INOM
EXAMENSARBETE TEKNIK,
GRUNDNIVÅ, 15 HP ,
STOCKHOLM SVERIGE 2016
Functionalized capillaries for
label-free bio-molecule-detection
utilizing the streaming current
method
MATILDA TAUBE
KARL STENLUND
KTH
Functionalized capillaries for label-free
bio-molecule-detection utilizing the streaming
current method
KARL STENLUND, MATILDA TAUBE
Bachelor’s Thesis at ICT Supervisor: Apurba Dev
iii
Abstract
This report aims to make a contribution in optimizing a bio-sensor which exploits the changes in streaming current in a fused silica capillary as tar-get molecules bind to immobilized receptors on the inner surface of the cap-illary. In this study we investigate the temporal characteristics of the sensor response as a function of capillary-dimension for different concentrations of target biomolecules. Using Heceptine-z domain interaction pairs and silica capillaries with different inner diameters in the range of 25-55 μm and length in the range of 3-9 cm we show that the reduction in capillary cross-section re-sults in a faster sensor response. The rere-sults show qualitative agreement with an analytical model grounded on diffusion-based mass transport under laminar flow condition and offer possibility to achieve a lower limit of detection with thinner capillaries.
Abstract
Contents
Contents iv
1 Introduction 1
2 Theory 3
2.1 Electric Double Layer . . . 3
2.2 Streaming Current . . . 3
2.3 Apparent Zeta Potential . . . 4
2.4 Diffusion time . . . 5
3 Method 7 3.1 Experimental Setup . . . 7
3.2 Procedure . . . 8
4 Results and Discussion 11 4.1 Control Values . . . 11
4.2 Detection Experiments . . . 12
4.3 Conclusion . . . 15
4.3.1 Future Implementations . . . 16
4.3.2 Optimizing the system . . . 16
4.3.3 Choice of Antibody Protein Pair . . . 17
4.3.4 Issues with our Data . . . 17
A Appendix 19 A.1 Debye Length . . . 19
A.2 Pressure Calculation . . . 19
Bibliography 21
Chapter 1
Introduction
A point-of-care device which would only need a small droplet of blood to show which diseases and infections you are suffering from would be a fine thing. It would be inexpensive, usable by professionals and patients alike in homes and hospitals. This has been the vision that we have kept in mind when working on this project, when optimizing a modern bio-molecule sensor with implementations from the latest dis-coveries within nanotechnology and biochemistry. Many efforts and projects are working towards this goal [2, 5], and market demands such a sensor to be reliable, easy to use and inexpensive. It should be easy to adapt to detect a wide range of proteins and molecules, or have multiplexed detections. This need has driven tech-nology toward developing a lab-on-chip compatible biosensor and thereby offering possibilities for point-of-care applications. One promising way of arriving at such a sensor is by using semiconductor based nanotechnology, as it can offer modules with nanometer resolution arrays [10, 3]. This decrease in size of sensory components increases its sensing range down to picomolar levels of the target protein as we shall see in this report.
When designing sensors at the micrometer scale, it is vital that we have a reliable system to contain our sample and transport it without contamination towards the target while only requiring a small sample to work on. Therefore, we arrive at the conclusion that microfluidic implementations are required [9], which can offer all previously stated demands with the addition of parallel testing on the same sample, to detect multiple target molecules. The microfluidic systems also greatly minimize the distance which our targets need to diffuse until meeting with the liquid/solid interface where the reaction takes place. Although both the nanoelectronic sensors and the microfluidic systems have shown great promises, their integration has been a threshold to cross [1]. One challenge is the interference of the electrokinetic effect, which occurs due to the flow of electrolyte in a microfluidic channel.
In this thesis we investigated the temporal characteristics of such a capillary sensor and the dependence of response time on the capillary cross-section area. In these experiments we are forcing the liquid to move by applying a hydrodynamic pressure which causes a potential to build up [8]. The interaction is more specifically
2 CHAPTER 1. INTRODUCTION
Chapter 2
Theory
The main part of the theory in this study is based in the area of electro kinetics and especially the streaming potential phenomenon that occurs when pressure driven fluid moves through a capillary.
2.1
Electric Double Layer
Charged particles and atoms contained in a colloid liquid will when in contact with a charged surface create an Electric Double Layer (EDL). Composed of the charged surface and a particle layer bonded to the surface due to chemical inter-actions. The surface charge will create an electrostatic field that will attract the opposite charged particles, or counterions. These ions will either be bonded to the surface or stay there by the electric force creating a shielding layer closest to the surface called the Stern Layer which will partially weaken the electric field. By only partial shielding, the electric field creates a gradually decreasing concentration of counterions, the diffuse layer, and further away from the surface the number of counterions equals the number of co-ions (the ions with the same charge as the sur-face). The length where the surface charge no longer can interact with the liquid is called the Debye Hückel length, this length is shown in figure 2.1 and can be calculated using equation 2.1.
λd = 1 q 4πlb P iρizi2 (2.1)
This length is dependent on P
iρizi2 which is the ionic concentration and the
Bjerrum length lbwhich in aqueous solutions at room temperature is 0.7 nm [13, 11].
2.2
Streaming Current
By applying a hydrodynamic pressure to a capillary containing a liquid where an electric double layer is present, the liquid will flow through the capillary and because
4 CHAPTER 2. THEORY
Figure 2.1. Simplified model showing the different parts and properties of the electric double layer at a charged surface.
of the non neutral charge distribution the counterions motion will start to build up a charge inside the capillary as seen figure 2.2. If there is a path available to electrically short circuit the beginning and end of the capillary, an electric current will start to level out the differences [4, 14]. This is the so called streaming current which is used in this thesis to calculate the apparent zeta potential. The current is proportional to the applied pressure to the capillary, the viscosity of the liquid, the cross section of the capillary and the potential at the shear plane.
2.3
Apparent Zeta Potential
In a system with EDL and hard and flat surfaces the potential at the shear plane shown in 2.1 is called zeta potential and denoted with the greek letter ζ. As a system with long molecules and soft proteins have neither hard or flat surfaces the position of the shear plane may vary in position and size which makes it hard to define exactly.
The zeta potential may be calculated using the Smoluchowski equation (eq 2.2) for known values of permittivity εε0 and viscosity η [6, 7, 12].
ζ = ∆Is ∆P η εε0 L A (2.2)
In this equation ∆Is is the difference in streaming current at two different
2.4. DIFFUSION TIME 5
Figure 2.2. Descriptive figure of how the potential builds up radially.
The surface charge can change for two different reasons, one is if the pH changes inside the capillary, the other being if something is bonded to the surface.
The potential in the capillary is described by the thermodynamic surface poten-tial Ψ0 which linearly changes across the Stern Layer the potential of which being
denoted by Ψδ. After this the potential decays exponentially through the diffuse part following the formula:
Ψ = Ψ−κxδ (2.3) where κ is the Debye-Hückel parameter and x is the distance from the capillary surface [4]. The AZP lies right next to the potential at the Stern Layer and for a negatively charged surface like the one for a silica capillary the Stern Layer potential is positively charged as counterions bind to it.
2.4
Diffusion time
The theory we have for AZP does not account for a laminar flow. This is comple-mented by the streaming current theory, which both includes a laminar flow and a gradually increasing charge distribution in the direction of flow. Currently, neither of these formulas have time as a parameter. The main theory that we wish to test in this report is if a small cross-area gives a faster response, which is based on diffusion. We will here try to explain the main factors in this theory.
When a molecule enters a capillary with receptors immobilized to the inner surface, the molecule needs to diffuse a distance toward the inner capillary area (in radial direction) before being flushed out, to have a chance to bond to the inner surface. The molecules that will have time to diffuse to the bonding sites will enter the capillary within an outer shell of the liquid body, determined by the effective diffusion length (Lef f). Factors important in calculating Lef f are travel time along
6 CHAPTER 2. THEORY
Figure 2.3. A figure describing how the molecule with the longest path travels to the surface, where d is the diameter.
If all the molecules from this shell (Lef f) bind to the surface we get a num/min as a bonding speed. The number of molecules needed to cover the area compared to the number of molecules in the effective shell and its expected bonding time can give us an expected time for a specific concentration of molecules to cover the solid/liquid surface. This is an overestimation, as the number of bonding sites may be spread out and the probability of bonding taking place is less than 100%. Nevertheless, this is enough to get a qualitative understanding and is in good agreement with experimental results.
The travel time for the molecule can be estimated by solving 2.4 and 2.5.
L = t Z 0 1 4µ ∆p ∆x[R 2− (r +√Dt)2]dt (2.4) τ = (R − r) 2 D (2.5)
Where τ is the time it takes for a molecule to travel the longest trajectory to the surface without exiting the capillary, µ is the viscosity of the main liquid,∆p/∆x is the pressure gradient, r is the distance from the center, R is the capillary radius, D is the diffusion coefficient for IgG and L is the length if the capillary.
By using an approximation done in [1] we find that at least one hour is needed to saturate our different capillaries when using a concentration of 1 nm. On the other hand, as we have approximated and rounded of many factors this is an extreme case we have calculated on. There are much less sites where bonding can take place, and therefore will be saturated much faster.The functionalization is not perfect and the IgG is much bigger than the antibody immobilized, making it impossible for all sites to be bonded to if they are packed along the surface.
Chapter 3
Method
3.1
Experimental Setup
To be able to accurately run a liquid through a capillary the setup has to be rigid and without leakage. A 1 mL precision syringe was used in a programmable infu-sion syringe pump which was connected to the capillary using tubes and fitting for microfluidics as seen in 3.1. The Labsmith UPS pressure sensor was connected to the system using a three way cross connector between the syringe and the capillary. To measure the current two platinum tubes are connected, one between the pressure sensor and the capillary and one directly after the capillary. A sourcemeter is con-nected to the platinum tubes with two silver alligator clips. The computer is used to control the infusion pump as well as to collect the data from the pressure sensor and the source-meter. A Faraday cage was used to shield the setup from external electrical fields.
Figure 3.1. The microfluidic setup surrounding the functionalized capillary showing the position of the pressure sensors, the platinum tubes and the capillary.
8 CHAPTER 3. METHOD
3.2
Procedure
Testing the Setup
To make sure the setup is working properly a performance test was conducted using solutions containing a balance between HCl and NaOH with different pH. This has been done by many other researchers so we can easily compare our performance. The solution is pushed through a clean silica capillary until a constant streaming current is observed which is recorded as presented in 4.1. The solution is then switched to another pH and the difference in current is recorded. The different solutions are then run through the system for 20 minutes each. [6] [12]
Functionalization
The interaction proteins used are Staphylococcal protein A-derived domain Z which bind strongly to Immunoglobulin G type 1 which is an active component in a cancer treatment drug. To functionalize the capillaries they were first cleaned thoroughly using a solution containing 5 parts Milli-Q water, one part 30% H2O2 and one part 25% NH4OH, and rinsed with milli-Q water. This process leaves a thin oxide on the inside of the capillaries which then is coated with 3-aminopropyltriethoxysilane through silanization. This is followed by another rinsing using clean water and ethanol before the capillaries are stored in a nitrogen atmosphere until receptor immobilization. The receptor is injected into the capillaries and allowed to bind to the surface, the remaining receptors are deactivated and removed by rinsing the capillaries with 1x PBS [1].
Preparation for Protein Measurements
To be able to dilute the antibody in a liquid without harming them a cell friendly buffer is used, in this case phosphate buffered saline (PBS). The PBS buffer is freshly prepared in the lab and diluted using DI water to 0.1x as a higher concentrated buffer gives a short debye length. Because of the inhomogeneous nature of the immobilizing process of the receptor, there are slight gaps where untreated silica is unprotected. This may lead to non-specific interaction with many molecules in the solution, and is undesired. To prevent this a solution of casein is mixed using a casein protein powder mixed in PBS solution, which is then run through the capillary at a flow rate of 10 μL/ min for 30 minutes. This completely seals off any silica still exposed to the liquid in the capillary, while still having its functionalization.
Measurements
3.2. PROCEDURE 9
2 and 10 μL/ min. The flow rates are controlled by the computer program which alternates between them, measuring the current and pressure difference to calculate the apparent zeta potential using equation 2.2. As antibodies bind to the immobi-lized receptors the potential saturates and the PBS solution is injected once more to decrease the influence of non-specific bonded antibodies.
Control
Chapter 4
Results and Discussion
4.1
Control Values
Figure 4.1. Testing the setup with different pH solutions.
The results from the pH-experiments are presented in figure 4.1. The step dif-ference between the pH values corresponds well to previous experiments where the difference in zeta potential varies with 10 mV for a difference in pH by one unit. Our result is not as free from noise as desired, but it shows the expected behavior when switching between pH solutions that is required when calculating the AZP.
In figure 4.2 we present data from three measurements where the length of the capillary varies. The deviant points are disturbances caused by experimental error. The data is synchroized in such a way that the injection of antibody happens at the 20 min mark in all three datasets. When comparing the formula for the pressure
12 CHAPTER 4. RESULTS AND DISCUSSION
Figure 4.2. Length dependence
between the capillary ends (equation figure A.1) with the equation for the AZP (equation 2.2), we see that they both contain the variable for the length, and cancel out. However according to the theory, we do expect a linear increase in the signal strength. This is canceled out by the increased pressure in the capillary, which is increasing at the same rate as the signal would have increased. This can be seen in 4.2 where the response time difference is negligible, when comparing 3 cm to 9 cm long capillaries. We also assume that we have some leakage in the system, because when we look at the pressure values they are too similar to be true. This points to that our setup has a limit in what pressure it can contain, and leaks when that value is passed.
Figure 4.3 shows us what signal we are to expect without any functionalization in the capillaries, or if the capillary has become saturated; there is a faint signal when the protein is within the capillary. This is expected and disappears as soon as we flush with PBS-solution as the glass capillary surface and the antibodies are faintly charged. We get an interaction but no bonds creating a faint APZ. A good example of this phenomenon in the detection experiments is by looking at figure 4.4, the 1 nM line. At about 160 min the protein is switched back to an antibody free PBS solution, and the decrease in signal strength is visible.
4.2
Detection Experiments
4.2. DETECTION EXPERIMENTS 13
Figure 4.3. The signal strength where there is no functionalization of the capillary. There is a faint signal when protein flows in the capillary, at time 30-60 min in the figure. At time 0-30 and 60-100 min there is no protein in the capillary, only buffer solution. The data is from a 43 μm capillary and a 10 nM protein solution.
not impossible to see an increase in the APZ.
A clear trend is that smaller inner diameter gives faster response, and this makes us go towards smaller capillaries. In figure 4.5 we can clearly see how the 25 μm capillary has both a clearer and faster response.
14 CHAPTER 4. RESULTS AND DISCUSSION
Figure 4.4. Four different experimental datasets for the 25 μm capillary, with different concentrations of the antibody in a PBS solution. Arrows indicate where we switched back to PBS buffer from the protein solutions.
for the functionalization. This would increase the number of antibodies being able to bind to the surface, but it might not increase the rate at which the apparent zeta potential increases as that is related to the electric double layer. This would then not directly make the streaming current any higher as that is dependent on the radial influence of the electric double layer on the liquid flow. However this is something that needs to be tested in future studies.
What can also be seen from the results is that a smaller radius gives a much clearer measurement with less noise, due to the increase of current and pressure in the system which makes the noise from the pump and external sources negligible. One way to increase the accuracy would be to increase the length or the viscosity as seen in equation 2.2. However increasing either of these would increase the pressure equally and make the current smaller, so the optimal solution would be to find a point where the current and the pressure have the same amount of relative noise. As mentioned before, this becomes hard for smaller capillaries as the setup is not good enough to connect to too short capillaries.
The measured data is summarized in figure 4.6 as a function of concentration by taking the first incline of the zeta potential with time when injecting antibodies. All slope data are plotted together with their standard deviation. Lines are fitted to the data and grouped by capillary diameter.
Slope of Apparent Zeta Potential (mV/ min) Capillary diameter (μm)
0,31854 25
0,19185 43
4.3. CONCLUSION 15
Figure 4.5. The results from three experiments at the same antibody concentration but with different capillaries.The concentration in the experiments presented in the figure is 1 nM.
These values are the slopes from the linear approximations, as displayed in figure 4.6. We can clearly see a trend where the smaller capillaries give faster response and that the different slopes vary linearly from each other, both agreeing with previously stated theory.
The average noise in the potential measurement lies around 0.6 mV but can be as low as 0.3 mV, so any change measured being lower or equal to this would be indistinguishable and therefore not possible to detect. If a test is running for 120 minutes, we can see how great the change must be to be distinguishable from the noise, which is 0.3 / 120 = 0.0025 mV/ min. This would give a signal with the same strength as the noise for the optimal case so doubling this gives us 0.005 mV/ min for the optimal case and 0.01 mV/ min for the most common noise level during our experiments. With this approximation of noise, we predict that we can detect protein concentrations down to 30 pm , with the lowest amount of noise. We want to test this in the future, with a similar setup.
4.3
Conclusion
16 CHAPTER 4. RESULTS AND DISCUSSION
Figure 4.6. All slopes with a linear approximation of the trend.
sensitivtiy. We see lots of potential in this field and project, and in this concluding chapter we discuss the different possibilities and difficulties for this sensor.
4.3.1 Future Implementations
Being able to detect label free proteins down to picomolar concentrations opens up a whole new possibility for low cost immunosensors, including detection of low concentrations of proteins in blood. This would be a on-site method for quickly measuring exact concentrations of different antibodies, with the possibility for a multiplexed sensor. For this to be possible the system would have to be further developed to eliminate the noise appearing in both pressure and current. We expect to get a lower noise level in the pressure, when only checking with one flow rate since the system needs to stabilise after each switch in flow rate. This would be possible for a more complete device where the data collected is compared to other known data. If the setup was to be made smaller than currently, a shielding cage surrounding the whole system would be easy to make resulting in less noise in the current.
4.3.2 Optimizing the system
4.3. CONCLUSION 17
we designed a holder for the system, two plastic parts mounted on an aluminium beam to both stabilize the system and to align everything. The design was made in Autodesk fusion 360 and the parts were then cut in plastic using a CNC milling machine. While the system took a bit more time to set up, the problem with leakage did not occur as often and the possibility to move the system without putting strain on the valves was important in minimizing the leakages.
4.3.3 Choice of Antibody Protein Pair
In previous studies [1] other smaller proteins such as dibarnase, barstar, avidin and streptavidin have been used causing smaller response in the setup as the smaller size causes less interaction to the liquid when bonded to the surface. These proteins were about 50 kDa in size and the antibodies used in this experiment are around 150 kDa, with three times the size we see a much larger response [2]. For the research to be as practical as possible an antibody from a non-human source was modified to resemble the human IgG as much as possible.
4.3.4 Issues with our Data
Due to a lack of time and the amount of problems with the experimental setup, we could not use all of our data. There were several factors that forced us to completely discard certain data, such as:
• leakage in the system • pH of the PBS buffer • age of the casein
• faulty functionalization of capillaries • badly connected silver clamps
18 CHAPTER 4. RESULTS AND DISCUSSION
Chapter A
Appendix
A.1
Debye Length
The Debye length is a parameter describing the length of which a charge carrier’s electrostatic interaction is possible. This is not simply calculated with Coulomb’s Law as the carrier is screened by other ions and charge carriers. Equation 2.1 gives an approximated value for the Debye length using the concentration of ions, summing both concentration and valencies, and the Bjerrum length. The Bjerrum length is the longest distance between two elementary charges where they still interact with each other, this distance is 0.7 nm in an aqueous solution with low ionic concentration at room temperature. Because the Debye Length differs with the concentration of the solution, diluting the solution 1:10 will increase the Debye Length by √10 as seen in A.1.
A.2
Pressure Calculation
The capillary is a fused silica construction with an inner diameter smaller than a hair’s width, coated with a polyimide to protect it from breaking. Poiseuille’s Law tell us that the pressure difference ∆P between the capillary ends is described by:
∆P = F 8ηL
πr4 (A.1)
Concentration Debye length (nm)
1x 0.76 0.1x 2.41 0.01x 7.61
Table A.1. Debye length in different concentrations
20 CHAPTER A. APPENDIX
Diameter (μm) Cross section area (μm2) Mean pressure difference ∆P (kPa) 8 50.265 24000 25 490.874 256 43 1452.20 29 55 2375.83 16
Table A.2. Mean pressures in capillaries with different diameters
Bibliography
[1] Dev. Apurba, J. Horak, A. Kaiser, X. Yuan, A. Perols, P. Bj?rk, A. Eriks-son KarlsEriks-son, P. Kleimann, and J. Linnros. Electrokinetic effect for molecular recognition: A label-free approach for real-time biosensing. Biosensors and Bioelectronics, 82:55–63, August 2016.
[2] Piet Bergveld. Development, operation, and application of the ion-sensitive field-effect transistor as a tool for electrophysiology. IEEE Transactions on Biomedical Engineering, BME-19(5):342–351, September 1972.
[3] A. De, J. van Nieuwkasteele, E. T. Carlen, and A. van den Berg. Integrated label-free silicon nanowire sensor arrays for (bio)chemical analysis. Analyst, 138:3221–3229, March 2013.
[4] D. Fairhurst. An overview of the zeta potential - part 1: The concept. American Pharmaceutical Review, February 2013.
[5] J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H.-J. G?ntherodt, Ch. Gerber, and H.-J. K. Gimzewski. Translating biomolecular recognition into nanomechanics. Science, 288:316–318, April 2000.
[6] Brian J. Kirby and Ernest F Jr. Hasselbrink. Zeta potential of microfluidic substrates: 1. theory, experimental techniques, and effects on separations. Elec-trophoresis, 25:187–202, 2004.
[7] Brian J. Kirby and Ernest F Jr. Hasselbrink. Zeta potential of microfluidic substrates: 2. data for polymers. Electrophoresis, 25:203–213, 2004.
[8] Sabine. Koch, Peter. Woias, Leonhard K. Meixner, Stephan. Drost, and Hans Wolf. Protein detection with a novel isfet-based zeta potential analyzer. Biosen-sors and Bioelectronics, 14(4):413–421, April 1999.
[9] J. Melin and S. R. Quake. Microfluidic large-scale integration: The evolution of design rules for biological automation. Annual Review of Biophysics and Biomolecular Structure, 36:213–231, February 2007.
[10] E. Stern, James F. Klemic, David A. Routenberg, Pauline N. Wyrembak, Daniel B. Turner-Evans, Andrew D. Hamilton, David A. Lavan, and Mark A.
22 BIBLIOGRAPHY
FahmyTarek M. Reed. Label-free immunodetection with cmos-compatible semi-conducting nanowires. Nature, 445:519–522, 2006.
[11] E. Stern, F. J. Wagner, R. Sigworth, R. Breaker, T. M. Fahmy, and Mark A. Reed. Importance of the debye screening length on nanowire field effect tran-sistor sensors. Nano Letter, 11:3405–3409, November 2007.
[12] V. Tandon, S. K. Bhagavatula, W. C. Nelson, and Brian J. Kirby. Zeta potential and electroosmotic mobility in microfluidic devices fabricated from hydrophobic polymers: 1. the origins of charge. Electrophoresis, 29:1091–1101, 2008.
[13] A. Vacic, Jason M. Criscione, N. K. Rajan, E. Stern, T. M. Fahmy, and Mark A. Reed. Determination of molecular configuration by debye length modulation. Journal of the American Chemical Society, 133:13886–13889, August 2011. [14] He. Xie, T. Saito, and Michael A Hickner. Zeta potential of ion-conductive
TRITA -ICT-EX-2016:163
