CDM cortical organoids model MBNL-driven brain defects

What the work claims

This is a primary preclinical study that uses human cortical organoids to model the neurodevelopmental arm of congenital DM1 (CDM), the most severe form of myotonic dystrophy type 1. The authors generated cortical organoids from patient-derived pluripotent stem cells carrying more than 1000 CTG repeats in the 3' untranslated region of DMPK, the expansion that causes CDM. They report that these organoids recapitulate the disease's molecular hallmarks, including nuclear DMPK RNA foci, MBNL protein sequestration and splicing abnormalities previously seen in patient brain tissue. They also show that MBNL loss of function is sufficient to reproduce key neurodevelopmental phenotypes, and that two clinical-stage compounds, tideglusib and erythromycin, reduce RNA foci and rescue progenitor proliferation.¹

The central claim is that cortical organoids can serve as a human-relevant preclinical platform for CNS-targeted therapies in DM1, a disease whose drug-development efforts have focused almost entirely on skeletal muscle. The evidence spans molecular markers, cell-type-specific phenotypes, genetic perturbation and pharmacological rescue.

How it works

DM1 is caused by an expanded CTG repeat in DMPK that is transcribed into a toxic CUG-repeat RNA. That RNA accumulates in nuclear foci and sequesters Muscleblind-like (MBNL) proteins, which are RNA-binding regulators of alternative splicing. Loss of MBNL function produces widespread splicing dysregulation, and disease severity rises with repeat length. CDM patients carry thousands of repeats and present at birth with severe hypotonia plus cognitive and behavioral impairments, pointing to early brain involvement.

The authors differentiated three DM1 human pluripotent stem cell lines, all carrying more than 1300 CTG repeats, and three control lines into cortical organoids using the Paşca 2015 protocol. At days 30 and 60 both control and CDM organoids contained SOX2-positive/PAX6-positive neural progenitor cells and HuC/D-positive/TUJ1-positive postmitotic neurons, showing that early cortical patterning occurred. By day 120, GFAP-positive and S100β-positive astrocytes were present, indicating that the neurogenic-to-gliogenic transition took place even in CDM organoids.¹

Disease hallmarks appeared early and persisted. Intranuclear CUG RNA foci were detected in roughly 70 percent of nuclei in CDM organoids at both day 30 and day 120, with a higher number of foci per cell in SOX2-positive progenitors than in HuC/D-positive neurons. MBNL2 sequestration was visible in foci at day 30, and western blotting of 120-day organoids showed reduced MBNL2 protein in CDM samples across three independent batches. Splicing analysis of MBNL2 exon 5, ATP2A1 exon 22 and ITGA6 exon 27 revealed abnormalities in CDM organoids that matched patterns reported in DM1 patient brain tissue.¹

The cellular phenotype was a progenitor proliferation defect. CDM organoids at day 30 had fewer SOX2-positive neural progenitors, but similar cleaved-caspase-3 staining, suggesting impaired proliferation rather than increased apoptosis. EdU incorporation and PHH3 staining confirmed reduced proliferation. To test whether MBNL loss of function was the driver, the authors analyzed organoids from MBNL2 knockout and MBNL1/2/3 triple-knockout induced pluripotent stem cells. Both models reproduced the progenitor proliferation defect and the later neuronal-glial imbalance, with reduced CTIP2-positive deep-layer and SATB2-positive superficial-layer neurons and increased GFAP-positive and NFIA-positive glial cells. At day 120, NFIA-positive cells carried more RNA foci than neurons, and magnetic sorting of GLAST-positive astrocyte-lineage cells showed a higher foci burden in CDM organoids.¹

For pharmacology, two CDM iPSC lines were treated from day 10 to day 25 with 25 µM tideglusib, a GSK3β inhibitor in phase II evaluation for CDM cognitive endpoints, or 25 µM erythromycin, a macrolide reported to displace MBNL proteins from RNA foci. Tideglusib lowered DMPK transcript levels; erythromycin did not change transcript abundance, consistent with its proposed mechanism. Both compounds, however, markedly reduced nuclear DMPK RNA foci and rescued EdU incorporation in SOX2-positive progenitors without increasing cleaved-caspase-3 signal. Tideglusib also normalized elevated GSK3β protein and restored RBM45, an RNA-binding protein upregulated in DM1 and regulated by GSK3 activity, to control levels.¹

Where a skeptic should push

The authors themselves identify the main causal gap: the study compares CDM patient lines with unrelated control lines, and the magnitude of the phenotypes could partly reflect donor-specific genetic background rather than the CTG expansion alone. They state that a formal demonstration of causality would require CRISPR-corrected isogenic CDM lines. That caveat is load-bearing. Until isogenic correction is done, one cannot be sure that the observed neuronal-glial imbalance is fully attributable to the DMPK expansion rather than to background variation that happened to segregate with the CDM genotype.

Sample size is also limited. The progenitor and drug-treatment data are shown for two or three CDM lines, with individual data points representing the mean of three organoids from one independent differentiation experiment. The drug rescue experiments used two CDM iPSC lines. That is enough to establish feasibility, but it is not a powered pharmacology study. The concentrations, 25 µM for both compounds, are high for erythromycin relative to achievable plasma levels, and the treatment window covers early differentiation rather than a mature neuronal readout. Whether the proliferation rescue translates into lasting correction of cortical architecture or function is not shown.

A third caution is interpretation of the MBNL knockout phenotypes. MBNL2 knockout and MBNL1/2/3 triple knockout reproduced key CDM phenotypes, supporting MBNL loss of function as a major contributor. However, complete MBNL knockout may be a stronger perturbation than the partial sequestration seen in CDM, and it removes functions that may not be equally impaired in patients. The phenotypic similarity is therefore suggestive of mechanism but not a precise quantitative model of disease severity.

What CDM brain organoids change for CNS drug models

For organoid-based drug discovery, the most important message is that a neurodevelopmental disease mechanism can be modeled and pharmacologically manipulated in human cortical tissue before any clinical exposure. Most DM1 drug development has targeted skeletal muscle because muscle phenotypes are accessible and clinically urgent. The brain phenotypes have been harder to reach because there was no human model that captured early corticogenesis. This paper provides one, and it immediately flags a translational gap: two compounds moving through clinical trials for muscle or cognitive endpoints can now be asked whether they also correct the early brain developmental phenotypes the model reproduces.

The non-obvious implication is that CNS efficacy and systemic toxicity can now be decoupled earlier. A compound that reduces RNA foci in muscle may or may not cross into the developing brain at sufficient concentration, engage the same MBNL-sequestration mechanism in neural progenitors, or act during the right developmental window. An organoid platform lets those questions be asked in human tissue, with molecular and cellular readouts, before committing to a pediatric neurodevelopmental trial. That is particularly important for CDM, where the window for protecting corticogenesis may be prenatal or early postnatal.

The opportunity is to make congenital neurodevelopmental disorders a mainstream indication for brain organoid pharmacology. The threat is over-interpretation of rescue in a reductionist model. Reduced RNA foci and restored progenitor proliferation are encouraging, but they are not the same as rescued cognition, cortical layering or circuit function. If regulators or investors treat organoid rescue as stronger evidence than it is, the field could advance compounds that fail in patients because the model missed later developmental steps, immune interactions or whole-brain pharmacokinetics. The right use of this platform is as a filter and mechanism-discovery tool, not as a surrogate endpoint for clinical efficacy.

The bottom line

Established: cortical organoids from CDM patient stem cells recapitulate nuclear DMPK RNA foci, MBNL sequestration, splicing abnormalities, reduced neural progenitor proliferation and altered neuronal-glial composition, and these phenotypes are reproduced by MBNL knockout. Both tideglusib and erythromycin reduce RNA foci and rescue progenitor proliferation in the model. Unestablished: whether the phenotypes are fully caused by the CTG expansion rather than donor background, whether the drug effects persist and translate into corrected cortical architecture or function, and whether the concentrations and treatment windows are clinically relevant. What would confirm the platform's value is replication with isogenic corrected lines and a prospective comparison of organoid pharmacology with clinical CNS outcomes. What would break the optimistic reading is evidence that the rescue is restricted to early progenitor readouts and does not track with any clinically meaningful neurological improvement. For the foundry, this is a strong example of how organoid models can expose a neglected organ-system target, provided the model is not mistaken for the patient.