Mutations in the fused in sarcoma/translated in liposarcoma gene (FUS/TLS, FUS) have been identified in sporadic and familial forms of amyotrophic lateral sclerosis (ALS). FUS is an RNA-binding protein that is normally localized in the nucleus, but is mislocalized to the cytoplasm in ALS, and comprises cytoplasmic inclusions in ALS-affected areas. However, it is still unknown whether the neurodegeneration that occurs in ALS is caused by the loss of FUS nuclear function, or by the gain of toxic function due to cytoplasmic FUS aggregation. Cabeza (Caz) is a Drosophila orthologue of human FUS. Here, we generated Drosophila models with Caz knockdown, and investigated their phenotypes. In wild-type Drosophila, Caz was strongly expressed in the central nervous system of larvae and adults. Caz did not colocalize with a presynaptic marker, suggesting that Caz physiologically functions in neuronal cell bodies and/or their axons. Fly models with neuron-specific Caz knockdown exhibited reduced climbing ability in adulthood and anatomical defects in presynaptic terminals of motoneurons in third instar larvae. Our results demonstrated that decreased expression of Drosophila Caz is sufficient to cause degeneration of motoneurons and locomotive disability in the absence of abnormal cytoplasmic Caz aggregates, suggesting that the pathogenic mechanism underlying FUS-related ALS should be ascribed more to the loss of physiological FUS functions in the nucleus than to the toxicity of cytoplasmic FUS aggregates. Since the Caz-knockdown Drosophila model we presented recapitulates key features of human ALS, it would be a suitable animal model for the screening of genes and chemicals that might modify the pathogenic processes that lead to the degeneration of motoneurons in ALS.