The brain’s most sophisticated cognitive functions are powered by the same metabolic machinery that, over decades, creates the conditions for their destruction. A new study from the Technical University of Munich provides direct evidence for this trade-off, using a novel imaging approach that simultaneously measures where brain regions connect and how much energy they consume.
The study, led by Valentin Riedl and published June 22 in PNAS, introduces the metabolism-weighted connectome (MwC), a brain network map in which both the connections between regions and the metabolic activity of each region carry weight. The results show that the same cortical hubs that sustain higher-order cognition, memory, attention, self-referential thought, are disproportionately vulnerable to Alzheimer’s pathology later in life.
Beyond standard connectomics
Conventional connectomics treats all brain regions as functionally equivalent nodes. It maps which regions are connected and how strongly, but it ignores the intensity of local neural activity: a hub that fires at high rates and consumes large amounts of glucose is treated the same as a quiet hub with the same connection count.
The Munich team addressed this by using integrated PET/MRI scanners, hybrid machines that acquire functional MRI and FDG-PET data simultaneously in the same subject. The fMRI captures the BOLD signal, from which functional connectivity is derived. The FDG-PET measures the cerebral metabolic rate of glucose, a direct proxy for local synaptic activity. By multiplying the functional connectivity of each region by the metabolic activity of its neighbors, the researchers produced a fully weighted graph in which a region achieves high MwC not by being densely connected, but by being preferentially coupled to regions that themselves sustain high energy consumption.
The metric explains 39.9% of the variance in regional glucose metabolism across the brain, roughly four times better than standard weighted degree centrality (9.6%), a statistically significant improvement (p = 0.01).
The vulnerable hubs
High-MwC regions cluster in the brain’s higher-order cognitive networks, the default mode network (posterior cingulate, precuneus, medial prefrontal cortex, angular gyrus), the salience network, and the cingulo-opercular network. MwC increases monotonically from sensory networks (visual, somatomotor) through association networks to these cognitive networks.
The connection to Alzheimer’s was validated three ways. Transcriptomic analysis using the Allen Human Brain Atlas showed that high-MwC regions express genes enriched for oxidative phosphorylation, mitochondrial function, and synaptic signaling, six of the ten top KEGG pathways linked to neurodegenerative disease. Amyloid-PET data from 224 participants in the Alzheimer’s Disease Neuroimaging Initiative confirmed that beta-amyloid plaques preferentially accumulate in high-MwC regions. And SV2A PET, which measures synaptic vesicle density, confirmed increased synaptic density in the same hubs.
The metabolic burden hypothesis
The authors propose that lifelong high metabolic throughput in these regions generates increased oxidative stress and mitochondrial demand. Over decades, this renders them selectively vulnerable to proteinopathy and neurodegeneration. The pattern matches the known topography of early Alzheimer’s: the default mode network is the first network to show both hypometabolism and amyloid deposition.
“This suggests that the same metabolic machinery that supports cognitive complexity also incurs a biological cost,” Riedl said. The same energy-intensive hubs that enable integrative cognition, remembering, planning, reflecting, are metabolically exposed and disease-prone.
The study involved 40 healthy subjects across two scanning sites (Munich and Vienna), with data parcellated using the Human Connectome Project’s 334-region atlas. The findings were consistent across both sites and a replication cohort.
Source: Ashrafi M, Fraticelli L, Castrillón G, Riedl V. Metabolism-weighted brain connectome reveals synaptic integration and vulnerability to neurodegeneration. PNAS. 2026;123(26):e2531706123. doi:10.1073/pnas.2531706123