Microfluidic-Based Point-Of-Care Testing for Global Health

Microfluidic-based Point-of-Care Testing for Global Health

Tassaneewan Laksanasopin

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy In the Graduate School of Arts and Sciences

Columbia University

2015

ABSTRACT

Microfluidic-based Point-of-Care Testing for Global Health

Tassaneewan Laksanasopin

Point-of-care (POC) tests can improve the management of infectious diseases and clinical outcomes, through prompt diagnosis and appropriate delivery of treatments for preventable and treatable diseases, especially in resource-limited settings where health care infrastructure is weak, and access to quality and timely medical care is challenging. Microfluidics or lab-on-chip technology is appropriate for

POC tests when general design constraints such as integration, portability, low power consumption, automation, and ruggedness are met. Although many POC tests have been designed for use in developed countries, they might not be readily transferable to resource-limited settings. These new technologies need to be accessible, affordable and practical to be implemented at resource-limited settings to save lives in developing countries.

The overall goal of this dissertation is to develop microfluidic diagnostic devices which are practical and reliable for global health. We first focused on immunoassays, an important class of diagnostic tests which utilize antibodies to quantify host immunity or pathogen protein markers. We developed and evaluated a rapid, accurate, multiplexed, and portable microfluidic immunoassay for diagnosis of HIV and syphilis on hundreds of archived specimens (whole blood, plasma, and sera). Our assay exhibited performance equal to lab-based immunoassays in less than 20 minutes. In addition, our technique quantified signals using a handheld instrument, allowing for objective measurements as opposed to current rapid HIV tests which require subjective interpretation of band intensities. We further integrated three important off-chip processes in a diagnostic test - liquid handling, optical signal detection, and data communication – in a low-cost, versatile, handheld instrument capable of performing immunoassays on reagent-loaded ( i.e. “ready-to-run”) cassettes at high analytical performance characteristic of ELISA but

with the speed, portability and ease-of-use of a rapid test. We also evaluated this immunoassay device in

Rwanda on archived samples and achieved analytical performance comparable to that of benchtop

standards.

To simplify the user interface and reduce the cost of the diagnostic device, we integrated our microfluidic immunoassay with a smartphone to replace computers or high-cost processors for diagnostic devices in low-resource settings. Our low-cost ($34), smartphone-supported device for a multiplexed immunoassay detected three antibody markers from HIV, treponemal- and non-treponemal syphilis from fingerstick whole blood simultaneously in 15 minutes. This device was designed to eliminate the number of manual steps, through the use of lyophilized secondary antibodies and anti-coagulant, preloaded reagents

on cassette, and an automatic result readout. A step-by-step user guide was included on the smartphone

to make the device simple enough to be used by an untrained operator. The analytical performance of the

device was evaluated in Rwanda by local health care workers. We also accessed user experiences for

improvement of the device in future.

While immunoassays offer rapid and accurate diagnosis for infectious diseases, various infections

cannot be confirmed using protein markers. Due to increasing clinical demand for detection of DNA and

RNA signatures for diagnosis and monitoring of patients in resource-limited settings, we also explored how

microfluidic and nanoparticle technologies can improve nucleic acid amplification test at the point of care.

Nucleic acid tests are arguably some of the most challenging assays to develop due to additional steps

required for sample pre-treatment ( e.g. cell sorting, isolation, and lysis, as well as nucleic acid extraction),

signal amplification (due to low physiological concentrations, target contamination, and instability) and

product detection. Here we developed a sputum processor to isolate and lyse mycobacteria (M.smegmatis )

from a more complex sample matrix, using magnetic beads-based target isolation to replace the need of a

centrifuge or other complicated sample preparation technique. We also investigated a technique to detect

Mycobacterium tuberculosis using multiplex polymerase chain reaction (PCR) and silver-gold amplification

detection.

Table of Contents

List of Figures ...... III List of Tables ...... IX Acknowledgements ...... X Chapter 1 Overview ...... 1 1.1 Introduction ...... 1 1.1.1 Overview of global health issues ...... 1 1.1.2 Point-of-care tests for resource-limited settings ...... 3 1.1.3 Current approaches to point-of-care tests ...... 4 1.2 Objectives ...... 8 1.3 Thesis outline ...... 10 1.4 My role in each chapter and relevant publications ...... 12 Chapter 2 Accurate, rapid, and portable immunoassays for HIV and syphilis ...... 14 2.1 Introduction ...... 14 2.2 Methods ...... 19 2.2.1 Microfluidic-based Immunoassay ...... 19 2.2.2 Modeling gold-silver signal amplification ...... 24 2.2.3 Custom-built devices ...... 28 2.2.4 Limit of detection on human IgG immunoassay ...... 32 2.2.5 Dual HIV/syphilis assay evaluation on commercial specimen panels ...... 33 2.2.6 Assay evaluations on archived specimens at the POC settings ...... 34 2.3 Results and Discussion ...... 37 2.4 Conclusion ...... 56 Chapter 3 A smartphone accessory for portable immunoassays ...... 60 3.1 Introduction ...... 60 3.1.1 Smartphone-based healthcare technology ...... 60 3.1.2 Smartphone-based POC immunoassay ...... 61 3.1.3 User acceptability and new technology implementation ...... 64 3.2 Methods ...... 65 3.2.1 Smartphone dongle ...... 65 3.2.2 Cassette preparation ...... 67 3.2.3 Field trial ...... 71 3.3 Results and Discussion ...... 76 3.3.1 Dongle design...... 76 3.3.2 Multiplex immunoassay ...... 84 3.3.3 Field testing with target end-users ...... 86 3.3.4 Feedback from participants enrolled in the study...... 90

3.4 Conclusion ...... 91 Chapter 4 Development of a low-cost and point-of-care sputum processor for a downstream nucleic acid amplification test ...... 95 4.1 Introduction ...... 95 4.2 Methods ...... 104 4.2.1 Target isolation and concentration ...... 104 4.2.2 Fast multiplex PCR ...... 106 4.2.3 PCR product detection using gold-silver amplification ...... 107 4.3 Results and Discussion ...... 110 4.3.1 Sputum processor ...... 110 4.3.2 Target amplification and detection ...... 114 4.4 Conclusion and future directions ...... 117 Chapter 5 Conclusion and Future Directions ...... 119 References ……………………………………………………………………………………………………….122

List of Figures

Figure 1.1. Global burden of disease. Death rate by broad cause group in high-income countries (top) versus low- and middle- income countries (bottom) . Adapted from [1]...... 1

Figure 1.2. Crude death rate by broad cause group, 2000 and 2012 by WHO region [2]...... 2

Figure 1.3 . Disability-adjusted life years (DALYs) for infectious and parasitic diseases in 2005. Adapted from [3]...... 2

Figure 1.4. Two wide range condition of laboratories commonly found in mid-level healthcare centers in the developing countries [3]...... 4

Figure 1.5. An example of current rapid tests for HIV diagnosis (Trinity Biotech’s Uni-Gold® HIV rapid test, which is FDA-approved and CLIA-waived). Step-by-step illustration of a lateral flow test, (1) a drop of whole blood, from fingerstick or venipuncture, is applied to the specimen well, (2) four drops of sample diluent buffer are added, and fluids travel up the nitrocellulose membrane by capillary action. Impregnated below the test region are recombinant HIV envelope proteins, conjugated to a color reagent such as colloidal gold or dyed latex particles, which are solubilized by the running sample and buffer and bind to HIV antibodies if present. The mixture continues up the membrane to the test region, where a band of immobilized antigen binds to antibody-labeled antigen complexes if present, producing a visible line (step 3, left ) or no visible line if antibody-antigen complexes are absent (step 3, right ). A control line with immobilized proteins specific to the detection conjugated proteins should produce a visible line if the test strip is functioning correctly. . 5

Figure 1.6. Commercially available dual HIV and syphilis tests using an immunochromatographic technology for the qualitative detection of antibodies to HIV-1/2 and/or syphilis Treponema palladium (TP) simultaneously in human serum, plasma, or whole blood: (A) SD Bioline HIV/Syphilis Duo, and (B) Chembio DPP® HIV-Syphilis Assay...... 6

Figure 1.7. Examples of commercially available or in development microfluidics-based POC tests: (A) i- STAT (Abbott), Abbott’s i-STAT, a handheld blood analyzer that uses microfabricated thin-film electrodes + + - to measure levels of electrolytes (Na , K , Cl ), general chemistries (pH, urea, glucose), blood gases (pCO 2, pO 2) and hematology (hematocrit). Electrochemical detection system includes amperometry, voltammetry, and conductance, depending on the analyte. (B) Epocal (Alere), a portable blood chemistry analyzer using self-contained cards (Flexcards™), patterned electrodes for sensing, wireless data transmission (C) Abaxis, a compact analyzer (Piccolo® xpress) using injection-molded plastic discs with no sample pre- processing to measure blood chemistries ( e.g. metabolites, lipid, electrolytes, gases). (D) Dakari Diagnostics, a handheld counter for CD4+ T-cell using label-free electrochemical sensing of captured cell lysate by impedance spectroscopy instead of conventional flow cytometry technique. (E) Cepheid, disposable cartridges with benchtop analyzer (GeneXpert®) for PCR-based nucleic acid detection ( e.g. respiratory infections (bacterial and viral) and cancer). (F) Biosite (Alere), a porable reader (Triage® meter) with a disposable microcapillary-driven microfluidic test to detect captured cardiac marker analytes (troponin I, creatine kinase-MB, and myoglobin). (G) Diagnostics for All, instrument-free tests based on paper, capillary-driven microfluidics, with a colorimetric readout to measure liver function (liver damage from HIV/AIDS medication). (H) Claros Diagnostics (OPKO Health), a portable analyzer with a disposable injection-molded plastic cassette to detect protein markers for urological maladies and infectious diseases. Taken from [21]...... 7

Figure 2.1. HIV prevalence rates of adults ages 15 – 49 in 2013 [23]...... 14

Figure 2.2. Percentage of antenatal care attendees test for syphilis at first visit (latest update, 2013) [32]...... 15

III

Figure 2.3. Microfluidic cassettes. (A) Picture of a 7-channel microfluidic cassette. Each cassette can accommodate seven samples (one per channel), with molded holes for coupling of reagent-loaded tubes. (B) Picture of a single-run microfluidic cassette, with areas for reagent storage, analyte capture and detection, and waste storage. (C) Transmitted light micrograph of channel meanders. Scale bar is 1 mm...... 19

Figure 2.4. Assay preparation and setup. (A) Schematic diagram of reagent delivery. A plastic tubing is pre-loaded with reagents separated by air bubbles in a defined sequence. At the time of assay, the plastic tubing is inserted onto the plastic chip, and the syringe is pulled back to displace a volume of air to create a vacuum. The train of reagents passes over a series of four detection zones, each characterized by dense meanders coated with capture proteins, before exiting the chip to the disposable syringe. (B) Illustration of biochemical reactions in detection zones at different immunoassay steps. Silver enhancement (i.e. reduction of silver ions on gold nanoparticle-conjugated antibodies bound on the surfaces of meandering zones) yields signals which can be read with low-cost optics (for quantification) or examined by eye. (C) (Top ) Image sequence demonstrating technique of loading small-diameter plastic tubes with reagent plugs drawn from test tubes and separated by air gaps. Marks are drawn on the tubing to indicate length ( i.e. volume) of reagent plugs. Arrows (in white) indicate reagent plugs, with three plugs positioned in series (far right ). ( Bottom ) Manually loading reagents in tubes using a 1 mL syringe, with arrows (in white) indicating direction of air and fluid movement. (D) Overall assay setup. Picture of cassette with a tube filled with sequence of reagent plugs (here, colored dye) and syringe for generating vacuum. No other peripherals are needed to run the mChip...... 23

Figure 2.5. Modeling of silver development. a, Parameters for modeling silver reduction. tOD,mid , a, OD mid , OD min , OD max and tn are determined from best-fit curves from experimental data; fragAb and SAu,I are estimates based on literature. b, Silver enhancement of zone functionalized with 1:2 antibody to goat IgG antibody : BSA physisorption ratio. Data points are mean absorbance values, and dashed line is best-fit curve (four parameter logistic equation). Parameters are indicated accordingly. c, Dependence of tn on gold nanoparticle density captured on the surface, with best-fit curve (exponential decay) as dashed line and best-fit parameters listed in adjacent table. d, Experimental kinetic data of silver enhancement for various antibody to goat IgG antibody : BSA physisorption ratio. Data points indicate mean values, error bars indicate one s.d. Dashed lines are best-fit curves. e, Computational modeling results (solid lines) superimposed with experimental data points. The difference (expressed as normalized objective function) between the model and experiment is 9.2%...... 39

Figure 2.6. Optical density reader for signal detection. (A) A photo of the custom-build reader showing the user interface and detection unit for 4-detection zone. (B) Schematic of the instrument components showing optics, electronics and component structure. (C) The gradients of optical density and coefficient of variance calculated from five optical densities of each analyte zone...... 41

Figure 2.7. A fully integrated device. (A) A picture of mChip instrument (with cassette inside). (B) Inside view of our POC instrument. Our design includes three main modules: liquid handling (highlighted in orange), signal detection (highlighted in red), and data communication (highlighted in green)...... 42

Figure 2.8. Measured drawn current (black) and absorbance in the HIV zone (red) over time, using less than 1 µL of HIV-positive blood sample. The current is predominately at baseline value of 48 mA, with sharp spikes powering diaphragm pump before entry of blood and silver reagents and powering data transmission (here, via satellite network). Small peaks at beginning and end of silver indicate transmittance readings taken by LEDs and photodetectors (not labeled). The device validates positive HIV status of sample, as absorbance at the end of silver enhancement is above cutoff value of 0.10 for whole blood assays...... 45

Figure 2.9. Control of flowrate inside microfluidic chip by adjusting vacuum pressure. Error bar is one standard deviation (n=3)...... 46

freshly prepared cassettes. (B) Thermal stability of HIV immunoassay. Cassettes were prepared and stored for over 3 months at 4 oC, then ran on the instrument. Shown are signal-to-cutoff values for a positive specimen (n=5) and negative specimen (n=3) validated from a commercial supplier (Seracare). Dashed lines represent s/co values from specimens run on freshly prepared cassettes...... 47

Figure 2.11. Limit of detection using human IgG as a standard marker. Sensitivity is in pM range. Dashed line is background signal from a 0 pM human IgG sample. Errors bars show one standard deviation ( n=3)...... 48

Figure 2.12. Results of immunoassays performed at Columbia University on commercial specimen panels. (A) Test results for HIV (left) and syphilis (right) antibodies. Vertical scatter plots of silver absorbance (normalized by cutoff values) for positive and negative serum/plasma specimens (each patient sample is represented by one filled circle for HIV or cross for syphilis). (B) Receiver operating characteristic (ROC) for HIV and syphilis tests...... 50

Figure 2.13. Field results of HIV immunoassay collected in Muhima Hospital in Rwanda using less than 1 µL of unprocessed pre-collected whole blood sample. ( Left) Signal-to-cutoff ratios for positive and negative samples. ( Right) ROC curve...... 51

Figure 2.14. Field results of a HIV and syphilis duplex immunoassay collected in Projet Ubuzima in Rwanda, using 7 m L of archived plasma or sera. ( Left ) Signal-to-cutoff ratios of sera or plasma specimens which are positive and negative for HIV (circles) and syphilis (crosses). Signal-to-cutoff values greater than 10 are shown at 10, and those less than 0.1 are shown at 0.1 (both with arrows). ( Right ) ROC curves for HIV and syphilis...... 52

Figure 2.15. mChip integrates point-of-care diagnostics with data communication. Here, a step-by-step illustration: (1) fingerprick sample collection, (2) metering onto microfluidic cassette and button-push operation of instrument, (3) display of data in less than 20 minutes of optical density in four analysis zones, (4) option to communicate data by satellite (5a), which sends an encoded message to an email address, and/or by GSM/GPRS (5b), which sends an encoded message to a cellular phone. Shown in red inset is the message, decoded and translated from short-burst form to meaningful data values (test ID, patient ID, and OD values of the four detection zones)...... 54

Figure 2.16. On-the-field performance of POC diagnostics instrument in Rwanda. (A) Vertical scatter plots of silver absorbance (normalized by cutoff values) for whole blood samples positive (filled circles) and negative (open circles) for HIV. (B) Contingency table and evaluation of test accuracy (n = 40). (C) Completed (filled bars) and incompleted (checkered bars) transmission of HIV results via satellite and SMS. For each sample tested, we attempted data transmission with both modes (see Methods for further information). (D) HIV results, communicated remotely, are displayed onto TRACnet server, an electronic medical records system used by the Rwandan Ministry of Health...... 56

Figure 3.1. The processor speed comparison between mobile phones and computers developed over time. Taken from [89]...... 60

Figure 3.2. Smartphone based microfluidic and lab-on-chip technology (a) colorimetric analysis for serum cholesterol detection (b) smartphone microscopy (c) genetic testing and (d) electrochemistry analysis. Taken from [95]...... 61

Figure 3.3. Common patterns of serological reactivity in syphilis patients. Taken from [37]...... 63

Figure 3.4. Principle of Frequency Shift Keying (FSK) modulation. Digital information (data) is transmitted through discrete frequency changes of a carrier wave (FSK modulated signal). Taken from [114]...... 66

Figure 3.5. Robot-assisted manufacturing for cassette preparation...... 67

Figure 3.6. Schematic of multiplex immunoassay on the dongle device and the biochemical reactions at each step of reagents flowing through the microfluidic channel. Five zones are individually treated with proteins: Stabilcoat for internal negative reference, HIV antigen for capturing anti-HIV antibodies (Ab), treponemal syphilis (TP syph) antigen for capturing anti-treponemal antibodies, cardiolipin for capturing anti-cardiolipin antibody (non-treponemal biomarker), and rabbit anti-goat antibody for capturing gold (Au)- labeled goat antibodies for internal positive reference. The 4 th zone for non-treponemal syphilis is coated with Poly-L-Lysine prior the cardiolipin attachment. Whole blood samples are loaded in the antibody holders. Pre-loaded washes on reagent cassette are flowed through the antibody holders to resolubilize lyophilized Au-labeled IgG and IgM antibodies, followed by the silver reagents...... 69

Figure 3.7. Reagent cassette. The cassette for pre-stored reagents needed for the assay features two main parts for on-board reagent storage: washes and silver reagents...... 70

Figure 3.8. Step-by-step illustration of dongle testing. (A) User starts the application and enters patient ID number. (B) User mixes 1 µl of whole-blood sample with 9 µl of diluent and inserts the cassette into the dongle. (C) User presses the bulb fully to initiate vacuum, places 2 µl of the mixed sample into the cassette and attached the antibody holder into the cassette. (D) The application displays step-by-step instructions for user to follow, as well as remaining assay time. (E) After completing the assay, user selects “View Results” to display test results. Another passcode could be added at this step for an extra security and privacy. (F) Screen displays results for each disease marker. The app can be set to show absorbance values or “positive” and “negative” diagnoses, in the study we displayed absorbance values. User clicks “Finish” button to prompt back to the first screen for the next test...... 75

Figure 3.9. A smartphone dongle. (A) An image of the dongle with a microfluidic cassette connected to an iPod touch. (B) A two-layer disposable cassette with antibody holder. The reagent cassette (top layer) contains pre-stored reagents (washes (yellow), silver nitrate (blue), and reducing agent (green)) and the test cassette (bottom layer) contains five detection zones. Reagents are numbered in the order they flow through the test cassette. The power-free vacuum chamber connects to the cassette outlet, drawing fluids from the inlet towards the waste pad...... 77

Figure 3.10. Mechanism of the smartphone dongle. (A) Schematic diagram of dongle highlighting a power- free vacuum generator. Sub-figure shows vacuum activation. (B) The audio jack connector on the dongle is used for audio-based powering and FSK data transmission to a smartphone...... 78

Figure 3.11. User-activated negative pressure–driven flow. Average time to flow a total sequence of 6 washes: two 2-μl washes of 0.05% Tween-PBS and four 2-μl washes of DI water, are shown. Two of the three users were not device developers (i.e. unfamiliar with the device). Data are averages, ± 1 SD ( n = 4). n.s., not significant, one-way ANOVA ( P = 0.149)...... 80

Figure 3.12. Power consumption of dongle (black) and OD of the HIV zone (red) during the assay. High OD is observed while whole blood flows through the zones, with low OD during gold-labeled antibodies (Au-Ab) and washes. At minute 5, OD increases as silver develops...... 80

Figure 3.13. Circuit diagram of: (A) Power harvesting from audio jack connector, (B) MSP430f1611 connections, (C) LED circuit, and (D) photodiode with signal amplification. In (C) and (D) only one example circuit is shown, but analogous circuits are connected to LED 2-5 and PHOTO 2-5 respectively...... 82

Figure 3.14. Optical densities readout using a smartphone dongle. (A) An HIV-positive whole-blood sample was run in triplicate, where measurements were taken with the iPod audio jack–powered dongle and the benchtop analyzer (OPKO Diagnostics, without the temperature-control system, in order to mimic the dongle) reading the same cassette. Data are averages ± 1 SD ( n = 3). n.s., not significant, two-tailed Student’s t-test. (B) Non-trepnonemal syphilis quantitative assay by detecting anti-cardiolipin antibodies. Serial dilution of RPR positive (1:128) serum to mimic lower RPR titers (1:1 to 1:64) as conventional RPR quantification. Data are averages ±1 SD ( n=3) and plotted with a linear regression fit and correlation. .... 83

Figure 3.15. An additional of gold-labeled anti-human IgM for syphilis detection. Comparison of signal measurements obtained by additional of gold-labeled anti-hIgM to gold-labeled anti-hIgG and gold-labeled anti-hIgG alone as detection antibodies for negative, weak positive non-treponemal syphilis (RPR titer 1:2), and strong positive non-treponemal syphilis (RPR titer 1:32) plasma samples. Data are averages ±1 SD (n=4 for anti-hIgG and n=3 for anti-hIgG:anti-hIgM)...... 85

Figure 3.16. Comparison of signal from gold-labeled anti-hIgG/anti-hIgM antibodies lyophilized in a plastic antibody holder and freshly prepared in solution. Detection zones were functionalized with human IgG, human IgM and rabbit anti-goat antibodies (positive ctrl). Data are averages ±1 SD ( n=3). n.s. not significant, Student’s t-test. Comparison of signal from gold-labeled anti-hIgG/anti-hIgM antibodies lyophilized in a plastic antibody holder and freshly prepared in solution...... 85

Figure 3.17. Stability of functionalized protein on microfluidic cassette. Cassettes were prepared using StabilCoat Immunoassay Stabilizer (Surmodics). Microfluidic surfaces were stored for 21 days at 60⁰C, after which signals on HIV 1/2, treponemal syphilis, and cardiolipin zones were measured with stored plasma samples. Data are averages ( n=2), with individual data shown as dots (•). n.s., not significant, two- tailed Student’s t-test...... 86

Figure 3.18. Testing of dongle using clinical fingerprick whole-blood specimens in the field by third party running the test. (Left) Vertical scatter plot of the dongle device signal-to-cutoff ratios for HIV, treponemal syphilis, and non-treponemal syphilis positive (Pos) and negative (Neg) samples using fingerprick whole blood compared to gold standard tests (HIV ELISA, TPHA, and RPR). (Right) A receiver operating characteristic (ROC) curve for each disease marker...... 88

Figure 3.19. Testing of dongle using clinical venipuncture whole-blood specimens in the field by our team. (Left) Vertical scatter plot of the dongle device signal-to-cutoff ratios for HIV, treponemal syphilis, and non- treponemal syphilis positive (Pos) and negative (Neg) samples using fingerprick whole blood compared to gold standard tests (HIV ELISA, TPHA, and RPR). (Right) A receiver operating characteristic (ROC) curve for each disease marker ...... 89

Figure 3.20. Satisfaction survey from participants in this study...... 91

Figure 4.1. Estimated HIV prevalence (%) in new and relapse TB cases, 2013 [136]...... 95

Figure 4.2. Estimated TB incidence rates, 2013 [136]...... 96

Figure 4.3. Cepheid’s GeneXpert test platform (left) integrates sample processing and PCR in a disposable plastic cartridge (right) containing reagents for cell lysis, nucleic acid extraction, amplification and amplicon detection. This system has been used to detect drug-resistant tuberculosis cases by amplifying Mycobacterium tuberculosis specific sequence of the rpoB gene and probing with molecular beacons for mutations conferring rifampicin resistance [163] ...... 101

Figure 4.4. Nanosphere Verigene system and capture schematic. (A) Capture and detection of target sequence in the Nanosphere system which uses gold nanoparticles functionalized with oligonucleotides. Following silver development, the amount of light scattered is measured using a scanner-based device. (B) Picture of a Nanosphere cartridge with loaded reagent wells. (C) Picture of Verigene readers with touch- screen control panels...... 102

Figure 4.5. Target amplification using biotinylated primers in a fast multiplex PCR...... 107

Figure 4.6. PCR product detection using gold-silver amplification. ( Left ) Illustration of biochemical reactions in detection zones at different signal detection steps after PCR. ( Right ) Silver enhancement yields signals which can be read with low-cost optics or examined by eye...... 109

VII

Figure 4.7. PCR product detection using gold-silver amplification with enzyme cleavage to remove DNA mismatches...... 110

Figure 4.8. Bacteria isolation from artificial sputum using coated magnetic beads. Schematic of cross section (side view) of the device and an image of capture chamber filled with magnetic beads observed under stereoscope from top view. A permanent magnet is aligned underneath the capture chamber. The magnetic beads can be seen to be captured efficiently, with very little going to the outlet. Scale bar is 2 mm...... 112

Figure 4.9. Performance of a handheld sputum processor. (A) Percentage of beads captured from 3 mL and 0.5 mL of magnetic beads at flow rates of 0.5 and 2.0 mL/min, as measured by optical density (OD750) of beads in solution before and after capturing. (B) Comparison of conventional sputum processing by conventional centrifugation and our USP at various dilution of bacteria ( M.smegmatis as a laboratory model of M.tuberculosis ) spiked in artificial sputum (porcine mucin from Sigma Aldrich, a commonly used proxy for clinical sputum [172]), as measured by intensity of bands on electrophoresis gel of PCR products of bacterial DNA. Data are averages ±1 SD ( n=3). Dashed line indicates band intensity of no template control ...... 113

Figure 4.10. Multiplex PCR. Gel electrophoresis shows each target and multiplex PCR using in-house primer design with a fast PCR mix, NTC is no-template control...... 114

Figure 4.11. Probe hybridization and gold-silver amplification detection of M.tuberculosis. Optical densities of silver reduction on surface from PCR products of no template control (NTC), wild type genomic DNA, and mutated rifampin resistance-determining region (RRDR). Data are averages ±1 SD ( n=2)...... 115

Figure 4.12. Successful detection of nucleic-acid targets using the gold-silver amplification chemistry. (Left) Optical densities of silver signal from target DNA bound to probes specific to MSSA and MRSA genes for differentiation of two S. aureus strains. Target DNA was generated through multiplex PCR reaction which amplifies SA and MecA from whole genomes. SA gene is present in both S. aureus strains while MecA is present only in MRSA strain. (Right) Preliminary limit of detection of S. aureus using gold-silver amplification from PCR products. Data are averages ±1 SD ( n=3)...... 116

Figure 4.13. Probe hybridization with enzyme cleavage prior signal amplification. Optical densities of silver signal from target DNA bound to probes specific to rifampin-sensitive strain (rpoB gene) after enzyme cleavage to remove DNA mismatch for differentiation of two M.tuberculosis strains: rifampin-resistant (RIF- R) and rifampin-sensitive (RIF-S) strains. Data are averages ±1 SD ( n=3)...... 116

VIII

List of Tables

Table 4.1. Comparison of conventional microbiological approaches with microfluidic approach for nucleic acid detection (adapted from [149])...... 98

Table 4.2. Target genes and primers...... 106

Acknowledgements

This thesis dissertation would not have been possible without the tremendous help of many people.

I thank the Royal Thai Scholarship for the financial support and facilitating for my study abroad.

I would like to thank my thesis advisor, Prof. Samuel Sia, for his invaluable support, guidance, and critical feedback throughout my time at Columbia University. His passion and vision for global health has always been inspired me. I am grateful to have gotten an opportunity to work on this project to create a meaningful impact. Sam’s mentorship is not limited to research, he has taught me many other skills including entrepreneurship, teamwork, proposal writing, critical review, presentation, etc. I really appreciate these experiences which have helped prepared for my career.

I thank graduate, undergraduate, and exchanged students in the Sia Lab who worked with me on my projects and/or had good time in and outside the laboratory. Special thanks to: Dr. Curtis Chin, one of my best coworkers, for his guidance, support, and teamwork on immunoassay development, and for numerous trips to Rwanda to evaluate the immunoassay and instrument; Dr. Yuk Kee Cheung-Poh, for her incredible help on many electronics, hardware and smartphone application development as well as general guidance to work in the lab; Tiffany Guo, for her expertise on clinical discussion, great help on the smartphone accessory, and trips to Rwanda; Archana Sridhara, for her endless work on immunoassay;

Samiksha Nayak, for her help on Rwanda field evaluation; Sau Yin Chin, for her scientific advice and discussion as well as shipping all supplies to Rwanda when requested; Robert Houghtaling, for his dedicated help on sputum processor; and Claire Duvallet, for her help on target and primer selection.

I thank many people outside of the Sia Lab for enabling my work and providing critical feedback.

Special thanks to: David Steinmiller, Dr. Vincent Linder, and coworkers at OPKO Diagnostics; Dr. Ruben

Sahabo, Dr. Veronicah Mugisha, Elisaphane Munyazesa and colleagues at ICAP-Rwanda; Prof. Janneke van de Wijgert and colleagues at Projet Ubuzima; Dr. Etienne Karita and colleagues at Projet San

Francisco; Jean-Marie Uwimana and colleagues at Rwanda National Reference Laboratory; Keith Yeager

(previously) at BME Columbia University; Dr.Arnold Castro at CDC; Dr. Barry Kreiswirth at PHRI; Susie

Kim at RASCAL Columbia University; Prof.Kam’s Lab and Prof.Jacobs’ Lab at BME Columbia University. I also thank my thesis committee, Prof. Gordan Vunjak-Novakovic, Prof. Kam Leong, Prof. Jinyue Ju, and

Prof. Anne-Catrin Uhlemann, for their insightful feedback on my dissertation.

Last but not least, I would like to thank my family and close friends for their continual support throughout my graduate study and life in NYC.

Tassaneewan Laksanasopin

April, 2015

Chapter 1 Overview

1.1 Introduction

1.1.1 Overview of global health issues

Non-communicable diseases were responsible for 66% of all deaths globally in 2012, up from

58% in 2000, however, communicable diseases or infectious diseases are still the leading cause of death in low- and middle-income countries [1] (Fig. 1.1). Infectious diseases cause an enormous burden of death and disability in low-income countries, especially, Africa (Fig. 1.2). Diarrhea, malaria, HIV/AIDS, tuberculosis, preterm birth complications, and congenital disorders are in the top ten that account for almost a third of the healthy life years lost to premature death and to disability in sub-Saharan Africa [4] (Fig. 1.3) . Among the most vulnerable population groups are young children and women of reproductive age, as well as those living in remote rural areas and in areas of conflict who have poor access to clinics [5].

Figure 1.1. Global burden of disease. Death rate by broad cause group in high-income countries (top) versus low- and middle- income countries (bottom) . Adapted from [1].

Figure 1.2. Crude death rate by broad cause group, 2000 and 2012 by WHO region [2].

Figure 1.3. Disability-adjusted life years (DALYs) for infectious and parasitic diseases in 2005. Adapted from [3].

In the absence of diagnostic tests in resource-limited settings, disease is often treated based on clinical symptoms and local prevalence of disease. While syndromic management approach may capture most patients requiring treatment, it also unnecessarily treats patients who do not require treatment and results in overtreatment, wasted resources and, potentially, increased antimicrobial resistance [6, 7]. Equally important, this latter group of patients is not being treated for their disease

[8]. In other cases, diagnostic tests are needed where asymptomatic infection is common, clinical features are nonspecific but the consequences of infection are serious, or treatment is potentially toxic and/or difficult to administer [6, 7]. Appropriate diagnostic tests for early detection will guide treatment, reduce the risk of development of long-term complications, and prevent onward transmission.

1.1.2 Point-of-care tests for resource-limited settings

Innovation in science and technology can reduce the health disparity between the developed world and developing world [9, 10]. In particular, new technologies are needed to provide healthcare beyond the centralized laboratory and into health clinics at the community level (defined here as

“point-of-care” settings). Point-of-care (POC) tests can improve the management of infectious diseases and clinical outcomes, especially in resource-limited settings where health care infrastructure is weak, and access to quality and timely medical care is a challenge, by delivering appropriate and prompt diagnosis and treatments to those preventable and treatable diseases [11-

13] (Fig. 1.4). The World Health Organization Sexually Transmitted Diseases Diagnostics Initiative

(SDI) has developed a guideline called ‘‘ASSURED’’ (Affordable, Sensitive, Specific, User-friendly,

Rapid and robust, Equipment-free or minimal equipment and Deliverable to end-users) for diagnostic tests in these settings [9]. General design constraints, for example, integration, portability, low power consumption, automation, and ruggedness are important to POC tests. Ideally, the diagnosis should be made at the point of care, so there would be no delay in treatment initiation, and should not depend on the availability of a laboratory or highly trained staff [7].

Figure 1.4. Two wide range condition of laboratories commonly found in mid-level healthcare centers in the developing countries [3].

1.1.3 Current approaches to point-of-care tests

Current approaches to POC diagnostics in resource-limited settings use established technologies such as immunochromatography ( i.e. lateral-flow) and immunofiltration ( i.e. flow- through). Lateral flow tests, where flow is driven by capillary forces on a membrane or paper strip without user intervention together with a colorimetric readout (gold colloids or dyed latex), is widely used to detect protein markers present in blood, urine, or saliva specimens. Although the test is simple to perform and requires short assay time, producing reproducible, quantitative, and sensitive results could not be done under the single-flow action. Unlike the multi-step procedures of gold standard laboratory-based assays (e.g. enzyme-linked immunosorbent assay or ELISA), which provide high sensitivity through multiple washing steps and signal amplification. This leads to a recognized limitation of most rapid tests of subjective determination of visible bands, which is difficult with weakly positive samples, and has led to unacceptably high level of false positive and negative diagnoses in standard HIV rapid tests in multiple objective evaluations at resource-limited settings

[14-17].

Figure 1.5. An example of current rapid tests for HIV diagnosis (Trinity Biotech’s Uni- Gold® HIV rapid test, which is FDA-approved and CLIA- waived). Step-by-step illustration of a lateral flow test, (1) a drop of whole blood, from fingerstick or venipuncture, is applied to the specimen well, (2) four drops of sample diluent buffer are added, and fluids travel up the nitrocellulose membrane by capillary action. Impregnated below the test region are recombinant HIV envelope proteins, conjugated to a color reagent such as colloidal gold or dyed latex particles, which are solubilized by the running sample and buffer and bind to HIV antibodies if present. The mixture continues up the membrane to the test region, where a band of immobilized antigen binds to antibody-labeled antigen complexes if present, producing a visible line (step 3, left ) or no visible line if antibody-antigen complexes are absent (step 3, right ). A control line with immobilized proteins specific to the detection conjugated proteins should produce a visible line if the test strip is functioning correctly.

Another limitation of rapid tests is the difficulty to detect multiple markers in a single test, a cost- effective strategy for simultaneously testing a family of conditions relevant to a particular setting or target population. Note that there are few companies who rolled out duplex test for HIV and syphilis in the past few years (Fig. 1.6). Finally, rapid tests are not integrated with electronic health records, which are being introduced to increase efficiency of managing health services, such as the rapid expansion of HIV antiretroviral treatment programs. While several commercial readers are available to measure results of a lateral flow strip and provide quantitative readout, it can be difficult to reliably adjust for positioning, illumination accuracy, and dynamic range and hence influence accurate quantitation [18]. Typically, precise liquid handling and metering is challenging when an untrained user executes the test, which can result in test errors [19, 20]. In spite of the achievements of rapid tests in providing diagnostic services closer to underserved patients in resource-limited settings, there is significant room for improvement of rapid POC tests for infectious diseases.

Microfluidics or lab-on-a-chip technology offers advantages including small reagent volumes, fast turnaround time, small size, low cost, and low power consumption while maintaining high performance lab testing that can be engineered to serve this purpose. Examples of some current microfluidic diagnostic devices in the marketplace or in development for near patient testing are shown in Figure 1.7. Collectively, these innovations demonstrate great versatility in detecting a range of analytes: clinical chemistry markers (Fig. 1.5A-C), cells (Fig. 1.7D), nucleic acids (Fig.

1.5E), and proteins (Fig. 1.5F-H).

Figure 1.7. Examples of commercially available or in development microfluidics-based POC tests: (A) i-STAT (Abbott), Abbott’s i-STAT, a handheld blood analyzer that uses microfabricated thin-film electrodes to measure levels of electrolytes (Na +, K +, Cl -), general chemistries (pH, urea, glucose), blood gases (pCO 2, pO 2) and hematology (hematocrit). Electrochemical detection system includes amperometry, voltammetry, and conductance, depending on the analyte. (B) Epocal (Alere), a portable blood chemistry analyzer using self-contained cards (Flexcards™), patterned electrodes for sensing, wireless data transmission (C) Abaxis, a compact analyzer (Piccolo® xpress) using injection-molded plastic discs with no sample pre-processing to measure blood chemistries ( e.g. metabolites, lipid, electrolytes, gases). (D) Dakari Diagnostics, a handheld counter for CD4+ T-cell using label-free electrochemical sensing of captured cell lysate by impedance spectroscopy instead of conventional flow cytometry technique. (E) Cepheid, disposable cartridges with benchtop analyzer (GeneXpert®) for PCR-based nucleic acid detection ( e.g. respiratory infections (bacterial and viral) and cancer). (F) Biosite (Alere), a porable reader (Triage® meter) with a disposable microcapillary-driven microfluidic test to detect captured cardiac marker analytes (troponin I, creatine kinase-MB, and myoglobin). (G) Diagnostics for All, instrument-free tests based on paper, capillary-driven microfluidics, with a colorimetric readout to measure liver function (liver damage from HIV/AIDS medication). (H) Claros Diagnostics (OPKO Health), a portable analyzer with a disposable injection-molded plastic cassette to detect protein markers for urological maladies and infectious diseases. Taken from [21].

Despite these examples, very few have been actually tested in resource-limited settings. Many

POC tests have been designed for use in developed countries, and might not be readily transferable to resource-limited settings [6]. Testing in the field can exhibit markedly different performance from tests run in a laboratory owing to variations in clinical specimens, local environmental conditions

(including temperature and humidity), and variations in how the tests are run by individual users.

With many advances in individual microfluidics components but less substantial progress in real working devices, a tremendous engineering challenge for microfluidic devices for global health diagnostics is the integration of appropriate LOC procedures onto a portable and low-cost device

[3, 13]. These new technologies need to be accessible, affordable and practical to be implemented at resource-limited settings. By increasing access to appropriate diagnosis and treatment, POC tests for infectious diseases could save many lives in developing countries.

1.2 Objectives

The goal of this research is to develop microfluidic diagnostic devices which are practical and reliable for global health applications. We first focus on immunoassays, an important class of diagnostic tests which utilize antibodies to quantify host immunity or pathogen protein markers. We aim to develop and evaluate a rapid, accurate, multiplexed, and portable immunoassay for diagnosis of HIV and sexually transmitted infections (STIs) on the field (aim 1). While rapid diagnostic tests or lateral flow assays have been widely used as POC diagnostic tools, they are limited by a restricted number of markers detected in a single test, not sufficient sensitivity of the test, subjective user interpretation of band intensities, and lack of precise sample and reagent used in each test. We miniaturized the ELISA assay, a multistep immunoassay which serves as the clinical “gold standard” for detecting most protein-based biomarkers, on a plastic cassette which can diagnose HIV and syphilis in less than 20 minutes. We further integrated three important off-chip processes in a

8 diagnostic test - liquid handling , optical signal detection, and data communication – in a low-cost, versatile, handheld instrument capable of performing immunoassays on reagent-loaded ( i.e. “ready- to-run) cassettes at high analytical performance characteristic of ELISA but with the speed, portability and ease-of-use of a rapid test. We evaluated this immunoassay in Rwanda on hundreds of patient samples and achieved analytical performance comparable to that of benchtop standards.

To simplify user interface and reduce the cost of the diagnostic device, we move towards an integration with a smartphone (aim 2). Mobile phone accessibility is increasing and also transforming healthcare delivery to mobile health worldwide. Smartphone serves as “low-cost processor” with friendly user interface and easily connect with the external devices, which is the best option to replace computers in low-resource settings. We developed a low-cost, smartphone supported diagnostic device for a multiplexed immunoassay detecting antibody markers from HIV, treponemal- and non-treponemal syphilis from fingerstick whole blood. This device is designed to eliminate a number of manual steps through the use of lyophilized secondary antibodies and anti-coagulant, preloaded reagents on cassette, and an automatic result readout. A step-by-step user guide was included on the smartphone to make the device simple enough to be used by an untrained operator.

The analytical performance of the device was evaluated in Rwanda by local health care workers.

We also accessed user experiences for improvement of the device in future.

Due to increasing clinical demand for detection of DNA and RNA signatures for diagnosis and monitoring of patients in resource-limited settings, we also aim develop a device for a rapid bacterial isolation from large volume of sputum to be used for downstream nucleic acid tests at the point of care (aim 3). While immunoassay offers rapid and accurate diagnosis for infectious diseases, various infections cannot be confirmed using protein markers. Nucleic acid tests are arguably some of the most challenging assays to develop due to additional steps required for sample pre-treatment

(e.g. cell sorting, isolation, and lysis, as well as nucleic acid extraction), signal amplification (due to low physiological concentrations, target contamination, and instability) and product detection. Here

9 we developed a sputum processor to isolate and lyse mycobacteria ( M.smegmatis ) from a more complex sample matrix, using magnetic beads-based target isolation to replace the need of a centrifuge or other complicated sample preparation technique. We also investigated a technique to detect Mycobacterium tuberculosis using multiplex polymerase chain reaction (PCR) and silver-gold amplification detection to identify drug-resistant tuberculosis.

1.3 Thesis outline

The contents of this dissertation are organized in the following manner:

Chapter 1: The overview of the needs for global health and point-of-care diagnostics is discussed along with the currently available technologies and their limitations.

Chapter 2: Undiagnosed and untreated STIs cause large morbidity and mortality, including birth defects and stillborn babies. Since most STI’s have known treatments, the largest barriers for treating patients include high cost of transporting specimens to central labs and lack of access to diagnostic testing. To respond to the challenge of rapid, easily accessible and easy-to-use diagnostic tests, we developed and tested a portable and low-cost microfluidics device for point-of- care diagnosis of HIV and syphilis combination. In order to reduce the cost and size of the assay while maintaining high performance, we incorporated microfluidic designs such as single-use plastic microfluidic cassettes, a passive method for delivering reagents, and an amplification chemistry using gold nanoparticles that can be detected using low-cost optics to replicate all steps of ELISA for POC diagnosis in resource-limited settings. We demonstrate an integrated strategy for miniaturizing complex laboratory assays using microfluidics and nanoparticles to enable POC diagnostics and early detection of infectious diseases in remote settings. We further integrated microfluidics, optics, and mobile communication to perform all main functions of ELISA in a low-cost

10 handheld device. The results can be reliably synchronized in real time with electronic health records on a remote server, using wireless communication via a cell-phone network or satellites in orbit. The device operates autonomously with minimal user input, produces each result in less than 20 minutes. We evaluated our technologies in Rwanda using archived sera/plasma and whole blood.

Chapter 3: Mobile phone accessibility is increasing and also transforming health care delivery worldwide. The capability of new medical devices and smartphone as well as the consumer needs can go beyond commercially available devices for glucose monitoring, vital signs and other well- being applications. This work demonstrates that a full laboratory-quality immunoassay can be run on a smartphone accessory. This low-cost device replicates all mechanical, optical, and electronic functions of lab-based ELISA without requiring any stored energy; all necessary power is drawn from a smartphone. This device performed a triplexed immunoassay not currently available in a single test format: HIV antibody, treponemal-specific antibody for syphilis, and non-treponemal antibody for active syphilis infection from a fingerprick whole blood in 15 minutes. We asked local health care workers to perform the test on 96 enrolled patients in Rwanda. We interviewed feedback from participants regarding user experience and preference regarding POC diagnostic test. This work suggests coupling microfluidics with recent advances in consumer electronics can make certain lab-based diagnostics accessible to those hard-to-reach population.

Chapter 4: Nucleic acid testing for infectious diseases is usually performed in centralized laboratories where high-end instruments and skilled personnel are located. Nucleic acid-based POC tests for infections which are less complex and cheaper can address the needs for diagnosis and management in some of the endemic infectious diseases, for example, tuberculosis. Tuberculosis diagnosis has been relied mainly on smear microscopy or culture in developing countries. The

Cepheid GeneXpert, a fully automated PCR-based system, the only available platform is not quite accessible at resource-limited settings where the endemic is. Sample processing is the most underdeveloped module for nucleic acid detection. Conventional sputum processing, in particular,

11 involves centrifugation, multiple manual steps for adding and removing buffer and waste, and a biosafety cabinet if the sputum is a possible tuberculosis [22]. Here, we developed a technique to isolate mycobacteria from large volume (mL) of viscous sample using magnetic beads with liquefying and disinfecting specimen in one step to minimize user workload and prevent exposure to infectious pathogens. We also explored tuberculosis identification and drug resistant genes detection using multiplex PCR and gold-silver amplification adopted from the previous development

(Chapter 2-3). Focusing on sputum processing in this work, future development towards a fully integrated device for POC nucleic acid detection was discussed.

Chapter 5: The overall summary of this dissertation highlighting what we accomplished towards our goal for global health and point-of-care diagnostics. We also discussed the future developments and requirements of this work.

1.4 My role in each chapter and relevant publications

Device design and development for global health is a team effort from various background people. Here, I would like to clarify my role in each study. For Chapter 2, my role was to develop both HIV and HIV-syphilis immunoassay tests, a benchtop reader and perform preclinical evaluation in Rwanda. For Chapter 3, my role in this study was to lead the multiplex immunoassay development, guide the smartphone device prototyping, design questionnaires to assess user feedback and experience, and perform evaluation in Rwanda. For Chapter 4, my role was to develop the new platform for nucleic acid detection focusing on tuberculosis diagnosis. Manuscripts related to these sections are listed below:

Chapter 2:

T. Laksanasopin , C.D. Chin, H. Moore, J. Wang, Y.K. Cheung, and S.K. Sia, "Microfluidic point- of-care diagnostics for resource-poor environments", Proc. SPIE 7318, 73180E (2009).

C.D. Chin, T. Laksanasopin , Y.K. Cheung, D. Steinmiller, V. Linder, et. al., “Microfluidics-based diagnostics of infectious diseases in the developing world”, Nature Medicine 17:1015-U138 (2011).

C.D. Chin*, Y.K. Cheung*, T. Laksanasopin , M.M. Modena, S.Y. Chin, et. al. , “Mobile Device for Disease Diagnosis and Data Tracking in Resource-Limited Settings”. Clinical Chemistry 59:629-

40 (2013). (* denotes equal contribution)

Chapter 3:

T. Laksanasopin *, T.W. Guo*, S. Nayak, A.A. Sridhara, S. Xie, O.O. Olowookere, et. al. , “A smartphone dongle for diagnosis of infectious diseases at the point of care”. Science Translational

Medicine 7:273re1 (2015). (* denotes equal contribution)

Chapter 4:

T. Laksanasopin, R. Houghtaling, and S.K. Sia, “Point-of-Care Sputum Processor” (Manuscript in preparation.)

Chapter 2 Accurate, rapid, and portable immunoassays for HIV and syphilis.

2.1 Introduction

Increasing access to appropriate treatment for infectious diseases could have a major clinical impact in developing countries since infectious diseases continue to cause an enormous burden of death and disability in those resource-limited settings [6]. HIV/AIDS and malaria accounted for a larger proportion of disability in sub-Saharan Africa than the world as a whole [4]. HIV prevalence rates remain above 10% in many countries in sub-Saharan Africa as of 2013 (Fig. 2.1) and HIV primarily affects those in their most productive years. About 40% of new infections are among those under the age 25 [23]. To slow down the spread of HIV/AIDS, it is important to diagnose and treat other sexually transmitted infections (STI’s), which significantly increase the transmission of HIV

[24]. Since there are often no obvious clinical symptoms or non-specific symptoms, accurate diagnostic tests are crucial to stop the transmission of these diseases.

Figure 2.1. HIV prevalence rates of adults ages 15 – 49 in 2013 [23].

Pregnant mothers can pass HIV and syphilis infections onto their children if left untreated, which can lead to severe health consequences for both mother and child [25-27]. Additionally, HIV and syphilis share the same routes of transmission, and the presence of syphilis has been associated with an increased risk of HIV co-infection [24]. It is estimated that between 2.5% and 17% of pregnant women in sub-Saharan Africa are infected with syphilis; recent estimates suggest that more than 535,000 pregnancies occur in women with active syphilis each year [28, 29]. Intervention to prevent HIV mother-to-child transmission and treatment of syphilis in diagnosed pregnant women are cost-effective, especially, in low- and middle-income countries where the prevalence of disease is high [30, 31]. Although syphilis testing of all pregnant women is part of the recommended basic antenatal care package recommended by WHO, less than 50% of antenatal care attendees are tested for syphilis in over 40 countries (Fig. 2.2) [32]. Our idea is that a low-cost multiplex HIV and

STIs (such as syphilis) test will help increase access to diagnosis in antenatal care to prevent significant adverse health consequences to both mothers and their children.

Figure 2.2. Percentage of antenatal care attendees test for syphilis at first visit (latest update, 2013) [32].

Immunoassays make use of the binding interactions between antigens and antibodies to detect protein markers from either pathogen or host immune responses. They are main diagnostics for infectious diseases in screening and confirmation. Early diagnosis and treatment of HIV, syphilis, and other sexually transmitted diseases in pregnant mothers reduces significant adverse health consequences to both mothers and their children [27]. Detection of host antibodies against HIV is well known and widely used in adults and adolescents on both ELISA and rapid tests. ELISA is the

“gold standard” in evaluating performance of HIV immunoassays because of their high sensitivity and specificity [33, 34] and p24 antigen detection can shorten the serological window period by 5 days [35]. 4th generation HIV ELISA test detecting both HIV antibodies and p24 antigens are currently used widely (as a screening test followed by a western blot as a confirmation test in developed countries and a reference test in some developing countries). In many developing countries where HIV prevalence remains high, a series of HIV rapid tests are used as a screening and confirmation tests.

For syphilis screening in resource-limited settings is typically performed by flocculation assays, such as the rapid plasma reagin (RPR) test and venereal disease research laboratory (VDRL) test, which detect presence of anti-cardiolipin antibodies (nontreponemal specific marker, released during cellular damage that occurs from the syphilis spirochete) and can be performed in a peripheral laboratory setting [24]. Nontreponemal test may produce false positives in case of viral infections (Epstein-Barr, hepatitis, varicella, measles), lymphoma, tuberculosis, malaria, endocarditis, connective tissue disease, pregnancy, autoimmune diseases, intravenous drug abuse, or contamination [36, 37]. Treponema-specific antibody-detection tests, such as the agglutination- based TPHA/TPPA, fluorescent treponemal antibody absorption (FTA-ABS), and enzyme immunoassays, generally used as confirmatory tests for syphilis [24, 38]. Nontreponemal test tends to be used more in syphilis endemic regions as a screening test and treponemal test is used as a confirmatory test. On the other hand, a reverse algorithm which treponemal test is used as a

16 screening test and quantitative nontreponemal is used as a confirmation test is recommended and currently used [39]. The use of only one type of serologic test for syphilis may not be sufficient for diagnosis, because each type of test has limitations [40]. Treponemal antibodies appear earlier than nontreponemal antibodies, however, treponemal antibodies tests cannot distinguish between current and past infection or treated versus untreated infections since the antibody level remains high for years with or without treatment [37, 41] except in patients treated early for primary syphilis

[42]. Patients without non-treponemal syphilis test can be treated repeatedly for syphilis for lifetime which could lead to penicillin resistant. The United States Centers for Disease Control and

Prevention (CDC) is currently recommending to perform non-treponemal (e.g. RPR) test on all anti- treponemal Enzyme Immunoassay (EIA) reactives if EIA is used for screening [43]. Particularly, pregnant women with reactive treponemal screening tests should have confirmatory testing with non-treponemal tests with titers and serologic titers should also be repeated at 28–32 weeks’ gestation and at delivery [40]. Pregnant women who are seropositive should be considered infected unless an adequate treatment history with decline in serologic antibody titers is provided [40]. The dual nontreponemal/treponemal syphilis point-of-care test may help save cost in sub-Saharan Africa where disease prevalence (and loss to follow-up) is high, while substantially reducing overtreatment in antenatal care to prevent stillbirths, neonatal deaths [44, 45]. Nonetheless, these antibody- detection tests need to be performed in a laboratory setting equipped with various instrument and constant electric supply, for example, a plate washer and a plate reader for ELISAs, a rotator and refrigerator for RPR (for VDRL, a light microscope instead of a rotator), and an incubator (or rotator) for TPHA/TPPA. These tests also require trained users to perform and/or interpret test results.

Moreover, ELISA, in particular, is done in a large batch (>80 samples) and takes several hours to complete which can lead to sample degradation, delay in treatment, or lost to follow-up.

Many clinical diagnostic methods can be miniaturized into handheld microchip-based devices, which offer many potential advantages over the same tests done in centralized laboratories. These

17 advantages include small reagent volumes, fast turnaround time, small size, low cost, low power consumption, and potential for parallel operation and for integration with other devices [3, 11, 46].

These capabilities are particularly well-suited for POC diagnostics, which can help deliver appropriate and prompt treatments and improve clinical outcomes [11-13]. We present an innovative strategy for an integrated microfluidic-based diagnostic device that can perform complex laboratory assays at the cost as low as lateral flow tests. We focus on miniaturizing the complex ELISA test, which serves as the clinical “gold standard” for detecting most protein-based biomarkers. ELISA achieves a substantial reduction in background and non-specific binding through serial washings and amplification of signal through enzyme reaction. In our assay, called “mChip”, a heterogeneous immunoassay performed on an easily manufacturable substrate, delivery of multiple reagents, and final optical-based detection of signal. We integrated new procedures for manufacturing, fluid handling and signal detection in microfluidics into a POC assay for HIV and syphilis diagnosis that replicates all steps of ELISA into a simple device that can be performed in the remote regions of the world.

Real-time synchronization of test results to patients’ health records or centralized database would enhance the healthcare delivery and efficiency as it offers reduction in human-caused error in data transcription, rapid transmission of results to health experts, improved monitoring of disease outbreaks [47], and increased effectiveness in allocating medications to communities [48]. Despite these potential advantages, it remains challenging for POC devices, which are powerful precisely because they can be operated in remote settings, to be easily integrated with patient records which are typically stored in a central location [49, 50]. Here, we integrated all modules of ELISA test into an automated format with a data communication module for a real-time synchronization of patient health record data using satellite and text messaging.

2.2 Methods

2.2.1 Microfluidic-based Immunoassay

Microfluidic cassettes.

Cyclic Olefin Copolymer (COC), transparent plastic similar to polystyrene (the same material in multiwell plates for ELISA), was used to fabricate microfluidic cassettes by injection molding

(designed and manufactured by OPKO Diagnostics). The plastic cassette has overall dimensions of 5.4 cm by 8.5 cm by 3 mm. Two designs were used in this chapter: 1) a high-throughput cassette

(perform 7 runs in parallel) (Fig. 2.3A), and 2) a single run cassette with reagent storage (Fig. 2.3B).

Each channel has dimensions 120 µm width by 50 µm depth by 138 mm length with rounded cross- sections (to reduce formation of tiny bubbles, which can obstruct flow if accumulated), and widens at both ends to a width of 240 µm to through holes of diameter 1.54 mm. Each channel has four meandering sections positioned in series, with each meandering section covering a rectangular area of 6.19 mm 2 (2.75 mm x 2.25 mm) (Fig. 2.3C) and separated by a distance of 9 mm. The meandering zones can be functionalized with different capture proteins to create a multiplexed test. A single run cassette (Fig. 2.3B) contains regions for pre-loading reagents off-site to reduce the number of steps required during each test.

Figure 2.3. Microfluidic cassettes. (A) Picture of a 7-channel microfluidic cassette. Each cassette can accommodate seven samples (one per channel), with molded holes for coupling of reagent- loaded tubes. (B) Picture of a single-run microfluidic cassette, with areas for reagent storage, analyte capture and detection, and waste storage. (C) Transmitted light micrograph of channel meanders. Scale bar is 1 mm.

Surface modification.

Similar to ELISA assays using plastic 96-well plates, we used direct physisorption of proteins on our plastic microfluidic cassettes (Fig. 2.4). We selected gp41-gp36 envelope chimera for HIV-

1/2 (Biolink International ) (see schematic below) and 17 kDa outer membrane protein of treponema pallidum for syphilis (Lee Labs ) based on initial performance comparison among other commercial antigens on known HIV and syphilis status specimens (data not shown).

Schematic diagram of HIV-1. Trimeric gp120–gp41 complexes are embedded in the membrane. The transmembrane glycoprotein gp41 and the external envelope glycoprotein gp120 are depicted in non-covalent association. The capsid protein, p24, makes up the cone-shaped core, which contains two positive-strand RNA copies of the HIV-1 genome that are surrounded by the nucleocapsid protein (yellow). Taken from [51]

As an internal positive, we chose a secondary anti-goat IgG antibody ( Invitrogen ) and no protein treatment for an internal negative control. Concentrations for physisorption were 2 – 4 µg/mL for

HIV env, 15 µg/mL for TpN17, and 5 – 10 µg/mL for anti-goat IgG Ab in bicarbonate buffer solution.

15 µL of antigen solution was pipetted on each detection zone. After physisorption of capture protein for 1 hour in a humid chamber at room temperature, we aspirated the antigen solution, washed each detection zone three times with 1x phosphate buffer saline (PBS), rinsed off the entire cassette surface with deionized water, and dried using nitrogen gas. We sealed the cassettes using a proprietary adhesive as a backing (OPKO Diagnostics ). Filter papers ( Whatman ) were used to adsorb liquid waste to eliminate the presence of biohazard waste for both environmental contamination and safety of handling in the single-channel design cassettes. Sealed cassettes then incubated the channels with blocking buffer (1% w/v bovine serum albumin (BSA) in filtered PBS)

(Sigma Aldrich ) for 45 min at room temperature to prevent nonspecific protein adsorption as well as hydrophilize plastic surface to minimize bubble forming within microfluidic channel, and then cleared the channels. The cassettes can be stored in a dry chamber at 4 degrees Celsius for long term storage.

Reagent loading.

We loaded a series of reagents manually using a 1 mL syringe to draw reagents into the PE tube which reagent plugs were separated by air spacers (Fig. 2.4A, C). For sera/plasma testing, the reagent sequence consists of one lead wash buffer plug (0.3 cm long, ~ 1.3 µL), one plug of neat or diluted sample (1.5 cm long, ~ 6.7 µL), four small plugs of washing buffer (~ 1.3 µL each), one plug (2.5 cm long, ~ 11.2 µL) of gold nanoparticle-conjugated goat anti-human IgG antibody (1.45

µg/mL in 3% BSA/0.2% Tween-20 in filtered PBS), two small plugs (~ 1.3 µL each) of washing buffer, and four small plugs (~ 1.3 µL each) of distilled water. For testing with whole blood, we metered neat sample (0.2 cm long, ~ 0.9 µL) not in the PE tubing but rather in a short thin-walled polycarbonate (PC) tube which was connected to the pre-loaded PE tube. We also added an extra wash buffer plug in between sample and gold conjugate plugs, and eliminated the lead wash plug.

In addition, the concentration and volume of gold-labeled anti-human IgG antibody was 0.4 µg/mL and 2.2 µL.

For testing using the single channel cassette (Fig. 2.3B), pre-loaded reagents in PE tube was loaded in the same manner and then loaded the solution train into the reagent storage on the cassette prior the test. Silver reagents (silver nitrate and reducing agent) were also loaded on board in two separate area and mixed by pulling through the venting port at beginning of silver reduction of each assay.

Assay operation.

We connected tubes to the inlet ports using a short PC tube connector, which was used as a metering device for whole blood testing. To generate a vacuum on a 7-channel cassette, we connected a 60 mL syringe to the outlet port using PE tube and pulled back the syringe a set distance to generate a vacuum with a metal rod to hold the piston at the set position (Fig. 2.4D).

The magnitude of the vacuum was determined by relative displacement; up to 100 kPa (~1 atm) of vacuum was possible using this method. With a vacuum pressure of 20 kPa (in the syringe), the residence times of the plugs were 2.5 min for sera/plasma samples (1 min for whole blood samples),

3.5 min for the gold-labeled anti-human IgG antibody, and 25 seconds per wash. After the final water wash, we mixed silver solutions A and B (composed of silver nitrate and reducing agent, among other chemicals) (supplied by OPKO Diagnostics) and immediately loaded the silver into the tubes.

We ran silver reagents continuously for 5 min (2.5 min for whole blood testing) and then immediately quenched the reaction with distilled water, and then left the channels dry. The overall assay time was 20 minutes or less. A single-channel cassette was operated using a handheld device with a diaphragm pump (see below).

Figure 2.4. Assay preparation and setup. (A) Schematic diagram of reagent delivery. A plastic tubing is pre-loaded with reagents separated by air bubbles in a defined sequence. At the time of assay, the plastic tubing is inserted onto the plastic chip, and the syringe is pulled back to displace a volume of air to create a vacuum. The train of reagents passes over a series of four detection zones, each characterized by dense meanders coated with capture proteins, before exiting the chip to the disposable syringe. (B) Illustration of biochemical reactions in detection zones at different immunoassay steps. Silver enhancement (i.e. reduction of silver ions on gold nanoparticle- conjugated antibodies bound on the surfaces of meandering zones) yields signals which can be read with low-cost optics (for quantification) or examined by eye. (C) (Top ) Image sequence demonstrating technique of loading small-diameter plastic tubes with reagent plugs drawn from test tubes and separated by air gaps. Marks are drawn on the tubing to indicate length ( i.e. volume) of reagent plugs. Arrows (in white) indicate reagent plugs, with three plugs positioned in series ( far right ). ( Bottom ) Manually loading reagents in tubes using a 1 mL syringe, with arrows (in white) indicating direction of air and fluid movement. (D) Overall assay setup. Picture of cassette with a tube filled with sequence of reagent plugs (here, colored dye) and syringe for generating vacuum. No other peripherals are needed to run the mChip.

2.2.2 Modeling gold-silver signal amplification

We aimed to predict extent of silver formation under a range of surface gold density and silver development time in order to minimize total assay time. To guide model development (in forming a general mechanism and determining best-fit values of kinetic parameters e.g. rate constants), we first studied a simplified experimental system of physisorbing different amounts of capture antibody to produce a range of surface gold density. We controlled the amount of antibodies to goat IgG antibody physisorbed to the surface by adding varying amounts of BSA as a competitor. We used the procedure described previously to functionalize one detection zone per channel with a particular ratio of antibodies to goat IgG antibody:BSA. We loaded all tubes in a manner described previously with the following sequence of reagents: one lead wash buffer plug (~ 1.3 µL), one plug (~ 11.2 µL) of gold nanoparticle-conjugated goat antibodies to human IgG antibody, two small plugs (~ 1.3 µL each) of washing buffer, and four small plugs (~ 1.3 µL each) of distilled water ( i.e. a reagent sequence which lacked human sample and trailing buffer washes). We ran all tests as described previously, but collected a transmittance reading of the target zone every second during silver development to generate absorbance curves over time.

The kinetics of silver reduction in our microfluidic setup reveal a sigmoid-shaped response, with the presence of an induction period followed by rapid growth of signal and termination of signal growth (Fig. 2.5d). For curve fitting, we used a variation of the four parameter logistic equation,

- = + ODmax ODmin OD ODmin - 1+ ea *(tOD ,mid t)

where OD , OD min , OD max are the optical density, minimum and maximum values respectively, t is time of silver reduction and tOD,mid is the time at point of inflection, and a is a curvature parameter

(Fig. 2.5b). Curve fitting was performed using GraphPad Prism software for nonlinear regressions.

R-squared values for best-fit curves were 0.92 (antibody to goat IgG antibody only), 0.99 (1:1 antibody to goat IgG antibody:BSA), 0.99 (1:2), 0.96 (1:4), 0.96 (1:8), and 0.71 (BSA only).

Together with the development of silver precipitates (established by AFM [52, 53] and SEM [54] in related systems), the silver reduction in mChip is believed to start similarly with the catalytic formation of in situ silver nanoclusters around gold particles [55, 56] and undergo a general mechanism involving a nucleation step followed by a autocatalytic surface-growth step[57, 58]:

The mechanism for the termination of signal growth may be due to non-linearity between absorbance and silver at high silver density, or reduced silver deposition due to agglomeration of nanoclusters to catalytically-inactive bulk metallic silver [57, 58].

We focused on the first 5 minutes of silver reduction, during which sufficient signal is generated and where nucleation and autocatalytic growth are prominent (Fig. 2.5d, box outlined in red ). We assume excess reducing agent ( e.g. hydroquinone), irreversible reactions, fast adsorption of reactants onto gold surface, even distribution of gold density across the detection zone, and negligible silver precipitate desorption. Based on this reaction mechanism we formed the following rate of silver formation:

d = + [Ag(0)] k1SAu [Ag(I)] k2 [Ag(I)][Ag(0)] dt

with rate constant of nucleation ( k1), rate constant of growth ( k2), concentration ([ ]), and active surface density of gold catalyst SAu (in units of moles of gold nanoparticles per square meter of substrate surface). Due to attachment of silver precipitate around gold nanoparticles, the number of

25 active catalytic sites on the gold nanoparticles diminishes as the reaction progresses. We therefore express SAu as a function of initial gold surface density bound ( SAu,i ) and the extent of nucleation reaction, x

S = S (1- x ) Au Au,i

To estimate SAu,i , we assumed (1) equal rates of physisorption between antibody to goat IgG antibody and BSA, (2) a surface density of 0.5 m g/cm 2 of antibody to goat IgG antibody (with molecular weight 150 kDa) for the antibody-only experimental condition [59], (3) a 1:1 capture ratio of antibody to goat IgG antibody : gold-conjugated goat antibody to human IgG antibody, (4) a 1:1 labeling ratio of gold nanoparticle : goat antibody to human IgG antibody conjugate. Values of SAu,i for each experimental condition are given in Fig. 2.5a. To relate the amount of captured antibody with SAu,i , we used an average gold nanoparticle diameter of 10 nm:

S = f * 3.3x10- 8 Au,i ragAb

where fragAb is the percentage of antibody to goat IgG antibody (relative to total protein) in the physisorption solution, and SAu,i is expressed.

For estimating x , we define a time tn, beyond which negligible silver is formed from nucleation on gold nanoparticles (i.e. since all the surfaces of gold nanoparticles are already covered by reduced silver), by finding the intersection between lines tangent to best-fit curve of reduced silver formation at tOD,mid and t~0 (Fig. 3b). We then express active catalyst surface area as

 t  S = S 1 -  Au Au,i  t  n

The dependence of tn on gold nanoparticle density captured on the surface is shown in Fig. 2.5c; we generalize the relationship with an exponential decay fit

- t = (t - t )e k *SAu + t n n,max n,min n,min

where tn, max , tn,min are tn at 100% and 0% of antibody to goat IgG antibody surface coverage and k is a curvature parameter. R-squared value is 0.97.

We modeled the effect of flow parameters on the kinetic of reduction, by coupling the convection/diffusion equation with the rate of silver formation at the boundary layer using weak-form formulation (and assuming a parabolic velocity profile in the microchannel):

¶ + - + = - d [Ag(I)] ( DAg(I) [Ag(I)] [Ag(I)]u) [Ag(0)] ¶ t dt

 2   y - 0.5h u = u 1-    0   0.5h   

-10 2 -1 -11 3 -1 with diffusion constant DAg(I) (1 x 10 m -s ), flow rate Q (5 x 10 m -s ), channel height h (50 x

-6 -6 10 m), channel width w (100 x 10 m), velocity u and max velocity u0 (= 3* Q/(2* h*w)), and initial concentration of Ag(I) (0.01 mol-m-3).

We then tuned the parameters by comparing the model output with the experimental data for 5 different gold concentrations. Nucleation rate constant ( k1) and autocatalytic rate constant ( k2) were determined according to minimization of the model error. There are several optimization algorithms available to minimize the magnitude of an objective function ( e.g. the sum of the errors between the model output and the experimental data over time and over the five different gold concentrations).

We chose “ pattern search algorithm ” (Direct search toolbox, Matlab) since it handles the constrained nonlinear optimization problems in a reasonable timeframe and does not require the function to be differentiable and continuous. Minimizing the objective function, we determined the nucleation rate

-6 -1 3 -1 -1 constant ( k1) and autocatalytic rate constant ( k2) to be 10 s and 20 m mol s respectively. (We

27 converted silver surface density to optical density by assuming a linear relationship with scaling

OD = factor of t 5min ). [ Ag(0)]t=5min

2.2.3 Custom-built devices

Optical density reader

The amount of analyte is quantified by reduction of silver ions onto gold nanoparticles. The custom-built reader was designed for analyzing the optical signals from microfluidic chips. We used two reader prototypes: one which can take measurements of a single detection zone at once [60], and another which can take simultaneous measurements of four detection zones (shown in Fig.

2.6). The dimensions are 5.5 x 5.5 x 4.5 inches and 6.5 x 6.5 x 5.5 inches, respectively ( l x w x h ).

Both readers consist of a detection unit and a display unit, with all components controlled by an

Atmel Mega32 microcontroller (Digikey ). The detection unit consists of a super bright LED (660 nm) as a light source and a photodetector (photo sensitivity at 660 nm is 0.36 A/W) (Hammamatsu ) to measure the light passing through the silver film on microfluidic chips and have black Delrin plastic plates (McMaster-Carr ) with 2 mm diameter pinholes (aligned above each detection zone) to reduce noise from ambient light. The top plate is attached to linear bearings for adjusting height when switching between cassettes. Both readers also use a custom-designed microcontroller which is mounted underneath the bottom plastic plate and contains an Atmel Mega32 microcontroller (for converting analog signals of transmitted light to digital values), among other components. Readings are taken upon a button push. The transmittance values ( i.e. mean values over a set of 32 rapid measurements over 1 s) are displayed on a liquid crystal display (LCD) (Digikey ) and are logged manually or automatically log the values and variance of transmittance in a laptop computer through a USB port.

The absorbance (optical density) can be determined from [61]

  = -  I  Al log10   I  0  , where I is the intensity of light at a specified wavelength λ that passes through a silver opaque and

I0 is the intensity of the light that passes through a blank channel. To reduce noise from the surrounding light, black Delrin plastic plates are used to cover the microfluidic chips and hold the electronic and optic components in place. Furthermore, a 2 mm diameter pinhole is aligned on top of the photodetector to eliminate stray light that is outside the detection zone.

The result or disease status is determined by comparing to the OD threshold or cutoff. Result is determined to be positive when the value is above the threshold and negative when it is below the threshold. ODs of internal negative and positive controls are used to determine validity of each run.

We also note that although hard threshold values were used for purpose of evaluating sensitivity and specificity, in practice, ODs close to the threshold values could be interpreted as an indeterminate result and would motivate additional testing in a centralized laboratory.

Validation of the reader. We compared measurements taken from the reader with those taken from a flatbed scanner ( Hewlett Packard ) as trends in normalized values should be similar despite differences in light source (red vs. white light) and sensor position (detecting transmitted vs. reflected light). We analyzed image intensity using ImageJ ( National Institutes of Health ).

Fully integrated device for immunoassay

We integrated three important off-chip processes in a diagnostic test - liquid handling, optical signal detection, and data communication – in a low-cost, versatile, handheld instrument for performing immunoassays at the point of care. The technology encompasses three main steps of a

29 diagnostic test: (1) a small diaphragm pump for controlling fluid flow on a microfluidic cassette that is preloaded with plugs of different reagents and samples, (2) a set of robust and low-cost LEDs and photodetectors for quantifying absorbance of silver development, and (3) a compact unit for communicating results in real time via satellite or SMS.

Liquid handling. This module includes a diaphram micropump (Fig. 2.7B, component 7) to generate negative pressure, a pressure regulator (Fig. 2.7B, component 6) to regulate the pressure applied to the fluidic chip, and two pressure sensors to monitor pressure at the outlet of the pump and the inlet of the chip in real-time. As the micropump is turned on, the pressure in the first vacuum chamber (Fig. 2.7B, component 4) decreases. It shuts off automatically (usually after turning the micropump on for about 20s) when the negative pressure reaches a set value (e.g. -50kPa).

Meanwhile, the vacuum regulator maintains a constant negative pressure of smaller magnitude (e.g.

-20kPa) in the second vacuum chamber (Fig. 2.7B, component 5) to which the outlet of fluidic cassette connects. The micropump is turned on again when the negative pressure in the first vacuum chamber increases beyond the set value. This technique helps reduce the amount of power consumption of the micropump by minimizing the power-on time. With the use of pressure sensors, we are able to precisely control the pressure drop and flow rate in the microfluidic channel.

Signal detection. This module consists of four pairs of low-cost super-bright red LEDs (Fig.

2.7B, component 9) and photodetectors (Fig. 2.7B, component 8) and a signal amplifying circuit similar to the optical density reader described previously. Each sets of LEDs and photodetectors are aligned with the four detection zones of the cassette and fixed in place, hence, no adjustment or alignment is required for each reading. The absorbance value is obtained and computed similar to above.

Results communication. We integrated a compact satellite modem (Iridium 9601 SBD

Transceiver, Fig. 2.7B, component 10) and a GMS/GPRS module (LinkSprite SM5100B, Fig. 2.7B,

30 component 11) to our instrument to communicate diagnostic results from a remote setting to a centralized database. The satellite modem operates with a data service called “short burst data”

(SBD) provided by the Iridium satellite network. SBD has low operational cost, with minimal latency in data transmission (approximately 5 seconds) and global coverage. The basic architecture of the

Iridium system includes the satellite network, the ground network (Earth gateways), and a satellite modem. When a user initiates a data transmission from the field, the modem connects to an overhead satellite, and the data is relayed among satellites around the globe until it reaches a satellite that is above the appropriate Earth gateway, which downloads the data and send it as an email to a pre-designated email address via the Internet. We also connected an extendable antenna to our instrument to enable RF communication with the satellite.

The GSM/GPRS module can connect to any portable network operating in the EGSM900,

GSM850, DCS1800 or PCS1900 band frequency, according to the inserted SIM card. The module is able to communicate with the microcontroller, receive the coded results and send it as a text message to a pre-stored recipient number. The cost of the text message depends on the network provider, usually about 10 cents.

Messaging protocol. In order to minimize the size of data to be transmitted (and hence reduce the cost of data transmission) and security and privacy of data, we formatted all the data, including date/time of test, patient’s information, and test results, into a 15-byte binary string. As opposed to sending data on a character-by-character basis, our messaging format treats each type of data (e.g. date of test, gender) as a numerical value and transmits it as a binary string. We believe this format will allow us to significantly reduce data size while maintaining the amount of information being transmitted. For instance, transmitting the test date “123109” (ie. 12/31/2009) as 8-bit characters will require 48 bits, whereas transmitting it as a binary string “11110000011100101” (binary representation of 121309) will require only 17 bits. Our streamlined messaging format is key to maintaining a secure and low operating cost to transmit POC diagnostics results.

Automation with a microcontroller. We used an 8-bit microcontroller to control and integrate all the three modules together. All the electronic components (including the microcontroller, power regulator, and pressure sensors) were mounted onto a small custom designed printed circuit board that operates on a 9V battery. We programmed a software routine in C into the microcontroller of our POC instrument such that: 1) it prompts the user to initiate a test; 2) controls the liquid handling and signal detection modules with precise timing; 3) at the end of the test, it computes absorbance values, displays results to the user on a liquid crystal display (LCD), and prompts the user to perform data transmission; 4) when a button is pushed subsequently, the microcontroller formats the test data into a string of binary numbers and initiate a transmission session with the satellite modem or the GSM/GPRS module using AT commands (a set of standard instructions for controlling modem);

5) and reports the result of data transmission to the user. We used an Atmel AVR© STK500 development board along with the AVR Studio 4 IDE to program the microcontroller (Atmel

ATmega32).

2.2.4 Limit of detection on human IgG immunoassay

To determine the lowest possible concentration of human antibodies detected, we developed an immunoassay to detect human IgG antibodies. In order to prepare the microfluidic assay, goat- anti-human IgG antibody 4 µg/mL in carbonate-bicarbonate buffer was spotted on plastic cassette and incubated in humid chamber for 1 hr at room temperature. After that we washed each spot three times with PBS, sealed the cassette with an adhesive tape and blocked with 1%BSA-0.05%Tween-

20 in PBS (45 minutes at room temperature or overnight at 4 ºC). To perform the assay, we loaded human IgG antibody (6.7 pM – 67 nM) in 1%BSA-0.05%Tween-20 in PBS, followed by four plugs of PBS/Tween-20, gold conjugated anti human IgG antibody (55 times dilution in 3%BSA-

0.2%Tween-20 in PBS), two plugs of PBS/Tween 20 and four plugs of deionized water. Next, we

32 loaded the silver solution (a mixture of silver salts and reducing agent) to the channel and flowed for 5 minutes. The silver film intensities were measured by the instrument and calculated to absorbance values.

2.2.5 Dual HIV/syphilis assay evaluation on commercial specimen panels

For single-analyte testing performed at Columbia University, 6.7 µL of sera/plasma from panels purchased from Zeptometrix (catalog numbers CN6710 and K-ZMC002) and Seracare (catalog numbers QHV711, PRB204, QSS701, and VRZ602) in a total of 67 samples were used. We performed each assay as described in previous sections. Samples used for evaluating HIV test performance had been validated with commercial ELISA tests and/or rapid tests by third-party suppliers; in some cases, additional testing with Western Blot was also provided by the supplier.

We reported HIV status based on vendor-provided results from reference test(s): positive (i.e. pos), negative (i.e. neg), and indeterminate (for samples which had conflicting results from independently- performed commercial reference tests). Nine samples from commercial specimens exhibited conflicting reference results. Since the true disease states of these specimens are not clear, we did not include them in the analysis. For syphilis testing, we ran positive samples that were previously independently validated with treponema-specific tests (and, for several of these, the RPR status was also known); for syphilis negative samples, RPR results were available and used as the reference standard.

2.2.6 Assay evaluations on archived specimens at the POC settings

First field evaluation using simplify immunoassay setup in Rwanda

The study was reviewed and approved by Rwanda National Ethics Review Committee, Rwanda

Ministry of Health and Columbia University Institutional Review Boards (IRB). We ran whole blood samples (validated for HIV) at Muhima Hospital, a district-level hospital in Kigali, Rwanda. The reference testing algorithm for HIV (schematic below) is a series of rapid tests (First Response, Uni-

Gold, and Capillus); HIV positive samples must be positive for all three or else discrepancies are resolved by ELISA, which is unavailable at Muhima and therefore performed off-site (at the National

Reference Laboratory). Samples which test negative are only tested once (with First Response, the most sensitive of the three rapid tests).

Muhima experiences regular frequency (~5-20 samples per day) of HIV testing for PMTCT and

VCT counseling as well as for HIV/AIDS monitoring. This allowed testing of patient whole blood samples which were recently collected via venipuncture (our ethics protocol allowed only for testing

34 of archived, validated specimens). We tested 70 patient samples, of which 42 were positive and 28 were negative. Gender was not known for HIV-positive samples (most of which were collected from patients getting CD4 counts). There were about equal numbers of women and men for HIV-negative sample group. All patient samples were coded.

In addition we ran HIV-syphilis duplex tests at Projet Ubuzima, an international NGO registered in Kigali, Rwanda. They are running a HIV/STI incidence study on female commercial sex workers and female clients of public HIV testing centers in Kigali, and have access to validated blood samples for both HIV and syphilis. Of a total 67 patient samples tested, 66 were sera and one was plasma. HIV positive samples were first validated with a series of rapid tests (First Response, Uni-

Gold, Capillus) and confirmed with ELISA (Abbott Ag/Ab Combo Murex). Syphilis positive samples are first validated with RPR (SpinReact) and confirmed with TPHA (Human). For RPR and TPHA, titers are expressed as the highest dilution factor in which samples yield a positive result. All patient samples were coded. There were no criteria for selection other than specimen availability, which was the main constraint as these samples were also used for other studies.

Second field evaluation using a fully-integrated device in Rwanda

We demonstrated the full capacity of our POC instrument on immunodiagnostic assays on 40 whole blood samples from both HIV positive and negative patients, and transmitted the diagnostic results via satellite and/or SMS immediately after assay completion in Rwanda. We tested samples collected at Muhima hospital, Rwanda. We performed each test in the following manner: we drew less than one microliter of whole blood in a capillary tube (e.g. a fingerprick volume) (step 1), connected the sample-loaded tube onto a reagent-loaded cassette and inserted into instrument

(step 2), started assay with a single button push, and after completing the assay, we displayed the optical density of each meandering zone on the LCD screen (step 3). After prompted to transmit

35 data packet (step 4), we sent both by satellite, which sends an email message with the data as an attachment to a pre-designated address (step 5a), and by SMS, which delivers the message to a pre-designed cellphone number (step 5b).

Evaluation of test performance

Performance of test is reported in term of sensitivity and specificity compared to reference test.

Given the number of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN), the performance metrics are:

TP sensitivity = FN +TP

TN specificity = FP +TN

95% confidence intervals are calculated using the formula:

p(1- p) p – 1.96 * n where p is sensitivity (or specificity) expressed as a proportion and n is the number of samples analyzed [62]. For cases where 100% sensitivity or specificity was reported, a p-value of 0.999 was used. Scatter plots and receiver operating characteristic (ROC) curves were created using

GraphPad Prism.

2.3 Results and Discussion

Injection-molded disposable cassettes

Traditional manufacturing technologies for microfluidics (such as those for silicon, glass, or polydimethylsiloxane (PDMS) materials) are useful for prototyping, but are too expensive and slow for mass production of cassettes. Injection molding is an easily scaleable strategy for manufacturing hard-plastic components with features on the order of millimeter and above. Here, we manufactured microfluidic cassettes with spatially precise features, ranging from 1 µm to 1 mm, in transparent cyclic olefin copolymer (Fig. 2.3). Like lateral flow tests, each cassette costs less than 10 cents in material cost and can be easily produced at a high throughput (1 chip every ~40 seconds). Plastic surface (meandering zones) is functionalized with capture protein using direct physisorption similar to ELISA well-plate.

Automated delivery of multiple reagents for multi-step reactions

In ELISA, multiple additions of reagents and washings ensure high reproducibility and signal- to-noise ratio. In the mChip, we integrated a bubble-based method of reagent delivery (in which metered plugs of reagents are introduced sequentially into a tubing separated by air spacers) to deliver 14 separate reagents (consisting of antibodies, washing solutions, and signal development solutions; Fig. 2.4A). Hand-loading of regents can be performed by local workers at a location and time close to when and where the test will be conducted (Fig. 2.4C). We can also load the reagents and prepare the microfluidic cassettes in an automated fashion with the aid of robots in manufacturing facility. The use of passive reagent delivery with stored plugs has not previously been demonstrated to be compatible with whole-blood specimens, which can dramatically increase number of non-specific interactions. Also, this passive delivery method, with a succession of plugs

37 of reagents and air, has not been previously shown to work smoothly in microchannels that meander with sharp bends. This method can be performed in the field to deliver fluids with precise volumes and incubation times that rivals the bulky pipetting robots used in laboratory-based ELISA.

Signal amplification and detection using minimal instrumentation

ELISA achieves a strong signal due to enzyme-mediated signal amplification. Rather than forgoing signal amplification as in most lateral flow assays, we made use of reduction of silver ions onto gold nanoparticles in an immunosandwich [63, 64], a procedure that allows a signal to be amplified on a solid substrate under continuous flow of fluid (Fig. 2.4B). We also used a “meandering channels” design (Fig. 2.3C) to convert a signal developed from nanoparticles inside microchannels to a millimeter-sized scale that can be detected with a wide optical beam without lenses or fine optical alignment [63]. The optical density of the silver film can be measured using low-cost and robust optics such as light-emitting diodes and photodetectors, which we incorporated into a simple reader (see below in section 2.3.2) while most microfluidics-based assays require a microscope or bulky sensitive detector to detect low signals generated inside microchannels. We used this signal amplification strategy in microfluidics to detect antibody markers against HIV and syphilis. The silver reduction generates sufficient signal within 5 minutes, with only minimal detectable background development.

Modeling of gold-silver signal amplification

To characterize the influence of gold and time on silver formation, we collected experimental data using a modified assay which controlled the surface density of captured gold-labeled antibodies

(Fig. 2.5d). The kinetics of silver reduction revealed a sigmoid-shaped response which agrees with

38 observations in related systems [56], and confirmed that considerable signal could be generated in

5 minutes or less over a range of surface gold density. We used a transport-reaction model, which approximates empirical rates of silver growth to within 10% (Fig. 2.5), to guide designs to minimize total assay time.

SAu,I are estimates based on literature. b, Silver enhancement of zone functionalized with 1:2 antibody to goat IgG antibody : BSA physisorption ratio. Data points are mean absorbance values, and dashed line is best-fit curve (four parameter logistic equation). Parameters are indicated accordingly. c, Dependence of tn on gold nanoparticle density captured on the surface, with best-fit curve (exponential decay) as dashed line and best-fit parameters listed in adjacent table. d, Experimental kinetic data of silver enhancement for various antibody to goat IgG antibody : BSA physisorption ratio. Data points indicate mean values, error bars indicate one s.d. Dashed lines are best-fit curves. e, Computational modeling results (solid lines) superimposed with experimental data points. The difference (expressed as normalized objective function) between the model and experiment is 9.2%.

Optical density reader.

The optical density (OD) of the silver film can be measured using low-cost and robust optics such as light-emitting diodes and photodetectors ($0.50 and $6.00 per unit, respectively), which we incorporated into a low-cost compact device. We used two reader prototypes: one which can take measurements of a single detection zone at once [60], and another which can take simultaneous measurements of four detection zones (shown in Fig. 2.6). The dimensions are 5.5 x 5.5 x 4.5 inches and 6.5 x 6.5 x 5.5 inches, respectively ( l x w x h ).The 5.5 x 5.5 x 4.5 inches device is operated by a 9 V battery (or an AC adaptor), which is suitable for use in resource-poor environments (without ground electricity). Figure 2.6C shows the OD values from an immunoassay detecting different concentrations of captured human IgG as well as the reproducibility of the reader. The optical densities calculated from the reading values are proportional to the concentrations of captured human IgG (and more directly to bound secondary gold-conjugated anti-human IgG antibodies), as silver film opacity is a function of the concentration of the analyte. We differentiated disease status

(positive and negative) of samples based on threshold absorbance values. The coefficient of variance (CV) (calculated from five optical densities of each analyte zone) shows that lower OD we get, the higher CV present. As CV is calculated by the ratio of the standard deviation to the mean, it provides a high value at the mean value close to zero. However, the standard deviation is low enough (< 0.01) to distinguish the OD at the low value. This OD range can be applied as a semi- quantitative measure for POC diagnostics. This reader may be applied to detect signals from any

40 microfluidic platforms which leads to POC diagnostics and global health development for resource- poor environments.

Figure 2.6. Optical density reader for signal detection. ( A) A photo of the custom-build reader showing the user interface and detection unit for 4-detection zone. ( B) Schematic of the instrument components showing optics, electronics and component structure. ( C) The gradients of optical density and coefficient of variance calculated from five optical densities of each analyte zone.

Fully integrated device for immunoassay

We have previously described how immunoassays can be run on a microfluidic cassette via a series of manual steps: loading a low volume of whole blood into the card, connecting an external tubing that contains all the assay reagents, and using an external syringe to draw the sample and reagents through the card (section 2.3.1, Fig. 2.4D). However, the requirement for manual steps in diagnostic tests can lower their accuracy when performed by minimally trained users as well as decrease access in settings without trained users [65-67].

Figure 2.7. A fully integrated device. ( A) A picture of mChip instrument (with cassette inside). ( B) Inside view of our POC instrument. Our design includes three main modules: liquid handling (highlighted in orange), signal detection (highlighted in red), and data communication (highlighted in green).

We designed a handheld instrument that mimics the major functions of pipetting machines, microplate reader, desktop computer, and communication hardware, all without the need for grid- based power and at a fraction of the instrumentation cost of benchtop ELISA (Fig. 2.7) to operate a

POC diagnostic test on a disposable microfluidic cassette (Fig. 2.3B). The device includes three sets of components: (1) a small diaphragm pump and vacuum regulator for controlling fluid flow in the microfluidic card, (2) robust and low-cost light-emitting diodes (LEDs) and photodetectors for quantifying the optical signal, and (3) a communication module with cell-phone towers and globe- orbiting satellites that synchronizes test results in real time to a centralized server from anywhere in the world. The modules are controlled by an on-board eight-bit microcontroller and a custom designed circuit.

To replace the myriad function of multiple benchtop instruments at sufficiently low cost and energy consumption to be suitable for resource-limited settings, we integrated components and concepts from a variety of sources of advanced technologies, including non-medical applications.

To ensure a low cost of components for signal detection, we used red LEDs as light source, and photodetectors to measure absorbance. For data communication with a remote server, we incorporated two wireless technologies: a satellite transceiver using an inexpensive short-burst data

(SBD) service with the Iridium satellite service to ensure global data coverage for our device, and an inexpensive GSM/GPRS transceiver to allow communication via local cell-phone towers.

Because our data transmission requires only short data strings, we can use the inexpensive SBD mode [68] of the robust satellite communication networks at a similar cost as SMS messages via cell-phone network. Finally, the mobile device runs on disposable microfluidic cards that can be mass manufactured in plastic for pennies in material per piece [65].

Handheld mobile devices that are simple to charge and use, like cell phones, can gain rapid adoption in the developing world. All components in the mChip device are powered by a single 9V battery. The single component that consumes the most power (27% of total power consumed) is the

43 diaphragm pump. To drastically minimize the power consumption of the device, we connected the pump to a vacuum chamber and a tunable vacuum regulator; the output of the regulator was set to

~-20 kPa. This setup conserved energy by minimizing the time the pump was turned on. The pressure inside the vacuum chamber was continuously monitored by a pressure sensor which output pressure readings to the on-board microcontroller; the pump was automatically activated when the pressure in the vacuum chamber rose above a set value (~ -30 kPa) and for only a brief period of time (~30 seconds) until the pressure dropped below a threshold (~ -50 kPa).

To ensure simplicity of use, we packaged the components inside a casing (24 x 11 x 11 cm) with a single push button lit either green (ready-to-start) or orange (assay in progress). After running each test, the results were displayed on a liquid crystal display. We decided on implementing these and other aspects of the design after consultation with end users in Rwanda, accompanied by an industrial design partner.

Baseline operation of device on microfluidics card. We characterized the baseline performance of the device on a microfluidics card by measuring two parameters: power consumption and control over flow rate in the microfluidic chip. First, we measured the amount of current drawn by the device during a clinically relevant assay (Fig. 2.8). In this assay, a whole blood sample with known HIV- positive status was validated correctly by the device in 17 minutes. All the incubation and washing steps of benchtop ELISA were faithfully replicated in our mobile device; in the absorbance trace, high absorbance during sample incubation resulted from opacity of whole blood, while the absorbance peaks in between washes were due to air gaps between plugs of reagents. The instantaneous drawn current remained largely at 48 mA throughout the assay, punctuated only by current spikes due to operation of pump (initiated during start of assay and prior to entry of silver reagents in analysis zones), absorbance measurement, and data transmission. The average power

44 consumption per test (including data transmission) was 0.62 W, comparable to an average power consumption of 0.75 W for a mobile phone and much less than 20 to 60 W for a laptop computer

[69].

Flow rates in microfluidic systems can affect assay performance [70]. In benchtop ELISA, control of incubation times, volumes, and mixing is achieved by pipetting robots. Here, we could adjust the strength of the vacuum pump in the device to precisely control the flow rate in the microfluidic card from 2.5 to 20 µL/min (Fig. 2.9).

Figure 2.9. Control of flowrate inside microfluidic chip by adjusting vacuum pressure. Error bar is one standard deviation (n=3).

Validation of a fully integrated device for POC testing

We assessed two factors important for POC testing: robust outdoor operation (to test environmental ruggedness) and temperature stability. As a demonstration of the feasibility of outdoor operation, we tested a HIV-positive and a HIV-negative plasma sample three times on the battery-powered mobile device (Fig. 2.10A). We saw little difference between results from battery- powered operation performed outdoors compared to operation performed indoors using ground electricity (dashed lines in Fig. 2.10A). For a HIV-positive sample, outdoor operation yielded signal- to-cutoff result of 2.2 ± 0.3 (mean ± SEM), similar to indoor signal-to-cutoff result of 2.5. For a HIV- negative sample, outdoor operation gave a signal-to-cutoff result of 0.4 ± 0.0, compared to indoor signal-to-cutoff result of 0.4. Also, we investigated whether or not long-term storage of pre-treated cassettes was possible (Fig. 2.10B). After storing pre-treated cassettes at 4 oC for three months, we ran tests using the same HIV-positive and HIV-negative sample and found little difference between results from stored cassettes and freshly-prepared cassettes (dashed lines in Fig. 2.10B). For a

HIV-positive sample, running stored cassettes yielded signal-to-cutoff result of 2.6 ± 0.3, compared to a freshly prepared cassette which gave signal-to-cutoff result of 2.5; for a HIV-negative sample, outdoor operation gave a signal-to-cutoff result of 0.3 ± 0.0, compared to a freshly prepared cassette

46 which gave a signal-to-cutoff result of 0.4. Together, this data suggests that the HIV immunoassay on mChip can be operated outdoors and on battery, and that cassettes can be prepared and stored until on-site use.

Figure 2.10. Assay characterization on outdoor operation and temperature stability. ( A) Outdoor HIV testing on battery-operated instrument. Shown as signal-to-cutoff values from the same positive specimen (n=3) and negative specimen (n=3). Dashed lines represent signal-to-cutoff values from specimens run on freshly prepared cassettes. (B) Thermal stability of HIV immunoassay. Cassettes were prepared and stored for over 3 months at 4 oC, then ran on the instrument. Shown are signal-to-cutoff values for a positive specimen (n=5) and negative specimen (n=3) validated from a commercial supplier (Seracare). Dashed lines represent s/co values from specimens run on freshly prepared cassettes.

Limit of detection on human IgG immunoassay

To characterize analytical performance, we first used the mChip device to detect a standard marker, human IgG [71, 72]. For this experiment, we used a disposable plastic tube that was loaded with samples containing 26 µL of human IgG with concentrations ranging from 6.7 pM to 67 nM, along with 15 separate reagents for washing and signal amplification. Compared to procedures described previously which used tube-based loading of reagents [73, 74], this device enabled automatic alignment of analysis zones to the optics (which gave more reproducible values of optical

47 density as compared to manual alignment), activation of vacuum using a single button-push, and on-cassette retainment of waste fluids. For this antibody-antigen pair, the limit of detection of our

POC device was 167 pM (Fig. 2.11), a sensitivity level that is comparable to conventional ELISA

(52 pM for double antibody sandwich, HRP-labeled, colorimetric, total human IgG antibody kit from

Immunology Consultants Laboratory). The dynamic range of mChip was greater than that achieved in commercial ELISA (400-fold compared to 64-fold), which can avoid the need for repeated measurements to adjust sample dilution for bring analyte concentration into range of calibration, a step commonly performed for human IgG measurements. Additionally, the results are achieved in less than 20 minutes, as opposed to several hours, because heterogeneous reactions often take place more quickly in microchannels than in multiwell plates [75].

Dual HIV/syphilis assay evaluation on commercial specimen panels

Our test simultaneously detects antibodies against HIV and treponema pallidum (the causative agent of syphilis) from needle-prick sample volumes of blood samples ( i.e. sera, plasma and whole

48 blood). We chose a HIV/syphilis combination test because HIV and syphilis are treatable in diagnosed pregnant mothers [76], for whom short-course antiretroviral prophylaxis reduces transmission of HIV [26], and treatment with penicillin reduces congenital syphilis [66], which can be fatal for the newborn.

In the mChip, we used an envelope antigen for capturing anti-HIV antibodies (gp41 and gp36), and the 17 kDa outer membrane antigen (TpN17) for capturing anti-treponema antibodies (Fig.

2.4B) (these antigens produced the best results from 20 different antigens screened; other optimization steps such as selection of an appropriate internal positive control and validation of optical reader were also completed, data not shown). To investigate the performance of these antigens in a research laboratory, we conducted two parallel sets of assays on validated commercial sera and plasma sample panels, one for HIV (70 samples), and one for syphilis (38 samples). We used the method in Fig. 2.4, which is simple to operate, requiring only a syringe in additional equipment. We used a low-cost handheld reader (section 2.3.2) to measure the optical absorbance for each zone. We differentiated positive and negative samples based on threshold absorbance values. The test sensitivities, for detection of anti-HIV and anti-treponema antibodies, respectively, were 100% (with 95% confidence interval of 99.2-100.0) and 94% (86.4-100.0), with specificities of

100% (99.2-100.0) and 81% (68.5-93.5) (Fig. 2.12). The performances of both tests were comparable to that in current commercial ELISA kits (~100% sensitivity and 98-100% specificity for tests detecting anti-HIV antibodies [77], and 82-100% sensitivity and 97-100% specificity for syphilis antibody tests [24]); for syphilis, non-ELISA tests such as RPR and TPHA serve as laboratory “gold standards” but cannot currently be miniaturized into POC tests.

Assay evaluations on archived specimens at the POC settings

First field evaluation using simplify microfluidic immunoassay in Rwanda

Many technologies work well in a lab but not in the field, which exhibits additional complexities not replicated in a research laboratory: local specimen strain and subtypes, disease prevalence, and real POC infrastructures, environmental conditions, and testing protocol [62]. We tested our hypothesis that advanced miniaturization can ultimately simplify a complex assay to an extent that it can be operated in the most remote regions of the world, where some of the most clinically dire patients reside. In particular, we tested our device in Rwanda (in a collaboration approved by the

Rwanda National Ethics Committee and CU IRB), where the HIV prevalence nationally is ~3% and as high as 8% for women in Kigali [78]. First, we tested the device in Muhima Hospital in Kigali, which lacks access to on-site ELISA assays. Current turnarounds for results take days or weeks due to the need to send the blood samples back to the National Reference Laboratory. In this assay, we used less than one microliter of unprocessed whole blood (pre-collected whole blood in either

50 heparin or EDTA vacutainers and stored in 4 ⁰C for 1 – 7 days prior to use). (Such a low volume enables collection of blood from a variety of patient types, including infants, for whom even 10 µL is difficult to obtain from a needle prick. Additionally, whereas many ELISA assays can run on whole blood, processing of whole blood in microfluidic devices is often non-trivial due to clogging and coagulation). The assay took less than 15 minutes to complete (since the sample volume is much smaller than the sera/plasma assay). Out of a total of 70 specimens with known HIV status (about half male and half female), only one tested false, resulting in overall sensitivity of 100% (99.2-100) and specificity of 96% (91.4-100) (Fig. 2.13), rivaling the accuracy of lab-based HIV testing. Hence, this experiment verified the ability of our microfluidic device to perform test on whole blood, but unlike ELISA, the assay was performed on-site, faster, and consumed lower volumes of sample and reagents.

Figure 2.13. Field results of HIV immunoassay collected in Muhima Hospital in Rwanda using less than 1 µL of unprocessed pre-collected whole blood sample. (Left ) Signal-to-cutoff ratios for positive and negative samples. (Right ) ROC curve.

We also investigated the ability of mChip to perform a combined HIV/syphilis test. ELISA exhibits an attractive feature of testing multiple biomarkers in parallel, by testing specimens in separate wells using different reagents. This feature is difficult to achieve in lateral flow assays,

51 where a common buffer is often inappropriate and non-optimal for detecting more than one biomarker [20]. Here, we ran HIV-syphilis duplex tests at Projet Ubizima in Rwanda (Fig. 2.14), using samples collected at the clinic for a separate research study on commercial female sex workers in Kigali. In the duplex version of the mChip, separate zones on the same microfluidic cassette contained different capture reagents to allow for detection of more than one biomarker.

We analyzed 67 sera and plasma samples, and used the threshold absorbance values determined previously (in Fig. 2.12) to identify which samples were positive or negative for both diseases. Our duplex test exhibited a sensitivity of 100% (99.1-100) and 94% (88.3-99.7) and a specificity of 95%

(89.8-100.0) and 76% (65.8-86.2) for HIV and syphilis, respectively (Fig. 2.14). The syphilis test was able to detect anti-treponemal antibodies in samples ranging from weakly reactive (low titer) to strongly reactive (high titer). Hence, the duplex test as conducted in the field exhibited performances similar to lab-based reference tests [24, 77] for these two diseases.

52 values greater than 10 are shown at 10, and those less than 0.1 are shown at 0.1 (both with arrows). (Right ) ROC curves for HIV and syphilis.

Second field evaluation using a fully-integrated device in Rwanda

To minimize the number of steps needed for running a POC test in resource-limited settings, we used an HIV immunoassay with on-cassette reagent storage and that allowed testing of whole blood, as shown in Fig. 2.7. We tested 40 whole blood samples pre-collected at Muhima hospital, a district-level government-run clinic in Kigali, Rwanda, which does not normally perform ELISAs and relies solely on rapid tests for diagnosing HIV. Whole-blood samples were collected from patients recently presenting to the clinic for antenatal care, voluntary counseling, and/or HIV/AIDS testing.

We performed each test in the following manner (Fig. 2.15): we drew less than one microliter of whole blood in a capillary tube (e.g. a fingerprick volume) (step 1), connected the sample-loaded tube onto a reagent-loaded cassette and inserted into instrument (step 2), started assay with a single button push, and after completing the assay, we displayed the optical density of each meandering zone on the LCD screen (step 3). After prompted to transmit data packet (step 4), we sent both by satellite, which sends an email message with the data as an attachment to a pre- designated address (step 5a), and by SMS, which delivers the message to a pre-designed cellphone number (step 5b).

Figure 2.15. mChip integrates point-of-care diagnostics with data communication. Here, a step- by-step illustration: (1) fingerprick sample collection, (2) metering onto microfluidic cassette and button-push operation of instrument, (3) display of data in less than 20 minutes of optical density in four analysis zones, (4) option to communicate data by satellite (5a), which sends an encoded message to an email address, and/or by GSM/GPRS (5b), which sends an encoded message to a cellular phone. Shown in red inset is the message, decoded and translated from short-burst form to meaningful data values (test ID, patient ID, and OD values of the four detection zones).

We used 0.2 µL of whole blood (pre-collected whole blood in either heparin or EDTA vacutainers and stored in 4 ⁰C for 1 – 7 days prior to use) for each test. Each test, starting from sample metering and ending with data transmission, required less than 20 minutes (Fig. 2.15). The sensitivity and specificity (with 95% confidence intervals) of our POC device in classifying the HIV status for whole blood samples, based on 40 tests, were 100% ± 1.2% and 100% ± 1.7%, respectively (Fig. 2.16A and B). As shown by the protocol for evaluating new diagnostic tests set out by the World Health

Organization, the high test sensitivity and specificity result in tight 95% confidence intervals [62].

The ability to test whole blood samples on-site is advantageous as it avoids issues concerning transportation and storage of whole blood. In fact, most benchtop ELISAs require a sample centrifugation step so that it runs on serum or plasma rather than unprocessed whole blood.

At the end of a test, the mobile device could transmit the test result from any remote setting to a central computer server. All the relevant test data were encoded into a small data packet. The mobile device prompted the user to push the button to begin data transmission using one of the two

54 available transmission modes: satellite or SMS. Via satellite, the data were sent to a pre-designated email address along with the global position coordinates of the test; via SMS, a message with the data was sent to a pre-designated phone number. All results were automatically time-stamped. In our testing, the device successfully sent results either through satellite or SMS for all samples tested

(Fig. 2.16C). The device successfully transmitted 38 of 40 test results on the initial try using GSM

(the failures were due presumably to high traffic in the cell phone network). Also, 33 of 40 test results were successfully sent on the first try using satellite (the failures were due presumably to poor weather conditions such as excessively cloudy sky or an unclear line of sight of the sky). In practice, we used the satellite system as a backup in the event of cell-phone network failure. In contrast to the cell-phone network, the satellite system also allows us to send results from every region of the world.

TRACnet (www.tracnet.rw) is The United States President's Emergency Plan for AIDS Relief

(PEPFAR)-funded mobile health network being developed by the Rwandan Ministry of Health, in partnership with the company Voxiva, to monitor issues of patient adherence to treatment regimens, drug resistance, and availability of HIV drug and lab supplies [79]. TRACnet has been deployed for over 40,000 patients in Rwanda. As a step towards integrating POC devices with a scalable electronic health-record system, we uploaded HIV diagnostic data, collected in Rwanda and transmitted by both satellites and SMS, onto a component of the TRACnet database custom-built for the mChip. The complete test record was displayed within a secure section of the TRACnet electronic patient health records database (Fig. 2.16D).

Figure 2.16. On-the-field performance of POC diagnostics instrument in Rwanda. (A) Vertical scatter plots of silver absorbance (normalized by cutoff values) for whole blood samples positive (filled circles) and negative (open circles) for HIV. ( B) Contingency table and evaluation of test accuracy (n = 40). ( C) Completed (filled bars) and incompleted (checkered bars) transmission of HIV results via satellite and SMS. For each sample tested, we attempted data transmission with both modes (see Methods for further information). (D) HIV results, communicated remotely, are displayed onto TRACnet server, an electronic medical records system used by the Rwandan Ministry of Health.

2.4 Conclusion

The ELISA test serves as a gold standard for diagnosing many diseases, but it can be performed only in a centralized lab. Here, we designed the mChip assay to serve as a miniaturized ELISA in resource-limited settings, as it exhibited performance equal to lab-based immunoassays, with versatility in type of blood sample (whole blood, plasma, and sera) and ability to detect more than one type of marker at once. In particular, the high sensitivities of the mChip for both HIV and syphilis are attractive for screening: in remote settings, potentially infected patients can be diagnosed immediately, and their samples can be confirmed by tests with greater specificity, different antigen

56 preparations, test principles and/or biological targets ( e.g. anti-cardiolipin antibodies for distinguishing active from past or latent syphilis) [24]. The mChip exhibits advantages over the lateral flow assay ( i.e. immunochromatographic tests or “dipsticks”) [20], which are not sufficiently sensitive for detecting many important protein markers [80] and are challenging to multiplex [81]. In addition, the mChip quantifies signals using a handheld instrument, allowing for objective measurements as opposed to current rapid HIV tests which require subjective interpretation of band intensities by the user that can result in false positives in real-world settings [14, 82]. Furthermore, the sensitivity of some commercially available HIV rapid tests has been documented to be lower in sample matrices such as finger-pricked whole blood than for serum, resulting in false-negative tests

[83]. The mChip retains the most attractive features of lateral flow assays, as it can be operated in the field with no external infrastructure or electricity and minimal training, with the individual tests costing pennies in material.

Moreover, integration of individual components into a single device has proven to be one of the most difficult tasks in building lab-on-a-chip systems [3, 84]. Often, components are incompatible with each other; even if integrated, they are not field-deployable due to cost, difficulty of use, or excessive instrumentation [85]. Engineering has enabled the use of traditionally complex technologies in the remote regions of the world. Such technologies include communications (such as cell phones) and diagnostics (such as glucose meters). Here, we also demonstrate that three main functions of benchtop instrumentation - liquid handling, signal detection, and data communication - can be combined into a compact and versatile device for operating microfluidics- based diagnostics. The instrument contains, among other components, a small diaphragm pump, robust and low-cost optics, and satellite transceiver and GSM/GPRS modem. This device is 1000- fold cheaper than benchtop counterparts, can run multiple immunoassays on a 9V battery, is as sensitive as commercial HIV ELISA tests, but completes a test in less than twenty minutes in a hands-free manner. In addition, we communicated all results successfully via satellite and/or SMS

57 to a computer or cell phone. We also transferred all field data to an electronic medical records system currently used by the Rwandan Ministry of Health for patient monitoring and disease surveillance. E-health technologies have shown promise in developing countries [86]. This study demonstrates the feasibility of miniaturizing the complex functions of benchtop diagnostic devices, including all peripheral instrumentation, into a low-cost and portable format compatible with communication technologies used for healthcare in developing countries.

An ultimate goal of this work is to develop a device for infectious-disease screening of pregnant women located in remote areas to prompt early treatment. This test can also be used outside of the antenatal care clinic, as HIV and syphilis combination testing is important in pre-screening of blood donations [8] as well as epidemiological surveillance (as STIs are particularly poorly monitored in developing countries). Also, this test is relevant for use by primary care physicians and outpatient clinics in developed countries, as on-site diagnosis can lead to higher rates of correct treatment and lower rates of unnecessary over-treatment. Finally, the immunoassay can be extended to other clinically useful markers for STI’s such as chlamydia and gonorrhea, for which few adequate POC options exist [66]. Our platforms provided clinical performance similar to other rapid tests when tested at resource-limited settings, however, our tests had been tested using limited number of archived specimens. Reagents used in this part (gold-labeled secondary antibodies and silver reagents) were in solution form and might not be stable at room temperature for long period of time

(> 6 months). Further assay development is needed to improve sensitivity, specificity and shelf-life of our test to match the gold standard or WHO requirement (e.g . sensitivity >99% and specificity

>98%).

Overall, we demonstrate a strategy by which microfluidics and nanoparticles can be fundamentally re-designed within an integrated device to achieve POC diagnosis of clinically relevant infectious diseases to improve clinical care in resource-limited settings by removing

58 subjective visual interpretation of signals (as in lateral flow tests), increasing sensitivity and accuracy to match those of ELISA, and enabling automatic tracking of test results.

Chapter 3 A smartphone accessory for portable immunoassays

3.1 Introduction

3.1.1 Smartphone-based healthcare technology

Smartphones are being adopted at a breathtaking pace, including in developing countries [87,

88]. They offer fast computing as a “low-cost processor”, a friendly user interface, and connectivity to the cloud or external devices, all at falling prices, which would be the best option to replace computers in low-resource settings [89] (Fig. 3.1). The use of smartphones, in particular, expands an implementation of Electronic Medical Record (EMR) and Personal Health Record (PHR) systems as well as increases the integration of new medical devices with smartphone towards point-of-care or consumer oriented technologies. While they are increasingly being adapted for health diagnostics, the most common applications have leveraged individual components and functions

(cameras [90], data communication [91], or data processing [92]) (Fig. 3.2), rather than replicating any complete diagnostic assay performed in clinical laboratories [93, 94].

Figure 3.1. The processor speed comparison between mobile phones and computers developed over time. Taken from [89].

3.1.2 Smartphone-based POC immunoassay

The clinical goal is to improve poor pregnancy outcomes due to mother-to-child transmission of sexually transmitted infections. As such, we sought to build on previous work in miniaturizing diagnostics hardware [93, 96, 97]. Here, we seek to engineer all the capabilities of a bench-top

ELISA instrument into a small accessory or “dongle” that attaches to a smartphone. The dongle runs assays on disposable plastic cassettes with pre-loaded reagents where disease-specific zones will provide an objective read-out (14), similar to an ELISA microplate assay (but with gold nanoparticles and silver ions performing the amplification step instead of enzymes and substrate). Few parameters we considered to engineer a smartphone accessory for POC diagnostic test.

First, we evaluated how the device connects and communicates to a smartphone. Seamless data communication between smartphones and medical devices relies mainly on Bluetooth, Wi-Fi, or radio frequency with trade-offs in terms of cost, bandwidth, power consumption (from an external

61 battery), and reliability [98]. Wireless communication requires power on both medical device and smartphone to control the device and transfer data. Micro USB or Apple 30-pin or lightning connector can offer reliable data transmission and power, however, are different for each phone model. The audio jack is a universal connector for mobile phones and tablets that has remained ubiquitous and unchanged throughout technological advances. Communication between the smartphone and accessory via audio jack can be done using frequency shift keying (FSK) [99].

Second, to minimize power consumption of the device, an electrical pump for fluid flow actuation and control should be eliminated. A controllable pressure-driven flow in microfluidics with a simple syringe setup can be achieved for resource-limited settings as we showed in the previous chapter which involves multiple manual steps. Self-powered disposable chip for bioassays by degas-driven flow have been demonstrated [100], however, this requires more complicated manufacturing and storage and may shorten shelf-life due to vacuum lost over time (package leaks over time).

Integrating a novel power-free mechanical pump can minimize power consumption, material cost and maintenance of the device while maintaining the fluid control capability.

Third, the shelf-life of the assay kit is also crucial for both consumer and global health applications. Various stabilizers have been added to immobilize protein on plastic surface for immunoassay applications to enhance the stability. Lyophilization of protein necessary for assay amplification and detection can lengthen the shelf-life while minimizing the need for cold storage

[101]. Remaining reagent needed for the test is wash buffer which is stable in various environmental conditions.

Recently, some manufacturers (Chembio, SD Bioline, and MedMira) have developed dual

HIV/treponemal-syphilis tests, but these tests rely on lateral-flow or immunofiltration technologies which could limit their performance and ability to add additional tests [102]. For our assay targets, we chose HIV and treponemal syphilis antibody tests from our previous work [103], while detecting

62 anti-cardiolipin antibody as a third target as a non-treponemal syphilis marker. Non-treponemal antibody detection is added to help differentiate past and current syphilis infection (Fig. 3.3). A triplex test with HIV, treponemal syphilis, and non-treponemal syphilis results is currently unavailable but offers the potential advantage of helping to characterize the infection as active or inactive since treponemal syphilis antibody level remains high for life [37, 104], hence, the ability to perform both treponemal and non-treponemal simultaneously at low cost would save time and simplify workflow

[38, 105, 106]. In addition, we added IgM as a secondary antibody for early syphilis detection [37].

Anti -tp IgG

Anti -tp IgM Anti -cardiolipin IgG / IgM

Figure 3.3. Common patterns of serological reactivity in syphilis patients. Taken from [37].

Testing in the field can exhibit markedly different performance from tests run in a laboratory due to variations in local clinical specimens, local environmental conditions (including temperature and humidity), and variations in how the tests are run by users. In the field, sensitivities have been reported to be as low as 82% and specificities as low as 85% for the widely used HIV RDTs [107-

109], 64–96% sensitivity and 97–99% specificity for treponemal syphilis antibody tests (Determine,

SD Bioline, Syphicheck, VisiTect, and Chembio) [110-112], and 85% sensitivity and 96% specificity

[112] for non-treponemal syphilis antibody test (Chembio). Here, we sought to demonstrate the field performance of three point-of-care ELISA tests run simultaneously on a smartphone dongle by local health care workers on fingerprick whole blood collected fresh from patients in a blinded manner where the reference-lab results were unknown until testing of all subjects had concluded.

3.1.3 User acceptability and new technology implementation

Product development and field trials should be carried out in parallel. In the previous chapter, we have shown our simple setup with some training required of end users and a handheld automated device that can also upload data to the cloud to perform ELISA in the field. We proved that the technology worked in the field but in order to assure implementation, the simplicity of the test needs to be improved to limit the potential for error. Development of new technology for POC is important, however, its implementation at POC also requires a tremendous consideration and work. Many interventions provided an excellent performance in research studies fail to translate into meaningful health outcomes due to unsuccessful implementation [113]. Technology-centered design could lead to “human error” or design-induced error as there is a limited effort users can adapt to the new technology. Cost is an obvious distinguishing feature in point-of-care tests intended for developed countries (which demand low-cost) vs. developing countries (which demand extremely low-cost). Due to vastly smaller budgets via public financing, developing countries are limited compared to developed countries. Thus, the cost of the microfluidic device (which includes both the material and the manufacturing process) must be kept extremely low in POC testing in developing countries; the fixed instrument must be portable and cheap, and the disposable must be extremely cheap. All components of the device (including the instrument and disposable) must be robust and rugged under a variety of environmental conditions.

In this chapter, the development process towards user-center design includes simplifying the user interface, miniaturizing processes, reducing power consumption and cost, and making a test more rugged to prevent those foreseen issues and remove frustration of new technologies from users. User feedback is also assessed to provide informative evaluation and lead to effective implementation of new technology.

3.2 Methods

3.2.1 Smartphone dongle

The smartphone dongle accomplishes power-free fluid flow and objective signal read-out with powering and signal transmission solely via audio jack connection.

Power-free fluid flow. In previous studies, we found electrical vacuum pumping to consume significant amounts of power [93]. Here, we designed a power-free mechanically-activated vacuum source. Using a simple mechanics of rubber bulb and one-way valve, the dongle generates a reliable, repeatable vacuum at the time of the assay, while keeping the consumables simple to manufacture.

Audio jack powering and signal transmission. We use the audio jack for both power delivery to the device and signal transmission from the device, as has been shown by Kuo, et. al. [99]. A 19kHz audio signal is sent through the left audio channel and is converted to a 3.0V DC signal and used to power the device. Signals acquire from the device are sent through the microphone line via frequency-shift keying (FSK) to a smartphone (Fig. 3.4).

Signal read-out. In the device each testing zone is sandwiched between an LED and a photodiode with a signal amplifying circuit. The light intensity is measured before (I o) and after (I) silver development, and absorbance value (optical density, or OD) is calculated by:

OD = −log(I ) I

The amount of analytes captured determines the amount of silver development, which in turn correlates with the absorbance value.

Custom printed circuit boards (PCBs) were designed in Altium and printed from PCB Universe.

LEDs and photodiodes were precisely aligned with the cassette slot so testing zones align without manual effort. 1-mm pinholes made of 1-mm thick black Delrin (McMaster-Carr) above each photodiode. The dongle casing was designed in SolidWorks and printed in-house (Objet 24 3D-

Printer, Stratasys). Vacuum chamber was created with a one-way umbrella valve (Minivalve), a rubber bulb from a 140-mL syringe (Becton Dickinson), and a conical spring (Century Spring Corp) inside to aid re-expansion. Silicone rubber o-rings and sheets (McMaster-Carr) were used to connect to outlet and seal the venting port.

User-friendly app interface. We designed a user-friendly app on an iPod Touch to assist in assay operation, power the device to take intensity readings, and demodulate the FSK signals sent by the device. The app provides step-by-step directions with pictures to assist the user through steps necessary to perform the assay. Once the assay has been completed, the results are not displayed until the user presses “show results”. This allows the user to ensure privacy before revealing any sensitive information on the screen. The results page clearly indicates a positive or negative diagnosis for each of the three tests performed. The app will also save the raw data, diagnosis, time and date of testing, and patient ID number for review at a later time. An additional passcode can be required for results viewing if increased security is needed. The option to transmit test results via cloud or short message to a predesignated email address or phone number could be incorporated.

3.2.2 Cassette preparation

Automated liquid dispenser for spotting reagents. Robot-assisted manufacturing (OPKO

Diagnostics) for reproducible and high-throughput cassette preparation (Fig. 3.5) was used to prepare microfluidic cassettes at Columbia University prior shipping to study sites in Rwanda.

Figure 3.5. Robot-assisted manufacturing for cassette preparation.

Protein markers . We used recombinant multi-epitope chimeric antigens (gp41, gp36, and O-

IDR) for an HIV 1/2 (Biolink International) marker (same as chapter 2), a 17-kDa recombinant outer membrane protein TpN17 (Lee Labs) [103, 115] for a treponemal syphilis marker (same as chapter

2), and synthetic cardiolipin prepared from plant source [116] provided by CDC for a non-treponemal syphilis marker. An antibody against cardiolipin is used as a non-treponemal marker for syphilis.

Cardiolipin, a lipoidal material, is released as a result of damage to the host cells because of the active infection and also from the cell surface of Treponema pallidium itself [117]. It indicates active infection as well as a reinfection and is helpful in tracking the effectiveness of treatment (reduction in RPR titers), especially if patients are allergic to penicillin and must take other treatment instead.

An anti-goat IgG antibody, binds to gold-labelled goat anti-human antibody (Life Technologies), was selected as an internal positive control [103]. For an internal negative control (provide background signal), surface was not functionalized with any protein but treated with blocking agent.

Surface modification of microfluidic cassettes. Like many ELISA assays using plastic 96-well plates, we use direct physisorption of antigens onto our plastic microfluidic cassettes [103]: 2 µg/mL of HIV chimeric antigens, 15 µg/mL of TpN17, and 10 µg/mL of anti-goat IgG Ab in bicarbonate buffer solution were spotted on the detection zones using a robotic arm as an automated fluid dispenser (Fig. 3.5). Different surface chemistries were used for cardiolipin functionalization on plastic surface. Cardiolipin was covalently attached to the plastic surface using EDC-Sulfo-NHS reaction, which activates the carboxylate groups on cardiolipin for binding with amine-groups on poly-L-lysine coated plastic. This attachment technique was selected as it yielded the best performance (Fig. 3.6).

Figure 3.6. Schematic of multiplex immunoassay on the dongle device and the biochemical reactions at each step of reagents flowing through the microfluidic channel. Five zones are individually treated with proteins: Stabilcoat for internal negative reference, HIV antigen for capturing anti-HIV antibodies (Ab), treponemal syphilis (TP syph) antigen for capturing anti-treponemal antibodies, cardiolipin for capturing anti-cardiolipin antibody (non-treponemal biomarker), and rabbit anti-goat antibody for capturing gold (Au)-labeled goat antibodies for internal positive reference. The 4th zone for non-treponemal syphilis is coated with Poly-L-Lysine prior the cardiolipin attachment. Whole blood samples are loaded in the antibody holders. Pre-loaded washes on reagent cassette are flowed through the antibody holders to resolubilize lyophilized Au-labeled IgG and IgM antibodies, followed by the silver reagents.

Blocking/stabilizing agents. To preserve the conformation and reactivity of the adsorbed antigens, we have tested and analyzed the performance of three blocking/stabilizing agents: casein

(Thermo Scientific), Sea Block (Thermo Scientific) and StabilCoat Immunoassay Stabilizer®

(Surmodics Inc.,). StabilCoat was found to be most effective in our initial testing (data not shown) and used in all experiment. Cassettes were functionalized with the various protein markers as described above. For this experiment, fluid was dispensed by manual pipetting onto cassette surfaces. The blocking agent was then spotted on each detection zone and incubated for 1 hour at room temperature in a humid chamber. After incubation, the blocking agent solution was aspirated and plastic cassettes were placed in a vacuum desiccator for 20-30 minutes. The vacuum sealed chamber containing the microfluidic cassettes was then placed in a 30°C oven for 4-6 hours of secondary drying. Cassettes were sealed with clear adhesive tape (OPKO Diagnostics) and stored

69 at 4°C until use. An HIV /Syphilis co-infected sera sample as well as a disease negative sera sample was tested on cassettes prepared with each blocking agent.

In evaluating the blocking agents, the optimization goals for each detection zone were as follows: (1) minimize noise on the negative control zone; (2) minimize non-specific binding on target zones for tests with disease-negative samples; (3) maintain optimal conditions for antibody binding on target zones for tests with disease positive samples, and (4) maintain high positive control signals for all tests.

Reagent cassette. Two PBS-T and four water washes as well as Silver reagent A (silver nitrate) and B (hydroquinone as reducing agent) were loaded to the reagent cassette (Fig. 3.7) manually by pipetting and sealed using an adhesive tape (OPKO Diagnostics). Industrial robotic techniques for loading reagents and apply adhesives can be employed for a high-throughput manufacturing.

Figure 3.7. Reagent cassette. The cassette for pre-stored reagents needed for the assay features two main parts for on-board reagent storage: washes and silver reagents.

Gold-labeled IgG/IgM antibodies. Gold-labeled antibody chemistry is similar to our previous work with additional anti-human IgM antibodies for early detection of diseases (see Fig. 3.6 for a

70 schematic of the biochemical reaction). Anti-cardiolipin antibodies are commonly found IgM antibodies, and therefore, the addition of gold-labeled IgM offers enhanced sensitivity.

Lyophilized gold-labeled antibodies. The shelf-life of the assay kit is also crucial for both consumer and global health applications. Gold-labeled antibodies as well as an anti-coagulant were lyophilized onto the antibody holder for long term storage and ease of use for end-user (outsourced to OPKO Diagnostics).

3.2.3 Field trial

Study sites and populations. In order to guide field evaluation of this study, site selection consideration took place between our team, ICAP-Columbia University and the Rwanda Bio-Medical

Center (RBC). Study sites were selected from the highest HIV/syphilis prevalence in the Preventing

Mother-to-Child Transmission (PMTCT) group, based on 2013 routine data from the HIV division at

RBC (unpublished data). Three community-level health centers (HC) in Kigali, Rwanda were selected for this study: 1) Kimironko HC, serving 56,000 people, 2) Biryogo HC, serving 50,000 people, and 3) Gahanga HC, serving 28,000 people.

Study subjects were primarily recruited from patients currently enrolled in PMTCT and Voluntary

Counseling and Testing (VCT) programs in selected sites by health care workers (HCWs) at each site using a recruitment script provided by the study team. Patients who showed interest in the study were referred to the study team (Rwandan study facilitator) for further study information and receive the consent. Patients were enrolled on a first come, first served basis and his/her participation was completely voluntary and did not affect the received services at any health facilities. After informed consent, patients were subjected to two blood draws by trained HCWs: venipuncture for routine

71 laboratory tests as they originally came to the clinic for, and an additional fingerprick (less than 5

µL) specifically for the dongle assay testing.

Sample size. We aimed for a small-scale field trial to provide meaningful results in the field with target end-users; not necessarily for a definitive assessment of sensitivity and specificity. We chose a sample size of ~100, which allowed us to focus on incorporating device testing with clinic flow, getting user feedback, and assessing patient reception.

Smartphone dongle operation training. HCWs who agreed to participate in the study were trained at the beginning of the study at each study site. The training was carried out using a printed info sheet handout to provide background and basic information of the study and hands-on training on how to perform the dongle tests using fingerprick whole blood. The training session was approximately 30 min at each site.

Routine testing in each population groups. HIV testing at each study site was done for all participants using blood plasma from venipuncture blood collection (routine laboratory testing as they originally came to clinic for). In accordance with Rwanda’s HIV testing algorithm (see schematic below), Gold Colloidal HIV 1/2 Ab (KHB Shanghai Kehua Bio-engineering, China) was used as a screening test, followed by Alere Determine HIV-1/2 Ag/Ab (Alere, Japan) as a confirmatory test, and a Uni-Gold Recombigen HIV 1/2 Ab (Trinity Biotech, Ireland) was performed if results of first two rapid tests were discordant. An aliquot of plasma (~ 0.5 mL) from each specimen was transferred to Rwanda National Reference Lab (NRL) for a confirmation test using Vironostika HIV

Uni-Form II Ag/Ab (bioMérieux, The Netherlands). All tests were performed in a blinded manner by local lab technicians. Syphilis testing at each study site was done for only for PMTCT participants with RPR test (Spinreact, Spain). Aliquots of each plasma specimen at NRL were tested using TPHA

(Spinreact, Spain) and RPR tests as reference tests for treponemal-syphilis and non-treponemal syphilis, respectively.

Testing using fingerprick whole blood. At the select sites, local HCWs who agreed to participate in this study were trained to operate dongle by our team on how to perform a fingerstick blood collection and how to operate the dongle using fingerprick whole blood samples. Patients currently enrolled in PMTCT and VCT programs were informed of the study by the HCWs. HCWs assessed eligibility of interested patients and those that were eligible processed the informed consent step.

After consent, these patients were subjected to two blood draws by HCWs: venipuncture for routine laboratory tests as they originally came to the clinic for, and an additional fingerprick for the dongle testing. The fingerprick donation was less than 20 µL.

To perform the test, the user mixed 1 µl of whole-blood sample obtained via conventional finger pricking with 9 µl of diluent (1% BSA - 0.05% Tween-20 in PBS), placed ~2 µl of the mixed sample

73 into the cassette, attached the antibody holder into the cassette, inserted the cassette into the dongle, pressed the bulb fully to initiate vacuum, and pressed “start assay” on the app to start Phase

1 of the test. After 5 minutes, all reagents downstream of the venting port (gold-labeled antibodies, and washes) will have passed through the chip (Fig. 3.9B). For silver development, or Phase 2, the user was then prompted to slide the toggle to close a venting port, to initiate flow and mixing of silver

A and B. The two silver reagents were stable separately and thus stored separately on the reagent cassette; the venting port design allowed mixing of the silver reagents immediately prior to use, minimizing silver auto-catalysis. To prevent exposure of chemicals to the user, sample and reagents were contained in a membrane filter within the cassette, and the antibody holder was securely connected to the cassette. Optical density readings were taken before and after silver development, and at the end of the assay (15 min) results for all markers are available and clearly displayed on the app interface. Step-by-step illustration of dongle testing is shown in Figure 3.8.

Figure 3.8. Step-by-step illustration of dongle testing. ( A) User starts the application and enters patient ID number. ( B) User mixes 1 µl of whole-blood sample with 9 µl of diluent and inserts the cassette into the dongle. ( C) User presses the bulb fully to initiate vacuum, places 2 µl of the mixed sample into the cassette and attached the antibody holder into the cassette. ( D) The application displays step-by-step instructions for user to follow, as well as remaining assay time. (E) After completing the assay, user selects “View Results” to display test results. Another passcode could be added at this step for an extra security and privacy. ( F) Screen displays results for each disease marker. The app can be set to show absorbance values or “positive” and “negative” diagnoses, in the study we displayed absorbance values. User clicks “Finish” button to prompt back to the first screen for the next test.

Testing using venipuncture whole blood. Each patient venipuncture whole blood was tested using the dongle by our team in parallel with the fingerprick blood testing. Whole blood was processed in the same manner as the fingerprick whole blood.

Privacy and confidentiality. All data collected during interviews were labeled with a sequential unique study ID number by an interviewer. Names or other individually identifiable information were not record on any data collection document. The written consent forms, containing the subject name and signature were the only document with a patient identifier but there was no link to the data collected in the interview.

Statistics. Averages, standard deviations, linear fit, and two-sided Student’s t-tests (α = 0.05) were calculated with Microsoft Excel. Student’s t-test was chosen to compare two small sets of quantitative data when data in each sample set were related. Vertical scatter plots, sensitivity, specificity, 95% confidence intervals, ROC curves, and McNemar’s test were created in GraphPad

Prism.

3.3 Results and Discussion

3.3.1 Dongle design

The smartphone dongle accomplishes power-free fluid flow and objective signal read-out with powering and signal transmission solely via audio jack connection. The dongle (Fig. 3.9A) consists of two main innovations to achieve low power consumption: 1) power-free fluid flow and 2) audio jack powering and signal transmission.

Figure 3.9. A smartphone dongle. (A) An image of the dongle with a microfluidic cassette connected to an iPod touch. ( B) A two-layer disposable cassette with antibody holder. The reagent cassette (top layer) contains pre-stored reagents (washes (yellow), silver nitrate (blue), and reducing agent (green)) and the test cassette (bottom layer) contains five detection zones. Reagents are numbered in the order they flow through the test cassette. The power-free vacuum chamber connects to the cassette outlet, drawing fluids from the inlet towards the waste pad.

Power-free fluid flow. In previous studies, we found electrical vacuum pumping to consume significant amounts of power [118]. Here, we eliminated the power-consuming electrical pump by using a “one-push vacuum”, where a user mechanically activates a negative-pressure chamber (Fig.

3.10A) to move a sequence of reagents pre-stored on a cassette (Fig. 3.9B).

Figure 3.10. Mechanism of the smartphone dongle. (A) Schematic diagram of dongle highlighting a power-free vacuum generator. Sub-figure shows vacuum activation. (B) The audio jack connector on the dongle is used for audio-based powering and FSK data transmission to a smartphone.

The simple vacuum chamber is created with a rubber bulb, with one port connected to the assay cassette outlet, and one port to a silicone one-way valve (Fig. 3.10A). When the bulb is depressed, air exits out the one-way valve, and a spring aids the bulb in re-expansion, creating a negative

78 pressure within the chamber that pulls liquids through the channels. Previous methods of power- free microfluidic flow have primarily been achieved with degassing the microfluidic chip and sealing under vacuum [100, 119], however, these methods can add significant complexity to manufacturing, packaging, and storing. By contrast, our setup mechanically generates the vacuum source at the time of assay; this procedure is durable (similar to a bulb for a manual sphygmomanometer), requires little user training, and does not require maintenance and additional manufacturing to pre- package a vacuum source (which can also leak over time).

Each test included a) microfluidic cassette with reagent holder on top and b) sample collector with lyophilized gold (Fig. 3.9B). Once the sample collector was attached, the microfluidic cassette and reagent holder were connected. When the microfluidic chip is fully inserted into the dongle, the microfluidic output connects with the vacuum chamber sealed with a rubber O-ring. The user would initiate flow by pressing the bulb. This movement evacuated air from the vacuum chamber via the one-way valve, and as the bulb re-expanded, negative pressure was generated in the vacuum chamber to enable flow of fluid (Fig. 3.10A). After the sample and washes passed through (~6 min), the user was prompted to slide the toggle to seal the venting port. The user is prompted to re-initiate the vacuum by pressing the bulb, which initiated movement and mixing of the two silver reagents and flow over the testing zones (see Fig 3.8 for step-by-step illustration.) The mechanically-activated vacuum showed no significant difference in flow time between three different users, two of which were not device developers (Fig. 3.11).

The dongle measures the optical density (absorbance) of silver enhancement on each assay as described previously. We designed the dongle so that power is only consumed during optical density readings (8.5 mW) and information transfer back to the smartphone (1.5 mW) (Fig. 3.12); no power is consumed by the dongle while sample and washes are flowing.

Audio jack powering and signal transmission. Other components in the dongle, including robust and low-cost LEDs, photodetectors [93, 103], and a microcontroller, consume very low power. We use the audio jack for both power delivery to our device and signal transmission from our device

(Fig. 3.10B), as has been shown by Kuo, et. al [120]. Others have used a similar technique for heart rate monitoring [121], measuring breath carbon monoxide in smokers [122], electrocardiogram acquisition [123], and electrochemical sensing of glucose and other analytes [124]. A 19 kHz audio signal is sent through the left audio channel and is converted to a 3.0V DC signal and used to power the device (Fig. 3.10B). The device has no internal battery and uses only the power delivered by the audio signal. The dongle measures the optical density (absorbance) of silver enhancement on each assay as described previously. After the optical density at each zone is measured, the microcontroller, sends the information through the microphone line via frequency-shift keying (FSK).

The decimal integer is converted into binary signal and each digit is sent as either 1600 Hz (1) or

800 Hz (0) at a rate of 1 bps. Signal transmission rates can be sped up in the future, and our initial implementation was focused on fidelity of signal. To test the accuracy of the signal, we programmed the microcontroller to send a pattern of alternating “1” and “0”, or high frequency and low frequency signals. We observed 100% accuracy of 12160 bits tested. The circuit diagram is shown in Fig. 3.13.

We designed the dongle so that power is only consumed during optical density readings (8.5 mW) and information transfer back to the smartphone (1.5 mW) (Fig. 3.11). No power is consumed by the dongle while sample and washes are flowing. We found an average power consumption of 8.54 mW while taking intensity readings (8 sec) and an average of 1.49 mW while transmitting information and latent time during silver development (8 min). The total power consumption by the device was

0.22 mWh per test. Over the course of the 15-minute assay, the dongle has an average power consumption of 1.6 mW. By comparison, an iPhone 4 uses 751 mW on a 3G network and 17.5 mW on a standby mode [93]. While using an iPod (4 th Gen), we found that each run consumed about

2.4% of the battery, and therefore could run about 41 runs in a single charge. Newer devices with fuller battery capacity or devices with a larger battery capacity (iPod 4th Gen holds 930 mWh) would

81 be able to run even more tests on a full battery charge. Using commercially available electronic components with an injection-moldable case, our device would have a manufacturing cost of $34

USD, in comparison to $18,450 USD of a typical ELISA equipment.

The power harvested from the audio jack was stable and sufficient for reliable optical density

(OD) measurements compared with a benchtop analyzer (Fig. 3.14A). In order to compare the OD

82 readings of the iPhone powered dongle to the benchtop analyzer, an HIV seropositive whole blood sample was tested ( n=3) on the assay platform, with light intensity measurements taken by both devices for direct, head-to-head comparison. Optical density values were calculated on three detection zones for analysis: negative control, R17 target, and HIV target to show a range of signal readings. For both target zones, there was no significant difference in OD measurements taken by the dongle and benchtop analyzer. The negative-control zone showed a small but significant difference ( p = 0.03). Importantly, the dongle produced optical-density readings that correlated to a serial dilution of a strongly RPR positive (1:128 titer) syphilis sample (R 2 = 98.9%, Fig. 3.14B).

Figure 3.14. Optical densities readout using a smartphone dongle. (A) An HIV-positive whole- blood sample was run in triplicate, where measurements were taken with the iPod audio jack– powered dongle and the benchtop analyzer (OPKO Diagnostics, without the temperature-control system, in order to mimic the dongle) reading the same cassette. Data are averages ± 1 SD ( n = 3). n.s., not significant, two-tailed Student’s t-test. (B) Non-trepnonemal syphilis quantitative assay by detecting anti-cardiolipin antibodies. Serial dilution of RPR positive (1:128) serum to mimic lower RPR titers (1:1 to 1:64) as conventional RPR quantification. Data are averages ±1 SD ( n=3) and plotted with a linear regression fit and correlation.

We designed a user-friendly app on an iPod Touch to assist in assay operation, power the device to take intensity readings, and demodulate the FSK signals sent by the device (as described above). The app provides step-by-step directions with pictures to assist the user through steps necessary to perform the assay. Once the assay has been completed, another level of security ( e.g. password required) can be added to allow only authorized users to view the results to ensure 83 patient’s privacy. The results page clearly indicates a positive or negative diagnosis for each of the three tests performed. The app will also save the raw data, diagnosis, time and date of testing, and patient ID number for review at a later time. An additional passcode can be required for results viewing if increased security is needed. While our app was coded in iOS, it can easily be adapted to the Android platform.

3.3.2 Multiplex immunoassay

Further, in terms of assay chemistry and format, we built on our previous work for HIV/syphilis detection [103] in five significant ways. First, we expanded the detection zones in the disposable microfluidic cassette to five zones, for detecting HIV, treponemal syphilis, and non-treponemal syphilis antibodies simultaneously with internal negative and positive control (Fig. 3.6 and 3.9B).

Non-treponemal syphilis marker was added as our third disease marker due to its clinical significance as differentiation between active and past infection is particularly valuable in endemic areas (see also section 3.1.2).

Second, gold-labeled IgM antibodies were added to the assay (Fig. 3.15). The biochemistry of our immunoassay is similar to our previous work (chapter 2) with an additional anti-human IgM antibodies as detection antibodies for early detection of diseases (Fig.3.6). Anti-cardiolipin antibodies are commonly found in IgM antibodies, and therefore, the addition of gold-labeled IgM offers enhanced sensitivity of non-treponemal syphilis.

Figure 3.15. An additional of gold-labeled anti-human IgM for syphilis detection. Comparison of signal measurements obtained by additional of gold-labeled anti-hIgM to gold-labeled anti-hIgG and gold-labeled anti-hIgG alone as detection antibodies for negative, weak positive non-treponemal syphilis (RPR titer 1:2), and strong positive non-treponemal syphilis (RPR titer 1:32) plasma samples. Data are averages ±1 SD ( n=4 for anti-hIgG and n=3 for anti-hIgG:anti-hIgM).

Third, to improve long term stability in shipping and storage, we lyophilized gold-conjugated secondary antibodies inside the antibody holder (Fig. 3.9B), along with a stabilizer and anticoagulant and packed the holder in an individual moisture barrier bag prior shipping them to Rwanda. Previous studies have also shown that lyophilization enabled the secondary antibodies to be stable for over

6 months at room temperature [101, 103], and showed comparable performance as gold-conjugated antibodies freshly diluted in buffer from a refrigerated stock solution (Fig. 3.16).

Figure 3.16. Comparison of signal from gold-labeled anti-hIgG/anti-hIgM antibodies lyophilized in a plastic antibody holder and freshly prepared in solution. Detection zones were functionalized with human IgG, human IgM and rabbit anti-goat antibodies (positive ctrl). Data are averages ±1 SD (n=3). n.s. not significant, Student’s t-test. Comparison of signal from gold-labeled anti-hIgG/anti- hIgM antibodies lyophilized in a plastic antibody holder and freshly prepared in solution.

Fourth, to further mimic real testing conditions, we prepared the test cassettes ahead of time at

Columbia University before transporting them to the testing sites in Rwanda. By using stabilizing agent during physisorption of capture proteins, we found the protein to retain its original function over 3 weeks at 60°C (Fig. 3.18) (equivalent to roughly 28 weeks at 25°C according to Arrhenius- like approximations [125]). Further stability experiment should be carried out at room temperature with a control humidity packaging for long-term stability of prepared cassettes.

Fifth, the wash buffers and silver reagents were pre-loaded on the reagent cassette (Fig. 3.7 and 3.9B) each day prior to testing. In the future, robot-assisted loading of reagent cassettes will be implemented as reagents are known to be stable for over 6 months at room temperature [103] and wash plugs have been shown to stay separated after airborne shipping [126]. These conditions replicated real shipping and transportation conditions, minimized user steps, and increased field- readiness to enable a “plug-and-play” operation for the user.

3.3.3 Field testing with target end-users

The dongle app presented a user-friendly interface to aid the user through each test, step-by- step pictorial directions, built-in timers to alert the user to next steps, and records of test results for

86 later review (Fig. 3.8). Given the simplicity of running the test, training of HCWs took approximately

30 minutes via step-by-step instructions included in the custom app. Five HCWs (lab technicians with no experience in ELISAs) tested fresh fingerprick whole blood from 96 patients, whose disease statuses were unknown until the reference-lab tests were performed at the end of the study.

The testing procedure was run as follows. At three health centers (HC) in Kigali the capital of

Rwanda, HCWs recruited patient volunteers enrolled in PMTCT and VCT programs for the study, with our team providing further information about the study to participants as needed. Consent forms were translated to Kinyarwanda (the principal local language), and obtained by a third-party translator fluent in English and Kinyarwanda. Details of the recruitment and testing process are discussed in section 3.2.3. Fingerprick whole blood specimens were collected and coded with a study ID number with no link to access other health information to protect their privacy. For reference tests, HIV RDTs were completed at each site, while HIV ELISA, syphilis Treponema pallidum hemaglutination (TPHA), and syphilis RPR reference tests were performed at Rwanda National

Reference Laboratory using plasma.

For this batch of pre-shipped cassettes, a 10-fold dilution of fingerprick whole blood was used, and we incorporated this dilution step into the testing procedure. Subsequent to the field trial, we found that cassettes with enhanced concentration of coated proteins can take neat whole blood and produce accurate results (data not shown). Cut-off values to determine if samples are reactive or non-reactive for each marker were selected by using receiver-operating characteristic (ROC) curves. While a final product will offer preset cut-off values, in this development work, we identified cut-off values retrospective to data collection that maximize sensitivity (minimize false negatives) since our test is targeted towards screening applications. Cut-off values for internal negative and positive controls were also applied to verify validity of test results; no tests were excluded based on these criteria. An indeterminate range (e.g. if OD is within±10-20% of cut-off) [127] can be implemented for future tests, and indicate the need to rerun the test. The test results for detection

87 of each marker are compared with the gold standards of lab-based HIV ELISA, TPHA, and RPR, and are presented in terms of signal-to-cutoff of each target relative to its reference test displayed as vertical scatter plots and ROC curves.

From fingerprick whole blood performed by local HCWs, the detection of HIV antibodies had a sensitivity of 100% (95% CI: 59–100%) and specificity of 87% (95% CI: 78–93%). Sensitivity for detection of treponemal antibodies was 92% (95% CI: 64–100%) with specificity of 92% (95% CI:

83–97%). Sensitivity for detection of anti-cardiolipin antibodies was 100% (95% CI: 48–100%) with specificity of 79% (95% CI: 69–87%). ROC curve shows AUC of 0.96 for HIV, 0.90 for treponemal syphilis, and 0.92 for non-treponemal syphilis (Fig.3.18).

Figure 3.18. Testing of dongle using clinical fingerprick whole-blood specimens in the field by third party running the test. ( Left ) Vertical scatter plot of the dongle device signal-to-cutoff ratios for HIV, treponemal syphilis, and non-treponemal syphilis positive (Pos) and negative (Neg) samples using fingerprick whole blood compared to gold standard tests (HIV ELISA, TPHA, and RPR). ( Right ) A receiver operating characteristic (ROC) curve for each disease marker.

On venipuncture whole blood, the dongle yielded a sensitivity and specificity of 100% (CI, 59.0

– 100), 91% (CI, 83.0 – 96.0) for HIV, 77% (CI, 46.2– 95.0), 89% (CI, 80.4 – 95.0) for treponemal syphilis, and 80% (CI, 28.4 – 99.5), 82% (CI, 73.0 – 89.6) for non-treponemal syphilis (Fig. 3.19).

There is no statistical difference of dongle assay performance on fingerprick and venous whole blood with p-values of 0.45, 1.0, and 0.33 for HIV, treponemal syphilis, and non-treponemal syphilis, respectively using McNemar’s test.

Figure 3.19. Testing of dongle using clinical venipuncture whole-blood specimens in the field by our team. ( Left ) Vertical scatter plot of the dongle device signal-to-cutoff ratios for HIV, treponemal syphilis, and non-treponemal syphilis positive (Pos) and negative (Neg) samples using fingerprick whole blood compared to gold standard tests (HIV ELISA, TPHA, and RPR). ( Right ) A receiver operating characteristic (ROC) curve for each disease marker

The performances of our tests were comparable to those of current commercial RDTs on whole blood (fingerprick or venipuncture) performed by trained staff in regular clinical settings: 1) HIV antibody tests (Clearview COMPLETE, Clearview STAT-PAK, OraQuick Advance, and Uni-Gold

Recombigen): 97 – 100% sensitivity and ~100% specificity [108], 2) Determine HIV-1/2 showed lower specificity (85.2%) in one field evaluation [109], 3) Treponemal syphilis antibody tests

(Determine, SD Bioline, Syphicheck, VisiTect, and Chembio): 64 – 96% sensitivity and 97 – 99% specificity [110-112], and 4) Non-treponemal syphilis antibody test (Chembio): 85% sensitivity and

96% specificity [112].

12 out of 96 samples tested for HIV were false positive from our test. Although, high specificity of the test is preferable in high disease prevalence regions, we chose the cut-off values retrospective to data collection to maximize sensitivity (minimize false negatives) since our test is targeted towards screening applications. A confirmation test is required prior to starting treatment. Comparing assay performance from this chapter to the previous chapter, this current work showed lower specificity.

Few assay parameters were changed from the previous chapter which may lead to the higher amount of nonspecific binding of antibodies on surface and/or lower the signal of target antibodies;

1) no washes after the sample, 2) smaller surface area for capture antigens, 3) lyophilized gold- labeled antibodies, 4) an additional of gold-labeled anti-hIgM, 5) new blocking agent, and 6) no anticoagulant used in fingerprick whole blood. Another parameter which could lead to a false positive is patient’s health and clinical conditions, for example, pregnancy, and flu vaccination. We asked participants for their medical history (self-reported) including pregnancy status, previous HIV/syphilis infections, risk behaviors, recent vaccinations and current infections and illness. However, we could not find any correlation between false results and patient health status (data not shown). A larger samples size is needed to draw a significant conclusion.

With the test performance below a requirement of WHO for HIV (>99% sensitivity and >98% specificity), our test could be used as an initial screening test which requires a confirmation test(s).

Further improvement has been ongoing to improve the clinical performance (see Chapter 5 for more discussion). For syphilis test, since the treatment ( e.g. penicillin) is cheap and less toxic, treatment can be carried out without a confirmation test in pregnant women to prevent still births.

3.3.4 Feedback from participants enrolled in the study

To assess user feedback, a third-party interviewer fluent in English and Kinyarwanda conducted the surveys to reduce any biases from the study team. The questions posed to the participants focused on dongle/fingerprick preference versus traditional venipuncture (and why), desired test time, and whether the dongle would be recommended (and why) (Fig. 3.20). A vast majority of patients (97%) would recommend the dongle to others because of the fast turnaround time (57%), potential to offer results for multiple diseases (44%), and simplicity of procedure (29%). Fingerprick blood collection was preferred to conventional venipuncture by 95% of patients because: it is less painful (98%), takes shorter time (60%), the HCW had trouble with the needle collection (55%), the patient is scared of a needle (43%), and a fingerprick takes less volume of blood (42%). However, 90

2% of patients preferred venipuncture because they trust the result more with venous blood. HCWs appreciated the lack of user interpretation to read the result or external power to operate, and also noted in surveys that the dongle could be useful in low-volume testing sites (e.g. VCT or mobile visit) or serve as a backup test for high-volume patients’ clinics in case of power outage, which we experienced during testing.

Figure 3.20. Satisfaction survey from participants in this study.

3.4 Conclusion

This dongle presents new capabilities for users ranging from health care providers to consumers. First, for HCWs, the dongle enables an ELISA-quality non-treponemal syphilis test to be performed at the point of care. The current procedure in Rwanda (and many developing countries) calls for single lab-based qualitative RPR test, such that all patients with positive non- treponemal results are recommended for treatment. However, this can lead to overtreatment, given the intrinsic lower specificity of non-treponemal assays [38, 104]. A number of countries (including

Brazil, China, Peru, Tanzania, Uganda, and Zambia) are adopting treponemal-specific RDTs [128].

However, blanket treatment of patients testing positive with a treponemal-specific RDT can lead to overtreatment and penicillin resistance, since treponemal antibodies remain even after infection has cleared. A dual-syphilis dongle would empower HCWs to follow guidelines [38, 104] which

91 recommend treatment only if both tests are positive or in the case of treponemal-specific positive results with no previous infection or symptoms or history consistent with new infection (such as new rash or unprotected sex with infected partner). Differentiation between active from past infection is particularly valuable in endemic areas, including Rwanda [129]. Among our subject group, 13 patients exhibited positive TPHA results, and would all be treated under guidelines with rapid treponemal-specific tests, even though only four exhibited positive RPR results. Note that our field trial assay only provided qualitative results for anti-cardiolipin marker.

Second, our test detects IgM in addition to IgG antibodies, thereby enabling early detection of syphilis, as anti-treponemal IgM antibodies appear earlier than anti-treponemal IgG by two weeks [37].

Our platform has the flexibility to expand by using labeled IgG/IgM for all disease markers compared to tests that use labeled antigens specific to each disease. Together, the addition of a non-treponemal test and detection of IgM – performed alongside HIV which patients are accustomed to testing – moves another step towards improved access to a complete POC multiplex antenatal-care panel.

Performing three individual commercially-available tests can cost up to $8.50 USD ($0.80 – $5 for

HIV RDT [130], $1 – $3 for treponemal RDT [104], and $0.50 for RPR [131]). Material and reagents cost per test for our triplex test is $1.44 USD, leaving room for a significantly lower anticipated market price. Further, our diagnostic only requires 2 µL for all three tests, where HIV and syphilis

RDTs using whole blood require anywhere from 5 – 60 µL each. It is critical for true POC tests that they work on easy-to-collect clinical samples; as was demonstrated with this study’s use of fingerprick samples rather than venous blood. Previous health-impact modeling has shown that a new syphilis test with only 70–80% sensitivity and specificity, but one that can be performed truly at the POC, can reduce deaths by 10-fold over a hypothetically perfectly accurate lab-based test [132] by increasing detection of infections (high sensitivity is most important for large-scale screening.)

Compared to syndromic management [37], POC syphilis tests also help HCWs avoid overtreatment and improve cost-effectiveness.

Third, for HCWs, a quantitative non-treponemal measurement can be valuable for monitoring disease progression. For example, following treatment, a four-fold reduction in antibody titer is an indicator of successful treatment [38]. Our dongle showed strong correlation of optical density against a serial dilution of a strongly RPR positive syphilis sample. While several commercial readers are available to measure results of a RDTs flow strip and provide quantitative readout, it can be difficult to reliably adjust for positioning, illumination accuracy, and dynamic range and hence influence accurate quantitation [18]. Nevertheless, adoption of a smartphone-based test must present clear advantages over low-cost RDTs that exhibit high sensitivities and specificities on single disease targets. While our smartphone-based diagnostic test involves more instrumentation than RDTs that require no instrument, it offers automation of assay, objective readout of signal, and quantitation. It also offers multiplexed detection (in this case, avoiding multiple finger pricks, a point appreciated by the patients in the survey), and can also ultimately reduce cost compared to purchasing multiple RDTs. Precise liquid handling and metering are challenging when an untrained user executes the test, which can result in test errors. By contrast, our system contains precise injection molded cassettes, pre-loaded reagents, highly optimized optics, and exact alignment that can offer a rapid and sensitive quantitation while reducing user variability.

In a context of rapidly increasing consumer adoption of health monitoring devices, this dongle also takes a step towards coupling microfluidics with advances in consumer electronics. The hardware of the dongle exhibits characteristics similar to familiar consumer electronics devices through a number of technical innovations: low power (using a power-free, continuous-flow vacuum and requiring no power), durable components (using LEDs and photodetectors), portability and low cost (less than a pair of headphones). The disposable cassettes can be robustly manufactured and pre-coated with proteins with stable reagents before shipping to the use location (with an automated fluid dispenser that can spot and block 12 cassettes in two minutes). These hardware specifications suggest that new consumer-oriented medical devices are on the precipice of moving beyond

93 glucose monitoring, vital signs, and wellness into clinical diagnostics. This trend is expected to accelerate aided by advances in low-cost on-demand fabrication and cloud computing.

Finally, in terms of running a microfluidics device at the POC, we have taken measures to significantly simplify the user experience, a paramount consideration in the adoption of new devices.

The most important feature behind a “plug-and-play” experience is the reduction of manual steps, enabled by integrating advances in fluidics, mechanical, optical, and electronic components. In addition, we built a touch-activated pictorial software interface that allowed for training in 30 minutes, a step-by-step software guide on the smartphone contemporaneous with assay operation, and reporting of results (yes/no or quantitative titers) without user interpretation. Indeed, HCWs commented that result bands on RDTs can be faint or unclear (necessitating blood specimens to be sent to a reference lab and delaying results for days) and on limitations of the lab-based non- treponemal RPR test (which also could not be performed during power outages). Patients also had a positive experience with the dongle, citing the 15-minute turnaround time as the biggest benefit.

Currently, more than half of patients have to wait for results without knowing how long it takes, while a third have to wait more than 2.5 hours to receive their results (informal interviewed with patients).

Although our current test performance may not be up to the standard approval of in vitro diagnostic test, in the future, this type of smartphone dongle approach can be further integrated with electronic medical records and personal health records, with several layers of security built to protect patient identity. Overall, to respond to the challenge of rapid, accessible and easy-to-use clinical diagnostic tests, we developed and integrated numerous technological innovations to miniaturize multiple ELISA-quality tests into a power-free microfluidic dongle, and demonstrated its ability to run on a smartphone for simultaneous multiplex detection of infectious diseases from a drop of fingerprick whole blood.

Chapter 4 Development of a low-cost and point-of-care sputum processor for a downstream nucleic acid amplification test

4.1 Introduction

Although immunoassay is a powerful diagnostic test, some disease diagnoses cannot be done using protein markers, for example, tuberculosis. Current tuberculosis serological test has been shown to be neither accurate nor cost-effective [133, 134]. A major public health threat, tuberculosis is extremely widespread in developing countries [135]. An estimated 9 million peopled developed tuberculosis and 1.5 million people died from the disease in 2013 [136]. The more recent emergence of multidrug-resistant tuberculosis strains (MDR-TB) and rising rates of HIV/TB co-infection have led to significant new healthcare challenges. An estimated one third of people living with HIV in

African region is co-infected with tuberculosis with rates reaching over 50% in parts of southern

African countries (Fig. 4.1) [136].

Figure 4.1. Estimated HIV prevalence (%) in new and relapse TB cases, 2013 [136].

Majority of new tuberculosis cases occur in low- and middle-income countries (Fig.4.2). Drug- resistant strains can be transmitted the same way as regular tuberculosis. Patients can have multiple infection strains. Therefore, rapid identification of MDR- / XDR-TB patients for immediate respiratory isolation and treatment optimization are required [137]. Tuberculosis is also a leading killer of HIV infected patients.

Figure 4.2. Estimated TB incidence rates, 2013 [136].

Conventional techniques, for example, smear microscopy, growth detection in culture, and drug susceptibility testing (DST) require experienced lab technicians and extended incubation time

(culture and DST), and some specimens are very difficult to grow because of low abundance and/or lack of viability due to quality of specimen collection, storage, and transportation [138]. The use of nucleic acid amplification (NAA) tests to detect DNA extracted directly from clinical specimens facilitates the identification of these pathogens and reduces the overall turnaround time by 2 to 6 weeks compared to conventional growth detection. It also eliminates the need of biosafety-level 3 facility for culture. CDC recommends that “NAA testing be performed on at least one respiratory specimen from each patient with signs and symptoms of pulmonary TB for whom a diagnosis of TB

96 is being considered but has not yet been established, and for whom the test result would alter case management or TB control activities, such as contact investigations” [139]. FDA-approved, commercially available NAA tests, Cobas Amplicor MTB (Roche Diagnostics) and Amplified

Mycobacterium Tuberculosis Direct Test (MTD; Gen-Probe) have been used for MTB detection based on polymerase chain reaction (PCR) and transcription-mediated amplification (TMA), respectively.

Drug resistance detection in addition to MTB detection is also needed. Even though lab-based gold standard is a culture based (growth detection), molecular assay offers fast turnaround time.

WHO has endorsed GeneXpert MTB/RIF (Cepheid) [140] – a real-time PCR system for M. tuberculosis and rifampicin resistance detection. Around 10% of patients susceptible to RIF will have

INH resistance which cannot be detect by GeneXpert [136]. Each MTB/RIF test machine currently costs US$17,000–$62,000. More importantly, each disposable test-cartridge costs US$10 – $120

[141, 142]. Even though GeneXpert is a very promising molecular diagnostic test, its clinical impact is limited if it cannot reach patients who need it. It is cost-effective but not affordable in many countries/settings. Xpert requires stable and uninterrupted power and restricted operating temperature range to complete each test [138]. Early implementation in nine countries showed more than 25% of failed tests were due to the power failure during the test [143]. Computer interface was one of the challenging in training and implementation [143]. An automated yet simple molecular diagnostic test for TB is still needed for use in peripheral laboratory facilities in resource-limited settings. A hand-held battery-powered device which can detect multiple drug resistance mutations could greatly advantage global health. If an improved rapid NAAT is adopted worldwide, it could help avert more than 15 million tuberculosis-related deaths by 2050 [144].

In order to perform nucleic acid testing on a microfluidic chip, the main functionalities that have to be integrated on-chip include sample preparation, nucleic acid amplification and the detection of the amplified product to avoid contamination issues which is critical in NAAT laboratories, reduce

97 worker steps, and deliver rapid results. While these steps can be easily integrated together and performed in a general laboratory, miniaturization of such disparate processes onto a single microfluidic chip is still a topic of intense research [46, 145, 146]. A comparison of conventional approach into microfluidic approach is summarized in Table 4.1.

Table 4.1. Comparison of conventional microbiological approaches with microfluidic approach for nucleic acid detection (adapted from [147]).

Nucleic Acid Detection Conventional Microfluidic Technology Steps Techniques - Magnetic capture - Micromixers Microbiological cell - Dielectrophoresis - Functionalized magnetic beads isolation - Mechanical - Dielectrophoresis chips filtration - Microfilters - Chemical lysis - Micromixers - Mechanical lysis - Minisonicators Cell lysis - Thermal lysis - Microfilters - Electroporation - Microelectrode-based electrical lyses - Silica-based resins - On-chip channels of cellulose, silica beads, and Nucleic acid extraction - Glass matrix silica resins and purification - Capture - Silicon dioxide pillar arrays membranes - Photoactivated PC surfaces Polymerase chain - Microfabricated reaction chambers Nucleic acid reaction - Flow-through PCR chips amplification - Resistive heaters - Convection-driven PCR chips - Optical On-chip detection using fluorescence, SPR, SERS, Product detection - Electrochemical nanoparticles, electrochemical methods, QCM, and - Mechanical microcantilevers

Sample preparation. After biological samples are collected, cell isolation and lysis followed by nucleic acid extraction, purification and pre-concentration may be performed [148]. This “sample processing” or “sample preparation” step has been less developed than other assay steps because of its intrinsic complexity. Moreover, contamination, inhibitors for subsequent amplification steps and

98 nucleic acid degradation are also critical and influence diagnostic testing as these factors impede quantitative assessment of the analyte in question, leading to misinterpretation of results [146, 147].

Therefore, sample preparation tends to be performed off-chip (using laboratory equipment such as centrifuge), while amplification and detection can be accomplished in microfluidic systems [149].

Integration of sample pre-treatment with analysis could lead to improvements in sensitivity (as less sample is lost in between steps) and convenience [11].

Signal Amplification. Since the amount of nucleic acid acquired from either the preparation step or from the raw sample is usually low for immediate identification and quantification [150], a method of amplification is needed to obtain a sufficiently strong nucleic acid detection signal. The most common technique is the polymerase chain reaction (PCR), for which miniaturization promotes the ability to reduce the reagent consumption, reduce the cycle time, and automate the process [11].

Miniaturization of PCR has many advantages, such as decreased cost of fabrication and operation, decreased reaction time for DNA amplification, reduced cross-talk of the PCR reaction, and ability to perform large numbers of parallel amplification analyses on a single PCR microfluidic chip. Also, microfluidics allows for increased portability and integration of the PCR device.

One of the first silicon-based stationary PCR chip was described several years after the introduction of PCR itself [151]. Since then, many research groups began to develop microchip- based PCR devices. Most of these devices are based on silicon and glass, but more recently, polymer materials such as PDMS [152], PMMA, polycarbonate, SU-8, polyimide, poly(cyclic olefin) and epoxy are being used.

The three main design concepts for PCR microfluidics are a chamber stationary PCR, flow- through PCR and thermal convection-driven PCR [151]. For a stationary design, a droplet-based microfluidic PCR, which is performed similarly to conventional PCR methods, has been demonstrated [153]. In this design, the PCR fluid is kept stationary while the temperature of the

99 reaction chamber is cycled between three different temperatures (melting, annealing and extension phases). The authors demonstrated the use of a microfluidic device to perform PCR on aqueous- in-oil droplets with volumes in the nanoliter range for potential use at the point of care. Flow-through

PCR microfluidics allow for more flexibility in changing reaction rates and times due to a time-space conversion concept. For example, IMM (Institut für Mikrotechnik Mainz GmbH, Mainz, Germany) is developing a micropump-based system that uses a ferrofluidic actuator and magnets to move magnetic fluids throughout the microchip to react with specific reagents and access different temperature zones [154]. Therefore, the duration of each cycle can be controlled by the fluid velocity.

This design allows for rapid heat transfer and thermal cycling and the run-time for such assays is in the order of minutes. HandyLab (HandyLab, Inc., Michigan, USA. HandyLab is now part of Becton

Dickinson) has also developed a disposable microfluidic chip that implements heat and pressure gradients to move microliter-sized plugs via valves and gates through different temperature zones within the chip [99, 154]. Flow-through designs also decrease the possibility of cross-contamination between samples as well as allow for the incorporation of many other functions, which is appealing as it leads towards the development of a micro total analysis system. Thermal convection-driven

PCR microfluidics involves driving the reaction solutions through two temperature zones. This concept is similar to the flow-through design except that the driving force of the sample is buoyancy.

In this manner, a temperature gradient is generated across a reaction vessel which drives thermal convection to circulate the solution between the hot and cold regions for PCR [155]. As PCR is a temperature-controlled, enzyme catalyzed biochemical reaction system, the method in which the different temperature zones are generated and maintained is crucial to the design of PCR microfluidics. Various heating methods have been employed and they can be broadly categorized into contact and non-contact heating methods. Contact heating has been implemented in PCR microfluidics using integrated thin-film platinum resistors as both the heating and sensing elements on these chips [156] and non-contact methods include hot-air cycling [157]. Both methods have low power consumption and are amenable for use at the point of care.

100

Isothermal techniques have been developed to perform nucleic acid amplification without thermal cycling [158, 159]. This feature removes the need of using different temperatures by making use of enzymes to perform amplification at a single temperature. For example, Biohelix (Beverly,

MA, USA) has developed a method for performing isothermal amplification called helicase- dependent amplification that is amenable for point-of-care use [160].

Product Detection: There are many methods to detect nucleic acids, which one of the primary methods is quantifying nucleic acids by ultraviolet light (260 nm) absorption. Nonetheless, fluorescence-based techniques remain the most commonly employed due to their high level of sensitivity and low background noise [147]. Fluorescent dyes can either bind to non-specific locations (general interactions) or specific locations of molecule depending on the application.

Cepheid Inc., for example, has developed a real-time fluorescent PCR detection (fluorescently- labeled probes) requiring instruments that may be used in some but not all point-of-care settings.

Cepheid’s GeneXpert test platform (Figure 4.3) has been tested for clinical trials in four developing countries and showed promising results for detecting tuberculosis [161]. However, this system requires uninterrupted and stable electrical power supply and annual validation of the system, and generates considerable more waste than microscopy technique [162] which may lead to problems of waste management in resource-limited settings.

Figure 4.3. Cepheid’s GeneXpert test platform (left) integrates sample processing and PCR in a disposable plastic cartridge (right) containing reagents for cell lysis, nucleic acid extraction,

101 amplification and amplicon detection. This system has been used to detect drug-resistant tuberculosis cases by amplifying Mycobacterium tuberculosis specific sequence of the rpoB gene and probing with molecular beacons for mutations conferring rifampicin resistance [161] .

At the point of care, electrochemical methods may also be suitable due to their compatibility with low-cost and portable analyzers [150]. For example, Nanosphere is building a scanner-based detector to detect DNA via nanoparticle probes [163]. Gold nanoparticles functionalized with oligonucleotides are used as probes for DNA sequences complementary to the sequences of those oligonucleotides (Figure 4.4). Visualization of the gold nanoparticle and, hence DNA content, is performed using a signal amplification method in which silver is reduced at the surface of the gold and a scanner is used to measure amount of light scattered.

102

Another commercial effort towards a point-of-care nucleic acid device includes Spartan RX

CYP2C19, by Spartan Bioscience, which is the first point-of-care genetic testing system to receive the CE marking [140].

Few other microfluidic technologies (under development) have shown promising results.

Magnetic nanoparticles and nuclear magnetic resonance have been used by the Weissleder group in a barcode assay for genetic detection of Mycobacterium tuberculosis from unprocessed sputum specimens [164]. They developed a platform for the detection of MTB-DNA based on a magnetic barcoding strategy without a bacterial isolation step. PCR-amplified mycobacterial genes are sequence-specifically captured on microspheres, labeled by magnetic nanoprobes and detected by nuclear magnetic resonance. The Soh Lab has also utilized magnetic particles for isolation and concentration in nucleic acid assays [165]. The Landers group has developed a visual DNA detection by magnetic beads aggregation resulting from DNA/bead interactions driven either by the presence of a chaotrope (a nonspecific trigger for aggregation) or by hybridization with oligonucleotides on functionalized beads (sequence-specific) [166].

For tuberculosis, microfluidic technologies can provide a next generation of diagnostic devices, especially for the management of tuberculosis near patients and control of diseases in communities.

This technology should be able to cope with the challenging engineering tasks of 1) handling large volume (1 – 30 mL) of complex specimens, 2) reducing the sample volume or pre-concentrate analyte, and 3) integrating mechanical, thermal, and optical detection processes, to ensure an adequate analytical sensitivity, clinical validity, and user-friendliness of the test [167]. Here we focus on the sample preparation module which is the least developed.

103

4.2 Methods

4.2.1 Target isolation and concentration

Mycobacterium culture. Mycobacterium tuberculosis culturing requires biosafety level 3 containment and it is a slow-growth culture (2 – 4 weeks) [168]. Mycobacterium smegmatis is used as a surrogate for Mycobacterium tuberculosis because of its similar cell wall property [169].

Mycobacterium smegmatis (mc 2155, ATCC ) was incubated in Middlebrook 7H9 broth with

Middlebrook ADC Enrichment ( Sigma-Aldrich ) in an incubator shaker (200 rpm) at 37 ºC for 2 – 3 days or until OD 600 ~ 0.8 - 1.0. Cell suspension was pellet using centrifugation (3000 rpm for 5 min), supernatant was discard and cell pellet was resuspended in 1x tris-EDTA (TE) buffer pH 8.0 ( Sigma-

Aldrich ).

Artificial sputum. Porcine mucin, a commonly used proxy for clinical sputum, was used as an artificial sputum. Porcine mucin solution was freshly prepared for each experiment from dissolving porcine mucin ( Sigma Aldrich ) in water to make 25 % (w/v) [170]. Porcine mucin solution was mixed with M.smegmatis suspension in TE buffer at 4 to 1 ratio to mimic bacterial spiked sputum.

Liquefying and decontamination. 4% (w/v) NaOH with 0.5% (w/v) N-acetyl-L-cysteine (NALC) or NALC-NaOH solution is used for digestion and decontamination of sputum samples [171]. NaOH and NALC ( Sigma Aldrich ) were prepared in deionized (DI) water.

Target cell isolation using coated magnetic beads. TB beads ( Microsens Medtech ), coated magnetic beads with ligand specific to lipoarabinomannan on mycobacterial cell wall [172, 173], were used to isolate mycobacteria from sputum samples. This type of beads have been used in microscopy to concentrate mycobacterium [174]. TB beads stock solution was diluted 2 – 200 times in NALC-NaOH solution. Sputum was mixed with working concentration of coated magnetic beads in liquefying agent and incubated at room temperature for 15 min.

104

Microfluidic chip for cell isolation and concentration. To deliver a rapid target isolation and concentration, we developed a disposable device consisting of multi-layer polydimethylsiloxane

(PDMS, Dow Corning ) on flat substrate placed over either an external permanent magnet or a solenoid (Fig.4.8) for a separation of magnetic bead-bound particles (bacteria) from complex sample matrices. A PDMS chip containing channel, as shown in Figure 4.8, was fabricated by conventional soft-lithography. Small holes were punched with a needle at the end of channels as inlet and outlet ports. A 10-mm hole was cored out from the bottom layer of PDMS to form a capture chamber. The device was formed by an irreversible bonding of PDMS layers and glass substrate. PE tubes (Fisher

Sci ) were inserted to inlet and outlet for liquid flow. Sample was flowed through a vertical chamber for target separation and enrichment via the bottom channel and exit at the top channel using a syringe pump ( Kent Scientific) at high flow rate (~mL/min). Magnetic beads were trapped in the chamber and allowed the rest of particles or buffer pass through the outlet. Trapped beads were then washed to remove sputum residual and unwanted particles and ready for the downstream analysis.

Cell lysis and quantification. To quantify bacterial isolation and capture efficiency, PCR and gel electrophoresis were performed as follow. Trapped beads in the capture chamber were washed once with 1x TE buffer and the chamber was filled with 10x TE buffer. The disposable chip was then placed on hot plate or thin film heater set to 85⁰C for 5 min and cooled down immediately afterwards.

PCR was performing using primers targeting one of glycosyltransferase (gtf ) genes (forward primer: 5’- ATG TTC CAC TCG TCA CGC TG -3’, reverse primer: 5’- ACC AAT CGA ACG CGG

TCA TC -3’) [175] and GeneAmp Fast PCR kit ( Invitrogen ). Cell lysate (1 µL) from the chamber was used in 20 µL direct PCR reaction run on a thermo cycler (MyCycler, Bio-Rad ) for 10 sec at 95⁰C,

35 cycles of 94⁰C (1 sec) and 62⁰C (25 sec), and 72⁰C for 10 sec. PCR products were quantify using standard gel electrophoresis on 2% agarose gel with 1x SYBR green I dye (Invitrogen ) in a

Tris/Acetate/EDTA (TAE) buffer at 100 V for 50 min. After the electrophoresis was completed, PCR

105 products were visualized using a gel-imager under ultraviolet light. Band intensities were quantified using ImageJ and normalized with the band intensity of the ladder at the appropriate length.

4.2.2 Fast multiplex PCR

Target DNA was amplified using fast PCR which combines annealing and extension steps into one. Target genes for tuberculosis detection and drug resistant identification are IS6110 , a M. tuberculosis complex-specific insertion sequence for M. tuberculosis infection detection, rifampicin

(RIF) - resistant strain with a common rpoB mutation, and katG mutant isoniazid (INH)-resistant strain [176] (Table 4.2). Primers for inhA mutant INH-resistant strain were also designed and tested.

Following parameters should to be considered when designing primers for multiplex PCR; primer length, amplicon length, primer melting temperature, optimum annealing temperature, primer melting temperature mismatch, GC content, primer secondary structure, and primer dimer. Various ratios of each primer sets were compared to achieve the highest yield for each target.

Table 4.2. Target genes and primers.

Gene Product Target Forward primer Reverse primer involved length (bp)

TB 5’-GAG CGT AGG CGT 5’-GCT TCG GAC CAC IS6110 105 identification CGG TGA CAA AGG-3’ CAG CAC CTA ACC-3’

Rifampin 5’-TCA CAC CGC AGA 5’-CAC GCT CAC GTG rpoB 152 resistance CGT TGA T-3’ ACA GAC C-3’

Isoniazid 5’-TCG TAT GGC ACC 5’- CAG CTC CCA CTC katG 123 resistance GGA ACC-3’ GTA GCC-3’

Our designed primers amplify both wild type and mutated strains (Fig.4.5). Biotinylated primers were used to accompany the following detection technique (discuss below). Purified genomic DNA

106 of M.tuberculosis (ATCC ) was resuspended in nuclease free water ( Fisher Sci ). Primers were design using primer3 and custom synthesized ( Sigma-Aldrich ). 1 µL of purified M.tuberculosis was used in

20 µL direct PCR reaction run on a thermo cycler for 10 sec at 95⁰C, 35 cycles of 94⁰C (1 sec) and

62⁰C (25 sec), and 72⁰C for 10 sec. PCR products were visualized using standard gel electrophoresis on 2% agarose gel with 1x SYBR green I dye in a Tris/Acetate/EDTA (TAE) buffer at 100 V for 75 min. After the electrophoresis was completed, PCR products were visualized using a gel-imager under ultraviolet light. Rifampin-resistant DNA strains were received from Dr. Barry

Kreiswirth, Public Health Research Institute (PHRI) New Jersey Medical School.

Figure 4.5. Target amplification using biotinylated primers in a fast multiplex PCR.

4.2.3 PCR product detection using gold-silver amplification

Probe functionalization. The plastic surface was treated with 0.1% w/v poly-L-lysine (PLL,

Sigma-Aldrich ), incubated for two hours at room temperature on an orbital shaker (low speed,

Thermo Sci ), and rinsed once with 0.2 M 2-(N-morpholino)ethanesulfonic acid (MES) buffer (Sigma-

Aldrich ).

107

Capture oligos or single-stranded DNA (ssDNA) specific to wild-type genes were then immobilized on the treated plastic surface via EDC/NHS coupling. Capture oligos were designed and custom synthesized with carboxylic acid group at the 5’ end (Tib MOLBIOL ). 0.6 µM of carboxyl- oligo was prepared in MES buffer (200mM, pH=3) with 12 mM of N-(3-dimethylaminopropyl)-N′- ethylcarbodiimide hydrochloride (EDC, Sigma-Aldrich ) and 5 mM of N-hydroxysulfosuccinimide

(sulfo-NHS, Thermo Sci ), mixed on a vortex mixer (Scientific Industries ) at a low setting for 15 minutes, and spotted the mixture on the treated surface. (Note that EDC and sulfo-NHS required to be prepared right before the reaction.) After an hour of incubation at room temperature, the solution was aspirated and surface was rinsed with 200 mM MES buffer. Treated plastic surface was blocked with 1% BSA in PBS to prevent non-specific binding as well as hydrophilize surface.

Hybridization. After the PCR was completed, PCR products were mixed with hybridization buffer (PerfectHyb™, Sigma-Aldrich ) at 1:2 ratio. The mixture was heated at 95⁰C for 10 min and chilled on ice immediately for 2 min to denature dsDNAs (double-stranded DNA or PCR products) into ssDNAs for hybridization. Denatured PCR products were then incubated with probes for 1.5 hr at 37⁰C and rinsed once with hybridization buffer and twice with 0.1x saline-sodium citrate (SSC,

Sigma-Aldrich, or 15 mM NaCl in 1.5 mM sodium citrate).

108

Figure 4.6. PCR product detection using gold-silver amplification. ( Left ) Illustration of biochemical reactions in detection zones at different signal detection steps after PCR. (Right ) Silver enhancement yields signals which can be read with low-cost optics or examined by eye.

Gold-silver amplification. Gold nanoparticle conjugated streptavidin ( Nanoprobes ) in 1%BSA in

0.05%Tween/PBS was added to plastic surface after the hybridization step was completed. A mixture of gold conjugated streptavidin and gold conjugated biotin can be used to increase the amount of gold nanoparticles available on surface for silver reduction reaction. After the surface was incubated with gold conjugated streptavidin, silver reagents (silver salts and reducing agents, OPKO

Diganostics ) were mixed and flowed through the surface of oligos and gold nanoparticles. Signal can be readout by simple optics as previously discussed (Fig. 4.6).

109

Enzyme cleavage. T7 endonuclease I (New England Biolabs ) was used to cleave non-perfectly matched DNA after hybridization of probe due to some of non-specific binding of mutant gene products to probes (Fig. 4.7).

Figure 4.7. PCR product detection using gold-silver amplification with enzyme cleavage to remove DNA mismatches.

4.3 Results and Discussion

4.3.1 Sputum processor

Up to now, sputum processing has been complex: it requires proper facilities and well-trained staff for infection and contamination control, extended incubation time for culture, and infrastructure for quantitatively analyzing sputum cell counts. However, such information is extremely valuable for administering targeting antibiotics and monitoring of patient responses to specific treatments. 110

Accurate sputum analysis requires that the specimen is tested fresh within 2 hours [177]. However, conventional sputum processing involves multiple manual steps for adding and removing buffer and waste and centrifugation, including vigorous shakes and a precise manual pipetting due to its viscosity. DNA isolation from sputum requires many steps involved in multiple equipment and worker actions which leads to high cost and is time consuming. Sputum is heterogeneous and collected in large volume. World-to-chip interface is also a major problem due to the small volume sampled out from patient sputa which might not be homogeneous. A paper-based viscous sample preparation has been used for DNA extraction from E.coli-spiked mucin by using a microfluidic origami device with dry lysis buffer on a storage pad [170]. While this technique does not need an external power to operate, it requires long processing time (1 – 1.5 hr), large buffer volume (6 times of sample volume) which will be difficult to handle with a small paper device (1 – 3 mL of sputum), and easily leads to cross-contamination. Other microfluidic platforms for sample processing are handling small volume (~uL) [178]. We have developed a microfluidic design that can process large volumes (mL) of sputum and isolate small numbers of bacteria, and be used in-line with downstream microfluidic detection and manipulation.

Target-specific isolation is performed by adding coated magnetic beads to samples before flowing it through the device which has external permanent magnet aligned with the capture chamber, therefore, non-target substances would not be captured and isolated out (Fig. 4.8).

111

After captured, magnetic beads inside the chamber can be washed by flowing wash buffer through while beads are trapped in place with magnetic force and resuspend in smaller volume and release from chamber for the following step. The DNA or RNA extraction can be done with this device by including heating element at the bottom of chamber for heat lysis. Magnetic beads will also act as heat transfer material to enhance heat exchange. Our design forced the fluids to move vertically over a confined capture zone, where the beads can be easily trapped. The trapped beads were efficiently captured (Fig. 4.8, right; Fig. 4.9A), and washed to remove sputum residual and unwanted particles. Our design can trap large volume of beads without expanding the footprint, lowering the capture efficiency, clogging flow channel, or shearing targets off from bead surface under high flow rate (mL/min range) when compare to traditional one-layer microfluidic designs for capturing beads or particles. Our platform can be modified to work on diverse targets by changing the marker coated on magnetic beads, and can incorporate multiple washes to eliminate non- specific binding to beads.

We directly compared the use of this rapid, microchip-based sputum processor compared to conventional sputum processing by conventional centrifugation (Fig. 4.9B). As a model for bacteria

112 in a clinical specimen, we used various dilutions of Mycobacterium smegmatis spiked in an artificial sputum that has been commonly used as a proxy for clinical sputum [170] (porcine mucin). 0.5 mL of bacteria, at dilutions spanning five orders of magnitude, spiked in artificial sputum were mixed with either 1 mL of liquefying solution for conventional sputum processing or 1 mL of magnetic beads in liquefying solution prior flowing to our device. Coated magnetic beads with ligand specific to lipoarabinomannan on mycobacterial cell wall were used. Isolated bacteria from both methods were rinsed with buffer, and lysed using heat. 1 µL of lysate was used for PCR in both methods, and 10

µL of amplified products were analyzed by gel electrophoresis. Band intensities were quantified using ImageJ and normalized with intensity of DNA ladder at the product length. The results were encouraging that our simple, rapid USP could isolate bacteria from a complex matrix as well as the laborious processes of conventional centrifugation to statistical significance, and across five orders of magnitude in bacterial dilution. A minimal band intensity from no template control sample suggested that there is no cross-contamination or carryover particles presented when performing sputum processing using our device.

Figure 4.9. Performance of a handheld sputum processor. (A) Percentage of beads captured from 3 mL and 0.5 mL of magnetic beads at flow rates of 0.5 and 2.0 mL/min, as measured by optical density (OD750) of beads in solution before and after capturing. (B) Comparison of conventional sputum processing by conventional centrifugation and our USP at various dilution of bacteria (M.smegmatis as a laboratory model of M.tuberculosis ) spiked in artificial sputum (porcine mucin from Sigma Aldrich, a commonly used proxy for clinical sputum [170]), as measured by intensity of bands on electrophoresis gel of PCR products of bacterial DNA. Data are averages ±1 SD (n=3). Dashed line indicates band intensity of no template control .

113

4.3.2 Target amplification and detection

Fast multiplex PCR. Target DNA was amplified using fast PCR which combines annealing and extension steps into one. We chose PCR for our multiple targets detection because PCR is more robust than other isothermal amplifications. M. tuberculosis genomic DNA is difficult to amplify due to its high GC content. Other amplification techniques might be able to amplify one target of this

DNA but primer design for more than one target could be complicated and might not work well. Our initial target genes are IS6110, an M. tuberculosis complex-specific insertion sequence for

M.tuberculosis infection detection, an RIF-resistant strain with a common rpoB mutation, and an

INH-resistant strain with a common katG mutation. After 35 cycles (< 1hr), we are able to detect M.

Tuberculosis purified genomic DNA and then quantify it by gel electrophoresis (Fig.4.10). Both wild- type and mutated strains were amplified using these primer sets which anneal outside the area of interest since there are more than one mutation sites for each drug resistance.

Figure 4.10. Multiplex PCR. Gel electrophoresis shows each target and multiplex PCR using in- house primer design with a fast PCR mix, NTC is no-template control.

We designed primers and probes such that amplified products from wild-type strains would hybridize to detection probes. For example, if IS6110 product is detectable but rpoB is not, the sample will be TB positive with rifampin resistant strain. If we designed probes specific to each

114 mutations, many probes would be needed to detect one drug resistance. For instance, rifampin has a minimum of five most frequent codons reported as drug resistant strains.

Gold-silver amplification. Biotinylated primers were used in our multiplex PCR to simplify our detection system. We adopted gold-silver amplification which we used in our previous chapters in this nucleic acid detection application. Gold conjugated streptavidin was added after amplified products bind to capture oligos on surface. Gold conjugated biotin could also be added to further increase the amount of gold nanoparticles on surface, however, the binding capacity of streptavidin would be lower than theoretically 4 binding sites after streptavidin bound to biotin (not all 4 binding sites can be occupied by biotin molecules). Due to the high GC content of M.tuberculosis DNA, the designed probes cannot differentiate between the wild-type and drug resistant strains (Fig. 4.11).

2.5

1.5

Optical density 0.5

0 NTC wild type RRDR DNA sample

However, to show that this technique works well on other genomic DNA, we demonstrated the detection of Staphylococcus aureus (Fig. 4.12) where we differentiated methicillin-resistant S. aureus (MRSA) from methicillin-sensitive S. aureus (MSSA).

115

Figure 4.12. Successful detection of nucleic-acid targets using the gold-silver amplification chemistry. ( Left ) Optical densities of silver signal from target DNA bound to probes specific to MSSA and MRSA genes for differentiation of two S. aureus strains. Target DNA was generated through multiplex PCR reaction which amplifies SA and MecA from whole genomes. SA gene is present in both S. aureus strains while MecA is present only in MRSA strain. ( Right ) Preliminary limit of detection of S. aureus using gold-silver amplification from PCR products. Data are averages ±1 SD ( n=3).

Enzyme cleavage. T7 endonuclease I was used to cleave non-perfectly matched DNA after hybridization of probe due to some of non-specific binding of mutant gene products to probes. Gold- silver amplification was carried out afterwards for signal detection. Signal from rpoB probe would need to be normalized with IS6110 probe or other control probes to be able to compare and identify as drug resistance.

1.40 1.20 1.00 0.80 No enzyme 0.60 2 units of T7 0.40 Optical Density 0.20 0.00 RIF-R RIF-S

Figure 4.13. Probe hybridization with enzyme cleavage prior signal amplification. Optical densities of silver signal from target DNA bound to probes specific to rifampin-sensitive strain (rpoB gene) after enzyme cleavage to remove DNA mismatch for differentiation of two M.tuberculosis strains: rifampin-resistant (RIF-R) and rifampin-sensitive (RIF-S) strains. Data are averages ±1 SD ( n=3).

116

4.4 Conclusion and future directions

Unless there is a universal treatment for tuberculosis, a test for drug resistant TB would still be crucial to minimize transmission and improve health outcomes of TB patients. MDR-TB and XDR-

TB diagnostic tests would also be in great need in the near future, especially for low- and middle- income countries.

To simplify sputum processing for nucleic acid amplification testing for TB, we designed a low- cost, vertical chamber device with the outlet higher than the inlet, which can process large volume of specimens without expanding the footprint, thereby lowering the capture efficiency, clogging flow channel, or shearing targets off from bead surface under high flow rate (mL/min range) when compare to traditional one-layer microfluidic designs for capturing beads or particles. We investigated fast multiplex PCR and gold-silver amplification techniques for detecting three target genes from M.tuberculosis gDNA. An additional design component would also be required to seamlessly transferred patient’s sputum from sputum collector to the device to prevent the cross- contamination and protect the test operator from exposing to M.tuberculosis (no BSL-3 infrastructure required).

Future work is still required to further develop and improve the technology in this chapter towards the final product for TB and drug-resistant TB detection at the point of care. For example, other targets including inhA mutant isoniazid-resistant strain, gyrA mutant fluoroquinolone-resistant strain, embB mutant ethambutol-resistant, and rrs mutant aminoglycoside-resistant; resistance to isoniazid (1st line drug), fluoroquinolone, ethambutol, and aminoglycoside (2nd line drugs) could be added to the panel. Integration of all steps into one device which operated using minimal power consumption, a small number of hands-on steps and short assay time would be ideal. Other techniques for DNA amplification or detection ( e.g. isothermal amplification, real-time PCR, sequencing, etc.) could be adopted to this work as well. Once the prototype is ready, clinical

117 specimens for both wild-type (drug-sensitive) and mutated (drug-resistant) strains should be tested to determine the cutoff for disease diagnosis and assess clinical performance of the technique.

118

Chapter 5 Conclusion and Future Directions

This dissertation investigates microfluidic technologies for a point-of-care testing focusing on global health applications. We first developed a miniaturized ELISA for duplex HIV/syphilis test which can be used in resource-limited settings by replacing conventional microfluidic strategies

(including materials, advanced fluid handling, and signal amplification and detection) with low-cost, portable, easy-to-operate materials and techniques. We evaluated our test in Rwanda and it exhibited performance equal to lab-based immunoassays, with versatility in type of blood sample

(whole blood, plasma, and sera). Although this format provided the HIV/syphilis test results within

20 minutes, the test required a few manual steps to perform the test. An ideal POC test would perform all the functions of laboratory-based ELISA, but require as few manual steps as possible for the end user (equivalent in simplicity to a CLIA-waived test). Hence, we integrated three important off-chip processes in a diagnostic test - liquid handling , optical signal detection, and data communication – in a low-cost, versatile, handheld instrument capable of performing microfluidic immunoassays on reagent-loaded cassettes with the speed, portability and ease-of-use of a lateral- flow test or a rapid test. Our test result is not subject to user interpretation, a feature which might counteract the observed lower accuracy of current rapid tests when practiced in the field, as well as reduce variability of results between highly trained and minimally trained users. Another clinical benefit is automated tracking of data results, to eliminate mistakes in keeping patient records.

Many technologies work well in research settings but not in the field, which exhibits additional complexities not replicated in a research laboratory: local specimen strain and subtypes, disease prevalence, and real POC infrastructures, environmental conditions, and testing protocol. We have demonstrated that our microfluidic devices work well in the field, however, all the testing were carried out by the development team. One of our goals is to get our diagnostic devices to be used independently by the end users. We then re-designed our microfluidic immunoassay, which can be

119 performed by layperson, by simplifying user interface, minimizing manual steps, and scaling it down into a small smartphone accessory. We engineered it such that it reduced the cost of the hardware and power consumption by replacing an electrical pump with a simple mechanic pump. The electronics drew power directly from a smartphone via an audio port. We also incorporated conditions replicated real shipping and transportation conditions, minimized user steps, and increased field-readiness to enable a “plug-and-play” operation for the user. A clinical pilot study was carried out using this smartphone accessory in Rwanda by local health care workers on enrolled patients to assess user feasibility and acceptability for future design improvement.

Overall, our microfluidic-based immunoassay served as a miniaturized ELISA in resource- limited settings. In particular, the high sensitivities of our test for both HIV and syphilis are attractive for screening: in remote settings, potentially infected patients can be diagnosed immediately, and their samples can be confirmed by tests with greater specificity, different antigen preparations, test principles and/or biological targets. This test can also be used outside of the antenatal care clinic, as HIV/syphilis combination testing is important in pre-screening of blood donations as well as epidemiological surveillance (as STIs are particularly poorly monitored in developing countries).

Also, this test is relevant for use by primary care physicians and outpatient clinics in developed countries, as on-site diagnosis can lead to higher rates of correct treatment and lower rates of unnecessary over-treatment.

To increase sensitivity and specificity of our immunoassay, few assay parameters can be reconsidered. For instance, adding or changing the surface antigens to include other markers or adjust the surface chemistry for antigen attachment (e.g. covalent bonds or biotin chemistry) could reduce the nonspecific binding of protein to surface. Gold-labeled disease antigens could be used instead of gold-labeled anti-hIgG and anti-hIgM to increase specificity of the test. Washing steps could be changed to include more wash volume or plugs and/or various types of wash buffer to help remove nonspecific binding of proteins on surface. P24 antigen assay could be incorporate to

120 increase clinical performance of HIV test. On-chip blood separation to remove hematocrit could also enhance the assay performance. Other improvements, for example, long-term storage conditions of test cassettes, device ruggedness, and reproducibility of the test should be done prior a large scale clinical trial to obtain an approval (countries, FDA, or WHO). The immunoassay can be extended to other clinically useful markers for STI’s such as hepatitis C, chlamydia, and gonorrhea, for which few adequate POC options exist. Finally, this technology can be adapted for consumer health applications ( e.g. hormone and vitamin levels, etc.) in developed countries.

Nucleic acid test is another important tool in diagnostics. There are few other clinical applications which required this type of testing which can provide more information than protein- based test. Populations in resource-limited settings, in particular, are suffering from lack of access to nucleic acid tests due to its limited availability (both hardware and lab technicians) and high-cost per test. Seamless integration, high throughput, fast turnaround time, and low cost microfluidic- based test will improve the clinical outcomes at the point of care.

121

References

1. World Health Organization. The top 10 causes of death . [cited 2014; Available from: http://www.who.int/mediacentre/factsheets/fs310/en/index4.html# .

2. World Health Organization. Causes of death, by WHO region . Global Health Observatory Data Repository 2014 [cited 2015 March 15].

3. Yager, P., T. Edwards, E. Fu, et al., Microfluidic diagnostic technologies for global public health. Nature, 2006. 442 (7101): p. 412-418.

4. G. B. D. Mortality Causes of Death Collaborators, Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet, 2014.

5. WHO, UNICEF, and UNAIDS, Towards universal access: scaling up priority interventions in the health sector: progress report 2009. 2009, World Health Organization: Geneva.

6. Peeling, R.W. and D. Mabey, Point-of-care tests for diagnosing infections in the developing world. Clin Microbiol Infect, 2010. 16 (8): p. 1062-9.

7. Peeling, R.W. and S. Nwaka, Drugs and Diagnostic Innovations to Improve Global Health. Infectious Disease Clinics of North America, 2011. 25 (3): p. 693-+.

8. Yager, P., G.J. Domingo, and J. Gerdes, Point-of-care diagnostics for global health. Annu Rev Biomed Eng, 2008. 10 : p. 107-44.

9. Peeling, R.W., K.K. Holmes, D. Mabey, et al., Rapid tests for sexually transmitted infections (STIs): the way forward. Sexually Transmitted Infections, 2006. 82 : p. V1-V6.

10. Genomics and World Health: Report of the Advisory Committee on Health Research . 2002, World Health Organization: Geneva.

11. Chin, C.D., V. Linder, and S.K. Sia, Lab-on-a-chip devices for global health: Past studies and future opportunities. Lab on a Chip, 2007. 7(1): p. 41-57.

12. Yager, P., G.J. Domingo, and J. Gerdes, Point-of-care diagnostics for global health. Annual Review of Biomedical Engineering, 2008. 10 : p. 107-144.

13. Sia, S.K. and L.J. Kricka, Microfluidics and point-of-care testing. Lab on a Chip, 2008. 8(12): p. 1982-1983.

14. Gray, R.H., F. Makumbi, D. Serwadda, et al., Limitations of rapid HIV-1 tests during screening for trials in Uganda: diagnostic test accuracy study. British Medical Journal, 2007. 335 (7612): p. 188-190.

15. Klarkowski, D.B., J.M. Wazome, K.M. Lokuge, et al., The Evaluation of a Rapid In Situ HIV Confirmation Test in a Programme with a High Failure Rate of the WHO HIV Two-Test Diagnostic Algorithm. Plos One, 2009. 4(2).

16. Kroidl, I., P. Clowes, W. Mwalongo, et al., Low Specificity of Determine HIV1/2 RDT Using Whole Blood in South West Tanzania. Plos One, 2012. 7(6).

122

17. Dagnra, A.Y., S. Dossim, M. Salou, et al., Evaluation of 9 rapid diagnostic tests for screening HIV infection, in Lome, Togo. Medecine Et Maladies Infectieuses, 2014. 44 (11-12): p. 525-529.

18. Faulstich, K., R. Gruler, M. Eberhard, et al., Handheld and Portable Reader Devices for Lateral Flow Immunoassays. Lateral Flow Immunoassay, 2009: p. 157-183.

19. Mark, D., S. Haeberle, G. Roth, et al., Microfluidic Lab-on-a-Chip Platforms: Requirements, Characteristics and Applications. Microfluidics Based Microsystems: Fundamentals and Applications, 2010: p. 305-376.

20. Posthuma-Trumpie, G.A., J. Korf, and A. van Amerongen, Lateral flow (immuno) assay: its strengths, weaknesses, opportunities and threats. A literature survey. Analytical and Bioanalytical Chemistry, 2009. 393 (2): p. 569-582.

21. Chin, C.D., V. Linder, and S.K. Sia, Commercialization of microfluidic point-of-care diagnostic devices. Lab on a Chip, 2012. 12 (12): p. 2118-2134.

22. Parsons, L.M., A. Somoskovi, C. Gutierrez, et al., Laboratory Diagnosis of Tuberculosis in Resource-Poor Countries: Challenges and Opportunities. Clinical Microbiology Reviews, 2011. 24 (2): p. 314-+.

23. UNAIDS, GAP Report . 2014.

24. Herring, A., R. Ballard, D. Mabey, et al., Evaluation of rapid diagnostic tests: syphilis. Nature Reviews Microbiology, 2006: p. S33-S40.

25. Lackritz, E.M., N. Shaffer, and C. Luo, Prevention of mother-to-child HIV transmission in the context of a comprehensive AIDS agenda in resource-poor countries. Journal of Acquired Immune Deficiency Syndromes, 2002. 30 (2): p. 196-199.

26. De Cock, K.M., M.G. Fowler, E. Mercier, et al., Prevention of mother-to-child HIV transmission in resource-poor countries - Translating research into policy and practice. Jama- Journal of the American Medical Association, 2000. 283 (9): p. 1175-1182.

27. Genc, M. and W.J. Ledger, Syphilis in pregnancy. Sex Transm Infect, 2000. 76 (2): p. 73- 9.

28. Newman, L., M. Kamb, S. Hawkes, et al., Global Estimates of Syphilis in Pregnancy and Associated Adverse Outcomes: Analysis of Multinational Antenatal Surveillance Data. Plos Medicine, 2013. 10 (2).

29. Gomez, G.B., M.L. Kamb, L.M. Newman, et al., Untreated maternal syphilis and adverse outcomes of pregnancy: a systematic review and meta-analysis. Bulletin of the World Health Organization, 2013. 91 (3): p. 217-226.

30. Johri, M. and D. Ako-Arrey, The cost-effectiveness of preventing mother-to-child transmission of HIV in low- and middle-income countries: systematic review. Cost Eff Resour Alloc, 2011. 9: p. 3.

31. Terris-Prestholt, F., D. Watson-Jones, K. Mugeye, et al., Is antenatal syphilis screening still cost effective in sub-Saharan Africa. Sexually Transmitted Infections, 2003. 79 (5): p. 375-381.

32. World Health Organization, Syphilis in pregnancy , in Global Health Observatory data 2015.

123

33. Nkeze, J.N., N.T. Constantine, and R.Y. Zhao, Laboratory Testing for HIV Infection: Advances After 28 Years. Molecular Diagnostics: 12 Tests That Changed Everything, 2014: p. 81- 106.

34. Respess, R.A., M.A. Rayfield, and T.J. Dondero, Laboratory testing and rapid HIV assays: applications for HIV surveillance in hard-to-reach populations. Aids, 2001. 15 : p. S49-S59.

35. Miedouge, M., M. Greze, A. Bailly, et al., Analytical sensitivity of four HIV combined antigen/antibody assays using the p24 WHO standard. Journal of Clinical Virology, 2011. 50 (1): p. 57-60.

36. Rusnak, J.M., C. Butzin, D. McGlasson, et al., False-positive rapid plasma reagin tests in human immunodeficiency virus infection and relationship to anti-cardiolipin antibody and serum immunoglobulin levels. J Infect Dis, 1994. 169 (6): p. 1356-9.

37. Peeling, R.W. and H. Ye, Diagnostic tools for preventing and managing maternal and congenital syphilis: an overview. Bull World Health Organ, 2004. 82 (6): p. 439-46.

38. Centers for Disease Control and Prevention, Sexually Transmitted Diseases Treatment Guidelines, 2010 . 2010, Centers for Disease Control and Prevention: Atlanta, GA.

39. Binnicker, M.J., Which algorithm should be used to screen for syphilis? Current Opinion in Infectious Diseases, 2012. 25 (1): p. 79-85.

40. CDC. Sexually Transmitted Diseases Treatment Guidelines, 2010 . 2010 [cited 2014 Mar 5]; Available from: http://www.cdc.gov/std/treatment/2010/ .

41. Sena, A.C., B.L. White, and P.F. Sparling, Novel Treponema pallidum Serologic Tests: A Paradigm Shift in Syphilis Screening for the 21st Century. Clinical Infectious Diseases, 2010. 51 (6): p. 700-708.

42. Young, H., J. Pryde, L. Duncan, et al., The Architect Syphilis assay for antibodies to Treponema pallidum: an automated screening assay with high sensitivity in primary syphilis. Sexually Transmitted Infections, 2009. 85 (1): p. 19-23.

43. Ratnam, S., The laboratory diagnosis of syphilis. Can J Infect Dis Med Microbiol, 2005. 16 (1): p. 45-51.

44. Owusu-Edusei, K., T.L. Gift, and R.C. Ballard, Cost-Effectiveness of a Dual Non- Treponemal/Treponemal Syphilis Point-of-Care Test to Prevent Adverse Pregnancy Outcomes in Sub-Saharan Africa. Sexually Transmitted Diseases, 2011. 38 (11): p. 997-1003.

45. Kuznik, A., M. Lamorde, A. Nyabigambo, et al., Antenatal Syphilis Screening Using Point- of-Care Testing in Sub-Saharan African Countries: A Cost-Effectiveness Analysis. PLoS Med, 2013. 10 (11): p. e1001545.

46. Lii, J., W. Hsu, W. Lee and S.K. Sia, Microfluidics, in Kirk-Othmer Encyclopedia of Chemical Technology . 2006, John Wiley and Sons.

47. CDC. Flu Activity & Surveillance. . 2009; Available from: http://www.cdc.gov/flu/weekly/fluactivity.htm .

48. Dzenowagis, J., Connecting for health : global vision, local insight. , G. Kernen, Editor. 2005, World Health Organization: Geneva.

124

49. NIH. A Summit on Mobile Technologies as a Platform for Health Research and Healthcare Delivery . 2009; Available from: http://www.fic.nih.gov/news/events/mhealthsummit.htm .

50. Schreiner, S., 2009 Inaugural mHealth Summit Successfully Brings Together Health, Technology and Policy Communities to Advance Technological Innovation in Global Healthcare , in Business Wire . 2009.

51. Hedestam, G.B.K., R.A.M. Fouchier, S. Phogat, et al., The challenges of eliciting neutralizing antibodies to HIV-1 and to influenza virus. Nature Reviews Microbiology, 2008. 6(2): p. 143-155.

52. Sia, S.K., V. Linder, B.A. Parviz, et al., An integrated approach to a portable and low-cost immunoassay for resource-poor settings. Angewandte Chemie-International Edition, 2004. 43 (4): p. 498-502.

53. Gupta, S., S. Huda, P.K. Kilpatrick, et al., Characterization and optimization of gold nanoparticle-based silver-enhanced immunoassays. Analytical Chemistry, 2007. 79 (10): p. 3810- 3820.

54. de la Escosura-Muniz, A., M.M.D. Costa, and A. Merkoci, Controlling the electrochemical deposition of silver onto gold nanoparticles: Reducing interferences and increasing the sensitivity of magnetoimmuno assays. Biosensors & Bioelectronics, 2009. 24 (8): p. 2475-2482.

55. Widegren, J.A. and R.G. Finke, A review of the problem of distinguishing true homogeneous catalysis from soluble or other metal-particle heterogeneous catalysis under reducing conditions. Journal of Molecular Catalysis a-Chemical, 2003. 198 (1-2): p. 317-341.

56. Patakfalvi, R., S. Papp, and I. Dekany, The kinetics of homogeneous nucleation of silver nanoparticles stabilized by polymers. Journal of Nanoparticle Research, 2007. 9(3): p. 353-364.

57. Besson, C., E.E. Finney, and R.G. Finke, A mechanism for transition-metal nanoparticle self-assembly. Journal of the American Chemical Society, 2005. 127 (22): p. 8179-8184.

58. Watzky, M.A. and R.G. Finke, Transition metal nanocluster formation kinetic and mechanistic studies. A new mechanism when hydrogen is the reductant: Slow, continuous nucleation and fast autocatalytic surface growth. Journal of the American Chemical Society, 1997. 119 (43): p. 10382-10400.

59. Wild, D., Separation systems , in The Immunoassay Handbook, 2nd Ed , D. Wild, Editor. 2001, Nature Publishing Group: New York. p. 149-158.

60. Laksanasopin, T., C.D. Chin, H. Moore, et al. Microfluidic point-of-care diagnostics for resource-poor environments . in SPIE Proceedings . 2009.

61. Palik, E.D., Handbook of Optical Constants of Solids . 1985, New York: Academic Press.

62. Banoo, S., D. Bell, P. Bossuyt, et al., Evaluation of diagnostic tests for infectious diseases: general principles. Nat Rev Microbiol, 2006. 4(12 Suppl): p. S20-32.

63. Sia, S.K., V. Linder, B.A. Parviz, et al., An integrated approach to a portable and low-cost immunoassay for resource-poor settings. Angew Chem Int Ed Engl, 2004. 43 (4): p. 498-502.

64. Nam, J.M., C.S. Thaxton, and C.A. Mirkin, Nanoparticle-based bio-bar codes for the ultrasensitive detection of proteins. Science, 2003. 301 (5641): p. 1884-6.

125

65. Chin, C.D., T. Laksanasopin, Y.K. Cheung, et al., Microfluidics-based diagnostics of infectious diseases in the developing world. Nature Medicine, 2011, in the press.

66. Peeling, R.W., D. Mabey, A. Herring, et al., Why do we need quality-assured diagnostic tests for sexually transmitted infections? Nature Reviews Microbiology, 2006: p. S7-S19.

67. Beebe, D.J., G.A. Mensing, and G.M. Walker, Physics and applications of microfluidics in biology. Annual Review of Biomedical Engineering, 2002. 4: p. 261-286.

68. Deep, U.D., B.R. Petersen, and J. Meng, A Smart Microcontroller-Based Iridium Satellite- Communication Architecture for a Remote Renewable Energy Source. IEEE Transactions on Power Delivery, 2009. 24 (4): p. 1869 - 1875

69. Mulhern III, J. Approximate Desktop, Notebook, & Netbook Power Usage . 2010 6/28/2010 [cited 2011; Available from: http://www.upenn.edu/computing/provider/docs/hardware/powerusage.html .

70. Parsa, H., C.D. Chin, P. Mongkolwisetwara, et al., Effect of volume- and time-based constraints on capture of analytes in microfluidic heterogeneous immunoassays. Lab on a Chip, 2008. 8(12): p. 2062-2070.

71. Linder, V., E. Verpoorte, N.F. de Rooij, et al., Application of surface biopassivated disosable poly(dimethylsiloxane)/glass chips to a heterogeneous competitive human serum immunoglobulin G immunoassay with incorporated internal standard. Electrophoresis, 2002. 23 (5): p. 740-749.

72. Linder, V., E. Verpoorte, W. Thormann, et al., Surface biopassivation of replicated poly(dimethylsiloxane) microfluidic channels and application to heterogeneous immunoreaction with on-chip fluorescence detection. Analytical Chemistry, 2001. 73 (17): p. 4181-4189.

73. Sia, S.K., V. Linder, B.A. Parviz, et al., An integrated approach to a portable and low-cost immunoassay for resource-poor settings. Angew Chem Int Ed Engl., 2004. 43 (4): p. 498-502.

74. Linder, V., S.K. Sia, and G.M. Whitesides, Reagent-loaded cartridges for valveless and automated fluid delivery in microfluidic devices. Anal Chem, 2005. 77 (1): p. 64-71.

75. Wild, D., The Immunoassay Handbook . 3rd edition ed. 2005: Elsevier Science.

76. Miranda, A.E., M.C. Alves, R.L. Neto, et al., Seroprevalence of HIV, hepatitis B virus, and syphilis in women at their first visit to public antenatal clinics in Vitoria, Brazil. Sex Transm Dis, 2001. 28 (12): p. 710-3.

77. Beelaert, G., G. Vercauteren, K. Fransen, et al., Comparative evaluation of eight commercial enzyme linked immunosorbent assays and 14 simple assays for detection of antibodies to HIV. Journal of Virological Methods, 2002. 105 (2): p. 197-206.

78. Gatarayiha, J.P., A. Mukanyonga, E. Kayirangwa, et al., Rwanda Demographic and Health Survey 2005 . 2006, Institut National de la Statistique du Rwanda (INSR) and ORC Macro: Calverton, Maryland, U.S.A.

79. Svoronos, T., I.A. Jillson, and M.M. Nsabimana, TRACnet's absorption into the Rwandan HIV/AIDS response. International Journal of Healthcare Technology and Management, 2008. 9(5): p. 430-445.

126

80. Chin, C.D., V. Linder, and S.K. Sia, Lab-on-a-chip devices for global health: past studies and future opportunities. Lab Chip, 2007. 7(1): p. 41-57.

81. Martinez, A.W., S.T. Phillips, E. Carrilho, et al., Simple telemedicine for developing regions: Camera phones and paper-based microfluidic devices for real-time, off-site diagnosis. Analytical Chemistry, 2008. 80 (10): p. 3699-3707.

82. Klarkowski, D.B., J.M. Wazome, K.M. Lokuge, et al., The Evaluation of a Rapid In Situ HIV Confirmation Test in a Programme with a High Failure Rate of the WHO HIV Two-Test Diagnostic Algorithm. Plos One, 2009. 4(2): p. -.

83. Pavie, J., A. Rachline, B. Loze, et al., Sensitivity of Five Rapid HIV Tests on Oral Fluid or Finger-Stick Whole Blood: A Real-Time Comparison in a Healthcare Setting. Plos One, 2010. 5(7): p. -.

84. Whitesides, G.M., The origins and the future of microfluidics. Nature, 2006. 442 (7101): p. 368-73.

85. Malkin, R.A., Design of health care technologies for the developing world. Annual review of biomedical engineering, 2007. 9: p. 567-87.

86. Blaya, J.A., H.S.F. Fraser, and B. Holt, E-Health Technologies Show Promise In Developing Countries. Health Affairs, 2010. 29 (2): p. 244-251.

87. Ericsson, Ericsson Mobility Report: on the pulse of the networked society . 2014.

88. Fitzpatrick, A. Mozilla's $25 Smartphone Is a Brilliant Gamechanger. Time, 2014.

89. Ozcan, A., Mobile phones democratize and cultivate next-generation imaging, diagnostics and measurement tools. Lab on a Chip, 2014. 14 (17): p. 3187-3194.

90. Wei, Q., H. Qi, W. Luo, et al., Fluorescent imaging of single nanoparticles and viruses on a smart phone. ACS Nano, 2013. 7(10): p. 9147-55.

91. Zurovac, D., R.K. Sudoi, W.S. Akhwale, et al., The effect of mobile phone text-message reminders on Kenyan health workers' adherence to malaria treatment guidelines: a cluster randomised trial. Lancet, 2011. 378 (9793): p. 795-803.

92. Freifeld, C.C., R. Chunara, S.R. Mekaru, et al., Participatory epidemiology: use of mobile phones for community-based health reporting. PLoS Med, 2010. 7(12): p. e1000376.

93. Chin, C.D., Y.K. Cheung, T. Laksanasopin, et al., Mobile device for disease diagnosis and data tracking in resource-limited settings. Clin Chem, 2013. 59 (4): p. 629-40.

94. Oncescu, V., M. Mancuso, and D. Erickson, Cholesterol testing on a smartphone. Lab Chip, 2014. 14 (4): p. 759-63.

95. Erickson, D., D. O'Dell, L. Jiang, et al., Smartphone technology can be transformative to the deployment of lab-on-chip diagnostics. Lab on a Chip, 2014. 14 (17): p. 3159-3164.

96. Nemiroski, A., D.C. Christodouleas, J.W. Hennek, et al., Universal mobile electrochemical detector designed for use in resource-limited applications. Proceedings of the National Academy of Sciences, 2014. 111 (33): p. 201405679.

127

97. Lee, D.S., B.G. Jeon, C. Ihm, et al., A simple and smart telemedicine device for developing regions: a pocket-sized colorimetric reader. Lab Chip, 2011. 11 (1): p. 120-6.

98. Clifford, G.D. and D. Clifton, Wireless Technology in Disease Management and Medicine. Annual Review of Medicine, Vol 63, 2012. 63 : p. 479-492.

99. Kuo, Y.S., T. Schmid, and P. Dutta, Hijacking Power and Bandwidth from the Mobile Phone’s Audio Interface. Proc. Int. Symp. Low Power Electronics and Design, 2010.

100. Dimov, I.K., L. Basabe-Desmonts, J.L. Garcia-Cordero, et al., Stand-alone self-powered integrated microfluidic blood analysis system (SIMBAS). Lab on a Chip, 2011. 11 (5): p. 845-850.

101. Wang, W., Lyophilization and development of solid protein pharmaceuticals. International Journal of Pharmaceutics, 2000. 203(1-2): p. 1-60.

102. Hasenbank, M.S., T. Edwards, E. Fu, et al., Demonstration of multi-analyte patterning using piezoelectric inkjet printing of multiple layers. Anal Chim Acta, 2008. 611 (1): p. 80-8.

103. Chin, C.D., T. Laksanasopin, Y.K. Cheung, et al., Microfluidics-based diagnostics of infectious diseases in the developing world. Nature Medicine, 2011. 17 (8): p. 1015-U138.

104. Sena, A.C., B.L. White, and P.F. Sparling, Novel Treponema pallidum serologic tests: a paradigm shift in syphilis screening for the 21st century. Clin Infect Dis, 2010. 51 (6): p. 700-8.

105. Owusu-Edusei, K., Jr., T.L. Gift, and R.C. Ballard, Cost-effectiveness of a dual non- treponemal/treponemal syphilis point-of-care test to prevent adverse pregnancy outcomes in sub- Saharan Africa. Sex Transm Dis, 2011. 38 (11): p. 997-1003.

106. Castro, A.R., J. Esfandiari, S. Kumar, et al., Novel point-of-care test for simultaneous detection of nontreponemal and treponemal antibodies in patients with syphilis. J Clin Microbiol, 2010. 48 (12): p. 4615-9.

107. Plate, D.K. and H.I.V.T.E.W.G. Rapid, Evaluation and implementation of rapid HIV tests: the experience in 11 African countries. AIDS Res Hum Retroviruses, 2007. 23 (12): p. 1491-8.

108. Delaney, K.P., B.M. Branson, A. Uniyal, et al., Evaluation of the performance characteristics of 6 rapid HIV antibody tests. Clin Infect Dis, 2011. 52 (2): p. 257-63.

109. Kagulire, S.C., P. Opendi, P.D. Stamper, et al., Field evaluation of five rapid diagnostic tests for screening of HIV-1 infections in rural Rakai, Uganda. Int J STD AIDS, 2011. 22 (6): p. 308- 9.

110. Li, J., H.Y. Zheng, L.N. Wang, et al., Clinical evaluation of four recombinant Treponema pallidum antigen-based rapid diagnostic tests for syphilis. J Eur Acad Dermatol Venereol, 2009. 23 (6): p. 648-50.

111. Jafari, Y., R.W. Peeling, S. Shivkumar, et al., Are Treponema pallidum specific rapid and point-of-care tests for syphilis accurate enough for screening in resource limited settings? Evidence from a meta-analysis. PLoS One, 2013. 8(2): p. e54695.

112. Yin, Y.P., X.S. Chen, W.H. Wei, et al., A dual point-of-care test shows good performance in simultaneously detecting nontreponemal and treponemal antibodies in patients with syphilis: a multisite evaluation study in China. Clin Infect Dis, 2013. 56 (5): p. 659-65.

128

113. Damschroder, L.J., D.C. Aron, R.E. Keith, et al., Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science. Implementation Science, 2009. 4.

114. Brixen, E.B. Digital wireless and mics – it’s digital so why bother? [cited 2015 April 9]; Available from: http://www.dpamicrophones.dk/da/MikrofonUniversitet/Tech-Guide/Digital- wireless-and-microphones.aspx .

115. Chin, C.D., Y.K. Cheung, T. Laksanasopin, et al., Mobile Device for Disease Diagnosis and Data Tracking in Resource-Limited Settings. Clinical Chemistry, 2013. 59 (4): p. 629-640.

116. Castro, A.R., W.E. Morrill, W.A. Shaw, et al., Use of synthetic cardiolipin and lecithin in the antigen used by the Venereal Disease Research Laboratory Test for serodiagnosis of syphilis. Clinical and Diagnostic Laboratory Immunology, 2000. 7(4): p. 658-661.

117. Belisle, J.T., M.E. Brandt, J.D. Radolf, et al., Fatty-Acids of Treponema-Pallidum and Borrelia-Burgdorferi Lipoproteins. Journal of Bacteriology, 1994. 176 (8): p. 2151-2157.

118. Chin, C.D., Y.K. Cheung, T. Laksanasopin, et al., Mobile device for disease diagnosis and data tracking in resource-limited settings. Clinical chemistry, 2013. 59 (4): p. 629-640.

119. Cira, N.J., J.Y. Ho, M.E. Dueck, et al., A self-loading microfluidic device for determining the minimum inhibitory concentration of antibiotics. Lab on a Chip, 2012. 12 (6): p. 1052-1059.

120. Kuo, Y.-S., S. Verma, T. Schmid, et al. Hijacking power and bandwidth from the mobile phone's audio interface . in Proceedings of the First ACM Symposium on Computing for Development . 2010. ACM.

121. Nirjon, S., R.F. Dickerson, Q. Li, et al. Musicalheart: A hearty way of listening to music . in Proceedings of the 10th ACM Conference on Embedded Network Sensor Systems . 2012. ACM.

122. Meredith, S.E., A. Robinson, P. Erb, et al., A mobile-phone-based breath carbon monoxide meter to detect cigarette smoking. Nicotine & Tobacco Research, 2014. 16 (6): p. 766- 773.

123. Depari, A., A. Flammini, E. Sisinni, et al. A wearable smartphone-based system for electrocardiogram acquisition . in Medical Measurements and Applications (MeMeA), 2014 IEEE International Symposium on . 2014. IEEE.

124. Nemiroski, A., D.C. Christodouleas, J.W. Hennek, et al., Universal mobile electrochemical detector designed for use in resource-limited applications. Proceedings of the National Academy of Sciences, 2014: p. 201405679.

125. Anderson, G. and M. Scott, Determination of product shelf life and activation energy for five drugs of abuse. Clin Chem, 1991. 37 (3): p. 398-402.

126. Linder, V., S.K. Sia, and G.M. Whitesides, Reagent-loaded cartridges for valveless and automated fluid delivery in microfluidic devices. Analytical Chemistry, 2005. 77 (1): p. 64-71.

127. He, J. and S. Parker, Qualitative immunoassays . The Immunoassay Handbook, 4th Edition: Therory and applications of ligand binding and related techniques, ed. D. Wild. 2013: Elsevier Science.

128. Mabey, D.C., K.A. Sollis, H.A. Kelly, et al., Point-of-care tests to strengthen health systems and save newborn lives: the case of syphilis. PLoS Med, 2012. 9(6): p. e1001233.

129

129. Centers for Disease Control and Prevention, Notice to readers: Recommendations regarding screening of refugee children for treponemal infection , M.M.M.W. Rep, Editor. 2005: MMWR Morb Mortal Wkly Rep. p. 933-934.

130. Rational Pharmaceutical Management Plus Program, HIV test kits listed in the USAID Source and Origin Waiver: procurement information document. 5th Ed. Edited by A. Johnson. , R.P.M.P. Program, Editor. 2009, U.S. Agency for International Development by the Rational Pharmaceutical Management Plus Program: Arlington, VA.

131. Binnicker, M.J., J.D. Yao, and F.R. Cockerill, 3rd, Non-treponemal serologic tests: a supplemental, not confirmatory testing approach. Clin Infect Dis, 2011. 52 (2): p. 274-5; author reply 275-6.

132. Aledort, J.E., A. Ronald, M.E. Rafael, et al., Reducing the burden of sexually transmitted infections in resource-limited settings: the role of improved diagnostics. Nature, 2006. 444 Suppl 1: p. 59-72.

133. Steingart, K.R., L.L. Flores, N. Dendukuri, et al., Commercial Serological Tests for the Diagnosis of Active Pulmonary and Extrapulmonary Tuberculosis: An Updated Systematic Review and Meta-Analysis. Plos Medicine, 2011. 8(8).

134. Dowdy, D.W., K.R. Steingart, and M. Pai, Serological Testing Versus Other Strategies for Diagnosis of Active Tuberculosis in India: A Cost-Effectiveness Analysis. Plos Medicine, 2011. 8(8).

135. Steingart, K.R., S. Jotblad, K. Robsky, et al., Higher-dose rifampin for the treatment of pulmonary tuberculosis: a systematic review. International Journal of Tuberculosis and Lung Disease, 2011. 15 (3): p. 305-316.

136. World Health Organization, Global Tuberculosis Report 2014 . 2014: Geneva, Switzerland

137. McNerney, R. and P. Daley, Towards a point-of-care test for active tuberculosis: obstacles and opportunities. Nature Reviews Microbiology, 2011. 9(3): p. 204-213.

138. Parsons, L.M., A. Somoskovi, C. Gutierrez, et al., Laboratory diagnosis of tuberculosis in resource-poor countries: challenges and opportunities. Clin Microbiol Rev. 24 (2): p. 314-50.

139. [Anon] and Cdc, Updated Guidelines for the Use of Nucleic Acid Amplification Tests in the Diagnosis of Tuberculosis (Reprinted from MMWR, vol 58, pg 7-10, 2009). Jama-Journal of the American Medical Association, 2009. 301 (10): p. 1014-1016.

140. WHO, Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF system . Policy statement . 2011, World Health Organization: Geneva.

141. Evans, C.A., GeneXpert-A Game-Changer for Tuberculosis Control? Plos Medicine, 2011. 8(7).

142. Menzies, N.A., T. Cohen, H.H. Lin, et al., Population health impact and cost-effectiveness of tuberculosis diagnosis with Xpert MTB/RIF: a dynamic simulation and economic evaluation. PLoS Med, 2012. 9(11): p. e1001347.

143. Creswell, J., A.J. Codlin, E. Andre, et al., Results from early programmatic implementation of Xpert MTB/RIF testing in nine countries. BMC Infect Dis, 2014. 14 : p. 2.

130

144. Small, P.M. and M. Pai, Tuberculosis diagnosis--time for a game change. N Engl J Med, 2010. 363 (11): p. 1070-1.

145. Dineva, M.A., L. MahiLum-Tapay, and H. Lee, Sample preparation: a challenge in the development of point-of-care nucleic acid-based assays for resource-limited settings. Analyst, 2007. 132 (12): p. 1193-9.

146. Chen, L., A. Manz, and P.J. Day, Total nucleic acid analysis integrated on microfluidic devices. Lab Chip, 2007. 7(11): p. 1413-23.

147. Lui, C., N.C. Cady, and C.A. Batt, Nucleic Acid-based Detection of Bacterial Pathogens Using Integrated Microfluidic Platform Systems. Sensors, 2009. 9(5): p. 3713-3744.

148. Ausubel, e.a., Short Protocols in Molecular Biology: A Compendium of Methods from Current Protocols in Molecular Biology . 1992, New York: John Wiley & Sons.

149. Crevillen, A.G., M. Hervas, M.A. Lopez, et al., Real sample analysis on microfluidic devices. Talanta, 2007. 74 (3): p. 342-357.

150. Lee, T.M.H. and I.M. Hsing, DNA-based bioanalytical microsystems for handheld device applications. Analytica Chimica Acta, 2006. 556 (1): p. 26-37.

151. Zhang, C.S., J.L. Xu, W.L. Ma, et al., PCR microfluidic devices for DNA amplification. Biotechnology Advances, 2006. 24 (3): p. 243-284.

152. Cady, N.C., S. Stelick, M.V. Kunnavakkam, et al., Real-time PCR detection of Listeria monocytogenes using an integrated microfluidics platform. Sensors and Actuators B-Chemical, 2005. 107 (1): p. 332-341.

153. Wang, F. and M.A. Burns, Performance of nanoliter-sized droplet-based microfluidic PCR. Biomed Microdevices, 2009.

154. Holland, C.A. and F.L. Kiechle, Point-of-care molecular diagnostic systems - past, present and future. Current Opinion in Microbiology, 2005. 8(5): p. 504-509.

155. Braun, D., PCR by thermal convection. Modern Physics Letters B, 2004. 18 (16): p. 775- 784.

156. Liao, C.S., G.B. Lee, J.J. Wu, et al., Micromachined polymerase chain reaction system for multiple DNA amplification of upper respiratory tract infectious diseases. Biosensors & Bioelectronics, 2005. 20 (7): p. 1341-1348.

157. Lee, D.S., S.H. Park, H.S. Yang, et al., Bulk-micromachined submicroliter-volume PCR chip with very rapid thermal response and low power consumption. Lab on a Chip, 2004. 4(4): p. 401-407.

158. Van Ness, J., L.K. Van Ness, and D.J. Galas, Isothermal reactions for the amplification of oligonucleotides. Proc Natl Acad Sci U S A, 2003. 100 (8): p. 4504-9.

159. Tan, E., J. Wong, D. Nguyen, et al., Isothermal DNA amplification coupled with DNA nanosphere-based colorimetric detection. Anal Chem, 2005. 77 (24): p. 7984-92.

160. Mabey, D., R.W. Peeling, R. Ballard, et al., Prospective, multi-centre clinic-based evaluation of four rapid diagnostic tests for syphilis. Sexually Transmitted Infections, 2006. 82 : p. V13-V16.

131

161. Boehme, C.C., P. Nabeta, D. Hillemann, et al., Rapid Molecular Detection of Tuberculosis and Rifampin Resistance. New England Journal of Medicine, 2010. 363 (11): p. 1005-1015.

162. WHO, Roadmap for rolling out Xpert MTB/RIF for rapid diagnosis of TB and MDR-TB . 2010, World Health Organization: Geneva.

163. Taton, T.A., C.A. Mirkin, and R.L. Letsinger, Scanometric DNA array detection with nanoparticle probes. Science, 2000. 289 (5485): p. 1757-1760.

164. Liong, M., A.N. Hoang, J. Chung, et al., Magnetic barcode assay for genetic detection of pathogens. Nature Communications, 2013. 4.

165. Ferguson, B.S., S.F. Buchsbaum, T.T. Wu, et al., Genetic Analysis of H1N1 Influenza Virus from Throat Swab Samples in a Microfluidic System for Point-of-Care Diagnostics. Journal of the American Chemical Society, 2011. 133 (23): p. 9129-9135.

166. Leslie, D.C., J.Y. Li, B.C. Strachan, et al., New Detection Modality for Label-Free Quantification of DNA in Biological Samples via Superparamagnetic Bead Aggregation. Journal of the American Chemical Society, 2012. 134 (12): p. 5689-5696.

167. Bissonnette, L. and M.G. Bergeron, Infectious Disease Management through Point-of- Care Personalized Medicine Molecular Diagnostic Technologies. Journal of Personalized Medicine, 2012. 2(2): p. 50-70.

168. Reyrat, J.M. and D. Kahn, Mycobacterium smegmatis: an absurd model for tuberculosis? Trends in Microbiology, 2001. 9(10): p. 472-473.

169. Hett, E.C. and E.J. Rubin, Bacterial growth and cell division: a mycobacterial perspective. Microbiology and Molecular Biology Reviews, 2008. 72 (1): p. 126-+.

170. Govindarajan, A.V., S. Ramachandran, G.D. Vigil, et al., A low cost point-of-care viscous sample preparation device for molecular diagnosis in the developing world; an example of microfluidic origami. Lab on a Chip, 2012. 12 (1): p. 174-181.

171. Yajko, D.M., J.J. Madej, M.V. Lancaster, et al., Colorimetric Method for Determining Mics of Antimicrobial Agents for Mycobacterium-Tuberculosis. Journal of Clinical Microbiology, 1995. 33 (9): p. 2324-2327.

172. Gevorkian, G., E. Segura, G. Acero, et al., Peptide mimotopes of Mycobacterium tuberculosis carbohydrate immunodeterminants. Biochemical Journal, 2005. 387 : p. 411-417.

173. Dhiman, R.K., P. Dinadayala, G.J. Ryan, et al., Lipoarabinomannan Localization and Abundance during Growth of Mycobacterium smegmatis. Journal of Bacteriology, 2011. 193 (20): p. 5802-5809.

174. Wilson, S., A. Lane, R. Rosedale, et al., Concentration of Mycobacterium tuberculosis from sputum using ligand-coated magnetic beads. International Journal of Tuberculosis and Lung Disease, 2010. 14 (9): p. 1164-1168.

175. Mukherjee, R. and D. Chatterji, Evaluation of the role of sigma B in Mycobacterium smegmatis. Biochemical and Biophysical Research Communications, 2005. 338 (2): p. 964-972.

176. Volokhov, D., V. Chizhikov, S. Denkin, et al., Molecular Detection of Drug-Resistant Mycobacterium tuberculosis with a Scanning-Frame Oligonucleotide Microarray , in Mycobacteria Protocols , T. Parish and A.C. Brown, Editors. 2009, Humana Press. p. 395-417.

132

177. Jayaram, L., K. Parameswaran, M.R. Sears, et al., Induced sputum cell counts: their usefulness in clinical practice. European Respiratory Journal, 2000. 16 (1): p. 150-158.

178. Liu, D.Y., G.T. Liang, Q. Zhang, et al., Detection of Mycobacterium tuberculosis Using a Capillary-Array Microsystem with Integrated DNA Extraction, Loop-Mediated Isothermal Amplification, and Fluorescence Detection. Analytical Chemistry, 2013. 85 (9): p. 4698-4704.

133