In Vitro Model of Cerebrospinal Fluid Outflow through Human Arachnoid Granulations

Abstract

purpose. To describe and validate an in vitro model of the arachnoid granulation (AG) outflow pathway for cerebrospinal fluid (CSF), by using human AG cells grown on a filter membrane support and perfused in a modified Ussing chamber at pressures analogous to normal human intracranial pressures.

methods. Human AG cells were grown, characterized, seeded onto filter membranes, and perfused in the physiologic (basal to apical, B→A) or nonphysiologic (apical to basal, A→B) directions. Cells were fixed under pressure after perfusion and prepared for electron microscopy (EM).

results. The average cellular hydraulic conductivity in the B→A direction (10 total) was 4.52 ± 0.43 μL/min per mm Hg/cm² with an average transcellular pressure decrease of 3.13 ± 0.09 mm Hg. The average cellular hydraulic conductivity in the A→B direction (six total) was 0.29 ± 0.16 μL/min per mm Hg/cm² with an average transcellular decrease in pressure of 3.33 ± 0.16 mm Hg. Cells perfused nonphysiologically showed a large number of dead and dying cells. EM postperfusion analysis showed that AG cells were integrally attached to the underlying filter membrane. Large extracellular cisternal spaces were visible between overlapping AG cells and vacuoles within the cytoplasm. It is possible that these spaces within and between cells represent pathways for transcellular and paracellular transport of fluid.

conclusions. The results demonstrate that AG cells in vitro show a statistically significant greater flow rate and cellular hydraulic conductivity when perfused in the physiologic versus the nonphysiologic direction under normal intracranial pressures. These results suggest that this in vitro model of the AGs can accurately replicate the unidirectional flow of CSF in vivo.

The arachnoid granulations (AGs) project through the dura mater into the superior sagittal sinus and venous lacunae (Fig. 1) . ^{1 2} They are the interface between the CSF and venous blood. It is generally accepted that they function to return CSF to the systemic circulation. ^{3 4 5} Our understanding of CSF movement into the venous sinuses is limited. 

Idiopathic intracranial hypertension (IIH) is a disorder of increased intracranial pressure (ICP) with or without papilledema. Papilledema can lead to progressive irreversible visual loss and optic atrophy. An increased resistance to outflow may impair CSF egress in IIH. ^{6 7 8 9} To study the role of AG cells in CSF outflow and IIH, we have developed an in vitro model of the CSF outflow pathway across the arachnoid granulations that can be applied to conditions of increased intracranial pressure, such as IIH. By monitoring the pressure decrease and flow rate across the cell layer, it is possible to calculate the hydraulic conductivity (L _p) of the cell layer. 

AG cells are mesothelial cells, expressing both epithelial and fibroblast cell properties. They line the outermost aspect of the AG and are believed to be involved in regulation of CSF outflow. AGs are believed to be multilobular arachnoid villi, which are the initial growth of the arachnoid membrane through the dura. The distinction between arachnoid granulations and arachnoid villi is merely one of nomenclature. Both represent projections of the arachnoid membrane through the dura mater into the venous sinuses, and function in the outflow of CSF. However AGs are often described as macroscopic and visible to the naked eye, whereas arachnoid villi are microscopic. It has been speculated that arachnoid villi are present at infancy and ultimately enlarge and lobulate under increasing CSF pressure ¹⁰ to form AGs, although experimental studies in this area are lacking. The term arachnoid cell simply refers to the arachnoid cells that line the AG. This cell layer is continuous with the underlying arachnoid membrane and covers the surface of the granulation, where it clusters at the apex of the granulation, forming the cap cell cluster. For review of the structure of a human AG, see Kida et al. ¹¹  

It has been shown that human AG cells in vivo express some of the cytoskeletal proteins ^{12 13 14 15 16} and junctional complexes, ^{13 14 15 16 17} that are important in mediating cell-cell adhesion and communication, allowing arachnoid cells to form a barrier to CSF egress, regulating the return of CSF to the venous circulation. ^{11 13 14 15 16 17 18 19 20 21} The expression of many of these proteins has been further confirmed by our group, using in vitro cultures of human AG cells. ¹⁶ The next step in verifying the validity of this model is to demonstrate that human AG cells in vitro can replicate the unidirectional flow of CSF in vivo (i.e., from the subarachnoid space through the AGs to the venous sinuses). 

The mechanism by which the AG cells facilitate fluid transport is still debated, though several reports have noted the similarity to drainage of aqueous humor across Schlemm’s canal. ^{22 23} Such studies suggest that AGs, like Schlemm’s canal endothelial cells, regulate fluid flow by the formation of transcellular, basal-to-apical vacuoles, and function as one-way valves, ultimately fusing with the apical membrane and draining fluid to the venous lumen. Ultrastructural studies of human AGs by Kida et al. ¹¹ and Yamashima ^{19 20} confirmed this vacuolization mechanism. 

This article describes the validation of the in vitro model of human AG cells grown on filter membrane supports and perfused in both the physiological and nonphysiological directions in a specialized perfusion chamber under pressures that mimic normal intracranial pressures. 

Materials and Methods

AG Cell Culture

Brain tissue was obtained within 24 hours after death from the Ohio State University Regional Autopsy Center. Collection was in accordance with guidelines and regulations set forth by the Office of Responsible Research Practices Institutional Review Board for human subjects at The Ohio State University and The Declaration of Helsinki. At autopsy, AGs were collected from the superior sagittal sinus and adjacent tissue. Samples were placed into sterile phosphate-buffered saline with penicillin and streptomycin. Before explantation, the tissue was washed three times in sterile Dulbecco’s phosphate-buffered saline (D-PBS; Cellgro Mediatech, Herndon, VA), then plated into fresh culture medium of Dulbecco’s modified Eagle’s medium/Ham’s F-12 nutrient medium (50:50 vol/vol), with l-glutamine, penicillin-streptomycin, amphotericin B (all from Cellgro Mediatech), and 10% newborn calf serum (Invitrogen-Gibco, Carlsbad, CA). 

At explantation guided by a dissecting microscope, an individual granulation was secured adjacent to the apical cap cell portion of the granulation and cut just below the cap of the granulation to ensure that the cap cell cluster was explanted while excluding the fibrous capsule and central core. The AG cap was removed, washed again in medium, and placed in a 24-well culture plate coated with a fibronectin solution (30 μg human fibronectin [Sigma-Aldrich, St. Louis, MO] per millimeter of M-199 culture medium [Cellgro Mediatech]). Culture medium was added to each well, and the explants were incubated undisturbed for 3 to 4 days. The medium was changed every 3 to 4 days. When confluent, cells were washed with D-PBS, removed from their wells with 0.05% trypsin EDTA in Hanks’ buffered saline (Cellgro Mediatech) and spun down at 1300 rpm. The cell pellet was resuspended and plated to a T-25-cm² culture flask. In subsequent passages, the cells were grown in T-75-cm² flasks. 

Immunocytochemistry

The following monoclonal antibodies were used for immunofluorescence microscopy at the dilutions indicated: mouse anti-human cytokeratin antibody clones AE1/AE3 (1:50; DakoCytomation, Carpinteria, CA), mouse anti-human desmoplakin-1 and -2 antibody (1:40; Chemicon International, Temecula, CA), and FITC-conjugated mouse anti ZO-1 antibody (1:50; Zymed, South San Francisco, CA). The secondary antibody used was an Alexa Fluor 555–conjugated donkey anti-mouse IgG1 antibody (Invitrogen, Eugene, OR) at a 1:50 dilution for 45 minutes at 37°C and an FITC-conjugated goat anti-mouse IgG1 antibody (Sigma-Aldrich). All primary and secondary antibodies were diluted in 10% calf serum in D-PBS. 

Second- or third-passage cells were seeded onto fibronectin-coated coverslips (BD Biosciences, Franklin Lakes, NJ) and grown to confluence. Cell cultures were tested at 1 to 1.5 weeks after confluence for the presence of cytokeratins, desmoplakin-1 and -2, and ZO-1. The cells were washed three times with sterile D-PBS, fixed with 3.7% paraformaldehyde for 10 minutes, and permeabilized with 0.2% Triton X-100 (Sigma-Aldrich) in PBS at 37°C for 5 minutes. The cells were incubated for 30 minutes in 10% calf serum in D-PBS to block nonspecific binding of the primary antibody and incubated with the primary antibodies at the dilutions indicated earlier for 45 minutes at 37°C. Cells were then washed in D-PBS and incubated with the secondary antibody for 45 minutes, washed again in D-PBS, counterstained with 4′,6′-diamino-2-phenylindole (DAPI), and mounted with an antifade reagent (Prolong Gold with DAPI; Invitrogen) onto slides for visualization. The cell cultures were visualized with an inverted microscope (Axiocam; Carl Zeiss Meditec, Inc., Dublin, CA) equipped with DAPI, FITC, and Cy3 filter sets. As a negative control, cells grown on a coverslip were stained according to the same procedures and incubated with the secondary antibody only. The lack of staining (data not shown) indicated a high specificity of the primary antibody. 

Perfusion System

Second- or third-passage cells from human AG explants were seeded onto polycarbonate cell culture inserts (12 mm filter membrane diameter, 0.4 μm pore size, 0.6 cm² effective growth area; Millipore, Bedford, MA) at a density of 5 × 10⁵ cells/cm². Cells were seeded at near confluent densities and grown on the filter support for 10 to 14 days to ensure the formation of tight junctions. After 10 to 14 days, cells on filters were inserted into the perfusion system shown in Figure 2 . The cell perfusion system consisted of a modified Ussing chamber, connected through tubing to a pair of fluid reservoirs, one for each side of the perfusion chamber, the downstream reservoir being open to the atmosphere, horizontal, and at the same height as the cell layer. The upstream reservoir held culture medium and provided the hydrostatic pressure gradient. To maintain viability of the cells over the 16 to 20 hours of perfusion, the system was placed in an incubator at 37°C. 

The pressure decrease across the cell layer was continuously monitored by an in-line pressure transducer (Transpac IV; Abbott Critical Care Systems, North Chicago, IL), interfaced with a data-acquisition computer equipped with software to record pressure in real time (Labview; National Instruments, Austin, TX). The setup, calibration, and calculation of pressure values were described by Rivera et al. ²⁴ Care is taken during setup so that the cellular layer is never exposed to a decrease in pressure to less than atmospheric or greater than the target experimental pressure. In addition to monitoring the cell layer pressure, a second pressure transducer monitored atmospheric pressure during perfusion. The atmospheric pressure was subtracted from the cell layer pressure measurements to remove the atmospheric back pressure to obtain the resultant transcellular pressure decrease across the cell layer plus filter. 

To determine the volumetric flow rate across the cell layer, we collected the perfusate from the sampling port on the downstream side of the cell layer. The collection tube was empty at the beginning of the experiment and was left open to the atmospheric environment in the incubator during the experiment. At the conclusion of the perfusion, this volume was measured and then divided by the time of the run to get an average volumetric flow rate. To measure evaporation from this open collection tube, we placed a known amount of medium (2 mL) in an open tube, next to and identical with the perfusate collection tube. After the run, this volume was remeasured and the evaporation rate calculated. The result was then added to the average volumetric flow rate to correct for evaporation. 

To test whether human AG cells in vitro could accurately replicate the unidirectional flow of CSF in vivo, we perfused cells on filter membranes in both the physiological direction, mimicking normal CSF flow in vivo, and the opposite direction, simulating conditions of increased pressures in the venous sinuses. For cells perfused in the physiological direction, the fluid first flowed through the filter membrane, then the basal side of the cell layer, and finally the apical cell membrane. This path is referred to as the basal-to-apical (B→A) direction. For cells perfused in the nonphysiological apical-to-basal (A→B) direction, the fluid first passed through the apical cell membrane, then the basal membrane, and finally the filter membrane. 

Calculation of Cellular Hydraulic Conductivity

The hydraulic conductivity (L _p) is a measure of the amount of fluid that passes through the cell layer or membrane and depends on the surface area of the confluent cell layer on the filter, volumetric flow rate, and pressure decrease. The resistance of the cell layer to flow is inversely related to the cellular hydraulic conductivity. As the resistance across the cell layer increases the cellular hydraulic conductivity decreases. The resistance of the cell layer is also inversely related to the area across which the fluid flows. These variables are related by equation 1 .  

\[L_{\mathrm{p}}{=}\frac{Q}{{\Delta}P{\cdot}A}.\]

To determine the resistance of the cell layer to flow, it is useful to draw parallels between fluid flow and basic electrical circuit theory. In electrical circuit theory,  

\[V{=}IR,\]

where V is the voltage or electromotive force, I is the current, and R is the resistance. The electromotive force is the difference in potential or voltage between two points and is analogous to the pressure difference, ΔP, between the height of the fluid reservoir and the cell layer in the perfusion system that provides the driving force for fluid flow. Current is the rate of flow of electrical charge past a given point and can likewise be compared to the volumetric flow rate, Q, across the cell layer. Therefore, this expression can be rewritten for fluid flow as  

\[{\Delta}P{=}QR.\]

Combining equations 1 and 3gives  

\[R{=}\frac{{\Delta}P}{Q}{=}\frac{1}{L_{\mathrm{p}}A}\]

and  

\[\frac{1}{L_{\mathrm{p}}}{=}\frac{{\Delta}PA}{Q}.\]

In calculating accurately the L _p of the cell layer alone, the resistance of the filter membrane must be accounted for. Although more permeable than the cell layer, it is possible that the filter membrane still has a significant resistance to fluid flow. In accounting for this, the L _p of the empty filter alone is calculated experimentally and accounted for in the electrical circuit analogy. In calculating how the hydraulic conductivity of the filter affects the total resistance, the cells and filters can be considered a pair of resistors in series. Drawing again from electrical circuit theory, the total resistance of resistors in series is given by the sum of the individual resistances  

\[R_{\mathrm{total}}{=}R_{1}{+}R_{2}{+}R_{3}{+}{\ldots}\]

or  

\[R_{\mathrm{cells}{+}\mathrm{filter}}{=}R_{\mathrm{cells}}{+}R_{\mathrm{filter}}\]

and  

\[\frac{1}{L_{\mathrm{p,cells}{+}\mathrm{filter}}}{=}\frac{1}{L_{\mathrm{p,cells}}}{+}\frac{1}{L_{\mathrm{p,filter}}}.\]

Combining equations 5 and 8and solving for L _p,cells gives  

\[L_{\mathrm{p,cells}}{=}\frac{Q_{\mathrm{cells}{+}\mathrm{filter}}}{{\Delta}PA\left(1{-}\frac{Q_{\mathrm{cells}{+}\mathrm{filter}}}{Q_{\mathrm{filter}}}\right)}.\]

From this equation, the cellular hydraulic conductivity can be calculated because the surface area of the confluent cell layer is known and assumed to be equal to the filter. The transcellular pressure drop can be calculated based on the pressure across the cell layer plus filter, and atmospheric pressure (ΔP _{transcellular} = P _{cell layer +filter} − P _atmospheric). The flow rate across the cells and filter, Q _cells+filter, is calculated, and the flow rate across the filter, Q _filter, is measured experimentally by perfusing empty filters without cells and calculating the average volumetric flow rate and L _p across the filter alone. 

Viability Assay

Perfused cell layers were stained with a cell viability assay to determine whether the perfusion conditions caused cell necrosis or apoptosis. Cell viability was assessed with a kit (Live/Dead viability/cytotoxicity kit; Invitrogen) that shows diffuse cytoplasmic green fluorescence for live cells or localized nuclear red fluorescence for dead or dying cells. 

After perfusion, cells on filter were removed from the perfusion chamber, washed three times with D-PBS and incubated with a solution of 2 μM calcein and 4 μM ethidium homodimer in PBS for 40 minutes. The filter membrane was then excised from its casing, and the cells on the filter were inverted onto a microscope slide, coverslipped, and viewed on an inverted microscope (Axiocam; Carl Zeiss Meditec, Inc.) equipped with a FITC long-pass filter. 

Electron Microscopy

To gain insight into the mechanism by which AG cells regulate flow, perfused cells on filters were fixed under pressure, and their cellular ultrastructure was examined by EM. After perfusion, the upstream chamber was flushed and replaced with a 3% glutaraldehyde solution in 0.1 M phosphate buffer. The reservoir height was unchanged so that cells were fixed at the same pressure as perfusion. Cells on filters were fixed at pressure for 45 minutes and then removed from the perfusion chamber and fixed for an additional 30 minutes. For comparing the effects of the direction of perfusion, cells were perfused and fixed under B→A pressure and also under A→B pressure. 

After glutaraldehyde fixation, the cells were postfixed in 1% osmium tetroxide in 0.1 M phosphate buffer containing 0.1 M sucrose (pH 7.4) for 1 hour. They were then rinsed, stained (2% uranyl acetate for 30 minutes), and dehydrated in a graded series of ethanol. Samples were incubated in two changes of hydroxypropyl methacrylate and rinsed five times in 100% resin (Polybed 816), then embedded in resin (Polybed 812; both from Polyscience, Warrington, PA) in a round silicon mold and polymerized at 60°C for 24 hours. Embedded samples were sectioned on an ultramicrotome (EM UC6; Leica) at 70 nm. Sections were picked up on 100-mesh formvar-coated grids stained with 2% uranyl acetate and Reynolds lead citrate and finally examined under a transmission electron microscope (CM12; Phillips, Eindhoven, The Netherlands) at 80 kV. 

Results

Cell Culture

AGs were collected at autopsy from the superior sagittal sinus and lateral lacunae (Fig. 3A)and explanted into culture plates. Cell migration from human AG explants was seen within 7 to 10 days (Fig. 3B) . The AG cap cells in culture grew in monolayers, exhibited a polygonal morphology, and packed in densely, assuming the cobblestone-like appearance characteristic of epithelial cell types (Fig. 3C) . 

Identification of Phenotype

Second-passage cultures were immunoreactive to the anti-human cytokeratin antibody (AE1/AE3), which recognizes a wide range of human cytokeratins (Moll’s designation 1–8, 10, 13–16, 19). ² A subpopulation of cells expressed cytokeratin intermediate filaments in a perinuclear pattern, with long filaments surrounding the nucleus in a basketlike structure (Fig. 4A) . AG cells expressed the desmosomal junctional proteins desmoplakin-1 and -2 at cell borders (Fig. 4B) , where desmosomes mediate cell-cell attachments. Finally, human AG cells were immunoreactive to the anti-ZO-1 antibody expressed at tight junctions (Fig. 4C) . 

Perfusion Results

Table 1provides a summary of the average pressure, average volumetric flow rate, and average cellular hydraulic conductivity for the perfusion runs in both the physiological and nonphysiological directions. Figure 5shows the average cellular hydraulic conductivity plotted versus the average transcellular pressure drop across the cell layer and filter. The cellular hydraulic conductivity averaged over all perfusions in the A→B direction (six total) was 0.28 ± 0.16 μL/min per mm Hg/cm², with an average transcellular pressure drop across the cell layer and filter of 3.33 ± 0.16 mm Hg. The cellular hydraulic conductivity averaged over all perfusions in the B→A direction (17 total) was 4.50 ± 0.53 μL/min per mm Hg/cm² with an average transcellular pressure drop across the cell layer and filter of 3.15 ± 0.08 mm Hg. A Student’s t-test was performed taking P < 0.05 as statistically significant. For the average cellular hydraulic conductivity, P = 2.09 × 10⁻⁷ indicating a statistically significant difference in average cellular hydraulic conductivity between the A→B and B→A perfusions. The average transcellular pressure drop across the cell layers in both directions was statistically equivalent. In comparison, the average hydraulic conductivity of the empty conditioned filters (no cells) in the A→B direction (three total) was 4.28 ± 0.29 μL/min per mm Hg/cm², with an average transcellular drop in pressure across the filter of 3.85 ± 0.5 mm Hg and average flow rate of 9.95 ± 1.9 μL/min. 

Viability

Figure 6shows representative images of cells stained with the viability assay (Live/Dead; Invitrogen). Figures 6A and 6Bshow two different filters perfused in the A→B direction and Figures 6C and 6Dshow two filters perfused in the B→A direction. Although the images are in black and white, it is still possible to see that the filters perfused in the A→B direction show a larger number of dead and dying cells by fluorescence localized to the cell nuclei. Filters perfused in the B→A direction show a more diffuse fluorescence throughout the cytoplasm, indicating that most of the cells on the filter are still viable. 

Electron Microscopy

EM of AG cells on filter membranes fixed under pressure are shown in Figures 7 8 and 9 . Figure 7shows a highly magnified (10,000×) view of AG cells fixed under B→A pressure. This image depicts several overlapping cell processes, with arrows and arrowheads indicating cell-cell junctions. The arrow denotes a desmosomal junction between overlapping cells, and the arrowheads show a tight junction between cells. These images demonstrate that human AG cells grown on filter supports express some of the cell-cell junctional complexes that have been identified in human AG tissue. Figure 8shows AG cells on a filter, fixed under B→A pressure with several overlapping cells on the filter membranes. Visible is a close apposition of the cells to the underlying filter as well as cellular extension into the pores of the filter. These observations suggest that the cells are attached to the filter membrane and are not dislodged when perfused under physiological B→A pressure. 

In addition, in Figure 8several large intracellular vacuoles (stars) are visible. A magnified view in Figure 8Bshows vacuoles self-contained within the cell cytoplasm (open star) along with micropinocytotic vesicles of fluid (arrowheads). Also shown in Figure 8Aare several large extracellular cisternal spaces between overlapping cells (C). These spaces may represent a paracellular transport between adjacent cells. In cells on filter fixed under A→B pressure and shown in Figure 9 , there was still close apposition of the cell to the underlying filter and cellular material within the filter’s pores. However, cells shown in Figure 9that were fixed under A→B pressure did not show any large vacuoles and only a few small extracellular spaces between cells. 

Discussion

Hydraulic Conductivity of AG Cells under Physiological and Nonphysiological Conditions

In normal humans, the CSF pressure is approximately 3 to 5 mm Hg greater than the pressure in the venous sinuses, creating a driving force for the outflow of CSF from the subarachnoid space, across the basal membrane of the cells lining the AGs, and finally across the apical membrane into the venous sinuses. ^{25 26} Normal flow of CSF across the AG cells is basal to apical (B→A). Were there an increase in venous pressure as with thrombosis or transverse sinus wall collapse, the CSF driving force would be diminished or reversed. In such a case, even though CSF outflow would be impaired, venous blood would not flow into the CSF, further supporting the suggestion that physiologic CSF flow is unidirectional. 

After demonstrating that human AG cells in vitro express many of the intercellular junctional proteins necessary to regulate fluid flow, it was also necessary to demonstrate that these cells could mimic some of the physiological function of the AG cells in vivo. A major step in this process was to show that human AG cells in vitro could replicate the unidirectional flow of CSF found in vivo. 

To test the behavior of the cells under similar conditions, cells were perfused at pressure decreases of 3 mm Hg in both the physiological (B→A) and nonphysiological (A→B) directions. Six cell layers were perfused in the nonphysiological direction and 10 in the physiological direction. Three of the six had no flow rate across the cell layer (Q _cells+filter = 0) and mimic the in vivo response of these cells. Cells perfused in the physiological direction (B→A) had a cellular hydraulic conductivity at least an order of magnitude greater than the nonphysiologic cellular hydraulic conductivity. Calculations of the average cellular hydraulic conductivity showed a statistically significant difference between cells perfused physiologically versus nonphysiologically (P = 4.08 × 10⁻⁵) despite perfusion at the same pressure (P = 0.31). The results of these perfusion studies indicate that human AG cells in vitro simulate the unidirectional flow of CSF. 

The assumption that the surface area of the insert is equal to the surface area of the cells is necessary for the simplification of the L _p calculation. This convention is common in many different cell culture models and has been used extensively in transport and permeability models of intestinal epithelial cells (CACO-2 models) as well as blood-brain barrier endothelial cells grown on permeable culture inserts. ²⁷ In perfusion models, more closely related to the one presented here, this convention has also been used to calculate L _p of perfused trabecular meshwork and Schlemm’s canal endothelial cells. ^{28 29 30 31 32}  

This system represents only one isolated cell type involved in the outflow of CSF through the AGs. The arachnoid membrane and arachnoid granulations are complex tissues consisting of multiple cell types supported by protein structures. As a first effort to try to understand the mechanism of CSF outflow through these tissues, we have isolated only one cell type and are trying to understand how it functions in an isolated environment. Interactions between other cell types, ECM proteins, and humoral factors most likely occur, and at this point are not being considered in this model. This includes any interaction with the connective tissue matrix core that includes fibroblasts interspersed in a connective tissue matrix that likely includes elastin, collagen, fibronectin, and fibrin, and may have implications when considering the dynamic nature of this system. 

In vivo, the AGs are in a curved arrangement overlying a central core composed of fibroblasts and a connective tissue matrix. The arrangement of AG cells on this connective tissue matrix likely provides a resiliency to the AG cells allowing them to withstand mechanical forces that would break apart an inflexible cell layer. It is not likely, however, that this connective tissue core provides a functional resistance to fluid flow. The arrangement of this connective tissue matrix is loose and porous, allowing fluid and proteins to percolate through to the barrier of AG cells. The mature cell-cell junctions of the AG cells lining the granulation provide the major resistance to fluid passage. By isolating the AG cells in vitro, we believe we have isolated the cells that represent the major resistance to outflow. 

In addition, because the AG cells have properties similar to fibroblasts, they are capable of synthesizing some of the ECM proteins found within the underlying connective tissue matrix including collagen IV, fibronectin, and laminin. ³³ This property of AG cells allows them to lay down ECM in culture as well, imitating their in vivo microenvironment. This curved structural relationship between the AG cells and their in vivo environment is not modeled in our system; however, because we are focusing on the microscopic inter- and extracellular scale, these differences should not significantly affect the results. We realize these assumptions do not represent the complete physiologic environment, but believe that they are a critical step in fully understanding the functionality of the arachnoid granulations as completely as possible. 

In vivo, the basal cell side is physiologically in contact with CSF which has a low concentration of serum proteins, whereas our experimental setup uses media with serum. The in vitro culture of arachnoidal cells may alter their expression levels of some proteins compared with levels as the cells express in vivo; however, we have shown that the expression of the critical junctional proteins is maintained. ¹⁶ In addition, the passage number of the cultured cells is monitored closely and maintained at the optimum passage for in vivo protein expression. 

Postperfusion Analysis of Cell Viability

To evaluate the difference in hydraulic conductivity between the cells perfused in each direction, the viability of the cell layer was studied immediately after perfusion. Cell viability was assessed with an assay of intracellular esterase activity and membrane integrity. More dead/dying cells were observed in the cells perfused in the nonphysiological (A→B) direction shown by dye located at the cell nucleus and fluorescing red. These results were demonstrated in tissues from multiple donors, perfused at different times. Figures 6A and 6Cshows cells from the same tissue donor, seeded onto filters at the same time, and perfused on the same day. In this case, the cells perfused nonphysiologically show a much larger number of dead cells than does the accompanying filter perfused physiologically. 

There were no variables other than perfusion pressure and direction. Cells were typically perfused in a large incubator maintained at 37°C. The levels of CO₂ in the incubator were not controlled, but this is not likely to have had an effect on the cell’s health because the time of the run was <24 hours, and there was no visible change in phenol red in the perfusion media, which confirms a stable pH. Nonperfused cells on the filter were incubated for equivalent times as an additional control. In these experiments, sets of three sister cultures were seeded on filters and grown to confluence. One filter was perfused physiologically, one filter was perfused nonphysiologically, and the third was merely placed in medium in the incubator for the duration of the perfusion experiment without a hydrostatic pressure drop (static culture). At the conclusion, all cells perfused or nonperfused and static cultures were examined for cell viability. Cells that were not perfused (static) and cells perfused physiologically, showed predominantly live, healthy cells, whereas those cells perfused nonphysiologically had many dead or dying cells. 

Postperfusion Electron Microscopy

Cells on filter were examined after perfusion using electron microscopy. At the conclusion of perfusion runs, in both the physiologic and nonphysiologic directions, the growth medium in the perfusion chamber on the upstream side was flushed and replaced with fixative. Cells were fixed at the same pressure as when perfused, to maintain the same ultrastructure. Analysis of these cells on filter using TEM revealed several important details. The close apposition of the cells to the filter, as well as the presence of cellular material within the pores of the filter, indicated that these cells were strongly attached to the filter after (B→A) perfusion. During physiological (B→A) perfusion the pressure driving force acted to push the cells off of the filter. Because the cells remained attached to the filter in both cases, the flow rates during the perfusion were most likely a result of flow through or between cells and not of cells being “blown off” of the filter. 

EM images suggest mechanisms by which these cells allow the passage of fluid. From these images, it was possible to compare the mechanisms of transport with those that had been postulated from animal and human electron microscopy studies of whole AG tissue. The mechanism of fluid transport across AGs has been investigated extensively in different species, including rats, ³⁴ dogs, ³⁵ sheep, ^{36 37 38 39} and monkeys. ^{22 23 40 41} From these investigations, several mechanisms have been proposed, including micropinocytotic vesicles, paracellular transport by the widening of the spaces between overlapping endothelial cells, ^{38 39 40} and large intracellular vacuoles fusing to form tortuous transendothelial channels. ^{22 42 43}  

Showing such mechanisms in human AGs has been difficult, probably because of the difficulty of using tracer molecules and fixing tissue under pressure in situ. Nevertheless, Kida et al. ¹¹ and Yamashima. ^{19 20} have attempted to explain the transport mechanisms in human AGs. Their studies examined human AGs using light and electron microscopy. In human subjects with subarachnoid hemorrhage, red blood cells collect in the AGs and cause proliferation of AG cap cells and chronic hydrocephalus. ⁴⁴ Massicotte and Del Bigio ⁴⁴ showed that AGs form an intact arachnoid cell layer directly in contact with the venous lumen, suggesting that the passage of CSF across the arachnoid cell layer constitutes the last barrier between the subarachnoid space and venous lumen. Yamashima ^{19 20} and Kida et al. ¹¹ showed intracellular vacuoles and micropinocytotic vesicles in cells lining the AGs, as well as large extracellular cisternal spaces between the apical AG cells. They suggested two mechanisms for the transport of CSF across human AGs: transcellular transport through pinocytosis and vacuolization and paracellular transport through extracellular cisterns. 

Based on electron microscopic examination of human AG cells on filters fixed under pressure (B→A) after perfusion, our consistent findings suggest the same transport mechanisms suggested by Yamashima. ^{19 20} Figure 8shows AG cells fixed under pressure and reveals the presence of large intracellular vacuoles, suggesting a transcellular pathway for fluid flow. Large extracellular spaces have also been shown between overlapping AG cells comparable to the extracellular cisterns described by Yamashima. These spaces may represent paracellular routes of transport and may involve transient alterations of intercellular tight junctions. 

Conclusion

This article reports our initial efforts to model the outflow of CSF across human arachnoid granulations. We have demonstrated that AG cells in vitro show a statistically significant increase in flow rate and cellular hydraulic conductivity when perfused in the physiologic versus the nonphysiologic direction under normal pressure. The results of these perfusion studies suggest that this in vitro model of the AGs can accurately replicate the unidirectional flow of CSF in vivo. 

 Supported by The Ohio Lions, The Davis Medical Research Grant, and the Theodore Lyman Moor Graduate Student Research Fund.

 Submitted for publication July 18, 2005; revised January 21 and February 23, 2006; accepted May 23, 2006.

 Disclosure: D.M. Grzybowski, None; D.W. Holman, None; S.E. Katz, None; M. Lubow, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Cerebrospinal fluid is produced by the choroid plexus in the lateral, third, and fourth ventricles where it circulates to the subarachnoid space and eventually returns to the venous blood via the arachnoid granulations. (Illustration reprinted, with permission, from Fishman, RA. Cerebrospinal Fluid in Diseases of the Nervous System. 2nd ed. Philadelphia; WB Saunders Co., Harcourt Brace Jovanovich, Inc.; 1992.) ¹ (A) AGs are projections of the arachnoid membrane into the dural venous sinuses and lateral lacunae. (Illustration reprinted, with permission, from Warwick R, Williams PL. Grey’s Anatomy. Philadelphia: WB Saunders; 1973.) ³ (B).

Figure 2.

 

AG cells were perfused in a modified Ussing chamber. Cells were seeded onto filter membranes and perfused in the physiological (B→A) or nonphysiological (A→B) direction (see inset). A hydrostatic pressure head was created by adjusting the height of the fluid reservoir relative to the cell layer. An in-line strain gauge pressure transducer interfaced with a data acquisition computer allows for real-time monitoring of the pressure decrease across the cell layer. Volumetric flow rate was measured by collecting the perfusate from a downstream sampling port.

Figure 3.

 

Human arachnoid granulations were collected at autopsy within 24 hours after death (A). Individual AGs were excised with microsurgical scissors and explanted to fibronectin-coated culture dishes. Cell growth from the explant was seen within 7 to 10 days (B). AG cells in culture showed polygonal cell morphology and when confluent and packed in densely, assuming the cobblestone-like appearance typical of epithelial cells in culture (C).

Figure 4.

 

Immunocytochemical methods were used to identify phenotype. AG cells were grown on fibronectin-coated coverslips and labeled with antibodies to broad-spectrum cytokeratins (A). AG cells were also labeled with antibodies to the desmosomal plaque protein desmoplakin-1 and -2 (B) and the tight junction protein ZO-1 (C). Positive expression of these proteins was also seen at cell-cell borders. Primary antibody expression was visualized by labeling the cells with a secondary antibody conjugated to a fluorescent molecule. Secondary antibodies were either anti-IgG FITC (green) or anti-IgG Alexa Fluor 555 (red).

Table 1.

 

Summary of Results for AG Cells Perfused in the Physiological (B→A) and Nonphysiological (A→B) Directions

	Pressure (mm Hg)	Flow Rate (μL/min)	Average Cellular L p (μL/min per mmHg/cm2)
Average A → B	3.33 ± 0.16	0.52 ± 0.29	0.28 ± 0.16
Average B → A	3.15 ± 0.08	4.30 ± 0.25	4.50 ± 0.53
P-value	0.35	7.40E-08	2.09E-07

 Data are average cellular hydraulic conductivity (L _p).

Figure 5.

 

Plot of average cellular hydraulic conductivity versus average pressure for AG cells perfused in the B→A or A→B directions.

Figure 6.

 

After perfusion cells were examined with a cell viability assay. Cells perfused nonphysiologically (A, B) showed a larger number of dead and dying cells as evidenced by the greater amount of dye that located to the nuclei. Cells perfused physiologically (C, D) showed more diffuse cytoplasmic staining, indicating a larger number of viable cells.

Figure 7.

 

High-magnification view of AG cells on a filter that were perfused and fixed under B→A pressure. Several cell-cell junctions were visible, including a tight junction (arrowhead) and a desmosome (arrow). Magnification, ×20,000. Scale bar, 1 μm.

Figure 8.

 

AG cells were perfused and fixed under 3-mm Hg B→A pressure. Electron microscopic analysis showed a very close apposition between the AG cells and the underlying filter membrane (A, arrowheads) as well as the presence of what appears to be cellular material within the filter pores (A, arrows). Taken together, these observations suggest that the cells were not dislodged from the filter during perfusion. Also visible were large extracellular cisternal spaces (C) between overlapping AG cells. It is possible that these spaces between cells represent a pathway for paracellular transport of fluid. Finally, several large vacuoles (★) within the cytoplasm were visible. Closer examination of an individual vacuole (B, ☆) shows this structure to be self-contained with the cytoplasm of a cell process. These vacuoles may represent a trans-cellular pathway for fluid flow. Also visible in (B) are cell junctions (arrows) between overlapping cell processes and micropinocytotic vesicles of fluid (arrowheads). Scale bars, 1 μm.

Figure 9.

 

AG cells were perfused and fixed under 3-mm Hg pressure in the A→B direction. Again, there was a close apposition between the cells and filter, suggesting that the cells remain tightly attached during perfusion and fixation. Unlike the cells perfused physiologically, these cells did not show any intracellular vacuoles or large extracellular spaces. However, some smaller spaces between cells were visible (arrows), although further work is necessary to determine whether the spaces are an artifact of fixation. Scale bar, 10 μm.

The authors thank Jennifer L. Hodges (The Ohio Lions Eye Research Facility), and Kathy Wolken (The Ohio State University Campus Microscopy and Imaging Facility) for technical assistance.