Automated System for the Cell-Free Protein Microarray Synthesis and the Label-Free Molecule-Protein Interaction Analysis

Automated system for the cell-free protein microarray synthesis and the label-free molecule-protein interaction analysis

Dissertation zur Erlangung des Doktorgrades der Technischen Fakultät der Albert-Ludwigs-Universität Freiburg

vorgelegt von

Jürgen Burger

Freiburg im Breisgau, Juli 2017

Tag der mündlichen Prüfung: 28.11.2017

Dekan

Prof. Dr. Oliver Paul (Universität Freiburg)

Referenten

Prof. Dr. Gerald Urban (Universität Freiburg)

Prof. Dr. Günter Gauglitz (Universität Tübingen)

Betreuer

Dr. Günter Roth (Universität Freiburg)

Erklärung

Ich erkläre hiermit, dass ich die vorliegende Arbeit ohne unzulässige Hilfe Dritter und ohne Benutzung anderer als der angegebenen Hilfsmittel angefertigt habe. Die aus anderen Quellen direkt oder indirekt übernommenen Daten und Konzepte sind unter Angabe der Quelle gekennzeichnet. Insbesondere habe ich hierfür nicht die entgeltliche Hilfe von Vermittlungs- oder Beratungsdiensten (Promotionsberaterinnen oder Promotionsberatern oder anderer Personen) in Anspruch genommen. Niemand hat von mir unmittelbar oder mittelbar geldwerte Leistungen für Arbeiten erhalten, die im Zusammenhang mit dem Inhalt der vorgelegten Dissertation stehen. Die Arbeit wurde bisher weder im In- noch im Ausland in gleicher oder ähnlicher Form einer anderen Prüfungsbehörde vorgelegt. Ich erkläre hiermit, dass ich mich noch nie an einer in- oder ausländischen wissenschaftlichen Hochschule um die Promotion beworben habe oder gleichzeitig bewerbe.

______Datum/Date Unterschrift / Signature

iii

Abstract

Protein microarrays are essential tools for high-throughput proteome interaction and binding kinetics analysis. Recent technical developments have resulted in approaches for the generation of protein arrays by in situ cell-free expression of DNA array templates [1]. We combined and enhanced these ideas by creating innovative microfluidic flow cells in standard microscope slide format and such improved system efficiency, operability, robustness and controllability. Two of these flow cell designs proved to be most suitable, the booklet design and the structured PDMS slide design. Both incorporate the idea of having a highly parallel microfluidic gap of ~ 30 µm between a template DNA microarray and a protein capture surface. The in situ protein expression is initiated by priming the microfluidic gap with ~ 20 μl of cell-free protein expression mix, after an incubation of 30 to 90 minutes at a temperature of ~ 37 °C the protein microarray is created. Despite free diffusion of the proteins they mainly immobilize highly specific by fusion- tag as rather sharp-edged spots opposite of their DNA spot origin, such creating a protein image of the DNA array. Compared to the most advanced cell-free in situ protein microarray generation technique DAPA [2], we obtained protein copies of similar quality however, the system robustness could be significantly enhanced. Furthermore, the microfluidic system decreases the consumption of the rather expensive cell-free system to one-fifth and cut down the copy process time to half. Our new system yields easy and quick operability with slide exchange times of 1-2 minutes versus > 10 minutes for DAPA.

After a first proof-of-concept for the in situ expression of protein microarrays in our newly devised microfluidic flow cells, an automated setup suitable for both the in situ expression as well as the interaction analysis of protein microarrays by flow-injection was designed, built and evaluated. The reflecto-interferometric principle patented by Biametrics GmbH, Tübingen, Germany, was selected to enable a label-free imaging detection. Such, major deficiencies of most prominent methods for imaging label-free interaction analyses could be circumvented: sensitivity of SPR against temperature drifts and plasmon crosstalk; low number of parallel measurements (< 384) typical for imaging surface plasmon resonance (SPRi), quartz crystal microbalance (QCM) or bio-layer interference (BLI). Aside label-free detection the automated system provides fluidic and temperature steering elements controlled by a programmable process protocol defining parameters for assays with up to 44 reagents. Matlab based software enables automated spot identifications followed by evaluations of these spots resulting in binding curves and finally the determination of binding constants. The total system is validated physically with a bBSA Streptavidin based assay. System sensitivity with baseline noise of < 3 × 10–9 RSS/min and baseline drift of < 0.003 ‰/min proves to be suitable for various biochemical iv

assays. Spotting pattern analyses revealed effects of reagent depletion depending both on concentrations of ligand or reagents as well as spot geometry. Due to available spotting technologies microarrays have been analyzed with a spatial resolution of 2 spots/mm2 (max. 540 spots for applied array of size 18 mm x 15 mm); up to 600 spots/mm2 , in total 162,000 spots at a spatial resolution of ~ 40 µm could be detected. Further biochemical system evaluation by thrombin aptamer assays resulted in a KD of ~ 100 nM, which is in correspondance to literature values. Various antibody antigene assays and a rabbit sera immunization assay were proving this system to be an efficient tool for interaction screenings.

Finally both, the protein microarray expression followed by the microarray interaction analysis was performed sequently in the same microfluidic flow cell. Such the created protein spots have never been in contact with air: this is a major advantage, by now only realized in the PING (Protein In situ Network Generator) chip, a large scale microfluidic integration chip, which is significantly more complex to built and less robust to work with. Moreover, with our microfluidic system multiple expressions of a single DNA array with remarkable reproducibilities have been performed in 2017 by Normann Kilb and Tobias Herz, both working in our research group.

Next the following topics will be tackled for improving the system additionally:

 detailed protein expression analyses considering diffusion processes and depletion effects upon various DNA spotting patterns with different concentrations, densities, geometries.  improving flow cell morphology for a more homogeneous flow on the microarray.  increasing SNR by enhancing the optical system with modified reflective layers to enable the detection and analysis of low molecular binders or very weak interactions.  automation of washing procedures for further increasing system robustness and overall process reproducibility.  deepening biochemical analysis of protein synthesis, e.g. codon usage, IRES sites, transcription factors, ribosome stalling.  expanding applications to the screening of phage / ribosome displays or pre-defined DNA arrays for allergens analysis and the determination of vaccination status or autoimmunities.

Considering all aspects being realized or being latently available, this system has a huge potential for both high-density cell-free protein microarray synthesis and versatile high-throughput proteomic analyses.

Zusammenfassung

Mikroarrays sind ein Grundpfeiler der Hochdurchsatz-Proteom-Interaktionsanalyse und unerlässlich bei der quantitativen Analyse von Bindungskinetiken. Zu den neuesten technologischen Entwicklungen gehören vor allem die Methoden zur direkten Generierung von Protein-Arrays aus DNA-Arrays mittels zellfreier in situ Expression [1]. Diese diversen Lösungsansätze haben wir nun kombiniert und innovativ erweitert, indem wir mikrofluidische Flusszellen im Standard-Objektträgerformat entwickelten, die ein System mit höherer Effizienz, verbesserter Handhabbarkeit, Robustheit und Regelbarkeit realisierten. Zwei der Designentwürfe haben sich als besonders gelungen erwiesen, das Booklet Design und das Design mit einem strukturierten PDMS Slide. Beide Designentwürfe basieren auf der Idee, einen präzise parallelen, mikrofluidischen Spalt mit ~ 30 µm Höhe zwischen dem DNA-Mikroarray und der Protein- Fängeroberfläche zu schaffen. Die in situ Proteinsynthese beginnt zeitlich exakt definiert mit dem Befüllen des mikrofluidischen Spaltes mit ~ 20 µl des zellfreien Proteinsynthesesystems. Nach einer 30 – 90 minütigen Inkubation bei einer Temperatur von ~ 37 °C ist das Protein Mikroarray hergestellt. Trotz der freien Diffusion der Proteine immobilisieren diese mittels eines Fusions- Tags hochspezifisch als bemerkenswert scharfkantige Spots gegenüber ihrer originären DNA Spots, somit wird ein Proteinabbild des DNA Arrays geschaffen. Im Vergleich zu DAPA [2], der wohl fortschrittlichsten Methode zur Erzeugung von Protein-Mikroarrays aus einem DNA- Mikroarray, konnten wir Proteinkopien von vergleichbarer Qualität erzeugen, die Robustheit des Systems jedoch deutlich verbessern. Darüber hinaus hat das entwickelte mikrofluidische System den Verbrauch teurer Biochemikalien auf ein Fünftel reduziert und die Zeit zur Erzeugung einer Protein-Kopie halbiert. Unser System ist einfach und schnell zu handhaben, einer Objektträger- Wechselzeit von 1-2 Minuten stehen mehr als 10 Minuten bei DAPA gegenüber.

Nach einem ersten Machbarkeitsnachweis für die Synthese von Protein-Mikroarrays in einer mikrofluidischen Flusszelle wurde ein komplett automatisiertes Gerät für diese Protein-Synthese und die nachfolgende Interaktionsanalyse mittels Echtzeit-Kinetik-Messungen entworfen, gebaut und getestet. Das reflekto-interferometrische Prinzip (patentiert von Biametrics GmbH, Tübingen) wurde hierbei als label-freies Detektionssystem ausgewählt, da es wesentliche Nachteile der bekannten, bildgebenden, label-freien Interaktionsanalysesysteme nicht aufweist: SPR ist auf Temperaturveränderungen und Plasmoneneinkopplung sehr anfällig; weitere Methoden erlauben nur eine geringe Anzahl von parallelen Messungen (meist < 384) wie z.B bildgebende Oberflächenplasmonenresonanz (SPRi), Quarzkristall-Schwingsystem (QCM) oder die Bio-Layer-Interferenz (BLI). Neben der label-freien Detektion verfügt das automatisierte Gerät über Fluidik- sowie Temperatur-Steuer- und Regeleinheiten, die über eine vi

programmierbare Prozesssteuerung zur Durchführung von Assays mit bis zu 44 Reagenzien parametriert werden können. In Matlab programmierte Module erlauben die automatisierte Spotidentifikation und die darauf folgende Evaluierung dieser Spots, die zur Bestimmung der Bindekurven und zuletzt zur Ermittlung der Bindungskonstanten führt. Das Gesamtsystem wurde physikalisch mittels eines bBSA-Streptavidin-Assays validiert. Es wurde eine Systemsensitivität mit einem Grundlinienrauschen von < 3 × 10–9 RSS/min und einem Grundliniendrift von < 0.003 ‰/min ermittelt, was sich für eine Vielzahl von Assays als hinreichend sensitiv erwiesen hat. Detaillierte Analysen von Arrays offenbarten Verarmungseffekte der Reagenzien in Abhängigkeit von der Reagenz- oder Ligandenkonzentration und der Spotgeometrie. Aufgrund der verfügbaren Spottingtechnologien wurden Mikroarrays mit einer räumlichen Auflösung von 2 Spots/mm2 (max. 540 Spots für die verwendete Arraygröße von 18 mm x 15 mm) analysiert. Das implementierte Detektionssystem könnte jedoch bis zu 600 Spots/mm2 und damit bis zu 162.000 Spots mit einer räumlichen Auflösung von ~ 40 µm detektieren. Die biochemische Systemevaluierung erfolgte mit einem Thrombin-Aptamer-Assay, das KD-Werte von ~ 100 nM lieferte, was den in der Literatur angegebenen Werten entspricht. Vielfältige Antikörper-Antigen- Assays und ein Immunisierungs-Assay (durchgeführt mit Hasenseren) belegen die Eignung des Systems als effiziente Methode für das Screening Protein-basierter Interaktionen.

Final wurde die Protein-Mikroarray-Synthese gefolgt von Interaktionsmessungen auf dem frisch erzeugten Mikroarray in ein und derselben mikrofluidischen Flusszelle durchgeführt. Somit waren die erzeugten Proteinspots nie in Kontakt mit Luft – ein beträchtlicher Vorteil, der bisher nur im PING (Protein In situ Network Generator) Chip, einem hochkomplexen mikrofluidischen Chip mit aufwändiger Herstellung und geringer Robustheit, realisiert wurde. Darüber hinaus konnte von Normann Kilb und Tobias Herz (beide sind Doktoranden der AG Roth) im ersten Quartal 2017 gezeigt werden, dass die DNA-Arrays mehrfach zur Synthese von Protein-Arrays mit bemerkenswerter Reproduzierbarkeit in unserem mikrofluidischen System genutzt werden können.

Zur weiteren Verbesserung des Systems stehen nun die folgenden Aufgaben an:

 Die detaillierte Analyse der Proteinexpression unter Berücksichtigung der Diffusionsprozesse und der Verarmungseffekte in Abhängigkeit der Spottingmuster mit unterschiedlichen DNA-Konzentrationen, Spotabständen, Spotgeometrien.  Die weitere Verbesserung der Flusszellendesigns zur Erzielung homogenerer Flussraten.  Die Erhöhung der Sensitivität des Detektionssystems, um auch Interaktionen mit kleinen Molekülen und schwachen Bindern nachweisen zu können.  Die Erweiterung der Automatisierung des Gerätes um Reinigungsprogramme zur Steigerung der Robustheit des Gesamtsystems und der Reproduzierbarkeit der Assays.

vii

 Tiefgehende Analysen der Biochemie der Proteinerzeugung wie z.B. ‘Codon usage‘, ‘IRES sites‘, Transkriptionsfaktoren und ‘Ribosome stalling‘.  Umsetzung weiterer Applikationen wie Screening von Phagen oder Ribosome Displays, Allergenanalysen, Impfstatusermittlung, Autoimmundiagnostik.

Betrachten wir alle bereits realisierten oder latent verfügbaren Aspekte dieses Systems, erkennen wir ein enormes Potential für die integrierte zell-freie Herstellung von hochdichten Proteinmikroarrays und die Hochdurchsatzanalyse für vielfältige proteomische Anwendungen.

viii

List of Publications

Publications in Peer Reviewed Journals

 Lab-on-a-chip solutions designed for being operated on standard laboratory instruments [3]

 Protein microarray generation by in situ protein expression from template DNA [1]

 Automated setup for label-free detection and analysis of molecule- protein interactions on microarrays by imaging RIf [4]

Posters / Talks at National and International Conferences

 MST Kongress Berlin 2009

 µTAS Groningen 2010

 Ideenwettbewerb 2010, Stuttgart (Talk)

 Functional Genomics 2011, Frankfurt (Talk)

 MMB 2011, Luzern

 Transducers 2011, Peking (Talk)

 Ideenwettbewerb 2012, Stuttgart (Talk)

 Molecular Interactions 2013, Berlin (Talk)

 TechConnect 2014, Washington

 Vision Pharma 2015, Stuttgart

 BITE 2017, Riva del Garda/Italy

Supervised diploma / bachelor / master theses

 David Lämmle, 2011 [5]

 Suleman Shakil, 2012 [6]

 Michael Fakler, 2012 [7]

 Linda Rudmann, 2013 [8]

 Nessim Ben Ammar, 2013 [9]

Table of Content

ERKLÄRUNG ...... III

ABSTRACT...... IV

ZUSAMMENFASSUNG ...... VI

LIST OF PUBLICATIONS ...... IX

TABLE OF CONTENT ...... X

NOMENCLATURE ...... XIII

1 INTRODUCTION ...... 15

1.1 METHODS FOR PROTEIN MICROARRAY SYNTHESIS ...... 15 1.2 OPTICAL DETECTION OF MOLECULE-PROTEIN INTERACTIONS ...... 20 1.3 OBJECTIVES OF THE THESIS ...... 23 1.3.1 Objective 1 - protein microarray synthesis in microfluidic incubator ...... 23 1.3.2 Objective 2 – automated real-time label-free detection system ...... 25 1.4 STRUCTURE OF THESIS ...... 27

2 THEORETICAL BACKGROUND...... 28

2.1 MICROFLUIDIC SYSTEMS ...... 28 2.1.1 Capillary pressure in rectangular flow cell ...... 28 2.1.2 Pinning on horizontal edge ...... 28 2.1.3 Laminar flow...... 29 2.1.4 Taylor disperison ...... 30 2.1.5 Flow boundary condition ...... 31 2.1.6 Navier Stokes equation ...... 31 2.2 DIFFUSION OF MOLECULES ...... 33 2.2.1 Characteristic diffusion time ...... 33 2.2.2 Anomalous diffusion ...... 34 2.3 BIOSENSOR TECHNOLOGY ...... 35 2.3.1 Intermolecular forces ...... 35 2.3.2 Antibody and antigene ...... 36 2.3.3 Assay formats ...... 37 2.4 BIOMOLECULAR INTERACTION KINETICS ...... 38 2.4.1 Receptor and ligand in solution [105] ...... 39 2.4.2 Ligand immobilized on substrate [105, 106] ...... 39 2.5 DETERMINATION OF DISSOCIATION CONSTANT KD ...... 40 2.5.1 Fitting of binding curves ...... 40 2.5.2 Determination of rate constants ...... 42 2.5.3 Depiction by scatter blotting ...... 43 2.6 MODELLING OF MOLECULE INTERACTION IN FLOW CELL ...... 44 2.6.1 Kinetically controlled condition ...... 45 x

2.6.2 Diffusion limited condition ...... 45 2.7 REFLECTOMETRIC INTERFEROMETRY ...... 48 2.7.1 Basics of optical detection systems ...... 48 2.7.2 Sensitivity of optical detection systems ...... 51 2.7.3 RIfS detection technology ...... 54 2.8 BIOCHEMISTRY ...... 57 2.8.1 Expression-ready DNA ...... 57 2.8.2 DNA amplification and analysis ...... 58 2.8.3 Cell-free protein expression ...... 59

3 MATERIALS AND METHODS ...... 61

3.1 FLOW CELL DESIGN AND FABRICATION ...... 61 3.1.1 PMMA flow cell (concept 1) ...... 61 3.1.2 PDMS flow cell (concept 2) ...... 63 3.1.3 Booklet for 2 glass slides (concept 3) ...... 79 3.2 OPTICAL DETECTION SYSTEM ...... 82 3.2.1 Camera and lens system ...... 83 3.3 AUTOMATED SETUP ...... 85 3.3.1 iRIf prototype for proof-of-concept ...... 85 3.3.2 First prototype of automated system ...... 87 3.3.3 Elaborated prototype of automated system ...... 90 3.4 IMAGE PROCESSING ...... 95 3.4.1 Digital filtering...... 95 3.4.2 Spot identification ...... 96 3.4.3 Binding curve evaluation ...... 97 3.5 SURFACE CHEMISTRY ...... 97 3.5.1 PDITC (for DNA immobilization) ...... 98 3.5.2 GOPTS (for DNA or protein immobilization) ...... 98 3.5.3 Ni-NTA (for protein immobilization by His tag) ...... 98 3.6 BIOCHEMISTRY ...... 99 3.6.1 DNA amplification and analysis ...... 99 3.6.2 Microarray spotting...... 99 3.6.3 In situ protein microarray synthesis ...... 102 3.6.4 Aptamers ...... 102

4 EXPERIMENTS AND RESULTS ...... 103

4.1 BASIC ANALYSES AND PREPARATIONS ...... 103 4.1.1 Contact angles ...... 103 4.1.2 Experimental flow analysis ...... 103 4.1.3 Analysis of temperature control and homogeneity ...... 108 4.1.4 DNA array spotting ...... 108 4.2 CELL-FREE PROTEIN MICROARRAY SYNTHESIS IN FLOW CELL ...... 110 4.2.1 Concept 1 ...... 110 4.2.2 Concept 2 ...... 111

4.2.3 Concept 3 ...... 112 4.2.4 Benchmarking of concept 2 with DAPA ...... 113 4.2.5 Evaluation of concepts ...... 115 4.3 MOLECULE-PROTEIN INTERACTION ANALYSIS BY REFLECTOMETRIC INTERFEROMETRY ...... 116 4.3.1 Basic evaluations of iRIf method ...... 116 4.3.2 System noise evaluation of iRIf setup ...... 124 4.3.3 Bio-physical evaluations of elaborated iRIf setup ...... 135 4.3.4 Biological applications with automated iRIf setups ...... 141

5 CONCLUSION AND OUTLOOK ...... 150

5.1 CONCLUSION ...... 150 5.1.1 Objective 1 - protein microarray synthesis in microfluidic incubator ...... 150 5.1.2 Objective 2 – automated real-time label-free detection system ...... 152 5.2 OUTLOOK ...... 154

REFERENCES ...... 157

ACKNOWLEDGEMENTS ...... 168

CURRICULUM ...... 169

APPENDIX ...... 171

xii

Nomenclature

1λ-iRIf Single wavelength imaging reflectometric interferometry AMD Aminodextrane API Application programing interface bBSA biotinylated BSA BLI Biolayer interferometry BSA Bovine Serum Albumine CCD Charge coupled device c-Myc tag Polypeptide protein tag from the c-Myc-gene CRP C-Reactive Protein DIC N,N’-Diisopropyl-carbodiimide DI Bi-distilled water DMF N,N’-Dimethylformamide DNA Deoxyribonucleic acid EDTA Ethylenediaminetetraacetic acid ELISA Enzyme-linked immuno sorbant assay erDNA expression ready DNA GA Glutaric anhydride GFP Green fluorescent protein GOPTS 3-Glycidyloxypropyltriethoxysilane GST tag Glutathione S-transferase tag GUI Graphical user interface Halo tag Hydrolase with genetically modified active site binding to chloroalkane linker His tag Polyhistidine tag HPLC-MS High performance liquid chromatography – mass spectrometry IF glass Interference glass (standard RIfS transducer) JAR Java archive JDK Java development kit iRIfS imaging reflectometric interference spectroscopy LED Light emitting diode LUT Lookup table NHS N-Hydroxysuccinimide Ni-NTA Nickel-nitrilotriacetic NMR Nuclear magnetic resonance PBS Phosphate buffered saline PCR Polymerase chain reaction PDMS Polydimethylsiloxane (silicone) PDITC 1,4- Phenylendiisothiocyanate PID Proportional-integral-differential (controller) QCM Quartz crystal microbalance RIf Reflectometric interferometry RIfS Reflectometric interference spectroscopy RNA Ribonucleic acid ROI Region of interest

xiii

SNR Signal to noise ratio SPR Surface plasmon resonance Strep-Cy5 Cy5-Streptavidin conjugate TACN 1,4,7-Triazacyclononane TIR Total internal reflection

xiv 1.1 Methods for protein microarray synthesis

1 Introduction

1.1 Methods for protein microarray synthesis

Genome analysis is nowadays used for a variety of applications, such as biomarker identification for diseases including cancer, development of pharmaceutical products, analysis of mutations and detection of drug resistances. Established technologies for the generation of DNA microarrays allow the examination of up to 1 million sequences within one analysis [10], nearly covering the whole human genome. However, genes exert their function at the level of their encoded proteins, and therefore genome analysis does not give the functional insights that proteome analysis can give [11]. On the molecular level, protein- protein and molecule-protein interactions orchestrate the complex network of cellular metabolism [12]. Functional proteome analysis therefore is a complex task which can be only tackled at the protein level, for example by the use of protein microarrays [1, 13–16]. The standard method to create a protein microarray is to express recombinant proteins in cells, purify them from the lysed cells and dispense the protein solution onto a microarray matrix [17]. This has to be carried out for each protein individually. Even mass-produced recombinant protein microarrays such as the HuProt Array from CDI-lab with ~ 20 000 proteins and the ProtoArray from LifeTechnologies with ~ 9000 proteins cost more than 1000 USD (price in 2014) per array [1]. Another strategy to generate protein microarrays is the stepwise chemical in situ synthesis directly on the array by chemical linking of amino acids [18]. Due to low synthesis yield the generated molecules are mere peptides, limited to length of 20 to 60 amino acids. However, these peptide arrays can be used in applications such as epitope mapping or phosphorylation site analysis [19].

Re-thinking the protein arraying process, several approaches were developed for the in situ expression of proteins on the array by cell-free expression systems [1, 20–24]. A basic requirement for in situ cell-free expression of proteins is a template DNA providing several functional elements in addition to the actual protein coding sequence. These elements include a promoter and a terminator for mRNA transcription and a ribosome binding site for initiation of translation [25]. Such functional DNA is designated in this work as ‘expression ready DNA’ (erDNA).

Presented in 2001, Protein In Situ Array (PISA) [1, 26], generates protein microarrays by mixing erDNA with the cell-free expression system directly before dispensing (Fig. 1.1, A). Small droplets of the mixture are subsequently transferred onto a Ni-NTA surface, where

15 1.1 Methods for protein microarray synthesis protein expression takes place in each droplet. An encoded His tag ensures the specific binding of the in situ expressed protein.

Figure 1.1: evolution of cell-free protein microarray generation systems. All systems are based on a protein capture surface, which immobilizes the expressed proteins. (A) PISA contains in a droplet of liquid both free DNA and cell-free system. (B) NAPPA uses pre- spotted, immobilized DNA and adds cell-free system later at any time in a thin liquid layer. (C) DAPA separates the microarray of immobilized DNA from the protein microarray via a membrane soaked with cell-free system. Several protein microarray copies can be made from one DNA microarray. [1] (image by Dr. Günter Roth)

While the PISA system circumvents the effort of standard recombinant expression in cells and purification, mixing each template DNA sequence with cell-free expression system prior to the transfer onto the microarray matrix is still tedious. The dispensing of thousands of spots considering concentrations and time could only be realized reliably with a pipetting robot. In 2006 a modified PISA method was used to express 384 erDNA constructs [27]. First an erDNA microarray is spotted and possibly stored. Next, cell-free expression system is spotted on top of each erDNA spot; this technique allows to create microarrays of 13,000 spots per slide.

A system first published in 2004, Nucleic Acid Programmable Protein Array (NAPPA) [1, 28], is immobilizing biotinylated erDNA and biotinylated anti-GST antibody on a Streptavidin coating (Fig. 1.1, B). For protein expression the complete erDNA microarray is covered with a cell-free expression system. As the expressed proteins contain a GST tag they get bound by the co-spotted antibodies. In 2008 a NAPPA array with ~ 1000 proteins was published [29]. A so-called HaloTag-NAPPA was introduced recently (in 2016) having created a high-density protein array comprising 12,000 Arabidopsis ORFs to query protein–protein interactions for a set of 38 transcription factors and transcriptional regulators (TFs) that function in diverse plant hormone regulatory pathways [30] . NAPPA in comparison to modified PISA, allows to create a protein array without the need for a precise second spotting run. A disadvantage is that NAPPA uses a dual capture surface system (Streptavidin for the erDNA and antibodies for the proteins) which increases the probability of unspecific binding [31] and reduces possible spot density.

16 1.1 Methods for protein microarray synthesis

In 2008 He et al. published the DNA Array to Protein Array (DAPA) system (Fig. 1.1, C) [1, 2]. DAPA was the first system which allowed to re-use an erDNA microarray several times as master to generate several protein microarray copies. In the DAPA method, a membrane, soaked with a cell-free system, is placed onto the erDNA microarray and then superimposed with a Ni-NTA protein capture surface. His tagged proteins expressed from the erDNA microarray diffuse through the membrane and bind to the Ni-NTA surface. Finally, the sandwich is disassembled, both DNA and protein microarray are washed and ready for further use. Due to the diffusion of mRNA and protein, the protein spots are by principle larger than the DNA spots. Compared to PISA and NAPPA the most obvious advantage of the DAPA system is its inherent capability of using the erDNA microarray as master for generating several protein microarray copies. Nevertheless, the DAPA procedure has some significant drawbacks:

1) The physical contact of the solid membrane and the glass slides can violate both the erDNA and the protein microarray. Such it is potentially limiting the lifetime of the DNA microarray and compromising the later use of the protein microarray.

2) The membrane is typically a non-woven, inhomogeneous material, which restricts the free diffusion of both, proteins and cell-free system resources; such it slows down the cell- free protein synthesis.

3) The inhomogenous structure of the membrane inherently limits the reproducibility of the copying process as each copy process needs a new membrane with different local inhomogeneities, leading to differences in protein expression, diffusion and final copying pattern.

4) Due to diffusion the protein spot diameter increases with the thickness of the membrane, therefore a thinner membrane would be favorable but more difficult to handle. A trade-off between good handling characteristics and thinness has to be chosen, limiting the maximal resolution.

5) The protein expression immediately starts as the membrane soaked with cell-free system is touching the erDNA microarray. This happens obligatory during the assembly of the erDNA microarray, the membrane and the Ni-NTA protein capture slide. Temporal separation of assembly and reaction start is therefore not possible.

6) In practical terms, assembling the membrane sandwich is challenging (Fig. 1.2). Drying-out of the soaked membrane and enclosure of air bubbles can hardly be avoided however, both lead to defects in the resulting protein array copy [32].

7) For every 1 cm² of the membrane 10 µl of cell-free expression system is needed, in total > 180 µl/slide. [33]

17 1.1 Methods for protein microarray synthesis

Figure 1.2: assembly of DAPA system with 1 and 9, outer Aluminium pressure plates; 2 and 8, rubber spacers; 3 and 7, layers of parafilm for airtight seal; 4, protein capture slide; 5, Durapore membrane; 6 DNA array slide [33].

Despite listed drawbacks the DAPA system is considered as being state of the art of in situ DNA to protein microarray expression technology capable to realize protein microarrays of 128 different human proteins, inherent problems are revealed by expression results depicted in Fig. 1.3:

Figure 1.3: 128 different human proteins expressed cell-free, each protein with 4 spots at 1 mm pitch; inhomogeneities between equal neighbouring spots, efficiency gradient from left to right (s. red boxes). Damage by mechanical abrasion (orange ellipse). [34]

DAPA’s capability of creating multiple protein microarrays of one DNA template is remarkable too however, lacking robustness as 2 of 20 sucessive expressions failed and missing reproducibility as expressed arrays look rather different (Fig. 1.4):

18 1.1 Methods for protein microarray synthesis

Figure 1.4: multiple protein microarrays expressed from a single DNA array template. A series of consecutive replicates by DAPA, incubation and detection conditions were identical for each print. 18 out of 20 attempted repeats were successful (two failed due to handling reasons) however, reproducibility and homogeneity are rather poor. Blue array in left upper corner is PCR-amplified DNA encoding GFP directly labeled with Cy5. Red, immobilized GFP is detected by immunostaining. Protein arrays are mirror imaged to match DNA array. [2]

The protein expression reproducibility of 408 identical template spots of the DNA binding domain of transcription factor E2F6, His- and c-Myc tagged is analysed [34] – again the basic functionality is proven but lacking reproducibility:

A B C D

Figure 1.5: (A) array of 408 identical template spots of the DNA binding domain of transcription factor E2F6, His- and c-Myc tagged, direct fluorescence label on the DNA array. (B - D) 3 successive DAPA protein array replications expressed from the left DNA array, secondary immunofluorescence against c-Myc tag on the protein array, identical scanner settings for all protein arrays. [34]

19 1.2 Optical detection of molecule-protein interactions

1.2 Optical detection of molecule-protein interactions

Label-free technologies for the analysis of molecule-molecule interactions have the intrinsic advantage of no requirement for an additional label, e.g. an isotope, a fluorescent dye, a fluorescent protein, an enzyme, a quantum dot, or a nanoparticle. These labels could at least make the system more artificial [35]. Several label-free detection methods, e.g. HPLC-MS, NMR, calorimetry allow the determination of affinity. Especially biosensor based methods, e.g. SPR [36, 37], QCM [38], BLI [39], and RIfS [35] determine affinity and provide detailed insights into molecular kinetics on single spots or even microarrays. However, biosensors implicate one reactant to be immobilized while maintaining its initial activity [35]. While the SPR systems are based upon the measurement of changes of the refractive index via surface plasmon resonance [40] relying on the evanescent field of electromagnetic radiation, the RIfS System is based on measuring the changes of the so-called optical thickness of the reagent binding layer, that is the product of the physical thickness and the refractive index. Such it is a combination of refractometry and reflectometry [41].

A major advantage of the RIfS technology is the low bias to temperature changes [42, 43], as temperature changes effect the change of the refractive index and the change of the physical layer thickness vice versa. Purely refractometric methods as SPR are very sensitive to temperature variations because of the high temperature impact on the refractive index. It is dependent on the density given from the Clausius–Mossotti equation, and the density is dependent on the thermal expansion; the refractive index decreases with increasing temperature. Therefore changing temperatures during purely refractometric measurements result in measurement errors if systems are not stabilized with an excellent temperature control system [43, 44].

The spatial detection range of evanescent field techniques in comparison to RIfS technology is much lower: they monitor the distance from the surface of the transduction element with an exponential decay, e.g. visibility is reduced to 1/e at a distance of ∼ 250 nm from the transducer. Reflectometric interference spectroscopy (RIfS) however demonstrates a linear dependence of the physical thickness in homogeneous media [43].

RIfS technology as a label-free detection method offers further advantages in comparison to evanescent field techniques [45]. The analysis of the spectroscopical measurement can even distinguish between a change in the optical thickness n*d as function of the wavelength shift of the curve minimum and the change in refractive index n as function of the intensity variation [46]. However, due to the significant calculation times for the evaluation by curve- fitting techniques it cannot be used for on-line monitoring of molecular immobilization

20 1.2 Optical detection of molecule-protein interactions

processes, especially if several binding events of a microarray should be monitored in parallel.

We aim to have an imaging system for microarrays consisting finally of thousands of spots. Therefore the RIfS system inherent capability of distinguishing optical thickness and refractive index is abandoned for the significant advantage of having an imaging detection system suitable to monitor in real-time binding events on microarrays and to save the data for later analysis. 1λ-iRIf is not any more analyzing the shift of the white light spectrum but measuring the intensity change due to molecule adsorption at one specific wavelength [47] (Fig. 1.6).

A B

Figure 1.6: (A) Single wavelength imaging reflectometric interferometry (1λ-iRIf) evolved from RIfS measures the intensity change due to molecule adsorption at a specific wavelength. The computation time for the analysis is reduced significantly [47]. (B) A LED is used as light source, reflective physical layers as thin film stack cause partial reflections and interference - a glass slide of 1 mm thickness and refractive index of 1.52, a semi-reflective layer of Ta2O5 with thickness of 10 nm and refractive of index 2.2, a SiO2 layer with thickness of 330 nm and refractive index of 1.46. Reflection at the outer interface is affected by the refractive index of the liquid (e.g., PBS with a refractive index of 1.33). Reflected waves I1, I2 and I3 interfere, I0 and possibly reflection from rear side of fluidic channel cause noise.

As depicted in Fig. 1.6 the relative intensity of the reflected beam is depending on the selected wavelength. The change of relative intensity due to molecule adsorption further depends on the appropriate selection of the reflective layers. This system of wavelength and layers should be designed such that the change of the optical thickness by molecule adsorption is leading to a high reflectivity change [48]. For quantifying measurements e.g. concentration or affinity measurements it is essential that the reflectivity changes are linear within the measuring range [49]. This technology allows to perform high-throughput measurements on microarrays by using a digital camera as detector [50, 51] however, the state-of-the-art prototype system design of 2010 lacked functionality and integration:

21 1.2 Optical detection of molecule-protein interactions

1) Rudimentary flow cell based upon PMMA plate with O-ring lacks precision and structural flexibility, e.g. no flow guides for priming. Rather high flow chamber leads to high reagent consumption. Rear of flow chamber lacks absorption and hence, scatter light by non-coherent reflection increases (Fig. 1.7).

Figure 1.7: simple flow cell for proof-of-concept. Brackets clamp glass slide with immobilized molecule spots together with PMMA plate, chamber made and sealed by oval O-ring. [5]

2) Flow control is purely manual by Peristaltic pump neglecting the demand for continuous flow during a sequence of buffer and reagents. (Fig. 1.8).

Fig. 1.8: prototype manual fluidic control system. [5]

3) Lack of integrated process control and data processing (Fig. 1.9)

a. In order to reduce the amount of data, areas with assumed spots need to be marked before the experiment starts  no precise identification of rather dense arrays possible.

b. The average data of these assumed spots is saved after the experiment for post-analysis  no real-time depiction of binding event possible.

c. Implementation of data acquisition and processing with NI Vision and LabView  no licence free run time environment; limited functionality.

22 1.3 Objectives of the thesis

Figure 1.9: data flow of state-of-the-art iRIf system lacking real-time ability, efficient data flow and precise spot identification. [5]

1.3 Objectives of the thesis

The task to realize an “Automated system for the cell-free protein microarray synthesis and the label-free molecule-protein interaction analysis” was split into two major objectives: first create a microfluidic system enabling protein microarray synthesis, second realize a label- free detection system as part of an automated setup suitable for protein interactions analysis within that microfluidic system.

Beyond this thesis the protein microarray generation should be monitored and controlled by the same label-free monitoring system as developed for the interaction analysis. Such the vision of an integrated, fully controllable system for proteome analysis will be realized.

1.3.1 Objective 1 - protein microarray synthesis in microfluidic incubator

Summary:

First objective of this thesis is to devise, realize and evaluate a technical system allowing to express a microarray of immobilized DNA strands into a microarray of immobilized proteins with local correspondence in a microfluidic gap: o Creating generic microfluidic flow cells with temperature control. o Analysing and establishing protocols for DNA and protein immobilization. o Analysing and establishing protocols for DNA amplification and protein expression. o Establishing proof-of-concept for protein expression in microfluidic flow cell. Details:

To circumvent the significant drawbacks of the state-of-the-art DAPA protein microarray synthesis system, the basic system design of DAPA will be replaced by a microfluidic concept eliminating all drawbacks of the DAPA membrane and allowing for microfluidic control. A

23 1.3 Objectives of the thesis system will be created that combines the robustness and simplicity of NAPPA, the controllability and precision of microfluidic systems as PING [52] and the capability of generating multiple protein microarray copies from a single DNA microarray according to DAPA. By now several approaches for cell-free protein synthesis in microfluidic devices have been already described [53–55] however, they focus on the optimization of the protein production yield.

This thesis introduces generic flow cell systems for the robust, quick and cost efficient generation of protein microarrays expressed from DNA microarray templates. Basically a microfluidic gap between a DNA template array slide and a protein capture slide needs to be created and sealed. Aside simple approaches for the proof-of-concept, especially a design consisting of 2 glass slides and a design consisting of one Polydimethylsiloxan (PDMS) slide and one glass slide will be realized and analysed. The moulded PDMS slide will be microstructured by photo-lithographical processes applied to the mould master, enhancing stability, bubble-free priming and fluid control in order to increase the reproducibility of the microarray synthesis process. The systems will be equipped with a temperature control unit to optimize the cell-free expression process and to allow for general process control, e.g. cooling to decelerate the reaction process. The microfluidic gap, that is the distance between the DNA microarray (Fig. 1.10, d) and the protein capture surface (Fig. 1.10, e), will be realized with heights of 20 – 100 µm. Such the system allows to be automized rather easily and incorporates both a high controllability and robustness. The flow cells will be designed generic in some essential aspects in order to be adaptable and optimizable during the iterative process development. They should either be easy to clean or exchangeable at low costs and time efforts. The design of the flow cell clamping system is essential for the quick and easy assembly. Finally flow cell’s suitability for the protein microarray synthesis needs to be demonstrated as proof-of-concept.

(b) (d) (a) (c) (f) (e)

Figure 1.10: (a) DNA template; (b) RNA transcribed from DNA template is translated by ribosome; (c) protein translated from RNA; (d) DNA microarray; (e) protein microarray; (f) microfluidic chamber filled with cell-free system. (drawing by G. Roth)

24 1.3 Objectives of the thesis

1.3.2 Objective 2 – automated real-time label-free detection system

Summary:

Second objective of this thesis is to build an imaging label-free interaction analysis system based upon the patented reflecto-interferometric principle [56] (re-branded to SCORE (Single Colour Reflectometry) in 2016 [57]): o Generic microfluidic flow cells compatible to objective 1. o Design of a fully automated, user-friendly platform incorporating the single wavelength iRIf principle for detection. o Physical analysis, especially noise and sensitivity evaluation of setup. o Suitability analysis of total system by assays of biological impact e.g. antigen-antibody binding assay, selection of aptamer binders, detection of serum immunizations. o Beyond this thesis but kept in mind: Formation of protein microarray (objective 1) should be monitored by label-free detection system (objective 2) such that the formation process can be controlled e.g. by temperature. Protein creation and analysis should be integrated to one system, such that proteins do not denaturate between their in situ creation and the subsequent binding analysis. Details: This work now introduces an automated 1λ-iRIf system with a generic microfluidic flow cell, an electrically controlled fluidic system and an optical detection unit ready for the detection of binding events on microarrays being immobilized on glass slides in standard format.

Fluidic control system

Figure 1.11: schematic of total analysis system setup - fluidic control system, microfluidic flow cell with temperature controller, optical detection system and personal computer for system control, data acquisition and processing.

25 1.3 Objectives of the thesis

The flow cell design has to meet the demands both of the in situ protein expression system according to thesis objective 1 (chapter 1.3.1) and of the total analysis system. Such it basically consists of a microscope glass slide sealing a microfluidic chamber with inlet and outlet. The chamber should be of very low height (< 100 µm) for low reagent consumption, but high enough to prevent collaps during assembly.

The detection system will consist according to the state-of-the-art iRIf technology of a digital camera, a telecentric lens system with single wavelength LED incorporated and semi- reflective layers coated onto the microscope glass slide. The rear side of the flow cell should be absorbing such that non-coherent reflected light will not significantly add to the system noise level. Any binding of proteins to pre-spotted molecules should be detected with high SNR.

Figure 1.12: optical detection system consists of an optical detector (digital camera), a telecentric lens system with single wavelength LED excitation and a semi-reflective microscope slide sealing a microfluidic flow cell; partial reflection occurs at 4 interfaces, most right interface moves by adsorption of molecules in flow cell.

The fluidic system should be capable to realize a continuous flow of buffer and several reagents during the experiment. Rather precise flow rates of some nl/min to several ml/min should be adjustable. Reagent volumes of few microliters to 1.5 ml should be selectable. The systems tubing should be suitable for biological reagents and the dead volumes should be < 60 µl. Taylor dispersion should be kept minimal. The process control should allow user- friendly programming of flexible protocols for selecting > 20 reagents from sampler, parameterizing liquid flow, liquid volume and temperature. Several applications such as in situ protein synthesis and/or complex molecule-protein interaction analysis should be feasible. The image processing modules should allow the real-time visualization of the binding event and the (semi-)automated spot identification for the final analysis of the interactions on each spot resulting in a kinetic curves plot with curve fittings for the determination of the kinetic rate constants and the binding constants KD.

26 1.4 Structure of thesis

1.4 Structure of thesis

This thesis is basically composed of 5 parts:

1) Introduction

It gives an insight into state-of-the-art protein microarray generation methods being especially relevant for this thesis. It introduces optical detection methods for the molecule- protein interaction analysis, especially the patented reflecto-interferometric principle as label-free method selected for the setup developed within this work. It reveals the objectives of the thesis in a final section.

2) Theoretical background

Physical, chemical and biological essentials relevant for the major parts of this thesis, i.e., microfluidics, molecule transport in a microfluidic flow cell, diffusion of the molecule to its immobilized target, binding to this target, optical detection of the binding process by reflecto-interferometry, digital data acquisition and processing as well as the analysis of the binding curves for the binding constants determination are explained to a relevant extent.

3) Materials and Methods

This chapter introduces all aspects about the design of the flow cell, the flow cells clamping mechanism and the overall design of the label-free monitoring prototype with mechanical, fluidical, thermical, optical, process controlling and image processing aspects. Further it describes the surface coating procedures for molecules immobilization and protocols for the DNA amplification and the protein expression.

4) Experiments and Results

The cell-free DNA microarray expression in the newly devised microfluidic cells is analyzed and compared to the DAPA system. The automated setup is evaluated physically, e.g. noise evaluation, sensitivity analysis, and by applications with biological impact, e.g. antibody-antigen binding, Thrombin-aptamer binding, serum immunization analysis.

5) Conclusion and Outlook

Comparison of introduced objectives / specifications with the final results. Potential applications and hints for further optimizations are given.

27 2.1 Microfluidic systems

2 Theoretical Background

2.1 Microfluidic systems

2.1.1 Capillary pressure in rectangular flow cell

The capillary pressure between two parallel surfaces of different contact angle (Fig. 2.1) is given by equation 2.1 [58]. The height of channel is assumed to be very small in comparision to the width of the channel.

휎 푝 = ∗ (cos 휃 + cos 휃 ) (2.1) 푐푎푝 ℎ 1 2 where

휃1 = advancing contact angle of surface 1

휃2 = advancing contact angle of surface 2 h = height [m]

휎 = surface tension [N/m]

푝푐푎푝 = priming pressure [Pa] Figure 2.1: capillary pressure between two parallel surfaces of different contact angle.

2.1.2 Pinning on horizontal edge

The wettability of microstructured surfaces can be affected by geometrical structures. Such the static advancing contact angle of the planar surface can be modified significantly [59, 60]. We now consider horizontal structures as pinning elements, e.g. phase guides [61]:

Figure 2.2: contact line pinning at horizontal line, e.g. phase guide. Simplified depiction of equal 휃 values for both top and bottom plate. The apparent contact angle 휃푎 at the edge can attain values of

28 2.1 Microfluidic systems

휃 < 휃푎 < 훼 + 휃 (2.2)

Having different materials for top, bottom and side walls, equation 2.1 expands to

cos 휃 cos 휃 cos 휃 푝 = 휎 (2 ∗ 푠 + 푡표푝 + 푏표푡푡표푚) (2.3) 푐푎푝 푤 ℎ ℎ where

h = height of channel [m]

휃푠 = contact angle at side plates

휃푡표푝 = contact angle at top plate

휃푏표푡푡표푚 = contact angle at bottom plate

Concus Finn condition

Menisci are unstable in corners, if the sum of contact angle 휃 and half corner angle α is smaller than 90° [62]. The edge is called critical, as liquid would run endless. Considering an orthogonal corner, the meniscus is unstable for contact angles smaller 45°.

2.1.3 Laminar flow

In laminar flow the flow profile of the liquid is parabolic. Stream lines are smooth without any appearance of vortex. Laminar flow is highly predictable and mostly apparent in microfluidic devices. In opposite to laminar flow there is turbulent flow which is merely statistically predictable (Fig. 2.3), the flow profile flattens with increasing Reynolds number [58, 63].

Figure 2.3: laminar versus turbulent flow between two flat boundaries. Increasing Reynolds numbers cause flattening profiles.

In stationary laminar flow, pressure driven force is equal and opposite of the frictional force.

29 2.1 Microfluidic systems

Figure 2.4: steady laminar flow between two fixed plates.

Applying the boundary conditions for velocity in y and z direction to be zero, maximal fluid velocity is calculated by

−ℎ2푑푝 푢 = (2.4) 푚푎푥 2휇 푑푥 where

h = half height of channel [m] dp = pressure drop [Pa] dx = length of channel [m] µ = dynamic viscosity [Pa*s]

푢푚푎푥 = maximal velocity [m/s]

2.1.4 Taylor disperison

Within the channels the flow profile has a parabolic shape with the maximum flow velocity in the center and zero flow at the channel wall while applying the no-slip boundary condition. This flow profile leads to different local flow velocities resulting in different passage times of targets depending on their position in the channel. Thus, a sample plug broadens axially while flowing through the channel. Furthermore, diffusion will additionally broaden the plug both in axial and lateral directions smearing out the concentration distribution (Fig. 2.5). The combination of these two effects is called Taylor dispersion [64, 65]. As timing information is lost, the Taylor dispersion limits the applicability of continuous- flow analyses.

30 2.1 Microfluidic systems

Figure 2.5: Taylor dispersion: a) sample plug in a microfluidic channel with a parabolic flow profile. b) Due to the flow profile the sample plug will be deformed and smeared out downstream. c) Diffusion additionally broadens the sample plug resulting in reduced concentration of sample distributed over a larger volume. [65]

2.1.5 Flow boundary condition

The no-slip condition for viscous fluids states that at a solid boundary the fluid will have zero velocity relative to the boundary. The assumption that a liquid adheres to a solid boundary is one of the central tenets of the Navier-Stokes theory. However, there are situations wherein this assumption does not hold. Usually, the slip is assumed to depend on the shear stress at the wall. However, a number of indicators suggest that the slip velocity also depends on the normal stress. In regions where the slip velocity depends strongly on the normal stress, the flow field in a channel is not fully developed and rectilinear flow is not possible [66, 67]. In general, traditional methods such as the Mooney method cannot be used for calculating the slip velocity [68]. Slip determination by experiments requires utmost care as pointed out by Mackay and Henson [69]. Studies of wall slip indicate increasing flow velocities in microchannel, the slip velocities on wide wall are larger than those on narrow wall [70]. When flow boundary conditions change from stick or weak slip to strong slip at any location along the length of the wall, undamped periodic oscillations in pressure and mean velocity have been observed in PDMS flow cells [71]. No-slip boundary condition is found in all wetting situations; on strongly hydrophobic surfaces, water undergoes finite slippage that increases with hydrophobicity [72].

2.1.6 Navier Stokes equation

The Navier-Stokes (NS) equations are considered to be the central differential equations describing the motion of incompressible Newtonian fluids. Basic assumptions are continuous media and continuum mechanics [73]. These assumptions are fulfilled in macrofluidics as well as

31 2.1 Microfluidic systems in microfluidics down to geometrical sizes of 10-100 nm for liquids. Simplifications reduce the Navier-Stokes equation to

∇푃 = µ∇2푣 (2.5)

This equation is also known as Poisson equation, which shows that in stationary laminar flow, pressure driven force is equal and opposite of the frictional force. Considering further the velocity in y and z direction to be zero leads from equation 2.5 to the single partial differential equation

∂p ∂2u(z) = µ ( ) (2.6) ∂x ∂z2

휕푝 Partial integration with respect to z while is treated as a constant with respect to y 휕푥

푧 휕푝 푢(푧) = ∬ 휕푧2 (2.7) 0 휇∗휕푥

1 휕푝 푢(푧) = 푧2 + 푐 푧 + 푐 (2.8) 2µ 휕푥 1 2

As boundary condition we apply the no-slip conditions i.e., the fluid is "stuck" to the plates leading to u = 0 at z = ±h; such the coefficients are

푐1 = 0 (2.9)

ℎ2 휕푝 푐 = − (2.10) 2 2µ 휕푥

The velocity at any position between the center (z=0) and the boundary (z=h) is given by

1 휕푝 푢(푧) = (푧2 − ℎ2) (2.11) 2µ 휕푥

As depicted in Fig. 2.4 the laminar flow between 2 plates reaches a maximal velocity in the middle of the channel at z=0 according to equation 2.4.

Integrating the parabolic function of equation 2.11 from the lower to the upper plate leads to the total flow area, the normed volumetric flow

ℎ ℎ 1 휕푝 ℎ3 휕푝 푞 = ∫ 푢휕푧 = ∫ (푧2 − ℎ2)휕푧 = − (2.12) −ℎ −ℎ 2µ 휕푥 3µ 휕푥

Please note that the flow is negative, i.e. to the left, for a positive pressure gradient, dp/dx, this is due to the gradient definition where decreasing pressure to the right is negative. The total volumetric flow is the integral of q along the width of the flow cell

32 2.2 Diffusion of molecules

푤 푄 = 푞휕푦 (2.13) ∫−푤

Considering q being a constant along the width of the flow cell leads to

푄 = 푞푦 (+푐1) (2.14) In general q is a function of y and as such needs to be integrated. Only if w >> h equation 2.14 is true for the major areas of the flow cell, i.e. the areas at a significant distance d >> h off the lateral borders. Areas closer to the lateral borders within distances of dy < 2 times of the flow cell height should be analysed separately in case of interest [74, 75].

2.2 Diffusion of molecules

Diffusion is the spontaneous spreading of molecules driven by a gradient in concentration. It leads to uniform distributions of initially non-homogenous separated molecules such that the entropy is maximized. Every molecule performs random motion due to atomic collisions, the so called Brownian motion [76, 77].

Any concentration gradients in an enclosed volume will be counteracted according to Fick’s law’s [78]. The system always seeks for homogeneity, the state of highest probability. The flux jN is a function of the concentration gradient dc/dx and the reagent specific diffusion coefficient D [m²/s].

휕푐 휕푗 휕2푐 = − 푁 = 퐷 (2.15) 휕푡 휕푥 휕푥2

2.2.1 Characteristic diffusion time

Assuming a low Reynolds number, low solvent viscosity and a molecule size > 5 times larger than the solvent molecule size the diffusion constant can be calculated by the Stokes- Einstein equation

푘푇 퐷 = (2.16) 6휋휂푅

The characteristic diffusion time 푡퐷 for spheric diffusion is calculated by

푙 2 푡 = 퐷 (2.17) 퐷 퐷 where k = Boltzmann constant (1.38064852 × 10-23 m2 kg s-2 K-1) T = temperature

33 2.2 Diffusion of molecules

η = liquid viscosity (water (300 K) ~ 1 mPas) R = molecule radius (IgG ~ 4 nm)

푙퐷 = diffusion length

Hence, an antibody of radius R = 4 nm has a diffusion constant of D = 0.05 × 10-9 m2 s-1 at T =

300 K. Assuming a diffusion length of lD = 30 µm (~ flow cell height), a diffusion time 푡퐷 = ~18 푠 is resulting.

2.2.2 Anomalous diffusion

Anomalous translational diffusion and subdiffusion, respectively, is a breakdown of the laws of mass action [79]. As opposed to normal translational diffusion, in which the movement of molecules is not significantly correlated with their previous position, anomalous translational diffusion molecules are spatially and temporally correlated [80, 81]. This spatio-temporal correlation reflects a fundamentally different behavior compared with the one in dilute or very dilute solution, e.g. it is a general phenomenon (also called molecular crowding) in living cells and their compartments such as cytoplasm, nucleoplasm and membranes [79, 80, 82– 87]. There is solid evidence for analyzing fluorescence correlation and dual color fluorescence crosscorrelation spectroscopy data (FCS and dual color FCCS) in cellular applications by equations based on anomalous subdiffusion [88]. Using equations based on normal diffusion causes artifacts of the fitted biological system response parameters and of the interpretations of the FCS and dual color FCCS data in the crowded environment of living cells. Equations based on normal diffusion are not valid in living cells. If even a single binding site is present, there is a very short, almost artifactual, period of anomalous subdiffusion, but a hierarchy of binding sites extends the anomalous regime considerably [89]. Without choosing the proper equations for anomalous subdiffusive behavior, the fitted biological system response parameters are likely meaningless and wrong, respectively [90].

In uniform Euclidean systems, the mean-square displacement MSD of a random walker is proportional to the time t for any number of spatial dimensions d. However, in disordered systems, this physical law is not valid in general, and the diffusive law becomes anomalous and subdiffusive, respectively, as given by

α α 푀푆퐷(푡) = Γα ∗ 푡 ∝ 푡 (2.18)

A linear dependence, i.e. α =1, is a fingerprint of normal translational diffusion in dilute or very dilute solution. Any other value of α corresponds to anomalous translational diffusion. Subdiffusive motion (0 < α < 1) can appear due to geometric and/or energetic disorders [83].

34 2.3 Biosensor technology

2.3 Biosensor technology

Biochemical sensors add to the nonselective transducers of physical sensors (thermometers, photodiodes, etc.) a recognition layer on top that allows to selectively detect specific analytes. This recognition layer is shielded by a biopolymer layer being tailored to reduce nonspecific binding and to increase the number of recognition elements by surface enhancements [43]. Aside the recognition elements and the shielding layer a biosensor relies mainly on the transduction principle. The transducer surface, e.g. glass, metal is usually silanized, onto the siloxane groups e.g. a hydrogel (aminodextran or carboxydextran) is covalently bound. This provides good shielding effects and multiple sites for the immobilization of recognition elements, ~ 20 ng/mm² binding capacity. Polyethylene glycol (PEG) is a useful shielding layer too. It has a lower protein binding capacity of ~ 6 ng/mm² due to monolayer formation. For complex matrices as blood, PEG better reduces nonspecific binding. Further biopolymer layers as biotin-(strept)avidin biolayers, polyelectrolyte layers, histidine tags, halo tags [30], membranes, or even biomimetic membranes are known [49, 91–94]. Specifically synthesized hydrogel materials consisting of water-swollen polymer networks exhibit a large number of specific properties highly attractive for a variety of optical biosensor applications [94]. Renewable sensor surfaces, e.g. based upon magnetic nanoparticles contribute to the design of flexible and robust biosensors [95].

2.3.1 Intermolecular forces

Attraction or repulsion forces between neighboring atoms, molecules, or ions are weak in comparison to intramolecular forces keeping a molecule together, e.g. the covalent bond involving the sharing of electron pairs between atoms. Still they are an essential part of force fields frequently used in molecular mechanics. Quantification of intermolecular forces is obtained either by macroscopic measurements of properties like viscosity, pressure-volume- temperature data or by microscopic aspects given by virial coefficients and Lennard-Jones potentials. Life is based upon biomolecular interactions. Ligands attach to the binding site of a receptor molecule, binding is realized by a couple of rather weak attraction forces being a function of distance and surrounding media, e.g. water:

35 2.3 Biosensor technology

Table 2.1: most relevant biomolecular interaction types and their binding energy in aqueous solutions [96]. type of interaction distance dependency Energy [kJ/mol]

electrostatic interactions 1/d² 12 – 20

hydrogen bonding 1/d6 10 – 30

van-der-Waals forces 1/d10 0,5 – 5

hydrophobic interactions - 0 – 40

2.3.2 Antibody and antigene

Antibodies, also called immunoglobulins are glycoproteins synthesized by the B- lymphocyctes of vertebrates in response to detection of "foreign" substances [97]. The foreign substances, called antigens, may be proteins, polysaccharides, or nucleic acids (bacterias or viruses). Each lymphocyte (white blood cell) and all its decendants synthesize the same antibody; over time, as the vertebrade is exposed to a wide variety of antigens, a large number of antibodies develop. When an antigen binds to an antibody on the surface of a lymphocyte, the cell is stimulated to make more copies of the antibody for release into the blood stream. These "free" antibodies bind additional antigens, forming insoluble complexes that mark the antigens for destruction by proteases or lymphocytes. If an antigen is exposing several binding sites (epitopes), a mixture of antibodies with different specificities is created (polyclonal antibodies). The most abundant antibodies in humans are the immunoglobulin type G, IgG antibodies. They all have the same general structure as depicted in Fig. 2.6:

Figure 2.6: structure of immunoglobulin type G, IgG antibody [98].

Immunoglobulins are roughly Y-shaped molecules or combination of such molecules consisting of two heavy (MW ~ 50000 g/mol) and two light (MW ~ 25000 g/mol) chains.

36 2.3 Biosensor technology

Their structures are divided into two regions - the variable (V) region (top of the Y) defining antigen binding properties and the constant (C) region (stem of the Y), interacting with effector cells and molecules. Immunoglobulins can be divided into five different classes IgA, IgD, IgE, IgM, and IgG based on their C regions, respectively designated as a, d, e, m, and g (five main heavy-chain classes). Most IgGs are monomers, but IgA and IgM are respectively, dimers and pentamers linked by J chains [99].

2.3.3 Assay formats

Certain assays formats allow for the immobilization of antibodies and antigens on solid surfaces, e.g. the direct assay, the sandwich assay, the competitive test format and the binding inhibition test (Fig. 2.7).

Figure 2.7: (a) direct assay: the recognition element is immobilized to the shielding layer, allowing direct detection of the analyte. (b) sandwich assay: interaction between the immobilized recognition element, the analyte, and a secondary recognition element (antibody). (c) competitive test format. (d) binding inhibition test, processes 2a and 3a: no analyte concentration in the sample; processes 2b and 3b: high concentration of analyte in the sample. [43]

The direct assay (Fig. 2.7, a) just immobilizes the recognition layer on top of the shielding layer and requires no other reagent. This simple assay can be used for the detection of either labeled or large enough analytes for label-free detection methods.

If no direct assay is possible, a secondary analyte can be used, a process that is comparable to the first steps of an enzyme-linked immunosorbent assay (ELISA) approach (Fig. 2.7, b). Within a competitive assay the interaction of a recognition element at the surface with a

37 2.4 Biomolecular interaction kinetics labeled analyte (e.g., biofluorescence marker) is substituted by the analyte (Fig. 2.7, c). This depends on the relative concentration of both partners and the relative binding constants. Accordingly, an inverse signal is achieved: A high concentration of the analyte to be detected causes a low signal, whereas for a low concentration, the labeled analyte at the surface is not replaced, exciting a high signal. Because this process depends to a large degree on the relative binding constant, a modification of this assay, called the binding inhibition assay (Fig. 2.7, d), is preferable. In a preincubation step, the sample with the analyte is mixed with a recognition element in the homogeneous phase. In parallel, a derivate of the analyte is immobilized to the surface of the shielding layer. Then, the homogeneous phase is flashed across the interaction partner at the shielding layer, in most cases in a flow-injection type system. If only a few analyte molecules are present in the sample, they block only a small number of recognition elements. The nonblocked recognition elements (e.g., antibodies) can be transported via diffusion to the surface and will interact with the immobilized analyte derivatives. This results in a high signal (Fig. 2.7, d, processes 2a and 3a). If, however, the amount of analyte in the sample is large, then most of the recognition elements are blocked during the preincubation step and cannot go to the surface. Accordingly, no signal or only a poor one will be monitored (Fig. 2.7, d, processes 2b and 3b). This assay also results in an inverse signal-to-concentration relationship [43].

Labeling is time consuming and expensive, it even effects the molecule interaction or hinders the detection. A biosensor should best be label-free, highly sensitive and suitable for high- throughput screenings, e.g. of drugs [100]. Label-free detection methods have been basically developed already in the early 20th century by R.M. Wood discovering the anormal diffraction of light at gratings due to surface plasmon waves stimulation [101], which was later called electromagnetical evanescent wave on thin metal layer surfaces [102].

2.4 Biomolecular interaction kinetics

One basic property of proteins is their ability to specifically target and form non-covalent complexes with other proteins. Such protein–protein interactions play key roles in all cellular processes and functions. Identifying and characterizing protein interactions and entire interaction networks (‘interactomes’) is therefore prerequisite to understanding these processes on a molecular and biophysical level. An average of five interaction partners per protein has been estimated, predicting a problem that far exceeds the complexity of the genome. Furthermore, dissection of the interactome is highly challenging because of the physicochemically diverse properties of proteins and the very different characteristics of

38 2.4 Biomolecular interaction kinetics

protein– protein interactions: equilibrium dissociation constants of protein complexes vary over several orders of magnitude. Further key parameters are the oligomeric state of the interaction partner, the stoichiometric ratio in the complex and the nature of the interaction sites [103]. In this work we analyse all interactions by one to one Langmuir reaction [104].

2.4.1 Receptor and ligand in solution [105]

For both ligand and receptor in solution, the binding follows the law of mass action displayed in equation (2.19), where ka is the association and kd the dissociation rate constant. The concentrations of receptor cR, ligand cL and receptor-ligand complex cRL are in equilibrium linked by the binding constant Kaff, describing the thermodynamical stability of the complex.

푘푎 푐푅퐿 퐾푎푓푓 = = (2.19) 푘푑 푐푅∗푐퐿 The rate for the formation of a receptor-ligand complex is given in equation (2.20). The formation of the complex is dependent on both the concentration 푐퐿 of the ligand as well as nd the concentration 푐푅 of the receptor, such it is a 2 order bimolecular reaction. The dissociation occurring simultaneously is conversely only dependent on the concentration of st the receptor-ligand complex 푐푅퐿 hence, only a 1 order monomolecular reaction [105].

푑푐 푅퐿 = 푘 ∗ 푐 ∗ 푐 − 푘 ∗ 푐 (2.20) 푑푡 푎 푅 퐿 푑 푅퐿 Molecular interactions in solutions are called homogenous reactions, whereas interactions of reagents with ligands being immobilized on surfaces are heterogenous, as one or more physical features of the sensitive layer change.

2.4.2 Ligand immobilized on substrate [105, 106]

Having immobilized ligands the receptor-ligand complex concentration cRL is now called surface load Γ, with Γmax representing the full amount of available binding sites and Γ(t) the amount of receptor-ligand complexes at time t. The ligand concentration cL is becoming the difference of Γmax − Γ(t). Such equation 2.20 for the formation rate of the receptor-ligand complex is becoming

푑Γ(푡) = 푘 ∗ 푐 ∗ (Γ푚푎푥 − Γ(푡)) − 푘 ∗ Γ(푡) (2.21) 푑푡 푎 푅 푑

being transformable into

푑Γ(푡) = 푘 ∗ 푐 ∗ Γ푚푎푥 − (푘 ∗ 푐 + 푘 ) ∗ Γ(푡) (2.22) 푑푡 푎 푅 푎 푅 푑

39 2.5 Determination of dissociation constant KD

This differential equation can be solved with the initial condition Γ (0) = 0 as no receptor- ligand complexes are present at t = 0. This results in equation (2.23) depicting that both association and dissociation affect the kinetics of forming a receptor-ligand complex.

푘 ∗푐 ∗Γ푚푎푥 Γ(푡) = 푎 푅 (1 − e−(푘푎∗푐푅+푘푑)∗푡) (2.23) 푘푎∗푐푅+푘푑

The surface load in the final equilibrium Γeq(t→∞) results in

푘푎∗푐푅∗Γ푚푎푥 Γeq = (2.24) 푘푎∗푐푅+푘푑

Combining equations (2.23) and (2.24) leads to the simplified kinetic equation

−(푘푎∗푐푅+푘푑)∗푡 Γ(푡) = Γeq(1 − e ) (2.25)

Finally a pseudo first order kinetic equation can be created by introducing the observable rate constant kobs

푘표푏푠 = 푘푎 ∗ 푐푅 + 푘푑 (2.26)

Inserted into equation (2.25) leads to

−푘표푏푠∗푡 Γ(푡) = Γeq(1 − e ) (2.27)

2.5 Determination of dissociation constant KD

2.5.1 Fitting of binding curves

Having a constant receptor concentration 푐푅 the observed rate constant kobs is determined by the exponential part of the binding curve. This exponential part of the measured binding curve needs to be identified by analyzing the curvature [65, 105] .

40 2.5 Determination of dissociation constant KD

bi-exponential evaluable evaluable

diffusion behaviour

] limited region region

a.u.

[

(t)/dt

I relative I intensity relative

time [s] Ir Figure 2.8: (left) binding curve separated for evaluation in 3 regions - evaluable region lies between intial part with diffusion limitation (due to rather high ligand concentration, that is low surface load) at the beginning, and final part with bi-exponential behaviour (less accessible ligand binding sites are bound by receptors too, such multiple association rate constants superimpose). (right) The evaluable region can be identified systematically by plotting the signal derivative Ir (t)/dt over the signal Ir (t) and selecting the linear region [8].

The first part of the increasing binding curve is predominantly linear caused by the limitation of mass transport of the receptor through limited diffusion; the surface load is small – rather few receptors have bound to the highly concentrated immobilized ligand (Fig. 2.8, left). The final part of the binding curve, while having a high surface load (such a low remaining ligand concentration) reveals a bi-exponential behavior, the receptors bind to less accessible regions too, the total association can be described by the sum of the association rate constant ka and a further association rate constant ka, bi-exp [106], such multiple association rate constants superimpose [100]. Only the middle part allows to fit rather properly an exponential function to the measured values according to the theory of the pseudo first order kinetic equation [107–111], appropriate for 1:1 reactions only:

−푘표푏푠∗푡 퐼푟(푡) = 퐵 − 퐴 ∗ 푒 (2.28) where

Ir (t) … mean intensity change at time t

B-A … y shift (Ir (0) = B-A)

B … Ir (t→∞)

kobs … observable rate constant

41 2.5 Determination of dissociation constant KD

Graphically the middle part of the function graph is depicted by plotting the signal derivative over the signal and selecting the linear region (Fig. 2.8, right), mathematically by

푑퐼푟(푡) 푑푡 = const. (2.29) 퐼푟(푡)

2.5.2 Determination of rate constants

Method 1 (rate constants with at least 2 kobs values)

From kobs both the association rate constant ka and the dissociation rate constant kd can be determined if at least binding curves of two different receptor concentrations c exist. Fig. 2.9

(left) shows kobs values referring to 3 concentrations, Fig. 2.9 (right) depicts the method of binding rates determination: the fitting line intersects the ordinate at the value kd, the gradient according to equation (2.26) is ka.

association dissociation

]

a.u.

]

[

obs

k intensityI

time concentration Figure 2.9: (left) association[s] and dissociation curves of 3 different receptor concentrations; (right) determination of ka (gradient) and kd (ordinate intersection) by plotting kobs values to corresponding receptor concentrations. [8]

The dissociation constant KD can be calculated as quotient of kd/ka.

푘 퐾퐷 = 푑 (2.30) 푘푎

Method 2 (rate constants with one kobs and one kdiss value):

As alternative to measuring the association curve of multiple receptor concentrations, performing one dissociation experiment allows to determine the dissociation rate constant kd [65].

푑훤(푡) = − 푘 ∗ 훤(푡) (2.31) 푑푡 푑 With the initial condition Γ (0) = 1 follows

42 2.5 Determination of dissociation constant KD

훤(푡) = e−푘푑푡 (2.32)

As the dissociation is not affected by the receptor concentration, the dissociation constant can be determined by fitting the measured curve to the function of equation 2.32. For the dissociation experiment it needs to be considered that back bonding would negatively affect the measurement and therefore should be suppressed – this could be achieved by adding a competitor to the buffer or at least by measuring the dissociation rate starting from a very high surface load [100].

Method 3 (rate constants with set of binding curves depicting association and dissociation):

TM Proprietary Biacore3000 Software allows depicting average kass and kdiss values from a bundle of curves with association and dissociation tracks resulting in average KD value.

2.5.3 Depiction by scatter blotting

Usually a set of experimental data is analysed. If both association and dissociation rate constants have been determined, scatter blotting allows visual depiction of the KD value cloud and finally calculation of the average KD value.

] 1

1 -

rate constant rate constant nM

[

ass k

association

-1 dissociation rate constant kdiss [s ]

Figure 2.10: blue dots cloud depicting KD values (dissociation rate constant kd / association rate constant ka) allowing to depict average KD value.

Non-first order bindings need to be analyzed in different manners, their identification requires systematic approaches [110].

The calculation of the standard errors of each parameter of the first order nonlinear fitting function is derived from the method Origin applies [112]. Both the theory and the implementation with Matlab is depicted in the Appendix A5.

43 2.6 Modelling of molecule interaction in flow cell

2.6 Modelling of molecule interaction in flow cell

Often theoretical models of molecular interactions, biochemical and chemical reactions are described on the single-molecule level, although our knowledge about the biochemical / chemical structure and dynamics primarily originates from the investigation of multiple- molecule systems. There are four experimental platforms to observe the movement and the behavior of single fluorescent molecules: wide-field epi-illumination, near-field optical scanning, laser scanning confocal and multiphoton microscopy. The platforms are combined with analytical methods such as fluorescence resonance energy transfer (FRET), fluorescence auto- or two-color cross-correlation spectroscopy (FCS), fluorescence polarizing anisotropy, fluorescence quenching and fluorescence lifetime measurements. Counting and characterization of freely diffusing single molecules in a single-phase like a solution or a membrane without hydrodynamic flow or immobilization can be achieved e.g. by fluorescence cross-correlation spectroscopy [113]. Labeling molecules can severly change their interactions expecially when the marker is bound close to the interaction center. We focus on the modeling of the molecule interactions of receptors in solution in a microfluidic flow cell and ligands immobilized on a substrate sealing that flow cell. The total process of the receptor-ligand interaction can be splitted into three vital parts: the transport of molecules by bulk flow to the diffusion layer, the diffusion of the molecules to the immobilized ligands and the binding kinetics of the receptors R and the immobilized ligands L (Fig. 2.11).

Figure 2.11: transport of molecules by bulk flow to diffusion layer, diffusion of molecules to immobilized ligands, bindings of receptors to ligands according to law of mass action. [49]

We distinguish two essential scenarios, first the kinetically controlled system with a constant receptor concentration, second the diffusion-limited condition with insufficient bulk flow or diffusion rate for keeping the surface receptor concentration at a constant level.

44 2.6 Modelling of molecule interaction in flow cell

2.6.1 Kinetically controlled condition

If mass transport of receptors from the bulk flow to the surface with immobilized ligand is much faster than the binding to the ligand, the concentration of receptors at the surface will be the same as that in the bulk. In this case, the measured binding will exclusively reflect binding kinetics according to the theory of chapter 2.4.2. Ensuring no receptor depletion neither by low bulk flow nor by insufficient diffusion rates can be achieved by a low concentration of ligand being immobilized on the substrate and a correspondingly high concentration of receptor molecules in solution. The reaction is not diffusion-limited but according to the bi-molecular mechanism, decay is mono-molecular [49, 114, 115].

2.6.2 Diffusion limited condition

If mass transport is rate limiting, obtained binding curves will depict the mass transport, which depends on the active analytes concentration, the diffusion coefficient of the analyte, the flow rate, and the dimensions of the flow cell. This scenario can be obtained by a high concentration of the ligand being immobilized on the surface, and a receptor concentration in the solution being low. Such, each receptor molecule diffusing towards the substrate interacts with an immobilized ligand. The diffusion constant of the receptor, the receptor concentration gradient from surface to bulk and the height of the diffusion layer define the receptor flux according to Fick’s first law

푐 푐 푗 = 퐷 푅푆 − 푅퐵 = 푘 ∗ (푐 − 푐 ) (2.33) 푅 푅 훿 퐷 푅푆 푅퐵 with

-2 -1 jR = flux of receptor molecules in mol m s

2 -1 DR = diffusion constant of receptor in m s

-3 cRB = concentration of receptor at bulk in mol m

-3 cRS = concentration of receptor nearby surface in mol m 훿 = diffusion layer thickness in m

kD = diffusion rate constant in m/s

and such mainly determine the bio-molecular interaction speed being proportional to the receptors concentration gradient.

For the modelling of the diffusion time we estimate the diffusion layer thickness by the layer approximation [105, 116].

45 2.6 Modelling of molecule interaction in flow cell

Figure 2.12: diffusion layer thickness approximation by Nernst - dotted straight lines indicate the construction method of the fictitious diffusion layer thickness perpendicular to the sensor surface. This thickness 훿 is called the effective or equivalent thickness of the diffusion layer. cRB is the bulk concentration, cRS the sensor layer concentration [117].

Table 2.2: diffusion layer data for proteins [105] reagent Diffusion constant Diffusion layer Diffusion rate

Dr (at 300 K) thickness constant

[m²/s] δ [µm] kD [µm/s]

proteins in water 10-11 - 10-9 ~ 10 1 - 100

Kinetics are controlled by the concentration gradient between sensor surface and bulk flow

[105]. Assuming flux of receptor molecules jR (eq. 2.33) being equal to the binding reaction 푑Γ(푡) speed (eq. 2.21) gives equation: 푑푡

푘퐷 ∗ (푐푅퐵 − 푐푅푆) = 푘푎 ∗ (Γ푚푎푥 − Γ(푡)) ∗ 푐푅푆 − 푘푑 ∗ Γ(푡) (2.34)

Solving this equation for the receptor concentration at surface 푐푅푆 leads to

푘퐷∗푐푅퐵+푘푑∗Γ(푡) 푐푅푆 = (2.35) 푘푎∗(Γ푚푎푥 − Γ(푡))+ 푘퐷

Inserting this expression of 푐푅푆 into equation 2.21 results in

푑Γ(푡) 푘퐷∗푐푅퐵+푘푑∗Γ(푡) = 푘푎 ∗ ∗ (Γ푚푎푥 − Γ(푡)) − 푘푑 ∗ Γ(푡) (2.36) 푑푡 푘푎∗(Γ푚푎푥 − Γ(푡))+ 푘퐷

46 2.6 Modelling of molecule interaction in flow cell

being transformed to

푑Γ(푡) 푘 ( 푘 ∗ 푐 ∗ (Γ푚푎푥 − Γ(푡))− 푘 ∗Γ(푡)) = 퐷 푎 푅퐵 푑 (2.37) 푑푡 푘푎∗(Γ푚푎푥 − Γ(푡))+ 푘퐷

There is no analytical solution of this general differential equation, numerical methods can be applied. Certain initial conditions allow analytical solutions, e.g. for [105]

Γ푚푎푥 − Γ(푡) ~ Γ푚푎푥 (2.38) are resulting in

푑Γ(푡) 푘 ( 푘 ∗ 푐 ∗Γ푚푎푥 − 푘 ∗Γ(푡)) = 퐷 푎 푅퐵 퐷 (2.39) 푑푡 푘푎∗Γ푚푎푥∗푐푅퐵

Integration leads to

푘 푘 − 퐷 ∗ 푎 ∗푡 푘 ∗Γ푚푎푥 + 푘 Γ(푡) = ΓGG(1 − e 푑 퐷 ) (2.40) The Onsager coefficients [105] determine the reaction speed on the surface

퐿푅 = 푘푎 ∗ (훤max − 훤(푡)) (2.41) and the mass transport to the surface

퐿퐷 = 푘퐷 (2.42)

As the diffusion parameters and the associaton constant is given by the system consisting of receptor and ligand, the ratio of the Onsager coefficients is only affected by the maximal

퐿퐷 load 훤max . Only for ≥ 10 the evaluation of the binding curve is not affected by the mass 퐿푅 transport [105].

In opposite, having a high ratio of ligand to receptor concentration, no steric competition and no other types of inflictions will occur. Therefore the affinity constants determined under diffusion-limited conditions can even be higher and closer to reality. When active analyte concentration is the only unknown parameter, this can be determined without prior knowledge of the binding kinetics [118]. This method requires enough ligand to be immobilized for the mass transport to be totally rate limiting, which is not always easily obtained. A simple analytical expression for the mass transport coefficient exists when mass transport is totally rate limiting [119, 120]. Conditions of partial mass transport limitation where this expression can be used, have been found by numeric computer simulations. The active concentration of the analyte does not to be known but, a relatively pure ligand should

47 2.7 Reflectometric interferometry be available and the molecular weight and diffusion coefficient of the analyte should be known [119].

Tailoring the number of recognition sites to the surface can allow one to determine the kinetic constants of the equilibrium at the surface or to use the linear slope of the diffusion process for direct measurement of the concentration of the analyte in the homogeneous phase. In any case, in a graph ‘signal versus logarithm of concentration’, a sigmoid calibration curve is formed to which a multiparameter curve can be fitted [43].

2.7 Reflectometric interferometry

This thesis focuses on the physical measuring principles of thin layer reflectometry and refractometry [103] for the analysis of protein interactions. Both use the dependency on the thickness of the layer and/or the refractive index, which influences the phase and/or amplitude of the electromagnetic radiation penetrating this layer or being reflected. Generally the properties of electromagnetic radiation can be characterised by amplitude, frequency (wavelength), phase, polarisation state, and time dependence [41].

2.7.1 Basics of optical detection systems

2.7.1.1 Interference and coherence Interference of different light waves demands a sufficiently coherent light source. The relation between coherence length and phase shift defines whether interference occurs or not. Coherence time and coherence length in the direction of propagation are directly related to the spectral width of the light source or the filters used in the setup [121]. For an LED the coherence length in air is between 3 and 4 μm. In a different medium it is reduced by the amount of the refractive index. For glass (n = 1.5) it is reduced to 2 - 2.6 μm.

2.7.1.2 Reflection and transmission

If light passes from one medium into another one, the fraction 푃푅 of the wave will be reflected, others (푃푇) transmitted (refracted) or (푃퐴) absorbed. Due to conservation of energy the incident power is the sum of the fractions power:

푃0 = 푃푅 + 푃푇 + 푃퐴 (2.43) These fractions are defined by the absorption coefficient k and the refractive indices n.

Assuming no absorption (k=0), an angle of the incident light of φ1 and refractive indices n1

48 2.7 Reflectometric interferometry

and n2 with n1 < n2, some fraction of the ray will be reflected with an angle of φ1, the other will be transmitted into the bulk material according to the law of Snellius [121]:

푛1 ∗ sin 훩1 = 푛2 ∗ sin 훩2 (2.44)

Figure 2.13: incident ray is split into refraction and transmission fraction The Fresnel coefficients [121] enable to calculate the fractions of the complex-valued amplitudes of the reflected (rp + rs) and the transmitted (tp + ts) wave, each for the components orthogonal (‘s’) to the incident plane and parallel (‘p’).

푛2∗푐표푠휙1−푛1∗푐표푠휙2 푟푝,12 = (2.45) 푛1∗푐표푠휙2+푛2∗푐표푠휙1

푛1∗푐표푠휙1−푛2∗푐표푠휙2 푟푠,12 = (2.46) 푛1∗푐표푠휙1+푛2∗푐표푠휙2

2∗푛1∗푐표푠휙1 푡푝,12 = (2.47) 푛1∗푐표푠휙2+푛2∗푐표푠휙1

2∗푛1∗푐표푠휙1 푡푠,12 = (2.48) 푛1∗푐표푠휙1+푛2∗푐표푠휙2

When light reflects off a material denser (with higher refractive index) than the external medium, it undergoes a polarity inversion. In contrast, a less dense, lower refractive index material will reflect light in phase. This is an important principle in the field of thin-film optics.

2.7.1.3 Wave-Transfer-Matrix Method When light makes multiple reflections between two or more parallel surfaces, the multiple beams of light generally interfere with one another, resulting in net transmission and reflection amplitudes that depend on the light's wavelength. The interference, however, is seen only when the surfaces are at distances comparable to or smaller than the light's coherence length, which for ordinary white light is few micrometers; it can be much larger for light from a laser. A quantitative analysis of these effects is based on the Fresnel equations, but with additional calculations to account for interference. The wave-transfer- matrix method [121], or the recursive Rouard method [122] can be used to solve multiple- surface problems.

49 2.7 Reflectometric interferometry

Light as an electromagnetic wave can be described by the complex function U and is dependent upon the position r = (x, y, z) and the time t. It is a function that satisfies the partial differential equation called the wave equation [121] which, in addition, depends on the speed c of the wave. For light, c as the speed of light depends on the speed of light in vacuum c0 and the refractive index n of the medium with c = c0/n.

1 휕2푈 ∇2푈 − ∗ = 0 (2.49) 푐2 휕푡2

The intensity of the wave is given by the square of the absolute value of the complex U: I = |푈| 2 (2.50) By splitting U(r,t) into U(r)*U(t) and replacing U(t) with ei2πft as a solution for the time dependent part of the wave equation, we get the time independent Helmholtz equation 1 ∇2푈 − 푘 = 0 (2.51) 푐2 with 2휋푓 휔 k = = (2.52) 푐 푐

One solution for the Helmholtz equation is the plane wave, which travels into the direction k

= (kx, ky, kz) with A as a constant for the intensity. U(r) = A ∗ e−푖푘푟 (2.53)

In linear optics each wave can be regarded as a superposition of plane waves. In the rather easy case of a 1-dimensional system, where a plane wave travels in the x direction only and is transmitted and reflected at a boundary to a layer with a different refractive index, the wave in each layer can be described by the superposition of an ingoing and an outgoing wave.

Figure 2.14: superposition of ingoing and outgoing waves.

50 2.7 Reflectometric interferometry

With the condition that at each boundary the waves are continous and continuously differentiable, it is possible to describe the transfer of a wave through a boundary with the Wave-Transfer-Matrix (transmission matrix) M 푈 푈 ( 2+) = 푀 ( 1+) (2.54) 푈2− 푈1−

To receive a value for the reflectivity of this system, U1− as the outgoing wave on the left side should be expressed in terms of the incoming wave U1+. U2− is 0 in our case, due to the fact that no light is coupled into the system from the right side. This two-layer system can be expanded to a multi-layer system. The wave-transfer-matrices M of each boundary can be multiplied in order to receive an equation that describes the whole setup. The sequence of the matrices is important, as matrix multiplication is not commutative.

푈푛+ 푈1+ ( ) = (푀푛 ∗ 푀푛−1 ∗ 푀푛−2 ∗ … 푀1) ( ) (2.55) 푈푛− 푈1−

Each matrix Mx is a 2 x 2 matrix. 퐴 B 푀 = ( ) (2.56) 퐶 퐷 The elements of the matrix M can be derived from the scattering matrix S which consist of physically related elements, the fresnel coefficients for reflection r and transmission t. The indices describe the direction of incidence e.g. 12 refers to incident light from layer 1 to layer 2 [121]. 푡 푟 푆 = ( 12 21) (2.57) 푟12 푡21 However, as a multi-layer system can not be depicted as the product of all S matrices, we convert these S matrices into M matrices by the formula [121]

퐴 B 1 푡12푡21 − 푟12푟21 푟21 푀 = ( ) = ( ) (2.58) 퐶 퐷 푡21 −푟12 1 For a multi-layer system these M matrices can now be multiplied, the resulting M matrix can be converted back to a S matrix to get the fresnel coefficients for the total system by

푡12 푟21 1 AD − BC B 푆 = ( ) = ( ) (2.59) 푟12 푡21 D −C 1

2.7.2 Sensitivity of optical detection systems

Both the sensitivity and the reproducibility of detection systems is biased to signal noise and drift. Noise in optical systems is either of systematical or statistical nature. Systematical

51 2.7 Reflectometric interferometry noise is caused by system inherent sources, e.g. 50 Hz noise evolving from the use of alternating current, oszillations at eigenfrequencies. It can be eliminated by filters or by referencing. Signal drift is generally of systematical type, e.g. temperature drift, LED intensity drift, bulk effects [100]. Statistical noise is inherent to the optical detection system and caused by the particle nature of light. This type of noise is called Poisson or Shot noise.

2.7.2.1 Statistical noise Light being viewed as a wave with a specific amplitude and frequency does not explain all characteristics of light, e.g. the photoelectric effect can only be understood with the assumption that light is transported by particles with quantized amounts of energy, the photons. For a finite numbers of photons the amount of reflected photons needs to be described with statistics. Independent photons follow the Poisson distribution. The standard deviation 휎 of this function can be calculated and expressed by the number of photons n [121]:

휎 = √푛 (2.60)

Basically the signal-to-noise ratio SNR is expressed as

푛 푆푁푅 = (2.61) 휎 and can be simplified to

푆푁푅 = √푛 (2.62)

By spatial averaging over different pixel of one frame or by temporal averaging of one pixel over different frames the noise can be reduced. Averaging over j values results in a mean photon number 푛̅ of

1 ∑푗 푛 = 푛̅ (2.63) 푗 푖=0 푖

The noise of that mean number of photons is

푖∗푛̅ √푛̅ 휎 = √ = (2.64) 푖 √푖

Hence shot noise reduces by the factor √푖 by averaging i times.

2.7.2.2 Systematical noise Further noise is generated by the camera and its read-out electronics. It is of the type systematical noise, not related to the amount of photons. Cooling CCD cameras, either by air or water decreases the temperature drift. Dark current can be subtracted using a dark frame as reference picture. Combining multiple images is averaging out random pixel variations. Flat fielding smoothes out pixel non-uniformity caused by variations in the pixel-

52 2.7 Reflectometric interferometry

to-pixel sensitivity of the detector and/or by distortions in the optical path. Any further noise can be classified and interpreted by Fourier Transforms (FFTs).

There are many potential sources of noise in a CCD camera. The read noise inherent in a CCD tends to be random, uncorrelated noise. All cameras add additional noise when reading the image from the CCD and converting it into a digital image. The camera designer’s goal is to add as little additional noise as technically possible, and make sure that no noise is introduced that can’t be reliably reduced using standard image calibration techniques. Combining multiple frames is useful for smoothing out random noise.

Fig. 2.15 compares the highest possible SNR caused by shot noise with the combined noise sources of the pco 1600 camera measured with monochrome light at 525 nm peak wavelength according to EMVA standard 1288 [123]:

Figure 2.15: maximum theoretically attainable SNR, caused only by shot noise, compared to real SNR as combination of all noise soures measured with monochrome light of 525 nm with the CCD camera pco 1600. [123]

2.7.2.3 Resolution For optical systems, consisting of a lens and a digital camera, two factors can limit the resolution of the system: The lens itself and the resolution of the imaging sensor. The resolution of the sensor is defined by the amount of pixels. The resolution of the whole system can be described with the unit linepairs/image-height. It describes, how many linepairs, consisting of a black and a white horizontal line can be imaged with the system. This characterization does not say anything about the smallest object that can be made visible. The magnification and the distance of the object need to be known for that. For telecentric lens systems the distance of the object does not play a role anymore. The magnification of the lens is the only parameter that describes the relation between the size of the object and the size of the image. With this parameter, the size of one pixel can be used to calculate the smallest feature size. Further the Nyquist theorem has to be applied. It says that the sampling frequency has to be twice as big as the highest spatial frequency. This means, that the smallest measurable object 푠obj_min needs to have the size of 2 pixels 푠pixel divided by the magnification 푚lens of the lens system:

53 2.7 Reflectometric interferometry

푠obj_min = 2 ∗ 푠pixel / 푚lens (2.65)

The quality of the lens system additionally needs to be taken into account.

2.7.2.4 Total system noise quantification For the noise evaluation of the total system we take a set of data at the final stage of the baselining or data of the final washing step after the experiment; we consider 2 critieria:

1) Baseline drift: slope of linear fitting to data set in ‘‰/min’. 2) Baseline noise: residuals squared sum of data set in ‘RSS/min’.

Both criteria are biased to setup adjustments, these are especially the camera exposure time defining the well load, such the shot noise; further the number of frames being averaged: our standard exposure time is 20 ms, the standard averaging is 64 frames.

2.7.3 RIfS detection technology

Reflectometry has been introduced many decades ago as ellipsometry using polarized light. Reflectometric interference spectroscopy measures the changes of amplitude and phase of at least two polarized beams of radiation being reflected at two or more thin layer interfaces, these could be physical layers as well as (bio-)chemical layers.

A B relative intensity relative

I I1 I2 I3 400 500 600 700 800 E

wavelength [nm]

C D n*d thickness changeof optical

time relative intensity relative

I I1 I2 I3 400 500 600 700 800 wavelength [nm] Figure 2.16: (A) schematic of the RIfS detection principle with light beams being reflected at different physical layers. (B) The superimposition of these partially reflected light beams causes wavelength dependent interference. (C) Molecule binding at the outer layer changes the optical thickness on the sensor surface, (D) causing a shift of the interference spectrum, (E) finally a binding curve as the result of the steadily changing optical thickness. [56][47]

If the coherence length of the partially reflected beams of radiation is longer than the optical path through the layers, they interfere and form an interference pattern depending on the

54 2.7 Reflectometric interferometry

optical thickness of the layers, the incident angle and the refractive index of the surrounding medium [124]. In case of perpendicular incidence, non-absorbing layers, and low reflectances, the reflectance IR is given by

1/2 IR = I1 + I2 + I3 + 2 (I2I3) cos(4πnd / λ) (2.66) where I1, I2 and I3 denote the Fresnel reflectance at three interfaces, d is the physical thickness of the film, n its refractive index, and λ the wavelength of incident light. A typical interference pattern showing the modulation of reflectance with cos(1/λ) is given in Fig. 2.16, B. The optical thickness nd can be determined from the position of an extremum with a given order value m by nd = mλ / 2 (2.67) To evaluate the binding signal, the locus of the extremum is tracked over time; thus, the change of the interference spectrum results in a time-resolved binding curve representing the binding of the analyte molecule to the sensor surface (Fig. 2.16, E) [56].

The binding of an analyte molecule or particle to a sensor surface modifies the bio-chemical layer by permeation or/and by adsorption. In the case of adsorption, any difference in the refractive index of the adsorbed layer will introduce the additional effect of Fresnel reflectivity, influencing the intensity of the reflected partial beam. Penetration of the layer by an analyte can cause a pure swelling resulting in a change of the physical path length; in addition, due to its density dependency the refractive index can change too. Spectral interferometry allows these effects to be discriminated to a certain extent [124]. Pure adsorption will vary the refractive index at the interface to the coating layer. Thus, the reflectivity at this interface changes, diminishing the amplitude in the interference spectrum. Any change in the optical path length (n*d) influences the distance between extrema (Fig. 2.17).

reference molecule adsorption

change of optical thickness relative intensity [a.u.] intensity relative wavelength λ [nm]

Figure 2.17: molecule adsorption varies refractive index causing diminishing amplitude, varied optical thickness changes distance of extrema. [124]

55 2.7 Reflectometric interferometry

Quantification of molecule adsorption and / or penetration detected by reflectometric interference spectroscopy depends on the precision of the extrema positions determination on the wavelength scale. The better the wavelength of the extremum can be determined, the more precise the optical path length can be calculated. Dispersion, noisy data and low integration times complicate and limit the exact determination of the optical path length. Three methods could be applied to determine rapidly the extrema wavelengths: differentiation of the interference spectrum, Fourier analysis, and overall curve fit. Curve fitting gives best results, Fourier analysis poor ones. Differentiation in combination with polynomial fitting is both fast and supplies sufficient data resolution. Due to the significant calculation times, the overall evaluation by curve-fitting technique cannot be used for on- line monitoring of the molecular immobilization processes [124].

56 2.8 Biochemistry

2.8 Biochemistry

2.8.1 Expression-ready DNA

DNA is encoding the protein according to the DNA sequence. This DNA sequence is first transcriped into RNA, then the RNA sequence is translated into proteins. This process is called the central biochemical dogma upon which all life is based. The DNA sequence is a type of work procedure for the protein expression. The RNA polymerase binds to the promotor and starts the transcription into the RNA sequence; this procedure ends at the transcription terminator. Then the ribosomes bind to the RBS (ribosome binding site) and proceed with the translation of the RNA codons (Fig. 2.18) according to the RNA codon table. The polyA sequence indicates the number of cycles the ribosomes are expressing the RNA sequence and such the number of proteins it generates from one RNA sequence. The encoding part of the DNA sequences carries information for the protein and its 5‘- and 3‘- tags, e.g. c-myc tag.

(A)

(B)

(C)

Figure 2.18: (A) design of expression-ready DNA sequence: the promotor enables the RNA polymerase to bind to the DNA sequence at the 5’ end. It transcribes the DNA into RNA until it reaches the transcription terminator at the 3’ end. (B) the RNA sequence is translated by the ribosome into a protein starting at the ribosome binding site. (C) The start codon identifies the begin of the protein translation, each triple of base pairs is translated into an amino acid according to the RNA codon table. Depending on the sequence of amino acids the protein finally folds itself.(drawing by G. Roth)

Due to evolutionary optimisaton processes the RNA polymerase reaches a precision of 100 ppm, that is one error within 10.000 transcriptions; transcription speed is 50 RNA bases per second. Mammals ribosomes consist of a conglomerate of 82 single proteins and 4 RNA strands (in total 6872 RNA strands), such they are belonging to the most complex macromolecules. Witin each cell 105 to 107 ribosomes work in parallel for the continous protein production; 300 amino acids are transcriped within ~ 20 seconds, within another 20- 40 seconds they are translated into proteins.

57 2.8 Biochemistry

Expression-ready DNA (erDNA) sequence design – recipee:

1) T7 promotor a. Generally starts with TATA sequence – TAATACGACTCACTATAGGG b. Binds the RNA polymerase complex and initiates the transscription towards the 3’end of the DNA sequence by the RNA polymerase II. 2) Ribosome binding site a. Often starts with AGGAGGT, assembles protein expressing ribosome from subunits onto corresponding RNA sequence. 3) Start codon a. ATG as DNA sequence indicates the begin of the protein, it is transcribed into AUG as RNA sequence. The ribosome starts the translation of the RNA sequence into a sequence of amino acids after AUG. 4) Protein sequence a. Protein encoding DNA describing the amino acid sequence on the DNA level b. The open reading frame (ORF) encompasses the part of the DNA sequence (including Start-/Stopcodon) which creates the protein with linkers and tags. 5) Capture Tag a. Protein encoding sequence which can be detected by an antibody on the protein level, e.g. Strep tag, His tag, myc tag, GST tag. Often tags are kept in distance of the amino acid sequence of the main protein by linkers, either before or after the protein sequence. 6) Linker a. DNA sequence which separates main protein and tag. 7) Stop codon a. TAA stops the protein synthesis by ribosome; double TAA increases efficiency. 8) polyA a. number of A’s determines the number of proteins created by one RNA sequence. 9) Terminator a. DNA sequence rich of GC combinations which causes the termination of the transcription.

2.8.2 DNA amplification and analysis

Generation of erDNA templates was performed as published previously by assembly PCR [32]. For the work presented within this thesis, four different erDNA templates were prepared. Each construct encoded a double hexahistidine tag (2His6) for immobilization on Ni-NTA surfaces, three of the constructs also encoded a N-terminal triple c-myc tag (3myc)

58 2.8 Biochemistry

for antibody detection. Constructs were GFP-2His6, 3myc-Transcription factor DNA binding domain-2His6 (3myc-TF-DBD-2His6), 3myc-p53-2His6, 3myc- fragment of Notch1 – 2His6. erDNA for immobilisation was labeled by incorporation of Cy3- or Cy5-modified primers to allow detection on the erDNA arrays. Except the protein sequence itself any linker or tags were added by the polymerase chain reaction (PCR) considered as standard procedure [125]. Often protein sequences are designed with linkers and capture tags at the C terminal region [126].

2.8.3 Cell-free protein expression

The production of small quantities of proteins can be performed quickly and economically by cell-free systems. This leads to their adaptability to high-throughput experiments, in which high numbers of experiments are desired. In cell-free expression systems, no cell viability concerns are necessary. Therefore, toxic proteins, which would destroy the cell metabolism or simply the expression apparatus, can be produced. Before cell-free expression techniques were developed, biochemical and structural characterization of membrane proteins for example have been in its infancy. Nowadays, sufficient amounts of functional membrane proteins even for crystallography and biochemical analysis are producible. In cell-free expression systems additives like detergents, metal ions, cofactors or binding partners can simply be added to the expression reaction. This can only be accomplished by owning the “open” system characteristic. Additionally, incorporating isotopic labels and non-natural amino acids into the peptide chain of the produced proteins is easy. Even the simultaneous expression of more than one protein in one cell-free reaction is possible. A disadvantage of cell-free expressions is the relatively low protein yield, depending on the expression system. Additionally, cell-free expression can be expensive, depending on the used system and if it is commercially purchased. One has to keep in mind, that the in vitro reaction has no sustained metabolism to convert cheap energy sources like sugars. Recently, glucose or polymeric carbohydrates were successfully used in cell-free expressions as an energy supply. Another disadvantage, which is solving itself by the years, is the low degree of characterization of cell-free expression systems and the rather low degree of usage experience in laboratories compared to organisms like E.coli. Cell-free systems are purified from cell cultures and basically consist of two solutions: one provides the enzymes for the synthesis, the other one the energy with ATP’s and the amino acids as building blocks. Mixing these solutions and adding DNA leads to the protein expression. Depending on the origin (cell type) of the cell- free system glycosylation and phosphorylation occurs. Protein synthesis preferably happens at 30 – 37 °C. All cell-free systems consist of elementary items:

59 2.8 Biochemistry

 RNA polymerase / effectors o bind to DNA and create mRNA.  Ribonucleid acid triphosphate (rNTPs) o These monomeres are consumed by the RNA polymerase, while pyrophosphate is separated, RNA is created.  Ribosomes o These macromoleculare complexes create proteins. They require mRNA as template, consume beloaded tRNA and produce pyrophosphate, while ribosomes synthesize the protein.  Beloaded transferRNA (tRNA) o Each beloaded tRNA carries an active amino acid, which is build specifically into the protein sequence. Empty tRNA’s are reloaded with new amino acids.  Loading system o Complex enzyme system which recognizes empty tRNA and reloads it with the correct amino acids while consuming ATP’s (energy). Often protein synthesis expires due to energy limitations.  Adenosin-triphosphat (ATP) and further energy providers o Energy suppliers for enzymes, often by separation phosphate groups.  Release factors o Release the ribosome at end of mRNA after having created the proteins.  Ressources o Amino acids, ions like calcium or magnesium, as well as further carriers for the activation and functionality of the cell-free expression system.

60 3.1 Flow cell design and fabrication

3 Materials and Methods

3.1 Flow cell design and fabrication

We want to design a flow cell suitable for the cell-free expression of a protein microarray from a DNA microarray followed by microfluidic protein interaction analysis based upon microfluidic flow injection and the label-free iRIf detection method. Major parts of the specification for the flow cells are low reagent volumes in the range of some 10 µl and a design considering at least the use of 1 microscope slide in standard format as protein microarray substrate, such that the result of the microfluidic assay could be analysed by established end-point measurement methods too. The materials in touch with the reagents should be non-inhibitive and coatable by surface chemistry for the immobilization of DNA, oligonucleotides and proteins. All flow cells and moulding tools were designed with the CAD tool SolidWorks.

3.1.1 PMMA flow cell (concept 1)

3.1.1.1 Design and fabrication For a very first analysis of the microfluidic cell-free expression of DNA microarrays into protein microarrays a flow cell consisting of an upper epoxy coated microscope slide, a lower Ni-NTA coated microscope slide and a spacer encompassing a microfluidic gap was designed (Fig. 3.1) and fabricated. The spacer is made of laser-cut polyester adhesive foils of 100 µm thickness. The adhesive foils were sticked onto epoxy coated microscope glass or polycarbonate slides and cut by laser PLS3.60 at 40 W. The excess foil was removed such that only the frame of the flow cell with a width of 2 mm was remaining. Sealing this first slide with the adherent spacer by a second microscope slide with Ni-NTA coating realized a first microfluidic flow cell. Both slides were held together by an upper and a lower plate, both laser-cut of a 4 mm PMMA plate and screwed together by 6 screws of type DIN 963 A2 M3 x 10. Inlet and outlet was realized either by holes through the upper PMMA plate and the epoxy coated DNA slide (Fig. 3.2) or laterally through a gap in the polyester spacer (Fig. 3.3).

61 3.1 Flow cell design and fabrication

epoxy coated DNA slide

polyester spacer

Ni-NTA coated protein capture surface Figure 3.1: schematic of flow cell with polyester spacer between two glass slides.

(A) upper plate (B) (C)

spacer

2 cm lower plate Figure 3.2: (A) design of PMMA flow cell subtype 1 with laser-cut self-adhesive polyester foil of ~100 µm as spacer. (B) design of assembly. (C) Realized system, priming through upper PMMA plate via holes in top slide.

(A) upper plate (B) lateral 2 cm support spacer

(C)

protein capture surface DNA matrix

lower plate

Figure 3.3: (A) design of PMMA flow cell subtype 2 for lateral priming. Spacer could be realized by laser-cutting of polyester adhesive or by spin-coating of PDMS onto a silicon wafer, followed by laser-cutting and vacuum transfer onto the coated glass slide. (B) design of assembly. (C) Realized system, lateral priming via opening in spacer (red arrow).

(A) (B) (C)

2 cm 2 cm 2 cm

Figure 3.4: design study for enhancing bubble free priming; (A) double trapezoidal layout with support lane; (B) standard double trapezoidal layout; (C) opening channel. All design have lateral openings for inlet and outlet for PMMA flow cell subtype 2.

Priming via holes from top is convenient however, only suitable while using a laser-cut polycarbonate slide as DNA matrix as drilling through glass is cumbersome and expensive,

62 3.1 Flow cell design and fabrication

such controversary to the idea of having low-cost consumables. Lateral priming is tricky but worked rather fine. Bubble-free priming remained a challenge for all types of designs.

3.1.2 PDMS flow cell (concept 2)

3.1.2.1 PDMS (Polydimethylsiloxane) Amorphous polymers as PDMS (Polydimethylsiloxane) have been widely used in the fabrication process of microsystems including microfluidic systems. PDMS consist of repeating monomer [SiO(CH3)2] units [127]. It is bio-compatible for most biological assays. PDMS with a range of impressive physical and chemical features is a favourite material for the low-cost fabrication of micro-structured flow cells. Before polymerization it is a viscous liquid conforming to any moulds surface rather conveniently, after polymerization at 70 °C for 2 hours it can be easily peeled off the mould keeping even tiny features up to microns scale without any damage. It is very suitable for the replication of narrow microfluidic channels [128].

(a) (b)

Figure 3.5: (a) PDMS is mould onto master and cured at 70 °C. (b) Even tiny micro- structures (< 50 µm) can be peeled off easily, any sticking could be avoided by pre- silanisation of the master before moulding.

3.1.2.2 Concept and pre-analysis We choose a flow cell design allowing spacer heights in versatile dimensions and high precision considering eveness. For priming and flow control rather any 3-dimensional structure for the flow cell ground could be designed. Above all the flow cell needs to be coatable for DNA immobilization and the outer size should be of standard microscope format, such that the cavity can be sealed with a standard microscope slide (Fig. 3.6). We consider PDMS as appropriate material for the fabrication of structured flow cell inlays.

coating for DNA array

PDMS slide with incorporated spacer microscope glass slide with protein capture surface Figure 3.6: schematic of flow cell with PDMS inlay enclosing flow cell cavity being sealed by microscope glass slide. PDMS inlay can be 3D-structured for realizing an appropriate flow chamber geometry and for adding flow control structures to bottom of flow chamber.

63 3.1 Flow cell design and fabrication

For the pre-analysis of some geometrical structures and dimensions we perform some basic fluidic simulations with SolidWorks. Appropriate to the standard microscope slide dimensions a main cavity with a length of 20 mm and a width of 15 mm is designed; low flow cell heights of 20 - 100 µm shall reduce the reagent volume consumption. Flow uniformity and flow cell stability for a high reproducability can be supported by certain 3-dimensional structures, e.g. double trapezoidal pillars have been pre-analyzed by simulations with SolidWorks:

Figure 3.7: double trapezoidal layout leads to rather homogenous flow profile in the centre area of the flow cell. Streamlined pillars as support structures would affect the flow homogeneity especially if used in short distances.

3.1.2.3 Protoype fabrication methods A first mould master was micro-milled as negative of the later PDMS structure out of a polycarbonate slide of 76 x 26 x 1 mm³ (Fig. 3.8, A) with simply a sealing edge of width 1 mm and heights between 60 µm and 100 µm. A PTFE spacer of 1 mm thickness framing the outer dimensions of the later PDMS inlay was superimposed, finally sandwiched together with a standard microscope glass slide and firmly clamped into a moulding station compatible with luer-lock connectors to standard syringes (Fig. 3.8, B). Either transparent Sylgaard 184 PDMS or red RT 607 PDMS was prepared in accordance to the manufacturers’ standard protocol, pressed into the mould via inlet hole in the mould master with a 5 ml syringe and cured at 70 °C for 1 h. The priming of the PDMS could be visually monitored by a rear window (Fig. 3.8, C), finally the moulded PDMS inlay was released (Fig. 3.8, D). To immobilize erDNA this PDMS surface was activated by PDITC chemistry.

64 3.1 Flow cell design and fabrication

Figure 3.8: PDMS inlay fabrication method: (A) for first prototype inlays a master structure is milled into a 1 mm PMMA slide of standard microscope slide format. (B) moulding station with luer-lock connector for priming PDMS by syringe. (C) rear window to visually monitor the priming of PDMS. (D) PDMS inlay after curing at 70 °C for 1 h.

First microfluidic applications have been performed by simply adhering the PDMS flow cell onto a microcope glass slide (Fig. 3.9).

Figure 3.9: PDMS inlay activated with APTES/PDITC for DNA immobilization rolled onto a microscope glass slide as simplest microfluidic system. Spotted droplet is enclosed by PDMS cavity.

A simple clamping mechanism with laser-cut polypropylen (PP) clips allows first to assemble the PDMS inlay and the glass slide, then priming via holes.

Figure 3.10: 6 laser-cut polypropylene clips slightly press the PDMS inlay and the glass slide together and allow priming the created flow cell via inlet.

Assembled flow cell solutions depicted above did not integrate any temperature control system being necessary for the cell-free expression of proteins. A modification of the

65 3.1 Flow cell design and fabrication moulding tool (Fig. 3.8) integrates the flow cell with stable inlet and outlet connectors (Omni-Lok Inverskonus) and a Peltier based temperature control system.

Figure 3.11: design of clamping system for flow cell consisting of PDMS inlay and glass slide with adjustable pressure; integrated Peltier based temperature control unit.

The backbone of the PDMS inlay is now a copper plate connecting the flow cell with a Peltier element (250 V, 8 A). Onto the open copper plate (Fig. 3.11, bottom left) first a Peltier element is sticked via heat transfer paste, next a passive aluminium cooling block with fan. Such this system provides the capability of controlling the protein expression process by heating to 30 - 37 °C and by cooling to 8 °C for expression stop. The bottom side offers a window (Fig. 3.12 (B)) for the in-process monitoring of the protein expression and/or the protein interaction analysis.

(A) (A) (B)

Figure 3.12: (A) PDMS flow cell clamping unit with copper backbone for Peltier temperature control, with inlet / outlet connectors for tubes of 1/16”. (B) bottom side monitoring window.

This system now allows both manual and automated operation. Manual operation starts with manually priming the flow cell by pipette via inlet followed by putting the whole system

66 3.1 Flow cell design and fabrication

to an oven for protein expression at 37 °C. For automated operation inlet and outlet are connected to any pump for priming, then tempering by the integrated Peltier element and finally washing by buffer is performed.

3.1.2.4 PDMS inlay designs [6] The flow chamber (Fig. 3.13, centre grey area) is surrounded by an active sealing edge (red lane) followed by a passive sealing edge (grey lane, groove). The outer parts (red areas) support the active sealing edge against deformation at higher pressure of the glass slide towards the PDMS for proper sealing. Such the basic design of the PDMS inlay consists of a 2 level structure, which can be rather easily fabricated in negative shape as a structure on a Si wafer being used as mould master. The high degree of flatness of the wafer and the very low roughness of the wafer surface is used for getting a very flat and smooth outer boundary of the PDMS inlay ensuring high adhesion to the glass slide. Inlet / outlet holes punched through the PDMS at the very left and right side of the flow chamber avoid fluidic dead ends and allow connecting the flow chamber from the rear side. The rear side of the PDMS inlay has to be very flat too, for tightening the inlet / outlet connectors.

active sealing outer boundary edge

outlet inlet passive sealing edge

Figure 3.13: PDMS inlay – basic design in microcope slide format (76 mm x 26 mm). Active and passive sealing edges surround the flow chamber with outlet / inlet holes at the left and right end of the incubation chamber. The plane of the outer boundary at the same level as the active sealing edge ensures adhesion of the sealing glass slide and avoids deformations of the elastic PDMS inlay to the decrease of the flow chamber height even at higher sealing pressures initiated from external clamping elements.

PDMS inlay with pillars:

The basic PDMS inlay design could have the major deficit of allowing the inner parts of the flow chamber to collapse. Therefore we introduce support structures which will maintain the flatness and the distance between the PDMS and glass surface. Pillars are designed in streamline shape according to the simulation results (Fig. 3.7). The length of each pillar structure is 0.75 mm, the width is 0.3 mm.

67 3.1 Flow cell design and fabrication

Figure 3.14: design with rhombic pillars. Length of flow chamber reduced to further decrease risk of collapse. Outlet side of different geometry is suggested for analysis of flow behaviour.

Design with channels

Channels both support the inlay against collapse and should enhance the bubble-free priming. Channel walls have a width of 0.2 mm, a length of 30 (28) mm.

Figure 3.15: design with channels of 28 (short) and 30 (long) mm length.

Design with geometrical valves:

At the end of each channel geometrical valves were placed. Geometrical valves ensure the complete priming of each channel before the valves are bursting; such the enclosing of an air bubble in the major flow cell region is avoided.

Figure 3.16: design with geometrical valves at channel ends ensure bubble-free priming in the centre flow cell region.

The pressure drop at the valve is calculated using equation 3.1 derived from [60] but considering 2 different materials for bottom and top sealing:

cos 휃 cos 휃 cos 휃 ∆푝 = 휎 ∗ (2 ∗ 푝 + 푃퐷푀푆 + 𝑔푙푎푠푠) (3.1) 푤 ℎ ℎ Following data according to the physical system is taken for the calculation:

68 3.1 Flow cell design and fabrication

ℎ = 30 µm

휃푃퐷푀푆 = 105°

휃푔푙푎푠푠 = 60°

휃푝 = 180°

휎 = 72 mN/m

The value of 휃푃 is 180° due to pinning at side walls of PDMS with effective contact angle of min (휃푃퐷푀푆 + 훼, 180°), with 훼 = 90°. For the first design we chose a valve width of 0.08 mm resulting in a pressure demand of 1200 Pa according to Fig. 3.17:

Figure 3.17: design of valve width according to pressure drop – 0.08 mm is selected for a first design, that results in a pressure drop to be overcome of 1200 Pa.

Design with phase guides:

Initial bubble-free priming is supported by phase guides, a type of horizontal geometrical valves separating the total chamber into compartments. These compartments are expected to be filled one after the other, such avoiding trapped bubbles. Pillars on each phase guide serve as support structures according to streamlined pillars introduced before. The total height of the flow chamber is designed to 40 µm, each phase guide to half of it, 20 µm. The width of each phase guide is 0.2 mm. Pillars have square shape with the edge towards the flow. Each side of pillars has length of 0.11 mm and height of 20 um. Such pillars are smaller than the width of the phase guide and do not affect the phase guide behaviour. Pillars are placed at the front end of each phase guide, such the liquid is merging before moving into the next compartment. Width of each chamber is designed to 1.8 mm, distance of pillars to 4 mm.

69 3.1 Flow cell design and fabrication

Figure 3.18: design with phase guides in distance of 2 mm with height of 20 µm at a total flow cell height of 40 µm. Pillars on top of phase guides avoid the collapse of the flow chamber.

To prevent bubbles in the corners of the phase guides due to the hydrophobic surface of the PDMS (convex shape of meniscus expected) each phase guide was rounded at the edges. Fig. 3.19 depicts a design with a compartment width of 2.8 mm.

Figure 3.19: design with curved edge phase guides enhances bubble-free priming despite hydrophobic surfaces.

Phase guides work as horizontal geometrical valves by the principle of pinning. The sudden change in geometry creates the pinning barrier. Liquid needs to overcome the increase in pressure to advance the contact line. From equation 2.3 with 푤 ≫ ℎ, ‘w’ is neglected:

cos 휃 cos 휃 ∆푝 = 휎 ∗ ( 푃퐷푀푆 + 𝑔푙푎푠푠) (3.2) ℎ ℎ

The value of 휃푃퐷푀푆 is changed to 180° as min (휃푃퐷푀푆 + 훼, 180°), with 훼 = 90°.

Figure 3.20: pinning pressure at phase guide in relation to remaining gap (width w). A phase guide gap width of 20 µm results in a pinning pressure of -1800 Pa.

70 3.1 Flow cell design and fabrication

Design of plane flow cell:

Despite versatile structures for enhanced stability, support and prevention of bubbles an inlay design of equal dimensions with a plane flow chamber is created. First analyses revealed that proper adhesion of the flat PDMS rear side to a very flat solid backbone could be sufficiently stable, bubble-free priming of the flow cell of height ~ 30 µm with water or water-like buffers is likely.

Figure 3.21: design of plane flow chamber in dimensions as structured ones.

Design of flow cell with partly open micro-cavities:

By this open cavity design the basic priming features should be kept, but lateral diffusion especially for the application of in situ protein expression should be drastically limited. Each cavity could be used for the expression of separate protein spots; as alternative tiny cavities would just be put like a net onto a DNA microarray; adjustment wouldn’t be necessary. For initial experiment 2 designs are considered, the hexagonal and the rhombus cavity type.

Hexagonal cavities:

Cavities are designed in a honey comb structure. Each cavity is separated by a wall of 0.08 mm width. Such ~ 1600 cavities could be realized on a flow chamber area of 18 x 15 mm², a first prototype is designed with 300 cavities with rounded corners to prevent bubbles.

Figure 3.22: design of partly open hexagonal cavity flow chamber. Gates allow priming after sealing however, lateral diffusion is reduced significantly.

71 3.1 Flow cell design and fabrication

Rhombus cavities:

Rhombus shaped cavities with a length of 0.243 mm and width of 0.129 mm are more stream-lined, ~ 8500 cavities could be realized on a flow chamber area of 18 x 15 mm².

Figure 3.23: design of partly open rhombus shaped cavities. Streamlined shapes promise a better priming than the honey comb structure.

3.1.2.5 Advanced master moulds The micro-milling of the PDMS mould masters according to chapter 3.1.2.3 enabled to quickly generate first PDMS inlay prototypes, but lacked accuracy, precision and resolution. Decreasing the flow cell height towards 30 µm at a required evenness of few microns, and replacing the positive sealing edge by a negative one in order to potentially increase the sealing pressure demands a different fabrication method for the moulding master and finally an appropriate and robust moulding tool. Therefore, 4” Si wafers were either laminated with dry resist TMMF S2030 of thickness 30 µm (Tokyo Ohka Kogyo Co., Ltd.), dry resist Ordyl SY 330 of 20 µm thickness (Elga Europe, Italy) or spin-coated with negative SU-8 3000 photoresist (MicroChem) in thicknesses between 5 µm and 15 µm. After photo- lithographically structuring them they are used as mould master in centrifuges for moulding PDMS inlays of high precision in a cost efficient manner.

Mask for photo-lithography:

The SolidWorks CAD file is converted to DXF file format. This file is used to draw the mask structures in AutoCAD, it is converted into *.GBR format using LinkCAD 5. This format is suitable to manufacture foil masks at the physical department of the University of Freiburg.

Figure 3.24: photomask for flow cell master mould created by LinkCAD 5.

72 3.1 Flow cell design and fabrication

Single layer system with dry resist TMMF (30µm):

A negative dry film resist TMMF S2030 of thickness 30 µm is laminated on Si wafer. TMMF is a permanent dry film resist composed of 5 % antimony compound and 95 % novolak type epoxy resin [129]. It is a negative-tone photopolymer designed to be applied with hot roll lamination. Hot and pressurized rollers are used to laminate the substrate (temp = 80 °C, pressure = 1.38 bar). 5 min after the substrate reached room temperature, the protective layer is removed and substrate is soft baked for 20 min at 60 °C. Cracks could appear because of the difference of thermal coefficient of expansion between substrate and resist. Slow heating and cooling is done to avoid cracks. Exposure is performed using mask aligner by UV light. UV light is exposed 6 times for 5 seconds with an interval of 10 seconds using I- Line filter. Post exposure bake is performed at 90 °C. After 10 minutes the substrate is heated up to 150 °C for 40 min with slow heating and finally cooling gradients. Development is done in SU-8 developer and isopropanol bath for 3 minutes each, respectively. Substrate is rinsed in distilled water, dried in centrifuge (2000 rpm). Wafer shows accurate structures.

Summary of protocol:

Table 3.1: recipe for fabrication of structured Si wafer with dry resist. 1) Lamination of 4” Si wafers with TMMF at 80 °C with 1.38 bar. 2) Softbake at 60 °C for 20 min. 3) UV exposure with I-line filter, 6 x for 5 s with interval of 10 s. 4) Post exposure bake at 90 °C for 10 min, then 150 °C for 40 min. 5) Developing in SU8 / Isopropanol bath, each 3 min. 6) Rinse with DI / drying in centrifuge at 2000 rpm.

Multi-layer systems with 2 layers of Ordyl (20 µm):

A negative dry film resist Ordyl of 20 µm thickness is used to make the multilayer structure [130]. Resist is protected by two layers, polyester layer (PET) on one side and polyethylene layer (PE) on the other side. The PE layer is removed and resist is laminated onto substrate with hot elastic rollers (Fig. 3.25, A). Substrate with resist on top is exposed to UV light using mask aligner (Fig. 3.25, B). Exposed substrate is post-baked for 20 sec at 100 °C. PET layer is removed. The second layer of dry film resist Ordyl of thickness 20 µm is laminated, aligned, exposed and post-baked using same procedure as described above (Fig. 3.25, D + E). PET layer is removed. Substrate is developed in Ordyl developer for 45 seconds and later in Isopropanol for 1 minute. Substrate is rinsed in distilled water and dried in centrifuge.

73 3.1 Flow cell design and fabrication

Summary of protocol:

Table 3.2: recipe for fabrication of structured Si wafer with dry resist. 1) Remove polyethylene (PE) layer on one side of dry resist Ordyl. 2) Lamination of 4” Si wafers with Ordyl at 100 °C with 3 bar / 0.5 m/min. 3) Softbake at 100 °C for 5 min. 4) UV exposure for 5.5 s. 5) Post exposure bake at 100 °C for 20 s (could be skipped). 6) Remove polyester (PET) from first layer, PE from second layer. 7) Lamination of second layer 8) Alignment; UV exposure for 7.5 s. 9) Post exposure bake as (5); remove PET layer. 10) Developing in Ordyl developer for 45 s / then in Isopropanol for 1 min. 11) Rinse with DI / drying in centrifuge at 2000 rpm.

(A) (B)

Figure 3.25: fabrication of 2 layer structure on Si wafer. (A) Lamination of first layer. (B) first exposure. (C) Lamination of second layer. (D) second exposure after alignment. (E) development and rinsing.

The dry film resist Ordyl is the simpler, faster and cheaper alternative to SU-8 and especially leads to accurate structures on the Si wafer mould. Measured by profilometer over a length of ~ 1 mm the 2 layer structure shows clear depths of 20 µm and 40 µm, a flatness with an accuracy of ~ 2 µm. The edges are sharp, such the proper function of phase guides can be expected (Fig. 3.26).

Figure 3.26: phase guide with pillar fabricated with 2 times 20 µm Ordyl dry resist; contour measured with profilometer shows accurate profile.

74 3.1 Flow cell design and fabrication

3.1.2.6 Moulding tool and process The manual moulding of PDMS inlays based upon the milled polycarbonate masters in microscope slide format is replaced by a semi-automated process using a structured Si wafer as mould master and a Hettich PDMS centrifuge [131] for the final bubble-free moulding.

Aside the increase in reproducibility we aim to increase the flatness and parallelism of the PDMS inlay. We design an insert for a given standard Hettich centrifuge mould for a quick and cheap solution. An aluminum frame adapts the standard moulding tool, a laser-cut rear side supports the disassembly and finally allows to remove the PDMS inlay from the wafer.

PMMA spacer

structured Si wafer sealing O-ring

Al adapter

6x rectangular PDMS gate rear plate easing disassembly after moulding Figure 3.27: design of wafer adapter with rear side access by metal sheet in order to avoid breaking of the wafer during disassembly. Thickness of PDMS inlay can be adjusted by variation of height of metal sheets inner part.

Moulding process:

Mould is assembled with structured wafer and sealing O-rings. PDMS is poured into the tanks, the centrifuge rotates with 400 rpm at 70 °C for 2 hours. PDMS enters the moulding chamber by 6 rectangular gates of width 0.7 mm in the Al adapter. By centrifugal pressure the PDMS primes the cavities of the structured wafer without any air bubbles.

(A) (B) (C)

Figure 3.28: (A) standard mould with insert before. (B) PDMS primed to mould. (C) insert disassembled after PDMS moulding.

75 3.1 Flow cell design and fabrication

(A) (C)

(B)

Figure 3.29: moulded PDMS inlays. (A) one level design with rhombic pillars. (B) 2 level design with phase guides and pillars. (C) enlargement of yellow rectangle of B.

Summary of moulding process:

Table 3.3: recipe for moulding PDMS inlay with structured Si wafer as mould master. 1) Assembly of wafer, Al moulding tool, O-rings and standard Hettich mould. 2) Inserting assembled mould into Hettich centrifuge. 3) Pouring PDMS into loading reservoirs. 4) Running centrifuge at 400 rpm for 2 hours at 70 °C. 5) Disassembly and cleaning.

3.1.2.7 Flow cell carrier After having designed and realized several variations of PDMS flow cells according to the previous chapters we now elaborate the flow cell carrier. Major challenges are

1) Highly parallel clamping of the glass PDMS sandwich to keep the constant flow chamber height with deviations of ‘few’ micrometers across the flow chamber of 40 mm x 20 mm. 2) Leak tight clamping up to flow rates of 1.5 ml/min. 3) Easy exchange of flow cell sandwich between experiments. 4) Compatibility to iRIf detection system. 5) Flow chamber tempering between 8 °C and 40 °C for protein expression control.

We choose a carrier design based upon the prototype introduced in chapter 3.1.2.3, but with thin stainless ferromagnetic metal sheets with low heat capacities as backbones. Such the tempering will be more dynamical, clamping by levers is replaced by a magnetical clamping. Magnets are equally distributed across the flow cell area hence, the pressure distribution is uniform. The carrier consists of two ferromagnetic plates of material number 1.4021 arranged parallel to each other. Each plate has a dimension of 90 * 40 * 1 mm³. Magnets are positioned by structured PMMA jig fabricated by laser-cutting. Magnets are of cube shape

76 3.1 Flow cell design and fabrication

with side length of 5 mm. Maximum force of each magnet is 10.8 N. One metal sheet offers a laser-cut window for optical detection.

setup adapter top plate

magnets magnets jig

frame for slide

connectors for bottom plate inlet / outlet

Figure 3.30: re-design of carrier for PDMS glass sandwich. Magnetic clamping allows easy and robust exchange of sandwich, while clamping pressure is homogenous across whole flow cell.

PDMS flow cell is placed onto bottom plate inside rectangular frame made out of PMMA. The glass slide is placed on the top of the flow cell. PDMS is self-adhesive to the metal plate and glass surface. This supports leak tightness at inlet, outlet and sealing edge of the flow cell. The rectangular frame also guides the glass slide to its exact position. When both plates are put together, magnets attached to the top plate press onto the glass surface, such sealing the boundary edge of the flow cell. Inlet and outlet tube connectors are assembled to the rear plate. Each connector has a length of 5 mm and an outer diameter of 7 mm. The holes of inlet and outlet are in a distance of 48 mm in standard PDMS inlay size, max. 60 mm apart for variations. The carrier offers an adapter for integration into an automated setup with detection system. The front plate has a window for monitoring the flow processes by camera. Handles on each plate allow easy opening of the clamped system.

This basic magnet type carrier is now enhanced by a rear side tempering system based upon a Peltier element as actor for flow chamber tempering between 8 °C and 40 °C with low gradients across the flow cell. Further we introduce a hinge system for the easy exchange of PDMS inlays and glass slides, and tube pins as inlet and outlet connectors being laser-welded to the metal plate for an increased robustness (Fig. 3.31).

77 3.1 Flow cell design and fabrication

Figure 3.31: (A) design of hinge system for easy exchange, patterned jig for cube magnets of side length 5 mm. (B) design of system with passive cooler for Peltier element, laser-welded inlet / outlet pins and adapter to iRIf setup. (C) fabricated and assembled flow cell carrier with detection window and rear side Peltier tempering unit with fan.

3.1.2.8 Magnetic force versus priming pressure The pressure required to avoid any leakage through the sealing edge depends on the flow rate inside the flow cell. A higher flow rate leads to an increase in pressure loss inside the fluidic chamber, requiring increased pumping pressure. The pressure loss inside the fluidic chamber can be calculated using the Navier-Stokes equations (chapter 2.1.6). Using the condition of laminar flow, the Navier-Stokes equations are simplified to derive a relationship between total volumetric flow and the pressure inside the fluidic chamber according to equation 2.14. With this equation we determine the relationship between the magnetic pressure and the total volumetric flow (Fig. 3.32). Assumed length of the channel is 48 mm, width 10 mm and height 30 µm; the maximal dynamic viscosity of water is 1 mPas.

] 4000 3500 3000 2500 2000 1500 1000 500

0 total volumetric flow [µl/min 0 5 10 15 20 25 30 35 pressure [kPa]

Figure 3.32: pressure versus total volumetric flow in chamber. With 16 magnets we could have a maximal total force of ~ 170 N, due to the air gap of ~ 1 mm we only get 25 N in total. The total pressure with the standard PDMS inlays (effective

78 3.1 Flow cell design and fabrication

area of ~ 1000 mm²) results in ~ 25 kPa. That would allow theoretically maximum volumetric flows of 2800 µl/min according to Fig. 3.32.

3.1.3 Booklet for 2 glass slides (concept 3)

3.1.3.1 Design and fabrication Despite manifold opportunities to even 3-dimensionally structure the PDMS inlays as introduced in chapter 3.1.2 there are reasons to realize a flow cell made of 2 glass slides:

1) Solid surfaces without bias to oszillations on both sides. 2) Manifold surface coatings for both slides possible. 3) Standard array matrix for in situ protein expressions on both sides.

In order to fully exploit the desired usability of a flow cell mainly consisting of 2 glass slides, the assembly design needs to offer a very simple and robust exchange of the glass slides. In comparison to the PDMS inlay design there will be some advantage in tempering this type of flow cell due to the better thermal conductivity of glass (~ 0.9 W/(m*K)) in comparison to PDMS (0.2 W/(m*K)). However, any structuring of the flow chamber for flow control will not be possible. The schematic of the design considers a glass slide for the DNA array, a glass slide for the protein array and an elastic spacer, e.g. PDMS sealing the flow chamber.

DNA array

PDMS frame with spacer and microfluidics

protein array Figure 3.33: schematic of flow cell consisting of 2 glass slides serving as DNA array and protein array matrix. Glass slides are sealed by elastic element, e.g. PDMS determining height of microfluidic gap. Inlet and outlet needs to be incorporated to spacer / scaffold for lateral priming.

Certain challenges need to be met for a fully functional flow cell:

1) Sealing of microfluidic chamber made of 2 glass slides and spacer of ~ 30 µm thickness. 2) Fully leak tight lateral priming of the microfluidic gap. 3) Easy and robust exchange of the glass slides.

First a prototype proves the feasibility of this concept with very cheap and simple methods:

79 3.1 Flow cell design and fabrication

(A) (B)

Figure 3.34: (A) schematic of mould made of laser-cut elements for casting PDMS spacer scaffold. (B) design of sandwich with PDMS scaffold and 2 glass slides.

(A) (B)

Figure 3.35: (A) design of clamping system for flow cell concept 3. (B) first prototype of flow cell system with PDMS scaffold and 2 glass slides. Flow chamber of ~ 60 µm height created by PDMS spacer and glass slides. Priming after clamping (by low-cost metal office clamps) via inlet and microchannels incorporated in PDMS scaffold.

Few experiments proved the basic feasibility of the concept. Hence the very ordinary design based upon utmost cheap elements is revised to enhance the usability, robustness and the reproducibility. Especially the PDMS spacer with incorporated inlet and outlet is too fragile however, the sealing was rather well. Hence we look for a non-corrosive spacer with similar sealing features but significantly better stability and precise parallelism. A stainless steel foil coated with some elastic polymer would meet these demands. We choose a foil of thickness 20 µm of material 1.4301 from Hasberg with a thickness precision of < 3 µm.

inlet outlet Figure 3.36: spacer of 20 µm foil of stainless steel 1.4301, spark-eroded. As elastic sealing layer we choose the hydrophobic PTFE coating. We need to estimate next an appropriate thickness of the layer such that under limited pressure (appropriate to the glass slides) the parallelism is kept precise while deforming the elastic layer to a certain

80 3.1 Flow cell design and fabrication

extent to make the system leak tight. Linear deformations at a constant pressure reveal the necessity of having a polymer layer thickness of >10 µm for an elastical deformation of >1 µm; calculation is based upon modulus of elasticity of 420 MPa for PTFE and a pressure resulting of screw tightening torque of 0.4 Nm. For PTFE coating we passed the spark-eroded metal foils in thicknesses between 5 µm and 25 µm to Nanosol AG, Balzers/Lichtenstein.

Figure 3.37: (A) clamping unit for sandwich of 2 glass slides kept apart from each other by PTFE coated stainless steel (1.4301) foil. (B) glass slides are shifted by ~ 4 mm such that one opens the inlet, the other the outlet. Glass slides overlap the spacer laterally by ~ 1 mm on each side creating a flow chamber of width 23 mm and length 67 mm. Sealing of inlet / outlet to glass slide by vertical sealing cord (purple). (C) PMMA plates press the sandwich together via 6 stainless steel screws tightened each with 0.4 Nm (glass slides withstand limited pressure only). Omnifit connectors allow dispensing manually or via pump. Major enhancement of robustness and usability is realized within a thesis [7] at IMTEK supervised by myself, resulting in the following booklet system with mechanical clamping unit applying significant sealing pressure to sandwich of 2 glass slides separated and sealed by Teflon coated 20 µm metal sheet (Fig. 3.38). 1 2 3 4 1

Figure 3.38: instruction to insert glass slides: (1) open booklet, put first glass slide into frame; (2) turn frame, load second glass slide; (3) close booklet; (4) insert booklet into clamping unit (4); seal flow chamber by turning top screw until stop. [7]

3.1.3.2 Temperature control unit Both heating and cooling primarily by a Peltier element as energy source and heat conduction elements is designed within the bachelor thesis of Nessim Ben Ammar [9] at 3 IMTEK supervised by myself. For cooling heat pipes are transfering the thermal energy from 81 3.2 Optical detection system the top of the Peltier element to the dissipation elements, a passive heat sink and a radiator. The Peltier element is sandwiched between copper plates.

Figure 3.39: design of temperature control system using Peltier element for heating / cooling. Heat sink connected by heat pipes with peltier element dissipates rear side heat from Peltier element while cooling.

The booklet top plate is opened for a Peltier element, 4 sensors for the temperature homogeneity analysis are introduced to the design (Fig. 3.40).

(A) (B) (C)

Figure 3.40: (A) booklet with integrated Peltier based tempering system; (B) Peltier element contacted via temperature distributing elastic graphite foil to copper element; (C) distribution of sensors for temperature homogeneity analysis.

3.2 Optical detection system

The detection principle of the iRIf method is based upon reflecto-interferometry. A LED beam enters the optical system via polarizer, beam splitter [132], λ/2 wave plate and hits the semi-reflective layers of a coated glass slide as part of the flow cell. Reflections at several layers cause interference – the reflected beam (with linear polarized light shifted by the λ/2 wave plate) passes the beam splitter and is detected by the sensor of the camera (Fig. 3.41). Any bio-chemical immobilization onto the outer surface of the glass slides changes the optical density and results in an intensity change of the reflected beam, such the immobilization of any molecules to the surface can be detected.

82 3.2 Optical detection system

Figure 3.41: schematic of the optical detection system. A LED sends light via polarizer, beam splitter and λ/2 wave plate to a thin film stack. Light is partially reflected at thin film layer interfaces, such that interference of the reflected beams occurs. The interfering reflected beam with polarization shifted by 90° by the λ/2 wave plate now passes the beamsplitter and finally enters the camera detector, a CCD or CMOS sensor. Bio-chemical layers on the outer surface of the thin film stack affect the optical thickness and such the intensity of the reflected beam.

3.2.1 Camera and lens system

The prototype iRIf setup is based upon the CCD camera pco.1600 [133] and the Sill Optics lens system S5LPL2060 [134] incorporating the polarizer, the beam splitter and attached the λ/2 wave plate, a LED as light source of 530 nm (optional 470 nm) wavelength and a precise LED current controller (Appendix A4).

The detection unit for the enhanced iRIf setup uses the CMOS camera pco.edge 5.5 [135] and the telecentric lens system S5LPL1005 from Sill Optics [136] with zoom factor 0.926 incorporating the polarizer, the beam splitter and attached the λ/2 wave plate. LED and current controller remained (Appendix A4).

3.2.1.1 Essential camera parameters Dynamic range A/D: gives the resolution of the A/D converter; prototype setup with pco.1600 has a dynamic range of 14 bit (= 16384 increments), the pco.edge provides 16 bit camera (= 65536 increments); e.g. when at least 100 increments are desired to represent the maximal signal changes of 0.3 % (~ 1 % for Streptavidin binding to bBSA on IF glass), in total 33333 (= 100 / 0.003) increments would be needed.

Sensor pixel size: determines the object resolution aside the magnification of the lens.

Sensor size: increasing sensor sizes provide the possibility of having a bigger object size (observed flow cell area) without decreasing the spatial resolution by lens systems; e.g. pco.1600 with sensor size of 8.8 mm x 6.6 mm is inferior to pco.edge with 12.2 mm x 9 mm.

83 3.2 Optical detection system

Pixel scan rate (readout time): a faster readout can be used to increase the sensitivity by averaging more frames or it can be used to build a faster system.

Full-well-capacity: The amount of photons that are available per frame are directly correlated to the shot noise of the system and can, therefore, be used to increase the sensitivity of the system.

Let’s compare the selected cameras systems considering these most significant aspects:

Table 3.4: comparison of essential parameters of pco.1600 and pco.edge. pco.1600 (CCD) pco.edge 5.5 (CMOS) Dynamic range A/D 14 bit (16384) 16 bit (65536) Sensor pixel size 7.4 µm x 7.4 µm 6.5 µm x 6.5 µm Sensor size 12.2 mm x 9 mm 16.6 mm x 14 mm (Sensor pixel resolution) 1600 x 1200 2520 x 2160 Pixel scan rate 2x 40 MHz 286 MHz (Frame rate) max. 30 fps max. 100 fps Full-well capacity 40000 e- 30000 e-

Table 3.5: comparison of essential parameters of lenses for pco.1600 and pco.edge. S5LPL2060 (for pco.1600) S5LPL1005 (for pco.edge) Length 161.1 157.8 Focus length 117 100 Magnification 0.502 0.924 Supported sensor size 8.8 mm x 6.6 mm 12.8 mm x 9.6 mm Distortion 0.3 % 0.2 % (Supported object size) 17.6 mm x 13.2 mm 13.8 mm x 10.3 mm Polarization filter S5SET1199/060 S5SET1199/060

The pco.edge has several advantages – dynamic range, sensor size, resolution, frame rate however, due to a lower full-well capacity the sensitivity affected by shot noise is basically worse in comparison to the pco.1600. The lens affects the sensitivity especially considering homogeneity due to distortions and resolutions determined by magnification. Both the pco.1600 and the pco.edge sensor size can only be used by ~ 70%, as otherwise distortions of bigger 0.3 % [134] or 0.2 % [136] are exceeded; lens magnifications result in supported object sizes of 17.6 mm * 13.2 mm for the pco.1600 system, of 13.8 mm * 10.3 mm for the pco.edge system. The resolution of the pco.edge system (~ 7 µm * 7 µm) is more than twice bigger than the resolution of the pco1600 (~ 15 µm * 15 µm).

84 3.3 Automated setup

3.3 Automated setup

According to objective 2 the analysis of molecule protein interactions by flow injection should run completely automatic for the purpose of robustness, reproducibility and efficiency. The microfluidic flow cell being introduced in the previous chapter in various types as location of the protein molecule interaction is therefore encompassed by technical components for fluidical control, temperature control and optical detection. All individual components interact under the supervision of a programmable process controller. The biosensor data is opto-electronically acquired and digitally processed by software modules.

3.3.1 iRIf prototype for proof-of-concept

The optical setup is mounted onto a steel breadboard with suitable posts (Fig. 3.42, A). The detection of single pixels of 15 µm * 15 µm requires a proper fixation. Camera, lens and flow cell are assembled at lowest possible height, the breadboard is placed on a vibration- cushioned stone plate protecting against external vibrations. The flow cell fixed onto a linear stage (Fig. 3.42, B) is adjusted to the focal plane.

(A) (B)

Figure 3.42: (A) proof-of-concept prototype system with camera, lens, flow cell and peristaltic pump [5]. Lens system incorporates LED, beam splitter, polarization filter, λ/2 plate. (B) Flow cell assembled to linear stage is adjusted to focal plane.

This very basic setup uses a flow cell consisting of PMMA plates and O-ring already available at our lab. At 2.5 rpm the peristaltic pump generates a flow of ~ 100 μl/min. Black PVC tape is attached to the back of the flow cell to prevent light scatter by reflections. This setup enabled us to get first 1λ-iRIf measurements in our lab. An image processing software based upon ImageJ was realized for real-time monitoring and post data processing with following modules:

 Live view of quotient images.

85 3.3 Automated setup

 Binding curves depicting relative intensity changes over time for predefined spots.  Evaluation of the binding curves for arbitrary regions after the experiment.  Generation of movies as sequence of quotient images after the experiment.

Figure 3.43: data acquisition by camera / camware; live view by real-time data processing with ImageJ macros.

This system proved to be suitable for the basic analysis of molecule protein interactions on microarrays. Both real-time depiction and post-experimental analysis could be performed by the ImageJ evaluation modules e.g. binding of Streptavidin-Cy5 to bBSA (50% dilution of 1 mg/ml bBSA in PBS) spotted to GOPTS coated IF glass slides, 1 ml of Streptavidin-Cy5 was pumped over the array at 100 µl/min for 10 minutes .

(A) (B) (C)

Figure 3.44: (A) primed flow cell after 2 minutes of injection of Streptavidin-Cy5 with flow rate of 100 µl/min, no binding visible, only bBSA spots outside of flow cell due to oszillations. (B) binding of Streptavidin-Cy5 to bBSA spots occurs after 4 minutes injection according to flow direction from bottom to top. (C) binding saturated after 8 minutes of injection. [5]

System lacks

 microfluidic versatility considering fluidic control and reagent consumption.  surface coatings for top side of the flow chamber.  automation with complete fluidic and thermic control.

86 3.3 Automated setup

3.3.2 First prototype of automated system

Based upon the proof-of-concept prototype an automated prototype is devised applying a rather simple system architecture with basic fluidic components, but sophistated flow cell systems according to chapter 3.1 with integrated thermic control. A process controller for the control of all components, that are syringe pump, 9-fold fluidic valve and temperature controller is realized in C#, the image processing module is integrated as Java based software for open and versatile processing of digital camera data.

Figure 3.45: schematic of 1st automated prototype based upon single syringe pump and 9-fold valve connecting the flow cell with the reagent storage. All components including the tempering unit is controlled by the supervising process control, the camera as digital biosensor provides data of the molecule protein interaction process, visualized by Java modules.

The flow cell could be primed either by pulling the reagent from the storage tube by the syringe pump into the flow cell or by pressing it into the flow cell from a sample loop (Fig. 3.46, A). The pulling method supports bubble-free priming however, it demands a high degree of leak tightness due to underpressure.

(A) (B)

Figure 3.46: (A) priming of flow cell by pushing reagent from sample loop into flow cell. Tube between valve and pump could be washed during incubation. (B) housing for syringe pump, valve, temperature controller and reagent storage.

87 3.3 Automated setup

As syringe pump we choose the product Cavro XE1000 [138]. For the selection of up to 6 reagents and buffer we choose the product Cavro SVP valve with 9 port distribution adapter [139], Teflon plug and PCTFE body. Fig. 3.46, B shows design of housing with front board for syringe pump, 9-fold valve and temperature controller; energy supply and communication from rear side. Left plate for 6 reagents and 1 buffer. Fittings as connectors of type Omni-Lok Inverskonus are used [140], in particular article no. 2145 (diameter 1/16” of material PEEK).

Flow cell tempering by Peltier element of size 30 x 30 mm², 70 W heating and 34 W cooling power controlled by Omron E5CN temperature controller via RS485 interface with resolution of 0.1 K. Peltier connected to flow cell by copper adapter, thermically sealed by PTFE element. Electrical Reverser Unit (ERU) as switching amplifier realizes change of current flow direction for cooling or heating. Platinum resistance thermometer (PT100) as temperature sensor grants for high accuracy (absolute and relative), low drift and wide operating range (Fig. 3.47, A). The performance of the system is depicted in chapter 4.1.3.

(A) Electrical Reverser Unit (B)

Omron E5CN

Peltier element

Temperature sensor PT100

Figure 3.47: (A) temperature control system with PT100 sensor, digital controller, electrical reverser unit (ERU), Peltier element. (B) Flow cell with Peltier based tempering system assembled to camera setup, rear side cooling of Peltier element by fan.

Fig. 3.48 shows final assembly with reagent storage on the left side, 9-port distribution valve, syringe pump and Omron temperature controller. The user interface of the process control according to Fig. 3.49 allows to create a protocol in a language similar to ‘C’ with loops, conditions and variables. This protocol determines the sequence of actions and parameters for flow rates, reagent volumes, incubation times or flow cell temperatures. Both the process control software written in C# as well as the mechatronical components proved to be suitable as automated system. The piping system lacks operability as pipes need to be inserted to each reagent tube manually after washing routines between experiments. Further, due to one syringe pump only synchronous aspiration and dispensing can not be realized, such there is no continuous flow between reagents possible.

88 3.3 Automated setup

Figure 3.48: prototype of automated system with Tecan syringe pump, Tecan 9-fold valve and Omron temperature controller. Flow cell without temperature control attached, further flow cell types with standardized adapters exposed on top of housing can be selected. Insert (top left): flow cell and lens covered by light scatter protection.

Figure 3.49: user interface (UI) of process controller. On the left side, the process protocol is depicted. It can be written sequentially with conditions, loops and variables; syntax is similar to C programming language. Bottom right offers variables to be created and assigned for protocol templates.

89 3.3 Automated setup

3.3.3 Elaborated prototype of automated system

This system eliminates the disadvantages of the automated system prototype introduced in the previous chapter and further integrates the camera based optical detection system to the process control system. The main arrangement of the components is designed according to Fig. 3.50. On the right rear side the electrically controlled fluidic system (1) with the housing for a Vici Pump M6, a Hamilton syringe pump PSD/4, a Hamilton MVP 6 port loop valve and a Hamilton MVP 4 port loop valve. This housing further serves as electrical cabinet for the whole system e.g. temperature controller, current control for the LED, voltage supply. In front of this cabinet there is a sampler (2) in MTP format assembled onto a Festo EXCM x-y stage, possibly cooled by Peltier elements. A Festo CMMO axis enables a cantilever moving vertically and taking the reagent samples. Left to the sampler we have the microfluidic flow cell (3) mount onto a stone plate (4). This stone plate further fixes the optical detection system, a pco.edge camera and a telecentric lens system (5).

5 1

4 2 3

Figure 3.50: arrangement of iRIf system components: fluidic control system and electrical cabinet (1), reagent sampler (2), microfluidic flow cell (3), damped stone plate (4) and optical detection system (5). Such the tubing length is minimized for low dead volumes, the whole system remains portable with a weight of ~ 25 kg.

This arrangement allows to minimize the tube length of the fluidic system and such the dead volumes within the tubes. The fluidic system is designed to follow the idea of having a constant flow of different reagents or buffer through the flow cell (Fig. 3.51). While buffer is dispensed, the next reagent sample is aspirated. During the dispension of the reagent the reagent sampling tip is washed in order to avoid cross-contamination. The system allows to have sample volumes of up to 1100 µl, as the sample loop tube provides a total volume of

90 3.3 Automated setup

1147 µl. The continuous dispensing is realized by a Vici pump M6 allowing flow rates between 5 nl/min and 5 ml/min, the aspiration of the samples into the sample loop by a Hamilton PSD/4 syringe pump with 2.5 ml syringe. For the fluidic control we use 2 Hamilton MVP valves, a 4-port and a 6-port loop valve. In mode 1 (as depicted in Fig. 3.51) the Vici is pumping buffer through the flow cell while the PSD/4 syringe pump first washes the sample loop with buffer taken via valve port 3I, then aspirates the next reagent into the sample loop. From the sampler tip to the sample loop there is a total dead volume of ~ 40 µl. In mode 2 both valves turn one step to the right. Now the Vici pump dispenses into valve port 4.6, out through port 4.1 into valve port 5.4, out through 5.3 via sample loop into 5.1, out through 5.2 into 4.4, then out through 4.5 into the flow cell. By this arrangement the Taylor dispersion is minimized as the flow direction is changed between aspiration of the reagent into the sample loop and dispensing the reagent into the flow cell.

Figure 3.51: fluidic system for constant flow of different reagents or buffer through flow cell. While buffer is dispensed, the next reagent sample is aspirated. Sample volumes of up to ~1100 µl possible. Vici pump M6 (1) with flow rates of 5 nl/min to 5 ml/min for continuous dispensation, aspiration of samples and washing of the fluidic system by a Hamilton PSD/4 syringe pump (2) with 2.5 ml syringe and 3-port valve (3). 2 Hamilton MVP valves (a 6-port (4) and a 4-port (5) loop valve) control the fluids. Technical data of valves and tubings are listed in Appendix A4.

Any cross-contamination has to be avoided. After each reagent step the sample loop is cleaned by a washing procedure. The sampler is moving to the washing station (Fig. 3.52). The sample loop and the tubing are washed by pumping 1.2 ml (programmable) of buffer through the system into the washing stations waste (Fig. 3.52, 1). Then the sampling tips outside is cleaned by diving successively into a primary (Fig. 3.52, 2) and secondary (Fig. 3.52, 3) washing tube. Each 500 µl tube is primed with 1 ml of buffer causing first the filling up of the tube until a certain level but then emptying it via a bottom hole after the gravitation force is overcoming the pinning force.

91 3.3 Automated setup

2 3

Figure 3.52: washing station with 50 ml waste container (1) being easily exchangeable. Washing of sampling tips outer part by diving successively into primary (2) and secondary (3) washing tube – pumping of buffer causes filling up of tubes but regular emptying via bottom hole after overcoming pinning force by gravitation force.

Total system assembly:

According to the components arrangement of Fig. 3.53 the optical detection unit with camera, telecentric lens system and flow cell is assembled on an Impala stone plate and damped with Bilz Damper B13W/B8 (Fig. 3.53, 1). Rear right, there is the electrical cabinet (Fig. 3.53, 2) housing the electrically controlled fluidic system consisting of a Vici Pump M6, a Hamilton syringe pump PSD/4 with 2.5 ml syringe, a Hamilton MVP 6 port loop valve and a Hamilton MVP 4 port loop valve. Each of this component is controlled via serial-USB interface. Further this cabinet houses the current control for the LED and the voltage distribution panel. In front of this cabinet there is the sampler made of the Festo EXCM x-y stage of range 150 mm x 110 mm and the Festo CMMO axis for vertical movements of 40 mm (Fig. 3.53, 3). Each Festo axis is controlled via TCP/IP interface. The vertical axis carries a cantilever for holding the sampler tip. The washing station is attached to the sampler, the 50 ml waste container can be easily emptied by pulling it out of a laser-cut clamp. Essential features of this setup are listed in table 3.6, further data as parts list is given in the Appendix A4.

92 3.3 Automated setup

1 3

Figure: 3.53: iRIf system setup. The optical detection unit with camera, telecentric lens system and flow cell with semi-reflective layers on glass slide assembled on an Impala stone plate and damped (1). The electrical cabinet for power supply, LED current controller, temperature controller (USB interface) for rack and flow cell, controllers and fluidic mechanics of Vici pump M6 (RS485 interface), Hamilton PSD/4 syringe pump (RS485 interface), Hamilton MVP 4-port loop (2x RS485 interface) and 6-port loop valve (2). Sampler for 33x 500 µl tubes and 11x 1.5 ml tubes mounted on Festo EXCM x-y stage (TCP/IP interface), range of 150 mm x 110 mm. Cantilever for sampling tube on Festo CMMO z-axis (TCP/IP interface), range of 40 mm (3). Washing station as depicted in Fig. 3.52. User interfaces for process control, image acquisition and image processing (4); excel chart as UI for flow and temperature protocol is depicted in Appendix A4.

Table 3.6: essential technical data of enhanced iRIf setup. weight 35 kg size 500 x 500 x 200 mm³ no. of reagent tubes 33 x 500 µl; 11 x 1.5 µl flow rates 5 nl/min to 5 ml/min continuous flow dead volumes < 100 µl frame rate < 50 fps object resolution 7 µm baseline drift // baseline noise < 3 * 10-3 ‰/min // < 3 * 10-9 RSS/min

A process controller programmed in C# (Fig. 3.53, 4) is running on a standard personal computer. It controls the fluidic components Vici pump M6, Hamilton syringe pump PSD/4, Hamilton valves MVP 4-port loop and MVP 6-port loop via serial-to-USB interface, the Festo x-y-z axes via TCP/IP interface and the temperature controllers for the sampler and the flow

93 3.3 Automated setup cell via USB interface. The process controller SW is open for upgrading the system by further hardware components and allows them to communicate via any standard ports. Process protocols can be edited or programmed by skilled users in a simplified C code, which is being interpreted at run time. Loops, conditions and variables help to realize logical operations for any interaction of control components. The outer shell of the user interface allows any user the easy operation of the setup by simply entering parameters data for standardized protocols into an excel chart e.g. flow rates, priming times, temperatures (Appendix A4). While the fluidic protocol is running, the camera observes the immobilization processes according to the 1λ-iRIf principle. The image processing interface allows to view the camera picture in a live view mode, or the scaled camera picture in the live view iRIf mode. The picture is scaled by any reference picture the camera has taken already. As the intensity changes by the reagent surface bindings are rather low of max. 5% for good binders, only by applying a look-up table to the scaled picture the spots can be made visible (Fig. 3.54).

Figure 3.54: UI of iRIf setup for image processing and analysis. Monitoring functions as camera live view or iRIf live view are provided; further post-processing functions for movies of any sequence, quotient pics or binding curves calculation.

After completion of the process protocol, the data can be pre-viewed by creating quotient pictures or making movies with selectable reference pictures. For the detailed analysis and especially quantification of any binding interactions, process time-resolved binding curves are calculated and evaluated according to the following chapter.

94 3.4 Image processing

3.4 Image processing

3.4.1 Digital filtering

3.4.1.1 Calculation of the mean The mean value is the average value of a range of values e.g. pixel intensities in a region of interest (ROI). The mean takes every value into account but yields a smoothened average value. Such an average value across a ROI can be considered as a spatial average.

3.4.1.2 Calculation of the median The median of a range of values e.g. pixel intensities, is determined by sorting these values and taking the middle value in case of an odd number of values, or the average of the two middle values in case of an even number of values. Taking the median of a set of frames now means to take the median intensity of each pixel of these images and building up a new image, the median of that set of images. The median works as a band filter eliminating value peaks (max or min) caused by noise.

3.4.1.3 Calculation of quotient pics Only by calculating the pics double quotient by ‘pic/picref divided by bg/bgref’ low intensity variations during the binding process can be visualized. The range of the LUT is essential for visual analysis.

Figure 3.55: pure camera pics (top row) can not recognize any binding; only the calculation of quotient pics and the application of an appropriate look-up table (LUT, e.g. 0.995 to 1.05) makes small signal changes visible. [8]

For each frame, the mean pixel intensities of the spots are divided by the mean intensities of their local backgrounds for normalization. Any two of these normalized values can now be

95 3.4 Image processing divided by each other to get the relative spot intensity, that is the mean intensity change of spots pixel values from time t1 to time t2:

푛2푏𝑔 (∑푛2 (퐼 )/푛 ) / (∑ (퐼 )/푛 ) 1 2 2 1 2푏𝑔 2푏𝑔 퐼푟 (푡2, 푡1) = 푛1 (3.3) (∑푛1 (퐼 )/푛 ) / (∑ 푏𝑔 (퐼 )/푛 ) 1 1 1 1 1푏𝑔 1푏𝑔 with

Ir (t2, t1) … mean intensity change from t1 to t2 of normalized spot

I1, I2 … pixel intensities at times t1 and t2

n1, n2, n1bg, n2bg … number of pixels of data and background spots

For increasing the SNR by reducing noise, e.g. shot noise [141] a median reference picture can be calculated from a range of frames. The implemented method calculates the median of each pixel of the selected ROI’s.

3.4.2 Spot identification

[8] Each spot of interest and its local background is identified manually or by spot identification routines programed in Matlab. For a first approach to automated spot identification we discuss 2 methods:

1) Determination of spotted grid by manually selecting 3 spots. All spot centers could be calculated, independent of any threshold. Spots ROI’s would always have the same size, non-symetric arrays can not be detected.

Figure 3.56: simple spot region determination by calculating regular grids based upon 3 manually marked spots. Fragmented spots or out of place ones are not hit well. [8]

2) Applying algorithm on binary image to identify spot borders after digital noise filtering. Irregular spotting patterns could be detected, weak spots below threshold couldn’t; manual editing would allow amendments.

96 3.5 Surface chemistry

Figure 3.57: border following algorithm identifies any spots above threshold. Low intensity spots could be marked manually after the automated spot identification. [8]

Method 2 is superior due to no limitation to grid or spot shape irregularities and therefore implemented with Matlab routines. A modified method 1 could be applied thereafter by analyzing spotting pattern of the detected spots and interpolating (even extrapolating) it.

3.4.3 Binding curve evaluation

To determine the binding coefficient KD, an exponential function is fitted to each binding curve according to chapter 2.5 by a Matlab program based upon the least-squares fitting method (Appendix A5) [142].

3.5 Surface chemistry

Our applications, both the protein microarray synthesis and the molecule-protein interaction analysis require molecules to be immobilized on flat substrates, either glass slides or PDMS slides in microscope slide format. The surfaces of these substrates therefore need to be activated.

The protein microarray synthesis requires a DNA microarray to be immobilized on either a glass or a PDMS slide, further a protein capture surface for the trapping of the expressed proteins diffusing through the microfluidic gap.

The molecule-protein interaction analysis requires activated surfaces immobilizing the initial target molecules array.

97 3.5 Surface chemistry

3.5.1 PDITC (for DNA immobilization)

Table 3.7: PDITC functionalization of PDMS slides; DNA amino linker binds to thiocyanate group. [143] 1) Rinse PDMS slide with EtOH, 3 times, blow dry with nitrogene. 2) Surface activation in hydrogene plasm, 4 min, 100 W. 3) Incubation for 2 hours in 3-aminopropyltriethoxysilane (APTES). 4) Rinse with EtOH, 2 times. 5) Curing at 70 °C for 1 hour. 6) Incubation over night in 1,4-phenylenediisothiocyanate (PDITC). 7) Rinse with EtOH, 2 times.

3.5.2 GOPTS (for DNA or protein immobilization)

Table 3.8: GOPTS functionalization of glass slides; DNA amino linker binds to epoxy group. 1) Rinse glass slide with EtOH, 3 times, blow dry with nitrogene. 2) Surface activation in hydrogene plasm, 4 min, 100 W. 3) Incubation for 1 hour in 3-Glycidyloxypropyl-tri-methoxysilane (GOPTS). 4) Rinse with Acetone.

3.5.3 Ni-NTA (for protein immobilization by His tag)

Table 3.9: Ni-NTA functionalization of glass slides. 1) GOPTS protocol according to table 3.8. 2) Polymer network with amino-dextrane (14 mg AMD / 86 µl DI; over night) 3) Rinse with DI. 4) Transfunctionalisation by Glutaric anhydride (68 mg GA / 32 mg water-free DMF; >= 4 h) 5) Rinse with water-free DMF. 6) NHS-ester activation (11.52 mg NHS / 100 µl water-free DMF; 76.77 ml of this solution / 23.23 µl DIC; >= 2 h) 7) Rinse with water-free acetone. 8) Ni-NTA (10 mg TACN / 100 µl water-free DMF; 1.5 h) 9) Rinse with DMF, then with DI.

10) Incubate in NiCl2 (10 mM NiCl2 in DI; 1 h). 11) Rinse with DI.

98 3.6 Biochemistry

3.6 Biochemistry

This chapter introduces methods for the preparation of expression-ready DNA and its amplification, the spotting of DNA into arrays, the cell-free expression of protein arrays from these DNA arrays and the labelling of the protein array spots by antibodies for detection.

3.6.1 DNA amplification and analysis

PCR protocol using Qiagen HotStar Taq polymerase for the amplification of A07 (862 bp), D07 (1294 bp) templates provided by project partner O. Stoevesandt / Babraham Bioscience Technologies Ltd. (sequences in Appendix A2) and GFP (~ 1000 bp, inherited construct, detailed sequence not known):

Table 3.10: PCR protocol for the amplification of DNA templates A07 and D07 (Appendix A2; provided by O. Stoevesandt, BBT Cambridge) with Qiagen HotStar Taq polymerase. 30 cycles of 3 steps with denaturalization at 94 °C, primer hybridization at 54 °C and elongation at 72 °C have been selected.

3.6.2 Microarray spotting

Protein microarray synthesis

Arrays were spotted according to protocol 2 (Appendix A2.4) for benchmarking DAPA with microfluidic in situ protein expression (chapter 4.2.4). Arraying was carried out on standard epoxy activated microarray glass slides from Schott-Nexterion for DAPA and on PDITC (chapter 3.5) coated PDMS flow cells for the expression in microfluidic flow cells.

99 3.6 Biochemistry

Bio-physical evaluation of the automated setup

Microarrays were plotted with the Scienion sciFLEXARRAYER S3/S5 [144] provided by the chair of Chemistry and Physics of Interfaces, IMTEK, University of Freiburg. During the printing procedure spots are kept humid by humidity control. The spotter uses a non-contact spotting technology, the sample volume for one droplet ranges from 50 pl to 800 pl per droplet, we used 700 pl/droplet for our microarrays. 20 clusters of 6 spots (5 columns x 4 rows) of bBSA are printed according to Fig. 3.58. Each row consists of 5 clusters with concentrations of 4.5 / 15 / 50 / 15 / 4.5 % from left to right, each column of 4 clusters with spot volumes of 0.7 / 2.1 / 6.3 / 18.9 nl.

Figure 3.58: spotting pattern of bBSA microarrays with clusters consisting of 3 rows each of 2 spots, the upper and lower row spots shifted to the right. From left to right the clusters change in bBSA / BSA (each 1 mg/ml in PBS) concentration (4.5 / 15 / 50 / 15 / 4.5 %). From top to bottom the clusters change in spot volumes (0.7 / 2.1 / 6.3 / 18.9 nl/spot).

For the concentration series we mixed bBSA with BSA, both in concentrations of 1 mg/ml in PBS (phosphate buffered saline, pH 7.4). For the immobilization of the bBSA spots we coated the glass slides with GOPTS.

First assay with biological impact

For analyses of assays with biological impact epitope mapping of antibodies was applied. We choose CRP antibodies on NHS-Ester functionalised surfaces for the analysis. After some basic experiments with manually spotted dilution series of antibodies and antigens, we use concentrations according to our stock solutions for the analysis, that is 200 μg/ml for the CRP antigen and 500 μg/ml for the capture (αCRP1) and the detection antibody (αCRP2 with biotin). Three rows of each antibody and antigen were immobilised with equal sample volumes on the left, with different sample volumes ranging from 400 pl to 4 nl on the right side; further 1 mg/ml bBSA (equals 100% bBSA) control spots.

100 3.6 Biochemistry

Figure 3.59: spotting pattern of CRP1-, CRP2-antibodies, CRP antigen and bBSA control spots with 3 rows of each reagent in constant concentrations but in different volumes from 400 pl (1 drop) to 4 nl (10 drops). [8]

Thrombin aptamer assay

Thrombin (M = 39 kDa) is the major protease for blood coagulation. It splits the protein fibrinogen while releasing fibrin, which itself polymerases and such triggers the coagulation. We now look for Thrombin inhibitiors which suppress the blood coagulation. These could be peptides or even smaller haptenes [100] - our objective is to find aptamers as synthetic inhibitors which could have a major impact on therapeutics. Experiment series with different concentrations of Thrombin are performed upon aptamer microarrays with 76 dots of 2.1 nl spotted with Scienion sciFLEXARRAYER S3/S5 (provided by the chair of Chemistry and Physics of Interfaces (CPI), IMTEK, University of Freiburg) according to Fig. 3.60, consisting of 10 different aptamers (Appendix A2) and bBSA spots as positive control:

Figure 3.60: array layout of aptamers spotted by IMTEK/Lab of CDI – 2.1 nl/spot (3 shots of 0.7 nl), incubation in humid atmosphere sealed with parafilm over night (protected from light), immersed 30 min in 10% Ethanolamine, flushed with DI water, immersed 2x for 5 min in DI water @70 °C, blown dry with N2, stored sealed with parafilm in Nitrogen atmosphere, blocked in 1% BSA diluted in PBS for 15 min (10 µg/ml), flushed slide holder with DI water and washed each slide with DI water, blown dry with nitrogen (image by Christin Rath, AG Roth).

101 3.6 Biochemistry

3.6.3 In situ protein microarray synthesis

According to the assembly technology introduced in chapter 2.8 constructs as the template A07 can be expanded by N-terminal 6His tag and C-terminal flag. Cell-free expression mixtures were prepared according to the manufacturer’s instructions, using RTS 100 E.coli HY kit [145] or the Qiagen EasyXpress kit [25]. Ni-NTA coated glass slides from Xenopore [146] were used as protein capture surfaces. Assembly of DAPA setups with membrane was carried out as described previously [31]. The microfluidic system essentially consists of a erDNA array matrix, e.g. PDMS inlay, which is sealed directly by the Ni-NTA glass slide, allowing in situ expression:

1) Priming microfluidic flow chamber with ~ 30 µl of cell-free system. 2) Cell-free protein expression at 37 °C in incubator. 3) Disassembly of setups in bath of 0.05% Tween20 in PBS.

4) Rinsing erDNA microarray with H2O; drying and storing at 4 °C. 5) Blocking protein array slides in 1% BSA, 0.05% Tween20 in PBS for 30 minutes; drying in centrifuge. 6) Visualizing the immobilized proteins by immunofluorescence staining with an antibody mixture: a. Materials: i. mouse anti-p53 (f.c. 250 ng/ml), anti-mouse-cy3 (f.c. 500 ng/ml). ii. rabbit anti-c-myc (f.c. 250 ng/ml) and anti-rabbit-Cy5 (f.c. 500 ng/ml). b. Dispensing 100 µl of this mixture onto protein arrays, superimposing cover slides onto the liquid. c. Incubation at 4 °C overnight. d. Removing cover slides; washing protein arrays in 0.05% Tween20 in PBS 3

times 5 minutes while rocking, rinsing them with H2O, drying in centrifuge. 7) Scanning protein slides with Cy3- and Cy5-filters.

3.6.4 Aptamers

For aptamer binding analysis thrombin aptamers - the fibrinogen-binding-site and the heparin-binding-site aptamer [147] was used. Mold fungus aptamers (provided by Christine Reinemann, Helmholtz-Zentrum Leipzig) were used as negative binding controls. As positive target control the bBSA-Streptavidin-Cy5 binding was used. To ensure immobilization we also printed controls with fluorescent labels on the microarrays for iRIf measurements; please take details from Appendix A2.3.

102 4.1 Basic analyses and preparations

4 Experiments and Results

4.1 Basic analyses and preparations

4.1.1 Contact angles

Contact angles of relevant liquid surface combinations have been measured (table 4.1 + 4.2). Contact angles on glass are significantly lower than on PDMS. Adding BSA to PBS buffer is not essentially changing the contact angle.

Table 4.1: advancing contact angles on PDMS with some relevant liquids. measurement PBS 0.3% BSA in PBS 2.5% BSA in PBS 1 107 106 100 2 104 102 103 3 101 107 97 average 104 105 100 Table 4.2: advancing contact angles on glass with some relevant liquids. measurement PBS 0.3% BSA in PBS 2.5% BSA in PBS 1 63 58 57 2 62 62 59 3 61 57 61 average 62 59 59

4.1.2 Experimental flow analysis

[6] Now we fluidically evaluate the PDMS flow cell designs with priming studies. We choose PBS as test liquid. We aim to have a bubble-free priming, no leackages, homogenous flow profiles, no encasing of air bubbles.

Flow cell with rhombic pillars:

Figure 4.1: design with rhombic pillars in flow chamber of height 30 µm.

103 4.1 Basic analyses and preparations

Figure 4.2: priming happens without any visible bubbles. Pillars do not seem to disturb the meniscus, which however is not completely planar. [6]

Channel type:

Figure 4.3: design with channels in flow chamber of height 30 µm.

Figure 4.4: priming happens inhomogenous in each channel. Liquid starts receding backwards into left channel and encloses air bubble. [6]

Geometric valve type:

Figure 4.5: design with channels and geometric valves in flow chamber of height 30 µm.

Aim of this design is to completely fill each channel and to stop the backward flow of liquid. Only when all channels are filled the liquid moves towards the outlet.

104 4.1 Basic analyses and preparations

Figure 4.6: priming happens inhomogenous in each channel but geometric valves avoid receding flow and such trapping of air bubbles within the channels. Only outside the channels remain tiny air spots. [6]

Each channel is filled with slightly different rate but without remaining air bubble. Geometric valves open before complete priming of the middle channel, the pinning pressure of ~ 1200 Pa shoud be increased for future designs.

Phase guides type 1 (sharp edges): Phase guides and pillars have a height 20 µm each, the flow chambers total height is 40 µm.

Figure 4.7: design with phase guides with sharp edges of height 20 µm. Flow chamber total height is 40 µm. Pillars of height 20 µm on top of phase guides stabilize flow cell.

Figure 4.8: priming happens along phase guides. No visible air bubbles trapped in corners of phase guides. [6]

Phase guides control the flow of liquid. Each chamber is filled one by one. Pillars do not seem to disturb.

105 4.1 Basic analyses and preparations

Phase guides type 2 (round edges):

Figure 4.9: design with phase guides with curved edges of height 20 µm. Flow chamber total height is 40 µm. Pillars of height 20 µm on top of phase guides stabilize flow cell.

Figure 4.10: priming happens along phase guides. No visible air bubbles trapped in corners of phase guides. Pillars slightly separate flow. [6]

Liquid follows the phase guides. As the width of the pillars is smaller than the width of the phase guides, liquid join together on phase guide before bursting. Round edges do not give a visible advantage but could have positive effects at higher priming rates.

Plane type:

Figure 4.11: design with plane flow chamber of height 30 µm.

Figure 4.12: priming happens rather homogenous. Collapse of flow chamber due to missing support structures (pillars, lines) did not happen. [6]

106 4.1 Basic analyses and preparations

Fabrication of flow cell by silicon master results in highly even flow chambers of equal height. The clamping is maintaining the flatness of the flow cell. Further designs with semi- open cavities have been realized for pre-analyses (see chapter 3.1.2). The analysed flow cell designs offer characteristics for a range of applications, e.g. flow-through measurements, in situ protein expressions. Experiments indicated non-cavity flow cells kept their leak tightness up to flow rates of 700 µl/min, cavity types up to 100 µl/min; due to envisaged applications that is sufficient and no real limitation. The fabrication of master moulds based upon multi- lamination of dry-resists onto Si wafers proved to be both robust and versatile.

Inlet and outlet geometry:

Both inlets and outlets should be designed such that any air bubble trapping is avoided. Inlets should support the bubble-free priming by appropriate geometries, e.g. opening channel supports pinning while priming:

Figure 4.13: (left) inlet hole at very right position eliminates dead volume. (right) opening geometry leads to pinning and such controlled priming with stable meniscus.

While avoiding the trapping of any bubbles, outlets should smoothly guide the parallel flow profile back to the round outlet hole.

trapped air bubble

Figure 4.14: outlet hole should be utmost right to minimize probability of trapping a bubble (top right). Oval shaped outlet geometry leads fluids smoothly to outlet hole.

107 4.1 Basic analyses and preparations

4.1.3 Analysis of temperature control and homogeneity

The temperature control system according to chapter 3.3.2 is now tested with the sensor placed aside the Peltier element on the metal plate. The temperature gradient while heating is ~ 1 K/s. The lower set point of 4 °C and the upper setpoint of 37 °C are hit well with some oscillations of ~ 1 K.

~ 1 K

temperature temperature [°C]

time [s] Figure 4.15: temperature control system allows tempering the flow cell but reveals some instability with oscillations of ~ 1 K.

According to the sensor arrangement described in chapter 3.1.3.2 the resulting temperature curves follow the simulations performed within the associated master thesis of Nessim Ben Ammar; temperature gradients of less than 0.8 K across the flow cell chamber have been measured.

4.1.4 DNA array spotting

Basic analysis of DNA immobilization by amino linkers (Appendix 2.4, protocol 1) of templates A07 and D07 with Cy3 label and template ‘97’ (unknown sequence) with FITC label.

A07 blocked D07 denaturated 97

Figure 4.16: spotting pattern with depiction of finally blocked and denatured region for analysis of immobilization and denaturation of DNA A07 and D07 with Cy3 label, template ’97’ with FITC label.

108 4.1 Basic analyses and preparations

(A) (B)

Figure 4.17: analysis of immobilization of DNA A07 and D07 with Cy3 label, template ’97’ with FITC label: (A) Cy3 scan (500 ms) after washing with SSC buffer. (B) as ’A’, but blocked and denaturated. (C) FAM scan (500 ms) after washing with SSC buffer. (D) as ’C’, but blocked and denaturated.

Immobilization visualized by Cy3 and FITC label could be realized for all spots; washing with 0,2 M NaCl + 0,125 M NaOH for denaturation of double-stranded DNA, 15 min denatured A07 and D07 significantly, template ‘97’ hardly; blocking did not result in any visible effects.

4.1.4.1 Spotting of DNA arrays by TopSpot Arrays of DNA D07 spotted by TopSpot [148] according to protocol 1 (Appendix A2.4) onto PDITC coated PDMS flow cell (Fig. 4.18) and PDITC coated glass slide (Fig. 4.19):

3 mm

Figure 4.18: array of 16 clusters of DNA template D07 (100 ng/µl) spotted with ~ 1 nl/spot onto PDITC coated PDMS flow cell in 2 concentrations, 3:1 (1 column / cluster) and 6:1 (2 columns / cluster), dilution by Nexterion spotting buffer [149]; detected by Cy3 label at 200 ms exposure time.

3 mm

Figure 4.19: array with 3 rows of 16 clusters of template D07 (100 ng/µl) spotted with ~ 1 nl/spot onto PDITC coated glass slide in 2 concentrations, 3:1 (1 column / cluster) and 6:1 (2 columns / cluster), dilution by Nexterion spotting buffer [149]; detected by Cy3 label at 200 ms exposure time.

109 4.2 Cell-free protein microarray synthesis in flow cell

Spotting of DNA D07 onto both PDITC coated PDMS and glass resulted in arrays being well immobilized and such ready for the cell-free expression. Deficiencies as e.g. irregular spot distances, planets need to be eliminated by optimization of the spotting procedure.

4.2 Cell-free protein microarray synthesis in flow cell

Next we analyse the suitability of the flow cell designs (concepts 1-3) for in situ protein expression of immobilized DNA arrays.

4.2.1 Concept 1

Joint experiments with Oda Stoevesandt: an epoxy coated microscope slide (NEXTERION® Slide E from Schott) [150] is used as DNA microarray matrix; such the spotted DNA is immobilized via amino linker. As protein capture slide a Ni-NTA coated slide from Xenopore Cooperation [146] is selected. The microfluidic chamber is primed with cell-free expression system RTS 100 [145] by pipette. During 3 hours at 37 °C proteins are expressed and finally immobilized by their His tag to the Ni-NTA surface as protein microarray. Two types of DNA have been spotted by the GeSIM Nano-Plotter 2.1 [151] at spot distances of ~ 1 mm with a volume of ~ 5 nl/spot: Cy3-TF15myc-2His6 (Fig. 4.20, A, green spots) and Cy5- GFP-2His6 (Fig. 4.20, A, red spots). Protein spots are visualized via immunostaining with rabbit-a-myc / a-rabbit-Al532 for protein 1, with a-GFP-bio / Strep-Al633 for protein 2 by fluorescence scanning.

(A) (B)

3 mm 3 mm

Figure 4.20: expression of DNA microarray (A) consisting of DNA Cy3-TF15myc-2His6 (green) and Cy5-GFP-2His6 (red) into protein microarray (B) by 12 µl of cell-free expression system RTS100 [145] in microfluidic flow cell of concept 1. Immunostaining of protein 1 by rabbit-a-myc and a-rabbit-Al532, of protein 2 by a-GFP-bio and Strep-Al633 lead to the same colour appearance of the proteins as the corresponding DNA (experiment performed by Oda Stoevesandt, BBT Cambridge, 2010; flow cell provided by Juergen Burger, AG Roth). Distortions of protein spots by leackage (orange region) and evaporation through outlet (purple region).

110 4.2 Cell-free protein microarray synthesis in flow cell

Each protein spot was expressed from the DNA spot and immobilized approximately opposite of it by diffusion. Fig. 4.20, B reveals distortions of the protein array pattern. They suppose to be caused by streams during the protein expression process – that is transcription of the immobilized DNA into mRNA in solution, and translation of the mRNA into protein in solution. Both the mRNA and the proteins could have drifted laterally before immobilization due to evaporation caused by leackages or, less likely by temperature gradients. While providing the proof-of-concept for cell-free protein arraying in microfluidic devices, some deficiencies with this concept became obvious:

(1) The self-adhesive polyester film affected the epoxy-coating of the glass slide, and reduced the immobilization of DNA.

(2) The holder has to be closed very firmly to properly seal the incubation chamber.

(3) Bubble-free priming of flow cell is a challenge and evaporation has to be avoided for prevention of smears by convection.

(4) Drilling holes into glass slides for top priming is rather cumbersome and lateral priming is tricky to handle, both not appropriate for daily lab work.

4.2.2 Concept 2

For the elimination of these deficits we devise flow cell concept 2. This concept is based upon the idea of creating three-dimensionally structured PDMS slides. The microfluidic gap is now an integrated part of the PDMS slide which carries the DNA microarray. The microfluidic gap is directly sealed by the Ni-NTA coated glass slide being protein capture surface and finally protein microarray matrix.

Figure 4.21: flow cell concept 2 according to chapter 3.1.2; microfluidic gap is part of a 3-dimensionally structured PDMS slide, which itself carries the DNA microarray. By sealing this gap with the protein capture glass slide an incubation chamber for the DNA to protein expression is realized. The chamber geometry can be variated in 3 dimensions allowing for priming control and protein expression (efficiency) analyses.

Most convenient priming can be realized by simply rolling the PDMS slide onto the protein capture glass slide after having dispensed manually a drop of ~ 20 µl of cell-free system onto

111 4.2 Cell-free protein microarray synthesis in flow cell the middle of the glass slide. The PDMS slide seals itself purely by adhesion. The protein expression and immobilisation takes place at 37 °C for 90 minutes with RTS100 [145] or at 37 °C for 30 minutes with EasyXpress [25]. Thereafter the sandwich is disassembled, the protein slides washed and immunostained.

(A) DNA (B) proteins

1 mm 1 mm

Figure 4.22: (A) erDNA (A07) manually spotted in 2 volumes with 30 / 15 / 15 x 106 dsDNA/spot onto PDITC coated PDMS slide. (B) Proteins expressed from DNA spots by cell-free system EasyXpress for 30 minutes at 37 °C and immobilized onto Ni-NTA coated glass slide [146]. First results without leackages and evaporation streams.

In line with the manufacturer’s specifications for both systems, the EasyXpress system allowed shorter reaction times than the RTS100 system, however, background fluorescence, likely caused by unspecific binding of kit components to the capture slides (seen as haze in Fig. 4.22, B), was higher with the EasyXpress kit.

4.2.3 Concept 3

According to chapter 3.1.3 the design of concept 3 creates a microfluidic gap by 2 glass slides and a microfluidic spacer, such both arrays are immobilized with rather versatile surface coatings. The spacer needs to be rather sophisticated for allowing lateral priming of the microfluidic gap while ensuring complete leak tightness. For basic analyses priming of the (B) proteins microfluidic gap (height ~ 60 µm) with ~ 20 µl of cell-free system (EasyXpress) was realized manually by pipette laterally. After incubation at 37 °C for 25 minutes proteins were expressed and immobilized, finally immunostained:

(A) DNA (B) proteins

1 mm 1 mm

6 mm Figure 4.23: (A) erDNA (A07) manually spotted in 2 volumes with 30 / 15 / 15 x 106 dsDNA/spot onto GOPTS coated glass slide. (B) Proteins expressed from DNA spots by cell-free system EasyXpress for 25 minutes at 37 °C and immobilized onto Ni-NTA coated glass slide [146]; microfluidic gap of heigt ~ 60 µm.

112 4.2 Cell-free protein microarray synthesis in flow cell

4.2.4 Benchmarking of concept 2 with DAPA

Joint experiments with Oda Stoevesandt: after having realized 3 concepts for microfluidic protein microarray printing we next compare our most versatile system according to concept 2 with the DAPA membrane setup. We choose 2 types of PDMS flow cells with width of 8 mm and 15 mm, both with a height of ~ 60 µm. We produce 3 flow cells of each type, all surface coated for the DNA immobilization with PDITC at IMTEK in Freiburg. All protein expressions are performed in the laboratory of BBT Cambridge together with Oda Stoevesandt, the DNA arrays consisting of template p53 and template notch are spotted in chess pattern with shortest distance of 1 mm by GeSIM Nano-Plotter 2.1. For the comparison we start with 3 DAPA expressions using the cell-free expression systems EasyXpress and RTS100:

(A) (B) (C)

Figure 4.24: DAPA expressions of DNA templates p53 (green) and notch (red). 3 expressions in sequence A (with RTS 100, 3 hours), B (with RTS 100, 3 hours) and C (with EasyXpress, 1.5 hours). Experiment A reveals rather big regions with no proteins due to dried membrane. Experiment B looks rather well with some local inhomogeneities. Experiment C reveals haze using EasyXpress. DNA was spotted onto epoxy coated glass slides, Nexterion Ni-NTA slides were used as protein capture surfaces. For visualization the proteins have been immunostained after expression / immobilization with Cy3 labelled antibodies (green) and Cy5 labelled antibodies (red).

Missing spots illustrate typical problems introduced by local inhomogeneities of the membrane, missing regions especially in experiment A reveals the sensitivity of the membrane towards drying out while assembly or incubation. Experiment C hints to the difficulty of handling the membrane, while disassembly the membrane dropped onto the slide.

We continue with protein expression in the PDMS flow cells of width 8 mm and 15 mm, again with above mentioned cell-free systems:

113 4.2 Cell-free protein microarray synthesis in flow cell

(A) (B) (C) (D) (E) (F)

Figure 4.25: Expressions of DNA templates p53 (green) and notch (red). 6 expressions in sequence: A + D with RTS 100, 1.5 hours; B + E with EasyXpress, 0.5 hours; C + F with RTS 100, 1.5 hours, cooled during disassembly. A shows rather big streams; B some haze but otherwise a proper protein spot pattern; C, D, E (small width) show no haze but significant inhomogeneities; F is similar to A but reveals lower degree of haze. DNA was spotted onto epoxy coated glass slides, Nexterion Ni-NTA slides were used as protein capture surfaces. For visualization the proteins have been immunostained after expression / immobilization with Cy3 labelled antibodies (green) and Cy5 labelled antibodies (red).

The benefits of protein transfer without a membrane became apparent through a reduction of incubation times by factor 2 with RTS100 or even 3 using EasyXpress. The microfluidic system results reveal streams (possibly due to evaporation / convection, e.g. caused by temperature gradients) and spot pattern distortions as negative effects however, prove the system to be capable of realizing rather clear protein ‘copies’ of the DNA microarray (Fig. 4.25, B, C, D). Let’s compare the best result of each method, these are arrays Fig. 4.24, C of DAPA and array Fig. 4.25, B of the microfluidic system:

(A) (B)

Figure 4.26: comparison of protein microarrays: A realized by DAPA (Fig. 4.24, C) and B the microfluidic system (Fig. 4.25, B) - DAPA shows violated areas (white arrows) and haze (red arrow); the microfluidic system reveals an irregularity (white arrow) possibly caused by an air bubble and a haze or drift at the entrance / exit (red arrow).

Both protein arrays are of rather good, similar quality. Deficits reveal potential for improvements, the microfluidic system creates the protein microarray 3 times faster with only 20% consumption of cell-free system in comparison to the DAPA.

114 4.2 Cell-free protein microarray synthesis in flow cell

4.2.5 Evaluation of concepts

Concept 1:

Pro

 Established proof-of-concept with rudimentary means.  Simplest incubation chamber o laser-cut polyester adhesive o epoxy-coated glass slide for DNA array o Ni-NTA coated glass slide for protein array

Con

 Self-adhesive foil lacks robustness o Sealing insufficient o Thickness >= 100 µm o Adhesive possibly inhibits DNA immobilization

Concept 2:

Pro

 Functional, good sealing  Variable chamber height down to possibly few microns  3 dimensional structures for fluid control possible  Coating (e.g. PDITC) easy to apply  Rapid (low-cost) replication by moulding

Con

 Not suitable for all surface coatings (minor, as DNA binds well to PDITC)

Concept 3:

Pro

 Functional  Variable chamber heights down to ~ 20 µm  Suitable for 2 glass slides with arbitrary coatings

Con

 Sealing rather difficult

115 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3 Molecule-protein interaction analysis by reflectometric interferometry

According to chapter 3.3 several prototypes of automated setups for the label-free detection of molecule-protein interactions in microfluidic flow cells have been realized. Basic physical evaluations have been made with the ‘iRIf prototype for proof-of-concept’. Finally the elaborated system described in chapter 3.3.3 resulted. This system is evaluated considering both the technical-physical aspects as well as the abilities of measuring molecule-protein interactions of biological impact.

4.3.1 Basic evaluations of iRIf method

4.3.1.1 Validation of LED wavelength with IF slide [5] With the standard RIfS system from Biametrics GmbH / Tübingen, spectral, non-imaging measurements were performed in order to verify the expected spectral behavior, such ensuring the selection of the optimal LED wavelength. We compared IF glass slides and standard microscope slides. The binding of Streptavidin-Cy5 onto a bBSA spot was measured with the flow-through method [106] as described in chapter 2.6. bBSA (20 μl) was immobilized on the whole area by physisorption (incubation for 1 h).

Figure 4.27: spectral analysis of the Strep-Cy5 binding to biotinylated BSA. The graph on the left shows the relative change of intensity as end-point measurement throughout the spectrum of white light. The graph on the right shows the change of relative intensity at wavelength of 530 nm during the experiment. [5]

This experiment reveals that the LED with wavelength 530 nm is suitable for IF glass slides. For reasons of comparison Cy5 is adequate for label-free iRIf measurements and end-point fluorescence measurements.

116 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.1.2 Analysis of LED homogeneity within field of view [5] For the prototype setup the available camera pco.1600 has a sensor size of 12.2 x 9 mm² however, our lens system with a zoom factor of ~ 0.5 was designed for a sensor of 8.8 mm x 6.6 mm. Therefore outside of this area there are distortions and non-homogenous illumination expected. For analysis a glass slide without flow cell is assembled to the holder, a white background outside of the focal plane in the back should help to generate a homogeneous background. An image is then taken with adapted exposure time of 10 ms. The illustration in false colors identifies differences in illumination (Fig. 4.28). The field of

view is rather homogeneous in the centre area but not in the outer regions.

13.2 mm

17.6 mm

Figure 4.28: variation of illumination over the whole sensor surface of camera pco.1600. The rectangle shows the size of the object supported by the lens system lens with zoom factor 0.5; due to distortions even within that size especially in one corner there are significant inhomogeneities. [5]

4.3.1.3 Analysis of spatial homogeneity of iRIf measurement [5] Aside the illumination other aspects could affect the homogeneity of the binding signal. For analysis 100% bBSA was spotted by PipeJetTM [152] onto GOPTS functionalized IF glasses, 1 ml of Strep-Cy5 was flow-injected to the array at a speed of 100 µl/min leading to the following iRIf quotient pics sequence:

(A) (B)

117 4.3 Molecule-protein interaction analysis by reflectometric interferometry

(E)

Figure 4.29: Streptavidin binding to bBSA spots immobilized via GOPTS onto IF slides detected by iRIf; picture sequence shows increasing signal intensities from A to E with increasing times having passed since reference picture, 2 / 4.5 / 6 / 8 / 13 minutes. [5]

During the experiment, the spot intensities vary to a certain degree. Especially the spots in the middle of the channel and close to the inlet increase faster in intensity. Finally however, all spots reach an equilibrium. Evaluation of spots in a horizontal and a vertical line gives

deeper insights into spot positions bias to interactions (Fig. 4.30 and Fig. 4.31).

relative intensity [a.u.] intensity relative

spot no.

Figure 4.30: horizontal uniformity analysis; evaluation of the spots shown on the left for the iRIf images in Fig. 4.29, E. The error bars show the standard deviation of the homogeneous area in the middle of each spot. The rings (donuts) with higher non- uniform intensities on the outside of every spot are neglected. [5]

118 4.3 Molecule-protein interaction analysis by reflectometric interferometry

spot no. spot

relative intensity [a.u.] Figure 4.31: vertical uniformity analysis; evaluation of the spots shown on the left for the iRIf images in Figure 4.29, E. The flow from bottom to top causes an increasing depletion of the concentration from bottom to top. [5]

Results prove that spot positions affect the dynamic binding behaviour significantly in this arrangement. Presumably, inhomogenous flow velocities, especially higher flow rates along the middle of the cell and reagent concentration depletion from inlet towards outlet are main affects. Nevertheless, detection homogeneity is proven by this setup, as the equilibrium can be reached at the end-point according to the rather straight lines in both horizontal (Fig. 4.30) and vertical (Fig. 4.31) direction after 13 minutes.

4.3.1.4 Spatial resolution analysis Elastomeric stamps [153] with feature sizes between 1 μm and 100 µm allowed to determine the total spatial system resolution of the camera based detection system in comparison to the fluorescence microarray scanner [154]. Stamped spots of bBSA are arranged in hexagons, strep-Cy5 was flow-injected.

119 4.3 Molecule-protein interaction analysis by reflectometric interferometry

Figure 4.32: comparison of resolution of iRIf (pco.1600, top) and microarray fluorescence scanner ( [154], bottom) measurements. The numbers marked in orange indicate the spot diameters. [5]

Results reveal that the iRIf method detects feature sizes smaller than 50 μm, but not 15 μm spots (Fig. 4.32, top); fluorescence scanning depicts feature sizes down to 15 μm. The resolution of the system can also be approximated mathematically. The smallest feature size

(xmin) or the highest spatial frequency (1/xmin) of the setup can be calculated using the

Nyquist theorem, without regarding the quality of the lens. The size of a sensor pixel (xPixel) and the magnification (m) of the lens are the only parameters needed – e.g. for pco.1600 camera based system:

2∗ 푥 2∗ 7.4 µ푚 푥 = 푃𝑖푥푒푙 = = 29.6 µ푚 (4.1) 푚푖푛 푚 0.5 Lens specification shows a spatial frequency at the sensor of

1/xLensSensor = 30 linepairs/mm (4.2) resulting in a resolution at the object of

xLensObject = xLensSensor/m = 66 μm (4.3)

The real resolution is expected to be somewhere between the minimum feature size supported by the sensor and the feature size calculated for the worst case of the lens. Calculations correspond with the resolution measured by iRIf rather good. Feature sizes of at least 50 μm are both visible and theoretically possible.

120 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.1.5 Comparison and enhancement of optical detection systems Next we verify total resolution of the components camera and lens as major elements of the optical detection system by a PS20 Universal Calibration Slide [155]. Applying the pco.edge camera system with the lens S5LPL1005 for the enhanced iRIf setup improves the resolution in comparison to the pco1600 camera system with lens S5LPL2060:

2∗ 푥 2∗ 6.5 µ푚 푥 = 푃𝑖푥푒푙 = = 14.1 µ푚 (4.4) 푚푖푛 푚 0.92 The calculated object resolution could be doubled with the pco.edge, comparison of the pictures of the calibration slide according to Fig. 4.33 revealed a total increase of resolution by ~ 5/3 including the lens systems affects.

Figure 4.33: resolution comparison of pco.1600 camera / lens S5LPL2060 (left) and pco.edge camera / lens S5LPL1005 with PS20 Universal Calibration Slide (1 small dash in scale = 100 µm). The resolution increased by factor 5/3 from ~33 µm/pixel (pco.1600) to ~20 µm/pixel (pco.edge).

4.3.1.6 Comparison of different reflective layer systems [5] Next we analyse the sensitivity of different glass slide layer systems in combination with appropriate LED’s. These slides were GOPTS activated, spotted with a dilution series of bBSA and primed by flow-injection with Streptavidin-Cy5. The IF glass was measured with a green LED (530 nm), the Goethe glass with a blue LED (465 nm). The end point intensity changes are depicted in Fig. 4.34. Results show that Goethe glass with blue LED leads to highest intensity changes.

121 4.3 Molecule-protein interaction analysis by reflectometric interferometry

non-coated glass, blue LED non-coated glass, blue LED

non -coated glass, green LED IF glass, green LED

Goethe glass, blue LED relative intensity [a.u.] intensity relative

bBSA / BSA conc. [%] Figure 4.34: comparison of glass slides with different reflective layer systems. Experiment performed with bBSA array of different concentrations primed by flow- injection with streptavidin-Cy5. Averaged intensity changes of four different experiments plotted. Goethe glass shows highest intensity changes. [5]

4.3.1.7 Fluorescence scanning compared to iRIf [5] Comparing iRIf measurements with fluorescence scans should affirm the linearity of the iRIf method within a certain range. Fluorescence measurement data depicted in all graphs is

taken from experiment with glass slide, green LED.

fluorescent signal fluorescent signal

IF, green LED Goethe, blue LED

[a.u.]

fluorescence

intensity

intensity relative intensity [a.u.] intensity relative [a.u.] intensity relative bBSA / BSA conc. [%] bBSA / BSA conc. [%]

fluorescent signal fluorescent signal

glass, blue LED glass, green LED [a.u.]

[a.u.]

fluorescence

intensity

intensity relative intensity [a.u.] intensity relative [a.u.] intensity relative bBSA / BSA conc. [%] bBSA / BSA conc. [%]

Figure 4.35: comparison of iRIf with reference fluorescence scanning measurements; exposure time of 200 ms. [5]

For all glasses, the differences in signal intensity changes between fluorescence and iRIf measurements are rather small. The direct correlation between the two signals is presented in Fig. 4.36. These graphs reveal the basic linearity of the correlation between fluorescence scanning and the iRIf measurements however, at high signal intensities curves seem to bend.

122 4.3 Molecule-protein interaction analysis by reflectometric interferometry

That could be caused by quenching effects at high fluorescence signals, which cause a decrease of the fluorescence signal but not of the iRIf signal. For low intensities the signal of the Goethe glass is the only one not converging to zero. This effect is puzzling, but can be

caused by the drying and swelling of the covalently bound and dried BSA.

relative intensity [a.u.] intensity relative

fluorescence intensity [a.u.]

Figure 4.36: correlation between fluorescent signal (at 200 ms) and iRIf measurement [5]. Especially Goethe glass with blue LED points to quenching effects.

4.3.1.8 Spot homogeneity analysis Based upon a dilution series of spotted bBSA the Streptavidin-Cy5 signal curves along a line of spots of equal concentrations are depicted by ImageJ plugin. Signal curves of 100% bBSA spots have rather steep flanks, 50% bBSA spots are quite inhomogenous.

bBSA 50% bBSA 100%

50% bBSA

100% bBSA [a.u.] intensity relative

1 2 3 4 5 6 7 8 9 10 Figure 4.37: (left) signal intensities of Streptavidin-Cy5 having bound to dilution series of spotted bBSA (bottom to top: 100/ 50/ 20/ 10/ 5/ 2/ 1/ 0.5). (right) intensities along horizontal lines through 50% (blue) and 100% (purple) bBSA spots showing spot morphology and intensity homogeneity within each spot; determined by ImageJ/Plugins/Compile and Run/GetOneRow.java.

123 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.2 System noise evaluation of iRIf setup

4.3.2.1 Classification of noise sources The total setup can be split in major components - optical, fluidical, mechatronical, most of them are biased to influence the SNR of the total system. There are dynamical noise sources occurring systematically or randomly during the measurement, and static ones affecting the homogeneity and such the comparability of the measurement.

Table 4.3: noise sources and their classification. Estimation of impact with high (++), medium (+), low (-); values in brackets indicate potential of noise source. No. Component Subtype Noise source Noise type Impact Counteraction

1 Optical LED Fluctuation of light Dynamic, systematic - (+) Improved LED emission system

2 Optical Camera Read-out noise, Dynamic, systematic + Improved dark current camera

3 Optical Camera shot noise Dynamic, random + (++) Spatial/ temporal averaging

4 Optical Lens Distortion Static + (++) Improved lens

5 Optical Slides Inhomogenic Static - Quality reflective layers standards

6 Optical Light Environment Dynamic, random - (+) Shielding scatter

7 Fluidical Flow Oscillations of Dynamic, systematic + Stiffening cell elastic elements – elements; tubes, PDMS slides fluidic protocol

8 Mechatronical Valves, Inherent vibrations Dynamic, systematic - Damping pumps during movements

9 Mechatronical Camera Cooling fan Dynamic, systematic - Water cooling / vibrations damping

10 Mechatronical Base Surrounding Dynamic, random - Damping vibrations

124 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.2.2 LED power stability analysis The optical multimeter OMM-6810B [156] from the Lab of Microoptics (University of Freiburg / IMTEK) was used to identify any optical power fluctuation or temperature drift during a time similar to standard measurement times, that is < 4 hours. The OMM was set to a wavelength of 530 nm, the LED was put directly in front of the sensor head in order to eliminate stray light effects.

40 35

] 30

mW 25

2 -

10 20 Series2 15 Series1

power power [ 10 5 0 0 20 40 60 80 100 120 140 160 180 200 220 240 time [min] Figure 4.38: LED power changes over time, especially during the first hour after having switched on the LED. The system is close to steady-state and such ready for measurements after 90 minutes; continuous LED power drifts of ~ 0.15% / hour should be considered for very precise measurements. In opposite to the graph depicting the modulus of the measured LED power change, the power is decreasing by time; hence the signal of the iRIf data of an unsettled system would decrease by this effect. Measuring series 2 has been shifted by ~15 minutes (vertical red line), as it starts as unsettled system due to having switched on/off the LED during the first 15 minutes 4 times.

Measurement series 1 shows data taken from the LED system after a long period of > 24 hours settling time to room temperature. During the first 20 minutes we observe a high gradient, presumably the temperature drift of the LED / LED controller. The data of series 2 is taken from an unsettled LED system having been switched on/off for 4 times during 15 minutes (vertical red line) before the measurements were logged. The measured data is shifted in time by 15 minutes, in power by 21 mW in order to better compare the LED systems state to measurement series 1. In opposite to the graph depicting the modulus of the measured LED power change, the power is decreasing by time; hence the signal of the iRIf data of an unsettled system would decrease by this effect. The initial LED power of series 1 was 7.18 mW. A total power change of 0.34 mW as measured in series 1 therefore means a change of ~ 5%, that would be more than standard signal changes (e.g. Streptavidin onto 20% bBSA) measured by iRIf however, after warming up for 1.5 hours the change decreases to ~ 0.01 mW / h, that is ~ 0.15%. For very precise measurements even these tiny drifts should be considered and compensated.

125 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.2.3 Camera noise sources [5] Detailed information to the inherent noise sources read-out noise, dark current and possibly temperature drifts give the technical manuals for the pco.1600 [133] and the pco.edge [135]. Cooling fan vibrations could be extracted from measurements by Fourier analyses. Random noise as shot noise is analysed now. First the signal of single pixels is evaluated for 5000 frames. A probability density function for the signal of one pixel and the comparison of signals of different pixels are used to find the origin of the noise. In Fig. 4.39 the signal intensity of three adjacent pixels is depicted for 20 frames, revealing, that the

fluctuations are random, therefore not caused by any static or systematic noise sources.

intensity [a.u.] intensity pixel 1 pixel 2

pixel 3 relative

frame no. Figure 4.39: shot noise analysis - signal behavior of three adjacent pixels for 20 measurements. [5]

Therefore, the measured noise is now compared with the noise resulting of the probabilistic behavior of photons, the so-called shot noise. Initially, the amount of photons as a function of the intensity value of the camera has to be calculated. This is done with the conversion factor 2.1 for the pco.1600 [133]. By forming a probability density function of the amount of photons for one of the pixels, one can show how often a specific number of photons is measured during the 5000 measurements. This function can be compared to the theoretical probability density distribution of the shot-noise of light with the same amount of photons. For large amounts of photons, the Poisson distribution which describes the density distribution of photons can be approximated with a normal distribution. For a total number of measurements ‘a’ with 푛̅ as the mean amount of photons, the amount of measurements that have a result between x − 1/2 and x + 1/2 is given by:

1 푥−푛̅ 푎 − ( )² 푓(푥) = ∗ e 2 √푛̅ (4.5) √2휋푛̅ The resulting function for a = 5000 and 푛̅ = 19408 can be directly compared to the measurements:

126 4.3 Molecule-protein interaction analysis by reflectometric interferometry

number ofcounts number measurements with equal equal with measurements

photons per measurement

Figure 4.40: probability density function for 5000 measurements compared with density function of the theoretical shot noise. [5]

As illustrated in Fig. 4.40 the calculated and measured distributions fit well together. The standard deviation of the measurement with 131 photons is quite close to the calculated value of 139 photons. Having already verified that shot noise is the major noise source, one can calculate the signal-to-noise ratio in dependence of the amount of averaging [5]. Assuming a maximum signal change of 0.3% while using the IF glass and 0.1% for normal

microscope slides, one can calculate the SNR for any kind of averaging (Fig. 4.41).

SNR

number of averaged frames

Figure 4.41: calculated SNR’s for bBSA Streptavidin-Cy5 assay and initial intensity of 10000; measured with pco.1600, max. intensity value of 16384. [5]

As the PCO camware only allows powers of two for the averaging of frames, the averaging of 128 frames per image (SNR of 4.9) or 64 (SNR of 3.5) is recommened while applying IF glass, for slow processes an averaging of 256 increases the SNR to 6.9.

4.3.2.4 Noise by non-coherent reflections One of the inherent noise sources is light scatter, especially non-coherent reflections by the flow cell background decreases the SNR additionally. The background grey scale values of three different flow cell types were determined for different exposure times. We applied

127 4.3 Molecule-protein interaction analysis by reflectometric interferometry exposure times of 5, 10, 15, 20, 22.5, 25, 27.5, 28 and 30 ms. For analysis the flow cells were primed with PBS however, measurements were made at flow rate 0 µl/min for elimination of disturbances by liquid flow. Flow cells according to chapter 3.1 (concepts 1 to 3) were

compared. Each flow cell was measured 3 times for statiscal reasons.

value

mean grey scale scale meangrey concept 1 (PMMA flow cell) concept 2 (PDMS flow cell) concept 3 (booklet type flow cell)

exposure time [ms] Figure 4.42: grey scale values of raw images of different flow cell types. Flow cell according to concept 3 reveals strong background reflections. Measures taken with pco.1600 in prototype of automated system (chapter 3.3.2).

Table 4.4: grey scale values of different flow cells / background types at exposure time of 20 ms [8]. Measures taken with pco.1600 in prototype for automated system. (chapter 3.3.2). Flow cell type Background Mean grey scale +/- stdev concept 1 PMMA 12773 +/- 535 concept 2 PDMS, red 11929 +/- 58 concept 2 PDMS, black 11823 +/- 182 concept 3 Glass slide 16383 +/- 0.1311 (sat.) concept 3 IF slide 16383 +/- 0.0116 (sat.) concept 3 IF slide + PDMS on rear side 13790 +/- 256 concept 3 IF slide + black PMMA slide 12698 +/- 292

The background grey scale values for flow cells according to ‘concept 1’ and ‘concept 2’ are in the same range. Nevertheless, the standard deviations depicted for flow cells of ‘concept 2’ are about ten times smaller than for ‘concept 1’ - the background of the red PDMS flow cell (‘concept 2’) is most homogenous and therefore most suitable. The reflection at the background of flow cells ‘concept 3’ is almost twice as high; the standard deviation is high too – hence flow cells according to ‘concept 3’ are not suitable for precise measurements and need to be redesigned considering the background: for preliminary tests a red PDMS inlay attached to the rear side of the glass slide or a black plastic slide instead of a glass slide were used. Both variants resulted in lower degrees of reflections (table 4.2). ‘Concept 3’

128 4.3 Molecule-protein interaction analysis by reflectometric interferometry

background variations are only suitable for preliminary tests and need to be better incorporated into the flow cell design.

4.3.2.5 Flow cell comparison Flow cells according to ‘concept 2’ and ‘concept 3’ are compared; ‘concept 3’ with a black plastic slide lowering the background reflections. Referencing with the fluorescent signal of each slide we compare the iRIf data acquired of a bBSA Streptavidin assay, measured with averaging of 128 frames. Flow cell ‘concept 2’ reaches a relative signal intensity being 30%

higher than flow cell type ‘concept 3’.

concept 2

concept 3

relative intensity [a.u.] intensity relative

bBSA concentration [%] Figure 4.43: comparison of iRIf binding signals using flow cell types ’concept 2’ and ’concept 3’; ’concept 2’ results in higher intensities. Measures taken with pco.1600 in prototype for automated system (chapter 3.3.2).

4.3.2.6 Influence of exposure time to SNR For analysis we again choose a bBSA Streptavidin binding experiment, having bBSA spotted in concentrations of 100%, 50%, 20%, 10%, 5%, 2% and 1%. Flow cell type ‘concept 2’ with red PDMS was selected as it was proven to be most suitable concerning background reflectivity and homogeneity. The results for all exposure times ranging from 5 to 25 ms are in the same range. Only for the highest exposure time of 27.5 ms the resulting change of the signal decreases largely due to saturation of the camera wells.

129 4.3 Molecule-protein interaction analysis by reflectometric interferometry

bBSA conc. 100% 50% 20% 10% 5% 2%

1% relative intensity [a.u.] intensity relative

exposure time [ms] Figure 4.44: resulting signal changes for different exposure times and different bBSA concentrations. Measures taken with pco.1600 in prototype system (chapter 3.3.2).

Error bars indicate rather good reproducibility such that we can conclude that exposure times in good distance to full well capacity of the camera are suitable and do not directly affect the measuring result. Nevertheless, an exposure time of 20 ms is suggested to be the standard exposure time due to an enhanced SNR according to shot noise.

4.3.2.7 Noise analysis of enhanced iRIf setup Quantification of noise in elaborated automated setup (chapter 3.3.3) with fan cooled pco.edge 5.5 is based upon assay with Streptavidin binding to 3 concentrations of spotted bBSA – 100%, 50% and 20%.

Spotted bBSA binding Streptavidin 1.004

1.003

1.002

1.001

1 relative intensity [a.u.] intensity relative

0.999 3500 4000 4500 5000 5500 6000 time [s] Spot Nr. 1 Spot Nr. 2 Spot Nr. 3 Spot Nr. 4 Spot Nr. 5 Spot Nr. 6 Spot Nr. 7 Spot Nr. 8

Figure 4.45: flow-injection measurement with Streptavidin binding to bBSA spots, detected by iRIf system consisting of fan cooled pco.edge 5.5, GOPTS coated IF glass slide and red PDMS flow cell of ’concept 2’. Bunches of curves belonging to each concentration are clearly visible however, spot 6 reaches lower saturation value than other 100% spots – that is caused by local surface inhomogeneity (dark stripe within spot).

130 4.3 Molecule-protein interaction analysis by reflectometric interferometry

We now take the 50% bBSA spots and analyse them from time 4800 seconds to time 6000 seconds, being referenced with their values at time 4800 seconds (Fig. 4.46). Baseline drift and baseline noise are calculated according to chapter 2.7.2.4 using Origin with exposure time of 20 ms, 64 frames being averaged, and total measurement time of 20 minutes:

baseline drift: < 3 * 10-3 ‰/min

baseline noise: < 3 * 10-9 RSS/min

relative intensity [a.u.] intensity relative

time [s] Figure 4.46: zoom of 50% curves (spots 3, 4 and 5) from Fig. 4.45 normalized with their values at time 4800 s resulting in a baseline drift of < 3 * 10-3 ‰/min and a baseline noise of < 3 * 10-9 RSS/min.

4.3.2.8 Noise damping by spatial averaging Spatial averaging is an essential and appropriate means to significantly reduce shot noise. We evaluate the impact of spatial averaging and introduce recommendations for minimal spot sizes being big enough for stable analyses. Analysis is made upon rectangular ROI’s with sizes beween 1 pixel and 1024 pixel resulting in standard deviations depicted in Fig. 4.47. The analysis is based upon bBSA Streptavidin assay during baselining from time 1300 s to 1600 s.

131 4.3 Molecule-protein interaction analysis by reflectometric interferometry

0.001

0.0008

0.0006

0.0004

0.0002 stddev intensity [a.u.] intensity stddev

1 2 4 1 1 1 2 2 2 4 8 8

16 64 16 32 32 64

128 128 256 512 1024 ROI's of different sizes [pixel no.]

Figure 4.47: data acquired with fan cooled pco.edge 5.5, relation of ROI size to standard deviation of spatially averaged ROI intensities. Data resulted from bBSA streptavidin assay during baselining from time 1300 s to 1600 s with rectangular ROI’s of different sizes of 1 pixel to 1024 pixel; standard deviations reduce by increasing number of pixel per ROI.

Average values of standard deviations of signal intensities according to Fig. 4.47 result in function according to Fig. 4.48.

0.01

0.001 -0.467 stddevI = 0.0011np

0.0001 stddev intensity (a.u.) intensity stddev 0.00001 1 10 100 1000 number of pixel (np) Figure 4.48: noise reduces by ~1/pixel no.1/2, spot size of >200 pixel (~14x14 pixel or diameter ~16 pixel (~110 µm)) recommended.

Next we replace the pco.edge 5.5 with fan cooling (standard of enhanced iRIf setup according to chapter 3.3.3) by a water cooled pco.edge 5.5 for noise comparison. For analysis we again choose a bBSA Streptavidin binding assay with 11 spots in 3 concentrations: 4.5%, 15% and 50%. We focus the analysis on spot 5 (15% bBSA) in the middle of the flow cell, such that effects as lens distortions, inhomogenous excitations, geometrical affects by the flow cell or reagent depletion are mostly eliminated. We select ROI’s of size 1, 4, 16, 64, 256, 1024 pixel (Fig. 4.49):

132 4.3 Molecule-protein interaction analysis by reflectometric interferometry

Figure 4.49: experiment with water cooled pco.edge 5.5; quantification of noise based upon bBSA streptavidin assay with 4 columns of manually spotted bBSA (4.5%, 15%, 50%, 15%). Spot 5 (15% bBSA) is separated for analysis in 4x 1 pixel, 2x 4 pixel, 2x 16 pixel, 2x 64 pixel, 1x 256 pixel, (1x 1024 pixel not depicted, surrounds ROI 2 (256 pixel) in equal distances.

Considering each ROI as spot results in the binding curves of Fig. 4.50 – ROI’s consisting only of 1 pixel are not depicted due to high noise:

1.007 1.006

1.005 Spot Nr. 1 1.004 Spot Nr. 2 1.003 Spot Nr. 3 1.002 Spot Nr. 4 1.001 Spot Nr. 5 1

relative intensity [a.u.] intensity relative 0.999 Spot Nr. 6 0.998 Spot Nr. 7 0.997 Spot Nr. 8 600 800 1000 1200 1400 1600 1800 2000 time [s]

Figure 4.50: binding curves of ROI’s of spot 5 of Fig. 4.49 without ROI’s of size 1 pixel. Even with ROI sizes of only 4 pixel SNR would allow a binding evaluation.

The relation of standard deviations of signal intensities to the number of spot pixels is depicted in Fig. 4.51:

133 4.3 Molecule-protein interaction analysis by reflectometric interferometry

0.01

0.001 -0.426 stddevI = 0.0009np

0.0001 stddev intensity (a.u.) intensity stddev 0.00001 1 10 100 1000 number of pixel (np) Figure 4.51: data acquired with water cooled pco.edge 5.5: evaluation of curve Fig. 4.50 from time 1550 to 1750 seconds. ROI size of 1 pixel leads to standard deviation of signal intensities of ~ 10-3; value reduces to 5*10-4 for ROI sizes of 4 pixel; ROI sizes of > 256 pixel lead to standard deviations of < 10-4. Error bars of standard deviation of spots of equal size are not visible due to size of < 8*10-5 even for spots consisting of 1 Pixel only.

Comparing data of fan cooled (Fig. 4.48) with water cooled (Fig. 4.51) pco.edge inidcates that the water cooled camera does not reduce the standard deviations of spots relative intensities; spots consisting of equal pixel numbers have rather similar standard deviation values.

4.3.2.9 Noise of non-primed chamber Next we perform a noise analysis with a non-primed chamber, that is an empty flow cell with a glass air interface as outer boundary. Such the refractive index at the outer boundary changes from 1.33 of PBS (water) to ~ 1 of air hence, towards more reflectivity. Within this experiment series measures taken by the fan cooled pco.edge 5.5 and the water cooled pco.edge 5.5 are compared; PDMS flow cell and IF slide remained the same for both experiments, while green (530 nm) LED is used. 6 oval shaped spots of equal size are spread diagonal across the flow chamber from upper left to lower right corner, referenced to constant, global background.

134 4.3 Molecule-protein interaction analysis by reflectometric interferometry

1.000002 1.0000015 1.000001 Spot Nr. 1 1.0000005 Spot Nr. 2 1 Spot Nr. 3 0.9999995 Spot Nr. 4

0.999999 Spot Nr. 5 relative intensity [a.u.] intensity relative 0.9999985 Spot Nr. 6 0.999998 1200 1210 1220 1230 1240 1250 time [s]

Figure 4.52: measured by water cooled pco.edge 5.5 – total random noise is smaller than 3*10-6. Single spots reveal random noise of ~ 1*10-6.

1.000002 1.0000015 1.000001 Spot Nr. 1 1.0000005 Spot Nr. 2 1 Spot Nr. 3 0.9999995 Spot Nr. 4

0.999999 Spot Nr. 5 relative intensity [a.u.] intensity relative 0.9999985 Spot Nr. 6 0.999998 1200 1210 1220 1230 1240 1250 time [s]

Figure 4.53: measured by fan cooled pco.edge 5.5 – total random noise is slightly bigger than 3*10-6. Single spots reveal random noise of ~1.5*10-6.

Water cooled camera seems to reduce random noise by ~ 30%, further experiments need to confirm this result.

4.3.3 Bio-physical evaluations of elaborated iRIf setup

The IF slide spotted according to chapter 3.6.2, Fig. 3.58 we assembled with a PDMS flow cell (trapezoidal pillar type) and run the experiment with the following process protocol:

135 4.3 Molecule-protein interaction analysis by reflectometric interferometry

Table 4.5: parameters for flow-injection protocol; PBS (1x, pH 7.2) used as buffer, BSA with 5 mg/ml, Streptavidin in concentrations of 0.5 / 1 / 2 / 5 / 10 µg/ml. flow rate reagent name step no. [µl/min] priming time [s] Puffer 1 60 960 BSA 2 60 300 Puffer 3 60 300 Streptavidin-Cy5 4 60 300

With bBSA spots being incubated > 1 night at 4 °C in humid (1x PBS) environment, a sequence of experiments with Streptavidin-Cy5 concentrations of 0.5 / 1 / 2 / 5 / 10 µg/ml performed with the elaborated iRIf setup (chapter 3.3.3) lead to the following quotient pictures, reference taken after the PBS step 3:

0.5 µg/ml Streptavidin-Cy5 1 µg/ml Streptavidin-Cy5 2 µg/ml Streptavidin-Cy5

Slide 033 Slide 027 Slide 022

5 µg/ml Streptavidin-Cy5 10 µg/ml Streptavidin-Cy5

Slide 011 Slide 028

Figure 4.54: quotient pictures of experiments with different concentrations of Streptavidin-Cy5, protocol as listed in table 4.3, spotting pattern as depicted in chapter 3.6.2; performed with elaborated iRIf setup; fluorescence scans shown in Appendix A6.

Both the concentrations series of bBSA and of Streptavidin-Cy5 are clearly visible. Backgrounds are differing in brightness and homogeneity however, due to local referencing this affects the comparison not essentially. Quotient pics in jpg format allow to identify the spots via Matlab procedure: pics are first passed through median filter, then converted to binary pics with global thresholding; next spots are identified as oval shaped ROI’s, oval shaped local backgrounds are added to each spot; regions are exported as ImageJ ROI file (Fig. 4.55, A), then superimposed to quotient pic file (Fig. 4.55, B) in ImageJ; ROI’s could be amended or added manually.

136 4.3 Molecule-protein interaction analysis by reflectometric interferometry

(A) (B)

Figure 4.55: (A) spots identification by Matlab e.g. for slide 022 (2 µg/ml Streptavidin- Cy5) with oval shaped ROI’s and their local backgrounds; yellow rectangle surrounds 6.3 nl spots. (B) An ImageJ ROI file is created and superimposed to the quotient pic.

Next we apply a Java routine for the calculation of the binding curves of each spot. The results are saved as Excel file, spot numbers are added to the ROI file for assigning spots to correspondent curves. For the reason of visibility the analysis depicted is reduced to spots of volume 6.3 nl (Fig. 4.55, A, yellow rectangle).

(6x)

(12x)

(12x) relative intensity [a.u.] intensity relative

time [s]

Figure 4.56: binding curves determined by Java routine for spots of 6.3 nl volume of Fig. 4.55 with assay protocol according to table 4.3. Each set of spots with equal bBSA concentrations (50%, 15% and 4.5%) creates a set of curves. BSA binds to background, therefore it causes slight signal decrease.

According to table 4.3 PBS baselining for 960 seconds follows blocking with BSA (5 mg/ml) for 300 seconds - the curve goes down slightly as BSA binds to the background as desired.

137 4.3 Molecule-protein interaction analysis by reflectometric interferometry

After washing with PBS for another 300 seconds, 2 µg/ml of Streptavidin-Cy5 is flow-injected resulting in 3 sets of curves belonging to the 3 concentrations of bBSA spots - 50%, 15% and 4.5%. Each bBSA concentration step, that is 10/3 * the lower concentration, leads to an increase of ~ 0.4% of the relative signal intensity. Within each set of curves, that is within each bBSA concentration we have an intensity variation at the endpoint of ~ 0.05%, the variation for the lowest concentration of 4.5% is slightly higher.

Analysis of curve shifts due to spot positions is depicted in Fig. 4.57. Spots in proximity of each other should see the same flow rate as due to the very high ratio of flow cell width / flow cell height of 500 (width of ~15 mm / height of ~30 µm) the parabolic flow profile is expected only within the first 50 µm from the side walls [74]. However, we see significant binding delays of 2 to 12 seconds for spots in distance of 1 mm, despite a flow rate of 60 µl/min leading to a flow rate vector in the flow cell of 2.22 mm/s; for a spot distance of 1 mm that would be ~ 0.5 seconds only.

푣푠푦푠 푣푓푐 = /60 (4.6) 푓푐푤∗ 푓푐ℎ (A) with

vfc … flow vector in flow cell in mm/s

vsys … system flow rate in µl/min

fcw, fch … flow cell width and height in mm

(B) 1.0117 Spot Nr. 13 1.0097 Spot Nr. 14

1.0077 Spot time shift of successive spots at Nr. 15 1.0057 measuring level (red arrow) Spot 1.0037 Nr. 16 16 --> 18 3 seconds Spot relative intensity [a.u.] intensity relative 1.0017 13 --> 15 2 seconds Nr. 17 14 --> 17 12 seconds Spot 0.9997 Nr. 18 1630 1680 1730 1780 1830 1880 time [s] Figure 4.57: (A) zoom of Fig. 4.55, A; 6.3 nl spots of 50% bBSA (13-18) with their local backgrounds (43-48). Distance between spot pairs 16-18, 13-15 and 14-17 is 1 mm. (B) significant binding delays of 2 to 12 seconds for spots in distance of 1 mm point to e.g. depletion of the reagent, local inhomogeneities of flow cell geometry and surface coating.

Hence the binding curve delay as measured is not only caused by the flow vector - we mainly assume depletion of the reagent; inhomogeneous flow rate vectors due to local

138 4.3 Molecule-protein interaction analysis by reflectometric interferometry

inhomogeneities of flow cell geometry and surface coating could be further reasons. This result should be confirmed by a set of experiments analyzing each parameter.

Next we fit exponential functions according to chapter 2.5 to the data curves by a Matlab routine which visualizes the fitting result (Fig. 4.58, left column), depicts the signal derivative (Fig. 4.58, right column), and finally creates a chart with kobs values.

Figure 4.58: (left column) exponential functions fitted to the evaluable regions of the binding curves; (right column) corresponding derivations. From top to bottom: 4.5%, 6.3 nl; 15%, 6.3 nl; 50%, 6.3 nl - increasing concentration of the bBSA spot leads to a growing SNR; all fittings are well made.

According to chapter 2.5 the dissociation constant KD can be calculated from the association and dissociation rate contants ka and kd for different reagent concentrations. We choose

139 4.3 Molecule-protein interaction analysis by reflectometric interferometry reagent concentrations of 0.5 / 1 / 2 / 5 and 10 µg/ml Streptavidin-Cy5 in PBS. We select the 50% spots of the 6.3 nl row for the determination. Such we have 6 spots for each concentration, we repeat the experiment to get 2 data sets. Finally we get the dissociation rate constants 0.0048 s-1 and 0.0053 s-1, the association rate constants of 0.1097 M-1s-1 and 0.0745 M-1s-1, leading to an averaged KD value of 57 nM (Fig. 4.59 and table 4.6).

0.025 y = 0.1097x + 0.0048 R² = 0.8846 0.02

0.015 kobs1 y = 0.0745x + 0.0053 Kobs2

R² = 0.5108 kobs Linear (kobs1) 0.01 Linear (Kobs2)

0.005

0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 conc [µM]

Figure 4.59: determination of the rate constants by linear fitting of the kobs values for concentrations of 0.5 / 1 / 2 / 5 / 10 µg/ml Streptavidin bound to 6.3 nl, 50% bBSA spots (2x 6 spots of 2 experiments). The gradients of the linear fitings correspond with the association rate constants, the y intercepts with the dissociation rate constants.

Table 4.6: calculation of dissociation constant KD by division of rate constants kd / ka. Data sets according to Appendix Table A4.2.5.

KD [µM] kd [1/s] ka [1/µM*s] data set 1 0.04377 0.0048 0.1097 data set 2 0.07114 0.0053 0.0745 average data 0.05744 0.00505 0.0921

140 4.3 Molecule-protein interaction analysis by reflectometric interferometry

4.3.4 Biological applications with automated iRIf setups

The total system is aimed to analyze assays of biological impact. Beyond the purely technical and physical aspects therefore its suitability for some exemplary applications is tested e.g. biotinylated peptide Streptavidin assay, thrombin aptamer assay, serum immunization assay, general antibody antigen assay.

4.3.4.1 Antibody antigen assay [8] Performed with first prototype of automated setup (chapter 3.3.2): binding analysis is carried out with the following protocol according to table 4.7 - all analytes were flow- injected with a concentration of 1 μg/ml at a flow rate of 100 µl/min. For specificity control Streptavidin-Cy5 should bind to the biotinylated reagents, finally allowing an end-point measurement by fluorescent scanning for comparison. Arrays are spotted onto NHS-ester coated IF slides according to chaper 3.6.2, Fig. 3.59.

Table 4.7: fluidic protocol with αCRP2, CRP and αCRP1 followed by Streptavidin control after baselining and blocking with BSA. flow rate priming time reagent name conc. step no. [µl/min] [min] PBS 1x 1 100 10 BSA 50 mg/ml 2 100 10 PBS 1x 3 100 10 Streptavidin-Cy5 in PBS/BSA 5 µg/ml 4 100 5 PBS 1x 5 100 10 αCRP2 in PBS 1 µg/ml 6 100 5 PBS 1x 7 100 10 CRP (in 20 mM Tris-HCL) 1 µg/ml 8 100 5 PBS 1x 9 100 10 αCRP1 in PBS 1 µg/ml 10 100 5 PBS 1x 11 100 10 Streptavidin-Cy5 in PBS/BSA 5 µg/ml 12 100 5 PBS 1x 13 100 15

141 4.3 Molecule-protein interaction analysis by reflectometric interferometry

(A) (B)

intensity [a.u.] intensity

relative relative relative intensity [a.u.] intensity relative

time [s] time [s]

relative intensity [a.u.] intensity relative relative intensity [a.u.] intensity relative

time [s] time [s]

Figure 4.60: binding curves measured on reagents spotted to NHS-ester surfaces. (A) Bindings to 100% bBSA spots result in ~20 times higher signals than bindings to antibody or antigene spots. (B) bindings to CRP spots. (C+D) bindings to αCRP1 and αCRP2 spots. Curves of (A) are referenced after second PBS step at 1800 seconds, other curves are referenced after Streptavidin step at 2700 seconds [8]. Data acquired with prototype of automated system.

Streptavidin binds well to bBSA as positive binding control with a signal change of ~ 2%; following bindings of antibodies and antigenes are significantly lower but visible (Fig. 4.60, A). Biotinylated αCRP2 attaches to the bound Streptavidin, and itself slightly binds CRP followed by αCRP1 (Fig. 4.60, B). Bindings to spotted antibodies are weak and irregular, possibly antibodies are not immobilized well or denaturated (Fig. 4.60, C+D).

Next an antibody antigene assay with Streptavidin-GFP, HA-GFP and Streptavidin-mCherry antigens spotted onto PDITC coated IF glass slides, all in 2 concentrations (1:100 and 1:1000) is realized with the elaborated setup (chapter 3.3.3). Multi-step interactions with antibodies αHA, αStrep, αGFP-biotinylated and Streptavidin-Cy5 as positive control (all in concentrations of 5 µg/ml) could be visualized label-free by iRIf technology as depicted in Fig. 4.61; high antigene concentrations yield high intensity signals, low antigene

142 4.3 Molecule-protein interaction analysis by reflectometric interferometry

concentrations yield signals at the limit of detection. Streptavidin-Cy5 binds to bBSA spot as

positive control and to the biotinylated αGFP.

relative intensity [a.u.] intensity relative

time [s]

Figure 4.61: antigens Streptavidin-GFP, HA-GFP and Streptavidin-mCherry spotted onto PDITC coated IF glass slides in 2 concentrations (1:100; 1:1000) interact with antibodies αHA, αStrep, αGFP-biotinylated (each 5 µg/ml) and Streptavidin-Cy5 as positive control; initial BSA blocking is followed by baselining with HBS buffer; assay performed with elaborated prototype of automated system (chapter 3.3.3).

4.3.4.2 Thrombin Aptamer assay Flow-injection analysis was performed with the elaborated setup (chapter 3.3.3) using the protocol of table 4.8.

Table 4.8: protocol for aptamer thrombin binding analysis by flow-injection. flow rate reagent name conc. step no. [µl/min] priming time [min] BB(KC), pH 7.6 1x 1 32 30 BSA 5 mg/ml 2 32 15 BB(KC), pH 7.6 1x 3 32 15 Human thrombin 5 µg/ml 4 32 15 BB(KC), pH 7.6 1x 5 32 15 Streptavidin-Cy5 1 µg/ml 6 60 5 BB(KC), pH 7.6 1x 7 60 5

A first glimpse to the iRIf quotient pics (Fig. 4.62, spotting pattern according to chapter 3.6.2, Fig. 3.60) suggests that all 15 experiments lead to thrombin aptamer bindings of different 143 4.3 Molecule-protein interaction analysis by reflectometric interferometry intensities; experiment 6 was partly affected by a bubble, all others performed well. Background colour variations indicate that reflectivities vary, presumably caused by e.g. physical layer thickness variations of the glass slides, surface coating inhomogeneities.

1 2 3 4 5 6

7 8 9 10 11 12

13 14 15

Figure 4.62: quotient pics of thrombin binding to aptamer array (spotting pattern according to chapter 3.6.2, Fig. 3.60), reference pic of median from pic 2400 to 2430; final pic is median of 2900 to 2930; lookup table from 0.998 to 1.017. From left to right column concentrations series of 10 µg/ml (272 nM; 1+7+13), 6.7 µg/ml (181 nM; 2+8+14), 4.5 µg/ml (120.7 nM; 3+9+15), 3 µg/ml (80.5 nM; 4+10), 2 µg/ml (53.7 nM; 5+11), 1.33 µg/ml (35.8 nM; 6+12). Spots within orange rectangles (top, aptamer A13; bottom, aptamer A14; see Appendix A2.3) will we be further analysed.

Now we compare the association and dissociation curves depending on the thrombin concentrations, e.g. experiments 13 (272 nM), 14 (181 nM) and 15 (120.7 nM). The intensities at saturation decrease as expected from higher to lower thrombin concentrations; despite rather high variations of signal intensities among equal spots an averaged signal intensity decrease of factor ~ 1.5 can be determined, corresponding to the decrease of thrombin concentration.

144 4.3 Molecule-protein interaction analysis by reflectometric interferometry

1.002 experiment 13 6 14 1.0015

1.001 8 11 1.0005

1 7 12

relative relative intensity [a.u.] 0.9995 5 13

0.999 2 15 3400 3900 4400 4900 5400 time [s] 3 9

Spot Nr. 1 Spot Nr. 2 Spot Nr. 3 Spot Nr. 5 1 10 Spot Nr. 6 Spot Nr. 7 Spot Nr. 8 Spot Nr. 9 Spot Nr. 10 Spot Nr. 11 Spot Nr. 12 Spot Nr. 13 Spot Nr. 14 Spot Nr. 16 4 16

1.002 experiment 14 1 9 1.0015

1.001 2 10 1.0005

1 3 11

relative relative intensity [a.u.] 0.9995 4 15

0.999 5 12 3400 3900 4400 4900 5400 time [s] 6 13

Spot Nr. 1 Spot Nr. 2 Spot Nr. 3 Spot Nr. 4 7 14 Spot Nr. 5 Spot Nr. 6 Spot Nr. 7 Spot Nr. 8 Spot Nr. 9 Spot Nr. 10 Spot Nr. 11 Spot Nr. 12 Spot Nr. 13 Spot Nr. 14 Spot Nr. 15 Spot Nr. 16 8 16

145 4.3 Molecule-protein interaction analysis by reflectometric interferometry

1.002 experiment 15 2 9 1.0015

1.001 1 11 1.0005 6 14 1

relative relative intensity [a.u.] 0.9995 7 15

0.999 3 12 3400 3900 4400 4900 5400 time [s] 4 13

Spot Nr. 1 Spot Nr. 3 Spot Nr. 4 Spot Nr. 5 5 10 Spot Nr. 6 Spot Nr. 7 Spot Nr. 8 Spot Nr. 9 Spot Nr. 10 Spot Nr. 11 Spot Nr. 12 Spot Nr. 13 Spot Nr. 14 Spot Nr. 15 Spot Nr. 16 8 16

Figure 4.63: binding curves of thrombin to spotted aptamers; thrombin from top to bottom in concentrations of 272 nM (13), 181 nM (14) and 120.7 nM (15). 272 nM concentration did not reach complete saturation within 15 minutes. Spots are autodetected by Matlab routine (chapter 3.4.2), spot numbers therefore not ascending.

Next we determine the KD values of aptamer A14 at various positions (Fig. 4.64): we plot the association and dissociation curves (Fig. 4.65), apply fitting methods according to chapter 2.5 and calculate dissociation constants KD with values of 32 -208 nM, averaged to 101 nM (Fig. 4.66), which matches the reported KD of typical thrombin aptamer bindings of 25-200 nM [4].

8 13 10

Figure 4.64: 20 spots of A14 (fibrinogen-binding-site aptamer elongated with a T20 spacer) at various positions (framed yellow) are analysed.

146 4.3 Molecule-protein interaction analysis by reflectometric interferometry

1.0025

1.002

1.0015

1.001

1.0005 relative intensity [a.u.] intensity relative

0.9995 0 300 600 900 1200 1500 1800 2100

time [s] Figure 4.65: association of thrombin to A14 followed by dissociation; evaluable region for association determined from time 520 s to 750 s; for dissociation from time 1500 s to 1850 s.

Figure 4.66: blue dots cloud depicting KD values (dissociation rate constant kd / association rate constant ka) with average KD of 101 nM (orange dot); ’irregular’ values e.g. 8, 10, 13 with lower KD or 1, 5 with higher KD occurred.

4.3.4.3 Serum immunization analysis Basic analyses of immunizations of rabbit sera indicates possibilities for applications in vaccines verification. The protocol according to table 4.9 was logged by the process controller (chapter 3.3.3); after initial buffer step and blocking there is non-immunized serum (negative control), followed by immunized serum, α-rabbit, αStrep and αHA for specificity analysis, finally Streptavidin-Cy5 as positive control. Buffer between each reagent step aims for washing and dissociation.

147 4.3 Molecule-protein interaction analysis by reflectometric interferometry

Table 4.9: logged data of protocol for serum immunization assay flow priming tube tube logged reagent step rate time position volume step logged time MTP name no. µl/min seconds code µl Buffer 1 60 200 0 220 1 7/30/2015 14:12 BSA 2 60 300 D12 320 2 7/30/2015 14:15 Buffer 3 60 300 0 320 3 7/30/2015 14:20 neg. Serum 4 60 300 C12 320 4 7/30/2015 14:25 Buffer 5 60 300 0 320 5 7/30/2015 14:30 pos. Serum 6 30 600 B12 320 6 7/30/2015 14:35 Buffer 7 60 300 0 320 7 7/30/2015 14:45 anti Rabbit 8 30 600 D11 320 8 7/30/2015 14:50 Buffer 9 60 300 0 320 9 7/30/2015 15:01 anti Strep 10 30 600 C11 320 10 7/30/2015 15:06 Buffer 11 60 300 0 320 11 7/30/2015 15:16 anti HA 12 30 600 B11 320 12 7/30/2015 15:21 Buffer 13 60 300 0 320 13 7/30/2015 15:31 Strep-Cy5 14 60 300 D10 320 14 7/30/2015 15:36 Buffer 15 60 300 0 320 15 7/30/2015 15:41

This protocol is applied to manually spotted IF slides coated with PDITC. Array design is according to the spots of Fig. 4.67 – it shows the final quotient pic after the complete protocol was run; further the choosen ROI’s and corresponding backgrounds, including an area with no spot as negative control. Spots were marked manually in this experiment.

1 4 7 4 4 4

2 5 8 4 4 4 10

3 6 9 4 4 4 11 Figure 4.67: quotient pic of rabbit serum immunization test depicting both the final intensity variations of the spots and the position of the various types of spots according to the table on the right.

148 4.3 Molecule-protein interaction analysis by reflectometric interferometry

serum serum ab- ab- Strep- BSA control positiv rabbit Strep ab-HA Cy5 1.025

HIS-HALO-GFP

1.02 HA-GFP-2x6HIS

Strep-GFP-2x6HIS HA-GFP-HALO-HIS 1.015

1.01 bBSA

bBSA relative intensity [a.u.] intensity relative

1.005 Strep-mCHERRY-2x6HIS HA-DbpA-HALO-HIS HA-mCHERRY-HALO-HIS 1 BSA no spot

0.995 300 900 1500 2100 2700 3300 3900 4500 5100 5700 6300 6900 time [s]

Figure 4.68: curves depict the binding of each reagent as expected: no binding of serum control; rabbit serum immunized against GFP binds to 4 GFP spots; ab-rabbit strongly binds to these spots, slightly to Strep-mCherry; ab-Strep binds to Strep-mCherry, slightly to Strep-GFP and HA-mCherry; ab-HA binds well to HA-DbpA and HA-mCherry, slightly to HA-GFP (2x); Streptavidin-Cy5 as positive control binds strongly to bBSA control spots, further to GFP spots, as ab-rabbit was biotinylated.

Experiment clearly shows the potential of the iRIf setup to depict quickly rather complex serum analyses. Various proteins could be spotted and immobilized onto PDITC coated glass slides, keeping their natural activity, that is their affinity and binding capability towards antibodies and antigens. These series of experiments finally prove the iRIf setups potential for highly parallel, quick molecule-protein interaction analyses on microarrays.

149 5.1 Conclusion

5 Conclusion and Outlook

5.1 Conclusion

5.1.1 Objective 1 - protein microarray synthesis in microfluidic incubator

In situ generation of protein microarrays via cell free mix from DNA microarrays was an innovative step in protein microarray generation. Since 2008 DAPA [31] was the most advanced method as it generates multiple full length protein microarrays from one single DNA microarray. However, it is restrictive in its efficiency, operability, robustness and controllability. Most limiting is its very unfavorable membrane:

1) causing inherent inhomogeneities due to its anisotropic porous structure. 2) lacking operability as a porous sheet of ~ 100 µm thickness needs to be soaked with cell-free system and put between template DNA and capture surface potentially violating the surface coatings, the DNA and the protein microarray. 3) missing robustness as drying-out during the assembly process can hardly be avoided; reproducibility is restricted due to pure manual assembly. 4) decreasing protein expression efficiency by hindering free diffusion of cell-free system ressources and created proteins.

In parallel several microfluidic devices addressing the efficiency of the protein expression have been introduced [53–55]; protein microarray generation by micro intaglio printing [157] or by PING chip [52, 158] improved controllability however, all systems lack simplicity and robustness in their technical design and in their operability.

In this thesis a novel approach was developed to circumvent the drawbacks of DAPA while keeping the multiple copy technology, and to harvest the advantages of microfluidic systems. The system design focuses on simplicity and robustness in design, operability and controllability in order to achieve a high acceptance among biologists and biochemists: a microfluidic gap is generated between a DNA array and a protein capture surface as part of a microfluidic flow cell omitting the DAPA membrane; sophisticated flow cell clamping systems allow the easy exchange of both the DNA microarray and the protein microarray slides in less than 2 minutes; washing routines allow for easy decontamination of all items. A Peltier element attached to the rear side of the flow cell enables temperature steering and control

150 5.1 Conclusion

during the protein expression. Such, a high degree in system robustness and operability could be created.

Compared to DAPA, our microfluidic gap reduces the amount of consumed cell-free expression mix to one-fifth, whilst cutting the copy process time in half. Such, the system efficiency is increased significantly.

Free diffusion of all components towards the DNA templates proves to work well, finally expressed proteins could be immobilized with high local correspondence opposite of the DNA spot. Spot distortions mainly occurred due to evaporations via inlet / outlet ports and leackages especially during the early states of the development; final PDMS flow cells with rear side structures prove to be enhanced in this respect and provide protein microarrays with minor deformations only.

The alternative booklet type flow cell design consisting of 2 standard glass slides with PTFE coated metal spacer meets complete compatibility with laboratory standards however, needs to be improved in robustness beyond this thesis; in comparison to the PDMS flow cell design it is especially more challenging considering leak tightness and flow control.

Appropriate protocols for the amplification of DNA and the cell-free in situ protein expression have been developed in our lab. Immobilization protocols for DNA with amino- linker onto PDITC coated glass or PDMS slides have been established. Protein capture by Ni- NTA coated glass slides (Xenopore, US) and self-made Ni-NTA coatings (work of Normann Kilb, AG Roth) proved to work well; further immobilization systems for covalent protein binding are being developed (HaloBind system, Tobias Herz, AG Roth).

Summary:

The proof-of-concept for the generation of protein microarrays from DNA microarrays with local correspondence in a microfluidic flow cell by cell-free protein expression and free protein diffusion could be established, benchmarking with the most advanced in situ protein microarray generation approach - DAPA - revealed major advantages:

o system without membrane improves handling significantly. o no physical abrasion of microarrays due to eliminated membrane. o unrestricted diffusion of enzymes, RNA and proteins in solution enhances reproducibility and leads to 2-fold increase in expression speed. o low microfluidic gap of 30 µm reduces consumption of cell-free system to 1/5. o integrated system with Peltier tempering and automation of liquid handling grants for operability and controllability; it opens up further assay and screening applications.

151 5.1 Conclusion

Generation of high-density protein microarrays via cell-free protein expression and free protein diffusion looks feasible however, certain aspects have to be analysed in detail:

o sharpness of spot rim due to free lateral diffusion. o depletion of cell free mix in dependence of flow cell height. o molecular crowding whilst protein synthesis. o bleeding of non-covalently bound proteins during washing steps.

Beyond that, different cell-free mixes should be applied for the analysis of expression efficiency, protein folding and special features as IRES-site screening, codon usage or transcription factor influences.

Eventually, aspired system features according to objectives could be achieved: multiple protein copy technology from DAPA meets controllability of microfluidic systems as e.g. PING; the lack of all existing systems in terms of robustness and operability is eliminated by a sophisticated system design ruled by simplicity; full-length protein microarrays can be created based upon their DNA microarray counterpart.

5.1.2 Objective 2 – automated real-time label-free detection system

Based upon the patented reflecto-interferometric principle an automated setup for label- free real-time imaging and analysis of binding interactions was realized. The criterion ‘real- time imaging’ demanded to sacrifice the highly sensitive method of spectroscopical analysis (RIfS) towards the lesser sensitive single wavelength system (1λ-iRIf) for the benefit of significantly lower image processing times allowing frame rates up to 100 Hz. Such as robust detector, a rather common digital camera with telecentric lens, a single wavelength LED with diffusor and a polarizer remain as major hardware parts of the sensor system. For signal generation a special glass slide with buried high-refractive layer is applied as microarray surface.

As first major task within this thesis an image processing software was developed, initially based upon ImageJ for proof-of-concept, later implemented with Java allowing free programming of customized functionalities. Both types of implementations proved to be suitable for real-time imaging followed by post experimental analysis, creating quotient pictures for any selected time interval, videos of binding events and determining the association and dissociation curves. For the purpose of automated evaluations Matlab programs for the spot identification and the binding curve analysis have been implemented. Spot identifications via given grids could not be applied, as distortions of the spotting pattern could not be detected reliably. The spot identification by spot border detection proved to work well, as long as contrasts are sufficient. This allows to automatically identify good

152 5.1 Conclusion

binders, generally weak binders are not of main interest. The Matlab based binding curve evaluation with visual depiction of curve fittings and evaluation range as well as calculation of the rate constants permit efficient analyses of 1:1 reactions; multiple binding site reactions, molecular crowding effects or reagent diffusion limitations are not met yet automatically however, some effects could be read by naked eye from the signal course.

As second major task a generic microfluidic flow cell as essential element of an electrically controlled fluidic system was realized. Glass slides in standard format but semi-reflective according to the iRIf principle were sealed by PDMS slides – basically the same system as for objective 1 was realized however, versatile structures for the priming and flow control have been evaluated and finally lead to the result, that a precisely manufactured plane flow cell with appropriate inlet and outlet geometries meets the requirements. The red PDMS revealed good absorbence of the 530 nm light, such ensuring that non-coherent reflected light is minimized for the benefit of a higher SNR. The resolution of > 5 million pixel for an image of size 18 x 15 mm2 (object resolution of ~ 7 µm per pixel) enables to monitor simultaneously > 10.000 spots at sizes of 200 pixel (spot diameter of ~ 110 μm) with standard deviation of < 10-4 by spatial averaging only. Even more spots of lower size can be detected with equal SNR if pixel intensity averaging or median value determination over time is applied additionally. The fluidic system design permits a continuous flow of buffer and several reagents during the experiment. A sequence of flow rates from some nl/min to several ml/min with reagent volumes up to 1.5 ml and arbitrary buffer volumes can be determined for up to 44 reagents in an Excel sheet as process parameters – this enables complex flow-injection assays with biological impact to be performed, e.g. multi-step antigene antibody analysis, thrombin aptamer analysis including competition assays and secondary staining with antibodies up to serum immunization analysis upon an array of antigens. Thrombin aptamer interaction analyses resulted in the determination of KD values of ~ 100 nM, which is in accordance to literature, bindings at concentrations as low as 50 nM of thrombin by spatial averaging of 200 pixel could be measured. Multi-step ELISA-like interactions on antigen arrays with several antibodies reveal the systems potential in microarray based kinetic data acquisition. Experiments with rabbit sera show the systems benefit for analyzing the efficiency of vaccines, e.g. for personalized medical studies. The physical system analysis exhibited inherent flow-injection effects, e.g. inhomogenous flow profiles or reagent depletion, influencing the KD determination. That needs to be addressed while establishing the spotting pattern and the flow-injection protocol. Baseline noise of < 3*10-9 RSS/min and a baseline drift of < 0.003 ‰/min allow sensitive long term measurements. Hence, this system is competitive to known label-free detection systems, e.g. SPR, SPRi, QCM or BLI systems [system comparison paper of AG Roth in preparation] considering sensitivity, yet it is far superior in imaging capacity with tens of thousands of

153 5.2 Outlook spots being detectable in parallel. Such, this system allows to make high-throughput analyses in a very efficient manner as imaging system based upon digital camera technology. Combined with the protein microarray expression it opens up a complete new field of analytics and biochemistry for the screening of phage and ribosome displays, pre-defined DNA arrays for analysis of allergens, or the determination of vaccination status or autoimmunities. A patent with more than 40 applications described has been filed (WO 2013/ 186359 A1).

5.2 Outlook

Finally the merging of the in situ protein expression and the protein interaction analysis will be tackled to enable a complete workflow – from copy to thorough interaction analysis. A proof-of-concept for the integrated system could be established (Fig. 5.1). Protein expression was realized in the tempered flow cell, the expression was observed via labelled proteins by microarray scanner. Consecutive interactions of the immobilized proteins with antibodies could be detected label-free within the iRIf setup.

The deposition of the in situ generated proteins onto the surface can hardly be detected yet, probably due to a lack of SNR caused by the influence of the flow cell tempering and the massive protein interactions of the cell-free mix with all surrounding surfaces. Further analyses have shown that sheer forces applied by tubes, connectors or valves onto the cell- free mix is a critical issue leading to an inactivation of the cell-free system. The fluidic protocol therefore needs to be optimized, tube coatings and tubes inner dimensions have to be selected appropriately.

154 5.2 Outlook

(A) (B) (C)

Figure 5.1: (A) expressed protein spots in solution (GFP (green), top; mCherry (red), bottom), detected in closed flow cell by fluorescence microscopy [159] during the protein expression in the flow cell. (B) label-free iRIf detection of binding of antibodies to immobilized protein spots (anti-GFP antibody, top; anti-mCherry antibody, bottom). (C) end-point detected antibodies with fluorophores bound to proteins [154]. (experiments performed by Tobias Herz and Normann Kilb, AG Roth; setup enhanced in 2017 by Jürgen Burger, applying microscope iRIf slides with novel coatings and enhanced LED system provided by Biametrics GmbH, Tübingen, Germany.)

First experiments towards multiply expressing protein microarrays from a single DNA template microarray in the microfluidic setup have already been successfully performed (Fig. 5.2). Results clearly reveal feasibility and high yield, further development and detailed analyses will be done within the dissertations of Normann Kilb and Tobias Herz, AG Roth.

Figure 5.2: first and second protein microarray expressed from single DNA template microarray, end-point detected by fluorescence scanning via labelled antibodies. (experiment performed by Tobias Herz and Normann Kilb, AG Roth)

155 5.2 Outlook

Latest proof-of-concepts for the integrated system hinted to a range of parameters which need to be addressed for reaching a high process standard:

1) Inhibition, denaturation or inactivation of cell-free system by tubings, flow cell. 2) Temperature homogeneity during expression enhancing reproducibility of copy process. 3) Robustness of protein immobilization minimizing bleeding. 4) Suitability of DNA templates for cell-free in situ expression. 5) Minimization of lateral diffusion of proteins for getting tiny, sharp protein spots. 6) Several capture surfaces and protein tags providing flexibility and oriented protein immobilization addressing steric artefacts. 7) Sensitivity of label-free detection system.

Aside these topics the flow cell geometry could be enhanced. Further decrease in flow cell heights down to 10 μm would decrease the lateral protein spot growth, would reduce the consumption of cell-free mix to less than 10 μl per protein array copy. In addition, the height of the flow cell could be designed such that the cell-free mix will use up all its ressources to generate a complete microarray without any surplus of proteins; such being equivalent to the carbon paper copying process. Nevertheless, lower flow cell heights will also increase reagent depletion during subsequent binding assays. Hence, a trade-off between material consumption whilst protein microarray copying and good conditions during binding kinetics analysis has to be found.

Recently, a comfortable App “Anabel” programed in ‘R’ for the curve analysis including drift compensation was created by Stefan Krämer, AG Roth. This tool primarily aims to increase robustness, operability and precision of the binding constants determination. Next, we aspire to enhance this tool by a routine for the automated identification of the evaluable binding curve region, representing the 1:1 reaction. Further studies need to reveal intelligent methods for the analysis of multiple site binding interactions.

156

References

1. Kilb N, Burger J, Roth G (2014) Protein microarray generation by in situ protein expression from template DNA. Eng. Life Sci.(14):352–364 2. He M, Stoevesandt O, Palmer EA, Khan F, Ericsson O, Taussig MJ (2008) Printing protein arrays from DNA arrays. Nature methods 5(2):175–177 3. Mark D, Focke M, Lutz S, Burger J, Müller M, Riegger L, Rombach M, Hoffmann J, Roth G, Piepenburg O, Park Y, Zengerle R, Stetten Fv (2010) Lab-on-a-chip solutions designed for being operated on standard laboratory instruments. Procedia Engineering 5:444–447 4. Burger J, Rath C, Woehrle J, Meyer PA, Ben Ammar N, Kilb N, Brandstetter T, Pro ll F, Proll G, Urban G, Roth G (2016) Low-Volume Label-Free Detection of Molecule- Protein Interactions on Microarrays by Imaging Reflectometric Interferometry. Journal of laboratory automation:1–10 5. Lämmle D (2011) Imaging of Immobilization Processes of Molecules with iRIfS. Master's thesis, University of Freiburg 6. Suleman Shakil (2012) Design and evaluation of tempered PDMS flow cell for protein microarray generation_Final.docx. Master's Thesis, Albert-Ludwigs- Universität 7. Fakler M (2012) Entwicklung und Konstruktion einer mikrofluidischen Flusszelle. Technikerarbeit 8. Rudmann L (2013) Label-free Detection and Analysis of Protein Interactions on Microarrays in Microfluidic Flow Cells. Master Thesis, University of Freiburg 9. Ben Ammar N (2013) Temperature Control Unit for Protein Microarray Copier. Bachelor Thesis, Universität Freiburg 10. Heller MJ (2002) DNA microarray technology: devices, systems, and applications. Annual review of biomedical engineering 4:129–153 11. Eads CA (2000) MethyLight. A high-throughput assay to measure DNA methylation. Nucleic acids research 28(8):e32 1-8 12. Benfey PN, Mitchell-Olds T (2008) From genotype to phenotype: systems biology meets natural variation. Science (New York, N.Y.) 320(5875):495–497 13. Hall DA, Ptacek J, Snyder M (2007) Protein microarray technology. Mechanisms of ageing and development 128(1):161–167 14. Kamphampati D (2004) Protein Microarray Technology. WILEY-VCH Verlag, Weinheim 15. Stoll D (2002) Protein microarray technology. Front Biosci 7(1-3):815–822

157

16. Zhu H, Qian J (2012) Applications of functional protein microarrays in basic and clinical research. Advances in genetics 79:123–155 17. MacBeath G, Schreiber SL (2000) Printing Proteins as Microarrays for High- Throughput Function Determination. Science(289):1760–1763 18. Gao X, Pellois JP, Na Y, Kim Y, Gulari E, Zhou X (2004) High density peptide microarrays. In situ synthesis and applications. Mol Divers 8(3):177–187 19. Frank R (2002) The SPOT-synthesis technique Synthetic peptide arrays on membrane supports—principles and applications. J. Immunol. Methods(267):13–26 20. Baranov VI, Morozov IY, Ortlepp SA, Spirin AS (1989) Gene expression in a cell-free system on the preparative scale. Gene 84(2):463–466 21. PELHAM HRB, JACKSON RJ (1976) An Efficient mRNA-Dependent Translation System from Reticulocyte Lysates. Eur J Biochem 67(1):247–256 22. Ashaari NS, Ramarad S, Khairuddin D, Akhir NAM, Hara Y, Mahadi NM, Mohamed R, Nathan S (2015) Development of repeatable arrays of proteins using immobilized DNA microplate (RAPID-M) technology. BMC research notes 8:669:1–9 23. Karthikeyan Kea (2016) A contra capture protein array platform for studying post- translationally modified auto-antigenomes. The American Society for Biochemistry and Molecular Biology)(Manuscript M115.057661):1–54 24. Timm C, Niemeyer CM (2013) On-chip protein biosynthesis. Angewandte Chemie (International ed. in English) 52(10):2652–2654 25. QIAGEN (2005) EasyXpress Protein Synthesis Handbook 26. He M, Taussig MJ, He M (2001) Single step generation of protein arrays from DNA by cell-free expression and in situ immobilisation (PISA method). Nucleic acids research 29(15 e73):1–6 27. Angenendt P (2006) Generation of High Density Protein Microarrays by Cell-free in Situ Expression of Unpurified PCR Products. Molecular & Cellular Proteomics 5(9):1658–1666 28. Ramachandran N, Hainsworth E, Bhullar B, Eisenstein S, Rosen B, Lau AY, Walter JC, Labaer J (2004) Self-assembling protein microarrays. Science 305(5680):86–90 29. Ramachandran N, Raphael JV, Hainsworth E, Demirkan G, Fuentes MG, Rolfs A, Hu Y, Labaer J (2008) Next-generation high-density self-assembling functional protein arrays. Nat Meth 5(6):535–538 30. Yazaki J, Galli M, Kim AY, Nito K, Aleman F, Chang KN, Carvunis A, Quan R, Nguyen H, Song L, Alvarez JM, Huang SC, Chen H, Ramachandran N, Altmann S, Gutierrez RA, Hill DE, Schroeder JI, Chory J, Labaer J, Vidal M, Braun P, Ecker JR (2016) Mapping transcription factor interactome networks using HaloTag protein arrays. Proceedings of the National Academy of Sciences of the United States of America 113(29):E4238-E4247

158

31. He M, Stoevesandt O, Taussig MJ (2008) In situ synthesis of protein arrays. Current Opinion in Biotechnology 19(1):4–9 32. Stoevesandt O, Vetter M, Kastelic D, Palmer EA, He M, Taussig MJ (2011) Cell free expression put on the spot: advances in repeatable protein arraying from DNA (DAPA). New Biotechnology 28(3):282–290 33. Stoevesandt O, Taussig MJ, He M (2011) Producing Protein Microarrays from DNA Microarrays. Methods in Molecular Biology 785:265–276 34. Stoevesandt O, Vetter M, Kastelic D, Palmer EA, He M, Taussig MJ (2011) Cell free expression put on the spot: advances in repeatable protein arraying from DNA (DAPA). New Biotechnology 28(3):282–290 35. Fechner P, Bleher O, Ewald M, et al. (2014) Size does matter! Label-free detection of small molecule-protein interaction. Anal. Bioanal. Chem. 406(17):4033–4051 36. Malmqvist M (1993) Biospecific interaction analysis using biosensor technology. Nature 361(6408):186–187 37. Mattei Bea (2004) Biomolecular interaction analysis and MS. American chemical Society:18–25 38. 3T analytik Quartz crystal microbalances. http://www.3t-analytik.de/. Accessed 4 April 2016 39. ForteBIO Octet RED96 System. http://www.fortebio.com/. Accessed 4 April 2016 40. Homola J, Yee SS, Gauglitz G (1999) Surface plasmon resonance sensors: review. Sens. Actuators, B 54(1-2):3–15 41. Gauglitz G (2005) Direct optical sensors: principles and selected applications. Anal. Bioanal. Chem. 381(1):141–155 42. Pröll F, Möhrle B, Kumpf M, Gauglitz G (2005) Label-free characterisation of oligonucleotide hybridisation using reflectometric interference spectroscopy. Anal. Bioanal. Chem. 382(8):1889–1894 43. Gauglitz G (2014) Point-of-care platforms. Annu. Rev. Anal. Chem. 7:297–315 44. GE Healthcare (2007) Biacore T100 Instrument Handbook BR-1006-47 Edition AD. Handbook BR-1006-47 Edition AD(1):1–82 45. Pröll F, Fechner P, Proll G (2009) Direct optical detection in fragment-based screening. Analytical and Bioanalytical Chemistry 393(6-7):1557–1562 46. Kraus G, Klotz A, Seemann J, Spaeth K, Gauglitz G (1995) Application and comparison of algorithms for the evaluation of interferograms. Fresenius J Anal Chem(352):426–430 47. Ewald M (2014) Entwicklung und Charakterisierung einer portablen Biosensorplattform zur markierungsfreien Multi-Analyt-Bestimmung. Dissertation, Eberhard-Karls-Universität Tübingen 48. Frank R (2005) Reflektometrische und integriert optische Sensoren für die Bioanalytik. Dissertation, Eberhard-Karls-Universität Tübingen

159

49. Markovic G (2008) Einsatz von Nanopartikeln und optimierten Schichtsystemen für die Signalverstärkung in reflektometrischen Biosensoren. Dissertation, Eberhard- Karls-Universität Tübingen 50. Pröll F (2010) iRIfS-imaging Reflectrometric Interference Sensor : ein System für die hochparallele und markierungsfreie Untersuchung biomolekularer Wechselwirkungen. Dissertation, Eberhard-Karls-Universität Tübingen 51. Bleher O, Schindler A, Yin M, Holmes AB, Luppa PB, Gauglitz G, Proll G (2014) Development of a new parallelized, optical biosensor platform for label-free detection of autoimmunity-related antibodies. Anal. Bioanal. Chem. 406(14):3305– 3314 52. Gerber D, Maerkl SJ, Quake SR (2008) An in vitro microfluidic approach to generating protein-interaction networks. Nat Meth 6(1):71–74 53. Kinpara T (2004) A Picoliter Chamber Array for Cell-Free Protein Synthesis. Journal of Biochemistry 136(2):149–154 54. Mei Q, Fredrickson CK, Simon A, Khnouf R, Fan ZH (2007) Cell-free protein synthesis in microfluidic array devices. Biotechnology progress 23(6):1305–1311 55. Meier M, Kennedy-Darling J, Choi SH, Norstrom EM, Sisodia SS, Ismagilov RF (2009) Plug-Based Microfluidics with Defined Surface Chemistry to Miniaturize and Control Aggregation of Amyloidogenic Peptides. Angew. Chem. 121(8):1515–1517 56. Proll G, Markovic G, Steinle L, Gauglitz G (2009) Reflectometric interference spectroscopy. Methods in molecular biology (Clifton, N.J.) 503:167–178 57. Biametrics GmbH SCORE (Single Colour Reflectometry) measurement principle. http://biametrics.com/technology/. Accessed 26 May 2017 58. Bruus H (2008) Theoretical microfluidics. Oxford master series in condensed matter physics, vol. 18. Oxford University Press, Oxford 59. Forsberg PSH, Priest C, Brinkmann M, Sedev R, Ralston J (2010) Contact line pinning on microstructured surfaces for liquids in the Wenzel state. Langmuir : the ACS journal of surfaces and colloids 26(2):860–865 60. Cho H, Kim H, Kang JY, Kim TS (2007) How the capillary burst microvalve works. Journal of Colloid and Interface Science 306(2):379–385 61. Vulto P, Podszun S, Meyer P, Hermann C, Manz A, Urban GA (2011) Phaseguides: a paradigm shift in microfluidic priming and emptying. Lab on a Chip 11(9):1596– 1602 62. Concus P, Finn R On capillary free surfaces in the absence of gravity. Acta Mathematica 1974(132):pp 177-198 63. Purcell EM (1976) Life at low Reynolds number. Physics and our World. American Institute of Physics

160

64. Taylor F (1953) Dispersion of soluble matter in solvent flowing slowly through a tube. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences.:186–204 65. Kerstin Laenge (2000) Einsatz der Fließ-Injektions-Analyse in markierungsfreien Bindungsassays. Dissertation, Eberhard-Karls-Universität Tübingen 66. Bäumchen O, Jacobs K (2010) Slip effects in polymer thin films. Journal of physics. Condens matter 22((3) 033102):1–21 67. Sochi T (2011) Slip at Fluid-Solid Interface. Polymer Reviews 51(4):309–340 68. Rao IJ, Rajagopal KR (1999) The effect of the slip boundary condition on the flow of fluids in a channel. Acta Mechanica 135:113–126 69. Mackay ME, Henson DJ (1998) The effect of molecular mass and temperature on the slip of polystyrene melts at low stress levels. J. Rheol. 42(6):1505–1517 70. Lei W, Jiankang W, Bo C (2009) Analytic Solution of Liquid Flow in Rectangular PDMS-GLASS Microchannel with Wall Slip and Electro-Viscous Effects. Applied Mathematical Sciences(3):2195–2214 71. Tang HS, Kalyon DM (2008) Unsteady circular tube flow of compressible polymeric liquids subject to pressure-dependent wall slip. J. Rheol. 52(2):507–525 72. Cottin-Bizonne C, Steinberger A, Cross B, Raccurt O, Charlaix E (2008) Nanohydrodynamics: the intrinsic flow boundary condition on smooth surfaces. Langmuir : the ACS journal of surfaces and colloids 24(4):1165–1172 73. Noll W (1958) A mathematical theory of the mechanical behavior of continuous media. Arch. Rational Mech. Anal.(2):197–226 74. Meyer PA (2014) Design and characterization of microfluidic flow cells for label-free reflectrometric analysis of molecular binding kinetics. Master Thesis, Albert- Ludwigs-Universität Freiburg 75. Hong Qea (1991) Single particle tracking. Analysis of diffusion and flow in two- dimensional systems. Biophys. Journal(60):910–921 76. Hänggi Pea (2008) 100 Years of Brownian motion. http://arxiv.org/pdf/cond- mat/0502053v1.pdf. Accessed 7 January 2016 77. Chapman S, Cowling TG (1970) The mathematical theory of non-uniform gases: an account of the kinetic theory of viscosity, thermal conduction and diffusion in gases. Cambridge university press, Cambridge 78. Fick A (1855) On liquid diffusion. Philosophical Magazine Series 4(10:63):30–39 79. Baumann G, F. Place R, Foldes-Papp Z (2010) Meaningful Interpretation of Subdiffusive Measurements in Living Cells (Crowded Environment) by Fluorescence Fluctuation Microscopy. CPB 11(5):527–543 80. Foldes-Papp Zea Fluorescence Molecule Counting for Single-Molecule Studies in Crowded Environment of Living Cells without and with Broken Ergodicity. Current Pharmaceutical Biotechnology 2011(12):824–833

161

81. Losa, Gabriele A. et al. (ed) (1998) Fractals in Biology and Medicine. Springer Basel AG 82. Foldes-Papp Z (2013) Measurements of Single Molecules in Solution and Live Cells Over Longer Observation Times Than Those Currently Possible: The Meaningful Time. CPB(14):1–4 83. Baumann G, Kinjo M, Földes-Papp Z (2014) Anomalous Subdiffusive Measurements by Fluorescence Correlations Spectroscopy and Simulations of Translational Diffusive Behavior in Live Cells. J Biol Methods 1(1) 84. Feder, Toni J. et al. (1996) Constrained Diffusion or Immobile Fraction on Cell Surfaces: A New Interpretation. Biophysical Journal(70):2767–2773 85. Digman MA, Gratton E (2009) Imaging barriers to diffusion by pair correlation functions. Biophysical Journal 97(2):665–673 86. Digman MA, Gratton E (2011) Lessons in fluctuation correlation spectroscopy. Annual review of physical chemistry 62:645–668 87. Cardarelli F, Gratton E (2010) In vivo imaging of single-molecule translocation through nuclear pore complexes by pair correlation functions. PLoS ONE 5(5 e10475):1–12 88. Schwille P, Korlach J, Webb WW (1999) Fluorescence correlation spectroscopy with single-molecule sensitivity on cell and model membranes. Cytometry 36(3):176– 182 89. Saxton MJ (2007) A biological interpretation of transient anomalous subdiffusion. I. Qualitative model. Biophysical Journal 92(4):1178–1191 90. Foldes-Papp Z (2015) Individual Macromolecule Motion in a Crowded Living Cell. CPB 16(1):1–2 91. Mehne J, Markovic G, Pröll F, Schweizer N, Zorn S, Schreiber F, Gauglitz G (2008) Characterisation of morphology of self-assembled PEG monolayers: a comparison of mixed and pure coatings optimised for biosensor applications. Analytical and Bioanalytical Chemistry 391(5):1783–1791 92. Proll G, Markovic G, Steinle L, Gauglitz G (eds) (2008) Reflectometric interference spectroscopy. In Biosensors and Biodetection: Methods and Protocols. Optical- Based Detectors, 1st edn. Humana Press, New York 93. Hilbig U, Bleher O, Le Blanc A, Gauglitz G (2012) A biomimetic sensor surface to detect anti-β2-glycoprotein-I antibodies as a marker for antiphospholipid syndrome. Analytical and Bioanalytical Chemistry 403(3):713–717 94. Mateescu A, Wang Y, Dostalek J, Jonas U (2012) Thin hydrogel films for optical biosensor applications. Membranes 2(1):40–69 95. Dominguez RB, Alonso GA, Muñoz R, Hayat A, Marty J (2015) Design of a novel magnetic particles based electrochemical biosensor for organophosphate

162

insecticide detection in flow injection analysis. Sensors and Actuators B: Chemical 208:491–496 96. Alberts Bea (ed) Molekularbiologie der Zelle, 4th edn. WILEY-VCH Verlag, Weinheim 97. Roitt IM, Male D, Brostoff J (1996) Immunology, 4th edn. Mosby-Year Book 98. Nelson DL, Cox MM (2008) Lehninger Principles of Biochemistry. http://images.google.de/imgres?imgurl=http%3A%2F%2Fchemistry.umeche.main e.edu%2FCHY431%2FAntibody2.jpg&imgrefurl=http%3A%2F%2Fchemistry.umec he.maine.edu%2FCHY431%2FChorismate5.html&h=804&w=866&tbnid=HKMgeFR 6dC8djM%3A&vet=1&docid=D5ZKZnp3I5EdFM&ei=R4hqWLnjA4KtUdiBvNgM&tb m=isch&iact=rc&uact=3&dur=837&page=4&start=100&ndsp=29&ved=0ahUKEwj5 ornv-KPRAhWCVhQKHdgAD8s4ZBAzCA8oDTAN&bih=782&biw=1370. Accessed 19 May 2017 99. Wang W, Singh S, Zeng DL, King K, Nema S (2007) Antibody structure, instability, and formulation. Journal of pharmaceutical sciences 96(1):1–26 100. Hänel C (2003) Parameteroptimierung für zwei Verfahren zur markierungsfreien Analyse biomolekularer Wechselwirkungen. Dissertation, Eberhard-Karls- Universität Tübingen 101. Wood RW (1912) XXVII. Diffraction gratings with controlled groove form and abnormal distribution of intensity. Philosophical Magazine Series 6 23(134):310– 317 102. Burstein Eea (1974) Surface polaritons—propagating electromagnetic modes at interfaces. J. Vac. Sci. Technol(11 6):1004–1019 103. Piehler J (2005) New methodologies for measuring protein interactions in vivo and in vitro. Current Opinion in Structural Biology 15(1):4–14 104. Karlsson R, Fält A (1997) Experimental design for kinetic analysis of protein-protein interactions with surface plasmon resonance biosensors. J. Immunol. Methods(200):121–133 105. Gauglitz G, Reichert M Affinitätsreaktionen: Beschreibung von Immunoassays, Testformaten, Massentransport, Bindungskinetik. http://www.chemgapedia.de/vsengine/vlu/vsc/de/ch/13/vlu/kinetik/affinitaet/a ffinitaetsreaktionen.vlu/Summary.html. Accessed March 2015 106. Länge K (2000) Einsatz der Fließ-Injektions-Analyse in markierungsfreien Bindungsassays. Dissertation, Eberhard-Karls-Universität Tübingen 107. Maass‐Moreno R (2005) Fitting Models to Biological Data Using Linear and Nonlinear Regression: A Practical Guide to Curve Fitting. By Harvey Motulsky and Arthur Christopoulos, vol. 80. Oxford University Press, Oxford and New York 108. O'Shannessy, D. J., et al. (1996) Interpretation of Deviations from Pseudo-First- Order Kinetic Behavior in the Characterization of Ligand Binding by Biosensor Technology. Analytical Biochemistry(236):275–283

163

109. Hall DR, Gorgani NN, Altin JG, Winzor DJ (1997) Theoretical and experimental considerations of the pseudo-first-order approximation in conventional kinetic analysis of IAsys biosensor data. Analytical Biochemistry 253(2):145–155 110. Edwards PR, Maule CH, Leatherbarrow RJ, Winzor DJ (1998) Second-order kinetic analysis of IAsys biosensor data: its use and applicability. Analytical Biochemistry 263(1):1–12 111. Azizian S (2004) Kinetic models of sorption: a theoretical analysis. Journal of Colloid and Interface Science 276(1):47–52 112. Origin Nonlinear Curve Fitting. https://www.originlab.com/doc/Origin- Help/Nonlinear-Curve-Fit. Accessed 2 January 2017 113. Foldes-Papp Zea (2004) Counting and Behavior of an Individual Fluorescent Molecule without Hydrodynamic Flow, Immobilization, or Photon Count Statistics. Current Pharmaceutical Biotechnology(5):163–172 114. Möhrle BP, Kumpf M, Gauglitz G (2005) Determination of affinity constants of locked nucleic acid (LNA) and DNA duplex formation using label free sensor technology. The Analyst 130(12):1634–1638 115. Piehler J, Brecht A, Giersch T, Hock B, Gauglitz G (1997) Assessment of affinity constants by rapid solid phase detection of equilibrium binding in a flow system. Journal of Immunological Methods 201(2):189–206 116. Amatore C, Szunerits S, Thouin L, Warkocz J (2001) The real meaning of Nernst's steady diffusion layer concept under non-forced hydrodynamic conditions. A simple model based on Levich's seminal view of convection. Journal of Electroanalytical Chemistry 500(1-2):62–70 117. Ibl N (1981) Nomenclature for transport phenomena in electrolytic systems. Pure and Applied Chemistry 53(10):1827–1840 118. Karlsson R, Michaelsson A, Mattsson L (1991) Kinetic analysis of monoclonal antibody-antigen interactions with a new biosensor based analytical system. J. Immunol. Methods 145(1-2):229–240 119. Christensen LL (1997) Theoretical analysis of protein concentration determination using biosensor technology under conditions of partial mass transport limitation. Analytical Biochemistry 249(2):153–164 120. Sjölander S, Urbaniczky C Integrated Fluid Handling System for Biomolecular Interaction Analysis. Anal Chem. 1991(63):2338–2345 121. Saleh BEA, Teich MC (1991) Fundamentals of photonics. Wiley series in pure and applied optics. John Wiley @ Sons, Inc., New York 122. Heavens OS (1955) Optical Properties of Thin Films. Academic Press 123. PCO AG pco.camera 1600 data sheet. https://www.pco.de/fileadmin/user_upload/db/download/MA_DCOPIE_0412b.pdf. Accessed May 2017

164

124. Gauglitz G, Brecht A, Kraus G, Mahm W (1993) Chemical and biochemical sensors based on interferometry at thin (multi-) layers. Sensors and Actuators B: Chemical 11(1-3):21–27 125. Mok J, Im H, Snyder M (2009) Global identification of protein kinase substrates by protein microarray analysis. Nat Protoc 4(12):1820–1827 126. He M, Taussig MJ (2001) Single step generation of protein arrays from DNA by cell- free expression and in situ immobilisation (PISA method). Nucleic Acids Res. 29(15 e73):1–6 127. Ratner BD (2004) Biomaterials science: An introduction to materials in medicine. https://books.google.de/books?hl=en&lr=&id=9PMU1iYGe34C&oi=fnd&pg=PR5&d q=B.+D.+Ratner,+Biomaterials+science:+An+introduction+to+materials+in+medici ne,+2nd+Ed,+London:+Elsevier+Academic+press,+2004.&ots=gmVdqljvEv&sig=NP zgzY0FkmhMYqoY58O8tAwz0y0#v=onepage&q=B.%20D.%20Ratner%2C%20Bio materials%20science%3A%20An%20introduction%20to%20materials%20in%20 medicine%2C%202nd%20Ed%2C%20London%3A%20Elsevier%20Academic%20 press%2C%202004.&f=false. Accessed May 2017 128. Tang S, Whitesides GM (eds) (2010) Basic Microfluidic and Soft Lithographic Techniques. Optofluidics: Fundamentals, Devices and Applications. Chapter 2. Biophotonics series. McGraw-Hill, New York 129. Wangler N, Gutzweiler L, Kalkandjiev K, Müller C, Mayenfels F, Reinecke H, Zengerle R, Paust N (2011) High-resolution permanent photoresist laminate TMMF for sealed microfluidic structures in biological applications. J. Micromech. Microeng. 21(9):1–9 130. Vulto P, Huesgen T, Albrecht B, Urban GA (2009) A full-wafer fabrication process for glass microfluidic chips with integrated electroplated electrodes by direct bonding of dry film resist. J. Micromech. Microeng. 19(7 077001):1–5 131. Andreas Hettich GmbH Manual Hettich Zentrifuge Universal 320R. http://www.hettichlab.com/web/global/prod/Mikroliterzentrifugen/502. Accessed May 2017 132. Tyan R, Sun P, Scherer A, Fainman Y (1996) Polarizing beam splitter based on the anisotropic spectral reflectivity characteristic of form-birefringent multilayer gratings. Opt. Lett. 21(10):761–763 133. PCO AG pco.1600 14 bit CCD camera system. http://www.pco.de/fileadmin/user_upload/db/products/datasheet/pco1600_2008 0805.pdf. Accessed 19 May 2017 134. Sill Optics S5LPL2060/LED Correctal® TA/0.5. http://www.silloptics.de/fileadmin/user_upload/Downloads/Datasheet/S5LPL206 0-LED.pdf. Accessed 19 May 2017

165

135. PCO AG pco.edge 5.5 scientific CMOS camera. http://www.pco.de/fileadmin/user_upload/db/download/BR_pco_edge55_102.pdf. Accessed 19 May 2017 136. Sill Optics S5LPL1005/LED Correctal TA/0.92. http://www.silloptics.de/fileadmin/user_upload/Downloads/Datasheet/S5LPL100 5-LED.pdf. Accessed 19 May 2017 137. Novoa A, Eierhoff T, Topin J, Varrot A, Barluenga S, Imberty A, Römer W, Winssinger N (2014) A LecA ligand identified from a galactoside-conjugate array inhibits host cell invasion by Pseudomonas aeruginosa. Angewandte Chemie (International ed. in English) 53(34):8885–8889 138. Tecan Group Ltd. Cavro XE1000 syringe pump. http://ww3.tecan.com/platform/content/element/1649/Tecan_XE1000_DataSheet .pdf. Accessed 19 May 2017 139. Tecan Group Ltd. Cavro SVP valve. http://ww3.tecan.com/platform/content/element/6302/BR_Cavro_Product_Overvi ew_395351_v3-1.pdf. Accessed 19 May 2017 140. Msscientific Chromatographie-Handel GmbH Fitting - Omni-Lok Fitting Inverskonus. http://www.diba-industries.de/fittings_omnifit_omnilok_inverskonus.htm. Accessed 19 May 2017 141. Tyson T ShotNoise.dvi. http://123.physics.ucdavis.edu/shot_files/ShotNoise.pdf. Accessed 4 March 2015 142. The MathWorks Inc ©1994-2014 The MathWorks Inc., Least-Squares Fitting - MATLAB, 2014b ed. http://de.mathworks.com/help/curvefit/least-squares- fitting.html. Accessed Jan 2017 143. Hoffmann J, Hin S, Stetten Fv, Zengerle R, Roth G (2012) Universal protocol for grafting PCR primers onto various lab-on-a-chip substrates for solid-phase PCR. RSC Adv. 2(9):3885–3889 144. Wulfetange C Scienion Enabling Life Science- sciFLEXARRAYER Ultra-Low Volume Dispensers. http://www.scienion.com/fileadmin/user_upload/pdf/11_sciFLEXARRAYER_SX_20 140903.pdf. Accessed 4 March 2015 145. 5 PRIME RTS E.coli System and Technology. https://www.5prime.com/applications/protein-technologies/rts/technologies/e- coli.aspx. Accessed 14 June 2014 146. Xenopore Corporation Nickel chelate slides. http://www.xenopore.com/products/nickel-chelate-on-glass/. Accessed 19 May 2017 147. Tasset, D. M. et al. Oligonucleotide Inhibitors of Human Thrombin that Bind Distinct Epitopes. J. Mol. Biol. 1997(272):688–698

166

148. BioFluidiX GmbH TopSpot® Technology. http://www.biofluidix.com/en-products- topspot-topspottechnology.html. Accessed 19 May 2017 149. Schott NEXTERION® Spot. https://appliedmicroarrays.com/product/nexterion- spot-100-ml/. Accessed 7 April 2017 150. Schott Epoxysilane coating. http://www.us.schott.com/nexterion/english/products/coated_slides/epoxysilane /epoxysilane.html. Accessed 19 May 2017 151. GESIM GeSIM Nano-Plotter. http://gesim-bioinstruments-microfluidics.com/wp- content/uploads/2014/11/nano-plotter-microarraying-picolitre-pipetting-gesim- prospekt-de.pdf. Accessed 19 May 2017 152. BioFluidiX PipeJet Technology. http://www.biofluidix.com/en-products-pipejet- pipejettechnology.html. Accessed 19 May 2017 153. Roth G (2006) Entwicklung neuer Hochdurchsatz-Assay-Technologien und Identifikation eines Ubiquitin-Liganden. Dissertation, Eberhard-Karls-Universität Tübingen 154. Molecular Devices GenePix® 4000B Microarray Scanner. https://www.moleculardevices.com/systems/microarray-systems/genepix-4000b. Accessed Jan 2017 155. Pyser SGI Stage Micrometers Calibration Standards. PS20 calibration slide. http://www.pyser-sgi.com/graticules/stage-micrometers-calibration-scales- grids/multi-image-calibration-slide-ps20/ps20-universal-calibration-slide-detail. Accessed 19 May 2017 156. newport OMM-6810B User's guide. https://assets.newport.com/webDocuments- EN/images/70034103_OMM-6810B_IX.pdf. Accessed 27 November 2015 157. Biyani M, Moriyasu J, Tanaka Y, Sato S, Ueno S, Ichiki T (2013) Microintaglio Printing of In situ Synthesized Proteins Enables Rapid Printing of High-Density Protein Microarrays Directly from DNA Microarrays. Appl. Phys. Express 6(8 87001):1–6 158. Glick Y, Avrahami D, Michaely E, Gerber D (2012) High-throughput Protein Expression Generator Using a Microfluidic Platform. JoVE(66):1–6 159. Keyence Deutschland GmbH Kompaktes Fluoreszenz-Mikroskop HS Modellreihe BZ- 9000 (BIOREVO). Accessed 19 May 2017 160. Balamurugan S, Obubuafo A, McCarley RL, Soper SA, Spivak DA (2008) Effect of Linker Structure on Surface Density of Aptamer Monolayers and Their Corresponding Protein Binding Efficiency. Anal. Chem. 80(24):9630–9634 161. DNA/RNA Copy Number Calculator. http://www.endmemo.com/bio/dnacopynum.php. Accessed 19 May 2017

167

Acknowledgements

I am very grateful to the former representatives of the HSG-IMIT, Prof. Dr. Reinecke, Prof. Dr. Manolis and Prof. Dr. Zengerle who accepted me in 2008 for academic research at HSG-IMIT in the area of microfluidic Lab-on-a-chip systems, after having spent 15 years in industrial R&D, mainly mechatronical and software engineering, since having accomplished my master thesis in 1993. During approximately 2 years, I could get familiar with some essential aspects of microsystem technologies (special thanks to Ludwig Gutzweiler and the IMTEK RSC team), especially for realizing microfluidics for point-of-care applications, and with fluorescence based read-out technologies for assay analysis.

During this time I met Dr. Günter Roth with his visionary idea of ‘Microarray Copying’. Fascinated by this idea, at the same time not being capable to assess the impact of this technology, I got trapped more and more and finally got the chance to completely work in his ‘Microarray group’ first at the IMTEK/APP (Applications Laboratory at Institute for Microsystems Technology, University of Freiburg), later, from 2013 at the ZBSA (Centre for Biological Systems Analysis, University of Freiburg) in his own group. Günter allowed me to build very first, basic microfluidic devices for protein copying, he encouraged me to leave stable ground and move into biological topics like PCR, protein expression, genetic engineering, blotting and the huge area of surface chemistry. I am very happy and grateful to Günter, that with probably all his patience I could dive into bio-technology and such being capable to tremendously broaden my knowledge. He steadily encouraged me for this thesis, finally I hope it will be to the benefit of him and our whole group.

I like to thank all group members of the AG Roth, Normann Kilb, Tobias Herz, Philipp Maier, Johannes Wöhrle, Stefan Krämer and Christin Rath with their various backgrounds for encouraging discussions and the daily possibility of learning to analyse and solve problems in very different manners; this multi-disciplinary group is generating such a lot of creativity and ensuring not to be trapped by narrow thoughts. I especially thank all the students who contributed with their diploma, bachelor and master thesis to this work, namely David Lämmle as the pioneer of RIfS in our group, Suleman Shakil with his broad engineering skills for microfluidics and electronics, Michael Fakler with his elaborate capability of designing and realizing precise mechanical systems, Linda Rudmann for creating the first Java based image processing application and pioneering spot identification, Nessim Ben Ammar for simulating and elaborating our temperature control system. Beyond, I thank all members of our cooperation partner Biametrics GmbH for providing their RIfS technology and supporting us throughout the years with their knowledge, engagement and trainings at their site; further I thank Maya Temerinac-Ott for assisting in Matlab programming and Bence Melicuty supporting with advanced mathematics.

I thank my family, Yesim, Selina and Denis, my parents and relations for love and support – especially for tolerating my ambitions and my fascination for the research work.

Respectfully and sincerely, I am grateful to my assessors Prof. Dr. Gerald Urban and Prof. Dr. Günter Gauglitz for their engagement and support of this dissertation.

168

Curriculum

Name: Jürgen Burger

Address: Hinterhofstraße 39, 79211 Denzlingen

Contact: [email protected]

Date / place of birth: 23.09.1966 / Freiburg

Nationality: German

Education

1983 - 1986 Apprenticeship as Precision Mechanic, Sick AG Waldkirch.

1989 - 1992 Studies Precision Eng., Fachhochschule Esslingen, degree in 7/1992.

1992 - 1993 MSc studies Information Eng., City University London, degree in 11/1993.

Employment:

1987 - 1988 Precision Mechanic - Sick AG; military service.

1994 - 2008 Mechatronical / software engineer - Daimler-Benz-Aerospace AG, Niles GmbH, Buro GmbH.

2009 - 2010 R&D at HSG-IMIT in Freiburg for microfluidic lab-on-a-chip systems, design of experimental setups.

2011 - 2017 R&D / Dissertation at University of Freiburg, IMTEK/Sensors and ZBSA

Automated system for the cell-free protein microarray synthesis and the label- free molecule-protein interaction analysis

Teaching:

2010 - 2012 Cleanroom student trainings at IMTEK / University of Freiburg.

2009 - 2014 Supervision of diploma / bachelor / master theses at University of Freiburg.

169

Grant:

Burger et al., Entwicklung eines einfachen Handgerätes zur Herstellung von Protein- Mikroarrays aus DNA-Mikroarrays, Ideenwettbewerb 2011 (first phase – proof-of-concept) / 2012-2015 (second phase – R&D towards commercialisation), Stuttgart, (https://www.ptj.de/Ideenwettbewerb)

 major financial support for dissertation

VIP+ Förderung durch das BMBF (FKZ 03VP01200, 08/2016-07/2019) “Entwicklung eines Biomolekül-Kopierers zur Erzeugung neuartiger Next Generation Microarrays”

 financial support for myself and the whole team

Patent:

Patent for microfluidic handheld for protein expression (WO 2012/104399A3, submitted via University of Freiburg, co-inventor Dr. Günter Roth)  granted for Europe, China, US+Canada, (Japan pending)

170

Appendix

A1 Materials for surface chemistry

Table A1: materials for surface activations. component component description supplier Acetone 2-Propane Carl Roth Isopropanol Isopropyl alcohol Carl Roth

H2O2 Hydrogene peroxide 30 % Carl Roth

H2SO4 Sufuric acid 95-98 % Carl Roth DMF N,N’-Dimethylformamide Sigma-Aldrich Detergent Neodisher Laboclean A8 Bruno Kummer GmbH DI Bi-distilled water Carl Roth GOPTS (3-Glycidyloxypropyl)triethoxysilane Sigma-Aldrich AMD Amine-dextrane, 100 kDA, 50 % functionalized Innovent e.V., D-Jena GA Glutaric anhydride Sigma-Aldrich DIC N,N’-Diisopropyl-carbodiimide Sigma-Aldrich NHS N-Hydroxysuccinimide Sigma-Aldrich TACN 1,4,7-Triazacyclononane Sigma-Aldrich

NiCl26H2O Nickel(II)-chloride hexahydrate Carl Roth

A2.1 DNA data

1) A07 complete construct

insertion_seq 96..103 /vntifkey="14" /label=sequence\introduced\by\PET7/F promoter 20..36 /vntifkey="30" /label=T7-Promoter misc_feature 75..83 /vntifkey="21" /label=g\10 misc_feature 86..91 /vntifkey="21" /label=RBS primer_bind 1..15 /vntifkey="28" /label=RTST7/B insertion_seq complement(79..95) /vntifkey="14" /label=binding\portion\of\PET7/F misc_feature 176..205 /vntifkey="21" /label=c-MYC\tag misc_feature 140..169 /vntifkey="21" /label=c-MYC\tag misc_feature 104..133 /vntifkey="21" /label=c-MYC-tag /note="EQKLISEEDL" primer_bind complement(184..223) /vntifkey="28" /label=myc-NcoL/F CDS 674..694 /vntifkey="4" /label=linker:\LEGGGSA CDS 695..763 /vntifkey="4" /label=2\(His6)\-Tag terminator 810..846 /vntifkey="43" /label=T7\terminatior misc_signal 764..769 /vntifkey="87" /label=StopStop primer_bind complement(866..883) /vntifkey="28"

172

/label=Tterm/F /note="Alternative downstream primer (without single primer overhang)" insertion_seq complement(884..898) /vntifkey="14" /label=Single\primer\overhang\=\RTST7/B\seq primer_bind 674..709 /vntifkey="28" /label=Link-2H/B /note="Upstream Primer" primer_bind complement(866..898) /vntifkey="28" /label=T-term-tag/F /note="downstream primer" CDS 242..655 /gene="CRABP2" /codon_start=1 /protein_id="CAG29353.1" /db_xref="GI:47496661" /db_xref="GDB:134819" /db_xref="GOA:P29373" /db_xref="HGNC:2339" /db_xref="InterPro:IPR000463" /db_xref="InterPro:IPR000566" /db_xref="InterPro:IPR012674" /db_xref="PDB:1BLR" /db_xref="PDB:1BM5" /db_xref="PDB:1CBQ" /db_xref="PDB:1CBS" /db_xref="PDB:1XCA" /db_xref="PDB:2CBS" /db_xref="PDB:2FR3" /db_xref="PDB:2FRS" /db_xref="PDB:2FS6" /db_xref="PDB:2FS7" /db_xref="PDB:2G78" /db_xref="PDB:2G79" /db_xref="PDB:2G7B" /db_xref="PDB:3CBS" /db_xref="UniProtKB/Swiss-Prot:P29373" /vntifkey="4" /label=CRABP2 gene 242..655 /gene="CRABP2" /vntifkey="60"

173

source 242..655 /organism="Homo sapiens" /mol_type="mRNA" /db_xref="taxon:9606" /clone="RZPDo834B093D" /clone_lib="Human Full ORF Clones Gateway(TM) - RZPD" /lab_host="DH10B" /vntifkey="98" /note="Vector: pDONR201, Site_1: attP1; Site_2: attP2" BASE COUNT 269 a 205 c 249 g 175 t ORIGIN 1 gatctcgatc ccgcgaaatt aatacgactc actataggga gaccacaacg gtttccctct 61 agaaataatt ttgtttaact ttaagaagga gatatccacc atggagcaaa agctcatttc 121 tgaagaggac ttgaatggag aacagaaatt gatcagtgag gaagacctca acggtgagca 181 gaagttaata tccgaggagg atcttggctc tagtgccatg gctggcaccc ctggtccagg 241 tatgcccaac ttctctggca actggaaaat catccgatcg gaaaacttcg aggaattgct 301 caaagtgctg ggggtgaatg tgatgctgag gaagattgct gtggctgcag cgtccaagcc 361 agcagtggag atcaaacagg agggggacac tttctacatc aaaacctcca ccaccgtgcg 421 caccacagag attaacttca aggttgggga ggagtttgag gagcagactg tggatgggag 481 gccctgtaag agcctggtga aatgggagag tgagaataaa atggtctgtg agcagaagct 541 cctgaaggga gagggcccca agacctcgtg gaccagagaa ctgaccaacg atggggaact 601 gatcctgacc atgacggcgg atgacgttgt gtgcaccagg gtctacgtcc gagagccagg 661 cggcgaccca gctctcgagg gtggcggtag cgcacatcac catcaccatc actctagagc 721 ttggcgtcac ccgcagttcg gtggtcacca ccaccaccac cactaataaa aaaaaaaaaa 781 aaaaaaaaaa aaaaaaaccg ctgagcaata actagcataa ccccttgggg cctctaaacg 841 ggtcttgagg ggttttttgc tgaaaggagg aactatatcc ggacgcggga tcgagatc //

2) D07 (only protein encoding sequence available) LOCUS CR456895_TNFSF10 846 bp DNA linear PRI 12-MAY-2009 DEFINITION Homo sapiens full open reading frame cDNA clone RZPDo834F0111D for gene TNFSF10, tumor necrosis factor (ligand) superfamily, member 10; complete cds, incl. stopcodon. ACCESSION CR456895 VERSION CR456895.1 GI:48145906 KEYWORDS Full ORF shuttle clone, Gateway(TM), complete cds. SOURCE Homo sapiens (human). ORGANISM Homo sapiens Eukaryota; Metazoa; Chordata; Craniata; Vertebrata; Euteleostomi; Mammalia; Eutheria; Euarchontoglires; Primates; Haplorrhini; Catarrhini; Hominidae; Homo. REFERENCE 1 (bases 1 to 846) AUTHORS Ebert,L., Schick,M., Neubert,P., Schatten,R., Henze,S. and Korn,B. TITLE Cloning of human full open reading frames in Gateway(TM) system entry vector (pDONR201)

174

JOURNAL Unpublished REFERENCE 2 (bases 1 to 846) AUTHORS Ebert,L., Schick,M., Neubert,P., Schatten,R., Henze,S. and Korn,B. TITLE Direct Submission JOURNAL Submitted (03-JUN-2004) RZPD Deutsches Ressourcenzentrum fuer Genomforschung GmbH, Im Neuenheimer Feld 580, D-69120 Heidelberg, Germany COMMENT RZPD; RZPDo834F0111D, ORFNo 1231 www.rzpd.de/cgi-bin/products/cl.cgi?CloneID=RZPDo834F0111D RZPDLIB; Human Full ORF Clones Gateway(TM) - RZPD (kan-resist.) RZPD LIB No. 834 www.rzpd.de/cgi-bin/products/showLib.pl.cgi/response?libNo=834 www.rzpd.de/products/orfclones/ Contact: Inge Arlart RZPD Deutsches Ressourcenzentrum fuer Genomforschung GmbH, Heubnerweg 6, D-14059 Berlin, Germany Tel: +49 30 32639 100 Fax: +49 30 32639 111 www.rzpd.de This clone is available from RZPD; contact RZPD ([email protected]) for further information. This CDS clone is a part of a collection of human full length expression clones generated by RZPD. This CDS has been cloned incl. stopcodon. The CDS has been inserted into pDONR201 via a BP Clonase(TM) reaction. Additional sequence has been added in front of the start codon: att..AAAAAA GCA GGC (ATG). The last base of the last coding triplett has been changed to T, which might lead to an amino acid change at the C terminus of the polypeptide. The stop codon has been set to TAA followed by TTAACCCAGCTTTCTT..att. Compared to the reference sequence NM_003810 we did not find any amino acid exchanges. Clone distribution: http://www.rzpd.de/products/orfclones/. COMMENT This file is created by Vector NTI http://www.invitrogen.com/ COMMENT ORIGDB|GenBank COMMENT VNTDBDATE|518116495| COMMENT LSOWNER| COMMENT VNTNAME|CR456895_TNFSF10| FEATURES Location/Qualifiers source 1..846 /organism="Homo sapiens"

175

/mol_type="mRNA" /db_xref="taxon:9606" /clone="RZPDo834F0111D" /clone_lib="Human Full ORF Clones Gateway(TM) - RZPD" /lab_host="DH10B" /vntifkey="98" /note="Vector: pDONR201, Site_1: attP1; Site_2: attP2" gene 1..846 /gene="TNFSF10" /vntifkey="60" CDS 1..846 /gene="TNFSF10" /codon_start=1 /protein_id="CAG33176.1" /db_xref="GI:48145907" /db_xref="GOA:Q6IBA9" /db_xref="HGNC:11925" /db_xref="InterPro:IPR006052" /db_xref="InterPro:IPR008983" /db_xref="InterPro:IPR017355" /db_xref="UniProtKB/TrEMBL:Q6IBA9" /vntifkey="4" /label=TNFSF10 BASE COUNT 273 a 167 c 193 g 213 t ORIGIN 1 atggctatga tggaggtcca ggggggaccc agcctgggac agacctgcgt gctgatcgtg 61 atcttcacag tgctcctgca gtctctctgt gtggctgtaa cttacgtgta ctttaccaac 121 gagctgaagc agatgcagga caagtactcc aaaagtggca ttgcttgttt cttaaaagaa 181 gatgacagtt attgggaccc caatgacgaa gagagtatga acagcccctg ctggcaagtc 241 aagtggcaac tccgtcagct cgttagaaag atgattttga gaacctctga ggaaaccatt 301 tctacagttc aagaaaagca acaaaatatt tctcccctag tgagagaaag aggtcctcag 361 agagtagcag ctcacataac tgggaccaga ggaagaagca acacattgtc ttctccaaac 421 tccaagaatg aaaaggctct gggccgcaaa ataaactcct gggaatcatc aaggagtggg 481 cattcattcc tgagcaactt gcacttgagg aatggtgaac tggtcatcca tgaaaaaggg 541 ttttactaca tctattccca aacatacttt cgatttcagg aggaaataaa agaaaacaca 601 aagaacgaca aacaaatggt ccaatatatt tacaaataca caagttatcc tgaccctata 661 ttgttgatga aaagtgctag aaatagttgt tggtctaaag atgcagaata tggactctat 721 tccatctatc aagggggaat atttgagctt aaggaaaatg acagaatttt tgtttctgta 781 acaaatgagc acttgataga catggaccat gaagccagtt ttttcggggc ctttttagtt 841 ggttaa //

176

A2.2 DNA amplification and expression

Figure A2.2.1: PCR for A07 construct generation using long primers both with HiFi and Taq polymerase. Amplification of A07 with both polymerases visible but stronger band with Taq polymerase (blots by Dr. Oda Stoevesandt resulting from common experiments at BBT, Cambridge in August 2011).

Figure A2.2.2: in-vitro expressed erDNA templates with control DNA. Western blots with ab-myc and ab-His. (blots by Dr. Oda Stoevesandt resulting from common experiments at BBT, Cambridge in August 2011).

177

A2.3 Aptamers Aptamers are short, single-stranded DNA (ssDNA) or RNA (ssDNA) oligonucleotides (typically < 100 bp) which became increasing interest and importance in many research and medical applications as substitutes for antibodies (e.g. analysis, diagnostics, therapeutics, drug discovery, biomarker discovery).

Table A2.3.1: sequences of aptamers. A14 as fibrinogen-binding-site aptamer elongated with a T20 spacer [160] will be especially analyzed within this thesis.

178

A2.4 DNA arraying and immobilization

Protocol 1 (established at IMTEK/APP in 2009/10):

1) Materials a. Spotting buffer i. sodium buffer (2x), 5 µl ii. RNAse free water, 4 µl b. Control DNA 97, FITC labelled (1µg/µl) c. erDNA A07 / D07 (> 100 ng/µl recommended from O. Stoevesandt / BBT Ltd.) 2) Spotting with pipette or PipeJet [152]or TopSpot [148] 3) Storage with wet cloth in petri dish sealed with parafilm 4) Incubation over night (12h) 5) Washing with PBS or SSC buffer (few minutes at room temperature) 6) Immobilization control by fluorescence scanning

Protocol 2 (from Oda Stoevesandt / BBT Ltd.):

1) Materials a) Spotting buffer - phosphate buffer (6x, pH 8.5), 50 mM; made of Natriumdihydrogenphosphat-monohydrat (NaH2PO4*H2O), and Gibco water b) e.g. template DNA p53 (from BBT Ltd.); 1627 bp, 86 ng/µl i. 1/6 spotting buffer  DNA spotting concentration of 72.5 ng/µl ii. 10x 0.4 nl/spot  4 nl/spot  0.288 ng  1.64 x 10^8 dsDNA [161] c) e.g. template DNA notch (from BBT Ltd.); (unkown) bp, 64 ng/µl i. 1/6 spotting buffer  DNA spotting concentration of 53.3 ng/µl ii. 7x 0.4 nl/spot  2.8 nl/spot  0.149 ng  (unknown) dsDNA 2) Spot erDNA solution at 65% humidity onto matrix by GeSIM Nano-Plotter 2.1 [151] a) Chess board pattern, distance between spot centres of 1.06 mm in both directions (diagonal distance of 1.5 mm) 3) Incubate over night at RT in humid atmosphere (box with buffer cloth) 4) Incubate 30 min at 60 °C (strengthens immobilization) 5) Wash on rocking platform in a) 0.1 % Triton X100 (RT, 5 min) b) 1 mM HCl (RT, 2x 2 min) c) 100 mM KCl (RT, 10 min)

d) H2O (RT, 1 min) e) Block remaining epoxy groups 50 mM Ethanolamin, 0.1 M Tris, pH9, 60 °C, 15 min

6) Rinse matrix with milliQ H2O 7) Dry by centrifugation or with nitrogen, not too long – DNA should not dry out! 8) Immobilization control by fluorescence scanning

179

A3 Materials for flow cells

A3.1 PMMA flow cell (concept 1)

Table A3.1: materials for PMMA flow cell (concept 1). material component description supplier Polyester Adhesive foil, 100 µm thickness 3M PMMA Poly(methyl methacrylate), plates for Märtin, Freiburg holder PC Polycarbonate, microscope slide format Märtin, Freiburg of 75x25x1 mm³ Epoxy coated glass Microcope slide of 76x26 mm², epoxy Schott / Peqlab introduces DNA binding sites Ni-NTA coated glass Microcope slide of 76x26 mm², Ni-NTA SMS introduces protein binding sites 1.4305 Screws DIN 963 A2 M 3x10 Elpe, Freiburg Laser cutter PLS3.60 / 40 W Universal laser system inc.

A3.2 PDMS flow cell (concept 2)

Table A3.2: materials for flow cells made of PDMS and glass slides; fabrication systems. material component description supplier PTFE Polytetrafluoroethylene Märtin, Freiburg PDMS, red-brown, Polydimethylsiloxane, RT607 Reiff Technische Produkte A/B, 55A GmbH (Wacker Chemie AG) PDMS, transparent Polydimethylsiloxane, Sylgaard 184 Sigma-Aldrich Co. LLC. Matrix array 1 PDITC coated PDMS inlay in standard self-made microscope slide format (DNA matrix) Matrix array 2 Ni-NTA coated standard microscope self-made glass slide (protein matrix) TMMF S2030 Dry film resist, 30 µm thickness Tokyo Ohka Kogyo Co., Ltd Ordyl SY 330 Dry film resist, 20 µm thickness Elga Europe, Italy SU-8 3000 Negative photo resist MicroChem Corp. Centrifuge Hettich For moulding PDMS into cavity made of Andreas Hettich GmbH & Co. 320R structured 4” Si wafer sealed by O-rings KG and cap

A3.3 Booklet flow cell (concept 3)

Table A3.3: materials of booklet type flow cell. material component description supplier PDMS, red-brown, Polydimethylsiloxane, RT607 Reiff Technische Produkte A/B, 55A GmbH (Wacker Chemie AG) Matrix array 1 PDITC coated standard microscope self-made glass slide (DNA matrix) Matrix array 2 Ni-NTA coated standard microscope self-made glass slide (protein matrix) Basic clamping unit Aluminium (AlMgSi0,5) part shaped by HSG-IMIT / 78052 Villingen- milling; anodized Schwenningen

180

Specific connection Stainless steel (1.4305) parts shaped by Sutter Medizintech. GmbH / elements turning 79110 Freiburg Standardized Stainless steel A2 (1.4301) Heinrich Kipp Werk KG 72168 mechanical elements Sulz/Neckar Srews and nuts Stainless steel A2 (1.4301) Elpe GmbH, Freiburg Frames Stainless steel (1.4310) Wagner Lasertechnik GmbH 78083 Dauchingen Eccentric, ledge Brass (CuZn40Pb3) HSG-IMIT / 78052 VS Cover PMMA HSG-IMIT / 78052 VS Gears, stops, bushes POM IMTEK, 79110 Freiburg Spacer Stainless steel foil (1.4310) Fa. Hasberg GmbH 83230 Bernau; RS Präzisionstechnik 79211 Denzlingen PTFE Polytetrafluoroethylene, coating for IMTEK, 79110 Freiburg; spacer (Teflon) Nanosol AG / 9496 Balzers Lichtenstein PEEK Polyether ether ketone, material of Omnifit / England fittings

A3.4 Glass slides

Table A3.4: glass slides as iRIf transducers. name substrate layer 1 layer 2 supplier Standard microscope slide unknown no no Carl Roth Interference glass (IF) D263 Ta2O5, n500 = 2.2, SiO2, n500 = 1.46, d Biametrics GmbH n500 = 1.52 d = 10 nm = 330 nm / Schott AG Goethe glass BK7 Ta2O5, n500 = 2.2, SiO2, n500 = 1.46, d Biametrics GmbH n500 = 1.52 d = 45 nm = 20 nm / Schott AG

A4 Materials for automated systems

A4.1 Optical components

Table A4.1.1: technical data of major detection components for prototype iRIf setup. Data leads to an image resolution of 15 µm at an image size of 24 mm x 18 mm; due to increasing distortions of the lens evaluable object size should not exceed 17.5 mm x 13 mm (distortion < 0.3 %). Supplier Component technical data Article no. pco.1600 C-mount, CameraLink, 1600x1200, 14 bit, 30 fps, sensor PCO size 12.2 mm x 9 mm 80108000080 Telecentric lens correctal TL, 0.502, 117 mm focus length, 0.3 % distortion for object size of 17.5 mm x 13.1 mm Sill Optics --> image resolution of 15 µm S5LPL2060 λ/2 waveplate, adapter D 60 mm, quartz crystalline, no preferred direction, 90% transmission @ 400-800 nm, Sill Optics aperture of 48 mm. S5SET1199/060 Thorlabs Holder for lens system AB90

181

Highpower LED 1W grün, Typ "WTN-1W-75g" 1 Watt, ca. World Trading Net 525nm, ca. 75 lm, 3,5V/350 mA, ca.120° 77900050 Highpower LED 1W blau, Typ "WTN-1W-20b" 1 Watt, ca. World Trading Net 470 nm, ca. 20 lm, 3,5V/350 mA, ca. 120 77900060 LedStar GmbH (Energy SAVE GmbH) LED current controller HKO350 HKO350

Table A4.1.2: technical data of optical detection components for enhanced iRIf setup. Data leads to an image resolution of 7 µm at an image size of ~17 mm x 15 mm; due to increasing distortions of the lens evaluable object size should not exceed 13.8 mm x 10.3 mm (distortion < 0.2 %). Supplier Component technical data Article no. pco.edge 5.5 C-mount, CameraLink, 2560x2160, 16 bit, 100 fps, sensor size 16.6 PCO mm x 14 mm 80108000150 Telecentric lens correctal TL, 0.926, 100 mm focus length, 0.2 % distortion for object of 13.8 x 10.3 mm² Sill Optics --> image resolution of 7 µm S5LPL1005 λ/2 waveplate (type 1) adapter D 60 mm, quartz crystalline, no preferred direction, 90% Sill Optics transmission @ 400-800 nm, aperture of 48 mm; S5SET1199/060

A4.2 Elaborated prototype of automated system Table A4.2.1: technical data of fluidic system tubes, survey of dead volumes. Tube Tube name / connectors Tube inner Tube length Tube volume location diameter [mm] [mm] [µl] A Sample loop / 5.1 – 5.3 1.0 1460 1147 B Reagent tube / 4.2 - Reagent 0.3 320 23 C Buffer output / Vici output – 4.6 1.0 320 251 D Flow cell tube / 4.5 – flow cell 0.3 387 27 E … / Buffer – Vici-Input 1.0 330 259 F … / Syringe system output – 4.3 1.0 196 154 G Valve connector / 5.2 – 4.4 0.3 160 11 H Valve connector / 5.4 – 4.1 0.3 112 8 I … / System fluid - Syringe input 1.0 285 224 J … / Syringe output - waste 1.0 410 322

182

Table A4.2.2: survey of dead volumes of valves. Valve Valve name / order no. Type Tube length Valve total dead [mm] volume [µl] 4-4 Loop Flow Path 4-ports / 36654 HVM 1.5 14 6-6 Loop Flow Path 6-ports/ 36781 HVM 1.0 7.5 FCI Flow cell inlet 2

Table A4.2.3: fluidic and electro-mechanical components. Supplier Component technical data Article no. Syringe pump PSD/4, CER VAL 3-5 ALU BODY PSD4, 54848-01 / H7991-01 Hamilton SYR, 1005.5TLL, 2.5ML, 30MM / H8250-45 Hamilton Valve / MVP, 6 Port loop 159698/36781 Hamilton Valve / MVP, 4 Port loop 159698/36766 Vici Pump M6 (incl. controller + power supply) CP2-4841-100M1 CP M SNT 250W 24V Weidmüller (Conrad) Power supply PRO-M 10A UW Vertical axis, EGSK-20-25-1P / CMMO-ST-C5-1-DION 562758 / 1512317 / Festo / EMMS-ST-28-L-SE, 50 mm 1430663 Festo X-Y-table, EXCM10, 150 x 110 mm² 1801920 Fischer Stone plate Impala, 370 x 120 x 50 mm³ VFK-000977 Lupberger / Malzew Base plate 450 x 450 x 20 mm³, AlMg4,5Mn, ALPLAN own design Handheld with magnetic clamping system, sheet AG Roth metal 1.4021 own design Flow cell temperature control - peltier elements, AG Roth PT100 sensor, controller own design Sample cooler - Peltier elements, PT100 sensor, AG Roth controller, Al-block for Eppendorf tubes, housing own design Bilz Vibration Tech. Damper B13W/B8 (2-fach), 50x50x13 / 8 mm³ B13-W/50X50# msscientific 1/16" Peek, connector 002145 + 001544 Tubes, Teflon, OD 1,6 mm, ID 1,0 mm 008T16-100-20 msscientific Tubes, Teflon, OD 1,6 mm, ID 0,3 mm 008T16-030-20 Screw, Inch, 1/4"-20; UNC933A41/45/8; Screw, Inch, 6-32; UNC933A263/8; Dunken (Händler) Screw, Inch, 8-32; UNC933A283/8 Malzew Mechanical parts for own designs AG Roth Plastic parts for waste container + washing station self made

183

Table A4.2.4: standardized protocols for the binding interaction analysis or for in situ protein expression can be parameterized via Excel chart – flow rates and priming times can be chosen individually for buffers and reagents, further the flow cell temperature within a range from 8 °C to 40 °C can be selected. The sampler cooler can be switched on/off, the tube position on the sampler is coded according to the MTP plate positions. The volume of each reagent to be provided is indicated (column H).

Table A4.2.5: experiment numbers of series JB100. 0.5 µg/ml 1 µg/ml 2 µg/ml 5 µg/ml 10 µg/ml data set 1 15 27 21 17 28 data set 2 16 25 22 11 29

184

A5 Mathematical algorithms

A5.1 Curve fitting by non-linear least-square method with function f(xdata) = B - A .* exp(-kobs * xdata)

Matlab code:

function [estimates, model] = fitcurve_jb4(xdata, ydata,satval)

% Calculate appropriate starting point for fminsearch

STP_B = satval;

start_point = [ydata(1) 0.01 STP_B];

model = @expfun;

% Calculate appropriate starting point for fminsearch

[estimates] = fminsearch(model, start_point, optimset('MaxFunEvals', 10e4, 'TolX',1e-16, 'MaxIter', 10e4, 'TolFun', 1e-16));

function [sse, FittedCurve] = expfun(params)

A = params(1);

kobs = params(2);

B = params(3);

FittedCurve = B - A.* exp(-kobs * xdata);

ErrorVector = FittedCurve - ydata;

sse = sum(ErrorVector .^ 2);

end

A5.2 Non-linear least squares method A nonlinear model is defined as an equation that is nonlinear in the variables, or a combination of linear and nonlinear in the variables, e.g. ratios of polynomials, power functions. In matrix form, nonlinear models are given by the formula

푌 = 푓(푥, 훩) + 휀 (1) where

Y is an n-dimensional vector of responses

185

f is a function of x and 훩

훩 is a m-dimensional vector of parameters

x is the n-by-m design matrix of model variables

ε is an n-dimensional vector of errors

The aim of nonlinear fitting is to estimate the parameter values which best describe the data. The standard way of finding the best fit is to choose the parameters that would minimize the squared deviations of the theoretical curve(s) from the experimental points. This method is called chi-square minimization, defined as follows [112]:

푛 2 푌푖 − 푓(푥푖, 훩̂) 휒2 = ∑ ( ) − −> 푚푖푛 (2) 휎푖 푖=1 where

휒2 is the summed squared error (SSE)

σi is the weighting factor

To estimate the 훩̂ value with the least square method, we need to solve the normal

2 equations which are set to be zero for the partial derivatives of Χ with respect to each 훩̂푝.

푛 2 휕휒 1 휕푓(푥푖′, 훩̂) = −2 ∑ [푌 − 푓(푥 ′, 훩̂)][ ] = 0 (3) ̂ 2 푖 푖 ̂ 휕훩푝 휎푖 휕훩푝 푖=1

Since there are no explicit solutions to the normal equations, we employ an iterative strategy to estimate the parameter values:

1. Determination of some initial values Θ0 given by heuristic approaches or random values on the interval [0,1]. 2. With each iteration, the 휒2 value is computed and then the parameter values are adjusted to reduce the 휒2. This computation involves the calculation of the Jacobian of f(x, 훩), which is defined as the matrix of partial derivatives, taken with respect to the parameters 훩. Adjust the parameters 훩 and determine whether the fit improves. The direction and magnitude of the adjustment depend on the fitting algorithm, e.g. Matlab and Origin provide: 186

Trust-region — This is the Matlab default algorithm and must be used if you specify parameter constraints. It can solve difficult nonlinear problems more efficiently than the other algorithms and it represents an improvement over the popular Levenberg-Marquardt algorithm.

Levenberg-Marquardt — This is the Origin default algorithm which has been used for many years and has proved to work most of the time for a wide range of nonlinear models and starting values. If the trust-region algorithm does not produce a reasonable fit, and you do not have parameter constraints, you should try the Levenberg-Marquardt algorithm.

3. Iterate the process by repeating step 2 until the fit reaches the specified convergence criteria. When the difference in 휒2 values computed in two successive iterations are small enough (compared with the tolerance), we can say that the fitting procedure has converged

Because of the nature of the approximation process, no algorithm is foolproof for all nonlinear models, data sets, and starting points. Therefore, if you do not achieve a reasonable fit using the default starting points, algorithm, and convergence criteria, you should experiment with different options. As nonlinear models can be particularly sensitive to the starting points, this should be the first fit option you modify [142].

A5.3 Parameter standard error calculation The calculation of the standard errors of each parameter of the nonlinear fitting function is derived from the method Origin applies [112]. During L-M iteration the partial derivatives matrix F is calculated, with its element in ith row and jth column:

휕푓푖(푥, 훩̂) 퐹푖푗 = (4) 휎푖휕훩̂푗

Next the mean residual variance, or the Deviation of the Model s2 is calculated:

RSS 푠² = (5) 푛 − 푝 where

RSS is the summed squared error (SSE)

n is the number of discrete values

187

p is the number of parameters

Then we can get the Variance-Covariance matrix for parameters 훩̂ by:

퐶 = (퐹′퐹)−1푠2 (6)

The square root of a main diagonal value of this matrix is the Standard Error of the corresponding parameter

푠훩𝑖 = √푐푖푖 (7) where

cii is the element in i-th row and i-th column of the matrix C.

The parameter standard errors give us an idea of the precision of the fitted values. Typically, the magnitude of the standard error values should be lower than the fitted values. If the standard error values are much greater than the fitted values, the fitting model may be over parameterized [112].

Instructions for the implementation:

i. Determination of nonlinear fitting function:

f(x) = B – A*exp(-C*x) with parameters A, B, C

ii. Calculation of parameters partial derivations matrix F: a. df(x)/dA = -exp(-C*x); b. df(x)/dB = 1 c. df(x)/dC = -A*(-x)*exp(-C*x) = A*x*exp(-C*x) iii. Derivatives matrix with no weighting (σ = 1):

F = [-exp(-C*x), 1, A*x*exp(-C*x)]

with each matrix element being a n-dimensional vector of discrete elements

for x, e.g. ith row of F = [-exp(-C*xi), 1, A* xi *exp(-C* xi)]

iv. Calculation of mean residual variance (model deviation): s² = sse/(no_values-no_parameters)

188

where

no_values = length of x values vector

no_parameters = number of parameter

sse = summed squared errors

v. Determination of Variance-Covariance-Matrix C by multiplication of the inverse matrix of the vector product of the transposed matrix F and the matrix F with the model deviation s²:

C = (F'F)-1s²

vi. Standard error matrix sTheta is square root of main diagonal values of Var-

CoVar-matrix C (cii):

1/2 푠훩𝑖 = cii

Parameter standard error determination of function f(xdata) = B - A .* exp(-kobs * xdata)

Matlab code (equal data, no weighting):

function [std_err] = param_stderr(params, sse)

A = params(1);

kobs = params(2);

B = params(3);

no_param = size(params,2);

% partial derivatives with respect to parameters

df_dA = -exp(-kobs*xdata);

df_dB = ones(size(df_dA));

df_dkobs = A*xdata.* exp(-kobs * xdata);

df_dC = df_dkobs;

% derivations matrix with no weighting (sigma = 1)

F = [df_dA, df_dB, df_dC];

% calculation of mean residual variance (model deviation)

mr_var = sse/(size(xdata,1)- no_param);

189

% determination of Variance-Covariance-Matrix co_var by multiplication of the inverse matrix of the vector product of the transposed matrix F and the matrix F with the model deviation

co_var = inv(F'*F)*mr_var;

% calculation of standard errors of corresponding parameters std_err as square root of main diagonal values of matrix C (cii)

std_err = sqrt(diag(co_var));

end

A6 Fluorescent scans belonging to Fig. 4.54

0.5 µg/ml Streptavidin-Cy5 1 µg/ml Streptavidin-Cy5 2 µg/ml Streptavidin-Cy5

Slide 033 Slide 027 Slide 022

5 µg/ml Streptavidin-Cy5 10 µg/ml Streptavidin-Cy5

Slide 011 Slide 028

Figure A6: fluorescence end-point measurements of streptavidin-Cy5 bound to bBSA spots.

190