A regime shift is a sudden and drastic change in natural ecosystems. Ecological theory suggests that such shifts, potentially triggered by interactions among a few species, can propagate throughout an entire system via interspecific relationships. However, empirical evidence for these cascading regime shifts remains limited. In this study, we applied nonlinear time series analysis to a 37-year-long dataset, which includes weekly monitoring of 500 phytoplankton species in Lake Biwa, Japan. Our findings reveal that the dramatic changes in community composition and chlorophyll-a concentration observed in the early 1990s can be attributed to successive stability shifts across multiple taxonomic groups. Furthermore, network analysis provided a more detailed perspective, showing that the regime shift originated with the genus Staurastrum and gradually spread to other species in the community network based on their proximity. Over five years, this shift ultimately affected more than 15 genera. A deeper understanding of cascading regime shifts in ecological networks can enhance predictions of biodiversity responses to global change and contribute to more effective ecosystem management.