NEW EXPERIMENTAL
BLOOD-BRAIN BARRIER MODELS FOR ALZHEIMER'S DISEASE STUDIES
Jonna Niskanen
Master of Science thesis Master´s Degree Programme in Biomedicine
University of Eastern Finland Faculty of Health Sciences School of Medicine
3.6.2019
University of Eastern Finland, Faculty of Health Sciences, School of Medicine Master´s Degree Programme in Biomedicine
Jonna Niskanen: New experimental models for blood-brain barrier for studies of Alzheimer's disease
Master of Science thesis; 38 pages, supplement; 4 pages Supervisors: Lehtonen Šarka 1,2 & Sonninen Tuuli-Maria 1,2
1University of Eastern Finland
2A.I. Virtanen Institute for Molecular Sciences 3.6.2019
Keywords: Alzheimer’s disease, blood-brain barrier, hiPSC, co-culture, astrocytes, endothelial cells, pericytes
Abstract
Blood-brain barrier (BBB) protects the brain by creating a physical barrier and regulating the transport of molecules across it. A well-functioning BBB is essential for maintaining healthy brain tissue and BBB breakdown or dysfunction is associated with a variety of neurological diseases, including Alzheimer’s disease (AD). To further study BBB in AD, several experimental models were generated using human induced pluripotent stem cells (hiPSCs) derived from AD patients and healthy subjects. As BBB is mainly formed by endothelial cells (ECs), astrocytes, and pericytes, these cell types were used for the models. Throughout our project, we discovered that the cell count affects the differentiation efficacy and utilised this information to derive several cell types simultaneously. The differentiated cells were characterized by morphology and based on their gene (qPCR) and protein expressions (ICC).
The experimental models were formed combining these cells into different co-cultures.
Astrocytes affected the gene expression of ECs in the non-contact model by increasing tight junction protein expression in the control while decreasing it in the AD model. However, in the contact model with EC-astrocyte co-cultures, the ECs lost their barrier forming abilities.
The spheroids were successfully formed directly in the media, but they self-assembled in reverse order compared to studies previously described in the literature. Therefore, we suggest some changes to the implemented protocols and conclude, that with few modifications, these hiPSCs derived BBB models could be utilized in discovering AD pathology in the future.
1
Introduction
The blood-brain barrier (BBB) protects the brain. It forms a physical barrier that regulates the flow of substances to and from the brain. These substances can be essential such as nutrition and oxygen, or unwanted ones like pathogens and neurotoxins (Winkler et al, 2014). EC, pericytes, and astrocytes are the three main cell types forming the BBB (Fig. 1), and together with neurons and microglia they form a neurovascular unit (Erickson & Banks, 2013). ECs, pericytes and astrocytes were used to model the BBB in this study (Fig. 1).
Figure 1 A model of the blood-brain barrier. Between the brain and the bloodstream, a blood-brain barrier is formed from interactions of endothelial cells (purple), pericytes (orange), and astrocytes (green).
ECs cover the inner lumen of blood vessels. They create a physical barrier to shield the brain by forming different paracellular junctions such as tight junctions and adherens or intermediate junctions (Stamatovic et al, 2008). Tight junctions constitute the vascular seal when intermediate junctions initiate and maintain EC cell-cell contact. Tight junctions enable a high trans-endothelial electrical resistance (TEER) that is specific for brain ECs (Sweeney et al, 2018). TEER can be used to visualize the ability of a cellular layer to conduct electrical current – better the barrier formation, higher the TEER value (unit: Ω cm2). Several different junction proteins work together in the BBB formation, proteins such as occludin (OCLN), claudin-5 (CLD5), tight junction protein 1 or zonula adherens 1 (TJP1 or ZO1), vascular endothelium
2 cadherin (CDH5), and junctional adhesion molecules (JAMs) just to name a few (Stamatovic et al, 2008). OCLN regulates paracellular barrier permeability along with calcium-dependent cell-cell adhesion protein CDH5 (Gavard 2014). CDH5 functions in the intermediate junctions below the tight junctions and regulate endothelial intermediate junction assembly and maintenance. ZO1 stabilizes tight junctions and functions in signal transduction. It coordinates the binding of transmembrane proteins, cytosolic proteins and filamentous actin to the basement membrane (Klock et al, 2015). CLD5 is the component of tight junction strands. It helps to tighten the junctions and obliterate any intercellular space (Stamatovic et al, 2008).
Previous studies have shown, that especially CLD5 is essential for fully functioning BBB; mice lacking CLD5 died at birth (Sohet & Daneman 2013).
As the microvascular ECs form an efficient barrier, the BBB needs to enable influx and efflux across the barrier. Influx refers to the transport of useful substances, like nutrients, into the brain. Efflux is the transportation of harmful substances, such as cellular metabolism waste products, out of the brain. For the transportation, BBB utilizes specific mechanisms including passive diffusion, receptor-mediated transcytosis, and carrier-mediated transcytosis (Agrawal et al, 2017). The ECs express multiple transporters, for example glucose transporters (GLUT1), efflux transporters, like p-glycoprotein (ABCB1) and low-density lipoprotein receptor-related protein 1 (LRP1), and influx transporters, such as receptor of advanced glycation end products receptor (RAGE). ABCB1 is an ATP binding cassette transporter. This efflux transporter carries various xenobiotics (with an exceptionally vast size range, ranging from 300 to 4000 Da) across the BBB (Miller et al, 2008). LRP1 protects the vasculature by transporting serum lipoproteins from the bloodstream and it also plays a role in regulating angiogenesis (Lillis et al, 2008). RAGE is a multiligand transporter that has an important role in regulating cell signaling (Rouhiainen et al, 2013).
Pericytes are in direct contact with ECs, located in intervals along the walls of capillaries (Klock et al 2015; Erickson & Banks, 2013). In the human BBB the ratio between these cells varies from 1:1 to 1:3, in peripheral capillaries far fewer pericytes are found. The higher pericyte density correlates with tighter barrier formation. Pericytes are important for angiogenesis and maintenance of the BBB. They have a role in controlling the brain blood flow and regulating the immune cell entry to the central nervous system (CNS) (Attwell et al, 2016;
Erickson & Banks, 2013). Characterization of these cells can be a little tricky. First, there are several different subclasses of pericytes even in the BBB (Attwell et al, 2016). Second, only a
3 few pericyte markers have been in use and they are rarely exclusive to pericytes (Smyth et al, 2018). The pericyte markers in use include proteins that are involved in angiogenesis and brain microcirculation; platelet-derived growth factor receptor beta (PDGFRβ), α-smooth muscle actin (α-SMA), and caldesmon (CAL).
PDGFRβ is a tyrosine kinase receptor that takes part in blood vessel development by regulating cell proliferation, growth and differentiation (Kloc et al, 2015). It recruits pericytes and smooth muscle cells to ECs during angiogenesis. PDGFRβ is modulated by LRP1. LRP1 reduces the PDGFRβ activation, completes the angiogenic process and thus supports vascular integrity (Lillis et al, 2008). The regulation of PDGFRβ is crucial. Prior studies with LRP1 knock-out mice have demonstrated significant increase of abnormal PDGFRβ activation, leading into formation of aneurysms and atherosclerotic lesions. Capillary pericytes, especially in junctional blood vessels, express α-SMA. α-SMA is a key component in regulating brain microcirculation (Alarcon-Martinez et al, 2018). The pericytes increase blood flow to the brain by relaxation (Attwell et al, 2016, Erickson & Banks, 2013). The contractility of the pericytes is regulated by CAL. CAL is a calmodulin-binding actin-regulatory protein, that modulates the access of myosin to actin (Kim et al, 2008). Pericytes also affect the BBB permeability and vascular integrity. They express some metalloproteinases such as aminopeptidase N (ANPEP) and matrix metalloproteinase 9 (MMP9) that take part in weakening the adherens of the ECs and pericytes on the BBB, enabling the EC migration during angiogenesis (Kloc et al, 2015;
Zozulya et al, 2008; Winkler et al, 2014). In the later stages of angiogenesis, pericytes start to produce fibronectin that downregulates and/or inhibits these destabilizing matrix metalloproteinases and by doing so concludes the angiogenic process. Loss of pericytes can lead to vascular instability, dilatation and rupture, and cause BBB leakage (Winkler et al, 2014).
Astrocytes are largest and most prevalent type of glial cells in the CNS. They are active participants in the neuronal energy supply (Belanger et al, 2011) and play an essential role in neurovascular coupling. They connect the neurons to BBB by extending fine perisynaptic processes to neurons, and vascular processes to surround intracerebral blood vessels. They take part in functional hyperemia, where neuronal activity increases the cerebral blood flow, by causing both vasodilatation and vasoconstriction (Belanger et al, 2011). Pericytes play a significant role in facilitating these changes of the blood flow (Attwell et al, 2016, Erickson &
Banks, 2013). Astrocytes have multiple functions. They contribute to the formation and
4 preservation of BBB and participate in the maintenance of extracellular ionic and chemical homeostasis. Astrocytes respond to injury and inflammation, and they affect neuronal development and plasticity (Garwood et al, 2017). Perivascular astrocytes regulate the osmotic balance through aquaporin 4 (AQP4) -channels and provide energy to the neurons. They intake glucose through the BBB (GLUT1), metabolize it into lactate, and then feed it to the neurons (Belanger et al, 2011). GFAP, or glial fibrillary acid protein, is a structural protein found in the main branches of fibrillary astrocytes. According to prior in vivo studies conducted with rats, loss of GFAP-positive astrocytes causes a widespread loss of tight junction proteins CLD5, OCLN and ZO-1 (Garwood et al, 2017).
This study focuses on Alzheimer’s disease (AD). AD is a debilitating progressive neurodegenerative disease. Neurodegenerative diseases in general are a growing problem in the world. According to the World Alzheimer Report 2018, there were roughly 33 million people suffering from Alzheimer’s disease. The number is expected to almost double every 20 years. This means that by 2030 there will be over 70 million people suffering from dementia.
As one can imagine this also has a huge economic impact – in 2018 worldwide cost of dementia was 1 trillion US dollars (Patterson 2018). People are living longer than ever, especially in Scandinavia. In Finland 2017, dementia and Alzheimer’s disease was the third most significant cause of death with 9 390 deaths, which is 17.5 % from all the cases (Suomen virallinen tilasto 2019) and in England and Wales it already is the most significant cause of death (Patterson 2018). Pathways involving AD onset and progression are still unresolved. This devastating disease is affecting a growing population in the world but has no cure, thus further studies are required.
Pathologically, AD is characterized by loss of neurons and synapses. Disease progression results in gradual progressive memory impairment and cognitive decline. Molecular hallmarks of the disease include the formation of neurofibrillary tangles of tau-protein and clusters of aggregated amyloid-β-peptide (Aβ). BBB dysfunction is likely contributing to AD pathology already at early stage (Zenaro et al, 2017; Yamazaki & Kanekiyo, 2017). Despite having several different hypotheses on the root cause of AD, no consensus has been settled on. The most popular theory for the past few decades has been the amyloid cascade hypothesis.
According to this hypothesis, the pathological sequence of events leading to AD starts with the accumulation of the Aβ to the brain, followed by the deposition of neurofibrillary tangles of tau (Heppner et al, 2015). Main Aβ influx and efflux receptors, RAGE and LRP1, are
5 associated with Aβ accumulation in AD. These receptors are responsible for transporting the Aβ through BBB (Lyros et al, 2014). In AD pathology an increase in the Aβ influx (RAGE) and decrease in Aβ efflux (LRP1) has been depicted.
In contrast to the amyloid cascade hypothesis the two-hit-vascular hypothesis states that the pathological accumulation of Aβ is secondary to a primary vascular damage. According to this hypothesis, the reduced cerebral blood flow from microcirculation disturbances together with hypoxia and BBB dysfunction lead to AD progression (de la Torre 2018; Lyros et al, 2014).
BBB disruption is implicated in many chronic neurodegenerative diseases, AD included (Sweeney et al, 2018). Several in vivo (Joo et al, 2017) and clinical studies support this hypothesis (Erickson & Banks, 2013; van de Haar et al, 2016). Transgenic rat models of AD display early neurovascular dysfunction, with reduced vascular contractility and Aβ-accumulation (Joo et al, 2017). According to a clinical study, using dynamic contrast material–enhanced magnetic resonance imaging, patients with early AD displayed global BBB leakage that could be associated with cognitive decline (van de Haar et al, 2016). Cerebral hyperfusion, or leakage of BBB, is considered to be an indicator for early AD. Different neuroimaging applications can be applied to detect these microleakages and to predict preclinical dementia (de la Torre, 2018).
Preclinical and clinical studies have also established a link between inflammation and AD pathogenesis. The inflammatory hypothesis suggests, that the inflammatory response that is associated with AD pathology contribute to and exacerbate the disease (Heppner et al, 2015).
Traditional neuroinflammatory diseases e.g. multiple sclerosis, are mainly driven by leucocytes, B- and T-lymphocytes, when neuroinflammation in AD is primary driven by the the brain's intrinsic myeloid cells, microglia (Heppner et al, 2015). In previous studies with in vivo AD models the disease pathology and progress could be altered by manipulating some of the molecules of the innate immune system or their respective pathways. This indicates that targeting components of the innate immune system could potentially ameliorate the course of the disease. The microglia are innate immune cells residing in the brain. They can be activated by either external microbial stimulus or by endogenous ligands released by tissues under destruction (Mariathasan & Monack 2007). The activated microglia secrete e.g.
interleukin 1 Beta (IL-1β) and tumor necrosis factor alpha (TNF-α) (Wang et al, 2015). TNF-α has been linked to AD. Astrocytes and microglia release large amounts of TNF-α in pathological conditions. This phenomenon is known as neuroinflammatory response.
6 Neuroinflammation triggers and sustains neurodegenerative processes such as AD (Olmos &
Llandó 2014).
Chronic inflammation is closely linked with disturbances in glucose metabolism. This brings us to another hypothesis; AD as ‘type 3 diabetes’ (Kandimallaa et al, 2017). There is substantial epidemiological evidence suggesting that there is a link between type 2 diabetes (T2D) and cognitive impairment (Barbagallo & Dominguez, 2014). T2D and AD appear to share different pathological mechanisms. They interlink insulin resistance, inflammatory response, oxidative stress, glycogen synthase kinase 3β (GSK3β) signaling, Aβ formation, and tau formation (Kandimallaa et al, 2017). According to this hypothesis, the over-activity of GSK3 accounts for memory impairment, tau hyper-phosphorylation, increased Aβ production, degreased tight junction formation in the BBB ECs, and local plaque-associated microglial-mediated inflammatory responses (Ramirez et al, 2013; Hooper et al, 2008). The brain requires 20% of the total glucose energy supply (Garwood et al, 2017). Glucose transporters, such as GLUT1 and GLUT4, provide energy across the BBB to the neurons. GLUT1 is the most prominent insulin independent glucose transporter on the BBB. It is essential for feeding glucose to the brain and enabling the normal neurological function that is consuming vast amounts of energy (Patching, 2017). GLUT4 is an insulin dependent glucose transporter that can be found especially in the hypothalamic BBB where it appears to be involved in glucose homeostasis (Ngarmukos et al. 2001). Disturbances in these transporters cause energy deficit in the brain that can have devastating effects (Lyros et al, 214). Insulin is regulating glucose homeostasis, but it also has a broader pleiotropic nature. Insulin receptor (INSR) belongs to the tyrosine kinase superfamily. It regulates glucose synthesis, activates a complex signaling network through IRS proteins and the canonical PI3K & ERK cascades (De Meyts et al, 2000) and it also modulates the aforementioned GSK3β kinase (Ramirez et al, 2013) making it an important component in the glucose metabolism in the brain.
To further study the root cause of AD, human models of the BBB are required. As the access to human primary brain material is very limited, alternate tools to model BBB must be discovered. Non-human animal primary cells have been used previously, but this model is prone to interspecies variations, same as in vivo models (Cho et al, 2017). Humane primary cells could be harvested from the CNS of embryos, fresh cadavers or from brain biopsies, but these procedures have ethical problems and a limited source. The immortalized cell lines currently readily available, do not fully display normal cellular function. Almost all
7 immortalized EC-lines form barriers with a transendothelial electrical resistance (TEER) below 150 Ω cm2, when expected physiological TEER values exceed 1 500 Ω cm2.
Stem cells are self-renewable cells that can be differentiated into several different mature somatic cell types. They provide a virtually unlimited cell source for a vast variety of cells (Appelt-Menzel et al, 2017). Stem cells can be harvested from e.g. the bone marrow, but their differentiation potential is tied to the origin of the stem cells. Pluripotent stems cells on the other hand have the capacity to differentiate into any cell of a human body. They can differentiate into three layers of an embryo: ectoderm (skin and nervous system), endoderm (gastrointestinal and respiratory tracts, endocrine glands, liver, and pancreas) or mesoderm (bone, cartilage, most of the circulatory system, muscles, and connective tissue). Both ECs and pericytes can be derived from the mesoderm. The human induced pluripotent stem cells (hiPSCs) can be reprogrammed from e.g. skin sample fibroblasts with Sendai-virus using Yamanaka factors (Lehtonen et al, 2018). Using hiPSCs from patients with genetic diseases open the possibilities to discover differences between the characteristics of diseased and healthy cells.
By combining ECs, pericytes, and astrocytes, multicellular interactions of BBB can be observed in vitro. Co-culture creates a more realistic microenvironment of in vivo metabolism than the traditional monocultures. These systems enable to study the interactions between different cell populations. Cells communicate by juxtacrine or contact signaling and by paracrine or non-contact signaling, where the signaling molecules are secreted from cells (Kook et al, 2017). Co-culture systems include these intercellular signaling factors. Juxtacrine signaling can be observed in the contact co-cultures and paracrine signaling in non-contact co-cultures. Transwell co-cultures are the most commonly used 3D co-culture models in previous studies. This model is suitable for both contact and non-contact co-culturing.
Transwell co-cultures are time- and cost-effective, but best suited for short-term culturing, since the cells in transwells display loss of BBB characteristics in longer cultures (Cho et al, 2017). The model is well suited for permeability assays, such as Lucifer Yellow and TEER measurements (Ruck et al, 2015), as this form enables access to both apical and basolateral side. A spheroid model of a BBB on the other hand gives a unique opportunity to study the cell interactions when placed freely together in media (Cho et al, 2017). The cells spontaneously self-arrange and form a contact co-culture. The model offers a way of studying the unrestricted interactions between all the cell types used.
8 Jari Koistinaho’s group from A.I. Virtanen Institute for Molecular Sciences (AIVI) have previously differentiated hiPSC cells into endothelia cells and astrocytes (Oksanen et al, 2017).
The first aim of this study was to optimize the differentiation and co-culturing protocols for hiPSC-derived models of the BBB. The second aim was to see, if there were any differences between the healthy and AD co-cultures. We discovered that the number of hiPSCs plated at the beginning of the differentiation effects the differentiation efficacy. The differentiated monocultures were characterized for their anatomical protein expression and gene expression.
The ECs and pericytes formed did express the expected cellular markers. Some differences were discovered between healthy and disease models. There were some technical difficulties with the contact co-cultures and the results did not reflect the previous studies. The spheroid model was successfully created using 96-well U-bottom plate, but ICC imaging revealed that they had self-assembled in a reverse order compared to previous studies (Cho et al, 2017). The spheroids were also characterized for their gene expression. This study gives indications on how to proceed with further studies: selecting appropriate hiPSC number for the differentiation, controlling the differentiated cell types and for instance collecting a larger number of spheroids for analysis. With few modifications to the protocols this type of hiPSC derived BBB models have prospects of shedding light on the AD pathogenesis.
Materials and methods
Cell cultures
The cell cultures in this study were created from hiPSCs. The hiPSCs included healthy control-line, a causative AD mutation (PSEN ∆E9), and a protective mutation (APP A673T, Islandic) (Tab. 1S). These cells were used to differentiate ECs, pericytes, and astrocytes (Tab. 2S). The hiPSCs were previously reprogrammed from fibroblasts with Sendai virus (Lehtonen et al, 2018). Fibroblasts were transduced with vectors including the four Yamanaka factors OCT-3/4, KLF-4, SOX-2 and c-MYC.
The dermal biopsies had been collected from patients with informed consent, with the approval from both the Research Ethics of Northern Savo Hospital District (license no. 123/2016, AD-lines) and the Ethics Committee of the Helsinki University Hospital District, (license no.
262/EO/06, control-lines). The hiPSCs used in this study were generated and characterized in
9 the Stem Cell Laboratory of Molecular Brain Research Group at the University of Eastern Finland as previously described (Lehtonen et al, 2018;Oksanen et al, 2017).
Essential 8 cell culture media (Life Technologies) was used to culture the hiPSCs. Matrigel basement membrane matrix (growth factor reduced, phenol red -free, by Corning), was used to coat the 3.5 cm dishes (Sarstedt) used for culturing. When passaging, the hiPSCs EDTA 0.5 M pH 8.0 (Life Technologies) with 2.5mM ROCK-inhibitor (Selleckchem) was used.
Differentiating ECs was done modifying a protocol previously published in Stem Cells -journal (Harding et al, 2017). The hiPSC were cultured until they were all with consistent morphology and in the same growth phase, the dishes ~75% confluent. The cells were grown in incubator 37˚C and 5% CO2.
Table 1. Mediums for the pericyte culturing. The mixture of Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12), Penicillin-Streptomycin (P/S), GlutaMax, MEM Non-essential Amino Acid Solution (MemNEAA) with 10% fetal bovine serum (FBS), was the most successful one. Endothelial cell growth medium MV2 (ECGM MV2) was also experimented on.
DMEM/F12 P/S GlutaMax MemNEAA FBS%
x x - - 5
x x - - 10
x x x - 5
x x x - 10
x x x x 10
ECGM MV2 x - - -
As the aim was to optimize the differentiating protocol, different number of hiPSCs were used for the initial passaging. In the first set there were ~250 000-300 000 hiPSCs/3.5 cm dish and in the second set ~100 000-200 000 hiPSCs/3.5 cm dish. Different mediums, StemDiff APEL (StemCell Technologies) and N2B27 medium (1:1 Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 DMEM/F12 (Gibco), 1:1 Neurobasal medium (Gibco), 0.5x B27, 0.5x N2, 0.5x GlutaMax, Penicillin-Streptomycin 50 IU/50 µg (P/S, Invitrogen), 0.05 mM β-mercaptoethanol) were experimented on. The hiPSCs were differentiated with a selective GSK-3 inhibitor, CHIR 6µM (Cayman), for two days to give rise of mesoderm. From day 3 to 5, growth factors bone morphogenetic protein 4 (BMP4, by Peprotech, 25 ng/ml), fibroblast growth factor (FGF, by Peprotech, 10 ng/ml), vascular endothelial growth factor (VEGF, by Peprotech, 50 ng/ml) were used. At day 6, the ECs were harvested with TrypLE
10 Express (Gibco) by incubating in 37°C for 1-2 minutes. Endothelial cell growth medium MV2 (EM, by PromoCell) was used to culture the ECs with added VEGF (50 ng/ml) to mature them.
The cells were purified by passaging. The ECs were cultured in EM on 3.5 cm or 6 cm dishes (Sarstedt). The media was changed every 2 to 3 days and the passaging done with TrypLE.
In optimizing the pericyte differentiation, previous studies were utilized (Orlova et al, 2014).
Pericyte differentiation followed the EC-protocol from the beginning up until day 6. Same mediums, StemDiff APEL and N2B27, with same growth factors were used (see above). After the cells were lifted with TrypLE, the ones still attached on the base of the dish were used to differentiate pericytes. The commercial StemDiff was selected as the differentiation media.
The medium experiments extended also to the pericyte culturing. Several different mediums were tested to find the optimal growth media for the pericytes (Tab. 1). Pericyte medium (PM:
DMEM/F12 with MEM Non-essential Amino Acid Solution (MemNEAA) 1x1 10µl/ml, GlutaMax 1x1 10µl/ml, 50 IU/50 µg P/S, FBS 10%) and EM were selected as medium of choice for the second differentiation batch. Growth factors, transforming growth factor beta-3 (TGF3β, 2 ng/ml) and platelet-derived growth factor-beta (PDGF-BB, 4 ng/ml), were used to mature the pericytes for 4 to 6 days.
The astrocytes were differentiated by the scientist in Jari Koistinaho’s research group Stem Cell Lab – team with the protocol presented in prior publications (Oksanen et al, 2017). The astrocytes were between 5.5 to 7 months old when used for co-culturing. These cells were grown as astrospheres with astrocyte medium (AM: 1:1 DMEM/F12, 50 IU/50 µg P/S, 0.5x N2, 0.5x GlutaMax, 0.5x Mem NEAA), with growth factors FGF (10 ng/ml) and EGF (10 ng/ml). For the co-cultures, astrospheres were collected into 1.5 ml Eppendorf tubes.
The estimated cell count was ~ 20 000 to 50 000 cells in one sphere and number of harvested spheres was estimated according to the needed cell count in different co-cultures. Medium was removed, the cells washed with 200 µl of PBS. Cells were detached with Accutase (200 µl, Gibco) for 8 min in 37°C. Accutase was removed and cells rinsed with AM (200 µl). Then the cells were suspended into single cells in 1 ml of AM.
Co-culture models
Both transwell and spheroid co-culture models were experimented on. Transwell co-cultures were established on 24-well cell culture plates (Cellstar, Greiner Bio-one) along with TC-inserts with 3 µm pore size (Sarstedt). The cells were grown in non-contact and contact
11 formations (Fig. 2). In the non-contact co-culture, one cell layer was plated onto the bottom of the 24-well plate and another into the basolateral side of the insert. In the contact co-culture, the cell layers were on opposite sides of the insert. This way both paracrine and juxtacrine signaling could be observed.
Figure 2 Co-culture models produced in transwells. E. Monoculture of ECs on the apical side of the insert.
Double-co-cultures A. ECs on the basolateral side of the insert and pericytes at the bottom of the well. B. Pericytes on the basolateral side of the insert and ECs on the bottom of the well. C. Astrocytes on the basolateral side of the inserts and ECs on the bottom of the well. D. Both pericytes and astrocytes at the bottom of the well. F. ECs on the apical side of the insert and astrocytes on the basolateral side. Triple-co-culture G. ECs on the apical side of the insert and both pericytes and astrocytes on the basolateral side.
ECs and pericytes of the AD-line were used to form a non-contact, double-co-cultures. One cell layer was on the bottom of the well (200 000 cells/ECs or 15 000 cells/pericytes) and another on the basolateral side of the inserts (3 000 cells/pericytes or 50 000 cells/ECs) (Fig. 2 A & B). Three different medium mixtures were used to see how the cells endure serum-free conditions. The serum levels were gradually reduced (Tab. 2). The human endothelial serum-free media (hESF) had 50 IU/50 µg P/S.
Table 2. The serum level for the co-cultures was gradually reduced to see how the cells endure serum-free conditions. Endothelial medium (EM), human endothelial serum-free media (hESF), astrocyte medium (AM), and pericyte medium (PM) in different mixtures were experimented on. (FBS =fetal bovine serum)
DAY 0 DAY 1-3
½ EM + ½ hESF hESF
½ EM + ½ hESF ½ AM + ½ hESF.
½ PM + ½ hESF ½ PM without FBS + ½ hESF
12 The pericytes were cultivated with these 3 different media. At day 8 astrocytes (50 000 cells) of the control-line were added to the AD pericytes and pericytes (15 000 cells) and astrocytes (50 000 cells) of the control-line were plated simultaneously (Fig. 2 D). No maturation factors were used for the astrocytes. Three different mediums were then applied: ½ hESF+ ½ AM,
½ PM w/o FBS + ½ AM, and AM. These co-cultures were cultivated for 7 days and then fixed for ICC.
Transwell, contact co-culture models were formed on Matrigel-coated inserts. First astrocytes (25 000 cells) were plated on the basolateral side of the insert. From AD-line, a combination of astrocytes (25 000 cells) together with pericytes (5 000 cells) was also plated on the basolateral side of the insert. After 4 h, the inserts were reversed and ECs (125 000 cells) were plated onto the apical side. These co-cultures were grown in ½ EM + ½ hESF. On day 3 the medium was changed to hESF and the co-cultures were used for TEER and permeability testing the next day. Transwell, non-contact co-cultures were formed with ECs (175 000 cells, on the bottom of 24-well plate) and astrocytes (25 000 cells, on the apical side of the inserts) (Fig. 2C).
The astrocytes were matured using AM with BMP4 (10ng/ml) and CNTF (10ng/ml) for 4 days before forming the EC+A co-cultures. Co-cultures were grown using ½ EM + ½ AM for 5 days.
The spheroid co-cultures using ECs, pericytes, and astrocytes, were formed using prior studies as a starting point (Cho et al. 2017). Control, AD and protective cell lines were used to create spheroids on low attachment 96-well U-bottom plates (Thermo Scientific). Two different mediums (EM and ½ EM + ½ AM), two different mounting styles (with or without Matrigel) and two different cell numbers (1500 or 3000 cells/cell type) were experimented on (Fig. 3).
The cells were pooled and pipetted directly onto the plate, either freely into the medium (100µl medium and 30 µl of cell mixture) or into Matrigel that had been allowed to set for few minutes (50 µl of Matrigel and 30 µl of cell mixture). The media was changed every 2-3 days and spheroids were maintained for 11 days. The growth and spheroid formation were monitored by taking bright field images (Zeiss Observer.z AX10) and measuring the size with ImageJ.
13
Figure 3 The 96-well U-bottom plate setup for the spheroids. EM=endothelial medium AM=astrocyte medium. Non-MG=cells placed directly into the medium MG= cells pipetted into the slightly coagulated Matrigel. cntr=control, AD=diseased MAD= protective cell line. 1500/3000 cells/cellular type were plated. The cell types used were ECs, pericytes, and astrocytes.
Immunocytochemistry
EC, pericyte and astrocyte samples were fixed with 4% formaldehyde (20 min at room temperature (RT)) and stored in 4°C with phosphate buffered saline (PBS, Gibco)-solution.
The cells that were stained with intracellular antibodies were permeabilized with Triton-X (Sigma-Aldrich). Different stains were permeabilized with different concentrations for 30 min at RT. 0.1% Triton-X for PDGFRβ, 0.25% Triton-X for pericyte-astrocyte co-cultures, 0.4% Triton-X for α-SMA, vWF, and Cal stainings. Blocking was performed by incubating the cells in 5% normal goat serum (NGS, Chemicon) for 1h at RT. Primary antibodies (Tab. 3S) were diluted to 5% NGS blocking buffer (except for AQP4 where 1% bovine serum albumin, (Sigma) was used), and the samples incubated overnight at 4°C. Secondary antibodies were diluted in 1:300 PBS and incubated at RT for 1h. The nuclei were stained with 4′,6-diamidino- 2-phenylindole 1:2000 (DAPI, Sigma) and incubated at RT for 5 minutes. Fluoromount-G (Invitrogen) was used to mount the samples onto glass slides. Some of the primary antibodies were conjugated with fluorescence probe and work with those was done in dark room conditions as well as with the secondary antibodies.
The spheroids were collected into 500µl Eppendorf tubes where they were first fixed and then stained. The same protocol was followed as before, except the incubation steps were performed on a reciprocating shaker. After staining, the spheroids were placed onto coverslips and fixed
14 with Fluoromount-G onto glass slides. The stained samples were imaged with Zeiss Vert.A1 AX10 (protective ECs and all co-cultures), Zeiss Imager M2 AX10 (spheroids) and Olympus Bx51 (control and AD ECs and all pericytes). List of both primary and secondary antibodies with the working dilutions and manufacturers details can be found in the supplementary material (Tab. 3S).
qRT-PCR
The RNA extraction was performed according to the QIAGEN RNeasy Mini Kit protocol and RNA-concentrations were measured with DeNovix Ds-11 Fx spectrophotometer/fluorometer.
cDNA synthesis was performed following Thermo Fisher’s Maxima RT protocol. RNA concentration varied between 2.2 (control spheroid) to 262 ng/µl (EC, protective line). The spheroids were pooled for the RNA extraction, 6x3000 cells/cell type. The final concentration of the cDNA was 2.5 ng/µl, except the spheroids that had only 1.25 ng/µl due to poor yield.
A different protocol was used for the samples depending on their original RNA concentration.
When the RNA concentration was high >50 ng/µl, the sample volume was adjusted to 12.5 µl and when it was low < 50 ng/µl a larger sample volume 27.5 µl was used. The primers used in these assays were TaqMan Gene Expression Assays (Applied Biosystems by Thermo Fisher Scientific). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an endogenous control. Genes of interest (Tab. 4S) were grouped into four categories: tight junction proteins, transporter proteins, biomarkers and miscellaneous. qRT-PCR was performed on the StepOnePlus Real-Time PCR System (AB Applied Biosystems).
TEER-measurement
TEER was measured from the contact co-cultures using CellZscope from NanoAnalytics. EC monocultures and EC-astrocyte co-cultures from control, AD, and protective-line were measured. A triple co-culture of ECs-pericytes-astrocytes from AD-line was measured. The manufacturer’s protocol was followed. The function of the equipment was tested prior to the actual measurement. The inserts with the contact co-cultures and control monocultures, were moved to the wells in CellZscope. The device was placed in the incubator at 37˚C with 5% CO2.
The cells were let to adjust for 30 min before the single time point measurement.
15 Permeability testing
Paracellular permeability was tested with Lucifer Yellow (LY, Sigma) using the inserts from TEER-measurement. A set of standards were made (1 µM, 0.5 µM, 0.25 µM, 0.125 µM, 0.0625 µM, 0.03125 µM, 0.015625 µM, 0 µM). The LY buffer was produced by combining Hanks’ Balanced Salt solution (HBSS) x1 (Gibco), HEPES 1 M (Gibco), MgCl2, and CaCl2. The buffer was pH balanced to 7.45. and was added and the solution filtered with 0,2 µm PES-filter. The inserts from the CellZscope were transferred back to the 24-well plate, and 600µl LY buffer was added into the wells and 200µl of LY buffer with 100µM LY was added into the inserts. The cells were placed into the incubator for 80 min and then medium samples were drawn from the bottom of the wells, onto 96-well plate. The fluorescence from the samples and standards were measured with the Wallac Victor2 1420 Multilabel counter using excitation wavelength 485 nm and emission wavelength 535 nm.
Cytokine exposure
Spheroids and AD pericytes were treated for 24 h with IL-1β (10 ng/ml) and TNF-α (50 ng/ml).
The pericytes were treated with both cytokines separately, but the spheroids were treated with both cytokines at the same time. Medium samples were collected pre- and post-treatment for protein quantification with cytometric bead array (CBA). RNA was extracted from the samples and qRT-PCR was used to visualize gene expression of few genes of interest.
CBA
Cytometric bead array (CBA) was used to detect the cellular response of the treated spheroids and AD pericytes. CBA uses fluorescence detection of flow cytometry together with antibody-coated beads to capture chosen analytes. BC CBA protocol was followed. The analytes of interest were cytokines/chemokines typically expressed under immunological stress intracellular adhesion molecule 1 (ICAM-1), vascular cell adhesion molecule 1 (VCAM-1), interleukin-6 (IL-6), colony stimulating factor 2 (GM-CSF), C-C Motif Chemokine Ligand 5 (RANTES), and C-C Motif Chemokine Ligand 2 (MCP-1). Beckman Coulter CytoFlexS flow cytometer was used for the CBA measurement.
Data analysis
Microsoft Excel, ImageJ and GraphPad Prism were used to quantify the results. Outliers were removed from the endothelial cell qRT-PCR results with using Grubbs’ or extreme studentized
16 deviate test, before the statistical analysis. One-way-ANOVA with non-parametric Kruskal-Wallis test was used to detect statistical significance, Dunns test was used as a post-hoc, to better distinguish the significance. Significant p-values were coded with stars:
p<0.05 *, p<0.01 **, and p<0.001 ***.
Results
Cell count influences the differentiating efficacy
The differentiating protocol for endothelial cells was optimized, by selecting an optimal differentiation medium and the number of hiPSCs to be used for the initial passaging (Fig. 4A, 4B). The commercial StemDiff APEL medium had a better yield of ECs compared to the N2B27 medium. Pericyte differentiation from the vascular progenitors benefitted from using PM. Adding the non-essential amino acids seemed to improve the survival of the pericytes. When using these other media mixtures, the pericytes suddenly started to die and produce twig-like debris around them. Using PM in the initial passages, killed remaining ECs and purified the cultures effectively. It was also noted that the pericytes cultured solely with EM, started to turn back into cells resembling EC morphology.
The number of cells in the initial passage had an impact on the final yield of different cells.
When using 250 000 hiPSCs/3.5 cm dish, the yield of EC’s was lower. From this batch more pericytes could be collected. Using 100 000 hiPSCs/3.5 cm dish gave a greater yield of EC’s but not as many pericytes, as the differentiation was almost complete. To differentiate both pericytes and ECs it is best to do them from two separate dishes with different cell counts.
Differentiated cells displayed the expected characteristics
Cells from both control and AD lines were stained together with a protective line (differentiated earlier) to characterize the ECs. EC markers and cellular junction proteins; von Willebrand factor (vWF), platelet endothelial cell adhesion molecule (CD31), ZO1, CLD5, and CDH5 were visualized (Fig. 5A). The cells expressed these EC markers. The protective line differentiated on a prior differentiation had more distinct expression of tight junction proteins CDH5 and CLD5 to the control and AD lines on this differentiation batch. This could be caused by some differences in the differentiation process. There were some technical issues with the
17 laboratory equipment and for instance the incubator CO2-level was 6 to 6.5% in the beginning when culturing the hiPSCs.
Figure 4 ECs and pericytes were differentiated from the hiPSCs. A. Illustration of the differentiation steps.
From vascular progenitors onwards, the differentiation protocol was different between ECs and pericytes.
B. Representative images from the differentiation of hiPSCs into ECs and pericytes. Brightfield images were taken from the control-line. I. hiPSCs. Scalebar 50 µm. II. Day 3 of differentiation: mesoderm. Scalebar 50 µm.
III. Day 6 of differentiation: vascular progenitors. Scalebar 100µm. IV. Day 10 of differentiation: EC p.0.
Scalebar 100µm. V. Day 10 of differentiation: P p.0. Scalebar 100 µm. VI. EC p.2. Scalebar 100 µm.
VII. Pericytes p.1. Scalebar 100 µm.
18
Figure 5 ECs differentiated from hiPSCs displayed the expected cellular markers. A. Representative fluorescence images of ECs from control, AD and protective lines stained for von Willebrand Factor (vWF) and (CD31), tight junction proteins zonula adherens 1 (ZO1), VE-cadherin (CDH5), and claudin-5(CLD5). Cells from the protective line were differentiated prior to the control and AD lines. DAPI (blue) was used to stain the nuclei.
Control & AD: Scalebar 100µm for the vWF, 50µm for CD31, ZO1, CDH5 and CLD5. Protective: Scalebar 50µm for vWF, CD31 and ZO1, 100 µm for CDH5 and CLD5. (B-D.) Gene expression from the control, AD and protective line. The gene expression was normalized to housekeeping gene GAPDH. The data is presented as fold change to control (mean +/-SD). Statistical significance analysed with one-way ANOVA. * p<0.05, **p<0.01 B.
Tight junction protein gene expression. CDH5 (n: ctr=10, AD=8, prt=14), CLD5 (n: ctr=9, AD=8, prt=14), OCLN (occludin. n: ctr=9, AD=7, prt=13), TJP1 (tight junction protein1 or zonulla adherens 1 n: ctr=10, AD=8, prt=13).
C. Gene expression of AD associated transporter proteins ABCB1 (p-glycoprotein, n: ctr=9, AD=8, prt=14), LRP1 (low density lipoprotein receptor-related protein 1, n: ctr=9, AD=8, prt=12), MMP9 (matrix metalloproteinase 9, n: ctr=3, AD=3, prt=3) and RAGE (receptor for advanced glycation end-products, n: ctr=10, AD=8, prt=14). D.
Gene expression of genes associated with glucose metabolism. GSK3β (glycogen synthase kinase 3 beta, n: ctr=10, AD=8, prt=14), GLUT1 (glucose transporter 1, n: ctr=10, AD=8, prt=14), GLUT4 (glucose transporter 4, n: ctr=9, AD=7, prt=14), INSR (insulin receptor, n: ctr=10, AD=8, prt=14)
19 To analyze the EC gene expressions, 12 genes were selected for analysis, among them tight junction proteins that were visualized with ICC as well. Additionally, to CDH5, CLD5 and TJP1 (=ZO1), OCLN mRNA expression level was analyzed (Fig.5B). The results revealed no statistical significance in the differences between the study groups, except for AD-ECs that had a higher expression of CLD5 (statistical significance p<0.05) and TJP1 (p<0.001) compared to the control. However, the difference in gene expression between the lines was not seen in the ICC images displaying CLD5 and ZO1 (=TJP1) (Fig.5A).
When analyzing the AD-associated transporter protein gene expression (Fig.5C), no statistical significance was observed between the groups of ABCB1 and LRP1, but the expression of RAGE was elevated on AD ECs when compared to the protective ECs group (p<0.05). When separating the different cell-lines within the protective group (Fig. 1S), statistical significance was also observed. For example, the ABCB1 expression had seven-fold increase in expression compared to the control and AD (p<0.05) on protective line Mad12cl2. Due to the high SD of the groups, no statistical significance was observed with AD associated matrix metalloproteinase 9, MMP9, even though there was a tendency of higher expression on the AD-line ECs. The third group of genes of interest was involved in glucose metabolism (Fig. 5D). No statistical significance was observed on GSK3β, INSR and GLUT1. The GLUT4 expression was low, but statistically significant between the protective and control groups (p<0.05). GLUT4 expression was increased on the protective group.
ICC was utilized also in the characterization of the pericytes. Pericyte markers PDGFRβ, α-SMA and caldesmon were expressed in hiPSC derived pericytes (Fig. 6A). The AD pericytes appeared to be larger in size as seen from both the fluorescence and brightfield images. To analyze the mRNA expressions q-RT-PCR was applied (Fig. 6B). Pericyte marker PDGFRβ along with glucose transporters GLUT1 and GLUT4, appeared to have a lower expression on the AD pericytes. All the groups expressed ANPEP pericyte markers. MMP9 expression appeared to be higher on AD pericytes.
Pericytes were exposed to cytokines to see how they respond to immunological stress. 24 h exposure of IL-1β and TNF-α was used to simulate this stress. Due to the lack of cellular material, only the AD pericytes were used, and the untreated cells served as control. The response was analyzed with CBA and qRT-PCR. The cytokine exposure successfully flared up response in all treatment groups compared to untreated controls (Fig. 7A). IL-1β elevated IL-6
20 secretion on the AD pericytes by ten-fold and GM-CSF secretion eight-fold. Both cytokines had an impact on the secretion of the analytes, but the selected analytes were secreted more from TNF-α stimulus than IL-1β stimulus. Secretion of V-CAM 1 and RANTES were elevated slightly by TNF-α, but it elevated ICAM-1 and MCP-1 secretion even more. The cytokine exposure also altered the gene expression (Fig. 7B). The TNF-α treatment accounted for the greatest changes between the groups. The TNF-α stressed pericytes had a tendency to downregulate their GLUT4 expression and upregulate their MMP9 expression.
Figure 6 Pericytes expressed typical cellular markers when analysed with ICC for their anatomical features and qRT-PCR for their gene expression. A. Representative ICC images of control and AD pericytes. Pericyte markers platelet derived growth factor β (PDGFRβ, red, left), α-smooth muscle action (α-SMA, green, middle) and caldesmon (CAL, red, right) were used. DAPI (blue) was used to stain the nuclei. Scalebar 50µm for α-SMA, 100µm for PDGFRβ. B. Pericyte gene expression of ANPEP (aminopeptidase N), GLUT1 (Glucose transporter 1), GLUT4 (Glucose transporter 4), MMP9 (matrix metalloproteinase 9) and PDGFRβ were analyzed. Gene expression normalized to GAPDH, fold change from control.Data presented as mean +/-SD. (n: control=1, AD=5)
21
Figure 7 AD-pericytes modulated their gene expression after 24 h exposure of cytokines IL-1β 10 (ng/ml) and TNF-α (50 ng/ml). A. Analytes inspected were signalling molecules that the cells typically express as a response to immunological stress; intracellular adhesion molecule 1 (ICAM-1), vascular cell adhesion molecule 1 (VCAM-1), interleukin-6 (IL-6), colony stimulating factor 2 (GM-CSF), C-C Motif Chemokine Ligand 5 (RANTES), and C-C Motif Chemokine Ligand 2 (MCP-1). IL-1β elevated IL-6 and GM-CSF secretion when TNF-α elevated ICAM-1, MCP-1, RANTES and V-CAM-1 in AD pericytes. Data analysed with CB array. (n=2 on all groups) B. The stimulated AD-pericytes were analysed for their gene expression of ANPEP (aminopeptidase N), GLUT1 (Glucose transporter 1), GLUT4 (Glucose transporter 4), MMP9 (matrix metalloproteinase 9) and PDGFRβ (platelet derived growth factor β). AD-pericytes displayed the tendency to downregulate the GLUT4 expression and upregulate the MMP9 expression upon TNF-α stimulus. Gene expression normalized to GAPDH, fold change from control. Data presented as mean +/-SD. (n=2 on all groups)
Transwell co-cultures from the control-line indicated a tendency of increasing their tight junction protein gene expression compared to monocultures
The ECs, pericytes and astrocytes differentiated from hiPSCs were used to form co-cultures.
In preliminary testing non-contact co-cultures in transwells were formed from ECs and pericytes. Different mediums were tested, and co-cultures grown in EM and ½EM+½AM seemed most viable. These mediums were used also for spheroid testing. To test how long the cells could survive without supplementary serum, non-contact co-cultures from ECs and pericytes (AD-line) were formed and grown in serum-free media. The aim was to demonstrate
22
Figure 8 The ECs changed their gene expression when co-cultured with different cell types.
A. Representative bright field images of pericyte-astrocyte co-cultures on day 3. Left: Control line, both cell types plated simultaneously. Right: AD line, pericytes 7 days old when control astrocytes plated on top. Medium on both ½ Pericyte Media + ½ Astrocyte Media. Scale bar 100 µm. B. Pericyte-astrocyte co-cultures x10 and x20.
Pericytes (control/AD) stained with α-SMA (α-smooth muscle actin, green) and the astrocytes (control) with GFAP (glial fibrillary protein, red) or AQP4 (aquaporin 4, red). DAPI (blue) was used to stain the nuclei. The cultures were established at different times with different treatments. (C-E.) Gene expression from the non-contact
23
co-cultures. Gene expression normalized to GAPDH, fold change from control. Data presented as mean +/-SD.
(n: control=1, control EC+A=2, AD4=2, AD4 EC+A=3) C. Gene expression of tight junction proteins VE-cadherin (CDH5), claudin-5 (CLD5), occludin (OCLN) and tight junction protein 1 (TJP1). D. Gene expression of transporter proteins p-glycoprotein (ABCB1), low density lipoprotein receptor-related protein 1 (LRP1) and receptor for advanced glycation end-products (RAGE). E. Genes involved in glucose metabolism, glucose transporter 1 (GLUT1), glucose transporter 4 (GLUT4), glycogen synthase kinase 3 beta (GSK3β) and insulin receptor (INSR).
if the different cell types could provide each other enough supporting signals for survival. The serum level was gradually reduced to serum-free media. In the beginning, there was no visible difference between the mediums. On the third day, the ECs started to die, but the pericytes seemed unaffected. This test indicated that the ECs could not survive long periods without serum. The pericytes, on the other hand, appeared to be quite resilient in the serum-free media.
The EC’s were discarded at this point and the pericytes that were plated on the bottom of the well were cultured for contact pericyte-astrocyte co-culture.
Control pericytes and astrocytes were plated on the bottom of a 24-well plate, and control astrocytes were plated on top of the AD pericytes. These cultures were grown with different media mixes. No apparent difference was observed between the mediums. The control-line had pericytes and astrocytes plated simultaneously, with only a few of the pericytes attached. On day 3 the astrocytes were starting to form branches (Fig. 8A). On the simultaneously plated co-cultures dead cells were stuck to the pericytes. On day 7 (14 days after the original plating) the AD pericytes started to die. At this point these pericyte-astrocyte co-cultures were fixed for further ICC imaging. This test indicated that it is possible to plate the cells successfully either simultaneously or pericytes first. The ICC images revealed differences in the GFAP-branching between the co-cultures (Fig. 8B). The astrocytes grown with AD pericytes had less GFAP-positive branches than the ones grown with controls. The AQP4 and αSMA stained equally on both AD- and control-lines.
The mRNA expression was analyzed from non-contact co-cultures of ECs and astrocytes (EC+A), it was normalized to GAPDH and compared to respective monocultures. These co-cultures had the astrocytes on the basolateral side of the insert and ECs on the bottom of the well (Fig. 2C). The astrocytes were matured for 4 days on the inserts before forming the co-cultures. Co-culturing appeared to slightly increase the tight junction protein expression on control line (Fig. 8C) when on the AD line co-cultures, the gene expression appeared to reduce
24 upon co-culturing. The gene expression of AD-associated transporter proteins showed that the co-culturing upregulated the ABCB1 expression on the control line (Fig. 8D). When inspecting the expression of genes involved in glucose metabolism, control and AD lines had lower expression of GLUT4 and INSR on the co-cultures compared to the monocultures.
Contact co-cultures were analyzed for their tight junction and barrier forming abilities with TEER and permeability testing. EC monocultures were compared to double (EC+A) and triple co-cultures (EC+A+P). Empty, Matrigel-coated inserts were used for comparison to remove the background. TEER is a quantitative technique for measuring the integrity of tight junction dynamics in cell culture models (Srinivasan et al, 2015). It measures the transendothelial electrical resistance: the higher the measurement, the better the barrier formation. TEER measurement (Fig. 2S) was performed in a single timepoint from 3-day-old co-cultures. All the co-culture groups (control, AD and protective) had lower TEER-values compared to the monocultures. The AD monoculture had the highest TEER value, 60 Ω cm2, control monoculture only had 7.3 Ω cm2. EC+A co-cultures of AD had 21.7 Ω cm2, thus adding astrocytes with the ECs dropped the electrical resistance of the AD line by 35%.
Permeability testing was performed directly after TEER the measurement (Fig. 2S). In permeability testing a small molecule substance with a fluorescent stain – such as Lucifer Yellow (LY) – is applied inside the insert. If a physical barrier formation is successful, it would hinder the movement of this molecule. The fluorescence is then detected from the medium and measured with microplate reader. The results of this assay reflected the results of the TEER measurement. The co-cultures displayed lower barrier formation abilities by having a higher concentration of LY passing into the basolateral medium at the 80 min timepoint. The AD monoculture showed the highest barrier formation, only 0.4 µM was leaked through, when 9.1 µM LY was detected in the basolateral medium of control monoculture.
Spheroid formation was successful directly into the media
ECs, pericytes and astrocytes from control, AD and protective cell lines were used to create spheroid BBB models. The aim was to see how the cells self-organize into spheroids and how the intercellular signaling affects the stress response and gene expression. ECs, pericytes, and astrocytes were used. Two different cell numbers were experimented on: 1500 or 3000 cells/cell type. Spheroids were formed by mounting the cell mixture into Matrigel or injecting them freely into the medium. Best spheroid formation was achieved when the differentiated
25 cells were plated freely into the medium (Fig. 9 A & B). After 24 h loose collections of cells were formed, but after 48 h cells were forming denser spheroids. After a week of cultivation, noticeable debris was observed outside the spheroids. On day 10 the spheroids had more defined lines and had a tightly packed form.
The cells that were embedded in the half-set Matrigel did not form spheres as well as the cells placed in the medium (Fig. 9 C). The Matrigel trapped the cells separately in the gel and only a few clusters were formed. The cells continued to grow in the Matrigel, but the desired spheroid shape was lacking. No noticeable difference was observed among the cell lines regarding medium used (Fig. 9 D). AD-line had the best growth, these spheroids were the largest regardless medium contents. The protective line had better growth than the control line, thus the control spheroids were the smallest.
To characterize the spheroids, different methodological approaches were used. The spheroids were visualized for their protein expression, mRNA expression, and their response to cytokine stimulus were analysed. The spheroids were stained with EC, pericyte, and astrocyte markers to localize the different cell types within the spheres (Fig. 9 E). EC-markers, CD31 and CDH5 (VE-Cad) localized in the center of the spheroid, whereas astrocyte marker AQP4 and pericyte marker α-SMA were localized on the surface layers. To analyze the mRNA expression levels from the spheroids, several spheroids were combined, however the yield of mRNA was still low. AD-spheroid RNA concentration was 5.8 ng/µl and control spheroids 2.2 ng/µl. RNA collected from the control sample had low concentration and inadequate purity, and most of the qRT-PCR results from the control were unmeasurable. Since results from the control spheroids were not available, the AD-line gene expression was normalized to GAPDH using Q-gene (Fig. 9 F). Expression of astrocyte marker AQP4 along with EC tight junction genes CDH5, OCLN, and TJP1 along with Aβ-efflux pump LRP1 were detected.
To study the response to inflammatory stimulus, the spheroids were exposed simultaneously to IL-1β and TNF- α for 24 h. The samples were divided into untreated and treated groups and processed with CBA analysis (Fig. 9G). Both control and AD spheroids responded to the treatment by releasing chemokines and cytokines. The AD spheroids had a stronger response, especially in the case of chemokine monocyte chemotactic and activating factor (MCP-1) and, a vascular cell adhesion molecule 1 (V-CAM-1), that mediates leukocyte-endothelial cell adhesion and signal transduction.
26
Figure 9 The spheroids were formed without mounting media. Spheroid formation was experimented with plating the cell mixture directly into media or inside Matrigel mounting. ICC, qRT-PCR and cytokine exposure testing were utilized in characterizing the spheroids. A-C. Representative brightfield images of spheroid co-cultures. Scalebar 100 µm. A. Spheroids were grown without Matrigel with EC-medium. B. Spheroids were grown without Matrigel with ½ EC Medium + ½ Astrocyte Medium. C. Spheroids were grown in Matrigel. Both EM and ½EC Medium + ½ Astrocyte Medium. D. Size development of the spheroids created from 3000 cells/cell type. Size measured from images on day 1, 2 and 10 and analyzed with ImageJ. E.Representative immunofluorescence images of spheroids stained for EC and astrocyte markers CD31 and AQP4 (Aquaporin 4), respectively (left), and EC and pericytes markers VE-cad (VE-cadherin or CDH5) and α-SMA (α-smooth muscle
27
actin), respectively (right). Scalebar 100µm. F. Q-gene was used to normalize the AD spheroids gene expression of AQP4, CDH5, lowdensity lipoprotein receptor-related protein 1 (LRP1) and tight junction protein 1 (TJP1) to housekeeping gene GAPDH (n=2, except LRP1 n=1). Q-gene used to normalize the expression to GAPDH. G.
Release of IL-6, GM-CSF, VCAM-1, MCP-1, ICAM-1 and RANTES quantified from media after 24 h exposure to IL-1β (10 ng/ml) and TNF-α (50 ng/ml), analyzed with CB array. (n: control= 2, control treatment=3, AD=2, AD treatment n=3). The signaling molecules analyzed wereintracellular adhesion molecule 1 (ICAM-1), vascular cell adhesion molecule 1 (VCAM-1), interleukin-6 (IL-6), colony stimulating factor 2 (GM-CSF), C-C Motif Chemokine Ligand 5 (RANTES), and C-C Motif Chemokine Ligand 2 (MCP-1).
Discussion
The aim of this study was to differentiate hiPSCs to the cells of the BBB. These cells were then characterized, and co-cultures were formed from them to discover the intercellular interactions within the BBB. We discovered, that the density of hiPSCs that were plated for the initial differentiation effected the differentiation efficacy. While the lower cell count produced more ECs, the higher initial hiPSC count produced more pericytes. The protocol for differentiating pericytes was established and some characterizations performed for these cells. Different cell lines that were used in this project had different growth rates already at the hiPSC-stage, with AD-line having the best growing abilities. Isogenic control -lines were the most fragile, they differentiated poorly and had the slowest growth rates. Due to this, these cells were left out from the final testing.
The differentiation protocols could still be improved. The ECs and pericytes were purified by passaging, but to increase the purity of the monocultures e.g. MACS cell separator beads or flow cytometry could be utilized (Orlova et al, 2014). On the other hand, it is worth considering that these methods also require a lot more resources, both time and money, whereas passaging is a relatively fast and inexpensive way of purifying the cells albeit not as effective. Some other methods could be used when differentiating ECs. One could experiment with the transforming growth factor -β (TGF-β) pathway inhibitor SB431542. SB431542 supports EC expansion by inhibiting the anti-proliferative effects of endogenous TGF-β (Orlova et al, 2014). It could be used for the ECs from the vascular progenitor step to the endothelial progenitors together with VEGF.
In AD pathology the flow of Aβ from plasma to brain increases and decreases from brain to plasma. These changes are contributing to the Aβ accumulation in the brain (Lyros et al, 2014).
When observing the AD-associated transporter proteins in EC monocultures, LRP1 (efflux) expression was elevated together with RAGE (influx) expression. RAGE expression was
28 statistically higher (p>0.5) on the AD ECs compared to the protective ECs. On EC-astrocyte non-contact co-cultures LRP1 was increased on the control co-culture versus control monoculture, when it was lowered on the AD co-culture versus AD monoculture. RAGE expression was unaltered on the control cultures but appeared to decrease the expression on the AD co-cultures versus AD monocultures. For the future studies triple-co-cultures of ECs, pericytes and astrocytes could be implemented. Also, co-culturing control ECs with AD astrocytes and AD ECs with control astrocytes. This would give more knowledge of the co-operation of these three cell types in the BBB. Also, one could look at the EC-pericyte co-cultures and try and see if the PDGFRβ expression correlates with the LRP1.
EC tight junction mRNA expression was observed between control, AD, and protective-lines.
The expressions appeared to be quite uniform on CDH5 and OCLN, but AD ECs had higher CLD5 (p>0.05) and TJP1 (p>0.01) expression compared to control. This did not directly reflect the results of previous studies where AD models have been depicted with loss of tight junction proteins (Lyros et al, 2014). The matter could be further studied by performing a protein expression assay to see if the elevated gene expression also translates into elevated protein expression. Another point to consider is the fact that the AD line had the best growth of all three cell lines. The elevated expression could be due to their better growth. None the less, further studies are required. According to previous studies, co-culturing with pericytes and/or astrocytes increases tight junction gene expression on ECs (Cho et al, 2017; Sansing et al, 2012). Co-culturing was expected to support the full maturation of the cells by providing direct and indirect intercellular signals (Kook et al, 2017). With the control non-contact astrocyte-EC co-cultures, the gene expression of tight junction protein coding CDH5, CLD5, OCLN, and TJP1 was increased compared to the respective monoculture. With the AD-line, co-culturing decreased the gene expression of tight junction protein-coding genes, apart from CDH5, that had slightly elevated expression. Further studies are needed to reveal if this change was due to technical issues or if it is caused by the genetic disposition of AD.
Chronic inflammation and type 2 diabetes have been linked to AD pathogenesis. In this study glucose metabolic changes such as glucose transporter, insulin receptor, and GSK3β gene expression were studied from ECs, pericytes, and non-contact EC-astrocyte co-cultures. The general expression for GSK3β, INSR, and GLUT4 were low on all cultures. No statistical significance was observed between AD and control-line EC monocultures. The protective ECs had a higher expression of GLUT4 (statistical difference p<0.5) and GLUT1, along with a
