Uncertainties about the ultimate biologic consequences of using live virus vaccines to confer immunologic protection against CMV have focused attention on the use of noninfectious subunit vaccines. At least two classes of such preparations have been demonstrated to be effective in other systems. The first is virus particles bearing the relevant antigens but lacking nucleic acid (eg, hepatitis vaccine [31]). And the second class is biologically or chemically synthesized proteins or peptides with appropriate immunogenicity (eg, foot and mouth disease virus vaccine [32]). In this paper, two noninfectious CMV particles and several viral proteins have been discussed in view of their potential for use as such a vaccine. The two noninfectious virus particles discussed are referred to as dense bodies and NIEPs. The use of dense bodies for vaccine purposes has been suggested by others [14], but the simplicity of their composition has only recently been established [17]. Two characteristics of these particles make them attractive prospects for vaccine purposes. First, neither contains more than trace amounts of DNA or infectivity (ie, less than or equal to 0.1% that of virions). Thus, the concerns about possible adverse consequences of introducing DNA with the vaccine are greatly reduced. Second, both NIEPs and dense bodies contain all of the glycoprotein species present in virions and in approximately the same relative amounts. If, as anticipated, these proteins are important in eliciting the immune response to CMV, then NIEPs and dense bodies may be as effective as virions in that capacity. The fact that NIEPs contain the full complement of virion proteins, and in approximately the same relative amounts, suggests that they may produce a more complete immunologic response than dense bodies, which lack all of the capsid and most of the tegument proteins of the virion. Although NIEPs normally represent only a small percentage of the extracellular particles (eg, less than 1%), we have found that strain AD169 produces them in amounts nearly equivalent to virions. More importantly, we have shown here that NIEP production is essentially unaffected following treatment of infected cells with a concentration of hydroxyurea that reduced virion and dense body production by more than 90% (Fig. 6). Thus, by using strain AD169 to infect cells and hydroxyurea treatment for selective enrichment, it is possible to produce NIEPs in relatively large amounts and with theoretically very low levels of contaminating virions (ie, less than or equal to 0.01% infectivity of equivalent amount of virions).