Mechanisms for ensuring replisome processivity
The replisome includes separate polymerase machineries for the leading and lagging strands, and the replicative helicase complex (Cdc45-MCM-GINS or CMG) which removes obstacles and unwinds DNA in front of the polymerases, as well as numerous auxiliary and regulatory factors (reviewed in (Pellegrini, 2023)). In unperturbed replication, the polymerase machineries are connected to the CMG to ensure that the replisome travels as a unit and to allow regulation of fork progression rate (Gambus et al., 2006; Jones et al., 2021; Simon et al., 2014).
When the replisome encounters an obstacle, either the CMG complex is uncoupled and continues along the template while one or both of the polymerases pause, or the whole replisome pauses. The former outcome results in the characteristic signature of replication stress (Figure 1B) - an accumulation of single stranded DNA ahead of the replication fork - observed with treatments that cause base lesions (eg: MMS) or impair DNA polymerase activity (eg: HU).  Persistent accumulation of ssDNA around the replication fork is recognised by the kinase Mec1 (the yeast homolog of ATR) via Ddc2 and the 9-1-1 complex, which then phosphorylates histone H2A at S129 in neighbouring chromatin, forming a focus of ɣH2A (Feng et al., 2006; Lopes et al., 2006; Namiki and Zou, 2006; Pardo et al., 2017; Puddu et al., 2011).
For base lesions such as 8-oxo-2’-deoxyguanosine (8-oxo-dG), the DNA polymerases in the replisome are replaced with trans-lesion synthesis enzymes Rad30 (Polη), Rev3/7 (Polζ) and/or Rev1, allowing DNA replication to continue past the lesion while (hopefully) inserting the correct base, with the lesion itself being left in place for NER or BER to cure post-replication (reviewed in (Powers and Washington, 2018; Zhao and Washington, 2017)) (Figure 1B). Alternatively, template switching can be used to bypass lesions. In this process, the stalled polymerase temporarily switches to using the other newly synthesised DNA strand as a template to bypass obstacles (reviewed in (Ripley et al., 2020)) (Figure 1B). Both strategies aim to restart efficient DNA synthesis on both strands at the replication fork, resulting in a functional replication fork that is no longer associated with the replicative helicase, an equivalent situation to a fork paused for example by a short pulse of HU (Petermann et al., 2010). In both cases, replication appears to resume as normal suggesting that these replication forks re-couple to the CMG helicase, probably facilitated by the slower progression of CMG that is uncoupled from the replicative polymerases (Graham et al., 2017; Petermann et al., 2010).
The CMG helicase can traverse almost all replisome obstacles, and can switch from single and double stranded binding modes when uncoupled and back to single stranded binding once re-coupled(Wasserman et al., 2019). However, programmed replication fork barriers can block CMG progression resulting in a stable fork arrest, personified by the RFB in S. cerevisiae ribosomal DNA and RTS1 at the S. pombemating-type locus (reviewed in (Labib and Hodgson, 2007)). This arrest is mediated not by a physical block to the replisome but by the Fork Protection Complex (FPC), which travels with the CMG complex (reviewed in (Shyian and Shore, 2021)). Similarly, the DNA replication checkpoint, which responds to global replication stress, can pause and stabilise replication forks without uncoupling also through the FPC (Noguchi et al., 2004; Pardo et al., 2017), though at least in human cells and likely in yeast the forks become progressively less stable with time (Petermann et al., 2010).
Only if bypass and restart mechanisms fail, and additionally the fork is not resolved by an oncoming replication fork from another origin, must further more potentially mutagenic options be explored. Once disconnected from the CMG helicase, a stalled fork gains the capability to reverse by re-annealing of the nascent leading and lagging strands to yield a 4-pronged structure akin to a Holliday Junction and capable of migration over significant distances, a remodelling process assisted by Rad5 in budding yeast (Blastyak et al., 2007; Toth et al., 2022; Unk et al., 2010) (Figure 1C). Reversed replication forks gain a free double stranded end, which can be resected and initiate homologous recombination (Cotta-Ramusino et al., 2005; Lemacon et al., 2017). Homologous recombination can restart the replication process, albeit in a different form known as Break-Induced Replication (BIR) in which the leading strand is copied as a migrating D-loop, then the lagging strand copied from the newly synthesised leading strand (Kramara et al., 2018; Liu and Malkova, 2022; Malkova and Ira, 2013) (Figure 1C). Therefore, whereas in canonical DNA replication both newly formed DNA helices contain one strand from the original DNA template (known as ’semi conservative replication’), in BIR both strands of the daughter DNA helix are newly synthesised (’conservative replication’). BIR uses a different DNA polymerase complement and relies on other helicases for processivity (notably Pif1) (Kramara et al., 2018; Liu and Malkova, 2022; Lydeard et al., 2010); movement of this complex is slow and error prone when forced to proceed over long distances, but in almost all instances this problem is avoided as the BIR fork is resolved by a replication fork coming from the other direction (Liu et al., 2021; Mayle et al., 2015). Because BIR is innately slow and error prone, replication forks are actively protected from resection so that transient fork pausing can be overcome with a non-recombinational system that likely does not involve a loss of the replisome structure or function (Brambati et al., 2018).  Single strand gaps are also converted to double strand breaks when encountered by the replisome and could enter a similar recombinational restart pathway (Vrtis et al., 2021). However, this must again be a last resort, and mammalian cells use PARP to guard against this outcome and it is speculated that yeast use the FPC, at least for Top1-mediated damage (Ray Chaudhuri et al., 2012; Westhorpe et al., 2020).
BIR forks have much more potential for introducing mutations than normal replication forks (Deem et al., 2011; Pardo and Aguilera, 2012; Sakofsky et al., 2014); the lack of the CMG complex makes these prone to further stalling and recombinational repair if unresolved (Liu et al., 2021; Smith et al., 2007). Furthermore, proper sister chromatid cohesion requires replisome factors including the FPC, meaning that recombinational repair during BIR is much more likely to result in non-allelic homologous recombination (van Schie and de Lange, 2021). Template switching based on microhomology is also more frequent, resulting in complex structural variations showing multiple switches with more or less homology at break points (Pardo and Aguilera, 2012; Sakofsky et al., 2015). These properties make the slow progression of BIR forks actively advantageous to the cell by providing the highest chance of resolution by a normal replication fork coming from the other direction and limiting the chances of mutation.
Therefore, stalled forks are protected from introducing further mutations at multiple levels. Firstly, these forks are protected from processing by the recombination machinery for as long as possible, secondly the recombination events are tightly restrained and thirdly the low processivity of BIR forks favours resolution by oncoming forks.