Microfluidic study of plasticity in pancreatic beta cell heterogeneity

MICROFLUIDIC STUDY OF PLASTICITY IN PANCREATIC β -CELL HETEROGENEITY TAN CHERNG-WEN, DARREN (B.Sc (Hons.), NUS) A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY GRADUATE PROGRAMME IN BIOEGNINEERING YONG LOO LIN SCHOOL OF MEDICINE NATIONAL UNIVERSITY OF SINGAPORE 2009 ACKNOWLEDGMENTS The author wishes to thank A/P Partha Roy for his invaluable guidance and generous support in the course of this work The author also wishes to express deep appreciation for the technical guidance provided by A/P Michael Raghunath and A/P Dieter Trau Appreciation is also due A/P Hanry Yu, A/P Lanry Yung and A/P Zhang Yong for their helpful insights and suggestions during the Qualifying Examination Critical technical assistance was provided by Mr Shashi Ranjan, Mr Ganesh Balasubramanian, Mr Vedula Sri Ram Krishna and Mr Tok Wee Lee, without which many of the obstacles in the course of this work would never have been surmounted Finally, the author wishes to express infinite gratitude to his wife, who shared these five years of long days and endless nights, whose patience, tolerance, encouragement and unflagging support provided him with the fuel to persevere and see the task to its end TABLE OF CONTENTS Table of Contents ii SUMMARY vi LIST OF TABLES ix LIST OF FIGURES x ABBREVIATIONS .xiv PART I Chapter Introduction and Literature Survey 1.1 Introduction 1.2 The islets of Langerhans 1.3 Insulin secretion and glucose homeostasis 1.4 Glucose sensing and activation of -cell function 1.5 Effect of glucose on first phase insulin secretion 1.6 Effect of glucose on second phase insulin secretion 1.7 -cell heterogeneity 10 1.8 The hypothesis 12 1.9 Three questions raised to address the hypothesis .13 1.10 Microfluidic channel cell culture 15 1.11 Generating solute concentration gradients in microfluidic channels 16 1.12 Organisation of the thesis 19 Chapter .21 Development of the hydrogel-assisted gradient generator (HAGG) .21 2.1 The proposed design 21 2.2 Microfluidic gradient generators 22 2.3 Overview of microfluidic system structure 28 2.4 Overview of fabrication process 30 ii 2.5 Master mold fabrication 30 2.6 Micromolding 31 2.7 Microfluidic channel surface treatment with TPM 32 2.8 Hydrogel formation 33 2.9 Developing and disinfecting 39 Chapter .40 Application of the HAGG .40 3.1 Gradient generation 40 3.2 Cell culture in the HAGG .48 3.3 Gradient generation in the presence of cells .55 3.4 Selecting assays to address the three questions 58 3.4.1 Metabolism of 2-NBDG 59 3.4.2 Glucose-induced quinacrine secretion 60 3.4.3 Ca2+-induced intracellular Fura-2/acetoxymethylester fluorescence 60 3.5 -TC-6 characterization 61 3.5.1 -TC-6 metablism of 2-NBDG 61 3.5.2 Glucose-induced quinacrine secretion in -TC-6 64 3.5.3 Fura-2/AM staining of -TC-6 67 3.5.4 -TC-6 glucose uptake rate 70 3.6 Addressing the three questions 70 3.6.1 Ascertaining mercury lamp intensity variation 79 3.6.2 Net GLUT2 transporter/glucokinase activity under an imposed extracellular glucose gradient 80 3.6.3 Intracellular vesicle density after exposure to an imposed extracellular glucose gradient 87 3.6.4 Intracellular vesicle density after exposure to an imposed extracellular glucose gradient and glucose challenge 91 Chapter .95 Understanding the results 95 4.1 Microfluidic channel design 95 4.2 Transient-diffusion model of mass transport in the HAGG 96 iii 4.3 Understanding gradient generation in the HAGG 100 4.3.1 Transient diffusion of 2-NBDG 100 4.3.2 Steady-state diffusion of 2-NBDG 102 4.3.3 Applying the model for experimental design 103 4.4 Cell culture in the HAGG 106 4.5 Gradient generation in presence of cells 110 4.6 -TC-6 characterization 113 4.7 Concerns over the use of UV excitation 118 4.8 Addressing the three questions 119 Chapter 124 Conclusion and future directions 124 PART II 128 Chapter 129 Microfluidic competition assay via equilibrium binding 129 6.1 Analyte detection 129 6.2 Competition assays 129 6.3 Microfluidic channel-based competition assays 130 6.4 Convection-diffusion model of microfluidic competition assay 130 6.5 Materials and Methods 134 6.6 Labeling of proteins with fluorescein 135 6.7 Conjugation of anti-insulin IgG with biotin 136 6.8 Microchannel fabrication 137 6.9 Microchannel surface blocking and IgG immobilization 137 6.10 Flow cell fabrication 137 6.11 Microchannel binding assays 138 6.12 Results 139 6.12.1 Comparison of elution profiles for aprotinin-FITC and insulin-FITC 139 6.12.2 Determination of insulin-FITC equilibrium dissociation constant, K d,L 140 6.12.3 Determination of unlabeled insulin equilibrium dissociation constant, K d , I 142  iv  6.13 Discussion 146 6.14 Conclusion 152 Literature cited 153 Appendix – Fluorescein labeling kit 160 Appendix – Chemical structure of reagents used 161 Appendix – INSTECH P625/10K.143 peristaltic pump 162 Appendix – Lists of P-values for the 2-sample t-tests 163 Appendix – Abstracts of work presented at various conferences 164 v SUMMARY This dissertation is divided into two parts - a major part and a minor part In the major part comprising five chapters, a microfluidic system is developed and applied to the study of pancreatic -cells Immunohistochemical studies of insulin and glucokinase in the rat islets of Langerhans have revealed functional heterogeneity among the -cell population along with reversibility Anatomical studies of islets have shown them to be more vascular and more permeable than exocrine tissue Islets receive more than five times the blood flow in relation to their volume and fill first with infusion The rodent islet has a non-random cell distribution with a core of -cells and a mantle of -, - and PP cells The microvasculature of the islets confers a directionality of blood flow outwards from the core, and establishes gradients of nutrients In view of these findings, a hypothesis is proposed that the establishment of spatial gradients of extracellular glucose and promotion of -cell interactions are adequate for development of glucose concentrationmapped functional heterogeneity in -cell populations Three main questions are posed within the context of this hypothesis and addressed with the help of two specific assays The two assays are tested by designing an in vitro microfluidic system that allows the establishment and control of glucose gradients over the micro-scale length of islets An immortalized insulin secreting cell line derived from transgenic mice, -TC-6 is applied in this test The gradient generating section of the microfluidic system comprises a source microchannel and a sink microchannel aligned in parallel These are connected by 24 perpendicular cross-channels, and converge at the end of the series of cross-channels to form a single eluent channel The main and cross-channels are about 105 m and 310 m in width respectively Each cross-channel is about 317 m in length All the microchannels are 50 m in height The final product includes poly(ethylene glycol) diacrylate hydrogel barriers at either end of each cross-channel Each hydrogel has a vi height of 50 m and a width of 82 m This microsystem is designated the hydrogelassisted gradient generator or HAGG The fluorophore Alexa Fluor 488 and a fluorescent glucose analog (2-NBDG) are used as probes to illustrate the generation of stable, reproducible and linear, probe concentration gradients in the absence of cells A method is developed for estimating the diffusivity and hydrogel permeability of a solute from in situ imaging data Concentration gradients are also generated in the presence of TC-6 to demonstrate the compatibility of the system for our study The three questions are concerned with the establishment of microscale glucose gradients over a population of -TC6 cells to determine the resultant mapping of (a) 2-NBDG accumulation, (b) insulin storage, and (c) insulin secretion The two specific assays test for -cell response by probing for net GLUT-2/glucokinase activity that leads to intracellular 2-NBDG fluorescence, and intracellular vesicle density revealed by quinacrine fluorescence Our results show that (a) net GLUT-2/glucokinase activity is heterogeneous and does not map to extracellular glucose gradient, (b) quinacrine uptake is homogeneous with no mapping of insulin storage to glucose gradient, and (c) insulin secretion is not influenced by the imposed glucose gradient In the second part of this dissertation, a microfluidic competition assay valid for the case of equilibrium binding between a receptor and competing ligands is developed A mathematical model describes the transient, convection-dispersion of solutes, undergoing equilibrium binding to immobilized receptors, while entrained in a low Reynolds number incompressible fluid flowing through a microchannel The proposed method involves monitoring the elution profile of a reference molecule and ligand in the presence of a competitor The time difference between the two breakthrough curves provides a measure of the unknown concentration of the competitor Theoretical results illustrate the general method for determining the equilibrium dissociation constant ( K d ) of the ligand and competitor, as well as the competitor concentration Experimental data is vii presented for the binding of fluorescein-labeled insulin and unlabeled insulin to a monoclonal antibody It is found that the unlabeled insulin binds with higher affinity ( K d = 0.17 M) than the labeled insulin ( K d = 0.76 M) The potential advantages of the method and further improvements in the model are discussed viii LIST OF TABLES Number Page Table Program for spin-coating of PDMS onto hydrogel barrier photomask 34 Table Program for spin-coating second coat of PDMS onto hydrogel barrier photomask 35 Table Summary and rationale for each step of the experimental protocol for studying the three questions 71 Table Concentrations of glucose used for the control and test samples in the HAGG assays 76 Table Table showing the molecular weight of various probes and reagents 108 Table Experimentally measured values of equilibriumm constants Kd,L and Kd,I and other related data 146 ix In this paper, we present measurements of dispersion coefficients, in PDMS microchannels, for three model proteins with large differences in molecular weight and size The surface interactions are minimized in our experiments by employing a suitable set of blocking molecules The next section discusses experimental methodology and the fiber optic fluorescence sensor The results and discussion section provides a comparison between the theoretical and experimental values of dispersivity In concluding, we make some observations about sources of error for our experimental system and the applicability of these results in assessing the performance of PDMS devices EXPERIMENTAL MATERIALS AND METHODS Protein labeling Insulin (Sigma-Aldrich, USA), bovine serum albumin, BSA (Sigma-Aldrich, USA) and fibrinogen (Fluka, Buchs, Switzerland) were conjugated to FITC using a standard labeling kit from Molecular Probes (6SFX kit, Eugene, OR) Concentration of each protein was measured using the bicinchoninic acid assay (BCATM Protein Assay kit, Pierce) Phosphatebuffered saline with azide (PBSA) was used as the washing buffer Protein solutions were prepared in PBSA, polyethylene glycol, and Tween 20 Avidin at μM was used to coat the microchannels Microchannel fabrication Polydimethylsiloxane base and curing agent (Sylgard 184, silicone elastomer kit, Dow Corning, Midland, MI), at a ratio of 10 parts of PDMS base to one part of curing agent, were micromolded against a template to form microchannels with a width (W) of 240 μm, a height (H) of 30 μm and a length (L) of 3.5 cm A mm diameter perforation of the PDMS was made at one end of each microchannel to form the outlets An additional layer of PDMS was prepared and both pieces of PDMS were exposed to reactive air plasma (Harrick Scientific, New York) for minutes before being bonded together to seal the microchannels A mm diameter perforation was then made through both layers of PDMS to form the inlets Fiber optic fluorescence probe The fluorescence sensor was assembled from individual components purchased separately As shown in Figure 1(a), the fiber optic light source is a blue LED (Ocean Optics, USA) or a tungsten lamp coupled to a digital monochromator (Optometrics, USA) The emitted light is guided via a bifurcated fiber into the emission monochromator and photomultiplier tube (Hamamatsu Corp, Japan) Light intensity data as a function of time is recorded by PC software Experimental setup A perforation about mm wide was made cm from one end of a poly(ether ether ketone) (PEEK) tube The nearest end of the PEEK tube was then sealed using acrylate glue The perforated PEEK tubing was then inserted into the inlet until the perforation was in line with the microchannel The other end of the PEEK tube led to the syringe pump via suitable adapters The T-shaped flow cell was secured to the outlet of the microchannel by inserting one end of a PEEK tube into the outlet while the fiber probe was placed in contact with the outer surface of the PEEK tube Experimental protocol Following assembly, coating buffer was perfused into each microchannel at μL/min for one hour and then left in the microchannels for a further one hour at room temperature, before being removed by perfusion with PBSA at 10 μL/min for hr The microchannels were further washed by perfusion with PBSA at maximal velocity, during which the fluorescence signal detected at the outlet was used as the baseline for the subsequent measurements of fluorescence intensity Thereafter, 300 nM insulin, 300 nM BSA or 100 nM fibrinogen was perfused into the microchannels at the flow velocities shown in Table The increase in fluorescence signal from the baseline at the outlet of the microchannels was monitored until no further increase was observed The protein solution was then replaced with PBSA and perfusion recommenced at the same flow rate The resultant reduction in fluorescence signal was monitored until no further decrease in fluorescence signal was observed RESULTS AND DISCUSSION Our experiments were designed to establish low Reynolds number (< 1) laminar flow of an aqueous buffer in a microchannel of rectangular cross-section All data reported here, was obtained in microchannels of width to height ratio of The main advantage of using a high aspect ratio is that the velocity profile is nearly plug flow along the width and parabolic along the height, as demonstrated by theoretical and experimental evidence Thus convective dispersion occurs primarily due to non-uniform velocity along the height of the channel and the theoretical Taylor-Aris result given by Brenner and co-workers may be applied [5], ⎡ (Pe)2 ⎤ Dl = Dm ⎢1 + ⎥ ⎣ 105 ⎦ (1) The Péclet number (Pe) is defined as UH/2Dm The Taylor-Aris result for the long-time asymptotic value of the dispersion coefficient is applicable when the average channel velocity has an upper limit as given in Table A transient convection dispersion model was solved by the finite difference method and the results matched the values presented by Brenner [8] Figure 1(b) depicts the typical elution profile obtained at the microchannel outlet when the proteinfree buffer was used to displace protein solution from the microchannel This data was fitted to the computational result via a statistical nonlinear fitting procedure to yield the experimental values of Dl presented in Table The dispersion coefficient as predicted by Eqn is also given in Table and estimated from the experimental velocity (U) and published molecular diffusivities (Dm) The experimental dispersion coefficient for insulin is larger than the theoretical value by a factor of 2.3, while that for BSA is larger by a factor of Fibrinogen exhibits the lowest difference of less than a factor of In all cases, the theoretical value is found to be less than the experimental value This difference may be partially explained by (1) noting that the velocity in Table somewhat exceeds the limiting value, (2) some additional dispersion occurs along the width of the channel and is neglected in Eqn 1, and (3) other experimental details discussed below 1.2 Fibrinogen Insulin BSA 1.0 Flow cell Light source Monochromator PMT Detection system 0.8 C/C0 Monochromator 0.6 0.4 0.2 0.0 Eluent Readout system PC 0.0 0.5 1.0 1.5 2.0 dimensionless time (a) (b) Figure (a) Experimental setup (b) Concentration versus time data at the microchannel outlet for the three proteins listed in Table (one data set each) Concentration is normalized with initial protein concentration Time is normalized by residence time Elution profiles were also obtained for the case of protein displacing buffer in the microchannel However, the results showed substantial deviation from the expected elution curves predicted by the computational model (data not shown) This could arise from protein adsorption, although Table Experimental and theoretical (Eqn 1) values of the dispersion coefficient and other parameters for the three proteins studied Experimental values are averages of three data sets each Protein Diffusivity, Dm cm2/s DmL/H2 cm/s Insulin MW 5800 BSA MW 66000 Fibrinogen MW 330000 12.9x10-7 0.50 6.3x10-7 0.25 2.0x10-7 0.08 Dl (Expt) cm2/s 0.0368± 0.002 0.0112± 0.001 0.0056± 0.0002 Dl (Eqn.1) cm2/s 0.016 U cm/s Pe 0.694 807 0.0036 0.231 550 0.0029 0.116 870 fluorescence microscopy did not reveal any fluorescence in the microchannel, thereby reaffirming our blocking strategy The main limitation of our current method is the fairly large outlet tubing volume (and residence time) that rearranges the boundary layer and velocity profiles and may contribute to increased dispersion This may be corrected by direct microscopic imaging of the microchannel near the outlet CONCLUSION Dispersivity of three proteins undergoing pressure-driven flow in PDMS microchannels is reported here Our method for preventing protein adsorption was more successful during protein displacement from the microchannel as compared to protein uptake This data will be useful in designing microfluidic devices ACKNOWLEDGEMENT This study was funded by NUS FRC grant no R-397-000-015-112/101 REFERENCES [1] Gunther A and Jensen K.F., "Multiphase microfluidics: from flow characteristics to chemical and materials synthesis", Lab On A Chip, Vol 6, No 12, (2006), pp 1487-1503 [2] Squires T.M and Quake S.R., "Microfluidics: Fluid physics at the nanoliter scale", Reviews Of Modern Physics, Vol 77, No 3, (2005), pp 977-1026 [3] Dutta D., Ramachandran A., and Leighton D.T., "Effect of channel geometry on solute dispersion in pressure-driven microfluidic systems", Microfluidics and Nanofluidics, Vol 2, No 4, (2006), pp 275-290 [4] Dorfman K.D and Brenner H., "Generalized Taylor-Aris dispersion in discrete spatially periodic networks: Microfluidic applications", Physical Review E, Vol 65, No 2, (2002) [5] Rush B.M., Dorfman K.D., Brenner H., and Kim S., "Dispersion by pressure-driven flow in serpentine microfluidic channels", Industrial & Engineering Chemistry Research, Vol 41, No 18, (2002), pp 4652-4662 [6] Bello M.S., Rezzonico R., and Righetti P.G., "Use Of Taylor-Aris dispersion for measurement of a solute diffusion-coefficient in thin capillaries", Science, Vol 266, No 5186, (1994), pp 773-776 [7] Belongia B.M and Baygents J.C., "Measurements on the diffusion coefficient of colloidal particles by Taylor-Aris dispersion", Journal of Colloid and Interface Science, Vol 195, No 1, (1997), pp 19-31 [8] Brenner H., "The Diffusion Model of Longitudinal Mixing in Beds of Finite Length Numerical Values", Chemical Engineering Science, Vol 17, (1962), pp 229-243 Microfluidic Channel Design For Continuous Concentration Gradients Across Static Fluid Domains C W Tan1, P Roy2 Graduate Program In Bioengineering, National University of Singapore Division of Bioengineering, National University of Singapore Abstract Myriad methods have been established for generating continuous solute concentration gradients in a microfluidic system These employ electrokinetic processes, wherein a voltage is applied to effect solute distribution, or methods of diffusive mixing wherein the diffusion of solutes across laminar flows of differing solute concentration is used to establish the gradients However, electrokinetic methods confine the use of such systems to charged solutes Furthermore, these techniques require laminar flow, and as such, the concentration gradients cannot be generated across a static region This makes these techniques inappropriate for test systems that employ uncharged particles, and in which flow across the test domain is undesirable To address this need, we propose to exploit diffusion across static fluid domains connecting two laminar streams of equal fluidic pressure, but differing solute concentrations, to establish these gradients instead This report describes the use of a microfluidic channel system comprising two parallel microchannels connected by regularly spaced cross-channels for setting up steady-state solute concentration gradients across the length of each cross-channel Keywords Concentration gradient, microfluidic, electrokinetic, diffusive mixing microchannel, laminar flow, List of Topics Biomedical Imaging Bioinformatics Biomaterials Biomechanics Biomolecular Engineering Biomedical Instrumentation Computational Bioengineering Biosignal Processing Nanobioengineerin g Micro Biomedical Engineering Systems Tissue Engineering Others Determination of Affinity Constant from Microfluidic Binding Assay D Tan1, P Roy2 Graduate Programme in Bioengineering, National University of Singapore, Singapore Division of Bioengineering, National University of Singapore, Singapore Abstract— We present a method of determining the affinity constant of a receptor-ligand pair using equilibrium binding in a microchannel This technique involves the immobilization of the receptor to the microchannel surface and perfusion of the ligand at a fixed flow rate The ligand concentration in the eluent is monitored in real time The break-through time of the ligand is then determined and compared with that of an appropriate, non-binding, reference molecule The difference in break-through time can then be used in a mathematical model we have developed to determine the affinity constant of the receptor-ligand pair Using this method, we have determined the equilibrium dissociation constant, KD, of a rat anti-insulin immunoglobulin and a fluorescein-conjugated human insulin Fluorescein-conjugated aprotinin was used as the nonbinding reference molecule Our experiments yielded a KD value of 0.76 μM for this anti-insulin IgG and insulin-FITC pair Keywords—Receptor-ligand, competition binding, microfluidic, equilibrium constant, transport model I INTRODUCTION Molecules in solution that undergo specific binding are described as receptor-ligand pairs Since these binding events not result in covalent bond formation, but rely rather upon the formation of hydrogen bonds, van der Waal’s forces as well as hydrophobic interactions, they are reversible and in the absence of perturbation and mass change, result in an equilibrium binding state [ ] This state comprises the free receptor, free ligand and the receptorligand complex co-existing at concentrations characteristic of the system [ ] These concentrations can be expressed as a single parameter – the equilibrium dissociation constant, KD – which is a useful means of characterizing the receptorligand system Consider a receptor, R, undergoing a binding event with the ligand, L: R+L k k-1 RL At equilibrium, the forward rate of reaction is equal to the reverse rate of reaction Defining the forward and reverse reaction rate constants as k and k-1 respectively, the equilibrium dissociation constant, k−1 (KD) can be expressed k as follows: KD = [R][L] [RL] (1) As such, the equilibrium dissociation constant of a system can be determined if the concentration of free ligand, free receptor and receptor-ligand complexes can be measured Conventional systems for determining KD include dialysis or solid-phase binding assays [ ] wherein the system is allowed to reach equilibrium, after which the free components are separated from the complexes Where dialysis allows the binding event to occur completely in solution, the solid-phase assays typically involve immobilisation of the receptor on a solid surface, such as the column resin used in affinity chromatography [ ] In the latter case, the ligand would have to diffuse from the bulk solution to the reaction surface, whereupon they would bind the immobilized receptors Subsequently, the concentration of either of the free components, or that of the complexes may be determined To facilitate this determination, either the receptor or the ligand might be labeled with an enzyme or a fluorescent probe Unlike enzyme probes, which require the addition and enzymatic modification of a substrate to yield a detectable product, a fluorescent probe allows direct measurement of concentration Mass conservation would then allow the concentration of the other components, and subsequently, the KD to be calculated However, these assays require the use of quantities of receptor or ligand which may be prohibitive in the case where either or both may not be readily available Adaptation of the solid-phase assay for 96-well plates reduces the necessary reaction volume to 100 μL per microwell and increases the surface area to volume ratio, so improving the efficacy of the system However, this improvement can be taken a step further by adapting the assay for a microfluidic system [ ] Microfluidic channels are capillary tubes, typically made of optically transparent material, with at least two of their dimensions in the micron range A microchannel of 35 μm (H) x 240 μm (W) x cm (L) would have a volume more than 297 times less than that in a microwell This considerably reduces the volume of reagent needed for each assay Typically, the receptor molecules are immobilized onto the surface of the microchannel and the ligand in solution is perfused through the microchannel, during which binding occurs Parameters such as flow rate and perfusion time can then be selected to ensure sufficient time for systems with fast-binding reaction kinetics to attain equilibrium Using a fuorescence-labeled ligand, the bound fraction can easily be quantitated by measuring the fluorescence intensity retained in the microchannels Alternatively, one could monitor the eluent fluorescence intensity and exploit changes in the breakthrough time to calculate KD In the absence of a receptor, the ligand resident time in the microchannel should be dictated only by the flow rate used The resultant elution profile is a dispersion-induced sigmoid In contrast, when the receptor is immobilized on the surface of the microchannel walls, binding of the ligand to the receptors increases the residence, so delaying the breakthrough time This delay is a function of the system KD value A model was developed to define the relationship between the shift in breakthrough time, ΔT50%, immoblised receptor surface concentration (RT), antigen concentration and KD Describe model Here, we report the use of a polydimethyl siloxane (PDMS) microchannel system to determine the KD of FITCconjugated recombinant human insulin and a monoclonal rat anti-insulin immunoglobulin (IgG) FITC-conjugated aprotinin was used as a reference II MATERIALS AND METHODS A Labeling of proteins with fluorescein Recombinant, human, insulin (Sigma-Aldrich, USA) and bovine lung aprotinin (Fluka Biochemika, Switzerland), each at 10 mg/mL were prepared in 0.1 M sodium bicarbonate buffer (pH 8.3) Fluorescein isothiocyanate (FITC) succinimidyl ester (Molecular Probes, USA) at 10 mg/mL in dimethylsulfoxide was then added to each protein solution at a FITC to protein molar ratio of 5:1 The reaction mixes were then incubated at room temperature for one hour with gentle agitation at 165 rpm on an orbital shaker Unreacted FITC was then removed from each reaction mix by ultrafiltration (Vivaspin 500, kDa MWCO, Sartorious, Germany) The protein solutions were then diluted to 500 µL using phosphate-buffered saline supplemented with 0.05% (w/v) sodium azide (PBSA) and the protein concentration of each solution was measured using the bicinchoninic acid assay (Pierce, USA) B Conjugation of anti-insulin IgG with biotin To immobilize the anti-insulin immunoglobulin (IgG) onto the avidin-coated microchannel surface, it is conjugated with the amine-reactive biotin ester, NHS-PEO-Biotin (Pierce, Rockford, IL) This biotin derivative was selected because it includes a poly(ethylene glycol) (PEO) linker of 29Å between the biotin moiety and the conjugation site [ ] This long and flexible spacer would extend the IgG further into the bulk flow, so minimizing the effects of steric hindrance on the antigen-antibody binding event Furthermore, the PEO segment would improve the solubility of the biotin succinimidyl ester in aqueous solutions, so allowing the IgG to be conjugated under conditions that would not jeopardize its structure and function To prepare the reaction mix, 92 μL of D6C4 anti-rodent insulin IgG (Hytest, Finland) at 1.1 mg/mL was added to μL of biotin succinimidyl ester at 0.5 mg/mL (biotin to protein molar ratio of 10:1) The mixture was then incubated at room temperature for one hour, with gentle agitation at 165 rpm on an orbital shaker The reaction mix was then diluted to 1, 000 µL with PBSA and the unreacted biotin was removed from the mix by ultrafiltration at 12, 000 rcf for 20 using two Vivaspin 500 ultrafiltration units (50 kDa MWCO, Vivascience, Sartorious, Germany) Each retentate was then diluted to 500 µL and filtered again This is repeated one more time to ensure complete removal of the unreacted biotin The final retentates were diluted to a total volume of 400 µL using PBSA The concentration of the biotinylated IgG was then determined using absorbance at 280 nm C Microchannel fabrication Microchannels with a width of 240 μm, a height of 35 μm and a length of 3.7 cm were micromolded from a template using Sylgard 184 (polydimethyl siloxane, PDMS, Dow Corning, USA) at a ratio of 10 parts of PDMS base to one part of cross-linker A one-millimeter diameter perforation of the PDMS at one end of each microchannel was made to form the outlets A mm layer of PDMS was separately prepared and both pieces of PDMS were exposed to reactive air plasma for minutes before being bonded together to seal the microchannels A mm diameter perforation was then made through both layers of PDMS to form the inlets A piece of PDMS was used to fabricate a flow cell with a PEEK tube from the outlet passing through it and placed in contact with a fiber optic cable via a T-junction D Microchannel surface blocking and IgG immobilization Immediately after sealing, avidin (Sigma-Alrich, USA) in PBSA at µM was perfused into each microchannel at μL/min for 60 The avidin solution was then left in the microchannels for a further one hour at room temperature The resident blocking buffer was then removed by perfusing the microchannel with PBSA at 10 μL/min for 6.5 hrs to ensure thorough removal of unadsorbed avidin Biotinylated IgG at 600 nM was then perfused into the microchannels at μL/min for 60 min, and left to incubate in the microchannels for a further 60 The microchannels were then reperfused with 600 nM biotinylated IgG at μL/min for hr, followed by a PBSA wash (1hr at 10 μL/min) to remove IgG that had not bound to the surface adsorbed avidin The value of RT was determined to be 10 pmoles/cm2 [ ] E Microchannel binding assay Excitation of the eluent in the flow cell was maintained at 490 nm from a LED source (Instech, USA) Fluorescence from the eluent was filtered at 520 nm using a monochromator (DongWoo Optron, Korea) and detected using a photon multiplier tube The data was then acquired by a PC using the counting board software (Hamamatsu, Japan) Each microchannel was perfused with wash buffer (polyethylene glycol in PBSA) at 0.20 mL/hr for 30 before being used for each assay The reference molecule was then perfused into the microchannel at 0.02 mL/hr and the increase in fluorescence intensity of the eluent was monitored until no further increase was observed The reference protein was then washed from the microchannel by perfusing wash buffer at 0.20 mL/hr and monitoring the decrease in fluorescence intensity until no further decrease was observed The sample solution was then perfused into the microchannel at 0.02 mL/hr and monitored as before, for the reference results presented here correspond to straight microchannels of rectangular cross section where the Taylor-Aris dispersion theory remains valid [ ] (U ≤ 0.39 cm/s) All experiments were designed to adhere to the conditions prescribed by the theoretical model in order to achieve the best possible agreement Aprotinin, with a molecular weight of Da, displays an almost identical elution profile to that of human insulin at 5, 800 Da, likely due to their similar diffusivities For this reason, aprotinin was used as the reference for elucidating the transport characteristics of the antibody-antigen pair, in the absence of receptor binding In the presence of surfaceimmobilised antibody, the model predicts a right-shift of the elution profile, indicating an increased residence time Each elution profile has the characteristic sigmoidal shape whose maximum slope is a function of the solute dispersion coefficient [ ] Since dispersion distorts the profile, so making determination of breakthrough time difficult, the time at which the eluent concentration is 50% of that of the inlet was used instead Any difference between the ligand and reference breakthrough times is then expressed as ΔT50%, which is the difference between the ligand and reference breakthrough times at 50% of the respective inlet concentrations Mathematical simulation of insulin and aprotinin transport phenomena in microchannels coated with anti-insulin IgG demonstrates this trend and also demonstrates how ΔT50% changes with L0/KD (constant RT) (Fig 2) It is seen that an increase in affinity constant increases ΔT50% because more binding takes place at the leading front, so depleting the bulk flow to a greater extent Kd,L 2Kd,L ΔT50% Another PEEK tube, at the inlet, was carefully perforated and aligned with the microchannel to eliminate impinging flow III RESULTS AND DISCUSSION The microfluidic competition assay works by entraining the ligand and competitor in low Reynolds number laminar flow through a microchannel whose surfaces contain the immobilized receptor The mathematical model presented above, incorporates the basic physics describing the microchannel transport and equilibrium binding of competing solutes Although the theory is generally applicable, the 0.001 0.01 0.1 10 100 1000 L0/Kd,L Fig Binding curves obtained from the numerical simulation showing the change of ΔT50% with L0/Kd,L for two different affinity constants 1.2 1.0 0.6 0.4 1.2 1.0 0.8 z=1/L0 z=1/L0 0.8 0.2 0.6 0.4 0.2 0.0 0.0 10 12 Time -0.2 10 15 20 25 Time Fig Average concentration at the outlet (z=1) normalized with the inlet concentration (L0) as a function of dimensionless time, obtained during measurement of KD corresponding to ΔT50%=2.7 (Table 1) with aprotinin-FITC (black dots) as the reference and insulin-FITC (green dots) as the ligand (Inset) Theoretical breakthrough curves for the reference (black, Pe=0.8 is dashed, Pe=8 is solid) and ligand (green, Pe=0.8 is dashed, Pe=8 is solid) RT and KD are constant Figure shows the experimental data for FITCconjugated human insulin binding to rat monoclonal antiinsulin IgG immobilized in the microchannel The reference used is FITC-conjugated aprotinin, with similar diffusivity The breakthrough curve for insulin is shifted to the right of the aprotinin curve as expected from the theoretical results (Fig inset) The aprotinin breakthrough occurs at t > 1, unlike the theory in Fig (inset), due to an additional outlet volume in our experiments However, ΔT50% would be unaffected by this if the outlet volume does not contain any immoblised antibody Furthermore, to minimize the variation between different microchannels, each assay comprising a reference perfusion as well as a ligand perfusion was performed in the same microchannel Our measured value for ΔT50% was 2.7±1.3 Our experimentally determined ΔT50% allows us to estimate KD from a plot of ΔT50% vs KD for a fixed L0 = μM The value of KD was found to be 0.76 μM Further confirmation of this KD may be achieved by experiments with other L0 values in the linear part of the binding curve IV CONCLUSIONS We have developed a microfluidics-based method of determining the equilibrium dissociation constant of any re- ceptor-ligand pair This method relies upon determination of the difference in breakthrough times between that of a reference molecule and that of the ligand The theoretical results presented here may also be extended to the study of equilibrium binding (or systems with fast binding kinetics) between the immobilized receptor and competing molecules in solution The proposed method has the following advantages: 1) works with complex samples, 2) uses less reagents than standard ELISA, 3) allows a larger selection of detection techniques, 4) relatively quicker and, 5) has better mass transfer characteristics than well plates The apparent disadvantage of using a reference molecule may be overcome by applying, a ‘virtual’ reference instead of a ‘real’ reference, after the assay has been standardized The advances in immobilization methods and linker designs together with low cost microfabrication, makes this a viable and promising technique for large scale adoption ACKNOWLEDGMENT The authors would like to thank Song Ying for assistance with the experimental protocols and Chaitanya Kantak for helping with microchannel fabrication This study was funded by NUS-FRC research grant R397000015112/101 REFERENCES Smith J, Jones M Jr, Houghton L et al (1999) Future of health insurance N Engl J Med 965:325–329 South J, Blass B (2001) The future of modern genomics Blackwell, London Smith J, Jones M Jr, Houghton L et al (1999) Future of health insurance N Engl J Med 965:325–329 DOI 10.10007/s002149800025 Lock I, Jerov M, Scovith S (2003) Future of modeling and simulation, IFMBE Proc vol 4, World Congress on Med Phys & Biomed Eng., Sydney, Australia, 2003, pp 789–792 IFMBE at http://www.ifmbe.org Use macro [author address] to enter the address of the corresponding author: Author: Institute: Street: City: Country: Email: Solute Concentration Gradients in A Hydrogel-assisted Gradient Generator C W Tan1, P Roy2 Graduate Program In Bioengineering, National University of Singapore Division of Bioengineering, National University of Singapore Abstract We report the design, fabrication and testing of a microfluidic device capable of generating linear Sink channel biomolecule concentration gradients across a static, fluid-filled domain The system comprises a source microfluidic channel and a sink microfluidic channel aligned in parallel These are connected by 24 perpendicular cross-channels, and converge at the end of the series of cross-channels to form a single eluent channel The main and cross-channels were about 105 μm and 310 μm in width respectively Each cross-channel was about 317 μm in length All the microfluidic channels were 50 μm in height The final product includes hydrogel barriers at either end of each cross-channel Each hydrogel had a height of 50 μm and a width of 82 μm The fluorophores, Alexa Fluor 488 and 2-NBDGwere used as probes for testing the system Either fluorophore at μM and 10 μM was used, while phosphate buffered saline (PBS) served as the sink The gradient generator was able to establish linear probe concentration gradients across the length of the cross-channels, except near the boundary of the hydrogel where some optical interference was evident The linearity was independent of the concentrations of probe used The gradients were also stable and reproducible throughout the system Using 2-NBDG as the probe yielded similar data, although the optical interference by the hydrogel appeared more prominent This interference was less significant at higher concentrations of the probe, presumably due to a higher signal to noise ratio Our system offers the advantages of (i) being easily fabricated so that dedicated laboratories as well as expertise are not crucial for its use; (ii) being able to generate biomolecule concentration gradients over length-scales mimicking those found in vivo; and most importantly; (iii) being able to so without exposing the cells to shear These advantages make the HAGG an easily accessible and easily operated device for biologists and biengineers alike Keywords Concentration gradient, microfluidic, microchannel, diffusive mixing List of Topics Biomedical Imaging Bioinformatics Biomaterials Biomechanics Biomolecular Engineering Biomedical Instrumentation Computational Bioengineering Biosignal Processing Nanobioengineerin g Micro Biomedical Engineering Systems Tissue Engineering Others World Academy of Science, Engineering and Technology 53 2009 Glucose-dependent Functional Heterogeneity In β-TC-6 Murine Insulinoma Darren C-W Tan, and Partha Roy allow the optimization of experimental setups and protocols The murine SV40 T-antigen-transformed pancreatic β-cell line, β-TC-6, was selected for our purpose because it retains major functional characteristics of pancreatic β-cells, in addition to being easier to maintain in culture [16] However, it was necessary to determine if β-TC-6 retains the functional heterogeneity observed in primary pancreatic β-cells To ascertain this, we performed three distinct functional assays to observe: (i) its ability to endocytose and metabolize the fluorescent glucose analog, 2-[N-(7-nitrobenz-2-oxa-1,3diaxol-4-yl)amino]-2-deoxyglucose (2-NBDG) [17]–[20]; (ii) its ability to respond to a glucose challenge by increasing membrane permeability to calcium ions [7]; (iii) as well as its ability to secrete quinacrine, and hence insulin, dosedependently [7] Finally, the Michaelis-Menten constant, as well as the maximum glucose uptake rate of β-TC-6 was measured for future reference Abstract—To determine if the murine insulinoma, β-TC-6, is a suitable substitute for primary pancreatic β-cells in the study of βcell functional heterogeneity, we used three distinct functional assays to ascertain the cell line’s response to glucose or a glucose analog These assays include: (i) a 2-NBDG uptake assay; (ii) a calcium influx assay, and; (iii) a quinacrine secretion assay We show that a population of β-TC-6 cells endocytoses the glucose analog, 2NBDG, at different rates, has non-uniform intracellular calcium ion concentrations and releases quinacrine at different rates when challenged with glucose We also measured the Km for β-TC-6 glucose uptake to be 46.9 mM and the Vm to be 8.36 x 10-5 mmole/million cells/min These data suggest that β-TC-6 might be used as an alternative to primary pancreatic β-cells for the study of glucose-dependent β-cell functional heterogeneity Keywords—2-NBDG, Fura-2/AM, functional heterogeneity, quinacrine I INTRODUCTION II MATERIALS AND METHODS ANCREATIC β-cells in the Islets of Langerhans play a pivotal role in maintaining homeostatic concentrations of glucose in the systemic circulation When exposed to elevated levels of glucose in the blood, they endocytose the molecule, whose metabolism ultimately results in the release of insulin from cytoplasmic granules [1], [2] This glucose-dependent response is based on the cells’ ability to sense glucose and respond accordingly, the failure of which typically results in diabetic pathologies [1], [3] Studies have demonstrated that although the β-cells in an islet respond in a synchronized manner to a glucose challenge, not all of the population secretes insulin [4]–[9] Of those that do, variation is observed in the enzymatic activity as well as quantity of insulin released This functional heterogeneity is a characteristic of glucose-dependent pancreatic β-cell function and allows the tissue fine control over the response Interestingly, this heterogeneity is plastic and appears to be modulated by glucose itself [3], [7], [10]–[15] Our laboratory is interested in the study of pancreatic β-cell functional heterogeneity However, owing to the difficulty in isolating, purifying and maintaining primary β-cell cultures, a hardier and more easily managed alternative was sought to P A β-TC-6 culture β-TC-6 cells were purchased from American Type Culture Collection (ATCC) and cultured in Dulbecco’s Modified Eagle Medium Cultures were maintained at 37 °C and under a % CO2 atmosphere inside a humidified incubator The growth medium was changed every three days B 2-NBDG uptake and Metabolism Assay β-TC-6 at 1.0 x 105 cells/mL was seeded onto a 24-well plate and allowed to attach, spread and proliferate to near confluence at 37 °C, % CO2 The complete medium was then removed and the cells were washed with PBS The PBS was then replaced with 2.5 mM glucose in basal medium comprising Dulbecco’s Modified Eagle medium (DMEM) with neither glucose nor pyruvate (Product No 11966, GIBCO, Invitrogen, USA), supplemented with L-glutamine and 10 % (v/v) fetal bovine serum (final serum glucose concentration of about 0.25 mM) Conditioning of the cells proceeded at 37 °C, % CO2 for 60 mins The conditioning medium was then removed and replaced with 200 μM 2-[N(7-nitrobenz-2-oxa-1,3-diaxol-4-yl)amino]-2-deoxyglucose (2-NBDG, Invitrogen, USA) in basal medium The cells were then incubated further at 37 °C, % CO2 for 30 mins to allow them to endocytose the glucose analog The 2-NBDG in basal medium was then removed and the cells were washed with Darren C-W Tan is with the Graduate Program in Bioengineering, National University of Singapore (email: g0402807@nus.edu.sg) Partha Roy is with Division of Bioengineering, National University of Singapore (phone: (65) 6515 1624; e-mail: biepr@nus.edu.sg) 368 World Academy of Science, Engineering and Technology 53 2009 were taken at t = 0, 10, 20, 30, 40, 50, 60, 80, 100 and 120 mins for glucose quantitation using an ACCU-CHEK glucose quantitation kit (Roche Diagnostics) Finally, the cells in each well were harvested by trypsinisation and counted using a haemocytometer PBS They were then observed for intra-cellular fluorescence at excitation and emission wavelengths of 467 nm and 542 nm respectively C Glucose-dependent Calcium Ion Influx The complete medium was removed from a T-25 culture of β-TC-6 at about 50 % confluence and the cells were washed with PBS The PBS was then replaced with 2.5 mM glucose in basal medium comprising Dulbecco’s Modified Eagle medium (DMEM) with neither glucose nor pyruvate (Product No 11966, GIBCO, Invitrogen, USA), supplemented with Lglutamine and 10 % (v/v) fetal bovine serum (final serum glucose concentration of about 0.25 mM) Conditioning of the cells proceeded at 37 °C, % CO2 for 60 mins The conditioning medium was then removed and replaced with μM fura-2/acetoxymethyl ester (Fura-2/AM, Sigma-Aldrich, USA) in basal medium supplemented with 10 % (v/v) fetal bovine serum The cells were then incubated at 37 °C, % CO2 for 30 mins after which they were washed twice with PBS 25 mM glucose in PBS supplemented with mM CaCl2 was added and the cells were observed for intra-cellular fluorescence at excitation and emission wavelengths of 360 nm and 500 nm respectively III RESULTS A 2-NBDG uptake and Metabolism Assay The data show that β-TC-6 can endocytose the fluorescent glucose analog, 2-NBDG, and also shows increased intracellular 2-NBDG fluorescence in distinct groups of cells (Fig 1) Fluorescence appears to be restricted to the cytoplasm As the time allowed for 2-NBDG uptake should be sufficient for 2-NBDG uptake and metabolism to stabilise [20], this heterogeneous intracellular 2-NBDG fluorescence suggests that β-TC-6 retains the heterogeneous glucose uptake activity of native β-cells (a) (b) D Quinacrine Secretion Assay β-TC-6 at 1.0 x 105 cells/mL was seeded onto a 24-well plate and allowed to attach, spread and proliferate to near confluence at 37 °C, % CO2 The complete medium was then removed and the cells were washed with PBS The PBS was then replaced with 500 μL of glucose-free DMEM supplemented with 0.1 % bovine serum albumin (BSA) Conditioning of the cells proceeded at 37 °C, % CO2 for 30 mins The basal medium was then replaced with a fresh aliquot Incubation at 37 °C, % CO2 was resumed for a further 30 mins The basal medium was then replaced with 100 nM quinacrine dihydrochloride (Aldrich, USA) and the cells were incubated at 37 °C, % CO2 for 15 mins The cells were then washed twice with PBS and the center of each well was imaged under fluorescence excitation at 360 nm and emission at 500 nm for 0.5 s The PBS was then replaced with DMEM supplemented with 0.1 % BSA and 0.5 mM, 1.0 mM, 2.8 mM, 5.6 mM or 16.5 mM glucose The cells were then incubated at 37 °C, % CO2 for 60 mins Subsequently, the cells were washed with PBS and imaged under fluorescence excitation at 360 nm and emission at 500 nm for 0.5 s The fluorescence intensity of the cells in these images was then analysed using ImagePro Plus software Fig Brightfield (a) and fluorescence (b) images of β-TC-6 cells in 24-well plate loaded with 2-NBDG showing heterogeneous 2-NBDG uptake and metabolic activity B Glucose-dependent Calcium Ion Influx Loading β-TC-6 with the calcium chelator, Fura-2/AM and then challenging them with glucose resulted in an increase in intracellular fluorescence at 500 nm (Fig 2) compared to controls (not shown) This demonstrated that Fura-2/AM was able to penetrate the β-TC-6 cell membrane and detect an increase in intracellular calcium ion concentration as previously reported by Holz and co-workers [21] As the intracellular Fura-2/AM is not uniform, β-TC-6 appears to display different calcium ion influx rates in different cells (a) E β-TC-6 Glucose Uptake Rate (b) Fig Brightfield (a) and fluorescence (b) images of β-TC-6 cells in T-25 flask loaded with Fura-2/AM showing intracellular fluorescence following a glucose challenge β-TC-6 at 1.0 x 10 cells/mL was seeded onto a 24-well plate and allowed to attach, spread and proliferate at 37 °C, % CO2 The complete medium was then removed and the cells were washed with PBS The wells were then filled with 500 μL of basal media containing 5.8 mM or 2.9 mM glucose The plate was then incubated at 37 °C, % CO2 for 120 mins with gentle agitation at 100 rpm Aliqouts of 10 μL from each well C Quinacrine Secretion Assay Quinacrine dihydrochloride was able to penetrate and accumulate in the β-TC-6 cells This resulted in a marked increase in intracellular fluorescence at 500 nm (Fig 3) 369 World Academy of Science, Engineering and Technology 53 2009 Michaelis-Menten type equation This allowed us to determine the Michaelis-Menten constant to be, Km = 46.9 mM, and the maximum glucose uptake rate to be, Vm = 8.36 x 10-5 mmole/million cells/min Furthermore, the intracelluar fluorescence appears to be punctate and the specks of increased fluorescence are likely to be quinacrine-loaded insulin granules Non-uniform intracellular fluorescence suggests that β-TC-6 has retained heterogeneous insulin content IV DISCUSSION (a) The glucose analog, 2-NBDG, has been used to determine cell viability as well as to estimate glucose uptake rates in a variety of cell types [17]-[20] As β-TC-6 is a β-cell derivative, it is expected to be able to endocytose the fluorescent analog This ability is not uniform and certain groups of cells were observed to have higher uptake rates or lower metabolic rates, which resulted in regions of increased intracellular 2-NBDG fluroescence This demonstates the heterogeneous nature of the 2-NBDG, and hence glucose, processing machinery in the cell line This heterogeneity was also observed, albeit less distinctly, when the cells were stained with quinacrine Quinacrine is a fluorescent probe that readily diffuses across cell membranes and accumulates in regions of low pH, such as secretory vesicles In β-TC-6 cells, it is expected to co-localise with the insulin granules and has been reported to so in primary pancreatic β-cells Our data showing non-uniform quinacrine staining demonstrates the heterogeneous insulin content among β-TC-6 cells The secretion assay data also confirms Poitout’s characterisation of glucose-dependent insulin secretrion in β-TC-6 [16] In pancreatic β-cells, glucose-dependent insulin secretion has been shown to be concomitant with a calcium ion influx [14], [18] The calcium ion chelator Fura-2/AM, used in these studies, readily penetrates cell membranes and fluoresces when in the presence of Ca2+ Our data showing non-uniform intracellular Fura-2/AM fluorescence, following a glucose challenge, suggests that β-TC-6 retains heterogeneous insulin secretion rates among its cells The ability of β-TC-6 to respond appropriately to a glucose challenge supports its use as a viable alternative to primary pancreactic β-cells in the study of glucose-dependent functional heterogeneity Our estimate of the Michaelis-Menten constant for β-TC-6 appears to be higher than reported values for the predominant GLUT2 receptor as well as the metabolic enzyme, glucokinase in pancreatic β-cells [3], [15] Our assay does not distinguish between the affinity of either GLUT2 or glucokinase for glucose, but does provide a value that might be a resultant value for both activities A more rigorous measurement of the Km for β-TC-6 might yield results closer in agreement to those reported However, it should be noted that β-TC-6, being a transformed derivative of primary pancreatic β-cells, might not have retained its native affinity for glucose Furthermore, our cultures were maintained in growth media containing 21 mM glucose, a supra-physiological concentration which might have altered β-TC-6 glucose dependence We aim to carry out further studies of glucose-dependent functional heterogeneity in β-TC-6 in a microfluidic channel system This would make it possible for us to employ microfluidic phenomena to manipulate the extracellular (b) Fig Brightfield (a) and fluorescence (b) images of β-TC-6 cells in T-75 flask loaded with quinacrine Percentage of pre-stimulation fluorescence Furthermore, subsequent incubation with medium containing glucose resulted in a decrease in intracellular fluorescence intensity, suggesting that the quinacrine had localised within the insulin granules and been co-secreted with the insulin upon glucose stimulation of the β-TC-6 cells Quantitative assays performed in 24-well plates also showed that there was a dose-dependent decrease in intracellular quinacrine fluorescence (Fig 4) The release profile so-obtained is similar to the insulin secretion profile described by Poitout and co-workers for β-TC-6 cells, with the half-maximal seceretion rate at about 0.5 mM glucose [16] 95% 90% 85% 80% 75% 70% 65% 60% 0.0 5.0 10.0 15.0 20.0 Concentration of glucose (mM) Fig Graph showing glucose-dependent secretion of quinacrine from β-TC-6 (P