DNA interstrand crosslinks (ICLs) can be induced by multiple crosslinking agents and are a severe form of DNA damage that prevents strand separation during DNA replication and transcription. Failure to repair endogenous ICLs in S phase is thought to underlie the bone marrow failure syndrome Fanconi anemia, and up-regulation of ICL repair is one of the causes of tumor chemoresistance against widely used cancer drugs. In metazoans, a major pathway of ICL repair is coupled to DNA replication and requires the Fanconi anemia pathway. In most current models, collision of a single DNA replication fork with an ICL is sufficient to initiate repair. In contrast, we show that in Xenopus egg extracts, two DNA replication forks must converge on an ICL to trigger repair. When only one fork reaches the ICL, the replicative DNA helicase, CMG, fails to unload from the stalled fork, and repair is blocked. Arrival of a second fork, even when substantially delayed, rescues repair. We conclude that ICL repair requires a replication-induced X-shaped DNA structure surrounding the lesion, and we speculate how this requirement helps maintain genomic stability in S phase. Next, we focus on how psoralen-induced ICLs are repaired. Although psoralen-ICL repair still requires replication fork convergence, downstream repair events are completely different. Unlike cisplatin-ICLs, which are resolved via incisions in the parental strands with formation of double-strand breaks (DSBs), psoralen-ICLs are resolved via cleavage of a N-glycosyl bond that forms part of the ICL. Unlike cisplatin-ICL repair, psoralen-ICL repair does not require CMG helicase unloading or FANCI-FANCD2. When N-glycosyl bond cleavage is inhibited, the psoralen-ICL is repaired via FANCI-FANCD2-dependent incisions. Our work identifies a novel S phase ICL repair mechanism that is independent of the Fanconi anemia pathway and avoids DSBs. ; Medical Sciences