Clostridium tyrobutyricum is the main spoilage agent of late blowing defect (LBD) in Grana Padano and Parmigiano-Reggiano cheeses; LBD is characterized by openings and holes and is sometimes accompanied by cracks and an undesirable flavor. Even a very few spores remaining in the cheese curd may cause LBD; thus, it is essential to eradicate them during milk natural creaming. By this process, most of the bacteria, somatic cells, and spores rise to the top of the milk, together with the fat globules, and are removed with the cream. Previous studies suggested that milk immunoglobulins mediate the interactions between fat globules and bacteria that occur upon creaming but no direct evidence for this has been found. Moreover, other physical chemical interactions could be involved; for example, physical entrapment of spores among globule clusters. To maximize the efficiency of the natural creaming step in removing Cl. tyrobutyricum, it is essential to understand the nature of spore–globule interactions. With this aim, raw milk was contaminated with spores of Cl. tyrobutyricum before going to creaming overnight at 8°C, after which spore and bacteria removal was >90%. The obtained cream was analyzed by light interference contrast and fluorescence microscopy and by transmission electron microscopy (TEM). Results showed that most of the vegetative cells and spores, which were stained with malachite green before addition to milk, adhered tightly to the surface of single fat globules, the membranes of which appeared heterogeneous when stained with the fluorescent dye DilC18(3)-DS. Using the same dye, we observed transient and persistent interactions among globules, with formation of clusters of different sizes and partial coalescence of adhering membranes. Transmission electron microscopy examination of replicates of freeze-fractured cream allowed us to observe tight adhesion of spores to fat globules. Ultrathin sections revealed that this adhesion is mediated by an amorphous, slightly electron-opaque material, sometimes granular in appearance. Bacteria also adhered to different fat globules, linking them together, which suggests that adhesion was strong enough to maintain a stable contact. Although we cannot exclude physical entrapment of bacteria among fat globule clusters, we show for the first time that most of the bacteria are adhered to fat globules by an electron-opaque material whose nature has yet to be determined. Immunoglobulins are certainly the best candidates for adhesion but other compounds may be involved.