Computational hypotheses about brain information processing can be expressed in neural network models. Neuroscientists have begun to compare such models in terms of their alignment with neural and behavioural data. The high parametric capacity of these models is essential to their ability to capture cognitive processes but also enables them to approximate arbitrary functions, making distinct models difficult to discriminate experimentally. Model comparisons using stimuli sampled from the training distribution often fail to reveal differences. This challenge can be met by optimizing stimulus sets for model discrimination and by leveraging out-of-distribution generalization as a severe test. This Review explains the emerging methods for optimizing stimuli to adjudicate among neural network models. These methods seek stimulus sets that are controversial among the models in that they make the models disagree in their predictions of the experimental data. We discuss the choices researchers must make, including a prior over candidate stimuli (such as naturalistic images), a measure of the power to discriminate among alternative models and a procedure for selecting or synthesizing stimulus sets that maximize model-comparison power. Historically, researchers have chosen either natural or artificial stimuli for a given study, prioritizing ecological validity or model-comparison power, respectively. Tempered by a prior, controversial stimuli offer a synthesis of these classical approaches, combining the greater ecological validity of naturalistic stimuli with the greater power for model comparison enabled by artificial stimuli. We offer a unified perspective on current work, drawing connections to Bayesian optimal experimental design.
A new study reports that olfactory receptor genes in the mosquito Aedes aegypti can be coexpressed in olfactory neurons and expressed in multiple subtypes of these neurons.
Early life stress shapes brain development during sensitive developmental windows through persistent epigenomic changes that influence gene regulation, neural circuit maturation and susceptibility to neuropsychiatric disorders. Evidence from animal models and human studies demonstrates that early adversity remodels DNA methylation, histone post-translational modifications and chromatin organization, leading to lasting alterations in transcriptional programmes. These responses are highly cell-type specific, brain-region specific and developmental-stage specific, with neurons and glia exhibiting distinct molecular adaptations that contribute to behavioural outcomes. Emerging evidence also implicates non-coding RNAs as important regulators of stress-induced gene expression, although their long-term roles remain less well defined than those of other epigenetic mechanisms. Peripheral epigenetic signatures associated with early life stress may provide accessible biomarkers of exposure and disease risk. Integrating epigenomic, transcriptomic, circuit-level and behavioural approaches will be essential for understanding biological embedding and identifying mechanisms that promote resilience and improve therapeutic interventions.
Most mammalian offspring are born helpless and depend on sustained parental care for survival, making caregiving one of the most conserved and essential behavioural repertoires in mammals. Yet infant-directed behaviours can be remarkably variable, ranging from care to neglect and aggression, even within the same individual. How neural circuits support both the robustness and the flexibility of parental behaviour remains poorly understood. Here we review recent advances in our understanding of the neural mechanisms of mammalian caregiving, placing circuit-level insights from rodent models into the broader context of parental diversity across mammals. Emerging evidence indicates that internal state and social experience dynamically reshape parental circuits across timescales from hours to weeks and that these circuits adapt their function to generate life stage-appropriate behavioural output. Rather than acting as fixed control systems for instinctive actions, parental circuits are highly plastic and context dependent. Parental behaviour therefore provides a powerful model for understanding how neural circuits are reconfigured to meet changing behavioural demands.
Cerebellar rhythms provide frequency-specific support for motor, cognitive and affective functions. These oscillations are not epiphenomenal but rather dynamically regulated, spatially organized control signals that contribute to the coordination of prediction, error correction, learning and internal model updating. By synchronizing neuronal activity across cerebellar and distributed brain networks at multiple timescales, cerebellar rhythms enable precise and adaptable behaviour and coordination across neural systems. Accordingly, they offer biologically grounded targets for network-level diagnostics and therapeutic neuromodulation. At the circuit level, cerebellar rhythms within distinct frequency bands, ranging from theta and beta to gamma and very high-frequency oscillations, arise from specific microcircuit mechanisms within the inferior olive, the granular and molecular layers of the cerebellar cortex, and the deep cerebellar nuclei. These rhythms structure spike timing and help shape synaptic plasticity windows, forming a frequency-organized substrate for learning and control. Here we describe a frequency-function-modulation framework that links cerebellar oscillations to their behavioural roles and to neuromodulatory interventions. By integrating evidence from animal studies, computational models and non-invasive stimulation studies, we position cerebellar oscillations as a bridge between cerebellar circuit dynamics and systems-level coordination, thereby providing a mechanistic rationale for precision neuromodulation across motor and cognitive domains in neurological and psychiatric conditions.
Microglia, the resident macrophages of the CNS parenchyma, are recognized as highly plastic, transcriptionally diverse cells whose phenotypes are moulded by development, region, sex, age, genotype and environment. Advances in single-cell and single-nucleus transcriptomics, chromatin accessibility profiling, and spatial multi-omics have negated binary frameworks of 'resting versus activated' or 'M1 (pro-inflammatory) versus M2 (anti-inflammatory)' and revealed a multidimensional state space that supports brain development, homeostasis and adaptive responses to perturbation. Building on the foundational concepts of the microglial sensome, homeostatic and disease-associated signatures, microglia exhibit transcriptomic state transitions in neurodegeneration, demyelination, infection and systemic inflammation. Moreover, a mechanistic framework for more 'hidden' microglial states has emerged, in which latent programmes that appear homeostatic at baseline are revealed by challenges and are instructed through innate immune training or tolerance. We argue that these covert reprogrammed states, which are shaped by ageing, genotype, sex, location and prior exposures such as sepsis or viral infection, help explain interindividual variability in disease trajectories. We conclude by outlining priorities for unifying state annotation across species and modalities, and for translating state-resolved insights into biomarkers and interventions.
In this Journal Club, Guilhem Ibos reflects on a 2011 study linking the neurophysiology of attention with the neuromodulation of cognitive functions.
Spontaneous brain activity is experimentally unconstrained. Embracing this freedom unlocks rich opportunities beyond task-based and naturalistic paradigms, the traditional Herculean pillars of constrained cognition. Emerging data-driven approaches can extract cognitively meaningful and translationally relevant insights from spontaneous brain activity: across profoundly altered states, across the human lifespan and across species.
Cognition and behaviour arise from computations in neural circuits, which can differ in their readiness for recruitment or in the computations and behavioural outputs that they generate. Mitochondria contribute to both circuit properties and their variability by shaping the cellular processes on which circuit function depends. Across neurons and glia, mitochondria provide bioenergetic support, regulate Ca2+ dynamics and reactive oxygen species levels, influence neurotransmitter synthesis and turnover, and sustain quality control programmes that preserve cellular integrity. The capacity of mitochondria to provide this support and their plasticity have been linked to circuit architecture, engagement and adaptation, with implications for learning and memory, reward and reinforcement, state-trait anxiety and motivation. Here we describe two complementary modes of mitochondrial support: a baseline mode, in which mitochondria sustain circuit architecture and physiological properties over long timescales, and an activity-evoked mode, in which local mitochondrial outputs support synaptic transmission and plasticity. Distinguishing these two modes helps to explain how behavioural modulators, including stress hormones, immune activity and metabolic signals, can shape behaviour by altering either baseline mitochondrial control of circuit readiness or activity-evoked mitochondrial support during circuit engagement.
In this Journal Club, Laura Bradfield discusses a 1998 study showing that scrub jays exhibit episodic-like memory.
Motor neuron diseases (MNDs) are caused by the progressive loss of motor neurons and eventually lead to paralysis and death. Once viewed as primarily neurocentric, MNDs are now recognized to be driven by intertwined cell-autonomous and non-cell-autonomous mechanisms. Dissecting these interactions is essential for developing effective therapies. Here, we describe induced pluripotent stem cell-derived 3D models that can be used to capture distinct aspects of MND pathology. We show that spinal cord organoids can be used to investigate cell-autonomous mechanisms and motor neuron-glia interactions (with axially elongated spinal cord organoids being particularly useful to study developmental vulnerability) as well as in 3D muscle and combined neuromuscular models to dissect muscle pathology and neuromuscular junction dismantling. In parallel, we discuss advances in bioengineering, machine learning and human trunk-like models, which together can begin to reproduce the coordinated co-development and spatial organization of the multiple tissues affected in MNDs. We discuss how these systems have advanced our understanding of disease mechanisms and highlight opportunities for drug repurposing. Finally, we propose a mechanism-informed and phenotype-informed framework to guide 3D model selection for future research and to prioritize promising avenues for therapeutic development.
A multisynaptic pathway from the piriform cortex to the spleen drives learned fear avoidance behaviour in mice.
Acetylcholine release in the hippocampus has been associated with diverse neural functions in learning and memory, including novelty, uncertainty detection, error correction, arousal and hidden state inference, while also modulating theta oscillations. Confronted with this plurality of roles, a unifying framework for interpreting cholinergic function is lacking. Recently, predictive models have emerged as a normative lens for understanding neural function, viewing the brain as learning to predict environmental dynamics. Within this framework, we propose that hippocampal acetylcholine serves as a fundamental learning signal, encoding state transition prediction errors - the magnitude of mismatches between predicted and actual transitions in environmental states. We suggest that, analogously to how dopamine signals reward prediction errors to guide value learning, acetylcholine guides structural learning, signalling state transition prediction errors. By condensing behaviour onto timescales amenable to spike-timing-dependent plasticity, theta sequences provide a potential substrate for calculating state transition prediction errors and driving the synaptic updates that revise the internal world model. Within the septo-hippocampal circuit, we propose that theta sequences relay predictions to the septum for comparison against observed state transitions, with the resulting cholinergic feedback gating plasticity in proportion to the mismatch. Plausibly, this mechanistic account unifies the diverse roles of acetylcholine, from novelty detection to hidden state inference, as aspects of a single computational principle - learning predictive internal world models from precisely-timed neural sequences.
Astrocytes are morphologically complex and functionally diverse glial cells that play central roles in neural circuits and disease. Although transcriptomic and physiological analyses have advanced understanding of astrocyte function, the intricacies of the relationship between gene expression and protein levels remains poorly understood. Proteins, rather than transcripts, execute the molecular processes that define astrocyte function, often within specialized subcellular compartments whose molecular composition has remained largely unresolved. Recent advances with genetically encoded proximity-dependent biotinylation have enabled mapping of astrocyte proteomes and subproteomes, revealing the spatial organization of protein networks that underpin astrocyte identity and function. In this Review we summarize strategies for defining astrocyte proteomes and subproteomes, emphasizing methodological innovations that permit cell-type-specific and spatially resolved proteomic profiling within intact neural tissue. The use of these approaches is shaping understanding of astrocyte biology by uncovering the molecular nature of their distributed physiology and the molecular basis of their interactions with neurons and other cells. Viewing astrocytes through a proteomic lens is beginning to help elucidate the molecular determinants of their morphology, signalling, and roles in homeostasis and disease. Such insights should enable detailed mechanistic understanding of astrocyte function and identify new avenues for targeting astrocytic contributions to brain disorders.
Epitranscriptomic regulation of cellular RNAs is a major mechanism of gene expression control in the brain. N6-Methyladenosine (m6A) is installed on thousands of mRNAs and non-coding RNAs, where it functions as a context-dependent regulator of RNA-protein interactions to control the amplitude and kinetics of gene expression. In the nervous system, m6A is critical for neurodevelopment, synaptic plasticity and adaptive responses to physiological stimuli, and its dysregulation has been linked to various brain disorders. In this Review, we present a comprehensive synthesis of how m6A is deposited, interpreted and dynamically regulated, and integrate recent advances to present a unified framework for its function in neural cells. We discuss how m6A coordinates RNA stability, translation, localization and chromatin-associated processes across developmental and adult contexts and how disruption of these pathways contributes to neurological disease. Finally, we explore challenges and future directions for the field.
In this Tools of the Trade article, Omid Yaghmazadeh describes the proof-of-concept development of transcranial radio frequency stimulation in mice, a potentially scalable and effective therapeutic platform for non-invasive deep-brain stimulation.
In this Journal Club, M. Jerome Beetz highlights a study published 2021 that examined hippocampal representation of large environments in flying bats.
AMPA receptors (AMPARs) mediate the majority of fast excitatory neurotransmission in the mammalian brain. Recent structural, functional and proteomic advances have reshaped our understanding of how these receptors assemble, gate and diversify within distinct synaptic environments. Notably, AMPAR function is governed by an interlocking set of regulatory layers that include alternative splicing and RNA editing within regions of the receptor subunits responsible for gating and permeation, and direct allosteric coupling with different families of auxiliary proteins. The assembly of AMPARs through a dedicated endoplasmic reticulum biogenesis pathway ensures accurate tetramer formation and provides a regulatory checkpoint for synaptic receptor abundance. Brain development introduces additional layers of control as editing, splicing and auxiliary-subunit expression shift from embryonic to adult states, whereas interactions with extracellular proteins contribute to the organization and diversification of synaptic architecture across circuits. Thus, AMPARs are dynamic macromolecular assemblies whose diversity underlies the breadth of glutamatergic signalling in health and disease. This Review synthesizes these emerging principles, highlighting how AMPARs transition from their molecular 'birth' to their deployment at functionally specialized synapses.
In this Tools of the Trade article, Colleen McLaughlin describes endocytome profiling, a systematic and quantitative approach for monitoring of cell-surface protein remodelling in the intact brain.
