notes

Notes

1

preface and acknowledgements

Preface and acknowledgements This thesis is the product of a 1 year experimental project, marking the end of a 2 years Biochemistry Master’s Degree at the University of Copenhagen. Besides having focused on my qualification profile within the field of protein chemistry, this project involves immunology and biophysical aspects of biochemical interactions. This project was performed from April 2012 to April 2013 at the Bartholin Institute, Rigshospitalet (Denmark) under supervision of Professor Carl-Henrik Brogren, The Bartholin Institute, with Professor Anders Woetmann Andersen, University of Copenhagen, as academic supervisor. The data from this project were presented at several conferences in Denmark, Sweden, Germany and Poland. The last mentioned was acknowledged with a 3rd prize in the MST Conference Award 2012. I would like to thank all the people who have helped me both professionally and personally throughout the course of this project. This in particular includes friends and colleagues at the Bartholin Institute, Rigshospitalet (Denmark). A special thanks to my supervisor Carl-Henrik Brogren for help and guidance, interesting scientific discussions and for letting me unfold my scientific creativity. Also a special thanks to my dear friend and fellow student Kathrine Louise Jensen for both personal and scientific support. Thanks to Mathilde Voetmann, Christina Engmose, Sarah Christensen and Mie Sefeld from the IC2 group, and to Postdoc Kåre Engkilde, Professor Karsten Buschard and Professor Knud Josefsen for scientific help and advice. I am also exceedingly grateful to Professor Marité Cárdenas, head of the Membrane Properties group at the Nano-Science Center, University of Copenhagen (Denmark), who let me use the QCM-D instrument and gave me help and guidance during this project. I was lucky to get help and lots of good advice from Tania Kjellerup Lind and Anna Åkesson as well as the other members of the Membrane Properties group. I owe a lot of gratitude to Research Director 1st class Anne Houdusse, head of the Structural Motility Group at Institute Curie, Paris (France), for welcoming me into her group and for letting me use her Monolith.NT. Also a special thanks to Ingénieur de Recherche Carlos Kikuti for help and guidance with both acquisition and interpretation of data, and for giving me a unique insight into the field of crystallization and structural biochemistry. Another big thank you goes out to CEO Teodor Aastrup and all coworkers at Attana AB in Stockholm (Sweden) for hosting me and letting me perform experiments on the Attana Cell 200. A special thanks to Application Specialist Camilla Käck for kind help and guidance, both during my stay and after my departure. A kind thanks to Professor Jens Stougaard and Postdoc Mickael Blaise at The Department of Molecular Biology and Genetics at Aarhus University (Denmark) for use of the Monolith.NT and for help and guidance with interpretation and evaluation. I would also like to thank NanoTemper Technologies GmbH and Q-sense for the help and guidance I have received during this project.

2

preface and acknowledgements

I received important financial support through a summer fellowship from Federation of Biochemical Societies which made it possible for me to travel and stay in Paris (France) for almost 2 months while microscale thermophoresis experiments was performed. Finally, I want to thank my beloved family, friends and amazing girlfriend for supporting me, accepting my devotion to this project and for always being there for me. I love you all.

Copenhagen, April 2013 Ida Dalgaard Pedersen

Front page illustration: Confocal fluorescence microscopic staining with IC2-IgM on a single INS-1E pancreatic beta-cell (Pedersen et al., 2011).

3

abstract

Abstract Type 1 diabetes mellitus is a chronic disorder resulting from autoimmune destruction of the insulinproducing pancreatic β-cells in the islet of Langerhans. The cause of type I diabetes is thought to be a multistep precipitation of environmental factors in genetically susceptible individuals leading to an imbalanced T-cells distribution. A monoclonal autoantibody, IC2, has a unique specificity for the insulin secreting pancreatic β-cell both in vitro and in vivo. The unique properties of IC2 as a functional biomarker makes it an obvious candidate for non-invasive imaging of the functional β-cell mass. The exact epitope for IC2 on the pancreatic β-cell surface still remains largely unknown. The IC2 antigen was found to be trypsinsensitive, suggesting that a protein is involved. Surprisingly, the antigen was later found to be of lipogenic nature, characterized as a sulfatide, suggesting that the epitope of IC2 could be a protein-lipid complex constituted on the surface of the β-cell. Recently, sphingomyelin, phosphatidylcholine, and possibly the lysoforms of these, were found to be involved in the binding of IC2. In addition, it was shown that the loss of cholesterol leads to a loss of IC2 binding. In a modified functional NKT-hybridoma assay, IC2 was found to inhibit type I NKT cells and possibly also type II NKT cells. These results strongly suggest the involvement of the CD1d receptor which is known to stimulate NKT-cells. The primary aim of this study was to obtain a better understanding of the identity of the IC2 autoantigen. This was to be done by performing affinity measurements of IC2 towards intact cells, plasma membranes and different lipids. For this, cellular quartz crystal microbalance, quartz crystal microbalance with dissipation and microscale thermophoresis was used. Also immunoblotting was performed. A secondary aim of this study was to compare the affinity of different IC2 antibody formats (IgM, rhIgG, F(ab’)2 and Fab) towards intact β-cells, to determine the best suited format for further imaging trials. Surprisingly, interactions between IC2 and intact cells were not successfully measured in this project, although interactions with the used β-cell lines had been observed several times in the past. Interaction was observed between IC2-F(ab’)2 and sonicated β-cell plasma membrane. With quartz crystal microbalance with dissipation, interaction was observed between IC2 and sphingomyelin, sphingosylphosphorylcholine, cholesteryl sulphate and possibly also sulfatide. Furthermore, interaction between IC2 and lyso-sulfatide was observed using microscale thermophoresis. By theoretical comparison of the found IC2 lipid targets, it is suggested that IC2 recognition is spatially specific for lipids with smaller extruding carbohydrate or choline groups, and not charge nor the presence of particular functional groups. It is suggested that the natural autoantigen of IC2 is a formation of a special CD1d-lipid complex on the surface of pancreatic β-cells. The involved lipid must be specific for pancreatic βcells either by involving a hitherto unknown lipid with origin in the active insulin producing β-cell, or in carrying a special β-cell imprint such as a bound zinc-ion. The affinity of this β-cell specific complex is proposed to be higher than binding of IC2 to other CD1d-lipid complexes. Therefore the IC2-CD1d-lipid interactions are lost after a certain period of clearance, after which only the β-cell specific IC2 binding is present. It is suggested that the current findings be supported by additional interaction experiments. Furthermore, purification and identification of the IC2 autoantigen by immuno thin-layer chromatography and lipid mass spectrometry is suggested. To further investigate the role of the CD1d receptor in IC2 recognition, confocal fluorescence microscopy and competition experiments using flow cytometry and ELISA are suggested.

4

resumé

Resumé Diabetes mellitus type I er en kronisk lidelse, som skyldes autoimmun destruktion af insulinproducerende β-celler i de Langerhanske øer i pancreas. Årsagen til type I diabetes menes at være en ubalance i fordelingen af T-celler forårsaget af en kombination af en række miljømæssige faktorer i genetisk disponerede individer. Et monoklonalt autoantistof, IC2, har en unik specificitet over for insulinsekrerende pankreatiske β-celler, både in vitro og in vivo. IC2s unikke egenskaber som en funktionel biomarkør gør det til en oplagt kandidat til brug ved ikke-invasiv afbildning af den funktionelle β-celle masse. Præcis hvilken epitop, IC2 genkender på overfladen af den pankreatiske β-celle, er stort set ukendt. IC2 antigenet har vist sig at være trypsinsensitivt, hvilket indikerer, at et protein er involveret. Senere hen fandt man, at IC2antigenet var af lipogen oprindelse, og det blev karakteriseret som værende et sulfatid. Sammen indikerer dette, at IC2 epitopen kan være et protein-lipid komplex på overfladen af β-celler. Senest har sphingomyelin, phosphatidylcholin samt muligvis også deres lysoformer vist sig at være involveret i IC2bindingen. Herudover blev det vist, at elimerering af cholesterol også medfører tab af IC2-binding. Ved en modificeret funktionel NKT-hybridom analyse fandt man, at IC2 kunne hæmme type I og muligvis også type II NKT celler. Disse resultater peger stærkt på, at CD1d receptoren, som er kendt for at stimulere NKT-celler, er involveret i bindingen af IC2. Formålet med dette studie var primært at opnå en mere grundig forståelse for IC2 autoantigenets identitet. Dette blev hovedsageligt gjort ved affinitetsmålinger mellem IC2 og hele celler, plasmamembraner samt diverse lipider. Til dette blev følgende teknikker brugt: cellulær quartz crystal microbalance, quartz crystal microbalance with dissipation, microscale thermophoresis, samt immunblotting. Et andet formål med dette studie var at sammenligne affiniteter af bindingen mellem forskellige IC2 antistofformater (IgM, rhIgG, F(ab’)2 og Fab) og hele β-celler for at finde det bedst egnede format til fremtidige billeddiagnostiske studier. Overraskende nok sås der i dette studie ikke binding mellem IC2 og intakte β-celler på trods af, at dette har været påvist utallige gange tidligere. Der blev dog observeret binding mellem IC2-F(ab’)2 og sonikerede βcelle plasmamembraner. Med quartz crystal microbalance with dissipation blev der observeret binding mellem IC2 og sphingomyelin, sphingosylphosphorylcholin, cholesteryl sulfat og muligvis også sulfatid. Derudover blev der ved microscale thermophoresis set binding mellem IC2 og lyso-sulfatid. Ved teoretisk sammenligning af de fundne IC2-lipid antigener vurderes det, at IC2 binding er rumligt afhængigt af lipider med mindre fremtrædende kulhydrat- eller cholin-grupper, og ikke af ladning eller tilstedeværelsen af speficikke funktionelle grupper. Det foreslås, at det naturlige IC2 autoantigen er en formation af et specielt CD1d-lipid kompleks på overfladen af pankreatiske β-celler. Det involverede lipid kan være specifikt for pankreatiske β-celler, enten idet det er et hidtil udkendt lipid med oprindelse i den aktivt insulinsekrerende β-celle, eller ved at det bærer et specielt β-celle-aftryk som f.eks. en bundet zinkion. Affiniteten for dette β-celle specifikke kompleks menes at være højere end bindingen af IC2 til andre CD1d-lipid komplekser, hvorfor andre IC2-CD1d-lipid interaktioner dissocierer efter en vis periode med klarering, hvorefter IC2 bindingen er β-celle specifik. Det foreslås, at nærværende opdagelser bakkes op med yderligere affinitetsmålinger. Herudover foreslås det at oprense og identificere IC2 autoantigenet ved hjælp af immun tyndt-lags kromatografi og lipid massespektrometri. For at undersøge CD1d-receptorens rolle i bindingen af IC2 nærmere foreslås konfokal fluorescens mikroskopi samt competition-forsøg ved brug af flow cytometri og ELISA.

5

table of contents

Table of contents List of abbreviations ........................................................................................................................................ 12 1. Introduction ................................................................................................................................................. 14 1.1. Type 1 diabetes..................................................................................................................................... 14 1.1.1. A chronic autoimmune disorder .................................................................................................... 14 1.1.2. Cause of type I diabetes ................................................................................................................ 15 1.1.2.1. T-cells ...................................................................................................................................... 15 1.1.2.2. Genetics .................................................................................................................................. 15 1.1.2.3. Viruses and bacteria ............................................................................................................... 16 1.1.2.4. Polyclonal activation ............................................................................................................... 16 1.1.3. Silent immune events .................................................................................................................... 18 1.2. Pancreas and the Islet of Langerhans ................................................................................................... 18 1.2.1. Anatomy ........................................................................................................................................ 18 1.2.2. Insulin ............................................................................................................................................ 19 1.2.2.1. Production of insulin .............................................................................................................. 19 1.2.2.2. Secretion of insulin ................................................................................................................. 19 1.3. IC2 ......................................................................................................................................................... 20 1.3.1. The monoclonal autoantibody IC2 ................................................................................................ 20 1.3.2. Imaging and various IC2 formats ................................................................................................... 22 1.3.3. The IC2 autoantigen ...................................................................................................................... 23 1.4. NKT-cells and CD1d............................................................................................................................... 26 1.4.1. Natural killer T-cells (NKT-cells) ..................................................................................................... 26 1.4.2. NKT subsets ................................................................................................................................... 26 1.5. CD1d ..................................................................................................................................................... 27 1.5.1. The CD1d molecule ........................................................................................................................ 27 1.5.2. CD1d folding and binding of lipids ................................................................................................. 28 1.6. Classification of lipids ........................................................................................................................... 29 1.6.1. Simple, complex and derived lipids ............................................................................................... 29 1.6.2. Glycolipids...................................................................................................................................... 29 1.6.2.1. Sphingolipids and glycosphingolipids ..................................................................................... 29 1.6.3. Selected lipids ................................................................................................................................ 30 1.6.3.1. The galactosylcerebrosides α-GalCer and β-GalCer ............................................................... 30

6

table of contents

1.6.3.2. Sulfatides ................................................................................................................................ 30 1.6.3.3. Lyso-sulfatides ........................................................................................................................ 31 1.6.3.4. Sphingomyelins....................................................................................................................... 31 1.6.3.5. Lyso-sphingomyelins .............................................................................................................. 31 1.6.3.6. Phosphatidylcholines .............................................................................................................. 32 1.6.3.7. Sodium cholesteryl sulphate .................................................................................................. 32 1.7. Antibody interactions and kinetics ....................................................................................................... 32 1.7.1. Affinity and avidity......................................................................................................................... 32 1.7.2. Kinetics........................................................................................................................................... 33 1.8. Affinity measurements ......................................................................................................................... 34 1.8.1. Prevailing techniques form measuring antibody-antigen affinities .............................................. 34 1.8.2. Quartz crystal microbalance (QCM) .............................................................................................. 35 1.8.3. Microscale thermophoresis (MST) ................................................................................................ 36 1.8.4. Choice of methods for this project ................................................................................................ 38 2. Study plan .................................................................................................................................................... 39 3. Materials and methods ............................................................................................................................... 40 3.1. Materials ............................................................................................................................................... 40 3.1.1. Chemicals and reagents ................................................................................................................. 40 3.1.2. Cell lines ......................................................................................................................................... 40 3.1.3. Materials and reagents for cell culture ......................................................................................... 41 3.1.4. Antibodies and serum.................................................................................................................... 41 3.1.5. Lipids .............................................................................................................................................. 41 3.1.6. Other reagents............................................................................................................................... 42 3.2. Methods ............................................................................................................................................... 42 3.2.1. Cell culturing in culture flasks........................................................................................................ 42 3.2.2. Preparation and isolation of plasma membranes ......................................................................... 42 3.2.2.1. Cells from culture flasks ......................................................................................................... 42 3.2.2.2. Cells from cell factories .......................................................................................................... 42 3.2.2.3. Isolation of plasma membranes ............................................................................................. 43 3.2.3. Cellular quartz crystal microbalance experiments (cellular QCM) ................................................ 43 3.2.3.1. Growth experiment of adherent cells in 24 well plates ......................................................... 44 3.2.3.2. Culturing of adherent cells on COP-1 sensor chips in 24 well plates ..................................... 44 3.2.3.3. Fixation of cells on COP-1 sensor chips .................................................................................. 44

7

table of contents

3.2.3.4. Cellular Attana Cell 200 experiments ..................................................................................... 44 3.2.4. Quartz crystal microbalance with dissipation experiments (QCM-D) ........................................... 45 3.2.4.1. Preparations ........................................................................................................................... 45 3.2.4.2. Sample preparation for QCM-D experiments ........................................................................ 45 3.2.4.2.1. Lipid samples ................................................................................................................... 45 3.2.4.2.2. Living cell samples ........................................................................................................... 46 3.2.4.2.3. Plasma membrane samples ............................................................................................. 46 3.2.4.2.4. Other samples and blocking solutions............................................................................. 46 3.2.5. Microscale thermophoresis (MST) ................................................................................................ 46 3.2.5.1. Calibration .............................................................................................................................. 47 3.2.5.2. Preparations and experiments ............................................................................................... 47 3.2.5.3. Sample preparation ................................................................................................................ 47 3.2.5.3.1. Lipid samples ................................................................................................................... 47 3.2.5.3.2. Plasma membrane samples ............................................................................................. 47 3.2.5.3.3. Other samples ................................................................................................................. 48 3.2.6. Immunoblotting ............................................................................................................................. 48 4. Results ......................................................................................................................................................... 49 4.1. Quartz crystal microbalance (QCM) ..................................................................................................... 49 4.1.1. Attana Cell 200 experiments ......................................................................................................... 49 4.1.1.1. 24 well growth experiment .................................................................................................... 49 4.1.1.2. Growth of cells on COP-1 sensors .......................................................................................... 49 4.1.1.3. Cellular QCM experiments...................................................................................................... 49 4.1.2. Quartz crystal microbalance with dissipation monitoring (QCM-D) ............................................. 51 4.1.2.1. Living cells ............................................................................................................................... 51 4.1.2.2. Plasma membrane sonicate ................................................................................................... 53 4.1.2.2.1. β- and α-cell plasma membrane sonicate ....................................................................... 53 4.1.2.2.2. A20 plasma membrane vesicles ...................................................................................... 55 4.1.2.3. Lipids ....................................................................................................................................... 57 4.1.2.3.1. Sphingomyelin ................................................................................................................. 57 4.1.2.3.2. Sphingosylphosphorylcholine .......................................................................................... 58 4.1.2.3.3. Sulfatide ........................................................................................................................... 60 4.1.2.3.4. Lyso-sulfatide................................................................................................................... 60 4.1.2.3.5. Other glycolipids .............................................................................................................. 61

8

table of contents

4.1.2.4. IC2 binding to sulphated monosaccharides ........................................................................... 62 4.2. Microscale thermophoresis (MST) ....................................................................................................... 63 4.2.1. β-cell plasma membrane ............................................................................................................... 63 4.2.2. Sphingomyelin and sphingosylphosphorylcholine ........................................................................ 64 4.2.3 Sulfatide and lyso-sulfatide ............................................................................................................ 65 4.2.3.1. Sulfatide .................................................................................................................................. 65 4.2.3.2. Lyso-sulfatide.......................................................................................................................... 65 4.2.4. Other lipids .................................................................................................................................... 69 4.2.5. Sulphated monosaccharides.......................................................................................................... 69 4.3. Immunoblotting on SphingoStrip ......................................................................................................... 69 5. Discussion .................................................................................................................................................... 70 5.1. Discussion of results ............................................................................................................................. 70 5.1.1. Binding of IC2 to intact cells and plasma membrane .................................................................... 70 5.1.1.1. Fixated and living cells ............................................................................................................ 70 5.1.1.1.1. Cellular quartz crystal microbalance (cellular QCM) ....................................................... 70 5.1.1.1.2. Quartz crystal microbalance with dissipation (QCM-D) .................................................. 71 5.1.1.2. Plasma membranes ................................................................................................................ 71 5.1.1.2.1. β-cells............................................................................................................................... 71 5.1.1.2.1.1. IC2-IgM interaction with α- and β-cells .................................................................... 71 5.1.1.2.1.2. IC2-F(ab’)2 interaction with β-cells ........................................................................... 72 5.1.1.2.2. A20-CD1d cells ................................................................................................................. 73 5.1.2. Interaction between IC2, intact cells and plasma membranes ..................................................... 74 5.1.3. Binding of IC2 to lipids ................................................................................................................... 75 5.1.3.1. Sphingomyelin ........................................................................................................................ 75 5.1.3.2. Sphingosylphosphorylcholine ................................................................................................. 76 5.1.3.3. Sulfatide .................................................................................................................................. 76 5.1.3.4. Lyso-sulfatide.......................................................................................................................... 77 5.1.3.4.1. Sigma lyso-sulfatide ......................................................................................................... 77 5.1.3.4.2. Matreya lyso-sulfatide ..................................................................................................... 77 5.1.3.5. Other lipids ............................................................................................................................. 78 5.1.3.5.1. Sodium cholesteryl sulphate ........................................................................................... 78 5.1.3.5.2. β-galactosylcerebroside .................................................................................................. 78 5.1.3.6. Combinations of lipids ............................................................................................................ 78

9

table of contents

5.1.3.7. Immunoblotting ...................................................................................................................... 79 5.1.3.8. IC2 seem to recognize several diverse lipids .......................................................................... 79 5.1.4. IC2 binding to sulphated monosaccharides .................................................................................. 80 5.1.4.1. IC2 does not bind to sulphated monosaccharides alone ....................................................... 81 5.2. Evaluation of methods.......................................................................................................................... 81 5.3. The IC2 antigen ..................................................................................................................................... 82 5.3.1. The molecular characteristics of IC2 recognized lipids ................................................................. 82 5.3.1.1. What do IC2 recognized lipids have in common? .................................................................. 82 5.3.1.1.1. Hydrophobic part of the lipids......................................................................................... 82 5.3.1.1.2. Polar head group of the lipids ......................................................................................... 82 5.3.1.1.3. IC2 might recognize a certain structure rather than certain sequences ......................... 83 5.3.1.2. The lipid chain seems to be essential for IC2 recognition ...................................................... 83 5.3.2. Lipids as targets for IC2 - but something is missing....................................................................... 84 5.3.2.1. IC2 might recognize a special protein-lipid complex in vivo .................................................. 84 5.3.2.2. Could IC2 bind to a slightly different CD1d expressed only on β-cells? ................................. 85 5.3.2.3. A specific lipid coupled to insulin production and/or secretion could be the IC2 antigen .... 85 5.3.2.3.1. A hitherto unknown lipid could be the IC2 antigen ........................................................ 85 5.3.2.3.2. Binding might depend on binding of a certain ion to charged regions of the lipid complex ........................................................................................................................................... 85 5.3.3. The NKT inhibition experiments reveals details about the IC2 binding to CD1d complexes ........ 86 5.3.4. IC2 and the β-cell specificity .......................................................................................................... 87 5.3.4.1. IC2 may have a higher affinity for CD1d-lipid complexes carrying a special β-cell imprint ... 87 5.3.4.2. Lipids as targets of future therapeutics.................................................................................. 88 6. Conclusions and perspectives...................................................................................................................... 89 6.1. Conclusions ........................................................................................................................................... 89 6.2. Perspectives and future experiments................................................................................................... 89 6.2.1. Determining the IC2 autoantigen .................................................................................................. 89 6.2.1.1. Investigation of interactions ................................................................................................... 89 6.2.1.2. A new direction ...................................................................................................................... 90 6.2.1.2.1. Mass spectrometry .......................................................................................................... 90 6.2.1.2.2. Bioinformatics.................................................................................................................. 90 6.2.1.2.3. The role/involvement of the CD1d receptor ................................................................... 91 6.2.2. Other perspectives ........................................................................................................................ 91

10

table of contents

6.2.2.1. Imaging and determination of IC2-format for future imaging trials ...................................... 91 6.2.2.2. Antiidiotypic antibody ............................................................................................................ 91 6.2.2.3. Further explore the functional abilities of IC2........................................................................ 92 7. References ................................................................................................................................................... 93 Appendix I ...................................................................................................................................................... 100 Appendix II ..................................................................................................................................................... 101 Appendix III .................................................................................................................................................... 102 Appendix IV.................................................................................................................................................... 105 Appendix V..................................................................................................................................................... 110

11

list of abbreviations

List of abbreviations α-GalCer AC ADP ATP β-GalCer BB BLI Breg BSA CELISA

KD kd LED LSM LSU MHC MQ MRI MST NIR

Dissociation constant Dissociation rate constant Light-emitting diode Lyso-sphingomyelin Lyso-sulfatide Major histocompatibility complex Mili-Q water Magnetic resonance imaging Microscale thermophoresis Near-infrared

CFM CRIA D-Gal-3-Sul D-Gal-4-Sul D-Glu-3-Sul DAPI DC DLS DMSO DP EDTA ELISA ER FCM FITC FMT GAD GlcCer GLuc GLUT2

α-galactosyl-ceramide Alternating current Adenosine diphosphate Adenosine triphosphate β-galactosyl-ceramide Biobreeding Bioluminescence imaging Regulatory B-cell Bovine serum albumin Cellular enzyme-linked immuno sorbent assay Confocal fluorescence microscopy Cellular RIA D-galactose-3-sulphate D-galactose-4-sulphate D-glucose-3-sulphate 4’-6-diamidino-2-phenylindole Dendritic cell Dynamic light scattering Dimethyl sulfoxide Diabetes prone Ethylenediaminetetraacetic acid Enzyme-linked immuno sorbent assay Endoplasmic reticulum Flow cytometry Fluorescein isothiocyanate Fluorescence molecular tomography Glutamic acid decarboxylase Glucosylceramide Gaussian luciferase Glucose transporter 2

NK cell NKT NOD PAT PBS PET PFA PP-cells PVP-40 QCM QCM-D RBC rhIgG RIA RPM RT SCS SM SNARE SPECT

HLA HRP Hz IAA IA2 IB ICC IHC IL-2 ITC iTLC + K /ATP KA ka

Human leukocyte antigen Horseradish peroxidase Hertz Insulin antibody Islet-cell antibody Immunoblotting Immunocytochemistry Immunohistochemistry Interleukin 2 Isothermal titration calorimetry Immuno thin-layer chromatography ATP sensitive potassium ion channel Association constant Association rate constant

SPR SUL T1DM TBS TCR Th Th1 Th2 Treg ZnT8

Natural killer cell Natural Killer T-cell Non-obese diabetic Photoacoustic tomography Phosphate-buffered saline Positron emission tomography Paraformaldehyde Polypeptide producing cells Polyvinylpyrrolidone-40 Quartz crystal microbalance Quartz crystal microbalance with dissipation Red blood cell Recombinant immunoglobulin G Radioimmunoassay Revolutions per minute Room temperature Sodium cholesteryl sulphate Sphingomyelin Soluble NSF attachment protein receptor Single-photon emission computed tomography Surface Plasmon resonance Sulfatide Type I diabetes mellitus Tris-buffered saline T-cell receptor T-helper cell Type 1 T-helper cell Type 2 T-helper cell Regulatory T-cell Zinc transporter 8

12

13

introduction

1. Introduction Type 1 diabetes mellitus (T1DM) is a chronic disorder resulting from autoimmune destruction of the insulinproducing β-cells. Globally, the T1DM incidence has been rising during the past decades. The cause of type 1 diabetes is thought to be a combination of genetic and environmental factors. Both type 1 and type 2 diabetic patients would benefit tremendously from the deduction of non-invasive imaging studies, because a unique insight into the pathology and dysfunctionality of the β-cell mass would be obtained. Besides leading to a better characterization of the diabetic process and an early diagnosis, these studies could have the potential of showing results of drug interventions, thus helping with decisions both regarding type of therapy and treatment follow-up in clinical settings. This study relates to the autoantigen of the pancreatic β-cell specific monoclonal autoantibody IC2. Because of its specificity, IC2 is a potential candidate as biomarker for non-invasive imaging of the functional β-cell mass. The exact epitope that IC2 recognizes on the surface of pancreatic β-cells is unknown. Determination of the major IC2 binding epitope is an important task, and a task, that has to be completed if IC2 is to be used as a biomarker in clinical settings. Thus, the aim of this study has been to: 1. Examine the strength of interactions between IC2 and intact cells, plasma membranes, monosaccharides and lipid targets. 2. Investigate the affinity of interactions of the various IC2 formats towards intact β-cells.

1.1. Type 1 diabetes 1.1.1. A chronic autoimmune disorder T1DM is a chronic disorder resulting from autoimmune destruction of the insulin-producing pancreatic βcells in the islet of Langerhans. T1DM is characterized by a marked and progressive inability of the pancreas to produce and secrete insulin because of autoimmune destruction of the β-cells. As β-cell mass declines due to ongoing immune destruction, insulin secretion eventually decreases until the available insulin is no longer adequate to maintain normal blood glucose levels. After approximately 80-90 % of the β-cells are destroyed, hyperglycemia develops and diabetes may be diagnosed (Van Belle et al., 2011)(Van Belle et al., 2011)(Van Belle et al., 2011)(Van Belle et al., 2011). When the destruction is sufficient to reduce and eventually completely eliminate insulin production, it leads to hyperglycaemia and several accompanying complications in affected individuals (Khardori et al., 2007, Van Belle et al., 2011, Di Gialleonardo et al., 2012). The complications of diabetes include hyperglycemia, increased risk of infections, microvascular complications (such as retinopathy and nephropathy), neuropathic complications and macrovascular disease causing heart disease and stroke (Khardori et al., 2007, Van Belle et al., 2011). Affected individuals are dependent on exogenous insulin. Since the 1920s, diabetes has been treated by insulin replacement which, in the ideal case, only shortens life expectancy by app. 10 years. Even more recent technology such as continuous glucose monitoring and slow release insulin further reduces the risk of life-threatening

14

introduction

hypoglycaemic events from insulin overdose. Therefore a future immune based intervention should ideally be effective, long-lasting and have minimal side effects to replace insulin treatment with a cure (Van Belle et al., 2011). To study T1DM, murine animal models that become diabetic spontaneously have been developed. The most widely used is the non-obese diabetic (NOD) mice which are genetically predisposed in multiple loci to develop T1DM at an incidence of 60-80% for females and 20-30% in males (Anderson and Bluestone, 2005). Another murine model is the diabetes prone (DP) biobreading (BB) rat. Dependent on the individual colony, around 90% of DP-BB rats develop T1DM equally in both sexes (Whalen et al., 2001, van den Brandt et al., 2010). Globally, the T1DM incidence has been rising during the past decades. If the present trend continues, a doubling of new cases of T1DM in European children younger than 5 years is predicted for 2005-2020. Also the prevalence of cases in individuals younger than 15 will rise by 70 % which implies that the onset of T1DM in general moves towards an earlier age (Van Belle et al., 2011). The increase in incidence is faster than what could be accounted for by genetic change, and this is not due to improved diagnosis (Zaccone and Cooke, 2013b).

1.1.2. Cause of type I diabetes The cause of T1DM is thought to be a multiplicity of environmental factors, such as certain viral or bacterial infections, nutrition or chemical factors, in genetically susceptible individuals leading to an imbalanced Tcells distribution. The wide gap between initiation and detection of T1DM poses a problem in the search for what triggers the disease. Another complicating factor for the determination of the cause of T1DM is that it might require multiple environmental factors to set off autoimmunity (Van Belle et al., 2011).

1.1.2.1. T-cells T1DM is generally believed to be a T-cell dependent disease and if the T-cells are inactivated like in athymic nude mice, T1DM will not develop. There is accumulating evidence for the presence and functionality of HLA(human leukocyte antigen)-A*02 -restricted CD8+ T-cells reacting against antigens on β-cells. Transgenic mice have been generated expressing HLA-A*02 molecules and their accelerated diabetes onset provides functional evidence for the involvement of this particular class I allele (Marron et al., 2002).

1.1.2.2. Genetics Patients with T1DM are thought to have genetic susceptibility to development of the disease, however concordance rates between monozygotic twins are only 50 % (dizygotic twins are 10 %), and there are strong divergencies in terms of the time it takes to develop diabetes (Van Belle et al., 2011). It seems that genetic susceptibility persists for life and progression to diabetes is usually preceded by a longer period of anti-islet autoantibody expression (Van Belle et al., 2011). Exposure to a trigger of viral or environmental origin seems to stimulate immunologically mediated destruction of the β-cells (Atkinson and Eisenbarth, 2001). Autoimmune diabetes is only rarely of monogenic forms, in which case diabetes is caused by mutational defects in a single gene. When this is the case, diabetes is typically accompanied by multiple other autoimmune conditions due to a disruption of common pathways (Liston et al., 2003, Van Belle et al., 2011).

15

introduction

1.1.2.3. Viruses and bacteria One of the most prominent type 1 diabetes linked viruses is the coxsackie virus B, an enterovirus belonging to the picornavirus family (Van Belle et al., 2011). Extensive circumstancial data states enteroviruses and more specifically coxsackie viruses, as main viral candidates that can cause T1DM. It is a fact though, that given the current knowledge on genetic contribution, viral infection alone does not cause T1DM. It is believed that β-cell tropic infection upregulates both HLA class I and IFN-α and thereby leaves a molecular viral signature. This could explain why the immune response is directed specifically against the islets (Van Belle et al., 2011). Data from the genetically T1DM prone NOD mice show that existing insulitis is required for coxsackie virus to induce diabetes. Translating this to the human situation, susceptible individuals may have an ongoing subclinical insulitis for years until viral challenge triggers an acceleration of β-cell destruction and hyperglycemia. Other viruses have been associated with T1DM but the causal relation has not been proven (Van Belle et al., 2011). The bacterial composition of the intestine has long been acknowledged as an important variable affecting T1DM development. Both in T1DM models and in patients, the intestinal wall does not seem to have the same capacity to form a coherent barrier separating luminal bacteria and the immune system as compared to controls. This so-called ‘leaky-gut’ phenotype is thought to enhance the exposure of bacterial antigens to the immune system (Van Belle et al., 2011). Among other potential environmental triggers are potential agents to cause leaky gut such as cow’s milk (especially its albumin component), wheat proteins such as gluten, and lack of vitamin D (Van Belle et al., 2011).

1.1.2.4. Polyclonal activation Based on the increased incidence of not only T1DM as described earlier, but also of allergy and autoimmune conditions such as systemic lupus erythematosus, an alternative hypothesis to the occurrence could be a lack of early exposure of infections to our immune system. Our immune system has co-evolved with infectious agents, and it is thought that the lack of interplay between certain infectious agents and our immune system can contribute to the observed increase of autoimmunity (Zaccone and Cooke, 2013b), serving as a possible explanation of why the incidence of autoimmune diseases is higher in developed countries. The explanation for this is the Hygiene Hypothesis (Strachan, 1989). This hypothesis depicts that the human immune system, by virtue of many years of exposure to and thereby also co-evolution with ubiquitous pathogens, is adapted to the exposure of these pathogens and that their immunomodulatory products, has helped to condition the polarization of responses against other foreign pathogens. With the increasing cleanliness of our surrounding environments we are no longer exposed to these infections during periods of life that is critical for tuning the immune system, which allows hypersensitivity and reactions to develop against self antigens and harmless environmental substances, thus leading to a greater likelihood of allergic and autoimmune responses (Murphy et al., 2008). This ability to downregulate the inflammation would be a benefit for both host and parasite (Zaccone and Cooke, 2013b). It has been proposed that deliberate infection with certain infectious agents early in life, could prime the immune system and thereby affect the immune system to an improved regulation of inflammatory responses (Zaccone and Cooke, 2013b). A deliberate infection with parasitic worms, or helminths, has shown to have exactly this effect (Cooke, 2012). In NOD mice, infection with helminths was shown to actually prevent autoimmune disease (Cooke et al., 1999, Saunders et al., 2007). Later it was shown that infection with helminths could also ease the course of illness in other murine models with induced

16

introduction

autoimmune diseases, such as experimental autoimmune encephalomyelitis, Graves’ hyperthyroidism, and collagen induced arthritis (La Flamme et al., 2003, Nagayama et al., 2004, Osada et al., 2009). Helminth infections affect many different immune cells including innate and regulatory cell populations as well as epithelial cells. Thus to better understand the molecules and the complex interactions involved in prevention and amelioration of autoimmune diseases with infectious agents such as helminths, more studies are required. During helminth infections, regulatory T-cells (Tregs) as well as other regulatory cell types such as regulatory B-cells (Bregs) and natural killer T-cells (NKT-cells) are induced. These cell types control the responses of T-helper cells (Th), which stimulate cytotoxic T-cells (Zaccone and Cooke, 2013b). Helminth products are a complex mix of glycoproteins, proteins, glycolipids and polysaccharides able to modulate signaling through pathogen recognition receptors on the host’s innate cells. This induces major changes which drive the excretion of type 2 T-helper cell (Th2) response as well as the expansion of regulatory cell populations (Zaccone and Cooke, 2013b). Especially Tregs and type I NKT-cells are thought to be implicated in the helminth mediated prevention of T1DM. Impairment of immunoregulatory mechanisms of these cells might result in disproportional activity of diabetogenic T-cells with the consequence of breakdown of peripheral tolerance and resulting autoimmunity (Zaccone and Cooke, 2013a). Early work showing that helminth infections can prevent type 1 diabetes was based on the hypothesis that the Th2 cell bias, as induced by infection, would control the type 1 T-helper cell (Th1)-mediated autoimmune diseases such as diabetes (figure 1). In recent years data generated from the NOD mouse has added complexity to the simple idea of Th2 shift as prevention of Th1-autoimmunity, by suggesting that other cytokines are responsible for the diabetes response (Zaccone and Cooke, 2013a).

Figure 1. Model proposed by Zaccone and Cook for the involvement of Tregs and NKT-cells in helminth antigenmediated diabetes protection. In T1DM prevention, immunity to helminth antigens involves several classes of immune cells. B-cells, tolerogenic dendritic cells (DCs), alternatively activated macrophages and innate lymphoid cells provide the right cytokine environment to induce Th2, NKT and Treg expansion. In turn, cytokine secretion from NKT, Th2 and Tregs feeds back and maintains the tolerogenic and alternatively activated phenotypes of antigen presenting cells. As a result, Th1 (diabetogenic) cells become regulated, inducing only a non-destructive insulitis. Figure from (Zaccone and Cooke, 2013a).

17

introduction

1.1.3. Silent immune events Several silent immune events occur in advance of clinical symptoms of T1DM. Importantly, autoantibodies are produced which reflects the consequence of an underlying sustained autoimmune process (Van Belle et al., 2011). IA2 (islet-cell antibody) and IAA (insulin antibody) levels as well as autoantibodies to GAD (glutamic acid decarboxylase) are all indicators of type 1 diabetes status (Khardori et al., 2007). Elevated serum concentrations of the glycolipid lyso-phosphatidylcholine precede the appearance of each autoantibody (Van Belle et al., 2011). Furthermore, self-reactive lymphocytes become activated and infiltrate the pancreas (insulitis) to destroy the β-cells. This persistent targeted destruction may go undetected for years, hence resulting in the clinical symptoms not being apparent until a majority of the βcells have been destroyed or turned dysfunctional (Van Belle et al., 2011). To get a complete picture of the mechanisms behind T1DM, the anatomy and structure of the pancreas including a description of the endocrine pancreatic cells such as β-cells are described in the following sections. Also information regarding the production, secretion and actions of the peptide hormone insulin will be introduced.

1.2. Pancreas and the Islet of Langerhans 1.2.1. Anatomy The pancreas is located behind the stomach and the parietal peritoneum and is joined to the duodenum with the pancreatic duct through which it transports digestive juice consisting of a mixture of different enzymes to the intestines. The pancreas consists of two types of secretory tissue and thus has a dual function as being both an exocrine gland, and an endocrine gland that produce and release important polypeptides and regulatory hormones. The endocrine portion of the pancreas consists of groups of cells that are closely associated with blood vessels. These cells, called α-, β-, δ- and PP(polypeptide-producing)cells, are arranged in between each other in groups of cells called islets of Langerhans or pancreatic islets (figure 2)(Shier et al., 2009, Skelin et al., 2010).

Figure 2. The pancreas is joined to the duodenum with the pancreatic duct through which it transports digestive juice. The pancreas consists of two types of secretory tissue exocrine and endocrine. The endocrine pancreatic cells are closely associated with blood capillaries. These cells, called α-, β-, δ- and PP-cells, are arranged in between each other in groups of cells called islets of Langerhans or pancreatic islets. Modified from (Efrat and Russ, 2012, Webpage, http://www.jltacademystudents.com/).

18

introduction

All endocrine pancreatic cells are evolved from the same lineage and thus share many morphologic traits. Their functions are different, but all of high importance. The peptide hormone insulin is produced and secreted by β-cells. The insulin secreting β-cells make up the primary part of the islet of Langerhans (65-90 %) and are also the most frequently studied. The β-cells are stimulated to proliferation as well as destruction both by mitogenic signalling and glucose. Glucose thereby affects and maintains the adequate amount of β-cells that are important for overall regulation of metabolism. The α-cells make up 15-20 % of the islet of Langerhans. α-cells produce and release the peptide hormone glucagon which affects the liver to perform glycogenolysis (convert glycogen to glucose) and later glyconeogenesis. It thereby raises the blood glucose level and thus has the adverse effect as that of insulin (Shier et al., 2009). 3-10 % of the islets of Langerhans consists of somatostatin-producing δ-cells, and around 1 % is PP-cells (Skelin et al., 2010).

1.2.2. Insulin 1.2.2.1. Production of insulin Insulin is a vital hormone since it promotes facilitated diffusion of glucose across the membrane of certain cells throughout the body (Shier et al., 2009). The biosynthesis of insulin occurs in the endplasmic reticulum (ER) of β-cells from the precursors preproinsulin and proinsulin (Ashcroft and Ashcroft, 1992). Upon cleavage of the signal sequence from preproinsulin, proinsulin is transported to the golgi where it is mounted as hexamers in zinkand calcium-rich secretory vesicles for temporary storage. Inside the secretory vesicles, the C-peptide is cleaved from the proinsulin and it is thereby converted to insulin. This removal of the C-peptide significantly alters the solubility of the insulin, causing crystallization of the insulin hexamer inside the vesicle (figure 3)(Ashcroft and Ashcroft, 1992). This crystallized insulin hexamer is an inactive storage form that must be converted to monomers for biological Figur 3. Structure of an insulin hexamer activity. The hexamer formation and crystallization seems to have crystallized with zinc (Chang et al., the effect of stabilizing the insulin thereby preventing the 1997). degradation within the storage vesicles (Dunn, 2005)

1.2.2.2. Secretion of insulin Glucose is the principal physiological insulin secretagogue and a potent regulator of β-cell activity, because insulin is produced and secreted by pancreatic β-cells in response to an increase in extracellular glucose levels. This increase in extracellular glucose levels results in production of ATP (adenosine triposphate) which closes the ATP sensitive potassium ion channels (K+/ATP). This causes a depolarization of the plasma membrane and a subsequent opening of the voltage-gated calcium channels causes an influx of calcium amplified by depolarization of the endoplasmic reticulum, which results in the transport of insulincontaining vesicles to the cell membrane and the subsequent release of insulin (figure 4). Only a fraction of the stored β-cell insulin is released during stimulation (Rorsman and Renstrom, 2003, Skelin et al., 2010). The release of insulin occurs as the vesicle membrane fuses with the plasma membrane of the β-cell. The insulin crystal is released into the intracellular space where it dissolves to insulin monomers (Dunn, 2005). A group of proteins referred to as SNARE (soluble NSF attachment protein receptor) proteins play an

19

introduction

important role in membrane fusion as they facilitate vesicle transport as well as exocytosis by bringing the vesicle membrane in close contact with the plasma membrane in a way analogous to a zipper. The conformational change occurring during this process is thought to provide the energy required for membrane fusion (Rorsman and Renstrom, 2003). Figur 4. Secretion of insulin. An increase in extracellular glucose activates the GLUT2 (glucose transporter 2). The glucose causes an increase in intracellular ATP concentration, closing ATP + sensitive K /ATP. This causes a depolarization of the plasma membrane and a subsequent opening of voltage-gated calcium channels causes an influx of calcium. The calcium levels are amplified by depolarization of ER. The increased calcium concentration results in the transport of insulin-containing vesicles to the cell membrane and the subsequent release of insulin.

1.3. IC2 1.3.1. The monoclonal autoantibody IC2 The monoclonal autoantibody, IC2, has a unique specificity for the insulin secreting pancreatic β-cell (figure 5)(Brogren et al., 1986). It is interspecies specific in multiple species including human (Brogren, 1986) and has been shown specific to pancreatic β-cells both in vitro and in vivo (Brogren et al., 1984, Poussier et al., 1984, Brogren et al., 1986, Buschard et al., 1988, Aaen et al., 1990, Moore et al., 2001, Bronsart et al., 2011, Kavishwar et al., 2011). The most important findings regarding IC2 are described here and briefly summarized in table 2 at the end of this section. IC2 was raised from a diabetes-prone BB-rat by screening against RIN-5F insulinoma cells (Brogren and Andersen, 1983, Brogren et al., 1986). Initially, the relevant IC2-producing hybridoma was selected by screening the culture supernatants on ELISA (enzyme-linked immuno sorbent assay) plates coated with isolated native plasma membranes of RINm5F cells and further characterized by ELISA using intact islet cells and immunofluorescence studies on various tissues and cells (Brogren, 1986). Interestingly, the expression of the IC2 autoantigen shows a dependency to the functional state of the β-cell with decreased expression following fasting or insulin administration, which, however, might be a general feature for many cell surface antigens (Buschard et al., 1988, Aaen et al., 1990). Since β-cells release insulin when glucose stimulated, it is plausible that the increased expression of IC2 autoantigen following glucose stimulation could somehow be correlated to the production, transport or release of insulin (figure 6).

20

introduction

This β-cell specificity was first shown by immuno fluorescence and electron microscopy (Brogren et al., 1986). The affinity of the interaction between radioisotope labelled IC2-IgM and pancreatic β-cells Figure 5. Immunoperoxidase Figure 6. Confocal immunofluorescence was measured using cellular cytochemical staining with IC2 on an microscopy with IC2 on and INS-1E cell. radio immunoassay (RIA) INS-1E cell population attached to a Confocal immunofluorescence LigandTracer (Ridgeview microscopic slide. Notice the microscopy of IC2 on the surface of a pronounced heterogeneity on single INS-1E β-cell. IC2 marks Technologies, New Hampshire, functionally expressed IC2 autoantigen autoantigen located in patches which are USA) and showed a mediumin the cell suspension. (Pedersen et al., associated with the cell surface lipid rafts 2011) (Pedersen et al., 2011). high affinity of an estimated dissociation constant (KD) of app. 18nM (Desai, 2009) (figure 7). The affinity of IC2 towards the pancreatic β-cell line INS-1E was measured by cellular quartz crystal microbalance (cellular QCM) in a pilot experiment. Here the interaction was estimated to a high affinity with a KD of 1.39 nM (figure 8)(Pedersen et al., 2011). Beside the ability of IC2 to bind to pancreatic β-cells, brand new data from flow cytometry studies show that IC2 recognizes a small population of pancreatic primary islet cells marked with the leukocyte common antigen CD45. CD45 is exclusively expressed on all haemotolymphoid cells, including precursor cells, mature B- and Tlymphocytes, and monocytes (Murphy et al., 2008). Also it was observed, that IC2 was found to recognize a percentage of rat thymocytes in the thymus, but not in thymus from mouse. The percentage of IC2 positive cells was found to decline with age from 32 % at birth to only 4 % at age 60 days (Poussier et al., 1986). Figure 7. Interaction between IC2 and intact cells measured by LigandTracer. LigandTracer affinity study of the 123 interaction between IIC2 tracer and INS-1E cells. 123 The I-IC2 tracer was added at increasing concentrations. The KDvalue was derived by the LigandTracer software using the Langmuir Interaction model (1:1) (Desai, 2009)

Figure 8. Interaction between IC2 and intact cells measured by cellular QCM. Pilot experiment showing IC2 interaction with INS-1E rat β-cells measured by cellular QCM using the Attana Cell 200 (Attana AB, Stockholm). INS-1E cells were fixated with formaldehyde after 120 h of growth on a COP-1 sensor. The kinetics for the interaction was derived from the sensogram. IC2 has a very slow off-rate and high association rate. The estimated affinity for the interactions is KD = 1.39 nM indicating a high affinity (Pedersen et al., 2011)

21

introduction

1.3.2. Imaging and various IC2 formats The unique properties of IC2 as a functional biomarker, labelling insulin-secreting β-cells with high specificity (Buschard et al., 1988, Aaen et al., 1990, Moore et al., 2001, Bronsart et al., 2011), makes it an obvious candidate for non-invasive imaging of the functional β-cell mass (Saudek et al., 2008, Di Gialleonardo et al., 2012). Non-invasive β-cell imaging using modern medical imaging such as single-photon emission computed tomography (SPECT) positron emission tomography (PET), magnetic resonance imaging (MRI) and photoacoustic tomography (PAT) could provide very high sensitivity in humans, but exhibits some difficulties. First, single-cell resolution cannot yet be achieved to enable differentiation between scattered islets, single β-cells, or surrounding tissue. Second, these imaging techniques rely on the existence of a molecular marker with high specificity, the lack of which most potential molecular markers for non-invasive determination of β-cell mass have suffered from in the past (Saudek et al., 2008, Di Gialleonardo et al., 2012). Part of this problem emanates from the fact that β-cells share the same lineage as other cells in the pancreas, making it difficult to elucidate unique targets on the β-cell surface (Kavishwar et al., 2011). Third, a challenging requirement for imaging agents to be retained by β-cells at least 100-fold more strongly bound than by exocrine cells is to be met (Sweet et al., 2004). In 2001, IC2 was used to conduct nuclear imaging biodistribution studies of the pancreatic β-cell mass in mice (Moore et al., 2001). This study was the first to exploit the properties of IC2, by proof of principle, for estimation of β-cell mass using an imaging approach. Even though the IC2-IgM affinity is applicable for imaging purposes, the IgM format is large (900 kDa) causing the excess of unbound tracers to circulate in the blood for days and weeks. Fragments of the IC2 antibody (F(ab)'2, Fab, scFv & diabodies) have been produced in order to reduce kidney clearance time. Also a chimeric humanized recombinant IC2-IgG molecule was produced (rhIgG) (figure 9). Table 1 provides an overview of the current native and recombinant IC2-antibody formats. Figure 9. IC2 formats. The native IC2 format is an IgM antibody. Proteolytic fragments of IC2, F(ab’)2 and Fab has been produced. A recombinant diabody and scFv has been produced. Also a recombinant chimeric humanized (rhIgG) IgG antibody was produced. The sizerelationship of the molecules in the figure is not authentic. Modified from (Peer et al., 2007). Table 1. Overview of current IC2 antibody formats Native

App. size

Site of production

Recombinant

App. Size

Site of production

IC2-IgM

900 kDa

Bartholin Institute, Copenhagen

rhIgG

150 kDa

Modiquest, Nijmegen

F(ab')2

110 kDa

Bartholin Institute, Copenhagen

Diabody-Gluc

70 kDa

Stanford School of Medicine, Palo Alto

Fab

50 kDa

Bartholin Institute, Copenhagen

Diabody (Fv)2

50 kDa

Stanford School of Medicine, Palo Alto

scFab

50 kDa

Millegen, Toulouse

scFv

25 kDa

Millegen, Toulouse

22

introduction

In 2010 a biodistribution study in rats using 125I-IC2-IgM showed a factor 12 in pancreas specificity for IC2IgM, F(ab’)2 and the Fab format (Wismann, 2010). In 2011, an engineered diabody format of IC2 coupled to Gaussian luciferase (GLuc) served as biomarker in a non-invasive bioluminescent (BLI) study of the β-cell mass in mice with factor 3 in pancreas specificity in vivo (figure 10)(Bronsart et al., 2011). The substrate used has later been known to be internalized by certain types of cells (Bronsart, 2012). Therefore, what we observe here could possibly be the result of internalization of substrate of all types of pancreatic cells and thus not imaging of the β-cells only.

Figure 10. Non-invasive IC2-GLuc imaging in mice. Non-invasive imaging and bioluminescent biodistribution studies in mice using a recombinant engineered IC2 diabody-GLuc fusion protein. The biodistribution showed explicit bioluminescence from the isolated pancreas (Bronsart et al., 2011).

In 2012, the F(ab’)2 and Fab fragments were used for near infrared imaging of the pancreatic β-cell mass by fluorescence molecular tomography (FMT) in mice for the first time in a collaboration between Animascope (Geneve, Switzerland) and the Bartholin Institute (Sefeld et al., 2013). Though a distinct signal was seen from both liver and stomach, the quantification of the fluorescence showed that the IC2-F(ab’)2 fragment was 15 times more specific to the pancreatic β-cells than the rat-IC2 control. The IC2-Fab fragment was an astonishing 26 times more specific to the pancreatic β-cells than the rat-IC2 control and thus the most specific of the IC2-formats measured so far. This is very convenient since the Fab-fragment is also the smallest of the formats tested and thus size wise the most applicable for future imaging. The poor penetration of near infrared (NIR) light makes the imaging applicable only in smaller animals and this study cannot readily be applied to larger animals. A project which aims to use IC2 and fragments as biomarker using PAT imaging is now being initiated. For clinical use PET is preferred.

1.3.3. The IC2 autoantigen The exact epitope for IC2 on the pancreatic β-cell surface still remains largely unknown. Already in the 1980’s it was found, that the IC2 target was trypsin-sensitive, suggesting that a protein is involved in the binding (Brogren et al., 1984, Poussier et al., 1984, Brogren et al., 1986). Surprisingly, in 1989 another target was found to be of lipogenic nature and it was further characterized by immuno thin layer chromatography (iTLC) as a sulfatide (figure 11) (Brogren et al., 1989, Spitalnik, 1989). Together these findings suggested that the epitope of IC2 could be a protein-lipid complex constituted on the surface of the β-cell. This initiated a search for sulfatide binding proteins on the surface of β-cells.

23

introduction

For a while it was thought that the IC2 autoantigen could somehow be related to the ZnT8 (zinc transporter 8) structure, because Zn2+ ions would have a strong affinity to the negative sulfatides involved in insulin packaging in secretory vesicles, but then it was shown by double immunostaining in confocal microscopy that the IC2 plasma membrane specificity and the anti-ZnT8 insulin granula specifity does not overlap with each other (figure 12)(Spitalnik, 1989). In the search for the protein counterpart to the sulfatide epitope, the focus then turned to certain t-SNARE proteins and other β-cell specific cell surface Figure 11. Sulfatides characterized by iTLC. Binding expressed proteins which could involve sulfatides. Since IC2 of monoclonal antibody IC2 to sulfatides: only recognizes the functional insulin producing cells, and Glycolipids were analyzed by iTLC. Lanes 1-3 were stained with orcinol. Lanes 4-7 were since lipids are known to be involved in the release of insulin immunostained with IC2. Lanes 1, 4-6: galactosyl (Fox and Kester, 2010), and sulfatides in the packaging and sulfatide (2, 1, 5 and 10 μg, respectively). Lanes 2, 7: lactosyl sulfatide (2 μg). Lane 3: human brain release of insulin (Osterbye et al., 2001), the theory of a acidic glycolipid standards (Spitalnik, 1989). protein-lipid complex on β-cell surfaces indirectly coupled to the production or release of insulin, could perhaps explain the current findings. In a recent study by Kavishwar et al., sphingomyelin, phosphatidylcholine, and possibly the lyso-forms of these, were found to be involved in the β-cell specific binding of IC2 (Kavishwar et al., 2011). The trypsin Figure 12. Immunostaining with IC2 and ZnT8 antibody. Confocal microscopy imaging of sensitivity, however, was not the top of the cells with double indirect immunostaining for ZnT8 (red) and IC2 detected in this study. In autoantigen (green) on permeabilized RIN-5AH cells: The IC2 autoantigen is expressed exclusively on the β-cell surface, while the ZnT8, which is located in the membrane of addition it was shown that the the insulin vesicle, is distributed equally throughout the cytoplasm. Antibody capping loss of cholesterol leads to a and lipid rafting of the IC2 monoclonal autoantibody is observed (Spitalnik, 1989). loss of IC2 binding, which indicates that the binding to the autoantigenic epitope is somehow coupled to cholesterol as well (Kavishwar et al., 2011). This could indicate a connection to plasma membrane lipid rafts, which is supported by the binding of IC2 in ‘patches’ as seen on the confocal immunofluorescence IC2 staining of a single pancreatic β-cell (figure 6) (Lang et al., 2004, Pedersen et al., 2011). While Kavishwar et al. payed special attention to the determination of IC2 autoantigen, the functional aspects of IC2 were investigated in a pilot experiment in San Diego, California. In a modified functional NKThybridoma assay (Roy et al., 2008) IC2 was found to inhibit interleukin 2 (IL-2) secretion of type I and possibly also type II NKT-cells. In this experiment, microwell plates were coated with soluble CD1d molecules and to wells were added the glycolipids sulfatide or α-galactosyl-ceramide (α-GalCer) either directly, or after a period of preincubation. Then NKT-cells were added and the IL-2 concentration measured. The inhibition was shown to be most pronounced when IC2 was preincubated with the glycolipid before it was added to the CD1d coated plate showing a striking 98 % inhibition of the IL-2 secreted by the type I NKT-cells (appendix II)(Jensen et al., 2012). These results strongly suggest the

24

introduction

involvement of the plasma membrane located CD1d receptor which is known to potently stimulate NKT-cell cytokine release (Godfrey and Kronenberg, 2004). The CD1d molecule has a deep hydrophobic antigenbinding cleft in which lipid antigens are loaded. It is essential for stimulation of NKT-cells that glycolipids and the CD1d molecule form an epitope accessible for the NKT T-cell receptor (TCR) binding. Our theory is that IC2 recognizes a certain part of this CD1d-lipid assembly and thereby blocks binding of the NKT TCR ligation hindering NKT stimulation. The different lipids that IC2 seems to recognize could act as minor- and major targets for IC2 recognition. Table 2 below provides selected land mark references in relation to the study of IC2 in chronological order. Table 2. Chronological overview of Important land marks in the history of IC2 Important findings

Reference

Review of monoclonal antibodies in diabetes research.

(Brogren and Lernmark, 1982)

BB-rat rat-rat hybridoma monoclonal autoantibody methods established.

(Brogren and Andersen, 1983)

IC2 was presented for the first time as a monoclonal autoantibody from BB-rat with specificity to the (Brogren et al., 1984, Poussier et al., pancreatic β-cell surface 1984) The original IC2 paper. IC2 target is found to be trypsin sensitive and surface specific to islet cells

(Brogren et al., 1986)

Interspecies specificity is proven to multiple species including human islet cells.

(Brogren, 1986)

Minor subpopulation of rat thymocytes reactive with IC2.

(Poussier et al., 1986)

IC2 is found to be specific to functional β-cells only. Functional stage dependency of the autoantigen (Buschard et al., 1988) expression. Explicit specificity of IC2 to mouse and rat islet β-cells. The role of polyclonal activation in autoimmunity and sulfatides are described as part of a trypsin sensitive (Brogren et al., 1988) IC2 autoantigenic epitope. Terminal galactose-3-sulphate as autoantigenic determinant for the IC2 monoclonal antibody.

(Brogren et al., 1989)

Explicit pancreatic β-cell specificity of IC2, and electron microscopic evidence for explicit plasma membrane (Aaen et al., 1990) anchored autoantigen expression. IC2 used for SPECT biodistribution studies in mice.

(Moore et al., 2001)

Because of its specificity, IC2 is thought to be the best suitable candidate for in vivo imaging.

(Saudek et al., 2008)

Pancreatic cell line and plasma membrane lipid IC2 autoantigen studies by iTLC .

(Mia, 2009)

Dissociation constant of IC2-IgM towards β-cells are estimated to KD=18nM.

(Desai, 2009)

Proteolytic fragmentation of IC2–IgM, creation of F(ab’)2 and Fab fragments.

(Gao, 2010)

Rat biodistribution studies with

125

I-IC2,

125

I-F(ab’)2,

125

I-Fab

125

I-A2B5 and

125

I-K14D10.

Biodistribution and pancreas islet β-cells specificity in vivo.

(Wismann, 2010) (Brogren et al., 2010)

Serological titration of IC2 autoantibodies in human and BB-rat serum measured by competitive ELISA on (Jacobsen, 2010, Jørgensen, 2010) RIN-5AH plasma membrane coated plates. IC2 specificity shown non-invasively in vivo by BLI using diabody-GLuc injected in mice.

(Bronsart et al., 2011)

Sphingomyelin, phosphatidylcholine and the lyso-forms of these are proposed as antigen for IC2 with (Kavishwar et al., 2011) cholesterol af stabilizing agent. IC2 is in preliminary data found to inhibit stimulation of NKT type I and perhaps also NKT type II.

(Jensen et al., 2012)

Project to develop an antiidiotypic antibody against IC2 is initiated.

(Jensen et al., 2012)

A review of imaging of the β-cell mass in T1DM describes IC2 as an ideal imaging candidate.

(Di Gialleonardo et al., 2012)

CD1d found by flow cytometry (FCM) to be expressed on the surface of β-cell lines.

(Voetmann, 2013)

Using FCM, IC2 is found to bind to CD1d transfected macrophage and B-lymphoma cells and not to their Unpublished data by non-transfected counterparts. Engmose Pedersen, 2013 Imaging of the pancreatic β-cell mass with IC2 using NIR imaging.

(Sefeld et al., 2013)

A literature review of the role of Type II NKT-cells as regulatory cell in autoimmune diabetes.

(Sørensen et al., submitted)

Christina

Attempt to elucidate the antigen of IC2 by QCM and microscale thermophoresis (MST) studies of the binding Present study, 2013 of IC2 to lipids, plasma membranes and β-cells.

25

introduction

1.4. NKT-cells and CD1d 1.4.1. Natural killer T-cells (NKT-cells) NKT-cells are a subset of T-cells that share some characteristics with natural killer (NK) cells such as expressing NK lineage receptors, along with expressing an invariant T-cell receptor specific for CD1d(Godfrey et al., 2004, Bendelac et al., 2007, Girardi and Zajonc, 2012). NKT-cells are reactive to the major histocompatibility complex (MHC) class-I-like molecule CD1d. They are stimulated by a CD1d-lipid complex when their TCR interacts with this complex. The CD1d-lipid-TCR complex can be formed with a relatively wide palette of different lipids dependent on NKT-cell subset (Brigl et al., 2003). These lipids can be of both microbial and endogenous origin (Girardi and Zajonc, 2012). The precise distribution of NKT-cells in the lymphoid organs are unknown but in mice they reside in large numbers within the liver sinusoids which implies that they function as a first line of defense (Bendelac et al., 2007). An issue that is still unresolved is the identity of the lipid ligands presented in CD1d that drive the selection of NKT-cells during their thymic development. The same, or different, ligands may also be presented by antigen presenting cells and contribute to activation of NKT-cells during microbial infections and autoimmunity (Brigl et al., 2003). Stimulation of NKT-cells makes them release a wide variety of cytokines which dependent on the length of the lipid chain, primarily drives a Th2 or Th1 response. The mechanisms underlying the Th1 versus Th2 response are debated and may be diverse (Bendelac et al., 2007, Sørensen et al., submitted).

1.4.2. NKT subsets NKT-cells are divided into three subsets called NKT-like cells, the NKT type I (NKT I) and NKT type II (NKT II) cells (Godfrey et al., 2004). The NKT-like subset of cells are CD1d independent NK1.1 cells and will not be discussed further here (Godfrey et al., 2004). The most studied and thus best characterized of the subsets, is the NKT I subset. It is characterized by the expression of an invariant T-cell receptor in which the α-chain is invariant (Vα14-Jα18 in mice and Vα24-Jα18 in humans), why they are also known as invariant NKT-cells (Godfrey et al., 2004). These invariant NKT-cells recognize CD1d when it forms complexes with the selfantigen isoglobotrihexosylceramide and non-mammalian lipids such as the α-GalCer (figure 13). The most pronounced stimulation of NKT I cells is achieved with α-GalCer which has been used intensively to study the NKT I cells (Bendelac et al., 2007). In addition to glycosphingolipids, other classes of lipids such as phospholipids and diacylglycerol-based microbial lipids are also ligands for type I NKT-cells (Rhost et al., 2012). The NKT II subset includes another CD1d dependent T-cell population. (Godfrey et al., 2004). The self- and foreign antigens recognized by NKT II cells remain to be identified, but sulfatide is known to stimulate these NKT-cells (figure 14)(Bendelac et al., 2007). Since sulfatide-reactive NKT-cells remain in mice lacking sulfatide (Jahng et al., 2004) other CD1d-restricted ligands must be involved in the thymic selection as well as the peripheral activation of sulfatide-reactive type II cells. Also lyso-sulfatide and β-galactosyl-ceramide (β-GalCer) have been shown to activate NKT II cells when complexed with CD1d. Thus, the same NKT II cells showed a certain degree of promiscuity as they were activated by several ligands. Lyso-forms have been shown to be the more potent isoforms of the glycosphingolipid ligands. Lyso-form activated NKT II cells can be blocked by anti-CD1d antibodies demonstrating a CD1d dependent activation (Rhost et al., 2012).

26

introduction

Figure 13. Antigen recognition by the NKT I TCR. The complex between CD1d (grey) and α-GalCer (yellow) forms an epitope recognized by the TCR of NKT I cells. The NKT I TCR interacts with the cyan, green and orange regions of the CD1d (Girardi and Zajonc, 2012).

Figure 14. Antigen recognition by the NKT II TCR. The complex between CD1d (grey) and lyso-sulfatide (yellow) forms an epitope recognized by the TCR of NKT II cells. The NKT II TCR interacts with the purple, blue and metallic regions of the CD1d (Girardi and Zajonc, 2012).

A major difference between NKT II and NKT I cells is that NKT I cells require endosomal trafficking of CD1d and intact lysosomal functions for presentation at the cell surface, whereas NKT II ligands are normally presented by a tail-truncated CD1d, which is defective in endosomal trafficking and therefore likely presents antigens loaded in the secretory pathway (Chiu et al., 1999).

1.5. CD1d 1.5.1. The CD1d molecule CD1d is a MHC class I-like molecule which bind and present lipid antigens from self or microbial origin to NKT-cells. It belongs to the family of CD1 protein isoforms which differs in sequence and 3D structure so that their heavy-chains form antigen-binding grooves of differing size and shape (Moody and Porcelli, 2003). The CD1d molecule forms a binding groove beneath two anti-parallel α-helices placed above a platform constituted of eight-stranded β-sheets which is mainly lined by hydrophobic residues. The CD1d antigenbinding groove is characterized by two pockets denominated A’ and F’, which roughly corresponds to the position of the A and F pockets in MHC I molecules. The deeply buried A’ pocket adopts a unique donut-like shape, while the F’ pocket is rather straight and less deep. Each pocket can accommodate one alkyl chain from the antigen (figure 15) (Zeng et al., 1997, Gadola et al., 2002, Zajonc et al., 2003, Scharf et al., 2010, Girardi and Zajonc, 2012). Dependent on the characteristics of the bound lipid, such as length and number of alkyl chains, charge etc., a unique epitope will form. The solvent accessible parts of the CD1d and this epitope are able to serve as binding site for the TCR of certain NKT-cells and can affect the response from NKT-cells after T-cell ligation (figure 16). CD1d is constitutively expressed by antigen presenting cells such as dencritic cells, macrophages and Bcells. CD1d is also expressed on cortical thymocytes, where it is essential for NKT-cell development (Bendelac et al., 2007). Recently, analysis by flow cytometry (FCM) showed that the β-cell lines RIN-5AH and βTC-tet express CD1d, which according to the investigator Voetmann could be general for pancreatic βcells (Voetmann, 2013).

27

introduction

Figure 15. Top view looking into the CD1d in complex with sulfatide. The CD1d molecule (purple) forms a binding groove beneath two anti-parallel α-helices placed above a platform constituted of eight-stranded β-sheets. The sulfatide (yellow) fatty acid is bound in the larger A’ pocket, and the sphingosine backbone is bound in the F’ pocket. The deeply buried A’ pocket adopts a unique donut-like shape, while the F’ pocket is rather straight and less deep. Modified from (Zajonc et al., 2005).

Figure 16. Front view of the CD1d in complex with sulfatide. Dependent on the characteristics of the bound lipid (yellow), such as length and number of alkyl chains, a unique epitope will form between CD1d (purple) and the bound lipid. The solvent accessible parts of the CD1d and this epitope are able to serve as binding site for the TCR of certain NKT-cells. Modified from (Zajonc et al., 2005).

1.5.2. CD1d folding and binding of lipids CD1 proteins initially fold in the ER, exit through the secretory pathway, and traffic to endosomes via direct or surface mechanisms that require the binding of cytoplasmic sequences to adaptor proteins. These subcellular trafficking mechanisms expose the nascent CD1 proteins to a wide variety of diverse endogenous cellular lipids in ER, Golgi, and endosomal compartments (Moody and Porcelli, 2003, De Libero and Mori, 2007, Huang et al., 2011). CD1d rapidly reaches the plasma membrane within 30 minutes after biosynthesis and undergoes extensive internalization and recycling between the plasma membrane and endosomal compartments (Bendelac et al., 2007). Co-precipitation experiments of microsomal triglyceride transfer protein with CD1d suggest, that this protein may assist in the folding of CD1d by loading of lipids into the lipid binding groove (Bendelac et al., 2007). The CD1d protein in particular, is strongly influenced by cellular pH in the capture and release of lipids. It is relatively resistant to exogenous antigen loading at neutral pH conditions normally found in the secretory pathways, but exchange antigen more rapidly at pH 4-6 which is characteristic of the late endosomal or lysosomal environment. Whereas later antigen exchange events within endosomes are extensively characterized, the early events in which newly folded CD1 proteins load self-lipids are less well understood (Huang et al., 2011). Studies of mouse and human CD1d have lead to detection of several types of phospho- and sphingolipid ligands (Cox et al., 2009, Fox et al., 2009, Yuan et al., 2009), suggesting a great diversity of lipid ligands for CD1d. Early studies of CD1d immunoprecipitates obtained from cell detergent lysates suggested a predominance of phospholipids, but these results have later been questioned because of the methods used in these studies (Bendelac et al., 2007).

28

introduction

1.6. Classification of lipids 1.6.1. Simple, complex and derived lipids Lipids are classified into three main groups: simple, complex and derived lipids (table 3). The simple lipids are certain alcohols (such as glycerol), or esters of Table 3. Classification of lipids Complex lipids Derived lipids fatty acids (such as triacylglycerols). The complex Simple lipids Glycerol Lipoproteins Cholesterol lipids are esters of fatty acids or alcohols but also Triacylglycerol Glycolipids Retinol contain other groups, which is why this group of Neutral lipids Phospholipids Cholecalciferol lipids is also known as conjugated lipids. The derived Waxes Sphingolipids Ketone bodies lipids differ from the two others as they are High MW alcohols Sulfolipids Fatty acids Prostaglandins composed of hydrocarbon rings and a long Steroid hormones hydrocarbon side chain (Fahy et al., 2005, Fahy et al., Ecosanoids 2009).

1.6.2. Glycolipids Glycolipids have either a ceramide (glycosphingolipids) or a glycerol backbone (glycerolipids). Whatever backbone, the glycolipid consists of a hydrophobic tail region composed by fatty acid chains, and a hydrophilic head composed of monosaccharides. Thus, glycolipids are amphipatic (figure 17). Glycolipids can be hydrolysed into the lyso-form of the concerned lipid, by removing the fatty acid chain (Fahy et al., 2005). Glycerolipids are simple lipids named according to the number of fatty acid chains it contains. Monoacylglycerol has one fatty acid chain, diacylglycerol has two, and triacylglycerol has three. Glycerolipids with saccharides are called glyceroglycolipids and in cases where another functional group such as phosphate is incorporated in the glyceroglycolipid, it is called e.g. a glycerophospholipid (Fahy et al., 2005).

1.6.2.1. Sphingolipids and glycosphingolipids Sphingolipid are lipids consisting of a sphingosine backbone. When added a fatty acid is added to the backbone, it is also called a ceramide (figure 17). If a phosphate containing group is attached to the ceramide, it is called a sphingomyelin. If this functional group for example is phosphorylcholine, it is classified further as a ceramidephosphorylcholine. Sphingolipids containing a saccharide is called a glycosphingolipid.

Figure 17. Glycolipids are amphipathic and consists of a hydrophilic head group and a hydrophobic lipid tain region. Sphingolipids consists of a sphingosine backbone. If a fatty acid is added, it is called a ceramide. Sphingolipids containing a saccharide such as galactose is called a glycosphingolipid.

29

introduction

Glycosphingolipids are either neutral or negative at pH 7 and are differentiated hereafter (figure 18). The neutral glycosphingolipids are either ceramide monosaccharides (cerebrosides) or ceramide oligosaccharides (globosides). Cerebrosides have a single sugar moiety which can be either glucose or galactose. The two major types of cerebrosides are thus named glucocerebrosides and galactocerebrosides. The negatively charged glycosphingolipids can either be charged ceramide oligosaccharides containing one or more sialic acids (gangliosides) or sulfatides which are glycosphingolipids that contain a sulphate group. Thus, the sulfatides are cerebrosides with sulfate groups (Fahy et al., 2005, Fahy et al., 2009). Figure 18. A classification of glycosphingolipids. Glycosphingolipids with a monosaccharide is called a cerebroside. Dependent on the saccharide, these are either gluco- og galactocerebrosides. Glycosphingolipids with oligosaccharides are called globosides. Oligosaccharides with one or more sialic acids are called gangliosides. Glycosphingolipids with a sulphur group are called sulfatides. Glycosphingolipids can either be neutral (blue) or negatively charged (red) at pH 7.

1.6.3. Selected lipids 1.6.3.1. The galactosylcerebrosides α-GalCer and β-GalCer α-GalCer is a slightly modified version of the α-branched galactosyl-ceramide originally extracted from the marine sponge Agelas mauritianus (figure 19), thus it is not an endogenous lipid (Kobayashi et al., 1995). αGalCer has been found to be a potent stimulator of NKT I cells. As a surrogate ligand of very high activity both in vitro and in vivo, α-GalCer has been intensively used to study CD1d NKT I interactions (Bendelac et al., 2007). β-GalCer is a β-anomeric galactocerebroside (figure 19). In contrast to α-GalCer, this cerebroside is an endogenous lipid (Brown and Mattjus, 2007). It is known to bind to CD1d in smaller amounts and only weakly stimulates NKT-cells compared to α-GalCer (Zajonc et al., 2005).

1.6.3.2. Sulfatides Sulfatides such as, β-galactosyl-3’-sulfate ceramides, are negatively charged glycosphingolipids containing sulfate esters on their saccharide chain (figure 20). In mammals, sulfatides are present primarily in nervous tissue, testis, kidney, erythrocytes, platelets and granulocytes (Ishizuka, 1997). Furthermore, sulfatides are one of the major lipids in serum (Zhu et al., 1991). Sulfatides were among the first self-glycosphingolipids demonstrated to stimulate T-lymphocytes (Shamshiev et al., 1999). Sulfatide is associated with insulin and is present at the surface of β-cells (Buschard, 2011). Sulfatide decrease cytokine (Buschard et al., 1996) and chemokine (Roeske-Nielsen et al., 2004) secretion, thereby reducing the destructive actions of cytokines on β-cells. Sulfatide stimulates regulatory NKT-cells when in complex with CD1d (Jahng et al., 2004). The presence of sulfatide in the central nervous system, mainly in myelin, has shown to be interesting. Myelin is a target of the autoimmune process during multiple sclerosis. In multiple sclerosis it was found that sulfatide injections protected CD1d deficient mice from the disease, consistent with the theory that sulfatide-reactive NKT-cells could suppress the autoimmune process (Jahng et al., 2004).

30

introduction

1.6.3.3. Lyso-sulfatides A sulfatide with the fatty acid chain removed is called a lyso-sulfatide (figure 20). The lyso-sulfatide sphingosine-1-galactoside-3-sulfate has also been found to stimulate NKT II cells (Roy et al., 2008). It is not clear how the lyso-forms bind in CD1d. Although with ceramide lipids the sphingoid base is bound in the F’ pocket, in the absence of the fatty acid, the long chain base potentially could anchor in either of the A’ or F’ pockets. The lack of the fatty acid may also influence the positioning and/or flexibility of the head group (Rhost et al., 2012) Since lyso-sulfatide has only one hydrophobic chain it can be expected to complex with the CD1d molecule in a different manner than the sulfatide with two hydrophobic chains. It is unclear whether the sphingosine binds in the A’ of F’ pocket of CD1d (Zajonc et al., 2003).

Figure 19. Structure of α-GalCer (upper) and β-GalCer (lower). (Brown and Mattjus, 2007).

Figure 20. Sulfatide (upper) and lyso-sulfatide (lower) (Webpage, www.lipidworld.com).

1.6.3.4. Sphingomyelins Sphingomyelin, or phosphosphingolipid, is a sphingolipid where the ceramide is linked to a phosphate containing group such as phosphocholine or phosphoethanolamine by an ester linkage (figure 21). Sphingomyelin is one of the major components of the lipid bilayer and is believed to be the only phospholipid not derived from glycerol (Chalfant and Del Poeta, 2010). Sphingomyelin has been suggested as the autoantigen of IC2 (Kavishwar et al., 2011).

Figure 21. Sphingomyelin (Webpage, www.themedicalbiochemistrypage.org).

Figure 22. Sphingosylphosphorylcholine (Webpage, www.trc-canada.com).

1.6.3.5. Lyso-sphingomyelins As it is the case with sulfatide, sphingomyelin with the fatty acid chain removed by hydrolysis is called a lyso-sphingomyelin. Another nomenclature for lyso-sphingomyelin is sphingosylphosphorylcholine (figure 22).

31

introduction

1.6.3.6. Phosphatidylcholines Phosphatidylcholines are a class of phospholipids derived from glycerol with choline the polar head group (figure 23). They are a major component of biological membranes. Phosphatidylcholine and sphingomyelin have the same polar head group but differ in the backbone where phosphatidylcholine has a glycerol backbone and sphingomyelin has a ceramide backbone (Fahy et al., 2005).

Figure 23. Phosphatidylcholine (Webpage, www.wikipedia.org).

1.6.3.7. Sodium cholesteryl sulphate Sodium cholesteryl sulphate is a sulphated cholesterol and thus it belongs to the group of derived lipids, cf. Table 3 (figure 24). It is present in high concentrations in human plasma and is also a component of the cell membranes where it has a stabilizing role. Because of the many carbohydrate rings and the carbon tail, the molecule is overall hydrophobic but has a negatively charged sulphate group (Abrahamsson et al., 1977, Strott and Higashi, 2003).

Figure 24. Sodium cholesteryl sulphate (Webpage, www.thsci.com).

1.7. Antibody interactions and kinetics One of the aims of this project was to determine whether one or several of the lipids discussed in section 1.6.3 could be part of the autoantigen for the IC2 antibody, perhaps in a complex with the CD1d molecule. If interaction is observed between IC2 and the lipids, which is what has been observed earlier, affinity measurements of the interactions could help to clarify which of the lipids that are minor and major autoantigens, defined by the strength of the interaction. First I will briefly describe the difference between affinity and avidity. Then, I will describe the basic kinetic parameters that can be used to evaluate interactions. State of the art techniques for measuring antibodyantigen affinities are then touched upon, before a more thorough description of two techniques of particular interest to this project.

1.7.1. Affinity and avidity Antibodies are produced in order to bind and retain potentially harmful substances in the organism of the host organism until these substances can be destroyed by other cells in the immune system. Therefore it is essential that antibodies display a very high affinity towards the specific antigen. Because of the power of

32

introduction

the immune system in response to substances bound by antibodies, it is also necessary that the antibodies display a high stability and high specificity towards the antigen, to avoid unspecific and potential autoimmune reactions (Murphy et al., 2008). The IgM isotype, a pentameric antibody, is known to have several low-affinity binding sites normally compensated by multiplicity of binding (Murphy et al., 2008). This combined synergistic strength of an antibody-antigen interaction, avidity, depends on both the affinity of binding sites and the valency of interactions (Abbas et al., 2007, Murphy et al., 2008). In this project we have been dealing with several different IC2 formats where almost all are oligovalent (section 1.3.2.). Therefore, when applying the theory of monovalent interactions, as we are going to, the affinity is only an approximation to the real system since it deals with affinity and not avidity. When attempting to do comparable analyses it is therefore very important that the affinities measured by the same technique are truly comparable. Therefore the environment such as physical and chemical parameters of the sampling should be as identical as possible throughout all experiments. It is also important to be aware of factors such as the aforementioned multivalent binding when doing the comparison. Since the assembly of antigens is ranging from intact cells, over plasma membranes to single lipids, we have in this project been limited to use affinity measurement techniques applicable to these very diverse subjects.

1.7.2. Kinetics Several kinetic parameters can be used to mathematically describe antibody-antigen interactions. In this project primary focus was on the dissociation constant (KD) and the association and dissociation rate constants (ka and kd), whenever these were obtainable. Also the affinity constant, KA, can be used to describe the interactions. The rates and constants are given by the law of mass action, which describes the binding kinetics of monovalent interactions. The law of mass action and the involved kinetic constants are shown in equation 1, whereas the formation and dissociation of the complex with ka (M-1s-1) as the association rate constant and kd (s-1) as the dissociation rate constant is described in equation 2 (Björkelund et al., 2011).

equation 1

equation 2

These interaction parameters can be used to describe and relatively compare the binding of different IC2antibody formats to each other. This way one can efficiently distinguish which of the measured antigens IC2 displays a higher affinity for, indicating that this antigen could be the, or a part of the, in vivo IC2autoantigen. These parameters can also be used to examine if the smaller formats of IC2 than the full IgM have maintained the high affinity for the autoantigen, indicating high affinity interactions on all binding sites, or if the high affinity observed is the result of the combined synergistic strength of multiple lowaffinity binding sites on IgM.

33

introduction

1.8. Affinity measurements 1.8.1. Prevailing techniques form measuring antibody-antigen affinities Until recently, primarily indirect competition methods based on ELISA, FCM and RIA have been used in antibody-antigen interaction studies (Radbruch, 1992, Friguet et al., 1993, Goldberg and DjavadiOhaniance, 1993). These methods all require labelling of the antibody to be investigated or the use of a labelled detection of antibody or antigen. For control experiments it might also be necessary to label the antigen (Fagerstam et al., 1990). The indirect ELISA measurements are both simple and cheap and require only small amounts of antigen and antibody. The drawback of this method is that it requires both direct enzyme labelling and immobilization of one of the components (Bobrovnik, 2003). FCM can do sensitive measurements of molecular interactions as well as affinity measurements, but only on intact cells, and thus this method is not readily suitable for this project (Radbruch, 1992, Nolan and Sklar, 1998). In recent years, a new method for cellular affinity measurements based on real-time cellular RIA has emerged. This technique quantifies tracer-cell interactions in real-time by using a radioisotope-labelled biomarker tracer which is known as real-time cellular radio immunoassay (CRIA) or in a semi-automated format LigandTracerTM technology (Ridgeview Technologies, New Hampshire, USA). This technology is practically based on the principle of a rotating RIA on cells (Bjorke and Andersson, 2006). CRIA in the format of LigandTracer can record kinetic data and calculate affinity from the measurement of different concentrations of radioactively labelled ligand tracer interacting with an area of attached target cells. The affinity of IC2 towards intact β-cells from different cell lines has been successfully measured by LigandTracerTM real-time CRIA. This technique is, however, only applicable to intact cells. Furthermore, LigandTracerTM requires radioactive labelling and can only be applied in a small scale screening which makes it unsuitable for this project where the proposed screening is of relatively large scale (Bjorke and Andersson, 2006, Desai, 2009). Another method, isothermal titration calorimetry (ITC), uses a calorimetric approach to directly extract the thermodynamics of an interaction. It is a true label-free method, which by virtue of the direct measurements extract both the dissociation constant and the stoichiometry of an interaction (Jelesarov and Bosshard, 1999, Privalov and Dragan, 2007, Ladbury, 2010). ITC provides a direct real-time measure of dissipated or absorbed heat of a reaction. However, to gain a measurable amount of heat a remarkably high concentration in the order of milligram pr. millilitre is required for at least one binding partner (Bunde et al., 1998, Kurosawa et al., 2004, Jerabek-Willemsen et al., 2011). Label-free measuring methods, such as Surface Plasmon Resonance (SPR), have the advantage of being able to measure the kinetics of biomolecular interactions in real-time. For SPR, samples with milimolar protein concentration is required, and the technique is well suited to measure the majority of high-affinity reactions (KD from 10μM-10nM). SPR is a quick method to use after immobilization of one component, e.g. the IC2 antibody, and a screening of a variety of samples would be possible. SPR has many advantages over the previously mentioned methods such as high sensitivity, being label-free and delivering data in realtime. The major drawback of SPR is that it requires immobilization, raising issues of steric hindrance and molecular activity. SPR suffers from mass transport limitations (Schuck et al., 2001). Furthermore, strong hydrophobic antigens like lipids are inconvenient in SPR (Fagerstam et al., 1990, Schuck et al., 2001).

34

introduction

1.8.2. Quartz crystal microbalance (QCM) Another label-free technique capable of measuring real-time affinities of molecular interactions is quartz crystal microbalance (QCM). QCM uses an acoustic, or piezoelectric, sensing technology which is a unique method for in situ observations of change in contact mechanics, interfacial dynamics, surface roughness or mass in soft-matters (Kurosawa et al., 2004, Shen et al., 2008, Becker and Cooper, 2011). The QCM apparatus has a thin quartz crystal chip sandwiched between a pair of detecting electrodes. An alternate current (AC) voltage is applied across the electrodes, causing the crystal to excite and oscillate. If the frequency of the applied Figure 25. The principle of QCM. A QCM consists of a quartz crystal voltage matches the crystal's resonance (purple) coupled to a flow system. When the ligand (blue) is flushed frequency (or multiples hereof called through the flow channel, it binds to a molecule (black) immobilized or adsorbed to the surface of the crystal. The change in mass bound to the overtones), a standing wave is generated quartz crystal, leads to an equivalent shift In the oscillation frequency of inside the crystal. Depending on the cut of the quartz(Cooper and Singleton, 2007). the crystal relative to its crystallographic axes, different kinds of oscillation may arise. For example, certain crystals vibrate in a so-called thickness-shear mode where the two surfaces move in an anti-parallel fashion. In liquids and gases, shear-waves decay rapidly, making QCM interface-specific. This oscillation is picked up by a transducer which converts and transmits the signal into an electrical signal (Shen et al., 2008, Jerabek-Willemsen et al., 2011). The frequency of this excitation can thus be measured. The frequency of the vibrating crystal is altered by either adding or removing molecules from the surface of the crystal, leading to a frequency shift (figure 25). Thus, the crystal is used as a sensitive microbalance, recording a decrease in frequency corresponding to mass increase on the sensor surface during interaction (figure 26)(Kurosawa et al., 2004). Figure 26. A QCM sensogram. The QCM sensogram shows the frequency (blue) and dissipation (red) of the vibrating crystal as a function of time. The frequency is altered by either adding or removing molecules from the surface of the crystal, leading to a frequency shift. The QCM crystal (yellow) is calibrated in an appropriate running solution. As IgM antibodies (green) are injected into the flow cell, a shift in frequency and dissipation is observed. After rinsing with PBS, the injection of an IgM specific antibody (purple) causes an additional shift in frequency together with a more modest shift in dissipation.

35

introduction

One of many ways to perform QCM measurements is to use the ring-down scheme. Here the external driving voltage is turned off in intervals and the oscillations are left to decay freely. Because quartz is piezoelectric, a voltage is generated during these decaying mechanical oscillations. The signal is then recorded, giving two different parameters per overtone: the resonance frequency and the dissipation. The dissipation of the crystal oscillation reveals the viscoelasticity of the material (Höök et al., 1998). If the layer changes from a rigid conformation to a more loosely bound layer, the dissipation will increase (figure 26 and 27)(Reviakine et al., 2011). The quartz crystal microbalance with dissipation (QCM-D) (Q-sense, Sweden) is capable of measuring dissipation (Rodahl et al., 1995) thereby greatly increasing the informational content of the data. Like SPR, QCM requires immobilization but because of the piezoelectric basis of the technique. It has rapid response time, low operation costs, and the ability to work directly in complex liquids (Kurosawa et al., 2004, Webpage, www.q-sense.com). There are no size restrictions and multichannel systems exist to make several recordings at a time. QCM detects more properties than just mass, including changes in liquid viscosity, density, temperature and pressure. This, of course, also makes the QCM data more prone to erroneous conclusions which should be kept in mind when analysing data. The large Figure 27. Dissipation. As the layer bound layer changes from drawback of QCM is, as with SPR, the challenge rigid (red) to softer with the binding of a larger and more associated with immobilization procedures and flexible component (green), the dissipation signal changes from optimization, which can be very time consuming a slow and homogeneous decay (red) to a faster and less consistent decay (green) (Webpage, www.q-sense.com). and thus it might not be suited for projects with only short research periods (Cooper and Singleton, 2007, Shen et al., 2008, Becker and Cooper, 2011). In addition, when measuring molecular interactions with SPR and QCM there is a need for a certain change in mass to obtain a high enough signal-to-noise ratio. Finally, the excisting QCM instruments consume relatively large volumes of sample material (Bunde et al., 1998).

1.8.3. Microscale thermophoresis (MST) Microscale thermophoresis (MST) is a method for the analysis of a broad range of molecular interactions. It measures the mobility of a fluorescently labelled molecule in a laser induced microscopic temperature gradient field (thermophoresis) and the temperature dependent change of fluorescence (MST T-jump) (Webpage, www.nanotemper.de). The MST consists of an infrared laser coupled to the path of fluorescence excitation/emission using a dichroic mirror. The infrared laser is focused into the sample through the same optics as used for fluorescence detection, causing local heating of a solution with very high precision (figure 28)(Jerabek-Willemsen et al., 2011).

36

introduction

During MST measurements, fluorescently labelled molecules or particles are initially distributed evenly and diffuse freely in solution. By switching on the infrared laser, the molecules experience a thermophoretic force caused by the temperature gradient, and move out of the heated spot. In steady state, this molecular flow is counterbalanced by ordinary mass diffusion. After turning off the laser, the particles diffuse back towards a homogeneous distribution. By recording thermophoresis, back diffusion and changes in fluorescence, information about the affinity and mechanisms of binding can be derived (figure 29 and 30).

Figure 28. MST setup. The MST consists of an infrared laser coupled to the path of fluorescence excitation/emission using a dichroic mirror. The infrared laser is focused into the sample causing a very high precision localized heating of the samples in capillaries (Baaske et al., 2010).

Figure 29. The coupling between fluorescence and molecular flow during a MST measurement. In the initial state there is a homogeneous distribution of the molecules. The laser heats the sample leading to a change in fluorescence induced by thermophoresis. After a steady state has been established, the laser is turned off. As the sample cools, the molecules diffuse back in to a homogeneous distribution, causing a change in fluorescence. All these individual phases contain information about the affinity and mechanism of binding. (Baaske et al., 2010, Webpage, www.nanotemper.de).

Figure 30. An example of a MST plot. A direct interaction study between a fluorescently labeled protein kinase and a small molecule inhibitor. The concentration of the labeled protein kinase is kept constant and the small molecule is titrated from 25000-0.3 nM. The difference in normalized fluorescence was plotted against the concentration of the small molecule. The drop in fluorescence with increasing small molecule concentrations is indicative of interaction. The KD of this interaction was fitted to 6-2 nM (JerabekWillemsen et al., 2011).

37

introduction

The thermophoretic effect is very sensitive to the molecule-solvent interface which allows for quantification of biomolecule interactions as they are induced by a binding event. The difference in the molecule's thermophoresis is utilized to quantify the binding in titration experiments under constant buffer conditions. The approach is sufficiently sensitive to measure the interactions of low-molecular weight compounds and proteins, as well as the interactions between proteins and liposomes (Baaske et al., 2010, Jerabek-Willemsen et al., 2011). MST is, like QCM, well suited to measure a wide range of interactions, from small molecule-protein interactions, to quantification of affinities of multi-subunit complexes and liposomes with essentially no limitations on molecular size or weight. MST is, because of the thermophoretic nature of the technique, exceptionally sensitive, thus combining various aspects of the above mentioned methods. Apart from being easy to use, very quick, and not requiring immobilization of molecules, MST also has very low sample consumption (only a few millilitres at nanomolar concentrations). Measurements are performed within minutes in free solution in a buffer and temperature of choice, or in a complex biological context like cell extract or blood serum. Unlike any other interaction analysis technique, the signal-to-noise ratio does not significantly depend on the molar weight ratio in thermophoresis. High-affinity in the nanomolar regime is equally well measured as low-affinities in the high micromolar range (Bunde et al., 1998, JerabekWillemsen et al., 2011, Wienken, 2011). The drawback of this method is that it requires fluorescent labelling of one of the components which, like in isotope-labelling, can change affinity by structural changes. A new labelfree version of MST has recently been produced. It uses the intrinsic fluorescence of a protein to follow the migration of molecules (Webpage, www.nanotemper.de). MST provides only endpoint measurements, not real-time, leading to several blind spots one being the lack of information on association rate and time to equilibrium (Björkelund et al., 2011).

1.8.4. Choice of methods for this project When it comes to which affinity measurement method to choose for the IC2 antigen (affinity measurements in protein-lipid systems), the advantages of especially the intrinsic label-free nature of QCM, the rapid response time, and the requirement for only a single analyte-specific antibody gives good reasons to explore this technique as a substitute of ELISA or FCM. Also SPR is definitely a technique of choice especially to compare and confirm experiments performed with QCM, but unfortunately we did not have access to an SPR in this project. But the MST technique, which does not have any restrictions regarding immobilization, was then chosen in addition for this project, as it was expected to provide interesting affinity measurements for comparison in free solution, and it might be the method of choice for screening of potential small molecule autoantigens.

38

study plan

2. Study plan The studies I planned to do in order to fulfil the overall aim of my project (as discussed above) were the following: Measure the affinity of the IC2 antibody to a range of potential IC2 antigens especially of lipid origin. Compare the affinity of the various IC2 antibody formats towards intact β-cells to try to find the best suited format for further imaging trials. As regards methodology I decided to use primarily QCM-D to obtain real-time measurements of affinity and to compare the potential antigens with each other. I wanted to supplement with MST in order to perform the same measurements in a different system with other strengths and weaknesses.

39

materials and methods

3. Materials and methods 3.1. Materials 3.1.1. Chemicals and reagents In this section, the buffers and reagents used and their abbreviations are listed. Table 4. List of buffered salines Abbreviation Buffer PBS Phosphate-buffered saline TBS Tris-buffered saline Table 5. List of other reagents Abbreviation Reagent BSA Fatty acid free bovine serum albumin, fraction V Hellmanex 2 % Hellmanex in mili-Q water MQ Mili-Q water Percoll Percoll

Manufacturer F. Hoffmann-La Roche AG, Switzerland Hellma GmbH & Co. KG, Germany Merck Milipore, Massachusetts, USA GE Healthcare, United Kingdom

PVP-40

Polyvinylpyrrolidone weight 40.000

molecular

Sigma-Aldrich, Missouri, USA

Tween-20 HPBS

Tween-20 Hank’s Balanced Salts with calcium, magnesium, sodium bicarbonate and phenol red Dulbecco’s Phosphate Buffered Saline without calcium and magnesium Biowhitaker PBS EDTA pH 7.5 BioWhitaker Trypsin EDTA Sodium azide

Sigma-Aldrich, Missouri, USA BioSera, France

DPBS PBS EDTA PBS trypsin NaN3

with

average

BioWest SAS, France Lonza Biosciences, Switzerland Lonza Biosciences, Switzerland Merck-Schuehardt, Germany

3.1.2. Cell lines Six different cell lines were used. The insulinoma derived mouse β-cell line βTC-tet (Efrat et al., 1988, MiloLandesman and Efrat, 2002) and the insulinoma derived mouse α-cell line αTC1-6 (Powers et al., 1990) were kindly provided by Professor S. Efrat, Department of Human Genetics and Molecular Medicine, Tel Aviv University, Israel. The insulinoma derived β-cell line Min-6, originating from a transgenic C57BL/6 mouse (Miyazaki et al., 1990) was kindly provided by Professor Miyazaki, Institute for Medical Genetics, Kumamoto University Medical School, Japan. The insulinoma derived rat β-cell line RIN-5AH (Billestrup, 1985) was kindly provided by Professor J. Høiriis Nielsen, Department of Biomedical Sciences, University of Copenhagen, Denmark. The insulinoma derived rat β-cell line INS-1E (Asfari et al., 1992) was kindly provided by Professor P. Maechler, University of Geneva, Switzerland. The B-lymphoma derived mouse cell line A20 transfected with CD1d, A20-CD1d (Lang et al., 2004), was kindly provided by Dr. Mitchell Kronenberg, Division of Developmental Immunology, La Jolla Insitute for Allergy & Immunology, California, USA.

40

materials and methods

3.1.3. Materials and reagents for cell culture All cells were grown in 25-175 cm2 Cellstar Cell Culture Flasks for adherent cells (Greiner Bio-one GmbH, Germany) with standard media (BioWhittaker RPMI 1640 media with Ultraglutamine (Lonza Biosciences, Switzerland) added an extra 10 % Gibco Fetal Bovine Serum (Life Technologies Corporation, California, USA), 1% Gibco Pen Strep (Life Technologies Corporation, California, USA) and 0.1% Gibco 2Mercaptoethanol (Life Technologies Corporation, California, USA). The cells were grown under standard conditions at 37 oC with 5 % CO2 in a Heraeus HeraCell 240 incubator (Thermo Fisher Scienfitic Inc., Massachusetts, USA). Hank’s Balanced Salts with calcium, magnesium, sodium bicarbonate and phenol red (BioSera, France) or Dulbecco’s Phosphate Buffered Saline without calcium and magnesium (BioWest SAS, France) was used to wash the cells. For detachment of cells were used cold Biowhitaker PBS EDTA pH 7.5 (Lonza Biosciences, Switzerland) or BioWhitaker Trypsin EDTA (Lonza Biosciences, Switzerland). 15 ml Cellstar Tubes cat. No.: 188-271 (Greiner Bio-one GmbH, Germany) or 50 ml Cellstar Tubes cat. No.: 227261 (Greiner Bio-one GmbH, Germany) was used for centrifugation.

3.1.4. Antibodies and serum Monoclonal autoantibody IC2-IgM (IC2) and Fluorescein isothiocyanate (FITC) labeled monoclonal autoantibody IC2-IgM (IC2-FITC), the proteolytically digested formats IC2-F(ab’)2 and IC2-Fab, as well as the human-rat chimeric rhIC2 were all provided by ImmunoSigns, Denmark. Goat-anti-rat-HRP (horseradish peroxidase) used in immunoblotting was bought from Southern Biotech, Alabama, USA. As controls in these experiments were used Biowest Rat Serum (cat. no. S2150-010) and Wistar rat serum kindly donated by the department of experimental medicine (section 10.3) at the University of Copenhagen, Denmark.

3.1.5. Lipids Table 6 provides an overview of the lipids that were used. Table 6. Overview of lipids Abbreviation Lipid

Solvent

Origin

Manufacturer

SM

Sphingomyelin

chloroform:methanol 2:1

Porcine RBC

SM

Sphingomyelin

chloroform:methanol 2:1

Chicken egg

SM

Sphingomyelin

chloroform:methanol 2:1

Bovine buttermilk

SM

Sphingomyelin

chloroform:methanol 2:1

Bovine spinal chord

LSM

chloroform:methanol 2:1

M.LSU

Sphingosylphosphorylcholin e (lyso-sphingomyelin) + Lyso-sulfatide (NH4 salt)

S.LSU

Lyso-sulfatide

Ethanol

SUL

chloroform:methanol 2:1

β-GalCer

Sulfatide containing crude material Galactocerebrosides ~ 99 %

Semisynthetic bovine Bovine brain sulfatide Bovine brain

chloroform:methanol 1:1

Bovine brain

α-GalCer

α-galactosyl-ceramide

DMSO

Synthetic

SCS

Sodium cholesteryl sulphate

Methanol

Matreya, Pennsylvania, USA Matreya, Pennsylvania, USA Matreya, Pennsylvania, USA Matreya, Pennsylvania, USA Matreya, Pennsylvania, USA Matreya, Pennsylvania, USA Sigma-Aldrich, Missouri, USA Kindly donated by Dr. J. Portoukalian Sigma-Aldrich, Missouri, USA Funakoshi Company, Japan Sigma-Aldrich, Missouri, USA

chloroform:methanol 2:1

Catalog number 1328 1332 1329 1051 1321 1904

C4905

C9523

41

materials and methods

3.1.6. Other reagents The lectin Jacalin (Vector Laboratories, California, USA) was kindly donated by Attana AB, Sweden. The sulphate monosaccharides D-galactose-3-sulphate (Lectinity Holdings Inc., Russia), D-galactose-4-sulphate with sodium salt (Sigma-Aldrich, Missouri, USA) and D-glucose-3-sulphate (Sigma-Aldrich, Missouri, USA) were all dissolved in water.

3.2. Methods 3.2.1. Cell culturing in culture flasks RIN-5AH, βTC-tet, αTC1-6 and INS-1E cells, all adherent, were grown in culture flasks with standard media at 37 oC with 5 % CO2 in an incubator. The cells were grown until approximately 80-90% confluency before they were split. Before splitting or harvest of cells, the media was discarded and the cells were gently washed twice with HBS or DPBS. After incubation with cold PBS EDTA or PBS trypsin for 20-60 min, the flasks were shaken softly, and the cell suspension was transferred to 15 ml or 50 ml cellstar tubes and centrifuged at 2000 RPM for 3 min in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany). The supernatant was discarded and pellet loosened before HPBS or DPBS was added. The centrifugation was repeated and the supernatant was discarded. This was repeated twice. Then the cell pellet was resuspended in standard media and the harvested cells were added to new cell flasks.

3.2.2. Preparation and isolation of plasma membranes 3.2.2.1. Cells from culture flasks 12 175 cm2 culture flasks with βTC-tet of 80 % confluency and another 12 175 cm2 culture flasks with αTC16 of almost 100 % confluency, all grown in standard media (not glucose stimulated), were emptied for media and washed twice with DPBS. The cells were incubated with 15 ml PBS EDTA for 30 min at 37 oC with 5 % CO2 in incubator. After mechanical shaking of the flasks, the cells suspension was transferred to 50 ml tubes and centrifuged at 2000 RPM for 3 min in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany). This was followed by three washes and subsequent centrifugations with DPBS.

3.2.2.2. Cells from cell factories 12 175 cm2 culture flasks with RIN-5AH and 12 175 cm2 culture flasks with INS-1E were grown to 80 % confluency. The media was discarded and the cells were washed with DPBS and then incubated with cold PBS EDTA for 30-45 min at 37 oC with 5 % CO2 in an incubator. The cell suspension was transferred to 50 ml tubes and centrifuged at 2000 RPM for 3 min in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany), followed by three washes and subsequent centrifugations with DPBS. The cells were then transferred to 10 trays NUNC cell factories (Thermo Scientific, Massachusetts, USA) together with 2 l of standard media (not glucose media) in each cell fabric. The cells were grown for 3 days at 37 oC with 5 % CO2 in incubator. Medium from both cell factories was discarded and the cell fabrics were washed with 500 ml DPBS before the cells were incubated with 500 ml PBS EDTA for 30 min at 37 oC with 5

42

materials and methods

% CO2 in incubator. To optimize the yield, the cell fabrics were added and incubated with another 100 ml PBS EDTA for 20 min at 37 oC with 5 % CO2 in incubator. The cell suspensions were transferred to 50 ml tubes and centrifuged at 2000 RPM for 3 min in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany). The supernatant was discarded and the cell pellets were resuspended in DPBS and centrifuged at 2500 RPM for 3 min in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany). This washing procedure was repeated three times.

3.2.2.3. Isolation of plasma membranes Plasma membranes were isolated using the Blanchet method (Blanchet, 1974). The supernatant was discarded and the pellet resuspended in 0.54% NaCl 0.01M Tris-HCl buffer pH 7.2 in which a “Complete protease inhibitor cocktail tablet” (F. Hoffmann-La Roche AG, Switzerland) had been dissolved. This cell solution was left on ice for approximately 15 minutes for the cells from the cell fabric and 45 min for the cells from the culture flasks, before it was subjected to 40 bar pressure using a French Pressure cell (American Instrument Company, Maryland, USA). The homogenate was then spun down for 8 min at 1500xg in a Biofuge Primo with rotor head #7591 (Heraeus Instruments GmbH, Germany). The supernatant was then transferred to new tubes and centrifuged for 15 min at 7800xg (= 8070 RPM) using a Dupont Sorvall SS34 rotor head (Sorval Instruments,). In the case of the cells originating from the cell fabric, the supernatant was kept on ice before next step. In the case of cells originating from the culture flasks, the supernatant was transferred to new tubes and stored at -20 oC for 4 days, before it was thawed and layered on a 15 ml 20 % (w/v) sucrose bed in 0.01 M Tris-HCl and 5 mM KCl pH 7.2 to approximately 20 ml lipid solution and centrifuged at 135000xg for 30 min in a SW27 rotorhead in a Beckman Coulter Optima LE-80K (Beckman, California, USA). The pellets were homogenized with 0.01 M Tris-HCl and 5 mM KCl pH 7.2 before layered over a linear continuous 30-70 % (w/v) sucrose bed. They were then centrifuged for 5 h at 105000xg using a SW27 rotorhead in a Beckman Coulter Optima LE-80K (Beckman, California, USA). The plasma membrane preparations were stored at -20 oC.

3.2.3. Cellular quartz crystal microbalance experiments (cellular QCM) The cellular quartz crystal microbalance experiments were performed using an Attana Cell 200 (Attana AB, Sweden). Cells are grown on small quartz crystal sensors are loaded in the instrument. Other components e.g. antibodies can then be injected over the sensor surface (figure 31).

Figure 31. The setup of measurements using the Attana Cell 200. 1) The cell (red) grown sensors (yellow) are inserted into the instrument and calibrated in running solution. 2) A component (blue) is injected into the flow cell. 3) If interaction between the cells and the injected component occurs, the component is retained on the surface of the cell. This gives rise to a change in frequency. Unbound components are flushed out of the flow cell.

43

materials and methods

3.2.3.1. Growth experiment of adherent cells in 24 well plates To determine the best seeding concentration for each of the cell types RIN-5AH, αTC1-6 and βTC-tet, a growth experiment in 24 well plates was conducted. Cells were grown, harvested and counted as described in section 3.2.1. 2 ml of cell suspension in standard media added 4.5 g/L glucose, with concentrations made as 3xdilutions ranging between 0.5*10^5 and 13.5*10^5 cells pr. ml, was added to the wells of a sterile Cellstar 24 Well Culture Plate (Greiner Bio-one GmbH, Germany). The cells were grown over night at 37 oC with 5 % CO2 in a Heraeus HeraCell 240 (Thermo Electron Corporation, Massachusetts, USA) incubator. Each well was inspected using an Olympus CK2 inverted microscope (Olympus Corporation, Japan) coupled to an Olympus ULWCD 0.30 phase contrast condenser (Olympus Corporation, Japan) and photographed using a Leica DC200 (Leica Microsystems GmbH, Germany) digital imaging system.

3.2.3.2. Culturing of adherent cells on COP-1 sensor chips in 24 well plates Cells were grown, harvested and counted as described in section 3.2.1. One COP-1 H213245 sensor chip (Attana AB, Sweden) was placed with the front side up in each well of a sterile 24 well culture plate. A suitable number of cells, determined by the previous cell growth experiments, were grown diluted in 2 ml of standard media, and added to the wells of the plate. The cells were grown over night at 37 oC with 5 % CO2 in incubator.

3.2.3.3. Fixation of cells on COP-1 sensor chips Cells grown on COP-1 H213245 sensor chips was inspected in the wells using an Olympus CK2 inverted microscope (Olympus Corporation, Japan) coupled to an Olympus ULWCD 0.30 phase contrast condenser (Olympus Corporation, Japan) and photographed using a Leica DC200 (Leica Microsystems GmbH, Germany) digital imaging system. Media was discarded and washed with cold PBS 1 ml 1 % paraformaldehyde in PBS was added to each well and incubated at 4 oC for 10 min. The fixing solution was removed and washed 3 x 5 minutes with cold PBS. The COP-1 sensor chips were stored in PBS at 4 oC. After approximately 24 hours, the fixated cells were inspected by 4',6-diamidino-2-phenylindole (DAPI) staining. The cells were incubated with 50 ul 1:5000 Invitrogen Molecular Probes DAPI Nucleic Acid Stain 5 mg/ml (Invitrogen, California, USA) for 4 min and washed with PBS before the cells were inspected at 461 nm in a Nicon Eclipse 80i (Nicon, Japan). The cells were photographed through a 4x objective.

3.2.3.4. Cellular Attana Cell 200 experiments The COP-1 sensor chips were dried on the bottom and carefully placed on top of gold conductors in a specially built plastic cartridge (Attana AB, Sweden). The sensor chip cartridges were placed inside an Attana Cell 200 (Attana AB, Sweden) and left to equilibrate in PBS, first with a flowrate of 100 µl/min for 5 min, and then with a constant flowrate of 20 µl/min over night. 55 nM IC2-IgM (ImmunoSigns, Denmark), 55 nM IC2-F(ab’)2 (ImmunoSigns, Denmark) or 0.8-2 µM of the lectin Jacalin (Vector Laboratories, California, USA) were injected with a sensor contact time of 150 seconds. The sensor surfaces were regenerated between measurements with one pulse of 200 µl glycine buffer pH 3.0 or three pulses of 200 µl glycine buffer pH 2.0.

44

materials and methods

3.2.4. Quartz crystal microbalance with dissipation experiments (QCM-D) Quartz crystal microbalance experiments with dissipation monitoring was performed using a Q-Sense E4 (Q-sense, Sweden). Using this instrument, IC2 is adsorbed to a gold coated quartz crystal sensor surface and the other components can then be added by flow (figure 32).

Figure 32. The setup of QCM-D experiments using Q-sense E4. 1) The bare gold coated sensor (yellow) is inserted into the chamber and a stable baseline is obtained with running solution. 2) A component (blue) is added in flow. 3) The component is adsorbed on the gold sensor surface causing a change in frequency and dissipation. 4) Another component is added in flow (white). 5) Interactions between the flushed component and the adsorbed component, causes a change in frequency and dissipation. Unbound components are flushed away.

3.2.4.1. Preparations o All Quartz Crystal Microbalance with Dissipation experiments were performed at 37 C using a QSense E4 (Q-Sense, Sweden) with continuous flow varied between 0.075-0.1 ml/min provided by the peristaltic pump, Ismatec IPC-N 4 (IDEX Health & Science SA, Switzerland). All measurements were performed on gold coated AT cut quartz crystal surfaces with frequency 4.95 MHz +/- 50 kHz from Q-Sense. The flow modules (Q-sense, Sweden) were cleaned with MQ water (Merck Milipore, Massachusetts, USA) and ethanol 96 % and carefully dried with nitrogen. O-rings (Q-sense, Sweden) were placed in 2 % Hellmanex (Hellma GmbH & Co. KG, Germany) for 10 min. The o-rings were then thoroughly washed with water and dried completely with nitrogen, before placed in the flow modules. Surfaces were placed in 2 % Hellmanex for 10 min and subsequently rinsed thoroughly with 15 cycles of Mili-Q water and 96 % ethanol. Surfaces were dried with nitrogen and placed in a UV/Ozone Procleaner (Bioforce Nanosciences, Iowa, USA) for 10 min or in a PDC002 plasma cleaner (Harrick Plasma, New York, USA) for 1 min. The surfaces were placed in the flow modules which were assembled before in- and outlet tubings were connected. All QCM-D measurements were performed at 37 oC. Overtones were found in MQ or in running buffer PBS with 0.1% NaN3 at same temperature and at a flowrate similar to the flowrate for each experiment. A stable baseline was obtained in running buffer both before the initiation of experiments and in between loading of samples.

3.2.4.2. Sample preparation for QCM-D experiments 3.2.4.2.1. Lipid samples The dissolving agent was evaporated from the lipids, before they were added PBS with 0.1% NaN3. The samples were dissolved mechanically before sonicated for 60 min in approximately 40 oC water in a Branson 1210 bath sonicator (Branson Ultrasonics Corporation, Connecticut, USA) and stored at 4 oC over night. Just prior to use, the lipid samples were sonicated for 75 min in 37 oC water in a Branson 2510 bath sonicator (Branson Ultrasonics Corporation, Connecticut, USA). The lipids were kept in 37 oC water until the moment they were loaded.

45

materials and methods

3.2.4.2.2. Living cell samples One culture flask with INS-1E cells were grown to 80 % confluency at 37 oC with 5 % CO2 in incubator. The cells were changed to glucose standard media with a final concentration of 61 mM glucose. After 8 hours of continued growth, the cells were harvested as described in section 3.2.1. The harvested cells were prepared and suspended in DPBS to a final concentration of 6 x 106 cells pr. ml. It was attempted to get the cells in single cell suspension. The cell-PBS suspensions were kept on ice the whole time until use (app. 30 min).

3.2.4.2.3. Plasma membrane samples Plasma membranes from various cell types were isolated as described in section 3.2.2. A weighed amount of plasma membranes containing an unknown concentration of Percoll were dissolved in PBS with 0.1% NaN3. Just prior to use, the plasma membranes were sonicated on ice for 4-10 min with 6-7 seconds pulses at 100 amplitude using a VibraCell tip sonicator (Sonics, Connecticut, USA) being careful that the plasma membrane mixture did not become overheated. In the experiment where A20 plasma membranes were incubated with α-GalCer, 2 µl of 1 mg/ml α-GalCer in DMSO was diluted to a total volume of 200 µl PBS with NaN3 and sonicated for 70 min in app. 37 oC water using a Branson 1210 bath sonicator (Branson Ultrasonics Corporation, Connecticut, USA). The α-GalCer sonicate was mixed with sonicated A20 plasma membranes with an unknown concentration of percoll diluted to a final concentration of 16.5 mg/ml (handled as described above), and incubated for 1 h in a 37 o C water bath.

3.2.4.2.4. Other samples and blocking solutions Monoclonal autoantibody IC2-IgM, IC2-F(ab’)2, IC2-Fab, IC2-rhIgG aliquots were diluted to the specified concentrations in PBS and 0.1% NaN3. β-D-galactose-3-sulphate, D-galactose-4-sulphate, D-galactose-6sulphate and D-glucose-3-sulphate were diluted in water. Different blocking solutions were prepared with either 1% PVP-40, 1% Percoll, 1% tween-20 or 1 % BSA in PBS with NaN3.

3.2.5. Microscale thermophoresis (MST) The microscale thermophoresis measurements were performed using MonolithNT (NanoTemper Technologies GmbH, Germany) and MonolithNT.labelfree (NanoTemper Technologies GmbH, Germany). In MonolithNT a fixed amount of FITC-labelled IC2 antibody is mixed with different a range of potential antigen made in 2 times dilutions. The FITC-label makes it possible to follow the migration of the molecules (figure 33).

Figure 33. 1) The FITC-labeled antibody and the potential antigen is mixed. A laser induces a heat spot (2) in the samples and the molecules migrate away from the heated area as a consequence of thermophoretic forces (3 and 4). When the laser is turned off, the molecules diffuses back (5). By virtue of the size dependent migration rate, the amount of bound molecules can be derived. When labelfree MST measurements are performed, IC2 is unlabeled. In that case, the migration of molecules is observed by the intrinsic fluorescence of IC2 (Webpage, www.nanotemper.de).

46

materials and methods

3.2.5.1. Calibration To obtain the optimal experimental conditions for the microscale thermophoresis measurements, the concentration of fluorescently labeled protein that gives rise to an optimal fluorescent background signal of 300-1000 fluorescent units was determined. A series of experiments were performed to determine the appropriate concentration of the FITC-labeled antibodies for the microscale thermophoresis experiments. Antibodies were diluted to a range of different concentrations in different buffers. All calibration experiments were performed at 25 oC with a laser-on period of 30 seconds and a laser-off period of 5 seconds. The data was obtained at 100 % laser power and 100 % LED power. The optimal final concentration of labeled IC2-FITC antibody was determined to 2.1 nM in PBS based buffers.

3.2.5.2. Preparations and experiments The appropriate concentration of antibody was diluted in selected buffer in 16 eppendorf tubes with a 2x dilution pr. tube giving a final volume of 10 µl in each tube. Subsequently 10 µl of IC2-FITC with a final concentration of 2.11 µM (determined by calibration) were added to each eppendorf tube and the samples were mixed by gentle spinning in a VWR Galaxy Mini-Centrifuge C1213 (VWR, Pennsylvania, USA). The samples were incubated at room temperature for a standard app. 60 min. Approximately 4-5 µl sample were by capillary force sucked into K002 Monolith NT.115 Standard Treated Capillaries (NanoTemper Technologies GmbH, Germany). The capillaries were placed in the loading tray and loaded into the microscale thermophoresis instrument Monolith NT.115 with blue/green laser (NanoTemper Technologies GmbH, Germany). In the case of the experiment with lyso-sulfatide and unlabeled IC2-IgM, the experiments were performed in a Monolith NT.labelfree (NanoTemper Technologies GmbH, Germany). All experiments were performed at 25 oC with a laser on period of 30 seconds and a laser off period of 5 seconds. For each experiment, data was obtained at 20, 50 and 80 % laser power. The light-emitting diode (LED) power was varied from 60-100 % to give a working fluorescent signal of 300-1000 fluorescent units in all experiments.

3.2.5.3. Sample preparation 3.2.5.3.1. Lipid samples The dissolving agent was in some cases evaporated from the lipids, or used with the dissolving agent, before a specific amount of lipids were added PBS with or without a varying tween-20. The samples were dissolved mechanically before use. In some cases, the samples were sonicated for 5-20 min in approximately 40 oC water in a Branson B200 Ultrasonics Cleaner (Branson Ultrasonics, Connecticut, USA) or ElmaSonics S40/(H) (Elma Hans Schmidbauer GmbH & Co. KG, Germany) just prior to use. The lipids were kept in 37-40 oC water until the moment they were mixed with the FITC-conjugated antibody.

3.2.5.3.2. Plasma membrane samples Plasma membranes from various cell types were isolated as described in section 3.2.2. A weighed amount of plasma membranes were dissolved in PBS with or without tween-20. In some cases, the plasma membranes were sonicated in in approximately 40 oC water in a Branson B200 Ultrasonics Cleaner (Branson Ultrasonics, Connecticut, USA) or ElmaSonics S40/(H) (Elma Hans Schmidbauer GmbH & Co. KG, Germany) for 10 min just prior to use.

47

materials and methods

3.2.5.3.3. Other samples IC2-FITC was diluted to the specified concentrations in PBS with or without Tween-20. D-galactose-3sulphate, D-galactose-4-sulphate and D-glucose-3-sulphate were diluted in water.

3.2.6. Immunoblotting SphingoStrip Membranes (Invitrogen, California, USA) was placed in a petridish and blocked in 10 ml blocking solution (TBS with 3% BSA and 1% PVP-40 for 1h). The membrane was incubated at 4 oC over night with 5 ml 1 µg/ml IC2 dilution in the above mentioned blocking solution. After incubation, the membrane was washed twice with blocking buffer followed by three 10 ml washes with 10 min wash at agitation in blocking buffer. The membrane was then incubated for 1h at RT with 1 µl Goat-anti-rat-HRP in 5 ml blocking buffer. After incubation, the membrane was washed twice with blocking buffer followed by three 10 ml washes with 10 min wash at agitation in blocking buffer. The Membrane was blotted with EZ-ECL chemiluminescence detection kit for HRP (Biological Industries, Israel) and incubated for 3-5 min before the blot was detected using an ImageQuant RT ECL (GE Healthcare, Ohio, USA) with an interval of 5 seconds.

48

results

4. Results 4.1. Quartz crystal microbalance (QCM) IC2 has long been known to interact specifically with the surface β-cells (Brogren et al., 1986). To show this interaction on intact cells, experiments using both cellular QCM and QCM-D monitoring were performed.

4.1.1. Attana Cell 200 experiments 4.1.1.1. 24 well growth experiment To ensure proper confluency of the cells by determining the optimal seeding concentration for growth on the Attana COP-1 sensors, a growth experiment with the chosen cells lines were performed in 24 well plates (Appendix III). αTC1-6 and βTC-tet cells were seeded with concentrations of 1*105-2.7*106 cells pr. well. RIN-5AH cells were seeded at concentrations of 1*105-9*105 cells pr. well. The growth of cells were found to be optimal at different concentrations for the different cell lines. The optimal seeding concentration for the αTC1-6 cells were found to be 9*105 cells pr. well. For both βTC-tet and RIN-5AH cells, 3*105 cells pr. well. Subsequently the cells were fixated in 1% paraformaldehyde (PFA).

4.1.1.2. Growth of cells on COP-1 sensors After growth and fixation of the cells in appropriate concentrations, the cells were kept in PBS at 4 oC and transported to Stockholm, Sweden. At arrival, the fixated cells were dyed with DAPI and accessed by microscopy (appendix III). The confluency was found to be as expected with the βTC-tet and RIN-5AH coated sensors. The αTC1-6 coated sensors had a lower confluency than expected. This may be due to potential differences in growth of this particular type of cells on the COP-1 sensor surface layer compared to the surface layer in the 24 well plates. The low confluency compared to the RIN-5AH and βTC-tet cells could potentially bias the comparability of the data, as the control-sensors with αTC1-6 cells consequently could display a lower signal because of the lower confluency.

4.1.1.3. Cellular QCM experiments To examine the interaction between IC2 and intact β-cells, the RIN-5AH- and βTC-tet cell coated COP-1 sensors were loaded in the Attana Cell 200. IC2-IgM and IC2-F(ab’)2 was injected over the sensors.

49

results

Figure 34 og 35. IC2 injected over intact β-cells Figure 34. Sensogram showing the frequency shift in hertz (Hz) as a function of time is seconds (s) for injection of IC2-IgM over the β TC-tet coated sensor surface A. After buffer calibration, 27.8 nM of IC2IgM was injected over the surface (light blue). The injections gave rise to an extremely low, if any, frequency shift hardly distinguishable from random noise.

Figure 35. Sensogram showing the frequency shift in hertz (Hz) as a function of time is seconds (s) for injection of IC2-IgM over the RIN-5AH coated sensor surface A. After buffer calibration, 55 nM of IC2IgM was injected over the surface (light blue). The injections gave rise to an extremely low, if any, frequency shift hardly distinguishable from random noise.

The injections of IC2-IgM (figure 34 and 35) as well as IC2-F(ab’)2 (not shown) gave rise to extremely low frequency shifts which could not be distinguished from random noise meaning that no binding was observed. This was the case for all sensors coated with β-cell lines and this with concentrations more than twice as high as what was found to be successful in the pilot experiment with INS-1E cells (Pedersen et al., 2011).

50

results

To check for system malfunction and whether something could have happened to the expression of cell surface antigens in general, the lectin Jacalin was injected (figure 36-37). Figure 36. Lectin injected over intact RIN-5AH-cells

Figure 37. Lectin injected over βTC-tet cells

Figure 36 and 37. Unifying plot showing the frequency shift in hertz (Hz) as a function of time is seconds (s) for two separate injections of antibodies over the Rin-5AH coated sensor D (figure 36) and the βTC-tet coated sensor C (figure 37). After calibration, 55 nM of IC2 was injected over the surface (light blue). The surface was regenerated with one pulse of 200 µl glycine buffer pH 3.0 before 862 nM Jacalin was injected (pink). No binding was observed with IC2-IgM whereas a clear binding signal was detected with the lectin Jacalin.

Whereas no interactions were detected even at high IC2 concentrations, a clear binding signal was observed with Jacalin on both cell types, meaning that O-glycoproteins was present on the surface of the βcells. Thus it was concluded that the IC2 antigen was not present on the surface of the fixated β-cells.

4.1.2. Quartz crystal microbalance with dissipation monitoring (QCM-D) Before initiation of the actual series of experiments, experiments were performed to determine a concentration of IC2 giving rise to a good signal-to-noise ratio. It was determined that the 50 µg/ml IC2 concentration, giving rise to an increase in dissipation and a frequency shift of 65-80 Hz, was sufficient to give a good biding signal while avoiding non-specific binding effects (appendix IV)(Wolny et al., 2010). Also, experiments were performed to find the best suited blocking agent. The following blocking agents were used: 1 % PVP-40, 1 % BSA, 10 % Percoll (a suspension of colloidal silica particles of 15-30 nm in size coated with PVP (Pertoft et al., 1978)), and 1 % tween-20. PVP-40 and BSA seemed to be reversible blockers and Percoll had a partially reversible blocking effect (appendix IV).

4.1.2.1. Living cells After these initial experiments, experiments with IC2 and living INS-1E β-cells were performed. IC2 was adsorbed on the cell surface and a flow of glucose stimulated living INS-1E cells was injected into the flow cell (figure 38). As a control, the sensor surface was with different components from both normal rabbit immunoglobulin fraction, control rat serum, instead of IC2 (figure 39).

51

results

Figure 38. IC2 and living cells Figure 38. Adsorption of IC2 and flow of living INS-1E cells. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and unspecific binding was blocked with a flow of 10 % percoll in 6 PBS. After this, a flow of 1 ml 1*10 living glucose stimulated cells INS-1E cells was injected into the flow cell. A stable baseline was obtained in PBS before and in between injection of samples.

Figure 39. Control rat serum and living cells

Figure 39. Adsorption of components from control serum and flow of living INS-1E cells. The plot shows the frequency (blue) and dissipation (orange) for resonance 7 as a function of time. Components from 1 ml normal rabbit immunoglobulin fraction mixed 1:1 with PBS, and 1 ml control rat serum diluted 1:4 in PBS was adsorbed on the gold surface of the sensor and a flow of 1 % BSA in PBS was injected into the flow cell to further block the sensor with different serum components . After this, a flow of 1 ml 6 1*10 living glucose stimulated INS-1E cells was injected into the flow cell. A stable baseline was obtained in PBS before and in between injection of samples.

Upon injection of living INS-1E cells (figure 38), a shift in frequency of approximately 25 Hz was observed. Since we only see very modest shifts in frequency and dissipation, it is not likely that we are observing an interaction between IC2 and plasma membrane vesicles. An interaction does occur, but this is likely to be binding of smaller components such as cytokines or medium components such as ions, that has not been properly removed before measurements. In figure 39, a frequency shift of approximately 15 Hz is observed when living INS-1E cells are added. Only 10 Hz less than what is observed with IC2 in figure 38. In both cases the event causes a shift in dissipation of less than 5 units. Therefore it can be concluded that in this case, no binding of living INS-1E cells to IC2 was observed, but again rather to other components present in the cell medium. Rat serum produced a large F shift of ~25 Hz but this shift is partially reversible upon washing with PBS, and thus further absorption of serum components might occur when the cells are injected.

52

results

4.1.2.2. Plasma membrane sonicate 4.1.2.2.1. β- and α-cell plasma membrane sonicate In ELISA, IC2 has previously been shown to interact specifically with isolated plasma membranes from βcells. Therefore, QCM-D experiments were performed with IC2 and sonicated glucose-stimulated plasma membranes from the β-cell line MIN6 (figure 41-42), as a control the α-cell line αTC1-6 (figure 40), and the CD1d-transfected A20-CD1d cell line (figure 43-44). In figure 41 IC2 was adsorbed on the sensor surface and a flow of sonicated plasma membrane from the βcell line MIN6 was injected. In figure 40 IC2 was adsorbed on the sensor surface and a flow of αTC1-6 plasma membrane sonicate was injected. Figure 40. IC2 and αTC1-6 plasma membrane

Figure 40. Adsorption of IC2-IgM and flow of sonicated αTC1-6 plasma membranes. The plot shows the frequency (blue) and dissipation as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a flow of 1 % BSA in PBS was injected to further block the sensor. After this, the running buffer was changed to 10 % Percoll in PBS and a stable baseline was obtained. 2 ml of an unknown concentration of αTC1-6 plasma membranes with an unknown concentration of Percoll in PBS (total conc 40 mg/ml) was sonicated for 10 min on ice with a tip sonicator, before the sonicate was injected into the flow cell. A stable baseline was obtained in running buffer before and in between change of samples.

Figure 41. IC2 and MIN6 plasma membrane Figure 41. Adsorption of IC2-IgM and flow of sonicated MIN6 plasma membranes. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in 10 % Percoll in PBS. 2 ml of an unknown concentration of MIN6 plasma membranes with an unknown concentration of Percoll in PBS (total concentration 30 mg/ml) was sonicated for 4 min on ice with a tip sonicator before the sonicate was injected into the flow cell. A stable baseline was obtained in running buffer before and in between change of samples.

53

results

In figure 40, as expected, we observe that binding of IC2 to the chip gives a frequency shift of app. 75 Hz. When αTC1-6 plasma membrane sonicate is injected, a frequency shift of approximately 20 Hz occurs and a large increase in dissipation is observed, as it is expected for vesicles (Tellechea et al., 2009). As the buffer is changed back to pure PBS, the frequency slowly increases and the dissipation decreases. After app. 1.5 h a stable baseline with no difference in frequency and dissipation compared to the start of the injection. Thus, there seems to be binding of components from the sonicate to IC2, but this binding has a rather high off-rate since it is completely gone after 1.5 h of continuous wash with PBS, meaning that the affinity of the interaction relatively low. Furthermore, the low shift in dissipation implies that the bound layer is rather rigid, which does not correspond to binding of vesicles but rather a smaller and more dense component (Tellechea et al., 2009). In figure 41, injection of IC2 results in a frequency shift of 90 Hz as expected. Injection of Percoll induces a drop to -100 Hz. When the MIN6 sonicate is injected a frequency shift of approximately 30 Hz, accompanied by a very large increase in dissipation, is observed. As the running buffer is changed to pure PBS and upon extensive rinsing with PBS for 6 h, the frequency rises and the total shift after addition of the sonicate decreases to 15 Hz. This interaction for MIN6 has a somewhat lower dissociation rate than with the αTC1-6 cells. The shift in dissipation is 2 units for MIN6, which is very low as it is the case with αTC1-6. This suggests that there is no interaction between IC2 and sonicated MIN6 vesicles. In a serial connected QCM-D experiment, IC2-F(ab’)2 was adsorbed on the sensor surface and the outflow of the sensor from the experiment in figure 41 was coupled to the inflow of another sensor (figure 42). Thus, when a flow of MIN 6 vesicles was injected into the flow cell of figure 41, it was also flushed over the sensor on figure 42. This was done as an experiment to reduce the amount of plasma membrane sonicate used per experiment. Figure 42. IC2-F(ab’)2 and MIN6 plasma membrane

54

Figure 42. Adsorption of IC2F(ab’)2 and flow of sonicated MIN6 plasma membranes. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonances 7. 1 ml 24 µg/ml IC2-F(ab’)2 was adsorbed on the gold surface of the sensor and a stable baseline was obtained in a flow of 10 % Percoll in PBS. This sensor was coupled in flow with the experiment on figure 41 to save as much sample as possible. The outflow from the above sensor was the inflow of this sensor. 2 ml of an unknown conc. of MIN6 plasma membranes in a buffer with an unknown conc. of Percoll was injected in to the liquid flow cell. A stable baseline was obtained in running buffer before and in between change of samples.

results

Before the serial coupling, a flow of IC2-F(ab’)2 gave rise to a frequency shift of less than 5 Hz and almost no response in dissipation. The smaller change in frequency and dissipation as compared to IC2-IgM is due to differences in molecular weight. A flow of Percoll in PBS caused a frequency shift of additional 5-10 Hz. As the serial coupling was performed the baseline stayed stable at -10 Hz. The MIN6 plasma membrane sonicate was flushed over the first sensor (figure 41) and then into the current sensor (figure 42). This gave rise to a large frequency shift of 95 Hz, and an increase in dissipation of 15 units. As the running buffer was changed to pure PBS, this slowly started to reverse. When the experiment was stopped after additional 2-3 h, the total frequency shift amounted to 85 Hz. The dissipation stabilized with a total shift of a 10-15 units. The increase in dissipation and the rather large shift in frequency all suggest binding of vesicles. The low off-rate also suggests that the interaction is strong.

4.1.2.2.2. A20 plasma membrane vesicles New unpublished FCM data has shown that IC2, besides being specific for pancreatic β-cells, also binds to both the mice CD1d-transfected cell lines CD1d-A20 and CD1d-RAW, but not to the corresponding nontransfected cell lines which do not express CD1d (Unpublished data. Appendix I). Therefore, QCM-D experiments were performed with A20-CD1d plasma membrane sonicate (figure 43-44). In figure 43 IC2 was adsorbed on the sensor surface and A20 plasma membrane sonicate was injected into the flow cell. To see whether the binding of A20-CD1d plasma membrane sonicate to IC2 would improve if the CD1d receptors in these cell membranes were loaded with an NKT-cell activating agent, and hence perhaps also an IC2-antigenic, the A20-CD1d plasma membrane sonicate was incubated with α-GalCer prior to injection (figure 44). Figure 43. IC2 and A20-CD1d plasma membrane

Figure 43. Adsorption of IC2IgM and flow of sonicated A20 plasma membranes. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonances 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in a flow of 10 % Percoll in PBS. An unknown concentration of A20 plasma membranes with an unknown concentration of Percoll in PBS (total concentration 40 mg/ml) was sonicated for 4 min on ice using a tip sonicator before the sonicate was injected in the liquid flow cell. A stable baseline was obtained in running buffer before and in between change of samples.

55

results

Figure 44. IC2 and A20-CD1d plasma membrane incubated with α-GalCer

Figure 44. Adsorption of IC2IgM and flow of sonicated A20 plasma membranes incubated with α-GalCer. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonances 7. 1 ml 50 µg/ml IC2-IgM was absorbed on the gold surface of the sensor and a stable baseline was obtained in a flow of 10 % Percoll in PBS. An unknown concentration of A20 plasma membranes with an unknown concentration of Percoll in PBS (total concentration 16.5 mg/ml) was sonicated for 5 min on ice using a tip sonicator and mixed and o incubated for 1 h at 37 C with newly sonicated αGalCer. A stable baseline was obtained in running buffer before and in between change of samples.

In the experiment on figure 43, a flow of IC2 gave rise to a frequency shift of app. 75 Hz. With injection of PBS with 10 % Percoll, a stable baseline was obtained before a flow of A20-CD1d plasma membrane sonicate was injected. This gave rise to a frequency shift which, after rinsing with PBS buffer, amounted to only 10 Hz and no significant increase in dissipation. This change is even smaller than the modest frequency shift of 20 Hz observed with the MIN6 β-cell line on figure 41 even though the concentrations of the A20CD1d plasma membrane sonicate were higher than that of the MIN6 sonicate (40 mg/ml and 30 mg/ml, respectively). When rinsing with PBS, the frequency continues to increase and as the experiment is ended after approximately 3 h, the total shift in frequency caused by the addition of the A20-CD1d plasma membrane sonicate has shrinken to only 5 Hz. As it is the case with most of the other experiments with plasma membrane sonicate, the change in frequency and dissipation is too small to be caused by the binding of plasma membrane vesicles (Tellechea et al., 2009). The observed interaction is more likely unspecific interactions between an inadequately sensor surface and other components from the plasma membrane sonicate, or specific interactions between IC2 and components from the A20-CD1d plasma membrane sonicate such as small micelles of detached lipids and proteins. In figure 44, the injection of IC2 gave rise to a frequency shift of approximately 70 Hz. With the injection of PBS with 10 % Percoll, a stable baseline was obtained before A20-CD1d plasma membrane sonicate incubated with α-GalCer was injected over the surface. This gave rise to a frequency shift of only 5 Hz and only 2 units in dissipation. It is important to consider that a concentration of only 16.5 mg/ml A20-CD1d plasma membrane with an unknown concentration of Percoll was used in this experiment, compared to 40 mg/ml in figure 41 where a 10 Hz shift in frequency was observed. Thus, though the differences in concentration are considered, no notable effect of α-GalCer in regard to IC2s binding of A20-CD1d plasma membranes is observed.

56

results

4.1.2.3. Lipids The CD1d receptor loads endogenous lipids in endosomal compartments. These are transported to the surface of the plasma membrane where the CD1d molecule presents the lipid to NKT-cells (Moody and Porcelli, 2003). IC2 has been found to inhibit the stimulation of NKT-cells from CD1d-lipid complexes, thus it is plausible that the IC2 antigen binds to part of this CD1d molecule or the lipid that is bound. Therefore, the binding of IC2 to various lipids was investigated. Since CD1d-molecules are known to bind the lysoforms of lipids (Rhost et al., 2012), experiments with the lyso-forms of sphingomyelin (sphingosylphosphorylcholine) and sulfatide (lyso-sulfatide) were performed.

4.1.2.3.1. Sphingomyelin Sphingomyelin was proposed as the antigen of IC2 by Kavishwar et al. (Kavishwar et al., 2011). Therefore, QCM-D experiments were performed with IC2 adsorbed on the sensor surface and sonicated sphingomyelin from bovine buttermilk in flow (figure 45). To see if the interaction is concentration dependent, an experiment with only one fifth of the initial sphingomyelin concentration was performed (figure 46). As a control, components from control rat serum was immobilized on the sensor surface instead of IC2 (figure 47). Figure 45. IC2 and sphingomyelin Figure 45. Adsorption of IC2-IgM and flow of sonicated sphingomyelin (from bovine buttermilk). The plot shows the frequency (blue) and dissipation (orange) in Hz for resonances 7 as a function of time. 1 ml 50 µg/ml IC2-IgM was immobilized on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. 0.5 mg/ml sonicated sphingomyelin in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples. Figure 46. IC2 and sphingomyelin (decreased concentration) Figure 46. Adsorption of IC2-IgM and flow of sonicated sphingomyelin (from bovine buttermilk). The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was absorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. 0.1 mg/ml sonicated sphingomyelin in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

57

results

Figure 47. Control rat serum and sphingomyelin

Figure 47. Adsorption of components from control rat serum and flow of sonicated sphingomyelin (from bovine buttermilk). The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. Components of rat serum was absorption on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. 0.5 mg/ml sonicated sphingomyelin in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

In figure 45 a flow of IC2 gave rise to a frequency shift of approximately -80 Hz. With the addition of 0.5 mg/ml sphingomyelin from bovine buttermilk a frequency shift of additional -180 Hz and a change of 40 units in dissipation was observed. The frequency did not rise considerably over the next 1.5 h when experiment was ended. The same was observed in an experiment with 0.5 mg/ml sonicated sphingomyelin from chicken egg (appendix V) and from bovine spinal chord (not shown). In figure 46 a lower concentration of sphingomyelin, the IC2-IgM gave rise to a frequency shift of 60 Hz. The subsequent flow of 0.1 mg/ml sonicated sphingomyelin from bovine buttermilk resulted in a frequency shift of additional -145 Hz and a change of 40 units in dissipation. After 1.5 h of rinsing with PBS, the total frequency shift amounts to a total shift of 110 Hz. In the control experiment on figure 47 the injection of control rat serum gave rise to a frequency shift of 20 Hz. A flow of 0.5 mg/ml sonicated sphingomyelin did not give rise to an additional decrease in frequency or dissipation. A control experiment with only 0.1 mg/ml sonicated sphingomyelin on control rat serum showed the same (appendix 5). Together these experiments indicate that IC2 interacts with sphingomyelins (from various sources) and that components from control rat serum does not. Based on the large shift in both frequency and dissipation, IC2 seems to bind to sphingomyelin in a concentration dependent manner.

4.1.2.3.2. Sphingosylphosphorylcholine Since CD1d-molecules are known to bind the lyso-forms of glycolipids, experiments with the lyso-form of sphingomyelin (sphingosylphosphorylcholine) was performed (figure 48). IC2 was adsorbed to the sensor surface and sonicated sphingosylphosphorylcholine was injected into the flow cell. To verify this interaction, a control experiment was performed in which control rat serum was used instead of IC2 (figure 49).

58

results

Figure 48. IC2 and sphingosylphosphorylcholine (LSM) Figure 48. Absorption of IC2-IgM and flow of sonicated LSM. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 9. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. Then, 1ml 0.1 mg/ml sonicated LSM in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

Figure 49. Control rat serum and sphingosylphosphorylcholine (LSM) Figure 49. Absorption of components from control rat serum and flow of sonicated LSM. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. Components from control rat serum was absorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. Then, 1 ml 0.1 mg/ml sonicated LSM in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

In figure 48, flow of IC2 gave rise to a frequency shift of -55 Hz. As sonicated sphingosylphosphorylcholine was injected, a frequency shift of an additional 65-70 Hz was observed, but as the running solution was changed to PBS, the frequency gave a small kink and then kept on declining for another 15 Hz. Similarly, the dissipation rose from 19 to 25 units when the running solution was changed to PBS. Then the frequency started to increase slowly, the dissipation started to drop, and as the experiment was ended after 1.5 h, the frequency had only almost reached the point of where the initial sphingosylphosphorylcholine frequency drop ended. This observation was reproducible. If we do not take this extra drop in frequency into consideration, we observe a frequency shift of app. 75 Hz which indicates a possible interaction between IC2 and the sphingosylphosphorylcholine sonicate. In figure 49 the control rat serum induced a change in frequency to -65 Hz. As sphingosylphosphorylcholine sonicate was added an additional shift in frequency of 30 Hz was observed. The frequency stabilized and

59

results

slowly started to increase upon rinsing with PBS. The kink on the frequency curve observed in figure 48 was not observed in this experiment. No binding between the components and the rat serum and the sphingosylphosphorylcholine sonicate was observed.

4.1.2.3.3. Sulfatide Already in 1989, sulfatide was suggested as antigen of IC2 (Spitalnik, 1989). Since sulfatide is known to be implicated in the development of type I diabetes (Buschard, 2011), the interaction between IC2 and sulfatide is very interesting. Because of lack of material and information about the sulfatide available, only few experiments were performed with sulfatide using QCM-D. IC2 was adsorbed to the sensor surface and an unknown concentration of sonicated sulfatide was injected into the flow cell. Figure 50. IC2 and sulfatide Figure 50. Adsorption of IC2IgM and flow of sonicated sulfatide. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. An unknown concentration of sonicated sulfatide in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

In the experiment seen on figure 50, injection of IC2 resulted in a frequency shift of 80 Hz and a subsequent flow of 1% BSA in PBS resulted in a small additional drop in frequency. When an unknown concentration of sonicated sulfatide in PBS was flushed over the sensor surface, an additional frequency shift of a little more than 10 Hz was observed accompanied by an increase of 2 units in dissipation. After the running buffer was shifted to pure PBS, the frequency started to rise, and as the experiment was terminated after 1.5 h, the total frequency change amounted to a little less than 10 Hz. An interaction is observed but considering the the frequency shift and dissipation, this does not look like the binding of sulfatide vesicles, but rather some other smaller component from the sonicate (Tellechea et al., 2009).

4.1.2.3.4. Lyso-sulfatide Also experiment with the lyso-form of sulfatide, lyso-sulfatide, was performed. IC2 was adsorbed to the sensor surface and sonicated lyso-sulfatide was injected into the flow cell.

60

results

Figure 51. IC2 and lyso-sulfatide Figure 51. Adsorption of IC2IgM and flow of sonicated lysosulfatide. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 9. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding, before 0.1 mg/ml sonicated lyso-sulfatide in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

In figure 51, flow of IC2 resulted in a frequency shift of 65 Hz. A subsequent flow of sonicated lyso-sulfatide initially resulted in a drop in frequency (-35 Hz) but as soon as pure PBS was flushed into the flowcell, a large desorption occurred. The frequency stabilized after 1h with at the same level as before lyso-sulfatide injection. Thus, no binding of lyso-sulfatide to IC2 was observed at the given concentration of lyso-sulfatide.

4.1.2.3.5. Other glycolipids Kavishwar et al (Kavishwar et al., 2011) found that the presence of cholesterol in the plasma membrane was necessary for the binding of IC2 to β-cells. To elucidate that IC2 actually binds to cholesterol itself, IC2 was adsorbed on the sensor surface and sonicated sodium cholesteryl sulphate was injected into the flow cell. Figure 52. IC2 and sodium cholesteryl sulphate Figure 52. Adsorption of IC2-IgM and flow of sonicated sodium cholesteryl sulphate. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was immobilized on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. Then 1 ml 0.1 mg/ml sonicated sodium cholesteryl sulphate in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

61

results

In figure 52, injection of IC2 gave rise to a frequency shift of 80 Hz. As a flow of sonicated sodium cholesteryl sulphate was injected, the frequency shifted with an additional 20 Hz and a change of app. 5 units in dissipation. As the running solution was changed to PBS, the frequency slowly started to increase. When the experiment was terminated after 30 min, the total shift in frequency with the addition of sodium cholesteryl sulphate amounted to 15 Hz. This indicates that sodium cholesteryl sulphate binds to IC2. Also the cerebroside β-GalCer was tested using QCM-D. No interaction was observed (data not shown).

4.1.2.4. IC2 binding to sulphated monosaccharides When a lipid is bound in a CD1d molecule, the hydrophobic tail of the lipid is buried deep within the lipid binding groove of the CD1d receptor. Thus, the only part of the lipid that is actually surface exposed is the polar head group. Therefore it would be interesting to examine the binding of IC2 to various sulphated monosaccharides. D-Galactose-3-sulphate has previous been found to be the determinant of IC2 binding (Brogren et al., 1989). The binding of IC2 to D-glucose-3-sulphate, D-galactose-3-sulphate and D-galactose4-sulphate was investigated. IC2 was adsorbed on the sensor surface and the monosaccharides were injected into the flow cell. Figure 53. IC2 and sulphated monosaccharides Figure 53. Adsorption of IC2 and flow of various sulphated monosaccharides. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. A stable baseline was obtained in MiliQ water before 1 mg D-glucose-3-sulphate, 0.5 mg Dgalactose-4-sulphate and 0.5 mg/ml β-D-galactose-3-sulphate (all in water) was injected. A stable baseline was obtained in the relevant running solution before and in between change of samples.

On figure 53, a flow of IC2 gave rise to a frequency shift of 70 Hz. When the running solution was changed from PBS to MQ water, the frequency rose from -70 to only -35 Hz. Because problems with the baseline had been observed in previous experiments when shifting from PBS to MQ water if the sensor surface had been blocked with 1 % BSA in PBS (data not shown), no blocking agent was used in this experiment. Part of the observed desorption, seen as an increase in frequency and a drop in dissipation, is due to changes in the viscosity of the fluid. After a stable baseline was obtained in MQ water, 1 mg of D-glucose-3-sulphate was injected. This resulted in a transient frequency shift of only a few Hz and this shift was reversed as the running solution was shifted back to MQ water. With the injection of 0.5 mg D-galactose-4-sulphate sodium salt showed a similar pattern, with a small shift in frequency which was reversed as soon as the solution was shifted to MQ.

62

results

Subsequently, 0.5 mg of β-D-galactose-3-sulphate was injected. This caused a relatively larger frequency shift of around 8 Hz but again the shift was reversed as the buffer was changed back to MQ. It can thus be concluded that neither D-glucose-3-sulphate, D-galactose-4-sulphate nor β-D-galactose-3-sulphate was bound by IC2.

4.2. Microscale thermophoresis (MST) 4.2.1. β-cell plasma membrane IC2 has with ELISA previously been shown to interact specifically with isolated plasma membranes from βcells. Therefore, in solution experiments with FITC conjugated IC2 and non glucose-stimulated plasma membranes isolated from the β-cell lines INS-1E (not shown) and RIN-5AH (figure 54) were performed using MST. No binding was observed between FITC-IC2-IgM and INS-1E vesicles (data not shown). A titration series of sonicated RIN-5AH plasma membrane was incubated with a constant concentration of FITC-IC2. Figure 54. FITC-IC2 and RIN-5AHplasma membranes

Figure 54. A) Temperature jump as a function of plasma membrane concentration. B) Thermophoresis as a function of plasma membrane concentration. C) The normalized fluorescence signal as a function of time. D) Fluorescence signal as a function of plasma membrane concentration. RIN-5AH plasma membranes were suspended in PBS and sonicated for 10 min before titrated in o PBS heated to approximately 40 C. The titration gave a 2 times dilution with concentrations in the of range 40000-1.2 nM. A fixed concentration of FITC-labeled IC2-IgM in PBS was added to all samples before they were incubated in the dark for 60 min.

In the plot of temperature jump as a function of plasma membrane concentration (figure 54A), there seems to be a drop in fluorescence starting at plasma membrane concentrations between 10 0 and 101 which then flattens out around 102-103. However, this total shift in fluorescence is only about 4 fluorescence units and

63

results

the curve does seem rather bumpy. The fluorescence over time (figure 54C) all starts in the same point of about 1,00 fluorescence units, and then migrates with slightly different shifts in fluorescence, which could indicate interaction. When inspecting the plot of thermophoresis (figure 54B) no drop in fluorescence similar to that seen on figure 54A is observed. Also the fluorescence does seem to increase with concentration of RIN-5AH plasma membrane (figure 54D), indicating that the fluorescence drop observed in figure 54A, could be a result of a precipitation problem and not an actual binding. Since the drop in fluorescence was not directly reproducible, the results of this experiment are evaluated as being random noise and the picture of an interaction between FITC-IC2 and the RIN-5AH plasma membranes. This experiment was also performed both with and without sonication of plasma membranes before titration, and with and without tween-20 in concentrations varying from 0.05-1%, but no interaction was observed. Thus, no interaction was observed between the plasma membranes from the β-cell lines tested (INS-1E and RIN-5AH) though this has previously been shown with ELISA.

4.2.2. Sphingomyelin and sphingosylphosphorylcholine Experiments using MST were performed, but no interaction was observed between FITC-IC2 and sphingomyelin using this technique (data not shown). Interactions between FITC-IC2 and sphingosylphosphorylcholine were sought measured (figure 55). Figure 55. FITC-IC2 and sphingosylphosphorylcholine

Figure 55. A) Temperature jump as a function of sphingosylphosphorylcholine concentration. B) Thermophoresis as a function of sphingosylphosphorylcholine concentration. C) The normalized fluorescence signal as a function of time. D) Fluorescence signal as a function of sphingosylphosphorylcholine concentration. Sphingosylphosphorylcholine was suspended in PBS with 0.05 % tween-20 before it was titrated in PBS and 0.05 % tween-20 to give a 2 times dilution in the range 53763-1.6 nM. A fixed concentration of FITC-labeled IC2-IgM in PBS with 0.05 % tween-20 was added to all samples before they were incubated in the dark for 80 min. With a simple KD fit (NanoTemper software) this curve corresponds to a KD value of 234 µM.

A drop in fluorescence is observed in the temperature jump plot (figure 55A). The plot of thermophoresis (figure 55B) is similar, but it is not as pronounced as the temperature jump plot. The drop in fluorescence is

64

results

approximately 12 fluorescence units, which is relatively small, but still enough to indicate a binding. While the data shows a drop in fluorescence, the flattening of the curve is not seen. The plot of fluorescence over time (figure 55C) show that there is some difference in the fluorescence of the samples, which again could indicate binding. However, in the plot of fluorescence versus antigen concentration (figure 55D), the two last points show a fluorescence that is somewhat higher than the rest. When looking at the plot of temperature jump (figure 55A) and thermophoresis (figure 55B) it is exactly these two points that indicates interaction. It is therefore estimated that this data does not show an interaction between FITC-IC2 and sphingosylphosphorylcholine.

4.2.3 Sulfatide and lyso-sulfatide In 1988 sulfatide was first described as a potential part of the IC2 autoantigen (Spitalnik, 1989). Therefore, MST experiments with sulfatide and its lyso-form, lyso-sulfatide, were performed. Lyso-sulfatide from two different providers was examined. Sigma lyso-sulfatide (Sigma-Aldrich, Missouri, USA) and Matreya lysosulfatide (Matreya, Pennsylvania, USA).

4.2.3.1. Sulfatide In spite of previous findings using other techniques, no binding was observed between FITC-IC2 and sulfatide (data not shown).

4.2.3.2. Lyso-sulfatide A titration series of Sigma lyso-sulfatide was incubated with a constant concentration of FITC-IC2. Figure 56. FITC-IC2 and Sigma lyso-sulfatide

Figure 56. A) Thermophoresis as a function of lyso-sulfatide concentration. B) The normalized fluorescence signal as a function of time. Lyso-sulfatide (Sigma-Aldrich, Missouri, USA) dissolved in ethanol was suspended in PBS + 0.1% tween-20 and titrated in PBS giving a 2 times dilution in the range 20000-0.6 nM. A fixed concentration of FITC-labeled IC2-IgM in PBS was added to the samples before they were incubated in the dark for 30 min. With a simple Kd fit (NanoTemper Software) the KD value was determined to 1 µM. This fit, however, was made with the supposed concentration of 20 nM of FITC-labeled IC2-IgM and not 25 nM as it should have been. The data of this experiment have since then been lost due to a computer exchange, why a plot of temperature jump and fluorescence could not be obtained for this experiment.

On the thermophoresis plot (figure 56A) a clear drop in fluorescence is observed with an antigen concentration of around 102. This drop then flattens with an antigen concentration around 104.The thermophoresis of almost 100 fluorescence units gives a good signal to noise ratio, making the results very reliable. It would, however, have been nice with a few more data points in the high antigen concentrations to see that the fluorescence does in fact stay stabilized. The plot of normalized fluorescence (figure 56B) shows that all samples have the same initial fluorescence of 1.00 and then have different levels of fluorescence over time, indicating populations with different rates of migration. It would have been nice to see the plot of fluorescence to make sure that fluorescence does not increase with the antigen concentration, which could indicate a precipitation problem.

65

results

The interaction between FITC-IC2 and Sigma lyso-sulfatide was measured with other experimental conditions with only 0.05 % tween-20 in the dilution buffer (figure 57 and 58).

Figure 57. FITC-IC2 and Sigma lyso-sulfatide

Figure 57. A) Temperature jump as a function of lyso-sulfatide (Sigma-Aldrich, Missouri, USA) concentration. B) Thermophoresis as a function of lyso-sulfatide concentration. C) The normalized fluorescence signal as a function of time. D) Fluorescence signal as a function of lyso-sulfatide concentration. Lyso-sulfatide (Sigma-Aldrich, Missouri, USA) was suspended in PBS and 0.05 % tween-20 before it was titrated in PBS with 0.05 % tween-20 to give a 2 times dilution in the range 2000-0.06 nM. A fixed concentration of FITC-labeled IC2-IgM in PBS with 0.05 % tween was added to all samples before they were incubated in the dark for 70 min. With a simple KD fit (NanoTemper software) this interaction shows a KD value of 3.4 µM.

66

results

Figure 58. IC2-FITC and Sigma lyso-sulfatide

Figure 58. A) Temperature jump as a function of lyso-sulfatide concentration (Sigma-Aldrich, Missouri, USA). B) Thermophoresis as a function of lyso-sulfatide concentration. C) The normalized fluorescence signal as a function of time. D) Fluorescence signal as a function of lyso-sulfatide concentration. Lyso-sulfatide (Sigma-Aldrich, Missouri, USA) was suspended in PBS with 0.05 % tween-20 before it was titrated in PBS with 0.05 % tween-20 to give a 2 times dilution in the range 5000-0.15 nM. A fixed concentration of FITC-labeled IC2-IgM in PBS with 0.05 % tween-20 was added to all samples before they were incubated in the dark for 75 min. With a simple KD fit (NanoTemper software) this curve corresponds to a KD value of 22 µM.

In the experiment on figure 57, the plot of temperature jump (figure 57A) shows a drop in fluorescence which continuous to drop with higher antigen concentrations. No flattening of the curve is observed. When looking at the thermophoresis (figure 57B) the same sudden drop is observed, though the data points differs more from each other than seen in the temperature jump (especially the first data point and the third from right). The total drop in fluorescence is of about 50 fluorescence units. The fluorescence plot (figure 57C) shows that the fluorescence of all samples starts at 1.00. A large population of the samples gives rise to the same fluorescence intensity, whereas 3-4 samples migrates differently. There is a large variation in fluorescence and the curves are crossing each other, which is a sign of poor data quality. The plot of fluorescence versus antigen concentration (figure57D) shows that the first and third data point from the right have a higher level of fluorescence than the others which could be the cause of the drop in fluorescence that is observed in figure 57A and B. All together these factors indicate that we are probably not watching an interaction. In the experiment seen on figure 58 where the concentration of lyso-sulfatide is slightly higher, the plot of the temperature jump (figure 58A) again shows a drop in fluorescence similar to what we saw in figure 57A. When we look at the thermophoresis of these measurements (figure 58B), no drop in fluorescence is observed. The fluorescence as function of time (figure 58C) shows that all samples have the same initial fluorescence of 1.00. The difference in fluorescence as time passes, is not very large and the one sample that actually does show a larger drop in fluorescence with time, is an outlier in the plot of fluorescence as a

67

results

function of antigen concentration (figure 58D). The position of this sample in the fluorescence plot could be caused by deviations in concentration, aggregation or impurities on the capillary containing the sample. Together this points towards that the drop in fluorescence seen at the high antigen concentrations in figure 36A is more likely random noise than an interaction. By mistake, the samples in this experiment were only left to incubate for 10 min, which clearly could have a major impact on the results. Unfortunately there was not enough lyso-sufatide (Sigma-Aldrich, Missouri, USA) to repeat this experiment with longer incubation time. As regards the experiment in figure 35, the antigen concentrations are not high enough to obtain the full s-shaped curve. These last two experiments (figures 57 and 58) differ from the first (figure 56) in that the titration of the samples was performed in PBS with 0.05 % tween-20, where it was done in pure PBS in the first experiment (figure 56). The interaction between the Matreya lyso-sulfatide and unlabeled IC2 was measured using MonolithNT.labelfree (NanoTemper Technologies GmbH, Germany), a labelfree MST system, which uses the intrinsic fluorescence of a molecule to follow the thermophoresis (figure 59). Figure 59. FITC-IC2 and Matreya lyso-sulfatide

Figure 59. The Hot/Cold ratio (thermophoresis) as a function of lyso-sulfatide concentration (Sigma-Aldrich, Missouri, USA). Lysosulfatide was suspended in PBS and 0.05 % tween-20 before it was titrated in PBS and 0.05 % tween-20 to give a 2 times dilution in the range 100.000-3.05 nM. A fixed concentration of non-labeled IC2-IgM in PBS with 0.05 % tween-20 was added to all samples before they were incubated in the dark for 60 min. With a simple KD fit (NanoTemper software) this curve corresponds to a KD value of 15.9 µM.

The hot/cold ratio (thermophoresis) of the labelfree experiment (figure 59A) shows a drop in fluorescence from about 878 to 864 which is 14 fluorescence units which is not a lot, but enough to indicate an interaction. When looking at the normalized fluorescence curve (figure 59B) it is seen that all samples have the same initial starting point at 1,00 and then differs in terms of change in fluorescence, which again could indicate an interaction. The fluorescence curves are rather bumpy and several curves cross eachother which indicates poor data quality. With data like this, a drop in fluorescence of 14 units is probably not enough to talk about interactions. Because of problems with different versions of the NanoTemper Software, we were not able to obtain the curve of temperature jump versus antigen concentration, nor fluorescence versus antigen concentration, which is very unfortunate. The KD of this interaction was calculated to 15.9 µM which is a lot higher that the 1 µM interaction derived from the data in figure 56.

68

results

4.2.4. Other lipids No interaction was detected between FITC-IC2 and sodium cholesteryl sulphate or FITC-IC2 and β-GalCer (data not shown).

4.2.5. Sulphated monosaccharides No interaction was detected between FITC-IC2 and D-glucose-3-sulphate, D-galactose-4-sulphate or β-Dgalactose-3-sulphate (data not shown).

4.3. Immunoblotting on SphingoStrip Figure 60. IC2 blotted on Sphingostrip

To support our findings, an IC2 blotting assay was performed using an EZ-ECL chemiluminescence kit on the commercially available SphingoStrips (Invitrogen, California, USA). On SphingoStrip blotted with IC2 (figure 60), a clear staining is seen in the sphingomyelins spot, whereas no staining was observed in any of the other spots, including sulfatide, ceramide and phosphatidylcholine. Due to instrument malfunction, no control experiment has been performed yet.

Figure 60. Sphingostrip blotted with IC2 using Goat-anti-rat-HRP and an EZ-ECL chemiluminescence detection kit for HRP. The blot shows a clear staining of the sphingomyelins spot with IC2-IgM whereas no spots are stained using a control IgM antibody.

69

discussion

5. Discussion 5.1. Discussion of results 5.1.1. Binding of IC2 to intact cells and plasma membrane 5.1.1.1. Fixated and living cells IC2 has long been known to interact specifically with the surface of pancreatic β-cells. The affinity of this interaction had prior to this project been measured twice. Using CRIA the interaction was estimated to a medium-high affinity with a KD of 18 nM (Desai, 2009). In a pilot experiment using cellular QCM, the affinity was estimated to a high affinity KD of 1.39 nM (figure 8)(Pedersen et al., 2011).

5.1.1.1.1. Cellular quartz crystal microbalance (cellular QCM) To obtain an estimate of the affinity of IC2 towards other β-cell lines and to estimate and compare the affinity of the different IC2 formats (section 1.3.2.), the affinity of IC2 was again sought measured with cellular QCM using the Attana Cell 200 (Attana AB, Sweden). The β-cell lines RIN-5AH and βTC-tet cells have previously shown an expression of IC2-autoantigen comparable to that of INS-1E cell line that was tested in the cellular QCM pilot experiment (Brogren et al., 1986). Therefore the binding of IC2 to these cells was expected to be similar to what we observed in the pilot experiment. For comparison it would have been highly interesting to have tested the INS-1E cell line as well, but because of lack of time and an unexpected slow growth rate of the cell line, it was not possible. No binding of IC2 to the β-cell lines RIN-5AH or βTC-tet was observed whether in the case of the native IC2IgM or the IC2-F(ab’)2 fragment (figure 34 and 35). The IC2-Fab fragment and the recombinant IC2-rhIgG format were not tested. Because of the previous results with INS-1E (figure 8), this was highly unexpected. To make sure that there were no system malfunctions and the other cell surface antigens were present, the lectin Jacalin was injected (figure 36-37). Whereas no IC2 binding could be detected even at high concentrations at the two β-cell lines, a clear binding signal was seen with Jacalin on both cell types, which means that O-glycoproteins were present on the surface of the β-cells. Thus it was concluded that the IC2 antigen was not present on the surface of the β-cells. The absence of IC2 antigen could be due to inefficient fixation. A total fixation of cells is a difficult matter and not all molecular elements are equally well fixated. As described in a study by Tanaka et al. (Tanaka et al., 2010) the fixation of proteins, and in particular lipids, on the cell surface is very often not successful with up to 95 % of the plasma membrane lipids being mobile after fixation. That the IC2 autoantigen is not detected on the fixated β-cells, is actually be consistent with the hypothesis of the supposed antigen of IC2 being lipogenic. As described in section 1.5., the CD1d receptor is a lipid binding protein. If the outer membrane lipids are not properly fixated, the lipid could migrate out of the CD1d receptor, thereby dissolving the IC2 epitope. In the pilot experiment, the cells were grown and fixated shortly before the measurements were performed. With these present experiments, the cells were fixated and kept in PBS for a period of approximately 36 hours before measurements. This leaves plenty of time for not properly fixated glycolipids to diffuse away. Unfortunately, we did not have time or resources to grow

70

discussion

new cells on the sensors on location. Therefore these experiments could not be repeated and the hypothesis neither confirmed nor denied. Also reloading of the CD1d molecules on the fixated cells were discussed, but the resources were not available. Since binding of IC2 to the β-cells were not detected, the affinity of the different IC2-formats could not be measured and compared.

5.1.1.1.2. Quartz crystal microbalance with dissipation (QCM-D) Using QCM-D (Q-sense, Sweden) the inversed setup was used to detect interaction between IC2 and a flow of living INS-1E cells (figure 38). To my knowledge, this kind of experiments with living cells has never before been performed using QCM-D. In the experiment seen on figure 38, there was a shift in frequency of approximately 25 Hz upon injection of 106 living INS-1E cells. Since intact cells are large and heavy, interactions of cells with IC2 on the sensor surface is expected to cause a much larger shift in frequency and dissipation (Tellechea et al., 2009). Since shifts we observe in frequency and dissipation are modest, it is likely to be the image of another interaction. This could be unspecific binding of cell products or medium components such as salts that have not been properly removed from the cell sample. In the control experiment without IC2 (figure 39), a frequency shift of approximately 15 Hz is observed when 106 living INS-1E cells pr. ml are added, which is only 10 Hz less than what is observed with IC2 in figure 38. In both cases the event causes a shift in dissipation of less than 5 units. Therefore it can be concluded that in this case, no binding of living INS-1E cells to IC2 was observed, but rather unspecific interactions to other components present in the cell medium. In this case a sample with growth media from the cells which had gone through the same cleaning procedure as the cells, could have served as a control. Even though we did expect to see binding of the cells to IC2, this is a pioneering experiment and further experiments will be needed to determine whether interactions between antibodies and living cells can be detected this way.

5.1.1.2. Plasma membranes IC2 has previously been shown to interact specifically with isolated plasma membranes from β-cells (Brogren et al., 1986). Therefore, QCM-D experiments were performed with IC2 and sonicated plasma membranes from the β-cell line MIN6 (figure 41 and 42). Plasma membranes from the α-cell line αTC1-6 served as control (figure 40). Also binding to the CD1d-transfected A20-CD1d cell line (figure 43-44) was tested. In these experiments, IC2 was adsorbed to the sensor surface and the sonicated plasma membranes were injected into the flow cell. In solution experiments with FITC conjugated IC2 and non glucose-stimulated plasma membranes isolated from the β-cell lines INS-1E (not shown) and RIN-5AH (figure 54) were performed using MST.

5.1.1.2.1. β-cells 5.1.1.2.1.1. IC2-IgM interaction with α- and β-cells Injection of plasma membrane sonicate from the α-cell line, αTC1-6, initially gave rise to a large shift in frequency and dissipation. As the buffer was changed back to pure PBS, the frequency quickly increased and the dissipation decreased. After app. 1.5 h a stable baseline was obtained with no difference in frequency and dissipation compared to the start of the injection.

71

discussion

The observed frequency shift after rinsing with PBS is low and probably not the observation of plasma membrane vesicles, as these are presumably large and heavy and thus would give rise to a somewhat larger and longer lasting shift both in frequency and dissipation, depending on the vesicle coverage and the size of the vesicles (Tellechea et al., 2009). Instead, what we observe here could be transient interactions with other components from the sonicate such as detached saccharides, protein parts, ions etc. The full reversibility upon PBS wash suggests that this binding is highly transient. IC2 has previous been shown not to bind α-cells so this was expected. Using the same assay, the binding of plasma membrane sonicate from the β-cell line MIN6 was examined (figure 41). After injection of the MIN6 sonicate a frequency shift of approximately 30 Hz, accompanied by a large increase in dissipation, was observed. After extensive rinsing with PBS for 6 h, the total shift after addition of the sonicate amounts to 15 Hz in frequency and 2 units in dissipation. In contrast to αTC1-6 (figure 40) the binding observed for the β-cell line MIN6 sonicate is more persistent. The frequency shift is slightly larger in spite of the differences in concentration (40 mg/ml of αTC1-6 and 30 mg/ml of MIN6). Together this suggests increased binding of MIN6 sonicate than αTC1-6 sonicate, but the bound layer is not vesicles or larger species. In the microscale thermophoresis experiments, no binding was observed between FITC-IC2-IgM and INS-1E vesicles (data not shown). For the RIN-5AH plasma membrane experiments, the results were more ambiguous (figure 54) though they were evaluated as being random noise rather than an interaction. Because glucose stimulation is essential for a high level of IC2-antigen expression (Buschard et al., 1988) , the fact that RIN-5AH and INS-1E were glucose stimulated before the plasma membranes were extracted, could explain the lack of interaction. The ratio of expressed IC2-antigen compared to amount of plasma membrane might simply have been too low to be detected using MST. 5.1.1.2.1.2. IC2-F(ab’)2 interaction with β-cells In a serial connected QCM-D experiment, the outflow of the sensor from the MIN6 experiment in figure 41 was coupled to the inflow of another sensor (figure 42) to reduce the use of plasma membrane sample. Before the serial coupling, IC2-F(ab’)2 was adsorbed on the sensor surface. The smaller change in frequency and dissipation with adsorption of IC2-F(ab’)2 compared to IC2-IgM, is due to differences in molecular weight. The injection of MIN6 plasma membrane sonicate gave rise to a large frequency shift of 95 Hz and an increase in dissipation of 15 units. Together this large shift in both dissipation and frequency indicates a binding event of larger particles such as MIN6 vesicles to IC2-F(ab’)2. The frequency shift that is observed here is large and a little surprising, since only a very low shift in frequency was observed in the experiment where IC2-IgM was immobilized (figure 42). The binding sites of the IC2-F(ab’)2 fragment, are also present in the experiment on figure 41, as IC2-F(ab’)2 is a fragmented version of the IC2-IgM antibody. However, the F(ab’)2 fragment is much smaller than IgM weighing 110 kDa and 900 kDa respectively. Thus, a much larger amount of IC2-F(ab’)2 can be immobilized on the same area than IgM. The F(ab’)2 has a higher amount of the antigen specific binding site compared to its size than IC2, and therefore, we expect to see an increased binding when using the F(ab’)2.

72

discussion

Since this is a serial experiment, the concentration of MIN6 plasma membrane sonicate that reaches the second sensor (figure 42) is potentially lower than the concentration in the first sensor (figure 41). It cannot be excluded that the source of the frequency shift is not really MIN6 plasma membranes, but other molecules, such as the IC2-IgM antibody, that have detached from the previous sensor. What is observed here is interesting, but to be able to conclude further on this, this experiment will have to be repeated without the serial connection. Also, an experiment to show the binding of vesicles to the gold surface coated with Percoll only would be interesting. The current data suggests a persistent binding of MIN6 vesicle to IC2-F(ab’)2.

5.1.1.2.2. A20-CD1d cells Recent experiments using FCM has shown that IC2, besides being specific for pancreatic β-cells, also binds to CD1d-transfected cell lines from mice, such as A20-CD1d. IC2 does not interact with the regular A20 cell line which does not express CD1d (Unpublished data. Appendix I). Therefore, QCM-D experiments were performed with A20-CD1d plasma membrane sonicate (figure 43-44). IC2 was adsorbed on the sensor surface before a flow of A20-CD1d plasma membrane sonicate was injected into the flow cell (figure 43). This gave rise to a frequency shift which after rinsing with PBS amounted to only 10 Hz and no change in dissipation. This is even less than the modest frequency shift observed with IC2-IgM and the MIN6 β-cell line even though the concentrations of the A20-CD1d plasma membrane sonicate were higher than that of the MIN6 sonicate (40 mg/ml and 30 mg/ml, respectively). As it is the case with most of the other plasma membrane experiments, it is not likely that the observed change in frequency is caused by the binding of plasma membrane vesicles. Rather we are observing unspecific interactions between other components from the plasma membrane sonicate with the sensor surface, or specific interactions with components from the A20-CD1d plasma membrane sonicate such as small micelles of lipids and proteins. In order to see if loading of the NKT-I cell activating glycolipid α-GalCer in the CD1d receptors of the A20CD1d plasma membrane could improve IC2 binding, the A20-CD1d plasma membrane sonicate was incubated with α-GalCer prior to injection (figure 44). When the incubate was injected into the flow cell, a frequency shift of only 0-5 Hz and no change in dissipation was observed after rinsing with PBS. Again, it is important to consider that a concentration of only 16.5 mg/ml A20-CD1d plasma membrane with an unknown concentration of Percoll was used in this experiment, compared to 40 mg/ml in the experiment on figure 43 which gave rise to a 10 Hz shift in frequency. However, if loading of α-GalCer in the CD1d molecules on the A20-CD1d plasma membrane should actually be a determinant for binding of the plasma membrane sonicate to IC2, a rather dramatic effect would be expected. That this is not observed could be caused by several factors. Of course, it is possible that the amount of αGalCer used for incubation is not sufficient to load the CD1d molecules in the A20-CD1d plasma membrane which explains why IC2 binding does not increase. Another option is that IC2 binds to α-GalCer and therefore is saturated by the unbound excess of α-GalCer from the CD1d-α-GalCer solution when it is injected over the surface. However, if this would be the case a larger shift in both frequency and dissipation would have been expected. Recently, I learned that the loading of α-GalCer to CD1d molecules requires a slightly acidic pH (Huang et al., 2011). Thus, a good explanation for our observation is that α-GalCer was not loaded in CD1d and therefore no improvement in IC2 binding was observed compared to A20-CD1d sonicate without α-GalCer. Thus further experiments should be performed with correct α-GalCer loading

73

discussion

procedure to determine whether α-GalCer is determinant for the binding of A20-CD1d plasma membrane sonicate to IC2. Also the interaction between IC2 and sonicated α-GalCer should be examined further.

5.1.2. Interaction between IC2, intact cells and plasma membranes Table 7 and 8 provides an overview of the results from interaction experiments between IC2 and intact αand β-cells, and plasma membranes, respectively. Table 7. Overview of binding of IC2 to intact α- and β-cell lines Living Fixated QCM-D CRIA CELISA FCM C-QCM ICC CFM IHC INS-1E β-cells βTC-tet RIN-5AH α-cells αTC1-6 Islet β-cells cells References Current (Desai, (Brogren et (Brogren et Current (Brogren et (Buschard (Brogren et study 2009) al., 1988) al., 1988) study al., 1988) et al., 1988) al., 1988) Table 7. Detection of interaction between IC2 and the given cell types. Green indicates that interaction has been observed. Yellow indicates that interaction may have been observed, but that important controls are missing. Red indicates that no interaction was observed. Grey means that the experiment have not been performed. Methods marked with blue refers to this current project whereas methods with black is from previous studies. C-QCM is cellular quartz crystal microbalance, QCM-D is quartz crystal microbalance with dissipation, CRIA is cellular radioimmunoassay, FCM is flow cytometry, CELISA is cellular ELISA, ICC is immunoperoxidase cytochemical staining, CFM is confocal immunofluorescence microscopy and IHC is immunohistochemistry. Table 8. Overview of binding of IC2 to plasma membranes QCM-D MST INS-1E βTC-tet β-cells RIN-5AH MIN6 α-cells αTC1-6 A20-CD1d B-lymphoma A20-CD1d + α-GalCer References Current study Current study

RIA

ELISA

(Desai, 2009)

(Brogren et al., 1988) Table 8. Detection of interaction between IC2 and the plasma membrane of given cell types. Green indicates that interaction has been observed. Yellow indicates that interaction may have been observed, but that important controls are missing. Red indicates that no interaction was observed. Grey means that the experiment have not been performed. Methods marked with blue refers to this current project whereas methods with black is from previous studies. QCM-D is quartz crystal microbalance with dissipation, MST is microscale thermophoresis, RIA is radioimmunoassay, ELISA is enzymelinked immunoadsorbant assay.

Interactions between IC2 and intact cells were not successfully measured in this project. That no interaction was detected when using the cellular QCM is probably a matter of incomplete fixation of important IC2antigen components. Interactions between these cell lines have been observed several times in the past (table 7) and therefore this should be investigated further. In QCM-D experiments with living cells, as well as with plasma membrane sonicate, interactions were observed, but the size of the shifts in frequency and dissipation implies that it was not interaction between IC2 and living cells nor plasma membranes. Only exception to this is with IC2-F(ab’)2 and the MIN6 β-cells.

74

discussion

The reasons for this could be that the large cell and plasma membrane components are not able to bind in a flow system if the amount of IC2 antigen compared to the total size of the particles is too small. This could in explain why binding is observed with IC2-F(ab’)2 were the concentration of the binding regions of IC2 are more concentrated. If this is the case, interactions would be easier to detect using other methods where a flow is not applied. MST does not apply flow, but in the MST experiments performed here, the plasma membranes that were used were from cells that had not been glucose stimulated cells. Thus, the amount of expressed IC2 autoantigen compared to the total amount of plasma membrane might have been too low to be detected at the given concentrations. In the case of the plasma membrane from the CD1d-transfected A20 cells, the experiment with α-GalCer loading in CD1d should be performed using correct pH conditions in order to show whether binding of IC2 can be improved.

5.1.3. Binding of IC2 to lipids The CD1d receptor loads endogenous lipids in endosomal compartments and are then transported to the surface of the plasma membrane where the CD1d molecule presents the lipid to NKT-cells (De Libero and Mori, 2007). Since IC2 has been found to inhibit the stimulation of NKT-cells from CD1d-lipid complexes (Jensen et al., 2012), it is plausible that such a lipid is part of the IC2 antigen. Therefore, the binding of IC2 to various lipids was investigated (figures 45-52 and 55-59). Since CD1d-molecules are known to bind the lyso-forms of lipids (Roy et al., 2008), also binding of IC2 to the lyso-forms of sphingomyelin (sphingosylphosphorylcholine) and sulfatide (lyso-sulfatide) was investigated using both quartz crystal microbalance with dissipation monitoring (figure 48-49 and 51) and microscale thermophoresis (figure 55 and 56-59). To further support our findings, immunoblotting on Sphingostrips was performed (Invitrogen, California, USA)(figure 60).

5.1.3.1. Sphingomyelin Based on a combination of iTLC, competitive ELISA and confocal fluorescence microscopy (CFM), sphingomyelin was proposed as the antigen of IC2 by Kavishwar et al. (Kavishwar et al., 2011). Therefore, QCM-D experiments were performed with IC2 on the sensor surface and sonicated sphingomyelin in flow (figure 45-47). Also experiments using MST were performed, but no interaction was observed between FITC-IC2 and sphingomyelin using this technique (data not shown). Figure 45 shows the sensogram of the injection of sphingomyelin from bovine buttermilk over an IC2 adsorbed sensor surface. After rinsing with PBS, a frequency shift of 180 Hz and a change of 40 units in dissipation were observed. The frequency and dissipation did not change considerably over the next 1.5 h where the experiment ended. An experiment with reduced concentrations of sphingomyelin showed that the frequency shift, but not the dissipation, was concentration dependent (figure 46). The same pattern was observed in experiments with sphingomyelins from other sources (appendix V). In a control experiment without IC2 (figure 47) the injection of sonicated sphingomyelin did not give rise to a shift in frequency or dissipation. Together these experiments indicate that IC2 interacts with sphingomyelins (from various sources) and that components from control rat serum do not. The affinity of this interaction is yet to be determined. To further support our findings, the interaction between IC2 and sphingomyelin could be performed at a third concentration of sphingomyelin. The size of the vesicles should be examined, preferably using

75

discussion

dynamic light scattering (DLS), since different sizes of vesicles give rise to different shifts in both frequency and dissipation. Further experiments should be performed to examine whether IC2 is specific for sphingomyelin and not the structurally similar sphingosylphosphorylcholine and phosphatidylcholine.

5.1.3.2. Sphingosylphosphorylcholine Sphingosylphosphorylcholine is the lyso-form of sphingomyelin, thus they share polar head groups but differ in that sphingosylphosphorylcholine has only one fatty acid chain. Interaction between FITC-IC2 and sphingosylphosphorylcholine was examined by MST (figure 55) but no interaction was observed. In QCM-D, IC2 was adsorbed on the sensor surface and sonicated sphingosylphosphorylcholine was injected into the flow cell (figure 48). When sonicated sphingosylphosphorylcholine was injected (Figure 48) a frequency shift of 65-70 Hz was observed. As the running solution was changed to PBS, the frequency made a small kink and Figure 61. Proposed rearrangement of sphingosylphosphorylcholine bound to then kept on declining for another 15 IC2 at the change to pure PBS. 1) Sphingosylphosphorylcholine micelles (white) Hz before it flattened out. Similarly, are flushed over the sensor surface with adsorbed IC2 (blue). 2) Upon binding the dissipation rose from 19 to 25 to IC2 the micelles are squished flat. At the change of running buffer, the micelles folds out to its globular shape. 3) The increased volume of the micelles units as the running solution was cause a shift in frequency and the increase in flexibility of the bound layer gives changed to PBS. After this, the a shift in dissipation. The cause of this supposed rearrangement is unknown. frequency started to increase slowly and the dissipation started to drop indicating removal of mass from the sensor surface. In a control experiment without IC2 (figure 49) the sphingosylphosphorylcholine sonicate gave rise to a frequency shift of 30 Hz and a 2 units shift in dissipation. The kink on the frequency curve observed in figure 48 was not observed in the control experiment figure 49. If we do not take this frequency kink into consideration, we observe a frequency shift of app. 75 Hz with IC2 and only 30 Hz in the control experiment. Thus, sphingosylphosphorylcholine micelles interact with IC2. Sphingosylphosphorylcholine has only one lipid chain and forms micelles in polar solution. Micelles are considerably smaller than vesicles, and therefore the size of the frequency shift is expected to be smaller than with vesicles. To see whether this interaction is concentration dependent, further experiments have to be performed. What causes the frequency kink at the transition between lipid sonicate and PBS is unknown, but since this effect comes with a relatively pronounced effect also in dissipation, and since it is not observed in the control experiment, it could be a rearrangement of the bound sphingosylphosphorylcholine micelle layer bound on the IC2 (figure 61).

5.1.3.3. Sulfatide Already in 1988, sulfatide was suggested as antigen of IC2 (Brogren et al., 1988). Since then sulfatide has been suggested to be implicated in the development of type I diabetes in several regards (Buschard, 2011). Because of lack of material and information about the owned sulfatide, only few experiments were performed with sulfatide using QCM-D (figure 59) and microscale thermophoresis. In spite of previous

76

discussion

positive experiments using other techniques such as ELISA and RIA (table 9), no binding was observed between FITC-IC2 and sulfatide in the microscale thermophoresis experiments (data not shown). In QCM-D, IC2 was adsorbed on the sensor surface and sonicated sulfatide was injected into the flow cell (figure 50). The sensogram show how an unknown concentration of sonicated sulfatide gave rise to a frequency shift of a little more than 10 Hz accompanied by an increase of 2 units in dissipation. Since the concentration of sulfatide is unknown, not much information regarding the interaction can be deduced. A frequency shift of 10 Hz and small change in dissipation could be the reflection of a binding, however both the shift in frequency and dissipation are expected to be larger with the binding of vesicles (Tellechea et al., 2009). The experiment should be repeated with a higher concentration of sulfatide to see if a larger shift in frequency and dissipation can be obtained.

5.1.3.4. Lyso-sulfatide In this project, lyso-sulfatide from two different sources was used: Lyso-sulfatide from Sigma and Lysosulfatide from Matreya. Interactions between IC2 and the Matreya lyso-sulfatide was examined using QCMD (figure 51). The interaction between FITC-IC2 and both the Sigma lyso-sulfatide and the Matreya lysosulfatide was examined by MST (figure 56-59).

5.1.3.4.1. Sigma lyso-sulfatide Using MST, interaction between FITC-IC2 and Sigma lyso-sulfatide was observed (figure 56). With a simple KD fit (NanoTemper Software) the KD value was determined to 1 µM. This fit, however, was made with the supposed concentration of 20 nM of FITC-labelled IC2 and not 25 nM as it should have been. This was not possible to redo, since the data unfortunately was lost due to exchange of computers. Since the concentration of the labeled protein affects the KD value, a KD value calculated with 25 nM instead of 20 nM would have given a slightly lower KD value, so the real affinity estimate of this interaction is slightly lower than the otherwise calculated 1 µM. The measurement of the interaction between FITC-IC2 and the Sigma lyso-sulfatide (figure 56) was attempted replicated. The experiment was not directly replicable (data not shown) using the experimental conditions used in the experiment on figure 56. One of the reasons for that could be that only small amounts of this Sigma lyso-sulfatide were available and that the product had been discontinued (personal communication, Sigma-Aldrich). This made it impossible to obtain the required high antigen concentrations. A pattern resembling interaction was observed using other experimental conditions, with only 0.05 % tween-20 in the dilution buffer (figure 57 and 58). In both of these experiments, the antigen concentrations was not high enough to obtain the full s-shaped curve and this together with other parameters, resulted in that these experiments are evaluated as not showing an interaction. By mistake, the samples in one of these experiments (figure 58) were left to incubate for only 10 min, which clearly could have a major impact on the results. Unfortunately there were not enough of the Sigma lysosufatide left to repeat this experiment with longer incubation time.

5.1.3.4.2. Matreya lyso-sulfatide Instead, interactions between FITC-IC2 and a lyso-sulfatide from Matreya was sought measured using MST with starting concentrations ranging from 10000 nM to 1 mM with and without 0.05 % tween-20 in PBS. No binding was observed (data not shown). Using the labelfree MST, Monolith NT.Labelfree (NanoTemper Technologies GmbH, Germany), which uses the intrinsic fluorescence of a molecule to follow the thermophoresis, the interaction between the Matreya

77

discussion

lyso-sulfatide and unlabeled IC2 was measured (figure 59). The fluorescence curves are rather bumpy and several curves cross each other which indicate poor data quality. Thus, the drop of 14 fluorescence units is evaluated to not being enough to indicate an interaction, and therefore this experiment should definitely be repeated. Because of problems with different versions of the NanoTemper Software, we were not able to obtain the curve of temperature jump versus antigen concentration, or fluorescence versus antigen concentration, which is very unfortunate. In QCM-D experiments, IC2 was adsorbed on the sensor surface and sonicated Matreya lyso-sulfatide was injected into the flow cell (figure 51). The sonicated lyso-sulfatide initially resulted in a frequency shift of 35 Hz but as soon as pure PBS was flushed into the cell, a large desorption occurred, and the frequency stabilized after 1 h at the IC2 starting frequency of -65 Hz. Thus, no binding of lyso-sulfatide to IC2 was observed at the given concentration of lyso-sulfatide. Since a lasting larger drop in frequency is initially observed that terminates as soon as the lyso-sulfatide solution is exchanged by PBS, and thus quickly return to the IC2 frequency, the concentration of lyso-sulfatide is thought to be large enough not to limit the interaction. The steep dive in frequency could be reflecting the presence of large lipid micelle or gel species which could be a result of an incomplete sonication. This experiment was reproduced and gave consistent results. Together these results indicate that the interaction between IC2 and lyso-sulfatide is highly dependent on the exact lyso-sulfatide used.

5.1.3.5. Other lipids 5.1.3.5.1. Sodium cholesteryl sulphate Kavishwar et al (Kavishwar et al., 2011) found the presence of cholesterol in the plasma membrane to be necessary for the binding of IC2 to β-cells. To elucidate whether IC2 binds to cholesterol itself, QCM-D (figure 52) and microscale thermophoresis experiments was performed. No binding was observed between FITC-IC2 and sonicated sodium cholesteryl sulphate using MST (data not shown). IC2 was adsorbed on the sensor surface and sonicated sodium cholesteryl sulphate was injected into the flow cell (figure 52). The flow of sonicated sodium cholesteryl sulphate gave rise to a frequency shift of 20 Hz and a change of approximately 5 units in dissipation. As the experiment was terminated total shift in frequency had halved. Thus, sodium cholesteryl sulphate might bind to IC2, but as in the case of sulfatide, this should be further elucidated by performing the experiment with higher concentrations of sodium cholesteryl sulphate. Furthermore, a control experiment with rat serum components instead of IC2 is needed.

5.1.3.5.2. β-galactosylcerebroside Also β-GalCer was tested using both QCM-D and MST. No interaction was observed with this glycolipid (data not shown).

5.1.3.6. Combinations of lipids The complex composition of lipids in the cell membrane inspired to do experiments with FITC-IC2 and a combination of certain lipids. Kavishwar et al suggested the presence of cholesterol in the plasma membrane as a necessary determinant for the binding of IC2 to β-cells. To test whether the presence of

78

discussion

cholesterol would affect the binding of IC2, MST experiments were performed with FITC-IC2 and a sonicated mixture of sphingomyelin and sodium cholesteryl sulphate, and lyso-sulfatide and sodium cholesteryl sulphate. No binding was observed in any of the experiments (data not shown), but it is likely that a certain ratio of cholesterol is needed to obtain the right stabilizing environment. These experiments should also be performed in QCM-D where interaction between sphingomyelin and IC2 has been observed.

5.1.3.7. Immunoblotting To support our findings, an IC2 blotting assay was performed using the commercially available SphingoStrips (Invitrogen, California, USA) (figure 60). A clear staining is seen in the sphingomyelins spot, whereas no staining was observed in any of the other spots, including sulfatide, sphingosylphosphorylcholine and phosphatidylcholine. Due to instrument malfunction, no control experiment has been performed yet. Kavishwar et al. (Kavishwar et al., 2011), also supported their findings of IC2-recognition of sphingomyelins by immunoblotting on SphingoStrips. While using a more sensitive substrate, they found that IC2 bound to sphingomyelin and phosphatidylcholine. We do not see the binding to phosphatidylcholine, but this could be because of the less sensitive blotting substrate. Since sphingomyelin and phosphatidylcholine share the same polar head group, IC2 interaction to phosphatidylcholine would mean that the binding of IC2 is independent of the hydrophobic part of the lipids. If this is the case, it is surprising that Kavishwar et al. do not observe binding to lyso-phosphatidylcholine as well. It could be because the binding of IC2 requires that these lipids are oriented in a special way. This may occur for certain lipids only when they are held in micelles and vesicles, or when they are complexed with a lipid binding protein such as the CD1d receptor. That would explain why interaction is not observed with immunoblotting.

5.1.3.8. IC2 seem to recognize several diverse lipids Table 9 provides an overview of the results from interaction experiments between IC2 and lipids. Table 9. Overview of IC2 binding to lipids QCM-D MST IB SM LSM SUL S.LSU M.LSU LSU SCS β-GalCer α-GalCer PC Current Current Current References study study study

IB

iTLC

iTLC

RIA

Comp. ELISA

NKT inhib

(Kavishwar (Spitalnik, (Kavishwar (Spitalnik, (Kavishwar (Jensen et et al., 1989, Mia, et al., 2011) 1989) et al., 2011) al., 2012) 2011) 2009) Table 9. Detection of interaction between IC2 and lipids. Green indicates that interaction has been observed. Yellow indicates that interaction may have been observed, but that important controls are missing. Red indicates that no interaction was observed. Grey means that the experiment have not been performed. Methods marked with blue refers to this current project whereas methods with black is from previous studies. QCM-D is quartz crystal microbalance with dissipation, MST is microscale thermophoresis,IB is immunoblotting, iTLC is immune thin layer chromatography, RIA is radioimmunoassay, Comp.ELISA is competition enzyme-linked immune sorbent assay, NKT inhib is NKT inhibition studies.

79

discussion

None of the examined lipids have shown IC2-reactivty using all methods. When looking at the overview of previous and current findings (table 9) especially the interaction between IC2 and sphingomyelin has been found repeatedly. In QCM-D, sphingomyelin from various sources showed a clear binding and this series of experiments, are in my opinion the most reliable and best supported in this project, since the interaction is concentration dependent and show distinct differences between the positives and the control. Surprisingly, this interaction was not observed with MST. Our experiments show that IC2 binds to both sphingomyelin and its lyso-form, suggesting that the interaction depend on the polar head group rather than the fatty acid chains. Also the immunoblotting experiments by Kavishwar et al. where they find binding of IC2 to both sphingomyelin and phosphatidylcholine, suggest that the interaction does not depend on the hydrophobic part of the lipid. Also the interaction between IC2 and sulfatide has been described several times, but not conclusively in this project. In this project the sulfatide material was very limited and this could be the reason why no interaction is observed. Interactions between IC2 and lyso-sulfatide was observed using MST, but only with the Sigma lyso-sulfatide. This indicates that small differences in the structure of the lyso-sulfatide e.g. charge position or length of fatty acid chain, strongly affects the interaction with IC2. To further investigate the binding of IC2 to lyso-sulfatide, experiments could be performed with a broader range of lysosulfatides. Another reason for the observed differences in binding of IC2 to the different types of lysosulfatide, could be that these lipid solutions are not totally pure and that certain impurities could be present. These impurities might give the Sigma lyso-sulfatide a unique feature that is not present in the Matreya lyso-sulfatide. To examine whether the lipidbinding CD1d receptor is involved in the interaction between IC2 and lipids, experiments with lipids in combination with CD1d should be performed. The interaction between IC2 and lipids might be dependent on a certain stabilization or orientation of the lipid, and not necessarily the CD1d receptor molecule, and therefore further experiments with lipids in combination with cholesterol should be performed using both MST and QCM-D.

5.1.4. IC2 binding to sulphated monosaccharides When a lipid is bound in a CD1d molecule, the lipid tail is buried deep within the hydrophobic binding groove of the CD1d receptor (Zajonc et al., 2003). Thus, the only part of the lipid that is actually exposed to the solvent and thus accessible for interaction is the polar head group. Likewise, when lipids are in a polar solution, micelles or vesicles will form, and the only part exposed to the polar solvent is the polar head group. Since IC2 has been found to interact with sulphated glycolipids, we found it relevant to investigate the binding of IC2 to various sulphated monosaccharides. It has previous been shown by RIA that sulfurcontaining glycolipids such as galactose-sulphate is determinant for binding of IC2 (Brogren et al., 1989). Therefore, the binding of IC2 to D-glucose-3-sulphate, D-galactose-3-sulphate, D-galactose-4-sulphate was investigated using QCM-D (figure 53) and microscale thermophoresis. No binding was observed between FITC-IC2 and any of the sulphated monosaccharides in the microscale thermophoresis experiments (data not shown). In QCM-D, IC2 was immobilized on the sensor surface and the sulphated monosaccharides

80

discussion

were injected into the flow cell figure 53. Neither D-glucose-3-sulphate, D-galactose-4-sulphate nor Dgalactose-3-sulphate was bound by IC2.

5.1.4.1. IC2 does not bind to sulphated monosaccharides alone Table 10 provides an overview of the results from interaction experiments between IC2 and sulphated monosaccharides. Table 10. Overview of IC2 binding to sulphated monosaccharides No interaction between IC2 and these QCM-D MST RIA sulphated monosaccharides was detected D-Glu-3-Sul using QCM-D and MST. Since the lipids D-Gal-3-Sul D-Gal-4-Sul found to be bound by IC2 have extensive References Current study Current study (Brogren et al., 1989) variations in the hydrophobic regions, the Table 10. Detection of interaction between IC2 and sulphated monosaccharides. Green indicates that interaction has been observed. part of the lipids thought to be implicated in Red indicates that no interaction has been observed. Methods marked IC2 binding is the polar head group. In with blue refers to this current project whereas methods with black is from previous studies. QCM-D is quartz crystal microbalance with sulfatide and lyso-sulfatide, this polar head dissipation, MST is microscale thermophoresis, RIA is radioimmunoassay. group is exactly D-galactose-3-sulphate. Therefore the results are a little surprising, but it can be concluded that these sulphated monosaccharides are not enough for IC2 recognition when they are not attached to a lipid.

5.2. Evaluation of methods The results of this study and the comparison with previous findings have been thoroughly described in the previous sections. The methods used in this project differ markedly from those earlier used to detect interactions between IC2 and its antigen, in that they are able to directly obtain kinetics from the measured interactions. This easy access to kinetics was also the basis for the choice of exactly these methods with the aim of comparison of potential autoantigens based on the affinity of the interactions. The estimated KD value was easily derived from the interactions detected by MST. Unfortunately, the software for calculation of kinetics in the QCM-D was not as well developed as initially thought and therefore deriving the kinetics from the interactions measured by QCM-D is still a task that has to be fulfilled. Kinetics could also easily be derived from the cellular QCM studies, but unfortunately no binding of IC2 was observed using this technique. An issue that complicates the outcome of this project, is that the antigen of IC2 is unknown and therefore it is of course difficult to do affinity measurements between the antibody and the antigen. This is particularly problematic, because we do not have a positive control when new experimental settings are tested. Therefore, antigens previously shown to bind to IC2 such as intact cells and plasma membrane preparations were used as positive controls in this project. Using kinetic parameters such as KD (on- and off-rate) to compare the affinity of IC2 towards lipids and lipid specimens relative to each other, is only possible if the size and distribution of micelles and vesicles are uniform, and if they consist of the same amount of potential IC2-antigens. A matter that is hard to resolve, when the antigen is unknown. Because of the difference in size, it definitely matters whether the sonicated lipids have formed micelles or vesicles, since the signal when using the QCM-D method is weight

81

discussion

dependent. The size and distribution of the product of the sonication can and should have been checked with DLS. In spite of the methods chosen, we ended up with only few conclusive results. It is very important to keep in mind that experiments done in vitro rarely corresponds to what actually happens in vivo. Therefore, it is important to try to replicate the in vivo conditions as much as possible. In this project, the experiments were performed at 37 oC to mimic the physiological settings. The buffer used was primarily PBS which is a buffer with physiological pH and salts. With that said, there is still a long way from working with sonicated lipids in flow system, to what actually happens in vivo. Because of these issues, this way of searching for the IC2-antigen by affinity measurements might not be the right path when the supposed antigen is a lipid. If the potential suspected antigen had been a protein (only) this method of determining the right autoantigen by an affinity-screening project, would probably have been a better choice. The IC2 autoantigen has previously been elucidated by purification and iTLC but the results from this study are not consistent with the current observations, so something must be missing.

5.3. The IC2 antigen 5.3.1. The molecular characteristics of IC2 recognized lipids 5.3.1.1. What do IC2 recognized lipids have in common? When comparing the lipids that IC2 recognizes on a molecular level, they are all amphipatic. Besides from that, they display some significant differences. Lipids found to interact with IC2 are: sulfatide and lyso-sulfatide, sphingomyelin and sphingosylphosphorylcholine, the exogenous α-GalCer observed with previous NKT-cell inhibition experiments (Jensen et al., 2012), sodium cholesteryl sulphate, and phosphatidylcholine. Table 9 shows the experimental basis for these discoveries. The molecular structures of these lipids are shown in the introduction in section 1.6.3.

5.3.1.1.1. Hydrophobic part of the lipids The lipid lyso-forms (lyso-sulfatide and sphingosylphosphorylcholine) have the characteristics of only one fatty-acid chain, whereas sphingomyelin, sulfatide, phosphatidylcholine and α-GalCer has two fatty acid chains. The lengths of these fatty acid chains can differ greatly. The hydrophobic part of phosphatidylcholine differs from the others in that it is glycerolbased. Sodium cholesteryl sulphate has an even more deviant structure. Instead of a long carbohydrate tail, it has a group of carbohydrate rings which makes it hydrophobic. Together these observations indicate that the number, length and composition of the fatty acid chain play a minor part in IC2 recognition. Since the binding of IC2 to these glycolipids has all been observed in experiments in polar solutions where the glycolipids are thought to form either micelles or vesicles and thus is inaccessible for IC2, this is not surprising.

5.3.1.1.2. Polar head group of the lipids When discussing the molecular differences between the polar head groups of these lipids, it is important to remember that the lyso-forms of the lipids, in this case lyso-sulfatide and sphingosylphosphorylcholine only

82

discussion

differ in the number of fatty acid chains, and thus has the same head groups as those of sulfatide and sphingomyelin, respectively. At a first glance, there are some important structural similarities between α-GalCer and sulfatide. They both have a hexagonal carbohydrate ring linked to the sphingosine base by an ether bond. A crucial difference between these two is the orientation of the carbohydrate ring, which is α-linked in α-GalCer and β-linked in sulfatide. The β-linked version of α-GalCer is β-GalCer. No interaction between IC2 and β-GalCer has been observed. We know that the difference in orientation structurally can be very important, since this is one of the differing points in NKT-cells CD1d-glycolipid complex recognition where exactly the orientation of carbohydrate group (α-GalCer vs. β-GalCer and sulfatide) results in big differences in stimulation of NKTcells (Bendelac et al., 2007). Another important difference is that sulfatide contains a sulfur group in position 3’ of the carbohydrate ring. The presence of the sulphate group adds a negative charge to sulfatide. This, of course, is the same with lyso-sulfatide. Since IC2 binds to both α-GalCer and sulfatide, but not to β-GalCer, it seems that the charged sulphur group subordinates the orientation of the carbohydrate. Sphingomyelin and phosphatidylcholine differ from the others in that they have completely different head groups both functionally and structurally. The sphingomyelin and phosphatidylcholine head group consists of a choline linked by a phosphate group to the hydrophobic part of the lipid. The choline carries a negative charge. Again, this is the same with sphingosylphosphorylcholine. Thus both sphingomyelin and sulfatide are potentially charged but with a positive and negative charge, respectively. Also the sodium cholesteryl sulphate carries a negative charge by virtue of its sulphate group. Apparently, the presence of charge is not a point to which IC2 distinguishes binding. This could be because ions from buffers neutralize the charge prior to binding. If there is a difference between the affinity of the IC2-lipid interactions, it could readily be caused by the differences in charge as these could somehow mediate or strengthen the interaction.

5.3.1.1.3. IC2 might recognize a certain structure rather than certain sequences As described in the previous section, there are apparent differences between the lipids that IC2 recognize. Thus it is likely that IC2 recognizes a certain structural pattern rather than a certain sequential feature or functional group. The polar group of all of the lipids extrudes from the hydrophobic part to a greater or lesser extent, and thus has a certain spatial structure. Cholesteryl Sulphate and phosphatidylcholine differs from the other lipids in they do not bind in CD1d. From this comparison of the molecular parts of the IC2 bound lipids, other lipids such as sulphated αGalCer, glucosylceramide (GlcCer) and glycolipids with smaller extruding carbohydrate groups are suggested as potential targets of IC2 as well.

5.3.1.2. The lipid chain seems to be essential for IC2 recognition In this project, IC2 was not found to bind to the sulphated monosaccharides D-galactose-3-sulphate, Dgalactose-4-sulphate and D-glucose-3-sulphate. Since the polar head group on sulfatide is D-galactose-3sulphate, this suggests that also a lipid part is crucial for IC2 binding. This dependency does not have to be directly implicated in the IC2 recognition, but could instead contribute to the binding by stabilizing the actual IC2 antigen epitope in vesicles or micelles, or in the case of the CD1d receptor, by enabling display of the polar head group.

83

discussion

5.3.2. Lipids as targets for IC2 - but something is missing As depicted earlier (table 9) several different methods has confirmed that IC2 binds to lipids. Besides, IC2 is known to be specific to the β-cells, but this does not immediately add up. Several of these lipids are very common in cell plasma membranes all over the body, and sphingomyelin is dispersed throughout the body since sphingomyelin constitutes the membranous myelin sheaths that surround nerve cell axons (Shier et al., 2009). How can an antibody that binds to lipids such as sphingomyelin and sulfatide be specific for βcells along with only few other cell populations? And why do we see binding to these lipids in vitro but not in vivo? Remember IC2 can be used to image the β-cell mass in vivo. Could this be explained by the fact that IC2 binds with a relatively low affinity to these lipids, but with a high affinity if another component is involved in the interaction? The previously observed high affinity of IC2 towards intact β-cells using a CRIA of 18 nM (figure 7) and the cellular QCM pilot experiment (figure 8) of 1.39 nM is of high affinity, whereas the affinity observed with lyso-sulfatide using MST is of medium affinity with a KD of only 1 µM. The affinity of the interaction between IC2 and sphingomyelin has not been derived yet. This difference in affinity could be caused by the involvement of a so far unknown additional molecule.

5.3.2.1. IC2 might recognize a special protein-lipid complex in vivo Our theory is that the lipid constitutes part of the IC2 autoantigen together with a lipidbinding protein. The latter is based on the trypsin sensitivity of the IC2 antigen (Brogren et al., 1986). This protein may be necessary for IC2 to obtain high affinity binding because the proximity of certain amino acid residues from the protein increases the strength of the interaction. It is imaginable that IC2 recognizes an antigenic epitope similar to those shown in figure 13 and 14. This lipid binding protein could very well be CD1d as indicated by recent findings (appendix I)(Jensen et al., 2012, Voetmann, 2013). Another factor contributing to improved binding of IC2 to a CD1d-lipid target, could be that the lipid has to be held in a certain position and have a certain orientation in order to provide a high affinity binding epitope. The CD1d receptor loads its lipids in the endosomal compartment and the CD1d-lipid complex is then transported to the surface of the plasma membrane situated in a lysosomal compartment or exosome vesicle (Bendelac et al., 2007). As the vesicle fuses with the plasma membrane to release its cargo out into the extracellular matrix the CD1d-lipid complex becomes situated on the plasma membrane, possibly in a lipid raft stabilized by other lipids such as cholesterol (Lang et al., 2004). When located here at the surface of the plasma membrane, the CD1d receptor function largely as a MHC molecule by displaying endogenous lipids to NKT-cells. The NKT-cells respond to the display by binding to an epitope involving the CD1d receptor and the polar head groups (Girardi and Zajonc, 2012). Exactly how these groups are oriented and which parts of the molecule that are accessible for the NKT-cell TCR depend on the structure of the lipid, in particular the number of lipid chains (only one for lyso-forms) and their length, but also the orientation and properties of the polar head group, including the presence of functional and/or charged groups (Girardi and Zajonc, 2012). By virtue of the inhibitory effects of IC2 on the stimulation of type I and type II NKT-cells from the CD1d-αGalCer and CD1d-sulfatide (Jensen et al., 2012), we believe that IC2 recognizes, if not exactly the same

84

discussion

epitope as NKT-cells, then parts of the same epitope. Thus, when IC2 binds to its antigenic epitope on the surface of the β-cells, it causes IC2 to block the binding of the NKT T-cell receptor to the CD1d-lipid complexes, which causes the NKT-cells to secrete another cocktail of cytokines and interleukins than it would otherwise have done if it had received stimulation from the cells.

5.3.2.2. Could IC2 bind to a slightly different CD1d expressed only on β-cells? It has recently been shown by FCM (Voetmann, 2013) that CD1d-receptors are in fact located on β-cells. An obvious thought is that the CD1d expressed on the surface of β-cells differs slightly compared to CD1ds expressed by other cells, and that this could explain the IC2 β-cell specificity. However, the binding of IC2 to CD1d-transfected A20 and RAW cells (Unpublished data. Appendix I) argues against this, as the sequences used for this transfection are most likely not originally derived from a β-cell CD1d, and therefore there is nothing β-cell specific about it.

5.3.2.3. A specific lipid coupled to insulin production and/or secretion could be the IC2 antigen 5.3.2.3.1. A hitherto unknown lipid could be the IC2 antigen The β-cells have the special property of producing insulin. Insulin is produced in the endoplasmic reticulum and is then dimerized into hexamers which are crystallized in the presence of zinc and stored in secretory vesicles until they are transported to the plasma membrane as the insulin is secreted (Ashcroft and Ashcroft, 1992, Dunn, 2005). It is conceivable that a special lipid, it could be a byproduct of this insulin production or one that it is somehow involved in the storage or secretion of insulin, perhaps a lipid that is not yet known, could be bound in the CD1d receptor in the endosomal compartment and that this CD1dlipid complex that is expressed on the surface of the β-cells is special in such a way, that it differs from the CD1d-lipid complexes that are otherwise present on the other cell types in the body. IC2 is a marker of insulin secreting β-cells, though the recognition is not to insulin itself. Thus, at least some component of the IC2-antigen must be specific for β-cells only. A number of microbial glycolipids have been identified as binders of CD1d, however the self lipids largely remain unknown, and it is therefore not totally unimaginable that a hitherto unknown lipid could have stimulating effects on the NKT-cells (Sørensen et al., submitted).

5.3.2.3.2. Binding might depend on binding of a certain ion to charged regions of the lipid complex Another possibility is that IC2 recognizes an epitope consisting of the CD1d-receptor with a bound lipid that has a certain property, that makes it specific for insulin secreting β-cells only. Sulfatide, which is known to be an endogenous self-lipid and a stimulator of type II NKT-cells (Bendelac et al., 2007), has been shown to have the ability to bind calcium enabled by the negative charge provided by the sulphate group (Koshy et al., 1999, Merten et al., 2003). Often this cation binding site is occupied by something that is readily available both intracellularly and extracellularly such as calcium. In the process of insulin storage, it has been shown that zinc is an important stabilizing agent mediating crystallization (Osterbye et al., 2001). Zinc is located inside the insulin storage vesicles as it is transported out to the plasma membrane surface. As insulin is secreted extracellularly, it is possible that these zinc molecules are bound by plasma membrane located sulfatides or other similar lipids with a cation binding site. Thus, a sulfatide with zinc bound in the cation binding site could be a signature for the active insulin producing β-cell and perhaps something the IC2 would bind to.

85

discussion

Also sulfatide plays an important role in insulin storage as sulfatide both promotes the folding of insulin, and mediates crystallization and the subsequent monomerization. Beside insulin, the insulin storage vesicle may also contain different extracellular components of the plasma membrane such as lectins, integrals and perhaps also the CD1d molecule which is known to be recycled (De Libero and Mori, 2007). It is thus imaginable that CD1d complexes with endogeneous sulfatide or a similar lipid, could be loaded with zinc in the secretory vesicle. When the CD1d-sulfatide-Zn2+ complex reaches the plasma membrane, it would constitute a β-cell specific NKT-stimulatory, and perhaps IC2-binding epitope, unique for insulin secreting βcells. The binding of IC2 to CD1d-transfected cells (A20 and RAW) has recently been shown by FCM (Unpublished data. Appendix I) weakens this hypothesis. The A20 is a B-cell lymphoma cell line, and it therefore appears to have nothing in common with β-cells and therefore it is highly interesting that IC2 binds to the CD1dtransfected cells. That we observe a binding of IC2 to the CD1d-transfected cells show that the IC2 antigen must also be present on the surface of this B-lymphoma cell line (A20) as well as the macrophage cell line (RAW) when CD1d is expressed. A possible explanation for this consistent with the current hypothesis is that the CD1d in the transfected cells is loaded with this special β-cell specific lipid from the growth media and the fetal calf serum in which it is grown. Other explanations for this are sought.

5.3.3. The NKT inhibition experiments reveals details about the IC2 binding to CD1d complexes In the NKT inhibition experiments, IC2 is found to inhibit the secretion of IL-2 from NKT-cells in vitro. For a more thorough description, view section 1.3.3. and appendix II. From these experiments it can primarily be deduced, that IC2 under the conditions used in that experiment, must be able to bind to both α-GalCer and sulfatide, which are the lipids used in these experiments. Again, this raises the question of why the IC2 is found to be β-cell specific in vivo, when a glycolipid such as sulfatide, which is dispersed throughout the body especially in the brain, is not at all specific towards β-cells, but still we see that IC2 binds to this in vitro. Since this experiment is in vitro, and under ordered conditions with relatively few steps, no zinc or any other unordinary cations are present in the experiment and thus the glycolipids such as the sulfatide, is not “β-cell specific” as it is hypothesized. Thus, that IC2 actually inhibits the NKT-cells activity by preincubation with sulfatide and α-GalCer is surprising. The IC2 binding towards these lipids alone must be strong enough to almost totally prevent binding of NKTcell TCR receptors from the highly potent stimulatory target, the α-GalCer-CD1d complex. In this case, the inhibition of the NKT-cells is most effective, when IC2 was preincubated with α-GalCer before it is added to the CD1d receptor, but also some inhibition was seen without the preincubation, which indicates that IC2 binds more readily to the CD1d receptor when IC2 is already bound to the glycolipid, than if IC2 has to bind to the glycolipid loaded CD1d receptor. Two different scenarios could be the cause of this inhibition. Either, IC2 binds to more or less all of the glycolipids in a way that abolishes the ability of loading of glycolipids into the CD1d. No glycolipids are loaded in CD1d, leading to no stimulation of NKT-cells. Another option is that IC2 binds to the glycolipids in a way that loading of glycolipids to the CD1d complex is still possible. Thus, the IC2-glycolipid complex is

86

discussion

loaded in the CD1d, forming a CD1d-glycolipid-IC2 complex. The IC2 hinders the binding of the TCR to the glycolipid-CD1d complex, which leads to no stimulation of NKT-cells. Which of these scenarios that occur must depend on the attractive forces involved in the loading of glycolipid to CD1d molecules. The IC2-IgM used in this experiment is a large molecule and the binding of this to a small glycolipid could have a huge impact on the onward progress. Thus, I find it most likely that the first scenario occurs, depicting that the glycolipid is simply not loaded to the CD1d receptor and that this is the cause of the observed inhibition/lack of stimulation of NKT-cells. Assuming that the second scenario is the actual one, then it is possible that the binding of IC2 in vivo also takes place in this order, and that could explain why it is possible to also measure binding between IC2 and glycolipids without the presence of CD1d. In the case of IC2 binding in vivo, it is possible that IC2 binds loosely to a specific lipid on the surface of the β-cells e.g. a zinc-loaded sulfatide, and that this lipid-IC2 complex then diffuses around in the plasma membrane and into a CD1d receptor, where the lipid exchanges with the already bound glycolipid. When the CD1d complex comes in proximity of the IC2 glycolipid complex, this exchange occurs, because also attracting forces from IC2 works towards the formation of a CD1d-glycolipid-IC2 complex. These attracting forces could be caused by the attraction between certain amino acids from IC2 and the CD1d complex.

5.3.4. IC2 and the β-cell specificity 5.3.4.1. IC2 may have a higher affinity for CD1d-lipid complexes carrying a special β-cell imprint Our hypothesis is that IC2 recognizes a neoautoantigen consisting of the CD1d receptor loaded with a lipid which has a feature that makes it unique for insulin secreting β-cells. Such a feature could be a sulfatide with a zinc-ion bound in its cation binding site, or a hitherto unknown lipid. This hypothesis is weakened by the binding of IC2 to the CD1d-transfected cells. Also in the NKT inhibition experiment, IC2 is found to block the stimulation of NKT-cells both in the case of α-GalCer and in the case of sulfatide. Lipids which in this case are not β-cell specific in any way. α-GalCer is not an endogenous lipid and even though IC2 displays an apparently high affinity for α-GalCer in this experiment, it is not naturally present in mammals. With the sulfatide, it is different. Sulfatide is present in vivo and therefore IC2 would be expected to recognize CD1dsulfatide complexes all over the body and not just on β-cells. But since IC2 can be used for β-cell imaging, something must be missing. The inhibitory effect of IC2 on NKT-cells are not as high with sulfatide as with α-GalCer. Only about 20-40 % inhibition is seen compared to the app. 98 % inhibition with α-GalCer. Given that the effect of inhibition is an indirect measure of the affinity of the interaction between IC2 and CD1d-lipid, a relatively weak affinity of IC2 towards the CD1d-sulfatide could have the effect of only a temporary binding of IC2 in vivo, and that a binding of IC2 to a β-cell specific CD1d-sulfatide e.g. with zinc would give a higher affinity, and thereby retain the antibody much longer. It has been shown that IC2 is optimal to use in vivo after approximately 72 hours after injection, since by then the excess of the IC2 coupled to a tracer has been removed from the blood (Wismann, 2010). It is possible to imagine, that a binding to CD1d-sulfatide complexes to other cell types occurs within these 72 hours but that the affinity of this interaction is not high, causing IC2 to detach

87

discussion

and be cleared from the system together with the unbound IC2. Left after 72 hours is then only the IC2 that with high affinity has bound to the β-cell specific CD1d-sulfatide+Zn2+ or similar CD1d-lipid-ion complexes on the surface of the pancreatic β-cells.

5.3.4.2. Lipids as targets of future therapeutics Given that IC2 might have the ability to block for NKT-cell stimulation, either generally or to some degree, the presence of IC2 or other IC2-like antibodies could alter the general signaling of NKT-cells in an individual. Exactly what the effect of this could be would only be a guess, since the effect of NKT-cell released cytokines and interleukins on other cells in the immune system such as helper T-cells type I and II (TH1 and TH2, respectively) are still relatively unresolved (Sørensen et al., submitted). It is possible that a lack of NKT-cell stimulation could lead to an increased Th1 cell population, which could result in a generally more aggressive population of cytotoxic T-cells (figure 1). This would definitely increase the risk of autoimmunity. Thus, the presence of IC2 or IC2-like antibodies might potentially increase autoimmunity. Given the location of the IC2-antigen, this autoimmune destruction could lead to type I diabetes. If this is the case, then an antiidiotypic antibody raised against the complementarity determining regions (CDRs) on these antibodies could bind and retain IC2-like antibodies before they were to reach their target on the β-cell surface, thereby eliminating the inhibitory, and potentially autoimmune, actions of these antibodies. In an ongoing study, such antiidiotypic antibodies are being produced and we will soon be able to see whether these will have an effect on the incidence and development of type I diabetes in experimental animals.

88

conclusions and perspectives

6. Conclusions and perspectives 6.1. Conclusions In this thesis, interactions between IC2 and intact cells, plasma membrane sonicates, glycolipids and sulphated monosaccharides were sought measured. IC2 was found to bind to lyso-sulfatide, sphingosylphosphorylcholine and sphingomyelin. By theoretical comparison of the suggested IC2 glycolipid targets, it is suggested that IC2 interaction seems to be spatially specific for lipids with smaller extruding carbohydrate or choline groups. IC2 recognition does not seem to rely on charge nor the presence of certain functional groups. It is suggested that the natural autoantigen of IC2 is a formation of a special CD1d-lipid complex on the surface of pancreatic β-cells. This glycolipid is specific for pancreatic β-cells either in involving a hitherto unknown lipid of β-cell origin, or carrying a special β-cell imprint such as a bound zinc-ion. The affinity of this β-cell specific complex is proposed to be higher than binding of IC2 to other CD1d-lipid complexes, thus the IC2-CD1d-lipid interactions are lost within 72 hours, after which IC2 can be used as a marker for imaging of the β-cell mass.

6.2. Perspectives and future experiments 6.2.1. Determining the IC2 autoantigen 6.2.1.1. Investigation of interactions Though a lot of results have been obtained in this project, there is still a lot to do. A certain amount of follow-up on this project is possible to obtain a higher degree of security of the obtained results. Another control with the flow of phosphatidylcholine instead of sphingomyelin would be relevant to show that the observed reaction with sphingomyelin is actually specific to the ceramide based sphingomyelin. Also, the sonication of the sphingomyelin samples should be checked by DLS to make sure that vesicles are formed and that they are homogeneous in size – also in between the different lipid species to allow for comparability. Furthermore, experiments which access the interaction between α-GalCer-loaded A20-CD1d plasma membranes should be repeated at the right loading conditions. Also an experiment of IC2 and αGalCer could be performed to elucidate the binding of IC2 to this glycolipid alone. Another important experiment would be to do affinity measurements between IC2 and sulfatide incubated with zinc. A major unexplored treasure chest in this project is the data curves from the QCM-D experiments where the affinities of the interactions remains to be further addressed and calculated. This is not an easy task since this monoclonal IgM molecule contains no less than 10 identical binding sites, but as discussed in the introduction, a relative basis for comparison of the affinity of IC2 towards different lipids could be obtained using a standard affinity calculation for only one binding site. This would yield data that are comparable only to each other, but this would still be useful to further explore the interactions of IC2. Also, these experiments could be performed with IC2-F(ab’)2 or IC2-fab fragment which we already know is able to bind directly to the gold sensor surface. Here, the calculation of the kinetics would be more straight forward as Fab has only one binding site and F(ab’)2 has only two binding sites.

89

conclusions and perspectives

For the subsequent comparative analyses and comparison of the different formats, I would suggest to use a kinetic distribution ranking system (Canziani et al., 2004), as a basis for comparison (Figure 62). For each of the antigens the antibody formats could be ranked by their KD. Figure 62. Kinetic distribution ranking system for ranking of antibody affinities. By applying a logarithmic scale, the affinities of the antibody/antigen complexes are plotted by their kinetic constants. Diagonal lines represent affinity isotherms by KD values. The highest affinity complexes are clustered in the upper right corner of the plot. I suggest to use a modification of this ranking system for future affinity ranking (Canziani et al., 2004).

For future more extensive experiments using QCM, or perhaps SPR, it would be a good idea to covalently attach the IC2 antibody to the chip. This would allow for several continuous measurements on either the same or a different potential antigen, which would allow for screening for a larger amount of antigens and for affinity measurements at different concentrations of antigen. In between the measurements, it would be necessary to check that the antibody had not detached from the gold sensor surface, and that it had not been partially or totally denatured. The results obtained from the MST measurements showing interaction between IC2 and lyso-sulfatide will now be repeated to validate and to give a more precise estimate of the K D value of the interaction with the addition of statistic parameters. Also MST and QCM-D experiments with the CD1d-molecule alone and in complex with different glycolipids would be interesting to do as soon as this CD1d molecule is obtained, as it is likely that CD1d is involved in the IC2 binding interaction somehow. A large production of CD1d is to be established at the Bartholin Institute using a transfected insect cell line.

6.2.1.2. A new direction 6.2.1.2.1. Mass spectrometry Regarding the current studies, the quest for determining the natural autoantigen for IC2 will proceed in a new direction. Currently, a project is being set up where the IC2 antigen is attempted elucidated by using another approach. By shaving of IC2-binding cells with proteases, the IC2 autoantigen will be attempted purified by the use of mild detergents, affinity chromatography and iTLC. The extracted lipids will then be analyzed by lipid mass spectrometry (Gode and Volmer, 2013).

6.2.1.2.2. Bioinformatics A totally different direction to try to elucidate the IC2 autoantigen is to, by use of bioinformatics, not just search for different potential targets, but also to simulate the binding of different potential antigens. When the IC2 autoantigen has been elucidated, it could be interesting to try to crystallize this molecule when interacting with IC2 to by X-ray elucidate the structure and thereby further characterize the interaction.

90

conclusions and perspectives

6.2.1.2.3. The role/involvement of the CD1d receptor As I see it, the evidence that the CD1d receptor is involved with the IC2-autoantigen is piling up: First, pilot experiments showed that IC2 was able to inhibit stimulation of NKT type I and possibly type II cells (Jensen et al., 2012), which indicated that IC2 could interact with a part of the CD1d receptor. It was shown by Voetmannby immunostaining, that CD1d is present of the surface of β-cells (Voetmann, 2013). We have performed FCM experiments which showed that IC2 bound to two CD1d-transfected cell lines, but not to their non-transfected equals (Unpublished data. Appendix I). It has been found by FCM, that IC2 binds to a small population of CD45+ cells from the pancreas of a NOD mouse and that this binding was not observed in cells from the pancreas of a CD1d knockout mouse (Simoni, 2012). Furthermore, IC2 has been found to bind to a small fraction of cells in thymus (Poussier et al., 1986), which is where NKT-cells are developed and thus also selected to be reactive towards CD1d-lipid complexes. All of this points towards an involvement of CD1d in the binding of IC2. The problem with the involvement of the CD1d receptor is that is does not explain the β-cell specificity that IC2 exhibits. CFM staining with the IC2 antibody and CD1d antibodies could be used to see if IC2 binds to a structure related to the CD1d receptor. If the marking with these antibodies shows to be convergent, the actual link between these two structures could be further investigated by inhibition and competition experiments. These inhibition experiments could be performed in a modification of the assay used to observe the inhibitory effect of IC2 on NKT-cells. Here it could be interesting to do the experiments with other lipids also known to stimulate NKT-cells and to quantify the amount of stimulation with and without IC2. Another idea is to do competition experiments with IC2 and different CD1d antibodies known to interact with different parts of the CD1d receptor. These competition experiments could be performed using FCM or ELISA on plasma membrane coated plates.

6.2.2. Other perspectives 6.2.2.1. Imaging and determination of IC2-format for future imaging trials Extensive imaging trials are currently being performed in murine animal models with different approaches. Recently, near infrared imaging was performed with IC2-Fab and IC2-F(ab’)2 fragments (Sefeld et al., 2013). In the near future IC2 will be used in photoacousting imaging trials in murine animal models using nanorods (Su et al., 2010). To do imaging trials is a rather costly affair both in regards to sample consumption, the use of animal models, and labor time. Therefore, it could be worth the time to perform affinity measurements of the different IC2-formats towards intact β-cell lines and thereby find the best suited format for imaging. These experiments could be performed using the Attana Cell 200 QCM. This was an aim of this project, but unfortunately no binding of IC2 to the β-cells were detected probably due to the diffusion of an important part of the IC2 antigen. Thus, growing the cell on site and doing measurements immediately after cell fixation could be the method to measure the affinity of the different IC2 formats towards intact β-cells to find the best suited candidate for further imaging trials.

6.2.2.2. Antiidiotypic antibody A project based on the observation of the NKT-cell inhibitory action of IC2 in vitro is currently ongoing. If certain populations of NKT-cells are inhibited, it may cause a shift in the release of Th1 and Th2 response enhancing cytokines and interleukines, thus shifting the balance of the T-cell population towards a more

91

conclusions and perspectives

aggressive T-cell population which could cause autoimmune destruction. Therefore, an antiidiotypic antibody towards IC2 is currently being developed to see whether elimination of IC2 and IC2-like antibodies, will be able to prevent type 1 diabetes (Jensen et al., 2012). In this project, the methods for affinity measurements used in this project and the experiences obtained in this project, may come in handy for calculation of and ranking of affinity between IC2 and the newly developed antiidiotypic antibodies.

6.2.2.3. Further explore the functional abilities of IC2 Whatever the outcome of the antiidiotypic antibody project, the use of IC2 in development of a therapeutic approach by interference in the NKT-cell regulatory system is a very interesting project. If the right path is taken, IC2 might in the future play a role in the development of regulation of autoimmunity by NKT-cells.

92

references

7. References Abbas AK, Lichtman AH, Pillai S (2007) Cellular and Molecular Immunology, 7th edition. Chapter: Structural and Chemical Basis of Antigen Binding. Book. Abrahamsson J, Abrahamsson S, Hellqvist B, Larsson K, Pascher I, Sundell S (1977) Cholesteryl sulphate and phosphate in the solid state and in aqueous systems. Chemistry and physics of lipids 19:213-222. Anderson MS, Bluestone JA (2005) The NOD mouse: a model of immune dysregulation. Annual review of immunology 23:447-485. Asfari M, Janjic D, Meda P, Li G, Halban PA, Wollheim CB (1992) Establishment of 2-mercaptoethanoldependent differentiated insulin-secreting cell lines. Endocrinology 130:167-178. Ashcroft FM, Ashcroft SJH (1992) Insulin Molecular Biology to Pathology. New York: Oxford University Press. Atkinson MA, Eisenbarth GS (2001) Type 1 diabetes: new perspectives on disease pathogenesis and treatment. Lancet 358:221-229. Becker B, Cooper MA (2011) A survey of the 2006-2009 quartz crystal microbalance biosensor literature. Journal of Molecular Recognition 24:754-787. Review. Bendelac A, Savage PB, Teyton L (2007) The biology of NKT cells. Annual review of immunology 25:297-336. Billestrup N (1985) Human growth hormone binding and stimulation of insulin biosynthesis in cloned rat insulinoma cells. Biomedica biochimica acta 44:85-87. Bjorke H, Andersson K (2006) Measuring the affinity of a radioligand with its receptor using a rotating cell dish with in situ reference area. Appl Radiat Isot 64:32-37. Björkelund H, Gedda L, Andersson K (2011) Avoiding false negative results in specificity analysis of proteinprotein interactions. Blanchet JP (1974) Chicken erythrocyte membranes: comparison of nuclear and plasma membranes from adults and embryos. Experimental cell research 84:159-166. Bobrovnik SA (2003) Determination of antibody affinity by ELISA. Theory. J Biochem Biophys Methods 57:213-236. Brigl M, Bry L, Kent SC, Gumperz JE, Brenner MB (2003) Mechanism of CD1d-restricted natural killer T cell activation during microbial infection. Nature immunology 4:1230-1237. Brogren CH (1986) Interspecies crossreactivity of the pancreatic islet cell surface specific rat monoclonal antibody IC-2. Diabetologia 29:A522. Abstract. Brogren CH, Andersen PS (1983) BB Wistar Rat Monoclonal Autoantibodies. Metabolism - Clinical & Experimental 32:165. Brogren CH, Gao W, Mia GKAM, Buschard K, Bachmann MH, Contag C, Renaut L, D. C, N. S, Mondon P (2010) Development of the pancreatic beta-cell surface-specific monoclonal autoantibody, IC2, for accessing beta-cell mass in vivo. Diabetes 59:A59. Abstract. Brogren CH, Hirsch F, Wood P, Druet P, Poussier P (1986) Production and characterization of a monoclonal islet cell surface autoantibody from the BB rat. Diabetologia 29:330-333. Brogren CH, Hoofar J, Lassen A, Buschard K (1988) Polyclonal activation by dietary immunogens and mitogens - a possible nutritional regulator for the onset of diabetes in BB rats. Diabetologia 31:A474. 424 th EASD annual meeting, Paris, France. Abstract. Brogren CH, Lernmark A (1982) Islet cell antibodies in diabetes. Clin Endocrinol Metab 11:409-430. Brogren CH, Poussier P, Hirsch F, Druet P (1984) Production in the BB-rat of a monoclonal antibody specific for the surface of the rat insulin producing cells. Diabetologia 27:A260. EASD 220 th Annual Meeting, London. Abstract. Brogren CH, Spitalnik SL, Shienvold FL, Buschard K (1989) Terminal Galactose-3-sulfate as autoantigenic determinant for the monoclonal autoantibody IC-2 isolated from diabetic BB-rat. 7th International Congress of Immunology, July 30 - August 5 Berlin, Germany. Abstract.

93

references

Bronsart L (2012) The novel application of the luciferase substrate coelenterazine for imaging beta-cell mass. SS89. Poster. World Molecular Imaging Congress, September 2012, Dublin, Ireland. Bronsart L, Bachmann MH, Stokes C, Schmidt TL, Cochran F, Brogren CH, Renaut L, Mondon P, Contag C (2011) Genetic Engineering of the IC2 Autoantibody for Molecular Imaging of the Pancreatic BetaCell Mass. In: Annual World Molecular Imaging Congress Poster. Brown RE, Mattjus P (2007) Glycolipid transfer proteins. Biochimica et biophysica acta 1771:746-760. Bunde RL, Jarvi EJ, Rosentreter JJ (1998) Piezoelectric quartz crystal biosensors. Talanta 46:1223-1236. Buschard K (2011) What causes type 1 diabetes? Lessons from animal models. APMIS Supplementum 132:1-19. Buschard K, Brogren CH, Ropke C, Rygaard J (1988) Antigen expression of the pancreatic beta-cells is dependent on their functional state, as shown by a specific, BB rat monoclonal autoantibody IC2. APMIS 96:342-346. Buschard K, Diamant M, Bovin LE, Mansson JE, Fredman P, Bendtzen K (1996) Sulphatide and its precursor galactosylceramide influence the production of cytokines in human mononuclear cells. APMIS 104:938-944. Baaske P, Wienken CJ, Reineck P, Duhr S, Braun D (2010) Optical thermophoresis for quantifying the buffer dependence of aptamer binding. Angew Chem Int Ed Engl 49:2238-2241. Canziani GA, Klakamp S, Myszka DG (2004) Kinetic screening of antibodies from crude hybridoma samples using Biacore. Analytical Biochemistry 325:301-307. Chalfant C, Del Poeta M (2010) Sphingolipids as Signaling and Regulatory Molecules. vol. 688 (Bach, N. et al., eds), p 287 Texas, USA: Landes Bioscience. Chang X, Jorgensen AM, Bardrum P, Led JJ (1997) Solution structures of the R6 human insulin hexamer. Biochemistry 36:9409-9422. Chiu YH, Jayawardena J, Weiss A, Lee D, Park SH, Dautry-Varsat A, Bendelac A (1999) Distinct subsets of CD1d-restricted T cells recognize self-antigens loaded in different cellular compartments. The Journal of experimental medicine 189:103-110. Cooke A (2012) Parasitic worms and inflammatory disease. Current opinion in rheumatology 24:394-400. Cooke A, Tonks P, Jones FM, O'Shea H, Hutchings P, Fulford AJ, Dunne DW (1999) Infection with Schistosoma mansoni prevents insulin dependent diabetes mellitus in non-obese diabetic mice. Parasite immunology 21:169-176. Cooper MA, Singleton VT (2007) A survey of the 2001 to 2005 quartz crystal microbalance biosensor literature: applications of acoustic physics to the analysis of biomolecular interactions. J Mol Recognit 20:154-184. Review. Cox D, Fox L, Tian R, Bardet W, Skaley M, Mojsilovic D, Gumperz J, Hildebrand W (2009) Determination of cellular lipids bound to human CD1d molecules. PLoS One 4:e5325. De Libero G, Mori L (2007) Structure and biology of self lipid antigens. Current topics in microbiology and immunology 314:51-72. Desai MN (2009) Radioiodination of the monoclonal autoantibody IC2 - a biomarker for pancreatic betacells. A study by cellular real-time radioimmunoassay of in-vitro specificity and affinity measurement. Master Thesis from University of Copenhagen and Rigshospitalet. Di Gialleonardo V, de Vries EF, Di Girolamo M, Quintero AM, Dierckx RA, Signore A (2012) Imaging of betacell mass and insulitis in insulin-dependent (Type 1) diabetes mellitus. Endocrine reviews 33:892919. Dunn MF (2005) Zinc-ligand interactions modulate assembly and stability of the insulin hexamer -- a review. Biometals : an international journal on the role of metal ions in biology, biochemistry, and medicine 18:295-303. Efrat S, Linde S, Kofod H, Spector D, Delannoy M, Grant S, Hanahan D, Baekkeskov S (1988) Beta-cell lines derived from transgenic mice expressing a hybrid insulin gene-oncogene. Proc Natl Acad Sci U S A 85:9037-9041.

94

references

Efrat S, Russ HA (2012) Making beta cells from adult tissues. Trends in endocrinology and metabolism: TEM 23:278-285. Fagerstam LG, Frostell A, Karlsson R, Kullman M, Larsson A, Malmqvist M, Butt H (1990) Detection of antigen-antibody interactions by surface plasmon resonance. Application to epitope mapping. J Mol Recognit 3:208-214. Fahy E, Subramaniam S, Brown HA, Glass CK, Merrill AH, Jr., Murphy RC, Raetz CR, Russell DW, Seyama Y, Shaw W, Shimizu T, Spener F, van Meer G, VanNieuwenhze MS, White SH, Witztum JL, Dennis EA (2005) A comprehensive classification system for lipids. J Lipid Res 46:839-861. Fahy E, Subramaniam S, Murphy RC, Nishijima M, Raetz CR, Shimizu T, Spener F, van Meer G, Wakelam MJ, Dennis EA (2009) Update of the LIPID MAPS comprehensive classification system for lipids. J Lipid Res 50 Suppl:S9-14. Fox LM, Cox DG, Lockridge JL, Wang X, Chen X, Scharf L, Trott DL, Ndonye RM, Veerapen N, Besra GS, Howell AR, Cook ME, Adams EJ, Hildebrand WH, Gumperz JE (2009) Recognition of lysophospholipids by human natural killer T lymphocytes. PLoS biology 7:e1000228. Fox TE, Kester M (2010) Therapeutic Strategies for Diabetes and Complications: A Role for Sphingolipids? In: Sphingolipids as Signaling and Regulatory Molecules (Poeta, C. C. M. D., ed), pp 206-216: Landes Bioscience and Springer Science + Business Media. Friguet B, Fedorov A, Serganov AA, Navon A, Goldberg ME (1993) A radioimmunoassay-based method for measuring the true affinity of a monoclonal antibody with trace amounts of radioactive antigen: illustration with the products of a cell-free protein synthesis system. Anal Biochem 210:344-350. Gadola SD, Zaccai NR, Harlos K, Shepherd D, Castro-Palomino JC, Ritter G, Schmidt RR, Jones EY, Cerundolo V (2002) Structure of human CD1b with bound ligands at 2.3 A, a maze for alkyl chains. Nature immunology 3:721-726. Gao W (2010) Fragmentation of the pancreatic beta-cell surface-specific antibody IC2 - a potential biomarker for beta-cell mass imaging in vivo. Master thesis from Uppsala University, Sweden. Girardi E, Zajonc DM (2012) Molecular basis of lipid antigen presentation by CD1d and recognition by natural killer T cells. Immunological reviews 250:167-179. Gode D, Volmer DA (2013) Lipid imaging by mass spectrometry - a review. The Analyst 138:1289-1315. Godfrey DI, Kronenberg M (2004) Going both ways: immune regulation via CD1d-dependent NKT cells. J Clin Invest 114:1379-1388. Godfrey DI, MacDonald HR, Kronenberg M, Smyth MJ, Van Kaer L (2004) NKT cells: what's in a name? Nat Rev Immunol 4:231-237. Goldberg ME, Djavadi-Ohaniance L (1993) Methods for measurement of antibody/antigen affinity based on ELISA and RIA. Curr Opin Immunol 5:278-281. Review. Huang S, Cheng TY, Young DC, Layre E, Madigan CA, Shires J, Cerundolo V, Altman JD, Moody DB (2011) Discovery of deoxyceramides and diacylglycerols as CD1b scaffold lipids among diverse grooveblocking lipids of the human CD1 system. Proc Natl Acad Sci U S A 108:19335-19340. Höök F, Rodahl M, Brzezinski P, Kasemo B (1998) Energy Dissipation Kinetics for Protein and Antibody−Antigen Adsorption under Shear Oscillation on a Quartz Crystal Microbalance. Langmuir 14:729-734. Ishizuka I (1997) Chemistry and functional distribution of sulfoglycolipids. Progress in lipid research 36:245319. Jacobsen I (2010) Udvikling af et kompetitivt immunoassay baseret på det monoklonale autoantistof IC2 i en serologisk testning for type 1 diabetes. Bachelor thesis from University of Copenhagen. Jahng A, Maricic I, Aguilera C, Cardell S, Halder RC, Kumar V (2004) Prevention of autoimmunity by targeting a distinct, noninvariant CD1d-reactive T cell population reactive to sulfatide. The Journal of experimental medicine 199:947-957. Jelesarov I, Bosshard HR (1999) Isothermal titration calorimetry and differential scanning calorimetry as complementary tools to investigate the energetics of biomolecular recognition. J Mol Recognit 12:3-18. Review.

95

references

Jensen KL, Pedersen ID, Åkesson L, Lernmark Å, Klöting I, Niessner F, Svakgaard J, Engkilde K, Buschard K, Brogren CH (2012) Natural Killer T-cell IL-2 cytokine release is inhibited by the monoclonal autoantibody IC2 and might be restored by its anti-idiotypic antibodies. Diabetologia 55:S202. The 248nd European Association for the Study of Diabetes Annual Meeting, Berlin, Germany. Jerabek-Willemsen M, Wienken CJ, Braun D, Baaske P, Duhr S (2011) Molecular interaction studies using microscale thermophoresis. Assay Drug Dev Technol 9:342-353. Jørgensen CV (2010) Udvikling af et kompetitivt immunoassay baseret på det monoklonale autoantistof IC2 i en serologisk testning for type 1 diabetes. Bachelor thesis from University of Copenhagen. Kavishwar A, Medarova Z, Moore A (2011) Unique sphingomyelin patches are targets of a beta cell specific antibody. Journal of Lipid Research 52:1660-1671. Khardori R, Votey SR, Peters AL, Schraga ED, Talavera F, Bessen HA, Halamka JD, Brenner BE (2007) Diabetes mellitus type I - a review. Medscape Drugs diseases and procures http://emedicine.medscape.com/article/117739-overview. Kobayashi E, Motoki K, Uchida T, Fukushima H, Koezuka Y (1995) KRN7000, a novel immunomodulator, and its antitumor activities. Oncology research 7:529-534. Koshy KM, Wang J, Boggs JM (1999) Divalent cation-mediated interaction between cerebroside sulfate and cerebrosides: an investigation of the effect of structural variations of lipids by electrospray ionization mass spectrometry. Biophysical journal 77:306-318. Kurosawa S, Nakamura M, Park JW, Aizawa H, Yamada K, Hirata M (2004) Evaluation of a high-affinity QCM immunosensor using antibody fragmentation and 2-methacryloyloxyethyl phosphorylcholine (MPC) polymer. Biosens Bioelectron 20:1134-1139. La Flamme AC, Ruddenklau K, Backstrom BT (2003) Schistosomiasis decreases central nervous system inflammation and alters the progression of experimental autoimmune encephalomyelitis. Infection and immunity 71:4996-5004. Ladbury JE (2010) Calorimetry as a tool for understanding biomolecular interactions and an aid to drug design. Biochem Soc Trans 38:888-893. Review. Lang GA, Maltsev SD, Besra GS, Lang ML (2004) Presentation of alpha-galactosylceramide by murine CD1d to natural killer T cells is facilitated by plasma membrane glycolipid rafts. Immunology 112:386-396. Liston A, Lesage S, Wilson J, Peltonen L, Goodnow CC (2003) Aire regulates negative selection of organspecific T cells. Nature immunology 4:350-354. Marron MP, Graser RT, Chapman HD, Serreze DV (2002) Functional evidence for the mediation of diabetogenic T cell responses by HLA-A2.1 MHC class I molecules through transgenic expression in NOD mice. Proc Natl Acad Sci U S A 99:13753-13758. Merten M, Motamedy S, Ramamurthy S, Arnett FC, Thiagarajan P (2003) Sulfatides: targets for antiphospholipid antibodies. Circulation 108:2082-2087. Mia GKAM (2009) Characterization of the pancreatic beta-cell autoantigen targeted by the IC2 monoclonal autoantibody. Master thesis from University of Linköbing, Sweden. Milo-Landesman D, Efrat S (2002) Growth factor-dependent proliferation of the pancreatic beta-cell line betaTC-tet: an assay for beta-cell mitogenic factors. International journal of experimental diabetes research 3:69-74. Miyazaki J, Araki K, Yamato E, Ikegami H, Asano T, Shibasaki Y, Oka Y, Yamamura K (1990) Establishment of a pancreatic beta cell line that retains glucose-inducible insulin secretion: special reference to expression of glucose transporter isoforms. Endocrinology 127:126-132. Moody DB, Porcelli SA (2003) Intracellular pathways of CD1 antigen presentation. Nat Rev Immunol 3:1122. Moore A, Bonner-Weir S, Weissleder R (2001) Noninvasive in vivo measurement of beta-cell mass in mouse model of diabetes. Diabetes 50:2231-2236. Murphy K, Travers P, Walport M (2008) Janeway's Immunobiology. Book.: Garland Science, Taylor & Francis Group, LLC.

96

references

Nagayama Y, Watanabe K, Niwa M, McLachlan SM, Rapoport B (2004) Schistosoma mansoni and alphagalactosylceramide: prophylactic effect of Th1 Immune suppression in a mouse model of Graves' hyperthyroidism. Journal of immunology 173:2167-2173. Nolan JP, Sklar LA (1998) The emergence of flow cytometry for sensitive, real-time measurements of molecular interactions. Nature biotechnology 16:633-638. Osada Y, Shimizu S, Kumagai T, Yamada S, Kanazawa T (2009) Schistosoma mansoni infection reduces severity of collagen-induced arthritis via down-regulation of pro-inflammatory mediators. International journal for parasitology 39:457-464. Osterbye T, Jorgensen KH, Fredman P, Tranum-Jensen J, Kaas A, Brange J, Whittingham JL, Buschard K (2001) Sulfatide promotes the folding of proinsulin, preserves insulin crystals, and mediates its monomerization. Glycobiology 11:473-479. Pedersen ID, Mia GKAM, Desai MN, Nalla A, Prats C, Wismann P, Larsen J, Danley M, Spitalnik SL, Popa J, Portoukalain J, Ilyas A, Chen Z-W, Mondon P, Renaut L, Bronsart L, Bachmann MH, Contag C, Eliasson L, Herrmann T, Cardell S, Engkilde K, Buschard K, Brogren CH (2011) The pancreatic betacell surface specific BB-rat monoclonal autoantibody IC2 binds to sulphated and phosphorylated glycolipid epitopes on sulfatides, sphingomyelins and part of a more complex and trypsin sensitive neoautoantigen. The EASD Islet Cell Working Group Symposium ”Programming Beta Cell Development, Impairment and Regeneration”, October 23-26, 2011, Helsingør, Denmark. Peer D, Karp JM, Hong S, Farokhzad OC, Margalit R, Langer R (2007) Nanocarriers as an emerging platform for cancer therapy. Nature nanotechnology 2:751-760. Pertoft H, Laurent TC, Laas T, Kagedal L (1978) Density gradients prepared from colloidal silica particles coated by polyvinylpyrrolidone (Percoll). Anal Biochem 88:271-282. Poussier P, Brogren CH, Hirsch F, Druet P (1984) Monoclonal antibody from the BB-rat against rat insulin producing cells. Diabetes Annual Meeting of the American Diabetes Association, Las Vegas, USA. Abstract. Poussier P, Legendre C, Wood P, Guttmann RD, Brogren CH (1986) Identification of a common epitope on the surface of the rat islet cells and thymic lymphocytes-T using a monoclonal autoantibody (MAB) produced from the BB rat. Diabetes 35:A25. Abstract. Powers AC, Efrat S, Mojsov S, Spector D, Habener JF, Hanahan D (1990) Proglucagon processing similar to normal islets in pancreatic alpha-like cell line derived from transgenic mouse tumor. Diabetes 39:406-414. Privalov PL, Dragan AI (2007) Microcalorimetry of biological macromolecules. Biophys Chem 126:16-24. Review. Radbruch A (ed.) (1992) Flow cytometry affinity studies. Berlin, Heidelberg: Springer. Reviakine I, Johannsmann D, Richter RP (2011) Hearing what you can't see and visualizing what you hear: interpreting quartz crystal microbalance data from solvated interfaces. Anal Chem. Rhost S, Lofbom L, Rynmark BM, Pei B, Mansson JE, Teneberg S, Blomqvist M, Cardell SL (2012) Identification of novel glycolipid ligands activating a sulfatide-reactive, CD1d-restricted, type II natural killer T lymphocyte. European journal of immunology 42:2851-2860. Rodahl M, Hook F, Krozer A, Brzezinski P, Kasemo B (1995) Quartz crystal microbalance setup for frequency and Q factor measurements in gaseous and liquid environments. Rev Sci Instrum 66:3924-3930. Roeske-Nielsen A, Fredman P, Mansson JE, Bendtzen K, Buschard K (2004) Beta-galactosylceramide increases and sulfatide decreases cytokine and chemokine production in whole blood cells. Immunology letters 91:205-211. Rorsman P, Renstrom E (2003) Insulin granule dynamics in pancreatic beta cells. Diabetologia 46:10291045. Roy KC, Maricic I, Khurana A, Smith TR, Halder RC, Kumar V (2008) Involvement of secretory and endosomal compartments in presentation of an exogenous self-glycolipid to type II NKT cells. Journal of immunology 180:2942-2950.

97

references

Saudek F, Brogren CH, Manohar S (2008) Imaging the Beta-cell mass: why and how. The Review of Diabetic Studies 5:6-12. Review. Saunders KA, Raine T, Cooke A, Lawrence CE (2007) Inhibition of autoimmune type 1 diabetes by gastrointestinal helminth infection. Infection and immunity 75:397-407. Scharf L, Li NS, Hawk AJ, Garzon D, Zhang T, Fox LM, Kazen AR, Shah S, Haddadian EJ, Gumperz JE, Saghatelian A, Faraldo-Gomez JD, Meredith SC, Piccirilli JA, Adams EJ (2010) The 2.5 a structure of CD1c in complex with a mycobacterial lipid reveals an open groove ideally suited for diverse antigen presentation. Immunity 33:853-862. Schuck P, Boyd LF, Andersen PS (2001) Measuring Protein Interactions by Optical Biosensors. Currrent Protocols in Protein Science Chapter 20, Unit 20.2. Sefeld M, Pedersen ID, Jensen KL, Voetmann ME, Engmose CP, Gao W, Briat A, Buschard K, Brogren CH (2013) Proteolytic fragmented and chimeric IC2 monoclonal autoantibody spefific against the pancreatic beta-cell surface chemically labeled with near-infrared fluorochromes Vivotag 680 and 750 and applied for near-infrared cellular and pancreatic beta-cell imaging. Scandinavian journal of immunology 77:302. Shamshiev A, Donda A, Carena I, Mori L, Kappos L, De Libero G (1999) Self glycolipids as T-cell autoantigens. European journal of immunology 29:1667-1675. Shen D, Kang Q, Wang YE, Hu Q, Du J (2008) New cut angle quartz crystal microbalance with low frequencytemperature coefficients in an aqueous phase. Talanta 76:803-808. Shier D, Butler J, Lewis R (2009) Hole's essentials of human anatomy & physiology. Boston: McGraw-Hill Higher Education. Simoni Y (2012) Personal communication. Skelin M, Rupnik M, Cencic A (2010) Pancreatic beta cell lines and their applications in diabetes mellitus research. Altex 27:105-113. Spitalnik SL (1989) Personal communication. Strachan DP (1989) Hay fever, hygiene, and household size. Bmj 299:1259-1260. Strott CA, Higashi Y (2003) Cholesterol sulfate in human physiology: what's it all about? J Lipid Res 44:12681278. Su JL, Wang B, Wilson KE, Bayer CL, Chen YS, Kim S, Homan KA, Emelianov SY (2010) Advances in Clinical and Biomedical Applications of Photoacoustic Imaging. Expert Opin Med Diagn 4:497-510. Sweet IR, Cook DL, Lernmark A, Greenbaum CJ, Krohn KA (2004) Non-invasive imaging of beta cell mass: a quantitative analysis. Diabetes Technol Ther 6:652-659. Sørensen J, Brogren CH, Buschard K (submitted) The preventive role of type 2 NKT cells in type I diabetes development. APMIS. Tanaka KA, Suzuki KG, Shirai YM, Shibutani ST, Miyahara MS, Tsuboi H, Yahara M, Yoshimura A, Mayor S, Fujiwara TK, Kusumi A (2010) Membrane molecules mobile even after chemical fixation. Nature methods 7:865-866. Tellechea E, Johannsmann D, Steinmetz NF, Richter RP, Reviakine I (2009) Model-independent analysis of QCM data on colloidal particle adsorption. Langmuir 25:5177-5184. Van Belle TL, Coppieters KT, Von Herrath MG (2011) Type 1 Diabetes: Etiology, Immunology, and Therapeutic Strategies. Physiol Rev 91:79-118. Review. van den Brandt J, Fischer HJ, Walter L, Hunig T, Kloting I, Reichardt HM (2010) Type 1 diabetes in BioBreeding rats is critically linked to an imbalance between Th17 and regulatory T cells and an altered TCR repertoire. Journal of immunology 185:2285-2294. Voetmann ME (2013) CD1d expression on pancreatic β-cells, Bachelor thesis from University of Copenhagen. Webpage (http://www.jltacademystudents.com/) JL Academy homepage. Accessed April 2013. Webpage (www.lipidworld.com) Lipids in Health and Disease Journal. Accessed April 2013. Webpage (www.nanotemper.de) NanoTemper homepage. Accessed April 2013. Webpage (www.q-sense.com) Q-sense homepage. Accessed November 2011.

98

references

Webpage (www.themedicalbiochemistrypage.org) The Medical Biochemistry webpage. Accessed April 2013. Webpage (www.thsci.com) Tetrahedron hompage. Accessed March 2013. Webpage (www.trc-canada.com) Toronto Research Chemicals Inc. homepage. Accessed April 2013. Webpage (www.wikipedia.org) Wikipedia webpage. Accessed March 2013. Whalen BJ, Mordes JP, Rossini AA (2001) The BB rat as a model of human insulin-dependent diabetes mellitus. Current protocols in immunology / edited by John E Coligan [et al] Chapter 15:Unit 15 13. Wienken CJ (2011) Biotechnological applications of thermophoresis. In: der Fakultät für Physik: phD thesis, der Ludwid-Maximillians-Universität München. Wismann P (2010) Biodistribution of the pancreatic beta-cell specific antibody IC2 - a possible early diagnostic for type-1 diabetes. Bachelor thesis from University of Copenhagen. Wolny PM, Spatz JP, Richter RP (2010) On the adsorption behavior of biotin-binding proteins on gold and silica. Langmuir 26:1029-1034. Yuan W, Kang SJ, Evans JE, Cresswell P (2009) Natural lipid ligands associated with human CD1d targeted to different subcellular compartments. Journal of immunology 182:4784-4791. Zaccone P, Cooke A (2013a) Helminth mediated modulation of Type 1 diabetes (T1D). International journal for parasitology 43:311-318. Zaccone P, Cooke A (2013b) Vaccine against autoimmune disease: can helminths or their products provide a therapy? Curr Opin Immunol. Zajonc DM, Elsliger MA, Teyton L, Wilson IA (2003) Crystal structure of CD1a in complex with a sulfatide self antigen at a resolution of 2.15 A. Nature immunology 4:808-815. Zajonc DM, Maricic I, Wu D, Halder R, Roy K, Wong CH, Kumar V, Wilson IA (2005) Structural basis for CD1d presentation of a sulfatide derived from myelin and its implications for autoimmunity. The Journal of experimental medicine 202:1517-1526. Zeng Z, Castano AR, Segelke BW, Stura EA, Peterson PA, Wilson IA (1997) Crystal structure of mouse CD1: An MHC-like fold with a large hydrophobic binding groove. Science 277:339-345. Zhu XH, Hara A, Taketomi T (1991) The existence of galactosylceramide I3-sulfate in serums of various mammals and its anticoagulant activity. Journal of biochemistry 110:241-245. Aaen K, Rygaard J, Josefsen K, Petersen H, Brogren CH, Horn T, Buschard K (1990) Dependence of antigen expression on functional state of beta-cells. Diabetes 39:697-701.

99

appendix

Appendix I Binding of IC2 to CD1d-transfected cell lines IC2 binding to A20 WT and A20mCD1d

Count

Unpublished data from experiments by Christina Petersen Engmose 65 (The Bartholin Institute, Rigshospitalet) showing the binding of IC2 and the CD1d-specific antibody WT to CD1d-transfected A20 cells 49 (A20mCD1d)(figure A1). These experiments were also performed with a CD1d-transfected derivative of the macrophage cell line RAW (RAWmCD1d)(not shown). IC2 binding33was also observed with RAWmCD1d though not as pronounced as with CD1d-A20. 16

 

0

IC2 binding to A20 WT and A20mCD1d

0

10 10

65

2

10

10

4

PE-A A20mCD1d control A20 WT w ith IC2-PE A20mCD1d w ith IC2-PE

49

Count

3

33

  16 0 10

0

10

2

10

3

10

4

10

5

PE-A Figure A1. Binding of IC2 to the CD1d transfected A20 cell line. Unpublished data by Christina Petersen Engmose, 2013, The A20mCD1d control A20 WT w ith IC2-PE Bartholin Institute, Rigshospitalet. A20mCD1d w ith IC2-PE

100

10

5

appendix

Appendix II NKT-cells are inhibited by IC2 In preliminary results from a modified functional NKThybridoma assay, IC2 was found to inhibit IL-2 release from NKT-cells. Figure A2 shows a drawing of the experimental setup. Figure A3 shows two different plots of the inhibition of NKT I cells. Data regarding NKT II cells are not shown here (Jensen et al., 2012).

Figure A2. Modified functional NKThybridoma assay [6]. Microwell plates were first coated with soluble CD1d. As a control NKT cells were added to the wells without glycolipid loaded to the CD1d, and the IL-2 concentration was measured from the supernatant. The other wells were filled with either α-GalCer alone or with IC2+α-GalCer either preincubated or added directly to the wells before the NKT cells were added and supernatant concentrations of IL-2 were measured.

Figure A3. Preliminary data from the functional hybridoma assay showing that IC2 inhibits the IL-2 secretion of NKT type I cells done in triplicates. A) Histogram showing that 4 ug/ml IC2 leads to a 97.98% inhibition, whereas addition of 4ug/ml IgM only shows a 4.44 % inhibition compared to the IL-2 secretion when no antibody is added. B) Shows that the IL-2 secretion is inhibited by IC2 in all tested concentrations, whereas the secretion is only inhibited by IgM when the IgM concentration is extremely high. Data with lower concentrations have not been yet obtained.

101

appendix

Appendix III Growth experiments for Attana Cell 200 experiments 24 well growth experiment To determine the optimal seeding concentration for growth on the Attana COP-1 sensors, a cell growth experiment in 24 well plates was conducted. αTC1-6 and β TC-tet cells were seeded at concentrations of 1*105-2.7*106 cells pr. well. RIN-5AH cells were seeded at concentrations of 1*105-9*105 cells pr. well (figures A4-A6).

Figure A4. β TC-tet cells grown at different seeding concentrations in a 24 well plate. The different concentrations gave rise to a confluency of 40 %, 70 %, 90 % and 100 %, respectively.

Figure A5. RIN-5AH cells grown at different seeding concentrations in a 24 well plate. The different concentrations gave rise to a confluency of 30 %, 70 % and 100 %, respectively.

102

appendix

Figure A6. αTC1-6 cells grown at different seeding concentrations in a 24 well plate. The different concentrations gave rise to a confluency of 5 %, 15 %, 70 % and 100 %, respectively.

The growth of cells were found to have the optimal, given by a close but not too close growth, at different concentrations for the different cell lines. The optimal seeding concentration of the αTC1-6 cells were found to be 9*10^5 cells. The optimal seeding concentration for both β TC-tet and RIN-5AH cells were found to be 3*10^5 cells. Subsequently the cells were fixated in 1% PFA. After fixation, the cells looked the same in the microscope (not shown).

Growth of cells on COP-1 sensors After growth and fixation of the cells in appropriate concentrations, the cells were transported to Stockholm, Sweden, but kept at 5 oC during the transport. Before use, the growth and fixation of cells were accessed by DAPI staining and microscopy (figures A7-A9).

Figure A7. DAPI stain of βTC-tet cells on COP-1 sensor chips. The cells showed a confluency of approximately 80 % for sensor A and B, and 70 % for sensor C and D. Figure A8. DAPI stain of RIN-5AH cells on COP-1 sensor chips. The cells showed a confluency of approximately 50 % for sensor A and D, 60 % for sensor C and 80 % for sensor B. Figure A9. DAPI stain of αTC1-6 cells on COP-1 sensor chips. The cells were less confluent than in the growth experiment, now only about 15-20 % for both sensor A and B.

103

appendix

The confluency was found to be more or less as expected according to the 24 well cell growth experiment on figure A4 and A5 when grown with β TC-tet cells and RIN-5AH cells, respectively. The αTC1-6 cells grown on the COP-1 sensor (figure A9) had a confluency quite different from the expected and observed in the 24 well growth experiment (figure A6). The confluency was found to be more or less as expected according to the 24 well cell growth experiment (figure A4-A6) when grown with β TC-tet cells and RIN-5AH cells (figure A7-A8). The αTC1-6 cells grown on the COP-1 sensor (figure A9) had a confluency quite different from the expected and observed in the 24 well growth experiment (figure A6). This difference may be due to potential differences in growth of this particular type of cells on the COP-1 sensor surface layer compared to the surface layer in the 24 well plates. Another explanation could be that either cell counting or pipetting has been imprecise or that the cells were not equally distributed in the cell solution when a certain amount of suspension was seeded in the wells. Anyway, the low confluency compared to the RIN-5AH and β TC-tet cells could potentially bias the measurements since the αTC1-6 cells, which are supposed control-cells, consequently will show a lower signal because of the fewer cells, making the data uncomparable.

104

appendix

Appendix IV Adsorption and blocking agent experiments for QCM-D IC2 adsorption experiments In order to determine the concentration of IC2-IgM antibody that saturates the gold coated sensor, an adsorption experiment was performed by flushing the sensor with increasing concentrations of IC2 (figure A10). It was determined that 1 ml 50 µg/ml IC2-IgM concentration which gives rise to a frequency shift of 65-80 Hz and a rise in dissipation, was sufficient to give a good biding signal while avoiding non-specific binding effects.

Figure A10. Adsorption at different IC2 concentrations

Figure A10. IC2 adsorption. Plot of frequency (blue) and dissipation (orange) as a function of time with increasing concentrations of IC2 for resonances 5, 7 and 9. A stable baseline was obtained in PBS before 1 ml 20 µg/ml IC2-IgM was injected into the liquid cell followed by a 1 ml 50 µg/ml IC2-IgM injection. Subsequently extensive rinsing with PBS was performed.

Blocking agent experiments To avoid unspecific binding to the surface of the gold coated sensor, a series of experiments with blocking agents was performed using 1 % PVP-40 (figure A11), 1 % BSA (figure A12), 10 % Percoll (figure A13), and 1 % tween-20 (figure A14). Also after immobilization of IC2 on the gold surface, blocking experiments with these three blocking agents were performed (figure A13).

105

appendix

Figure A11. Blocking with PVP-40.

Figure A11. Blocking experiment with 1 % PVP-40. The plot shows the frequency (blue) and the dissipation (orange) as a function of time for resonance 5, 7 and 9. A stable baseline was obtained in PBS before PBS buffer with 1 % PVP-40 was injected into the liquid cell, followed by extensive rinsing with PBS was performed.

For 1 % PVP-40 in PBS (figure A11) there is an adsorption of substance to the sensor, but after approximately 30 min of rinsing with PBS, the bound PVP-40 molecules desorbed and the frequency shift and dissipation returned to the level before PVP-40 injection, indicating a reversible binding. Figure A12. Blocking with BSA

Figure A12. Blocking experiment with 1 % BSA. The plot shows the frequency(blue) and dissipation (orange) as a function of time for resonance 7. A stable baseline was obtained in PBS before PBS buffer with 1 % BSA was injected into the liquid cell followed by extensive rinsing with PBS was performed.

106

appendix

For 1 % BSA in PBS (figure A12) adsorption to the sensor is observed, but after approximately 1h and 30 min of rinsing with PBS, the bound BSA desorbed and the frequency as well as the dissipation returned to the level before BSA injection. Both the frequency- and dissipation curve in figure A12 shows some odd fluctuations every 20-40 min which by virtue of the repetitivity could be pump effects. Figure A13. Blocking with Percoll

Figure A13. Blocking experiment with 10 % Percoll. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. A stable baseline was obtained in PBS before PBS buffer with 10 % Percoll was injected into the liquid cell followed by extensive rinsing with PBS was performed.

Percoll is a suspension of colloidal silica particles of 15-30 nm in size coated with PVP (Pertoft et al., 1978). In figure A13, adsorption is observed when 10 % percoll in PBS is injected over the sensor surface. This binding is partially reversed when the flow is changed to pure PBS, but a significant portion of the bound percoll remains bound after 15 min of rinsing with PBS, and the curve seems to saturate. The dissipation rises to a rather high level with the injection of Percoll and stabilizes at a somewhat lower level after rinsing with PBS, which indicates the adsorption of a rather compact layer of Percoll.

107

appendix

Figure A14. Blocking with tween-20

Figure A14. Blocking experiment with 1 % Tween 20. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. A stable baseline was obtained in PBS before PBS buffer with 1 % Tween 20 was injected into the liquid cell followed by extensive rinsing with PBS was performed.

In figure A14, the decrease in frequency and increase in dissipation indicates that adsorption of tween-20 to the sensor surface is observed. After the flow is changed to pure PBS, we obtain a very stable baseline which is totally stable for at least 1 h. The dissipation with the adsorption of tween-20 is large, which indicates a diffuse layer. In the experiment with different blocking agents after the immobilization of IC2 on the gold sensor chip (figure A15), the same pattern is observed as in the experiments on the bare gold chip. PVP and BSA seem to be reversible blockers and percoll has a partially reversible blocking effect. Figure A15. Blocking with different blocking agents on an IC2 adsorbed sensor surface

Figure A15. Blocking experiment with 3 different blocking agents after immobilization of IC2. The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonances 5, 7 and 9. 1 ml 50 µg/ml IC2-IgM was immobilized on the gold surface of the sensor (not shown) and a stable baseline was obtained in PBS before PBS buffers with 1 % PVP-40, 1 % BSA, or 10 % Percoll was injected into the liquid cell. Subsequently extensive rinsing with PBS was performed. Between the injections of blocking agents, a stable baseline with PBS was obtained.

108

appendix

From the blocking agent experiments on the bare gold coated sensor surface (figure A11-A14) it can be concluded that 1 % tween-20 in PBS is the best blocking agent. Also 10 % Percoll have a blocking effect, and it is surprising that these Percoll particles have a better blocking effect than PVP-40, since PVP is the surface material of these particles. This could be caused by the density of the rather large Percoll particles compared to the 40 kDa PVP-40 polymers used in figure A11. We find 1 % tween-20 to be the best blocking agent, but this agent may cause a problem as a blocking agent in many of these experiments, as it is a detergent and hence should not be used together with lipids, plasma membranes of living cells. Moreover, the dissipation for percoll decreases significantly after PBS wash while this is not the case for 1 % tween-20. This suggests that percoll after PBS wash gives a compact layer while tween gives a rather diffuse layer. For these experiments a compact layer will be most favorable, as the more diffuse layer may be able to hinder interactions IC2 interactions. In the experiment with different blocking agents after the immobilization of IC2 on the gold sensor chip (figure A15), the same pattern is observed with PVP and BSA as reversible blockers and percoll with a partially reversible blocking effect. Perhaps the lack of blocking effect arises from the very little accessible surface area due to the presence of already absorbed IC2. It is important to make sure that these blockers are actually filling areas without prior adsorption on the sensor surface and not actually attaching to IC2.

109

appendix

Appendix V QCM-D experiments with IC2 and sphingomyelin QCM-D experiments were performed with IC2 on the sensor surface and sonicated sphingomyelin from chicken egg in flow (figure A16). Figure A16. Sphingomyelin from chicken egg

Figure A16. Adsorption of IC2-IgM and flow of sonicated sphingomyelin (from egg, chicken). The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonance 7. 1 ml 50 µg/ml IC2-IgM was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. 0.5 mg/ml sonicated sphingomyelin in PBS was injected into the flow cell. A stable baseline was obtained in PBS before and in between change of samples.

Here the flow of IC2-IgM resulted in a frequency shift of -70 Hz. The addition of 0.5 mg/ml sphingomyelin from chicken egg gave rise to a frequency shift of additional -150 Hz and a 40 units rise in dissipation. When the experiment was ended after approximately 1.5 h, the frequency had rised with 5 Hz. A control experiment with adsorption of control rat serum on the sensor surface, only one fifth of the initial sphingomyelin concentration (figure A17) was performed.

110

appendix

Figure A17. Wistar Rat serum + 0.1 mg/ml sphingomyelin (Bovine buttermilk) (blocking BSA 1%)

Figure A17. Adsorption of components of control rat serum and flow of sonicated sphingomyelin (from bovine buttermilk). The plot shows the frequency (blue) and dissipation (orange) as a function of time for resonances 7. Components of rat serum was adsorbed on the gold surface of the sensor and a stable baseline was obtained in PBS before a flow of 1 % BSA in PBS was injected to block for unspecific binding. 0.1 mg/ml sonicated sphingomyelin in PBS was injected. A stable baseline was obtained in PBS before and in between change of samples.

Where addition of control rat serum in flow gave rise to a frequency shift of -65 Hz. This is considerably more than what was observed in figure 20 (-20 Hz). However, the source of these two rat sera samples was different and thus the concentrations of components with sensor surface binding abilities in these sera could vary greatly. With the addition of 0.1 mg/ml sonicated sphingomyelin, no additional shift in frequency nor dissipation was observed, which indicates no binding of sphingomyelin to the rat serum components.

111