Molecular hallmarks of hydrocephalus

Abstract

Hydrocephalus (HC) is a failure of brain and cerebrospinal fluid (CSF) homeostasis often associated with dilation of the CSF-filled ventricles (ventriculomegaly). Hallmarks of HC include aberrant CSF dynamics, neural stem cell dysfunction resulting in impaired neurogenesis and corticogenesis, biomechanical instability at the brain-CSF interface, and disrupted synaptogenesis and neural circuitry. Pleiotropic mechanisms, including genetic and environmental insults to the brain, contribute to neurodevelopmental comorbidities. Hypothesis generation from genome-wide, single cell multi-omic analyses coupled to experimental validation using induced pluripotent stem cell-derived cerebral organoids will refine molecular classification of HC subtypes and may lead to precision-based surgical and pharmacologic treatments.

One-Sentence Summary

The molecular hallmarks of hydrocephalus provide a framework for mechanistic studies to re-classify human disease subtypes and develop precision therapies.

INTRODUCTION

Hydrocephalus (HC) is the most common indication for childhood brain surgery ¹. HC is a disorder of aberrant brain development and cerebrospinal fluid (CSF) homeostasis that is often associated with dilation of the cerebroventricular system (ventriculomegaly), increases in intracranial pressure or both ^{1, 2}. HC includes primary (genetic or developmental) ^{3, 4} and acquired forms ⁵ across the age spectrum ^6–8. Systemic and intracranial infections are the leading causes of HC across the world. However, numerous molecular mechanisms and genetic factors underlie congenital HC ^{6, 7}. Current HC treatments are limited to surgical procedures such as ventriculoperitoneal (VP) shunting, endoscopic third ventriculostomy (ETV), or ETV with choroid plexus cauterization (CPC) ^{2, 9}. Although age, etiology, treatment history and other factors may help to predict rates of surgical success ¹⁰, failure rates remain unacceptably high with approximately 25% requiring re-operation within the first post-operative year ¹¹. Some children undergo more than 100 neurosurgical procedures over the course of their abbreviated life spans ^11,12. Rational design of pharmacologic treatments could substantially alleviate morbidity and mortality associated with HC but these efforts have been impeded by an incomplete mechanistic understanding of the disease.

The path of CSF production and flow is classically represented as secretion by the choroid plexus (ChP) (Fig. 1A), with unidirectional flow from the lateral ventricles through the foramina of Monro to the third ventricle, through the cerebral aqueduct of Sylvius into the fourth ventricle, followed by CSF reabsorption into the subarachnoid space through the foramina of Luschka and Magendie (Fig. 1B). While useful as a general framework for ventricular system anatomy, this model is widely challenged and neglects emerging contributors to CSF flow, including SVZ neural stem cells (NSCs), ependymal cilia, and the glia-lymphatic (glymphatic) system (Fig. 1C), including the CSF connection to the peri-neural space ¹². For example, defects in ependymal cilia can cause HC in animal models through disruption of cilia organization, beat frequency, and alterations in CSF flow dynamics with pleiotropic effects on brain development and function ¹³. However the role of cilia as a primary pathophysiologic driver of human HC remains contested. In addition, the contribution of the glymphatic system ¹⁴ to HC is attributed to attenuated CSF waste clearance from germinal matrix hemorrhage and/or inflammation ^{14, 15}, but additional mechanistic studies are needed. Finally, the nasopharyngeal lymphatic plexus and meningeal lymphatic vessels may represent alternative routes of CSF drainage in mice ^{16, 17}, but the role of these systems in human pathophysiology remain unexplored. These emerging mechanisms challenge dogmatic models of CSF flow and highlight our incomplete understanding of the relationships connecting CSF flow, brain development, and HC pathophysiology.

Figure 1. Overview of the human cerebroventricular system.

(A) Illustration of the choroid plexus (ChP) epithelium with component cell types. (B) Systemic movement of cerebrospinal fluid (CSF) produced by the ChP epithelium. CSF is thought to traverse the subarachnoid space and to be reabsorbed into the systemic circulation via the arachnoid granulations (B) or the cranial meninges and the glymphatic system (C). Abbreviations: aqueduct of Sylvius (AS), cerebrospinal fluid (CSF), ChP (CP), foramen of Monro (FM), foramen of Luschka (FM), fourth ventricle (4V), lateral ventricle (LV), interstitial fluid (ISF), third ventricle (3V), and subventricular zone (SVZ).

Molecular studies reveal HC as a phenotypically heterogenous disorder with impact on both “drain and brain.” Here we outline molecular hallmarks of human HC, to serve as a framework for understanding human brain development, HC pathobiology, and to identify remaining knowledge gaps. The hallmarks of HC include genetic predisposition, NSC dysfunction, alterations in corticogenesis, perturbation of brain volume and growth, biomechanical disruption of brain architecture, defects in synaptogenesis and neural circuitry, aberrant CSF dynamics and secretion, and augmentation of ChP developmental trajectories (Figure 2). Much of our molecular understanding of HC is derived from in vivo and in vitro model systems. Although valuable, these approaches do not fully recapitulate the diverse genetic, ancestral, physiologic, or phenotypic characteristics of human HC. Recent advances in generation and study of human induced pluripotent stem cell (iPSC)-derived cerebral organoids suggest that many mechanisms of HC will soon be directly testable in human systems. We will discuss in detail relevant considerations in leveraging this model system to advance our understanding of HC.

Figure 2. The hallmarks of human hydrocephalus,

defined as neural stem cell (NSC) dysfunction, alterations in corticogenesis, brain volume and growth perturbation, biomechanical disruption, defects in synaptogenesis and brain circuitry, aberrant cerebrospinal fluid dynamics and choroid plexus development, genetic predisposition, and ventriculomegaly and comorbid traits.

GENETIC PREDISPOSITION

The numerous challenges to understanding genetic contributions to HC include locus heterogeneity, the sporadic nature of HC and phenotypic complexity among HC patients. Moreover, HC is commonly accompanied by neurodevelopmental traits (for example autism spectrum disorders (ASD), epilepsy, cognitive dysfunction, etc.) with pleiotropic mechanisms. Strategies to overcome these challenges include Mendelian genetic analysis using whole-exome sequencing (WES) of rigorously phenotyped human cohorts to identify inherited and de novo mutations, as well as copy number and other structural variations (SV) across the genome. This approach enables precise identification of disease-causing mutations (at single base pair resolution in select cases), and has been successful in identifying causative genes for HC ^{3, 18}, ASD ¹⁹, and epilepsy ²⁰. WES is unbiased, systematic and reproducible, enabling robust characterization of implicated loci. HC risk genes TRIM71, SMARCC1, PTEN, FOXJ1, and PIK3CA have been identified using this approach ²¹. As larger cohorts are sequenced and additional genetic mechanisms are explored, diagnostic refinement of congenital HC by genetic subtype and endophenotype might enable precision medicine-based treatment, as WES was diagnostically informative in approximately 38% of cases ²².

Genome-wide and transcriptome-wide association studies (GWAS, TWAS) identify single nucleotide polymorphisms (SNPs) and gene-based associations, respectively. GWAS typically requires a large sample size and best captures associations with common cording or non-coding variants (minor allele frequency > 0.1%), typically low effect size. In addition, GWAS typically relies on array-based SNP detection and imputation methods based on reference population genetic architecture ²³. In contrast, TWAS (which aggregates the effect of single variants across a gene locus), benefits from a higher p-value threshold, reflecting the lower number of statistical tests (for thousands of genes vs. millions of variants). TWAS also accounts for tissue-specific gene regulatory information (benefitting from reference tissue platforms such as the Genotype-Tissue Expression (GTEx) project ²⁴), and can infer molecular mechanism owing to directionality of the gene-tissue pair association with the phenotype ²⁵. While GWAS identifies single SNPs, often in non-coding regions of often obscure functional impact, TWAS can suggest tissue- and gene-level directionality to guide hypotheses to elucidate underlying molecular mechanisms. Integration of these human genetic approaches with ‘omic’ technologies such as neuroimaging linked to genetics and scRNA, epigenetic, and proteomic atlases of the developing brain can provide complementary and convergent means to define HC mechanisms. For example, a TWAS of HC identified maelstrom (MAEL) in the cortex as the gene-tissue pair reaching experiment-wide significant association with HC. This finding was supported by rare variant analysis, integrative genomic analysis of neuroimaging phenotypes, ChP transcriptome analysis in a mouse model of HC, and CSF proteomics from patients with HC ⁴. In summary, use of multiple parallel approaches is and will continue to be paramount for productive study of human cohorts.

NEURAL STEM CELL DYSFUNCTION

The global consequences to the developing brain of molecular errors within NSCs include HC ^{26, 27}. Human genetics studies of HC have identified de novo and inherited mutations, copy number variations and other alterations in genes with roles in NSC maintenance, renewal and development, such asTRIM71, PTEN, PIK3CA, SMARCC1, FOXJ1, and MAEL ^{4, 18, 21}. Mutations in these genes can disrupt NSC function through divergent mechanisms, as illustrated by the examples below. HC-associated mutations in TRIM71, an RNA binding protein and E3 ubiquitin-ligase, alters NSC differentiation and commitment to neural lineage through repression of β-catenin and lysine demethylase 1a (LSD1) translation ^{28, 29}. Long non-coding RNA (lnRNA) Trinc1 (an inhibitor of Trim71 gene expression) blocks TRIM71-dependent fibroblast growth factor/extracellular signal-regulated kinase (FGF/ERK) signaling and NSC self-renewal ³⁰. PTEN phosphatase and PIK3CA kinase regulate the production and subcellular localization of phosphatidyl-inositol 3,4,5-triphosphate (PIP₃), thereby controlling NSC fate, proliferation, and identity. These enzymes also play pleiotropic roles in brain overgrowth phenotypes, among other neurodevelopmental pathologies ^{31, 32}. SMARCC1 is a key regulator of transcription through chromatin remodeling. Smarcc1 deficiency in mice results in embryonic lethality due to exencephaly ³³, and SMARCC1 dysfunction in mouse telencephalon results in aberrant epigenetic regulation in NSCs, causing global alterations in transcription and chromatin state ^{34, 35}. Loss of function Smarcc1 mutations in Xenopus also result in midbrain overgrowth and CSF obstruction, inducing ventriculomegaly without forebrain abnormalities ³⁵. Transcription factor FOXJ1 is required for cilia formation and differentiation of ChP epithelial and ependymal cells in the postnatal brain, but is not expressed in RGCs ³⁶. These data together suggest that convergent, disruptive pathways of NSC differentiation broadly underly congenital HC.

A TWAS of human HC identified MAEL, an evolutionarily divergent gene that regulates piRNA-mediated transposon silencing, expression in the brain cortex as an experiment-wide predictor of HC across all gene-tissue pairs tested. Gene set enrichment analysis of statistically significant HC-associated genes identified NSC regulation as a shared mechanism ⁴. MAEL may contribute to HC risk by regulating transcription through rearrangement of transposons ^{37, 38}, potentially inducing mutations to shape cell identity and augment genome size by altering the genomic loci of DNA sequences. Loss of MAEL can recruit RNA polymerase II to increase transcription and steady-state levels of transposon RNAs. The accompanying increase in heterochromatin spreading despite modest changes in Histone-3 lysine-9 trimethylation (H3K9me3) patterns ³⁹ suggests that MAEL may act independently or downstream of H3K9me3. Indeed, H3K9me3-dependent changes were among the most significantly enriched gene sets (gene ontology) and H3K9me3 methyltransferases among the genes most highly differentially expressed comparing HC cases and controls ⁴. However, the precise mechanism(s) by which H3K9me3-dependent alterations cause HC remain unelucidated. These data collectively implicate NSC function, proliferation, differentiation, and renewal as key mechanistic drivers of HC.

NORMAL DEVELOPMENT AT THE CSF-CORTICAL INTERFACE

Normal development of the cerebral cortex and ventricular system reflects a highly coordinated series of molecular, genetic, and cell-signaling events at the brain-CSF interface. The brain parenchyma and the CSF-producing ChP share a common tissue origin in the NSC at the dorsal midline and rhombic lip ^{40, 41}. Dorsal pallial NSCs give rise to neurons and glia of the cerebral cortex and the lateral ventricle ChP.^42–44. Distinct NSC subsets can self-renew through symmetric cell division and then differentiate to generate de novo all neural and neuroglial cell lineages ⁴⁵. Defects within the NSC compartment or its precursor lineages, including RGCs, oRGCs and IPCs can cause alterations in cortical patterning, folding, and cell-type identity ⁴⁶, with global functional consequences in cognition and neurodevelopment ⁴⁷. Epigenetic alterations within the NSC compartment may also impact corticogenesis and drive HC pathogenesis ⁴⁸.

Human neuroepithelial cells arise from neuroectoderm after neural tube closure (between days 28–32 post-conception) and give rise to much of the brain parenchyma (Figure 3A). Neuroepithelial cells generate multiple progenitor cell types, including radial glia cells (RGCs) beginning around gestational week (GW) 7 ⁴⁹, ChP epithelial cells around GW8 ^{50, 51}, and ependymal cells around GW20 ⁵². RGCs rapidly proliferate to expand around GW10 to generate neurons and intermediate progenitor cells (IPCs) that in turn generate immature neurons ⁵³. RGCs also produce more specialized RGC subtypes such as outer radial glia cells (oRGCs) by GW18 ^{44, 54}. Together, RGCs, oRGCs, and IPCs are the progenitors responsible for producing cortical neurons through the stereotyped inside-out process in which deep-layer neurons are born before the neurons that reside in more superficial layers ⁵⁵. After generating neurons and neuronal progenitors, RGCs switch to a gliogenic role, producing astrocytes and oligodendrocyte precursor cells by the second trimester ⁵⁶.

Figure 3.

Overview of human neocortical development and HC-associated genes. (A) Cellular differentiation during cortical development between gestational weeks (GW) 4–40. NSCs successively expand, generate progenitors, and differentiate to form the principal components of the human cortex and ventricular system ^{43, 48}. (B) Summary of cell types and their HC-associated gene expression. (C) UMAP clustering of HC-associated gene expression in neural progenitor cells (RGCs, oRGCs, IPCs, ORGs) and early differentiated neural cell types during human brain development. Processed scRNA-seq data of human neocortical samples was obtained from Wang, et al. 2024 ⁵⁸. The data subset was of cells originating from pre-natal samples (including first, second, and third trimester). Cell types were re-annotated to reflect stages of neural differentiation by combining annotated cell-types from the original publication using SCANPY ¹⁴⁷. Expression values reflect log normalized UMI counts with a size factor of 10,000. Known markers of the annotated cell types (PAX6, TBR2, NeuN, GFAP) are plotted to demonstrate annotation specificity, and are compared with expression plots of 10 HC-linked genes.

Gliogenesis is a protracted process of differentiation that extends following neurogenesis through early childhood. Cerebrovascular and immune cell development in the human brain, though of possible relevance to HC, are beyond the scope of this review.

NSCs are the common precursor of neurons, neuroglia, ChP epithelial cells, and ependymal cells. Disruption of this lineage may impede CSF homeostasis by several mechanisms. For example, mutations that affect the earliest-stage NSCs could induce overgrowth and ventricular obstruction, alter CSF production and regulation, or perturb ependymal cells in the CSF-brain barrier. Forebrain expansion, neuroepithelial differentiation into oRGCs, and disproportionately long gliogenic periods are highly species-specific, dynamic processes essential for cortical formation, maturation, and growth. For example, mutations in L1CAM have been hypothesized to cause X-linked HC through aberrant radial neuronal migration ⁵⁷. In contrast, mutations that affect later-born, more highly specified progenitors (including RGCs, neuroblasts, or glioblasts) may regulate particular cell types implicated in disease.

The timeline of human cortical development at the cellular level is central to HC. Identification of cell types expressing HC-associated genes during development may therefore help to clarify disease mechanisms. The expression of individual, congenital HC-associated genes, identified through unbiased human genetic studies ^{3, 4, 18}, is enriched in divergent cell subpopulations in the brain neocortex (Figure 3B–C), suggesting cell-type specific roles for HC-associated genes ⁵⁸. We surmise that mapping expression patterns of human HC-associated genes ^{3, 4, 18} to the developing human brain may yield mechanistic insights. For example, MAEL expression is highest in cortex (GTEx, Human Protein Atlas, and ⁴), while other HC-associated genes (PLOD2, PTCH1, PIK3CA, SMARCC1, PTEN, SGSM3) map onto specific subpopulations of cells in the human prenatal brain neocortex ⁵⁹ (Figure 3C). Although these human studies did not validate that NSCs express HC-associated genes, mouse studies suggest that many of the genes expressed by human RGC/oRGCs could also be expressed in NSCs ⁶⁰. Other HC-associated genes (SHH, FOXJ1), although not strongly captured in the prenatal brain dataset ⁵⁹, display cell-type specific expression. SHH ligand is detectable in CSF, and source cells include epithelial cells, blood vessels postmitotic neurons, Cajal-Retzius cells, and ChP ⁶¹. FOXJ1 is expressed by multiciliated cells, including ChP epithelial cells and ependymal cells that line the brain ventricles (GTEx, Human Protein Atlas). Mapping HC risk genes to epigenetic atlases of developing human brain can reveal additional mechanistic insights into how genetic errors cause HC and comorbidities ^{4, 62–66}. Moreover, to our knowledge no whole genome sequencing (WGS) studies of HC have been performed to elucidate the disease contributions of non-coding, regulatory regions or SVs. The highly human-specific and elaborately orchestrated genetic and molecular events underlying NSC development, CSF regulation/dynamics and corticogenesis will require human models to delineate HC mechanisms.

BRAIN VOLUME AND GROWTH PERTUBATION

The goal of treatment for pediatric patients with HC is to optimize brain growth and cognitive development, both primary outcomes in randomized controlled trials of surgical treatments for HC ^{9, 67}. Brain growth trajectories in HC correlate with neurodevelopmental outcomes in HC ⁶⁷. However, determining ‘normal’ brain growth patterns is challenging. Understanding how alterations in HC risk genes contribute to growth of brain, white matter, grey matter ⁶⁸ and cortical patterning ⁶⁹ is essential to inform precision treatments and to define successful outcomes. For example, patients with a genetic syndrome characterized by attenuated cortical development and mild ventriculomegaly in the setting of normal intracranial pressure may not require CSF diversion but, rather, referral to early intervention services for neurocognitive development. On the other hand, permanent CSF shunting may be indicated for patients with loss-of-function mutations in a key regulatory protein involved in CSF clearance. Precision medicine approaches for HC ⁷⁰ have the potential to reduce morbidity and mortality associated with the disease and treatment complications. Indeed, integrative neuroimaging and genomics analyses have identified alterations in brain architecture as a key feature of neurodevelopmental diseases comorbid with HC ^{68, 69, 71, 72}. The top gene-level associations with HC include statistically significant enrichment for genes involved in total brain and white matter volumes, with MAEL among the topmost statistically significant genes ⁴. However, these population genetic models were developed from human genetic studies linked to neuroimaging phenotypes ⁷³ lacking both birth-age data and adjustments for development. Brain charts coupled with multi-omics analysis across age, anthropometric, and population scales will further refine classification of HC subtypes.

HC rarely occurs in isolation, and the extent to which pleiotropic genetic mechanisms contribute to comorbid neurodevelopmental phenotypes is unknown. Detailed neuroimaging studies linked to genetic information may resolve these questions. Ventriculomegaly, but not necessarily HC, has been observed in patients with ASD, cognitive dysfunction, and neuropsychiatric disease ^74–77. ‘Normal’ ventricular size is highly subjective, and while brain charts across the age spectrum have been developed (with significant effort) ⁷⁸, these models have not been widely incorporated into clinical practice. For example, disruptive mutations in the PI3K-Akt-mTOR signaling pathway have been observed in patients with HC, macrocephaly, developmental delay, and ASD. As dysregulation of this pathway can alter corticogenesis, ventriculomegaly in these cases is likely a harbinger of a primary structural brain disorder, rather than a primary driver of HC.

BIOMECHANICAL DISRUPTION OF BRAIN PARENCHYMA

Emerging data on perturbations in the development of brain architecture support direct roles for biomechanical disruption underlying subtypes of congenital HC. For example, HC-associated de novo mutations in Trim71 and disruption of TRIM71 activity in the murine NSC compartment cause cortical hypoplasia and hypercompliant cortex with compensatory ventricular enlargement ^{79, 80}. Mice harboring Trim71 mutations displayed impaired CSF outflow secondary to acquired aqueductal stenosis that could be ameliorated by CSF diversion. In contrast, the aqueduct in patients with L1CAM mutations remained stenotic even after CSF shunting ⁷⁹. These data suggest a distinct mechanism by which altered cortical compliance and biomechanics may cause HC, such that the genetic etiology of HC determines, in large part, surgical treatment response. Independent mechanistic studies of Trim71 further supporting the importance of defects in the NSC compartment have implicated impaired epigenetic mechanisms ^{28, 29, 81} as responsible for phenotypes of Trim71-deficient mice (and humans harboring Trim71 loss-of-function mutations ^{79, 80}). These findings disrupt the classical paradigm of HC, in which defects of CSF production and absorption were primary pathologic drivers of disease, and suggest primary disruption of brain parenchymal structural integrity as a causative mechanism for development of HC

SYNAPTOGENESIS AND CIRCUIT DYSRUPTION

It remains unclear whether increased intracranial pressure and subsequent changes in CSF dynamics after CSF diversion are primary causes of circuit disruptions associated with HC; or, alternatively, if HC-associated genetic errors result in pleiotropic effects leading to ‘passenger’ phenotypes. For example, mutations in genes within the PI3K signaling pathway disrupt corticogenesis and are associated with the HC comorbidities of brain overgrowth and epilepsy ^{82, 83}. Genomic alterations within the PI3K pathway have also been identified by molecular phenotyping of congenital HC patients ^{3, 18}, suggesting a pleiotropic basis for the wide phenotypic spectrum observed in humans with HC. Many independent mechanistic studies of the PI3K-PTEN signaling axis, a critical driver of synapse development ⁸⁴, have identified defects in synaptogenesis and brain circuitry alterations caused by disruptions in this pathway. Disruptions in NSCs may delay gliogenesis or cause downstream defects in glial differentiation, impair processes required for synaptogenesis (astrocytes), and produce defects in circuit transduction through disrupted myelin production caused by oligodendrocyte dysfunction. Finally, aberrant cortical neurogenesis leads to imbalance between excitatory neurons and subpallial-generated inhibitory neurons, or to defects in neuronal migration, both known mechanisms contributing to epilepsy ⁸⁵.

Until recently, the mechanism(s) by which PI3K-PTEN alterations cause both HC and comorbid disorders such as ASD, epilepsy, and impaired cognition have remained elusive. Mice harboring de novo Pten mutations observed in patients with ASD and HC recapitulate the cerebral ventriculomegaly seen in humans ^{86, 87}. Pten-mutant mice developed aqueductal stenosis from proliferation of medial ganglionic eminence Nkx2.1 neural precursors and CSF hypersecretion ^{86, 87}. Unbiased brain network analysis identified similar neural networks in Pten-mutant mice with HC, reflecting diminished activity of Nkx2.1 NSC-derived inhibitory interneurons. These phenotypes were rescued by postnatal pharmacologic inhibition of mTORC1 with Everolimus, which restored cortical architecture and decreased ventricular size ^{86, 87}. These data allow proposal of a mechanism by which primary alterations in brain circuitry and HC are caused by genetic disruption of PI3K-PTEN signaling. They demonstrate a possible pharmacological strategy to ameliorate both circuit-level and ventriculomegaly phenotypes through shared mechanisms. Refined molecular characterization of HC endophenotypes will enable clinical application of such precision-based approaches.

ABERRANT CSF DYNAMICS, SECRETION, AND CHOROID PLEXUS DEVELOPMENT

Precise regulation of CSF dynamics plays an important role in cerebral cortical development, and impairment of CSF homeostasis has been implicated in the pathophysiology of HC. The ChP epithelium is the primary source of CSF ^{88, 89}, based on observations from as far back as 1918, when neurosurgeon Walter Dandy first described surgical removal of ChP to decrease CSF production for HC treatment ⁹⁰. Scarff later described endoscopic approaches to treat HC through ChP cauterization ⁹¹. However, the complex mechanisms of ChP-mediated ion and water transport remain debated ^{92, 93}. The tightly regulated, concerted action of multiple integral membrane ion transport proteins such as Na⁺/K⁺-ATPase, Na⁺ K⁺ Cl⁻ transporter NKCC1 and aquaporin water channels mediates transepithelial movement of solutes and water across the ChP epithelium ^{88, 92}. Other enzymes such as carbonic anhydrase secondarily contribute to mature ChP secretion without direct mediation of trans-membrane transport. CSF secretion is achieved by the polarized localization of transport proteins in the ChP epithelium. For example, CSF [Na⁺] is regulated by the Na⁺/K⁺-ATPase and NKCC1 localized atypically on the ChP apical membrane, in contrast to their classically basolateral localization in other epithelia ⁹⁴. Inhibitors of key ChP proteins such as ouabain (Na⁺/K⁺-ATPase), bumetanide and furosemide (NKCC1), and acetazolamide (carbonic anhydrase) decrease CSF secretion rates in vivo across model systems.

Increased cell mass or cell number due to ChP hyperplasia ⁹⁵, ChP papilloma ⁹⁶, and ChP carcinoma ⁹⁷ can augment CSF production and obstruct CSF flow ⁹⁸, leading to HC. Adult models of acquired HC secondary to hemorrhage ⁵, infection ⁹⁹, and tumor ¹⁰⁰ have demonstrated that increased CSF secretion in response to ChP inflammation is associated with up-regulated phosphorylation and activation of the SPAK kinase-regulated NKCC1 cotransporter, which exists in a protein complex with the Na⁺/K⁺-ATPase on the apical ChP membrane. Genetic or pharmacologic inhibition of Toll-like receptor 4 -associated ChP inflammation, as well as of the SPAK-NKCC1 complex, prevents development of ventriculomegaly in adult models of hemorrhagic and infectious HC ^{99, 101–103}. NKCC1 activity is required for CSF secretion by the adult ChP ^{99, 101, 102}, consistent with the decreased epithelial secretion and “slit-ventricle” brain morphology indicative of decreased intraventricular CSF volumes in patients with loss-of-function NKCC1 mutations ¹⁰⁴. However, the vector of net NaCl transport by NKCC1 (into or out of the ChP epithelial cell) reflects prevailing transepithelial ion gradients and its bidirectional and electroneutral transport mechanism ¹⁰⁵, but remains debated. In developing mouse CSF in which ion concentrations have not yet achieved adult levels, NKCC1 may also play a role in CSF K⁺ clearance ^{106, 107}. Interestingly, genetic overexpression of NKCC1 early in development through adeno-associated viral (AAV) transduction attenuates post-hemorrhagic HC and obstructive HC in neonatal mice, which may be secondary to clearance of K^{+ 107, 108}. Germline NKCC1 knockout animals display normal ventricular size ¹⁰⁹, underscoring the complex, species-specific nature of CSF volume regulation. The lack of identified mutations in the NKCC1-SPAK driven pathway in human genetic studies of HC is not surprising, as the pathophysiologic and mechanistic basis of acquired and congenital HC are distinct ⁶. Importantly, the redundant pathways, molecular machinery, and mechanisms regulating ChP ion transport remain incompletely characterized. These data suggest NKCC1 acts as a “rheostat” to mediate CSF homeostasis, albeit through incompletely understood mechanisms.

The molecular toolkit required to address fundamental questions in ChP biology has, until recently, been lacking. Large-scale developing brain atlases from multiple species, offer increased resolution of NSCs and the ChP across neurodevelopment and enable unbiased insights into the function of these cell types ^110–112. For example, FOXJ1 mutations may cause HC through disruption of ependymal cell differentiation and ciliary biosynthesis in the absence of obstruction of CSF flow ¹¹³. Although primary disruption in ciliary function due to impaired ependymal cell development can disrupt CSF bulk flow and the transepithelial ion transport required for CSF production ¹¹⁴, primary ciliopathies are not classically associated with HC in humans. Moreover, FOXJ1 in ChP is expressed predominantly at early developmental stages in non-motile, multi-ciliated cells ¹¹⁵, suggesting that defects in FOXJ1-dependent pathways may also alter CSF production and homeostasis. Radial glial over-production defects in SHH signaling ¹¹⁶ have also been shown to contribute to development of the hindbrain ChP ^{117, 118}, as well as to HC ¹¹⁹ and to HC-related macrocephaly ¹²⁰.

Single cell and spatial sequencing atlases ¹¹⁰ have identified a shared progenitor cell for both ChP epithelial and neuronal cells in early ChP development, confirmed by Wnt1 lineage tracing ¹¹⁰ and consistent with previous Nestin1 lineage tracing ¹²¹. Cell type composition, cell-cell interactions, and transition states shift considerably across the age spectrum. However, these analyses are currently restricted to the mouse brain, and the extent to which these paradigms apply to human ChP remains to be investigated. Further studies are needed to elucidate mechanisms of bi-directional CSF effects on neural development ⁸⁸. Creation of a human model system could help resolve these mechanistic questions.

Cerebral organoids as a model to elucidate mechanisms underlying HC

Many in vitro, ex vivo, and in vivo (Xenopus ^{35, 122, 123}, fish ^124–126, rat ^{101, 127}, mouse ^{128, 129}, and pig ¹³⁰) models have been developed to elucidate fundamental mechanisms underlying hallmark features of HC including, among others, alterations in ependymal barrier function ¹³¹, cortical development ¹³² and CSF dynamics ¹³³. In view of the varied mechanisms and pathways implicated in HC and its comorbid phenotypes, a human model system at least partially recapitulating these complexities may enable a more detailed and translationally relevant mechanistic understanding of HC. Moreover, a human model system could serve as a screening platform for pharmacologic development. Human iPSC-derived cerebral organoids, in vitro 3-dimensional structures resembling developing human brain, have recently emerged as a promising experimental tool to answer fundamental questions about normal and abnormal brain development and disease ¹³⁴. Cerebral organoids are self-organizing, recapitulate the spatial organization of the brain, and display transcriptional and epigenetic states that mirror brain development ¹³⁵. These organoids have been critical tools to aid our mechanistic understanding of HC-related phenotypes including microcephaly ¹³⁶, macrocephaly ¹³⁷, ASD ¹³⁸, host-pathogen interactions ¹³⁹, altered CSF production and selectively modified transport of small molecules ¹⁴⁰. Moreover, cerebral organoids form ventricular zones that resemble cerebral ventricles ¹⁴¹. Disruption of ZEB2, a gene crucial for neuroepithelial cell development, results in enlarged neuroepithelial buds resembling ventriculomegaly in cortical organoids ¹⁴¹. Leveraging this model system to study HC pathogenic features would be a logical next step to bridge the gap between mechanisms and patients.

Genetic forms of HC are driven primarily by disruption of the NSC compartment and subsequent defects in cortical growth, patterning, and development. Extended culture protocols will aid our understanding of the long-term consequences of genetically encoded early developmental dysfunction on downstream molecular consequences ¹⁴². Genetically tractable iPSCs can be used to understand the functional impact of patient-specific mutations using gene editing technologies. Such cell lines can also elucidate the consequences to cortical development of genetic errors at the NSC stage, and the impact on neuronal synaptic function through organoid-wide electrophysiological recording. Modulation of tissue morphology has also been shown to influence brain organoid development, a feature which may mirror the effect of permanent CSF shunting on brain structural architecture ¹⁴³. Important limitations of cerebral organoids include the absence of circulating immune cells, limited nutrient and oxygen diffusion through vascular networks, absent myelination, low reproducibility from batch effects, and variation in iPSC lines necessitating stringent controls ¹³⁵.

The mechanistically distinct acquired HC is driven by alterations in bulk CSF flow (production/absorption mismatch) and dysregulation of CSF dynamics. ChP-like cerebral organoids recapitulate the epithelial and stromal cell types observed in vivo. These organoids form a blood-CSF-barrier-like tight junction interface, secrete CSF-like fluid, and produce a protein signature like human CSF, recommending organoids for study of the effects of hemorrhage or infection on HC. Drawbacks to ChP organoids as an experimental model for acquired HC include the lack of a vascular network and circulating immune cells, lack of the physiological contribution of adjacent brain structure to CSF composition, and inability to analyze human ChP from surgical samples owing to safety concerns (for example, bleeding, stroke, etc.). Nonetheless, composition of CSF-like fluid from ChP organoids appears to resemble that of human CSF, and transcriptomic studies of iPSC-derived ChP mirror human ChP datasets ¹⁴⁰. Finally, ChP organoids have been used to study infectious diseases like SARS-CoV-2 ¹⁴⁴, highlighting their utility in understanding brain infections. Reconstitution experiments introducing genetically modified and/or additional cell types (such as microglia) into organoid cultures will provide a platform for well-controlled, and hypothesis-driven experiments not feasible in other systems. Rigor and reproducibility of such experiments will require the broader scientific community to adhere to strict, clearly defined nomenclature and methods of organoid generation, culture, and assembly ^{145, 146}.

Concluding remarks

We define the molecular hallmarks of human HC as a framework for future mechanistic studies, but many open questions remain. The most urgent unmet need is identification of therapeutic targets that reduce the requirement for and/or guide choice of neurosurgical interventions. We hypothesize these targets will emerge from: (1) careful study of cell-type and stage-specific roles of HC-associated genetic aberrations in human brain tissue; (2) molecular characterization of implicated genetic mechanisms in improved human and animal models; and (3) identification and validation of HC-associated genes in carefully phenotyped human cohorts. Such approaches will enable personalized characterization of genetic contributions to HC, will enhance validation of emerging pathogenic causes of HC in defined ancestral genomic backgrounds, and will provide resources for the broader HC research community.