Drosophila is an excellent model organism for studying human neurodegenerative diseases (NDs). However, there is still almost no experimental system which could directly observe the degeneration of neurons and automatically quantify axonal degeneration. In this study, we created MeDUsA (a ‘method for the quantification of degeneration using fly axons’), a standalone executable computer program based on Python that combines a pre-trained deep-learning masking tool with an axon terminal counting tool. This software automatically quantifies the number of retinal R7 axons in Drosophila from a confocal z-stack image series. Using this software, we were able to directly demonstrate that axons were degenerated by the representative causative genes of NDs for the first time in Drosophila. The fly retinal axon is an excellent experimental system that is capable of mimicking the pathology of axonal degeneration in human NDs. MeDUsA rapidly and accurately quantifies axons in Drosophila photoreceptor neurons. It enables large-scale research into axonal degeneration, including screening to identify genes or drugs that mediate axonal toxicity caused by ND proteins and diagnose the pathological significance of novel variants of human genes in axons.

In human neurodegenerative diseases, neurons undergo axonal degeneration months to years before they die. Here, we developed a system modeling early degenerative events in Drosophila adult photoreceptor cells. Thanks to the stereotypy of their axonal projections, this system delivers quantitative data on sporadic and progressive axonal degeneration of photoreceptor cells. Using this method, we show that exposure of adult female flies to a constant light stimulation for several days overcomes the intrinsic resilience of R7 photoreceptors and leads to progressive axonal degeneration. This was not associated with apoptosis. We furthermore provide evidence that loss of synaptic integrity between R7 and a postsynaptic partner preceded axonal degeneration, thus recapitulating features of human neurodegenerative diseases. Finally, our experiments uncovered a role of postsynaptic partners of R7 to initiate degeneration, suggesting that postsynaptic cells signal back to the photoreceptor to maintain axonal structure. This model can be used to dissect cellular and circuit mechanisms involved in the early events of axonal degeneration, allowing for a better understanding of how neurons cope with stress and lose their resilience capacities. SIGNIFICANCE STATEMENT Neurons can be active and functional for several years. In the course of aging and in disease conditions leading to neurodegeneration, subsets of neurons lose their resilience and start dying. What initiates this turning point at the cellular level is not clear. Here, we developed a model allowing to systematically describe this phase. The loss of synapses and axons represents an early and functionally relevant event toward degeneration. Using the ordered distribution of Drosophila photoreceptor axon terminals, we assembled a system to study sporadic initiation of axon loss and delineated a role for non-cell-autonomous activity regulation in the initiation of axon degeneration. This work will help shed light on key steps in the etiology of nonfamilial cases of neurodegenerative diseases.

Background: Drosophila is an excellent model organism for studying human neurodegenerative diseases (NDs), and the rough eye phenotype (REP) assay is a convenient experimental system for analysing the toxicity of ectopically expressed human disease genes. However, the association between REP and axonal degeneration, an early sign of ND, remains unclear. To address this question, we developed a method to evaluate axonal degeneration by quantifying the number of retinal R7 axons in Drosophila; however, it requires expertise and is time-consuming. Therefore, there is a need for an easy-to-use software that can automatically quantify the axonal degeneration. Result: We created MeDUsA (a method for the quantification of degeneration using fly axons), which is a standalone executable computer program based on Python that combines a pre-trained deep-learning masking tool with an axon terminal counting tool. This software automatically quantifies the number of axons from a confocal z-stack image series. Using this software, we have demonstrated for the first time directly that axons degenerate when the causative factors of NDs (αSyn, Tau, TDP-43, HTT) were expressed in the Drosophila eye. Furthermore, we compared axonal toxicity of the representative causative genes of NDs and their pathological alleles with REP and found no significant correlation between them. Conclusions: MeDUsA rapidly and accurately quantifies axons in Drosophila eye. By simplifying and automating time-consuming manual efforts requiring significant expertise, it enables large-scale, complex research efforts on axonal degeneration, such as screening to identify genes or drugs that mediate axonal toxicity caused by ND disease proteins.