MAGNETIC ISOLATION OF FECAL INDICATING BACTERIA USING BIO- FUNCTIONALIZED MAGNETIC MICRODISCS FOR WATER QUALITY MONITORING

By

KEISHA YARIE CASTILLO-TORRES

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2020

© 2020 Keisha Yarie Castillo-Torres

To my beloved parents

ACKNOWLEDGMENTS

I would like to thank the University of Florida (UF) Office of Technology Licensing

(OTL) and the US Army Medical Research and Materiel Command under SBIR Contract led by Innovative Space Technologies, LLC for their financial support for this project.

I also thank my supervisory committee: Dr. Y.K. Yoon, Dr. Carlos Rinaldi, Dr. Eric

McLamore, and Dr. David Arnold for their support, guidance, and ideas throughout this process. Special thanks to Dr. Eric McLamore for his continuous help and guidance on the biological concepts and experiments. Also, my immense gratitude goes to my supervisory committee chair and advisor, Dr. David Arnold for giving me the opportunity to work in his group and for his continuous and valuable support, guidance, advice, mentorship, and patience.

I would also like to thank staff from the University of Florida (UF) Nanoscale

Research Facility (NRF) for their assistance during the microfabrication process and the staff from the Electron Microscopy core in the UF Interdisciplinary Center for

Biotechnology Research (ICBR) for their assistance during confocal imaging.

Special thanks go to my lab mates and friends from Dr. Arnold’s group: Dr.

Nicolas Garraud, Dr. Camilo Velez, Dr. Alexandra Garraud, Dr. Xiao Wen, Yuzheng

Wang, Connor Smith, and Beatriz Jimenez, whose help, support, feedback, friendship, and/or positive vibes have helped throughout this path. Also, special thanks to Dr. Fan and his students, specially Jingling Zhang, Kangfu Chen, Jose Varillas, Pablo Dopico, and Minh-Chau Le for their help and availability to help, teach, and let me use their fluorescence scope, and microfluidic fabrication tools. As well as Dr. Rinaldi and his students, especially Angelie Rivera-Rodríguez and Melissa Cruz-Acuña for their help.

Finally, special thanks go to my “cubicle row-mates” and friends from the

4

Interdisciplinary Microsystems Group (IMG), specifically, Brittney, Justin, Ashley, and

Brett, whose support, friendship, and/or positive vibes have also helped on a daily basis.

Also, I would like to deeply thank professors and mentors that encouraged me to consider, pursue, and/or continue my doctoral degree, specifically Dr. Nayda Santiago,

Dr. Domingo Rodríguez, Dr. Néstor Rodríguez, and Dr. María Vera.

Special thanks go to my long-distance friends (Mónica, Kathyria, Sharlene,

Débora, Lorraine, and Deneiz) for their support, motivation, and positive vibes when I needed them the most (even when miles away). But most specially, I would like to thank my dearest friend Sylmarie Dávila-Montero, for her unconditional and constant support and feedback, and for walking this path with me from the very beginning when we were lab mates until today when we are miles away pursuing our PhD degrees.

Immeasurable thanks to my parents, Irma L. Torres-Plaza and Luis A. Castillo-

Dávila, whose endless love, understanding, and support have meant the world to me throughout every single step of my personal, academic, and professional paths.

Also, I will be forever grateful with my loving husband, and best friend, Adail A.

Rivera-Nieves, whose unconditional love, motivation, understanding, and support have made this path easier by just listening, celebrating even the smallest of my victories, and for walking this path right by my side while working towards his own PhD degree.

Finally, and most importantly, I would like to thank God for giving me the strength and persistence to start, continue, and complete my PhD degree.

5

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 9

LIST OF FIGURES ...... 10

LIST OF ABBREVIATIONS ...... 15

ABSTRACT ...... 16

CHAPTER

1 INTRODUCTION ...... 18

1.1 Overview ...... 18 1.2 Motivation ...... 18 1.3 Previous Works on Bacteria Detection ...... 20 1.3.1 Standard Methods ...... 20 1.3.2 Bio-Nano-Magnetic Methods ...... 22 1.3.3 Microfluidic Magnetic Separation Technologies ...... 28 1.4 Research Approach ...... 31 1.4.1 Objectives ...... 32 1.4.2 Methodology ...... 32 1.5 Dissertation Overview ...... 34

2 BACKGROUND ...... 36

2.1 Overview ...... 36 2.2 Biosensors Overview ...... 36 2.3 Magnetic Materials Overview ...... 41 2.4 Spin-Vortex Magnetic Behavior ...... 45 2.5 Magnetic Forces and Torques Overview ...... 49 2.6 Summary ...... 52

3 BIO-FUNCTIONALIZED MAGNETIC MICRODISCS ASSAY TO DETECT -COATED PARTICLE OR BACTERIAL TARGETS ...... 53

3.1 Overview ...... 53 3.2 Microfabrication of Gold-Coated Magnetic Microdiscs ...... 55 3.3 Surface Bio-Functionalization of Magnetic Microdiscs ...... 58 3.3.1 Surface Bio-Functionalization for Avidin-Coated Particles ...... 58 3.3.2 Surface Bio-Functionalization for Bacterial Targets ...... 59 3.4 Sample Preparation ...... 60 3.4.1 Protein-Coated Particle Target Samples ...... 60

6

3.4.2 Bacterial Target Samples ...... 61 3.5 Bio-Functionalized Microdiscs Targeting Particles and/or Bacteria ...... 62 3.6 Viability Assay Preparation for Bacterial Targets ...... 62 3.6.1 SYTO9 / Propidium Iodide (PI) Viability Labelling ...... 63 3.6.2 Other Fluorescent Labels ...... 63 3.7 Magnetic Isolation and Detection of Target Bacteria/Particles Overview ...... 64 3.8 Fluorescence Inspection ...... 64 3.9 Summary ...... 65

4 METHODS FOR THE MAGNETIC SEPARATION OF MICRODISCS...... 66

4.1 Overview ...... 66 4.2 Magnetic Microdisc Capture Efficiency Quantification ...... 67 4.3 Multiphysics Simulations ...... 71 4.3.1 Magnetic Simulations ...... 72 4.3.2 Microfluidic and Particle Tracing Simulations ...... 76 4.4 Magnetic Separation Configurations ...... 78 4.4.1 Single Magnet Configuration Setup...... 78 4.4.2 Dual Magnet (Attracting and Repelling) Configuration Setups ...... 79 4.4.3 Column Array of Magnets with Alternating Polarizations ...... 80 4.4.4 Microfluidic Magnetic Separation (µFMS) Device ...... 82 4.4.4.1 µFMS device fabrication ...... 83 4.4.4.2 µFMS device experiments ...... 85 4.4.4.3 Estimation of magnetic capture efficiency of µFMS device ...... 89 4.4.4.4 Magnetic capture efficiency vs. flow rate for µFMS device ...... 91 4.5 Filtering MNPs from Large-Volume Samples Using µFMS Device ...... 94 4.6 Summary ...... 96

5 MAGNETIC ISOLATION OF PARTICLE/BACTERIAL TARGETS ...... 98

5.1 Overview ...... 98 5.2 Evaluation of Multiple Fluorescence Labels for Target Particles/Bacteria ..... 99 5.3 Exploration of Microdisc Bio-Functionalization Target Specificity ...... 101 5.4 Demonstration of Magnetic Isolation of Target Particles/Bacterial Cells ..... 108 5.4.1 Magnetic Isolation of Avidin-Coated Particles Using the µFMS Device ...... 109 5.4.2 Magnetic Isolation of E. coli Using the µFMS Device ...... 111 5.4.3 Magnetic Isolation of Coliforms / E. coli Using Macrofluidic Setups .. 113 5.5 Evaluation of LODs for Water Quality Monitoring Applications ...... 115 5.6 Exploration of Microdisc Rotation to Enhance Target Binding Rate ...... 120 5.6.1 Magnetic Rotation (or Stirring) Experimental Setup ...... 121 5.6.2 Magnetic Stirring Preliminary Experiments ...... 124 5.6.2.1 Rotation of microdiscs in samples with E. coli cells ...... 124 5.6.2.2 Rotation of microdiscs in samples with avidin-coated particles.. 132 5.7 Summary ...... 134

6 CONCLUSIONS ...... 137

7

6.1 Research Contributions ...... 137 6.2 Future Perspectives ...... 139 6.2.1 Other Bacterial Targets ...... 139 6.2.2 Microdisc Rotation...... 140 6.2.3 Portable Microscopy ...... 140 6.2.4 Other applications ...... 143

APPENDIX

A GRAM-POSITIVE AND GRAM-NEGATIVE BACTERIA CELL WALL STRUCTURES ...... 145

B SURFACE BIO-FUNCTIONALIZATION OF MAGNETIC MICRODISCS: SELECTIVITY ...... 146

C CONFIRMATION TESTS OF BACTERIA GROWTH FROM MICROBIOLOGICS E. COLI/COLIFORMS KITS ...... 151

D COMSOL CAPTURE EFFICIENCY OF DISCS ...... 155

E VSM DATA FOR MICRODISC QUANTIFICATION ...... 156

F CELL CULTURE COLONY COUNTING CONFIRMATION RESULTS ...... 157

G IMAGEJ EXAMPLE IMAGES FROM IMAGE POST-PROCESSING ...... 158

H SMARTPHONE FLUORESCENCE MICROSCOPE PLATFORM ...... 159

LIST OF REFERENCES ...... 160

BIOGRAPHICAL SKETCH ...... 170

8

LIST OF TABLES

Table page

1-1 Standard methods comparison ...... 22

1-2 Biosensing characteristics literature review ...... 27

1-3 Summary of microfluidic magnetic separation recent works...... 30

2-1 Key biosensing terminologies: definitions and examples...... 38

3-1 Summary of water samples prepared for experiments ...... 55

4-1 Summary of Capture Efficiency (%) Calculations for Microdiscs ...... 89

4-2 Summary of Capture Efficiency (%) Calculations for IONs...... 90

5-1 Summary of samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using E. coli cells as target bacteria and aptamers as capture probes for bio-functionalization)...... 125

5-2 Summary of samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using E. coli cells as target bacteria and lectins as capture probes for bio-functionalization)...... 128

5-3 Summary of samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using avidin-coated polystyrene particles as target particles and biotin as capture probes for bio- functionalization)...... 132

6-1 Key contributions ...... 138

B-1 Lectin selectivity for coliforms...... 146

9

LIST OF FIGURES

Figure page

1-1 Summary of typically used methods for bacteria detection (limits of detection and test times)...... 21

1-2 Data from Web of Science with the number of publications per year on magnetic particles “and” cells...... 23

1-3 Spin-vortex magnetic behavior of the magnetic microdiscs...... 25

1-4 Overall magnetic isolation of target bacteria...... 33

1-5 Overall concept of this research...... 34

2-1 Summarized application areas of biosensors...... 37

2-2 Diagram explaining the overall mechanism of biosensors...... 38

2-3 Summarized bio-application areas of magnetic materials...... 42

2-4 Magnetic moment arrangement for different magnetic materials...... 43

2-5 Hysteresis loop of soft and hard ferromagnetic materials...... 44

2-6 Magnetic moments creating a circular vortex...... 45

2-7 Magnetic behavior of discs with different dimensions (radius and thickness). .... 46

2-8 OOMMF simulation results for Permalloy microdiscs with dimensions of 1.5 µm in diameter and 70 nm in thickness...... 48

3-1 Microfabrication process schematic...... 56

3-2 SEM images of patterned on substrate and lifted-off microdiscs...... 57

3-3 Optical profilometer data confirming a thickness of ~ 70 nm for our array of magnetic microdiscs...... 57

3-4 Magnetic microdiscs microfabrication and bio-functionalization...... 59

3-5 Overall experimental procedure for disc functionalization, staining, and sample preparation for imaging...... 60

4-1 Capture efficiency terminology example diagram...... 68

4-2 Example of VSM data before and after filtration (i.e. stock and filtrate solutions, respectively)...... 71

10

4-3 COMSOL simulation results for single magnet configuration...... 74

4-4 COMSOL simulation results for pair of (attracting) magnets configuration...... 75

4-5 COMSOL simulation results for pair of (repelling) magnets configuration...... 75

4-6 COMSOL simulation results for configuration using a column array of magnets with alternating polarization...... 75

4-7 COMSOL simulation results for 3x3 array of magnets with alternating polarization configuration and microfluidic device setup...... 76

4-8 COMSOL geometry used for simulations of the µFMS device, isolating ‘microdisc’ particles with a Dhd of 618 nm...... 77

4-9 Single permanent magnet setup diagram...... 78

4-10 Isolated discs from a small-volume sample (~5 mL) using a simple permanent magnet. Photo courtesy of author...... 78

4-11 Pair of magnet configurations (attracting and repelling)...... 80

4-12 Magnetic separation setup using a column array of magnets with alternating polarizations...... 81

4-13 Setup showing discs “stuck” in the tubing after removing magnets and multiple washes. Photo courtesy of author...... 82

4-14 Diagram of the PDMS device attached to a glass slide (ready-to-inspect) setup with detachable magnet array...... 84

4-15 Magnetic separation microfluidic device experimental setup diagram and image...... 87

4-16 µFMS device with alternating polarization array of magnets (top) and magnetically separated discs on the device (bottom)...... 88

4-17 Example image of discs magnetically separated using µFMS device...... 88

4-18 Vibrating sample magnetometer (VSM) data obtained (before and after filtration) from 0.2-mL samples filtered using the microfluidic magnetic separation (µFMS) device at different flow rates : 5 µL/s, 15 µL/s, 30 µL/s, 60 µL/s, and 120 µL/s...... 90

4-19 VSM data obtained (before and after filtration) from 0.2-mL IONs samples filtered using the microfluidic magnetic separation (µFMS) device at different flow rates: 5 µL/s and 120 µL/s...... 91

11

4-20 Capture efficiency study for (A) microdiscs and (B) iron-oxide nanoparticles (experiments and COMSOL simulations)...... 93

4-21 Example image of magnetic microdiscs captured from a 100-mL sample at 120 µL/s using the µFMS device...... 95

4-22 Experiment results from 100-mL samples using the µFMS device...... 95

4-23 Summarized results on the magnetic re-capture of magnetic particles (i.e. microdiscs or nanoparticles) using the different magnet configurations...... 97

5-1 Fluorescence tagging of E. coli was demonstrated using three kinds of labels...... 100

5-2 Magnetic isolation of avidin-coated particles was demonstrated using biotin- functionalized magnetic microdiscs...... 102

5-3 Magnetic isolation of E. coli cells was demonstrated using aptamer- functionalized magnetic microdiscs...... 102

5-4 Proof-of-concept results demonstrating selectivity of bio-functionalized microdiscs in lab-prepared water samples...... 104

5-5 Proof-of-concept results demonstrating selectivity of bio-functionalized microdiscs in a complex environmental sample (i.e. vegetable broth), (valuable for field testing applications)...... 105

5-6 Bio-flocs containing coliforms isolated using Con A-functionalized magnetic microdiscs...... 106

5-7 Control samples...... 108

5-8 Isolation of fluorescent avidin-coated particles from 100-mL samples using biotin-functionalized magnetic microdiscs...... 110

5-9 Example image of aptamer-functionalized microdiscs and bacteria conjugates isolated using µFMS device and a flow rate of 120 µL/s...... 112

5-10 More example images of aptamer-functionalized microdiscs and bacteria conjugates isolated using µFMS device and a flow rate of 120 µL/s...... 112

5-11 Example images of lectin-functionalized microdiscs and coliform conjugates isolated dual-magnet setup with repelling configuration of magnets at a flow rate of ~83 µL/s...... 114

5-12 Example images of aptamer-functionalized microdiscs and E. coli conjugates isolated dual-magnet setup with repelling configuration of magnets at a flow rate of ~83 µL/s...... 115

12

5-13 Confocal microscopy images showing isolation of various target coliform cells using lectin-functionalized magnetic microdiscs...... 117

5-14 Confocal microscopy images showing isolation of various target E. coli cells using aptamer-functionalized magnetic microdiscs...... 118

5-15 Example (bright and fluorescent fields) images of samples containing aptamer-functionalized discs exposed to different concentrations of E. coli. .... 119

5-16 Schematic (top) and image (bottom) of proposed pathogen detector apparatus incorporating a magnetic actuation and an optical measurement by transmission...... 122

5-17 Magnetic microdiscs acting as an "optical shutter” when actuated by a rotating magnetic field...... 123

5-18 Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable)...... 126

5-19 Quantified results using ImageJ (green pixels/dark pixels). Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). 127

5-20 Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable)...... 129

5-21 Quantified results using ImageJ (green pixels/dark pixels). Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). 130

5-22 Quantified results using culture colony counting (plating) of the filtrate (not captured cells)...... 130

5-23 Plates for culture colony counting for each of the samples. Colonies counted correspond to cells not captured during the experiments (decanted supernatant)...... 131

5-24 Experimental and control results on the magnetic isolation of avidin-coated particles after rotation (if applicable)...... 133

5-25 Experimental and control results from ImageJ on the magnetic isolation of avidin-coated particles after rotation (if applicable)...... 134

6-1 Potential applications of this method include water quality monitoring in recreational, fisheries, and irrigation waters...... 139

6-2 Diagram of envisioned imaging platform setup...... 142

6-3 Current and envisioned future setup for the magnetic isolation and fluorescence inspection of bacterial targets...... 143

13

A-1 Structure and composition of Gram-positive and Gram-negative cell walls...... 145

B-1 Selectivity of aptamer for E. coli over other Gram-negative and Gram-positive cells...... 147

B-2 Electrochemical impedance spectroscopy analysis of thiolated aptamers on a gold electrode toward E. coli 25922...... 149

B-3 Selectivity of aptamer toward E. coli 25922 over other Gram negative bacteria...... 149

B-4 Cartoon representation of secondary structure predicted using mfold...... 150

B-5 EIS analysis of scrambled aptamer shows no selectivity toward E. coli 25922. 150

C-1 Confirmation test: -Methylumbelliferyl β-D-galactopyranoside (MUG) assay. ... 151

C-2 Confirmation test: Fluorescent emission of MUG assay at 445 nm measured using fiber optic sensor...... 152

C-3 Confirmation test: Culture colony counting...... 152

C-4 E. coli and Coliforms kits used for lab-prepared samples...... 153

D-1 Plot of simulated capture efficiencies (COMSOL) as a function of change in hydrodynamic diameter of IONs, considering possible aggregation of particles...... 155

E-1 VSM data obtained (before and after filtration) from 50-mL sample containing iron-oxide nanoparticles (IONs) at a concentration of 0.1 mg/mL filtered using the µFMS device at 120 µL/s...... 156

E-2 Image of 50-mL sample before and after filtration using the µFMS device at 120 µL/s. Photo courtesy of author...... 156

F-1 Plating results for E. coli sample enriched for 23 hours in TSB. Photos courtesy of Adam B. Grossman...... 157

G-1 Example images after post-processing in ImageJ “color thresholding.” ...... 158

H-1 Prototype for smartphone fluorescence microscopy platform along with two different filters for green-fluorescing samples. Photos courtesy of author...... 159

H-2 Filter placement and setup powered. Excitation filter: ~480 nm, Emission filter (LED): ~500 nm. Photos courtesy of author...... 159

H-3 Example images of different concentrations of Biotin-4-fluorescing obtained using the prototype with two different filters. Photos courtesy of author...... 159

14

LIST OF ABBREVIATIONS

Con A Concanavalin A

DI deionized

DNA deoxyribonucleic acid

E. coli Escherichia coli

PBS Phosphate buffered saline

TBS Tryptic soy broth

VSM vibrating sample magnetometer

15

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

MAGNETIC ISOLATION OF FECAL INDICATING BACTERIA USING BIO- FUNCTIONALIZED MAGNETIC MICRODISCS FOR WATER QUALITY MONITORING

By

Keisha Yarie Castillo-Torres

May 2020

Chair: David P. Arnold Major: Electrical and Computer Engineering

This work provides a new method for water quality monitoring via the isolation of fecal indicating bacteria in relatively large water samples, up to 100 milliliters (mL). This method demonstrated suitability for rapid bacteria detection in drinking water samples (<

8 hr) and recreational water samples (< 2 hr). Here, bacteria isolation is done using bacteria-sized (1.5 µm in diameter, 80 nm in thickness) gold-coated soft ferromagnetic

(Ni80Fe20) microdiscs.

The overall process for bacteria detection is described as follows, magnetic microdiscs are bio-functionalized with capture probes that selectively (or preferentially) bind to bacterial targets. Examples of capture probes include, Concanavalin (Con A) lectins and deoxyribonucleic (DNA) aptamers when targeting total coliforms and

Escherichia coli cells, respectively. Then, the already bio-functionalized microdiscs target bacterial cells suspended while incubated in water samples for 20-30 mins. After incubation, samples are filtered using a fluidic device containing “magnetic traps” to magnetically isolate microdisc/bacteria conjugates into a smaller and easier-to-image substrate (i.e. glass slide). At this point, fluorescence labeling is performed using commercially available live/dead viability stains, such as, SYTO9 and propidium iodide

16

(PI). Finally, bacteria detection is assessed via fluorescent imaging using either epi- fluorescent or confocal microscopy.

Proof of concept results have demonstrated capture of viable and non-viable cells using bio-functionalized magnetic microdiscs. Also, results have demonstrated selective isolation of general coliforms and E. coli cells using lectin-functionalized and aptamer-functionalized microdiscs, respectively. Additionally, it was observed that fluorescence signals are directly correlated to bacteria concentrations and low signal is apparent in control samples (i.e. no bacteria).

Bacteria isolation results were obtained at detection levels as low as 2-10 CFU /

100 mL for total coliforms and 1-10 CFU / 100 mL for E. coli with 6-hour enrichment steps. However, lower detection levels for non-enriched samples resulted in 100

CFU/100 mL. These detection limits represent maximum contamination levels of coliform/E. coli bacteria in safe drinking water (1 CFU/100 mL) and recreational water

(126 CFU/100 mL) systems according to the U.S. Environmental Protection Agency

(EPA).

Additionally, results on the magnetic isolation of other targets (i.e. avidin-coated particles) was demonstrated using biotin-functionalized microdiscs. Hence, future works involving the evaluation of this method’s capability to target other bacterial targets, such as, Enterococci, with appropriately bio-functionalized magnetic microdiscs are encouraged. Also, further studies on the effect of rotating the magnetic microdiscs when exposed to bacterial targets to increase binding events, as well as, the development of a smartphone-based fluorescent imaging platform for detection are encouraged for future works.

17

CHAPTER 1 INTRODUCTION*

1.1 Overview

This chapter will introduce the importance of bacteria isolation in water samples

(drinking, recreational, and irrigation waters). Also, an overview about previous and current works, including standard and bio-magnetic methodologies used to detect bacteria, will be discussed and compared. Finally, an overview of this research objectives and methodologies will be introduced and discussed.

1.2 Motivation

Detection of bacteria indicative of fecal contamination is central to water quality monitoring for ensuring safe water for human contact and/or drinking. Globally, more than 900,000 deaths are reported every year related to foodborne and waterborne diseases [1], [2]. Also, as reported in the National Outbreak Reporting System (NORS) from the Center for Disease Control and Prevention (CDC), between 2007 and 2017 there were a total of 401 outbreaks associated with exposure to fecal contaminated

(Escherichia, Enterococcus, and Streptococcus) water (drinking, recreational, environmental, and undetermined), which resulted in 7,095 cases of illness, 1,344 hospitalizations, and 23 death in the United States [3], [4].

Historically, health effects (such as gastrointestinal illnesses) have been associated with fecal indicating bacteria exposure [5], [6]. Specifically, epidemiological studies by the U.S. Environmental Protection Agency (EPA) have shown strong

* Portions of this chapter are reprinted with permission from (i) K. Y. Castillo-Torres, D. P. Arnold, and E. S. McLamore, “Rapid isolation of Escherichia coli from water samples using magnetic microdiscs,” Sensors and Actuators: B chemical, 291 (2019), 58-66. Copyright 2019 Elsevier B.V. and (ii) K. Y. Castillo-Torres, E. S. McLamore, and D. P. Arnold, “A High-Throughput Microfluidic Magnetic Separation (µFMS) Platform for Water Quality Monitoring,” Micromachines, vol. 11, no. 1, p. 16, Dec. 2020.

18

correlations between illnesses and bacteria concentrations of Enterococci and E. coli in fresh and marine waters [5]. Hence, ensuring safe water still represents a global concern and an ongoing societal challenge and monitoring recreational, irrigation, and drinking water sources plays a vital role in reducing these numbers.

The U.S. EPA uses total coliforms in water samples as an indicator of the presence of other pathogenic microorganisms that might be harmful to humans [7], [8].

Total coliform bacteria represent a group of microorganisms present in the intestines of humans and animals. A subgroup of total coliform bacteria is fecal coliform bacteria, which includes non-spore forming and Gram-negative bacteria, such as Escherichia coli

(E. coli).

For safe human contact, it is recommended that recreational waters have a geometric mean (GM) below 35 CFU/100 mL for Enterococci and below 126 CFU/100 mL for E. coli, as specified by the Recreational Water Quality Criteria from the U.S. EPA

[5]. This limit (126 CFU/100 mL) is identical to the hazard threshold set by the Food and

Drug Administration for the irrigation of fresh produce according to the Food Safety

Modernization Act (FSMA) [9]. In drinking water, the maximum level is set at 1 CFU/100 mL for E. coli [7], which is near the detection limit of many detection systems.

While there has been much work in bacteria detection for water and food samples [10], [11], there are operational aspects that limit detection schemes for rapid, point of use applications [12]. Prominent limitations include long analysis times (typically

24 hr including enrichment of bacteria), a need for complex lab tools/equipment, and/or the requirement for a highly skilled operator. The long-term aim of this work is to rapidly

19

(< 2 hr) detect bacterial indicators of fecal contamination in 100 mL water samples pertinent to drinking water, recreational water, and food safety monitoring applications.

1.3 Previous Works on Bacteria Detection

1.3.1 Standard Methods

Existing methods for bacteriological water analysis (assessment of total coliform bacteria and E. coli) suffer from one or more critical limitations prohibiting their applicability for portable, rapid, and low-cost field-testing.

Culture methods (such as, multiple tube method, plate count, membrane filtration, and pour plate) are extremely accurate and can discriminate viable from non- viable cells, but the test requires long incubation times and typically highly skilled personnel. Immunoassays, such as enzyme-linked immunosorbent assay (ELISA) is a highly specific quantitative method that permits species or serotype level confirmation, but it is expensive, low throughput, demands highly skilled personnel, and requires long- time frame to obtain results (usually at least 24 hours). Labelling techniques using flow cytometry, laser scanning, luminometry, or epifluorescence are rapid (typically < 1 hour), but fluorescent labels are expensive, susceptible to photobleaching, and some are cytotoxic. In addition, these acquisition systems are cost-prohibitive for large scale or high throughput field monitoring campaigns.

Molecular methods for monitoring nucleic acids, including polymerase chain reaction (PCR), reverse transcriptase PCR (RT-PCR) and nucleic acid sequence-based amplification (NASBA) have been used extensively, but the techniques are time consuming (e.g. 8-24 hr for sample filtration, DNA amplification, and PCR detection)

[13]. Although a few recent techniques have used PCR to distinguish viable from non- viable pathogens (known as molecular viability analyses), the variable persistence of

20

nucleic acids in cells post-death leads to low accuracy in “real” environmental samples, which adds to the analysis time and induces error [14]. In addition, molecular methods require that viability markers be measured over long periods of time, extending the detection time considerably and deterring development of “one pot” rapid testing schemes. Table 1-1 and Figure 1-1 summarize some biosensing characteristics of different standard methods used to target and detect coliforms/E. coli, as summarized mostly in [15]. Ideal characteristics of these methods for water quality monitoring applications include low detection limits (1-126 CFU/100 mL) and fast response (<24 hr). In terms of viability assessment, in some cases viability discrimination can provide valuable information since poor biosensor accuracy may result in disposal/closure of otherwise safe drinking/recreational water or illness, adding significant logistical burden and increased disability adjusted life years (DALYs).

Figure 1-1. Summary of typically used methods for bacteria detection (limits of detection and test times).

21

Table 1-1. Standard methods comparison

Method Examples LOD* Response Viability Ref. (CFU/100 Time (hr) assess- mL) ment? Culture multiple tube method, 100-102 ≥24 YES [8], [15] methods plate count, membrane filtration, and pour plate Immunoassay enzyme-linked 103-105 8-24 NO [15] methods immunosorbent assay (ELISA) Molecular polymerase chain 100-105 2-24 YES [15][16] methods reaction (PCR), reverse transcriptase PCR (RT- PCR) and nucleic acid sequence-based amplification (NASBA) Labelling flow cytometry, laser >102 <2.5 YES [17] techniques scanning, luminometry, or epifluorescence *LOD: limit of detection

1.3.2 Bio-Nano-Magnetic Methods

Interest in merging biology, nanoscience, and magnetics has been growing over the last two decades [18]–[20]. Most notably, magnetic nanoparticles (MNPs) have been used for magnetic imaging contrast enhancement agents, magnetic hyperthermia, drug delivery, and magnetic separation of cells [18]–[21]. MNPs are attractive for biological applications for their biocompatibility and their ability to be non-invasively actuated and/or interrogated using magnetic fields.

Figure 1-2 summarizes data from Web of Science with the number of publications per year on magnetic particles and cells [22]. Data was compiled for cells

“and” different key words including magnetic particles, magnetic nanoparticles, magnetic beads, magnetic microparticles, and magnetic microdiscs (which will be used in this work).

22

Figure 1-2. Data from Web of Science with the number of publications per year on magnetic particles “and” cells. From the compiled data it can be observed that very few publications related to magnetic microdiscs and cells are available, hence the labelled numbers in red. Data accessed from Web of Science [22].

A key feature of MNPs is their superparamagnetic behavior (ideally zero magnetic remanence), which inhibits inter-particle magnetic dipole interactions and thus unwanted magnetic agglomeration of particles when in suspension [18]. However, for a magnetic particle to exhibit superparamagnetic behavior, the particle diameter is limited to a few tens of nanometers. For example, the ubiquitously used superparamagnetic iron oxide nanoparticles (SPIONs) are superparamagnetic up to only ~20 nm, above which the particles are considered magnetically blocked and hence retain a small dipole moment even in the absence of a magnetic field. Upon application of a field, the

SPIONs are typically driven to magnetic saturation with a saturation magnetization of typically 0Ms = 0.58 T (90 emu/g) [23]–[26]. Because the net magnetic moment of a

23

single ~20-nanometer particle is too small for many applications, the MNPs are often mixed with polymers (e.g. polystyrene) and used to form larger micron-sized magnetic beads. However, the magnetic volume fraction can be up to 70% [27], which means the effective saturation magnetization of a single bead can be only ~0.41 T (~63 emu/g).

These MNPs have been used to magnetically isolate cells, including circulating tumor cells (CTC), CD34+ cells, and E. coli cells [18]. However, detection limits of magnetic isolation methods for water/food samples using MNPs (103‒109 CFU/100 mL) are still far away from the 126 CFU/100 mL and 1 CFU/100 mL criteria from the FSMA and EPA standards for safe recreational/irrigation and drinking waters, respectively

[28]–[31].

An alternative to MNPs for bio-nano-magnetic applications is the use of metal or metal alloy (typically Ni80Fe20) “spin-vortex” discs. A spin-vortex disc is a disc-shaped nanostructure, that under the right dimensions, exhibits a self-enclosing vortex arrangement of the atomic moments in the absence of a magnetic field [32], [33]. Like a

SPION, these discs are characterized by near-zero remanence (Figure 1-3), but much higher saturation magnetization, 0Ms = 1.00 ± 0.01 T (91.16 ± 1.34 emu/g). The magnetic microdiscs used in this work (1.5 µm in diameter, 80 nm in thickness) are lithographically patterned, which provides reliable control of size, shape, structure, and function. Furthermore, the magnetic microdiscs are already micron-sized and fully magnetic. The saturation magnetic moment of one disc is up to 6 orders of magnitude larger compared to typical SPIONs (~200,000 times larger than typical magnetic beads), and therefore these particles can impart much higher magnetic forces and/or make use

24

of smaller magnetic fields and field gradients. Another distinguishing advantage is that the magnetic discs can impose torques, whereas SPIONs and magnetic beads cannot.

Figure 1-3. Spin-vortex magnetic behavior of the magnetic microdiscs. As can be observed from this plot, soft-ferromagnetic microdiscs exhibit near-zero magnetic moment in the absence of an external magnetic field, however, when an external magnetic field is applied, the magnetic microdiscs respond with a high magnetic moment. This behavior causes large forces exerted on the magnetic microdiscs when an external magnetic field is applied, (i.e. a simple permanent magnet).

All these advantages make magnetic microdiscs of interest for bio-applications, especially for (micron-sized) cell targeting and magnetic isolation. For example, use of these discs have been explored and demonstrated to target and trigger cancer cell apoptosis and magnetic hyperthermia by introducing an alternating magnetic field that creates a rotating magneto-mechanical stimulus on the discs and therefore on the cells

[21], [34]. Similarly, previous works on these discs have demonstrated the ability to characterize fluidic sample properties (i.e. viscosity) by also actuating the discs with a

25

rotating magnetic field [35]. Preliminary results for selective targeting and isolation of bacteria using these discs was reported in 2016 [36], [37].

Table 1-2 summarizes the biosensing characteristics of different methods that have implemented the use of magnetic particles to target and detect E. coli. It can be observed that most of the magnetic particles used consisted of iron oxide particles with diameters raging between 10-150 nm.

As was mentioned before, these MNPs (tens of nanometers) are in many cases coated and/or combined with polymers and/or polysterene NPs (i.e. polyaniline, polystyrene beads, and poly-ethylene glycol (PEG)), which when compared to our magnetic microdiscs (1.5 µm in diameter, 70 nm in thickness), still possess up to

236,000 times smaller volumes. Also, in Table 1-2, details on the assay selectivity (or preferentially) to E. coli was described in terms of the capture probes used. Here, a variety of , amines, aptamers, and others were presented. Other details, such as, sample volumes, limits of detection, response times, and the integration of a viability step were considered and summarized in the table. Here, only one other method [38] considered sample volumes of up to 100 mL, the rest ranged between 0.1-10 mL. Also, only one other method [39] added a viability step. Limits of detection ranged from

3.0×100 to 1.5×109 CFU per 100 mL, while response times ranged between 0.35-2.5 hr.

When compared with the method proposed here, the limit of detection is inside that same range (102 CFU/100 mL). However, preliminary results have been obtained for limit of detection of down to 1 CFU/100 mL in enriched samples. Similarly, the response time of the proposed method can be less than an hour and up to 8 hours (if enrichment is needed for 1 CFU/100 mL concentrations).

26

Table 1-2. Biosensing characteristics literature review Method Magn. Selectivity Sample LOD Resp. Viability Ref. part. (capture volume (CFU time step? (material probe) (mL) /100 (hr) and size) mL) Impedance SA-coated biotin- 0.5 7.4 0.58 No [28] 6 biosensor Fe3O4 labeled ×10 coupled with NPs poly-clonal MNP-antibodies (145 nm) goat antibodies (E. coli) Immuno- PA-coated Mono-clonal 0.1 6.0×102 1.08 No [40] magnetic gamma anti-E. coli separation (IMS) Fe3O4 O157:H7 and NPs antibodies electrochemical (20‒100 (EC) detection nm) 0 IMS and EC MAX E. coli 1-100 3.0×10 2.00 No [38] detection Dyna- O157:H7 beads and poly- styrene beads (2.8 µm) Bacteria Fe O Cetyl- 5.0 9 1.00 No [41] 3 4 1.5×10 magnetic capture NPs with trimethyl- and disinfection CTAB ammonium (10 nm) bromide (CTAB) 4 Bacteria removal Fe3O4 / Solvo- 1-2.05 1.0×10 >0.50 Yes [39] using magnetic graphene thermal graphene synthesis and Fe3O4 composite and NPs MNPs (10‒25 nm) Magneto-phoretic Platinum- half- 10 10×102 0.5 No [29] chromato-graphy coated fragments of (milk) and colorimetric Fe3O4 E. coli detection with clusters antibodies MNPs (150 nm) Electro-chemical Poly- monoclonal 0.5 10×102 < 1 No [30] immune-sensor aniline- antibodies using MNPs coated γ - Fe2O3 NPs (50‒100 nm)

27

Table 1-2. Continued Method Magn. Selectivity Sample LOD Resp. Viability Ref. part. (capture volume (CFU time step? (material probe) (mL) /100 (hr) and size) mL) Nuclear gamma E. coli 0.1 7.6×102 0.35- No [42] magnetic iron oxide O157:H7 0.52 resonance NPs antibody (50‒100 nm) 9 MTT Cobalt amine- 5.0 1.5×10 1.00 No [43] (colorimetric) ferrite NPs functiona- assay using 7 nm lized MNPs MNPs Magnetic Magnetic 2- 10 1.0×103 0.50 Yes [44] reduced particles nitrodopa- graphene oxide (MPs) mine (rGO) anchored modified nanoheaters on rGO MPs for selective nanocom- isolation of E. posites coli Bacterial PEG aptamer- 0.1 1.0×104 2.5 No [45] capture and hydrogel functiona- detection using inverse lized hydrogel opal barcodes magnetic particles barcodes (150 nm) 4 Magnetic Ni80 Fe20 DNA 1-100 1.0×10 - < Yes This isolation using micro-discs aptamers 1.0×105 0.50 work spin-vortex (Ø: 1.5 µm, magnetic thickness: microdiscs + 70 nm) fluorescent assay

1.3.3 Microfluidic Magnetic Separation Technologies

Microfluidic devices possess advantages in terms of their size, low-cost fabrication, and the possibility of parallel device operation [46]. Different microparticle/cell separation (or sorting) microfluidic technologies have been developed for large number of particles/cells using acoustic, dielectric, thermal, or magnetic properties, among others as reviewed by Y. Shen, et. al, and T. Zhang, et. al in [47] and

[48], respectively. Size-based microfluidic devices using deterministic lateral

28

displacement (DLD) arrays for high-throughputs have been developed, using flow rates of up to 167 µL/s [49], [50]. However, DLD structures usually consist of complex microfabrication process. Also, these DLD example devices possess an enrichment step and use multiple pumps [49] or have been tested to process up to 5-mL samples with a 91% targeted cell capture efficiency [50]. Microfluidic devices that make use of magnetic field gradients to enhance selectivity and increase throughput in cell separation, sorting, and trapping applications have been developed [46], [51]–[56].

Microfluidic magnetic separation technologies have aimed to reduce the total analysis time by avoiding long enrichment steps by isolating/concentrating magnetically- tagged bacteria using various magnetic field apparatuses [57]. Most magnetic separation biosensing systems for bacteria detection have been tested with sample volumes not larger than 10 mL, with limits of detection ranging 3.0×100 – 1.5×109

CFU/100 mL and analysis times ranging 0.35 – 2.5 hr [57]. Previous works generally use magnetic beads having diameters 50 – 250 nm, where the beads comprise superparamagnetic iron oxide nanoparticles embedded in a polymer matrix (e.g. polystyrene). The net magnetic volume fraction of the bead is typically less than 15

%vol. [27], [38], [41], [42], [45], [53], [58]. In contrast, the magnetic microdiscs used in this work are highly magnetic (88 %vol), bacteria-sized discs (1.5 µm in diameter, 80 nm in thickness) and include a 5-nm layer of gold on each side, making them well-suited for magnetic separation of bacteria. In a previous work [57], aptamer-functionalized microdiscs were used to isolate E. coli at levels as low as 100 CFU/100 mL in less than

45 min. However, the isolation (magnetic trapping) step was performed using a bulky apparatus that required multiple successive passes through the device to achieve high

29

capture efficiencies. Here, we present the use of a microfluidic device for faster sample filtering, convenient sample preparation, and ultimately better performance.

Table 1-3. Summary of microfluidic magnetic separation recent works. Sample Flow Rate Ref. Lab-prepared Sample Type Volume (µL/s) (target particle/cell) (µL) [59] Water (magnetic particles) 5 0.25 [53] Blood (E. coli) 10 0.011 [52] Water (E. coli and Acinetobacter sp.) 25 0.017 [51] Water (E. coli) nr* 0.833 [54] Water (magnetic particles) nr* 0.017 [60] MACS (magnetic particles) 200 0.278 [61] Buffer (Jurkat cells) 1,000 0.333 [62] Blood (E. coli) 2,000 16.67 [55] Blood (circulating tumor cells) 10,000 2.780 [56] Blood (Candida albicans fungi) 10,000 5.556 up to This Water (avidin-coated particles and E. coli) 120 100,000 work *nr: not reported

Main challenge of magnetic separation microfluidic devices is their typically small volume capacity. As summarized in Table 1-3, most of these magnetic separation microfluidic devices used sample volumes, ranging from a few µL to no more than 10 mL [52], [53], [55], [56], [59]–[62]. For these volumes, relatively low flow-rates, typically less than 20 µL/s, were sufficient to achieve results in short time [51]–[54], [59]–[62]. For example, Zanini, et al. developed a microfluidic device with an integrated array of micromagnets with alternating polarities for the separation of magnetic nanoparticles, which resulted in >94% particle capture efficiencies (with 0.25 µL/s flow rate) [28].

30

However, for water quality monitoring, there is need for processing much larger 100-mL samples in a short time period, which serves as motivation for this work.

In this work, we demonstrate the use of a much-improved microfluidic magnetic separation (µFMS) device that offers flow rates of up to 120 µL/s with capture efficiencies of ~94%. This device is capable of analyzing 100-mL water samples in less than 15 min, a significant advancement towards rapid bacteria detection. Instead of integrating the magnets into the microfluidic platform as in [52], [59], [61], our device utilizes an external magnet array placed below the microfluidic platform, drastically simplifying the device fabrication and lowering the per-unit test cost. This proof-of- concept demonstrates isolation of microdisc/particle or microdisc/bacteria conjugates using a µFMS device, which uses a single filtration step protocol, providing an imaging- ready substrate for subsequent fluorescent microscopy.

Figure 5-1 describes the overall concept of the biosensing system for bacteria detection. First, magnetic microdiscs bio-functionalized with specific capture probes

(e.g. aptamers, antibodies, ) are used to separate bacterial/particle targets (e.g.

E. coli or other target particles/cells). After co-incubation of these bio-functionalized microdiscs with a 100 mL water sample containing the target particle or cell, the µFMS device is used to isolate the microdisc/target conjugates in a localized area for imaging.

Target cells can optionally be stained/labeled with a variety of fluorescent tags, and analysis is carried out using standard fluorescence microscopy.

1.4 Research Approach

In this approach, bacteria-sized magnetic discs (1.5 µm in diameter and 80 nm in thickness) are used. Compared to nanoparticle-based concentration approaches, these unique magnetic microdiscs have volumes up to 200,000 times larger, and thereby

31

enable the application of much higher magnetic forces without damaging cells, resulting in incredibly enhanced concentration of target bacteria. By localizing target bacteria from water samples, of up to 100-mL, into a compact field of view, microscopic imaging can be performed more quickly and with enhanced sensitivity.

1.4.1 Objectives

This work aims to extend the study on bacteria targeting using surface- functionalized gold-coated magnetic microdiscs to detect E. coli in recreational, irrigation, and/or drinking water and other complex matrices samples. It is envisioned that coliforms/E. coli isolation is performed using functionalized magnetic microdiscs at detection levels of at least 100 CFU/100 mL in a short period of time (< 2 hr) using different 1) capture probes to target different target cells and 2) fluorescent tags for bacteria detection during fluorescence imaging.

A long-term goal of this method is to develop a water monitoring system that serves as a rapid indicator of presence of fecal material in water samples. Once the detection of coliforms/E. coli is performed, the water sample can be subject to analysis under a more specific biosensor. The significance of this technique will rely on its capability to provide a platform for rapid screening for bacteria with limits of detection reaching as low as 100 CFU/100 mL using simple fluorescent imaging. Although a less than 2-hr period is envisioned to obtain results, if combined with rotational stimulus of discs to enhance bacteria binding and a smartphone fluorescent imaging platform, one can envision an even faster and portable bacteriological screening tool.

1.4.2 Methodology

The bacteria-sized magnetic microdiscs are used to isolate target bacteria (i.e. general coliforms or E. coli). The overall process consists of pre-concentration steps to

32

localize all target bacteria from up to 100-mL samples into a compact field of view (~400 mm2), where microscopic fluorescent imaging can be performed more quickly and with enhanced sensitivity.

Figure 1-4 illustrates how bio-functionalized magnetic microdiscs are added to a water sample, mixed, and incubated for at least 20 minutes so that magnetic microdiscs diffuse and bind to target cells (i.e. coliforms and/or E. coli). This incubation time was chosen by considering previous works [63], [64] on bacteria detection using aptamers and/or aptasensors, which ranged between 10 and 45 min, but also keeping in mind the ideal total time required for detection with our method. Next, an external magnetic field is applied using a simple permanent magnet (or a magnetic filtering setup) to isolate the cells bound to the magnetic microdiscs in suspension to an area of a few millimeters allowing easy discharge of supernatant. At this point, cells bound to the microdiscs can be either rinsed, re-suspended in a desired buffer, or retrieved by removing the permanent magnet and pipetting them out.

Figure 1-4. Overall magnetic isolation of target bacteria. A) water sample containing target cells are exposed to functionalized magnetic microdiscs, incubated and isolated with a permanent magnet, B) magnetically isolated cells are stained using fluorescent tags, and C) optical inspection is done by fluorescent microscopy.

Figure 1-5 shows the concept of our microfabricated magnetic microdiscs, their surface functionalization, and bacteria isolation from water samples. In summary, DNA

33

aptamers are used as the capture probes to functionalize our gold-coated magnetic microdiscs to isolate E. coli from water samples and detect presence and viability of E. coli using different fluorescent labels (i.e. GFP, SYTO9/PI, and CDs).

Figure 1-5. Overall concept of this research. (From left to right) 1) SEM image of magnetic microdiscs, 2) concept of surface bioconjugation using thiolated DNA-aptamers, 3) and fluorescent microscope image showing isolation of GFP E. coli from a water sample. Adapted from [57] and reprinted with permission of 2019 Elsevier B.V.

1.5 Dissertation Overview

The structure of this dissertation will be fragmented into 6 chapters. First,

Chapter 1 with the Introduction, which focused on the motivation, objectives, and previous related works of the research. Then, Chapter 2 will discuss the Background of relevant terminologies and concepts of this research, namely related to biosensors and magnetic materials. Next, Chapter 3 will cover an overview on the Bio-functionalized

Magnetic Microdisc Assay to Detect Protein-coated Particle or Bacterial Targets. Here, details on the microfabrication of the gold-coated microdiscs, their surface bio- functionalization, sample preparations, particle/bacteria targeting, and viability assay/labelling will be discussed. Later, Chapter 4 will discuss the Methods for the

Magnetic Separation of Microdiscs by themselves from water samples of up to 100 mL.

34

Followed by Chapter 5, where the Magnetic Isolation of Particle/Bacterial Targets will be described using the methods discussed in Chapter 4. Finally, Chapter 6 will focus on the Conclusions and Future Perspectives of this research.

35

CHAPTER 2 BACKGROUND

2.1 Overview

This chapter will introduce some important concepts related to biosensors and magnetic materials. In the case of biosensors, a brief overview about their mechanism and targeting will be presented. For the magnetic materials, main focus will be given to how they react when exposed to external magnetic fields. First, the classification of common magnetic materials, such as, diamagnetic, paramagnetic, and ferromagnetic will be discussed. Then, the special cases of superparamagnetic nanoparticles and the

“spin-vortex” magnetic microdiscs to be used in this work will be discussed. Finally, some details regarding the magnetic characterization and actuation of the spin-vortex magnetic microdiscs will be introduced.

2.2 Biosensors Overview

As defined by the International Union of Pure and Applied Chemistry, a biosensor is “a device that uses specific biochemical reactions mediated by isolated enzymes, immunosystems, tissues, organelles or whole cells to detect chemical compounds usually by electrical, thermal or optical signals” [65]. In summary, they can be considered as devices that “translate” a biological or chemical reaction to a measurable/near-quantitative (thermal, electrical, or optical) signal.

Although the concept of a biosensor was first introduced in the 1980’s by Dr.

Leland C Clark, Jr, who is considered one of the founding fathers of biosensors [66].

Interest and research related to biosensors have been increasing over the last decades.

In fact, according to [67], in the period of 1972-2014 publications containing the word biosensor showed an exponential increase trend. There are many applications in which

36

biosensors have demonstrated promise, including: water/food/environmental monitoring, drug discovery, and disease detection. More areas of interest or applications of biosensors are summarized in Figure 2-1.

Figure 2-1. Summarized application areas of biosensors. Adapted from [68].

The overall mechanism of biosensors can be divided in three steps: biorecognition, transduction, and signal acquisition. The biorecognition step refers to the interactions between a macromolecular structure (i.e. capture probes) on the sensor and the target (i.e. molecules, viruses, or bacterial cells). The transduction step essentially describes the bio-chemical process that causes property changes (i.e. mass, electrochemical, or optical) [15]. The signal acquisition step, as its name implies, refers

37

to the methodology used to measure the transduction step property changes. Here, a transducer provides a final “translation” (reaction to signal) that results in sensing and/or detection “interpretation” using an output interface. In Figure 2-2, the overall working mechanism of biosensors is depicted as introduced here. In Table 2-1, some key definitions and examples of biosensing related terms mentioned in Figure 2-2 will be summarized [69].

Table 2-1. Key biosensing terminologies: definitions and examples. Terminology Definition Example(s) Recognition molecules that Ligand/capture Antibodies, peptides, bind to metal surfaces probes aptamers, lectins, etc. and/or other molecules. As a target, it is a Viruses, cells (e.g. Analyte substance of interest which bacteria), DNA, enzymes, will be detected. proteins, etc. Molecule that can be attached chemically to Green fluorescent protein Fluorescent label other biomolecules to (GFP), SYTO9, propidium facilitate fluorescent iodide, etc. detection

Figure 2-2. Diagram explaining the overall mechanism of biosensors. A) example of a direct detection biosensor (no labels needed) and B) example of an indirect detection biosensor (labels needed). Adapted from [69].

38

In relation to this research, two main capture probes will be used for the biorecognition step: DNA aptamers and Con A lectins. As discussed in [70], the word aptamer has origins in the Latin aptus and mer, which respectively mean “to fit” and

“region”. Nucleic acid aptamers are short single-stranded (ss) DNA- or RNA-based oligonucleotides, which can be chemically-modified and designed using a DNA or RNA sequencing/selection process [70]. This sequencing or selection process usually consists of a systemic evolution of ligands by exponential enrichment, also known as

SELEX process. SELEX process typically consists of multiple sequences to ensure recognition of a target (i.e. cell-surface molecules (Lipopolysaccharides) or O-antigen) from a random ssDNA or ssRNA library [70]. Con A is a lectin (carbohydrate-binding) protein—derived from jack bean (Canavalia ensiformis)—that contains four binding sites specific to sugars, such as D-glucose and D-mannose [71]. Considering the mannose- terminated surface glycans on coliforms, ConA is a viable option to bind general coliforms as reviewed by [71], [72].

Another important detail about the capture probes, especially when the final goal is to attach the target to a gold surface, is that the capture probes need a ‘link’ to the gold surface. This ‘link’ can be introduced using thiols, which are organic compounds containing sulfur. Thiols have been well-studied to create self-assembled monolayers

(SAM) on gold surfaces creating strong bounds [73]. Both, DNA aptamers and Con A can be thiol-terminated prior surface-functionalization steps.

The targets of interest in this work are fecal indicating bacteria commonly used to detect and/or estimate levels of fecal contamination in food and/or water, which might represent a public health risk. Here, target cells to be studied are general coliforms and

39

E. coli, which are Gram-negative bacteria. However, other target cells such as,

Enterococci (also commonly used to monitor water quality in recreational waters) which is Gram-positive, can be evaluated if appropriate capture probes are utilized. In general, the main difference between Gram-negative and Gram-positive bacteria is how their cell walls are structured. See Appendix A for more information and Figure A-1 for details regarding Gram-negative and Gram-positive bacteria cell wall structures.

One last topic for which cell wall structures are of importance is the fluorescence staining process, particularly when viability (live/dead) information is desired. For example, commercially available nucleic acid stains, such as SYTO9 (green) and propidium iodide (red), target two different structures of cells. SYTO9 (green) can penetrate cell membranes that are intact as well as the ones that have suffered damages. However, PI (red) will only penetrate damaged membranes, which will reduce the green signal [74]. During the detection stage, fluorescent staining plays an important role in identifying the target analyte (i.e. cells). In the case of viable (i.e. live) cells, which can be tagged using SYTO9, green fluorescence is expected during fluorescence microscopy inspection, while for non-viable cells (tagged with PI) red fluorescence is expected.

In the methodology proposed in this research, the main transduction property change is the fluorescence exhibited by the (SYTO9/PI) stained bacteria attached to the functionalized microdiscs after the bio-recognition step. Referring to the terminology from Figure 2-2, the “target analytes” will be the bacterial cell (surface molecules), the

“bio-recognition element” will be the DNA aptamer- or Con A-functionalized microdiscs, and the “label” will be the fluorescence stains (SYTO9/PI). Now, the acquisition step will

40

be mostly dependent on the isolation of these gold-coated magnetic microdiscs/bacteria conjugates via magnetic capture followed by fluorescence imaging.

The next section will discuss some important concepts regarding magnetic materials and their interactions with magnetic fields. Then, in following chapters, examples of magnetic isolation setups and an envisioned portable fluorescence imaging platform will be discussed.

2.3 Magnetic Materials Overview

Magnetic materials/particles have been of interest in the last decades for different biological/biomedical applications. Some bio-applications in which magnetic materials have been studied in the past, include (but are not limited to), magnetic cell separation, magnetic resonance imaging, immunoassays, minimally invasive surgeries, and drug delivery (refer to Figure 2-3 for some more examples) [75]. This research aims to use magnetic particles (microdiscs) for bio-applications, more specifically, magnetic bacterial cell separation as part of a biosensing system. In this chapter, concepts regarding magnetic materials and particles will be introduced aiming to explain why the microdiscs behave as they do when exposed to magnetic fields and how it is useful for magnetic separation or isolation.

In order to compare different kinds of magnetic materials, and how they react when exposed to applied magnetic fields (퐻⃗⃗ ), attention must be given to their magnetic moment arrangement, their magnetization, and other characteristic properties such as, susceptibility and permeability. In magnetostatics, a magnetic moment, 푚⃗⃗ , refers to the smallest quantity of magnetism and is measured in A·m2. The magnetization, 푀⃗⃗ , refers to the magnetic dipole moment per unit volume which is measured in A/m [76], [77].

41

Figure 2-3. Summarized bio-application areas of magnetic materials [75].

Magnetic materials can be classified as diamagnetic, paramagnetic, and ferromagnetic. Classification of magnetic materials can also be made considering their bulk magnetic susceptibility (휒), or the degree of the magnetization (푀⃗⃗ ) of a magnetic material to its applied magnetic field (퐻⃗⃗ ). Diamagnetic materials (휒 ≈ −10−5) possess weak magnetization, which is aligned in parallel but opposite direction to the applied 퐻⃗⃗ field. Paramagnetic materials (휒 ≈ 10−3 − 10−5) are characterized by having weak coupling between atoms and random arrangement of magnetic moments until exposed to an applied 퐻⃗⃗ field, in which case they rotate to align in parallel to the 퐻⃗⃗ field. In the case of ferromagnetic materials (휒 = 50 − 10,000), they are characterized by strong

42

coupling between atoms and large parallelly aligned magnetic moment domains with the applied 퐻⃗⃗ field [76], [77].

Ferromagnetic materials can be divided in two general categories: hard ferromagnetic materials and soft ferromagnetic materials. Hard ferromagnetic materials are distinguished by their suitability as permanent magnets due to their ability to remain in a magnetized state once they have been magnetized previously by an applied magnetic field strong enough (퐻⃗⃗ ≥ 푀⃗⃗⃗⃗⃗푆 ). On the other hand, soft ferromagnetic materials can also get to a saturation magnetization state when an applied magnetic field is enough (퐻⃗⃗ ≥ 푀⃗⃗⃗⃗⃗푆 ), but as soon as the applied magnetic field is removed, the material loses its magnetization [77].

Figure 2-4. Magnetic moment arrangement for different magnetic materials. A) Diamagnetic, B) paramagnetic, C) ferromagnetic, D) antiferromagnetic, and E) ferrimagnetic.

Specific physical characteristics of soft ferromagnetic materials over hard ferromagnetic materials are their low coercivity and low remanence. Magnetic properties such as, coercivity, remanence, and saturation magnetization can be extracted from hysteresis curves. Coercivity of a material is the field strength needed to reverse the magnetic material to zero magnetic induction (퐵⃗ ) or magnetization (푀⃗⃗ ). Remanence of a magnetic material refers to the remanent induction (퐵⃗ ) or magnetization (푀⃗⃗ ) on a

43

material after an applied magnetic field has been reduced to zero (퐻⃗⃗ → 0). Saturation magnetization refers to the maximum magnetization when the applied field is increased indefinitely (퐻⃗⃗ → ∞) [76].

Hysteresis loops can be obtained using a vibrating sample magnetometer (VSM).

The VSM gives a direct measurement of magnetization (푀⃗⃗ ) or magnetic induction (퐵⃗ ) of a sample vibrating at a fixed frequency and taking measurements of the sample at different field strengths (퐻⃗⃗ ) [76]. Figure 2-5 shows the hysteresis loop examples for soft and hard ferromagnetic materials.

Figure 2-5. Hysteresis loop of soft and hard ferromagnetic materials. Reprinted with permission from D. Lisjak and A. Mertelj, “Anisotropic magnetic nanoparticles: A review of their properties, syntheses and potential applications,” Progress in Materials Science, vol. 95. Elsevier Ltd, pp. 286–328, 01-Jun-2018 [78]. Permission obtained via Copyrights Clearance Center.

Since special focused will be given to ferromagnetic materials, other magnetic materials that have been related to ferromagnetic materials include antiferromagnetic, ferrimagnetic, and superparamagnetic materials. Antiferromagnetic materials consist of

44

neighbouring magnetic moments that are aligned parallel but in opposite directions when exposed to a 퐻⃗⃗ field. Similarly, ferrimagnetic materials possess antiparallel magnetic moments but in this case with different magnitudes [76], [77]. Refer to Figure

2-4 for a summarized diagram of the magnetic moment behavior for different magnetic materials exposed to an applied 퐻⃗⃗ field. Lastly, superparamagnetic materials consist of small (<20 nm) ferromagnetic particles that behave as a single domain due to the exchange energy. These particles behave like paramagnetic materials in the sense that their net magnetization will be near zero due to the thermal energy even when 퐻⃗⃗ → 0.

However, when larger fields are applied, these particles will reach a saturation magnetization, similar to ferromagnetic material.

2.4 Spin-Vortex Magnetic Behavior

Another type of magnetic particle that has gained interest for biomedical applications are the spin-vortex magnetic microdiscs. Under specific dimensions

(thickness/diameter ratio) these microdiscs exhibit a unique behavior in which magnetic moments (spins) form a circular vortex, as seen in Figure 2-6.

Figure 2-6. Magnetic moments creating a circular vortex.

45

Figure 2-7 shows the relationship between the dimensions and the particle behavior. In general, it can be observed that for radius over 20 nm (up to 60 nm) and thicknesses over 40 nm (up to 120 nm) exhibited the spin-vortex behavior. However, in other works [34], [35], discs with dimensions in the micrometer range (similar to what is used in this work) have also exhibited spin-vortex behavior.

Figure 2-7. Magnetic behavior of discs with different dimensions (radius and thickness). Red arrow depicts the region that will fit the dimensions of the microdiscs being fabricated in this work. Reprinted/adapted with permission from M. Goiriena-Goikoetxea, A. García-Arribas, M. Rouco, A. V Svalov, and J. M. Barandiaran, “High-yield fabrication of 60 nm Permalloy nanodiscs in well- defined magnetic vortex state for biomedical applications,” Nanotechnology, vol. 27, no. 17, p. 175302, Apr. 2016 [79]. Permission obtained via Copyrights Clearance Center.

It is important to explain the two classifications for this kind of microdisc: 1) single-domain phase (in-plane or out-of-plane) and 2) vortex phase (spins forming a

46

circular vortex in-plane). The single-domain phase is characterized by all magnetic moments behaving as a single “giant” spin with high magnetization. On the other hand, the vortex phase, is where there is a nucleation (or center core) in the microdisc surrounded by the spins forming a vortex causing the microdiscs to have zero net magnetization. As has been showed in literature [32], [80], according to the hysteresis loops of these magnetic microdiscs there has been observations of a nucleation field and an annihilation field. In brief, the annihilation field (typically high) is the field required to “destroy,” or move towards the disc edges, the core of the spin-vortex formation on the disc. The nucleation field (typically low) refers more to the field required to form the core or the spin-vortex.

As was described in [80], the annihilation field in submicron ferromagnetic arrays, can be defined as

1 푅 ℎ (훽, 푅) = 2 (2휋퐹 (훽) − ( 0)). 푎푛 1 2 푅 (2-1)

In this Equation, 훽 is the aspect ratio of the disc thickness and its radius (L/R),

∞ 푑푡 푅 = √퐶/푀2 is the exchange length, and 퐹 = ∫ 푓( 훽푡)퐽2(푡), where 푓(푥) = 1 − [1 − 0 푠 1 0 푡 휇

2 exp(−푥)]/푥 and 퐽휇(푡) is the first order Bessel function. Similarly, the nucleation field can be described as in Equation 2-2, as defined in [80], and where 퐹(훽) = 퐹1(훽) − 퐹2(훽).

1 푅 2 ℎ (훽, 푅) = 4휋 (퐹(훽) − ( 0) ) 푛 휋 푅 (2-2)

From Equations 2-1 and 2-2, it can be observed that both, annihilation and nucleation fields, are directly dependent on the microdiscs’ dimensions (thickness and radius). In general, as observed from [80], as the diameter of the microdiscs increase, both the annihilation and nucleation fields decrease.

47

In order to obtain results for microdiscs with dimensions closer to the microdiscs to be used in this work (1.5 µm in diameter and 70 nm in thickness), a micromagnetic simulation using OOMMF (Object Oriented Micromagnetic Framework) has been implemented.

Figure 2-8. OOMMF simulation results for Permalloy microdiscs with dimensions of 1.5 µm in diameter and 70 nm in thickness. (a)-(e) show the magnetic moment configurations on the discs for each region labeled in the hysteresis curve.

48

For this simulation a saturation magnetization of 800 kA/m was used as well as a cell size of 5 nm and an initial vortex magnetization, refer to Figure 2-8 for results. Here, it was confirmed that for applied magnetic fields equal to zero, no net magnetic moment, while for higher applied magnetic fields, magnetization was observed. It is also convenient to consider the magnetic moments arrangements on the microdisc structure from the simulations (Figure 2-8a – e), which are labeled in the plot with their respective net magnetic moment and estimated applied field location. Here, (c) represents the vortex phase (after nucleation or vortex formation) with near-zero magnetization, while

(a) and (e) represent the single-domain phase (after annihilation or vortex destruction).

2.5 Magnetic Forces and Torques Overview

Another important term to be considered when working with magnetic micro/nano-particles is the magnetic force and torque that can be exerted on them for actuation. In the case of the spin-vortex magnetic microdiscs, high-magnetic forces and torques can be applied due to their high volume and hence higher magnetic moment 푚⃗⃗ .

Therefore, magnetic forces and torques play a vital role in the remote actuation and capture of magnetic micro/nano-particles. Here, the magnetic force and torque will be defined. First, as discussed earlier, the smallest quantity of magnetism in a material is the magnetic moment, 푚⃗⃗ , which has units of A·m2. Now, if we consider a magnet equivalent to a circulating electric current, as suggested by Ampère and explained in

[77], then, a magnetic moment can be represented as a very small current loop. If this

“small current loop” possess an area 퐴 (m2) and a circulating current I (A), then, the magnetic moment can be defined as

49

푚⃗⃗ = 퐼 ∫ 푑퐴 . (2-3)

As discussed in [77], [81], if we consider ∫ 푑퐴 = 푆 , then, the magnetic moment can be defined as

푚⃗⃗ = 퐼푆 (2-4) where 푆 is the total vector area enclosed in the current loop. Now, if we consider a system with multiple “small current loops” or magnetic moments, then the magnetic moment can be defined as

푚⃗⃗ = ∑ 푚푗 = ∑ 퐼푗푆⃗⃗⃗𝑗 푗 푗 (2-5) where 푚⃗⃗ is now the vector sum of all the individual magnetic moments present in the system [81].

Now, as described in Chapter 19 from [81], if we consider the magnetic interaction energy, 푈푚, of the system under an external magnetic induction or magnetic flux density, 퐵⃗ , then

⃗ 푈푚 = 퐼 ∫ 퐵 ∙ 푑퐴. (2-6) 푆

If multiple “small current loops” or magnetic moments are considered again as in

Equation (2-5), then Equation (2-6) can be expressed as

푈 = ∑ 퐼 ∫ 퐵⃗ ∙ 푑퐴⃗⃗⃗ . 푚 푗 𝑗 (2-7) 푗 푆푗

If the magnetic induction, 퐵⃗ , is approximated as constant and the Equation (2-5) is considered, then Equation (2-8) can be simplified to

50

푈 = 퐵⃗ ∙ ∑ 퐼 ∫ 푑퐴⃗⃗⃗ = 퐵⃗ ∙ ∑ 퐼 푆 = 푚⃗⃗ ∙ 퐵⃗ . 푚 푗 𝑗 푗 푗 (2-8) 푗 푆푗 푗

Finally, if we consider the relationship between energy and the force, as described and derived in Chapters 18-19 from [81], we obtain that the magnetic force,

퐹⃗⃗⃗푚⃗ , on a magnetic dipole, or a particle, is defined as

퐹⃗⃗⃗⃗ = ∇푈 = ∇(푚⃗⃗ ∙ 퐵⃗ ). 푚 푚 (2-9)

This expression has been used interchangeably with 퐹⃗⃗⃗푚⃗ = (푚⃗⃗ ∙ ∇)퐵⃗ . The reasons for this have been discussed and explained in [82] and summarized here. If we used vector identity (∇(퐴 ∙ 퐵⃗ )) for ∇(푚⃗⃗ ∙ 퐵⃗ ), we obtain

퐹⃗⃗⃗⃗ = ∇(푚⃗⃗ ∙ 퐵⃗ ) = (푚⃗⃗ ∙ ∇)퐵⃗ + (퐵⃗ ∙ ∇)푚⃗⃗ + 푚⃗⃗ × (∇ × 퐵⃗ ) + 퐵⃗ × (∇ × 푚⃗⃗ ) 푚 (2-10) As explained by T. H. Boyer in [82], if we take as an assumption that 푚⃗⃗ has no spatial dependence, then Equation 2-10 is simplified to

퐹⃗⃗⃗⃗ = ∇(푚⃗⃗ ∙ 퐵⃗ ) = (푚⃗⃗ ∙ ∇)퐵⃗ + 푚⃗⃗ × (∇ × 퐵⃗ ) 푚 (2-11)

From Equation 2-11, term (∇ × 퐵⃗ ) can be substituted by the current density relationship from Ampere’s law, where 휇0퐽 = ∇ × 퐵⃗ . Here, if we assume that there is no current density, which leads to 휇0퐽 = ∇ × 퐵⃗ = 0, then Equation 2-11 can be simplified to

⃗⃗⃗⃗ ( )⃗ 퐹푚 = 푚⃗⃗ ∙ ∇ 퐵. (2-12) As a summary, it can be observed that the magnetic force on a particle is directly related and proportional to the magnetic gradient of the magnetic flux density.

Now, the magnetic torque present on a magnetic dipole in the presence of a magnetic field is defined as 휏 = 푚⃗⃗ × 퐵⃗ , where 푚⃗⃗ is also the magnetic moment and 퐵⃗ the

51

magnetic flux density. This implies that a torque will be generated as 퐵⃗ tries to align the dipole in order to have the magnetic moment, 푚⃗⃗ , in alignment with 퐵⃗ [76].

2.6 Summary

This chapter discussed some fundamental terminology regarding biosensors and magnetic materials. Additionally, applications of biosensors and magnetic materials in bio-relate fields were mentioned. Also, definitions of most commonly used magnetic materials, including the unique spin-vortex magnetic behavior of the microdiscs used in this work, were introduced. Lastly, details related to magnetic forces and torques were described, which will be of interest for the magnetic isolation methods.

52

CHAPTER 3 BIO-FUNCTIONALIZED MAGNETIC MICRODISCS ASSAY TO DETECT PROTEIN- COATED PARTICLE OR BACTERIAL TARGETS*

3.1 Overview

In this chapter, details regarding the microdisc microfabrication, their surface bio- functionalization, particle/bacterial target sample preparations, viability assay procedures, and magnetic isolation of targets from water samples will be introduced.

Here, protein-coated particles and bacteria were evaluated as biological targets. Also, different options of bio-functionalization of microdiscs for those targets were studied.

More specifically, targets studied included: avidin-coated polystyrene particles and general coliforms, and E. coli; while bio-functionalization capture probes studied included: biotin, lectins, and aptamers, respectively.

The main reason why avidin-coated polystyrene particles were selected as targets was the fact that thiolated biotin could be used to bio-functionalize our magnetic microdiscs and use as a proof-of-concept for our biosensing application [83]. Also, biotin-avidin interactions have been extensively studied and it has been demonstrated that the biotin-avidin bond is considered one of the strongest bonds known for biological interactions (between a ligand and a protein) [84]. Now, after this proof-of-concept using biotin-avidin, more specific examples were chosen for water quality monitoring applications (i.e. bacterial targets). As was discussed in previous chapters, for water quality monitoring applications there is interest to detect fecal coliforms, including E. coli in water samples (drinking and recreational). Therefore, in this work, general coliforms

* Portions of this chapter have appeared in [57] and are reprinted (adapted) with permission from K. Y. Castillo-Torres, D. P. Arnold, and E. S. McLamore, “Rapid isolation of Escherichia coli from water samples using magnetic microdiscs,” Sensors and Actuators: B chemical, 291 (2019), 58-66. Copyright 2019 Elsevier B.V.

53

including E. coli, K. variicola, and P. mirabilis, were evaluated as the bacterial targets.

Now, in the case of these bacterial targets, lectins and aptamers were chosen depending on what bacteria was being targeted, as was discussed in previous chapters.

As a summary, lectins (i.e. Con A) contain binding sites specific to sugars, such as D- glucose and D-mannose, which are present on general coliforms [71], [72], while nucleic acid aptamers are short single-stranded (ss) DNA- or RNA-based oligonucleotides, which can be chemically-modified and designed using a DNA or RNA sequencing/selection process to ensure recognition of a specific target, in this case E. coli [70]. Therefore, from a proof-of-concept using proteins-coated particles, we continued to study the targeting and isolation of a general population of bacterial targets

(i.e. general coliforms, including E. coli, K. variicola, and P. mirabilis), and finished with a very specific bacterial target from the same population (i.e. E. coli).

In this chapter, another topic that will be discussed is the fluorescence labelling, which only applies to bacterial targets, since avidin-coated particles were purchased with fluorescent tags embedded. Also, a brief description of bacteria targeting and magnetic isolation of targets using magnetic separation setups during experiments will be introduced.

In the meantime, Table 3-1 summarizes the samples prepared and used for the magnetic capture experiments using bio-functionalized magnetic microdiscs and particle/bacterial targets. Here, biosafety level 1 (BSL-1) lab-prepared samples with E. coli only, general coliforms (E. coli, K. variicola, and P. mirabilis), an activated sludge with a mix of particles and coliforms, and avidin-coated polystyrene particle, were

54

evaluated. Additionally, some of the lab-prepared samples were also evaluated by diluting them in a complex matrix (i.e. vegetable broth).

Table 3-1. Summary of water samples prepared for experiments Water sample Capture Limit of Analyzed Sample Type source probes Detectability sample volumes Lab-prepared protein- Spherotech avidin- Biotin Up to 300×109 0.5-100 mL coated particles cated polystyrene particles/100 mL (avidin) particles Lab-prepared coliforms EZ Hydro Shot DNA aptamers 2-10 CFU/100 mL 100 mL (E. coli, K. variicola, coliform/E. coli kit and and P. mirabilis) (BSL-1) Con A lectins Lab-prepared E. coli EZ CFU E. coli kit DNA aptamers 1-10 CFU/100 mL 0.5 mL (BSL-1) and 100 mL Activated sludge (E. Domestic Con A lectins N/A 0.5 mL coli and coliforms) wastewater (BSL-1)

3.2 Microfabrication of Gold-Coated Magnetic Microdiscs

The magnetic microdiscs were microfabricated using standard microfabrication techniques. Refer to Figure 3-1 for a detailed diagram of the microfabrication process. A densely packed array of circular polymer pillars (1.5 µm in diameter) were formed on a

100-mm-diameter silicon substrate similar to [36], [37], [57]. Starting with the cleaning process, <100> p-type Silicon test wafers were submerged in positive resist stripper

(PRS3000) at 70°C for 3 minutes, rinsed three times in DI water and dried with a nitrogen gun. Then, wafers were exposed to buffer oxide etching (BOE) for 30 seconds, rinsed, dried, and placed in the oven at 150°C for 3 minutes for complete de-hydration.

Then, a 100-nm layer of tungsten was deposited as a sacrificial layer, at a deposition rate of 1 Å/s, using magnetron sputtering (Kurt J. Lesker multi-source RF and DC sputter system) followed by spin-coating a 300-nm thick layer of lift-off resist (LOR 3A) and a 800-nm thick layer of a positive photoresist (AZ 1512). Polymer resists (LOR and

AZ 1512) were patterned by standard UV exposure (vacuum contact for 11.5 seconds

55

and developed in AZ300MIF for 1 min). Then, gold and permalloy (Ni80Fe20) metals were deposited also by magnetron sputtering followed by an ultra-sonicated lift-off process to obtain the permalloy magnetic microdisc array on the sacrificial layer using

AZ400K diluted in water (1:4). The thickness of the magnetic material (permalloy) was

70 nm and it was sandwiched between two gold layers of 5 nm each. Finally, the microdiscs were released by dissolving a tungsten sacrificial layer using 30% hydrogen peroxide and rinsed three times using deionized (DI) water and a permanent magnet.

Figure 3-1. Microfabrication process schematic. a) 100-nm layer of tungsten as a sacrificial layer, 300-nm thick layer of LOR 3A, and 800-nm thick layer of AZ 1512, b) metal stack (gold and Ni80Fe20) deposition c) ultra-sonicated lift-off process to obtain the permalloy magnetic microdisc array on the sacrificial layer, and d) release of discs by dissolving the tungsten sacrificial layer.

56

Figure 3-2. SEM images of patterned on substrate and lifted-off microdiscs.

Figure 3-3. Optical profilometer data confirming a thickness of ~ 70 nm for our array of magnetic microdiscs.

Figure 3-2A and 3-2B present scanning-electron microscopy (SEM) images of magnetic microdiscs a) patterned on the sacrificial layer of the substrate and b) after

57

disc release. Also, in Figure 3-3 an image obtained from an optical profilometer is presented, where the total thickness of the microdiscs was approximately 62 nm for that particular sample, from which 80 nm thickness was expected. This disagreement might have been affected by the sensitivity of the tool used (Contour GT-I Bruker Optical

Profilometer) or by some microfabrication effects, most likely during metal sputtering.

3.3 Surface Bio-Functionalization of Magnetic Microdiscs

In order to make gold-coated magnetic microdiscs useful for biosensing applications, as has been mentioned, the microdisc surfaces were bio-modified

(“decorated”) using different bio-molecules or capture probes. The capture probes selected were carefully considered to make the microdiscs selective to the specific biological target (e.g. avidin-coated particles or bacteria). As already mentioned, capture probes to be used in this work include, biotin, lectins, and aptamers. In Appendix B, more details regarding the selectivity of lectins and aptamers towards target bacteria are described.

3.3.1 Surface Bio-Functionalization for Avidin-Coated Particles

For avidin-coated polystyrene particles, thiolated biotin was used for surface bio- functionalization of the gold-coated microdiscs. Biotin-thiols (5 kDa, Biotin-PEG-SH) were purchased from NanoCS, Inc. (Boston, MA). For bio-functionalization, microdiscs

(~5 million microdiscs) were suspended overnight in biotin-thiols solution (~300 µL) with a concentration of 1 mg/mL. Finally, after overnight storage, discs were rinsed three times with DI water using a permanent magnet to decant the supernatant.

As a reminder, biotin was selected as a capture probe for these particular portion of the work, considering that biotin-avidin bond is considered one of the strongest bonds known and studied for biological interactions [84].

58

Figure 3-4. Magnetic microdiscs microfabrication and bio-functionalization using A) Con A (lectins), B) DNA aptamers, and C) biotin.

3.3.2 Surface Bio-Functionalization for Bacterial Targets

Microdiscs were bio-functionalized for selective binding to a biological target (i.e. general coliforms or E. coli) using thiolated capture probes (i.e. Concavalin A lectins or

DNA aptamers). For specific capture of general coliforms, thiol-terminated mannose binding lectin Concavalin A (Con A) were used as the capture probes, which binds mannose-terminated surface glycans on coliforms as reviewed by [71], [72]. Figure 3-4 depicts bio-functionalization concepts for avidin-coated particles, general coliforms, and

E. coli.

Con A lectins used (Con A from Sigma Aldrich) were diluted in DI water (1.5 mg/mL). Aptamers that specifically bind to E. coli 25922 cells [63] (P12-55; 88 mer;

MW=11.8 kDa; KD=0.83 nM synthesized by GeneLink) were reduced as described in the Thiol modified Oligo Disulfide Reduction protocol by GeneLink [85]. The detailed sequence of the aptamer used is depicted in Appendix B along with the structure of a control scrambled aptamer (Figure B-4) tested to compare the specificity of the P12-55 aptamer towards the E. coli ATCC 25922 strain.

59

Experimentally, the process of bio-functionalizing the magnetic microdiscs consists of suspending a desired number of discs (e.g. 1-5 million discs) in 200-300 µL of capture probe solutions overnight at room temperature. Similarly, after overnight storage, discs were rinsed three times with DI water using a permanent magnet to decant the supernatant. Figure 3-5 shows a diagram with the overall process followed for experiments. Here, bio-conjugation of gold-coated discs, bacteria targeting, fluorescent staining and imaging are described.

Figure 3-5. Overall experimental procedure for disc functionalization, staining, and sample preparation for imaging.

3.4 Sample Preparation

In general, samples processed were lab prepared. In the case of protein-coated particles, surface modified particle targets (i.e. avidin-coated polystyrene particles) were purchased from Spherotech, Inc. For bacterial targets, EZ hydro shot E. coli/Coliform and EZ CFU Escherichia coli kits were purchased from Microbiologics, Inc. Some confirmation test results of bacteria growth from samples prepared with these kits are presented in Appendix C.

3.4.1 Protein-Coated Particle Target Samples

Protein-coated polystyrene samples were prepared using avidin-coated fluorescent particles (yellow, 0.7-0.9 µm) from Spherotech, Inc. (Lake Forest, Illinois).

60

Stock concentration of these particles was of 3.07×109 particles/mL, as specified from manufacturer. Avidin-coated particles samples were diluted down to 300 particles/mL for preliminary experiments at volumes up to 100 mL, which will be discussed in following chapters.

3.4.2 Bacterial Target Samples

The stock concentration for the lab-prepared samples used for the E. coli/Coliform kit was reported by the manufacturer to be 20-100 CFU/mL, while for E. coli kit it was 100-1,000 CFU/mL. To obtain these stock concentrations, the respective protocols for each kit from Microbiologics were followed, which are also described in

Figure C-4 from Appendix C. In brief, for the E. coli/Coliforms kit, one lyophilized pellet of each kind of bacteria available in the kit (i.e. Eschericia coli, Klebsiella variicola, and

Proteus mirabilis) were transferred into three individual vials containing 1 mL of DI water each and vortexed. This process resulted in a concentration of 20-100 CFU/mL, which was then used for our experiments. Similarly, for the E. coli kit, lyophilized pellets were equilibrated at room temperature for 30 minutes while the corresponding hydrating fluid is warmed to ~34°C. Then, using sterile tweezers, 2 pellets were transferred into the 2 mL of warmed hydrating fluid, mixed and incubated for 30 minutes in ~34°C. After this,

1 mL of the solution is transferred to 9 mL of phosphate buffer (7.2 pH) and vortexed. At this point, the working solution was ready with a concentration of 100-1,000 CFU/mL, which then was diluted to 20-200 CFU/mL (1 mL of working solution into 5 mL of 0.85%

NaCl buffer).

Viable cells were obtained as described previously. However, non-viable target cells were obtained by centrifuging the desired volume of cell solution for 3 minutes at

10,000×g (10,319 rpm), removing supernatant and substituting it with 70% ethanol. The

61

solution was then vortexed and incubated for 30 mins at room temperature. Then, the sample was centrifuged again under the same conditions and supernatant (ethanol) was substituted with desired buffer.

3.5 Bio-Functionalized Microdiscs Targeting Particles and/or Bacteria

After surface bio-functionalization of the magnetic microdiscs and preparing the biological samples (containing either avidin-coated particles or bacterial targets), the microdiscs were exposed to the samples for about 20 minutes. Typically, 5-10 million bio-functionalized microdiscs were used per sample (up to 100 mL). This process was done at room temperature and occasional stirring was applied. After microdiscs were exposed to the biological samples containing the target particles or cells, a magnet (or magnetic setup) was used to rinse and decant supernatant. After rinsing, if needed, sample was fluorescently tagged right before microscopy inspection as will be described in the next sections.

3.6 Viability Assay Preparation for Bacterial Targets

In this section, the process of fluorescence labeling and/or viability labeling will be described. As mentioned before, avidin-coated particles were fluorescently tagged by the manufacturer (Spherotech, Inc), therefore, no additional fluorescence/viability labeling steps were needed for these samples. In the case bacterial targets (i.e. coliforms and E. coli from Microbiologics, Inc.), which were not fluorescently labeled prior to the experiments, and which viability state was of interest for water quality monitoring, an additional step for fluorescence tagging was added. Additionally, as a proof-of-concept, green fluorescence protein (GFP) tagged E. coli (from EnCor

Biotechnology, Inc.) and carbon quantum dots (CDs) tagged E. coli were evaluated,

62

which also were fluorescently tagged prior to experiments, so no additional fluorescence labeling steps were needed.

3.6.1 SYTO9 / Propidium Iodide (PI) Viability Labelling

For the samples requiring fluorescence and/or viability labeling, the

LIVE/DEAD™ BacLight™ Bacterial Viability and Counting Kit, for flow cytometry from

Microbiologics was used. Here, a membrane-permeable stain (i.e. SYTO9) and a membrane-impermeable stain (i.e. propidium iodide) were used to label target cells to be analyzed using fluorescence microscopy.

Experimentally, as described by the protocol in the BacLight kit, staining of bacterial targets was done by adding 3 µl of SYTO9 nucleic acid stain and 3 µl of 20 mM propidium iodide (PI) per every 1 mL of solution containing the magnetic microdiscs bound/unbound to target cells (viable or non-viable) [74]. After a 15-minute incubation, the solutions were washed again by magnetically concentrating the microdiscs with the permanent magnet, removing the supernatant and re-suspending in new buffer. Finally, once the bound/unbound to target cells (viable or non-viable) were stained and rinsed, the microdisc-conjugates were magnetically concentrated and retrieved from the vial using a 10 µl pipette and deposited on a glass slide for microscopic imaging and inspection.

3.6.2 Other Fluorescent Labels

Carbon dots (CD) were produced as auto-fluorescent tags, in the McLamore lab, following the methods by Sun et al. in [86], with slight modification. A Nd:YAG laser

(1064 nm, 10 Hz) was used to repetitively ablate graphite powder in the presence of water vapor (from a nebulizer) and argon gas. The solution was collected from a stainless-steel gas-liquid separator, refluxed with a nitric acid solution (2.0 M), and

63

finally mixed with PEG2000 as a capping agent/passivation agent to activate the fluorescence as described by Sun et al. in [86]. For CD labeling of viable E. coli, the methods reported by Luo et al. (2006) in [87] were used. Briefly, E. coli 25922 cells were grown in TSB (Difco) at 37 ºC for 6 hours and then washed in PBS buffer. A cell suspension (1 mL) was mixed with an aliquot of CD (75 µL) and vortex mixed. The mixture was incubated on a shaker table (30 rpm) at 37 ºC for 6 hours. Samples were then added to disc solutions for confocal microscopy as described below.

GFP E. coli cells were grown to a concentration of 1011 CFU/100 mL by the manufacturer in EnCor Biotechnology, Inc. and diluted prior to our experiments and imaging.

3.7 Magnetic Isolation and Detection of Target Bacteria/Particles Overview

After targeting bacteria, the magnetic microdiscs could be actuated using an external magnetic field (e.g. a simple permanent magnet) to attract them to a specific concentrated area for retrieval. Here, a first pre-concentration step consists of isolating all the bacteria-bound and -unbound bio-functionalized magnetic microdiscs using permanent magnets. If large volume samples were being examined, these bacteria- bound and -unbound microdiscs were re-suspended in a smaller volume (1-2 mL), exposed to viability stains (if needed), and isolated using a permanent magnet as well.

At this point, the magnetic microdiscs were retrieved using a magnet and a 10-µL pipette and deposited on a glass slide for optical inspection under the microscope (Fig.

3-4).

3.8 Fluorescence Inspection

The last stage of this biosensing system is detecting the presence of biological targets using fluorescence. Fluorescence microscopes used here include epi-

64

fluorescence microscopes as well as a confocal microscope. More details regarding the setups used for the magnetic isolation of bacteria will be presented in Chapter 4. In the meantime, a general overview on targeting and detecting bacteria using the spin-vortex magnetic microdiscs is presented here.

3.9 Summary

This chapter summarized the overall protocol for the magnetic microdisc assay for the detection of biological targets, specifically, protein-coated particles and bacteria.

First, a description of the surface bio-functionalization process for magnetic microdiscs was introduced for three different particle/bacterial targets: avidin-coated polystyrene particles, general coliforms, and E. coli. Capture probes used for each of these targets were biotin, lectins, and aptamers, respectively. Also, description of microdisc fabrication and sample preparations was provided. Finally, details regarding the magnetic separation/isolation of microdiscs and biological targets (i.e. particles or bacteria) were provided. In Chapter 4, more details regarding the methods used for magnetic isolation of microdisc conjugates will be presented.

65

CHAPTER 4 METHODS FOR THE MAGNETIC SEPARATION OF MICRODISCS*

4.1 Overview

As was discussed in Chapter 3, where the magnetic microdisc assay protocol was described for the detection of protein-coated particles and/or bacterial targets, the magnetic isolation is a crucial step for this biosensing system. In this chapter, a more detailed discussion regarding the different setups used for magnetic separation of the magnetic microdiscs is presented. This chapter will mostly focus on the magnetic separation of microdiscs by themselves and the capture efficiency of the microdiscs from each of the setups presented. Therefore, one of the primary goals is to understand which platforms or setups provide a more efficient magnetic separation (or capture) of microdiscs, so that when biological targets are present in water samples a higher capture efficiency can be achieved. Also, considerations like time to filter the water samples containing the microdiscs and the sample volumes will be discussed.

The overall method followed for magnetic separation of microdiscs from up to

100 mL samples consists of at least one magnetic separation (or isolation) step.

Compared to nanoparticle-based magnetic separation approaches, the magnetic microdiscs used in this work have volumes 10,000 times larger, and thereby enable the application of much higher magnetic forces without damaging cells, resulting in an enhanced concentration of target particle/bacteria.

The first magnetic separation method to be described uses a single permanent

* Portions of this chapter are reprinted with permission from (i) K. Y. Castillo-Torres, E. S. McLamore, and D. P. Arnold, “A High-Throughput Microfluidic Magnetic Separation (µFMS) Platform for Water Quality Monitoring,” Micromachines, vol. 11, no. 1, p. 16, Dec. 2020 and (ii) K. Y. Castillo-Torres, N. Garraud, D. P. Arnold, and E. S. McLamore, “Investigation of magnetic microdiscs for bacterial pathogen detection,” 2016, vol. 9863, p. 98630G.

66

magnet (NdFeB), which is mostly used for small volume samples (i.e. <5 mL). This method was used for rinsing steps of magnetic microdiscs during bio-functionalization and rinsing steps to decant supernatant, as discussed in Chapter 3. Then, four additional magnetic separation methods evaluated using multiple magnets, to increase magnetic field gradients or “traps,” will be described. The primary focus on these four setups was to process large volume samples (i.e. 100 mL), for water quality monitoring applications, in less than 30 mins. Thus, flow rate control was introduced.

In the following sections, various magnetic capture setups will be described and depicted with simulations and diagrams. Additionally, preliminary experiments and results of magnetic separation and the capture efficiency of these setups will be summarized. Most of the prototypes were built with easy to access, disposable, and low-cost materials (such as, pipettes and intravenous (IV) lines) and reusable NdFeB magnets. Therefore, for most of these prototypes, no advanced fabrication processes were needed. However, as will be seen in the last (and most optimized version of the magnetic separation devices) a more detailed fabrication procedure will be described.

4.2 Magnetic Microdisc Capture Efficiency Quantification

One crucial factor to be evaluated for each of the magnetic separation methods is their capture efficiency, or how effectively each of the setups can magnetically separate the microdiscs from a water samples. Here, quantification of capture efficiencies was evaluated using a vibrating sample magnetometer (VSM) to measure the amount of magnetic material before and after “filtration.” More specifically, the VSM was used to measure the magnetic content in the “stock” solution (before filtration) and then either in the solution containing the “trapped” discs or the solution containing the

“not-trapped” discs (after filtration). Then, capture efficiency was calculated as

67

푣표푙 − 푣표푙 퐶. 퐸. (%) = 1 − 푠 푐 × 100 푣표푙푠 (4-1) where 푣표푙푠 is the volume of the magnetic content (i.e. magnetic particles or microdiscs) in the initial or stock concentration (i.e. sample before filtration) and 푣표푙푐 is the volume of the magnetic content in the concentrate solution after re-suspension or after flowing fresh water through the setups once microdiscs were “trapped” and magnets were removed (i.e. large-volume sample after filtration, pre-concentrated into a smaller known volume: 0.5-1 mL). Alternatively, capture efficiency was calculated as

푣표푙푠 − 푣표푙푓 퐶. 퐸. (%) = × 100 푣표푙푠 (4-2)

where 푣표푙푓is the volume of the magnetic content corresponding to particles that were not captured by the setup and hence were present in the filtrate solution (i.e. small- volume sample after filtration).

Figure 4-1. Capture efficiency terminology example diagram. Here, the solution before magnetic separation or “filtration” (i.e. “stock”) and the solutions after magnetic separation (i.e. “filtrate” and “concentrate”) are defined.

In the case of large-volume samples, “trapped” microdiscs were captured and re- suspended in 0.5 – 1 mL of fresh water, which can be defined as the “concentrate” (as observed in Figure 4-1 example diagram). The concentrate, ideally contains all the

68

captured or “trapped” microdiscs in a known volume of fresh water, which was analyzed/quantified instead of the “filtrate”, which in such large volumes would have had a significantly smaller concentration of magnetic content per unit volume (e.g. 100 mL) of water samples. The concentrations of magnetic content in the samples were of importance since the VSM possess limitations in terms of sensitivity (~ 10-7 A·m2 smallest measurable magnetic moment). Therefore, considering that typical sample preparations for VSM measurements consisted of volumes of up to 100 µL (dried droplet on a Silicon substrate), a high concentration of magnetic content (> 10-7 A·m2) was needed. Therefore, “concentrate” solutions were analyzed to determine capture efficiencies of large volume samples (100 mL), as described in Equation 4-1. Now, in the case samples with smaller volumes (up to 0.5 mL), capture efficiency quantification was calculated considering the “filtrate” solutions with “not-trapped” discs, as described in Equation 4-2.

In general, the sample preparation for quantification steps consisted of depositing and letting dry a 100 µL droplet of each filtered sample’s filtrates on a silicon substrate and the saturation magnetic moment was measured, which was then compared with the stock sample (typically 5-10 million microdiscs/mL). Then, considering that the magnetization (푀⃗⃗ ) refers to the magnetic dipole moment per unit volume as

푚⃗⃗ 푀⃗⃗ = , 푣표푙 (4-3)

which is measured in A/m [76], [77], as introduced in Chapter 2, then, the volume of the sample could be estimated from the VSM data.

As a reminder from Chapter 2, the VSM gives a direct measurement of the magnetic moment (푚⃗⃗ ) of samples vibrating at a fixed frequency, taking measurements

69

of the sample at different field strengths (퐻⃗⃗ ) [76]. From this data, saturation magnetization can be extracted, and the volume of the magnetic content can be inferred from Equation 4-3 as follows

푚⃗⃗ 푠푎푡 푣표푙 = 푚푒푎푠푢푟푒푑 ⃗⃗ (4-4) 푀푠푎푡푚푎푡푒푟푖푎푙

where 푚⃗⃗ 푠푎푡푚푒푎푠푢푟푒푑 is the saturation magnetic moment of the sample (measured in the

⃗⃗ VSM) and 푀푠푎푡푚푎푡푒푟푖푎푙 is the saturation magnetization value, which varies with materials

(for Ni80Fe20 it was been estimated as 800 kA/m). Figure 4-2 provides an example of

VSM data and the datapoints used to estimate the mean of the saturation magnetic moment of a sample can be defined as

퐻푚푎푥 1 |푚푠푎푡 (퐻)| 푣표푙̅̅̅̅ = ∑ 푚푒푎푠푢푟푒푑 (4-5) 푁 푀푠푎푡푚푎푡푒푟푖푎푙 |퐻|=퐻0 and the sample standard deviation can be defined as

퐻푚푎푥 2 1 |푚 (퐻)| 푠푎푡푚푒푎푠푢푟푒푑 ̅̅̅̅ 푠푣표푙 = √ ∑ ( − 푣표푙) . (4-6) 푁 − 1 푀푠푎푡푚푎푡푒푟푖푎푙 |퐻|=퐻0

Here, 푁 represents the number of datapoints, 퐻 is the applied magnetic field, where 퐻0 and 퐻푚푎푥 are the minimum and maximum magnetic field values were the samples are in saturation, represented in Figure 4-2 as the red and green dashed lines, respectively.

Additionally, the uncertainty (95% confidence interval; N ranging 40 – 56 datapoints) was computed using the corresponding t-value and the definition of uncertainty as

푡(0.95,푁−1)푠푣표푙 푈푣표푙̅̅̅̅̅ = . √푁 (4-7)

70

Figure 4-2. Example of VSM data before and after filtration (i.e. stock and filtrate solutions, respectively). Black dashed lines represent the average value of the saturation magnetic moment measured and the gray bands represent ± one standard deviation.

4.3 Multiphysics Simulations

In order to get more insight regarding the magnetic field, magnetic field gradient, and magnetic force in different setups, 2D and/or 3D finite element model (FEM) simulations were computed using COMSOL Multiphysics 5.5. Macrofluidic setups

(single magnet, dual magnet, and column array of magnets configurations) were simulated in 3D, while the microfluidic device setup was simulated in 2D due to

71

computational memory limitations (very small features). For all cases (2D and 3D),

Maxwell’s equations were solved within the ‘magnetic fields, no currents’ module to compute the magnetic field, field gradient, and magnetic force. While for the microfluidic device setup (2D) simulations, the Navier-Stokes’ equations were solved within the

‘laminar flow’ and ‘particle tracing for fluid flow’ modules to obtain the flow and particle velocities. Also, for these 2D simulations, the drag force acting on the particles and their trajectories were studied within the ‘particle tracing for fluid flow’ module.

Now, in more detail, as was explained in Chapter 2, the magnetic force that will act on the magnetic microdiscs can be expressed 퐹⃗⃗⃗푚⃗ = (푚 ∙ ∇)퐵⃗ , where 푚 is the magnetic moment of the magnetic microdiscs when in saturation. Also, if we consider

⃗⃗⃗⃗ 퐹푚 = 퐹푚푥푥̂ + 퐹푚푦푦̂ + 퐹푚푧푧̂ , then each component can be defined as

휕퐵 휕퐵 휕퐵 (4-8) 퐹 = (푚 푥 + 푚 푥 + 푚 푥) ̂푥 푚푥 푥 휕푥 푦 휕푦 푧 휕푧

휕퐵푦 휕퐵푦 휕퐵푦 (4-9) 퐹 = (푚 + 푚 + 푚 ) 푦̂ 푚푦 푥 휕푥 푦 휕푦 푧 휕푧

휕퐵푧 휕퐵 휕퐵 (4-10) 퐹 = (푚 + 푚 푧 + 푚 푧) 푧̂ 푚푧 푥 휕푥 푦 휕푦 푧 휕푧

Then, the magnitude of the magnetic force can be obtained and plotted in

COMSOL as a normalization of the 푥, 푦, and 푧 components as

2 2 2 (4-11) |퐹⃗⃗⃗푚⃗ | = √퐹푚푥 + 퐹푚푦 + 퐹푚푧

which is for 2D simulations would consider 퐹푚푧 = 0.

4.3.1 Magnetic Simulations

The first set of simulations modeled the first prototypes used for microdisc magnetic separation. Namely, a single magnet, a pair of magnets (attracting and

72

repelling configurations), and a column array of magnets with alternating polarizations.

The results of these simulations were expected to provide information regarding the magnetic flux density 퐵⃗ , magnetic field gradient ∇퐵⃗ , and the magnetic force 퐹⃗⃗⃗푚⃗ that would be exerted on the magnetic microdiscs when in suspension.

Figures 4-3 to 4-7 show simulations results for all the prototypes explored in this work. Here, results on the B-field produced from each of these setups as well as the magnetic force to be exerted of the magnetic microdiscs are presented. Other information such as the direction of the B-field is also plotted (grey arrows) as well as the magnetization direction of the permanent magnets (white arrows).

It can be observed that the magnetic force was plotted, instead of the magnetic field gradient. However, the magnetic force is directly proportional to the magnetic field gradient, as has been discussed: 퐹⃗⃗⃗푚⃗ = (푚0 ∙ ∇)퐵⃗ . From results it can be observed that the magnetic force is higher where magnetic gradients are expected (increase or decrease of B-field). Higher magnetic forces can be exerted on magnetic microdiscs in the red areas from observed in Figures 4-3 to 4-7, which we could consider the

‘magnetic traps.’ Magnetic forces for all setups ranged in the 4.3 × 10−10 푁 to

33.8 × 10−10 푁, or magnetic field gradients of 4.0 × 103 푇/푚 to 34.2 × 103 푇/푚.

Previous works on magnetic separation, using micromagnets, have reported magnetic field gradients ranging up to 2.5 × 105 푇/푚 [52]. However, using much larger bulk magnets was an attractive option in this work for re-usability purposes. Hence, with magnetic forces higher than 0.4 nN, the setups considered and simulated in this work, represented an interesting and viable tool for effective magnetic capture of microdiscs.

73

In the next sections, more details regarding experiments for the evaluation of these setups will be introduced. In brief, experiments were performed to determine the capture efficiency of magnetic microdiscs in samples ranging from 0.2 to 100 mL in the different magnetic separation setups. The magnetic separation setups include configurations using a single magnet, a pair of magnets (attracting and repelling configurations), a column array of magnets with alternating polarizations, and a microfluidic device with a chessboard 3-by-3 magnet array also with alternating polarizations.

Moreover, challenges with the ‘bulky’ or macrofluidic setups during experiments will be described, which led to the use of microfluidic devices combined with the use of magnetic traps (array of magnets). Additionally, more details on the microfluidic devices, for both multiphysics simulations and experiments, will be discussed. Here, high flow rates (up to 120 µL/s) were evaluated in order to see their effect on magnetic particle capture efficiency. Additionally, high flow rates were desired for water quality monitoring applications, specifically to process large-volume samples (i.e. 100 mL) in a short period of time.

Single magnet

B-field (퐵⃗ ) Magnetic Force (퐹⃗⃗⃗푚⃗ = (푚0 ∙ ∇)퐵⃗ )

Figure 4-3. COMSOL simulation results for single magnet configuration.

74

Dual magnet (attracting configuration) B-field Magnetic Force

Figure 4-4. COMSOL simulation results for pair of (attracting) magnets configuration.

Dual magnet (repelling configuration) B-field Magnetic Force

Figure 4-5. COMSOL simulation results for pair of (repelling) magnets configuration.

Column array of magnets B-field Magnetic Force

Figure 4-6. COMSOL simulation results for configuration using a column array of magnets with alternating polarization.

75

Microfluidic device B-field Magnetic Force

Figure 4-7. COMSOL simulation results for 3x3 array of magnets with alternating polarization configuration and microfluidic device setup.

4.3.2 Microfluidic and Particle Tracing Simulations

The second set of simulations consisted of more details on the magnetic particle capture efficiency using the 3-by-3 array of magnets with alternating polarizations and the microfluidic device. Magnetic capture of particles using 2D finite element modeling in COMSOL Multiphysics® was performed at different flow rates. Here, the cross- sectional view presented in Figure 4-8 was designed, consisting of a single microfluidic channel (length: 45 mm, and height: 60 µm), 3 magnets (6.35 mm x 6.35 mm) with alternating polarizations, and a glass-slide spacing (1 mm) in between. Free triangular meshes were defined with minimum element sizes of 3 µm (for the microfluidic channel), 31.8 µm (for the glass slide), and 76.2 µm (for remaining structure). Here, two different kinds of particles were studied, IONs (Sigma-Aldrich <50 nm) and our custom magnetic microdiscs (1.5 µm in diameter, 70 nm in magnetic material thickness), which were simulated as spherical particles with three different hydrodynamic diameters (Dhd):

70 nm (lower bound), 618 nm (equivalent magnetic volume of microdisc: 1.24 × 10-19 m3), and 1.5 µm (upper bound). In the case of the IONs, they were simulated with hydrodynamic diameters of 50 nm as defined by Sigma-Aldrich (<50 nm) and 124 nm

76

considering the hydrodynamic diameter measured in [37], as well as many other diameters accounting for particle aggregation. In Appendix D, Figure D-1, all Dhd simulated for IONs are shown with their respective capture efficiencies. Figure 4-7 shows more details on the simulated geometry. Simulations for each of the particles

(IONs and microdiscs) were performed by changing the flow rates (Q = 5, 15, 30, 60, and 120 µL/s), to compare with later experiments. Here, 1,000 particles were released in the inlet and simulated for 8 s with time steps of 0.01s. Finally, after each simulation, particles in the outlet were counted using the global evaluation (‘total number of particles in selection’ option), and capture efficiency was computed as follows

#푃푎푟푡푖푐푙푒푠 − #푃푎푟푡푖푐푙푒푠 퐶. 퐸. % = 푏푒푓표푟푒 푓푖푙푡푟푎푡푖표푛 푎푓푡푒푟 푓푖푙푡푟푎푡푖표푛 × 100 (4-12) #푃푎푟푡푖푐푙푒푠푏푒푓표푟푒 푓푖푙푡푟푎푡푖표푛 where the number of particles before filtration were the number of particles released at the inlet (i.e. 1,000) and number of particles after filtration are the number obtained from

‘total number of particles in selection’ at the outlet.

Figure 4-8. COMSOL geometry used for simulations of the µFMS device, isolating ‘microdisc’ particles with a Dhd of 618 nm.

77

4.4 Magnetic Separation Configurations

4.4.1 Single Magnet Configuration Setup

Magnetic separation of microdiscs from small-volume samples (i.e. less than 5 mL) was performed using a simple permanent magnet (NdFeB), as depicted in Figures

4-9 – 4-10.

Figure 4-9. Single permanent magnet setup diagram.

Figure 4-10. Isolated discs from a small-volume sample (~5 mL) using a simple permanent magnet. Photo courtesy of author.

The specific magnets used were nickel-plated neodymium magnets, from K&J magnetics (Pipersville, PA). Here, the samples contained in a small vial and after localized in a small area using the permanent magnet, a 10 µL pipette was used to

78

retrieve the microdiscs and deposit on a glass slide for inspection under a microscope.

In cases where samples were intended to be suspended in smaller volumes (e.g. from 1 mL to 0.5 mL), the same procedure was performed to decant supernatant and resuspend microdiscs in the desired final volume. Capture of discs was performed successfully and microdiscs were imaged using optical microscopy. However, for more efficient capture of microdiscs and hence of target particles or bacteria from larger volumes of samples (i.e. 100 mL) other methods were considered, which will be discussed in the following sections.

4.4.2 Dual Magnet (Attracting and Repelling) Configuration Setups

Considering the need for processing higher volumes of samples in short times, a fluidic platform was prototyped using multiple magnets. Dual magnet configuration setups consisted of a pair of magnets in two different configurations: attracting and repelling. Also, for the fluidic part of this prototype, an intravenous (IV) line was used as the fluidic channel that allowed flow rate adjustment as depicted in Figure 4-11.

Samples of up to 100 mL were “filtered” three times at a flow rate of around 0.42 mL/s using these setups. After filtering steps, magnets were removed and fresh water

(0.5 – 1 mL) was passed through the tubing to rinse/retrieve the discs “trapped,” which was then used to study capture efficiencies. Figure 4-11 shows a diagram of both setups.

From quantification results, obtained from the fresh water retrieving the

“captured” discs in the setups, it was observed that around 40-60% of the discs were being “captured”. This led to more detailed experiments to try to increase this capture efficiency. Some details tested were the treatment of the tubing using Fluoropel to

79

create a hydrophobic surface coating in the tubing walls and decrease surface tension forces, as well as increasing flow rates and rinsing steps. However, capture efficiencies seemed to remain close to 40-60%, as summarized in plot from Figure 4-23.

Figure 4-11. Pair of magnet configurations (attracting and repelling).

4.4.3 Column Array of Magnets with Alternating Polarizations

In this iteration of the design, a column consisting of 8 magnets with alternating polarizations was tested, with the idea of increasing the magnetic field gradient spots that will act as magnetic “traps” for the microdiscs. Here, the same tools were used:

NdFeB magnets and an IV line to control the flow rate. The difference of this setup was that the magnets were arranged in a column with alternating polarizations as shown in

Figure 4-12.

As mentioned, quantification of the magnetic material present in the solutions 1) before filtration (from stock concentration of microdiscs) and 2) after filtration (from fresh sample of water that was used to rinse the “trapped” microdiscs from the tubing) was

80

performed as described in Equation 4-1. Quantification results showed that a significant number of the magnetic microdiscs were getting “stuck” in the setup tubing, even after consecutive washing steps, which might explain the 40-60% capture efficiency from previous setups (i.e. dual magnet). Figure 4-13 shows the “stuck” magnetic microdiscs in the tubing.

Figure 4-12. Magnetic separation setup using a column array of magnets with alternating polarizations. Here, discs are incubated with water samples (containing bacteria). Then, sample is filtered using the magnetic capture setup ad finally discs/bacteria are retrieved by removing external magnets and flowing fresh water. Finally, bacteria conjugates (bound to discs) are stained with SYTO9/PI and retrieved again for sample preparation and/or imaging.

81

Figure 4-13. Setup showing discs “stuck” in the tubing after removing magnets and multiple washes. Photo courtesy of author.

However, even with the discs “stuck” to the walls, these setups demonstrated to successfully filter (isolate and retrieve) a portion of the magnetic microdiscs in the samples (40-60%) using the last 3 methods (i.e. dual magnets and column array of magnets with alternating polarizations).

From these setups, it was concluded that a setup that allowed sample preparation on a glass slide that provided a ready-to-inspect platform under a fluorescence microscope was desired, which will be discussed in the next section.

4.4.4 Microfluidic Magnetic Separation (µFMS) Device

An important observation from all the previous setups was that some of the captured magnetic microdiscs were “stuck” in the tubing walls even after multiple washing steps, potentially because of surface tensions or other attraction forces. This represented a challenge when there is interest in capturing and inspect most of the microdiscs and its conjugates (i.e. target cells) from large volume samples for water

82

quality monitoring applications. Therefore, in order to localize microdiscs and potentially the conjugated target bacteria from 100-mL samples in a compact field of view so that microscopic imaging could be performed more quickly and with enhanced sensitivity, the following method using a microfluidic device bonded to a microscope glass slide and attached to a 3-by-3 array of magnets with alternating polarizations was tested.

In the next subsections, more details regarding the microfluidic magnetic separation (µFMS) device will be provided. Here, the fabrication of the microfluidic device will be described as well as the complete setup used for sample processing.

Also, a description of the magnetic isolation of magnetic microdiscs will be provided as well as details regarding the experiments and results on quantification of the capture efficiency of the µFMS device for two different magnetic particles, namely, the microfabricated gold-coated magnetic microdiscs (Ni80Fe20, 1.5 µm in diameters and 80 nm in thickness) and commercially available magnetic nanoparticles (Fe2O3 nanopowder, <50 nm particle size ) purchased from Sigma-Aldrich (St. Louis, MO).

4.4.4.1 µFMS device fabrication

The microfluidic device was fabricated using a standard soft-lithography process

[88]. Briefly, polydimethylsiloxane (PDMS) was prepared by mixing the base and curing agent solutions at a ratio of 10:1. The PDMS was then degassed using a vacuum chamber, deposited on a SU-8-silicon master mold (containing the desired microfluidic channel design), and degassed again. Next, the mold was cured in an oven at 65 ºC for

4 hr, treated with ultraviolet ozone (UVO) for 5 minutes to activate bonding between glass slide and PDMS, bonded, and placed in the oven for another 30 min to ensure chemical bonding between (PDMS and glass) surfaces. The master mold used here was fabricated using SU-8 and standard photolithography techniques on a silicon

83

substrate. Each microfluidic device was designed with 8 microfluidic channels with dimensions of 2.1 mm x 45 mm and a height of 60 µm [89], [90]. These 8 microchannels were originally connected to a single inlet and a single outlet. However, due to the high- throughput needs (100-mL samples, up to 120 µL/s), a total of 4 outlets were used to avoid pressure-induced delamination of the PDMS from the glass slide.

Figure 4-14. Diagram of the PDMS device attached to a glass slide (ready-to-inspect) setup with detachable magnet array. Reprinted from [91] with permission from K. Y. Castillo-Torres, E. S. McLamore, and D. P. Arnold, “A High-Throughput Microfluidic Magnetic Separation (µFMS) Platform for Water Quality Monitoring,” Micromachines, vol. 11, no. 1, p. 16, Dec. 2019.

The magnetic assembly comprised a 3x3 array of neodymium (NdFeB) magnets

(each a 6.35 mm x 6.35 mm x 6.35 mm cube) arranged with alternating polarizations in a chess-board pattern. The maximum field gradients (and hence magnetic forces acting

84

on the magnetic discs) occurred at the boundaries between the magnets. The magnet array was attached to the glass slide of the microfluidic device using removable adhesive film. A syringe pump was used to control the flow rate of sample solutions into the microfluidic device during experiments. Figure 4-14 shows an image of the magnetic separation microfluidic device experimental setup.

4.4.4.2 µFMS device experiments

In this work, two main experiments were performed to evaluate capture efficiencies of the microfluidic magnetic separation (µFMS) device using both microdiscs and iron-oxide nanoparticles (IONs): 1) separation of magnetic particles from 0.2-mL samples at various flow rates and 2) large-volume separation of magnetic particles from

50 – 100 mL samples. Considering the relationship between magnetic forces acting on particles, the magnetic moment of particles, and their volumes (Equations 2-12 and 4-

4), our microdiscs, larger in size and magnetic material volume (diameter: 1.5 µm, thickness: 70 nm, volume: 1.24×10-19 m3) than typical individual IONs (diameter: 50 –

124 nm, volume: 6.54×10-23 – 99.8×10-23 m3), should experience higher magnetic forces resulting in better capture efficiencies during µFMS.

First, experiments with microfabricated magnetic microdiscs were performed in both small- and large-volume samples (0.2- and 100-mL). Concentrations of magnetic microdiscs solutions tested were 6.5 µg/mL and 0.065 µg/mL for 0.2 and 100-mL samples, respectively. The concentration of the 100-mL sample with microdiscs represented an immeasurable magnetic volume sample in the VSM, but it was limited by microfabrication costs. Therefore, experiments with IONs were replicated for large- sample volumes with much higher concentrations of magnetic particles (12.5 – 100

µg/mL) in an effort to obtain a measurable magnetic volume sample for VSM

85

quantification (as will be described in the next sections), as well as to evaluate possible clogging of microchannels of the µFMS device.

Small-volume sample solutions were prepared with 1) microfabricated microdiscs and 2) IONs (Fe2O3 nanopowder, <50 nm particle size) purchased from Sigma-Aldrich

(St. Louis, MO). Then, five 0.2-mL samples of the disc solution, with a fixed concentration of ~6 million discs/mL, was filtered through the µFMS device at different flow rates: 5, 15, 30, 60, and 120 µL/s (to compare with COMSOL Multiphysics simulations). Additionally, two 0.2-mL samples of the IONs solution, with a fixed concentration of ~3,000 million particles/mL, were filtered through the µFMS device at two different flow rates: 5 and 120 µL/s.

Capture of microdiscs from large-volume samples (up to 100 mL) was assessed by filtering the samples with the µFMS device at a concentration of 0.065 µg/mL as a proof-of-concept. Then, capture of IONs was performed in 50- and 100-mL samples with concentrations of 100 µg/mL and 12.5 µg/mL, respectively. The importance of this experiment relied on our goal of isolating magnetic microdisc/bacteria conjugates from

100-mL samples in a short time period. Hence, these samples were filtered at a flow rate of 120 µL/s (up to 14-min filtering times) using a syringe pump with a 10-mL syringe

(up to 10 times, for 100-mL samples).

Figure 4-15 shows the experimental setup used for the µFMS device experiments on capture efficiency. Here, the µFMS device is shown along with a syringe pump used to regulate the flow rate of the fluid. Also, Figures 4-16 shows closer images of the µFMS device before and after filtration, where grey spots are observed after

86

filtration in the channels, which are the microdiscs trapped by the already removed magnets.

Figure 4-15. Magnetic separation microfluidic device experimental setup diagram and image. Photo courtesy of author.

Finally, Figure 4-17 shows an optical microscope image confirming the magnetic separation of microdiscs after magnetic separation using the µFMS device. In this case, the magnetic microdiscs which were suspender in a 100 mL water sample were filtered using the µFMS device at a flow rate of 120 µL/s. These represented promising results for processing of large-volume samples for water quality monitoring applications.

87

Figure 4-16. µFMS device with alternating polarization array of magnets (top) and magnetically separated discs on the device (bottom). Photos courtesy of author.

Figure 4-17. Example image of discs magnetically separated using µFMS device.

88

4.4.4.3 Estimation of magnetic capture efficiency of µFMS device

The total magnetic moment of discs in solution was measured using a VSM, similar to a method described by C. M. Earhart, et al. [29]. Also, similar to what was described earlier in this chapter, the volume of the magnetic content in the sample before and after filtration were estimated (푚 = 푣표푙 ∙ 푀푆), and the capture efficiency was calculated using Equation 4-1.

Sample preparation for the VSM measurements consisted of drying a pipetted droplet of known volume (10 to 500 µL) of the sample containing magnetic particles on a silicon substrate (~25 mm2). For example, in the case of the first experiment (0.2-mL samples), 0.1-mL filtrate droplets were dried on the substrate. On the other hand, for the second experiment (50- and 100-mL samples) 50-µL and 300-µL filtrate droplets were used for sample preparation.

Figures 4-18 and 4-19 show example VSM data corresponding to samples before and after filtering with µFMS device for small-volume samples of microdiscs and

IONs, respectively. Additionally, Tables 4-1 and 4-2 summarize the magnetic particle capture efficiency calculations for each of the experiments using the µFMS device. Also, the uncertainty (confidence interval of 95%) results are presented.

Table 4-1. Summary of Capture Efficiency (%) Calculations for Microdiscs Sample Saturation Magnetic Moment (10-7 A·m2) Capture Mean Std. Uncertainty (95% CI; Efficiency (%) Dev. N=56) (95% CI) stock 0.745 0.038 0.199 - filtrate (5 µL/s) 0.041 0.029 0.008 94.5 ± 1.8 filtrate (15 µL/s) 0.050 0.036 0.010 93.3 ± 2.2 filtrate (30 µL/s) 0.040 0.031 0.008 94.6 ± 1.8 filtrate (60 µL/s) 0.037 0.033 0.009 95.0 ± 1.8 filtrate (120 µL/s) 0.034 0.026 0.007 95.4 ± 1.6

89

Figure 4-18. Vibrating sample magnetometer (VSM) data obtained (before and after filtration) from 0.2-mL samples filtered using the microfluidic magnetic separation (µFMS) device at different flow rates : 5 µL/s, 15 µL/s, 30 µL/s, 60 µL/s, and 120 µL/s. Table 4-1 summarizes the saturation magnetic moment for each sample, as well as the capture efficiency (%).

Table 4-2. Summary of Capture Efficiency (%) Calculations for IONs. Sample Saturation Magnetic Moment (10-7 A·m2) Capture Mean Std. Uncertainty (95% CI; Efficiency (%) Dev. N=40) stock 0.972 0.046 0.015 - filtrate (5 µL/s) 0.051 0.041 0.013 94.7 ± 1.3 filtrate (120 µL/s) 0.054 0.034 0.011 94.4 ± 1.1

90

Figure 4-19. VSM data obtained (before and after filtration) from 0.2-mL IONs samples filtered using the microfluidic magnetic separation (µFMS) device at different flow rates: 5 µL/s and 120 µL/s. Table 4-2 summarizes the saturation magnetic moment for each sample, as well as the capture efficiency (%).

4.4.4.4 Magnetic capture efficiency vs. flow rate for µFMS device

As explained in the previous section and as shown in Figures 4-18 – 4-19, capture efficiency data was analyzed for different flow rates. In Figure 4-20, experimental measurements (‘cross’ data points) and simulation results (dashed lines) for various hydrodynamic diameters are presented. Here, simulated and experimental capture efficiency results for the (A) microdiscs and (B) IONs are summarized. From the COMSOL simulations (Figure 4-20A), it was predicted that capture efficiencies of

91

magnetic microdiscs (simulated as spherical particles) with Dhd of 618 nm and 1.5 µm as 100%, while for the smaller diameter (70 nm) capture efficiencies decreased as the flow rates increased. Figure 4-20B shows how capture efficiencies of IONs in the µFMS device decreased with higher flow rates for Dhd smaller than 124 nm, which experimentally was not matched. However, in order to match experiments, capture efficiency simulations were close to 100% for Dhd of 294 nm or higher, which may serve as an indication of particle aggregation during experiments.

Experimental results for microdiscs (Fig. 4-20A), showed capture efficiencies of

94.5 ± 1.8%, 93.3 ± 2.2%, 94.6 ± 1.8%, 95.0 ± 1.8%, and 95.4 ± 1.6% for 5, 15, 30, 60, and 120 µL/s flow rates, respectively. These results closely match the 100% COMSOL simulation results for discs (>618 nm,). Similarly, results for IONs (Fig. 4-20B), showed capture efficiencies of 94.7 ± 1.3% and 94.4 ± 1.1% for 5 and 120 µL/s flow rates, respectively, which closely matches COMSOL simulation results for particles with Dhd

>294 nm. All percentages are represented as estimated mean with 95% confidence interval. Tables 4-1 and 4-2 summarized the estimated means, standard deviations, and

95% confidence interval values for experimental data.

An important consideration from the simulations is that even though the experiments for each of the particles (IONs and microdiscs) were performed by changing the flow rates (Q = 5, 15, 30, 60, and 120 µL/s), when simulating a single channel in 2D, flow rates were modified. Therefore, considering that flow rates set experimentally (e.g. using syringe pump) are defined for a single inlet that then divides into 8 parallel channels (our microfluidic device design), then in the simulations (for a

92

single microfluidic channel) the flow rates were defined as: Qc = Q/8 ≈ 0.63, 1.9, 3.8,

7.5, and 15 µL/s.

Figure 4-20. Capture efficiency study for (A) microdiscs and (B) iron-oxide nanoparticles (experiments and COMSOL simulations).

93

4.5 Filtering MNPs from Large-Volume Samples Using µFMS Device

For this experiment, a 100-mL sample containing ~6 million magnetic microdiscs was filtered at a flow rate of 120 µL/s using the µFMS device and imaged using an optical microscope. Figure 4-21 shows successful confirmation under the microscope of magnetic microdiscs trapped in the µFMS device. It is important to note that the concentration of the discs in solution was very low (0.065 µg/mL) compared to the solutions containing IONs (12.5 – 100 µg/mL) (Figure 4-22). Therefore, capture efficiency estimation (using the VSM) was not possible at this low concentration level.

From the large-volume samples experiments with IONs, the magnetic separation microfluidic device was capable of processing >50-mL samples. A 50-mL sample (IONs concentration of 100 µg/mL) was filtered in ~7 min at a flow rate of 120 µL/s and the estimated capture efficiency was 70.0 ± 2.3% (results in Appendix E: Figures E-1 and

E-2). Similarly, a 100-mL sample (IONs concentration 12.5 µg/mL, 100-mL) was filtered in ~15 min at a flow rate of 120 µL/s and the estimated capture efficiency was 72.2 ±

2.0%. Figure 4-22 show 100-mL sample experiment images of the solutions before filtering (A), after filtering (B), the µFMS device with all captured IONs (C), and the VSM data to estimate capture efficiency for 100-mL sample (D).

VSM data in Figure 4-22D shows the magnetic moment measured for the stock concentration (before µFMS filtration, orange data points) and filtrate (after µFMS filtration, blue data points) for the 100-mL sample. Additionally, the black dashed line represents the mean saturation magnetic moment for each of the samples (1.52 ×10-7

A·m2 for the 300-µL from the stock sample, and 4.2 ×10-8 A·m2 for the 300-µL from the filtrate sample), which resulted in the 72.2 ± 2.0% capture efficiency.

94

Figure 4-21. Example image of magnetic microdiscs captured from a 100-mL sample at 120 µL/s using the µFMS device.

Figure 4-22. Experiment results from 100-mL samples using the µFMS device. (A)-(B) 100-mL sample comparison: before filtration, after filtration, (C) IONs captured in the magnetic separation microfluidic device, and (D) vibrating sample magnetometer data corresponding to 300-µL dried droplet on a substrate from the 100-mL sample filtered using the µFMS device at 120 µL/s. In panel (D), black dashed lines represent the average value from the saturation magnetic moment measured and the gray bands represent ± one standard deviation. Photos courtesy of author.

95

These experimental results demonstrated that the µFMS device is capable of filtering >50-mL samples at flow rates of 120 µL/s. Now, it can be observed that the capture efficiencies measured in these large-volume experiments (~72%) are lower than the ~93% reported for the experiments performed on 0.2-mL. However, it is important to consider that the 0.2-mL sample experiments consisted of a lower concentration (~6.47

µg/mL) of particles (microdiscs). Therefore, it is possible that this decreased capture efficiency was related to clogging of the µFMS device microchannels with such high concentrations of IONs in the 50- and 100-mL samples (100 and 12.5 µg/mL, respectively) (Figure 4-22C). This decreased capture efficiency might be related to the decreased free flowing cross-sectional area in the microchannels for the particles to move after clogging starts. This will cause an increase in the velocities experienced by the particles (inversely proportional to the cross-sectional area of the channel) and consequently increasing the drag forces acting on them, which may surpass the magnetic force. Even though a breakthrough curve to compare the highest number of particles that we could filter and how the capture efficiency would change, the number of particles suggested (~10 million microdiscs or ~10 µg per sample) for the method being followed in this work, saturation of the device is not an anticipated concern.

4.6 Summary

This chapter summarized the magnetic separation setups evaluated for the isolation of magnetic microdiscs. This chapter mostly focused on the isolation of microdiscs by themselves, but preliminary experiments and results for the selectivity of bio-functionalized microdiscs (i.e. biotin, lectins, and aptamers) towards targets (i.e. avidin-coated particles, coliforms, and E. coli) and their magnetic isolation were performed and will be presented in Chapter 5. Additionally, capture efficiency of each of

96

these setups was evaluated in order to choose a method that not only possessed the ability of processing large-volume samples, but also to do it efficiently and in a timely manner. A summary of the magnetic separation setups presented in this chapter were using: 1) a single magnet (for small-volume samples), 2) dual-magnet configurations

(attracting and repelling), 3) alternating magnets, and 4) alternating magnets (with a microfluidic device). Figure 4-23 summarizes the capture efficiency of all the different setups discussed in this chapter. The last magnetic separation method using an alternating array of magnets with alternating polarizations and a microfluidic device was chosen as the most effective platform to capture microdiscs and its conjugates in a localized ready-to-inspect microscope slide.

Figure 4-23. Summarized results on the magnetic re-capture of magnetic particles (i.e. microdiscs or nanoparticles) using the different magnet configurations. Bubble size represent the volume of the sample filtered. More details regarding how this capture efficiencies are obtained can be found in Appendix E.

97

CHAPTER 5 MAGNETIC ISOLATION OF PARTICLE/BACTERIAL TARGETS*

5.1 Overview

In Chapters 2 and 3, a summary of the bio-functionalization concept and process of the magnetic microdiscs to target avidin-coated particles or bacteria were introduced.

In brief, the overall target particle/bacterial assay consists of bio-functionalizing the gold- coated magnetic microdiscs, an incubation period, magnetic separation, and fluorescence labelling (if needed). In Chapter 4, more details regarding the magnetic separation of microdiscs was provided. Several magnetic separation magnet configurations and setups were evaluated to find an optimum setup to ‘filter’ samples of up to 100 mL in a timely manner (<30 min). From these setups, it was found that the

‘bulky’ or macrofluidic setups were capable of filtering large-volume samples but microdiscs could not be retrieved entirely once they were trapped in the channels.

Therefore, a microfluidic magnetic separation (µFMS) device was proposed, evaluated, and used to provide a ready-to-inspect platform, where no retrieval of microdiscs was needed.

Although Chapter 4 presented magnetic separation results of microdiscs by themselves, in Chapter 5 (this chapter), experiments and results pertaining to avidin- coated particles and bacteria will be presented. Here, an introductory experiment to evaluate the use of multiple fluorescence labels will be discussed before describing all the particle/bacterial target magnetic isolation experiments in detail. Primary goals of the

* Portions of this chapter are reprinted with permission from (i) K. Y. Castillo-Torres, D. P. Arnold, and E. S. McLamore, “Rapid isolation of Escherichia coli from water samples using magnetic microdiscs,” Sensors and Actuators: B chemical, 291 (2019), 58-66. Copyright 2019 Elsevier B.V. and (ii) K. Y. Castillo-Torres, E. S. McLamore, and D. P. Arnold, “A High-Throughput Microfluidic Magnetic Separation (µFMS) Platform for Water Quality Monitoring,” Micromachines, vol. 11, no. 1, p. 16, Dec. 2020.

98

experiments to be described in this chapter include: 1) to demonstrate target specificity of bio-functionalized microdisc, 2) to demonstrate the magnetic isolation of targets using bio-functionalized microdiscs, 3) to evaluate the limits of detection using bio- functionalized magnetic microdiscs at water quality monitoring levels, and 4) to explore the enhancement of binding rates (binding events per unit time) between the particle/bacterial targets and bio-functionalized microdiscs via the application of rotational magnetic fields to samples during incubation.

5.2 Evaluation of Multiple Fluorescence Labels for Target Particles/Bacteria

One of the earliest experiments performed was the testing of different fluorescent labels for cell imaging. In this case, green fluorescent protein (GFP), carbon quantum dots (CDs) (obtained from Dr. Eric McLamore’s lab), and SYTO9 green fluorescent nucleic acid stains were evaluated. The primary goal of this experiment was to evaluate and demonstrate the ability to use multiple fluorescent labels and determine which label to use moving forward with the experiments. In the case of GFP-transformed bacteria, cells were obtained from EnCor, Inc (Gainesville, FL). The CDs were fabricated and provided by Dr. McLamore’s lab (Agricultural and Biological Engineering Department,

University of Florida). Finally, SYTO9 was purchased within the LIVE/DEAD BacLight

Bacterial Viability Kit from ThermoFisher Scientific (Waltham, MA).

The experiments consisted of using aptamer-functionalized microdiscs to target and isolate E. coli from water samples (up to 5 mL). Here, microdiscs were incubated with E. coli (concentrations up to 108 CFU/mL) for 15-20 min, magnetically isolated using a single magnet and stained with three different fluorescent techniques, namely:

GFP-transformed cells, CD-labeled viable cells, and SYTO9 stained cells. Figure 5-1 shows representative fluorescent images after E. coli capture using aptamer-

99

functionalized magnetic microdiscs. GFP-transformed E. coli (Fig. 5-1A) were visible along the edges of the discs, and the rod-like structure is apparent with little debris or artifact. CD-labeled E. coli (Fig. 5-1B) were apparent from the rod-like shape in the appropriate size range, with also little debris, which is attributed to the uptake of CD by viable cells and the limited aggregation onto environmental DNA, polymers, or other cell debris. SYTO9/PI-tagged E. coli (Fig. 5-1C) were visible, and there was significant debris and artifact in all samples, which is a common challenge in SYTO9 staining as any extracellular DNA would be labeled and may produce false positives. From these results, CD-labeled E. coli (Fig. 5-1B) was one of the best candidates among the three experiments for use as a rapid biosensor in field testing. However, for the lab-prepared samples to be used in the majority of the benchtop experiments performed in this dissertation, the commercially available SYTO9/PI viability kit was chosen as the primary fluorescent labeling technique. In the future, for field testing, it will be valuable to consider other labels, such as, CDs.

Figure 5-1. Fluorescence tagging of E. coli was demonstrated using three kinds of labels: A) green-fluorescence protein (GFP), B) carbon quantum dots (CDs), and C) SYTO9/propidium iodide (PI). Also, imaging was performed using an epi-fluorescence microscope for sample A, while a confocal microscope was used for samples B and C.

100

5.3 Exploration of Microdisc Bio-Functionalization Target Specificity

A next primary goal was to demonstrate the ability of the microdiscs to selectively target different particles or cells when bio-functionalized with appropriate capture probes. This set of experiments consisted of bio-functionalizing microdiscs with biotin,

ConA lectins, and DNA-aptamer to target avidin-coated particles, coliforms, and E. coli, respectively.

These experiments consisted of introducing bio-functionalized magnetic microdiscs to the sample containing the target particles or bacterial cells for about 30 min. Then, samples were rinsed using water and microdisc-target conjugates were retrieved using a single permanent magnet and a pipette, followed by sample preparation and microscope inspection, as described in Chapter 3.

Figure 5-2 shows results on the isolation of avidin-coated particles using biotin- functionalized magnetic microdiscs. The particle concentration in this sample consisted of 109 particles/mL. Here, a 0.5 mL sample was exposed to ~5x106 microdiscs.

Therefore, as can be observed, most of the microdiscs seem to be bound to target particles (more particles than microdiscs when in suspension). Figure 5-2A-C show the bright field, green channel, and composite images, respectively. The green-fluorescent particles represent the target avidin-coated polystyrene particles.

Now, in Figure 5-3, results on the isolation of E. coli cells using aptamer- functionalized magnetic microdiscs is presented. Similar to avidin-coated particles experiment, the bio-functionalized magnetic microdiscs (~5 5x106 microdiscs) were exposed to a 0.5-mL solution with a concentration of 10-100 CFU/mL (Microbiologics,

Inc). Again, Figure 5-3A-C show the bright field, green channel, and composite images, respectively, where the green-fluorescent particles represent the target E. coli cells.

101

Figure 5-2. Magnetic isolation of avidin-coated particles was demonstrated using biotin- functionalized magnetic microdiscs. A) bright-field microscope image, B) green-channel fluorescence image, and C) composite image. Here, avidin- coated polystyrene particles were already labeled with green fluorescence, discs are the dark dots in A).

Figure 5-3. Magnetic isolation of E. coli cells was demonstrated using aptamer- functionalized magnetic microdiscs. A) bright-field microscope image, B) green-channel fluorescence image, and C) composite image. Here, E. coli cells were labeled using SYTO9, discs are the dark dots in A).

102

Another proof-of-concept experiment was performed to demonstrate the specificity of aptamer-functionalized microdiscs compared to lectin-functionalized microdiscs. Here, water samples containing general coliforms (E. coli, K. variicola and

P. mirabilis) were prepared and exposed to 1) lectin-functionalized microdiscs and 2) aptamer-functionalized microdiscs. Results in Figures 5-4A – B showed more fluorescence in the samples exposed to lectin-functionalized microdiscs than in the sample exposed to aptamer-functionalized microdiscs. These results align with the idea that lectin-functionalized microdiscs should bind to many types of general coliforms (E. coli, K. variicola and P. mirabilis) while aptamer-functionalized microdiscs should show more specificity towards E. coli only.

Similarly, coliform samples (E. coli, K. variicola and P. mirabilis) were prepared in a more complex environment (i.e. vegetable broth, instead of DI water), as a preliminary evaluation of selectively targeting bacteria in the presence of organic interferents.

Results regarding bacteria isolation from vegetable broth samples using lectin- and aptamer-functionalized microdiscs are presented in Figures 5-5 A – B. From results in

Figure 5-5A, more fluorescence was observed in samples that were exposed to lectin- functionalized microdiscs when compared to samples exposed to aptamer- functionalized microdiscs. Although, more detailed experiments can be conducted regarding non-specific binding, the promising results obtained from these experiments are of value for when the biosensing system is intended to be used to analyze field samples (i.e. recreational waters from beaches or lakes), which can be considered a more complex environmental sample (containing other unknown particles and/or cells).

103

Figure 5-4. Proof-of-concept results demonstrating selectivity of bio-functionalized microdiscs in lab-prepared water samples. A) lectin-functionalized microdiscs towards coliforms (E. coli, K. variicola and P. mirabilis) and B) aptamer- functionalized discs towards E. coli cells in DI water samples.

104

Figure 5-5. Proof-of-concept results demonstrating selectivity of bio-functionalized microdiscs in a complex environmental sample (i.e. vegetable broth), (valuable for field testing applications). A) Con A-functionalized microdiscs towards coliforms (E. coli, K. variicola and P. mirabilis) and B) aptamer- functionalized discs towards E. coli cells.

105

Next, a set of positive and negative control experiments were performed. Here, positive control samples were collected from a local municipal wastewater treatment plant in Gainesville, FL (UF Water Reclamation Facility). These samples were collected from the effluent wier of the secondary clarifier and analyzed on the same day of collection. From this sample collected, half was treated with ethanol, to induce cell death as was described in Chapter 3. Then, both samples (0.5-mL each) were exposed to Con A lectin-functionalized discs for about 20 min, rinsed, and stained using

SYTO9/PI. Figure 5-6A was a sample analyzed as it was when collected, while Figure

5-6B shows results for ethanol-treated sample. Since SYTO9 targets intact or healthy cell membranes, while PI targets damaged membranes in a cell as explained in Chapter

2, more red fluorescence was expected from the ethanol-treated sample. From Figure

5-6B, confirmation of more PI presence (red fluorescence) in the ethanol-treated sample can be observed. Therefore, results provided promising results on the discrimination of viable from non-viable bacteria bound to magnetic microdiscs

Figure 5-6. Bio-flocs containing coliforms isolated using Con A-functionalized magnetic microdiscs. This sample was obtained from the UF wastewater reclamation facility and contained an estimated bacteria concentration of 4,000 CFU/mL.

106

In terms of negative controls, where no cells were added to the samples, multiple cases of water samples containing microdiscs were tested to evaluate possible non- specific binding of the fluorescence labels (i.e. SYTO9). The goal here was to determine if there was non-specific binding towards other particles present in the water samples or towards the capture probes attached to the bio-functionalized microdiscs (i.e. lectins or aptamers). Therefore, not-functionalized and bio-functionalized microdiscs in water were stained with SYTO9 and rinsed using a permanent magnet to decant supernatant.

Also, not functionalized microdiscs, not exposed to SYTO9 were evaluated as an additional control. As with previous positive control sample, sample preparation consisted of using a single magnet to localize microdiscs and retrieved using a pipette and deposited on a microscope glass slide for imaging inspection.

All results from control experiments are summarized in Figure 5-7. Here, row (A) shows results on the not functionalized microdiscs that were not exposed to fluorescent labels (SYTO9). Then, rows (B)-(D) show results from samples with not-functionalized, lectin-functionalized, and aptamer-functionalized microdiscs that were exposed to green fluorescent labels (SYTO9), respectively. In general, it was observed that the presence of fluorescence was significantly smaller than when bacterial targets were present in the samples. However, there seemed to be some particle contamination or autofluorescence that was captured as fluorescence in some of these controls, which as was already mentioned in previous sections, is a challenge when working with SYTO9.

Therefore, for lab-prepared samples mostly SYTO9 and PI were used, but for future experiments, a more detailed evaluation of fluorescent labels can be considered (e.g. carbon quantum dots, CDs).

107

Figure 5-7. Control samples/ Four sets of controls were evaluated: 1) aptamer- functionalized microdiscs + SYTO9/PI, 2) Con-A-functionalized microdiscs + SYTO9/PI, 3) not-functionalized microdiscs + SYTO9/PI, and 4) not- functionalized microdiscs with no stains.

5.4 Demonstration of Magnetic Isolation of Target Particles/Bacterial Cells

In this section more results will be presented as a demonstration of the use of the different macrofluidic and microfluidic magnetic separation setups presented in Chapter

4. All results presented in this chapter show isolation of particle/bacterial targets from

108

water samples, whereas the results in Chapter 4 were only focused on capture of the microdiscs (without cells). Additionally, larger sample volumes are studied, ranging from

0.5 mL to 100 mL. Even though capture efficiencies of some of the setups were ≤60%, a few proof-of-concept experiments were performed with bio-functionalized microdiscs and target particles and/or bacterial cells in up to 100-mL samples. Particle/bacteria concentrations in the samples evaluated as proof-of-concept ranged in 102 – 109 particles (or CFU) per 100 mL. In this section, experiments on the magnetic isolation of targets and their respective results will be presented starting with samples evaluated with higher concentrations of targets. In the subsequent section, experiments with lower bacterial concentrations—which are more representative of water quality monitoring levels of detection—will be introduced.

5.4.1 Magnetic Isolation of Avidin-Coated Particles Using the µFMS Device

Targeting and isolating of target particles or bacteria was assessed following the same protocol that has been described in Chapter 3. For example, as a proof-of- concept experiment, targeting and isolation of avidin-coated particles was assessed by adding (~6 million) biotin-functionalized microdiscs to a 100-mL particle solution for 30 min at room temperature (protecting from light) and applying occasional stirring. Here, two 100-mL samples with different particle concentrations were prepared, one sample contained 3,000 million particles/mL and the second sample contained 3,000 particles/mL. Then, samples were filtered using the magnetic separation setup (in this case, the µFMS device at a flow rate of 120 µL/s) in less than 15 min. Finally, magnet array was removed from the µFMS device and sample was inspected under the fluorescence microscope (Dr. Hugh Fan’s lab).

109

Considering that the avidin-coated polystyrene particles were pre-labeled with green fluorescence, more fluorescence was expected from samples exposed to higher concentrations of particles. Results in Figure 5-8 confirmed magnetic isolation of target particles and also a higher fluorescence in the sample with (A-B) higher concentration of particles (3,000 million particles/mL), when compared to sample with (C-D) lower concentration of particles (3,000 particles/mL) can be observed. These results using biotin/avidin in 100-mL samples provided a valuable insight regarding the use of µFMS device for the biosensing application of interest: bacteria detection for water quality monitoring, which led to the preliminary experiment to be described next.

Figure 5-8. Isolation of fluorescent avidin-coated particles from 100-mL samples using biotin-functionalized magnetic microdiscs. (A-B) 100-mL sample containing 3,000 million particles and (C-D) 100-mL sample containing 3,000 particles.

110

5.4.2 Magnetic Isolation of E. coli Using the µFMS Device

As a preliminary experiment, the same procedure was followed to isolate

SYTO9-labeled E. coli cells at a concentration of 100 CFU/mL from a 0.5-mL sample using aptamer-functionalized microdiscs and the µFMS device at a flow rate of 120

µL/s. Here, the magnetic microdiscs were exposed to the 0.5-mL E. coli sample for about 30 mins, rinsed, and fluorescently labeled with SYTO9. After fluorescence staining, the sample was filtered using the µFMS device, which took less than 5 seconds to filter (at a flow rate of 120 µL/s). Lastly, magnet array was removed from the

µFMS device and inspection was performed using a fluorescence microscope.

Figures 5-9 – 5-8 shows confirmation results, via fluorescent microscopy, that isolation of microdisc/bacteria conjugates was achieved in the µFMS device. These results showed promise towards the scalability of this technique for rapid separation of

100-mL water samples in as fast as ~15 minutes when a flow rate of 120 µL/s is used.

Also, these results demonstrated that bio-functionalized discs (i.e. aptamers) can be used to target E. coli, but in the future, fluorescence label step can be done after magnetic isolation by flowing diluted SYTO9 for 10 mins through the µFMS device, followed by fluorescence imaging for detection. This process would result in a detection timeframe of less than 2 hours, which is still inside the timeframe of interest for water quality monitoring applications. Typical bacteria-isolation methods (including immunomagnetic separation) can generally process volumes of up to 100 mL within timeframes up to 24 hours and require complex lab equipment. More recent magnetic isolation methods using microfluidic devices have demonstrated capture efficiencies of

~94% but flow rates implemented are typically less than 15 µL/min which would result in

100+ hours for the magnetic separation of 100-mL samples [92].

111

Figure 5-9. Example image of aptamer-functionalized microdiscs and bacteria conjugates isolated using µFMS device and a flow rate of 120 µL/s.

Figure 5-10. More example images of aptamer-functionalized microdiscs and bacteria conjugates isolated using µFMS device and a flow rate of 120 µL/s.

112

5.4.3 Magnetic Isolation of Coliforms / E. coli Using Macrofluidic Setups

Next, experiments using the macrofluidic magnetic separation setups (i.e. dual magnet or magnet-column array) which used easy-to-access and disposable materials

(i.e. pipettes and IV lines) were performed to evaluate if isolation of cells was possible.

Even with lower capture efficiencies (compared to µFMS), easy access to materials and assembly process of the setup was attractive for proof-of-concepts experiments.

Experimentally, the protocol described in Chapter 3 was followed. To recap briefly, 100-mL water samples containing coliforms of E. coli cells were exposed to 5-10 million bio-functionalized microdiscs. Again, bio-functionalization was determined depending on the bacterial target to be used (i.e. lectins for coliforms and aptamers for

E. coli). General coliform samples containing E. coli, K. variicola and/or P. mirabilis were prepared with concentrations of 2,000-10,000 CFU/ 100 mL. While in the case of E. coli samples, similar protocol was followed and concentrations of 2,000-20,000 CFU/100mL were obtained. In both cases, half the samples were treated with ethanol for viability tests. Once the samples were prepared, bio-functionalized microdiscs were exposed to the samples for about 30 mins and then filtered using the magnetic separation setup at a flow rate of ~83 µL/s, which was adjusted via the clamp of the IV line. After magnetic separation, the microdiscs and conjugates (i.e. coliforms and/or E. coli) were retrieved and resuspended in a smaller volume (~1 mL) and fluorescently labeled using SYTO9 and PI. Finally, after fluorescence labeling, samples were prepared using a pipette and depositing the conjugated microdiscs on a glass slide for inspection in the fluorescence microscope.

In Figures 5-11 and 5-12, magnetic isolation of bacteria using these setups was apparent. Specifically, in Figure 5-11, lectin-functionalized microdiscs bound to

113

viable/non-viable coliforms (i.e. E. coli, K. variicola and/or P. mirabilis) were observed.

Here, green fluorescence corresponds to viable coliforms and red fluorescence corresponds to non-viable (or membrane-damaged) coliforms. Similarly, in Figure 5-12, aptamer-functionalized discs can be observed bound to viable/non-viable E. coli. These results demonstrated successful discrimination of viable and non-viable bacteria from the magnetically isolated of microdisc-conjugates (i.e. viable/non-viable coliforms) at concentrations of 1,000-20,000 CFU/100 mL. These successful confirmation results led to experiments at even lower concentrations, as will be described in the next section.

Figure 5-11. Example images of lectin-functionalized microdiscs and coliform conjugates isolated dual-magnet setup with repelling configuration of magnets at a flow rate of ~83 µL/s.

114

Figure 5-12. Example images of aptamer-functionalized microdiscs and E. coli conjugates isolated dual-magnet setup with repelling configuration of magnets at a flow rate of ~83 µL/s.

5.5 Evaluation of LODs for Water Quality Monitoring Applications

As demonstrated from previous section, the magnetic isolation of target particles/bacterial cells was achieved at detection levels of down to 1,000 CFU/100 mL, even when using macrofluidic setups (i.e. dual magnets with repelling configuration and column-array of magnets), smaller concentrations were of interest for water quality monitoring applications. Therefore, this section will focus on experiments performed in large volume samples and at concentrations ranging 1-126 CFU/100 mL, which are between the detection levels of interest for both drinking and recreational water quality

115

monitoring, as suggested by the U.S. EPA [5], [7].

For these experiments, EZ Hydro Shot coliform and EZ CFU E. coli kits from

Microbiologics, Inc. were also used for the preparation of samples. From each of the protocols, concentrations expected were 200-1,000 CFU/100 mL for coliforms and 100-

1,000 CFU/100mL for E. coli. In order to obtain lower concentrations, of interest for water quality monitoring applications, dilutions were needed followed by an enrichment step. After sample dilutions (i.e. 2-10 CFU/100 mL for coliform samples and 1-10

CFU/100 mL for E. coli samples), the enrichment step consisted of incubating all samples in Tryptic Soy Broth (TSB) at 37°C for 6 hours prior to analysis. After the 6- hour incubation, concentrations were expected to be close to 128-640 CFU/100 mL in coliform samples and 64-640 CFU/100 mL in E. coli samples [93]. At this point, samples were ready to be analyzed, hence, were exposed to bio-functionalized magnetic microdiscs for about 30 mins. Then, magnetic separation was performed in each sample also using the magnetic separation setup at a flow rate of ~83 µL/s, which was adjusted via the clamp of the IV line. After magnetic separation, the microdiscs and conjugates (i.e. coliforms and/or E. coli) were retrieved and resuspended in a smaller volume (~1 mL) and fluorescently labeled using SYTO9 and PI. After fluorescently labeled, sample preparation was performed using a pipette and deposited on a glass slide for fluorescence microscopy inspection.

It is important to note that these experiments did not study the viability discrimination at this lower detection limits, hence no additional step to damage cell membranes were added, as was done in the previous experiment. Therefore, only green fluorescence was expected to be expressed in the samples.

116

Now, Figure 5-13 shows ConA lectin-functionalized microdiscs bound to viable coliforms (green fluorescence corresponding to viable E. coli, K. variicola and/or P. mirabilis) in the sample with a starting concentration of 2-10 CFU/100 mL (before enrichment) and 128 – 640 CFU/100 mL after 6-hr TSB enrichment. Similarly, in Figure

5-14, it can be observed that aptamer-functionalized discs were bound to viable E. coli

(green-fluorescence), demonstrating successful magnetic isolating of microdisc- conjugates (i.e. E. coli) using another capture probe (i.e. aptamers). These results demonstrated successful magnetic isolation of microdisc-conjugates (i.e. viable coliforms and/or E. coli). Further studies were performed to compare results from different concentrations of samples (from 0 to 104 CFU/100 mL), experiment which will be described next.

Figure 5-13. Confocal microscopy images showing isolation of various target coliform cells using lectin-functionalized magnetic microdiscs.

117

Figure 5-14. Confocal microscopy images showing isolation of various target E. coli cells using aptamer-functionalized magnetic microdiscs.

A final experiment to evaluate the limits of detection of the biosensing system that has been described in this dissertation, multiple samples with different concentrations were evaluated. Here, a starting concentration of E. coli samples was

104 CFU/100 mL, which was prepared similar to the previous experiments, using the same EZ CFU E. coli kit from Microbiologics, Inc. This sample was diluted to 102

CFU/100 mL and 1 CFU/100 mL. Also, a control sample, which contained no cells was tested. These 100-mL samples were also filtered using macrofluidic devices (i.e. dual magnet with repelling configuration), retrieved and resuspended in smaller volumes (1 mL) followed by fluorescence staining using SYTO9. Then, sample preparation was done by using permanent magnet to localize all microdiscs, which were then retrieves using a pipette and deposited on a glass slide for fluorescence inspection and imaging. in fluorescence

118

Figure 5-15. Example (bright and fluorescent fields) images of samples containing aptamer-functionalized discs exposed to E. coli with concentration of A) 0 CFU/100 mL (no cells, control), B) 100 CFU/100 mL, C) 102 CFU/100 mL, and D) 104 CFU/100 mL. All samples were tagged using SYTO9/PI. Adapted from [57] and reprinted with permission from 2019 Elsevier B.V.

119

From results in Figure 5-15, it is observed that as the concentration of bacteria decreased, less fluorescence intensity was observed in the images. Although the work being presented here is focusing of the use of this technology only for presence detection (qualitative), this might serve as an indication that bacteria quantification/estimation might be possible considering the amount of fluorescence observed from samples. Also, it is important to note that these results were obtained using a magnetic separation setup that was evaluated for capture efficiency of magnetic microdiscs by themselves, and results showed up to 60% capture efficiency. However, promising results were still obtained in low concentration samples of bacteria, specifically at levels of interest for water quality monitoring.

With all the results obtained in mind, the biosensing system being presented here should successfully isolate bacterial cells from large-volume samples using most of the magnetic separation setups described in Chapter 4. However, the preferred method, because of the high capture efficiency and the convenient ready-to-inspect characteristic is the µFMS device. Preliminary results with the µFMS device have already demonstrated the feasibility of its use to process samples of 100-mL in less than 15 minutes and to magnetically isolate target particles or bacterial cells. Therefore, it can be concluded that the setup should be capable of replicating all these results on lower concentrations of target bacterial cells.

5.6 Exploration of Microdisc Rotation to Enhance Target Binding Rate

Previous experiments described in this chapter have focused on the demonstration of targeting multiple particle or bacterial targets by changing bio- functionalization of the microdiscs, magnetically isolating targets (particles or bacteria) using multiple magnetic separation setups, as well as evaluating the limits of detection

120

of the biosensing system as a whole. Now, in this section more focus will be given towards the enhancement (or increase) of binding events with targets when bio- functionalized microdiscs are incubated in samples containing the targets (i.e. particles or bacterial cells). Here, enhancement of binding rates between bio-functionalized microdiscs and targets was proposed by applying a rotating magnetic field to the samples with the microdiscs in suspension, which would cause a torque and hence rotation of the free-floating magnetic microdiscs potentially causing a ‘micromixing’ effect in the surrounding environment (i.e. sample with particle/bacterial targets).

Microstirring or micromixing has been predominately studied using microfluidic devices, and some of these have been reviewed in [94]. Previous works have shown successful mixing in microfluidic channels with interest in biological and lab-on-a-chip applications. Some examples of developed methods of magnetic micromixing, include the work by Lu, et al. (2002), where a magnetic bar micromixer for microchannel applications was developed [95]. Also, works by Rida, et al. (2003) showed the ability to used self-assembled structures using external alternating magnetic fields and magnetic microbeads [96]. Other applications in which magnetic bead-based micromixers have been developed or designed include cell sorting and particle capture [97]–[99].

5.6.1 Magnetic Rotation (or Stirring) Experimental Setup

As has been previously published from Dr. Arnold’s group, specifically by N.

Garraud, an experimental setup combining a magnetic actuation with an optical detection has been used to actuate these magnetic microdiscs [35]. As a summary, two orthogonal magnetic Helmholtz coils creating a homogeneous rotating magnetic field to control the microdiscs rotation were used (Figure 5-16). For the optical interrogation, a laser and a photodiode were used to monitor the microdisc angular rotation via

121

transmitted light intensity modulation (Figures 5-16 and 5-17). Here, the relative light variation ∆I and the phase shift φ were extracted by post-processing the light modulation. As demonstrated in [35], monitoring the phase shift was a key aspect to monitor the microdisc surrounding environment while being independent of light interferences and microdisc concentration. Additionally, this apparatus and overall approach was used to successfully characterize fluids with different viscosities [35].

Figure 5-16. Schematic (top) and image (bottom) of proposed pathogen detector apparatus incorporating a magnetic actuation and an optical measurement by transmission. Diagram/photo courtesy of Nicolas Garraud. Both reproduced from [35] with permission of AIP Publishing.

122

Figure 5-17. Magnetic microdiscs acting as an "optical shutter” when actuated by a rotating magnetic field. Diagram/photos courtesy of Nicolas Garraud. Reproduced from [35] with permission of AIP Publishing.

Another application that was considered using this apparatus was detection of bacteria in water samples, but it was found that for such low concentrations of bacteria in water quality monitoring applications, the optical detection approach was not sensitive enough [36], [100]. Hence, it is hypothesized that applying a rotating magnetic field to samples containing the microdiscs and bacterial targets in water samples could enhance capture rates, which will be described in this section. The apparatus, as is, holds up to 2-mL vials. However, for future and more in-depth experiments, the apparatus could be modified for larger volume samples, if the enhancement of binding rates is concluded. In preliminary experiments to enhance binding rates of targets via the rotation of the free-floating microdiscs, samples of 0.5 mL were studied. Lab- prepared samples tested consisted of avidin-coated particles as proof-of-concept, and

E. coli cells.

123

Some modifications to the apparatus described in [35] were implemented before the experiments to be described here. These modifications include: the use of a different 1) data acquisition system (e.g. Analog Discovery 2, Digilent Inc., Pullman,

WA) and 2) power amplifier (e.g. Crown XLS2500 high-density power amplifier, Elkhart,

IN). Another primary modification of the setup, compared to the one presented in [35], was the elimination of the optical interrogation part (laser and photodiode). For the preliminary experiments to be performed here, only the rotation of the microdiscs was of interest. Therefore, only the 2 pairs of Helmholtz coils were used to create the rotating magnetic field.

5.6.2 Magnetic Stirring Preliminary Experiments

5.6.2.1 Rotation of microdiscs in samples with E. coli cells

Preliminary experiments consisted of exposing 0.5-mL samples to an external rotating magnetic field, using the slightly modified apparatus mentioned before and used in [35]. This time, the apparatus was set to produce a rotating magnetic field of ~1 mT at a frequency of 5 Hz for 10 minutes. Experimental bacterial samples were prepared using EZ CFU E. coli kit (10-100 CFU/mL) enriched for 20-23 hours, with expected concentrations ranging 10 x106 – 8.4x108 CFU/mL. However, after plating the samples for confirmation (results in Appendix F), concentrations ranged 2.3x107 – 1.2x109

CFU/mL (as detailed in Tables 5-1 – 5-3). These experimental samples were exposed to bio-functionalized microdiscs and then rotated using the parameters mentioned above (~1 mT, 5 Hz, 10 min). Duplicates of these samples were conserved as controls, which were not exposed to a rotating magnetic field. Also, other controls studied were samples with microdiscs (not bio-functionalized) but rotated using the apparatus to

124

determine how the functionalization step also enhances target particle/bacteria binding rates. Table 5-1 summarizes the samples studied in this experiment.

Table 5-1. Summary of (0.5 mL) samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using E. coli cells as target bacteria and aptamers as capture probes for bio- functionalization). Target Bio- Applied Time under Sample Particle/Bacteria functionalized Rotating Applied Field Concentration^ Microdiscs Magnetic Field* (min) Experimental A 1.2x109 CFU/mL YES YES 10 Control A 1.2x109 CFU/mL YES NO - Experimental B 2.3x107 CFU/mL YES YES 10 Control B 2.3x107 CFU/mL YES NO - ^23-hr TSB enrichment (plating results); *applied field: ~1 mT, 5 Hz, 10 mins

As can be seen in Figure 5-18, control and experimental preliminary results are presented for the two different concentrations tested. Here, it can be noted that the (A-

B) control samples (not exposed to the rotating magnetic field) show less fluorescence compared (C-D) experimental results. In order to quantify fluorescence in each of these images to determine if in fact there was more fluorescence in experimental samples when compared to controls, ImageJ was used for post-processing. Briefly, quantification in ImageJ was performed by counting the dark pixels (corresponding to the magnetic microdiscs) and the bright green pixels (corresponding to SYTO9-labeled cells).

Additionally, size of the bright green clusters was taken into account considering that they could be non-specific binding of SYTO9 on other particles in solution as was tested. For the pixel counting, “color threshold” option in ImageJ was used to define green and dark pixels, while the “analyze particles” option was used to define the maximum size of interconnected pixels, which was defined as 50. Example images after post-processing in ImageJ are presented in Appendix G.

125

Figure 5-18. Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). Refer to Table 5-1 for summary of samples. Scale bar is 20 µm.

Preliminary results on the quantification of green fluorescence and presence of microdiscs in each sample are presented in Figure 5-19. Here, it can be observed that control samples seemed to have less fluorescence when comparted to the corresponding experimental samples. Also, it can be noted that the sample with higher concentration of cells had more fluorescence when compared to the sample with lower concentration of cells. It is important to mention that the number of green pixels

(proportional to the number of cells) were normalized with respect to the number of dark pixels present in the image (proportional to the number of microdiscs).

126

Figure 5-19. Quantified results using ImageJ (green pixels/dark pixels). Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). Refer to Table 5-1 for summary of samples.

With these promising results in mind, a second experiment with a fixed concentration of bacterial cells but now being exposed to the rotating magnetic field for three different periods of time were tested, as summarized in Table 5-2. Here, controls were also studied, where no rotating magnetic field was applied to the samples. After the experiments, images (Figure 5-20) were also analyzed using ImageJ (Figure 5-21) but this time, a secondary confirmation test was performed: culture colony counting of the decanted supernatant of each sample. Hence, results from the culture colony counting (Figures 5-22 and 5-23) would correspond to bacterial cells not captured by the bio-functionalized microdiscs. Culture colony counting was performed by Adam

Grossman from the Microbiology department.

127

Table 5-2. Summary of samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using E. coli cells as target bacteria and lectins as capture probes for bio-functionalization). Target Bio- Applied Time under Sample Particle/Bacteria functionalized Rotating Applied Field Concentration^ Microdiscs Magnetic Field* (min) Control C 18x108 CFU/mL NO NO - Control D 18x108 CFU/mL NO YES 10 Control E 18x108 CFU/mL YES NO - Experimental C 18x108 CFU/mL YES YES 5 Experimental D 18x108 CFU/mL YES YES 10 Experimental E 18x108 CFU/mL YES YES 20 ^20-hr TSB enrichment (plating results); *applied field: ~1 mT, 5 Hz

Results obtained from ImageJ, which quantified the green (fluorescence) pixels from the cells and dark pixels from the magnetic microdiscs, were inconclusive. Here, no obvious trends were observed that might imply the enhancement of binding rates

(events per unit of time) with bacterial targets when a rotating magnetic field was applied.

Now, from the culture colony counting plating results, it was observed that the bacteria concentration in the samples were larger than expected (18x108 CFU/mL), after

20-hour TSB enrichment. The expected concentration of bacteria was close to 10x106 –

10x107 CFU/mL. Therefore, comparing the number of microdiscs in the sample (~10 million discs/mL) versus the number of bacteria in the samples (up to 9x108 CFU, in 0.5- mL samples), it can be concluded that there were not enough microdiscs to capture all

(or most) bacterial cells. As a final preliminary/proof-of-concept study, an experiment was performed using avidin-coated particles as targets and biotin-functionalized microdiscs which were exposed to a rotating magnetic field. This last experiment will be described in the next subsection.

128

Figure 5-20. Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). Refer to Table 5-2 for summary of samples. Scale bar is 20 µm.

129

Figure 5-21. Quantified results using ImageJ (green pixels/dark pixels). Experimental and control results on the magnetic isolation of E. coli after rotation (if applicable). Refer to Table 5-2 for summary of samples.

Figure 5-22. Quantified results using culture colony counting (plating) of the filtrate (not captured cells). Refer to Table 5-2 for summary of samples.

130

Figure 5-23. Plates for culture colony counting for each of the samples. Colonies counted correspond to cells not captured during the experiments (decanted supernatant). A) Initial solution of cells, B) not functionalized or rotated microdiscs, C) not functionalized microdiscs rotated for 10 min , D) bio- functionalized microdiscs not rotated, E) bio-functionalized microdiscs rotated for 5 min, and F) bio-functionalized microdiscs rotated for 10 min, and G) bio- functionalized microdiscs rotated for 20 min. Refer to Table 5-2 for summary of samples. Photos courtesy of Adam B. Grossman.

131

5.6.2.2 Rotation of microdiscs in samples with avidin-coated particles

This final experiment followed the same procedure as the last two experiments, where the only difference was that avidin-coated particles were used as targets (instead of E. coli cells). The samples were prepared using the same avidin-coated particles from Spherotech, Inc, as in earlier experiments. Here, the concentrations of samples were 18x106 particles/mL. Experimental and control samples used in this experiment are summarized in Table 5-3. In this set of experiments samples were exposed to ~6 million microdiscs, exposed to the rotating magnetic field (if applicable) for 10 minutes, and the microdiscs were retrieved using a single magnet and a pipette. Followed by sample preparation that consisted of depositing the microdiscs and conjugates (i.e. particles) on a microscope slide for inspection.

Table 5-3. Summary of samples studied in the rotation of magnetic microdiscs as a possible enhancement step for water quality monitoring (using avidin-coated polystyrene particles as target particles and biotin as capture probes for bio- functionalization). Target Bio- Applied Time under Sample Particle/Bacteria functionalized Rotating Applied Field Concentration Microdiscs Magnetic Field* (min) Control F.1 18x106 part./mL NO NO - Control F.2 18x106 part./mL NO YES 10 Control F.3 18x106 part./mL YES NO - Experimental F 18x106 part./mL YES YES 10 *applied field: ~1 mT, 5 Hz, 10 mins

Once the fluorescence images were obtained, the images were also analyzed with ImageJ to count the number of pixels corresponding to particles (also bright green) and those corresponding to the microdiscs (dark pixels). In Figure 5-24 and 5-25, results show higher fluorescence in images corresponding to the experimental samples

(exposed to rotated magnetic field) when compared to the controls. Also, higher

132

fluorescence is observed in samples that had bio-functionalized microdiscs, which could be an indication of higher capture due to bio-functionalization and rotation of the microdiscs when in suspension.

Figure 5-24. Experimental and control results on the magnetic isolation of avidin-coated particles after rotation (if applicable). Refer to Table 5-3 for summary of samples.

These preliminary results provide promising proof-of-concept results regarding the enhancement of binding rates with target particles and/or bacteria when exposed to

133

a rotating magnetic field, or ‘micromixing.’ However, further evaluation is needed, specifically quantifying the number of particles or cells not being captured instead of only relying on image post-processing. For example, a possible quantification experiment would involve the quantification of particles using a flow cytometer and compare with the image post-processing results.

Figure 5-25. Experimental and control results from ImageJ on the magnetic isolation of avidin-coated particles after rotation (if applicable). Refer to Table 5-3 for summary of samples.

5.7 Summary

This chapter focused on the experimental demonstration of the biosensing system described in this dissertation. From bio-functionalization to magnetic isolation of bacterial cells, experiments were described here. First, an introductory experiment to

134

evaluate the use of multiple fluorescence labels was discussed, in which fluorescence labelling was performed using GFP, CDs, and SYTO9/PI. Then, the primary goals of the experiments were described in the following order: 1) to demonstrate target specificity of bio-functionalized microdisc, 2) to demonstrate the magnetic isolation of targets using bio-functionalized microdiscs, 3) to evaluate the limits of detection using bio- functionalized magnetic microdiscs at water quality monitoring levels, and 4) to explore enhancement of binding rates between target particles/bacteria and bio-functionalized microdiscs via the application of rotational magnetic fields to the samples during incubation. Each of these goals were demonstrated with different experiments. More specifically, the demonstration of target specificity was successfully demonstrated using biotin-, lectin-, aptamer-functionalized microdiscs towards different target particles/cells.

Then, the magnetic separation of microdiscs that was described in Chapter 4 was applied to microdisc-conjugates (i.e. avidin-coated particles or bacteria), where the magnetic isolation of the microdisc-conjugates was demonstrated. Additionally, the magnetic isolation was pushed to the limits of detection of interest for water quality monitoring applications, where results showed the magnetic isolation of bacteria from

100-mL samples at levels as low as 1 CFU/100 mL with a 6-hr enrichment step. It is important to understand that this enrichment step might be considerable reduced if higher concentrations of bacteria are being evaluated. For example, for drinking water samples, it is necessary to reach such low limits of detection, however, for recreational waters, limits of detection can be closer to 100 CFU/100 mL. Finally, the exploration of applying a rotating magnetic field to the samples during incubation for the enhancement of binding rates by ‘micromixing’ samples with microdiscs and target particles/cells was

135

described. Here, promising results were obtained, in which higher fluorescence

(quantified using ImageJ) was observed in samples in which microdiscs were exposed to rotating magnetic fields. However, further experiments are encouraged in the future in regard to the time of samples exposure to rotating magnetic fields, frequency of rotation, and sample sizes closer to 100 mL.

136

CHAPTER 6 CONCLUSIONS

The work presented here represents a promising step towards a rapid bacteria detection biosensing system for water quality monitoring applications. This work proposed and successfully demonstrated the use of bacteria-sized magnetic microdiscs to target fecal indicating bacterial targets (i.e. general coliforms and/or E. coli). Here, the magnetic isolation of fecal indicating bacteria was demonstrated using bio- functionalized microdiscs in levels of detection of interest for: 1) recreational water quality monitoring (126 CFU/100 mL) in less than 2 hours and 2) drinking water quality monitoring (1 CFU/100 mL) in less than 8 hours (adding a 6-hr enrichment step). These analysis times include filtering of 100-mL water samples using the proposed microfluidic magnetic separation device in less than 15 minutes. Therefore, this method represents a promising tool when compared to other standard methods for water quality monitoring applications, which typically require 12-24 hr of analysis.

6.1 Research Contributions

Primary contributions of this work include the fabrication of ~600 million magnetic microdiscs on each wafer at a material cost of ~$100/wafer, translating to ~$0.17/million microdiscs. Based on the proposed detection method, an estimate of 1–100 million microdiscs will be needed for up to 100-mL samples, equating to a cost ranging from

$0.17 to $17 per test (at current research-scale manufacturing).

Other contributions of this work include the successful demonstration of bio- functionalization of magnetic microdiscs to target specific particles or cells. Also, their magnetic isolation in less than 20 min was successfully demonstrated using multiple magnetic separation setups with capture efficiencies of up to 95%. Additionally,

137

fluorescence labeling was demonstrated using multiple fluorescence stains (i.e. green fluorescence protein, carbon quantum dots, and SYTO9/propidium iodide). Results on presence detection (qualitative) of 1) general coliforms using lectin-functionalized microdiscs and 2) E. coli using aptamer-functionalized discs in samples was demonstrated in samples up to 100 mL with bacteria concentrations as low as 102

CFU/100 mL in less than 2 hr. Also, successful demonstration bacteria detection not only performed in lab-prepared water samples but also, in more complex environmental samples (i.e. vegetable broth and wastewater samples). Table 6-1 summarizes the key contributions of the work done.

Table 6-1. Key contributions Contributions Significance/Benefit Technology demonstration for the low- Batch fabrication of magnetic microdiscs cost fabrication of the magnetic (~600 million microdiscs per substrate at microdiscs (~$0.17-$17.00 per sample) current research-scale manufacturing). at a research-level production. Demonstration of detection and viability Technology demonstration for rapid of E. coli down to 1-10 CFU/100 mL detection and viability of E. coli. within 8 hr test time Demonstration of detection and viability Technology demonstration for rapid of coliforms down to 2-10 CFU/100 mL detection and viability of coliforms. within 8 hr test time Demonstration of the feasibility of using Carbon dots provide improved shelf-life carbon quantum dots for live/dead and reduced cytotoxicity compared to viability staining commercial laboratory stains Demonstration of the ability to bio- Technology demonstration for versatile functionalize microdiscs according to and rapid detection/isolation of other other target particles (e.g. avidin-coated target particles/cells. polystyrene particles) Demonstration of the feasibility of a Technology demonstration with high- microfluidic magnetic separation throughput capabilities for in-the-field (µFMS) platform to process high-volume applications. samples in short periods of time (e.g. 15 min). Preliminary demonstration Preliminary technology demonstration for particle/bacterial target magnetic the enhancement of binding events by isolation after microdisc induced rotation creating a ‘micromixing’ effect. Further via the application of a rotating studies encouraged (field exposure times magnetic field. and different rotation frequencies).

138

6.2 Future Perspectives

Future perspectives include 1) the use of other bioreceptors (e.g. lectins, antibodies, phages, or peptides) to isolate other target cells (i.e. Enterococci), 2) further studies on the rotational dynamics of the microdiscs to enhance binding interactions with bacteria and hence their magnetic isolation, and 3) the integration of the magnetic separation microfluidic device with a portable microscopy platform along with a image- processing algorithm for near-quantification of bacterial targets and microdiscs in field samples.

Figure 6-1. Potential applications of this method include water quality monitoring in recreational, fisheries, and irrigation waters. Photos reprinted with permission from Jorge Ramírez Portela – Grupo SIN, http://www.noticiassin.com (April 26, 2019) [101], Ana Martínez – Metro PR, http://www.metro.pr (April 26, 2019) [102], Shannon Tompkins – Houston Chronicle, https://www.chron.com (April 26, 2019) [103] , and iStock / Getty Images Plus – The Pew Charitable Trusts, https://www.pewtrusts.com (April 26, 2019) [104], respectively.

6.2.1 Other Bacterial Targets

The method developed here has demonstrated a rapid (< 2 hour) isolation and detection of E. coli with no enrichment steps with a limit of detection down to 100

139

CFU/100 mL. This detection limit is below the U.S. EPA threshold of 126 CFU/100 mL for recreational waters. Therefore, the development of this bacterial isolation technology for on-site detection of fecal indicating bacteria (i.e. E. coli and Enterococci) in recreational waters (lakes, beaches, rivers, etc) is a potential application. Figure 6-1 show images of a few applications envisioned for this work.

6.2.2 Microdisc Rotation

As presented in Chapter 5, preliminary results on the exploration of magnetic microdisc rotation induced via the application of rotating magnetic fields for the enhancement of particle/bacterial targets binding interactions were obtained. Here, promising results showing higher fluorescence (quantified using ImageJ) was observed in samples in which microdiscs were exposed to rotating magnetic fields. Here, it is hypothesized that the rotation of the microdiscs will create a ‘mixing’ effect, which will enhance the interactions of the microdiscs with its environment (particles or bacterial cells in suspension). However, further experiments are encouraged to study the effect of

1) longer exposures to rotating magnetic fields and 2) different frequencies of rotation.

Additionally, other quantification techniques, such as, flow cytometry as a secondary confirmation to determine if the binding interactions (or isolation) of particle/bacterial targets was enhanced are recommended. Finally, if isolation (or binding events) of targets is confirmed to be enhanced, modification of the current apparatus to hold higher sample sizes closer to 100 mL is recommended.

6.2.3 Portable Microscopy

Portable microscopy has been explored previously and even some commercially available setups are (or will be) available from companies (or institutions) such as

AmScope, Foldscope, and CellScope [105]–[107]. Now, in terms of fluorescence

140

microscopy in the last few years several works have been trying to develop some portable systems to obtain fluorescent images, which can be applied to biological applications [108]–[113], including that on Figure 6-2A.

Now, considering that the biosensing system detection mechanism presented in this work consists of fluorescence microscopy inspection, for field applications a portable platform would be ideal. In fact, using a mobile device (e.g. a smartphone) for the data acquisition step could lead to making this detection system a sensor analytics point solutions (SNAPS) as discussed in [114]. Where the biosensing system data

(obtained from imaging) could meet analytics using a smartphone-based platform.

The overall future perspective or idea related to a smartphone-based portable fluorescence microscopy platform for this work is described in Figure 6-3. In brief, the bio-functionalized microdiscs would be incubated in a lab-prepared and/or collected water sample to selectively bind fecal indicating bacterial targets (20-30 min). Then, magnetic isolation will occur over a small area (20 mm × 20 mm) by flowing water samples through a microfluidic device with “magnetic traps” consisting of an array of external permanent magnets (Figure 6-3A), as described in Chapter 4. Finally, isolated microdiscs/bacteria conjugates would be fluorescently tagged by flowing fluorescent stains through the same channel and inspected using fluorescent microscopy. Figure 6-

3B shows the method used in this work, while Figure 6-3C depicts the idea of a portable, smartphone-based, fluorescent imaging platform for field applications.

Design of a fluorescent imaging setup for the inspection of magnetically isolated bacteria samples is envisioned using a smartphone device and a 3D printed platform that holds lenses, filters, and the sample, similar to that presented by S. Shrivastava in

141

reference [113] (Figure 6-2A). Regarding the fluorescent tags, possible stains to be used could include green fluorescein protein (GFP), SYTO9, and/or green-fluorescing carbon quantum dots (CDs), for which the excitation filter needed will range in the 480 nm and an emission filter or light emitting diode (LED) close to 500 nm.

Figure 6-2. Diagram of envisioned imaging platform setup - adapted from reference [113] and B) Example fluorescence images of discs and bacteria conjugates (manually traced circles/ovals represent visualized quantification detection of discs using image processing tools).

As has been shown in Figure 5-15, the fluorescence signal seems to be directly correlated to bacteria concentrations. Therefore, image post-processing methods to analyze and potentially quantify fluorescence to correlate it with bacteria concentrations could be explored for this smartphone based fluorescence imaging platform. It can be hypothesized that as higher fluorescence is detected by the designed fluorescent microscopy platform, higher concentration of bacteria will be present in the samples. A first prototype idea is presented in Appendix H (Figures H-1 – H3), where some tests

142

were performed to detect different concentrations of a fluorescing protein (i.e. Biotin-4- fluorescein).

Figure 6-3. Current and envisioned future setup for the magnetic isolation and fluorescence inspection of bacterial targets. A) Magnetic isolation setup, B) current microscopy inspection method, and C) smartphone based fluorescent imaging setup.

6.2.4 Other applications

In this work, interest in isolating particle or bacterial targets in concentrated area

(µFMS device) was based on the need of sample inspection to determine presence of the target as a detection system. Another potential application could be the isolation of unwanted particles or cells from a sample of interest (e.g. water, blood, or others). For example, commercially available systems like MACS® (magnetic-activated cell sorting),

143

have explained the use of their technology not only for cell sorting, but also for depletion of unwanted cells [115].

For the isolation of unwanted particles or bacteria in a sample, the use of the bio- functionalized magnetic microdiscs and the µFMS device could be used to target and isolate these unwanted particles or bacterial cells. In that case, the filtrate of the processed sample in the reservoir (Figure 6-3A) would be ready to use for the intended purposes/analysis (after unwanted particles or cells have been removed).

Another potential application for the system proposed here involves the use of the magnetic patterning inside the microfluidic channels of the µFMS device (similar to

[52]) to potentially speed up the microscopy inspection process. It is important to consider that the fabrication process will change and the cost might increase, but further evaluation is encouraged for a potentially faster inspection of samples once the bacterial targets have been isolated in the µFMS device (and potentially provide a fluorescence cell counting system).

144

APPENDIX A GRAM-POSITIVE AND GRAM-NEGATIVE BACTERIA CELL WALL STRUCTURES

Figure A-1. (A) Structure and composition of Gram-positive and Gram-negative cell walls. (From Figure 4-11, p 83 of TORTORA, GERALD, MICROBIOLOGY: INTRODUCTION, 3rd Edition, © 1989. Reprinted with permission of Pearson Education, Inc., Upper Saddle River, NJ.) (B) Penicillin binding protein 2 (PBP2) from S. aureus (beyond the focus of this research). For more information see the originally accessed source: [116]. Copyright © 2020 McGraw-Hill Education. All rights reserved.

From figure A-1, it can be observed that both Gram-positive and Gram-negative cell walls are composed of peptidoglycan (glycan) layers, which is targeted by Con A lectins. In the case of E. coli, the O-antigen inside the LPS layer is what will be targeted by the DNA aptamers.

145

APPENDIX B SURFACE BIO-FUNCTIONALIZATION OF MAGNETIC MICRODISCS: SELECTIVITY

Previous works also by the McLamore group at UF, have tested 9 different lectins against various bacteria using impedimetric and plasmonic sensors, and confirmed the results with Colifast kits (MU-D-galactosidase assay). Three examples are shown in Table B-1 using a mannose binding lectin (Con A), a C type lectin (SIGN-

R1), and an F-type lectin (FUC). These results suggest Con A as the best candidate for selectivity to total coliforms. Figure 3-4 depicts bio-functionalization concepts for both general coliforms and E. coli.

Table B-1. Lectin selectivity for coliforms (Courtesy of Dr. McLamore). Total Coliform (TC) Listeria innocua TC + L. innocua LOD LOD LOD [CFU/100 [CFU/100m [CFU/100m Sensor mL] Colifast L] Colifast L] Colifast Con A (impedimetric) 3 ± 2 positive ND negative 4 ± 3 positive Con A (plasmonic) 2 ± 2 positive ND negative 3 ± 3 positive *SIGN-R1 (impedimetric) 11 ± 4 positive 45 ± 8 negative 112 ± 18 positive *SIGN-R1 (plasmonic 15 ± 6 positive 69 ± 9 negative 135 ± 41 positive FUC (impedimetric) 8 ± 4 positive 7 ± 5 negative 26 ± 12 positive FUC (plasmonic) 2 ± 3 positive 8 ± 4 negative 15 ± 8 positive ND=no detection, *10mM CaCl2 added to sample

For specific capture of E. coli., DNA aptamers are used as the capture probes, as demonstrated in previous works [36], [37]. For coating gold-sputtered microdiscs with aptamer, a thiol (-SH) group is first inserted at the 3’ terminal end with a six-carbon spacer using Traut’s reagent, which ensures adequate mobility for bacteria binding. The

DNA aptamers function as a capture probe that discriminates E. coli among other

Gram-negative bacteria and interferents by forming a covalent bond with cell surface O- antigen specific to E. coli. The 67mer binds at stem-loop structures distal to the thiol

146

group at the 3’ end. In other work by the McLamore lab at UF using impedimetric sensors and surface plasmon resonance sensors, these aptamers have shown strong selectivity to E. coli K12 and E. coli O157:H7 over other Gram-negative and Gram- positive cells, as shown in Fig. A-2. Also, validation of this aptamer against an IgG antibody to confirm selectivity was performed, and there no significant difference between affinity constant (KD=240 nM) for the aptamer versus the antibody was observed.

]

]

25 120 -1 -1 Plasmonic sensor Impedimetric sensor 20 100

80 /CFU-mL 15 60 10 40 5 20

0 0

-5 -20

Plasmonic sensitivity [RU/CFU-mL

Impedimetric sensitivity [k E coli K12 Salmonella E coli O157:H7 Listeria inoocua Campylobacter jejuni Lilsteria monocytogenes

Figure B-1. Selectivity of aptamer for E. coli over other Gram-negative and Gram- positive cells. (Courtesy of Dr. McLamore).

Additional tests to determine the affinity of aptamers to gold were performed in the McLamore lab by conjugating aptamers to gold electrodes following the methods in

Burrs et al. (2016) and Rong et al. (2018). Briefly, a disposable microwell was fixed to the electrode and the de-protected aptamer was drop cast into the well, then allowed to bind for 6 hours. The microwell fixture was subsequently removed and the surface rinsed three times in DI prior to analysis.

147

A commercial gold disc electrode (1.6 mm diameter) was used to confirm aptamer binding chemistry both for aptamer P12-55 and the scrambled aptamer for targeting E. coli 25922. Nyquist plots show that charge transfer resistance (Rct) significantly increased after addition of thiolated aptamer, as expected. In preliminary studies, it was determined that the optimum concentration of aptamer was 200 nM, resulting in a surface coverage of 40 nM/mm2. Using this optimum concentration, P12-

55 had a strong affinity toward E. coli 25922 in a buffer solution at pH 7.1 with 4 mM

K4FeCN6 and 1 M KNO3 (Faradaic impedance) (Figure A-3). The change in Rct (22  5 k/CFU/mL) was linear over the range of 20 to 500 CFU/100 mL. This represents the first report of aptamer P12-55 when thiolated and conjugated to a metal surface and shows that addition of the C6 spacer and thiol tag had no detrimental effect. Although the specific molecular target of aptamer P12-55 was not reported by Marton et al.

(2016), the authors estimate that the aptamer has a high affinity for E. coli 25922 cells, with an estimated 14,950 binding sites on a single E. coli cell. On the other hand, the scrambled aptamer (three single bp substitutions in the stem-loop structures) did not respond to E. coli 25922 at any concentration; although the Rct increased significantly after conjugating the aptamer, indicating the thiol-gold bond was intact (Figure B-2).

This result confirms the results by Marton et al. (2016) in studies of aptamer-cell suspensions and validates that single bp substitutions to the stem-loop region significantly alter binding affinity, likely resulting from a change in secondary structure

[63].

To further challenge this result, the selectivity of aptamer P12-55 and the scrambled aptamer were tested with various Gram negative cells and Gram positive

148

cells (including mixtures) (Figure B-3). Aptamer P12-55 was 92% selective over

Pseudomonas and 97% selective over Salmonella. Electrodes were also tested in municipal wastewater samples collected from a secondary clarifier, which was used as a positive control during initial sensor development.

Figure B-2. Electrochemical impedance spectroscopy analysis of thiolated aptamers on a gold electrode toward E. coli 25922. (A) Bode plots for P12-55 88mer; inset shows linear calibration plot using impedance at a cutoff frequency of 6.31 Hz (Faradaic impedance with 4mM K4FeCN6) (B) Cutoff frequency was performed between 1-21.54 Hz for sensitivity toward E. coli and linear regression coefficient. (Courtesy of Dr. McLamore).

Figure B-3. Selectivity of aptamer toward E. coli 25922 over other Gram negative bacteria. (Courtesy of Dr. McLamore).

149

scramble

scramble

KEY THYMINE 5’ 5’ ADENINE 3’ 3’ CYTOSINE GUANINE SH SH Figure B-4. Cartoon representation of secondary structure predicted using mfold. (left) P12-55 aptamer by Marton et al (2016) (right) Location of three single bp mismatches in P12-55 aptamer (i.e., scrambled aptamer). (Courtesy of Dr. McLamore).

Figure B-5. EIS analysis of scrambled aptamer shows no selectivity toward E. coli 25922. (A) Bode plots. B) Response curve shows no sensitivity toward E. coli 25922. (Courtesy of Dr. McLamore).

150

APPENDIX C CONFIRMATION TESTS OF BACTERIA GROWTH FROM MICROBIOLOGICS E. COLI/COLIFORMS KITS

Figure C-1. Confirmation test: -Methylumbelliferyl β-D-galactopyranoside (MUG) assay. Photos courtesy of Dr. Eric McLamore.

MUG assay was performed with a Colilert microwell assay to confirm E. coli growth before and after capture by magnetic discs. Here, “stock” sample refers to the E. coli stock solution (2.4 × 108 CFU/100mL) used for the dilutions for the magnetic isolation process. “Filtrate 1:1” refers to the supernatant of the incubated and filtered sample using magnetic microdiscs at a ratio of 1:1 (supernatant E. coli solution:PBS).

Lastly, “Filtrate 1:2” refers to the supernatant of the incubated and filtered sample using magnetic microdiscs but this time at a ratio of 1:2 (supernatant E. coli solution:PBS).

151

Figure C-2. Confirmation test: Fluorescent emission of MUG assay at 445 nm measured using fiber optic sensor. (Courtesy of Dr. McLamore).

Fluorescent emission of the MUG assay was quantified with a fiber optic sensor

(Ocean Optics USB4000) which was used to measure intensities at λ=454 nm under excitation at 405 nm for four different dilutions of “stock” sample. Average fluorescence intensity for the samples is shown (n=3) and error bars represent standard deviation.

Figure C-3. Confirmation test: Culture colony counting. Photo courtesy of Dr. McLamore.

A duplicate of the diluted E. coli solution (1 CFU/100 mL) used for our incubation and isolation process with magnetic microdiscs was cultured for 24 hours and an isolated colony was confirmed.

152

Figure C-4. E. coli and Coliforms kits used for lab-prepared samples. Photos/diagram courtesy and reprinted with permission of Microbiologics, Inc., https://www.microbiologics.com/ [117].

153

In this work, samples were classified under the biosafety level 1 (BSL-1). BSL-1 indicates that the microbes or cells in the samples are “not known to cause disease in healthy adults and present minimal potential hazard to laboratories and the environment” [118].

As a summary, for BSL-1 the personal and laboratory practices still require the use of adequate personal protective equipment (such as, lab coats, gloves, and eye protection, as needed), following of standard microbiological practices even though work can be performed in open laboratory benches and/or tables. Also, in terms of the facility, there should be a sink available for hand washing [118].

154

APPENDIX D COMSOL CAPTURE EFFICIENCY OF DISCS

Figure D-1. Plot of simulated capture efficiencies (COMSOL) as a function of change in hydrodynamic diameter of IONs, considering possible aggregation of particles. It is shown how the capture efficiency increases as the hydrodynamic/magnetic diameter increases.

155

APPENDIX E VSM DATA FOR MICRODISC QUANTIFICATION

Figure E-1. VSM data obtained (before and after filtration) from 50-mL sample containing iron-oxide nanoparticles (IONs) at a concentration of 0.1 mg/mL filtered using the µFMS device at 120 µL/s. The black dashed line represents the average saturation magnetic moment for each of the samples (8.10 ×10-8 A·m2 for the 20-µL from the stock sample, and 2.43 ×10-8 A·m2 for the 20-µL from the filtrate sample), which resulted in the 70.0% capture efficiency.

Figure E-2. Image of 50-mL sample before and after filtration using the µFMS device at 120 µL/s. Photo courtesy of author.

156

APPENDIX F CELL CULTURE COLONY COUNTING CONFIRMATION RESULTS

PLATING COURTESY OF ADAM GROSSMAN (UF MICROBIOLOGY, RICE LAB)

Figure F-1. Plating results for E. coli sample enriched for 23 hours in TSB. Photos courtesy of Adam B. Grossman.

157

APPENDIX G IMAGEJ EXAMPLE IMAGES FROM IMAGE POST-PROCESSING

Figure G-1. Example images after post-processing in ImageJ “color thresholding.” Here, bright green pixels and dark gray pixels are counted for latter comparison. Considering that the number of pixels of each are proportional to the number of fluorescence cells and microdiscs, respectively.

158

APPENDIX H SMARTPHONE FLUORESCENCE MICROSCOPE PLATFORM

Figure H-1. Prototype for smartphone fluorescence microscopy platform along with two different filters for green-fluorescing samples. Photos courtesy of author.

Figure H-2. Filter placement and setup powered. Excitation filter: ~480 nm, Emission filter (LED): ~500 nm. Photos courtesy of author.

Figure H-3. Example images of different concentrations of Biotin-4-fluorescing obtained using the prototype with two different filters. Photos courtesy of author.

159

LIST OF REFERENCES

[1] World Health Organization, “Food safety,” 2017. [Online]. Available: http://www.who.int/news-room/fact-sheets/detail/food-safety. [Accessed: 03-Dec- 2018].

[2] World Health Organization, “Drinking-water,” 2017. [Online]. Available: http://www.who.int/news-room/fact-sheets/detail/drinking-water. [Accessed: 03- Dec-2018].

[3] R. P. McClung et al., “Waterborne Disease Outbreaks Associated With Environmental and Undetermined Exposures to Water — United States, 2013– 2014,” CDC MMWR. Morb. Mortal. Wkly. Rep., vol. 66, no. 44, pp. 1222–1225, Nov. 2017.

[4] “National Outbreak Reporting System (NORS) Dashboard | CDC.” [Online]. Available: https://wwwn.cdc.gov/norsdashboard/. [Accessed: 05-Nov-2019].

[5] U. S. Environmental Protection Agency, “OFFICE OF WATER 820-F-12-058 Recreational Water Quality Criteria.” [Online]. Available: https://www.epa.gov/sites/production/files/2015-10/documents/rwqc2012.pdf. [Accessed: 20-Mar-2019].

[6] K. Pond and World Health Organization, “Water Recreation and Disease Plausibility of Associated Infections: Acute Effects, Sequelae and Mortality.” [Online]. Available: www.iwapublishing.com. [Accessed: 20-Mar-2019].

[7] U.S. Environmental Protection Agency, “Revised Total Coliform Rule (RTCR): A Quick Reference Guide,” 2013.

[8] N. R. Council, Indicators for Waterborne Pathogens. National Academies Press, 2004.

[9] C. M. Rock et al., “Review of water quality criteria for water reuse and risk-based implications for irrigated produce under the FDA Food Safety Modernization Act, produce safety rule,” Environ. Res., pp. 616–629, May 2019.

[10] R. T. Noble and S. B. Weisberg, “A review of technologies for rapid detection of bacteria in recreational waters,” Journal of Water and Health, vol. 3, no. 4. IWA Publishing, pp. 381–392, 2005.

[11] D. Ivnitski, I. Abdel-Hamid, P. Atanasov, and E. Wilkins, “Biosensors for detection of pathogenic bacteria,” Biosensors and Bioelectronics, vol. 14, no. 7. pp. 599– 624, Oct-1999.

[12] J. Vidic et al., “Point-of-need DNA testing for detection of foodborne pathogenic bacteria,” Sensors (Switzerland), vol. 19, no. 5. MDPI AG, 01-Mar-2019.

160

[13] J.-Y. Yoon and B. Kim, “Lab-on-a-chip pathogen sensors for food safety.,” Sensors (Basel)., vol. 12, no. 8, pp. 10713–41, 2012.

[14] G. A. Cangelosi and J. S. Meschke, “Dead or Alive: Molecular Assessment of Microbial Viability,” Appl. Environ. Microbiol., vol. 80, no. 19, pp. 5884–5891, Oct. 2014.

[15] D. C. Vanegas, C. L. Gomes, N. D. Cavallaro, D. Giraldo-Escobar, and E. S. McLamore, “Emerging Biorecognition and Transduction Schemes for Rapid Detection of Pathogenic Bacteria in Food,” Compr. Rev. Food Sci. Food Saf., vol. 16, no. 6, pp. 1188–1205, Nov. 2017.

[16] H. Y. Tsen, C. K. Lin, and W. R. Chi, “Development and use of 16S rRNA gene targeted PCR primers for the identification of Escherichia coli cells in water,” J. Appl. Microbiol., vol. 85, no. 3, pp. 554–560, Sep. 1998.

[17] B. Li, Q. Yu, and Y. Duan, “Fluorescent labels in biosensors for pathogen detection,” Crit. Rev. Biotechnol., vol. 35, no. 1, pp. 82–93, Jan. 2015.

[18] I. Ŝafařík and M. Ŝafaříková, “Invited Review Magnetic Nanoparticles and Biosciences,” 2002.

[19] Q. A. Pankhurst, C. J, S. K. Jones, and J. Dobson, “Applications of magnetic nanoparticles in biomedicine,” J. Phys. D. Appl. Phys., vol. 36, pp. 167–181, 2003.

[20] Q. A. Pankhurst, T. K. Thanh, S. K. Jones, and J. Dobson, “Progress in applications of magnetic nanoparticles in biomedicine,” J. Phys. D. Appl. Phys., vol. 42, p. 15, 2009.

[21] V. Novosad and E. Rozhkova, “Ferromagnets-Based Multifunctional Nanoplatform for Targeted Cancer Therapy,” in Biomedical Engineering, Trends in Materials Science, InTech, 2011.

[22] “Web of Science [v.5.34] - Web of Science Core Collection Result Analysis.” [Online]. Available: http://wcs.webofknowledge.com/RA/analyze.do?product=WOS&SID=7FhqZPxRf KbzIIfNBoe&field=TASCA_JCRCategories_JCRCategories_en&yearSort=false. [Accessed: 02-Mar-2020].

[23] C. P. Hunt, B. M. Moskowitz, and S. K. Banerjee, “Magnetic Properties of Rocks and Minerals,” Am. Geophys. Union, 1995.

[24] A. S. Teja and P.-Y. Koh, “Synthesis, properties, and applications of magnetic iron oxide nanoparticles,” Prog. Cryst. Growth Charact. Mater., vol. 55, pp. 22–45, 2009.

161

[25] M. Hofmann-Amtenbrink, B. Von Rechenberg, and H. Hofmann, “Chapter 5: Superparamagnetic nanoparticles for biomedical applications,” in Transworld Research Network, 2009.

[26] W. Jiang et al., “The effect of [Fe 3+ ]/[Fe 2+ ] molar ratio and iron salts concentration on the properties of superparamagnetic iron oxide nanoparticles in the water/ethanol/toluene system,” J Nanopart Res, vol. 13, pp. 5135–5145, 2011.

[27] V. Sundaresan, J. U. Menon, M. Rahimi, K. T. Nguyen, and A. S. Wadajkar, “Dual-responsive polymer-coated iron oxide nanoparticles for drug delivery and imaging applications,” J. Pharm., vol. 466, pp. 1–7, 2014.

[28] M. Varshney and Y. Li, “Interdigitated array microelectrode based impedance biosensor coupled with magnetic nanoparticle-antibody conjugates for detection of Escherichia coli O157:H7 in food samples,” Biosens. Bioelectron., vol. 22, pp. 2408–2414, 2007.

[29] D. Kwon, S. Lee, M. Mo Ahn, I. S. Kang, K.-H. Park, and S. Jeon, “Colorimetric detection of pathogenic bacteria using platinum-coated magnetic nanoparticle clusters and magnetophoretic chromatography,” Anal. Chim. Acta, vol. 883, pp. 61–66, 2015.

[30] Y. Wang and E. C. Alocilja, “Gold nanoparticle-labeled biosensor for rapid and sensitive detection of bacterial pathogens,” J. Biol. Eng., vol. 9, no. 16, 2015.

[31] E. C. Alocilja, P. Jain, and K. Pryg, “Immunosensor for rapid extraction/detection of enteric pathogens,” TECHNOLOGY, vol. 04, no. 03, pp. 194–200, Sep. 2016.

[32] R. P. Cowburn, D. K. Koltsov, A. O. Adeyeye, M. E. Welland, and D. M. Tricker, “Single-Domain Circular Nanomagnets,” 1999.

[33] W. Scholz et al., “Transition from single-domain to vortex state in soft magnetic cylindrical nanodots,” J. Magn. Magn. Mater., vol. 266, pp. 155–163, 2003.

[34] D.-H. Kim et al., “Biofunctionalized magnetic-vortex microdiscs for targeted cancer-cell destruction,” Nat. Mater., vol. 9, 2009.

[35] N. Garraud and D. P. Arnold, “Characterization of fluids via measurement of the rotational dynamics of suspended magnetic microdiscs,” Cit. J. Appl. Phys., vol. 117, pp. 17–320, 2015.

[36] K. Y. Castillo-Torres, N. Garraud, D. P. Arnold, and E. S. McLamore, “Investigation of magnetic microdiscs for bacterial pathogen detection,” 2016, vol. 9863, p. 98630G.

[37] K. Y. Castillo-Torres, N. Garraud, D. P. Arnold, and E. S. McLamore, “Towards pathogen detection via optical interrogation of magnetic microdiscs,” Tech. Dig. Solid-State Sensors, Actuators, Microsystems Work. (hilt. Head 2016), 2016.

162

[38] H. Jayamohan et al., “Highly Sensitive Bacteria Quantification Using Immunomagnetic Separation and Electrochemical Detection of Guanine-Labeled Secondary Beads,” Sensors, vol. 15, no. 5, pp. 12034–12052, May 2015.

[39] S. Zhan et al., “Highly Efficient Removal of Pathogenic Bacteria with Magnetic Graphene Composite,” 2015.

[40] E. B. Setterington and E. C. Alocilja, “Electrochemical Biosensor for Rapid and Sensitive Detection of Magnetically Extracted Bacterial Pathogens,” Biosensors, vol. 2, no. 1, pp. 15–31, Jan. 2012.

[41] Y. Jin, J. Deng, J. Liang, C. Shan, and M. Tong, “Efficient bacteria capture and inactivation by cetyltrimethylammonium bromide modified magnetic nanoparticles,” Colloids Surfaces B Biointerfaces, vol. 136, pp. 659–665, Dec. 2015.

[42] Y. Luo and E. C. Alocilja, “Portable nuclear magnetic resonance biosensor and assay for a highly sensitive and rapid detection of foodborne bacteria in complex matrices,” J. Biol. Eng., vol. 11, no. 1, p. 14, Dec. 2017.

[43] R. A. Bohara, N. D. Throat, N. A. Mulla, and S. H. Pawar, “Surface-Modified Cobalt Ferrite Nanoparticles for Rapid Capture, Detection, and Removal of Pathogens: a Potential Material for Water Purification,” Appl. Biochem. Biotechnol., vol. 182, no. 2, pp. 598–608, Jun. 2017.

[44] F. Halouane et al., “Selective isolation and eradication of E. coli associated with urinary tract infections using anti-fimbrial modified magnetic reduced graphene oxide nanoheaters †,” J. Mater. Chem. B, vol. 5, p. 8133, 2017.

[45] Y. Xu, H. Wang, C. Luan, Y. Liu, B. Chen, and Y. Zhao, “Aptamer-based hydrogel barcodes for the capture and detection of multiple types of pathogenic bacteria,” Biosens. Bioelectron., vol. 100, pp. 404–410, Feb. 2018.

[46] K. Kant et al., “Microfluidic devices for sample preparation and rapid detection of foodborne pathogens,” Biotechnology Advances, vol. 36, no. 4. Elsevier Inc., pp. 1003–1024, 01-Jul-2018.

[47] Y. Shen, Y. Yalikun, and Y. Tanaka, “Recent advances in microfluidic cell sorting systems,” Sensors and Actuators, B: Chemical, vol. 282. Elsevier B.V., pp. 268– 281, 01-Mar-2019.

[48] T. Zhang et al., “Focusing of sub-micrometer particles in microfluidic devices,” Lab Chip, 2019.

[49] Z. Liu et al., “High throughput capture of circulating tumor cells using an integrated microfluidic system,” Biosens. Bioelectron., vol. 47, pp. 113–119, Sep. 2013.

163

[50] K. Loutherback, J. D’Silva, L. Liu, A. Wu, R. H. Austin, and J. C. Sturm, “Deterministic separation of cancer cells from blood at 10 mL/min,” AIP Adv., vol. 2, no. 4, 2012.

[51] D. Royet et al., “Using injection molding and reversible bonding for easy fabrication of magnetic cell trapping and sorting devices,” J. Magn. Magn. Mater., vol. 427, pp. 306–313, 2017.

[52] J. Pivetal et al., “Selective isolation of bacterial cells within a microfluidic device using magnetic probe-based cell fishing,” Sensors Actuators, B Chem., vol. 195, pp. 581–589, May 2014.

[53] N. Xia et al., “Combined microfluidic-micromagnetic separation of living cells in continuous flow,” Biomed. Microdevices, vol. 8, no. 4, pp. 299–308, Dec. 2006.

[54] R. Zhou and C. Wang, “Microfluidic separation of magnetic particles with soft magnetic microstructures,” Microfluid. Nanofluidics, vol. 20, no. 3, pp. 1–11, Mar. 2016.

[55] K. Hoshino et al., “Microchip-based immunomagnetic detection of circulating tumor cells,” Lab Chip, vol. 11, no. 20, pp. 3449–3457, Oct. 2011.

[56] C. W. Yung, J. Fiering, A. J. Mueller, and D. E. Ingber, “Micromagnetic- microfluidic blood cleansing device,” Lab Chip, vol. 9, no. 9, pp. 1171–1177, 2009.

[57] K. Y. Castillo-Torres, D. P. Arnold, and E. S. McLamore, “Rapid isolation of Escherichia coli from water samples using magnetic microdiscs,” Sensors Actuators, B Chem., vol. 291, 2019.

[58] Y. Chen et al., “One-step detection of pathogens and viruses: Combining magnetic relaxation switching and magnetic separation,” ACS Nano, vol. 9, no. 3, pp. 3184–3191, Mar. 2015.

[59] L. F. Zanini, N. M. Dempsey, D. Givord, G. Reyne, and F. Dumas-Bouchiat, “Autonomous micro-magnet based systems for highly efficient magnetic separation,” Appl. Phys. Lett., vol. 99, no. 23, Dec. 2011.

[60] C. M. Earhart, R. J. Wilson, R. L. White, N. Pourmand, and S. X. Wang, “Microfabricated magnetic sifter for high-throughput and high-gradient magnetic separation,” J. Magn. Magn. Mater., vol. 321, no. 10, pp. 1436–1439, May 2009.

[61] O. Osman et al., “Microfluidic immunomagnetic cell separation using integrated permanent micromagnets,” Biomicrofluidics, vol. 7, no. 5, Sep. 2013.

[62] J.-J. Lee et al., “Synthetic Ligand-Coated Magnetic Nanoparticles for Microfluidic Bacterial Separation from Blood,” 2013.

164

[63] S. Marton, F. Cleto, M. A. Krieger, and J. Cardoso, “Isolation of an Aptamer that Binds Specifically to E. coli,” PLoS One, vol. 11, no. 4, p. e0153637, Apr. 2016.

[64] W. Wu et al., “Aptasensors for rapid detection of Escherichia coli O157:H7 and Salmonella typhimurium,” Nanoscale Res. Lett., vol. 7, no. 1, p. 658, 2012.

[65] B. Nagel, H. Dellweg, and L. M. Gierasch, “GLOSSARY FOR CHEMISTS OF TERMS USED IN BIOTECHNOLOGY (IUPAC Recommendations 1992),” Pure Appl. Chem, vol. 64, no. 1, pp. 143–168, 1992.

[66] W. R. Heineman and W. B. Jensen, “Leland C. Clark Jr. (1918–2005),” Biosens. Bioelectron., vol. 21, no. 8, pp. 1403–1404, Feb. 2006.

[67] G. Rodriguez J and R. M, “Importance of Biosensors,” Biosens. J., vol. 04, no. 01, Jul. 2015.

[68] N. Bhalla, P. Jolly, N. Formisano, and P. Estrela, “Introduction to biosensors.,” Essays Biochem., vol. 60, no. 1, pp. 1–8, 2016.

[69] A. Rasooly and K. E. Herold, Optical-based detectors. Humana Press, 2009.

[70] T.-H. Ku et al., “Nucleic Acid Aptamers: An Emerging Tool for Biotechnology and Biomedical Sensing,” Sensors, vol. 15, no. 7, pp. 16281–16313, Jul. 2015.

[71] B. Wang, J. Anzai, B. Wang, and J. Anzai, “Recent Progress in Lectin-Based Biosensors,” Materials (Basel)., vol. 8, no. 12, pp. 8590–8607, Dec. 2015.

[72] O. Neth, D. L. Jack, A. W. Dodds, H. Holzel, N. J. Klein, and M. W. Turner, “Mannose-Binding Lectin Binds to a Range of Clinically Relevant Microorganisms and Promotes Complement Deposition,” Infect. Immun., vol. 68, no. 2, pp. 688– 693, Feb. 2000.

[73] E. Pensa et al., “The Chemistry of the Sulfur–Gold Interface: In Search of a Unified Model,” Acc. Chem. Res., vol. 45, no. 8, pp. 1183–1192, Aug. 2012.

[74] Molecular Probes, “LIVE/DEAD BacLight Bacterial Viability Kits,” 2004.

[75] R. V. Ramanujan, “Magnetic Particles for Biomedical Applications,” in Biomedical Materials, Boston, MA: Springer US, 2009, pp. 477–491.

[76] D. Jiles, Introduction to magnetism and magnetic materials. Chapman and Hall, 1998.

[77] J. M. D. Coey, Magnetism and Magnetic Materials. Cambridge: Cambridge University Press, 2010.

165

[78] D. Lisjak and A. Mertelj, “Anisotropic magnetic nanoparticles: A review of their properties, syntheses and potential applications,” Progress in Materials Science, vol. 95. Elsevier Ltd, pp. 286–328, 01-Jun-2018.

[79] M. Goiriena-Goikoetxea, A. García-Arribas, M. Rouco, A. V Svalov, and J. M. Barandiaran, “High-yield fabrication of 60 nm Permalloy nanodiscs in well-defined magnetic vortex state for biomedical applications,” Nanotechnology, vol. 27, no. 17, p. 175302, Apr. 2016.

[80] K. Y. Guslienko, V. Novosad, Y. Otani, H. Shima, and K. Fukamichi, “Magnetization reversal due to vortex nucleation, displacement, and annihilation in submicron ferromagnetic dot arrays.”

[81] R. K. Wangsness, Electromagnetic fields. Wiley, 1986.

[82] T. H. Boyer, “The force on a magnetic dipole,” Am. J. Phys., vol. 56, no. 8, pp. 688–692, Aug. 1988.

[83] “Avidin-Biotin Interactions,” 2008.

[84] R. P. Haugland and W. W. You, “Coupling of Antibodies With Biotin.”

[85] Gene Link, “Thiol Modified Oligo Disulfide Reduction.” .

[86] Y.-P. Sun et al., “Quantum-Sized Carbon Dots for Bright and Colorful Photoluminescence,” J. Am. Chem. Soc., vol. 128, no. 24, pp. 7756–7757, Jun. 2006.

[87] P. G. Luo et al., “Carbon ‘quantum’ dots for optical bioimaging,” J. Mater. Chem. B, vol. 1, no. 16, p. 2116, Mar. 2013.

[88] D. Qin, Y. Xia, and G. M. Whitesides, “Soft lithography for micro- and nanoscale patterning,” Nat. Protoc., vol. 5, no. 3, pp. 491–502, Mar. 2010.

[89] K. Chen et al., “Tumor cell capture patterns around aptamer-immobilized microposts in microfluidic devices,” Biomicrofluidics, vol. 11, no. 5, p. 054110, Sep. 2017.

[90] J. I. Varillas et al., “Microfluidic Isolation of Circulating Tumor Cells and Cancer Stem-Like Cells from Patients with Pancreatic Ductal Adenocarcinoma The CTC subtypes (EpCAM + CK + CD45-DAPI + CD133 + and EpCAM + CK + CD45- DAPI + CD133-) and CSC subtypes (CD133 + CK + CD45-DAPI + EpCAM + and CD133 + CK + CD45-DAPI + EpCAM,” Theranostics, vol. 9, no. 5, pp. 1417– 1425, 2019.

[91] K. Y. Castillo-Torres, E. S. McLamore, and D. P. Arnold, “A High-Throughput Microfluidic Magnetic Separation (µFMS) Platform for Water Quality Monitoring,” Micromachines, vol. 11, no. 1, p. 16, Dec. 2019.

166

[92] L. F. Zanini, N. M. Dempsey, D. Givord, G. Reyne, and F. Dumas-Bouchiat, “Autonomous micro-magnet based systems for highly efficient magnetic separation,” Appl. Phys. Lett., vol. 99, no. 23, p. 232504, Dec. 2011.

[93] “CELL DIVISION IN THE PRESENCE OF A GROWTH FACTOR.” [Online]. Available: http://www.tiem.utk.edu/~gross/bioed/webmodules/growthfactor.htm. [Accessed: 09-Feb-2020].

[94] N.-T. Nguyen and Z. Wu, “Micromixers—a review,” J. Micromechanics Microengineering, vol. 15, no. 2, pp. R1–R16, Feb. 2005.

[95] Liang-Hsuan Lu, Kee Suk Ryu, and Chang Liu, “A magnetic microstirrer and array for microfluidic mixing,” J. Microelectromechanical Syst., vol. 11, no. 5, pp. 462– 469, Oct. 2002.

[96] A. R. and and M. A. M. Gijs*, “Manipulation of Self-Assembled Structures of Magnetic Beads for Microfluidic Mixing and Assaying,” 2004.

[97] H. Suzuki, C.-M. Ho, and N. Kasagi, “A Chaotic Mixer for Magnetic Bead-Based Micro Cell Sorter,” J. Microelectromechanical Syst., vol. 13, no. 5, pp. 779–790, Oct. 2004.

[98] R. Ghosh, G. A. Buxton, O. B. Usta, A. C. Balazs, and A. Alexeev, “Designing Oscillating Cilia That Capture or Release Microscopic Particles,” Langmuir, vol. 26, no. 4, pp. 2963–2968, Feb. 2010.

[99] H. Masoud and A. Alexeev, “Harnessing synthetic cilia to regulate motion of microparticles,” Soft Matter, vol. 7, no. 19, p. 8702, Sep. 2011.

[100] D. Castillo-Torres, Keisha; Garraud, Nicolas; Mclamore, Eric; Arnold, “Towards pathogen detection via optical interrogation of magnetic microdiscs,” Tech. Dig. Solid-State Sensors, Actuators, Microsystems Work. (hilt. Head), no. HH2016, p. 2, 2016.

[101] Photo by Jorge Ramírez Portela from GFR Media, “Aguas contaminadas y sucias en el Balneario de Luquillo, Puerto Rico.” [Online]. Available: https://noticiassin.com/isla-sin-playas-a-3-82-la-inversion-por-playa-en-puerto- rico/201703403_prhoy_manejo_costas_playas_jrp_/. [Accessed: 26-Apr-2019].

[102] Photo by Ana Martínez from Metro PR, “Inundado de basura islote metropolitano | Metro.” [Online]. Available: https://www.metro.pr/pr/noticias/2013/03/13/inundado- basura-islote-metropolitano.html. [Accessed: 26-Apr-2019].

[103] Photo by Shannon Tompkins, “Where to go fishing in the Houston area - Houston Chronicle.” [Online]. Available: https://www.chron.com/life/travel/explore/article/Where-to-go-fishing-in-the- Houston-area-12901755.php. [Accessed: 26-Apr-2019].

167

[104] Photo from iStock Getty Images Plus, “Romaine Lettuce Contamination Reinforces Need for Agricultural Water Quality Rule | The Pew Charitable Trusts.” [Online]. Available: https://www.pewtrusts.org/en/research-and- analysis/articles/2019/01/07/romaine-lettuce-contamination-reinforces-need-for- agricultural-water-quality-rule. [Accessed: 26-Apr-2019].

[105] Amscope, “AMSCOPE-KIDS 20X 40X Cordless LED Portable Stereo Microscope with Smartphone Holder.” [Online]. Available: https://www.amscope.com/amscope-kids-20x-40x-cordless-led-portable-stereo- microscope-with-smartphone- holder.html?medium=tsa&gclid=EAIaIQobChMIo_aLqu7f4QIVg6_ICh1UWw50EA kYECABEgKBy_D_BwE&gclsrc=aw.ds. [Accessed: 26-Apr-2019].

[106] “Foldscope Instruments | Magnify Your Curiosity.” [Online]. Available: https://www.foldscope.com/. [Accessed: 26-Apr-2019].

[107] “CellScope – UC Berkeley.” [Online]. Available: https://cellscope.berkeley.edu/. [Accessed: 26-Apr-2019].

[108] A. R. Miller et al., “Portable, Battery-Operated, Low-Cost, Bright Field and Fluorescence Microscope,” PLoS One, vol. 5, no. 8, p. e11890, Aug. 2010.

[109] Q. Wei et al., “Fluorescent Imaging of Single Nanoparticles and Viruses on a Smart Phone,” ACS Nano, vol. 7, no. 10, pp. 9147–9155, Oct. 2013.

[110] J.-H. Kim, H.-G. Joo, T.-H. Kim, and Y.-G. Ju, “A smartphone-based fluorescence microscope utilizing an external phone camera lens module,” BioChip J., vol. 9, no. 4, pp. 285–292, Dec. 2015.

[111] Y. Sung, F. Campa, and W.-C. Shih, “Open-source do-it-yourself multi-color fluorescence smartphone microscopy,” Biomed. Opt. Express, vol. 8, no. 11, p. 5075, Nov. 2017.

[112] A. Shanmugam, M. Usmani, A. Mayberry, D. L. Perkins, and D. E. Holcomb, “Imaging systems and algorithms to analyze biological samples in real-time using mobile phone microscopy,” PLoS One, vol. 13, no. 3, p. e0193797, Mar. 2018.

[113] S. Shrivastava, W.-I. Lee, and N.-E. Lee, “Culture-free, highly sensitive, quantitative detection of bacteria from minimally processed samples using fluorescence imaging by smartphone,” Biosens. Bioelectron., vol. 109, pp. 90–97, Jun. 2018.

[114] E. S. McLamore et al., “SNAPS: Sensor Analytics Point Solutions for Detection and Decision Support Systems,” Sensors, vol. 19, no. 22, p. 4935, Nov. 2019.

168

[115] “Magnetic cell separation - Cell separation - Miltenyi Biotec technologies - MACS Handbook - Resources - Miltenyi Biotec - USA.” [Online]. Available: https://www.miltenyibiotec.com/US-en/resources/macs-handbook/macs- technologies/cell-separation/magnetic-cell-separation.html. [Accessed: 08-Mar- 2020].

[116] B. L. Hilal-Dandan R, Ed., “Penicillins, Cephalosporins, and Other β-Lactam Antibiotics,” in Goodman and Gilman’s Manual of Pharmacology and Therapeutics, 2nd ed., New York: AccessPharmacy | McGraw-Hill Medical.

[117] “Microbiologics │ Reference Strains for Microbiological QC Testing.” [Online]. Available: https://www.microbiologics.com/. [Accessed: 26-Apr-2019].

[118] “CDC LC Quick Learn: Recognize the four Biosafety Levels.” [Online]. Available: https://www.cdc.gov/training/QuickLearns/biosafety/. [Accessed: 27-Feb-2020].

169

BIOGRAPHICAL SKETCH

Keisha Y. Castillo-Torres grew up in Carolina, Puerto Rico and attended the

University High School in Río Piedras, San Juan, Puerto Rico. In 2009, Keisha started her Bachelor of Science (BS) degree in electrical engineering at the University of Puerto

Rico, taking courses at the Río Piedras and Bayamón Campuses (Articulated Transfer

Program of the College of Engineering of Mayagüez). In 2011, Keisha transferred to the

University of Puerto Rico at Mayagüez, from where she obtained her electrical engineering BS degree in May 2015.

Keisha started graduate school in August 2015, when she became a graduate

(Ph.D.) research assistant within the Interdisciplinary Microsystems Group (IMG) and the Electrical and Computer Engineering department at the University of Florida (UF). At

UF, she has been working under the co-advisement of Dr. Arnold (Electrical and

Computer Engineering Department) and Dr. McLamore (Agricultural and Biological

Engineering Department) on the application of magnetic microdiscs for detection of waterborne and foodborne bacterial pathogens.

Within the IMG labs, Keisha served as Organizer of the biweekly IMG seminar series for three consecutive years, and currently serves as a student member of the

IMG leadership council. Keisha also served as the Social Chair within the Electrical and

Computer Engineering Graduate Student Organization for one semester and she is currently a Student Member of the IEEE (IEEE Young Professionals and IEEE Women in Engineering) and a member of Tau Beta Pi.

170