Sequential emergence and contraction of epithelial subtypes in the prenatal human choroid plexus revealed by a stem cell model

Masters, Haley; Wang, Shuxiong; Tu, Christina; Nguyen, Quy; Sha, Yutong; Karikomi, Matthew K.; Fung, Pamela Shi Ru; Tran, Benjamin; Martel, Cristina; Kwang, Nellie; Neel, Michael; Jaime, Olga G.; Espericueta, Victoria; Johnson, Brett A.; Kessenbrock, Kai; Nie, Qing; Monuki, Edwin S.

doi:10.1038/s41467-025-60361-9

Download PDF

Article
Open access
Published: 03 June 2025

Sequential emergence and contraction of epithelial subtypes in the prenatal human choroid plexus revealed by a stem cell model

Nature Communications volume 16, Article number: 5149 (2025) Cite this article

3084 Accesses
1 Citations
Metrics details

Subjects

Abstract

Despite the major roles of choroid plexus epithelial cells (CPECs) in brain homeostasis and repair, their developmental lineage and diversity remain undefined. In simplified differentiations from human pluripotent stem cells, derived CPECs (dCPECs) display canonical properties and dynamic motile multiciliated phenotypes that interact with Aβ uptake. Single dCPEC transcriptomes over time correlate well with human organoid and fetal CPECs, while pseudotemporal and cell cycle analyses highlight the direct CPEC origin from neuroepithelial cells. In addition, time series analyses define metabolic (type 1) and ciliogenic dCPECs (type 2) at early timepoints, followed by type 1 diversification into anabolic-secretory (type 1a) and catabolic-absorptive subtypes (type 1b) as type 2 cells contract. These temporal patterns are then confirmed in independent derivations and mapped to prenatal stages using human tissues. In addition to defining the prenatal lineage of human CPECs, these findings suggest dynamic models of ChP support for the developing human brain.

Choroid plexus apocrine secretion shapes CSF proteome during mouse brain development

Article Open access 28 May 2025

Modeling preeclampsia using human induced pluripotent stem cells

Article Open access 15 March 2021

AUTS2 disruption causes neuronal differentiation defects in human cerebral organoids through hyperactivation of the WNT/β-catenin pathway

Article Open access 22 August 2024

Introduction

Inside the brain’s ventricles, the choroid plexus (ChP) lies at the interface between two circulating fluids¹ – the blood and cerebrospinal fluid (CSF). At this interface, the ChP receives a disproportionate share of cerebral blood flow^2,3 via a rich network of fenestrated capillaries¹ and mediates a robust two-way exchange of blood and CSF components. In turn, the ChP produces the CSF (400–600 mL per day in humans)¹, which equilibrates with interstitial fluid of the brain and spinal cord across permeable linings (pia and ependyma) and via glymphatics⁴. CSF then returns to the peripheral circulation via arachnoid villi, granulations, and lymphatics in olfactory, dural, and basal regions⁵. These circulations and equilibria give the ChP and CSF access to CNS cells behind the blood-brain barrier, providing the rationale for CSF-based delivery of neurotherapeutics that cannot cross this barrier^6,7.

ChP functions at the blood-CSF interface are largely executed by its epithelial cells (CPECs)¹. Among their roles, CPECs produce the CSF, secrete most of the CSF proteome, gate and transport molecules and immune cells between the blood and CSF, detoxify the circulating fluids, and form the anatomical blood-CSF barrier. Despite this broad physiologic and therapeutic relevance, remarkably little is known about human CPECs. The myriad of secretory and absorptive functions attributed to CPECs are thought to be carried out by a single epithelial cell type¹ that advances through developmental stages and asynchronous phases (“dark” and “light” cells^8,9). Among CPECs, lineage differences relative to the embryonic roof plate^10,11,12, and molecular differences based on ventricular origin have been described¹³, and recent studies from mice¹⁴ and from human ChP organoids¹⁵ suggest a mammalian CPEC subtype specialized for ciliogenesis. However, formal delineation of CPEC subtypes and their lineage relationships remain lacking.

To address these basic and applied gaps, we previously developed methods^16,17 to generate derived CPECs (dCPECs) from mouse embryonic stem cells (ESCs) using an aggregate-organoid approach. (Note: “dCPEC” is used when referring specifically to the derived cells in vitro). Together with studies from mouse explants¹⁸, the earlier derivations established fundamental developmental principles in mouse CPEC development, including BMP4 sufficiency as a CPEC morphogen^16,17 and the temporal restriction of CPEC competency to pre-neurogenic neuroepithelial cells (NECs) rather than radial glia¹⁶. For human dCPECs, we used human ESCs and the neural rosette method¹⁹ to establish initial proof-of-concept¹⁶, but this method had limited utility due to its high complexity, inefficiency, and inconsistency.

In this study, we describe and validate a human dCPEC protocol with enhanced simplicity, efficiency, consistency, and scalability. Using this protocol, we then define multiple forms of CPEC diversity in humans. We first describe reciprocal multiciliated and Aβ uptake phenotypes, as well as cilia motility, which change over time. We then use single cell RNA sequencing (scRNA-seq) across timepoints in vitro to define a human CPEC lineage tree with two bifurcations. Using independent derivations and perinatal tissues, we confirm and map the two bifurcations to midgestation and preterm periods in humans, then discuss the implications of this dynamic lineage to human brain development via two models.

Results

Simple, efficient, and accelerated human dCPEC generation from H1 ESCs

Our proof-of-concept method using human ESCs (H1 cells) yielded relatively few dCPECs (5–10%)¹⁶. To improve efficiency and simplicity, we optimized a feeder-free monolayer system involving the seeding of small ESC clumps on Matrigel followed only by media changes (Fig. 1A, B; Supplementary Fig. 1A, B). Compared to an alternative high-density method, which also worked efficiently (Supplementary Fig. 1C), the low-density seeding protocol provided ~30X greater scalability and up to 30 million dCPECs from 2 million starting ESCs.

**Fig. 1: dCPEC derivation from H1 ESCs.**

To accelerate dCPEC generation, “rapid” neural induction media (NIM; Thermo Fisher) was introduced after ESC colonies reached 150–200 um in diameter (typically 2–3 days). As advertised, NIM led to ~100% dual positivity by immunocytochemistry (ICC) for neural progenitor markers (NESTIN and SOX2) and negativity for the pluripotency marker OCT4 by 7 days (Supplementary Fig. 1D). Timing and duration of NIM and BMP4 (10 ng/mL) were then co-optimized. Consistent with the early and short window of NEC competency for mouse CPEC fate¹⁶, BMP4 addition 1 day in vitro (1 div) after NIM initiation, but not before (0 div) or after (2 div), resulted in substantial dCPEC induction (Fig. 1C). With BMP4 addition at 1 div, NIM for 5 days was most effective (Fig. 1D), after which the media was switched to CPEC media with BMP4. Like the mouse cells¹⁶, induction efficiency was similar over a range of BMP4 exposure times (10–30 days; Supplementary Fig. 1E); 15 day BMP4 exposures were used for subsequent studies. Using this protocol, H1 ESCs, via an NEC progenitor, gave rise to dCPEC sheets and islands that routinely covered 40% or more surface area by 5–6 weeks (Fig. 1E; Supplementary Fig. 1F) and could remain adherent and viable past 1 year.

dCPEC validations of canonical CPEC properties and functions

Like mouse dCPECs¹⁶, human dCPEC islands formed three dimensional (3D) folds²⁰ and ridges over time and expressed established markers of several CPEC compartments by ICC (secretory apparatus/TTR²¹, apical membrane/AQP1²², basolateral membrane/AE2²³, cilia/ARL13B²⁴, nucleus/OTX2²⁵, tight junctions/CLDN1 and CLDN5²⁶) in a stereotypical order (Fig. 1F; Supplementary Fig. 1G). The dCPECs also had a uniform apicobasal polarity, a well-known property of derived¹⁶ and in vivo CPECs¹ (Supplementary Fig. 1H), with apical AQP1+ membranes facing the media while basolateral AE2+ membranes faced Matrigel or abutted one another within 3D folds (Fig. 1G). Comparable derivation efficiencies were achieved with multiple iPSC lines (Supplementary Fig. 1I–K). As in mouse cells²⁷, early cell cycle withdrawal was evident in human dCPECs, with relatively low EdU labeling after 21 div (Supplementary Fig. 1L) and almost none after 31 div (Fig. 1H).

Consistent with the high mitochondrial content of CPECs in vivo²⁸, dCPECs displayed high mitochondrial mass (ATPB; Supplementary Fig. 1M) and activity (MitoTracker Orange; Fig. 1I) compared to neighboring cells. CPECs are also well known to produce and secrete TTR via the classical secretory pathway²⁹. By 1–2 h after washouts, TTR was readily detectable in culture media by ELISA (Fig. 1J) and blocked over a wide concentration range (0.1–10 ug/mL) of Brefeldin A (BFA), which inhibits the classical secretory pathway³⁰. BFA blockade was also fully reversible (Fig. 1J), ruling out toxicity as an explanation for the TTR reductions.

Dynamic multiciliated CPEC phenotypes in vitro and in vivo

CPECs are relatively unique among CNS cells in being multiciliated³¹. Even at low magnification, the cilia marker ARL13B readily distinguished monociliated non-dCPECs and neural rosettes³² (Supplementary Fig. 2A) from the islands of dCPECs with multiple apical cilia per cell (Fig. 2A, B). Qualitative impressions of cilia mass increases per dCPEC between 38 and 80 div (Fig. 2B; Supplementary Fig. 2B, C) were confirmed using a metric based on ARL13B and ZO1 staining (Fig. 2C; Supplementary Fig. 2D; see Methods). In addition, cilia occupied increasing percentages of dCPEC surface area over this time period (Fig. 2B, D).

**Fig. 2: Dynamic and interacting multiciliated and Aβ_1-42 uptake phenotypes.**

To assess cilia distributions in vivo, we evaluated whole mount preparations of human postmortem ChP using the same two antibodies (ARL13B and ZO1). Around midgestation (23 postconceptional weeks/pcw), multiciliated CPECs were abundant, and their cilia were generally clustered. Near term (38 pcw) and into early adulthood (19 years of age), CPECs remained multiciliated, but their cilia were widely dispersed (Fig. 2E; Supplementary Fig. 2E). Thus, cilia clustering and dispersion in vivo paralleled that seen in dCPECs over time in vitro.

Interacting multiciliated and Aβ uptake phenotypes

Robust uptake (absorption) is another fundamental property of CPECs¹. Uptake of Aβ peptides associated with Alzheimer’s disease has been demonstrated for rat ChP³³, but not yet for human CPECs. After treatment with fluorescently-tagged Aβ_1-42 (AnaSpec), fluorescence in 64-div dCPECs was evident within 30 min, peaked by 6 h, then decreased markedly by 24 h (Fig. 2F, G; Supplementary Fig. 2F) across a range of concentrations (500nM-4uM). Aβ_1-42 fluorescence was intracytoplasmic (Supplementary Fig. 2G) within EEA1+ endosomes (Supplementary Fig. 2H), indicating rapid and robust Aβ_1-42 uptake into the early endosomes of human dCPECs. EEA1 colocalization also decreased over time (Supplementary Fig. 2H, I), suggesting Aβ_1-42 transfer from early endosomes to other intracellular compartments.

Across concentrations and timepoints, intracellular Aβ_1-42 appeared as variably-sized puncta and often in circular arrangements (Fig. 2F). ARL13B costaining revealed cilia at the center of these arrangements with Aβ-positive dCPECs having cilia that were generally more clustered (Fig. 2H). 1 uM or 2 uM Aβ_1-42 application itself resulted in dCPECs having more clustered cilia than vehicle controls, with peak clustering 1 h after application (Fig. 2I) and less by 2 h (Supplementary Fig. 2J). Moreover, Aβ_1-42 signal intensities were inversely correlated with cilia dispersion (Fig. 2J). These findings suggest reciprocal interactions between cilia clustering and Aβ_1-42 uptake in human dCPECs.

Direct dCPEC origin from tripotent neuroepithelial cells: Pseudotemporal scRNA-seq analysis

We then acquired four scRNA-seq datasets from 31 div (after apparent dCPEC cell cycle exit; Fig. 1H) to 75 div when dCPEC changes in morphology and immunocytochemical profile appeared to decelerate (Fig. 1F). Timepoints were paired (31–46 and 55–75 div pairs; Fig. 3A) for culturing and processing through a 10x Genomics-Illumina pipeline to reduce batch effects, which we were unable to discern across the four datasets (Supplementary Fig. 3A, B). Datasets were processed primarily through SoptSC, which performs unsupervised clustering, pseudotemporal ordering, and lineage inference in parallel³⁴, as well as Seurat³⁵.

**Fig. 3: Direct dCPEC origin from tripotent NECs and comparison to human organoid and fetal CPECs.**

SoptSC analyses suggested six cell clusters at 31 div (Supplementary Fig. 3C), which increased to eight by 75 div (Fig. 3B). Differentially-expressed genes (DEGs) and cell-type markers^13,14,15,16 identified single clusters of dCPECs, neurons, and neuroepithelial cells (NECs) along with multiple clusters of neural progenitors – three at 31 and 46 div, four at 55 div, and five at 75 div (Fig. 3B, C). The same populations, with similar DEGs, were identified by Seurat (Supplementary Fig. 3D–F). All defined cell clusters were neural in origin; no non-neural cell clusters were identified in any dataset, underscoring the efficiency and consistency of neural induction and differentiation using the simplified protocol.

Consistent with NECs being stem progenitors, NEC fraction was highest at the earliest time point (31 div), while dCPEC and neuron fractions peaked at 46 div before declining (Fig. 3C). At 46 div, dCPECs comprised ~30% of cells. By 55 and 75 div, dCPEC fractions decreased as additional classes of proliferating neural progenitors appeared, expanded, and contracted in an orderly fashion (Fig. 3C).

Pseudotemporal analyses of individual timepoints suggested three classes of NEC progeny. At all four timepoints, dCPECs arose directly from NECs as a direct and distinct end-branch (pink TTR-expressing cluster in Fig. 3D–F), as did neurons. Neural progenitors constituted the third progeny class, which had more varied relationships to each other and to NECs (Fig. 3D,F) consistent with their significant transcriptomic similarities overall. Among the three classes of NEC progeny (Fig. 3G), dCPECs had a consistently direct lineage relationship to NECs based on pseudotemporal ordering.

Orderly dCPEC transcriptome changes compared to human fetal and organoid CPECs

The four scRNA-seq datasets were then compared to published human ChP organoid (55 div)¹⁵ and fetal CPECs (8 and 14 pcw)³⁶. With batch correction by reciprocal principal component analysis (RPCA)³⁷, CPECs from all sources clustered together and away from neurons and NECs by UMAP (Supplementary Fig. 4A) and principal component analysis (PCA) (Supplementary Fig. 4B). Batch correction using SCTransform³⁸ increased the differences among CPEC clusters, which nonetheless continued to cluster and correlate well by UMAP and PCA (Fig. 3H, I). Pearson correlations between sources, using either batch correction method, supported dCPEC identity (Fig. 3J; Supplementary Fig. 4C, D). These analyses also revealed increasing dCPEC-to-fetal CPEC correlations with increasing time in vitro (Fig. 3H–J; Supplementary Fig. 4D), while the organoid CPECs displayed an intermediate differentiation profile (Fig. 3I, J) consistent with their intermediate derivation time (55 div)¹⁵.

dCPEC subtypes and bifurcations: Pseudotemporal and time-series scRNA-seq analyses

Cells comprising the dCPEC lineage (dCPECs and NECs) were then re-run through SoptSC (Supplementary Fig. 5A, B). At the two early timepoints (31 and 46 div), two dCPEC sub-clusters were evident (Fig. 4A–C). The sub-cluster with greater similarity to NECs was designated CPEC “type 1” (C1); the other sub-cluster was designated CPEC “type 2” (C2) (Fig. 4B; Supplementary Fig. 5C). C2 fraction was highest at 31 div (~40% of dCPECs compared to ~10% at 46 div) (Supplementary Fig. 5D), and C2 cells were present at the two later timepoints (Fig. 4D; Supplementary Fig. 5F–H), but were too few in number to define as clusters computationally (see Methods).

**Fig. 4: SoptSC and STITCH analyses across timepoints suggest a branching dCPEC lineage tree.**

Type 1 dCPECs were present at all four timepoints, but branched into two sub-clusters in the last 75-div timepoint (Fig. 4A–E). The sub-cluster with higher Pearson correlation (98%) to 55-div C1 cells was designated CPEC “type 1a” (C1a); the other sub-cluster (82%) was designated CPEC “type 1b” (C1b) (Supplementary Fig. 5I). These correlations suggested the emergence of a distinct C1b subtype from the C1-C1a lineage. Similar sub-clusters and pseudotemporal relationships were obtained using Seurat (Supplementary Fig. 5J–L).

For time-series analysis, dCPEC and NEC sub-clusters were analyzed using STITCH³⁹. The roughly-triangular STITCH dot plots (Fig. 4F–H; Supplementary Fig. 6A) suggested one broad dCPEC and two broad NEC groups. The dCPEC group contained dCPECs from all four timepoints (Fig. 4F–H; Supplementary Fig. 6A). Additional STITCH analyses illustrated clear C1-C2 and C1a-C1b separations, supporting their designations as distinct subtypes (Supplementary Fig. 6B). STITCH lineage trees were then formally constructed. At all timepoints, C2 cells were distinguished from the C1 lineage and NECs. The C1 lineage spanned all four timepoints, then bifurcated into C1a and C1b subtypes at the last timepoint (Fig. 4F–H, lower panels). Taken together, the SoptSC and STITCH analyses suggest two temporally-separated bifurcations within a branched dCPEC lineage tree.

We explored whether previously-described differences among CPECs could account for the dCPEC subtypes. CPECs in mice^11,12 and humans⁴⁰ have differing lineage relationships with the embryonic roof plate, but roof plate markers¹¹ were not selectively associated with a dCPEC branch or subtype (Supplementary Fig. 6C). Distinct anteroposterior (ventricular) CPEC identities have been described in mice^13,14, but all dCPEC sub-clusters were mixtures of cells with lateral ventricle (LV), fourth ventricle (4V), or unspecified identities (Supplementary Fig. 6E). The favoring of LV over 4V identity (Supplementary Fig. 6E) would be consistent with the “default” model of neural induction, which favors anterior specification⁴¹, as well as the lack of Sonic hedgehog (SHH) agonism in our protocol, which is implicated in the specification of 4V ChP identity⁴². Regardless, neither roof plate lineage nor ventricular identity explained dCPEC branches or subtypes.

Unique dCPEC ‘G0’ signature and direct origin from tripotent NECs: Cell cycle analyses

C1a-C1b lineage bifurcation in vitro occurred well after the apparent cell cycle withdrawal of dCPECs (Fig. 1H). Using Seurat³⁵, which uses S, G2/M, and G0/G1 phase gene lists from mice⁴³, dCPECs displayed G0/G1 or G1/G2M signatures (Supplementary Fig. 7D); the G1/G2M dCPECs correspond to C2 cells due to the extensive overlap between ciliogenesis and G2/M genes⁴⁴ (see next section for more on ciliogenesis). Individual dCPECs from all timepoints formed a cluster distinct from NECs, neurons, and neural progenitors by UMAP (Supplementary Fig. 7D). When grouped by timepoint, dCPEC clusters separated from the other cell types by PCA (Supplementary Fig. 7E). Interestingly, the three branches of NEC progeny (Fig. 3) were also clearly delineated by this cell cycle transcriptomic analysis alone (Supplementary Fig. 7E-F).

We then used human cell cycle genes that distinguish a ‘neural’ G0 phase from G1⁴⁵. The dCPECs from all stages had predominant G0 signatures, as did neurons (Fig. 5A, B; Supplementary Fig. 7A, B), but dCPECs and neurons were clearly distinguished from one another, as well as from NECs and neural progenitors, by UMAP or PCA (Fig. 5A–D). Some C2 cells again displayed a G2M signature that nonetheless clustered with C1 cells (Fig. 5B,D). Notably, as with Seurat, human neural G0 analysis alone delineated the three progeny classes of tripotent NECs (Fig. 5C–E; Supplementary Fig. 7C). Neural G0 analysis of the earliest dataset alone (31 div) delineated these classes (Fig. 5F).

**Fig. 5: Cell cycle analyses highlight NEC tripotency and reveal a distinctive ‘neural G0’ dCPEC signature.**

To confirm their non-cycling status, we stained derived and perinatal CPECs for KI67, which labels cycling cells⁴⁶. At a derivation timepoint (49 div) before C1a-C1b lineage bifurcation (Fig. 4H), KI67 clearly stained non-CPECs, but neither dCPECs nor neurons (Fig. 5G; Supplementary Fig. 7I). Similarly, in midgestation tissue (23 pcw) before C1a-C1b lineage bifurcation (see next section), KI67-positive cells could be seen in ChP stroma, but not in CPECs (Fig. 5H).

Characterizations and confirmations of early type 1 and type 2 CPEC subtypes

All of the shared KEGG pathways⁴⁷ and GO terms⁴⁸ enriched in the top-300 DEGs of C1 and C2 cells at 31 div (when C2 cells were most abundant) were neurodegenerative diseases (Alzheimer, Huntington, Parkinson, and prion diseases; Fig. 6A). Selective C1 enrichments were associated with metabolism (e.g., metabolic pathways, non-alcoholic fatty liver disease) and energy (e.g., oxidative phosphorylation, thermogenesis), while C2 enrichments centered on ciliogenesis and cilia maintenance (Fig. 6A, B). Inspection of the top-10 individual DEGs at 31 and 46 div further highlighted the strong ciliogenesis signature of C2 cells (Fig. 6C).

**Fig. 6: Early type 1 and type 2 CPECs, with later type 2 contraction.**

Multiciliated CPECs in zebrafish and mice are motile^49,50, and time-lapse videomicroscopy revealed human dCPECs with highly motile cilia (Supplementary Movies 1, 2). Tracking of individual 42-div dCPEC cilia revealed heterogeneity in beat orientations, magnitudes (Fig. 6D), and frequencies (0–10 Hz). In contrast, cilia motility was negligible in otherwise healthy-appearing dCPECs by 110 div (Fig. 6E; Supplementary Movie 3). In transmission electron micrographs of 40-div dCPECs, 103 of 131 identified ciliary axonemes possessed the central pair of singlet microtubules that distinguishes the 9 + 2 microtubule arrangement of motile cilia⁵¹ (Fig. 6F) along with other ultrastructural features of multiciliation (Supplementary Fig. 8I).

Master regulators of motile ciliogenesis⁵², particularly FOXN4⁵³ and MCIDAS⁵⁴, were collectively higher in C2 compared to C1 cells at all four timepoints (Fig. 6G, H; Supplementary Fig. 8F). Secondary analysis of scRNA-seq data from human ChP organoids¹⁵ also identified a CPEC subpopulation that coexpresses FOXN4 and MCIDAS (Supplementary Fig. 6F). FOXN4 and MCIDAS protein expression were then examined in independent derivations. Both were expressed at high (Supplementary Fig. 8A) and highly-correlated levels (Supplementary Fig. 8E) in early dCPECs (31 div) with FOXN4 demonstrating particularly discrete “salt and pepper” expression (Fig. 6I; Supplementary Fig. 8A, B). Over time, FOXN4-expressing C2 cells remained detectable, but significantly reduced in fractional positivity and expression level (Fig. 6I; Supplementary Fig. 8C). Likewise, MCIDAS expression levels decreased over derivation time (Supplementary Fig. 8A).

We then examined perinatal ChP tissues using the same two antibodies (FOXN4 and MCIDAS). Around midgestation (23 pcw), MCIDAS levels correlated positively with those of FOXN4 (Supplementary Fig. 8F) with MCIDAS demonstrating particularly discrete and mosaic expression in a substantial CPEC fraction (Fig. 6J). Near term (42 pcw), MCIDAS-expressing C2 cells were detectable, but significantly reduced in fractional positivity and expression level (Fig. 6J; Supplementary Fig. 8D). Similarly, FOXN4 expression levels decreased from midgestation to term (Supplementary Fig. 8G). Consistent with the scRNA-seq and ICC findings in vitro, the perinatal tissues demonstrate higher type 2 CPEC prevalence at midgestation, which decreases significantly by term.

Characterizations and confirmations of late type 1a and type 1b CPEC subtypes

Based on the top-300 C1a and C1b DEGs, C1b cells were uniquely enriched for endocytotic, stress response, and catabolic KEGG pathways (Fig. 7A) and Gene Set Enrichment Analysis (GSEA)⁵⁵ terms (Fig. 7B). Conversely, while also sharing several pathways with the earlier-stage C1 cells (Fig. 7A–C), C1a cells were selectively enriched for several secretory, water transport, and anabolic pathways based on manually-curated GSEA gene lists and GSEA plots (Supplementary Fig. 9A). GSEA analyses also revealed additional enrichments in C1b cells, such as immune cell signaling interactions and steroid hormone biosynthesis (Supplementary Fig. 9B; Supplementary Data 1). C1a and C1b cells continued to share neurodegenerative disease pathways (Fig. 7A). Pearson correlations between top C1a and C1b DEGs were similar between 31 and 46 div, increased for a subset at 55 div, then strongly diverged by 75 div (Fig. 7D, E). Thus, 55-div C1 cells had features of a primed or mixed intermediate in normal development and transdifferentiation⁵⁶. Re-examinations of the human fetal³⁶ and organoid datasets¹⁵ did not reveal C1a and C1b subtypes (Supplementary Fig. 6F) consistent with C1a-C1b emergence after the stages sampled in these studies.

**Fig. 7: Late type 1a and type 1b CPECs with prenatal lineage summary.**

Specification of C1a and C1b subtypes was then examined in paired derivations and human tissues using antibodies against CLDN5 (C1a) and SC5D (C1b) (Fig. 7E). While also known for its expression by endothelial cells, CLDN5 has been described in human⁵⁷ and zebrafish CPECs⁵⁸, although not in mouse CPECs¹⁴, raising the possibility of species-specific differences. At 61 div, correlation between CLDN5 and SC5D trended positively, then became significantly negative by 79 div (Fig. 7F–H; Supplementary Fig. 9D–H, L). By 79 div, C1a and C1b subtypes also appeared clustered rather than randomly distributed (Fig. 7F, G). Similarly, in the human tissues, CLDN5 and SC5D expression were positively correlated at 23 pcw, then became negatively correlated by 42 pcw (Fig. 7I–K; Supplementary Fig. 9I–K,M) and also appeared clustered (Fig. 7I, J; Supplementary Fig. 9J). Together with the scRNA-seq analyses, these studies indicate a C1a-C1b lineage bifurcation that occurs between 61–75 div in vitro and between 23–42 pcw in vivo.

Discussion

In this report, we present an improved derivation of human CPECs, which differentiate with substantial cell autonomy and confirm several developmental principles from other species and systems^1,16 (Fig. 1; Supplementary Fig. 1). The human dCPECs also have dynamic multiciliated phenotypes that reciprocally interact with Aβ uptake (Fig. 2; Supplementary Fig. 2), while pseudotemporal, time series, and cell cycle scRNA-seq analyses reveal their direct origin from tripotent NECs as well as substantial CPEC subtype diversity (Figs. 3–5; Supplementary Figs. 3–7). Non-cycling type 1 (metabolic-energetic) and type 2 CPECs (ciliogenic) are present early in vitro and at midgestation in vivo (Fig. 6; Supplementary Fig. 8). By term, type 2 CPECs contract (Fig. 6; Supplementary Fig. 8) while type 1 CPECs bifurcate into type 1a (anabolic-secretory) and type 1b subtypes (catabolic-absorptive) (Fig. 7; Supplementary Fig. 9). Transcriptomic comparisons within this human NEC-CPEC lineage model (Fig. 7L) suggest the ontogenetic and phylogenetic emergence of type 2 and type 1b subtypes from a primordial NEC-type 1-type 1a lineage.

Molecular, cellular, functional, and transcriptomic findings established the identity and quality of dCPECs in the derivation protocol, which is more efficient, reproducible, scalable, and simple than our earlier methods^16,17. In addition, dCPEC differentiation is relatively rapid and associated with phase-contrast milestones that are predictive and straightforward to assess (Fig. 1B; Supplementary Fig. 1). Published CPEC derivation methods^15,59,60 vary in their apicobasal topologies - i.e., whether apical or basolateral CPEC surfaces face the media. Using this protocol, dCPEC apical surfaces are media-facing (Fig. 1G; Supplementary Fig. 1J); culture media therefore corresponds to the CSF compartment, which can be advantageous. While BMP4 suffices for human dCPEC induction, others have coapplied BMP4 with a WNT signaling activator (CHIR 99021) to induce CPEC fate^15,59,60. In our hands, BMP4 alone induced dCPEC WNT expression (Supplementary Fig. 6C), and CHIR 99021 provided no obvious additional benefit.

The early BMP4 requirement (Fig. 1C) for the direct NEC-CPEC lineage and its diversification (Figs. 3–7) support NECs being the CPEC-competent progenitor rather than radial glia^16,61, BMP4 sufficiency as a CPEC morphogen^16,17,18, the cell-intrinsic nature of NEC responses to BMP4^18,62, and the significant cell autonomy of CPEC differentiation^16,17, including subtype diversification. At a minimum, absence of non-neural cells in the cultures (Fig. 3; Supplementary Fig. 1D) indicates autonomy from mesenchymal, vascular, blood, and blood-borne influences, while the prominent 3D dCPEC folding (Fig. 1B, G; Supplementary Fig. 1G, K) is consistent with ChP morphogenesis being driven by CPECs rather than mesenchyme⁶³. Early dCPEC differentiation (Fig. 3) and cell cycle exit (Figs. 1H, 5) also support CPECs being among the earliest differentiating cells in the human brain^11,40,64. The presence of light and dark dCPECs (Supplementary Fig. 8I) suggests the autonomy of these ultrastructural and likely phasic, rather than developmental, differences among CPECs^8,9. Our protocol also provides some insights, albeit limited, into neural induction. Absence of dCPECs upon BMP4 coapplication with NIM (0 div) is consistent with inhibition of BMP signaling being required to initiate neural induction, while the high efficiency of neural induction upon early BMP4 addition after NIM (1 div) indicates that complete BMP signaling inhibition may not be required to establish and maintain neural fate.

BMP4 and MSX1 gene expression by dCPECs^18,62 well after the cessation of exogenous BMP4 application (Supplementary Fig. 6C), the autocrine/paracrine BMP signaling by dCPECs predicted by CellChat⁶⁵ (Supplementary Fig. 6D), and dCPEC island formation (Fig. 1; Supplementary Fig. 1) suggest that human CPEC induction is homeogenetic (“like inducing like”), as postulated over a century ago⁴⁰ and as described for the roof plate^66,67 and telencephalic CPECs in mice^11,12. BMPs, and BMP4 in particular, have been implicated in these homeogenetic phenomena, which expand the space and time over which morphogens can act during development⁶⁸. By 55 div, type 1 dCPECs have a mixed identity (Figs. 4B, 7D; Supplementary Fig. 5H) typical of proliferative intermediates in normal development and transdifferentiation⁵⁶ (Fig. 7D), but were non-cycling G0 cells (Figs. 1H, 5). Thus, unlike the irreversible G0 of most neurons⁶⁹, the G0 state of type 1 CPECs is presumably reversible.

The variably-high motility of early-stage dCPEC cilia (Fig. 6D; Supplementary Movies 1, 2) diminishes over time in vitro (Fig. 6E; Supplementary Movie 3). Zebrafish⁴⁹ and mice⁵⁰ also have motile CPEC cilia, and interestingly, this motility in mice diminishes postnatally⁵⁰. For human dCPECs, decreased motility coincides with the loss of cilia clustering in vitro and in vivo (Fig. 2B–E; Supplementary Fig. 2B, C, E). We found no literature precedent for such losses, but for multiciliated cells, cilia motility is strongly associated with cilia clustering⁷⁰. As in mice⁵⁰ and along with type 2 subtype contraction (see below), our motility and cilia clustering findings portend transience of CPEC cilia motility in humans.

Conversely, the enhancement of cilia clustering by exogenous Aβ_1-42 (Fig. 2I; Supplementary Fig. 2J) raises the possibility that Aβ uptake promotes cilia motility. To our knowledge, this study provides the initial demonstration of Aβ uptake by human CPECs, a potential Aβ clearance mechanism involved in Alzheimer’s disease initiation and pathogenesis⁷¹. Preferential Aβ uptake into early endosomes (Supplementary Fig. 2H, I) surrounding clustered cilia (Fig. 2H) suggests that CPEC cilia serve as docking sites for uptake, as seen in other cell types^72,73. While Aβ dispositions following this uptake remain to be fully described, these findings – together with the enrichment of Alzheimer’s and other neurodegenerative disease pathways in all dCPEC subtypes (Figs. 6A, 7A) – raise the possibility that CPECs and the ChP contribute to the developmental origins of Alzheimer’s disease^74,75 and potentially other neurologic and neuropsychiatric disorders that manifest later in life^76,77.

While dCPECs are generally multiciliated (Fig. 2A–E) and both type 1 and type 2 CPECs are enriched for ciliogenesis genes compared to non-dCPECs (Supplementary Fig. 8F), ciliogenesis gene enrichment further distinguishes type 2 cells (Fig. 6A–C). Type 2 CPECs correspond to the ciliogenesis-enriched CPECs described in developing mice¹⁴ and human ChP organoids¹⁵, with our study adding their early emergence, then contraction during prenatal human development (Figs. 4, 6) as well as the cilia dynamics just discussed. The prenatal contraction of type 2 CPECs draws comparisons to other transient cell types of the developing brain, such as Cajal-Retzius cells⁷⁸ and subplate neurons⁷⁹, and adds to the cell types with unique roles in early brain development. Our lineage analyses do not resolve whether NECs give rise to C1 and C2 cells directly or via a bipotential C1-C2 intermediate, or the symmetry/asymmetry of this event(s). Although pseudotemporal analyses did not reveal such an intermediate, C1 and C2 clusters were already present at the earliest stage profiled. Thus, additional studies will be needed to further resolve this initial CPEC lineage branch.

Along with the transience of cilia motility and clustering, contraction of ciliogenic type 2 cells suggests a conserved CPEC-to-ependymal transition in multiciliated ventricular cells that promote CSF flow and/or mixing. Nonami et al.⁵⁰ noted that mouse CPEC cilia motility—which peaks around birth before declining over the first two postnatal weeks—precedes the emergence of ependymal cells⁸⁰, whose cilia propel CSF⁸¹, but do not mature until weeks after birth^82,83. Likewise, in humans, ependymal multiciliation and maturation are largely postnatal and continue into adolescence⁸⁴, while CPEC differentiation⁶⁴, CPEC cilia clustering (Fig. 2B–E; Supplementary Fig. 2B, C, E), and type 2 cells (Figs. 4, 6) are evident much earlier in utero. The prominent “bloom” of telencephalic ChP in cynomolgus monkeys and early first-trimester humans^64,68,85 would further facilitate CPEC-mediated CSF flow and/or mixing before ependymal cells are able to.

Type 1a-1b bifurcation does not occur until later in vitro or near term in vivo (Figs. 4, 7), and this late prenatal timing suggests preparation for birth as the ontogenetic driver. Other than conception and death, birth is the most significant physiologic transition in human ontogeny due to the abrupt separation and loss of maternal, placental, and amniotic fluid supports for fetal energy flux (energy substrate delivery, usage, and waste clearance)^86,87. Two well-known adaptations that support this dramatic transition in energy flux are the specification of surfactant-producing alveolar type II cells (AT2 or AEC2) in the lung and globin gene switching in the bone marrow. Both initiate during the late prenatal period with autonomy from the human birth process itself ^88,89 and directly impact the fluids for energy flux (air and blood) in neonates and adults - i.e., transport of oxygen, glucose, other energy sources, carbon dioxide, heat, and other forms of metabolic waste to and from tissues and the body.

Notably, CPECs also produce and directly regulate a fluid (CSF), albeit one that selectively subserves the brain. Hence, type 1a and 1b CPECs likely evolved to meet the extrauterine energy flux demands of the human brain, including the replacement of in utero functions provided by the placenta, which shares many of the same top enriched pathways as type 1b cells (e.g., waste removal, immune regulation, and steroid hormone production; Supplementary Fig. 9B)⁹⁰. While AT2 cells and globin gene switching are conserved among mammals^91,92, type 1a and 1b CPECs have not been described previously. As mentioned, the human fetal³⁶ and ChP organoid datasets¹⁵ lack these subtypes (Supplementary Fig. 6F), but were sampled before type 1a-1b specification occurs. Regardless, while additional studies will be needed to define CPEC subtype phylogeny, brain phylogeny provides ample rationale for the ontogeny and phylogeny of unique CPEC subtypes in humans. Like AT2 cells and globin genes, fundamental CPEC roles center on energy flux and homeostasis needs, and these needs became massive in human brains. The absolute and relative expansion of hominin and human brains, especially their neocortices^93,94, includes an exponential relative burst in Homo sapiens^93,95 and associated changes in energy allocation that were just as dramatic. Despite representing only 2% of body weight, the adult human brain uses 20–25% of the energy budget compared to 8–10% in non-human primates and 3–5% in other mammals⁹⁶, with the neonatal human brain expending a remarkable 60–65%⁹⁷.

While CPEC subtype evolution to support the increased energy flux demands of human brains has not been described previously, molecular and physiological system adaptations have been. Notably, many of these rely on CPECs for their implementation in the brain. These adaptations include the thyroid hormone (TH) system, which also co-evolved with human encephalization⁹⁸; the basal metabolic rate⁹⁹ regulated by TH¹⁰⁰; and the TH carrier and distributor system¹⁰¹, including TTR^102,103, the most abundant CPEC product¹³ whose expression is enriched in type 1 and type 1a CPECs (Fig. 4C). These and other evolutionary adaptations would align with the preterm specification of specialized CPEC subtypes to support the uniquely energy-demanding human brain.

Methods

Ethical regulation of research

These studies were carried out under the supervision and approval of local ethics committees (UCI IBC, IRB, and hSCRO). Postmortem ChP tissue studies were determined to be non-human subjects research by the UCI IRB. Research was conducted only after sufficient laboratory and safety training had been provided to all researchers.

Pluripotent stem cell culture

H1 embryonic stem cells (ESCs; WiCell Research Institute, Madison, WI) and iPSCs (APOE3/3 and APOE4/4 isogenic pairs of UCI ADRC Line 6, Mount Sinai TCW line 1, Mount Sinai TCW line 2, and Coriell WTC-GFP) were cultured on growth factor-reduced Matrigel (Corning #354230) diluted in DMEM (Thermo Fisher #11995-065) and maintained in E8 media (Thermo Fisher #A1517001) at 37 °C in 5% CO₂ with daily full media changes. After reaching 80% confluence, PSCs were dissociated using ReLeSR following manufacturer protocol (StemCell Technologies #05872), passaged using E8 with 10uM Y-27632 (ROCK inhibitor; StemRD #Y-025). Y-27632 was removed after 12–24 h. For quality controls, PSCs passed array CGH testing (Cell Line Genetics, Madison, WI) and pluripotency testing using a four-antibody kit (Thermo Fisher #A24881). Cell culture media were sterile filtered (Fisher Scientific #SCGP00525 and #0974063D), but no antibiotics or antimycotics were used per best practice guidelines¹⁰⁴. Contamination was monitored by phase-contrast microscopy and mycoplasma detection kits (Thermo Fisher #M7006).

CPEC derivation

Glass chamber slides (Millipore #PEZGS0816) were coated with growth factor-reduced Matrigel (0.085 mg/mL per chamber slide well). PSCs were dissociated, as above, and with trituration using a p1000 pipette tip to achieve clumps averaging 50 um in diameter. Cell solution was diluted to achieve 10 clumps/mm² density before being distributed evenly in the wells and incubated at 37 °C in 5% CO₂. ROCK inhibitor was removed after 24 h. Once PSC colonies reached ~150 um in diameter (usually 2–3 days post-seeding), PSC media was fully replaced with Neural Induction Media (NIM; Thermo Fisher #A1647801), which marked the start of derivation (“N0”). Twenty-four hours later (“N1”) media was replaced fully with NIM with 10 ng/mL BMP4 (R&D Systems #314-BP-050). On day N5, media was replaced fully with CPEC media composed of DMEM/F12 (Thermo Fisher #11330-032), NEAA (Thermo Fisher #07980), N2 (Thermo Fisher #17502048), and Heparin (StemCell Technologies #07980), supplemented with 10 ng/mL BMP4; day N5 became “C5”. After 15 days of BMP4 exposure (day C16), cultures were given daily half-media changes with CPEC media without BMP4 for the remaining time in culture. Derivations with H1, UCI ADRC line 6 isogenic pair, and Coriell WTC-GFP isogenic pair were characterized with immunostaining and scRNA-seq while the other lines were confirmed for CPEC derivation with immunostaining only.

Optimization of derivation variables

Variables involved in CPEC differentiation involved seeding density, seeding clump size, Matrigel thickness, glass or plastic surface, BMP4 addition, BMP4 concentration, BMP4 duration, neural induction start, neural induction duration, neural induction method, maturation additives, and length of culture. Seeding densities tested ranged from 2–20 clumps/mm². Seeding clump sizes tested ranged from 5–100 um/clump. Matrigel thicknesses tested ranged from 0.04–0.68 mg/mL. Plastic, glass coverslip, and glass slides were tested for culturing. BMP4 additions tested ranged from 1 day prior to the start of neural induction to 3 days post start of neural induction. BMP4 concentrations tested ranged from 5–20 ng/mL. BMP4 durations tested ranged from 5–35 days. Neural induction initiations tested ranged from 1 day post seeding to 5 days post seeding. Neural induction durations tested ranged from 3–15 days. Neural induction method was compared between Thermo Fisher’s product and StemCell Technologies’ product. Maturation additives tested were with or without N2 supplementation, with or without cyclopamine, and with or without CHIR 99021. Length of cultures tested ranged from 10 to 100 days. These factors were tested one at a time initially and then in various combinations. In addition to phase-contrast morphology and TTR ICC, CPEC differentiation efficiency was secondarily based on AQP1, CLDN1, AE2, and ARL13B ICC. For alternative neural induction (Supplementary Fig. 1C), SMADi Neural Induction kit (StemCell Technologies #08581) was applied 48 h after seeding fully dissociated H1 ESCs for 5 days. ESCs were dissociated using ReLeSR to obtain a single cell state, suspended in E8 + RI (10uM), diluted to 500,000 cells/mL, and then 1 mL of the solution was added to a single well of a 4-well chamber slide. Rock inhibitor was removed after 24 h (P1). Fourty-eight hours after initial seeding (P2 = N0), the cells were given SMADi neural induction media (StemCell Technologies #08581) for 5 days. 10 ng/mL BMP4 was added 48 h after kit application followed by daily half-media changes with CPEC media. After the 5 days of neural induction (N5 = C5), the media was fully replaced with CPEC + BMP4 media and after a total of 15 days exposure of BMP4 (C17), cultures were given daily half media changes with CPEC media without BMP4 for the remaining time in culture and yielded >90% CPECs.

EdU (5-ethynyl-2’-deoxyuridine) incorporation

The Click-iT EdU Alexa Fluor 555 kit (Thermo Fisher #C10338) was used per manufacturer protocol. Cultures were washed 2x with CPEC media, then given a full media change with 40 uM EdU in CPEC media, followed by daily half-media changes with EdU-CPEC media. At termination, cultures were aspirated, washed 1x with DPBS, then fixed with 4% PFA (Fisher Scientific #50-980-487) for 15 min at room temperature (RT). Click-iT reactions were performed following manufacturer protocol prior to immunostaining, Hoechst counterstaining, and coverslipping.

MitoTracker Orange (MTO) labeling

MTO dye (Thermo Fisher #M7510) was reconstituted in DMSO (Thermo Fisher #D12345) per manufacturer protocol and used with minimal light exposure. Biological triplicate wells were gently washed 2x with CPEC media, then given full media changes with 150 nM MTO in CPEC media or vehicle in CPEC media. After 20 min at 37 °C in 5% CO₂, media was aspirated, cultures washed once with DPBS, then fixed with 4% PFA prior to immunostaining.

Aβ uptake

FAM-labeled Aβ_1-42 peptide (AnaSpec #AS-23526-01) was reconstituted following manufacturer protocol and used with minimal light exposure. Stocks were diluted to 500 nM-4 uM working solutions in CPEC media. Prior to treatment, cultures were gently washed 1x with DPBS, then CPEC-Aβ_1-42 media was added to triplicate wells and incubated at 37 °C in 5% CO₂ for 30 min-48 h. To address slide-to-slide variability, conditions were randomized across wells of different chamber slides. All conditions were stopped at the same time by aspiration, gentle washing 2x with DPBS, and fixation with 4% PFA for 20 min at RT.

TTR ELISA with brefeldin A (BFA) inhibition

After 3x washes with warm CPEC media, 1 ml of CPEC media was reapplied to each well of a 4-well chamber slide, then 20 ul of conditioned CPEC media was longitudinally collected from individual wells over time (0–10 hrs starting with 1 ml, ending with ~820 ul in each well). To inhibit secretion, BFA (Invitrogen #B7450) or vehicle (DMSO) was added directly to wells. To test later for BFA reversibility, media was aspirated, then cells were washed 3x before replacing with CPEC media with or without BFA. Longitudinally collected media were assayed using a human TTR ELISA kit (AssayPro #EP3010-1) per manufacturer protocol and read on an HT microplate reader (BioTek Synergy). Background media-only values were subtracted, and concentrations were inferred from values normalized to the 2 h timepoint using standard curves measured in parallel.

Immunostaining

Antibody information is provided in Table 1.

Table 1 Information on antibodies used in this study

Full size table

ICC of adherent cultures

Cultures were gently washed 2x with DPBS at RT, fixed with 4% PFA at RT for 20 min, then washed 3x with DPBS. Cultures were incubated in blocking solution (5% donkey serum (Jackson ImmunoResearch #017-000-121) in PBS + 0.03% Triton X-100 (MilliporeSigma #9036-19-5)) at RT for 1 h, followed by incubation in primary antibody solution (primary antibody diluted in 1% donkey serum in PBS + 0.03% Triton) at 4 C overnight. The next day, cultures were washed 3x followed by incubation in secondary antibody solution (secondary antibody diluted in 1% donkey serum in PBS + 0.03% Triton) at RT for 1 h in the dark, washed 3x, then incubated in Hoechst solution (2ug/ml in PBS) (Thermo Fisher #H3570) for 5 min at RT in the dark, then washed 3x. The chamber slide apparatus was then removed per manufacturer protocol and slides were mounted in Fluoromount-G (Southern Biotech #0100-01) and coverslipped (Thermo Fisher #102460).

IHC of sectioned ChP

Paraffin-embedded human ChP tissue, fixed in 10% formalin at RT, were sectioned at 5um thickness by histology service (Experimental Tissue Resource). Sections were heated at 65 ºC for 30 min, deparaffinized in xylene (Sigma-Aldrich #214736), then rehydrated through a descending alcohol series. Antigen retrieval using sodium citrate pH 6 (MilliporeSigma #6132-04-3) for 20 min in a vegetable steamer at 100 ºC was followed by 1x ddH20 wash for 10 min at RT. Hydrophobic Pap-Pen barriers (Thermo Fisher #R3777) were drawn before fluorescent immunostaining and coverslipping, as described above.

IHC of ChP whole mounts

Postmortem human ChP tissue from the lateral ventricle was obtained at autopsy, fixed and stored in 10% formalin at RT or 4 C. Under stereo dissecting microscope, multiple small pieces (<1 mm diameter) were randomly picked and placed in a mesh basket (Corning #CLS3478) in a 12-well plate. Free-floating immunostaining was performed as described above for sections using a rocker on low settings. Stained tissue pieces were placed on glass slides containing spacers (Invitrogen #S24735), then coverslipped.

Image acquisition and modification

Phase contrast images of live cultures were acquired with an EVOS XL Core microscope (Advanced Microscopy Group). Fluorescent images were acquired with Olympus FV3000 confocal microscope, Nikon Eclipse E400 epifluorescent microscope with Nuance spectral deconvolution software, or Keyence BZ-X810 epifluorescent microscope for stitched images. All images were processed in ImageJ (v2.0.0). For all images compared qualitatively or quantitatively, identical image acquisition settings were used, and image adjustments in ImageJ were done in parallel and restricted to Brightness/Contrast.

Transmission electron microscopy

Human dCPEC cultures were fixed with a 2% glutaraldehyde and 4% paraformaldehyde solution for 20 min at room temperature followed by cutting out the CPEC islands with a scalpel blade. Samples were kept in fixative for 5 days at 4 C followed by transfer to the UCI Experimental Pathology Core for processing and imaging.

Image analysis and quantification

MitoTracker Orange

Confocal Z-stacks of cultures were converted to maximal projections. Membranous ZO1 signal was used to define TTR+ (CPEC) and TTR-negative (non-CPEC) ROIs, and total integrated MitoTracker Orange (MTO) signal was measured. Negative controls (no primary antibody) were imaged, then compartment-specific background signals were averaged and subtracted from experimental values. ROIs were drawn in blinded fashion prior to measuring integrated MTO values.

Cilia, Aβ uptake, and early endosomes

Confocal Z-stacks of cultures were converted into maximal projections. For cilia mass measurements, membranous ZO1 was used to define cell ROIs, then total integrated ARL13B (cilia) and/or fluorescent Aβ_1-42 signals were measured. To assess cilia dispersion, outer contours of ARL13B signal were drawn and used as cilia ROIs. Cilia coverage was calculated by dividing cilia ROI area by cell ROI area. Negative controls were used for compartment-specific background subtraction, as above. For EEA1-Aβ colocalization measurements, the ImageJ co-loc2 tool was used. Due to inadequate signal using single cell ROIs, ROIs containing multiple TTR+ CPECs (5-10 cells) were drawn in blinded fashion before EEA1-Aβ colocalization measurements were taken.

CPEC subtype markers

Confocal Z-stacks of cultures and tissues were converted to maximal projections. For C2 markers (FOXN4 and MCIDAS), Hoechst channels were used to outline nuclear ROIs in ImageJ, which were then applied to the C2 marker channels to measure total fluorescent signal. For C1a and C1b markers, membranous CLDN5 signal was also used to outline cell ROIs. As above, ROIs were drawn in blinded fashion, and negative controls were used for compartment-specific background subtraction. Sample numbers are reported in Figure Legends; for the expanded in vitro and in vivo series in Supplementary Fig. 8C, D, all imaged and identifiable nuclei were scored.

Videomicroscopy of motile cilia

Live dCPEC cultures were imaged in an Olympus FV3000 confocal microscope equipped with CO2 and temperature regulation. Plane of focus was set to the apical surfaces of dCPEC islands. DIC movies were acquired using resonance scanning at 200 frames/second for 5 min intervals. Cilia tracking was performed in ImageJ with the Manual Tracking plugin. Using default settings, the tip coordinates of an in-focus cilium (one per dCPEC) were manually recorded every 100 frames, then plotted in two dimensions (X-Y graph).

scRNA-seq data acquisition and quality controls

Multiple dissociation methods were tested per 10x Chromium guidelines. For each timepoint, two wells of a 4-well chamber slide were dissociated and pooled. The selected protocol involved dissociating with warm TrypLE (Thermo Fisher #12605010) for 25 min at 37 ºC, resuspending in FACS buffer at RT (0.02% BSA in DPBS), then propidium iodine was added for 10 min (Thermo Fisher #P1304MP) before placing on ice. FACS for single live cells was performed by the UCI Flow Cytometry Core using a FACSAria II (BD Biosciences). Single cell capture with the 10x Chromium platform, cDNA library preparation, and cDNA QC analysis (Qubit dsDNA HS Assay kit (Life Technologies Q32851) and high sensitivity DNA chips (Agilent 5067-4626)) were performed by the UCI Genomics Research and Technology Hub (GRT Hub). Library generation was done with a V2 chemistry kit and an average capture of ~10,000 barcoded cells, per user guide (CG00052 Rev B). The cDNA libraries were sequenced using the HiSeq 4000 platform (Illumina) to achieve an average depth of 50,000 reads per cell and 2,500 genes per cell. FASTQ files were processed and mapped to the reference genome GRCh38 through Cell Ranger. scRNA-seq datasets were then bioinformatically filtered to remove low quality cells and reads as previously described^14,34,35,36. Briefly, genes expressed in <10 cells were excluded, and cells with 1000–6000 genes/cell and ≤12.5% mitochondrial gene expression were included. The four time points were tested for their batch effect through two batch correction methods, RPCA and SCTransform with the Seurat platform. Datasets post quality control were combined with the merge function in Seurat while the time point of origin was retained in the metadata. RPCA batch correction was done with default settings of features to consider (2000) and plotted as a UMAP with color coding by time point of origin showing similar display of cells to the non-corrected. SCTransform batch correction was done with default settings of features to consider (all) and plotted as a UMAP with color coding by time point of origin showing similar display of cells to the non-corrected. Due to the similar grouping of cells in batch and non-batch corrected data, non-batch corrected data were used for further processing with the SoptSC platform.

scRNA-seq data analysis

SoptSC clustering and analysis

These were performed as described³⁴. We applied SoptSC (implemented in MATLAB R2019b) for clustering, marker gene identification, and lineage inference for each timepoint in the study. We selected 2000 highly variable genes (HVGs) as input to SoptSC for clustering and lineage estimation at each timepoint. Based on the number of zero eigenvalues and the largest eigengap of the clustering consensus matrix Laplacian, SoptSC provided a range of cluster-number estimates for each time-point. Cluster number per time point was done by finding the largest eigengap, picking the lower boundary of the gap, and generating a DEG heatmap to confirm the uniqueness of each cluster. For all time points the lower boundary was chosen, except for the 55 div set where the upper boundary of the eigengap showed cleaner clustering based on DEG heatmaps. We further determined the cluster number by verifying the expression of marker genes expressed by the clustered cells. For each specific time point, SoptSC infers lineage from cell-cell graph and abstract cluster-cluster graph using minimum spanning tree and was further validated by the biomarkers of the progenitor cell types.

Seurat clustering

Seurat³⁵ v4.1 analysis was performed per online Seurat tutorial. Each timepoint was processed independently. Highly variable features with standard deviations ≥2.5 were used for PCA generation. Resolution ranged from 0.1–0.3, and DEG heatmaps were used to validate clustering and identity similar to SoptSC.

STITCH analysis

For STITCH³⁹, highly variable genes were computed using Fano factors and ranking all genes by an above-Poisson noise statistic. Using default settings, top-2000 variable genes were filtered to include only genes whose single-cell transcript counts were minimally correlated (correlation coefficient >0.3) to at least one other variable gene. Gephi software was used for visualization.

Neural G0 analysis

ScRNA-seq datasets were merged, then Seurat was used initially for cell cycle assignment. Using the same Neural G0 classifier⁴⁵, either the mouse S and G2M gene lists from Seurat or the human G0, S, and G2M gene lists from Neural G0 were used. For Fig. 5D, cell clusters were assigned their predominant cell cycle phase unless a second phase exceeded 30%, in which case a “dual” phase was designated (see legend for Supplementary Fig. 7B).

CellChat analysis

For CellChat⁶⁵, preliminary imputation using Drimpute was performed to account for dropouts, and the imputation output was used as input. The CellChat vignette was followed to create various communication and summary plots (https://github.com/jinworks/CellChat).

CPEC comparisons across datasets

Raw reads and metadata of published datasets^15,36 were obtained from the UCSC Cell Browser. Datasets were processed through Seurat, as done by the originating groups, using the same quality control thresholds used for dCPECs. Cell types of interest (CPECs, neurons, NECs) were merged across datasets, followed by batch correction with either RPCA³⁷ or SCTransform³⁸ before clustering, Pearson correlations, and PCAs were performed in Seurat.

Gene enrichment and pathway analyses

Differential (top 300) and highly-expressed genes (top 10%) were used as input for pathway enrichment studies, then TopGO, KEGGprofile, and ReactomePA packages returned terms that reached statistical significance (p < 0.05 with Bonferroni correction). GSEA enrichment analysis compared all genes expressed in ≥25% of cells, also returning terms with p < 0.05 with Bonferroni correction. Heatmaps of manually-curated gene lists for specific CPEC subtype functions (Supplementary Fig. 9A, B) were generated after mining the GSEA human molecular signature database (https://www.gsea-msigdb.org/gsea/msigdb).

Correlation and aggregate analysis

C1a-C1b gene coexpression and divergence were explored by correlating the expression of every pairwise combination (15 in total) of the top 3 C1a and C1b DEGs across type 1 CPECs only at every timepoint with the FeatureScatter function to yield Pearson correlation values. Additionally, evidence for early emergence of late subtypes (C1a and C1b) and persistence of early subtypes (C1 and C2) was explored by aggregating the CPEC and NEC subtypes across the four timepoints and subjecting them to clustering, which generated 6 clusters supported by DEG heatmap. UMAP graphs of the aggregate dataset were color coded and/or split by timepoint or cluster of origin. See Supplementary Information for additional details.

Statistics and reproducibility

Specific statistical tests for each experiment are mentioned in Figure Legends and Supplementary Fig. Legends. All t-tests were unpaired and two-tailed, while all ANOVAs were one-way. Both Pearson and Spearman correlation tests were two-tailed. Pearson correlations were used to compare transcriptome datasets and fluorescent levels. For rank-order analyses, Spearman correlations were used. Quantification of fluorescent signal in postmortem human tissue used multiple tissue sections across multiple glass slides. When possible, samples from different patients were used. Qualitative assessment of fluorescent signal also involved multiple tissue sections across slides per patient and images were representative across sections. Quantitative studies of dCPECs, apart from the TTR ELISA and MTO treatment, were performed with multiple chamber slides in at least three independent CPEC derivations. MTO studies were performed in two independent derivations. TTR ELISA was performed once but involved longitudinal sampling across multiple slides and wells. Movies of cilia motility were performed on cells derived from the same initial plating and are representative of movies imaged from multiple wells and chamber slides. For studies tracking changes across the age of dCPECs, timepoints were from the same initial plating. Immunostaining and imaging were performed once all replicates were generated. Quantification involved blinding using three different methods. Multiple users were employed who were blinded to the ages, conditions, and marker identity of the fluorescent and DIC images. While measurements were taken, users were blinded to other fluorescent channels as well.

Graphing and AI attestation

GraphPad Prism8 was used for graphing and statistical analysis of the ELISA, fluorescent measurements, and transcriptome correlations. Custom illustrations (Fig. 7L) were created in BioRender. Masters, H. (2025) https://BioRender.com/cpj4ynj. All studies were conducted without the use of deep learning algorithms or platforms. Additionally, no form of AI was used for data analysis or writing this report.

Inclusion and ethics

This study involved various local and regional researchers across multiple career stages. The collaborators in this study qualify for authorship as outlined by Nature Communications and are listed as authors. Their involvement and contribution were discussed prior to the start of the research. This work is locally relevant as determined by all collaborating researchers. The research was neither restricted by the setting of the researchers nor resulted in stigmatization, incrimination, discrimination, or personal risk to participants. Local and regional research relevant to our study was addressed properly in the citations and the report.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The single cell sequencing data generated in this study have been deposited in the NCBI GEO database under accession code GSE296691. All codes used in this study were publicly available through Bioconductor [https://www.bioconductor.org/] or through the institutions that published the platforms Seurat [https://satijalab.org/seurat/articles/pbmc3k_tutorial]³⁵, STITCH³⁹ [https://github.com/wagnerde/STITCH], and SoptSC³⁴ [https://github.com/WangShuxiong/SoptSC?tab=readme-ov-file]. Source data are provided with this paper.

References

Lun, M. P., Monuki, E. S. & Lehtinen, K. Development and functions of the choroid plexus–cerebrospinal fluid system. Nat. Rev. Neurosci. 16, 445–457 (2015).
Article CAS PubMed PubMed Central Google Scholar
Faraci, F. M., Mayhan, W. G. & Heistad, D. D. Effect of serotonin on blood flow to the choroid plexus. Brain Res. 478, 121–126 (1989).
Article CAS PubMed Google Scholar
Faraci, F. M., Baumbach, G. L. & Heistad, D. D. Cerebral circulation: humoral regulation and effects of chronic hypertension. J. Am. Soc. Nephrol. 1, 53–57 (1990).
Article CAS PubMed Google Scholar
Hablitz, L. M. & Nedergaard, M. The glymphatic system: a novel component of fundamental neurobiology. J. Neurosci. 41, 7698–7711 (2021).
Article CAS PubMed PubMed Central Google Scholar
Simon, M. J. & Iliff, J. J. Regulation of cerebrospinal fluid flow in neurodegenerative, neurovascular and neuroinflammatory disease. Biochimica. Biophysica. Acta Mol. Basis Dis. 1862, 442–451 (2016).
Article CAS Google Scholar
Kouzehgarani, G. N. et al. Harnessing cerebrospinal fluid circulation for drug delivery to brain tissues. Adv. Drug Deliv. Rev. 173, 20–59 (2021).
Article Google Scholar
Pardridge, W. M. Blood-brain barrier and delivery of protein and gene therapeutics to brain. Front. Aging Neurosci. 11, 373 (2020).
Article PubMed PubMed Central Google Scholar
Dohrmann, G. J. Dark and light epithelial cells in the choroid plexus of mammals. J. Ultrastruct. Res. 32, 268−73 (1970).
Johanson, C. E. et al. AVP V ₁ Receptor-mediated decrease in Cl ⁻ efflux and increase in dark cell number in choroid plexus epithelium. Am. J. Physiol. Cell Physiol. 276, C82–C90 (1999).
Article ADS CAS Google Scholar
Liddelow, S. A. Development of the choroid plexus and blood-CSF barrier. Front. Neurosci. 9, 32 (2015).
Currle, D. S., Cheng, X., Hsu, C. M. & Monuki, E. S. Direct and indirect roles of CNS dorsal midline cells in choroid plexus epithelia formation. Development 132, 3549–3559 (2005).
Article CAS PubMed Google Scholar
Cheng, X. et al. Central roles of the roof plate in telencephalic development and holoprosencephaly. J. Neurosci. 26, 7640–7649 (2006).
Article CAS PubMed PubMed Central Google Scholar
Lun, M. P. et al. Spatially heterogeneous choroid plexus transcriptomes encode positional identity and contribute to regional csf production. J. Neurosci. 35, 4903–4916 (2015).
Article CAS PubMed PubMed Central Google Scholar
Dani, N. et al. A cellular and spatial map of the choroid plexus across brain ventricles and ages. Cell 184, 3056–3074 (2021).
Article CAS PubMed PubMed Central Google Scholar
Pellegrini, L. et al. Human CNS barrier-forming organoids with cerebrospinal fluid production. Science 369, 6500 (2020).
Article Google Scholar
Watanabe, M. et al. BMP4 sufficiency to induce choroid plexus epithelial fate from embryonic stem cell-derived neuroepithelial progenitors. J. Neurosci. 32, 15934 (2012).
Article CAS PubMed PubMed Central Google Scholar
Watanabe, M. et al. BMP4 acts as a dorsal telencephalic morphogen in a mouse embryonic stem cell culture system. Biol. Open 5, 1834–1843 (2016).
Article CAS PubMed PubMed Central Google Scholar
Srinivasan, S., et al. A BMP-FGF morphogen toggle switch drives the ultrasensitive expression of multiple genes in the developing forebrain. PLoS Comput. Biol. 10, e1003463 (2014).
Li, X. J. & Zhang, S. C. In vitro differentiation of neural precursors from human embryonic stem cells. Methods Mol. Biol. 331, 169–177 (2006).
PubMed Google Scholar
Javed, K., Reddy, V., and Lui, F. Neuroanatomy, Choroid Plexus. (StatPearls Publishing, 2023).
Johnson, B. A. et al. Accurate, Strong, and Stable Reporting of Choroid Plexus Epithelial Cells in Transgenic Mice Using a Human Transthyretin BAC. Fluid Barriers CNS. 15, 22 (2018).
Speake, T., Freeman, L. J. & Brown, P. D. Expression of aquaporin 1 and aquaporin 4 water channels in rat choroid plexus. Biochimica. Biophysica. Acta (BBA) Biomembranes 1609, 80–86 (2003).
Article CAS Google Scholar
Lindsey, A. E. et al. Functional expression and subcellular localization of an anion exchanger cloned from choroid plexus. Proc. Natl. Acad. Sci. 87, 5278–5282 (1990).
Article ADS CAS PubMed PubMed Central Google Scholar
Li, Q. et al. Disruption of GMNC-MCIDAS multiciliogenesis program is critical in choroid plexus carcinoma development. Cell Death Differ. 29, 1596–1610 (2022).
Article CAS PubMed PubMed Central Google Scholar
Johansson, P. A. et al. The transcription factor Otx2 regulates choroid plexus development and function. Development 140, 1055–1066 (2013).
Article CAS PubMed Google Scholar
Lippoldt, A. et al. Organization of choroid plexus epithelial and endothelial cell tight junctions and regulation of claudin-1, -2 and -5 expression by protein kinase C. Neuroreport 11, 1427–1431 (2000).
Article CAS PubMed Google Scholar
Barkho, B. Z. & Monuki, E. S. Proliferation of cultured mouse choroid plexus epithelial cells. PLoS ONE. 10, e0121738 (2015).
Cornford, E. M., Varesi, J. B., Hyman, S., Damian, R. T. & Raleigh, M. J. Mitochondrial content of choroid plexus epithelium. Exp. Brain Res. 116, 399–405 (1997).
Article CAS PubMed Google Scholar
Thouvenot, E. et al. The proteomic analysis of mouse choroid plexus secretome reveals a high protein secretion capacity of choroidal epithelial cells. Proteomics 6, 5941–5952 (2006).
Article CAS PubMed Google Scholar
Misumi, Y. et al. Novel blockade by brefeldin A of intracellular transport of secretory proteins in cultured rat hepatocytes. J. Biol. Chem. 261, 11398–11403 (1986).
Article CAS PubMed Google Scholar
Narita, K. and Takeda, S. Cilia in the choroid plexus: their roles in hydrocephalus and beyond. Front. Cell. Neurosci. https://doi.org/10.3389/fncel.2015.00039 (2015).
He, Q. et al. Primary cilia in stem cells and neural progenitors are regulated by neutral sphingomyelinase 2 and ceramide. Mol. Biol. Cell 25, 1715–1729 (2014).
Article PubMed PubMed Central Google Scholar
Crossgrove, J. S., Li, G. J. & Zheng, W. The choroid plexus removes β-amyloid from brain cerebrospinal fluid. Exp. Biol. Med. 230, 771–776 (2005).
Article CAS Google Scholar
Wang, S., Karikomi, M., MacLean, A. L. & Nie, Q. Cell lineage and communication network inference via optimization for single-cell transcriptomics. Nucleic Acids Res. 47, e66–e66 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e29 (2021).
Article CAS PubMed PubMed Central Google Scholar
Nowakowski, T. J. et al. Spatiotemporal gene expression trajectories reveal developmental hierarchies of the human cortex. Science 358, 1318–1323 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902.e21 (2019).
Article CAS PubMed PubMed Central Google Scholar
Durmaz, A. & Scott, J. G. Stability of SCRNA-Seq analysis workflows is susceptible to preprocessing and is mitigated by regularized or supervised approaches. Evolut. Bioinforma. online 18, 11769343221123050 (2022).
Article Google Scholar
Wagner, D. E. et al. Single-cell mapping of gene expression landscapes and lineage in the zebrafish embryo. Science 360, 981–987 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Bailey, P. Morphology of the roof plate of the fore-brain and the lateral choroid plexuses in the human embryo. J. Comp. Neurol. 26, 79–120 (1916).
Article Google Scholar
Ozair, M. Z., Kintner, C. & Brivanlou, A. H. Neural induction and early patterning in vertebrates. Wiley Interdiscip. Rev. Dev. Biol. 2, 479–498 (2013).
Article CAS PubMed Google Scholar
Huang, X. et al. Sonic hedgehog signaling regulates a novel epithelial progenitor domain of the hindbrain choroid plexus. Development 136, 2535–2543 (2009).
Article CAS PubMed PubMed Central Google Scholar
Nestorowa, S. et al. A single-cell resolution map of mouse hematopoietic stem and progenitor cell differentiation. Blood 128, e20–e31 (2016).
Article CAS PubMed PubMed Central Google Scholar
Plotnikova, O. V., Pugacheva, E. N. & Golemis, E. A. Primary cilia and the cell cycle. Methods Cell Biol. 94, 137–160 (2009).
Article CAS PubMed PubMed Central Google Scholar
O’Connor, S. A. et al. Neural G0: A Quiescent-like State Found in Neuroepithelial-derived Cells and Glioma. Mol. Syst. Biol. 17, e9522 (2021).
Scholzen, T. & Gerdes, J. The Ki-67 protein: from the known and the unknown. J. Cell Physiol. 182, 311–322 (2000).
Article CAS PubMed Google Scholar
Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27–30 (2000).
Article CAS PubMed PubMed Central Google Scholar
Ashburner, M. et al. Gene ontology: tool for the unification of biology. Nat. Genet. 25, 25–29 (2000).
Article CAS PubMed PubMed Central Google Scholar
D’Gama, P. P. et al. Diversity and function of motile ciliated cell types within ependymal lineages of the zebrafish brain. Cell Rep. 37, 109775 (2021).
Article PubMed PubMed Central Google Scholar
Nonami, Y., Narita, K., Nakamura, H., Inoue, T. & Takeda, S. Developmental changes in ciliary motility on choroid plexus epithelial cells during the perinatal period. Cytoskeleton 70, 797–803 (2013).
Article CAS PubMed Google Scholar
Loreng, T. D. & Smith, E. F. The central apparatus of cilia and Eukaryotic Flagella. Cold Spring Harb. Perspect. Biol. 9, a028118 (2017).
Article PubMed PubMed Central Google Scholar
Choksi, S. P., Lauter, G., Swoboda, P. & Roy, S. Switching on cilia: transcriptional networks regulating ciliogenesis. Development 141, 1427–1441 (2014).
Article CAS PubMed Google Scholar
Campbell, E. P., Quigley, I. K. & Kintner, C. Foxn4 promotes gene expression required for the formation of multiple motile cilia. Development 143, 4654–4664 (2016).
Article CAS PubMed Google Scholar
Arbi, M., Pefani, D. E., Taraviras, S. & Lygerou, Z. Controlling centriole numbers: geminin family members as master regulators of centriole amplification and multiciliogenesis. Chromosoma 127, 151–174 (2018).
Article CAS PubMed Google Scholar
Wu, G. & Haw, R. Functional interaction network construction and analysis for disease discovery. Methods Mol. Biol. 1558, 235–253 (2017).
Article CAS PubMed Google Scholar
Reid, A. & Tursun, B. Transdifferentiation: do transition states lie on the path of development?. Curr. Opin. Syst. Biol. 11, 18–23 (2018).
Article PubMed PubMed Central Google Scholar
Bergen, A. A., Kaing, S. & ten Brink, J. B. Netherlands Brain Bank; Gorgels TG, Janssen SF. Gene expression and functional annotation of human choroid plexus epithelium failure in Alzheimer’s disease. BMC Genomics 16, 956 (2015).
Article PubMed PubMed Central Google Scholar
van Leeuwen, L. M. et al. A transgenic zebrafish model for the in vivo study of the blood and choroid plexus brain barriers using claudin 5. Biol. Open 7, bio030494 (2018).
Article PubMed PubMed Central Google Scholar
Jacob, F. et al. Human pluripotent stem cell-derived neural cells and brain organoids reveal SARS-CoV-2 neurotropism predominates in choroid plexus epithelium. Cell Stem Cell 27, 937–950.e9 (2020).
Article CAS PubMed PubMed Central Google Scholar
Sakaguchi, H. et al. Generation of functional hippocampal neurons from self-organizing human embryonic stem cell-derived dorsomedial telencephalic tissue. Nat. Commun. 6, 8896 (2015).
Article ADS CAS PubMed Google Scholar
Thomas, T. & Dziadek, M. Capacity to form choroid plexus-like cells in vitro is restricted to specific regions of the mouse neural ectoderm. Development 117, 253–262 (1993).
Article CAS PubMed Google Scholar
Hu, J. S. et al. Border formation in a Bmp gradient reduced to single dissociated cells. Proc. Natl. Acad. Sci. USA 105, 3398–3403 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Wilting, J. & Christ, B. An experimental and ultrastructural study on the development of the avian choroid plexus. Cell Tissue Res. 255, 487–494 (1989).
Article CAS PubMed Google Scholar
Bayer, S. A. & Altman, J. The human brain during the early first trimester. In Atlas of the Human Central Nervous System Development, Vol. 5 (eds Bayer, S. A. & Altman, J.) (CRC Press, Boca Raton, 2008).
Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12, 1088 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Liem, K. F., Tremml, G. Jr & Jessell, T. M. A role for the roof plate and its resident TGFbeta-related proteins in neuronal patterning in the dorsal spinal cord. Cell 91, 127–138 (1997).
Article CAS PubMed Google Scholar
Alexandre, P. & Wassef, M. The isthmic organizer links anteroposterior and dorsoventral patterning in the mid/hindbrain by generating roof plate structures. Development 130, 5331–5338 (2003).
Article CAS PubMed Google Scholar
Monuki, E. S. Induction and patterning in the telencephalon. In Neuroscience in the 21st Century (eds. Pfaff D. W., Volkow N. D., Rubenstein J.) 1–24 (Springer, 2022).
Nandakumar, S., Rozich, E. & Buttitta, L. Cell cycle re-entry in the nervous system: from polyploidy to neurodegeneration. Front Cell Dev. Biol. 24, 698661 (2021).
Article Google Scholar
Marshall, W. F. & Kintner, C. Cilia orientation and the fluid mechanics of development. Curr. Opin. Cell Biol. 20, 48–52 (2008).
Article CAS PubMed PubMed Central Google Scholar
Hampel, H. et al. The amyloid-β pathway in Alzheimer’s disease. Mol. Psychiatry 26, 5481–5503 (2021).
Article CAS PubMed PubMed Central Google Scholar
Hilgendorf, K. I., Johnson, C. T. & Jackson, P. K. The primary cilium as a cellular receiver: organizing ciliary GPCR signaling. Curr. Opin. Cell Biol. 39, 84–92 (2016).
Article CAS PubMed PubMed Central Google Scholar
Naslavsky, N. & Caplan, S. Endocytic membrane trafficking in the control of centrosome function. Curr. Opin. Cell Biol. 65, 150–155 (2020).
Article CAS PubMed PubMed Central Google Scholar
Arendt, T., Stieler, J. & Ueberham, U. Is sporadic Alzheimer’s disease a developmental disorder?. J. Neurochem 143, 396–408 (2017).
Article CAS PubMed Google Scholar
Gauvrit, T., Benderradji, H., Buée, L., Blum, D. & Vieau, D. Early-life environment influence on late-onset alzheimer’s disease. Front Cell Dev. Biol. 10, 834661 (2022).
Article PubMed PubMed Central Google Scholar
Amgalan, A., Andescavage, N. & Limperopoulos, C. Prenatal origins of neuropsychiatric diseases. Acta Paediatr. 110, 1741–1749 (2021).
Article PubMed Google Scholar
Rinaudo, P. F. & Lamb, J. Fetal origins of perinatal morbidity and/or adult disease. Semin Reprod. Med 26, 436–445 (2008).
Article PubMed Google Scholar
Elorriaga, V., Pierani, A. & Causeret, F. Cajal-retzius cells: Recent advances in identity and function. Curr. Opin. Neurobiol. 79, 102686 (2023).
Article CAS PubMed Google Scholar
Molnár, Z., Luhmann, H. J. & Kanold, P. O. Transient cortical circuits match spontaneous and sensory-driven activity during development. Science 370, eabb2153 (2020).
Article PubMed PubMed Central Google Scholar
Delgehyr, N. et al. Ependymal cell differentiation, from monociliated to multiciliated cells. Methods Cell Biol. 127, 19–35 (2015).
Article CAS PubMed Google Scholar
Dur, A. H. et al. In Xenopus ependymal cilia drive embryonic CSF circulation and brain development independently of cardiac pulsatile forces. Fluids Barriers CNS 17, 72 (2020).
Article CAS PubMed PubMed Central Google Scholar
Spassky, N. et al. Adult ependymal cells are postmitotic and are derived from radial glial cells during embryogenesis. J. Neurosci. Off. J. Soc. Neurosci. 25, 10–18 (2005).
Article CAS Google Scholar
Tissir, F. et al. Lack of cadherins Celsr2 and Celsr3 impairs ependymal ciliogenesis, leading to fatal hydrocephalus. Nat. Neurosci. 13, 700–707 (2010).
Article CAS PubMed Google Scholar
Coletti, A. M. et al. Characterization of the ventricular-subventricular stem cell niche during human brain development. Development 145, dev170100 (2018).
Article PubMed PubMed Central Google Scholar
Smart, I. H., Dehay, C., Giroud, P., Berland, M. & Kennedy, H. Unique morphological features of the proliferative zones and postmitotic compartments of the neural epithelium giving rise to striate and extrastriate cortex in the monkey. Cereb. Cortex 12, 37–53 (2002).
Article PubMed Google Scholar
Hillman, N. H., Kallapur, S. G. & Jobe, A. H. Physiology of transition from intrauterine to extrauterine life. Clin. Perinatol. 39, 769–783 (2012).
Article PubMed PubMed Central Google Scholar
Morton, S. U. & Brodsky, D. Fetal physiology and the transition to extrauterine Life. Clin. Perinatol. 43, 395–407 (2016).
Article PubMed PubMed Central Google Scholar
Bard, H. Postnatal fetal and adult hemoglobin synthesis in early preterm newborn infants. J. Clin. Invest 52, 1789–1795 (1973).
Article CAS PubMed PubMed Central Google Scholar
Jacob, A. et al. Differentiation of human pluripotent stem cells into functional lung alveolar epithelial cells. Cell Stem Cell 21, 472–488.e10 (2017).
Article CAS PubMed PubMed Central Google Scholar
Burton, G. J. & Fowden, A. L. The placenta: a multifaceted, transient organ. Philos. Trans. R. Soc. Lond. B Biol. Sci. 370, 20140066 (2015).
Article PubMed PubMed Central Google Scholar
Daniels, C. B. & Orgeig, S. The comparative biology of pulmonary surfactant: past, present and future. Comp. Biochem Physiol. A Mol. Integr. Physiol. 129, 9–36 (2001).
Article CAS PubMed Google Scholar
Rohlfing, K., Stuhlmann, F., Docker, M. F. & Burmester, T. Convergent evolution of hemoglobin switching in jawed and jawless vertebrates. BMC Evol. Biol. 16, 30 (2016). 1.
Article PubMed PubMed Central Google Scholar
Striedter, G. F. Principles of Brain Evolution. (Sunderland, 2005).
Striedter, G. F., Srinivasan, S. & Monuki, E. S. Cortical folding: when, where, how, and why?. Annu Rev. Neurosci. 38, 291–307 (2015).
Article CAS PubMed Google Scholar
Kappelmann, J. The evolution of body mass and relative brain size in fossil hominids. J. Hum. Evol. 30, 243–276 (1996).
Article Google Scholar
Wang, H., Kim, M., Normoyle, K. P. & Llano, D. Thermal regulation of the brain-an anatomical and physiological review for clinical neuroscientists. Front Neurosci. 9, 528 (2016).
Article PubMed PubMed Central Google Scholar
Steiner, P. Brain fuel utilization in the developing brain. Ann. Nutr. Metab. 75, 8–18 (2019).
Article CAS PubMed Google Scholar
Levy, S. B. & Bribiescas, R. G. Hierarchies in the energy budget: thyroid hormones and the evolution of human life history patterns. Evolut. Anthropol. 32, 275–292 (2023).
Article Google Scholar
Pontzer, H. et al. Metabolic acceleration and the evolution of human brain size and life history. Nature 533, 390–392 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Kim, B. Thyroid hormone as a determinant of energy expenditure and the basal metabolic rate. Thyroid 18, 141–144 (2008).
Article CAS PubMed Google Scholar
McLean, T. R., Rank, M. M., Smooker, P. M. & Richardson, S. J. Evolution of thyroid hormone distributor proteins. Mol. Cell. Endocrinol. 459, 43–52 (2017).
Article CAS PubMed Google Scholar
Schreiber, G. The evolutionary and integrative roles of transthyretin in thyroid hormone homeostasis. J. Endocrinol. 175, 61–73 (2002).
Article CAS PubMed Google Scholar
Richardson, S. J. Evolutionary changes to transthyretin: evolution of transthyretin biosynthesis. FEBS J. 276, 5342–5356 (2009).
Article CAS PubMed Google Scholar
Ludwig, T. E. et al. ISSCR standards for the use of human stem cells in basic research. Stem Cell Rep. 18, 1744–1752 (2023).
Article CAS Google Scholar

Download references

Acknowledgements

We thank the UCI Genomics Research and Technology Hub (GRT Hub) for assistance with scRNA-seq; the Sue and Bill Gross Stem Cell Research Center for access to FACS, core equipment, and stem cell culture support; Julia TCW and Mt. Sinai, the ADRC iPSC Core, and Coriell Institute for iPSCs; the UCI Experimental Tissue Resource and autopsy service of the Department of Pathology & Laboratory Medicine for access to postmortem human tissues, histology, and transmission electron microscopy services; and members of the Monuki lab for support and feedback on the work and manuscript. This work was supported by NIH R21 AG064640, NIH R21 MH109036, and UCI SoM-SoBS pilot award (E.S.M.); NIH R25 GM055246, NIH T32 NS082174, and Rose Hills Foundation (H.M.); NSF DMS1763272 and Simons Foundation 594598 (QN); and CIRM EDUC2-08383 (VE, OGJ).

Author information

Authors and Affiliations

Department of Pathology & Laboratory Medicine, University of California Irvine, Irvine, CA, USA
Haley Masters, Pamela Shi Ru Fung, Benjamin Tran, Cristina Martel, Nellie Kwang, Michael Neel, Olga G. Jaime, Victoria Espericueta, Brett A. Johnson & Edwin S. Monuki
Department of Developmental & Cell Biology, University of California Irvine, Irvine, CA, USA
Haley Masters, Yutong Sha, Qing Nie & Edwin S. Monuki
Department of Mathematics, University of California Irvine, Irvine, CA, USA
Shuxiong Wang, Yutong Sha, Matthew K. Karikomi & Qing Nie
Sue and Bill Gross Stem Cell Research Center, University of California Irvine, Irvine, CA, USA
Christina Tu & Edwin S. Monuki
Department of Biological Chemistry, University of California Irvine, Irvine, CA, USA
Quy Nguyen & Kai Kessenbrock

Authors

Haley Masters
View author publications
Search author on:PubMed Google Scholar
Shuxiong Wang
View author publications
Search author on:PubMed Google Scholar
Christina Tu
View author publications
Search author on:PubMed Google Scholar
Quy Nguyen
View author publications
Search author on:PubMed Google Scholar
Yutong Sha
View author publications
Search author on:PubMed Google Scholar
Matthew K. Karikomi
View author publications
Search author on:PubMed Google Scholar
Pamela Shi Ru Fung
View author publications
Search author on:PubMed Google Scholar
Benjamin Tran
View author publications
Search author on:PubMed Google Scholar
Cristina Martel
View author publications
Search author on:PubMed Google Scholar
Nellie Kwang
View author publications
Search author on:PubMed Google Scholar
Michael Neel
View author publications
Search author on:PubMed Google Scholar
Olga G. Jaime
View author publications
Search author on:PubMed Google Scholar
Victoria Espericueta
View author publications
Search author on:PubMed Google Scholar
Brett A. Johnson
View author publications
Search author on:PubMed Google Scholar
Kai Kessenbrock
View author publications
Search author on:PubMed Google Scholar
Qing Nie
View author publications
Search author on:PubMed Google Scholar
Edwin S. Monuki
View author publications
Search author on:PubMed Google Scholar

Contributions

H.M., B.A.J., and E.S.M. conceived the study. H.M., C.T., B.A.J., and E.S.M. contributed the theoretical basis for protocol optimizations, while H.M. and C.T. performed the optimizations. H.M. and Q.Nguyen acquired the scRNA-seq data. H.M., S.W., Y.S., M.K.K., N.K., K.K., Q.Nie, and E.S.M. contributed to scRNA-seq analysis. H.M., P.S.R.F., B.T., and C.M. contributed to fluorescent imaging and quantification. H.M., V.E., B.A.J., and E.S.M. contributed to statistical analysis. H.M., M.N., and O.G.J. contributed to cell culture treatments. H.M., B.A.J., and E.S.M. contributed to experimental planning and data analysis. H.M. performed the live videomicroscopy and analyzed the TEM photomicrographs, while H.M. and E.S.M. wrote and revised the manuscript. All authors approved the final manuscript.

Corresponding author

Correspondence to Edwin S. Monuki.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Hideya Sakaguchi, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Description of Additional Supplementary Information

Supplementary Movie 1

Supplementary Movie 2

Supplementary Movie 3

Supplementary Dataset 1

Reporting Summary

Transparent Peer Review file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Masters, H., Wang, S., Tu, C. et al. Sequential emergence and contraction of epithelial subtypes in the prenatal human choroid plexus revealed by a stem cell model. Nat Commun 16, 5149 (2025). https://doi.org/10.1038/s41467-025-60361-9

Download citation

Received: 22 June 2024
Accepted: 21 May 2025
Published: 03 June 2025
DOI: https://doi.org/10.1038/s41467-025-60361-9