HCMV Drives Distinct Neural and Viral States in Human Cortical Organoids
Human cytomegalovirus infection did not produce a single uniform response in cortical tissue. It fractured the biology of the organoid into multiple cellular and viral substates, with changes visible across infected and uninfected populations after HCMV exposure to healthy iPSC-derived human cortical organoids.
The study used single-cell RNA sequencing to map how infection altered the developing cortical environment. In uninfected organoids, the investigators defined 16 cell type clusters using the Human Neural Organoid Cell Atlas, establishing a baseline neural architecture before viral challenge. After infection with GFP-expressing HCMV, the cluster structure shifted in ways that extended beyond the cells carrying detectable viral signal. Some changes appeared in GFP-positive cells, while others also emerged in GFP-negative populations, suggesting that infection altered the surrounding tissue ecosystem rather than remaining confined to visibly infected cells.
A striking transition involved the Glutamatergic/GABAergic Neuron cluster, which appeared to move toward a CMV-Infected Cell cluster. This is biologically plausible in a developing neural system, where a virus that favors specific permissive states can redirect cells away from their expected maturation path and into a transcriptional program dominated by infection. In practical terms, the organoid did not simply host the virus; it appeared to reorganize around it.
Infection Was Not Uniform Across Cells or Time
The viral transcript patterns were heterogeneous enough that a single infection score could not capture the full spectrum of cellular states. Viral gene expression varied between individual cells and between cell types, forcing the investigators to introduce a TC indexing score to account for differing stages of infection. That approach matters because HCMV biology is not static. Early gene activity, late gene activity, and productive viral replication can occupy different transcriptional phases, and cells within the same tissue can sit at different points along that continuum.
Machine learning approaches layered onto the transcriptional data identified cells that were likely producing de novo HCMV, along with several viral substates. The distinction between a cell merely carrying viral material and one actively manufacturing new virions has major mechanistic implications. Productive infection typically demands broader rewiring of host transcription, metabolism, and cell state, which can explain why the tissue signature spread well beyond a single cell class.
One sentence captures the core result: the infected organoid contained more than one viral phenotype at the same time.
This is important for congenital neurotropism as well as for broader central nervous system disease hypotheses. If the same tissue can harbor multiple viral substates, then pathogenesis may depend less on one linear infection program and more on a shifting balance of permissive, transitional, and productive states. That framework fits a virus with lifelong persistence and a known capacity to disrupt neurodevelopment under conditions of insufficient immunity.
EMT-Linked Programs Appeared in Viral Substates
Several viral substates showed significant enrichment of genes involved in epithelial-to-mesenchymal transition, or EMT. In a cortical organoid context, EMT-associated signaling is not a superficial annotation. It points to altered adhesion, motility-related transcriptional programs, and broader developmental instability. Even in neural tissue, the appearance of EMT-linked signatures suggests that HCMV may push infected cells toward a state that is structurally and transcriptionally less anchored to normal neuronal differentiation.
EpCAM emerged as a particularly notable marker in this analysis. The study independently verified HCMV-induced EpCAM protein expression in cortical organoids and glutamatergic neurons. EpCAM is more often associated with epithelial biology, so its appearance in infected neural contexts implies a stark deviation from the usual cortical developmental program. In a developing brain model, such a shift may reflect a viral strategy that exploits cell plasticity while simultaneously disturbing the normal identity of the neuron.
The finding also reinforces that host-pathogen interactions in the brain cannot be reduced to neuron loss alone. A virus can distort cell identity before a cell dies or stalls in differentiation. In congenital infection, that kind of transcriptional rerouting may be enough to derail circuit formation, even without complete destruction of the tissue.
Why Cortical Organoids Matter for Congenital HCMV Biology
Human CMV remains a major concern in settings where immunity is limited, and congenital infection is among the most clinically consequential settings. Infection of the developing brain can produce microcephaly and sensorineural hearing loss, and the broader literature has increasingly implicated HCMV in adult CNS disease as well. Yet direct study in human brain tissue is constrained by host tropism and anatomy, leaving organoid models as a useful bridge between mechanistic virology and human neurodevelopment.
These cortical organoids offer a controlled human system in which viral entry, replication, and host response can be measured in parallel. The single-cell design matters because bulk RNA approaches would have blurred the separation between infected cells, bystander cells, and transitional cell states. Here, the cellular map exposed a more layered biology: one that included altered developmental clusters, divergent infection stages, and a viral transcriptional spectrum broad enough to require a new analytic index.
The work also points to a methodological lesson for future CNS virology studies. Complex infection phenotypes will probably continue to escape simple one-score, one-cluster analysis. By combining single-cell sequencing, temporal indexing, and machine learning, the investigators built a framework that can separate productive infection from adjacent developmental noise. That kind of analytic structure will be essential if organoids are going to do more than merely confirm that a virus infects the brain model.
The preprint stops short of clinical translation, and it does not answer whether these exact states occur in vivo at the same scale or sequence. It does, however, offer a biologically credible explanation for how HCMV can reshape developing cortical tissue without producing a single uniform lesion. The remaining challenge is to link these viral substates to developmental injury in human brain tissue and to determine which host programs are most vulnerable when infection begins to fragment neuronal identity.