCL-2 and spares the related guard BCL-xL.

Now take the machinery away, in a knockout mouse, one in which a chosen gene has been deleted. Mice without APAF-1 die before or around birth, and their cells resist many kinds of death signal. Mice without caspase-9 have badly malformed brains with too many cells. So the inside route matters. Yet these embryos get through gastrulation, the early stage at which the body’s layers form, and go on to build organs. The idea that caspase mutants make it past gastrulation suggests another pathway.

One such pathway has a name. When TNFR1 is engaged and caspase-8 is blocked, as some viruses arrange, two kinases named RIPK1 and RIPK3 assemble, and RIPK3 phosphorylates a protein called MLKL. MLKL moves to the plasma membrane and breaks it. The cell swells and bursts like a necrotic cell, but on instruction, and the process is called necroptosis. Caspase-8 normally restrains it, in part by cutting RIPK1. A mouse embryo without caspase-8 dies, and it survives if RIPK3 is removed as well.