Mouse Brain Study Reveals Shifting Protein Networks During Maze Learning

Every time an animal learns to navigate a familiar environment, its brain quietly rewires itself at the molecular level. A new study published in the journal Neuroinformatics offers one of the most detailed looks yet at how that rewiring unfolds over time in a brain region long associated with habit and action: the dorsal striatum. Researchers led by S. Gutman and Izhak Michaelevski of Ariel University, working with colleagues at Tel Aviv University and the Weizmann Institute of Science, tracked the protein composition of the mouse dorsal striatum across five days of radial-arm maze training. Their results, generated with ion-mobility-enhanced data-independent mass spectrometry and analyzed through two complementary network approaches, reveal a striking choreography of molecular change in which different protein systems take the stage at different points during learning.

The dorsal striatum sits at the heart of the brain’s basal ganglia, a cluster of structures that translate motivation into movement and, critically, help convert repeated experiences into habits. While the hippocampus has traditionally dominated research into spatial memory, decades of work have shown that striatal memory systems contribute substantially to navigation, particularly when animals rely on familiar routes and stimulus-response associations. The radial-arm maze, in which animals must retrieve food rewards from the ends of multiple arms radiating from a central platform, engages spatial learning, motivation, motor execution, and reward processing simultaneously. This makes it a rich but complicated behavioral paradigm for molecular analysis, because changes in brain chemistry could reflect any combination of these processes.

In the new study, the researchers trained male mice in the radial-arm maze and collected dorsal-striatal tissue after trials on Days 0, 1, 3, and 5 of the training schedule. Crucially, they included a control group of habituated, food-restricted mice that never encountered the maze, designated the NoRAM group, which was collected before the first trial. This design allowed the team to separate a broad, maze-associated state difference from more subtle changes that emerged as training progressed. The study was approved under standard animal research reporting guidelines, and all data supporting the findings are available within the paper and its supplementary materials.

On the technical side, the team employed ion-mobility-enhanced data-independent acquisition LC-MS/MS, a powerful proteomic strategy that fragments and measures peptides systematically rather than relying on the stochastic selection of ions. This approach enables deep, reproducible quantification of thousands of proteins across many samples, which is essential when the goal is to track coordinated molecular programs over time. Rather than simply asking which individual proteins changed, the researchers applied two complementary network-level frameworks: exploratory factor analysis, which identifies dominant axes of variation across samples, and weighted gene co-expression network analysis, or WGCNA, which groups proteins into modules based on correlated abundance patterns and relates those modules to experimental conditions.

The first major finding was the dominance of a single, broad difference between trained and untrained animals. Mice exposed to the radial-arm maze showed a dorsal-striatal protein landscape strongly enriched in ribosomal proteins and translation-initiation factors, proteasome components, and proteins involved in synaptic-vesicle trafficking. This triad of protein synthesis, protein degradation, and vesicle transport makes biological sense: learning is widely understood to require new protein production, the regulated turnover of existing proteins, and the delivery of molecular cargo to synapses where memories are thought to be encoded. The prominence of translation machinery echoes earlier findings from the same laboratory showing dynamic protein expression in the hippocampus during long-term spatial memory formation.

Beneath this dominant maze-versus-no-maze difference, the two analytical frameworks resolved distinct, time-dependent patterns within the trained series. Comparing Day 1 with Day 3, the researchers found that the emphasis shifted toward actin and cytoskeletal regulation, membrane trafficking, mitochondrial respiration, and ATP-generating processes. This middle phase of training appears to coincide with structural remodeling of neuronal architecture, since the actin cytoskeleton underpins the growth and reshaping of dendritic spines, the tiny protrusions where synaptic connections form. The concurrent rise of mitochondrial and energy-related proteins suggests that this remodeling is metabolically expensive, consistent with prior evidence that cognitive demand draws heavily on local energy supplies in the brain.

By Day 5, a different molecular program had come to the fore. Later training contrasts emphasized postsynaptic-density organization, vesicle recycling, protein quality control, and metabolic support. The postsynaptic density is a dense protein complex beneath the receiving side of a synapse, anchoring neurotransmitter receptors such as NMDA receptor subunits and scaffolding proteins like PSD-95 and SHANK3, both of which appeared among the study’s central proteins. Enrichment of vesicle-recycling and quality-control systems suggests that as training matures, the striatum may be stabilizing and refining synaptic connections rather than building new ones, a transition that fits conceptual models of memory consolidation in which initial plasticity gives way to maintenance.

When the researchers integrated their factor-analysis and network-analysis results, they identified 65 shared central proteins that link synaptic signaling and receptor organization with proteostasis, energy production, and structural remodeling. Computational enrichment of candidate transcription-factor and microRNA targets, using databases such as TRANSFAC, TransmiR, and miRTarBase, nominated plausible regulatory players, including factors associated with nuclear respiratory control and proteasome regulation, although the authors stress that these regulatory associations were not directly measured. The team is also careful about the limits of interpretation: because the NoRAM controls were collected before task exposure and no matched control for activity, reward, or repeated testing was included, the observed changes cannot be assigned specifically to spatial learning or to any particular navigational strategy. In addition, bulk tissue analysis of pooled striatum precludes conclusions about specific cell types, subregions, or individual synapses.

Despite these caveats, the study delivers what the authors describe as a biology-centered, hypothesis-generating map of the molecular systems engaged by repeated spatial-task exposure. By showing that the dorsal striatum does not simply switch on a single learning program but instead recruits translation, cytoskeletal, energetic, synaptic, and proteostatic modules in a temporally ordered sequence, the work provides a framework for future experiments designed to test these programs directly. It also prioritizes specific molecular systems, from ribosomal and proteasomal machinery to mitochondrial respiration and postsynaptic scaffolding, for temporal and functional validation in targeted studies. As proteomics continues to mature into a tool for watching the brain change in real time, studies like this one bring the field closer to understanding not just where memories live in the brain, but how their molecular foundations are laid, remodeled, and maintained day by day.

Subject of Research: Temporal proteomic dynamics of the dorsal striatum in mice during repeated radial-arm maze training

Article Title: Temporal Dynamics of Dorsal-Striatal Protein Networks During Radial-Arm Maze Training

Article References: Gutman, S., Borovok, N., Kirby, M., Levin, Y., & Michaelevski, I. (2026). Temporal Dynamics of Dorsal-Striatal Protein Networks During Radial-Arm Maze Training. Neuroinformatics, 24(4), Article 63. https://doi.org/10.1007/s12021-026-09819-9

Image Credits: AI Generated

DOI: 10.1007/s12021-026-09819-9

Keywords: dorsal striatum, radial-arm maze, proteomics, spatial navigation, WGCNA, exploratory factor analysis, LC-MS/MS, synaptic plasticity, protein networks, mice, memory consolidation, temporal proteomics