The research team, led by Lei Shi of the Kunming Institute of Zoology at the Chinese Academy of Sciences, exploited an unusual comparative tool: the tree shrew. This small, squirrel-like mammal is not a primate, but it sits close enough to the primate branch on the evolutionary tree to serve as an outgroup, a reference point that allows scientists to distinguish what is genuinely primate-specific from what is shared more broadly among mammals. By comparing the developing brains of humans, macaques, tree shrews and mice, the researchers could trace which genetic changes coincided with the expansion of cortical territory that defines our order.

Their first discovery was that CCNB1IP1, a gene that had little or no activity in the developing brains of other mammals, acquired expression in the primate brain. More strikingly, the level of that expression climbed progressively across the evolutionary ladder, from lower primates through monkeys to apes and finally humans. That gradient mirrored the gradient in cortical size and gyrification, the technical term for the degree of cortical folding, suggesting a quantitative relationship rather than a simple on-off switch. Statistical analysis confirmed that both the gyrification index and brain weight correlate with evolutionary divergence time across species, consistent with a gradual, cumulative process rather than a sudden leap.

Where did the increased expression come from? The answer, according to the study, lies not in the protein-coding sequence of the gene itself but in its regulatory DNA. Comparative analysis of cis-regulatory elements, the short stretches of DNA near a gene that control when and how strongly it is transcribed, revealed evolutionary changes that ratcheted up CCNB1IP1 activity. The researchers focused on two candidate regulatory regions, CRE1 and CRE2, and examined histone modifications such as H3K4me3 and H3K27ac, chemical tags on chromatin that mark active regulatory elements in human neural progenitor cells. Sequence alignments across twenty mammalian species highlighted sites that changed specifically in the human lineage, and one region, CRE1, contained sub-elements associated with the gradual expression increase during primate evolution.

To test whether the gene could actually reshape a brain, the team turned to experiment. When they overexpressed CCNB1IP1 in the embryonic mouse cortex using in utero electroporation, they saw a cascade of cellular effects: more cells entering DNA synthesis, an expanded pool of basal progenitors, and increased neuronal output. Basal progenitors are the self-amplifying stem cells of the subventricular zone, a germinal layer that is disproportionately well developed in primates and is widely regarded as the engine of cortical expansion. In the human fetal cortex, single-cell analysis showed that CCNB1IP1 is expressed most strongly in neural progenitor cells, and higher in basal progenitors than in apical progenitors at the peak of neurogenesis, precisely where extra rounds of cell division would matter most.

The mechanistic heart of the paper concerns what CCNB1IP1 actually does inside those cells. The researchers demonstrated that it represses E2F1, a transcription factor that drives the cell cycle, by targeting it for ubiquitin-mediated degradation, the cellular waste-disposal pathway that tags proteins for destruction by the proteasome. By damping E2F1, CCNB1IP1 alters the tempo of the cell cycle in neural progenitors. Slower, differently timed cycles are a long-standing theme in cortical evolution: since the classic work of Pasko Rakic and others, developmental biologists have argued that small changes in the duration and timing of progenitor cell divisions, repeated over thousands of rounds, can yield enormous differences in final neuron number. CCNB1IP1 offers a concrete molecular handle on that process.

The most dramatic experiment came next. Using CRISPR-Cas9, the team created conditional knock-in mice carrying the human CCNB1IP1 gene at the Rosa26 locus, activated in the excitatory neuronal lineage via Emx1-Cre. Ordinary mice have smooth, lissencephalic brains; their cortices lack the folds seen in humans and other gyrencephalic species. Remarkably, the knock-in mice developed recognizable cortical folding, along with an expanded cortical area, an increased proportion of upper-layer neurons, and measurable gains in cognitive performance on behavioral tests such as novel object recognition and the Morris water maze. In effect, a single regulatory change was sufficient to push a rodent brain partway toward a primate-like architecture.

The findings fit into a broader and increasingly rich picture of human brain evolution. Previous work has identified other human-specific genetic innovations that expand cortical progenitors, including the NOTCH2NL gene family, which enhances Notch signaling, and ARHGAP11B, a human-specific gene that promotes basal progenitor amplification and can even induce folding in mouse embryos. What distinguishes the new study is its emphasis on gradualism. Rather than a single human-specific invention, CCNB1IP1 shows a stepped increase in expression across the primate tree, implying that the enlargement and folding of the cortex were built incrementally, species by species, through cumulative regulatory tweaks rather than one dramatic mutation.

It is also a study in methodological craft. The authors combined laser microdissection of individual cortical layers in tree shrew embryos, bulk and single-cell transcriptomics across species, comparative genomics of regulatory elements, protein interaction assays, and transgenic mouse modeling into a single coherent argument. Data from brain organoids derived from humans, chimpanzees, bonobos and gorillas reinforced the expression gradient, and postnatal expression data from public brain atlases showed that the gene’s activity persists across multiple brain regions during development. All sequencing data have been deposited in public repositories, allowing other groups to scrutinize and extend the results.

Caveats remain, as they always do. Correlation between gene expression and cortical complexity across species does not by itself prove causation for every step of primate evolution, and the knock-in mouse, while striking, models only one component of a process that surely involved many genes. CCNB1IP1 has also been implicated in other contexts, including cancer biology, where it modulates ubiquitination of proteins such as MYCN, so its functions are likely broader than neurodevelopment alone. Still, the study delivers something rare: a gene whose evolutionary upregulation tracks the growth of the primate cortex, a plausible molecular mechanism linking it to progenitor behavior, and proof that boosting it can fold a smooth brain. In the long story of how our ancestors’ minds swelled and wrinkled into their modern form, CCNB1IP1 now stands as one of the clearest characters yet identified.

Subject of Research: Evolutionary changes in CCNB1IP1 gene regulation and their role in primate cortical expansion and folding

Article Title: CCNB1IP1 regulatory evolution underlies gradual increases in cortical size and folding in primates

Article References: Hu, T., Ma, P., Kong, Y., Tan, Y., Sun, X., Wang, J., Xiang, K., Mao, B., Wu, Q., Yi, S. V., & Shi, L. (2026). CCNB1IP1 regulatory evolution underlies gradual increases in cortical size and folding in primates. Nature Genetics. https://doi.org/10.1038/s41588-026-02773-x

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02773-x

Keywords: CCNB1IP1, primate evolution, cortex, cortical folding, neurogenesis, basal progenitors, cis-regulatory elements, E2F1, ubiquitin degradation, tree shrew, knock-in mice, brain evolution

One of the most striking exceptions is the common marmoset (Callithrix jacchus), a small South American monkey that has become an increasingly important model organism in biomedical research. Despite being a genuine primate, the marmoset carries a brain that is smooth and virtually unfolded, a feature that sets it apart from the majority of its relatives. For developmental neuroscientists, this raises an obvious and compelling question: if marmosets and humans descend from a shared primate lineage, what cellular changes during embryonic development cause the marmoset brain to grow less, remain smaller, and never acquire the pronounced folds seen in other species?

A team at the German Primate Center (DPZ) — the Leibniz Institute for Primate Research in Göttingen — set out to answer exactly that question. In a study published in Science Advances, the researchers traced the developmental divergence between smooth and folded primate brains back to specific differences in neural progenitor cells, the self-renewing precursors from which all neurons arise. Their central finding is that the marmoset brain does not start out different. In early embryogenesis it displays the typical structure and composition of a large, folded primate brain, and only later do particular processes intervene to slow the production of nerve cells.

“We wanted to understand which changes at the cellular level cause the marmoset brain to grow less and form fewer folds,” explains Lidiia Tynianskaia, one of the study’s two first authors and a PhD student in the Junior Research Group Brain Development and Evolution at the DPZ. “At the beginning of development, the common marmoset brain exhibits the typical structure and composition of a large, folded primate brain. As development progresses, processes must therefore occur that effectively slow down the production of nerve cells.” That framing matters, because it reframes the smooth marmoset brain not as a primitive starting point but as the outcome of active developmental regulation imposed partway through embryogenesis.

To investigate those processes, the researchers turned to brain organoids — small, three-dimensional cell cultures grown from stem cells that recapitulate key aspects of early brain development. Comparative organoid experiments using marmoset and human cells allowed the team to focus on the brain’s progenitor cells, the population whose behavior largely determines how many neurons the cortex will ultimately contain. Since neuron number is one of the decisive factors governing both the size of a brain and the degree to which its surface folds, differences in progenitor behavior are a natural place to look for the roots of anatomical divergence.

The comparisons revealed that several distinct mechanisms, operating at different levels and in different progenitor cell types, converge to reduce the size and folding of the marmoset cerebral cortex. “Our investigations have shown that certain progenitor cells in the common marmoset divide significantly more slowly than in humans,” says César Mateo Bastidas Betancourt, the study’s other first author and also a PhD student in the Junior Research Group. “Other progenitor cells have a simpler structure than their human counterparts, with fewer processes, and are therefore less proliferative. Both of these factors ultimately result in fewer nerve cells, which contributes to a smaller size and less folding of the cerebral cortex in marmosets.”

These two mechanisms are complementary rather than redundant. Slower cell division directly limits how quickly the progenitor pool expands, while a simpler morphology — fewer of the cellular processes that in more proliferative progenitors help sustain repeated rounds of division — appears to constrain the cells’ proliferative capacity itself. In folded primate brains, progenitor cells in the developing neocortex typically undergo many rounds of division, amplifying the pool of neurons that will populate the expanding cortical sheet. When that amplification is dialed down, as it is in marmosets, the downstream consequence is a thinner, smaller cortex with less surface area to fold.

Timing emerged as a third, crucial dimension of the story. The scientists showed that certain characteristics and behaviors of marmoset progenitor cells appear at altered time points compared with their human counterparts, with the net effect that marmoset progenitors have a shorter overall window during which they can proliferate rapidly. In developmental biology, such shifts in timing are a well-known engine of evolutionary change: modest adjustments to when developmental programs switch on or off can produce large differences in final organ size without requiring wholesale redesign of the underlying machinery. The marmoset data suggest that a compressed proliferative window is one such adjustment, quietly capping the neuron output of the developing cortex.

Because most of the experiments relied on organoids, establishing that the cultures faithfully mirrored real development was essential. “The study combines the advantages of in vivo and in vitro methods,” says Michael Heide, head of the Junior Research Group Brain Development and Evolution. “Organoids are well-suited for obtaining statistically robust results because such sample sizes are not feasible in primates. We subsequently repeated some key experiments in fetal brain tissue to confirm the results from the organoids.” The team found that 50-day-old marmoset organoids closely resemble natural brain development at day 90, a correspondence that allowed them to align the developmental clock of the organoids with that of the actual marmoset embryo and interpret their results with confidence. That validation step is what elevates the findings from an interesting culture phenomenon to a credible account of what happens inside a developing primate.

The broader significance of the work runs in two directions. Evolutionarily, it offers a concrete, mechanistic scenario for how the striking diversity of primate brain sizes and shapes could have arisen: not through the appearance of entirely new cell types, but through coordinated changes in how existing progenitor cells divide, mature, and keep to their developmental schedule. Medically, the same cellular programs that scale a brain up or down are the ones that can go awry in developmental disorders, and the study provides an important framework for understanding both the processes themselves and the consequences of their disruption in the human brain. For a small monkey with a smooth brain, the common marmoset is now teaching scientists a great deal about what it took to build a folded one.

Subject of Research: Cellular mechanisms of neural progenitor development underlying smooth versus folded primate brains

Article Title: Smooth instead of folded—brain development in primates

Article References: Smooth instead of folded—brain development in primates. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: marmoset, brain development, neural progenitor cells, brain organoids, cortical folding, primate evolution, cerebral cortex, neurogenesis, German Primate Center, Science Advances, developmental neuroscience, brain size

The framework, described in Nature Computational Science by Yingjie Zhao and Zhiping Xu of Tsinghua University together with Yicheng Song and Fan Xu of Fudan University, is called physics-transfer learning, and its central insight is deceptively simple: the governing equations of nonlinear elasticity do not care whether they are describing a rubber sheet, an engineered shell or living brain tissue. If a neural network can first learn the mechanics of folding in simple, analytically tractable geometries where the mathematics is fully understood, it can then carry that knowledge across a carefully staged sequence of increasingly complex systems, ultimately arriving at the brain itself. Each step along this chain of physics-anchored domains reinforces the same underlying mechanical principles, so the network never drifts far from the physical laws that actually govern morphogenesis.

The starting point of this transfer chain is deliberately humble. The researchers began with elastic structures whose buckling behavior can be derived in closed form, embedding the classical theory of nonlinear elasticity into the network’s architecture and training objectives as explicit constraints rather than leaving the model to infer them from examples. From there, the team constructed what they call chains of physics-anchored domains, a progression of model systems that gradually increases in complexity and fidelity while preserving the shared mechanical substrate. This staging matters because transfer learning in the standard machine learning sense has long been criticized for moving knowledge between domains with no guarantee of success. By anchoring every step to consistent governing laws, the researchers could go further than intuition and actually prove something remarkable: a theoretical generalization bound that explains why and when the transferred knowledge remains reliable on the unfamiliar target domain.

That theoretical foundation sets the work apart from the crowded field of physics-informed machine learning. Since the introduction of physics-informed neural networks, researchers have sought to weave partial differential equations into deep learning, producing impressive results in fluid dynamics, materials discovery and beyond. But these approaches often face what practitioners describe as an accuracy-performance dilemma, and a 2024 analysis in Nature Machine Intelligence warned that weak baselines and reporting biases have inflated optimism in the field. The new study confronts that skepticism head-on by deriving its generalization bound from the physics-transfer theory itself, turning what is usually an empirical claim into a mathematical statement about how closely the source and target domains share their governing laws. In effect, the bound quantifies the value of physical consistency, providing a principled answer to the question every applied scientist asks: how well will my model work on data it has never seen?

To test the framework against reality, the researchers turned to clinical data, validating their models against magnetic resonance imaging atlases of the developing human brain, including the spatio-temporal surface atlas of the fetal brain generated by the Developing Human Connectome Project. The models were asked to perform two distinct tasks. The first was descriptive: characterizing geometric features of the brain surface, such as curvature, that encode the signature of the folding process. The second was predictive: forecasting the progression of morphological development, essentially predicting how the folding landscape will evolve over developmental time. Compared with purely supervised learning models trained on the same limited data, the physics-transfer approach demonstrated markedly stronger performance on both tasks, a result the generalization bound anticipated. The comparison makes clear that when labeled examples are scarce, physics is worth more than data.

Beyond raw predictive accuracy, the framework delivers something arguably more valuable for scientists: understanding. The trained networks yield reduced-dimensional evolutionary representations, compact mathematical descriptions that distill the essential physics of brain morphogenesis into a low-dimensional space. In this compressed representation, the bewildering complexity of a folding cortex collapses onto a small set of coordinates that trace the trajectory of development, in much the same way that principal component analysis reveals dominant patterns in high-dimensional data. The researchers’ analysis of the internal structure of the neural networks shows that these reduced representations are not statistical artifacts but carry genuine mechanical meaning, connecting the information bottleneck of deep learning with the physical instabilities, such as buckling and wrinkling, that mechanicians have long studied in shrinking and growing shells.

The implications reach well beyond developmental neurobiology. The same group had previously applied physics-transfer learning to discover high-strength alloys, demonstrating that the framework generalizes across materials science, and the new work extends the recipe to soft matter and biomechanics. For clinicians, the prospect of physics-aware digital twins of the developing brain is particularly compelling. A digital twin in this context is a patient-specific computational model that evolves in lockstep with its biological counterpart, offering a virtual laboratory in which development can be monitored, diagnosed and, in principle, intervened upon. Malformations of cortical development, a class of congenital conditions arising when the folding process goes awry, are currently classified largely by descriptive criteria; a predictive mechanical model could identify deviations from normal development earlier and connect them to their underlying biomechanical causes.

The team has been unusually generous with the products of the research, releasing digital libraries of brain models and simulation data through figshare, the fetal brain surface atlas data through G-Node, and all scripts for reproducing the results through GitHub, Zenodo and Code Ocean. This openness matters because reproducibility has been a persistent sore point in machine learning for the physical sciences, and it invites the broader community to stress-test the generalization bounds and probe the limits of the domain transfer. It also lowers the barrier for groups in resource-limited settings, where large labeled datasets of fetal imaging are precisely the kind of resource that is hardest to obtain. A method that achieves strong performance from scarce data, grounded in physics rather than brute-force annotation, democratizes access to high-fidelity developmental modeling.

The work also speaks to a deeper conversation about how machine intelligence should engage with the physical world. Data-driven approaches have achieved spectacular successes, but critics note that pure pattern recognition can fail catastrophically outside its training distribution, an unacceptable risk in medicine. Theory-grounded approaches, by contrast, trade some flexibility for reliability, encoding invariants that hold regardless of the examples on hand. Physics-transfer learning offers a bridge between these poles: it exploits data where data exist, leans on physical law where they do not, and, crucially, comes with a mathematical guarantee that quantifies the trade. As Zhao, Song, Xu and Xu suggest, that combination of accuracy, interpretability and theoretical assurance may be exactly what physics-aware digital-twin technologies need to move from promising demonstrations to clinical tools for understanding, diagnosing and ultimately intervening in the developing and diseased brain.

For a species whose intelligence is written in the folds of its cortex, there is a satisfying symmetry in the fact that it took a fusion of physics and machine learning to finally read that script. The folding of the brain is neither pure genetics nor pure mechanics but the inseparable entanglement of the two, and any model that captures only one side of the story was always destined to fall short. By teaching a neural network the elastic grammar that all folding matter obeys, and then letting it climb the ladder of complexity from simple shells to living tissue, the Tsinghua and Fudan team has shown that the shortest path to understanding nature’s most intricate forms may run through her simplest ones. If the framework continues to validate against clinical imaging, the folding patterns on a fetal scan could one day speak not just of anatomy, but of mechanics, prognosis and possibility.

Subject of Research: Physics-transfer learning for predicting brain morphogenesis and cortical folding from limited data

Article Title: Predicting brain morphogenesis via physics-transfer learning

Article References: Zhao, Y., Song, Y., Xu, F., & Xu, Z. (2026). Predicting brain morphogenesis via physics-transfer learning. Nature Computational Science. https://doi.org/10.1038/s43588-026-01040-7

Image Credits: AI Generated

DOI: 10.1038/s43588-026-01040-7

Keywords: physics-transfer learning, brain morphogenesis, cortical folding, gyrification, nonlinear elasticity, physics-informed machine learning, neural networks, generalization bound, digital twin, fetal brain imaging, Developing Human Connectome Project, computational mechanics