Quantification of Amino Acid Transporter Interactions

Zurich Open Repository and Archive University of Zurich Main Library Strickhofstrasse 39 CH-8057 Zurich www.zora.uzh.ch

Year: 2018

Quantification of amino acid transporter interactions

Taslimifar, Mehdi

Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-158149 Dissertation Published Version

Originally published at: Taslimifar, Mehdi. Quantification of amino acid transporter interactions. 2018, University of Zurich, Faculty of Science. Quantification of Amino Acid Transporter Interactions

Dissertation

zur

Erlangung der naturwissenschaftlichen Doktorwürde (Dr. sc. nat.)

vorgelegt der

Mathematisch-naturwissenschaftlichen Fakultät

der

Universität Zürich

von

Mehdi Taslimifar

aus

Iran

Promotionskommission

Prof. Dr. Vartan Kurtcuoglu (Vorsitz und Leitung der Dissertation)

Prof. Dr. François Verrey

Prof. Dr. Christian von Mering

Dr. Marianne Schmid Daners

Dr. Stefano Buoso

Zürich, 2018

Abstract

The homeostatic regulation of amino acid (AA) concentration is essential for cellular maintenance, growth, and function. AA passage into and out of cells is mediated by the complex interactions between various amino acid transporters (AATs) at cell membranes. While numerous in vitro and in vivo studies have been carried out to investigate mechanisms underlying the transport systems, there are still many fundamental questions about their physiologic function that remain challenging to be addressed by available techniques. In the framework of this thesis, we have combined as well as applied computational and experimental approaches to shed light on the control mechanisms of cellular amino acid transport pathways, and to advance our understanding of their cooperation. In particular, we first established a systems biology approach to quantify the contributions of specific players in a complex cellular network of AATs functioning at single membranes. This enables the prediction of trans-membrane transport in response to various extracellular conditions in the presence of competitive substrates and/or inhibitors. We used this approach to characterize L-leucine (Leu) and L-phenylalanine (Phe) transporter systems of the Xenopus laevis oocyte expression system. We then developed a new computational model based on a mathematical representation of the nonlinear mechanistic kinetics of AATs to investigate the dynamic interactions between large neutral amino acid (LNAA) transporters located at the neurovascular unit (NVU) cell membranes. We incorporated published in vivo LNAA microdialysis (MD) measurements performed in the rat brain upon intraperitoneal administration of L-tyrosine and L-phenylalanine into our model to evaluate existing hypotheses about LAT1-mediated trans- endothelial transport of LNAAs across the blood brain barrier (BBB). We thereby demonstrated that either strong asymmetric bi-directional kinetics of LAT1 in microvascular brain endothelial cells (MBECs) (lower affinity inside the MBECs) and/or an asymmetric distribution of LAT1 at both membranes of MBECs (lower expression at the abluminal membrane of the BBB) is required to reproduce in vivo observations. Finally, we employed our computational model to investigate whether and how the dynamics of LNAAs in the NVU might be affected by abnormal fluctuations of plasma Phe in phenylketonuria, which has so far not been addressable using standard experimental techniques. We showed that fluctuations of Phe concentration in the plasma can propagate across the BBB and change the dynamics of LNAAs, however, the induced effects vary between NVU individual compartments and they largely depend on indices of plasma Phe fluctuation such as mean and frequency for which we have provided a quantitative representation.

Zusammenfassung

Die homöostatische Regulierung der Konzentration von Aminosäuren (AS) ist essentiell für zellulären Unterhalt, Wachstum und Funktion. AS Ein- und Austritt aus Zellen wird hauptsächlich über komplexe Interaktionen zwischen verschiedenen Aminosäuretransportern (AST) an Zellmembranen vermittelt. Obwohl eine Vielzahl an in vitro und in vivo Studien durchgeführt wurden, um die den Transportsystemen zugrundeliegenden Mechanismen zu untersuchen, gibt es immer noch viele fundamentale Fragen über deren physiologische Funktion, welche mit den verfügbaren Techniken noch immer schwierig zu untersuchen bleiben. Im Rahmen dieser Doktorarbeit haben wir computergestützte und experimentelle Ansätze sowohl kombiniert als auch angewandt, um ein Licht auf die Kontrolle der zellulären Aminosäuretransportwege zu werfen und um unser Verständnis über ihre Kooperation zu erweitern. Im Detail etablierten wir zunächst einen Systembiologie- Ansatz, um die Beteiligungen der verschiedenen Akteure im komplexen zellulären Netzwerk von an einzelnen Membranen funktionierenden AST zu quantifizieren. Dies ermöglicht die Vorhersage von Transmembrantransport als Antwort auf verschiedene extrazelluläre Bedingungen in Anwesenheit von kompetitiven Substraten oder Inhibitoren. Wir benutzten diesen Ansatz, um die L-Leucin (Leu)- und L-Phenylalanin (Phe)-Transportersyteme im Xenopus laevis Oozyten-Expressionssystem zu charakterisieren. Wir entwickelten danach ein neues Computermodell basierend auf einer mathematischen Repräsentation der nichtlinearen mechanistischen Kinetik von AST, um die dynamische Interaktion zwischen den an den Zellmembranen der neurovaskulären Einheiten (NVE) befindenden Transportern von grossen, neutralen Aminosäuren (GNAS) zu untersuchen. Wir integrierten publizierte in vivo GNAS Mikrodialyse-Messungen (MD), die im Rattengehirn nach intraperitonealer Verabreichung von L-Tyrosin und L-Phenylalanin durchgeführt worden waren, in unser Model, um bestehende Hypothesen über LAT1-vermittelten Transendotheltransport von GNAS durch die Blut-Hirn-Schranke (BHS) zu evaluieren. Wir zeigten dadurch, dass entweder eine starke asymmetrische Kinetik von LAT1 in mikrovaskulären Hirnendothelzellen (MHEZ) (geringere Affinität innerhalb der MHEZ) und/oder eine asymmetrische Verteilung von LAT1 an beiden Membranen von MHEZ (geringere Exprimierung an der abluminalen Membran der BHS) nötig ist, um die in vivo-Beobachtungen zu reproduzieren. Zuletzt benutzten wir unser Computermodell, um zu untersuchen, ob und wie die Dynamiken der GNAS in den NVE von den abnormalen Fluktuationen des Blutserum-Phe bei Phenylketonurie beeinflusst werden könnten, was bisher nicht mit experimentellen Standardtechniken untersucht werden konnte. Wir zeigten, dass sich Fluktuationen der Phe- Konzentration im Blutserum durch die BHS hindurch fortpflanzen und die Dynamiken der GNAS verändern können, allerdings variieren die induzierten Effekte zwischen den individuellen Kompartimenten der NVE und hängen hauptsächlich von Kennzahlen der Blutserum-Phe-Fluktuation wie Durchschnitt und Frequenz ab, für die wir eine quantitative Repräsentation vorgegeben haben.

iii

Acknowledgements

First and foremost, I would like to express my sincere gratitude to my supervisor, Prof. Vartan Kurtcuoglu and my co-supervisor, Prof. François Verrey who gave me the opportunity to work in this interdisciplinary project, directed this research, and consistently supported me to develop myself as a researcher in the best possible way. Vartan and François, you have been tremendous mentors for me. I greatly appreciate my co-supervisors, Dr. Victoria Makrides, Dr. Ufuk Olgaç, and Dr. Stefano Buoso who provided invaluable guidance throughout this research and helped me with their fruitful comments and edits. Many thanks to my thesis committee members, Prof. Christian von Mering and Dr. Marianne Schmid Daners for the constructive discussions and their valuable suggestions for this project.

I would like to thank all my colleagues and friends at The Interface Group and The Epithelial Transport Group, Dr. Mahdi Asgari, Dr. Virginia Meskenaite, Dr. Anastasios Marmaras, Dr. Diane de Zélicourt, Dr. Kartik Jain, Dr. Andreas Spiegelberg, Dr. Claudia Danzer, Dr. David Granjon, Dr. Manuel Gehlen, Mr. Willy Kuo, Ms. Lena Wiegmann, Mr. Norman Juchler, Ms. Helen Girard, Dr. Elena Dolgodilina, Ms. Lalita Oparija, Ms. Eva Hänsenberger, Dr. Simone Camargo, Dr. Ian Forster, Dr. Monica Patti, Dr. Samyuktha M. Pillai, Dr. Nadège Poncet, Dr. Christian Feuerstacke, Ms. Anuradha Rajendran, Ms. Brigitte Herzog, Ms. Eva Roth, Ms. Jin Chen, Dr. Evelyne Kuster, Dr. Julia Jando, Dr. Emilia Boiadjieva, Dr. Liviu Vanoaica and all members of J-floor for creating a friendly and productive work environment.

I would like to take this opportunity to express profound gratitude from the bottom of my heart to my beloved parents, parents-in-law and also to my sweet little brother, sister and sister-in-law for their love and continuous support through this journey. Most importantly, my special thanks to my lovely wife who has been extremely supportive throughout this entire process and has made countless sacrifices to help me get to this point. Thank you Niloo, I could never have accomplished this without your love, patience, and support.

Finally, I am very grateful to SystemsX.ch - The Swiss Initiative in Systems Biology for the financial support of this project and also to my PhD program, Systems Biology, for providing interesting courses, symposiums, and retreats.

Contents ABSTRACT ...... II

ZUSAMMENFASSUNG ...... III

ACKNOWLEDGEMENTS ...... IV

CHAPTER 1 INTRODUCTION ...... 7

AMINO ACID TRANSPORT ...... 7 ...... 10 EXPERIMENTAL APPROACHES FOR STUDYING AMINO ACID TRANSPORT NETWORKS ...... 10 MODELING OF TRANS-MEMBRANE AMINO ACID TRANSPORT PROCESSES ...... 11 THE COMBINED USE OF IN SILICO, IN VITRO AND IN VIVO APPROACHES TO STUDY AMINO ACID TRANSPORT NETWORKS ...... 11 QUANTIFICATION OF SINGLE CELL AMINO ACID TRANSPORT NETWORK THROUGH A COMBINED IN SILICO AND IN VITRO APPROACH ...... 12 QUANTIFICATION OF MULTI-CELL AMINO ACID TRANSPORT NETWORK THROUGH AN INTEGRATED IN SILICO AND IN VIVO APPROACH...... 13 INVESTIGATION OF PERTURBED AA TRANSPORT PATHWAYS IN PHENYLKETONURIA DISEASE ...... 15 CHAPTER 2 QUANTIFYING THE RELATIVE CONTRIBUTIONS OF DIFFERENT SOLUTE CARRIERS TO AGGREGATE SUBSTRATE TRANSPORT ...... 16

ABSTRACT ...... 16 INTRODUCTION ...... 17 METHODS ...... 20 RESULTS ...... 28 DISCUSSION ...... 35 SUMMARY ...... 38 SUPPLEMENTARY INFORMATION ...... 39 CHAPTER 3 FUNCTIONAL POLARITY OF MICROVASCULAR BRAIN ENDOTHELIAL CELLS SUPPORTED BY NEUROVASCULAR UNIT COMPUTATIONAL MODEL OF LARGE NEUTRAL AMINO ACID HOMEOSTASIS ...... 39

ABSTRACT ...... 40 INTRODUCTION ...... 40 METHODS ...... 43 RESULTS ...... 47 DISCUSSION ...... 59 SUMMARY ...... 61 SUPPLEMENTARY INFORMATION ...... 63 CHAPTER 4 PROPAGATION OF PLASMA L-PHENYLALANINE CONCENTRATION FLUCTUATIONS TO THE NEUROVASCULAR UNIT IN PHENYLKETONURIA: AN IN SILICO STUDY...... 69

ABSTRACT ...... 69 INTRODUCTION ...... 70 METHODS ...... 72 RESULTS ...... 76 DISCUSSION ...... 82 SUMMARY ...... 83 SUPPLEMENTARY INFORMATION ...... 85 CHAPTER 5 CONCLUSION AND OUTLOOK ...... 89

CONCLUSION ...... 89 v

OUTLOOK...... 91 REFERENCES ...... 94

LIST OF PUBLICATIONS ...... 103

CURRICULUM VITAE ...... 104

Chapter 1 Introduction Amino acid transport

Amino acids (AA) are essential molecules in the synthesis of proteins and play crucial roles in energy, neurotransmitter, and other essential metabolic pathways whose homeostasis is tightly maintained [1]. There are twenty proteogenic AAs characterized with common chemical structure with both a carboxyl and an amino group attached to a central carbon (α-Carbon), but with the difference in the side group (side chain) attached to the α-Carbon (Fig. 1.1) [2]. AAs have traditionally been categorized into groups of essential, nonessential and conditionally essential. The nine essential AAs (L-phenylalanine (Phe), L-tryptophan (Trp), L-leucine (Leu), L-isoleucine (Ile), L-lysine (Lys), L-valine (Val), L-methionine (Met), L-threonine (Thr), and L-histidine (His)) must be taken from foods. The five nonessential AAs (L-alanine (Ala), L-aspartic acid (Asp), L-asparagine (Asn), and L-serine (Ser) and L-glutamic acid (Glu)) can be synthesized in the body. On the other hand, the synthesis of six conditionally essential AA (L-arginine (Arg), L-cysteine (Cys), L-glycine (Gly), L-glutamine (Gln), L-proline (Pro), and L-tyrosine (Tyr)) is limited under specific pathophysiological conditions.

Carboxylic acid group Amino group

H H O

N C C H O

H α-Carbon Side chain group

Figure 1. 1. Basic structure of an amino acid. AAs consist of an amino group, a carboxylic acid group, a hydrogen atom (H) and a side chain (R group) attached to the central alpha carbon (α-Carbon) that differs between the AAs.

In general, since AAs cannot freely diffuse into or out of the cells, specialized membrane-spanning proteins, the so-called solute carriers (SLCs), have evolved to

mediate their movements across barrier membranes [3]. More than 75 assigned SLC proteins are characterized as amino acid transporters (AATs) [4]. AATs display a broad substrate selectivity and have often been named according to their first or most important characterized AA substrate. For example, L-type AATs (i.e., LAT1/SLC7A5 and LAT2/SLC7A8) prefer large neutral amino acids (LNAAs) (i.e., Leu, Tyr, Phe, etc.) or ASC-type AATs (i.e., ASCT1/SLC1A4 and ASCT2/SLC1A5) consider Ala, Ser, and Cys as substrate. In addition, AATs have been classified into three groups based on their mechanism: Uniporters that mediate the transfer of AAs with the direction of their concentration gradient, antiporters that exchange AAs for others across the membrane and symporters that cotransport AAs together with ions along the ions’ electrochemical gradient (i.e. sodium-dependent symporter) (Fig. 1.2). The functional cooperation between different uniporters, antiporters and symporters with overlapping substrate specificities constitutes an integrated network at single cell level in which one AA can inhibit and/or facilitate the movements of the others (Fig. 1.2). An overview of the epithelial AATs with the information about their substrate specificity, mechanism, and ion dependency is shown in Table 1.1 [1].

Amino acids

Uniporter

Figure 1. 2. Diagram of the interactions between different classes of amino acid transporters. The diagram represents the underlying transport mechanisms for uniporter, antiporter and symporters and their interactions at single cell level.

Table 1.1 Overview of epithelial amino acid transport systems

Type cDNA SLC Amino Acid Substrates Mechanism Ions Gly,Pro,Ala,Ser,Cys,Gln,Asn, A SNAT2 SLC38A2 His,Met Symporter Na+ Gly, Ala,Ser,Cys, Gln, Asn, SNAT4 SLC38A4 Met, AA+ Symporter Na+ ASC ASCT1 SLC1A4 Ala,Ser,Cys Antiporter Na+ ASCT2 SLC1A5 Ala,Ser,Cys,Thr, Gln Antiporter Na+ asc 4F2 hc/asc1 SLC3A2/SLC7A10 Gly,Ala,Ser,Cys,Thr Antiporter B0 B0AT1 SLC6A19 AA0 Symporter Na+ B0AT2 SLC6A15 Pro, Leu, Val, Ile, Met Symporter Na+ 0,+ 0, 0 + + - B ATB SLC6A14 AA , AA , b-Ala Symporter Na , Cl 0,+ 0,+ b rBAT/b AT SLC3A1/SLC7A9 Arg,Lys,O,cystine Antiporter b TauT SLC6A6 Tau, b-Ala Symporter Na+ , Cl- Gly XT2 SLC6A18 Gly NR NR IMINO IMINO SLC6A20 Pro, HO-Pro Symporter Na+ , Cl- His,Met,Leu, L 4F2hc/LAT1 SLC3A2/SLC7A5 Ile,Val,Phe,Tyr,Trp Antiporter 4F2hc/LAT2 SLC3A2/SLC7A8 AA0 except Pro Antiporter LAT3 SLC43A1 Leu, Ile, Met,Phe Uniporter LAT4 SLC43A2 Leu, Ile, Met, Phe Uniporter N SNAT3 SLC38A3 Gln, Asn, His Symporter Na+(Symporter),H(antiporter) SNAT5 SLC38A5 Gln Asn, His,Ser, Gly Symporter Na+(Symporter),H(antiporter) PAT (Imino acid) PAT1 SLC36A1 Pro, Gly,Ala GABA, b-Ala Symporter H+ PAT2 SLC36A2 Pro, Gly,Ala Symporter H+ T TAT1 SLC16A10 Phe, Tyr, Trp Uniporter X AG EAAT2 SLC1A2 Glu, Asp Symporter Na+ , H+ (symporter), K+ (antiporter) EAAT3 SLC1A1 Glu, Asp Symporter Na+,H+ (symporter), K+ (antiporter) x- c 4F2 hc/xCT SLC3A2/SLC7A11 Glu, Cys Antiporter y+ CAT-1 SLC7A1 Arg,Lys,O, His Uniporter 4F2hc/y+LAT1 SLC3A2/SLC7A7 Lys, Arg,Gln, His,Met,Leu Antiporter Na+ (symporter with AA0) + y L Lys, Arg, Gln, 4F2hc/y+LAT2 SLC3A2/SLC7A6 His,Met,Leu,Ala,Cys Antiporter Na+ (symporter with AA0) Adapted from Broer [1]. NR, not reported, AA0, neutral amino acids; AA neutral amino acids; AA+, cationic amino acids. O, ornithine; HO-Pro, hydroxyproline The AA transport network can become more difficult to understand when it comes into the interplay between transporters at multi-cell level (Fig. 1.3). These complexities arise from the fact that multiple AATs expressed at different cell membranes can be involved in the process of transporting a single AA. Characterizing the functional role of individual transporters within networks is necessary to understand AA transport related dysfunctions and to design appropriate treatment strategies.

Amino acids

Figure 1. 3. Diagram of amino acid transport network at multi-cell level. The diagram represents different types of AAs and AATs that interact across multiple cell membranes

Experimental approaches for studying amino acid transport networks The mammalian AATs are low-abundance membrane proteins whose function are affected by post-translational modifications. Hence, it can be a difficult task to isolate them in abundant quantities and identify their function in an active state. Over the last decades, the emergence of technologies such as radiotracer techniques, analytical quantification techniques (i.e. paper and liquid chromatography), DNA cloning and cellular overexpression methods, and also the advances in determining AATs structure has allowed researchers to gain more insights into the function of AATs [2, 5]. A large number of experimental approaches have been conducted to study physiological function of mammalian AA transport systems. These approaches investigate the role of endogenously expressed AATs using in vivo animal models and/or ex vivo organs/tissues/cells models and/or they characterize the activity of heterogeneously expressed AATs using in vitro models such as Xenopus laevis oocyte [6-11]. Although in vitro approaches provide the possibility of a direct investigation of AATs function as compared to the ex vivo and in vivo assays, they might not reflect the in vivo situation due to the high-level sensitivity of some AATs to the culture conditions. On the other hand, the in vivo and ex vivo experiments typically provide an overall picture of the AA transport across the tissues or organs. Furthermore, the intrinsic overlapping substrate 10

specificity of multiple AATs present in the biological sample could be of a critical challenge for data interpretation in all approaches. For example, the interference from endogenous AATs can provide significant noise for the investigation of exogenous AATs using Xenopus laevis oocyte expression system. The review by Souba et al. discusses the strengths and limitations of different approaches [12]. Modeling of trans-membrane amino acid transport processes The process of transport of a single amino acid by a simple uniporter transporter consists of four key steps: association of AA with the AAT and construction of AA-AAT complex, conformational change of the complex from one side to the other side of the membrane, dissociation of AA from the complex, and eventually reverse conformational change of the unbound carrier to its initial position. In general, trans- membrane transport processes can be represented by the law of mass action as applied to enzymatic reactions [13, 14]. The Michaelis-Menten rate law (MMRL) [15] and its variations [16] are the most widely used forms of the mass-action law after being simplified using steady state approximations [17]. The MMRL assumes a rapid binding of the substrate (AA) to the transporter (rapid equilibrium approximation), it considers concentration of transporter to be negligible as compared to the AA concentrates (free ligand approximation), and it holds where the formation of transported AA is linear with time and also the concentration of AA-AAT complex remains constant (steady state approximation) (i.e. during which the intracellular AA does not limit the transport of extracellular AA)[18-20]. Hence, the MMRL fairly represents the steady state behavior of an isolated transporter-amino acid couple. However, it is not sufficient to adequately describe the dynamic behavior of amino acid transporters where the AA uptake rate is not linear with time. To overcome this limitation, more recently, Panitchob et al. developed a mechanistic ''carrier'' type model to represent the kinetics of transporters with various transport mechanisms (i.e., uniporter, antiporter, and symporters) [21, 22]. Their mathematical model is based on the law of mass action under quasi-steady state assumptions and provides an abstract representation of the process of transport of AAs with different classes of AATs [23]. The combined use of in silico, in vitro and in vivo approaches to study amino acid transport networks Given the inherent functional complexities associated with the transport systems, it can be very challenging to characterize their function by relying on only pure experimental 11

approaches. On the other hand, in silico methods have shown to be of assistance to unravel these complexities.

In this thesis, we integrated various computational and experimental approaches to gain more insights into the function of amino acid transport systems. In chapter 2, we introduce a combined in vitro and in silico methodology to quantify the function of individual AATs within a complex single cell AA transport network . We thereby predicted the aggregate single cell transport responses for various extracellular assays in the AAs, ions and competitive inhibitors under steady state conditions. We further establish an in silico framework based on the ''carrier'' type models to investigate AAT interactions through multiple cell membranes. In chapter 3, using our in silico framework, we describe the dynamic interactions between the large neutral amino acid (LNAA) transporters expressed at the membrane of different cell types composing the neurovascular unit (NVU) system (i.e., microvascular brain endothelial cells, astrocytes and neurons). We further explain how we combined our in silico model with published in vivo microdialysis measurements to address fundamental physiological questions about the bi-directional kinetics and expression pattern of LAT1 at the BBB which could not be investigated by available in vitro and in vivo techniques. We further employed our NVU-LNAA model to gain more insights into the pathophysiology of Phenylketonuria (PKU) disorder. In chapter 4, we provide a quantitative representation for the interactive cooperation between LNAA transporters involved in this disorder and further describe the therapeutic impact of LNAA supplementation in the control of disturbed concentrations of LNAAs in the cells of NVU in PKU disorder. In the following sections, we introduce how these integrations of in silico, in vitro and in vivo approaches could lead to answering fundamental biological questions about the AAT processes which could not be understood by available experimental techniques.

Quantification of single cell amino acid transport network through a combined in silico and in vitro approach The cellular AA transport response is highly sensitive to extracellular AA conditions, and it can be influenced by changes in the concentrations of different AAs, ions, and inhibitors (stimuli) outside of the cell. On the other hand, various types of AATs with overlapping substrate specificity can be involved in the process of transporting AA into and out of the cell. Hence, it can be a difficult and challenging task to accurately 12

determine the roles of individual AATs from the aggregate transport of a given substrate. This further motivated us to establish a systems biology methodology with the capability to quantify the contribution of individual AAT transporter to the overall transport of a given substrate. We applied our strategy to quantify Leu and Phe AAT systems expressed in the Xenopus leavis oocyte in vitro cell model. In chapter 2, we discuss our approach in more detail. In summary, the first step in our approach is to design appropriate extracellular conditions based on prior knowledge on the kinetics of individual Leu and Phe transporters (i.e. information about substrate specificity and/or ion sensitivity of individual AATs). The next step is to monitor the cellular transport response under designed extracellular medium conditions (i.e. in the presence/absence of competitive substrates and/or inhibitors and/or ions). The following step in our methodology is to construct a stimuli-response model based on the kinetics of individual transporters. The subsequent step is to collect stimulus- response measurements to estimate kinetic parameters of individual transporters based on the global fitting method in which we consider the AAT kinetic parameters shared among all assay conditions. Given the estimated model parameters, the next step is to quantify the activity of individual AATs and finally to predict the overall cellular transport responses to extracellular conditions in the presence of competitive substrates and/or inhibitors.

Quantification of multi-cell amino acid transport network through an integrated in silico and in vivo approach The introduced approach in the second chapter of this thesis, however, is limited to characterize transport processes across single membranes under steady state kinetic assays (MMRL assumptions). In the second phase of this project, we develop an in silico model, beyond the validity region of MM approximation, to investigate dynamic behavior of transporters interacting through multiple cell membranes. The interplay between AATs expressed in the cells of the neurovascular unit (NVU) (i.e., microvascular brain endothelial cells (MBECs), astrocytes and neurons) is of great importance for proper brain function. It has been shown that the Na+-independent antiporter SLC7A5 (LAT1) in microvascular brain endothelial cells (MBEC) [24-26], the Na+-independent antiporter SLC7A8 (LAT2) in astrocytes [27-29], and the Na+- dependent symporter SLC6A15 (B0AT2) in neurons [30-32] are the main regulators of

large neutral amino acid (LNAAs) homeostasis in the NVU (Fig. 1.4). LAT1 is the central element of NVU-LNAA system which plays a crucial role in controlling the homeostasis of LNAAs in the brain (Fig 1. 4). However, its bi-directional kinetic as well as its distribution pattern at the BBB [24-26], could not be well understood so far by available in vivo and in vitro standard assays [26, 33-35]. Additionally, the dynamic interactions between the NVU-LNAA transporters could not be sufficiently captured so far by experimental means. To overcome these limitations, we established a new computational model of NVU-LNAA homeostasis and incorporated it with the published in vivo microanalysis measurements carried out in rat models. This has allowed us to evaluate existing hypotheses about the LAT1-mediated trans-endothelial transport of LNAAs across the blood brain barrier (BBB). In particular, in chapter 3, we describe that either strong asymmetrical bi-directional kinetics of LAT1 in microvascular brain endothelial cells (MBECs) and/or an asymmetric distribution of LAT1 at both membranes of MBECs is required to reproduce the experimentally measured brain ISF responses to intraperitoneal (IP) L-tyrosine (Tyr) and L-phenylalanine (Phe) injections. Additionally, we employed our model to capture changes in LNAA levels in MBEC, astrocytes, and neurons upon perturbations of plasma LNAA concentrations.

Figure 1. 4. Schematic representation of the neurovascular unit (NVU) and the therein expressed dominant large neutral amino acid (LNAA) transporters. The dominant LNAA transporters in the NVU include the Na+-independent antiporter LAT1

(SLC7A5) in MBECs, the Na+-independent antiporter LAT2 (SLC7A8) in astrocytes and the Na+-dependent symporter B0AT2 (SLC6A15) in neurons. The arrows specify the LNAA transmembrane pathways.

Investigation of perturbed AA transport pathways in Phenylketonuria disease In chapter 4, we further describe the employment of our in silico NVU-LNAA model (introduced in chapter 3) to get more insights into the perturbed dynamics of LNAAs in Phenylketonuria (PKU) disorder. PKU is an inherited disease characterized by abnormally high concentrations of the essential amino acid L-phenylalanine (Phe) in the systemic blood. It is caused by a deficiency in activity of phenylalanine hydroxylase (PAH) which also leads to the abnormally high fluctuations of plasma Phe concentration as compared to normal physiological conditions [36, 37]. While numerous studies have verified a negative correlation between blood Phe accumulation and cognitive outcomes, they have not clarified whether Phe fluctuations in plasma could affect brain functions [38]. This requires a deep understanding of complex and interactive competitions between Phe and LNAAs taking place through NVU-LNAA transporters, which could not be achieved precisely by available in vitro and in vivo standard methods so far. To bridge this gap, we quantified the interactive dynamics of Phe and competitive LNAAs (CL) from plasma into NVU individual cells. As described in chapter 4, we suggest that plasma Phe fluctuations can potentially propagate into the NVU and change the dynamics of LNAA concentrations. However, the induced effects vary between NVU individual compartments and they largely depend on plasma fluctuation Phe indices such as mean, fundamental frequency and amplitude-to-mean ratio for which we provide a quantitative representation. Finally, we employed the in silico model to explain the therapeutic impact of LNAA supplementation in attenuation of disturbed concentrations of Phe and CL in the cells of NVU in PKU disorder.

Chapter 2 Quantifying the relative contributions of different solute carriers to aggregate substrate transport

This chapter has been published as:

Mehdi Taslimifar , Lalita Oparija, François Verrey, Vartan Kurtcuoglu, Ufuk Olgac, and Victoria Makrides. "Quantifying the relative contributions of different solute carriers to aggregate substrate transport." Scientific reports 7 (2017): 40628.

Abstract

Determining the contributions of different transporter species to overall cellular transport is fundamental for understanding the physiological regulation of solutes. We calculated the relative activities of Solute Carrier (SLC) transporters using the Michaelis-Menten equation and global fitting to estimate the normalized maximum transport rate for each transporter (Vmax). Data input were the normalized measured uptake of the essential neutral amino acid (AA) L-leucine (Leu) from concentration- dependence assays performed using Xenopus laevis oocytes. Our methodology was verified by calculating Leu and L-phenylalanine (Phe) data in the presence of competitive substrates and/or inhibitors. Among 9 potentially expressed endogenous X. laevis oocyte Leu transporter species, activities of only the uniporters SLC43A2/LAT4 (and/or SLC43A1/LAT3) and the sodium symporter SLC6A19/B0AT1 were required to account for total uptake. Furthermore, Leu and Phe uptake by heterologously expressed human SLC6A14/ATB0,+ and SLC43A2/LAT4 was accurately calculated. This versatile systems biology approach is useful for analyses where the kinetics of each active protein species can be represented by the Hill equation. Furthermore, its applicable even in the absence of protein expression data. It could potentially be applied, for example, to quantify drug transporter activities in target cells to improve specificity.

Keywords: Systems biology, solute carrier, amino acid transporter, in vitro cell model

Introduction Solute carriers (SLCs) represent a large group of eukaryotic membrane transport proteins that control the uptake and efflux of a wide range of substrates such as inorganic ions, nucleotides, amino acids (AAs), neurotransmitters, sugars, purines, fatty acids, and thus, also drug molecules [3]. Solute carriers are ubiquitously expressed in all tissue and cell types, and in most organelles including lysosomes and mitochondria. The activities of SLC species are often highly redundant, and furthermore, the regulation of SLC expression and activity is frequently complex and influenced by numerous stimuli. Therefore, it can be difficult to accurately determine the roles of a particular species of SLC in the aggregate transport of a substrate. The goal of the work at hand was to establish a methodology that enables the quantification of the relative contributions to the overall transport of a given substrate by specific SLC species based on their enzymatic characteristics. Amino acids by virtue of their indispensable roles in protein, energy, neurotransmission, and other crucial metabolic pathways, are key physiological molecules. Since AAs cannot passively diffuse through intact cell membranes, movement across biological membranes is largely mediated by a subclass of SLCs, the amino acid transporters (AATs). Due to their control over AA transport across barrier membranes, AATs perform crucial roles in AA homeostasis. By mediating intestinal absorption and renal reabsorption, AATs are among the cornerstone regulators of AA bioavailability in humans and other mammals [1, 2, 5, 39]. To date, of 52 assigned families of SLCs, eight (SLC 1, 6, 7, 12, 16, 25, 38, 43) are known to have members transporting AAs [4] . In total more than 75 SLC protein species are recognized as AATs [4]. All AATs function mechanistically by either simple facilitative diffusion (passive transport), or by sym- and/or anti-port of co-substrates such as ions (secondary active transport), and/or the obligatory exchange of AA pairs [2, 3]. The driving force for vectorial transport is provided by chemical and/or electrical gradients. Additionally, functional interactions between transporters operating by different mechanisms can give rise to cooperative amino acid transport [40, 41]. For example, it was shown by exogenous expression in Xenopus laevis oocytes that an obligatory exchanger, SLC7A8/LAT2 (LAT2), efﬂuxes intracellular AAs to AA free buffer only in the presence of a co-expressed facilitative transporter, SLC16A10/TAT1 (TAT1). Cooperative transport is accomplished when TAT1 recycles to the outside a LAT2 17

uptake substrate, e.g. L-phenylanalanine (Phe), against which LAT2 can efﬂux in exchange another intracellular AA[42]. The physiological functions of mammalian AATs (and of SLCs in general) have been commonly studied by probing responses of endogenous transporters in vivo in whole animals, ex vivo in organs, tissues, or cells, or by testing cloned wild-type or mutated transporters heterologously expressed using in vitro cell models [6-11]. While these approaches have yielded a large body of knowledge, for many studies, such as on AAT regulation or interactions, data interpretation can be confounded by the intrinsic complexity of the involved biological networks. This complexity, and the consequent difficulties for data analyses, arises from the fact that there are over 20 physiologically relevant AAs, and the ubiquitous cellular expression of multiple AAT protein species with overlapping AA specificities and non-mutually exclusive transport mechanisms. Meaningful analysis of these complex processes would be aided by a method based on AAT kinetic characteristics to determine the relative contributions of specific transporter species to overall substrate transport. In this study, our aims were to develop a strategy to (1) quantify the relative function of specific SLC species within a system of transporters with a variety of substrate affinities and transport mechanisms, and (2) experimentally verify the calculations for total transport and for responses to new stimuli (extracellular conditions) by individual transporter species. As a biological model we chose transporters expressed in Xenopus laevis oocytes for the essential neutral amino acid L-leucine (Leu) (Table 2. 1). As an important model organism X. laevis has been extensively physically and biochemically characterized. Furthermore, X. laevis oocytes have been widely used for molecularly identifying SLCs (including AATs) and for characterizing transport kinetics including substrate specificities and transport mechanisms [26, 43-54]. However, while good data exists for expression of X. laevis oocyte endogenous AAT (xAATs) mRNA, little data exists about their protein expression. For the developed approach, the activity of specific AAT species in different experimental conditions were represented by the Michaelis-Menten (MM) equation. We estimated unknown transporter parameters by a global fitting method [55, 56] in which they were considered as shared parameters among all assay conditions. Model calculations were verified by comparison to measured AA uptake rates. Using our approach we (1) identified the relative activity of endogenous X. laevis AATs Leu

transporters in a variety of experimental conditions, (2) extended our predictions to successfully characterize Phe transporters, and (3) accurately calculated the behaviour of the heterologously expressed human sodium-dependent (Na+-dep) symporter SLC6A14/ATB0,+ (hATB0,+) and sodium-independent (Na+-indep) uniporter SLC43A2/LAT4 (hLAT4). Our approach is shown to provide a robust and versatile tool for unraveling the contributions of specific players in a complex cellular network of transporters and substrates.

Table 2. 1. Model input parameters (Vmax and Km) for L-leucine and L-phenylalanine transporters Symporters Antiporters Uniporters SLC no. SLC6A14 SLC6A19 SLC7A5 SLC7A8 SLC7A7 SLC7A6 SLC1A5 SLC16A10 SLC43A1 SLC43A2 Alias ATB0,+ B0 AT1 LAT1 LAT2 y+LAT1 y+LAT2 ASCT2 TAT1 LAT3 LAT4

Accessory coll 4F2hc 4F2hc 4F2hc 4F2hc

protein TMEM27 SLC3A2 SLC3A2 SLC3A2 SLC3A2

Vmax (nmol/h)

Vmax1 Vmax2 Vmax1 Vmax2

0.014a 0.539 0.162 0.402 0.38 1.60 1.01 NA 0.0138 0.766 0.0204 3.516 Leu [57] [47] [58] [59] [60] [61] [62] [43] [43] [45] [45]

10 0.034a 1.118 0.147 0.27 0.0076 0.75 0.0204 3.516 Phe [57] [47] [58] [59] NA NA NA 0 [43] [43] [45] [45] [53]

Km (mM)

Km1 Km2 Km1 Km2

Leu 0.012 1.1 0.032 0.048 0.0317 0.236 0.367 NA 0.0842 1.024 0.103 3.733 [63] [64] [65] [66] [60] [61] [62] [43] [43] [45] [45]

Phe 0.017 4.7 0.74 0.0122 NA NA NA 36 0.0658 1.206 0.178 4.694 [63] [64] [65] [66] [53] [43] [43] [45] [45]

Ala 0.099 4.1 NA 0.167 NA 4.12 0.0184 NA NA NA NA NA [63] [64] [66] [61] [62]

Arg 0.104 NA NA NA 0.34 0.177 NA NA NA NA NA NA [63] [60] [61]

Trp 0.026 >12 0.0214 0.0576 NA NA NA NA NA NA NA NA [63] [64] [58] [59]

0.055 0.023 0.055b 2.5ab 0.055b 28ab 0.52a NA NA NA NA NA BCH [28] [28] [28] [43] [28] [43] [63]

Val 0.036 1.53a 0.0472 0.124 NA NA 0.522 NA 0.0306 1.885 0.0472b 36ab [63] [47] [58] [59] [62] [43] [43] [28] [43]

Sodium yes yes no no yes yes yes no no no dependent Log 10 4 0.6 1 2.5 3.4 3 3.5 0.8 2.4 1.5 mRNA

References for model input are given. As indicated all Vmax rates were reported as nmoles per hour (nmol/h). If the Vmax rates were not provided in units a of nmol/h then these values were recalculated from literature reported units. Vmax and Km values were calculated using the Michaelis-Menten equation. b The high affinity component was assumed to be equal to LAT1. NA (not applicable) indicates the amino acid was not reported to be a substrate for the

transporter. The Log10 mRNA expression is taken from the reported microarray data for unfertilized X. laevis oocytes (Xenbase). The mRNA expression Log10 values for the accessory proteins, TMEM27/collectrin and SLC3A2/4F2hc, are 1.5 and 1.0 respectively. Standard three letter abbreviations are used for amino acid names; BCH is the Leu analog b(−)2-aminobicyclo[2,2, 1]heptane-2-carbocyclic acid.

Methods

An overview of the systems biology approach we developed for this study and the generally applicable method for characterizing enzyme activities are shown in Fig. 2. 1. The method can be applied for calculating the contributions of specific enzyme species to the bulk measured enzymatic activity if the MM (or Hill) equation can be used to describe the kinetics of the target enzyme species. In other words this method can be applied for analyses of steady-state (or rapid equilibrium) measurements of a saturable activity under conditions where (1) the enzyme concentration is well below the MM binding constant ( ), and (2) substrate binding (and dissociation) ocurrs much more rapidly than product퐾푚 formation (i.e. transport) [67].

Figure 2.1. Diagram of the method used for quantifying activity of specific transporter species. The left panel is a schematic of the steps taken in the current study to establish and verify the method for amino acid transport in Xenopus laevis oocytes. AAT refers to Solute Carrier amino acid transporters for L-leucine and L- phenylalanine. The source for the mRNA expression data was Xenbase as described in Methods and Results. Uptake assay protocols are described in Methods. Defined uptake buffers are described in Methods, Results, and Fig. 2. The input parameters for the calculations are given in Table 2. 1. The rationale for the specific equations used to construct the current model are given in the methods. The right panel is a generalized series of steps that could be applied to any system where the kinetics of each active enzyme species can be represented by the Hill equation.

Experimental Procedures

Preparation of human SLC6A14/ATB0,+ and SLC43A2/LAT4 cRNA

The cDNA for hATB0,+ (cloned in the pSPORT1 vector) was kindly provided by Andreas Werner (University of Newcastle upon Tyne, Newcastle, UK). Linearization for cRNA preparation was carried out using NOT1 restriction enzyme digestion (Thermo Scientific). Flag-tagged hLAT4 was prepared from hLAT4 cDNA (in pTLN vector) kindly provided by Manuel Palacin (IRB Barcelona, Barcelona Spain). The cDNA was initially cloned in a pLenti6-EGFP vector (Invitrogen) using NruI and MluI restriction sites and subsequently transferred into aFastBac-FLAG(C) vector (kindly provided by Thierry Hennet, University of Zurich) with the following primers: 5’-CAT GGC GCC CAC CCTGGC CAC TG-3’ (for) and 5’-CTA CAC GAA GGC CTC CTG GTT G-3’ (rev). The hLAT4-FLAG(C) insert was then cloned into a pSDeasy vector using XbaI / NotI restriction sites. FLAG-tagged hLAT4 in pSDeasy (referred to in text as hLAT4) cDNA was linearized with Pvu I restriction enzymes (Thermo Scientific) for cRNA preparation. The cRNAs for both human transporters were synthesized (according to the manufacturer’s protocol) using the MEGAscript high yield transcription kit (Ambion) and the T7 (hATB0,+) or SP6 (hLAT4) RNA polymerases.

Xenopus laevis oocyte preparation

Stage VI oocytes [68] were treated with collagenase A at room temperature in Ca2+- free buffer (82.5 mM NaCl, 2 mM MgCl2 and 10 mM HEPES, pH 7.4) for 50-60 minutes. Remaining follicular layers were removed manually and non-injected (NI) oocytes were either tested naïve or injected with 10 ng of hLAT4 or 25 ng of hATB0,+ cRNA. Oocytes were incubated for three days at 16oC in modified Barth’s solution (88 mM NaCl, 1 mM

KCl, 0.82 mM MgSO4, 0.41 mM CaCl2, 0.33 mM Ca(NO3)2, 2.4 mM NaHCO3, 10 mM HEPES, 5 mg/l Gentamicin, 5 mg/l Doxycycline) before assaying [49].

Xenopus laevis oocyte L-amino acid radiolabelled tracer uptake assay

Oocytes were pre-equilibrated for 2 minutes at 25°C in a sodium containing solution + (100 mM NaCl, 2 mM KCl, 1 mM CaCl2, 1 mM MgCl2, 10 mM HEPES pH 7.4 (+Na )) or in the uptake assay buffers (uptake buffers) specified in Fig. 2. 2. Uptake buffers for specific substrates (e.g., Leu or Phe) contained defined sodium concentrations (i.e.

with 100 mM sodium (+Na+) vs. without Na+ (Na+free) in which Na+ was replaced with equimolar N-Methyl-D-glucosamine (NMDG)). Leu uptakes were carried out in uptake buffers without and with the addition of the following 10 mM competitive substrates: L- alanine (+Ala), L-arginine (+Arg), 2-aminobicyclo [2.2.1]-heptane-2-carboxlate (+BCH), and L-tryptophan (+Trp). For uptakes, oocytes were incubated in 100 µl of uptake buffers containing unlabeled AA in concentrations as indicated (see text, Fig. 2. 2 or figure legends). The [3H]-L-radiolabeled AAs (Leu or Phe) at a concentration of 20 μCi/ml (for experiments with only non-injected (NI) oocytes), or 5 μCi/ml (for experiments with exogenously expressed AATs including NI controls) were added as tracers (Hartmann Analytic, Braunschweig, Germany) for the indicated assay times (at 25°C). To stop the reactions, uptake buffers were removed, oocytes washed six times with ice-cold +Na+, and lysed individually in 2% SDS. Scintillation fluid (Emulsifier- Safe™) was added and radioactivity counted in a liquid scintillation counter (TRI-CARB 2900TR, Packard Instrument, Meriden, CT).

Uptake data normalization

The measured Leu and Phe uptake rates were normalized to the uptake rate of 1 mM Leu and 10 mM Phe in +Na+ uptake buffers, respectively. Experiments were routinely repeated for 3 or more oocyte batches (as indicated in the figure legends). Additionally, some experiments were performed over more than one day using the same batch of oocytes, therefore, normalization was carried out for data from each day experiments were performed, as well as for each batch of tested oocytes.

Figure 2. 2. Inhibition of L-leucine transporters in tested uptake buffers L-leucine (Leu) uptake was tested in buffers containing 10 mM added competitors (BCH or amino acids). The Leu transporters (AATs) with affinity (Km (mM) values shown in Table 2. 1) for the added competitors are indicated in the filled cells of the same color as the text for the named competitors. For Na+free uptake buffers, the Na+-dependent Leu AATs

(i.e. require Na+ for Leu uptake) are indicated by ND. Additionally, responses of the Leu transporters for the competitor listed first for each uptake buffer are shown in the upper row and for the competitor listed second are shown in the lower row of filled cell(s). Clear boxes indicate AAT Leu uptake is unchanged due to addition of the indicated competitor and are used for AATs that transport Leu with affinities listed in a Table 2. 1 but do not transport the added competitor. Km (mM) value is for low affinity kinetic component.

Mathematical model

An AAT kinetic component (AATi) is defined as each unique MM binding affinity ( ) and maximum transport rate ( ) exhibited by an AAT species for a given substrate.퐾푚 In X. laevis oocytes there are푉푚푎푥 11 AATi for Leu transport among the 9 known Leu endogenous AAT (xAAT) species (Table 2. 1). This is because SLC43A1/LAT3 (LAT3) and LAT4 each display a high and low affinity for Leu and thus 2 AATi, while the remaining 7 xAAT species each have 1 AATi for Leu [43, 45]. The uptake rate per oocyte in a defined assay condition (a) carried out by all xAAT species ) is given by the sum of the uptake rates of the active xAAT species as (Vendo,a

ni 1 endo,a i,a where is the cumulative xAATV uptake= ∑ ratei=1 V for, a given substrate (e.g., Leu) in a definedV endo,aassay condition (including substrate concentration), is the rate of transport by each exhibited AATi of the active xAAT species, and isV thei,a total number of AATi for a given substrate displayed by all active xAAT species.ni Assuming non-cooperative binding, (i.e. a Hill coefficient of 1), under steady-state conditions the transport rate for a given xAAT species can be defined by the MM equation as

= , 2 [S]a i,a 푚푎푥,i 푚,i [ ]a Where V 푉 is the퐾 maximal+ S transport rate by a AATi of an xAAT species, provided that all AAT푉i 푚푎푥of that,i species are bound to the given AA, with the corresponding MM binding constant; and is the concentration of the AA in 퐾the푚,i given assay condition. In the presence of different[S]푎 competitive inhibitors, the rate of transport for a given assay condition by eachnI xAATi is given by

= , 3 [S]a i,a 푚푎푥,i nI V 푉 [I]a 퐾푚,i (1+∑ ( ) )+[S]a 퐾퐼 j where and are inhibitorj=0 concentration and inhibitor dissociation constant of each competitive[I]푎 inhibitor퐾퐼 for a given AAT type (Table 2. 1), respectively. Using from equation (3), then equation (1) for cumulative uptake rate per oocyte in a givenVi,a uptake buffer ( ) can be defined as Vendo,a

ni [ ]a endo,a 푚푎푥,i S V = ∑i=1 푉 nI , 4 [ ]a 푚,i ( ∑ ( I ) ) [ ]a 퐾 1 + 퐼 + S j=0 퐾 j Global fitting

In equation (4), , is a free parameter that is shared between assay conditions

푚푎푥,i and therefore assumed 푉 to be equal for all uptake buffers. Using global fitting (also referred to as shared parameter fitting) a single (global) value is estimated as the best fit for each free parameter ( ) for all uptake buffers [55, 56]. The fitting process

푚푎푥,i seeks to minimize the loss function 푉 given by the mean squared residual error (MRE)

na 1 2 ∑ (Vendo,a − V̅̅endo,a̅̅̅̅̅̅̅) , na a=1 where is the number of assay conditions (i.e. for 2 uptake buffers and 7 Leu

a concentrations,n =14), is the cumulative uptake rate calculated based on

a endo,a equation (4) andn Vis the experimentally measured uptake rate by all xAATs for

endo,a each assay condition.V̅̅̅̅̅̅̅ ̅ ̅

To estimate the free parameters , the cumulative xAAT uptake rate, , was 푚푎푥,i ̅̅endo,a̅̅̅̅̅̅̅ measured for Leu transport in two 푉 uptake buffers (+Na+ and Na+free) at 7 extracellularV

Leu concentrations (0 – 1000 µM). In addition, for each experimental condition, the

cumulative uptake rate by xAATs, , was calculated based on equation (4) using

endo,a the literature reported values for V each AAT species (Table 2. 1). The MRE (equation

푚 (5)) was minimized simultaneously퐾 for all datasets using least squares optimization in

Origin (OriginLab (2016), Origin: An industry-leading scientific graphing and data analysis software. Northampton, MA, United States. URL http://originlab.com/), yielding the unknown parameters . Based on the estimated and the

푚푎푥,i 푚푎x,i 푚,i values reported in the literature for 푉 each AAT species (Table 2. 1), 푉the uptake rate퐾 by all xAAT species with previously reported mRNA expression (Xenbase) was calculated for 0 – 1000 µM extracellular Leu using equation (2). As a validation step, based on equation (4) and using the estimated and reported values (Table 2. 1), the

푚푎푥,i 푚,i cumulative uptake rate by xAATs for six 푉 uptake buffers was퐾 simulated for 0 – 1000 µM extracellular Leu.

Relative expression of amino acid transporters

The maximum transport rate for a given AA ( ) by an AAT species (or for AAT species with multiple kinetic components the푉푚푎푥 dominant ) is related to the (dominant) maximum turnover rate by a single AAT molecule푉푚푎푥,푖 as follows 푚푎푥,s (푉 ) 6

푚푎푥 푚푎푥,s 푉 = N 푉 where is the total number of expressed molecules of an AAT species. To calculate

N, we N assumed that for given substrate the (dominant) maximum transport rate by single AAT molecule of given xAAT species is equal to that of the exogenously expressed AAT ortholog Therefore,푉푚푎푥,s,endo using equation (6), the level of expression of exogenous relative to푉 푚푎푥endogenous,s,exo. AATs (R) is

7 Nexo 푉푚푎푥,exo Where R = N endoand= 푉푚푎푥,endo are, the total number of exogenous and endogenous AAT molecules Nexo of the givenNendo AAT species, respectively. The values (corresponding 26 푉푚푎푥,endo

to dominant ) were calculated using global fitting (to solve equation (5)), whereas 푉푚푎푥 for,푖,endo each xAAT ortholog was taken from the literature (Table 2. 1). 푉푚푎푥,exo Calculation of human SLC6A14/ATB0,+ activity

0,+ Total uptake rate by oocytes expressing hATB (Vt,a) is given by

= + 8 0,+ Vt,a Vendo,a VhATB ,a , where is the rate of uptake in a given assay condition by all xAAT species (from

0,+ equationVendo,a (4)) and is the hATB uptake rate. Based on the MM equation 0,+ (equation (2)), then VequationhATB (8) can be written as

= + 9 a 0,+ [S] t,a endo,a V V 푉푚푎푥,hATB 퐾푚,hATB0,++[S]a, To estimate , was determined in 2 assay conditions (10 µM and 1000 0,+ µM Leu in +Na). 푉푚푎푥 Subsequently,,hATB Vt,a equation (9) was fit to the data using nonlinear least squares fitting, employing (equation (4)) and the reported (Table 2. 0,+ 1) as the known parameters.Vendo,a The maximum transport rate by hATB 퐾푚,hATB0,+ ( ) 0,+ was calculated, and used in the MM equation with (Table 2. 1) 푉 to푚푎푥 calculate,hATB 0,+ 0,+ the hATB uptake rates from 0 – 3000 µM Leu. 퐾푚,hATB

Calculation of human SLC43A2/LAT4 L-phenylalanine uptake rate

Based on equation(7), the relationship between the maximum transport rates for Leu and Phe by exogenous and endogenous AATs is:

= , 10 푉푚푎푥,i,endo,Phe 푉푚푎푥,i,exo,Phe where 푉푚푎푥,i,endo,Leu 푉is푚푎푥,i,exo, estimatedLeu by global fitting (equation (5)). The values for

푉푚푎푥 ,i,endo,and Leu were taken from the literature for each xAAT species ortholog 푉푚푎푥,i,exo, (TableLeu 2. 1) 푉 푚푎푥to calculate,i,exo,Phe for the 6 xAAT species transporting both Leu and Phe (Table 2. 푉푚푎푥 1). ,i,endo, BasedPhe on the Xenbase database[69], xSLC16A10/TAT1 (xTAT1) mRNA is expressed by unfertilized oocytes. Furthermore, it is the only xAAT species potentially expressed by oocytes that transports Phe but not Leu. The maximum transport rate for xTAT1 was estimated using cumulative uptake by NI oocytes measured for 2 assay conditions (1000 µM and 10000 µM Phe 27

in +Na; equation (4)). The estimated values for xSLC6A19/B0AT1 (xB0AT1), xLAT4, and xTAT1 were used with the 푉respective푚푎푥,푖 reported values for Phe for the orthologous AAT species (Table 2. 1) in the MM equation to calculate퐾푚 the uptake rates from 0 – 10,000 µM Phe.

For the total uptake rate by hLAT4 expressing oocytes (as for hATB0,+), equation (8) can be written as

= + , 11 [S]푎

t,a endo,a LAT4 푚푎푥,xLAT4 퐾m,hLAT4+[S]푎 V V R ∙ 푉 where is the cumulative uptake rate by the 9 xAATi for 7 xAAT species transporting Vendo,a Phe (equation(3)), and is the expression of hLAT4 relative to xLAT4 based on equation (7). To estimate RLAT4, was determined by uptake experiments + using 2 assay conditions (1000 µMRLAT4 and V 10,000t,a µM Phe in +Na ). Subsequently, equation (11) was fit to the data using nonlinear least squares fitting, employing

(equation (1)), (global fitting solution of equation (5)), and the reportedVendo,a (Table 푉 푚푎푥 2. 1),xLAT4 as the known parameters, and as the free parameter. Based 퐾푚,hLAT4 on the estimated , the hLAT4 maximum transportRLAT4 rate ( ) was calculated and used in theRLAT4 MM equation with the reported 푉 (Table푚푎푥,hLAT4 2. 1) to calculate hLAT4 uptake rates from 0 – 10,000 µM Phe. All calculations퐾푚,hLAT4 were carried out using Origin (OriginLab (2016), Origin: An industry-leading scientific graphing and data analysis software. Northampton, MA, United States. URL http://originlab.com/''.)

Results

Xenopus laevis oocyte endogenous L-Leucine transport

The National Institutes of Health (NIH) supports the web based resource Xenbase [69- 71] that reports, among other information, Xenopus laevis mRNA expression data from microarray studies. In this study we used the Xenbase microarray data to determine the set of potentially expressed xAAT species in unfertilized oocytes. This search identified 9 SLC genes that code for Leu xAATs with mRNA expression (Table 2. 1). We first determined the appropriate experimental conditions for which the transport activity of these AAT protein species can be described by MM kinetics [67]. We observed that under our assay conditions, and consistent with previous reports, Leu

uptake rates by oocytes increased linearly for at least 8 minutes (Suppl. Fig. 2. 1). Therefore, subsequent experiments were carried out at 3 minutes as indicated in figure legends and/or text. Leu concentration-dependence (0−1000 µM) uptake assays were performed using +Na+ or Na+free uptake buffers. Leu uptake in +Na+ buffers corresponds to uptake by both Na+-dep and Na+-indep AAT species, while the uptake measured in Na+free uptake buffers indicates the activity of Na+-indep AAT species alone (Tables 2.1 and Fig. 2. 2). A simultaneous fit (global fitting method) to the measured uptake responses in +Na+ and Na+free buffers was used to calculate the maximum transport uptake rates by the active Leu xAAT species (Fig. 2. 3A). Since there are no reported kinetic studies for xAAT species, values used in the calculations were taken from published reports of mammalian퐾푚 orthologs assayed by heterologous expression in X. laevis oocytes (Table 2. 1). Based on the values calculated for each xAAT species (Table 2. 2) and the respective reported푉 푚푎푥 values for the orthologs (Table 2. 1), the activity of xAATs was determined usin퐾g푚 the MM equation (Fig. 2. 3B). Although mRNA for 9 Leu xAAT species has been detected in unfertilized oocytes (Xenbase), we calculated that the activity of only two low affinity transporter species accounted for total Leu uptake. These were the sodium-dependent symporter xB0AT1 and the sodium-independent uniporter xLAT4, while the remaining xAATs were calculated to contribute minimally to Leu uptake. The activity of xLAT4 was found to predominate throughout the range of Leu concentrations tested (10 – 1000 µM). The calculated normalized (± SD) for xB0AT1 and xLAT4 (low affinity kinetic component [43, 45]) was 30.7 푉±푚푎푥 5.1 and 383.4 ± 8.7, respectively, indicating that the relative contribution to Leu uptake by xLAT4 was approximately 12 times that of xB0AT1 (Fig. 2. 3B).

Table 2. 2. Comparison of endogenous Xenopus laevis L-leucine transporter maximum turnover rates estimated using model input data from assays in various combinations of uptake buffers.

Original Test

Leu+Na+ Leu+Na++Ala+Arg Experimental buffer Leu+Na+ Leu+Na++Ala+Arg Leu+Na++BCH Leu+Na++Ala+Arg conditions used as Leu+Na++BCH model input Leu+Na+free Leu+Na++BCH Leu+Na+free+Ala Leu+Na+free Leu+Na++BCH+Ala Leu+Na+free+Ala

0 Vmax,xB AT1 30.7 ± 5.1 28.2 ± 8.1 39.2 ± 4.3 27.9 ± 4.8 34.3 ± 2.5

Vmax,xLAT4 383.4 ± 8.7 405.3 ± 16.6 372.1 ± 8.2 406.5 ± 9.4 382.0 ± 3.2

Results are amino acid transporter maximum transport rates (normalized to uptake rate at 1 mM Leu in +Na+ buffer) with standard deviations ( ± SD). The model estimations were calculated using data generated from different combinations of uptake buffers as indicated in푚푎푥,푖 the column headings. Data generated from assays in the “Original” buffer combination was used for model estimations of Vmax. The푉 dependence of the model calculations on the data from different buffer combinations used as input was probed using the 0 “Test” buffer combinations. The mean ± SEM for the four ”Test” buffer combinations is 32.4 ± 2.7 for xB AT1 and 391.5 ± 5.2 for xLAT4. 푉푚푎푥,푖

Model verifications

Several model verifications were performed as follows: (1) for the xAATs species

was calculated from concentration-dependence assays carried푉푚푎푥,푖 out in different uptake buffers (Table 2. 2), and (2) Leu uptake rate was measured in uptake buffers with 10 mM added competitive substrates (Fig. 2. 2) and compared with the calculated model output. To test the dependence of the model output relative to the input data, the

values calculated from 4 combinations of uptake buffers were compared with v푉alues푚푎푥,푖 from Leu uptake response in ± Na+ buffers. The mean normalized values for

0 xB AT1 and xLAT4 in the 4 new buffer combinations were calculated as푉푚푎푥 32.4,i ± 2.7 and 391 ± 5, respectively (Table 2. 2). Since these values compared favorably with the original calculations in ± Na+ buffers (see above), further simulations were carried out using the normalized values calculated from ± Na+ Leu uptake response.

푉푚푎푥,i Next, we checked the model calculations for Leu uptake responses by xAATs species in the presence of uptake buffers containing excess competitive inhibitors (Fig. 2. 2). Calculations of cumulative uptake rates were carried out using the calculated

for each xAAT species and the reported values for the mammalian orthologs푉푚푎푥,i (Table 2. 1). The calculated cumulative퐾푚,i uptake curves were compared with

experimental data for the tested uptake buffers (Figs. 2.3 and 2.4). For example, in Na+free buffer, if the model calculations were correct, then only the Na+-indep xLAT4 and not the Na+-dep B0AT1 could have contributed to total Leu uptake. As we predicted, excess added Ala ± Trp did not inhibit Leu uptake relative to uptake in Na+free buffer alone (since neither AA is a substrate for LAT4) (Figs. 2.2 and 2.3C). There are several other Na+-indep xAAT species (LAT1, 2, 3) with reported mRNA expression in oocytes. Since Ala is a LAT2 substrate, and Trp is a substrate for LAT1 and LAT2, if either AAT species were significantly active in

Figure 2. 3. L-leucine uptake by endogenous Xenopus laevis oocyte transporters. Three minute uptakes by non-injected (NI) oocytes of 10 – 1000 µM L- leucine (Leu) in uptake buffers containing (+Na+) and without sodium (Na+free) were tested using radiolabeled amino acid (AA) tracers. In all panels uptake data (pmol/3 min per oocyte) were normalized to uptake in 1 mM Leu, +Na+ uptake buffer for each batch of oocytes and experimental day. Panel (A) shows normalized Leu uptake rate data (expt) vs. the simultaneous fit (fit) for endogenous Leu uptake rates. Panel (B) shows model calculations (model) for contributions to uptake of 10 – 1000 µM Leu by various endogenous Xenopus laevis oocyte (xAAT) species. Model output was calculated based on the calculated xAAT and the reported for each xAAT ortholog (Table 2. 1). The 95% confidence limits for transporter activities 푉푚푎푥,i 퐾푚 are shown (dotted lines) bracketing the model predictions. Panels (C) and (D) show 31

the calculated Leu uptake rates in +Na+ and Na+ free uptake buffers containing excess competitive inhibitors. Panel (C) shows the concentration dependence (0 – 1000 µM Leu) of normalized cumulative uptake data by NI oocytes in Na+free uptake buffers containing 10 mM each of the following competitors: L-alanine (Ala) (Na+free+Ala), Ala and L-tryptophan (Trp) (Na+free +Ala +Trp), L-valine (Val) (Na+free +Val) vs. calculated cumulative endogenous uptake results for the respective uptake buffers. For uptakes in Na+free+Ala, and Na+free+Ala+Trp uptake buffers, the calculated values were virtually indistinguishable from the global fit for the Na+free data, therefore a single line was used for graphing all three data sets. Panel (D) shows the concentration dependence (0 – 1000 µM) of total Leu uptake rates in +Na+ uptake buffers containing 10 mM each of the following competitors: Ala, Arg (+Na++Ala+Arg), 2-aminobicyclo- (2,2,1)-heptane-2-carboxylic acid (BCH), (+Na++BCH), and BCH and Ala (+Na++BCH+Ala) (Fig. 2. 2). n = 6 – 8 ooyctes each experiment, for 3 independent experiments. The experimental data are shown as the mean ± SEM for the measured uptake rates and the calculated cumulative endogenous uptake rates were based on the previously calculated and reported values for the kinetic components exhibited by each AAT species (Table 2. 1). 푉푚푎푥 퐾m oocytes, then addition of excess Ala or Trp would have competitively inhibited Leu uptake (Fig. 2. 2). Furthermore, our predicted decrease in total Leu uptake in the presence of added Val (a LAT4 substrate) was confirmed (Fig. 2. 3C).

The cumulative Leu uptake rates by xAATs measured in +Na+ with 10 mM of the competitive substrates (1) +Ala+Arg, (2) +BCH, or (3) +BCH+Ala vs. the calculated results is shown in Fig. 2. 3D. Briefly, Ala is a substrate for B0AT1 but not LAT4, while excess Arg does not compete for Leu uptake by either AAT species, and BCH competitively inhibits LAT4 but not B0AT1 (Fig. 2. 2). Therefore, if the model calculations were correct, we predicted (as observed) that the total Leu uptake from highest to lowest would be: +Na+ (100%) > Na++Ala+Arg (Ala inhibits only the lesser active xB0AT1) > Na++BCH (BCH inhibits the highly active xLAT4) > Na++BCH+Ala (inhibits both xB0AT1 and xLAT4 activity) (Fig. 2. 3D). Overall, total Leu uptake curves were accurately calculated for all tested uptake buffers, supporting the validity of the model output (Figs. 2.2 and 2.3D, E).

Calculating the relative activity exogenously expressed transporters

The utility of the model for calculating the activity of heterologously expressed AATs distinct from xAATs was tested for exogenously expressed hATB0,+. The uptake measured from non-injected (NI) and hATB0,+ injected oocytes was compared with

calculated uptake rates (Fig. 2. 4A). Using the total uptake measured from hATB0,+ injected oocytes (NI+hATB0,+) for two Leu concentrations (10 and 1000 µM), the maximum normalized transport rate for exogenously expressed hATB0,+ transporters ( ) was calculated as 139.7 ± 26.9 (equation (9), Fig. 2. 4A). This value was 0,+ used 푉푚푎푥 in,hATB the MM equation together with the reported value (Table 2. 1) to 0,+ 푚,hATB calculate the rate of uptake by hATB0,+ transporters 퐾 from 0 – 3000 µM Leu (Fig. 2. 4B). The hATB0,+ transport rate was calculated to plateau at ≈ 100 µM as expected from the high affinity of hATB0,+ for Leu (Table 2. 1, Fig. 2. 4B). Furthermore, with increased Leu concentrations the total uptake in both hATB0,+ expressing and NI oocytes was increasingly due to activity of xLAT4 (and to a lesser extent xB0AT1) transporters (Figs. 2.3C and 2.4B). In hATB0,+ expressing oocytes, the relative contribution of xLAT4 transporters to total Leu uptake rate was calculated to increase from less than 20% of hATB0,+ activity at 100 µM to approximately 40% higher than that of hATB0,+ at 3 mM Leu (Fig. 2. 4B).

Figure 2. 4. L-leucine and L-phenylalanine uptake by endogenous Xenopus laevis oocyte and exogenously expressed human sodium-dependent symporter SLC6A14/ATB0,+ and sodium-independent uniporter SLC43A2/LAT4 transporters. L-Leucine (Leu) uptake rates (pmol/3 min per oocyte) from non-injected (NI) and oocytes expressing human ATB0,+ (hATB0,+) were compared with model calculations. Measured uptake data (pmol/10 min per oocyte) for each batch of oocytes and experimental day was normalized to uptake rates by NI oocytes in 1 mM Leu, +Na+

uptake buffer. Panel (A) shows measured Leu (10 – 3000 µM) uptake rates for NI oocytes (NI, expt), and for oocytes expressing hATB0,+ without subtraction of NI uptake data (NI+hATB0,+, expt) vs. the calculated model outputs for the respective oocytes (NI, model; and NI+ATB0+, model). Panel (B) shows experimental data for total Leu (0 – 3000 uM) uptake rates by hATB0,+ expressing oocytes with subtraction of NI oocyte uptake rates (+hATB0,+, expt) vs. calculated Leu uptake by endogenous Xenopus laevis oocyte Leu transporters (xB0AT1, model; and xLAT4, model) and exogenous hATB0,+ transporters (+hATB0,+, model ). n = 6 − 8 oocytes per experiment for 7 independent experiments. Panels (C) and (D) show L-phenylalanine (Phe) uptake data (pmol/10min per oocyte) in Na+containing uptake buffer vs. model calculations for endogenous Phe transporters (xB0AT1, xLAT4, xTAT1) and human LAT4 (hLAT4) transporters. Experimental uptake (pmol/10 min per oocyte) for all oocytes was normalized to uptake in 10 mM Phe, Na+ containing- uptake buffer (Na+). Panel (C) shows normalized total Phe uptake rates by NI oocytes (NI, expt) vs. oocytes with expressed hLAT4 without subtraction of NI oocyte uptake (NI+hLAT4, expt) vs. calculated data for the respective oocyte uptakes (NI, model; and NI+hLAT4, model). Panel (D) shows the normalized data for 0 – 10 mM Phe uptake rates by oocytes exogenously expressing hLAT4 with subtraction of NI uptake (hLAT4, expt) vs. model calculations for hLAT4 (hLAT4, model) and xAAT Phe transporters (xB0AT1, model; xLAT4, model; and xTAT1, model). n = 6 − 8 oocytes each experiment for 4 independent experiments. For panels (B) and (D), 95% confidence limits for predicted transporter activities are shown (dotted lines) bracketing the model predictions for each AAT activity.

Prediction of L-phenylalanine transport from L-leucine data

The versatility of the model was further probed by calculating the uptake of a second AA (Phe) based on results for Leu. Specifically, experimental data for Phe uptake by NI and hLAT4 expressing oocytes were compared with calculations of Phe uptake based on the xAAT expression previously determined from Leu data. Six of the potentially expressed xAAT species transport both Leu and Phe (xATB0,+, xB0AT1, xLAT1, xLAT2, xLAT3, xLAT4). Phe uptake rates by xAAT species were calculated using the reported Phe values for the respective exogenously expressed orthologs (Table 2. 1) and푉 푚푎푥 the previously calculated Leu values for the xAAT 0 species (Table 2. 2). As expected, in NI oocytes only activity푉푚푎푥 of xB AT1 and xLAT4 transporters contributed to Phe uptake. In addition, oocytes express mRNA (Xenbase) for the TAT1 transporter species, which transports Phe but not Leu. Using NI uptake measured at 2 Phe concentrations (1000 and 10000 µM), the maximum transport rate 34

for xTAT1 transporters was calculated (equation (4)). Relative Phe for xB0AT1, xLAT4, and xTAT1 transporters (estimated as 2.39±0.39, 14.34±0.34푉푚푎푥 and 408±9.5, respectively) were used together with the corresponding reported Phe values

(Table 2. 1) in the MM equation to calculate the uptake rates from 0 – 10000퐾푚 µM Phe (Fig. 2. 4C). These results indicated that the activity of xTAT1 transporters predominated at all Phe concentrations. The of xTAT1 for Phe transport was

0 approximately 28 times that of xLAT4 and 170 푉times푚푎푥 that of xB AT1 (Fig. 2. 4D). The of xLAT4 for Phe transport was approximately 6 times that of xB0AT1.

0 Furthermore,푉푚푎푥 for both xLAT4 and xB AT1 the calculated for Phe transport was approximately 13X and 27X lower than for Leu, respectively푉푚푎푥 (Figs. 2.3B and 4D). The calculated cumulative uptake rates of Phe by xAATs was found to compare well with the data measured from NI oocytes (Fig. 2. 4C,D).

The hLAT4 to xLAT4 expression ratio was estimated from uptake data for 2 Phe concentrations (1000 µM and 10,000 µM) as 18.56 ± 0.44. The for Phe transport by hLAT4 was estimated (see Methods) and used with reported푉푚푎푥,hLAT4 hLAT4 for Phe (Table 2. 1) to calculate uptake from 0 – 10 mM Phe. The model predictions퐾푚 for cumulative uptake of Phe in both NI and hLAT4 expressing oocytes were found to be in good agreement with the experimental data. For Phe, xTAT1 transporters, in addition to xLAT4 transporters, were calculated to have contributed significantly to uptake by NI oocytes. Overall, these findings support the conclusion that model calculations for relative AAT activity based on Leu data were generalizable to describe the transport of Phe and potentially other substrates (Fig. 2. 4C,D).

Discussion

Using a systems biology strategy, we developed an accurate, robust, and versatile method to calculate the relative contributions of different SLC species in the total cellular transport of a given substrate (Fig. 2. 1). Here we demonstrated this approach as applied to the characterization of Leu and Phe AAT activities in X. laevis oocytes (Figs. 2.3 and 2.4). Using our method the transport rates in diverse uptake buffers by endogenous and exogenous transporters were accurately calculated as verified by the measured data. Quantification of relative endogenous transporter activities

It has long been hypothesized that the AA transport systems B0,+, L, ASC, asc, b0,+ are active in X. laevis oocytes [72]. In this study the activities of potentially expressed endogenous oocyte Leu and Phe AAT species (based on Xenbase mRNA data) were calculated. All of the xAAT species proposed by Van Winkle and co-workers to be expressed were reported by Xenbase to have some level of mRNA expression in unfertilized oocytes. However, since many factors influence protein expression and activity, no conclusions can be drawn about specific xAAT protein activities on the basis of mRNA expression alone. Indeed, based on model output we propose that only two xAAT species, the low affinity Na+-indep uniporter xLAT4, and the low affinity Na+- dep symporter xB0AT1 contribute to endogenous Leu uptake (Fig. 2. 3A,B). Two kinetic components for transport by LAT4 have been reported [45]. From our calculations, only the xLAT4 low affinity component contributed to transport under all assay conditions.

Model parameters

There is a very little information about Xenopus AAT kinetics, therefore, the substrate affinities of human or mouse orthologs were used for modeling (Table 2. 1 and Fig. 2. 2). Given that many AATs are essential and highly evolutionarily conserved, differences in ortholog kinetics that might affect model output, may be relatively minor [73, 74]. Furthermore, physiological parameters such as blood AA levels support the idea that Xenopus transporters likely evolved in similar environments as mammalian AATs. For example, Leu concentration in Xenopus plasma was reported as 150 µM[75] which is consistent with values reported for humans[2] and mice[76]. Additionally, in vitro studies using Xenopus intestines demonstrate similar drug permeabilities as human intestines indicating frog transporter kinetic and substrate specificities mimic human transporter activity[77].Salmon SLC6A19/B0AT1 expressed in Xenopus oocytes was found to have similar kinetics as mouse B0AT1[78]. Additionally, Xenopus Glutamate ionotrophic transporters were found to function, with only minor differences, kinetically like their rat homologs[79]. Taken together these various data support the hypothesis that Xenopus transporters may likely display similar kinetic parameters as their mammalian counterparts.

However, if Xenopus AAT or values differ significantly from reported literature values for mammalian orthologs,푉푚푎푥 퐾 then푚 the calculated activity of xAATs would be correspondingly impacted. Substrate determines the fractional occupancy of a

퐾푚36

transporter for a given substrate concentration, and therefore, calculations for AAT activity. In particular regarding calculations for xLAT3 and xLAT4 transporter activity, it is important to note that the best fit (lowest MRE) results when the xAAT with the lowest affinity is active. Given the mammalian Km values this is assumed to be xLAT4. Since the Leu Km values for xLAT3 and xLAT4 are not definitively known, it is possible that that either xLAT4 or xLAT3 is the predominant endogenous Leu transporter. However, the minimal competition of Leu uptake by 10 mM L-valine (Val) in Na+free buffer is more consistent with xLAT4 than xLAT3 activity; assuming that like their mammalian orthologs xLAT4 has an approximately 20 times lower affinity for Val than xLAT3 (Table 2.1, Figs.2.2, and 2.3C, D).

In general while SLC orthologs are well-conserved structurally and functionally, for some transporters interspecies kinetic differences do exist [80]. Furthermore, while the model calculates the best fit to the overall Leu uptake data as resulting from the predominant activity of xLAT4 and xB0AT1, it is possible that other transporters may have minimal activity. Nevertheless, the strong congruence between model calculations and independently generated experimental data for endogenous and exogenous AATs supports the validity of the model output for X. laevis Leu and Phe transporters (Figs. 2.3 and 2.4).

Versatility of the method

Although, we based our calculations on data generated from Leu concentration- dependence uptake assays carried out in ± Na+ buffers, calculations using other buffer combinations reached similar values (SEM < 10% of the mean). This result demonstrates the versatility of the approach for application to a variety of assay conditions (Table 2. 2). Additionally, our conclusions regarding the activities and expression of transporters are supported by the close concurrence of calculated responses with measured data for different stimuli. Model output was compared with uptake rates in the presence of added excess competitive substrates (Fig. 2. 3C, D), as well as, exogenously expressed human transporters (Fig. 2. 4). Furthermore, the model, which was initially constructed for calculating Leu transporter activities, was shown to be extrapolatable to report activity of Phe transporters (Fig. 2. 4C,D).

Applications

Our systems biology approach provides a method to quantitatively calculate the aggregate cellular transport and relative contributions of specific transporter activities for given substrates in different assay conditions. Some AA, peptide, and other SLC transporters have been shown to carry a broad variety of drugs (e.g., levodopa, antibiotics, ACE inhibitors etc) [81]. Thus, our approach could potentially be helpful in modeling SLC transporter responses to drugs even when the protein expression of candidate transporters is not known. For example, the relative contributions of specific transporters in different cell types for a given drug or combination of drugs could be quantified. Thereby, potential strategies could be assessed for increasing treatment efficacy by modulating drug influx to targeted tissues or tumors. Additionally, therapeutics could be evaluated to decrease unwanted side effects by, for example, increased exclusion and/or efflux of drugs from spurious or deleterious targets. Furthermore, this framework is not limited to SLC transporters and theoretically could be applied to a variety of enzymatic systems. Taken together, the availability of a platform that provides the simultaneous calculation of the contributions of specific enzymes in an enzymatic system and their overall responses to new stimuli promises to be an asset for basic and applied research.

Summary

We established an adaptable systems biology approach to characterize the contributions of specific enzyme species to overall catalytic activity based on limited experimental input. Using this approach, we could accurately calculate responses to new stimuli. In this study, we applied the strategy to characterize Leu and Phe SLC AAT species using the X. leavis oocyte in vitro cell model. However, given the appropriate assay conditions, this approach is applicable to many other enzymes, substrates, and/or cellular systems.

Acknowledgements

We kindly acknowledge the financial support for this work from SystemsX.ch - The Swiss Initiative in Systems Biology through the QuantX project, and from the Swiss National Science Foundation through NCCR Kidney.CH.

Author Contributions

MT, VK, FV, UO and VM conceived and designed the project. MT did the experiments with help from LO. MT analyzed the data with help of UO and VM. MT, VM, UO, VK and FV wrote the manuscript. All authors read and approved the final version.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary information

Supplementary Figure 2. 1. Time course of L-leucine uptake by non-injected Xenopus laevis oocytes. L-leucine uptake rates by non-injected oocytes was tested by radiolabeled amino acid tracer experiments at ten time points (1, 2, 3, 4, 8, 12, 13, 15, 21, 25 minutes) with 1 mM unlabeled L-leucine in 100mM sodium containing uptake buffer (+Na+) The linear fit to the uptake data (pmole per oocyte) for 1-8 minutes was calculated and the data graph prepared using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA) (n = 6 − 8 oocytes per time point for one experiment).

Chapter 3 Functional polarity of microvascular brain endothelial cells supported by neurovascular unit computational model of large neutral amino acid homeostasis

This chapter has been published as: 39

Mehdi Taslimifar, Stefano Buoso, François Verrey, and Vartan Kurtcuoglu. "Functional Polarity of Microvascular Brain Endothelial Cells Supported by Neurovascular Unit Computational Model of Large Neutral Amino Acid Homeostasis." Frontiers in Physiology 9 (2018): 171.

Abstract

The homeostatic regulation of large neutral amino acid (LNAA) concentration in the brain interstitial fluid (ISF) is essential for proper brain function. LNAA passage into the brain is primarily mediated by the complex and dynamic interactions between various solute carrier (SLC) transporters expressed in the neurovascular unit (NVU), among which SLC7A5/LAT1 is considered to be the major contributor in microvascular brain endothelial cells (MBEC). The LAT1-mediated trans-endothelial transport of LNAAs, however, could not be characterized precisely by available in vitro and in vivo standard methods so far. To circumvent these limitations, we have incorporated published in vivo data of rat brain into a robust computational model of NVU-LNAA homeostasis, allowing us to evaluate hypotheses concerning LAT1- mediated trans-endothelial transport of LNAAs across the blood brain barrier (BBB). We show that accounting for functional polarity of MBECs with either asymmetric LAT1 distribution between membranes and/or intrinsic LAT1 asymmetry with low intraendothelial binding affinity is required to reproduce the experimentally measured brain ISF response to intraperitoneal (IP) L-tyrosine and L-phenylalanine injection. On the basis of these findings, we have also investigated the effect of IP administrated L-tyrosine and L-phenylalanine on the dynamics of LNAAs in MBECs, astrocytes and neurons. Finally, the computational model was shown to explain the trans-stimulation of LNAA uptake across the BBB observed upon ISF perfusion with a competitive LAT1 inhibitor.

Keywords: blood brain barrier, neurovascular unit, amino acid transporter, large neutral amino acid, SLC7A5/LAT1

Introduction

The blood-brain barrier (BBB) is a truly dynamic interface separating the brain from the bloodstream. It is formed by highly specialized microvascular brain endothelial cells (MBECs) connected by tight junctions forming brain capillaries. The BBB endothelium together with the astrocytes and neurons are the fundamental elements of the neurovascular unit (NVU) system. 40

Numerous solutes move across the NVU cell membranes with various transport mechanisms. While small lipophilic molecules can diffuse, larger and hydrophilic solutes, such as amino acids (AAs), need the assistance of specialized carrier proteins to cross the membrane, for instance amino acid transporters (AATs) [82]. NVU-AATs are expressed at both luminal and abluminal membranes of the MBECs, as well as on astrocytes and neurons. The NVU-AATs mediate the transfer of particular amino acids with different transport mechanisms: antiporters, for example, exchange some AAs for others across the membrane, while symporters cotransport AAs together with ions along the ions’ electrochemical gradient [83]. Taken together, different classes of NVU-AATs constitute an integrated dynamic system controlling the homeostasis of AAs such as large neutral amino acids (LNAAs: L-tyrosine, L-leucine, L- isoleucine, L-phenylalanine, L-histidine, L-valine, L-tryptophan and L-methionine) in the brain interstitial fluid (ISF). The homeostasis maintenance of LNAA concentrations, which have been shown to be asymmetrically distributed in the plasma and individual NVU compartments [76, 84, 85], is of particular importance due to their crucial role in the central nervous system (CNS), for instance as precursors of key neurotransmitters such as Dopamine, Serotonin, and Histamine. Figure 2. 1 illustrates a simplified model of the adult rat NVU that includes the dominant LNAA transporter of each cell membrane. The NVU-LNAAs have been shown to be transported mainly, but not exclusively, by SLC7A5 (LAT1), SLC6A15 (B0AT2) and/or SLC7A8 (LAT2). LAT1 associated with the accessory subunit 4F2hc (SLC3A2) functions as a Na+-independent antiporter and plays a dominant role at the luminal and abluminal membranes of the MBECs [24-26]. B0AT2 is a Na+-dependent symporter which has been shown to be the dominant uptake pathway for LNAAs in neurons [30-32]. A number of studies have shown that the Na+- independent antiporter LAT2 also associated with 4F2hc is the major mediator of LNAA transport in primary astrocytes [27-29]. While it has to be mentioned that comparably high LAT1 mRNA levels have been detected by Zhang Y, et al. [86] in freshly isolated astrocytes, the functional contribution of this transporter remains unclear [29]. In vivo assays and in vitro measurements carried out on freshly isolated cells have shown that the expression of other AATs, such as y+LAT2/SLC7A6 (SLC7A6) and ASCT2/SLC1A5 (SLC1A5), is very low in adult brain compared to the aforementioned AATs [87-89]. Therefore, based on the available evidence in the literature, we consider LAT1, LAT2, and B0AT2 to be the predominant LNAA transporters in MBECs, astrocytes and neurons, respectively. Taken together, these transporters co-operate as a highly complex and integrated dynamic system to predominantly control the homeostasis of LNAAs in the brain ISF. For example, LNAAs in the brain ISF can be taken up by B0AT2 localized in neurons, and/or they can be exchanged with other LNAAs of astrocytes (mediated by LAT2) and/or be transported back into MBECs and eventually into

the bloodstream via LAT1 expressed at the abluminal and luminal membranes of the MBECs (Fig. 3. 1).

Figure 3 1. Diagram of the dominant LNAA transporters expressed in cells of the neurovascular unit (NVU). The diagram represents the major compartments of the brain with the dominant NVU carrier-mediated LNAA transport pathways from brain capillary plasma (input) across blood brain barrier (BBB) microvascular endothelial cells (MVEC) into the interstitial fluid (ISF) and from there into astrocytes and neurons. The abbreviations used for the NVU-SLC transporters are LAT1 (SLC7A5) and LAT2 (SLC7A8), both Na+-independent large neutral amino acid antiporters, and B0AT2 (SLC6A15), a Na+-dependent large neutral amino acid symporter. The arrows indicate the transmembrane pathways of LNAAs via these transporters into and out of the NVU cells. TL and CL represent test and competing large neutral amino acids, respectively.

Among the aforementioned dominant transporters, LAT1 is the central element of the NVU that is involved in the regulation of LNAA homeostasis in the brain ISF. However, despite its importance, its bi-directional kinetic behavior across the BBB has not been characterized yet. We have previously investigated the bi-directional kinetics of LAT1 using the Xenopus laevis oocyte expression system, and observed strongly asymmetric bi-directional kinetics (high extra-cellular versus low intra-cellular binding affinity) [26, 33], a finding that has recently been confirmed by reconstitution experiments in proteoliposomes [90]. However, it remains unclear

whether this bi-directional asymmetry is dependent on the cell type, and whether it may be influenced by the regulatory/modulatory function of gene products absent in Xenopus laevis oocytes and possibly present in other cell types such as MBECs [26, 33]. In vivo tracking of LNAAs from MBECs toward blood plasma and ISF could provide information on the bi- directional kinetic behavior of MBEC LAT1. However, there is currently no suitable in vivo technique available to achieve this. While bi-directional uptake and efflux assays using in vitro models of MBECs could be used, they may not reflect the in vivo situation because of the high sensitivity of the expression level of AATs to culture conditions [34]. In addition to the unclarity regarding the bi-directional kinetics of LAT1 in MBECs, the abluminal to luminal expression ratio of LAT1 at the BBB is not well known yet. Only a study carried out in isolated vesicles has characterized the relative expression of LAT1 at the BBB [91], but this approach may not reflect the situation in vivo [35]. Taken together, for the above mentioned reasons, the bi-directional kinetic behavior of LAT1 in MBECs as well as its distribution pattern at the luminal and abluminal membranes of the BBB could so far not be addressed satisfactorily. To circumvent these limitations, we have developed a robust computational model of LNAA homeostasis in the NVU based on a mathematical description of the nonlinear mechanistic kinetics of the dominant individual NVU-LNAA transporters in conjunction with published in vivo LNAA microdialysis (MD) measurements performed in the rat brain ISF upon intraperitoneal administration of L-tyrosine and L-phenylalanine [92, 93]. This has allowed us to explore potential asymmetries of LAT1 bi-directional kinetics and expression in MBECs. Our computations support the hypothesis that MBECs exhibit a functional polarity for LNAAs due to an asymmetry in either bi-directional kinetics and/or expression of LAT1 in MBECs. In addition, we have employed our model to capture changes in LNAA levels in MBEC, astrocytes, and neurons upon perturbations of plasma LNAA concentrations. Finally, we employed the computational model to explain the trans-stimulation of LNAAs upon ISF perfusion of MBEC LAT1 competitive inhibitor.

Methods

Transport model

The NVU is represented by four interacting compartments for MBEC, ISF, astrocytes, and neurons, each with a homogeneous mixture of LNAAs. The plasma conditions are prescribed as dynamic inputs to the NVU (Fig. 3. 1). Carrier-mediated transport of LNAAs between the compartments is represented by fluxes dominantly mediated by AATs located at the interface between compartments [21, 22, 94]. Following these modeling assumptions, temporal changes in the test LNAA (TL) concentration within the individual NVU compartments are given by 43

MBEC , d[TL] 1 P→MBEC MBEC→ISF 1 MBEC LAT1,lum LAT1,abl dt = V (f − f )

ISF 2 d[TL] 1 MBEC→ISF ISF→Ast ISF→Neu0 ISF LAT1,abl LAT2 B AT2 dt = V (f − f − f ) ,

Neu ISF→0 Neu d[TL] f B AT2 3 = , dt VNeu Ast ISF→Ast d[TL] f LAT2 4 = Ast , wheredt andV represent the concentration of the test LNAA in the compartment i and the i volume [TL of ] that compartment,Vi respectively. The carrier-mediated flux of test LNAA from compartment to compartment is denoted with , and and refer to i→j plasma, microvasculari brain endothelialj cell, brainf AAT interstitialP, MBEC, fluid, ISF, neuron Neu, andAst astrocyte, respectively. Subscript and refer to the luminal and abluminal membranes of the

MBEC, respectively. lum abl

The fluxes of LNAAs between NVU compartments depend on the mechanism of the individual transporters and their dependence on (or independence of) sodium ions. LAT1 and LAT2 are sodium independent antiporters, while B0AT2 functions as a sodium dependent symporter [26, 32]. The fluxes mediated by these transporters are given by [21, 22, 94, 95]

, P MBEC MBEC P 5 P→MBEC 2Vmax,LAT1,lum,TL ([TL] [CL] −[TL] [CL] ) LAT1,lum K P P P MBEC m,LAT1,TL P MBEC f = Km,LAT1,TL ([TL+CL] +[TL+CL] ) + ( MBEC+1) ([TL+CL] [TL+CL] ) Km,LAT1,TL , MBEC ISF ISF MBEC 6 MBEC→ISF 2Vmax,LAT1,abl,TL([TL] [CL] −[TL] [CL] ) LAT1,abl K MBEC f = MBEC MBEC ISF m,LAT1,TL MBEC ISF Km,LAT1,TL ([TL+CL] +[TL+CL] )+( ISF +1)([TL+CL] [TL+CL] ) Km,LAT1,TL

ISF Ast Ast ISF 7 ISF→Ast 2Vmax,LAT2,TL ([TL] [CL] − [TL] [CL] ) f LAT2 = ISF , ISF ISF Ast Km,LAT2,TL ISF Ast Km,LAT2,TL ([TL + CL] + [TL + CL] ) + ( Ast + 1) ([TL + CL] [TL + CL] ) Km,LAT2,TL 0 ISF→Neu 2Vmax,B AT2,TL ′ ISF Neu ISF Neu Neu ISF f B0AT2 = (휀휀 [Na] [Na] ([TL] [CL] − [TL] [CL] D ′ ISF ISF Neu Neu + 휀 [TL] [Na] Km,B0AT2,CL Km,B0AT2,Na Neu Neu ISF ISF − 휀 [TL] [Na] Km,B0AT2,TL Km,Na ), ISF Neu ′ ISF Neu Neu Neu ISF D = [Na] [Na] (휀 [TL + CL] ( [TL + CL] + Km,CL ) + 휀 [TL + CL] ( [TL + CL] + 8 0 ISF ISF 0 Neu 0 Neu ISF ′ Km,B AT2,TL )) + [Na] Km,B AT2,CL Km,B AT2,Na [TL + CL] (휀 + 1) + Neu 0 ISF 0 ISF Neu [Na] Km,B AT2,TL Km,B AT2,Na [TL + CL] (휀 + 1) + 0 ISF 0 Neu ISF 0 Neu Neu 0 ISF Km,B AT2,TL Km,B AT2,CL ([Na] Km,B AT2,Na + [Na] , Km,B AT2,Na ) + 0 ISF 0 Neu 0 ISF 0 Neu 2 Km,B AT2,TL Km,B AT2,CL Km,B AT2,Na Km,B AT2,Na = β z F and = (β−1)z F , ( R T Δψ) ′ ( R T Δψ ) 휀 푒 휀 푒 44

Where represents the concentration in compartment i of LNAAs competing with the test i LNAA, and[CL] and are the maximum transport rate of the AATs for the test and competing Vmax,AAT, LNAATL (competitive Vmax,AAT inhibitors,CL of test LNAA), respectively. In Eq. (8), and are ′ the electrical potential-induced biases for forward and backward transport rates, respecti휀 휀vely, and , , , , , and represent potential difference, electrical bias constant, Faraday constant,Δψ β sodiumF z Rcharge,T gas constant and absolute temperature, respectively [22, 94, 95]. and are, respectively, the AAT apparent Michaelis-Menten binding i i constantsKm,AAT,TL for theK m,AAT, test CL and competing LNAAs in the presence of competitors. They are determined by [25, 96]

, i i i [CL] m,AAT,TL m,abs,AAT,TL i 9 K = K (1 + Km,abs,AAT,CL ) , i i i [TL] m,AAT,CL m,abs,AAT,CL i whereK = K and (1 + Km,abs,AAT,TL are,) respectively, the AATs absolute Michaelis-Menten i i bindingK constantsm,abs,AAT,TL for testK andm,abs,AAT, competingCL LNAAs in the absence of competitors [25, 96]. For simplicity, the competing LNAAs are treated as a single-entity component, representing the overall concentration of the mixture of individual competing LNAAs (Fig. 3. 1). The maximum transport rate and the overall absolute Michaelis-Menten binding constant for the competing LNAA, and , respectively, are given by [97, 98]: i max,AAT, V CL Km,abs,AAT,CL

n max,AAT,CLk k k=1 V [CL ] ∑ ( m,abs,AAT,CLk ) max,AAT,CL K V = n k , 10 k=1 [CL ] ∑ ( m,abs,AAT,CLk) K n 푖 j=1 k m,abs,AAT,CL ∑ [CL ] K = n k , j=1 [CL ] ∑ ( m,abs,AAT,CLk) where and K represent, respectively, the concentration and the absolute

Michaelis-Menten[CLk] K bindingm,abs,AAT, constantCLk of the individual competing LNAAs within the considered mixture (see Suppl. Table 3. 1), and where n is the total number of individual competing LNAAs.

The MBEC LAT1 bi-directional kinetics are modeled as

MBEC P(ISF) 11 Km,abs,LAT1 = RKLAT1 Km,abs,LAT1 , where is the LAT1 bi-directional kinetic constant, which represents the absolute

Michaelis-Menten RKLAT1 binding constant for LAT1 in MBECs relative to the corresponding value at

the outside of MBECs in the ISF and in plasma. The LAT1 expression ratio in MBECs is modeled as

Vmax,LAT1,abl = RELAT1 Vmax,LAT1,lum, where represents the relative ratio for the maximum transport rate of LAT1 at the abluminal RE membraneLAT1 of the MBECs to the corresponding value at the luminal membrane. Eqs. (5-12) and Eqs. (1-4) can be combined to describe the intercompartmental rate of change in

the concentration of the test LNAAs, i), as a system of nonlinear ordinary differential d[TL] equations of the following general form:( dt

i 13 d[TL] i i i i = function ([TL] , [CL] , Vi , Km,AAT,TL , Km,AAT,CL ,Vmax,AAT,TL,Vmax,AAT,CL, RKLAT1, RELAT1) dt The intra-compartmental concentration change rate of the competing LNAAs ( i) can be d[CL] formulated similarly. Values for kinetic parameters of individual AATs dt , i and ) and volumes of compartments ( used (Km,abs,AAT, in Eqs. TL(1- i 13) Km,abs,AAT, are listedCL ,Vinmax Table,AAT, 3.TL 1. Vmax,AAT,CL Vi)

Model initialization and numerical model

To capture the responses of individual NVU compartments (MBEC, ISF, astrocyte, and neuron) to perturbations in plasma LNAA concentration, the baseline (pre-stimulus or pre- injection) state of the system needs to be determined. To this end, we first obtain the steady-

state solution of Eq. 13 ( i = i ) by prescribing constant plasma concentrations of d[TL] d[CL] LNAAs as NVU system input dt (Fig. dt3.1 =and 0 Suppl. Table 3. 2) and solving the resulting system of equations whose unknowns are the baseline LNAA concentrations in the individual compartments. To do so, we are required to initialize the LNAA concentrations in individual NVU compartments. The LNAA concentrations in the ISF, astrocytes, and neurons are initialized according to baseline values reported in the literature ( and ) (Suppl. Table i i 3. 2). Such information is not available for MBECs, however.[TL Therefore,]b [CL ] wbe initialize the corresponding LNAA concentration based on a parametric study obtained with random values of the initial baseline concentration (Suppl. Table 3. 2). It has to be noted that once the LNAA concentrations in the different compartments have been prescribed, the solution of the steady- state problem is constrained in the total amount of LNAAs in the NVU. We examined whether this amount reflects in vivo conditions by extrapolating the calculated compartmental LNAA concentrations to the brain as a whole and comparing these values to experimental results

reported in [85, 99] , finding very good agreement (Suppl. Table 3. 2). Once the baseline or pre-stimulus state of the NVU system is determined, we calculate the post-stimulus state of the NVU in response to perturbations of LNAA concentrations in the plasma.

All amino acid transport models were implemented in Matlab (R2015a). To calculate the concentration of LNAAs in the individual NVU compartments (pre- and post-stimulus states), we performed the time integration of Eq. 13 using the ode23s function (Bogacki–Shampine method) [100, 101]. The source code from Panitchob et al [94] has been used as a starting point for our implementation.

Results

Computational model combined with in vivo brain ISF measurements support a functional polarity of MBECs

To discriminate, using our new computational model of the NVU, the hypothesized effects of asymmetry on bi-directional kinetics and expression of LAT1 in MBECs (see Introduction), we first searched the literature for kinetic parameters of LNAA transporters of the individual NVU compartments (Table 3. 1). Most carefully measured kinetic parameters of transport at the endothelial barrier reported by Smith et al.[25] were obtained by using in situ brain perfusion with short uptake times and thus likely represent the kinetics of the first step of LNAAs transport that is into MBECs across their luminal membrane [76, 92, 93, 102] and are thus not representative of steady-state trans-MBEC transport. Using these kinetic parameters, we first considered the bidirectional kinetics of LAT1 to be symmetric in MBECs ( =1) and also assumed LAT1 to be symmetrically expressed at the luminal and abluminalRKLAT1 membranes of MBECs ( =1). Under these assumptions of symmetry, we compared the output of our computationalRELAT1 model with in vivo measurements made by Bongiovanni et al.[92]. In their study, they had increased the plasma level of L-tyrosine (test LNAA) by intraperitoneal (IP) injection in awake rats and simultaneously measured the post-stimulus response in the brain ISF by microdialysis. Using their measured plasma-stimulus profiles of the test LNAA L-tyrosine ( ) and of the L-tyrosine competing LNAAs (competitive inhibitors) ( ) as input to the P P model[TL] (Fig. 3. 2A, results reported as a percentage of baseline),[ CL we] calculated the corresponding post-stimulus responses in the brain ISF and found a significant mismatch between the measured and our calculated results which showed a larger excursion due to a much faster transport rate across MBECs (see results of statistical analysis in Fig. 3. 2B and Suppl. Table 3. 4).

Figure 3 2. Plasma concentration and corresponding brain ISF concentration response after intraperitoneal injection of L-tyrosine and L-phenylalanine. Panel (A) shows the plasma concentration of L-tyrosine (TL) and L-tyrosine competing LNAAs (CL) after intraperitoneal administration of 200 mg/kg L-tyrosine as measured by Bongiovanni et al. [92] and used as input for the model calculation. Panels (B) and (C) show the experimental data for the L-tyrosine (Tyr) post-stimulus response in the brain ISF, measured in the prefrontal cortex (PFC). Panel (B) shows the model calculations for various ratios of the bi-directional kinetic constant of MBEC LAT1 ( , Eq.11) with symmetric distribution of LAT1 at both luminal and abluminal membranes of the BBB ( ). Panel (C) shows the model RKLAT1 calculations for various abluminal to luminal expression distribution ratios of LAT1 ( , RELAT1 = 1 Eq.12) with symmetric bi-directional kinetics ( ). The model results and experimental RELAT1 data are represented as percent of the baseline value. In panel (A), the plasma baseline value RKLAT1 = 1 48

for L-tyrosine and L-tyrosine competing LNAAs (constant input) are 112 and 535 µM [84, 92], respectively. In panels (B) and (C), the ISF baseline value for L-tyrosine is 1.0 and 1.1 µM (Suppl. Table 3. 2), respectively. Each experimental data point represents the mean ± SD for three (plasma) and four to eight (ISF) animals [92]. In panel (A), the CL refers to a mixture of L-tyrosine competing LNAAs (L-leucine, L-isoleucine, L-phenylalanine, L-tryptophan, L-valine, L-histidine and L-methionine). The error bars associated with model calculations indicate standard deviation with respect to concentrations obtained with the nominal model parameter set (see Methods). Panel (D) shows the measured plasma concentration of L-phenylalanine (TL) and L-phenylalanine competing LNAAs (CL) after intraperitoneal administration of 200 mg/kg L-phenylalanine as measured by [93, 103]. Panels (E) and (F) show the experimental data for the L-phenylalanine (Phe) post-stimulus response in the brain ISF, measured in the prefrontal cortex (PFC) versus model calculations for different ratios for the bi-directional kinetic constant of MBEC LAT1 ( , Eq.11), assuming symmetric distribution for LAT1 at luminal and abluminal membranes of the BBB ( ) and the model calculations for RKLAT1 various abluminal to luminal expression distribution ratios of LAT1 ( , Eq.12), assuming RELAT1 = 1 symmetric bi-directional kinetics of MBEC LAT1 ( ). In panels (E) and (F), the ISF RELAT1 baseline value for L-phenylalanine is 0.4 µM (Suppl. Table 3. 2). The data are represented as RKLAT1 = 1 percent of baseline. In panel (D), the plasma baseline value for L-phenylalanine and L- phenylalanine competing LNAAs (constant input) are 77 and 562 µM [84, 92], respectively. In panel (D), the CL refers to a mixture of LNAAs competing with the test amino acid L- phenylalanine (L-leucine, L-isoleucine, L-tyrosine, L-tryptophan, L-valine, L-histidine and L- methionine). In Panels (B), (C), (D) and (E), the differences between the concentrations calculated with the symmetric model ( and ) and the experimental measurements are statistically significant at all post-stimulus time points (p<0.001, Suppl. RKLAT1 = 1 RELAT1 = 1 Table 3. 4). In contrast, there is no significant difference between the experimental measurements and the model calculations with and (Panel B), and (Panel C), and (Panel E) and RKLAT1 = 160 RELAT1 = 1 and (Panel F) with the exception of the 30 min post-stimulus time RKLAT1 = 1 RELAT1 = 0.18 RKLAT1 = 80 RELAT1 = 1 point in Panels E and F (Suppl. Table 3. 4). RKLAT1 = 0.11 RELAT1 = 1

Table 3. 1. Model input parameters -L tyrosinea L-phenylalanineb Parameters Value Unit Ref. LAT1 (MBEC)

64 11 µM [25] P(ISF) Km,abs,LAT1,TL 0.175 0.075 µmol/min [25, 104] c Vmax,LAT1,lum,TL 37c 52.9 µM [25] P(ISF) c Km,abs,LAT1,CL 0.086c 0.129 µmol/min [25, 104] VLAT2max,LAT1,lum, (Astrocyte)CL 294d 110.2d µM [28] ISF(Ast) Km,abs,LAT2,TL 0.1128 0.1128 µmol/min [59, 105] c c Vmax,LAT2,TL 163.6 185.9 µM [28] ISF(Ast) c Km,abs,LAT2,CL 0.1452c 0.1494 µmol/min [59, 105] 0 VBmax,LAT2,AT2 (Neuron)CL NA 1050 µM [32] 0 ISF(Neu) Km,abs,B AT2,TL NA 0.0086 µmol/min [32, 106] 0 Vmax,B AT2,TL 123.5c 126.2c µM [32] ISF(Neu) 0 c Km,abs,B AT2,CL 0.0184c 0.0186 µmol/min [32, 106] 0 Vmax,B AT2,CL 1050 1050 µM [107] 0 ISF(Neu) ΔΨKm,B AT2,Na -70 -70 mV [108] β 0.6 e 0.6 e mV [94, 107] 141 141 mM [109] ISF [Na] 40 40 mM [110] Neu [VolumeNa] 3.5 µl [109, 111]

VMBEC 352.6 µl [104, 112] VISF 742 µl [113, 114] VAst 441.7 µl [114-116] a b VInNeu this column, TL and CL represent L-tyrosine and L-tyrosine competing LNAAs, respectively. In this column, TL and CL represent L- phenylalanine and L-phenylalanine competing LNAAs, respectively. cThe kinetic parameters for the mixture of L-tyrosine and L- phenylalanine competing LNAAs are calculated based on Eq.10 (Suppl. Table 3. 1). d The kinetic parameters are calculated based on Michaelis-Menten equation. e Estimated based on data by [107] , fig 7.D. NA (not applicable) specifies the large neutral amino acid was not reported to be a substrate for the transporter. For calculation of Vmax values, the total rat brain weight, volume and protein content are considered 1.81 g [117] , 1737 µl [104] and 105 mg protein /g brain [118], respectively.

We then evaluated whether asymmetric bi-directional kinetics of LAT1 in MBECs could explain the slower and less important impact of plasma L-tyrosine perturbation on its ISF concentration observed in vivo, compared to our first calculations made assuming symmetric transport properties of LAT1. To this end, we varied the ratio of extracellular to intracellular Michaelis- Menten binding constants of LAT1 in MBECs, named here , from 1 (representing the symmetric bi-directional kinetic) to 1300 (highly asymmetric bi-directi RKLAT1 onal kinetics as described for LAT1 in Meier et al. [26] and considered LAT1 to be symmetrically distributed at the BBB ( =1). We calculated the post-stimulus LNAA concentration response and compared the results RELAT 1with the in vivo measurements (shown as percentage of baseline in Fig. 3. 2B). Under consideration of asymmetric bi-directional kinetics for LAT1 in MBECs, the numerical results agreed well with in vivo experimental data, best for a bi-directional kinetic constant of =160. Thus, the results obtained with our model support the hypothesis that LAT1

RKdisplaysLAT1 a strong asymmetry in bi-directional kinetics in MBECs.

We then evaluated the alternative or complementary hypothesis that an asymmetry of LAT1 expression at the luminal and abluminal membranes of MBECs could explain the observed equilibration kinetics. To this end, we varied the LAT1 expression constant at the BBB ( between 0.01 to 10 (representing highly symmetric abluminal to luminal expression ratio)RELAT an1d) compared the numerical calculations with the in vivo measurements assuming symmetric bi- directional kinetics of the MBEC LAT1 ( ) (plotted as percentage of baseline in Fig.

3. 2C). The error bars associated with modelRKLAT1 simulations= 1 are calculated based on sensitivity studies (see Sensitivity analysis section). In contrast to the symmetric case, the numerical results obtained for asymmetric transporter expression agreed well with in vivo experimental data, best for an expression kinetic constant of =0.18 (see Fig. 3. 2C). These results are compatible with the hypothesis of a strong asymmetryRELAT1 in the expression of the LAT1 in MBECs with lower expression at the abluminal membrane. Taken together, the computational model, combined with in vivo measurements supports a functional polarity of MBECs with either asymmetry in bi-directional kinetics and/or expression distribution of LAT1 in MBECs.

Cross-substrate versatility

We next evaluated whether our conclusion on the functional polarity of MBECs described in the previous section depends on the substrate by comparing our calculations with in vivo data published by Goldstein 28 and Bongiovanni et al.18 in which the ISF response was measured after IP administration of L-phenylalanine in awake rats (Fig. 3. 2F). Just as with the L-tyrosine case, the model failed to reproduce the experimental measurements when assuming symmetric bi-directional kinetics for LAT1 in MBECs, whereas we found a close match between

our model calculations and experimental measurements assuming asymmetric LAT1 bi- directional kinetics (best with 80). Similarly, the numerical results obtained when assuming an asymmetric transporter RKLAT1 expression= also agreed well with in vivo experimental data, best for an expression kinetic constant of = 0.11. Taken together, the comparison of model output with experimental measurementsRELAT supports1 the hypothesis that the MBECs show a functional polarity for both L-tyrosine and L-phenylalanine which could be explained by either asymmetric distribution of LAT1 at the luminal and abluminal membranes of the MBECs (lower abluminal expression) and/or a strong asymmetry in its bi-directional kinetics in MBECs (lower intracellular affinity) as previously shown in Xenopus oocytes.

To further evaluate the dependence of our results on the asymmetric function of LAT1 suggested for MBECs, we checked whether considering LAT1 as dominant astrocytic AAT instead of LAT2 would modify our conclusion on the functional polarity of the MBECs. Calculations presented in the Supplementary material (Supp. Fig. 3. 1) showed that this is not the case.

Calculating the post-stimuli responses in MBECs, astrocytes, and neurons

The in vivo standard methods have so far not been able to address the effects of plasma LNAA perturbations on the dynamics of LNAA concentrations in individual NVU compartments. To close this gap, we employed the computational model considering either an asymmetry in bi- directional kinetics of LAT1 or an asymmetry in the expression pattern of LAT1 in MBECs as determined for the best cases in Fig. 3. 2. The dynamic responses of L-tyrosine (TL) and L- tyrosine competing LNAAs (CL) and of L-Phenylalanine (TL) and its competitor LNAA (CL) in MBECs, ISF, astrocytes, and neurons are shown in Fig. 3. 3 and Fig. 3. 4 for asymmetric bi- directional kinetics and asymmetric expression of LAT1, respectively. The same plasma perturbations of the test LNAAs (L-tyrosine or L-phenylalanine) used also for Fig. 3. 2 are shown to first propagate into the MBECs (Figs. 3.3A, E and 3.4A, E). The dynamics of this propagation depend on the competitions between the test and competing LNAAs through MBEC LAT1 and its kinetics for each substrate. Since MBEC LAT1 functions as an antiporter, the elevated level of the test LNAA in the MBECs leads to an initial reduction in the MBEC level of the competing LNAAs (Figs. 3.3A, E and 3.4A, E). Subsequently, the test and competing LNAAs compete for efflux via LAT1 across the abluminal membrane of the BBB MBECs and eventually gain entry into the brain ISF in exchange for competing LNAA of the ISF (Figs. 3.3B, F and 3.4B, F). The observed delayed response in the concentration of the test LNAAs in brain ISF in response to the plasma perturbations is mainly due the low inter- endothelial affinity of LAT1 (Fig. 3. 3B) and/or a low expression of LAT1 at the abluminal

membrane of the BBB (Fig. 3. 4B) both of which would strongly limit the trans-endothelial transport of LNAAs across the BBB. Once the test (and competing) LNAAs enter the brain ISF, they are differentially co-transported together with sodium ions into neurons via B0AT2 (Figs. 3.3D, H and 3.4D,H) and exchanged back into the MBECs (Figs. 3.3A,E and 4A,E) and astrocytes (Figs. 3.3C,G and 3.4C,G) via LAT1 and LAT2, respectively. The rate of these transports depends on LNAA concentration, on that of competitor LNAAs and on the kinetics of the transporters expressed at the interface to the other NVU compartments (Table 3. 1). For example, LNAA transport from ISF to MBECs is comparably low due to the relatively low concentration of the LNAAs compared to their Michaelis-Menten binding affinities (Table 3. 1 and Suppl. Table 3. 2). As shown in Figs. 3.3C, G and 3.4C, G, astrocytic elevation of the test LNAAs is associated with the reduction of intracellular competing LNAAs. This behavior is due to the exchange mechanism of LAT2 localized at the membrane of astrocytes. In Figs. 3.3D and 3.4D, while L-tyrosine transport is shown not to be mediated by B0AT2 (Table 3. 1), this LNAA could nonetheless be transported to some extent into neurons by other, less expressed transporters (see Discussion section). Taken together, the difference between the response of L-tyrosine and L-phenylalanine results from various factors such as their differing original perturbation dynamics in plasma (Figs.3.2 A and D) and the transport kinetics differences of the NVU-AATs for these substrates and their competitors (Eq. 13 and Table 3.1).

Figure 3.3. The post-stimulus response in MBECs, ISF, astrocytes and neurons after intraperitoneal administration (IP) of L-tyrosine and L-phenylalanine for asymmetric bi- directional kinetics of LAT1 in MBECs. Panels (A),(B),(C) and (D) and (E),(F),(G) and (H) show the model calculations for the post-stimulus responses in the NVU individual compartments after IP administration of L-tyrosine 160 and 1) and L- phenylalanine 80 and 1), respectively. The error bars associated with (RKLAT1 = RELAT1 = (RKLAT1 = RELAT1 = 54

model calculations indicate standard deviation with respect to concentrations obtained with the nominal model parameter set. In panels (A), (B), (C) and (D), CL refers to a mixture of L- tyrosine competing LNAAs (L-leucine, L-isoleucine, L-phenylalanine, L-tryptophan, L-valine, L- histidine and L-methionine). In panels (E), (F), (G) and (H), CL indicates a mixture of L- phenylalanine competing LNAAs (L-leucine, L-isoleucine, L-tyrosine, L-tryptophan, L-valine, L- histidine and L-methionine). The ISF post-stimulus response for TL in panels (B) and (F) are replotted from Figs. 3.2B and 3.2E, respectively. In all panels, the baseline concentration for L-tyrosine, L-tyrosine competing LNAAs, L-phenylalanine and L-phenylalanine competing LNAAs are reported in Suppl. Table 3. 2.

Figure 3.4. The post-stimulus response in MBECs, ISF, astrocytes and neurons after intraperitoneal administration (IP) of L-tyrosine and L-phenylalanine for asymmetric expression distribution of LAT1 at luminal and abluminal membranes of the BBB. Panels (A,B,C,D) and (E,F,G,H) show the model calculations for the post-stimulus responses in the NVU individual compartments after IP administration of L-tyrosine 0.18 and 1) and L-phenylalanine 0.11 and 1), respectively. The error bars associated with (RELAT1 = RKLAT1 = LAT1 LAT1 (RE = RK =56

model calculations indicate standard deviation with respect to concentrations obtained with the nominal model parameter set. In panels (A), (B), (C) and (D), CL refers to a mixture of L-tyrosine competing LNAAs (L-leucine, L-isoleucine, L-phenylalanine, L-tryptophan, L-valine, L-histidine and L-methionine). In panels (E),(F),(G) and (H), CL indicates a mixture of L-phenylalanine competing LNAAs (L-leucine, L-isoleucine, L-tyrosine, L-tryptophan, L-valine, L-histidine and L- methionine). The ISF post-stimulus response for TL in panels B and F are replotted from Figs. 3.2C and 3.2F. In all panels, the baseline concentration for L-tyrosine, L-tyrosine competing LNAAs, L-phenylalanine and L-phenylalanine competing LNAAs are reported in Suppl. Table 3. 2.

The trans-stimulation of the test LNAA uptake across the BBB upon ISF perfusion with a LAT1 competitive inhibitor

Finally, we employed the established computational model to investigate the induced effects of brain ISF perfusion with 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid (BCH, a transported competitive inhibitor of LAT1 and LAT2) on the dynamics of test LNAAs in the brain ISF [83]. We have shown recently that continues perfusion of 20 mM BCH (~ 2mM local concentration near the perfusion probe [76]) into the brain ISF of freely moving mice trans- stimulates the LAT1 functions at the BBB and consequently changes the dynamics of LNAAs in the brain ISF in exchange for the perfused BCH [76]. To mimic the experimental conditions, we have prescribed the brain ISF concentration of BCH as constant input to the model (considering BCH as competing LNAA (CL) with the same kinetics [83]) and consequently calculated the post-stimulus responses in the concentration of test LNAAs. Considering the fact that measuring the global concentrations of BCH in the entire brain ISF compartment is experimentally challenging, we compared the numerical calculations with the in vivo measurements for different values for the global concentrations of BCH which are much lower than the local BCH concentrations near the probes. The computational results for the dynamic changes of L-tyrosine and L-phenylalanine calculated using the kinetic and expression ratios of LAT1 ( and ) determined above are plotted in Fig. 3. 5A and 5B, respectively as percentageRKLAT1 of the baseline. RELAT1 The error bars associated with model simulations are calculated based on sensitivity studies described below. As shown in all panels, the elevation of perfused BCH concentration leads to increased stimulation of the transport of test LNAAs into the brain ISF which is due to the stimulated exchange of the perfused BCH with the test LNAAs via MBEC LAT1 and astrocyte LAT2 (trans-stimulation of efflux from these cells). The model calculations for the stimulated test LNAAs eventually reach a plateau consistent with our previous experimental observations. The best match between model and experimental measurements was observed for global BCH concentrations of 17-30 µM in the brain ISF. It has to be noted that our model, by assuming a homogenous mixture of LNAAs within the individual NVU compartments, disregards the delayed diffusion time of the perfused BCH from 57

the probe site into the ISF which already explains the initial difference between model calculations and experimental measurements in all panels (see Discussion).

A C

B D

Figure 3.5. Trans-stimulation of the test LNAA uptake across the BBB during ISF perfusion with BCH. Figure 5 shows the ISF concentration of the test LNAAs during ISF perfusion with 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid (BCH) started at time zero. In all panels, the experimental data are measured by Dolgodilina et al. [76] for trans- stimulation of test LNAA (L-valine) during 170 minute continues ISF perfusion with 20 mM BCH into a group of freely moving mice (four animals). Panels (A) and (B) show the model calculations for L-tyrosine trans-stimulations upon perfusion of BCH with different global concentration levels. In panels (A) and (B), the bidirectional kinetic constant and the expression ratio of LAT1 are considered, 160, ) and 1, 0.18), respectively. Panels (C) and (D) show the model calculations for L- (RKLAT1 = RELAT1 = 1 (RKLAT1 = phenylalanine trans-stimulations during perfusion of BCH with different global RELAT1 = concentration levels in the entire brain ISF compartment. In panels (A) and (B), the bi- directional kinetic constant and the expression ratio of LAT1 are considered, 80, ) and 1, 0.11), respectively. The model simulations and the (RKLAT1 = experimental data are represented as percent of the baseline value. The error bars associated RELAT1 = 1 (RKLAT1 = RELAT1 = with model calculations indicate standard deviation with respect to concentrations obtained with the nominal model parameter set. For all panels, the calculated baseline concentrations of the test LNAAs are reported in Suppl. Table 3. 2. The differences between the experimental measurements and model calculations with BCH=10 and 100 µM (Panels (A), (B)) as well as BCH=5 and 100 µM (Panels (C), (D)) are statistically significant at all post-stimulus time points (p<0.001, Suppl. Table 3. 4). In contrast, model calculations with BCH=30 µM (Panels (A), (B)) 58

and BCH=17 µM (Panels (C), (D)) are not significantly different from the experimental measurements with the exception of the 20 min post-stimulus time point (Suppl. Table 3. 4).

Sensitivity analysis and statistical testing

We assessed the sensitivity of the reported results with respect to the choice of literature- reported values of model parameters. To accomplish this goal, we simultaneously varied the nominal model input parameters (Michaelis-Menten binding constant, maximum transport rate of AATs (Table 3. 1) and the initialized baseline concentration of LNAAs in individual compartments (Suppl. Table 3. 2) within realistic bounds (± 20% for each parameter), and then assessed the model output for 100 random parameter sets. The results of the sensitivity analysis are presented in Suppl. Tables 3.2 and 3.3, as well as in Figs. 3.2-5 and Suppl. Fig. 3.1, where error bars indicate standard deviation of the computed concentrations from those obtained under nominal parameter conditions. We then assessed whether differences in the set of calculated and experimentally measured concentration profiles are statistically significant. To this end, we performed at each post-stimulus time point Student’s unpaired t- test with Holm-Sidak correction for multiple comparisons using GraphPad Prism 5.0 (GraphPad Software, USA). P values < 0.01 were considered indicative of statistical significance (see Suppl. Table 3. 4 for test results).

Discussion

In this study, using a computational model and experimental input data, we obtained results that strongly suggest a functional polarity of MBECs for the trans-endothelial transport of LNAAs, and characterized a potential strong asymmetry in bi-directional kinetics and/or an asymmetry in membrane expression of MBEC LAT1, which could so far not be addressed with current standard in vitro and in vivo methods. The robust computational model of NVU-LNAA transport we have built and used in this study is based on the fluxes mediated by the respective dominant transporters expressed in MBECs, astrocytes and neurons, namely LAT1, LAT2 and B0AT2. This allowed us to test different symmetric and asymmetric hypotheses about the bi- directional kinetics and/or the expression of LAT1 in MBECs. The comparison of our computational results with published in vivo microdialysis measurements obtained in rat brain supports the hypothesis that MBEC LAT1 either exhibits strong asymmetric bi-directional kinetics for LNAAs (lower affinity inside the MBECs) and/or is asymmetrically expressed at the BBB (lower expression level at the abluminal membrane relative to the luminal membrane of the BBB). This observation is shown to be independent of the substrate considered (i.e. L- tyrosine and L-phenylalanine).

After the characterization of the functional polarity of MBECs, we aimed at understanding the response of the individual NVU cells to IP administration of LNAAs, which has not been addressed so far by in vivo standard methods. To accomplish this, we employed the computational model to calculate the changes in the concentrations of NVU-LNAAs in response to IP administration of L-tyrosine and L-phenylalanine, considering asymmetry in either bi-directional kinetics and/or expression distribution of LAT1 in MBECs. We thereby captured the interactive dynamics of LNAAs as they traverse the blood-brain barrier (BBB) from the capillary lumen into the brain interstitial fluid and from there eventually into astrocytes and/or neurons. Finally, we employed the model to explain also the trans-stimulation of LNAA uptake across the BBB upon ISF perfusion with BCH, a competitive inhibitor of LAT1.

LAT1 is the primary entry way to the brain for a broad range of the essential LNNAs and their analogs, such as L-DOPA, gabapentin and L-melphalan [24, 97, 119]. Hence, our finding of its asymmetric kinetics and/or expression in MBECs provides novel insight that may help advance our understanding of LAT1-mediated prodrug delivery (i.e. meta-substituted phenylalanine prodrugs) to the brain [120-122]. In addition, our computational model could be employed to provide insight into the amino acid transport processes in brain disorders associated with perturbations of LNAAs in the plasma, e.g. phenylketonuria (PKU) or Maple syrup urine disease (MSUD). This could be achieved by using the plasma LNAA perturbations observed in patients as input to the model to calculate the corresponding responses in the NVU-LNAA concentrations, which are challenging to measure experimentally [123].

We note a number of simplifying assumptions made for the development of our computational model. For instance, the assumption of a homogenous mixture of LNAAs within the individual NVU compartments disregards the local differences in the intra-compartmental concentration of LNAAs. In reality, however, regional distribution of amino acids in NVU compartments may affect the binding of LNAAs to the corresponding transporters, and therefore also the local transport fluxes. Moreover, we have considered competitive LNAAs as a single entity rather than accounting one by one each individual competitor for the transport of L-phenylalanine or L-tyrosine, such as L-leucine, L-tryptophan and others (Suppl. Table 3. 1). This assumption, however, has already been experimentally validated for multi-substrate enzymatic reactions [124]. In addition, we focused on a single carrier per NVU compartment membrane, specifically on the antiporters (obligatory exchangers) LAT1 and LAT2 for MBECs and astrocytes, respectively, and a symporter (cotransporter B0AT2) for neurons, not taking into consideration diffusive pathways which have been shown, however, to be of lesser importance for LNAAs in the NVU [96]. Moreover, we did not include LNAA metabolism, which is not completely known and understood in the CNS [125, 126]. However, it has been shown that the brain metabolic 60

fluxes of LNAAs (such as L-phenylalanine, L-histidine, etc) are small compared to the carrier- mediated fluxes [127]. Additionally, our model relies on literature-reported parameter values, which are inevitably associated with the reported uncertainty. Nevertheless, our sensitivity analysis has shown that the conclusions drawn in this study hold within reasonable parameter variations. Furthermore, it has to be pointed out that the established computational model takes only into account the interactions between the dominant NVU-LNAA transporters mentioned above and disregards the contribution of other transporters, such as for instance y+LAT2 and ASCT2, which have been shown to be expressed in adult brains, though at a lower level [87- 89]. Beyond that, it has to be highlighted that the structure and function of many (SLC) transporters have yet to be fully characterized and that some of them may also transport LNAAs, such that further research is required [119]. The contribution of newly discovered NVU transporters could then be included in the computational model upon sufficient characterization of their kinetics.

Summary

We have characterized a functional polarity for MBECs which are the key NVU element for the control of LNAA homeostasis in the brain ISF. For this purpose, we have developed a robust computational model of NVU-LNAA homeostasis and combined it with published in vivo measurements obtained in rat brain. We have shown that either strong asymmetrical bi- directional kinetics of LAT1 in MBECs and/or an asymmetric distribution of LAT1 at both membranes of MBECs is required to reproduce available in vivo measurements. This conclusion is strengthened by the fact that it is supported by data obtained for two tested LNAAs, L-tyrosine and L-phenylalanine. Important characteristics of LAT1 function in MBECs have not been tested satisfactorily up to now by experimental means. In addition, based on our findings on the functional polarity of MBECs, we employed our computational model to investigate the dynamic behavior of LNAAs in astrocytes and neurons in response to IP- administered L-tyrosine and L-phenylalanine, values which are challenging to determine experimentally. Finally, we used the model to explain the trans-stimulation of LNAA uptake across the BBB upon ISF perfusion with a LAT1 competitive inhibitor. While we employed our computational platform to answer fundamental physiological questions about homeostatic regulation of LNAAs in the NVU, it could also be used to test strategies designed to improve the treatment and management of LNAA-related brain disorders.

Conflict of interest

The authors declare that they have no conflict of interest.

Author Contributions

M.T. implemented the computational model and performed the calculations with S.B.. F.V. and V.K. contributed equally to this article and directed the research. All authors conceived and designed the study, analyzed the data, wrote the manuscript and approved the final version.

Funding

This work was supported by SystemsX.ch - The Swiss Initiative in Systems Biology (MT), by the Swiss National Science Foundation through NCCR Kidney.CH (VK) and grant 310030_166430 (FV), and by the Zurich University Hospital (SB).

Acknowledgments

We are grateful to Dr. Victoria Makrides and Dr. Virginia Meskenaite for the fruitful discussions in preparation of the computational model.

Supplementary information Supplementary Table 3.1: The kinetic parameters of the dominant transporters and concentration of individual LNAAs in the different brain compartments

Kinetic parameters Concentration

LAT1 LAT2 B0AT2 LAT1 LAT2 B0AT2e Astrocyte and Plasma ISF (MBEC) (Astrocyte) (Neuron) (MBEC) (Astrocyte) (Neuron) neuronf

Km,abs (µM) V max (µmol/min) Concentration (µM) 11 1 10.2a 1050 0.075 0.1128 0.0193 77 0.73 52.1 L-phenylalanine [25] [28] [32] [25, 104] [59, 105] [32, 106] [84] [84] [85] 210 1 10.2a 190 0.089 0.1427 0.0193 211 1.22 86.5 L-valine [25] [28] [30, 31] [25, 104] [59, 105] [32, 106] [84] [84] [85] 29 224.2 81 0.107 0.184 0.0193 149 1.35 66.7 L-leucine [25] [28] [32] [25, 104] [59, 105] [32, 106] [84] [84] [85] 56 1 39.2a 58 0.109 0.131 0.0154 102 0.77 26.1 L-isoleucine [25] [28] [32] [25, 104] [59, 105] [32, 106] [84] [84] [85] 100 3 27a 0.111 0.173 52 1.12 68.8 L-histidine NA NA [25] [28] [25, 104] [59, 105] [84] [84] [85] 64 2 94a 0.175 0.113 112 0.76 76.1 L-tyrosine NA NA [25] [28] [25, 104] [59, 105] [92] [84] [85] 15 1 10.2a 0.099 0.131 17 0.15 13.5 L-tryptophan NA NA [25] [28] [25, 104] [59, 105] [84] [84] [99] 40 4 30.7a 40 0.046 0.156 0.0193 96 0.37 38.6 L-methionine [25] [28] [32] [25, 104] [59, 105] [32, 106] [84] [84] [85] CLbc 37 163.6 123.5 0.0859 0.1452 0.0193 704 5.7 352.2

CLbd 59.2 185.9 126.2 0.0973 0.1494 0.0193 739 5.7 376.2

aThe values are calculated based on Michaelis-Menten equation [28, 128] .bThe kinetic parameters for the mixture of competing LNAAs are calculated based on Eq.10. c In this row, CL represent the mixture of L-tyrosine competing LNAAs (mixture of L-leucine, L-isoleucine, L-phenylalanine, L-tryptophan, L-valine, L-histidine and L-methionine). dIn this row, CL represent the kinetic parameters for the mixture of phenylalanine competing LNAAs (mixture of L-leucine, L-isoleucine, L-tyrosine, L-tryptophan, L-valine, L-histidine and L-methionine). eThe volume of astrocyte is considered 742 µl [113, 114] (Table 1). eThe reported Vmax values for individual LNAAs are based on the Leucine measurements. f The concentration of LNAAs in the neuron and astrocyte compartments are based on measurements in the brain tissue [105]. NA (not applicable) specifies that the large neutral amino acid was not reported

to be a substrate for the transporter. For calculation of Vmax values, the total rat brain weight, volume and protein content and the volume of astrocyte and neuron are considered 1.81 g [117], 1737 µl, 105 mg protein/g brain [118], 742 µl [113, 114] and 441.7 µl [114-116] (Table 1), respectively.

Supplementary Table 3. 2: The initialization and model calculations for the baseline concentration of LNAAs in the individual NVU compartments.

-L tyrosinea -L phenylalanineb Initialized Initialized baseline Calculated baseline Parameter baseline Calculated baseline concentration Unit concentrati concentration concentrationc onc Microvascular brain endothelial celld

[0-76.1] 274.0 ± 20.9 16.6±1.1 [0-52.1] 146.8 ± 24.6 106.5 ± 13.6 µM MBEC [85] [85] b [TL] [0-352.2] 1722.2 ± 131.5 104.5±6.6 [0-376.2] 1408.8 ± 235.7 1022.4 ±130.7 µM MBEC [85, 99] [85, 99] b [BrainCL] interstitial fluide 0.8 1.0±0.1 1.2±0.1 0.7 0.4 ± 0.03 0.4 ± 0.03 µM ISF [84] [84] b [TL] 5.7 6.5±0.9 7.3±0.9 5.7 3.8 ± 0.3 3.8 ± 0.3 µM ISF [84] [84] b [AstrocyteCL] f 76.1 61.5 ± 7.0 66.1±6.7 52.1 40.6 ± 4.8 40.2 ± 4.6 µM Ast [85] [85] b [TL] 352.2 386.5± 43.7 415.5±41.8 376.2 389.7 ± 46.0 385.7 ± 44 µM Ast [85] [85, 99] b [NeuronCL] f 76.1 75.1 ± 9.0 77.1±9.3 52.1 53.4 ± 7.0 54.3 ± 6.0 µM Neu [85] [85] b [TL] 352.2 316.6 ± 41.4 354.8±41.3 376.2 368.7 ± 46.9 375.2 ± 39.8 µM Neu [85, 99] [85, 99] b [TheCL] bi-directional and expression constant of MBEC LAT1

- 160 1 - 80 1 -

RKLAT1 - 1 0.18 - 1 0.11 - RELAT1

Evaluation of brain ISF post-stimuli responses under the assumption of LAT1 as the dominant LNAA transporter in astrocytes

As mentioned in the Introduction section, several studies have shown that LAT2 is the dominant LNAA transporter in primary astrocyte cells [27-29]. However, Zhang Y, et al. [86] showed that freshly isolated astrocytes specifically express higher levels of LAT1 mRNA compared LAT2 mRNA [86]. Even though the mRNA expression does not necessarily correspond to protein abundance and dominance [83], and although there is no report on the astrocytic transport activity of LAT1 in non-cultured cell assays, we have additionally checked whether considering LAT1 instead of LAT2 as dominant astrocytic AAT would modify our conclusion on the functional polarity of MBECs with either asymmetric bi-directional kinetics

and/or asymmetric distribution of LAT1 in MBECs. To this end, we assumed LAT1 to be the dominant astrocytic LNAA transporter, replacing LAT2 in the astrocyte. In this situation, given that the intra-compartmental fluxes depend on the choice of the dominant transporters, we simply substituted the parameters of Eq. (7) (Methods section) with those of LAT1 for astrocytes (reported in Suppl. Table 3. 3) and similarly calculated the baseline (pre-stimulus) state of the NVU system as reported in Suppl. Table 3. 3. We then calculated the post-stimulus response of LNAA concentrations (Suppl. Fig. 3.1) upon perturbation of plasma L-tyrosine and L-phenylalanine concentrations (Figs. 3.2A and 3.2C). All model parameters except for kinetic parameters specific to LAT1 in the astrocyte (Suppl. Table 3. 3) remain the same as in the nominal model (Table 3. 1). The bi-directional kinetic constant for astrocyte LAT1 is considered equal to the corresponding value in the MBEC LAT1 ( . The time evolution of the plasma concentration of L-tyrosine and L-phenylalanine and RK ofLAT1 competing) LNAAs is plotted as percentage of the baseline values in Suppl. Fig. 3. 1. The error bars given for the baseline concentrations were calculated based on sensitivity studies as described in the Sensitivity analysis section. We compared the model predictions for the ISF response to IP L-tyrosine and L-phenylalanine injection with results of in vivo measurements [92, 93], in the range from 1 (symmetric case) and 1300 (highly asymmetric bi-directional kinetics[26]) (Suppl. Fig. 3.1A,C). We found a close agreement between our model calculations and experimental measurements assuming asymmetric MBEC LAT1 kinetics (best with 220 and 45, respectively, for

L-tyrosine and L-phenylalanine IP injection cases), while RKLAT the1 =model failed to reproduce the experimental data when the bi-directional kinetics of LAT1 were assumed to be symmetric in MBECs (Suppl. Fig. 3. 1A, C).

To evaluate the hypothesis of asymmetric distribution of LAT1 at the BBB, we varied the abluminal to luminal expression ratio of LAT1, , between 0.01 to 10 (representing highly asymmetric abluminal to luminal expression ratio) RELAT while1 we assumed the bi-directional kinetics of LAT1 to be symmetric ( ). The numerical results obtained with asymmetric transporter expression agreed RK LAT1 well = with 1 in vivo experimental data, best for an expression kinetic constant of =0.12. Taken together, our results show that assuming LAT1 rather than LAT2 as the dominantRELAT1 astrocytic AAT does not affect our conclusion on functional polarity of MBECs with either strong asymmetric kinetics of LAT1 and/or its expression at the BBB. Further exploration of the correlation between astrocyte mRNA is required to characterize the function of LAT1 in astrocytes in vivo.

Supplementary Table 3. 3: The kinetic parameters when assuming LAT1 as the dominant LNAA transporter in astrocytes, and the calculated baseline concentration of individual LNAAs in the different brain compartments L-tyrosinea L-phenylalanineb Parameters Value Unit LAT2 (Astrocyte) 64 11 µM ISF(Ast) [25] [25] m,abs,LAT1,TL K 0.132 0.184 µmol/min [105] [105] Vmax,LAT1,TL 37 52.9 µM ISF(Ast) [25] [25] m,abs,LAT1,CL K 0.184c 0.178c µmol/min [105] [105] Vmax,LAT1,CL Calculated baseline

concentrationd 281.4 ± 23.9 16.8±1.1 142.9 ± 24.9 99.7 ±14.7 µM MBEC [TL]b 1770.9 ±150.3 105.8±6.8 1371.7 ± 238.9 957 ±140.7 µM MBEC [CL]b 1.0 ± 0.1 1.1±0.1 0.4 ± 0.03 0.4 ±0.03 µM ISF [TL]b 6.4 ± 0.8 7.1±0.8 3.8 ±0.3 3.9 ±0.3 µM ISF [CL]b 63.0± 7.5 62.4±7 39.6 ± 4.7 38.4 ±4.4 µM Ast [TL]b 395.7 ± 42.3 392.4±44.1 379.5 ±45.3 368.8 ±42.7 µM Ast [CL]b 74.6 ± 8.8 75.4±8.7 54.4 ± 6.6 57.3 ±6.4 µM Neu [TL]b 308.8 ± 40.8 342.9±39.2 376.1 ± 44.0 395 ±43 µM Neu The[CL ]bi-directionalb and expression constant of MBEC LAT1 220 1 45 1 -

RKLAT1 1 0.12 1 0.12 - a b REInLAT1 this column, TL and CL represent L-tyrosine and L-tyrosine competing LNAAs, respectively. In this column, TL and CL represent L- phenylalanine and L-phenylalanine competing LNAAs, respectively. cThe kinetic parameters for the mixture of L-tyrosine and L-phenylalanine competing LNAAs are calculated based on Eq.10 (Suppl. Table 3. 1) and the reported Vmax values for individual LNAAs are based on the Leucine measurements dThe initialization of baseline concentrations is described in Suppl. Table 3. 2.

Statistical analysis

Supplementary Table 3.4: P-values of Student’s unpaired t-test for differences between calculated and measured concentrations Model parameters Post-stimulus time (min)

30 60 90 120 150 180 210 240 [-] [-] LAT1 LAT1 RK1 RE1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 2.38.10-8 10 1 3.79.10-7 2.45.10-4 7.95.10-3 0.056 0.057 0.057 0.057 0.011 Fig. 2B 160 1 0.387 0.035 0.906 0.906 0.906 0.387 0.387 0.387 1300 1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.01 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.18 <10-15 0.235 0.847 0.329 0.149 0.629 0.847 0.149 Fig. 2C 1 1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 2.38.10-8 1 10 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 5.01.10-10 1.90.10-5 1 1 <10-15 <10-15 <10-15 <10-15 1.36.10-5 0.972 <10-15 <10-15 10 1 1.77.10-8 9.60.10-12 1.56.10-11 6.70.10-9 0.013 0.807 8.45.10-5 0.121 Fig. 2E 80 1 8.96.10-8 0.990 0.749 0.990 0.047 0.990 0.018 0.826 1300 1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.01 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.11 0.001 0.720 0.720 0.720 0.435 0.720 0.231 0.720 Fig. 2F 1 1 <10-15 <10-15 <10-15 <10-15 1.36.10-5 0.972 <10-15 <10-15 1 10 <10-15 <10-15 <10-15 5.15.10-12 0.024 <10-15 <10-15 <10-15

System input Post-stimulus time (min) [BCH] [µM] 20 50 80 110 170 10 <10-15 <10-15 <10-15 <10-15 <10-15 Fig. 5A 30 2.94.10-12 0.121 0.662 0.023 0.646 100 <10-15 <10-15 <10-15 <10-15 <10-15 10 <10-15 <10-15 <10-15 <10-15 <10-15 Fig.5.B 30 1.50.10-8 0.971 0.045 0.971 0.045 100 <10-15 <10-15 <10-15 <10-15 <10-15 5 <10-15 <10-15 <10-15 <10-15 <10-15 Fig.5.C 17 <10-15 0.135 0.470 0.151 0.492 10 <10-15 <10-15 <10-15 <10-15 <10-15 5 <10-15 <10-15 <10-15 <10-15 <10-15 Fig.5.D 17 <10-15 0.549 0.071 0.571 0.062 100 <10-15 <10-15 <10-15 <10-15 <10-15

Model parameters Post-stimulus time (min)

[-] [-] 30 60 90 120 150 180 210 240 LAT1 LAT1 RK1 RE1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 Suppl. 10 1 3.32.10-7 5.64.10-5 6.63.10-3 0.056 0.057 0.057 0.057 0.011 Fig.1.A 220 1 0.139 2.34.10-4 0.029 0.472 0.737 0.472 0.394 0.246 1300 1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.01 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.12 3.42.10-12 0.623 0.627 0.623 0.011 0.627 0.503 0.503 Suppl. -15 -15 -15 -15 -15 -15 -15 -15 Fig.1.B 1 1 <10 <10 <10 <10 <10 <10 <10 <10 9.66.10- 1 10 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 3.31.10-11 6 6.17.10- <10-15 <10-15 <10-15 <10-15 1 1 8.83.10-7 0.012 0.262 4 Suppl. 1.58.10- 10 1 . -11 . -8 . -8 . -5 . -4 . -4 4 Fig.1.C 7.21 10 2.40 10 1.71 10 1.03 10 1.62 10 2.90 10 0.018 45 1 0.032 4.35.10-9 0.223 0.032 0.045 0.013 0.535 0.017 1300 1 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 1 0.01 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 <10-15 0.088 Suppl. 1 0.12 0.025 0.011 0.758 0.719 0.758 0.204 0.758 6.17.10- Fig.1.D <10-15 <10-15 <10-15 <10-15 1 1 8.83.10-7 0.012 0.262 4 1 10 <10-15 <10-15 <10-15 7.36.10-11 0.023 0.030 9.05.10-11 <10-15

Supplementary Figure 3.1. The brain ISF concentration response after intraperitoneal injection of L-tyrosine and L-phenylalanine under the assumption of LAT1 as the dominant LNAA transporter in astrocytes. Panels (A) and (B) show the experimental data for L-tyrosine (Tyr) concentration in the brain ISF (prefrontal cortex (PFC)) [92] in response to IP administration of 200 mg/kg L-tyrosine (as plotted in Fig. 3.2B) compared to the model calculations assuming LAT1 as the dominant transporter in astrocytes (model input as shown in Fig. 3.2A) for various ratios of the bi-directional kinetic constant of LAT1 in MBEC and astrocytes ( ) as well as different abluminal to luminal expression ratio of LAT1 ( ). In Panels (A) and (B), the ISF baseline values for L-tyrosine are 1.0 and 1.1 µM (Suppl. Table RKLAT1 RELAT1 3. 3), respectively. Each experimental data point represents the mean ± SD for three (plasma) and four to eight (ISF) animals [92]. The error bar associated with model calculations indicate the standard deviation determined based on the sensitivity analysis results. Panel (C) and (D) show the experimental data for L-phenylalanine (Phe) concentration in the brain ISF (prefrontal cortex (PFC)) [93] in response to IP administration of 200 mg/kg L- phenylalanine (as plotted in Fig. 3.2D), versus the model calculations (model input is shown in Fig. 3.2D) for various ratios of the bi-directional kinetic constant of LAT1 in MBEC and astrocytes ( ) as well as different abluminal to luminal expression ratios of LAT1 ( ). In panels (C) and (D), the RKLAT1 ISF baseline value for L-phenylalanine is 0.4 µM as reported in Suppl. Table 3. 3. In Panels RELAT1 (A), (B), (C) and (D), the differences between the output of the symmetric model ( and ) and experimental measurements are statistically significant at all post- RKLAT1 = 1 stimulus time points (p<0.001, Suppl. Table 3. 4) with the exception of 180 and 210 min in RELAT1 = 1 Panels (C) and (D). In contrast, there is no significant difference between the experimental measurements and the model calculations with 220 and (Panel A), and (Panels (B) and (D)), and and (Panel RKLAT1 = RELAT1 = 1 C), with the exception of the 30 min post-stimulus time point in Panel (B) and the 60 min point RKLAT1 = 1 RELAT1 = 0.12 RKLAT1 = 45 RELAT1 = 1 in Panels (A) and (C) (Suppl. Table 3. 4).

Chapter 5 Conclusion and outlook

Conclusion The focus of this doctoral thesis is to establish a combined computational and experimental framework of AAT interactions to investigate the regulation of amino acids (AAs) across cell membranes. The first part of the thesis (chapter 2) introduced an integrated in vitro and in silico approach that enables unraveling the contributions of specific players in a complex cellular network of AAs and AATs at single membranes. The second part (chapter 3) described a combined in vivo and in silico approach to investigate the dynamic interactions between large neutral amino acid (LNAA) transporters located at the membranes of neurovascular unit (NVU) cells. The third part (chapter 4), presented an application for the in silico model of large neutral amino acid (LNAA) homeostasis in the NVU to gain some insights into the regulatory role of brain LNAA transporters involved in Phenylketonuria (PKU) disorder.

In chapter 2, we provided a robust and versatile tool based on a systems biology approach to characterize the contributions of specific AAT to overall cellular transport, using limited experimental input. Using this approach, we could accurately predict the cellular transport responses to new stimuli in presence and absence of extracellular sodium ions and different competitive inhibitors. We applied this strategy to quantify L- leucine and L-phenylalanine individual AATs expressed in X. leavis oocyte in vitro. However, given the appropriate assay conditions, this methodology is also applicable to other enzymes, substrates, and/or cellular systems where the kinetics of each active enzyme species in the system can be described by the Michaelis–Menten equation. While the main strength of our approach is its capability to quantify the activity of individual transporters that function within a complex network of transporters, it is limited to the investigation of the trans-membrane transport processes through single membranes and under steady state conditions where the MM assumptions are valid.

In chapter 3, we further extended our mathematical model beyond the Michaelis– Menten kinetics to represent dynamics of AATs interactions through multiple cell membranes. We have targeted the NVU-LNAA transport system to get more insight on mechanisms that maintain the homeostasis of LNAAs in the NVU individual cells, such as microvascular brain endothelial cells (MBECs), astrocytes and neurons. By

incorporation of published in vivo microdialysis measurements obtained in rat brain into our newly developed in silico model, we have shown that either strong asymmetrical bi-directional kinetics of LAT1 in MBECs (lower affinity inside the MBECs) and/or an asymmetric distribution of LAT1 at both membranes of MBECs (lower expression at the abluminal membrane of the BBB) is required to reproduce published in vivo measurements. Our conclusion is supported by data obtained for two different LNAAs, L-tyrosine and L-phenylalanine. This vital characteristic of LAT1 function in MBECs could not be addressed satisfactorily up to now by available in vivo and in vitro standard assays. LAT1 plays a crucial role in regulating the import and export of a large number of substrates including some prodrugs such as L-DOPA in and out of the brain. Therefore, our observations for its asymmetric kinetics and/or expression could shed lights on LAT1 mediated prodrug delivery across the BBB [120-122]. On the basis of our findings about the function of LAT1 across the BBB, we furthure investigated the dynamics of LNAAs concentrations in MBECs, brain ISF, astrocytes and neurons in response to IP-administered L-tyrosine and L-phenylalanine, values which are difficult and challenging to determine experimentally. Finally, we used our model to explain the trans-stimulation of LNAA uptake across the blood brain barrier (BBB) upon interstitial application of BCH (2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid), a LAT1 competitive inhibitor. Taken together, our developed in silico model of NVU-LNAA homeostasis is shown to circumvent some limitations associated with the available in vivo and in vitro experimental approaches. Our computational model, however, is not limited to study the cases we reported and it could be further served to mimic more experimental situations. For instance, it can be used to capture the impact of inhibition of LAT1 at the luminal and abluminal membranes of the BBB and/or LAT2 in astrocytes and/or B0AT2 in neurons, through supplementation of different physiological and non- physiological substrates and thereby to provide a quantitative representation for the corresponding dynamic changes in concentrations of LNAAs in cells of NVU.

In chapter 4, using our NVU-LNAA homeostasis model, we further extended our study to unravel the regulatory role of individual LNAA-NVU transporters in the propagation of abnormal perturbations of Phe from plasma into the NVU individual cells in Phenylketonuria (PKU) disease. We suggested that plasma Phe fluctuations can potentially propagate into the NVU and change there the concentration of LNAAs, with the highest magnitude of this effect at low frequency and high amplitude-to-mean ratio 90

of the plasma Phe concentration fluctuations. Failure of AA homeostasis in general harmfully affects cell function [150]. In the NVU individual cells, Phe and CL are involved in a variety of processes, including the synthesis of essential neurotransmitters such as serotonin (5 hydroxytryptamine; 5 HT) and dopamine (3- hydroxytyramine) [151-154], and the production‐ of proteins ‐ and lipids (i.e., myelin) [125, 126], although their function in the latter processes is not completely characterized. Therefore, depending on the full set of function of Phe and CL, fluctuations of their concentrations in the NVU could have a negative impact on the brain function. Finally, we explained the therapeutic impact of LNAAs supplementation in attenuating the disturbed concentrations of Phe and CL in the MBECs, ISF, astrocytes and neurons in PKU disorder towards normal physiologic levels, something out of reach of in vivo techniques. It has to be pointed out while our in silico model has been employed in PKU disease, it could potentially be applied to provide new insights into the pathophysiology of other metabolic disorders affecting the brain, such as Maple syrup urine disease (MSUD), that are sensitive to the plasma levels of LNAAs [123].

In conclusion, given the inherent functional complexities associated with the behaviour of AAT systems, there are technological hurdles that need to be overcome to enable corresponding experiments. As an alternative approach, this thesis demonstrates the capabilities of integration of in silico with in vitro and in vivo methods to go around these hurdles and thereby to provide novel insight into the physiological aspects of AAT systems.

Outlook The computational models we have developed in this thesis as any other model have their simplifications and limitations that are clearly mentioned in each chapter. In particular, the computational models rely on literature-reported parameter values, which are inevitably associated with the reported uncertainty. We handled this limitation through performing sensitivity analysis for each model and demonstrated that the conclusions drawn in this thesis hold within the reported reasonable parameter variations. Also, as it is a common characteristic of biological models, we have built our in silico frameworks based on a simplified anatomy of understudied transport

systems. For instance, we have modeled the NVU systems as four compartments (i.e. MBECs, brain ISF, astrocyte and neuron) in which the distribution of LNAAs is considered to be homogenous. In reality, however, regional distribution of AAs may affect their binding efficiency to the corresponding AATs, and therefore also the local transport fluxes. On the other hand, there is hope that upon the progress in imaging and biochemistry techniques more information about the local distribution of AAs as well as the cell morphology will be provided in the near future. This evolution might enhance our knowledge about different AAT systems and would allow for more detailed investigation of these systems in more precise biological morphologies. Furthermore, we focused on a single dominant LNAA transporter per membrane of NVU cells (i.e. LAT1 for MBECs, LAT2 for astrocytes and B0AT2 for neurons), and thus disregarded the pathways related to diffusion and also NVU transporters with low levels of expression, whose impacts have been shown, however, to be insignificant for LNAAs in the NVU [96, 127]. Moreover, it has to be pointed out that our computational models are developed based on the available literature about the identified AATs and the structure and kinetics of many SLC transporters, including AATs, have yet to be fully characterized, such that further research is required. The contribution of newly discovered AATs, upon appropriate investigation of their kinetic functions, could then be included in our developed computational models.

It has to be emphasized that the approaches developed in this thesis are not limited to the study AAT systems, but they theoretically could be applied to a variety of SLC transport systems functioning in different cell types. For example, our systems biology approach described in chapter 2) could be employed to quantify the relative contribution of a specific SLC peptide (or drug) transporter within a complex network of peptides (or drugs) and SLC transporters functioning at single membranes. Our integrated in silico and in vivo approach described in chapter 3, for example, could potentially be applied to get more insights into the regulatory roles of SLC glucose (or drug) transporters in the maintenance of glucose (or drug) homeostasis in the NVU. Moreover, our computational approaches can potentially be used to study SLC transport processes across various biological systems such as kidney, intestine, etc. Similar to the situation we faced with NVU-LNAA transport systems, this would significantly depend on the availability of sufficient in vivo experimental data including

results of dynamic time course measurements of concentration changes of the considered substrates, such as AAs, peptide, etc as well as on accurate information about the kinetics of the expressed SLC transporters. Taken together, our studies provide several examples of integrating computational and experimental approaches for studying SLC transport complex systems and demonstrate how this integration could be an excellent asset for basic and applied research.

References

1. Broer, S., Amino acid transport across mammalian intestinal and renal epithelia. Physiol Rev, 2008. 88(1): p. 249-86. 2. Makrides, V., S.M. Camargo, and F. Verrey, Transport of amino acids in the kidney. Compr Physiol, 2014. 4(1): p. 367-403. 3. Hediger, M.A., et al., The ABCs of solute carriers: physiological, pathological and therapeutic implications of human membrane transport proteinsIntroduction. Pflugers Arch, 2004. 447(5): p. 465-8. 4. Hediger, M.A. BioParadigms. 2004; ]. Available from: http://www.bioparadigms.org/. 5. Verrey, F., et al., Kidney amino acid transport. Pflugers Arch, 2009. 458(1): p. 53-60. 6. Verrey, F., et al., Novel renal amino acid transporters. Annu Rev Physiol, 2005. 67: p. 557-72. 7. Silbernagl, S., The renal handling of amino acids and oligopeptides. Physiol Rev, 1988. 68(3): p. 911-1007. 8. Sigrist-Nelson, K., H. Murer, and U. Hopfer, Active alanine transport in isolated brush border membranes. J Biol Chem, 1975. 250(14): p. 5674-80. 9. Evers, J., H. Murer, and R. Kinne, Phenylalanine uptake in isolated renal brush border vesicles. Biochim Biophys Acta, 1976. 426(4): p. 598-615. 10. Samarzija, I. and E. Fromter, Electrophysiological analysis of rat renal sugar and amino acid transport. III. Neutral amino acids. Pflugers Arch, 1982. 393(3): p. 119-209. 11. Fromter, E., Electrophysiological analysis of rat renal sugar and amino acid transport. I. Basic phenomena. Pflugers Arch, 1982. 393(2): p. 179-89. 12. Souba, W.W. and A.J. Pacitti, How amino acids get into cells: mechanisms, models, menus, and mediators. Journal of Parenteral and Enteral Nutrition, 1992. 16(6): p. 569-578. 13. Steinfeld, J.I., J.S. Francisco, and W.L. Hase, Chemical kinetics and dynamics. 1989, Englewood Cliffs, N.J.: Prentice Hall. xii, 548 p. 14. Gillespie, D.T., The chemical Langevin equation. Journal of Chemical Physics, 2000. 113(1): p. 297-306. 15. Michaelis, L. and M.L. Menten, Die kinetik der invertinwirkung. Biochem. z, 1913. 49(333-369): p. 352. 16. Chou, I.-C. and E.O. Voit, Recent developments in parameter estimation and structure identification of biochemical and genomic systems. Mathematical biosciences, 2009. 219(2): p. 57-83. 17. Segel, L.A., On the validity of the steady state assumption of enzyme kinetics. Bulletin of mathematical biology, 1988. 50(6): p. 579-593. 18. Briggs, G.E. and J.B.S. Haldane, A note on the kinetics of enzyme action. Biochemical journal, 1925. 19(2): p. 338. 19. Michaelis, L. and M.L. Menten, Die kinetik der invertinwirkung. 2007: Universitätsbibliothek Johann Christian Senckenberg. 20. Berg, J.M., J.L. Tymoczko, and L. Stryer, Biochemistry. 2002, New York: WH Freeman.

21. Panitchob, N., et al., Computational modelling of amino acid exchange and facilitated transport in placental membrane vesicles. Journal of theoretical biology, 2015. 365: p. 352-364. 22. Panitchob, N., et al., Computational modelling of placental amino acid transfer as an integrated system. Biochimica et Biophysica Acta (BBA)- Biomembranes, 2016. 1858(7): p. 1451-1461. 23. Bisi, M., F. Conforto, and L. Desvillettes, Quasi-steady-state approximation for reaction-diffusion equations. BULLETIN-INSTITUTE OF MATHEMATICS ACADEMIA SINICA, 2007. 2(4): p. 823. 24. Killian, D.M. and P.J. Chikhale, Predominant functional activity of the large, neutral amino acid transporter (LAT1) isoform at the cerebrovasculature. Neuroscience letters, 2001. 306(1): p. 1-4. 25. Smith, Q.R., et al., Kinetics of neutral amino acid transport across the blood brain barrier. Journal of neurochemistry, 1987. 49(5): p. 1651-1658. 26. Meier, C., et al., Activation of system L heterodimeric amino acid exchangers‐ by intracellular substrates. EMBO J, 2002. 21(4): p. 580-9. 27. Yudkoff, M., et al., Astrocyte leucine metabolism: Significance of branched chain amino acid transamination. Journal of neurochemistry, 1996. 66(1): p. 378-385. ‐ 28. Kim, D.K., et al., System L-amino acid transporters are differently expressed in rat astrocyte and C6 glioma cells. Neuroscience Research, 2004. 50(4): p. 437-446. 29. Braun, D., et al., Developmental and cell type specific expression of thyroid hormone transporters in the mouse brain and in primary brain cells. Glia, 2011. 59(3): p. 463-471. ‐ 30. Bak, L.K., et al., Valine but not leucine or isoleucine supports neurotransmitter glutamate synthesis during synaptic activity in cultured cerebellar neurons. Journal of neuroscience research, 2012. 90(9): p. 1768-1775. 31. Yudkoff, M., et al., Neuronal metabolism of branched chain amino acids: Flux through the aminotransferase pathway in synaptosomes. Journal of neurochemistry, 1996. 66(5): p. 2136-2145. ‐ 32. Bröer, A., et al., The orphan transporter v7-3 (slc6a15) is a Na+-dependent neutral amino acid transporter (B0AT2). Biochemical Journal, 2006. 393(1): p. 421-430. 33. Verrey, F., System L: heteromeric exchangers of large, neutral amino acids involved in directional transport. Pflügers Archiv, 2003. 445(5): p. 529-533. 34. Lyck, R., et al., Culture-induced changes in blood–brain barrier transcriptome: implications for amino-acid transporters in vivo. Journal of Cerebral Blood Flow & Metabolism, 2009. 29(9): p. 1491-1502. 35. Duelli, R., et al., Expression of large amino acid transporter LAT1 in rat brain endothelium. Journal of Cerebral Blood Flow & Metabolism, 2000. 20(11): p. 1557-1562. 36. Blau, N., F.J. van Spronsen, and H.L. Levy, Phenylketonuria. The Lancet, 2010. 376(9750): p. 1417-1427. 37. Madden, M., Phenylketonuria: Defects in amino acid metabolism. Mol Med (SCJMM), 2004. 5: p. 57-61. 38. Cleary, M., et al., Fluctuations in phenylalanine concentrations in phenylketonuria: a review of possible relationships with outcomes. Molecular genetics and metabolism, 2013. 110(4): p. 418-423. 95

39. Camargo, S.M., D. Bockenhauer, and R. Kleta, Aminoacidurias: Clinical and molecular aspects. Kidney Int, 2008. 73(8): p. 918-25. 40. Chillaron, J., et al., Obligatory amino acid exchange via systems bo,+-like and y+L-like. A tertiary active transport mechanism for renal reabsorption of cystine and dibasic amino acids. J Biol Chem, 1996. 271(30): p. 17761-70. 41. Verrey, F., et al., CATs and HATs: the SLC7 family of amino acid transporters. Pflugers Arch, 2004. 447(5): p. 532-42. 42. Ramadan, T., et al., Recycling of aromatic amino acids via TAT1 allows efflux of neutral amino acids via LAT2-4F2hc exchanger. Pflugers Arch, 2007. 454(3): p. 507-16. 43. Babu, E., et al., Identification of a novel system L amino acid transporter structurally distinct from heterodimeric amino acid transporters. J Biol Chem, 2003. 278(44): p. 43838-45. 44. Bauch, C. and F. Verrey, Apical heterodimeric cystine and cationic amino acid transporter expressed in MDCK cells. Am J Physiol Renal Physiol, 2002. 283(1): p. F181-9. 45. Bodoy, S., et al., Identification of LAT4, a novel amino acid transporter with system L activity. Journal of Biological Chemistry, 2005. 280(12): p. 12002- 12011. 46. BRÖER, A., B. HAMPRECHT, and S. BRÖER, Discrimination of two amino acid transport activities in 4F2 heavy chain-expressing Xenopus laevis oocytes. Biochemical Journal, 1998. 333(3): p. 549-554. 47. Camargo, S.M., et al., Steady-state kinetic characterization of the mouse B0AT1 sodium-dependent neutral amino acid transporter. Pflügers Archiv, 2005. 451(2): p. 338-348. 48. Danilczyk, U., et al., Essential role for collectrin in renal amino acid transport. Nature, 2006. 444(7122): p. 1088-91. 49. Guetg, A., et al., Essential amino acid transporter Lat4 (Slc43a2) is required for mouse development. J Physiol, 2014. 50. Kanai, Y., et al., Expression cloning and characterization of a transporter for large neutral amino acids activated by the heavy chain of 4F2 antigen (CD98). Journal of Biological Chemistry, 1998. 273(37): p. 23629-23632. 51. Pfeiffer, R., et al., Luminal heterodimeric amino acid transporter defective in cystinuria. Mol Biol Cell, 1999. 10(12): p. 4135-47. 52. Pfeiffer, R., et al., Functional heterodimeric amino acid transporters lacking cysteine residues involved in disulfide bond. FEBS Lett, 1998. 439(1-2): p. 157-62. 53. Ramadan, T., et al., Basolateral aromatic amino acid transporter TAT1 (Slc16a10) functions as an efflux pathway. J Cell Physiol, 2006. 206(3): p. 771- 9. 54. Singer, D., et al., Orphan transporter SLC6A18 is renal neutral amino acid transporter B0AT3. J Biol Chem, 2009. 284(30): p. 19953-60. 55. Herman, P. and J.C. Lee, The advantage of global fitting of data involving complex linked reactions. Allostery: Methods and Protocols, 2012: p. 399-421. 56. Meisl, G., et al., Molecular mechanisms of protein aggregation from global fitting of kinetic models. Nat Protoc, 2016. 11(2): p. 252-72. 57. Bridges, C.C. and R.K. Zalups, System B-0,B-+ and the transport of thiol-S- conjugates of methylmercury. Journal of Pharmacology and Experimental Therapeutics, 2006. 319(2): p. 948-956.

58. Yanagida, O., et al., Human L-type amino acid transporter 1 (LAT1): characterization of function and expression in tumor cell lines. Biochimica Et Biophysica Acta-Biomembranes, 2001. 1514(2): p. 291-302. 59. Segawa, H., et al., Identification and functional characterization of a Na+- independent neutral amino acid transporter with broad substrate selectivity. Journal of Biological Chemistry, 1999. 274(28): p. 19745-19751. 60. Pfeiffer, R., et al., Amino acid transport of y(+)L-type by heterodimers of 4F2hc/CD98 and members of the glycoprotein-associated amino acid transporter family. Embo Journal, 1999. 18(1): p. 49-57. 61. Broer, A., et al., The heterodimeric amino acid transporter 4F2hc/y(+)LAT2 mediates arginine efflux in exchange with glutamine. Biochemical Journal, 2000. 349: p. 787-795. 62. Utsunomiya-Tate, N., H. Endou, and Y. Kanai, Cloning and functional characterization of a system ASC-like Na+-dependent neutral amino acid transporter. J Biol Chem, 1996. 271(25): p. 14883-90. 63. Sloan, J.L. and S. Mager, Cloning and functional expression of a human Na+ and Cl--dependent neutral and cationic amino acid transporter B0+. Journal of Biological Chemistry, 1999. 274(34): p. 23740-23745. 64. Bohmer, C., et al., Characterization of mouse amino acid transporter B(0)AT1 (slc6a19). Biochemical Journal, 2005. 389: p. 745-751. 65. Mastroberardino, L., et al., Amino-acid transport by heterodimers of 4F2hc/CD98 and members of a permease family. Nature, 1998. 395(6699): p. 288-291. 66. Rossier, G., et al., LAT2, a new basolateral 4F2hc/CD98-associated amino acid transporter of kidney and intestine. Journal of Biological Chemistry, 1999. 274(49): p. 34948-34954. 67. Segel, I.H., Enzyme kinetics. Vol. 957. 1975: Wiley, New York. 68. Dumont, J.N., Oogenesis in Xenopus laevis (Daudin). I. Stages of oocyte development in laboratory maintained animals. J Morphol, 1972. 136(2): p. 153-79. 69. Bowes, J.B., et al., Xenbase: gene expression and improved integration. Nucleic Acids Res, 2010. 38(Database issue): p. D607-12. 70. James-Zorn, C., et al., Xenbase: Core features, data acquisition, and data processing. Genesis, 2015. 53(8): p. 486-97. 71. Karpinka, J.B., et al., Xenbase, the Xenopus model organism database; new virtualized system, data types and genomes. Nucleic Acids Res, 2015. 43(Database issue): p. D756-63. 72. Van Winkle, L.J., Endogenous amino acid transport systems and expression of mammalian amino acid transport proteins in Xenopus oocytes. Biochim Biophys Acta, 1993. 1154(2): p. 157-72. 73. Fredriksson, R., et al., The solute carrier (SLC) complement of the human genome: phylogenetic classification reveals four major families. FEBS letters, 2008. 582(27): p. 3811-3816. 74. Schiöth, H.B., et al., Evolutionary origin of amino acid transporter families SLC32, SLC36 and SLC38 and physiological, pathological and therapeutic aspects. Molecular aspects of medicine, 2013. 34(2): p. 571-585. 75. Taylor, P.M., et al., Amino-acid-dependent modulation of amino acid transport in Xenopus laevis oocytes. Journal of experimental biology, 1996. 199(4): p. 923-931.

76. Dolgodilina, E., et al., Brain interstitial fluid glutamine homeostasis is controlled by blood–brain barrier SLC7A5/LAT1 amino acid transporter. Journal of Cerebral Blood Flow & Metabolism, 2015: p. 0271678X15609331. 77. Franco, M., et al., Frog intestinal sac as an in vitro method for the assessment of intestinal permeability in humans: application to carrier transported drugs. International journal of pharmaceutics, 2008. 352(1): p. 182-188. 78. Margheritis, E., et al., Amino acid transporter B0AT1 (slc6a19) and ancillary protein: impact on function. Pflügers Archiv-European Journal of Physiology, 2016: p. 1-12. 79. Schmidt, C., C. Klein, and M. Hollmann, Xenopus laevis oocytes endogenously express all subunits of the ionotropic glutamate receptor family. Journal of molecular biology, 2009. 390(2): p. 182-195. 80. Forster, I.C., et al., Electrogenic kinetics of a mammalian intestinal type IIb Na(+)/P(i) cotransporter. J Membr Biol, 2006. 212(3): p. 177-90. 81. Camargo, S.M., et al., The molecular mechanism of intestinal levodopa absorption and its possible implications for the treatment of Parkinson's disease. J Pharmacol Exp Ther, 2014. 351(1): p. 114-23. 82. Abbott, N.J., L. Rönnbäck, and E. Hansson, Astrocyte-endothelial interactions at the blood-brain barrier. Nature reviews. Neuroscience, 2006. 7(1): p. 41. 83. Taslimifar, M., et al., Quantifying the relative contributions of different solute carriers to aggregate substrate transport. Scientific reports, 2017. 7: p. 40628. 84. Currie, P.J., et al., Microdialysis as a tool to measure dietary and regional effects on the complete profile of extracellular amino acids in the hypothalamus of rats. Life sciences, 1995. 57(21): p. 1911-1923. 85. Kandera, J., G. Levi, and A. Lajtha, Control of cerebral metabolite levels: II. Amino acid uptake and levels in various areas of the rat brain. Archives of biochemistry and biophysics, 1968. 126(1): p. 249-260. 86. Zhang, Y., et al., An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. Journal of Neuroscience, 2014. 34(36): p. 11929-11947. 87. Utsunomiya-Tate, N., H. Endou, and Y. Kanai, Cloning and functional characterization of a system ASC-like Na+-dependent neutral amino acid transporter. Journal of Biological Chemistry, 1996. 271(25): p. 14883-14890. 88. Deitmer, J.W., A. Bröer, and S. Bröer, Glutamine efflux from astrocytes is mediated by multiple pathways. Journal of neurochemistry, 2003. 87(1): p. 127-135. 89. Gliddon, C.M., et al., Cellular distribution of the neutral amino acid transporter subtype ASCT2 in mouse brain. Journal of neurochemistry, 2009. 108(2): p. 372-383. 90. Napolitano, L., et al., LAT1 is the transport competent unit of the LAT1/CD98 heterodimeric amino acid transporter. The international journal of biochemistry & cell biology, 2015. 67: p. 25-33. 91. del Pino, M.M.S., D.R. Peterson, and R.A. Hawkins, Neutral amino acid transport characterization of isolated luminal and abluminal membranes of the blood-brain barrier. Journal of Biological Chemistry, 1995. 270(25): p. 14913- 14918. 92. Bongiovanni, R., et al., Pharmacokinetics of systemically administered tyrosine: a comparison of serum, brain tissue and in vivo microdialysate levels in the rat. Journal of neurochemistry, 2003. 87(2): p. 310-317.

93. Bongiovanni, R., et al., Relationships between large neutral amino acid levels in plasma, cerebrospinal fluid, brain microdialysate and brain tissue in the rat. Brain research, 2010. 1334: p. 45-57. 94. Panitchob, N., Computational modelling of amino acid transfer interactions in the placenta. 2015, University of Southampton. 95. Pradhan, R.K., et al., Carrier-Mediated Transport Through Biomembranes- Chapter 5. 96. Smith, Q.R. and Y. Takasato, Kinetics of amino acid transport at the blood brain barrier studied using an in situ brain perfusion technique. Annals of the New York Academy of Sciences, 1986. 481(1): p. 186-201. ‐ 97. Cundy, K.C., et al., XP13512 [(±)-1-([(α-isobutanoyloxyethoxy) carbonyl] aminomethyl)-1-cyclohexane acetic acid], a novel gabapentin prodrug: II. Improved oral bioavailability, dose proportionality, and colonic absorption compared with gabapentin in rats and monkeys. Journal of Pharmacology and Experimental Therapeutics, 2004. 311(1): p. 324-333. 98. Thorn, M., A method for determining the ratio of the Michaelis constants of an enzyme with respect to two substrates. Nature, 1949. 164(4157): p. 27-29. 99. Amorini, A.M., et al., Severity of experimental traumatic brain injury modulates changes in concentrations of cerebral free amino acids. Journal of cellular and molecular medicine, 2017. 21(3): p. 530-542. 100. Shampine, L.F. and M.W. Reichelt, The matlab ode suite. SIAM journal on scientific computing, 1997. 18(1): p. 1-22. 101. Bogacki, P. and L.F. Shampine, A 3 (2) pair of Runge-Kutta formulas. Applied Mathematics Letters, 1989. 2(4): p. 321-325. 102. Pardridge, W.M., Introduction to the blood-brain barrier: methodology, biology and pathology. 2006: Cambridge University Press. 103. Goldstein, F.B., Biochemical studies on phenylketonuria I. Experimental hyperphenylalanemia in the rat. Journal of Biological Chemistry, 1961. 236(10): p. 2656-2661. 104. Tilgmann, C., et al., Expression of recombinant soluble and membrane bound catechol O methyltransferase in eukaryotic cells and identification of the respective enzymes in rat brain. European Journal of Biochemistry,‐ 1992. 207(2): p. 813-821.‐ 105. Shank, R.P. and G.L. Campbell, Amino acid uptake, content, and metabolism by neuronal and glial enriched cellular fractions from mouse cerebellum. The Journal of neuroscience, 1984. 4(1): p. 58-69. 106. Rao, K.R., M.C. Vemuri, and C.R. Murthy, Synaptosomal transport of branched chain amino acids in young, adult and aged rat brain cortex. Neuroscience letters, 1995. 184(2): p. 137-140. 107. Takanaga, H., et al., Characterization of a branched-chain amino-acid transporter SBAT1 (SLC6A15) that is expressed in human brain. Biochemical and biophysical research communications, 2005. 337(3): p. 892-900. 108. Smith, Q.R., C.E. Johanson, and D.M. Woodbury, Uptake of 36Cl and 22Na by the Brain Cerebrospinal Fluid System: Comparison of the Permeability of the Blood Brain and Blood Cerebrospinal Fluid Barriers. Journal of neurochemistry,‐ 1981. 37(1): p. 117-124. 109. Mori, K., et ‐al., Temporal profile‐ of changes in brain tissue extracellular space and extracellular ion (Na+, K+) concentrations after cerebral ischemia and the

effects of mild cerebral hypothermia. Journal of neurotrauma, 2002. 19(10): p. 1261-1270. 110. Fedoroff, S. and A. Vernadakis, Astrocytes: Biochemistry, Physiology, and Pharmacology of Astrocytes, vol. 2. 1986, Academic Press, Orlando. 111. Licinio, J. and M.-L. Wong, Pharmacogenomics: the search for individualized therapies. 2009: John Wiley & Sons. 112. Syková, E., et al., Changes in extracellular space size and geometry in APP23 transgenic mice: a model of Alzheimer's disease. Proceedings of the National Academy of Sciences of the United States of America, 2005. 102(2): p. 479- 484. 113. Anderova, M., et al., Cell Death/Proliferation and Alterations in Glial Morphology Contribute to Changes in Diffusivity in the Rat Hippocampus after Hypoxia—Ischemia. Journal of Cerebral Blood Flow & Metabolism, 2011. 31(3): p. 894-907. 114. Ren, J., et al., Quantitative analysis of neurons and glial cells in the rat somatosensory cortex, with special reference to GABAergic neurons and parvalbumin-containing neurons. Experimental brain research, 1992. 92(1): p. 1-14. 115. Hosseini-Sharifabad, M. and J.R. Nyengaard, Design-based estimation of neuronal number and individual neuronal volume in the rat hippocampus. Journal of neuroscience methods, 2007. 162(1): p. 206-214. 116. Setou, M., T. Hayasaka, and I. Yao, Axonal transport versus dendritic transport. Journal of neurobiology, 2004. 58(2): p. 201-206. 117. Stewart, C., Weights of various parts of the brain in normal and underfed albino rats at different ages. Journal of Comparative Neurology, 1918. 29(5): p. 511- 528. 118. Banay-Schwartz, M., et al., Protein content of various regions of rat brain and adult and aging human brain. Age, 1992. 15(2): p. 51-54. 119. Rautio, J., M. Gynther, and K. Laine, LAT1-mediated prodrug uptake: a way to breach the blood–brain barrier? Therapeutic delivery, 2013. 4(3): p. 281-284. 120. Gynther, M., et al., Large neutral amino acid transporter enables brain drug delivery via prodrugs. Journal of medicinal chemistry, 2008. 51(4): p. 932-936. 121. Peura, L., et al., Large amino acid transporter 1 (LAT1) prodrugs of valproic acid: new prodrug design ideas for central nervous system delivery. Molecular pharmaceutics, 2011. 8(5): p. 1857-1866. 122. Rautio, J., et al., Prodrug approaches for CNS delivery. The AAPS journal, 2008. 10(1): p. 92-102. 123. Dixon, M., et al., Disorders of amino acid metabolism, organic acidaemias and urea cycle disorders. Clinical Paediatric Dietetics, 2015: p. 381-525. 124. Alberty, R.A., Enzyme kinetics. Adv Enzymol Relat Areas Mol Biol, 1956. 17: p. 1-64. 125. Sperringer, J.E., A. Addington, and S.M. Hutson, Branched-Chain Amino Acids and Brain Metabolism. Neurochemical Research, 2017. 42(6): p. 1697- 1709. 126. Yudkoff, M., Interactions in the Metabolism of Glutamate and the Branched- Chain Amino Acids and Ketoacids in the CNS. Neurochemical research, 2017. 42(1): p. 10-18. 127. Sadasivudu, B. and A. Lajtha, Metabolism of amino acids in incubated slices of mouse brain. Journal of neurochemistry, 1970. 17(8): p. 1299-1311.

100

128. Segel, I.H., Biochemical calculations. 1975: Wiley. 129. GÜttler, f., e.s. Olesen, and e. Wamberg, Diurnal variations of serum phenylalanine in phenylketonuric children on low phenylalanine diet. American Journal of Clinical Nutrition, 1969. 22: p. 1568-1570. 130. Moats, R.A., et al., Brain phenylalanine concentrations in phenylketonuria: research and treatment of adults. Pediatrics, 2003. 112(Supplement 4): p. 1575-1579. 131. Burgard, P., et al., Phenylalanine hydroxylase genotypes, predicted residual enzyme activity and phenotypic parameters of diagnosis and treatment of phenylketonuria. European Journal of pediatrics, 1996. 155: p. S11-S15. 132. Williams, R.A., C.D. Mamotte, and J.R. Burnett, Phenylketonuria: an inborn error of phenylalanine metabolism. The Clinical Biochemist Reviews, 2008. 29(1): p. 31. 133. Romani, C., et al., The impact of phenylalanine levels on cognitive outcomes in adults with phenylketonuria: Effects across tasks and developmental stages. Neuropsychology, 2017. 31(3): p. 242. 134. Hood, A., et al., Variability in phenylalanine control predicts IQ and executive abilities in children with phenylketonuria. Molecular genetics and metabolism, 2014. 111(4): p. 445-451. 135. Viau, K.S., et al., Correlation of age-specific phenylalanine levels with intellectual outcome in patients with phenylketonuria. Journal of inherited metabolic disease, 2011. 34(4): p. 963-971. 136. Hood, A., et al., Prolonged exposure to high and variable phenylalanine levels over the lifetime predicts brain white matter integrity in children with phenylketonuria. Molecular genetics and metabolism, 2015. 114(1): p. 19-24. 137. Taslimifar, M., et al., Functional polarity of microvascular brain endothelial cells supported by neurovascular unit computational model of large neutral amino acid homeostasis. Frontiers in Physiology, 2018. 9: p. 171. 138. Anastasoaie, V., et al., Stability of blood phenylalanine levels and IQ in children with phenylketonuria. Molecular genetics and metabolism, 2008. 95(1): p. 17- 20. 139. van Vliet, D., et al., Therapeutic brain modulation with targeted large neutral amino acid supplements in the Pah-enu2 phenylketonuria mouse model. The American journal of clinical nutrition, 2016. 104(5): p. 1292-1300. 140. van Vliet, D., et al., Large neutral amino acid supplementation exerts its effect through three synergistic mechanisms: proof of principle in phenylketonuria mice. PLoS One, 2015. 10(12): p. e0143833. 141. Schindeler, S., et al., The effects of large neutral amino acid supplements in PKU: an MRS and neuropsychological study. Molecular genetics and metabolism, 2007. 91(1): p. 48-54. 142. van Spronsen, F.J., et al., Large neutral amino acids in the treatment of PKU: from theory to practice. Journal of inherited metabolic disease, 2010. 33(6): p. 671-676. 143. Möller, H.E., et al., Kinetics of phenylalanine transport at the human blood– brain barrier investigated in vivo. Brain research, 1997. 778(2): p. 329-337. 144. Ferguson, C. and A. Morris, Changes in serum phenylalanine after overnight fasts in youngsters with phenylketonuria. Journal of human nutrition and dietetics, 1999. 12(3): p. 213-218.

101

145. MacDonald, A., et al., Does a single plasma phenylalanine predict quality of control in phenylketonuria? Archives of disease in childhood, 1998. 78(2): p. 122-126. 146. Mitchell, J.J., Y.J. Trakadis, and C.R. Scriver, Phenylalanine hydroxylase deficiency. Genetics in medicine, 2011. 13(8): p. 697. 147. Michals-Matalon, K., et al., Response of phenylketonuria to tetrahydrobiopterin. The Journal of nutrition, 2007. 137(6): p. 1564S-1567S. 148. Hargreaves, K. and W. Pardridge, Neutral amino acid transport at the human blood-brain barrier. Journal of Biological Chemistry, 1988. 263(36): p. 19392- 19397. 149. Martynyuk, A., F. Van Spronsen, and E. Van der Zee, Animal models of brain dysfunction in phenylketonuria. Molecular genetics and metabolism, 2010. 99: p. S100-S105. 150. Suraweera, A., et al., Failure of amino acid homeostasis causes cell death following proteasome inhibition. Molecular cell, 2012. 48(2): p. 242-253. 151. Fernstrom, J.D., Dietary amino acids and brain function. Journal of the American Dietetic Association, 1994. 94(1): p. 71-77. 152. Lieberman, H.R., Amino Acid and Protein Requirements: Cognitive Performance, Stress, and Brain Function. 2000, ARMY RESEARCH INST OF ENVIRONMENTAL MEDICINE NATICK MA. 153. Fernstrom, J.D., Branched-chain amino acids and brain function. The Journal of nutrition, 2005. 135(6): p. 1539S-1546S. 154. Cansev, M. and R. Wurtman, 4 Aromatic Amino Acids in the Brain, in Handbook of neurochemistry and molecular neurobiology. 2007, Springer. p. 59-97.

102

List of publications

1) Mehdi Taslimifar, Stefano Buoso, François Verrey, and Vartan Kurtcuoglu." Propagation of plasma L-phenylalanine concentration fluctuations to the neurovascular unit in phenylketonuria: An in silico study". This manuscript (chapter 4 of the thesis) has been submitted to the Journal of the Royal Society Interface.

2) Mehdi Taslimifar, Lalita Oparija, François Verrey, Vartan Kurtcuoglu, Ufuk Olgac, and Victoria Makrides. "Quantifying the relative contributions of different solute carriers to aggregate substrate transport." Scientific reports 7 (2017): 40628.

3) Mehdi Taslimifar, Stefano Buoso, François Verrey, and Vartan Kurtcuoglu. "Functional Polarity of Microvascular Brain Endothelial Cells Supported by Neurovascular Unit Computational Model of Large Neutral Amino Acid Homeostasis." Frontiers in Physiology 9 (2018): 171.

103

Curriculum Vitae

Mehdi Taslimifar (born: 15th May 1988, Iran)

Education APR2014- Jun2018 Ph.D., Life Science Zurich Graduate School University of Zurich, Zurich, Switzerland Thesis title: Quantification of amino acid transporter interactions Advisors: Prof. Vartan Kurtcuoglu and Prof. François Verrey

SEP2010- NOV2012 M.S., Mechanical Engineering Sharif University of Technology, Tehran, Iran GPA: 18.55/20 Thesis title: Computational modeling, design optimization and experimental investigation of nanofluidic pulsating heat pipe systems Advisors: Prof. M.B. Shafii and Prof. M.H. Saidi

SEP2006- SEP2010 B.S., Mechanical Engineering Sharif University of Technology, Tehran, Iran GPA: 18.13/20 Thesis title: Dynamic modeling and design optimization of micro-CHP systems Advisor: Prof. S.Jani

104

Work experience

APR2014- Jun2018 Research assistant, University of Zurich, Switzerland

Responsibilities: • Quantification of carrier mediated trans-membrane transport across single membranes • Computational modeling of dynamical interactions between large neutral amino acid transporters in the central nervous system • Computational modeling of perturbed large neutral amino acid transport pathways in Phenylketonuria disease • Teaching assistant for 4 medical practical courses

JAN2012--APR2014 Research assistant, Sharif University of Technology, Iran

Responsibilities: • Computational modeling, design optimization and experimental investigation of PHPs • Teaching assistant in the computational fluid dynamic course

JAN2012-JA2011 Research assistant, Sharif University of Technology, Golpayegan, Iran

Responsibilities: • Computational research consultant for about 800 bachelor students in engineering science • Teaching assistant for 6 courses in advanced mathematics and mechanics

Honors and awards • Received the 2017-18 extension grant from SystemsX.ch, The Swiss Initiative in Systems Biology, November 2015. • Selected member of Iran's National Elites Foundation (INEF), January 2014. • Ranked 2nd among ~100 BSc students of Department of Mechanical Engineering, Sharif University of Technology, Golpayegan, 2010. • Ranked 18th among ~20,000 students in Iranian National Entrance Exam, 2010. Skills • Computer Skills: – Data analyzing: Origin, GraphPad Prism, Systems biology softwares – Programming Languages: C++, Python, MATLAB, R, Fortran • Soft skills: – Analytical thinking and problem-solving skills – Team working skills in multidisciplinary projects – Writing skills • Languages: - English: Fluent - German: A2 level - Persian: Mother language

Publications – Taslimifar, M., Buoso, S., Verrey, F., Kurtcuoglu, V., (2018). Functional polarity of microvascular brain endothelial cells supported by neurovascular unit computational model of large neutral amino acid homeostasis, Frontiers in physiology 9, 171. – Taslimifar, M., Buoso, S., Verrey, F., Kurtcuoglu, V., (2018). Propagation of plasma L-phenylalanine concentration fluctuations to the neurovascular unit in phenylketonuria: An in silico study, manuscript submitted. – Taslimifar, M., Oparija, L., Verrey, F., Kurtcuoglu, V., Olgac, U., & Makrides, V. (2017). Quantifying the relative contributions of different solute carriers to aggregate substrate transport. Scientific reports, 7, 40628. – Taslimifar, M., Mohammadi, M., Afshin, H., Saidi, M. H., & Shafii, M. B. (2013). Overall thermal performance of ferrofluidic open loop pulsating heat pipes: an experimental approach. International Journal of Thermal Sciences, 65, 234-241. – Mohammadi, M., Taslimifar, M., Saidi, M. H., Shafii, M. B., Afshin, H., & Hannani, S. K. (2014). Ferrofluidic open loop pulsating heat pipes: efficient candidates for thermal management of electronics. Experimental Heat Transfer, 27(3), 296-312. – Mohammadi, M., Taslimifar, M., Haghayegh, S., Hannani, S. K., Shafii, M. B., Saidi, M. H., & Afshin, H. (2014). Open-loop pulsating heat pipes charged with magnetic nanofluids: powerful candidates for future electronic coolers. Nanoscale and Microscale Thermophysical Engineering, 18(1), 18-38. 105