ABSTRACT Title of Dissertation: A TISSUE-ENGINEERED PLACENTAL BARRIER MODEL FOR TOXICOLOGY AND PHARMACOLOGY APPLICATIONS Navein Arumugasaamy, Doctor of Philosophy, 2019 Dissertation directed by: John P. Fisher, Fischell Family Distinguished Professor & Department Chair, Fischell Department of Bioengineering Throughout history, there have been two major instances where a substance caused thousands of birth defects, yet it took a few years for the causation to be noted: thalidomide, in the late 1950s and early 1960s, and Zika Virus, just recently in 2014 to 2016. In both instances, the developing fetus was indirectly exposed to the substance through the placental barrier. Pregnant women took thalidomide as a medication or were stung by mosquitos and exposed to Zika Virus. These examples clearly show why models of the placental barrier and downstream fetal tissues are critically needed. Herein, I present our work on the development and utilization of a biomimetic placenta-fetus model. The three objectives in this work were to: (1) develop and validate the tissue-engineered BPB model through study of biologically relevant substances; (2) assess the effects of SSRIs on the BPB’s cells and evaluate the drugs’ transport profile across the barrier; and, (3) assess how SSRIs influence cardiomyocyte signaling and injury biomarker release following passage through the BPB. We suggest that this work provides a critically needed and biologically relevant placenta-fetus model, useful as a method to assess pharmacology and toxicology properties of medications and other substances. Moreover, the knowledge gained through the studies performed may hopefully improve clinical care of pregnant women through enhanced understanding of how a medication impacts both the pregnant mother-to-be and her developing fetus. A TISSUE-ENGINEERED PLACENTAL BARRIER MODEL FOR TOXICOLOGY AND PHARMACOLOGY APPLICATIONS by Navein Arumugasaamy Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2019 Advisory Committee: Professor John P. Fisher, Chair Professor Peter C.W. Kim Associate Professor Helim Aranda-Espinoza Assistant Professor Kimberly M. Stroka Associate Professor Kan Cao, Dean’s Representative © Copyright by Navein Arumugasaamy 2019 Acknowledgements I would like to first acknowledge my advisors, John P. Fisher, PhD, and Peter C.W. Kim, MD, PhD, for their mentorship throughout this challenging process. I would also like to acknowledge my lab group and colleagues, both past and present, in the Tissue Engineering and Biomaterials Laboratory at the University of Maryland for their continued feedback and assistance throughout graduate school. In addition, I give my sincere thanks and gratitude to my colleagues and the staff at the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children’s National Medical Center for making the three years I spent there an enjoyable experience. In particular, Stéphanie Val, PhD and the Fernandes Lab group (Rohan Fernandes, PhD, Lizie Sweeney, PhD, Juliana Cano-Mejia, and Rachel Burga) for their constant support and assistance. Lastly, I would like to acknowledge the efforts of my undergraduate mentees in helping to complete these studies and for making me a better mentor: Leila E. Ettehadieh, Alana Gudelsky, and Amelia C. Hurley-Novatny. I acknowledge that I could not have completed this work on my own, and so, for everyone I have met along the way who has helped to shape my graduate school experience, I am grateful. ii Table of Contents Acknowledgements ....................................................................................................... ii Table of Contents ......................................................................................................... iii List of Tables ................................................................................................................ v List of Figures .............................................................................................................. vi Chapter 1: Introduction ................................................................................................. 1 1.1 Significance......................................................................................................... 1 1.2 Thesis objectives ................................................................................................. 3 Chapter 2: Models for Studying Epithelial Tissue Barriers .......................................... 5 2.1 Introduction ......................................................................................................... 5 2.2 Skin Epithelium .................................................................................................. 9 2.3 The Gastrointestinal (GI) Tract ......................................................................... 17 2.4 Lungs and the Blood-Air Barrier ...................................................................... 22 2.5 Endothelium and the Blood-Brain Barrier ........................................................ 26 2.6 Placenta and the Blood-Placenta Barrier .......................................................... 30 2.7 Comparison of Tissues and Barriers ................................................................. 35 2.8 Conclusions and Summary ............................................................................... 41 Chapter 3: The Human Placenta and Modeling Approaches ...................................... 42 3.1 In Vivo Human and Animal Placental Barriers ................................................ 42 3.2 In Vitro Placental Barrier Models ..................................................................... 43 Chapter 4: A Placenta-Fetus Model for Evaluating Maternal-Fetal Transmission and Fetal Neural Toxicity of Zika Virus............................................................................ 45 4.1 Introduction ....................................................................................................... 45 4.2 Materials and Methods ...................................................................................... 47 4.3 Results ............................................................................................................... 57 4.4 Discussion ......................................................................................................... 67 4.5 Summary ........................................................................................................... 71 Chapter 5: Assessing Placental Barrier Phenotypic Response to SSRIs Fluoxetine and Sertraline ..................................................................................................................... 73 5.1 Introduction ....................................................................................................... 73 5.2 Materials and Methods ...................................................................................... 78 5.3 Results ............................................................................................................... 83 5.4 Discussion ......................................................................................................... 93 5.5 Summary ........................................................................................................... 99 Chapter 6: Assessing the Impact of Fluoxetine and Sertraline on Cardiomyocytes Downstream of Placenta Using Placental Barrier Model ......................................... 100 6.1 Introduction ..................................................................................................... 100 6.2 Materials and Methods .................................................................................... 104 6.3 Results ............................................................................................................. 108 6.4 Discussion ....................................................................................................... 117 6.5 Summary ......................................................................................................... 122 Chapter 7: Summary, Contribution, and Future Directions ..................................... 123 7.1 Summary ......................................................................................................... 123 7.2 Contributions................................................................................................... 126 iii 7.3 Future Directions ............................................................................................ 127 Appendix A: Determination the Diffusion Coefficient Experimentally ................... 129 Bibliography ............................................................................................................. 133 iv List of Tables Table 2.1: Summary of key physiological and modeling information for each tissue. Type of barrier, key barrier cells, and supporting barrier cells are all biological information. Models used are different approaches discussed, where Transwell includes semi-permeable membranes and 3D Tissue Gels includes bioprinting approaches. Cells used refers to both primary cells and cell lines, and model properties outlines common characteristics of models……………………………..36 v List of Figures Fig. 2.1. The four types of interfacial barriers discussed in this review. (A) The air-tissue interface shows molecular transport (purple circles) from air, across the barrier layer, into the tissue space (i.e. skin or lungs). (B) The air-liquid interface shows molecular transport (yellow diamonds) from air, across the barrier layer(s) and tissue, into liquid, which is often the vasculature of the body, represented by blue/red vessels (i.e. skin). (C) The liquid-tissue interface shows molecular transport (orange circles) from liquid, across the barrier layer, into the tissue space (i.e. the brain, from vasculature into tissue). (D) The liquid-liquid interface shows molecular transport (green diamonds) from liquid, across the barrier layer(s) and tissue, into liquid (i.e. the placenta, from one vascular source into another)………………………………....8 Fig. 2.2. Models of transport through skin. Traditional models for skin transport include (A) liquid-to-liquid interface systems, usually Franz diffusion cells where the donor chamber acts as the apical side of the membrane and the receiver side acts as the basolateral side. Skin models still widely rely on variation of the transwell inserts (B-C), where the exposure of the barrier to air allows assessing the physiological behavior of keratinocytes and the external role of skin. (B) Topical application is modeled by applying the treatment on the air-exposed (apical) side of the barrier, while (C) systemic application is modeled by delivering the treatment via the liquid- exposed (basolateral) side……………………………………………………………14 Fig. 2.3. The GI tract barrier and models. (A) The GI tract barrier is composed of mucus and a multi-cellular layer. Enterocytes are depicted in beige while additional cell types are depicted in green, with only a small portion of the entire barrier depicted here. (Adapted from Peterson, et al., Nat Rev Immunol, 2014.) (B) Cells seeded within a transwell insert are commonly used to model the pathway between the apical and basolateral side of the GI tract. (C) Cells within an organoid, depicted as blue cells here, have been utilized to evaluate transport from the exterior of the organoid towards the interior………………………………………………..………18 Fig. 2.4. The blood-air barrier within the lungs. (A) Alveolar epithelial cells lie within the bronchioles, at the interface of air within the lungs and the local vasculature. The two types of alveolar cells are depicted in orange and blue. (B) A co- culture model of multiple cell types within the lung containing alveolar epithelial cells (yellow), endothelial cells (orange), and immune cells (gray and green). (Adapted from Klein et al., Part Fibre Toxicol, 2013.)……………………………...23 Fig. 2.5. The Blood-Brain Barrier. (A) Schematic depicting molecules (yellow) crossing the vasculature to enter into the brain space with numerous brain cells (depicted in green, yellow, and blue). Molecules within the lumen of the vasculature must traverse the brain endothelial cells and basement membrane before reaching the brain space. (B) Illustration of a microfluidic platform to mimic the BBB, where micropillars are used within the cell-laden area to better mimic the cellular niche. (Adapted from Prabhakarpandian, et al., Lab Chip, 2013.)………………………....27 vi Fig. 2.6. The Placental Barrier. (A) Physiology of the placental barrier, with maternal blood interacting with the syncytiotrophoblast and fetal endothelial cells interacting with fetal blood. (B) A common microfluidic approach to building a cell- model, with media inputs and outputs shown. Cells are seeded in the middle portion of the device, where both apical and basolateral media interact with cells. (C) An Ussing chamber schematic, where a tissue section is held between two compartments, creating an apical and basolateral chamber for measurement of ion transport. A current source and voltmeter is used to regularly measure differences between the two chambers………………………………………………………………………..…..32 Fig. 4.1. Overview of BPB Model and Placenta-Fetus System. (A) Maternal decidua, placenta, amniotic fluid, and fetus. (B) From the maternal side, the decidual vasculature feeds blood into the placenta that interacts with fetal capillary trees, forming the blood-placenta barrier. Fetal blood within these capillaries feed blood into the umbilical cord. (C) Blood-placenta barrier composition where maternal blood interacts with the syncytiotrophoblast. Underneath is a layer of villous cytotrophoblasts, followed by basement membrane containing placental macrophages, called Hofbauer cells, and then fetal endothelial cells in direct contact with fetal blood. (D) Schematic of organization within our multi-layer model: BeWo b30 cells are encapsulated is GelMA hydrogel with more BeWo b30 cells seeded on one side and HUVECs seeded on the other side. (E) Image of a BPB model. (F) Models are cast with BeWo b30 cells and crosslinked on Day 0. Following 7 days of BeWo b30 media, HUVECs are seeded and models fed HUVEC media until the transport study and other assays are performed on Day 10 PS. (G-K) Transport assay (schematic, G, and image, K) with BPB model (gray in G) and plug (H/I) within a transwell insert (J). Using this plug (I), transport is limited to passing through the gel (yellow arrows, G)……………………………………………………………..…..58 Fig. 4.2. Formation of a Cell-Cell Barrier. (A) Cross-section view of BPB model via 3D rendering of confocal z-stack of nuclei cells, showing two distinct layers of cells (one trophoblast and one endothelial cell layer) within a single construct. Scale bar = 200µm. (n=3). (B) ZO-1 expression (green) with cells’ nuclei (cyan). The merge indicates intercellular ZO-1 expression for HUVECs on day 9 following seeding. Scale bar = 50µm. (n=3) (C) E-Cadherin (BeWo b30) and VE-Cadherin (HUVEC) expression (both green) with cells’ nuclei (cyan). The merge indicates intercellular cadherin expression for HUVECs on day 9 following seeding. Scale bar = 50µm. (n=3) (D) TEER testing for each layer of the multilayer BPB model, with the blank gel (black) having the lowest value. The next highest is BeWo b30 cells (red) seeded on top of the gel, followed by HUVECs (blue) seeded on top. BeWo b30 cells encapsulated within the hydrogel (green) and the full multilayer model (purple) showed the highest TEER values over 9 days. * indicates significance (p<0.05) for the group at a given time point, compared to the blank gel. (n=3) (E) Fluorescent staining showing nuclei (cyan) and actin (magenta) at 3, 5, 7, and 9 days post- seeding, showing cell proliferation within each layer, towards the formation of a monolayer. Scale bar = 50µm. (n=3)…………………………………….………60 vii Fig. 4.3. Cellular Bioactivity, Specificity, and Model Functionality. (A) Live-Dead images showing the vast majority of cells are alive (green) with little to no dead cells (red) at day 9 PS. Scale bar = 200µm. (n=6). (B) Positive expression of human endogenous retroviral protein (HERV) for BeWo b30 cells used within the model. Scale bar = 100µm. (n=4). (C/D/E) Secreted human chorionic gonadotropin (hCG) (C), progesterone (D), and vascular endothelial growth factor (VEGF) (E) concentrations for BeWo b30 cells and HUVECs, on tissue culture polystyrene (TCPS), a GelMA hydrogel, or within co-culture as the full model. (n≥4, groups that do not share a letter are significantly different, p<0.05) (F/G) Glucose (F) or IgG (G) was added to the apical compartment and the concentration in both the apical compartment (donor; shown with filled-in circles) and the basolateral compartment (receiver; shown with empty circles) was measured for a blank acellular gel (black) or a full multilayer cell-laden model (red). * indicates significant (p<0.05) difference between the blank gel and full model gel at the same time point. (F, n=3; G, n=4)…………………………………………………………………..……….…….63 Fig. 4.4. ZIKV Infection, Transmission Across BPB Model, and NPC Outcomes. (A) Fold change in ZIKV infection (MOI 1) indicating trophoblast cells (BeWo) are more susceptible to infection than endothelial cells (HUVEC). Data is normalized to the BeWo group. (n=2) (B) Representative flow cytometry plots of ZIKV infection or mock control treatment for trophoblast cells (BeWo) and endothelial cells (HUVEC) 4 days post infection. (n=2) (C) ZIKV present within the media in the fetal (basolateral) compartment at each time point was determined by qRT-PCR for blank gels, full model and full model treated with chloroquine (n=3). Standards with known PFU concentrations were included in the qRT-PCR to calculate the values from the samples. The full model gels show peak fetal PFU at 12 hours, whereas blank and chloroquine-treated samples do not. (n=3) (D) Full model gels with or without chloroquine at 4 days post infection showing Zika M protein. Scale bar = 100µm. (n=3) (E) Viability of neural progenitor cells (NPCs) in the fetal compartment of our system 7 days after infection. ZIKV added to either a blank acellular gel, a full model cell-laden gel, or directly to NPC without model present. **p<0.01. (n=3)………………………………………………………………………………..65 Fig. 4.5. Chloroquine dose effects on cell viability. Microscopy images depicting live (green) or dead (red) cells, either BeWo b30 or HUVEC, following exposure to chloroquine for 4 days. A lack of image indicates no cells present. Scale bar = 50µm (n=4).……………………………………………………………………….…..….66 Fig. 5.1. Evaluation of Endothelial Cell Type and Placental ECM. (A) TEER values increased over 14 days with minimal differences between HBMEC or HUVEC in co-culture with BeWo b30 cells. (B) Fluorescein sodium salt shows similar transport profiles across the barrier models using either HUVEC or HBMEC in the model. (C) Permeability for fluorescein sodium salt is slightly higher when using HUVECs in the model instead of HBMEC, but is comparable when evaluating immunoglobulin G transport. (D) The top 10 protein families present within the viii solubilized placental ECM after proteomics analysis, by peptide fraction. (E) Relative DNA and protein content for the solubilized ECM compared to the dry tissue recovered after lyophilization of the decellularized tissue. (F) TEER values for BeWo b30 cells grown on GelMA with either 1% or 10% (v/v) ECM in the hydrogel, showing minimal differences between the two concentrations. (G) TGFβ secretions for HUVECs and BeWo b30 cells on GelMA-coated dishes with (+) or without (-) placental ECM (10% v/v) included in the coating. Addition of ECM significantly decreases TGFβ secretions for BeWo b30 cells………………………………….…85 Fig. 5.2. Transport profile for SSRIs crossing the BPB. (A) Relative fluoxetine concentration over the first 8 hours following addition of fluoxetine to the apical (donor) side, with and without inhibitor (elacridar) treatment prior to the addition of the drug. (B) Relative sertraline concentration over the first 8 hours following addition of sertraline to the apical side, with and without inhibitor treatment prior to the addition of the drug. (C) Relative fluoxetine concentration over the first 8 hours of day 3 (i.e. hours 48 to 56) following addition of the drug to the apical side of the barrier, with or without inhibitor treatment prior to the addition of the drug. Here, after multiple days of treatment, the basolateral side (receiver) is not at a relative concentration of 0%. (D) Calculated partition coefficients for each drug, and diffusion coefficients for the transport profiles within the first 8 hours following treatment at days 1 and 3, with and without inhibitor. (E) Positive expression of P-glycoprotein (green) for BeWo b30 cells and HUVECs, with a nuclear counterstain (blue). Scale bar = 100µm. (F) Positive expression of breast cancer resistance protein (green) for BeWo b30 cells and HUVECs, with a nuclear counterstain (blue). Scale bar = 100µm……………………………………………………………………………...87 Fig. 5.3. Barrier Dose Response and Direct Drug Comparison. (A/B) ICAM secretions for the coculture model (BeWo b30 and HUVEC) at different dosages of (A) fluoxetine or (B) sertraline after 1 and 3 days of exposure. (C-E) Coculture model secretions for (C) ICAM, (D) TGFβ, or (E) VCAM when exposed to no drug or fluoxetine or sertraline, at a drug dosage of 10,000 ng/mL following 1 day of treatment. (F) TEER values over 9 days following initial exposure (day 7 post- seeding cells) to fluoxetine or sertraline at 10,000 ng/mL with fresh exposure daily. Groups that do not share a letter are significantly (*p<0.05) different……………..89 Fig. 5.4. Secretions of CAM molecules by BPB cell type. (A/B) ICAM secreted by BeWo b30 cells in response to varying dosages of (A) fluoxetine or (B) sertraline after 1 and 3 days of repeated exposure. (C/D) ICAM secreted by HUVECs in response to varying dosages of (C) fluoxetine or (D) sertraline after 1 and 3 days of repeated exposure. (E/F) VCAM secreted by HUVECs in response to varying dosages of (E) fluoxetine or (F) sertraline after 1 and 3 days of repeated exposure. Groups not sharing the same letter within a single timepoint are significantly (p<0.05) different and * indicates p<0.05 across two time points for the same group……....91 Fig. 5.5. Secretion of TGFβ by BPB cell type. (A/B) TGFβ secreted by BeWo b30 cells in response to varying dosages of (A) fluoxetine or (B) sertraline after 1 and 3 ix days of repeated exposure. (C/D) TGFβ secreted by HUVECs in response to varying dosages of (C) fluoxetine or (D) sertraline after 1 and 3 days of repeated exposure. Groups not sharing the same letter within a single timepoint are significantly (p<0.05) different and * indicates p<0.05 across two time points for the same group………92 Fig. 6.1. Cardiomyocyte calcium handling response to SSRI exposure. Calcium intensity for cardiomyocytes was quantified following direct exposure to SSRIs (either fluoxetine or sertraline) using multiple metrics: (A) the peak-to-peak time (i.e. the period of the oscillations), (B) the frequency of oscillations, (C) the time to reach the peak, (D) the decay time, from peak to a minimum value, (E) the duration time at 30% calcium reuptake, and (F) the duration time at 80% calcium reuptake. *p<0.05 for the two groups indicated………………………………………………………..109 Fig. 6.2. Cardiomyocyte calcium handling response following SSRI exposure through BPB model. Calcium intensity for cardiomyocytes in the basolateral compartment was quantified following exposure to SSRIs (either fluoxetine or sertraline) through a BPB model, using these metrics: (A) the peak-to-peak time (i.e. the period of the oscillations), (B) the frequency of oscillations, (C) the time to reach the peak, (D) the decay time, from peak to a minimum value, (E) the duration time at 30% calcium reuptake, and (F) the duration time at 80% calcium reuptake. *p<0.05 for the two groups indicated……………………………………………………..…111 Fig. 6.3. Cardiomyocyte Injury Marker Secretions in Response to Drug Exposure. (A/B) Creatine Kinase MB secretions for cardiomyocytes either (A) directly exposed to the drugs (fluoxetine, F, or sertraline, S) at one of two dosages (100 or 1000 ng/mL), with a non-exposed (NoE) negative control, or (B) indirectly exposed to the drugs, via additional to apical compartment of placental barrier and secretions in basolateral compartment assessed. (C/D) Troponin T secretions for cardiomyocytes either (C) directly exposed to the drug, or (D) indirectly exposed to the drug, with annotation as stated above. *p<0.05 for the two groups indicated……………………………………………………………………………113 Fig. 6.4. Cardiomyocyte secretions of NT-proBNP following exposure to SSRIs. (A) Secretions of NT-proBNP from cardiomyocytes following 1 day of exposure to SSRIs (either sertraline, S, or fluoxetine, F) at one of three dosages (10, 100, or 1000 ng/mL), or no drug exposure (None). (B) Secretions in the basolateral compartment, where cardiomyocytes were indirectly exposed to the drug following passage through a BPB model. (C) Secretions in the apical compartment, indicating protein that was secreted in the basolateral compartment and passed through the BPB model. *p<0.05 for the two groups indicated…………………………………………………….…115 Fig. 6.5. Secretions of VCAM following exposure to SSRIs. (A) Secretions of VCAM from cardiomyocytes following 1 day of exposure to SSRIs (either sertraline, S, or fluoxetine, F) at one of three dosages (10, 100, or 1000 ng/mL), or no drug exposure (None). (B) Secretions in the basolateral compartment, where cardiomyocytes were indirectly exposed to the drug following passage through a BPB x model. (C) Secretions in the apical compartment, indicating protein that was secreted in the basolateral compartment and passed through the BPB model or protein secreted by the cells within the BPB model in response to drug exposure. *p<0.05 for the two groups indicated………………………………………………………………….116 xi Chapter 1: Introduction 1.1 Significance In the late 1950s and early 1960s, there was an increase in the number of babies being born throughout the world with severe limb malformations. Unfortunately, it took a couple of years for physicians and researchers to realize that the cause of limb malformations were resulting from pregnant women being prescribed thalidomide, a medication used as a sedative and meant to help alleviate morning sickness in pregnant women1. Decades later, in 2015, physicians noted a rise in the cases of microcephaly in Brazil. Within the next year or so, the association between Zika Virus exposure and the resulting microcephaly in newborns was realized2–4, even though the first viral outbreaks in French Polynesia were reported in late 2013. In both instances, a substance (medication and virus, respectively) was able to cross the maternal-fetal interface, termed the blood-placenta barrier (BPB), and impact fetal development. Both of these historical events indicate that, despite decades of advancements in medical science, we still are severely lacking is methods for assessing fetal safety during pregnancy. The lack of methods to assess medication safety during pregnancy is a growing concern given that nearly 90% of pregnant women report taking at least one medication throughout the course of pregnancy5. Some of these medications include antibiotics and the influenza vaccine, whereas others include antidepressants, such as selective serotonin reuptake inhibitors (SSRIs)5,6. Notably, it is widely accepted that SSRIs, such as sertraline and fluoxetine, readily cross the BPB and enter into the fetal 1 compartment, with fetal concentrations from a fraction of that found in maternal serum to concentrations exceeding maternal serum values7. Moreover, SSRIs are thought to influence fetal development, including neurological development and cardiovascular development8–13. Controversially though, others have argued that SSRIs carry no risks significantly greater than normal14,15. Ultimately, this will be resolved through greater understanding of how SSRIs influence individual cell types, and whether the passage of SSRIs through the BPB can lead to the drugs impacting healthy fetal development. To better understand the extent of these drugs’ impacts, improved “tools” are necessary. Examples like thalidomide could be prevented with improved tools to determine whether medications cross the BPB. Similar tools could be utilized to assess potential effects of a virus, such as Zika Virus, on the placental barrier and downstream fetal tissue. Ideally, these “tools” would be rigorous in vivo studies evaluating medication and substance safety. However, given the obvious ethical limitations to conducting these studies in humans, and the physiological differences between humans and other animals in placenta and fetal development16, biomimetic in vitro models are critically needed. Moreover, models that recapitulate both the BPB and downstream fetal tissue would enable probing of maternal-fetal transport of a substance, as well as evaluation of phenotypic effects within the developing fetus. Herein, I describe our approach for developing, validating, and using a tissue- engineered placental barrier model to study Zika Virus and prescription antidepressants. 2 1.2 Thesis objectives The overall goal of this work was to develop a biomimetic placenta-fetus model, and utilize it to understand how drugs and substances influence the placental barrier and fetal cells downstream of the barrier. The work described herein had three objectives: (1) develop and validate the tissue-engineered BPB model through study of biologically relevant substances; (2) assess the effects of SSRIs on the BPB’s cells and evaluate the drugs’ transport profile across the barrier; and, (3) assess how SSRIs influence cardiomyocyte signaling and injury biomarker release following passage through the BPB. In the first objective, we utilized a gelatin methacrylate (GelMA) hydrogel with a trophoblast cell line and primary endothelial cells to fabricate a BPB model. This was validated through transepithelial electrical resistance (TEER) testing, staining for intercellular junctions (cadherins and zonula occludens-1), assessing transport of small and large molecules (glucose and immunoglobulin G, respectively), and analysis of bioactive secretions (human chorionic gonadotropin, progesterone, and vascular endothelial growth factor). We then utilized this to study maternal-fetal Zika Virus transport. In the second objective, we utilized this model to assay selective serotonin reuptake inhibitors, as they have been shown to impact endothelial cell health and readily cross the placental barrier. We assessed the effect of endothelial cell choice and addition of extracellular matrix (ECM). Then, we evaluated the maternal-fetal transport concentration profiles for both fluoxetine and sertraline, characterizing their partition coefficient and diffusion coefficient. We assessed how different dosages and 3 lengths of exposure impact secretions of cell adhesion molecules, transforming growth factor beta, and TEER values. We further characterized how both drugs influence endothelial cells and trophoblast cells differentially, to better understand the extent to which each cell type within the BPB is impacted by these drugs. In the third objective, we utilized this model with cardiomyocytes to assay how these SSRIs influence calcium handling and cardiomyocyte cell secretions. We quantified and characterized the calcium handling in terms of period, oscillation, time to peak, decay time, and duration time at 30% and 80% calcium transients. We also assayed cardiac injury biomarkers (creatine kinase MB, troponin T, and N-terminal pro-B-type natriuretic peptide) and vascular cell adhesion molecule in response to drug exposure. To decouple the effect of the BPB model from the drugs, we characterized the effect of the drugs when the cardiomyocytes were directly exposed to the drugs and when the drug exposure occurred following the drug’s passage through the BPB. Through the work done for each objective, we achieved the overall goal and have added knowledge in the fields of tissue engineering, placental biology, virology, pharmacology, and toxicology. I begin with a chapter related to in vitro models of epithelial tissue and a chapter on placental barrier models, before discussing the research performed to address the three objectives stated above. 4 Chapter 2: Models for Studying Epithelial Tissue Barriers1 2.1 Introduction Epithelial tissues are key regulators of physiologic homeostasis within humans, providing protection from foreign substances while also regulating transport of nutrients necessary for survival. Epithelial cells comprise the largest tissue in the human body, ranging from skin and vasculature to the placenta. Some instances of the physiological barrier are multilayer and multicell-type in nature and significantly influence the regulation of transport across the barrier. These added complexities are particularly important in various instances. In the placenta, for example, multicellular barriers are critical in regulating nutrient transfer between mother and fetus. Therefore, understanding the tissue-specific physiology and regulation will allow for improving the design process of therapeutics and biomedical treatments. Additionally, by comparing epithelial tissues that are seemingly unrelated, we find similarities in approach in development and validation of the model, which could in turn improve the therapeutic development process through the use of in vitro models in place of animal studies. This is particularly true for the use of tissue models to gain new knowledge about substance transport across these tissues, which is invaluable in drug development for optimizing delivery kinetics, testing efficacy, and potentially reducing necessary dosages, potential side effects, and cost. With regard to endogenous substances or exogenous drugs, the use of tissue models provides a 1 Adapted from: N Arumugasaamy, J Navarro, JK Leach, PCW Kim, and JP Fisher. In Vitro Models for Studying Transport Across Epithelial Barriers. Annals of Biomedical Engineering. In Press. 5 robust and rigorous method for investigating molecular interactions, and in some instances, such as for the placenta, can enable studies that otherwise have key ethical limitations. Recent efforts to model epithelial tissues have utilized both two-dimensional (2D) and three-dimensional (3D) approaches17–20. Traditionally, 2D models are cells seeded on a transwell insert, creating apical and basolateral compartments as occurs in vivo21. Over the past few years, 2D-like organ-on-a-chip models have grown in prominence. These platforms facilitate the biomimicry of key aspects relevant to the tissue-specific interface, such as contractions and the resulting biomechanical stimulation on lung cells within the lung-on-a-chip17,18. Furthermore, these models can incorporate convective flow to mimic barriers interacting with fluid, such as occurs in the lungs with air or the epithelium with blood. Organ-on-a-chip models are more extensively discussed in a 2016 review by Sakolish et al.22. Also in recent years, 3D models have gained prominence due to the underlying premise that cells behave differently in 3D and 2D environments23. An engineered 3D environment can provide an improved spatial, biomechanical, and biochemical niche compared to 2D environments23, though it should be noted that engineered 2D approaches could also provide some of these cues. In particular, interest in tissue organoids has grown over the past few years24. Tissue organoids are 3D systems that mimic the spatial arrangement of cells, which may be more realistic and useful for certain tissues19,20, though these models are not without their concerns of reproducibility and limitations of cellular complexity24,25. Regardless of the approach taken, an emphasis on 6 biomimicry drives many of the approaches we discuss and will help to refine models used within each of these fields. The focus on this chapter is to highlight some recent work done on five epithelial tissues: skin, the gastrointestinal (GI) tract, the lungs and blood-air barrier, the endothelium and blood-brain barrier, and the placenta. A key question arises from collective comparison of these epithelial tissues: do the differences in each of the tissue-specific interfaces ultimately affect how tissue models are developed, and if so, how? These tissues represent multiple interfaces: air-to-tissue, air-to-liquid, liquid-to- tissue, and liquid-to-liquid (Fig. 2.1). Though these interfaces are often considered unique, they may be more similar than many realize. With each tissue, we discuss the relevant physiology and review recent models. We searched Web of Science for studies primarily from the past five years that have utilized a tissue model to evaluate molecular transport across these epithelial tissues (skin, GI tract, lungs, blood-brain barrier, and placenta), including studies that fit this criteria or were notable within the field, such as being the first model of that type within the field. We excluded studies that did not investigate molecular transport across an epithelial model, such as studies that investigated cellular uptake of molecules or impact of a molecule on barrier development. In some instances, there may have been unique approaches utilized for a single study. We chose to include those as they present approaches that others may find useful for their own application within that tissue. Two recent reviews took a similar approach, though one focused on nanoparticles and the other on organ-on-a- chip approaches. By comparison, our focus was on presenting multiple approaches to models and evaluating these approaches across various tissues22,26. To that end, we 7 compare tissues and approaches, and then summarize with conclusions drawn from each tissue and commentary on the future of tissue models. Fig. 2.1. The four types of interfacial barriers discussed in this review. (A) The air-tissue interface shows molecular transport (purple circles) from air, across the barrier layer, into the tissue space (i.e. skin or lungs). (B) The air-liquid interface shows molecular transport (yellow diamonds) from air, across the barrier layer(s) and tissue, into liquid, which is often the vasculature of the body, represented by blue/red vessels (i.e. skin). (C) The liquid-tissue interface shows molecular transport (orange circles) from liquid, across the barrier layer, into the tissue space (i.e. the brain, from vasculature into tissue). (D) The liquid-liquid interface shows molecular transport (green diamonds) from liquid, across the barrier layer(s) and tissue, into liquid (i.e. the placenta, from one vascular source into another). 8 2.2 Skin Epithelium As the outermost layer, skin is the first and largest barrier the body has to defend against external mechanical and biochemical agents. Mammalian skin is composed of the epidermis, dermis, and hypodermis layers. In skin anatomy the epithelium refers to the epidermis, the external layer of skin, which ranges in thickness from 0.04 to 2mm, is avascular, and composed of ~95% keratinized stratified squamous epithelial cells: keratinocytes. The epidermis is further stratified into the innermost stratum basale, stratum spinosum, stratum granulosum, and the outermost stratum corneum. As they move from the basale to corneum strata, keratinocytes increasingly organize, form desmosome junctions, and secrete keratin and lipids reinforcing the mechanical and biochemical barrier27,28. The epidermis and the dermis are attached via the intermediate basement membrane, a mesh of fibronectin, laminins, collagen IV, and proteoglycans that allows water retention and anchors skin appendages such as hair follicles and sweat glands27,29. The dermis, responsible for the mechanical properties of skin, is a 1 to 4 mm thick layer that consists mostly of fibroblasts synthesizing extracellular matrix composed of collagen (~70%), elastin, and proteoglycans27–29. Last, the innermost hypodermis is generally neglected in skin models as fat storage for thermal regulation. However, this complex lipid barrier has significant contributions to the overall function of skin. It is the layer from which nerves and larger blood vessels permeate the upper layers and is a rich source of stem cells, hormones, and growth factors, which are key players in re- epithelization, wound healing, and angiogenesis28,30–32. 9 The combination of properties between these distinct layers of skin results in a highly efficient barrier. Any damage to it results in immediate compromised thermoregulation, massive fluid shifts and risk of bacterial sepsis. Understanding the transport through the skin barrier has been pursued to understand not only homeostasis and hydration, but also how medications can be optimized for topical or systemic delivery and how the body is shielded against, or responds to, pathogens and radiation. Ultraviolet (UV) radiation, for example, is an unavoidable agent that has been studied with quantum dot penetration33 or percutaneous absorption models34 and proven to change the skin barrier and immune functions. Throughout the dermal layers, the permeability function is defined by different combinations of extracellular matrix (ECM) and lipids (ceramides, cholesterol, cholesterol esters, and free fatty acids). The lipids assemble into multi-lamellar sheets that fill the spaces between cells and the ECM, the main route of permeation for most compounds29,35. Models of skin have struggled to mimic the characteristic distribution of layers, ECM, lipids, and cells that form the dermal barrier. Because of the complex layering and its multiple functions, in vitro models have generally been too simple to adequately capture all aspects of the physiologically relevant transport phenomena through skin. The earliest approaches to understand the barrier functions were based on ex vivo human and animal skin, particularly pig, mouse, rat, guinea pig, and snake models33,35–39. Pig skin is the traditional model of human skin, as it has a comparable stratum compactum and broadly the same lipid content and ratios35,40. Even with the advantage of being a natural barrier, explanted tissues are limited by donor availability, dermal differences between species, lack of control over specific cells or 10 pathways, and the overall problem of maintaining a living tissue in optimal condition to render adequate results. Studies have further elucidated the limitations of animal models, particularly on the higher permeability of their skin due to differences such as thinner stratum corneum, different sweat gland or hair follicle density, and lipid composition37,41. Furthermore, the use of animals and animal skin for testing of cosmetic ingredients has been banned since 2009 by the European Union, motivating a critical need for alternative models to mimic the dermal barrier. Tissue engineering has been used to construct models based on specific questions, separating or combining the different dermal layers, cell populations, matrix compositions, and growth factors as needed. Such constructs have reached commercial grade as “living skin equivalents” or “dermal equivalents”. Products such as Episkin®42–44, EpiDermTM 37,45,46, MatriDerm®, and Graftskin are bi-layered scaffolds constructed by cultivating fibroblasts inside a hydrogel (generally collagen type I), upon which a layer of keratinocytes is seeded37,46. The constructs are cultivated submerged in growth media until the populations are mature, then the level of media is decreased to expose the keratinocytes to air (effectively an air-liquid interface), stimulating them to proliferate, stratify, and keratinize37,46. Made from human cell lines and natural materials and grown together, commercial skin equivalents provide the biological processes and metabolism of native human skin and are generally considered accurate representatives of skin29. Nevertheless, the barrier function has been proven to be more permeable in skin equivalents, a difference commonly linked to the subtle variations in lipid content in living equivalents, particularly a higher content of di- and triglycerides37,47. Fleischli 11 et al. reported a comparative study between Episkin® and human skin, assessing the transport of topically applied substances like caffeine, benzoic acid, salicylic acid, and octyl methoxycinnamate43. Differences between the tissues included the changes in the local concentration profiles in the stratum corneum, the water profile, the lipid content, and the presence of natural moisturizing factor in living equivalents43. Models including the hypodermis fat layer are rarely proposed and are not commercial30–32,48. In particular cases, collagen, fibrin, or silk gels have been used to encapsulate adipocytes or adipose-derived stem cells, which are cultured separately and then stacked with bi-layered scaffolds in a process that can take up to 35 days30,32. The addition of the adipose layer has been used to study potential complex drug absorption models32 or the role of skin in adipose tissue metabolism31. Such approaches using complex layers are compared against standardized controls that include synthetic membranes (PDMS, Silescol®, silicone, or dialysis membranes) or naturally-derived gels (Matrigel, cellulose)34,49,50. These systems are generally used as controls for transport variables or cell response, yet they cannot fully model the cell- lipid-matrix barrier complexity of skin. These engineered living skin equivalents are used throughout literature as skin barrier models, becoming a popular alternative to explanted tissues37,42–47,51. As illustrated in Fig. 2.2, traditional systems used for modeling transport across the skin can be broadly divided into three categories: i) a liquid-to-liquid interface system, usually Franz diffusion cells (Fig. 2.2A), ii) topical application using an air-to-liquid interface system (Fig. 2.2B), or iii) systemic application in an air-to-liquid interface system (Fig. 2.2C). The Franz diffusion cells are standardized systems for permeation 12 studies and allow the use of living equivalents47 or the inclusion of split-thickness (epidermis only) or full-thickness (epidermis and dermis) human52 or animal tissue40,41,49,53,54. The air-to-liquid interface systems more accurately replicate transport through the dermal barrier, mimicking the exposure of the stratum corneum to air and environmental factors, the transport across the layers, and the blood or systemic components with the liquid/media chamber. Inclusion of living equivalents in these systems fully completes a skin model that can be grown and used in vitro. The topical and systemic exposure variations have been used to study the permeation of medications and active molecules across the dermal barrier to further optimize drug delivery systems51, or study phenomena such as cytotoxicity of compounds generally applied to skin37,45,46, dermal pathologies and wound healing42,44, or the effects and treatment of radiation33,51. The Curren group used EpiDermTM in air-to- liquid interface systems to develop the reconstructed skin micronucleus assay. Here, the induction of micronuclei in a dose-dependent fashion was considered the initial steps in developing an in vitro assay for chromosomal damage in human skin, bypassing the need for in vivo genotoxicity tests of cosmetics ingredients in animals37,45. Though multiple commercial skin equivalents exist, a key challenge that remains is the polarization of the skin epithelium to better mimic the native barrier55. 13 Fig. 2.2. Models of transport through skin. Traditional models for skin transport include (A) liquid-to-liquid interface systems, usually Franz diffusion cells where the donor chamber acts as the apical side of the membrane and the receiver side acts as the basolateral side. Skin models still widely rely on variation of the transwell inserts (B-C), where the exposure of the barrier to air allows assessing the physiological behavior of keratinocytes and the external role of skin. (B) Topical application is modeled by applying the treatment on the air-exposed (apical) side of the barrier, while (C) systemic application is modeled by delivering the treatment via the liquid- exposed (basolateral) side. 14 Common air-to-liquid interface systems generally provide the necessary setup to perform electrical resistance (ER) or trans-epithelial electrical resistance (TEER), tritiated water flux (TWF), and trans-epidermal water loss (TEWL) assays without major modifications54,56. Variations of these systems usually refer to modifications of the tissue layer, ranging from the use of synthetic membranes or laminated dialysis membranes49,50 to novel 3D printed (3DP) constructs. 3DP technologies have been applied to skin modeling to exploit the spatiotemporal placement of cell lines and materials into stratified constructs such as specialized skin equivalents or stratified epidermis57–59. More robust 3DP approaches include the Integrated Composite tissue/organ Building System (ICBS) presented by the Cho group60 or the Laser- assisted BioPrinting (LaBP) system reported by the Chichkov group61. The ICBS, in particular, allows printing a poly(caprolactone) support and then printing the layered skin construct on top, producing a lifted system that both allows the differentiation of keratinocytes exposed to air and the possibility of studying topical and systemic deliveries as an air-to-liquid interface60. Most recently, the development of organ-on-a-chip technologies has allowed the introduction of skin-on-a-chip systems: miniaturized models of dermal cell populations, structures, and functions62–65. Mori et al. presented a method for fabricating perfusable, endothelialized vascular channels within a cultured collagen skin-equivalent chip. This system was intended for drug development and validated with the percutaneous absorption of caffeine and isosorbide dinitrate. Simulation of topical exposure indicated that both molecules first reached the vascular channel and then the bottom of the skin-equivalent at rates consistent with those of human skin 15 and conventional models, successfully introducing the highly relevant parameter of vascular transport into dermal models66. Alternatively, Wurfuer et al. developed a multilayered skin-on-a-chip system with epidermal, dermal and endothelial components67. The microfluidic device was designed for co-culture of human skin cells, and each layer was separated by using porous membranes to allow interlayer communication. This novel approach allows modeling of skin inflammation and edema by disrupting the layers with tumor necrosis factor alpha (TNF-α) and subsequent therapeutic drug testing. Tissue engineering principles have successfully modeled the multilayered structure of skin, including its complex combinations of cells, lipids, and ECM proteins. The progression of these constructs into liquid-to- liquid and air-to-liquid interface systems allows one to model the broad behavior of transport phenomena through skin for topical and systemic medications or the effects of radiation. Other specific parameters of skin transport have been studied, including the viability of tight junctions and permeability of the endothelium67. Organ-on-a- chip or 3DP approaches have enabled increased complexity of the models. For example, future efforts in modeling skin transport can include elements of the immune system, alterations in skin microstructure due to scarring and open-wound healing, and interactions of skin with underlying tissues. A more in-depth review by Abaci et al. discusses the use of induced pluripotent stem cell (iPSC) technology and multi-tissue organ-on-a-chip platforms for drug and chemical permeability, both additional areas for future endeavors in skin transport modeling68. 16 2.3 The Gastrointestinal (GI) Tract The GI tract regulates entry into the human body for orally consumed substances, such as nutrients from foods and drugs from modern medicines. The bulk of the absorption of these molecules occurs through the small intestine, which is composed of three key structural parts: the duodenum, the jejunum, and the ileum. Though each of these vary in their specific function, most models of the small intestine use a single cell type to model molecular absorption though the GI tract. Enterocytes are an intestinal adsorptive cell type and compose nearly 90% of the small intestine’s apical surface69,70, suggesting they are the primary regulator of transport across the GI tract (Fig. 2.3A). However, other cell types reside within the GI barrier region including goblet cells, Paneth cells, stromal cells, macrophages, and enteroendocrine cells, among others71. Thus, it may be overly simplistic and limiting to utilize only a single cell type in modeling this barrier, particularly given that intestinal co-culture has existed for nearly two decades and recent efforts have focused on optimization of these approaches for permeability studies72–74. In addition to a complex cellular population, the small intestine also contains a mucus layer which serves to regulate molecular transport from the apical to basolateral sides of the small intestine69,71,75,76. The mucus layer varies in thickness and ratio of firmly-to- loosely adherent mucus along the length of the GI tract. This finding suggests that the mucus plays an important role in regulating substance absorption along the GI tract at different locations, such as absorption within the duodenum being unique to absorption within the ileum75. Surprisingly, the cellular complexity and mucus layer are often neglected or ignored in in vitro models for evaluating transport across the 17 intestinal barrier, albeit some approaches do include multiple cells types and/or a mucus layer72,73,77. Regardless, in mimicking this barrier, numerous systems have been used, from transwell inserts to, more recently, 3D printing19,77–80. Fig. 2.3. The GI tract barrier and models. (A) The GI tract barrier is composed of mucus and a multi-cellular layer. Enterocytes are depicted in beige while additional cell types are depicted in green, with only a small portion of the entire barrier depicted here. (Adapted from Peterson, et al., Nat Rev Immunol, 2014.) (B) Cells seeded within a transwell insert are commonly used to model the pathway between the apical and basolateral side of the GI tract. (C) Cells within an organoid, depicted as blue cells here, have been utilized to evaluate transport from the exterior of the organoid towards the interior. Transwell insert-based systems remain the most common system for assessing transport across the intestinal barrier21,81–87 (Fig. 2.3B). In particular, the Caco-2 cell line derived from human epithelial colorectal adenocarcinoma cells is generally the accepted model of the intestinal barrier in these approaches81,87,88. Devriese et al. suggested that Caco-2 cells are better representative of enterocytes, whereas T84 cells, an intestinal epithelial cell line derived from colorectal adenocarcinoma, are 18 more representative of colonocytes, a second key epithelial cell type lining both the small and large intestines21,89. This distinction is important since, as noted above, the composition of the GI tract varies throughout its length and depth. Other recent studies have shown that Caco-2 cells in co-culture with other cell types (i.e., HT29, HT29-MTX, and Raji B cells) can lead to a more biomimetic barrier with mucus production in the barrier, though this does not yet seem to be the accepted approach within the field74,90–92. Regardless, continuity in the use of the Caco-2 cell line and enterocytes, as key elements to the biological relevance of the model, allows for consistency and better comparison between multiple studies. This continuity also facilitates mechanistic studies to understand specific receptors or transporters involved, such as has been done for polybrominated diphenyl ethers or fructosazine85,87. These mechanistic studies improve the collective understanding of absorption through particular segments of the GI tract that can inform therapeutic design with the intent on controlling transport across this barrier. Moreover, the consistency in cell type within the field enables the application of common methods to assess barrier formation, such as stating a specific TEER value that must be reached to validate that a monolayer has formed21,82,83,87,88. Surprisingly, there is variability in the minimum TEER values necessary to ensure monolayer formation, with minimum values in the range of 300-700 Ω∙cm2 reported as sufficient21,83,87. Notably, co-culture of additional cell types with Caco-2 cells leads to decreased TEER values, suggesting that high TEER values are not necessarily the most representative of the in vivo scenario90,91. These studies show that while TEER can be used to assess barrier formation, it should not be the only mechanism by which 19 barrier formation is determined since results can be highly variable with no consensus threshold. As previously mentioned, one key critique in utilizing this transwell insert- based approach is that these models often lack a mucus layer, a collection of mucin proteins that adds 3D complexity to the barrier75. Therefore, the use of a three- dimensional approach, rather than a 2D approach, would be beneficial to better mimicking the intestinal barrier. Organoids and other 3D systems have become more common in recent models of the intestinal barrier (Fig. 2.3C). Zietek et al. developed intestinal organoids in order to assess nutrient transport and hormone secretion19. In this approach, a mixture of murine cells was grown in Matrigel and uptake of different compounds assessed. Here, substances were co-delivered with and without inhibitor to study the spatial distribution of substance in the extracellular vs. intracellular space. The objective of the work was to determine whether particular receptors on the cell surfaces impact the uptake and/or passage of substances, thus studying the cellular mechanism of the compounds tested19. Zhao et al. utilized organoid models to assess uptake and later release of the substance of interest (Rhodamine 123)93. Though this organoid approach enables some mechanistic insight to be studied within a 3D cellular environment, it does not directly measure transport of substances across a barrier, but instead, it simply allows for understanding of whether a substance is taken up by the cell. Costello et al. developed a 3D model to assess bacterial adhesion and invasion, yet this strategy is limited by similar challenges in determining whether the substance of interest crosses from one side of the barrier to the other side94. Alternative 3D models have been developed that would enable this determination. Approaches by Li 20 et al. and Leonard et al. use co-culture and three-dimensional gels cast within a transwell insert to achieve a multi-cellular system with apical and basolateral compartments77,95. This allows for both determining substance permeability and validating the model through traditional, quantitative measurements such as TEER values77,95. Additionally, and critically, this approach allows for multiple cell types to be used, thereby artificially creating more realistic cell-cell interactions within the barrier model. Given the cellular complexity of the intestinal barrier, the incorporation of multiple cell types into the barrier model improves the biological relevance through the consideration of cell-cell interactions71. Other approaches have moved beyond static conditions to incorporate flow into their models78,79,96. Flow leads to shear forces on the model surfaces, resulting in mechanical stimulation of the cells. Importantly, the cells within the intestinal barrier experience multiple forces applied both laterally and normal to the surface including shear, strain, and pressure. The inclusion of these forces within an epithelial model improves biological relevance and is more mimetic of the biomechanical forces the cells experience in vivo97,98. Pocock et al. reported morphological changes in a Caco- 2 monolayer due to flow within their intestine-on-a-chip model79. In particular, they observed that the application of flow accelerated the formation of a 3D-like monolayer, confirmed by F-actin staining, in as early as 5 days compared to 21 days when using a traditional transwell approach79. Moreover, the investigators reported greater than an order of magnitude difference for the apparent permeability of caffeine and atenolol under dynamic and static conditions, confirming the importance of fluid flow on molecular transport across an epithelial barrier79. Thus, future 21 endeavors should consider strategies to simulate the mechanical forces these cells experience in vivo, together with the contribution of cellular complexity and mucus, as each of these factors may influence transport of molecules across the epithelial barrier. 2.4 Lungs and the Blood-Air Barrier The respiratory tract supports efficient transfer of gases, such as oxygen and carbon dioxide, between the airstream and circulating blood. Air moves from the upper respiratory tract, through the trachea and bronchi, eventually reaching the lungs. Within the lungs, the regulated exchange of molecules between the airstream and blood occur at the alveoli. Our focus here is on this regulated barrier, specifically the alveolar-capillary barrier or the blood-air barrier99,100 (Fig. 2.4A). A key function of this barrier is to prevent the mixing of gaseous air and liquid blood by controlling the interactions between airborne substances, such as bacteria or particulate contaminants, and bloodborne substances, such as blood cells and other molecules within the bloodstream. This lung barrier consists of alveolar epithelial cells (also called pneumocytes), microvascular endothelial cells, and a very thin (<1 µm) basement membrane101,102. Alveolar cells are further subdivided into alveolar epithelial type I and type II cells, each with unique properties99. The unique properties for each cell type have not been thoroughly studied due to technical difficulties in culturing them in vitro, though it should be noted that recent efforts have yielded success in culturing primary alveolar type II cells103,104. Generally, more is known about type II cells than type I cells99,105,106. Alveolar type I cells are thought to be more important in exchange of gases and seem to have limited capacity to proliferate, 22 whereas type II cells contribute in surfactant synthesis and secretion as well as proliferation and differentiation in response to acute lung injury99,105. Both cell types express intercellular proteins, indicative of their role as an epithelial barrier99. Microvascular endothelial cells within this barrier primarily aid in fluid management by preventing blood from entering the airway, though they also regulate diffusive gas transport from blood into the airstream. To recreate this barrier, numerous models have been used with up to four different cell populations within a co-culture system101,107–109. Many groups utilize an air-liquid interface in culture, similar to that for skin, wherein cells are simultaneously exposed to both cell culture media and air, as opposed to being fully submerged by cell culture media alone110. Fig. 2.4. The blood-air barrier within the lungs. (A) Alveolar epithelial cells lie within the bronchioles, at the interface of air within the lungs and the local vasculature. The two types of alveolar cells are depicted in orange and blue. (B) A co- culture model of multiple cell types within the lung containing alveolar epithelial cells (yellow), endothelial cells (orange), and immune cells (gray and green). (Adapted from Klein et al., Part Fibre Toxicol, 2013.) 23 In the simplest of these systems, cells are used to mimic the lung epithelium in the presence or absence of endothelial cells. Lung epithelium has been modeled using multiple cell lines including A549, NCI-H441, and Calu-3111,112, all of which are derived from lung adenocarcinoma. A549 and NCI-H441 cells are more commonly used, though the limited characterization of their receptors and low TEER values are critiques against their use18,109,111,113–116. In contrast, the Calu-3 cell line is considered more biomimetic based on reported correlation to in vivo permeability, but is derived from the bronchi, not alveoli, and thus has questionable relevance for the blood-air barrier112,117,118. In addition, isolated primary cells and immortalized primary cells have also been used, an approach that is generally accepted as more biomimetic113,117. For studies using a single cell type, lung epithelial cells are generally seeded within a transwell insert and grown to confluence, a process taking 7 to 13 days and sometimes requiring the addition of dexamethasone (0.2-1 µM) to culture medium to enhance barrier tightness, as determined by TEER, and improve expression of intercellular junction proteins, such as ZO-1 and claudin-3111,115,116. For co-culture studies with an endothelium, the transwell insert is flipped and an endothelial cell population is seeded on the bottom of the insert. Numerous endothelial cell types have been used, including both primary cells and cell lines, but a common characteristic is that these cells are often microvascular cells17,109,114,116. Since pulmonary microvascular cells are commercially available, they may be the most relevant cell type for lung models other than primary isolated cells, as they are vascular cells from the same tissue17,114. Other approaches to building a co-culture model of the lung epithelium containing two cell types include organ-on-a-chip and 24 bioprinted lung models17,18,109. For example, Huh et al. induced cyclical mechanical stretching to mimic breathing in a novel lung-on-a-chip17,18. The FlexCell® apparatus is among the only commercially available alternatives for inducing cyclic strain. While this system has been used to apply cyclic strain to lung epithelial cells119,120, it has not been utilized in assessing transport across strain-induced epithelial barriers to our knowledge. We speculate that the inclusion of mechanical stimulation may be feasible in bioprinted models. Horváth et al. reported that barrier integrity was improved in cell-printed models compared to manually seeded cells, suggesting that bioprinting may be an advantageous route for building lung epithelium models in the future109. The use of three or more cell types adds complexity to the co-culture system as previously reported101,107,108,121 (Fig. 2.4B). In these models, isolated immune cells such as macrophages, dendritic cells, and mast cells are included101,107,108,121. Macrophages tend to be the most commonly incorporated immune cell because macrophages in the alveolar space are well-regulated and responsible for clearing any pathogens that have not been cleared higher up in the respiratory tract101,102. Interestingly, Lehmann et al. reported that co-culture of lung epithelial cells (A549 cell line, 16HBE14o- cell line, or isolated human alveolar type II epithelial cells) with both macrophages and dendritic cells reduces TEER values, and the lung epithelial cell morphology changes depending on which immune cell type and epithelial cell type are co-cultured together108. This suggests that isolated alveolar type II cells are the closest to what occurs in vivo, which might be anticipated. Kasper et al. cultured either A549 or NCI-H441 lung epithelial cells with microvascular endothelial cells 25 (ISO-HAS-1) to study the response of different macrophages (i.e., M1 or M2 macrophages) 101. Macrophage phenotype did not affect the construction of the model, which allowed its use for studying the inflammatory response at the blood-air barrier and interrogating the role of M1 vs. M2 macrophages101. Moreover, Klein et al. established a model using lung epithelial cells (A549), endothelial cells (EA.hy926 cell line), macrophages (differentiated THP-1 cell line), and mast cells (HMC-1)107. The participation of two immune cell populations suggests that the immune response, though perhaps quiescent more often than fully active, is important within models of the alveolar-capillary barrier107. We speculate that the inclusion of immune cells into blood-air barrier models composed of endothelial and lung epithelial cells, together with advances in applying mechanical stimulation and bioprinting109,122, will enable improved models of this critical interface and allow for more complex studies of this barrier and lung diseases in the future. 2.5 Endothelium and the Blood-Brain Barrier The body’s vasculature provides a conduit for nutrient and oxygen transport throughout the body. Given the size and extent of this vasculature, it should come as no surprise that the system is highly regulated123. Of particular interest is the vasculature within the brain that forms the blood-brain barrier (BBB), a topic that has been extensively reviewed in and of itself124,125. Thus, our focus here is on the recent work related to the brain, including the use of co-culture systems, stem cells, organ- on-a-chip approaches and spheroid models to mimic the BBB126–129. Anatomically, the brain endothelial cells (BECs) are the primary regulator of transport between blood and the interstitial space and neural cells within the brain130 (Fig. 2.5A). These 26 BECs regulate passage of the vast majority of small and large molecules, through passive and active transport mechanisms, with few small molecules capable of diffusing across the barrier130,131. These barrier properties provide superior, necessary protection for the brain from injury, yet prevents the successful delivery of many potential therapies by blocking entry into the brain space130,131. These properties have motivated the development of improved BBB models to better understand how to develop therapeutics that can effectively cross the BBB and reach the interstitial brain space130,131. Fig. 2.5. The Blood-Brain Barrier. (A) Schematic depicting molecules (yellow) crossing the vasculature to enter into the brain space with numerous brain cells (depicted in green, yellow, and blue). Molecules within the lumen of the vasculature must traverse the brain endothelial cells and basement membrane before reaching the brain space. (B) Illustration of a microfluidic platform to mimic the BBB, where micropillars are used within the cell-laden area to better mimic the cellular niche. (Adapted from Prabhakarpandian, et al., Lab Chip, 2013.) 27 As in other tissues, culturing cells on a semi-permeable insert represents one of the most common approaches to model the BBB 127,128,140–148,132–139. Common cell types used in these studies include both primary cells, such as porcine BECs, and cell lines, including hCMEC/D3, BB19, transfected human brain microvascular endothelial cells (hBMECs), TY10, and b.End3 cells133–136,149. While many of the cell lines are simply stable brain microvascular endothelial cells, Stins et al. reported that hBMECs, tranfected with oncogenic sequences, performed similarly to primary cells150. A comparative study by Eigenmann et al. suggested that transfected hBMECs were the most promising for establishing a BBB model, as determined by barrier tightness (TEER values up to 10 Ω∙cm2 higher than other groups at 80 hours) and paracellular permeability136. These findings suggest that the choice of cell line is key for the development of BBB models, particularly when evaluating mechanistic details of molecular transport across the barrier. Alternatively, BBB models have used brain endothelial cells derived from induced pluripotent stem cells (iPSCs) and hematopoietic stem cells127,132,138,147. The use of iPSC-derived BMECs in co-culture with neural progenitor cells resulted in TEER values in the range of 5,000 Ω∙cm2 after optimization132, which are extremely high compared to values reported in vivo (approximately 1,500 Ω∙cm2)127,132. As TEER is a key quantitative metric for assessing barrier formation in BBB models, this suggests that stem cell-derived neural endothelial cells, such as iPSC-derived BMECs, may provide valuable flexibility to tune intercellular adhesion for use in models of the BBB, either to assess effects on the barrier or evaluate transport across the barrier138,144. 28 Fluidics and microfluidics represent another common approach to model the BBB129,151–154. For example, Prabhakarpandian et al. developed a microfluidic device with pillars and posts that mimic the sub-micron and micron-scale membranes and pores within a transwell153 (Fig. 2.5B). This design enables testing of converging and diverging bifurcations, mimicking the blood vessels within the body, which may be more relevant given the branching of vasculature within the body153. Booth and Kim reported low TEER values in static transwell inserts, yet these values were significantly increased by an order of magnitude in a dynamic microfluidic BBB model151. Wang et al. demonstrated that the co-culture of BMECs with astrocytes, as opposed to monoculture of either alone, also increased TEER values nearly an order of magnitude within their microfluidic device129. Similarly, Partyka et al. reported that fluid flow in a mechanically compliant microfluidic device enhances TEER values compared to static samples152. Moreover, the results of this study confirmed that the type of flow (i.e., steady vs. pulsatile flow) results in differences in pressure at the inlet port of the fluidic device, the velocity profile within the inner lumen, and radial strain at two positions within the wall152. Additionally, fluid flow in the model, with or without stretch, influenced the permeability compared to static conditions152. Intuitively, it is also likely that fluid flow, and the resultant shear stress, may impact endothelial cell behavior and the observed transport characteristics. This more generally suggests that taking into account the various stimuli present within a system (i.e., mechanical, chemical, biological) is important to ensure the model best reflects the in vivo scenario. These approaches together show that microfluidics may be the 29 future of BBB models given their versatility in design, ability to support co-culture, and their mechanical compliance. The evaluation of molecular transport into the brain has also been modeled with the use of spheroids, which are dense cellular aggregates that better mimic the 3D environment compared to cells in monolayer culture126. Cho et al. formed spheroids containing astrocytes, pericytes, and either primary HBMECs or the hCMEC/D3 cell line 126. Interestingly, permeability into their model was modulated by vascular endothelial growth factor (VEGF), a potent mitogen and growth factor involved in angiogenesis126. Moreover, inhibition of P-glycoprotein, a drug efflux pump responsible for pumping molecules out of a cell, influenced the extent of transport in these spheroids, providing further evidence that the spheroid model mimics biological responses in a bulk-tissue manner126. From this model and other approaches, it is clear that improving the complexity of BBB models over current transwell approaches, whether by co-culture, architecture, or addition of fluid flow, provides an exciting opportunity to increase our understanding of BBB function. 2.6 Placenta and the Blood-Placenta Barrier The placental epithelium refers to the maternal-fetal barrier, also known as the blood-placenta barrier (BPB), that separates maternal blood and fetal blood. The key transport regulator is the syncytiotrophoblast, a multi-nucleated giant cell that is directly in contact with the maternal blood that also provides endocrine function for the placenta155,156 (Fig. 2.6A). Underlying the syncytiotrophoblast is villous cytotrophoblasts, progenitor cells that help maintain placental structure and syncytiotrophoblast function and viability by fusing into the 30 syncytiotrophoblast155,156. Additionally, the placenta contains a basement membrane and placental-specific macrophages termed Hofbauer cells that are in contact with fetal endothelial cells157. The precise function of the Hofbauer cells within the placenta is unknown, but these cells are not thought to contribute to the placental transport barrier157. Fetal endothelial cells are the most external cell layer within this barrier, directly contacting fetal blood and thus the fetal circulation. Within the placental transport field, most in vitro models focus on mimicking the syncytiotrophoblast, often neglecting the fetal endothelium. However, fetal endothelial cells have been incorporated into more recent models since the endothelium is highly regulated and adds a layer of transport resistance both in the placenta and throughout the rest of the body158–161. The use of ex vivo perfused placenta models, wherein the term-delivered placenta is kept viable and perfused with fluid, has convincingly demonstrated that the cellular complexity of the tissue is necessary to recapitulate the path of a molecule in traversing from maternal blood to fetal blood162–164. The placenta places maternal blood in direct contact with an epithelial cell (the syncytiotrophoblast), whereas blood otherwise contacts endothelial cells throughout the rest of the body. This suggests that approaches in the placenta field derive from other epithelial tissue models, not necessarily from models of the endothelium. 31 Fig. 2.6. The Placental Barrier. (A) Physiology of the placental barrier, with the syncytiotrophoblast interacting with maternal blood and fetal endothelial cells interacting with fetal blood. Cells depicted within the basement membrane layer are Hofbauer cells interspersed in that layer. (B) A common microfluidic approach to building a cell-model, with media inputs and outputs shown. Cells are seeded in the middle portion of the device, where both apical and basolateral media interact with cells. (C) An Ussing chamber schematic, where a tissue section is held between two compartments, creating an apical and basolateral chamber for measurement of ion transport. A current source and voltmeter is used to regularly measure differences between the two chambers. Traditional models utilizing transwell inserts have still largely been used for investigating placental transport165–173. Historically, these models use trophoblast cell lines for mimicking the syncytiotrophoblast, most commonly the BeWo cell line and the b30 clone modified from BeWo165,174–177. The main criticisms of this approach is that cell lines may not be representative of the in vivo condition, diminishing the relevance of the findings, and that BeWo cells lack some transporters that may alter cellular transport16,178,179. A recent proof-of-concept study by Huang et al. utilized primary cells instead of immortalized cell lines to demonstrate that primary trophoblast cells, although difficult to grow in vitro, can be used for placental 32 transport models166. Interestingly, the primary cells were seeded at an extremely high density in this study (1,000,000 cells/cm2), compared to the common seeding densities used for BeWo b30 cells (100,000 cells/cm2)164–166,174. However, one key limitation to this approach is that primary trophoblast cells derived from term placenta behave differently than trophoblasts at earlier stages of development (i.e. less invasive and migratory180), indicating limitations of the relevance of the data obtained over the entire course of pregnancy. Additionally, challenges in maintaining cell viability for more than a few days must be resolved to enable common use of primary cells. Following the excitement of other organ-on-a-chip reports, the placenta-on-a- chip has utilized both cell lines and primary cells to create trophoblast and endothelial cell co-culture sytems158,181 (Fig. 2.6B). As in other “on-a-chip” technologies, this approach uses small amounts of cells and resources and allows for perfused flow to mimic the blood flow in the placenta. In one study, the authors reported that the glucose transfer rate across this membrane was high when acellular (67.8-99.5%), and glucose transport was dramatically reduced in a JEG-3/HUVEC co-culture system (17.3-39.1%)181. When using fused BeWo b30 cells and primary human placental villous endothelial cells in a placenta-on-a-chip model, the glucose transfer rate (34.8%) was similar to that of ex vivo perfused placenta (26.5-38.3%), the gold standard in the placental transport field. When compared to a standard transwell co- culture system, glucose transport was reduced (22.5%), suggesting that the “on-a- chip” technology was more physiologically relevant than transwell models158. 33 Together, these studies showcase the versatility of this approach and suggests this strategy has advantages compared to traditional in vitro approaches. In contrast to these technologies, a tissue engineering approach has been used to build a placental barrier model159,160. Here, an amniotic membrane was used with trophoblast (JEG-3) and endothelial (HUVEC) cells seeded on both sides of the membrane, housed within a custom-built perfusion chamber. Interestingly, this approach yielded a cell-laden membrane that was similar in thickness to the placental barrier (~20 µm). Membrane thickness is a key parameter to effectively mimic the diffusive properties of a membrane, as thickness is inversely proportional to the rate of diffusion. Conversely, transwells and “on-a-chip” technologies, each of which utilize cells on a thin membrane, place the emphasis on cell-cell contact for forming the barrier. Although these technologies place less emphasis on ensuring the overall model thickness, the thickness of the barrier in these modeling approaches, containing one or two cell layers on a thin semi-permeable membrane, can be assumed to be comparable to the in vivo condition. Furthermore, this tissue engineering model incorporated perfusion160, as in the placenta-on-a-chip and ex vivo perfused placenta, providing an opportunity to better model the native tissue and induce physiologically relevant shear stress (0.001 – 0.1 dyn/cm2)182 that may lead to changes in transport. Due to the opportunity to control the cellular composition, architecture, and fluid forces, the tissue engineering approach provides a platform that better approximates the in vivo placental barrier, compared to other approaches. Other approaches of note include the use of an Ussing chamber and computational models to evaluate transport183–186. An Ussing chamber is a 34 commercial apparatus that requires a membrane, which is placed in the middle of the chamber creating two compartments183 (Fig. 2.6C). Song et al. used placental slices as their membrane and demonstrated that they could measure time-dependent diffusion with sufficient placental activity for short times (<5 hours)183. However, the manner in which the membrane is sliced is key to obtaining relevant transport information. Computational models have also been recently used for evaluating both drug and nutrient transport within the placenta184–186. These approaches require experimental data to develop predictive models that can complement, inform, and guide wet lab experiments. The placental transport field is advancing with new technologies and improving our understanding of substance transfer across the blood- placenta barrier. 2.7 Comparison of Tissues and Barriers The key to understanding molecular transport across any epithelial barrier is to know the starting point, the end point, and the pathway between. This pathway ultimately dictates what must be included in a model that is intended to recapitulate molecular transport across a barrier. For example, air is a common starting point when studying molecular transport across skin, the pathway is through dermal layers, and the end point can be either the dermal tissue or the vasculature. Thus, either only a few or all dermal layers will need to be included in the model to adequately recapitulate the physiologic barrier presented to molecules as they traverse this pathway. Broadly, this conceptual approach is applicable to all the tissues we have discussed herein. A summary of tissue characteristics and model information is presented in Table 2.1. 35 36 Tissue interfaces that regulate transport can be defined in multiple ways: (1) air-liquid, (2) air-tissue, (3) liquid-liquid, and (4) liquid-tissue. This yields the most similarities between approaches in building tissue models. Air-liquid barriers, such as the lung, GI tract, and skin, are increasingly modeled using air-liquid interfaces in cell culture. The use of this culture technique provides a more realistic pathway of a molecule from air, where transport occurs through a gaseous medium, to interactions with cells and ECM before reaching liquid, where transport is distinct from that in a gaseous phase. Liquid-liquid barriers, such as the placenta, are often capable of being recapitulated using a flow system that induces fluid shear stress, such as a bioreactor or organ-on-a-chip approach. The use of flow provides increased stimulation similar to the in vivo condition, particularly when using similar rates of shear and other biomechanical forces. Air-tissue and liquid-tissues, such as the brain and skin, can often be modeled using organoid approaches, a modeling approach that may not be as suitable for other types of interfaces. This is due to the organoid containing a core of tissue-specific cells in these models, as would be found in vivo when a substance is transported into a tissue. Ultimately, the normal physiology of the tissue dictates the approach, but these characteristics of a transport pathway, including cell type and phenotype, presence of supporting cells and/or key ECM proteins, biochemical stimuli, and biomechanical stimuli, are important factors that should be considered when developing a biologically relevant model. While there are biological differences between each of the tissues discussed throughout, approaches for constructing these models and assessing how well they reflect the in vivo scenario are often similar between tissues. Molecules often have 37 two approaches for traversing a barrier: a transcellular approach, by passing through the cells of the barrier, or a paracellular approach, by passing between the cells of the barrier. Cells generally self-regulate the transcellular pathway, although the mechanism for how cells regulate this pathway is largely dependent upon the microenvironment. For example, differences in mechanical forces such as shear or hydrogel matrix stiffness can impact cell morphology, and in turn, the spatial arrangement within the microenvironment of membrane proteins187,188. Additionally, as discussed throughout this review, cells behave differently in 2D and 3D environments, as well as when there are multiple cell types present (i.e., co-culture of two or more cell types in a single model). In certain cases, such as the GI tract, co- culture of cells leads to changes in protein (i.e., mucin) secretions, which modifies the microenvironment experienced by the epithelial cell and may impact available transport pathways for a molecule. Therefore, while it may often be assumed that an appropriate cell choice will be sufficient in recapitulating this pathway, care must be taken in realizing how the cellular environment will impact the cell. This is particularly true for transporter proteins that shuttle a molecule across a cell, such as the GLUT1 transport protein transporting glucose158. The paracellular pathway is equally important and is also heavily influenced by the microenvironment. The paracellular pathway is often composed of intercellular junction proteins such as cadherins, claudins, and occludens95,143, where expression of these proteins localized at points of cell-cell contact may or may not be sufficient for recapitulating the desired tissue barrier. Similar to the transcellular pathway, mechanical forces can impact intercellular junction protein expression187,188. More quantitatively, TEER is 38 often used as a metric to assess barrier formation, but this method may provide limited characterization given the wide range of values for some tissues, the variability due to the presence or absence of multiple cell types, and the lack of comparative values from human tissues in vivo. A more thorough review by Srinivasan, et al, highlights a number of limitations and variabilities that should be considered when evaluating TEER values189. Ultimately, the in vitro barrier model should be compared with the in vivo epithelial barrier to determine whether or not the barrier is sufficient to conduct the intended experiments. This includes confirming the phenotype of cell(s) within the barrier, confirming that relevant necessary ECM components are produced (e.g., mucin protein), and, as best as possible, mimicking biomechanical and biochemical stimuli. Beyond the physical model, the mathematics and equations used to describe the transport phenomena, such as calculating the (apparent) permeability or diffusion coefficient, provides a relatively simple metric for comparison across tissues and between models. Experimentally, the diffusion coefficient and (apparent) permeability can be determined be measuring concentration changes, on both sides of the barrier, over time. A common mathematical approach is to use Fick’s first law relating flux with the diffusion coefficient and concentration gradient190,191: j = −D∇C Where j is the flux [mass per area per time, i.e. g/(cm2∙s)], D is the diffusion coefficient [area per time, i.e. cm2/s], and ∇C is the concentration gradient [mass per volume per length, i.e. (g/cm3)/cm = g/cm4] of the species across the barrier. Though other approaches have been utilized, such as rate of transfer158, the diffusion 39 coefficient and/or (apparent) permeability is more widely used and suitable for comparison of engineered systems. For additional guidance on the mathematics of diffusion and calculation of these constants, we refer the reader to widely available texts192,193. Further, a derivation of how to calculate the diffusion coefficient experimentally is included in appendix A of this thesis. Broadly, the application of these models toward therapeutic development is a commonly pursued end goal in these areas of study17,78,132. The promise is that these models could provide a method for assessing toxicity of any substance to these tissues and for determining the extent to which these substances may cross these barriers. For example, the former may be important in studies for drug formulations and cosmetic products that may be sufficiently evaluated in a reductionist model consisting of tissue-specific cells. The latter may be valuable in developing therapeutics to target the brain or novel detection mechanisms for probing the fetus and may require more complex, multi-component systems that sufficiently mimic molecular transport into these tissues. It is critically important to improve our baseline understanding of how a substance interacts with and traverses across a particular tissue to advance the development of improved molecular and diagnostic approaches. Moreover, the use of in vitro models are markedly more reproducible and cost effective compared to animal studies, while obviating persistent ethical concerns. Arguably, the present could be one of the best times to enter the field, given that in 2017 the U.S. Food and Drug Administration announced a multi-year agreement with a company to utilize organ-on-a-chip technologies for assessing the effectiveness of this technology. This announcement has large implications on the potential commercial and clinical 40 implications of this field, while emphasizing the promise of these technologies to unearth a deeper understanding of molecular transport across barriers throughout the human body. 2.8 Conclusions and Summary Tissue models of epithelial barriers continue to evolve as cell-based technologies mature. Though often considered distinctive, there is some truth to saying all epithelial tissues are similar. The cells often have similar functions and structures, with additional similarities in the classes of proteins expressed by the cells in these barriers. As emphasized herein, strategies to develop and analyze barrier models exhibit similarities between tissues, and the emergence of organ-on-a-chip, organoid, and tissue engineered models reinforce the need for continued study and development of these approaches. Moreover, the use of primary cells and/or multi- cellular co-culture systems pervading the tissues confirms the widespread interest in building better models to enhance our understanding of these tissues. Though beyond the scope of this review, computational models of epithelial barrier transport have been reported and are likely to expand in the future, aided by and informing experimental data from in vitro models. 41 Chapter 3: The Human Placenta and Modeling Approaches 3.1 In Vivo Human and Animal Placental Barriers The human placenta is a unique barrier that is unlike many other animals16,194. The human placental barrier is considered hemochorial, where the maternal blood is directly contact the placental epithelium (i.e. the syncytiotrophoblast). There are also endotheliochorial and epitheliochorial placental barriers within other animals. Endotheliochorial barriers occur when the maternal blood is surrounded by a maternal epithelium, meaning that there is a maternal endothelium, a placental epithelium, and a fetal endothelium separating maternal and fetal blood, as is the case for dogs and cats16,194. Epitheliochorial barriers build upon this by adding a secondary epithelium to the placental barrier, meaning that maternal and fetal blood are separated by a maternal endothelium, two placental epithelium layers, and a fetal endothelium, as is the case for cows, pigs, and horses16,194. These three barriers are nicely drawn out within the commentary piece by PrabhuDas, et al194. Given the differences between these animals, it becomes clear that the human placenta is drastically different from those in most other animals. Therefore, it would inappropriate to utilize animal models to study the human placenta, where the placental structure does not mimic the human placenta. The most commonly used animal model for placental studies is the rodent, since both are considered hemochorial16,194,195. However, there are notable differences between humans and rodent placental barriers. In particular, the mouse placenta is widely used, in spite of the numerous differences between them. Some hormones that 42 are synthesized in the human placenta, such as estrogen, is absent in mice195. Structurally, the human placenta maintains a high amount of intervillous space, where the fetal villous trees are essentially bathing in maternal blood, while the mouse placenta is described as a “placental labyrinth”, where there are a number of interconnected cavities rather than the open space in human placentas195–197. Further, within the syncytiotrophoblast layers, the human placenta has a single syncytiotrophoblast layer whereas mouse placentas have two syncytiotrophoblast layers197, indicating that there would likely be significant differences in utilizing mouse placentas for human placental molecular transport studies. Given the widespread differences between human and mouse placentas, it is critical to ensure that molecular transport studies, such as those described later in this thesis, be conducted on a model that mimics the multilayer structure of the human placenta and not studied utilizing mouse placentas, which are anatomically different than the human placental barrier. 3.2 In Vitro Placental Barrier Models Given that in vivo approaches (i.e. animal models) to model the human placental barrier for molecular transport studies are largely not relevant to the human placenta, in vitro approaches are largely the most biomimetic approach. As discussed in the previous chapter, a number of approach are emerging as novel techniques to study transplacental transport, including placenta-on-a-chip approaches and custom built perfusable bioreactor systems. However, the vast majority of placental transport studies utilize one of two approach: the transwell model and the ex vivo perfused placenta. In the transwell approaches, cells are seeded within the transwell in order to 43 create a monolayer of trophoblast cells164,174, with distinct apical and basolateral compartments. The limitation with this approach is that most often, only trophoblast cells are utilized, ultimately neglecting the role that the fetal endothelium plays in regulating transport across this barrier161,164,165. This is detrimental to the overall benefit of each transport study as numerous studies, including those presented within later chapters in this thesis, have shown that the endothelium contributes to the overall transport resistance and diffusion of molecules across the placental barrier161,198. By comparison, the ex vivo perfused placenta approach, where the vasculature within a placenta delivered at term is connected to perfusion equipment, maintains all the layers within the placental barrier16,199. However, some limitations on this approach include damage to the tissue resulting from delivery, thus raising concerns about the validity of the data obtained, and the temporal state of the tissue, where studies carried out on a tissue at full term (i.e. the end of the third trimester) does not adequately represent the placental barrier earlier in the pregnancy16. Despite these limitations, the ex vivo perfused placenta remains the gold standard approach for transplacental transport studies, in part because fewer advanced systems, such as in vitro perfusion models and placenta-on-a-chip approaches158–160,200, have only come about within the past decade, if not the past few years. Ultimately, in vitro approaches that are biomimetic to the human placenta are largely needed to enable high throughput and rigorous studies for the transplacental transfer of drugs and substances. In particular, the field needs 3D models that recapitulate the multilayer architecture of the placental barrier better than current 2D approaches, and models that allow for multicellularity, supporting coculture and cellular barrier formation. 44 Chapter 4: A Placenta-Fetus Model for Evaluating Maternal- Fetal Transmission and Fetal Neural Toxicity of Zika Virus2 4.1 Introduction Zika virus (ZIKV) has emerged as a global health epidemic with significant concern for pregnant women due to a link between viral infection and birth defects in newborn infants3. Much of the current literature on ZIKV focuses on clinical outcomes and pathogenesis, yet few studies, to our knowledge, have evaluated maternal-fetal transmission of ZIKV across the maternal-fetal interface201–203. This is concerning given that ZIKV has been linked to tissue damage in the placenta and fetal brain201. There is a critical need for understanding ZIKV transport across the maternal-fetal interface, termed the blood-placenta barrier (BPB), and the resulting fetal outcomes158,159,166. Moreover, no models exist that adequately recapitulate placental biology to allow investigation of ZIKV transport across the BPB and determination of how ZIKV influences fetal outcomes. Current models of the BPB fail to recapitulate the complex physiology of this tissue. The BPB is composed of syncytiotrophoblast, villous cytotrophoblast, Hofbauer cells, fetal endothelial cells, and basement membrane156,157. The syncytiotrophoblast is considered the predominant regulator of placental transport, though fetal endothelial cells add another highly regulated layer156,204. Villous 2 Adapted from: N. Arumugasaamy, LE Ettehadieh, CY Kuo, D Paquin-Proulx, SM Kitchen, M Santoro, JK Placone, PP Silveira, RS Aguiar, DF Nixon, JP Fisher, and PCW Kim. Biomimetic Placenta-Fetus Model Demonstrating Maternal-Fetal Transmission and Fetal Neural Toxicity of Zika Virus. Annals of Biomedical Engineering. (2018). 45 cytotrophoblast also aid in keeping the syncytiotrophoblast alive, whereas the Hofbauer cells do not seem to add any additional transport barrier to the placenta156,157. In vitro models of the BPB are often trophoblast cells seeded within a transwell insert, ignoring the highly regulated endothelial cell population16. In contrast, most animals have different placental biology than that of humans16. A more biomimetic approach is tissue engineering, where natural proteins are used to create a 3D environment that recapitulates tissue-level physiology159. Tissue-engineered models can be transformative in areas such as reproductive health and pregnancy, where it is extremely difficult, if not impossible, to conduct rigorous studies of healthy and diseased tissues due to increasing safety and ethical concerns205,206. Moreover, for pregnancy-related studies, researchers must consider both maternal and fetal health to ensure minimal, if any, harm to both the pregnant mother and her fetus205,206. Thus, biomimetic models developed using human cells that recapitulate the physiology of pregnancy, including maternal reproductive tissue, the placenta, and fetal tissue, provide a clear opportunity for better understanding reproductive biology, and developing and testing potential therapeutics. To address these shortcomings, we have developed a biomimetic placenta- fetus model that incorporates a tissue-engineered BPB model and a compartment for fetal cells to determine effects on the BPB and fetal neural cells following exposure to ZIKV. We developed this BPB model to mimic the placenta’s multilayer organization and evaluated barrier formation, cellular functionality, and molecular transport across the model. We then evaluated ZIKV infection of placental cells, viral transport across the model, viral inhibition by chloroquine, and neural cell viability resulting from 46 viral exposure. Our results from these ZIKV studies suggest that the placenta, though a preferential location for viral replication203, may modulate viral exposure and could account for the high variability in observed fetal damage. This biomimetic modeling approach supports the findings from in vivo approaches203 and enables assessment of fetal cell changes that, to our knowledge, have not previously been possible. 4.2 Materials and Methods Cell culture BeWo b30 cells were obtained with a signed material transfer agreement from Erik Rytting (University of Texas Medical Branch, Galveston, TX), who received this cell line from Lisbeth E. Knudsen (University of Copenhagen, Denmark), who had obtained the cells from Margaret Saunders (Bristol Haematology and Oncology Centre, Bristol, UK) with permission granted to Dr. Rytting by Alan Schwartz (Washington University, St. Louis, MO). Clone b30, intended for improved monolayer formation, was derived by limiting dilution177 from the parent BeWo cell line established by Roland Patillo and George Gey176, and was grown in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with fetal bovine serum (FBS), antibiotic-antimycotic (A/A), non-essential amino acids, and L-glutamine. BeWo cells were purchased from American Type Culture Collection (ATCC) and grown in DMEM with FBS and Penicillin-Streptomycin (P/S). Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and grown in EGM BulletKit (Lonza, Basel, Switzerland). Further, EGM BulletKit was used throughout BeWo b30/HUVEC co-culture studies. Human induced pluripotent stem cells (iPSC)- derived neural progenitor cells (NPCs) were generously provided by Dr. Kazue 47 Hashimoto-Torii (Children’s National Health System) and were grown in DMEM/F12 with N-2 Supplement, B-27 Supplement minus vitamin A, A/A, Laminin, and FGF2. Additionally, NPCs were grown on polyornithine and laminin- coated plates, as described in literature207. Throughout the text, BeWo and BeWo b30 cells are referred to as trophoblast cells, HUVECs as endothelial cells, and iPSC- derived NPCs as fetal neural cells. BPB model fabrication Our BPB model was fabricated via casting gelatin methacrylate (GelMA) and crosslinking to form a hydrogel. Gelatin type A (Sigma-Aldrich) was functionalized by addition of a methacrylate group, as previously reported208. Briefly, Type A gelatin from porcine skin was mixed (10% w/v) into phosphate buffered saline (PBS) at 50°C for 20 minutes. Methacrylic anhydride (Sigma-Aldrich) was added to this solution, under vigorous stirring, at a ratio of 0.6g methacrylic anhydride per gram gelatin, and the reaction allowed to proceed for one hour. The solution was then centrifuged at 1000g for 2 minutes to remove any unreacted components. The supernatant was diluted 1:1 with PBS and dialyzed in cassettes (10 kDa MWCO, ThermoFisher) for 3 days at 50°C. The pH of the dialyzed product was adjusted to between 7.3 and 7.4, and then lyophilized for 1 week prior to use or storage. Lyophilized GelMA was kept at -80°C for long term storage until use. Hydrogel precursor solution was made by dissolving lyophilized GelMA in cell culture media. 2-hydroxy-1-(4-(2-hydroxyethoxy)phenyl)-2-methyl-1-propanone (Irgacure 2959, BASF, Ludwigshafen, Germany) was added as a photoinitiator at 48 50°C for 15 minutes. The precursor solution was then allowed to cool at room temperature for 30 minutes, cast in either transwell inserts or in a well plate, and then allowed to cool for an additional 30 minutes. For cell-laden hydrogels, cells were encapsulated at a density of 106 cells/mL within the precursor solution prior to casting. Following the second cooling phase, gels were exposed to UV light for 2 minutes via UV box to form a crosslinked GelMA hydrogel and kept in cell culture media throughout experiments. BeWo b30 cells seeded on top, at a density of 105 cells/cm2, were added after crosslinking. HUVECs were seeded on the other side of the gel, at the appropriate time and a density of 105 cells/cm2, by inverting the hydrogel using a metal spatula. To ensure cells were present on both sides of the hydrogel, samples were fixed on day 10 in 4% formaldehyde for 4 hours at room temperature. Samples were washed with PBS and stained with DAPI for 5 minutes. Samples were again washed prior to imaging via confocal microscope (Olympus FV1000), allowing for reconstruction of a cross-section view (Z-stack) of the sample. Hydrogel thickness was via image analysis. Transepithelial electrical resistance (TEER) testing Transepithelial electrical resistance (TEER) testing was performed to determine barrier resistance over time. Hydrogels were cast in transwell inserts, BeWo b30 cells seeded on day 0, and HUVECs seeded on day 7, post-casting (Fig. 4.1). TEER values were recorded daily, beginning on day 1. Values were normalized against values obtained for PBS and multiplied by the hydrogel’s surface area, 49 assumed to be equal the transwell insert growth area. For the blank gel, the value is reported as an average of all time points. To visualize TEER data, samples from every other day, beginning on day 3 post seeding, were fixed, as described above, and stained to visualize cellular nuclei and actin. First, samples were washed with PBS and permeabilized with 0.25% Triton-X for 5 minutes. Samples were then washed three times with PBS on a shaker plate, blocked with 5% goat serum, then incubated with Texas Red-X Phalloidin for 30 minutes. Samples were then washed three times with PBS and incubated with 4’,6- Diamidino-2-Phenylindole, Dihydrochloride (DAPI) for 5 minutes. Samples were washed three times with PBS and transferred to glass bottom dishes for imaging via confocal microscope. Viability Stain Viability of cells within our model was determined qualitatively using a Live/Dead Viability/Cytotoxicity Kit for mammalian cells according to the manufacturer’s protocol. Cells were grown for a total of 10 days then stained using 2µM calcein-AM and 4µM ethidium homodimer-1 in PBS for 1 hour. Samples were washed prior to imaging on a confocal microscope (Olympus FV1000) or epifluorescent microscope (Nikon Ts2R). Intercellular junction immunofluorescent staining Immunostaining experiments were performed to qualitatively determine cell monolayer formation. Epithelial (E-) and vascular endothelial (VE-) cadherin are 50 intercellular adherens junction markers expressed by BeWo b30 cells and HUVECs, respectively158. Zonula occludens-1 (ZO-1) is a tight junction protein expressed by both BeWo b30 cells and HUVECs123,159. BPB models were fixed as described above, washed with PBS, and permeabilized with 0.25% Triton-X for 5 minutes. Samples were washed with PBS, blocked with 5% goat serum, then incubated with primary antibody against E-Cadherin, VE-Cadherin, or ZO-1 overnight at 4°C. The following day, samples were washed with PBS for at least 1 hour, followed by incubation with an appropriate secondary antibody for 1 hour, then washed with PBS. Samples were counterstained with DAPI and washed prior to fluorescent imaging via confocal microscope. HERV Stain Expression of HERV by BeWo b30 cells was determined via fluorescent microscopy. Samples were immunostained using a similar protocol to that used for intercellular junction immunofluorescent staining. The primary antibody used was Anti-HERV (Abcam, ab71115) while the secondary antibody used was Goat Anti- Rabbit IgG H&L Alexa Fluor 488 (ab150077, Abcam). Enzyme-linked immunosorbent assays (ELISA) ELISA were performed to determine secretion levels of human chorionic gonadotropin (hCG), vascular endothelial growth factor (VEGF), and progesterone. Cell culture supernatant was collected after 2 days in culture for each group, then frozen and stored at -80°C. Samples were thawed to room temperature prior to 51 running the assay. Kits against human hCG, human VEGF, and progesterone were used according to the manufacturer’s protocol. Absorbance was measured via microplate reader. Transport assays Diffusion across the model was determined using glucose and IgG. In both instances, basal media for HUVECs supplemented with additional glucose (15mM) or IgG (100µg/mL) was added to the apical compartment while basal media was added to the basolateral compartment. At each time point, small samples were taken from both compartments and stored at -80°C. Plugs, designed in SolidWorks and 3D printed using an Objet500 Connex3 3D-Printer with the biocompatible material BIOMED610 (Stratasys), were press-fit into the apical compartment to ensure no leakage. Concentrations from all samples were determined simultaneously using a colorimetric glucose assay kit (BioVision, Milpitas, CA, USA) or IgG ELISA kit (Invitrogen). Absorbance was measured via microplate reader. To determine the diffusion coefficient, D, for glucose transport through the BPB model, we used a quasi-steady state approach. In this approach, the concentration profile across the membrane is assumed to be at steady state and thus does not change over time. Further assumptions applied to our setup include that the BPB and the basolateral compartment are at the same concentration in a pure solvent and that the apical compartment is a known concentration. Generally, the pure solvent is a concentration of 0mM for the molecule of interest. Here, as noted above, the basal concentration is 5mM while the concentration in the apical compartment is 52 15mM. Using this approach, flux across the diaphragm can be incorporated into the mass balance equations, yielding a simple solution (fully derived in appendix A): C     ln donor Creceiver  PtA 1 1   C o C o  L     Vdonor V  donor receiver receiver  Where C is the concentration, P is the permeability, t is time, A is the area of the membrane, L is the length of the membrane in the direction of transport, and V is the volume. Using experimental data, the permeability can be determined from the above equation. Then, using the relationship P = D*K, the diffusion coefficient can be determined if the partition coefficient, K, is known. Mathematically, K is given by the equation: Cgel Cinitial C final Vsolution K   Csolution C finalVgel To calculate the partition coefficient, we allowed an acellular gel to reach equilibrium in a solution of 5mM glucose. After 24 hours, the solution was replaced with a solution containing 15mM glucose. The concentration of the solution containing the gel was measured until equilibrium was reached. Since we knew the volumes used for the gel and solution, we could solve for the partition coefficient using the equation above. This was assumed to be the same, regardless of whether the gel was acellular or cell-laden. Finally, knowing the partition coefficient, we could solve for the diffusion coefficient for the system. Virus stocks ZIKV was isolated from the blood of a patient from Brazil and cultivated in Vero cells (ATCC) at a multiplicity of infection (MOI) of 0.1 and incubated for 4-6 53 days209. The supernatant was centrifuged at 1500rpm for 5 min, filtered (0.45 µm, cellulose acetate membrane), aliquoted, and stored at –80°C. The viral titer was determined using plaque assays on Vero cells as previously described210. Briefly, virus stocks were serially diluted and adsorbed to confluent monolayers. After 3 hours, the inoculum was removed and cells were overlaid with semisolid medium containing 1% carboxymethyl cellulose. Cells were further incubated for 5 days, fixed with 4% paraformaldehyde, and stained with 0.5% aqueous crystal violet solution (Sigma-Aldrich) for plaque visualization. Titers were expressed as plaque forming units (PFU) per milliliter. The protocol of the present study was approved by the local internal review board (IRB) under the number CAAE 52888616.4.0000.5693, in compliance with the requirements of Resolution 466/2012 of the National Health Council. All the patients enrolled in this study were only admitted if they voluntarily agree to participate, signing an informed consent form allowing virus diagnostic and isolation. Cell line infections BeWo cells and HUVECs were cultured at 37ºC. ZIKV (or mock) was added to the media at a MOI of 1. On day 4 post infection, cells were detached using EDTA, spun down, and split in two. Half of the cells were dissolved in TRIzol before being stored at -80ºC and the other half were analyzed by flow cytometry. 54 RNA isolation and RT-qPCR Following the manufacturer’s protocol, RNA was isolated from the samples in TRIzol. cDNA synthesis was performed using the High-Capacity cDNA Reverse Transcription Kit. ZIKV was detected by RT-qPCR performed on a ViiA7 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with TaqMan Universal Master Mix II with UNG and ZIKV specific primers and probe211 (Integrated DNA Technologies, Coralville, IA, USA). Flow cytometry BeWo cells and HUVECs were fixed with Cytofix/Cytoperm and then stained with 4G2 mouse monoclonal hybridome anti-envelope (produced in house)/FITC Goat Anti-Mouse IgG/IgM polyclonal (BD Biosciences) antibodies212. For NPC viability assays, NPCs were seeded and incubated at 37ºC for 4 days. Samples were then either infected directly or infected via addition to the apical compartment. NPCs were then collected following treatment with Accutase 7 days after infection. Samples were stained with LIVE/DEAD Fixable Aqua Stain and then fixed with Cytofix/Cytoperm. Data were acquired on a BD LSRFortessa instrument and analyzed using FlowJo Version 10 software. Zika virus transport assay Full model gels, containing both trophoblast (BeWo b30) and endothelial (HUVEC) cells, were fabricated in a transwell insert as described above and incubated at 37ºC for 10 days total prior to infection. ZIKV was added to the apical 55 compartment of transwell inserts at a MOI of 1. For viral inhibition experiments, chloroquine was also added to the apical compartment (50µM final concentration, as previously determined to inhibit ZIKV replication and to maintain cell viability210). Samples were then incubated at 37ºC. At each time point, cell culture media was collected from the basolateral compartment and mixed 1:1 with TRIzol before being stored at -80ºC. RNA isolation, cDNA synthesis, and RT-qPCR were performed as described above. Zika PFU was determined by comparing cycle threshold (CT) values with dilutions of known PFU/mL viral stocks. Transport gel staining Following completion of the transwell assay, BPB models were fixed with 4% paraformaldehyde for 4 hours. Gels were then permeabilized and immunostained as described above, using Zika Virus PrM Protein Rabbit Polyclonal Antibody (GeneTex, Irvine, CA, USA) and an appropriate secondary antibody. Samples were imaged via confocal microscope. Statistical analysis Quantitative results are reported as a mean ± standard deviation. Standard deviations were determined using Excel, based upon the dataset as the entire population. Significance for TEER testing was determined using a two-sample t-test via Minitab, comparing the group at each time point to the blank gel. Significance for ELISAs was determined using ANOVA with Tukey’s modification via Minitab. Significance for transport assays were determined using a two-sample t-test via 56 Minitab, comparing the full model to the blank gel, for donor or receiver groups at each time point. Significance for NPC viability was determined using an unpaired t- test. For all statistical tests, a p<0.05 was assumed to be significant. * indicates p<0.05, ** indicates p<0.01. 4.3 Results Design and development of the blood-placenta barrier model Architecturally, the placental barrier is a multi-layer barrier composed of syncytiotrophoblast, villous cytotrophoblasts, Hofbauer cells, fetal endothelial cells, and basement membrane (Fig. 4.1A-C). Our multilayer BPB model (schematic in Fig. 4.1D) has multiple layers, mimicking the trophoblast and endothelial layers that combine to form the full BPB model. We were able to cast GelMA hydrogels that supported cell growth and could be used in transwell inserts for transport studies (photo in Fig. 4.1E). Average gel thickness was determined, via image analysis, to be 200±50µm. This thickness, though larger than the physiological barrier, enabled the hydrogel to be handled for seeding both trophoblast and endothelial cells, without damaging the bulk structure of the gel. Evaluating cell growth from studies in literature, we grew the model over 10 days, allowing for cell proliferation and monolayer growth (timeline in Fig. 4.1F). In transport studies, we sought to ensure molecules must pass through the barrier model and used a transwell insert with a plug to prevent solutes from passing around the barrier instead of through it (Figs. 4.1G- K). As indicated, transport only occurs through the model (yellow arrows in Fig. 4.1G) by the inclusion of a 3D printed plug (CAD model in Fig. 4.1H, photo in Fig. 4.1I). This plug fit securely and tightly within a 12-well format transwell insert (Fig. 57 4.1J/K). The gel, not clearly visible, is within the transwell insert but below the 3D printed plug (Fig. 4.1K). Fig. 4.1. Overview of BPB Model and Placenta-Fetus System. (A) Maternal decidua, placenta, amniotic fluid, and fetus. (B) From the maternal side, the decidual vasculature feeds blood into the placenta that interacts with fetal capillary trees, forming the blood-placenta barrier. Fetal blood within these capillaries feed blood into the umbilical cord. (C) Blood-placenta barrier composition where maternal blood interacts with the syncytiotrophoblast. Underneath is a layer of villous cytotrophoblasts, followed by basement membrane containing placental macrophages, called Hofbauer cells, and then fetal endothelial cells in direct contact with fetal blood. (D) Schematic of organization within our multi-layer model: BeWo b30 cells are encapsulated is GelMA hydrogel with more BeWo b30 cells seeded on one side and HUVECs seeded on the other side. (E) Image of a BPB model. (F) Models are cast with BeWo b30 cells and crosslinked on Day 0. Following 7 days of BeWo b30 media, HUVECs are seeded and models fed HUVEC media until the transport study and other assays are performed on Day 10 PS. (G-K) Transport assay (schematic, G, and image, K) with BPB model (gray in G) and plug (H/I) within a transwell insert (J). Using this plug (I), transport is limited to passing through the gel (yellow arrows, G). 58 Validation of the formation of the blood-placenta barrier We sought to demonstrate two distinct, yet adjacent trophoblast and endothelial cell layers (Fig. 4.2A). This was supported by morphological differences between trophoblast and endothelial cells within each layer (Fig. 4.2B/C). Next, to confirm that cells formed an intercellular barrier, we evaluated expression of zonula occludens-1 (ZO-1) for both cell types, E-Cadherin for BeWo b30 cells, and VE- Cadherin for HUVECs. We observed the expression of ZO-1 for both cell types grown on a GelMA hydrogel (Fig. 4.2B). ZO-1 also localized between HUVEC nuclei (Fig. 4.2B). Similarly, we saw cadherin expression throughout both cell types and localized between HUVEC nuclei (Fig. 4.2C). To further assess the progression and formation of an intercellular barrier more quantitatively, we performed transepithelial electrical resistance (TEER) testing, evaluating each of the layers within our multilayer model as well as a blank acellular gel and the full multilayer model (Fig. 4.2D). We observed an increase in TEER values, from initial seeding through day 9 post-seeding (PS), for two groups (BeWo b30 on Top and HUVEC) where cells were only on the surface of the gel (Fig. 4.2D). Two groups (BeWo b30 Encapsulated and Full Model) that had cells encapsulated within the hydrogel maintained the highest TEER values but showed modest, if any, change in TEER values over time (Fig. 4.2D). These results were further corroborated with nuclei and actin staining showing cell proliferation and progression of monolayer formation (Fig. 4.2E). 59 Fig. 4.2. Formation of a Cell-Cell Barrier. (A) Cross-section view of BPB model via 3D rendering of confocal z-stack of nuclei cells, showing two distinct layers of cells (one trophoblast and one endothelial cell layer) within a single construct. Scale bar = 200µm. (n=3). (B) ZO-1 expression (green) with cells’ nuclei (cyan). The merge indicates intercellular ZO-1 expression for HUVECs on day 9 following seeding. Scale bar = 50µm. (n=3) (C) E-Cadherin (BeWo b30) and VE-Cadherin (HUVEC) expression (both green) with cells’ nuclei (cyan). The merge indicates intercellular cadherin expression for HUVECs on day 9 following seeding. Scale bar = 50µm. (n=3) (D) TEER testing for each layer of the multilayer BPB model, with the blank gel (black) having the lowest value. The next highest is BeWo b30 cells (red) seeded on top of the gel, followed by HUVECs (blue) seeded on top. BeWo b30 cells encapsulated within the hydrogel (green) and the full multilayer model (purple) showed the highest TEER values over 9 days. * indicates significance (p<0.05) for the group at a given time point, compared to the blank gel. (n=3, data reported as mean ± standard deviation) (E) Fluorescent staining showing nuclei (cyan) and actin (magenta) at 3, 5, 7, and 9 days post-seeding, showing cell proliferation within each layer, towards the formation of a monolayer. Scale bar = 50µm. (n=3) 60 Cellular bioactivity and model functionality We assessed the bioactivity of cells within this model and the transport characteristics of the model. We first determined viability, qualitatively, of both cell types within this model (Fig. 4.3A). The vast majority of cells appear to be alive (green) with minimal, if any, dead (red) cells throughout the 3D construct (Fig. 4.3A). We next assessed cellular specificity and functional bioactivity by staining for human endogenous retroviral protein (HERV), a protein expressed by trophoblast, and measuring secretions of human chorionic gonadotropin (hCG), vascular endothelial growth factor (VEGF), and progesterone. hCG and progesterone are hormones that are key biomarkers of placental health while VEGF is a proangiogenic growth factor expressed by healthy endothelial cells. Here, we utilized hCG and progesterone as markers of trophoblast function and VEGF as a marker of endothelial function. BeWo b30 cells are positive for HERV expression (Fig. 4.3B). hCG and progesterone secretions are significantly higher (p<0.05) for BeWo b30 on GelMA and the full model, compared to BeWo b30 on TCPS (Fig. 4.3C,D). Similarly, VEGF secretions are significantly higher (p<0.05) for BeWo b30 on GelMA and the full model, compared to BeWo b30 on TCPS and HUVEC on TCPS or GelMA (Fig. 4.3E). The full model gel has a decrease in progesterone and VEGF secretions, compared to BeWo b30 on GelMA, though hCG remains similar. Why this occurs is unclear, though we speculate that it may be due to co-culture of these cells causing changes in cell secretions213. Regardless, these data suggest that the cells within the full model secrete bioactive proteins, recapitulating key aspects of the in vivo phenotype. 61 To evaluate transport across our barrier model, we used both glucose and immunoglobulin G (IgG) as example molecules. Glucose has been extensively studied in numerous other placental transport models, allowing us to draw comparisons between this model and others using either the diffusion coefficient of glucose through the model or the percent rate of transfer158,166. IgG has also been studied and is a much larger molecule that should not readily cross the placental barrier214,215. Glucose added to the apical compartment reaches equilibrium with glucose in the basolateral compartment after 12 hours in blank acellular gels (Fig. 4.3F), whereas equilibrium is reached in 24 hours with a full multilayer cell-laden gel (Fig. 4.3F). These studies were performed utilizing plugs to prevent leakage around the gel. IgG passively diffuses across acellular gels but has significantly (p<0.05) lower basolateral concentrations when using full multilayer cell-laden gels (Fig. 4.3G). Additionally, IgG remains far from reaching equilibrium, even at 8 hours (Fig. 4.3G). More rigorously, we determined the partition coefficient of glucose and an acellular GelMA hydrogel to be 0.166, though this would change when using a cell- laden model. The time-average diffusion coefficient for transport through an acellular gel was 6.97x10-6 cm2/s and with a permeability of 1.16x10-6 cm2/s. These values decreased for a full model, with a time-average diffusion coefficient of 2.88x10-6 cm2/s and permeability of 4.79x10-7 cm2/s. Additionally, we determined the percent rate of transfer, as defined in literature158, to be 66.2% for a blank acellular gel and 35.9% for the full model gel. 62 Fig. 4.3. Cellular Bioactivity, Specificity, and Model Functionality. (A) Live-Dead images showing the vast majority of cells are alive (green) with little to no dead cells (red) at day 9 PS. Scale bar = 200µm. (n=6). (B) Positive expression of human endogenous retroviral protein (HERV) for BeWo b30 cells used within the model. Scale bar = 100µm. (n=4). (C/D/E) Secreted human chorionic gonadotropin (hCG) (C), progesterone (D), and vascular endothelial growth factor (VEGF) (E) concentrations for BeWo b30 cells and HUVECs, on tissue culture polystyrene (TCPS), a GelMA hydrogel, or within co-culture as the full model. (n≥4, groups that do not share a letter are significantly different, p<0.05) (F/G) Glucose (F) or IgG (G) was added to the apical compartment and the concentration in both the apical compartment (donor; shown with filled-in circles) and the basolateral compartment (receiver; shown with empty circles) was measured for a blank acellular gel (black) or a full multilayer cell-laden model (red). * indicates significant (p<0.05) difference between the blank gel and full model gel at the same time point. (F, n=3; G, n=4). Data reported as mean ± standard deviation. 63 Zika virus pathogenesis We began our ZIKV work by evaluating the susceptibility of trophoblast (BeWo) and endothelial cells (HUVECs) to ZIKV infection using qRT-PCR and flow cytometry. We used a ZIKV strain isolated from Brazil, as this strain is more relevant for the recent outbreak in the Americas than the commonly used MR766 strain from Africa209. Through both qRT-PCR and flow cytometry, we found that trophoblasts (BeWo) were more susceptible to infection than endothelial cells (Fig. 4.4A/B). We then evaluated ZIKV transport across the BPB model, containing BeWo b30 cells and HUVECs, to determine how much virus crosses the placental barrier and is thus exposed to fetal cells. ZIKV was added to the apical (maternal) compartment of the model and plaque forming units (PFUs) in the basolateral (fetal) compartment measured over 72 hours (Fig. 4.4C). In all experiments, PFUs reached a peak earlier and of higher magnitude when using the full BPB model, compared to acellular blank gels. At the peak, PFUs measured for the full BPB model were higher than the initial amount of ZIKV used to infect the model, suggesting ZIKV replication in the BPB model. Further, chloroquine treatment reduced the PFUs in the fetal compartment and infection of the cells of the BPB, both quantitatively (Fig. 4.4C) and qualitatively (Fig. 4.4D). Chloroquine is currently clinically used as an anti-malaria treatment, and has previously been shown to inhibit ZIKV infection due to its activity as an endocytosis inhibitor and preventing endosome acidification210 and to protect fetal mice from microcephaly216,217, while maintaining cell viability (Fig. 4.5). Together, this suggests that viral replication within placental cells is 64 responsible for the transient increase in ZIKV on the fetal side when using the full BPB model compared to the acellular blank gels. Fig. 4.4. ZIKV Infection, Transmission Across BPB Model, and NPC Outcomes. (A) Fold change in ZIKV infection (MOI 1) indicating trophoblast cells (BeWo) are more susceptible to infection than endothelial cells (HUVEC). Data is normalized to the BeWo group. (n=2) (B) Representative flow cytometry plots of ZIKV infection or mock control treatment for trophoblast cells (BeWo) and endothelial cells (HUVEC) 4 days post infection. (n=2) (C) ZIKV present within the media in the fetal (basolateral) compartment at each time point was determined by qRT-PCR for blank gels, full model and full model treated with chloroquine (n=3). Standards with known PFU concentrations were included in the qRT-PCR to calculate the values from the samples. The full model gels show peak fetal PFU at 12 hours, whereas blank and chloroquine-treated samples do not. (n=3) (D) Full model gels with or without chloroquine at 4 days post infection showing Zika M protein. Scale bar = 100µm. (n=3) (E) Viability of neural progenitor cells (NPCs) in the fetal compartment of our system 7 days after infection. ZIKV added to either a blank acellular gel, a full model cell-laden gel, or directly to NPC without model present. **p<0.01. (n=3). Data reported as mean ± standard deviation. 65 Fig. 4.5. Chloroquine dose effects on cell viability. Microscopy images depicting live (green) or dead (red) cells, either BeWo b30 or HUVEC, following exposure to chloroquine for 4 days. A lack of image indicates no cells present at that concentration for that cell type. Scale bar = 50µm (n=4). Zika virus-induced neural cell death We used human iPSC-derived NPCs207 to determine the viral effects downstream of the barrier, as NPCs have been well-validated to be affected and infected by ZIKV218. For this purpose, iPSC-derived NPCs were seeded in the basolateral compartment and viability of the cells was determined 7 days after addition of ZIKV or a mock-infected negative control. As a positive control, ZIKV was added directly to fetal neural cells in absence of the BPB model. This direct addition of ZIKV to the fetal neural cells in absence of BPB model led to a significant (p<0.01) decrease in viability of the cells as a result of ZIKV infection, compared to mock-infection (Fig. 4.4E). A significant (p<0.01) reduction in cell viability of fetal neural cells was also observed when ZIKV was added to blank gels. Interestingly, sustained viability of fetal neural cells was observed when the full BPB model was used (Fig. 4.4E). This indicates that NPC death occurs following ZIKV exposure, but 66 suggests that the placental cells may influence the extent of NPC death that occurs. Given that the rate of neural damage is wide-ranging (6-40+%4,201,219), we suggest that placenta, a site of preferential viral replication203, provides some protective effect by “soaking up” ZIKV, thereby limiting fetal exposure. 4.4 Discussion Zika virus has been linked to microcephaly and other birth defects at highly different rates4,201,202,219. Studies to explain the observed effects have been conducted in vitro using primary cells and cell lines202,220, utilizing bead-based models221, in mice203,222, and in non-human primates223. However, despite progress towards understanding the mechanisms and biomolecules related to ZIKV infection or lack thereof, there remain questions regarding viral transmission from mother to fetus. To address this critical question, we sought to develop a biomimetic, easy-to-use model of the BPB that could be used to investigate ZIKV transmission across the BPB and the effects of ZIKV on fetal neural tissue. We developed this model with the focus on recapitulating the 3D organizational structure of cells and extracellular matrix of the BPB. We utilized a 3D GelMA hydrogel with BeWo b30 cells and HUVECs, comparable to trophoblast and endothelial cells, respectively, utilized in placenta-on-a-chip and other approaches158,159, to mimic the multilayer structure within the BPB. Our inclusion of endothelial cells within this model is supported by their presence in the BPB, their highly regulated permeability, and previous work indicating changes in placental transport and cell bioactivity in trophoblast-endothelial cell co-culture systems, compared to trophoblast-only systems123,158,159,166. Additionally, ZIKV can infect 67 endothelial cells, suggesting that their inclusion leads to increased viral loads exposed to the fetus224. We also utilized neural progenitor cells to mimic fetal neural cells, allowing us to evaluate ZIKV-induced neural outcomes in vitro. Although other in vitro models could utilize fetal cells in their approach, many lack endothelial cells and are thus less relevant to the human BPB161. To our knowledge, no other studies have evaluated effects on both the BPB and fetal cells in vitro. Given that we could recapitulate the multilayer structure of the in vivo placental barrier and that we utilized key cell types relevant to the study of Zika Virus infection of the placenta and downstream fetal tissue, we consider this model to be a valid approach in recreating the BPB to study the effects described throughout this study. The 3D model presented herein is comparable to other 2D and 3D models of the BPB recently described in literature. One notable approach is the placenta-on-a- chip microfluidic platform, a 2D approach that allows for perfused flow across cell barriers. In the study by Blundell, et al., they noted a glucose transfer rate of 34.8% for the placenta-on-a-chip and 26.5-38.3% for the ex vivo perfused placenta, the gold standard within the placental transport field158. The rate of transfer for the 3D model discussed here is 35.9%, which is comparable to that of ex vivo perfused placenta, even though the model discussed herein does not utilize flow nor does it have syncytialized trophoblast cells. Poulsen, et al, also found that a BeWo b30 monolayer without flow or syncytialized cells correlated well with the perfused placenta approach, though the BeWo b30 monolayer took longer to reach maternal-fetal equilibrium164. This is interesting, as flow would likely impact the placental barrier model through shear-induced effects on the fetal endothelial cells, as well as 68 impacting transport through changes in the residence time for a molecule at the interface between flow and the maternal-facing trophoblast. Additionally, syncytialized trophoblast are considered the primary barrier to placental transport, and have been shown to respond differently than BeWo cells to stimuli, such as ZIKV infection202,220. Thus, it is likely that both flow and syncytialized would impact the observed transport of any molecule or ZIKV across the BPB, even though our data suggest comparable rates of transport without these two factors. Additionally, Levkovitz, et al, reported a diffusion coefficient of 3.9-4.5 x 10-7 cm2/s for glucose using an amniotic membrane with cells and a reported thickness of 27-43µm159,160. By comparison, we calculated a time-average diffusion coefficient of 2.88x10-6 cm2/s, an order of magnitude higher, even though this model is an order of magnitude greater in thickness. Similarly to flow, it is likely that thickness would impact the residence time of a molecule, with more molecules being taken up by thicker membranes. Further, given that the in vivo thickness has been reported at 2-100µm159, dependent on the stage of pregnancy, it would be interesting to further investigate the role of thickness in influencing transport across the barrier. Additionally, utilizing a biological membrane, such as done by Levkovitz, could lead to a dramatic difference in diffusion through the membrane, compared to utilizing a fabricated hydrogel. The extent to which flow, syncytialized cells, and membrane thickness impacts molecular transport would be of interest to elucidate more rigorously, particularly as the field transitions from 2D to 3D models. This model allowed us to study how in utero transmission of ZIKV occurs and what effect ZIKV has on fetal neural cells. Previous in vitro studies convincingly 69 show ZIKV infection of placental cells and neural cells202,218,220,224, but none have shown viral transmission or neural cell outcomes following transmission. Interestingly, we saw the highest viral load in the fetal compartment at 12 hours post- infection and a subsequent reduction, suggesting that the placenta may modulate the viral exposure at longer periods of time. Our findings are in agreement with those by Miner, et al., suggesting that ZIKV may preferentially interact with the placenta (the “soaking up” effect mentioned earlier)203. This is critically important as how and why the placenta, and its constituent cells, are affected by ZIKV is not well recognized202,220. Moreover, the role the placenta plays in regulating ZIKV passage to the fetus remains a largely unanswered question, though some studies have begun to identify factors involved203,220–222. Given that our findings agree with those found in vivo, this 3D model may be better suited than existing BPB models to address some these questions. Here, we saw NPC death from ZIKV exposure, but surprisingly, ZIKV transmission through the full model gel does not lead to significantly increased NPC death, compared to the mock-infected group after 7 days. This again hints at a preferential interaction between ZIKV and the placenta, though more detailed experimentation is needed to clarify this interaction. It is also possible that soluble factors produced by the cells in the full model, such as interferons220,221 or differentially expressed cytokines, provide a protective effect to the NPCs. In vivo, ZIKV is detected in the amniotic fluid, sometimes in absence of virus in the maternal blood. Our results suggest that viral replication in the BPB could contribute to this phenomenon. The decreased in viral load that we observed after 12 hours could be because there was no cells in the basolateral compartment to sustain viral replication 70 after the virus had crossed the model. The data presented here provides some insight into in utero transmission of ZIKV, and we suggest that this model can be used to study in utero transmission in more mechanistic detail and to assess fetal neural cell effects resulting from ZIKV exposure. Although this model provides a biomimetic approach to understanding maternal-fetal transport, ZIKV pathogenesis, and in utero virus transmission, there are some limitations. One key limitation is the lack syncytialized trophoblast and Hofbauer cells. Syncytiotrophoblast can produce interferons, which are thought to impact ZIKV transmission and fetal outcomes220–222. ZIKV can infect Hofbauer cells, thus it is possible that our observations would be influenced with their inclusion220,225. Further, although newer culture methods have demonstrated longer-term viability of primary trophoblast cells166, we used established cells lines herein as a proof-of- concept. Given the versatility of our approach, primary cells can be utilized in our model system to ultimately better recapitulate the function of the syncytiotrophoblast. There are noted differences in cell response between immortalized cell lines and primary cells220, which would be of interest to study further. Additional future work involves deeper investigation into ZIKV’s effects on placental and neural cells to further elucidate some of the effects we have discussed here. 4.5 Summary In summary, we have developed a model of the maternal-fetal interface that recapitulates key aspects of human placental biology allowing us to investigate the effects of ZIKV on the BPB and downstream neural cells. Interestingly, we observed that placental cells may function as a modulator of the viral load to downstream fetal 71 tissues, though additional studies are necessary to elucidate this effect. This biomimetic approach and model provide potentially scalable tools for assessing placental transport of substances in the future, including ZIKV therapeutics, while improving our understanding of maternal-fetal environment. 72 Chapter 5: Assessing Placental Barrier Phenotypic Response to SSRIs Fluoxetine and Sertraline 5.1 Introduction The placental barrier, termed the blood-placenta barrier (BPB), is a physiologic membrane that regulates bi-directional maternal-fetal substance exchange. Generally, this barrier limits the exposure of harmful substances to a fetus. Historically though, there have been instances of where this barrier did not prevent fetal damage, such as the recent Zika Virus epidemic leading to neural abnormalities, or the period in the 1960s where thalidomide, a prescription medication, led to severe limb malformations1,2,201. In both instances, there was no a priori knowledge about whether these substances would cross the BPB, leading to significant consequences in fetal development. Even more alarming, in the case of thalidomide, doctors prescribed medication during pregnancy with no knowledge of its teratogenic effects on the fetus. The failure to test for safety and toxicity continues to be a high concern since nearly 90% of pregnant women report taking at least one medication during the course of pregnancy5,226. Therefore, there is a critical unmet need for efficacious models to assess the transport and toxicity of substances across the BPB to ensure fetal safety. In particular, antidepressant use during pregnancy poses a significant health safety issue as up to 13% of pregnant women use antidepressants and its incidence continues to rise227. The most commonly prescribed class of antidepressants are Selective Serotonin Reuptake Inhibitors (SSRIs), with up to 6% of newborns 73 having prenatal exposure to SSRIs6,227. Given the increasing usage of SSRIs, the wide acceptance that they cross the BPB, and the limited knowledge regarding the drugs’ effects on a fetus, additional studies are critically needed to better understand the potential effects of taking these drugs during pregnancy5–7,227,228. Two of the most common SSRIs prescribed during pregnancy are fluoxetine (1.9% of pregnancies) and sertraline (2.1% of pregnancies)5,227, more commonly known by their brand names Prozac® and Zoloft®, respectively. Although initially approved as antidepressants in the mid-1980s (fluoxetine) and the early 1990s (sertraline), studies evaluating their effects when used during pregnancy didn’t occur until the mid-to-late 1990s, where they primarily focused on neonatal outcomes and the overall well-being of the baby229. Since then, numerous studies have evaluated how SSRIs lead to the intended neurological effects, as well as the unintended side effects on other tissues, including neurodevelopmental effects9,230, cardiovascular- related effects231–234, effects on endothelial cells (ECs)231,235–238, and, more recently, effects on the placenta itself and downstream fetal tissue8,239–241. As it is widely accepted that SSRIs readily pass through the placenta and can impact fetal development, studies related to the role of the placenta and drug-induced effects on the placental barrier have become increasing important7,206,228,232. To our knowledge, no study has yet to explain the transporter proteins involved in SSRI trans-placental passage. Moreover, it would be difficult, if not impossible, to show a specific mechanism given the numerous transporters in the placenta with overlapping functionality178. In spite of this, some studies have shown that SSRIs, particularly sertraline, can interact with and potentially inhibit P-glyocoprotein (P-gp)236, a drug 74 efflux pump that, with breast cancer resistance protein (BCRP), limits fetal drug exposure by transporting drugs form the placenta to maternal circulation242. Therefore, further elucidating these drug-transporter protein interactions is critically important to limiting fetal exposure and adverse effects. Even more so, it may be important if SSRIs impact transport or metabolism of other drugs, leading to adverse drug-drug interactions during pregnancy236. Adverse effects arise following a drug’s passage across the BPB since it immediately enters into fetal circulation and can lead to drug-induced effects throughout the entire developing fetus243,244. Previous work has shown that prenatal or perinatal exposure to SSRIs may lead to preterm birth, low birth weight, congenital heart issues, or potentially autism8,233,245. Conflictingly though, other studies have argued the incidences of these health risks are no different between SSRI-exposed and non-exposed babies, and that SSRI exposure may actually be protective against some of these risks14,245. On a cellular level, SSRIs have previously been shown to modulate cell adhesion molecule (CAM) expression mediated by a pro-inflammatory response and in depressed patients238,246. Additionally, CAMs were previously shown to be altered in placental disease (i.e. pre-eclampsia and villitis)247,248, suggesting we need to better understand the extent to which SSRIs influence CAM expression to ensure drug safety during pregnancy. Similarly, transforming growth factor beta (TGFβ) expression has been shown to change during treatment for depression249. Interestingly, it has been suggested that increased TGFβ leads to damage to the BPB through the destruction of tight junctions250, thereby making the BPB less effective at limiting fetal exposure to drugs. Given these observations and with the rise in use of SSRIs during pregnancy, 75 understanding the cell-level phenotypic changes that result from SSRIs, the potential drug-drug interactions, and the questions regarding general medication safety for both maternal health and fetal development are becoming increasingly important to answer immediately5,227. Moreover, as the placenta regulates transport and given that placental signaling pathways play a role in development8,251, it is clear that we need a biomimetic and robust in vitro system capable of probing drug-induced effects on both the placenta and downstream fetal tissue. As previously discussed, current in vitro systems for modeling placental transport include the ex vivo perfused placenta model and in vitro models, including transwell approaches and placenta-on-a-chip approaches16,158,165. The ex vivo perfused placenta approach, though the most biomimetic, is dependent on the tissue, which can be damaged during delivery and have a compromised barrier16. Through in vitro models to evaluate placental transport of drugs exist, they often lack the cellular complexity and 3D nature of the placental barrier158,165. In particular, extracellular matrix (ECM) proteins have rarely been incorporated into these models, even though studies have shown that ECM proteins (i.e. laminin) are important in placental biology180. Similarly, to our knowledge, no studies have evaluated whether endothelial cell phenotype influences the placental barrier model or molecular permeability across the model, even though previous literature has shown differences in permeability and phenotypic heterogeneity when comparing macrovascular (i.e. umbilical vein) and microvascular (i.e. brain or dermal microvascular) ECs158,181,200,252,253. Therefore, understanding the role of both ECM and EC type 76 within a placental barrier model is important to ensuring robustness, reliability, and relevance of the model to in vivo studies. In the previous chapter, we described developing our model system to study maternal-fetal transmission of Zika Virus and fetal neural toxicity, where we observed similarities between our model and in vivo studies203,243. This model mimics the multilayer complexity of the BPB and enables probing of both the placenta and downstream fetal tissue. Given that the model includes both trophoblast and endothelial cells, both thought to be impacted by SSRIs236,240,246, we consider this model to be a valid approach in studying the effects of these drugs on the placental barrier cells. Herein, we utilize this model with the goal of understanding: (1) the time-dependent concentration profile of SSRIs (fluoxetine and sertraline) in maternal- fetal transmission and transporter proteins that alter this profile, and (2) to what extent the BPB barrier proteins (i.e. CAMs) and vascular markers (i.e. transforming growth factor beta, or TGFβ) are impacted, indicating intercellular adhesion and vascular health of the BPB. Given the widespread usage of fluoxetine and sertraline, we also wanted to assess whether two drugs within the same class show the same maternal- fetal transport profile and elicit the same alterations to barrier proteins and vascular markers, or whether each drug’s effect would be unique. We hypothesized that both of these drugs would freely diffuse across the BPB with similar profiles, but that the impact on barrier proteins and vascular markers would be dependent upon the specific drug used, the drug concentration, and the length of drug exposure. Herein, we report on: (1) the effects of EC type and ECM on the placental barrier model; (2) the trans- placental transport profile for both drugs and the extent to which inhibition of drug 77 efflux pumps impacts these profiles; and (3) the response of the BPB model and each cell type there within to fluoxetine and sertraline at multiple dosages and lengths of exposure. In total, we suggest that this work furthers our understanding of the transplacental transport concentration profiles of these drugs and the differential effect between these drugs on cells within the BPB. 5.2 Materials and Methods Cell culture and assays BeWo b30 cells (trophoblast) were both obtained from Erik Rytting (University of Texas Medical Branch, Galveston, TX) and purchased from AddexBio (San Diego, CA), and were grown in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco) with fetal bovine serum (FBS, Gibco), antibiotic-antimycotic (A/A, Gibco), non-essential amino acids (Gibco), and L-glutamine (Gibco) supplemented. Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and grown in Endothelial Cell Growth Medium (Promocell, Heidelberg, Germany). HUVEC medium was used for BeWo b30/HUVEC co-culture studies. Human brain microvascular endothelial cells (HBMECs) were purchased from Cell Systems (Kirkland, WA) and grown in RPMI-1640 supplemented with FBS, endothelial cell growth supplement (ECGS), heparin, and A/A. To create the 3D BPB model, we utilized a similar approach as previously described in the previous chapter, with minor modifications regarding composition and timing. 3D hydrogel constructs were fabricated using gelatin methacrylate (GelMA, 10% w/v) and decellularized placental extracellular matrix (pECM, 10% v/v). This composition was also used for monoculture studies, where the hydrogel 78 precursor solution was coated onto multi-well dishes for 1 hour at room temp and washed thoroughly with phosphate buffered saline (PBS, Gibco) prior to seeding cells. Regarding timing changes, constructs were grown for 7 days in total prior to assaying, with ECs being seeded on day 4 following construction fabrication. All assays, including those in monoculture, followed this same timeline. Placental decellularization and characterization Placental tissue was collected and decellularized similar as previously described180. In brief, ex vivo term placental tissue from normal pregnancies were collected and frozen until ECM isolation (MedStar Washington Hospital Center, IRB# 2015-131). Surgical tools were utilized to mechanically digest the whole placental tissue, excluding the umbilical cord. The digested tissue was washed using PBS until the effluent ran clear, wherein the tissue was placed in 1% A/A and 0.1% (w/v) Sodium Azide (NaAzide, Sigma-Aldrich) in PBS and stirred overnight. The following day, the tissue was placed in a solution containing sodium dodecyl sulfate (SDS, 0.1%) and 0.1% NaAzide in PBS, under stirring at room temperature for 48 hours, to decellularize the tissue. The tissue was repeatedly washed over 2 days using 0.1% NaAzide in DI Water, until being frozen and lyophilized for 5 days. Following lyophilization, dry tissue mass was recorded and the tissue mechanically broken down prior to pepsin digestion. To solubilize ECM proteins, the dry tissue was placed in acetic acid (0.5M, 1% w/v) with pepsin (10% w/w), the solution mixed, and left at 4°C for 48 hours. Following pepsin digestion, the solution was pH-adjusted to 7.4, spun to pellet the tissue, and the supernatant run through a cell strainer (40µm). This 79 solubilized protein solution was the sterile filtered, with the final product being the decellularized placental ECM (pECM) added into assays. To characterize the pECM, samples were run through PicoGreen (Life Technologies) and Pierce® BCA (ThermoFisher) assays to determine DNA and protein content, respectively, as per manufacturer’s instructions. The digested ECM was concentrated via centrifuge tube (Pall Corporation, 10kDa MWCO) prior to being prepared and run on liquid chromatography-mass spectrometry (LC-MS), as previously described180. Peptide fraction was calculated as the number of peptides within a given protein family divided by the total number of peptides counted for the sample. Molecular transport assays Diffusion across the model was studied by adding a known concentration to the apical compartment and taking samples from both the apical and basolateral compartments at multiple timepoints. Molecules assayed include: fluorescein sodium salt (fluorescein, 1µg/mL, Sigma-Aldrich), immunoglobulin G (IgG, 100µg/mL, Sigma-Aldrich), fluoxetine (1,000 ng/mL, Sigma-Aldrich), and sertraline (1,000 ng/mL, Sigma-Aldrich). Fluorescein concentrations were determined by generating a known standard and measuring fluorescence intensity for all samples relative to the standard. Samples were kept in the dark until being assayed via plate reader, within 4 hours of initiating the transport study. IgG samples were collected and frozen at - 80°C until being assayed, where concentrations were determined via ELISA (Invitrogen) according to manufacturer’s protocols. Fluoxetine and sertraline were 80 detected via forensic kits (Neogen) specific to each molecule and their metabolite, with all samples diluted, as appropriate, to within the linear range of the detection kit. Samples were stored at -80°C until being assayed, simultaneously, following manufacturer’s instructions using absorbance on a plate reader. For all studies, 3D printed plugs were used, as previously described243. Calculations for permeability and diffusion coefficient were also performed as previously described in detail243. Drug exposure and cytokine measurements For drug exposure experiments, fluoxetine hydrochloride and sertraline hydrochloride (Sigma-Aldrich) were solubilized in dimethyl sulfoxide (DMSO) and diluted in cell culture media to the stated concentration/dosage. DMSO was controlled in all experiments such that higher drug concentrations did not necessarily indicate higher DMSO dosages, with all media containing <1% DMSO. For all exposure experiments, media was replenished daily to simulate clinical administration of drugs. To determine concentrations of intercellular adhesion molecule-1 (ICAM), vascular cell adhesion molecule-1 (VCAM), and transforming growth factor beta-1 (TGFβ), ELISAs were performed. DuoSet kits (R&D Systems) for each target were purchased and used per manufacturer’s protocols. At each time point, cell culture supernatant was collected, spun at 10,000g for 10 min at 4°C, then frozen at -80°C. Samples were thawed at room temperature prior to running each assay. Absorbance was measured via plate reader. 81 Transepithelial electrical resistance (TEER) testing Transepithelial electrical resistance (TEER) testing was performed to determine barrier resistance over time. In all instances, barrier models were fabricated in transwell inserts using the process described above. Co-culture studies involved seeding BeWo b30 cells during construct fabrication with ECs seeded later, with measurements taken following seeding all cells. For pECM experiments, BeWo b30 cells were seeded on hydrogels (105 cells/cm2) containing 1% or 10% (v/v) pECM in GelMA hydrogels. For exposure experiments, samples were exposed to drug- containing media (10,000 ng/mL dosage), with daily media changes. All values were normalized against values obtained for PBS and multiplied by the hydrogel’s surface area, assumed to be equal the transwell insert growth area. Acellular GelMA values were averaged over initial and final time points and presented as such. Immunofluorescent staining Immunostaining experiments were performed to visualize protein expression. BPB models (i.e. ‘thick’ hydrogels) were fixed with 4% paraformaldehyde for 4 hours followed by repeated washes with PBS. Samples were blocked with 5% goat serum, then incubated with primary antibodies against P-gp or BCRP overnight at 4°C. The following day, samples were extensively washed, followed by incubation with the appropriate secondary antibody for 1 hour. Following additional washes, 82 samples were counterstained with DAPI and washed prior to imaging via fluorescent microscope. Statistical analysis Quantitative results are reported as a mean ± standard deviation, calculating using Excel. Significance for ELISAs was determined using either ANOVA with Tukey’s modification, as represented by letters, or using a two-sample t-test, where represented by asterisk, via Minitab. In graphs with multiple days and dosages, ANOVA was used across dosage for the same timepoint while a t-test was performed across timepoints for the same dosage. For all statistical tests, a p<0.05 was assumed to be significant. * indicates p<0.05. 5.3 Results ECM, Not EC Type, Influence BPB Function In comparing HUVEC and HBMEC cell types, macrovascular and microvascular cell types, respectively, both of which often used in barrier models, we evaluated TEER values and molecular transport and permeability, hypothesizing that the microvascular cells (i.e. HBMEC) would have higher TEER values and reduced permeability, given that HBMEC are utilized in the highly regulated BBB, as noted in Chapter 2. In initially evaluating TEER measurements, we observed no significant differences (p>0.05) between HBMEC and HUVEC coculture samples over a period of 14 days in culture (Fig. 5.1A). We further evaluated transport profiles where fluorescein diffusion curves showed minimal differences between each cell type (Fig. 83 5.1B). This was further evident through permeability calculations, where IgG permeability was within the standard deviation for each cell type, though fluorescein showed reduced permeability when using HBMEC (Fig. 5.1C). In total, there appeared to be modest, if any, differences between these cell types within the context of this BPB model, suggesting either cell type is appropriate when developing models of the BPB. Simultaneously, we characterized decellularized pECM and evaluated its impact on cell phenotype. Following digestion and proteomics analysis, we found that multiple ECM protein families were present within the pECM, including laminin, collagen, fibrinogen, and fibronectin (Fig. 5.1D). We quantified DNA and protein content, finding minimal DNA (<100 ng DNA/mg dry tissue) and modest protein content (30 µg protein/mg dry tissue) within the solubilized pECM (Fig. 5.1E). When evaluating TEER values, there were no significant differences (p>0.05) based on relative pECM content (Fig. 5.1F). Last, when evaluating phenotypic changes in cells through TGFβ secretions, we observed pECM addition had no impact on HUVECs but significantly decreased (p<0.05) BeWo b30 secretions of TGFβ, suggesting that pECM influences BeWo b30 secreted growth factors and phenotype (Fig. 5.1G). 84 Fig. 5.1. Evaluation of Endothelial Cell Type and Placental ECM. (A) TEER values increased over 14 days with minimal differences between HBMEC or HUVEC in co-culture with BeWo b30 cells. (n=3). (B) Fluorescein sodium salt shows similar transport profiles across the barrier models using either HUVEC or HBMEC in the model. (n=3). (C) Permeability for fluorescein sodium salt is slightly higher when using HUVECs in the model instead of HBMEC, but is comparable when evaluating immunoglobulin G transport. (D) The top 10 protein families present within the solubilized placental ECM after proteomics analysis, by peptide fraction. (n=1). (E) Relative DNA and protein content for the solubilized ECM compared to the dry tissue recovered after lyophilization of the decellularized tissue. (n=3). (F) TEER values for BeWo b30 cells grown on GelMA with either 1% or 10% (v/v) ECM in the hydrogel, showing minimal differences between the two concentrations. (n=3). (G) TGFβ secretions for HUVECs and BeWo b30 cells on GelMA-coated dishes with (+) or without (-) placental ECM (10% v/v) included in the coating. Addition of ECM significantly (p<0.05 for groups that do not share a letter) decreases TGFβ secretions for BeWo b30 cells. (n=3). Data, excluding panel D, reported as mean ± standard deviation. 85 BPB Uptakes and Slowly Releases SSRIs We evaluated trans-placental transport for both fluoxetine and sertraline within the BPB model to better understand their transport profiles, similar to what has been done with SSRIs and the blood-brain barrier254,255. When adding drug to the donor side, we found a decrease on the donor side with no corresponding increase on the receiver side, for either fluoxetine (Fig. 5.2A) or sertraline (Fig. 5.2B). In both instances, adding elacridar as a P-gp/BCRP inhibitor led to further reduced donor concentrations (Fig. 5.2A/B), showing that both SSRIs are substrates for these transporter proteins. Following repeated daily exposures, we found the transport profile on day 3 was comparable in characteristics for day 1, though the receiver side appeared to have accumulated drug over time with repeated dosages (Fig. 5.2C). In determining the transport-related coefficients, we found fluoxetine had a much higher partition coefficient than sertraline while addition of inhibitor led to modest increases in the diffusion coefficient (Fig. 5.2D). This suggests that inhibiting P-gp and BCRP leads to increased diffusion across the barrier, which is expected given that these efflux pumps work to pump drugs from the placenta back into maternal blood256. Upon staining and imaging for P-gp (Fig. 5.2E) and BCRP (Fig. 5.2F) expression within the cells, we found BeWo b30 cells more highly expressed these proteins than HUVECs, though both cell populations showed positive expression (Fig. 5.2E/F). 86 Fig. 5.2. Transport profile for SSRIs crossing the BPB. (A) Relative fluoxetine concentration over the first 8 hours following addition of fluoxetine to the apical (donor) side, with and without inhibitor (elacridar) treatment prior to the addition of the drug. (n=3). (B) Relative sertraline concentration over the first 8 hours following addition of sertraline to the apical side, with and without inhibitor treatment prior to the addition of the drug. (n=3). (C) Relative fluoxetine concentration over the first 8 hours of day 3 (i.e. hours 48 to 56) following addition of the drug to the apical side of the barrier, with or without inhibitor treatment prior to the addition of the drug. Here, after multiple days of treatment, the basolateral side (receiver) is not at a relative concentration of 0%. (n=3). (D) Calculated partition coefficients for each drug, and diffusion coefficients for the transport profiles within the first 8 hours following treatment at days 1 and 3, with and without inhibitor. (E) Positive expression of P- glycoprotein (green) for BeWo b30 cells and HUVECs, with a nuclear counterstain (blue). Scale bar = 100µm. (F) Positive expression of breast cancer resistance protein (green) for BeWo b30 cells and HUVECs, with a nuclear counterstain (blue). Scale bar = 100µm. (n=3). Data reported as mean ± standard deviation. 87 SSRIs Induce Barrier Function Changes We evaluated phenotypic changes by the BPB model in response to differing drugs and dosages. We found BPB model cells show a dosage dependent response, in terms of cell secretions, to a wide range of fluoxetine (Fig. 5.3A) and sertraline (Fig. 5.3B) dosages. Further, this response was time-dependent with fluoxetine leading to the highest ICAM secretions after 3 days of exposure at a dosage of 10,000 ng/mL (Fig. 5.3A). Sertraline exposed samples showed no dosage effect at the day 3 time point (Fig. 5.3B). These increases in CAM secretions suggest reduced cell-cell adhesion, and thus a reduced barrier function and potential for increased fetal drug exposure. In directly comparing high concentrations (10,000 ng/mL) of the drugs, we observed no differences in ICAM secretions in response to each drug (Fig. 5.3C). However, sertraline led to significantly decreased (p<0.05) TGFβ secretions (Fig. 5.3D) while both fluoxetine and sertraline led to significantly (p<0.05) decreased VCAM secretions (Fig. 5.3E), compared to a no exposure control. Both decreases suggest a protective effect to the barrier, with limited drug-induced damage. TEER values showed no significant changes (p>0.05) over time, though sertraline exposure appeared to reduced TEER values over extended exposure (Fig. 5.3F), again suggesting a reduced barrier function. In total, the BPB model phenotype is impacted by these drugs, in a dose-dependent and drug-dependent manner. 88 Fig. 5.3. Barrier Dose Response and Direct Drug Comparison. (A/B) ICAM secretions for the coculture model (BeWo b30 and HUVEC) at different dosages of (A) fluoxetine or (B) sertraline after 1 and 3 days of exposure. (n=3). (C-E) Coculture model secretions for (C) ICAM, (D) TGFβ, or (E) VCAM when exposed to no drug or fluoxetine or sertraline, at a drug dosage of 10,000 ng/mL following 1 day of treatment. (n=3). (F) TEER values over 9 days following initial exposure (day 7 post- seeding cells) to fluoxetine or sertraline at 10,000 ng/mL with fresh exposure daily. (n=3). Data reported as mean ± standard deviation. Groups that do not share a letter are significantly (*p<0.05) different. 89 CAM Secretions of ECs, not Trophoblast, are SSRI-Influenced We evaluated drug and dosage effects for each cell type to assess which BPB cell type, either trophoblast or ECs, is more impacted in cell adhesion due to exposure to these drugs. We found that BeWo b30 cells did not dramatically change in their ICAM secretions in response to changing length of exposure or dosage for either fluoxetine (Fig. 5.4A) or sertraline (Fig. 5.4B), although sertraline did lead to slightly increased ICAM secretions. By comparison, extended exposure to fluoxetine led to HUVECs secreting more ICAM (day 3 vs. day 1), with the highest dosage (10,000 ng/mL) leading to the most dramatic change (Fig. 5.4C). This temporal trend carried through for HUVECs exposed to sertraline, though, interestingly, a middle dosage of 100 ng/mL led to the highest ICAM secretions (Fig. 5.4D). Additionally, when directly comparing each cell type and each drug, it becomes clear that HUVECs secrete dramatically more ICAM in response to either drug than BeWo b30 cells do (Fig. 5.4A-D), suggesting ECs, a supporting cell type within the BPB, are impacted to a much greater extent than trophoblast cells, the primary regulator of transport at the BPB. VCAM secretions by HUVECs are higher after 3 days of exposure instead of 1 day of exposure to fluoxetine, though this effect is reversed at the highest fluoxetine concentration (Fig. 5.4E). By comparison, sertraline concentration has modest impact on VCAM secretions, though the highest concentration similarly seems to reduce these secretions by HUVECs (Fig. 5.4F). In sum, it appears that both drugs elicit a greater effect on HUVECs than BeWo b30 cells when evaluating CAM secretions, suggesting endothelial cell adhesion is more impacted by these drugs than trophoblast cell adhesion. 90 Fig. 5.4. Secretions of CAM molecules by BPB cell type. (A/B) ICAM secreted by BeWo b30 cells in response to varying dosages of (A) fluoxetine or (B) sertraline after 1 and 3 days of repeated exposure. (n=3). (C/D) ICAM secreted by HUVECs in response to varying dosages of (C) fluoxetine or (D) sertraline after 1 and 3 days of repeated exposure. (n=3). (E/F) VCAM secreted by HUVECs in response to varying dosages of (E) fluoxetine or (F) sertraline after 1 and 3 days of repeated exposure. (n=3). Groups not sharing the same letter within a single timepoint are significantly (p<0.05) different and * indicates p<0.05 across two time points for the same dosage group. Data reported as mean ± standard deviation. 91 Sertraline, not Fluoxetine, Impacts TGFβ Secretions by BPB Cells In addition to CAM secretions, we evaluated TGFβ secretions in response to each drug for each cell type within the model. We observed minimal differences in response to fluoxetine dosage or length of exposure for BeWo b30 cells (Fig. 5.5A), though sertraline exposure led to slightly higher TGFβ secretions (Fig. 5.5B), compared to fluoxetine. Similarly, HUVEC secreted TGFβ in response to fluoxetine, with a slight decline at the highest dosage over 3 days of exposure (Fig. 5.5C), but sertraline induced even greater secretions of TGFβ (Fig. 5.5D), compared to fluoxetine. Overall, this suggests that sertraline has a more prominent impact on BPB cells, particularly ECs within the barrier, compared to fluoxetine. Fig. 5.5. Secretion of TGFβ by BPB cell type. (A/B) TGFβ secreted by BeWo b30 cells in response to varying dosages of (A) fluoxetine or (B) sertraline after 1 and 3 days of repeated exposure. (n=3). (C/D) TGFβ secreted by HUVECs in response to varying dosages of (C) fluoxetine or (D) sertraline after 1 and 3 days of repeated exposure. (n=3). Groups not sharing the same letter within a single timepoint are significantly (p<0.05) different and * indicates p<0.05 across two time points for the same dosage group. Data reported as mean ± standard deviation. 92 5.4 Discussion Numerous studies have shown SSRIs cross the BPB, yet few studies have evaluated the time-dependent concentration profiles of these drugs across the barrier7,256,257. Furthermore, even fewer studies, if any, have investigated how these medications impact the placenta and BPB, the key regulator limiting exposure of toxins to the fetus. Therefore, the overall goal of this study was to better understand the time-dependent concentration profiles of SSRIs crossing the BPB and how these drugs influence BPB phenotype. We chose to compare fluoxetine and sertraline given their widespread usage during pregnancy5. We hypothesized that both of these drugs would readily diffuse across the barrier, but that the barrier phenotype would be impacted dependent upon the specific drug used, the drug concentration, and the length of drug exposure. We evaluated endothelial cell type and placental ECM as others have shown these factors may influence overall phenotype. Groups have shown that microvascular cells, such as HBMEC, differ from macrovascular cells, such as HUVEC, in their response to angiogenic factors and how ECs take up substrates258,259. Comparing EC type, we observed the largest difference in small molecule permeability, though even this difference was relatively modest, where permeability was still within the same order of magnitude. In evaluating placental ECM, we were able to decellularize the tissue to an extent similar to other studies, with low DNA content and similar protein composition180,260,261. Given that protein composition was comparable to reports in literature, we chose to proceed utilizing a non-pooled source of placental ECM. We further observed modest differences in 93 barrier formation, which is comparable to previous findings that ECM proteins have minimal impact on permeability characteristics, albeit studied for ECs262. Interestingly, we observed that addition of ECM influence TGFβ secretions of trophoblast, but not ECs. Previous studies have established an interplay between TGFβ secretions and extracellular matrix components, where ECM components may inactivate TGFβ263–265. In this case, it is surprising that trophoblast have significantly reduced TGFβ secretions (p<0.05) while ECs do not, suggesting that, though TGFβ is secreted by numerous cell types throughout the body, the extent of this interplay is specific to each cell type. Given that SSRIs readily cross the BPB, we sought to better understand the time-dependent concentration profiles of these drugs in transplacental transport, as well as evaluate whether P-gp/BCRP influence these concentration profile. Previously studies have generally shown that, while SSRIs cross the BPB, the amount of drug (or metabolite) that ends up on the fetal side is a fraction of the maternal side concentration7,228. Further looking at sertraline and fluoxetine exposure, it becomes apparently that, generally, fluoxetine leads to a higher relative fetal concentration, compared to maternal concentration, than does sertraline, suggesting a drug- dependent effect. From our results, we observed three key transport trends for both drugs: (1) the BPB model uptakes drug from the maternal compartment more quickly than it releases that drug into the fetal compartment, (2) over extended exposure (i.e. 3 days vs. 1 day), there is accumulation of drug in the fetal side suggesting that BPB cells do not degrade the drug as it passes through, and (3) use of elacridar to inhibit P- gp and BCRP leads to a more rapid maternal-side concentration decrease, suggesting 94 that both of these drugs are removed from the BPB by these drug efflux pumps. We did observe a differential in the partition coefficient between the drugs (i.e. the ratio of drug concentration in the gel vs. solution, as described in section 3.2), where fluoxetine has a much higher partition coefficient than sertraline. This suggests that fluoxetine interacts with the hydrogel more than sertraline, and that it may stay in the BPB model at a higher relative concentration than sertraline. Comparing our results to others, we agree that SSRIs do cross the BPB and that P-gp and BCRP are involved in limiting fetal exposure. However, we propose that the reason the fetal concentration is a fraction of the maternal concentration is due to the BPB cells uptaking the drug and slowly releasing it. Further studies are needed to assess this suggestion as to whether placental cells uptake drug and maintain it intracellularly for extended periods of time. Additionally, it would be prudent to perform these studies utilizing another biomaterial to ensure that this uptake is not in response to the biomaterial used but rather is due to the cells within the barrier. More mechanistically, we observed that both of these drugs lead to changes in CAM secretions. Given that CAMs are intercellular molecules, secretions of these proteins into solution indicates delocalization of the CAMs from cell-cell junctions to the surrounding solution and thus damage to cell-cell junctions246,266. A study by Lopez- Vilchez, et al., showed that, over time, SSRIs reduced both cell surface ICAM expression and soluble VCAM expression in depressed patients back towards a healthy control, where both of these markers were significantly elevated over the healthy control with no treatment246. We found that both fluoxetine and sertraline exhibited a dose-dependent effect in causing ICAM secretions, though the two drugs 95 were not significantly different in their extent to impact either ICAM or VCAM secretions from the whole BPB model. When we assessed each cell type, we found that these drugs primarily cause CAM secretions through ECs. By comparison, trophoblast CAM secretions were almost an order of magnitude lower than the ECs, suggesting that SSRIs have a minimal, if any, impact on these secretions. Moreover, to our knowledge, SSRI impacts on trophoblast secretions were not previously studied, although Clabault, et al., found that SSRIs, particularly sertraline, appear to impact trophoblast syncytialization in a dose-dependent manner240, suggesting that SSRIs may impact the trophoblast layer and its barrier properties, though likely not through CAMs. Similar to the study by Clabault, et al., sertraline seemed to illicit a more dramatic response than fluoxetine, further suggesting that sertraline may be more detrimental to the BPB than fluoxetine. Interestingly, we observed fluoxetine led to increased ICAM levels with a simultaneous decrease in VCAM levels over time at higher dosages, suggesting some specificity to these individual CAMs. Previous studies have suggested a role for each of these CAMs in atherosclerosis, with some hints as to a time-dependent effect for each molecule267,268. Further, studies have suggested a protective role of SSRIs in atherosclerosis, primarily thought to occur from SSRIs inducing production of pro-inflammatory cytokines234,269. Based on our results, we suggest that CAM modulation may play a partial role in SSRI-induced protection against atherosclerosis, though further experimentation in vivo would be needed to test this hypothesis. Given the expected changes within the barrier function, we sought to evaluate changes in angiogenesis since the BPB is essentially a network of capillary trees 96 interfacing with maternal blood. We evaluated TGFβ because of its role in angiogenesis and in major depressive disorder, an instance where SSRIs would be prescribed as the course of treatment249,270,271. When evaluating the whole model, we observed that sertraline, not fluoxetine, significantly reduced TGFβ expression (p<0.05), suggesting some level of drug specificity. This supports the hypothesis that there are drug-specific effects, where each SSRI elicits a unique change in cell physiology, which has been previously shown for placental cells236,240,241. For example, Clabault, et al, showed that sertraline, but not fluoxetine or its metabolite norfluoxetine, increased syncytialization of primary trophoblast (i.e. fusion of cells towards the syncytiotrophoblast) and reduced human chorionic gonadotropin (hCG) secretions, a key pregnancy hormone, in BeWo cells, a choriocarcinoma cell line240. Similarly, Kapoor, et al, observed that sertraline, not fluoxetine, impacts P-gp function in brain ECs236. When we evaluated drug-induced TGFβ expression in each cell population, we observed that sertraline elicited greater secretions than fluoxetine for both cell types, but that ECs had higher secretions than trophoblast cells. Interestingly, there was no significant dose-response to either drug for trophoblast cells, while ECs appeared to have a dose-response, where low drug doses caused higher than baseline TGFβ secretions while higher dosages reduced these secretions to a baseline level over time. This time and dose-dependent response is intriguing as others have previously shown that SSRI treatment changes cell physiology, both in vivo in depressed patients and in vitro following exposure to serum from depressed patients, where SSRI treatment leads to a return to baseline levels over time246. Given that TGFβ is important for survival, homeostasis and barrier function of ECs, as well 97 as angiogenesis, it’s interesting that these drugs can elicit a time and dose dependent effect270,272,273. Together, this suggests that depression and SSRIs may induce changes in TGFβ signaling, leading to a change in placental angiogenesis and barrier formation. This is further supported by a previous study showing TGFβ1 downregulates VE-cadherin, again suggesting that SSRI-induced changes in TGFβ signaling may lead to a change, potentially even a partial compromise, of the BPB274. The work presented herein has allowed us to better understand the concentration profiles of these drugs crossing the BPB and learn more about drug- specific and dose-dependent effects on the placental barrier utilizing a biomimetic BPB model, where we could not do so using in vivo approaches. Although this work adds to our knowledge of SSRI-induced effects on the BPB, it is not without limitations. One of the limitations is that this is an in vitro model, thus some of the specific effects of these drugs in this context in vivo may be different, though there are obviously considerable ethical limitations to performing those studies given the lack of suitable animal models of the human placenta16. However, given that others have observed similarities between in vitro and in vivo studies, we anticipate that the trends would still be similar246. Additionally, we utilized trophoblast cell lines in co- culture with primary umbilical vein ECs. As other studies have shown, there are differences based upon the trophoblast cell line used or if primary trophoblast are used, and thus some of the effects may not be observed when using cell lines240,275. Additional further studies utilizing primary trophoblast would help to alleviate these concerns and elucidate the observed effects. 98 5.5 Summary In summary, we utilized a biomimetic BPB model to assess the concentration profiles of SSRIs transported across the BPB and assay what effects these drugs, fluoxetine and sertraline, have on BPB cells. We observed a loading-accumulation- release effect of the drugs within the BPB, suggesting a potential build-up of bioactive SSRI on the fetal side of the barrier over time. Further, we observed changes in CAM and TGFβ secretions, where ECs are more impacted by SSRIs than trophoblast cells are. This is one of a few studies to evaluate effects of SSRIs on the placenta, an area that should be further studied with increasing use of antidepressants during pregnancy to ultimately ensure the safety and health of both pregnant women and their future babies. 99 Chapter 6: Assessing the Impact of Fluoxetine and Sertraline on Cardiomyocytes Downstream of Placenta Using Placental Barrier Model 6.1 Introduction In the 1960s, there was a sharp increase in the number of children born with severe limb malformations and it took a few years before this was discovered to be resulting from a prescription medication, thalidomide1. Historically, there has been a lack of a priori knowledge about whether medications, such as thalidomide, would lead to issues with fetal development as a result of the drug’s off-target effects. With greater than 90% of pregnant women taking at least one medication during pregnancy5,226, the lack of knowledge regarding medication safety during pregnancy, particularly as it relates to fetal development, raises concerns. Further complicating this are medications prescribed for clinical conditions occurring simultaneously during pregnancy, such as depression. The risk of depression is estimated at up to 20%, where pharmacologic treatment with selective serotonin reuptake inhibitors (SSRIs) is the primary treatment276. Fetal exposure rates to SSRIs are estimated at 6- 8%, and continuing to rise6,251. Moreover, there is widespread acceptance that SSRIs readily cross the maternal-fetal interface, termed the blood-placenta barrier (BPB)7,228, a barrier intended to limit fetal exposure to toxins prenatally. As the SSRIs cross into the fetal compartment, there is the possibility that it influences fetal development, as has been previously shown233. With the rates of usage and fetal 100 exposure to SSRIs on the rise, and given that SSRIs can readily pass into the fetal compartment during development, additional studies are needed to better understand the potential effects resulting from these medications being taken during pregnancy. In particular, SSRIs have been linked to congenital heart disease (CHD), though there are arguments claiming SSRIs cause no deleterious effects to the developing fetus and others showing significant consequences and increased risks for the developing fetus14,232,233,244,276. The two most commonly prescribed SSRIs are sertraline and fluoxetine, both being prescribed in approximately 2% of pregnancies5,227. Previous studies and reviews have summarized the effects of both of these drugs, in the context of changes in cardiovascular physiology11–13. Both sertraline and fluoxetine have been linked to inhibition of ion channels (Na+, Ca2+, and K+) and depressant effects on cardiovascular cells12,13,277. In particular, both drugs have been observed to inhibit L-type Ca2+ current in cardiomyocytes, though to different extents277. In vivo, SSRIs have been suggested to lead to QT prolongation and syncope (i.e. reduced blood flow to the brain)276,278, sertraline has been shown to lead to small left heart syndrome233, and fluoxetine has been linked to bradycardia11 (i.e. a slower than normal heart rate), suggesting SSRIs can lead to abnormal heart rhythms. Since QT interval, action potential duration, and calcium transients (CaT) all generally correlate279,280, though notably they do not directly compare to each other, it is perhaps intuitive that these in vitro and in vivo results agree. Interestingly, though these studies are all suggestive of potential congenital heart issues related to signaling, few studies, if any, to our knowledge, have evaluated whether these drugs lead to changes in heart-specific biomarkers for cardiac injury281. Specifically, 101 creatine kinase-MB (CKMB) and troponin T have been suggested as biomarkers of cardiovascular defects within newborns and children, based upon their usage as biomarkers for adults281–283. Similarly, N-terminal pro-B-type natriuretic peptide (NT- proBNP) has been suggested as a biomarker of cardiac load and heart failure, and in some instances, may be an alternative when other biomarkers show no changes in expression levels281,284. Another marker critically important in development is vascular cell adhesion molecule (VCAM), and though not specific to cardiomyocytes, it has previously shown to be impacted by SSRIs246,285,286. Thus, understanding whether these biomarkers are related to cardiovascular signaling changes may help to provide predictive measures to improve clinical care of women taking SSRIs during pregnancy. Moreover, understanding the extent to which SSRIs influence cardiovascular development is critical to improve maternal and fetal safety of these drugs during pregnancy. The gold standard approach in understanding and alleviating these concerns regarding fetal development is the use of animal models. However, though there are animal models of developmental cardiovascular disease and defects, the vast majority of these models are for evaluation of gene expression lethality287,288, and do not adequately reflect drug-induced changes resulting from pregnancy. Moreover, animal models do not recapitulate the key features of the human placenta, raising questions as to their relevance to human physiology16. Thus, a critical need for improving the safety and efficacy of SSRIs during pregnancy, to ultimately maximize maternal benefit and minimize fetal harm, is the development and utilization of biomimetic in vitro models of the human placenta and developing fetus. 102 In chapter 4, we reported using a biomimetic placenta-fetus model system to study maternal-fetal transmission of Zika Virus, and to assess fetal neural toxicity, observing that our findings were in agreement with in vivo studies203,243. The model system mimics the complexity of the human BPB and allows for investigating the effects of drugs on the placenta, as done in chapter 5, and downstream fetal tissue. Herein, we have utilized this model with the goal of assaying: (1) whether the SSRIs fluoxetine and sertraline lead to changing in cardiomyocyte calcium transients, through direct exposure and indirect exposure following passage through the BPB; and (2) do the biomarkers CKMB, troponin T, NT-proBNP, and VCAM show differences in response to drug exposure that correlates with changes in calcium handling by the cardiomyocytes. Additionally, given that fluoxetine and sertraline are in the same class of drugs (as SSRIs), but that unique phenomena have been observed from each drug, we wanted to assess whether the two drugs lead to the same, or unique, effects in cardiomyocytes. We hypothesized that both drugs would lead to similar elongations in calcium transients, and that all biomarkers would trend in a similar fashion towards higher levels being secreted with increasing drug dosage. Herein, we report on: (1) the calcium transients for cardiomyocytes exposed to each drug at three dosages, (2) the biomarkers secreted by cardiomyocytes in response to drug exposure, and (3) the extent to which the BPB changes the response by cardiomyocytes. In sum, we suggest that this work improves our understanding these drugs’ impact on cardiomyocytes and the BPB’s influence in altering the cardiomyocyte response to these drugs. 103 6.2 Materials and Methods Cell culture BeWo cells (trophoblast) were purchased from American Type Culture Collection (Manassas, VA) and grown in F12-K medium (ATCC) supplemented with fetal bovine serum (FBS, Gibco, Gaithersburg, MD) and antibiotic-antimycotic (A/A, Gibco). Human umbilical vein endothelial cells (HUVECs, or endothelial cells) were purchased from Lonza and grown in Endothelial Cell Growth Medium (Promocell, Heidelberg, Germany). Induced pluripotent stem cell derived-cardiomyocytes (iPSC- derived cardiomyocytes) were purchased from Fujifilm Cellular Dynamics (Madison, WI) and grown in plating and maintenance media, as per manufacturer’s protocols. For co-culture studies involving trophoblast and endothelial cells, HUVEC media was utilized. For indirect co-culture studies involving BPB models, fabricated as described below, and iPSC-derived cardiomyocytes, media appropriate for each cell type was used within the apical and basolateral compartments of the co-culture setup. BPB Model Fabrication BPB models were fabricated using gelatin methacrylate (GelMA, 10% w/v), decellularized placental extracellular matrix (pECM, 10% v/v), trophoblast cells, and endothelial cells, using a similar approach as previously described in literature243 and in chapter 5. Placental decellularization was also performed using previously established protocols as described in literature180 and in chapter 5. In brief, frozen whole placental tissue purchased from Zen-Bio (Research Triangle Park, NC) was thawed 104 and mechanically digested via surgical tools. Next, tissue was washed with antibiotic overnight and decellularized via sodium dodecyl sulfate (SDS, 0.1%) at room temperature over 48 hours. The decellularized tissue was repeatedly washed over 2 days, then frozen, and lyophilized. Following lyophilization, the tissue was digested via pepsin (10% w/w in 0.5M acetic acid) at 4°C for 48 hours. This solution was pH- adjusted to 7.4, centrifuged to pellet the tissue, and supernatant run though cell strainer (40µm) and sterile filter (0.2µm), where the final product was the pECM added into BPB models. BPB Model-Cardiomyocyte Exposure Assays Cardiomyocytes were exposed to fluoxetine and sertraline through one of two methods: (1) directly, where drug was included in maintenance media given to cardiomyocytes at the start of drug exposure; or, (2) indirectly, where drug was added to the apical compartment of a co-culture setup (see Fig. 6.2) where the BPB model, in a transwell insert, separated the apical and basolateral compartments and cardiomyocytes were seeded in the basolateral compartment. For all assays, glass bottom dishes (MatTek, Ashland, MA) were coated with fibronectin (Gibco) prior to seeding cardiomyocytes in plating media (Fujifilm Cellular Dynamics). After 4 hours, plating media was replaced with maintenance media, which was changed every 2 days thereafter. Cardiomyocytes were grown for 7 days prior to inclusion in indirect exposure co-culture studies involving the BPB model. For exposure studies, fluoxetine hydrochloride or sertraline hydrochloride (Sigma-Aldrich) was solubilized in dimethyl sulfoxide (DMSO) and added to either cardiomyocyte maintenance media 105 or HUVEC media, for direct and indirect exposure studies, respectively, at the appropriate dosage (10, 100, or 1,000 ng/mL, final concentration). These dosages were chosen based upon previously reported values for SSRIs found in the umbilical vein7. A control with DMSO but without drug was used as a no exposure control. Further, DMSO content was controlled across all groups, with all media having a concentration <1% DMSO. In all instances, the drug-containing media was given to cells on day 7 post-seeding. For indirect exposure studies, drug-containing media was used in the apical compartment while cardiomyocyte maintenance media was used in the basolateral compartment. Calcium Assays and Quantification Following drug-exposure, Fluo-4 calcium dye (Invitrogen, Carlsbad, CA) was given to cells in maintenance media and incubated for 1 hour at 37°C. Dye was removed and replaced with maintenance media prior to live-cell imaging via confocal microscopy (Olympus FV3000) with incubation chamber. Resonant scanning was performed, imaging samples for 1500 frames at a rate of 15 frames per second (fps). For each sample, calcium intensity was quantified using FIJI image analysis software289. From this intensity, calculations were performed in Excel to determine: oscillation period, oscillatory frequency, time to peak, decay time, and calcium transients at 30% and 80% reuptake (CaT30 and CaT80, respectively). 106 Cardiomyocyte Secretions and ELISAs Enzyme-linked immunosorbent assays (ELISAs) were used to determine secretions of cardiac biomarkers and intercellular adhesion markers. Assay kits for Creatine Kinase MB and Troponin T were purchased from Ray Biotech (Norcross, GA). DuoSet assay kits for N-terminal pro b-type natriuretic peptide (NT-proBNP) and vascular cell adhesion molecule-1 (VCAM) from R&D Systems (Minneapolis, MN). All kits were used as per manufacturer’s protocols. At each time point, cell culture supernatant was collected, spun at 10,000g for 10 min at 4°C, then frozen at - 80°C. Samples were thawed at room temperature prior to running each assay. Absorbance was measured via plate reader. Statistical analysis Results are reported as a mean ± standard deviation, calculated using Excel. Significance was determined using a two-sample t-test via Minitab, where p<0.05 was considered significant and is shown using an asterisk (*). 107 6.3 Results SSRI-Induced Calcium Signaling Changes in Cardiomyocytes We initially evaluated changes in calcium signaling when cardiomyocytes were exposed to SSRIs (Fig. 6.1), given its correlation with action duration potential and QT interval in the myocardium279,280. In quantifying these results, we found that sertraline at higher dosages (100 and 1000 ng/mL) led to significantly different (p<0.05) changes in the period and frequency of the calcium transient, but that lower sertraline dosages (10 ng/mL) and all doses of fluoxetine did not elicit this effect (Fig. 6.1A/B). To further characterize this, the time to peak (Fig. 6.1C) and decay time (Fig. 6.1D) was characterize, where these phases correspond with release and reuptake of calcium, respectively, into the sarcoplasmic reticulum of cardiomyocytes. Increasing dosages of sertraline led to modest but significant (p<0.05) increases in time to peak (Fig. 6.1C), suggesting a dose-dependent effect in time for calcium release. We also observed that a low fluoxetine dosage (10ng/mL) resulted in a decrease time to peak, suggesting faster calcium release into the cell (Fig. 6.1C). In evaluating decay time, indicative of the reuptake of calcium, we found no significant differences among groups, though there appears to be a trend that increasing sertraline dosage leads to increased decay time (Fig. 6.1D). We also evaluated duration time for the calcium transient at 30% reuptake and 80% reuptake280,290. We observed a significant (p<0.05) increase in the CaT30 duration time for sertraline (1000 ng/mL) and a significant (p<0.05) decrease for fluoxetine (10 and 100 ng/mL) (Fig. 6.1E), suggesting opposite effects from each drug. Interestingly, we observed significant (p<0.05) increases in CaT80 duration time for both sertraline (100 and 108 1000 ng/mL) and fluoxetine (10ng/mL) (Fig. 6.1F). Overall, these data suggest sertraline prolongs the cardiomyocyte calcium signaling cycle in a dose-dependent manner, whereas fluoxetine impacts this cycle only at low (10 ng/mL) dosages. Fig. 6.1. Cardiomyocyte calcium handling response to SSRI exposure. Calcium intensity for cardiomyocytes was quantified following direct exposure to SSRIs (either fluoxetine or sertraline) using multiple metrics: (A) the peak-to-peak time (i.e. the period of the oscillations), (B) the frequency of oscillations, (C) the time to reach the peak, (D) the decay time, from peak to a minimum value, (E) the duration time at 30% calcium reuptake, and (F) the duration time at 80% calcium reuptake. *p<0.05 for the two groups indicated. (n=3 for all groups/studies). Data reported as mean ± standard deviation. 109 BPB Modifies SSRI-Induced Cardiomyocyte Calcium Signaling Changes Following understanding the extent to which each drug directly impacts cardiomyocyte calcium handling, we sought to evaluate whether these changes would occur when the cardiomyocytes were only exposed to drug after it had passed through the placental barrier (Fig. 6.2). As before, we evaluated the period (Fig. 6.2A) and frequency (Fig. 6.2B) of calcium oscillations, finding significant (p<0.05) differences between sertraline dosages (10 and 1000 ng/mL) and fluoxetine dosages (100 and 1000 ng/mL) (Fig. 6.2A/B), though only fluoxetine at a dosage of 100 ng/mL led a significant (p<0.05) difference from the control group. Notably, the period for all groups increased (Figs. 6.2A/1A) when the BPB model was present, compared to without the BPB model, along with a reduction in frequency (Figs. 6.2B/1B). When evaluating time to peak, there were differences between dosage of sertraline (10 and 100 ng/mL) and fluoxetine (100 and 1000 ng/mL), but only fluoxetine at 100 ng/mL led to a significant (p<0.05) decrease (Fig. 6.2C). In evaluating decay time, we found differences between sertraline dosages (10 and 1000 ng/mL) and all fluoxetine doses, though surprisingly, the middle dose of 100 ng/mL led to decrease in decay time, whereas the low (10 ng/mL) and high (1000 ng/mL) doses led to no change, and perhaps even an increase (Fig. 6.2D). When evaluating the CaT30 duration time, we also observed a significant (p<0.05) decrease from fluoxetine at 100 ng/mL but also a significant (p<0.05) increase from fluoxetine at 1000 ng/mL (Fig. 6.2E). When evaluating the CaT80 duration time, we found differences between sertraline dosages (10 and 1000 ng/mL), differences between all three fluoxetine dosages, and a significant (p<0.05) decrease from exposure to fluoxetine at 100 ng/mL (Fig. 6.2F). 110 Overall, these data suggest that the BPB model leads to an increase in the period of calcium oscillations for cardiomyocytes, that fluoxetine dosage differentially impacts these oscillations in a non-dose-dependent manner, and that these changes primarily occur in the reuptake phase of calcium and likely in the late-phase calcium reuptake. Fig. 6.2. Cardiomyocyte calcium handling response following SSRI exposure through BPB model. Calcium intensity for cardiomyocytes in the basolateral compartment was quantified following exposure to SSRIs (either fluoxetine or sertraline) through a BPB model, using these metrics: (A) the peak-to-peak time (i.e. the period of the oscillations), (B) the frequency of oscillations, (C) the time to reach the peak, (D) the decay time, from peak to a minimum value, (E) the duration time at 30% calcium reuptake, and (F) the duration time at 80% calcium reuptake. *p<0.05 for the two groups indicated. (n=3 for all groups/studies). Data reported as mean ± standard deviation. 111 Myocardial Injury Biomarkers Show Drug-Dependent Effect We sought to further characterize the extent of cardiomyocyte damage resulting from SSRI exposure by determining secretions levels of known markers of myocardial injury, looking specifically at Creatine Kinase MB and Troponin T281 (Fig. 6.3). We observed significant (p<0.05) differences in creatine kinase MB secretions, where only direct exposure to fluoxetine at a dosage of 100 ng/mL led to changes (Fig. 6.3A). This differed significantly (p<0.05) from fluoxetine at a higher dose (1000 ng/mL) and sertraline at the same dose, whereas none of the other groups did (Fig. 6.3A). When evaluating secretions in the basolateral compartment following exposure through the placental barrier, we found no significant differences, though a trend was observed of low dosages leading to higher secretions (Fig. 6.3B). These trends carried over when evaluating troponin T secretions (Figs. 6.3C/D), a protein key to cardiomyocyte function and only secreted due to cardiomyocyte damage. Following direct exposure, almost all of the groups fell below the detection limit of the assay kit (0.35 ng/mL) suggesting no deleterious effects to the cardiomyocytes (Fig. 6.3C). However, these levels increased following exposure through the BPB model, though did not differ significantly between groups (Fig. 6.3D). This data suggest that fluoxetine (100 ng/mL) may cause some myocardial damage, but that these drugs do not lead to deleterious effects within the developing fetal heart. 112 Fig. 6.3. Cardiomyocyte Injury Marker Secretions in Response to Drug Exposure. (A/B) Creatine Kinase MB secretions for cardiomyocytes either (A) directly exposed to the drugs (fluoxetine, F, or sertraline, S) at one of two dosages (100 or 1000 ng/mL), with a non-exposed (NoE) negative control, or (B) indirectly exposed to the drugs, via additional to apical compartment of placental barrier and secretions in basolateral compartment assessed. (C/D) Troponin T secretions for cardiomyocytes either (C) directly exposed to the drug, or (D) indirectly exposed to the drug, with annotation as stated above. *p<0.05 for the two groups indicated. (n=3 for all groups/studies). Data reported as mean ± standard deviation. 113 We also evaluated NT-proBNP as a cardiac biomarker given that it is not co- expressed with CKMB and Troponin T, and provides a wider view of cardiac injury (Fig. 6.4). Counter to CKMB and Troponin T, we found that direct exposure both sertraline and fluoxetine lead to dose-dependent decreases in NT-proBNP secretions from cardiomyocytes (Fig. 6.4A). Interestingly, when cardiomyocytes were exposed to drug following passage through the BPB model, we observed a trend of increasing sertraline dosage leading to increased NT-proBNP secretions in the basolateral compartment (Fig. 6.4B). However, we found fluoxetine at a dosage of 100 ng/mL led to a significant (p<0.05) decrease in NT-proBNP secretions whereas other dosages of fluoxetine (10 and 1000 ng/mL) did not (Fig. 6.4B). When we evaluated expression in the apical compartment, we found NT-proBNP secretions do not differ significantly across drug or dosage groups (Fig. 6.4C). This suggests that NT-proBNP secretions, though impacted by the drugs, do not cross from fetal into maternal compartment. Further, the BPB seems to influence whether increasing drug dosage leads to increased or decreased secretions. 114 Fig. 6.4. Cardiomyocyte secretions of NT-proBNP following exposure to SSRIs. (A) Secretions of NT-proBNP from cardiomyocytes following 1 day of exposure to SSRIs (either sertraline, S, or fluoxetine, F) at one of three dosages (10, 100, or 1000 ng/mL), or no drug exposure (None). (B) Secretions in the basolateral compartment, where cardiomyocytes were indirectly exposed to the drug following passage through a BPB model. (C) Secretions in the apical compartment, indicating protein that was secreted in the basolateral compartment and passed through the BPB model. *p<0.05 for the two groups indicated. (n=3 for all groups/studies). Data reported as mean ± standard deviation. Myocyte Cell Adhesion Impacted by SSRIs Previous work has shown that SSRIs modulate vascular cell CAM expression246,285,286. Given that VCAM, in particular, is important for cardiomyocyte function291, we sought to assess whether iPSC-derived cardiomyocytes are similarly impacted by SSRIs. We found that cardiomyocytes directly exposed to sertraline show no difference in VCAM secretions, but that exposure to low dosages of fluoxetine (10 and 100 ng/mL) significantly (p<0.05) increased VCAM secretions (Fig. 6.5A). When cardiomyocytes were indirectly exposed to sertraline and fluoxetine through a BPB model, and VCAM secretions in the basolateral compartment quantified, we again observed no differences amongst sertraline, compared to the control and across dosages, but that fluoxetine at a high dose (1000 ng/mL) led to a reduction in VCAM secretions (Fig. 6.5B). Important to note, these protein levels may be reflective of the cardiomyocytes in the basolateral compartment 115 or the endothelial cells in the apical compartment. When we assessed secretions in the apical compartment though, we found all are dramatically reduced compared to the basolateral compartment (Fig. 6.5C), with the only significant difference (p<0.05) being a reduction from exposure to fluoxetine at 10 ng/mL. Overall, this suggestions that sertraline has minimal impact on cardiomyocyte VCAM secretions, but that fluoxetine may hurt cardiomyocytes upon direct exposure and provide some type of protective effect when indirectly exposed via BPB, though additional studies are needed to clarify this effect. Fig. 6.5. Secretions of VCAM-1 following exposure to SSRIs. (A) Secretions of VCAM-1 from cardiomyocytes following 1 day of exposure to SSRIs (either sertraline, S, or fluoxetine, F) at one of three dosages (10, 100, or 1000 ng/mL), or no drug exposure (None). (B) Secretions in the basolateral compartment, where cardiomyocytes were indirectly exposed to the drug following passage through a BPB model. (C) Secretions in the apical compartment, indicating protein that was secreted in the basolateral compartment and passed through the BPB model or protein secreted by the cells within the BPB model in response to drug exposure. *p<0.05 for the two groups indicated. (n=3 for all groups/studies). Data reported as mean ± standard deviation. 116 6.4 Discussion The impact of SSRIs on fetal cardiac development ranges from no negative effects to detrimental development effects, including congenital heart issues, with no definitive consensus in the field15,233,292. Therefore, the overall goal of this study was to better understand whether SSRIs hinder cardiomyocytes, and to assess the extent to which the BPB impacts the observed response. Given that sertraline and fluoxetine has seen relative increases in their usage during pregnancy, and sertraline in particular has been associated with cardiac issues, we investigated whether there were indeed drug-specific differences5,13,233. We hypothesized that both drugs would lead to inhibited calcium signaling and increased secretion of damage biomarkers, but that sertraline would be significantly worse than fluoxetine in all instances. We evaluated changes in the calcium transient as a metric for changes to the action duration potential and QT interval, as these generally, though not directly, correlate279,280. Though these metrics have been previously studied in response to fluoxetine and sertraline12,13,277, those studies utilized rat cardiomyocytes and we are unaware of any studies utilizing human iPSC-derived cardiomyocytes to investigate these effects stemming from SSRIs. We observed that the iPSC-derived cardiomyocytes, in response to sertraline but not fluoxetine, had a dose-dependent increase in period and time to peak, a corresponding decrease in frequency, but no significant changes in decay time (i.e. p>0.05 for all comparisons). Interestingly though, though decay time did not significantly increase, the CaT80 duration time increased, and nearly doubled with a high dose (1000 ng/mL) of sertraline. Together, this suggests that the early-phase calcium reuptake is modestly impacted, but that the 117 late-phase calcium reuptake may actually be more impacted than the early-phase. The sodium-calcium exchanger (NCX) has been implicated as being more prominent in late-phase reuptake293,294, though additional studies are necessary to show these drugs inhibit the NCX and that is why the change occurs. Other studies have evaluated how these drugs impact cardiomyocytes and tried to pinpoint the mechanism at play for the observed effects. Park, et al., found that fluoxetine prolonged the action potential duration at 50% and that both fluoxetine and sertraline inhibited L-type Ca2+ current277, the latter of which is a mechanism for involved in cardiomyocyte contractility and electrical activity, and has been shown to be important in preventing cardiac arrhthmias295,296. This suggests that both drugs lead to elongation of the calcium transient, whereas we observed that sertraline and only low doses (10 ng/mL) of fluoxetine leads to elongation. By comparison, fluoxetine at higher doses (100 and 1000 ng/mL) seem to have no effect. In another study by Lee, et al., it was observed that sertraline inhibited sodium, potassium, and calcium ion channel currents, at concentrations comparable to the study by Park, et al, both of which were between 100 and 1000 ng/mL13,277. As before, this suggests sertraline at a high dosage (1000 ng/mL) should negatively impact the calcium transient, which we did observe when directly exposing the cardiomyocytes to drug. Beyond changes in calcium handling, we evaluated whether SSRI exposure led to an increase in cardiac biomarker secretions. These biomarkers would ultimately provide a comparatively easy metric by which effects could be measured perinatally, and possibly prenatally as well. We observed that fluoxetine, not sertraline, led to changes in CKMB, with a lower dosage (100 ng/mL) showing an increase. 118 Interestingly, though there were no significant differences between drugs and dosage when evaluating these biomarkers with drug exposure occurring following passage through the BPB, we found there appeared to be a modest trend that lower dosages of either drug led to increased secretions of both CKMB and troponin T. Trends were somewhat similar when we evaluated VCAM expression in response to cardiomyocyte direct drug exposure, where sertraline had no effect while fluoxetine led to higher secretions at lower dosages. Moreover, when exposed following passage through the BPB, sertraline slightly trended towards higher dosages leading to higher secretions, whereas fluoxetine had highest secretions at a moderate dosage (100 ng/mL). This suggests that fluoxetine’s effects on cells are highly dependent on the specific dosage of drug experience by the cells, where the drug may be able to both increase and decrease oscillations dependent upon drug binding to cell receptors, though additional studies are needed to further address this question. When evaluating NT-proBNP though, we found a trend opposite from the other biomarkers. Increasing dosages of sertraline led to significant (p<0.05) decreases in NT-proBNP secretions, while fluoxetine held a similar, though not significant, trend. Conversely, when cardiomyocytes were exposed to drug following passage through the BPB model, increasing drug dosage generally led to increased secretions. Moreover, fluoxetine- induced secretions were higher than those induced by sertraline. Regardless though, it appeared minimal amounts of this protein, if any, crossed the placental barrier back from fetal to maternal side. Interestingly, multiple studies have observed that NT- proBNP in maternal serum is higher when the pregnancy is complicated by disease, such as pre-eclampsia297,298. This suggests that NT-proBNP can be useful as a 119 biomarker of cardiac stress during pregnancy, but that maternal and fetal measurements may be necessary to distinguish between maternal cardiac concerns and fetal cardiac concerns. In addition, as NT-proBNP elicited trends running opposite and parallel to the trends observed by other biomarkers, it becomes apparent that multiple proteins should be screened for to assess the damage resulting from SSRI exposure. Though these trends are beneficial, whether they correlate to injury in vivo is something that still needs to be determined. Nearly a decade ago, Kocylowski, et al., sought to establish reference values for these markers, indicating cardiovascular injury, in postpartum cord blood281. They found that cardiac injury does indeed lead to non-normal levels of these biomarkers281, supporting their use as biomarkers of congenital heart issues. A key next step to improving in vitro studies would be to correlate the values of these biomarkers determined in vivo with those that can be studied in vitro, thereby dramatically improving the predictive value of in vitro systems. One aspect of the studies presented herein that was quite surprising was the role that the BPB played in cardiomyocyte response to these drugs. As previously mentioned, directly exposing cardiomyocytes to fluoxetine and sertraline at multiple dosages indicated changing in the calcium handling by these cells. Interestingly though, these effects were much more dramatic when using the placental barrier. As before, sertraline trended where increased drug dosage led to significantly increased (p<0.05) period, decay time, and CaT80 transient duration, and a corresponding decrease in frequency. Fluoxetine, by comparison, showed similar trends for low and high dosages (10 and 1000 ng/mL), but that a moderate dosage (100 ng/mL) led to 120 decreases in every measure, except for an increase in frequency corresponding to a decrease in period. Additionally, across comparable dosages, the two drugs did not seem to dramatically differ, suggesting that fluoxetine has as much of an effect as sertraline does. This suggests that the BPB plays a more prominent role than we anticipated in controlling drug exposure. Further, it is possible that the effects observed in the previous chapter, where these SSRIs impacted placental barrier cell phenotypes leading to secretions of bioactive molecules by the placental barrier cells, caused an additional effect on the cardiomyocytes, outside of those elicited solely by exposure to the drugs. Additional studies utilizing conditioned media are needed to investigate these questions further, as to whether the drugs or the drug-induced effects from other cells are more impactful towards the cardiomyocyte response. Although this work provides additional information regarding fluoxetine and sertraline’s impact on cardiomyocytes and the role that the BPB plays, it is not without limitations. As mentioned above, one interesting facet was the extent to which the BPB caused changes in the cardiomyocyte response. Potentially, this was due to the specific cell types used (trophoblast cell line, primary endothelial cells, and iPSC-derived cardiomyocytes). It begs the question of whether alternative cell types, particularly primary trophoblast, would lead to similar observations, given that these cell types generally do not hinder fetal development and other studies have shown different responses between primary trophoblast and cell lines240. Additionally, calcium handling studies do not directly indicate changes in electrical signaling and myocardial contraction280,299, thus future studies to investigate these results are necessary. Future studies should also investigate the proteins involved and the 121 mechanism behind the observed changes in calcium handling. Lastly, concentrations used and analysis performed assumed passage of drug through the BPB and subsequent contact directly with the fetal heart, though it’s possible that fetal endothelial tissue provides added resistance between fetal blood and the cardiovascular tissue. This fetal resistance was not considered in these studies, and future endeavors should assess whether this will impact the drug-tissue interactions. 6.5 Summary In summary, we assessed the extent to which iPSC-derived cardiomyocyte calcium transients and biomarker secretions are influenced by SSRIs, evaluating direct drug exposure and indirect exposure following passage through the BPB. We observed that sertraline had a larger influence on calcium transients than fluoxetine when the cells were directly exposed to both drugs, but that passage through the BPB seemed to moderate these effects and showed both drugs to be generally comparable. Further, we found that both drugs result in biomarker secretions, with a differential effect by drug. Fluoxetine appeared to influence CKMB, Troponin T, and VCAM secretions whereas sertraline was more influential in NT-proBNP secretions. Dose- dependent effects appeared modulated by the BPB, supporting the placenta’s role in limiting fetal effects. To our knowledge, this is one of the first studies evaluating these parameters using iPSC-derived cardiomyocytes, and evaluating the extent to which the placental barrier influences these effects. This will ultimately improve about ability to understand how SSRIs influence fetal development, and maybe enable improved clinical care through the knowledge gained. 122 Chapter 7: Summary, Contribution, and Future Directions 7.1 Summary The overall goal of this work was to develop a biomimetic placenta-fetus model, and utilize it to understand how drugs and substances influence the placental barrier and fetal cells downstream of the barrier. This was achieved through multiple steps, initially characterizing and validating the model as it was developed, then utilizing it to study viral transport, SSRI transport, SSRI-induced phenotypic changes in the BPB and each cell type within the BPB, and evaluating effects on neural and cardiovascular cells downstream of the placental barrier. The first objective of this work was to develop and validate the tissue- engineered BPB model through study of biologically relevant substances. We were able to show that we could create a 3D tissue-engineered construct that mimicked the multilayer architecture of the BPB. We characterized the intercellular junction protein expression and quantified barrier resistance (90-100 Ω∙cm2), showing progressive barrier formation. We further characterized the model, showing that trophoblast and endothelial cells within the model maintain bioactivity in co-culture and that small molecules, such as glucose, and larger proteins, such as IgG, show different time- dependent concentration profiles when diffusing across the placental barrier, where glucoses reaches equilibrium within 24 hours while IgG does not. Moreover, we showed cells are key to this barrier, characterizing glucose through an acellular gel (diffusion coefficient of 6.97 x 10-6 cm2/s) and through a cell-laden gel (diffusion coefficient of 2.28 x 10-6 cm2/s). We then studied Zika Virus, finding it infects both 123 cell types within the BPB model and it crosses the placental barrier, though this can be inhibited by chloroquine. We then showed that Zika Virus leads to neural cell death (70%+ for control vs. <30% for ZIKV), but that the BPB model may actually modulate the extent of this neural cell death (70-80% for control and ZIKV exposure). These findings mimicked those found in vivo, helping to validate that this model is relevant in studying biologically relevant molecules. Moreover, given that we could recapitulate the multicellular organization and multilayer architecture of the BPB, we considered this model to be valid for the questions asked, hypotheses posed, and studies performed in all of the objectives described within this dissertation. The second objective was to assess the effects of SSRIs on the BPB’s cells and evaluate the drug’s transport profile across the barrier. We found that both SSRIs, fluoxetine and sertraline, have a transport profile suggestive of three phases: uptake by cells, accumulation within the BPB, and slow release into the fetal compartment. We found that the BPB is impacted by both drugs, where higher dosages (i.e. 10000 ng/mL) lead to increase secretions of ICAM and VCAM, and that sertraline modestly reduces barrier resistance (72.8 Ω∙cm2 for sertraline vs 102.2 Ω∙cm2 for no exposure control at day 9 following daily drug exposure). We showed these drugs primarily impact secretions from endothelial cells, not trophoblast (ICAM secretions of 980 pg/mL vs. 63 pg/mL for ECs and trophoblast, respectively, following 3 days of exposure to fluoxetine at 10000 ng/mL). Further, within ECs, sertraline elicits a greater effect across dosages than fluoxetine does (ICAM secretions of 2300 pg/mL vs 420 pg/mL after 3 days of exposure at 100 ng/mL for each drug). We also found TGFβ secretions are impacted, though again to a greater extent in ECs, not 124 trophoblast, by sertraline than fluoxetine (secretions of 1430 pg/mL vs 950 pg/mL for comparable dosages and lengths of exposure to sertraline and fluoxetine, respectively). The third objective was to assess how SSRIs influence cardiomyocyte signaling and injury biomarker release following passage through the BPB. We found that the calcium transients of cardiomyocytes were impacted to a greater extent by sertraline (33% increase in oscillation period) than fluoxetine (2% increase in oscillation period) when the cells were directly exposed to the drug, but that the two drugs had a more comparable effect when the cardiomyocytes were exposed to the drugs following their passage through the BPB (32% and 30% relative increase in oscillation period, for sertraline and fluoxetine, respectively). In evaluating biomarkers, we found that CKMB, Troponin T, and VCAM secretions increased in response to fluoxetine, not sertraline. Conversely, sertraline had a greater impact on NT-proBNP secretions than fluoxetine did. This suggests that cardiomyocytes are impacted by the drugs in different ways, meaning there are drug-specific effects. More importantly, the effects we observed with the BPB model were different than those without the model, indicating the important of utilizing this model to assess drug effects on fetal cells downstream of the barrier. In conclusion, these studies demonstrate that we were able to successfully recapitulate the placental barrier utilizing a tissue engineering approach to develop a 3D biomimetic model. We utilized this model to advance our knowledge regarding the role of the placenta in maternal-fetal transmission of Zika Virus, the influence of SSRIs on placental barrier properties and the cells within the BPB, and how SSRIs 125 influence calcium handling and biomarker release by cardiomyocytes downstream of the placental barrier. In sum, we suggest that this work provides a biomimetic model to assess pharmacology and toxicology properties of medications to ensure safety of both mother and fetus during pregnancy. 7.2 Contributions My research has resulted in 7 publications published or in press (3 first- author), along with 5 additional publications in preparation or in the submission and peer-review process. I have co-authored 16 conference abstracts and presented at 5 conferences, including internationally at TERMIS World Congress 2018 in Kyoto, Japan, where my oral presentation was awarded 3rd place among students and young investigators presentations. More broadly, through the work presented herein, we have developed a tissue-engineered model of the placental barrier and utilized it to study Zika Virus and SSRIs. To my knowledge, this is one of the first tissue engineering approaches to developing a placental barrier model. Additionally, this is the first in vitro model to study Zika Virus transport and show the placenta both as a preferential location for viral replication and as a modulate of fetal viral exposure. To my knowledge, the concentration profiles for both fluoxetine and sertraline were not previously established, where we showed that both drugs seem to go through an uptake-accumulation-release profile over time. Additionally, we showed drug- dependent and dose-dependent effects for fluoxetine and sertraline in placental cells and cardiomyocytes, something which has not been directly compared before. Therefore, in sum, I suggest that this work is novel and has contributed to the fields of tissue engineering, placental biology, virology, pharmacology, and toxicology. 126 7.3 Future Directions The work presented here addresses many biological questions, and through the completion of this work, has allowed us to ask further questions along the way. One of the first questions would be whether primary cells would improve the model, and whether we could make observations that we wouldn’t otherwise be able to. Other groups have utilized primary trophoblast and primary placenta-derived endothelial cells in the context of placental transport158,166. However, I am unaware of any transport-related studies that directly compare these cell types, other than the work completed in this dissertation. Interestingly though, some studies have shown biological responses to primary cells that were not observed when utilizing trophoblast cells lines, and vice versa240,241. Thus, one direction for future work would be to investigate how molecular transport through primary cells compares to cell lines, and evaluate which would be more applicable to each trimester of pregnancy. Another line of questioning revolves around Zika Virus, and whether this model could be used to assess further phenotypic effects beyond neural cell viability. Neural cells have their own injury biomarkers300,301. However, questions regarding Zika Virus-influenced expression of these biomarkers have yet to be answered. Further, understanding the time component (i.e. short-term vs. long-term) would be interesting. Our observations were generally kept within a few days, but extending this to a few weeks may lead to changes that we did not anticipate. Being able to correlate these in vitro studies with published in vivo work would help validate the use of this model, and may enable us to probe further into the unknowns of Zika Virus interactions with neural cells. 127 A third set of questions revolves around the mechanisms for fluoxetine and sertraline. We observed differences between the two drugs in eliciting phenotypic changes within cells. However, intuition suggests that there should be almost no differences in how these drugs interact. They are from the same class of drugs (SSRIs), and are very similar in molecular weight. Probing to assess which proteins they interact with in endothelial cells and cardiomyocytes would help to address some of these differences. Additionally, understanding their interactions with the various membrane proteins on trophoblast cells would help to elucidate the transport mechanisms are play for SSRIs, improving our understanding of maternal-fetal drug transport and perhaps allowing for the development of improved antidepressant medications that ultimately have a reduced risk of injury to the developing fetus. 128 Appendix A: Determination the Diffusion Coefficient Experimentally The diffusion coefficient, D, is a proportionality constant that describes molecular diffusion of some molecule A through some species B by relating flux to a concentration profile. Throughout portions of the work presented herein, molecule A is often a molecule of interest (i.e. glucose, fluoxetine, and/or sertraline, amongst others) while species B is the BPB model used in the experiment, thus describing the molecular diffusion of these molecules of interest through the BPB model that we have developed. Generally, this equation would be: 𝑑𝐶 𝑗 = 𝐷 𝑑𝑧 𝑑𝐶 Where j is flux, D is the diffusion coefficient, and is the concentration profile of a 𝑑𝑧 molecule, assumed to vary only along the z-direction. Experimentally determining the diffusion coefficient required assumptions to be made regarding the system, such as characterizing our cell-laden BPB model as a porous membrane with no reactions or degradation to the molecules of interest. We also assumed that each side of the barrier model was a well-stirred and homogeneous solution, flux across the membrane quickly reaches steady state, the donor compartment contains a solution with a known concentration of species A, and the receiver compartment contains a solution free of species A. This quasi-steady state approach allows us to define the flux of across the BPB model as: 𝐷𝐾 𝑗 = [ ] (𝐶 𝐿 𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) 129 Where K is the partition coefficient, L is the length of the membrane in the direction of transport, and C is the concentration for the donor and receiver compartments. Similarly, we can define mass balance equations for each compartment: 𝑑𝐶 𝑉 𝑑𝑜𝑛𝑜𝑟𝑑𝑜𝑛𝑜𝑟 = −𝐴𝑗 𝑑𝑡 𝑑𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 = +𝐴𝑗 𝑑𝑡 Where A is the area of the membrane and V is the volume in the donor and receiver compartments. If each equation is divided by the respective volume, and the equations are subtracted from one another, we get: 𝑑 1 1 (𝐶 𝑑𝑡 𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = −𝐴𝑗 ( + ) 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Substituting in the previous equation defined for flux, j, we get: 𝑑 𝐷𝐾 1 1 (𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = −𝐴 [ ] (𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) ( + ) 𝑑𝑡 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 More succinctly, we get: 𝑑 𝐴𝐷𝐾 1 1 (𝐶 − 𝐶 𝑑𝑡 𝑑𝑜𝑛𝑜𝑟 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 ) = − ( + ) (𝐶 𝐿 𝑉 𝑉 𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) 𝑑𝑜𝑛𝑜𝑟 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 This equation can be solved, with the known initial condition: 𝑡 = 0, 𝐶 𝑜 𝑜𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 = 𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Rearranging: 1 𝐴𝐷𝐾 1 1 𝑑(𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = − ( + ) 𝑑𝑡 (𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Integrating this equation and solving: 1 𝐴𝐷𝐾 1 1 ∫ 𝑑(𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶(𝐶 − 𝐶 ) 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 ) = ∫ − ( + ) 𝑑𝑡 𝑑𝑜𝑛𝑜𝑟 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 130 𝐴𝐷𝐾 1 1 ln(𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = − ( + ) 𝑡 + (𝑈𝑛𝑘) 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Where Unk is the unknown constant from integrating. Plugging in the initial condition: 𝐴𝐷𝐾 1 1 ln(𝐶𝑜𝑑𝑜𝑛𝑜𝑟 − 𝐶 𝑜 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = − ( + ) 0 + (𝑈𝑛𝑘) 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 ln(𝐶𝑜 − 𝐶𝑜𝑑𝑜𝑛𝑜𝑟 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = 𝑈𝑛𝑘 Replacing Unk with the known constant: 𝐴𝐷𝐾 1 1 ln(𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = − ( + ) 𝑡 + ln(𝐶 𝑜 − 𝐶𝑜 𝐿 𝑉 𝑑𝑜𝑛𝑜𝑟 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 ) 𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Rearranging and simplifying: 𝐷𝐾𝑡𝐴 1 1 ln(𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) − ln(𝐶 𝑜 𝑜 𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟) = − ( + ) 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝐷𝐾𝑡𝐴 1 1 ln ( 𝐶𝑜 𝑜 ) = − ( + ) 𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 This equation contains two unknown values that need to be determined experimentally, specifically the diffusion coefficient, D and the partition coefficient, K. These are related by permeability, P, where P = DK. This can be substituted, yield the equation previously described in the text: 𝐶𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝑃𝑡𝐴 1 1 ln ( ) = − ( + ) 𝐶𝑜 𝑜𝑑𝑜𝑛𝑜𝑟 − 𝐶𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 𝐿 𝑉𝑑𝑜𝑛𝑜𝑟 𝑉𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 In this equation, concentrations are determined experimentally over time through a diffusion experiment, while membrane area, length, and volumes are known. This leaves the permeability as the only unknown, and it can therefore be calculated from experimental data. Once the permeability is known, the diffusion coefficient can be calculated. This is, however, reliant upon knowing the partition coefficient, K, which 131 itself can be calculated experimentally. As previously described, it is given by the equation: 𝐶𝑔𝑒𝑙 (𝐶𝑖𝑛𝑖𝑡𝑖𝑎𝑙 − 𝐶𝑓𝑖𝑛𝑎𝑙)𝑉𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 𝐾 = = 𝐶𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 𝐶𝑓𝑖𝑛𝑎𝑙𝑉𝑔𝑒𝑙 To determine the partition coefficient, the gel (i.e. an acellular BPB model) is placed into a solution without species A. After the gel has reached equilibrium in that solution, the initial solution is replaced by same solution with a known concentration of species A, and solution concentration is measured until equilibrium is reached. From these concentrations, and known volumes for the solution and gel, the partition coefficient can be calculated. As mentioned above, this partition coefficient can be used in conjunction with the permeability of a molecule interest determined experimentally to calculate the diffusion coefficient. 132 Bibliography 1. Mcbride, W. G. Thalidiomide and congenital abnormalities. Lancet 278, 1358 (1961). 2. Rasmussen, S. A., Jamieson, D. J., Honein, M. A. & Petersen, L. R. Zika Virus and Birth Defects — Reviewing the Evidence for Causality. N. Engl. J. Med. 1–7 (2016). doi:10.1056/NEJMsr1604338 3. Cugola, F. R., Fernandes, I. R., Russo, F. B., Freitas, B. C., Dias, J. L. M., Guimarães, K. P., Benazzato, C., Almeida, N., Pignatari, G. C., Romero, S., Polonio, C. M., Cunha, I., Freitas, C. L., Brandão, W. N., Rossato, C., Andrade, D. G., Faria, D. de P., Garcez, A. T., Buchpigel, C. A., et al. The Brazilian Zika virus strain causes birth defects in experimental models. Nature 534, 267–271 (2016). 4. Johansson, M. A., Mier-y-Teran-Romero, L., Reefhuis, J., Gilboa, S. M. & Hills, S. L. Zika and the Risk of Microcephaly. N. Engl. J. Med. 375, 1–4 (2016). 5. Mitchell, A. A., Gilboa, S. M., Werler, M. M., Kelley, K. E., Louik, C. & Hernández-Díaz, S. Medication use during pregnancy, with particular focus on prescription drugs: 1976-2008. Am. J. Obstet. Gynecol. 205, 51.e1-51.e8 (2011). 6. Andrade, S. E., Raebel, M. A., Brown, J., Lane, K., Livingston, J., Boudreau, D., Rolnick, S. J., Roblin, D., Smith, D. H., Willy, M. E., Staffa, J. A. & Platt, R. Use of antidepressant medications during pregnancy: a multisite study. Am. J. Obstet. Gynecol. 198, (2008). 7. Hendrick, V., Stowe, Z. N., Altshuler, L. L., Hwang, S., Lee, E. & Haynes, D. 133 Placental passage of antidepressant medications. Am. J. Psychiatry 160, 993– 996 (2003). 8. Velasquez, J. C., Goeden, N. & Bonnin, A. Placental serotonin: implications for the developmental effects of SSRIs and maternal depression. Front. Cell. Neurosci. 7, 1–7 (2013). 9. El Marroun, H., White, T., Verhulst, F. C. & Tiemeier, H. Maternal use of antidepressant or anxiolytic medication during pregnancy and childhood neurodevelopmental outcomes: a systematic review. Eur. Child Adolesc. Psychiatry 23, 973–992 (2014). 10. Gemmel, M., Bögi, E., Ragan, C., Hazlett, M., Dubovicky, M., van den Hove, D. L., Oberlander, T. F., Charlier, T. D. & Pawluski, J. L. Perinatal selective serotonin reuptake inhibitor medication (SSRI) effects on social behaviors, neurodevelopment and the epigenome. Neuroscience and Biobehavioral Reviews 85, 102–116 (2018). 11. Pacher, P. & Kecskemeti, V. Cardiovascular Side Effects of New Antidepressants and Antipsychotics: New Drugs, old Concerns? Curr. Pharm. Des. 10, 2463–2475 (2004). 12. Pacher, P., Ungvari, Z., Kecskemeti, V. & Furst, S. Review of Cardiovascular Effects of Fluoxetine, a Selective Serotonine Reuptake Inhibitor, Compared to Tricyclic Antidepressants. Curr. Med. Chem. 5, 381–390 (1998). 13. Lee, H. A., Kim, K. S., Hyun, S. A., Park, S. G. & Kim, S. J. Wide spectrum of inhibitory effects of sertraline on cardiac ion channels. Korean J. Physiol. Pharmacol. 16, 327–332 (2012). 134 14. Alwan, S., Friedman, J. M. & Chambers, C. Safety of Selective Serotonin Reuptake Inhibitors in Pregnancy: A Review of Current Evidence. CNS Drugs 30, 499–515 (2016). 15. Casper, R. C. Use of selective serotonin reuptake inhibitor antidepressants in pregnancy does carry risks, but the risks are small. J. Nerv. Ment. Dis. 203, 167–9 (2015). 16. Sastry, B. V. R. Techniques to study human placental transport. Advanced Drug Delivery Reviews 38, 17–39 (1999). 17. 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 (80-. ). 328, 1662–1668 (2010). 18. 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. 4, 159ra147-159ra147 (2012). 19. Zietek, T., Rath, E., Haller, D. & Daniel, H. Intestinal organoids for assessing nutrient transport, sensing and incretin secretion. Sci. Rep. 5, 16831 (2015). 20. Fatehullah, A., Tan, S. H. & Barker, N. Organoids as an in vitro model of human development and disease. Nat. Cell Biol. 18, 246–54 (2016). 21. Devriese, S., Van den Bossche, L., Van Welden, S., Holvoet, T., Pinheiro, I., Hindryckx, P., De Vos, M. & Laukens, D. T84 monolayers are superior to Caco-2 as a model system of colonocytes. Histochem. Cell Biol. 148, 85–93 (2017). 135 22. Sakolish, C. M., Esch, M. B., Hickman, J. J., Shuler, M. L. & Mahler, G. J. Modeling Barrier Tissues In Vitro: Methods, Achievements, and Challenges. EBioMedicine 5, 30–39 (2016). 23. Duval, K., Grover, H., Han, L.-H., Mou, Y., Pegoraro, A. F., Fredberg, J. & Chen, Z. Modeling Physiological Events in 2D vs. 3D Cell Culture. Physiology 32, 266–277 (2017). 24. Little, M. H. Organoids: a Special Issue. Development 144, 935–937 (2017). 25. Huch, M., Knoblich, J. A., Lutolf, M. P. & Martinez-Arias, A. The hope and the hype of organoid research. Development 144, 938–941 (2017). 26. Braakhuis, H. M., Kloet, S. K., Kezic, S., Kuper, F., Park, M. V. D. Z., Bellmann, S., van der Zande, M., Le Gac, S., Krystek, P., Peters, R. J. B., Rietjens, I. M. C. M. & Bouwmeester, H. Progress and future of in vitro models to study translocation of nanoparticles. Archives of Toxicology 89, 1469–1495 (2015). 27. Markeson, D., Pleat, J. M., Sharpe, J. R., Harris, A. L., Seifalian, A. M. & Watt, S. M. Scarring, stem cells, scaffolds and skin repair. J. Tissue Eng. Regen. Med. 9, 649–668 (2015). 28. Bannasch, H., Momeni, A., Knam, F., Stark, G. B. & Föhn, M. Tissue engineering of skin substitutes. Panminerva Med. 47, 53–60 (2005). 29. Zhang, Z. & Michniak-Kohn, B. B. Tissue engineered human skin equivalents. Pharmaceutics 4, 26–41 (2012). 30. Sugihara, H., Toda, S., Yonemitsu, N. & Watanabe, K. Effects of fat cells on keratinocytes and fibroblasts in a reconstructed rat skin model using collagen 136 gel matrix culture. Br. J. Dermatol. 144, 244–53 (2001). 31. Bellas, E., Seiberg, M., Garlick, J. & Kaplan, D. L. In vitro 3D Full-Thickness Skin-Equivalent Tissue Model Using Silk and Collagen Biomaterials. Macromol. Biosci. 12, 1627–1636 (2012). 32. Monfort, A., Soriano-Navarro, M., García-Verdugo, J. M. & Izeta, A. Production of human tissue-engineered skin trilayer on a plasma-based hypodermis. J. Tissue Eng. Regen. Med. 7, 479–490 (2013). 33. Mortensen, L. J., Jatana, S., Gelein, R., De Benedetto, A., De Mesy Bentley, K. L., Beck, L. A., Elder, A. & Delouise, L. A. Quantification of quantum dot murine skin penetration with UVR barrier impairment. Nanotoxicology 7, 1386–1398 (2013). 34. Huong, S. P., Bun, H., Fourneron, J. D., Reynier, J. P. & Andrieu, V. Use of various models for in vitro percutaneous absorption studies of ultraviolet filters. Ski. Res. Technol. 15, 253–261 (2009). 35. Abd, E., Yousef, S. A., Pastore, M. N., Telaprolu, K., Mohammed, Y. H., Namjoshi, S., Grice, J. E. & Roberts, M. S. Skin models for the testing of transdermal drugs. Clin. Pharmacol. Adv. Appl. 8, 163–176 (2016). 36. Bond, J. R. & Barry, B. W. Limitations of hairless mouse skin as a model for in vitro permeation studies through human skin: hydration damage. J. Invest. Dermatol. 90, 486–9 (1988). 37. Curren, R. D., Mun, G. C., Gibson, D. P. & Aardema, M. J. Development of a method for assessing micronucleus induction in a 3D human skin model (EpiDermTM). Mutat. Res. Toxicol. Environ. Mutagen. 607, 192–204 (2006). 137 38. Rissmann, R., Oudshoorn, M. H. M., Hennink, W. E., Ponec, M. & Bouwstra, J. a. Skin barrier disruption by acetone: Observations in a hairless mouse skin model. Arch. Dermatol. Res. 301, 609–613 (2009). 39. Scheuplein, R. J. Mechanism of percutaneous absorption. II. Transient diffusion and the relative importance of various routes of skin penetration. J. Invest. Dermatol. 48, 79–88 (1967). 40. Teodorescu, F., Quéniat, G., Foulon, C., Lecoeur, M., Barras, A., Boulahneche, S., Medjram, M. S., Hubert, T., Abderrahmani, A., Boukherroub, R. & Szunerits, S. Transdermal skin patch based on reduced graphene oxide: A new approach for photothermal triggered permeation of ondansetron across porcine skin. J. Control. Release 245, 137–146 (2017). 41. Van Bocxlaer, K., Yardley, V., Murdan, S. & Croft, S. L. Drug permeation and barrier damage in Leishmania-infected mouse skin. J. Antimicrob. Chemother. 71, 1578–1585 (2016). 42. Hsia, E., Johnston, M. J., Houlden, R. J., Chern, W. H. & Hofland, H. E. J. Effects of topically applied acitretin in reconstructed human epidermis and the rhino mouse. J. Invest. Dermatol. 128, 125–130 (2008). 43. Fleischli, F. D., Morf, F. & Adlhart, C. Skin Concentrations of Topically Applied Substances in Reconstructed Human Epidermis (RHE) Compared with Human Skin Using <I>in vivo</I> Confocal Raman Microscopy. Chim. Int. J. Chem. 69, 147–151 (2015). 44. Rodrigues, F., Pereira, C., Pimentel, F. B., Alves, R. C., Ferreira, M., Sarmento, B., Amaral, M. H. & Oliveira, M. B. P. P. Are coffee silverskin 138 extracts safe for topical use? An in vitro and in vivo approach. Ind. Crops Prod. 63, 167–174 (2015). 45. Mun, G. C., Aardema, M. J., Hu, T., Barnett, B., Kaluzhny, Y., Klausner, M., Karetsky, V., Dahl, E. L. & Curren, R. D. Further development of the EpiDermTM 3D reconstructed human skin micronucleus (RSMN) assay. Mutat. Res. - Genet. Toxicol. Environ. Mutagen. 673, 92–99 (2009). 46. Wills, J. W., Hondow, N., Thomas, A. D., Chapman, K. E., Fish, D., Maffeis, T. G., Penny, M. W., Brown, R. A., Jenkins, G. J. S., Brown, A. P., White, P. A. & Doak, S. H. Genetic toxicity assessment of engineered nanoparticles using a 3D in vitro skin model (EpiDermTM). Part. Fibre Toxicol. 13, 1–21 (2016). 47. Branka, R., Mirjana, G., Estelle, T.-T., Fabrice, P. & Francoise, F. Simultaneous absorption of vitamins C and E from topical microemulsions using reconstructed human epidermis as a skin model. Eur. J. Pharm. Biopharm. 72, 69–75 (2009). 48. Trottier, V., Marceau-Fortier, G., Germain, L., Vincent, C. & Fradette, J. IFATS Collection: Using Human Adipose-Derived Stem/Stromal Cells for the Production of New Skin Substitutes. Stem Cells 26, 2713–2723 (2008). 49. Garland, M. J., Migalska, K., Tuan-Mahmood, T. M., Raghu Raj Singh, T., Majithija, R., Caffarel-Salvador, E., McCrudden, C. M., McCarthy, H. O., David Woolfson, A. & Donnelly, R. F. Influence of skin model on in vitro performance of drug-loaded soluble microneedle arrays. Int. J. Pharm. 434, 80–89 (2012). 139 50. Oshima, S., Suzuki, C., Yajima, R., Egawa, Y., Hosoya, O., Juni, K. & Seki, T. The Use of an Artificial Skin Model to Study Transdermal Absorption of Drugs in Inflamed Skin. Biol. Pharm. Bull. 35, 203–209 (2012). 51. Flamand, N., Marrot, L., Belaidi, J. P., Bourouf, L., Dourille, E., Feltes, M. & Meunier, J. R. Development of genotoxicity test procedures with Episkin®, a reconstructed human skin model: Towards new tools for in vitro risk assessment of dermally applied compounds? Mutat. Res. - Genet. Toxicol. Environ. Mutagen. 606, 39–51 (2006). 52. Grégoire, S., Ribaud, C., Benech, F., Meunier, J. R., Garrigues-Mazert, A. & Guy, R. H. Prediction of chemical absorption into and through the skin from cosmetic and dermatological formulations. Br. J. Dermatol. 160, 80–91 (2009). 53. Carreras, N., Alonso, C., Martí, M. & Lis, M. J. Mass transport model through the skin by microencapsulation system. J. Microencapsul. 32, 358–363 (2015). 54. Davies, D. J., Heylings, J. R., Gayes, H., McCarthy, T. J. & Mack, M. C. Further development of an in vitro model for studying the penetration of chemicals through compromised skin. Toxicol. Vitr. 38, 101–107 (2017). 55. Paz, A. C., Trieu, D., Soleas, J., Poon, J. C. H., Waddell, T. K. & McGuigan, A. P. Challenges and Opportunities for Tissue-Engineering Polarized Epithelium. Tissue Eng. Part B Rev. 20, 56–72 (2013). 56. Ali, S. & Rytting, E. Influences of nanomaterials on the barrier function of epithelial cells. Adv. Exp. Med. Biol. 811, 45–54 (2014). 57. Ananta, M., Brown, R. A. & Mudera, V. A Rapid Fabricated Living Dermal Equivalent for Skin Tissue Engineering: An In Vivo Evaluation in an Acute 140 Wound Model. Tissue Eng. Part A 18, 353–361 (2012). 58. Guo, Z., Higgins, C. A., Gillette, B. M., Itoh, M., Umegaki, N., Gledhill, K., Sia, S. K. & Christiano, A. M. Building a microphysiological skin model from induced pluripotent stem cells. Stem Cell Res Ther 4 Suppl 1, S2 (2013). 59. Reijnders, C. M. A., van Lier, A., Roffel, S., Kramer, D., Scheper, R. J. & Gibbs, S. Development of a Full-Thickness Human Skin Equivalent In Vitro Model Derived from TERT-Immortalized Keratinocytes and Fibroblasts. Tissue Eng. Part A 21, 2448–2459 (2015). 60. Kim, B. S., Lee, J. S., Gao, G. & Cho, D. W. Direct 3D cell-printing of human skin with functional transwell system. Biofabrication 9, (2017). 61. Koch, L., Deiwick, A., Schlie, S., Michael, S., Gruene, M., Coger, V., Zychlinski, D., Schambach, A., Reimers, K., Vogt, P. M. & Chichkov, B. Skin tissue generation by laser cell printing. Biotechnol. Bioeng. 109, 1855–1863 (2012). 62. Ataç, B., Wagner, I., Horland, R., Lauster, R., Marx, U., Tonevitsky, A. G., Azar, R. P. & Lindner, G. Skin and hair on-a-chip: in vitro skin models versus ex vivo tissue maintenance with dynamic perfusion. Lab Chip 13, 3555 (2013). 63. Lee, S., Jin, S.-P., Kim, Y. K., Sung, G. Y., Chung, J. H. & Sung, J. H. Construction of 3D multicellular microfluidic chip for an in vitro skin model. Biomed. Microdevices 19, 22 (2017). 64. van den Broek, L. J., Bergers, L. I. J. C., Reijnders, C. M. A. & Gibbs, S. Progress and Future Prospectives in Skin-on-Chip Development with Emphasis on the use of Different Cell Types and Technical Challenges. Stem Cell Rev. 141 Reports 13, 418–429 (2017). 65. Li, Y., Wang, S., Huang, R., Huang, Z., Hu, B., Zheng, W., Yang, G. & Jiang, X. Evaluation of the Effect of the Structure of Bacterial Cellulose on Full Thickness Skin Wound Repair on a Microfluidic Chip. Biomacromolecules 16, 780–789 (2015). 66. Mori, N., Morimoto, Y. & Takeuchi, S. Skin integrated with perfusable vascular channels on a chip. Biomaterials 116, 48–56 (2017). 67. Wufuer, M., Lee, G. H., Hur, W., Jeon, B., Kim, B. J., Choi, T. H. & Lee, S. H. Skin-on-a-chip model simulating inflammation, edema and drug-based treatment. Sci. Rep. 6, 1–12 (2016). 68. Abaci, H. E., Guo, Z., Doucet, Y., Jacków, J. & Christiano, A. Next generation human skin constructs as advanced tools for drug development. Experimental Biology and Medicine 242, 1657–1668 (2017). 69. Kararli, T. T. Comparison of the gastrointestinal anatomy, physiology, and biochemistry of humans and commonly used laboratory animals. Biopharmaceutics & Drug Disposition 16, 351–380 (1995). 70. Carr, K. E. & Toner, P. G. Morphology of the intestinal Mucosa. in Pharmacology of Intestinal Permeation 1–3 (Springer, Berlin, Heidelberg, 1984). doi:10.1007/978-3-642-69505-6_1 71. Peterson, L. W. & Artis, D. Intestinal epithelial cells: regulators of barrier function and immune homeostasis. Nat. Rev. Immunol. 14, 141–53 (2014). 72. Hilgendorf, C., Spahn-Langguth, H., Regårdh, C. G., Lipka, E., Amidon, G. L. & Langguth, P. Caco-2 versus Caco-2/HT29-MTX co-cultured cell lines: 142 Permeabilities via diffusion, inside- and outside-directed carrier-mediated transport. J. Pharm. Sci. 89, 63–75 (2000). 73. Mahler, G. J., Shuler, M. L. & Glahn, R. P. Characterization of Caco-2 and HT29-MTX cocultures in an in vitro digestion/cell culture model used to predict iron bioavailability. J. Nutr. Biochem. 20, 494–502 (2009). 74. Pan, F., Han, L., Zhang, Y., Yu, Y. & Liu, J. Optimization of Caco-2 and HT29 co-culture in vitro cell models for permeability studies. Int. J. Food Sci. Nutr. 66, 680–685 (2015). 75. Ensign, L. M., Cone, R. & Hanes, J. Oral drug delivery with polymeric nanoparticles: The gastrointestinal mucus barriers. Advanced Drug Delivery Reviews 64, 557–570 (2012). 76. Kelly, J. R., Kennedy, P. J., Cryan, J. F., Dinan, T. G., Clarke, G. & Hyland, N. P. Breaking down the barriers: the gut microbiome, intestinal permeability and stress-related psychiatric disorders. Front. Cell. Neurosci. 9, 392 (2015). 77. Li, N., Wang, D., Sui, Z., Qi, X., Ji, L., Wang, X. & Yang, L. Development of an Improved Three-Dimensional In Vitro Intestinal Mucosa Model for Drug Absorption Evaluation. Tissue Eng. Part C Methods 19, 708–719 (2013). 78. Esch, M. B., Mahler, G. J., Stokol, T. & Shuler, M. L. Body-on-a-chip simulation with gastrointestinal tract and liver tissues suggests that ingested nanoparticles have the potential to cause liver injury. Lab Chip 14, 3081–3092 (2014). 79. Pocock, K., Delon, L., Bala, V., Rao, S., Priest, C., Prestidge, C. & Thierry, B. Intestine-on-A-Chip Microfluidic Model for Efficient in Vitro Screening of 143 Oral Chemotherapeutic Uptake. ACS Biomater. Sci. Eng. 3, 951–959 (2017). 80. Costello, C. M., Phillipsen, M. B., Hartmanis, L. M., Kwasnica, M. A., Chen, V., Hackam, D., Chang, M. W., Bentley, W. E. & March, J. C. Microscale Bioreactors for in situ characterization of GI epithelial cell physiology. Sci. Rep. 7, 1–10 (2017). 81. Farrell, T. L., Gomez-Juaristi, M., Poquet, L., Redeuil, K., Nagy, K., Renouf, M. & Williamson, G. Absorption of dimethoxycinnamic acid derivatives in vitro and pharmacokinetic profile in human plasma following coffee consumption. Mol. Nutr. Food Res. 56, 1413–1423 (2012). 82. Wu, X.-W., Wei, W., Yang, X.-W., Zhang, Y.-B., Xu, W., Yang, Y.-F., Zhong, G.-Y., Liu, H.-N. & Yang, S.-L. Anti-Inflammatory Phenolic Acid Esters from the Roots and Rhizomes of Notopterygium incisium and Their Permeability in the Human Caco-2 Monolayer Cell Model. Molecules 22, 935 (2017). 83. Kobayashi, T., Koizumi, T., Kobayashi, M., Ogura, J., Horiuchi, Y., Kimura, Y., Kondo, A., Furugen, A., Narumi, K., Takahashi, N. & Iseki, K. Insulin stimulates transport of organic anion compounds mediated by organic anion transporting polypeptide 2B1 in the human intestinal cell line Caco-2. Drug Metab. Pharmacokinet. 32, 157–163 (2017). 84. Rani, P., Vashisht, M., Golla, N., Shandilya, S., Onteru, S. K. & Singh, D. Milk miRNAs encapsulated in exosomes are stable to human digestion and permeable to intestinal barrier in vitro. J. Funct. Foods 34, 431–439 (2017). 85. Yu, Y., Wang, M., Zhang, K., Yang, D., Zhong, Y., An, J., Lei, B. & Zhang, X. The transepithelial transport mechanism of polybrominated diphenyl ethers 144 in human intestine determined using a Caco-2 cell monolayer. Environ. Res. 154, 93–100 (2017). 86. Kimura, O., Kotaki, Y., Hamaue, N., Haraguchi, K. & Endo, T. Transcellular transport of domoic acid across intestinal Caco-2 cell monolayers. Food Chem. Toxicol. 49, 2167–2171 (2011). 87. Bhattacherjee, A., Hrynets, Y. & Betti, M. Transport of the Glucosamine- Derived Browning Product Fructosazine (Polyhydroxyalkylpyrazine) Across the Human Intestinal Caco-2 Cell Monolayer: Role of the Hexose Transporters. J. Agric. Food Chem. 65, 4642–4650 (2017). 88. Lozoya-Agullo, I., González-Álvarez, I., González-Álvarez, M., Merino- Sanjuán, M. & Bermejo, M. In Situ Perfusion Model in Rat Colon for Drug Absorption Studies: Comparison with Small Intestine and Caco-2 Cell Model. J. Pharm. Sci. 104, 3136–3145 (2015). 89. Noah, T. K., Donahue, B. & Shroyer, N. F. Intestinal development and differentiation. Experimental Cell Research 317, 2702–2710 (2011). 90. Araújo, F. & Sarmento, B. Towards the characterization of an in vitro triple co- culture intestine cell model for permeability studies. Int. J. Pharm. 458, 128– 134 (2013). 91. Béduneau, A., Tempesta, C., Fimbel, S., Pellequer, Y., Jannin, V., Demarne, F. & Lamprecht, A. A tunable Caco-2/HT29-MTX co-culture model mimicking variable permeabilities of the human intestine obtained by an original seeding procedure. Eur. J. Pharm. Biopharm. 87, 290–298 (2014). 92. Antunes, F., Andrade, F., Araújo, F., Ferreira, D. & Sarmento, B. 145 Establishment of a triple co-culture in vitro cell models to study intestinal absorption of peptide drugs. Eur. J. Pharm. Biopharm. 83, 427–435 (2013). 93. Zhao, J., Zeng, Z., Sun, J., Zhang, Y., Li, D., Zhang, X., Liu, M. & Wang, X. A Novel Model of P-Glycoprotein Inhibitor Screening Using Human Small Intestinal Organoids. Basic Clin. Pharmacol. Toxicol. 120, 250–255 (2017). 94. Costello, C. M., Sorna, R. M., Goh, Y. L., Cengic, I., Jain, N. K. & March, J. C. 3-D intestinal scaffolds for evaluating the therapeutic potential of probiotics. Mol. Pharm. 11, 2030–2039 (2014). 95. Leonard, F., Collnot, E.-M. & Lehr, C.-M. A 3-dimensional co-culture of enterocytes, macrophages and dendritic cells to model the inflamed intestinal mucosa in vitro. Mol. Pharm. 7, 2103–2119 (2010). 96. Villenave, R., Wales, S. Q., Hamkins-Indik, T., Papafragkou, E., Weaver, J. C., Ferrante, T. C., Bahinski, A., Elkins, C. A., Kulka, M. & Ingber, D. E. Human gut-on-a-chip supports polarized infection of coxsackie B1 virus in vitro. PLoS One 12, e0169412 (2017). 97. Gayer, C. P. & Basson, M. D. The effects of mechanical forces on intestinal physiology and pathology. Cellular Signalling 21, 1237–1244 (2009). 98. Kim, H. J., Huh, D., Hamilton, G. & Ingber, D. E. Human gut-on-a-chip inhabited by microbial flora that experiences intestinal peristalsis-like motions and flow. Lab Chip 12, 2165–2174 (2012). 99. Bhattacharya, J. & Matthay, M. A. Regulation and Repair of the Alveolar- Capillary Barrier in Acute Lung Injury. Annu. Rev. Physiol. 75, 593–615 (2013). 146 100. Rackley, C. R. & Stripp, B. R. Building and maintaining the epithelium of the lung. J. Clin. Invest. 122, 2724–2730 (2012). 101. Kasper, J. Y., Hermanns, M. I., Unger, R. E. & Kirkpatrick, C. J. A responsive human triple-culture model of the air–blood barrier: incorporation of different macrophage phenotypes. J. Tissue Eng. Regen. Med. 11, 1285–1297 (2017). 102. Lambrecht, B. N. Alveolar Macrophage in the Driver’s Seat. Immunity 24, 366–368 (2006). 103. Mao, P., Wu, S., Li, J., Fu, W., He, W., Liu, X., Slutsky, A. S., Zhang, H. & Li, Y. Human alveolar epithelial type II cells in primary culture. Physiol. Rep. 3, (2015). 104. Bove, P. F., Dang, H., Cheluvaraju, C., Jones, L. C., Liu, X., O’Neal, W. K., Randell, S. H., Schlegel, R. & Boucher, R. C. Breaking the in vitro alveolar type II cell proliferation barrier while retaining ion transport properties. Am. J. Respir. Cell Mol. Biol. 50, 767–776 (2014). 105. WARD, H. E. & NICHOLAS, T. E. ALVEOLAR TYPE I AND TYPE II CELLS. Aust. N. Z. J. Med. 14, 731–734 (1984). 106. Castranova, V., Rabovsky, J., Tucker, J. H. & Miles, P. R. The alveolar type II epithelial cell: A multifunctional pneumocyte. Toxicol. Appl. Pharmacol. 93, 472–483 (1988). 107. Klein, S. G., Serchi, T., Hoffmann, L., Blömeke, B. & Gutleb, A. C. An improved 3D tetraculture system mimicking the cellular organisation at the alveolar barrier to study the potential toxic effects of particles on the lung. Part. Fibre Toxicol. 10, 31 (2013). 147 108. Lehmann, A. D., Daum, N., Bur, M., Lehr, C. M., Gehr, P. & Rothen- Rutishauser, B. M. An in vitro triple cell co-culture model with primary cells mimicking the human alveolar epithelial barrier. Eur. J. Pharm. Biopharm. 77, 398–406 (2011). 109. Horváth, L., Umehara, Y., Jud, C., Blank, F., Petri-Fink, A. & Rothen- Rutishauser, B. Engineering an in vitro air-blood barrier by 3D bioprinting. Sci. Rep. 5, 7974 (2015). 110. Nichols, J. E., Niles, J. A., Vega, S. P., Argueta, L. B., Eastaway, A. & Cortiella, J. Modeling the lung: Design and development of tissue engineered macro- and micro-physiologic lung models for research use. Exp. Biol. Med. 239, 1135–1169 (2014). 111. Ren, H., Birch, N. P. & Suresh, V. An Optimised Human Cell Culture Model for Alveolar Epithelial Transport. PLoS One 11, e0165225 (2016). 112. Ong, H. X., Traini, D., Bebawy, M. & Young, P. M. Ciprofloxacin is actively transported across bronchial lung epithelial cells using a calu-3 air interface cell model. Antimicrob. Agents Chemother. 57, 2535–2540 (2013). 113. Kuehn, A., Kletting, S., De Souza Carvalho-Wodarz, C., Repnik, U., Griffiths, G., Fischer, U., Meese, E., Huwer, H., Wirth, D., May, T., Schneider-Daum, N. & Lehr, C. M. Human alveolar epithelial cells expressing tight junctions to model the air-blood barrier. ALTEX 33, 251–260 (2016). 114. Nalayanda, D. D., Wang, Q., Fulton, W. B., Wang, T. H. & Abdullah, F. Engineering an artificial alveolar-capillary membrane: a novel continuously perfused model within microchannels. J. Pediatr. Surg. 45, 45–51 (2010). 148 115. Kasper, J. Y., Feiden, L., Hermanns, M. I., Bantz, C., Maskos, M., Unger, R. E. & Kirkpatrick, C. J. Pulmonary surfactant augments cytotoxicity of silica nanoparticles: Studies on an in vitro air-blood barrier model. Beilstein J. Nanotechnol. 6, 517–528 (2015). 116. Bengalli, R., Mantecca, P., Camatini, M. & Gualtieri, M. Effect of nanoparticles and environmental particles on a cocultures model of the air- blood barrier. Biomed Res. Int. 2013, (2013). 117. Zeng, L., Yang, X., Li, H., Li, Y., Yang, C., Gu, W., Zhou, Y., Du, J., Wang, H., Sun, J., Wen, D. & Jiang, J. The cellular kinetics of lung alveolar epithelial cells and its relationship with lung tissue repair after acute lung injury. Respir. Res. 17, 164 (2016). 118. Mathias, N. R., Timoszyk, J., Stetsko, P. I., Megill, J. R., Smith, R. L. & Wall, D. A. Permeability characteristics of Calu-3 human bronchial epithelial cells: In vitro-in vitro correlation to predict lung absorption in rats. J. Drug Target. 10, 31–40 (2002). 119. Nayak, P. S., Wang, Y., Najrana, T., Priolo, L. M., Rios, M., Shaw, S. K. & Sanchez-Esteban, J. Mechanotransduction via TRPV4 regulates inflammation and differentiation in fetal mouse distal lung epithelial cells. Respir. Res. 16, 60 (2015). 120. Patel, H. & Kwon, S. Interplay between cytokine-induced and cyclic equibiaxial deformation-induced nitric oxide production and metalloproteases expression in human alveolar epithelial cells. in Cellular and Molecular Bioengineering 2, 615–624 (Springer US, 2009). 149 121. Rothen-Rutishauser, B. M., Kiama, S. G. & Gehr, P. A three-dimensional cellular model of the human respiratory tract to study the interaction with particles. Am. J. Respir. Cell Mol. Biol. 32, 281–289 (2005). 122. Arumugasaamy, N., Baker, H. B., Kaplan, D. S., Kim, P. C. W. & Fisher, J. P. Fabrication and Printing of Multi-material Hydrogels. in 3D Printing and Biofabrication 1–34 (Springer International Publishing, 2016). doi:10.1007/978-3-319-40498-1_13-1 123. Mehta, D. Signaling Mechanisms Regulating Endothelial Permeability. Physiol. Rev. 86, 279–367 (2006). 124. Czupalla, C. J., Liebner, S. & Devraj, K. In Vitro Models of the Blood–Brain Barrier. in Methods in molecular biology (Clifton, N.J.) 1135, 415–437 (2014). 125. Naik, P. & Cucullo, L. In vitro blood-brain barrier models: Current and perspective technologies. Journal of Pharmaceutical Sciences 101, 1337–1354 (2012). 126. Cho, C.-F., Wolfe, J. M., Fadzen, C. M., Calligaris, D., Hornburg, K., Chiocca, E. A., Agar, N. Y. R., Pentelute, B. L. & Lawler, S. E. Blood-brain-barrier spheroids as an in vitro screening platform for brain-penetrating agents. Nat. Commun. 8, 15623 (2017). 127. Appelt-Menzel, A., Cubukova, A., Günther, K., Edenhofer, F., Piontek, J., Krause, G., Stüber, T., Walles, H., Neuhaus, W. & Metzger, M. Establishment of a Human Blood-Brain Barrier Co-culture Model Mimicking the Neurovascular Unit Using Induced Pluri- and Multipotent Stem Cells. Stem Cell Reports 8, 894–906 (2017). 150 128. Wang, Y., Wang, N., Cai, B., Wang, G. Y., Li, J. & Piao, X. X. In vitro model of the blood-brain barrier established by co-culture of primary cerebral microvascular endothelial and astrocyte cells. Neural Regen. Res. 10, 2011– 2017 (2015). 129. Wang, Y. I., Abaci, H. E. & Shuler, M. L. Microfluidic blood–brain barrier model provides in vivo-like barrier properties for drug permeability screening. Biotechnol. Bioeng. 114, 184–194 (2017). 130. Woodworth, G. F., Dunn, G. P., Nance, E. A., Hanes, J. & Brem, H. Emerging Insights into Barriers to Effective Brain Tumor Therapeutics. Front. Oncol. 4, 1–14 (2014). 131. Banks, W. A. Characteristics of compounds that cross the blood-brain barrier. BMC Neurol. 9, S3 (2009). 132. Lippmann, E. S., Al-Ahmad, A., Azarin, S. M., Palecek, S. P. & Shusta, E. V. A retinoic acid-enhanced, multicellular human blood-brain barrier model derived from stem cell sources. Sci. Rep. 4, 4160 (2015). 133. Mohamed, L. A., Zhu, H., Mousa, Y. M., Wang, E., Qiu, W. Q. & Kaddoumi, A. Amylin Enhances Amyloid-β Peptide Brain to Blood Efflux Across the Blood-Brain Barrier. J. Alzheimer’s Dis. 56, 1087–1099 (2017). 134. Liew, K. F., Hanapi, N. A., Chan, K. L., Yusof, S. R. & Lee, C. Y. Assessment of the Blood-Brain Barrier Permeability of Potential Neuroprotective Aurones in Parallel Artificial Membrane Permeability Assay and Porcine Brain Endothelial Cell Models. J. Pharm. Sci. 106, 502–510 (2017). 135. Siupka, P., Hersom, M. N., Lykke-Hartmann, K., Johnsen, K. B., Thomsen, L. 151 B., Andresen, T. L., Moos, T., Abbott, N. J., Brodin, B. & Nielsen, M. S. Bidirectional apical–basal traffic of the cation-independent mannose-6- phosphate receptor in brain endothelial cells. J. Cereb. Blood Flow Metab. 0271678X1770066 (2017). doi:10.1177/0271678X17700665 136. Eigenmann, D. E., Xue, G., Kim, K. S., Moses, A. V, Hamburger, M. & Oufir, M. Comparative study of four immortalized human brain capillary endothelial cell lines, hCMEC/D3, hBMEC, TY10, and BB19, and optimization of culture conditions, for an in vitro blood–brain barrier model for drug permeability studies. Fluids Barriers CNS 10, 33 (2013). 137. Hoff, D., Sheikh, L., Bhattacharya, S., Nayar, S. & Webster, T. J. Comparison study of ferrofluid and powder iron oxide nanoparticle permeability across the blood-brain barrier. Int. J. Nanomedicine 8, 703–710 (2013). 138. Lippmann, E. S., Azarin, S. M., Kay, J. E., Nessler, R. A., Wilson, H. K., Al- Ahmad, A., Palecek, S. P. & Shusta, E. V. Derivation of blood-brain barrier endothelial cells from human pluripotent stem cells. Nat. Biotechnol. 30, 783– 791 (2012). 139. Liu, Z., Mi, J., Yang, S., Zhao, M., Li, Y. & Sheng, L. Effects of P- glycoprotein on the intestine and blood-brain barrier transport of YZG-331, a promising sedative-hypnotic compound. Eur. J. Pharmacol. 791, 339–347 (2016). 140. Freese, C., Hanada, S., Fallier-Becker, P., Kirkpatrick, C. J. & Unger, R. E. Identification of neuronal and angiogenic growth factors in an in vitro blood- brain barrier model system: Relevance in barrier integrity and tight junction 152 formation and complexity. Microvasc. Res. 111, 1–11 (2017). 141. González-Burgos, E., Carretero, M. & Gómez-Serranillos, M. In vitro Permeability Study of CNS-Active Diterpenes from Sideritis spp. Using Cellular Models of Blood-Brain Barrier. Planta Med. 79, 1545–1551 (2013). 142. Song, Y., Du, D., Li, L., Xu, J., Dutta, P. & Lin, Y. In Vitro Study of Receptor- Mediated Silica Nanoparticles Delivery across Blood–Brain Barrier. ACS Appl. Mater. Interfaces 9, 20410–20416 (2017). 143. Biemans, E. A. L. M., Jäkel, L., de Waal, R. M. W., Kuiperij, H. B. & Verbeek, M. M. Limitations of the hCMEC/D3 cell line as a model for Aβ clearance by the human blood-brain barrier. J. Neurosci. Res. 95, 1513–1522 (2017). 144. Mantle, J. L., Min, L. & Lee, K. H. Minimum Transendothelial Electrical Resistance Thresholds for the Study of Small and Large Molecule Drug Transport in a Human in Vitro Blood-Brain Barrier Model. Mol. Pharm. 13, 4191–4198 (2016). 145. Lelu, S., Afadzi, M., Berg, S., Aslund, A., Torp, S., Sattler, W. & de L. Davies, C. Primary porcine brain endothelial cells as in vitro model to study effects of ultrasound on blood-brain barrier function. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 64, 1–1 (2016). 146. Maherally, Z., Fillmore, H. L., Tan, S. L., Tan, S. F., Jassam, S. A., Quack, F. I., Hatherell, K. E. & Pilkington, G. J. Real-time acquisition of transendothelial electrical resistance in an all-human, in vitro , 3-dimensional, blood–brain barrier model exemplifies tight-junction integrity. FASEB J. fj.201700162R 153 (2017). doi:10.1096/fj.201700162R 147. Cecchelli, R., Aday, S., Sevin, E., Almeida, C., Culot, M., Dehouck, L., Coisne, C., Engelhardt, B., Dehouck, M. P. & Ferreira, L. A stable and reproducible human blood-brain barrier model derived from hematopoietic stem cells. PLoS One 9, (2014). 148. Hersom, M., Helms, H. C., Pretzer, N., Goldeman, C., Jensen, A. I., Severin, G., Nielsen, M. S., Holm, R. & Brodin, B. Transferrin receptor expression and role in transendothelial transport of transferrin in cultured brain endothelial monolayers. Mol. Cell. Neurosci. 76, 59–67 (2016). 149. Shi, D., Sun, L., Mi, G., Sheikh, L., Bhattacharya, S., Nayar, S. & Webster, T. J. Controlling ferrofluid permeability across the blood–brain barrier model. Nanotechnology 25, 075101 (2014). 150. Stins, M. F., Badger, J. & Sik Kim, K. Bacterial invasion and transcytosis in transfected human brain microvascular endothelial cells. Microb. Pathog. 30, 19–28 (2001). 151. Booth, R. & Kim, H. Characterization of a microfluidic in vitro model of the blood-brain barrier (μBBB). Lab Chip 12, 1784 (2012). 152. Partyka, P. P., Godsey, G. A., Galie, J. R., Kosciuk, M. C., Acharya, N. K., Nagele, R. G. & Galie, P. A. Mechanical stress regulates transport in a compliant 3D model of the blood-brain barrier. Biomaterials 115, 30–39 (2017). 153. Prabhakarpandian, B., Shen, M.-C., Nichols, J. B., Mills, I. R., Sidoryk- Wegrzynowicz, M., Aschner, M. & Pant, K. SyM-BBB: a microfluidic blood 154 brain barrier model. Lab Chip 13, 1093 (2013). 154. Cho, H., Seo, J. H., Wong, K. H. K., Terasaki, Y., Park, J., Bong, K., Arai, K., Lo, E. H. & Irimia, D. Three-Dimensional Blood-Brain Barrier Model for in vitro Studies of Neurovascular Pathology. Sci. Rep. 5, 15222 (2015). 155. Lager, S. & Powell, T. L. Regulation of nutrient transport across the placenta. Journal of Pregnancy 2012, 179827 (2012). 156. Knipp, G. T., Audus, K. L. & Soares, M. J. Nutrient transport across the placenta. Adv. Drug Deliv. Rev. 38, 41–58 (1999). 157. Tang, Z., Abrahams, V. M., Mor, G. & Guller, S. Placental Hofbauer cells and complications of pregnancy. Ann. N. Y. Acad. Sci. 1221, 103–108 (2011). 158. Blundell, C., Tess, E. R., Schanzer, A. S. R., Coutifaris, C., Su, E. J., Parry, S. & Huh, D. A microphysiological model of the human placental barrier. Lab Chip 16, 3065–3073 (2016). 159. Levkovitz, R., Zaretsky, U., Gordon, Z., Jaffa, A. J. & Elad, D. In vitro simulation of placental transport: Part I. Biological model of the placental barrier. Placenta 34, 699–707 (2013). 160. Levkovitz, R., Zaretsky, U., Jaffa, A. J., Hod, M. & Elad, D. In vitro simulation of placental transport: Part II. Glucose transfer across the placental barrier model. Placenta 34, 708–715 (2013). 161. Elad, D., Levkovitz, R., Jaffa, A. J., Desoye, G. & Hod, M. Have We Neglected the Role of Fetal Endothelium in Transplacental Transport? Traffic 15, 122–126 (2014). 162. Conings, S., Amant, F., Annaert, P. & Van Calsteren, K. Integration and 155 validation of the ex vivo human placenta perfusion model. J. Pharmacol. Toxicol. Methods 88, 25–31 (2017). 163. Grafmueller, S., Manser, P., Diener, L., Diener, P. A., Maeder-Althaus, X., Maurizi, L., Jochum, W., Krug, H. F., Buerki-Thurnherr, T., von Mandach, U. & Wick, P. Bidirectional transfer study of polystyrene nanoparticles across the placental barrier in an ex vivo human placental perfusion model. Environ. Health Perspect. 123, 1280–1286 (2015). 164. Poulsen, M. S., Rytting, E., Mose, T. & Knudsen, L. E. Modeling placental transport: Correlation of in vitro BeWo cell permeability and ex vivo human placental perfusion. Toxicol. Vitr. 23, 1380–1386 (2009). 165. Cartwright, L., Poulsen, M. S., Nielsen, H. M., Pojana, G., Knudsen, L. E., Saunders, M. & Rytting, E. In vitro placental model optimization for nanoparticle transport studies. Int. J. Nanomedicine 7, 497–510 (2012). 166. Huang, X., Luthi, M., Ontsouka, E. C., Kallol, S., Baumann, M. U., Surbek, D. V. & Albrecht, C. Establishment of a confluent monolayer model with human primary trophoblast cells: Novel insights into placental glucose transport. Mol. Hum. Reprod. 22, 442–456 (2016). 167. Ali, H., Kalashnikova, I., White, M. A., Sherman, M. & Rytting, E. Preparation, characterization, and transport of dexamethasone-loaded polymeric nanoparticles across a human placental in vitro model. Int. J. Pharm. 454, 149–157 (2013). 168. Correia Carreira, S., Walker, L., Paul, K. & Saunders, M. The toxicity, transport and uptake of nanoparticles in the in vitro BeWo b30 placental cell 156 barrier model used within NanoTEST. Nanotoxicology 9, 1–14 (2013). 169. Ikeda, K., Ueda, C., Yamada, K., Nakamura, A., Hatsuda, Y., Kawanishi, S., Nishii, S. & Ogawa, M. Carrier-mediated placental transport of cimetidine and valproic acid across differentiating JEG-3 cell layers. Pharmazie 70, 471–476 (2015). 170. Kloet, S. K., Walczak, A. P., Louisse, J., van den Berg, H. H. J., Bouwmeester, H., Tromp, P., Fokkink, R. G. & Rietjens, I. M. C. M. Translocation of positively and negatively charged polystyrene nanoparticles in an in vitro placental model. Toxicol. Vitr. 29, 1701–1710 (2015). 171. Lopalco, A., Ali, H., Denora, N. & Rytting, E. Oxcarbazepine-loaded polymeric nanoparticles: Development and permeability studies across in vitro models of the blood–brain barrier and human placental trophoblast. Int. J. Nanomedicine 10, 1985–1996 (2015). 172. Li, H., Rietjens, I. M. C. M., Louisse, J., Blok, M., Wang, X., Snijders, L. & van Ravenzwaay, B. Use of the ES-D3 cell differentiation assay, combined with the BeWo transport model, to predict relative in vivo developmental toxicity of antifungal compounds. Toxicol. Vitr. 29, 320–328 (2015). 173. Araújo, J. R., Pereira, A. C., Correia-Branco, A., Keating, E. & Martel, F. Oxidative stress induced by tert-butylhydroperoxide interferes with the placental transport of glucose: In vitro studies with BeWo cells. Eur. J. Pharmacol. 720, 218–226 (2013). 174. Bode, C. J., Jin, H., Rytting, E., Silverstein, P. S., Young, A. M. & Audus, K. L. In Vitro Models for Studying Trophoblast Transcellular Transport. Methods 157 Mol Med 122, 225–239 (2006). 175. Albekairi, N. A., Al-Enazy, S., Ali, S. & Rytting, E. Transport of digoxin- loaded polymeric nanoparticles across BeWo cells, an in vitro model of human placental trophoblast. Ther. Deliv. 6, 1325–1334 (2015). 176. Pattillo, R. A. & Gey, G. O. The Establishment of a Cell Line of Human Hormone-synthesizing Trophoblastic Cells in Vitro. Cancer Res. 28, 1231– 1236 (1968). 177. Wice, B., Menton, D., Geuze, H. & Schwartz, A. L. Modulators of cyclic AMP metabolism induce syncytiotrophoblast formation in vitro. Exp. Cell Res. 186, 306–316 (1990). 178. Kitano, T., Iizasa, H., Hwang, I.-W., Hirose, Y., Morita, T., Maeda, T. & Nakashima, E. Conditionally immortalized syncytiotrophoblast cell lines as new tools for study of the blood-placenta barrier. Biol. Pharm. Bull. 27, 753– 759 (2004). 179. Mannelli, C., Ietta, F., Avanzati, A. M., Skarzynski, D. & Paulesu, L. Biological Tools to Study the Effects of Environmental Contaminants at the Feto-Maternal Interface. Dose. Response. 13, 1559325815611902 (2015). 180. Kuo, C. Y., Guo, T., Cabrera-Luque, J., Arumugasaamy, N., Bracaglia, L., Garcia-Vivas, A., Santoro, M., Baker, H., Fisher, J. & Kim, P. Placental basement membrane proteins are required for effective cytotrophoblast invasion in a three-dimensional bioprinted placenta model. J. Biomed. Mater. Res. - Part A 106, 1476–1487 (2018). 181. Lee, J. S., Romero, R., Han, Y. M., Kim, H. C., Kim, C. J., Hong, J.-S. & Huh, 158 D. Placenta-on-a-chip: a novel platform to study the biology of the human placenta. J. Matern. Neonatal Med. 29, 1046–1054 (2016). 182. Miura, S., Sato, K., Kato-Negishi, M., Teshima, T. & Takeuchi, S. Fluid shear triggers microvilli formation via mechanosensitive activation of TRPV6. Nat. Commun. 6, 8871 (2015). 183. Song, D., Guo, J., Han, F., Zhang, W., Wang, Y. & Wang, Y. Establishment of an in vitro model of the human placental barrier by placenta slice culture and ussing chamber. Biosci. Biotechnol. Biochem. 77, 1030–4 (2013). 184. Takaku, T., Nagahori, H., Sogame, Y. & Takagi, T. Quantitative Structure– Activity Relationship Model for the Fetal–Maternal Blood Concentration Ratio of Chemicals in Humans. Biol. Pharm. Bull. 38, 930–934 (2015). 185. Zhang, Y.-H., Xia, Z., Yan, L. & Liu, S. Prediction of Placental Barrier Permeability: A Model Based on Partial Least Squares Variable Selection Procedure. Molecules 20, 8270–8286 (2015). 186. Plitman Mayo, R., Charnock-Jones, D. S., Burton, G. J. & Oyen, M. L. Three- dimensional modeling of human placental terminal villi. Placenta 43, 54–60 (2016). 187. Kohn, J. C., Zhou, D. W., Bordeleau, F., Zhou, A. L., Mason, B. N., Mitchell, M. J., King, M. R. & Reinhart-King, C. A. Cooperative effects of matrix stiffness and fluid shear stress on endothelial cell behavior. Biophys. J. 108, 471–478 (2015). 188. Kraning-Rush, C. M. & Reinhart-King, C. A. Controlling matrix stiffness and topography for the study of tumor cell migration. Cell Adhesion and Migration 159 6, 274–279 (2012). 189. Srinivasan, B., Kolli, A. R., Esch, M. B., Abaci, H. E., Shuler, M. L. & Hickman, J. J. TEER Measurement Techniques for In Vitro Barrier Model Systems. J. Lab. Autom. 20, 107–126 (2015). 190. GEHRKE, S. H., FISHER, J. P., PALASIS, M. & LUND, M. E. Factors Determining Hydrogel Permeability. Ann. N. Y. Acad. Sci. 831, 179–184 (2006). 191. Gaur, R., Mishra, L. & Gupta, S. K. Sen. Diffusion and Transport of Molecules In Living Cells. in Modelling and Simulation of Diffusive Processes 27–48 (Springer, Cham, 2014). doi:10.1007/978-3-319-05657-9 192. Crank, J. The Mathematics of Diffusion. (Oxford University Press, 1980). 193. Cussler, E. L. Mass Transfer in Fluid Systems. (Cambridge University Press, 2009). 194. PrabhuDas, M., Bonney, E., Caron, K., Dey, S., Erlebacher, A., Fazleabas, A., Fisher, S., Golos, T., Matzuk, M., McCune, J. M., Mor, G., Schulz, L., Soares, M., Spencer, T., Strominger, J., Way, S. S. & Yoshinaga, K. Immune mechanisms at the maternal-fetal interface: Perspectives and challenges. Nature Immunology 16, 328–334 (2015). 195. Schmidt, A., Morales-Prieto, D. M., Pastuschek, J., Fröhlich, K. & Markert, U. R. Only humans have human placentas: Molecular differences between mice and humans. J. Reprod. Immunol. 108, 65–71 (2015). 196. Georgiades, P., Fergyson-Smith, A. C. & Burton, G. J. Comparative developmental anatomy of the murine and human definitive placentae. 160 Placenta 23, 3–19 (2002). 197. Dilworth, M. R. & Sibley, C. P. Review: Transport across the placenta of mice and women. in Placenta 34, S34–S39 (W.B. Saunders, 2013). 198. Vähäkangas, K. & Myllynen, P. Drug transporters in the human blood- placental barrier. Br. J. Pharmacol. 158, 665–678 (2009). 199. Hutson, J. R., Garcia-Bournissen, F., Davis, A. & Koren, G. The human placental perfusion model: A systematic review and development of a model to predict in vivo transfer of therapeutic drugs. Clinical Pharmacology and Therapeutics 90, 67–76 (2011). 200. Blundell, C., Yi, Y. S., Ma, L., Tess, E. R., Farrell, M. J., Georgescu, A., Aleksunes, L. M. & Huh, D. Placental Drug Transport-on-a-Chip: A Microengineered In Vitro Model of Transporter-Mediated Drug Efflux in the Human Placental Barrier. Adv. Healthc. Mater. 7, 1700786 (2018). 201. Mlakar, J. . J., Korva, M. . M., Tul, N. . N., Popović, M. M. ., Poljšak-Prijatelj, M. M. ., Mraz, J. . J., Kolenc, M. . M., Rus, K. R. ., Vipotnik, T. V. ., Vodušek, V. F. ., Vizjak, A. . A., Pižem, J. . J., Petrovec, M. . M., Zupanc, T. A. ., Resman Rus, K., Vesnaver Vipotnik, T., Fabjan Vodušek, V., Vizjak, A. . A., Pižem, J. . J., et al. Zika Virus Associated with Microcephaly. N. Engl. J. Med. 374, 951–8 (2016). 202. Tabata, T., Petitt, M., Puerta-Guardo, H., Michlmayr, D., Wang, C., Fang- Hoover, J., Harris, E. & Pereira, L. Zika Virus Targets Different Primary Human Placental Cells, Suggesting Two Routes for Vertical Transmission. Cell Host Microbe 20, 155–166 (2016). 161 203. Miner, J. J., Cao, B., Govero, J., Smith, A. M., Fernandez, E., Cabrera, O. H., Garber, C., Noll, M., Klein, R. S., Noguchi, K. K., Mysorekar, I. U. & Diamond, M. S. Zika Virus Infection during Pregnancy in Mice Causes Placental Damage and Fetal Demise. Cell 165, 1081–1091 (2016). 204. Leach, L. The phenotype of the human materno-fetal endothelial barrier: Molecular occupancy of paracellular junctions dictate permeability and angiogenic plasticity. J. Anat. 200, 599–606 (2002). 205. Blehar, M. C., Spong, C., Grady, C., Goldkind, S. F., Sahin, L. & Clayton, J. A. Enrolling Pregnant Women: Issues in Clinical Research. Women’s Heal. Issues 23, e39-45 (2013). 206. Gedeon, C. & Koren, G. Designing Pregnancy Centered Medications: Drugs Which Do Not Cross the Human Placenta. Placenta 27, 861–868 (2006). 207. Brennand, K. J., Simone, A., Jou, J., Gelboin-Burkhart, C., Tran, N., Sangar, S., Li, Y., Mu, Y., Chen, G., Yu, D., McCarthy, S., Sebat, J. & Gage, F. H. Modelling schizophrenia using human induced pluripotent stem cells. Nature 473, 221–225 (2011). 208. Kuo, C. Y., Eranki, A., Placone, J. K., Rhodes, K. R., Aranda-Espinoza, H., Fernandes, R., Fisher, J. P. & Kim, P. C. W. Development of a 3D Printed, Bioengineered Placenta Model to Evaluate the Role of Trophoblast Migration in Preeclampsia. ACS Biomater. Sci. Eng. 2, 1817–1826 (2016). 209. Garcez, P. P., Nascimento, J. M., de Vasconcelos, J. M., Madeiro da Costa, R., Delvecchio, R., Trindade, P., Loiola, E. C., Higa, L. M., Cassoli, J. S., Vitória, G., Sequeira, P. C., Sochacki, J., Aguiar, R. S., Fuzii, H. T., de Filippis, A. M. 162 B., da Silva Gonçalves Vianez Júnior, J. L., Tanuri, A., Martins-de-Souza, D. & Rehen, S. K. Zika virus disrupts molecular fingerprinting of human neurospheres. Sci. Rep. 7, 40780 (2017). 210. Delvecchio, R., Higa, L., Pezzuto, P., Valadão, A., Garcez, P., Monteiro, F., Loiola, E., Dias, A., Silva, F., Aliota, M., Caine, E., Osorio, J., Bellio, M., O’Connor, D., Rehen, S., de Aguiar, R., Savarino, A., Campanati, L. & Tanuri, A. Chloroquine, an Endocytosis Blocking Agent, Inhibits Zika Virus Infection in Different Cell Models. Viruses 8, 322 (2016). 211. Faye, O., Faye, O., Diallo, D., Diallo, M., Weidmann, M. & Sall, A. Quantitative real-time PCR detection of Zika virus and evaluation with field- caught Mosquitoes. Virol. J. 10, 311 (2013). 212. Henchal, E. A., Gentry, M. K., McCown, J. M. & Brandt, W. E. Dengue virus- specific and flavivirus group determinants identified with monoclonal antibodies by indirect immunofluorescence. Am. J. Trop. Med. Hyg. 31, 830– 836 (1982). 213. Costa, M. A. The endocrine function of human placenta: An overview. Reproductive BioMedicine Online 32, 14–43 (2016). 214. Baba, U., Ashir, G., Mava, Y., Elechi, H., Saidu, G. & Kaleb, A. Transplacental transfer of macromolecules: Proving the efficiency of placental transfer of maternal measles antibodies in mother: Infant pairs. Ann. Med. Health Sci. Res. 4, 298 (2014). 215. Palmeira, P., Quinello, C., Silveira-Lessa, A. L., Zago, C. A. & Carneiro- Sampaio, M. IgG placental transfer in healthy and pathological pregnancies. 163 Clinical and Developmental Immunology 2012, 985646 (2012). 216. Li, C., Zhu, X., Ji, X., Quanquin, N., Deng, Y. Q., Tian, M., Aliyari, R., Zuo, X., Yuan, L., Afridi, S. K., Li, X. F., Jung, J. U., Nielsen-Saines, K., Qin, F. X. F., Qin, C. F., Xu, Z. & Cheng, G. Chloroquine, a FDA-approved Drug, Prevents Zika Virus Infection and its Associated Congenital Microcephaly in Mice. EBioMedicine 24, 189–194 (2017). 217. Cao, B., Parnell, L. A., Diamond, M. S. & Mysorekar, I. U. Inhibition of autophagy limits vertical transmission of Zika virus in pregnant mice. J. Exp. Med. 214, 2303–2313 (2017). 218. Souza, B. S. F., Sampaio, G. L. A., Pereira, C. S., Campos, G. S., Sardi, S. I., Freitas, L. A. R., Figueira, C. P., Paredes, B. D., Nonaka, C. K. V., Azevedo, C. M., Rocha, V. P. C., Bandeira, A. C., Mendez-Otero, R., dos Santos, R. R. & Soares, M. B. P. Zika virus infection induces mitosis abnormalities and apoptotic cell death of human neural progenitor cells. Sci. Rep. 6, 39775 (2016). 219. Honein, M. A., Dawson, A. L., Petersen, E. E., Jones, A. M., Lee, E. H., Yazdy, M. M., Ahmad, N., Macdonald, J., Evert, N., Bingham, A., Ellington, S. R., Shapiro-Mendoza, C. K., Oduyebo, T., Fine, A. D., Brown, C. M., Sommer, J. N., Gupta, J., Cavicchia, P., Slavinski, S., et al. Birth Defects Among Fetuses and Infants of US Women With Evidence of Possible Zika Virus Infection During Pregnancy. JAMA 317, 59 (2017). 220. Bayer, A., Lennemann, N. J., Ouyang, Y., Bramley, J. C., Morosky, S., Marques, E. T. D. A., Cherry, S., Sadovsky, Y. & Coyne, C. B. Type III 164 Interferons Produced by Human Placental Trophoblasts Confer Protection against Zika Virus Infection. Cell Host Microbe 19, 705–712 (2016). 221. Corry, J., Arora, N., Good, C. A., Sadovsky, Y. & Coyne, C. B. Organotypic models of type III interferon-mediated protection from Zika virus infections at the maternal–fetal interface. Proc. Natl. Acad. Sci. 114, 201707513 (2017). 222. Yockey, L. J., Jurado, K. A., Arora, N., Millet, A., Rakib, T., Milano, K. M., Hastings, A. K., Fikrig, E., Kong, Y., Horvath, T. L., Weatherbee, S., Kliman, H. J., Coyne, C. B. & Iwasaki, A. Type I interferons instigate fetal demise after Zika virus infection. Sci. Immunol. 3, eaao1680 (2018). 223. Nguyen, S. M., Antony, K. M., Dudley, D. M., Kohn, S., Simmons, H. A., Wolfe, B., Salamat, M. S., Teixeira, L. B. C., Wiepz, G. J., Thoong, T. H., Aliota, M. T., Weiler, A. M., Barry, G. L., Weisgrau, K. L., Vosler, L. J., Mohns, M. S., Breitbach, M. E., Stewart, L. M., Rasheed, M. N., et al. Highly efficient maternal-fetal Zika virus transmission in pregnant rhesus macaques. PLoS Pathog. 13, e1006378 (2017). 224. Liu, S., Delalio, L. J., Isakson, B. E. & Wang, T. T. AXL-Mediated Productive Infection of Human Endothelial Cells by Zika Virus. Circ. Res. 119, 1183– 1189 (2016). 225. Rosenberg, A. Z., Weiying, Y., Hill, D. A., Reyes, C. A. & Schwartz, D. A. Placental pathology of zika virus: Viral infection of the placenta induces villous stromal macrophage (Hofbauer Cell) proliferation and hyperplasia. Arch. Pathol. Lab. Med. 141, 43–48 (2017). 226. Sinclair, S. M., Miller, R. K., Chambers, C. & Cooper, E. M. Medication 165 Safety During Pregnancy: Improving Evidence-Based Practice. Journal of Midwifery and Women’s Health 61, 52–67 (2016). 227. Cooper, W. O., Willy, M. E., Pont, S. J. & Ray, W. A. Increasing use of antidepressants in pregnancy. Am. J. Obstet. Gynecol. 196, 544.e1-544.e5 (2007). 228. Rampono, J., Simmer, K., Ilett, K. F., Hackett, L. P., Doherty, D. A., Elliot, R., Kok, C. H., Coenen, A. & Forman, T. Placental transfer of SSRI and SNRI antidepressants and effects on the neonate. Pharmacopsychiatry 42, 95–100 (2009). 229. Baum, A. L. & Misri, S. Selective Serotonin-Reuptake Inhibitors in Pregnancy and Lactation. Harv. Rev. Psychiatry 4, 117–125 (1996). 230. Boukhris, T., Sheehy, O., Mottron, L. & Bérard, A. Antidepressant Use During Pregnancy and the Risk of Autism Spectrum Disorder in Children. JAMA Pediatrics 170, 1 (2015). 231. Chittaranjan, A., Chethan, K. B. & Sandarsh, S. Cardiovascular mechanisms of SSRI drugs and their benefits and risks in ischemic heart disease and heart failure. Int. Clin. Psychopharmacol. 28, 145–55 (2013). 232. Morrison, J. L., Riggs, K. W. & Rurak, D. W. Fluoxetine during pregnancy: Impact on fetal development. Reproduction, Fertility and Development 17, 641–650 (2005). 233. Haskell, S. E., Hermann, G. M., Reinking, B. E., Volk, K. A., Peotta, V. A., Zhu, V. & Roghair, R. D. Sertraline exposure leads to small left heart syndrome in adult mice. Pediatr. Res. 73, 286–293 (2013). 166 234. Wozniak, G., Toska, A., Saridi, M. & Mouzas, O. Serotonin reuptake inhibitor antidepressants (SSRIs) against atherosclerosis. Med. Sci. Monit. 17, RA205- RA214 (2011). 235. Ofek, K., Schoknecht, K., Melamed-Book, N., Heinemann, U., Friedman, A. & Soreq, H. Fluoxetine induces vasodilatation of cerebral arterioles by co- modulating NO/muscarinic signalling. J. Cell. Mol. Med. 16, 2736–2744 (2012). 236. Kapoor, A., Iqbal, M., Petropoulos, S., Ho, H. L., Gibb, W. & Matthews, S. G. Effects of Sertraline and Fluoxetine on P-Glycoprotein at Barrier Sites: In Vivo and In Vitro Approaches. PLoS One 8, 3–8 (2013). 237. Hantsoo, L., Czarkowski, K. a, Child, J., Howes, C. & Epperson, C. N. Selective serotonin reuptake inhibitors and endothelial function in women. J. Womens. Health (Larchmt). 23, 613–8 (2014). 238. Lekakis, J., Ikonomidis, I., Papoutsi, Z., Moutsatsou, P., Nikolaou, M., Parissis, J. & Kremastinos, D. T. Selective serotonin re-uptake inhibitors decrease the cytokine-induced endothelial adhesion molecule expression, the endothelial adhesiveness to monocytes and the circulating levels of vascular adhesion molecules. Int. J. Cardiol. 139, 150–158 (2010). 239. Bhuiyan, M., Petropoulos, S., Gibb, W. & Matthews, S. G. Sertraline Alters Multidrug Resistance Phosphoglycoprotein Activity in the Mouse Placenta and Fetal Blood-Brain Barrier. Reprod. Sci. 19, 407–415 (2012). 240. Clabault, H., Flipo, D., Guibourdenche, J., Fournier, T., Sanderson, J. T. & Vaillancourt, C. Effects of selective serotonin-reuptake inhibitors (SSRIs) on 167 human villous trophoblasts syncytialization. Toxicol. Appl. Pharmacol. 349, 8– 20 (2018). 241. Clabault, H., Cohen, M., Vaillancourt, C. & Sanderson, J. T. Effects of selective serotonin-reuptake inhibitors (SSRIs) in JEG-3 and HIPEC cell models of the extravillous trophoblast. Placenta 72–73, 62–73 (2018). 242. Young, A. M., Allen, C. E. & Audus, K. L. Efflux transporters of the human placenta. Adv. Drug Deliv. Rev. 55, 125–132 (2003). 243. Arumugasaamy, N., Ettehadieh, L. E., Kuo, C. Y., Paquin-Proulx, D., Kitchen, S. M., Santoro, M., Placone, J. K., Silveira, P. P., Aguiar, R. S., Nixon, D. F., Fisher, J. P. & Kim, P. C. W. Biomimetic Placenta-Fetus Model Demonstrating Maternal–Fetal Transmission and Fetal Neural Toxicity of Zika Virus. Ann. Biomed. Eng. 46, 1963–1974 (2018). 244. Wichman, C. L., Moore, K. M., Lang, T. R., St Sauver, J. L., Heise, R. H. & Watson, W. J. Congenital heart disease associated with selective serotonin reuptake inhibitor use during pregnancy. Mayo Clin. Proc. 84, 23–7 (2009). 245. Malm, H., Sourander, A., Gissler, M., Gyllenberg, D., Hinkka-Yli-Salomäki, S., McKeague, I. W., Artama, M. & Brown, A. S. Pregnancy complications following prenatal exposure to SSRIs or maternal psychiatric disorders: Results from population-based national register data. Am. J. Psychiatry 172, 1224–1232 (2015). 246. Lopez-Vilchez, I., Diaz-Ricart, M., Navarro, V., Torramade, S., Zamorano- Leon, J., Lopez-Farre, A., Galan, A. M., Gasto, C. & Escolar, G. Endothelial damage in major depression patients is modulated by SSRI treatment, as 168 demonstrated by circulating biomarkers and an in vitro cell model. Transl. Psychiatry 6, e886 (2016). 247. Goksu Erol, A. Y., Nazli, M. & Yildiz, S. E. Significance of platelet endothelial cell adhesion molecule-1 (PECAM-1) and intercellular adhesion molecule-1 (ICAM-1) expressions in preeclamptic placentae. Endocrine 42, 125–131 (2012). 248. Labarrere, C. A., Ortiz, M. A., Sosa, M. J., Campana, G. L., Wernicke, M., Baldridge, L. A., Terry, C. & DiCarlo, H. L. Syncytiotrophoblast intercellular adhesion molecule-1 expression in placental villitis of unknown cause. Am. J. Obstet. Gynecol. 193, 483–488 (2005). 249. Lee, K. M. & Kim, Y. K. The role of IL-12 and TGF-β1 in the pathophysiology of major depressive disorder. Int. Immunopharmacol. 6, 1298–1304 (2006). 250. Tossetta, G., Paolinelli, F., Avellini, C., Salvolini, E., Ciarmela, P., Lorenzi, T., Emanuelli, M., Toti, P., Giuliante, R., Gesuita, R., Crescimanno, C., Voltolini, C., Di Primio, R., Petraglia, F., Castellucci, M. & Marzioni, D. IL-1β and TGF-β weaken the placental barrier through destruction of tight junctions: An in vivo and in vitro study. Placenta 35, 509–516 (2014). 251. Shea, A. K., Oberlander, T. F. & Rurak, D. Fetal serotonin reuptake inhibitor antidepressant exposure: Maternal and fetal factors. Can. J. Psychiatry 57, 523–529 (2012). 252. Kelly, J. J., Moore, T. M., Babal, P., Diwan, A. H., Stevens, T. & Thompson, W. J. Pulmonary microvascular and macrovascular endothelial cells: 169 differential regulation of Ca2+ and permeability. Am. J. Physiol. 274, L810-9 (1998). 253. Lang, I., Pabst, M. A., Hiden, U., Blaschitz, A., Dohr, G., Hahn, T. & Desoye, G. Heterogeneity of microvascular endothelial cells isolated from human term placenta and macrovascular umbilical vein endothelial cells. Eur. J. Cell Biol. 82, 163–73 (2003). 254. O’Brien, F. E., Dinan, T. G., Griffin, B. T. & Cryan, J. F. Interactions between antidepressants and P-glycoprotein at the blood-brain barrier: Clinical significance of in vitro and in vivo findings. British Journal of Pharmacology 165, 289–312 (2012). 255. Rochat, B., Baumann, P. & Audus, K. L. Transport mechanisms for the antidepressant citalopram in brain microvessel endothelium. Brain Res. 831, 229–236 (1999). 256. Ewing, G., Tatarchuk, Y., Appleby, D., Schwartz, N. & Kim, D. Placental Transfer of Antidepressant Medications: Implications for Postnatal Adaptation Syndrome. Clin. Pharmacokinet. 54, 359–370 (2015). 257. Wang, J.-S., Newport, D. J., Stowe, Z. N., Donovan, J. L., Pennell, P. B. & DeVane, C. L. The emerging importance of transporter proteins in the psychopharmacological treatment of the pregnant patient. Drug Metab. Rev. 39, 723–46 (2007). 258. Sobrevia, L., Abarzúa, F., Nien, J. K., Salomón, C., Westermeier, F., Puebla, C., Cifuentes, F., Guzmán-Gutiérrez, E., Leiva, A. & Casanello, P. Review: Differential placental macrovascular and microvascular endothelial 170 dysfunction in gestational diabetes. Placenta 32, S159–S164 (2011). 259. Brouillet, S., Hoffmann, P., Benharouga, M., Salomon, A., Schaal, J. P., Feige, J. J. & Alfaidy, N. Molecular characterization of EG-VEGF-mediated angiogenesis: differential effects on icrovascular and macrovascular endothelial cells. Mol. Biol. Cell 21, 2832–2843 (2010). 260. Sackett, S. D., Tremmel, D. M., Ma, F., Feeney, A. K., Maguire, R. M., Brown, M. E., Zhou, Y., Li, X., O’Brien, C., Li, L., Burlingham, W. J. & Odorico, J. S. Extracellular matrix scaffold and hydrogel derived from decellularized and delipidized human pancreas. Sci. Rep. 8, 10452 (2018). 261. Chen, C.-P. & Aplin, J. D. Placental extracellular matrix: gene expression, deposition by placental fibroblasts and the effect of oxygen. Placenta 24, 316– 25 (2003). 262. Nooteboom, A., Hendriks, T., Ottehöller, I. & Van Der Linden, C. J. Permeability characteristics of human endothelial monolayers seeded on different extracellular matrix proteins. Mediators Inflamm. 9, 235–241 (2000). 263. Streuli, C. H., Schmidhauser, C., Kobrin, M., Bissell, M. J. & Derynck, R. Extracellular matrix regulates expression of the TGF-beta 1 gene. J. Cell Biol. 120, 253–60 (1993). 264. Hinz, B. The extracellular matrix and transforming growth factor-β1: Tale of a strained relationship. Matrix Biology 47, 54–65 (2015). 265. Roberts, A. B., McCune, B. K. & Sporn, M. B. TGF-β: Regulation of extracellular matrix. Kidney Int. 41, 557–559 (1992). 266. De Caterina, R., Basta, G., Lazzerini, G., Dell’Omo, G., Petrucci, R., Morale, 171 M., Carmassi, F. & Pedrinelli, R. Soluble vascular cell adhesion molecule-1 as a biohumoral correlate of atherosclerosis. Arterioscler. Thromb. Vasc. Biol. 17, 2646–2654 (1997). 267. Cybulsky, M. I., Iiyama, K., Li, H., Zhu, S., Chen, M., Iiyama, M., Davis, V., Gutierrez-Ramos, J. C., Connelly, P. W. & Milstone, D. S. A major role for VCAM-1, but not ICAM-1, in early atherosclerosis. J. Clin. Invest. 107, 1255– 1262 (2001). 268. Hwang, S. J., Ballantyne, C. M., Sharrett, A. R., Smith, L. C., Davis, C. E., Gotto, A. M. & Boerwinkle, E. Circulating adhesion molecules VCAM-1, ICAM-1, and E-selectin in carotid atherosclerosis and incident coronary heart disease cases: The Atherosclerosis Risk In Communities (ARIC) study. Circulation 96, 4219–4225 (1997). 269. Nezafati, M. H., Eshraghi, A., Vojdanparast, M., Abtahi, S. & Nezafati, P. Selective serotonin reuptake inhibitors and cardiovascular events: A systematic review. Journal of Research in Medical Sciences 21, 13–19 (2016). 270. Carmeliet, P. Mechanisms of angiogenesis and arteriogenesis. Nature Medicine 6, 389–395 (2000). 271. Chen, D. B. & Zheng, J. Regulation of Placental Angiogenesis. Microcirculation 21, 15–25 (2014). 272. Walshe, T. E., Saint-Geniez, M., Maharaj, A. S. R., Sekiyama, E., Maldonado, A. E. & D’Amore, P. A. TGF-β is required for vascular barrier function, endothelial survival and homeostasis of the adult microvasculature. PLoS One 4, e5149 (2009). 172 273. Ferrari, G., Cook, B. D., Terushkin, V., Pintucci, G. & Mignatti, P. Transforming growth factor-beta 1 (TGF-β1) induces angiogenesis through vascular endothelial growth factor (VEGF)-mediated apoptosis. J. Cell. Physiol. 219, 449–458 (2009). 274. Cheng, J. C., Chang, H. M. & Leung, P. C. K. Transforming growth factor-β1 inhibits trophoblast cell invasion by inducing snail-mediated down-regulation of vascular endothelial-cadherin protein. J. Biol. Chem. 288, 33181–33192 (2013). 275. Rothbauer, M., Patel, N., Gondola, H., Siwetz, M., Huppertz, B. & Ertl, P. A comparative study of five physiological key parameters between four different human trophoblast-derived cell lines. Sci. Rep. 7, 5892 (2017). 276. Marchocki, Z., Russell, N. E. & Donoghue, K. O. Selective serotonin reuptake inhibitors and pregnancy: A review of maternal, fetal and neonatal risks and benefits. Obstetric Medicine 6, 155–158 (2013). 277. Park, K. S., Kong, I. D., Park, K. C. & Lee, J. W. Fluoxetine Inhibits L-Type Ca2+and Transient Outward K+Currents in Rat Ventricular Myocytes. Yonsei Med. J. 40, 144–151 (1999). 278. Yekehtaz, H., Farokhnia, M. & Akhondzadeh, S. Cardiovascular considerations in antidepressant therapy: An evidence-based review. Journal of Tehran University Heart Center 8, 169–176 (2013). 279. Vaughan Williams, E. M. QT and action potential duration. Br. Heart J. 47, 513–514 (1982). 280. Wang, K., Terrar, D., Gavaghan, D. J., Mu-u-min, R., Kohl, P. & Bollensdorff, 173 C. Living cardiac tissue slices: An organotypic pseudo two-dimensional model for cardiac biophysics research. Prog. Biophys. Mol. Biol. 115, 314–327 (2014). 281. Kocylowski, R. D., Dubiel, M., Gudmundsson, S., Sieg, I., Fritzer, E., Alkasi, Ö., Breborowicz, G. H. & von Kaisenberg, C. S. Biochemical tissue-specific injury markers of the heart and brain in postpartum cord blood. Am. J. Obstet. Gynecol. 200, 273.e1-273.e25 (2009). 282. Neves, A. L., Cabral, M., Leite-Moreira, A., Monterroso, J., Ramalho, C., Guimarães, H., Barros, H., Guimarães, J. T., Henriques-Coelho, T. & Areias, J. C. Myocardial Injury Biomarkers in Newborns with Congenital Heart Disease. Pediatr. Neonatol. 57, 488–495 (2016). 283. Neves, A. L., Cabral, M., Leite-Moreira, A., Monterroso, J., Ramalho, C., Guimarães, H., Barros, H., Guimarães, J. T., Henriques-Coelho, T. & Areias, J. C. Time-dependence of cardiac biomarker levels in newborns with congenital heart defects: Umbilical cord versus peripheral newborn blood. Int. J. Cardiol. 214, 412–414 (2016). 284. Sugimoto, M., Kuwata, S., Kurishima, C., Kim, J. H., Iwamoto, Y. & Senzaki, H. Cardiac biomarkers in children with congenital heart disease. World Journal of Pediatrics 11, 309–315 (2015). 285. Kwee, L., Baldwin, H. S., Shen, H. M., Stewart, C. L., Buck, C., Buck, C. A. & Labow, M. A. Defective development of the embryonic and extraembryonic circulatory systems in vascular cell adhesion molecule (VCAM-1) deficient mice. Development 121, 489–503 (1995). 174 286. Dawood, T., Barton, D. A., Lambert, E. A., Eikelis, N. & Lambert, G. W. Examining endothelial function and platelet reactivity in patients with depression before and after SSRI therapy. Front. Psychiatry 7, (2016). 287. Moon, A. Mouse Models of Congenital Cardiovascular Disease. Current Topics in Developmental Biology 84, 171–248 (2008). 288. Conway, S. J., Kruzynska-Frejtag, A., Kneer, P. L., Machnicki, M. & Koushik, S. V. What cardiovascular defect does my prenatal mouse mutant have, and why? Genesis 35, 1–21 (2003). 289. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., Tinevez, J. Y., White, D. J., Hartenstein, V., Eliceiri, K., Tomancak, P. & Cardona, A. Fiji: An open-source platform for biological-image analysis. Nature Methods 9, 676–682 (2012). 290. Zhu, R., Millrod, M. A., Zambidis, E. T. & Tung, L. Variability of Action Potentials Within and among Cardiac Cell Clusters Derived from Human Embryonic Stem Cells. Sci. Rep. 6, 18544 (2016). 291. Iwamiya, T., Matsuura, K., Masuda, S., Shimizu, T. & Okano, T. Cardiac fibroblast-derived VCAM-1 enhances cardiomyocyte proliferation for fabrication of bioengineered cardiac tissue. Regen. Ther. 4, 92–102 (2016). 292. Bérard, A., Zhao, J. P. & Sheehy, O. Antidepressant use during pregnancy and the risk of major congenital malformations in a cohort of depressed pregnant women: An updated analysis of the Quebec Pregnancy Cohort. BMJ Open 7, e013372 (2017). 175 293. Crespo, L. M., Grantham, C. J. & Cannell, M. B. Kinetics, stoichiometry and role of the Na-Ca exchange mechanism in isolated cardiac myocytes. Nature 345, 618–621 (1990). 294. Lang, D., Holzem, K., Kang, C., Xiao, M., Hwang, H. J., Ewald, G. A., Yamada, K. A. & Efimov, I. R. Arrhythmogenic remodeling of β2 versus β1 adrenergic signaling in the human failing heart. Circ. Arrhythmia Electrophysiol. 8, 409–419 (2015). 295. Satin, J. & Schroder, E. A. Autoregulation of Cardiac L-type calcium channels. Trends in Cardiovascular Medicine 19, 268–271 (2009). 296. Splawski, I., Timothy, K. W., Decher, N., Kumar, P., Sachse, F. B., Beggs, A. H., Sanguinetti, M. C. & Keating, M. T. Severe arrhythmia disorder caused by cardiac L-type calcium channel mutations. Proc. Natl. Acad. Sci. 102, 8089– 8096 (2005). 297. Junus, K., Wikström, A. K., Larsson, A. & Olovsson, M. Placental expression of proBNP/NT-proBNP and plasma levels of NT-proBNP in early- and late- onset preeclampsia. Am. J. Hypertens. 27, 1225–1230 (2014). 298. Sadlecki, P., Grabiec, M. & Walentowicz-Sadlecka, M. Prenatal clinical assessment of NT-proBNP as a diagnostic tool for preeclampsia, gestational hypertension and gestational diabetes mellitus. PLoS One 11, e0162957 (2016). 299. Kaufmann, R. L., Antoni, H., Hennekes, R., Jacob, R., Kohlhardt, M. & Lab, M. J. Mechanical response of the mammalian myocardium to modifications of the action potential. Cardiovasc. Res. 5, 64–70 (1971). 176 300. Erta, M., Quintana, A. & Hidalgo, J. Interleukin-6, a major cytokine in the central nervous system. International Journal of Biological Sciences 8, 1254– 1266 (2012). 301. Fregnan, F., Muratori, L., Simões, A. R., Giacobini-Robecchi, M. G. & Raimondo, S. Role of inflammatory cytokines in peripheral nerve injury. Neural Regen. Res. 7, 2259–2266 (2012). 177