27. DNA Lesions IV: Fork Reversal

Fork reversal

If a DNA lesion isn't addressed by TLS or repriming in the first 20-30 minutes following RPA-ssDNA detection, fork reversal will begin to unfold. In this process, cell proteins rearrange the replication fork into a shape that makes the problematic lesion easier to repair. More precisely, the lesion is removed from the crowded active replisome environment with its clamps, polymerases, etc. and returns it back to the context of dsDNA, where it will be much easier to access and then to fix.

Let's start with an analogy for fork reversal. Very simply, it's like zipping a zipper. Normally, in genome replication, the CMG helicase flies down dsDNA, separating it into single strands (i.e., unzipping a zipper). The actual handled zipper is the helicase, separating one entity (the closed zipper, or the double helix) into two (the open zipper, or the two template single strands). Given that the analogy for helicase unwinding is the unzipping of a zipper, the analogy for fork reversal would be the re-zipping of that zipper.

Let's pause and reset the scene before we jump whole hog into fork reversal (also called "fork regression"). We pick up the story after a lesion on the leading strand template has reached DNA polymerase epsilon, causing it to stall. Recall that this is often because the lesioned nucleotide can no longer fit in the enzyme's active site. The polymerase remains arrested in this position without fork reversal starting for about 20-30 minutes after the advent of ATR-CHK1 signaling.

But at that point, several highly specialized proteins are going to collaborate to load the an amazing protein called RAD51 at the ssDNA/dsDNA junction (i.e., near DNA polymerase epsilon where the nascent leading strand ends). The first critical protein to arrive will be BRCA1. It's possible you've heard of this protein's gene. It is one of two well-known breast cancer predisposition, or risk, genes (the other being BRCA2, which I'll introduce momentarily). Certain mutations in BRCA1 and BRCA2 increase the risk of breast and ovarian cancer significantly. But the proteins' normal roles in the cell include involvement in replication fork reversal.

BRCA1 arrives at the site of the stall first and effectively "decides" whether it is an appropriate site for RAD51 activity. This decision step is important because RAD51, as we'll see, is a powerful protein if employed in the right setting. But it can cause significant genome damage if it acts in an inappropriate setting. BRCA1 makes sure that doesn't occur.

How does BRCA1 evaluate the site? It integrates several criteria, most of which are communicated by chemical modifications to the protein. The three main questions it "asks" and the signals that communicate the answers are as follows:

First, BRCA1 asks if the problem real and persistent. DNA copying often pauses, but most lesions are fixed quickly by TLS or repriming. BRCA1 knows the stall is persistent based on the status of the signaling cascade: RPA-ssDNA formation leads to ATR phosphorylation, which leads to CHK1 phosphorylation, which leads to BRCA1 phosphorylation. The final phophorylation of BRCA1 "tells" it that the signaling cascade is in effect.

Second, BRCA1 asks if the timing is safe--that is, if the cell is in S/G2 when it's replicating DNA. This timing is essential because RAD51 works by comparing damaged DNA to an identical backup copy, and a nearby copy will only be present during S/G2. BRCA1 knows that it is in S/G2 because CDK2 (aided by Cyclins E/A) phosphorylates it at a specific amino acid: a different amino acid than the target of CHK1.

Third, BRCA1 asks if the DNA structure is organized enough for fork reversal (or do we risk making the situation worse)?"If the answer is "yes" to the first question, BRCA1 initiates RAD51 recruitment and loading. If not, RAD51 is kept out and a safer option is used. BRCA1 assesses DNA architecture indirectly based on the proteins present around the DNA and how they are chemically modified, rather than by directly "feeling" the shape of the DNA.

Assuming BRCA1's criteria are met, the next step unfolds: the protein PALB2 arrives and attaches to BRCA1. Soon it will connect BRCA1 and BRCA2.

When BRCA2 arrives, it is in a complex with multiple RAD51 proteins. BRCA2 has several RAD51 binding sites. This BRCA2 loaded with RAD51s physically connects to BRCA1 through the PALB2 bridge and then performs its main task: loading its cluster of RAD51s at the ssDNA/dsDNA junction.

In fork reversal, RAD51 has different roles at two different points in time. Here we're focusing on the first roles it adopts at a stalled replication fork--that is, prior to fork reversal. Yes, it coats and protects ssDNA located near the lesion. But it is also going to license the next critical protein, the 3'-5' exonuclease MRE11, to do its job, which is to chew back the nascent leading strand about 10-50 nucleotides. This is referred to as controlled strand "resection." And it is the RAD51s that are doing the controlling. If RAD51 wasn't present, MRE11's resection would be thousands of bases long and would severely damage the replication fork and the genome.

MRE11 resection of the nascent leading strand also releases DNA polymerase epsilon from the ssDNA-dsDNA junction. It remains near the replication fork, though, since it is tethered to the CMG helicase. Think of this releasing of the polymerase as, effectively, disengaging the fork from an unproductive configuration--breaking a log jam. The stall has persisted and the lesion has not been fixed. Something must change. Chewing back the DNA and releasing the polymerase allows the next step to take place.

And the next step is fork reversal. Motor proteins--proteins with the ability to move along DNA and to move DNA--are going to perform something called a "branch migration." As the name suggests, in branch migration a branch point in DNA--like the branch point in the normal 3-way replication fork--is going to be moved along the DNA and repositioned.

In this case, three translocases--SMARCAL1, ZRANB3 and HLTF--are literally going to push the stalled fork backwards, in the direction away from the CMG helicase, thereby rezipping the zipper. This migrates the junction, causing the two template strands that were originally separated by the helicase to re-anneal based on their complementarity.

When the two template strands re-anneal, it causes their respective nascent strands to be displaced from these template strands. The nascent strands are complementary to each other, since they were built off of complementary template strands. They have a proclivity to anneal to each other, but in this context they'll need the help of the three previously mentioned translocases to do so. Once they do, the result is a reversed, or regressed, fork.

Let's pause and consider the fork at this point. When the branch migrates and the two nascent strands self-anneal, they transform the three-way replication fork (the double helix being one arm and the two separated strands base-paired to the synthesized complementary strands being the other two) into a four-way fork called a "chicken foot structure." The fourth arm of the chicken-foot is made up of the two nascent strands annealed to each other. This regressed arm is fragile and so will be protected and stabilized in a correctly base-paired alignment by RAD51s lined up along its length.

Fork reversal is difficult to picture... without a picture! So take a look at the figure below.

Fork reversal. In fork reversal, motor proteins remodel the replication fork. They push the normal 3-way replication fork backward (i.e., to the right in the figure) which requires that the two strands being synthesized anneal to each other to form the fourth leg of the 4-way junction or "chicken foot structure." The reversed fork places the lesion back in the context of dsDNA where it is more easily repaired by the cell.

When a fork reverses, the leading template strand containing the lesion gets pushed back into a double-stranded context. There, it will be repaired. The next section discusses how one kind of large legion called a DNA-protein crosslink (DPC) would be repaired during a fork reversal.

In addition to the lesion, the nascent strands of the regressed arm also need to be repaired prior to fork restart. There are two problems. First, because of MRE11s resection, the leading nascent strand is generally shorter than the lagging nascent strand. The cell would like the two strands of the regressed arm to be roughly symmetrical, so of roughly equal length. Second, give how the nascent lagging strand is synthesized, it likely has some ssDNA gaps between Okazaki fragments. These gaps are best filled prior to fork restart.

With RAD51 stabilizing the paired nascent strands, the cell calls on DNA polymerase delta to make the repairs. This kind of synthesis is referred to as "repair synthesis." Unlike normal synthesis, repair synthesis requires RAD51 stabilization of the two DNA strands. Repair synthesis isn't--and doesn't need to be--high fidelity. Errors can be fixed later. The main goal is to make the DNA strands physically intact enough for fork restart. It makes sense that DNA polymerase delta is used in repair synthesis since it is adept at handling single-stranded gaps.

Once both the repairs have been made, fork restart is initiated. At the center of fork restart is a forward branch-migration move (now we're again unzipping the zipper) driven by the helicase protein RECQ1. So while the roles of SMARCAL1, ZRANB3 and HLTF are to reverse a fork (zip the template strands and create the regressed arm), RECQ1 restores the normal three-way replication fork junction so that genome synthesis can resume.

Repairing a DPC

One general category of large lesion is DNA-protein crosslinks (DPCs). These are proteins or, in some case, very large protein complexes that have become permanently "glued to" a nucleotide via a chemical bond. Most DPC get past the CMG helicase but stall at DNA polymerase epsilon (although some very large ones do stall at the helicase). Let's pick up this story post-fork reversal, after the DPC has been positioned back in dsDNA.

Proteases are proteins that chew up and destroy other proteins. When a DPC is present in dsDNA near a replication fork following fork reversal, the protease SPRTN appears. SPRTN is only active in S phase during replication. This talented protein: (1) detects stalled replication forks using a complex set of clues... not just RPA-ssDNA, (2) chews the protein that's embedded in dsDNA into many tiny peptides, and (3) stays inactive elsewhere to avoid harming the cell.

SPRTN chews up almost the entire attached protein into small peptides that are then broken down and recycled. But it can't remove the small peptide or single amino acid attached to the nucleotide. This remaining piece of the protein is refered to as a Peptide-DNA adduct. It will be removed by standard nucleotide excision repair (NER) once SPRTN is finished digesting the bulk of the protein. This often happens immediately after SPRTN digestion but some smaller adducts might pass through the polymerase to be repaired by NER later. So, DPC's are repaired byia a two-step hybrid repair system that combines SPRTN proteolysis (i.e., protein destruction) with NER.

Finally, I'll note that this SPRTN-based DPC repair pathway isn't a "nice to have." It's critical in humans. Without it, we literally wouldn't survive. In fact, mutations in SPRTN are known to cause Ruijs-Aalfs syndrome, a human disease that results in early liver cancer, chromosomal instability, premature aging and a very short lifespan.

A bit more about RAD51

In addition to its role in fork reversal, when DNA in the genome is broken or severely damaged, the cell sometimes repairs it using a process called homologous recombination (HR). Instead of guessing how to fix the damage, the cell uses an undamaged copy of the same DNA sequence—usually the sister chromatid—as a template. This makes homologous recombination the most accurate of the DNA repair pathways.

Let me use an analogy to help explain homologous recombination. I liken it to a scenario in which you have two photocopies of the same book page, but one of them has missing text. You want to fill in that text, so you place the good sheet behind the ripped sheet and use the information from the good sheet to fill in the incomplete sheet.

In HR, a chromosome with missing information is repaired using information from its sister chromatid. Thus, HR occurs mainly during S phase and G2, when sister chromatids are present and held in close proximity, allowing damaged DNA to be repaired using an identical template. It will be clear why this is important in a moment.

At the center of homologous recombination is RAD51, which has an extraordinary ability: it can help one DNA strand find and align with another DNA sequence that is identical. RAD51 coats a single DNA strand from the damaged chromatid and allows it to invade and “test” double-stranded DNA from the other chromatid for sequence similarity. Pairing is stabilized only when the bases match correctly; mismatched sequences fall apart. Once the complete and the damaged chromatids are aligned, aDNA polymerase fills in the missing information on the damaged chromatid.

This ability to recognize and stabilize matching DNA sequences is unusual—most proteins bind DNA based on shape or charge, not genetic identity. Importantly, the same homology-sensing property that allows RAD51 to guide accurate repair during homologous recombination also underlies its role at stalled replication forks. There, RAD51 helps preserve correct strand alignment and prevents the cell from making dangerous rearrangements while DNA replication is temporarily paused.