Permeability of Epithelial/Endothelial Barriers in Transwells and Microﬂuidic Bilayer Devices

micromachines

Article Permeability of Epithelial/Endothelial Barriers in Transwells and Microﬂuidic Bilayer Devices

Timothy S. Frost 1,* , Linan Jiang 2, Ronald M. Lynch 1,3 and Yitshak Zohar 1,2 1 Department of Biomedical Engineering, University of Arizona, Tucson, AZ 85721, USA 2 Department of Aerospace and Mechanical Engineering, University of Arizona, Tucson, AZ 85721, USA 3 Department of Physiology, University of Arizona, Tucson, AZ 85721, USA * Correspondence: [email protected]

Received: 25 June 2019; Accepted: 11 August 2019; Published: 13 August 2019

Abstract: Lung-on-a-chip (LoC) models hold the potential to rapidly change the landscape for pulmonary drug screening and therapy, giving patients more advanced and less invasive treatment options. Understanding the drug absorption in these microphysiological systems, modeling the lung-blood barrier is essential for increasing the role of the organ-on-a-chip technology in drug development. In this work, epithelial/endothelial barrier tissue interfaces were established in microfluidic bilayer devices and transwells, with porous membranes, for permeability characterization. The effect of shear stress on the molecular transport was assessed using known paracellular and transcellular biomarkers. The permeability of porous membranes without cells, in both models, is inversely proportional to the molecular size due to its diffusivity. Paracellular transport, between epithelial/endothelial cell junctions, of large molecules such as transferrin, as well as transcellular transport, through cell lacking required active transporters, of molecules such as dextrans, is negligible. When subjected to shear stress, paracellular transport of intermediate-size molecules such as dextran was enhanced in microfluidic devices when compared to transwells. Similarly, shear stress enhances paracellular transport of small molecules such as Lucifer yellow, but its effect on transcellular transport is not clear. The results highlight the important role that LoC can play in drug absorption studies to accelerate pulmonary drug development.

Keywords: barrier permeability; epithelial–endothelial interface; paracellular/transcellular transport

1. Introduction Drug discovery is becoming slower and more expensive over time, a trend referred to as Eroom’s law (Moore’s law spelled backward), despite major progress in technologies such as high-throughput screening and computational drug design [1]. Following this trend, the number of new drugs produced per billion US $ has halved every decade since 1950 [1,2]. The increasing diﬃculty and cost of developing a new compound can be traced back to current methods of new compound screening. In vitro cell cultures and animal models are the major means used to select promising drugs for clinical trials. Experience has shown that these models are poor predictors for compound success in human clinical trials. In vitro cell models in particular have had a limited impact, since they are overly simple failing to recreate the complex tissue microenvironment in vivo, especially the interactions between multiple cell types [3,4]. Animal models, while presenting in vivo tissue-level complexity, fail to provide the proper tissue microenvironment because animals do not possess the same anatomy or physiology as humans; consequently, these models have less than 8% successful translation to therapies in some cancer trials [5,6]. These serious shortcomings underline the critical need to develop new in vitro biomimetic systems that better represent the in vivo human physiological conditions in eﬀort of hastening medical innovation [7].

Micromachines 2019, 10, 533; doi:10.3390/mi10080533 www.mdpi.com/journal/micromachines Micromachines 2019, 10, 533 2 of 18

Many pharmaceutical companies are turning to microfluidic devices as a means to streamline the pre-clinical phases of new compound screening and development [8,9]. Among this new generation of microfluidic devices are microphysiological systems (MPSs), often called organs-on-chips, which have begun to add additional insights as well as increased physiological accuracy to the screening of new compounds [10]. MPSs are in vitro models that capture important aspects of in vivo organ function through the use of specialized culture microenvironments [11]. These microfluidic systems, incorporating small plastic devices, are designed to model certain human organs to trigger more accurate cellular responses [12,13]. Indeed, MPS technology has begun to accelerate medical research across several fields. MPSs are used for modeling of diseases, such as cancer, which is emerging as a prominent application driving development of complex systems with higher-order tissue functions [11]. In drug discovery, these models are utilized to perform high-throughput assays to assess drug viability, optimizing clinical trials, and potentially reducing R&D costs to develop new compounds [14]. MPSs can be used to evaluate the entire life cycle of a compound inside a human body by culturing different human cell types in different chambers on the chip [15,16], and are often used to provide a unique window into the behavior of a tissue that cannot be observed in detail in vivo [17–19]. MPSs are particularly useful in exploring the dynamic permeability across barrier-tissue interfaces such as blood–brain, vascular-endothelial, and lung-blood barrier. Tissue barrier permeability is essential in the selective transport across the interface, is maintained through tight cell–cell junctions, and is controlled by growth factors, cytokines, and other stress related molecules [20]. Understanding the drug life cycle in the body begins with the absorption of a drug into the bloodstream. The intricate process of cellular transport is highly tailored to each organ and is dependent on the cell types being used as well as the expression of any proteins involved in the transport of a drug into the blood stream. Pulmonary drug delivery has recently become an attractive route for medical therapy as it is both minimally invasive and able to quickly interact with a large blood volume [21]. In order to study pulmonary absorption, static transwell inserts have been used as a standard assay for many years [22]. However, these static conditions lack the fluid-flow induced shear stress to mimic blood or air flow in human body. Media flow is an essential characteristic of lung-on-a-chip (LoC) devices allowing the introduction of mechanical forces in an effort to provide a more physiologically relevant model system [23]. Transport of bio-species ranging from small molecules such as water and ions, to large proteins, and even whole cells is significantly more complex in systems with media flow. On top of diffusion, convection is added as another mass transport mechanism. Furthermore, the shear stress exerted on the cellular layer may modify the tissue barrier properties through alteration of inter-cellular junctions and cellular interactions with the extracellular matrix [24–26]. With the growing interest in microfluidic pulmonary models to study drug absorption, a thorough evaluation of the lung–blood barrier is needed. Recently, some permeability measurements were reported in a small airway-on-a-chip model [23], and vascular-endothelial barrier permeability was characterized using a biomimetic microfluidic blood vessel model [27]. To date, little has been done to analyze the transport of larger molecules within an organ-on-a-chip device. In this work we utilize microfluidic membrane bilayer devices, typically used for LoC applications, to form epithelial/endothelial cell barriers. The permeability of various molecules, via different transport pathways, is characterized with and without epithelial/endothelial monolayers or bilayers, and the results are compared with corresponding data collected in traditional transwell inserts. In the current work, both epithelial and endothelial cell layers are submerged in liquid media although, for human lung models, an air-liquid interface is an important feature affecting molecular diffusion in the epithelial side of the interface and subsequently the permeability.

2. Methods Two experimental setups have been utilized in this work: (i) commercially available transwell inserts under static condition to eliminate the effect of shear stress, and (ii) in-house fabricated microfluidic devices under fluid flow accounting for flow-induced shear stress. Epithelial and endothelial cell layers were cultured in the transwells inserts and microfluidic devices to characterize Micromachines 2019, 10, 533 3 of 18 the permeability of various bio-species based on their concentration measurements. The experimental error analysis associated with these measurements has been reported elsewhere [28].

2.1. Transwell Insert Models Micromachines 2019, 10, x FOR PEER REVIEW 3 of 18 Either epithelial or endothelial cells were cultured in the top compartments (often called apical or luminal)theof permeability commercially-available of various bio-species 12-well based on transwells their concentration (Corning, measurements. Corning, NY,The experimental USA) on polyester membraneserror with analysis 0.4 associatedµm diameter with these pores measur andements 0.5% porosity.has been reported Once theelsewhere cultured [28]. cell monolayers had reached confluency, media in the top compartments were replaced with 500 µL solutions of cell media 2.1. Transwell Insert Models mixed with various fluorescently-labeled molecules at pre-determined initial concentrations. After filling the bottomEither compartments epithelial or endothelial (often calledcells were basolateral cultured in orthe abluminal) top compartments with 1500(often µcalledL fresh apical cell media, or luminal) of commercially-available 12-well transwells (Corning, Corning, NY, USA) on polyester the transwellsmembranes were with placed 0.4 µm in diameter an incubator pores and for 0.5% two porosity. hours. Once Then, the cultured liquid cell samples monolayers of 100 hadµ L were removedreached from both confluency, compartments, media in the 5 fromtop compartments the top and were 15 fromreplaced the with bottom, 500 µL and solutions the corresponding of cell fluorescentmedia intensities mixed with of allvarious samples fluorescently-labeled were measured molecules using at a pre-determined BioTek Synergy initial 2 Plateconcentrations. Reader (BioTek, Winooski,After VT, filling USA). the Molecular bottom compartments concentrations (often incalled the ba collectedsolateral or samples abluminal) were with then 1500 estimatedµL fresh cell based on media, the transwells were placed in an incubator for two hours. Then, liquid samples of 100 µL were the fluorescentremoved intensity from both measurements compartments, 5 usingfrom the pre-established top and 15 from calibrationthe bottom, and curves. the corresponding The average value among thefluorescent samples intensities collected offrom all samples each were compartment measured using was a taken BioTek as Synergy the molecular 2 Plate Reader concentration. (BioTek, The experimentsWinooski, were VT, repeated USA). Molecular without concentrations cell monolayers in the in collected the transwells. samples were Both then the estimated initial concentrationbased and incubationon the timefluorescent were intensity determined measurements to keep the using concentration pre-established measurements calibration curves. within The theaverage linear range of the calibrationvalue among curves the samples to avoid collected signal from saturation. each compartment was taken as the molecular concentration. The experiments were repeated without cell monolayers in the transwells. Both the initial concentration and incubation time were determined to keep the concentration measurements within 2.2. Microfluidic Devices the linear range of the calibration curves to avoid signal saturation. Microfluidic devices were fabricated using soft-lithography techniques. Polydimethylsiloxane 2.2. Microfluidic Devices (PDMS, Sylgard 184 Silicone Elastomer, Dow Corning, Midland, MI, USA) resin was poured over an Microfluidic devices were fabricated using soft-lithography techniques. Polydimethylsiloxane aluminum mold and allowed to cure for 24 h at 55 ◦C. The cured PDMS substrate with grooves, each (PDMS, Sylgard 184 Silicone Elastomer, Dow Corning, Midland, MI, USA) resin was poured over an about 35 mm long and cross-section area of 500 µm° 1000 µm, was peeled off the aluminum mold. aluminum mold and allowed to cure for 24 h at 55 C.× The cured PDMS substrate with grooves, each A polyesterabout track 35 mm etched long membrane,and cross-section about area20 of µ500m µm thick × 1000 with µm, 1% was porosity peeled off and the 0.8aluminumµm diameter mold. pores, was plasmaA polyester treated track and soakedetched membrane, in a solution about of20 5%µm thick (v/v) with (3-Aminopropyl)triethoxysilane 1% porosity and 0.8 µm diameter pores, (APTES, MP was plasma treated and soaked in a solution of 5% (v/v) (3-Aminopropyl)triethoxysilane (APTES, Biomedicals, Solon, OH, USA) in water at 55 ◦C for 1 hour for leakage-free sealing [29]. While the MP Biomedicals, Solon, OH, USA) in water at 55°C for 1 hour for leakage-free sealing [29]. While the membrane was drying, pairs of microchannel grooves were oxygen-plasma treated for 60 s at 1000 mTorr membrane was drying, pairs of microchannel grooves were oxygen-plasma treated for 60 s at 1000 (Harrick Plasma,mTorr (Harrick Ithaca, Plasma, NY, USA). Ithaca, Each NY, USA). pair ofEach PDMS pair of microchannels, PDMS microchannels, coated coated to prevent to prevent small small molecule absorptionmolecule [30], was absorption bonded [30], together was withbonded a membranetogether with sandwiched a membrane between sandwiched the overlappingbetween the channel segmentsoverlapping about 20 mm channel in length. segments Once about firm 20 mm bonding in length. had Once been firm achieved,bonding had tubing been achieved, adapters tubing were placed over the punchedadapters were holes placed to serve over the as punched inlet/outlet holes connectorsto serve as inlet/outlet to the external connectors fluid to the handling external fluid system. An handling system. An image of a fabricated device along with a schematic of a device operation is image of ashown fabricated in Figure device 1. along with a schematic of a device operation is shown in Figure1.

(a) (b)

Figure 1. FigureA microfluidic 1. A microfluidic bilayer bilayer device: device: (a )(a A) A photograph photograph of of a afabricated fabricated and andpackaged packaged microfluidic microfluidic bilayer device with blue dye in the top and red dye in the bottom microchannel, and (b) a schematic bilayer device with blue dye in the top and red dye in the bottom microchannel, and (b) a schematic of of a microfluidic bilayer device in operation; the top epithelial microchannel is separated from the a microfluidicbottom bilayer endothelial device microchannel in operation; channel the by top a poro epithelialus membrane. microchannel A syringe is pump separated is used from to drive the bottom endothelialcell microchannel medium mixed channel with fluorescent by a porous molecular membrane. markers A through syringe the pump top ischannel used toand drive only cellcell medium mixed with fluorescent molecular markers through the top channel and only cell medium through the

bottom channel at equal rates; the ﬂuorescent markers are transported across the membrane, from the top to the bottom microchannel and downstream along both microchannels. Micromachines 2019, 10, 533 4 of 18

2.3. Cell Cultures Epithelial and endothelial cell lines were chosen to represent human lung–blood barrier interface. The epithelium was composed of carcinoma alveolar epithelial cells (A549; ATCC Manassas, VA, USA), cultured in RPMI media supplemented with 10% fetal bovine serum (FBS, Gibco, Waltham, MA, USA) (v/v) and 1% penicillin streptomycin (Sigma Aldrich, St. Louis, MO, USA). The A549 cell line has characteristic features of pulmonary epithelium, which has a role in the oxidative metabolism of drugs in the lung. This cell line has indeed been extensively utilized in toxicology studies, including its properties on permeable supports, because of its potential target for drug delivery of macro-molecules [31].The endothelium was composed of human umbilical vein cells (HUVEC; ATCC Manassas, VA, USA), cultured in F-12K media (Kaighn’s modification, Gibco, Waltham, MA, USA) with 10% FBS, 1% penicillin streptomycin, 50 µg/mL endothelial cell growth supplement (ECGS, Sigma Aldrich, St. Louis, MO, USA) from bovine neural tissue, and 100 µg/mL heparin (Sigma Aldrich, St. Louis, MO, USA). HUVEC cells were selected for this characterization phase of LoC models, since they are most commonly utilized to represent an endothelium based on human derived primary endothelial cells [32]. The cell passage number of the A549 and HUVEC culture was 40–60 and 20–30, respectively. Cell media inside the transwells and microfluidic devices was replenished every 48 h. All cell cultures were maintained at 37 ◦C, 5% CO2, 95% air, and 100% relative humidity. Epithelial and endothelial cells were seeded in either transwells or microfluidic devices at a density of 2 106 cells/mL. HUVEC cell attachment and proliferation required membrane surface coating × with collagen. A 200 µg/mL type 1 collagen solution (Gibco, Waltham, MA, USA) was prepared in 0.01 M acetic acid and perfused across the membrane surface. After one hour of curing at an ambient temperature, the collagen solution was replaced with 1x HBSS followed by 30 min incubation time in the cell culture environment to adjust the system pH to a biocompatible level. The HUVEC media for cells cultured in transwell or microfluidic devices had an elevated ECGS concentration (150 µg/mL) to promote a fully confluent monolayer [33].

2.4. Tracer Molecules Molecules can be transported through confluent cell monolayers via two major routes paracellular and transcellular as illustrated in Figure2. A confluent monolayer refers to cells in culture, which are in contact forming a cohesive sheet of adhering neighboring cells. It is thought that contact inhibition of proliferation is a characteristic of a confluent monolayer, where cells stop proliferating upon contact formation [34]. Adherens junctions (AJs) and tight junctions (TJs) comprise two modes of cell–cell adhesion that provide different functions. While AJs initiate cell–cell contacts and mediate the maintenance of the contacts, TJs regulate the paracellular pathway for the movement of ions and solutes in-between cells [35]. In contrast, transcellular transport involves the passage of solutes through the cells crossing both the apical and basolateral membranes. The transcellular transport includes passive diffusion through ion channels, active carrier mediated transportation such as organic anion/cation transporters, and transcytosis of macromolecules such as transferrin and insulin [36]. Various fluorescently labeled molecules were used to study both molecular pathways in the transwells and in the microfluidic devices. Fluorescently labeled tracer molecules commonly used in permeability studies were chosen as biomarkers for paracellular and transcellular transport. Lucifer yellow (0.44 kDa; Invitrogen, Carlsbad, CA, USA) has previously been used as a small molecule paracellular transport marker [31,37]. However, it should also be noted that a growing body of literature suggests that transcellular passage of Lucifer yellow may also occur via organic anion transporters [38]. Dextrans, non-digestible sugars, were selected to represent typical agents that are transported via the paracellular route through cell tight junctions [27,39]. Expression by epithelial or endothelial cells of membrane receptors mediating transcytosis of dextran has not been reported thus far; however, pinocytosis of dextran was observed [40], in which molecules are captured in vesicles on one side of the cell, drawn across the cell, and ejected on the other side. Since pinocytosis is a very slow process, transcellular transport of dextran Micromachines 2019, 10, 533 5 of 18 is negligible compared to its paracellular transport [41]. To explore potential molecular-size effect, 4-dextran (4 kDa) and 70-dextran (70 kDa) (Invitrogen, Carlsbad, CA, USA) were chosen to represent the intermediate and large-size agents, respectively. FITC-transferrin (78 kDa; Rockland Antibodies, Limerick, PA, USA) was used to represent a typical protein that is transported via the transcellular route, transcytosis, where epithelial A549 and endothelial HUVEC cells uptake the molecules by binding transferrin to its membrane receptor expressed by both cell lines [31,39,42,43]. FITC-transferrin and 70-dextran molecules are about the same in size and, therefore, may share similar paracellular

Micromachinestransport level. 2019, 10, x FOR PEER REVIEW 5 of 18

AJ TJ

FigureFigure 2. AA schematic illustrating two cellular transpor transportt mechanisms: Paracellular Paracellular (green) (green) occurs occurs as as moleculesmolecules didiffuseffuse through through adherens adhere (AJ)ns and(AJ) tightand junctionstight junctions (TJ) between (TJ) cellsbetween because cells of concentrationbecause of concentrationgradient, and gradient, transcellular and (purple) transcellular takes (purple) place when takes cells place uptake when and cells release uptake molecules and release through molecules their throughbodies via their transport bodies proteinsvia transp orort as proteins vesicles. or as vesicles. 2.5. Concentration Measurements in Microfluidic Devices Fluorescently labeled tracer molecules commonly used in permeability studies were chosen as biomarkersEpithelial or for endothelial paracellular cell and monolayers transcellular were culturedtransport. with Lucifer their yellow specific (0.44 medium, kDa; describedInvitrogen, in SectionCarlsbad, 2.3, on CA, the USA) separation has previously membranes been in used the as top a small channels. molecule Co-cultures paracellular of cell transport bilayers marker were constructed[31,37]. However, by growing it epithelialshould also and beendothelial noted that cellsa growing on opposite body of surfaces literature of thesuggests separation that membranes.transcellular HUVEC passage cell suspensions of Lucifer yellow were first may seeded also withoccur endothelial via organic medium anion intransporters the top channels [38]. on collagen-coatedDextrans, non-digestible surface membranes. sugars, were Once selected the HUVEC to represent cell monolayers typical agents reached that are full transported confluency, typicallyvia the within paracellular 1–2 days, route the bottom through channels cell tight were junctions perfused with[27,39]. suspensions Expression of by A549 epithelial cells in theor endothelialendothelial medium. cells of The membrane tubing adapters receptors were mediating temporarily transcytosis plugged, of dextran and the has microfluidic not been reported devices werethus flipped far; however, upside down. pinocytosis The inverted of dextran devices was observed were suspended [40], in which over a molecules small water are bath captured for 24 in h allowingvesicles the on establishment one side of ofthe epithelial cell, drawn and endothelialacross the cell, confluent and ejected monolayers. on the Theother devices side. Since were thenpinocytosis flipped back is toa anvery upright slow process, position transcellular and the tubing transport adapters of were dextran un-plugged. is negligible Live compared cell staining to wasits performed paracellular using transport CellTracker [41]. To fluorescent explore potential probes (Invitrogen, molecular-size Carlsbad, effect, 4-dextran CA, USA), (4 within kDa) and the microfluidic70-dextran devices, (70 kDa) and (Invitrogen, cells were immunostained Carlsbad, CA, USA) for known were chosen intercellular to represent adhesion the biomarkers intermediate to confirmand confluencylarge-size agents, of the cell respective layers. ly. FITC-transferrin (78 kDa; Rockland Antibodies, Limerick, PA,In all USA) cases, was with used or withoutto represent cells, a the typical top and protein bottom that channels is transported were filled via withthe transcellular cell media solution route, mixedtranscytosis, with various where tracer epithelial molecules A549 at pre-determined and endothelial initial HUVEC concentrations cells uptake and cell the media molecules with zero by molecularbinding concentration, transferrin to respectively. its membrane Both receptor microchannels expressed were by then both connected cell lines to a[31,39,42,43]. PHD Ultra syringe FITC- pumptransferrin (Harvard and Apparatus, 70-dextran Holliston, molecules MA, are USA) about to drive the same the media in size in and, each therefore, channel at may the sameshare constant similar flowparacellular rate to minimize transport trans-membrane level. pressure differences. A549 cell medium was flown through both channels only when epithelial monolayers were tested. In all other conditions, including cell bilayers, 2.5.HUVEC Concentration cell medium Measuremen was flownts in through Microfluidic both Devices channels. The temporal fluorescent intensities because of decreasingEpithelial concentrationsor endothelial cell at themonolayers top and increasingwere cultured concentrations with their specific at the medium, bottom microchannel described in Sectionoutlets, 2.3, with on and the without separation cells, membranes were recorded in the at fivetop minuteschannels. intervals. Co-cultures The of channel cell bilayers outlets were were constructedexposed to a by light growing source forepithelial a short time,and lessendothelial than 5 s, cells to limit on photo opposite bleaching, surfaces and of calibration the separation curves membranes. HUVEC cell suspensions were first seeded with endothelial medium in the top channels on collagen-coated surface membranes. Once the HUVEC cell monolayers reached full confluency, typically within 1–2 days, the bottom channels were perfused with suspensions of A549 cells in the endothelial medium. The tubing adapters were temporarily plugged, and the microfluidic devices were flipped upside down. The inverted devices were suspended over a small water bath for 24 h allowing the establishment of epithelial and endothelial confluent monolayers. The devices were then flipped back to an upright position and the tubing adapters were un-plugged. Live cell staining was performed using CellTracker fluorescent probes (Invitrogen, Carlsbad, CA, USA), within the microfluidic devices, and cells were immunostained for known intercellular adhesion biomarkers to confirm confluency of the cell layers.

Micromachines 2019, 10, x FOR PEER REVIEW 6 of 18

In all cases, with or without cells, the top and bottom channels were filled with cell media solution mixed with various tracer molecules at pre-determined initial concentrations and cell media with zero molecular concentration, respectively. Both microchannels were then connected to a PHD Ultra syringe pump (Harvard Apparatus, Holliston, MA, USA) to drive the media in each channel at the same constant flow rate to minimize trans-membrane pressure differences. A549 cell medium was flown through both channels only when epithelial monolayers were tested. In all other conditions, including cell bilayers, HUVEC cell medium was flown through both channels. The temporal fluorescent intensities because of decreasing concentrations at the top and increasing concentrations at the bottom microchannel outlets, with and without cells, were recorded at five Micromachines 2019, 10, 533 6 of 18 minutes intervals. The channel outlets were exposed to a light source for a short time, less than 5 s, to limit photo bleaching, and calibration curves were used to relate the measured fluorescent intensitieswere used toto relatemolecular the measured concentrations. fluorescent For intensities steady-state to molecular experiments, concentrations. similar to For the steady-state transient experiments, similarmicrofluidic to the devices transient with experiments, and without microfluidic confluent devices cell monolayers with and withoutor bilayers confluent were connectedcell monolayers to the orsyringe bilayers pump were under connected several to thecons syringetant flow pump rates. under Following several the constant development flow rates. of steady-stateFollowing the molecular development concentration of steady-state distributions, molecular 500 µL concentration effluent samples distributions, from both 500microchannelsµL effluent weresamples collected from bothand microchannelsthe corresponding were fluorescent- collected andlight the intensities corresponding were measured fluorescent-light using the intensities BioTek Synergywere measured 2 well plate using reader. the BioTek The Synergysample 2collection well plate time reader. was The25, 10, sample 5, and collection 2.5 h for time 20, was50, 100, 25, 10,and 5, 200and µL/h 2.5 h flow for 20,rate, 50, respectively, 100, and 200 withµL /theh flow microfluidic rate, respectively, devices placed with thein while microfluidic the syringe devices pump placed kept outsidein while of the an syringe incubator. pump As keptin th outsidee transwell of an experiments, incubator. Asthe in same the transwellcalibration experiments, curves were the used same to relatecalibration the curveslight intensities were used to to relatemolecular the light concentrations. intensities to molecularThe initial concentrations. inlet concentrations The initial in inletthe microfluidicconcentrations temporal in the microfluidic and steady temporalexperiments and steadywere the experiments same as those were theapplied same in as the those transwell applied experiments.in the transwell It experiments.was important It wasto determine important the to determinetime to reach the time steady to reach state steady from statethe temporal from the experimentstemporal experiments to ensure to collection ensure collection of effluent of esamplesffluent samples under steady-state under steady-state conditions. conditions. The tested The tested flow rateflow range rate range was 20–200 was 20–200 µL/hµ withL/h with a corresponding a corresponding wall wall shear shear stress stress range range of of0.0012–0.012 0.0012–0.012 dyne/cm dyne/cm2. 2.

3. Results Results and and Discussion Discussion A549 and and HUVEC cell cell monolayers monolayers as as well well as as A549/HUVEC A549/HUVEC bilayers were were successfully cultured inin microfluidic microfluidic devices as as shown by the bright-fieldbright-field images in Figure3 3a–c.a–c. SinceSince itit isis didifficultfficult to distinguish betweenbetween didifferentfferent cell cell types types in bright-fieldin bright-field images images of bilayers of bilayers cultured cultured on opposite on opposite surfaces surfacesof a membrane, of a membrane, live cell staininglive cell wasstaining performed was performed within a microfluidicwithin a microfluidic device to confirmdevice to successful confirm successfulestablishment establishment of an A549 of/HUVEC an A549/HUVEC cell co-culture. cell co-culture.

(a) (b) (c)

Figure 3.3. MicroscopicMicroscopic bright-field bright-field images images of: (of:a) confluent(a) confluent A549 epithelialA549 epithelial cell monolayer, cell monolayer, (b) confluent (b) confluenthuman umbilical human veinumbilical cells (HUVEC) vein cells endothelial (HUVEC) cell endothelial monolayer, cell and monolayer, (c) confluent and A549 (/cHUVEC) confluent cell A549/HUVECbilayer cultured cell in bilayer microfluidic cultured devices; in micr scaleofluidic bars devices; are 250 µ scalem. bars are 250 μm.

Cell layer confluency confluency is typically verified verified via staining particular proteins involved in cell–cell adhesion. PreviousPrevious work work has has suggested suggested that A549that A549 form tighterform barrierstighter withbarriers the presencewith the of presence intercellular of intercellularjunction proteins junction [44,45 proteins]. Tight [44,45]. junctions Tight comprise junctions transmembrane comprise transmembrane proteins (occludin, proteins claudins, (occludin, JAM), cytoplasmic attachment proteins (ZO-1, ZO-2, ZO-3), and cytoskeleton protein F-actin [46]. The ZO-1 cytoplasmic protein was selected for staining as a biomarker for tight junctions in the A549 cells. The abundance of the red signal in Figure4a, corresponding to the stained ZO-1 proteins (Proteintech, Rosemont, IL, USA), is evidence for the formation of tight junctions and confluency of the A549 cell monolayer. Adherens junctions in endothelial cells are mainly composed of cadherins and catenins [47]. The vascular endothelial cadherin (VE-cadherin, Cell Signaling Technology, Beverly, MA, USA), a calcium-dependent cell–cell adhesion transmembrane glycoprotein, was selected for staining as a biomarker for adherens junctions in the HUVEC cells. The green signal in Figure4b, corresponding to Micromachines 2019, 10, x FOR PEER REVIEW 7 of 18 claudins, JAM), cytoplasmic attachment proteins (ZO-1, ZO-2, ZO-3), and cytoskeleton protein F- actin [46]. The ZO-1 cytoplasmic protein was selected for staining as a biomarker for tight junctions in the A549 cells. The abundance of the red signal in Figure 4a, corresponding to the stained ZO-1 proteins (Proteintech, Rosemont, IL, USA), is evidence for the formation of tight junctions and confluency of the A549 cell monolayer. Adherens junctions in endothelial cells are mainly composed ofMicromachines cadherins2019 and, 10 , 533catenins [47]. The vascular endothelial cadherin (VE-cadherin, Cell Signaling7 of 18 Technology, Beverly, MA, USA), a calcium-dependent cell–cell adhesion transmembrane glycoprotein, was selected for staining as a biomarker for adherens junctions in the HUVEC cells. The greenthe stained signal VE-cadherins, in Figure 4b, corresponding demonstrates theto the formation stained VE-cadherins, of Adherens junctions demonstrates and confluency the formation of the of AdherensHUVEC cell junctions monolayer. and confluency of the HUVEC cell monolayer.

(a) (b)

Figure 4. FluorescenceFluorescence microscope microscope images images of: ( (aa)) A549 A549 epithelial epithelial cells cells immunostained for for ZO-1 cytoplasmic proteins (red) as a biomarkerbiomarker for tight junctions, and (b) HUVECHUVEC endothelialendothelial cells immunostained forfor VE-cadherin VE-cadherin transmembrane transmembrane proteins proteins (green) (green) as a biomarker as a biomarker for adherens for adherens junctions withjunctions standard with Hoechststandard staining Hoechst (blue) staining of both (blue) A549 of both and HUVECA549 an celld HUVEC nuclei. cell The nuclei. images The were images taken werefrom slide-mountedtaken from slide-mounted membranes, membranes, which were removedwhich were from removed the microﬂuidic from the devices microfluidic following devices cell followingﬁxing and cell staining. fixing and staining.

3.1. Molecular Transport in Transwells DiffusionDiffusion in transwells is a process continuing untiluntil an equilibrium state is reached, at which the concentration at the bottom equals the concentrationconcentration at the top compartment with zero gradient. The parameter used for assessing molecular transport is the relative temporal concentration at the bottom compartment, C(t), defineddefined as:as: C(t) CB(t) = (1) 𝐶(𝑡) 𝐶(𝑡) C0 CT(t) + 3 CB(t) = · (1) 𝐶 𝐶(𝑡) +3∙𝐶(𝑡) where CB(t) and CT(t) are the average molecular concentrations measured at the bottom and top wherecompartments, CB(t) and respectively, CT(t) are the with average initial conditions molecularC Bconcentrat(t = 0) = 0ions and C measuredT(t = 0) = C at0. Initialthe bottom concentrations and top compartments,for Lucifer yellow, respectively, 4-dextran, with 70-dextran, initial conditions and transferrin CB(t = 0) molecules = 0 and CT( weret = 0) =C C0 0=. Initial100, 60, concentrations 1, and 1 µM, forrespectively. Lucifer yellow, Since the4-dextran, bottom 70-dextr compartmentan, and volume transferrin is three molecules times larger were than C0 = the100, top 60, compartment1, and 1 µM, volume,respectively. the equilibriumSince the bottom concentration compartment is expected volume to is be three a quarter times of larger the top than initial the top concentration, compartment i.e. Cvolume,B(t )the= C equilibriumT(t ) = C concentration0/4, independent is expected of molecule to be or a cell quarter type. of Therefore, the top initial to elucidate concentration, the effect i.e. of →∞ →∞→∞ CdiBff(terent) cell= CT monolayers(t ) = C0/4, on independent the transport of of molecule various molecules, or cell type. experiments Therefore, were to elucidate terminated the aftereffect two of differenthours to ensurecell monolayers that the concentration on the transport at the of bottom various compartment, molecules, experiments while high enough were terminated to allow reliable after twomeasurements, hours to ensure is still that smaller the concentration than the equilibrium at the bottom level. compartment, while high enough to allow reliableThe measurements, measured relative is still concentrations smaller than the of threeequilibrium paracellular-transport level. markers, Lucifer yellow, 4-dextran,The measured and 70-dextran, relative and concentrations one transcellular-transport of three paracellular-transport marker, transferrin, markers, are compared Lucifer inyellow, Figure 4-5 withdextran, and and without 70-dextran, confluent and cellone monolayers.transcellular-transport All relative marker, concentrations transferrin, measured are compared at the in bottom Figure 5compartments with and without are smaller confluent than cell the equilibriummonolayers. concentration, All relative concentrationsCt2 = C(t = 2 h) measured< 0.25C0, indicatingat the bottom that compartmentsthe diffusion process are smaller is still than ongoing the equilibrium after the two-hours concentration, incubation Ct2 = time. C(t = Without2 h) < 0.25 cells,C0, indicating black columns, that the relativediffusion concentration process is still decreases ongoing with after increasing the two-hours molecular incubation size. Based time. on Without the Stokes–Einstein cells, black equation, molecular diffusivity depends on its size as follows [48,49]:

!1/3 kB T 3M D = · with r = (2) 6πµr 4πρN Micromachines 2019, 10, 533 8 of 18

where kB is the Boltzmann constant, T is the temperature, µ and ρ are the ﬂuid viscosity and density, respectively; r is a radius calculated based on the molecular weight, M, and the Avogadro number, N. Micromachines 2019, 10, x FOR PEER REVIEW 9 of 18 Following the convection-diﬀusion equation: through the tight junctions of the confluent∂c monolayers. The transferrin relative concentration with + u c = D 2c (3) HUVEC cells is about double that with A549∂t cells,·∇ suggesting∇ either higher transport rate or higher receptor density in HUVEC cells.

Figure 5.5. RelativeRelative concentrationsconcentrations of of four four molecular molecular markers, markers, Lucifer Lucifer yellow, yellow, 4-dextran, 4-dextran, 70-dextran, 70-dextran, and transferrin,and transferrin, measured measured at the at the bottom bottom compartments compartments of of transwells transwells with with and and without without conﬂuentconfluent A549 epithelial or HUVEC endothelial cell monolayer after aa 2-h2-h incubationincubation time.time. SigniﬁcanceSignificance determined by Student’s tt-test;-test; **P < 0.05; *** P < 0.001;0.001; nn >> 3.3.

3.2. MolecularThe spatiotemporal Convection–Diffusi concentrationon in Microfluidic is directly Devices proportional to the molecular diffusion coefficient, D, where u is the velocity vector. Lucifer yellow is the smallest molecule in this study with the The convection–diffusion experiments10 2 in the microfluidic device start with initial conditions of highest diffusion coefficient, 5.0 10− m /s, while the diffusion coefficients of the large 70-dextran zero molecular concentration at× the bottom channel 11inlet2 and a uniform 11concentration2 at the top and transferrin molecules are much smaller, 5.6 10− m /s and 5.85 10− m /s, respectively, and channel inlet, C0. Molecular transport from the top× to the bottom channel× starts after imposing the the diffusion coefficient of 4-dextran is intermediate about 1.35 10 10m2/s. Thus, in the absence of same constant flow rate through both channels. The time-dependent× − molecular concentrations are cells, the measured relative concentration decreases because of decreasing diffusivity with increasing estimated from the average concentration measurements, c(x,y,z;t), which were recorded at the top molecular size. and bottom channel outlets yielding: In the presence of cell monolayers, either epithelial A549 or endothelial HUVEC, the measured / relative concentrations, Ct2/C0, are smaller than those measured in the absence of cells (Figure5). Lucifer yellow and both dextran𝐶(𝑡) molecules= have 𝑐(𝑥=𝐿/2,𝑦,𝑧;𝑡 been widely considered)𝑑𝑦𝑑𝑧/(𝑊 as𝐻 paracellular) markers [27,(4)31 ]. Hence, their mass transfer rate is generally/ not aided by active membrane transport proteins; it is rather a passive diffusion through the cellular junctions and membrane pores due to local concentration and gradients. The total mass transferred from the top to the bottom compartment, and the resulting bottom concentration, is the integral over time/ of the product of molecular flux and flux area. Therefore, since the cell monolayers present𝐶( additional𝑡) = resistance 𝑐(𝑥=𝐿/2,𝑦,𝑧;𝑡 connected in)𝑑𝑦𝑑𝑧 series/( with𝑊𝐻) the membrane resistance(5) to molecular transport through the/ combined barrier, the molecular flux decreases in the presence of cells and with it the molecular concentration. Furthermore, in the case of the large 70-dextran, the where H, W, and L are the channel height, width, and overlapping length, respectively, and h is half relative concentration with cells is at the background level suggesting that the confluent monolayers the membrane thickness. The origin of the coordinate system is chosen to be at the center of the device essentially blocked molecular transport between the two compartments. For the smaller molecules, such that the fluid domain is confined to the top: −L/2 ≤ x ≤ L/2, h ≤ y ≤ H+h, −W/2 ≤ z ≤ W/2, and the concentrations measured in the presence of cells were found to be about half of the measured bottom channel: −L/2 ≤ x ≤ L/2, −H−h ≤ y ≤ −h, −W/2 ≤ z ≤ W/2. Because of the convection effect, a concentrations without cells but clearly above background levels. Since neither A549 nor HUVEC cells steady-state concentration distribution is developed before complete mixing of the two fluid flows are known to uptake these molecules, they penetrated either one of the confluent monolayers via the such that C∞ = CB(t→∞) < C0/2. Initial concentrations in microfluidic devices were identical to those in paracellular route. The junctions of the confluent A549 and HUVEC cell monolayers appear to be tight the transwell experiments, i.e., C0 = 100, 60, 1, and 1 µM for Lucifer yellow, 4-dextran, 70-dextran, and enough to block the transfer of large proteins but leaky to small proteins amounting to a reduction of transferrin molecules, respectively. the separation-membrane porosity. HUVEC cell layers are known to be leaky with low paracellular resistance3.2.1. Transient [50], andResponse the leaky Characterization junctions are associated with cell proliferation or turnover (mitosis) and The temporal measurements of Lucifer yellow concentrations at the bottom channel outlets of the microfluidic devices with and without cells, under a flow rate of 20 µL/h, are summarized in Figure 6. The measured concentrations are normalized by the steady-state concentration, C∞ = 0.2C0, and a characteristic time scale of τ = 25 min. In all cases, the system had reached a steady state within

Micromachines 2019, 10, 533 9 of 18 cell death (apoptosis) [18,51]. A549 cell layers have also been reported to be much leakier than other conventional epithelial cell lines [52]. Indeed, the leakiness of A549 cell monolayers limits their utility in examining the transport of low molecular weight drugs [31]. Therefore, it is not surprising that for these three paracellular markers, no clear difference has been observed between paracellular transport through a confluent endothelial HUVEC and an epithelial A549 monolayer. The size and diffusivity of FITC-transferrin and large 70-dextran are about the same. Therefore, in the absence of an alternative transport mechanism, paracellular transport of both molecules through a confluent cell monolayer should be similar. Indeed, the transferrin concentration measurements with cells are smaller than the measured concentration without cells. However, while the concentration level of 70-dextran with cells is nearly zero, the measured concentrations of transferrin are statistically significant above background level. Assuming that a confluent monolayer of either A549 or HUVEC cells renders paracellular transport of transferrin negligible, similar to 70-dextran, the elevated transferrin concentration measurements can be attributed to the transcellular transport not available for dextran molecules. Hence, it seems that transferrin receptors expressed by both A549 and HUVEC cells actively facilitate transcellular transport of transferrin molecules not through the tight junctions of the confluent monolayers. The transferrin relative concentration with HUVEC cells is about double that with A549 cells, suggesting either higher transport rate or higher receptor density in HUVEC cells.

3.2. Molecular Convection–Diffusion in Microfluidic Devices The convection–diffusion experiments in the microfluidic device start with initial conditions of zero molecular concentration at the bottom channel inlet and a uniform concentration at the top channel inlet, C0. Molecular transport from the top to the bottom channel starts after imposing the same constant flow rate through both channels. The time-dependent molecular concentrations are estimated from the average concentration measurements, c(x,y,z;t), which were recorded at the top and bottom channel outlets yielding:    ZW/2 HZ+h    C (t) =  c(x = L y z t)dydz (W H) T  /2, , ; / (4)   · W/2 h −

and    ZW/2 Z h   −  C (t) =  c(x = L y z t)dydz (W H) B  /2, , ; / (5)   · W/2 H h − − − where H, W, and L are the channel height, width, and overlapping length, respectively, and h is half the membrane thickness. The origin of the coordinate system is chosen to be at the center of the device such that the fluid domain is confined to the top: L/2 x L/2, h y H+h, W/2 z W/2, and − ≤ ≤ ≤ ≤ − ≤ ≤ bottom channel: L/2 x L/2, H h y h, W/2 z W/2. Because of the convection effect, a − ≤ ≤ − − ≤ ≤ − − ≤ ≤ steady-state concentration distribution is developed before complete mixing of the two fluid flows such that C = CB(t ) < C0/2. Initial concentrations in microfluidic devices were identical to those ∞ →∞ in the transwell experiments, i.e., C0 = 100, 60, 1, and 1 µM for Lucifer yellow, 4-dextran, 70-dextran, and transferrin molecules, respectively.

3.2.1. Transient Response Characterization The temporal measurements of Lucifer yellow concentrations at the bottom channel outlets of the microﬂuidic devices with and without cells, under a ﬂow rate of 20 µL/h, are summarized in Figure6. The measured concentrations are normalized by the steady-state concentration, C = 0.2C0, and a ∞ characteristic time scale of τ = 25 min. In all cases, the system had reached a steady state within one hour. More importantly, the collapse of the data indicates that the bottom-channel concentration increase with time, CB(t), in the absence of cells is very much the same as in the presence of cells, Micromachines 2019, 10, x FOR PEER REVIEW 10 of 18

Micromachines 2019, 10, 533 10 of 18 one hour. More importantly, the collapse of the data indicates that the bottom-channel concentration increase with time, CB(t), in the absence of cells is very much the same as in the presence of cells, eithereither anan A549A549 oror HUVECHUVEC monolayermonolayer oror anan A549A549/HU/HUVECVEC bilayer.bilayer. Thus,Thus, withinwithin experimentalexperimental error, eacheach ofof thethe threethree cell-culturecell-culture combinationscombinations hashas negligiblenegligible eeffectﬀect onon thethe transporttransport raterate ofof LuciferLucifer yellowyellow moleculesmolecules fromfrom thethe toptop toto thethe bottombottom channel.channel. InIn contrast,contrast, inin thethe transwelltranswell staticstatic experiments,experiments, thethe concentrationconcentration atat thethe bottombottom compartmentcompartment withwith cellscells waswas foundfound toto bebe aboutabout halfhalf ofof thatthat withwith nono cells,cells, andand isis discusseddiscussed next.next.

FigureFigure 6.6. Normalized Lucifer yellow concentration measured at the bottombottom channelchannel outletoutlet ofof aa microfluidicmicrofluidic devicedevice asas aa functionfunction ofof time,time, underunder a a 20 20µL/hµL/h flowflow raterate inin eacheach channel,channel, withwith andand withoutwithout confluentconfluent cellcell monomono/bilayers;/bilayers; thethe experimentalexperimental datadata (symbols)(symbols) areare fittedfitted withwith aa linearlinear (red(red dashdash line)line) andand exponentialexponential functionfunction (black(black solidsolid line).line).

3.2.2. Steady-state Molecular Transport 3.2.2. Steady-state Molecular Transport The flow rate, Q, is the most dominant parameter controlling the molecular transport rate between The flow rate, Q, is the most dominant parameter controlling the molecular transport rate the two channels as it can vary from a very low rate resulting in complete mixing of the two streams, between the two channels as it can vary from a very low rate resulting in complete mixing of the two i.e., C C0/2 as→Q 0, to a flow→ rate high enough to render cross-membrane diffusion negligible streams,∞→ i.e., C∞ C→0/2 as Q 0, to a flow rate high enough to render cross-membrane diffusion such that C 0 as Q → . Therefore,→∞ the average molecular concentration at the bottom channel negligible such∞→ that C→∞∞ 0 as Q . Therefore, the average molecular concentration at the bottom outlet was measured under different flow rates once steady state concentration distributions had been channel outlet was measured under different flow rates once steady state concentration distributions established within the microfluidic devices. The relative concentrations of the four molecules, C /C0, had been established within the microfluidic devices. The relative concentrations of the∞ four in the absence and presence of cell cultures are detailed in Figure7 as a function of the flow rate. In molecules, C∞/C0, in the absence and presence of cell cultures are detailed in Figure 7 as a function of general, in all cases with and without cells, the concentration is roughly inversely proportional to the the flow rate. In general, in all cases with and without cells, the concentration is roughly inversely flow rate decreasing with increasing flow rate. Under high flow rate, Q > 100 µL/h, the measured proportional to the flow rate decreasing with increasing flow rate. Under high flow rate, Q > 100 µL/h, concentrations are so low such that it is very difficult to elucidate the effect of the cell cultures on the the measured concentrations are so low such that it is very difficult to elucidate the effect of the cell ensuing molecular transport because of the large experimental error. Phase contrast imaging revealed cultures on the ensuing molecular transport because of the large experimental error. Phase contrast no changes in the confluence or morphology of monolayers cultured under the tested flow rate range, imaging revealed no changes in the confluence or morphology of monolayers cultured under the Q = 30–200 µL/h. tested flow rate range, Q = 30–200 µL/h. The steady concentrations measured at the bottom channel outlets under a 20 µL/h flow rate, however,0.3 are high enough to facilitate a meaningful discussion0.3 and are depicted in Figure8. In 0 0

C 4-Dex No Cells C / LucY No Cells /

microﬂuidic∞ devices, similar to the results in the transwells, the concentrations of the 4-dextran ∞

C LucY A549 monolayer 4-Dex A549 monolayer C molecules in the presence of cellsLucY areHUVEC smaller monolayer than the concentration without cells;4-Dex namely, HUVEC inmonolayer both cases, 0.2 0.2 leaky junctions of the confluentLucY cell A549/HUVEC layers bilayer failed to block paracellular transport4-Dex of A549/HUVEC intermediate-size bilayer molecules but rather presented an added resistance akin to decreasing membrane porosity. Also, the concentrations0.1 of the 70-dextran in the presence of cells0.1 are within the background noise indicating that the cellular junctions are tight enough to prevent paracellular transport of large molecules. On the Relative Concentration, other hand, in contrast with the transwells findings, theRelative Concentration, Lucifer yellow relative concentrations with 0 0 and without0 cells 50 are about 100 the same 150 within 200 experimental 250 error.0 Since 50 Lucifer 100 yellow 150 has been 200 utilized 250 Flow Rate, Q [µL/h] Flow Rate, Q [µL/h] as a paracellular marker, its uptake by A549 and HUVEC cells is assumed to be negligible. Thus, if indeed its transcellular transport(a) is insignificant, it seems that the shear stress(b) enhances paracellular

Micromachines 2019, 10, x FOR PEER REVIEW 10 of 18 one hour. More importantly, the collapse of the data indicates that the bottom-channel concentration increase with time, CB(t), in the absence of cells is very much the same as in the presence of cells, either an A549 or HUVEC monolayer or an A549/HUVEC bilayer. Thus, within experimental error, each of the three cell-culture combinations has negligible effect on the transport rate of Lucifer yellow molecules from the top to the bottom channel. In contrast, in the transwell static experiments, the concentration at the bottom compartment with cells was found to be about half of that with no cells, and is discussed next.

Figure 6. Normalized Lucifer yellow concentration measured at the bottom channel outlet of a microfluidic device as a function of time, under a 20µL/h flow rate in each channel, with and without confluent cell mono/bilayers; the experimental data (symbols) are fitted with a linear (red dash line) and exponential function (black solid line).

3.2.2. Steady-state Molecular Transport The flow rate, Q, is the most dominant parameter controlling the molecular transport rate between the two channels as it can vary from a very low rate resulting in complete mixing of the two streams, i.e., C∞→C0/2 as Q→0, to a flow rate high enough to render cross-membrane diffusion negligible such that C∞→0 as Q→∞. Therefore, the average molecular concentration at the bottom channel outlet was measured under different flow rates once steady state concentration distributions had been established within the microfluidic devices. The relative concentrations of the four molecules, C∞/C0, in the absence and presence of cell cultures are detailed in Figure 7 as a function of the flow rate. In general, in all cases with and without cells, the concentration is roughly inversely proportionalMicromachines 2019to the, 10 flow, x FOR rate PEER decreasing REVIEW with increasing flow rate. Under high flow rate, Q > 100 µL/h, 11 of 18 Micromachines 2019, 10, 533 11 of 18 the measured concentrations are so low such that it is very difficult to elucidate the effect of the cell 0.3 0.3 cultures0 on the ensuing molecular70-Dex No transport Cells because of0 the large experimental error. Phase contrast

C C Transferrin No Cells / /

∞ 70-Dex A549 monolayer ∞ C imagingtransport revealed of Lucifer no yellowchanges to in such the a confluence level practically or morphology diminishingC of themonolayers resistanceTransferrin cultured of A549 the monolayer cell under layers the to 70-Dex HUVEC monolayer 0.2 0.2 Transferrin HUVEC monolayer testedsmall-molecule flow rate range, transport. Q = 30–20070-Dex A549/HUVECµL/h. bilayer Transferrin A549/HUVEC bilayer

0.3 0.3 0

0.1 0 0.1

C 4-Dex No Cells C / LucY No Cells / ∞ ∞

C LucY A549 monolayer 4-Dex A549 monolayer C LucY HUVEC monolayer 4-Dex HUVEC monolayer 0.2 0.2

Relative Concentration, LucY A549/HUVEC bilayer Relative Concentration, 4-Dex A549/HUVEC bilayer 0 0 0 50 100 150 200 250 0 50 100 150 200 250 Flow Rate, Q [µL/h] Flow Rate, Q [µL/h] 0.1 0.1 (c) (d) Relative Concentration, 0Figure 7. Flow rate effect on the steady-state relativeRelative Concentration, 0concentrations of: (a) Lucifer yellow, (b) 4- 0 50 100 150 200 250 0 50 100 150 200 250 dextran, (c) 70-dextran,Flow Rate, Qand[µL/h] (d) transferrin markers measured at the Flowbottom Rate, Qchannel[µL/h] outlets of Micromachinesmicrofluidic 2019, 10 devices,, x FOR PEER because REVIEW of molecular transport from top to bottom channel, with and without 11 of 18 confluent cell mono/bilayers;(a) dash lines are functions inversely proportional(b to) the flow rate fitted to 0.3the no cell experimental data with R2 = 0.9. 0.3

0 70-Dex No Cells 0

C C Transferrin No Cells

/ / ∞ 70-Dex A549 monolayer ∞ C C Transferrin A549 monolayer The steady concentrations70-Dex HUVECmeasured monolayer at the bottom channel outlets under a 20 µL/h flow rate, 0.2 0.2 Transferrin HUVEC monolayer 70-Dex A549/HUVEC bilayer however, are high enough to facilitate a meaningful discussion and areTransferrin depicted A549/HUVEC in Figure bilayer 8. In microfluidic devices, similar to the results in the transwells, the concentrations of the 4-dextran molecules0.1 in the presence of cells are smaller than the0.1 concentration without cells; namely, in both cases, leaky junctions of the confluent cell layers failed to block paracellular transport of intermediate-

sizeRelative Concentration, molecules but rather presented an added resistanceRelative Concentration, akin to decreasing membrane porosity. Also, 0 0 the concentrations0 50 of 100 the 70-dextran 150 200 in the 250presence 0of cells 50 are within 100 the 150 background 200 noise 250 Flow Rate, Q [µL/h] Flow Rate, Q [µL/h] indicating that the cellular junctions are tight enough to prevent paracellular transport of large molecules. On the other(c) hand, in contrast with the transwells findings,( dthe) Lucifer yellow relative concentrations with and without cells are about the same within experimental error. Since Lucifer yellowFigureFigure has 7. 7. been Flow utilized raterate eeffectffect as on ona theparacellular the steady-state steady-state marker relative relative, concentrationsits concentrationsuptake by of: A549 ( aof:) Lucifer and(a) Lucifer HUVEC yellow, yellow, ( bcells) 4-dextran, (isb )assumed 4- dextran, (c) 70-dextran, and (d) transferrin markers measured at the bottom channel outlets of to be(c negligible.) 70-dextran, Thus, and (ifd )indeed transferrin its transcellular markers measured transp atort the is insignificant, bottom channel it outletsseems that of microfluidic the shear stress microfluidic devices, because of molecular transport from top to bottom channel, with and without enhancesdevices, paracellular because of moleculartransport transportof Lucifer from yellow top toto bottomsuch a level channel, practically with and diminishing without confluent the resistance cell confluent cell mono/bilayers; dash lines are functions inversely proportional to the flow rate fitted to of themono cell/ bilayers;layers to dash small-molecule lines are functions transport. inversely proportional to the flow rate fitted to the no cell theexperimental no cell experimental data with Rdata2 = with0.9. R2 = 0.9.

The steady concentrations measured at the bottom channel outlets under a 20 µL/h flow rate, however, are high enough to facilitate a meaningful discussion and are depicted in Figure 8. In microfluidic devices, similar to the results in the transwells, the concentrations of the 4-dextran molecules in the presence of cells are smaller than the concentration without cells; namely, in both cases, leaky junctions of the confluent cell layers failed to block paracellular transport of intermediate- size molecules but rather presented an added resistance akin to decreasing membrane porosity. Also, the concentrations of the 70-dextran in the presence of cells are within the background noise indicating that the cellular junctions are tight enough to prevent paracellular transport of large molecules. On the other hand, in contrast with the transwells findings, the Lucifer yellow relative concentrations with and without cells are about the same within experimental error. Since Lucifer yellow has been utilized as a paracellular marker, its uptake by A549 and HUVEC cells is assumed to be negligible. Thus, if indeed its transcellular transport is insignificant, it seems that the shear stress enhances paracellular transport of Lucifer yellow to such a level practically diminishing the resistance of theFigure Figurecell layers 8.8.Steady-state Steady-state to small-molecule relativerelative concentrationsconcentrations transport. ofof fourfour molecularmolecular markers,markers, LuciferLucifer yellow,yellow, 4-dextran,4-dextran, 70-dextran,70-dextran, and and transferrin, transferrin, measured measured at at the the bottom bottom channel channel outlets outlets of of microfluidic microfluidic devices devices with with and and withoutwithout confluent confluent cell cell mono mono/bilayers/bilayers under under a 20 µ La /h20 flow µL/h rate flow in each rate channel. in each Significance channel. determinedSignificance bydetermined Student’s tby-test; Student’s *P < 0.05; t-test;n > *3.P < 0.05; n > 3.

TheThe concentrationconcentration ofof transferrintransferrin isis signiﬁcantlysignificantly largerlargerthan than that that of of 70-dextran. 70-dextran. BothBothmolecules molecules areare similarsimilar inin sizesize and,and, sincesince appliedapplied shearshear stressstress hadhad nono eeffectﬀect onon transporttransport ofof largelarge 70-dextran70-dextran

Figure 8. Steady-state relative concentrations of four molecular markers, Lucifer yellow, 4-dextran, 70-dextran, and transferrin, measured at the bottom channel outlets of microfluidic devices with and without confluent cell mono/bilayers under a 20 µL/h flow rate in each channel. Significance determined by Student’s t-test; *P < 0.05; n > 3.

The concentration of transferrin is significantly larger than that of 70-dextran. Both molecules are similar in size and, since applied shear stress had no effect on transport of large 70-dextran

Micromachines 2019, 10, 533 12 of 18 molecules through cellular junctions, paracellular transport of transferrin is also expected to be negligible. Therefore, the measured transferrin concentration can be attributed to molecular transport via the transcellular route, i.e., transcytosis, as was observed in transwells. It is difficult to discern the contributions of the different cell layers to the transferrin transport because of the relatively large experimental error. However, in comparison with the transwells results, the small difference between transferrin concentrations with and without cells suggest that the shear stress in microfluidic devices can perhaps enhance transferrin transcellular transport by the active protein transporters.

3.3. Barrier Permeability Characterization Barrier-tissue interfaces are semi-permeable allowing selective transportation of molecules such as water, ions, and nutrients across the interface. Such a permeability barrier is maintained through cell–cell junctions and is controlled by growth factors, cytokines, and other stress-related molecules. Barrier permeability is a dynamic process influenced by many factors including exposure to inflammatory agents and gene products. Permeability is a measure of the average molecular flux across a barrier interface area. In microfluidic devices, under steady-state conditions, the apparent permeability is given by: Pam = (C /C0)(Q/A) (6) ∞ where A is the barrier interface area. However, in transwells, experiments are terminated prior to reaching an equilibrium of uniform concentration in top and bottom compartment; therefore, the permeability is calculated as follows:

Pat = (Ct2/C0)(V/A)(1/t) (7)

Ct2 is the measured concentration at the bottom compartment with a volume of V = 1500 µL following diffusion time of t = 2 h. The permeability values estimated based on Equations (6) and (7) in the range of 6 5 10− –10− cm/s are in close agreement with previously published data [12,23,27,31]. The molecular size effect on the permeability, mainly because of paracellular transport, is demonstrated in Figure9. The estimated permeability in transwells after two hours of diffusion (Figure9a) and in microfluidic devices at steady-state under a 20 µL/h flow rate (Figure9b) are plotted as a function of the molecular weight. The separation membrane permeability—without cells—is inversely proportional to the molecular size because of decreasing diffusivity with increasing size. The barrier permeability—with cell monolayers cultured on the separation membrane—decreases roughly logarithmically with increasing molecular size within experimental error. In static transwells, with no shear stress, the permeabilities of all three paracellular markers with a confluent A549 or HUVEC cell monolayer are markedly smaller than without cells; no clear difference between the epithelial- and the endothelial-monolayer permeability can be discerned. The trends in microfluidic devices with shear stress are similar to those in transwells, except for the Lucifer yellow permeability in microfluidic devices which is about the same with and without the cell layers. The permeability of the 70-dextran is close to zero in both systems. Since its transcellular transport is negligible, its paracellular transport through the cell TJs is also negligible because of its large size. Transcellular transport of 4-dextran is assumed to be negligible, similar to 70-dextran, independent of dextran molecular size. Therefore, because of its smaller size, the 4-dextran permeability in both transwells and microfluidic devices is likely due to paracellular transport through leaky TJs. The effect of shear stress on permeability can be delineated from comparing the transwells with the microfluidic results. However, the microfluidic permeability was estimated under steady molecular flux independent of time, while the molecular flux in transwells is time-dependent decreasing gradually from its maximum at the start of the diffusion process to zero at equilibrium when the concentrations at the top and bottom compartments are the same. Therefore, a direct comparison between the permeability results obtained in transwells and microfluidic devices is not appropriate. Instead, it is more appropriate to compare normalized permeability results for each system. The ratio between the permeability with and without cell layers in transwells and microfluidic devices, PC/PN, is shown in Micromachines 2019, 10, x FOR PEER REVIEW 13 of 18

systems are similar as the permeability ratio decreases with increasing molecular size indicating that the paracellular transport is adversely affected as the size of the molecules increases. However, the microfluidics slope is significantly higher than the transwells slope highlighting the additional effect of shear stress. The results suggest that the paracellular-transport dependence on molecular size can be regulated to a certain degree by applying shear stress. Indeed, it is known that TJs form regulated, selectively permeable barriers between two distinct compartments. TJs do not just represent static structural elements but they are dynamically regulated to control paracellular solute and ion transport in diverse physiologic states [53].

Micromachines 2019, 10, 533 13 of 18 Micromachines 2019, 10, x FOR PEER REVIEW 13 of 18 systemsFigure 10 are as similar a function as the of permeability molecular size. ratio The decrease trendss in with both increasing systems aremolecular similar size as the indicating permeability that theratio paracellular decreases withtransport increasing is adversely molecular affected size indicatingas the size that of the the molecules paracellular increases. transport However, is adversely the microfluidicsaffected as the slope size ofis thesignificantly molecules higher increases. than However,the transwells the microfluidics slope highlighting slope is the significantly additional highereffect ofthan shear the stress. transwells The results slope highlighting suggest that the the additional paracellular-transport effect of shear dependence stress. The resultson molecular suggest size that can the beparacellular-transport regulated to a certain dependence degree by applying on molecular shear size stress. can Indeed, be regulated it is known to a certain that TJs degree form by regulated, applying selectivelyshear stress. permeable Indeed, it barriers is known between that TJs two form distinct regulated, compartments. selectively permeableTJs do not barriers just represent between static two structuraldistinct compartments. elements but TJsthey do are not justdynamically represent regulated static structural to control elements paracellular but they aresolute dynamically and ion transportregulated in to diverse control physiologic paracellular(a) states solute [53]. and ion transport in diverse physiologic(b) states [53].

Figure 9. Molecular size effect on the permeability of paracellular markers, Lucifer yellow, 4-dextran, and 70-dextran, with and without cells in: (a) transwells after 2-hrs incubation time, Pat, and (b)

microfluidic devices under a 20µL/h flow rate, Pam. Dash lines are logarithmic functions fitted to the corresponding data sets with R2 > 0.9.

While there is a general consensus that shear stress affects paracellular transport, the actual resulting effect is not clear. Shear stress was observed to enhance barrier function in human airway epithelial cells reducing paracellular permeability [54], and the permeability of bovine aortic endothelial monolayers was reported to be enhanced by application of shear [55]. Elsewhere it was suggested that the effect depends on the shear stress level as high shear stress suppresses mitosis and apoptosis while low shear stress supports both processes [51]. Thus, cellular turnover rates and apoptosis rates, and by (a)association the prevalence of leaky junctions and permeability,(b) will be greater in low shear stress regions. In a recent study, permeability was found to increase at the onset of flow FigureFigure 9. 9. MolecularMolecular size size effect effect on on the the permeability permeability of of paracellular paracellular markers, markers, Lucifer Lucifer yellow, yellow, 4-dextran, 4-dextran, and slowly plateaus to a baseline value [27]. Similarly, here the paracellular permeability of smaller andand 70-dextran, 70-dextran, with with and and without without cells cells in: ( (aa)) transwells transwells after after 2-hrs 2-hrs incubation incubation time, time, PPatat, ,and and ( (bb)) molecules is enhanced as the permeability ratio of 4-dextran almost doubled, increasing from about microfluidicmicrofluidic devices devices under under a a 20µL/h 20µL/h flow flow rate, Pam.. Dash Dash lines lines are are logarithmic logarithmic functions functions fitted fitted to to the the 1/3 in static transwells to about 2/3 in microfluidic devices, due to the applied shear stress. correspondingcorresponding data data sets sets with with RR22 >> 0.9.0.9.

Figure 10.10. Molecular sizesize eeffectffect on on the the ratio ratio between between the the permeability permeability with with and and without without cells, cells,PC/ PNC/,P ofN, paracellularof paracellular markers, markers, Lucifer Lucifer yellow, yellow, 4-dextran, 4-dextran, and and 70-dextran, 70-dextran, in transwells in transwells after after 2-h incubation 2-h incubation time and microfluidic devices under a 20 µL/h flow rate. Dash lines are logarithmic functions fitted to the corresponding data sets with R2 > 0.9.

While there is a general consensus that shear stress affects paracellular transport, the actual resulting effect is not clear. Shear stress was observed to enhance barrier function in human airway epithelial cells reducing paracellular permeability [54], and the permeability of bovine aortic endothelial monolayers was reported to be enhanced by application of shear [55]. Elsewhere it was suggested that the effect depends on the shear stress level as high shear stress suppresses mitosis and apoptosis while low shear stress supports both processes [51]. Thus, cellular turnover rates and apoptosis rates, and by association the prevalence of leaky junctions and permeability, will be greater in low shear stress regions. In a recent study, permeability was found to increase at the onset of flow and slowly plateaus to a baselineFigure 10. value Molecular [27]. Similarly,size effect on here the the ratio paracellular between the permeability permeability ofwith smaller and without molecules cells, is P enhancedC/PN, of paracellular markers, Lucifer yellow, 4-dextran, and 70-dextran, in transwells after 2-h incubation

Micromachines 2019, 10, 533 14 of 18 as the permeability ratio of 4-dextran almost doubled, increasing from about 1/3 in static transwells to Micromachines 2019, 10, x FOR PEER REVIEW 14 of 18 about 2/3 in microfluidic devices, due to the applied shear stress. The permeabilitytime and microfluidic results devices of Lucifer under ayellow 20 µL/h inflow microfluidic rate. Dash lines devices are logarithmic are somewhat functions enigmatic.fitted In transwells,to itsthe permeability corresponding data with sets either with A549R2 > 0.9. or HUVEC cell monolayer is about half of its permeability without cells, which can be attributed to the effect of paracellular transport similar to the 4-dextran. However,The the permeability results of Lucifer of Lucifer yellow yellow in microfluidic in microfluidic devices devices is about are somewhat the same withenigmatic. and without In transwells, its permeability with either A549 or HUVEC cell monolayer is about half of its cell mono or bilayers. While Lucifer yellow has been widely used as a paracellular transport marker, permeability without cells, which can be attributed to the effect of paracellular transport similar to previous work suggested that it could be transported via a transcellular route as well. Cells with the 4-dextran. However, the permeability of Lucifer yellow in microfluidic devices is about the same organicwith anion and without transporters cell mono (OATs) or bilayers. may uptakeWhile Lucifer Lucifer yellow yellow has been since widely it is anionic used as a in paracellular solution. The presencetransport of OATs marker, on previous A549 epithelial work suggested and HUVEC that it could endothelial be transported cells has via beena transcellular reported route along as with the suggestionwell. Cells with that someorganic transcellular anion transporters transport (OATs) of Lucifermay uptake yellow Lucifer may yellow occur since [56, 57it is]. anionic To determine in whethersolution. OATs The play presence a significant of OATs role on inA549 Lucifer epitheli yellowal and molecularHUVEC endothelial transport, cells a common has been OATreported blocker (probenecid,along with Santa the Cruzsuggestion Biotechnology, that some transcellular Dallas, TX USA) transport was of used Lucifer to hinder yellow any may OAT occur activity [56,57].through To competitivedetermine inhibition whether [58OATs]. Confluent play a significant A549 monolayers role in Lucifer cultured yellow in molecular a transwell transport, top compartment a common and a microfluidicOAT blocker device (probenecid, top channel Santa wereCruz exposedBiotechnology, to media Dallas, solutions TX USA) mixed was used with to1 hinderµM Lucifer any OAT yellow activity through competitive inhibition [58]. Confluent A549 monolayers cultured in a transwell top and 1 µM probenecid. Lucifer yellow concentrations were then measured at the transwell bottom compartment and a microfluidic device top channel were exposed to media solutions mixed with 1 compartment,µM Lucifer after yellow two-hour and 1 µM incubation probenecid. time, Lucifer and atyellow the deviceconcentrations bottom were channel then outlet, measured after at reaching the steadytranswell state, similar bottom to compartment, previous experiments. after two-hour incubation time, and at the device bottom channel Asoutlet, shown after inreaching Figure steady 11, within state, experimentalsimilar to previous error, experiments. the measured relative concentrations in both transwellsAs and shown microfluidic in Figure 11, devices within areexperimental about the error, same the with measured and without relative concentrations probenecid treatment in both in presencetranswells of a A549 and cellmicrofluidic monolayer. devices This are may about be consistentthe same with with and more without recent probenecid reports that treatment OATs arein not expressedpresence by HUVEC of a A549 cells cell [ 59monolayer.,60]. Nevertheless, This may be the consistent contribution with ofmore OATs recent to Lucifer reports yellow that OATs transcellular are transportnot expressed is negligible by HUVEC with or cells without [59,60]. shear Nevertheless, stress. Paracellular the contribution seems of to OATs be the to dominant Lucifer yellow transport routetranscellular resulting in transport reduced is permeability negligible with through or wi thethout membrane shear stress. with Paracellular confluent seems A549 monolayerto be the in dominant transport route resulting in reduced permeability through the membrane with confluent transwells, PC/PN < 1. In microfluidic devices, on the other hand, the Lucifer yellow permeability A549 monolayer in transwells, PC/PN < 1. In microfluidic devices, on the other hand, the Lucifer yellow ratio is about one, P /P 1. Thus, the shear stress could only enhance the paracellular transport permeability ratioC is Nabout one, PC/PN ≅ 1. Thus, the shear stress could only enhance the paracellular as observedtransport for as 4-dextran,observed for with 4-dextran, vanishing with evanishinffect ofg theeffect cell of monolayer, the cell monolayer, or augment or augment the enhanced the paracellularenhanced transport paracellular by activatingtransport by a activating transcellular a transcellular transport tr routeansport other route than other OATs. than OATs.

FigureFigure 11. Lucifer 11. Lucifer yellow yellow relative relative concentrations concentrations withwith and without without a aconfluent confluent A549 A549 cell cellmonolayer, monolayer, with andwith withoutand without exposure exposure to probenecid,to probenecid, measured measured at atthe the bottombottom compartments compartments ofof transwells, transwells, after after a 2-h incubationa 2-h incubation time, time, and and at the at the bottom bottom channel channel outlets outlets of microfluidic microfluidic devices, devices, under under a 20a µL/h 20 µ flowL/h flow rate inrate each in each channel. channel. Significance Significance determined determined byby Student’sStudent’s tt-test;-test; *P * P< <0.05;0.05; **P**P < 0.01;< 0.01; n > 3.n > 3.

Finally, the permeability ratio for transferrin also increases from PC/PN = 0.35 in transwells Finally, the permeability ratio for transferrin also increases from PC/PN = 0.35 in transwells (Figure(Figure5) to 5) about to about 0.55 0.55 in microfluidicin microfluidic devices devices (Figure(Figure 88).). Here Here again, again, the therole role of shear of shear stressstress in transcellular transport is not very clear. Physiologically laminar shear stress was observed to in transcellular transport is not very clear. Physiologically laminar shear stress was observed to downregulate the expression of transferrin receptor reducing transcytosis [61], but elevated steady downregulate the expression of transferrin receptor reducing transcytosis [61], but elevated steady shear stress has also been shown to increase intracellular uptake enhancing endocytosis [62]. Fluid Micromachines 2019, 10, 533 15 of 18 shear stress is known to stimulate both apical and basolateral expression and trafficking of protein transporters [63], which may account for the higher transferrin permeability ratio in microfluidic devices with shear stress.

4. Conclusions Epithelial A549 and endothelial HUVEC cell were successfully co-cultured in a microfluidic membrane bilayer device to model the lung–blood barrier interface. The microfluidic devices enabled characterization of the epithelial/endothelial barrier permeability using measured concentrations resulting because of paracellular and transcellular molecular transport. The bilayer interface permeability for four different molecules was compared with the permeability in microfluidic devices, at steady state, and transwells, after two-hours incubation time, without and with either A549 or HUVEC cell monolayers. The microfluidic transient response reveals that a steady-state molecular distribution is achieved as a balance between convection and diffusion in less than an hour. The molecular flux through the porous membrane in microfluidic devices and transwells, without cells, is inversely proportional to the molecular size because of decreasing diffusivity with increasing size. Paracellular transport of 70-dextran is negligible because of its large size and, in the absence of an active dextran transporter, its slow pinocytosis can also be neglected. As a result, the measured 70-dextran concentrations in microfluidic devices and transwells were within the background noise. Transcellular transport of 4-dextran is similar to that of 70-dextran; however, because of its smaller size, the paracellular transport via tight junctions is not negligible. The permeability of 4-dextran with cells in microfluidic devices and transwells reduces to about two-thirds and one-thirds, respectively, in comparison to the permeability with no cells. The higher ratio in microfluidic devices can be attributed to the flow-induced shear stress known to enhance the leaky tight junctions. Paracellular transport of transferrin is similar to that of 70-dextran; however, because of the expression of transferrin membrane receptors by A549 and HUVEC cells, its transcytosis is significant. Similar to 4-dextran, transferrin permeability ratio in microfluidic devices is higher than in transwells indicating that the transferrin transport pathway may be more active when exposed to shear stress. Lucifer yellow permeability through confluent A549 cell monolayers in transwells and microfluidic devices with and without probenecid treatment, known to block organic anion transporters, is about the same. This suggests that transcellular transport of Lucifer yellow is negligible and, as a small molecule, paracellular is its dominant transport mechanism. Lucifer yellow permeability in transwells with cells is reduced to about half of the permeability without cells. However, in microfluidic devices with shear stress, the permeability with and without cells is about the same. In summary, the ability to precisely monitor transport properties of bio-species in microfluidic devices with multiple layers of cells, in a more physiological microenvironment, has been demonstrated. Using this technology, we can evaluate the shear stress effect on permeability that cannot be observed in standard static transwell inserts.

Author Contributions: Conceptualization T.F., L.J., Y.Z.; Data Curation T.F., Y.Z.; Formal Analysis T.F., Y.Z.; Funding Acquisition T.F., Y.Z.; Methodology T.F., L.J., Y.Z.; Resources T.F., L.J.; Supervision L.J., R.L., Y.Z.; Validation T.F., Y.Z.; Visualization T.F., Y.Z.; Writing T.F., L.J., R.L., Y.Z. Funding: This work was supported by the Arizona Biomedical Research Commission through grant ABRC ADHS14-082983, the UA GPSC Research and Project grant, and by the UA/NASA Space Grant Graduate Fellowship. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Scannell, J.W.; Blanckley,A.; Boldon, H.; Warrington, B. Scannell 2012 diagnosing the decline in pharmaceutical R&D eﬃciency. Nat. Rev. Drug Discov. 2012, 11, 191–200. [PubMed] 2. Van Norman, G.A. Overcoming the Declining Trends in Innovation and Investment in Cardiovascular Therapeutics: Beyond EROOM’s Law. JACC Basic Transl. Sci. 2017, 2, 613–625. [CrossRef][PubMed] Micromachines 2019, 10, 533 16 of 18

3. Huh, D.; Hamilton, G.A.; Ingber, D.E. From 3D cell culture to organs-on-chips. Trends Cell Biol. 2011, 21, 745–754. [CrossRef][PubMed] 4. Brandon, E.F.A.; Raap, C.D.; Meijerman, I.; Beijnen, J.H.; Schellens, J.H.M. An update on in vitro test methods in human hepatic drug biotransformation research: Pros and cons. Toxicol. Appl. Pharmacol. 2003, 189, 233–246. [CrossRef] 5. Mak, I.W.; Evaniew, N.; Ghert, M. Lost in translation: animal models and clinical trials in cancer treatment. Am. J. Transl. Res. 2014, 6, 114–118. [PubMed] 6. Sena, E.; Bart van der Worp, H.; Howells, D.; Macleod, M. How can we improve the pre-clinical development of drugs for stroke? Trends Neurosci. 2007, 30, 433–439. [CrossRef][PubMed] 7. Chuang-stein, C.; D’Agostino, R. Statistics in Drug Development. In Pharmaceutical Statistics Using SAS: A Practical Guide; Dmitrienko, A., Chuang-Stein, C., Eds.; SAS: Cary, NC, USA, 2007; pp. 1–6. 8. Maguire, T.J.; Novik, E.; Chao, P.; Barminko, J.; Nahmias, Y.; Yarmush, M.L.; Cheng, K.-C. Design and Application of Microfluidic Systems for In Vitro Pharmacokinetic Evaluation of Drug Candidates. Curr. Drug Metab. 2009, 10, 1192–1199. [CrossRef] 9. Muthaiyan Shanmugam, M.; Subhra Santra, T. Microfluidic Devices in Advanced Caenorhabditis elegans Research. Molecules 2016, 21, 1006. [CrossRef] 10. Bhatia, S.N.; Ingber, D.E. Microfluidic organs-on-chips. Nat. Biotechnol. 2014, 32, 760–772. [CrossRef] 11. Edington, C.D.; Chen, W.L.K.; Geishecker, E.; Kassis, T.; Soenksen, L.R.; Bhushan, B.M.; Freake, D.; Kirschner, J.; Maass, C.; Tsamandouras, N.; et al. Interconnected Microphysiological Systems for Quantitative Biology and Pharmacology Studies. Sci. Rep. 2018, 8, 4530. [CrossRef] 12. Huh, D.; Matthews, B.D.; Mammoto, A.; Montoya-Zavala, M.; Hsin, H.Y.; Ingber, D.E. Reconstituting organ-level lung functions on a chip. Science 2010, 328, 1662–1668. [CrossRef] 13. Huh, D.; Leslie, D.C.; Matthews, B.D.; Fraser, J.P.; Jurek, S.; Hamilton, G.A.; Thorneloe, K.S.; McAlexander, M.A.; Ingber, D.E. A human disease model of drug toxicity-induced pulmonary edema in a lung-on-a-chip microdevice. Sci. Transl. Med. 2012, 4, 159ra147. [CrossRef] 14. Marx, U.; Walles, H.; Hoffmann, S.; Lindner, G.; Horland, R.; Sonntag, F.; Klotzbach, U.; Sakharov, D.; Tonevitsky, A.; Lauster, R. ‘Human-on-a-chip’ Developments: A Translational Cutting- edge Alternative to Systemic Safety Assessment and Efficiency Evaluation of Substances in Laboratory Animals and Man? Lab. Anim. 2012, 40, 235–257. [CrossRef][PubMed] 15. Zhao, Y.; Kankala, R.; Wang, S.-B.; Chen, A.-Z. Multi-Organs-on-Chips: Towards Long-Term Biomedical Investigations. Molecules 2019, 24, 675. [CrossRef][PubMed] 16. Visone, R.; Gilardi, M.; Marsano, A.; Rasponi, M.; Bersini, S.; Moretti, M. Cardiac Meets Skeletal: What’s New in Microfluidic Models for Muscle Tissue Engineering. Molecules 2016, 21, 1128. [CrossRef] 17. Benam, K.H.; Dauth, S.; Hassell, B.; Herland, A.; Jain, A.; Jang, K.-J.; Karalis, K.; Kim, H.J.; MacQueen, L.; Mahmoodian, R.; et al. Engineered In Vitro Disease Models. Annu. Rev. Pathol. Mech. Dis. 2015, 10, 195–262. [CrossRef] 18. Higuita-Castro, N.; Nelson, M.T.; Shukla, V.; Agudelo-Garcia, P.A.; Zhang, W.; Duarte-Sanmiguel, S.M.; Englert, J.A.; Lannutti, J.J.; Hansford, D.J.; Ghadiali, S.N. Using a Novel Microfabricated Model of the Alveolar-Capillary Barrier to Investigate the Effect of Matrix Structure on Atelectrauma. Sci. Rep. 2017, 7, 11623. [CrossRef] 19. Kimura, H.; Sakai, Y.; Fujii, T. Organ/body-on-a-chip based on microfluidic technology for drug discovery. Drug Metab. Pharmacokinet. 2018, 33, 43–48. [CrossRef] 20. Lal, B.K.; Varma, S.; Pappas, P.J.; Hobson, R.W.; Durán, W.N. VEGF Increases Permeability of the Endothelial Cell Monolayer by Activation of PKB/akt, Endothelial Nitric-Oxide Synthase, and MAP Kinase Pathways. Microvasc. Res. 2001, 62, 252–262. [CrossRef] 21. Tiwari, G.; Tiwari, R.; Sriwastawa, B.; Bhati, L.; Pandey, S.; Pandey, P.; Bannerjee, S.K. Drug delivery systems: An updated review. Int. J. Pharm. Investig. 2012, 2, 2–11. [CrossRef] 22. Ghaffarian, R.; Muro, S. Models and Methods to Evaluate Transport of Drug Delivery Systems Across Cellular Barriers. J. Vis. Exp. 2013, 50638. [CrossRef] 23. Benam, K.H.; Villenave, R.; Lucchesi, C.; Varone, A.; Hubeau, C.; Lee, H.-H.; Alves, S.E.; Salmon, M.; Ferrante, T.C.; Weaver, J.C.; et al. Small airway-on-a-chip enables analysis of human lung inflammation and drug responses in vitro. Nat. Methods 2016, 13, 151–157. [CrossRef] Micromachines 2019, 10, 533 17 of 18

24. Shyy, J.Y.-J.; Chien, S. Role of Integrins in Endothelial Mechanosensing of Shear Stress. Circ. Res. 2002, 91, 769–775. [CrossRef] 25. Ingber, D.E. Mechanical Signaling and the Cellular Response to Extracellular Matrix in Angiogenesis and Cardiovascular Physiology. Circ. Res. 2002, 91, 877–887. [CrossRef] 26. Mehta, D.; Malik, A.B. Signaling Mechanisms Regulating Endothelial Permeability. Physiol. Rev. 2006, 86, 279–367. [CrossRef] 27. Thomas, A.; Wang, S.; Sohrabi, S.; Orr, C.; He, R.; Shi, W.; Liu, Y. Characterization of vascular permeability using a biomimetic microfluidic blood vessel model. Biomicrofluidics 2017, 11, 024102. [CrossRef] 28. Frost, T.S.; Estrada, V.; Jiang, L.; Zohar, Y. Convection-diffusion molecular transport in a microfluidic bilayer device with a porous membrane. Submitt. Microfluid. Nanofluidics 2019. 29. Song, K.-Y.; Zhang, H.; Zhang, W.-J.; Teixeira, A. Enhancement of the surface free energy of PDMS for reversible and leakage-free bonding of PDMS–PS microfluidic cell-culture systems. Microfluid. Nanofluidics 2018, 22, 135. [CrossRef] 30. van Meer, B.J.; de Vries, H.; Firth, K.S.A.; van Weerd, J.; Tertoolen, L.G.J.; Karperien, H.B.J.; Jonkheijm, P.; Denning, C.; IJzerman, A.P.; Mummery, C.L. Small molecule absorption by PDMS in the context of drug response bioassays. Biochem. Biophys. Res. Commun. 2017, 482, 323–328. [CrossRef] 31. Foster, K.A.; Oster, C.G.; Mayer, M.M.; Avery, M.L.; Audus, K.L. Characterization of the A549 Cell Line as a Type II Pulmonary Epithelial Cell Model for Drug Metabolism. Exp. Cell Res. 1998, 243, 359–366. [CrossRef] 32. Man, S.; Ubogu, E.E.; Williams, K.A.; Tucky, B.; Callahan, M.K.; Ransohoff, R.M. Human Brain Microvascular Endothelial Cells and Umbilical Vein Endothelial Cells Differentially Facilitate Leukocyte Recruitment and Utilize Chemokines for T Cell Migration. Clin. Dev. Immunol. 2008, 2008, 1–8. [CrossRef] 33. Maciag, T.; Hoover, G.; Weinstein, R. High and Low Molecular Weight Forms of Endothelial Cell Growth Factor. J. Biol. Chem. 1982, 257, 5333–5336. 34. McClatchey, A.I.; Yap, A.S. Contact inhibition (of proliferation) redux. Curr. Opin. Cell Biol. 2012, 24, 685–694. [CrossRef] 35. Hartsock, A.; Nelson, W.J. Adherens and tight junctions: Structure, function and connections to the actin cytoskeleton. Biochim. Biophys. Acta Biomembr. 2008, 1778, 660–669. [CrossRef] 36. Considine, R.V. Plasma Membrane, Membrane Transport, and Resting Membrane Potential. In Medical Physiology: Principles for Clinical Medicine; Williams & Wilkins: Philadelphia, PA, USA, 2013; ISBN 1-4511-1039-1. 37. Nemoto, R.; Yamamoto, S.; Ogawa, T.; Naude, R.; Muramoto, K. Effect of Chum Salmon Egg Lectin on Tight Junctions in Caco-2 Cell Monolayers. Molecules 2015, 20, 8094–8106. [CrossRef] 38. Masereeuw, R.; Moons, M.M.; Toomey, B.H.; Russel, F.G.M.; Miller, D.S. Active Lucifer Yellow Secretion in Renal Proximal Tubule: Evidence for Organic Anion Transport System Crossover. J. Pharmacol. Exp. Ther. 1999, 289, 8. 39. Azizi, P.M.; Zyla, R.E.; Guan, S.; Wang, C.; Liu, J.; Bolz, S.-S.; Heit, B.; Klip, A.; Lee, W.L. Clathrin-dependent entry and vesicle-mediated exocytosis define insulin transcytosis across microvascular endothelial cells. Mol. Biol. Cell 2015, 26, 740–750. [CrossRef] 40. Matsukawa, Y.; Lee, V.H.L.; Crandall, E.D.; Kim, K.-J. Size Dependent Dextran Transport across Rat Alveolar Epithelial Cell Monolayers. J. Pharm. Sci. 1997, 86, 305–309. [CrossRef] 41. Michael Danielsen, E.; Hansen, G.H. Small molecule pinocytosis and clathrin-dependent endocytosis at the intestinal brush border: Two separate pathways into the enterocyte. Biochim. Biophys. Acta Biomembr. 2016, 1858, 233–243. [CrossRef] 42. Kobayashi, S.; Kondo, S.; Kazuhiko, J. Permeability of Peptides and Proteins in Human Cultured Alveolar A549 Cell Monolayer. Pharm. Res. 1995, 12, 1115–1119. [CrossRef] 43. Fotticchia, I.; Guarnieri, D.; Fotticchia, T.; Falanga, A.P.; Vecchione, R.; Giancola, C.; Netti, P.A. Energetics of ligand-receptor binding affinity on endothelial cells: An in vitro model. Colloids Surf. B Biointerfaces 2016, 144, 250–256. [CrossRef] 44. Li, Y.; Gao, A.; Yu, L. Monitoring of TGF-β 1-Induced Human Lung Adenocarcinoma A549 Cells Epithelial-Mesenchymal Transformation Process by Measuring Cell Adhesion Force with a Microfluidic Device. Appl. Biochem. Biotechnol. 2016, 178, 114–125. [CrossRef] Micromachines 2019, 10, 533 18 of 18

45. Nalayanda, D.D.; Puleo, C.; Fulton, W.B.; Sharpe, L.M.; Wang, T.-H.; Abdullah, F. An open-access microfluidic model for lung-specific functional studies at an air-liquid interface. Biomed. Microdevices 2009, 11, 1081–1089. [CrossRef] 46. Feng, S.; Zou, L.; Wang, H.; He, R.; Liu, K.; Zhu, H. RhoA/ROCK-2 Pathway Inhibition and Tight Junction Protein Upregulation by Catalpol Suppresses Lipopolysaccaride-Induced Disruption of Blood-Brain Barrier Permeability. Molecules 2018, 23, 2371. [CrossRef] 47. Harris, E.S.; Nelson, W.J. VE-cadherin: at the front, center, and sides of endothelial cell organization and function. Curr. Opin. Cell Biol. 2010, 22, 651–658. [CrossRef] 48. Erickson, H.P. Size and Shape of Protein Molecules at the Nanometer Level Determined by Sedimentation, Gel Filtration, and Electron Microscopy. Biol. Proced. Online 2009, 11, 32–51. [CrossRef] 49. Ravetto, A.; Hoefer, I.E.; den Toonder, J.M.J.; Bouten, C.V.C. A membrane-based microfluidic device for mechano-chemical cell manipulation. Biomed. Microdevices 2016, 18, 31. [CrossRef] 50. Bischoff, I.; Hornburger, M.C.; Mayer, B.A.; Beyerle, A.; Wegener, J.; Fürst, R. Pitfalls in assessing microvascular endothelial barrier function: impedance-based devices versus the classic macromolecular tracer assay. Sci. Rep. 2016, 6, 23671. [CrossRef] 51. Tarbell, J.M. Shear stress and the endothelial transport barrier. Cardiovasc. Res. 2010, 87, 320–330. [CrossRef] 52. Fujita, T.; Kawahara, I.; Yamamoto, M.; Yamamoto, A.; Muranishi, S. Comparison of the Permeability of Macromolecular Drugs Across Cultured Intestinal and Alveolar Epithelial Cell Monolayers. J. Pharm. Pharmacol. 1995, 1, 231–234. 53. Bruewer, M.; Nusrat, A. Regulation of Paracellular Transport across Tight Junctions by the Actin Cytoskeleton. In Tight Junctions; Gonzalez-Mariscal, L., Ed.; Springer US: Boston, MA, USA, 2006; pp. 135–145. ISBN 978-0-387-36673-9. 54. Sidhaye, V.K.; Schweitzer, K.S.; Caterina, M.J.; Shimoda, L.; King, L.S. Shear stress regulates aquaporin-5 and airway epithelial barrier function. Proc. Natl. Acad. Sci. USA 2008, 105, 3345–3350. [CrossRef] 55. Kang, H.; Cancel, L.M.; Tarbell, J.M. Effect of shear stress on water and LDL transport through cultured endothelial cell monolayers. Atherosclerosis 2014, 233, 682–690. [CrossRef] 56. Seki, S.; Kobayashi, M.; Itagaki, S.; Hirano, T.; Iseki, K. Contribution of organic anion transporting polypeptide OATP2B1 to amiodarone accumulation in lung epithelial cells. Biochim. Biophys. Acta Biomembr. 2009, 1788, 911–917. [CrossRef] 57. Ito, S.; Osaka, M.; Higuchi, Y.; Nishijima, F.; Ishii, H.; Yoshida, M. Indoxyl Sulfate Induces Leukocyte- Endothelial Interactions through Up-regulation of E-selectin. J. Biol. Chem. 2010, 285, 38869–38875. [CrossRef] 58. Cole, L.; Coleman, J.; Kearns, A.; Morgan, G.; Hawes, C. The organic anion transport inhibitor, probenecid, inhibits the transport of Lucifer Yellow at the plasma membrane and the tonoplast in suspension- cultured plant cells. J. Cell Sci. 1991, 99, 545–555. 59. Mishima, M.; Hamada, T.; Maharani, N.; Ikeda, N.; Onohara, T.; Notsu, T.; Ninomiya, H.; Miyazaki, S.; Mizuta, E.; Sugihara, S.; et al. Effects of Uric Acid on the NO Production of HUVECs and its Restoration by Urate Lowering Agents. Drug Res. 2016, 66, 270–274. [CrossRef] 60. Tumova, S.; Kerimi, A.; Williamson, G. Long term treatment with quercetin in contrast to the sulfate and glucuronide conjugates affects HIF1α stability and Nrf2 signaling in endothelial cells and leads to changes in glucose metabolism. Free Radic. Biol. Med. 2019, S0891584919302576. [CrossRef] 61. Wasserman, S.M.; Mehraban, F.; Komuves, L.G.; Yang, R.-B.; Tomlinson, J.E.; Zhang, Y.; Spriggs, F.; Topper, J.N. Gene expression profile of human endothelial cells exposed to sustained fluid shear stress. Physiol. Genom. 2002, 12, 13–23. [CrossRef] 62. Kudo, S.; Ikezawa, K.; Ikeda, M.; Oka, K.; Tanishita, K. Albumin concentration profile inside cultured endothelial cells exposed to shear stress. Am. Soc. Mech. Eng. Bioeng. Div. 1997, 35, 547–548. 63. Duan, Y.; Weinstein, A.M.; Weinbaum, S.; Wang, T. Shear stress-induced changes of membrane transporter localization and expression in mouse proximal tubule cells. Proc. Natl. Acad. Sci. USA 2010, 107, 21860–21865. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).