4-1BBL(-/-) mice exhibit normal primary CD8 T cell responses to influenza virus, but show decreased CD8 T cell numbers late in the primary response as well as decreased secondary responses. In contrast, CD28(-/-) mice are defective in initial CD8 T cell expansion. Using agonistic anti-4-1BB Ab to replace the CD28 or 4-1BB signal, we examined the timing of the required signals for CD28 vs 4-1BB costimulation. A single dose of agonistic anti-4-1BB Ab added only during priming restores the secondary CD8 T cell response in CD28(-/-) mice. Once the T cell numbers in the primary response reach a minimum threshold, a full secondary response is achieved even in the absence of CD28. In contrast, anti-4-1BB added during priming fails to correct the defective secondary response in 4-1BBL(-/-) mice, whereas addition of anti-4-1BB during challenge fully restores this response. Thus, there is a switch in costimulatory requirement from CD28 to 4-1BB during primary vs recall responses. Adoptive transfer studies show that T cells primed in 4-1BBL(-/-) or wild-type mice are equally capable of re-expansion when rechallenged in wild-type mice. These studies rule out a model in which signals delivered through 4-1BB during priming program the T cells to give a full recall response and suggest that 4-1BB-4-1BBL interactions take place at later stages in the immune response. The results indicate that anti-4-1BB or 4-1BBL therapy will be most effective during the boost phase of a prime-boost vaccination strategy.