To determine the type of cell(s) that contain latent varicella-zoster virus (VZV) DNA, we prepared pure populations of neurons and satellite cells from trigeminal ganglia of 18 humans who had previously had a VZV infection. VZV DNA was present in 34 of 2,226 neurons (1.5%) and in none of 20,700 satellite cells. There was an average of 4.7 (range of 2 to 9) copies of VZV DNA per latently infected neuron. Latent VZV DNA was primarily present in large neurons, whereas the size distribution of herpes simplex virus DNA was markedly different.Varicella-zoster virus (VZV) is present in a latent form in sensory ganglia of humans who have had a primary infection (varicella) with VZV (11). This is demonstrated by the reactivation of VZV in these individuals as herpes zoster and by the presence of VZV DNA and proteins in ganglia recovered at autopsy (2,3,5,9,10,12,13,15,16,19,20,(22)(23)(24)26). Different laboratories have ascribed the site of latency to either neurons (12-14, 18, 22) or perineuronal satellite cells (5, 26); some laboratories suggest that neurons are the primary site of latency together with a smaller proportion of satellite cells containing the VZV genome (16,20). The primary aim of the experiments described here was to determine the type of cell in trigeminal ganglia that harbors latent VZV and also to obtain data on the proportion of these cells that contain latent VZV. One reason why the site of VZV latency has remained an open question is that neurons and satellite-supporting cells are tightly associated in ganglia and the in situ methods used to locate VZV DNA did not clearly resolve the source of the hybridization signal. This resolution was easier for latent HSV because a strong hybridization signal to HSV facilitated its localization to the nucleus of the neuron. In contrast, the hybridization signal from VZV DNA is considerably weaker and dispersed over both nucleus and cytoplasm. While the hybridization methodology has improved, these considerations remain (21, 30). To reduce the ambiguities inherent in the in situ methods, we developed a procedure to separate cell types prior to analysis. The separation procedure has been verified with a study of HSV (1). This report presents data on the cellular localization of latent VZV, the frequency of latency, and the number of viral genomes present in latent cells.