Signature Redacted Author

Rational Hydrogel Design for Point-of-Care

Bioassays MASSACHUSETTNSTITNTE by Sarah Jane Shapiro SEP 112019 B.S., University of Oklahoma (2013) LIBRARIES M.S., University of Oklahoma (2014) S.M., Massachusetts Institute of Technology (2017) Submitted to the Department of Chemical Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemical Engineering at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY September 2019 @Massachusetts Institute of Technology 2019. All rights reserved.

Signature redacted Author ...... Department of Chemical Engineering August 28, 2019 Certified by...... Signatureredacted Patrick S. Doyle Robert T. Haslam (1911) Professor of Chemical Engineering Thesis Supervisor

Accepted by ...... Signatureredacted Patrick S. Doyle Robert T. Haslam (1911) Professor of Chemical Engineering Chairman, Committee for Graduate Students 2 Rational Hydrogel Design for Point-of-Care Bioassays by Sarah Jane Shapiro

Submitted to the Department of Chemical Engineering on August 28, 2019, in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemical Engineering

Abstract As the global disease burden shifts increasingly towards chronic diseases, there is a need for improved diagnosis and monitoring so that patients can get the care they need. This is particularly evident in the developing world, where many people live far from diagnostic laboratories. Point-of-care diagnostics are tests that can be run in doctor's offices, clinics, and in patient homes. These tests must be rapid, so that they can be run quickly while the patient waits for results. Established point-of-care technologies are largely centered on lateral flow assays. Hydrogel microparticles have been used extensively for bioassays due to their nonfouling nature and ability to be functionalized with different types of biomolecules. Here, we use polyethylene glycol hydrogel particles to develop point-of-care bioassays. We focus on two different biomarkers: proteins and microRNA. Proteins are well established clinical biomarkers that are regularly tested to diagnose a number of different diseases. miRNA are emerging biomarkers that were discovered within the past thirty years and have dysregulation patterns that are implicated in a wide variety of diseases. The aim of this thesis is to enable hydrogel-based point-of-care detection of miRNA and proteins by developing and applying theory to aid in rational design of the bioassay. First, we establish a theoretical framework to investigate the key factors that influence bioassay signal for hydrogel-based rapid bioassays. By developing scaling arguments for the flux of target into the hydrogel, we find that the key factors that influence bioassay signal are the reaction rate constant, the diffusion coefficient of the target in the gel, the probe concentration, the target concentration, the assay time, and the shape of the hydrogel. By changing the hydrogel particle shape, we are able to decrease the limit of detection of a protein assay by a factor of six. We then apply the theory we developed for hydrogel signal to an assay for microRNA. Using the theory, we are able to design the hydrogels to enable muultiplexed detection of miRNA directly from serum in a 40-minute assay, with a clinically-relevant limit of detection. This assay only requires minimal preprocessing of the serum, making it useful for point-of-care applications. Leveraging our theoretical knowledge, we also develop a new assay format by incorporating hydrogels into fibrous substrates such as nitrocellulose, glass fiber membranes, and silk fabric and demonstrating their utility

3 for bioassays. We demonstrate that these constructs can be used for detection of both miRNA and proteins. This work combines the fields of flexible fibrous materials and lithographic patterning to directly pattern hydrogels of varying shape and function within fibrous substrates. The work presented in this thesis demonstrates the utility of hydrogels for point-of- care applications. We believe that this work can be leveraged in the future to develop tests for additional biomarkers and can be combined with advances in fluorescence imaging and portable heating to create point-of-care devices that can quickly and reliably quantify proteins and miRNA from complex samples, in order to enable earlier diagnosis of disease.

Thesis Supervisor: Patrick S. Doyle Title: Robert T. Haslam (1911) Professor of Chemical Engineering

4 Acknowledgments

A PhD is a very long road and I am grateful for the many people whose support has helped bring me to this point. First of all, I would like to thank my advisor, Professor Patrick Doyle, whose intelligence and piercing questions shaped who I am as a researcher. Pat, thanks for always asking me hard questions and for pushing me to be the best researcher I could be. Thank you also for building a supportive lab culture; it has been an honor to work in your lab for the past four and a half years and I count many of my labmates among my close friends. Next, I would like to thank my thesis committee members: Professor Hadley Sikes and Professor Joel Voldman. Hadley, thanks for always being supportive of my career development and for asking insightful questions about the applications of the technology I was building. Joel, thanks for always being excited about the research I have done and for pushing me to set concrete goals for completion. The questions you both asked during committee meetings made this thesis better and your support helped me navigate the challenges of graduate school.

To the Doyle Lab: Thanks for not only being labmates but also becoming some of my closest friends. I appreciate the way you are not only supportive of successes but are also always willing to exchange stories of failures. The lab you choose defines much of your graduate school experience and I am very fortunate to have spent the past five years in such a supportive place. In particular, I would like to thank Hyewon Lee, Beatrice Soh, Lynna Chen, Maxwell Nagarajan, Li-Chiun Cheng, Augusto Tentori, Nidhi Juthani, Alexander Klotz, Jae Jung Kim, Ankur Gupta, Vivek Narsimhan, Lillian Hsaio, Hyundo Lee, Jeremy Schieferstein, Liam Chen, George Kapellos, Yuan Tian, Trystan Domenech, Signe Lin Kuei Vehusheia, Abu Zayed Md Badruddoza, Doug Godfrin, Gaelle Le Goff, Ben Renner, Seung Goo Lee, Alona Birjiniuk, and Rathi Srinivas. A special thanks to Hyewon and Rathi for teaching me when I joined the lab.

I would like to offer a special thanks to the Tata Center at MIT that provided funding for the majority of my PhD. They provided me with a unique experience

5 to not only work on developing diagnostics but also to partner with a startup in India, Achira Labs, and travel to India many times to gain an understanding of the state of diagnostics there and work on building and testing technology. I also thank Dr. Dhananjay Dendukuri for supporting me on many visits to Achira and providing many useful suggestions as well as the many people at Achira who made my travels there so enjoyable, including Mithila Ja, Dr. Bhavna Goyal, Dr. Purbasha Halder, Jayeeta Pai, Dr. Lokanathan Arcot, Raghavendra Katti, Gokul Rajamanickam, Tr- isha Manna, Rakesh Sharma, Dr. Satish Kalme, Sakul, Chethana, and Damu among others. Everyone at Achira was very hospitable during my visits and answered my frequent questions.

I would like to thank my family and friends for being so supportive during my time in graduate school. Thanks for always being there when I needed to laugh, cry, or vent about the stressors of graduate school. Thanks also to all my colleagues on the Graduate Student Advisory Board and the Graduate Student Council for working with me to try to make the graduate student experience better. And a thank you to the Chemical Engineering Communication Lab for being great colleagues and helping me improve my communication skills. I would like to thank my practice school buddies for remaining close friends throughout the graduate school experience: J. C., T.-C. (J.) C., S. D., Y. K., M. L., Y. M., Y. R., L. V., and S. V. Particular thanks to the many friends who have helped me destress over the past year, including Teresa and Chad Ratashak (and George!), Jennifer Herrmann, Lisa Volpatti, Mark Goldman, Stephanie Doong, and Yamini Krishnan. A special thanks to Barbara and Tom Herrmann, who have become a second family to me. Finally, I would like to thank my family, who have been incredibly supportive my entire time in graduate school. To my parents, Kendall and LinMarie Stephenson: Dad, thanks for encouraging a love of science and math at a young age and for being the best teacher I have ever had. Mom, thanks for being a female engineer role model for me and so many others. Thank you both for always encouraging me to pursue my dreams, even when it took me far away from you. I am grateful to my sister Rebecca Kaufman and her husband Ryan for always being willing to listen and for encouraging

6 me to keep going when life was difficult. To Nicole Stephens, thanks for letting me invade your space so many times and for being a patient listener. To my brother, Matthew Stephenson, thanks for the many long phone calls where you listened to me vent, for letting me crash at your apartment so many times, and for distracting me with hiking and camping trips. "The one who called you is faithful, and he will do it." SDG

7 8 Contents

1 Introduction 21 1.1 Biomolecules ofinterest ...... 22 1.1.1 P roteins ...... 22 1.1.2 microRNA ...... 22

1.2 Point-of-care detection methods ...... 24

1.2.1 Point-of-Care Detection of Proteins ...... 26 1.2.2 Point-of-Care Detection of microRNA ...... 27

1.3 Hydrogel-based detection of biomolecules ...... 29

1.4 Thesis Organization ...... 30

2 Materials and Methods 33 2.1 M aterials ...... 33 2.1.1 B uffers ...... 33

2.1.2 Reagents for monomers ...... 34 2.2 M ethods ...... 34 2.2.1 Making microfluidic channels ...... 34 2.2.2 Projection lithography ...... 35 2.2.3 Imaging ...... 35 2.2.4 Determining limit of detection ...... 35

3 Design of Hydrogel Particle Morphology for Rapid Bioassays 39

3.1 Introduction ...... 39 3.2 Experimental ...... 42

9 3.2.1 Particle Synthesis ...... 42 3.2.2 Antibody Functionalization ...... 43 3.2.3 Hydrogel Optimization ...... 43 3.2.4 Detection of Thyroid-Stimulating Hormone ...... 44 3.3 Results and Discussion ...... 45 3.3.1 Analytical Modeling of Bioassay Signal ...... 45 3.3.2 Experimentally Manipulating Hydrogel Imaging Area . .. 53 3.3.3 Application to Detection of TSH 54 3.4 Conclusions ...... 57

4 Rapid, Multiplex, On-chip Detection of microRNA Directly from Serum 59 4.1 Introduction ...... 5 9 4.2 Experimental ...... 62 4.2.1 Channel preparation ...... 62 4.2.2 Hydrogel polymerization ...... 63 4.2.3 Channel pretreatment ...... 63 4.2.4 Detection of miRNA ...... 64 4.3 Results and discussion ...... 65 4.3.1 Assay design ...... 65 4.3.2 Hydrogel optimization ...... 68 4.3.3 Assay optimization ...... 70 4.3.4 Calibration curves for miRNA detection ...... 71 4.3.5 Detection in serum ...... 72 4.4 Conclusions ...... 75

5 Incorporation of Hydrogels into Fibrous Substrates for Quantitative, Multiplexed Bioassays 77 5.1 Introduction ...... 77 5.2 Results and Discussion ...... 79 5.2.1 Patterning in substrates ...... 81

10 5.2.2 Hydrogel Motifs in Glass Fiber ...... 82 5.2.3 Optimization of Hydrogel-Functionalized Glass Fiber Substrates for B ioassays ...... 84 5.2.4 Detection of Biomolecules ...... 86 5.3 C onclusion ...... 89

5.4 Experimental ...... 90

5.4.1 Substrates ...... 90

5.4.2 Monomer preparation ...... 90

5.4.3 Patterning hydrogels in fibrous substrates ..... 91 5.4.4 Biotin-streptavidin assay ...... 92 5.4.5 miRNA assay ...... 92

5.4.6 Assay for thyroid-stimulating hormone . . 93

5.4.7 Im aging ...... 94

6 Conclusions and Outlook 95 6.1 Conclusions ...... 95 6.2 Future Work ...... 96

6.2.1 Design of Hydrogel Morphology ...... 96

6.2.2 Point-of-Care Detection of miRNA ...... 97

6.2.3 Integration of Hydrogels into Fibrous Substrates . 98

6.3 O utlook ...... 99

A Supplementary Information: Design of Hydrogel Particle Morphol- ogy for Rapid Bioassays 101 A.1 Estimating the Forward Reaction Rate Constant ...... 103 A.2 Estimating the Target Depletion ...... 105 A.3 Varying Hydrogel Height ...... 106 A.4 Signal from Particles with Overlapping Boundary Layers ...... 106 A.5 Wheel Particles with More Uniform Signal ...... 110

11 B Supplementary Information: Rapid, Multiplex, On-chip Detection of miRNA Directly from Serum 113 B.1 Nucleic Acid Sequences ...... 115 B.2 Validation of multiplexing ...... 115 B.3 Validation of buffer ...... 116 B.4 Comparison of assay conditions ...... 117

C Supplementary Information: Incorporation of Hydrogels into Fi- brous Substrates for Quantitative, Multiplexed Bioassays 119 C.1 Comparison of light attenuation through all three substrates ..... 120 C.2 Comparison of signal within and outside of glass fiber ...... 120 C.3 Exposure time studies ...... 121 C.4 Information on silk used in this study ...... 122 C.5 Discussion of limit of detection ...... 122 C.6 Nucleic acid sequences ...... 123

12 List of Figures

1-1 Previous studies for point-of-care protein detection...... 26 1-2 Previous studies for point-of-care miRNA detection ...... 28 1-3 Previous studies for biomarker detection in hydrogels include detection of miRNA[1], DNA[2], proteins[31, and small molecules[4]...... 31

2-1 Determination of limit of detection. A line is fit to the data and extrapolated to the point at which the net signal-to-noise ratio is equal to three...... 37

3-1 Fabrication of hydrogels and bioassay procedures...... 46 3-2 Hydrogel signal variation as a function of Damkihler number ..... 48 3-3 Signal variation with probe concentration...... 50 3-4 Hydrogel signal increase with internal boundary layers ...... 55 3-5 Calibration curves for detection of thyroid-stimulating hormone (TSH). 56

4-1 (A) Fabrication of hydrogel posts using projection lithography within a microfluidic channel. (B) Schematic of miRNA reaction. First, the miRNA target hybridized to the DNA probe covalently incorporated into the hydrogel. Next, a biotinylated universal linker was ligated to the hybridized miRNA for 10 minutes at room temperature. Finally, fluorescent streptavidin-phycoerythrin binds to the biotin for visualization. (C) Brightfield image of 10 pm radius miR-21 posts after reaction with serum. Scale bar is 20 tm. (D) Fluorescent image of posts in (C). Scale bar is 20 pm ...... 61

13 4-2 Optimization of hydrogels for point-of-care detection of miRNA. (A) Signal from assay with synthetic miR-21 as a function of exposure time used to polymerize hydrogels. Each data point is the average of 3-5 hydrogel posts and error bars represent one standard deviation. (B) Signal from assay with miR-21 as a function of the radius of the hydrogel post. The height of each post is 50 pn. Each point is the average of 4-5 hydrogel posts and error bars represent one standard deviation. (C) Signal from assay with miR-21 with different probe concentrations in the monomer. Each data point is the average of five hydrogel posts and error bars represent standard deviations...... 66

4-3 Optimization of assay for point-of-care miRNA detection. (A) Signal from miR-21 assay as a function of ligation time. Each data point is the average from five hydrogel posts and error bars represent one standard deviation. (B) Signal from miR-21 assay as a function of labeling time with streptavidin-phycoerythrin. Each point is the average from four hydrogel posts and error bars represent one standard deviation. ... 69

4-4 Calibration curves for detection of miRNA. (A) Calibration curve for detection of miR-21. Each point represents the average of five posts. Error bars represent one standard deviation. (B) Calibration curve for detection of miR-451. Each point represents the average of five posts and error bars represent one standard deviation. R 2 >0.99 for both curves. LOD indicates limit of detection. SNR indicates signal to noise ratio...... 72

4-5 Detection of miRNA in serum. Detection of two endogenous miRNAs: miR-451 and miR-21 and a negative control (cel-miR-54) and positive control (cel-miR-39). The posive control was spiked into the reaction mixture at 3.2 pM. Each point is the average of 3-5 hydrogel posts and error bars represent one standard deviation...... 73

14 5-1 Methods for bioassays in fibrous substrates. (A) Fabrication of hydrogels in fibrous substrates. (B) Image of hydrogel particle fabricated in glass fiber. Scale bar is 100 pm. (C) Detection of proteins on hydrogels in glass fiber. The first step is target capture for 15 minutes, followed by labeling with detection antibody for 5 minutes. (D) Detection of miRNA on hydrogels in glass fiber. The first step is hybridization of the miRNA target for 90 minutes at 55 °C, followed by ligation with a universal linker sequence for 30 minutes at 21.5 °C and labeling with streptavidin-phycoerythrin for 45 minutes at 21.5 °C...... 80 5-2 In situ fabrication of hydrogel particles in various substrates...... 83

5-3 Fabrication of different hydrogel motifs in glass fiber substrates. .. . 85 5-4 Optimization of bioassay in glass fiber using biotin-functionalized particles ...... 87 5-5 Detection of miRNA on hydrogels in glass fiber...... 88

A-1 Determination of ka. (A) Experimental setup. Streptavidin-phycoerythrin was added to a microfluidic channel containing 10 pm hydrogel posts with a biotinylated probe. (B) Plot of the ratio of the hydrogel post fluorescence to the channel fluorescence as a function of time. A rate constant was extracted from the linear region of the graph. The lines show linear fits for each post...... 104 A-2 The mean signal increases with increasing ratio of the height to the radius. Data shows mean signal from four biotinylated particles of varying heights after reaction for 15 minutes with 50 pg/L streptavidin- phycoerythrin. The radius of all particles was 100 pm. Error bars represent standard deviations...... 107 A-3 Average signal in the particle increases with decreased radius in the 1-D case (theory) and the 3-D case (COMSOL Multiphysics simulations). At low DaR, there is lower marginal signal increase from reducing the radius...... 107

15 A-4 Experimental signal from posts with overlapping boundary layers (A) Average signal in the hydrogel post increases with decreased radius. (B) Fluorescent images of the posts that visually show the overlapping boundary layers. Scale bar is 20 pm...... 110

A-5 Wheel particles with different monomer composition. The images were thresholded to different values to better illustrate the degree of signal uniformity for each particle. (A) Hydrogel particle with original composition after reaction with 10 pIU/mL TSH. (B) Hydrogel particle with reduced polyethylene glycol after reaction with 2.5 AU/mL TSH. Scale bar is 50 pm...... 111

B-1 Comparison of hydrogel signal from different sets of hydrogel posts. n=5 posts for each data point and error bars represent one standard deviation...... 116

B-2 Comparison of hydrogel signal from serum assay and neat buffer. . .. 117

B-3 Comparison of hydrogel signal from different assay conditions. Soak refers to leaving the channel in 1X TE buffer prior to the assay. "Pre- incubation with Proteinase K" refers to incubating the serum with Proteinase K prior to beginning the assay. Each data point is an average of 4-5 hydroogel posts and error bars represent one standard deviation...... 118

C-1 Setup for measuring light intensity ...... 121

C-2 Analysis of signal in biotinylated particles. (A) Fluorescent image of hydrogel particle overhanging the edge of glass fiber substrate. Scale bar is 100 pm. Yellow line indicates location of cross cut for fluorescence analysis. (B) Plot of fluorescence at the location indicated by the line in (A). The signal is similar within and outside of the glass fiber substrate...... 121

16 C-3 Fluorescence image of glass fiber substrate with biotin-functionalized hydrogels after reaction with streptavidin-phycoerythrin. The hydrogels were polymerized at different exposure times (100-500 ms). .. . 122 C-4 Label from silk used in this study ...... 123

17 18 List of Tables

2.1 List of buffers used in this thesis ...... 33 2.2 List of monomer components used in this thesis ...... 34 2.3 List of fluorescent filters ...... 36

A.1 Monomer composition for ka studies ...... 103 A.2 Parameters for 1-D calculations and 3-D COMSOL simulations . ... 108 A.3 Parameters for calculating Damk6hler number for posts with overlapping boundary layers ...... 109 A.4 Monomer composition for TSH wheel particles ...... 110

B.1 Nucleic acid sequences. /5Acryd/ indicates a 5' Acrydite modification, /3InvdT/ indicates a 3' inverted dT modification, /5Phos/ indicates a 5' phosphorylation, and /3Bio/ indicates a 3' biotin modification ... 115

C.1 Nucleic acid sequences used in this study ...... 124

19 20 Chapter 1

Introduction

As the quality of medical care has improved and the ability to treat specific diseases

is increasing, diagnosis is becoming increasingly important to reduce the burden of

disease. Point-of-care diagnostics are becoming increasingly popular as a method

of reducing the time to diagnosis. This is particularly important in situations where

time-to-diagnosis is critical for patient outcomes. As an example, incorporating point-

of-care tests in emergency rooms can reduce turnaround time by around an hour[5].

However, there is interest not only in improved diagnostic devices for use in hospitals

in the developed world but also in developing robust devices that can be used in

doctors offices, rural clinics, and by individual patients. More than half of the world's

population lacks sufficient access to healthcare[6]. While this lack encompasses more

than just diagnostics, noncommunicable diseases are becoming an increasing fraction

of the global disease burden and disproportionately affect low income countries[7,

8]. These diseases require frequent monitoring and early diagnosis improves disease

outcomes, indicating a need for improved diagnostics. Here, we focus particularly on

developing protein and miRNA point-of-care detection methods.

In this thesis, we design hydrogels for rapid bioassays, enabling detection of biomarkers at point-of-care relevant timescales. First, we use transport theory to understand the key factors that influence hydrogel bioassay signal and develop a theoretical framework that can be applied to different bioassays. Then, we the theory we developed to optimize an assay for microRNA, reducing the total assay time by a

21 factor of five to enable detection from serum in 40 minutes. Finally, we use our knowledge of hydrogel design to develop a new assay format that integrates hydrogels into fibrous substrates and demonstrate its applicability for protein and microRNA-based assays.

1.1 Biomolecules of interest

For disease diagnosis, a number of different analytes are important and range from small molecules such as glucose to proteins and nucleic acids. Here, we focus on two particular analytes for detection: proteins and miRNA.

1.1.1 Proteins

Proteins are well-established biomarkers for disease diagnosis and are used extensively clinically to diagnose and monitor the progress of disease. Proteins are formed from the translation of mRNA, after it is transcribed from genomic DNA, and as a result protein concentrations are transient[9] and analysis of these concentrations gives a snapshot into current conditions or disease states. As of 2008, 109 different protein targets had been approved by the FDA for clinical assays[10]. 30% of all diagnostic tests were antibody-based assays[11] as of 2009, indicating the importance of proteins as clinical diagnostic targets. Immunoassays are also well-established clinically, with the first immunoassay reported in 1959[121. Clinically, immunoassays for protein assays are typically run on large-scale analyzers such as Roche Cobas analyzers[13]. Although these systems have high throughput, large-scale immunoassay platforms typically cost hundreds of thousands of dollars[14], making them unaffordable for small clinics in developing countries.

1.1.2 microRNA microRNA (miRNA) are short, non-coding segments of RNA that are involved in post-transcriptional gene regulation. They function primarily to negatively regulate

22 gene expression by either preventing translation of mRNA into proteins or triggering mRNA degradation through binding to the untranslated region of the mRNA[15]. Over 2500 different mature miRNAs have been identified and cataloged in miRBase, the online microRNA database[16, 17]. These short ( 22 nucleotide) RNA segments have a normal levels of expression in a healthy human being, but their expression levels change substantially in the presence of various diseases[18, 19], making them promising diagnostic targets. miRNAs were discovered relatively recently, first observed in C. elegans in 1993 as part of the lin-4 gene[20]. Although at first their discovery did not garner much interest from the scientific community, after the first miRNA was discovered in humans in 2000[21] research in the field increased substantially. The ongoing research in the field has uncovered much information about their origin, function, and diagnostic capabilities. Mature miRNAs function primarily to prevent translation of mRNAs into proteins, although in some special cases they can work to increase translation [22]. In the cytosol, the mature miRNA is coupled to the RNA-induced Silencing Complex, or RISC, which consists of Dicer, Argonaute2, and TRBP proteins in humans [23]. The miRNA can then either repress translation of its target mRNA or signal its degradation [24]. This primarily happens when nucleotides 2-7 of the 5' end of the miRNA, known as the seed sequence, bind to the 3' untranslated region of the target mRNA [25]. In general, if the full miRNA is closer to being completely complementary, the mRNA is degraded by the Argonaute protein, while if the matching is less perfect, translation is merely inhibited [26]. Outside the cell, circulating miRNA are complexed with proteins or stored in small extracellular vesicles, protecting them from RNAses and making them stable in blood and other body fluids[27, 28, 29, 30]. Because of their involvement in regulating protein expression, panels of miRNA dysregulation have been discovered in various diseases, including cancer[31], heart disease[32], and diabetes complications, including diabetic retinopathy[33]. Depend- ing on the specific disease, certain miRNAs are up- or down-regulated from their base- line levels. In cancer, the dysregulated miRNAs can have a role that is oncogenic or

23 tumor supressive[34]. Detecting these changes can be difficult because miRNA represents only a small fraction of the total RNA in a cellular sample [35]. miRNAs present promising diagnostic targets due to their stability in serum and plasma[28, 29], but also present detection challenges due to their sequence homology and short length[35]. Currently, miRNA content is quantified primarily with microarrays, sequencing, or RT-PCR[36]. These methods typically require RNA isolation prior to analysis and sometimes rely on reverse transcription into DNA before detection, which wastes precious sample and can induce sequence bias [37, 36, 38]. RNA isolation alone typically takes 3-8 hours to complete[37], resulting in lengthy protocols. Additional techniques such as northern blotting are cumbersome and low-throughput[39]. Needed are methods that can overcome the specific challenges of miRNA detection such as a wide range of abundance, high sequence homology, short length, and low concentration in raw samples[35, 39]. Recently, there has been great interest in developing improved methods of miRNA detection, including such strategies as incorporating nanoparticles or using microfluidics [40].

1.2 Point-of-care detection methods

Point-of-care technologies have become of increasing interest in recent years, in part due to technological advantages and a desire to make diagnosis more accessible[41]. The term point-of-care (POC) is broad and means different things in different settings. The most basic application is tests that require no external equipment or trained operators and can be run in patient homes or in very rural areas, a common example of which is the pregnancy test[42]. Point-of-care can also refer to testing in doctors offices or clinics or at patient bedsides in hospitals in emergency rooms[43, 44]. These applications are typically encountered in situations when the time delay of diagnosis could result in significant negative impact to the patient, such as in suspected myocardial infarction[44]. Most established assays for POC testing are either lateral flow assays or dipstick assays[45, 46]. These tests use paper for sample wicking and typically have a col-

24 orimetric readout, resulting in a qualitative or semi-quantitative readout, but can be used in very low resource settings, without requirements for electricity[47, 48]. Point-of-care tests can also be combined with a reader to facilitate more quantitative output, as in the case of glucose meters[49, 50]. There are many commercially available lateral flow tests, including applications for infectious diseases, cardiac diseases, and cancer[47, 51].

Current research for developing point-of-care technologies is focused on three main areas: microfluidic technologies, paper-based assays, and technologies that utilize cellphones[52]. Microfluidic devices have attracted great attention for point-of-care diagnostics, due to their ability to work with small fluid volumes. More than 20 companies are developing microfluidic-based systems for point-of-care diagnosis[53], with eleven FDA-approved commercially available point-of-care tests as of 2015[47]. These tests typically consist of a disposable cartridge used in conjunction with either a handheld or benchtop reader[53, 47]. These readers typically give results within minutes to an hour and commercial systems are available for detection of blood and urine chemistry, viral nucleic acids, respiratory pathogens, and blood typing[47]. One commercially available example is the i-STAT, which combines electrochemical detection with microfluidics for blood chemistry analysis[47, 54]. Paper-based microfluidics assays are also being investigated that are more complex than lateral flow assays and use some type of patterning (with wax or other materials) to direct fluid flow along the paper. Because historically paper-based tests have suffered from low limits of detection, much research focuses on improving sensitivity either through improving capture molecules[55, 56], improving reporting schemes through signal amplification[57, 58], or combining paper-microfluidics with improved detection schemes based on electrochemistry, chemiluminescence, electrochemiluminescence, or plasmonic sensors[59, 60]. Either paper or traditional microfluidics are also commonly combined with smartphone-based detection to generate portable devices for assay and analysis. As smartphones are becoming more common in the developing world, an estimated 45% of people in emerging economies own smartphones[61]. Interest has grown in using smartphones to develop point-of-care devices. The applications range

25 glucose protein

ketones nitrite

sample inlet Paper microfluldics Conventional microfluldics Martinez et al., 2010 TaJudin etal, 2013

[Ad"*~u

Microfuldics+smartphone Microfluldics+ benchtop reader Laksanasopin et al., 2015 Kalme et al., 2019

Figure 1-1: Previous studies for point-of-care protein detection include paper microfluidics[68], conventional microfluidics[69], conventional microfluidics using a smartphone attachment[62], and conventional microfluidics with a benchtop reader[70]. from using them as power sources and for data analysis[62, 63] to using them for imaging purposes in conjunction with a POC device or attachment[64, 65, 66, 67]. In the following sections, we describe previous studies for point-of-care protein and miRNA detection in more detail.

1.2.1 Point-of-Care Detection of Proteins

To make diagnostic tests more accessible, point-of-care protein tests are being developed. An early point-of-care detection device for proteins is the urinary dipstick, developed in 1957[49]. Most devices that are commercially available are lateral flow assays[11]. These assays are typically sandwich assay based, with one antibody for capture and a second conjugated to some type of visualization molecule, such as gold

26 nanoparticles for a colometric readout[48]. However, these assays suffer from low sensitivity[71]. Current research focuses on developing tests for new antigens (often using lateral flow formats)[72], and on developing more sensitive or portable tests using paper or conventional microfluidics[73]. The Whitesides group pioneered the use of patterned paper for portable bioassays for detection of a number of analytes, including proteins[74]. Paper-based microfluidics can use a variety of different capture molecules including antibodies, binding proteins, or aptamers and can utilize a number of different detection methods including optical detection, fluorescence, or chemiluminescence[71]. Recent studies have investigated improving the sensitivity through improving labeling strategies, signal amplification schemes, or improved imaging[71, 75, 76, 77]. Switching to microfluidics from laminar flow assays can reduce device-to-device variability in material properties[78]. Conventional microfluidic devices typically offer lower limits of detection than paper-based tests, due to better flow control and lower background, but often require external pumps to drive fluid flow[79]. Microfluidic strategies vary from incorporation of microarrays or functionalized beads into microfluidic channels to complex microfluidic chips that enable blood separation and protein detection on a single device[79, 69, 80, 81]. Microfluidic chips have also been incorporated into readers that can control fluid flow and analysis to allow for more automated operation[70, 62]. Figure 1-1 illustrates some previous studies. Although commercial systems exist for some targets, conventional and paper microfluidics can be leveraged to develop assays for additional disease targets.

1.2.2 Point-of-Care Detection of microRNA

Due to the relatively recent discovery of miRNAs as biomarkers, no tests are currently part of standard clinical workflows, although recently the first miRNA-based tests have been approved for theraputic management[82]. For point-of-care detection, research into miRNA sensing is considerably less advanced than proteins. Because of the low abundance of miRNA in clinical samples[29, 35], technologies are needed that can rapidly detect miRNAs at low concentrations. Additionally, since one miRNA can influence more than one mRNA, panels of multiple miRNA are typically required

27 10

4• 10

0 10 10 o102

0 100 200 300 Assay time (minutes)

Figure 1-2: Previous studies for point-of-care miRNA detection. The data for the graph was assembled from a table in a recent review[82] as well as an additional study that is relevant[83]. The y axis is a logarithmic scale to enable visualization of a broad range of limits of detection. Black dots are studies that were conducted in complex samples while grey dots detected either synthetic miRNA or extracted RNA. The dashed line indicates 100 fM, an upper limit for average concentrations in serum[29]. for disease diagnosis[35, 84, 85, 861, resulting in a need for multiplexing. Common approaches for developing point-of-care tests for miRNA have included isothermal amplification, lateral flow assays, bead-based systems, electrochemical systems, and microfluidics[821. In general, amplification strategies offer improved limits of detection but are subject to sequence bias[87 and challenging for multiplexing[82]. Lateral flow assays have been used either with or without signal amplification, but suffer from either a long assay time or high limits of detection and have only demonstrated single-plex detection[82, 88, 89]. Bead-based studies can reach good limits of detection but suffer from long assay times and often require complex equipment for readout[90, 91]. Electrochemical systems have the potential for achieving very low limits of detection, but typically require more than an hour for measurements[92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104]. One recent electrochemical study

28 demonstrated a 0.1 fM limit of detection in 10 minutes, but only demonstrated sin- gleplex detection and had a high background in serum[83]. Additionally, these studies used bulky electrochemical equipment for measurement and have not translated the detection to a point of care device. Microfluidic devices offer the opportunity to detect miRNA from small sample volumes. One recent study did report miRNA detection from serum in 30 minutes, but did not demonstrate any multiplexing capabilities or report a limit of detection[105]. For point-of-care miRNA detection, as with any bioassay, there is a tradeoff between sensitivity and reaction time, as using strategies such as rolling circle amplification can reduce the limit of detection down to the zmol range, but require many hours to run[106, 1071. Figure 1-2 shows previous studies for point-of-care miRNA detection, plotted as limit of detection vs. assay time. This graph was created from a table in a recent review of point-of-care miRNA detection[82] and we added an additional study[83]. For detection in serum, abundant miRNA concentrations average around 10-100 fM[29]. Few studies can reach a limit of detection in this range with assay times under an hour, and those that can have either not worked with complex samples or have not demonstrated multiplexing[105, 83, 108].

1.3 Hydrogel-based detection of biomolecules

Hydrogels have found large application in bioassays due to their nonfouling nature in complex samples and their ability to be easily functionalized with capture molecules[109. These characteristics enable low limits of detection from very complex samples such as serum, raw cell lysate, and formalin-fixed paraffin-embedded tissue[106, 110, 111, 112]. Hydrogels have advantages over planar assays due to their 3D structure, which results in higher effective surface concentrations of functional groups[113]. Several different hydrogel chemistries have been applied for biomolecule detection including polyacrylamide[116, 117,118] and polyethylene glycol diacrylate[1, 3, 114, 115].In these varying chemistries, hydrogels are able to be functionalized with biomolecules, either by covalent incorporation during polymerization, by entrapment,

29 or by post-functionalization after fabrication[119, 120, 121, 122, 115]. Technologies such as stop flow lithography[123], micromolding[124], and 3D printing[125] enable precise control over hydrogel shape during fabrication. Additionally, hydrogels can be combined with microfluidics for detection either by insertion into microfluidic channels for amplification or detection purposes[126, 701 or by functionalization of the channels to enable attachment during polymerization[127, 128]. Previously, hydrogels have been applied to detect a number of different analytes[109], including mRNA[121], miRNA[1], proteins[129, 3], small molecules[4], and DNA[130]. These assays typically suffer from long assay times, as is shown in Figure 1-3. Although hydrogels have demonstrated effective detection of many types of biomolecules in complex samples, optimization is needed to reduce the assay times to lengths relevant for point-of-care applications.

1.4 Thesis Organization

This thesis develops hydrogel-based point-of-care detection methods for miRNA and proteins and is organized in the following manner: Chapter 1 provides a general introduction to the thesis, including an introduction to proteins and miRNA, a description of methods of point-of-care detection, and an introduction to the use of hydrogels in biosensing applications. Chapter 2 describes the methods that are shared between the studies in this thesis, including projection lithography, and a list of common buffers and reagents. Chapter 3 explores how hydrogel shape influences bioassay signal from point- of-care bioassays. Using a transport-based theoretical analysis, we develop scaling arguments for the key factors that influence hydrogel signal and manipulate the particle shape to decrease the limit of detection of a protein assay by a factor of six. Chapter 4 uses the scaling arguments developed in Chapter 3 to optimize an assay for point-of-care microRNA detection. In the optimized assay, we are able to obtain multiplexed microRNA concentration from human serum from a 40-minute assay.

30 .IRNA de ctio AcryiI" nONA

i-iW' a Rat

DNA detection Chapin et al., 2011 Baeissa et al., 2010 >165 minute assay >120 minute assay

Unce -nY (10~ S

PS-A Apt-linker * 9uco AuNPs IPs-a complex

ProteIn detection Small molecule detection Appleyard et aL,2011 Ma et al., 2018 >210 minute assay >100 minute assay

Figure 1-3: Previous studies for biomarker detection in hydrogels include detection of miRNA[1], DNA[2], proteins[3], and small molecules[4].

31 Chapter 5 details the incorporation of hydrogels into fibrous substrates such as glass fiber, nitrocellulose, and silk for bioassays. Fibrous substrates are commonly used for lateral flow assays and other paper microfluidic point-of-care systems. Hydrogel-functionalized substrates combine the advantages of fibrous substrates such as moisture wicking and spatial encoding with the easy functionality of hydrogel posts. We demonstrate the use of these substrates for both protein and microRNA detection. Chapter 6 summarizes the work completed in this thesis and provides an outlook on future directions of the field.

32 Chapter 2

Materials and Methods

This chapter contains common materials and methods utilized in the various chapters of this thesis. Each chapter also contains a brief description of experimental procedures, but this chapter provides more detail on common reagents and techniques.

2.1 Materials

2.1.1 Buffers

For running the bioassays described in this thesis, many buffers were required, which are listed in Table 2.1.

Table 2.1: List of buffers used in this thesis Name Composition PBS Phosphate-buffered saline (Corning) PBST PBS with 0.1% Tween-20 PBST-PEG PBST with 1% polyethylene glycol, MW 400 (Sigma) 1X TE 10 mM Tris, 1 mM EDTA, purchased at OOX (EMD Millipore) IX TET IX TE with 0.05% Tween-20 (Sigma) Rinse buffer IX TET wtih 50 mM sodium chloride Hybridization IX TET wtih 350 mM sodium chloride buffer

33 Table 2.2: List of monomer components used in this thesis Component Polyethylene glycol diacrylate MW 700 (Sigma) Polyethylene glycol MW 200 (Sigma) Polyethylene glycol MW 600 (Sigma) 2-hydroxy-2-methylpropiophenone (Darocur 1173, Sigma) 3X TE (purchased at 10OX from EMD Millipore) PBS (Corning)

2.1.2 Reagents for monomers

All monomers used to polymerize the hydrogels in this thesis were polyethylene glycol diacrylate (PEGDA) based. These monomers also contained unfunctionalized polyethylene glycol (PEG) as a porogen and used 2-hydroxy-2-methylpropiophenone (also known as Darocur-1173, Sigma) as a photoinitiator. The specific monomer composition(s) used is listed in each chapter. Table 2.2 lists the various components.

2.2 Methods

2.2.1 Making microfluidic channels

We used two different types of microfluidic channels in this thesis: glass channels and polydimethyl siloxane (PDMS) channels. The glass channels were purchased commercially from Hilgenberg GmbH and functionalized to enable hydrogel attachment, as is described in more detail in Chapter 4. To fabricate PDMS channels, we used silicon wafers with SU-8 photoresist as a mold. To make the PDMS, we mixed a 10:1 ratio of base to curing agent (Sylgard 184 Silicone Elastomer Kit, Dow Corning Corporation) until frothy, then allowed to rest for at least 45 minutes until all bubbles had disappeared. We then poured the PDMS over the mold and placed in a 65 °C oven overnight to cure. After curing, we cut the PDMS from the mold, punched inlet and outlet holes through the PDMS, and bonded to a partially cured PDMS-coated slide. To make the slide, we placed a drop of uncured PDMS on a cover slip (24 x 60 mm, # 1,5 Microscope Cover Glass, VWR) and sheared with a second cover slip

34 until the surface visually appeared uniform. We then placed in a 65 °C oven for 18-20 minutes to partially cure. We then placed the prepared PDMS channel on top of the partially-cured slide and allowed to rest for 5 minutes before returning to the 65 °C oven for overnight curing.

2.2.2 Projection lithography

The hydrogels in this thesis were polymerized through projection lithography. We ordered masks from FineLine Imaging (Colorado Springs, CO) and placed them in the field stop of a Zeiss Axio Observer inverted microscope. We used a 20x objective for all polymerization. To polymerize the monomer solution, UV light from a UV LED (ThorLabs M365L2-C4) was directed through the photomask, a UV filter set (Chroma Technology 11000v3), and the 20x objective to polymerize the hydrogels. Due to the microscope optics, features on the masks were 7.8 times larger than the final intended size of the gel. The exposure time was controlled by a custom Python script and a Vincent Associates Uniblitz VCM-D1 shutter driver.

2.2.3 Imaging

We imaged hydrogels on a Zeiss Axio Observer Al microscope. All fluorescent images utilized an Excelitas Technologies X-CITE LED120 as a fluorescent light source. The microscope was equipped with several fluorescent filter sets from Omega Optical and Semrock that enabled imaging at various excitation and emission wavelengths. The specific filter sets used in this thesis are listed in Table 2.3. We acquired grayscale images for quantitative purposes using an Andor Clara CCD camera and Andor SO- LIS software. The microscope was also equipped with a Nikon D7000 camera and Nikon Camera Control Pro 2 software for acquiring color images.

2.2.4 Determining limit of detection

The limit of detection (LOD) is defined as the point where the signal was three standard deviations above the control signal. To calculate this point, we plotted the

35 Table 2.3: List of fluorescent filters Manufacturer Filter set Semrock FF660-DiOl filter (dichroic) FF01-628/40 filter (exitation) FF01-692/40 filter (emission) Omega Optical XF101-2 filter set Omega Optical XF100-3 filter set

net signal to noise ratio (s/n) as a function of concentration to create a calibration curve and fit a line to the data, as shown in Figure 2-1. The net signal is defined as the difference between the signal at a given concentration and the control signal (no target added). The noise is the standard deviation of the control signal. The fitted line was then extrapolated to find the concentration at which the net signal to noise ratio was equal to 3. The calculation for the net signal-to-noise ratio is shown in Equation 2.1, where signal, is the signal at concentration "x" and o- represents the standard deviation.

s/n = signal, - signalcontrol (2.1) (7control

36 0

C 0 (U

Concentration LOD

Figure 2-1: Determination of limit of detection. A line is fit to the data and extrapolated to the point at which the net signal-to-noise ratio is equal to three.

37 38 Chapter 3

Design of Hydrogel Particle Morphology for Rapid Bioassays

The goal of this thesis was to develop guidelines for hydrogel design to enable point-of- care detection of biomolecules and to apply the framework to make new point-of-care tests. In this chapter, we develop a theoretical framework for hydrogel signal from a rapid bioassay and use the framework we developed to improve the limit of detection for a protein-based assay by a factor of six. This chapter is reproduced with permission from Sarah J. Shapiro, Dhananjay Dendukuri, and Patrick S. Doyle. Design of Hy- drogel Particle Morphology for Rapid Bioassays. Anal. Chem., 90(22):13572-13579, November 2018. Copyright 2018 American Chemical Society.

3.1 Introduction

Hydrogel microparticles have found widespread applications in biosensing[109]. Al- though hydrogels are typically biologically inert and non-fouling in the presence of biological materials, they can be functionalized with reactive molecules to enable their use as capture or release agents for diagnostic or therapeutic applications. The wide variety of materials for hydrogels[131, 132, 133] and of methods of hydrogel functionalization[134, 135, 136]has ensured their versatility for a wide range of applications. These particles vary in size from a few to hundreds of microns[137] and the

39 advent of technologies such as stop-flow lithography (SFL)[123], micromolding[138], and 3D printing[125] enable facile control over hydrogel shape and size in addition to function. The versatility of hydrogels has previously been utilized to develop point-of-care (POC) diagnostics[139, 140, 141] in an effort to make diagnostic tests more accessible. POC diagnostics enable diagnosis of disease wherever care is delivered to patients, including doctor's clinics, patient bedsides, and in patient's homes, facilitating more rapid treatment[142]. In contrast, sending samples to a centralized laboratory results in delays of at least 1-2 hours in European countries, typically requiring non-emergency patients to receive their test results at a later time or date[143]. In- corporating POC testing was previously shown to reduce mean test turn-around time from 59 minutes to 8 minutes[144]. In developing countries, POC testing has the potential to increase test accessibility more drastically, as only an estimated 28% of the population in Africa has access to advanced health care facilities[145]. Implementing POC testing is particularly important in regions where centralized laboratories are less prevalent or in acute conditions where time-to-diagnosis is critical[146, 147]. Protein-based assays are extremely important in diagnostics, with more than 200 different protein targets from serum or plasma analyzed in clinical laboratories, an estimated 10% of all proteins known to exist in plasma[148, 149]. Among immunoassays, tests for thyroid function are widely utilized, as thyroid diseases affect around 200 million people worldwide[150]. Because of the function of thyroid-stimulating hormone (TSH) in regulating thryroid hormone levels, it is used as a diagnostic marker either alone or in panels for a wide variety of thyroid disorders[151, 152]. TSH is typically measured in central laboratories using large-scale analyzers[153]. In countries where centralized laboratories are less prevalent, this can lead to reduced test availability and a lack of diagnoses. In India, for example, an estimated 42 million people are suffering from thyroid disease, with more than 10% of the adult population suffering from hypothyroidism, around a third of whom are undiagnosed [154, 155, 156]. Although hydrogels have been used extensively and there is a large body of lit- erature on their formulation and applications, little is known about the influence

40 of hydrogel shape on diagnostics. Previously, hydrogel shapes have been used as a method of enabling multiplexing [157, 158, 159], for enhancing cell capture[119], for increasing cellular uptake[160], and for generating vasculature in tissue engineering scaffolds[161]. Previous studies have also analyzed the time-dependent binding of target molecules in hydrogels for biosensing applications[162, 113]. However, these studies assumed a specific shape profile for the hydrogel and did not explore variations. To our knowledge, no studies have analyzed the impact of hydrogel particle morphology on bioassay signal. Our study uses poly(ethylene glycol) microparticles to explore the influence of shape on the fluorescent signal for point-of-care bioassays.

We explored methods of improving the limit of detection of point-of-care bioassays through manipulating the hydrogel shape. We observed that often, although the 3-D hydrogels are porous to enable target capture, due to the relative rates of reaction and diffusion only the edges of the hydrogel are utilized for binding[162, 163]. We determined that in order to maximize the fluorescent signal, it is optimal to maximize both diffusion and reaction rates while decreasing the cross-sectional area of the hydrogel parallel to the imaging plane to concentrate the signal in a small area. By calculating the flux of target into the hydrogel, we were able to obtain estimates of the final signal per unit imaging area at the end of the bioassay. To validate our theory, we performed experiments wherein we used projection lithography to form hydrogel microparticles with varying surface area and analyzed the resulting signal in the hydrogel microparticles after labeling a biotin probe with a fluorescent streptavidin molecule. By using ring structures instead of disk shapes, we were able to add additional surface area on the inside of the microparticle, thereby increasing signal in the bioassay. Finally, we applied the method to reduce the limit of detection in a protein-based assay for thyroid-stimulating hormone (TSH) by a factor of 6. In this study, we achieve a limit of detection of 0.056 pIU/mL, below the 0.1 pIU/mL threshold that distinguishes between grade 1 and grade 2 hyperthyroidism164], with a short 15.5 minute target incubation, surpassing second-generation TSH tests[165].

41 3.2 Experimental

3.2.1 Particle Synthesis

Hydrogel particles were synthesized using a variation of stop-flow lithography[135, 123]. The precursor solution was formed from a stock monomer solution consisting of 20% polyethylene glycol diacrylate (PEG-DA) 700, 40% polyethylene glycol 200, 5% Darocur 1173 photoinitiator (all from Sigma-Aldrich), and 35% buffer. For studies with biotinylated probe, 3X Tris-EDTA buffer (30 mM Tris, 3 mM EDTA) was used (EMD Millipore) while for antibody incorporation, 1X phosphate buffered saline (PBS) buffer was used (Corning). The stock monomer solution was diluted 9:1 with probe or antibody solution. The biotinylated probe consisted of a DNA strand (sequence: 5' ATA GCA GAT CAG CAG CCA GA 3') with an Acrydite modification on the 5' end and a biotin group on the 3' end (Integrated DNA Technologies). This probe was used in concentrations of 10 IM, 5 pM, 0.5 pM, or 0.05 pM in the prepolymer solution. For antibody-functionalized particles, anti-TSH 10-T25C antibody was purchased from Fitzgerald (Acton, MA) and conjugated to an acrylate group as described below for covalent incorporation into the hydrogels. The antibody was used at 0.4 mg/mL in the monomer solution. The final monomer solution was vortexed for 30 seconds before addition to a microfluidic chamber. Microfluidic channels 150 pm tall and 450 pm wide were constructed of polydimethyl siloxane (PDMS). PDMS was mixed in a 10:1 ratio of base to curing agent and poured over a silicon wafer mold created from SU-8 photoresist. The mixture was allowed to rest for 45 minutes to remove bubbles, then placed in a 65 °C oven overnight. To construct microfluidic channels, a small drop of PDMS was sheared between two glass coverslips to create a thin layer. The coverslips were then partially cured for 18 minutes at 65 °C, then removed from the oven and a PDMS channel from the silicon wafer mold was placed on top of the slide. The curing process was completed overnight at 65 C. Hydrogel particles were formed through a variant of SFL[123, 135] on a Zeiss Axio Observer microscope as shown in Figure 3-1A. A mylar mask (Fineline Imaging) was

42 placed in the field stop of the microscope and UV light from a UV LED (M365L2-C4, ThorLabs) illuminated the monomer mixture for 100 ms in the microfluidic channel through a UV filter set (11000v2, Chroma Technology) using a 20x objective. In order to maximize the number of particles per batch of antibody, after polymerization, rather than flushing the channel with fluid, the channel was moved to a different region to enable the formation of additional particles. The particles were subsequently flushed from the channel and washed five times. For biotinylated particles, the particles were washed and resuspended in IX Tris-EDTA buffer with 0.05% Tween-20 (1X TET buffer) and for antibody-functionalized particles, the particles were washed and resuspended in PBS with 0.1% Tween-20 (PBST buffer).

3.2.2 Antibody Functionalization

Capture antibodies from Fitzgerald (Acton, MA) were washed three times in PBS using centrifugal filters from EMD Millipore. The antibody concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific) and adjusted to 7.14 mg/mL in the final reaction mixture, which also contained 1.67 mM NHS-PEG- Acrylate (Creative PEGworks) in PBS. After a 1 hour incubation at room temperature on a tube rotator (MX-RD-E, SCILOGEX), tris-HCl solution was added from a 100 mM stock to bring the final concentration to 16.6 mM to quench the reaction. After quenching for 30 minutes at room temperature on the rotating mixer, the antibody was washed four times in PBS for 5 minutes at 14,000g. The final antibody concentration was adjusted to 10 mg/mL in PBS.

3.2.3 Hydrogel Optimization

For initial Damkdhler number and shape studies, a simplified system using the reaction of streptavidin-phycoerythrin (SAPE) with biotinylated particles was utilized, as described in Figure 3-1B. Particles were added to an Eppendorf tube and the buffer was removed down to 5 pL. 50 pL of 50 pg/L SAPE was then added to each tube. SAPE was purchased from Life Technologies and diluted in 1 X TET buffer. The

43 Eppendorf tubes were incubated on a thermoshaker (MultiTherm Shaker, Thomas Scientific) at 21.5°C for 15 minutes, then washed three times with 500 pL 1X TET buffer by vortexing followed by a 2 minute centrifugation step. The particles were then placed on a glass coverslip and imaged through a 20x objective on a Zeiss Axio Observer Al microscope using a broad spectrum LED (X-CITE 120LED, Excelitas Technologies) as a fluorescence source and a XF101-2 filter set from Omega Optical. Images were acquired using Andor SOLIS software and an Andor Clara CCD camera. In most cases, 5 particles were analyzed for each condition, except when fewer particles were found after the bioassay, in which case fewer particles were analyzed. In all cases, at least 3 particles were analyzed for each data point. After imaging, the images were cropped with ImageJ (National Institutes of Health)1661 and the fluorescent signal was analyzed using custom Matlab scripts. A fluorescent image of particles after a bioassay is shown in Figure 3-1C.

3.2.4 Detection of Thyroid-Stimulating Hormone

For detection of thyroid-stimulating hormone (TSH), - 10 particles per assay condition were added to an Eppendorf tube. The assay protocol is illustrated in Fig- ure 3-1D. After removal of the buffer down to 5 pL, 50 pL of TSH standard (from Monobind TSH Accubind ELISA kit) was added. PBS was used as a negative control. 0.5 pL of 10% Tween-20 were then added to prevent particles from sticking together or to the edges of the tube. The particles were incubated with the TSH target for 15.5 minutes on a thermoshaker (MultiTherm Shaker, Thomas Scientific) at 1500 rpm and 21.5°C. Immediately after the target incubation, 500 iL of PBST were added and the particles were centrifuged for 2 minutes. After removal of the supernatant, the particles were washed 2 times by addition of 500 pL PBST, vortexing for 2 minutes, and centrifugation for 2 minutes. It was observed that longer vortexing periods could reduce the fluorescent accumulation in the negative sample (data not shown), but to minimize assay time, a 2 minute vortexing step was used. After rinsing, the supernatant was removed down to 5 pL, and 50 pL of detection antibody was added at a final concentration of 10 pg/mL. The detection antibody

44 was purchased from Biospacific (Anti-TSH 5409 SPTNE-5) and labeled with Dyomics DY647P4 NHS Ester as described below. The particles were incubated with the labeling antibody for 5 minutes on the thermoshaker at 1500 rpm and 21.5°C. After the incubation, 500 pL of PBST with 1% polyethylene glycol MW 400 (PBST-PEG) was added and the particles were centrifuged for 2 minutes. The particles were then rinsed 3 times in PBST-PEG and once in PBS before resuspension in PBST for imaging. The particles were then imaged as described previously, using a filter cube set from Semrock consisting of a FF660-DiOl filter (dichroic), a FF01-628/40 filter (exitation), and a FF01-692/40 filter (emission). The images were cropped and the wheel particle images were aligned in ImageJ then analyzed with custom Matlab scripts. The DY647P4 NHS Ester was suspended in DMSO at 10 mg/mL for long-term storage at -20 °C. The antibody was buffer-exchanged into PBS using filter columns. After buffer-exchange, the concentration was measured using a NanoDrop spectrophotometer and adjusted to 4 mg/mL using PBS. The conjugation reaction proceeded in a buffer containing 2.7 mg/mL antibody, 68.5 mM sodium bicarbonate in PBS and 0.73 mg/mL DY647P4 NHS ester, with the remaining volume made up with PBS. The reaction proceeded for one hour at room temperature on a tube rotator. After completion of the reaction, the antibody was washed six times in PBS using a filter column and the concentration was adjusted to 1.7 mg/mL using PBS.

3.3 Results and Discussion

3.3.1 Analytical Modeling of Bioassay Signal

In this study we sought to optimize hydrogel performance for point-of-care bioassays by maximizing assay signal. Hydrogels have been used extensively for bioassay applications due to their nonfouling nature and the ability to covalently incorporate biomolecules for sensing applications[109, 1671. For point-of-care (POC) diagnostics, the timing of the assay becomes critical and slow kinetics cannot be circumvented by

45 A B 10°P SA-PE $Blotlnylated5 Probe Microscope Objective Gel Matrix Labeling with M Photomask Streptavidin-Phycoerythrin 15 min, 21.5 °C

Detection IC |I D Antibody Tret

capture -

GelMatrix Target Incubation Labeling with 15.5 min, 21.5 °C Ab-Fluorophore 5 min, 21.5 °C Figure 3-1: Fabrication of hydrogels and bioassay procedures. (A) Fabrication of hydrogel particles using projection lithography in a microfluidic channel. The particles were polymerized using UV light shining through a photomask placed in the field stop of the microscope. (B) Labeling assay with streptavidin-phycoerythrin. A DNA probe with a biotin molecule was covalently incorporated into the hydrogel particles during polymerization. The biotin reacted with streptavidin-phycoerythrin (SAPE) during a 15 minute incubation at 21.5°C. (C) Fluorescent image of disk particles and ring particles with 60 pm inner radius showing particle uniformity after reaction with SAPE. Scale bar is 100 pm. (D) Sandwich assay protocol for antibody assay for protein detection. Capture antibody was immobilized within the hydrogel during polymerization. The gels were then incubated with the target antigen for 15.5 minutes at 21.5°C. After washing, the particles were labeled by incubation with a fluorophore-conjugated detection antibody for 5 minutes at 21.5C.

46 long incubation times. For this reason, detailed understanding and optimization of the hydrogel is crucial for POC applications. During a hydrogel-based bioassay, target molecules diffuse into the hydrogel and react with immobilized probe molecules, resulting in a reaction-diffusion process. Previously, reaction-diffusion mechanisms in hydrogels have been studied for long-time scale incubations for both protein and nucleic acid assays[162, 113]. For the case of POC diagnostics, the key consideration is maximizing the signal obtained from short target incubation times, as FDA-approved POC technologies take from 30 seconds to 1 hour[47]. Typically, these assays operate in a regime where the target concentration in solution is not significantly reduced over the assay timescales (calculations in SI).

We can estimate the signal from a bioassay by considering a hydrogel particle in a well-mixed tube. The goal of any POC bioassay is to maximize the signal from the target captured within the hydrogel over the short assay time, which is related to the flux of target into the hydrogel, J. The flux can be described by Fick's law[1681, where Dge is the diffusivity of the target in the hydrogel matrix and T is the target concentration:

J = -DgeVT (3.1)

Many bioassays are performed under conditions where the rate of reaction within the hydrogel is much faster than the rate of diffusion[162, 163]. The ratio of these rates is known as the Damkbhler number koPoL 2 ka (Da)[168] (Da= Dgel ). represents the forward rate constant, Po the initial probe concentration, and L the distance from the center of the particle to the edge. Under conditions with high Da, where the reaction is much faster than diffusion, the target reacts with embedded probe before reaching the center of the hydrogel, and the binding is confined to a region around the edges of the hydrogel-the so-called boundary layer. This process is illustrated in Figure 3-2A. Under these conditions, the target concentration within the hydrogel quickly decays to zero, and the flux can be approximated by Equation 3.2, where o is the boundary layer thickness and T,,O is the target concentration at the edge of the

47 A Hydrogel! Solution BC m Da=9 Da9 2.5 -Da=0.9 aDa=0.09

P 0 50 100 150 Radial Distance (pm) 0

Figure 3-2: The signal variation within the hydrogel is dependent on the Damk6hler number (Da). (A) Schematic representing reaction of target with immobilized probe within particle. At high Damkbhler number (Da), the signal is localized around the edges of the hydrogel, while at low Da the signal is nearly uniform throughout the gel. (B) Imaging of 3-D hydrogels, resulting in a 2-D projection of the 3-D signal profile. The images show actual biotinylated particles at various Da, thresholded to different values to illustrate the boundary layer profile. Scale bar is 100 pm. (C) Signal profile of biotinylated hydrogels at various Da after reaction with streptavidin- phycoerythrin for 15 minutes. The graph shows the radially-averaged signal in the hydrogels, normalized to the center of the disk. Each trace is the average of five particles. Da was modified by changing the probe concentration in the hydrogels. hydrogel:

J ~ Dge Ts(3.2)

For reaction-diffusion mechanisms with high Da, the target never reaches the center of the particle and the boundary layer thickness is dependent on the relative rates of reaction and diffusion: 6 ~ Da-2L[168]. Substituting this equation into Equation 3.2, we obtain:

1 1 1 J ~ Pol T,,o (3.3)

At high Da, the flux into the particle is independent of the particle size. However, at low Da (Da < 1), the full hydrogel is utilized for the reaction and the concentration in the center of the hydrogel is non-zero, Teenter, resulting in an estimated flux of:

J ~ Dgel T,O - Tcenter (3.4) J~D9 ~1 L

48 Comparing Equation 3.2 and Equation 3.4, the flux into a particle at high Da will always be higher than the flux into a particle at low Da, since 6 is less than L. To maximize hydrogel signal, then, a higher Da is preferred.

So far we have concerned ourselves with 1-D flux approximations, but, ultimately, we are concerned with the amount of target bound at the end of the bioassay, as measured by the signal. Bioassays often use a fluorescent signal per unit imaging area as a quantification metric, taking a 2-D image of the 3-D particle, as shown in Figure 3-2B. Da values were calculated assuming a 10% incorporation of biotinylated probe into the hydrogel[162], a forward rate constant of 10' M- 1 s-1 (measured in Appendix A), a diffusivity of 5.66 x 10-" m 2/s, (previously measured in the hydrogels for a fluorescently labeled antibody{3]), and a radius of 100 im. At high Da, the average signal is higher, but the capture is concentrated around the edges of the hydrogel particle, as is shown in Figure 3-2C.

Assuming a constant flux, and that the assay operates at high Da, the signal per unit imaging area can be calculated by multiplying the flux by the external surface area of the particle and the time, to obtain the total amount of target in the hydrogel:

Signal Areasurface Areasurface 1 1 ~ J =DSelPoka TSlot (3.5) Areaimaging Areaimaging Areaimaging

To verify Equation 3.5, we used cylindrical particles arid varied the probe concentration from 5-500 nM in the hydrogel while keeping the shape of the hydrogel and the rest of the hydrogel composition constant. The resulting signal in the hydrogel particles is plotted against the square root of the probe concentration in Figure 3-3. Probe concentrations in the hydrogel were calculated by assuming a 10% incorporation rate from the monomer concentration [1621. The plot of signal versus P2 is expected to be linear at Damkbhler numbers greater than one, since Equation 3.5 was developed by assuming the presence of a boundary layer in the flux calculation. At low Da, the flux is expected to decrease due to a non-zero target concentration in the center of the hydrogel, resulting in a lower measured signal than predicted by Equation 3.5 and a deviation from linearity. However, we observe a linear signal decrease down

49 500

400 .0 300

200

C t 100 C.) 0CD 0O C 0 2 4 6 8x10 P1" (M/) Figure 3-3: Mean background-subtracted signal increases with probe concentration in the hydrogel. The initial probe concentration was varied experimentally from 5- 500 nM in order to change the flux into the particle. The line is a linear least squares fit to the data. Each point is the mean of 3-4 particles and error bars represent standard deviations. to the lowest probe concentrations, equivalent to Da = 0.09, suggesting that at this Damkbhler number the flux has not decreased substantially below the value predicted by Equation 3.3. Based on Equation 3.5, we can determine the parameters that influence the measured bioassay signal. Typically, the target concentration and the incubation time are set by required assay parameters, placing them outside of the researcher's control. To increase the signal, it is advantageous to increase the target diffusivity, the initial probe concentration, the reaction rate, and the ratio of the surface area to the imaging area. Depending on the specific target, ka is usually set by the choice of target-probe pair, giving little control over that parameter. Po and Dei can be influenced by manipulating the hydrogel prepolymer mixture, but often these are already optimized to a large extent during assay development. They are also sometimes coupled, making them difficult to control. For example, increasing the pore size of the hydrogel can result in higher diffusivity but lower probe incorporation[162, 121]. As a result, changing the ratio of the surface area to the imaging area is the easiest

50 way to influence assay signal. This parameter varies depending on the shape of the hydrogel. In a sphere, for example, the surface area is 41R2 and the imaging area is 7rR2 , resulting in a ratio of surface area to imaging area that is independent of the size of the particle. However, with particles with more complex shapes such that the height is independent of the planar imaging area, there is an opportunity to tune the surface area to imaging area ratio in order to increase signal. In a cylinder imaged through the planar surface as in Figure 3-2B, for example, the surface area to imaging area ratio is 27rR2 +2rRH = 2(1+E). We can therefore increase the signal by increasing the ratio of the height to the radius of the cylindrical particle (demonstrated in Figure A-2 in Appendix A).

So far, we have considered non-overlapping boundary layers in the theoretical analysis. As the hydrogel particle size is decreased, however, the boundary layers will come closer together, until they eventually overlap when the distance from the center of the particle to the edge is smaller than the boundary layer. It is of interest to know what happens to the flux during this scenario. To determine that, we must develop a model for the way the signal varies within the hydrogel. For simplicity, we begin with a 1-D cylindrical system.

For the general case, the system can be described by species conservation equations, where [T] is the target concentration, [P] is the probe concentration, [TP] is the concentration of the target-probe complex, t represents time, De is the diffusivity of the target in the hydrogel, and ka is the forward reaction rate. For a typical bioassay, the forward reaction rate is much faster than the reverse reaction rate (typical antibody KDvalues are at least 10-7 s-1 )[169], so the reverse reaction is neglected for the model. In addition, typically, the probe concentration is in great excess of the target (Po - 1 p M, T,,o ~ 10-10 M), leading to a nearly constant probe concentration during the short assay time ([P] ~ Po).

[T]= DgeiV2 [T] - ka [P][T] (3.6) at

This equation can be nondimensionalized by using = ,T T= , and r=

51 -. For point-of-care bioassays, the target concentration is typically not significantly kaPo depleted during the assay, but the assay is longer than the characteristic reaction or diffusion time, leading to a pseudosteady target concentration within the hydrogel

and resulting in the following nondimensionalized equation, where DaR - k"PoR2

1 a aT 0- - DaRT (3.7) r jBi iFr With this assumption, the 1-D system can be solved for the target concentration as a function of position, using the boundary conditions

T(i = 1)= 1 (3.8)

dT (3.9) d =0) = 0 The solution to these equations for the target concentration in the hydrogel uses modified Bessel functions:

Io(Da fi) R= (3.10) Io(DaA) Ultimately, we are interested in the signal per unit imaging area, which we can calculate from the flux into the hydrogel. To find the flux into the radial face of the hydrogel, we can differentiate the dimensional version of Equation 3.10 with respect to r, in order to calculate the 1-D flux.

T ,1(Da r)Da J = Dgelr =Dgel R 1 oR (3.11) (9r Io(Dal)R We evaluated the flux at the boundary where r = R in order to calculate the expected signal per unit imaging area:

Signal 27rRHtJ Ii(Da )Dal T5,o (3.12) Areaimaging ~ r R 2 = 2 H tDgel Io(Da2 )R2

52 Equation 3.12 shows that as the radius decreases, the signal per unit imaging area increases, even in the case of overlapping boundary layers, although at very small values of DaR there is diminishing signal increase from further reduction of the radius. A plot of this equation can be viewed as Figure A-3 in Appendix A. The plot illustrates that this diminishing increase in signal occurs around DaR - 0.1. We also verified that the trend of increasing signal with decreasing radius holds in the 3-D case using COMSOL Multiphysics in Appendix A (Figure A-3). To maximize the signal measured from a bioassay it is therefore advantageous to decrease the cross-sectional area of the particle, down to at least Da = 0.1. We also verified this point experimentally by creating hydrogel posts with overlapping boundary layers (Figure S4). Because imaging particles with height larger than the diameter can prove challenging due to particle toppling, in the following sections we explore more complex shapes that reduce the imaging area of the particle while still maintaining a large outer radius.

3.3.2 Experimentally Manipulating Hydrogel Imaging Area

We sought to apply our knowledge of flux into the hydrogels to increase signal from a bioassay, by decreasing the imaging area of hydrogel particles relative to the surface area. Technologies such as stop-flow lithography enable facile manipulation of particle shape and can provide a simple route to increase signal in POC bioassays. We began with cylindrical particles and increased fluorescent signal by adding internal features to create 3-D extruded rings. This added an additional boundary layer on the inside of the ring, increasing the surface area to imaging area ratio. Images of these particles are shown in Figure 3-4A. These particles had a constant outer radius of 100 pm and an inner radius varying from 0 to 80 pm. To assess the relative fluorescence of each of the shapes, the radially-averaged signal was plotted as a function of the radial distance (Figure 3-4B). To facilitate comparison, this signal was normalized to the inner signal in the disk particles and was plotted for the entire particle, including the region in the center of the rings. As the inner radius is increased, the boundary layers on the inner and outer ring surface begin to overlap, increasing the average signal within

53 the particle. We then analyzed average signal within the particle. We predicted that some signal increase could be obtained by improved image processing through only analyzing the region on the edge of the disk particles, since doing so increases the effective ratio of surface area to imaging area during the image analysis step. For example, for a cylinder imaged through one face, the ratio of the surface area to imaging area is 2(1+ ). By eliminating the center of the cylinder from the analysis and only analyzing a ring around the outside of the cylindrical particle, the ratio increases to 2(1+ )where R is the outer radius of the cylinder andrinneis R the radius of the inner circle eliminated from the analysis. By constructing particles of more complex shapes, such as 3-D extruded rings, however, we can add an additional boundary layer around the inside of the ring, increasing the surface area to imaging area ratio to 2(1+ ,,i,). The analysis method for the disk particles is shown in Figure 3-4C. The disks were analyzed both using the full area and by analyzing a ring-shaped region around the edges of the disk, neglecting the center. The results are plotted in Figure 3-4D, comparing the ring and disk shapes. Focusing the analysis on the outer regions of the disk-shaped particles results in a small improvement to the overall signal as the inner radius is increased from 0 pm (disk) to 80 pm. However, a much greater improvement in mean signal can be obtained by creating ring particles, adding an internal boundary layer. The rings with 80 pm inner radius show a more than two-fold increase in mean signal above the signal in the disk particles. We observed occasional breakage of the 80 pm inner radius rings during the bioassay, so we did not fabricate rings with larger inner radii due to issues with structural integrity. However, we added additional inner features in the following section that could enable fabrication of thinner structures.

3.3.3 Application to Detection of TSH

We anticipated that increased signal in a bioassay would translate to reduced limit of detection (LOD), so we applied our strategy for increasing signal to POC detection of thyroid-stimulating hormone (TSH), a ~ 30 kDa serum protein. We used a sandwich assay format to detect TSH, capturing TSH from solution on immobilized

54 A B - C Full2DImage D 3 3000 -8 mInner R dus ' Disk Pa 'dess 5 -0mlnnrRdiu 2500 Rk Parboce

4pm Inner Radus 20

Rada ance 20 AnalyzedRegion Disk 40pmIR 60pmIR 0 pmIR

Figure 3-4: Adding internal boundary layers increases average signal within the hydrogels. (A) Fluorescent images of biotinylated hydrogel disk particles and rings with 40 pm,60 and80minner radiiafterreactionwith50g/Lstreptavidin- phycoerythrin.Scalebaris100 m. (B) Radially-averagedsignalinringanddisk particles, normalized tothe signal in the center of the disk, plotted as afunction of the radial distance. To obtain an accurate representation of the signal in the center, the signal in the inner10 pmnof the disk was averaged. Each line represents the average of five particles. (C) Analysis of disk particles. Either the full shape of the disk was analyzed or the analysis region was restricted to aring shape of varying inner radius. (D) Comparison of signal between disk particles and ring particles. The disk particles were analyzed either by considering the full area orby eliminating acircle of the given inner radius in the center of the disk. Each barrepresents five particles and error bars represent standard deviations. capture antibodies within the hydrogel in the first step and labeling with afluorescent detection antibody in the second step. Because the concentration of labeling protein can be increased during the assay in order to reduce incubation times, we focused on optimization of the target capture step. To dothis, we created "wheel" particles with 20 pm feature thickness. We adopted the wheel motif due to observations that the rings with 80 pm inner radius often deformed during imaging or broke during the bioassay. By adding in the cross bars, we increased the rigidity of the particles.

By using these wheel particles for detection of TSH, we both increased the fluorescent signal during the bioassay and decreased the limit of detection. Figure 3-5 shows calibration curves for the TSH assay. We constructed acalibration curve for TSH using disk particles (Figure 3-5A). The limit of detection was calculated as the point where the net signal was three times the noise (defined as the standard deviation of the control sample). The mean net signal refers tothe difference between the background-subtracted signal and the background-subtracted signal of the control particles. We anticipated that some signal increase could be obtained by improved

55 A Disk particles B Wheel particles L)D=0.21 pIlU/mL 14000 -i 5000 60 pm LOD=0.29 plU/mL I 12000 CdLOD=O.33 plU/mL ( 4000 LODO.35pIU/mL 10000

3000 8000

820000

1000 M 2000

0 5 10 5 10 TSH Concentration (plU/mL) LOD TSHConcentraton(plU/mL) 0.21-0.35 pU/mL 0.056MpU/mL

Figure 3-5: Calibration curves for detection of thyroid-stimulating hormone (TSH). (A) Calibration curve for disk particles. Each data point represents the average of 3-5 particles and error bars represent one standard deviation. Each disk was analyzed as the mean signal in either the full disk or in a ring-shaped region around the edge of the disk. Limit of detection (LOD) is calculated as the point where the signal is three standard deviations above the control. Inset figure is an image of a disk particle at 10 pAU/mL. Scale bar is 100 pm. (B) Calibration curve for wheel particles. Each data point represents the average mean signal of 3-5 particles and error bars represent one standard deviation. LOD represents point where signal is three standard deviations above control. Inset figure shows wheel particle after reaction with 10 pIU/mL TSH. Scale bar is 100 pm.

analysis methods and only utilizing the signal at the edges of the hydrogel for analysis. Reducing the analysis region to a ring with inner radius of 80 pm and outer radius of 100 pm improves the limit of detection slightly, from 0.35 to 0.21pIU/mL.

However, we anticipated that greater improvements in limit of detection could be achieved by adding internal surface area. For the "wheel" particles, the resulting limit of detection was 0.056 pAU/mL, a six-fold reduction in LOD (Figure 3-5 B). In addition to improved signal in the "wheel" particles, we anticipate that the reduced diffusion lengthscale also improves washing in the particles, contributing to the LOD improvement. We observed that the standard deviation of the control particles was lower for the wheel particles than the disk particles by about a factor of two, which we attribute to better washing. We also note that the signal in the particles has a splotchy appearance, which we attribute to antibody phase separation and can be improved by using monomer compositions with reduced PEG content (Figure S5).

56 3.4 Conclusions

We have developed a method of increasing signal for point-of-care bioassays by increasing the ratio of the surface area to the imaging area in hydrogel microparticles. By changing the particle shape, we added an additional boundary layer on the inside of the particle and increased the mean signal within the hydrogel. We derived equations to estimate the signal after a bioassay from the flux of target into the hydrogel that can provide guidelines for hydrogel shape design and we applied those guidelines to increase assay signal from a model biotin-streptavidin assay. Finally, we applied this technique to an assay for thyroid-stimulating hormone, resulting in a sixfold decrease in the assay limit of detection, simply through increasing the surface area relative to the imaging area. We anticipate that this strategy could be applied to many types of POC bioassays, enabling LOD reductions through changing particle shape, without needing to modify other assay parameters. This is particularly advantageous since often the researcher has limited control over the other parameters for POC bioassays.

57 58 Chapter 4

Rapid, Multiplex, On-chip Detection of microRNA Directly from Serum

The aim of this thesis is to enable point-of-care detection of biomolecules on hydrogels. This chapter details the application of the theory developed in Chapter 3 to a detection scheme for microRNA, enabling detection of microRNA from serum in a 40-minute assay.

4.1 Introduction microRNA (miRNA) are short, non-coding RNA that contribute to post-transcriptional regulation of gene expression[15]. Due to their dysregulation patterns in a wide variety of diseases, including heart disease[32], stroke[170], and many cancers[31], there has been recent interest in using circulating miRNA signatures as biomarkers for diagnosing disease. Additionally, because of their stability in blood and body fluids[29, 30], they present attractive diagnostic targets. However, current established methods for detecting miRNA such as Northern Blots, microarrays, and RT-PCR involve lengthy protocols and expensive equipment and typically require isolation of total RNA from the sample before beginning the assay[171, 37, 172]. Although RT-PCR is the most rapid of these options, it can be subject to sequence bias due to target amplification[173, 174]. Furthermore, total RNA isolation alone takes 3-8 hours[37],

59 resulting in a significant delay between sample draw and results.

Point-of-care technologies have shown great promise as a way of decreasing time- to-diagnosis in various settings including doctor's offices, emergency rooms, and in patient homes[147, 144, 175]. In emergency rooms in developed countries, use of point- of-care technologies can reduce turnaround time by around an hour[144, 5]. There is particular interest in developing point-of-care technologies for low-resource settings, to increase test availability in developing countries[41, 146]. However, development of such tests comes with additional challenges such as limited availability of complex equipment to run the test and the presence of minimally-trained operators. Although miRNAs have shown great promise for disease diagnosis and have recently begun to be used in clinical settings for identification of tumor origin[176], due to long assay times and the need for complex equipment, miRNA assays have not been available for point-of-care settings.

Recent studies have focused on developing rapid miRNA diagnostics for point-of- care applications using colorimetric detection, electrochemical systems, and microfluidics. Studies have shown paper-based colorimetric systems for miRNA detection from body fluids, but these systems either require multiple hours for signal amplification or have limits of detection above 50 pM[177, 178, 179], much higher than the 10-100 fM average concentration in serum for highly abundant miRNA[29]. Elec- trochemical systems have been used to detect miRNA with subfemtomolar limits of detection[83, 180] but require bulky electrochemical workstations. Many microfluidic strategies have also been developed, but these result in assay times longer than an hour or have high limits of detection[181, 182, 107]. Incorporation of microfluidics with molecular beacon probes can achieve rapid detection of miR-21 but suffers from a lack of multiplexing[105]. In general, there is a tradeoff between assay time and limit of detection, since many strategies for reducing the limit of detection require increasing incubation times or adding additional amplification steps, making direct detection of miRNA in bodyfluids difficult at point-of-care timescales. For point-of- care detection in serum, technologies are needed that can detect miRNA down to the fM range with limited preprocessing and minimal external equipment.

60 Hydrogels have found great usefulness in biosensing because of their 3D crosslinked structure that allows them to be easily functionalized with capture agents for different types of biomolecules[109, 114, 129]. Due to their non-fouling nature, hydrogels have been used successfully for detection of miRNA from complex samples such as cell lysate, and FFPE. tissue[1, 106, 110, 128, 111, 112]. In raw cell lysate, hydrogel microparticles can achieve a limit of detection of 50 fM for miR-21 from a more than three hour assay[110]. For point-of-care detection of miRNA in serum, the LOD must be similar, but from an assay that lasts a third of the time. A B 0 Microfluidic Channel

Microscope Objective Bi - Blotin SA Streptavidin

Photomask4

light...... UU

Figure 4-1: (A) Fabrication of hydrogel posts using projection lithography within a microfluidic channel. (B) Schematic of miRNA reaction. First, the miRNA target hybridized to the DNA probe covalently incorporated into the hydrogel. Next, a biotinylated universal linker was ligated to the hybridized miRNA for 10 minutes at room temperature. Finally, fluorescent streptavidin-phycoerythrin binds to the biotin for visualization. (C) Brightfield image of 10 pm radius miR-21 posts after reaction with serum. Scale bar is 20 pm. (D) Fluorescent image of posts in (C). Scale bar is 20 pm.

In this study, we report direct detection of miRNA from serum in a 40-minute assay using polyethylene glycol diacrylate (PEGDA)-based hydrogel posts in a microfluidic channel. Using this format, we are able to obtain clinically-relevant LODs: a 73 f M limit of detection for miR-451 and a 94 fM limit of detection for miR-21. We also demonstrate multiplexing by detecting and quantifying miR-451 and miR-21 directly from serum along with positive and negative C. elegans controls. Both of

61 these miRNAs have been suggested for cancer diagnosis. miR-21 has strong anti- apoptotic functions and is overexpressed in a wide variety of different cancers[183, 184, 185, 186]. miR-451 is involved in tumorigenesis and is also dysregulated in many different cancers[187]. The assay is run under gravity-driven flow in a microfluidic channel, eliminating the need for external pumping equipment. The only equipment required is a hot plate and a fluorescent imaging setup. We optimize the assay for point-of-care applications by applying transport theory to improve the hydrogels and the bioassay, utilizing previous analysis of the key factors that influence bioassay signal[188]. By using a microfluidic format with high Peclet number, reagents are constantly delivered to the functionalized hydrogels. However, we maintain the flow rate low enough such that only 100-200 iL of serum are used per assay. The only pretreatment required for the serum are short (15 minute total) preincubations with lysis buffer to degrade endogenous RNAses to enable addition of a positive control and to begin the protein degradation process to release the miRNA for binding. This assay enables rapid detection of miRNA directly from minimally-processed serum and could be applied to diagnosis of a variety of disease conditions.

4.2 Experimental

4.2.1 Channel preparation

50 pm tall microfluidic channels were purchased from Hligenberg (Malsfeld, Ger- many). To create inlets and outlets to the channels, a gasket was created by punching 1.5 mm holes in a sheet of polydimethylsiloxane (Sylgard 184 Silicone Elastomer Kit, Dow Corning Corporation, Midland, MI, mixed at a base to curing agent ratio of 10:1). The gasket was then bonded to the glass channel after plasma treatment for 20-25 seconds in a Harrick PDC-32G plasma cleaner. Cut pipette tips inserted into the gasket holes served as inlet reservoirs. For channel activation, the channels were treated with 1 M NaOH for at least an hour, then rinsed three times with nuclease-free water (Affymetrix, Cleveland, OH)

62 and dried with argon gas (Airgas). The channels could be reused, in which case they were treated overnight with 1 M NaOH to dissolve the hydrogel posts. The channels were then treated with a solution of 2% 3-(trimethoxysilyl)propyl acrylate (Sigma), 73.5% ethanol (Decon Labs, Inc., King of Prussia, PA), and 24.5% phosphate-buffered saline (Corning) for 30 minutes, then rinsed with ethanol, dried, and placed in an 80 °C oven for 20 minutes.

4.2.2 Hydrogel polymerization

Monomer stock solution was made from 20% polyethylene glycol diacrylate 700 (Sigma- Aldrich), 5% 2-Hydroxy-2-methylpropiophenone (Sigma-Aldrich), 40% polyethylene glycol 600 (Sigma-Aldrich), and 35% 3X Tris-EDTA (TE) buffer (EMD Millipore). This stock solution was then diluted 9:1 with probe to form the final monomer mixture. In most cases, the final probe concentration in the monomer was 1 mM, although for studies with lower probe concentration, the concentration was 100 PM. ~ 2 pL of monomer was then added to a microfluidic channel. Hydrogel posts were formed through projection lithography (Figure 4-1A) as described previously[128] using mylar masks ordered from Fineline Imaging (Colorado Springs, CO) and a Zeiss Axio Observer microscope. The light source was a ThorLabs M365L2-C4 UV LED and the filter set was a Chroma Technology 11000v3-UV filter set. After polymerization, the channel was washed three times with ~ 150 pL of IX TE buffer. For studies with multiplexing, the next monomer (~ 5 ,L volume) was then introduced to the channel using a 1 mL syringe attached to a cut pipette tip.

4.2.3 Channel pretreatment

After polymerization, the hydrogels were treated with 500 pM potassium permanganate (Sigma) in 0.1 M Tris-HCl (Purchased at 1.5 M from Teknova) for five minutes under gravity-driven flow using a cut pipette tip as an inlet to the channel. The channels were then flushed two times with ~ 150 pL 1X TE buffer, then rinsed in 1X TE. For optimization studies the channels were rinsed for five minutes, but for serum

63 studies and calibration curves the channels were allowed to sit overnight in IX TE. The channels were then blocked with 3% Pluronic F-108 in nuclease-free water for 30 minutes under flow conditions. After blocking, the channels were flushed two times with IX TE buffer, then rinsed under flow with IX TE buffer for at least 30 minutes. After the pretreatment steps the channel could be stored for future use.

4.2.4 Detection of miRNA

The scheme for detection of miRNA is shown in Figure 4-1B. First, the hybridization buffer was added to the channel. The hybridization time was 20 minutes at 55 °C unless otherwise specified. For studies with synthetic miRNA, the RNA sequence was ordered from IDT (Coralville, IA) and the buffer consisted of 1X TE buffer with 0.05% Tween-20 and 350 mM NaCl. All nucleic acid sequences used in this study are listed in Table B.1. For studies in serum, the buffer composition[110] consisted of IX TE buffer with 0.05% Tween-20, 2% (w/v) sodium dodecyl sulfate (Sigma-Aldrich), 28 U/mL Proteinase K (New England Biolabs), and 350 mM NaCl. We calculated the NaCl concentration assuming 150 mM physiologic NaCl concentration in serum[189] as was done previously[106]. All serum assays were conducted in a biosafety cabinet. Serum assays also contained a spike-in of synthetic cel-miR-39, a C. elegans miRNA that is not present in human samples. The spike-in was present at a concentration of 3.2 pM. We verified that we obtained similar signal from synthetic miRNA from both serum spike-ins and using neat buffer (Fig. B-2). Serum was purchased commercially from BioIVT. After thawing the serum, it was mixed with all components of the hybridization/lysis buffer except the Proteinase K and the synthetic spike-in and heated to 90 °C for 10 minutes to deactivate RNAses. Although we could detect miRNA from serum without the heating step, it was necessary to deactivate the RNAses to prevent degradation of the synthetic spike-in. After the heating step, the solution was cooled to room temperature and the Proteinase K was added. The solution was then incubated for 5 minutes on a thermoshaker at 55 °C and 1500 rpm to begin denaturing proteins. The solution was then added to the channel for the hybridization step.

64 After hybridization, the channels were rinsed three times with ~ 150 pL of rinse buffer (IX TE buffer with 0.05% Tween-20 and 50 mM NaCl). After rinsing, the ligation buffer, consisting of IX TE with 0.05% Tween-20, 10% (v/v)) NEBuffer 2 (New England Biolabs), 250 tM ATP (New England Biolabs), 40 nM A12 universal linker (Integrated DNA Technologies, sequence given in Table B.1), and 798 U/mL T4 DNA Ligase (New England Biolabs) were added. The ligation step lasted for 10 minutes at room temperature unless otherwise specified. The labeling buffer was immediately introduced to the channel, without rinsing, after completion of the ligation step. The labeling buffer consisted of rinse buffer with 20 pg/mL streptavidin-R-phycoerythrin (Life Technologies). After labeling, the channel was rinsed three times with - 150 pL of rinse buffer and then imaged. For imaging, a Zeiss Axio Observer Al inverted microscope was used with an X-CITE 120LED broad spectrum LED as a light source and an Omega Optical XF101-2 fluorsecent filter set. Images were acquired on an Andor Clara CCD camera using Andor SOLIS software. The images were analyzed using ImageJ (National Institutes of Health)[166] and MATLAB scripts. The mean signal was calculated as the total fluorescence intensity of a circular region containing the hydrogel post divided by the area of the circular region. The mean background-subtracted signal was calculated by subtracting the mean signal of a region of the channel near the hydrogel post (the background) from the mean signal. Brightfield (Fig. 4-1C) and fluorescence (Fig. 4-iD) images of hydrogel posts after a serum bioassay show uniform signal.

4.3 Results and discussion

4.3.1 Assay design

For hydrogel particles in a well-mixed tube, the flux into the particle during short assay times is described by[188]:

J ~ D ekaPo T8 0, (4.1)

65 A B C 12 0 140 , , 0 120. 100 80 4 60 0 ~-100 I 8(j,

a 40 -40 20. .

20 C 20

200 300 400 500 00 700 0 10 20 30 4'0 60 > 10 Exposure time (ms) Post radius (pm) 0.1 mM 1 mM

Figure 4-2: Optimization of hydrogels for point-of-care detection of miRNA. (A) Signal from assay with synthetic miR-21 as a function of exposure time used to polymerize hydrogels. Each data point is the average of 3-5 hydrogel posts and error bars represent one standard deviation. (B) Signal from assay with miR-21 as a function of the radius of the hydrogel post. The height of each post is 50 pm. Each point is the average of 4-5 hydrogel posts and error bars represent one standard deviation. (C) Signal from assay with miR-21 with different probe concentrations in the monomer. Each data point is the average of five hydrogel posts and error bars represent standard deviations.

Dj is the diffusivity of the target in the hydrogel, ka is the forward reaction rate constant of the reaction of the target with the probe, Po is the initial probe concentration in the hydrogel, and To is the initial target concentration present at the edge of the hydrogel. For particles in a microfluidic channel, the external mass transfer is set by the Peelet number (Pe), or the ratio of the convective to diffusive flux[168]:

Pe UL (4.2) where U is the average velocity in the channel, L is the characteristic length scale (in this case either the distance from the post to the edge of the channel or the minimum post spacing transverse to the direction of fluid flow, whichever is less), and Dflyid is the diffusivity of the target in the fluid. In the limit of high Pe, when convection is much faster than diffusion, no external boundary layer forms and the concentration at the edge of the hydrogel is approximately the same as that in solution, assuming a partition coefficient of one. Under these conditions, where reagents are constantly being delivered to the hydrogel, the external concentration remains approximately constant and Equation 4.1 applies. In samples where volume is not limited, it is advantageous, then, to run flow-based tests at high Pe, such that the concentration

66 at the edge of the sensor is always maximized. In situations where the sample is very precious, reducing the flow rate lowers the throughput, resulting in a higher percentage capture of the target but a lower rate of reaction. The optimal operating conditions depend on the type of sample being processed.

In serum samples, blood is typically collected in a tube with volume > 1 mL, resulting in a large sample with low concentrations of miRNA. For optimal capture, then, we should operate at high Pe, to maximize the flux of target into the hydrogel. The flow rate in microfluidic devices varies with the size of the channel. Previously, in a channel 1000 pm wide by 50 pm tall, the flow rate was estimated to be 1 - 5 upL/min[128]. If flow rates are on this order of magnitude, a 20 minute assay should require well under 100 pL of sample volume to run, which is a small volume comprable to some fingerstick systems[190]. To maximize capture for a POC assay, we desire to keep the target concentration at the edge of the hydrogel as high as possible, to drive flux into the hydrogel. Assuming a partition coefficient of one, the flux into the hydrogel can then be described by Equation 4.1.

In order to have no external boundary limiting mass transport to the hydrogel posts, we need Pe» 1, so we take as the basis for our calculations a desired Peclet number of 100. Assuming a diffusivity of 2 x 10100m2 /s for miRNA in water at 55 °C (Calculated from the Stokes-Einstein equation assuming a radius of gyration of 3 nm for miRNA[121]) and a minimum post spacing of 100 tm, the minimum fluid velocity required for a Peclet number of 100 is

Pe * Dluid 100 x 2 x 10-1° m Umin L 100 106 2 x 10 (4.3)

This velocity gives us a minimum average velocity to obtain high Pe in the channels. The velocity is also bounded on the upper end by the maximum target volume we desire to use for the assay. Because the signal in the hydrogels increases with increasing height[188], we desire tall hydrogel posts to concentrate the signal.

High flow through the channel will result in constant delivery of target to the posts within the channel, maintaining the target concentration at the edge of the

67 posts equal to the concentration in solution. In order to concentrate the signal as much as possible, the projected area of the posts should be minimized[188]. However, the posts also must be stable under fluid flow. For this study, we selected 50 pm tall glass channels for the microfluidic assay. The use of glass channels enables covalent attachment of the hydrogel to both the top and bottom surface, increasing structural integrity. We anticipate that using taller channels could increase the signal further by increasing the height of the hydrogels, as long as structural integrity was maintained.

4.3.2 Hydrogel optimization

In order to design an assay for point-of-care miRNA detection, we first considered the factors that influence the resulting signal[188]:

Signal Areasurface Areamaging Areaimaging Dge is the diffusivity of the target in the hydrogel, ka is the forward reaction rate constant of the reaction of the target with the probe, Po is the initial probe concentration in the hydrogel, T,,o is the initial target concentration present at the edge of the hydrogel and t is the assay time. To optimize the hydrogel performance, we investigated three parameters: the exposure time (which indirectly influences the diffusivity and the probe incorporation through degree of cross-linking), the size of the hydrogel post (thereby manipulating the surface area to imaging area ratio), and the probe concentration in the monomer solution. To investigate the exposure time, we polymerized functional hydrogel posts using a 10 pr radius mask at varying UV exposure times during polymerization and tested by reacting with synthetic microRNA. We found that the signal after reaction with 1 pM synthetic miR-21 did not vary strongly with exposure time (Fig. 4-2A), suggesting a tradeoff between decreased diffusivity and higher probe incorporation at higher exposure times. We also observed that at higher exposure times the post size increased due to diffusion of the monomer during the polymerization process, which spread the signal over a larger area. We selected an exposure time of 400 ms for future study as we found that the posts polymerized with

68 A 70

-60 +

a 10

2 0 0 10 20 30 40 Ligation time (min) B

80 "a f C 05 060 tM . 0

. 20

40 F 20 4 6 8 10 12 14 16 Labeling time (min) Figure 4-3: Optimization of assay for point-of-care miRNA detection. (A) Signal from miR-21 assay as a function of ligation time. Each data point is the average from five hydrogel posts and error bars represent one standard deviation. (B) Signal from miR-21 assay as a function of labeling time with streptavidin-phycoerythrin. Each point is the average from four hydrogel posts and error bars represent one standard deviation.

69 lower exposure times often detached from the channel surface, but posts with longer exposure times were more stable. We also varied the post radius as we expected that we could increase the mean signal by concentrating the signal in a smaller area (Eq. 4.4). We found that by decreasing the post size from 50 pm to 10 pm we were able to increase the mean signal by a factor of four, so we selected 10 pm posts as our optimal size. We did not investigate smaller sizes due to fabrication and imaging constraints, but we anticipate that smaller post sizes would increase the mean signal further, as long as the posts were stable under fluid flow. Finally, we also increased our probe concentration in the monomer by an order of magnitude to increase the signal further (Fig. 4-2C). By increasing the probe concentration by an order of magnitude, we were able to obtain a factor of two increase in signal. If we desired a lower limit of detection, we could also increase the hybridization time to increase the signal. However, for this study, maintaining a fast hybridization time to be relevant to point-of-care applications was more important than decreasing the limit of detection, so we chose a faster hybridization time. Through optimization of the hydrogel and the assay format, we were able to achieve a 20 minute hybridization time, which is much less than the 90 minute hybridization commonly used previously[191, 192, 1, 128, 111, 112].

4.3.3 Assay optimization

Using our optimized hydrogel posts, we next sought to optimize the miRNA assay for point-of-care detection. Previous studies developed a three-step scheme for miRNA detection that involves hybridization of the miRNA, ligation of a universal linker, and labeling with streptavidin-phycoerythrin for fluorescent detection. This scheme was previously validated against RT-PCR[1. While previous studies have used a ligation time of 30 minutes and a labeling time of 45 minutes[1, 128, 111, 112], we worked to decrease these times to make the assay more applicable to point-of-care settings. We tested hydrogel posts by reaction with 1 pM synthetic miR-21 at various ligation times, keeping the hybridization time and labeling time constant at 20 and 5 minutes respectively. We found that the signal increased with increasing ligation time up to 30 minutes. Because we were concerned with shortening the assay time for this study, we

70 chose a ligation time of 10 minutes. The signal could be increased by lengthening the ligation time, in applications where the overall assay time was not a concern. We also anticipate that the ligation time could be decreased by increasing the concentration of the components of the ligation buffer, but we did not investigate that in this study. We also investigated the labeling time with 20 pg/mL streptavidin-phycoerythrin. Previous studies used a concentration of 2 pg/mL streptavidin-phycoerythrin and a 45 minute labeling time, but after increasing the concentration tenfold we found that varying the labeling time from 5-15 minutes did not impact the assay signal(Fig. 4-3B), suggesting that even faster labeling times could be possible without decreasing signal. For this study, we selected a labeling time of 5 minutes. In any miRNA detection technology, multiplexing capabilities are crucial. While miRNA are dysregulated in the presence of many diseases, many individual miRNA are dysregulated in multiple diseases, so most panels for disease diagnosis involve detection of 3 - 5 miRNA[84, 86, 193]. To enable multiplexing in this assay, we used spatial encoding and polymerized hydrogel posts functionalized for detection of a certain miRNA in a specific region of the channel. We then washed with buffer and introduced the next monomer into the channel before polymerizing again, enabling us to create posts with multiple different functionalities in the same channel. We verified that repeated alternation between two monomers resulted in similar signals by alternating between a biotin-functionalized probe and a negative control (Fig. B- 1). We obtained similar signal from all monomers of the same type, validating our multiplexing strategy.

4.3.4 Calibration curves for miRNA detection

To quantify the microRNA detection capabilities of our assay, we ran calibration curves with synthetic miRNA, varying the concentration of miRNA in each channel and measuring the resultant mean fluorescent signal. The net signal was calculated by subtracting the mean background-subtracted signal of the hydrogel posts at a given concentration from the mean background-subtracted signal of control posts from a channel with 0 fmol miRNA. The noise of the sample is defined as the standard

71 A B 10 3 10 3 LOD LOD 94 fM 73 fM 10 10 Ma

SN...... R = 3 ...... 10 *101 S N = 2 z

10 - 10SN 103 4 10 10 2 103 10 4 10 10 2 miR-21 concentration (fM) miR-451 concentration (fM)

Figure 4-4: Calibration curves for detection of miRNA. (A) Calibration curve for detection of miR-21. Each point represents the average of five posts. Error bars represent one standard deviation. (B) Calibration curve for detection of miR-451. Each point represents the average of five posts and error bars represent one standard deviation. R2 >0.99 for both curves. LOD indicates limit of detection. SNR indicates signal to noise ratio. deviation of the control sample. To find the limit of detection (LOD), we fit a line to a log-log plot of the net signal-to-noise ratio versus concentration and calculated the point at which the signal-to-noise ratio was equal to three. For miR-21, the limit of detection was 94 fM (Fig. 4-4A) while for miR-451 it was 73 fM (Fig. 4-4B). These LODs are comparable to the concentrations of abundant miRNA in serum[29]. We obtained similar LODs for both miRNAs studied.

4.3.5 Detection in serum

We desired to apply our technology for the detection of miRNA to detect miRNA directly in serum. Because we have previously shown that hydrogel particles are nonfouling even in raw cell lysate or serum[110, 106], we wanted to enable direct detection of miRNA with minimal sample pretreatment. In this study, we used both a positive and negative control and detected two miRNA: miR-21 and miR-451 from serum. To detect miRNA directly in serum, we used a lysis buffer containing SDS and Proteinase K that was previously shown to enable miRNA detection directly in cell lysate[110]. miRNA in serum is complexed either with exosomes[194] or proteins[28],

72 ? 400 --- - Mchannel 1 MChannel 21 350,

C, 300 ,

COI C miR-451 miR-21 cel- cel- miR-54 miR-39

Figure 4-5: Detection of miRNA in serum. Detection of two endogenous miRNAs: miR-451 and miR-21 and a negative control (cel-miR-54) and positive control (cel- miR-39). The posive control was spiked into the reaction mixture at 3.2 pM. Each point is the average of 3-5 hydrogel posts and error bars represent one standard deviation. so it is necessary to release the miRNA by denaturing the proteins and lysing exosomes before it can bind. Furthermore, RNAses are present in serum[195] and must be degraded so that the miRNA is not cleaved prior to analysis. For this reason, we included both SDS and Proteinase K in the reaction, to lyse exosomes and denature both the Argonaute2 complex and RNAses. We validated that synthetic miRNA had a similar signal in both neat hybridization buffer (containing only NaCl, Tris-EDTA buffer, and Tween-20) and the hybridization/lysis buffer that also contained SDS and proteinase K. We found that we obtained a similar signal in both buffers (Fig. B-2), indicating that the lysis buffer did not interfere with the assay. We also found that soaking the channels for at least four hours in Tris-EDTA buffer prior to running the assay increased the signal from the positive control (Fig. B-3), so for all calibration curves and all serum assays we soaked the channel overnight. We believe that the soaking period allows time for unreacted -probe and monomer to be eluted from the gels, and for the gels to swell to an equilibrium size after polymerization.

To optimize the assay for serum detection, we investigated the impact of preprocessing steps on the assay. We found that although it was possible to detect miRNA without any preprocessing of the serum other than thawing (data not shown), in

73 order to add a positive control it was necessary to degrade the RNAses prior to the beginning of the assay. For this reason, we implemented a 10 minute heating step at 90 °C to deactivate the RNAses prior to adding the synthetic control. We also performed this step prior to the addition of Proteinase K, so that the Proteinase K was not deactivated during the heating. We also investigated pre-incubating the serum with Proteinase K for five minutes after the heating step before beginning the assay, to begin denaturing the proteins prior to the start of the assay. We found that doing so impacted the assay signal (Fig. B-3), so we added an additional pre-incubation step after the addition of Proteinase K, resulting in 15 minutes total of pre-incubation prior to the start of the assay.

After optimizing the assay, we ran a multiplexed assay in serum. We selected serum for technology development due to its commercial availability. The serum was diluted 4:1 with the buffer components to maintain the miRNA concentration as high as possible. We ran a multiplexed assay detecting two endogenous miRNA (miR-21 and miR-451) and a positive and negative C. elegans miRNA control (Fig. 4-5). To assess variability, we ran two technical replicates with the same sample. We found a 12% difference between the positive controls in the two channels, indicating good reproducibility. Similarly, we measured a 14% difference in miR-21 and a 12% difference in miR-451. Although for bead and particle-based assays many particles are typically fabricated to ensure sufficient particles remain for analysis after washing steps[110, 112], because the hydrogel posts are covalently bound to the channel fewer posts need to be fabricated. The nonspecific signal obtained from the negative control is less than half of the signal obtained from miR-451 and miR-21, indicating that although there is some amount of nonspecific binding, this is substantially less than the signal obtained from the target miRNAs. By comparing the measured signal from serum with the calibration curves, we obtain a miR-451 concentration of 310 f M and a miR-21 concentration of 210 f M. Previously, miR-451 was measured to be on average 2-fold higher than miR-16 in human serum[196], which varies between 20,000-500, 000 copies per pL in normal serum[29]. Based on these values, we would expect a miR-451 concentration between 70 and 1700f M for normal donors, and the measured value of

74 310 fM is well within that range. miR-21 expression was previously measured to be on average three times higher than miR-16 in normal human serum[197], resulting in expected miR-21 concentrations from 100-2500 fM for human serum. The measured concentration is within the expected range, validating the assay. Future work will seek to apply this strategy to the detection of additional miRNA and apply the platform to measure miRNA panels for disease diagnosis.

4.4 Conclusions

In this article, we describe direct detection of miRNA from serum in a 40-minute assay. By optimizing the hydrogel posts used for miRNA capture, we are able to substantially reduce the assay time while maintaining a clinically-relevant limit of detection, in the 70-100 fM range. We obtain good technical reproducibility, with channel- to-channel variations around 12%. The measured miRNA concentrations are within previously reported ranges[196, 29, 197], validating the assay. The only pretreatment steps required are a preheating step to inactivate R.NAse and a pre-incubation step with the assay reagents, requiring only 15 minutes total and allowing for easy application to the point-of-care. The assay also uses gravity-driven flow in a microfluidic channel, eliminating the need for external pumping and only requiring a hot plate and fluorescence microscopy for external equipment. We anticipate that this technology could be combined with portable systems for fluorescence microscopy[198] and heating[199] to develop simple devices for detecting miRNA at the point-of-care. Fu- ture studies will focus on applying this technology to measure additional miRNA and to detect panels of miRNA for disease diagnosis. A couple possibilities are renal cell carcinoma, in which miR-21 is upregulated[200] and miR-451 is downregulated[196], or breast cancer, in which both miR-21 and miR-451 are upregulated[201]. The broad applicability of this technology can enable detection of miRNA for diagnosis of many different diseases at point-of-care-relevant timescales.

75 76 Chapter 5

Incorporation of Hydrogels into Fibrous Substrates for Quantitative, Multiplexed Bioassays

This thesis aims to develop hydrogels for point-of-care detection of biomarkers. This chapter describes a novel combination of hydrogels with fibrous substrates. Fibrous substrates are commonly used in paper microfluidics due to their fluid wicking properties. Here, we combine those advantages with the advantages of hydrogels and show that the hydrogel-functionalized fibrous substrates can be used for detection of microRNA and proteins.

5.1 Introduction

Fibrous substrates such as paper[45], silk[202], and glass fiber membranes[203, 204] have found great usefulness in point-of-care diagnostics due to their wicking properties with small sample volumes[74]. Bioassays can also be combined with sample processing steps to facilitate fingerstick assays from whole blood[205]. Fibrous substrates are also affordable[204] and can be easily disposed of through burning, eliminating the buildup of biohazardous waste[45], which are both important advantages for low- resource settings. On fibrous substrates, spatial encoding of different capture agents

77 can also easily enable multiplexing for bioassays[74, 68] and patterning of hydropho- bic regions facilitates fluid flow without the need for external pumps[74]. However, achieving high resolution patterning is expensive and more affordable technologies such as bioprinting require large solution volumes, typically at least 100 pL[206, 2071, creating challenges for expensive reagents such as antibodies. Furthermore, some substrates are challenging to functionalize. For example, nitrocellulose is used much more widely than glass fiber because it is easier to functionalize, even though glass fiber is more than an order of magnitude cheaper[208, 209, 2101. Hydrogels have been used extensively in bioassays for detection of proteins and nucleic acids[109, 128, 1, 129, 188]. The platform nature of the technology allows for incorporation of various detection molecules including aptamers, antibodies, and nucleic acid probes[1, 3, 121, 114]. Hydrogel advantages, including solution-like kinetics and a non-fouling nature, enable their use for sensitive bioassays from a wide variety of sample types, including raw cell lysate, serum, and tissue sections[109, 128, 112, 111, 106]. Additionally, their optical transparency enables highly sensitive fluorescent quantification. Previous bioassays involving hydrogels have used various formats including hydrogel particles in solution[1, 112, 188, 211], hydrogel posts in microfluidic channels[128, 127], hydrogel posts in wells[111], and hydrogel sheets[212]. Incorporating hydrogels into paper can improve the stability of biomolecules or aid in flow control for bioassays[213, 214]. Polymerization of hydrogels has also been used as a signal amplification strategy for bioassays on paper[57. Although hydrogel precursors have been printed onto paper or nitrocellulose using inkjet or solenoid printers for small molecule detection[215, 216, 217], such studies have only demonstrated resolution >100 pm. Similarly, pre-fabricated hydrogels can be threaded onto yarn substrates or spread on paper to keep reagents hydrated and aid in functionalization[218, 219]. Although these studies demonstrate the utility of hydrogels for bioassays in fibrous substrates, to our knowledge no studies have demonstrated precise patterning of functionalized hydrogels within fibrous substrates. Here, we report the direct polymerization of hydrogels in mask-defined shapes down to 50 pm sizes within fibrous substrates. We sandwiched the substrate and

78 monomer solution between glass slides and used projection lithography to polymerize hydrogels of various shape and functionality within the substrate. We then used these hydrogel-functionalized substrates to facilitate quantitative, multiplexed bioassays. The fibrous substrates serve as a handle for the hydrogel particles, to facilitate buffer exchange and reagent transfer steps which enable large volume exchange without the loss of particles. In this study we show that hydrogels polymerized into glass fiber membranes can be functionalized for detection of microRNA (miRNA) or proteins. miRNA are short (- 22 nucleotide), non-coding RNA that are of increasing interest for disease diagnosis due to their involvement in the regulation of gene expression and dysregulation patterns in many diseases[35]. We particularly focus on miR-210, an oncogenic miRNA that is upregulated in hypoxic conditions[2201, and miR-141, which is dysregulated in a wide variety of cancers[221]. We obtain a limit of detection of 28 amol for miR-210 on hydrogels in glass fiber membranes, which is comparable to some previously-published particle systems[191]. We also demonstrate the capability for multiplexed assays using spatial encoding for simultaneous detection of miR-210 and miR-141. Furthermore, we show that the format is compatible with protein detection by capturing thyroid-stimulating hormone (TSH) on antibody-functionalized hydrogels embedded in the glass fiber substrate. TSH detection is important due to the estimated 12% of people in the U.S. that will develop some type of thyroid dysfunction during their lifetime[222]. We believe that the use of these hydrogel-functionalized substrates for bioassays could be easily extended to other protein and miRNA targets as well as other types of biomolecules, such as mRNA, that have previously been detected on gel particles[1211.

5.2 Results and Discussion

A schematic diagram of the process for patterning hydrogels in fibrous substrates is shown in (Figure 5-1A). We placed the fibrous substrate on a slide coated with polydimethylsiloxane and added a 5 pL drop of monomer solution. We then used an additional slide to sandwich the substrate and placed it on an inverted microscope.

79 A Hyd ogel B C Capture Labeling GlassDetectiQ Slides.- antibody

.Ta rget 80-200 parge

Fibrous .. ctu Substrate antibody Mirscoenp jectivy D Hybridization Ligation Labeling

Target mniRNA

Linker

Capture Blotin Biotin probe SAPE UV LED

Figure 5-1: Methods for bioassays in fibrous substrates. (A) Fabrication of hydrogels in fibrous substrates. (B) Image of hydrogel particle fabricated in glass fiber. Scale bar is 100 im. (C) Detection of proteins on hydrogels in glass fiber. The first step is target capture for 15 minutes, followed by labeling with detection antibody for 5 minutes. (D) Detection of miRNA on hydrogels in glass fiber. The first step is hybridization of the miRNA target for 90 minutes at 55 °C, followed by ligation with a universal linker sequence for 30 minutes at 21.5 °C and labeling with streptavidin- phycoerythrin for 45 minutes at 21.5 °C.

Using projection lithography, we exposed the substrate to UV light through a photomask, to generate various shapes of hydrogels. A brightfield image of a circular hydrogel fabricated in glass fiber is shown in Figure 5-1B. After fabrication of the hydrogel-functionalized fibrous susbtrates, we applied these systems to the detection of proteins and microRNA. We enabled protein detection by incorporating a capture antibody into the hydrogel. The protein detection scheme is shown in Figure 5-1C. After capture of the target, we incubated with a detection antibody conjugated to a fluorescent target. We also enable microRNA detection by covalently incorporating a DNA capture probe into the hydrogel during polymerization. The assay is illustrated in Figure 5-1D and consists of a ligation step using a biotinylated linker followed by a labeling step with streptavidin conjugated to a fluorophore for fluorescence visualization.

80 5.2.1 Patterning in substrates

In this article, we demonstrate the ability to polymerize hydrogels directly into three different fibrous substrates: silk, glass fiber membrane, and nitrocellulose. All three of these materials have been previously used as a substrate for bioassays[202, 48, 2041. Nitrocellulose is commonly used as a substrate for both lateral flow and dipstick assays[48], while glass fiber is often used in the sample pad or conjugate pad of lateral flow assays or can be functionalized for microfluidic assays[45, 204, 203]. Due to its fluid flow properties, silk has also been used as a microfluidic platform for biomolecule detection[223]. Incorporating hydrogels into fibrous substrates can lever- age the wicking and spatial resolution advantages of the substrates and combine them with advantages of hydrogels, such as precise control over shape and the ability to incorporate a wide variety of capture moities into the gels[109]. The different structures of the three fibrous materials used in this study can be seen in Figure 5-2. The silk (Figure 5-2A) is highly ordered, while the glass fiber membrane (Figure 5-2B)) is more amorphous. The nitrocellulose (Figure 5-2C) also displays random pores, but of a smaller pore size than the glass fiber membrane. To test our ability to pattern hydrogels in these various substrates, we fabricated circular hydrogels using masks with projected radii of 250, 100, 50, and 25 (Figure 5-2D-F). For Figure 5-2D-F, brightfield images are on top with fluorescence images on the bottom. Although the shapes are visible in all three fluorescence images, the circles appear more well-defined in the silk and glass fiber substrates than in the nitrocellulose, indicating better patterning resolution. In Figure 5-2G-I, we show our ability to pattern triangular structures in all three fibrous substrates (silk (G), glass fiber membranes (H), and nitrocellulose (I)). Each panel contains a color brightfield image (top) and a greyscale fluorescence image (bottom). Although the triangular shape is identifiable in the fluorescence image of all three substrates, the triangles in the glass fiber membrane are visually the most well-defined of the three. To compare the three different substrates for direct polymerization of hydrogels, we plotted the radially-averaged fluorescence of the largest circle (250 pm mask). For each substrate, the fluorescence was background-

81 subtracted using an average background signal calculated from five images taken in different locations on the substrate. The hydrogels fabricated in the glass fiber membrane show the highest signal above background, with the signal from the hydrogels fabricated in silk and nitrocellulose are much lower. We attribute this to light scattering induced by the substrate during the polymerization process, which reduces the amount of UV light that penetrates the substrate and therefore the extent of polymerization. Because the fluorescent signal comes from fluorescent microspheres that are sterically trapped within the pores of the hydrogels, we expect gels with higher degree of cross-linking to have smaller pore sizes and therefore higher fluorescence. The particles fabricated in the glass fiber membrane also show the most well-defined shape when compared to the shape of the photomask (Figure 5-2K), which we believe also occurs due to lower light scattering. To verify this, we measured the percentage light transmittance for all three substrates. A schematic of the setup is shown in Ap- pendix C (Figure C-1). Glass fiber membrane and silk transmitted similar amounts of 365 nm light at 49% and 51%, respectively, while nitrocellulose only transmitted 22%. Although the glass fiber membrane transmitted slightly less light than the silk substrate, the thickness of the silk was less than half of the glass fiber, so the transmittance per unit depth is much higher for the glass fiber, resulting in better shape definition after polymerization.

5.2.2 Hydrogel Motifs in Glass Fiber

Because the glass fiber membrane showed the best performance for hydrogel fabrication, we investigated this substrate further, fabricating different motifs, which are all important for bioassays. On a large scale with a 20x objective we can fabricate ~ 100 pm-wide lines that stretch across the glass fiber substrate (Figure 5-3A), using only 5 pL of monomer. These strips are analogous to the test lines used in lateral flow and dipstick assays[48]. We anticipate that by using lower magnification or contact lithography[157] we could pattern even larger structures in the substrate. We can also pattern small structures in fine arrays, such as the 2x6 array of - 50 pm circles shown in Figure 5-3B. Small arrays of particles can be useful to generate detection

82 Nitrocellulose

Di E7 F

GitH L

J 4000 K Mask silk

3000 0 Nio

00 2 0 E '1000

0 100 200 300 Radius (pm) I Glass fiber Nitrocellulose

Figure 5-2: In situ fabrication of hydrogel particles in various substrates. (A) Brightfield image of silk substrate at 20x magnification. Scale bar is 100 pm. (B) Brightfield image of glass fiber substrate at 20x magnification. Scale bar is 100 pm. (C) Brightfield image of nitrocellulose substrate at 20x magnification. Scale bar is 100 pm. (D) Brightfield (top) and fluorescence (bottom) images of hydrogel circles of radii 250, 100, 50, 25 pm fabricated in silk. Scale bar is 100 Am. (E) Brightfield (top) and fluorescence (bottom) images of hydrogel circles of radii 250, 100, 50, 25 pm fabricated in glass fiber. Scale bar is 100 pm. (F) Brightfield (top) and fluorescence (bottom) images of hydrogel circles of radii 250, 100, 50, 25 pm fabricated in nitrocellulose. Scale bar is 100 pm. (G) Brightfield (top) and fluorescence (bottom) images of hydrogel triangles fabricated in silk. Scale bar is 20 pm. (H) Brightfield (top) and fluorescence (bottom) images of hydrogel triangles fabricated in glass fiber. Scale bar is 20 pm. (I) Brightfield (top) and fluorescence (bottom) images of hydrogel triangles fabricated in nitrocellulose. Scale bar is 20 pm. (J) Plot of background- subtracted fluorsecnt intensity of circles fabricated in various substrates using a 250 pm mask. Vertical line indicates projected mask edge. (K) Comparison of triangles fabricated in various substrates with projected mask size. Scale bar is 20 pm.

83 statistics from a single sample and enable spatial encoding for multiplexing. This hydrogel array shows good uniformity, with an interparticle coefficient of variation of the fluorescent intensity of 17.6%. The glass fiber substrate also allows for creating particles with different functionalities in the same substrate, using spatial encoding to differentiate between different types of particles, which can enable multiplexing for bioassays. Figure 5-3C (brightfield) and D (composite fluorescence image) show pic- tures of hydrogels functioned with red (left) or green (right) beads sterically trapped in the hydrogels during polymerization. The glass fiber membrane can also be utilized as a handle for hydrogel particles, to facilate moving the hydrogel particles for dipstick-like bioassays. This ensures that no particles are lost during buffer exchange steps. We polymerized 300 pm biotin-functionalized squares hanging off the end of the hydrogel substrate and reacted them with streptavidin phcoerythrin to generate a fluorescent signal (Figure 5-4E (brightfield) and F (fluorescence). Interestingly, the fluorescent signal in the area inside the glass fiber substrate is similar to the signal of the part of the particle overhanging the edge (Figure S2), indicating that the both the polymerization and signal acquisition are similar within and outside the glass fiber substrate.

5.2.3 Optimization of Hydrogel-Functionalized Glass Fiber Sub- strates for Bioassays

To show the applicability of these fibrous-hydrogel hybrid systems with bioassays we further optimized the system using biotin-functionalized hydrogels and a reaction with streptavidin-phycoerythrin. We varied the size of hydrogel particles to determine which size resulted in the best signal. We anticipated that there is an optimal particle size. At small sizes, the gel doesn't polymerize fully, while at large sizes, the signal from the bioassay is less concentrated (Figure 5-4A). We found that 50 pm radius particles generated a higher signal than either 20 pm or 100 pm particles (Figure 5-4B). We also investigated the exposure time, as we expected that at low exposure times the gels wouldn't polymerize fully, resulting in lower probe incorporation,

84 Figure 5-3: Fabrication of different hydrogel motifs in glass fiber substrates. (A) Fabrication of lines across glass fiber substrate. Scale bar is 100 pm. (B) Fabrication of array of circular hydrogels in glass fiber. Scale bar is 100 pm. (C) Brightfield image of two different particles with different functionalities. Right particle contains green fluorescence beads while left particle contains red fluorescent beads. Scale bar is 100 pm. (D) Fluorescence image of (C). (E) Brightfield image using glass fiber as a handle for rectangular particles functionalized with biotin. (F) Fluorescence image of biotin particles from (E) after reaction with streptavidin-phycoerythrin. Inset figure shows schematic of assay using hydrogels attached to glass fiber substrate.

85 while at higher exposure times, the gel would be too tightly cross-linked, limiting the diffusion of large strepvavidin-phycoerythrin[162]. We also anticipate that at longer exposure times, there is more diffusion of the monomer during the polymerization process, resulting in less well-defined shapes. Figure 5-4C shows fluorescence images of biotin-functionalized particles after reaction with streptavidin-phycoerythrin and Figure 5-4D quantifies the signal from these particles. The uncropped version of Figure 5-4C is shown in Figure C-3 in Appendix C. We found that a 300 ms exposure time resulted in the highest signal, so we selected that for further studies with bioassays.

5.2.4 Detection of Biomolecules

After optimization of the fibrous-hydrogel hybrid systems for bioassays, we sought to apply the technology to assays for proteins and miRNA. miRNA are becoming increasingly important diagnostic targets due to their dysregulation patterns in a wide variety of diseases[224, 225, 226] and their stability in blood and body fluids[30, 227]. Figure 5-5A shows a calibration curve for miR-210 in glass fiber substrate. The curve is linear over more than an order of magnitude change in concentration. The limit of detection of 28 amol is sufficient for the technology to find immediate use for detection of miRNA from cell lysate or tissue samples[35, 112, 110]. This level is also just above the level of circulating microRNA[35, 29] and is comparable to the limit of detection obtained with other non-amplification hydrogel particle systems[191, 228]. We can also use the hydrogel-functionalized glass fiber substrate for mutiplexing studies, by incorporating probes for different miRNA at different locations. After polymerizing the first monomer, we rinsed the substrate and placed it into a tube containing the next monomer, and in this way sequentially polymerized hydrogels of different functionalities. Figure 5-5B shows the results from simultaneous detection of 1000 amol miR-210 and miR-141. The negative control is a C. elegans miRNA that is not present in humans and was not spiked in to the buffer. We obtain a much higher signal from the miR-210 and miR-141 particles than the negative control. The difference in signal between the miR-210 and miR-141 hydrogel particles is likely

86 A B -3000 20 pm 2500

2000 50 pm E 1500

1000

j100pm 2 500 -0

20 pm radius 50 pm radius 100 pm radius C 500 ms D -1500 400 ms

t5 1000 300 ms

- i 500 200 ms

100 ms 00 200 400 600 Exposure time (ms) Figure 5-4: Optimization of bioassay in glass fiber using biotin-functionalized particles. (A) Comparison of different sizes of hydrogel particles in a glass fiber substrate using 20 pm, 50 pm, and 100 pm radius masks. Scale bar is 100 pm. (B) Quantifi- cation of the signal from biotin-functionalized hydrogel particles after reaction with streptavidin-phycoerythrin. (C) Image of hydrogel circles fabricated in glass fiber using 50 pm mask and varying UV exposure times for polymerization. Scale bar is 100 pm. (D) Comparison of signal from reaction of biotin-functionalized particles fabricated with different UV exposure times during polymerization.

87 A 10' B 3500 LOD- 28 anol - 3000 I 02 2500 102 0) 2000 0 1500 101 CW 1 1000 SNR=3 2 . 2 500 100, I S 0 10 10 m 10 10 4 cel-mir-238- miR-210 amount (amol)

C D_ 1500

1000

W 500

2 0 .

CU 0 M0 Figure 5-5: Detection of miRNA on hydrogels in glass fiber. (A) Calibration curve for miR-210 detection on hydrogel particles in glass fiber. Each point is the average of 4- 5 particles and error bars represent one standard deviation of the signal measurement divided by the noise. (B) Multiplex study showing detection of miR-210, miR-141, and cel-miR-238 (negative control) on the same glass fiber substrate. miR-210 and miR-141 were both present at 1000 amol. Each point is the average of five particles and error bars represent one standard deviation. (C) Fluorescence image showing multiplex study from (B). Scale bar is 100 pim. (D) Detection of protein target thyroid-stimulating hormone (TSH) on hydrogels in glass fiber. Each data point is the average of 4-5 particles and error bars represent standard deviation.

88 due to a difference in reaction rate between different probes; a previous study showed that miR-210 probes capture 1.5 times as much target as miR-141 probes on hydrogel particles[106]. Figure 5-5C shows a fluorescence image of the substrate with all three types of hydrogel particles. The hydrogels were spatially encoded by grouping similar types of gels. As expected, the signal from the negative control is much lower than the hydrogels functionalized for miR-210 and miR-141. In addition to detecting miRNA, we also wanted to use the substrates to detect protein targets, due to the wide prevalence of protein targets in clinical assays. To assess this, we chose thyroid-stimulating hormone (TSH) as a target. TSH is used clinically to diagnose thyroid diseases, which affect an estimated 200 million people[152, 150]. We verified our ability to detect protein targets by incorporating an antibody for TSH into the hydrogels and detecting the TSH protein. We tested with 20 pIU/mL TSH, which is a level that indicates the presence of hyperthyroidism[229]. We found that the signal from 20 pIU/mL TSH was much higher than the signal from the phosphate-buffered saline control (Figure 5-5D), indicating that hydrogels can be used to successfully functionalize the glass fiber substrate for protein detection of clinical analytes.

5.3 Conclusion

In this study we demonstrate direct spatial patterning of hydrogels into fibrous sus- bstrates and the use of these systems in bioassays. Flexible fibrous materials have shown great promise for development of portable diagnostic devices for low resource settings. This work combines the field of flexible fibrous materials with lithographic patterning of hydrogels, using control over hydrogel shape, location, and function to enable quantitative, multiplexed bioassays. Leveraging projection lithography, we demonstrate the fabrication of different hydrogel particle motifs within glass fiber substrates. The versatile nature of this approach indicates that by controlling the mask shape, virtually any 2-D extruded shape can be polymerized within the substrate. These hydrogel-functionalized substrates combine the advantages of fibrous sub-

89 strates such as moisture wicking with the easy functionalization, low background fluorescence, and nonfouling nature of hydrogels. We demonstrate the use of these materials for both protein and miRNA detection, suggesting wide applicability for disease diagnosis. By incorporating probes for other miRNA or antibodies or aptamers[114] for different protein targets, these systems could be used for the detection of a wide variety of diagnostic targets. We also anticipate that these systems could be used for detection of other targets such as mRNA that have been previously detected on hydrogel particles[121] as well as novel diagnostic targets. Due to the ability to control shape and function, these flexible substrates could also be used in additional applications that combine flexible substrates with a biofunctional element, such as wearable patches for disease diagnosis[230].

5.4 Experimental

5.4.1 Substrates

The three fibrous substrates used in this study were glass fiber membrane (G041 Glass Fiber Conjugate Pad Sheet, Millipore), nitrocellulose (FF120HP, Whatman), and silk (# 3760146, Mysore Saree Udyog, Bangalore, India). The silk was purchased in person in Bangalore and an image of the label is included in the Supporting Information (Figure C-4). The thickness of all three substrates was measured using a digital mi- crometer (Tormach) to be 202 pm for the glass fiber, 195 pm for the nitrocellulose (given by manufacturer to be 200 pm), and 85 pm for the silk.

5.4.2 Monomer preparation

The hydrogels used in this study were made from poly(ethylene glycol) diacrylate (PEGDA). For the studies without bioassay functionality, the monomer solution consisted of 35% PEGDA 700 (Sigma), 20% poly(ethylene glycol) 200 (Sigma), 5% photoinitiator (2-hydroxy-2-methylpropiophenone, Sigma), 35% 3X Tris-EDTA buffer (Calbiochem, Purchased at 1OOX from EMD Millipore), and 5% fluorescent mi-

90 crospheres (either carboxylated YG, diameter = 0.2 pm from Polyscience, Inc. or FluoSpheresTM carboxylate 0.5 pm (diameter) red from Invitrogen) For bioassay studies, the monomer consisted of 20% PEGDA 700, 40% PEG 200 or 600, 35% 3X Tris-EDTA buffer, and 5% photoinitiator. For studies with biotinylated probe or probes for miRNA PEG 600 was used, while for protein detection PEG 200 was used. The monomer solution was then diluted 9:1 with probe or antibody. The antibody was Fitzgerald 10-T25C, which was acrylated as described previously[188] prior to incorporation. For studies with biotinylated probe, the final probe concentration in the monomer was 10 pM, while for miRNA studies, the final probe concentration in the monomer was 100 pM. All probes were ordered from Integrated DNA Technologies and sequences are given in Table S1 in the Supporting Information.

5.4.3 Patterning hydrogels in fibrous substrates

To fabricate hydrogel motifs in the fibrous substrates, the substrate was placed on a slide coated with polydimethylsiloxane (PDMS, purchased as SYLGARD TM 184base and curing agent from the Dow Chemical Company). To prepare the slide, a drop of PDMS was placed on a cover slip (24 x 60 mm, # 1,5 Microscope Cover Glass, VWR) and spun on a spin coater (Model WS-650MMZ-23NPP, Laurell Technologies) for 30 seconds at 1500 rpm, then cured overnight in a 65 °C oven. After placement of the fibrous substrate on the slide, 5 pL of prepolymer mixture were pipetted onto the substrate and it was covered with another PDMS-coated glass slide. The slide was placed on an inverted microscope (Zeiss Axio Observer) and polymerized using projection lithography. To do this, a mask (Fineline Imaging) was placed in the field stop of the microscope and UV light (from a UV LED, ThorLabs M365L2-C4), was projected through the photomask and a Chroma Technology 11000v3-UV filter set to polymerize the monomer. Unless otherwise specified, an exposure time of 300 ms was used, using a Vincent Associates Uniblitz VCM-D1 shutter driver controlled by a Python script. After polymerization, the hydrogel-functionalized substrate was placed in 500 pL

91 1X Tris-EDTA buffer with 0.05% Tween-20 (IX TET buffer) and vortexed for 30 seconds to rinse. For particles functionalized with biotinylated probe or capture probe for miRNA, the substrates were rinsed a second time and the functionalized substrate was left at least overnight in 1X TET at 4 °C before running the assay. For studies with antibody-functionalized hydrogels, the substrates were rinsed two times in phosphate-buffered saline (Corning) with 0.1% Tween-20 (Sigma) (PBST buffer), then stored at 4 °C at least overnight before use. For multiplexing studies, the different hydrogels were fabricated sequentially in the same substrate. After fabrication of the hydrogels with the first monomer, the substrate was rinsed two times in 500 pL of IX TET buffer, then placed into 50 pL of the subsequent monomer. After vortexing for 30 seconds, the substrate sat in the monomer for 10 minutes, to allow time for diffusion into the substrate. The next set of hydrogels was then fabricated as described previously.

5.4.4 Biotin-streptavidin assay

For the assay with biotin-streptavidin, 50 pL of 1X TET buffer with 50 mM NaCl (Rinse buffer) were placed in a 0.7 mL Eppendorf tube and 5 pL of a 20 g/mL solution of streptavidin-phycoerythrin (purchased at 1 mg/mL from Life Technologies) in rinse buffer. The hydrogel-functionalized substrate was then added to the tube and the tube was placed on a thermoshaker (MultiTherm Shaker, Thomas Scientific) for 45 minutes at 21.5 °C and 1500 rpm. After incubation, the substrates were washed twice by removing all liquid in the tube and replacing with 500 IL of rinse buffer, then vortexed for 30 seconds. The substrates were then imaged.

5.4.5 miRNA assay

For running the miRNA assay, hydrogel features were fabricated in glass fiber as previously described, using a 300 ms exposure time for polymerization. After fabrication, they were rinsed two times with 500 jL IX TET buffer and stored in IX TET at 4 °C until use. The hydrogels were always fabricated on a different day from testing and

92 allowed to rest at least overnight before use. To conduct the hybridization step, the fibrous substrate was placed into a tube containing 40 pL of 427.5 mM NaCl and 5 IL of synthetic target in 1X TE. The tube was placed on a thermoshaker at 55 °C and 1500 rpm for 90 minutes. After hybridization, the substrate was rinsed by removing all volume in the tube and replacing with 500 pL of rinse buffer, then vortexing for 30 seconds. The substrate was rinsed a second time, then placed in a ligation buffer containing 0.05% Tween-20, 8.8 mM NaCl, 33 nM A12 universal linker (purchased from Integrated DNA Technologies, sequence given in Supporting Information), 8% (v/v) NEBuffer 2 (New England Biolabs), 200 pM ATP (New England Biolabs), and 658 U/mL T4 DNA Ligase (New England Biolabs). The tube was placed on a thermoshaker at 21.5 C for 30 minutes. After ligation, the substrate was rinsed two times in rinse buffer and added to the labeling buffer, which consisted of 1.8 pg/mL streptavidin-R-phycoerythrin (Life Technologies) in rinse buffer. The labeling step was conducted on a thermoshaker at 21.5 4°C and 1500 rpm for 45 minutes. After labeling, the solution was rinsed two times and imaged.

5.4.6 Assay for thyroid-stimulating hormone

For the protein assay for thyroid-stimulating hormone (TSH), 50 pL TSH standard (Monobind TSH Accubind ELISA kit) and 0.1 pL 10% T-20 were added to a 0.65 mL Eppendorf tube (VWR). Phosphate buffered saline (Corning) was used as a negative control. The glass fiber substrate containing the particles was added to the tube and the tube was incubated for 15 minutes at 21.5 °C on a thermoshaker at 1500 rpm. The substrate was then rinsed twice by removing as much volume as possible and replacing with 500 pL PBST buffer, then vortexing for 2 minutes. After the two rinse steps, as much volume as possible was removed from the tube and 50 PL of 11 pg/mL detection antibody were added. The detection antibody was Biospacific Anti-TSH 5409 SPTNE-5 and was conjugated to Dyomics DY647P4 NHS ester as described previously[188] before use. The substrate and the detection antibody were then incubated on a thermoshaker at 1500 rpm and 21.5 °C for five minutes. The substrate was then rinsed three times in PBST-PEG (PBST buffer with 1% PEG

93 400) and resuspended in PBST prior to imaging.

5.4.7 Imaging

The substrates were imaged on a Zeiss Axio Observer Al inverted microscope. For fluorescence images, a broad-spectrum LED (X-CITE LED120, Excelitas Technolo- gies) was used as a light source with a fluorescent filter set and an Andor Clara CCD camera and Andor SOLIS software was used for image acquisition. For studies with YG beads, the filter set was Omega Optical XF100-3, while for studies with red beads and protein studies, the filter set was a Semrock FF660-DiOl filter (dichroic), a FF01-628/40 filter (exitation), and a FF01-692/40 filter (emission). For all studies with biotinylated probe or miRNA capture probe, the fluorescent filter set used was Omega Optical XF101-2. Color brightfield images were acquired using a Nikon D7000 camera and Nikon Camera Control Pro 2 software.

94 Chapter 6

Conclusions and Outlook

In this thesis, we have developed hydrogels for point-of-care bioassays, incorporating both theoretical analyses of the key factors that influence bioassay signal for rapid assays and application of these principles to systems for both protein and miRNA detection.

6.1 Conclusions

In Chapter 3, we described an analysis of the flux of target molecule into the hydrogel and used this to develop relationships for the resulting bioassay signal. In contrast to previous studies that focused on long incubation times[162, 113], we focused on shorter timescales applicable for point-of-care diagnostics. We found that the key factors that influence signal from rapid bioassays are the rate of reaction, the diffusivity of target in the hydrogel, the amount of probe incorporated into the hydrogel, the target concentration, the assay time, and the ratio of surface area to imaging area. Of these, the easiest to manipulate is the area ratio. By changing this ratio, we were able to reduce the limit of detection in a protein assay by a factor of six. After developing theory for hydrogel signal from point-of-care bioassays, in Chap- ter 4 we developed a point-of-care assay for miRNA that could measure miRNA directly in serum in a 40 minute assay. Previous studies that were able to directly detect miRNA in serum have not demonstrated multiplexing, which is critical to be

95 able to detect panels of miRNA for disease diagnosis[83, 105]. We used the theoretical analysis from Chapter 3 to design the hydrogels, creating small hydrogel posts in a microfluidic channel. By designing the device to have a high flow rate, we ensured that the target concentration at the edge of the hydrogel remained as high as possible, which also allowed us to reduce the time required for the other assay steps. Using this approach, we were able to attain limits of detection below 100 fM for miR-21 and miR-451 and to measure both of these miRNA simultaneously in serum, along with positive and negative controls. This strategy only required preheating of the serum and did not require RNA extraction, making it useful for point-of-care applications. We then leveraged our experience with hydrogels in Chapter 5 to incorporate them into fibrous substrates for biomolecule detection. Directly polymerizing hydrogels into fibrous substrates takes the advantages of hydrogels such as easy functionalization and combines them with the advantages of fibrous substrates such as wicking and spatial encoding. Fibrous substrates are commonly used in point-of-care applications due to the capillary-driven flow within such substrates that eliminates the need for external pumping. In this thesis, we demonstrated direct polymerization of hydrogels of a variety of shapes and sizes within the substrates and showed that on a glass fiber substrate we could use the glass fiber as a handle for hydrogel particles to run a bioassay. We applied this strategy to demonstrate quantitative detection of miRNA on hydrogel-functionalized glass fiber and showed that hydrogels polymerized inside glass fiber can also be functionalized for protein detection, which we demonstrated by an assay for thyroid-stimulating hormone.

6.2 Future Work

6.2.1 Design of Hydrogel Morphology

In the future, the theoretical understanding of bioassay signal could be applied to additional assays and used to further improve hydrogel-based point-of-care assays. Although this study implicated the importance of diffusivity and probe incorpora-

96 tion for hydrogel signal, we did not investigate different hydrogel formulations. By changing the porogen type or concentration, the diffusivity could be increased. One challenge with this approach is that often with antibody-based studies, if the antibody is directly incorporated into the hydrogel, it is already near solubility limits in the monomer, and increasing the amount or molecular weight of porogen can cause the antibody to phase separate. This could be circumvented by functionalizing with the antibody after polymerization[122, 211]. Optimizing the hydrogel formulation requires balancing the tradeoffs of cross-linking density (reduced diffusivity at higher cross-linking densities but higher probe incorporation). This becomes particularly interesting at very small scales, as the polymerization process itself is increasingly important. As the gels are polymerized, there is a Damkhler number that governs the reaction-diffusion process of the inonomer[231]. At small length scales, the diffusion time becomes faster and the resulting hydrogel is less uniform, impacting its subsequent assay performance. This is also influenced by the intensity of the UV light used for polymerization as well as the exposure time. Optimizing these complex factors would be an interesting future study, as it could also give insight into better hydrogel design for probe incorporation, as current formulations only incorporate 10-20% of the probe in the monomer[162, 3].

6.2.2 Point-of-Care Detection of miRNA

In the future, this microfluidic miRNA detection method could be used to detect panels of miRNA for disease diagnosis. Currently, the limit of detection of less than 100 fM is sufficient to detect abundant miRNA in serum, but not to detect miRNA of lower abundance, such as miR-141, which is expressed below 3 fM in normal human serum. The limit of detection could be improved by further optimizing the ligation step of the assay. Currently, the step relies on a biotinylated nucleic acid linker that then binds to streptavidin-phycoerythrin for fluorescent detection. By switching to a streptavidin molecule attached to a linker rather than biotin and then labeling with biotin-phycoerythrin, the signal could be increased four-fold, as streptavidin has four binding sites for biotin. Additionally, by increasing the linker concentration

97 the ligation could proceed to completion in a shorter time, increasing the signal and reducing the time of the assay. In order to implement this strategy, the researcher would have to conjugate the universal linker DNA sequence to a streptavadin protein themselves, as commercial providers such as Integrated DNA Technologies do not produce streptavidin-functionalized nucleic acids[232. Manipulating the hydrogel formulation or further increasing the probe concentration could also increase the signal. However, in order to detect miRNA such as miR-141 in serum, orders of magnitude improvement would be required, which will require some type of signal amplification, such as the enzymatic amplification scheme that has been previously been demonstrated on hydrogel particles[128, 126]. The assay could also be applied in future work for cancer diagnosis. Some possibilities are breast cancer, in which average miR-451 concentrations are upregulated by around a factor of 10 and average miR-21 concentrations are upregulated by around a factor of three[201]. Breast cancer is a particularly interesting target due to its high prevalence and a lack of screening in developing countries[233]. In future work, this microfluidic device could also be integrated with a smartphone-based fluorescent reader[234] and resistive heating[199] to create a smartphone-powered point-of-care device for use in developing countries. Due to the nonfouling nature of the hydrogels, we believe that this device could also be utilized to enable detection of miRNA in whole blood, which could enable detection in locations where there is no access to a centrifuge.

6.2.3 Integration of Hydrogels into Fibrous Substrates

Hydrogel-functionalized substrates could be applied to enable additional functionalization options for lateral flow or dipstick assays, such as capture moities for mRNA, DNA, or small molecules, which have previously been detected with hydrogels[121, 2, 4]. Because we have previously demonstrated that hydrogels can be functionalized with nucleic acid probes[1], aptamers[114], or antibodies[3], a broad suite of diagnostic tests could be run on the hydrogel-functionalized fibrous susbtrates. In future work, both these new functionalizations and the hydrogels functionalized for miRNA

98 and proteins could be tested in lateral flow and dipstick formats. These constructs should also be tested in complex samples such as serum or whole blood and applied to disease diagnosis. Using such functionalized strips sequentially as dipstick assays could be used to detect multiple analytes (such as proteins and nucleic acids) from the same small sample (<100 pL), since combined panels can increase diagnostic effectiveness[235].

6.3 Outlook

In this thesis, we have developed hydrogel-based point-of-care detection schemes for both proteins and miRNA. Due to the nonfouling nature of hydrogels and their ability to be functionalized with different capture molecules at high density, they show great promise as a substrate for point-of-care diagnostics. As the field of point-of-care diagnostics progresses, there is a push towards increased sensitivity and better porta- bility. Although one can often be improved at the expense of the other, developing devices that can reliably diagnose diseases from complex samples without the need for external equipment can vastly improve health outcomes in both the developed and developing world. In the developing world, such devices can enable diagnosis and therefore treatment of diseases in regions that have been traditionally underserved by health facilities, increasing lifespans and quality of life. In the developed world, as chronic diseases become more and more prevalent, empowering patients to test for biomarkers in the comfort of their own homes can increase quality of life and expected lifespans substantially. By continuing to work to develop portable and sensitive diagnostic devices, researchers can make diagnostics more accessible and improve patient care.

99 100 Appendix A

Supplementary Information: Design of Hydrogel Particle Morphology for Rapid Bioassays

This appendix contains supplementary information for Chapter 3. Reproduced with permission from: Sarah J. Shapiro, Dhananjay Dendukuri, and Patrick S. Doyle. Design of Hy- drogel Particle Morphology for Rapid Bioassays. Anal. Chem., 90(22):13572-13579, November 2018. Copyright 2018 American Chemical Society.

101 Supporting Tables

Supporting Table 1 Monomer composition for ka studies Supporting Table 2 Parameters for calculation of target captured per imaging area Supporting Table 3 Parameters for calculation of Damkbhler number Supporting Table 4 Monomer Composition for TSH wheel particles

Supporting Figures

Supporting Figure 1 Determination of ka Supporting Figure 2 Varying hydrogel particle height Supporting Figure 3 Predicted target per unit imaging area in 1-D and 3-D cases Supporting Figure 4 Experimental signal from posts with overlapping boundary layers Supporting Figure 5 Wheel particles with different monomer composition

102 Table A.1: Monomer composition for ka studies

Component Composition (% v/v) Polyethylene glycol diacrylate 700 20 Polyethylene glycol 200 40 3X Tris-EDTA buffer 35 Darocur 1173 5

A.1 Estimating the Forward Reaction Rate Constant

To estimate the forward reaction rate, hydrogel posts of 10 pm radius were formed inside a commercial 50 pm tall glass microfluidic channel (Hilgenberg GmbH). Prior to hydrogel post formation, the channels were activated by the addition of 1 M sodium hydroxide for one hour, followed by incubation with a solution of 2% 3- (trimethoxysilyl)propyl acrylate (Sigma) 73.5% ethanol (KOPTEC) and 24.5% phosphate buffered saline (PBS, Cellgro). After rinsing with PBS, a monomer solution (Table A.1) was diluted 9:1 with 100 pM biotintylated probe (IDT, sequence: 5'- Acrydite - ATA GCA GAT CAG CAG CCA GA - Biotin - 3') and added to the channel. The solution was polymerized with a UV LED through a photomask (Fineline

Imaging) positioned in the field stop of the microscope to generate posts with - 10 pm diameter. We chose this size due to experimental observations that boundary layers for the reaction of streptavidin-phycoerythrin with biotinylated particles were > 10 pm. By creating posts with radii smaller than the boundary layer, the rate of reaction will be faster than the rate of diffusion, enabling us to measure the rate of reaction.

After washing with 1 X Tris-EDTA buffer, a solution of 50 pg/L streptavidin- phycoerythrin was added to the channel through a cut pipette tip (Figure A-1). After initial flushing with a syringe, the solution was allowed to flow under gravity while fluorescent images were captured on an Andor Clara CCD camera. A broad- spectrum X-Cite 120LED was used as the lightsource and passed through a dichroic filter set (XF101-2, Omega Optical). Images were acquired at 1.6 second intervals using MicroManager software to acquire images and turn on the LED simultaneously, to limit photobleaching.

103 5 A Bs Pst ::s.1

C 3

0 0 20 40 60 80 100 Time (s) Figure A-1: Determination of ka. (A) Experimental setup. Streptavidin- phycoerythrin was added to a microfluidic channel containing 10 pm hydrogel posts with a biotinylated probe. (B) Plot of the ratio of the hydrogel post fluorescence to the channel fluorescence as a function of time. A rate constant was extracted from the linear region of the graph. The lines show linear fits for each post.

At short reaction times, such that the forward reaction dominates, the formation of target-probe complex can be described by Equation A.1, assuming the reaction (and not diffusion) is rate limiting:

aTP]=ka[P][T] (A.1) at Since the probe is in great excess of the target, and the target is being constantly replenished, at short times, both [T] and [P] should be roughly constant, allowing integration of Equation A.1. Solving for the ratio of the target probe complex to the target in the channel, the value of the rate constant can be obtained from the slope of the linear region of a plot of [TP]/[T] vs. time, assuming the initial probe concentration is known:

[TP]= ka[P]t (A.2) [T] From the experimental data (Figure A-2B) we obtained a slope of 0.102 s-' for Post 1 and 0.107 s-1for Post 2, using the first 11 data points in each series. Using an estimated 10% incorporation of probe into the hydrogel[162], we obtain a rate constant

104 of 105 M-1s-1 for the reaction of streptavidin-phycoerythrin with the immobilized biotin. This is considerably smaller than the previously reported rate constant of 3.0 x 106 - 4.5 x 10 M-1 s-'[236], which we anticipate is due to the combined steric effects of the bulky fluorophore (- 240 kDa for phycoerythrin) and the hydrogel environment.

A.2 Estimating the Target Depletion

To estimate whether the target was significantly depleted over the course of the reaction, we can estimate the flux into each particle. The worst case for target depletion is that the solution concentration remains at the initial value, driving target into the hydrogels. By calculating the flux under these conditions, we can obtain an upper limit for the estimate. The flux is also highest under conditions of high Damkhler number, where a boundary layer forms around the edges of the particle. Under these conditions, the flux can be described by Equation A.3, where Dgel is the diffusivity of the target in the hydrogel, T,,o is the initial target concentration in solution (assuming a partition coefficient of one), Da is the Damkhler number, and L is the distance from the center of the particle to the edge.

T,Dai A3 J ~ Dge L(A.3)

To obtain the total amount of target that has been depleted from the solution, we can multiply the flux by the total surface area per particle, the number of particles, and the time, assuming a constant flux. We obtain Equation A.4, where Po is the initial probe concentration in the hydrogel, ka is the forward reaction rate constant, t is the assay time, and Np is the number of particles.

2 Depletion ~ (21rRH + 21rR )D 1elkaPo2Ts,oNpt (A.4)

We are ultimately interested in how the depletion compares to the total amount of target present initially in solution, which we can calculate by dividing by T,oV,

105 where V is the volume of the solution.

In our assay, t - 900 s, R ~ 100pm, H ~ 150pm, De 5 x 10-11 P0 106

3 M, T,o ~ 10-10 M, k 101 Mls-', V ~ 50 x 10-i m . Using these values, we obtain:

1 1 1 2 (2irRH + 21rR )D2 1ksPosNp t DepletionRatio ~ ~ 0.06 (A. 5)

The target concentration is depleted by less than 6% in the biotin/streptavidin phycoerythrin binding assay, indicating that a constant target concentration assumption is justified.

A.3 Varying Hydrogel Height

Biotinylated particles were fabricated as described in the main text. Microfluidic channels of various heights were used to fabricate hydrogels of different heights and constant radius, thereby maipulating the ratio of the height (H) to the radius (R). As expected, the signal increases with increasing H/R, as shown in Figure A-2.

A.4 Signal from Particles with Overlapping Bound- ary Layers

A plot of Equation 14 in the main text is shown below, illustrating that in the 1-D case the signal increases with decreasing radius, even as the boundary layers begin to overlap. Below DaR~ 0.1, there are diminishing returns for decreasing the radius further. Parameters used to calculate the signal per unit imaging area are listed in Table A.2. COMSOL Multiphysics 4.2 was used to calculate the 3D signal profile as a function of Damk6hler number. For all simulations, the Transport of Diluted Species module was used with a 2D axisymmetric model. For the target concentration, a Transport of Diluted Species model was used with diffusion enabled, while for the

106 250

200

03 .I ~150

100 .

( 50 2 0 0.5 1 1.5 2 H/R Figure A-2: The mean signal increases with increasing ratio of the height to the radius. Data shows mean signal from four biotinylated particles of varying heights after reaction for 15 minutes with 50 pg/L streptavidin-phycoerythrin. The radius of all particles was 100 pm. Error bars represent standard deviations.

400 6.43D coMsoL 350 . -- 1-D Theory

300

E .0 250

200

~150 0 4 1.6 3.6 6.4 10P DaR 0 20 40 60 80 100 120 R (pm) Figure A-3: Average signal in the particle increases with decreased radius in the 1-D case (theory) and the 3-D case (COMSOL Multiphysics simulations). At low DaR, there is lower marginal signal increase from reducing the radius.

107 Table A.2: Parameters for 1-D calculations and 3-D COMSOL simulations Component Value Height (H) 150 pm Forward reaction rate (ka) 102 M3 Reverse reaction rate (kd, used only for COMSOL) 2.4 x 106} Diffusivity (Dgei) 5 x 10-

Initial probe concentration (PO) 5 x1- 3 Time (t) 15 minutes Initial target concentration (T8,o) 5.56 x108 target-probe complex, an additional Transport of Diluted Species model was added with the diffusion set to zero. The probe concentration was calculated from a the amount of target-probe complex using a mass balance on the total amount of probe. To calculate the amount of target captured, a surface integral with volume enabled was used on the 2-D axisymmetric surface. For the simulations, the radius was varied while the parameters in Table A.2 were held constant.

Although for the 1-D case decreasing the radius as much as possible results in the highest signal, we desired to determine what would happen in a 3-D case, when both the top and bottom surfaces of the particle were exposed to the target, in addition to the sides. To do this, we used COMSOL Multiphysics simulations to determine what would happen to the amount of target bound as the boundary layers overlapped. In this case as the radius is decreased, in addition to reducing the flux through the sides of the particle, the flux through the top and bottom will also be reduced compared to the case with non-overlapping boundary layers, since the target concentration at the center of the particle is non-zero. The results of the COMSOL simulations are shown in Figure A-3. Similar to the 1-D case, the target bound per unit imaging area increases with decreasing radius, even at Da<1, when the boundary layers overlap. However, as the radius becomes smaller and smaller, decreasing Da much less than one, the impact of the decreasing radius on the signal diminishes, particularly below Da ~ 0.1. At high DaR, there is significant deviation between the 1-D theory and the 3-D simulations due to the fact that the theory does not consider flux through the top and bottom surfaces of the hydrogel, but the simulation considers flux through

108 Table A.3: Parameters for calculating Damkdhler number for posts with overlapping boundary layers Component Value Forward reaction rate (ka) 102 M Diffusivity (Dgei) 5.6 x 101"

Initial probe concentration ( 0 ) 1 x 10-4 m°°

all faces. At higher radius (corresponding to higher DaR), the deviation between the theory and the simulations is expected to increase due to larger areas of the top and bottom surfaces.

We also experimentally validated the trend of increasing signal with decreasing radius by creating hydrogel posts with overlapping boundary layers in a microfluidic channel. Posts of varying radii were formed inside a commercial 50 pm tall glass microfluidic channel (Hilgenberg GmbH) that was treated as described above. After rinsing with PBS, a monomer solution (Table A.1) was diluted 9:1 with 10 tM biotintylated probe (IDT, sequence: 5'- Acrydite - ATA GCA GAT CAG CAG CCA GA - Biotin - 3') and added to the channel. The solution was polymerized with a UV LED through a photomask (Fineline Imaging) positioned in the field stop of the microscope to generate posts with 100, 50, or 20 pm radius. After washing with 1 X Tris-EDTA buffer, a solution of 50 pg/L streptavidin-phycoerythrin was added to the channel through a cut pipette tip. After a 30 minute assay with the streptavidin- phycoerythrin, the channels were washed again with 1 X Tris-EDTA buffer and imaged. Fluorescent images were captured on an Andor Clara CCD camera using a broad-spectrum X-Cite 120LED lightsource and a dichroic filter (XF101-2, Omega

Optical). The signal from the posts increased with decreasing radius (Figure A-4A), even when the boundary layers were overlapping, as can be seen visually from Fig- ure A-4B). The parameters for calculating the Damkdhler numbers in Figure A-4 are shown in Table A.3.

109 A B 250 R=20 pm

. ~200 R=50 pm

150

R=100 pm 100 .

0.07 0.3 0.6 1.1 1.8 DaR 0 20 40 60 80 100 120 R (pm)

Figure A-4: Experimental signal from posts with overlapping boundary layers (A) Average signal in the hydrogel post increases with decreased radius. (B) Fluorescent images of the posts that visually show the overlapping boundary layers. Scale bar is 20 pm.

Table A.4: Monomer composition for TSH wheel particles

Component Original Composition Modified Composition Polyethylene glycol diacrylate 700 18 (% v/v) 18'(% v/v) Polyethylene glycol 200 36 (% v/v) 16 (% v/v) Darocur 1173 4.5 (% v/v) 1.5 (% v/v) Capture antibody 0.4 mg/mL 0.4 mg/mL

A.5 Wheel Particles with More Uniform Signal

We anticipated that the non-uniform signal observed in the TSH wheel particles was due to antibody phase separation in the monomer due to the high concentration of polyethylene glycol (PEG). To explore this, we reduced the amount of PEG in the monomer and compared the fluorescent particle images. Due to the role of PEG in increasing the solubility of the Darocur photoinitiator, we decreased the photoinitiator concentration simultaneously with the PEG concentration. The formulation of both particles is shown in Table A.4. Due to the reduced photoinitiator concentration, we increased the polymerization time from 100 ms to 400 ms for the new particles.

Reducing the PEG content of the monomer significantly improved the splotchy

110 A B

Figure A-5: Wheel particles with different monomer composition. The images were thresholded to different values to better illustrate the degree of signal uniformity for each particle. (A) Hydrogel particle with original composition after reaction with 10 pIU/mL TSH. (B) Hydro- gel particle with reduced polyethylene glycol after reaction with 2.5 pIU/mL TSH. Scale bar is 50 pm. signal in the particles, as is visible in Figure A-5. We note that the antibody used for the new particles was old, which may have contributed to the remaining splotchy signal. We anticipate that the signal variations could be removed entirely by reducing the PEG concentration further.

111 112 Appendix B

Supplementary Information: Rapid, Multiplex, On-chip Detection of miRNA Directly from Serum

This chapter contains supplementary information for Chapter 4.

113 Supporting Tables Supporting Table 1 Nucleic acid sequences

Supporting Figures

Supporting Figure 1 Multiplexing Supporting Figure 2 Validation of buffer Supporting Figure 3 Comparison of assay conditions

114 Table B.1: Nucleic acid sequences. /5Acryd/ indicates a 5' Acrydite modification, /3InvdT/ indicates a 3' inverted dT modification, /5Phos/ indicates a 5' phosphorylation, and /3Bio/ indicates a 3' biotin modification Title Sequence miR-21 5'- /5Acryd/ probe GAT ATA TTT TAT CAA CAT CAG TCT GAT AAG CTA /3InvdT/ -3' miR-451 5'- /5Acryd/ probe GAT ATA TTT TAA ACT CAG TAA TGG TAA CGG TTT /3InvdT/ -3' miR-19b 5'- /5Acryd/ probe GAT ATA TTT TAT CAG TTT TGC ATG GAT TTG CAC A /3InvdT/ -3' cel-miR-39 5'- /5Acryd/ probe GAT ATA TTT TAC AAG CTG ATT TAC ACC CGG TGA /3InvdT/ -3' cel-miR-54 5'- /5Acryd/ probe GAT ATA TTT TAC TCG GAT TAT GAA GAT TAC GGG TA //3InvdT/ -3' miR-21 5'- rUrArG rCrUrU rArUrC rArGrA rCrUrG rArUrG rUrUrG rA -3' target miR-451 5'- rArArA rCrCrG rUrUrA rCrCrA rUrUrA rCrUrG rArGrU rU -3' target miR-19b 5'- rUrGrU rGrCrA rArArU rCrCrA rUrGrC rArArA rArCrU rGrA -3' target cel-miR-39 5'- rUrCrA rCrCrG rGrGrU rGrUrA rArArU rCrArG rCrUrU rG -3' target Universal 5'- /5Phos/TAA AAT ATA TAA AAA AAA AAA A/3Bio/ -3' linker

B.1 Nucleic Acid Sequences

All nucleic acid sequences were ordered from Integrated DNA Technologies (Coralville, IA). The specific sequences used are listed in Table B.1.

B.2 Validation of multiplexing

To validate multiplexing in the microfluidic channel, we sequentially formed hydrogels with and without a biotinylated probe. We made six sets of hydrogels in total, starting with the biotinylated hydrogels, and alternating between the biotinylated

115 5000 4500 4000 i3500 S3000 S2500 9 2000 1500 Ca1000 500 0 Biotin3 Negative3 -500 BiotinI Negative1 BiotIn2 Negative2

Figure B-1: Comparison of hydrogel signal from different sets of hydrogel posts. n=5 posts for each data point and error bars represent one standard deviation. and negative hydrogels. We then tested the posts by flowing 20 pg/mL streptavidin- phycoerythrin through the channel for five minutes. The signal from the biotinylated posts were all similar, and the signal from all three sets of negative posts was close to zero, as shown in Figure B-1.

B.3 Validation of buffer

The hybridization step for serum assays was run in a custom buffer[110] with 10 mM Tris-HCl, 1 mM EDTA, 0.05% Tween-20, 2% (w/v) sodium dodecyl sulfate (Sigma- Aldrich), 28 U/mL Proteinase K (New England Biolabs), and 350 mM NaCl. To calculate the final salt concentration of 350 mM in the buffer, we assumed a 150 mM NaCl concentration in serum [1891 as has been done previously[106]. For the hybridization step for all other assays, the step was conducted in IX TE with 0.05% Tween-20 and 350 mM NaCl. To validate that we obtained a similar signal in both buffers, we compared the signal obtained from the cel-miR-39 spike-in

116 250 -

200

cc MC100

0 Serum Buffer

Figure B-2: Comparison of hydrogel signal from serum assay and neat buffer. positive control in serum with detection of synthetic cel-miR-39 in neat hybridization buffer. For both assays, the concentration of cel-miR-39 was 3.2 pM. We found that the signal obtained from the serum spike-in and the neat buffer was similar (Fig.

B-2).

B.4 Comparison of assay conditions

To investigate the impact of various assay parameters on bioassay signal directly from serum, we varied two different parameters. The first was the soaking time that we allowed the channels to sit in 1X TE buffer after oxidation of the posts with potassium permanganate and before the blocking step with Pluronic F-108.

We found that the detection of the positive control, cel-miR-39, was impacted by soaking time. Soaking for at least four hours increased the signal compared to running the assay immediately. The second parameter was a pre-incubation of the serum with Proteinase K. The proteinase K degrades the protein complexes that bind some mriRNA in serumn, releasing themn to bind with the target. For the pre-incubation, we

117 300 -

250

00 200 TNo soak, Pre-incubation with Proteinase K 150 m 4 hour soak, pre- incubation with o100 Proteinase K nOvernight soak, no pre- l50 . incubation with Proteinase K

CP CP

Figure B-3: Comparison of hydrogel signal from different assay conditions. Soak refers to leaving the channel in 1X TE buffer prior to the assay. "Pre-incubation with Proteinase K" refers to incubating the serum with Proteinase K prior to beginning the assay. Each data point is an average of 4-5 hydroogel posts and error bars represent one standard deviation.

added every component of the lysis buffer except the positive cel-miR-39 target and incubated for five minutes at 55 C in order to begin the protein cleavage prior to the assay. We found that this step had a large impact on the miR-451 signal, so we included this step as part of the assay protocol.

118 Appendix C

Supplementary Information: Incorporation of Hydrogels into Fibrous Substrates for Quantitative, Multiplexed Bioassays

This chapter contains supplementary information for Chapter 5.

119 Supporting Tables Supporting Table 1 Nucleic acid sequences

Supporting Figures Supporting Figure 1 Setup for measuring light intensity Supporting Figure 2 Analysis of signal in biotinylated particles Supporting Figure 3 Image of substrate from exposure time tests Supporting Figure 4 Image of label from commercially-purchased silk

C.l Comparison of light attenuation through all three substrates

To measure the light intensity after passing through each substrate, we used a Thor- Labs PM100USB light intensity meter. We placed the edges of, the light meter on top of two layers of Parafilm on a glass slide and placed the slide on an inverted microscope. Each sheet of Parafilm was 130 pm thick, so the combined thickness was thicker than any of the three substrates, the thickest of which was measured to be 200 pm. We first measured the intensity of the UV LED used for polymerization at 365 nm. To measure the intensity of the light that passed through a substrate, the substrate was placed on the glass slide between the parafilm sheets, with the light meter placed above so that the edges rested on the Parafilm. In this way the meter was a constant height above the slide. Under light from the ThorLabs LED used for polymerization, we measured the intensity with no substrate and with each of the substrates placed between the slide and the meter, in order to calculate the percent light transmittance. A schematic of the setup is shown in Figure C-1.

C.2 Comparison of signal within and outside of glass fiber

The signal from biotinylated particles after reaction with streptavidin-phycoerythrin is similar within and outside of glass fiber substrate. Figure C-2(A) shows a fluores-

120 Light meter -"Parafilm ' Slide

Microscope Objective

Figure C-1: Setup for measuring light intensity

A B 2000 ,_,_ ,

Cd 1500

1000

0 200 400 600 80 Distance (pm) Figure C-2: Analysis of signal in biotinylated particles. (A) Fluorescent image of hydrogel particle overhanging the edge of glass fiber substrate. Scale bar is 100 pm. Yellow line indicates location of cross cut for fluorescence analysis. (B) Plot of fluorescence at the location indicated by the line in (A). The signal is similar within and outside of the glass fiber substrate. cent image of the hydrogel particle. The glass fiber substrate is visible on the right half of the image. The yellow line indicates the location of the fluorescent plotted in Figure C-2(B). The boundary layer on the edge of the particle outside of the glass fiber substrate is brighter than the boundary layer on the inside, likely due to better mass transfer.

C.3 Exposure time studies

This section contains an image of the hydrogel-functionalized glass fiber substrate used for the exposure time tests. A cropped version of Figure C-3 is shown in Figure

121 Figure C-3: Fluorescence image of glass fiber substrate with biotin-functionalized hydrogels after reaction with streptavidin-phycoerythrin. The hydrogels were polymerized at different exposure times (100-500 ms).

4 in the main text.

C.4 Information on silk used in this study

The silk used in this study was purchased in-person at the Mysore Saree Udyog showroom (# 316 1st Floor, Mahaveer Mall, K Kamaraj Road, Bangalore 560042, India). An image of the label is shown in Figure C-4.

C.5 Discussion of limit of detection

The limit of detection of 28 amol is about an order of magnitude higher than what we have previously obtained with our hydrogel particle systems[1, 110], but comparable to other published particle systems for detection of miRNA[191]. We believe that the higher limit of detection is due to a combination of factors. The fibrous substrate induces light scattering during both the polymerization and fluorescence imaging processes, which can reduce the extent of polymerization and reduce the effective

122 8RE 8AEEUDYOG LLP 3760146 KPOJfhllII~ui BOOMAILSHAI

Figure C-4: Label from silk used in this study fluorescent signal, by scattering it outside the hydrogel. In addition, this scattering effect likely increases the noise; we noticed that the noise was higher than we typically see for particle-based assays. The fibrous substrate also adds an additional mass transfer limitation, as targets must diffuse into the fibrous substrate to reach the edge of the hydrogel.

C.6 Nucleic acid sequences

The sequences in Table C.1 were ordered from Integrated DNA technologies for use in this study.

123 Table C.1: Nucleic acid sequences used in this study Name Sequence Biotinylated 5'- Acrydite - ATA GCA GAT CAG CAG CCA GA - Biotin -3' Probe Universal linker 5'- Phosphorylation - TAA AAT ATA TAA AAA AAA AAA A - Biotin -3' miR-210 probe 5'- Acrydite - GAT ATA TTT TAT CAG CCG CTG TCA CAC GCA CAG - Inverted dT -3' miR-141 probe 5'- Acrydite - GAT ATA TTT TAC CAT CTT TAC CAG ACA GTG TTA - Inverted dT -3' miR-210 target 5'- rCrUrG rUrGrC rGrUrG rUrGrA rCrArG rCrGrG rCrUrG rA -3' rniR-141 target 5'- rUrArA rCrArC rUrGrU rCrUrG rGrUrA rArArG rArUrG rG -3'

124 Bibliography

[11 Stephen C. Chapin, David C. Appleyard, Daniel C. Pregibon, and Patrick S. Doyle. Rapid microRNA Profiling on Encoded Gel Microparticles. Angew. Chem. Int. Ed., 50(10):2289-2293, 2011. [2] Ajfan Baeissa, Neeshma Dave, Brendan D. Smith, and Juewen Liu. DNA- Functionalized Monolithic Hydrogels and Gold Nanoparticles for Colorimet- ric DNA Detection. ACS Appl. Mater.Interfaces, 2(12):3594-3600, December 2010. [3] David C. Appleyard, Stephen C. Chapin, and Patrick S. Doyle. Multiplexed Protein Quantification with Barcoded Hydrogel Microparticles. Anal. Chem., 83(1):193-199, January 2011. [4] Yanli Ma, Yu Mao, Yuan An, Tian Tian, Huimin Zhang, Jinmao Yan, Zhi Zhu, and Chaoyong James Yang. Target-responsive DNA hydrogel for non-enzymatic and visual detection of glucose. Analyst, 143(7):1679-1684, March 2018. [5] Byron H. Mainor, William E. Hardwick, and William D. King. Evaluation of a portable clinical analyzer in the pediatric emergency department: analysis of cost and turnaround time. South. Med. J., 95(6):634-638, June 2002. [6] World Health Organization and International Bank for Reconstruction and De- velopment. Tracking universal health coverage: 2017 global monitoring report. Technical report, The World Bank, Geneva, 2017. [7] Andre Terzic and Scott Waldman. Chronic Diseases: The Emerging Pandemic. Clin Transl Sci, 4(3):225-226, June 2011. [81 Mitchell Daniels, Thomas Donilon, and Thomas Bollyky. The Emerging Global Health Crisis: Noncommunicable Diseases in Low- and Middle-Income Coun- tries. Technical Report 72, Council on Foreign Relations, 2014. [9] Angelica M. Merlot, Danuta S. Kalinowski, and Des R. Richardson. Unravel- ing the mysteries of serum albumin-more than just a serum protein. Front. Physiol., 5, 2014. [101 N. Leigh Anderson. The Clinical Plasma Proteome: A Survey of Clinical Assays for Proteins in Plasma and Serum. Clinical Chemistry, 56(2):177-185, February 2010.

125 [11] Axel Warsinke. Point-of-care testing of proteins. Anal Bioanal Chem, 393(5):1393-1405, March 2009.

[12] Alan H. B. Wu. A selected history and future of immunoassay development and applications in clinical chemistry. Clinica Chimica Acta, 369(2):119-124, July 2006.

[13] Roche. cobas e 411 analyzer, July 2019.

[14] Clinical Lab Products. Tech Guide: Chemistry Analyzers, 2013.

[15] David P Bartel. MicroRNAs: Genomics, Biogenesis, Mechanism, and Function. Cell, 116(2):281-297, January 2004.

[16] http://www.mirbase.org/cgi-bin/mirna-summary.pl?org=hsa.

[17] Sam Griffiths-Jones. The microRNA Registry. Nucl. Acids Res., 32(suppl 1):D109-D111, January 2004.

[18] Ali M. Ardekani and Mozhgan Moslemi Naeini. The Role of MicroRNAs in Human Diseases. Avicenna J Med Biotechnol, 2(4):161-179, 2010.

[19] Yu Li and Kris V. Kowdley. MicroRNAs in Common Human Diseases. Ge- nomics, Proteomics & Bioinformatics, 10(5):246-253, October 2012.

[20] Rosalind C. Lee, Rhonda L. Feinbaum, and Victor Ambros. The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell, 75(5):843-854, December 1993.

[21] Amy E. Pasquinelli, Brenda J. Reinhart, Frank Slack, Mark Q. Martindale, Mitzi I. Kuroda, Betsy Maller, David C. Hayward, Eldon E. Ball, Bernard Degnan, Peter M6ller, Jiirg Spring, Ashok Srinivasan, Mark Fishman, John Finnerty, Joseph Corbo, Michael Levine, Patrick Leahy, Eric Davidson, and Gary Ruvkun. Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA. Nature, 408(6808):86-89, November 2000.

[22] Shobha Vasudevan, Yingchun Tong, and Joan A. Steitz. Switching from Re- pression to Activation: MicroRNAs Can Up-Regulate Translation. Science, 318(5858):1931-1934, December 2007.

[23] Richard I. Gregory, Thimmaiah P. Chendrimada, Neil Cooch, and Ramin Shiekhattar. Human RISC Couples MicroRNA Biogenesis and Posttranscrip- tional Gene Silencing. Cell, 123(4):631-640, November 2005.

[24] Marco Antonio Valencia-Sanchez, Jidong Liu, Gregory J. Hannon, and Roy Parker. Control of translation and mRNA degradation by miRNAs and siRNAs. Genes Dev., 20(5):515-524, March 2006.

126 [251 Eric Huntzinger and Elisa Izaurralde. Gene silencing by microRNAs: contribu- tions of translational repression and mRNA decay. Nat Rev Genet, 12(2):99-110, February 2011.

[26] David P. Bartel. MicroRNAs: Target Recognition and Regulatory Functions. Cell, 136(2):215-233, January 2009.

[27] Xi Chen, Yi Ba, Lijia Ma, Xing Cai, Yuan Yin, Kehui Wang, Jigang Guo, Yu- jing Zhang, Jiangning Chen, Xing Guo, Qibin Li, Xiaoying Li, Wenjing Wang, Yan Zhang, Jin Wang, Xueyuan Jiang, Yang Xiang, Chen Xu, Pingping Zheng, Juanbin Zhang, Ruiqiang Li, Hongjie Zhang, Xiaobin Shang, Ting Gong, Guang Ning, Jun Wang, Ke Zen, Junfeng Zhang, and Chen-Yu Zhang. Characteriza- tion of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Research, 18(10):997-1006, October 2008.

[28] Jason D. Arroyo, John R. Chevillet, Evan M. Kroh, Ingrid K. Ruf, Colin C. Pritchard, Donald F. Gibson, Patrick S. Mitchell, Christopher F. Bennett, Era L. Pogosova-Agadjanyan, Derek L. Stirewalt, Jonathan F. Tait, and Muneesh Tewari. Argonaute2 complexes carry a population of circulating mi- croRNAs independent of vesicles in human plasma. Proc Natl Acad Sci U S A, 108(12):5003-5008, March 2011.

[29] Patrick S. Mitchell, Rachael K. Parkin, Evan M. Kroh, Brian R. Fritz, Stacia K. Wyman, Era L. Pogosova-Agadjanyan, Amelia Peterson, Jennifer Noteboom, Kathy C. O'Briant, April Allen, Daniel W. Lin, Nicole Urban, Charles W. Drescher, Beatrice S. Knudsen, Derek L. Stirewalt, Robert Gen- tleman, Robert L. Vessella, Peter S. Nelson, Daniel B. Martin, and Muneesh Tewari. Circulating microRNAs as stable blood-based markers for cancer detection. PNAS, 105(30):10513-10518, July 2008.

[30] Christine Mall, David M Rocke, Blythe Durbin-Johnson, and Robert H Weiss. Stability of miRNA in human urine supports its biomarker potential. Biomark Med, 7(4):623-31, August 2013.

[31] Carlo M. Croce. Causes and consequences of microRNA dysregulation in cancer. Nat Rev Genet, 10(10):704-714, October 2009.

[32] Sadakatsu Ikeda, Sek Won Kong, Jun Lu, Egbert Bisping, Hao Zhang, Paul D. Allen, Todd R. Golub, Burkert Pieske, and William T. Pu. Altered microRNA expression in human heart disease. Physiological Genomics, 31(3):367-373, November 2007.

[33] Aaron D. McClelland and Phillip Kantharidis. microRNA in the development of diabetic complications. Clin. Sci., 126(2):95-110, January 2014. [34] Pai-Sheng Chen, Jen-Liang Su, and Mien-Chie Hung. Dysregulation of MicroR- NAs in cancer. Journal of Biomedical Science, 19(1):90, October 2012.

127 [351 Colin C. Pritchard, Heather H. Cheng, and Muneesh Tewari. MicroRNA profiling: approaches and considerations. Nat Rev Genet, 13(5):358-369, May 2012.

[36] Pieter Mestdagh, Nicole Hartmann, Lukas Baeriswyl, Ditte Andreasen, Nathalie Bernard, Caifu Chen, David Cheo, Petula D'Andrade, Mike DeMayo, Lucas Dennis, Stefaan Derveaux, Yun Feng, Stephanie Fulmer-Smentek, Bern- hard Gerstmayer, Julia Gouffon, Chris Grimley, Eric Lader, Kathy Y. Lee, Shu- jun Luo, Peter Mouritzen, Aishwarya Narayanan, Sunali Patel, Sabine Peiffer, Silvia Riberg, Gary Schroth, Dave Schuster, Jonathan M. Shaffer, Elliot J. Shelton, Scott Silveria, Umberto Ulmanella, Vamsi Veeramachaneni, Frank Staedtler, Thomas Peters, Toumy Guettouche, Linda Wong, and Jo Vandesom- pele. Evaluation of quantitative miRNA expression platforms in the microRNA quality control (miRQC) study. Nat Meth, 11(8):809-815, August 2014. [37] Chang-Gong Liu, George Adrian Calin, Stefano Volinia, and Carlo M. Croce. MicroRNA expression profiling using microarrays. Nat. Protocols, 3(4):563-578, April 2008. [38] Bin Wang, Paul Howel, Skjalg Bruheim, Jingfang Ju, Laurie B. Owen, Oystein Fodstad, and Yaguang Xi. Systematic Evaluation of Three microRNA Profiling Platforms: Microarray, Beads Array, and Quantitative Real-Time PCR Array. PLoS One, 6(2), February 2011. 139] Alastair W. Wark, Hye Jin Lee, and Robert M. Corn. Multiplexed Detection Methods for Profiling MicroRNA Expression in Biological Samples. Angewandte Chemie InternationalEdition, 47(4):644-652, January 2008. [40] Swathi Lingam, Madhu Beta, Dhananjay Dendukuri, and Subramanian Krish- nakumar. A focus on microfluidics and nanotechnology approaches for the ultra sensitive detection of microRNA. Microrna, 3(1):18-28, 2014. [41] Paul Yager, Gonzalo J. Domingo, and John Gerdes. Point-of-Care Diagnostics for Global Health. Annual Review of Biomedical Engineering, 10(1):107-144, 2008. [42] Mark A. Cervinski, Christina M. Lockwood, Angela M. Ferguson, Randall R. Odem, Ulf H. Stenman, Henrik Alfthan, David G. Grenache, and Ann M. Gronowski. Qualitative point-of-care and over-the-counter urine hCG devices differentially detect the hCG variants of early pregnancy. Clinica Chimica Acta, 406(1):81-85, August 2009. [43] Ralf Junker, Harald Schlebusch, and Peter B. Luppa. Point-of-Care Testing in Hospitals and Primary Care. Dtsch Arztebl Int, 107(33):561-567, August 2010.

[44] Madeleine H. E. Bruins Slot, Geert J. M. G. van der Heijden, Saskia D. Stelp- stra, Arno W. Hoes, and Frans H. Rutten. Point-of-care tests in suspected acute myocardial infarction: A systematic review. International Journal of Cardiology, 168(6):5355-5362, October 2013.

128 [45] Ali Kemal Yetisen, Muhammad Safwan Akram, and Christopher R. Lowe. Paper-based microfluidic point-of-care diagnostic devices. Lab Chip, 13(12):2210-2251, May 2013.

[46] Thomas R. Kozel and Amanda R. Burnham-Marusich. Point-of-Care Testing for Infectious Diseases: Past, Present, and Future. Journal of Clinical Microbiology, 55(8):2313-2320, August 2017.

[47] Shikha Sharma, Julia Zapatero-Rodriguez, Pedro Estrela, and Richard O'Kennedy. Point-of-Care Diagnostics in Low Resource Settings: Present Sta- tus and Future Role of Microfluidics. Biosensors, 5(3):577-601, August 2015.

[48] Helena de Puig, Irene Bosch, Lee Gehrke, and Kimberly Hamad-Schifferli. Chal- lenges of the Nano-Bio Interface in Lateral Flow and Dipstick Immunoassays. Trends in Biotechnology, 35(12):1169-1180, December 2017.

[49] Vladimir Gubala, Leanne F. Harris, Antonio J. Ricco, Ming X. Tan, and David E. Williams. Point of Care Diagnostics: Status and Future. Anal. Chem., 84(2):487-515, January 2012.

[50] Katarzyna M. Koczula and Andrea Gallotta. Lateral flow assays. Essays Biochem, 60(1):111-120, June 2016.

[51] Lawrence Kulinsky, Zahra Noroozi, and Marc Madou. Present Technology and Future Trends in Point-of-Care Microfluidic Diagnostics. In Gareth Jenkins and Colin D. Mansfield, editors, Microfluidic Diagnostics: Methods and Protocols, Methods in Molecular Biology, pages 3-23. Humana Press, Totowa, NJ, 2013.

[52] Sandeep Kumar Vashist, Peter B. Luppa, Leslie Y. Yeo, Aydogan Ozcan, and John H. T. Luong. Emerging Technologies for Next-Generation Point-of-Care Testing. Trends in Biotechnology, 33(11):692-705, November 2015.

[531 Curtis D. Chin, Vincent Linder, and Samuel K. Sia. Commercialization of microfluidic point-of-care diagnostic devices. Lab on a Chip, 12(12):2118, 2012.

[54] K. A. Erickson and P. Wilding. Evaluation of a novel point-of-care system, the i-STAT portable clinical analyzer. Clinical Chemistry, 39(2):283-287, February 1993.

[55] Ki-Joo Sung, Eric A. Miller, and Hadley D. Sikes. Engineering hyperther- mostable rcSso7d as reporter molecule for in vitro diagnostic tests. Mol. Syst. Des. Eng., 3(6):877-882, December 2018.

[56] E. A. Miller, M. W. Traxlmayr, J. Shen, and H. D. Sikes. Activity-based assessment of an engineered hyperthermophilic protein as a capture agent in paper-based diagnostic tests. Mol. Syst. Des. Eng., 1(4):377-381, November 2016.

129 [57] Abraham K. Badu-Tawiah, Shefali Lathwal, Kaja Kaastrup, Mohammad Al- Sayah, Dionysios C. Christodouleas, Barbara S. Smith, George M. Whitesides, and Hadley D. Sikes. Polymerization-based signal amplification for paper-based immunoassays. Lab Chip, 15(3):655-659, January 2015. [58] Wei Yin Lim, Boon Tong Goh, and Sook Mei Khor. Microfluidic paper-based analytical devices for potential use in quantitative and direct detection of disease biomarkers in clinical analysis. Journal of Chromatography B, 1060:424-442, August 2017. [59] Yuanyuan Yang, Eka Noviana, Michael P. Nguyen, Brian J. Geiss, David S. Dandy, and Charles S. Henry. Paper-Based Microfluidic Devices: Emerging Themes and Applications. Anal. Chem., 89(1):71-91, January 2017. [60] Snober Ahmed, Minh-Phuong Ngoc Bui, and Abdennour Abbas. Paper-based chemical and biological sensors: Engineering aspects. Biosensors and Bioelec- tronics, 77:249-263, March 2016. [61] Kyle Taylor and Laura Silver. Smartphone Ownership Is Growing Rapidly Around the World, but Not Always Equally. Technical report, February 2019. [62] Tassaneewan Laksanasopin, Tiffany W. Guo, Samiksha Nayak, Archana A. Sridhara, Shi Xie, Owolabi 0. Olowookere, Paolo Cadinu, Fanxing Meng, Na- talie H. Chee, Jiyoon Kim, Curtis D. Chin, Elisaphane Munyazesa, Placidie Mugwaneza, Alex J. Rai, Veronicah Mugisha, Arnold R. Castro, David Stein- miller, Vincent Linder, Jessica E. Justman, Sabin Nsanzimana, and Samuel K. Sia. A smartphone dongle for diagnosis of infectious diseases at the point of care. Science TranslationalMedicine, 7(273):273re1-273rel, February 2015. [63] Alexander C. Sun, Chengyang Yao, Venkatesh A.g., and Drew A. Hall. An effi- cient power harvesting mobile phone-based electrochemical biosensor for point- of-care health monitoring. Sensors and Actuators B: Chemical, 235:126-135, November 2016. [64] Dasheng Lee, Wen Pin Chou, Shiou Hwei Yeh, Pei Jer Chen, and Ping Hei Chen. DNA detection using commercial mobile phones. Biosensors and Bioelectronics, 26(11):4349-4354, July 2011.

[65] Manoj Kumar Kanakasabapathy, Hardik J. Pandya, Mohamed Shehata Draz, Manjyot Kaur Chug, Magesh Sadasivam, Shreya Kumar, Behzad Etemad, Vin- ish Yogesh, Mohammadali Safavieh, Waseem Asghar, Jonathan Z. Li, Athe M. Tsibris, Daniel R. Kuritzkes, and Hadi Shafiee. Rapid, label-free CD4 testing using a smartphone compatible device. Lab Chip, July 2017. [661 Jina Ko, Matthew A. Hemphill, David Gabrieli, Leon Wu, Venkata Yelleswarapu, Gladys Lawrence, Wesley Pennycooke, Anup Singh, Dave F. Meaney, and David Issadore. Smartphone-enabled optofluidic exosome diagnostic for concussion recovery. Scientific Reports, 6:31215, August 2016.

130 [67] Stephanie Knowlton, Ashwini Joshi, Philip Syrrist, Ahmet F. Coskun, and Savas Tasoglu. 3d-printed smartphone-based point of care tool for fluorescence- and magnetophoresis-based cytometry. Lab Chip, 17(16):2839-2851, August 2017. [681 Andres W. Martinez, Scott T. Phillips, Zhihong Nie, Chao-Min Cheng, Emanuel Carrilho, Benjamin J. Wiley, and George M. Whitesides. Programmable diagnostic devices made from paper and tape. Lab Chip, 10(19):2499-2504, Septem- ber 2010. [69] A. Ahmad Tajudin, K. Petersson, A. Lenshof, A.-M. Sward-Nilsson, L. Aberg, G. Marko-Varga, J. Malm, H. Lilja, and T. Laurell. Integrated acoustic immunoaffinity-capture (IAI) platform for detection of PSA from whole blood samples. Lab Chip, 13(9):1790-1796, April 2013. [70] Satish Kalme, Srinivasan Kandaswamy, Anusha Chandrasekharmath, Reeta Katiyar, Gokul Prasath Rajamanickam, Suraj Kumar, and Dhananjaya Den- dukuri. A hydrogel sensor-based microfluidic platform for the quantitative and multiplexed detection of fertility markers for point-of-care immunoassays. Anal. Methods, 11(12):1639-1650, March 2019. [71] Caitlin E. Anderson, Kamal G. Shah, and Paul Yager. Chapter Fourteen - Sensitive Protein Detection and Quantification in Paper-Based Microfluidics for the Point of Care. In Richard B. Thompson and Carol A. Fierke, editors, Methods in Enzymology, volume 589 of Enzymes as Sensors, pages 383-411. Academic Press, January 2017. [72] Irene Bosch, Helena de Puig, Megan Hiley, Marc Carre-Camps, Federico Perdomo-Celis, Carlos F. Narviez, Doris M. Salgado, Dewahar Senthoor, Made- line O'Grady, Elizabeth Phillips, Ann Durbin, Diana Fandos, Hikaru Miyazaki, Chun-Wan Yen, Margarita Gelvez-Ramirez, Rajas V. Warke, Lucas S. Ribeiro, Mauro M. Teixeira, Roque P. Almeida, Jos6 E. Mui6z-Medina, Juan E. Lud- ert, Mauricio L. Nogueira, Tatiana E. Colombo, Ana C. B. Terzian, Patri- cia T. Bozza, Andrea S. Calheiros, Yasmine R. Vieira, Giselle Barbosa-Lima, Alexandre Vizzoni, Jose Cerbino-Neto, Fernando A. Bozza, Thiago M. L. Souza, Monique R. 0. Trugilho, Ana M. B. de Filippis, Patricia C. de Sequeira, Ernesto T. A. Marques, Tereza Magalhaes, Francisco J. Diaz, Berta N. Restrepo, Ka- terine Marin, Salim Mattar, Daniel Olson, Edwin J. Asturias, Mark Lucera, Mohit Singla, Guruprasad R. Medigeshi, Norma de Bosch, Justina Tam, Jose Gomez-Mirquez, Charles Clavet, Luis Villar, Kimberly Hamad-Schifferli, and Lee Gehrke. Rapid antigen tests for dengue virus serotypes and Zika virus in patient serum. Science Translational Medicine, 9(409):eaanl589, September 2017. [731 Can Dincer, Richard Bruch, Andre Kling, Petra S. Dittrich, and Gerald A. Urban. Multiplexed Point-of-Care Testing - xPOCT. Trends in Biotechnology, 35(8):728--742, August 2017.

131 [74] Andres W. Martinez, Scott T. Phillips, Manish J. Butte, and George M. White- sides. Patterned Paper as a Platform for Inexpensive, Low Volume, Portable Bioassays. Angew Chem Int Ed Engl, 46(8):1318-1320, 2007. [75] Benjamin D. Grant, Chelsey A. Smith, Kristine Karvonen, and Rebecca Richards-Kortum. Highly Sensitive Two-Dimensional Paper Network Incor- porating Biotin-Streptavidin for the Detection of Malaria. Anal. Chem., 88(5):2553-2557, March 2016. [76] Michael D. Wilkins, Brendan L. Turner, Kristina R. Rivera, Stefano Menegatti, and Michael Daniele. Quantum dot enabled lateral flow immunoassay for detection of cardiac biomarker NT-proBNP. Sensing and Bio-Sensing Research, 21:46-53, November 2018. [77] Kamal G. Shah, Vidhi Singh, Peter C. Kauffman, Koji Abe, and Paul Yager. Mobile Phone Ratiometric Imaging Enables Highly Sensitive Fluorescence Lat- eral Flow Immunoassays without External Optical Filters. Anal. Chem., 90(11):6967-6974, June 2018. [78] Wooseok Jung, Jungyoup Han, Jin-Woo Choi, and Chong H. Ahn. Point-of-care testing (POCT) diagnostic systems using microfluidic lab-on-a-chip technologies. Microelectronic Engineering, 132:46-57, January 2015. [791 Ana I. Barbosa and Nuno M. Reis. A critical insight into the development pipeline of microfluidic immunoassay devices for the sensitive quantitation of protein biomarkers at the point of care. Analyst, 142(6):858-882, 2017. [80] Francesca Volpetti, Jose Garcia-Cordero, and Sebastian J. Maerkl. A Microflu- idic Platform for High-Throughput Multiplexed Protein Quantitation. PLOS ONE, 10(2):e0117744, February 2015. [81] Won Seok Lee, Vijaya Sunkara, Ja-Ryoung Han, Yang-Seok Park, and Yoon- Kyoung Cho. Electrospun TiO2 nanofiber integrated lab-on-a-disc for ultrasensitive protein detection from whole blood. Lab Chip, 15(2):478-485, December 2014. [82] Vivek Priy Dave, Tien Anh Ngo, Anna-Karin Pernestig, Diana Tilevik, Krishna Kant, Trieu Nguyen, Anders Wolff, and Dang Duong Bang. MicroRNA amplification and detection technologies: opportunities and challenges for point of care diagnostics. Laboratory Investigation, 99(4):452, April 2019. [831 Tanyu Wang, Emilie Viennois, Didier Merlin, and Gangli Wang. Microelectrode miRNA Sensors Enabled by Enzymeless Electrochemical Signal Amplification. Anal. Chem., 87(16):8173-8180, August 2015. [84] Yujuan Zhang, Donghong Zhang, Fei Wang, Danfei Xu, Ye Guo, and Wei Cui. Serum miRNAs panel (miR-16-2*, miR-195, miR-2861, miR-497) as novel non- invasive biomarkers for detection of cervical cancer. Scientific Reports, 5:17942, December 2015.

132 [85] Rimi Hamam, Arwa M. Ali, Khalid A. Alsaleh, Moustapha Kassem, Musaed Alfayez, Abdullah Aldahmash, and Nehad M. Alajez. microRNA expression profiling on individual breast cancer patients identifies novel panel of circulating microRNA for early detection. Scientific Reports, 6:srep25997, May 2016. [86] Yao Huang, Yao Huang, Jie Zhu, Jie Zhu, Wenshuai Li, Wenshuai Li, Ziqiang Zhang, Ziqiang Zhang, Panpan Xiong, Panpan Xiong, Hong Wang, Hong Wang, Jun Zhang, and Jun Zhang. Serum microRNA panel excavated by machine learning as a potential biomarker for the detection of gastric cancer. Oncology Reports, 39(3):1338-1346, March 2018. [87] Laura Maria Zanoli and Giuseppe Spoto. Isothermal Amplification Methods for the Detection of Nucleic Acids in Microfluidic Devices. Biosensors (Basel), 3(1):18-43, December 2012. [88] Shao-Yi Hou, Yi-Ling Hsiao, Ming-Shu Lin, Chun-Che Yen, and Chi-Sheng Chang. MicroRNA detection using lateral flow nucleic acid strips with gold nanoparticles. Talanta, 99:375-379, September 2012. [89] Xuefei Gao, Li-Ping Xu, Tingting Wu, Yongqiang Wen, Xinlei Ma, and Xueji Zhang. An enzyme-amplified lateral flow strip biosensor for visual detection of MicroRNA-224. Talanta, 146:648-654, January 2016.

[90] Federica Degliangeli, Prakash Kshirsagar, Virgilio Brunetti, Pier Paolo Pompa, and Roberto Fiammengo. Absolute and Direct MicroRNA Quantification Us- ing DNA-Gold Nanoparticle Probes. J. Am. Chem. Soc., 136(6):2264-2267, February 2014. [91] Yuanfeng Pang, Chongwen Wang, Jing Wang, Zhiwei Sun, Rui Xiao, and Shengqi Wang. Fe3o49Ag magnetic nanoparticles for microRNA capture and duplex-specific nuclease signal amplification based SERS detection in cancer cells. Biosensors and Bioelectronics, 79:574-580, May 2016. [92] Zhiqiang Gao and Zichao Yang. Detection of MicroRNAs Using Electrocatalytic Nanoparticle Tags. Anal. Chem., 78(5):1470-1477, March 2006. [93] Zhiqiang Gao and Yuan Hong Yu. A microRNA biosensor based on direct chemical ligation and electrochemically amplified detection. Sensors and Actuators B: Chemical, 121(2):552-559, February 2007. [94] Yanfen Peng and Zhiqiang Gao. Amplified Detection of MicroRNA Based on Ruthenium Oxide Nanoparticle-Initiated Deposition of an Insulating Film. Anal. Chem., 83(3):820-827, February 2011. [95] Haifeng Dong, Shi Jin, Huangxian Ju, Kaihong Hao, Li-Ping Xu, Huiting Lu, and Xueji Zhang. Trace and Label-Free MicroRNA Detection Using Oligonu- cleotide Encapsulated Silver Nanoclusters as Probes. Anal. Chem., 84(20):8670- 8674, October 2012.

133 [96] Guo-Jun Zhang, Jay Huiyi Chua, Ru-Ern Chee, Ajay Agarwal, and She Mein Wong. Label-free direct detection of MiRNAs with silicon nanowire biosensors. Biosensors and Bioelectronics, 24(8):2504-2508, April 2009.

[97] Yi Fan, Xiantong Chen, Alastair D. Trigg, Chih-hang Tung, Jinming Kong, and Zhiqiang Gao. Detection of MicroRNAs Using Target-Guided Formation of Conducting Polymer Nanowires in Nanogaps. J. Am. Chem. Soc., 129(17):5437- 5443, May 2007. [98] Christopher P6hlmann and Mathias Sprinzl. Electrochemical Detection of Mi- croRNAs via Gap Hybridization Assay. Anal. Chem., 82(11):4434-4440, June 2010.

[99] Zhimin Cai, Yanling Song, Yanfang Wu, Zhi Zhu, Chaoyong James Yang, and Xi Chen. An electrochemical sensor based on label-free functional allosteric molecular beacons for detection target DNA/miRNA. Biosensors and Bioelec- tronics, 41:783-788, March 2013. [100] Francesca Bettazzi, Ezat Hamid-Asl, Carla Lucia Esposito, Cristina Quin- tavalle, Nello Formisano, Serena Laschi, Silvia Catuogno, Margherita Iaboni, Giovanna Marrazza, Marco Mascini, Laura Cerchia, Vittorio De Franciscis, Gerolama Condorelli, and Ilaria Palchetti. Electrochemical detection of miRNA- 222 by use of a magnetic bead-based bioassay. Analytical And Bioanalytical Chemistry, 405(2-3):1025-1034, January 2013. [1011 Huanshun Yin, Yunlei Zhou, Haixia Zhang, Xiaomeng Meng, and Shiyun Ai. Electrochemical determination of microRNA-21 based on graphene, LNA integrated molecular beacon, AuNPs and biotin multifunctional bio bar codes and enzymatic assay system. Biosensors and Bioelectronics, 33(1):247-253, March 2012. [102] Tugba Kilic, Seda Nur Topkaya, Dilsat Ozkan Ariksoysal, Mehmet Ozsoz, Petek Ballar, Yasemin Erac, and Oguz Gozen. Electrochemical based detection of mi- croRNA, mir2l in breast cancer cells. Biosensors and Bioelectronics, 38(1):195- 201, October 2012. [103] Zhiqiang Gao and Yuan Hong Yu. Direct labeling microRNA with an electrocatalytic moiety and its application in ultrasensitive microRNA assays. Biosensors and Bioelectronics, 22(6):933-940, January 2007. [104] E. A. Lusi, M. Passamano, P. Guarascio, A. Scarpa, and L. Schiavo. Innovative Electrochemical Approach for an Early Detection of microRNAs. Anal. Chem., 81(7):2819-2822, April 2009. [105] Bindu Salim, M. V. Athira, A. Kandaswamy, Madhulika Vijayakumar, T. Sar- avanan, and Thiagarajan Sairam. Microfluidic device for novel breast cancer screening by blood test using miRNA beacon probe. Biomed Microdevices, 19(4):89, September 2017.

134 [106] Stephen C. Chapin and Patrick S. Doyle. Ultrasensitive Multiplexed MicroRNA Quantification on Encoded Gel Microparticles Using Rolling Circle Amplifica- tion. Anal Chem, 83(18):7179-7185, September 2011.

[1071 Hongmei Cao, Xin Zhou, and Yong Zeng. Microfluidic exponential rolling circle amplification for sensitive microRNA detection directly from biological samples. Sensors and Actuators B: Chemical, 279:447-457, January 2019.

[108] Hongxia Jia, Zhengping Li, Chenghui Liu, and Yongqiang Cheng. Ultrasensitive Detection of microRNAs by Exponential Isothermal Amplification. Angewandte Chemie International Edition, 49(32):5498-5501, 2010.

[109] Gaelle C. Le Goff, Rathi L. Srinivas, W. Adam Hill, and Patrick S. Doyle. Hydrogel microparticles for biosensing. European Polymer Journal, 72:386-412, November 2015.

[110] Hyewon Lee, Sarah J. Shapiro, Stephen C. Chapin, and Patrick S. Doyle. En- coded Hydrogel Microparticles for Sensitive and Multiplex microRNA Detection Directly from Raw Cell Lysates. Anal. Chem., 88(6):3075-3081, March 2016.

[111] Augusto M. Tentori, Maxwell B. Nagarajan, Jae Jung Kim, Wen Cai Zhang, Frank J. Slack, and Patrick S. Doyle. Quantitative and multiplex microRNA assays from unprocessed cells in isolated nanoliter well arrays. Lab Chip, 18:2410- 2424, July 2018.

[112] Maxwell B. Nagarajan, Augusto M. Tentori, Wen Cai Zhang, Frank J. Slack, and Patrick S. Doyle. Nonfouling, Encoded Hydrogel Microparticles for Mul- tiplex MicroRNA Profiling Directly from Formalin-Fixed, Paraffin-Embedded Tissue. Anal. Chem., 90(17):10279-10285, September 2018.

[113] D. A. Zubtsov, E. N. Savvateeva, A. Yu. Rubina, S. V. Pan'kov, E. V. Kono- valova, 0. V. Moiseeva, V. R. Chechetkin, and A. S. Zasedatelev. Comparison of surface and hydrogel-based protein microchips. Anal. Biochem., 368(2):205- 213, September 2007.

[114] Rathi L. Srinivas, Stephen C. Chapin, and Patrick S. Doyle. Aptamer- Functionalized Microgel Particles for Protein Detection. Anal. Chem., 83(23):9138-9145, December 2011.

[115] Sukwon Jung and Hyunmin Yi. Fabrication of Chitosan-Poly(ethylene glycol) Hybrid Hydrogel Microparticles via Replica Molding and Its Application toward Facile Conjugation of Biomolecules. Langmuir, 28(49):17061-17070, December 2012.

[116] Sukwon Jung, John H. Abel, Jesse L. Starger, and Hyunmin Yi. Porosity- Tuned Chitosan-Polyacrylamide Hydrogel Microspheres for Improved Protein Conjugation. Biomacromolecules, 17(7):2427-2436, July 2016.

135 [117] Dmitri Proudnikov, Edward Timofeev, and Andrei Mirzabekov. Immobi- lization of DNA in Polyacrylamide Gel for the Manufacture of DNA and DNA-Oligonucleotide Microchips. Analytical Biochemistry, 259(1):34-41, May 1998. [118] Augusto M. Tentori, Kevin A. Yamauchi, and Amy E. Herr. Detection of Isoforms Differing by a Single Charge Unit in Individual Cells. Angewandte Chemie InternationalEdition, 55(40):12431-12435, 2016. [119] Lynna Chen, Harry Z. An, Ramin Haghgooie, Aaron T. Shank, Joseph M. Mar- tel, Mehmet Toner, and Patrick S. Doyle. Flexible Octopus-Shaped Hydrogel Particles for Specific Cell Capture. Small, 12(15):2001-2008, April 2016. [120] Sami Park, Yeol Lee, Dae Nyun Kim, Sangphil Park, Eunji Jang, and Won-Gun Koh. Entrapment of enzyme-linked magnetic nanoparticles within poly(ethylene glycol) hydrogel microparticles prepared by photopatterning. Reactive and Functional Polymers, 69(5):293-299, May 2009. [121] Nak Won Choi, Jungwook Kim, Stephen C. Chapin, Thao Duong, Elaine Dono- hue, Pramod Pandey, Wendy Broom, W. Adam Hill, and Patrick S. Doyle. Multiplexed Detection of mRNA Using Porosity-Tuned Hydrogel Microparti- cles. Anal. Chem., 84(21):9370-9378, 2012. [122] Rathi Lakshmi Srinivas. Hydrogel-based microfluidic assays for multiplexed medical diagnostics. Thesis, Massachusetts Institute of Technology, 2015. [123] Dhananjay Dendukuri, Shelley S. Gu, Daniel C. Pregibon, T. Alan Hatton, and Patrick S. Doyle. Stop-flow lithography in a microfluidic device. Lab Chip, 7(7):818-828, June 2007. [124] Sukwon Jung and Hyunmin Yi. Facile Micromolding-based Fabrication of Biopolymeric-Synthetic Hydrogel Microspheres with Controlled Structures for Improved Protein Conjugation. Chem. Mater., May 2015. [125] Yong He, FeiFei Yang, HaiMing Zhao, Qing Gao, Bing Xia, and JianZhong Fu. Research on the printability of hydrogels in 3d bioprinting. Sci. Rep., 6:29977, July 2016. [126] Jae Jung Kim, Lynna Chen, and Patrick S. Doyle. Microparticle parking and isolation for highly sensitive microRNA detection. Lab Chip, 17(18):3120-3128, September 2017. [127] Rathi L. Srinivas, Stephen D. Johnson, and Patrick S. Doyle. Oil-Isolated Hydrogel Microstructures for Sensitive Bioassays On-Chip. Anal. Chem., 85(24):12099-12107, December 2013. [128] Hyewon Lee, Rathi L. Srinivas, Ankur Gupta, and Patrick S. Doyle. Sensitive and Multiplexed On-chip microRNA Profiling in Oil-Isolated Hydrogel Cham- bers. Angew. Chem., 127(8):2507-2511, February 2015.

136 [129] David C. Appleyard, Stephen C. Chapin, Rathi L. Srinivas, and Patrick S. Doyle. Bar-coded hydrogel microparticles for protein detection: synthesis, assay and scanning. Nat. Protocols, 6(11):1761-1774, November 2011. [130] Xinglu Jiang, Chenggui Zhao, Xiaobo Fan, Wei Xu, Rui Zhang, Hongbo Xu, and Guoqiu Wu. A DNA-modified hydrogel for simultaneous purification, concentration and detection of targeted cfDNA in human serum. RSC Advances, 9(6):3407-3415, 2019. [131] Olav Aarstad, Ellinor Bovre Heggset, Ina Sander Pedersen, Sindre Hove Bjornoy, Kristin Syverud, and Berit Lokensgard Strand. Mechanical Prop- erties of Composite Hydrogels of Alginate and Cellulose Nanofibrils. Polymers, 9(8):378, August 2017. [132] Jamadi Mahsa, Shokrollahi Parvin, Houshmand Behzad, Joupari Mortaza Daliri, Mashhadiabbas Fatemeh, Khademhosseini Ali, and Annabi Nasim. Poly (Ethylene Glycol)-Based Hydrogels as Self-Inflating Tissue Expanders with Tunable Mechanical and Swelling Properties. Macromol. Biosci., 17(8):1600479, August 2017. [133] J. Berger, M. Reist, J. M. Mayer, 0. Felt, N. A. Peppas, and R. Gurny. Structure and interactions in covalently and ionically crosslinked chitosan hydrogels for biomedical applications. Eur. J. Pharm. Biopharm., 57(1):19-34, January 2004. [134] Stephanie A. Fisher, Alexander E.G. Baker, and Molly S. Shoichet. Designing Peptide and Protein Modified Hydrogels: Selecting the Optimal Conjugation Strategy. J. Am. Chem. Soc., 139(22):7416-7427, June 2017. [135] Daniel C. Pregibon, Mehmet Toner, and Patrick S. Doyle. Multifunc- tional Encoded Particles for High-Throughput Biomolecule Analysis. Science, 315(5817):1393-1396, March 2007. [136] Eric Y. Liu, Sukwon Jung, and Hyunmin Yi. Improved Protein Conjugation with Uniform, Macroporous Poly(acrylamide-co-acrylic acid) Hydrogel Micro- spheres via EDC/NHS Chemistry. Langmuir, pages 11043-11054, October 2016. [137] Jae Jung Kim, Ki Wan Bong, Eduardo Reategui, Daniel Irimia, and Patrick S. Doyle. Porous microwells for geometry-selective, large-scale microparticle arrays. Nat Mater, 16(1):139-146, January 2017. [138] Judy Yeh, Yibo Ling, Jeffrey M. Karp, Jay Gantz, Akash Chandawarkar, George Eng, James Blumling III, Robert Langer, and Ali Khademhosseini. Micro- molding of shape-controlled, harvestable cell-laden hydrogels. Biomaterials, 27(31):5391-5398, November 2006. [139] Xiaofeng Wei, Tian Tian, Shasha Jia, Zhi Zhu, Yanli Ma, Jianjun Sun, Zhenyu Lin, and Chaoyong James Yang. Microfluidic Distance Readout Sweet Hydrogel Integrated Paper-Based Analytical Device (pDiSH-PAD) for Visual Quantita- tive Point-of-Care Testing. Anal. Chem., 88(4):2345-2352, February 2016.

137 [140] Brian J Taylor, Anita Howell, Kimberly A Martin, Dammika P Manage, Walter Gordy, Stephanie D Campbell, Samantha Lam, Albert Jin, Spencer D Polley, Roshini A Samuel, Alexey Atrazhev, Alex J Stickel, Josephine Birungi, An- thony K Mbonye, Linda M Pilarski, Jason P Acker, and Stephanie K Yanow. A lab-on-chip for malaria diagnosis and surveillance. Malar. J., 13:179, May 2014.

[1411 Ali K. Yetisen, Haider Butt, Lisa R. Volpatti, Ida Pavlichenko, Matjaz Hu- mar, Sheldon J. J. Kwok, Heebeom Koo, Ki Su Kim, Izabela Naydenova, Ali Khademhosseini, Sei Kwang Hahn, and Seok Hyun Yun. Photonic hydrogel sensors. Biotechnol. Adv., 34(3):250-271, May 2016.

[142] Gerald J. Kost, Nam K. Tran, and Richard F. Louie. Point-of-Care Testing: Principles, Practice, and Critical-Emergency-Disaster Medicine. In Encyclope- dia of Analytical Chemistry. American Cancer Society, 2008.

[143] Anders Larsson, Roman Greig-Pylypczuk, and Albert Huisman. The state of point-of-care testing: a european perspective. Upsala J. Med. Sci., 120(1):1-10, March 2015.

[144] W. W. Tsai, D. B. Nash, B. Seamonds, and G. J. Weir. Point-of-care versus central laboratory testing: an economic analysis in an academic medical center. Clin. Ther., 16(5):898-910, October 1994.

[145] Abou Tayoun, Ahmad N, Paul R. Burchard, Imran Malik, Axel Scherer, and Gregory J. Tsongalis. Democratizing Molecular Diagnostics for the Developing World. Am. J. Clin. Pathol., 141(1):17-24, January 2014.

[146] R. W. Peeling and D. Mabey. Point-of-care tests for diagnosing infections in the developing world. Clin. Microbiol. Infect., 16(8):1062-1069, August 2010.

[147] Augustine J. Sohn, John M. Hickner, and Fasika Alem. Use of Point-of-Care Tests (POCTs) by US Primary Care Physicians. J. Am. Board Fam. Med., 29(3):371-376, May 2016.

[148] Glen L. Hortin, Steven A. Carr, and N. Leigh Anderson. Introduction: Advances in Protein Analysis for the Clinical Laboratory. Clin. Chem., 56(2):149-151, February 2010.

[149] N. Leigh Anderson, Adam S. Ptolemy, and Nader Rifai. The Riddle of Protein Diagnostics: Future Bleak or Bright? Clin. Chem., 59(1):194-197, January 2013.

[150] The Lancet. Thyroid disease-more research needed. The Lancet, 379(9821):1076, March 2012.

138 [151] Mariusz W. Szkudlinski, Valerie Fremont, Catherine Ronin, and Bruce D. Wein- traub. Thyroid-Stimulating Hormone and Thyroid-Stimulating Hormone Re- ceptor Structure-Function Relationships. Physiol. Rev., 82(2):473-502, April 2002.

[1521 Douglas S. Ross. Serum Thyroid-Stimulating Hormone Measurement for Asess- ment of Thyroid Function and Disease. Endocrinol. Metab. Clin. North Am., 30(2):245-264, June 2001.

[153] Carole A. Spencer. Assay of Thyroid Hormones and Related Substances. In Leslie J. De Groot, George Chrousos, Kathleen Dungan, Kenneth R. Feingold, Ashley Grossman, Jerome M. Hershman, Christian Koch, Mdrta Korbonits, Robert McLachlan, Maria New, Jonathan Purnell, Robert Rebar, Frederick Singer, and Aaron Vinik, editors, Endotext. MDText.com, Inc., South Dart- mouth (MA), 2000.

[1541 Ambika Gopalakrishnan Unnikrishnan and Usha V. Menon. Thyroid disorders in India: An epidemiological perspective. Indian J. Endocrinol. Metab., 15(Suppl2):S78-S81, July 2011. [155] Sanjeet Bagcchi. Hypothyroidism in India: more to be done. Lancet Diabetes Endocrinol., 2(10):778, October 2014.

[1561 Ambika Gopalakrishnan Unnikrishnan, Sanjay Kalra, Rakesh Kumar Sahay, Ganapathi Bantwal, Mathew John, and Neeraj Tewari. Prevalence of hypothyroidism in adults: An epidemiological study in eight cities of India. Indian J. Endocrinol. Metab., 17(4):647-652, 2013.

[157] Gaelle C. Le Goff, Jiseok Lee, Ankur Gupta, William Adam Hill, and Patrick S. Doyle. High-Throughput Contact Flow Lithography. Adv. Sci., 2(10):1500149, October 2015.

[158] Jason E. Meiring, Matthew J. Schmid, Scott M. Grayson, Benjamin M. Rathsack, David M. Johnson, Romy Kirby, Ramakrishnan Kannappan, Kalpana Manthiram, Benjamin Hsia, Zachary L. Hogan, Andrew D. Ellington, Michael V. Pishko, and C. Grant Willson. Hydrogel Biosensor Array Platform Indexed by Shape. Chem. Mater., 16(26):5574-5580, December 2004.

[159] Eunae Kang, Sukwon Jung, John H. Abel, Allison Pine, and Hyunmin Yi. Shape-Encoded Chitosan-Polyacrylamide Hybrid Hydrogel Microparticles with Controlled Macroporous Structures via Replica Molding for Programmable Biomacromolecular Conjugation. Langmuir, 32(21):5394-5402, May 2016.

[160] Veronika Kozlovskaya, Jenolyn F. Alexander, Yun Wang, Thomas Kuncewicz, Xuewu Liu, Biana Godin, and Eugenia Kharlampieva. Internalization of red blood cell-mimicking hydrogel capsules with pH-triggered shape responses. A CS Nano, 8(6):5725-5737, June 2014.

139 [161] Ali Khademhosseini and Robert Langer. Microengineered hydrogels for tissue engineering. Biomaterials, 28(34):5087-5092, December 2007.

[162] Daniel C. Pregibon and Patrick S. Doyle. Optimization of Encoded Hydrogel Particles for Nucleic Acid Quantification. Anal. Chem., 81(12):4873-4881, June 2009.

[163] Jesse V. Jokerst, Jie Chou, James P. Camp, Jorge Wong, Alexis Lennart, Amanda A. Pollard, Pierre N. Floriano, Nicolaos Christodoulides, Glennon W. Simmons, Yanjie Zhou, Mehnaaz F. Ali, and John T. McDevitt. Location of Biomarkers and Reagents within Agarose Beads of a Programmable Bio-nano- chip. Small, 7(5):613-624, 2011.

[164] Bernadette Biondi, Luigi Bartalena, David S. Cooper, Laszlo Hegedils, Peter Laurberg, and George J. Kahaly. The 2015 European Thyroid Association Guidelines on Diagnosis and Treatment of Endogenous Subclinical Hyperthy- roidism. Eur. Thyroid J., 4(3):149-163, September 2015.

[165] Carole A. Spencer. Thyroid Profiling for the 1990s - Free T4 Estimate or Sensitive TSH Measurement. J. Clin. Immunoassay, 12(2):82-89, 1989.

[166] Caroline A. Schneider, Wayne S. Rasband, and Kevin W. Eliceiri. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods, 9(7):671-675, July 2012.

[167] J. Milton Harris. Introduction to Biotechnical and Biomedical Applications of Poly(Ethylene Glycol). In Poly(Ethylene Glycol) Chemistry, Topics in Applied Chemistry, pages 1-14. Springer, Boston, MA, 1992.

[168] William M. Deen. Analysis of transport phenomena. Topics in chemical engineering. New York : Oxford University Press, c2012., 2012.

[169] Huading Zhang, P. Stephen Williams, Maciej Zborowski, and Jeffrey J. Chalmers. Binding affinities/avidities of antibody-antigen interactions: Quan- tification and scale-up implications. Biotechnol. Bioeng., 95(5):812-829, De- cember 2006.

[170] Sugunavathi Sepramaniam, Jun-Rong Tan, Kay-Sin Tan, Deidre Ann DeSilva, Subramaniam Tavintharan, Fung-Peng Woon, Chee-Woon Wang, Fung-Lin Yong, Dwi-Setyowati Karolina, Prameet Kaur, Fu-Jia Liu, Kai-Ying Lim, Arun- mozhiarasi Armugam, and Kandiah Jeyaseelan. Circulating MicroRNAs as Biomarkers of Acute Stroke. Int J Mol Sci, 15(1):1418-1432, January 2014.

[171] Edyta Koscianska, Julia Starega-Roslan, Lukasz J Sznajder, Marta Olejniczak, Paulina Galka-Marciniak, and Wlodzimierz J Krzyzosiak. Northern blotting analysis of microRNAs, their precursors and RNA interference triggers. BMC Mol Biol, 12:14, April 2011.

140 [172] Evan M. Kroh, Rachael K. Parkin, Patrick S. Mitchell, and Muneesh Tewari. Analysis of circulating microRNA biomarkers in plasma and serum using quantitative reverse transcription-PCR (qRT-PCR). Methods, 50(4):298-301, April 2010.

[173] Carolin Coenen, Stefanie Liedtke, and Gesine Kogler. RNA Amplification Pro- tocol Leads to Biased Polymerase Chain Reaction Results Especially for Low- Copy Transcripts of Human Bone Marrow-Derived Stromal Cells. PLOS ONE, 10(10):e0141070, October 2015.

[174] Daniel Aird, Michael G Ross, Wei-Sheng Chen, Maxwell Danielsson, Timothy Fennell, Carsten Russ, David B Jaffe, Chad Nusbaum, and Andreas Gnirke. An- alyzing and minimizing PCR amplification bias in Illumina sequencing libraries. Genome Biol, 12(2):R18, 2011.

[175] Samiksha Nayak, Nicole R. Blumenfeld, Tassaneewan Laksanasopin, and Samuel K. Sia. Point-of-Care Diagnostics: Recent Developments in a Con- nected Age. Anal Chem, 89(1):102-123, January 2017.

[176] Per Hydbring and Gayane Badalian-Very. Clinical applications of microRNAs. F1000Res, 2:136, October 2013.

[177] Na Ying, Chuanjing Ju, Xiuwei Sun, Letian Li, Hongbiao Chang, Guangping Song, Zhongyi Li, Jiayu Wan, and Enyong Dai. Lateral flow nucleic acid biosensor for sensitive detection of microRNAs based on the dual amplification strategy of duplex-specific nuclease and hybridization chain reaction. PLoS One, 12(9):e0185091, September 2017.

[178] Wanli Zheng, Li Yao, Jun Teng, Chao Yan, Panzhu Qin, Guodong Liu, and Wei Chen. Lateral flow test for visual detection of multiple MicroRNAs. Sensors and Actuators B: Chemical, 264:320-326, July 2018.

[179] Umit Hakan Yildiz, Palaniappan Alagappan, and Bo Liedberg. Naked Eye De- tection of Lung Cancer Associated miRNA by Paper Based Biosensing Platform. Anal. Chem., 85(2):820-824, January 2013.

[180] Philip Gillespie, Sylvain Ladame, and Danny O'Hare. Molecular methods in electrochemical microRNA detection. Analyst, 144(1):114-129, December 2018.

[181] Somenath Roy, Jun Hui Soh, and Zhiqiang Gao. A microfluidic-assisted microarray for ultrasensitive detection of miRNA under an optical microscope. Lab Chip, 11(11):1886-1894, June 2011.

[182] Hideyuki Arata, Hiroshi Komatsu, Kazuo Hosokawa, and Mizuo Maeda. Rapid and Sensitive MicroRNA Detection with Laminar Flow-Assisted Dendritic Am- plification on Power-Free Microfluidic Chip. PLOS ONE, 7(11):e48329, Novem- ber 2012.

141 [183] Yin-Hsun Feng and Chao-Jung Tsao. Emerging role of microRNA-21 in cancer. Biomed Rep, 5(4):395-402, October 2016. [184] Jianhong Wu, Guangxin Li, Zeyou Wang, Yongliang Yao, Rui Chen, XiongYong Pu, and Jianjun Wang. Circulating MicroRNA-21 Is a Potential Diagnostic Biomarker in Gastric Cancer. Dis Markers, 2015:435656, 2015. [185] Mary Dillhoff, James Liu, Wendy Frankel, Carlo Croce, and Mark Bloomston. MicroRNA-21 is Overexpressed in Pancreatic Cancer and a Potential Predictor of Survival. J Gastrointest Surg, 12(12):2171-2176, July 2008. [186] Shichao Li, Xiaorong Yang, Jinmei Yang, Jiesheng Zhen, and Dechun Zhang. Serum microRNA-21 as a potential diagnostic biomarker for breast cancer: a systematic review and meta-analysis. Clin Exp Med, 16(1):29-35, February 2016. [187] Xuan Pan, Rui Wang, and Zhao-Xia Wang. The Potential Role of miR-451 in Cancer Diagnosis, Prognosis, and Therapy. Mol Cancer Ther, 12(7):1153-1162, July 2013. [188] Sarah J. Shapiro, Dhananjaya Dendukuri, and Patrick S. Doyle. Design of Hydrogel Particle Morphology for Rapid Bioassays. Anal. Chem., 90(22):13572- 13579, November 2018. [189] Junko Kohro-Kawata, Mark H. Wener, and Mart Mannik. The effect of high salt concentration on detection of serum immune complexes and autoantibodies to Clq in patients with systemic lupus erythematosus. The Journal of Rheuma- tology, 29(1):84-89, January 2002.

[190] Alexis F. Sauer-Budge, Samuel J. Brookfield, Ronald Janzen, Sarah McGray, Anna Boardman, Holger Wirz, and Nira R. Pollock. A novel device for col- lecting and dispensing fingerstick blood for point of care testing. PLOS ONE, 12(8):e0183625, August 2017. [191] Yoon Ho Roh, Sang Jun Sim, Il-Joo Cho, Nakwon Choi, and Ki Wan Bong. Vertically encoded tetragonal hydrogel microparticles for multiplexed detection of miRNAs associated with Alzheimer's disease. Analyst, May 2016. [192] H. Lee, Y. H. Roh, H. U. Kim, and K. W. Bong. Low temperature flow lithography. Biomicrofluidics, 12(5):054105, September 2018. [193] Guo-Kun Wang, Jia-Qi Zhu, Jun-Tao Zhang, Qing Li, Yue Li, Jia He, Yong- Wen Qin, and Qing Jing. Circulating microRNA: a novel potential biomarker for early diagnosis of acute myocardial infarction in humans. European Heart Journal, 31(6):659-666, March 2010. [194] Xiaojie Yu, Margarete Odenthal, and Jochen W. U. Fries. Exosomes as miRNA Carriers: Formation-Function-Future. Int J Mol Sci, 17(12):2028, December 2016.

142 [195] John J. Turner, Simon W. Jones, Sterghios A. Moschos, Mark A. Lindsay, and Michael J. Gait. MALDI-TOF mass spectral analysis of siRNA degradation in serum confirms an RNAse A-like activity. Mol. BioSyst., 3(1):43-50, December 2006.

[196] Martina Redova, Alexandr Poprach, Jana Nekvindova, Robert Iliev, Lenka Radova, Radek Lakomy, Marek Svoboda, Rostislav Vyzula, and Ondrej Slaby. Circulating miR-378 and miR-451 in serum are potential biomarkers for renal cell carcinoma. J Transl Med, 10:55, March 2012.

[197] Bing Wang and Qingyun Zhang. The expression and clinical significance of circulating microRNA-21 in serum of five solid tumors. J Cancer Res Clin Oncol, 138(10):1659-1666, October 2012.

[198] H. Zhu, 0. Yaglidere, T. Su, D. Tseng, and A. Ozcan. Wide-field fluorescent microscopy on a cell-phone. In 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pages 6801-6804, August 2011.

[199] Tamas Pardy, Indrek Tulp, Clemens Kremer, Toomas Rang, and Ray Stew- art. Integrated self-regulating resistive heating for isothermal nucleic acid amplification tests (NAAT) in Lab-on-a-Chip (LoC) devices. PLOS ONE, 12(12):e0189968, December 2017.

[200] Hamulati Tusong, Naibijiang Maolakuerban, Jin Guan, Mulati Rexiati, Wen- Guang Wang, Baihetiya Azhati, Yiliyaer Nuerrula, and Yu-Jie Wang. Func- tional analysis of serum microRNAs miR-21 and miR-106a in renal cell carcinoma. Cancer Biomarkers, 18(1):79-85, January 2017.

[2011 Enders K. 0. Ng, Rufina Li, Vivian Y. Shin, Hong Chuan Jin, Candy P. H. Leung, Edmond S. K. Ma, Roberta Pang, Daniel Chua, Kent-Man Chu, W. L. Law, Simon Y. K. Law, Ronnie T. P. Poon, and Ava Kwong. Circulating microRNAs as Specific Biomarkers for Breast Cancer Detection. PLoS One, 8(1):e53141, January 2013.

[202] Paridhi Bhandari, Tanya Naralari, and Dhananjaya Dendukuri. 'Fab-Chips': a versatile, fabric-based platform for low-cost, rapid and multiplexed diagnostics. Lab Chip, 11(15):2493-2499, July 2011.

[203] Xueen Fang, Shasha Wei, and Jilie Kong. Paper-based microfluidics with high resolution, cut on a glass fiber membrane for bioassays. Lab Chip, 14(5):911- 915, February 2014.

[204] Elizabeth A. Phillips, Anna K. Young, Nikolas Albarran, Jonah Butler, Kaira Lujan, Kimberly Hamad-Schifferli, and Jose Gomez-Marquez. Ampli: A Construction Set for Paperfluidic Systems. Advanced Healthcare Materials, 7(14):1800104, 2018.

143 [205] Xiaoxi Yang, Omid Forouzan, Theodore P. Brown, and Sergey S. Shevkoplyas. Integrated separation of blood plasma from whole blood for microfluidic paper- based analytical devices. Lab Chip, 12(2):274-280, December 2011.

[2061 Alexander B. Owczarczak, Stephen 0. Shuford, Scott T. Wood, Sandra Deitch, and Delphine Dean. Creating Transient Cell Membrane Pores Using a Standard Inkjet Printer. JoVE (Journal of Visualized Experiments), (61):e3681, March 2012.

[207] M. E. Pepper, C. A. Parzel, T. Burg, T. Boland, K. J. Burg, and R. E. Groff. Design and implementation of a two-dimensional inkjet bioprinter. In 2009 An- nual International Conference of the IEEE Engineering in Medicine and Biology Society, pages 6001-6005, September 2009.

[208] Arthur Yu, Jing Shang, Fang Cheng, Bradford A. Paik, Justin M. Kaplan, Rodrigo B. Andrade, and Daniel M. Ratner. Biofunctional Paper via Covalent Modification of Cellulose. Langmuir, 28(30):11265-11273, July 2012.

[209] Sartorious.

[210] G041 Glass Fiber Conjugate Pad Sheets GFCP203000.

[211] Hyun Jee Lee, Yoon Ho Roh, Hyeon Ung Kim, Sun Min Kim, and Ki Wan Bong. Multiplexed immunoassay using post-synthesis functionalized hydrogel microparticles. Lab Chip, November 2018.

[212] Kuniaki Nagamine, Kohei Okamoto, Shingo Otani, Hirokazu Kaji, Makoto Kan- zaki, and Matsuhiko Nishizawa. Hydrogel-based bioassay sheets for in vitro evaluation of contraction-dependent metabolic regulation in skeletal muscle cells. Biomaterials Science, 2(2):252-256, 2014.

[213] Nipapan Ruecha, Kentaro Yamada, Koji Suzuki, and Daniel Citterio. (Bio)Chemical Sensors Based on Paper. In Thiago Regis Longo Cesar Paixdo and Subrayal Medapati Reddy, editors, Materials for Chemical Sensing, pages 29-74. Springer International Publishing, Cham, 2017.

[214] Jane Ru Choi, Kar Wey Yong, Ruihua Tang, Yan Gong, Ting Wen, Hui Yang, Ang Li, Yook Chin Chia, Belinda Pingguan-Murphy, and Feng Xu. Lateral Flow Assay Based on Paper-Hydrogel Hybrid Material for Sensitive Point-of- Care Detection of Dengue Virus. Advanced Healthcare Materials, 6(1):1600920, January 2017.

[215] S. M. Zakir Hossain, Roger E. Luckham, Anne Marie Smith, Julie M. Lebert, Lauren M. Davies, Robert H. Pelton, Carlos D. M. Filipe, and John D. Bren- nan. Development of a Bioactive Paper Sensor for Detection of Neurotoxins Using Piezoelectric Inkjet Printing of Sol-Gel-Derived Bioinks. Anal. Chem., 81(13):5474-5483, July 2009.

144 [216] S. M. Zakir Hossain, Roger E. Luckham, Meghan J. McFadden, and John D. Brennan. Reagentless Bidirectional Lateral Flow Bioactive Paper Sensors for Detection of Pesticides in Beverage and Food Samples. Anal. Chem., 81(21):9055-9064, November 2009. [2171 Rabia Mateen, M. Monsur Ali, and Todd Hoare. A printable hydrogel microarray for drug screening avoids false positives associated with promiscuous aggregating inhibitors. Nature Communications, 9(1):602, February 2018. [2181 Meital Reches, Katherine A. Mirica, Rohit Dasgupta, Michael D. Dickey, Man- ish J. Butte, and George M. Whitesides. Thread as a Matrix for Biomedical Assays. ACS Appl. Mater. Interfaces, 2(6):1722-1728, June 2010. [2191 M. Monsur Ali, Sergio D. Aguirre, Yaqin Xu, Carlos D. M. Filipe, Robert Pelton, and Yingfu Li. Detection of DNA using bioactive paper strips. Chem. Commun., 0(43):6640-6642, October 2009. [220] Alberto Bavelloni, Giulia Ramazzotti, Alessandro Poli, Manuela Piazzi, En- rico Focaccia, William Blalock, and Irene Faenza. MiRNA-210: A Current Overview. Anticancer Res, 37(12):6511-6521, December 2017. [2211 Yanping Gao, Bing Feng, Siqi Han, Kai Zhang, Jing Chen, Chen Li, Rui Wang, and Longbang Chen. The Roles of MicroRNA-141 in Human Cancers: From Diagnosis to Treatment. CPB, 38(2):427-448, 2016. [222] American Thyroid Association. General Information/Press Room. [223] Rocktotpal Konwarh, Prerak Gupta, and Biman B. Mandal. Silk-microfluidics for advanced biotechnological applications: A progressive review. Biotechnology Advances, 34(5):845-858, September 2016. [224] Jayanth Kumar Palanichamy and Dinesh S. Rao. miRNA dysregulation in cancer: towards a mechanistic understanding. Front Genet, 5, March 2014. [225] Giovanni Lugli, Aaron M. Cohen, David A. Bennett, Raj C. Shah, Christo- pher J. Fields, Alvaro G. Hernandez, and Neil R. Smalheiser. Plasma Exosomal miRNAs in Persons with and without Alzheimer Disease: Altered Expression and Prospects for Biomarkers. PLoS ONE, 10(10):e0139233, October 2015. [226] Amit K. Pandey, Priyanka Agarwal, Kirandeep Kaur, and Malabika Datta. MicroRNAs in Diabetes: Tiny Players in Big Disease. Cellular Physiology and Biochemistry, 23(4-6):221-232, 2009. [227] Charlotte Glinge, Sebastian Clauss, Kim Boddum, Reza Jabbari, Javad Jab- bari, Bjarke Risgaard, Philipp Tomsits, Bianca Hildebrand, Stefan KKdb, Reza Wakili, Thomas Jespersen, and Jacob Tfelt-Hansen. Stability of Circulating Blood-Based MicroRNAs - Pre-Analytic Methodological Considerations. PLoS One, 12(2), February 2017.

145 [228] David Dannhauser, Filippo Causa, Edmondo Battista, Angela M. Cusano, Domenico Rossi, and Paolo A. Netti. In-flow real-time detection of spec- trally encoded microgels for miRNA absolute quantification. Biomicrofluidics, 10(6):064114, November 2016. [229] Layal Chaker, Antonio C. Bianco, Jacqueline Jonklaas, and Robin P. Peeters. Hypothyroidism. The Lancet, 390(10101):1550-1562, September 2017.

[230] Amay J. Bandodkar, William J. Jeang, Roozbeh Ghaffari, and John A. Rogers. Wearable Sensors for Biochemical Sweat Analysis. Annual Review of Analytical Chemistry, 12(1):1-22, 2019.

[231] Dhananjay Dendukuri, Priyadarshi Panda, Ramin Haghgooie, Ju Min Kim, T. Alan Hatton, and Patrick S. Doyle. Modeling of Oxygen-Inhibited Free Radical Photopolymerization in a PDMS Microfluidic Device. Macromolecules, 41(22):8547-8556, November 2008. [232] IDT. Attachment Chemistry / Linkers. [233] Sandeep Singh, Jyoti Priyadarshini Shrivastava, and Ankit Dwivedi. Breast cancer screening existence in India: A nonexisting reality. Indian J Med Pae- diatr Oncol, 36(4):207-209, 2015.

[234] Won-Il Lee, Younghyeon Park, Jaemin Park, Sajal Shrivastava, Young-Min Son, Hak-Jong Choi, Jaelin Lee, Byeungwoo Jeon, Heon Lee, and Nae-Eung Lee. A smartphone fluorescence imaging-based mobile biosensing system integrated with a passive fluidic control cartridge for minimal user intervention and high accuracy. Lab Chip, 19(8):1502-1511, April 2019.

[235] Bindhu Madhavan, Shijing Yue, Uwe Galli, Sanyukta Rana, Wolfgang Gross, Miryam M6ller, Nathalia A. Giese, Holger Kalthoff, Thomas Becker, Markus W Biichler, and Margot Z511er. Combined evaluation of a panel of protein and miRNA serum-exosome biomarkers for pancreatic cancer diagnosis increases sensitivity and specificity. Int. J. Cancer, 136(11):2616-2627, June 2015. [236] Monpichar Srisa-Art, Emily C. Dyson, Andrew J. deMello, and Joshua B. Edel. Monitoring of Real-Time Streptavidin-Biotin Binding Kinetics Using Droplet Microfluidics. Anal. Chem., 80(18):7063-7067, September 2008.

146