We show that bacteriophage T4 has two alternative mechanisms to initiate DNA replication: one dependent on Escherichia coli RNA polymerase (RNA nucleotidyltransferase, EC 2.7.7.6), and one dependent on general recombination. Continued DNA synthesis under recombination-defective conditions was sensitive to rifampin, an inhibitor ofRNA polymerase. On the other hand, DNA synthesis accelerated in spite of the presence of rifampin if recombination occurred.Replication ofbacteriophage T4 DNA is initiated on linear DNA molecules at one or several preferred origins (1-9). Soon after the first initiation, many more replication forks are initiated, leading to a rapid acceleration ofDNA synthesis (10, 11). During this acceleration period, a complex network of DNA is formed (12)(13)(14)(15)(16)(17). This process requires recombination functions (9, 18). Most mutations that inhibit recombination not only prevent for, mation of this network but also arrest DNA synthesis prematurely (19)(20)(21)(22). This indicates that recombination and the continuation ofDNA replication are interdependent (for review see ref. 23). The underlying reasons have remained unknown because of observations seemingly inconsistent with simple explanations: the DNA arrest phenotype ofcertain recombinationdefective mutants is overcome by additional mutations in genes 33 and 55 (20-22, 24), which code for RNA polymerase accessory proteins (25)(26)(27).To explain these and other apparently contradictory results, we have proposed that phage T4 uses different modes to initiate DNA replication: initiation from specific origin sequences, which we define as "primary" initiation, and subsequent "secondary" initiation from recombinational intermediates (7,8,28).For several reasons (7,8,29,30) we suspected that host RNA polymerase (RNA nucleotidyltransferase, EC 2.7.7.6) is required for primary initiation although it has been shown that late wild-type DNA replication does not depend on RNA polymerase (31, 32). After the onset of DNA replication, gene 33 and 55 products associate with RNA polymerase (25-27) to effect the switch to late gene expression (33-35) by altering promoter recognition (36). If unmodified host RNA polymerase were required for primary origin initiation, the association with gene 33 and 55 products would shut offreinitiation from primary origins and make the alternative recombinational initiation indispensable for growth of wild-type T4. This hypothesis accounts for the observations mentioned above: premature arrest of DNA synthesis in recombination-defective mutants (46-47-) and restoration of DNA synthesis by additional mutations in genes 33 and 55. Although the rate of DNA synthesis under these conditions is similar to that ofwild-type T4, no branched concatemers are formed (20)(21)(22). Instead, the DNA replicates as linear molecules of unit length, which we suspected to require continued initiation from primary origins by RNA polymerase. Therefore, this replication should remain sensitive to inhibitors of RNA polymerase at late...