The research team, led by Yue Niu, Zhao Xu, Zongpu Zhou, Xianru Jiao and Zhixian Yang of the Department of Pediatrics, screened pediatric patients evaluated between January 2018 and December 2024 for unexplained global developmental delay, intellectual disability, intractable epilepsy or structural brain malformations. To qualify for inclusion, each child had to show neurodevelopmental abnormalities and carry a rare variant in one of the four target genes, confirmed by trio-based whole-exome sequencing and validated by Sanger sequencing or quantitative methods. Children with acquired brain injuries, such as severe neonatal hypoxic-ischemic encephalopathy, or with additional pathogenic variants in other monogenic neurodevelopmental genes were excluded. Eleven probands ultimately met the criteria, and all were born at term or appropriately preterm, with birth weight, length and head circumference within normal gestational-age- and sex-adjusted percentiles.

The unifying clinical thread was global developmental delay, which was present in every single patient—eleven of eleven, or 100 percent. Beyond that shared feature, the picture fragmented dramatically. Epileptic seizures affected nine of the eleven children (81.8 percent), including eight with active epilepsy syndromes and one with only an isolated febrile seizure. Structural brain abnormalities appeared on neuroimaging in seven children (63.6 percent), while the remaining four—more than a third of the cohort—had entirely normal cranial MRI scans. That heterogeneity, the authors argue, is precisely why microtubule-related disorders resist simple genotype-phenotype rules and why careful case-level documentation remains valuable.

The four children with TUBA1A variants, three girls and one boy, presented with early-onset developmental delay beginning between three and six months of age. Two of them showed a particularly concerning feature: dynamic developmental regression. One child experienced motor deterioration after a febrile episode at six months, while another showed cognitive and motor decline after the age of two. Three of the four developed drug-resistant focal seizures, and one additionally had tonic seizures that fulfilled diagnostic criteria for Lennox-Gastaut syndrome. Despite multi-drug anti-seizure regimens, daily refractory seizures persisted in all three affected individuals, while the fourth remained completely seizure-free. Neuroimaging spanned the full severity spectrum: classical lissencephaly in one child, bilateral frontoparietal pachygyria in another, cerebellar hypoplasia in a third, and only mild periventricular leukomalacia in the fourth.

The TUBB2A group told a different story. All four children, again three girls and one boy, showed severe early developmental impairment, with onset ranging from the neonatal period—day three in one case—to four months. Independent walking and command comprehension were significantly impaired or delayed across the group, yet notably, none exhibited developmental regression during follow-up. All four experienced seizures, and three developed early-onset epilepsy syndromes: one with refractory focal and tonic seizures consistent with Lennox-Gastaut syndrome, and two with infantile epileptic spasm syndrome accompanied by hypsarrhythmia on EEG, both relapsing into drug-resistant epilepsy around age three. A striking detail was the presence of sound-induced reflex seizures in two children, triggered by voices or startle. Structural brain changes appeared in only two of the four probands—one showing deepened fronto-temporo-parietal sulci with progressive cerebral atrophy, the other multiple cerebellar gyral malformations—while the other two had normal MRI scans.

The two TUBB4A patients diverged sharply in age of onset, illustrating the phenotypic breadth of this gene. One child first showed symptoms at eleven years old, presenting with mild intellectual disability and autism spectrum disorder; at her most recent follow-up at age twenty-one, she walked with a slight limp, was thin, photosensitive and anemic, and carried an independent clinical diagnosis of systemic lupus erythematosus—an unusual co-observation the authors treat cautiously as a possible coincidence rather than evidence of an immune component of TUBB4A disease. The other TUBB4A patient, whose variant was actually a de novo 2.36-megabase duplication on chromosome 19p13.3–p13.2 encompassing TUBB4A and several neighboring genes, presented at one year and four months with severe developmental delay, a right lower limb limp and reduced muscle strength. Both children had focal seizures and shared a distinctive EEG signature: Rolandic discharges over the centroparietal regions, with discharge rates reaching up to 70 percent in one child—a feature the authors note is rarely described in the TUBB4A literature.

The single TUBGCP6 patient rounded out the cohort with the mildest profile. This boy walked at one year and three months and could say simple reduplicated words by age two, but at two years and six months his height, weight and head circumference all fell well below normal ranges—head circumference at minus three standard deviations, consistent with microcephaly. He could not yet speak simple sentences, but he had never had a seizure, and no other organ systems showed abnormalities. TUBGCP6 encodes a core subunit of the gamma-tubulin ring complex, which orchestrates microtubule nucleation and spindle pole integrity, and biallelic loss-of-function variants in this gene are classically associated with autosomal recessive microcephaly with chorioretinopathy. This child carried compound heterozygous variants—a maternally inherited missense change and a paternally inherited intronic variant near a splice acceptor—but both were conservatively classified as variants of uncertain significance pending transcript-level validation.

On the molecular side, the eleven patients carried twelve variants in total: ten missense variants, one non-canonical intronic alteration and one copy number duplication. Nine of the missense variants in TUBA1A, TUBB2A and TUBB4A were classified as pathogenic or likely pathogenic under ACMG/AMP criteria, and all were de novo—including two cases of post-zygotic mosaicism—and absent from the gnomAD population database. Three TUBA1A loci (P175L, M377I and Y172D) and three TUBB2A loci (D203E, R262H and R391H) had not been reported before, along with the TUBB4A G10S variant, expanding the known genotype-phenotype spectrum of these genes.

To probe how these substitutions might damage protein function, the team turned to computational structural biology. They predicted three-dimensional protein structures with AlphaFold3 and used FoldX to calculate the change in folding free energy, or ΔΔG, for each missense variant—a positive value indicating predicted destabilization. All ten missense variants showed some degree of increased free energy, but the magnitudes varied widely. The largest shift belonged to TUBA1A Y172D, at 5.15 kcal/mol, a residue sitting near the GTP-binding and intradimer interfaces where destabilization could theoretically interfere with longitudinal protofilament assembly. TUBB2A D203E and R262H showed substantial increases of 4.38 and 3.38 kcal/mol respectively, with the latter predicted to abolish side-chain hydrogen bonds with three neighboring residues. By contrast, TUBA1A T150I produced only a 0.50 kcal/mol shift, and the TUBB4A G10S substitution was predicted to introduce two entirely new hydrogen bonds with adjacent residues while retaining the original three.

The authors are careful to frame these structural results as hypotheses rather than conclusions. Because each novel variant was documented in a single patient, and because the magnitude of predicted destabilization did not correlate linearly with clinical severity—variants with high ΔΔG values appeared in children whose brain scans ranged from normal to lissencephaly—the computational models cannot serve as quantitative predictors of outcome. FoldX’s empirical force fields were parameterized for soluble globular proteins, not dynamic cytoskeletal polymers, and AlphaFold3 generates static structures that do not capture GTP hydrolysis states, chaperone interactions or the crowded cellular environment. The team calls for functional assays such as in vitro microtubule polymerization kinetics, patient-derived induced pluripotent stem cell neuronal models and larger international registries to test whether these theoretical biophysical perturbations translate into real cellular defects. For now, the study stands as a detailed clinical map of a rare and heterogeneous disease family—one that underscores both the power and the limits of computational modeling in rare disease genetics.

Subject of Research: Clinical and computational analysis of microtubule-related gene variants in children with neurodevelopmental disorders

Article Title: Clinical characterization and in silico structural analysis of microtubule-related gene variants in eleven patients

Article References: Niu, Y., Xu, Z., Zhou, Z., Jiao, X., & Yang, Z. (2026). Clinical characterization and in silico structural analysis of microtubule-related gene variants in eleven patients. Frontiers in Genetics, 17, 1939857. https://doi.org/10.3389/fgene.2026.1939857

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1939857

Keywords: microtubules, tubulinopathies, TUBA1A, TUBB2A, TUBB4A, TUBGCP6, epilepsy, global developmental delay, lissencephaly, AlphaFold3, FoldX, genotype-phenotype correlation