The existence of separate gelatinase granules in human neutrophils has been a matter of debate in recent years. We have demonstrated that the 135-kD form of neutrophil gelatinase is a complex of 92-kD gelatinase and a novel 25-kD protein termed neutrophil gelatinase-associated lipocalin (NGAL) that, in addition to being complexed with part of the gelatinase, is localized in free form in peroxidase-negative specific granules. Because this association was not appreciated in earlier studies, we decided to reassess the ultrastructural localization of gelatinase using specific antibodies without immunoreactivity towards NGAL. Double-labeling immunogold electron microscopy was performed on frozen thin sections of human neutrophils. Twenty-four percent of all peroxidase-negative granules were labeled with antigelatinase antibody, but not with antilactoferrin antibody. These granules are defined as gelatinase granules. Sixteen percent reacted with antilactoferrin antibody but not with antigelatinase antibody. The rest (60%) reacted with both antibodies. All granules labeling for lactoferrin are defined as specific granules. Gelatinase granules were observed as round and oval forms of considerably smaller size than specific granules, and were less electron dense. Isolated granules obtained by subcellular fractionation were also examined by immunoelectron microscopy. This demonstrated that peroxidase-negative granules comprise a continuum from the most dense granules that contain lactoferrin but no gelatinase to the lightest that contain gelatinase but no lactoferrin. Thus, gelatinase granules do exist as a subpopulation of peroxidase-negative granules and may allow for exocytosis of gelatinase during neutrophil diapedesis without substantial mobilization of other peroxidase-negative granules, ie, specific granules.