Checkpoints and Irreversible Transitions

Irradiate a budding-yeast cell that lacks a gene called RAD9 and it does not pause. It keeps dividing for several generations and dies. Hold the same cell back with a drug for a few hours, and it repairs the damage.

So RAD9 was not a repair gene. It was a gene for waiting. That result, from Ted Weinert and Leland Hartwell in 1988, opened a question you can ask of every step in the cycle: what makes one event hold off until another is done, and once the cell moves on, what stops it going back?

The idea: a checkpoint makes one event wait for another

In 1989 Hartwell and Weinert gave these controls a name. A checkpoint is a control mechanism that enforces dependency between cell-cycle events. Think of a train whose doors report to the driver: the train can run perfectly well with the doors open, so a separate circuit must hold it at the platform until they close. A checkpoint is that circuit. Remove it and the cell can still divide, but it divides on time instead of on completion.

The rad9 experiment shows how to test for one. Weinert and Hartwell irradiated haploid yeast. Wild-type cells arrested or delayed in G2, the gap between DNA synthesis and mitosis. rad9 cells did not delay. Then the authors supplied the missing wait themselves, using a microtubule poison to block irradiated rad9 cells in G2 for several hours, and repair could occur efficiently. The defect was in waiting, not in repair. Repair itself belongs to that lesson.

That gives you a rule for predicting phenotypes. A cell that loses a checkpoint usually grows normally when nothing goes wrong. It fails only when the event it should wait for is slow or broken, and then it moves on anyway. The cost shows up as broken or missing chromosomes in the next generation.

Checkpoints are not only emergency brakes. In a 2018 study of cultured human cells, Saldivar and colleagues reported that the kinase ATR, activated by a protein called ETAA1 during ordinary replication, holds back a CDK1-driven switch on the transcription factor FOXM1 until S phase ends; when ATR was inhibited, cells entered mitosis early with under-replicated DNA. The cyclin-CDK engine these controls act on is in the previous lesson.

Damage and replication: kinases that hold CDKs back

Both DNA checkpoints work the same way: a sensor finds an abnormal DNA structure, a protein kinase is activated there, and the kinase’s targets keep a cyclin-CDK switched off. Phosphorylation as a switch is covered in that lesson.

Broken DNA. At a double-strand break, the MRN complex (Mre11, Rad50 and Nbs1) recruits the kinase ATM. In a purified system, Lee and Paull showed that MRN and DNA together activated ATM to phosphorylate p53 and the kinase Chk2. In mammalian cultured cells, Chk2 activation after damage depended on ATM. Two arms follow.

• The G1 arm runs through p53, the transcription factor taught in that lesson. p53 induces the gene for p21, a CDK inhibitor; in biochemical assays p21 blocks cyclin E-Cdk2 from phosphorylating Rb. In isogenic human colon cancer lines that differed in p21 status, Waldman, Kinzler and Vogelstein found that loss of p21 completely abolished the G1 arrest after DNA damage. That result is in those cells, not in every cell type.
• The G2 arm acts on Cdc25C, the phosphatase that removes inhibitory phosphates from Cdk1 so the cell can enter mitosis. Human Chk1 phosphorylates Cdc25C on serine 216, and Chk2 can do so too. A Cdc25C mutant that cannot be phosphorylated there no longer binds 14-3-3, a small adaptor protein; 14-3-3 binding is thought to keep Cdc25C away from Cdk1, so Cdk1 stays off.

Stalled replication. When a polymerase stalls but the helicase keeps unwinding, single-stranded DNA appears and is coated by RPA, the single-strand binding protein. Zou and Elledge found that in human cells RPA was needed both for ATR to reach damage sites and for ATR to activate Chk1, and that RPA stimulates ATR’s partner ATRIP to bind single strands in vitro. The checkpoint also acts at the fork itself. In replisomes rebuilt from purified budding-yeast proteins, the yeast checkpoint kinase Rad53 restrained excess unwinding by the CMG helicase after DNA synthesis was blocked, and only kinase-active Rad53 did so. That is a yeast reconstitution; how closely human ATR and Chk1 match it is still being worked out.

The spindle checkpoint: one kinetochore says wait

Before anaphase, every pair of sister chromatids must be attached to opposite spindle poles. The attachment point is the kinetochore, a protein structure on each chromatid’s centromere that grips spindle microtubules. How it grips is left to the mitosis lesson.

The genes came from yeast. In 1991 Rong Li and Andrew Murray, and in the same issue of Cell Hoyt, Totis and Roberts, isolated budding-yeast mutants that kept going when a drug destroyed their microtubules. These were named mad (mitotic arrest deficient) and bub (budding uninhibited by benzimidazole).

Where does the signal come from? Conly Rieder’s group used PtK1 cells, flat rat-kangaroo cells whose chromosomes can be seen one by one. They waited for a cell with a single chromosome still attached to only one pole, then destroyed its unattached kinetochore with a laser. Anaphase followed 20 plus or minus 2.5 minutes later, although that chromosome never moved to the equator. The cell was not checking alignment. It was listening to a wait signal from an unattached kinetochore.

The signal is a protein complex. Unattached kinetochores promote assembly of the mitotic checkpoint complex (MCC): Mad2, Cdc20, BubR1 and Bub3. Cdc20 is the activator that APC/C needs to pick its mitotic substrates. In the cryo-EM structure of human APC/C bound to the MCC, BubR1 sequences that mimic a substrate’s degron sit in Cdc20’s degron-binding sites, and BubR1 also blocks the E2 enzyme UbcH10. It is a key broken off in the lock. Block Mad2 and the wait is gone: injecting an anti-Mad2 antibody into PtK1 cells or primary human keratinocytes in prophase or prometaphase brought on premature anaphase. Errors that escape this checkpoint produce the aneuploidy discussed in that lesson.

The block is not absolute. Brito and Rieder found that PtK and human RPE1 cells held in nocodazole can slip out of mitosis while the checkpoint is still active, because cyclin B is slowly destroyed by a proteasome-dependent route.

One-way doors: destruction by APC/C and SCF

A phosphate can be removed in seconds. A destroyed protein can come back only by being made again. That is why the cell’s biggest transitions are triggered by proteolysis.

When the last kinetochore attaches, the MCC releases its grip and APC/C-Cdc20 ubiquitinates two targets: securin and cyclin B. APC/C is the ubiquitin ligase; the proteasome does the destroying (in HeLa cells, the proteasome inhibitor MG132 blocked securin breakdown). Securin binds and inactivates separase, a protease. Freed separase cuts Scc1, the cohesin subunit that holds sister chromatids together, and anaphase begins. Losing cyclin B is a major way Cdk1 is turned off at mitotic exit. Ubiquitin chemistry is in that lesson.

The same logic runs at the budding-yeast G1/S transition through a second ligase, SCF. Sic1 is a CDK inhibitor. Once phosphorylated, it is bound by Cdc4 assembled with Skp1 and ubiquitinated. When the SCF subunit Hrt1 was inactivated, Sic1 stayed stable and cells arrested at G1/S. How such destruction turns a gradual input into an all-or-none switch is taught in that lesson.

Destruction is not the only lock. In cultured Xenopus and human cells, Potapova and colleagues forced exit from mitosis with a Cdk1 inhibitor while keeping cyclin B intact; washing out the inhibitor let many cells return to mitosis. A later study found that after mitotic exit, inhibitory phosphorylation of Cdk1 can hold it inactive in G1 even when cyclin B is preserved.

A worked problem: timing securin destruction. The following data are made up for practice, in the style of the live-imaging work by Hagting and colleagues, who showed in HeLa cells that securin destruction begins at metaphase once the checkpoint is satisfied. Suppose HeLa cells express fluorescent securin, and fluorescence (as percent of its metaphase value) is measured at 0, 5, 10, 15 and 20 minutes after metaphase: 100, 71, 50, 35, 25. Assume fluorescence is proportional to securin and synthesis is negligible.

F(t) = F0 × e^(−kt) ; half-life = ln 2 / k

1. Rate constant. At 10 minutes F = 50, so ln(100/50) = 0.693 = k × 10, and k = 0.0693 per minute. The half-life is 0.693 / 0.0693 = 10 minutes. Check another point: 100 × e^(−0.0693 × 15) = 100 × e^(−1.04) = 35. It fits.
2. Time to 10 percent. t = ln 10 / k = 2.303 / 0.0693 = 33 minutes.
3. Anaphase. If sisters separate at about 20 minutes, 25 percent of securin is still there. In these made-up data, separase does not need every securin molecule gone.
4. Re-imposing the checkpoint. Suppose a spindle poison such as nocodazole is added at 10 minutes and fluorescence stays near 50 for the next 20 minutes. Destruction has stopped, consistent with Hagting and colleagues’ finding that re-imposing the checkpoint rapidly stops securin destruction (they used taxol): the spindle is disrupted and the checkpoint is holding APC/C-Cdc20 back again.
5. A D-box mutant. The destruction box (D-box) is a short securin sequence that APC/C recognizes. A D-box mutant securin would not fall at metaphase, and in HeLa cells such a mutant, which cannot be degraded in metaphase, inhibited sister chromatid separation. (Hagting and colleagues found it was degraded later, in anaphase, through a KEN box.)

Design problem

Knocking down protein Z with RNAi changes how long mitosis takes. Design experiments to decide whether Z is a spindle-checkpoint component or is needed for kinetochores to attach to microtubules.