Common fragile sites have long been identified by cytogeneticists as chromosomal regions prone to breakage upon replication stress. They are increasingly recognized to be preferential targets for oncogene-induced DNA damage in pre-neoplastic lesions and hotspots for chromosomal rearrangements in various cancers. Common fragile site instability was attributed to the fact that they contain sequences prone to form secondary structures that may impair replication fork movement, possibly leading to fork collapse resulting in DNA breaks. Here we show, in contrast to this view, that the fragility of FRA3B--the most active common fragile site in human lymphocytes--does not rely on fork slowing or stalling but on a paucity of initiation events. Indeed, in lymphoblastoid cells, but not in fibroblasts, initiation events are excluded from a FRA3B core extending approximately 700 kilobases, which forces forks coming from flanking regions to cover long distances in order to complete replication. We also show that origins of the flanking regions fire in mid-S phase, leaving the site incompletely replicated upon fork slowing. Notably, FRA3B instability is specific to cells showing this particular initiation pattern. The fact that both origin setting and replication timing are highly plastic in mammalian cells explains the tissue specificity of common fragile site instability we observed. Thus, we propose that common fragile sites correspond to the latest initiation-poor regions to complete replication in a given cell type. For historical reasons, common fragile sites have been essentially mapped in lymphocytes. Therefore, common fragile site contribution to chromosomal rearrangements in tumours should be reassessed after mapping fragile sites in the cell type from which each tumour originates.

Chromosomal instability (CIN) is a hallmark of tumour initiation and progression. Some genomic regions are particularly unstable under replication stress, notably common fragile sites (CFSs) whose rearrangements in tumour cells contribute to cancer development. Recent work has shown that the Fanconi anaemia (FANC) pathway plays a role in preventing defective chromosome segregation and CIN under conditions of replication stress. Strikingly, FANCD2 is recruited to regions hosting CFSs on metaphase chromosomes. To decipher the mechanisms protecting CFSs in G2/M, we searched for proteins that co-localize with FANCD2 on mitotic chromosomes, and identified XPF-ERCC1 and MUS81-EME1, two structure-specific endonucleases. We show that depletion of either ERCC1 or MUS81-EME1 affects accurate processing of replication intermediates or under-replicated DNA that persist at CFSs until mitosis. Depletion of these endonucleases also leads to an increase in the frequency of chromosome bridges during anaphase that, in turn, favours accumulation of DNA damage in the following G1 phase.