The onco-functional reorganization of language network underlying metaplasticity induced by gliomas.

Background: Modern linguistic theories propose that language processing is supported by widely distributed large-scale modules that flexibly interact with domain-general networks. Progressive gliomas tend to reshape the language-associated systems causing dynamic reorganization that can be characterized by resting-state and task-based fMRI, thereby informing surgical decision-making. However, the interplay of neuropathological factors influencing reorganization patterns remains unclear, with ambiguity regarding the recruitment of cognitive resources across various levels of linguistic complexity. Methods: We included 100 patients with gliomas and 127 matched normal controls in this study. Activation analyses were performed on task fMRI data acquired during a picture-naming paradigm. Spectral DCM and connectome analyses were combined to examine state-dependent shifts between rest and task conditions. ANOVAs were used to assess relationships between clinicopathological factors and compensatory mechanisms. Mediation analyses were used to explore the mediated pathways among clinicopathological factors, topological indicators, and language performance. Results: First, we observed widely distributed activations associated with glioma-induced perilesional and remote reorganization patterns, exhibiting predominantly right-hemispheric lateralization in domain-general networks. VLSM analyses further showed that the spatial distribution of these activation clusters was significantly associated with tumor location and grade. Second, spectral DCM analyses indicated that task demands were associated with a greater number of positive effective connections across the reorganized language network, interactions among which were inherently dynamic varying with exogenous linguistic complexity and endogenous functional integrity. Third, ANOVAs suggested that alterations in language network topology were accompanied by flexible engagement of domain-general network components, which were associated with a partial restoration of the balance between integration and segregation under task conditions. Finally, we identified associations and potential mediation pathways among clinicopathological factors, topological properties, and language performance, consistent with the concept of glioma-related network metaplasticity. Conclusion: Our findings highlight that the dynamics of language reorganization depend on clinicopathological factors of gliomas and may thus open new perspectives for personalized surgical strategies for functional protection in the era of network neurosurgery. These group-level findings provide a modeling framework that may guide future research toward individualized network assessment and surgical planning.