Culture Model for Non-human Primate Choroid Plexus

While there are murine and rat choroid plexus epithelial cell cultures, a translationally relevant model for choroid plexus activation and function is still lacking. The rhesus macaque is the gold standard for modeling viral infection and activation of CNS, including HIV-associated neurocognitive disorders. We have developed a rhesus macaque choroid plexus epithelial cell culture model which we believe to be suitable for studies of inflammation associated with viral infection of the CNS. Epithelial morphology and function were assessed using vimentin, phalloidin, the tight junction protein zonula-occludens-1 (ZO-1), and focal adhesion kinase (FAK). Choroid plexus epithelial cell type was confirmed using immunofluorescence with two proteins highly expressed in the choroid plexus: transthyretin and α-klotho. Finally, barrier properties of the model were monitored using pro- and anti-inflammatory mediators (TNF-α, the TLR2 agonist PamCys3K, and dexamethasone). When pro-inflammatory TNF-α was added to the xCelligence wells, there was a decrease in barrier function, which decreased in a step-wise fashion with each additional administration. This barrier function was repaired upon addition of the steroid dexamethasone. The TLR2 agonist PAM3CysK increased barrier functions in TNF-α treated wells. We have presented a model of the blood-CSF barrier that will allow study into pro- and anti-inflammatory conditions in the brain, while simultaneously measuring real time changes to epithelial cells.


INTRODUCTION
The choroid plexus, comprising the blood-cerebrospinal fluid (CSF) barrier, lines the lateral, 3rd and 4th ventricles of the brain (Zheng and Zhao, 2002;Monnot and Zheng, 2013). It is responsible for secreting CSF, allowing the passive diffusion of water and oxygen into the brain, and facilitating the active transport of glucose and larger particles into the brain (Zheng and Zhao, 2002;Monnot and Zheng, 2013). It has low pinocytotic activity and also contains selective transporters including multidrug resistance proteins that are responsible for allowing drugs into, and out of, the brain (Rao et al., 1999;Varatharajan and Thomas, 2009;Schwerk et al., 2015). The choroid plexus also has a unique double-layer structure of relatively leaky fenestrated capillary endothelium and tight junction dense epithelium separated by a narrow space devoid of astrocytes and neurons (Liddelow, 2015). There are also unique populations of macrophages that line the endothelial blood vessels and choroid plexus resident macrophages (Bragg et al., 2002;Kim et al., 2006;Delery and Maclean, 2018).
Several rodent models of the choroid plexus exist, including immortalized rat Z310 and TR-CSFB, as well as primary mouse models (Kläs et al., 2010;Menheniott et al., 2010;Monnot and Zheng, 2013;Barkho and Monuki, 2015). Porcine choroid plexus epithelial cell lines have been used to study bacterial and drug infection of the brain (Haselbach et al., 2001;Angelow et al., 2004;Schwerk et al., 2011). However, the murine models have reported difficulty in proliferation in culture, the immortalized rat cultures have altered tight junction proteins compared to primary rat cultures, and none of the models is useful for studies of HIV infection (Angelow et al., 2004;Abel, 2009;Barkho and Monuki, 2015;Beck et al., 2017). To our knowledge, there currently is not a model that facilitates real-time measures of barrier function using label-free measurements. As viruses and viral proteins are likely to have effects over different time periods, such a model would be useful. Combined, there is a critical need for a translational choroid plexus model for neurovirology, neurodegeneration, and inflammaging studies. As the rhesus macaque is the gold standard animal model of studying viral pathogenesis in vivo, we designed a primary rhesus macaque choroid plexus epithelial cell culture.
Beginning with published methodologies for choroid plexus epithelium (Zheng and Zhao, 2002;Monnot and Zheng, 2013) and our previously established microvascular brain endothelial cells isolation protocol (MacLean et al., 2001(MacLean et al., , 2002Ivey et al., 2009b;Sansing et al., 2012), we designed and tested the following protocol for isolation, culture, and characterization of choroid plexus epithelial cells.

Primary Cell Culture Protocol
Rat tail collagen (Sigma-Aldrich, St. Loius, MO, United States) was dissolved in 0.1 M acetic acid to create the collagen-coated plates. Twenty-four hours prior to necropsy, 100 µL/well of the collagen solution was added to a 12 well plate. Gelling of collagen was induced by addition and changing M199 until the media did not change color. Plates were incubated at 37 • and 5% CO 2 overnight, with media on top in order to prevent the collagen from drying out.
Choroid plexuses from juvenile (1-5-year-old) Indianorigin Rhesus macaques were obtained at investigator-initiated necropsy at Tulane National Primate Research Center (TNPRC). Animals were housed and humanely euthanized according to standards set forth by the Office of Laboratory Animal Welfare (OLAW) and TNPRC's Institutional Animal Care and Use Committee (IACUC). A veterinary pathologist from TNPRC collected choroid plexus from the lateral, 3rd and 4th ventricles into PBS containing antibiotics/antimycotics (unless otherwise stated, all media was obtained from Gibco).
Using tweezers and a scalpel, the choroid plexus was minced and incubated with 2 mL collagenase/dispase (1 mg/ml, Roche, Indianapolis, IN, United States) at 37 • C and 5% CO 2 for 15-20 min. The digested tissue was centrifuged and washed twice with gentle trituration in 10 mL of complete media (450 ml M199, 50 mL exosome-depleted fetal bovine serum, 5 ml antibiotics/antimycotics, 1 mL Primocin (Invivogen) and 5 mL L-glutamine). Dissociated cells were resuspended in complete media and gently plated on top of the collagen gel (∼2 mL/well) and incubated at 37 • C and 5% CO 2 . Any contaminating fibroblasts in the primary cultures were inhibited with 3-5 days incubation with 25 µg/ml cis-4-hydroxy-Dproline in the complete media (Sigma-Aldrich, St. Louis, MO, United States). Reference Figure 1 for a step-by-step visualization of the protocol.
As cells approached confluence, they were sub-cultured using standard techniques (MacLean et al., 2001(MacLean et al., , 2002Ivey et al., 2009b;Sansing et al., 2012). Either collagen or gelatin was suitable for coating culture substrates beyond primary culture. In brief, cells were washed twice sequentially with PBS, then PBS-versene before trypsin-versene was added. Trypsin was stopped with complete media, and cells were washed and centrifuged twice prior to plating on the new plate or flask.
The media was pipetted off the coverslips, and cells were blocked for 15-60 min with 1% paraformaldehyde. Cells were permeabilized in PBS-FSG-Triton ×100 for 1 h before primary antibody was added at the previously reported concentrations. Cells were then washed with PBS-FSG for 5 × 5 min before secondary antibody was added for 1 h. For multiple stains the primary and secondary process was repeated. Coverslips were then placed on glass slides with DAPI mounting media.

Real-Time Monitoring of CSF Barrier Model
Cells were plated at a concentration of 10,000 cells/well on gelatin-coated xCELLigence 16 well E-plates (Acea Biosciences, San Diego, CA, United States). Cell indices were recorded continuously until plateaus were reached, indicating confluence Kumar et al., 2018). Wells were treated with one or a combination of: TNFα (100 U/ml), the TLR2 agonist Pam3CysK (1 µg/ml), or dexamethasone (1 µM, freshly prepared from a 10 mM stock dissolved in DMSO). Multiple controls were used, including media-only wells, and single-treated wells in the experiments where a combination of treatments were used. Four wells were used for each condition, and they were averaged together to produce our results using the inbuilt software. Error bars are ± standard deviation. Significance of increased or decreased impedance was calculated using Prism v8.2 using paired two-tailed t-tests.

Cell Culture
On initial plating, choroid plexus-derived cells had a circular morphology that became a mixed morphology after about 7 days (Figure 2A). By ∼14 days post isolation, there is evidence of aggregates of cells forming, with cuboidal/epithelial morphology ( Figure 2B). After 4-5 weeks, monolayers were evident ( Figure 2C). Cells were sub-cultured ∼1:4 after monolayer formation and reached confluence again after 2-3 weeks when grown on collagen coated coverslips ( Figure 2D). Although collagen appeared to improve the initial plating of primary choroid plexus cultures, choroid plexus epithelial cells grew as monolayers on either gelatin or collagen coated plates (not shown).
Expression of zonula occludens-1 (ZO-1) and FAK was confirmed in monolayer cultures on glass coverslips. Cells were positive for ZO-1 indicating the presence of tight junctions ( Figure 4A). We determined that the choroid plexus epithelial cells were immunopositive for FAK, which indicates cell-cell adhesions ( Figure 4B). The choroid plexus cells were also immunopositive for α-klotho, in a primarily perinuclear manner (German et al., 2012; Figure 4C).

Real-Time Model of Barrier Function
Choroid plexus epithelial cells at passage 3-4 were plated on xCELLigence E-plates, and allowed to adhere and spread overnight ( Figure 5A). Increased Cell Index is indicative of increased impedance . Once traces had reached a plateau, we modeled neuroinflammation by adding TNFα (100 U/ml) to half the wells (blue traces). This induced an initial slight increase in cell index, followed by a steady, sustained decrease ( Figure 5B, red = M199, blue = TNFα). A second administration of TNFα, 48 h later, induced a steeper and significant decrease in cell index (p < 0.005, n = 4). This indicates that TNFα induces barrier dysfunction of the rhesus choroid plexus epithelial cells.
Toll-like receptor (TLR2) is expressed on the choroid plexus of rhesus macaques (Delery et al., under review). We examined the effects of the TLR2 agonist PAM3CysK on choroid plexus barrier function. The greatest effect of PamCys3K was on cells pre-treated with TNFα, where we observed a sharp increase in cell index beginning approximately 10 h after incubation with PAM3CysK ( Figure 5C, blue = TNFα followed by M199, pink = TNFα followed by PamCys3K, p < 0.001, n = 4). PamCys3K also increased cell index when added to naïve control choroid plexus epithelial cells, compared to a baseline of M199 treatment; however, this increase in cell index was neither significant nor sustained ( Figure 5D, red = M199, green = PamCys3K, p = 0.597).
Finally, we examined the effect of the anti-inflammatory steroid dexamethasone on the barrier function. Following administration of 100 U/ml TNFα, the cultures were treated with dexamethasone (1 µM). Compared to TNFα treated cells only, the TNFα + dexamethasone treated cells saw a greater increase in cell index, with this effect increasing with each subsequent dose ( Figure 5E, blue = TNFα only, purple = TNFα followed by dexamethasone). Dexamethasone only modestly increased the cell index in naive choroid plexus monolayers, compared to control cells (Figure 5F, red = M199, green = dexamethasone).

DISCUSSION
While the choroid plexus is a critical brain structure that allows for the selective trafficking of immune cells into the brain during normal surveillance (Williams and Hickey, 1995), it is also an understudied viral entry point. Due to the highly fenestrated nature of the endothelial cells of the choroid plexus, it is quite possible that the choroid plexus could act as a "back-door" for HIV entry into the brain. As HIV does not have a translationally relevant small animal model, we believed it was critical to develop a choroid plexus cell culture model in a species that does allow studies of HIV neuropathology: the rhesus macaque (Ivey et al., 2009a). We believe this model will also be useful for the study of accelerated aging or "inflammaging, " as the choroid plexus is the main source of α-klotho, the "anti-aging hormone, " in the brain (Kuro-o et al., 1997;Takahashi et al., 2000;German et al., 2012). This cell culture model of the choroid plexus was verified by the presence of transthyretin, vimentin, phalloidin, ZO-1, FAK, and α-klotho. It is well established that transthyretin is a unique marker of choroid plexus epithelial cells (Zheng and Zhao, 2002;Baehr et al., 2006;Monnot and Zheng, 2013). By performing double-and triple-labeled immunocytochemistry with vimentin and phalloidin, both structural proteins (MacLean et al., 2005;Renner et al., 2013;Tavazzi et al., 2014), we determined the cuboidal morphology of the transthyretin positive cells. The cells also expressed the tight-junction protein, ZO-1, which is also of functional importance to choroid plexus barrier function (Howarth and Stevenson, 1995;Emerich et al., 2006;Szmydynger-Chodobska et al., 2007;Ivey et al., 2009b). Lastly, positive staining for the anti-aging protein α-klotho, further bolster this cell culture model of choroid plexus barrier (Kuro-o et al., 1997;Takahashi et al., 2000;German et al., 2012). These results were replicated in cultures derived from five animals.
Collagen-coated plates permitted selective attachment and growth of epithelial cells and allowed us to change the media and wash off debris with no adverse effects, including removal of poorly adhered cells, resulting in minimal-to-no fibroblast contamination. While growth was slower than other CNS cell types cultured in our labs, with this model taking about 6-8 weeks to reach a level of confluence to subculture, once established the choroid plexus epithelial cells were stable for at least five sub-cultures as confirmed by immunofluorescence with the aforementioned phenotypic markers. . After the addition of TNFα (100 U/mL), there was an initial increase in cell index peak followed by a sustained decrease (red = M199 media, blue = TNFα) (B). This effect was amplified with a second challenge of TNFα. There was an unexpected increase in cell index following treatment with the TLR2 agonist Pam3CysK in cells primed with TNFα (blue = TNFα, pink = TNFα followed by PamCys3K) (C). There was considerably less effect in naïve choroid plexus cells treated with PamCys3K versus media controls (red = media only, green = PamCys3K) (D). In separate experiments, cells were treated the cells with anti-inflammatory steroid dexamethasone after priming with TNFα. As anticipated, dexamethasone treatment increased the cell index compared to TNFα only treated cells (blue = choroid plexus cells + TNFα, purple = choroid plexus cells + TNFα + dexamethasone) (E). However, dexamethasone added to naive cells only induced a very modest increase in cell index compared to media alone (red = media only, green = naïve cells + dexamethasone) (F).
After sub-culturing onto xCelligence plates we monitored the barrier properties in real-time. TNF-α appeared to have a cumulative effect on the choroid plexus epithelial cells (Figure 5B), as cell index was subsequently lower after sequential TNFα challenges. This has interesting implications for multiple inflammatory challenges for overall brain health and repair both in vitro  and in vivo (Peterson et al., 2019).
We have previously shown increased TLR2 expression in the CNS in response to SIV, among other neuroinflammatory conditions (Lee et al., 2013a(Lee et al., ,b, 2014Snook et al., 2014;Inglis et al., 2016;Robillard et al., 2016;Chiu et al., 2019). Stimulation with TLR2 agonists would decrease BBB function both directly (Zhu et al., 2018), and indirectly, by activating astrocytes to secrete proinflammatory cytokines (Min et al., 2015) and induce leukocyte infiltration (Min et al., 2015). Indeed, increased TLR2 expression in parenchymal CNS cells has been linked to inflammaging (Wilhelm et al., 2017). In the current study, the TLR2 agonist PamCys3K increased cell index over time (Figures 5C,D). This effect was greatest in cells previously challenged with TNFα ( Figure 5C), however, PamCys3K administration still increased cell index in naive choroid plexus epithelial cells ( Figure 5D). The apparent dichotomy in our study is therefore most likely due to the epithelial nature of the cultured choroid plexus cells, rather than endothelial cells. Indeed, TLR2 increases barrier properties of epithelial cells in culture (Ragupathy et al., 2014). Our ongoing studies indicate that TLR2 is present in the choroid plexus epithelial cells of rhesus macaques, and this is lower in animals infected with SIV (Delery, under review).
As anticipated, the anti-inflammatory steroid dexamethasone increased cell index, and was even able to return TNFα treated cells close to normal cell index (Figures 5E,F). The effect was more pronounced in cells primed with TNFα than naïve choroid plexus cells, indicating a relatively intact barrier to begin with.
We have presented a model of the blood-CSF barrier that will allow study into pro-and anti-inflammatory conditions in the brain, while simultaneously measuring real time changes to choroid plexus epithelial cells. This rhesus macaque model could also allow us to study viral infections ex vivo. We propose using this model examine how multiple inflammatory challenges, or long-term exposure to anti-retroviral drugs could lead to early inflammaging.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript and/or, the supplementary files.

ETHICS STATEMENT
Animal Subjects: The animal study was reviewed and approved by the Tulane University IACUC.

AUTHOR CONTRIBUTIONS
ED performed the cell cultures under the guidance of AM who oversaw all procedures. ED and AM contributed to the writing of this manuscript.