Translational biomarkers of hypoxic brain injury uncovered in CSF secreting human choroid plexus organoids

Abstract

The choroid plexus-cerebrospinal fluid (ChP-CSF) interface regulates a microenvironment supporting neural stem cell growth, strongly affected by hypoxia through ChP function. From human induced pluripotent stem cells (hiPSCs), here we established and validated in vitro ChP organoid secreting CSF-like fluid (iCSF) and exposed them to low oxygen atmosphere for 24 h. Transcriptomic indicated major data on morphological and functional alterations in the ChP cells and shotgun proteomics revealed significant changes in proteins involved in energy metabolism and mitochondrial function. We found that H2AZ and ITM2B, involved in neurogenesis and neurite growth, were the key proteins downregulated in hypoxic iCSF and ChP organoids, respectively. Positive correlation analysis between hypoxia-induced mRNA expression of the neuronal progenitor biomarkers SOX2 and PAX6. Mature neuron MAP2 and H2AZ also confirmed impairment of neurogenesis. The results from this study suggest that ChP-CSF interface opens new opportunities to characterize hypoxic brain pathophysiology and discover novel biomarkers.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12987-025-00731-z.

Introduction

Hypoxic-Ischemic encephalopathy (HIE) is a result of insufficient blood flow to the brain combined with a lower-than-normal concentration of oxygen in arterial blood [1]. Regarding the incidence of HIE, One and a half per 1,000 infants experience HIE in developed countries and ranges from 2.3 to 26.5 per 1000 live births in developing countries at birth resulting in neurodevelopmental disabilities that place a lifelong burden on both parents and society [2]. This condition in the developing brain can lead to significant morbidity, mortality, and long-term neurological deficits including cerebral palsy, epilepsy, intellectual deficits and severe learning difficulties, motor, and behavioral disabilities.

Barriers between the blood and the central nervous system (CNS) encompass the endothelium of the cerebral microvessels, meningeal vessels, and the epithelium of choroid plexus (ChP). The ChP in the brain consists of specialized structures located primarily in the cerebral ventricles. It is primarily made up of a layer of ependymal cells, which are modified cells of the ependyma. These cells are arranged to form a selective barrier between the ventricles and the underlying tissues. The epithelium has tight junctions between the cells, tightly regulating the entry and exit of substances, thus helping maintain CSF homeostasis. The core of the epithelium is a dense network of capillaries surrounded by loose connective tissue. The main role of ChP is to produce cerebrospinal fluid (CSF), which serves to protect the brain while facilitating the removal of metabolic waste and the exchange of biomolecules [3].This tissue allows the exchange of nutrients, waste products, and other molecules between the blood and CSF [4]. The ependymal cells possess numerous cilia on their apical surface to promote the circulation of CSF throughout the ventricles. The epithelial cells express transporters, allowing precise control over the composition of CSF. These transporters facilitate the entry and exit of various ions, nutrients, and other molecules, and play a key role in regulating ionic balance and homeostasis [5]. The ChP-CSF interface is critical for the maintenance of the brain’s extracellular environment and the removal of metabolic waste needed for optimal brain function but also for the development and maintenance of CNS as ChP epithelial cells are signaling centers in early CNS development [4, 6, 7].

The CSF, a clear liquid secreted by ChP brain tissue, is home to hundreds of proteins and signaling activities, highlighting the crucial biological function for this complex fluid. Several growth factors are modulated in the brain and secreted in the CSF following injury [8, 9]. Hypoxic brain injury (HI) can compromise brain barriers including their integrity making the brain more vulnerable to injury [10]. Most studies have focused on the blood-brain barrier (BBB) while limited studies pointed out the potential alterations of the ChP-CSF interface. This interface between the choroid plexus and the cerebrospinal fluid (CSF) can be disrupted by ischemia [11, 12] or HI [13, 14]. For example, these injuries induce damage to ChP epithelial cells [15] and increase permeability to toxic compounds [16]. Infants experiencing HIE show higher levels of factors that promote this interface permeability in their CSF compared to controls [17].

Human cerebral organoids (COs) and choroid plexus (ChP) organoids are three-dimensional in vitro models that recapitulate key aspects of human brain development and architecture. The use of human induced pluripotent stem cells (hiPSCs) to generate ChP organoids enables the reproduction of the developmental and cellular complexity of the human ChP, including choroid plexus epithelial cells, and provides a more physiologically relevant platform to study the choroid plexus–CSF interface. However, hiPSC-derived ChP epithelial cells often exhibit an immature phenotype, which may influence barrier tightness, transporter expression, and responses to injury. In addition, organoids can undergo structural deterioration or show altered gene expression during prolonged culture, potentially affecting experimental reproducibility.

HI has been studied in in vitro models including COs and revealed severe neuronal damage, altered corticogenesis and decreased organoid size, loss of structure and changes in cellular behavior [18]. There is a significant gap in the knowledge regarding documentation of HI in the choroid plexus and CSF, thus the generation of human 3D ChP organoids appears as crucial to reveal pathways involved in and improve our understanding of HI pathophysiology. Thus, we established ChP organoids of 4 months from hiPSCs GMB7-1 cell line with secretion of CSF-like fluid in vitro (iCSF).

Proteomics was used to investigate molecular signatures of ChP-CSF organoids to validate the relevance of the model. After validation, they were exposed to HI for 24 h (1% O₂, 5% CO₂) followed by 24 h of normoxia. The response to HI was assessed in terms of impact on the HIF1α pathway, morphological and functional features of the ChP by transcriptomic analysis, as well as mitochondrial and metabolic functions by proteomics. In addition, shotgun proteomics associated with advanced bioinformatics data analysis reveals changes in the abundance of proteins that could be relevant biomarkers for HI. We provide insights into the dysregulation of the top significant proteins associated with the impairment of neurogenesis and epigenetic modification contributing to HI.

Results

Generation of human cerebral organoids (COs) and choroid plexus (ChP) organoids with a fluid-filled cavity

We generated organoids from the GMB7-1 cell line using the protocol of Lancaster & Knoblich [19] with minor modifications [20, 21] (Fig. 1A). The organoids generated from the GMB7-1 cell line spontaneously present a different structure as compared with other hiPSC cell lines (MAA3 and BXS0115).

Fig. 1.

Generation of human cerebral and ChP organoids with fluid-filled cavities. (A) Protocol timeline with images of ChP and COs over time. The black arrow indicates emerging ChP epithelium at day 9 and the arrowhead shows a later fluid-filled compartment. “d” indicates “day”, scale bar 150 μm. (B) Comparison of ChP organoid and COs (day 9 and day 120). The black arrow indicates ChP epithelium and the arrowhead fluid-filled compartment. Scale bar 150 μm. (C) Immunofluorescence of ChP (GMB7-1 cell line) and cerebral (BXS0115, MAA3) organoids. Identification of different cell types present in these organoids: mature neuron (MAP2 in red and NEUN in green), astrocytes (GFAP in green and GLAST in red), radial glia (PAX6 in green) and neural progenitor (SOX2 in red). Scale bar 150 μm, 10X objective for marker PAX6/MAP2; NEUN/SOX2 and 20X objective for GFAP/MAP2 and GLAST/NEUN. (D) Immunofluorescence of ChP (GMB7-1 cell line) organoids with tight junction markers CLDN5, ZO1 and MDR-1. Scale bar 50 μm, objective 40X and 63X. (E) Heatmap with different proteins expression between MAA3, BXS0115 and GMB7-1 cell line organoids explaining potential differentiation of GMB7-1 cell line into ChP organoid

Embryoid bodies (EBs) from MAA3, BXS0115 and GMB7-1 hiPSCs are quite similar in structure and size on days 2 and 4 (Fig. 1A). Structural differences appear at day 9 just before embedding in matrigel. Epithelium emerged on organoids from GMB7-1 hiPSC cell line (Fig. 1A and B), while MAA3 and BXS0115 organoids showed a bright circle around a dark center, indicative of early neuroepithelial patterning. GMB7-1 hiPSC organoids developed a fluid-filled cavity with a clear liquid inside along the 4 months of culture. This morphology suggests that differentiation of GMB7-1 leads to ChP organoids with a fluid-filled cavity secreting iCSF.

Characterization of the COs and ChP organoids reveals the presence of different cell types (Fig. 1C). ChP from cell line GMB7-1 contains mature neurons (NEUN, MAP2), astrocytes (GFAP, GLAST), radial glial cells (PAX6) and neuronal progenitors (SOX2). ChP organoids display tight junction proteins such as CLDN5 and ZO1 and transporters such as Multidrug Resistance Protein 1 (MDR-1) (Fig. 1D).

Modulation of Wnt signaling pathways triggers ChP organoids differentiation

To understand the spontaneous differentiation of the GMB7-1 cell line compared to the other cells MAA3 and BXS0115, proteomic analysis was performed for healthy (He) organoids generated from the three hiPSC cell lines. Particularly, we were interested to identify differentially abundant proteins (DAPS) between GMB7-1 and MAA3 and BXS0115 cell lines. We found over accumulation of several Wnt antagonists in GMB7-1 organoids compared to MAA3 and BXS0115 organoids, which could be one of the possible causes of spontaneous differentiation of cells and regulates cell fate decisions into ChP tissue. Bmp1 and Bmp7, antagonists of Wnt signaling, are more abundant expressed in GMB7-1 organoids compared to BXS0115 and MAA3 organoids. SMAD5, playing a central role in Bmp signaling, is down regulated in GMB7-1 organoids compared to MAA3 and BXS0115 cell lines. FGF2, which modulates Wnt signaling, is up-regulated in GMB7-1 organoids compared to the other cell lines. Finally, Dickkpof-1 protein (DKK1) is up-regulated in GMB7-1 organoids compared to MAA3 and BXS0115 cell lines (Fig. 1E).

Validation of the in vitro ChP-CSF interface model

In order to validate our human ChP in vitro model, we investigated our global proteome of ChP organoids in He condition (6794 proteins identified in total) and intersected them to the set of proteins with an elevated abundance in the human ChP compared to other regions of the brain which are retrieved from the proteome Atlas (371 proteins). One hundred proteins were found at the intersection between our human ChP in vitro model proteome and the human brain atlas ChP proteome data set (Fig. 2A). Among those common proteins some are considered as markers of the human ChP tissue such as transthyretin (TTR), Chloride Intracellular Channel 6 (CLIC6), folate receptor 1 (FOLR1), tight junction (claudin, zonula occluden), cell adhesion molecules (cadherin) and extracellular matrix (collagens). Enrichment analysis of the common proteins shows the involvement of matrisome, embryonic morphogenesis, connective tissue development, glial cell migration and growth factor stimulus (Fig. 2B).

Fig. 2.

Exploring similarities between human ChP proteome and ChP organoids secreting human CSF proteins. (A) Venn Diagram showing the intersection of proteins with an elevated expression in the human ChP compared to other regions of the brain (371 proteins retrieved from the proteome Atlas) and the in vitro ChP organoids data in He conditions (6794 proteins). (B) Enrichment analysis of the common proteins between in vitro ChP organoids data in He conditions and the regionally elevated protein expression in human ChP. (C) Venn Diagram showing the intersection of proteins identified in normal adult human CSF (Macron et al., 2020) (3174 proteins), and iCSF from ChP organoids (4451 proteins) in He conditions. (D) Enrichment analysis of common proteins between normal adult human CSF (Macron et al., 2020) and iCSF from ChP organoids in He conditions. (E) Venn Diagram showing the intersection of proteins in CSF of normal pediatric individuals (Guo et al., 2019) (342 proteins) and iCSF from ChP organoids (4451 proteins) in He conditions. (F) Enrichment analysis of common proteins between proteins identified in CSF of normal pediatric individuals and iCSF from ChP organoids in He conditions. (G) Venn Diagram showing the intersection of proteins identified in human embryonic CSF (Zappaterra et al., 2007) (185 proteins) and iCSF from ChP organoids in He conditions. (H) Enrichment analysis of common proteins between proteins identified in normal embryonic human CSF and iCSF from ChP organoids in He conditions. (I) Venn Diagram showing the intersection of proteins between iCSF from ChP organoids and human embryonic CSF, pediatric CSF and adult CSF

These ChP organoids from GMB7-1 cell line produce a clear liquid, a CSF-like fluid in vitro. Our proteomic data of the iCSF composition (4451 proteins identified) were compared with published datasets from human adult CSF [22] human pediatric CSF [23] and human embryonic CSF [24]. Proteomic comparative analysis revealed1539 common proteins between iCSF and in vivo human adult CSF representing 48% of similarities (Fig. 2C). Enrichment analysis of common proteins shows the involvement of these proteins in axon guidance molecules, cell morphogenesis, cell-cell adhesion and ECM organization (Fig. 2D).

Comparison with pediatric human CSF revealed 192 common proteins with iCSF showing 56% of similarities with 192 (Fig. 2E). Enrichment analysis of these 192 proteins suggests the involvement in ECM organization, axon development, gliogenesis and regulation of growth (Fig. 2F).

Finally, iCSF proteomic dataset has been compared to human embryonic CSF with 130 common proteins showing 70% of similarities (Fig. 2G). These proteins are involved in CNS development, ECM organization, blood vessel development, head development, axon regeneration and neuron projection development (Fig. 2.H).

All these common proteins between iCSF and the human in vivo CSF at different stages were crossed to see if there were any shared proteins among these proteins and different age of CSF in human (Fig. 2I). Only in vivo pediatric and adult CSF share common proteins (180 proteins).

Modeling HI on human ChP organoids

To model HI on human ChP organoids, the organoids were exposed to a low oxygen atmosphere (1% O₂, 5% CO₂) for 24 h followed by 24 h of normoxia. ChP organoids at 1% O₂ 24 h + 24 h normoxia (HI condition) and He condition were subjected to transcriptomics and proteomics (Fig. 3A, table EV1, table EV2). A total of 1673 differentially expressed genes (DEGs) were found by transcriptomics in HI ChP compared to He ChP (Table EV3).

Fig. 3.

Validation of HI protocol on human ChP organoids. (A) Global flow chart with generation of organoids, exposition to HI and proteomic transcriptomic analysis. (B) PCA Plot demonstrating linear class separation between HI and He states in GMB7-1 cell line. (C) Signaling pathways of HIF1α highlight the related differentially abundant proteins between HI and He condition in ChP samples. (D) Signaling pathways of HIF1α highlight the related differentially abundant proteins between HI and He condition in CSF samples

For evaluation of quality control of ChP samples in He and HI conditions, reproducibility plots, Principal Component Analysis (PCA) with proteomic analysis were generated using unsupervised hierarchical clustering and assessing the degree and quality of data separation between the different conditions and samples. The plot in Fig. 3B visualizes the separation of samples in the GMB7-1 cell line dataset across the first three principal components (PC1, PC2, and PC3) which collectively account for 56% of the total variance, as determined by linear PCA. The yellow color indicates He samples, and dark blue corresponds to HI, illustrating the clustering and distinction between these 2 conditions.

The differential abundance analysis of the normalized and filtered proteins identified 1341 significant proteins between HI ChP compared to He and 534 significant proteins between HI iCSF compared to He (Table EV.3).

Furthermore, the effective induction of HI was confirmed by exploring HIF1α pathways, a key transcriptional regulator of cellular and developmental response to HI [25]. Proteomic analysis between He ChP organoids and iCSF and HI reveals significant upregulation of enolase 1 (ENO1) and transferrin receptor (TFRC) (Fig. 3C and D). HI modulated the abundance of specific proteins associated with HIF1α pathway such as TFRC and ENO1 in ChP tissue and iCSF testifying the adaptation of cells in low oxygen condition.

Morphology-related proteins and transporters alterations in ChP organoids in HI

To identify the activated and significantly enriched pathways in HI ChP compared to He ChP, absolute GSEA (absGSEA) was performed using the DEGs and the Molecular Signature Database (MSigDB) including transcriptomics covering various pathways related to biological processes and molecular functions, cellular compartment, immune response and cell development. Sixty-seven pathways were significantly enriched from absGSEA in HI ChP compared to He ChP. The gene sets were then filtered to include gene sets that contain ≥ 20 genes and sorted according to their biological function (Table EV3). Significant DEGs with the top Log2 fold change in all GSEA are shown in Table EV4. Gene frequency was obtained by quantifying the occurrence of each gene across all analyzed pathways. These genes were derived from the significantly enriched gene sets identified in the analysis (Table EV5). The genes SLC28A2 (log2FC -4.74; frequency 9), SLC26A3 (log2FC -4.29; frequency 4), C17orf99 (log2FC -3.77; frequency 4), ANPEP (log2FC-3.23; frequency 5), POP1 (log2FC 2.10; frequency 4), ERCC6L (log2FC 2.14; frequency 6), SLC38A5 (log2FC 2.20; frequency 4), KIF4A (log2FC 2.22; frequency 5), and SFPQ (log2FC 2.55; frequency 4) in Tables EV5 and EV6 showed the most frequent and the most prominent fold change as upregulated or downregulated.

For the morphology-related proteins and transporter expression response to HI in ChP organoids, the annotation of the top leading-edge genes of the brush border of ChP gene set showed highly significant enrichment of three categories related to the organization of brush ChP, cells cytoskeleton and adhesion, and molecules transport systems (Fig. 4A). Absolute GSEA results revealed that a subset of the brush border of ChP genes was over-represented (p < 0. 001) in He ChP compared to HI ChP (Fig. 4B, C). In HI, chloride anions (SLC26A3), transport of neurotransmitters and nucleoside (SLC28A2), and amino acid (SLC38A5) (Fig. 4C) are dysregulated leading to impaired nutrient uptake, altered ion homeostasis, accumulation of toxic metabolites and disrupted redox balance. Proteomics corroborates this morphology-related proteins impairment with dysregulation of proteins involved in adherens junction (nectin and LAR proteins), actin cytoskeleton (gelsolin, myosin II and ERM proteins) and tight junction pathways (ROCK and MUPP1 proteins) (Fig. 4D, E, F).

Fig. 4.

Impairment of ChP morphology-related proteins and transporters alterations in the context of HI. (A)Volcano plots of DEGs. Genes that are significantly expressed in either HI or He ChP based on log2 fold change are shown. p < 0.05 are highlighted in red, p > 0.05 are highlighted in green, unchanged transcripts are highlighted in grey. (B, C) Significant enrichment pathways based on the frequency of significant genes in the brush border of the ChP gene set. (B) 108 genes of the brush border of ChP gene set were used in Metascape. The red arrows show the significant pathways involved in organization of brush ChP; the blue arrows show cells’ cytoskeleton and adhesion; and the black arrows show the molecules’ transport systems. The y-axis shows significantly enriched gene ontology (GO) terms of the target genes. The darker the color, the higher the gene counts included in that pathway. The x-axis shows the enrichment scores of GO terms. P-value is calculated according to the count of the provided genes found in the given pathway. Log10(P) is the P-value in log base 10. P ≤ 0.05 was considered as significant. (C) Leading edge analysis showed that 20 core genes accounted for the significant enrichment in ChP organoids (p < 0.001). The 13 leading-edge core genes are shown; the frequently and/or top Log2 fold < change genes are indicated in red. (D) Signaling pathways of adherens junctions highlight the related differentially abundant proteins between HI and He conditions in ChP samples by proteomic analysis. (E) Signaling pathways of tight junction highlight the related differentially abundant proteins between HI and He conditions in ChP samples by proteomic analysis. (F) Signaling pathways of the actin cytoskeleton highlight the related differentially abundant proteins between HI and He conditions in ChP samples by proteomic analysis

Mitochondrial dysfunction in ChP cells in HI

Analysis of the proteomic data comparing HI and He conditions in the ChP organoids identified modulation of the abundance of proteins involved in mitochondrial gene expression, mitochondrial translation, ATP metabolic process and mitochondrial translational elongation. The purine nucleotide metabolic process was also involved in the context of HI (Fig. 5A). The top enrichment of gene ontologies (GO) of significant DAPs between HI and He conditions in the ChP organoids are shown Fig. 5A. To further refine our analysis, a Cnetplot was generated, illustrating all proteins associated with the top 10 enriched GO biological processes in HI (Fig. 5B and Fig. EV1A). Our findings showed that mitochondrial large subunit ribosomal proteins (MRPL) are associated with mitochondrial pathways changes in HI, death associated protein (DAP), elongation factor mitochondrial 1 (GFM1), ribosomal subunit protein. These pathways involved in HI pathophysiology, are closely connected related to each other as shown on the enrichment map visualizing the relationships among the 10 enriched GO biological processes in the context of HI (Fig. EV2A). Mitochondrial inner membrane, mitochondrial matrix, and protein complex were identified among the top 10 significant GO cellular components in the context of HI encompassing potential proteins such as mitochondrial ribosomal protein, OPA1, and SLC transporters (Fig. 5C, Fig. EV1B). For a more global visualization, there is an interaction plot of top 10 significant GO cellular components in the context of HI showing the relationships between the pathways (Fig. EV2B). Finally, for molecular function GO enrichment, cadherin binding, nucleic acid activity, and ATPase activity are involved among the top 10 significant GO molecular functions in the context of HI (Fig. 5D, Fig. EV1C). Relationships between the top 10 significant GO molecular functions are shown on an enrichment map plot (Fig. EV2C).

Fig. 5.

Outcomes of HI in in vitro human ChP organoids and CSF. (A) The GO enrichment of significant DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in ChP organoids. (B) Top 10 significant GO biological processes showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in ChP organoids. (C) Top 10 significant GO cellular components showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in ChP organoids. (D) Top 10 significant GO molecular functions showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in ChP organoids. (E) The GO enrichment of significant DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in iCSF samples. (F) Top 10 significant GO biological processes showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in iCSF samples. (G) Top 10 significant GO cellular components showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in iCSF samples. (H) Top 10 significant GO molecular functions showing the potential involved DAPs between HI 1% O₂ 24 h + 24 h normoxia condition and He condition in iCSF samples

We investigated 10 major significant proteins dysregulated between He and HI ChP organoids in order to study the role of ChP in the context of HI (Fig. 6A). The selection of key proteins was based on their fold change effect between the HI and He conditions, with a threshold highlighting the most relevant proteins. Among them, Integral membrane protein 2B (ITM2B) demonstrated the highest fold (1.57), Acyl-coenzyme A synthetase mitochondrial ACSM3 (1.25) and receptor tyrosine-protein kinase erbB-3 (1.24) (table EV7A), reflecting strong downregulation in HI conditions. The interaction network reveals ITM2B (Fig. 6C) as a key node linked to several proteins such as MAP4K5 or DGS1, suggesting its involvement in pathways critical for cellular maintenance and neurogenesis. RT-qPCR of ChP organoids correlated proteomic analysis with ITM2B downregulation in HI (p < 0.01) and impact on neurogenesis by dysregulation of radial glial cells PAX6 (p < 0.05), mature neurons MAP2 (p < 0.05) and decreasing tendency of neural progenitors (SOX2) (Fig. 6E).

Fig. 6.

ITM2B and H2AZ are involved in neurogenesis impairment in HI ChP organoids and iCSF. (A) Top 10 up and down differentially abundant proteins between HI and He conditions in ChP organoids. (B) Top 10 up and down differentially abundant proteins between HI and He conditions in iCSF samples. (C) String scheme link between relevant proteins including ITM2B. (D) String scheme link between relevant proteins including H2AZ. (E) Real Time PCR (RT-qPCR) of genes ITM2B, H2AZ, PAX6, SOX2 and MAP2 in ChP organoids, N = 2; n = 5–6; T-test parametric (normality and lognormality test, Gaussian distribution) and non-parametric (Mann-Whitney) analysis; circle shape represents batch number 1 of ChP and triangle shape represent batch number 2 of ChP; *p < 0.1, **p < 0.01. (F) Correlation analysis using RT-qPCR data of ChP organoids. Positive Spearman correlation for H2AZ / PAX6 with coefficient of 0.6626, positive Spearman correlation H2AZ/SOX2 with coefficient of 0.9097; positive Spearman correlation for H2AZ/MAP2 with coefficient of 0.8596 and positive Spearman correlation for SOX2/MAP2 with coefficient of 0.8902, positive Spearman correlation for MAP2/ PAX6 with coefficient of 0.7370, positive Spearman correlation for SOX2/ PAX6 with coefficient of0.7703. Green points represent organoids samples for He condition (N = 2; n = 5–6) and red points are organoid samples for HI condition (N = 2; n = 5–6); circle shape represent batch number 1 of ChP and triangle shape represent batch number 2 of ChP. To account for multiple comparisons, Bonferroni correction was applied to adjust the p-values with the threshold for statistical significance set a 0.05. (G) Co-localization of H2AZ (red) and SOX2 (green) protein in He ChP organoids (GMB7-1 hiPSC cell line). Scale bar 50 μm and 25 μm and 40X objective, confocal microscopy, FIJI analysis

To go further in the investigation of mitochondrial dysfunction in ChP organoids, we deeply investigated the proteomics data related to the oxidative phosphorylation in the context of HI. We found an impairment of complex I (Fig. EV4A), complex IV (Fig. EV4B) and complex V (Fig. EV4C) in HI condition in ChP. This impairment of mitochondrial oxidative phosphorylation in HI is summarized in Fig. EV4D. Moreover, in the context of HI, the ChP organoid displayed a downregulation of Drp1, a key protein in mitochondrial fission process (Fig. EV4E). This decrease of Drp1 abundance in HI, can lead to a diminution of neuronal differentiation leading to an impaired neurogenesis contributing to HI.

Biomarkers of HI in the iCSF

GO enrichment of significant DAPs between HI and He iCSF shows for biological process part, participation of proteins involved in cellular trafficking (Fig. 5E). A Cnetplot for biological processes was generated to further explore the pathways implicated in HI (Fig. EV1D). The analysis revealed significant pathways associated with viral transcription, viral gene expression, and protein target to ER (Fig. 5F). Additionally, an enrichment map and interaction plot were constructed to elucidate the relationship among proteins linked to the top 10 enriched GO biological processes in the context of HI (Fig. EV3A). Regarding cellular components, cellular adhesion is impaired including collagen-containing ECM, cell substrate junction, focal adhesion which are important components in ChP physiology. Cellular architecture is impaired including cadherin binding, ECM constituent, cytoskeleton but also cell signaling antioxidant activity (Fig. 5G). Cellular component was investigated by Cnetplot also, demonstrating the role of ribosome, ECM, cell-substrate junction associated with specific protein (Fig. EV1E). Enrichment map and interaction also highlight all proteins from cellular component involved in HI in iCSF (Fig S3B). Finally, top 10 significant molecular functions include cadherin binding and antioxidant activity with protein associated (Fig. 5.H, Fig EV1F). Antioxidant activity could highlight a cellular defense to oxidative stress, to avoid more cellular damage in HI. Enrichment map and interaction of molecular function are shown in Fig EV3C.

We then analyzed the 10 major significant proteins between He and HI iCSF (Fig. 6B). These top dysregulated proteins revealed an impaired epigenetic regulation with downregulation of many histones in the context of HI: histone-lysine N-methyltransferase EHMT2 and histone H2AZ but also regulation of DNA binding of transcription activator (Table EV6B). RT-qPCR data on ChP organoids confirmed downregulation of H2AZ (p < 0.05), secreted in iCSF by epithelial cells in the context of HI and decrease of mature neurons MAP2 (p < 0.05) and decreasing trend of neural stem cell (SOX2) in the context of HI (Fig. 6E). Correlation analysis was conducted on the gene expression data, revealing a positive correlation between H2AZ and PAX6 with a Spearman coefficient of 0.6626 (p value < 0.05). Similarly, H2AZ and SOX2 exhibited a positive correlation with a Spearman coefficient of 0.9097 (p value < 0.001), also for H2AZ and MAP2 with a positive Spearman correlation of 0.8596 (p value < 0.001). SOX2 and MAP2 showed a positive correlation with a Spearman coefficient of 0.8902 (p value < 0.001) and SOX2 and PAX6 a positive correlation with a Spearman coefficient of 0.7703 (p value < 0.01) and finally a positive correlation of MAP2 and PAX6 with Spearman correlation of 0.7370 (p value < 0.01). (Fig. 6F). Immunofluorescence of SOX2 and H2AZ on He ChP organoids slices revealed by confocal microscopy, a co-localization and thus, close relationship in regulatory network (Fig. 6G).

Discussion

We developed a ChP-CSF interface using COs unguided protocol from hiPSCs. After characterization to confirm the relevance of the ChP-CSF modeling, organoids were exposed to HI to elucidate molecular signature of this insult on the specific brain structure. Analysis after exposure to HI reveal alterations of the HIFα pathway, morphological and functional alterations such as ChP brush border of expression of transporters, mitochondrial dysfunctions as well as alterations in the abundance of proteins involved in neurogenesis. Key findings are summarized in Fig. 7.

Fig. 7.

Impact of HI in ChP-CSF brain interface. After HI-brain injury, ChP-CSF interface display impairment of morphology with brush border of ChP epithelial cells and functionality with disruption of SLC transporters. Downregulation of ITM2B and H2AZ in ChP cells and iCSF respectively, highlighted a deficient neurogenesis associated with SOX2 and MAP2. Mitochondria is compromised in the context of HI-brain injury with energy failure (ATP metabolic process), compromised oxidative phosphorylation and fission process contributing to HI-brain injury

The differentiation leads spontaneously to ChP organoids. In the absence of specific guidance, some of the neuroectodermal progenitors in organoid culture may differentiate into cells resembling those of ChP [19, 26, 27]. This could be due to the plasticity of PSCs and their ability to differentiate into various cell types based on developmental cues and micro environmental conditions. The potential reason for this spontaneous differentiation into ChP tissue could be justified by the modulation of the Wnt antagonist in GMB7-1 organoids. Members of the bone morphogenetic protein (Bmp) family are known to regulate specification of ChP epithelium by inducing MSX1 and repressing LHX2/ FOXG1 levels [28]. Here we reported an upregulation of Bmp1 in GMB7-1 organoids, which is crucial in early brain development especially for processing ECM components such as collagen [29] and is an important constituent of ChP tissue [30]. Bmp7 is produced by ChP cells and cortical hem, low levels of Bmp7 expression extends into the medial cortex [31]. A study found a novel role of ChP in producing Bmp7 leading to the inhibition of the differentiation of cerebellar neural progenitors [32]. This inhibition mediated though Bmp signaling pathways suppresses the expression of Math1, a critical factor for the differentiation of these progenitors into neurons [33]. SMAD5 governs Bmp target gene expression in early mammalian embryos [34] and a study published in 2017 highlights the role of SMAD5 in the neural development process [35]. FGF2, is more abundant in GMB7-1 organoids and plays a crucial role in developing and mature tissues and is present in the early development of ChP tissue in humans [36]. FGF signaling pathways are involved in preserving the integrity and function of this specialized epithelium. DKK1 is a secreted protein that selectively blocks Wnt/β-catenin signaling by binding to the co-receptor LRP6 [37, 38]. The restricted expression of DKK1 in GMB7-1 organoids suggests a specialized role in modulating local Wnt signaling within this microenvironment and may be essential for regulating development of ChP.

Analysis by proteomics as well as immunostaining confirmed the presence of key characteristics of the ChP such as epithelial cells (CLIC6, FOLR1 and TTR) responsible of secretion of CSF, tight junctions (CLD, ZO) and cell adhesion cadherin to form a selective barrier and structural integrity, collagen to provide structural support, transporters to control movement of ions and nutrients. We identified by immunofluorescence, presence of tight junction Claudin-5, as a general indicator of barrier-associated junctional proteins, however claudin-1, -2 and − 3 are more characteristic of the choroid plexus epithelial tight junctions. We found presence of claudin-2 among 100 common proteins between our in vitro ChP organoids and in vivo human ChP (Fig. 2A), serving as a representative marker of ChP epithelial identity [39]. Our hiPSCs-derived organoids recapitulate keys aspect of human choroid plexus structure including epithelial cells, junctional proteins, extracellular matrix and CSF-related proteins. To complete the chacraterization of this ChP-CSF organoids, it would have been interesting to quantify TTR positive area or forkhead box protein J1 (FOXJ1) by immunofluorescence or by Western Blot and compare to improve protocol using patterning factors such as BMP4 described in the paper of Pellegrini et al. [40]. It is also crucial to compare key transporters such as MDR1 and monocarboxylate transporter 8 (MCT8) between our model of ChP-CSF organoids with an optimized protocol using BMP4 for example to validate validity of organoids differentiated. Comparison of barrier functionality using fluorescent dyes, along with global proteomic analysis, would further validate the fidelity of our model.

To strengthen the relevance of the model we compared proteins from iCSF to published datasets of three-brain stages (embryonic [24], pediatric [23] and adult [22]). In the same manner, the ChP organoids were compared to the currently established human proteome atlas. This analysis reveals strong similarities between in vitro and in vivo data sets and also highlights the critical role of iCSF in neuron growth, aiding axon guidance, gliogenesis (Fig. 2D, F,H) as reported previously [41]. This iCSF composition has more similarities with in vivo human embryonic and pediatric CSF confirming the relevance of the model to study neonatal HI.

We also further investigated role of the 1,635 unique proteins identified in adult in vivo CSF to better understand differences in composition and function between in vivo and in vitro CSF. These proteins are associated with specific pathways such as regulation of synapse structure, or activity, cell projection organization, neuron projection development, synaptic signaling and transmission, cell-cell adhesion likely reflect the ongoing dynamic maintenance and functional activity of the mature CNS. The 2,912 proteins uniquely identified in in vitro CSF were mainly associated with interferon signaling, endosomal and vesicle transport, membrane trafficking, RNA metabolism and catabolic and cell cycle processes. Together these pathways indicate an active secretory and regulatory environment within the organoids, reflecting high cellular turnover, vesicle-mediated communication and immune or stress-related signaling. The difference may arise because adult in vivo CSF is produced by a fully developed brain; where proteins mainly reflect established synaptic network and homeostatic signaling. In contrast, organoid-derived CSF comes from proliferating and differentiating cells, so its composition is dominated by proteins involved in secretion, vesicle trafficking, and regulatory processes necessary for growth and cellular remodeling. The 150 unique proteins in human pediatric CSF are involved in phospholipase activity, cytoskeleton and synapse organization, lipid metabolism, cell junctions and adhesion, neuron recognition and metalloproteinase activity. The presence of this protein indicates active processes of cell differentiation, synaptic formation and tissue organization, critical during early brain development. Regarding the iCSF from organoids, the 4,259 unique proteins are mostly involved in autophagy, lipid metabolism, membrane trafficking, viral processing, RNA metabolism, axon guidance and actin cytoskeleton organization and are essential for cell survival, differentiation, and communication during development. The 55 proteins found in human embryonic CSF are involved in key developmental pathways including growth regulation, protein digestion and absorption, cell-substrate adhesion, metabolic process, blood vessel development and vascular process, regulation of anatomical structure size, inflammatory response and extracellular matrix organization. These pathways are interconnected, forming a complex network that regulated embryonic development. For the 4,326 unique proteins found in iCSF from organoids, these proteins involved in cell cycle regulation and RNA metabolism, autophagy, membrane trafficking and intracellular protein transport, cytoskeleton organization and axon guidance reflect the active state of proliferating, differentiating cells in the brain development.

These differences between in vivo and in vitro CSF proteins and pathways may be partly explained by methodological factors, such as the use of different proteomic techniques, which could introduce bias. Additionally, the absence of CSF circulation in vitro may lead to protein accumulation and the 4 months maturation stage of the organoids could result in a mixture of proteins reflecting different developmental stage compared to human CSF.

To understand more deeply the impact of HI on the ChP-CSF interface, our model was exposed to a low oxygen atmosphere. We first focus on the HIFα pathway modulation since this is the key transcriptional regulator for the response to HI [42]. Several players of the HIF1α were highlighted as dysregulated when comparing normoxic and HI ChP organoids by transcriptomic. After exposition to HI, the analysis of the transcriptome revealed that ENO1, a major glycolytic enzyme, catalyzes the conversion of 2-phosphoglycerate to phosphoenolpyruvate in the glycolytic metabolic pathway is regulated by HIF1α [43–45]. This upregulation of ENO1 could highlight an increased metabolic demand, a compensatory mechanism to meet the increased energy demand during HI. This significant DEG could also reflect an inflammatory response and overexpression of ENO1 could help cells to adapt by promoting cell survival, tissue repair and reduced oxidative stress. TFRC is a cellular receptor required for cellular iron uptake by mediating endocytosis, important for cell growth and is dysregulated in the context of HI [46]. Upregulation of TFRC could serve several roles in ChP including metabolic support, increased iron acquisition and cellular protection. Transcriptomics revealed upregulation of proteins ENO1 and TFRC involved in energy metabolism and cellular protection, modulated by HIF1α pathways and confirmed our model of HI. HIF1-α is known to play a critical role in angiogenesis by regulating VEGF and in energetic metabolism [47]. The downregulation of VEGF as well as modulations of other players such as PAI-1 in context of HI [48, 49] are major vascular pathways dysfunction in HI. This downregulation of VEGF in HI can lead to a reduction of vascular permeability and impaired angiogenesis. In the ChP this may result in disrupted blood-CSF barrier function, but also weaken tight junction between ChP cells, increasing permeability. Modulation of proteins associated with HIF1α pathways and involved in energy metabolism such as LDHA and ALDOA further strengthen the implication of this pathway.

Next, we examined the morphology-related proteins and transporters expression changes in the HI ChP organoids to corroborate in vivo data [16, 50]. Consistently, transcriptomic data revealed dysregulation of brush border assembly and significant pathways involved in microvillus length, actin filament elongation, tight junction, and regulation cell shape and actin cytoskeleton organization. Impaired structure of ChP cells can significantly disrupt the integrity and function of the blood-CSF barrier in HI. This could lead to compromise selective permeability, affecting molecules transports. Disruption of actin filaments and tight junction can result in increased leakage of molecules into CSF, potentially causing inflammation and edema in the brain [51, 52].

Transporters expression of ChP is also impaired with downregulation of specific transporters including ion transporters, mineral absorption, vitamin transporter and SLC transporters (SLC28A2, SLC5A1, and SLC26A3). Transporters in ChP cells, especially SLC, play a crucial role in regulating the movement of ions, nutrients and waste products between CSF and the bloodstream. These transporters are essential for maintaining the brain’s microenvironment, barrier integrity and supporting processes of CSF production and composition [5]. This dysregulation of ChP transporters could have significant impact in barrier integrity and function, influx and efflux mechanism impairment could lead to an impaired detoxification or waste removal, inducing toxic accumulation of proteins. While our omics analyses of ChP organoids under healthy and hypoxic conditions provide comprehensive molecular profiles, it is important to recognize that changes in gene or protein expression do not always directly translate into functional alterations of the ChP–CSF interface. Furthers studies need to be done to evaluate functional aspect of ChP-CSF organoids in the context of hypoxic brain injury including assessment of barrier integrity using fluorescing dye (dextran), transporter activity or drug permeability assays.

Proteomics revealed changes of the abundance of proteins involved in energy metabolism and mitochondrial function. Downregulation of mitochondrial respiration highlighted by proteomics suggests enhanced oxidative stress, potential cell death, resulting in greater energy deficit and cellular stress, inability of the cell to maintain ion gradients, repair cellular damage and perform essential metabolic functions (Fig. EV4).

Bioinformatics analysis of the proteomic data set highlighted ITM2B as the top deregulated protein abundance in HI ChP. ITM2B is known to play a crucial role in neurite outgrowth and neuronal differentiation and its downregulation could affect neurogenesis [53, 54]. Proteins related to ITM2B, such as MAP4K5, DSG1, ERBB3 and APOA1 were also highly deregulated. Downregulation of MAP4K5, actor of the MAP kinase-signaling cascade, suggests impaired cell proliferation, survival, and apoptosis within the HI ChP [55]. Similarly, decrease ERBB3, a receptor tyrosine-kinase actor of the PI3K/AKT signaling, suggests impairment of neuronal cell survival and growth [56]. Interestingly, Henna Shaikh et al. also found a decrease of ERBB3 after 24 h and 48 h in the plasma of neonates affected with HIE [57]. Downregulation of APOA1, crucial for cholesterol transport and maintenance of lipid homeostasis in the brain, suggests detrimental lipid imbalance and altered signaling functions [58]. This interconnected network emphasizes how deficit of ITM2B could affect multiple pathways potentially leading to disrupts neurogenesis and cellular stability in HI. To our knowledge, deregulation of ITM2B and related pathways have never been investigated before in the context of HI.

In iCSF, the enrichment analysis of GO biological processes highlighted significant viral gene expression and transcription in the context of HI. These data could emphasize a viral mimicry in response to HI, that cells in HI are activating pathways that mimic viral infection due to cellular stress. This could lead to the upregulation of genes associated with viral gene expression or even inducing a viral-like pattern of gene expression without actual viral presence [59]. There is also a crosstalk between HIFs and pathways involved in immune responses, leading to the expression of genes associated with viral infection [60, 61]. Proteomic analysis of iCSF revealed several key proteins that could be considered as biomarkers of HI such as H2AZ. H2AZ, a variant of histone H2A, is a key player of epigenetic regulation involved in neurogenesis [62]. This epigenetic deregulation could lead to altered neuronal function, development and plasticity [63]. In addition, H2AZ is recruited to the promoters of genes involved in the HIF1-a pathways, promoting their transcription, H2AZ mediates the chromatin accessibility of HIF1α [64]. This further confirms the relevance of this marker in the context of HI.

RT-qPCR data confirmed proteomic analysis with downregulation of the expression of the genes encoding top protein ITM2B and H2AZ in HI. In addition, H2AZ is positively correlated with SOX2 and PAX6 demonstrating a key role of this histone variant in neurogenesis influencing neural stem cell and radial glial cells in ChP in the context of HI. What’s more, SOX2 and MAP2 are also positively correlated highlighting the crucial interaction between neural stem cells and mature neurons in HI contributing to brain lesions. This finding suggests that these proteins are involved in a shared regulatory network, which influence gene expression during neuronal differentiation and stem cell maintenance. This co-localization likely reflects functional relationship possibly involving chromatin remodeling and transcriptional regulation to influence neurogenesis [62].

However, this study may have few limitations. The use of a single hiPSC line may not capture the full diversity of cellular responses to experimental variables. Besides the choroid plexus epithelial cells, there are several others cell types that contribute to its structure and function including vascular cell types (endothelial cells and pericytes), immune cells or stromal cells. ChP-CSF organoids do not display these cell populations and need optimization to develop a more relevant platform to understand pathophysiology of hypoxic brain injury on ChP-CSF interface. In addition, sexual dimorphism in HI is a well described phenomenon that we could not address in this study. GMB7-1 hiPSC is a female cell line that spontaneously differentiate into ChP-CSF organoids using an unguided protocol of differentiation. Further research is needed by increasing the number of hiPSC cell lines from both sexes and a guided choroid plexus differentiation protocol described in the paper of Pellegrini et al. [40] with specific patterning factors such as BMP4 to understand the complete range of cellular response to HI at the ChP-CSF interface.

Conclusion

Altogether, our observations in a human in vitro interface of HI ChP-CSF are consistent with previous data and support altered barrier morphology and functions, mitochondrial/energy imbalance leading to modification of CSF composition and neuronal dysfunctions [15, 65, 66]. Equally, to the BBB, this understudied barrier appears as central in the context of HI since it ensures production of CSF and maintenance of brain homeostasis. The consistency of modulations observed in the ChP and the iCSF further strengthens the relevance of this model to identify new molecular signatures and biomarkers of HI.

The CSF presents promising potential as a source of biomarkers for the early detection and diagnosis of brain disorders. Changes in CSF composition due to HI could provide valuable diagnostic insights, enabling better disease monitoring and tailor therapeutic approaches. This study shows that the ChP-CSF interface is a relevant tool to study pathophysiology of HI.

Methods

Experimental design

In this study, we generated from hiPSCs choroid plexus organoids secreting CSF-like fluid in vitro, and exposed them to low oxygen atmosphere to study hypoxic brain injury. We investigated by transcriptomic and proteomic analysis molecular signature of choroid plexus and CSF fluid in vitro in hypoxic condition. Confocal microscopy and PCR were carried out as additional experiments to strengthen the results obtained.

In this material and methods section, we will describe human iPSC generation and characterization, cerebral and choroid plexus organoid generation, hypoxic modeling, immunofluorescence, RT-qPCR, proteomic and transcriptomic analysis and statistical analysis. All the reagents used in this study are described in supplementary material (table EV8).

Human IPSC generation and characterization

Reprogramming

Cell line MAA3 (SUi001-A) and GMB7-1 (RCNSi003-A) are registered in hPSCreg https://hpscreg.eu/ and MAA3 is already published (PMID: 33246498). Blood samples were collected from He donors with written informed consent, following the approval of the Human Reproduction Committee of the Hungarian Health Science Council (ETT HRB-Approval number: 42592-2/2016-EHR). Peripheral blood mononuclear cells (PBMCs) were isolated using sodium citrate cell separation tubes. 5 × 10^5 cells were plated in 24-well plates with 1 ml of PBMC medium, which included StemPro-34 medium, StemPro-34 Supplement, 2 mM GlutaMax, 10 µM β-mercaptoethanol, 1% Antibiotic-Antimycotic, and cytokines (20 ng/ml IL-3, 20 ng/ml IL-6, 100 ng/ml FLT-3, 100 ng/ml SCF). The medium was refreshed daily for 2 days. Reprogramming of PBMCs was conducted using Sendai virus vectors (CytoTune-iPS 2.0) expressing Oct3/4, Sox2, Klf4, and c-Myc. The reprogramming factors were removed the following day, and cells were replated in a fresh PBMC medium. After 24 h, enlarged cells were observed and subsequently seeded on mitomycin C-treated mouse embryonic fibroblast (MEF) cells in a supplemented StemPro-34 medium. Transition to hiPSC medium began by replacing half of the medium with KO-DMEM/F-12 supplemented with KO-Serum Replacement, MEM Non-Essential Amino Acids, GlutaMax, β-mercaptoethanol, bFGF, and Antibiotic-Antimycotic. The medium was changed daily. Once hiPSC colonies appeared, they were transferred to MEF culture for expansion, then to Matrigel-coated plates in mTeSR1 medium. Cells were passaged with Accutase at ~ 90% confluency and plated in mTeSR1 medium with 10 mM Rock inhibitor, with daily medium changes thereafter.

ATCC cell line BXS0115 human induced pluripotent stem cells were reprogrammed from parental cell line CD34- bone marrow cells using Sendai virus expression of oct4, klf4 and Myc genes.

Human iPSC culturing

Human iPSC cell lines: GMB7-1 and MAA3 were obtained from the Institute of Molecular Life Sciences, HUN-REN Research Centre for Natural Sciences, Hungary. hiPSC cell line BXS0115 was obtained from American Type Culture Collection (ATCC), ATCC number ACS-1029, lot number 0232 from parental cell line CD34⁺ Bone marrow cells, latino or hispanic female donor.

Human iPSC cell lines were maintained on Geltrex (Fisher Scientific, Geltrex™ LDEV-Free, hESC-Qualified, Reduced Growth Factor Basement Membrane Matrix) in mTESR1 medium (STEMCELL Technologies) and passages using ReLeSR (STEMCELL Technologies). iPSCs were negative for mycoplasma contamination (MycoAlert, Lonza). Cell passages were inferior to 40.

Assessment of pluripotency

ATCC cell line BXS0115 pluripotency was checked by surface antigen expression of stem cell markers with 96.26% SSEA4, 99.15% Tra-1-60 and 0.08% SSEA1 and germ layer differentiation with pluripotency scores: 29.41, 26.38 and novelty score: 1.34, 1.21. The pluripotency of the MAA3 and GMB7 induced pluripotent stem cell (iPSC) lines was assessed through several key analyses. The presence of pluripotency markers, specifically Oct4, Nanog, and SSEA4, was confirmed using immunocytochemical techniques. Additionally, the spontaneous differentiation potential of the iPSC lines was evaluated through immunocytochemistry and reverse transcription quantitative polymerase chain reaction (RT-qPCR) methods. Furthermore, the normal karyotype of the cell lines and no mycoplasma contamination were verified to ensure the integrity of the cultures (for further details see PMID: 33246498).

Cerebral and ChP organoids culture conditions

For the cerebral (COs) and ChP organoids culture conditions, we followed the protocol published by Lancaster & Knoblich [19] with minor modifications [20, 21]. Embryoid bodies (EBs) were prepared from single cell suspension from hiPSC as previously described. A suspension of 15.000 cells in 20µL of EBs medium containing Y-27,632 rock inhibitor was placed as hanging drops on a petri dish. After aggregation, the drops were collected and transferred into a non-treated 24-well plate (Sarstedt) for 4 days and moved to an incubator with an orbital shaker. After a few days in the EBs medium, neural induction was applied to EBs for neuroectoderm development and then EBs were embedded into an extracellular matrix Matrigel (STEMCELL Technologies) using an embedding sheet for expansion of neuroectoderm. Organoids were fed one time per week at the beginning and after 2 months of maturation, they were fed 3 times in 15 days. Organoids were used at 4 months old.

HI modeling

ChP and COs organoids were exposed to low oxygen conditions on day 120 by using a hypoxic chamber (C-Chamber Three Shelf, Biospherix) and an oxygen controller (Pro Ox C21 Oxygen CO2 Single Chamber Controller). ChP and COs organoids were cultured in fresh COs medium and transferred to a hypoxic chamber at 1% oxygen for 24 h. After hypoxic exposition, ChP and COs were transferred into an incubator (normoxic condition) for another 24 h (for reoxygenation).

Immunofluorescence

Organoids were fixed in 4% of formaldehyde for 45 min and then moved to 30% sucrose buffer overnight at 4 °C. Organoids were then embedded in optimal cutting temperature compound (OCT) and flash frozen at -80 °C. Sections of 20 μm were cut using cryostat (HM560 Microm Microtech). Organoid slides were permeabilized with 0.3% of Triton and blocked with 2% of goat serum (Gibco, 162210064) and 1% bovine serum albumin (BSA). Section were incubated at 4 °C overnight with following primary antibodies: SOX2 1:100 (Abcam, ab93689, Sigma-Aldrich SAB5300177), PAX6 1:100 (Sigma-Aldrich, AMAb91372) NeuN 1:100 (Sigma-Aldrich, MAB377), MAP2 1:100 (Sigma-Aldrich, ZRB2290), GFAP 1:200 (Sigma-Aldrich, G3893), CLD5 1:100 (Thermo Fisher Scientific, 35-2500), ZO1 (Thermo Fisher Scientific, 33-9100), GLAST 1:100 (Thermo Fisher Scientific, PA5-72895), MDR-1 (Santa-Cruz Technology, sc-13131) and H2AZ (Active Motif, 39943). After PBS washes, DAPI (100-1000ng/mL) and secondary antibodies labeled with Alexa-Fluor 488 or 594 1:1000 (Invitrogen, #A-11037, #A-11029) were incubated for 1 h. Finally, slices were mounted using Fluoromount Aqueous Mounting Medium (Sigma Aldrich). Images were acquired using a microscope Axio Observer (Carla Zeiss), objectives 10X, 20X, 40X and 63X. For co-localization of H2AZ and SOX2, Leica SP8 confocal microscopy was used at 40X objectives. Images were analyzed using FIJI software.

RNA extraction, RT-qPCR

Total RNA was extracted from ChP organoids with RNeasy Plus Universal Tissue Mini kit (Qiagen) and Precellys Evolution tissue homogenizer (Bertin). The concentration and purity of RNA samples were checked using the NanoDrop ND-1000 spectrophotometer at 260 and 280 nm (NanoDrop Technologies); the A260/280 ratio ranged from 1.8 to 2.2. One µg of total RNA was converted to cDNA with random primers using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystem) according to the manufacturer’s protocol. cDNA was diluted with sterile water RNase/DNase free to a final volume of 230 µL. Quantitative expression of markers were determined using 2 µl of diluted cDNA for each primer set at 10 µM using iQ SYBR Green Supermix (Biorad) in a final volume of 12 µl. Thermocycling was carried out in the CFX96 RT-PCR detection system (Bio-Rad) using SYBR green fluorescence detection. The amplification cycle used was as follows: 10 min at 95 °C, followed by 40 amplification cycles at 95 °C for 15 s, 60 °C for 60 s, and 72 °C for 30 s to reinitialize the cycle again. The specificity of each reaction was also assessed by melting curve analysis to ensure primer specificity. Relative gene expression values were calculated as 2^−ΔCT, where ΔCT is the difference between the cycle threshold (CT) values for genes of interest and housekeeping gene. Housekeeping genes were selected using refinder website by a comprehensive gene stability and by delta CT method (selection of human PPIA and EIF4A1 genes).

Shotgun proteomics

The MAA3, BXS0115 and GMB7-1 cell lines were explored for He condition to understand the differentiation process of GMB7-1 cell line into ChP organoid compared to COs (MAA3/ BXS0115). Nine organoid samples for BXS0115 and GMB7-1 He condition and four organoid samples for MAA3 He condition were used for the study. Only GMB7-1 cell line has been further explored for HI condition (1% O₂ 24 h + 24 h normoxia) in comparison with He condition (normoxia). For GMB7-1 cell line organoids, ChP body and fluid filled cavities: in vitro CSF (iCSF) were separated and analyzed by proteomic. For proteomic analysis of GMB7-1 cell line, nine organoid samples for He condition and 9 organoids samples for HI condition were used. Regarding iCSF, seven iCSF samples for He condition and 7 iCSF samples for HI condition were analyzed from GMB7-1 ChP organoids.

Regarding protein extraction, organoids were homogenized in fresh prepared lysis buffer of TBS1X (Biorad) supplemented with protease inhibitor cocktail 1X (cOmplete, Roche) and a mix of anti-phosphatases inhibitors 1X (ammonium molybdate, sodium glycerophosphate, sodium fluoride, sodium pyrophosphate, sodium orthovanadate) using a Precellys Evolution tissue homogenizer (Bertin). Samples were then centrifuged at 10.000 g for 20 min to obtain lysates for proteomic analysis. For iCSF, ChP organoids were washed with PBS1X and then fluid inside cavities was aspirate using a 300µL syringe and then centrifuge at 10.000 g for 10 min to remove cells debris (pellet). Proteins levels were quantified by Bradford assay (ThermoFisher Scientific).

Protein extracts (40 µg of total proteins in LDS 1X, Thermo) were heated for 5 min at 99 °C, and then subjected to NuPAGE electrophoresis for a short migration of 4 min. The proteomes were treated and subjected to trypsin proteolysis as previously described [67]. The resulting peptides were analyzed on an Orbitrap Exploris 480 (Thermo Scientific) high-resolution tandem mass spectrometer coupled to a Vanquish Neo UHPLC (Thermo Scientific). The instruments were operated as reported [68]. Peptides were desalted on a reverse-phase PepMap 100 C18 µ-precolumn (5 mm, 100 Å, 300 mm i.d. × 5 mm, Thermo Scientific) and separated on a 50-cm EasySpray column (75 mm, C18 1.9 mm, 100 Å, Thermo Scientific) at a flow rate of 0.250 µL/min using a 90-min gradient (5–25% B from 0 to 85 min, and 25–40% B from 85 to 90 min) of mobile phase A (0.1% HCOOH/100% H2O) and phase B (0.1% HCOOH/100% CH3CN). Tandem mass spectrometry results were acquired in Data-Independent Acquisition mode with the following parameters: m/z isolation window of 4, m/z window overlap of 1, a normalized collision energy type, an Orbitrap resolution of 30,0000, a precursor mass range between m/z 400 and m/z 1008, and a number of scan events of 151. The raw data were interpreted using the SwissProt human annotated genome (2024_02 release) with DIA-NN 1.8 as recommended [69]. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD063032. [The reviewers may access this currently private dataset using https://www.ebi.ac.uk/pride/ website with the username as reviewer_pxd063032@ebi.ac.uk and password as KuKYh6UDdIcx; or alternatively, reviewer access details with Project accession: PXD063032 and Token: dCVcxGOdrd4g].

[AJ1]https://pubmed.ncbi.nlm.nih.gov/23727365/.

[AJ2]https://pubmed.ncbi.nlm.nih.gov/38454512/.

[AJ3]https://pubmed.ncbi.nlm.nih.gov/32133743/.

Proteomics data analysis

Proteomic analysis comparing healthy (He) and hypoxia-treated (HI) conditions was performed on choroid plexus (ChP) organoids and isolated cerebrospinal fluid (iCSF) samples, each initially comprising 8,981 proteins. Proteins with zero mean expression in either group were excluded (644 in ChP; 2,489 in iCSF) to ensure analytical robustness. Normality was assessed using Shapiro-Wilk tests (α = 0.05), with normally distributed data analyzed via independent t-tests and non-normal data via Mann-Whitney U tests. Uncorrected significance thresholds (p ≤ 0.05) identified 1,341 differentially expressed proteins in ChP and 533 in iCSF. To prioritize biologically relevant candidates, proteins with median expression > 4 in both conditions were retained, and effect sizes were quantified as He/HI median fold change ratios, with results sorted by descending magnitude.

Differentially expressed proteins between HI and He conditions in ChP organoids and iCSF samples were widely annotated using several databases including Gene Ontology (GO) of Biological processes, Cellular Components and Molecular Functions and Kyoto Encyclopedia of Genes and Genomes (KEGG) database from MSigDB and Reactome gene sets at a significance level of p-value ≤ 0.01.

Transcriptomic

Total RNA extraction

Frozen ChP organoids were sectioned using Cryostat Cryotome (ThermoFisher Scientific, USA). Totally 90 µM was collected for RNA extraction. The total RNA was isolated using the RecoverAll total nucleic acid isolation kit (ThermoFisher Scientific, USA) according to the manufacturer’s instructions. Low quality RNA was purified using Zymo RNA Clean and Concentrator-5 (Zymo Research, USA), quantified using a Qubit HS RNA assay kit, and stored at -80 °C for downstream analysis of transcriptomics. Totally six biological replicas of ChP organoids were used for the transcriptomics analysis including three as HI ChP organoids and three as He ChP organoids.

Whole transcriptome sequencing

Genomic DNA removal from RNA samples was ensured by treating the RNA with Turbo DNase (ThermoFisher Scientific, USA). Next, whole transcriptome sequencing was performed using targeted RNA-Seq with Ion AmpliSeq whole transcriptome human gene expression kit (ThermoFisher Scientific, USA). Briefly, cDNA was synthesized using a SuperScript VILO cDNA Synthesis kit (ThermoFisher Scientific, USA) and amplified using Ion AmpliSeq human gene expression core panel primers. The amplified products were subjected to enzymatic shearing to get amplicons of ~ 200 bp, then ligated with the adapter and the unique barcodes. Next, the constructed library was purified using Agencourt AMPure XP Beads (Beckman Coulter, USA), quantified using an Ion Library TaqMan™ quantitation kit (Applied Biosystems, USA), further diluted to 100 pM, and pooled equally with 16 individual samples. The diluted library was amplified and enriched in the Ion Chef System (ThermoFisher Scientific) according to the manufacturer’s instructions. Ion 540 Chip was used to sequence the prepared template libraries using an Ion S5 XL Semiconductor sequencer (ThermoFisher Scientific).

RNA-Seq data analysis

RNA-seq data analysis was performed using the Torrent Mapping Alignment Program (TMAP) (https://github.com/iontorrent/TMAP). Alignment of the raw sequencing reads generated by the Ion Torrent sequencing data against the reference sequence derived from the hg19 (GRCh37) assembly was carried out using nested alignment. Each of the short alignment algorithms generates contigs of a certain size which is then passed to the next alignment algorithm generating longer contigs and eventually applying de novo alignment using the SMEM algorithm [70]. Raw read counts of the targeted genes were extrapolated using samtools (samtools view –c –F 4 –L bed_file bam_file). RNA-seq data was normalized using Fragments Per Kilobase Million (FPKM) normalization [71] prior to the quality control check. Differentially expressed genes (DEGs) analysis was performed using DESeq2 with p-value < 0.05. For displaying DEGs between He ChP and HI ChP organoids, a volcano plot was performed.

Gene set enrichment analysis for the DEGs between he and HI ChP organoids

The significant DEGs obtained were further analyzed to identify the activated and enriched cellular pathways in ChP in response to HI using the absolute GSEA, as previously described [72]. Absolute GSEA was performed on expression data using around 90,000 annotated cellular pathways obtained from the Broad Institute’s database (https://www.gsea-msigdb.org/gsea/index.jsp).

For significant enrichment pathways, a threshold of p < 0.05 and FDR < 0.25 was used. Enrichment pathways were further explored to identify the differentially enriched genes in HI ChP to He ChP by performing leading-edge analysis as previously described [73]. The resulting gene sets were further reduced by carrying out a systematic cross-reference of each gene enriched within statistically significant pathways. Finally, genes that are highly frequent across multiple significant pathways enriched and showed T-test p-value < 0.05 between the HI ChP and He ChP organoids were identified. The functional clustering and pathway analysis of the gene-set of interest from GSEA were identified using Metascape tool [74] available at https://metascape.org/gp/index.html#/main/step1.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 9.3 program. Experimental comparisons with two groups were analyzed using two-tailed student t-test. For correlation analysis, Pearson correlation coefficient was used to assess the strength and direction of linear relationships between continuous variables, assuming normal distribution. For variables that did not meet the assumption of normality, we applied Spearman’s rank correlation, a non-parametric test to examine monotonic relationship. To account for multiple comparisons, Bonferroni correction was applied to adjust the p-values with the threshold for statistical significance set a 0.05.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

absGSEA

Absolute GSEA

ACSM3

Acyl-coenzyme A synthetase mitochondrial 3

ATCC

American Type Culture Collection

BBB

Blood-Brain Barrier

BMP

Bone morphogenetic protein

BSA

Bovine serum albumin

ChP-CSF

Choroid plexus-cerebrospinal fluid

CLD

Claudin

CLIC6

Chloride Intracellular Channel 6

CNS

Central Nervous System

COs

Cerebral organoids

CT

Cycle threshold

DAPs

Differentially abundant proteins

DEGs

Differentially expressed genes

DKK1

Dickkpof-1 protein

EBs

Embryoid bodies

ECM

Extracellular matrix

EHMT2

Histone-lysine N-methyltransferase

ENO1

Enolase 1

ERBB3

Receptor tyrosine-protein kinase

FOLR1

Folate receptor 1

GFM1

Elongation factor mitochondrial 1

GO

Gene ontology

H2AZ

Histone 2 A.Z

He

Healthy

HI

Hypoxic brain injury

HIE

Hypoxic-ischemic encephalopathy

hiPSCs

Human induced pluripotent stem cells

iCSF

in vitro cerebrospinal fluid

ITM2B

Integral membrane protein 2

MRPL

Mitochondrial large subunit ribosomal proteins

MSigDB

Molecular Signature Database

OCT

Optimal cutting temperature compound

PBMCs

Peripheral blood mononuclear cells

PCA

Principal component analysis

RT-qPCR

Real time polymerase chain reaction

TFRC

Transferrin receptor

TTR

Transthyretin

ZO

Zonula occluden

Author contributions

AM was responsible for project conceptualization, administration, and ANR funding acquisition, data analysis and writing of the manuscript. RGB conducted the majority of the experiments, analyzed the data and prepared the figures and writing of the manuscript. AB, EA, JA conducted the proteomics data analysis, generated figures and contributed to the writing of the manuscript. AMA conducted the transcriptomics experiment, transcriptomics analysis, figures generation, writing and editing the manuscript. MS, RAH and RH were involved in bioinformatics data analysis. NC was involved in the Q-PCR experiments. CD was involved in the experiment supervision and the writing of the manuscript. AA and BS were involved in the generation of iPSC cell lines. All the authors have reviewed the manuscript.

Funding

The current study was supported by ANR (ANR-21-CE17-0019-01); MS and EA were supported by Russian Science Foundation grant N°25-71-30008.

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.The mass spectrometry proteomics data corresponding to the 60 nanoLC‒MS/MS runs have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD063032. [The reviewers may access this currently private dataset using https://www.ebi.ac.uk/pride/ (https://www.ebi.ac.uk/pride/) website with the username as reviewer_pxd063032@ebi.ac.uk and password as KuKYh6UDdIcx; or alternatively, reviewer access details with Project accession: PXD063032 and Token: dCVcxGOdrd4g].

Declarations

Ethics approval and consent to participate

GMB7-1 (RCNSi003-A) are registered in hPSCreg https://hpscreg.eu/ and MAA3 is already published (PMID: 33246498). Blood samples were collected from He donors with written informed consent, following the approval of the Human Reproduction Committee of the Hungarian Health Science Council (ETT HRB-Approval number: 42592-2/2016-EHR). BXS0115 was obtained from American Type Culture Collection (ATCC), ATCC number ACS-1029, lot number 0232 from parental cell line CD34⁺ Bone marrow cells, latino or hispanic female donor.

Declaration of generative AI and AI-assisted technologies

We did not use AI assisted technologies for this article.

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Dr Aloïse Mabondzo (aloise.mabondzo@cea.fr).

Competing interests

The authors declare no competing interests.