The core of the SV40 origin of replication (ori) contains a stretch of adenine (A) and thymine (T) residues. This region is very conserved among the papova viruses, and is known to be extremely sensitive to mutations. So far, mutations have been found to drastically reduce, and in most cases abolish, replication. The AT stretch has been shown to be the target for several host cellular proteins that belong to the replication machinery. We reasoned that, in this light, there might exist cellular DNA sequences that can substitute for the SV40 AT stretch. To study this possibility, we digested mammalian genomic DNA and inserted the fragments instead of the SV40 AT stretch in a plasmid carrying the SV40 ori core. The resulting pool was analyzed by a 'replication trap' in CosI cells. We present evidence that there are indeed several mammalian sequences that can substitute for the SV40 AT stretch. All of them are rich in adenines and thymines but, surprisingly, these sequences differ from the wild-type SV40 AT stretch to such extent that at first sight they would seem unlikely to replicate. This is all the more impressive if we consider that another AT-rich sequence from the yeast TRP1 gene, which also carries a similar variation, cannot substitute for the SV40 AT stretch.