In cluA- mutants of Dictyostelium, mitochondria are clustered near the cell center rather than being dispersed throughout the cytoplasm. We have examined two possible mechanisms that could account for this phenotype. First, we sought evidence that the cytoskeleton or a presumptive mitochondrion-cytoskeleton linkage was altered in mutant cells. We found that cytoskeletal structures in cluA- cells appeared normal by immunostaining, and that the distribution of peroxisomes in mutant cells was indistinguishable from that in wild type cells. Treatment of wild type cells with drugs that disrupted microtubules or actin filaments did not mimic the cluA- phenotype. Thus, cytoskeletal defects seemed unlikely to account for the mitochondrial clustering in cluA- cells. Observation of the movement of GFP-tagged mitochondria in wild type cells suggested that mitochondria are transported along microtubules, as in mammalian cells, rather than along actin filaments, as in budding yeast. Therefore, the similar phenotypes of cluA- Dictyostelium cells and clu1delta yeast cells argued against CluA/Clu1p acting as a mitochondrion-cytoskeleton linker. We next examined the ultrastructure of mitochondria in freeze-substituted, thin-sectioned cells. We found that the clustered mitochondria in cluA- cells are interconnected. Often, adjacent mitochondria are linked by narrow membranous strands, although sometimes the mitochondria are partially merged. The presence of narrow constrictions at presumptive division sites argues that the constriction step of division proceeds normally. Our data suggest that cluA- cells may be blocked at a very late step in fission of the outer mitochondrial membrane.