Graphene field-effect transistors functionalization for neurotransmitter biosensing

Universidade do Minho Escola de Engenharia Mafalda Inês Gonçalves de Abrantes Graphene field-effect transistors functionalization for neurotransmitter biosensing dezembro de 2021 UMinho | 2021 Mafalda Abrantes Graphene field-effect transistors functionalization for neurotransmitter biosensing Mafalda Inês Gonçalves de Abrantes Graphene field-effect transistors functionalization for neurotransmitter biosensing Dissertação de Mestrado Mestrado Integrado em Engenharia Física Dispositivos, microssistemas e nanotecnologias Trabalho efetuado sob a orientação do Professor Doutor Luís Ricardo Monteiro Jacinto e do Professor Doutor João Pedro dos Santos Hall Agorreta de Alpuim Universidade do Minho Escola de Engenharia dezembro de 2021 ii COPYRIGHTS AND TERMS OF USE OF WORK BY THIRD PARTIES This is an academic work that can be used by third parties as long as the internationally accepted rules and good practices are respected, concerning copyright and related rights. Thus, this work can be used under the terms set out in the license below. If the user needs permission to be able to make use of the work under conditions not provided for in the indicated license, he must contact the author, through the RepositóriUM of the University of Minho. License granted to users of this work CC BY--NC-SA https://creativecommons.org/licenses/by-nc-sa/4.0/ iii ACKNOWLEDGMENTS I would like to express my gratitude to my supervisors, Prof. Dr. Luís Jacinto and Prof. Dr. Pedro Alpuim, for the support, constant availability, and guidance throughout the thesis. Most of this thesis work was carried out at the International Iberian Laboratory, and so I must thank the 2DMD (2D Materials and Devices) group members for all valuable input. I also would like to thank the ICVS’ researchers (Life and Health Sciences Research Institute) that assisted me with experimental procedures on the later part of this thesis. iv STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Code of Ethical Conduct of the University of Minho. v RESUMO A compreensão de como o cérebro funciona é fundamental para o desenvolvimento de novos métodos de diagnóstico e terapêutica para as doenças que afetam o cérebro. O grafeno, devido às usas propriedades únicas, tem sido exponencialmente utilizado para novas aplicações biomédicas e é um material emergente para biossensores, incluindo no cérebro. A dopamina é um neurotransmissor, a mensagem química produzida pelas células no cérebro, envolvida em várias funções críticas como o controlo motor e a aprendizagem. Logo, alteração dos níveis de dopamina está relacionada com várias doenças que afetam o cérebro, incluindo Doença de Parkinson, esquizofrenia e abuso de substâncias. De forma a obter um método de deteção para dopamina mais seletivo e sensível, neste trabalho foi desenvolvido um novo sensor bio eletrónico. O canal de grafeno mono-camada de transístores de efeito de campo de porta líquida foi funcionalizado com uma sonda, o aptâmero de DNA específico para a dopamina. Quando o aptâmero se liga à dopamina, a redistribuição da carga eletrónica na superfície do grafeno gera uma mudança no campo elétrico que altera a condutividade eletrónica no canal e a resposta geral do dispositivo. Seguindo as alterações da condutividade com a adição de concentrações conhecidas de dopamina, definição das curvas de calibração dos sensores foi possível. Para tal, experiências de deteção de dopamina foram realizadas in vitro em solução salina de fosfato tampão e em líquido cefalorraquidiano artificial. No primeiro foi obtido um limite de deteção de 1 aM, o mais baixo reportado para um sensor de dopamina e no segundo o limite foi de 10 aM. A seletividade dos sensores também foi testada comparando as respostas da dopamina aos seus percursores, como L-Dopa e L-Tirosina, e a agentes interferentes como ácido ascórbico. O uso de um aptâmero específico para dopamina permitiu uma excelente seletividade, pois a resposta dos sensores a outras moléculas foi negligenciável. A excelente seletividade e sensibilidade dos nossos sensores permitiram também a realização de experiências ex vivo com amostras extraídas de murganhos. Os nossos sensores foram utilizados para demonstrar que a reserpina, uma droga que exauria a dopamina das sinapses neuronais, foi eficiente na redução do conteúdo de dopamina no líquido cefalorraquidiano de murganhos injetados com reserpina em 5 vezes quando comparado controlos. Foi também confirmada a habilidade dos nossos sensores em detetarem dopamina em meios complexos, através da adição de concentrações de dopamina artificial a homogeneizado de cérebro de murganhos tratados, tendo sido obtido um limite de deteção de 0.1 fM. Os resultados inovadores permitem-nos considerar experiências futuras num ambiente in vivo . Palavras chave: Aptasensor, dopamina, grafeno, transístor, baixo LOD vi ABSTRACT The understanding of how the brain works is fundamental for the development of new diagnostic and therapeutic methodologies for brain disorders. Graphene, due to its unique properties, is being increasingly used in a wide range of novel biomedical applications and is an emerging material for biosensing, including in the brain. Dopamine is a critical neurotransmitter, the chemical messages produced by brain cells, is dopamine, which is involved in various functions from motor control to learning. Thus, dopamine dysfunction underlies several brain disorders, including Parkinson’s Disease, schizophrenia and substance abuse. Striving to obtain a more sensitive and selective method of detection for dopamine, we engineered a new bioelectronic sensor for dopamine detection in this work. Electrolyte-gated graphene field-effect transistors’ monolayer graphene channel was functionalized with a molecular linker and a probe, a dopamine-specific DNA aptamer. When this receptor binds with dopamine, the redistribution of electronic charges on the graphene surface generates a change in the electric field which alters the electronic conductivity in the channel and overall device response. By tracking the changes in conductivity with the addition of known concentrations of dopamine calibration curves for our sensors were established. To do so, dopamine detection experiments were carried out in vitro in phosphate buffered saline solution and artificial cerebrospinal fluid electrolytic solutions. In PBS the obtained limit of detection (LOD) was of 1 aM, the lowest ever reported for a dopamine sensor, while in aCSF the LOD was of 10 aM. The sensors selectivity was also tested by comparing responses to dopamine with responses to dopamine precursors such as L-DOPA and L-tyrosine, and biological interferents as ascorbic acid. The use of a dopamine specific-aptamer allowed an outstanding selectivity, with the sensors producing negligeable responses to the other tested molecules. The excellent selectivity and sensitivity of our sensors allowed us the performance of ex vivo experiments with samples extracted from mice. Our novel sensors were used to show that reserpine, a drug that can deplete dopamine from brain synapses, was effective in reducing dopamine content in cerebrospinal fluid (CSF) extracted from reserpine injected mice by 5-fold compared with controls. Taking advantage of brain homogenate samples that were depleted of dopamine by the action of reserpine, the ability of our sensors to perform detection in such a complex media was confirmed by adding known concentrations of dopamine to these samples. Although an expected loss of sensitivity was observed, the LOD in these complex samples was of 0.1 fM. These ground-breaking results allow us to consider further experiments in an in vivo setting. Keywords: Aptasensor, dopamine, graphene, transistor, lowest LOD vii INDEX COPYRIGHTS AND TERMS OF USE OF WORK BY THIRD PARTIES .................................................................. ii ACKNOWLEDGMENTS .......................................................................................................................... iii STATEMENT OF INTEGRITY ................................................................................................................... iv RESUMO .............................................................................................................................................. v ABSTRACT .......................................................................................................................................... vi INDEX ................................................................................................................................................ vii FIGURE INDEX ..................................................................................................................................... ix LIST OF ABBREVIATIONS, INITIALS AND ACRONYMS ............................................................................... xii 1. INTRODUCTION ............................................................................................................................. 14 2. NEUROTRANSMITTER’S BIOSENSING .............................................................................................. 18 2.1. Dopamine .......................................................................................................................... 20 2.2. Dopamine detection state of the art and detection issues ......................................... 22 3. GRAPHENE FIELD-EFFECT TRANSISTORS ......................................................................................... 28 3.1. Graphene ........................................................................................................................... 28 3.2. Graphene field-effect transistor ..................................................................................... 30 3.3. Electrolyte-gated graphene field-effect transistor ....................................................... 32 3.4. Graphene field-effect transistors’ biosensing state of the art ................................... 34 4. FABRICATION OF GRAPHENE FIELD-EFFECT TRANSISTORS ................................................................. 38 4.1. Graphene growth and transfer ....................................................................................... 38 4.2. Wafer fabrication ............................................................................................................. 39 4.3. Chips’ layout ..................................................................................................................... 41 4.4. Chips’ packaging .............................................................................................................. 42 4.5. Signal acquisition system ................................................................................................ 45 4.6. The chip signal.................................................................................................................. 48 4.6.1. Transistor transfer curve physics ..................................................................................... 49 4.6.2. Ionic buffer solutions dependence ................................................................................... 50 4.6.3. Overtime buffer measurements ....................................................................................... 50 5. SURFACE FUNCTIONALIZATION AND PASSIVATION METHOD ............................................................... 54 5.1. Functionalization process ............................................................................................... 54 5.2. Assessment of the functionalization process ............................................................... 57 5.2.1. X-ray photoelectron spectroscopy ........................................................................................ 57 14 1. INTRODUCTION Neurons, the principal cells in the brain, use electrical impulses and chemical messengers to communicate with each other, encoding and processing complex information from the external world. Although technology to record and analyze electrical activity of the brain has been fast evolving over the past 60 years, the monitoring of neurotransmitters, the chemical messengers in the brain, remains a challenge and is an expanding area of active research [1]. Neurotransmitters are present in the nervous system at very low concentrations, they are responsible for proper brain function and their dysfunction underlies mental disorders. Being mixed with many other biochemical molecules and minerals makes their selective detection and measurement difficult. The precise detection of neurotransmitters and their dynamics would allow a better understanding of the brain encodes and processes information and ultimately of brain disorders, and the development of more efficient diagnostics and therapeutics. The first technologies used to detect neurotransmitters involved brain microdialysis, mass spectroscopy, high pressure liquid chromatography (HPLC), and capillary electrophoresis and were later replaced with imaging techniques, proton nuclear magnetic resonance and magnetic resonance imaging [4]. Recent approaches to measure neurotransmitters include biosensors and microelectrodes. However, due to brain’s complexity there are still many challenges like working with complex samples, obtaining good temporal resolution, and working at miniaturized sample preparations such as single neurons. Thus, for accurate detection and monitoring, a neurotransmitter has to be selectively identified among other biomolecules and other neurotransmitters ideally with a miniaturized and localized sensor. Dopamine is a neurotransmitter with critical roles in the human brain and body, and its dysfunction underlies several brain disorders. Dopamine’s function is crucial in the central nervous, renal, cardiovascular, and hormonal systems. High dopamine level indicates cardiotoxicity leading to rapid heart rates, hypertension, heart failure, and drug addiction [20]. However, low dopamine levels can cause stress, Parkinson’s disease [19], Alzheimer’s disease [21], and depression [30]. Despite dopamine’s critical roles, current sensors for its detection typically lack relevant selectivity or sensitivity, or are incapable of being miniaturized for localized detection, which has been hindering the development of reliable diagnostics and potentiating earlier and more efficient treatment for dopamine-related disorders. 15 The most frequently used dopamine biosensors are electrochemical because dopamine being an electroactive molecule can be easily oxidized without an enzyme, thus it can be simply measured using electrochemistry. Nevertheless, this means that dopamine is suffering a redox reaction and similar oxidation signals can be acquired from dopamine interferents and analogues in biological samples or tissues, meaning that these sensors lack the desired selectivity [1]. Dopamine detection through imaging techniques indirectly measure dopamine release with high spatiotemporal resolution but require the use of small identifier molecules (labels), such as fluorescent dyes which poses translational problems. Microdialysis combined with high-performance liquid chromatography is also an approach able to detect extracellular dopamine but lacks spatiotemporal resolution [27]. Different biosensors with enzymes and antibodies were also reported but such molecules lack stability, and present poor limits of detection. Aptamer-based sensors which include oligonucleotides or peptide molecules that are synthetically designed to bind to a specific target molecule as the recognition element, are a novel approach which has remarkable flexibility and convenient design of their structures has led to novel biosensors that have exhibited high sensitivity and selectivity [23]. To possibly overcome current limits and advance the state-of-the-art of dopamine detection, we engineered a novel solution combining a dopamine-specific DNA aptamer and graphene fieldeffect transistors. Graphene is a monolayer of carbon arranged in a hexagonal lattice. Graphene and graphenebased materials are being increasingly used in a wide range of fields from electronics to biomedical applications due to their unique properties. These properties vary from lightness, great flexibility, high thermo-electrical conductivity, low noise, optical transparency, and biocompatibility to functionalization. Making it an outstanding material for sensors, overcoming the flaws of silicon biological and chemical sensors like, wafer production that relies on toxic products which reduces their intrinsic biomedical applications, silicon’s reduced biocompatibility and small sensing surface area which hinders its performance in terms or efficiency and sensitivity. Another major disadvantage of silicon-based devices is their temperature dependence which makes them very limited when used for in vivo monitoring [73]. Graphene-based biosensors have been used for the detection of biologically relevant molecules such as glucose, DNA, or cholesterol [1] and are an emerging approach for brain sensing. This 2D material can be functionalized to detect very low concentrated analytes due to its one atom 16 thickness and excellent carrier mobility. Its non-toxicity, stability (due to strong covalent bonds in plane) and biocompatibility also make it a good contender for in vivo experiments in brain cells. A type of graphene-based biosensor is the graphene field-effect transistor (GFET). This device is based on the typical field-effect transistor effect where a gate contact modulates the electronic conductance of the graphene channel connecting the source and drain. Additionally, for biosensing, the graphene channel can be modified with an immobilized biorecognition probe for the target of interest. The EG-GFET is a particular class of GFETs that uses an electrolytic solution as the gate dielectric. This liquid medium is convenient for bringing the molecules into contact with graphene while operating the GFET at very low voltages (1 V max) thanks to the enormous capacitance resulting from the very thin electrical double layer (EDL) that form at the graphene-liquid interface. We used electrolyte-gated graphene field-effect transistors fabricated at the International Iberian Nanotechnology Laboratory (INL) and functionalized graphene’s channel with a dopaminespecific DNA aptamer probe. When dopamine binds to the aptamer on the graphene surface, the electronic charge redistribution changes the electric field at the graphene surface, resulting in ‘local gating’, i.e., in the local modulation of the electrostatic gate field. Local gating alters the electronic conductivity in the channel and the overall device response, which in graphene transistors is a “V”- shaped curve whose minimum is abusively called the Dirac point. The name stems from a singularity in the graphene electronic density of states (DOS), called the Dirac point, located at the point where the valence and conduction bands meet. At the Dirac point, the DOS is null, and therefore, the conductivity has a minimum when the Fermi level is close to that energy. In the EGGFET biosensor, the transistor Dirac point shift, induced by local gating, is proportional to the concentration of the target molecules bound to the immobilized probes. Since dopamine detection with this aptamer by a graphene transistor has never been reported before, a graphene functionalization method had to be developed and assessed. After that, dopamine detection experiments were carried out in vitro in electrolyte buffers and in ex vivo samples from mice models. Dopamine detection experiments in vitro were carried out in hundreds of sensors (transistors) and the lowest limit of detection ever reported for dopamine was obtained, at attomolar concentration, in phosphate buffered saline (PBS) and artificial cerebrospinal fluid (aCSF). Additionally, dopamine was also detected with great sensitivity in ex vivo experiments with cerebrospinal fluid and brain striatum homogenate samples extracted from mice. 17 Combining these two novel methods, target recognition by aptamers with graphene sensing could allow new limits of detection to be reached for dopamine, and therefore potentiate better understanding of how this neurotransmitter works in the brain. This thesis is organized as follows. In chapter 2, a review of neurotransmission, the roles of dopamine in the brain and methodologies for dopamine biosensing are presented. In chapter 3, a review of graphene, graphene field-effect transistors and their biosensing state of the art can be found. In chapter 4, the fabrication, assessment and characterization of our electrolyte-gated graphene field-effect transistors is described. In chapter 5, our surface functionalization and passivation method is presented as well as its characterization by spectroscopy techniques. In chapter 6, in vitro experiments for dopamine detection with our sensors in different electrolytes are described, as well as target selectivity experiments with dopamine analogues and interfering agents. In chapter 7, ex vivo experiments with dopamine detection in biological samples are described. 18 2. NEUROTRANSMITTER’S BIOSENSING A neuron is an electrically excitable cell that communicates with other cells through specialized connections, synapses. It is the functional and structural unit of our brain and the entire nervous system and consists of a cell body (soma), dendrites, and axon. Figure 1 shows an exemplificative illustration of a neuron’s structure. Figure 1. Exemplificative illustration of a neuron cell [30]. The soma is localized at the center of the neuron, it contains its nucleus and is where protein synthesis, and metabolic activity takes place. Dendrites are the numerous extensions in form of small branches from the neural body which relay inputs from other neurons to the soma. The axon is the main prolongation of the neuron, and it transports neuronal signals away from the soma to other neurons. The axon terminates in axonal terminals which contact the dendrites of other neurons. Neurons communicate with each other via electrical events called action potentials and chemical molecules called neurotransmitters. At the junction between two neurons, lies the synapse, a specialized structure, where the axon terminal from one neuron (pre-synaptic neuron) contacts with a dendrite of another neuron (post-synaptic neuron). Neurotransmission is the process by which neurotransmitters are released by the axon terminal of the presynaptic neuron and bind to and react with the receptors on the dendrites of the postsynaptic neuron. Neurotransmitters are released from synaptic vesicles, by incoming action potentials generated at the soma, into the synaptic cleft where they are received by neurotransmitter receptors on the target cell. The neurotransmitter can either excite or inhibit the other neuron from triggering its own action potential. Figure 2 shows a neurotransmission event. 19 Figure 2. Exemplificative illustration of a neurotransmission event. The presynaptic neuron (top) releases a neurotransmitter, which activates receptors on the nearby postsynaptic cell (bottom) [ 29] . A neurotransmitter is a signaling molecule secreted by a neuron and its purpose is to affect another cell through a synapse. To be considered a neurotransmitter a substance has to be synthesized in the neuron, be found in the presynaptic end and released to have an effect in the postsynaptic cell, then be imitated by exogenous application to the postsynaptic cell and have a specific mechanism for termination of its action [31]. As such, more than 100 neurotransmitters have been identified and can be categorized according to varying criteria. One exemplificative classification divides neurotransmitters in different categories like monoamines (e.g. dopamine, norepinephrine, epinephrine, histamine, serotonin), amino acids (e.g. glutamate, aspartate, D-serine, gamma-Aminobutyric acid) and peptides (e.g. oxytocin, somatostatin). A single neuron can synthetize and release different types of neurotransmitters and can contain receptors for several types of neurotransmitters. This creates a complex environment of chemical communication in the brain, critical for the encoding and processing of information. Many neurotransmitters are synthesized from simple and plentiful precursors such as amino acids, which are readily available and only require a small number of biosynthetic steps for conversion [3]. Because neurotransmitters play a fundamental role in neurotransmission, their dysfunction underlies a wide range of mental disorders including depression, drug dependence, schizophrenia, and degenerative diseases among many others [2]. Thus, the development of novel methods to detect neurotransmitters in the brain is paramount to the development of more efficient diagnostics and therapeutics for brain disorders. 20 2.1. Dopamine Dopamine, 3,4-dihydroxyphenethylamine, is a neurotransmitter with critical roles in the human brain and body. It is part of the monoamine neurotransmitters, and in particular of the catecholamine group, because it consists of a catechol structure (a benzene ring with two hydroxyl side groups) with one amine group attached via an ethyl chain. As such, dopamine is the simplest possible catecholamine. In fact, dopamine constitutes about 80% of the catecholamine content in the brain [32]. Figure 3. Dopamine’s chemical structure. Dopamine function is crucial in the central nervous, renal, cardiovascular, and hormonal systems. In blood vessels, it inhibits norepinephrine release and acts as a vasodilator; in the kidneys, it increases sodium excretion and urine output; in the pancreas, it reduces insulin production; in the digestive system, it reduces gastrointestinal motility and protects intestinal mucosa; and in the immune system, it reduces the activity of lymphocytes. With the exception of blood vessels, dopamine is synthesized locally and exerts its effects near the cells that release it. A significant amount of dopamine circulates in the bloodstream, mostly in the form of dopamine sulphate, but its functions there are not entirely clear [5]. Since dopamine is incapable of crossing the blood-brain barrier, it must therefore be synthesized inside the brain to perform its neuronal activity. Dopamine synthesis pathways involves several molecules. The main pathway for dopamine synthesis is through tyrosine, a non-essential amino acid, which is transported across the blood brain barrier. L-Tyrosine is converted into L-3,4dihydroxyphenylalanine (L-Dopa) by tyrosine hydroxylase. The amino acid L-Dopa is the direct precursor of dopamine because it can be converted into dopamine with help of the aromatic amino acid decarboxylase, by removing the carboxyl group [25]. 21 Figure 4. Dopamine’s synthesis pathways [34]. Homovanillic Acid (HVA); Tyrosine Hydroxylase (TH); Dopamine B-Hydroxylase (DBH); Monoamine Oxidase (MAO); Aldehyde Dehydrogenase (AH); Dopamine Decarboxylase (DDC); 3 Methoxytyramine (3-MT); 3,4 Dihydrophenylacetaldehyde (DHPA); 3-Methoxy-4-Hydroxyphenylacetaldehyde (MHPA). Dopamine is broken down into inactive metabolites by a set of enzymes, monoamine oxidase, catechol-O-methyl transferase, and aldehyde dehydrogenase, acting in sequence. Different breakdown pathways exist but the main end-product is Homovanillic acid, which has no known biological activity [5]. The brain includes several distinct dopamine pathways. One example is the mesolimbic pathway which transmits dopamine from the ventral tegmental area – the source of most dopaminergic neurons in the brain – to the striatum and plays a major role in the motivational component of reward-motivated behavior [32]. In this pathway, positive events and stimuli increase the levels of dopamine in the striatum, as well as the anticipation of this events and stimuli [32]. Thus, disruption of dopamine transmission in this pathway can lead to attention deficit disorder [33], schizophrenia [20] and substance addition [27]. Another important brain dopamine pathway is the nigrostriatal pathway which transmits dopamine from the substantia nigra pars compacta to the striatum and is involved in motor control [32]. Dysfunction of this pathway can lead to Parkinson’s disease and other motor related disorders [27]. Despite dopamine’s critical roles, current sensors for its detection typically lack relevant selectivity or sensitivity, or are incapable of being miniaturized for localized detection, which has 22 been hindering the development of reliable diagnostics and potentiating earlier and more efficient treatment for dopamine-related disorders. 2.2. Dopamine detection state of the art and detection issues Neurotransmitters are present in the nervous system at very low concentrations, and they are mixed with many other biochemical molecules and minerals, thus making their selective detection and measurement difficult. Measuring neurotransmitters in vivo is only possible on the extracellular space of the central nervous system and to do so stable and accurate measurements with high selectivity and selectivity are required. Neurotransmitter content and release are also studied in vitro through analysis of cells in culture and ex vivo tissue preparations such as brain slices. Although numerous techniques to do so have been proposed in the literature, neurotransmitter monitoring in the brain is still a challenge and the subject of ongoing research [1]. The most recent approaches to detect dopamine include biosensors and microelectrodes. However, there are still many challenges like working with complex samples (neurotransmitters can be mixed with other molecules), obtaining good temporal resolution, and working at miniaturized sample preparations such as single neurons. Thus, for accurate detection and monitoring, a neurotransmitter has to be localized and identified among other biomolecules and other neurotransmitters, as they can have similar behaviors. Earlier methods for dopamine detection included brain microdialysis, followed by HPLC, mass spectroscopy, and capillary electrophoresis. Extracellular brain fluid can be extracted by microdialysis through a probe implanted in the brain and the analytes collected over time can be analyzed by quantitative techniques. HPLC is a procedure for separating components from a mixture of chemical substances by pumping a pressurized liquid solvent, containing the sample mixture, through a column filled with a solid adsorbent material. Sample’s components interact differently with the adsorbent material, creating different flow rates for the different components which leads to the separation of the components as they flow out of the column [36]. Analytes separated in the column can then be analyzed by mass spectroscopy which is an analytical technique that is used to measure the mass-to-charge ratio of one or more molecules present in a sample [37]. This technique has been used extensively to study reward-mediated behaviors in 23 mice, such as drug addiction, and demonstrated an increase in extracellular dopamine but has poor spatiotemporal resolution [27]. Other earlier methods for dopamine detection included imaging techniques such as functional magnetic resonance imaging and [4] positron emission tomography (PET) [35]. It operates with positron emitting ligands that target specific receptors to supervise dopaminergic activity. Its diverse set of ligands and high sensitivity allow it to be used in humans. But the disadvantages surpass the advantages, ligands are expensive, difficult to synthesize, and can have short half-lives; poor spatiotemporal resolution makes it impossible to distinguish ligand signal from its metabolite signal. Function magnetic resonance imaging is non-invasive and can be used in humans. Ligands can be used to increase spatiotemporal resolution or highlight specific structures; however, information is largely structural and not functional [27]. Dopamine indirect imaging techniques involve fluorescent sensors that either co-transit with dopamine through the synaptic vesicle cycle or reversibly bind to dopamine resulting in a change in fluorescence. These methods offer spatial resolutions that provide synapseand circuitlevel detail. Fluorescent microscopy methods for dopamine detection are based on small-moleculeor protein-based sensors that indirectly measure dopamine release with high spatiotemporal resolution. There is a diverse set of fluorescent dyes and protein sensors but they have low biopenetrance and small wavelengths, can be phototoxic and have limited use in vivo [27]. In twophoton microscopy, fluorescent sensors are imaged using two-photon excitation. However in vivo imaging is expensive and difficult to implement [27]. Fluorescent imaging techniques are not practical for human application due to the extent of dopamine neurons within the brain, the limited bio penetrance of ultraviolet and visible light and, since the brain doesn’t work in the optical domain, to do so it is needed to deliver fluorescent molecules, and such is not allowed for humans. The most frequently used dopamine biosensors are electrochemical because dopamine being an electroactive molecule can be easily oxidized without an enzyme, thus it can be simply measured using electrochemistry. The sensitivity and selectivity for dopamine can be enhanced by modification of the electrode with nanomaterials and affinity ligands [10]. For example, fast-scan cyclic voltammetry (FSCV) was the main neurotransmitters detection method during many years that allowed us to better understand dopamine’s functions in distinct sub-regions of the striatum. It is based on fast waveform scans from −0.4 V to +1.3 V at 400 V/s, repeated at 100 ms intervals applied to a microelectrode to quickly oxidize and reduce electroactive species at the electrode’s 30 The Fermi energy, or Fermi level, EF (defined as the energy difference between the highest and lowest occupied single-particle states in a quantum system of non-interacting fermions at absolute zero temperature) [70] for intrinsic graphene is at the Dirac energy, i.e., the energy of the Dirac point. The Fermi energy can be considerably different from the Dirac energy in doped graphene devices. Since there is no band-gap, the Fermi level can move continuously between bands, which has significant consequences in the characteristics of graphene field-effect transistors, as discussed in the next section. After the clean-room fabrication process, the graphene channel in the GFETs is p-doped. As discussed in the following chapters, the doping level changes as different materials contact graphene throughout the work. Because of its unique band structure, graphene displays novel transport effects such as minimum conductivity and ambipolar field-effect transport, where carriers can be changed from electrons to holes and vice-versa by gating with an electric field, a possibility absent in most materials [41]. Graphene’s carrier mobility, as high as 105 cm2V-1s-1, is the largest ever reported for semiconductors or semimetals [46]. Even at the highest electric-field-induced carrier concentrations, mobility remains high and is barely affected by chemical doping [71]. In polycrystalline graphene deposited by CVD, the carrier mobility at room temperature is lower, typically in the range ~103-104 cm2V-1s-1. In the CVD graphene used in this work, the measured carrier mobility was in the range 2000-6000 cm2V-1s-1. 3.2. Graphene field-effect transistor The field-effect transistor (FET) is a three-terminal (gate, source, and drain) device where a gate voltage (VGS) controls the channel's conductance between source and drain. There is no current flow between gate and source, possibly apart from a small leakage current [74]. In the saturation regime, the source-drain current (IDS) is independent of the source-drain voltage (VDS) and only depends on VGS (figure 6). This regime is unattainable by the graphene FET, which never saturates because graphene is a zero bandgap semiconductor. To explain why it is necessary to recall that VGS modulation of the channel conductance is achieved by moving the Fermi level across the DOS of the channel material. In semiconductors, once VGS is large enough to place EF in the semiconductor forbidden energy bandgap, then there are no new states available for electron transport and IDS saturates (apart from second-order effects that we do not discuss here). In 31 graphene, EF moves continuously between the valence and conduction bands, and therefore the GFET never saturates, while it also never turns OFF (figure 7). Figure 7. GFET’s DOS and transfer curve. Dirac cones (left) and graphene transfer curve (right). When EF moves from the valence to the conduction band through the Dirac point – where the DOS is null – the carrier concentration derivative with respect to VGS changes sign but is never zero. Therefore, IDS exhibits a minimum when EF is at the Dirac point and increases again as EF penetrates the next band (the effect is symmetric and is observed when moving from valence to conduction band and vice-versa). Because of the linear energy-momentum dispersion relation for electrons occupying states close to the Fermi level, in the Dirac cones, the GFET transfer curve presents two linear symmetric branches relative to the point of the lowest channel conductance, also called the charge neutrality point (see figure 8, right). For comparison, figure 8 shows the characteristic transfer curve (IDS as a function of VGS) for a silicon field-effect transistor. Figure 8. FET’s VDS as function of IDS curve (left), FET’s VGS as function of ID curve (middle) and GFET’s VGS as function of IDS (right). 32 3.3. Electrolyte-gated graphene field-effect transistor Like its 3D semiconductor counterpart, the graphene field-effect transistor (GFET) can be made in the top and bottom gate geometries. Figure 9 is an exemplificative representation of a GFET. Figure 9. Exemplificative representation of a graphene field-effect transistor [68], drain (D), source (S) and gate (G). Bulk semiconductors such as silicon, used in standard FETs are three-dimensional, while the FET conduction channel is 2D and forms at the semiconductor surface. The surface of a 3D material is a locus of lattice reconstruction and defects, namely dangling bonds, because the surface atoms have a very different neighborhood than bulk atoms. Consequently, 3D semiconductors need a passivated interface to form the transistor channel. This passivation is obtained by covering the semiconductor surface with a dielectric material that, in most cases, acts as the gate dielectric. The quality of the interface Si/SiO2 is one of the reasons why today’s prevalent CMOS technology is based on silicon. Thus, surface defects, or dangling bonds, usually dominate in any non-passivated semiconductor surface, preventing a channel from forming for transistor action. In this context, the great advantage of graphene is that being only one atom thick, it suffers no surface reconstruction, having no or few dangling bonds. The GFET channel forms naturally only by placing graphene between the source and drain metallic contacts: a graphene layer is a channel. This property, allied with graphene’s high chemical stability, allows operating the GFET in ambient conditions, with the channel exposed to the environment, which is a remarkable fact that we explore extensively in biosensing devices, as discussed in the following paragraphs. Liquid electrolyte-gated graphene field-effect transistors (EG-GFETs) are a particular class of GFETs where a liquid electrolyte drop-casted on the transistors’ surface is used as the gate dielectric. In this case, the gate capacitance is dramatically increased compared to a solid-dielectric 33 transistor because the gate voltage drop occurs only at the electrical double layers (EDLs) that form at the solid-liquid interfaces, namely the electrode-electrolyte and the graphene-electrolyte interfaces. The EDLs are one Debye length thick, a quantity that depends on the ionic concentration in the electrolyte, but in normal conditions, is of the order of 1 to 10 nm. This huge capacitance allows operating the EG-GFET at very low gate voltage (< 1 V), which is essential in working with biomolecules that often possess secondary and tertiary complex structures, susceptible to electric fields and potentials. Moreover, since most biosensing is done in aqueous media, a low voltage is required to avoid water electrolysis (which starts nominally at 1.2 V). Working with biomolecules in aqueous media, it is crucial to consider the Debye-Hückel screening phenomenon where the electrical charge of the molecules is screened by the electrolyte counter-ions (those of opposite charge signal), forming EDLs that are similar to those mentioned above for the solid-liquid interfaces. Again, there is a specific characteristic length scale of this screening given by the Debye length, the distance from a molecule/surface where the electrostatic force is significantly reduced (to 1/e). Since the Debye length is dependent on the electrolyte’s ionic concentration and valence, it is essential to consider which electrolyte solution to use because they affect the electrostatics of the medium and, therefore, the EG-GFET operation. Figure 10 is an exemplificative illustration of the EDL on a EG-GFET. Figure 10. Exemplificative illustration of the electrical double layer (EDL) on a EG-GFET. The EG-GFET biosensor transducing mechanism is local electrostatic gating. The grapheneelectrolyte EDL is perturbed locally and changed by charged or polar species resulting from the biorecognition events between the probes immobilized on the graphene surface and the target biomolecules. 34 3.4. Graphene field-effect transistors’ biosensing state of the art Graphene’s chemical and electronic properties make it an appealing option for constructing biosensing devices as a biorecognition and transduction platform [52]. Graphene-based devices can enable rapid, label-free, high-sensitivity sensors for healthcare point-of-care diagnostics and have the potential to replace other technologies that are high cost, require complex fabrication processes, have low sensitivity, low specificity and lack fast response required for in situ biosensing. While graphene sensing and transducing capability are superlative, as discussed in the previous sections, it responds to any changes in its dielectric environment, and therefore is it not selective for particular molecular targets. Thus, the graphene surface must be functionalized with molecular probes that render it target-selective for biosensing. Generally, graphene does not bind with most materials. However, several surface chemistry strategies enable functionalization by forming binding sites on its surface. In these devices, graphene typically is exposed to permit functionalization of the channel surface and binding of receptor molecules. Bioreceptors such as amino acids, antibodies, enzymes, or aptamers can be added through a linker molecule attached to the graphene surface. Molecules can attach to these sites through covalent bonding, electrostatic, or Van der Waals forces, modifying the graphene dielectric environment or, in some cases inducing charge transfer to graphene [44]. Recently, several well-controlled chemical functionalization procedures compatible with GFETs have been developed. GFETs have been functionalized with proteins, chemical compounds, and DNA molecules to make sensors for various applications. Different chemistries appropriate for the functionalization of graphene devices were reported. They can be based on diazonium compounds that form a covalent bond to the graphene surface [56] or bifunctional pyrene compounds that interact with graphene through a π–π stacking interaction [57]. Due to the aromatic-ring containing amino acids on the surface, many proteins binding onto graphene surface by π–π stacking can be detected by doping effects. The first significant study between chemically-modified graphene transistor and biomolecules was published by Mohanty and Berry in 2008 [58]. Their single-stranded DNA probes were anchored on the surface of the graphene sheet, followed by hybridization with complementary 35 DNA strands functionalized with fluorescence dyes. The hybridization process would affect the current and the electrical field. Graphene has also been explored as a quencher for fluorescence-based detection. Chang et al. labelled an aptamer specific for thrombin with a fluorescent dye and used graphene as a substrate for non-specific adsorption achieving a LOD of 30 pM [53] by measuring fluorescence resonance energy transfer (FRET). Due to the non-covalent assembly between aptamer and graphene, fluorescence quenching of the dye takes place because of the FRET. Electrochemical detection was also performed with graphene materials. For example, Viswanathan et al. [54] reported the use of graphene for enzymatic detection of glucose, based on the reaction between the enzyme and glucose, and achieved millimolar LOD. Despite having graphene as the recognition element, the low sensitivity of this method still prevailed and was caused by the lack of selectivity associated with the redox reactions [50]. In terms of electrical transduction platforms, Dong et al. reported ta label-free use of graphene for electrical detection of single nucleotide mismatch in DNA with GFETs [55]. Graphene was patterned with gold nanoparticles that had the DNA probe attached to their surface, and when the complementary DNA was added, a shift to lower voltages of the Dirac point was observed at the lowest concentration of 100 nM. Guo et al. described the fabrication of graphene transistors with conventional photolithographic patterning process. The resulting devices also show high sensitivity in label-free detection of DNA [59]. Zhu et al. explored thionine-GFETs, a DNA sensor decorated with amino-substituted oligonucleotide as DNA probe through the arm linker, acquiring an excellent linear response and a low detection limit of 0.1 pM [61]. Campos et al. recently reported an electrolyte-gated GFET capable of detecting DNA hybridization with superior sensitivity at attomolar label-free recognition [62]. This work utilized a similar graphene transistor to the one we introduce in chapter 4. Ohno et al. produced label-free biosensors based on aptamer-modified GFETs for selective immunoglobulin E detection. In their experiment, the conductance of a GFET functionalized with immunoglobulin-specific aptamers decreased when the positively charged targets were introduced and bonded to the oppositely charged aptamers [60]. Her et al. detected alanine aminotransferase with a concentration range of 10 to 100 U/L using a graphene field-effect biosensor [63]. 36 Compared to chemical molecules detection, detecting living cells faces more difficulties and challenges because of the complex interaction between graphene and living cell membranes. A biosensor based on GFETs functionalized with E.coil antibodies was successfully fabricated by Huang et al . The detection limit was 10 cfu/m [64]. Because GFETs are highly sensitive to pH, the metabolic activities of living cells could be revealed in real-time through current change since it is affected by the changes of the local pH due to the release of metabolic organic acids. Ang et al. reported malaria-infected red blood cells detection with graphene-based transistors. The sensor array was integrated with microfluidics and achieved the single-cell flow sensing by executing a “flow-catch-release” [65]. An ultra-sensitive and flexible FET olfactory system was also demonstrated by Park et al. After O2 or NH3 plasma treatment to open the graphene bandgap, human olfactory receptors were bonded to amyl butyrate that was conjugated with a graphene platform. The results indicated that the sensitive sensor could achieve a LOD as low as 0.04 fM [66]. Graphene is also an emerging material for brain-sensing applications, and recent works have shown the use of graphene transistors for such applications. GFETs have been explored for detecting electroactive activity with a high signal-to-noise ratio from electrogenic cells, first in cardiomyocytes [81], and more recently with neuron. Blaschkle, Benno M., et al. [76] showed brain electrical activity mapping with a flexible EG-GFET in vivo in rats. It was demonstrated that graphene transistors offer additional advantages compared to existing state-of-the-art microelectrode-based recording technologies, such as intrinsic signal amplification and the possibility for down-scale and high-density integration. GFETs for the detection of action potential from electrogenic cells were also reported. Extracellular signals from cardiomyocytes were detected with a high signal-to-noise ratio [81]. The detection of neurotransmitters is exceptionally challenging due to a low background concentration. Sensors for the detection of neurotransmitters need to have the option of being miniaturized. Besides the device’s size, flexibility is also essential for in vivo applications where good contact between the device and the tissue is required. Thus, this is where GFETs are highly promising. Epinephrine neurotransmitter was detected by its oxidation on a FET gate electrode covered by graphene obtaining a LOD of 1 nM [80]. Dopamine sensing was performed with a reduced-graphene-oxide-based FET and obtained a LOD of 1nM with fast response [79]. 37 Acetylcholine neurotransmitter was detected on polymer-modified CVD-grown graphene transistors functionalized with enzymes. Detection of this neurotransmitter was only possible at 0.5 µM concentration [77]. Enzymes as the functionalization method most likely caused this method’s lack of sensitivity. An EG-GFET sensor based on electrochemical functionalization with enzymatic reaction also detected acetylcholine with a LOD of 2.3 µM with excellent selectivity [78]. GFETs used as biosensors are still in their early stages, however, provided the application of the inherent miniaturization potential of both FET devices and graphene channels, significant progress can be expected in the short future providing deeper results insights into biological processes. Hence, our interest in dopamine detection via GFETs. Graphene’s great qualities in biosensing make graphene-based devices more promising biomedical sensing contenders than silicon sensors. Silicon-based sensors’ most significant drawback is their wafer production that relies on toxic products which reduces their intrinsic biomedical applications and silicon’s reduced biocompatibility. Another major disadvantage of silicon-based devices is their temperature dependence which is very limited when used for in vivo monitoring [73]. 38 4. FABRICATION OF GRAPHENE FIELD-EFFECT TRANSISTORS Graphene biosensors dissemination is only possible with reproducible and scalable fabrication methods. Our transistors’ fabrication protocol has been optimized and is reproducible, allowing the fabrication of almost a thousand chips for biosensing in only one wafer with over 80% yield. The miniaturization of these chips allied with a small acquisition platform and interface are of relevance for in situ measurements. In this chapter we describe the fabrication of electrolyte-gated graphene field-effect transistors. Starting from graphene growth, wafer fabrication and everything related to our chips, from their layout and packaging to signal acquisition. 4.1. Graphene growth and transfer Continuous monolayer graphene films were grown at INL’s cleanroom facility by the 2D Materials and Devices group in an EasyTube ET3000 CVD system made by CVD Corp, USA. The continuous single-layer graphene was synthesized on copper (Cu) substrates by chemical vapor deposition (CVD). The 10 x 10 mm Cu foils were initially chemically treated using a solution of FeCl3, HCl and deionized water for one minute in ultrasound to reduce rugosities and organic contaminations. The copper (99.99 + % purity) foils were placed into a three-zone quartz tube furnace and the system was evacuated to approximately 2 mTorr and then filled with 250-sccm Argon (Ar, 99.999% purity) and 60-sccm Hydrogen (H, 99.999% purity) gas mixture. Once the growth temperature and pressure were reached, methane (0.5 sccm), the carbon precursor was introduced into the chamber. Graphene growth was carried out at 1040 ºC with 6 Torr for twentyfive minutes on both sides of the copper foil. For graphene transfer, a temporary poly(methyl methacrylate) (PMMA) substrate was used. PMMA (< 500 nm) was spin coated onto the top side of the graphene/Cu/graphene sample. Plasma ashing (PVA TePla GiGAbatch, 0.6 mTorr, O2:Ar 1:1, 250 W, 4 min) was performed to remove the graphene from the back side of the sample. Copper was then dissolved by dipping the PMMA/graphene/Cu into a 0.5 M FeCl3 solution for two hours. And finally, PMMA/graphene was 39 cleaned in 2% HCl solution to remove metal precipitates and further washed in deionized water. Figure 11 is a picture of monolayer graphene. Figure 11. Monolayer graphene film on optical microscope with 50 x magnification. 4.2. Wafer fabrication The fabrication sequence of the sensors’ wafer used in this work, was performed at INL’s cleanroom facility by the engineers of the 2D Materials and Devices group as previously reported [82]. The summarized sequence is as follows and is graphically depicted in figure 12. Each chip is composed of 20 sensors (EG-GFETs). A 200 mm silicon (Si) wafer (p-type doped with boron, B) with 100 nm of thermal oxide was used as substrate. The wafer was sputter-coated with Chromium (Cr, 3 nm) as adhesion layer, and Gold (Au, 35 nm) as the conductive layer, and an alumina Al2O3 (20 nm) capping. The source, drain and gate electrodes were patterned by optical lithography and ion milling. A sacrificial layer (TiWN, 5 nm; AlSiCu, 100 nm; TiWN, 15 nm) was sputtered and patterned by liftoff, for the material to be present where graphene is to be etched. 46 the current is converted in voltage which is measured by the Analog-to-Digital Converter (ADC). Figure 20 shows a photograph of the signal acquisition platform. Figure 20. Signal acquisition platform. The computer interface presents a real-time plot in Labview of the twenty sensors’ transfer curves and an overtime plot of the Dirac point, such as the one shown on figure 21. For our measurements a VS of 1 mV was set with a VGS sweep from 0 V to 1 V. It was avoided to insert voltages higher than 1 V not to reach water electrolysis at 1.23 V, and the voltage starts from 0 V because for most transistors, having negative voltages induces more reading hysteresis. Figure 21. Signal acquisition interface. 47 The “prepare for voltage” mode, before each sensor sweep, puts each sensor at the first voltage value from the VGS sweep range to stabilize the sensors for the desired starting voltage. The time to wait after each voltage sweep was 1000 ms, and the acquired data is an average of five ADC readings of gain ten. Data acquisition for a transistor follows the matrix represented in figure 22. “Voltage 1” is applied to the gate (C7), and a voltage sweep is applied. The source, C12, is connected to “Voltage 2” where the VS of 1 mV is fixed and the ADC reads the current between source and drain, C6, where current is converted to voltage. Figure 22. Data acquisition m atrix. The signal acquisition platform generates two files, one with the Dirac point of every measurement and the other with every point of the transfer curves. Data acquisition was performed with only one loop to assess the electronics’ time to process the measurements. With only one loop of data acquisition, we should be able to determine the lowest acquisition time possible, this is of interest to us since in a near future, we would like to detect neurotransmitters as soon they are released from synapses. A synaptic event will occur below 1 ms. The acquisition time presented on the generated file is a sum of the sweep time (time it takes to acquire data) and communication time (time it takes for the data to be sent to a excel file). With the help of a digital oscilloscope, we input a known squared signal wave with set frequency and amplitude in the platform. With the oscilloscope probes precisely placed on our platform, we found out the time it takes to acquire two points by subtracting the area of the output 48 squared curve signal to the input one. The reason why we take two points is because of Nyquist’s sampling theorem, which specifies that a sinusoidal function in time can be regenerated with no loss of information as long as it is sampled at a frequency greater than or equal to twice per cycle. We found out that our sampling time for 2000 points (measuring the highest possible number of points guarantees that there is no extra time breaking the points’ acquisition loop) is 530 ms and the real acquisition time is 322 ms. In conclusion, the Nyquist frequency for our system is 300 µs, but the time it takes to acquire one point is 0.15 ms, meaning that we can acquire data six times faster than the time a synapse takes to happen. 4.6. The chip signal In an electrolyte-gated GFET (EG-GFET) the solid-state gate dielectric is replaced by an aqueous solution with a given ionic strength. The gate voltage is transmitted through the electrolyte through electrical double layers. The electrical double layer (EDL) acts as a capacitor that charges when a gate voltage, is applied. A double layer is a structure that forms on a solid’s surface when it is exposed to a fluid due to the different chemical potentials of the species in both phases. In polar solvents, like water, it starts with a single layer of solvent molecules with their dipoles aligned with the electric field in the spacecharge region (the EDL). The first surface charge layer, either positive or negative depending on the balance between the chemical potentials, consists of counter ions firmly adsorbed onto the surface. The second charged layer is composed of ions attracted to the surface charge via the Coulomb force, electrically partially screening the first layer. A gate voltage leads to a redistribution of the interfacial charge accumulated between the electrolyte and the graphene channel. The space-charge region, i.e., the EDL, acts as a capacitor’s dielectric layer on top of graphene. The gate-voltage modulation of the space-charge layer changes its thickness and, therefore, its capacitance. Any change in the EDL capacitance is reflected in the transistor transfer characteristics. Because the second layer of ions does not entirely screen the charge in the first layer, due to Brownian motion inside the liquid, the space-charge region extends towards the electrolyte's bulk, a distance is usually represented by the Debye screening length, λD. An electrolyte with a considerable Debye length is often preferable for liquid-gated FETs when used in bio-sensing. A larger λD provides a larger volume to accommodate molecules that bind to the molecular probes 49 on the graphene surface. The EG-GFET cannot detect reactions far outside the EDL because of the Debye screening length. The following equation describes the Debye length (λD): Where 𝜀𝑟 represents the electrolyte’s constant; 𝜀0 represents the permittivity of free space; 𝑘𝐵 is Boltzmann constant; T is the temperature in Kelvin. NA is the Avogadro constant; e is the elementary charge. The smaller ionic strength gives rise to a larger λD, making the sensing probe-target interaction unscreened. Hence, as stated above, the sensing performance of EG-GFETs biosensors is susceptible to the ionic strength of electrolyte. 4.6.1. Transistor transfer curve physics Graphene’s transfer curve has a “V” shape where each branch represents the excursion of the Fermi level inside a different electronic band. According to the popular Drude theory, it is common to associate transport in the valence band (the left branch of the transfer curve) to hole carriers and the conduction band to electrons (the right branch of the transfer curve). Our as-fabricated graphene FET are p-doped, a consequence of lithographic processes, and consequently, the charge neutrality point is found at a positive gate-source voltage. For most chips, the transistors first measurement (bare graphene without any functionalization step) displayed a Dirac point around 0.6 V, shown in figure 23. Figure 23. Graphene transfer curve without functionalization, IDS as function of VGS. 𝜆 𝐷= √𝜀𝑟𝜀0𝑘𝐵𝑇 2𝑁𝐴𝑒2𝐼 50 4.6.2. Ionic buffer solutions dependence Since the ionic strength affects the Debye length and the electrical response of the electrolyte-gated biosensors, we studied the effect of using phosphate buffered saline (PBS) with different concentrations, i.e., different ionic strengths in the response of the transistor. It is reported in the technical literature that PBS 1x has an ionic strength of 162.7 mM and a λD of 0.76 nm; PBS 0.1x has an ionic strength of 16.27 mM and a λD of 2.41 nm [83]. Figure 24 shows the graphene’s transfer curve for two PBS concentrations. Figure 24. Graphene transfer curve, IDS as function of VGS for different electrolyte strengths. PBS 0.1x produces a positive Dirac point shift of fifty millivolts relative to PBS 1x. Since the ionic strength of the solution is lower, the Debye length is higher and the EDL capacitance is smaller. Consequently, higher voltage is required to accumulate the same number of charges in the channel, which means that the same current level is obtained at a higher VGS. 4.6.3. Overtime buffer measurements We studied the effect that water evaporation would have on our measurements. To that goal, a volume of 20 µL of phosphate-buffered saline 1x solution was drop cast on a chip and data were acquired overtime as the water evaporated from the droplet. The gate-source voltage was continuously swept from -0.2 V to 0.8 V. 51 Within one hour of evaporation and continuous data acquisition, some effects become noticeable, as the droplet saturates and the salt precipitation is potentiated, increasing the ionic strength of the medium and giving rise to a transfer curve drift to lower VGS, as shown in Figure 25, for the reasons discussed in the previous paragraph. Figure 25. Graphene’s transfer curve with IDS as a function of VGS overtime. Data were acquired continuously over a time interval of half an hour. Black arrow indicates the direction of the data acquisition. When the water droplet becomes flattened and salt deposition occurs substantially on the graphene channels, the transfer curve loses its “V” shape, and negative source-drain currents are observed. New peaks start to built-up at higher VGS becoming the new current maxima. When the droplet completely dries, the curves become since there is no transistor action because the gate dielectric is lacking. The transistors’ current flow is instantaneously restored as soon as the PBS droplet is renewed. However, because the chip’s surface is not cleaned before a new electrolyte volume is dropped, the solution’s ionic strength is higher than the initial one, and therefore new set of curves is centered at a different (lower) gate voltage. The explanation for the negative values of ISD and the overall new shape of the transistor transfer curve whence the water droplet is about to evaporate fully cannot be found in the acquired data. An IDS < 0 means that there is a current source in addition to the FET source contact. This additional source probably stems from redox reactions affecting the highly concentrated salts in the almost entirely evaporated water droplet. 52 The VDirac stability was tested for many VGS cycles in a constant voltage range, for -0.2 V < VGS < 0.8 V, -0.1 V < VGS < 0.9 V, and 0 V < VGS < 1 V in PBS 1x. The transfer curves were continuously acquired for 10 minutes, without allowing the liquid-gate to evaporate, for each VGS interval (see figure 26). VDirac is stable in time for the first two intervals, after a small initial drift. For the last interval, VDirac drifts in time, from ~0.70 V to ~0.82 V. Comparison between the time series shows that VDirac average values change from series to series. Figure 26. Dirac point time dependence as function of VGS. VGS sweep from -0.2 V to -0.8 V (blue), -0.1 V to -0.9 V (red) and 0 V to 1 V (black). We see no reason for the inter-time series Dirac point instability, but we suspect it is associated with the acquisition platform. In order to assess our sensor’s measurement error, we acquired transfer curves for six transistors of the same chip over time. By doing overtime measurements in a very controlled medium, minimizing external factors' interference such as temperature, droplet volume, pH alteration, and salt precipitation, we can determine the transistors’ measurement fluctuations and consider it our sensor’s error. This chip had a 20 µL droplet of PBS 1x as the gate electrolyte, and the volume was kept over time thanks to a controlled humidity chamber. Four sets of six VDirac measurements were acquired for set times, where each point is an average of 5 acquisitions. 53 The average difference from the first measurement to the last one, for the six sensors averages, was 4 mV, meaning that this should be considered our measurements’ error and should be taken into account when analyzing Dirac point shifts of similar magnitude. Such data can be found on figure 27. Figure 27. Dirac point time dependence for the same PBS 1x droplet. Each data point is a measurement from one transistor. 54 5. SURFACE FUNCTIONALIZATION AND PASSIVATION METHOD To achieve a biosensing system based on EG-GFETs, first, we need to functionalize the active transducing area of the biosensor, the graphene FET channel, with a selective biorecognition element. A dopamine-specific DNA aptamer was chosen for its ability to interact with the graphene channel within the Debye length and for high selectivity. As graphene is chemically stable and covalent modification of the material alters its electrical properties, a non-covalent method was used to modify graphene. This chapter describes the functionalization and passivation processes of our sensors and spectroscopy techniques to prove the successfulness of our method. 5.1. Functionalization process The full functionalization process can be found on figure 28. Figure 28. A graphical depiction of the full functionalization process. DDT passivation (first), PBSE probing (second), aptamer binding (third), ethanolamine blocking (fourth) and dopamine binding to the aptamer (fifth). Initially, to prevent non-specific interaction of biological sample and restrict functionalization to the graphene channel, the Au receded gate is passivated with 2 mM of 1Dodecanethiol (DDT), a gold blocking agent, in ethanol for 4 hours. Between every functionalization step the chip’s surface is cleaned with deionized water and blown dry with N2 flow. To facilitate the DNA immobilization to the graphene, a pyrene-derivative, 1-Pyrenebutanoic acid N-hydroxy-succinimide ester (PBSE) diluted in N, N-Dimethylformamide (DMF) was used as a bifunctional linker to create an interface between graphene and the biosensing probe, 20 µL of 10 55 mM of PBSE was incubated onto the graphene channel for 2 hours. The pyrene group of PBSE establishes a very stable non-covalent interaction with graphene (π-π interaction) due to its structural similarity with graphene and remains with a free ester group that will allow the NHydroxysuccinimide reaction (NHS) (reaction where ester-activated crosslinkers react with primary amines in physiologic to slightly alkaline conditions, pH 7.2 to 9, to yield stable amide bonds) with the amine group of the aptamer, the sensing probe. This way, the aptamer will be firmly bond with the PBSE linker without altering the electronic properties of graphene. Two different dopamine-specific DNA aptamer sequences were studied. After probe bonding to PBSE the following aptamer sequence produced the largest Dirac point shift in regard to the PBSE’s Dirac point. The 44-mer sequence of DNA aptamer with 5’ C6-aminolink modification (5' - CGACGCCAGTTTGAAGGTTCGTTCGCAGGTGTGGAGTGACGTCG -3') [15] was synthesized by STABvida. Despite being 57-mer long, the other DNA sequence [15] did not produce a big Dirac point shift as the shorter one for the same electrolyte solution. Having more nucleobases one would expect to see a larger Dirac point shift, since we have more charges the screening effect on the graphene would also be more intense and therefore bigger Dirac point shift. However, that was not the case, a possible reason is that for the same Debye length, only the charges located within such distance, contribute for the screening effect on graphene, and so, the rest of sequence’s length located above the Debye length does not contribute. The dopamine-specific aptamer was then prepared by diluting 20 µM solution of aptamer in MilliQ water and by heating the final solution to 95°C for 10 minutes. This heating procedure removes misfolds in aptamer structures and maximizes correctly folded secondary structures. The aptamer was only dropped on the chip after cooling down to room temperature. For incubation, 20 µL were added on top of the chip and kept for 16 hours in a humid chamber in the dark. Different aptamer concentrations and incubation times were studied, and this set of parameters (20 µM with 16 H incubation) have given the best PBSE-aptamer binding results since they displayed the biggest Dirac point shit in regard to previous functionalization step. After the aptamer immobilization, some free ester groups from PBSE can remain active on the surface, thus to minimize potential non-specific binding, 20 µL of ethanolamine (ETA) were dropped on the chips and incubated for 30 minutes. ETA is a small molecule with an amine group in one extremity that will react with PBSE’s free ester groups, and a hydroxyl group that blocks the reaction of the linker with dopamine and others. The non-ligand PBSE blocking step is the last stage from the surface functionalization process. 62 Figure 34. Oxygen 1s spectrum. Phosphorus 2 p spectrum analysis DNA contributes with PO43-, phosphate, from its phosphate-deoxyribose backbone, with characteristic binding energy at around 134 eV [84]. An additional intense peak was found at 129 eV which may be attributed to the phosphorus atom. Figure 35. Phosphorus 2p spectrum. 63 Phosphorus 2 p peak’s height increased, in regard to PBSE’s sample peak, from 489,15 CPS to 1655,57 CPS, area increased from 3105,26 CPS.eV to 12324,35 CPS.eV, and the atomic percentage increased from 0.3% to 1.25%. 5.2.2. Raman spectroscopy Confocal Raman spectroscopy (Alpha 300R, WITec) experiments were carried out at INL. Raman data analysis and imaging were done with Project FOUR+ software (WITec). Raman spectroscopy has historically been used to probe structural and electronic characteristics of graphite materials, providing useful information on the defects (D-band), in-plane vibration of sp2 carbon atoms (G-band), as well as the stacking orders (2D-band) [87]. The chip’s monolayer graphene channel and dopamine hydrochloride in powder form and diluted form spectrums were analyzed. In the figure below, the red spectrum represents monolayer graphene, where the 2D mode at 2700 cm-1 is bigger than the G mode at 1580 cm-1 indicating that our channel’s graphene is indeed monolayer [88]. Blue spectrum is dopamine’s Raman signature on glass substrate and pink, green and cyan are all dopamine diluted on PBS 1x on a graphene substrate. The wide peak at around 3400 cm-1 is characterized by water and the pink spectrum was taken a few minutes later after green and cyan, the peak has decreased which reflects the droplet evaporation. Figure 36. Monolayer graphene, dopamine in powder and diluted dopamine spectrums. Graphene (red), dopamine powder form (blue) and dopamine diluted in PBS (cyan, green and pink). 64 Such Raman measurements were able to confirm the presence of monolayer graphene in our transistors’ channels and to analyze dopamine’s Raman spectrum. However, the Raman acquisition system lacked the sensitivity to detect dopamine’s spectrum after washing and drying the chips after having dopamine incubated on them. Such experiment could have better results when carried out with surface-enhanced Raman spectroscopy, because it is a technique that offers orders of magnitude increases in Raman intensity. Dirac point voltage variation and x-ray photoelectron spectroscopy successfully proved aptamer binding to PBSE, and thus, our functionalization method. 65 6. DOPAMINE DETECTION IN VITRO Initial in vitro experiments with optimized buffers are fundamental to study biosensors’ responses in a controlled environment, as well as to establish calibration curves. Thus, in this chapter, dopamine detection in vitro is analyzed for phosphate buffer saline solution and artificial cerebrospinal fluid electrolyte solutions. Dopamine effects on our sensor’s surface with different passivation methods are presented, as well as our sensors response to dopamine’s analogues and biological interfering agents’. The data described in this chapter was presented in “NanoPt2021” conference (see abstract on Supplement 1). 6.1. Dopamine detection in phosphate buffer saline (PBS) Dopamine hydrochloride [Sigma-Aldrich] was weighed and diluted in phosphate buffer saline 1x into solutions of different dopamine concentration, ranging from 0.1 zeptomolar (10-22) to 0.1 millimolar (10-4). PBS (NaCl 137 mM, KCl 2.7 mM, Na2HPO4 10 mM, KH2PO4 1.8 mM) at 7.4 pH is commonly used as an electrolytic buffer solution in the experimental assessment of biosensors because is the pH found in biological fluids [89]. Measurements were performed with 20 µL droplets of each concentration pipetted on top of the transistors. Figure 37 shows dopamine detection as function of ΔVDIRAC. Plotted data is an average for 416 transistors that yielded reliable data from around 60 chips. Every dopamine concentration is an average value of at least 80 transistors. 66 Figure 37. Shift in Dirac point ( Δ VDIRAC) as a function of dopamine target concentrations in PBS 1x. The first stage of detection starts on 0.1 zeptomolar (10-22) and ends by 10 zeptomolar (1020). For the first concentration, we calculated, using the Avogadro constant, that there are no dopamine molecules present in the 20 µL volume. Thus, there is no shift in the Dirac point. For 1 (10-21) and 10 zeptomolar concentrations, although we have a +17 mV and +27 mV shift, we also determined that there are no molecules for these concentrations for the given 20 µL volume. However, we cannot explain this initial positive VDIRAC shift from 1 zM to 10 zM and thus, we attribute the cause to measurements’ artefacts. And the 0.1 attomolar (10-19) concentration has only a +2 mV shift from the previous one, which is half the error’s value (4 mV, see chapter 4), meaning that this concentration’s voltage shift cannot be considered as dopamine detection. Therefore, the limit of detection (LOD) was considered to be the 1 attomolar (10-18) concentration. The calculated number of dopamine molecules for this concentration in a 20 µL droplet was 12. With a VDIRAC shift of +50 mV, we were able to detect a very reduced number of dopamine molecules with great sensitivity. In fact, this is the lowest LOD ever reported for dopamine detection by a factor of 3. Additionally, it should be highlighted that this LOD was achieved with a label-free method as opposed to the current state-of-the-art LOD. The attomolar detection sets the start of our linear dopamine detection trend, this second stage is characterized by a linear detection behavior with a constant increase of VDIRAC as dopamine 67 concentrations increase until surface saturation occurs at 1 picomolar (10-12) concentration. The saturation occurs at a +80 mV increase from the baseline, however since the first concentration to be detected had already a shift of +50 mV, this means that our linear range has a response of 40 mV. Figure 38 shows the graphene transfer curve for different dopamine concentrations. Figure 38. Exemplificative transfer curve for one transistor for different dopamine concentrations in PBS 1x. This detection method is based on positive linear increases of Dirac point, and such can be explained by the aptamer’s conformational change every time it binds to dopamine. Before binding to dopamine, the aptamer is fully stretched and away from graphene, the aptamer’s segment is about 16 nm, only about 1 nanometer of its length is within the Debye length. By changing structural conformation upon dopamine binding, more charges fall within the Debye length and contribute to the observed changes in the transfer curve Dirac point. The electrolyte’s monovalent cations, like potassium and sodium are attracted to the aptamer’s phosphate-deoxyribose backbone because it has a single-bounded oxygen group available. With these stable bounds formed, Na-O and K-O, between the cations and the DNA, the overall charge of the DNA becomes less negative, however we believe that this electrolyte’s cations don’t have enough strength to alter the aptamer’s conformation, and so, it is only altered by the dopamine binding. 68 Initially we have two sets of separate electrical double layers. The first EDL is composed by the graphene counter-ions, which are anions like chloride because our graphene is p-doped (it has more positive charges). The second EDL is formed around the aptamers, where the counterions are cations like potassium and sodium, because of DNA’s negative backbone. We hypothesized that once dopamine binds to the dopamine-specific aptamer, the aptamer changes its conformation by wrapping around itself, coming closer to the graphene channel. With this surface alteration we have the aptamers close to graphene, which means we have more negative charges stacked to graphene’s electrical double layer. Consequently, for the same Debye length, we have more negative charges which increases the screening effect of positive charges on the graphene, producing a strong local gating effect. The positive charges induced on graphene are provided by the gold source and drain contacts to which graphene is connected to (see chapter 4). This detection mechanism is clarified in figure 39. Figure 39. Exemplificative illustration of aptamer’s conformational change with dopamine binding in PBS 1x electrolyte solution. On the left, the aptamer is fully stretched in PBS 1x without dopamine and the graphene is mostly p-doped. On the right, dopamine was diluted in PBS 1x and the aptamer binds to it changing its conformation and coming close to graphene which becomes more p-doped. The linear stage of dopamine detection sees a linear increase of ΔVDIRAC with increasing concentrations of dopamine. This is expected every time there is a dopamine concentration increase, there are more aptamers changing their conformation and getting close to graphene. Therefore, we have more negative charges on the electrical double layer which increases graphene’s screening effect of positive charges. The third stage of our calibration curve starts from 1 pM to 0.1 nM (10-10), where surface saturation occurs, meaning that there is a lack of Dirac point shift despite de addition of higher concentrations of dopamine. This can occur due to two different scenarios, either there are no 69 more aptamers for dopamine to bind to, meaning that we have reached aptamer saturation, or, surface saturation has occurred, where there is dopamine binding to the aptamers but there is not space left for them to change their conformation and come close to graphene. In either case, all the negative charges are already in the electrical double layer next to graphene within the Debye length and no more shifts of the Dirac curve are observed. The last part of the calibration curve is characterized by a negative Dirac point shift, starting from the 1 nM (10-9) to 0.1 mM dopamine concentration. Since we had previously reached either aptamer or surface saturation, we hypothesized that this observed negative response of -80 mV could be due to non-specific binding of dopamine to graphene. These results point to aptamer saturation as the most likely event that’s occurring for high dopamine concentrations. Considering that 40% of graphene is not passivated (see chapter 5), dopamine being a positively charged molecule would produce a negative screening effect by directly binding to graphene and a consequent negative Dirac point shift. This would more likely occur through direct charge transfer making graphene become less p-doped. To further investigate this effect and provide support to our hypothesis, several experiments were undertaken and are described in the next sub-chapter. 6.2. Dopamine and graphene direct interaction The observation that high concentrations of dopamine produced a negative shift of the gate voltage in our sensors, lead to the hypothesis that after aptamer saturation dopamine was directly binding to non-passivated graphene in the transistors channel. To test this hypothesis, we measured Dirac point shifts as a function of high dopamine concentrations in bare graphene and in different steps of the passivation/functionalization process. Additionally, we also tested our sensors response to high concentration of dopamine after blocking graphene with polysorbate 20 (Tween 20). Figure 40 shows results of dopamine response as function of ΔVDIRAC for different surface passivation’s. 70 Figure 40. Shift in Dirac point ( Δ VDIRAC) as a function of dopamine target concentrations in PBS 1x for different functionalization methods: bare graphene (pink), PBSE + ETA (blue), PBSE + APT + ETA (red), PBSE + ETA + TW20 (black) and PBSE + APT + ETA + TW20 (green). Our label-free dopamine-specific aptamer functionalization method (red data on figure 40), where we have the negative Dirac point shift trend that cannot be explained by local gating of the aptamer conformational change. For 0.01 µM (10-8) we still have a positive Dirac point shift regarding the baseline, because we started the experiment for high dopamine concentrations, meaning that the aptamers were dopamine-free and still bounded to it producing +40 mV shift. But for the next concentration, 0.1 µM (10-7) we already see a decrease of VDIRAC of -10 mV. The sensitivity of this chip was of 140 mV. To assess if this n-doping of the graphene was related to the aptamer-based detection or any additional interaction between non-bonded dopamine and surface aptamers, we removed the aptamer probe and incubated dopamine on a graphene channel functionalized with the blocked linker, i.e. PBSE bound to ethanolamine. Considering that PBSE only covers approximately 60% of the graphene surface, this means that exposed graphene was present. The progressive addition of higher concentration of dopamine also lead to an overall Dirac point shift of -50 mV (blue markers in figure 40). The first two concentrations did not produce a positive shift, but the response remained at the baseline value which is consistent with the fact that aptamers in this electrolyte give a positive shift when bounded to dopamine. The difference between the range of response 71 between both cases, with and without aptamer, is 50 mV, we were expecting to have a bigger response for the passivation without the probe but that was not the case. One possible theory is that the PBSE-ethanolamine binding occupies more area close to the graphene making it more difficult for dopamine to reach graphene. To further confirm our hypothesis that the observed shift is a product of dopamine directly interacting with graphene, we incubated a high concentration of dopamine (0.1 mM) in bare graphene, which resulted in a Dirac point shift of -210 mV. This was the highest observed response and 4 times higher than that observed for the previous experiment with PBSE and ETA. Since the transistors had all graphene in the channel exposed to the solution, and that other molecules involved in the functionalization process were not present, this experiment suggests that the observed negative shift is likely mediated by a direct interaction of dopamine and graphene. To test if we could block this negative shift response, we added polysorbate 20 (Tween 20) [Sigma-Aldrich] to passivate exposed graphene after PBSE and ETA functionalization steps (black markers in figure 40). Tween 20 is a non-ionic surfactant that strongly interacts with graphene [92]. It is composed of three chemical parts: aliphatic ester chains that prevent non-specific binding of biomolecules, three-terminal hydroxyl groups that are hydrophilic and an aliphatic chain that can easily be adsorbed on a hydrophobic surface like graphene by non-covalent interaction [93]. A previously reported graphene oxide sensor with aptamers combined with Tween 20 showed that the surfactant protected the graphene from cocaine’s non-specific binding [90]. Another graphene biosensor passivated with Tween 20 showed bacteria non-specific binding on graphene could be prevented [91]. The observed response had a linear range of 30 mV for all the incubated concentrations, 60 mV less than a complete functionalization process without Tween 20. This indicates that perhaps Tween 20 is able to block dopamine from reaching graphene. Although tween 20 contains aliphatic ester chains which may prevent dopamine molecules getting closer to the graphene channel, there was still a response of -15 mV for 0.1 mM of dopamine concentration. This is probably because the used concentration of Tween 20 was not enough to block all the graphene or electrostatic effects might be happening. Nevertheless, these results support our hypothesis that the negative shift observed for higher dopamine concentrations is due to a direct interaction of dopamine and graphene. However, this experiment cannot answer the question if this interaction is due to a binding process between dopamine and graphene or due to electrostatic interaction by proximity. 78 Figure 46. Shift in Dirac point ( Δ VDIRAC) as a function of Ascorbic Acid target concentrations for 36 transistors. L-Dopa’s very low response or lack of it for the femtomolar and picomolar concentrations proves that the aptamer and the functionalization process are very selective to dopamine only. For L-Tyrosine the biggest response was of only 10 mV. In figure 47 we have dopamine detection on the femtomolar concentration with a response of +100 mV, whereas L-Dopa gave a -4 mV response for 1 micromolar, despite being 9 times more concentrated than dopamine still gave a negligible response. 79 Figure 47. Comparative graph of dopamine’s response with its analogues and interfering agents’ response. Both Homovanillic Acid and Ascorbic Acid had a considerable positive response of around 20 mV that should be taken into account for more complex medium experiments. However, for the two amino acids precursors of dopamine, which are structurally very similar to dopamine, we barely observed any detection which is indicative of our sensor’s specificity. Our good results are derived by the fact that we used a highly selective biorecognition element method and avoided redox reactions typically used in electrochemical and catalytic detection methods [1], since dopamine and its analogues have similar redox potentials making it difficult to tell apart the neurotransmitters. Nevertheless, some electrochemical dopamine sensors were shown to be more selective to dopamine by modification of the electrodes with nanomaterials and affinity ligands [10], [51], [47], including with graphene [49]. A liquid crystal biosensor for dopamine with signal amplification through functionalization of gold nanoparticles, selectively detected dopamine between molecules with structural similarities such as ascorbic acid, glucose, epinephrine, norepinephrine, tyramine and serotonin but the limit of detection was on the micromolar dopamine concentration [51]. Other dopamine sensors based on the existing RNA [18] or DNA [15] dopamine-specific aptamers also reported high selectivity [16] [105], although it has also been reported that the DNA conversion of the RNA aptamer is not selective against other catecholamine neurotransmitters [104]. In this work we did not test other neurotransmitters from the monoamines groups, such as 80 serotonin and norepinephrine. However, Nakatsuka, Nako, et al. [15] showed that the same dopamine-specific DNA aptamer was highly selective when compared to these neurotransmitters and 5-HIAA, serotonin’s metabolite. Some different works on dopamine detection with a selective method but different recognition elements were also reported. For example, biosensors based on the enzyme tyrosinase have been successfully used for selective determination of dopamine in the presence of ascorbic and uric acids in a biological environment. But the main problem in the practical application of such biosensors is their poor long-term stability and reproducibility [50]. Maciejewska, J., et al. [75] developed a tyrosinase modified electrode, and demonstrated that the presence of these interferents does not affect the selectivity. However, this amperometric detection of dopamine lacked a low LOD (0.1–0.5 μM). Another amperometric biosensor, based on tyrosinase immobilized on a boron-doped diamond electrode, was also able to selectively detect dopamine but with a high LOD of 1.3 μM in presence of ascorbic acid [72]. 6.5. L-Dopa effect on passivation without aptamer To test if the dopamine is in fact binding to exposed graphene, we also used L-Dopa as a negative control in an additional experiment. Increasing concentrations of L-Dopa, from 0.01 µM to 0.1 mM were added in 20 µL droplets to a chip without aptamer but passivated with the blocked linker (i.e. PBSE + ETA). Figure 48 shows L-Dopa detection without aptamer. As L-Dopa is chemically similar to dopamine, only lacks the amine structure, we detected it on a chip without aptamer, functionalized only with the blocked linker, PBSE-ethanolamine. With this experiment we wanted to analyze what effect a molecule identical to dopamine would have on graphene. If we don’t have the negative voltage effect that is present on the dopamine detection for high concentrations, it can indicate that it’s the amine group that is indeed binding to graphene. 81 Figure 48. Shift in Dirac point ( Δ VDIRAC) as a function of L-Dopa target concentrations without aptamer. We observed a positive Dirac point shift with increasing concentrations of L-Dopa. This average response of +30 mV from the baseline, was the opposite of the negative response observed for high concentrations of dopamine (see subchapter 6.1.). This result suggests that dopamine may be binding to graphene through the amine group [98], since L-Dopa is chemically similar to dopamine, only lacking the amine group. 6.6. Dopamine detection in artificial cerebrospinal fluid (aCSF) In order to transition our in vitro detection to an ex vivo setting, first, we need to validate our functionalization process on a more complex electrolyte medium, such as artificial cerebrospinal fluid (aCSF). aCSF is a buffer solution prepared with a composition representative of cerebrospinal fluid (CSF, a clear fluid that surrounds the brain and spinal cord). aCSF is used experimentally to immerse isolated brains, brain slices, or exposed brain regions to maintain osmolarity, and to buffer pH at biological levels and commonly utilized for electrophysiology experiments to maintain the 82 neurons that are being studied. The reason why we use aCSF is because it is the closest extracellular medium to biological CSF. Fresh aCSF was prepared whenever experiments took place. aCSF solution is composed by 127 mM NaCl, 1.0 mM KCl, 1.2 mM KH2PO4, 26 mM NaHCO3, 10 mM D-glucose, 2.4 mM CaCl2 and 1.3 mM MgCl2. The following graph shows our dopamine detection in aCSF electrolyte for 54 transistors from 0.1 aM to 1 nM concentrations and figure 50 shows graphene transfer curve for the same dopamine concentrations. Figure 49. Shift in Dirac point ( Δ VDIRAC) as a function of Dopamine target concentrations for 54 transistors in aCSF medium. 83 Figure 50. Graphene transfer curve for different dopamine concentrations in aCSF 1x. For the first two concentrations, 0.1 aM and 1 aM, we have a Dirac point shift of +30 mV and +35 mV respectively. One of the reasons for these concentrations to have a positive response whereas higher dopamine concentrations have negative response, could be because for low dopamine concentrations the negative backbone of the DNA dictates the electrical double layer’s charges, while for higher concentrations the electrolyte’s interaction with the aptamer can no longer be neglected. Another possibility is that this positive shift is caused by measurements’ artefacts that we can not explain as of now. Therefore, we should exclude 0.1 aM and 1 aM as dopamine detection, making the 10 aM the limit of detection with a Dirac point shit of -45 mV in respect to the previous dopamine concentration. Since this electrolyte is much more complex than PBS, we were expecting to have a higher limit of detection, but even so, it is still lower than any previously reported LOD for dopamine detection in any electrolyte. Figure 50 is an exemplificative illustration of aptamer’s conformational change in aCSF 1x. 84 Figure 51. Exemplificative illustration of aptamer’s conformational change in aCSF 1x electrolyte solution. On the left, the aptamer is fully stretched in aCSF without dopamine and the graphene is mostly p-doped. On the right, dopamine was diluted in aCSF and the aptamer binds to it changing its conformation and coming close to graphene which becomes less p-doped due to divalent cation aptamer enhancement. A surprising observation was that the response of our sensors to dopamine in aCSF was different from the one observed in PBS. While PBS had a positive linear response with increasing dopamine concentration (p-doping), in aCSF we observed the exact opposite behavior, negative linear response (n-doping). And so, we tried to justify this difference considering the electrolyte’s compositions. One of the main differences between PBS and aCSF is that the latter contains divalent cations (Mg2+ and Ca2+) instead of monovalent cations (K+ and Na+). Divalent cations have bigger electrostatic interactions when compared to monovalent ions, which affect cooperative and competitive effects of counter-ions in solution [97]. DNA’s backbone has one possible binding site, a single bounded oxygen from the phosphate group, meaning that magnesium and calcium ions will bind to it, but since they are divalent, another possible single bound can be established. Mg+ and Ca+ can either attach to the next phosphate chain or remain available. Both elements have more atomic mass than sodium and potassium, meaning that not only they are heavier which can affect the aptamer’s conformational change but also, the electrical double layer composed of these counter-ions will stack on graphene’s EDL overshadowing the aptamer’s negative backbone. The EDL is then composed of positive charges meaning that the screening effect on graphene will produce negative charges. 85 This binding of divalent cations to the phosphate group of the aptamer’s backbone theory has been similarly explained by Korolev, Nikolay, et al. [94]. But the cations can also just be electrostatically imprisoned to the aptamer. Another possibility is that the presence of divalent cations changes the conformation of the aptamer even before it binds to dopamine, having for example, the aptamer close to the graphene before finding its target, producing different changes in surface charge rearrangement on the graphene channel. Such has been reported by Nakatsuka, Nako, et al. [96], showing that this dopamine aptamer exhibits different target binding affinities in presence of divalent cations since they influence the structure of the dopamine aptamer-target complex. Thus, they hypothesized that different aptamer conformational changes occur upon dopamine binding in the presence vs absence of the divalent cations. At 0.1 nM dopamine concentration surface saturation occurs at a Dirac point shift of -100 mV. The linear range of dopamine detection in aCSF is 140 mV, 60 mV more than detection in PBS 1x, surface saturation occurs two concentrations higher than PBS’. Like previously reported, higher receptor-target affinities and bigger response occur in the presence of divalent cations with this specific aptamer [95] [96]. This different detection response for the aCSF electrolyte solution allows us to better understand how our aptamer behaves on a more complex medium and confirms that it is possible do detect dopamine with great sensitivity (10 aM) in a very complex electrolyte, setting the bridge for dopamine detection in biological CSF. Detection of dopamine in biological CSF and in ex vivo brain samples can be found in the next chapter. 86 7. DOPAMINE DETECTION EX VIVO For ex vivo experiments, cerebrospinal fluid, blood and striatum homogenate were extracted from mice. The objective of these experiments was to assess our sensor’s selectivity and specificity in a very complex medium that is a real brain, in this case, a mouse brain. Since many different molecules can be found in the brain, to guarantee that we were in fact measuring dopamine, we used a mice model where dopamine could be acutely depleted from neuronal synapses. Such model was based in injections of a monoamine-depleting drug, reserpine, that irreversibly and non-selectively blocks the vesicular monoamine transporter [99]. Reserpine inhibits the activity of the vesicular monoamine transporter 1 and 2 (VMAT-1 and VMAT-2). Since VMAT-2 occurs in neurons, reserpine impairs dopamine uptake and storage in neuronal cells in the brain, including in the striatum brain region, leading to the depletion of dopamine in neuronal synapses [100]. First, we confirmed that reserpine could in fact lead to dopamine depletion in the brain by comparing cerebrospinal fluid (CSF) extracted from reserpine injected animals and controls. Then, to confirm that we could detect dopamine in a complex solution that closely mimics the extracellular space of the brain we used brain homogenates from dopamine-depleted animals to which dopamine was added. An additional preliminary experiment with dopamine detection in blood extracted from mice can also be found in Supplement 3. This chapter’s and previous chapter’s data was presented at the Portuguese Society for Neuroscience Meeting (“SPN2021”) conference (see abstract on Supplement 2). 7.1. Experimental procedures Animal experiments were carried out at the Life and Health Sciences Research Institute (ICVS), School of Medicine, University of Minho, and were conducted in accordance with European Union Directive 2016/63/EU and the Portuguese law on the protection of animals for scientific purposes (DL No 113/2013). This study was approved by the Portuguese National Authority for Animal Health (DGAV, 8519) and the Ethics Subcommittee for Life and Health Sciences of University of Minho (SECVS, 01/18). All procedures in animals were performed by researchers accredited by FELASA and DGAV. 87 Mice (n=5) were injected with reserpine (5 mg/Kg, i.p.) dissolved in 1% glacial acetic acid and diluted in 0.9% saline. Control mice (n=5) were injected with the vehicle (1% glacial acetic acid diluted in 0.9% saline). 8h post-administration, reserpine-injected mice displayed severe akinesia, postural instability, and tremors while control mice displayed normal behavior. Akinesia and postural instability are an indication of dopamine depletion since one major dopamine pathway, the nigrostriatal pathway, is involved in motor control. Mice were then anesthetized with avertin (tribromoethanol; 20 mg/mL, i.p.) placed in a stereotaxic frame for head fixation and CSF was extracted by cisterna magna puncture. Extracted CSF samples were immediately frozen in liquid nitrogen and stored at -80ºC. Mice were then perfused transcardially with 0.9% saline to remove all circulating blood and the brain was quickly removed. The striatum, on both hemispheres, was dissected, frozen in liquid nitrogen and stored at -80ºC. Before use, brain samples were thawed, and two samples were combined in a single Eppendorf tube with aCSF added at 5 µL/mg. Samples were then centrifuged at 14000 rpm, for 15 mins at 4ºC. 7.2. Dopamine detection in CSF CSF samples from 4 reserpine-injected and 4 control mice were thawed to room temperature before use. Each sample had an approximate volume of 2 µL and was incubated for 10 minutes. CSF from control mice was incubated for every chip for the same time interval as reserpine-injected CSF. Pairs of reserpine/control animals’ samples were incubated on the same chips under the same conditions so the only difference between the measurements would be the lack of monoamines, such as dopamine. The chips baseline was acquired with artificial cerebrospinal fluid incubation and produced a Dirac point shift similar to our chips’ natural error, around 5 mV, implying that there is barely any measurement difference between PBS 1x and aCSF 1x prior to aptamer-dopamine binding. 94 Figure 55. Graphene transfer curve for dopamine detection in striatum homogenate with measurements on PBS 1x. Figure 56. Graphene transfer curve for dopamine detection in striatum homogenate with measurements on striatum homogenate without dopamine. 95 7.3.1 Striatum homogenate control experiments To validate that we were detecting dopamine and not other non-target molecules present in the brain tissue or that our measurements were being affected by the process itself, we performed two additional control experiments. To control for the measurement and cleaning processes as well as for the incubation with multiple media (striatum homogenate, PBS 1x and striatum homogenate again) in a sequential fashion, we replicated the protocol described above but incubated striatum homogenate samples from a reserpine-treated animal without dopamine added only. Fresh 10 µL samples of striatum homogenate were incubated for 20 minutes and measured as described for the experiments above. After the measurements, a new sample of striatum homogenate was incubated again. This process was repeated 4 times. Such measurements can be found on figure 57, where we see measurements stabilization after 40 minutes. The three types of measurements have similar behavior overtime but symmetric to the baseline, meaning that there was no dopamine detected, otherwise all measurements would have the same negative linear shift trend. Figure 57. Overtime measurements of reserpine-injected striatum homogenate. In a second control experiment, we assessed the effect of measuring the same striatum homogenate sample over a long period of time in order to determine how long dopamine would take to bind to the aptamers. A 10 µL striatum homogenate sample from a control animal, 96 obtained as described for reserpine-treated animals, was incubated and measured on the same chip for 210 minutes in a humidity-controlled chamber to avoid evaporation and precipitation effects. This experiment’s data can be found in figure 58. At 30 minute mark, which corresponds to the first measurement taken, we observed a -30 mV shift of the Dirac point in respect to the baseline. At the 60 minute mark, an additional shift of -10 mV (-40 mV in respect to the baseline) was observed. The Dirac point then remained unchanged for the following 150 minutes. Meaning that there is no more dopamine to bind to the aptamers and that the linear range of detection for a control striatum homogenate is 40 mV. Based on these two experiments we conclude that dopamine binding to the aptamers was not time-dependent nor medium dependent (there is no continuous molecule deposition overtime). Figure 58. Overtime measurements of control striatum homogenate. Dopamine detection with our sensors proved to be possible in a very complex medium that is the brain, obtaining a response of +45 mV in cerebrospinal fluid and detection on 0.1 fM for striatum homogenate. These promising results allow us to consider detecting dopamine with our chips in an in vivo setting. 97 8. CONCLUSION Detecting neurotransmitters for better understanding of how the human brain works is of upmost importance, and in the past few years, recent advances in this field were facilitated with the help of novel biosensors. To current date, dopamine, a neurotransmitter with many important roles in the central nervous system, has been hardly detected below femtomolar concentrations. Not only some methods oxidize dopamine, but also use labelling agents, making detection impossible in a in vivo scenario. These and many other drawbacks of dopamine detection methods have been hindering the understanding of how dopamine works inside the brain and therefore, dopamine-related diseases. In this work, we combined the excellent properties of a 2D material, graphene, with a small, robust, and highly adaptive probe, a DNA aptamer. The biggest benefit that graphene introduces to the biosensing field, is the ability to reach limits of detection never reported before due to its one atom thickness. Thus, sparce events occurring on functionalized graphene can be felt and measured for low voltages and currents. Accordingly, label free highly sensitive and highly selective dopamine detection with electrolyte-gated graphene field-effect transistors was demonstrated in this work. It was the first time that our transistors had been functionalized with aptamers and our passivation method success was demonstrated based on the Dirac point voltage and x-ray photoelectron spectroscopy experiments. The detection method based on local gating of the electrical double layers allowed several electrolyte solutions to be studied. In phosphate buffered saline electrolyte, which is the common buffer solution used in biological experiments, we obtained a positive Dirac point shift for all dopamine concentrations and the limit of detection, 1 attomolar. This is the lowest LOD ever reported for dopamine detection. The linear range of detection was 40 mV and surface saturation occurred at 1 pM. Then, we moved on to a more complex electrolyte solution that could closely mimic the brain’s medium, the artificial cerebrospinal fluid. In aCSF we presented a divalent cation aptamer dependence that altered and improved dopamine detection, increasing the transistors’ linear range of detection from 40 mV to 140 mV and surface saturation occurred two concentrations higher, at 0.1 nM. The limit of detection was one concentration higher than in PBS, at 10 aM. 98 The selectivity or our sensors was also tested. The aptamer bonded specifically to dopamine only, giving barely any response to dopamine analogues and interfering agents like LDopa, L-Tyrosine, Homovanillic acid and Ascorbic acid. With this state-of-the-art contender limit of detection for in vitro experiments, we transitioned to more complex electrolytes. Dopamine detection was thus studied in ex vivo setting, where cerebrospinal fluid and striatum homogenate from mice were extracted and measured in our chips. In order to prove that our sensors’ Dirac point shift was caused by dopamine only, we compared the measurements from control mice to reserpine-injected mice, the later an acute animal model of dopamine depletion. In cerebrospinal fluid from control mice, our sensors gave a distinct response of around 45 mV more when compared to reserpine-injected mice. In striatum homogenate from dopamine depleted animals spiked with dopamine, the transistors gave a linear response for various dopamine concentrations with a linear range of 70 mV and the limit of 0.1 fM. The linear range for PBS 1x was 40 mV, for aCSF was 140 mV and for striatum homogenate was 70 mV. Our aptamer conformational change is enhanced by divalent cations, which explains why the response is bigger in aCSF and striatum homogenate than on PBS 1x (which has only monovalent cations), despite having slightly lower limits of detection. In conclusion, our graphene functionalized transistors are positioned in a state-of-the-art dopamine detection level and can be thoroughly improved with different graphene passivation methods to avoid non-specific binding effects, increase limits of detection and make it possible to progress into in vivo experiments. A possible future direction would be implementation of our graphene sensors in brains for better understanding of dopamine release through synaptic neurons and its acting mechanism inside the blood-brain-barrier. To do so, flexible graphene probes must be developed. 99 9. REFERENCES [1] Niyonambaza, Shimwe Dominique, et al. "A review of neurotransmitters sensing methods for neuro-engineering research." Applied Sciences 9.21 (2019): 4719. [2] Perry, Maura, Qiang Li, and Robert T. Kennedy. "Review of recent advances in analytical techniques for the determination of neurotransmitters." Analytica chimica acta 653.1 (2009): 122. [3] Hampel, Lena, and Thorsten Lau. "Neurobiological Principles: Neurotransmitters." NeuroPsychopharmacotherapy (2020): 1-21. [4] Hugo Cifuentes Castro, Victor, et al. "An update of the classical and novel methods used for measuring fast neurotransmitters during normal and brain altered function." Current neuropharmacology 12.6 (2014): 490-508. [5] Eisenhofer, G. "Kopin IJ, Goldstein DS." Catecholamine metabolism: a contemporary view with implications for physiology and medicine. Pharmacol Rev 56 (2004): 331-349. [6] Cosnier, Serge, et al. "An electrochemical method for making enzyme microsensors. Application to the detection of dopamine and glutamate." Analytical chemistry 69.5 (1997): 968-971. [7] Raghu, P., et al. "A novel horseradish peroxidase biosensor towards the detection of dopamine: A voltammetric study." Enzyme and microbial technology 57 (2014): 8-15. [8] Shervedani, Reza Karimi, and Akbar Amini. "Direct electrochemistry of dopamine on gold— Agaricus bisporus laccase enzyme electrode: Characterization and quantitative detection." Bioelectrochemistry 84 (2012): 25-31. [9] Njagi, John, et al. "Amperometric detection of dopamine in vivo with an enzyme based carbon fiber microbiosensor." Analytical chemistry 82.3 (2010): 989-996. [10] Liu, Xixia, and Juewen Liu. "Biosensors and sensors for dopamine detection." View 2.1 (2021): 20200102. [11] Choi, Jee-Hyun, et al. "Localized surface plasmon resonance-based label-free biosensor for highly sensitive detection of dopamine." Journal of nanoscience and nanotechnology 14.8 (2014): 5658-5661. [12] Alkhamis, Obtin, et al. "Innovative engineering and sensing strategies for aptamer-based smallmolecule detection." TrAC Trends in Analytical Chemistry 121 (2019): 115699. [13] Mannironi, Cecilia, et al. "In vitro selection of dopamine RNA ligands." Biochemistry 36.32 (1997): 9726-9734. 100 [14] Kim, Eunhye, and Insook Rhee Paeng. "Advantageous sensitivity in the DNA homolog of the RNA dopamine aptamer." Journal of Immunoassay and Immunochemistry 35.1 (2014): 83-100. [15] Nakatsuka, Nako, et al. "Aptamer–field-effect transistors overcome Debye length limitations for small-molecule sensing." Science 362.6412 (2018): 319-324. [16] Zheng, Yu, Yong Wang, and Xiurong Yang. "Aptamer-based colorimetric biosensing of dopamine using unmodified gold nanoparticles." Sensors and Actuators B: Chemical 156.1 (2011): 95-99. [17] Li, Bor-Ran, et al. "An ultrasensitive nanowire-transistor biosensor for detecting dopamine release from living PC12 cells under hypoxic stimulation." Journal of the American Chemical Society 135.43 (2013): 16034-16037. [18] Farjami, Elaheh, et al. "RNA aptamer-based electrochemical biosensor for selective and labelfree analysis of dopamine." Analytical chemistry 85.1 (2013): 121-128. [19] Napier, T. Celeste, Alana Kirby, and Amanda L. Persons. "The role of dopamine pharmacotherapy and addiction-like behaviors in Parkinson’s disease." Progress in NeuroPsychopharmacology and Biological Psychiatry 102 (2020): 109942. [20] Kaminski, Jakob, et al. "Association of cortical glutamate and working memory activation in patients with schizophrenia: A multimodal proton magnetic resonance spectroscopy and functional magnetic resonance imaging study." Biological psychiatry 87.3 (2020): 225-233. [21] Liao, Yang, et al. "Astrocytes in depression and Alzheimer’s disease." Frontiers of Medicine (2021): 1-13. [22] Bucher, Elizabeth S., and R. Mark Wightman. "Electrochemical analysis of neurotransmitters." Annual review of analytical chemistry 8 (2015): 239-261. [23] Song, Shiping, et al. "Aptamer-based biosensors." TrAC Trends in Analytical Chemistry 27.2 (2008): 108-117. [24] Kaur, Harmanjit, and Munish Shorie. "Nanomaterial based aptasensors for clinical and environmental diagnostic applications." Nanoscale Advances 1.6 (2019): 2123-2138. [25] Zhu, Qinchang, Ge Liu, and Masaaki Kai. "DNA aptamers in the diagnosis and treatment of human diseases." Molecules 20.12 (2015): 20979-20997. [26] White, Rebekah R., Bruce A. Sullenger, and Christopher P. Rusconi. "Developing aptamers into therapeutics." The Journal of clinical investigation 106.8 (2000): 929-934. [27] Post, Michael R., and David Sulzer. "The chemical tools for imaging dopamine release." Cell Chemical Biology (2021). 101 [28] Liang, Yuanying, et al. "Label-free split aptamer sensor for femtomolar detection of dopamine by means of flexible organic electrochemical transistors." Materials 13.11 (2020): 2577. [29] Miller, S. L., and H. H. Yeh. "Neurotransmitters and Neurotransmission in the Developing and Adult Nervous System." Conn's Translational Neuroscience (2017): 49-84. [30] Suresh, S., et al. "An Empirical Analysis of Leaky Integrate and Fire Neuron Model." [31] Cuevas, Javier. "Neurotransmitters and their Life Cycle." Elsevier (2019). [32] Hornykiewicz, Oleh. "Dopamine (3-hydroxytyramine) and brain function." Pharmacological reviews 18.2 (1966): 925-964. [33] Volkow, Nora D., et al. "Evaluating dopamine reward pathway in ADHD: clinical implications." Jama 302.10 (2009): 1084-1091. [34] Laatikainen, Linda M., et al. "Sexually dimorphic effects of catechol-O-methyltransferase (COMT) inhibition on dopamine metabolism in multiple brain regions." PLoS One 8.4 (2013): e61839. [35] Leenders, K. L., et al. "Brain dopamine metabolism in patients with Parkinson's disease measured with positron emission tomography." Journal of Neurology, Neurosurgery & Psychiatry 49.8 (1986): 853-860. [36] Sadek, Paul Charles, and Paul C. Sadek. “The HPLC solvent guide”. New York: Wiley, 1996. [37] Rajawat, Jyotika, and Gagandeep Jhingan. "Mass spectroscopy." Data Processing Handbook for Complex Biological Data Sources. Academic Press, 2019. 1-20. [38] Dunn, Matthew R., Randi M. Jimenez, and John C. Chaput. "Analysis of aptamer discovery and technology." Nature Reviews Chemistry 1.10 (2017): 1-16. [39] Cho, Il-Hoon, et al. "Chemiluminometric enzyme-linked immunosorbent assays (ELISA)-on-achip biosensor based on cross-flow chromatography." Analytica chimica acta 632.2 (2009): 247255. [40] Stojanovic, Milan N., Paloma de Prada, and Donald W. Landry. "Fluorescent sensors based on aptamer self-assembly." Journal of the American Chemical Society 122.46 (2000): 1154711548. [41] Cooper, Daniel R., et al. "Experimental review of graphene." International Scholarly Research Notices 2012 (2012). [42] Majee, Subimal, et al. "Scalable inkjet printing of shear-exfoliated graphene transparent conductive films." Carbon 102 (2016): 51-57. 102 [43] Amiri, Ahmad, et al. "A review on liquid-phase exfoliation for scalable production of pure graphene, wrinkled, crumpled and functionalized graphene and challenges." FlatChem 8 (2018): 40-71. [44] Cohen-Karni, Tzahi, et al. "Graphene and nanowire transistors for cellular interfaces and electrical recording." Nano letters 10.3 (2010): 1098-1102. [45] Zhan, Beibei, et al. "Graphene field‐effect transistor and its application for electronic sensing." Small 10.20 (2014): 4042-4065. [46] Weiss, Nathan O., et al. "Graphene: an emerging electronic material." Advanced materials 24.43 (2012): 5782-5825. [47] Rezaei, Behzad, Malihe Khalili Boroujeni, and Ali A. Ensafi. "Fabrication of DNA, ophenylenediamine, and gold nanoparticle bioimprinted polymer electrochemical sensor for the determination of dopamine." Biosensors and Bioelectronics 66 (2015): 490-496. [48] Stine, Rory, et al. "Fabrication, optimization, and use of graphene field effect sensors." Analytical chemistry 85.2 (2013): 509-521. [49] Zhu, Mingfang, Changqing Zeng, and Jianshan Ye. "Graphene‐Modified Carbon Fiber Microelectrode for the Detection of Dopamine in Mice Hippocampus Tissue." Electroanalysis 23.4 (2011): 907-914. [50] Jackowska, Krystyna, and Pawel Krysinski. "New trends in the electrochemical sensing of dopamine." Analytical and bioanalytical chemistry 405.11 (2013): 3753-3771. [51] Nandi, Rajib, et al. "Gold nanoparticle-mediated signal amplification of liquid crystal biosensors for dopamine." Analyst 144.4 (2019): 1110-1114. [52] Pumera, Martin. "Graphene in biosensing." Materials today 14.7-8 (2011): 308-315. [53] Chang, Haixin, et al. "Graphene fluorescence resonance energy transfer aptasensor for the thrombin detection." Analytical chemistry 82.6 (2010): 2341-2346. [54] Viswanathan, Sowmya, et al. "Graphene–protein field effect biosensors: glucose sensing." Materials Today 18.9 (2015): 513-522. [55] Dong, Xiaochen, et al. "Electrical detection of DNA hybridization with single‐base specificity using transistors based on CVD‐grown graphene sheets." Advanced Materials 22.14 (2010): 16491653. [56] Lu, Ye, et al. "Graphene-protein bioelectronic devices with wavelength-dependent photoresponse." Applied Physics Letters 100.3 (2012): 033110. 103 [57] Lerner, Mitchell B., et al. "Toward quantifying the electrostatic transduction mechanism in carbon nanotube molecular sensors." Journal of the American Chemical Society 134.35 (2012): 14318-14321. [58] Mohanty, Nihar, and Vikas Berry. "Graphene-based single-bacterium resolution biodevice and DNA transistor: interfacing graphene derivatives with nanoscale and microscale biocomponents." Nano letters 8.12 (2008): 4469-4476. [59] Guo, Shi-Rui, et al. "Label free DNA detection using large area graphene based field effect transistor biosensors." Journal of nanoscience and nanotechnology 11.6 (2011): 5258-5263. [60] Ohno, Yasuhide, Kenzo Maehashi, and Kazuhiko Matsumoto. "Label-free biosensors based on aptamer-modified graphene field-effect transistors." Journal of the American Chemical Society 132.51 (2010): 18012-18013. [61] Zhu, Limei, Liqiang Luo, and Zhenxin Wang. "DNA electrochemical biosensor based on thionine-graphene nanocomposite." Biosensors and Bioelectronics 35.1 (2012): 507-511. [62] Campos, Rui, et al. "Attomolar label-free detection of DNA hybridization with electrolyte-gated graphene field-effect transistors." ACS sensors 4.2 (2019): 286-293. [63] Her, Jim-Long, et al. "Label-free detection of alanine aminotransferase using a graphene fieldeffect biosensor." Sensors and Actuators B: Chemical 182 (2013): 396-400. [64] Huang, Yinxi, et al. "Graphene-based biosensors for detection of bacteria and their metabolic activities." Journal of Materials Chemistry 21.33 (2011): 12358-12362. [65] Ang, Priscilla Kailian, et al. "Flow sensing of single cell by graphene transistor in a microfluidic channel." Nano letters 11.12 (2011): 5240-5246. [66] Park, Seon Joo, et al. "Ultrasensitive flexible graphene based field-effect transistor (FET)-type bioelectronic nose." Nano letters 12.10 (2012): 5082-5090. [67] Nair, Rahul Raveendran, et al. "Fine structure constant defines visual transparency of graphene." Science 320.5881 (2008): 1308-1308. [68] Thingujam, Terirama, et al. "TCAD based modeling and simulation of Graphene Nanostructured FET (GFET) for High Frequency Performance." ADBU Journal of Engineering Technology 6.1 (2017). [69] Soldano, Caterina, Ather Mahmood, and Erik Dujardin. "Production, properties and potential of graphene." Carbon 48.8 (2010): 2127-2150. [70] Joyce, W. B., and R. W. Dixon. "Analytic approximations for the Fermi energy of an ideal Fermi gas." Applied Physics Letters 31.5 (1977): 354-356. 110 depleted animals down to attomolar level, and differences in dopamine concentration in CSF samples from control and dopamine-depleted animals. The measured detection limits are the lowest ever reported for dopamine both in vitro and in biological samples which allows us to consider further experiments in an in vivo setting. 111 SUPPLEMENT 3 – DOPAMINE DETECTION IN BLOOD A substantial amount of dopamine circulates in the bloodstream, but in humans, over 95% of the dopamine in the plasma is in the form of dopamine sulphate. Since dopamine does not cross the blood-brain barrier, its synthesis and functions in peripheral areas are independent of its synthesis and functions in the brain. Therefore, dopamine detection outside the brain is not of upmost interest. Blood samples from mice were measured in three different chips. On the first chip, chip 694, we incubated a blood centrifuged sample from a reserpine-injected animal and measured a +25 mV Dirac point shift. Figure A. Blood’s dopamine detection for chip 694. On the second chip, number 762, blood from a control animal gave a +30 mV Dirac point shift. 112 Figure B. Blood’s dopamine detection for chip 762. And for the last chip, we incubated two different blood samples, a reserpine-induced animal and control animal, giving +37 mV and +55 mV Dirac point shift respectively. Figure C. Blood’s dopamine detection for chip 775. Since the chip with reserpine sample and the chip with control sample gave similar response and the difference between both steps on the same chip lacked a considerable distinction. We think that non-specific binds might be happening, therefore our sensors don’t have a functionalization method for dopamine detection appropriate for a blood electrolyte.