Enabling sweat-based biosensors:
Solving the problem of low biomarker concentration in sweat
A dissertation submitted to the
Graduate School
of the University of Cincinnati
in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
in the Department of Biomedical Engineering
of the College of Engineering & Applied Science
by
Andrew J. Jajack
B.S., Biology, Wittenberg University, 2014
Committee Chairs: Jason C. Heikenfeld, Ph.D. and Chia-Ying Lin, Ph.D.
Abstract
Non-invasive, sweat biosensing will enable the development of an entirely new class of wearable devices capable of assessing health on a minute-to-minute basis. Every aspect of healthcare stands to benefit: prevention (activity tracking, stress-level monitoring, over-exertion alerting, dehydration warning), diagnosis (early-detection, new diagnostic techniques), and management (glucose tracking, drug-dose monitoring). Currently, blood is the gold standard for measuring the level of most biomarkers in the body. Unlike blood, sweat can be measured outside of the body with little inconvenience. While some biomarkers are produced in the sweat gland itself, most are produced elsewhere and must diffuse into sweat. These biomarkers come directly from blood or interstitial fluid which surrounds the sweat gland. However, a two-cell thick epithelium acts as barrier and dilutes most biomarkers in sweat. As a result, many biomarkers that would be useful to monitor are diluted in sweat to concentrations below what can be detected by current biosensors. This is a core challenge that must be overcome before the advantages of sweat biosensing can be fully realized.
The objective of this dissertation is to develop methods of concentrating biomarkers in sweat to bring them into range of available biosensors. This dissertation will encompass both the physiological understanding of how biomarkers enter sweat as well as two strategies for increasing concentration: technological and biological. The technological strategy involves a novel microfluidic-based biofluid preconcentration device. The biological strategy takes advantage of paracellular permeability enhancers in combination with reverse iontophoresis to increase biomarker flux into sweat, thereby increasing the concentration in sweat. Increasing the concentration of biomarkers in sweat is expected to improve the detection of previously hard-to-detect biomarkers, making sweat biosensing a more viable option for health monitoring.
ii Copyright Page
iii Acknowledgements
To my committee—thank you for your thoughtful feedback and direction, challenging me to grow as a researcher.
To my friends and colleagues of the Novel Device Lab—I have greatly appreciated working with each one of you. Our supportive, collaborative culture could be summed up as simply: work hard, play harder. Thanks for making the journey a blast.
To my advisor and mentor, Dr. Heikenfeld—you have taught me to think big, to challenge myself, and to trust my gut as I plunge into new adventures. Thanks for believing in me.
To my friends and family—thank you for helping me get to this point, supporting me as I grow, and keeping me grounded by reminding me of what matters most.
To my loving wife, Doreen—my rock, this journey would not have been possible without your love and support, and so it is only fitting that I dedicate this work to you.
Thank you, I love you, and I look forward to our next adventure with you by my side.
iv Table of Contents
Abstract...... ii Copyright Page ...... iii Acknowledgements ...... iv Table of Contents ...... v List of Figures ...... vii List of Tables ...... ix List of Acronyms ...... x Chapter 1: Introduction ...... 1 1.1 Background ...... 1 1.2 Sweat as an information-rich biofluid ...... 1 1.3 Catalytic and affinity biosensors ...... 2 1.4 Research aims...... 4 Chapter 2: Understand how biomarkers enter sweat and describe sweat-rate dependence of biomarkers ...... 10 2.1 Introduction ...... 10 2.2 The epidermis as an information barrier ...... 10 2.2.1 Epidermal structure ...... 11 2.2.2 Chemical impedance ...... 13 2.2.3 Chemical contamination ...... 13 2.3 Biofluid partitioning ...... 15 2.3.1 From blood to ISF ...... 15 2.4 Barrier to biomarker entry into sweat ...... 17 2.5 Experimental methods ...... 19 2.5.1 Custom testing device ...... 19 2.5.2 Sweat generation ...... 20 2.5.3 Avoiding sweat contamination ...... 23 2.5.4 Sweat collection ...... 24 2.5.5 Sweat glucose ...... 25 2.5.6 Blood glucose ...... 25 2.5.7 Sweat rate ...... 25 2.5.8 Glucose flux under normal conditions ...... 26 2.5.9 IRB protocol ...... 27 2.6 Results and discussion ...... 27 2.7 Conclusions ...... 28 Chapter 3: Continuously concentrate sweat samples within real-time device ...... 30 3.1 Introduction ...... 30 3.2 Experimental methods ...... 34 3.2.1 Membrane and draw molecule optimization ...... 34
v 3.2.2 Preconcentration device construction, characterizing, and modeling ...... 39 3.3 Results and discussion ...... 43 3.3.1 Membrane and draw molecule optimization ...... 43 3.3.2 Preconcentration device ...... 46 3.4 Conclusions ...... 50 Chapter 4: Increase concentration of target biomarkers in sweat using chemical permeability enhancement and reverse iontophoresis...... 51 4.1 Introduction ...... 51 4.2 RI likely does not induce electroporation under study conditions ...... 54 4.3 Experimental methods ...... 56 4.3.1 Glucose flux under modified conditions ...... 56 4.3.2 Statistical analysis ...... 56 4.3.3 Safety of chemical permeability enhancer ...... 56 4.3.4 pH considerations ...... 57 4.3.5 IRB protocol ...... 58 4.4 Results and discussion ...... 58 4.5 Conclusions ...... 60 Chapter 5: Conclusions ...... 62 5.1 Summary ...... 62 5.2 Applications ...... 63 5.3 Future work ...... 65 5.4 Conclusions ...... 67 References ...... 68
vi List of Figures
Figure 1.1. Cross-sectional and top-down illustration of the secretory coil and dermal duct. Figure 1.2. Example of aptamer-based biosensor folding in response to binding of a target. Figure 1.3. Possible routes from apical to basolateral sides of an epithelium. Figure 1.4. Cross section of the membranes of two neighboring cells. Figure 1.5. Diagram of continuous, real-time sweat concentrating device. Figure 1.6. Two citric acid molecules sequestering calcium ions. Figure 1.7. Electric double-layer formed by negative surface charges.
Figure 2.1. Diagrammatic cross-section of human skin. Figure 2.2. Morphology and histology of the sweat gland barrier for biomarker access. Figure 2.3. Normal modulation of the tight junctional strands allows shuttling of biomarkers. Figure 2.4. General biomarker routes of entry. Figure 2.5. Biomarker routes of entry based on their molecular properties. Figure 2.6. Custom two-part testing device with base and multiple twist-lock attachments. Figure 2.7. Demonstration cystic fibrosis diagnostic setup of the Wescor Nanoduct. Figure 2.8. Methods of natural and artificial sweat stimulation. Figure 2.9. Flow diagram of sweat stimulation. Figure 2.10. Flow diagram of sweat collection and analysis. Figure 2.11. Flow diagram of sweat rate measurements. Figure 2.12. Block diagram of setup for determining glucose flux under normal conditions. Figure 2.13. Sweat glucose concentration and sweat rates measured at various time points.
Figure 3.1. Operating principle and components of the microfluidic preconcentration device. Figure 3.2. Forward osmosis performance of various membranes and draw molecules. Figure 3.3. Schematic representation of dilutive and concentrative ICP. Figure 3.4. Concentrations of NaCl at discrete positions throughout straight-channel. Figure 3.5. Concentrations at various input flowrates for two biologically-relevant analytes. Figure 3.6. In-silico model predictions for required membrane surface area and time.
Figure 4.1. Calcium sequestration decreases trans-epithelial resistance. Figure 4.2. Calcium depletion induces internalization of adheren and tight junctional proteins. Figure 4.3. Conceptual model for glucose flux under various conditions. Figure 4.4. Molecular model of electroporation.
vii
Figure 4.5. Block diagram of methods for determining glucose flux under modified conditions. Figure 4.6. Total glucose flux measured before, during, and after treatments.
Figure 5.1. Examples of sweat biosensing application demos in the literature. Figure 5.2. Example of a preconcentrated point of care pregnancy test. Figure 5.3. Operating principle of the next generation of the preconcentration device.
viii List of Tables
Table 1.1. List of available aptamer equilibrium constants for biologically-relevant biomarkers.
Table 2.1. Evidence of contamination in sweat samples collected under various conditions.
ix List of Acronyms
Bovine serum albumin (BSA)
Cosmetic Ingredient Review (CIR)
Ethylenediaminetetraacetic acid (EDTA)
External concentration polarization (ECP)
Internal concentration polarization (ICP)
Interstitial fluid (ISF)
Molecular weight cut-off (MWCO)
Phosphate-buffered saline (PBS)
Transepidermal water loss (TEWL)
x
Chapter 1: Introduction
1.1 Background
The ways in which we approach wellness are changing—reactionary methods are being replaced with preventative measures, medicine is becoming personalized, and consumers are seeking more ways to monitor their day-to-day health. It is no surprise that the market for wearable biosensors has taken off in recent years. However, the sensors of today—the Fitbits and Apple Watches—barely scratch the surface with the data that they collect. These devices rely on decades-old optical- and electrical-based sensors and are largely limited to measuring pulse and tracking movement. While useful for some applications like fitness tracking, this level of detail is simply inadequate to meet the goal of providing truly useful and actionable health information.
Today, doctors run blood work to inform medical decisions and diabetics still rely on finger-prick tests to get at blood glucose. Why? Because, blood is the gold standard for measuring our biochemistry. Our biochemistry gives information about normal biological processes, pathogenic processes, and even pharmacologic responses to a therapeutic intervention1. The next generation of wearable devices must be able to continuously measure biochemistry in real-time.
However, continuous sampling is a problem for biofluids such as blood, which requires invasive, needle-based draws at discrete time points or urine, which requires a catheter prone to infection2. Sweat, on the other hand, can be measured outside of the body with little inconvenience unlike for tears or saliva and stimulated on-demand using cholinergic drugs3.
Sweat also contains far fewer digestive enzymes that would degrade biomarkers compared to tears or salvia4, making it an ideal biofluid to use for non-invasive wearable devices.
1.2 Sweat as an information-rich biofluid
But, isn’t sweat just salty water; how could it provide the same level of information as blood?
1 A few biomarkers are produced in the eccrine sweat gland itself (such as lactate); however, most are produced elsewhere in the body and must diffuse into sweat4. Sweat is generated in the secretory coil of the sweat gland. The secretory coil is heavily vascularized and is surrounded by a bed of capillaries feeding the gland. The purpose of capillaries is to exchange nutrients and signaling molecules and to remove waste, and so they are somewhat leaky by design5. This allows biomarkers to pass from blood into the fluid surrounding the extracellular space, interstitial fluid (ISF). Because of the leakiness of capillaries, ISF has similar concentrations of biomarkers compared to blood after some amount of time. For example, ISF glucose concentrations match that of blood after about 15 minutes.6,7 The sweat generated in the secretory coil must pass through a dermal duct before reaching the surface of the skin. The duct is a long tube about 2 mm long enclosed by a concentric ring of luminal cells surrounded by a ring of basal cells. These cells are arranged in a brick-and-mortar configuration (Fig. 1.1)8,9 and make up the sweat gland epithelium. Although surrounded by biomarker-rich blood and ISF, sweat is separated from these fluids by the sweat gland epithelium, which acts as a filter for biomarker entry into sweat, diluting the concentration of biomarkers of interest. For example, proteins are ~1000x more dilute in sweat compared to blood; glucose is ~100x more dilute4.
This makes measuring these biomarkers with currently available biosensors challenging.
1.3 Catalytic and affinity biosensors
There are two main types of biosensors: catalytic and affinity biosensors2.
Catalytic biosensors convert a biomarker into some other molecule using an enzyme. In these sensors, the biomarker is ‘consumed’ by the sensor. For example, glucose oxidase sensors couple the oxidation of glucose to gluconolactone with the reduction of a cofactor. The subsequent oxidation of the cofactor can be further coupled to an electron acceptor like oxygen, a mediator, or directly transferred to an electrode. The movement of electrons as a result reproduces a current that is measured amperometrically10. Catalytic sensors depend on mass
2 transfer and require the sample flow rate to be known to determine a concentration value. In addition, background current competes with the signal at low concentrations and practically limits these types of electrochemical biosensors to mM concentrations, but most biomarkers of interest are found at much lower concentrations in sweat4.
Figure 1.1. Cross-sectional and top-down illustration of the (a) secretory coil and (b) dermal duct.9
Affinity biosensors rely on the binding of biomarkers to receptors on the sensor to trigger a measurable response. Receptors can be antibodies or nucleic acids in the case of aptamer- based sensors. For example, in aptamer-based sensors, single-stranded DNA or RNA molecules bind to target biomarkers and undergo a conformational change to bring an attached redox couple closer to a gold-plated electrode2,4 (Fig. 1.2). Since these sensors are based on binding affinities, they are only effective within an order of magnitude above or below their equilibrium constant. Developing suitable aptamer sensors for a target biomarkers is an
Figure 1.2. Example of aptamer-based biosensor folding in response to binding of a target molecule.
3 expensive process, and so making sweat samples work with existing sensors would be advantageous. However, as seen in the following table, there is a mismatch between the concentrations of biomarkers of interest and the effective ranges for available aptamer-based sensors (Table 1.1).
Table 1.1. List of available aptamer equilibrium constants for biologically-relevant biomarkers.
As a result, many biomarkers that would be useful to monitor are diluted in sweat to concentrations below what can be detected by current biosensors. This is a core challenge that must be overcome before the advantages of sweat biosensing can be fully realized. The objective of this dissertation is to develop methods of concentrating biomarkers in sweat to bring them into range of available biosensors. This dissertation will encompass both the physiological understanding of how biomarkers enter sweat (Chapter 2) as well as two strategies for increasing concentration (Chapters 3 and 4). Increasing the concentration of biomarkers in sweat is expected to improve the detection of previously hard-to-detect biomarkers, making sweat biosensing a more viable option for health monitoring.
1.4 Research aims
Aim 1. Understand how biomarkers enter sweat and describe sweat-rate dependence of biomarkers. To understand how biomarkers diffuse into sweat, two routes entry must be
4 considered (Fig. 1.3)—transcellular (through the cells) or paracellular (between the cells). The
lipid bilayer forms a barrier for transcellular transport, while tight junctions form the barrier for
paracellular transport. Hydrophobic molecules such as cortisol and small, uncharged, polar
molecules such as urea and ethanol can easily pass through the lipid bilayer to enter sweat.
However, large, uncharged, polar molecules such as glucose and charged molecules including
ions or even proteins will have difficulty passing into sweat through the cells of the sweat gland
epithelium. These molecules will instead need to move through the space between the cells.
The cells are held together with cell-to-cell junctions that allow them to communicate with each
other (gap junctions), anchor to one another (adherens junctions), or partition fluids (tight
junctions). The space between the cells is 10’s of nanometers wide, leaving plenty of room for
biomarkers to pass11. However, tight junctions pinch neighboring cell membranes together
forming a seal that makes it difficult for biomarkers to pass through (Fig. 1.4). This explains why
most biomarkers are dilute in sweat.
Based on the ways biomarkers passively diffuse into sweat, the rate they do so is likely
independent of sweat rate. Therefore, it is hypothesized that as sweat rate decreases, less
fluid is generated in the sweat gland, and so the biomarkers that passively diffuse into sweat will
be less diluted. This means that the sweat rate must be considered when measuring biomarkers
Figure 1.3. Possible routes from apical to basolateral sides of an epithelium.
5 Figure 1.4. Cross section of the membranes of two neighboring cells showing the seal that tight junctions form between apical and basolateral sides of epithelial tissues. and could explain the large differences in biomarker concentrations found in the literature. In
Aim 1 (Chapter 2), the sweat-rate dependence of glucose will be examined. This is critical to understand first since it will help ensure that future experiments consider this important factor that could naturally alter biomarker concentrations.
Aim 2. Continuously concentrate sweat samples within real-time device. To accomplish the main objective of this dissertation, both Aims 2 and 3 (Chapters 3 and 4) will attempt to increase the concentration of biomarkers in sweat. Specifically, this aim will approach the problem starting once sweat reaches the surface of the skin. By removing water from the sweat sample, the concentration of biomarkers in sweat will increase. This aim will take advantage of advances in microfluidics and the principal of osmosis to develop a device able to continuously concentrate sweat in real-time. If a high osmotic draw solution is placed on one side of a membrane and a biofluid on the other, water will flow from the biofluid to the draw solution (Fig.
1.5). Semipermeable membranes are readily available with many different pore sizes and molecular weight cut-offs (MWCO). Unlike a simple wicking mechanism, water will only flow until the concentration of the biofluid equals that of the draw solution, preventing over concentration of the sample.
6 Figure 1.5. Diagram of continuous, real-time sweat concentrating device. Water from the sample is drawn into higher osmolarity draw solution in the pouch, concentrating the sample before sensing.
Aim 3. Increase concentration of target biomarkers in sweat using chemical permeability enhancement and reverse iontophoresis. This aim takes a biological approach to concentrating biomarkers and will attempt to increase the rate of biomarker flux into sweat by increasing the permeability of the sweat gland epithelium and actively flowing in biomarker-rich
ISF and blood.
Increasing permeability with chemical permeability enhancers. Until recently, tight junctions have been thought of as static junctions with the sole purpose of sealing and partitioning the apical and basal lateral portions of cells. However, it is now known that tight junctions are dynamic structures that are constantly remodeling12,13. This remodeling process is sensitive to outside signals such as low extracellular calcium14–16. Chelators such as citric acid and ethylenediaminetetraacetic acid (EDTA) have been used to modulate the paracellular permeability of epithelia to improve intestinal drug delivery13. Some of the same chemical agents that are used to aid in drug delivery may also improve the extraction of biomarkers in other epithelia such as the eccrine ductal epithelium.
Chelators bind metal ions and can be used to sequester extracellular calcium in the sweat lumen, causing a local drop in calcium concentration (Fig. 1.6). This drop in calcium
7 Figure 1.6. Two citric acid molecules sequestering calcium ions. concentration triggers an increase in clathrin-mediated endocytosis of tight junction proteins17.
EDTA and citric acid are two widely used chelating agents found in topical cosmetic formulations. Both EDTA and citric acid bind to calcium in a roughly 1:1 molar ratio depending on the pH of the solution. Since the extracellular calcium concentration within the sweat lumen is roughly 1-2 mM, only around 1-2 mM of either EDTA or citric acid would be needed to sufficiently sequester the extracellular calcium ions due to the tight affinity of chelators for divalent cations14,18. Both EDTA and citric acid (and its citrate derivatives) have been reviewed by the Cosmetic Ingredient Review (CIR) and are used in thousands of formulations19,20.
Active fluid flow with electroosmosis induced by reverse iontophoresis. By combining an increase in permeability with the principle of reverse iontophoresis induced electroosmosis, the movement of ISF and biomarkers into the sweat can be magnified21,22. The electrical double layer needed for electroosmosis is formed by the negatively-charged cell surface (due to the carboxylic acid moieties within glycans [including glycosaminoglycans] and proteins [including collagen]) interacting with the positively-charged ions in the extracellular matrix (including sodium and potassium)23. When an external electric field is applied, cations in both the double layer and bulk solution will move toward the cathode, an effect known as reverse
8 Figure 1.7. Electric double-layer formed by negative surface charges attracting cations in solution. As the cation sheath migrates toward the negative electrode, the bulk fluid follows. (Credit: Microfluidics/Nanofluidics Laboratory,
Cornell University) iontophoresis22,24,25. Electroosmosis describes the phenomenon in which the movement of the cations within the double layer causes the bulk solution to move toward the cathode21,26 (Fig.
1.7). Reverse iontophoresis and electroosmosis can be used to drive biomarkers out of blood/ISF into sweat glands and then finally out to the surface for measurement by sensors.
Therefore, it is hypothesized that sweat collected from glands treated with permeability enhancers and with active fluid flow will contain higher concentrations of target biomarkers.
Together, these three aims will tackle the problem of low biomarker concentration in sweat, a core challenge impeding the development of sweat biosensors. Non-invasive sweat sensing is poised to revolutionize the way we monitor health. This dissertation is expected to make it possible to measure more biomarkers of interest in sweat, helping the next generation of wearable devices to provide truly useful and actionable health information.
9 Chapter 2: Understand how biomarkers enter sweat and describe sweat-rate dependence of biomarkers
This chapter includes adapted text and figures from the following articles:
A. Jajack, M. Brothers, G. Kasting, J. Heikenfeld. Enhancing biomarker flux into sweat. PLoS
One. Submitted.
J. Heikenfeld, A. Jajack, B. Feldmen, S. Granger, S. Gaitonde, G. Begtrup, B. Katchman.
Peripheral Biochemical Monitoring: a Critical Review of Analyte Access for Interstitial fluid,
Saliva, and Sweat. Nature Biotechnology. In progress.
J. Heikenfeld, A. Jajack, J. Rogers, P. Gutruf, L. Tian, T. Pan, R. Li, M. Khine, J. Kim, J. Wang,
J. Kim. Wearable Sensors: Modalities, Challenges, and Prospects. Lab Chip, 2018, 18, 217-
248.
2.1 Introduction
An understanding of the ways in which biomarkers enter sweat is critical to allow for the back calculation of systemic biomarker concentrations from measured sweat concentrations. Since wearable sweat-based biosensors are worn on the skin, the epidermis as an information barrier and contamination source must first be discussed. Next, the sources of biomarkers through the sweat gland will be covered. Finally, with this understanding comes the postulate that for certain biomarkers, a sweat rate-dependence exists due limitations in diffusion. This is examined experimentally and discussed within this chapter.
2.2 The epidermis as an information barrier
That the epidermis is an information barrier is hardly surprising, since it is the first line of defense in our immune system, and because it serves as barrier to loss of water and circulating nutrients and solutes in blood. The epidermis also protects underlying tissue from damaging ultra-violet light. Furthermore, the stratum corneum is dry and oily, and therefore electrically
10 resistive. The epidermis is also soft, stretchy, and slides over underlying organs, dampening the effects of mechanical forces inside the body. For all these reasons and more, the epidermis generally is more of an information barrier than it is an information source when it comes to wearable sensing. In this section, the epidermal structure is described in detail, including sources of chemical contamination which must be considered for wearable sweat-based biosensors.
2.2.1 Epidermal structure
The epidermis is a stratified squamous epithelium with each of the strata serving an important role (Fig. 2.1). The deepest layer, the stratum basale, forms a continuous sheet of cells (largely keratinocytes, but also melanocytes, Langerhans cells, and Merkel cells) that separate the dermis from the epidermis. The highly proliferative keratinocytes in this layer divide and migrate upward to form the stratum spinosum. The keratinocytes of this layer actively synthesize fibrillar proteins that serve as the precursor to desmosomes, a type of cell-to-cell adhesion structure important for tissues to resist high shear stresses. These keratinocytes mature to form the stratum granulosum, which is responsible for inducing cell dehydration then cell death, cross- linking keratin fibers, and releasing lamellar bodies to form the intercellular hydrophobic barrier of the stratum corneum27. The tight junctions between cells of the stratum granulosum further impede the flow of water and solutes between the viable epidermis and the stratum corneum.
Some areas of thick skin possess a stratum lucidum, a region of several additional layers of keratinocytes found between the stratum granulosum and the stratum corneum. The stratum corneum is held together by corneodesmosomes. Proteases degrade these junctions and eventually cause the dead cells at the surface to shed in a process called desquamation. The tight junctions of the stratum granulosum and the organized intercellular lipid lamellae of the stratum corneum form the epidermal barrier13. Skin appendages such as hair, sebaceous
11 Figure 2.1. Diagrammatic cross-section of human skin, including a zoomed in view of the epidermis. Adapted from
Blausen 2014106.
glands, and sweat glands provide a natural pathway through the stratum corneum barrier, but still have layers of surrounding live cells that separate the outside world from the inside of the body.
Epithelia like the epidermis are common in other areas and organs of the body where a barrier function is required. The oral mucosa (mouth lining) is made up of both keratinized and non- keratinized stratified squamous epithelia. Keratinized regions are found in the masticatory mucosa where abrasion is common such as the surface of the tongue, hard palate, and gingiva.
The lining mucosa is largely non-keratinized and lacks a stratum corneum. The corneal and conjunctiva epithelia of the eye are also examples of similar structures. However, this
12 dissertation focuses on the epidermis and sweat glands because of ergonomic reasons and the reasons outlined above regarding sweats benefits over other biofluids.
2.2.2 Chemical impedance
As noted in the previous section, the skin is by design a barrier to transport of chemicals. The superficial layers of the epidermis, which include the tight junctions of the stratum granulosum and the interlamellar hydrophobic barrier of the stratum corneum are the major contributors to chemical impedance of the epidermis. Disrupting this epidermal barrier is possible and has been extensively studied for transdermal drug delivery purposes. The barrier can be disrupted by mechanical methods such as microneedles28, tape-stripping which removes the stratum corneum29, sonophoresis30, electroporation and reverse iontophoresis22,31, and chemical methods such as permeability enhancers that increase paracellular pathways13. The effectiveness of all these methods, and/or determining the integrity of the epidermis, is often assessed by measuring a change in the transepidermal water loss (TEWL)32. Of these techniques, only the invasive methods that form an actual physical pore can allow access to analyte concentrations at their blood and ISF levels. For all non-invasive methods, even with skin-permeability enhancers, the chemical impedance of the skin remains very high.
2.2.3 Chemical contamination
Not only does the skin serve as a barrier to analytes, but it can also contaminate analyte concentrations when collecting samples such as sweat, ISF, and blood. For example, estimates of the density of bacteria found on the skin are as high as 10 billion/cm2 33. Bacteria can consume analytes such as energy sources like glucose and secrete analytes such as proteins or cellular waste products. These alterations of levels of analytes by the microflora pose a challenge for chemical biosensing applications. In addition, sweat minerals have been shown to accumulate in the superficial layers of the epidermis and possibly in the sweat duct itself prior to
13 Table 2.1. Evidence of contamination in initial sweat samples collected from skin into a bag with: true sweat level
based on dripping sweat collection and an oil layer on skin to block contamination; dripping sweat collection
without an oil layer on skin to block contamination; scraping sweat collection without an oil layer to block
contamination. cAMP is cyclic adenosine monophosphate. Skin was washed/rinsed/dried before collection.
Adapted from Boysen et al. 1984 and Peng et al. 201634,36.
Analyte M.W. (Da) Wash & true level Wash & drip collect Wash & scrape collect
calcium 40 ~0.25 mM + 150% + 500%
urea 60 ~4 mM + 40% + 150%
cAMP 329 ~0.2 nM + 200% + 650%
protein 10’s k ~25 mg/dL + 60% + 150%
sweating events34. It can be assumed that similar accumulation may occur with other analytes, including proteins. For example, simply washing the skin surface does not mitigate contamination, as shown in Table 2.1 where even small analytes (calcium) to large analytes
(proteins) exist at concentrations high enough cause significant errors in the concentrations measured in sweat34. These contaminants also can cause significant errors for blood or ISF samples when the sample volume is very small and a needle is used to puncture the skin for fluid extraction. Finally, the skin surface is constantly being coated with proteases which aid in the shedding of dead skin cells and a mixture of triglycerides, wax esters, squalene, and metabolites from sebaceous glands27,35.
Chemical contamination does not always have to be a problem. For example, in non-invasive sweat sensing applications, epidermal contaminants can be avoided by preventing sweat from contacting the epidermis by coating the skin with an occluding layer of petroleum jelly or oil34,36.
Furthermore, with the growing awareness of the linkages between the microbiome and health
14 status, measuring the microbe-induced concentrations of analytes on the skin could represent a significant opportunity in itself33.
2.3 Biofluid partitioning
Analytes in blood are separated from ISF and sweat by thin cell-based barriers. The morphology and composition of these barriers determines the ease and route of diffusion. Here, the barriers for biomarker diffusion into ISF, saliva, and sweat are discussed.
2.3.1 From blood to ISF
Sweat glands are bathed in ISF (Fig. 2.2); therefore, understanding the relationship between capillary blood and ISF is essential before sweat can be understood. Biomarkers in ISF come from blood, specifically through capillaries. Since capillaries function to deliver nutrients, remove waste, and exchange signaling molecules from blood with the surrounding tissues, they are somewhat leaky by design. There are three ways in which biomarkers can enter ISF: (1) passive/carrier-mediated diffusion through the plasma membrane, (2) diffusion through the
Figure 2.2. Morphology and histology of the partitioning barrier for biomarker access through ISF and sweat.
15 space between cells, or (3) vesicular transport through the cell, called transcytosis. Passive diffusion through the plasma membrane is only possible for small, uncharged biomarkers, such as cortisol. Carrier-mediated diffusion requires specific transmembrane proteins and only certain types of biomarkers, such as electrolytes, can navigate this route. Most biomarkers of interest, such as proteins, cannot traverse these routes and must either be shuttled through the endothelium via transport vesicles or diffuse through the paracellular pathway. Continuous capillaries, the most common, are made up of a single layer of endothelial cells that are joined by a flexible network of tight junctional strands that continuously break, reseal, and branch (Fig.
2.3). The opening and closing compartments between the strands allow biomarkers to pass through the paracellular pathway. In addition, these capillaries have a small diameter of only 5-
10 μm. It is speculated that the high surface area to volume ratio, capillary density, low flow rate, high pressure difference, and combined transcytotic and paracellular entry routes for large biomarkers contributes to the similarities between blood and ISF concentrations of biomarkers.
Figure 2.3. Normal modulation of the tight junctional strands allows shuttling of biomarkers, reproduced38.
16 2.3.2 From blood and ISF to sweat
Most biomarkers in sweat originate from blood or ISF. Only the eccrine sweat glands (thermal regulation) are described herein, because advantages for apocrine glands have not yet been identified which supersede their disadvantages in placement (axilla, groin) and their much more lipid- and bacteria-rich makeup33. Most of the gland is in the dermis, a largely acellular and porous collagen matrix filled with ISF. The sweat glands are highly vascularized in the secretory coil which has 5-40 µm inner diameter and 2-5 mm length and a 1-3 cell-thick epithelium. The dermal duct has a 10-20 µm inner diameter and ~2mm length and a two-layer stratified cuboidal epithelium. A 200-300 µm coiled duct spans the remaining epidermis and stratum corneum9. The high surface-area-to-volume ratio of the sweat gland is conducive for analyte exchange with blood and ISF.
2.4 Barrier to biomarker entry into sweat
Sweat is separated from the surrounding biomarker-rich blood and ISF by a one- to two-cell thick epithelium9. This leaves only two routes of entry for biomarkers—transcellular (through the cells) or paracellular (between the cells). The cellular lipid bilayers form a barrier for transcellular diffusion, while the tight junctions between the cells form a barrier for paracellular diffusion (Fig
Figure 2.4. Biomarkers can pass through epithelial barriers through the cells (transcellularly) or between the cells
(paracellularly).
17 2.4). Moderately hydrophobic molecules (e.g., cortisol) and small (<100 Da), uncharged, polar
molecules (e.g., ethanol and urea) can pass through the lipid bilayer and enter sweat in
appreciable quantities (Fig. 2.5). However, larger, uncharged, polar molecules (e.g., glucose)
and charged molecules (e.g., ions or proteins) must move through the space between the cells.
Even though this space is 10’s of nanometers wide11, tight junctions pinch neighboring cell
membranes together to form a seal that makes it difficult for anything but small (<100 Da) ions
to pass.
Figure 2.5. Biomarker routes of entry are based on molecular properties. Some biomarkers can pass through the lipid bilayer, while others are blocked.
However, some biomarkers can still pass as a result of the dynamic nature of tight junctions.
Tight junctions make up a flexible network of strands that continuously break, reseal, and
branch37,38. While this reshaping can cause direct aqueous paths to form momentary, most
biomarkers diffuse from ISF to sweat between opening and closing compartments that fluctuate
between the strands37,38, creating a tortuous path. This is why many biomarkers are filtered
when entering sweat. Therefore, it is hypothesized that as sweat rate decreases, less fluid is
generated in the sweat gland, and so the biomarkers that passively diffuse into sweat will be
less diluted. This means that the sweat rate must be considered when measuring biomarkers
18 and could explain the large differences in biomarker concentrations found in the literature. In this chapter, the sweat-rate dependence of a set of target biomarkers will be examined experimentally. This is critical to accomplish first since it will help ensure that future experiments consider this important factor that could naturally alter biomarker concentrations.
2.5 Experimental methods
To determine how sweat rate affects biomarker concentration in sweat, sweat must be generated, collected, and then analyzed. In addition, a method of reliably measuring sweat rate is needed. For this and all future experiments with sweat, a device capable of performing these required tasks was needed.
2.5.1 Custom testing device
All tests were performed using a custom testing device—printed using a stereolithographic
(SLA) 3D printer (Formlabs Form 2) in UV-curable plastic resin—consisting of a base that adheres to the skin and a set of removable attachments that lock into the base (Fig. 2.6). The
Figure 2.6. Custom two-part device for sweat tests is shown with base and multiple twist-lock attachments.
base has an adhesive-covered (3M 1577) flange that firmly adheres to a ring of skin surrounding the testing area (circular region with 28 mm diameter). This, combined with a
Velcro strap, locks the testing area in place and prevents it from skewing as the participant
19 moves. The attachments mate with the base using a twist-lock mechanism. Attachments for stimulating sweat production, collecting sweat, and measuring sweat rate were used. Before attaching the device, participants were instructed to wash the testing area with soap and water.
The area was then cleaned with isopropyl alcohol, rinsed with deionized (DI) water, and then dried thoroughly by blotting with wipes (Kimwipes, Kimtech Science) and evaporating with compressed nitrogen.
2.5.2 Sweat generation
Sweat production can be induced naturally (via exercise or increased environmental temperature) or artificially (via parasympathetic mimetic drugs—carbachol or pilocarpine). Since artificially-induced sweating is more controlled and localized to the testing area, it is the preferred method of sweat stimulation. Artificially induced sweating follows a protocol like that of
Figure 2.7. Demonstration cystic fibrosis diagnostic setup of the Wescor Nanoduct Neonatal Sweat Analysis System.
(Credit: Wescor)
the Wescor Nanoduct Neonatal Sweat Analysis System (Fig. 2.7). The Nanoduct works by delivering pilocarpine, a cholinergic drug, to the sweat gland to stimulate sweating using a technique called iontophoresis3,39,40. Iontophoresis is a method of applying an electric current to deliver ionized drugs transcutaneously (Fig. 2.8). The Nanoduct uses circular, agarose gel disks
(1.89 cm2) containing the positively-charged drug. A gel disk is snapped into an electrode receptacle, which is strapped to the skin. A reference electrode is also applied to the skin. A
20 voltage is placed across the electrodes such that the positive lead is connected to the disk containing the positively-charged drug to be delivered. The Nanoduct controls the current to be
0.5 mA (0.26 mA/cm2). The result is that the positively-charged drug is repelled by the positive potential at the cathode and drawn down into the sweat gland following the electric field lines towards the anode. Pilocarpine is used by the neonatal Nanoduct for cystic fibrosis testing41.
Carbachol is typically used in eye drops to treat glaucoma42. Additionally, a previous study has shown the use of carbachol in iontophoretic stimulation of sweat43. For safety purposes, dosing was set to be less than that used in the FDA-approved Nanoduct (3% stimulant-gel disk, 0.5 mA for 2.5 minutes).
Figure 2.8. Natural and artificial sweat stimulation. Natural stimulation occurs in response to thermal loading or metabolic activity which triggers the release of acetylcholine. Artificial stimulation is possible via iontophoresis of a positively-charged acetylcholine look alike, carbachol. (Credit: Microchip Technology Inc.)
21 A first principles calculation can be performed to estimate the amount of charged molecules delivered to the skin using iontophoresis. The charge delivered, in Coulombs, is the product of the applied current (mA or mC/s) and the duration of iontophoresis. In the case of the Nanoduct protocol (0.5 mA for 2.5 minutes), the charge delivered would be 75 mC. The charge can be converted to moles of electrons using the Faraday constant (96,485 C/mol): 0.78 μmol of electrons in the case of the Nanoduct. Carbachol has a single positive charge at physiological pH. Assuming all the charge results in the delivery of only carbachol, a theoretical maximum delivery can be estimated to be 0.78 μmol of carbachol (or 142 μg). This calculation assumes that all charge is used to deliver the molecule of interest. Also, the charge, size, and shape of the molecule as well as the properties of the surrounding solutes and solvent affect the movement of charged molecules24,25,44. For example, carbachol competes with other positively charged species such as sodium. As a result, the first principles calculation represents a maximum amount delivered when in fact a significantly smaller dosage of the molecule of interest is likely delivered. These calculations are still helpful in determining if the upper-bound dosage is within safe limits.
All experiments discussed herein were performed using artificially stimulated sweat.
Carbachol—a slowly-metabolized cholinergic agent—has been shown to provide localized sweat stimulation when delivered iontophoretically45. A carbachol-containing gel disk (1% carbachol, 3% agarose, 6.1 cm2 x 6 mm thick) is placed into the testing device, and an attachment with a conductive-carbon layer locks into the device, pressing the disk against the skin (Fig. 2.9). A disposable counter electrode (commonly used for TENS stimulation) is then placed on the adjacent skin. A lab-grade potentiostat (Gamry 600) drives the carbachol into the skin by applying a constant current density of 0.02 mA/cm2 for 10 minutes. It takes about 15-30 minutes to produce an adequate sweat response in most people. After reaching a peak sweat
22 rate, sweat rate begins to decline. For carbachol, the sweat response lasts about 24 hours; for pilocarpine, the response last a few hours at most.
Figure 2.9. Flow diagram of sweat stimulation. First, the skin is cleaned with water and isopropyl alcohol.The base of the two-part device is secured with adhesive and strapped to the forearm of a subject. A carbachol-containing drug disk is loaded and a conductive attachment is locked into place. A constant current is then applied to iontophoretically deliver the sweat-stimulating drug.
2.5.3 Avoiding sweat contamination
Sweat needs to be collected on the skin, but the skin can contaminate samples by harboring microbes, accumulating analytes, and sloughing off cells. Both the density and diversity of the microflora (bacteria, fungi, and viruses) differ based on the region of the body and can differ widely between people and even within the same person over time33,46. As discussed above, estimates of the density of bacteria found on the skin are as high as 10 billion/cm2 47, and since bacteria can consume analytes such as energy sources like glucose and secrete analytes such as proteins or cellular waste products, this presents a challenge for accurate measurement of biomarker concentrations. In addition, sweat minerals have been shown to accumulate in the superficial layers of the epidermis and possibly in the sweat duct itself prior to sweating events
48,49. It can be assumed that similar accumulation may occur with other biomarkers, lowering the
23 time resolution of the sweat sample. Finally, the skin surface is constantly being coated with proteases which aid in the shedding of dead skin cells and a mixture of triglycerides, wax esters, squalene, and metabolites from sebaceous glands but cause a problem for protein detection50,51. Skin surface contaminates become an even larger issue for small samples. These issues can be avoided by preventing sweat from contacting the epidermis by coating the skin with an occluding layer of petroleum jelly or oil34,36 (Fig. 2.10).
Figure 2.10. Flow diagram of sweat collection and analysis. Petroleum jelly is applied to the testing area to form a barrier between sweat and epidermal contaminants. An absorbent, sweat collection disk is placed onto the testing area. A screwcap will be placed to prevent evaporation during collection. Sweat collection disk is spun down to collect liquid sample. Liquid sweat sample is analyzed using standard assays.
2.5.4 Sweat collection
A thin layer of petroleum jelly is applied to the testing area using a sterile cotton applicator.
Then, the testing area is cleaned with isopropyl alcohol, rinsed with DI water, and dried thoroughly. Any excess petroleum jelly is removed by blotting with a wipe. To allow for almost complete sample recovery, absorbent disks are made of a non-woven hydrophilic mesh (Sefar
Inc. Nitex 03-110/47) with low non-specific binding to analytes. Disks are laser cut (VLS3.50,
Universal Laser Systems) to a diameter of 28 mm. An absorbent disk is placed within the base
24 of the device and kept pressed against the skin using an attachment that locks into the base
(Fig. 2.10). At the end of the collection period, the absorbent disk is removed. The collected sweat is then extracted by centrifugation and stored in a -20°C freezer.
2.5.5 Sweat glucose
Sweat glucose is determined using an enzymatic glucose assay (Amplex Red Glucose Assay,
Thermo Fisher Scientific) read via a microplate reader (Synergy H1, BioTek). Before analysis, sweat samples are thawed to room temperature and are diluted by at least two-fold using the buffer provided with the assay to mitigate the effects of pH on enzyme activity.
2.5.6 Blood glucose
Blood glucose is measured using a fingerstick glucometer. Before each measurement, participants are instructed to wash their hands with soap and water. The finger to be tested is cleaned with isopropyl alcohol, rinsed with water, and then dried thoroughly. Blood is tested using Accu-Chek Aviva Plus test strips (Roche Diagnostics), because this brand had the best performance when compared to other commercially available strips52.
2.5.7 Sweat rate
Sweat rate is determined gravimetrically (Fig. 2.11). Absorbent disks (TX609 TichniCloth,
Texwipe) are laser cut to a diameter of 28 mm and individually weighed on an analytical balance
(ViBRA Shinko Denshi HT224R). Then, the surface of the skin is thoroughly dried. The same attachment used during sweat collection is used to keep the absorbent disk flat against the skin.
After 10 minutes, the absorbent disk is removed and weighed again. The difference in weights is converted to volume of water using the density of water (1 g/mL). The volume collected over the duration of collection yields the sweat rate.
25
Figure 2.11. Flow diagram of sweat rate measurements. Absorbent disks are weighed prior to being placed on the skin. After 10 minutes, the disks are reweighed. The difference in weights over time provides the sweat rate.
2.5.8 Glucose flux under normal conditions
For all experiments, participants are asked to fast for at least 8 hours prior to the start of the experiment and throughout the entire duration. All experiments are started in the morning at approximately 9AM. The custom testing device is fixed to the subject’s volar forearm approximately 15 cm proximal to the distal wrist crease. Sweat production is then stimulated as described above. After a 20-min wait period, the two experiments described below begin. Figure
2.12 outlines the procedure.
Sweat rate is measured gravimetrically for 10 mins. Midway through the sweat rate measurement, blood glucose is measured using the fingerstick test described above.
Immediately after the sweat rate measurement completes, sweat is collected for 45 mins as described above. This procedure of measuring sweat rate, blood glucose, and collecting sweat is repeated 6-8 times. The experiment ends with a final sweat rate and blood glucose
26 measurement. Both sweat rate and blood glucose measurements are averaged from before and after each collection period to determine the average sweat rate and blood glucose during the collection period. The sweat glucose concentration during the collection period is determined using the sweat analysis procedure described above.
Figure 2.12. Block diagram of experimental setup for determining glucose flux under normal conditions.
2.5.9 IRB protocol
This study was performed under University of Cincinnati IRB protocol study ID 2016-4769. The glucose flux experiment consisted of 6 healthy participants (5 male, 1 female) between approximately 20 to 40 yrs. old.
2.6 Results and discussion
Participants were artificially stimulated to sweat as described above and developed a peak sweating response that decayed over the course of several hours45. Sweat glucose concentrations and sweat rates were measured at various time points during this decay on the same site on the forearm (Fig. 2.13). As sweat rate increases, sweat glucose concentration decreases. This sweat rate dependence exists because glucose flux is largely independent of sweat rate. In fact, all participants had similar glucose fluxes: ~100 fmol/(min-gland) (Fig. 2.13).
Since participants were asked to fast prior to the study, their resting blood glucose values were both steady and similar, making the only difference the permeability of the sweat gland. This implies that a basal sweat gland permeability may exist and that high sweat rates simply dilute the concentration.
Under normal conditions, glucose flux into sweat is driven by Fick’s diffusion:
Eq. 2.1: 퐹푙푢푥퐹푖푐푘′푠 = 푝(∆푐)
27 Figure 2.13. Sweat glucose concentration and sweat rates measured at various time points following artificial sweat stimulation. Sweat glucose concentration is inversely proportional to sweat rate. The average glucose flux for all participants is roughly 100 fmol/(min-gland) (n = 6).
where flux is proportional to the concentration gradient (∆푐) with a proportionality constant, or permeability constant (푝). The glucose concentration in sweat is much lower than that in blood or ISF, and so the concentration gradient can be simplified.
Eq. 2.2: ∆푐 = 푐푏푙표표푑/퐼푆퐹 − 푐푠푤푒푎푡 ≈ 푐푏푙표표푑/퐼푆퐹
where 푐푏푙표표푑/퐼푆퐹 is the concentration of glucose in blood or ISF and 푐푠푤푒푎푡 is the concentration of glucose in sweat. The sweat glucose concentration is simply the total glucose flux (Fick’s diffusion) over the total water flux (sweat rate).
퐹푙푢푥퐹푖푐푘′푠 푝(푐푏푙표표푑/퐼푆퐹) Eq. 2.3: 푐푠푤푒푎푡 = = 푄푠푤푒푎푡 푄푠푤푒푎푡
Therefore, fitting an inverse function to the sweat glucose concentration and sweat rate relationship yields glucose flux.
2.7 Conclusions
28 This chapter represents a paradigm shift in the way biomarkers in sweat are analyzed by taking a flux-based rather than concentration-based approach. This approach is driven by an understanding of the ways in which biomarkers enter sweat as described above. The fundamental flaw with focusing on concentration alone is that it is dependent on the total water flux. Previous sweat biomarker studies ignored the importance of measuring sweat rate when measuring the concentration of biomarkers6,7,34. By focusing on biomarker flux, the effect of sweat rate on the concentration can be normalized. Under normal sweat collection conditions, this normalization is simple; the product of sweat rate and concentration yields biomarker flux. If both the flux and the permeability of the sweat gland is known, then the biomarker’s concentration in blood or ISF can be calculated.
In addition, this chapter highlighted a key phenomenon—glucose permeability may be similar from person to person. However, since basal glucose permeability was only measured from a total of six male and female participants, it is difficult to discern the impact of demographic factors, such as age or sex. Future work should include a larger, more diverse sample set.
29 Chapter 3: Continuously concentrate sweat samples within real-time device
This chapter includes adapted text and figures from the following article:
A. Jajack, I. Stamper, E. Gomez, M. Brothers, G. Begtrupe, J. Heikenfeld. Continuous, quantifiable, and simple osmotic preconcentration and sensing within microfluidic devices. Lab
Chip. Submitted.
3.1 Introduction
Advances in microfluidics have driven the miniaturization, integration, and automation of traditional analytical techniques for measuring biologically-relevant molecules. Lab-on-a-chip devices have targeted high-impact applications, such as point-of-care diagnostics53,54, but new frontiers, such as real-time health monitoring55, remain untapped. In many cases, the physiological concentrations of analytes found in non-invasively accessible biofluids (e.g., sweat, saliva, tears) are orders of magnitude below the range of detection of available analyte sensors4 or have binding and dissociation kinetics that cause equilibrium to take hours to achieve56,57. In this chapter, the mismatch between physiological concentration of analytes and the range of detection of currently available biomarkers is addressed using a technological solution.
One possible approach is to improve the range of detection of analyte sensors. To do this, the sensitivity (signal magnitude per unit concentration) must increase, and the limit of detection (a function of signal-to-noise ratio) must decrease. Unfortunately, no universal method exists to downshift the range of detection for all sensors. Downshifting the range of detection either requires the tedious development of new probe chemistries (antibodies, aptamers, etc.) repeated for each analyte of interest or enhancement of the signal produced from a single binding event. Multiple options exist for improving the latter; the most advanced sensor transducers, such as analyte probes on electronic graphene58, optically resonant surface plasmon59, or mechanically resonant micro-cantilevers60, downshift the range of detection by
30 increasing sensitivity rather than improving the binding interaction between the analytes and the probes. However, in all of these cases, enhancing the signal also enhances the noise, which is significant not only in complex biological media but also in electronic sensing.
An alternative approach, the one which will be the focus of this chapter, is to concentrate the sample before sensing. By bringing analytes into the detection range of sensors, concentration techniques can improve detection regardless of the sensing modality. For this reason, preconcentration methods (e.g., liquid-liquid extraction, solid phase extraction, crystallization, or precipitation) have been a staple of laboratory-scale analytical procedures but are usually time/labor-intensive and often must be performed in non-continuous batches61. On-chip concentration methods have been developed to overcome these limitations. Some on-chip methods take advantage of the electrokinetic properties of analytes to concentrate them using velocity differences or focusing effects61. However, these methods (e.g., field-amplified stacking, ion concentration polarization, isotachophoresis, or isoelectric focusing) require an applied electric field, often using high voltages; they are generally limited to concentrating charged analytes, such as ions and proteins; and conditions for concentration are highly sensitive to flowrate and to sample composition61,62. Other on-chip concentration methods rely on surface- binding effects, such as immunoprecipitation. These methodologies are inherently not real-time, as they require constant cycling between binding and resolubilization of the analyte of interest.
Also, many preconcentration methods provide difficult-to-quantify increases in concentration that make it challenging to correlate the measured concentration to the original concentration in the raw sample.61 There is, therefore, a clear need for continuous, quantifiable, and simple preconcentration methods that work across a wide range of analyte classes (e.g., ions, molecules, proteins, etc.) and solutions.
Concentration merely is solute per solvent. By focusing on reducing the amount of solvent
(usually water), broad-spectrum solutions become possible. Semi-permeable membranes are
31 well suited for this approach63. When suitably selected, a semi-permeable membrane can allow the passage of water while leaving analytes of interest in the channel. The movement of water can be driven by either a pressure (reverse osmosis) or an osmolarity (forward osmosis) gradient. Reverse osmosis requires active, pneumatic control of high pressures and is thus innately more difficult to miniaturize down to an on-chip device. On the other hand, forward osmosis requires no external peripherals. However, previous attempts to use forward osmosis for preconcentration were not continuous, quantifiable, or simple63,64.
This chapter describes the first membrane-based, microfluidic preconcentration device that is continuous, quantifiable, simple, and capable of working with any analyte. A cross-section of the microfluidic device is shown in Fig. 3.1a. A sample containing a dilute concentration of an analyte of interest flows into a microfluidic channel. The top of the channel is laminated with a semi-permeable membrane that rejects the analyte. A draw solution reservoir with a pre- determined high osmolarity sits on top of the membrane. Often, the sample will be a biofluid where the osmolarity is known to be within a tight range. The osmolarity gradient between the draw solution reservoir and sample solution will draw water into the draw solution reservoir, concentrating the sample until equilibrium is reached. Therefore, as long as the flowrate of the sample is below a critical threshold so that the sample osmolarity can equilibrate with the draw solution osmolarity, then the increase in concentration will be quantifiable. Furthermore, if the sample osmolarity is unknown or highly variable, placing Ag/AgCl-based Cl- sensors before and after the concentrating module can be used to potentiometrically quantify the amount of preconcentration. In either setup, sensors can then be placed downstream to detect the analytes of interest. Fig. 3.1b shows a working device with a small footprint and ready-for- manufacture design implemented by ALine Inc. Impressively, with <1 cm2 of exposed membrane area, complete 10-100x preconcentration occurs within mere minutes.
32 Figure 3.1. (a) Cross-sectional diagram demonstrating the operating principle of the microfluidic preconcentration device. (b) A miniaturized preconcentration device designed with form-factor and manufacturability in mind, produced by ALine Microfluidics. (c) SEM image of the FTS F2O Rainstick forward osmosis membrane, reproduced.107
This chapter will also cover the evaluation of the performance of multiple combinations of membranes and draw molecules to optimize for the device. While a few combinations have been studied,65–68 no extensive cross-comparison has been performed. This study is the first to perform a systematic optimization of the performance of several classes of membranes (i.e.,
33 forward osmosis, nanofiltration, and dialysis) in combination with a panel of differently-sized draw molecules (i.e., a salt, a sugar, and a polyelectrolyte).
Next, the design and construction of a proof-of-concept preconcentration device with sensors throughout its length is covered. In addition, a numerical model similar to that described by
Phuntsho et. al is presented69. A set of characterization experiments were conducted to allow the model to be fitted to the results. The selected membrane’s ability to concentrate two biologically relevant molecules—glucose and bovine serum albumin (BSA)—was evaluated.
Specifically, in the case of BSA, a traditionally sticky molecule, the membrane’s resistance to fouling was put to the test.
This novel preconcentration device will accelerate the development of next-generation lab-on-a- chip applications—enabling real-time, health-monitoring using non-invasive dilute biofluids (e.g., sweat, saliva, tears), allowing for the development of new point-of-care diagnostic tests, and impacting the detection of contaminants which are often measured in non-traditional fluid systems (e.g., food stocks, sewage, ocean water, etc.).
3.2 Experimental methods
The experimental methods are broken down into two main sections—membrane and draw molecules optimization and preconcentration device construction, characterization, and modeling.
3.2.1 Membrane and draw molecule optimization
First, the rationale behind the membranes and draw molecules selected for evaluation is explained. Then, the method of evaluating the performance of different combinations of membranes and draw molecules is detailed. In this section, a discussion on concentration polarization and the issue with wet vs dry membranes is also included.
34 Membrane and draw molecule candidates. Both the membrane and draw molecule are critical components that require careful consideration. The membrane must reject the analytes of interest and the draw molecule while allowing water to flow through at reasonable rates. The size and chemistry of a membrane’s pores determine the threshold of molecular sizes and types that will be rejected, and the density and length of these pores determine the resistance to water. The draw molecule must be large enough to be rejected by the selected membrane but small enough to have high solubility in water (osmolarity) and low viscosity65.
Membranes from three different classes were chosen—forward osmosis, nanofiltration, and dialysis—representing different pore sizes, porosities, constructions, and materials. All membranes were hydrophilic as this best supports water flux70. The forward osmosis membrane
(FTS Rainstick) has pores with a diameter of just 0.7 nm and is the only membrane selected that can reject NaCl by >99.9 %71. This thin-film composite membrane is designed to support high water flux and to resist fouling72–75. Several polyamide, nanofiltration membranes (Synder
Filtration) with various molecule weight cutoff (MWCO) ranges were also selected: NFS (100-
250 Da MWCO, 50 % NaCl rejection), NFX (150-300 Da MWCO, 40 % NaCl rejection), NFW
(300-500 Da MWCO, 20 % NaCl rejection), and NFG (600-800 Da MWCO, 10 % NaCl rejection). Finally, several cellulose-ester, dialysis membranes (Biotech, Spectrum Labs) with the following MWCO ranges were selected: 0.1–0.5 kDa, 0.5–1.0 kDa, and 3.5–5.0 kDa.
Draw molecules that have been previously demonstrated in forward osmosis systems were chosen: a salt (NaCl)65, a sugar (sucrose)67, and a polyelectrolyte (polyethylenimine)66. If a membrane fails to reject Na+ and Cl- (23 Da, 35 Da), then larger sucrose (342 Da) or very large polyelectrolytes like polyethylenimine (600 Da to 60 kDa) may be required by membranes with larger pores. While both NaCl and sucrose are similarly soluble in water (up to roughly 6 M),
NaCl is advantageous, as it is much less viscous at these higher concentrations. Each subunit of a polyelectrolyte contributes to its osmolarity, allowing it to produce a large osmotic pressure
35 while being easily rejected by most membranes. However, like sucrose, polyethylenimine is extremely viscous at the high concentrations needed to concentrate effectively.65,66 Branched polyethylenimines were used with both low (Mn ~600 Da; P/N: 408719, Sigma-Aldrich) and high
(Mn ~60 kDa; P/N: 181978, Sigma-Aldrich) molecular weights.
Membrane and draw molecule performance evaluation. The performance of each membrane and draw molecule combination is evaluated using water flux. Water flux is measured using a testing apparatus that clamps a membrane between two interlocking reservoirs (Fig. 3.2a inset). The reservoirs are printed using a high-resolution, stereolithographic
3D printer (Form 2, Formlabs) in clear photopolymer resin (FLGPCL02, Formlabs). Each reservoir holds roughly 10 mL of fluid and makes a fluid-tight seal against the membrane using an O-ring, leaving a 2.5 cm2 area of the membrane exposed for fluid transport. Membranes are wetted with deionized water for several minutes before being clamped within the testing apparatus. One reservoir is filled with 10 mL of deionized water. Then, the apparatus is weighed on an analytical balance. To begin the test, 10 mL of a draw solution are added to the other reservoir. The osmolarity of the draw solution was set to a constant, 1 OsM. However, the draw molecules being tested have different degrees of dissociation and so contribute differently to the osmolarity. NaCl dissociates into two ions, sucrose is nonionic, and polyethylenimine monomers dissociate into roughly two ions at pH 6.5. Therefore, the draw concentrations for NaCl, sucrose, and polyethylenimine were set to 0.5 M, 1 M, and 0.5 M, respectively. The apparatus is sealed using paraffin film to prevent any possible evaporative effects. After three hours, the draw solution is aspirated from its reservoir, and the apparatus is weighed again. Water flux is calculated as the difference in measured weights divided by both the duration and exposed membrane surface area. This procedure was performed in triplicate for all combinations of membranes and draw molecules.
36 Figure 3.2. (a) The forward osmosis performance of various combinations of membranes and draw molecules.
Performance metric is the average water flux over a 3 h period. Sample is 10 mL deionized water; draw osmolarity is
1 OsM. Overlay image depicts testing apparatus. (b) The MWCOs of each membrane and MW of each draw molecule is plotted together on the same graph to give insights on the size compatibility of each combination. Three membrane classes are represented: forward osmosis (FTS), nanofiltration (NFS, NFX, NFW, and NFG), and dialysis
(0.1-0.5 kDa, 0.5-1.0 kDa, and 3.5-5.0 kDa). Draw molecules are NaCl, sucrose, and both low and high MW branched polyethylenimine (PEI).
Concentration polarization. Forward osmosis performance can be affected by concentration polarization, a buildup of concentration gradients internal or external to a membrane76,77. This reduces the effective osmotic gradient across the active layer of the membrane thereby reducing the water flux. Internal concentration polarization (ICP) results from the fact that most
37 membranes are asymmetric, having a dense, active layer and a porous, support layer (Fig. 3.3).
Membranes can be oriented with the active layer facing either the feed or draw solution. When the active layer faces the feed solution, draw molecules from the bulk solution must diffuse into the support layer against the flow of incoming water. As a result, the draw solution is diluted at the interface between the active and support layer. Because the concentration of the draw solution is much larger than the feed solution, a dilution on the draw side dramatically reduces the osmotic gradient, and so the water flux is diminished. The effects of ICP are exacerbated when the thickness of the support layer increases, the porosity of the support layer decreases, water flux increases, or diffusivity of the draw molecule decreases. ICP can largely be avoided by flipping the orientation of the membrane. However, since the active layer is less prone to fouling than the support layer, membranes are oriented such that active layer always faces the sample. External concentration polarization (ECP) occurs at the surface of the membrane due to water being removed from the feed side causing a local region of concentration and water fluxing to the draw side causing a local region of dilution. However, the water fluxes involved in forward osmosis are so low that ICP dominates, and ECP can be ignored.
Figure 3.3. Schematic representation of (a) dilutive ICP and (b) concentrative ICP, reproduced76.
38 Wet vs. dry membranes. Membranes are stored either dry or wet, but once wet, a membrane should not be allowed to dry out, or else its properties may change. The forward osmosis and nanofiltration membranes are stored dry, but the dialysis membranes are pre-wetted. While not an issue for the membrane and draw molecule characterization experiments, pre-wetted membranes may be difficult to integrate into a device and were thus avoided.
3.2.2 Preconcentration device construction, characterizing, and modeling
In this sub-section, the construction of the preconcentration device is described. Then, the process of characterizing the concentration profiles of devices operated at various input flowrates is explained. The way in which the biological relevance of the device was evaluated is discussed next. Finally, the operating principle of the in-silico model, how it was fitted to experimental data, and the advantages and limitations of this approach are discussed.
Construction. The traditional soft lithography techniques for producing microfluidic devices are quickly being replaced by even more rapid prototyping techniques78–80. The well-documented, design-cut-assemble approach80 was used to produce the device layer-by-layer—a rigid support layer, a microfluidic channel patterned from double-sided tape, the selected semi-permeable membrane, and a draw solution reservoir. The entire device is built up from the rigid support layer. This layer is cut from a clear acrylic sheet (3 mm-thick, McMaster-Carr) using a laser cutter. Ag/AgCl ink (P/N: Cl-4001, Engineered Conductive Materials) is screen printed onto the acrylic support layer to provide six equally-spaced chloride-ion sensors throughout the length of the device. A straight microfluidic channel (150 mm long; 0.5 mm wide) is laser cut out of double-sided, microfluidic tape (90 um thick, 3M™ 9965). The patterned double-sided tape is aligned with the inlet and outlet ports of the acrylic sheet and then laminated using a heated-roll laminator. Similarly, the membrane (FTS Rainstick) is laminated on top of the tape. Strips of the same microfluidic tape are laminated across the inlet and outlet ports to ensure that only the straight channel is exposed to the draw solution. The draw solution reservoir must hold several
39 milliliters of draw solution (NaCl), and so a perimeter wall is cut out of acrylic and sealed to the top of the membrane with hot-melt adhesive. Evaporation can be prevented by covering the reservoir with paraffin film, plastic wrap, or another sheet of acrylic. Once all layers are assembled, ~20 cm lengths of polyetheretherketone (PEEK) tubing (0.03” ID, 1/16” OD) are pressed into the inlet and outlet ports that were cut on the acrylic support layer and then secured in place with UV curable epoxy (Loctite 352).
Concentration profiles at various input flowrates. A device is clamped to a stand so that the draw solution reservoir is facing up. The PEEK tubing of the inlet is attached to an empty, 1 mL glass syringe (Hamilton 1000 series) via a compression fitting, and the PEEK tubing of the outlet rests in a beaker containing 5 mM NaCl and a double-junction Ag/AgCl reference electrode (P/N
Z113107, Sigma-Aldrich). Only for experimental characterization (not needed in practice), the draw solution reservoir is initially filled with the same 5 mM NaCl so that the device channel can be filled without any water flux. A syringe press (KD Scientific Legato 111) withdraws the glass syringe at 20 µL/min until the 5 mM NaCl from the beaker is drawn through the channel and into the syringe. This withdrawal phase is critical to filling the channel without air bubbles.
The syringe is then removed from the syringe press. The syringe is disconnected from the inlet tubing by submerging the syringe and tubing in a large beaker of deionized water. The tubing is left submerged while the syringe is fully loaded with 5 mM NaCl. The syringe is primed, submerged, and then reattached to the tubing. This tedious, submersion process is critical to ensure that no air bubbles enter the channel. The syringe press infuses the 5 mM NaCl into the channel at 20 µL/min. The six screen-printed Ag/AgCl electrodes within the channel have an electrochemical potential that correlates to chloride ion concentration of the sample directly above the electrode when measured against the reference electrode in the outlet beaker. The six electrodes and the reference electrode are connected to a multi-channel, precision, high input impedance (1013 Ω) electrode interface (EMF6, Lawson Labs). The electrical potential of
40 each of the six electrodes is recorded. This process is repeated for 50 mM and 500 mM solutions to produce a three-point calibration curve for each electrode. Between each test, the draw solution is aspirated, rinsed with the new solution, then aspirated again before refilling.
A profile of the concentration as a function of position in the channel can then be determined for various input flowrates (Fig. 3.4). In this particular experiment, the device is configured to concentrate a 50 mM NaCl sample solution by 10x. The draw solution was replaced with 500 mM NaCl to provide the 10x osmolarity gradient, and the syringe was replaced with 50 mM
NaCl and then infused at various input flowrates (1 µL/min to 6 µL/min). The concentration at each of the six electrode positions was recorded after stabilizing. The calibration procedure is performed both before and after generating the profile to account for electrode drift. Reported concentration values are corrected to account for this drift and averaged from three independent runs each using a separate channel.
Figure 3.4. Concentrations of NaCl at discrete positions throughout straight-channel preconcentration devices
operating at various input flowrates. The sample solution is 50 mM NaCl, and the draw solution is 500 mM NaCl
which should concentrate samples by 10x. Concentrations are predicted by the in-silico model and plotted for
each flowrate (solid lines). A picture of an actual preconcentration device is depicted below the graph.
41 Biological relevance. To test if the device is capable of concentrating biologically relevant molecules without fouling, two molecules were selected: glucose, a small sugar, and BSA, a large sticky protein. A sample containing 100 µM glucose in 0.1x phosphate-buffered saline
(PBS) is concentrated by 10x using a 1x PBS draw solution, and a sample containing 0.39 mg/mL BSA in 1x PBS is concentrated by 10x using a 10x PBS draw solution. A device is prepared and pre-filled with the respective sample using the withdraw-based filling method as described above. Once the device is filled, the syringe is removed and filled with the respective sample using the submersion technique described above. Then, the syringe pump infuses the sample at various flowrates into the device. Concentrated samples from the outlet are collected and analyzed.
Glucose concentrations are determined using electrochemical glucose test strips designed for diabetes management. These test strips can be hooked up to a lab-grade potentiostat (Gamry
Instruments Reference 600+) rather than a glucometer to increase the precision of the reading81. Test strips are preferred over traditional assay techniques, because they require less than 1 µL of sample and provide results in less than a minute. BSA concentrations are determined by absorption at 280 nm using a UV-vis spectrophotometer (NanoDrop One
Microvolume, Thermo Scientific). This technique also requires less than 1 µL of sample and provides instantaneous results. Glucose concentration was replicated in four separate channels, and BSA concentration was replicated in two separate channels.
Predictive in-silico model. An in-silico model was developed that accurately predicts the behavior of the preconcentration device. The model simulates a sample solution (with a given osmotic concentration and input flowrate) flowing through a microfluidic channel (with a given length, width, and height). The sample solution is separated from a draw solution (with a given osmotic concentration) by a membrane (with a given resistance). The model, written in C++, performs the following on each discrete cycle, or tick:
42 1. A block of fluid is added to the channel inlet.
(푉표푙푢푚푒 표푓 푏푙표푐푘 푎푑푑푒푑) = (퐼푛푝푢푡 푓푙표푤 푟푎푡푒)(퐷푢푟푎푡𝑖표푛 표푓 푡𝑖푐푘)
2. Loop through each block and remove volume lost through the membrane due to
osmosis.
((퐷푟푎푤 푠표푙푢푡𝑖표푛 푐표푛푐. ) − (퐵푙표푐푘 푐표푛푐. ))(𝑖푅푇)(푀푒푚푏푟푎푛푒 푎푟푒푎) (퐹푙표푤 푟푎푡푒 푡ℎ푟표푢푔ℎ 푚푒푚푏푟푎푛푒) = (푀푒푚푏푟푎푛푒 푅푒푠𝑖푠푡푎푛푐푒)
(푁푒푤 푣표푙푢푚푒 표푓 푏푙표푐푘) = (푂푙푑 푣표푙푢푚푒 표푓 푏푙표푐푘) − (퐹푙표푤 푟푎푡푒 푡ℎ푟표푢푔ℎ 푚푒푚푏푟푎푛푒)(퐷푢푟푎푡𝑖표푛 표푓 푡𝑖푐푘)
3. If blocks have reached the outlet of the device, the model is finished. If not, begin
next tick.
The model outputs the flowrate and concentration at every block along the channel as well as the time needed to reach the outlet. Experimental outlet concentration data collected at various flowrates can be used to determine the membrane resistance. For simplicity, ICP ECP are not considered in this model. In forward osmosis, ECP is negligible, and ICP only changes when the membrane, draw molecule, or driving osmolarity gradient is changed. Since the membrane and draw molecule selection was optimized, they will not change. As a result, the membrane resistance term can be thought of as having the added resistance of ICP “baked in” for a set osmolarity gradient. Therefore, as long as the model is re-fit whenever the osmolarity gradient changes, it can be used as a useful, predictive tool. The trend lines shown in the figures represent the model output for those experiments.
3.3 Results and discussion
Within the results and discussion section, the results of the optimization experiments for the membrane and draw molecule are presented. Then, the discussion is focused on the preconcentration device itself.
3.3.1 Membrane and draw molecule optimization
43 In this sub-section, the performance of various combinations of membranes and draw molecules is attempted to be explained by considering pore sizes and molecules weights. Ultimately, these results are used to select the optimal combination. Then, the validity of this approach as well as its strengths and weaknesses are discussed.
Results of membrane and draw molecule performance evaluation. The water flux for each membrane and draw molecule combination is reported in Fig. 3.2a using a constant 1 OsM gradient. The NaCl draw solution is most effective when used with the salt-rejecting FTS membrane. However, with all other membranes, the NaCl draw solution is one of the poorest performers due to these membranes’ poor rejection of NaCl. Based on their MWCOs, the FTS,
NFS, and NFX membranes should reject sucrose, the NFW and the 0.1-0.5 kDa dialysis membranes may partially reject it, and all the rest should largely allow it to pass (Fig. 3.2b). The sucrose draw solution has a slightly lower water flux compared to the NaCl draw solution across the FTS membrane, possibly due to increased ICP as a result of the larger viscosity of the sucrose draw. The sucrose draw solution produces the lowest water flux through the NFS membrane, suggesting that either the NFS membrane may not be rejecting sucrose as expected or sucrose is fouling this membrane. The sucrose draw solution also produces low water fluxes in the other nanofiltration membranes (NFX, NFW, and NFG) compared to the FTS and dialysis membranes. The dialysis membranes have the largest pore sizes and exhibited large water fluxes when combined with a sucrose draw. Surprisingly, regardless of the draw solution, the water flux does not increase as expected as pore size increases within the nanofiltration (NFS, NFX, NFW, or NFG) or the dialysis (0.1-0.5 kDa MWCO, 0.5-1 kDa MWCO, or 3.5-5.0 kDa MWCO) membrane classes. The low molecular weight polyethylenimine should be rejected by the NFG nanofiltration, 0.5-1.0 kDa MWCO dialysis, and 3.5-5.0 kDa MWCO dialysis membranes, and the high molecular weight polyethylenimine should be rejected by all membranes. In all membranes, the low molecular weight polyethylenimine produces higher
44 water fluxes than the high molecular weight polyethylenimine, suggesting that all membranes likely reject these bulky, branched polyelectrolytes and that the higher viscosity of the high molecular weight polyethylenimine increases ICP. In summary, dialysis membranes generally produce higher maximum water fluxes than forward osmosis membranes, and nanofiltration membranes had the lowest maximum water fluxes.
A sucrose draw solution combined with the largest dialysis membrane (3.5-5.0 kDa MWCO) produces the most water flux of any membrane/draw molecule combination. However, the large
MWCO of this membrane limits the size of analytes that can be concentrated. For most proteins, this size limit is acceptable, but analytes such as metabolites, signaling molecules, or hormones are much too small. An NaCl solution combined with the tightest membrane (FTS) produces a comparably high water flux without the same size limitation for potential analytes.
While sucrose could also be used with the FTS membrane, NaCl is particularly advantageous because of its lower viscosity and ubiquity in biofluids. Therefore, NaCl and the FTS membrane were determined to be the preferred approach for the remainder of experimentation. It was initially speculated that the FTS membrane would excel, because it was specifically designed as a pouch for quickly generating potable water by forward osmosis even in a muddy puddle (e.g., must block small molecule biotoxins, etc.).
Validity of performance evaluation. This empirical approach of comparing water flux produced by combinations of membranes and draw molecules has some limitations. For example, if the membrane only partially rejects a draw molecule, water flux would start high and then quickly decrease as the draw molecule equilibrates on either side of the membrane, resulting in low average water flux. This is indistinguishable from the case where a draw molecule fouls the membrane, producing a high initial water flux followed by a quick decline.
However, both of these cases are undesirable, and so an exact explanation for poor performance is not needed. Also, this experimental setup includes performance limiting effects
45 such as ECP and ICP. Therefore, while not perfect, this approach provides realistic, real-world performance benchmarks for the selection of ideal membrane and draw molecule combinations.
3.3.2 Preconcentration device
In this sub-section, two operating regimes of preconcentration devices are defined to better explain results in the sections to follow. Next, results of the concentration profile experiments are discussed to better understand how devices concentrate with respect to position in the channel and input flowrate. Resistance to fouling is then inferred using the results of the biological relevance experiments. Finally, device design considerations are discussed.
Operating regimes. To aid the description of device function, two distinct operating regimes are defined: preconcentration-limited and transport-limited. During the preconcentration-limited regime, the sample undergoes preconcentration throughout the channel. At higher flowrates, the sample never reaches osmotic equilibrium, resulting in incomplete concentration (Fig. 3.4).
Interestingly, while the maximum osmotic gradient exists at the beginning of the channel, not much preconcentration occurs there. Rather, as the sample concentrates, the flowrate decreases and allows more time for water to be removed, increasing the rate of preconcentration further down the channel. However, as the sample osmolarity nears the draw solution osmolarity, the driving osmotic gradient decreases, and the rate of concentration again slows. Finally, the sample osmolarity equilibrates with the draw solution osmolarity, and preconcentration completes. The sample then exits at a fraction the input flowrate (1/10 for 10x,
1/100 for 100x, etc.). These slow, fast, then slow preconcentration rates result in the visually obvious sigmoidal profiles seen in Fig. 3.4. However, at low flowrates, samples do reach osmotic equilibrium. The transport-limited regime occurs after the sample has fully concentrated, reached its lowest flowrate, and then must merely continue to travel at this low flowrate until reaching the outlet.
46 Results of the concentration profile experiments. The in-silico model also predicts that the
NaCl concentrations throughout the channel follow a sigmoidal profile (Fig. 3.4). The membrane resistance calculated by the in-silico model is 3e12 Pa-s/m, and the standard errors of the regression for each flowrate in ascending order are as follows: 4.7 %, 3.9 %, 3.7 % 3.7 %, 2.1
%, and 1.1 %.
Results of the biological relevance experiments. The outlet concentrations at various input flowrates of both glucose and BSA also follow the sigmoidal profile predicted by the in-silico
Figure 3.5. The concentrations at various input flowrates are shown for two biologically-relevant analytes: glucose and BSA. Concentrations are predicted by the in-silico model and plotted for each analyte.
47 model (Fig. 3.5). Based on the in-silico model, input flowrates less than 1.9 µL/min for glucose and 2.5 µL/min for BSA will allow samples to be concentrated to >99 % of the expected 10x.
Experimentally, an input flowrate of glucose at 2 µL/min and BSA at 3 µL/min yields 97% and
77% completion, respectively. The membrane resistance calculated by the in-silico model is
1e12 Pa-s/m for glucose and 7.6e12 Pa-s/m for BSA, suggesting slight membrane fouling by
BSA. The standard error of the regression is 3.4 % for glucose and 4.7 % for BSA.
BSA may somewhat adhere to the channel walls, reducing recovery at the outlet thus causing a lower concentration to be measured. The lower input flowrates move slower through the channel and concentrate to a greater extent, causing further slowing. As a result, these lower flowrates likely experience greater BSA adhesion to the channel walls. This could explain the model’s difficulty in fitting this data. Sample analyte adhesion to the channel walls is a potential problem, but existing pre-treatment techniques82 could help but were not examined in this work.
It should be noted that neither glucose nor BSA was experimentally shown to reach 100% of the expected concentration, because neither was run at flowrates less than 2 µL/min and 3 µL/min, respectively. The time needed to collect an adequate sample at these lower flowrates would be prohibitively long without modifying device parameters.
Design considerations. Based on the in-silico model, membrane surface area is proportional to the maximum input flowrate that is capable of concentrating to 90 %. In Fig. 3.6a, this relationship is shown for two concentration gradients: 10x (50-500 mM) and 100x (50-5000 mM). The slope of the 100x gradient (47 µL/min per cm2) is roughly ten times greater than that of the 10x gradient (3.7 µL/min per cm2). While flowing at a flowrate less than the threshold flowrate will yield concentrations greater than 90 %, the time spent in the channel will increase.
For example, preconcentration of samples flowing at 10 µL/min by 90x and 1 µL/min by 9x requires a membrane surface area of just 0.22 cm2 and 0.27 cm2 and takes only 1.2 mins and
8.1 mins, respectively (Fig. 3.6b). However, by using a similar membrane surface area as the
48 straight channel (0.75 cm2), it would take both almost an hour to reach the outlet. The solid lines in Fig. 3.6b represent the short preconcentration-limited regime while the dashed lines represent the long transport-limited regime. Matching the membrane surface area to the range of flowrates that a device will experience during operation is critical to minimize the latency.
Figure 3.6. (a) The in-silico model predicts a linear proportionality between membrane surface area (cm2) and the maximum input flowrate that is capable of concentrating to 90 %. (b) The time (min) needed to for the sample to travel over a given membrane surface area (cm2) as predicted by the in-silico model.
49 3.4 Conclusions
This chapter demonstrated that forward osmosis can be used to preconcentrate samples within a microfluidic device both continuously and in real-time. The selected membrane showed good resistance to fouling from the biologically-relevant solutions tested. In addition, the developed in- silico model predicts device functionality based on design parameters, providing an engineering toolkit for future designs. This toolkit was used to optimize the original device, allowing for the development of an inexpensive, ready-for-manufacturing prototypes that preconcentrate 10-
100X within minutes.
Future Work. Preconcentration is just one of many other preprocessing steps such as desalting, buffering pH, removing interferents, and delivering reagents. For this reason, this preconcentration method may be combined with other steps to provide a complete solution for specific sensing applications. For example, concentrating analytes in a biofluid using a salt- rejecting membrane will also increase the salt concentration. Some chemical biosensors have altered performance when the salt concentration is outside of physiological ranges. For unbuffered fluids, pH changes could also be a challenge for some sensors. Fortunately, microfluidic approaches have been developed for modulating salt concentration83, and methods of buffering pH such as the addition of a buffer can be easily scaled down to a microfluidic device. Now that a microfluidic solution has been developed for continuous, quantifiable, and simple preconcentration, next-generation lab-on-a-chip devices targeting high-impact applications only need to focus on integrating specific sensing modalities, and if needed, additional modular sample preprocessing steps.
50 Chapter 4: Increase concentration of target biomarkers in sweat using chemical permeability enhancement and reverse iontophoresis
This chapter includes adapted text and figures from the following article:
A. Jajack, M. Brothers, G. Kasting, J. Heikenfeld. Enhancing biomarker flux into sweat. PLoS
One. Submitted.
4.1 Introduction
In this chapter, the biology of the sweat gland epithelium is exploited to enhance the biomarker flux into sweat. As described previously, many biomarkers of interest must pass into sweat via the paracellular pathway. However, tight junctions pinch neighboring cell membranes together, forming a barrier for biomarker enter. The dynamic nature of tight junctions allows some biomarkers to pass during normal modulation. If this modulation could be controlled, then so, too, could the paracellular permeability.
Figure 4.1. The addition of a chelating agent (EGTA) in the media of excised rabbit tracheal epithelium decreases the TER as measured using an Ussing chamber, reproduced15.
51 Several groups have artificially modulated the permeability of tight junctions in the intestinal epithelium to increase drug absorption12,13,16,84,85. Since tight junctions are calcium-sensitive, calcium chelators are effective tight junction modulators13. As seen in Figure 4.1, the addition of a chelating agent in the media of an excised rabbit tracheal epithelium triggered a decrease in trans-epithelial resistance (TER)86. In fact, evidence exists to suggest that chelators sequester calcium ions and then trigger endocytosis of tight junctional proteins14–16,18, opening up the paracellular pathway17 (Fig. 4.2). Therefore, this method of increasing paracellular permeability for intestinal drug delivery could also be used to improve the extraction of biomarkers in other epithelia such as the sweat gland epithelium.
Figure 4.2. Calcium depletion induces internalization of adheren and tight junctional proteins inside the cells of a T84 monolayer, reproduced17.
52 However, simply applying a chelator topically would not be effective, as the rate of diffusion into the gland would not be sufficient to counter the advective flow of sweat out of the gland. Most chelators are negatively charged in the pH range of sweat, and so instead, iontophoresis can be used to selectively drive chelators down into the lumen of the sweat gland. An additional benefit, the RI needed to drive a negatively-charged chelator also happens to induce electroosmotic flow of the surrounding ISF into sweat through the paracellular pathway22 (Fig 4.3). The negatively-charged carboxylic acid moieties within glycans and proteins on the cell surface provide the electrical double layer needed for electroosmosis. This electroosmosis has been
Figure 4.3. Conceptual model for glucose flux under various conditions: normal, RI + control (acetate), and RI + chelator (citrate). Normal. Tight junctions undergo constant and spontaneous remodeling. While this can cause direct aqueous paths to form momentary, the majority of glucose likely diffuses from ISF to sweat between opening and closing compartments that fluctuate between the tight junctional strands. RI + control (acetate). Electroosmosis
(enlarged) is the flow of fluid resulting from an applied electric field in the presence of an electrical double layer. The fixed negative charges needed to establish the double layer are due to the carboxylic acid moieties within glycans and proteins on the cell surface. Electroosmosis brings in biomarker-rich ISF through the direct aqueous paths that form spontaneously due to tight junction remodeling. RI + chelator (citrate). Chelation triggers reversible endocytosis of tight junctions (enlarged), producing many more direct aqueous paths, which close upon reintroduction of calcium. Therefore, it is expected that electroosmosis with chelation will facilitate much higher glucose flux compared to electroosmosis alone. Skin model has been adapted9.
53 previously used for diluted glucose extraction through the skin in an FDA-approved commercially-available product, the GlucoWatch Biographer87–89.
As depicted in Figure 4.3, it is speculated that RI triggers electroosmosis through the direct aqueous paths that form spontaneously due to tight junction remodeling. Also, it is speculated that chelation triggers endocytosis of tight junctions in the sweat gland, producing many more direct aqueous paths. Therefore, it is expected that RI in the presence of chelators will facilitate much higher flux of biomarker-rich ISF into sweat compared to RI without chelators.
Furthermore, because sweating provides an outward advective flow, the constant and 10X larger current (0.5 mA/cm2) of the GlucoWatch Biographer is not required.
Sweat glucose flux was examined under various conditions: normal (see Chapter 2) and RI both with and without a chelator (Fig. 4.3). A skin-safe20 calcium chelator, citrate (lemon/lime juice), is used as the permeability enhancer13, and acetate as the control. This study is the first to use a permeability enhancer for extraction rather than drug delivery and the first to combine permeability enhancement with electroosmosis. This proof of principal, demonstrated by the marked increase in glucose flux, shows promise for further optimization of paracellular permeability enhancement with electroosmosis as a viable option for overcoming detection challenges in sweat biosensing and enabling possibilities for enhanced drug delivery.
4.2 RI likely does not induce electroporation under study conditions
Electroporation refers to the creation of nanopores in cell membranes to increase transcellular permeability using an electric field (Fig. 4.4). Under normal conditions, the membrane fluctuates due to mechanical stresses resulting in nanoscale, hydrophobic pores which open and close rapidly throughout the membrane (Fig. 4.4a and 4.4b). When a sufficient transmembrane potential is applied, the energy state of the hydrophilic pore is lowered sufficiently such that energy barrier for a hydrophobic pore to shift to a hydrophilic pore can be overcome by the thermal energy of the system90,91. Electroporation is used medical for gene therapy as well as
54 Figure 4.4. Molecular model of pore formation in the cell membrane as a result of electroporation. tumor ablation92–95. Depending on the strength and duration of the electric field, electroporation can produce reversible pores, irreversible pores, or irreversible pores along with thermal damage to the cell31,96. Yarmush et al. has characterized the field strength and exposure durations that will trigger reversible electroporation. Electroporation has been associated with RI voltages to temporarily modulate the transcellular permeability of the ductal epithelium of the eccrine sweat glands97,98. However, in this study, the current density is much smaller, only 40
µA/cm2 over a 6.1 cm2 area. Using a gland density of around 100 glands/cm2 9,45, estimates of the surface area of the sweat gland lumen9, conductivity of the sweat gland epithelium99, and
Ohms law, then the sweat gland ductal epithelium can only support a voltage drop of about 50 mV. This is insufficient to trigger electroporation since the voltage required is at least an order of magnitude greater98,100. In addition, if electroporation does occurs, the resistance would drop even further, and so too will the voltage drop. On top of this, the method of increasing paracellular permeability will further reduce the resistance and thus the voltage drop. Therefore, the sweat gland epithelium is too conductive and thus not conducive to support electroporation under these study conditions.
55 4.3 Experimental methods
4.3.1 Glucose flux under modified conditions
The same procedure of measuring sweat rate, blood glucose, and collecting sweat as described in Chapter 2 is repeated for 8, 20-min collection periods: two before treatment, four with treatment, and two after treatment (Fig. 4.5). The treatment was iontophoresis of either acetate which should induce only electroosmotic flow or citrate which should induce both paracellular permeability enhancement and electroosmotic flow. The absorbent collection disks were pre- wetted with 30 μL of either a 500 mM acetate (pH 5.5) or citrate (pH 6.5) solution. Then, the pre- wetted disk was placed into the testing device and pressed against the testing area using a conductive attachment. Using the same counter electrode used for sweat stimulation, a lab- grade potentiostat drives the acetate or citrate into the skin by applying a constant current density of 0.04 mA/cm2 for the entire 20-min collection period. For the two collection periods before and after treatment, the absorbent disks were still pre-wetted, but no current was applied.
Figure 4.5. Block diagram of experimental methods for determining glucose flux under modified conditions.
4.3.2 Statistical analysis
Since each participant experienced both treatment types but on separate days and on separate arms, paired t-tests were used to examine the difference between acetate and citrate at each time point. Pairwise comparisons between time points of the same treatment (acetate or citrate) were conducted with a conservative Bonferroni correction.
4.3.3 Safety of chemical permeability enhancer
56 The chemical permeability enhancer used in this study, citric acid, was purposefully selected for its widespread use in topical cosmetic products and foods. The concentrations used in this study are less than the 10% concentration used in cosmetic formulations as reported by the
Cosmetic Ingredient Review. Citric acid has been designated as generally recognized as safe
(GRAS) by the FDA. It has also been approved for use in antacids as an active ingredient with a max daily dose of 8 g20.
4.3.4 pH considerations
The voltages applied by the electrodes may exceed the standard potential of water (-1.23 V) causing the electrolysis of water. At the cathode, reduction of water will occur to produce
− − hydrogen gas and hydroxide ions, increasing the basicity (2 H2O(l) + 2e → H2(g) + 2 OH (aq)).
At the anode, oxidation of water will occur to produce oxygen gas and hydrogen ions, increasing
+ − the acidity (2 H2O(l) → O2(g) + 4 H (aq) + 4e ). Due to the concentration of chloride ions in sweat, hypochlorite (bleach) will also likely be produced. It is important to consider how applied voltages will affect the pH to avoid skin irritation.
Iontophoretic delivery of the sweat stimulant will follow the Nanoduct protocol41. Since the
Nanoduct system is approved for use in neonates, pH was not a concern. However, iontophoresis of the chemical permeability enhancer was maintained below the total charge delivered by the Q-Sweat (600mC). For example, starting at pH 5 (physiological sweat pH), a charge of 600mC delivered via iontophoresis will result in the electrolysis of water such that the anode will be at pH = 0.2 and the cathode will be at pH = 13.8.
The Occupational Safety and Health Administration (OSHA) classifies substances as corrosive
(produces “irreversible damage to the skin…following the application of a test substance for up to 4 hours”) only if it has a pH ≤ 2 or a pH ≥ 11.5 unless there is existing evidence that the substance is corrosive. In order to avoid the buildup of acidity or basicity, the collection disk was
57 pre-wetted with a buffer. In the case of the control, the buffer was acetate. In the case of the treatment, citrate served as both the buffer and chelating agent.
4.3.5 IRB protocol
This study was performed under University of Cincinnati IRB protocol study ID 2016-4769. The glucose flux experiment consisted of 6 healthy participants (5 male, 1 female) between approximately 20 to 40 yrs. old.
4.4 Results and discussion
For this study, participants were artificially stimulated to sweat. Then, sweat samples were collected from the same site on the forearm for eight time points: two before treatment, four with treatment, and two after treatment. The treatment consisted of RI of either acetate (control), which should induce only electroosmotic flow, or citrate, which should induce both paracellular permeability enhancement and electroosmotic flow. Sweat rate and blood glucose were measured for each collected sample. Total glucose flux—the product of the measured sweat glucose concentration and the measured sweat rate—is plotted for each condition (Fig. 4.6).
Figure 4.6. Total glucose flux measured before, during, and after treatments: iontophoresis of either acetate (black) which should induce only electroosmosis or citrate (blue) which should induce both paracellular permeability enhancement and electroosmosis (n = 4).
58 Citrate produced significantly higher glucose fluxes compared to acetate at both the final treatment period (6) and the first recovery time point (7), p < 0.05. No significant differences exist at other time points. This suggests that the paracellular permeability enhancement induced by the delivery of citrate takes time to develop and once developed takes time to recover.
No significant differences exist between any of the acetate time points. This means that the glucose flux is not improved by electroosmotic flow alone. However, with citrate, the second to last and final treatment time points (5 and 6) were each significantly different compared to both of the initial baseline time points (1 and 2), p < 0.05. In addition, the final treatment time point (6) was significantly different from the final post-treatment baseline time point (8), p < 0.05. This again suggests that it takes time for the paracellular permeability enhancement to develop and once developed the recovery is not instantaneous.
Now with an understanding of the total glucose flux under normal conditions from Chapter 2, the total glucose flux in the presence of either electroosmotic flow or paracellular permeability enhancement with electroosmotic flow can be examined in detail. In these cases, both the total glucose flux and total water flux are more complicated than in Eq. 2.3 (see Chapter 2). There are two routes of entry for fluid into sweat—natural sweating (푄푠푤푒푎푡) and active flow of ISF into the gland (푄퐼푆퐹). Natural sweating brings water into the sweat lumen via aquaporins in the secretory coil101. Assuming a moderate sweat rate and sweat gland density for the forearm, the
2 flow of water due to natural sweating (푄푠푤푒푎푡) is ~500 nL/min-cm .
푛퐿 푔푙푎푛푑푠 푛퐿 Eq. 4.1: 푄 = (5 ) (100 ) = 500 푠푤푒푎푡 푚𝑖푛 − 푔푙푎푛푑 푐푚2 푚𝑖푛 − 푐푚2
Alternatively, electroosmosis results in active flow of biomarker-rich ISF into the sweat gland
25 (푄퐼푆퐹). Using an average value from literature for the volume per charge relationship and the
2 current density selected for this study (0.04 mA/cm ), the flow of ISF into sweat (푄퐼푆퐹) is calculated to be only ~6 nL/min-cm2.
59 푛퐿 푚퐴 푛퐿 Eq. 4.2: 푄 = (150 ) (0.04 ) = 6 퐼푆퐹 푚퐴 − 푚𝑖푛 푐푚2 푚𝑖푛 − 푐푚2
Because 푄퐼푆퐹 is much smaller than 푄푠푤푒푎푡, its contribution to the total water flux can be ignored.
However, its contribution to glucose flux is significant, as the ISF that passes into sweat is biomarker-rich. The glucose flux due to electroosmosis of ISF (퐹푙푢푥퐼푆퐹) is proportional to the flow of ISF (푄퐼푆퐹), the concentration of glucose in ISF (푐퐼푆퐹), and a proportionality constant (푘) that represents some resistance to glucose diffusion by the paracellular pathway.
Eq. 4.3: 퐹푙푢푥퐼푆퐹 = 푘(푐퐼푆퐹)(푄퐼푆퐹)
Taken together, the following is a constitutive equation that describes all three experimental cases: normal (푝 = 푝푛표푟푚푎푙; 푘 ≈ 0), electroosmotic flow (푝 = 푝푛표푟푚푎푙; 푘 > 0), and paracellular permeability enhancement with electroosmotic flow (푝 > 푝푛표푟푚푎푙; 푘 ≫ 0).
퐹푙푢푥퐹푖푐푘′푠 + 퐹푙푢푥퐼푆퐹 푝(∆푐) + 푘(푐퐼푆퐹)(푄퐼푆퐹) Eq. 4.4: 푐푠푤푒푎푡 = = 푄푠푤푒푎푡 + 푄퐼푆퐹 푄푠푤푒푎푡 + 푄퐼푆퐹
푝(푐푏푙표표푑/퐼푆퐹) + 푘(푐푏푙표표푑/퐼푆퐹)(푄퐼푆퐹) ≈ 푄푠푤푒푎푡
Rewritten, the following equation shows that the product of the measured sweat glucose concentration (푐푠푤푒푎푡) and the measured sweat rate (푄푠푤푒푎푡) yields total glucose flux
(퐹푙푢푥퐹푖푐푘′푠 + 퐹푙푢푥퐼푆퐹).
Eq. 4.5: 푐푠푤푒푎푡 (푄푠푤푒푎푡) = 퐹푙푢푥퐹푖푐푘′푠 + 퐹푙푢푥퐼푆퐹 = 푝(∆푐) + 푘(푐퐼푆퐹)(푄퐼푆퐹)
≈ 푝(푐푏푙표표푑/퐼푆퐹) + 푘(푐푏푙표표푑/퐼푆퐹)(푄퐼푆퐹)
4.5 Conclusions
Since the goal of sweat biosensing is to measure the concentrations of biomarkers in blood or
ISF, there must exist a way to relate the measured concentration in sweat to that of blood or
ISF. While this is simple for normal sweat collection conditions, RI with chelators provides additional routes that must be considered: passive paracellular diffusion and electroosmotic flow
60 of biomarker-rich ISF. Paracellular permeability enhancement increases both routes making it difficult to determine the extent of increased paracellular permeability vs increased fraction of
ISF in sweat. Based on the high conductivity of the sweat gland epithelium and low applied current density, it is speculated that enhancement due to electroporation can be ignored (see discussion above). Individual responses to this treatment differed substantially as seen by the large error bars. As a result, calculating the original concentration in blood or ISF is challenging.
One possible work-around is a ratio-metric approach that compares the sweat concentration of a target biomarker to another biomarker which is relatively stable in blood or ISF. For example, the sweat glucose to albumin ratio may allow better correlation with blood glucose. Future work should be focused on resolving this correlation issue.
This study represents the first step in the development of permeability enhancement with electroosmosis. Not only does this method increase biomarker flux in sweat, and thereby concentration, making it easier to detect biomarkers with currently available biosensors, but it also holds substantial promise for more effective delivery of topically applied therapeutics.
61 Chapter 5: Conclusions
5.1 Summary
In 2016, a critical review of the fundamental challenges holding back non-invasive sweat biosensing identified analyte accessibility as one of seven core issues4. While several groups have developed sweat sensing devices (Fig. 5.1), none have taken into account all of the fundamental challenges. As a result, these devices lack sufficient time resolution and poorly correlate back to blood. Instead of adding yet another application demo to the crowded literature, this dissertation took a more pointed approach to tackle a fundamental challenge
Figure 5.1. Examples of sweat biosensing application demos in the literature. None of these devices address all of the fundamental challenges facing sweat biosensing, reproduced4.
62 facing the entire biofluid sensing community: analytes are diluted in excreted biofluids, making them difficult to detect by currently available biosensors.
This dissertation broke down this challenge into first understanding how biomarkers enter sweat
(Chapter 2) and then developing technological and biological solutions (Chapters 3 and 4).
Chapter 2 began by including a thorough review of the literature on analyte partitioning into sweat and identified the key routes of entry through the transcellular and paracellular pathways.
Through this understanding, a paradigm shift occurred in the way biomarkers in sweat should be analyzed. Sweat biosensors should be concerned with flux rather than concentration as flux is what is ultimately useful in correlating back to systemic concentrations. In addition, this chapter carefully studied the basal flux of glucose into sweat in a small population, providing the first evidence that sweat gland permeability to biomarkers may be consistent from person to person. Next, Chapter 3 demonstrated the first membrane-based, microfluidic preconcentration device that is continuous, quantifiable, simple, and capable of working with any analyte. In the process, a much needed characterization of the forward osmosis performance of a broad selection of membranes and draw molecules was conducted to select the optimal combination.
In addition, the predictive in-silico model not only described the data but serves as an engineering toolkit to optimize future designs. Finally, Chapter 4 cleverly exploited a biological phenomenon of the sweat gland epithelium to increase the flux of glucose into sweat in a safe and reversible way. Together, the insights from this dissertation directly impact the sweat biosensing field by addressing the problem of low biomarker concentration in sweat.
5.2 Applications
Other non-invasively accessible biofluids such as saliva and tears can also benefit from background research conducted in Chapter 2, since the ways in which biomarkers enter are similar. The core benefit of non-invasively accessible biofluids is that they can be sampled continuously and in real time, impacting several key areas of healthcare—prevention (activity
63 tracking, stress-level monitoring, over-exertion alerting, dehydration warning), diagnosis (early- detection, new diagnostic techniques), and management (glucose tracking, drug-dose monitoring).
While the preconcentration device presented in Chapter 3 was developed to increase the concentration of analytes in sweat, it has broad scale applicability, especially in the area of point-of-care diagnostic tests. These tests are used for detecting infectious diseases and are prevalent in retail pharmacies and doctors' offices doctor’s office. Their low-cost, portability, and short time required for testing make them ideal for infections, especially RSV, influenza, and
HIV. Early and accurate detection remains crucial to guide appropriate and timely treatment.
The problem is that these tests have limited detection ranges and are often simply not sensitive enough to detect early stages of infection. Increasing the sensitivity is difficult, research intensive, and cost prohibitive. However, if samples could be concentrated prior to testing, then a single solution could be applied across the board. Therefore, the preconcentration device is well suited for this application (Fig. 5.2).
The biological method of increasing biomarker flux into sweat, presented in Chapter 4, provides a safe and reversible way of overcoming the problem of diluted analytes in sweat. Beyond
Figure 5.2. Example of a point of care pregnancy test used as directed (top) as well as after concentrating with a preconcentration device (bottom). The left bar is the test line; the right bar is the control line. The test line on the preconcentrated sample is significantly darker.
64 analyte extraction, chelators can be used for drug delivery purposes as well. Since reverse iontophoresis is needed to push the negatively-charged chelator down into the sweat lumen, negatively-charged drugs could also be delivered iontophoretically in tandem with the chelator.
These drugs may be able to be driven through the open paracellular route against the electroosmotic flow of ISF into the sweat gland. This could allow large drugs, such as protein- based drugs, to enter through the sweat gland epithelium.
5.3 Future work
The critical finding that participants had similar sweat gland permeabilities to glucose (Chapter
2) requires additional investigation. The sample size was small, only including six people. A larger, more diverse study should be conducted as a follow up. In its current form, the preconcentration device (Chapter 3) requires the sample to be inputted using a syringe press to both drive the sample into the device and keep the flowrate below the threshold flowrate. This is not a problem for sweat, because the device can be coupled directly to an area of activated sweat glands. The hydrostatic pressure of the generated sweat would drive the sweat into the device, and biological limits on the sweat generation rates would keep the flowrate within bounds of the device’s capabilities. However, to expand to other markets such as point of care diagnostic tests, a simpler method of driving the sample through is needed. Figure 5.3 shows an example of a modified design that uses an absorption-limited wick rather than a draw solution.
Here the sample is driven by capillary action of the channel and is concentrated until the wick is
Figure 5.3. Operating principle of the next generation of the preconcentration device designed to be more compatible with point of care diagnostic tests.
65 saturated. The membrane still provides analyte rejection just as in the original device. Additional work should be conducted to develop this new approach that is easier to couple with point of care diagnostic tests. Finally, the breakthrough results from Chapter 4 demonstrating an increase in glucose flux using RI in combination with chelators needs additional optimization. It is not yet clear how far down into the sweat gland the chelator is driven. The current density used to drive the iontophoretic delivery of the chelator is conservative for safety reasons, but more current would likely be more effective at opening more paracellular pathways. More work needs to be done to strike a better balance between safety and effectiveness. The reversibility was demonstrated over the course of the experiment, but long-term safety should be studied.
For example, if 100% effectiveness was achieved, the cells would no longer be anchored together and would slough off. Also, individuals had vastly different extents of improvements in glucose flux. This makes it difficult to correlate measured concentrations back to systemic concentrations. Additional work should be done to address this issue.
The need for a more diverse demographic for both the determination of glucose flux under normal conditions (Aim 1) and under enhanced extraction conditions (Aim 3) is further emphasized by research indicating that epithelial permeability may be linked to aging effects102,103. In these studies, inessential absorption in rats/mice increased as a function of age.
The participants in the studies of this dissertation were within a relatively narrow age range— between approximately 20 to 40 yrs. old. It is postulated that a similar age-related permeability increase may also occur in humans and in other epithelia such as the sweat gland epithelium.
In addition to age-related changes in permeability, the circadian clock has been shown in mice to modulate intestinal absorption104. This dissertation began all experiments at the same time of day for all participants (9 AM), and all participants fasted prior to beginning the study. This same level of rigor should be repeated in all future work to prevent clock-related confounding factors when assessing permeability of the sweat gland epithelium.
66 5.4 Conclusions
Healthcare is undergoing a revolution—a switch from the practice of medicine to the science of medicine—promising better outcomes, equality, and efficiency105. But, for this this evidence- based future to be realized, a key element is needed: data. Non-invasive biosensing provides a way to interrogate the body continuously and in real time, allowing us to paint a clearer picture of both well and sick states, track progression of illness, and gauge effectiveness of treatments.
The core concept behind this dissertation is to enable the potential of non-invasive biosensing by addressing the difficulties in detecting biomarkers. However, after a rigorous look at the problem, it became clear that the issue is in not only overcoming detection limitations but also correlating measured concentrations with meaningful systemic ones. This dissertation addresses both issues. For the latter issue, a flux-based rather than concentration-based approach is presented with an understanding of the partitioning model of how biomarkers enter sweat. For the former issue, two breakthrough solutions—technological and biological—are presented. The technological solution has broad applicability to other biosensing areas, such as point of care diagnostic tests, and is well aligned with the vision of the future of healthcare.
Therefore, this dissertation not only impacts sweat biosensing but forges the path for continued innovations that will enable the future of healthcare.
67 References
1. Strimbu, K. & Tavel, J. A. What are Biomarkers? Curr Opin HIV AIDS 5, 463–466 (2011). 2. Bandodkar, A. J. & Wang, J. Non-invasive wearable electrochemical sensors: a review. Trends Biotechnol. 32, 363–371 (2014). 3. Lamb, K., Bradshaw, C. M. & Szabadi, E. The responsiveness of human eccrine sweat glands to choline and carbachol. Eur. J. Clin. Pharmacol. 24, 55–62 (1983). 4. Heikenfeld, J. Non-invasive Analyte Access and Sensing through Eccrine Sweat: Challenges and Outlook circa 2016. Electroanalysis 28, 1242–1249 (2016). 5. Regittnig, W. et al. Assessment of transcapillary glucose exchange in human skeletal muscle and adipose tissue. Am. J. Physiol. - Endocrinol. Metab. 285, E241–E251 (2003). 6. Moyer, J., Wilson, D., Finkelshtein, I., Wong, B. & Potts, R. Correlation Between Sweat Glucose and Blood Glucose in Subjects with Diabetes. Diabetes Technol. Ther. 14, 398– 402 (2012). 7. Lee, H. et al. A graphene-based electrochemical device with thermoresponsive microneedles for diabetes monitoring and therapy. Nat. Nanotechnol. 11, 1–30 (2016). 8. Kohler, S., Couderc, V., O’Connor, R. P., Arnaud-Cormos, D. & Leveque, P. A versatile high voltage nano- and sub-nanosecond pulse generator. IEEE Trans. Dielectr. Electr. Insul. 20, 1201–1208 (2013). 9. Sonner, Z. et al. The microfluidics of the eccrine sweat gland, including biomarker partitioning, transport, and biosensing implications. Biomicrofluidics 9, 31301 (2015). 10. Kurnik, R. T., Berner, B., Tamada, J. & Potts, R. O. Design and Simulation of a Reverse Iontophoretic Glucose Monitoring Device. J. Electrochem. Soc. 145, 4119–4125 (1998). 11. Anderson, J. M., Itallie, C. M. Van & Van Itallie, C. M. Physiology and Function of the Tight Junction. Cold Spring Harb. Perspect. Biol. 1, a002584–a002584 (2009). 12. Salama, N. N., Eddington, N. D. & Fasano, A. Tight junction modulation and its relationship to drug delivery. Adv. Drug Deliv. Rev. 58, 15–28 (2006). 13. Deli, M. A. Potential use of tight junction modulators to reversibly open membranous barriers and improve drug delivery. Biochim. Biophys. Acta - Biomembr. 1788, 892–910 (2009). 14. Gonzalez-Mariscal, L. et al. Role of calcium in tight junction formation between epithelial cells. Am. J. Physiol. 259, C978-86 (1990). 15. Bhat, M. et al. Regulation of tight junction permeability by calcium mediators and cell cytoskeleton in rabbit tracheal epithelium. Pharm. Res. 10, 991–7 (1993). 16. Tomita, M., Hayashi, M. & Awazu, S. Absorption-Enhancing Mechanism of EDTA, Caprate, and Decanoylcarnitine in Caco-2 Cells. J. Pharm. Sci. 85, 608–611 (1996). 17. Ivanov, A. I. Endocytosis of Epithelial Apical Junctional Proteins by a Clathrin-mediated Pathway into a Unique Storage Compartment. Mol. Biol. Cell 15, 176–188 (2003). 18. González-Mariscal, L., Tapia, R. & Chamorro, D. Crosstalk of tight junction components with signaling pathways. Biochim. Biophys. Acta - Biomembr. 1778, 729–756 (2008). 19. Lanigan, R. S. & Yamarik, T. A. Final report on the safety assessment of EDTA, calcium disodium EDTA, diammonium EDTA, dipotassium EDTA, disodium EDTA, TEA-EDTA,
68 tetrasodium EDTA, tripotassium EDTA, trisodium EDTA, HEDTA, and trisodium HEDTA. Int. J. Toxicol. 21 Suppl 2, 95–142 (2002). 20. Fiume, M. M. et al. Safety Assessment of Citric Acid, Inorganic Citrate Salts, and Alkyl Citrate Esters as Used in Cosmetics. Int. J. Toxicol. 33, 16S–46S (2014). 21. Santi, P. & Guy, R. H. Reverse iontophoresis — Parameters determining electroosmotic flow: I. pH and ionic strength. J. Control. Release 38, 159–165 (1996). 22. Leboulanger, B., Guy, R. H. & Delgado-Charro, M. B. Reverse iontophoresis for non- invasive transdermal monitoring. Physiol. Meas. 25, R35–R50 (2004). 23. Pikal, M. J. The role of electroosmotic flow in transdermal iontophoresis. Adv. Drug Deliv. Rev. 46, 281–305 (2001). 24. Pikal, M. J. Transport mechanisms in iontophoresis. I. A theoretical model for the effect of electroosmotic flow on flux enhancement in transdermal iontophoresis. Pharm. Res. 7, 118–26 (1990). 25. Pikal, M. J. & Shah, S. Transport mechanisms in iontophoresis. III. An experimental study of the contributions of electroosmotic flow and permeability change in transport of low and high molecular weight solutes. Pharm. Res. 7, 222–9 (1990). 26. Uitto, O. D. & White, H. S. Electroosmotic pore transport in human skin. Pharm. Res. 20, 646–52 (2003). 27. Proksch, E., Brandner, J. M. & Jensen, J.-M. The skin: an indispensable barrier. Exp. Dermatol. 17, 1063–1072 (2008). 28. El-Laboudi, A., Oliver, N. S., Cass, A. & Johnston, D. Use of microneedle array devices for continuous glucose monitoring: a review. Diabetes Technol. Ther. 15, 101–15 (2013). 29. Pinkus, H. Tape stripping in dermatological research. A review with emphasis on epidermal biology. G. Ital. Dermatol. Minerva Dermatol. 107, 1115–26 (1966). 30. Mitragotri, S. Devices for overcoming biological barriers: The use of physical forces to disrupt the barriers. Adv. Drug Deliv. Rev. 65, 100–103 (2013). 31. Yarmush, M. L., Golberg, A., Serša, G., Kotnik, T. & Miklavčič, D. Electroporation-Based Technologies for Medicine: Principles, Applications, and Challenges. Annu. Rev. Biomed. Eng. 16, 295–320 (2014). 32. Levin, J. & Maibach, H. The correlation between transepidermal water loss and percutaneous absorption: an overview. J. Control. Release 103, 291–299 (2005). 33. Grice, E. A. & Segre, J. A. The skin microbiome. Nat. Rev. Microbiol. 9, 244–253 (2011). 34. Boysen, T. C., Yanagawa, S., Sato, F. & Sato, K. A modified anaerobic method of sweat collection. J. Appl. Physiol. 56, 1302–7 (1984). 35. Picardo, M., Ottaviani, M., Camera, E. & Mastrofrancesco, A. Sebaceous gland lipids. Dermatoendocrinol. 1, 68–71 (2009). 36. Peng, R. et al. A new oil/membrane approach for integrated sweat sampling and sensing: sample volumes reduced from μL’s to nL’s and reduction of analyte contamination from skin. Lab Chip 16, 4415–4423 (2016). 37. Tsukita, S., Furuse, M. & Itoh, M. Multifunctional strands in tight junctions. Nat. Rev. Mol. Cell Biol. 2, 285–293 (2001). 38. Steed, E., Balda, M. S. & Matter, K. Dynamics and functions of tight junctions. Trends
69 Cell Biol. 20, 142–9 (2010). 39. Buono, M. J., White, C. S. & Connolly, K. P. Cholinergic sensitivity of the eccrine sweat gland in trained and untrained men. J. Dermatol. Sci. 4, 33–37 (1992). 40. Kim, J. et al. Noninvasive Alcohol Monitoring Using a Wearable Tattoo-Based Iontophoretic-Biosensing System. ACS Sensors 1, 1011–1019 (2016). 41. Desax, M.-C. et al. Nanoduct® sweat testing for rapid diagnosis in newborns, infants and children with cystic fibrosis. Eur. J. Pediatr. 167, 299–304 (2008). 42. Gerber, S. L., Cantor, L. B. & Brater, D. C. Systemic drug interactions with topical glaucoma medications. Surv. Ophthalmol. 35, 205–218 (1990). 43. Riedl, B., Beckmann, T., Neundörfer, B., Handwerker, H. O. & Birklein, F. Autonomic failure after stroke--is it indicative for pathophysiology of complex regional pain syndrome? Acta Neurol. Scand. 103, 27–34 (2001). 44. Coston, A. F. & Li, J. K.-J. Iontophoresis: Modeling, Methodology, and Evaluation. Cardiovasc. Eng. An Int. J. 1, 127–136 (2001). 45. Simmers, P., Li, S. K., Kasting, G. & Heikenfeld, J. Prolonged and localized sweat stimulation by iontophoretic delivery of the slowly-metabolized cholinergic agent carbachol. J. Dermatol. Sci. 89, 40–51 (2018). 46. Davis, C. P. Normal Flora. Medical Microbiology (University of Texas Medical Branch at Galveston, 1996). doi:NBK7617 [bookaccession] 47. Sender, R., Fuchs, S. & Milo, R. Revised Estimates for the Number of Human and Bacteria Cells in the Body. PLoS Biol. 14, e1002533 (2016). 48. Liu, G. et al. A wearable conductivity sensor for wireless real-time sweat monitoring. Sensors Actuators, B Chem. 227, 35–42 (2016). 49. Ely, M. R. et al. Surface contamination artificially elevates initial sweat mineral concentrations. J. Appl. Physiol. 110, 1534–1540 (2011). 50. Sheu, H. M., Chao, S. C., Wong, T. W., Yu-Yun Lee, J. & Tsai, J. C. Human skin surface lipid film: an ultrastructural study and interaction with corneocytes and intercellular lipid lamellae of the stratum corneum. Br. J. Dermatol. 140, 385–91 (1999). 51. Cork, M. J. et al. New perspectives on epidermal barrier dysfunction in atopic dermatitis: Gene–environment interactions. J. Allergy Clin. Immunol. 118, 3–21 (2006). 52. Cha, K. H., Jensen, G. C., Balijepalli, A. S., Cohan, B. E. & Meyerhoff, M. E. Evaluation of Commercial Glucometer Test Strips for Potential Measurement of Glucose in Tears. Anal. Chem. 86, 1902–1908 (2014). 53. Wu, T.-F., Cho, S. H., Chiu, Y.-J. & Lo, Y.-H. in Handbook of Photonics for Biomedical Engineering (eds. Ho, A. H.-P., Kim, D. & Somekh, M. G.) 87–121 (Springer Netherlands, 2017). doi:10.1007/978-94-007-5052-4_10 54. Kozel, T. R. & Burnham-Marusich, A. R. Point-of-Care Testing for Infectious Diseases: Past, Present, and Future. J. Clin. Microbiol. 55, 2313–2320 (2017). 55. Heikenfeld, J. et al. Wearable sensors: modalities, challenges, and prospects. Lab Chip 18, 217–248 (2018). 56. Gold, L. et al. Aptamer-Based Multiplexed Proteomic Technology for Biomarker Discovery. PLoS One 5, e15004 (2010).
70 57. Olson, W. C., Spitznagel, T. M. & Yarmush, M. L. Dissociation kinetics of antigen- antibody interactions: studies on a panel of anti-albumin monoclonal antibodies. Mol. Immunol. 26, 129–136 (1989). 58. Deen, M. J., Marsal, L. F., Moldovan, O. & Iñiguez, B. Graphene electronic sensors – review of recent developments and future challenges. IET Circuits, Devices Syst. 9, 446– 453 (2015). 59. Homola, J., Yee, S. S. & Gauglitz, G. Surface plasmon resonance sensors: review. Sensors Actuators B Chem. 54, 3–15 (1999). 60. Datar, R., Passian, A., Desikan, R. & Thundat, T. Microcantilever Biosensors. in 2007 IEEE Sensors 37, 5–5 (IEEE, 2007). 61. Lin, C. C., Hsu, J. L. & Lee, G. Bin. Sample preconcentration in microfluidic devices. Microfluid. Nanofluidics 10, 481–511 (2011). 62. Giordano, B. C., Burgi, D. S., Hart, S. J. & Terray, A. On-line sample pre-concentration in microfluidic devices: A review. Anal. Chim. Acta 718, 11–24 (2012). 63. de Jong, J., Lammertink, R. G. H. & Wessling, M. Membranes and microfluidics: a review. Lab Chip 6, 1125 (2006). 64. Eijkel, J. C. T. & van den Berg, A. Water in micro- and nanofluidics systems described using the water potential. Lab Chip 5, 1202 (2005). 65. Cai, Y. & Hu, X. M. A critical review on draw solutes development for forward osmosis. Desalination 391, 16–29 (2016). 66. Jun, B.-M., Nguyen, T. P. N., Ahn, S.-H., Kim, I.-C. & Kwon, Y.-N. The application of polyethyleneimine draw solution in a combined forward osmosis/nanofiltration system. J. Appl. Polym. Sci. 132, n/a-n/a (2015). 67. Garcia-Castello, E. M., McCutcheon, J. R. & Elimelech, M. Performance evaluation of sucrose concentration using forward osmosis. J. Memb. Sci. 338, 61–66 (2009). 68. Yasukawa, M. et al. Effect of molecular weight of draw solute on water permeation in forward osmosis process. Ind. Eng. Chem. Res. 54, 8239–8246 (2015). 69. Phuntsho, S., Hong, S., Elimelech, M. & Shon, H. K. Osmotic equilibrium in the forward osmosis process: Modelling, experiments and implications for process performance. J. Memb. Sci. 453, 240–252 (2014). 70. Cath, T. Y., Childress, A. E. & Elimelech, M. Forward osmosis: Principles, applications, and recent developments. J. Memb. Sci. 281, 70–87 (2006). 71. Miller, S. Verification of Fluid Technology Solutions, Inc. Products for Microbiological Rejection by an Independent Laboratory. (2016). 72. Alturki, A. A. et al. Removal of trace organic contaminants by the forward osmosis process. Sep. Purif. Technol. 103, 258–266 (2013). 73. Jin, X., Shan, J., Wang, C., Wei, J. & Tang, C. Y. Rejection of pharmaceuticals by forward osmosis membranes. J. Hazard. Mater. 227–228, 55–61 (2012). 74. Nicoll, P. G. Forward Osmosis – a Brief Introduction. Int. Desalin. Assoc. World Congr. Desalin. Water Reuse 1–27 (2013). 75. Energy, T. et al. HTI starts producing thin-film composite FO membrane. Membr. Technol. 2012, 5–6 (2012).
71 76. Gray, G. T., McCutcheon, J. R. & Elimelech, M. Internal concentration polarization in forward osmosis: role of membrane orientation. Desalination 197, 1–8 (2006). 77. McCutcheon, J. R. & Elimelech, M. Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis. J. Memb. Sci. 284, 237– 247 (2006). 78. Do, J., Zhang, J. Y. & Klapperich, C. M. Maskless writing of microfluidics: Rapid prototyping of 3D microfluidics using scratch on a polymer substrate. Robot. Comput. Integr. Manuf. 27, 245–248 (2011). 79. Khashayar, P. et al. Rapid prototyping of microfluidic chips using laser-cut double-sided tape for electrochemical biosensors. J. Brazilian Soc. Mech. Sci. Eng. 39, 1469–1477 (2017). 80. Walsh, D. I., Kong, D. S., Murthy, S. K. & Carr, P. A. Enabling Microfluidics: from Clean Rooms to Makerspaces. Trends Biotechnol. 35, 383–392 (2017). 81. Cha, K. H., Qin, Y. & Meyerhoff, M. E. Origin of Low Detection Limit and High Selectivity of Roche Accu-Chek Test Strips that Enables Measurement of Tear Glucose Levels. Electroanalysis 27, 670–676 (2015). 82. Banerjee, I., Pangule, R. C. & Kane, R. S. Antifouling Coatings: Recent Developments in the Design of Surfaces That Prevent Fouling by Proteins, Bacteria, and Marine Organisms. Adv. Mater. 23, 690–718 (2011). 83. Roelofs, S. H., van den Berg, A. & Odijk, M. Microfluidic desalination techniques and their potential applications. Lab Chip 15, 3428–3438 (2015). 84. Lindmark T, Kimura Y, A. P., Lindmark, T., Kimura, Y. & Artursson, P. Absorption enhancement through intracellular regulation of tight junction permeability by medium chain fatty acids in Caco-2 cells. J. Pharmacol. Exp. Ther. 284, 362–9 (1998). 85. Ghartey-Tagoe, E. B., Morgan, J. S., Neish, A. S. & Prausnitz, M. R. Increased permeability of intestinal epithelial monolayers mediated by electroporation. J. Control. Release 103, 177–190 (2005). 86. Garcia, S. O. et al. Wearable Sensor System Powered by a Biofuel Cell for Detection of Lactate Levels in Sweat. ECS J. solid state Sci. Technol. JSS 5, M3075–M3081 (2016). 87. Tierney, M. J., Tamada, J. A., Potts, R. O., Jovanovic, L. & Garg, S. Clinical evaluation of the GlucoWatch® biographer: a continual, non-invasive glucose monitor for patients with diabetes. Biosens. Bioelectron. 16, 621–629 (2001). 88. Potts, R. O., A. Tamada, J. & J. Tierney, M. Glucose monitoring by reverse iontophoresis. Diabetes. Metab. Res. Rev. 18, S49–S53 (2002). 89. Tierney, M. J., Kim, H. L., Burns, M. D., Tamada, J. A. & Potts, R. O. Electroanalysis of Glucose in Transcutaneously Extracted Samples. Electroanalysis 12, 666–671 (2000). 90. Chizmadzhev, Y. A. et al. Electrical Properties of Skin at Moderate Voltages. Biophys. J. 74, 843–856 (1998). 91. Weaver, J. C. & Chizmadzhev, Y. A. Theory of electroporation: A review. Bioelectrochemistry Bioenerg. 41, 135–160 (1996). 92. Chen, T., Langer, R. & Weaver, J. C. Skin electroporation causes molecular transport across the stratum corneum through localized transport regions. J Investig Dermatol Symp Proc 3, 159–165 (1998).
72 93. Dev, S. B. S. B., Rabussay, D. P. D. P., Widera, G. & Hofmann, G. A. G. A. Medical applications of electroporation. IEEE Trans. Plasma Sci. 28, 206–223 (2000). 94. Prausnitz, M. R. A practical assessment of transdermal drug delivery by skin electroporation. Adv. Drug Deliv. Rev. 35, 61–76 (1999). 95. Kotnik, T. et al. Electroporation-based applications in biotechnology. Trends Biotechnol. 33, 480–488 (2015). 96. Weaver, J. C., Smith, K. C., Esser, A. T., Son, R. S. & Gowrishankar, T. R. A brief overview of electroporation pulse strength–duration space: A region where additional intracellular effects are expected. Bioelectrochemistry 87, 236–243 (2012). 97. Kasting, G. B. & Bowman, L. A. Electrical Analysis of Fresh, Excised Human Skin: A Comparison with Frozen Skin. Pharm. Res. An Off. J. Am. Assoc. Pharm. Sci. 7, 1141– 1146 (1990). 98. Chizmadzhev, Y. A. et al. Electrical properties of skin at moderate voltages: contribution of appendageal macropores. Biophys. J. 74, 843–856 (1998). 99. Brayden, D. J., Cuthbert, A. W. & Lee, C. M. Human eccrine sweat gland epithelial cultures express ductal characteristics. J. Physiol. 405, 657–75 (1988). 100. DeBruin, K. A. & Krassowska, W. Modeling Electroporation in a Single Cell. I. Effects of Field Strength and Rest Potential. Biophys. J. 77, 1213–1224 (1999). 101. Bovell, D. The human eccrine sweat gland: Structure, function and disorders. J. Local Glob. Heal. Sci. 2015, 5 (2015). 102. Katz, D., Hollander, D., Said, H. M. & Dadufalza, V. Aging-associated increase in intestinal permeability to polyethylene glycol 900. Dig. Dis. Sci. 32, 285–288 (1987). 103. Ma, T. Y., Hollander, D., Dadufalza, V. & Krugliak, P. Effect of aging and caloric restriction on intestinal permeability. Exp. Gerontol. 27, 321–333 (1992). 104. Pan, X. & Hussain, M. M. Clock is important for food and circadian regulation of macronutrient absorption in mice. J. Lipid Res. 50, 1800–13 (2009). 105. Khosla, V. 20-percent doctor included & Dr. Algorithm: Speculations and musings of a technology optimist. (2016). at
73