The advance of a DNA replication fork requires an unwinding of the parental
double helix. This in turn creates a positive superhelical stress, a (+)-D
elta Lk, that must be relaxed by topoisomerases for replication to proceed.
Surprisingly, partially replicated plasmids with a (+)-Delta Lk were not s
upercoiled nor were the replicated arms interwound in precatenanes. The ele
ctrophoretic mobility of these molecules indicated that they have no net wr
ithe. Instead, the (+)-Delta Lk is absorbed by a regression of the replicat
ion fork. As the parental DNA strands re-anneal, the resultant displaced da
ughter strands base pair to each other to form a four-way junction at the r
eplication fork, which is locally identical to a Holliday junction in recom
bination. We showed by restriction endonuclease digestion that the junction
can form at either the terminus or the origin of replication and we visual
ized the structure with scanning force microscopy. We discuss possible phys
iological implications of the junction for stalled replication in vivo.