Herpes simplex virus type 1 establishes latent infections in sensory neurons. During latency only one locus, the latency-associated transcript (LAT), is abundantly transcribed. Several lines of evidence suggest that this locus is required for the efficient reactivation from latency in experimental models. However, it is not yet clear whether this is a direct effect on the reactivation process per se or, as we have suggested, an indirect effect resulting from a decreased efficiency of establishment of latent infections. In this report wild-type and genetically engineered viral mutants were analyzed in a mouse model using a recently developed approach to precisely quantify latently infected neurons. It was found that strain KOS/M established latent infections, as defined by the presence of the viral genome, in about 30% of the neurons. Thirty-three percent of the mice with this latent viral burden reactivated in vivo following hyperthermic stress. In contrast, mutants in which either the basal LAT promoter or the 5' end of the LAT gene was deleted established latency in only 10% of trigeminal neurons (P < 0.00001), and these mice were impaired for reactivation. Repair of the locus resulted in wild-type levels of establishment and reactivation, mapping this function to the LAT region. Finer mapping demonstrated that a 2.3-kb fragment that contains the major LAT transcripts was sufficient to promote efficient establishment and subsequent reactivation when expressed in the context of a foreign gene. Hyperthermic stress applied during the first 3 days postinfection resulted in greatly increased numbers of neurons harboring the latent viral genome. This approach was found to increase the level of establishment of LAT-null mutants to that normally achieved by wild-type KOS/M. These establishment-repaired mice reactivated with wild-type efficiency. Thus, the LAT gene serves to increase the number of neurons in which latency is established, and no direct role for the LAT locus in reactivation could be demonstrated.