In herpesviruses, homologues of glycoprotein B (gB) are essential membrane proteins which are involved in fusion. However, there is no clear evidence regarding the location of the fusogenic domain on gB. By using bovine herpesvirus 1 (BHV-1) as a model, we studied the relationship between the structure and the fusogenic activity of gB. This was achieved by expressing genes of different gB derivatives containing specific truncations at the end of segments 2 or 3 of the transmembrane region in Madin-Darby bovine kidney cells under the control of the bovine heat-shock protein hsp70A gene promoter. All expressed gB products were structurally similar to authentic gB. One truncated form of gB, gBt, which contains residues 1-763, was efficiently secreted. However, gBtM (residues 1-807), which includes the first two segments at the carboxyl terminus, showed unstable retention on the cell surface, whereas gBtMA (residues 1 829), which contains all three membrane-spanning segments, was mostly intracellularly retained with some unstable surface anchorage. Another truncated gB, gBtDAF, which has gB residues 1-763 (gBt) and a human decay-accelerating factor (DAF) carboxyl tail, was also expressed. The DAF fragment provided a signal for the addition of a glycosyl phosphatidylinositol-based membrane anchor, which could target the gBt chimeric protein on the cell membrane. Immunofluorescence staining and pulse-chase kinetic studies support the theory that gBtM, gBtMA, and gBtDAF are retained on nuclear and cellular membranes via different segments of the transmembrane region or the DAF fragment, respectively. For the cells expressing gBt or gBtM, no cell fusion was observed, whereas cells expressing gBtMA clearly showed fusion. However, in gBtDAF cells, the overexpression and cellular accumulation of recombinant gB products did not cause fusion either, which supports our contention that the fusion phenomenon in gBtMA cells is caused by the fusogenic activity of the expressed gBtMA. With the help of sequence analysis, our results indicate that segment 2 of the transmembrane anchor region might be a fusogenic domain, whereas the real anchor is segment 3.