key: cord-0294431-hwelr3ob
authors: Kumarasamy, Murali; Sosnik, Alejandro
title: Multicellular Organoids of the Neurovascular Blood-Brain Barrier: A New Platform for Precision Neuronanomedicine
date: 2020-08-15
journal: bioRxiv
DOI: 10.1101/2020.08.14.249326
sha: 119d21251fec69557e837e8534900168f75ca39e
doc_id: 294431
cord_uid: hwelr3ob

The treatment of neurological disorders (NDs) is challenged by low drug permeability from the systemic circulation into the central nervous system (CNS) owing to the presence of the blood-brain barrier (BBB). Neuronanomedicine investigates nanotechnology strategies to target the brain and improve the therapeutic outcome in NDs. Two-dimensional adherent cell BBB models show substantial phenogenomic heterogeneity and their ability to predict the permeability of molecules and nanoparticles into the brain is extremely limited. Thus, the high-throughput screening of CNS nanomedicines relies on the use of animal models. To address this dearth, 3D organoids that mimic the in vivo physiology are under development. Still, there exist concerns about the standardization and scale-up of the production process, their proper characterisation, and their industrial application. In this work, we report on a novel multicellular organoid of the neurovascular blood–brain barrier (NV-BBB) that recapitulates the regulated syncytium of human endothelial cells and the function of the human BBB. For this, an advanced organoid comprising human brain microvascular endothelial cells, brain vascular pericytes and human astrocytes combined with primary neurons and microglia isolated from neonate rats is bio-fabricated without the use of an extracellular matrix. The structure and function are fully characterized by confocal laser scanning fluorescence microscopy, light sheet fluorescence microscopy, scanning transmission electron microscopy, cryogenic-scanning electron microscopy, western blotting, RNA-sequencing and quantitative gene expression by quantitative polymerase chain reaction analysis. This bulk of these self-assembloids is comprised of neural cells and microglia and the surface covered by endothelial cells that act as a biological barrier that resembles the BBB endothelium. In addition, the formation of neuron-microglia morphofunctional communication sites is confirmed. Analysis of key transcriptomic expressions show the up-regulation of selected BBB-related genes including tight junction proteins, solute carriers, transporters of the ATP-binding cassette superfamily, metabolic enzymes, and prominent basement membrane signatures. Results confirmed the more efficient cell-cell communication in 3D organoids made of multiple neural-tissue cells than in 2D endothelial cell monocultures. These multicellular organoids are utilized to screen the permeability of different polymeric, metallic, and ceramic nanoparticles. Results reveal penetration through different mechanisms such as clathrin-mediated endocytosis and distribution patterns in the organoid that depend on the nanoparticle type, highlighting the promise of this simple, reproducible and scalable multicellular NV-BBB organoid platform to investigate the BBB permeability of different nanomaterials in nanomedicine, nanosafety, and nanotoxicology.

Neurological disorders (NDs) cause approximately 17% of the deaths worldwide and an enormous economical and societal burden. [1, 2] A major limitation in the treatment of NDs is that most drugs do not cross the blood-brain barrier (BBB). [3] The BBB is formed by tightly bound endothelial cells and is an essential part of the neurovascular unit (NVU), a complex anatomical and functional multicellular structure comprised of a basal lamina covered with pericytes, smooth muscle cells, neurons, glia cells, an extracellular matrix (ECM), as well as a number of different neural stem/progenitor cells. [4] Understanding the central nervous system (CNS) pathways in health and disease as well as the evaluation of novel neurotherapeutics has been challenging due to the complexity of the NVU. [5] The use of nanotechnology to improve the delivery of neurotherapeutics to the CNS, a field coined neuronanomedicine, has emerged as one of the most dynamic research areas in nanomedicine. [6] Different strategies have been investigated to surpass the BBB by systemic (e.g., intravenous) and local (e.g., nasal) administration routes. [6, 7] More recently, nanotoxicology, the discipline that investigates the toxicity of nanomaterials, has devoted efforts to develop reliable models to assess the detrimental interaction of different nanomaterials with the CNS upon intentional or unintended exposure. [8] The systematic investigation of the biocompatibility, safety, permeability, and efficacy of neuronanomedicines remains mostly limited to in vivo experiments. However, the complex physiology of animal models challenges the conduction of permeability and mechanistic studies to understand the transport of nanoparticles (NPs) into the CNS. [9] For years, endothelial cell monolayers cultured on semipermeable membrane well-plates (e.g., Transwell ® ) have been the most commonly and widely used in vitro models of the BBB. [10] They utilize user-friendly setups, are scalable and enable high-throughput screening. However, they have been questioned because they cannot mimic the complex 3D cellular structure, the physiological microenvironment, the cellular phenotype and the cell-cell interactions in the NVU. [11] In addition, they exhibit edge effects that hinder cell growth, especially at the plate edges. Thus, the surface area of the semipermeable membrane may not be fully covered by cells, which artificially increases the permeability. More recently, the development of 3D cell culture models has gained attention to investigate the transport of different neurotherapeutics into the brain. [12] These models also represent a valuable tool to investigate pathophysiological pathways in the CNS. [13] Advantages of 3D organoids include easy and reproducible culture, miniature scale, small reagent volumes, low relative cost, reproducibility, and scalability. Furthermore, they reduce animal experimentation. [14] Most of these 3D models are one-cell or three-cell cultures and fail to fully recreate the NVU-BBB. [15] One usually missing cell type in these models is the so-called 'third element' of the CNS, which is in fact resident macrophages (microglia) that constitute 10-15% of the total cells in the brain. [16] We recently reported on the possible role of olfactory microglia in the nose-to-brain transport of different nanomaterials. [17] This mechanism might be also exploited by pathogens such as the coronavirus-19 to enter the CNS through the olfactory epithelium, [18] though the key cell players in this pathway are not known yet. In addition, glial cells (e.g., astrocytes), microglia and neurons are actively involved in neuro-haemostasis and regulate the transmission of electrical signals in the brain. [19] Despite recent advances in the development of cell spheroids and organoids of the BBB, many drawbacks persist.

They usually lack the essential BBB cellular milieu, including microglia, six distinct cortical layers, and endothelial vasculature. Moreover, the limited formation of microglia and mature neurons limits its utility for specific in vitro ND models. Thus, the application of these simplified 3D models to understand how different neurotherapeutics in general and neuronanomedicines in particular are transported across the BBB and interact with phagocytic cells likely involved in their uptake and eventually their clearance remains a significant scientific challenge. [5,12d] In this scenario, the development of more robust, predictive, and costeffective in vitro models that recapitulate better the complex cell-cell interactions in the BBB and the NVU function is called for.

In this work, we investigate a novel multicellular organoid comprised of human brain microvascular endothelial cells (hBMECs), human brain vascular pericytes (hBVPs) and human astrocytes (hAs) combined with primary neurons and microglia isolated from neonate rats that was produced by spontaneous cell self-assembly and without the use of an extracellular matrix (ECM) such as Matrigel ® . The structure of the organoids and the presence of cell-cell interactions is characterized through the integration of confocal laser scanning fluorescence microscopy (CLSFM), light sheet fluorescence microscopy (LSFM), scanning transmission electron microscopy (STEM) and cryogenic-scanning electron microscopy (cryo-SEM). [20] In addition, the up-regulation of selected key BBB-related genes that code for tight junction proteins, solute carriers (SLCs), transporters of the ATP-binding cassette superfamily (ABCs), metabolic enzymes and proteins of the basal membrane were screened by western blotting, RNA-sequencing (RNA-Seq) and gene expression by quantitative polymerase chain reaction (qPCR) analysis. Finally, these multicellular organoids were used to characterize the permeability of different model polymeric, metallic, and ceramic NPs. Overall results support the promise of this simple and scalable multicellular NV-BBB organoid platform to investigate the interaction of nanomaterials with the BBB in neuronanomedicine and nanotoxicology.

Multicellular neurovascular organoids are one of the most promising surrogate in vitro models in translational neuronanomedicine, overcoming some of the shortcomings of monocellular 2D and 3D models. [21] However, they do not incorporate microglia cells, which mediate immune responses in the CNS by acting as macrophages and clearing cellular debris, dead neurons, and up-taking foreign particles. In addition, they usually require complex fabrication procedures.

In previous studies, we used BBB endothelial and olfactory neuroepithelial cells isolated from adult and neonate rat to study the transport of different polymeric nanoparticles. [17, 22] The aim of the present work was to extend these investigations and to develop a platform of multicellular organoids as a tool to assess the interaction of different types of nanomaterials with the BBB. For this, we used a simple self-assembly method without ECM to bio-fabricate organoids that combine three human cell types, namely hCMEC/D3, hBVPs and hAs, and incorporated two primary rat cell types, neurons that form synapses and neuronal networks and microglia cells involved in the clearance of particulate matter ( Figure 1A) .

Before organoid fabrication, we characterized the five different neural-tissue cell types by immunocytochemical staining. hCMEC/D3 cells are derived from human temporal lobe endothelial microvessels and produce two characteristic proteins of adherens and tight junctions, VE-cadherin and CLDN5, respectively ( Figure 1B) . Primary hAs express the filament protein GFAP ( Figure 1C ) and hBVPs the NG2 proteoglycan ( Figure 1D ). Primary neurons ( Figure 1E ) and microglia (Figure 1F,G) from neurogenic and non-neurogenic regions of neonate rat brains express βIII-tubulin, which is a microtubule element almost exclusive of neurons, and Iba-1/AIF-1 and iNOS which are overexpressed in activated microglia. Primary neurons are also positive for MAP-2 (data not shown). In these experiments, microglia changed the morphology from ramified characteristic of quiescent cells to a more flattened macrophage-like one (Figure 1F,G) . 

In addition, we investigated the interaction of rat microglia and hCMEC/D3 in 2-cell spheroids. Qualitative analysis suggests that there are not detrimental interspecies interactions [23] such as bidirectional and permanent communications between them (Supplementary Figure S1) .

After the characterization of the individual cells, we used them to bio-fabricate multicellular organoids that resemble the complex brain structure and utilized them to assess the permeability of the nanoparticles. We hypothesized that the phenotype of hCMEC/D3 in multicellular organoids will mimic better their physiology and function in the BBB. To serve as a high throughput screening tool in neuronanomedicine, the bio-fabrication process needs to be simple, cheap, reproducible, robust, and eventually scalable.

We adapted a previously reported method in which all the cells are mixed and cultured on agarose gel or round bottom well-plates (see experimental section). After 2-3 days of incubation, spherical organoids formed, and they were fully characterized at day 5 ( Figure 2 ). 3D cultures were small spherical bodies with radially distributed cells, whereas monocultures in flat well-plates exhibited enlarged cell bodies with less and short processes. 

Upon production, organoids were immunostained to reveal the cellular architecture by CLSFM ( Figure   3A -H). hCMEC/D3 endothelial cells together with pericytes appeared to form a surface monolayer tightly encasing the rest of the cells in the spheroid and they expressed VE-cadherin ( Figure 3A ) and CLDN5 ( Figure 3B ) which are characteristic of adherens and tight junctions, respectively. In addition, organoids showed a strong immunoreactivity for GFAP ( Figure 3C ) that is characteristic of hAs, f-actin of key functional proteins and the AQP4 water channel at the astrocyte end-foot covering the whole organoid ( Figure 3D ) that is crucial for the function of the BBB, the NG2 proteoglycan of mature hBVPs ( Figure   3E ), βIII-tubulin and MAP-2 of primary rat neurons ( Figure 3F ,G, respectively), and Iba-1/AIF-1 of primary rat microglia ( Figure 3H) . Microglial processes at the neuro-glia junctions could potentially monitor and protect neuronal functions. Previous experiments demonstrated that CLSFM is not the most appropriate technique to monitor the inner cellular structure and the diffusion of fluorescently-labeled nanoparticles into the organoids because it only allows to scan a Z-stack depth of 100 µm. Imaging in deeper layers is time-consuming and not feasible.

The visualization of the whole multicellular organoid could be conducted more efficiently and in a short time by 3D tomography using LSFM; this method enables the detection of fluorescence signals and the imaging of the sample as deep as 1 mm, and thus of the organoid core. [24] LSFM evidenced that our organoids are a solid cellular structure ( Figure 3I ) where hCMEC/D3 endothelial cells cover almost completely and uniformly the surface, forming adherens junctions that are a fundamental structure to govern the permeability into the CNS (Figure 3J,K) . This observation was confirmed by CLSFM (Supplementary Figure S2 ). In addition, microglia cells express Iba-1/AIF-1 ( Figure 3K) , a microglia/macrophage-specific Ca 2+ -binding protein that participates in membrane ruffling and phagocytosis in activated microglia. [25] The staining of AQP4 confirmed the presence of abundant filamentous bundles characteristic of hAs in the spheroid core ( Figure 3L ). Perivascular regions are also essential to mimic the physiology and function of the BBB along with the vascular endothelial barrier and astrocytic end-feet. For instance, astrocytes restore their phenotype in a 3D culture. [26] In line with previous works, we confirmed that hAs cultured in multicellular organoids exhibit a ramified phenotype that resembles cortical astrocytic networks, as opposed to 2D cultures where this cell type exhibited enlarged cell bodies, with less and shorter processes ( Figure   1C ). This phenotype may contribute to regulate the permeability of the BBB, either independently or in concert with other neighbouring neuron-glia cells. [27] Furthermore, we performed Western blot analysis and confirmed unequivocally the expression of key proteins such as VE-cadherin (120 kD), β-III tubulin (55 kD), GFAP (52 kD) and Iba-1/AIF-1 (19 kD) (Supplementary Figure S3 ). To ensure that the level of total protein seeding in all the runs was similar, we stained the gels with Coomassie blue staining directly after running gel electrophoresis (data not shown).

To gain a deeper understanding on the organoid ultrastructure and fundamental cell-cell interactions, we Figure 5B ). In addition, in ultrastructure cross-sections, myelinated axons exhibited nearly circular profiles surrounded by a spirally wound multilamellar electrical insulator, which enables the electrical impulses between these biological wires to travel back and forth quickly ( Figure 5C ). Moreover, this analysis provided an experimental proof that microglia, a CNS macrophage, interacts with primary neurons and their synapses ( Figure 5D ). Figure S4 ).

One of the challenges in the production of multicellular NVU organoids is to achieve an endothelial cell phenotype that resembles the function in vivo because the BBB endothelium regulates the transport of soluble and particulate matter into the CNS. We anticipated that 3D co-culture with hBVPs and hAs would result in a more physiological endothelial cell phenotype.

To analyze whether our multicellular organoids exhibit physiological characteristics of the in vivo BBB and constitute a functional barrier or not, we evaluated and compared transcriptome expression by RNA-Seq at day 5. Owing to interspecies variabilities, [28] for these studies, we used 3-cell organoids comprised on only human cells, namely hCMEC/D3 endothelial cells, hAs and hBVPs, and compared them to 2D and 3D endothelial cell monocultures; endothelial cell monolayers are the most common in vitro model of BBB.

The quality of the extracted RNA was assessed by 1% agarose gel electrophoresis and the quantity and the purity (quality control) of the RNA samples by using the Qubit ® and TapeStation; all the samples showed Table S1 ).

To visualize transcriptomic differences between 3-human cell BBB organoids, endothelial cell ( The three groups showed close distance within samples. We assume that there is a different cell milieu and that in the 3-cell organoids, most transcripts stem from endothelial cells. Next, we confirmed the differentially expressed genes (DEGs) between the three different groups. We set the threshold to padj <0.05 Table S2 ). Due to the relevance of tight junction and gap junction proteins, ECM proteins, SLC influx transporters, ABC efflux transporters and metabolic enzymes to the barrier function of the BBB endothelium, a more detailed comparison and discussion of the expression of genes coding for these proteins in the three models is included below.

The expression of VE-cadherin and CLDN5 in endothelial cell 2D monocultures and 5-cell organoids was initially demonstrated by immunocytochemistry (Figures 1,3) . At this stage, we looked into the expression of genes of the CLDN family and ESAM that codes for the endothelial cell specific adhesion molecule (ESAM), a transmembrane junction protein with a similar structure to junctional adhesion molecules. [29] 3-Cell organoids showed the overexpression of CLDN5 which is by far the predominant CLDN in endothelium and that codes for the integral membrane tight junction protein CLDN5 and is a gatekeeper of neurological functions (Figure 6A,F) . [30] The expression of this gene was maximum in 3D endothelial cell monocultures. However, the number of endothelial cells in 3-cell organoids is smaller than that in endothelial cell monocultures. CLDN1 and CLDN12 were also expressed, though to a lower extent than in endothelial cell 2D monocultures (Figure 6A , F); CLDN12 is not required for BBB tight junction function.

In endothelial cell 2D monocultures, the expression of CLDN genes was in general lower than in both 3D systems except for CLDN1, CLDN11 and CLDN12 (Figure 6 A, F) . In all the systems, the expression of

CLDN17, CLDN18 and CLDN20 transcripts was relatively low (Figure 6A,F) ; these genes are more specific of epithelial (and not endothelial) tight junctions. [30, 31] A relatively high level of CLDN15 could be observed in the endothelial cell monocultures, while CLDN19 expression was detected in the 2D endothelial cell model, but not in 3D organoids (Figure 6A,B) . ESAM was also expressed at a lower level in 3-cell organoids than in endothelial cell 2D monocultures (Figure 6A ,F) because endothelial cells of mesoderm origin selectively encode the immunoglobulin family adhesion molecule ESAM, which mediates cell-cell adhesion through homophilic interactions. [32] Other genes upregulated in 3-cell organoids with respect to endothelial cell 2D and 3D monocultures are GJA1 that codes for the gap junction alpha-1 protein (GJA1) also known as connexin-43 (Supplementary Table S3 ); [33] connexin hemichannels and gap junctions contribute to maintain the physiology of the BBB, participate in paracrine communication, and mediate efficient and rapid bidirectional inter-cellular transmission of electrical and chemical signals. Similarly, we found the upregulation of VCAM1 that codes for the vascular cell adhesion molecule-1 (VCAM-1) protein, which mediates endothelial cell adhesion and VWF that codes for the von Willebrand factor (VWF), a glycoprotein that might be involved in brain homeostasis (Supplementary Table S3 ). [34] 

The ECM consists of multimeric proteins and proteoglycans that participate in cellular migration, differentiation, and function as a support system for endothelial cells and astrocytes. Figure S6) . In this context, we analyzed genes coding for key BM proteins.

Collagen type IV is the most abundant component of the BM. The -chain of this protein consists of three domains and it is thought that six -chains self-assemble into triple-helical molecules and form spider weblike scaffolds that interact with laminin. [36] COL4A1 and COL4A4 that code for collagen IV α-chains are upregulated in our 3-human cell organoids with respect to endothelial cell monocultures (Figure 6B,G) ; COL4A1 is a highly conserved protein across species and is involved in angiogenesis.

Laminin is a T-/cruciform-shaped trimeric protein composed of α, β and γ chains. Brain endothelial cells, pericytes and astrocytes produce different isoforms of laminin at the BBB. LAMA1, LAMA2 and LAMA4

(coding for laminin α1, α2 and α4, respectively), which regulate the maturation and function of the BBB, and LAMB2 and LAMC3 (coding for β and γ chains of laminin) were also upregulated in the multicellular organoids ( Figure 6B,G) . Other genes showing higher in 3-human cell organoids than in endothelial cell monocultures are HSPG2 (coding for perlecan, the core protein of the glycosaminoglycan heparin sulfate) and NID1 and NID2 (coding for nidogen-1 and 2, respectively) that serve as linker for collagen IV, laminin and other ECM proteins (Figure 6B,G) .

ABC transporters are highly expressed by the BBB endothelium and they play a key role in maintaining the brain homeostasis because they actively govern the entry of compounds from the bloodstream into the CNS.

[37] As expected, 3-cell organoids expressed a moderately higher amount of genes coding for Pglycoprotein (ABCB1), and several multidrug resistance proteins (MRPs) from the C subfamily (ABCC3, ABCC4, ABCC5, ABCC10, and ABCC11) (Figure 6C,H) ; it is noteworthy that in this system only ~1/3 of the cellular component in the multicellular spheroids is endothelial cells. MRP3 is a glycoprotein with a similar molecular mass as MRP2, with similar amino acid composition, and with overlapping substrate specificity. Human MRP3 is the only basolateral efflux pump shown to transport bilirubin glucuronides. In some cases such as MRP2 (ABCC2) deficiency, MRP3 (ABCC3) is strongly upregulated (Figure 6C,H) .

MRPs 1, 2, 7, 8 genes were not substantially expressed in any of the models (Figure 6C,H) .

The alanine, serine, and cysteine transporters belong to the SLC1A family of excitatory amino acid transporters (EAAT), Na + -dependent proteins that reside in the membrane of astrocytes, neurons and the abluminal (brain-facing) membrane of the BBB. [38] EAATs are involved in the efflux transport of glutamate across the BBB and ensure low levels of this neurotransmitter in the interstitial fluid of the brain.

Genes coding for EAAT1 (SLC1A3, GLAST), EAAT2 (SLC1A2, GLT1), EAAT3 (SLC1A1, EAAC1) and EAAT5 (SLC1A7) are upregulated in 3-cell organoids with respect to 2D and 3D endothelial cell monocultures ( Figure 6D ,I), most probably due to the contribution of hAs to the total expression. These results were in good agreement with previous works that showed their expression in endothelial cells isolated from brain capillaries. [39] 

Different influx transporters are involved in the transport of essential endogenous nutrients (e.g., amino acids, glucose) from the bloodstream into the CNS and are critical for the normal function of the brain. [40] Endothelial cell 3D monocultures expressed relatively high levels of SLC2A3 (GLUT3) encoding for glucose transporter-3 (GLUT3) (Figure 6D,I) , a pump that is more characteristic of all neurons, [41] and did not express SLC2A5 (GLUT5) that codes for the transporter GLUT5 (characteristic of enterocytes). [42] Our 3-cell organoids expressed high levels of genes coding for different glucose transporters (Figure 6D,I) . Of special interest is SLC2A1 (GLUT1) that codes GLUT1 which is crucial for the development of the cerebral microvasculature with BBB properties in vivo. [43] Glucose is the predominant energy source for the brain and heart; therefore brain is the most energy-demanding organ in which endothelium and astrocytes plays a major role in regulating their metabolism. [44] The transport of glucose across the BBB into the brain is mediated by the facilitative glucose transporter GLUT-1. This gene was upregulated in the 3-cell organoids with respect to endothelial cell 3D and 2D monocultures, which constitutes another confirmation of the more physiological phenotype of the endothelium in our multicellular model.

Other genes that were highly expressed in 3-cell organoids when compared to 2D and 3D endothelial cell monocultures are SLC16A2 (MCT2) and SLC16A6 (MCT6) that code for proton-coupled monocarboxylic acid transporters (MCTs) and SLC6A6 coding for Na + -and Cl --dependent taurine transporter (TauT) (Figure 6D,J) . TauT plays a key role in many biological pathways such as neurotransmission. SLC6A1, the gene coding for the voltage-dependent gamma-aminobutyric acid (GABA) transporter SLC6A1 was also upregulated in the 3-cell organoids. This transporter is responsible for the re-uptake of GABA from the synapse; GABA counterbalances neuronal excitation in the brain and any disruption of this balance may result in seizures. A similar trend was observed for ABCA2 and ABCA8, coding for ABCA2, an endolysosomal protein that plays an important role in the homeostasis of various lipids and Alzheimer's disease, and ABCA8 that regulates the lipid metabolism and is implicated in various CNS pathologies ( Figure   6E ,J). [45] SLC27A1 coding for the long-chain fatty acid transport protein 1 (FATP1) was also overexpressed in the 3-cell models (Figure 6D ,I).

Metabolizing enzymes in the BBB have a functional role in the local metabolism of drugs and other xenobiotics. [46] Thus, the overexpression of genes coding for them is an additional proof of the more physiological behaviour of an in vitro cellular model. 3-Cell organoids expressed high levels of cytochrome P450 genes including CYP2D6 and CYP2R1 that were low in 2D and 3D endothelial cell monocultures ( Figure 6E,J) . Conversely, the expression of genes such as GSTP1, SULT1A, and UGT1A1 coding for phase-II metabolic enzymes glutathione S-transferase π, sulfotransferase 1A1 and UDPglucuronosyltransferase, respectively, was low in all the specimens (Figure 6E,J) . In addition, SOD1, a gene coding for the apoptotic enzyme superoxide dismutase 1 (SOD1), and GTPBP10 for a mitochondrial protein were downregulated in 3-cell organoids, whereas upregulated in endothelial cell 2D cultures ( Figure 6E,J) . These results indicate that our organoids do not exhibit hypoxic conditions.

Overall, the comprehensive RNA-Seq analysis of our bio-fabricated 3-cell organoids confirmed high relative expression of endothelial cell specific genes involved in key signalling pathways that contribute to the establishment of a functional BBB.

Our previous investigations conducted with primary rat cells suggested that neither forebrain nor olfactory neurons internalize polymeric nanoparticles. [17] Conversely, they were internalized by primary microglia.

These studies were conducted in 2D monocultures. Cell-cell connections between different components of the CNS, including neurons and microglia, in 3D may contribute to generate a more physiological milieu, to the increase the integrity and the function of the BBB endothelium and reciprocally affect the phenotype of the other cells actively implicated in the interaction with particulate matter. [47] The transport of nanomaterials across the BBB endothelium is usually by transcytosis and is initiated by endocytosis, for which the size should be ≤200 nm.

[48] Depending on the shape and surface properties, particles as large at 500 nm could be transported to a more limited extent.

A main limitation of the existing in vitro models is that in general they do not include CNS macrophages.

Upon characterization of our multicellular organoids by immunocytochemistry, electron microscopy and RNA-Seq, and the confirmation that the recapitulate a more physiological behavior, we studied the ability of different polymeric, ceramic and metallic nanoparticles to cross the outer endothelial cell monolayer and reach the organoid bulk. The properties of the NPs used in this work are summarized in Supplementary  Table S4 . Before use, NPs were diluted under sterile conditions and mixed with the corresponding culture medium to the final desired concentration.

In this work, we used four polymeric NPs produced by the self-assembly of chitosan (CS)-, [49] poly(vinyl alcohol) (PVA)- [50] and hydrolyzed galactomannan (hGM)-based graft copolymers [51] synthesized by the hydrophobization of the polymer backbone with poly(methyl methacrylate) (PMMA) and displaying size between 92 ± 4 to 463 ± 73 nm and from positive to negative Z-potential (Supplementary Table S4 ); these two properties govern the interaction of nanoparticulate matter with cells [52] and were measured immediately before the biological experiments.

We hypothesize that owing to the cellular heterogeneity of the multicellular organoids, some immunocompetent cells (e.g., microglia) could be more susceptible to damage or, conversely, to uptake the NPs to a greater extent than others (e.g., neurons), as we previously demonstrated in 2D culture (31).

Differential cell compatibility and uptake by BBB cells has been extensively documented in the literature.

However, the effect of microglia in 3D multicellular systems has been never investigated before.

To address these questions, polymeric NPs were fluorescently labeled with FITC or RITC and their interaction (e.g., permeability) with 5-cell organoids after 24 h of exposure characterized by CLSFM and LSFM. In general, studies revealed that 0.1% w/v NPs do not cause any morphological damage in the organoids.

When 5-cell organoids were exposed to crosslinked mixed CS-PMMA30:PVA-PMMA17 NPs, most of them accumulated on the organoid surface and a smaller fraction could be found inside it, as shown in Figure 7A ,B by 2D and 2.5D CLSFM. However, cross-sectional CLSFM images cannot provide complete multi-view volumetric information of 3D organoids for which we need to detect the fluorescence intensity of each individual voxel. Cell uptake specificity was investigated by advanced LSFM. Images taken from different angles confirmed that, as opposed to CLSFM, some NPs permeate into the organoids and suggested the possible involvement of astroglia or microglia in the transport (Figure 7C,D) . In case of mild injury/disturbance, astrocytes become phagocytes which remove "foreign" material and produce antiinflammatory cytokines. Conversely, under excessive injury/insult, "reactive" astrocytes produce proinflammatory cytokines that recruit and activate microglia. [53] Both pathways could be involved in the uptake of the nanoparticles into the organoids bulk. Similar results were observed with fluorescently labeled CS-PMMA33 (Figure 7E-H) , crosslinked PVA-PMMA17 (Figure 7I-L) , and hGM-PMMA28 NPs (Figure 7M-P) . Furthermore, representation of the cells as dots (Figure 7D ,H,L,P) confirmed that these nanoparticles are not harmful to the cells and that the cell density was not majorly affected by exposure to the different NPs.

Both fluorescence microscopic analyses confirmed that the enhanced fluorescence of NP-exposed organoids stemmed from NPs that most probably accumulated within endosomes/lysosome compartment of the primary microglia or immunocompetent astrocytes. Permeability studies conducted with our 5-cells organoids and metallic and carbon nanoparticles by STEM further contributed to our understanding of the possible uptake pathways (see below).

The brain is rich in energy-demanding nerve cells that metabolize glucose as the main fuel. [54] Neurons consume glucose through glia cells, in which this nutrient is metabolized into lactate by the glycolytic pathway and transferred to axons and neuronal bodies when needed. To this end and to support their sentinel activity in the CNS, primary microglia overexpress GLUT1. [55] In a previous work, we demonstrated that the accumulation of hGM-PMMA28 nanoparticles in pediatric sarcomas correlates well with the overexpression of GLUT1. [51] Our LSFM results show that these inherently sugared nanoparticles are actively transported into the organoids, most probably by activated microglia and astrocytes. hGM-PMMA28 and other carbohydrate-based nanoparticles investigated in this work such as crosslinked mixed CS-PMMA30:PVA-PMMA17 and CS-PMMA33 could be also taken up through the mannose receptor that is expressed in microglia and astrocytes and that displays a carbohydrate recognition domain. Boric-acid crosslinked PVA-PMMA17 nanoparticles exhibit a boronated surface that may form complexes with microglia by toll-like receptor 4 (TLR4)-myeloid differentiation protein-2 (MD-2) signalling. [56] In addition, microglia expresses sialic acid which may also bind boron. [57] These findings are in good agreement with the active surveillance and homeostasis roles of microglia in the CNS microenvironment.

Together with the gene expression pattern observed for endothelial cells cultured in 3-cell organoids with respect to 2D and 3D monocultures, this more complex multicellular in vitro model would recapitulate better the physiology of the BBB and serve as a platform to assess the interaction of nanomedicines and nano-pollutants with the CNS. 

The use of metallic nanoparticles in nanomedicine is broad and varied. [58] For example, Au and Ag NPs have been proposed in anti-cancer therapy [59] and, upon injection, they could cross the BBB from the systemic circulation, and reach the CNS. [60] Several studies used Au NPs as shuttles and demonstrated that the smaller the size, the higher the permeability; e.g., 10-20 nm Au NPs resulted in the highest cellular distribution in the brain of mouse. [61] A main drawback of Au NPs is the limited ability of the CNS to clear them and the potential neurotoxicity associated with their accumulation.

To assess the performance of our new multicellular model, we exposed 5-cell organoids (5 days old) to ultra-small Au NPs (10 ± 2 nm, concentration of 1 x 10 6 NPs/mL, Supplementary Table S4) for 24 h and analyzed their possible endocytosis by STEM. After 24 h, Au NPs were readily taken up and distributed in the cytosol, and inside endosomes and lysosomes of primary microglia that were identified by the presence of multiple lipid droplets in the cytosol (Figure 8A,B) . The uptake mechanism is most probably energydependent, as shown elsewhere. [62] We reported the ultrastructural morphology of microglia phenotype and high accumulation of lipid bodies is linked to a surveillant state in which microglia actively monitors the surrounding environment.

Ag NPs have been also used in nanotherapeutics. Ag NPs (60 ± 13 nm, concentration of 1 x 10 6 NPs/mL, Supplementary Table S4) were produced by a chemical method and organoids exposed to them for the same time. STEM studies focused on intracellular vesicles and revealed the presence of only very few Ag NPs inside these organoids (Figure 8C-E) . This result might stem from the more limited cellular penetration of these NPs when compared to the much smaller Au counterparts. Another possible mechanism is that they undergo fast dissolution outside and inside the cells. These findings were consistent along different experiments. 

Among carbon nanomaterials, graphene-based ones are the most popular in the area of nanoneuroscience [63] due to their various applications in neuronanotechnology [63, 64] and nanosafety studies. [65] Here, we investigated the effect of graphene nanoplates (5 µm diameter and 10 nm thickness; 25 μg/mL, Supplementary Table S4 ) on our organoids. After 24 h, significant cell death could be observed, most

probably due to direct cell damage (Figure 8F) . Similarly, alkaline carbon dots (10 nm; 25 μg/mL, Supplementary Table S4 ) caused complete destruction of cellular structures (Figure 8G) . In recent years, carbon dots have become a rising environmental concern because of their extensive use in energy applications, massive emissions to the air and likely transport from the nasal mucosa into the CNS.

Our results strongly suggest the possible toxicity of these carbon NPs and the need for more systematic investigations that screen their detrimental effect on the CNS.

The interplay between other CNS-tissue-type cells, microglia and different nanomaterials by molecular level analysis which can provide sufficient sensibility to analyze the functional/phenogenotypic response of this brain's own-macrophage cell population, [66] was beyond the scope of the present work. In future research, this model will be characterized in other molecular level and individual cell specific aspects.

Several approaches have been utilized to develop preclinically and clinically relevant in vitro BBB models where human endothelial cells wrap up onto the surface of neural cell organoids. However, these models do not comprise a fundamental cellular player involved in the transport and elimination of particulate matter, microglia, and thus, their relevance in the study of the interaction of the CNS and nanomedicine and nano-pollutants is questioned. Previous models made of induced pluripotent stem cells showed that these cells cannot undergo differentiation into a complete set of neurovascular cells simultaneously and other cell types need to be added to create the complete brain environment.

To overcome the limitations in the field, in this work, we developed a multicellular organoid that combined human endothelial cells, pericytes and astrocytes with primary neurons and microglia isolated from rat brain in a ratio that mimics the cell composition of the brain tissue. The bio-fabricated organoids formed a layer of endothelial cells on the surface that formed adherens and tight junctions, two structures that govern paracellular permeability of molecules and particles at the BBB and thus, their bioavailability in the CNS.

They also express active efflux transporters which are associated with the low bioavailability of neurotherapeutics in the brain. The organoids are stable in culture for at least two weeks and after comprehensive characterization, they were used to study the toxicity and permeability of different types of

NPs. All the model polymeric NPs show good compatibility and undergo uptake. Similar results were observed with metallic nanoparticles, though the permeability depended on their size. In contrast, carbon NPs showed high toxicity, causing extensive cell death. The proposed multicellular organoid is easily biofabricated, and scalable to high-throughput capacity due to the simplicity of the method and the small number of reagents required. Furthermore, the screening throughput of this model can be increased even further through the possibility of integration with automated microscopy and robotics technologies. The ease of culture, cost-effectiveness and reproducibility of this model offer a very practical and attractive approach for researchers interested in studying BBB drug transport and developing brain-penetrant drugs for the treatment of CNS diseases. Taken together, our findings confirm that these multicellular systems form a functional BBB. Our model may serve as a valuable tool not only in nanotherapeutics, but in nanoneurosafety and nanoneurotoxicity.

The hCMEC/D3 cell line derived from human temporal lobe microvessels (hBMECs, EMD Millipore, Burlington, MA, USA) was maintained in EndoGRO Basal Medium (EMD Millipore) with supplements containing 5% fetal bovine serum (FBS), L-glutamine, vitamin C, heparin sulfate and recombinant human epidermal growth factor (rhEGF), all purchased from Sigma-Aldrich (St. Louis, MO, USA). [67] hAs (ScienCell Research Laboratories, Carlsbad, CA, USA) were grown in astrocyte growth medium (ScienCell Research Laboratories) containing 2% FBS supplemented with astrocyte growth factors, penicillinstreptomycin. [68] hBVPs (ScienCell Research Laboratories) were maintained in pericyte culture medium (ScienCell Research Laboratories) containing 2% FBS, pericyte growth supplement and penicillinstreptomycin. [69] All cells were incubated at 37°C in humidified 5% CO2/95% air.

The use of neonate Sprague-Dawley (SD) rats (P0-1) and the protocols utilized for the isolation of the primary CNS cells were conducted with the approval of the Animal Care Committee of the Technion-Israel

Institute of Technology. All the experimental procedures were in accordance with the guidelines set by the EU Council Directive (86/609 EEC) and according to the official protocol #IL-141-10-17 (expiry date 19

December 2021).

Primary neural, and neural stem/progenitor cells [17, 70] were isolated from P0-1 SD rats as described below.

Briefly, rats were sacrificed by decapitation, and whole brains were quickly removed aseptically prepared according to Saura et al. [71] Briefly, P0-1 SD rat brains were removed and rinsed in phosphate buffer saline (PBS, Sigma-Aldrich). After careful removal of the meninges and brains cortices, they were mechanically dissociated and trypsinized for 20 min. Cells were cultured in 6-well plates with DMEM containing 10% FBS at 37°C in humidified 5% CO2/95% air. The medium was exchanged twice weekly.

Microglia cells were isolated from mixed glia by mild trypsinization (30 min-2 h) with trypsin-EDTA 0.25% diluted 1:3 in serum-free DMEM, at 37°C. After detachment of astrocyte brown sheets, the firmly attached macrophages were further propagated in DMEM:F12 1:1 with 10% FBS and the cells replated in 24-well plates containing PLL-coated glass coverslips at a density of 50,000 cells/well (30) . For microglia stainings, cells were fixed and the immunostaining and imaging were performed according to the protocol described below.

Three-cell organoids comprising hCMEC/D3, hAs and hBVPs were cultured in EndoGRO Basal Medium supplemented with 5% FBS, L-glutamine, vitamin C, heparin sulfate and rhEGF, at 37°C in humidified 5% CO2/95% air using the liquid overlay culture system. We used the same method to produce 5-cell organoids incorporating primary neurons and microglia isolated from neonate rat (see above). This method is our adaptation of the aggregate cultures previously described. [12a,12b,72] 

Organoids and monocultures were collected at day 5, pooled into a 0. 

At the end of day 5, total RNA was extracted from BBB organoids using a T-series Ultraclear 1. as Phred values (10 log10P base call is wrong), i.e. values higher than 30 indicate a probability of less than 10 -3 of an incorrect base call (Figures 1,2 of Supplementary experimental file). Only unique mapped reads were counted to genes, using 'HTSeq-count' package version 0.6.1 with 'union' mode. Normalization and differential expression analyses were conducted using DESeq2 R package version 1.10.0. The reads were mapped to the Human genome (ftp://ftp.ensembl.org/pub/release-96/fasta/homo_sapiens/) using Tophat2 version 2.1.0 (65), with up to 2 mismatches allowed per read, the minimum and maximum intron sizes were set to 70 and 50,000, respectively, and an annotation file was provided to the mapper (Figure 3 of Supplementary experimental file). 

In this work, we utilized different polymeric, metallic and carbon nanoparticles to assess their permeability in 5-cell organoids.

Different polymeric NPs were obtained by the self-assembly of amphiphilic graft copolymers of CS, PVA and hGM with PMMA. The copolymers were produced by the free radical graft polymerization of methyl methacrylate initiated by cerium(IV) ammonium nitrated in acid aqueous medium, as reported elsewhere. [49b, 51] 

Au NPs (diameter of 10 nm, 741957; Sigma-Aldrich) and Ag NPs (diameter of 60 nm) synthesized from silver nitrate by reduction with sodium borohydride and stabilization with sodium citrate (all supplied by Sigma-Aldrich) [75] were utilized as prototypes of metallic NPs.

We used graphene nanoplatelets (diameter of ~5 µm and thickness of ~100 nm, GNN P0205, Ants Ceramics Pvt. Ltd., Vasai-Virar, India) and alkaline carbon dots produced by the reaction of acetone with sodium hydroxide, as reported by Hou et al. [76] The size (hydrodynamic diameter, Dh) and size distribution (polydispersity, PDI) of the different nanoparticles before the biological studies were measured by dynamic light scattering (DLS, Zetasizer Nano-ZS, Malvern Instruments, Malvern, UK) in 10 mm quartz cuvettes using a He-Ne laser (673 nm) as light source at a scattering angle of 173°, at 25°C. Data were analyzed using CONTIN algorithms (Malvern Instruments). The surface charge was estimated by measuring the zeta-potential (Z-potential) by laser Doppler micro-electrophoresis in the Zetasizer Nano-ZS. Each value is expressed as mean ± S.D. of at least three independent samples, while each DLS or Z-potential measurement is an average of at least seven runs.

For biological studies, copolymers were fluorescently-labeled with FITC or RITC, both from Sigma- 51] . The properties of the NPs used in this work are summarized in Supplementary Table S1 .

Before use, NPs were diluted under sterile conditions and mixed with the corresponding culture medium to the final desired concentration.

RNA-Seq data to compare between organoid groups was obtained from three biological replicates. Wald test parameters were used for pairwise comparisons. The statistical analysis was conducted and the log2 fold change and the adjusted p-values (padj) were indicated for significantly upregulated and downregulated genes; padj <0.05 was considered statistically significant. The differential expression analysis was conducted using 'DESeq2' R software. The Zen software was used to evaluate CLSFM and LSFM images. In some cases, the TIFF images acquired with IMARIS software were imported for subsequent measurement and 

38, 1667; b)

Fluids Barriers CNS 2020

Drug Efflux Pumps in Cancer Resistance Pathways: From Molecular Recognition and Characterization to Possible Inhibition Strategies in Chemotherapy

10, 478; b) I. Schlachet, A. Sosnik

2017, 12, 1533; b)

Fluids and Barriers CNS 2013

In Vitro 1972, 8, 26; b)