The research team, led by Nana Matoba, Jessica McAfee, Hyejung Won and Jason Stein at the University of North Carolina at Chapel Hill, deployed a massively parallel reporter assay, or MPRA, to interrogate 9,092 genetic variants associated with human cortical structure. The logic of the assay is elegant. Each candidate variant is synthesized as a short 150-base-pair segment of DNA, cloned into a plasmid upstream of a minimal promoter that drives a luciferase reporter gene, and tagged with a random 20-base-pair barcode. When the pooled library is introduced into cells, the amount of barcode RNA produced by each construct reflects the regulatory power of the DNA segment attached to it. By sequencing both the RNA transcripts and the plasmid DNA, researchers can calculate a transcription rate for every element, and by testing both alleles of a variant side by side, they can detect whether one version of the sequence drives expression more strongly than the other.

The cellular context matters enormously here. The team performed the assay in primary human neural progenitor cells, the fetal-derived stem cells that generate the neurons of the developing cortex. Previous work had shown that variants associated with cortical structure are enriched in regulatory regions active in these progenitors and colocalize with genetic effects on chromatin accessibility and gene expression in the same cells. This supports a model in which noncoding variants act during prenatal development, altering transcription factor binding and gene expression in progenitor cells, and thereby influencing decisions about cell proliferation and fate that ultimately determine cortical size in adulthood. The researchers transfected the library into 16 biological replicates, and after 24 hours split the cultures: half received a vehicle control, while half were treated with CHIR99021, a potent inhibitor of GSK3β that activates the Wnt signaling pathway, a key developmental signal known to influence brain size.

The results were striking in their scale and reproducibility. Regulatory activity measurements correlated between replicates with a median Pearson’s coefficient of 0.99, and the positive control promoters behaved exactly as expected. Among 17,837 elements tested under baseline conditions, 988 showed regulatory activity significantly above the negative controls, representing 918 variants drawn from 150 of the 198 cortical-structure-associated loci examined—meaning that 76 percent of the loci contained at least one variant with detectable regulatory function. More than half of the active variants also showed allelic effects, meaning one DNA letter change measurably altered transcription. The team dubbed these expression-modulating variants, or emVars, and they represent the strongest candidates for the true causal variants hiding within blocks of correlated genetic variation that have long frustrated geneticists.

Perhaps the most surprising finding came when the researchers asked what sequence features distinguished the active elements. Active elements were dramatically enriched in short interspersed nuclear elements, or SINEs—a class of retrotransposons, the so-called jumping genes that copy and paste themselves throughout the genome. Within the SINE family, the enrichment was driven almost entirely by Alu elements, primate-specific repeats that have expanded in prevalence alongside increases in brain size over human evolution. A typical Alu element carries two conserved sequences in its left arm, the A box and the B box, which function as promoters for RNA polymerase III. The most active MPRA elements sat precisely in this left arm, and elements containing both boxes showed the highest activity of all. Variants with the strongest allelic effects clustered near these promoter boxes, and even a scrambled negative control that accidentally contained a B-box-like sequence showed strong activity—a striking confirmation of the mechanism.

Skeptics might worry that such activity is an artifact of the episomal plasmid system, in which DNA is not packaged into chromatin as it would be in the native genome. The team addressed this concern with multiple lines of endogenous evidence. Fetal brain tissue shows among the lowest levels of repressive chromatin marks, such as H3K9me3 and H3K27me3, deposited on Alu elements of any tissue in the body, and the MPRA-active Alu elements are rarely repressed in fetal brain specifically. The active elements were also enriched for enhancer RNAs and promoter upstream antisense RNAs, noncoding transcripts previously shown to mediate enhancer-promoter communication through Alu sequences. Finally, using CUT&RUN profiling, the researchers showed that MPRA-active Alu elements are significantly enriched within binding sites of TFIIIC, a transcription factor complex associated with polymerase III and Alu activity. Together, these data suggest that the reporter activity reflects genuine regulatory potential in the developing brain.

The evolutionary implications are tantalizing. Alu elements can be dated by their accumulated mutations, measured as sequence divergence from the ancestral consensus. The younger Alu elements—those with fewer than 100 mismatches per thousand bases—showed significantly higher regulatory activity than older ones, particularly when located in the left arm of the element. Because Alu insertions have proliferated in the primate lineage in step with brain size expansion, the findings suggest that relatively recent retrotransposon insertions created new regulatory elements that influenced neurodevelopmental genes and may have contributed to the expansion of the human cortex. The authors caution that a neutral decay of regulatory potential in older elements over evolutionary time could also explain the pattern, but the convergence with known primate-specific biology makes the adaptive interpretation compelling.

The study also connected individual variants to specific transcription factors and, remarkably, to specific brain regions. Active elements were enriched for binding motifs of three transcription factor clusters—ZNF135/ZNF460, MEF2A-D and Zfx—and the magnitude of a variant’s allelic effect correlated strongly with how severely it disrupted a predicted motif. Disruption of RARA, RARG, ZNF135 and NR2F1 motifs produced the largest impacts. Intriguingly, the consensus sequences of recently evolved Alu subfamilies contain retinoic acid receptor binding sites, linking the transposon story to developmental signaling. Even more strikingly, when the team compared the regional pattern of a variant’s effect on cortical structure with the expression of the disrupted transcription factor across prenatal brain regions from the BrainSpan Atlas, they found significant correlations for seven variants. A variant that disrupts the RARG binding motif, for example, reduced surface area most strongly in regions where RARG expression was highest—suggesting that the regional specificity of genetic effects is governed by where the relevant transcription factors are actually present.

Wnt stimulation added another layer of complexity. Under CHIR treatment, the team detected a similar number of active elements, but 32 elements gained activity specifically under stimulation while seven lost it, and 12 variants showed allelic effects that differed significantly between conditions. One example, rs4670759, associated with the surface area of the inferior temporal cortex, showed enhanced activity and increased chromatin accessibility under Wnt stimulation; it sits in the promoter of CDC42EP3, a gene involved in actin cytoskeleton formation whose expression rises with Wnt activation, and the risk allele is predicted to disrupt a PLAGL2 motif that is itself upregulated by the pathway. Another variant, rs2802295 within an intron of FOXO3, showed an allelic effect only under stimulation near a Wnt-responsive element. The team then used CRISPR interference to silence that element in progenitor cells and confirmed that FOXO3 expression dropped—experimentally tying a cortical-structure-associated variant to its target gene in a condition-specific manner.

The study is not without limitations. The MPRA measures regulatory potential on episomal plasmids, and the overlap with endogenous chromatin accessibility quantitative trait loci was limited, indicating that native chromatin context modifies some effects. The analysis also focused on variants tested primarily in populations of European ancestry and did not address sex-dependent effects. Nevertheless, by experimentally separating correlated variants, identifying the transcription factors they disrupt, and demonstrating condition-dependent function, the work delivers a genome-scale functional framework for one of the most complex traits in human neuroimaging genetics. It transforms a catalog of statistical associations into a mechanistic hypothesis space—one in which ancient jumping genes, developmental signaling pathways and spatially patterned transcription factors jointly sculpt the most distinctive feature of the human brain.

Subject of Research: Functional characterization of noncoding genetic variants associated with human cortical structure using massively parallel reporter assays in neural progenitor cells

Article Title: Massively parallel assessment of gene regulatory activity at human cortical-structure-associated variants

Article References: Matoba, N., McAfee, J. C., Krupa, O., Beltran, A. A., Bell, J. L., Le, B. D., Valone, J. M., Min, H., Crawford, G. E., Raab, J. R., Won, H., & Stein, J. L. (2026). Massively parallel assessment of gene regulatory activity at human cortical-structure-associated variants. Nature Neuroscience. https://doi.org/10.1038/s41593-026-02454-2

Image Credits: AI Generated

DOI: 10.1038/s41593-026-02454-2

Keywords: cortical structure, massively parallel reporter assay, noncoding variants, Alu elements, neural progenitor cells, Wnt signaling, transcription factors, ENIGMA, gene regulation, brain development, retrotransposons, CRISPRi

The letter, authored by Bin Huang of the Department of Rehabilitation Medicine and Yuqian Shen of the Traditional Chinese Medicine Department at Suzhou Ninth People’s Hospital in Suzhou, China, takes aim at the growing practice of interpreting slow-wave slope measurements in patients with obstructive sleep apnea as direct markers of synaptic homeostasis. Writing in response to a recent population-based cohort study that reported sleep homeostasis impairment across the severity spectrum of obstructive sleep apnea, Huang and Shen do not dispute the clinical importance of the findings. Instead, they raise a methodological concern with potentially wide consequences: the electrophysiological signal that researchers are reading as a gauge of synaptic strength is entangled with multiple brain-altering processes in sleep apnea that have nothing to do with synaptic downscaling, and may therefore distort the measurement in ways that are difficult to disentangle.

The theoretical appeal of the slow-wave slope is easy to understand. Grounded in the synaptic homeostasis framework developed by Giulio Tononi and Chiara Cirelli, the reasoning holds that waking life potentiates synapses across the cortex, increasing the energetic cost of neural activity and saturating the capacity for further learning. Sleep, particularly slow-wave sleep, is proposed to renormalize this synaptic burden. High-density electroencephalography studies in humans, including the landmark work of Riedner and colleagues, demonstrated that the slope and amplitude of individual slow waves decline systematically across a night of sleep, tracking the dissipation of homeostatic sleep pressure. Because slope is less sensitive than amplitude to volume conduction and reference-electrode artifacts, many investigators consider it the cleaner measure of the underlying cortical synchronization, and by extension, of the synaptic substrate that generates it.

Obstructive sleep apnea, however, is not a neutral backdrop against which this synaptic story unfolds. The disorder imposes a cascade of stressors on the cortex that are capable of changing slow-wave morphology in their own right. Chronic intermittent hypoxia, repeated arousals that shatter the continuity of slow-wave sleep, blood pressure surges, carbon dioxide fluctuations, and systemic inflammation all accompany the apneic night. Beyond these acute disturbances, imaging work by Macey, Kumar, Harper and colleagues has documented structural changes in the brains of patients with obstructive sleep apnea, including alterations in gray matter volume in regions responsible for generating and propagating slow oscillations. If the cortical tissue itself has been reshaped by years of nocturnal suffocation, Huang and Shen argue, then any observed flattening of slow-wave slope in these patients could reflect neuronal loss, cortical thinning, or impaired network connectivity rather than a genuine deficit in the synaptic downscaling process that homeostatic theory describes.

This distinction matters because the two explanations lead to different conclusions about what is broken in the apneic brain. A flattened slow-wave slope interpreted as impaired synaptic homeostasis suggests that the fundamental sleep-dependent mechanism of synaptic renormalization is failing, with downstream consequences for memory consolidation, synaptic pruning, and cognitive resilience. The same measurement interpreted as a byproduct of structural brain injury or hypoxia-related neuronal dysfunction points instead toward damage that may or may not respond to restoring normal sleep architecture with continuous positive airway pressure therapy. Treatment studies that use slow-wave slope as an outcome measure would, in the second scenario, risk drawing misleading inferences about whether therapy has revived a dormant homeostatic process when the slope change might instead track the partial recovery of injured cortical networks or the resolution of hypoxic stress.

The letter also highlights problems that arise even before the brain’s pathophysiology enters the picture. Slow-wave detection and slope quantification are sensitive to the algorithms used, the electroencephalographic derivations examined, the thresholds for event detection, and the manner in which artifacts from respiration, movement, and arousal are handled. Patients with obstructive sleep apnea generate unusually noisy sleep recordings: respiratory artifacts contaminate the signal, arousals fragment epochs, and stage shifts compress the amount of analyzable slow-wave sleep. Population-based studies, which are invaluable for capturing the full severity spectrum of the disorder, often rely on limited electroencephalographic montages that may not sample the cortical regions where slow waves originate or where apnea-related damage concentrates. Slope estimates drawn from a small number of derivations may therefore not generalize to the whole-cortex synchronization dynamics that the homeostatic framework actually describes.

Compounding these technical issues is the question of what the slope is being compared against. The canonical decline of slow-wave slope across the night was established in healthy, typically young adults sleeping undisturbed. In a patient whose slow-wave sleep is repeatedly interrupted, the expected trajectory of homeostatic dissipation may be altered simply because sleep pressure dynamics themselves are disrupted by fragmentation, independent of any change in synaptic machinery. A lower slope at a given point in the night could indicate faster dissipation, impaired build-up of sleep pressure during the preceding wake period, or a cortex that synchronizes differently because its cells and connections have been remodeled by chronic disease. Without careful controls for apnea severity, oxygen desaturation burden, arousal frequency, age, and comorbid conditions, the letter contends, attributing slope differences specifically to synaptic homeostasis overreaches what the data can support.

None of this means that slow-wave slope should be abandoned as an investigative tool. Huang and Shen are careful to frame their argument as a call for methodological caution rather than a rejection of the measure. Slow-wave slope remains one of the most informative electrophysiological indices available in human sleep research, and the population-based evidence that sleep homeostasis is altered in obstructive sleep apnea is clinically significant in its own right, independent of the synaptic interpretation attached to it. The authors’ concern is that the field is at risk of building an explanatory edifice on a signal whose meaning in this patient population is confounded at multiple levels. The remedy they imply is methodological pluralism: combining slope measurements with complementary approaches such as high-density EEG mapping, measures of sleep pressure built from slow-wave activity, neuroimaging of cortical structure, and longitudinal designs that track patients before and after effective treatment of their breathing disorder.

For clinicians and researchers following the rapidly expanding literature on sleep apnea and brain health, the message is a sobering one. Obstructive sleep apnea affects hundreds of millions of people worldwide and is increasingly linked to cognitive decline, dementia risk, and impaired emotional regulation. Slow-wave sleep sits at the center of many proposed mechanisms connecting the disorder to these outcomes, and slow-wave slope has become a shorthand for the health of the sleeping brain. The letter from Suzhou reminds the field that a shorthand is not a mechanism. Until the confounding contributions of hypoxia, fragmentation, cortical structural change, and measurement methodology are systematically separated, the slope of a slow wave in a patient with sleep apnea should be read as an intriguing and imperfect signal, not as a direct readout of synaptic homeostasis. The nightly downscaling of synapses may well be impaired in obstructive sleep apnea—but proving it will require more than reading the steepness of the brain’s deepest waves.

Subject of Research: Methodological limitations of using EEG slow-wave slope as a marker of synaptic homeostasis in obstructive sleep apnea

Article Title: Methodological caveats in interpreting slow-wave slope as a marker of synaptic homeostasis in OSA

Article References: Huang, B., & Shen, Y. (2026). Methodological caveats in interpreting slow-wave slope as a marker of synaptic homeostasis in OSA. Journal of Clinical Sleep Medicine, 22(1), Article 168. https://doi.org/10.1007/s44470-026-00189-2

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00189-2

Keywords: obstructive sleep apnea, slow-wave slope, synaptic homeostasis hypothesis, slow-wave sleep, EEG, sleep homeostasis, intermittent hypoxia, synaptic downscaling, sleep fragmentation, cortical structure, neurophysiology, sleep medicine