How does ATM promote reversed replication fork processing during interstrand cross-link repair?

ATM drives resection rather than only protecting forks

The question is not whether ATM senses replication stress, but what it does to a reversed fork once that structure exists. Earlier work had linked ATM to constraining nuclease activity so reversed forks are not degraded. This study reports the opposite direction of control in replication-coupled interstrand cross-link (ICL) repair: ATM inhibition stabilised reversed fork intermediates, and ATM activity was needed for their nucleolytic processing.

The evidence comes from two systems used together: cell-free Xenopus egg extracts replicating plasmids that carry a cisplatin ICL, and human RPE-1 cells. In the extract, replication intermediates are resolved on native agarose gels, so the appearance and disappearance of figure-8 structures, reversed forks, sigma products and fully replicated supercoiled products can be followed over time. That design makes it possible to ask which enzyme changes the fate of a defined repair intermediate, rather than only whether cells survive a drug.

A CRISPR screen points to PP2A in FA pathway repair

The entry point was a genome-wide CRISPR dropout screen in human RPE-1 cells challenged repeatedly with psoralen plus UVA (PUVA), which creates psoralen ICLs. Screens were run in FANCD2-proficient and isogenic FANCD2-deficient cells, so genes acting in the same pathway as FANCD2 could be separated from genes acting in alternative repair routes. Sixteen known FANC genes scored as ICL resistance genes in FANCD2-proficient cells, and except for FANCJ none of them scored in FANCD2-deficient cells; NEIL3 and FAN1 scored in both.

PTPA, which encodes a factor that promotes PP2A phosphatase holoenzyme formation, emerged specifically in FANCD2-proficient cells. PTPA knockout cell pools were more sensitive to cisplatin, confirming a role in tolerance of cisplatin-induced damage. The screen therefore nominated PP2A activity as a contributor to FA pathway-dependent ICL repair, and the extract experiments were used to find out what PP2A was restraining.

Lifting PP2A inhibition accelerates fork destruction

Adding okadaic acid, a pan-phosphoprotein phosphatase inhibitor, five minutes after replication initiation did not appreciably affect replication of an undamaged plasmid or of a plasmid carrying an abasic-site ICL, which is repaired mainly by the NEIL3 pathway. It did disrupt repair of the cisplatin-ICL plasmid handled by the FA pathway. Reversed fork intermediates disappeared rapidly, fully replicated supercoiled products accumulated less, and faster-migrating cleavage species and well products appeared. The PP2A-selective inhibitor cytostatin produced the same destabilisation, and excess recombinant human PP2A-AC subcomplex largely suppressed the okadaic acid effects.

The authors checked whether this simply reflected failed FANCD2 activation, since PP2A-dependent FANCD2 dephosphorylation has been reported to support FANCD2 ubiquitylation. Okadaic acid did not inhibit accumulation of monoubiquitylated FANCD2 during cisplatin-ICL plasmid replication, and FANCD2 depletion generally stabilises reversed forks rather than destroying them. The okadaic acid phenotype is therefore distinct from loss of FANCD2 monoubiquitylation, which is why the paper turns to kinase signalling.

ATM inhibition traps reversed forks and blocks repair

Among the kinase inhibitors tested, pan-CDK inhibition with flavopiridol had no noticeable effect on formation or resolution of cisplatin-ICL reversed fork intermediates. DNA-PKcs inhibition caused a subtle accumulation, and ATR inhibition a modest one. ATM inhibition with KU-55933 produced a pronounced accumulation of reversed fork intermediates; AZD0156 and caffeine behaved similarly, and combining ATM and ATR inhibitors stabilised forks more strongly than either alone, suggesting distinct mechanisms.

ATM inhibition also suppressed the fork destabilisation caused by okadaic acid and delayed formation of fork cleavage products. Both ATM inhibition and PP2A inhibition reduced error-free repair products, so efficient ICL repair requires both activities. Consistent with PP2A opposing ATM, okadaic acid increased phosphorylation of ATM at S1981 and of the ATM targets MRE11 and CHK2, and ATM inhibitor attenuated these signals. ATM inhibition further increased nascent leading-strand stall products and reduced full-length extension, and stabilised nascent lagging strands.

DNA2 and EXO1 carry out the resection ATM promotes

The resection itself depends on known nucleases. ATM enhances chromatin recruitment of DNA2, and DNA2 acts alongside EXO1 to resect reversed forks formed during ICL repair. The paper's model is that the regressed arm of a reversed fork behaves like a one-ended double-strand break, activating ATM to promote resection of that arm, with MRN, DNA2, WRN and EXO1 contributing.

This is a mechanistic claim built from perturbation experiments in the extract and in human cells, not a direct measurement of a single molecule being cut. The distinction matters for readers: the evidence establishes that ATM activity is required for the normal timing and extent of reversed fork processing, and identifies DNA2 and EXO1 as the nucleases involved, while the precise division of labour between them at each intermediate is inferred from the combined perturbations.

PP2A sets how much resection is tolerated

The most decision-relevant comparison in the paper is between too little and too much ATM signalling. Inhibiting ATM stabilises reversed forks and reduces error-free repair. Inhibiting PP2A hyperactivates ATM, accelerates reversed fork processing and accumulates aberrant end-joining products. Both directions compromise repair, which is why the authors describe a phospho-regulatory circuit rather than a simple linear pathway.

That framing changes how the result should be read. ATM is not simply pro-repair here; its value lies in being correctly timed and bounded. The well products seen after okadaic acid are interpreted as possible end-joining of double-strand-break-containing intermediates, and the reappearance of fast figure-8 species as premature reversed fork resolution before unhooking incisions. These are interpretations of gel species, not sequenced repair outcomes, so the causal link between accelerated resection and aberrant joining remains a model supported by the perturbation pattern.

Scope of the ATM-PP2A circuit in ICL repair

The findings apply to replication-coupled repair of cisplatin and psoralen ICLs handled by the FA pathway, studied in Xenopus egg extracts and in RPE-1 and derivative human cell lines. The extract work used okadaic acid added after replication initiation, and the CRISPR screens were performed once, with the screen-based claims resting on that single genome-wide pass plus targeted validation. The paper does not establish that fork reversal is strictly required for ICL repair, and it notes that the enzyme mediating reversal in the FA pathway has not been identified.

Within those limits, the priority for readers is the direction of ATM's action on reversed forks in this repair context. Where earlier work emphasised ATM restraining nucleases to protect reversed forks, this study shows ATM promoting their resection during FA pathway ICL repair, with PP2A holding that activity in check. The practical threshold implied by the data is a window: ATM activity below the level needed for processing, or above the level PP2A can restrain, both reduce error-free repair products in the cisplatin-ICL plasmid assay.

Sources used in this answer