The Brain's Lipid Map: What a New Atlas Means for Neurofeedback

Every rhythm a neurofeedback clinician trains comes from ion currents crossing membranes and traveling between regions along myelinated axons. Both the membranes and the myelin are made mostly of fat. Yet for decades, brain mapping has cataloged cells and genes while treating the fatty layer as inert packaging. A 2026 study in Nature changes that picture by charting 172 lipids across the entire mouse brain at near-cellular resolution (Fusar Bassini et al., 2026).

The result is a map of 539 chemically distinct territories, which the authors call lipizones. These territories follow anatomy, trace long-range connections, reveal hidden diversity in white matter, and shift within weeks during pregnancy. The study is a basic-science resource, not a clinical tool. Even so, it speaks to questions that neurofeedback professionals ask every day about timing, connectivity, and plasticity. This post connects the two while marking the limits of a mouse study.

Every rhythm you train runs on fat

A lipid is a fat-like molecule that dissolves poorly in water. In the brain, lipids form the cell membrane, the thin boundary that separates a neuron from its surroundings. They also build myelin, the insulating wrap that an oligodendrocyte lays around an axon. O'Brien and Sampson (1965) measured lipids at 36% to 40% of gray-matter dry weight, 49% to 66% of white-matter dry weight, and 78% to 81% of myelin dry weight.

Three lipid families dominate these membranes. A phospholipid has a water-attracting head and two water-avoiding tails, so phospholipids line up in the two-layer sheet that forms every membrane. A sphingolipid belongs to a second family concentrated in myelin. Cholesterol stiffens and thickens the sheet, and myelin cannot form without it (Saher et al., 2005).

Membrane composition is not a passive setting for embedded proteins. An ion channel is a membrane protein that opens a pore for specific ions, and the currents through these pores generate every EEG rhythm. Levitan et al. (2010) reviewed evidence that membrane cholesterol regulates members of all major ion-channel families, altering their kinetics and voltage dependence. Change the lipid environment and the channel changes with it.

Myelin matters for the EEG in a different way. Conduction velocity, the speed at which an action potential travels along an axon, depends on myelin thickness and the spacing of its gaps. Long-range synchrony between two cortical areas requires that signals arrive on time, and myelin sets that timing. Pajevic et al. (2014) calculated that a delay of only 5 milliseconds can switch two coupled gamma oscillators from reinforcing to canceling each other.

The EEG is an electrical signal, but its generators are chemical structures built from lipids. Membrane composition shapes the channels that produce local rhythms, and myelin composition shapes the timing that links distant regions. A map of where each lipid sits is therefore a map of the hardware beneath every rhythm a clinician trains.

How the team mapped 172 lipids across a whole brain

Lipidomics is the identification and measurement of lipid molecules in a sample. Conventional lipidomics grinds tissue and reports its average composition, losing the location of every molecule. The atlas instead used MALDI-MSI, a form of mass spectrometry imaging. A laser releases and charges molecules from a thin tissue slice point by point, so each measurement retains its coordinates. The laser spot measured 5 micrometers, and the spots sat 25 micrometers apart, so each pixel sampled portions of one to four cells (Fusar Bassini et al., 2026).

The team collected 109 coronal sections from 11 adult mice and recorded 7.2 million pixels. They aligned every section to the Allen Brain Atlas Common Coordinate Framework, a shared three-dimensional reference, so that lipid maps could be compared directly with gene and cell-type maps. Scans from different days create a batch effect, a difference caused by processing rather than biology. They corrected it with a framework called uMAIA, developed in an earlier study (Schede et al., 2025). After quality control, they annotated 172 lipids with high confidence (Fusar Bassini et al., 2026).

The team then applied cluster analysis, grouping pixels with similar lipid profiles without telling the algorithm anything about anatomy. The resulting clusters were the lipizones. A lipizone is a brain territory defined only by its local lipid composition, and the 539 of them formed a hierarchy of 8 classes, 31 subclasses, and 222 supertypes. The first split separated gray matter, rich in cell bodies and synapses, from white matter, rich in myelinated axon bundles (Fusar Bassini et al., 2026).

Two limits apply to every finding below. Each pixel mixes cell bodies, axons, and extracellular material, so the atlas describes chemical neighborhoods rather than single cells. And MALDI-MSI reports a lipid's mass, not its exact structure, so some annotations are tentative and were excluded from biological conclusions (Fusar Bassini et al., 2026).

The atlas is a chemistry map registered to the same coordinates as the field's gene and cell-type maps. Its resolution is near-cellular, its coverage is the whole mouse brain, and its clusters were discovered without anatomical labels. Those three properties are what allow lipids to be compared, for the first time, with everything else we know about brain organization.

Lipids carry a shadow of the wiring diagram

The first test was whether lipizones meant anything anatomically. They did. Excluding fiber tracts and cerebellum, 76% of lipizones overlapped a known cell type territory, and several specific cell populations were rediscovered from lipids alone. Lipizones also tracked cytoarchitecture, the layered arrangement of cell bodies and fibers. They reproduced cortical layers and even separated the inner and outer halves of the cerebellar granule layer (Fusar Bassini et al., 2026).

The surprise came next. Gene-based maps mark cell bodies, because messenger RNA is concentrated there. Lipids fill the whole neuron, including the axon terminal that may lie millimeters away. For 106 lipizones, the team found two spatially separate clusters. One sat on a population of cell bodies, and the other sat where that population's axons end, with a documented pathway between them (Fusar Bassini et al., 2026).

A broader analysis confirmed the pattern. Pairs of brain regions with a known connection shared at least one lipizone 33.8% of the time, compared with 14.8% for unconnected pairs at similar distances. The hippocampus clustered with its input and output partners, including the lateral septum and entorhinal cortex. Not every circuit left a trace, however: the projection from the thalamus to cortical layer 4 was not visible in the lipid hierarchy. A shared lipid signature is a correlation, which may reflect shared axons or similar membrane needs rather than connectivity itself (Fusar Bassini et al., 2026).

For neurofeedback professionals, the relevance is conceptual rather than clinical. Coherence training and other connectivity-based protocols assume the trained sites are linked by physical pathways. The atlas suggests that those pathways leave a chemical footprint that can be read without tracers or genetic labels. Whether that footprint can be read in human tissue remains an open question.

Most of the lipid map agrees with the cell-type map, which is reassuring. The disagreement is the interesting part: lipids follow axons to their targets, so connected regions share chemical signatures that gene maps miss. The authors are careful to call this a correlation, and the thalamus-to-layer-4 pathway shows that the footprint is incomplete.

White matter is not one tissue

Transcriptomic surveys recognize fewer than ten oligodendrocyte subtypes in the mouse brain (Marques et al., 2016). The lipid atlas found far more chemical diversity than that. One branch of the hierarchy held 65 lipizones confined to fiber tracts, cerebellar white matter, and hindbrain, and another held 70 myelin-rich territories. Individual lipizones were restricted to structures such as the fimbria, internal capsule, optic tract, and corticospinal tract (Fusar Bassini et al., 2026).

Lipids restricted to oligodendrocytes explained more of the variation between white-matter lipizones than lipids restricted to neurons or astrocytes, so the authors concluded that the oligodendrocyte lipidome drives the pattern. Part of that pattern appears to record development. Myelination proceeds in waves after birth, reaching the lateral corpus callosum before the medial part, and in the adult atlas, anterior callosal lipizones are still sorted into lateral and medial pools. Antibody staining for three myelin proteins showed a zonation that resembled the lipizones (Fusar Bassini et al., 2026).

This matters because myelin is not fixed after development. Activity-dependent myelination is the process by which neuronal firing recruits new oligodendrocytes and new myelin onto active axons. Gibson et al. (2014) showed that optogenetic stimulation of mouse premotor cortex increased oligodendrocyte production and myelination within the stimulated circuit. McKenzie et al. (2014) found that mice learning a complex running wheel rapidly generated new oligodendrocytes, and that blocking this production prevented them from mastering the task.

The link to brain rhythms is becoming direct. Steadman et al. (2020) blocked learning-induced oligodendrocyte formation in adult mice, and fear learning then no longer increased the coupling between hippocampal sharp-wave ripples and cortical sleep spindles. Noori et al. (2020) modeled large-scale brain networks in which conduction delays adapt to activity and found that adaptive myelination let the network tune itself toward stable phase synchrony. In these accounts, the brain learns timing.

Neurofeedback research has begun to detect white-matter change after training. Ghaziri et al. (2013) randomized 30 healthy students to 40 sessions of beta-1 training at F4 and P4, to sham training, or to no training. Only the trained group showed increased fractional anisotropy in several tracts, including the cingulum and the splenium of the corpus callosum. Marins et al. (2019) reported increased fractional anisotropy in the sensorimotor corpus callosum after a single session of fMRI neurofeedback. Sampaio-Baptista et al. (2021) then showed that callosal fractional anisotropy rose or fell depending on the direction in which participants were trained to modulate sensorimotor activity.

These studies are small, with roughly ten to twenty participants per group. Diffusion tensor imaging cannot say whether a change in fractional anisotropy reflects myelin, axon packing, or glial swelling. The atlas does not resolve that ambiguity. What it adds is a chemically detailed baseline of the tissue those measures sample, and that baseline varies by tract and by developmental history (Fusar Bassini et al., 2026).

Myelin has been treated as a uniform insulator, with a handful of oligodendrocyte types doing the same job everywhere. The atlas shows instead that white matter is chemically regionalized, that oligodendrocytes drive the pattern, and that part of the pattern is a developmental record. For a field whose training targets include the timing between regions, the substrate of that timing turns out to be far more varied than assumed.

Genes predict lipids only halfway

A reasonable expectation is that lipid patterns simply follow the genes that make lipids. The transcriptome is the full set of RNA copies a cell makes from its genes, and the expression of a lipid-making enzyme should, in principle, predict where its product appears. It did not. Across 338 enzyme-product pairs, the average correlation between enzyme expression and lipid abundance was slightly negative, and only 81 pairs correlated above 0.5 (Fusar Bassini et al., 2026).

Broader models did better but not well. A flexible model using the full transcriptome explained 56% of the variance that could, in principle, be explained once measurement noise is removed. In the cortex, the team found 14 lipizones within layer 5 of the barrel field that no gene-expression map had detected. The authors describe the brain lipidome as irreducible to the transcriptome, meaning that lipids carry information that gene maps cannot supply (Fusar Bassini et al., 2026).

La Manno (2026) extends this argument in the essay accompanying his 2026 Eppendorf and Science Prize. He frames lipids as central players in cell diversity and organization, not passive building blocks. Capolupo et al. (2022) supplied the strongest experimental support so far, showing that rewiring sphingolipid metabolism drove human skin fibroblasts from one cell state to another. No equivalent manipulation has yet been done in the brain. The claim that lipids shape neuronal identity therefore remains a hypothesis with good circumstantial support.

Why should a neurofeedback professional care about a statistical gap between genes and lipids? Because the gap is where physiology lives. Gene expression defines what a cell is, whereas lipid composition reflects what the cell has been doing and what hormones and signals it has received. A layer of brain organization that responds to state rather than identity is, by definition, a layer that experience can move.

If lipids merely followed the enzymes that make them, the atlas would be a convenient summary of existing gene maps. Instead, genes explain about half of what can be explained, and whole territories exist that gene maps cannot see. The unexplained half is the part of brain chemistry most open to hormones, activity, and experience.

Pregnancy rewrites the map in weeks

To test whether lipid maps change with physiology, the team compared three mid-gestation pregnant mice with three non-pregnant females and three males. Sex differences were modest, affecting only 29 of 222 supertypes. Pregnancy produced far larger changes, altering 139 of 215 supertypes that could be matched between groups. The ventricular system, isocortex, fiber tracts, and olfactory areas changed most, whereas the thalamus and hypothalamus changed least (Fusar Bassini et al., 2026).

The most consistent change involved myelin. A hexosylceramide, a sphingolipid carrying a single sugar group, rose in 100 supertypes, especially across white matter. The authors identify it as a galactosylceramide, a major myelin lipid, and confirmed the rise by conventional lipidomics of dissected corpus callosum from four animals per group. They interpret the pattern as regionally selective adaptive myelination (Fusar Bassini et al., 2026). This fits Gregg et al. (2007), who found that pregnant mice generate new oligodendrocytes and remyelinate lesions faster than virgin mice, an effect driven by prolactin.

The cortical finding is the one most likely to interest EEG specialists. Within the cortex, the most altered territories lay in cortical layer 4, the layer that receives the main sensory input from the thalamus. The single most affected lipizone spanned layer 4 of the visual and auditory cortex and showed 34 changed lipids. Four of them were increased species of phosphatidylserine, a lipid class that regulates vesicle docking at synapses. The authors suggest these changes may tune sensory circuits for detecting infant cues, a speculation they did not test behaviorally (Fusar Bassini et al., 2026).

The thalamocortical loop that passes through layer 4 is also the circuit most closely tied to the rhythms clinicians record. Hughes and Crunelli (2005) reviewed evidence that occipital alpha, the Rolandic mu rhythm, and related rhythms depend on thalamic pacemaker mechanisms. Steriade et al. (1993) described the same loop as the generator of sleep spindles. The atlas did not record EEG, so it does not support any claim about pregnancy and alpha. It does show that the membrane chemistry of the thalamorecipient layer is hormonally responsive, which is a testable prediction for human EEG research.

To find a cause, the team modeled which hormone receptors could best explain the lipid shifts. For the most affected layer 4 territory, the best fit ran through the glucocorticoid receptor, the protein that responds to cortisol in humans and corticosterone in mice. Single-nucleus RNA sequencing and immunostaining of cortex from two pregnant and two control mice showed that layer 4 neurons had turned on glucocorticoid-response genes. Two animals per group, no blinding, and pregnancy sections scanned after the controls make this a promising candidate rather than an established mechanism (Fusar Bassini et al., 2026).

Human pregnancy also reshapes the brain. Hoekzema et al. (2017) reported gray-matter volume reductions that persisted two years after delivery. Pritschet et al. (2024) found widespread cortical thinning alongside increased white-matter integrity in one woman tracked from preconception through two years postpartum. Neither study measured lipids. Together with the mouse atlas, they make a strong case that reproductive state is a variable worth recording in any EEG dataset.

Within weeks, pregnancy shifted more of the mouse lipid map than sex did across a lifetime. White matter gained a myelin lipid, and the thalamic input layer of sensory cortex was remodeled through a pathway that points to stress-hormone signaling. The mechanism rests on two animals per group and awaits a blocking experiment, but the larger lesson stands: the lipid substrate of brain function is hormonally responsive.

From mouse map to human brain

The closest human counterpart is smaller but points the same way. Osetrova et al. (2024) measured 419 lipids across 75 regions from four cognitively healthy adult brains. Composition differed between regions for 93% of the lipids measured. The strongest organizing factor was myelin content estimated by MRI, with 65% of lipids rising and 17% falling as myelin increased. Lipid profiles also tracked each cortical region's position in the brain's information-processing hierarchy.

Four brains cannot define normal human variation, and the study had no disease comparison. Human lipid chemistry also changes with age in sex-dependent ways. Couttas et al. (2018) measured sphingolipids in 80 postmortem hippocampus samples from neurologically normal adults over 65. Ceramide, sphingomyelin, and sulfatide content tracked age in men but not in women. Neither study included a disease group, so both describe normal variation rather than a marker.

These studies establish that the human lipidome, like the mouse lipidome, is regionally structured and dominated by myelin. The mouse atlas authors make the same point about translation: the specific map will not transfer to humans, but the method will. Spatial lipidomics of human postmortem tissue can now be read against a near-cellular reference (Fusar Bassini et al., 2026).

A four-brain human atlas and an 80-sample aging study do not add up to a clinical lipid test. They do establish that human brain lipids are regionally organized, myelin-dominated, and sex-dependent in aging. The mouse atlas supplies the resolution these human studies lack, and the human studies supply the relevance the mouse atlas cannot claim.

What lipids mean for the signal you record

White matter also shapes the EEG before any rhythm is interpreted. Myelinated fiber bundles conduct current more easily along their length than across it, a property called anisotropy. Source localization methods such as LORETA estimate where scalp potentials originate by solving a head model, a mathematical description of how current spreads through scalp, skull, and brain. Most clinical head models treat white matter as a uniform conductor (Vorwerk et al., 2014).

Simulation studies show what that simplification costs. Wolters et al. (2006) found that white-matter anisotropy redirects return currents to flow along fiber tracts, with effects that grow with source depth. Haueisen et al. (2002) concluded that anisotropy changes the topography of the scalp field only modestly but substantially changes its amplitude, so source strength is misestimated more than source position. Neugebauer et al. (2017) estimated that modeling anisotropy improves beamformer localization by 5 to 10 millimeters, a smaller gain than modeling cerebrospinal fluid but not a trivial one. The lipid atlas does not change those rankings, but it does make vivid that the simplified tissue is chemically patterned by tract, by development, and by hormonal state.

Lipid-rich white matter is not neutral to the EEG. It conducts current differently along fibers than across them, bending the fields that reach the scalp and shifting estimated sources. Its myelin sets the delays that decide whether two regions can synchronize. The atlas does not change how an amplifier works, but it does change how a clinician should picture the tissue between a generator and an electrode.

What this means in the clinic today

The plain answer is that nothing in the atlas should change a treatment plan this year. It offers no diagnostic threshold, no lipid biomarker for any disorder, and no evidence that any supplement, diet, or medication reshapes the lipizones it describes. Claims that omega-3 products or other lipid supplements act on the brain chemistry mapped here go beyond anything the study tested. Clinicians should treat such claims as marketing until human data exist.

The atlas does change the mental model. First, the timing structure that EEG measures rests on a chemically diverse and developmentally patterned myelin substrate, not a uniform insulator. Second, that substrate responds to hormonal state within weeks, which argues for recording pregnancy, postpartum status, and menopausal status in case notes and research samples. Third, the layer of brain organization that genes explain least is the layer that activity and experience are most likely to reach. Neurofeedback studies that already collect diffusion imaging can now ask whether training-related white-matter change tracks myelin-sensitive measures against a chemically defined reference.

For the practitioner, the atlas is background, not instruction. It supplies a baseline against which future human studies of disease, aging, medication, and training can be read. It gives three reasons to think differently about the tissue behind the electrode. It supplies no reason to sell a supplement or change a protocol.

Integrative summary

The 2026 lipid brain atlas maps 172 lipids into 539 territories that follow anatomy, trace connections, expose chemical diversity in white matter, and shift during pregnancy. Genes explain about half of the explainable lipid pattern, leaving a state-dependent layer that hormones and activity can plausibly move. For neurofeedback, the lesson is that the membranes and myelin beneath every trained rhythm are varied, organized, and plastic. Human studies will decide whether that layer becomes measurable in patients.

Five key takeaways

1. Brain membranes and myelin are mostly lipid, and the 2026 atlas is the first whole-brain, near-cellular map of where 172 of those lipids sit.

2. Lipizones follow anatomy and cell types, but they also follow axons to their targets, so connected regions share chemical signatures that gene maps miss.

3. White matter is chemically regionalized, driven by oligodendrocyte lipids and possibly recording the order in which tracts are myelinated during development.

4. Genes capture only about half of the explainable lipid variation, which marks lipids as a state-dependent layer that hormones, activity, and experience can reach.

5. Pregnancy remodeled the mouse lipid map more than sex did, raising a myelin lipid across white matter and reshaping the thalamic input layer of sensory cortex; human relevance and clinical use remain untested.

Glossary

activity-dependent myelination: The process by which neuronal firing recruits new oligodendrocytes and new myelin onto active axons.

Allen Brain Atlas Common Coordinate Framework: A shared three-dimensional reference space for the mouse brain into which different kinds of data can be aligned and compared.

anisotropy: The property of conducting current, or allowing water to diffuse, more easily in one direction than another; white matter is anisotropic along its fibers.

axon: The long extension of a neuron that carries action potentials away from the cell body toward other cells.

axon terminal: The end of an axon, where a neuron releases neurotransmitter onto another cell.

batch effect: A difference in measured values caused by processing or scanning conditions rather than biology.

cell membrane: The thin lipid boundary that separates a cell from its surroundings and houses its receptors and ion channels.

cell type: A category of cells that share characteristic structure, gene expression, and function.

cholesterol: A ring-shaped lipid that thickens and stiffens cell membranes and is required for myelin formation.

cluster analysis: A computational method that groups observations with similar measurements without using predefined labels.

coherence: A measure of the consistency of the phase relationship between EEG signals recorded at two sites within a frequency band.

conduction velocity: The speed at which an action potential travels along an axon, set largely by axon diameter and myelination.

corpus callosum: The largest bundle of myelinated axons in the brain, connecting the two cerebral hemispheres.

correlation: A statistical association between two measurements that does not by itself show that one causes the other.

cortical layer 4: The middle layer of the cerebral cortex that receives the main sensory input from the thalamus.

cytoarchitecture: The arrangement of cell bodies and fibers within brain tissue, used to define cortical layers and regional boundaries.

diffusion tensor imaging: An MRI method that measures the directional diffusion of water to characterize white-matter microstructure.

enzyme: A protein that speeds up a specific chemical reaction, such as a step in lipid synthesis.

fractional anisotropy: A diffusion tensor imaging index, from 0 to 1, of how strongly water diffusion is confined to one direction within a voxel.

galactosylceramide: A sphingolipid carrying a galactose sugar group; the most abundant glycolipid in myelin.

glucocorticoid receptor: A protein that binds cortisol in humans and corticosterone in rodents and then regulates the activity of many genes.

gray matter: Brain tissue rich in neuronal cell bodies, dendrites, and synapses.

head model: A mathematical description of how electrical current spreads through scalp, skull, cerebrospinal fluid, and brain, used to estimate EEG sources.

hexosylceramide: A sphingolipid carrying a single sugar group; galactosylceramide is the main brain example.

immunostaining: A method that uses antibodies to mark specific proteins in tissue so that their location and amount can be seen.

ion channel: A membrane protein that opens a pore for specific ions; the currents through ion channels generate neuronal signals and the EEG.

lipid: A fat-like molecule that dissolves poorly in water; lipids form membranes, store energy, and act as signals.

lipidomics: The identification and measurement of the lipid molecules in a sample.

lipizone: A brain territory defined only by its local pattern of lipid composition.

MALDI-MSI: Matrix-assisted laser desorption/ionization mass spectrometry imaging, a method that uses a laser and a chemical coating to release and charge molecules from a tissue slice, then maps each molecule by its position.

myelin: The lipid-rich insulating sheath that oligodendrocytes wrap around axons, speeding conduction.

oligodendrocyte: The glial cell that produces myelin in the brain and spinal cord.

phosphatidylserine: A negatively charged phospholipid concentrated on the inner face of membranes that regulates vesicle docking and release.

phospholipid: A lipid with a water-attracting head and two water-avoiding tails that forms the two-layer sheet of every cell membrane.

pixel: One sampled location in an image; in the atlas, one 5-micrometer laser spot.

prolactin: A pituitary hormone that rises during pregnancy and lactation and promotes new oligodendrocytes in rodents.

sharp-wave ripples: Brief high-frequency hippocampal oscillations during rest and sleep that replay recent experience and support memory consolidation.

single-nucleus RNA sequencing: A method that reads the RNA inside individual cell nuclei to compare gene activity across cell populations.

sleep spindles: Bursts of 11 to 16 Hz activity during non-REM sleep generated by thalamocortical circuits.

source localization: Estimation of the brain locations that generate a recorded scalp EEG distribution.

sphingolipid: A family of lipids built on a sphingosine backbone, concentrated in myelin and in signaling domains of membranes.

thalamus: A paired group of nuclei that relays sensory and other information to the cerebral cortex and participates in cortical rhythms.

transcriptome: The full set of RNA molecules produced by a cell or tissue.

uMAIA: Unified Mass Imaging Analyzer, a computational framework for aligning and combining large mass spectrometry imaging datasets.

white matter: Brain tissue made mostly of myelinated axon bundles.

References

Capolupo, L., Khven, I., Lederer, A. R., Mazzeo, L., Glousker, G., Ho, S., Russo, F., Paz Montoya, J., Bhandari, D. R., Bowman, A. P., Ellis, S. R., Guiet, R., Burri, O., Detzner, J., Müthing, J., Homicsko, K., Kuonen, F., Gilliet, M., Spengler, B., . . . D'Angelo, G. (2022). Sphingolipids control dermal fibroblast heterogeneity. Science, 376(6590), Article eabh1623. https://doi.org/10.1126/science.abh1623

Couttas, T. A., Kain, N., Tran, C., Chatterton, Z., Kwok, J. B., & Don, A. S. (2018). Age-dependent changes to sphingolipid balance in the human hippocampus are gender-specific and may sensitize to neurodegeneration. Journal of Alzheimer's Disease, 63(2), 503–514. https://doi.org/10.3233/JAD-171054

Fusar Bassini, L., Schede, H. H., Capolupo, L., Alieh, L. H. A., Kaysudu, I., Venturi, F., Hochgerner, H., Valente, A., Droin, C., Trejo Banos, D., Khven, I., Maillat, J. A., Mahul-Mellier, A.-L., Asaro, A., Ülgen, D. H., Barahtjan, P., Asirim, E. Z., Nasrallah, A., Sandi, C., . . . La Manno, G. (2026). The lipidomic architecture of the mouse brain. Nature, 658(8134), 205–217. https://doi.org/10.1038/s41586-026-11050-0

Ghaziri, J., Tucholka, A., Larue, V., Blanchette-Sylvestre, M., Reyburn, G., Gilbert, G., Lévesque, J., & Beauregard, M. (2013). Neurofeedback training induces changes in white and gray matter. Clinical EEG and Neuroscience, 44(4), 265–272. https://doi.org/10.1177/1550059413476031

Gibson, E. M., Purger, D., Mount, C. W., Goldstein, A. K., Lin, G. L., Wood, L. S., Inema, I., Miller, S. E., Bieri, G., Zuchero, J. B., Barres, B. A., Woo, P. J., Vogel, H., & Monje, M. (2014). Neuronal activity promotes oligodendrogenesis and adaptive myelination in the mammalian brain. Science, 344(6183), Article 1252304. https://doi.org/10.1126/science.1252304

Gregg, C., Shikar, V., Larsen, P., Mak, G., Chojnacki, A., Yong, V. W., & Weiss, S. (2007). White matter plasticity and enhanced remyelination in the maternal CNS. The Journal of Neuroscience, 27(8), 1812–1823. https://doi.org/10.1523/JNEUROSCI.4441-06.2007

Haueisen, J., Tuch, D. S., Ramon, C., Schimpf, P. H., Wedeen, V. J., George, J. S., & Belliveau, J. W. (2002). The influence of brain tissue anisotropy on human EEG and MEG. NeuroImage, 15(1), 159–166. https://doi.org/10.1006/nimg.2001.0962

Hoekzema, E., Barba-Müller, E., Pozzobon, C., Picado, M., Lucco, F., García-García, D., Soliva, J. C., Tobeña, A., Desco, M., Crone, E. A., Ballesteros, A., Carmona, S., & Vilarroya, O. (2017). Pregnancy leads to long-lasting changes in human brain structure. Nature Neuroscience, 20(2), 287–296. https://doi.org/10.1038/nn.4458

Hughes, S. W., & Crunelli, V. (2005). Thalamic mechanisms of EEG alpha rhythms and their pathological implications. The Neuroscientist, 11(4), 357–372. https://doi.org/10.1177/1073858405277450

La Manno, G. (2026). Fats that think? Science, 394(6823), 499–500. https://doi.org/10.1126/science.ael6217

Levitan, I., Fang, Y., Rosenhouse-Dantsker, A., & Romanenko, V. (2010). Cholesterol and ion channels. In J. R. Harris (Ed.), Cholesterol binding and cholesterol transport proteins: Structure and function in health and disease (pp. 509–549). Springer. https://doi.org/10.1007/978-90-481-8622-8_19

Marins, T., Rodrigues, E. C., Bortolini, T., Melo, B., Moll, J., & Tovar-Moll, F. (2019). Structural and functional connectivity changes in response to short-term neurofeedback training with motor imagery. NeuroImage, 194, 283–290. https://doi.org/10.1016/j.neuroimage.2019.03.027

Marques, S., Zeisel, A., Codeluppi, S., van Bruggen, D., Mendanha Falcão, A., Xiao, L., Li, H., Häring, M., Hochgerner, H., Romanov, R. A., Gyllborg, D., Muñoz-Manchado, A. B., La Manno, G., Lönnerberg, P., Floriddia, E. M., Rezayee, F., Ernfors, P., Arenas, E., Hjerling-Leffler, J., . . . Castelo-Branco, G. (2016). Oligodendrocyte heterogeneity in the mouse juvenile and adult central nervous system. Science, 352(6291), 1326–1329. https://doi.org/10.1126/science.aaf6463

McKenzie, I. A., Ohayon, D., Li, H., Paes de Faria, J., Emery, B., Tohyama, K., & Richardson, W. D. (2014). Motor skill learning requires active central myelination. Science, 346(6207), 318–322. https://doi.org/10.1126/science.1254960

Neugebauer, F., Möddel, G., Rampp, S., Burger, M., & Wolters, C. H. (2017). The effect of head model simplification on beamformer source localization. Frontiers in Neuroscience, 11, Article 625. https://doi.org/10.3389/fnins.2017.00625

Noori, R., Park, D., Griffiths, J. D., Bells, S., Frankland, P. W., Mabbott, D., & Lefebvre, J. (2020). Activity-dependent myelination: A glial mechanism of oscillatory self-organization in large-scale brain networks. Proceedings of the National Academy of Sciences, 117(24), 13227–13237. https://doi.org/10.1073/pnas.1916646117

O'Brien, J. S., & Sampson, E. L. (1965). Lipid composition of the normal human brain: Gray matter, white matter, and myelin. Journal of Lipid Research, 6(4), 537–544. https://doi.org/10.1016/S0022-2275(20)39619-X

Osetrova, M., Tkachev, A., Mair, W., Guijarro Larraz, P., Efimova, O., Kurochkin, I., Stekolshchikova, E., Anikanov, N., Foo, J. C., Cazenave-Gassiot, A., Mitina, A., Ogurtsova, P., Guo, S., Potashnikova, D. M., Gulin, A. A., Vasin, A. A., Sarycheva, A., Vladimirov, G., Fedorova, M., . . . Khaitovich, P. (2024). Lipidome atlas of the adult human brain. Nature Communications, 15, Article 4455. https://doi.org/10.1038/s41467-024-48734-y

Pajevic, S., Basser, P. J., & Fields, R. D. (2014). Role of myelin plasticity in oscillations and synchrony of neuronal activity. Neuroscience, 276, 135–147. https://doi.org/10.1016/j.neuroscience.2013.11.007

Pritschet, L., Taylor, C. M., Cossio, D., Faskowitz, J., Santander, T., Handwerker, D. A., Grotzinger, H., Layher, E., Chrastil, E. R., & Jacobs, E. G. (2024). Neuroanatomical changes observed over the course of a human pregnancy. Nature Neuroscience, 27(11), 2253–2260. https://doi.org/10.1038/s41593-024-01741-0

Saher, G., Brügger, B., Lappe-Siefke, C., Möbius, W., Tozawa, R., Wehr, M. C., Wieland, F., Ishibashi, S., & Nave, K.-A. (2005). High cholesterol level is essential for myelin membrane growth. Nature Neuroscience, 8(4), 468–475. https://doi.org/10.1038/nn1426

Sampaio-Baptista, C., Neyedli, H. F., Sanders, Z.-B., Diosi, K., Havard, D., Huang, Y., Andersson, J. L. R., Lühr, M., Goebel, R., & Johansen-Berg, H. (2021). fMRI neurofeedback in the motor system elicits bidirectional changes in activity and in white matter structure in the adult human brain. Cell Reports, 37(4), Article 109890. https://doi.org/10.1016/j.celrep.2021.109890

Schede, H. H., Alieh, L. H. A., Rohde, L. A., Herrera, A., Schlaeppi, A., Valentin, G., Gargoori Motlagh, A., Dominguez Mantes, A., Jollivet, C., Paz-Montoya, J., Capolupo, L., Khven, I., Oates, A. C., D'Angelo, G., & La Manno, G. (2025). Unified mass imaging maps the lipidome of vertebrate development. Nature Methods, 22(9), 1981–1994. https://doi.org/10.1038/s41592-025-02771-7

Steadman, P. E., Xia, F., Ahmed, M., Mocle, A. J., Penning, A. R. A., Geraghty, A. C., Steenland, H. W., Monje, M., Josselyn, S. A., & Frankland, P. W. (2020). Disruption of oligodendrogenesis impairs memory consolidation in adult mice. Neuron, 105(1), 150–164. https://doi.org/10.1016/j.neuron.2019.10.013

Steriade, M., McCormick, D. A., & Sejnowski, T. J. (1993). Thalamocortical oscillations in the sleeping and aroused brain. Science, 262(5134), 679–685. https://doi.org/10.1126/science.8235588

Vorwerk, J., Cho, J.-H., Rampp, S., Hamer, H., Knösche, T. R., & Wolters, C. H. (2014). A guideline for head volume conductor modeling in EEG and MEG. NeuroImage, 100, 590–607. https://doi.org/10.1016/j.neuroimage.2014.06.040

Wolters, C. H., Anwander, A., Tricoche, X., Weinstein, D., Koch, M. A., & MacLeod, R. S. (2006). Influence of tissue conductivity anisotropy on EEG/MEG field and return current computation in a realistic head model: A simulation and visualization study using high-resolution finite element modeling. NeuroImage, 30(3), 813–826. https://doi.org/10.1016/j.neuroimage.2005.10.014