There is no effective vaccine for Marburg virus (MBGV) or any other filovirus, nor enough pertinent information to expedite rational vaccine development. To ascertain some of the minimal requirements for a MBGV vaccine, we determined whether whole inactivated MBGV, or a baculovirus-expressed virion subunit, could be used to immunize guinea pigs against a lethal infection. Baculovirus recombinants were made to express the MBGV glycoprotein (GP) either as a full-length, cell-associated molecule or a slightly truncated (5.4%) product secreted into medium; the latter, for its far greater ease in manipulation, was tested for its vaccine potential. Like MBGV GP, both the full-length and truncated GP expressed by baculovirus recombinants were abundantly glycosylated with both N- and O-linked glycans; differences in glycosylation were detectable, but these could not be shown to affect antigenicity with respect to available antibodies. The recombinant truncated glycoprotein elicited protection against lethal challenge with the MBGV isolate from which it was constructed and less effectively against an antigenically disparate MBGV isolate. Killed (irradiated) MBGV antigen was protective, in a reciprocal fashion, against both MBGV types. In a preliminary assessment of possible protective mechanisms, serum antibodies from immune animals were shown to be sufficient for protecting naive guinea pigs from lethal MBGV infections