Detection and Quantification of Pathogens, Proteins, and Molecules Using Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensors
A Thesis
Submitted to the Faculty
of
Drexel University
by
Gossett A. Campbell
in partial fulfillment of the
requirements for the degree
of
Doctor of Philosophy
April 2006
© Copyright 2006 Gossett A. Campbell. All Rights Reserved. ii
Dedications
To my mother,
Laverne Corroll Marshall
&
Father
Gossett A. Campbell Sr. iii
Acknowledgements
It was through the grace of God that I was able to acquire this great accomplishment.
Therefore, I first thank God for his blessing. I am grateful for the enormous support of
my advisor, Dr. Raj Mutharasan. Dr. Mutharasan has been the driving force of this
project. His guidance, ideas, experience, and commitment to quality research and hard
work have made this project a success. As a model researcher, I look up to him for
inspiration in my professional career.
Along the journey I was guided by my Ph.D. defense and thesis advisory committee,
so I would like to take the time to thank Dr. Charles Weinberger, Dr. Richard Cairncross,
Dr. Ryszard Lec, P. Mohana Shankar, and Dr. Margaret Wheatley for their valuable time,
and constructive criticism and suggestions. Their contributions improved the quality of
my dissertation. Also, I thank Dan Luu for the fabrication of sensor flow cells (SFC), and
the design of data acquisition programs that interface the computer and impedance analyzer. I appreciated the work of Dr. Scoles in the modification of the data acquisition program to simultaneously track multiple peaks, monitor peaks location, and increase data accuracy.
To my lab partners I say thanks for the laboratory support, suggestion, and ideas; this
includes Angela Labadessa, Angela Leung, Kishan Rijal, Raul Jackson, Andrew Detzel,
David Maraldo, and David DeLesdernier. I must extend thanks to people outside the
department such as Dr. Wan Shih for her assistance in the initial phase of the project in
the understanding of cantilever physics and Dr. Dee Breger’s Velcro for ESEM training,
Stephen Cox and his entire Louis Stokes Greater Philadelphia Region Alliance for
Minority Participation team at Drexel University for textbook supplies, and other iv financial support.
Special thanks to my Mom and Dad, sisters and brothers, my grandmother Elfreda
Campbell, aunts and uncles, Angela Labadessa, and entrepenuer Mr. Winston Williams for their encouragement, motivation, and for believing in me and my dream. Without these people my journey would be a lot more difficult to travel. Thank you all.
Finally, I would like to acknowledge the financial support of the Environmental
Protection Agency (EPA, Grant R8296041), the National Science Foundation (NSF,
Grant No. BES0120321), the National Institutes of Health (NIH, Grant 5R01EB000720), and the Department of Homeland Security/Department of Transportation (PA-26-0017-
00). v
Table of Contents
List of Tables ...... xvii List of Figures ...... xx Abstract ...... xxxi Chapter 1:Introduction...... 1
1.1 Background...... 1
1.2 Objectives ...... 2
1.3 Contributions ...... 4
Chapter 2:Literature review on cantilever sensors...... 5
2.1 Biosensors...... 5
2.2 Detection principles of cantilever sensors ...... 7
2.2.1 Deflection and surface stress ...... 7
2.2.2 Resonant frequency...... 8
2.3 Detection mechanisms of cantilever sensors ...... 9
2.4 Millimeter size versus micro and nanometer size cantilever sensors ...... 10
2.5 Application of cantilever sensors...... 11
2.5.1 Cantilever sensors for biochemical measurement...... 11
Chapter 3:Dynamics of piezoelectric-excited millimeter-sized cantilever (PEMC) sensors ...... 32
3.1 Piezoelectricity...... 32
3.2 General PEMC physics...... 33
3.2.1 Theory ...... 33 vi
3.2.2 Simulation...... 37
Chapter 4:PEMC sensor fabrication, characterization, and functionalization...... 41
4.1 General fabrication of stainless steel base cPEMC sensors...... 41
4.2 General fabrication of glass based cPEMC sensors...... 42
4.3 General fabrication of free-floating tip PEMC sensors ...... 42
4.4 Characterization of PEMC sensors...... 43
4.4.1 Resonant frequency...... 43
4.4.2 Mass change sensitivity ...... 46
4.4.3 Quality factor ...... 46
4.4.4 Functionalization techniques ...... 47
Chapter 5:Sensing of Liquid Level at Micron Resolution Using Self-Excited Millimeter- Sized PZT-Cantilever...... 57
5.1 Introduction...... 57
5.2 Cantilever Physics...... 58
5.3 Materials and Methods...... 59
5.3.1 Cantilever Fabrication...... 59
5.3.2 Experimental ...... 60
5.4 Results and Discussions...... 61
5.4.1 Resonance in Air...... 61
5.4.2 Selection of resonant mode...... 62
5.4.3 Effect of Immersion on Cantilever Resonance Frequency ...... 62
5.4.4 Effect of Immersion Depth change on Cantilever Resonance Frequency ...... 63 vii
5.4.5 Effect of Immersion Depth in a Volatile and Non-volatile Liquid...... 63
5.4.6 Experimental determination of the rate of liquid level change...... 65
5.4.7 Modeling of liquid level change ...... 65
5.5 CONCLUSIONS ...... 67
Chapter 6:Detection of Pathogen Escherichia coli O157:H7 Using Self-Excited PZT- Glass Microcantilevers...... 82
6.1 Introduction...... 82
6.2 Materials and Methods...... 83
6.2.1 Microcantilever Fabrication...... 83
6.2.2 Experimental Arrangement...... 84
6.2.3 Antibody Immobilization...... 84
6.2.4 Detection Experiments...... 85
6.2.5 Experimental Determination of Mass Sensitivity of Microcantilever ...... 85
6.2.6 Systematic Measurement Correction ...... 86
6.3 Results and Discussions...... 86
6.3.1 Resonant frequency in air ...... 86
6.3.2 Quality factor in air...... 87
6.3.3 Mass change sensitivity in air and under liquid immersion...... 88
6.3.4 Resonant frequency in liquid immersion ...... 89
6.3.5 Selection of flexural mode for detection...... 90
6.3.6 Detection of Pathogen E. coli O157:H7 ...... 91
6.3.7 Scanning Electron Micrograph ...... 92 viii
6.3.8 Kinetics of pathogen Attachment...... 92
6.3.9 Frequency response to different concentrations of E. coli O157:H7 in presence of a wild strain of E. coli...... 95
6.3.10 Frequency response to the release of the attached E. coli O157:H7...... 95
6.4 Conclusions...... 96
Chapter 7:Escherichia Coli O157:H7 Detection Limit of Millimeter-Sized PZT Cantilever Sensors is 70 cells/mL...... 109
7.1 Introduction...... 109
7.2 Sensor description...... 109
7.3 Experimental procedure...... 110
7.4 Results...... 110
7.5 Conclusion ...... 113
Chapter 8:Rapid Assessment of Escherichia coli by Growth Rate on Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensors...... 117
8.1 Introduction...... 117
8.2 Materials and Methods...... 119
8.2.1 Cantilever Dimensions...... 119
8.2.2 Cell culture preparation...... 119
8.2.3 Cantilever functionalization...... 119
8.2.4 Experimental Arrangement...... 120
8.3 Results...... 120
8.3.1 Resonance Spectrum in Air ...... 120
8.3.2 Resonance Spectra of the Functionalized Cantilever...... 121 ix
8.3.3 Active Growth Resonant Frequency Response...... 122
8.3.4 Modeling the Growth Kinetics...... 122
8.3.5 Stoichiometric and Sensitivity Analyses ...... 125
8.4 Conclusion ...... 127
Chapter 9:Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensors Detect Bacillus anthracis at 300 spores/mL...... 133
9.1 Introduction...... 133
9.2 Materials and Methods...... 135
9.3 Results and Discussion ...... 136
9.3.1 Resonance characterization of PEMC sensors...... 136
9.3.2 Selection of Antibody for the detection of Bacillus anthracis spores...... 137
9.3.3 PEMC Sensor Response to Spores of various concentrations ...... 138
9.3.4 Selective binding of Bacillus anthracis spores to antibody functionalized PEMC sensor ...... 141
9.3.5 Kinetics of antibody and spore binding ...... 142
9.4 Conclusion ...... 146
Chapter 10:Detection and Quantification of Proteins Using Self-Excited PZT-Glass Millimeter-Sized Cantilever...... 155
10.1 Introduction...... 155
10.2 Materials and Methods...... 157
10.2.1 Detection experiments...... 157
10.2.2 Protein quantification assay ...... 158
10.2.3 Systematic measurement correction...... 158 x
10.3 Results and Discussions...... 158
10.3.1 Characterization of PEMC sensor...... 158
10.3.2 Mass change sensitivity in air and under liquid immersion...... 159
10.3.3 Second flexural mode response to 1-mm liquid immersion ...... 160
10.3.4 Selection of flexural mode for detection...... 161
10.3.5 Binding of antibody to amine-terminal silane sensing glass surface...... 161
10.3.6 Kinetic modeling of protein binding to cantilever surface ...... 165
10.4 Conclusion ...... 169
Chapter 11:Monitoring of the Self-Assembled Monolayer of 1-Hexadecanethiol on Gold Surface at Nanomolar Concentration Using Piezo-Excited Millimeter-Sized Cantilever Sensor ...... 183
11.1 Introduction...... 183
11.2 Principle of Measurement...... 185
11.3 Materials and Methods...... 185
11.3.1 Cantilever fabrication...... 185
11.3.2 Reagents...... 186
11.3.3 Sample preparation ...... 186
11.3.4 Formation of alkanethiolate SAM on the gold-coated cantilever...... 186
11.4 Results and Discussions...... 187
11.4.1 Characterization of the PZT-gold coated macrocantilever ...... 187
11.4.2 Frequency response to 1-hexadecanethiol self-assembled monolayer formation...... 188
11.4.3 Characterization of the kinetics ...... 189 xi
11.5 Conclusion ...... 194
CHAPTER 12:Piezoelectric-excited Millimeter-sized Cantilever (PEMC) Sensors Measure Albumin Interaction with Self-assembled Monolayers of Alkanethiols having different Functional Head Groups ...... 202
12.1 Introduction...... 202
12.2 Materials and Methods...... 203
12.2.1 Cantilever fabrication...... 203
12.2.2 Reagents...... 203
12.2.3 Gold substrate, Monolayer formation, and HSA interaction ...... 204
12.2.4 PEMC sensor calibration ...... 205
12.3 Results and Discussions...... 206
12.3.1 Characterization of the PEMC sensor...... 206
12.3.2 SAM formation ...... 207
12.3.3 Human serum albumin adsorption ...... 209
12.3.4 Kinetics of HAS adsorption ...... 211
12.4 Conclusion ...... 212
Chapter 13:PEMC Sensor’s Mass Change Sensitivity is 20 pg/Hz under Liquid Immersion ...... 220
13.1 Introduction...... 220
13.2 Materials and Methods...... 221
13.2.1 PEMC Fabrication ...... 221
13.2.2 Antibody immobilization...... 221
13.2.3 Experimental arrangement and mass change sensitivity...... 222 xii
13.2.4 Detection experiments...... 223
13.3 Results and Discussions...... 223
13.3.1 Characterization of the PEMC sensors in air...... 223
13.3.2 Selection of mode for detection ...... 225
13.3.3 Mass change sensitivity in air and under liquid immersion...... 225
13.3.4 Detection of Group A Streptococcus (GAS) ...... 226
13.3.5 Scanning Electron Micrograph ...... 228
13.3.6 Kinetics of pathogen Attachment...... 229
13.4 Conclusion ...... 229
Chapter 14:Detection of Bacillus anthracis Spores and a model protein Using PEMC Sensors in a Flow Cell at 1 mL/min...... 237
14.1 Introduction...... 237
14.2 Materials and Methods...... 238
14.2.1 Chemicals...... 238
14.2.2 Flow cell fabrication ...... 238
14.2.3 PEMC sensor ...... 239
14.2.4 Experimental setup...... 239
14.2.5 Antibody immobilization in situ and antigen Detection ...... 240
14.3 Modeling fluid flow...... 240
14.4 Results and Discussion ...... 241
14.4.1 Resonance characterization of PEMC sensors...... 241
14.4.2 Effect of flow rates on resonant frequency ...... 241 xiii
14.4.3 Pressure map and velocity field ...... 243
14.4.4 Detection of Bacillus anthracis spore in SFC-2...... 244
14.4.5 Effect of flow rate on PEMC sensor response ...... 245
14.4.6 Protein detection and sensor regeneration ...... 247
14.4.7 Kinetics of spore binding ...... 248
14.5 Conclusion ...... 250
Chapter 15:Detection of Escherichia coli O157:H7 in Ground Beef Samples using Piezoelectric Excited Millimeter-Sized Cantilever (PEMC) Sensors...... 262
15.1 Introduction...... 262
15.2 Materials and Methods...... 263
15.2.1 Chemicals...... 263
15.2.2 PEMC Preparation ...... 263
15.2.3 Sample preparation ...... 264
15.2.4 Plating and counting of bacteria ...... 265
15.2.5 Experimental arrangement and procedure ...... 266
15.2.6 Scanning electron microscopy ...... 266
15.3 Results and discussions...... 267
15.3.1 Resonance characterization of PEMC sensors...... 267
15.3.2 Initial evaluation of PEMC sensor...... 268
15.3.3 Detection of E. coli O157:H7 in meat-free broths...... 269
15.3.4 Detection of E. coli O157:H7 in meat samples...... 270
15.3.5 Release of bound E.coli O157:H7...... 271 xiv
15.3.6 Detection of E. coli O157:H7 in broth containing irradiated meat samples 271
15.3.7 Scanning electron microscope image...... 273
15.3.8 Determination of E. coli concentration using the Most Probable Number Method (MPN)...... 273
15.3.9 Kinetics of E. coli O157:H7 binding ...... 275
15.4 Conclusions...... 277
Chapter 16:Detection of Staphylococcus Enterotoxin B at picogram levels using Piezoelectric-Excited Millimeter-Sized Cantilever Sensors...... 288
16.1 Introduction...... 288
16.2 Materials and Methods...... 290
16.2.1 PEMC Fabrication ...... 290
16.2.2 Experimental procedures ...... 290
16.3 Results and Discussion ...... 291
16.3.1 Resonance characterization of PEMC sensors...... 291
16.3.2 Detection of SEB at 50 ng/mL in SFC...... 292
16.3.3 The Stepwise Binding of SEB at the picogram levels in SFC...... 295
16.3.4 Kinetics of antibody and spore binding ...... 296
16.4 Conclusion ...... 297
Chapter 17:A Method for Measuring Bacillus Anthracis Spores in Presence of Copious Amounts of Bacillus Thuringiensis and Bacillus Cereus ...... 303
17.1 Introduction...... 303
17.2 Sensor fabrication ...... 305
17.3 Experimental...... 305 xv
17.4 Experimental determination of mass change sensitivity in a vacuum...... 306
17.5 Results and Discussion ...... 307
17.5.1 Resonance Characterization of FtPEMC Sensors...... 307
17.5.2 Mass change sensitivity ...... 308
17.5.3 Sensor response to BA spores binding from solutions containing various concentrations of non-antigenic bacillus species (BT and BC) ...... 309
17.5.4 Kinetics of BA spore binding...... 311
17.6 Conclusion ...... 313
Chapter 18:A Method for Measuring Escherichia Coli O157:H7 At 1 cell/mL in 1 Liter Sample Using Antibody Functionalized Piezoelectric-Excited Millimeter-Sized Cantilever Sensor...... 322
18.1 Introduction...... 322
18.2 Materials and Methods...... 323
18.2.1 Sensors ...... 323
18.2.2 Experimental Setup...... 323
18.2.3 Experimental procedure ...... 324
18.3 Results and Discussion ...... 325
18.3.1 Resonance Spectra ...... 326
18.3.2 Batch Detection of EC Using PEMC-a...... 328
18.3.3 Detection of EC Using PEMC-b At 1 EC/mL Under Various Flow Rates . 329
18.3.4 EC Detection In A Flow and Stop Modality...... 330
18.3.5 Kinetics of EC O157:H7 binding...... 332
18.4 Conclusion ...... 332 xvi
Chapter 19:Future Work ...... 339 List of References ...... 341 APPENDIX ...... 368
Appendix A: Cantilever sensors for physical and chemical measurements ...... 369
A.1 Physical measurement of cantilever sensors...... 369
A.1.1 Temperature ...... 369
A.1.2 Power ...... 372
A.1.3 Pressure ...... 372
A.1.4 Elastic properties...... 373
A.1.5 Stress ...... 373
A.1.6 Viscosity and Density ...... 376
A.1.7 Current ...... 381
A.1.8 Piezoelectric charge coefficient ...... 381
A.1.9 Resonant frequency...... 384
A.2 Chemical measurements cantilever sensors...... 386
A.2.1 Gas and vapor detection...... 386
A.2.2 Ion concentrations ...... 395
A.2.3 pH...... 397
A.2.4 Rate of reaction ...... 398
Vita ...... 405 xvii
List of Tables
Table 2.1: Microcantilever sensors for biochemical measurements…..………………...30
Table 4.1: Comarison of the experimental, simulation, and Euler Beam results of the first and second bending modes resonant frequencies……...…………………….50
Table 5.1: Dimensions of the piezoelectric-excited millimeter-sized stainless steel cantilever sensors free ends…………………..………………………….....69
Table 5.2: Resonance characterization of the first three bending mode peaks of cantilevers A and B………………………….…………….……………….70
Table 5.3: Measured and calculated values of model parameters for deionized water…71
Table 5.4: Cantilever performance parameters for the different liquid samples………..72
Table 6.1: Physical dimensions of PZT/Stainless Steel/ Glass cantilevers A and B……97
Table 6.2: Resonance characteristics of cantilevers A and B…………………….……..98
Table 6.3: Calculated parameters from experimental data obtained by the second flexural mode resonant frequency of cantilevers A and B………………....99
Table 8.1: Physical dimensions of the piezoelectric-excited millimeter-sized glass cantilever free-end……………..………………………………………….128
Table 9.1: Physical dimensions of PEMCa sensor………………………….…………147
Table 9.2: Composition of mixed spore samples containing Bacillus anthracis and Bacillus thuringiensis……….…………………………………………….148
Table 9.3: kobs values as a function of Bacillus anthracis (BA) spore concentration in both pure and mixed samples containing BA and BT……………..……...149
Table 10.1: Dimensions of the PZT-glass cantilever…………………………………..171 xviii
Table 10.2: Resonant peaks characteristics of the PZT-glass cantilever………………172
Table 10.3: Calculated values of the cantilever effective mass, spring constant, intrinsic damping, added mass and viscous damping in PBS solution at 1mm immersion depth for the second flexural mode resonant frequency………………………………………………...………………..173
Table 10.4: Data showing experimental predictions of mass attached and model prediction of the binding reaction rate constants. Experimental errors were within ±1% of the reported values……………………………………....174
Table 10.5: Data showing experimental and protein assay results of the mass of anti- rabbit IgG and captavidin that binds and unbinds during one hour………175
Table 11.1: Resonance peaks characteristics of the cantilever sensor used in the present study……………………………………………………………...……….196
Table 11.2: The results of observed binding rate constant, kobs , determined from the reversible first order Langmuir kinetic model………..…...……………...197
Table 12.1: Calculated parameters of the second flexural mode resonant frequency....213
Table 13.1: Resonance characteristics of PEMC sensors in both air and PBS at 1 mm immersion…………………………………………...……………...……...231
Table 14.1: kobs Values of anti-BA, BA, and BSA under stagnant and flow conditions………...... 252
Table 15.1: E. coli concentration in samples using the Most Probable Number Method (MPN)………………...…………………………………………………..278
Table 15.2: kobs Values as a function of E. coli O157:H7 incubation time…………..280 xix
Table 17.1: Composition of mixed spore samples containing Bacillus anthracis, Bacillus thuringiensis and Bacillus cereus, sensor response, and hindrance parameter….…………………………….………………………………...314
Table 17.1: Experimentally determined mass change sensitivity in a vacuum (60 mTorr) and the estimated liquid sensitivity using paraffin wax……………………315
Table A.1: Cantilever sensors for physical measurements………………………….....401
Table A.2: Microcantilever sensors for chemical measurements………..……….……403
xx
List of Figures
Figure 3.1A: Simulation results of a cPEMC sensor with dimensions of glass layer 6 x 2 x 2 mm3 and PZT layer2xmm3………….………...………………..39
Figure 3.1B: The resonant modes of frequency above 100 kHz of the cPEMC sensor investigated………………………………………………………………..40
Figure 4.1A: Two-dimensional cross sectional schematic of the stainless steel base cPEMC sensor…………….……………………………………………...51
Figure 4.1B: Cross section view of glass base cPEMC sensor………………………….51
Figure 4.1C: Schematic illustration of free-floating PEMC sensor……………………..51
Figure 4.2A: Van Dyke’s resonant equivalent circuit of a piezoelectric ceramic………52
Figure 4.2B: Resonant spectra of phase angle and impedance versus excitation frequencies………………………………………………………………52
Figure 4.3: Resonance spectrum, phase angle versus excitation frequency, and the simulation results of the sensor dynamics…………………………………53
Figure 4.4: Resonance spectra for PEMC1 and PEMC2. PEMC1 is 40% longer than PEMC2…………………………………………………….……………...... 54
Figure 4.5: Glass surface functionalization protocol..……………………...……………55
Figure 4.6: Gold surface functionalization protocol …………………………...………..56
Figure 5.1: Experimental arrangement showing the overhang stainless steel layer immersed one millimeter into the test liquid…………………………..….73
Figure 5.2: Resonant spectrum of phase angle versus frequency of cantilever A in air...... 74
xxi
Figure 5.3: Second bending mode resonant peaks in air (right) and in water (left) immersed at 1.0 mm depth…………………………………………………75
Figure 5.4A: Shifts in the second mode resonant peak of cantilever A, initially immersed at 1.0 mm in deionized water…………………...……………76
Figure 5.4B: The second mode resonant frequency response (±10 Hz) as a function of time…………………………..………………………………………..76
Figure 5.5A: Plot of resonant frequency change versus the level of deionized water on the tip of cantilever A, initially immersed at 1.0 mm…………...….....77
Figure 5.5B: The inverse square of resonant frequency versus immersion depth of cantilever A in deionized water……………………………..……………77
Figure 5.6: Second mode resonant frequency (±10 Hz) response of cantilever A Initially immersed at 1.0 mm in 50% ethanol-water solution……………..78
Figure 5.7A: The second flexural mode resonant frequency versus immersion depth (± 10 μm) of cantilever A in mineral oil………………………....79
Figure 5.7B: The reciprocal square of the resonant frequency versus immersion depth in mineral oil…………….………………………………………...79
Figure 5.8: Physical measurements of deionized water level (± 10 μm) as a function of time………….………………………………………………..80
Figure 5.9: Schematic of cross sectional view for the partial immersion of stainless steel cantilever in sample liquid……………….…………………………..81
Figure 6.1: Resonant spectra of phase angle versus frequency of cantilevers A (in panel A) and B (in panel B) in air…………………………….……...100
Figure 6.2A: Mass change sensitivity of cantilever A in air………………………….101
xxii
Figure 6.2B: The resonant frequency changes (±10 Hz) with mass additions (±0.12 μg) gives experimental measures of mass sensitivity……………………….101
Figure 6.3: Typical resonant frequency response (±10 Hz) upon immersion of a cantilever in liquid…………………………………….………………..102
Figure 6.4A: Second flexural mode resonant peaks of cantilever A…………………..103
Figure 6.4B: Second flexural mode resonant peaks of cantilever B…………………...103
Figure 6.5A: Attachment of pathogen at various sample concentrations……………...104
Figure 6.5B: Release of antigen upon immersion in low pH buffer subsequent to each attachment experiment………………………………………………….104
Figure 6.6A: Sample containing 7x107 cells/mL exposed to antibody immobilized glass cover slip (same material used to construct microcantilever) and at 10 minutes the cover slip was removed, rinsed and dried …...…..…...105
Figure 6.6B: Sample containing 7x106 cells/mL exposed to antibody immobilized glass cover slip (same material used to construct microcantilever) and at 10 minutes the cover slip was removed, rinsed and dried……………….…105
Figure 6.6C: Sample containing 7x104 cells/mL exposed to antibody immobilized glass cover slip (same material used to construct microcantilever) and at 10 minutes the cover slip was removed, rinsed and dried…………...……..105
Figure 6.7A: Kinetic analysis of 7x106 cells/mL experimental data for Cantilever A...106
Figure 6.7B: The results of 7x107 cells/mL is superimposed on the 7x106 cells/mL Result……………...…………………………………………………….106
Figure 6.7C: Kinetic analysis of 7x106 cells/mL experimental data for Cantilever B...107
Figure 6.8A: Attachment of pathogen E. coli O157:H7 in samples mixed with wild strain JM101…………………………………………………………………....108 xxiii
Figure 6.8B: Release of antigen upon immersion in low pH buffer subsequent to each attachment experiment………………………………..………………...108
Figure 7.1: Resonant spectra of phase angle versus frequency of cantilever in air……114
Figure 7.2: Resonant frequency response of the second flexural mode versus time for the binding of E. coli O157:H7 at various sample concentrations to the antibody functionalized cantilever………...……………………………………..…115
Figure 7.3: Resonant frequency response of the second flexural mode versus time when the clean (antibody-free) cantilever was immersed in E. coli O157:H7 at 7x107 cells/mL……………………………...……………………………...116
Figure 8.1: Two dimensional side view schematic of experimental setup…………….129
Figure 8.2: Resonant spectrum of phase angle versus frequency of the cantilever in air…………………………………………………..……………………..130
Figure 8.3: Plot of phase angle versus the fundamental mode frequency of the cantilever in air, with agar film, and initial E. coli inoculation, respectively……………………………………………….………………...131
Figure 8.4: Resonant frequency response of the fundamental mode to the growth of E. coli JM101………………………………………………………………...132
Figure 9.1A: Resonant frequency change as antibody is immobilized on the aminated PEMC sensor surface at 22 oC……………………………..……………150
Figure 9.1B: The binding of spores (3,000 spores/mL) to the antibodies on the sensor surface…………………...……………………………………………….150
Figure 9.2A: Resonant frequency shift of the second flexural mode of PEMCa sensor upon binding of Bacillus anthracis spores at various sample concentrations to antibody functionalized cantilever…………………………...……….151
xxiv
Figure 9.2B: The resonant frequency change at 20 minutes as a function of spore concentrations…………..……………………………………………….151
Figure 9.3A: Scanning electron micrograph of Bacillus anthracis spores at 25,000X…..…………………………………………………..……….152
Figure 9.3B: Scanning electron micrograph of Bacillus anthracis spores at 15,000X...152
Figure 9.4: The second flexural mode resonant frequency change of PEMCb sensor, antibody functionalized, for the binding of Bacillus anthracis spores from sample solutions mixed with Bacillus thuringiensis spore…………….....153
Figure 9.5A: The initial kinetic analysis of the various BA spores concentrations…….154
Figure 9.5B: A plot of kobs versus the log of the spore bulk concentrations…………...154
Figure 10.1: Resonant spectrum of phase angle versus frequency of the cantilever in air………………………………………………………...……………...176
Figure 10.2: In air measurement of mass change sensitivity…………………………..177
Figure 10.3: Second flexural mode resonance peaks in air (right) and in PBS solution (left) immersed 1 mm……………………………………………………178
Figure 10.4: Binding of antibody to E. coli O157:H7 at various sample concentrations…………...…………………………………………….179
Figure 10.5: Binding of antibody to Group A Streptococcus at various sample concentrations………………………….……………………………….180
Figure 10.6: Resonant frequency change versus time for the sequential binding of proteins………………………………………………………………..…181
Figure 10.7A: The kinetic analysis result for the binding reaction rate constant for the 0.1mg/mL anti-EC sample reacting with the amine-derivatized cantilever glass surface………………………………………………………….…182 xxv
Figure 10.7B: The binding kinetic results of the 1.0 mg/mL anti-GAS sample reacting with the amine functionalized sensor surface………………………..…182
Figure 11.1: Resonant spectra of phase angle versus frequency of the gold plated cantilever in air……………………..…………………………………...198
Figure 11.2: The third flexural mode resonant peak of the cantilever in air (right, solid line) and in ethanol solution immersed to a depth of 1.5 mm (left, broken line)…………………………………………………….………………...199
Figure 11.3: The third flexural mode resonant frequency response to various concentrations of 1-hexadecanethiol……………...…………………...200
Figure 11.4A: Langmuir kinetic analysis of thiol adsorption on PEMC sensor surface ………………………………………………………………...201
2.1E-03
1.6E-03 ] -1 s [
obs 1.1E-03 k Figure 11.4B: Concentration dependence of 6.0E-04 on the adsorption of 1-hexadecanethiol 1.0E-04 0 0.002 0.004 0.006 0.008 0.01 0.012
Thiol concentration, C0 [M] in ethanol onto a gold coated millimeter-sized cantilever mass sensor……………………………………………..…………………….201
Figure 12.1: Schematic illustration of the experimental setup………………………...214
Figure 12.2: Resonance spectrum of PEMC sensor in air (solid line) and in ethanol (broken line)………………………………………………………………215
Figure 12.3: Resonant frequency change for the adsorption of the various alkanethiols on the gold coated PEMC sensor surface………………………..…………..216
Figure 12.4A: A plot of resonant peak as known mass of wax was added to the PEMC sensor’s tip…………………………………….……………………….217
Figure 12.4B: A plot of known mass change versus the corresponding resonant frequency change……………………………………………………………….….217
xxvi
Figure 12.5: Resonant frequency change for the adsorption of human serum albumin (HSA) on the different self-assembled monolayer (SAM) terminated functional end groups………………………………………………...…218
Figure 12.6A: The initial adsorption kinetics for the adsorption of alkanethiols on to the gold plated sensor surface…………………………………………...…219
Figure 12.6B: Initial kinetic analysis for the binding of HSA to self-assembled thiol monolayers having different terminal end groups……………………..219
Figure 13.1: Experimental determination of mass change sensitivity of PEMCn1 and
PEMCo…………………………………………………………….……232
Figure 13.2: Resonant frequency responses for the binding of various Group A
Streptococcus (GAS) concentrations to anti-GAS functionalized PEMCn1 sensor……………………………………………………………………233
Figure 13.3: Resonant frequency response to the sequential binding and release of 7x107 cells/mL sample of Group A Streptococcus (GAS)……………………...234
Figure 13.4A: ESEM micrograph of antibody functionalized glass cover slip (same material used to fabricate PEMCs) after one hour exposure to 700 cells/mL of Group A Streptococcus……………………………..….…235
Figure 13.4B: ESEM micrograph of antibody functionalized glass cover slip (same material used to fabricate PEMCs) after one hour exposure to 7x107 cells/mL of Group A Streptococcus………………………………..….235
Figure 13.4C: ESEM micrograph of antibody functionalized glass cover slip (same material used to fabricate PEMCs) after one hour exposure to 7x109 cells/mL of Group A Streptococcus…………………………..……….235
Figure 13.5: Initial kinetic analysis for the binding of Group A Streptococcus……….236
Figure 14.1A: Cross sectional view of SFC-1………………………………………....253
xxvii
Figure 14.1B: Cross sectional view of SFC-2…………………………………………253
Figure 14.2: Flow circuit of experimental apparatus…………………………………..254
Figure 14.3A: Flow rate dependence of resonant frequency for SFC-1……………….255
Figure 14.3B: Flow rate dependence of resonant frequency for SFC-2……………….255
Figure 14.4A: SFC-1, the PEMC sensor is positioned with its width (1 mm) perpendicular to inflow………………………………………………256
Figure 14.4B: SFC-1, the width of the sensor is placed parallel to inflow…………….256
Figure 14.4C: SFC-2. Each panel pressure and flow maps at flow rates 1 and 17 mL/min……………………………………..………………………….256
Figure 14.5: The dependence of pressure drop across the sensor to increasing flow rate of the sample solution in the various sensor flow cells, and configurations…………………………………………………………..257
Figure 14.6A: The resonant frequency response to Bacillus anthracis (BA) spores at 300 spores/mL, and the release of the bound spores…………………..……258
Figure 14.6B: The exponential decrease in resonant frequency immediately after BA spores exposure…………………………………………………...…….258
Figure 14.7A: Antibody to Bacillus anthracis spores (0.01 mg/mL anti-spore) reacts with aminated PEMC sensor surface…………………...……………………259
Figure 14.7B: The binding of Bacillus anthracis spores, 300 spores/mL, to PEMC sensor…………………………………………………………………..259
Figure 14.8: Resonant frequency response in the detection of a model protein, bovine serum albumin (BSA) in SFC-2 at 1 mL/min……………………...……260
xxviii
Figure 14.9A: Anti-BA to the aminated PEMC sensor surface under both stagnant and flow conditions…………………………………………….…………..261
Figure 14.9B: Bacillus anthracis (BA) spores (300 spores/mL) to the anti-BA derivatized sensor in both flow and stagnant conditions………………...………….261
Figure 14.9C: Bovine serum albumin (BSA) to an anti-BSA functionalized sensor in flow configuration and in one cycle of regeneration………………………...261
Figure 15.1A: Typical resonant spectrum of PEMC sensors in air……………………281
Figure 15.1B: The fundamental resonant peak in air (right), and after immersion in the flow cell in PBS (left) flowing at 1mL/min………………………...…..281
Figure 15.2: Resonant frequency change as the antibody functionalized PEMC sensor was exposed to E. coli O157:H7 inoculated beef samples flowing at a rate of 3.2 mL/min and temperature of 22 oC……………….…………….…282
Figure 15.3: Resonant frequency change as a function of time for the binding of E. coli O157:H7 to antibody derivatized PEMC sensor from broth (B) samples flowing at 1 mL/min and was initially inoculated with 25 pathogens……………………………………………………….………..283
Figure 15.4: Resonant frequency change as a function of time for the binding of E. coli O157:H7 to antibody derivatized PEMC sensor from meat (M) samples flowing at 1 mL/min and was initially inoculated with 25 pathogens…………………………………………………………..…….284
Figure 15.5A: The attachment and release of E. coli O157:H7 from the 2 h meat (M) sample…………………………………………………………………..285
Figure 15.5B: The binding and unbinding of E. coli O157:H7 from the 6 h meat sample…………………………………………………………………..285
Figure 15.6A: Samples harvested at 2h, 4h, and 6h……………………………………286
xxix
Figure 15.6B: The results from two different experiments of irradiated meat-E.coli O157:H7 samples that were from the sample batch but contained more meat particles in one than the other………………………………...…286
Figure 15.7A: The initial kinetic analysis of the incubated E. coli O157:H7 sample of broth………………………………………...…………………………287
Figure 15.7B: The initial kinetic analysis of the incubated E. coli O157:H7 sample of broth and meat…………………………………………………………287
Figure 15.7C: The initial kinetic analysis of the incubated E. coli O157:H7 sample of broth and irradiated meat………………………………………………287
Figure 16.1: Resonant spectra of PEMC sensor’s fundamental resonant mode in air (right) and fully submerged in PBS flowing at 1 mL/min……………...299
Figure 16.2: Transient response of resonant frequency to the binding of 50 ng/mL SEB to an anti-SEB functionalized PEMC sensor and the subsequent release of the bound SEB………………………………………………..………….300
Figure 16.3: Transient response of PEMC sensor to the detection of increasing SEB concentration………………………………………...………………….301
Figure 16.4: Initial kinetic analysis of the binding of various SEB concentrations to an anti-SEB PEMC sensor………………………………………………….302
Figure 17.1: Schematic of PZT-anchored PEMC sensor……………………………....316
Figure 17.2A: Resonant spectrum, a plot of phase angle versus excitation frequency, of the PAPEMC sensor investigated……………………………….……..317
Figure 17.2B: Resonant peak of PAPEMC sensor in air (970 kHz, right) and totally submerged in PBS (835.6 kHz, left)…………………………………..317
Figure 17.3: Experimental determination of mass change sensitivity for the 970 kHz peak in a vacuum using paraffin wax………..………………………....318 xxx
Figure 17.4: Transient response of PZT-anchored PEMC sensor to the binding of 333 BA spores/mL from a solution containing various amounts of other Bacillus species…………………………………….…………………....319
Figure 17.5: Transient frequency response of PAPEMC sensor to the binding of 333 BA spores/mL and the release of the bound spores by exposure to low pH buffer, pH 1.8…………………………………………..……………..….320
Figure 17.6A: Langmuir kinetic analysis of BA binding on PAPEMC sensor surface...321
Figure 17.6B: Hindrance factor as a function of non-antigenic Bacillus species (BT and BC) to the transport of BA to the cantilever sensing surface………………………………………………………………….321
Figure 18.1: Schematic of experimental setup…………………………………………333
Figure 18.2: Resonant spectra of, phase angle versus excitation frequencies, PEMC-b sensorin air (blue) and fully immersed in PBS (pink)………..…………334
Figure 18.3: Frequency response of PEMC-a to the detection of EC O157:H7 in a batch configuration at 70, 10, and 1 EC in 1 mL of PBS buffer…………..…..335
Figure 18.4A: Resonant frequency change, of the 800.0 kHz peak under liquid, for the binding of EC (concentration 1EC/mL in 1 L buffer solution) to PEMC- b………………………………………………………………………..336
Figure 18.4B: Sensor response to the attachment of EC at sample flow rate of 1.5 mL/min and the subsequent release of the bound EC………………....336
Figure 18.5: Resonant frequency response of PEMC-b sensor to the binding of EC O157:H7 at concentration of 1 EC/mL (total sample volume was 1 L) for the dominant resonant modes investigated...... 337 Figure 18.6: Initial kinetic analysis of EC attachment at various flow rates…………..338
Figure A.1: Direction of excitation for a piezoelectric material……………………….400 xxxi
Abstract Detection and Quantification of Pathogens, Proteins, and Small Molecules Using the Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensor Gossett A. Campbell Raj Mutharasan, Ph.D., Advisor
In this study novel rectangular-shaped piezoelectric-excited millimeter-sized cantilever (PEMC) sensors were fabricated and shown to have picogram to femtogram sensitivity under liquid flow conditions (0.05 to 0.74 cm/s). Sensitivity is characterized by detecting pathogens, proteins, and small molecules: Escherichia Coli O157:H7,
Bacillus anthracis spores, and Group A Streptococcus at 1, 300, and 700 per mL, respectively. Staphylococcal enterotoxin B, a food toxin, was detected at 50 pg/mL, and the adsorption of 1-hexadecanethiol at 1 nM was measured. The sensors were operated under both batch and flow configurations in various sensor flow cell (SFC) geometries.
Flow was shown to double the detection sensitivity for the same sensor design.
Robustness of PEMC sensors was demonstrated under various flow rates, ranging of 1 to
17 mL/min. The sensors showed frequency fluctuations of ± 20 Hz over the flow rate range investigated, which is on the noise level of the device. The recognition layer
(antibody) was shown to be regenerated at least twice with HCl/PBS buffer (pH 2.0), using model protein BSA, without significant loss in the antibody activity. Selectivity was established by detecting a fix concentration of bacillus anthracis spores (333/mL) in the presence of varying amounts of other bacillus species. A constant total change in resonant frequency was observed in all the experiments with a decrease in binding rate with an increase in the non-antigenic species concentration. 1
Chapter 1: Introduction
1.1 Background
The invention of the atomic force microscope (AFM) in 1986,1 and its impact in the fields of biotechnology and nanotechnology have created a new modality of sensing: the cantilever. The growing interest in using cantilevers as biosensors is due to their high sensitivity. The cantilever sensor operates on three mechanisms: the deflection due to induced differential surface stress, resonant frequency shifts due to mass changes, and thermal changes that cause bending. Cantilevers have been used in numerous applications to detect physical properties (example density and viscosity), chemicals such as VOCs and heavy metal ions, biochemicals including cells, pathogen, toxins, DNA hybridization,
and proteins.
The rapid development of cantilever technology, so as to keep up with the fast
growing applications has favored reduction in its dimensions on the micro and nano-
scales. At these dimensions the sensors are extremely sensitive such that the detection of
a single virus and a few molecules is possible2-3. However, such detections were only
demonstrated in air and in vacuum in which the biologics were immobilized on the sensor
surface from an aqueous buffer followed by drying and subsequent measurement of the
resonant frequency shifts in air and vacuum. The main reason why micro and nano-sized
cantilevers cannot make direct measurement in liquids is that the liquid significantly
dampens the sensor’s dynamics at resonance and thus, the quality of resonance decreases
substantially. Since biological materials generally conform to their native state in a liquid
environment this technique may not reflect the true state of the biologics when taking
measurements. Furthermore, salt from residual buffer solution on the cantilever will 2
precipitate and add to the measured resonant frequency shift and further enhance the
cantilever’s deflection. Also, there is the risk of breaking the cantilever in transporting it
between different locations. In applications where the binding and unbinding kinetics are
needed and for real-time online monitoring of processes, detection in liquid is the
desirable modality. In addition, direct measurement in liquid will reduce false positive
and false negative responses. To circumvent the limitations of in air detection we have
developed the piezoelectric-excited millimeter-sized cantilever (PEMC) sensor that operates in the dynamic bending mode of vibration.
The PEMC sensor is a cost effective, compact, robust, easy-to-use, sensitive, specific
via surface modification, and a label-free light weight mass sensor that uses the resonant
frequency shifts as the measurement mechanism. The sensor is a composite structure of
two layers: a lead zirconate titanate (PZT) and a glass layer of a few millimeters in
length. The PZT layer acts both as an actuating and a sensing element. The detection of
biological entities requires the immobilization of a recognition molecule, such as an
antibody or a receptor, on the sensor surface. When the target of interest binds to the
cantilever’s sensing surface the effective mass of the cantilever increases which decreases
the cantilever’s resonant frequency. The resonant frequency change with time is used to
provide quantitative measurement of the analyte concentration.
1.2 Objectives
The goal of this research is to develop piezoelectric-excited millimeter-sized
cantilever sensors (PEMCS) for in-situ rapid direction and quantification of pathogens,
proteins, and molecules at very low concentrations. We increase the length of the
cantilever to millimeter size and tailor its geometry, so that the cantilever’s sensitivity is 3
comparable to the micro and nano-sized cantilever with the added advantage of being able to work under liquid without significant loss in resonance characteristics. We
hypothesize that by tailoring the cantilever in this way, detection of a single cell or 1
cell/mL and molecules on the attogram level can be detected with good accuracy and
reproducibility. The specific aims listed below address the application fundamentals that will lead to practical use of PEMC sensors in liquid.
1. Fabrication
a) To design and fabricate PEMC sensors with predominatly bending modes of
oscillation, high resonant frequencies, and good stability in liquids.
b) To design and fabricate sensor flow cell (SFC) with minimum hydrodynamic noise
and optimium analyte to sensor contact.
2. Sensor characterization
a) To characterize the sensors in both air and liquid in terms of resonant frequency,
phase angle and impedance curves, mass change sensitivity, resonance quality (Q-
factor), and flow stability.
b) To functionalize the sensor surface with recognition molecules for specific antigen
binding. For example glass-amination-antibody or gold-protein G-antibody.
3. Detection
a) To optimize immobilization protocol to enhance orientation of recognition layer
and antigen attachment.
b) To investigate the detection of low concentrations of pathogens, proteins, and small
molecules in both stagnant batch mode and flow configuration.
c) To validate antigen binding by releasing the bound analytes using pH alteration. 4
d) To characterize the binding process using kinetics models.
e) To investigate the simultaneous binding and unbinding of at least two antigen on a
single PEMC sensor.
1.3 Contributions
The novel aspects of this research work can be summarized as follows:
1) Design and fabrication of millimeter-sized cantilevers that as comparable sensitivity
under liquid flow conditions to that of microcantilever in air. Sensitivity was
demonstrated on the picogram level.
2) PEMC sensors are very robust; we have demonstrated the robustness of PEMC sensor
by operating in both stagnant and liquid flow conditions (at flow rates ranging from
1-17 mL/min).
3) PEMC sensors were shown to be selective to the target of interest. For instance, the
detected of bacillus anthracis spores were demonstrated in 10,000 fold of bacillus
thuringiensis and bacillus cereus.
4) We were the first to report the detection of nanomolar alkanethiolate monolayer
adsorption on a gold surface.
5
Chapter 2: Literature review on cantilever sensors
2.1 Biosensors
The current development in biosensing technologies relies heavily on fluorescence4, laser5, or fiber-optics-based methods, and amplification schemes such as polymerase chain reaction6, and binding to metal particles7. The major limitation of the current
approaches is that they are complex and requires extensive sample preparation. There are too many instances of false alarms, both positive and negative. Furthermore, most of the techniques are neither direct nor quantitative. They do not lend themselves to measurement of mass changes or forces due to biomolecular interactions. Another development of biosensing technologies relies on silicon-based microcantilevers8-12 due
to their availability and ease of integration with existing silicon based methodologies. All
silicon-based microcantilevers rely on external optical components for measuring
deflection. Because of the small size of the silicon-based microcantilevers, about 100 μm
long, they exhibit high mass-detection sensitivity. The mass change per Hz is about
∆m/∆f~10-12 g/Hz9,10, where ∆m and ∆f respectively denote the mass change and the
corresponding resonant frequency change due to the binding of the target molecules.
However, the required optical components are bulky, complex, and require precise
alignment. The vibration driver that is required to drive the microcantilever adds to the
weight and complexity of the sensor. Moreover, immersing silicon-based
microcantilevers in water reduces the resonance intensity by an order of magnitude;
reducing the quality factor (Q) to about one 11. This in turn, prohibits the use of silicon-
based microcantilevers for in-water detection. The main reason that silicon-based
microcantilevers cannot have high resonance signals in water is because of their size. As 6
the characteristic dimensions are on the micron scale, the Reynolds number of the
oscillating sensor is less than 1 (Re<1). As a result, viscous damping is significant and
resulting poor response.
The deflection at the tip of silicon-based cantilever is driven by the vibration driver located at the base of the cantilever. Driving a cantilever at its base is not the most effective way to generate deflections at the tip of the cantilever. While the relatively weak deflection signal generated by the vibration driver at the base is sufficient for in-air detection, it does not withstand the damping due to water. To circumvent the bulkiness and complexity of the silicon-based sensors the idea was to develop piezoelectric-based sensors. Piezoelectric devices are excellent transduction candidates because of their short response time and high piezoelectric coefficients. Because they are piezoelectric, both the driving and sensing of the mechanical resonance can be conveniently done electrically.
Piezoelectric biosensors are based on the commercially available quartz crystal microbalance (QCM), a disk device that uses the thickness shear-mode resonance for sensing13-15. Although quartz is a weak piezoelectric material, it is widely used as a layer
thickness monitor in part due to the availability of large quartz single crystals to make the
membranes. The typical mass detection sensitivity of a 5 MHz QCM, having a minimum
detectable mass density (DMD) of 10−9g/cm2, is 10−8g/Hz which is four orders of
magnitude less sensitive than the silicon-based microcantilevers. Furthermore, when
immersed in water the resonant peak intensity of QCM is reduced to less than one
twentieth of the in-air peak intensity due to viscous damping of water, thus limiting the
use of QCM in water. Between the two direct biosensors, the silicon-based
microcantilever sensors exhibit high mass-detection sensitivity, but they are bulky and 7
complex. The QCM-based sensors have the advantage of simple electrical driving and
electrical detection but they exhibit a much lower mass sensitivity than silicon-based
microcantilevers. The piezoelectric excited cantilevers have combined the merits of the
two sensors: exhibits high detection sensitivity, uses electrical means for both driving and
sensing, has minimal damping effect, and is compact, light-weight, and simple to operate.
2.2 Detection principles of cantilever sensors
The detection principles commonly used in cantilever sensors are based upon changes in surface stress, deflection, resonant frequency, capacitance, and resistance induced by the adsorption or attachment of analyte on the cantilever’s surface.
2.2.1 Deflection and surface stress
Cantilever sensors where the top and bottom surfaces are different, for example
bimetallic or only one side has a recognition layer, will develop a differential-induce
surface stress upon the adsorption of analyte or temperature change. The differential
surface stress will cause the cantilever to bend or deflect16,17,18,19,20. The adsorption-
induced surface stress can be determined from the radius of curvature of the cantilever
bending using the Stoney’s equation21,22.
1 ⎛1−υ ⎞ = 6⎜ ⎟()Δσ 1 − Δσ 2 (2-1) R ⎝ Et 2 ⎠ where R is the radius of curvature, t the thickness, E the Young modulus, υ is the
Poisson ratio, and ()Δσ 1 − Δσ 2 the differential surface stress. For a beam of length L that
is fixed at one end and free at the other, the tip deflection Δz can be expressed as:
2 ⎛1−υ ⎞ Δz = 3L ⎜ ⎟()Δσ 1 − Δσ 2 (2-2) ⎝ Et 2 ⎠ 8
Therefore, quantitative measures of surface stress can be determined directly from the measured cantilever deflection. Cantilever deflection measurements require a laser beam and a photosensitive detector. Briefly, visible light from a laser diode is shined on the cantilever tip and the reflected light is picked up by a photodetector. As the cantilever bends the reflected beam moves on the detector surface. The distance traveled by the beam on the detector surface is correlated to the deflection. This technique is commonly used in AFM; the optical lever method. More recently, a new optical method, interferometry, has been employed. This method uses two light sources, a reference laser, and a sample laser. The interference of the reference laser beam with the reflected light from the cantilever surface is the basis of measurements. Interferometry was shown to be more sensitive and gave absolute measurements of displacement23.
2.2.2 Resonant frequency
The frequency at which a cantilever resonates depends on its mass. The cantilever’s resonant frequency is characterized by material properties (Young’s modulus (E) and density ( ρ )) and its geometrical dimensions17. The classical expression for the resonant frequency was derived from the Euler-Bernoulli Beam equation.
1 ⎛ t ⎞ E f = ⎜ ⎟ (2-3) 2π ⎝ L2 ⎠ ρ where t is the thickness and L the cantilever’s length. The adsorption or binding of analyte to the sensor surface will increase its mass, which will decrease its resonant frequency. Since the Young’s modulus of cantilever’s material does not change significantly, this method is convenient in tracking the transient adsorbed mass of analyte on the sensor surface. Resonant frequency based detection has been illustrated in the 9 sensing of molecules74-76,87,88, proteins110,112,118, vapor91,92, humidity19, cells122, and pathogen114,120.
2.3 Detection mechanisms of cantilever sensors
The three most commonly reported mechanisms of operation of cantilever sensors are capacitance, piezoresistive, and piezoelectric. Capacitive sensors monitor deflection in a cantilever by the change in its capacitance. Capacitive cantilever sensors are used extensively in AFM devices24. Although this sensor is of high sensitivity and requires low operating power (mW) it can only function in a vacuum. In air the sensor’s performance is significantly damped as air gets between the electrodes. In liquid the sensor is prohibited by the Faradic currents between the capacitor plates as fabrication provides coating difficulty21,25. Piezoresistive cantilever sensors measure the change in electrical conductivity of the piezoresistor material (example, doped silicon) as it undergoes strain.
Unlike capacitive cantilevers, piezoresistive cantilevers can be use in ambient liquid due to the absence of Faradic currents. Another advantage of the sensor is that its surface temperature can be varied or maintained by altering the current through the resistor.
However, the cantilever fabrication is intricate and therefore, technology limits the size of the sensor. Piezoelectric cantilever sensors use the stress generated during the application of an alternating potential (converse piezoelectric effect) to produce an electric signal.
The piezoelectric cantilever has several advantages over the capacitive and piezoresistive cantilever sensors; these include four times less power (due to an higher impedance), requires no external oscillator, the piezoelectric material serves both as the actuator and sensing element, and they provide excellent performance in air with a small 10-15% reduction in the quality of resonance under liquid. 10
2.4 Millimeter size versus micro and nanometer size cantilever sensors
In the last three decades the monitoring of surface stresses in crystal growth and surface reconstruction in both air and vacuum have been done using millimeter size cantilevers (length and width)26. From Eq.(2), the longer the cantilever the smaller the differential-induced surface stress. That is, the stress change sensitivity increased with cantilever’s length. However, the longer a cantilever’s length the lower it’s resonant frequency and therefore, the higher its mass change sensitivity (see Eq. (7-1)).
Miniaturization (micro and nanometer size) of cantilever sensors provides better sensitivities (both mass and stress), lower mechanical noise, and higher resonant frequencies in air. The sensitivity of miniaturized cantilever is enough to detect a single virus and molecular layer 2,3. On the other hand, micro and nano sized cantilevers do not operate well under liquid due to the high viscous damping at resonance. The dynamics of a cantilever in liquid medium is affected by the viscosity and density of the liquid. As the cantilever oscillates in a liquid, the surrounding liquid offers resistance to the its motion: the added mass of liquid creates an inertial force proportional to the cantilever’s acceleration and the liquid viscosity causes dissipative losses proportional to the velocity of the cantilever’s motion. As discussed earlier, mass loading on the cantilever decreases the sensor’s resonant frequency relative to air and vacuum however, depending on the size of the sensor the viscous effects may or may not affect performance significantly.
The viscous loss dampens the cantilever’s motion and thus, decreases the resonance sharpness (boardening of resonance peak). Sader51 reported a detailed theoretical analysis of the frequency responses of macro and micro cantilever sensors in a liquid medium. He showed that the dynamics of a cantilever in a liquid are highly dependent on a 11 hydrodynamics function. The hydrodynamic term is a function of the Reynolds number.
For micro and nanometer size cantilever the hydrodynamic function needs to be accounted for since Reynolds number (Re) is between 0 and 1, however, in millimeter sized cantilever Re is in the millions and as a result the hydrodynamic function goes to one and the viscous damping is insignificant. Thus, millimeter-sized cantilevers can be operated as well in liquid as it does in air.
2.5 Application of cantilever sensors
Cantilevers have been increasingly used as sensors in a wide variety of applications requiring physical, chemical, and biochemical measurements. Several research groups have reported the use of cantilever sensors in the measurement of physical properties either by bending or resonant frequency change. These properties include temperature27-
42, power36-39, pressure33, Young’s and shear moduli40, viscosity and density47-53, current54, piezoelectric charge constant55-61, frequency61-71. Also, many different cantilever platforms have been developed and used in the measurements of several chemicals and chemical properties such as gases72-90 (which includes vapor concentration72,73, gas concentration84-88, nerve agent89, and explosive90), pH91, pesticide concentration93, ethanol/water concentration94, and ion concentration96-99. These physical, chemicals and chemical property measurements are summarized in Appendix A.
2.5.1 Cantilever sensors for biochemical measurement
The high sensitivity and surface modification chemistry of cantilever surfaces for both physical and chemical measurements have attracted a growing interest in the detection and quantification of biologics. In the last decade, several research groups have developed cantilever sensors for measurement of biologics such as DNA101-104, 12 proteins105-113, ligand-receptor adhesion114-119, vesicle120, and cells121-123. A list of some of the biochemical measurements done with cantilevers are summarized in Table 2.1 along with their resolution, cantilever type, principle of measurement, and references.
2.5.1.1 DNA
Cantilever sensors have been shown to be successful in the analysis of genomic material (DNA). Liu et al.101 reported on the flexoelectric effect induced cantilever bending theory and tried to explain the theory by fitting reported experimental data on the adsorption of single strand DNA (ssDNA) and DNA hybridization reaction125. Several research groups have studied the bending induced on a cantilever by biochemical reaction and explain the phenomenon by cantilever strain energy, intermolecular interaction, and conformational changes that create a differential surface stress124,125. Liu et al. purposed that bending in a cantilever is caused by curvature electricity effects and that the change in deflection is cause by the change of electrical potential across molecular layers. The authors used the asymmetric membrane shape equation to determine the deflection as a function of curvature. Then, they combined the polyelectrolyte theory and the correlation between curvature and potential to analyze DNA hybridization. The authors used their model to fit the experimental data reported by Wu et al.126. Their results showed that the cantilever bends upward initially (up to 5 nt ssDNA) and then, bend downwards upon further hybridization, which was contrary to the plot of Wu et al. (upward bending). The authors explain the phenomenon by the competition that exists between charge accumulation and monolayer concentration. That is, an increase in the total charge will cause downward bending, while an increase in the height of the hybridized layer will decreased the average monolayer coverage, which will result in upward bending. In 13 addition, the authors found a small relationship between cantilever bending, ssDNA length, and salt concentration. Although their model did not fit the experimental data it gave some insight on hybridization. The model may have failed because of the assumption that was made; they assumed that the charged biomolecules were distributed uniformly on the cantilever surface, which is impossible.
Kwak et al. demonstrated for the first time that an AFM operating in the pulsed-force- mode can be used to image the stretching of a single DNA molecule both in air and under water102. A glass cover slip was functionalized with a monolayer of silane compounds having two different terminal end groups; (7-octen-1-yl)trimethoxysilane (CH2=CH- terminated silane) and 3-(2-aminoethyl)aminopropyl)trimethoxysilane (NH2-terminated silane). The double stranded DNA molecules were immobilized on the modified surfaces in TE buffer. The 5´-phosphate binding of double stranded DNA onto the CH2=CH- terminated silane was reported by Allemand et al.126 However, when tried by the authors the surface density was significantly smaller (investigation was done by fluorescence microscope imaging). On the other hand, the formation of mixed silane terminated groups
(CH2=CH and NH2-terminated silanes, 1:10,000 v/v) on the glass surface created a high density of adsorbed DNA. The authors suggested that because the DNA strands are negatively charged in the aqueous solution at pH 7.6 (buffer) they bind to the positively charge NH2-terminals. In a preliminary study, the authors exposed a clean silicon nitride cantilever tip to the DNA substrate at varying humidity and map the stretching of the
DNA strands using pulse-mode AFM technique in air. They found the adhesive force between the DNA strands and the tip to increase up to 30% humidity. The authors indicated that the surface water layer thickness on the glass substrate increased with the 14 increase in humidity, which improved the adhesion between the cantilever tip and the
DNA, a phenomenon known as the capillary force effect. They also found that the contact mode AFM technique damaged some of the surface DNA in air. The author reduced this effect by carrying out the experiments in pure water. The gold coated AFM cantilever tip was modified with hexadecanethiol or 11-mercaotoundecanoic acid and the tip was brought in contact with the substrate surface. Their results showed that for the COO- terminated tip the adhesion force between the tip and the DNA was small, due to charge repulsion. In addition, they found that with the CH3-terminated cantilever tip
(hexadecanethiol) the adhesion with the tip was strong due to the hydrophobic interaction of the tip and the substrate surface. The authors concluded that stretched DNA molecules can be cleaved by strong adhesion of water on the cantilever tip and high stability of
DNA can be achieved under liquid.
Marie et al. described the adsorption kinetics and mechanical properties of a thiol modified DNA-oligo on a gold coated piezoresistive cantilever. An array of 150 μm long,
40 μm wide, and 1.3 μm thick cantilever (each cantilever has a reference sensor to cancel all background noise) and excited by 2 V were used. The measurement sensor was coated on one side with 60 nm thick gold to provide surface for immobilization. It is to be noted that all electrical connections were wax coated to protect against electrical shorts. The detection mechanism was based on the binding induced surface stress caused by reaction of the analyte with the cantilever surface. The experiments were carried out in a flow cell at a flow rate of 25 μL/min. Upon exposure of the gold coated cantilever to the sample solution (thiol-modified DNA-oligo) the sensor was under compressive stress (bend away from the gold coated side). The authors observed that the cantilever response was 15 exponential and the stress change stabilized in approximately 600 second at 0.9 Nm-1.
Their results showed an initial rapid rise and then, a slow increase followed by a slight decrease. The authors suggested that the rapid rise was cause by the monolayer adsorption, after which the oligos in the bulk started to adsorb on top of the monolayer via non-specific hydrogen bonding. The decrease was due to the release of the weakly bound analyte on the cantilever surface during rinsing. The authors fitted the initial responses of various oligo concentrations (1-25 μM) to the first order Langmuir kinetic model and found the association constant to be 2.5 x 103 M-1 s-1, the dissociation constant as 4.7 x 10-3 s-1, and the free energy as -34.6 kJ/mol for the double thiol-modified DNA- oligo.
Hansen et al. demonstrated a sensitive modality for measuring the oligonucleotide hybridization using gold coated thiol functionalized cantilevers and optical readout of their deflection104. Again a thin layer of gold was coated o one side of the cantilever to provide controlled immobilization of thiol-modified oligonucleotide. The authors conducted two sets of experiments; in the first experimental set the length of the complementary strands were varied, while in the second set the sequence of the complementary strands were varied (having a single base mismatch). The hybridization was detected and monitored using the optical deflection technique and the experiments were carried out in a flow cell at a sample flow rate of 1-2 mL/h. The concentration of probe DNA strand used was 25 μg/mL and the complementary strand concentration was
20 μg/mL. When the functionalized cantilever (having thiol-modified 20-mer- oligonucleotide) was exposed to complimentary strands of varying length, the sensor responded with increasing positive deflection with longer strands. Also, when 25-mer 16 thiol-modified oligonucleotides were exposed to complementary strands containing varying degrees of mismatch the authors were able to identify a singe base pair mismatch. Furthermore, they found that mismatch on short complementary strands gave a negative deflection, while those on longer complementary oligomers have positive deflection. In addition, complementary strands that have the same number of oligomer units as the thiolated oligomer (cantilever DNA) gave the largest positive deflection. A similar study was done by Fritz et al. in which they monitored the hybridization of oligonucleotides using an array functionalized cantilever sensor8. These authors also detected a single base pair mismatch using a 12-mer oligonucleotide with different degrees of complementarities.
2.5.1.2 Proteins
Calleja et al. reported on the fabrication of polymeric cantilevers and demonstration of the sensors sensitivity in the detection of cystamine105. The cantilever was fabricated from a 1.5 μm thick epoxy-based photoresist (SU-8) supported on a 400 μm thick SU-8 based. One side of the cantilever tip was gold coated allowing for deflection measurements via laser beam reflection or for surface stress based biochemical reaction.
The authors determined the spring constant of the SU-8 cantilevers of various dimensions and found it to increase with cantilever length, which was an expected result. Also, the elastic constant of the SU-8 sensors compared well with the soft silicon cantilever that is commercially available. The cantilevers were found to have deflection limit of 0.5 nm and a surface stress change sensitivity of 60 μN/m. Interestingly, the authors showed that a cantilever 200 μm long, 20 μm wide, and 1.5 μm thick has the same resonant frequency as one that was 200 μm long, 50 μm wide and 1.7 μm thick. However, 17 cantilevers that were 100 μm long, 20 μm wide, and 1.5 μm thick have resonance of four times the frequency of previous ones. The authors exposed the cantilever to cystamine solution and found the sensor to bend downward immediately (away from the gold surface), which is an indication that the self-assembled monolayer was formed on the sensor surface. The authors suggested that SU-8 sensors are convenient, inexpensive, and reliable for the fabrication of array cantilevers for multiple biochemical detection.
Moulin et al. investigated the adsorption of low-density lipoprotein (LDL) and the oxdised form (oxLDL) on a heparin modified cantilever106. LDL and oxLDL are responsible for the accumulation of cholesterol in the aortic intima and heparin is a strong negatively charged polysaccharide. The cantilevers were freshly coated with thin layer gold on one side and the gold surface was modified with amine-terminated thiol (2- aminoethanethiol hydrochloride). The amine end groups were covalently bonded to heparin via peptide linkage and subsequently, BSA was used to block non-specific binding sites. The adsorption of LDL and oxLDL on the cantilever surface was monitored by the surface stress induced deflection with optical readout. The authors showed that upon exposure of the heparin functionalized sensor to LDL (1.7 mg/mL in 10 μL) and oxLDL (0.3 mg/mL in 10 μL), respectively, the differential surface stress measurements increased for LDL and decreased for oxLDL. The adsorption response was initially rapid in the first few minutes and then slowly changes. Their results showed that in the first minute of LDL adsorption the stress was compressive which indicated that the adsorbed proteins electrostatically repel each other, while oxLDL adsorption change the charge distribution which lead to the observed tensile stress in the first minute. Furthermore, the authors observed that the adsorption happened on both sides of the cantilever (heparin 18
modified and the SiNx sides) and therefore, they suggested that the tensile stress observed in the oxLDL may be a result of oxLDL adsorbed on the non-gold coated side of the cantilever. However, the surface stress changes were slow in comparison to binding observed on surface plasmon resonance127,128. The authors indicated that this may be due to the structural conformation and/or packing configuration change of the adsorbed protein on the cantilever surface.
The same group of researchers106 reported on the non-specific adsorption of immunoglobulin G (IgG) and Bovine serum albumin (BSA) on a gold coated cantilever surface in a liquid buffer by measuring the stress generated during binding. Since the previous studies conducted by the authors showed that LDL adsorbed on the silicon side of the cantilever, they circumvent this problem by coating the silicon side with an inert thiol. Exposure of the cantilever to IgG and BSA at 0.195 mg/mL resulted in a compressive and a tensile stress response, respectively. The authors suggested that the compressive stress (downward bending) of the cantilever response to IgG may be caused by conformational change due to protein-surface interaction or proteins rearrangement caused by hydrophobic protein-protein interaction. The authors indicated that the tensile stress observed for BSA may have resulted from the weak surface-protein interaction, which may have caused the adsorbed protein to pack together tightly resulting in the cantilever bending upwards. The total stress upon IgG and BSA adsorption were determined as 0.22 and 0.10 Nm-1, respectively.
Kooser et al. investigated anti-bovine serum albumin (a-BSA) and BSA interaction in both air and liquid used a piezoresistive cantilever sensor107. Unlike the previous studies discussed, the cantilevers used by the authors were not functionalized with recognition 19 molecules. Instead, they were placed in contact with a recognition layer mechanically to detect swelling or expansion. The cantilevers have an embedded resistor and therefore, change its resistance occurs upon bending. In the liquid detection a glass cover slide was functionalized with 10% aminopropylsilane (APS) in acetate buffer followed by covalent linking of either BSA or a-BSA. The cantilever tip was then brought in contact with the functionalized glass surface and the setup was immersed in deionized water. If BSA was immobilized on the glass slide a-BSA was injected in the solution and visa versa. Their results showed an increase in the cantilever resistance (upward bending), which indicated that the recognition layer had expanded due to the interaction between BSA and a-BSA.
Although the response was small the authors confirm the a-BSA-BSA interaction by exposing the a-BSA layer to donkey serum albumin (DSA) and no response was observed. In the air detection, the authors combined BSA with a polymer (polyethylene oxide, PEO) in volume ratios of 3:1 (BSA:polymer). The polymer-BSA matrix was coated on the glass substrate and by delivering a-BSA in an aerosol of carrier nitrogen/water vapor across the coated glass substrate adsorption occurred. The sensor response was an immediate rapid increase in its resistance, followed by a constant value and then a decreasing response upon changing to pure nitrogen indicating the loss of water from the polymer. However, the decreasing response never returned to the value prior to a-BSA injection. The authors suggested that because a-BSA has a high affinity to
BSA, a-BSA still binds to the polymer matrix and therefore, the response was higher than previous. One of the limitations to this type of sensor is that the polymer will absorb water vapor along with other gasses and thus, the response may not be the true signature of the target analyte. Although the authors carried out experiments to determine the effect 20 of water vapor on the polymer matrix and showed that the response was less than the antibody-polymer matrix complex, there may be errors in the data presented since both experiment were carried out separately. The cantilevers were found to have a force constant of 1 Nm-1, a displacement sensitivity of 1-4 x 10-6 nm-1, and force sensitivity of
0.7 x 10-6 nN-1.
Atsushi et al. studied the adsorption behavior of BSA and IgG on the surface of phase-separated organosilane monolayers cantilever sensor108. A single monolayer and mixed monolayers were investigated. The organosilane monolayers were made with n- octadecyltrichlorosilane (OTS), 18-nonadecenyltriclorosilane (NTS), (2-
(perfluorooctyl)ethyl)trichlorosilane (FOETS), and various combinations of these were also prepared. The proteins were exposed to these surfaces and a modified gold coated
- AFM probe with negative charge groups (SH(CH2)9COO ) was used to characterized the protein layer. The authors found that both BSA and IgG (at pH 7.5) were preferentially adsorbed on the FOETS phase as a single layer and as a mixture of OTS/FOETS monolayer. The authors suggested that the preferential adsorption of the proteins was due to the minimum interfacial free energy between the bulk solution and the monolayer surface as well as the electrostatic repulsion amongst charged proteins.
Several research groups have reported the detection of prostate-specific antigen
(PSA)104, 109-111. Wu et al. reported on the sensitive and selective monitoring of prostate- specific antigen109. PSA is known to be the maker for early diagnosis of prostate cancer.
Two monoclonal specific antibodies to PSA were immobilized on the cantilever gold coated surface, to increase selectivity. Also, each experiment was conducted in a back grounded of BSA (1 mg/mL) or human serum protein (HP, 1 mg/mL). Exposing the 21 antibody functionalized surface to PSA, at concentrations ranging from 0.2 ng/mL to 60
μg/mL, causes the cantilever to deflect downwards with time and the total deflection increases with increasing PSA concentration. The binding response was confirmed by exposing the pure gold surface to PSA and the antibody functionalized sensor surface to buffer; in each case no response was observed. Therefore, the authors demonstrated that microcantilevers have the sensitivity and selectivity to diagnose prostate cancer.
In a similar study done by Lee et al. they demonstrated the detection of low levels of
PSA using a piezoelectric microcantilever110. The cantilever was a composite structure of
PZT and silicon nitride. The main advantage with such a sensor is that no external oscillator is required for actuation. The PZT acts as both the actuator and the sensing element. The detection principle relies on the change in resonant frequency upon mass load due to surface reaction. Therefore, as long as the adsorbed mass remains on the surface the resonant frequency remains constant. Cantilevers of dimensions (W x L) 100 x 300 μm2 and 50 x 150 μm2 were fabricated and used in the studies. An AC current was applied to the PZT layer to set the cantilever in resonance. The glass surface was coated with a 150 μm thick gold layer using a 30 μm thick chromium layer to increase the adhesion between the gold and glass. The gold surface was modified with a self- assembled monolayer of calixcrown to which monoclonal antibody to PSA (a-PSA) was covalently attached. Exposure of the cantilever of different dimensions caused a decrease of the sensors resonant frequencies. The authors determined that for the detection of 1 pg/mL PSA a frequency shift of -172 Hz and -273 Hz were observed for the100 x 300
μm2 and 50 x 150 μm2 cantilever, respectively. They also found the sensor to have a detection sensitivity of 10 pg/mL. 22
Wee et al. developed a piezoresistive self-sensing microcantilever for the sensing of
BSA and C-reactive protein (CRP)111. One side of the cantilever was gold coated by e- beam evaporation. The detection mechanism was based on the change in resistance as the cantilever bends upon surface reaction causing induced differential surface stress. The gold surface was thiol-modified and the antibody to BSA or CRP was immobilized.
Exposing of the antibody functionalized cantilever to various BSA and CRP concentrations (without antigen, 10 ng/mL, 100 ng/mL, and 1 μg/mL). The cantilever responded with an increase in output voltage with increasing antigen concentrations for both BSA and CRP. The authors suggested that the compressive stress observed was due to the binding of the antigens to the specific antibody surface. In another study by Lee et al. they report on the detection of CRP using the resonant frequency shift of a PZT monolithic film cantilever112. Lee et al. showed that the cantilevers were able to detect
CRP concentration on the ng/mL levels. The experimental setup was identical to their earlier work110.
Raiteri et al. reported the detection of 100 nM streptavidin using a silicon nitride biotin modified cantilever sensor21. The authors monitored the binding reaction using the deflection response. The high affinity between streptavidin and biotin was demonstrated in the large deflection response of the cantilevers within 10 min. The same group of researchers also reported the detection of 85 ng/mL myoglobin in solution using two cantilever sensors; one was functionalized with antibody and the other served as the reference sensor.
2.5.1.3 Receptor-ligand adhesion
Several research groups have investigated the force between ligand-receptor (antigen- 23 antibody)113-117. For instance, Gad et al. have reported mapping of the polysaccharides on living microbial cell surfaces113. The cell used in the study was Saccharomyces cerevisiae
(yeast). Yeast cells were immobilized on a glass substrate in a monolayer. The glass substrate was first coated with concanavalin (conA, approximately 102.3 kD), conA binds specifically to mannose residues found on the yeast cell wall. ConA was also immobilized on a gold coated AFM cantilever tip via thiol self-assembled monolayer
(1.56 nm long). The modified AFM tip was brought on contact with the cell sample for binding to occur. The distribution of mannose residues on the cell surface was mapped by force measurement using the force volume mode of AFM. The authors determined that a force of 75 to 200 pN was needed to break a single bond between conA and polysaccharide on the cell surface. They also found that the ligand-recptor bond stretched up to 600 nm before the tip was freed. Furthermore, their results indicated that the polysaccharides (mannan) on the cell surface were not uniformly distributed, because specific recognition only took place in a specific location.
Bowen et al. investigated the adhesion of yeast cells in the different life cycle stages
(exponential, stationary, and death phases) using an atomic force microscope cantilever114. The sensors used were the standard V-shaped AFM tipless cantilevers designed and fabricated by Thermomicroscopes. Yeast cells were cultured and harvested at the exponential and stationary phases. Also, cells from the culture were fixated with
5% glutaraldehyde solution. The electrical properties (zeta potential) and the hydrophobicity (Bath assay) of the cells were measured. Three different surfaces were prepared on a Mica; a hydrophobic (using dichlorodimethylsilane), hydrophilic, and protein functionalized (BSA). A single yeast cell from the different life cycles was glue 24 to the end of the cantilever; the immobilized cell was verified using scanning electron microscope (SEM). The difference between the active cells and the fixated ones were observed by staining the cells with methylene blue (the active cells reduce the dye to a colorless state). The cantilever (with the cell attached) was brought in contact with the various mica surfaces at a constant rate of 0.1 μm/s and then retracted. The adhesion of the cells with the mica surfaces was characterized from the retracted force-displacement data measured optically. The authors observed the largest adhesion force of the cell at the hydrophobic surface (46.6 ± 20.8 nN) and approximately the same adhesion for the hydrophilic and BSA modified surfaces 10.2 ± 9.0 and 9.7 ± 5.1 nN, respectively. They also determined that cells from the stationary phase have the largest adhesion to the freshly cleaved mica, then the fixated cells, and lastly the exponential ones. The adhesive force for the cells in the stationary phase was significant in comparison to those of the exponential phase, measurement of 32.5 ± 14.7 and 9.7 ± 5.1 nN, respectively. The authors suggested that the difference in adhesion of the cell from various phases was due to the difference in surface properties. Cells in the stationary phase have higher surface hydrophobicity and thus, will interact to a greater extent with the hydrophobic surface.
They also found the adhesion to increase on the hydrophilic surface for longer contact time (>5 min).
Ikai et al. developed a method to characterize the intra and intermolecular forces of proteins and polypeptides using an AFM microcantilever115. In their study, they used the monomeric protein carbonic anhydrase II (CAII) and polyglutamic acid (PGA) as the model compounds. The protein was modified at both the N- and C- terminals with cysteine residues, while PGA was only modified at the C-terminal. The modified 25 compounds were attached between the cantilever tip and a silicon substrate using bifunctional cross-linkers and force-displacement data were collected upon tip retraction.
The authors found that a large force peak appeared in the initial stage of stretching (100 to 600 pN) and they suggested that this was due to the non-specific binding of PGA to the silicon wafer. They then, introduced a spacer (polyethylene glycol, PEG) between PGA and the substrate and found the force peak to have shifted up to 25-30 nm. The authors concluded that the mechanical study of proteins and polypeptides can be accomplished by using a spacer to avoid non-specific interactions of the target with the substrate.
Kim et al. investigated the distribution of vitronectin receptors on living murine osteoblastic cells (MC3T3-E1) using an AFM cantilever117. Vitronectin is known as an adhesive lipoprotein that plays an important role in cell attachment. A carboxylated modified polystyrene microbead (10 μm in diameter) was attached to the cantilever tip, to increase contact area, and vitronectin was covalently bound to the bead using EDC/sulfo-
NHS chemistry. The tip was brought in contact with the active MC3T3-E1 cells and force-displacement data was recorded as the tip was retracted. A controlled experiment was carried out using a non-modified bead. The authors converted the force data into separation work and found that vitronectin receptors were localized on the cell surface and the controlled experiment showed significantly less adhesive force or separation work compared to the vitronectin modified beads. In addition, they found that by adding vitronectin directly to the cell solution before exposure to the modified cantilever the separation work reduce significantly. They suggested that this was due to the reduction in binding sites on the cell surface. Furthermore, the authors compared the AFM measurements with immunohistochemistry data and found the results to be in good 26 agreements. That is, the vitronectin receptors are localized on the cell surface. Work done by the same research group, showed that a type I collagen coated microbead attached to an AFM cantilever tip can be used to measure the effects of C6 glioma cells adhesion in presence of lipopolysaccharide (LPS) and phorbol-12-myristate-13-acetate (PMA)118.
LPS and PMA are known to increase the expression of fibronectin. The experimentation was similar to the authors’ previous study. They found the separation work of the collagen bead and the cells to be 0.41 ± 0.39 x 10-17 J/μm2 in the absence of LPS/PMA after 1 day seeding of the cells and in the presence of LPS/PMA the separation energy increased 1.7 times. However, after 4 days seeding of the cells, the separation work was
0.70 ± 0.66 x 10-17 J/μm2 in the absence to LPS/PMA and the separation work only increased by 1.6 times in the presence of LPS/PMA. The authors suggested that LS/PMA may have enhanced the expression of fibronectin from both mRNA and proteins.
2.5.1.4 Vesicles and Cells
Ghatnekar et al. developed a liquid based sensor for detecting phospholipid vesicle in-situ using the dynamic mode of oscillation119. Cantilevers 125 μm long, 35 μm wide, and 4 μm thick were used in this study. These sensors have usable resonant modes in air between 270-310 kHz and the resonant frequency shifts were monitored optically.
Actuation of the cantilever was provided by a piezoelectric layer (PZT). For the resonant peak investigated, its frequency decreased 142.3 kHz and the quality factor decreased from 240 to 60 in going from air to liquid, respectively. The significant reduction of peak quality was due to viscous damping and can be improved by changing the sensor dimensions and geometry. The silicon surface of the cantilever was cleaned with UV ozone before being immersed into the vesicle solution. The authors showed that upon 27 immersing the sensor into the sample solution the resonant frequency started to decrease after 1 minute, and stabilized in 8 minutes. They indicated that the frequency stabilized due to saturation of the sensor surface. The authors estimated the adsorbed mass of vesicles as 450 pg, which was in agreement with the total monolayer coverage of 400 pg.
Several research groups characterized the binding and growth of pathogens and cells on cantilever sensors120-122. For instance, Gfeller and coworkers studied the growth of
Escherichia coli using an array of sensors120. The array consisted of eight cantilever sensors. Each cantilever glass surface was aminated and then coated with agarose. Some of the sensors were inoculated with bacteria, while others were used as reference sensors.
The experiments were carried out in a controlled environment of 37 ± 0.2 °C and 93 ±
2% relative humidity. The resonant frequency was monitored optically by a position- sensitive detector (PSD). Frequency measurements were taken every 30 minutes. The authors observed no change in the resonant frequency of the reference sensors, while for the inoculated cantilevers the resonant frequency decreased exponentially over the first 5 h. The authors estimated the mass change sensitivity of the cantilever as 140 pg/Hz, from which they calculated a detected number of cells as 200. The authors did not provide any growth kinetics, nor did they try to determine growth rate.
In a study reported by Zhang and Ji121, a silicon-based microcantilever was developed for the in-situ detection of E. coli O157:H7. The mechanism of operation of the cantilever was through induced surface stress due to E. coli binding, which cause the cantilever to bend. The sensor deflection was measured optically. The silicon surface of the cantilever was gold coated and then, aminated and subsequently functionalized with E.coli specific antibody (anti-EC). The experiment was carried out in a fluid cell maintained at 20 ± 0.2 28
°C. The cantilever was exposed to various E. coli concentrations. The authors showed results of the cantilever bending immediately upon exposure of the anti-EC cantilever to
E.coli sample at 5 x 106 cfu/mL. However, the bending did not reach a maximum in 5 h and the authors indicated that this may be due to steric hindrance. In addition, when the
E. coli solution was replaced with a buffer, the cantilever’s deflection did not change.
Furthermore, the reference sensor showed little or no deflection in the E. coli sample. The experimental data was fitted to the first order Langmuir kinetic model and the reaction rate was calculated as 2.3 x 10-4 s-1. The authors determined the detection limit of the sensor to be 1 x 106 cfu/mL.
Yi et al. reported on the real-time liquid detection of yeast cells using a PZT/stainless steel cantilever123. The sensor was a composite structure of the two stated materials constructed such that the stainless steel layer overhangs the PZT layer. The PZT was used to actuate and sense the cantilever response, while the overhang stainless steel served as the immobilizing surface. The stainless steel layer was coated with a thin layer of gold, cleaned, and subsequently modified with 0.1% poly-lysine. Fundamental resonance of the cantilever occurred at approximately 22.3 kHz and resonance was monitored by phase angle and impedance using a HP impedance analyzer. The authors determined the cantilevers mass change sensitivity of 4 ± 0.5 x 10-7 and 2.3 ± 0.3 x 10-6g/Hz. Exposing the poly-Lysine functionalized sensor to yeast cells, at 1 and 2 mg/mL, decrease the resonant frequency at both concentration, however, the rate of binding of the yeast from the 2 mg/mL solution was faster. However, the total frequency change for both samples was approximately the same in 1h (120 Hz). Furthermore, they showed SEM photos of the cantilever surface exposed to both samples over several minutes. The photographs 29 indicated that for the 2 mg/mL sample the cantilever surface was saturated in 32 minutes, while for the 1 mg/mL sample saturation was achieved in approximately 61 minutes.
It is important to note that operation of AFM cantilevers in non-contact mode present a challenge in liquid. In this regard Grant and coworkers have reported on the optimization of non-contact mode scanning force microscope (SFM) for liquid imaging of biological samples123. Biological specimens are soft in liquid environment and therefore, contact mode analysis may distort and damage the biologics. The authors designed the scanning force microscope (SFM) in a modular form (two pieces) the top portion as the sample translator and scanner, whilst the bottom part as the sensors, liquid cell, and detection system. The cantilever was excited by a magnetic field, providing a wireless sensor, and a Q control circuit was incorporated that provides phase angle adjustment up to 180°. The problem encountered by the current design is that upon immersion of the cantilever under liquid, the quality of resonance decreases significantly due to viscous damping and therefore, the resolution of the image reduces. The performance of the optimized SFM Q controlled circuit was demonstrated with a cantilever of quality factor in air of 125, which reduced to 4 in liquid without the Q- control circuit, however, with the Q control the quality factor in liquid was 120. Also, the optimized SFM was demonstrated in the imaging of monoclonal antibody to human liver ferritin (anti-ferritin) in liquid, with and without the Q control circuit; the cantilever was magnetically excited in the non-contact mode. Their results showed that the Q-control circuit enhanced the sensor’s sensitivity and spatial resolution as the image was comparable with images obtained in air. 30
Table 2.1: Microcantilever sensors for biochemical measurements
Measurement Sensitivity/Range Cantilever type Detection References Parameters Principle DNA 6 x 1012 chains/cm2 [101] Flexoelectric Bending [101-104] 5 nN [102] [101] mode 1 μM [103] Optical [102,104] [101,102,104 3-8 μM [104] Piezoresistor ] [103] Static stress mode [103] Cystamine 1 mM Optical Frequency [105] [105] Lipoptotiens LDL (1.7 mg/mL) Optical Static stress [106] (LDL) & oxLDL (0.3 mg/mL) mode [106] (oxLDL) IGg 0.195 & 5.85 mg/mL Optical Static stress [106] mode [106] BSA 6 mg/mL [106] Optical [106, Static stress [106-108] 2 mg/mL [107] 108] mode [106] 0.1 mg/mL [108] Piezoresistive Resistance [107] [107] Bending mode [108] PSA 0.2 ng/mL-60 μg/mL [109] Optical [109,104] Bending [104,109- 10 pg/mL [110] Piezoelectric mode [109] 111] 10 ng/mL-1 μg/mL [111] [110] Frequency 6 ng/mL-60 μg/mL [104] Piezoresistive [110] [110] Resistance [111] 31
Table 2.1: Microcantilever sensors for biochemical measurements (continues).
Measurement Sensitivity/Range Cantilever type Detection References Parameters Principle C-reactive 10 ng/mL-1 μg/mL [111] Piezoresistive Resistance [111,112] protein 1 ng/mL [111] [111] Piezoelectric Frequency [112] [112] Biotin-avidin 100 nM Optical Bending [21] Ligand-receptor Concanavalin A-yeast 180 Optical Bending [113-117] force pN [113] mode Yeast to various surfaces [114] PGA-100 to 600 pN [115] Antifluorescyl-IgG to MASFM probe- 120-400 pN [116] Vitronection receptors- MC3T3-E1 cells (10-17 J/mm2) [117]
LPS/PMA to Separation work- 10-17 Optical Frequency [118] collagen J/μm2 adhesion Phospholipid 450 pg Optical Frequency [119] Vesicle Escherichia coli 140 pg/Hz Piezoelectric Frequency [120] E. coli O157:H7 1 x 106 cfu/mL Optical Bending [121] mode Yeast cells 400 ± 0.5 ng/Hz Piezoelectric Frequency [122] 32
Chapter 3: Dynamics of piezoelectric-excited millimeter-sized cantilever (PEMC) sensors
3.1 Piezoelectricity
Piezoelectric materials are classified under the general category of smart material.
Piezoelectricity is one phenomenon that is associated with these materials. A material that is piezoelectric develops a voltage across it’s boundaries when mechanically stressed, which is referred to as the direct piezoelectric effect. Conversely, the application of a potential across the piezoelectric material generates a strain, this is a phenomenon known as the converse piezoelectric effect. In 1880’s the Curie brothers studied the effects of pressure on crystals such as Quartz, Rochelle salt, and tourmaline and found them to generate an electrical charge under mechanical stress. The piezoelectric effect is caused by charge distribution within the crystal structure, upon the application of an external stress that results in an internal electric field. Therefore, the piezoelectric effect allows the conversion of mechanical energy into electrical energy and vice versa. The piezoelectric ceramic used in the construction of PEMC sensors is made from oxides of lead, zirconium, and titanium and is referred to as Lead zirconate titanate (PZT). The
PZT film gives a sensitive response to weak stresses due to the direct piezoelectric effect and generates high strain via the converse piezoelectric phenomenon when a potential is applied. The piezoelectric properties of PZT provide deflection within the cantilever.
The converse piezoelectric effect is used to excite the cantilever and the same PZT layer is used to sense the resulting response at resonance. 33
3.2 General PEMC physics
3.2.1 Theory
PEMC sensors are fabricated to provide predominantly the bending modes of vibration. The PZT film is bonded to a non-piezoelectric material (stainless steel, borosilicate glass, or quartz) forming a composite cantilever. An electric voltage applied across the thickness of the PZT film will lengthen or shorten the film depending on the polarity of the electric field. This change in dimension causes the structure to twist, buckle, or bend and also there could be a combination of the three modes of vibration randomly ordered. If the applied field is alternated periodically, the composite cantilever vibrates at the same frequency as the applied potential. The natural frequency of the cantilever depends on the flexural modulus and the mass density of the composite cantilever47,49,50,61,64. At resonance the cantilever undergoes a significantly higher level of vibration and larger stresses. As a consequence, the PZT layer exhibits a sharp change in impedance, and is conveniently followed by measuring phase angle122.
The natural frequency of a cantilever with a flexural rigidity of EI, where E is the modulus of elasticity and I the moment of inertia, can be obtained by solving the general equation representing transverse vibration129:
∂ 4 y ∂ 2 y ∂y EI + (ρwt) + (c ) =0 (3-1) ∂x 4 ∂τ 2 0 ∂τ
Here, y is the displacement along the thickness of the cantilever, x is the length along the cantilever, τ is time, and ρ is density. The term, c0, is the damping parameter intrinsic to the cantilever. The moment of inertia, I, of a rectangular cross section is wt3/12 where w 34 is the width and t is the thickness. A number of previous investigators have determined solutions to the above model47,85,130,131. A practical expression for the resonant frequency for sensing purposes is obtained when one considers the distributed mass of cantilever to be located at the tip, and is given as 47,64:
2 ν 'n K f n = (3-2) 2π M e
2 2 where ν n =ν 'n 3/ 0.236 , with ν n = 1.8751, 4.6941, 7.8548 and 10.9956 corresponding to the first four eigen values for a rectangular cantilever47. The parameter K is the effective spring constant which depends on the thickness, density, and modulus of the cantilever materials, namely both the non-piezoelectric material and the PZT layer. The solution of Eq. (3-1) can be found elsewhere47,51.
In this project two different geometries of PEMC sensors were investigated; (1) the conventional PEMC sensor (cPEMC) and (2) the free-floating tip PEMC sensor
(ftPEMC). In the cPEMC both the non-piezoelectric layer and the PZT film are anchor at one end and free at the other. Also, at the free end the length of the non-piezoelectric layer is longer than that of the PZT film. As a result the effective tip mass, Me, for a cPEMC of length L can be written as50:
M e = 0.236 (ρ p t p + ρ np t np )wL p + ρ np t np w (L − L p ) (3-3)
Also, the effective spring constant of cPEMC, neglecting the non-piezoelectric tip is50,68:
2 2 4 2 4 2 2 3w ()E pt p + Enptnp + 2E p Enpt ptnp (2t p + 2tnp + 3t ptnp ) K = 3 (3-4) 12L p ()E pt p + Enptnp 35
where subscripts p and np refer to PZT and non-piezoelectric layers, respectively. The ftPEMC sensors have only the PZT layer anchored at one end and at the other end a 1 x 2
2 mm non-piezoelectric layer is bonded. Therefore, the effective tip mass, Me, can be determined as:
M e = ρ p t p wL p1 + 0.236 (ρ p t p + ρ np t np )(L − L p1 ) (3-5)
Also, the effective spring constant of ftPEMC sensors, neglecting the non-piezoelectric layer can be expressed as50:
3E Iw K = p (3-6) L3
When a PEMC sensor is immersed in a liquid, the surrounding fluid offers resistance to its motion. Consequently, the sensor behaves as though an added mass of fluid is attached to it, which results in an additional inertial force in Equation (3-1). This force is in phase with the cantilever motion. In addition, a dissipative force proportional to the velocity of the cantilever should be included to describe the behavior in the liquid. The fluid does not respond instantaneously to cantilever motion, and cause a phase shift between the cantilever motion and the fluid motion. Inclusion of these two effects in Eq. (3-1) results in:
∂ 4 y ∂ 2 y ∂y EI +(ρwt + M ) + (c + c ) =0 (3-7) ∂x 4 a ∂τ 2 0 v ∂τ
where M a represents the added mass of liquid per unit length of cantilever and cv represents the added damping coefficient due to fluid motion. Thus Eq. (3-2) can be 36 rewritten as:
2 ' ν 'n K f nf = (3-8) 2π M e + M a
In detection applications, for instance the sensing of analytes in a fluid, the added mass consists of two terms. The first is due to the fluid surrounding the cantilever and the second is due to the attachment of the analyte at the cantilever’s tip. Thus Eq. (3-8) can be rewritten as:
2 ' ν 'n K f nf = (3-9) 2π M e + mae + Δm
' th where f nf is the resonant frequency of the n mode in fluid when analytes of mass , Δm,
is attached at the cantilever’s tip. In the above mae is the effective added mass of fluid at
the cantilever tip, similar to the effective cantilever mass M e . From the above equation one gets:
' 1 Δm f nf − f nf = f nf (3-10) 2 M ef
th where M ef = M e + mae and f nf is the resonant frequency of the n mode in fluid. That is, a change in resonant frequency represented by the left hand side of Eq. (3-10) is linearly dependent on the change in mass, at a particular resonant mode. The other parameters in Eq. (3-10) are constants for a given cantilever sensor. The effective mass
in fluid, M ef , can be determined experimentally from the resonant frequency under liquid immersion conditions. Thus, one can use Eq. (3-10) to calculate mass change of 37 cantilever due to analyte attachment.
3.2.2 Simulation
As mentioned earlier, the cantilever structure resonates at different modes of vibration. These include the bending, twisting, and buckling modes and also various combinations of the different modes. In order to get some insight on the cantilever vibration, and to predict which geometrical construction will yield predominantly the bending resonance modes, simulation of the various geometries were carried out. The analysis was conducted using FEMLAB® software. Here, the structural mechanics plane stress model was coupled with the piezoelectric plane stress model and using the appropriate sub-domain settings and boundary conditions the various vibrating modes can be determined. Below is shown the results of a model conventional PEMC sensor in 3- dimension. The dimensions of the non-piezoelectric layer (borosilicate glass) were 6 x 2 x 0.16 mm3 and that of the PZT layer (on top) was 2 x 2 x 0.127 mm3. The cantilever was excited by a potential of 0.1 V. The result presented in Figure 3.1A shows the first four modes of oscillation and it can be observed that the fundamental mode of vibration has a distinct resonance characteristic. That is, the first mode of oscillation resonates at a frequency far less than any other modes. Also, the fundamental mode was a bending mode; cantilever oscillates in the Z-direction. The second mode was also a bending mode that resonates approximately 4 times the frequency of the first. This suggests that each mode type has a subset of resonating modes. For example, the bending mode has 1st, 2nd,
3rd etc. modes. In close proximity to the second bending mode was the first twist mode, while approximately 7 times the fundamental frequency the first buckle mode appears.
The figure is a plot of stress level versus position along the sensor; red being the highest 38 stress point and dark blue the slowest. It is clear from the figures that the largest stress was generated in the bending vibration modes, which suggest that the bending mode is the most sensitive of the three mode types discussed.
In Figure 3.1B the next four vibrating modes are presented. Again, the 2nd twist mode and the 3rd bending mode are in close proximity (105.038 and 105.874 kHz, respectively). These were followed by the 3rd twist mode and the 4th bending mode, which were several thousand kHz apart. For the range of frequency investigated only the first buckle mode was observed. This indicates that the buckle modes are high frequency modes. Conversely, the bending modes were the dominant ones. However, it seems as if a twist mode is located next to each bending mode except the fundamental mode and the distance between these two modes widen as the mode number increases. 39
10,893 Hz 45,612 Hz st 1 Bend 1st Flex
41,727 Hz 78,834 Hz 2nd Bend 1st Buckling
A
Figure 3.1A: Simulation results of a cPEMC sensor with dimensions of glass layer 6 x 2 x 2 mm3 and PZT layer 2 x 2 x 2 mm3. The first four modes of oscillation are shown to be bending, twisting, and buckling. 40
105.038 Hz 186,187 Hz 2nd Flex 3rd Flex
105,874 Hz 214,964 Hz 3rd Bend 4th Bend
B
Figure 3.1B: The resonant modes of frequency above 100 kHz of the cPEMC sensor investigated. Each bending mode has a twist or flex mode in closed proximity. However, the frequency between these modes widen with increasing frequency. 41
Chapter 4: PEMC sensor fabrication, characterization, and functionalization
The resonance characteristic of PEMC sensor can be tailored by altering its geometry.
As discussed in Chapter 3, the bending mode of vibration generates a larger stress and is therefore the most sensitive. In this Chapter, the general sensor fabrication techniques, characterization methods, and functionalization procedures are presented.
4.1 General fabrication of stainless steel base cPEMC sensors
Conventional PEMC/stainless steel-based sensors are constructed from two main components: a 127μ m thick PZT single sheet (Piezo Systems Inc., Cambridge, MA) and a 50μm thick stainless steel foil (Alfa Aesar, Ward Hill, MA). A schematic is shown in
Figure 4.1A. The PZT and stainless steel layers, of identical width, are bonded with a conductive epoxy (GC Electronics, Rockford, IL) such that the stainless steel layer is of longer dimension and cured at room temperature for 24 hours. After this, the sides of the cantilevers are sanded down so as to ensure alignment of the two layers. Then, a LCR meter (Agilent) is used to check for short circuit of the PZT layer as indicated by resistance (typically 0.13 MΩ at 1 kHz), capacitance value (typically 1.23 nF at 1 kHz), and dissipation factor (typically 0.02 at 1 kHz). The sensing stainless steel end of the bilayer cantilever was then gold coated (~ 5 nm thick) with an adhesion layer of chromium at the University of Pennsylvania micro-fabrication laboratory to provide an inert surface for liquid immersion. The top nickel surface of the PZT serves as the top electrode, and the non-gold plated stainless steel end, serves as the bottom electrode, are connected using 30 gauge copper wire soldered to BNC couplers. The electrode end of the cantilever, including specific length of PZT and stainless steel layer) is encapsulated in a glass tube by a non-conductive epoxy (VS-101, Huntington Mechanical Laboratories 42
Inc., CA).
4.2 General fabrication of glass based cPEMC sensors
These cPEMC sensors are also fabricated manually as a composite structure of two layers: a 127μ m thick PZT single sheet and a 160μm thick cover glass slip or quartz
(Fisher Scientific). Bonding of the two layers is conducted in the same fashion as described in the above section with a non-conductive adhesive such that the glass layer overhangs the PZT layer to provide surface for antibody immobilization and antigen detection, see a cross section view in Figure 4.1B. However, the electrodes are made only on the top and bottom surfaces of the PZT layer, which is then encapsulated in a glass tube by a non-conductive epoxy for robustness.
4.3 General fabrication of free-floating tip PEMC sensors
It is well-established that if one lowers the cantilever length, the resonant frequency increases. The size of AFM-inspired microcantilevers is a good example of this relationship1-3. While in the case of PEMC this is also true, there is another significant design variable that enables us to achieve the same goal at a longer physical length; the geometry. The PZT-anchored piezoelectric-excited millimeter-sized cantilever
(PAPPEMC) sensors are of different geometry than those discussed in the earlier sections; see Figure 4.1C. The PAPEMC is fabricated from the same materials as the glass based cPEMC sensors: a 127 μm thick PZT single sheet and a 160 μm thick quartz cover square. The PZT layer is the base sensor platform. That is, the cantilever free end is designed with only the PZT layer anchored at one end and at the other end a 1 x 2 mm2 glass is attached. The anchored PZT end is electrode using a 30 gauge copper wire 43 soldered to BNC couplers and epoxied. The protruding PZT layer at the cantilever free end was insulated with a 20 μm thick polyurethane layer. It is worth noting that immobilization only takes place on the exposed glass surface (1 x 2 mm2).
4.4 Characterization of PEMC sensors
The construction of PEMC allows us to vary the relative length of PZT and the glass layer to maximize average stress levels in the PZT layer, which will enhance the sensor’s resonance characteristics, improves mass change sensitivity, and quality factor (Q-value).
4.4.1 Resonant frequency
The natural frequency of vibration of a resonating structure is traditionally used to determine mass change. The resonant frequency is a function of spring constant and effective mass, Eq. (3-2). The spring constant is a material property and is therefore a constant for a given PEMC sensor. Thus, the effective mass of the sensor is the only variable. That is, a change in effective mass, due to mass addition or remove, will alter the resonant frequency. The mechanical resonant frequency of PEMC sensors is measured using an impedance analyzer (HP4192A). The monitoring of resonant frequency is done using the phase angle. The phase angle is a measure of the phase shift between the potential applied to the PZT and the potential generated (signal) due to PZT deformation or it’s the arc-tangent of the ratio of reactance to the resistance of the circuit.
Resonance of the piezoelectric ceramic (PZT) can be represented by the Van Dyke’s equivalent circuit shown in Figure 4.2A. The model is recommended by the IEEE standard on piezoelectricity132-134. The circuit consists of a resistor, inductor, and a capacitor in series and a second capacitor in parallel, which is due to the electrodes on the top and bottom surface of the PZT. At resonance the reactance of the series capacitor and 44 inductor are equal in magnitude and opposite in direction and therefore, cancels each other. Thus, at resonance there is minimum impedance and maximum oscillation amplitude of the resonating structure. At a frequency above resonance the impedance of the circuit is at a maximum; known as the anti-resonant frequency. At anti-resonance the reactance of the inductor and that of the parallel capacitor are of equal magnitude and sign. Therefore, at frequencies below resonance and above anti-resonance the PZT behaves like a capacitor, and between these two frequencies the PZT behaves as an inductor. A plot of phase angle and impedance versus frequency at resonance and anti- resonance is shown in Figure 4.2B. The phase angle spectrum showed a curve at resonance and an inverted Z for the impedance. At resonance the impedance is minimum and maximum at anti-resonance, however the maximum point on the phase curve falls mid way between these two frequencies. This suggests that the frequency at the maximum phase is the point at optimum performance and therefore, the phase peak can be monitored to characterize binding and unbinding processes. Below the resonant frequency and above the anti-resonant frequency the phase angle is close to -90°. At both resonance and anti-resonance the circuit should be purely resistive; phase angle equals to
0°. However, this is not the case because the PZT ceramic is not a perfect capacitor
(larger than normal capacitance at resonance) and therefore, the phase is below 0°
(negative).
In Figure 4.3 the finite element stimulation results of the sensor investigated were in good agreement with the location of the resonant frequencies and the mode of vibration.
Clearly, the dominant mode of oscillation for the sensor was the bending mode.
However, not only can we tailor the geometry of PEMC sensor to a particular vibrating 45 mode, the geometry can be tuned to maximize the generated stress within the PZT and selectively enhance the phase angle of a particular bending mode. An example is shown in Figure 4.4. Here, the spectra of two conventional PEMC sensors are shown. PEMC1 has three peaks in the frequency range investigated with its tallest being the fundamental resonant peak. PEMC2 is of nearly identical dimensions with the exception of its length being 40% shorter than PEMC1. As a result the dominant peak shifted to the second bending peak. Therefore, by altering the cantilever’s dimensions one can select the dominant bending mode of vibration.
The experimental, numerical simulation, and the calculation (from Eq. 3-2) results of the first and second bending modes resonant frequencies for various glass and stainless steel based cantilevers are presented in Table 4.1. The results showed that the calculated resonant frequency values are in better agreement with the experimental ones, compared to the simulation values. For the glass-based cantilevers (A to D), the calculated first and second mode frequencies differ from the experimental values 3-10% and 2-20%, respectively. The simulation values showed a larger difference of 6-29% and 0.3-22% for the first and second modes, respectively. The larger difference in the simulation results may have been due to the thickness of epoxy layer being slightly off. A similar pattern was observed in the stainless steel based cantilevers investigated (1 and 2). The calculated valvues for the first and second bending modes differred by 10-12% and 1-
14%, respectively, while the simulation results showed a significantly large difference of
10-37% and 2-30%, respectively. One notes that the simulation values obtained for glass- based cantilevers falls in a small error range. This may be due to stainless steel kinks during handling and creates permanent indentations when an attempt was made to 46 straighten the material. Conversely, glass does not have such problems.
4.4.2 Mass change sensitivity
The mass change sensitivity is a measure of the smallest mass that can cause a unit change in resonant frequency. The mass sensitivity is represented by σ in units of grams per hertz (g/Hz). The sensitivity can be determined both experimentally and theoretically.
However, experimental determined sensitivity is preferred because it’s more accurate as no assumption is involved. The technique used involves the addition of known mass
(nanograms to picograms) to the sensor while monitoring the resonant peak(s) of interest.
A plot of the mass change versus the frequency change should give a straight line of which the slope is the mass change sensitivity in g/Hz. Manual attaching small masses
(nano and picograms) to the sensor seems to be a difficult task, however, this problem was circumvented by adding a volatile liquid containing a solute. The solute used was paraffin wax. The wax was dissolved into hexane and then nanoliter amounts were dispensed on the sensor surface. The mass of wax within the dispense volume was calibrated using a weighing technique. That is, the weight of several dispense volumes were determined from which the mass of wax in a single dispense volume was computed.
Measuring the frequency change by the added mass gives the mass change sensitivity.
4.4.3 Quality factor
The quality factor (Q-value) is a measure of the shape of a resonant peak. The numerical Q-value is calculated as the ratio of the resonant frequency to the peak width at half the peak height. Typically, the higher the Q-value the more suitable a peak is for detection because the resonant peak can be tracked a lot faster using a smaller frequency resolution. The quality factor can also be looked at as a measure of the dissipative losses 47 during bending. Dissipative losses are of two forms; inertial and viscosity. The viscous losses affect the height of the resonant peak and thus, alter the peak shape and the Q factor. Typically, in going from air to liquid both the resonant frequency and the peak height decreases. Also, the peak broadens causing the Q value to decrease. As a result of the high Reynolds number (Re>105), at the frequency of sensor operation, the viscosity of the liquid does not influence the resonant frequency shift. However, the effective mass of the sensor increases due to the added mass of liquid or target analyte which cause the resonant frequency to decrease as well as the Q value. Therefore, from the Q value the transient mass change of analyte bound to the sensor surface can be determined. For the dynamics represented in Eq. (3-7), under liquid immersion conditions the Q factor can be represented as47,135:
⎡ ρwt + mae ⎤ Q f =(2πf nf )⎢ ⎥ (4-1) ⎣ c0 + cv ⎦
th where f nf is the n mode resonant frequency, co, cv, and mae are constants for a given cantilever and the characteristic mode of vibration. Thus, once their values are known, any further changes in Qf values can be attributed to changes in mass.
4.4.4 Functionalization techniques
Before a PEMC sensor can be used for the detection of a particular analyte, the surface of the sensor (glass layer) has to be intelligent to the interested target. This is done through the immobilization of recognition molecules or ligands specific to the target. There are three functionalization schemes that are used in our laboratory but only one has been optimized. The three protocols are (1) derivatize the glass surface with an amine-terminal silane, (2) gold coat the glass surface with 5 to 10 nm thick layer and 48 then, immobilize protein A, C, or G and (3) functionalize the gold surface with an alkanethiolate monolayer containing specific terminated end groups. Schematics of the various techniques are shown in Figure 4.5 and 4.6. Figure 4.5 illustrates the glass surface functionality; amine-terminal silane is immobilized on the glass surface followed by covalently limking the antibody of interest to the aminated surface via the carboxylic group or the carbohydrate group from residual amino acids or sugars that hang from the antibody heavy region. The main disadvantage of this procedure is that the orientation of the bound antibody cannot be controlled. Figure 4.6 shows the functionalization methods on a gold coated platform. To the gold surface one method is to first attach protein A, C, or G (sulfur atom from cysteine fits into the cavity between three gold atoms that are arranged in <111> crystal structure, the binding energy approximately 35 kcal/mol)136 and subsequent attachment of the antibody. The advantage of this procedure is that protein A, C, and G orient the captured antibody correctly; Fab region is pointed away from the sensor surface. The gold surface can also be functionalized with an alkanethiol monolayer that has a carboxylic or an amine end group to which different chemistry can be applied to immobilize the antibody. In all the methods discussed above it is important to note that the attachment of recognition molecules or ligands to the sensor surface involves only covalent bonds and therefore, requires very harsh conditions to alter the recognition layer.
The recognition layer immobilization technique employed in this project is the amine-terminated silane followed by the antibody attachment via the carboxyl groups,
Figure 4.5. This procedure gives approximately 35 to 40% antigen surface coverage (Ref.
Chapters 6, 9, and 17). This method involves three main steps: cleaning, silanization, and 49 antibody immobilization. The sensing glass surface is cleaned sequentially with methanol-hydrochloric acid solution (1:1 v/v), concentrated sulfuric acid, hot sodium hydroxide, and finally boiling water136. The surface is rinsed between each washing step with deionized water. The cleaning procedure produces reactive hydroxyl groups on the glass surface. After cleaning, the glass surface is silanylated with 0.4% 3-aminopropyl- triethoxysilane (APTES; Sigma-Aldrich) in deionized water at pH 3.0 (adjusted by hydrochloric acid, 0.1 N) and 75oC for 2 hours. APTES reacts with glass leaving a free amine terminal for further reaction with carboxyl group to form a peptide bond. The carboxyl group present in the Fc region of the antibody is activated using the zero length cross linker 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC; Sigma-Aldrich) and promoted by sulfo-N-hydroxysuccinimide (Sigma-Aldrich). EDC converts carboxyl groups into reactive unstable intermediates susceptible to hydrolysis. Sulfo-NHS replaces the EDC, producing a more stable reactive intermediate that is susceptible to attack by amines. Covalent coupling of the stable intermediate with the silanylated glass surface is carried out at room temperature for 2 hours. The glass surface with the immobilized antibody is used to detect the antigen of interest. The immobilization protocol is taken from Bioconjugate Techniques137 and modified as per sample size. 50
Table 4.1: Comparison of the experimental, simulation (FEMLab), and Euler Beam predictions of the first and second bending modes resonant frequencies of various glass and stainless steel based millimeter-sized cantilever sensors.
51
BNC Connectors
A
30 Gauge Wire Glass Casing
Non-Conductive Epoxy Adhesive Stainless Steel PZT
Gold layer
BNC Connectors B
30 gauge copper electrodes Glass Casing
PZT ceramic
Non-Conductive Epoxy Adhesive
Polyurethane coating Glass
C
Glass PZT Epoxy
adhesive
Figure 4.1: Panel A: Two-dimensional cross sectional schematic of the stainless steel base cPEMC sensor. The stainless steel layer is gold coated with a 50 Å thick layer. Panel B: Cross section view of glass base cPEMC sensor. Panel C: Schematic illustration of free-floating PEMC sensor. 52
L1 R1 C1
A
C2
-50 30
-55 B Phase -60 Impedance 25
-65 Ω
-70 20
-75 Impedance, k
Phase Angle (Degrees) Angle Phase -80 15
-85
-90 10 15 16 17 18 19 20 21 22 23 Excitation Frequency, kHz
Figure 4.2: Panel A: Van Dyke’s resonant equivalent circuit of a piezoelectric ceramic. Panel B: Resonant spectra of phase angle and impedance versus excitation frequencies. 53
0
-15
-30
-45
-60
Phase Angle [degree] -75
-90 0 50 100 150 200 250
Frequency [kHz]
Figure 4.3: Resonance spectrum, phase angle versus excitation frequency, and the simulation results of the sensor dynamics. The simulation is in good agreement with the resonant mode type and frequency. 54
-10 -20 PEMC1 -30 PEMC2 -40 -50 -60 -70 Phase angle [degrees] -80 -90 0 50 100 150 200 250 Excitation frequency [kHz]
Figure 4.4: Resonance spectra for PEMC1 and PEMC2. PEMC1 is 40% longer than PEMC2. In PEMC1 the dominant bending mode is the fundamental or first resonance peak, and in PEMC2 it is the second resonant mode is dominant. Therefore, by tuning the cantilever’s dimension one can select the dominant bending mode of vibration.
55
Borosilicate glass/Fused quartz ~160 μm
APTES (silane)
-NH2 functional group
EDC/SulfoNHS NaIO4/NaCNBH3
Carboxylic Group on Glycosidic Group on antibody antibody
Figure 4.5: Glass surface functionalization protocol. The surface is cleaned and then amination with APTES to which the antibody is attached by carboxyl groups or glycosidic groups.
56
Borosilicate glass/Fused quartz ~160 μm
Sputter Cr
Gold <111>
Protein G, C or A Thiolic Acid(s)
EDC/SulfoNHS
Immobilize Ab Immobilize Ab
Figure 4.6: Gold surface functionalization protocol. The glas surface is sputtered with gold in <111> lattice configuration, cleaned and then functionailized with a monolayer of protein A, C, or G or thiolic acid. Subsequent immobilization of specific antibody then follows.
57
Chapter 5: Sensing of Liquid Level at Micron Resolution Using Self-Excited Millimeter-Sized PZT-Cantilever
5.1 Introduction
The measurement of liquid level with micron-level resolution is important in many applications, especially in bio-analytical laboratories and assays. Level detection is also important in many analytical processes where concentration is to be calculated. Liquid levels can be detected and continuously monitored by various techniques such as ultrasonic138,139, acoustic140-142, and optical143-153 techniques. These techniques are limited by scattering, operating frequencies, the type of liquid, and the materials used in the tank construction. None of these techniques has been demonstrated for micron-level liquid level sensing.
Tapered optical fibers have been used to carry out many biological assays154-159. In some of them, sample volume ranged from 150 μL to 400 μL153,154. Since most biological samples are aqueous based and these assays are carried out in semi-open systems, evaporation occurs resulting in changes in sample volume causing errors in concentration determination. It is desirable, therefore, to measure liquid levels in biological assays carried out in multi-well plates, and micro-bioreactors153,154 so that the appropriate volumetric corrections can be made.
In this chapter, we report on the detection of changes in liquid level at micron level using a lead zirconate titanate (PZT) actuated millimeter-sized cantilever. The sensors are a composite structure of two layers: PZT and stainless steel of a few millimeters in length. The PZT layer acts both as an actuating and a sensing element. When a 58 cantilever is transferred from air into a liquid, the liquid offers more resistance to its vibratory motion. The added mass contributed by the liquid and the dampening effect of liquid viscosity decreases the resonant frequency and the quality factor, respectively.
Liquid density and viscosity effects on the cantilever resonance have been reported in the literature 47,52,64,93,131,160. The change in liquid level alters the effective mass of a partially immersed cantilever, thus changing the cantilever’s resonant frequency. Therefore, monitoring of resonant frequency changes are a good measure of the liquid level change on the sensor.
5.2 Cantilever Physics
The effect of added liquid mass on resonant frequency was expressed in Eq. (3-8).
The added oscillating liquid mass is proportional to the displaced mass of the liquid135, and is given by the volume of the cantilever that is immersed and the density of the
liquid. Thus M a = αAρh , where α is the added mass coefficient, A the cross sectional area in the thickness direction, and h immersion depth. The terms α, A, and ρ are constants for a given cantilever and liquid. Therefore, Eq. (3-8) can be written as:
2 ' ν 'n K f nf = (5-1) 2π M e + βh where β = αAρ . Differentiating Eq. (5-1) with respect to h and rearranging, an expression for level change sensitivity (φ,μm Hz) can be obtained:
2()M e + βh φ = − ' (5-2) βf nf
59
The frequency response represented in Eq. (5-1), under liquid immersion condition can be rearranged as follows:
1 '2 = γM e + βγh (5-3) f nf 2 ⎛ 2π ⎞ 1 γ = ⎜ ⎟ (5-3a) ⎜ '2 ⎟ ⎝ vn ⎠ K In the above equation, the parameter γ is a constant cantilever parameter, while β contains cantilever dimensions, liquid property(ρ ), and the added mass coefficient (α ).
In general, the parameter ()α depends on Reynolds number135. A natural consequence of
⎛ 1 ⎞ Eq. (5-3) is that ⎜ ⎟ varies linearly with immersion depth. Experimentally, we can ⎜ '2 ⎟ ⎝ f nf ⎠
' measure f nf as a function of immersion depth ( h ). From these data, the cantilever property (γ ) and liquid property ( β ) can be obtained from a single plot. It should be noted that, since Eq. (5-1) is an approximation for small immersion depths, linear response to immersion depth can be expected only for small values of h . For large values of h , the resonant frequency, may deviate from linearity. Experiments were conducted to test the above relationships.
5.3 Materials and Methods
5.3.1 Cantilever Fabrication
Construction features of the PZT macrocantilever are shown schematically in Figure
4.1A. Several cPEMC sensors were fabricated and used in this study; however, the dimensions of the two cantilever sensors reported are listed in Table 5.1. The length of the PZT and stainless steel layer that protrudes at the cantilevers free ends were 2 and 3 60 mm for sensor A and 1.5 and 3 mm for cantilever B, respectively. The stainless steel layer was gold coated to create an inert surface and was longer than the PZT layer to allow for liquid immersion.
5.3.2 Experimental
Liquid level changes were investigated by two different experimental techniques: electro-mechanical sensing via the PZT-cantilever and by physical measurements using a vernier. The cantilever was mounted vertically on an XYZ-positioner (Optosigma
Corporation, Santa Ana, CA) installed on a vibration-free table (Newport RS 1000MT).
The XYZ manipulator permitted positioning of the cantilever in a one-milliliter liquid sample container at the desired immersion depths with ± 10 µm accuracy. The translation of the cantilever tip was made always in a downward direction, without reversing in order to avoid residual liquid film on cantilever surface, see experimental setup Figure 5.1.
The gold plated stainless steel sensing surface was thoroughly cleaned sequentially with 0.1 N sodium hydroxide, 0.1 N hydrochloric acid, and finally with 95% ethanol.
The surface was rinsed between each cleaning step with deionized water. The test liquids investigated were deionized water, ethanol-water solutions (50% by volume), and mineral oil (Johnson Baby Oil, NY). The change in resonant frequency resulting from the liquid level change was monitored by measuring amplitude ratio and phase angle using an impedance analyzer (HP4192A) with an excitation of 100 mV in 1 to 100 kHz range. For the experimental measurement of liquid level change, a mini-vernier (± 10 µm) was mounted on the inner side of the sample container of constant cross sectional area (0.8 cm2) and one milliliter of deionized water was dispensed into the container. The 61 container was placed adjacent to the cantilever experiment and the liquid level was monitored periodically.
5.4 Results and Discussions
5.4.1 Resonance in Air
Over forty cantilevers of similar dimension were fabricated and used in the various experiments. Since the cantilevers were made individually and manually, no two cantilevers had the same resonance characteristics. However, nearly similar results of liquid level change of the cantilevers were obtained under the same experimental conditions. For brevity, the detailed results of only one cantilever (Cantilever A) are reported. Summary results of Cantilever B are included. Each experiment was repeated at least three times and the data shown are typical of the results obtained with the forty cantilevers that we have examined. The resonance spectrum of Cantilever A (physical dimensions given in Table 5.1) in air is shown in Figure 5.2. Resonant frequency was measured by a sharp change in phase angle131. The first peak in Figure 5.2 is the fundamental frequency, and the higher modes occur at higher frequencies. Note that three resonant frequencies (6.1, 23, and 50.4 kHz) were found in the frequency range, 1 to 100 kHz. This is in contrast to our earlier results with PZT-glass millimeter sized cantilevers where typically only two resonant peaks were observed in the same frequency range163. Furthermore, resonance occurred at significantly higher frequencies in the PZT- glass cantilever, and this is due to the higher flexural rigidity of glass (160 μm thick) than stainless steel (50 μm thick) cantilever. The numerical results and key performance parameters, Q factor and liquid level change sensitivity, are given in Table 5.2. We list the properties of two cantilevers in Table 5.2, which represents typical results of the forty 62 that were used during the course of this study.
5.4.2 Selection of resonant mode
The liquid level change sensitivity depends strongly on the resonant frequency of the oscillating cantilever. The higher the resonant frequency, the more sensitive (lowφ ) the cantilever is for a given level change (Eq. 5-2). That is, the cantilever is able to detect a smaller liquid level change per unit change in resonant frequency if the resonant frequency selected is higher. Our experience with the millimeter-sized PZT cantilever is that the second flexural resonant mode offered higher level change sensitivity of the first three modes and, in most cases, a slightly higher Q factor than at the fundamental frequency. Often, the first and third modes were significantly dampened when the cantilever was immersed in liquid to a depth of one mm. In this study the second mode was selected and was used in all the experiments.
5.4.3 Effect of Immersion on Cantilever Resonance Frequency
In Figure 5.3 we compare the second mode resonant frequency of the cantilever in air with the response obtained when it was immersed at 1.0 mm depth in water. Note that both the resonant frequency and the peak height decrease, has been reported by others
89,122,163. The resonant frequency decreased from 23.05 ± 0.05 kHz in air to 20.25 ± 0.05
kHz in water. The added liquid mass (M a ) causes this decrease in the natural resonant frequency. The mechanical Q factor characterizes the sharpness of the peak. The cantilevers fabricated showed a range of Q factor values in air from 30 to 100163. Table
5.2 gives the Q values of the second flexural mode resonant frequency in both air and at an immersion depth of 1.0 mm in deionized water. The Q factor decreased from 38 ± 2
(n=4) in air to 30 ± 2 (n=4) in water. The reduction in Q value is due to the viscous 63 damping of the surrounding liquid.
5.4.4 Effect of Immersion Depth change on Cantilever Resonance Frequency
Upon immersing cantilever A to one millimeter in deionized water the resonant frequency decreased initially, but subsequently increased with time as shown in Figure
5.4A. In Figure 5.4B the change is plotted against time. The constant rate in resonant frequency change was due to continuous change in level as the water evaporated. The rate of change in resonant frequency was 19 ± 1 Hz/min (n=6). Physical measurements due to evaporation at the experimental conditions used were carried out at 22 oC, and the level changes were 5.1 ± 0.1 μm/min (n=6). The ratio of the rate of liquid level change to the rate of resonant frequency change gives the liquid level change sensitivity (φ) of the cantilever. The liquid level change sensitivity was 0.27 ± 0.01 μm/Hz (n=11) obtained with several fabricated cantilevers. Using the level change sensitivity and the observed change in resonant frequency, the change in liquid level can be determined.
Figure 5.5A shows a plot of the resonant frequency change as a function of immersion depth in deionized water. The frequency response of the cantilever is linear over the small immersion depth investigated. Applying Eq. (5-3) to the results given in
Figure 5.5A, the added mass coefficient was determined from a plot of the reciprocal square of the resonant frequency versus liquid level and is given in Figure 5.5B. The added mass coefficient was 0.021 ± 0.001 (n=4) for deionized water (see Table 5.4).
5.4.5 Effect of Immersion Depth in a Volatile and Non-volatile Liquid
In order to test if the continuous increase in resonance frequency is due to changing liquid level caused by evaporation, the test liquid was changed to one that was more volatile than water (50% ethanol) and one that was non-volatile (Johnson’s Baby oil). 64
The resonant frequency response of cantilever A to the change in the liquid level of 50% ethanol-water solution is shown in Figure 5.6. Here, the rate of change in resonant frequency is significantly higher than that of water. This is due to the higher volatility of ethanol compared to water at room temperature. The rate of resonant frequency change was 23 ± 1 Hz/min (n=4). A similar experiment conducted with Johnson’s Baby oil showed no change in resonant frequency over 75 minutes. The resonant frequency remained constant within ± 5 Hz, suggesting that there was no change in oil level. To alter the resonant frequency, the immersion depth was changed by sequentially moving the cantilever tip using the micromanipulator over a wide range of depths (see Figure
5.7A). As the immersion depth increased the resonant frequency decreased. For short immersion depths (0 – 1.0 mm) the resonant frequency decreased linearly, but for larger depths the change showed a nonlinear behavior.
A plot of the reciprocal square of the resonant frequency versus immersion depth in oil is shown in Figure 5.7B. It can be seen that beyond 1.0 mm immersion depth the plot deviates from linearity, which is in agreement with the deionized water results discussed earlier. A plot such as Fig. 5.7B provides rich data on both the cantilever and
the liquid. From Eq. (3-3) and (5-3a), the parameter (γM e ) can be computed and compared with the intercept value in Fig. 5.7B. More importantly, this value should remain constant on a plot such as Fig. 5.7B as it is only a function of the cantilever’s mass and spring constant. In addition, the slope of the line in Figure 5.7B is the parameter βγ , and thus the fluid property ( β ) can be obtained. The experimentally determined parameter values as well as theoretically calculated values are summarized in
Table 5.4. From the slope, the added mass coefficient (α ) of the mineral oil was 65 determined as 0.003 ± 0.0001, which is lower than for deionized water (0.02), presumably due to its lower density. The y-intercept value for water and oil was 1.56 x
-9 -9 -2 10 and 2.02 x 10 Hz , respectively. These values are within ± 15% of the γM e values predicted from theory (1.84 x 10-9 Hz-2). Therefore, the calculated γ values for both water and the oil experiments (4.32 x 10-4 and 5.55 x 10-4 kg-1 Hz-2) are in good agreement with theoretically determined value of 5.09 x 10-4 kg-1 Hz-2.
5.4.6 Experimental determination of the rate of liquid level change
Physical measurements of the rate of liquid level change due to evaporation were carried out by mounting a mini-vernier, resolution of 10 μm, to the inside of the sample container. The liquid level changes were recorded as a function of time as the liquid evaporated at experimental conditions. A plot of liquid level variation against time yields a straight line whose slope is the rate of liquid level change. Figure 5.8 shows a plot of the experimental level change for deionized water. The rate of level change was 5.1 ±
0.1 μm/min.
5.4.7 Modeling of liquid level change
Since we were able to measure continuously resonant frequency change, it is possible to model the characteristics of the level change and extract the liquid level change sensitivity from the cantilever response. The model is developed on the basis that the liquid level change is due to evaporation. A schematic of cantilever tip immersion and relevant dimensions are given in Figure 5.9. In an open container, the mass of liquid evaporates from the open surface depends on the concentration gradient in the vapor phase. Mass balance of the liquid results in: 66
dh − k (C − C ) = C S ∞ (5-4) dτ ρ
where h is the immersion depth, kC is the convective mass transfer coefficient, ρ is the
density of the sample liquid, C S is the surface concentration of the evaporating entity in
vapor phase, and C∞ is the concentration in the room . The mass transfer coefficient can be estimated from flow across a flat plate analysis. In the present experiments Reynolds number was determined to be less than 80,000 and therefore, the convective mass transfer coefficient is given by161:
1 1 ⎛ D ⎞ ⎛ dV ρ ⎞ 2 ⎛ μ ⎞ 3 AB ⎜ air ⎟ ⎜ ⎟ k c = 0 .664 ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ (5-5) ⎝ d ⎠ ⎝ μ ⎠ ⎝ ρ air D AB ⎠ where d is the diameter of the sample container, DAB is the mass diffusivity in air, and V,
ρair, and μ are the velocity, density, and viscosity of the circulating air, respectively.
These parameters (d, V, ρair, and μ ) are constant and their values are given in Table 5.3.
Integrating Eq. (5-4), the liquid level on the cantilever's sensing surface can be expressed as:
k (C − C ) *τ h(τ ) = h − C S ∞ (5-6) 0 ρ
For deionized water the above equation with parameters in Table 5.3, gives:
−03 h(τ ) = h0 − 5.5 x10 *τ (5-7)
67
where h0 is the initial immersion depth, h0 = 1mm . Eq. (5-7) contains only published values and parameters, and no measured values from our experiments. The above indicates that deionized water level changes at a rate of 5.5 x 10-3 mm/min due to evaporation at 22 0C, and is in good agreement with the experimental value of 5.1 ± 0.1 x
10-3 mm/min (see Figure 5.8). The rate of resonant frequency change corresponding to the rate at which the water level change was 19 ± 1 Hz/min. Therefore, the liquid level change sensitivity can be calculated as 0.26 ± 0.01 μm/Hz. Similar results were obtained with Cantilever B (see Table 5.2). Liquid level change sensitivity for the various cantilevers fabricated ranged from 0.2 to 0.35 μm/Hz, if the resonance occurred in the vicinity of 25 kHz.
5.5 CONCLUSIONS
In this chapter, we have shown that the piezoelectric-excited millimeter-sized stainless steel cantilever have the sensitivity for continuous monitoring of liquid level changes at the micron-level resolution in small sample volume. The liquid level change sensitivity of the sensor was 0.26 ± 0.01 μm /Hz, which suggests that a liquid level change of 0.26 μm is discernable with 1 Hz frequency resolution. The experimental results showed that the rate at which the resonant frequency change depended on the type of liquid tested. The level change of deionized water at room temperature and humidity was found to be 5.1 ± 0.1 μm/min, and is in good agreement with first principles model calculation of 5.5 μm/min. A model of the added liquid mass as proportional to displaced liquid mass provides a mathematical model that predicts this linear behavior. The model agrees with experimental values within 10% error. A plot of the reciprocal square of resonant frequency versus immersion depth provides a method for determining 68 cantilever’s spring constant and the added mass coefficient. Measurement with mineral oil showed that changes in resonant frequency vary linearly with immersion depths less than 1.0 mm. It was also shown that the resonant frequency change was non-linear for larger changes in immersion depths. The significance of these results is that sub-micron level changes can be measured quite accurately.
69
Table 5.1: Dimensions of the piezoelectric-excited millimeter-sized stainless steel cantilever sensors free ends.
Cantilever Dimensions PZT Stainless Steel
L (mm) 2±0.05 3±0.05
A W (mm) 2±0.05 2±0.05
t (μm) 127±5 50±0.50
L (mm) 1.5±0.05 3±0.05
B W (mm) 2±0.05 2±0.05
t (μm) 127±5 50±0.50
70
Table 5.2: Resonance characterization of the first three bending mode peaks of cantilevers A and B.
Cantilever Mode Resonant Q factor in Resonant Qf factor Liquid level frequency air frequency [kHz], 1.0 mm change sensitivity [kHz] in air 1.0 mm in water in water μm/Hz
1 6.1±0.05 25±2 5.2±0.05 ------A 2 23.0±0.05 38±2 20.2±0.05 30±2 0.26±0.01 3 50.4±0.05 ------1 12.4±0.05 30±2 11.4±0.05 ------B 2 33.6±0.05 42±2 31.0±0.05 36±2 0.31±0.01 3 57.6±0.05 58±2 50.0±0.05 ------
71
Table 5.3: Measured and calculated values of model parameters for deionized water.
Parameters Values Units
Mass diffusivity 0.26 cm2/s Wet bulb temp 16.50 oC Dry bulb temp 22.00 oC Air velocity 0.24 m/s Air density 1.20 kg/m3 Air viscosity 1.82(-5) Pa s Water density 1000 kg/m3 Oil density 822 kg/m3 Diameter of Container 0.01 m Convective mass transfer rate 0.018 m/s
3 Surface concentration (Cs) 0.019 kgH2O/m
3 Infinite concentration (C∞) 0.014 kgH2O/m 72
Table 5.4: Cantilever performance parameters for the different liquid samples.
γ β α M e
Parameters [x104 (kg Hz2)-1] [x107 (kg m-1)] [x102] [x106 (kg)]
Predicteda Experimentalb Experimentalc
Oil 5.09 5.55±0.02 2.59±0.01 2±0.1 3.61 Water 5.09 4.32±0.02 20.9±0.01 0.3±0.01 3.61 a From Eq (5-3a) b Calculated from data given in Figure 5.5(B) c Calculated from data given in Figure 5.7 (B)
73
Impedance Analyzer
XYZ Micromanipulator
Sample container
Vibration free Table
Figure 5.1: Experimental arrangement showing the overhang stainless steel layer immersed one millimeter into the test liquid. Note that the PZT layer was not immersed in liquid.
74
-60
-70
-80 Pase Angle (degree) Angle Pase
-90 0 20406080100 Frequency (kHz)
Figure 5.2: Resonant spectrum of phase angle versus frequency of cantilever A
in air. At resonance, the phase angle of the oscillating cantilever (100 mV
excitation) exhibits a sharp peak. 75
-60 Air
Water
-70
-80 Phase Angle [degree] Phase Angle
-90 18 19 20 21 22 23 24
Frequency [kHz]
Figure 5.3: Second bending mode resonant peaks in air (right) and in water (left) immersed at 1.0 mm depth. The liquid adhering to the cantilever surface became part of the oscillating mass, and thus reduces resonant frequency. One note the decrease and widening of the phase angle peak upon immersion. 76
-60 A t=0 t=60mins
-70
-80 Phase angle [degree] Phase
-90 19 19.5 20 20.5 21 21.5 22
Frequency [kHz]
1200
1000 B 800
600
400
200 Increase in Resonant frequency [Hz] Resonant frequency in Increase 0 0 10203040506070 Time [min]
Figure 5.4: Panel A: Shifts in the second mode resonant peak of cantilever A, initially immersed at 1.0 mm in deionized water. The resonant peak shifted in the direction of the arrow (to the right), in 10 minutes intervals, as the liquid level
decreased due to evaporation. Panel B: The second mode resonant frequency response (± 10 Hz) as a function of time. The linear change is due to a constant evaporation rate. The rate of change of resonant frequency was 19 ± 1 Hz/min. 77
1200
1000 A 800
600
400
200
Resonant frequency change [Hz] 0 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 Depth [mm]
2.5E-09
2.4E-09 B ] -2 2.3E-09 [Hz -1 ) nf 2 f ( 2.2E-09
2.1E-09 0.7 0.8 0.9 1 Depth [mm]
Figure 5.5: Panel A: Plot of resonant frequency change versus the level of deionized water on the tip of cantilever A, initially immersed at 1.0 mm. Panel B: The inverse square of resonant frequency versus immersion depth of cantilever A in deionized water. The slope is 9.02 x 10-10 Hz-2 m-1 and the y- intercept is 1.56 x 10-9 Hz-2. 78
1600 ]
1200
800
400 Resonant frequency change [Hz change frequency Resonant
0 0 10203040506070
Time [min]
Figure 5.6: Second mode resonant frequency (± 10 Hz) response of cantilever A initially immersed at 1.0 mm in 50% ethanol-water solution. The rate of resonant frequency change was 23 ± 1 Hz/min, which is larger than that of
deionized water under the same experimental conditions.
79
0 A -500
-1000 ncy change [Hz] change ncy -1500
-2000
-2500 Resonant freque 0.00 0.50 1.00 1.50 2.00
Immersion depth [mm]
2.7E-09
2.5E-09 ] -2
[Hz 2.3E-09 -1 ) nf 2 f ( 2.1E-09 B
1.9E-09 0.0 0.5 1.0 1.5 2.0
Depth [mm]
Figure 5.7: Panel A: The second flexural mode resonant frequency versus immersion depth (±
10 μm) of cantilever A in mineral oil. The increase in the mineral oil level on the cantilever tip increases the sensor’s effective mass resulting in the decrease in its resonant frequency. Panel
B: The reciprocal square of the resonant frequency versus immersion depth in mineral oil. The response is linear only for small level changes (1 mm) because the model is only for point mass analysis (mass at the tip). Therefore, as the mass distribution becomes significantly large on the cantilever free end and the frequency response became non-linear. The slope is 1.45 x 10-10 Hz-2 m-1 and the y-intercept is 2.02 x 10-9 Hz-2. 80
9 8
] 7 6 5 4 3 2 Liquid level change [mm 1 0 0 500 1000 1500 2000 Time [min]
Figure 5.8: Physical measurements of deionized water level (± 10 μm) as a function of time. Note that the change in liquid level is a linear function of time as is the resonant frequency change. The rate of liquid level change was 5.1 ± 0.1 μm/min. Experiment was conducted at 22 0C.
81
C infinity
N A
C s ho h h=0 Liquid (ρ)
Figure 5.9: Schematic of cross sectional view for the partial immersion of stainless steel cantilever in sample liquid.
82
Chapter 6: Detection of Pathogen Escherichia coli O157:H7 Using Self-Excited PZT-Glass Microcantilevers
6.1 Introduction
Enterrohemorrhagic Escherichia coli O157:H7 is an epidemiologically significant cause of food-borne disease originating from contaminated ground beef, milk and poultry products. It is easily transmitted not only via contaminated food, but also from person to person through the use of common facilities. There have been several cases of large scale food recall164 and tens of fatalities165. The US Food Safety Inspection Service has established a zero tolerance threshold for E. coli O157:H7 contamination of raw meat products166. The infectious dosage of E. coli O157:H7 is 10 cells and the Federal EPA standard in water is 40 cells per liter 165,167,168.
Resonant microelectromechanical systems (MEMS) have gained considerable interest over the last decade as sensors. A resonant MEMS relies on changes in its fundamental or higher mode resonant frequency changes to measure a number of different parameters, including absorption of biological47,101,122,169, gases50,85,86,130, chemicals 89,90,94,98,170 and others. More recently, applications in biological systems such as avidin-biotin171,172, antibody-antigen interaction173-175, and hybridization of complementary DNA strands171,176 have been reported. Our own work relating to cellular attachment was based on non-specific binding of yeast on polylysine-coated stainless steel microcantilever with lead zirconate titanate (PZT) actuation122. In general, detection of a biological molecule or an entire bacterium requires the immobilization of a recognition molecule, such as an antibody, on the sensor surface. When the target of interest binds to the cantilever’s sensing surface, the effective mass of the cantilever increases which alters the 83 cantilever’s resonant frequency. Monitoring the resonant frequency change with time provides quantitative measures of the analyte. In this chapter, we explore the application of PZT-glass microcantilevers for pathogen detection.
Very recently Zhang and Ji121 reported the use of antibody-immobilized silicon microcantilevers for detecting E. coli O157:H7 where they used cantilever deflection as the primary measurement of detection. They showed results that suggest that the sensitivity of detection is 1x106 CFU/mL. In this study, we report on the detection of the same pathogen, E. coli O157:H7, at a much lower concentration (700 bacteria/mL) using
PZT-glass cantilever. The amine-derivatized glass cantilever tip was immobilized with a monoclonal antibody (MAb) to the pathogen, and subsequently the tip was exposed to various concentrations of the pathogen. The significance of the results we report herein is that millimeter sized PZT cantilevers have the sensitivity of very low bacterial concentration and therefore may be useful in meat packing houses and other practical situations. Since the measurement involves impedance, and not nanometer scale deflection, it is quite robust and is less complex.
6.2 Materials and Methods
6.2.1 Microcantilever Fabrication
Construction features of the PZT-glass microcantilever is shown schematically in
Figure 4.1B. However, the cantilevers used in this study were fabricated from three main components: PZT ceramic, stainless steel foil, and glass. The PZT and silica glass cover slip were cut to 2 x 8mm2 and 2 x 12mm2, respectively. The stainless steel foil was cut into 2 x 15 mm2. The stainless steel film, which serves as the bottom electrode, was bonded to the PZT layer with a conductive epoxy. After this, the sides of the cantilevers 84 were sanded down and a LCR meter (Agilent) was used to ensure that there is no short circuit. The PZT-stainless steel composite was bonded to the glass and then clamped and cured overnight. The top nickel surface of the PZT, which serves as the top electrode, and the stainless steel were connectarized using 30 gauge copper wire soldered to BNC couplers. The electrode end of the cantilever was encapsulated in a glass tube by a non- conductive epoxy.
6.2.2 Experimental Arrangement
The fabricated cantilever was firmly attached to a XYZ position manipulator (Sigma
Corp., Santa Ana, CA) using a clamp. The manipulator was used to adjust the cantilever in the vertical direction from its reference point to various immersion depths in the sample containing beaker. Typical sample volume was one mL. The cantilever electrodes were connected to an impedance analyzer (Agilent, HP4192A) interfaced to a PC for obtaining impedance, phase angle and amplitude ratios at various frequencies in the range of 1 to 100 kHz with an excitation voltage of 100 mV. Typical Q factor obtained was in the range of 32 to 121 for the second mode of vibration.
6.2.3 Antibody Immobilization
The sensing glass surface was thoroughly cleaned and functionalized with APTES.
Then, monoclonal antibody (MAb) to E. coli O157:H7, IgG2 (Kirkegaard and Perry Lab
(KPL), Gaithersburg, MD), was covalently immobilized on the glass surface EDC/Sulfo-
NHS chemistry. The glass surface with the immobilized antibody was used to detect the pathogen E. coli O157:H7. The reader is referred to Chapter 4, section 4.4, for detail description on the immobilization technique.
85
6.2.4 Detection Experiments
Positive control E. coli O157:H7 (KPL, MD) was used to prepare stock solution
(7x109 cells/mL) using vendors rehydration protocol in 10 mM phosphate buffered saline solution pH 7.4 (PBS). Lower concentrations (7x107 cells/mL, 7x106 cells/mL, 7x104 cells/mL, and 7x102 cells/mL) of the antigen were also prepared in PBS by serial dilution.
One mL of this sample was loaded into a 1 mL container, and the functionalized sensing tip of the cantilever was immersed to a depth of 1mm using the XYZ-manipulator. The impedance and phase angle of the PZT-layer at 1 to 100 kHz was monitored and recorded. Subsequently, the cantilever was immersed to 1 mm in a low pH buffer,
HCl/PBS pH 2.0, for 1 hour to characterize the release of the antigen from the cantilever surface.
In order to evaluate the exclusive selectivity of the antibody-immobilized cantilever to the pathogenic bacteria, the cantilever response to non-pathogenic E. coli in solution was measured. Test samples containing both a non-pathogenic strain and the pathogen were prepared. Stock solution containing E. coli O157:H7 was mixed with a non- pathogenic strain of E. coli JM101 in proportions of 0%, 50%, and 100%. The total bacterial count was kept constant at 7x107 cells/mL. The detection experiments with these samples were carried out in the same manner as with pure E. coli O157:H7.
6.2.5 Experimental Determination of Mass Sensitivity of Microcantilever
Point mass of silicone oil (50 cp) was dispensed at the cantilever tip using a tapered
670 microns silica optical fiber. The tapering was done by flame drawing using a micro- torch (Microflame Inc., Plymouth, MN). The tapered tip, 20 microns in diameter, was dipped in the silicone oil to a prescribed depth, followed immediately by touching the 86 cantilever surface. This was executed by the same person, and with due practice reasonable reproducibility was achieved. Error in measuring mass of 0.02 mg was
± 0.0033 mg. The resonant frequency of silicone-added cantilever was measured in air.
The procedure was repeated for six mass changes, and several times for each mass change. Eq. (3-10) was used to determine the added mass of silicone oil. The resonant frequency changes resulting from these specific mass changes were plotted to obtain experimental measures of mass sensitivity of the cantilevers.
6.2.6 Systematic Measurement Correction
All experiments were carried out without the sample being covered or without constant temperature control. While temperature remained within 0.1 oC during a typical
1 hour experiment, the liquid level in the 1 mL sample container decreased by 4.5 microns per minute during an experiment. The linear change in liquid level with time resulted in a linear change in resonant frequency. Thus immediately before and after a detection experiment, the rate of change of resonant frequency due to liquid level change was measured and an average value of the two was used to correct the peak position in the detection experiment.
6.3 Results and Discussions
6.3.1 Resonant frequency in air
Several cantilevers were fabricated. Resonant frequency is conveniently followed by sharp change in phase angle121. The resonance spectra of the two typical cantilevers used in this study (labeled A and B; physical dimensions in Table 6.1) in air are shown in Fig.
6.1. The numerical results and key performance parameters are given in Table 6.2. Note that only two resonant frequencies (12.8 and 65.8 kHz for Cantilever A and 10.8 and 63.4 87 kHz for Cantilever B) were found in the frequency range, 1 to 100 kHz. The weak resonance at 41.7 kHz for Cantilever B is believed to be due to widthwise rather than lengthwise vibration of the cantilever. It is worth noting that for a slightly different characteristic geometry, length and width, the measured resonant frequencies are essentially similar in magnitude. It is also interesting to note that the ratios of second mode frequency to the first mode are respectively, 5.14 and 5.87 for cantilever A and B.
The relationship given for a rectangular cross-section cantilever (Eq. 3-2) predicts a value of 6.27. The difference is primarily due to the two-layer construction. Table 6.3 summarizes the parameters calculated from experimental data. It is important to note that the added mass of liquid on cantilevers A and B when immersed 1mm was calculated to be the same.
6.3.2 Quality factor in air
The mechanical quality factor (Q factor) characterizes the sharpness of the peak. As shown in Table 6.2, typical values range from 30 to 100, and this value does not deteriorate significantly upon immersion in water-like fluid. While there is a loss of about
10 to 20% in Q value, the decrease is far less than what one observes with silicon cantilevers177,178. In Table 6.2, cantilever B showed higher Q values compared to cantilever A. We believe that this is due to the thickness of the adhesive layer between the layers of PZT, stainless steel and glass. Further more, the length of the PZT layer in cantilever A is longer (Table 6.1) and so is the adhesive layer. As a result, the intrinsic damping of cantilever A is significantly larger (Table 6.3) and the Q factor is lower compared to cantilever B. 88
6.3.3 Mass change sensitivity in air and under liquid immersion
The mass change sensitivity of each cantilever in air was established by adding small masses (dip-touch silicone oil) to the cantilever tip and then observing changes in resonance behavior. A typical response for point mass changes ranging from 0.29 to 2.87
μg for cantilever A in second mode resonance is shown in Fig. 6.2A. The resonant frequency change when plotted against mass additions (Fig. 6.2B), yields a straight line
whose slope is termed mass change sensitivity in air,σ na , expressed in g/Hz. For the case
-8 illustrated in Figure 6.2, the mass change sensitivity in air (σ 2a ) is 1x10 g/Hz. This suggests that if resonant frequency resolution is 1 Hz, then mass changes of 10 ng are
discernable. The sensitivity under liquid immersion condition can be estimated fromσ na using Eq. (3-9). Differentiating Eq. (3-2) and Eq. (3-9) one gets the mass change sensitivity characteristics in air and liquid, respectively.
dM e − 2M e σ na = = (6-1) df n f n
dM ef − 2M ef σ nf = = (6-2) df nf f nf
th where σ na and σ nf are the mass change sensitivity of the n mode in air and in fluid,
respectively. It is clear from Eq. (6-1) and (6-2) that lower values of M e or M ef and
higher values of f n or f nf , result in greater mass change sensitivity for a given resonant frequency measurement. The ratio of Eq. (6-1) and Eq. (6-2) gives:
89
⎛ M f ⎞ σ = σ ⎜ ef . n ⎟ (6-3) nf na ⎜ ⎟ ⎝ M e f nf ⎠
In this study we have calculated values for M e , M ef , f n , and f nf (see Table 6.3). The
estimated values of σ nf were computed and are given in Table 6.2.
6.3.4 Resonant frequency in liquid immersion
Upon immersion of a cantilever in a liquid sample, one observes that the resonant frequency decreases, and subsequently it increases at a constant rate as shown in Figure
6.3. The constant change in resonant frequency is due to change in liquid level as it evaporates. Physical measurements due to evaporation at experimental conditions were carried out, and typical level changes were 4.5 μm/minute. The cantilever is sensitive enough to respond to this change. Since typical measurement time is about one hour or longer, the observed systematic change was characterized prior to and immediately after each detection experiment and correction to the observed resonant frequency was made to obtain the true resonant frequency change.
Figure 6.4 compares the resonant frequency of both cantilevers in both air and when it is immersed to a depth of one millimeter in PBS. Note that the resonant frequency decreases, and the peak height of the phase angle also decreases50,85,86,89,90,101,131.
Although Q-factor decreases, it does so only by about 10-25% for the cantilevers investigated. For oscillation in air, if we neglect viscous damping due to air, the cantilevers intrinsic damping parameter can be determined. For example, for cantilever A at its second resonant mode, the value of c0 is 46.5 kg/m-s, and for cantilever B, it is 6.55 kg/m-s. For the various cantilevers that have been fabricated in our lab A and B are 90 representative of the damping parameter obtained. Since material constants do not vary, it is believed that the rather high c0 value is due to the bonding epoxy layer.
In Fig. 6.4A the phase angle is plotted against frequency of the second flexural mode
(cantilever A). The resonant frequency decreases from 65.8 kHz in air to 57.5 kHz in
PBS. Resonant frequency is directly proportional to the inverse square root of the cantilevers’ mass. Therefore, added mass due to liquid causes a decrease in its natural resonant frequency. The quality factor of the resonant peak decreases from 33 in air to 32 in PBS. The second flexural mode of cantilever B (see Fig. 6.4B) showed a frequency decrease of 10.6 kHz, from 63.4 kHz in air to 52.8 kHz in PBS, when immersed to a 1 mm depth. The Q factor decreased from 121 in air to 100 in PBS. In these two cases, the corresponding mae and cv values can be computed if we assign changes in resonant frequency to changes in added mass only, and the change in damping to both mae and cv.
As shown in Table 6.2, the decrease in resonant frequency in PBS, and the consequential decrease in Q can be used to determine the damping parameters expressed in Eq. (4-1)
47,135. The added mass (mae) and the viscous damping parameter (cv) of cantilevers A and B immersed one millimeter in PBS was also determined. For cantilever A as 9.40 x
10-7 kg and 6.44 kg/m-s respectively, and for cantilever B as 6.70 x 10-7 kg and 2.04 kg/m-s, respectively (see Table 6.3). Note, that the added mass of liquid for both cantilevers immersed 1mm is in good agreement.
6.3.5 Selection of flexural mode for detection
Since the cantilever exhibits several resonant modes of vibration, selection of a particular mode for sensing is an important consideration. The mass change sensitivity depends highly on the resonant frequency of the oscillating cantilever. Eq. (6-1) and Eq. 91
(6-2) suggests that the higher the resonant frequency, the more sensitive the cantilever is for a given mass of cantilever. That is, the cantilever is able to detect a smaller mass change per unit change in resonant frequency if the resonant frequency selected is higher.
Our experience with the fabricated PZT-glass cantilever is that the second flexural resonant mode offers higher mass change sensitivity and, in most cases, higher Q factor than at the fundamental frequency. Our experience has been that often the first mode disappears when the cantilever is immersed in water to a depth of 1 mm. In this study the second mode was selected for in-liquid detection experiments.
6.3.6 Detection of Pathogen E. coli O157:H7
The response of cantilever A to samples containing the pathogen E. coli O157:H7 at concentrations 7 x 102, 7 x 104, 7 x 106, and 7 x 107cells/mL is shown in Fig. 6.5A. For high concentrations such as 7 x 107cells/mL, the resonant frequency decreases rapidly and reaches a constant value. Total change is approximately 600 Hz. At a much lower concentration, 7 x 102 cells/mL, the change is far slower and steady state was not achieved in 60 minutes. These results show that the rate of change in resonant frequency is concentration dependent. Using the mass change sensitivity of the cantilever under liquid immersion, a value 1.43 x 10-8 g/Hz, the mass of E. coli attached in the first ten minutes from the highest to the lowest concentration are, respectively, 7.66, 5.41, 3.78, and 2.77 μg. The attachment surface area is 4.64 mm2, and an average bacteria has a projected area of 1.4 x 10-6 mm2 (density of bacteria is 1.02 g/cm3). Thus, the maximum mass change that can be attained for complete monolayer coverage of the exposed surface is 5.2 μg. The micrographs described in the next section suggest multilayer attachment is possible and the higher observed values may be due to this phenomenon. 92
6.3.7 Scanning Electron Micrograph
In order to obtain a visual confirmation of the typical attachment density on the cantilever surface, glass cover slips were prepared for Mab immobilization by following the protocol described in section 6.2.3. The glass cover slips were then exposed to EC samples (7 x 104, 7 x 106, 7 x 107 cells/mL) for 10, 30 and 60 minutes. The resulting samples were rinsed in PBS, then in deionized water, and dried prior to mounting on a microscope stage. The samples shown in Figure 6.6 are at 10 minutes. Figures 6.6A,
6.6B, and 6.6C represent EC samples 7 x 107, 7 x 106, 7 x 104 cells/mL, respectively. One notes that the density of bacteria is not very high. In samples exposed for longer times
(60 minutes; data not shown) the coverage was slightly higher. E. coli cells are approximately 1 μm x 2 μm in dimension. In the micrograph, the longest dimension of each fragment is less than 1 μm. The samples of antigen from the vendor contained both whole cells and cell fragments. The micrographs in Figure 6.6 show these cell fragments.
The experimental determination of the mass of E. coli (EC) added for 7 x 107 cells/mL sample is greater than a monolayer. This is due in part to the high agglomeration of the
EC fragments, as can be seen in Fig. 6.6A. In order to ensure that the fragments were of cellular origin, energy dispersive spectroscopy (EDS) on the target yielded typical signatures of biological specimens, namely high carbon and nitrogen with small amounts of sodium, phosphorous, and calcium.
6.3.8 Kinetics of pathogen Attachment
The rate of attachment of E. coli O157:H7 to the cantilever surface depends on the bulk concentration of the pathogen in solution, and the number of available attachment 93 sites on the surface (antibody). Let N be the number of attachment sites and n is the number of sites that are occupied at any time, assuming one active site per cell.
Therefore, N is equivalent to the maximum number of cells the surface can hold. The bulk concentration of antigen in sample after n has attached is given by;
⎛V C 0 − n ⎞ ⎜ s b ⎟ (6-4) Cb = ⎜ ⎟ ⎝ Vs ⎠
0 where Cb is the initial bacterial concentration and Vs the sample volume. The rate of attachment is directly proportional to the bulk concentration (Cb) and available
0 0 attachment sites. WhenVsCb >> N , Eq. (6-4) reduces to Cb = Cb . Thus for samples containing high bacterial count (>106 cells/mL) a good approximation is:
dn = kC 0 ()N − n (6-5) dτ b
From the above, the number of attached cells can be determined from:
n 0 − kC b τ = 1 − e (6-6) N
where k is the attachment rate constant. If the mass of each bacteria is mb , the resonant frequency change is proportional to the attached mass of bacteria and therefore,
m n ()− Δf = b (6-7) σ where ()− Δf is the change in resonant frequency and σ is the mass sensitivity. At steady 94 state when all the attachment sites are filled, total change in resonant frequency is given by:
m N ()− Δf = b (6-8) ∞ σ
Hence the change in resonant frequency with time can be expressed in terms of bacterial
(antigen) concentration as follows:
− kC 0τ b ()()− Δ f = − Δ f ∞ (1 − e ) (6-9)
The above model was tested by plotting a semi-log plot of the ratio of resonant frequency change to the total resonant frequency change versus the product of bulk concentration and time. A sample result is illustrated in Fig. 6.7A for the case of 7 x 106 cells/mL sample. The quality of the fit is good, and from the slope the rate constant for attachment was determined as 3 x 10-9 min-1 (cell/mL)-1. In Fig. 6.7B, the result of 7 x 107 cells/mL is superimposed on the 7 x 106 cells/mL result. Although there is a larger scatter with the data set, the trend suggests that the two experiments give essentially the same attachment rate constant. In part, the scatter resulted from very rapid change in resonant frequency due to the large bacteria concentration. Results of other lower cell concentrations do not
0 satisfy the criterion VsCb >> N and thus were not analyzed in the same fashion. A similar analysis with cantilever B for 7 x 106 cells/mL data (see Fig. 6.7C) suggests an attachment rate constant of 5 x 10-9 min-1 (cell/mL)-1. We believe that the rate constant value will be useful in designing systems where response time is to be optimized. In the measurement arrangement used, there was no circulation or mixing of the sample and thus the attachment rate constant determined is probably the lowest possible value. Other contacting arrangements will be explored to increase the rate constant. 95
6.3.9 Frequency response to different concentrations of E. coli O157:H7 in presence of a wild strain of E. coli
Samples containing both the target pathogen E. coli O157:H7 and a wild strain
JM101 were prepared from stock solutions of each, 7 x 107cells/mL. These samples
(100%, 50%, and 0% wild strain JM101) were then exposed to the cantilever immobilized with MAb. Resulting responses given in Fig. 6.8A showed that the cantilever responds to the target antigen selectively. The change in resonant frequency is higher with higher proportion of the target pathogen, and when it is totally absent the average change in resonant frequency is 0 Hz although the total bacterial concentration is
7 x 107cells/mL (all JM101).
6.3.10 Frequency response to the release of the attached E. coli O157:H7
Attached E. coli to antibody can be released by changing either the pH or salt concentration of the medium. After the attachment experiment was concluded, the cantilever was immersed in a pH 2.0 buffer which caused the attached antigen to be released. The release was signified by a rapid increase of the resonant frequency. See Fig.
6.5B in which release data are given for 7 x 102, 7 x 104, 7 x 106, and 7 x 107cells/mL attached samples. Note that each release experiment was done for one hour. The rise in resonant frequency corresponded to the decrease observed during the detection experiments. This suggests that the mass attached was removed effectively and the largest portion of the mass was removed during the first five minutes. The release of the attached antigen from samples containing wild strain is shown in Fig. 6.8B. The 100% wild strain sample showed essentially no change in resonant frequency, although it fluctuates about 0 Hz. 96
6.4 Conclusions
In this Chapter, we showed that a composite self-excited cantilever made of a PZT film, stainless steel, and glass of a few millimeters in length with immobilized antibody can be used as a sensor for pathogen detection (E. coli O157:H7) under liquid immersion condition. Sensitivity achieved was in the order of tens of nanograms, and experimental results suggest detection of very low pathogen concentration, 700 bacteria/mL, can be successfully carried out. The immobilized antibody showed high selectivity to the pathogen (E. coli O157:H7) in the presence of a non-pathogenic strain and thus, demonstrated its usefulness for detecting water-borne pathogens. A kinetic model for attachment was proposed and verified using high concentration samples. The specific binding rate constant was found to be 3 x 10-9 min-1 (cell/mL)-1 to 5x10-9 min-1 (cell/mL)-
1. Current work is in progress to fabricate PZT-macrocantilevers that exhibits mass change sensitivity at the sub-picogram level so that detection of a single E. coli cell is possible. 97
Table 6.1: Physical dimensions of PZT/Stainless Steel/ Glass cantilevers A and B.
Cantilever Dimensions PZT Stainless Steel Glass
L (mm) 2±0.05 2±0.05 5±0.05 A w (mm) 2±0.05 2±0.05 2±0.05 t (μm) 127±5 50±1 160±5 L (mm) 1±0.05 1±0.05 3±0.05 B w (mm) 1.8±0.05 1.8±0.05 1.8±0.05 t (μm) 127±5 50±1 160±5
98
Table 6.2: Resonance characteristics of cantilevers A and B.
Cantilever Mode Resonant Q Sensitivity Resonant Sensitivity Qf values
frequency factor σ2a, frequency σ2f, in PBS [kHz] in air in air 109[g/Hz] [kHz], 1mm 109[g/Hz] 1mm in PBS immersion
1 12.8±0.05 64±2 230 ------A 2 65.8±0.05 33±2 10 57.5±0.05 14.3 32±2 1 10.8±0.05 83±2 ------B 2 63.4±0.05 121±2 3 52.8±0.05 4.8 100±2
99
Table 6.3: Calculated parameters from experimental data obtained by the second flexural mode resonant frequency of cantilevers A and B.
Cantilever Effective Spring Intrinsic Added mass, Viscous damping Binding rate
mass Me constant, damping mae, 1mm parameter, cv constant, k, 106(kg) K 10-3 parameter, immersion in [kg/m-s] at 1mm 109 [min-1 7 -1 [Eq.3] (N/m) co (kg/m-s) PBS 10 [kg] immersion in (cells/mL) ] [Eq.2] [Eq.8] [Eq.7] PBS [Eq.8]
A 3.71 16.60 46.50 9.35 6.44 3.00
B 1.99 8.27 6.55 6.65 2.04 5.00
100
-78
-80 A -82 -84 -86 -88 -90 0 20406080100 Phase Angle [degree] Frequency [kHz]
-78
-80 B -82 -84 -86 -88 -90 Phase Angle [degree] 0 20406080100 Frequency [kHz]
Figure 6.1: Resonant spectra of phase angle versus frequency of cantilevers
A (in panel A) and B (in panel B) in air. At resonance, the phase angle of the
oscillating cantilever (100 mV excitation) exhibits a sharp peak.
101
A
3.0E-06 B 2.5E-06 Slope = -1x10-08g/Hz 2.0E-06
1.5E-06
1.0E-06 Mass change [g]Mass 5.0E-07
0.0E+00 -200 -150 -100 -50 0 Resonant frequency change (Δf ) [Hz]
Figure 6.2: Mass change sensitivity of cantilever A in air. Point masses of silicone oil (50 cp) were dispensed at the cantilever tip by a tapered fiber tip 20 microns diameter and the resulting resonance spectrum of the second mode is shown in panel A. The resonant frequency changes (±10 Hz) with mass additions (±0.12 μg) gives experimental measures of mass sensitivity, shown in panel B.
102
600
500
400
300
200
100
0 0 20406080100 Resonant frequency change [Hz] Time [min]
Figure 6.3: Typical resonant frequency response (±10 Hz) upon immersion of a cantilever in liquid. The linear change is due to changes in liquid level of the sample as evaporation occurs.
103
-75 A -77 In Air In PBS -79
-81 -83
-85
-87 Phase Angle [degree] -89
50 55 60 65 70 75
Frequency [kHz]
-78 In Air -80
-82 In PBS B -84
-86
-88 Phase Angle [degree] Angle Phase -90 50 53 56 59 62 65 68 Frequency [kHz]
Figure 6.4: Second flexural mode resonant peaks of cantilever A (in panel A) and cantilever B (in panel B) in air (right) and in PBS solution (left) immersed to a depth of 1 mm. The liquid adjacent to the cantilever detection surface became part of its effective mass. As a result its resonant frequency decreased. 104
0 A -100 7E02 cells/ml -200 -300 7E04 cells/ml
-400 7E06 cells/ml [Hz] -500 7E07 cells/ml -600 -700 Resonant frequency change 0 20406080 Time [min]
B 800 7E07 cells/mL 700 600 500 7E06 cells/mL 400
300 7E04 cells/mL 200
100 7E02 cells/mL 0
Resonant frequency change [Hz] 0204060 Time [min]
Figure 6.5: Panel A: Attachment of pathogen at various sample concentrations. Panel B: Release of antigen upon immersion in low pH buffer subsequent to each attachment experiment. 105
A
B
C
Figure 6.6: Samples containing various concentrations (Panel A: 7 x 107 cells/mL, Panel B: 7 x 106 cells/mL, and Panel C: 7 x 104 cells/mL) were exposed to antibody immobilized glass cover slip (same material used to construct microcantilever) and at 10 minutes the cover slip was removed, rinsed and dried. Typical density of attached cell or cell fragments contained on the glass surface is shown at various magnifications. 106
0.0 -0.2 A ) ∞ f
Δ -0.4 )/ f
Δ -0.6 - ∞ f -0.8 Δ
ln(( -1.0 -1.2 0.E+00 1.E+08 2.E+08 3.E+08 4.E+08 5.E+08
0 Cbτ [min(cells / mL)]
0 7E06 cells/ml B -1 7E07 cells/ml ) ∞
f -2 Δ )/ f
Δ -3 - ∞ f Δ -4 ln((
-5
-6 0.0E+00 5.0E+08 1.0E+09 1.5E+09 2.0E+09 2.5E+09
0 Cbτ [min(cells/ mL)]
107
0.0 -0.5 C ) ∞
f -1.0 Δ )/
f -1.5 Δ - ∞
f -2.0 Δ -2.5 ln(( -3.0 -3.5 0.E+00 1.E+08 2.E+08 3.E+08 4.E+08 5.E+08
0 Cbτ [min(cells / mL)]
Figure 6.7: Natural log plots of the ratio of resonant frequency change to the total resonant frequency change versus the product of pathogen bulk concentration and time. Panel A and B is the results using cantilever A and panel C the result of cantilever B. Panel A: Kinetic analysis of 7 x 106 cells/mL experimental data. Slope is 3 x 10-9 min-1 (cells/mL)-1. Panel B: the results of 7 x 107 cells/mL is superimposed on the 7 x 106 cells/mL result. Panel C: (the results of cantilever B) the kinetics of 7 x 106 cells/mL sample is shown. Slope is 5 x 10-9 min-1 (cells/mL)-1. 108
300 100% Wild
0 A
-300 50% e.Coli 0157:H7, 50% Wild
-600
-900 100% e.Coli 0157:H7
-1200
-1500 Resonant frequency change [Hz] change frequency Resonant 0 20406080 Time [min]
1400 B 1200 100% E.Coli 0157:H7 1000 800 600 400 50% E.Coli 0157:H7, 50% Wild 200 0 100% Wild
Resonant frequency change [Hz] change frequency Resonant -200 0 10203040506070 Time [min]
Figure 6.8: Resonant frequency changes versus time for the second flexural mode of
cantilever B. Panel A: Attachment of pathogen E. coli O157:H7 in samples mixed with
wild strain JM101. Panel B: Release of antigen upon immersion in low pH buffer
subsequent to each attachment experiment.
109
Chapter 7: Escherichia Coli O157:H7 Detection Limit of Millimeter-Sized PZT Cantilever Sensors is 70 cells/mL
7.1 Introduction
In Chapter 6 we demonstrated the detection of various concentrations of E. coli
O157:H7 in a batch configuration. Now that we have some knowledge of the kinetics of
E. coli binding and unbinding, we can design experiments to determine the sensor’s detection limit. E. coli is a pathogen that is found in the intestine and can be a causative microbe of extra-intestinal infections165-168.
In this study, we explore the application of the PZT-glass cantilever for detection of the pathogen at very low concentration that is close to the Environment Protection
Agency (EPA) standards. Recently, Zhang and Ji121 reported a rather elegant experiment using silicon microcantilever for detecting E. coli O157:H7 at concentrations in the range of 1 x 105 to 2 x 107 cfu/mL. Their lowest concentration that was discernable was 1 x 106 cfu/mL. In this Chapter we report our success in using PZT-glass cantilever that shows sensitivity down to 70 cells/mL. We have used the same reagents, namely the antibody and the antigen, from the same source as did Zhang and Ji (2004).
7.2 Sensor description
The cantilevers were manually fabricated in our laboratory (Drexel University,
Philadelphia, PA) from PZT and glass. The PZT layer, 5 x 1.8 mm2 (length x width), was bonded to the glass, 6 x 1.8 mm2 (length x width) with a non-conductive adhesive, such that a 2 mm length of the PZT layer and 4mm of the glass layer sticks out at the cantilever free-end. The top and bottom nickel surfaces of the protruding PZT layer were connected using 30 gauge copper wire soldered to BNC couplers. The electrode end of 110 the cantilever was encapsulated in a glass tube by a non-conductive epoxy. See sensor schematic in Figure 3.1B.
7.3 Experimental procedure
The cantilever’s glass surface was cleaned and functionalized with antibody as was discussed in Chapter 4. Stock solution of a positive control E. coli O157:H7, 7 x 109 cells/mL, was prepared using the vendor’s instruction. Lower concentrations (7 x 107 to 1 cells/mL) of the pathogen were prepared in phosphate buffered saline (PBS) (10 mM, pH
7.4) by serial dilution. The detection experiments were carried out without the sample being covered or without constant temperature control. While temperature remained within ± 0.1 oC during a typical 1 hour experiment, the liquid level in the 1 mL sample container decreased by 4.5 microns per minute during an experiment. The linear change in liquid level with time resulted in a linear change in resonant frequency. Thus immediately before and after a detection experiment, the rate of change of resonant frequency due to liquid level change was measured and an average value of the two was used to correct the peak position in the E. coli O157:H7 detection experiment. One milliliter of the test pathogen solution was pipetted into a 1 mL sample container. The functionalized sensing tip of the cantilever was immersed to a depth of 1 mm using a
XYZ-positioner (Optosigma Corporation, Santa Ana, CA) installed on a vibration-free table. The change in resonant frequency resulting from the mass change was monitored by phase angle using an impedance analyzer (HP4192A) with an excitation of 100 mV in the frequency range of 1 to 100 kHz range.
7.4 Results
Several cantilevers were fabricated and used in the detection experiments. However, 111 for brevity only the results from one cantilever are reported so that comparison of responses on the effect of pathogen concentration can be made. Each experiment was repeated at least three times and the data shown are typical of the results obtained. The resonance spectrum (a plot of phase angle versus excitation frequency) of the cantilever in air, Fig. 7.1, showed a fundamental resonant frequency of 10.95 ± 0.05 kHz and a second mode resonance at 43.45 ± 0.05 kHz. Several repeated experiments showed that these two resonant frequencies are stable within ± 50 Hz. In any one experiment, the variance of resonance frequency was ± 5 Hz. In this study, the second mode was selected for in-liquid detection experiments because the first mode was significantly dampened when the cantilever was immersed in liquid.
Figure 7.2 shows the time profile of resonant frequency change of an antibody- functionalized cantilever in the detection of E. coli O157:H7 at concentrations of 70, 7 x
102, 7 x 103 and 7 x 107 cells/mL. For the highest E. coli concentration (7 x 107/mL), the resonant frequency decreased more rapidly initially and reached a lower constant value.
The more rapid rate observed is an expected response, as the binding rate is proportional to concentration. The total change was approximately 1010 ± 5 Hz (n=3). At 70 cells/mL the frequency change was the lowest. The total steady state change in resonant frequency was 58 ± 5 Hz (n=3). Lower concentrations (1 and 10 cells/mL) were investigated. The frequency profile for 1 cell/mL was identical to the control results, indicating that the possibility of a single, in 1 mL buffer solution, to gain contact with the sensor is very small. This experiment was repeated at least six times and in each case the data fluctuates about the zero change in frequency. At 10 cells/mL the sensor showed identical response 2 out of 5 experiments, which could mean that in the two successful 112 experiments the pathogen did bind to the sensor surface. However, the response was only
20 ± 5 Hz, which is in the noise level of the positive control (Fig. 7.3).
The mass change sensitivity of the cantilever under liquid immersion was determined as 1.2 x 10-11 g/Hz. The mass change sensitivity was determined using the silicone oil dip touch technique that we reported in Chapter 6. Using the mass change sensitivity, the mass of pathogen that was bound to the cantilever glass surface after one hour from the lowest to the highest pathogen concentrations were, respectively, 0.69 ± 0.05, 1.66 ±
0.03, 3.25 ± 0.05, and 12.12 ± 1.52 ng. Control experiment (negative control) was conducted with an antibody-functionalized cantilever, anti-E. coli O157:H7, immersed
1mm in PBS solution for 1 h at the same experimental conductions as the E.coli detection. No significant frequency response (± 5 Hz) was observed, suggesting that the change in resonant frequency shown in Figure 7.2 for the various concentrations of E
.coli O157:H7 were due to the binding of the pathogen to the cantilever functionalized glass surface. In order to determine if non-specific binding would occur, a second control experiment (positive control) was carried out and the results are shown in Figure 7.3. In this experiment, the cantilever without any antibody immobilized on its surface was immersed in a solution containing very high concentration of E. coli O157:H7 at 7 x 107 cells/mL. The results in Figure 7.3 show that the change in resonant frequency change was 0 ± 10 Hz. When no cells were present, the antibody-immobilized cantilever showed a variation of ± 5 Hz in PBS (Figure 7.2). Thus it is reasonable to conclude that non- specific attachment, if any, is weak and the variation in resonant frequency increased by
± 5 Hz, and this increase is significantly lower than that observed with 70 cells/mL 113
7.5 Conclusion
In this Chapter we have illustrated that a PZT-glass cantilever functionalized with anti-E. coli O157:H7 can be used to detect the pathogen E. coli O157:H7 under liquid immersion in real time, with a detection limit of 70 cells/mL. The millimeter sized PZT- actuated cantilever is a reasonably robust platform for practical use.
114
-74
-78
-82
-86 Phase Angle [degree] -90 0 20 40 60 80 100
Frequency [kHz]
Figure 7.1: Resonant spectra of phase angle versus frequency of cantilever in air. At resonance, the phase angle of the oscillating cantilever (100 mV excitation) exhibits a sharp peak. 115
200
Control 0 70 cells/ml
-200 700 cells/ml
7000 cells/ml -400
-600
-800 7x107 cells/ml
Resonant frequency change [Hz] -1000
-1200 0 10203040506070 Time [min]
Figure 7.2: Resonant frequency response of the second flexural mode versus time for the binding of E. coli O157:H7 at various sample concentrations to the antibody functionalized cantilever. Control was an antibody-immobilized cantilever immersed in PBS. 116
15
10
5
0 -5
-10
-15
change [Hz] Frequency -20
-25
0 10203040506070
Time [min]
Figure 7.3: Resonant frequency response of the second flexural mode versus time when the clean (antibody-free) cantilever was immersed in E. coli O157:H7 7 at 7 x 10 cells/mL. Normal variation observed in cantilever response is ± 5 Hz. Although the variation observed is slightly higher (± 10 Hz), no net decrease in resonant frequency is observed.
117
Chapter 8: Rapid Assessment of Escherichia coli by Growth Rate on Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensors
8.1 Introduction
Escherichia coli (E. coli) strains are the major bacterial cause of food and water borne diseases, and thus, the propagation of these micro-organisms is an important topic in the food industry. The growth rate of these bacteria has been studied extensively at different temperatures and media compositions179,180; specifically, the enterohemorrhagic serotype
O157:H7 due to its epidemiological outbreak of food borne diseases worldwide. E. coli
O157:H7 readily contaminates ground beef, milk and poultry products. In 1997 one of the largest recalls of ground beef was attributed to E. coli O157:H7 contamination164. The
US Food Safety Inspection Service has established a zero tolerance threshold for E. coli
O157:H7 contamination of raw meat products166. The infectious dosage of E. coli
O157:H7 is 10 cells and the Federal EPA standard in water is 40 cells per liter165.
Furthermore, diseases caused by the subset of the enterohemorrhagic serogroups have created a growing clinical concern. For example, the diarrheagenic E. coli strains which include the enterotoxigenic, enteropathogenic, enteroinvasive, and the eneteroaggregative strains are known to cause several cases of diarrhea in children annually181. Due to the pathogenic nature of E. coli, various kinetic models that predict the growth on meat and carcasses were developed182-186.
Piezoelectric cantilevers have been shown to detect biological entities101,169, chemicals90,98, liquid levels187, and gases85,130. In general, the lead zirconate titanate
(PZT) glass cantilever is a mass sensor whose resonant frequency depends on its effective mass. PZT is relatively inexpensive, and thus using it for exciting cantilever vibration is a 118 reasonable approach. The resonant frequency of the cantilever is inversely proportional to the square root of its effective mass. Therefore, monitoring of the resonant frequency change with time as the target mass changes provides quantitative measurements.
Routine monitoring of meat and food processing plants requires assaying for the presence of indicator microorganisms such as E. coli or coliform Enterobacteriaceae.
This process requires 24 hours to measure the development of colonies. If the growth of organisms can be measured quantitatively in a short time period such monitoring processes can be expedited considerably. Very recently Gfeller et al., (2004) reported using micro-fabricated micromechanical oscillators for the detection of the active growth of E. coli120. Their work was a demonstration of the ability to monitor growth on the cantilever surface. They made no attempt to quantify growth rate nor did they compare the growth rate obtained with values obtained using conventional methodology.
Resonance frequency (~32 kHz) was measured every 15 minutes, and their data showed a fair amount of scatter. Their results indicate that detection of the active bacterial growth can be obtained within a few hours. In this Chapter, we show that a millimeter-sized
PZT-glass cantilever can continuously monitor the growth of E. coli JM101 at 29 oC, and that the quality of data obtained provides an excellent basis for determining growth rate.
That is, we not only detect growth but determine the key metabolic parameter, specific growth rate. In addition, we develop a mathematical model for relating resonance frequency change to cell growth rate that enables interpreting sensor signal to cellular property, growth rate. The significance of the results is that a fairly simple and inexpensive device such as a millimeter-sized cantilever has the mass change sensitivity to rapidly assess E. coli contamination by measuring growth rate. Although the technique 119 we report is not specific to a particular pathogen, its very sensitive characteristics suggests that with suitable selective medium (not within the scope of this Chapter), one can use the technique reported here for pathogen detection. Thus, this device may be useful in food processing plants for measuring bacterial contamination.
8.2 Materials and Methods
8.2.1 Cantilever Dimensions
The cantilever was fabricated from PZT and glass of dimensions of 5 x 1 mm2; see
Figure 4.1B. The layers were bonded such that the cantilever free end had a 2 mm long
PZT layer and a 3 mm long glass layer. The glass layer at the cantilever free end is 1mm longer than the PZT layer to provide a sensing surface for the E. coli has they grows.
8.2.2 Cell culture preparation
E. coli JM101 was purchased from EDVOTEK (West Bethesda, MD). Ten milliliters of AC broth (Sigma-Aldrich) were inoculated with a few colonies of the bacteria which grew overnight at 37 oC before use. Three milliliters of the cell suspension were harvested by centrifugation (Genofuge 16M, Techne, 14,000 rpm, 10 min, room temperature). The supernatant was discarded and the pellet was resuspended in 1 mL AC broth. The inoculation and cell recovery was done in a biological hood.
8.2.3 Cantilever functionalization
The cantilever glass surface was cleaned with Piranha solution (7:3 volume ratio of concentrated H2SO4 and 30% H2O2) for 5 min followed by rinsing with sterile deionized water and ethanol. The cantilever was, then, placed in a sterile biological hood to dry until use. Luria broth (LB) agar (Sigma- Aldrich, PA) was prepared by dissolving LB 120 powder in boiling sterile deionized water at a concentration of 0.15 g/mL. The cantilever was positioned horizontally and immediately 2 μL of the warm agar was dispensed on the top side of the cantilever sensing glass surface and subsequently spread into a 200 μm thick film. The agar film was left to solidify, which was determined by reaching constant resonance frequency, indicating constant mass. The agar film was then inoculated with
0.5 μL of exponentially growing E. coli culture.
8.2.4 Experimental Arrangement
The cantilever loaded with E. coli was positioned horizontally via an O-ring into a temperature-controlled chamber that was mounted on a vibration free table (Newport RS
1000MT), Figure 8.1. The temperature was regulated by water that circulated through a jacket encompassing the chamber at 29 ± 0.2 oC. Two 3 mm diameter silicone tubes of six inches length were connected to the fermentation chamber that provided for oxygen transfer. The closed nature of the fermentation chamber provided for achieving constant humidity. The cantilever electrodes were connected to an impedance analyzer (Agilent,
HP4192A) interfaced to a PC for continuous measurements of impedance, phase angle and amplitude ratio at the fundamental frequency. The frequency was varied between 1 and 100 kHz with an excitation voltage of 100 mV.
8.3 Results
8.3.1 Resonance Spectrum in Air
The resonant air spectrum of the cantilever over the range of 0 to 100 kHz (a plot of phase angle versus frequency) is represented in Figure 8.2. The physical dimensions of said cantilever are listed in Table 8.1. The fundamental mode occurred at 31.08 ± 0.01 121 kHz and had a phase angle of -70.61 ± 0.01 degrees. Higher order modes were found at higher frequencies, but our interest was limited to the aforementioned range. A peak’s utility for detection is determined by its Q-factor and resonance. Typically, a higher resonant frequency leads to a more sensitive peak. In general, the Q-factor is directly proportional to the resonant frequency and inversely proportional to the width of the peak at the midpoint of its height and thus, indicates the sharpness of the resonance peak. A peak that is suitable for detection should have a high Q-value (typically > 20). The Q- factor for the fundamental mode shown in Figure 8.2 was 38.9 ± 0.1, which is in the range of suitable detection peaks. As a result, the fundamental mode was chosen for monitoring the E. coli growth.
8.3.2 Resonance Spectra of the Functionalized Cantilever
As discussed in Section 8.2.3, both an agar film and an E. coli suspension were sequentially added to the cantilever tip and the fundamental mode was monitored. Figure
8.3 is a plot of phase angle versus frequency at the fundamental mode in the two phases of functionalization. From right to left the resonant peaks represent the fundamental mode in air, with agar film, and E. coli inoculation, respectively. The resonant frequency decreased from the value in air when the agar film and E. coli was applied which, as defined by Eq. (3-9), signifies that the effective tip mass of the cantilever increased.
Since the agar and E. coli were added sequentially as suspensions, evaporation took place. As water evaporated from the cantilever tip, the mass decreased and, in turn, the resonant frequency increased. The frequency increase was monitored until the readings stabilized, result not shown. Typically this period of stabilization was between 30 and 60 minutes. The peaks plotted in Figure 8.2 for the agar and E. coli were recorded upon 122 confirming that frequency change, in essence evaporation, had concluded. The addition of the agar film resulted in a 2000 Hz change in resonant frequency, while E. coli addition caused a 100 Hz change. Therefore, the change in resonant frequency due to agar was 20 times greater than that of E. coli which, according to Eq. (3-8), means that the mass of agar on the cantilever tip was 400 times that of the E. coli inoculum.
8.3.3 Active Growth Resonant Frequency Response
In Figure 8.4 the change in resonant frequency with time as E. coli grew on the agar film is shown. Three distinct phases of growth are readily distinguishable. During the first hour, the resonant frequency remained relatively constant suggesting that the E. coli cells were adjusting to the new environmental conditions imposed by the new nutrient source, agar, and the test chamber. The initial period of non-growth is commonly referred to as the lag phase. For the next 4.5 h, the resonant frequency decreased in a relatively rapid fashion signifying an increase in mass upon the cantilever tip due to E. coli growth.
The total resonant frequency change during the 4.5 h was 5.08 kHz (28.86 kHz to 23.78 kHz). During the final stage (6.5 to 15 hours) further frequency change did not occur, suggesting near stoppage of cell growth.
8.3.4 Modeling the Growth Kinetics
E. coli growth is represented by:
dm c = μm (8-1) dt c
where μ is the specific growth rate, mc is the mass of E. coli, and t is the time.
Integrating and rearranging Eq. (8-1) gives:
123
m c μ t o = e (8-2) m c
o o where mc is the mass of E. coli at t = 0; thus, mc = mc + Δmc where Δmc is the mass change of E. coli. Thus Eq. (8-2) can be rewritten as:
Δ m c μ t 1 + o = e (8-3) m c
α Upon applying the agar film to the cantilever, the effective mass, M e , can be expressed by rearranging Eq. (3-2) as:
4 α K(v') M e = 2 2 (8-4a) 4π fα
α where f a is the resonant frequency corresponding to M e . Upon E. coli inoculation and
β after moisture stabilization the effective cantilever mass ( Me ) can be written as:
4 β K (v' ) M e = 2 2 (8-4b) 4π f β
β where f β is the resonant frequency corresponding to M e . It follows that the cantilever’s
γ effective mass due to the E. coli growth, M e , can be expressed as:
4 γ K (v' ) M e = 2 2 (8-4c) 4π f γ
γ where fγ is the resonant frequency corresponding to M e . Therefore, the initial mass of E.
o β α γ β coli, mc is ( M e - M e ), and the change in mass due to the growth, Δmc , is ( M e - M e ).
Δmc Hence, the mass ratio, o , in terms of frequency is: mc 124
Δm ⎛ f 2 ⎞⎛ f 2 − f 2 ⎞ c = ⎜ α ⎟⎜ β γ ⎟ (8-5) o ⎜ 2 ⎟⎜ 2 2 ⎟ mc ⎝ f γ ⎠⎝ fα − f β ⎠
Substituting Eq. (8-5) into Eq. (8-3) and rearranging one obtains:
⎡ ⎛ f 2 ⎞⎛ f 2 − f 2 ⎞⎤ ln⎢1 + ⎜ α ⎟⎜ β γ ⎟⎥ = μt (8-6) ⎜ f 2 ⎟⎜ f 2 − f 2 ⎟ ⎣⎢ ⎝ γ ⎠⎝ α β ⎠⎦⎥
The above equation suggests that a semi-log plot of the left hand side expression versus time will yield a straight line whose slope is the specific growth rate μ . Such a plot is presented in Figure 8.5 using the data shown in Figure 8.4. Here, the three phases of growth are distinguishable. The first phase, lasting one hour, is the lag phase and is characterized by negligible change in the semi-log frequency ratio. The second phase, lasting about 4.5 hours, is the log phase in which two distinct growth rates were observed.
The initial exponential growth rate was calculated to be 1.30 ± 0.05 h-1, while the late exponential growth rate was 0.55 ± 0.05 h-1as the culture entered the stationary phase.
The decrease in growth rate of the E. coli is primarily due to nutrient deprivation within the agar as the E. coli consumed it for growth188.
As was previously noted, the chamber in which the cantilever tip was enclosed was maintained at 29˚C. The exponential growth rate determined (1.30 ± 0.05 h-1) agrees well with results previously obtained in our laboratory of 1.28 ± 0.02 h-1189. In brief, E. coli JM101 was inoculated with the media described in section 8.3.2 in a 16-well microplate and kept in an incubator at 22, 32 and 37 oC. Samples were removed every thirty minutes, stained with 0.2% trypan blue and counted using heamocytometer. The results from cantilever experiments differ from the homogeneous culture experiments 125 only by 2.3%, which suggests that the resonant frequency change during the experiment was in response to E. coli growth.
Recently Fujikawa et al. (2004) developed a logistic model for E. coli growth and applied it to experimental data collected over a wide range of temperatures190. They reported the specific growth rate at 29˚C to be 1.5 h-1 which is in good agreement with the value obtained utilizing the cantilever. The difference in the growth rate values is likely due to the fact that Fujikawa et al. grew E. coli in a liquid nutrient broth whereas an agar film was used on the cantilever. Nutrient availability is inherently lower in agar, thus, the discrepancy in the values is expected.
8.3.5 Stoichiometric and Sensitivity Analyses
Method for determining both experimentally and theoretically mass change sensitivity of PEMC sensor was described in Chapter 6. Typically we find that the PEMC design used in this study has a mass change sensitivity of 50 pg/Hz under liquid immersed conditions. The sensitivity in air is higher. The mass change sensitivity of the particular cantilever used in growth experiment was established as 27 pg/Hz. Thus, the
100 Hz change due to inoculation of E. coli is estimated as 3,000 bacterial cells, as one bacterium is weighs approximately one picogram. Since the total frequency change due to growth was 5.08 kHz, the mass change on the cantilever tip was 0.15 μg, which corresponds to approximately 1.5 x 105 cells. The mass of agar, source of nutrients for the cells, was 0.30 μg which is from 2 μL of 0.15 g/mL concentration agar solution. That is, cell yield based on agarose place on the cantilever was 0.5 g/g. Cell yield on several carbon sources have been reported by several authors191,192, and it is typically 0.5 to 0.55 g/g on sugars. The yield obtained in this study is consistent with existing data reported in 126 the literature.
During growth, the cells uptake agarose and other nutrients from agar, and uptake oxygen from the atmosphere, and produce new cellular material while giving off carbon dioxide and water. Under aerobic conditions (this study), respiratory quotient is unity.
That is, moles of CO2 produced are equal to the moles of oxygen consumed. Because of this exchange, one needs to measure at least one more variable (example, CO2 evolved or agarose consumed) to establish from a fundamental perspective what the exact mass change is. This was not done in our investigation. However, mass increase due to growth was established experimentally by both Gfeller et al.120 on microcantilevers and by Chen et al.193 on QCM. In both cases the E. coli cells were grown on agar under constant temperature and humidity environment. In the case of Gfeller’s study, they found a frequency decrease of about 900 Hz, and their cantilever sensitivity was 140 pg/Hz and thus the total change in the number of cell was about 125,000 over a period of five hours. Their report did not give initial cell number as they used a dip method rather than dispensing a fixed number of cells onto the cantilever. The growth rate estimated from their graphical data suggests that the cells doubled every 90 minutes, which is equal to a specific growth rate of 0.46 h-1 at 37 °C.
Quartz Crystal Microbalance manufacturers (SRS, Inc, Sunnyvale, CA) report mass change sensitivity as 17 ng/cm2-Hz for a 5 MHz crystal. Given that the sensing area is of the order of 1 cm2, the sensitivity is about 17 ng/Hz. The PEMC used in this study has a mass sensitivity of about three orders of magnitude higher, at 27 pg/Hz. Thus, the
PEMC is a much more sensitive platform for measuring growth of cells. That is, PEMC requires a much shorter period of time to confirm the presence of the bacteria than a 127
QCM. For example, the noise level in our measurement was in the order of one Hz, and thus a change of 20 Hz would represent a signal confirming growth. This change based on the mass sensitivity of the cantilever sensor would be approximately 20 cells. If we have an inoculum of 20 cells, and the doubling time (typical for E coli) of the pathogen is
20 minutes, it will require just 20 minutes to establish the growth for detection purposes.
On the other hand, for a QCM, one would need (for the same SNR) a growth equivalent of 340,000 cells which would be formed in 3.2 hours. That is PEMC is approximately 10 times quicker in establishing growth.
This PEMC technology is most useful in detecting low levels of pathogen through selective growth. Although no selective growth was attempted in this study, the potential application of the PEMC sensor for such situations is worthy of further study, and is the focus of the author’s lab.
8.4 Conclusion
The real-time detection of E. coli growth on a millimeter sized PZT-glass cantilever at 29˚C was measured and compared with previously reported results. A model was developed to calculate specific growth rates from resonance frequency data. From the aforementioned model the specific growth rate (μ) of E. coli was in good agreement
(2.3% error) with the growth rate obtained for the same culture using actual counting of cells in our laboratory. This verifies that the observed resonant frequency change was, in fact, due to E. coli growth. 128
Table 8.1: Physical dimensions of the piezoelectric-excited millimeter-sized glass cantilever free-end.
Dimensions PZT Glass
L (mm) 2±0.05 3±0.05
w (mm) 1±0.05 1±0.05
t (μm) 127±5 160±5 129
Silicone tube
Fermentation Chamber
Cells Heating jacket
Silicone tube
Anti-vibration table O’ ring
Figure 8.1: Two dimensional side view schematic of experimental setup. The E. coli inoculated cantilever is mounted inside a temperature controlled chamber (at 29 oC) and the fundamental frequency was monitored periodically as the bacteria grew. The fermentation chamber reached a constant humidity due to limited exchange of air with the surroundings. 130
-70
-75
-80
Phase angle [degree] -85
-90 0 20406080100 Frequency [kHz]
Figure 8.2: Resonant spectrum of phase angle versus frequency of the cantilever in air. The fundamental resonance occurred at 31.08 ± 0.01 kHz. Resonance was followed by a sharp change in the phase angle. The cantilever was excitated with 100 mV. 131
-70
-74
E.coli Agar Air -78
-82 Phase angle [degrees] Phase angle -86
-90 27 28 29 30 31 32 33 Excitation frequency [kHz]
Figure 8.3: Plot of phase angle versus the fundamental mode frequency of the cantilever in air, with agar film, and initial E. coli inoculation, respectively. The resonant frequency decreased from 31.08±0.01 kHz in air to 29.08±0.01 kHz with agar film, and subsequently to 28.98±0.01 kHz with the initial E. coli inoculation. 132
30000
29000 Control (Agar; no inoculum)
28000
27000
26000 Resonant frequency [Hz] frequency Resonant 25000 E. coli on Agar film
24000
23000 0246810121416 Time [h]
Figure 8.4: Resonant frequency response of the fundamental mode to the growth of E. coli JM101. Initially the resonant frequency remained constant then, decreased after the first hour. The total change in frequency was approximately 5000 Hz. After the first six hours the frequency stabilized. These three phases of frequency response are attributed to the lag, exponential, and stationary phase during the E. coli growth. Control shown above is the cantilever prepared in the same fashion as in the growth experiment, except for inoculating E. coli cells. 133
Chapter 9: Piezoelectric-Excited Millimeter-Sized Cantilever (PEMC) Sensors Detect Bacillus anthracis at 300 spores/mL
9.1 Introduction
Bacillus anthracis (BA), a spore forming rod-shaped Gram positive and non-motile bacterium, is the etiology agent of anthrax. Under favorable growth conditions, the bacteria exist as vegetative cells. If the growth conditions deteriorate (extreme temperatures and/or nutrient deprived environment) the vegetative cells sporulate: forming intracellular endospores194-196. As the vegetative cells die the endospores are released as spores. Spores are biologically dormant structures that are highly resistive to extreme temperatures, physical damage, desiccation, and harsh chemicals. These properties allow the bacterial spores to survive for years in soil. Spores remain dormant until they encounter an ideal growth environment in which they germinate into their vegetative state. Spores are also airborne and cause respiratory infection, as was seen in the bioterrorism anthrax spores mailed in the United States of America in the Fall of
2001. There are three forms of human anthrax known to date: gastrointestinal, cutaneous, and inhalation anthrax. Bacillus anthracis spores can enter the body by ingestion, through the skin, and by inhalation. The inhalation anthrax is the most severe one because 99% casualties occurred in individuals who were not treated before symptoms developed197.
The threat of anthrax causing Bacillus anthracis spores as a bioterrorism agent has created an urgent need for a rapid real-time, highly selective and sensitive technique to detect the presence of anthrax spores. In response to the anthrax threat, various detection 134 techniques capable of providing reliable identification of anthrax spores are currently under development. These detectors include, evanescent wave fiber-optic biosensors198,
Love-wave biosensors199, real-time PCR200-205, phage display206, membrane-based on-line optical analysis systems207, electrostatic precipitator208, electrochemiluminescence209,210, and quartz crystal microbalance211. When confronted with the requirement of low concentration detection, the traditional approach involves the growing of the micro- organism on selective media for at least 24 h, followed by morphological and biochemical analysis212,213. The 24-hour incubation time is far too limiting, particularly in the context of public safety. Hence, there is a need for a simple and inexpensive method for the detection of Bacillus anthracis spores in real time. Furthermore, the United State
Postal Service (USPS) has required the development of a rapid detection method, which is cost effective, for the identification of bioterrorism threat agents214. In a similar manner the Department of Transportation (DOT) requires a detection system that will identify bio-terrorism agents within 20 minutes of exposure.
The objective of this Chapter is to explore the application of the piezoelectric-excited millimeter-sized cantilever (PEMC) sensors for the detection of anthrax-causing Bacillus anthracis spores at a concentration as low as 300 spores/mL in the liquid phase. The development of biosensors has been significantly enhanced over the past decade by biomedical101 and chemical applications90. However, the growing interest of millimeter- sized cantilever biosensors for biological detection is due to its high performance characteristics: high sensitivity, short response time, robustness, selectivity, resonance stability, and surface regeneration capabilities. It is worth noting that a complete biosensor system for the control and prevention of a bioterrorism attack requires a three- 135 step process: a detection step, an identification step, and finally a communication step215.
Here, we employ the PEMC sensor to the first step (detection).
9.2 Materials and Methods
Bacillus anthracis spores, the Sterne strain 7702, and Protein A purified Rabbit polyclonal antibody (rPC) in PBS was provided by Professor Richard Rest (Drexel
University College of Medicine). An anthrax combo Goat polyclonal antibody (gPC;
Catalog designation: AB8301) raised against a gamma inactivated spore mixture of
Ames, Sterne, and Vollum strains was purchased from Chemicon International
(Temecula, CA). All other chemical reagents were from Sigma-Aldrich.
The sensing glass surface was thoroughly cleaned using a combination of acids and bases. Then, the sensor surface was aminated and subsequently functionalized with antibody via EDC/sulfo-NHS chemistry. After each detection experiment, the sensor surface was renewed as per the protocol discussed in Chapter 4. That is, the sensor was immobilized with fresh antibody. Spore samples in concentrations ranging from 3 x 104 to 300 spores/mL were prepared in phosphate buffered saline (PBS) (10 mM, pH 7.4) by serial dilution of the master sample (3 x 106 spores/mL). The selectivity of the PEMC sensors was investigated using samples prepared by mixing Bacillus anthracis spores and
Bacillus thuringiensis (BT) spores. Bacillus thuringiensis spore powder was purchased from EDVOTEK (West Bethesda, MD) and a stock solution of 1.5 x 109 spores/mL was prepared in PBS solution, pH 7.4. The mixed samples were prepared in various volume ratios stock solutions of Bacillus anthracis (3 x 106/mL) and Bacillus thuringiensis spores
(1.5 x 109/mL) to yield spore concentrations labeled A, B, C, D and E whose BA concentration was 100%, 0.79%, 0.40%, 0.20% and 0 % with total spore count as given 136 in Table 9.2. The detection experiments were carried out in a chamber at a temperature of
22 ± 0.5 oC. During a typical 1 hour experiment, the liquid level in the sample container
(1 mL) decreased by 4.5 microns per minute. The linear change in liquid level with time resulted in a linear change in resonant frequency. Thus immediately before and after a detection experiment, the rate of change of resonant frequency due to liquid level change was measured and an average value of the two was used to correct the peak position in the detection experiment. One mL of the pathogen containing solution was pipetted into a
1 mL sample container. The functionalized sensing tip of the cantilever was immersed to a depth of 1.0 mm using a XYZ-positioner (Optosigma Corporation, Santa Ana, CA) installed on a vibration-free table. The change in resonant frequency resulting from the mass change was monitored by measuring phase angle and impedance using an impedance analyzer (HP4192A) with an excitation of 100 mV.
9.3 Results and Discussion
9.3.1 Resonance characterization of PEMC sensors
Three cantilevers were fabricated and used in the Bacillus anthracis spores detection experiments. For brevity and comparison of results of the various spore concentration results, only the data from one cantilever is presented here. Each experiment was repeated at least twice and the data shown are typical of the results obtained. The cantilever sensor used in detecting different sample concentrations of Bacillus anthracis spores is labeled PEMCa and the sensor used in the selectivity studies is labeled PEMCb.
The only difference in dimensions was the thickness of the adhesive layer. The thicknesses of the adhesive layer were 10 ± 1 and 25 ± 1 μm for PEMCa and PEMCb, respectively. Physical dimensions of PEMCa are given in Table 9.1. The resonance 137 spectra, phase angle versus frequency of PEMCa and PEMCb in air showed a fundamental resonant frequency of 14 ± 0.05 kHz and 12 ± 0.05 kHz, respectively.
Example spectra may be found in Chapter 8. The second mode resonant frequency of
PEMCa and PEMCb were respectively, at 62 ± 0.05 kHz and 58.5 ± 0.05 kHz. These spectra showed a base line phase angle of –89.4 degrees. Several repeated experiments showed that these resonant frequencies are stable within ± 50 Hz. Since the only difference in the sensors construction was the thickness of the adhesive layer the lower resonant frequencies observed in PEMCb was probably due to the higher mass of the adhesive layer, as resonant frequency is inversely proportional to the square root of the cantilever’s effective mass. In any one experiment, the fluctuation of the steady state resonant frequency was at most ± 10 Hz. In this study, the second mode was selected for in-liquid detection experiments because the sensitivity of a resonant mode is directly proportional to the resonant frequency and therefore, the second mode is of a higher sensitivity.
9.3.2 Selection of Antibody for the detection of Bacillus anthracis spores
In order to select a suitable antibody for BA detection, several experiments were carried out under the same conditions using the two polyclonal antibodies, gPC and rPC.
PEMC sensors were functionalized with APTES as previously described and then exposed to 10 μg/mL antibody solution at 22 oC in a batch measurement cell. The results shown in Figure 9.1A is typical of the second flexural mode resonant frequency change for the amide bond formation between the aminated sensor surface and the activated carboxylic groups on the antibody. The resonant frequency decreased exponentially, indicating rapid initial reaction and reaches a steady state value suggesting an equilibrium 138 surface coverage. Also, the steady state resonant frequency change was greater for the rPC (780 Hz) compared to gPC (350 Hz), suggesting that a larger mass of rPC was immobilized. In Figure 9.1B, the results of resonant frequency response to the binding of spores (3,000 spores/mL) to the sensor surfaces are shown. It can be seen that the equilibrium resonant frequency change obtained with the rPC antibody is 510 Hz compared to 165 Hz obtained with gPC antibody. Similar experiments conducted with
30,000 spores/mL (data not shown) resulted in 1040 Hz change with rPC compared with
510 Hz with gPC. Greater spore attachment with larger amount of antibody immobilized on sensor surface is an expected result. The reason why Rabbit polyclonal antibody was immobilized to a greater extent compared to gPC was not further investigated. We assumed that the concentration of the antibody provided by the suppliers were as given; no antibody assays were carried out. Since a larger response was obtained with rPC antibody, we used it in all further experiments reported in this paper.
9.3.3 PEMC Sensor Response to Spores of various concentrations
Figure 9.2A shows the time profile of the second mode resonant frequency change of
PEMCa antibody-functionalized sensor in the detection of Bacillus anthracis spores at concentrations of 3 x 102, 3 x 103, 3 x 104 and 3 x 106 spores/mL. In all cases, the response showed a rapid decrease during the first 5 to 15 minutes followed by a slower change reaching a constant resonant frequency. For the highest spore sample (3 x 106 spores/mL), the rate of decrease was more rapid compared to the lower concentration
(300 spores/mL) sample. This is an expected response, as the binding rate is proportional to concentration. The total change was 2696 ± 6 Hz (n=2) for 3 x 106/mL sample and 92
± 7 Hz (n=3) for the 300 spores/mL sample. Several control experiments with PEMCa 139 antibody-functionalized sensor was conducted in PBS solution for 1 h at the same experimental conditions as was used in the detection experiments. The response included in Figure 9.2, showed fluctuations in the steady state resonant frequency of ± 5 Hz, which is also the sweep frequency step size. That is, the fluctuations are likely to be even lower than this value. When one compares this with the response of 300 spores/mL sample, it is clear that steady state response to noise in measurement was in the order of 20 for the low spore count sample. For higher spore concentration samples, the steady state change in resonant frequency was significantly higher than measurement errors in resonant frequency. The observed resonant frequency fluctuation of ± 5 Hz for the final steady state value indicates that the cantilever resonance characteristic is quite stable under liquid immersion.
One notes that in Figure 9.1 the total change in the resonant frequency for the antibody immobilization was greater than the steady state resonant frequency change obtained for the corresponding spore attachment. One possible explanation is that the antibody forms a uniform layer on the sensor surface due to its smaller size compared to spores. Under liquid immersion conditions, liquid is readily entrained within the proteinous layer contributing to the oscillating mass. On the other hand, spore attachment was not dense, and was scattered and distributed. Such a low density surface coverage is not conducing to entrapping liquid mass, and thus a lower frequency change was observed. This interesting aspect of changes in the interfacial characteristics is currently under study, and is not within the scope the current chapter.
The mass change sensitivity of the sensor was determined using the silicone oil dip touch technique that we reported in Chapter 5. Briefly, known mass of silicone oil was 140 placed on the cantilever sensing-tip in air and the mass changes when plotted against the corresponding resonant frequency change of the second mode yielded a straight line whose slope is the mass change sensitivity in air, expressed in g/Hz. The mass change sensitivity in liquid was then calculated from the sensitivity in air. For the PEMCa sensor the mass change sensitivity under liquid immersion was determined as 1.56 x 10-11 g/Hz.
Using the mass change sensitivity, the mass of Bacillus anthracis spore bound to the sensor surface after one hour from the lowest to the highest spore concentrations were, respectively, 1.2 ± 0.05ng, 7.8 ± 0.10 ng, 16.2 ± 0.10 ng and 42.1 ± 1.50 ng.
The sensing requirement for bioterrorism agents suggested by several US Federal
Agencies (for example, Homeland Security Advanced Research Projects Agency) is 20 minutes detection time after exposure to bioterrorism agents. Hence it is useful to examine the results of our experiments in the context of this time frame. In Figure 9.2B, a plot of resonant frequency change following 20 minute exposure of sensor versus the log of spore concentration. Resonant frequency change increases in a nearly linear fashion with spore concentration, in the concentration range investigated. These results suggest that calibration relationships for estimating spore concentration can be stated as:
(−Δf ) + A log(C ) = 20 min (9-1) b0 B where A is y-intercept and B is the slope of the line Fig. 9.2B. The parameters A and B will depend on cantilever dimensions, antibody type, and immobilization method.
In order to obtain visual data on spore surface density on the sensor surface, antibody functionalized glass was exposed to spore sample 3 x 106 spores/mL for 1 h, then rinsed in deionized water three times (to remove residual salt) and dried at ~ 50 oC for 4 h and then examined under a Scanning Electron Microscope (SEM). About ten fields of 20 to 141
1280 μm2 area were examined, and the SEM photomicrograph shown given in Figure
9.3A is typical of the high surface density region. Spore surface density was not uniform when viewed at 20 μm2 area at 25,000X, but appeared uniform at lower resolution in the high surface density region, an area of 33 μm2 viewed at 15,000X. In none of the samples we examined, close packing of cells on sensor surface was observed. It is estimated that sensor surface covered by spores was always less than 40 %.
9.3.4 Selective binding of Bacillus anthracis spores to antibody functionalized PEMC sensor
The sensor selectivity to Bacillus anthracis (BA) spore was investigated by examining the response to mixed samples containing another spore, Bacillus thuringiensis (BT) spores. The detection experiments were carried out with PEMCb sensor since PEMCa sensing tip fractured during one of the cleaning steps. Figure 9.4 shows the second flexural mode resonant frequency responses to the various mixed samples. The 100% BA spore sample (3 x 106/mL; Sample A) gave the highest frequency change (2360 Hz; n=1) in 1 h. For the same sample, the PEMCa sensor gave a similar frequency response of 2696 ± 6 Hz (n=2). The closeness of the response of the two
PEMC sensors is due their similar resonance characteristics. As more BT spores were added to the pure BA spore sample (Samples B, C and D) the resonance frequency reached a lower steady state value: 1980 ± 10, 1310 ± 10, and 670 ± 10 Hz for the
Samples B, C and D, respectively. The fluctuation in resonant frequency of the final steady state value was ±10 Hz. It is to be noted that the pure BT spore sample (1.5 x
109/mL) gave only a 10 Hz change over a 1 h period indicating that non-specific binding, if any, was weak and was insignificantly small. Sample D contained 1.5 x 106 BA 142 spores/mL and therefore, one would expect a higher resonant frequency response in comparison to the pure Bacillus anthracis spore sample of 3 x 104 BA spores/mL (Figure
9.2A; using PEMCa sensor). However, the opposite result was observed. That is, the pure
BA sample (3 x 104 BA spores/mL) gave a much higher overall resonant frequency change (1030 Hz) than the Sample D containing 7.5 x 108 /mL BT spores (670 ± 10Hz).
These results suggest that the presence of large amount of BT spores hindered attachment of BA spores, and also reduced binding capacity. We hypothesize that BT in these experiments “crowd” the sensor surface due its very high relative concentration, which in effect reduces availability of binding sites for BA. Current work in our laboratory attempts to develop a better understanding of this phenomenon, and will be the focus of a future publication. From the results presented above, it is reasonable to conclude that the antibody functionalized PEMC sensors are selective to the Bacillus anthracis spores, but at a compromised sensitivity due to the presence of high BT spores.
9.3.5 Kinetics of antibody and spore binding
The adsorption or binding kinetics of microorganisms to solid surfaces is not a well studied topic in comparison to molecular adsorption and binding kinetics on solid substrate. There is a great body of literature dealing with molecular absorption. During the past decade researchers have characterized adsorption kinetics of alkanethiol to gold surfaces, and found it to follow the Langmuir models216-218. Although this is an adsorption process, the bond energy is high and is generally considered to be equivalent to a covalent bond formation. Since attachment of BA spores to sensor surface exhibits an exponential character, similar to alkanethiol to gold surface, it is reasonable to consider the Langmuir model to characterize the associated kinetics. The kinetic 143 parameters so derived will be useful in design of sensors, and evaluating various sensor configurations (for example, flow cell design), and possibly in the evaluation of one antibody immobilization technique versus another. In the following section we develop a general treatment of sensor response to obtain kinetic parameters.
At time = 0, there are no concentration gradients, and thus diffusion effects are absent. Furthermore, the bulk spore concentration is known accurately. Therefore, we consider here an initial rate analysis by fitting the antibody and spore frequency response data to Langmuir model. The Langmuir model can be expressed as (Yan et al., 2002):218
− k τ θ = 1 − e obs (9-2) where θ is the fraction of available reactive sites that are occupied ()0 ≤ θ ≤ 1 in time τ,
kobs is the observed binding rate constant, which depends possibly on bulk concentration
of the binding entity (Cb0 ); antibody or spores. Because the measured resonant frequency changes are a direct measure of antibody or spore attachment, the fractional coverage can be written in terms of frequency change as:
− k obs τ ()(Δf = Δf ∞ )(1 − e ) (9-3)
where ()Δf is the change in resonant frequency at time, τ , and (Δf ∞ ) is the steady state resonant frequency change. Eq.(9-3) can be rearranged to:
⎛ ()()Δf − Δf ⎞ ⎜ ∞ ⎟ ln⎜ ⎟ = −k obsτ (9-4) ⎝ ()Δf ∞ ⎠
The above suggests that the characteristic rate constant kobs during initial time (far from equilibrium) can be determined from a plot of the left hand side (LHS) versusτ of Eq.
(9-4) where we include only data obtained during the first few minutes. Then, a plot of 144
kobs versus Cb0 can be prepared to examine the dependence on Cb0 . In earlier studies an
adsorption on gold surfaces, kobs has been reported to depend on adsorbing species concentration216.
9.3.5.1 Kinetics of antibody reaction on aminated PEMC sensor surface
Since the reaction of antibody to the aminated sensor surface is through covalent bond formation, and experimentally desorption of the bound antibody was not observed in the frequency response, it may be assumed that antibody binding was irreversible. As a
result, kobs in Eq. (9-4) can be determined from the data presented in Figure 9.1. It is clear from Figure 9.1A that the observed binding rate of gPC was significantly slower than rPC. Fitting the data to Eq. (9-4) resulted in the observed binding rate of 3.6 x 10-2 min-1 for rPC and 1.8 x 10-2 min-1 for gPC. The antibody rPC appears not only to have attached one-third more rapidly than gPC but also the total amount bound to the sensor surface was three-fold higher (see Figure 9.1).
9.3.5.2 Kinetics of BA binding in pure samples
Fitting of the experimental data presented in Figure 9.2A to Eq. (9-4) gave straight lines with excellent correlation coefficients ranging from 0.96 to 0.99, and is illustrated in
Figure 9.5A. Limiting the initial rate analysis to the first five minutes appears to be a
reasonable approach to obtain kinetic constants. The values of kobs are listed in Table 9-
-1 -1 3. Here, one can see that kobs decreases from 0.105 min at 300 spores/mL to 0.043 min
at three million spores/mL. A plot of kobs versus log(Cb0 ), given in Figure 9.5B suggests
that the observed rate constant is a very weak function of Cb0 . Note that the decrease in
kobs is a factor of two for a spore concentration, Cb0 , change of four orders of magnitude. 145
The results in Figure 9.5B suggest that the overall binding rate may be independent of spore concentration in the concentration range investigated. Further studies are warranted to examine the kinetics of attachment at very low spore concentration.
9.3.5.3 Binding kinetics of BA spores in the presence of Bacillus Thuringiensis spores
When spores other than BA or other particulate matter are present in the sample, we expect the attachment kinetics to be affected. To what extent the binding rate is affected can be determined by examining the binding rate constant values when non-BA spores are present. The experimental data of the mixed samples (BA and BT) were analyzed in the same fashion as the pure BA sample and a summary of the analysis is given in Table
9.3. Again, the characteristic observed rate, kobs , decreases with increasing BA spore concentration in the mixed samples. The kinetic rate analysis suggests that the presence of BT spores slightly hinders BA spores in reaching the sensor surface.
9.3.5.4 Renewal of sensor surface
All the data reported in this paper was with a freshly immobilized antibody surface.
We have done limited experiments to determine the repeat application of the sensor.
Often, the sensing surface after exposure to low pH buffer loses about 10% of its sensitivity with one to two uses. Third re-use was usually with higher loss of activity, 20 to 30%. A systematic study of loss of activity, and methods to improve re-use of sensors are currently under investigation, and will be the focus of a future report from our laboratory.
146
9.4 Conclusion
In this Chapter we have illustrated the rapid detection of Bacillus anthracis (BA) spores under liquid immersed conditions at 300 BA spores/mL within 20 minutes. The selectivity of the antibody functionalized PEMC sensor was determined with BA samples mixed with Bacillus thuringiensis spore (BT). As the amount of BT spore increased in the mixed sample the steady state resonant frequency change decreased. Samples containing pure BT (1.5 x 109/mL) gave a total frequency change of 10 Hz in 1h, which is within the noise level of the sensor. This suggests that the antibody-functionalized sensor was highly selective to the targeted pathogen. The attachment kinetics of BA to the sensor surface was modeled with Langmuir model. The observed binding rate constant ranged from 0.105 to 0.043 min-1 in the spore concentration range of 300 to 3 million per mL. The presence of BT spores hinders slightly attachment of BA spores, and thus attenuates slightly the sensor performance. These results suggest that the PEMC sensor is a viable practical sensor for BA spores in liquid medium, and in presence of BT spores.
147
Table 9.1: Physical dimensions of PEMCa sensor.
Cantilever Dimensions PZT Adhesive Glass
L (mm) 1.5±0.05 1.5±0.05 3.5±0.05 PEMCa w (mm) 1.0±0.05 1.0±0.05 1.0±0.05 t (μm) 127±5 10.0±1 160±5 Where L = length, w = width, and t = thickness.
148
Table 9.2: Composition of mixed spore samples containing Bacillus anthracis and Bacillus thuringiensis.
Sample BA BT Total Ratio % BA spores in Designation Spores/mL Spores/mL Spores /mL BA:BT Sample
A 3 x 106 0 3.0 x 106 1:0 100.00% B 2.4 x 106 3.0 x 108 3.0 x 108 1:0.008 0.79% C 2.0 x 106 5.0 x 108 5.0 x 108 1:0.004 0.40% D 1.5 x 106 7.5 x 108 7.5 x 108 1:0.002 0.20% E 0 1.5 x 109 1.5 x 109 0:1 0.00% 149
Table 9.3: k obs values as a function of Bacillus anthracis (BA) spore concentration in both pure and mixed samples containing BA and BT.
BA Spores BT Spores k obs values in pure k obs values in mixed concentration concentration samples, PEMCa samples, PEMCb
(spores/mL) (spores/mL) (min-1) (min-1 )
3x102 0 0.105±0.03 ------3x103 0 0.102±0.01 ------3x104 0 0.062±0.02 ------1.50x106 7.5 x108 ------0.098±0.01 2.01x106 5.0 x108 ------0.111±0.04 2.40x106 3.0 x108 ------0.107±0.02 3.00x106 0 0.043±0.008 0.067±0.01 150
0 ] -100 Control. PBS -200 -300 -400 Goat polyclonal antibody -500 -600 -700 -800 A Rabbit polyclonal antibody -900 Resonant frequency change change [Hz frequency Resonant -1000 0 1020304050607080 Time [min]
100
] 10 μg/mL rPC - 0 spores/mL 0
-100 3 10 μg/mL gPC-3x10 spores/mL -200
-300
-400 3 10 μg/mL rPC-3x10 spores/mL -500 B Resonant frequency change [Hz change frequency Resonant -600 0 102030405060
Time [min]
Figure 9.1: Panel A: Resonant frequency change as antibody is immobilized on the aminated PEMC sensor surface at 22 oC. The concentration of both antibodies was
10μg/mL and the sample cell volume was 1mL. The Goat polyclonal antibody (gPC) showed lower steady state resonant frequency change. The control was an aminated cantilever immersed in PBS. Panel B: The binding of spores (3,000 spores/mL) to the antibodies on the sensor surface. A large steady state resonant frequency was observed for the rPC antibody-spore binding. The control was an rPC antibody-immobilized cantilever immersed in PBS. 151
500
] Control 0 3x102 -500 3x103
-1000 3x104
-1500
-2000
Resonant frequency change [Hz change frequency Resonant -2500 A 3x106 -3000 0 102030405060 Time [min]
3000
2500
2000 20mins
) 1500 f
(−Δ 1000
500 B
0 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07
Log Cb0, spores/mL
Figure 9.2: Panel A: Resonant frequency shift of the second flexural mode of PEMCa sensor upon binding of Bacillus anthracis spores at various sample concentrations to antibody functionalized cantilever. The results showed that the binding rate strongly depends on concentration. The control was an antibody- immobilized cantilever immersed in PBS. Panel B: The resonant frequency change at 20 minutes as a function of spore concentrations. 152
A
B
Figure 9.3: Scanning electron micrograph of Bacillus anthracis spores. Antibody functionalized cover slip glass pieces, 2 mm x 4 mm size, were exposed to 3x106 spores/mL for 1 h under identical experimental conditions as the detection experiment. The glass surface was triple rinsed with deionized water and dried. Panel A: sample at 25,000X. Panel B: the same sample at a lower magnification, 15,000X. The photomicrograph is typical of the high cell density region. Area of SEM photograph is 1.7 x 10-5 mm2. 153
0 9 100%BT (1.5x10 /mL) ] -500
0.20% BA spores -1000
-1500 0.40% BA spores
0.79% BA spores -2000
-2500 6 100% BA (3x10 /mL) Resonant frequency change [Hz frequency Resonant
-3000 0 102030405060 Time [min]
Figure 9.4: The second flexural mode resonant frequency change of PEMCb sensor, antibody functionalized, for the binding of Bacillus anthracis spores from sample solutions mixed with Bacillus thuringiensis spore. The stock samples of Bacillus anthracis and Bacillus thuringiensis spore concentrations were 3 x 106/mL and 1.5 x 109/mL, respectively. The frequency change in 1 h caused by the pure Bacillus thuringiensis sample (100% BT) was 10 Hz. As the amount of BA spore increased the steady state resonant frequency change increased. 154
0.00
-0.05 A -0.10
) -0.15 ? f
Δ
f)/ -0.20 Δ - ?
f 300 spores/mL Δ -0.25 3.0E03 spores/mL ln(( -0.30 3.0E04 spores/mL -0.35 3.0E06 spores/mL
-0.40 0.0 1.0 2.0 3.0 4.0 5.0
Time [min]
0.16
0.14 B
0.12
) 0.10 -1
0.08
(min obs
k 0.06
0.04
0.02 0.00 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Log10 Cb0
Figure 9.5: Langmuir kinetic analysis for the binding of Bacillus anthracis (BA) spores from pure samples to the PEMCa sensor surface. Panel A: The initial kinetic analysis of the various BA spores concentrations. Correlation coefficient ranged from 0.96 to 0.99.
The slope of each line gives the observed characteristic binding rate kobs. Panel B: A plot of kobs versus the log of the spore bulk concentrations. 155
Chapter 10: Detection and Quantification of Proteins Using Self-Excited PZT- Glass Millimeter-Sized Cantilever
10.1 Introduction
In this Chapter, we investigated the use of PEMC sensor, of millimeter scale dimension, for protein detection and quantification. The size, namely millimeter, is an important distinguishing feature of this work. Protein assays of target analyte were carried out, confirming the ability of the cantilever to detect protein-protein binding. The significance of the results we report is that millimeter sized PZT cantilevers have the sensitivity to detect protein interactions and, therefore, may be useful for monitoring proteins in body fluids.
The detection and quantification of protein is important in many medical and biomedical applications219,220. Protein detection has been investigated by various techniques such as atomic force microscopy (AFM)115, molecular affinity scanning force microscopy (MASFM)116, force amplified biological sensor (FABS)221, ellipsometry222, fluorescence spectroscopy223, acoustic plate mode sensors224, surface plasmon resonance
(SPR)225, and others. However, these techniques do not lend themselves to direct quantification of the target protein. Also, these techniques require sample preparation prior to detection. There are many different assays for measuring proteins; the most commonly used are BCA226, Bradford227, Lowry228, and ELISA229. These assays use a chemical reagent to react specifically with the protein to form colored products that can be measured spectrophotometrically. The fluorescence intensity of the colored sample is then compared to a calibration curve developed from known protein samples, to determine the protein concentration in the unknown samples. The protein purification and 156 quantification step significantly increases the cost and complexity of the process.
In Chapters 6 and 7, we described a means to detect the pathogen Escherichia coli (E. coli) O157:H7 and the detect limit using antibody functionalized lead zirconate titanate- glass (PZT-glass) cantilever sensors. The PEMC sensor we reported has very high mass sensitivity and also has the potential for real-time monitoring of mass changes. Mass change to a cantilever may be cause by adsorption or reaction and can be in the gas or liquid phase. The PZT layer attached to the cantilever’s base acts both as an actuating and a sensing element. In general, detection of a biological molecule requires the immobilization of antibody on the sensor’s surface. Antibodies can be covalently bound to amine groups derivatized on the sensor. When the target of interest binds to the cantilever’s sensing surface, the effective mass of the cantilever increases which alters the cantilever’s resonant frequency. Monitoring the resonant frequency change with time provides quantitative measures of the analyte using the mass change sensitivity of the device. Unlike optical methods that rely upon interaction of target molecules with an evanescent field, the detection phenomenon in this work occurs when the target binds to the cantilever’s sensing glass surface. Furthermore, PEMC sensors can detect target species in the presence of contaminating agents. Other advantages of the PEMC sensors include its low manufacturing cost, small size, and mechanical robustness.
Very recently Lee et al., (2004) reported the use of a PZT monolithic thin film cantilever for detecting C-reactive protein112. They used resonant frequency shift as the primary means of detection. Their results showed that sensitivity in the nanogram per milliliter level can be achieved successfully. Also, Kooser et al., (2003) have reported experiments where they used the piezoresistive microcantilever to study the interaction of 157 anti-bovine serum albumin (a-BSA) with bovine serum albumin (BSA) in solution107.
The deflection of the cantilever after the adsorption of the protein induces a change in the resistance of the piezoresistive channel. The authors observed that the temporal evolution of resistance change is characteristic of adsorption phenomena and is distinctly different from the response obtained due to the presence of the liquid media.
10.2 Materials and Methods
10.2.1 Detection experiments
Affinity purified rabbit IgG (Sigma Aldrich), anti-GAS (ViroStat, Portland, ME), and anti-EC (Kirkegaard and Perry Lab (KPL), Gaithersburg, MD) stock solutions of 1.0 mg/mL were prepared in 10 mM phosphate buffered saline (PBS) solution pH 7.4. A lower concentration (0.1 mg/mL) of the antibodies was also prepared in PBS by serial dilution. One mL of each sample was loaded into a 1 mL container, and the silanized sensing tip of the cantilever was immersed to a depth of 1mm using the XYZ- manipulator. The impedance and phase angle of the PZT-layer at 1 to 100 kHz was monitored and recorded for two hours. Subsequently, the cantilever bearing rabbit IgG was immersed to 1 mm in a separate container containing the antigen, anti-rabbit IgG
(biotin conjugated, Sigma Aldrich), for one hour. Then, the cantilever was immersed in captavidin for one hour to characterize the reaction of captavidin (Sigma Aldrich) to biotin. We deemed the transient to have achieved steady state if the change in resonance frequency remained within 3% or within 10 Hz over a 10 minute period. Natural variation of frequency value at steady state was about 10 Hz. From our experience and experimental observations, these criteria were sufficient for most experiments. After the binding experiments were concluded, the cantilever was immersed in a low pH buffer, 158
HCl/PBS pH 2.2, for 1 hour to characterize the release of captavidin and subsequently anti-rabbit IgG.
10.2.2 Protein quantification assay
The NanoOrange® kit N-6666 (Molecular Probes, OR) was used to quantify protein in the released samples. The NanoOrange® technique allows for accurate detection of proteins in solution at concentrations between 10 ng/mL and 10 μg/mL. Calibration and release samples were prepared as per the manufacturer’s protocol. A fluorometer (Photon
Technology International, PTI) was used to obtain fluorescence spectra from which a standard calibration curve was developed as a plot of intensity (counts/sec) versus protein concentration in ng/mL. The steady state fluorescence intensities of the samples were compared to the calibration curve to determine the protein concentration in each sample.
10.2.3 Systematic measurement correction
All experiments were carried out at room temperature without the sample being covered. This caused a constant change in resonant frequency due to liquid level change
(typically ~ 4.5 μm/min). The systematic error to resonant frequency change caused by the change in liquid level was corrected as previously reported in Chapter 5.
10.3 Results and Discussions
10.3.1 Characterization of PEMC sensor
Several cantilevers were fabricated and used in the various experiments reported in this Chapter. Since the cantilevers were made individually and manually, no two cantilevers had the same resonance characteristics. Thus, we used the same cantilever in all the experiments reported in this study so that suitable comparisons can be made. Each 159 experiment was repeated at least three times and the data shown are typical of the results obtained. The resonance spectra of the cantilever (physical dimensions shown in Table
10.1) in air is shown in Fig.10.1. The first peak in Fig. 10.1 is the fundamental frequency, and the higher modes occur at higher frequencies. Note that only two resonant frequencies (17.5 and 53.5 kHz) were found in the frequency range, 1 to 100 kHz. The numerical results and key performance parameters, Q factor and mass change sensitivity, are given in Table 10.2. It is also interesting to note that the ratio of second mode frequency to the first mode is 3.14. The relationship given for a uniform rectangular cross-section cantilever (Eq. 3-2) predicts a value of 6.27. The composite cantilever used in this study does not have a uniform cross section and thus, the frequency ratio differs due to the two-layer construction.
Mass detection of the macrocantilever is enhanced by a high quality factor (Q factor) of the resonant peak(s). The mechanical Q factor characterizes the sharpness of the peak.
The cantilevers fabricated showed a range of Q factor values in air from 30 to 100. Table
10-2 shows the Q values of the second flexural mode resonant frequency in both air and at an immersion depth of one millimeter in PBS solution. The Q factor decreased from 36 in air to 32 in PBS. The reduction in Q value is due to the viscous damping of the surrounding liquid. This small change in Q is a strong contrast when compared to the performance of silicon cantilevers131. The Q factor of PEMC sneosrs does not deteriorate significantly upon immersion in a liquid medium. If a higher Q value is desired, it can be achieved by using a Q-control suggested by Tamayo et al., (2001)93.
10.3.2 Mass change sensitivity in air and under liquid immersion
As described earlier, the mass change sensitivity of the cantilever in air is determined 160 by the dip-touch silicone oil technique. The resonant frequency changes plotted against mass additions, as shown in Fig 10.2, yields a straight line whose slope is termed mass
change sensitivity ()σ 2a in air of the second flexural mode, expressed in g/Hz. For the
-11 cantilever investigated, the mass change sensitivity (σ 2a ) was 5 x 10 g/Hz. That is, if the resonant frequency resolution is 1 Hz, then mass changes of 50 pg are discernable.
This parameter ()σ 2a numerically is lower for higher cantilever sensitivity. The mass
change sensitivity of the second flexural mode under liquid immersion (σ 2 f ) was
estimated from()σ 2a using Eq. (6-3). In this study the parameters M e , M ef , f n , and f nf
(see Table 10.2 and 10.3) were determined from resonant frequency values obtained
-11 experimentally. The calculated value of σ 2 f was 7.2 x 10 g/Hz.
10.3.3 Second flexural mode response to 1-mm liquid immersion
In Figure 10.3 we compare the second mode resonant frequency of the cantilever in air with the response obtained when it was immersed 1-mm in PBS. Note that both the resonant frequency and the peak height decreased, as was dicussed in Chapter 2 and 4.
The resonant frequency decreased from 53.0 kHz in air to 44.5 kHz in PBS. Resonant frequency is directly proportional to the inverse square root of the cantilever’s mass; therefore, an increase in mass due to the added oscillating liquid mass causes this decrease. Table 10.3 summarizes the parameters calculated for the second flexural mode resonance in air and in PBS. The value of the various parameters is within the range of values we reported earlier for cantilevers used in the detection of the pathogen
Escherichia coli O157:H7, Chapter 6. The value of the intrinsic damping parameter, co, was determined using Eq. (4-1) to be 11.57 kg/m-s. The added effective mass, mae, and 161
-6 the viscous damping parameter, cv, are 4.19x10 kg and 4.64 kg/m-s, respectively.
10.3.4 Selection of flexural mode for detection
The resonant mode of the cantilever used for detection depends on the mass change sensitivity in liquid and the Q factor. The greater the mass change sensitivity, and greater the Q factor, the more suitable is the peak for detection. Since the cantilever exhibits several resonant modes of vibration, selection of a particular mode for sensing is an important consideration. Eq. (6-1) and Eq. (6-2) suggests that the higher the resonant frequency, the more sensitive the cantilever for a given cantilever’s mass. That is, the cantilever is able to detect a smaller mass change per unit change in resonant frequency if a higher resonant mode is selected. Our experience with the fabricated PEMC sensor is that the second flexural resonant mode offers higher mass change sensitivity and, in most cases, a higher Q factor than at the fundamental frequency. Often, the first mode is significantly dampened when the cantilever is immersed in a liquid. That is, the resonant peak (phase angle vs. frequency) considerably reduced in height and broadening in width.
As a result, the apex of the resonant peak appears flat and therefore, introduces false positive responses during detection in liquid. In this study the second mode was selected and used in the experiments.
10.3.5 Binding of antibody to amine-terminal silane sensing glass surface
The reaction between amine nucleophiles on the glass surface and the carbonyl groups of the EDC/Sulfo-NHS-MAb ester form amide linkages. A covalent amide bond is strong and is stable under sensing conditions. However, extreme pH environment or proteases can hydrolyze this bond. Since the conditions used in this study were mild, the amide linkages did not hydrolyze. 162
10.3.5.1 Resonant frequency response to peptide bond formation between anti- EC and amines on the silanylated glass surface
The resonant frequency response of the cantilever to samples containing anti-EC at concentrations of 1.0 mg/mL and 0.1 mg/mL are shown in Fig. 10.4. The EDC/sulfo-
NHS activated anti-EC react with the amine groups on the glass surface, this causes a mass change of the cantilever. In essence, the response shown in Fig. 10.4 represents in real time the peptide bond formation on the cantilever surface. At higher antibody concentration (1 mg/mL), the resonant frequency decreased more rapidly initially and reached a lower constant value. The more rapid rate observed is an expected response, as reaction rate is proportional to concentration. The total change was approximately 1800
Hz. At the lower concentration, 0.1 mg/mL, the change was far slower and steady state was achieved at a lower value, 550 Hz. Using the mass change sensitivity under liquid
-11 immersion (σ 2 f ), 7.2 x 10 g/Hz, the mass of anti-EC that bind during the first two hours for the two concentrations are, respectively, 129 ± 1.30 ng and 39 ± 1.40 ng. The surface area available for binding is 4.5 mm2, and an average antibody has a projected area of 6.2 x 10-12 mm2, thus the maximum mass change that can be attained for complete monolayer coverage of the exposed glass surface is 190 ng. That is, a first order estimate suggests that the cantilever’s surface coverage by antibody was approximately 68% and
21%. It has been suggested in the literature224 that when a protein binds to artificial surfaces their three dimensional structure may change. In this slightly altered state they may occupy a larger projected surface area. Thus, the surface coverage estimates noted above can be thought of as the minimum estimates rather than actual values.
163
10.3.5.2 Resonant frequency response to the reaction of anti-GAS with amines on the silanylated glass surface
Fig. 10.5 shows the frequency response of the second flexural mode for the binding of anti-GAS at concentrations of 0.1 and 1.0 mg/mL to the glass tip bearing amine-terminal silane. Clearly, the largest MAb concentration, 1.0 mg/mL, has both the higher total resonant frequency change and the more rapid initial rate of change. Here, it is worth noting that during the first 60 minutes of the experiment the sample containing 1.0 mg/mL did not reach saturation. However, with a lower concentration sample, 0.1 mg/mL, it reached a constant value (640 Hz) in the first 40 minutes of the reaction. This corresponds to a cantilever mass change of 125 ± 1.20 ng and 46 ± 0.50 ng for the 1.0 mg/mL and 0.1 mg/mL anti-GAS samples, respectively. Note that these results are comparable to the earlier results reported here on anti-EC at the same concentration.
10.3.5.3 Resonant frequency response to protein reaction and interaction
The cantilever response to a sample containing rabbit IgG at a concentration of 0.1 mg/mL is given in Fig. 10.6 labeled as A. The resonant frequency decreases rapidly, and reaches a constant value during the first 18 minutes. Total change is approximately 300
Hz. Using the mass change sensitivity of the cantilever under liquid immersion, the mass of rabbit IgG that binds to the sensor surface is 22 ± 0.15 ng.
After the immobilization of rabbit IgG, the cantilever was exposed to anti-rabbit IgG sample, 0.1 mg/mL biotin conjugated, for one hour. The response to the protein interaction is shown in Fig. 10.6B. Clearly, the resonant frequency decreases immediately upon exposure. A steady change in resonant frequency was reached in the first 15
minutes. From the mass change sensitivity (σ 2 f ), the mass of anti-rabbit IgG attached was 18 ± 0.20 ng. The change in resonant frequency corresponded to the change observed 164 during the rabbit IgG detection experiment, suggesting a fairly good binding of the two proteins.
Once the biotin conjugated anti-rabbit IgG experiment was concluded, the cantilever was immersed in a captavidin (Nitroavidin, pH 7.2) 0.1 mg/mL sample for one hour (see
Fig. 10.6C). The resonant frequency decreased rapidly and reached steady state at approximately 1100 Hz. This suggests a change in resonant frequency of 500 Hz. The sensitivity predicted that the mass change due to captavidin-biotin binding was 36 ± 0.40 ng. The bonding between captavidin and biotin is reversible due to the modification of the binding site by a nitro group230. This decreases the association constant of natural avidin-biotin interactions from 1015 M-1 to 109 M-1, thus, the binding is more considerably reversible230,231.
Captavidin bound to the biotin conjugated anti-rabbit IgG can be released by changing the pH of the medium230. After the binding of captavidin, the cantilever was immersed in a pH 10 sodium carbonate buffer which caused the release of captavidin. As a result, the resonant frequency increased. In Fig. 10.6 during the time period D, the resonant frequency increased indicating the released of captavidin. The total increase of
460 Hz corresponds to captavidin release of 33 ± 1.3 ng. Suggesting that, approximately
92% of the bound captavidin was released.
The anti-rabbit IgG was released by exposing the cantilever to HCl/PBS solution (pH
2.2) after the captavidin release experiment was concluded. During the first 25 minutes the resonant frequency decreased, suggesting an increase in the attached effective mass.
This may have been caused by changes in the protein conformation due to the low pH which potentially can affect the rigidity of the antibody layer on the cantilever. After the 165 first 25 minutes the resonant frequency increased, indicating that the attached mass decreased, due to the unbinding of anti-rabbit IgG. The result (Fig. 10.6E) shows that the total increase was nearly identical to the antibody binding event (in Fig. 11.6B). The significance of these results is that serial binding of proteins and their subsequent release can be measured by the PZT-glass cantilever with good sensitivity. The mass released as
predicted from sensitivity parameter σ 2 f is 21 ± 1.00 ng. Since only 18 ng of the biotinylated anti-rabbit IgG was attached to the cantilever this suggest that the unreleased captavidin was released along with the antibody. It is useful to compare the mass change determined by the cantilever response with assay of the released protein. The results are summarized in Table 10.5. The data show mass change predicted from change in resonant frequency is in good agreement with protein assay carried out.
10.3.6 Kinetic modeling of protein binding to cantilever surface
Since the cantilever sensor continuously measures both protein reaction on its sensing surface and protein binding, it is possible to extract kinetic parameters associated with these processes. The net rate at which protein binds to the cantilever surface depends on the bulk concentration of the protein in solution, and the surface concentration of the reacting species. In this study, typical protein concentrations used in the experiments were 0.1 and 1.0 mg/mL. These sample concentrations correspond to protein molecules in the range of 4 x 1017 and 9 x 1021 molecules per milliliter of sample solution, respectively. The available sensing surface area for protein binding on the cantilever used was 4.5 mm2 and an average antibody has a projected area of 6.2 x 10-12 mm2. Thus, the total number of protein molecules for complete monolayer coverage of the exposed surface is 7 x 1011, which is small in comparison with available protein in solution. 166
0 Thus,C p = C p and hence, the change in resonant frequency with time can be expressed in terms of protein concentration as Eq. (9-4):
⎛ ()()− Δf − − Δf ⎞ ⎜ ∞ ⎟ 0 ln⎜ ⎟ = −k pC pτ (10-1) ⎝ ()− Δf ∞ ⎠
0 The above equation suggests that a plot of the left hand side (LHS) versus C pτ will yield
a straight line whose slope is (− k p ). In every experiment the cantilever was prepared in the same manner. Therefore, we expect near identical surface coverage of the amine- terminal silane (APTES). The theoretically calculated value of the total change in
resonant frequency ()− Δf ∞ for monolayer coverage of protein molecules on the amine- derivatized glass sensor surface is -3800 Hz. The total change in frequency in an
experiment ()− Δf ∞ was determined from the total monolayer mass coverage (see section
8.3.5.1) and the mass change sensitivity in liquid, (σ 2 f ). However, the experimental total change in resonant frequency for the different protein sample concentrations reported in this study is less than the theoretical value (see Table 10.4). That is, only a fractional
surface coverage was achieved. As a result, the steady state resonant frequency (− Δf ∞ ) in each experiment was used to calculate the binding reaction rate constant.
All the experimental data obtained in this study were analyzed using the above model. At low protein concentration (0.1 mg/mL) a single binding rate was seen.
However, at high protein concentration (1.0 mg/mL) two distinct binding rates were observed. A sample result is given in Fig. 10.7 for the reaction of anti-EC to the amine- derivatized cantilever glass surface. Fig. 10.7A shows that the binding rate for 0.1 mg/mL anti-EC samples was constant throughout the experiment conducted for two 167 hours. The binding rate constant was determined to be 1.28 x 10-1 ± 0.001 min-1
(mg/mL)-1. The time profile given in Fig. 10.7B shows two distinct binding rates: slow rate at short time period (30 minutes) and a higher rate at longer times. The binding rate constant at short time period, the initial binding rate constant, depends on the number of available reactive sites. Beyond the first 30 minutes, in both the antibody-amine reaction and the protein-protein binding experiments, the data deviate significantly, but systematically, from the initial rate. This suggests that beyond the initial time period other factors appear to affect the reaction and binding. For example, surface heterogeneity due to non-uniform distribution of antibodies will alter the binding rate constant. The initial binding rate constant (first 30 minutes) was 1.20 x 10-2 ± 0.0001 min-
1 (mg/mL)-1 for the 1.0 mg/mL sample. The quality of the fit was good (r2 = 0.99).
Similar binding kinetics was observed in all experiments where the concentration of the sample was 1.0 mg/mL, irrespective of the antibody. We believe that the apparent higher binding rate of the 1.0 mg/mL antibody samples beyond the 30 minutes may be due to protein aggregation. Aggregation may cause layer complex with a larger number of binding sites. However, it is important to point out that, even though aggregation may be taking place, the higher concentration of antibody during immobilization gives a more sensitive response for low antigen concentration.
Model analysis was done for each protein binding experiment (data not shown) to determine the binding rate constant. The initial binding rate constant of the 1.0 mg/mL anti-GAS sample to the amine-derivatized glass surface was 2.30 x 10-2 ± 0.0002 min-1
(mg/mL)-1. For the 0.1mg/mL anti-GAS sample the binding reaction rate constant was determined to be 3.83 x 10-1 ± 0.004 min-1 (mg/mL)-1. The binding of anti-EC and anti- 168
GAS to the amine derivatized cantilever glass surface showed higher binding rate constants for the 0.1 mg/mL sample. In Table 10.4, we have summarized the binding rate constants for the various protein reactions and binding experiments conducted in this study.
The binding rate constant of 0.1 mg/mL rabbit IgG to the amine functionalized cantilever glass surface was determined to be 1.46 ± 0.02 min-1 (mg/mL)-1. Here, the binding rate constant of 0.1 mg/mL rabbit IgG is significantly higher than both anti-EC and anti-GAS at the same concentration. This suggests that there are structural differences in the antibody molecules that result in different binding rates to the amine groups on the cantilever sensing surface.
The binding rate constant of 0.1 mg/mL anti-rabbit IgG (biotin conjugated) to the rabbit IgG bearing cantilever surface is given in Table 10.4. Note that the rate at which rabbit IgG reacts with the amine groups on the sensor surface is greater than the rate at which anti-rabbit IgG binds to rabbit IgG. This result is not surprising because the antigen-antibody interactions are specific to the binding sites on the antigen whereas the reaction of antibody (rabbit IgG) to the amine groups on the sensor surface is non- specific, because many EDC/sulfo-NHS activated carboxylic sites are present on the IgG surface.
The binding of the 0.1 mg/mL captavidin to biotin (on the anti-rabbit IgG) was also analyzed using the proposed model. The binding rate constant was 2.57 x 10-1 ± 0.003 min-1 (mg/mL)-1, which is significantly slower than the rate of anti-rabbit IgG binding to rabbit IgG (see Table 10.4). Since captavidin is much smaller in size (~63 kD) compared to IgG (~150 kD), its binding rate is expected to be more rapid. This difference may be 169 explained by the location of the binding sites on captavidin molecules. The binding sites are buried beneath the surface of the captavidin, thus requiring a longer time for orientation and binding.
10.4 Conclusion
In this Chapter we have showed that the PZT-glass cantilever with free amine groups immobilized on its surface can be used to detect protein binding and protein-protein interaction under liquid immersion in real-time. The mass change sensitivity of the sensor was in the order of tens of picograms. The experimental results showed that the steady state resonant frequency change depends on the bulk concentration of the protein sample.
That is, the greater the target protein concentration the larger is the total resonant frequency change and the rate of change. The results also indicate that the mass of protein attached and released were within experimental error. This suggests that serial binding of proteins and their subsequent unbinding can be detected with good sensitivity. Resonant frequency change was measured as protein reacted or bound with the sensing glass cantilever surface. Protein concentrations, 0.1 and 1.0 mg/mL, which resulted in nanogram mass change were successfully detected. The mass change sensitivity gave a total mass change of 54 ± 0.45 ng for the binding of anti-rabbit IgG (biotin conjugated) to rabbit IgG immobilized cantilever and the subsequent binding of captavidin. The unbinding of anti-rabbit IgG and captavidin gave a total mass change of 54 ± 1.70 ng.
Fluorescence based assays showed the combined mass of both proteins in the released samples as 54 ± 2.24 ng. The binding kinetics of the model proteins is modeled as first order. The initial binding rate constant of anti-rabbit IgG to rabbit IgG was 1.36 ± 0.02 min-1 (mg/mL)-1. The initial binding rate constant of captavidin to biotinylated anti-rabbit 170
IgG was 0.257 ± 0.003 min-1 (mg/mL)-1. The significance of the results we report here is that millimeter sized PZT actuated glass cantilevers have the sensitivity to measure in real-time protein-protein binding, and the binding rate constant. Current work is in progress to detect protein concentrations at the nanogram level in a flow cell arrangement.
171
Table 10.1: Dimensions of the PEMC sensor.
Dimensions PZT Glass
L (mm) 1±0.05 3±0.05
W (mm) 2±0.05 2±0.05
t (μm) 127±5 160±5 172
Table10.2: Resonant peaks characteristics of PEMC sensor.
Mode Resonant Q factor Sensitivity Resonant Qf factor Sensitivity
frequency in air σ 2a , frequency [kHz], 1mm in σ 2 f , [kHz] in air 1012[g/Hz] 1mm in PBS PBS 1012[g/Hz]
1 17.5±0.01 25±2 ---- 10.5±0.01 ------
2 53.0±0.01 38±2 50 44.5±0.01 30±2 72
173
Table 10.3: Calculated values of the cantilever effective mass, spring constant, intrinsic damping, added mass and viscous damping in PBS solution at 1 mm immersion depth for the second flexural mode resonant frequency.
Effective Spring Intrinsic damping Added mass, mae, Viscous damping
mass Me constant, K parameter, co 1mm immersion parameter, cv [kg/m-s] at 106x(kg) 10-3 x(N/m) (kg/m-s) in PBS 107x(kg) 1mm immersion in PBS [Eq.2] [Eq.1] [Eq.5] [Eq.4] [Eq.5]
2.02 5.99 11.57 4.19 4.64 174
Table 10.4: Data showing experimental predictions of mass attached and model prediction of the binding reaction rate constants. Experimental errors were within ±1% of the reported values.
Immobilized Target Concentration σ 2 f estimates of Binding reaction
surface [mg/mL] binding mass [ng] rate constant, kp, [min-1(mg/mL)-1] 0.1 46 0.383 aminated surface Anti-GAS 1.0* 125 0.023
0.1 39 0.128 aminated surface Anti-EC 1.0* 129 0.012
aminated surface rabbit IgG 0.1 22 1.456 anti-rabbit 1.362 rabbit IgG 0.1 18 IgG Biotinylated 0.257 captavidin 0.1 36 anti-rabbit IgG * represent the antibody concentration that showed two distinct binding rates. The reported value of the binding rate constants is the initial rate. 175
Table 10.5: Data showing experimental and protein assay results of the mass of anti-rabbit IgG and captavidin that binds and unbinds during one hour.
Samples Attached mass (1hr) Released mass (1hr) Average released determined from determined from mass (1hr) by protein cantilever response cantilever response assay [ng]
(σ 2 f )()− Δf [ng] (σ 2 f )(− Δf ) [ng]
Anti-rabbit 18±0.20 21±1.00 30±2.00 IgG
captavidin 36±0.40 33±1.30 24±1.00
Total mass 54±0.45 54±1.70 54±2.24
176
-70
-75
-80
Phase Angle [degree] Phase -85
-90 0 20 40 60 80 100
Frequency [kHz]
Figure 10.1: Resonant spectrum of phase angle versus frequency of the cantilever in air. At resonance, the phase angle of the oscillating cantilever (100 mV excitation) exhibits a sharp peak.
177
2.0E-07
1.6E-07 Slope = 5x10-11 g/Hz
1.2E-07
8.0E-08 Mass Change [g] Change Mass 4.0E-08
0.0E+00 0 1000 2000 3000 4000 Resonance frequency change (-Δf) [Hz]
Figure 10.2: In air measurement of mass change sensitivity. Point masses of silicone oil (50 cp) were dispensed at the cantilever tip by a tapered fiber tip 20 microns in diameter. The resonant frequency change (±10 Hz) with mass additions (±1.5 ng) gives experimental measures of mass sensitivity. 178
-74 In Air
-78
In PBS -82
-86 Phase Angle [degree]
-90 40 45 50 55 60 Frequency [kHz]
Figure 10.3: Second flexural mode resonance peaks in air (right) and in PBS solution (left) immersed 1 mm. The liquid adhering to the cantilever sensing surface became part of its effective mass. As a result its resonant frequency decreased.
179
0
0.1 mg/mL -400
-800
-1200 1.0 mg/mL -1600
Resonant frequency change [Hz] -2000 0 20 40 60 80 100 120 Time [min]
Figure 10.4: Binding of antibody to E. coli O157:H7 at various sample concentrations. The reaction of EDC/sullfo-NHS activated antibody reacts with amine groups on the glass surface to cause mass change of the cantilever. . 180
] 100
-200 0.1mg/mL -500
-800 -1100
-1400 1.0mg/mL -1700
Resonant frequency change [Hz change frequency Resonant -2000 0 20 40 60 80 100 120
Time [min]
Figure 10.5 : Binding of antibody to Group A Streptococcus at various sample concentrations. The reaction of EDC/sullfo-NHS activated antibody with amine groups on the glass surface causes a decrease in the second mode resonant frequency. The peptide bond formation increases the effective mass of the cantilever. 181
0
-200
-400
-600
-800
-1000 ABC D E Resonant Frequency Shift [Hz] Shift Frequency Resonant -1200 0 40 80 120 160 200 240 280 320 360 400 440 Time [min]
Figure 10.6: Resonant frequency change versus time for the sequential binding of proteins. Time period A: the reaction of rabbit IgG to amine groups on the silane- derivatized glass surface for 2hrs. The formation of peptide bonds increases the effective mass of the cantilever causing the decrease of its second flexural mode resonant frequency. Time period B: binding of biotin conjugate anti-rabbit IgG to the cantilever bearing rabbit IgG for 1hr. Time period C: the subsequent binding of captavidin to biotin on the cantilever sensing surface for 1hr. Time period D: the release of captavidin (1hr) upon cantilever exposure to pH 10.24 sodium carbonate buffer. Time period E: the release of anti-rabbit IgG (1hr) by immersing the cantilever one-millimeter in HCl/PBS buffer (pH 2.0). Note that the rise in resonant frequency due to the release of the protein entities corresponded to the decrease observed during the binding experiments. This suggests that the mass attached was removed effectively. Criterion for steady state was if the change in resonance frequency remained within 3% or within 10 Hz over a 10 minute period. Natural variation of frequency value at steady state was about 10 Hz. 182
0.0 -0.2 A ]
∞ -0.4 f Δ
)/ -0.6 f Δ
- -0.8 ∞
f -1.0 Δ -1.2 ln[( -1.4 -1.6 0 2 4 6 8 10 12
0 C pτ[min(mg mL)]
0.0
-0.5 B ] ∞
f -1.0 Δ )/
f -1.5 Δ - ∞ f -2.0 Δ
ln[( -2.5
-3.0 0 20 40 60 80 100 120
0 C pτ[min(mg mL)]
Figure 10.7: Natural log plot of the ratio of resonant frequency change to the total resonant frequency change versus product of protein bulk concentration and time. Panel A: the kinetic analysis result for the binding reaction rate constant for the 0.1mg/mL anti-EC sample reacting with the amine-derivatized cantilever glass surface. The binding rate constant is 0.128 ± 0.001 min-1 (mg/mL)-1. Panel B: the binding kinetic results of the 1.0 mg/mL anti-GAS sample reacting with the amine functionalized sensor surface. The initial binding rate constant is 0.012 ± 0.0001 min-1 (mg/mL)-1. 183
Chapter 11: Monitoring of the Self-Assembled Monolayer of 1-Hexadecanethiol on Gold Surface at Nanomolar Concentration Using Piezo-Excited Millimeter- Sized Cantilever Sensor
11.1 Introduction
In the previous Chapter we have reported the detection of various proteins. In this
Chapter, we report the adsorption of 1-hexadecanethiol in ethanol on a gold surface using a new method; namely, piezoelectric-excited millimeter-sized cantilever (PEMC) sensor.
The piezoelectric material we used was lead zirconate titanate (PZT) ceramic film. We have shown that PEMC sensors have sensitivity on the picogram level under liquid immersion conditions (Chapters 9 and 10). We became interested in this problem because the kinetics of thiol on gold has been investigated by a number of researchers, and there is a great body of literature on it. To the best of our knowledge alkanethiol adsorption on gold-coated PZT self-excited millimeter-sized cantilever mass sensor has not been reported, in particular nanomolar concentrations. Another important aspect we examine is adsorption over six orders of magnitude in concentration, and also under conditions where the available number of 1-hexadecanethiol molecules was comparable to the available sites for adsorption. An important advantage of this approach is that we measure resonant frequency change which is a direct function of mass change, which in turn is proportional to the attached mass to the sensor surface.
Self-assembled monolayer (SAM) of long chain alkanethiols on gold surfaces have attracted significant attention in recent years, largely, because they provide an easy method to engineer material surfaces to gain control over the molecular composition and the resulting surface properties232,233. SAMs are known to form stable and well organized spontaneous monolayer on gold surfaces136,234,235. Alkanethiols with more than nine 184 carbon atoms pack into an ordered structure with a tilt of 30o with respect to the gold surface136,214,236. Shorter alkanethiols are less orderly packed,237 and as a result long chain self- assembled monolayer of alkanethiols have been studied intensively as an excellent model system for the adsorption of proteins238-240. The formation of alkanethiol self- assembled monolayer has been investigated using numerous techniques, including
FTIR241, ellipsometry242, surface plasmon resonance243,244, quartz crystal microbalance214,245-247, atomic force microscope108,248,249, radiolabeling250,251, electrochemical impedance spectroscopy217, and others. In most of these studies, researchers focused on the kinetics of the alkanethiol adsorption on gold, and characterization of the adsorbed surface.
n-Alkanethiol is one of the more commonly studied SAM on gold surfaces241,252,253.
Although, there is a large body of information reported in the literature on HS(CH2)nCH3, experiments conducted at very low concentration in the range of nM or when the total number of thiol molecules is of the same order as the available sites for adsorption have not been reported. Further, the adsorption kinetics for the monolayer formation has been reported to be of varied characteristic time. For example, Pan et al.245 reported a characteristic time of 300 min for the monolayer adsorption of 5 mM HS(CH2)11CH3 from ethanol onto a quartz crystal microbalance (QCM), while Karpovich and
6 Blanchard, observed a very rapid assembly of HS(CH2)17CH3 (1-octadecanethiol) monolayer on gold from hexane and cyclohexane in the range of 4 to 30 seconds at 3 to
300 μM, respectively. These differences are thought to be due to the presence of pre- adsorbed contaminant245. It was suggested that solution phase hydrodynamics and the solvent used also affect the kinetics. 185
It has been reported that the formation of an alkanethiol monolayer follows a two step adsorption process235,245,247,252,100,254-256. The first step is rapid with a surface coverage of
60% to 90%, followed by a slower step of reorganization. The two stage adsorption process was described by Damos et al.257 that the thiol molecules first adsorbed and then reorganized to form a densely ordered monolayer. Hu and Bard100 suggested that the phenomenon was due to the repulsive interaction between adsorbed thiol. The kinetic profiles for the formation of alkanethiol self assembled monolayer on gold surfaces have reported to follow the first order Langmuir adsorption model by several authors100,217,245
252 ,258
11.2 Principle of Measurement
The PEMC sensor was a composite structure of two layers: PZT and a gold coated stainless steel foil, a few millimeters in length. Specific binding of a target molecule requires a recognition entity on the sensor surface. Gold (Au (111)) microcrystalline domains serve this purpose in the case of alkanethiol237,246,259,260, while antibody is used for biological entities.When the target of interest binds to the cantilever’s sensing surface, the effective mass of the cantilever increases which decreases the cantilever’s resonant frequency. Hence, by monitoring a cantilever resonant frequency change, quantitative measurements of the cantilever’s effective mass change can be measured in real time. A detailed development of the cantilever physics is available in Chapter 3.
11.3 Materials and Methods
11.3.1 Cantilever fabrication
The cantilever used in this work was fabricated from PZT and stainless steel; technique discussed in Chapter 4. The stainless steel layer (3.5 mm long) was coated with 186 an adhesive layer of chromium (approximately 5 nm thick) followed by vapor deposition of a 10 nm gold layer at the University of Pennsylvania micro-fabrication laboratory (U-
Penn, PA). The gold film was vapor deposited and yielded predominantly a polycrystalline (111) surface (90%).
11.3.2 Reagents
1-Hexadecanethiol (SH-(CH2)15CH3, 92%), concentrated sulfuric acid (H2SO4), hydrogen peroxide (H2O2), and ethanol (99.8%) were purchased from Sigma-Aldrich
(Allentown, PA). Deionized water used was from a Milli-Q plus ultra-pure water system
(18.2 M Ω-cm).
11.3.3 Sample preparation
A stock solution of 10mM 1-hexadecanethiol (SH-(CH2)15CH3) was prepared in absolute ethanol. Serial dilutions of the stock solution with ethanol were carried out to formulate the following concentrations 1nM, 10nM, 100nM, 1mM, 4mM, and 8mM.
These were kept in sealed containers at 4 oC until use. Care was taken to minimize evaporative loss of ethanol during handling.
11.3.4 Formation of alkanethiolate SAM on the gold-coated cantilever
Prior to each experiment the gold-coated sensing stainless steel surface was cleaned with a freshly prepared piranha solution (7:3 volume ratio of concentrated H2SO4 and
30% H2O2) for 10 min (Caution: this reaction is exothermic, and the solution reacts violently with many organic materials and thus, should be handle with great care). The sensing gold surface was then rinsed three times in deionized water, absolute ethanol, and then air dried and used immediately. The cantilever was immersed 1.5 mm into the alkanethiol ethanolic solutions for 1 h. The formation of the thiol SAM on the cantilever 187 gold surface was measured by the resonant frequency response. Before each adsorption experiment was carried out, the sensor was calibrated in absolute ethanol to ensure stability of the resonant peak. In short, this was a control experiment that was executed with a cleaned cantilever immersed in absolute ethanol in order to establish a baseline frequency change. The experiments were setup as discussed in previous Chapters.
11.4 Results and Discussions
11.4.1 Characterization of the PZT-gold coated macrocantilever
Each experiment was repeated at least three times: twice with the cantilever sensor reported in this study and at least once with a cantilever of similar resonance characteristics. For brevity and comparison of results only the data from one cantilever is presented here and the results are typical of the data obtained. Since the PEMC sensors were made manually, each had its own slightly different resonance spectra. Thus it became important to use the same PEMC sensor for all the experiments so that side-by- side comparison can be made. In the reported work, the same PEMC sensor was used in all the experiments. The resonance spectrum of the PEMC sensor in air is shown in
Figure 11.1. Note that only four resonance modes (3, 18.5, 45.5, and 71.5 kHz) were found in the frequency range, 1 to 100 kHz.
The quality factor (Q-value) in air and in ethanol measured for each resonant frequency is given in Table 11.1. Upon immersing the cantilever tip 1.5 mm in pure ethanol the resonant frequency and Q-factor decreased, with significant changes for the first, second, and fourth mode. The third flexural mode decreased by 6 kHz, see Figure
11.2. Also, for the third flexural mode the phase angle and the Q-factor decreased by 0.14 degrees and 5, respectively; see Table 11.1 for further comparison. A peak’s utility for 188 detection is determined by its Q-factor and resonant frequency. The Q-factor is a measure of the sharpness of the peak and is usually reduced by viscous damping when immersed in a liquid. Typically, a high resonant frequency and Q-value leads to a more sensitive peak for measurement. It is worth noting that the small change in Q-value of the third flexural resonant mode is an indication that viscous damping at this frequency is small.
Reynolds number135 of the third mode in ethanol was 6.5 x 106, which supports the experimentally observed small decrease in the Q-value. This is in part because the PEMC sensor’s characteristic dimension is in millimeters, which results in its Reynolds number being large and inertial liquid forces dominate as opposed to viscous forces. For the second and fourth resonant mode in ethanol the same is true, however, the resonant frequencies decreased by one-third and the Q-factor decreased by about one-half. In all the experiments reported in this paper, the third flexural resonance mode was used, because it was the sharpness peak after immersion in ethanol (Q-factor of 33 ± 1, see
Table 11.1) and had a relatively high resonant frequency.
11.4.2 Frequency response to 1-hexadecanethiol self-assembled monolayer formation
In Figure 11.3 we show how the third flexural mode resonant frequency changes as a function of time, during the SAM formation of 1-hexadecanethiol in ethanol at various concentrations, starting at 1 nM. The response shows an exponential decrease reaching a steady state resonance frequency after a period of about 40 minutes. Higher concentration samples showed higher overall resonant frequency decrease. The decrease is due to the binding of 1-hexadecanethiol molecules to the sensor surface, resulting in an increase of the effective mass of the oscillating cantilever. From the figure it can be seen that the rate at which the thiol adsorbed strongly depends on its concentration. In addition, all of the 189 three higher concentrated samples (4, 8 and 10 mM) reached the same overall steady state change (900 ± 10 Hz) within 1 h, suggesting that the resonant frequency change for maximum coverage was 900 Hz. At the lower concentrations, for example 1 to 1000 nM, the total change in resonant frequencies and the adsorption rates are lower and slower, respectively. In Table 11.2, the maximum frequency response as a function of concentration is summarized. From the raw data represented in Figure 11.3, the adsorption kinetics of 1-hexadecanethiol at the various concentrations can be calculated.
11.4.3 Characterization of the kinetics
Self-assembled monolayers of alkanethiols on gold have been shown to obey reversible Langmuir kinetics100,214,245. Several investigators have used an irreversible
Langmuir kinetic as well.217,218 In this Chapter, we will adopt a general approach by allowing reversibility, as such a model reduces to the irreversible one, by setting the desorption rate constant as zero. Here, we assumed that all the adsorption sites on Au have the same activity and is independent of the occupancy of the neighboring sites. That is,
B +Au ' Au•B (11-1) where B represents a thiol molecule. The rate of surface adsorption can be expressed as:
dn = k C 0 ()N − n − k n (11-2) dt a B d where n is the number of thiol molecules on the gold surface at time t, N is the total number of molecules that the surface can hold, therefore, (N-n) is the number of vacant adsorption sites, ka and kb are the intrinsic rate constants of adsorption and desorption,
0 respectively. CB is the bulk concentration of the adsorbed species. Integrating and 190 rearranging Eq. (11-2) gives:
0 n k C 0 a B − ()k a C B + k d t = 0 (1 − e ) (11-3) N k a C B + k d
At equilibrium the net rate of adsorption is taken to be zero and thus, Eq. (11-3) can be
0 rewritten as: k a C B (N − n eq ) = k d n eq , where neq is the number of thiol molecules adsorbed at equilibrium. Rearrangement gives:
n 0 eq k a C B = 0 (11-4) N k a C B + k d
If the mass of each thiol molecule is mB , the resonant frequency change is proportional to
m n the adsorbed mass of alkanethiol on the gold surface, ()− Δf = B , where ()− Δf is the σ
change in resonant frequency at any time t due to the adsorbed mass of thiol ()mB n , and σ is the mass change sensitivity. At steady state when all the attachment sites are filled the
m N total change in resonant frequency is given by: ()− Δf = B . Hence the change in ∞ σ resonant frequency with time can be expressed in terms of the thiol (1-hexadecanethiol) concentration, as:
(− Δf (t )) (− Δf eq ) − ()k C 0 + k t = (1 − e a B d ) (11-5) ()− Δf ∞ ()− Δf ∞
The above can be rearranged to determine the intrinsic rate constants of adsorption and desorption, respectively:
⎡ ()− Δ f ( t ) ⎤ ln ⎢1 − ⎥ = − k obs t (11-6) ⎣⎢ ()− Δ f eq ⎦⎥
191
0 where kobs = kaCB + kd and is therefore the observed reaction rate constant. The above equation suggests that a plot of the left hand side (LHS) versus t will yield a straight line
whose slope is(− kobs ). Once kobs is determined, the rate constants of adsorption and
0 desorption can be calculated from a linear plot of kobs versus bulk concentration (CB ). It should be noted that Eq. (11-6) remains mathematically the same if we had postulated
0 irreversible adsorption kinetics by setting kobs = kaCB . Subramanian and
Lakshminarayanan,214 concluded from their work that SAM formation form a bulk thiol concentration greater than 5 μΜ follows the Langmuir kinetic model, and at concentrations lower than 5 μΜ followed diffusion controlled Langmuir kinetics (DCL).
Therefore, at low concentrations diffusion may become important. That is, the overall adsorption rate may be purely diffusion controlled or follow DCL kinetics. Hence, we evaluated the experimental data at concentrations lower than 5 μΜ with the first order
Langmuir, DCL, and pure diffusion controlled adsorption models. DCL and the purely diffusion controlled models are given by, 214
k t DCL: θ ()t = (1 − e obs ) (11-7)
0 Diffusion: θ ()t = kC B t (11-8)
Δf whereθ ()t = is the surface coverage at any time (t) and k is the diffusion rate Δf ∞ constant. Eq. 11-7 can be rearranged as:
⎡ ()− Δ f ( t ) ⎤ ln ⎢1 − ⎥ = − k obs t (11-9) ⎣⎢ ()− Δ f eq ⎦⎥ 192
For the diffusion controlled Langmuir model a plot of the left hand side of Eq. (11-9) versus t should give a straight line and for the pure diffusion controlled model, Eq.
(11-8), a plot of θ ()t against t should show a linear relationship.
In every experiment the gold sensing surface was prepared in the same manner.
Therefore, we expect nearly identical surface reactivity of the 1-hexadecanethiol. The total frequency change for the adsorption of millimolar thiol concentrations (4, 8, and
10mM) was 900 Hz. Since the same steady state frequency value was reached for each millimolar concentrations, it would be reasonable to argue that the maximum frequency
change ()− Δf ∞ for the adsorption of N 1-hexadecanethiol molecules was 900 ± 10Hz.
The sulfur in alkanethiolate head group (4.5 Å) binds at the Au (111) crystal lattice sites with the closest member approximately 5 Å136,237. The PEMC sensor used in this study had a 6.25 mm2 sensing area. The sensing gold surface has approximately 9.6 x 1013 Au atoms assuming densely packed and gold atom diameter as 2.88 Å. There are nine three- atom hollow sites in the alignment of twelve Au atoms. Thus, approximately 7.2 x 1013 sites are available on the cantilever sensing area. Since, one sulfur head group can only occupy one hollow site; the maximum number of 1-hexadecanethiolate that can form is
7.2 x 1013. Unlike most of the experiments reported in the literature, the total sample volume in this study was 1 mL; used in a batch mode. Thus, at 1 mL the lowest thiol concentration used (1 nM) theoretically has enough thiol molecules (6.023 x 1014 molecules) to fully cover the sensing area, if we assume that the entire surface consists of
Au (111) sites and is defect free. As thiol molecules adsorb on the gold surface, the rate of adsorption slows and the surface reaches a constant equilibrium concentration. As a result, the bulk concentration also reaches equilibrium. Note that at all concentrations the 193 number of surface binding sites is far less than the number of thiol molecules. Since at the lowest thiol concentration, 1 nM, the sample contained 10-times more molecules than needed to fully saturate the sensor surface, the concentration of thiol in the bulk varies only very slightly during an entire experiment. Furthermore, during the first 7 minutes, the change in bulk concentration is even less and can be treated as a constant value. Thus, we chose to use initial rate analysis for obtaining kinetic data. The three adsorption kinetic models (Equations 11-6, 8 and 9) were applied to the frequency response data presented in Figure 11.3. Of the three models the reversible first order Langmuir kinetics gave the best correlation and the results are presented in Figure 11.4, Panel A. The correlation coefficient at 1 nM, 10 nM, 100 nM, 1μM, 4mM, 8mM, and 10 mM were, respectively 0.90, 0.97, 0.97, 0.97, 0.95, 0.99, and 0.99. The data fall quite close to a
straight line, and the slope values ( kobs ) obtained are summarized in Table 11.2. The other two models (data not shown) gave correlation coefficients ranging from 0.52 to
0.63 and 0.56 to 0.68 for the diffusion controlled Langmuir (DCL) and the purely diffusion controlled model, respectively. These results suggest that the adsorption of 1- hexadecanethiol in ethanol onto the gold coated sensor surface followed the reversible first order Langmuir kinetics, even at concentration as low as 1 nM.
The results in Table 11.2 show that, the observed adsorption rate constant (kobs ) decreased as thiol concentration decreased. Pan et al.245 reported on the adsorption of a shorter chain CH3-terminal n-alkanethiol (1-dodecanethiol) in ethanol on a quartz crystal
microbalance (QCM) at bulk concentration of 5 mM, and found kobs values to be similar to the values we obtained. Also Hu and Bard,100 examined the adsorption kinetics of 11-
mercapto-decanoic acid in an aqueous solution and observed kobs in the range of 0.045 to 194
0.02 min-1.
0 The dependence of kobs with bulk concentration (CB ) is illustrated in Figure 11-4,
-1 -1 Panel B. The straight line correlation gave a slope (k a ) 0.061 M s and an intercept
-4 -1 ()kd 3.61 x 10 s . The inset graph shows the relation of kobs to the bulk thiol concentration at low concentration values. The linearity suggests that the overall rate does depend on thiol concentration; somewhat similar to Pan’s results, although they experiment with 1-Dodecanethiol. The linearity of the fit using log10 of the bulk
0 concentrations, log10 (CB ), gave a correlation coefficient of 0.84. From the rate constants
of adsorption and desorption, the binding equilibrium constant (K eq ) was calculated and the free energy of adsorption for 1-hexadecanethiol monolayer was calculated as -12.7 kJ/mol or -3.04 kcal/mol. This value compares favorably with previous results of -5.6 and
-5.5 kcal/mol, reported for 1-octadecanethiol in n-hexane214.
11.5 Conclusion
Gold-coated millimeter-sized PZT cantilever sensor offers a convenient and sensitive method for measuring monolayer formation. The measured resonant frequency change is proportional to mass change of the cantilever. In this study we illustrated the approach with the case of ethanolic solution of 1-hexadecanethiol. The total resonant frequency change obtained for the 1nM, 10 nM, 100nM, 1μM, 4mM, 8mM, and 10 mM thiol concentrations were 116 ± 2 (n=2), 225(n=1), 270 ± 10 (n=2), 440± 10 (n=2), 900 ± 10
(n=2), 900 ± 10 (n=2), and 900 ± 10 (n=2) Hz, respectively. These results compare favorably to literature results in that the rate of the monolayer formation is concentration dependent and the exponential change during adsorption follows the reversible first order 195
Langmuir kinetic model. The rate constants of adsorption and desorption were 0.061 M-
1s-1 and 3.61 x 10-4 s-1, respectively. The significance of the results is that millimeter- sized PZT cantilevers can be used in real-time for characterizing self-assembly of monolayer formation at nano-molar concentration levels. In addition, at 1 nM the adsorption was found not to be diffusion limited. 196
Table 11.1: Resonance peaks characteristics of the cantilever sensor used in the present study.
Mode Resonant frequency Q factor in air Q factor in ethanol in air [kHz]
1 3 ± 0.01 6 ± 1 2 ± 1 2 18.5 ± 0.01 26 ± 1 12 ± 1
3 45.5 ± 0.01 38 ± 1 33 ± 1 4 71.5 ± 0.01 36 ± 1 20 ± 1
197
Table 11.2: The results of observed binding rate constant, kobs , determined from the reversible first order Langmuir kinetic model. The values are averages of three different measurements.
-1 Thiol concentrations (− Δf )∞ [Hz] kobs [min ]
[mM]
10 900 ± 10 0.0504 ± 0.0125 8 900 ± 10 0.0484 ± 0.0140
4 900 ± 10 0.0419 ± 0.0084
-3 1 x 10 440 ± 10 0.0364 ± 0.0152 1 x 10-4 270 ± 10 0.0385 ± 0.0108 -5 1 x 10 225 ± 10 0.0278 ± 0.0046
-6 1 x 10 116 ± 10 0.0227 ± 0.0022
198
-83
-84
-85
-86
-87
-88 Phase Angle [degree] -89 0 20406080100 Frequency [Hz]
Figure 11.1: Resonant spectra of phase angle versus frequency of the gold plated cantilever in air. The first peak represents the fundamental resonance mode and higher modes occurred at high frequency. The cantilever was excited with 100 mV. 199
-83.5 -84.0 In Ethanol In Air -84.5 -85.0 -85.5 -86.0 -86.5 -87.0
Phase angle [degrees] angle Phase -87.5 -88.0 -88.5 30 35 40 45 50 55 60
Excitation frequency [kHz]
Figure 11.2: The third flexural mode resonant peak of the cantilever in air (right, solid line) and in ethanol solution immersed to a depth of 1.5 mm (left, broken line). Damping of cantilever response was found to be small as Reynolds number is well over a one million. 200
100
0 Control
-100 1 nM
-200 10 nM 0.1 μM -300
-400 1 μM 4 mM -500 8 mM -600 10 mM
Resonant frequency change [Hz] -700
-800
-900 0 102030405060 Time [min]
Figure 11.3: The third flexural mode resonant frequency response to various concentrations of 1-hexadecanethiol. As the concentration of thiol increase the total change in resonant frequency increased. However, at concentrations of 4mM and above the total frequency change reached the same steady state value, suggesting maximum surface coverage. The control response shown is that of the cleaned cantilever immersed in ethanol, so as to establish the baseline frequency change of the sensor. 201
0.00 -0.05 A -0.10 -0.15 ) 1nM ∞ f -0.20 Δ 10nM
f)/ -0.25 Δ
- 100nM ∞
f -0.30 Δ 1000nM -0.35
ln(( 4mM -0.40 8mM -0.45 10mM -0.50 02468 Time,t [min]
0.0016 2.1E-03 0.0012 B ] -1
[s 0.0008 obs 1.6E-03 k 0.0004 ]
-1 0 [s 0.000 0.000 0.000 0.001 0.100 obs 1.1E-03 0 k Log, C [M ] B
6.0E-04
1.0E-04 0 0.002 0.004 0.006 0.008 0.01 0.012 0 Thiol concentration, C B [M ]
Figure 11.4: Panel A: Langmuir kinetic analysis of thiol adsorption on PEMC sensor surface. The initial kinetic analysis of the various 1-hexadecanethiol concentrations with correlation coefficient ranged from 0.90 to 0.99. The slope of each line gives the observed characteristic binding rate constant kobs. Panel B: Concentration dependence of kobs on the adsorption of 1- hexadecanethiol in ethanol onto a gold coated millimeter-sized cantilever mass sensor. The best fit line through the data points gave a slope (k ) and an intercept ()k . The inset shows a d the observed binding rate constant as a function of log base ten of the bulk thiol concentrations. 202
Chapter 12: Piezoelectric-excited Millimeter-sized Cantilever (PEMC) Sensors Measure Albumin Interaction with Self-assembled Monolayers of Alkanethiols having different Functional Head Groups
12.1 Introduction
Adsorption of proteins on solid surfaces and the evaluation of their characteristics for comparison to the native protein are of general interest in biomedical260-262 and biomaterials and in effort to design biocompatible surfaces263,254. Protein adsorption on synthetic material is a complex process that involves conformational changes, orientation and rearrangement, and protein detachment, which affects the protein activity233,265
Therefore, there is the need for a rapid real-time protein adsorption technique to characterize the adsorption process on various surfaces. The well characterized self- assembled monolayer (SAM) of alkanethiol with different terminal function groups on gold surfaces is a general model system for studying the adsorption of proteins and has been used by several researchers for biointerface design244,249,250,266. The functional end groups provide proper orientation and molecular organization when SAM is formed on gold surface. Several real-time techniques have been used to measure protein adsorption on solid surfaces via an alkanethiolate monolayer. They include acoustic plate mode sensors224, chronopotentiometry267, and piezoelectric quartz crystal sensors246,268.
Although these methods offer excellent platforms for characterizing protein adsorption on surfaces, their sensitivity per unit area is low135. On the other hand, piezoelectric-excited millimeter-sized cantilever sensor offers high sensitivity, and has been used successfully to monitor the self assembly of 1-hexadecanethiol (Chapter 11), and detection of protein
(Chapter 10). The electromechanical resonator operates on the converse piezoelectric effect where the application of an electric field creates a strain. 203
Recently Martins et al.267, used chronopotentiometry techniques to determine the adsorption of human serum albumin (HSA, 10 μg/mL) on SAMs of various functional end groups: carboxyl (COOH), hydroxyl (OH), and methyl (CH3). They concluded that the adsorption of HSA increased in the following order: -CH3 > -COOH> -OH. In this chapter, we report the adsorption behavior of HSA on self-assembled monolayer of the same set of end groups using a new modality: Piezoelectric-excited millimeter-sized cantilever (PEMC) sensor. We have used the same reagents, namely the alkanethiols and the HSA, from the same source as did Martins et al.267
12.2 Materials and Methods
12.2.1 Cantilever fabrication
PZT-stainless steel cantilevers were fabricated with free ends dimensions of PZT 1 x
2 mm2 and gold coated stainless steel layer 3.5 x 2 mm2 (L x w). For details see schematic illustration of the sensor geometry in Figure 4.1A. The experimental arrangement is shown in Figure 12.1.
12.2.2 Reagents
1-decanethiol (SH-(CH2)9CH3, 96%), 11-mercapto-1-undecanol (SH-(CH2)11OH,
97%), 11-mercaptoundecanoic acid (SH-(CH2)10COOH, 95%), concentrated sulfuric acid
(H2SO4), hydrogen peroxide (H2O2), and ethanol (99.8%) were purchased from Sigma-
Aldrich (Allentown, PA). Deionized water used was from a Milli-Q plus ultra-pure water system (18.2 MΩ-cm).
204
12.2.3 Gold substrate, Monolayer formation, and HSA interaction
The gold-coated sensing stainless steel surface was cleaned with a freshly prepared piranha solution (1:3 H2O2(30%) to concentrated H2SO4) for 2 minutes (Caution: this reaction is exothermic, and the solution reacts violently with many organic materials and thus, should be handled with great care). The sensing gold surface was then rinsed three times in deionized water, absolute ethanol, and then air dried and used immediately.
One milliliter stock solutions (1-decanethiol, 11-mercapto-1-undecanol, 11- mercaptoundecanoic acid) were diluted with absolute ethanol to a final concentration of 1 mM. The samples were kept in sealed containers at 4 oC until use. Care was taken to minimize evaporative loss of ethanol by conducting each experiment in a temperature controlled chamber, saturated with ethanol; humidity 92%, maintained at 25 ± 0.2 °C.
During a typical 1 hour experiment the liquid level in the sample container (1 mL) decreased slightly (approximately 4 μm/min). The linear change in liquid level with time resulted in a linear change in resonant frequency. Thus immediately before and after an experiment, the rate of change of resonant frequency due to liquid level change was measured and an average value of the two was used to correct the peak position in the detection experiment. Prior to each measurement the sample container (1 mL) was rinsed with the solution of interest. Self-assembled thiol monolayer was formed on the sensor surface by immersion of the clean sensor 1.5 mm into the alkanethiol solution (1 mL, of 1 mM thiol solution) for 1 h. The formation of the thiol monolayer on the cantilever gold surface was measured by the resonant frequency response.
Human serum albumin (HSA) solution was prepared in phosphate buffered saline
(PBS, pH 7.4) to a final concentration of 10 μg/mL. Upon formation of the alkanethiolate 205 monolayer on the sensor surface, the sensing area was rinsed in PBS and then, exposed to the HSA (10 μg/mL) solution until equilibrium was achieved as indicated by constant resonant frequency of PEMC. The SAM-protein interaction was monitored by tracking the measured resonant frequency of the sensor.
12.2.4 PEMC sensor calibration
The mass change sensitivity of the PEMC sensor in air was determined by adding known point mass of paraffin wax at the cantilever tip. A small fragment of the paraffin wax (0.23 mg) was dissolved in 4 mL of hexane solution. A 0.5 μL aliquot was dispensed into ten weighing dishes (the weight of each dish was recorded before dispensing the solution) and the dishes were left to dry in a fume hood for 30 minutes, after which each dish was re-weighed and the mass of wax dispersed was computed from the difference between the dish and dish plus wax mass measurement. The frequency of the resonant mode used in the detection experiment was measured in air. The mass change sensitivity of the sensor was determined by dispensing 0.5 μL of the paraffin-hexane solution on the sensor tip, dried for 30 minutes in the fume hood, and the resonant frequency of the detection mode was measured immediately. The procedure was repeated for six different mass changes, and five times for each mass change. The resonant frequency changes resulting from these specific paraffin mass changes were plotted to determine experimental measure of the mass change sensitivity of the PEMC sensor used in this study. 206
12.3 Results and Discussions
12.3.1 Characterization of the PEMC sensor
The resonance characteristics of the sensor are presented in Figure 12.2. The resonant spectrum, a plot of phase angle versus excitation frequency, exhibits resonance modes in air at 12.5, 27.5, and 45.5 kHz in the frequency range 1 to 100 kHz. Several repeat experiments showed that these resonant frequencies were within ± 50 Hz of each other when measured at different times. In any single experiment, the resonant frequency remained constant within ± 5 Hz suggesting that the basic features of the sensor were stable. The day to day variations (± 5 Hz) are thought to be due to changes in the local temperature. Upon immersion of the sensor in liquid ethanol, the resonant frequencies decreased (peaks shifted to the left) due to the increase in the effective mass of the sensor caused by the added mass of ethanol. Also, the peak heights decreased due to mass damping. The change in resonant frequency upon liquid immersion is a measure of peak sensitivity; the larger the change in resonant frequency the more sensitive the sensor. The change in resonant frequency from air to ethanol for the first, second, and third modes were 2, 2.5, and 6 kHz, respectively; see Figure 12.2, broken lines. In this study the third resonant mode, 39.5 kHz in ethanol, was used due to the sharpness of the peak under liquid immersed conditions and the greater response upon immersion. The sharpness of the peaks is defined by the quality factor (Q value) determined from the ratio of the resonant frequency to the peak width at half the peak height. The larger the Q value the more suitable a resonant peak is for detection. The Q values in air were 25.4 ± 0.1, 39.2 ±
0.1, and 50.9 ± 0.2 and in ethanol 17.5 ± 0.1, 27.7 ± 0.1, and 43.9 ± 0.2 for the first, second, and third modes, respectively. 207
12.3.2 SAM formation
In Figure 12.3 we show how the third flexural mode resonant frequency changes as a function of time, during the self-assembled monolayer formation of 1-decanethiol (SH-
(CH2)9CH3), 11-mercapto-1-undecanol (SH-(CH2)11OH), and 11-mercaptoundecanoic acid (SH-(CH2)10COOH) from a 1 mL, 1mM solution in absolute ethanol. Each experiment was repeated at least twice and the data shown are typical of the results obtained. The resonant frequency showed an exponential decrease before reaching a constant value. The steady state resonant frequency change observed suggests that a stable monolayer was formed on the gold coated sensing surface. 11-mercaptoundecanoic acid showed the largest change in resonant frequency followed by 11-mercapto-1- undecanol, and then 1-decanethiol. The total resonant frequency change obtained for SH-
(CH2)10COOH, SH-(CH2)11OH, and SH-(CH2)9CH3 were 885 ± 21 (n=2), 590 ± 14 (n=2), and 383 ± 10 (n=2) Hz, respectively. These results suggest that the monolayer formation depends on the functional head group of the alkanethiolate.
The mass change sensitivity of the sensor was determined using the known tip mass technique discussed in section 12.2.4. The mass of wax in 0.5 μL of the paraffin-hexane solution was determined as 47.8 ± 12.2 ng (n=5). As the known mass of wax was added to the cantilever tip the resonant frequency decreased, which is verified by the shifted in the resonant peak to the left (see Figure 12.4A). For clarity only the first, third, fifth, seventh, and eight mass additions are presented in Panel A; due to the small shifts in the resonant peak. A plot of the known mass change to the sensor tip against the corresponding resonant frequency change, Figure 12.4B, of the detection peak yielded a straight line whose slope is the mass change sensitivity in air, 6.49 ± 1.40 ng/Hz. This 208 suggests that if resonant frequency resolution is 1 Hz, then mass changes in the range
5.09 ng to 7.89 ng are discernable. The mass change sensitivity in liquid was then calculated from the sensitivity in air using a model equation developed in Chapter 6, Eq.
(6-3).
⎛ M f ⎞ σ = σ ⎜ ef . n ⎟ nf na ⎜ ⎟ ⎝ M e f nf ⎠
th where σ na and σ nf are the mass change sensitivity of the n mode in air and in fluid,
respectively. M ef is the effective mass of the sensor under liquid immersion and
therefore, M ef = M e + mae . Calculated values for M e , mae , K , and σ nf are given in
Table 12.1. The mass change sensitivity of the third mode (detection peak) under liquid
immersion (σ 3 f ) was determined as 9.74 ± 1.4 ng/Hz. Using the mass change sensitivity, the mass of adsorbed thiol molecules on the sensor gold surface after one hour were 3.73
± 0.45 μg, 5.75 ± 1.53 μg, and 8.62 ± 1.38 μg corresponding to 1-decanethiol, 11- mercapto-1-undecanol, and 11-mercaptoundecanoic acid, respectively.
The sensor used in this investigation has sensing area 6 mm2. For complete monolayer coverage of the alkanethiols, 7.2 x 1013 molecules were calculated to be absorbed on the gold sensing cantilever surface. The sulfur atom at the end of each thiol molecule binds at the Au (111) crystal lattice sites with the closest member approximately 5 Å136,237. In this study the sample volume was 1 mL and the concentration of each sample was 1mM. Therefore, theoretically there were enough thiol molecules
(6.023 x 1020 molecules) to fully cover the sensing area in each of the experiments. From the calculated value of the number of thiol molecules in a monolayer, the mass of the three monolayers were determined as 21, 24, and 26 ng corresponding to 1-decanethiol, 209
11-mercapto-1-undecanol, and 11-mercaptoundecanoic acid monolayer, respectively. It is clear from this calculation that the mass changes obtained from the sensitivity value are two orders of magnitude higher. In the literature the surfaces of self-assembled monolayers with different terminal end groups have been reported by several authors269,270. Faucheux et al.270, showed that a self-assembled monolayer with the following terminal head groups, methyl (CH3), hydroxyl (OH), and carboxyl (COOH), creates hydrophobic, wettable, and moderately wettable surfaces, respectively. Therefore, one possible explanation for the higher mass determined by the sensitivity is that the ethanol solution penetrates the monolayers and as a result adds to the mass change on the sensor. It is also suggested that the three monolayers allow different levels of penetration.
The hydrophobic nature of 1-decanethiol monolayer provides the least penetration due to the polar nature of the ethanol molecules. On the other hand, the amphipathic nature of
11-mercapto-1-undecanol and 11-mercaptoundecanoic acid allows greater penetration and more ethanol molecule in the monolayer, resulting in a larger resonant frequency change and the corresponding calculated mass change.
12.3.3 Human serum albumin adsorption
The response of the cantilever to human serum albumin (HSA) interaction to SAMs terminated with -CH3, -OH, and -COOH end groups are presented in Figure 12.5. The concentration of HSA used in each of the experiment was 10 μg/mL. Upon exposure of the monolayer to the serum protein the resonant frequency decreased immediately, and reached a steady state value within 5 to 30 minutes. For the three terminal groups, the total change in resonant frequency was 520, 290, and 210 Hz, respectively. In the case of
OH-terminated SAM the interaction was rapid and steady state was achieved in 6 210
minutes. On the other hand for -COOH and -CH3 functional groups the response was considerably longer, taking 20 to 30 minutes. All the responses were exponential (see
Figure 12.5) suggesting binding or phase transfer behavior. The interaction of HSA to
SAM terminated with -CH3 gave the largest total resonant frequency change of 520.8 ±
8.6 (n=3) Hz. The total resonant frequency change of the -COOH and -OH terminated self-assembled monolayer to HSA interaction was significantly less than -CH3–HSA response, and was slightly different from each other. The total frequency response of the -
COOH and -OH terminated SAM to HSA interaction were 290.4 ± 6.1 (n=2) Hz and
210.6 ± 8.1 (n=3) Hz, respectively. It is clear from Figure 12.5 that the initial adsorption kinetics for the different types of HSA interaction was approximately the same. It is worth noting that the CH3-terminated SAM formation (section 12.3.2) gave the lowest total resonant frequency change and however, showed the largest frequency response upon exposure to the serum protein solution. Protein conformation on artificial surfaces is an active research area244,271. Protein adsorption is complex and is difficult to predict272.
Monolayers have been shown to possess at least three different characteristics to prevent protein adsorption. These are monolayers should be hydrophilic, contain hydrogen-bond acceptors, and have an overall electrical neutrality273,274. Hydrophobic surfaces have been shown to exhibit high albumin adsorption, while hydrophilic surfaces have low albumin affinity267,243. Using the mass change sensitivity of the sensor in liquid, 9.74 ng/Hz, the mass changes per unit area of sensor surface to HSA adsorption on -CH3, -COOH, and –
OH terminated SAMs were determined as 0.84 ± 0.01, 0.47 ± 0.02, and 0.34 ± 0.01
μg/mm2, respectively. 211
12.3.4 Kinetics of HAS adsorption
The adsorption of 1-hexadecanethiol and Antibody-antigen binding on PEMC sensors have been shown to obey Langmuir kinetics (Chapters 10 and 11). Here, we examine if
HSA adsorption on the various thiol monolayers with different functional end groups may obey the same kinetics. At time close to τ = 0, HSA concentration gradient between the sensor surface and the bulk sample is small and thus, diffusion effects may be ignored. The model is expressed by Eq. (11-6):
⎛ ()()Δf − Δf ⎞ ⎜ ∞ ⎟ ln⎜ ⎟ = −k obsτ ⎝ ()Δf ∞ ⎠
The observed rate constant kobs during the initial time (far from equilibrium) can be determined from a plot of the left hand side (LHS) versusτ and is shown in Figure 12.6, where we have considered data obtained during the first 9 minutes. In Figure 12.6A, the
observed adsorption rate constant (kobs ) for the adsorption of the various alkanethiols to the sensor gold surface was determined as 0.041 ± 0.001, 0.069 ± 0.004, and 0.081 ±
-1 0.001 min corresponding to –CH3, -COOH, and –OH terminal alkanethiols. Note that the experiments were carried out at 1 mM. These rates are similar in magnitude to the values reported by Hu and Bard (1998) for mercaptoundecanoic acid on gold surfaces as
-1 100 0.045 ± 0.005 and 0.02 ± 0.003 min at 0.5 and 0.05 mM, respectively . Modeling the
HAS adsorption data presented in Figure 12.5 in the same fashion kobs was determined to be 0.163 ± 0.003, 0.248 ± 0.006, and 0.381 ± 0.001 min-1 corresponding to the adsorption of HAS to the –CH3, -COOH, and –OH self-assembly thiol monolayer functional end groups, Figure 12.6B. The quality of the fits was good with correlation coefficients of
0.96 to 0.98. This indicates that the adsorption rate of HAS on to alkanethiolate is about 212 an order of magnitude more rapid than the adsorption of thiol onto gold surface.
12.4 Conclusion
We described a new modality of measuring human serum albumin (HSA) adsorption continuously on -CH3, -COOH, and –OH terminated self-assembled monolayers (SAMs) of C11-alkanethiols and the direct quantification of the adsorbed amount. A gold-coated piezoelectric-excited millimeter-sized cantilever (PEMC) sensor of 6 mm2 sensing area was used in this investigation. Self-assembled monolayers (SAMs) of C11-thiols (in absolute ethanol) with different end groups was prepared on the PEMC sensor, and then exposed to buffer solution containing HSA at 10 μg/mL. The resonant frequency decreased exponentially and reached a steady state value within 30 minutes. The decrease in resonant frequency indicates that the mass of the sensor increased due to HSA adsorption onto the SAM layer. The frequency change obtained for the HSA adsorption on -CH3, -COOH, and –OH terminated SAM were 520.8 ± 8.6 Hz (n=3), 290.4 ± 6.1 Hz
(n=2), and 210.6 ± 8.1 Hz (n=3), respectively. These results confirm prior conclusions that albumin adsorption decreased in the order: CH3>COOH>OH. Observed binding rate constants were 0.163 ± 0.003, 0.248 ± 0.006, and 0.381 ± 0.001 min-1, for methyl, carboxylic and hydroxyl end groups, respectively. The significance of the results reported here, is that both the formation of self-assembled monolayers and adsorption of serum protein onto the formed layer can be measured continuously, and quantification of the adsorbed amount can be determined directly.
213
Table 12.1: Calculated parameters of the second flexural mode resonant frequency.
σ x109 σ x109 Spring constant, Effective mass mae, 1.5 mm immersion 3a 3 f -3 6 7 K x10 (N/m) Me x10 (kg) in ethanol 10 x[kg] (g/Hz) (g/Hz) [Eq.2] [Eq.3] [Eq.4] Experimental [Eq. 7]
6.35 2.97 9.71 6.49 9.74 214
Impedance PC Analyzer Constant temperature chamber
XYZ Micromanipul PEMC sensor ator
Sample container
Vibration free Table
Figure 12.1: Schematic illustration of the experimental setup.
215
-60
nd -65 2 modes
Air -70 Ethanol
-75 1st modes 3rd modes -80 Phase angle [degrees]
-85
-90 0 102030405060 Excitation frequency [kHz]
Figure 12.2: Resonance spectrum of PEMC sensor in air (solid line) and in ethanol (broken line). The spectrum is a plot of phase angle versus excitation frequency. The fundamental (1st), second, and third modes in air occurred at 12.5 ± 0.05 kHz, 27.5 ± 0.05 kHz, and 45.5 ± 0.05 kHz, and in ethanol were 10.5 ± 0.05 kHz, 25 ± 0.05 kHz, and 39.5 ± 0.05 kHz, respectively. Resonance is followed by a sharp change in the phase angle. The sensor was excitated with 100 mV. 216
100 0 Control -100 -200 -300 -400 CH3 -500 -600 OH -700 -800 -900 COOH
Resonant frequency change [Hz] change frequency Resonant -1000 0 10203040506070 Time [min]
Figure 12.3: Resonant frequency change for the adsorption of the various alkanethiols on the gold coated PEMC sensor surface. The terminal end groups were CH3, OH, and COOH. Control response was in thiol-free ethanol solution. 217
-79.5
-80
A -80.5 47.8 ng 143.4 ng -81 239.0 ng 334.4 ng
Phase angle [degrees] Phase angle -81.5 382.2 ng
-82 45.5 45.6 45.7 45.8 45.9
Resonant frequency [kHz]
450 400 B 350 300 250 200 150 Mass change [ng] Mass 100 50 0 0 10203040506070
-ΔF [Hz]
Figure 12.4: Panel A: A plot of resonant peak as known mass of wax was added to the PEMC sensor’s tip. For clarity the Figure shows the resonant peaks of the 47.8, 143.4, 239, 334.4, and 382.2 ng of added paraffin wax, respectively. Panel B: A plot of known mass change versus the corresponding resonant frequency change. The mass change sensitivity of the PEMC sensor in air was determined from the slope as 6.497 ng/Hz. 218
100
0 Control
-100
-200 OH-HSA
-300 COOH-HSA -400
-500
Resonant frequency change [Hz] change frequency Resonant CH3-HSA -600 0 10203040506070 Time [min]
Figure 12.5: Resonant frequency change for the adsorption of human serum albumin (HSA) on the different self-assembled monolayer (SAM) terminated functional end groups. The control response is typical of a SAM monolayer exposed to HSA-free PBS solution. 219
0
-0.2 A )) ∞ f
Δ
f)/ -0.4
Δ -CH3 - ∞ f -COH Δ -0.6 -COOH ln((
-0.8 0246810 Time, τ [min]
0.0
-0.5 B
-1.0 )) ∞
f -1.5 Δ f)/
Δ - -2.0 OH-HSA ∞ f Δ COOH- HSA -2.5 ln(( CH3-HSA -3.0
-3.5 0246810
Time , τ [min]
Figure 12.6: Panel A: The initial adsorption kinetics for the adsorption of alkanethiols on to the gold plated sensor surface. Correlation coefficient range from 0.98 to 0.99. Panel B: Initial kinetic analysis for the binding of HSA to self- assembled thiol monolayers having different terminal end groups. The fits are good with correlation coefficient ranged from 0.96 to 0.98. The slope of each line gives the observed characteristic binding rate kobs. 220
Chapter 13: PEMC Sensor’s Mass Change Sensitivity is 20 pg/Hz under Liquid Immersion
13.1 Introduction
Most cantilever sensors reported in the literature are sub-millimeter size, and normally do not operate under liquid immersion conditions. However, direct measurement of antigen concentration in liquid media is a desirable for practical applications. In Chapter 6, we reported that a piezoelectric-excited millimeter-sized cantilever (PEMC) sensor detected the pathogen E. coli O157:H7 in the concentration range of 7 x 102 to 7 x 107 cells/mL. The cantilever had a three layer composite structure consisting of lead zirconate titanate (PZT), stainless steel and borosilicate glass, with a mass change sensitivity of 1 x 10-8 g/Hz. We also showed that a PZT-stainless steel base cantilever sensor detected liquid level change of 0.26 μm (Chapter 5), which corresponds to a mass change sensitivity of 1.84 x 10-9 g/Hz. More recently, we investigated the detection limit of E. coli O157:H7 (Chapter 7) using a modified PEMC design consisting of two layers (glass and PZT) and found that it was 70 cells/mL. In this study, we characterized the modified PEMC and show that the revised two-layer design enhances mass change sensitivity by nearly three orders of magnitude. We demonstrate the sensitivity by detecting a model pathogen, Group A Streptococcus (GAS) pyrogens at various concentrations.
Group A Streptococcus (GAS) is a common human bacterial pathogen that infects the throat and skin and is capable of causing serious and life threatening illnesses275-277. The pathogen is easily transmitted from person to person by inhalation of aerosols emitted by an infected individual. Detection of GAS relies on tedious bacterial culture technique. 221
Recently, several tests to detect GAS were developed. These include enzyme-linked immunosorbent assays278, immuno-PCR assays279, and Gen-Probe test280,281. These are multi-step techniques requiring sample preparation. Recently, Liang et al., (2003) developed a highly sensitive immuno-PCR assay for detecting GAS with a sensitivity of
10 cells279.
13.2 Materials and Methods
13.2.1 PEMC Fabrication
The construction features of the glass-PEMC sensor were described in Chapter 6. In that design (PEMCo) we used 50 micron thick stainless steel between the lead zirconate titanate (PZT) film and the glass layer to provide for electrical connection to the bottom side of PZT. In the present design, we eliminated this layer and installed the PZT film to protrude beyond the glass so that access to both bottom and top sides are available for electrical connection. The procedure of manipulating small and fragile PZT pieces required some care, but was easily accomplished with good reproducibility. The PZT dimensions of the sensor investigated (PEMCn1) were 1 ± 0.05 x 1 ± 0.05 x 0.127 ± 0.005
3 mm (L x W x t). The resonance characteristics of PEMCn1 were compared to a sensor of the same geometry (PEMCn2) that was of dimensions 1 ± 0.05 x 2 ± 0.05 x 0.127 ± 0.005
3 mm . The dimensions of the overhanging glass at the free end of PEMCn1 and PEMCn2 were 3 ± 0.05 x 1 ± 0.05 x 0.160 ± 0.005 mm3 and 3 ± 0.05 x 2 ± 0.05 x 0.160 ± 0.005 mm3, respectively.
13.2.2 Antibody immobilization
The procedure used for antibody immobilization was the same as that we reported in
Chapter 6, except for the use of affinity purified monoclonal antibody to GAS; anti-GAS 222 was purchased from ViroStat, Portland, ME. Concentration of antibody used in all immobilization procedures was at 100 μg/mL in 10 mM phosphate buffered saline (PBS,
Sigma-Aldrich) solution adjusted to pH 7.4. The immobilization process was monitored for mass change, as the PEMC sensor has the sensitivity to track real time mass change.
One milliliter of the antibody solution was loaded into a 1 mL container placed in a temperature controlled chamber and the aminated glass tip of PEMC sensor was immersed in the sample to a depth of 1mm using an XYZ-manipulator. The impedance and phase angle of a resonant peak was monitored and recorded for two hours.
13.2.3 Experimental arrangement and mass change sensitivity
All experiments were done in a 1΄ x 1΄ x 2΄ temperature-controlled chamber maintained at 22 °C. The PEMC sensor was connected to an impedance analyzer (HP
4192A) interfaced to a PC that ran a custom designed LabView program for data acquisition and archiving. The program enabled a coarse sweep of desired frequency range, followed by a finer sweep (user specified) in the region of maximum phase angle.
The program has features to select any particular resonant peak, and a frequency resolution as low as 1 Hz. Often, a coarser resolution of 5 Hz was used. The data collected included impedance, equivalent circuit parameters, and phase angle in the range
of 1 kHz to 100 kHz. The mass change sensitivity of the second flexural mode (σ 2a ) of the cantilever in air was established by the dip-touch technique. The mass change
sensitivity under liquid immersion,σ 2 f , was determined from the mass change sensitivity
in air ()σ 2a using Eq.(6-3). 223
13.2.4 Detection experiments
A 50mL aliquot of heat inactivated Group A Streptococci suspension (GAS, 7 x 109 cells/mL) was purchased from Immuno Resources Inc., TX. Serial dilutes of the 7 x 109 cells/mL sample using 10 mM phosphate buffered saline (PBS) solution pH 7.4 were done to prepare lower stock solutions of 7 x 107, 7 x 106, 7 x 105, 7 x 103, and 700 cells/mL. PEMC sensor bearing anti-GAS was immersed to 1 mm in a container containing GAS cells for one hour. The impedance and phase angle of the second mode resonant peak was monitored and recorded for two hours. Following a binding experiment, the PEMC sensor was immersed in a low pH buffer, PBS adjusted to pH 1.5 with HCl, for 1 hour to obtain real time data on the release of the pathogen. Again, the impedance and phase angle of the second mode was monitored.
In all the experiments the PEMC sensor was partially immersed into the liquid sample without the sample being covered. This caused a constant change in resonant frequency due to liquid level change (typically ~ 4.5 μm/min). The systematic error to resonant frequency change caused by the liquid level change was corrected as previously discussed in Chapter 6.
13.3 Results and Discussions
13.3.1 Characterization of the PEMC sensors in air
Over 30 cantilevers of the old design, and about 60 of the new design were fabricated, and their resonant frequency spectra in air and in PBS were determined. The performance of one of the old design (labeled PEMCo) and two of the new design (labeled PEMCn1 and PEMCn2) are summarized in Tables 13.1. The three given in the Table are representative of the 90 cantilevers evaluated. Although there were small variations in 224 performance due to dimensional differences, by and large, they had similar resonance characteristics. Besides the elimination of the stainless steel layer, the new designs are slightly shorter, having 2 mm shorter glass, and 1 mm shorter PZT layer. PEMCn2 has the same overall width as PEMCo, but PEMCn1 is half as wide (1 mm). For all PEMC sensors only two resonant peaks were observed in the frequency range of 1 to 100 kHz.
The first (fundamental) mode and the second mode resonant frequencies of PEMCo in air were 12.8 and 65.8 kHz. By shortening the length from 5 mm to 3 mm and eliminating the stainless steel layer, these modes moved to 17.5 and 53 kHz, respectively. Upon narrowing the width from 2 to 1 mm, the modes increased to 18 kHz and 72.5 kHz.
Basically, the new designs have increased the fundamental mode by about 50%. Second mode change appeared to be width dependent.
In Table 13.1 we have tabulated the resonant frequencies for both PEMCo and PEMCn1 in air (fa) and in 1-mm immersion in PBS (fnf). The resonance spectra, a plot of phase angle versus excitation frequency, (figure not shown) showed PEMCn1 baseline phase angle closer to -90° compared to PEMCo suggesting that its capacitance characteristics is slightly better, possibly due to the elimination of the intervening stainless steel layer.
Also, the higher Q-value obtained in air (Qa) for the first mode indicated that the intrinsic damping in the PEMCn1 sensor was lower. Interestingly, the Q factor value in air for the first and second modes for PEMCn1 were, respectively, 70 and 38 while those of PEMCn2 were 25 and 38. These values were comparable to those obtained with the old wider and longer cantilever; namely, 63 and 33. Upon immersing the cantilevers to a depth of 1mm in PBS solution, both the resonant frequency and Q-value decreased, due to viscous damping. The decrease in Q-values was about 10% for all three PEMC sensors. See 225
Table 13.1.
13.3.2 Selection of mode for detection
The mass change sensitivity and the Q-value of a given resonant mode of a cantilever determine the suitability for detection. Since the cantilevers exhibit two modes, higher frequency was preferred as higher sensitivity is achieved with it. Hence second flexural resonant mode was used in all experiments. For most PEMC sensors that were fabricated the second mode exhibited a higher Q-factor than the fundamental mode.
13.3.3 Mass change sensitivity in air and under liquid immersion
The mass change sensitivity in air, see Figure 13.1, was determined by the dip-touch
- silicone oil technique. PEMCn1 exhibited second mode sensitivity(σ 2a ) in air of 1.56x10
11 g/Hz, see Table 13.2. That is, if the resonant frequency resolution is 1 Hz, then mass
changes of 15.6 pg is discernable. The sensitivity under liquid immersion (σ 2 f ) was
estimated from()σ 2a using Eq. (6-3). The parameter M ef was calculated from resonant frequency values obtained experimentally, and is listed in Table 13.1. The intrinsic and
viscous damping coefficients, c0 and cv were calculated from phase angle peak shape
using the relationships: QfMc= 2/π ne0 andQfMccf = 2/()π nf ef0 + v . The calculated
-11 -11 values of σ 2 f for PEMCn1, PEMCn2 and PEMC0 were 2.27 x 10 , 72 x 10 and 14.3 x
10-9 g/Hz, respectively. The significant increase in the sensitivity of the new design over
PEMC0 was due to removing the stainless steel layer, smaller dimension, and the reduction in the number of adhesive layers. The reduction in number of layers appears to have decreased intrinsic damping coefficient (80.3 and 11.6 kg/m-s compared to 372.6 kg/m-s) and reduced effective mass (1.01 and 2.02 μg compared to 3.71μg). Furthermore, 226
due to the smaller size of PEMCn1 sensor both the viscous damping coefficient (cv) and the increase in added mass due to liquid immersion were smaller; see Table 13.1. Since
PEMCn1 had superior sensitivity, it was selected for detecting Group A Streptococcus
(GAS).
13.3.4 Detection of Group A Streptococcus (GAS)
Model pathogen Group A Streptococcus (GAS) was detected at various concentrations with the PEMCn1 sensor to demonstrate its enhanced resonance characteristics. The second resonant mode frequency response of the cantilever to samples containing the pathogen (GAS) at concentrations 700, 7 x 103, 7 x 105, 7 x 106, 7 x 107, and 7 x 109 cells/mL is shown in Figure 13.2. For the highest concentration the resonant frequency decreased more rapidly and reaches a greater steady state frequency change. Total frequency change for the 700, 7 x 103, 7 x 105, 7 x 106, 7 x 107, and 7 x 109 cells/mL samples were 136, 509, 690, 1130, 1260, and 1782 Hz respectively. The lowest concentration, 700 cells/mL, showed the slowest binding kinetics and steady state was reached in the first 25 minutes. These results show that the rate of change in resonant frequency is strongly dependent on concentration. Using the mass change sensitivity of the cantilever under liquid immersion, a value 2.27 x 10-11 g/(Hz), the mass of GAS attached during the first hour from the lowest to the highest concentrations were, respectively, 3.51 ± 0.51, 11.6 ± 1.3, 15.7 ± 1.1, 25.77 ± 0.17, 28.6 ± 2.2, and 40.5 ± 3.1 ng (n=3 for all). The exposed surface area for attachment was 2.48 mm2 and assuming that an average bacterium has a cross sectional area of 3.85 x 10-7 mm2 (diameter 0.7μm), approximately 6 x 106 cells would form a monolayer surface coverage. Taking the density of the pathogen as 1.02 g/cm3, monolayer coverage would correspond to a mass 227 change of 1.1μg.
It can be observed from Figure 13.2 that there is a nonlinear relationship between
GAS concentration and resonant frequency response. The antigen frequency response at any time is shown to follow a logarithmic relationship. This is in part due to saturation kinetics typically exhibited by fixed surface sensors. That is, as the antigen binds to the surface, not only the number of binding sites decreases, but also their accessibility. Thus, the binding rate decreases. These factors result in the nonlinear response observed in
Figure 13.2. A plot of the steady state resonant frequency change versus the log of GAS bulk concentration (figure not shown) showed that the steady state resonant frequency change increased in a linear fashion with GAS concentration; in the concentration range investigated. These results suggest that calibration relationships for estimating GAS concentration can be stated as:
(−Δf ) + A log(C ) = ss (13-1) b0 B where A is y-intercept and B is the slope. The parameters A and B will depend on cantilever dimensions, antibody type, and immobilization method.
The cantilever response to a sample containing 7 x 107 cells/mL is shown in Fig. 13.3.
The resonant frequency for the binding of the pathogen to the cantilever functionalized with anti-GAS decreases rapidly and reaches a constant value of approximately 1260 Hz.
After the binding of the pathogen (bulk concentration 7 x 107 cells/mL), the cantilever was immersed in a pH 1.5 HCl/PBS buffer which caused the release of the attached GAS.
However, upon exposure of the cantilever bearing GAS cells to the release buffer the unbinding of the pathogen was not immediate, see Figure 13.3. The release of the pathogen started approximately 40 minutes after exposure to the buffer. It should be 228 noted that after immersing the GAS bearing cantilever into the release buffer the resonant frequency decreases and then remains constant for 40 minutes. This may be due to the low pH of the release buffer, which changes the conformation of the antibody on the cantilever surface causing a decrease in rigidity of the antibody layer through protein unfolding. As a result the cantilever responded with a decrease in its resonant frequency
(typical at pH 1.5). It is also worth noting that the total change in resonant frequency for the binding of the pathogen approximately equals to the total change for the unbinding of the pathogen. This suggests that the mass of GAS that binds to the cantilever was effectively released.
13.3.5 Scanning Electron Micrograph
In order to obtain visual confirmation environmental scanning electron microscope
(ESEM) pictures of the attached cells on glass surface were obtained. Glass cover slips were cut to cantilever dimension and prepared with anti-GAS following the protocol described in Section 13.2.2. The glass surfaces were then exposed to GAS samples (700,
7 x 107, and 7 x 109 cells/mL) for 60 minutes, rinsed with PBS, then with deionized water (to remove residual salt), and finally dried prior to the ESEM investigation. Figures
13.4 (a), (b), and (c) represent the electron micrograph of the GAS samples 700, 7 x 107, and 7 x 109 cells/mL, respectively. One notes that the density of the pathogen increases with sample concentrations. Also it is to be noted that the 700 cells/mL glass sample was significantly magnified because only a few spots containing cells were found. The distribution of attached cells was very non-uniform and no particular pattern was found.
However, the presence of cells on the glass surface at such a low sample concentration
(700 cells/mL) confirms, at least qualitatively, the observed resonant frequency change. 229
13.3.6 Kinetics of pathogen Attachment
The binding kinetics of microorganisms to surface immobilized antibody is not a well characterized area in comparison to molecular adsorption and binding kinetics on solid substrate. There is a great body of literature dealing with molecular absorption. Since attachment of GAS to PEMC surface exhibits an exponential response, similar to surface adsorption, it is reasonable to consider the Langmuir model to characterize the associated kinetics. The kinetic parameters so derived will be useful in design of sensors, and evaluating various sensor configurations. At time = 0, there are no concentration gradients and thus, diffusion effects are absent. Furthermore, the concentration of GAS is known accurately. Here, we considered the general treatment of sensor response to obtain kinetic parameters; Eq. (9-4).
Fitting the frequency response data presented in Figure 13.3 with the model, the
characteristic rate constant kobs can be determined. We include only data obtained during the first ten minutes. The results showed straight lines with excellent correlation coefficients ranging from 0.982 to 0.996, and is illustrated in Figure 13.5. Limiting the initial rate analysis to the first ten minutes appears to be a reasonable approach to obtain
3 5 6 7 kinetic constants. The values of kobs for 700, 7 x 10 , 7 x 10 , 7 x 10 , 7 x 10 , and 7 x
109 cells/mL samples were 0.166 ± 0.01, 0.140 ± 0.03, 0.080 ± 0.02, 0.076 ± 0.01, 0.067
-1 ± 0.02, and 0.051 ± 0.01 min , respectively. These results suggest the kobs dependence on bulk concentration of GAS.
13.4 Conclusion
In this Chapter, we showed that the mass change sensitivity of a piezoelectric-excited millimeter-sized cantilever (PEMC) sensor can be enhanced by two orders of magnitude 230 by reducing its length, and it’s damping coefficient through decreasing the number of layers. Also, the resonant frequency increased by more than 5 kHz. The effects of these design modifications were demonstrated with the detection of pathogen, Group A
Streptococcus. The sensitivity of the modified PEMC sensor is in the order of tens of picograms, and experiments show that successful detection of very low pathogen concentration, 700 cells/mL is possible with the revised geometry. The experimental results also showed that the steady state resonant frequency change depends on the bulk concentration of the pathogen. The resonant frequency change of the second mode at concentrations of 700, 7 x 103, 7 x 105, 7 x 106, 7 x 107, and 7 x 109 cells/mL resulted in, respectively, 3.51 ± 0.51, 11.6 ± 1.3, 15.7 ± 1.1, 25.77 ± 0.17, 28.6 ± 2.2, and 40.5 ± 3.1 ng (n=3 for all) of pathogen attachment. The release of the bound pathogen at pH 1.5 confirmed that the sensor response was indeed due to pathogen attachment. Langmuir first order model fits the observed attachment kinetics and provides a reasonable kinetic description of the sensor response. The observed binding rate constant was found to be in the range of 0.051 to 0.166 min-1. The significance of the results is that the modified
PEMC sensors have high mass sensitivity that pathogens can be detected at very low concentration under liquid immersion conditions. 231
Table 13.1: Resonance characteristics of PEMC sensors in both air and PBS at 1 mm immersion.
11 Cantilever Mode fa Qa σ2a, fnf σ2f, 10 x(g/Hz) kHz] 1011x(g/Hz) [kHz]
1 18 70 ---- 12.5 ----- PEMCn1 2 72.5 38 1.6 64.5 2.27 1 17.5 25 ---- 10.5 ----- PEMCn2 2 53.0 38 50 44.5 72 1 12.8 64 ------PEMCo 2 65.8 33 1000 57.5 1430
Cantilever Mode Qf Me K Co Maf Cv 6 -3 7 10 xKg 10 xN/m Kg/m-s 10 x(kg) Kg/m-s
PEMCn1 1 ------1.01 4.5 80.3 2.99 1.84 2 34
PEMCn2 1 ------2.02 5.99 11.57 4.19 4.64 2 30
PEMCo 1 ------3.71 16.6 372.6 9.35 6.44 2 32 232
3.5E-08 3.0E-06
PEMCo- 1.47E-08 g/Hz 3.0E-08 2.5E-06
2.5E-08 2.0E-06 2.0E-08 1.5E-06 mass change mass change [g] PEMCn1 - 5.02E-11 g/Hz o
n1 1.5E-08 1.0E-06 1.0E-08 PEMC PEMC
5.0E-09 5.0E-07
0.0E+00 0.0E+00 0 100 200 300 400 500 600 700
Resonance frequency change (-Δf ) [Hz]
Figure 13.1: Experimental determination of mass change sensitivity of PEMCn1 and -11 -8 PEMCo. The mass change sensitivities were 5.02 x 10 and 1.47 x 10 g/Hz for PEMCn1 and PEMCo, respectively. The first standard deviation of resonant frequency changes and mass additions were ± 10 Hz and ± 0.12 μ g for PEMCo, and ± 5 Hz and ± 0.5 ng for PEMCn1, respectively. 233
200 Control 0
700cells/mL -200
-400 7x103cells/mL -600 7x105cells/mL -800
-1000 7x106cells/mL -1200 7x107cells/mL -1400
Resonant frequency change [Hz] -1600
9 -1800 7x10 cells/mL
-2000
0 102030405060
Time [min]
Figure 13.2: Resonant frequency responses for the binding of various Group A
Streptococcus (GAS) concentrations to anti-GAS functionalized PEMCn1 sensor. The results showed that the binding rate strongly depends on concentration. The control was an antibody-immobilized cantilever immersed in PBS. 234
0
-200 Release of GAS using HCl/PBS pH 1.5
Binding of 7x107 -400 cells/mL to PEMC
-600
-800
-1000
-1200
Resonant frequency change [Hz] -1400
-1600 0 20406080100120 Time [min]
Figure 13.3: Resonant frequency response to the sequential binding and release of 7 x 107 cells/mL sample of Group A Streptococcus (GAS). 235
A
B
C
Figure 13.4: ESEM micrograph of antibody functionalized glass cover slip (same material used to fabricate PEMCs) after one hour exposure to various concentrations of Group A Streptococcus. Panel A: 700 cells/mL, Panel B: 7 x 107 cells/mL, and Panel C: 7 x 109 cells/mL. After binding, the cover slips were rinsed in deionized water, dried and analyzed under an ESEM. The cell density increased with pathogen concentration. 236
0.0
-0.2 -0.4
) -0.6 ∞ f Δ f)/ -0.8 Δ -
∞ 7E09 cells/mL f Δ -1.0 7E07 cells/mL
ln(( 7E06 cells/mL -1.2 7E05 cells/mL -1.4 7E03 cells/mL 700cells/mL -1.6
-1.8
0246810
Time, τ [min]
Figure 13.5: Initial kinetic analysis for the binding of Group A Streptococcus. Correlation coefficient ranged from 0.982 to 0.996. The slope of each line gives the observed characteristic binding rate constant, kobs.
237
Chapter 14: Detection of Bacillus anthracis Spores and a model protein Using PEMC Sensors in a Flow Cell at 1 mL/min
14.1 Introduction
In Chapters 10 and 13, we reported that detection sensitivity of piezoelectric-excited millimeter-sized cantilever (PEMC) sensors for pathogens and proteins is on the order of
20-50 pg/Hz. We also showed that E. coli O157:H7 can be detected using antibody immobilized PEMC sensors at such low concentration as 70 cells/mL in batch of 1 mL samples. We also showed that spores of Bacillus anthracis can be detected at 300 spores/mL in a batch cell using PEMC sensors, Chapter 9. While batch measurement is effective when sample volume is small, it is not effective for applications where sample volume is large and the pathogen count is very low. An example application is the case of drinking water and source water analysis for pathogen content. The tolerable level for
Cryptosporidium and Giardia is 0.2 per liter of drinking water282. Due to the size of the sensor (a few mm2), dipping it into a large volume of liquid is not an effective means of contacting the target analyte with the sensor surface, particularly particulate antigen such as the spores. Hence, contacting in a flow cell configuration is a natural next step of development. Hence, investigating the effects of flow rate on sensor performance, and the design configurations of flow cell for PEMC sensor that yields good sensor performance are worthwhile goals from a practical perspective. In this Chapter, we address these two issues.
Sensors that rely on mechanical resonance such as quartz crystal microbalance240,282,285,286, microcantilevers287, and plasma resonance (SPR) 229,288-290 have been used in the flow cell configuration. Typically the flow rates are in the order of 238
μL/min. For example, the recommended sample liquid flow rate for use in instrument grade quartz crystal microbalance (QCM) is 0.2 μL/min. Others, for example Pei et al.,
(2004) who used microcantilevers used low flow rates of 0.033 mL/min287. In general, performance of both QCM and microcantilevers deteriorate at high flow rates such as
100 μL/min (Sota et al., 2002). Alternate flow cell design was proposed by Sota et al.,
(2002) that enabled a 27 MHz QCM to operate satisfactorily at 100 μL/min.
In this investigation, we show that the PEMC sensor performance does not deteriorate at flow rates as high as 17 mL/min, which is approximately 200-fold higher than any of the cantilevers sensors.
14.2 Materials and Methods
14.2.1 Chemicals
All chemicals were purchased from Sigma-Aldrich (Allentown, PA). Deionized water used was from a Milli-Q plus ultra-pure water system (18.2 MΩcm).
14.2.2 Flow cell fabrication
Several sensor flow cells were designed, fabricated, and tested. In this Chapter we report on two sensor flow cells, labeled SFC-1 and SFC-2, which showed good promise at high flow rates. The flow cells were constructed of Plexiglas® and dimensional details are given in Figure 14.1. The central contacting chamber is cylindrical in shape of 7.0 mm diameter. Once PEMC sensor is installed the hold up volume was 500 μL and 300
μL in SFC-1 and SFC-2, respectively. The only difference between the two flow cells
(SFC-1 and SFC-2) is the position of the inlet and the outlet. In SFC-1 both the inlet and outlet are positioned horizontally at the lowest point on the sides of the cell, see Figure 239
14.1A. In SFC-2 the inlet and outlet are located at the bottom and on the side of the cell, respectively. The outlet (1.59 mm diameter) was located 4 mm above the inlet (also 1.59 mm diameter), see cross sectional view of SFC-2 Figure 14.1B. Constant temperature water (35 ± 0.1 °C) was circulated (17 mL/min) through a shell (not shown in figure) surrounding the flow cell. The shell (4 mm wide) was 2 mm from the cell inner surface.
At steady state, the temperature within the flow cell varied by ± 0.2 °C in the sample flow range of 1 to 17 mL/min.
14.2.3 PEMC sensor
PEMC sensors were fabricated and were immobilized with antibody to Bacillus anthracis (BA) spore (kindly provided by Dr. Richard Rest, Drexel University College of
Medicine) or Bovine serum albumin (BSA) as per Chapter 4.
14.2.4 Experimental setup
The experimental setup consisted of fluid reservoirs, peristaltic pumps, and a sensor flow cell (SFC-2) as shown in Figure 14.2. The apparatus consisted of five fluid reservoirs: one for PBS buffer pH 7.4, activated antibody solution, 10mM hydroxylamine solution, antigen sample, and release buffer (PBS/HCl; pH 2.0). The liquid reservoirs were connected to the flow cell via a five entrance port manifold, and the inflow enters the SFC through the bottom. The outlet of the flow cell is connected to a peristaltic pump, which controls the flow of the desired fluid into and out of the SFC. The APTES functionalized PEMC sensor was installed vertically into the cell filled with PBS and was secured in place. The cantilever electrodes were connected to an impedance analyzer
(Agilent, HP 4192A) interfaced to a data acquisition PC with LabVIEW application for obtaining impedance and phase angle measurements in the frequency range of 1 to 200 240 kHz with an excitation voltage of 100 mV. The constant temperature bath was set at 35 ±
0.1 °C in order to maintain the cell content at 25 °C. The fluidic system was first primed with PBS to remove any air bubbles. Valves located at the bottom of each fluid reservoir enabled selection of fluid for introduction into the flow cell or for circulation. Switching the outlet line from the peristaltic pump into the desired fluid reservoirs enabled total recirculation. All valves were manipulated manually. All detection experiments were carried out at a flow rate of 1 mL/min.
14.2.5 Antibody immobilization in situ and antigen Detection
The sensor was functionalized with 3-aminopropyl-triethoxysilane, and was installed in the SFC containing PBS. The resonant frequency of the fundamental mode was monitored until it stabilized. Then, the activated antibody solution was flowed into the cell (by opening valve V2) after which the flow was stopped to allow antibody immobilization under stagnant conditions for 3 h at 25 ± 0.1 °C. Hydroxylamine was then flowed through the cell (by opening V3), and then PBS (by opening valve V1 and closing
V3) to rinse out the lines. Upon completion of the rinsing step, the antigen solution was introduced into the sensor flow cell by opening V4 and closing V1. After the antigen attachment was concluded V4 was closed, and the cell was rinsed with PBS (by opening
V1) followed by flowing the release buffer (open V5 and close V1).
14.3 Modeling fluid flow
The flow patterns through the sensor flow cell were investigated by modeling the flow using finite element modeling platform, FEMLAB®. The two-dimensional Navier
Strokes equation (ρ = 1000 kg/m3 and μ = 0.001 Pa.s) was solved in conjunction with continuity equation for various inlet flow rates (1 to 17 mL/min). The dimensions of the 241 sensor flow cell used in the model were identical in shape and dimension to the actual ones used in the experiments within ± 0.05 mm.
14.4 Results and Discussion
14.4.1 Resonance characterization of PEMC sensors
Typical resonance spectra, a plot of phase angle and impedance versus excitation frequency, in air of PEMC sensor may be found in previous Chapters. For the PEMC sensor investigated the fundamental and second mode resonant frequencies were at 23 ±
1 kHz and 91 ± 2 kHz in air, respectively. Several repeat experiments showed that these resonant frequencies are stable within ± 10 Hz in any one particular experiment. In this study, the fundamental mode was used for detection because the peak was very stable and remained sharp under various flow rates. Each experiment was repeated at least twice and the data shown are typical of the results obtained.The sharpness of the peak is characterized by its quality factor (Q-factor). Typical Q-factor values ranged from 30 to
100, and do not deteriorate significantly upon water immersion. Our experience is that
PEMC sensors lose Q-value in the range of 10 to 20% under liquid immersed flow conditions. The Q-factor of the fundamental mode in air was 50, and in PBS it was 42.
14.4.2 Effect of flow rates on resonant frequency
In Figure 14.3 the response of PEMC sensor to flow rate changes are presented. In
Panel A, we show a flow sensitivity study that was conducted in SFC-1 with the sensor in two different configurations; namely, the width of the sensor (1 mm) was positioned parallel or perpendicular to the inflow. The flow rate was systematically varied from 1 to
17 mL/min in steps of 1 mL/min. The resonant frequency decreased monotonically, and the parallel or perpendicular positioning of the PEMC sensor did not appear to have 242 significant effects on the direction or magnitude of change in resonant frequency. The resonant frequency decreased linearly at a rate of 22 Hz/mL/min and appeared to decrease further beyond 17 mL/min; higher flow rates were not investigated due to the pumping limit of the peristaltic pump. When the sensor was placed parallel to the inflow, the decrease in resonant frequency with flow rate was linear up to 10 mL/min and further flow rate increase resulted in small changes in resonant frequency. One notes from Figure
14.3A the fluctuation of resonant frequency at any particular flow rate was small, and was in the range of 8 to 15 Hz over the entire flow rate investigated.
Because SFC-1 showed continual change in resonant frequency with flow rate, we searched for other flow cell geometries that would show low or no sensitivity to flow rate. Several geometries were explored and we found that SCF-2 geometry, given in
Figure 14.1B, gave the desired response. A natural variation on SFC-1 design was to introduce the inflow from the bottom of the flow cell. Unlike SFC-1, as the flow rate was initiated the resonant frequency increased initially in a step-like manner and then fluctuated around an average resonant frequency at each of the flow rates investigated in
1 to 17 mL/min. At each flow rate, the flow was kept constant for 20 minutes and the average resonant frequency value and the corresponding standard deviation during the 20 minutes time period are presented in Figure 14.3B. Note that the maximum and minimum variations at any particular flow rate were ± 7 and ± 20 Hz, respectively. This is slightly higher than the value we found under stagnant conditions, ± 5 Hz, but adequate for the sensing experiments. Over the entire range of flow rate investigated, 1 to 17 mL/min, the mean resonant frequency change was 217 Hz with a standard deviation of ± 20 Hz. These repeatable variations are believed to be characteristics of the flow geometry of SFC-2. It 243 is worth noting that the fluctuations in frequency response at any particular flow rate were lower in SFC-1. In the following section we have attempted to determine the reason for the difference in performance through simulation.
We believe that the increase in resonant frequency upon initiation of flow in SFC-2 is due to pressure change in the cell, and characterization of this phenomenon is currently under study. It is useful to obtain pressure map and velocity profile of SFC in order to compare their sensor interaction characteristics.
14.4.3 Pressure map and velocity field
The pressure map and velocity field in the flow cells at the various flow rates were investigated, and the results for 1 and 17 mL/min are presented in Figure 14.4. The figure shows surface plots of pressure (color coded), and the streamline of the velocity field.
Panel A shows the flow pattern at 1 and 17 mL/min, respectively, in SFC-1 for the case where the width (1 mm) of the sensor was positioned perpendicular to the inflow. Even at a low flow rate (1 mL/min) the flow map showed a small re-circulation region adjacent to the sensor on the inlet side. As the flow rate was increased to 17 mL/min the re- circulating domain increased progressively on the opposite side of the sensor, see Figure
14.4A. In Panel B, the flow map with sensor’s width positioned parallel to the inflow is shown. Unlike the perpendicular case, two re-circulation regions developed on both sides of the sensor even at low flow rate of 1 mL/min. As the flow rate was increased, the re- circulating region decreased on the inflow side, and the one on the outlet side increased significantly.
Unlike SFC-1, the flows in SFC-2 showed no increase or decrease in the size of the re-circulating region in the flow rate range investigated; see Figure 14.4C. However, as 244 the flow rate increased more of the inflow appeared to contact the sensor as indicated by the smaller streamline bandwidth at higher flow rate.
In order to determine the sensor response to flow rate, the pressure drop across the sensor was compared at various flow rates, and the results are plotted in Figure 14.5.
Note that the pressure drop across the sensor in SFC-1 was not affected significantly by the position of the sensor, and the pressure drop increased as flow rate was increased.
Furthermore, the pressure map shows higher pressure acting somewhat uniformly on the inflow side of the sensor; see color bar. In SFC-2 the pressure drop across the sensor increased slightly with flow rate, which may explain the observed fluctuations in resonant frequency around 217 Hz (Figure 14.3A) at higher flow rate.
14.4.4 Detection of Bacillus anthracis spore in SFC-2
Model pathogen Bacillus anthracis (BA) spores at 300 spores/mL was used to evaluate performance of PEMC sensor in SFC-2 design. In Figure 14.6 the change in the fundamental resonant frequency as a function of time for the capture of BA spore in solution is presented. Figure 14.6A shows the detection sequence including sensor preparation; namely reaction of sulfo-NHS activated anti-BA with the sensor surface amine groups, the binding of BA spores, and finally the release of the bound pathogen.
The anti-BA immobilization was carried out by filling SFC-2 with PBS containing anti-
BA at 10 μg/mL in stagnant conditions for approximately 3 h. The reaction (amide bond formation) of the activated anti-BA to the aminated surface began immediately resulting in a decrease in the resonant frequency, and ultimately reaching a steady state frequency change of 420 ± 3 Hz. In Figure 14.6 we show the resonant frequency response since the start of the experiment. Hydroxylamine was then flowed through the cell at 1 mL/min to 245 convert activated carboxylic groups on anti-BA back to normal carboxylic groups. It can be seen in Figure 14.6A that the initiation of hydroxylamine flow resulted in a step increase of 122 Hz and reached a value of -298 ± 3 Hz, with respect to the initial value:
See Figure 14.6B. This step response is similar to what was observed in Figure 14.3A.
Note that immediately after the initial rapid rise the resonant frequency decreased and reached a steady state value and remained constant, within ± 3 Hz until the flow of BA spores solution was initiated.
Exposure to BA spores caused a rapid decrease in the resonant frequency. The total frequency change was 160 ± 3 Hz and reached steady state in 20 minutes, which is similar to the time observed in the study done in Chapter 9. Following the conclusion of
BA spore binding, PBS was flowed at 1 mL/min for 10 minutes to rinse out SCF-2 and the tubing connections. During the PBS rinse, residues of the previous step and the weakly adsorbed antibody were removed. The change in resonant frequency was very small and was within the noise level (± 5 Hz). It is likely that the sensor does not respond to the weakly bound molecules. The exposure of the sensor to release buffer (pH 2.0) caused an immediate increase of the resonant frequency and reached a constant value
(300 ± 6 Hz), which was within 2 Hz of the resonant frequency value prior to BA spore flow. These responses suggest that all the bound BA spores were released by the release buffer. In this study, attempt to quantify the released spores was not pursued, nor did we characterize the regenerated sensor surface. Both of these issues are important for sensor development and are currently being investigated.
14.4.5 Effect of flow rate on PEMC sensor response
The total change in the measured resonant frequency due to BA spore binding 246 depends on the amount of antibody that is attached to the sensor and the accessibility of spores to the Fab region of the antibody. As a first step, we were interested in determining if the immobilization step itself depended on flow rate. Immobilization under flow conditions consumes 3 to 4 time larger amount of antibody and from a cost perspective it may not be desirable. However, if higher effective surface concentration is achieved higher sensitivity to BA spore detection is a likely outcome and is desirable.
Figure 14.7 shows the frequency change of the measured resonant peak for the attachment of anti-BA and the corresponding binding of BA spores both under stagnant and flow conditions. The measured resonant frequency change for the reaction of anti-BA
(10 μg/mL) to the aminated sensor surface is presented in Figure 14.7A. In stagnant conditions the steady state value of the measured resonant frequency change (230 Hz) was lower than that observed under 1 mL/min flow condition (430 Hz). Clearly, flow enhances transport of activated antibody to the sensor surface resulting in a higher level of binding. Also, the rate of frequency change was higher under flow conditions suggesting higher anti-BA immobilization on sensor surface.
In Figure 14.7B, we compare the stagnant and flow configuration results for the binding of BA spores (300 spores/mL) to the sensor surface in SFC-2. We find that the effect of flow on BA attachment can be characterized from three observed phenomena (1) binding kinetics or rate of change of resonant frequency response, (2) the total binding of spores or total change in resonant frequency, and (3) the response time or time to reach steady state. Both binding kinetics and total binding of spores increased under flow conditions. However, time of response appear to remain unaffected. For example, it took approximately 20 minutes to reach steady state under both flow and stagnant 247 measurements of BA binding. The initial rate at which the resonant frequency changed was about 40% higher under flow suggesting an increase in spore attachment kinetics.
Also, the total change in resonant frequency increased by approximately 100% compared to stagnant conditions, and can be seen in Figure 14.7B. One key observation of the results presented here is that the steady state frequency values for both anti-BA and the corresponding BA spore binding were approximately 100% higher under flow compared to stagnant conditions.
14.4.6 Protein detection and sensor regeneration
In order to examine the influence of flow with regard to detection of protein molecules we chose a model protein, bovine serum albumin (BSA). BSA was sequentially attached and released as shown in Figure 14.8. The sensor was initially exposed to anti-BSA (100 μg/mL) to immobilize the sensing molecule. The fundamental resonant frequency (23 kHz in air) was monitored for 2 h under stagnant conditions and the results in Figure 14.8 (labeled anti-BSA) shows that the resonant frequency decreased by approximately 900 ± 5 Hz. This value is higher in comparison to anti-BA, and is due to the higher concentration of anti-BSA used. Initiation of flow (1 mL/min) resulted in a step increase of resonant frequency. Flowing of solution containing BSA
(labeled as BSA-1 in Figure 14.8) caused a rapid initial decrease followed by an exponential decrease, and finally reached a steady state resonant frequency value of 396
± 10 Hz. The release of the bound BSA was carried out with release buffer (PBS/HCl, pH 2.0), labeled as HCl-1 in Figure 14.8. Exposure of the BSA-bearing sensor to release buffer caused an immediate rise in the resonant frequency and the frequency reached a constant value at the same value that was present prior to BSA exposure. This suggests 248 that the mass of BSA that was bound to the sensor was removed successfully without apparently destroying the underlying antibody layer. As a result the sensor was exposed to BSA again, labeled as BSA-2 in Figure 14.8. The resonant frequency decreased and reached a steady state value of 1137 ± 5 Hz, with total change in frequency of 393 ± 5
Hz, indicating that the antibody was fully regenerated without any loss of antibody activity. However, regenerating the surface the second time (labeled as HCl-2 in Figure
14.8) resulted in a higher resonant frequency change than was observed during the first regeneration suggesting that some of the antibody may have been released. The release profile, HCl-2, is also slightly different from HCl-1. In light of these results it is reasonable to suggest that PEMC sensors can be re-used a limited number of times.
Clearly, the composition and pH of release buffer will influence the number of reuses achievable.
14.4.7 Kinetics of spore binding
The effect of flow on the kinetics of binding is best quantitatively determined using
⎛ ⎛ (Δf )− (Δf )⎞ ⎞ the approach we reported in Chapter 9, Eq. (9-4) ⎜ln⎜ ∞ ⎟ = −k τ ⎟ . The model ⎜ ⎜ ⎟ obs ⎟ ⎝ ⎝ ()Δf ∞ ⎠ ⎠
was developed based on the assumption that the bulk concentration Cb0 remains constant
during a detection episode. The binding rate constant kobs can be determined from a plot of the left hand side (LHS) versusτ . The kinetic approach we employ here is the initial
rate analysis of the frequency response data to obtain kobs . This approach is particularly important when one deals with low concentration analyte. Furthermore, the bulk concentration of antibody and antigen are known accurately. Therefore, limiting the initial rate analysis to the first ten minutes is a reasonable approach. 249
As discussed in one of the earlier Chapter, antibody reaction with primary amines on glass results in covalent bonds and therefore, the binding can be modeled as an irreversible reaction. The exponential decay of the measured resonant frequency change in both the stagnant and flow measurements of the antibody-amine reaction provides useful kinetics information (see Figure 14.7A). Fitting the sensor response presented in
Figures 14.6A to Eq. (9-4) gave straight lines with excellent correlation coefficients of
0.96 and 0.92 for both the stagnant and flow methods, respectively, and is illustrated in
Figure 14.9A. The slope of each line gives the observed characteristic binding rate constant, kobs. The binding rate constant for anti-BA under both stagnant and flow conditions were 0.042 ± 0.008 min-1 and 0.213 ± 0.05 min-1, respectively. Here, it is clear from kobs values that flow causes approximately a five-fold increase in the characteristic observed rate constant. The higher binding rate observed in the flow modality may be explained by enhanced transport of anti-BA to the sensor surface, as it is well known that mass transfer coefficient increase as velocity is increased.
The experimental data of BA spore were analyzed in the same fashion as anti-BA (see
Figure 14.9B) and a summary of the analysis is given in Table 14.1. Again, the
characteristic sensor response rate constant, kobs , is higher in the flow measurement modality. The observed binding rate constants were 0.195 ± 0.02 min-1 and 0.263 ± 0.05 min-1, respectively, for stagnant and flow. That is, the kinetic rate analysis suggests that binding rate of the pathogen was 35% higher under flow conditions. Thus, at a constant signal-to-noise ratio, the higher kinetics of attachment under flow conditions can potentially enhance detection time as well as detection limit, and is currently under study.
The binding kinetics of protein to surfaces is a key parameter in many sensor 250 applications. However, the regeneration of the antibody surface is of great interest since it tells the status of the bound antibody. Fitting the frequency response data for BSA binding presented in Figure 14.8 (initial binding (BSA-1) and after regeneration (BSA-2))
-1 to Eq. (9-4) the observed rate constant, kobs was determined to be 0.111 ± 0.006 min and 0.103 ± 0.005 min-1, respectively. See kinetic analysis in Figure 14.9C. These results suggest that there was only a small loss in the antibody activity after the first surface regeneration with release buffer pH 2.0.
14.5 Conclusion
In this investigation we have shown that a flow cell system comprised of a PEMC sensor functionalized with the appropriate antibody can be used to detect low concentrations of proteins and pathogens in real time. Detection of pathogen (Bacillus anthracis (BA) at 300 spores/mL) and model BSA (1mg/mL) were investigated under both stagnant and flow conditions. Two flow cell designs were evaluated by characterizing flow-induced resonant frequency shifts. One of the flow cells labeled SFC-
2 (hold-up volume of 0.3 mL), showed small fluctuations (± 20 Hz) around a common resonant frequency response of 217 Hz in the flow rate range of 1 to 17 mL/min. The total resonant frequency change obtained for the binding of 300 spores/mL in 1 h was 90
± 5 (n=2), and 162 ± 10 (n=2) Hz under stagnant and flow conditions, respectively.
Binding of antibodies were more rapid under flow than under stagnant conditions. The sensor was repeatedly exposed to BSA with an intermediate release step. The first and second response to BSA was nearly identical. The total resonant frequency response to the model protein (BSA) was 388 ± 10 (n=2) Hz under flow conditions. The most significant result is that sensor response is not only more rapid for spore detection, but 251 also for protein binding. Furthermore, the total sensor response was significantly higher
(~ 100 %) for both Bacillus anthracis spores and protein (BSA) under 1 mL/min flow conditions compared to stagnant condition. 252
Table 14.1: k obs values of anti-BA, BA, and BSA under stagnant and flow conditions.
k values in stagnant k values in flow Sample obs obs -1 -1 (min ) (min ) Anti-BA 0.042±0.008 0.213±0.05
BA 0.195±0.02 0.263±0.05
BSA-1 ------0.111±0.006
BSA-2 ------0.103±0.005
253
PEMC sensor casing ID 6.2 mm A O-rings
SFC-1 ID 7mm 6 mm
Inlet Outlet 1/16˝ fitting 1/16˝ fitting
3 mm
PEMC sensor casing B ID 6.2 mm
O-rings
SFC-2 Outlet ID 7mm ID 1.4 mm
6 mm 3 mm Inlet ID 1.4 mm
Figure 14.1: Schematic illustration of sensor flow cell (SFC) design. Panel A: Cross sectional view of SFC-1. Panel B: Cross sectional view of SFC-2. The cells are 7 mm in diameter. SFCs 1 and 2 have hold up volumes of 500 and 300 μL, respectively, after sensor insertion. 254
Circulation tube
PBS Antibody Hydroxylamine Antigen Release buffer syringe syringe syringe syringe syringe
On/Off valves Peristaltic pump #2 V1 V2 V3 V4 V5 Sample outlet Sample
5-port manifold
Heating fluid inlet
Heating fluid outlet
Sample inlet
Peristaltic pump #1
Constant temperature bath
Figure 14.2: Flow circuit of experimental apparatus. Valves V1, V2, V3, V4, and V5 are on-off values. 255
0 A
-100
-200 [Hz]
-300 Flow perpendicular to width Flow parallel to width
Average resonant frequency change -400 0 5 10 15 20 Flowrate [mL/min]
300
250 A 217 ± 20 Hz 200
150 [Hz]
100 B 50 Average resonant frequency change 0 024681012141618
Flow rate [mL/min]
Figure 14.3: Flow rate dependence of resonant frequency. Panel A: SFC-1. Resonant
frequency was not affected by sensor orientation (parallel or perpendicular) to inflow. Panel B: SFC-2. Resonant frequency fluctuated within 7 to 20 Hz, after the initial step response. 256
A
1 mL/min 17 mL/min
B
1 mL/min 17 mL/min
C
1 mL/min 17 mL/min
Figure 14.4: Pressure map (color) and velocity map (lines) of SFC-1 and SFC-2. Panel A: SFC-1, the PEMC sensor is positioned with its width (1 mm) perpendicular to inflow. Panel B: SFC-1, the width of the sensor is placed parallel to inflow. Panel C: SFC-2. Each panel pressure and flow maps at flow rates 1 and 17 mL/min. 257
1.6E-04
1.4E-04
1.2E-04 SFC-1 perpendicular 1.0E-04 SFC-1 parallel SFC-2 8.0E-05
6.0E-05
4.0E-05
2.0E-05 Pressure drop across sensor [Pa]
0.0E+00 0 3 6 9 12 15 18
Flow rate [mL/min]
Figure 14.5: The dependence of pressure drop across the sensor to increasing flow rate of the sample solution in the various sensor flow cells, and configurations. 258
0 A anti-spores binding Spore binding -100 Spore unbinding
-200 PBS
-300
-400 Resonant frequencychange [Hz]
-500 0 100 200 300 400 500 600 Time [min]
-250 Hydroxylamine
PBS B -300
-350
Anti-BA BA spores (300 spores/mL) -400
-450 Resonant frequency change [Hz] change frequency Resonant -500 180 190 200 210 220 230 240 250 260 270 280 Time [min]
Figure 14.6: The measured resonant frequency change of PEMC sensor versus time in SFC-2. Panel A: Shows the resonant frequency response to Bacillus anthracis (BA) spores at 300 spores/mL, and the release of the bound spores. The antibody was immobilized in stagnant conditions, and spore attachment was at 1mL/min. Panel B: The exponential decrease in resonant frequency immediately after BA spores exposure. An exponential decrease in resonant frequency was observed immediately after BA spores exposure. 259
0 -50 -100 A -150 Stagnant, anti-BA 10 μg/mL -200 -250 -300 -350 Flow, anti-BA 10 μg/mL
Resonant frequency change [Hz] -400 -450 0 20406080100 Time [min]
B 20 Control 0 -20 -40 -60 Stagnant, 300 BA spores/mL -80 -100 -120
-140 Flow, 300 BA spores/mL Resonant frequency change [Hz] -160 -180 0 102030405060 Time [min]
Figure 14.7: Comparison of resonant frequency response under stagnant and flow conditions in SFC-2. Panel A: Antibody to Bacillus anthracis spores (0.01 mg/mL anti-spore) reacts with aminated PEMC sensor surface. Panel B: The binding of Bacillus anthracis spores, 300 spores/mL, to PEMC sensor. In the flow experiments, the sample flow rate was 1 mL/min. The control shown is the response of an antibody-functionalized cantilever sensor in SFC-2 with PBS at 1 mL/min. 260
0 anti-BSA BSA-1
-200 BSA-2 PBS-5 PBS-1 HCl-1 HCl-2 -400 PBS-3 PBS-2 -600 PBS- 4
-800
-1000 Resonant frequency change [Hz] change frequency Resonant
-1200
Hydroxylamine -1400 0 100 200 300 400 500 600 Time [min]
Figure 14.8: Resonant frequency response in the detection of a model protein, bovine serum albumin (BSA) in SFC-2 at 1 mL/min. Sequential binding and release of BSA from the sensor. The label indicates time of introduction and the number attached to the label represents the number of times the solution was introduced to the cell. For example, “PBS-4” means the fourth time PBS was pumped into the flow cell. 261 0.0
-0.4 Stagnant ) ? f
Δ
f)/ -0.8 Δ -
? Flow f
Δ -1.2 A ln((
-1.6 02468
Time, τ [min]
0.0 -0.5
) Stagnant
? -1.0 f
/Δ -1.5 f) Δ -2.0 - Flow ? f -2.5 B Δ -3.0 ln(( 0246810 Time, τ [min]
0.0
-0.2 BSA-1
) -0.4 BSA-2 ? f Δ -0.6 f)/
Δ - -0.8 ? f
Δ -1.0 C ln(( -1.2 0246810 Time, τ [min]
Figure 14.9: Initial binding rate analysis. Panel A: Anti-BA to the aminated PEMC sensor surface under both stagnant and flow conditions. Panel B: Bacillus anthracis (BA) spores (300 spores/mL) to the anti-BA derivatized sensor in both flow and stagnant conditions. Panel C: Bovine serum albumin (BSA) to an anti-BSA functionalized sensor in flow configuration and in one cycle of regeneration. The slope of each line gives the observed characteristic binding rate kobs. 262
Chapter 15: Detection of Escherichia coli O157:H7 in Ground Beef Samples using Piezoelectric Excited Millimeter-Sized Cantilever (PEMC) Sensors
15.1 Introduction
Escherichia coli (E. coli) O157:H7 is an enteropathogenic bacteria and the etiological agent of hemorrhagic colitis. E. coli O157:H7 is an epidemiologically significant cause of food-borne disease originating from contaminated ground beef, milk, juice, pork, lamb, and poultry products291-295. The organism is found in the intestine of animals, in particular cattle, and can be a causative microbe of extra-intestinal infections such as the life threatening hemolytic uremic syndrome296-298. The outbreaks of E. coli O157:H7 food poisoning in the USA over the past few decades299,300 and the sporadic worldwide outbreaks caused by contaminated ground beef such as hamburgers301 have raised growing interest in rapid pathogen identification on ground beef302-304 and beef carcasses304.
The United States Department of Agriculture has taken several actions to eradicate E. coli O157:H7 from raw meat and meat products. For example, a spot-checking program for the identification of E. coli O157:H7 on ground beef is carried out. This program is based on the Hazard Analysis and Critical Control Point (HACCP) guidelines, which analyzes critical point along a manufacturing process for contaminants and implements procedures to eliminate the contaminants305. The US Food Safety Inspection Service has established a zero tolerance threshold for E. coli O157:H7 contamination of raw meat products166. The infectious dosage of E. coli O157:H7 is 10 cells, and the Federal EPA standard in water is 40 cells per liter 167,168. Several detection techniques capable of providing reliable identification of E. coli O157:H7 at low cell concentrations on ground 263 beef and beef products are currently under development. These detection processes include but are not limited to the use of pulsed-field gel electrophoresis302, immuno- magnetic beads303,306,307, epifluorescence microscopy308, time-resolved fluoroimmunoassay294, and light addressable potentiometric sensor (LAPS)295. These techniques require pre-enrichment of the beef sample, since the number of pathogenic bacteria present on the meat is very low. While these methods are good in identifying the pathogen, they fall short in quantification. In this Chapter, we introduced a new modality of measurement in which both rapid, reliable detection and semi-quantitative estimation of E. coli O157:H7 can be done in a single measurement using the piezoelectric-excited millimeter-sized cantilever (PEMC) sensor.
15.2 Materials and Methods
15.2.1 Chemicals
Concentrated sulfuric acid (H2SO4), methanol, hydrochloric acid, sodium hydroxide,
3-aminopropyl-triethoxysilane (APTES), 1-ethyl-3-(3-dimethylaminopropyl)- carbodiimide (EDC), sulfo-N-hydroxysuccinimide (sulfo-NHS), phosphate buffered saline (PBS), hydroxylamine, and 2-mercaptoethanol. Deionized water used was from a
Milli-Q plus ultra-pure water system (18.2 MΩ cm).
15.2.2 PEMC Preparation
PEMC sensors were fabricated from 5 x 1 mm2 (length x width) PZT and silica layer.
Both layers were bond such that the cantilevers free end as 1 mm PZT and 3 mm of glass sticking out. The protruded PZT layer was coated with a thin layer of polyurethane (~20
μm) to protect it from short circuitry. The glass surface was cleaned, aminated, and functionalized with affinity purified polyclonal antibody to E. coli O157:H7. 264
15.2.3 Sample preparation
E. coli O157:H7 strain B1409 was used in all experiments. A mother culture of it was prepared by growing it in Brain-Heart Infusion media for 18 h at 37 C and 160rpm.
Cell count was determined as follows. Serially diluted bacterial suspensions were placed on a Petroff-Hauser counting chamber slide (Hauser Scientific Partnership, Horsham,
PA) with a center square of 0.2 x 0.2mm that was further divided into 16 squares.
Bacteria were counted in 5 of these squares randomly chosen. The average of these counts was used to estimate total cell concentration.
Samples containing the pathogen, E. coli O157:H7 strain B1409, were prepared by three methods. These are: (A) in growth media, (B) in raw beef prepared in Stomacher bags, and (C) in sterilized beef in Stomacher bags. For ease of reference these samples are labeled, respectively, (0, 2, 4 or 6) – (B, M or Mx) – (1 or 2). The first label, (0, 2, 4 or 6) refers to the time of sample collection in hours. The designation, (B,M or Mx) refers to broth, Stomacher raw beef, and Stomacher sterilized beef samples, respectively.
The last label, 1 or 2 refers to the sample number. Typically two samples were taken at each time, in each experiment. In order to develop a scope of experiments to be performed, we evaluated the PEMC sensor initially with two beef samples, with and without E. coli O157:H7 inoculation. In the first, 25 EC cells were mixed with 25 g of raw hamburger and then mixed with EC growth medium and stirred in a 250 mL flask.
In the second the same sample was agitated in a Stomacher bag equipped with 3-micron filter. Samples were drawn at 2, 4, and 6 h for analysis. These samples are designated as flask and bag samples with the appropriate time designation. Reference samples were 265 prepared in the same fashion as above without the addition of E. coli O157:H7.
Samples for the desired experiments were prepared as follows: E. coli O157:H7
(strain B1409) cells were added to 100mL of broth with novobiocin (20 mg/L) in a
Stomacher 400 bag containing a 3μm filter (Seward Lab Systems, Norfolk, England) with or without 25g of raw or sterilized hamburger. The bags were pummeled for 120 seconds on ‘normal’ speed. Bags were then placed in a 37°C incubator and set to oscillate at 160 rpm. Aliquots of filtered bacterial suspensions were drawn off at 0, 2, 4 and 6 h and immediately placed on ice. Samples that contained no beef were labeled as
“B” and those that contained beef as “M”. When sterilized beef was used, the sample was designated as “Mx”. In each one of the experiments above, reference samples were obtained by following the same procedure without adding E. coli O157:H7.
15.2.4 Plating and counting of bacteria
Using a 96-well microplate, aliquots of bacterial suspension in volumes of 200μL were added to the first well of each row in a 96 well plate (8 rows by 12 columns). To the next 7 wells in a row, 270 μL of 0.1M Tris Buffered Saline pH 7.5 (TBS) was added.
Then aliquots of 30μL was drawn off from wells in the 1st column and added to the adjacent wells in the 2nd column to achieve 10 folds dilution in bacterial concentration.
The 10-fold dilution process was repeated for the remaining wells of the 8th column.
Using a 6-channel multi-pipetter, aliquots of 70μL were drawn from the wells containing
102 to 107 folds of diluted bacterial samples. Six aliquots of 10 μL from each diluted sample were deposited sequentially onto a Brain Heart Infusion (BHI) agar plate (with
20mg novobiocin/L). Plates were incubated at room temperature overnight and colonies per spot were counted. Calculations were performed in Excel as per Chen et al (2003)340. 266
15.2.5 Experimental arrangement and procedure
The experimental arrangement, consisting of flow circuit, sensor flow cell (SFC) and the connection of associated electronic interfaces, is shown in a schematic form in Figure
14.2. The PEMC sensor was installed vertically in the SFC. The cantilever electrodes were connected to an impedance analyzer (Agilent, HP4192A) interfaced to a PC for obtaining impedance and phase angle measurements. The SFC’s hold up volume is 0.3 mL, and was maintained at 22 ± 0.1°C. All detection experiments were carried out at flow rates of 1 or 3.2 mL/min. Since the tubing and SFC have a hold-up volume of 3 mL, the sample concentration was in effect diluted by 50% of the original concentration during the measurement. The SFC operations were the same as discussed in Chapter 14 for rinsing, immobilization and detection.
15.2.6 Scanning electron microscopy
In two of the detection experiments, the cantilever tip fractured and they were preserved for examination under a microscope. The tip was first rinsed thrice with deionized water, and dried at 50°C for 5 h. The glass tip containing the immobilized bacteria was then platinum coated, and examined at 15 kV, 4.0 spot size, and 8,000X magnification using a field emission XL30 scanning electron microscope. In a separate experiment, the meat sample bearing the bacteria was dyed with viability stain (SYTO 9;
Molecular Probes) as per manufacturer’s instruction and examined in a fluorescence microscope. Micrographs at 400X were obtained for visual confirmation of bacterial concentration. 267
15.3 Results and discussions
15.3.1 Resonance characterization of PEMC sensors
Several cantilever sensors were fabricated and used in the various detection experiments. The PEMC sensors used in this study had approximately the same physical dimensions and therefore, their resonance characteristics were nearly identical. The resonant spectrum shown in Figure 15.1A is typical of the results obtained in air. The spectrum is a plot of phase angle versus excitation frequencies. The sensors showed fundamental and second flexural mode resonant frequency of 29.05 ± 0.05 kHz and
115.55 ± 0.05 kHz, respectively. In any single sensing experiment, the resonant frequency was monitored at a high resolution and thus noise due to measurement was less than ± 2 Hz suggesting that the structural and electrical features of the cantilever were stable and constant. The fundamental resonant mode had a larger change in phase angle and a narrower bandwidth compared to the second flexural mode. Upon immersing the sensor into the sensor flow cell, the fundamental resonant peak maintained its narrowness and was found to be suitable for monitoring the effective mass change on the cantilever.
On the other hand the second flexural peak became wider, and fluctuated a great deal
(±10 Hz) upon liquid immersion, and was deemed unsuitable for sensing. As a result, the first flexural resonant mode was used in all the detection experiments. In Figure 15.1B we compare the fundamental resonant frequency of the cantilever in air, and when it was submerged in PBS. Note that the resonant frequency decreased from 29.05 kHz in air to
13.05 kHz in PBS, and the peak height of phase angle also decreases. The decrease in phase angle is due to damping. The sharpness of the peaks is defined by the quality factor
(Q value) and the larger the Q value the more suitable a resonant peak is for detection. 268
The Q values in air and under liquid were 38 and 28, respectively.
15.3.2 Initial evaluation of PEMC sensor
In Figure 15.2 we compare the response of PEMC sensor to two beef samples prepared in flask and in a Stomacher bag. Samples from flask cultures were not filtered, while those from Stomacher bags had been filtered through the 3-micron filter. The former was visually more optically dense containing meat and fat particulate matter that readily settled in the sample tube. The latter was more homogenous and appeared to be less optically dense. During the initial 20 minutes, control or reference sample (no E. coli O157:H7 added) from one of the two sources (flask or bag) was flowed through the
SFC. The data shown in Figure 15.2 for time less than 20 minutes consists of three different reference samples, each obtained separately, and the data labeled control refers to the 4-hour flask beef sample that was not inoculated with E. coli O157:H7. Resonant frequency of the sensor remained constant when exposed to reference and control samples at its initial resonant frequency value. Variation observed was within a frequency band of ± 2 Hz.
The PEMC sensor resonant frequency change was larger with longer incubated samples. For example, the 4 h flask sample gave a total frequency change of 64 Hz compared to the 2 h sample which showed a 36 Hz change in 1 h. A slightly larger total change in the resonant frequency was observed for the 2 h bag sample compared to the 2 h flask sample, 42 Hz compared to 36 Hz. This may suggest that the bacteria grew slightly more favorably in the bag than in the flask. The difference may also be due to the presence of small beef particles in the bag sample. The results in Figure 15.2 suggest that the detection of EC at 2 h incubation is feasible. Thus, subsequent experiments were 269 designed with a starting sampling time of 2 hours.
15.3.3 Detection of E. coli O157:H7 in meat-free broths
To develop a basic set of PEMC response when no beef and particulate matter is present, EC was cultured in growth media. Samples drawn from these were exposed to
PEMC sensor and the resulting response is given in Figure 15.3. The sensor was exposed to reference and control solutions during the first 35 minutes flowing at 1 mL/min. Upon introducing the E. coli broth samples the resonant frequency decreased and reached a steady state value within 50 minutes. The data indicate that samples from longer incubation times gave rise to larger total resonant frequency decrease, similar to the results in flask cultures (Fig. 15.2). The steady state resonant frequency values for the 2 h, 4 h, and 6 h bag samples were respectively 16, 30, and 54 Hz. Here, it can be seen that the total frequency change of the 6 h sample is approximately 3 times that of the 2 h sample. One also notes in Figure 15.3 the initial slope of the response was slightly higher for the longer incubated samples. This is due to a larger number of EC cells in the 6 h sample compared to the 2 h sample. In order to obtain an actual cell count, samples were plated and incubated for 24 hours, and a summary of colony count is given in Table 15.1.
The 2 h sample yielded 96 ± 160 CFU/mL. The range of colony count was 33 to 380 in a total of five replicates. The response of 16 Hz shown in Figure 15.3 is well above the noise level in the measurement (± 2 Hz). In other words, detection of EC concentration at
100 cells/mL of the original sample by PEMC sensor is feasible. Note that only 3 mL of the sample was used, and thus a total of 300 cells were introduced into the sensor flow chamber. Since the measurement system dilutes the input by 50% due to the hold-up volume, the effective concentration in the measurement loop was 50 cells/mL. 270
15.3.4 Detection of E. coli O157:H7 in meat samples
To complement the results from meat-free samples, hamburger inoculated with E. coli O157:H7 were prepared as described earlier. In Figure 15.4 the response of PEMC sensor to the various meat samples is shown. The reference and control solutions were from samples prepared in an identical fashion without the inoculation of EC cells. The sensor response showed a decrease in the resonant frequency as the bacterial samples were flown through the sample cell. The total frequency change was proportional to the sample incubation time. The total resonant frequency change observed for the 2 h, 4 h, and 6 h, meat (M) samples were respectively 21, 38, and 70 Hz. Comparing the frequency change of the meat (M) and broth (B) samples harvested at the same time, a larger frequency change was observed for the meat containing broth samples. The meat samples contained substantial number of non-pathogenic EC cells in the order of thousands at time zero. Therefore, the sensor response to control shown in Figure 15.4 had not only meat particles, but also non-pathogenic EC in tens of thousands per mL (see
Table 15.1). For example, as shown in Table 15.1, the sample at 2 h had nearly seven thousand total cells per mL with approximately 3,800 pathogenic EC cells per mL. The sensor response of 21 Hz is indicative of its selectivity to the pathogen, E. coli O157:H7.
One also notes that in Figure 15.4, upon introduction of the pathogen containing samples, the initial rate of resonant frequency change was nearly the same in all three samples, suggesting that the initial response rate of the sensor is a weak function of pathogen concentration. However, steady state response is a function of pathogen concentration. A nearly similar conclusion can be drawn by examining the broth samples that are presented in Figure 15.3. A more detailed analysis of the kinetics is presented in 271 a later section.
15.3.5 Release of bound E.coli O157:H7
One way to confirm that the sensor response is due to the pathogen attachment is to release the bound EC cells and then check if the sensor resonant frequency returns to the pre-sample exposure resonant frequency value. Such a test was conducted in all sensing episode. Two meat sample results are given in Figure 15.5 to illustrate the typical release response. The release of the bound E. coli O157:H7 from the antibody functionalized
PEMC sensors was carried out using HCl/PBS solution pH 2.0. Our previous work on release using this buffer was quite successful (Chapter 6). At the end of a detection episode, the flow cell was rinsed with PBS buffer (pH 7.2) followed by the flowing of the release buffer at a rate of 1 mL/min. Release of bound EC caused a rapid rise of the resonant frequency. As shown in Figure 15.5, for the 2 and 6 h meat samples, there was full recovery of resonance frequency. Note that the rise in resonant frequency is exactly the same as the decrease observed during detection. This confirms that the sensor response was due to E. coli O157:H7 in the sample, and not due to extraneous material in the sample.
15.3.6 Detection of E. coli O157:H7 in broth containing irradiated meat samples
Since the raw meat contained various strains of microorganisms including wild E. coli, the sensor performance was characterized in presence of contaminating microorganisms. It is however useful to examine its performance when only meat and meat-related particles are present. This was accomplished by sterilizing 25 g of ground beef, followed by E.coli O157:H7 inoculation and measuring the samples obtained at various incubation times. The results from these experiments are presented in Figure 272
15.6. Figure 15.6A shows the frequency response to samples obtained from 2 h, 4 h, and
6 h incubation. The flowing of E. coli sample solution across the PEMC sensor showed an initial rapid decrease in resonant frequency in each of the experiments. The rate at which the resonant frequency changed and the total steady state frequency change depended on the beef sample incubation time. Longer incubation samples gave rise to larger total resonant frequency change. This is an expected result, since similar results were observed in the three different experiments presented earlier. Frequency change for the control, 2 h, 4 h, and 6 h irradiated samples were respectively 0 ± 2 (n=2), 22 ± 3
(n=2), 46± 5 (n=2), and 86 ± 5 (n=2) Hz. The raw meat samples of the same incubation time showed a smaller frequency change compared to the 2 h sample in Figure 15.6, suggesting that the presence of other microorganisms partially hindered the transport of
EC to the sensor surface. However, the difference in the total resonant frequency change between the sterilized and raw samples was within a small margin, suggesting that the
PEMC sensor is quite selective and sensitive to E. coli O157:H7 concentrations.
From the 4 h and 6 h sterilized meat experiments, samples were drawn from two different locations inside the stock container. These samples were labeled 1 and 2. The responses to these samples are compared in Figure 15.6, Panel B. We also include in the same graph, non-EC irradiated meat samples at 6 h served as the control. Note that samples drawn at the same incubation time, gave nearly the same overall frequency shift, but at a slightly different rate. The sample designated as 2 showed a slower response.
Visual inspection of the samples showed that sample 2 contained a larger quantity of meat particles, and was more optically dense compared to sample 1. The reduction in speed of detection is probably due to these extraneous particles. Presence of particulate 273 matter seems to retard attachment of EC on the sensor surface, but they do not appear to cause a large shift in ultimate steady state response.
In two of the experiments, the sensor tip fractured during removal, and was further examined in a scanning electron microscope. The micrographs showed non- uniform distribution of E coli cells attached to the surface, confirming the observed sensor response. Detailed quantification using micrographs was not pursued, as the handling of the sample may have been compromised. Raw beef samples dyed with the viability stain, SYTO 9, showed the presence of a fair number of wild E coli confirming the plating data summarized in Table 1. Again, quantification from 400X micrographs was not carried out
15.3.7 Scanning electron microscope image
In two of the experiments, the sensor tip fractured during removal, and was further examined in a scanning electron microscope. The micrographs showed non-uniform distribution of E. coli cells attached to the surface, confirming the observed sensor response. Detailed quantification using micrographs was not pursued, as the handling of the sample may have been compromised. Raw beef samples dyed with the viability stain,
SYTO 9, showed the presence of a fair number of wild E. coli confirming the plating data summarized in Table 15.1. Again, quantification from 400X micrographs was not carried out.
15.3.8 Determination of E. coli concentration using the Most Probable Number Method (MPN)
Each sample that was analyzed by the PEMC sensor, was also plated on agar with novobiocin, incubated and colonies counted after 24 hours. Each sample was plated a 274 minimum of five replicates. The results are expressed as CFU/mL, and summarized in
Table 16.1. Novobiocin retards significantly the growth of gram-positive bacteria and therefore, the plates were selective to all E. coli strains. The bacteria count in the meat samples without E. coli O157:H7 were due to endogenous bacteria, and the higher levels of bacteria counted in the meat samples inoculated with E. coli O157:H7 were the result of the added E. coli O157:H7. The results presented in Table 15.1 showed that the cells in broth plus meat grew significantly better than in broth alone, which is in agreement with the frequency response results obtained by the PEMC sensors. That is, a larger total resonant frequency change was obtained for the meat samples with E.coli O157:H7 in comparison to broth samples harvested at the same incubation time. The broth-meat samples without E. coli O157:H7 had a level of cells that did not change significantly in
2 h incubation time. The plating of broth samples reflects the actual numbers of pathogens added and grown. The high number of bacteria in the broth + beef plate counts is due to the endogenous bacteria that are not affected by the antibiotic novobiocin. The samples without E. coli O157:H7 were used as a reference for the samples in which E. coli O157:H7 was added. The zero resonant frequency change measured for the reference or control samples indicates the absence of E. coli O157:H7, which is in agreement with the plated results: broth medium and the broth containing sterilized meat samples. The 6-hour broth+hamburger + E. coli O157:H7 sample gave a high E. coli count (1.6 x 106 CFU/mL) compared to the 4-hour sample. However, the total resonant frequency change measured for the 6-hour sample was approximately twice that of 4-hour sample. One possible explanation is that the sensor detects only E. coli
O157:H7 and the high levels of finely ground beef retards the transport of E. coli 275
O157:H7 to the sensor surface.
It is useful to compare the results of this study with the values predicted by the 6-drop modified Most Probable Number (MPN) analysis developed for E. coli O157:H7 by US
Department of Agriculture www.fsis.usda.gov/Science/Microbiological_Lab_Guidebook/index.asp. In Table 15.1,
MPN estimate of E. coli O157:H7 concentration is given at various times using the aerobic growth model in the pathogen modeling program. Lag phase was assumed to be zero, and the initial concentration was set to 0.25 pathogen/mL as 25 EC were added to
100 mL of growth media. Results obtained are summarized in Table 15.1 along with the plating results. Comparison of responses obtained (Figures 15.3, 4 and 6) with MPN values in Table 15.1 suggests that detection sensitivity of the PEMC sensor is in the range of 50 to 100 cells/mL. The resonant frequency change was in the range of 20 to 40 Hz with a noise level of ± 2 Hz. Therefore, if an accurate measure of resonance frequency can be made with a resolution of 1 Hz, in principle, detection of less than 4 CFU/mL is possible. Further optimization of the flow field, and improved PEMC sensor design are needed to achieve detection at such a low concentration.
15.3.9 Kinetics of E. coli O157:H7 binding
The binding kinetics of E. coli 0157:H7 to PEMC sensor was shown to follow Langmuir model using initial rate analysis, Eq. (9-4). Fitting the sensor response presented in
Figures 15.3, 4, and 6 (Panel A) to Eq. (9-4) gave straight lines with excellent correlation coefficients ranging from 0.94 to 0.99, and is illustrated in Figure 15.7 (Panels A, B, and
C). Limiting the initial rate analysis to the first five minutes appears to be a reasonable
approach to obtain kinetic constants. The values of kobs determined are summarized in 276
Table 15.2. It is to be noted that kobs decreases in longer incubation samples of plain broth and raw beef, while it increases in EC inoculated sterile beef samples. The
-1 parameter kobs decreases from 0.01 to 0.05 min in broth samples, and from 0.19 to 0.03 min-1 in raw meat samples, while it more than doubles from 0.06 to 0.14 min-1in irradiated spiked meat samples. In the raw meat samples there were large number of particulate matter, and each aliquot of sample had its own characteristic optical density.
In other words, the samples from meat preparations were far more inhomogeneous than the broth samples. The reduction of binding rate with incubation time is probably due to crowding of the sensor surface by non-pathogenic E coli and inert particulate matter causing reduced access for the pathogen E. coli O157:H7. As shown in Table 15.1, in the raw beef samples wild type EC was present in 3,000 to 20,000 cells/mL and their presence would tend to hinder transport of pathogenic EC to sensor surface. In any case, further studies are needed to better characterize the kinetics of EC binding to PEMC sensor in the presence of meat particulate matter. It is however, interesting to note that although the E. coli O157:H7 concentration was very low in comparison to the other microbes, reliable level of binding did occur as evidenced by resonant frequency decrease. In the case of sterilized beef sample, there were no wild E. coli cells, but meat particles were present. It is believed that the meat particles tend to be much larger than the size of EC, and that may be the reason why we observed higher binding constant at 6 h than at 2 h. However, this value was lower than that observed in raw beef samples at 2 h. The results in Table 2, give us the range of values obtained for the binding rate constant, and is useful in future design of experiments and the sensor. Further studies are planned to quantify these aspects of sensing, so that better sensor design and improved 277 performance can be achieved.
15.4 Conclusions
In this paper we have shown that the PEMC sensor functionalized with anti-E. coli
O157:H7 can be used to detect the pathogen, E. coli O157:H7, in beef samples under flow conditions in real time. The most significant result is that the method does not require any sample preparation, and that the sensor is able to sense about 50 to 100 cells per mL of meat particle containing broths. The sensor is highly selective, and its response is reliable even when one hundred thousand non-EC cells per mL are present. The attachment kinetics of E. coli O157:H7 to the sensor surface was characterized using the
Langmuir model. The results showed a decrease in characteristics binding rate constant at increased E. coli O157:H7 concentration in broth and raw beef samples, while the rate constant increased in sterilized beef. 278
Table 15.1: E. coli concentration in samples using the Most Probable Number Method (MPN).
Samples 0 h count 2 h count 4 h count 6 h count
( CFU/mL ) ( CFU/mL ) ( CFU/mL ) ( CFU/mL )
Control: 0 0 0 0 100 mL Broth
100 mL Broth+25 E. coli 37±22 96±160 120±69 1,700±1,621 O157:H7 cells
Control: 100 mL 3.74x103±2.83x103 3.83x103±3.01x103 1.16x104±1.85x104 2.12x104±1.33x104 Broth+25 g Beef
100 mL Broth+25 g Beef+25 2.4x103±3.21x103 7.6x103±1.25x104 3.1x104±1.72x104 1.6x106±1.58x106 E. coli 157:H7 cells 279
Table 15.1: E. coli concentration in samples using the Most Probable Number Method (MPN) (Continues).
Samples 0 h count 2 h count 4 h count 6 h count
( CFU/mL ) ( CFU/mL ) ( CFU/mL ) ( CFU/mL )
Control: 100 mL Broth+25 g 0 0 0 0 of sterilized Beef
100 mL Broth+25 g of sterilized Beef+25 E. 33±81 200±120 6.79x104±7.79x104 8.21x105±5.8x105 coli O157:H7 cells
PMP70 Estimated of E. coli 0.25 16-58 847-7.05 x104 1.8 x104-9.95x104 O157:H7 concentration in broth 280
Table 15.2: k values as a function of E. coli O157:H7 incubation time. obs
E. coli O157:H7 k obs values in broth k obs values in meat k obs values in incubation time (h) samples (min-1) samples (min-1) irradiated meat samples (min-1 )
2 0.096±0.02 0.19±0.06 0.07±0.01 4 0.046±0.01 0.06±0.01 0.05±0.02 6 0.032±0.01 0.03±0.01 0.09±0.07
281
-70 -72 -74 A -76 -78 -80 -82 -84 Phase angle [degree] -86 -88 -90 0 25 50 75 100 125 150 175 200 Frequency [kHz]
-70 -72 -74 B -76 -78 -80 -82 -84 Phase angle [degree] -86 -88 -90 10 15 20 25 30 35 40 Frequency [kHz]
Figure 15.1: Panel A: Typical resonant spectrum of PEMC sensors in air. The spectrum is a plot of phase angle versus frequency. The fundamental and second resonance mode occurred at 29.5 ± 0.05 and 115 ± 0.05 kHz, respectively. Resonance is followed by a sharp change in the phase angle. The cantilever was excitated with 100 mV. Panel B: The fundamental resonant peak in air (right), and after immersion in the flow cell in PBS (left) flowing at 1mL/min. 282
10 Initiation of E. coli sample flow
0 control -10
-20 2h-Flask -30
-40
-50 2h-Bag
-60 4h-Flask Resonant frequency change [Hz] -70 0 1020304050607080
Time [min]
Figure 15.2: Resonant frequency change as the antibody functionalized PEMC sensor was exposed to E. coli O157:H7 inoculated beef samples flowing at a rate of 3.2 mL/min and temperature of 22 oC. The E. coli samples were harvested for 2h and 4h from a flask and a stomacher bag culture. The stomacher bag contains finely blended meat and the flask unblended beef, both samples were filtered to remove large beef particles. 283
10 Initiation of E. coli sample flow Control 0
-10 2h-25B
-20 4h-25B -30
-40
Resonat frequency change [Hz] frequency Resonat 6h-25B -50
-60 0 20 40 60 80 100 120 140 Time [min]
Figure 15.3: Resonant frequency change as a function of time for the binding of E. coli O157:H7 to antibody derivatized PEMC sensor from broth (B) samples flowing at 1 mL/min and was initially inoculated with 25 pathogens. The samples were harvested 2h, 4h, and 6h. The response labeled control was an antibody functionalized sensor immersed in broth samples in which no E. coli O157:H7 was initially added. 284
10 Initiation of E. coli sample flow Control 0
] -10 2h-25M -20
-30 4h-25M
-40
-50
Resonant frequency change [Hz change frequency Resonant -60 6h-25M
-70
-80 0 20 40 60 80 100 120 140 160 Time [min]
Figure 15.4: Resonant frequency change as a function of time for the binding of E. coli O157:H7 to antibody derivatized PEMC sensor from meat (M) samples flowing at 1 mL/min and was initially inoculated with 25 pathogens. The experiments were carried out at 22oC. Samples were harvested 2h, 4h, and 6h. The control was an antibody functionalized sensor immersed in samples in which no E. coli O157:H7 was initially added.
285
5
0 A
-5
Control -10 E. coli Attachment -15 E. coli Release
Resonant frequency change [Hz] change frequency Resonant -20
-25 0 50 100 150 200 250 Time [min]
10
] 0
-10 B
-20
-30
-40
-50 Control -60 E. coli Release Resonant frequency change [Hz change frequency Resonant -70 E. coli Attachment -80 0 50 100 150 200 250 300
Time [min]
Figure 15.5: The fundamental resonant frequency change as a function of time to the binding and the release of E. coli O157:H7 to the PEMC sensor surface. Panel A: The attachment and release of E. coli O157:H7 from the 2 h meat (M) sample. The total frequency change due the binding and unbinding were respectively 21.2 ± 2 Hz and 22.5 ± 2 Hz. Panel B: The binding and unbinding of E. coli O157:H7 from the 6 h meat sample. The steady state resonant frequency change during the attachment and the release of the bacteria was 70 ± 2 Hz. In both experiments the flow rate of the sample was 1mL/min. 286
10 Initiation of sterilized beef+E.coli sample
0 ] -10 Control:0h-25Mx-1
-20 2h-25Mx-1 -30
-40
-50 4h-25Mx-1 -60 A -70 Resonant frequency change [Hz change frequency Resonant 6h-25Mx-1 -80
-90 0 1020304050607080 Time [min]
10 0 Control:0h-25Mx-1 -10
-20 -30 4h-25Mx-2 -40
-50 4h-25Mx-1 -60
-70 Resonant frequency change [Hz] change frequency Resonant 6h-25Mx-2 -80 B 6h-25Mx-1 -90 0 10203040506070 Time [min]
Figure 15.6: Resonant frequency change as a function of time for the binding of E. coli
O157:H7 to antibody derivatized PEMC sensor from irradiated meat (Mx) samples that was flowing at 1 mL/min and was initially inoculated with 25 pathogens. The experiments were carried out at 22 °C. Panel A: Samples were harvested 2h, 4h, and 6h. The control was an antibody functionalized sensor immersed in irradiated meat sample in which no E. coli O157:H7 was initially added. Panel B: The results from two different experiments of irradiated meat-E.coli O157:H7 samples that were from the sample batch but contained more meat particles in one than the other. The 25Mx-2 sample contained more meat particles and appeared optically more dense than 25Mx-1. 287
0
-0.2 ) ∞ f
Δ -0.4 f)/ Δ 2h-0B-25B - -0.6 ∞
f 4h-0B-25B Δ -0.8 6h-0B-25B A ln(( -1 02468 Time,τ [min]
0.0
) -0.5 ∞ f Δ 2h-0M-25M
f)/ -1.0 Δ
- 4h-0M-25M ∞ f Δ -1.5 6h-0M-25M
ln(( B -2.0 02468 Time,τ [min]
0 -0.1 -0.2 ) ∞
f -0.3 Δ
f)/ -0.4 Δ 2h-0-25Mx-1 -
∞ -0.5 f Δ -0.6 4h-0-25Mx-1
ln(( -0.7 6h-0-25Mx-1 C -0.8 01234567 Time,τ [min]
Figure 15.7: Langmuir kinetic analysis of E. coli O157:H7 binding to the PEMC sensor surface. The initial kinetic analysis of the various incubated E. coli O157:H7 samples: Broth (Panel A), broth and meat (Panel B), and broth and irradiated meat (Panel C). Correlation coefficient ranged from 0.94 to 0.99. The slope of each line gives the observed characteristic binding rate constant kobs. 288
Chapter 16: Detection of Staphylococcus Enterotoxin B at picogram levels using Piezoelectric-Excited Millimeter-Sized Cantilever Sensors
16.1 Introduction
The detection of low toxin concentration is of great interest from a perspective of homeland security, medical, food, and environmental applications. Various methods have
341 been proposed for detecting biological toxins such as aflatoxins B1, B2, and G1, T-2 toxin,342 cholera toxin,341,343 and Shiga-like toxin (SLT).344 Our interest in examining
SEB detection stems from the development of highly sensitive mass change cantilever sensor which is based on resonant frequency measurements (Chapter 13). Given that
PEMC sensors measure adsorption of proteins, it is of interest to examine if it can be used to detect toxin molecules especially at the picogram levels using immobilized antibody as the sensing molecule.
Staphylococcal enterotoxin B (SEB), a causative agent of food poisoning, is a small protein toxin (28.4 kDa) produce by the Gram positive Staphylococcus aureus bacterium.
SEB pose a serious bioterrorism threat because the toxin can be used as a weapon in both food poisoning and aerosol309. Inhalation of SEB can result in the same symptoms as if ingested. Some of the symptoms are chest pain, nausea, diarrhea, vomiting, dyspnea, and toxin shock310,311. Because SEB pose such a potential threat to public health several sensor platforms have been developed to identify and detect SEB signature at low concentrations. Some of these sensor platforms are enzyme-linked immunosorbent assay
(ELISA)312-314, surface plasmon resonance (SPR) 315-319, optic fiber biosensors320-323, fluorescence-based sensors324-327, magnetoelastic immunosensor328, and piezoelectric crystal immunosensors329. These sensing platforms are useful in detecting high 289 concentration of SEB. For instance, Lin and Tsai (2003) showed that piezoelectric crystal immunosensors have detection limit in the range of 2.5 to 60 μg/mL329, and fiber optic surface plasmon resonance sensor has a detection limit of 10 ng/mL319. Most recently,
Raun et al., 2004 showed that magnetoelastic sensors have SEB detection limit of 0.5 ng/mL328. Peruski et al., 2002 showed that time-resolved fluorescence assays achieved detection sensitivity in the range of 4 to 20 pg/mL SEB325. To the best of our knowledge, this is the lowest concentration of SEB detection that has been reported. However, the technique requires labeled reagents that are expensive. Since it only requires a small dose of SEB to show symptoms of food poisoning there is a need for the rapid real-time detection of low SEB concentration. In this work, we investigate the detection of picogram SEB concentration using the highly sensitive piezoelectric-excited millimeter- sized cantilever (PEMC) sensor.
PEMC sensors are inexpensive, robust, selective, and sensitive mechanical resonators a few millimeters in length. The detection of biological species requires the immobilization of a recognition molecule, such as an antibody or a receptor. The recognition molecule provides the selectivity for the desired antigen. Detection measurements are monitored by the change in the sensor’s resonant frequency, which results due to changes in the cantilever’s effective mass caused by the binding of target analyte. The transient response of resonant frequency is used to provide quantitative measurement of the target concentration. In the earlier Chapters, we showed that PEMC sensors successfully detect low concentrations of pathogens, proteins, and molecules.
Detection of E.coli O157:H7 in batch configuration at 70 cells/mL (Chapter 7), proteins at 10 μg/mL (Chapter 10), self-assembly of 1-hexadecanethiol at 1 nM (Chapter 11), and 290 detection of Bacillus anthracis spores in both batch and flow configurations at 300 per mL were shown (Chapter 12 & 14). Our previous work with PEMC sensors for protein has been at high concentrations on the order of μg/mL. In this work we investigate if the
PEMC sensor can measure the presence of the model toxin, Staphylococcus Enterotoxin
B, under flow conditions at a much lower concentration of picograms
16.2 Materials and Methods
16.2.1 PEMC Fabrication
The design and construction of PEMC sensors are described in Chapter 4. The PEMC sensor used in this study has free-end dimensions of 1 ± 0.05 x 1 ± 0.05 x 0.127 ± 0.005 mm3 (L x W x t) PZT and 3 ± 0.05 x 1 ± 0.05 x 0.160 ± 0.005 mm3 glass. That is, 2 mm glass layer overhangs the PZT layer, and provides the surface for antibody functionalization for antigen detection. The PZT layer was polyurethane coated to prevent liquid contact.
16.2.2 Experimental procedures
All immobilization experiments were carried out under flow configuration. PBS solution was re-circulated through the SFC until the sensor’s resonant frequency reached a constant value, which was achieved within 10 minutes. Then, the activated antibody solution was flown into the cell (by opening valve V2 and closing V1) for 1 h to allow immobilization at 25 ± 0.2 °C. The antibody was affinity purified sheep polyclonal to staphylococcus enterotoxin B (SEB); anti-SEB and Staphylococcal enterotoxin B (SEB) were provided by Dr. Marjorie Medina (Agricultural Research Service, USDA), and were purchased from Toxin Technology (Sarasota, Florida). The anti-SEB concentration used in all immobilization procedures was 10 μg/mL in 10 mM phosphate buffered saline 291
(PBS, pH 7.4). Hydroxylamine was then flowed through the cell (by opening V3), and then PBS (by opening valve V1 and closing V3) to rinse out the lines. Upon completion of the rinsing step, the antigen solution was introduced by opening V4 and closing V1.
After the antigen was attached, V4 was closed and the cell was rinsed with PBS (by opening V1) followed by flowing of the release solution (open V5 and close V1).
Antigen solution used in each experiment consisted of 1 mL containing 50 pg/mL, 200 pg/mL, 1 ng/mL, and 50 ng/mL of SEB. The tubing and the flow cell has a total hold-up volume of 3 mL. Hence, the effective concentrations of SEB in the SFC were 12.5 pg/mL, 50 pg/mL, 250 pg/mL, and 12.5 ng/mL, corresponding to sample SEB concentration. One notes that the measured resonant frequency of the PEMC sensor was monitored until it stabilized before immobilizing any biological entities, followed by the initiation of the solution of interest. In all the experiments the temperature of the SFC was maintained at 25 ± 0.2 °C.
16.3 Results and Discussion
16.3.1 Resonance characterization of PEMC sensors
The PEMC sensors used in this study have two resonant peaks in the frequency range of 1 to 100 kHz. The fundamental and second mode resonant frequencies were located at
17.50 ± 0.01 kHz and 87.10 ± 0.01 kHz in air, respectively. Two PEMC sensors of nearly identical resonance characteristics were used, but for brevity only the results of one is presented. The fundamental mode was used to monitor the immobilization of anti-SEB and the detection of SEB, as it had better stability characteristics under liquid flow conditions than the second mode. Each experiment was repeated at least twice and the data shown are typical of the results obtained. Upon immersing the sensor in the SFC 292 containing PBS flowing at 1 mL/min, the fundamental frequency decreased from 17.50 kHz to 10.37 kHz; see Figure 16.1. The decrease in resonant frequency was due to the added oscillating mass of liquid on the sensor surface. Also, the peak height decreased slightly from -74° to -78°, and the peak shape remained somewhat constant. The shape of a resonant peak is characterized by its quality factor (Q-factor). The Q-factor is a measure of the sensor’s dissipative losses due to viscous damping and is determined as the ratio of the resonant frequency to the peak width at half the peak height. The Q-factor of the fundamental mode in air was 30, and in PBS it was 28. The small decrease suggests that damping due to the surrounding PBS solution was relatively small. This is quite unlike silicon microcantilevers that have low Q-factor (~ 1 to 2) under fully submerged conditions.11 The small viscous damping effect observed in PEMC sensors is due to the cantilever’s dimensions (length and width). As these dimensions are on millimeter scale the Reynolds number at the oscillation frequency was tens of thousands and thus, the hydrodynamic loading dissipative effects are negligible (liquid is considered inviscid).51
Our experience is that PEMC sensors lose Q-value in the range of 5 to 15% under liquid immersed flow conditions compared to values in air.
16.3.2 Detection of SEB at 50 ng/mL in SFC
In Figure 16.2 the transient frequency response of the PEMC sensor to the sequence of experiments consisting of anti-SEB immobilization and SEB detection is presented.
The functionalization of the sensor with APTES created free primary amines on the surface. The activated anti-SEB reacted with the amine groups forming peptide bonds.
The carboxyl groups on the antibody were activated with EDC and sulfo-NHS to form a stable intermediate137. As shown in Figure 16.2, the frequency was constant in PBS 293 flowing at 1 mL/min (baseline frequency) until the activated anti-SEB (10 μg/mL) flow was initiated. Exposure of the aminated sensor to anti-SEB caused an immediate decease of the resonant frequency and the frequency change reached a constant value of -212 Hz in approximately 1 h. It is observed from the Figure that the steady state frequency value
(-212 Hz) obtained for the anti-SEB immobilization is slightly above the minimum frequency change (-240 Hz) during the anti-SEB immobilization step. We believe that the increase in frequency (from -240 Hz to -212 Hz) was caused by the detachment of weakly bound anti-SEB from the sensor surface. Hydroxylamine was flown at 94 minutes for 4 minutes through the SFC to convert the activated carboxylic groups on anti-SEB that did not participate in the reaction on the sensor surface, back to normal carboxyl groups. Subsequently, PBS buffer was flown to rinse the SFC and the connecting tubes, and to achieve a new baseline for the preceding step. The injection of 1 mL of SEB at 50 ng/mL into the re-circulation buffer of 3 mL (effective concentration of 12.5 ng SEB/mL) caused an immediate decrease of resonant frequency upon sensor exposure. The frequency continued to decrease and reached a constant value of -419 ± 1 Hz. The total change in resonant frequency due to SEB binding was 208 Hz.
At the conclusion of SEB binding, PBS rinse was implemented for 10 minutes, during which residuals from the previous steps were removed from the flow circuit and achievement of the baseline for the step that follows was obtained. The change in resonant frequency during the rinsing step was very small and was within the noise level
(± 2 Hz) of the sensor. The sensor was then exposed to a release solution (HCl/PBS, pH
2.0) to detach the bound SEB from the sensor surface. Upon exposure, the sensor responded with an increase in its resonant frequency reaching the frequency value below 294 the one prior to the flow of the release solution. This indicates that the mass of the PEMC sensor decreased. As shown in Eq.(3-10), the resonant frequency change of a PEMC sensor is related linearly to changes in bound mass. Therefore, the increase in resonant frequency was caused by the detachment of SEB. The total change in frequency during
SEB attachment and release were 208 and 180 Hz, respectively. The difference may indicate that not all the bound SEB was released from the sensor surface. Another possible explanation for the difference is that at pH 2.0, there may have been changes to the antibody structure so that the resonant frequency change did not recover fully during the release. However, when the SFC was rinsed with PBS buffer the resonant frequency increased an additional 30 ± 4 Hz, which may support the idea of antibody undergoing conformational change. We are currently studying the effect of pH on resonant frequency of antibody immobilized sensors, and will be the subject of a future report. Furthermore, these results suggest that regeneration of the antibody recognition layer can be achieved after HCl/PBS wash by PBS buffer. Medina (2005) reported the high residual surface concentration of anti-SEB bound to the SEB-sensor after regeneration with 100 mM
317 9 -1 HCl . This can be attributed to the high affinity (KA = 10 M ) and low dissociation (KD
=10-10 M-1) of the anti-SEB with SEB. In this study, no attempt was made to quantify the released SEB nor did we characterize the regenerated sensor surface. Both of these issues are currently being investigated. It is worth noting that all the samples were made with
PBS buffer (pH 7.4) and therefore, have nearly identical viscosity. As a result, the sensor did not show any viscous respond to the various sample solution. This is an advantage of
PEMC sensor over other microcantilever sensors. 295
16.3.3 The Stepwise Binding of SEB at the picogram levels in SFC
This experiment was carried out in a similar fashion as described in the previous section except for the use of significantly lower concentration of SEB and the sequential increase in the SEB concentration. The purpose was to determine SEB detection limit of the PEMC sensors used. During the entire experiment the flow rate was kept constant at 1 mL/min. As shown in Figure 16.3, immobilization of anti-SEB resulted in a total resonant frequency change of 195 Hz, which is in good agreement with value obtained in the experiment given in Figure 16.2. That is, the antibody layer in both experiments had similar surface density. After the antibody was immobilized, the sensor was exposed to a
1 mL aliquot of 50 pg/mL SEB solution (effective concentration of 12.5 pg/mL) and the sensor did not show much of a response. This suggests that SEB concentration of 12.5 pg/mL was below the detection limit of the sensor. It is important to emphasize that the flow circuit always contain a hold up volume of 3 mL of PBS buffer before the injection of a 1 mL SEB sample. The response was within the noise level (±1 Hz). Next, the system was rinsed with PBS for 4 minutes to remove the entire 12.5 pg/mL sample.
During the rinse the resonant frequency increased by approximately 2 Hz, which may have caused by disturbance induced during sample reservoirs switching. However, a stable baseline was obtained for the following experiment. Then, 1 mL of 200 pg/mL
SEB sample was injected into the re-circulating buffer (effective concentration of 50 pg/mL), the sensor responded with a rapid decrease of its resonant frequency upon exposure, and ultimately reached a steady state frequency change of 31 ± 1 Hz. This indicates that SEB concentration of 50 pg/mL was within the detection limit of the sensor. Subsequently, the system was rinsed with PBS followed by the injection of 1 mL 296 of 1 ng/mL SEB solution (effective concentration of 250 pg/mL) into the re-circulating buffer (PBS pH 7.4), upon contact of the SEB with the cantilever the resonant frequency decreased by an additional 50 ± 1 Hz before reaching equilibrium in 35 minutes. The additional change in resonant frequency of 50 Hz was due to further binding of SEB to free anti-SEB on the surface. This suggests that an equilibrium exists between the antigen in suspension and antibody immobilized on the sensor.
In order to confirm SEB binding, the sensor was exposed to HCl/PBS release solution which caused a rapid increase of the resonant frequency. As shown in Figure 16.3, the total change in resonant frequency increase during the release process was 80 Hz and the total frequency response to SEB attachment was 81 Hz. This indicates that approximately all the bound SEB was released by the release solution. After release, the sensor flow cell was rinsed with PBS and the resonant frequency increased by approximately 10 Hz.
Therefore, the results of the two different experiments indicate that effective SEB binding and release, and recognition layer regeneration can be achieved on PEMC sensors.
Although we have demonstrated fully antibody surface recovery, we do not practice re- using the sensor after the release. Instead we re-functionalize the sensor with antibody in the same fashion each time. In Chapter 14, we reported that PEMC can be regenerated and re-used at least twice without significant loss in antibody activity.
16.3.4 Kinetics of antibody and spore binding
Antibody-antigen binding on PEMC sensors has been shown to obey Langmuir kinetics.
In this Chapter, we examine if SEB binding obeys the same kinetics. The model is
summarized in Chapter 9, Eq. (9-4). The observed rate constant kobs during the initial time (far from equilibrium) can be determined by fitting data obtained during the first 10 297 minutes. Fitting the antibody data presented in both Figure 16.3 and 16.4 to Eq. (9-4),
-1 kobs was determined as 0.173 and 0.134 min . The two values are within 22%, which indicates that the sensor surfaces were similar. The frequency response of SEB binding to the bound anti-SEB on the sensor surface was analyzed in the same fashion and the results are presented in Figure 16.4. The quality of the fits was good with correlation coefficients of 0.992, 0.996, and 0.987 corresponding to the effective SEB concentrations
of 0.05, 0.25, and 12.5 ng/mL. Here, one can see that kobs values decreased with increasing SEB concentration. The observed rate constant for SEB concentrations 0.05,
0.25, and 12.5 ng/mL was 0.104, 0.042, and 0.029 min-1, respectively. The 0.25 ng/mL concentration was fitted to the model based on the assumption that after the 0.05 ng/mL
SEB sample exposure, there was still significant free antibodies on the surface. The value
-1 of 0.042 min is in the range of kobs values obtained in the experiments where fresh sensors were used.
16.4 Conclusion
In this study we showed that piezoelectric-excited millimeter-sized cantilever
(PEMC) sensor can be used for the detection of picogram concentrations of staphylococcal enterotoxin B (SEB) under flow conditions. The binding of SEB to the antibody functionalized sensor exhibited an exponential decrease in resonant frequency.
The binding of SEB was confirmed by release of the attached SEB upon exposure to a low pH solution, and the sensor responded with an immediate increase in resonant frequency of a slightly lower magnitude to the decrease observed during SEB attachment.
However, rinsing the sensor with PBS buffer after the release episode fully regenerates the antibody or recognition layer. The attachment kinetics of SEB to the sensor surface 298 was modeled and the observed binding rate constant ranged from 0.104 to 0.029 min-1 at
SEB concentration of 50 pg/mL to 12.5 ng/mL. The experiments showed that the detection limit of PEMC sensor to SEB was greater than 12.5 pg/mL but less than 50 pg/mL.
299
-70
-74 In Air
-78 In PBS
-82 Phase [degrees] angle -86
-90 8 101214161820
Excitation frequency [kHz]
Figure 16.1: Resonant spectra of PEMC sensor’s fundamental resonant mode in air (right) and fully submerged in PBS flowing at 1 mL/min. The resonant frequency shifted approximately 7.13 kHz in going from air to PBS due to the increase in the sensor’s effective mass. 300
50 PBS 10 μg/mL anti-SEB 0 PBS PBS Release -50 solution Hydroxylamine -100 12.5 ng/mL SEB -150 PBS -200
-250
-300
Resonantfrequency change [Hz] -350
-400
-450 0 50 100 150 200 250 300 Time [min]
Figure 16.2: Transient response of resonant frequency to the binding of 12.5 ng/mL SEB to an anti-SEB functionalized PEMC sensor and the subsequent release of the bound SEB. 301
50 PBS PBS PBS PBS 0 10μg/mL anti-SEB PBS
12.5pg/mL Release -50 SEB solution 50pg/mL SEB -100 250pg/mL SEB PBS -150
-200 Resonant frequency change[Hz]
-250
-300 0 50 100 150 200 250 300 Time [min]
Figure 16.3: Transient response of PEMC sensor to the detection of increasing SEB concentration. The sensor did not respond to 12.5 pg/mL because of its higher detection limit. 302
0.00
-0.20
-0.40 ) ∞ f
Δ
f)/ -0.60 Δ - ∞ f 12.5 ng/mL Δ -0.80 250 pg/mL ln(( 50 pg/mL -1.00
-1.20 0246810 τ [min]
Figure 16.4: Initial kinetic analysis of the binding of various SEB concentrations to an anti-SEB PEMC sensor. Correlation coefficient ranged from 0.988 to
0.996. The slope of each line gives the observed characteristic binding rate kobs. 303
Chapter 17: A Method for Measuring Bacillus Anthracis Spores in Presence of Copious Amounts of Bacillus Thuringiensis and Bacillus Cereus
17.1 Introduction
Sensing of both foodborne and airborne pathogens is essential in the fight against bioterrorism. The Bacillus species present such threat; namely Bacillus anthracis (BA, airborne) is the etiology agent of anthrax330. Bacillus cereus (BC) a foodborne bacterium331,332, and Bacillus thuringiensis (BT) an insecticide toxin producer333,334 are closely related species. Anthrax spores are biologically dormant structures that are highly resistive to extreme temperatures, physical damage, desiccation, and harsh chemicals. As airborne pathogen anthrax spores can cause respiratory infection, as was done in the
United States Fall of 2001. In the case of an anthrax attack the Centers of Disease Control
(CDC) has estimated a cost of $ 26.2 billion per 100,000 people exposed335, and treatment of the disease can only be successful if the antibiotic is administered within a day of exposure336. The threat of a potential biowarfare agent release has created an urgent need for a low pathogen count detector that is selective, sensitive, robust, and gives a quick positive response in real-time.
In response to the potential biological warfare, several sensing platforms capable of providing reliable identification of airborne and foodborne biowarfare agents are currently under development. Some of these sensor platforms include, the enzyme-linked immunosorbent assays (ELISA)337, evanescent wave fiber-optic biosensors198, real-time
PCR,200,201,338 photoluminescence348, and quartz crystal microbalance (QCM)211. While these sensing platforms have been used in detection and identification of pathogens, no papers have appeared that show their selectivity and sensitivity (with the exception of 304
PCR) to very low pathogen concentrations in presence of large number of other bacterial species. In Chapter 9, we have reported the aqueous phase detection of Bacillus anthracis
(BA) spores at 300/mL using a partially immersed PZT/glass anchored cantilever in a batch configuration. We also have shown that the detection method using antibody (anti-
BA) immobilized cantilever sensor was selective to anthrax spores, by exposing the sensor to various mixtures of BA and Bacillus thuriengiensis (BT) at very high concentrations; BA at 106/mL and BT at 109/mL), and their mixtures. The results showed an increase in sensor response with increasing BA concentration. However, the detection of BA at very low concentration in presence of large amounts of other Bacillus species is important. In this Chapter, we investigate the binding kinetics of BA spores at very low concentration from samples containing copious amounts of both BT and BC.
The PZT-anchored piezoelectric-excited millimeter-sized cantilever (PAPEMC) sensors have a piezoelectric layer (lead zirconate titanate, PZT) as the base sensor platform to which a 1.5 mm2 glass layer is bonded at one end. The composite structure is a few millimeters in length. The sensor was designed so that the bending modes of vibration are the dominate modes. One advantage in using a piezoelectric layer is that it provides both actuation and signal sensing. Detection of spores requires the specificity of the sensor surface, which is accomplished by using immobilized antibody to the spore.
Binding of spores to antibody on the sensor increases the sensor’s mass and decrease its resonant frequency. Therefore, by tracking the change in resonant frequency the bound species concentration can be determined.
305
17.2 Sensor fabrication
The PEMC sensors used in this study are of a different geometry than those used in previous Chapters. The new geometry (PZT-anchored PEMC sensor, PAPEMC) is fabricated from the same materials: a 127 μ m thick PZT film and a 160 μm thick quartz cover square. The PZT layer is the base sensor platform, see Figure 17.1. The cantilever free end was designed with the glass layer, 1.50 ± 0.05 x 1 ± 0.05 mm2 (length x width), bonded at one end of the PZT layer, 4 ± 0.05 x 1 ± 0.05 mm (length x width), by a non- conductive adhesive. At the other end, 1.70 ± 0.05 mm length of the PZT layer was epoxied into a glass tube. As a result, the cantilever free end has 0.8 mm of exposed PZT layer. Top and bottom electrodes were made on the 1.7 mm long PZT layer, before it was expoxied, using a 30 gauge copper wire soldered to BNC couplers. The PZT layer at the cantilever free end was insulated with a 20 μm thick polyurethane layer. The glass layer,
1.5 ± 0.05 x 1 ± 0.05 mm2, provide surface for antibody immobilization.
17.3 Experimental
Bacillus anthracis (BA) spores, the Sterne strain 7702, and Protein A purified Rabbit polyclonal antibody in PBS was provided by Professor Richard Rest (Drexel University
College of Medicine, Philadelphia, PA). Bacillus thuringiensis (BT) was purchased from
EDVOTEK (West Bethesda, MD) and Bacillus cereus was provided by Dr. Shu-I Tu’s group (USDA, Eastern Regional Research Center, PA). All other chemical reagents were from Sigma-Aldrich. The procedure used to functionalize the glass surface with antibody was the same as that we report in previous Chapters. After each detection experiment, the sensor surface was renewed in the same fashion with antibody. That is, the sensor was immobilized with fresh antibody. 306
A stock sample of BA spore, concentration 2,000 BA/mL, was prepared in phosphate buffered saline (PBS) (10 mM, pH 7.4) by serial dilution of a master sample (2 x 105 spores/mL). Bacillus cereus (BC) and Bacillus thuringiensis (BT) spores of equal concentrations were mixed together to obtained a final concentration of 2 x 103, 2 x 104, 2 x 105, and 2 x 106 BC+BT/mL. The test samples were prepared by diluting 1 mL of BA sample (2,000 BA/mL) with 1 mL of BT+BC mixture. The 2 mL solution was, then, injected into a flow circuit containing a hold up volume of 4 mL PBS. Therefore, the concentration of BA in any one experiment was 333/mL. Selectivity of PAPEMC sensors was investigated using BA and BT+BC mixed samples of various concentration ratios;
0:1, 1:0, 1:1, 1:10, 1:100, and 1:1000 BA:BT+BC, which yield BA spore concentrations labeled A, B, C, D, E, and F corresponding to 0%, 100%, 50%, 9%, 1% and 0.01 % with total spore count as given in Table 17.1. All experiments were carried out in a flow configuration; the same fashion as was discussed in Chapter 16.
17.4 Experimental determination of mass change sensitivity in a vacuum
The mass change sensitivity of PAPEMC sensor was determined by measuring the changes in resonant frequency following the deposition of known amount of mass; a technique similar to the one reported in Chapters 10 and 13. The approach used the deposition of paraffin wax at 1 pg increments. A mass of 0.23 mg of wax was dissolved in 4 mL of hexane and 0.5 μL of the solution was dispensed into ten weighing plates that were already weighed. The plates were placed under a chemical fume wood for 15 minutes to remove all the solvent. Subsequently, each plate was re-weighed and the difference in weight between the plates with wax and the empty ones were used to compute the average mass of wax in 0.5 μL solution. An aliquot of the stock solution was 307 diluted to a final concentration of 1 pg of wax per 0.5 μL of solution. To determine the sensitivity of the cantilever, the sensor was cleaned, dried, and place in a vacuum chamber at 60 mTorr. The resonant frequency of the sensor was monitored and recorded.
The cantilever was removed from the chamber and 0.5 μL of the wax solution was deposited on its glass surface followed by drying under a fume wood for 15 minutes.
Subsequently, the cantilever was placed in vacuum at 60 mTorr and the resonant frequency was monitored and recorded until it stabilized. This procedure was repeated three times in successive mass addition and the sensitivity was determined as the slope from a plot of mass change against frequency change.
17.5 Results and Discussion
17.5.1 Resonance Characterization of FtPEMC Sensors
Several PAPEMC sensors were fabricated and the resonance spectrum shown, Figure
17.2A, is typical of the spectra obtained. However, only two of the cantilevers were used in the Bacillus anthracis spores detection experiments. For comparison of results, of the various detection experiments, only the data from one PAPEMC sensor is presented.
Each experiment was repeated at least twice and the data shown are typical of the results obtained. The resonant spectrum, a plot of phase angle versus excitation frequency, in air showed dominant bending mode resonant peaks at 102.15 ± 0.05, 970.05 ± 0.05, and
1810.05 ± 0.05 kHz, respectively. In this study, the 970.05 kHz peak was selected for in- liquid detection experiments because upon liquid immersion the resonant frequency shifted the largest (134.05 kHz) and flow initiation, at 1 mL/min, only decrease the frequency by an additional 450 Hz (a change of 134.5 kHz), see Figure 17.2B. The change in resonant frequency in going from air to liquid flow is a measure of sensitivity 308 and therefore, the 970 kHz peak is a sensitive peak. Also the sharpness of the peak did not deteriorate significantly in liquid, which is measured by the quality factor (Q-value).
The Q in air and liquid flow was 25 and 15, respectively.
17.5.2 Mass change sensitivity
Figure 17.3 shows a plot of mass change versus frequency change of the 970 kHz peak in a vacuum as paraffin wax was sequentially deposited at 1 pg increment. The mass of wax in 0.5 μL of the original stock solution was experimentally determined as 47.8 ±
6.4 ng. An aliquot of the stock solution was diluted to a final concentration of 1 pg of wax per 0.5 μL of solution. The addition of wax (1 pg) onto the cantilever glass surface
(1.5 mm2 area at sensor tip) caused the resonant frequency to decrease. The plot in Figure
17.3 gave a straight line with slope 1.47 ± 0.42 fg/Hz and correlation coefficients of
0.998. Several cantilevers, of nearly identical resonance characteristics, were used in the sensitivity study and the sensitivity values in vacuum of the dominant bending mode peaks shown in Figure 2A are summarized in Table 17.2. It is important to note that the sensitivity increased in a non-linear fashion with the increase in resonant frequency. The
mass sensitivity under liquid immersion condition,σ nf , can be estimated from the
sensitivity value in a vacuum(σ nv ). Differentiating and rearranging Eq. (3-2) one gets the mass change sensitivity in vacuum as:
dM e − 2M e σ nv = = (17-1) df nv f nv
where f nv is the resonant frequency in vacuum. It is clear from Eq. (17-1) and (6-2) that 309
lower values of M e or M ef and higher values of f nv or f nf , result in greater mass change sensitivity for a given resonant frequency measurement. The ratio of Eq. (6-3) and Eq.
(17-1) gives:
⎛ M f ⎞ σ = σ ⎜ ef . nv ⎟ (17-2) nf nv ⎜ ⎟ ⎝ M e f nf ⎠
The values calculated for M e and M ef were 2.14 mg and 3.62 mg, respectively. The
estimated values of σ nf were computed and are given in Table 17.2.
17.5.3 Sensor response to BA spores binding from solutions containing various concentrations of non-antigenic bacillus species (BT and BC)
In Figure 17.4, the transient resonant frequency responses to the binding of BA spores at concentration of 333 BA/mL are presented. Each experiment was repeated at least twice and the data shown are typical of the results obtained. In each experiment, the sensor responded with a rapid decrease in resonant frequency before reaching the same steady state value at different time periods. For the pure BA sample (333 spores/mL) the resonant frequency decreased most rapidly and reached steady state in 27 minutes. As the concentration of the non-antigenic Bacillus species (BT and BC) increased the rate of resonant frequency change decreased, and took a longer time to reach steady state changes. Steady states of 2742 ± 38 (n=3), 3053 ± 19 (n=2), 2777 ± 26 (n=2), 2953 ±
24 (n=2), and 3105 ± 27 (n=2) Hz were obtained for the 1:0, 1:1, 1:10, 1:100, and 1:1000
BA:BT+BC samples, respectively, in 27, 45, 63, 154, and 219 minutes correspondingly.
We note that the steady state responses yielded an average frequency decrease of 2926 ±
162 Hz. The deviation of 162 Hz from eleven separate sensor preparations and detection 310 experiments is small. For all practical purposes, the results indicate that the presence of non-antigenic components in the sample minimally affects the sensor response. We conclude that non-antigenic Bacillus species (BT and BC) hindered the transport of the
BA spores to the sensor surface, but never completely prevented attachment of antigenic spore. Corresponding to each experiment a control one was carried out. The control consisted of anti-BA functionalized cantilever exposed to a sample of only BT and BC at the same concentration and experimental conditions as the detection experiment. Upon exposing the sensor to the control sample the resonant frequency change fluctuated around zero. The sensor response to the BT+BC sample presented in Figure 17.4 is a typical result. Here, the response of the sensor was 14 ± 31 (n=11) Hz. We conclude that the control (BT + BC) exhibited no affinity to sensor surface.
Using the mass change sensitivity of the 836 kHz peak under liquid (2.89 ± 0.83 fg/Hz), the average mass of Bacillus anthracis spores bound to the sensor surface in any one experiment was 8.46 ± 2.46 pg. If we assume that a single BA spore weighs 1 pg, then only approximately 9 spores attached in any one experiment. However, in all the experiments there were 2,000 BA spores available for binding and therefore, the maximum total mass change that could have been observed was 2 ng. This indicates that only 0.5% of the total spores in the sample got attached to the sensor. The amplitude of oscillation of the sensor is estimated from the piezoelectric charge coefficient (d31) of
PZT and deflection relationship to be far less than 1 μm. The residence time of the sample in the SFC is about 7 second and the volume of the sample that comes in contact with the cantilever oscillation volume is calculated to be 0.044% of the total sample volume. If we take the case of the clean BA sample that was re-circulated for 311 approximately 50 minutes, the total sample volume that has contacted the sensor was ~22
μL (0.044% x 6 mL x 50 min/6 min. per cycle). Given the spore concentration value
(333/mL), the number of spores that could potentially attach is 7.33 (0.022 mL x 333 spores/mL). That is, to the first approximation, the number of attached spores predicted from the sensitivity value is in agreement with the number of spores that have access to the sensor surface.
The binding strength of antibody to antigen can be nullified by changing either the pH or ionic strength or both. The binding of BA spores to the PAPEMC sensor bearing anti-
BA was confirmed by exposure to a low pH (1.85) buffer and is given in Figure 17.5. The release was signified by a rapid increase of the resonant frequency to a value that was ~
13% smaller than the total frequency change observed during the attachment. One notes that after the spores were attached, flowing of PBS through the SFC caused no significant change in the resonant frequency, which suggests that there were no or low weakly bound
BA spores to the sensor. Conversely, rinsing of the SFC with PBS after the release step resulted in an increase of the resonant frequency to a value that was present prior to the attachment, which indicates that the antibody surface was, now, fully recovered.
17.5.4 Kinetics of BA spore binding
The binding kinetics of pathogens to PEMC sensors are characterized in previous
Chapters by the first order Langmuir kinetic model. In short, the initial rate analysis was used to analyze the binding process, since at time τ= 0 there are no concentration gradients, diffusion effects are absent. The Langmuir model can be expressed as:
− k τ θ = 1 − e obs (17-3) where θ is the fraction of available reactive sites that are occupied ()0 ≤ θ ≤ 1 in time τ, 312
kobs is the observed binding rate constant, which depends possibly on bulk concentration
of the binding entity (Cb0 ), BA spores. In Figure 17.4, we noted earlier that the total sensor response to BA spore was approximately the same in varying amounts of BT and
BC. Since only the time taken to reach the steady state frequency increased with increasing BT and BC concentrations, it is reasonable to modify the observed binding rate constant in Eq. (17.3) to account for the slower kinetics as:
k − obs τ θ = 1 − e α (17-4) where α is a parameter that characterizes the hindrance. Since the sensor response (Δf ) is proportional to mass of antigen attached the above can by written as:
Δf ⎛ k obs τ ⎞ ()⎜ α ⎟ θ = = ⎜1 − e ⎟ (17-5) ()Δf ∞ ⎝ ⎠
where ()Δf is the change in resonant frequency at time, τ , and (Δf ∞ ) is the steady state resonant frequency change. Eq.(17-5) can be rearranged to:
⎛ ()()Δf ∞ − Δf ⎞ k obs ln⎜ ⎟ = − τ (17-6) ⎝ ()Δf ∞ ⎠ α
From the above the characteristic binding rate constant kobs during initial time (far from equilibrium) can be determined by fitting the data of the pure BA sample, because the hindrance factor (α) by definition is unity. Fitting the experimental data of the pure BA binding presented in Figure 17.4 to Eq. (17-6) gave a straight line with a slope of 0.15 min-1 and correlation coefficient of 0.98, and is shown in Figure 17.6A. We limit the analysis to the first 10-12 minutes to avoid having to deal with diffusion effects. A summary of the analysis for the various cases is given in Table 17.1. In Figure 17.6B we 313 note that non-antigenic bacillus species (BT and BC) concentration increases α in a nearly exponential fashion. At low concentration of BT and BC (333 to 33300
BT+BC/mL) the hindrance effect seems to increase monotonically with concentration.
However, at higher concentration the hindrance factor increased more rapidly, suggesting that the transport of BA spores from the bulk solution to the sensor surface is a strong function of BT and BC concentration. The results we present here indicate that when BA spores are present in a matrix containing copious amounts of non-antigenic particulate matter the binding kinetics is affected rather strongly, while their effects on steady state response is relatively small.
17.6 Conclusion
In this Chapter we have shown the detection of Bacillus anthracis (BA) spores in presence of copious amounts of other bacillus species under liquid immersed flow conditions using PZT-anchored piezoelectric-excited millimeter-sized cantilever
(PAPEMC) sensor. The sensor showed an exponential decrease in its resonant frequency during BA spores binding. When copious amounts of non-antigenic bacillus species were present the attachment kinetics of BA spore decreased. However, the total change in resonant frequency remained the same .The presence of other bacillus species did not prevent binding but impeded transport. The attachment kinetics of BA to the sensor follows the Langmuir model. The observed binding rate constant was 0.15 min-1. The effect of non-antigenic concentration on the binding kinetics of BA spore was modeled by a hindrance factor ()α . The hindrance factor was shown to increase exponentially with an increase of BT and BC concentrations in the range investigated. 314
Table 17.1: Composition of mixed spore samples containing Bacillus anthracis, Bacillus thuringiensis and Bacllus cereus, sensor response, and hindrance coefficient.
Sample BA BT BC Δf∞ α Designation Spores/mL Spores/mL Spores /mL [Hz]
A 333 0 0 2,742 ± 38 1 B 333 166 166 3,053 ± 19 3.52 C 333 1,665 1,665 2,777 ± 26 4.53 D 333 16,650 16,650 2,953 ± 24 6.13 E 333 166,500 166,500 3,105 ± 27 11.04 F 0 166,500 166,500 14 ± 31 ------
315
Table 17.2: Experimentally determined mass change sensitivity in a vacuum (60 mTorr) and the estimated liquid sensitivity using paraffin wax.
Sensitivity (σ nv ) in Estimated Sensitivity in Peaks location vacuum by Wax liquid (σ nf ) [fg/Hz] [kHz] deposition, [fg/Hz] 102.05 11.01 ± 3.01 24.33 ± 6.25 970.05 1.47 ± 0.42 2.89 ± 0.83 1808.05 1.21 ± 0.02 2.11± 0.03 316
1.5 mm
Adhesive Glass PZT
m m 0 2.3 mm 1.
Figure 17.1: Schematic of PZT-anchored PEMC sensor. 317
50
fR=1.81 MHz 30 Q=10 A 10
-10 fR=102 kHz Q=20 fR=970 kHz -30 Q=25
-50 Phase angle [degrees]
-70
-90 4.0E+04 5.4E+05 1.0E+06 1.5E+06 2.0E+06 2.5E+06
Excitation frequency [Hz]
-10 In Air -20 B -30 fR=970 kHz -40 Q=25
-50 In PBS
-60 fR=835.5 kHz Q=15
Phase angle [degrees] -70
-80
-90 700 800 900 1,000 1,100 1,200
Excitation frequency [kHz]
Figure 17.2: Panel A: Resonant spectrum, a plot of phase angle versus excitation frequency, of the PAPEMC sensor investigated. The spectrum is typical of PAPEMC sensors. Resonant peak of 970 kHz was used in all the experiments. Panel B: Resonant peak of PAPEMC sensor in air (970 kHz, right) and totally submerged in PBS (835.6 kHz, left). 318
4.50E-12
4.00E-12
3.50E-12 Mass sensitivity (σ) = -1.47E-15 g/Hz
3.00E-12
2.50E-12
2.00E-12
1.50E-12 Mass change [pg] 1.00E-12
5.00E-13
0.00E+00 -3,000 -2,500 -2,000 -1,500 -1,000 -500 0
Resonant frequency change [Hz]
Figure 3: Experimental determination of the mass change sensitivity of the 970 kHz peak in a vacuum using paraffin wax. The sensitivity was determined as 1.47 ± 0.42 fg/Hz. 319
0.50 1:1 is 333 BA: 166 BT+166 BC cells/mL
0.00 1:0 BA:BT+BC 0:1 1:1 BA:BT+BC -0.50 1:10 BA:BT+BC 1:100 BA:BT+BC -1.00 1:1000 BA:BT+BC 0:1 BA:BT+BC -1.50
-2.00
-2.50 Resonant frequency change [kHz] Resonant
1:0 1:10 -3.00 1:100 1:1 1:1,000 -3.50 0 20 40 60 80 100 120 140 160 180 200 220 240 260 Time [min]
Figure 17.4: Transient response of PZT-anchored PEMC sensor to the binding of 333 BA spores/mL from a solution containing various amounts of other Bacillus species. The control response shown is that of the anti-BA functionalized PAPEMC exposed to a mixture of BT and BC spores, at concentrations of 166 BT/mL and 166 BC/mL, so as to establish the baseline frequency change of the sensor. 320
0.50 PBS PBS BA binding at 330/mL Release solution 0.00 PBS
-0.50
-1.00
-1.50
-2.00
-2.50
Resonant frequency change [kHz] change frequency Resonant -3.00
-3.50 0 20 40 60 80 100 120 140 Time [min]
Figure 17.5: Transient frequency response of PAPEMC sensor to the binding of 333 BA spores/mL and the release of the bound spores by exposure to low pH buffer, pH 1.8. 321
0.0
-0.2
-0.4 )
∞ -0.6 f
Δ
f)/ -0.8 1:0 BA:BT+BC Δ - ∞
f 1:1 BA:BT+BC Δ -1.0 1:10 BA:BT+BC
ln(( A -1.2 1:100 BA:BT+BC 1:1000 BA:BT+BC -1.4
-1.6 024681012
Time , τ [min]
12
10
8 B
α 6
4
2
0 100 1,000 10,000 100,000 1,000,000 BT & BC concentration [cells/mL]
Figure 17.6: Panel A: Langmuir kinetic analysis of BA binding on PAPEMC sensor surface. The initial kinetic analysis of the various BA detection experiments with correlation coefficient ranged from 0.95 to 0.99. The slope of each line gives the observed characteristic binding rate constant kobs. Panel B: Hindrance factor as a function of non- antigenic Bacillus species (BT and BC) to the transport of BA to the cantilever sensing surface. 322
Chapter 18: A Method for Measuring Escherichia Coli O157:H7 At 1 cell/mL in 1 Liter Sample Using Antibody Functionalized Piezoelectric-Excited Millimeter- Sized Cantilever Sensor
18.1 Introduction
The effect of flow rate on the detection of very low pathogen concentration has never been reported in literature and is important in applications where the sample volume is large. In this Chapter we investigate the detection of Escherichia Coli (EC) O157:H7 at 1 cell/mL in 1 liter sample at various flow rates, and at very high flow rate. Earlier, we reported the batch detection of Escherichia coli (EC) O157:H7 at 700 cells per mL with good certainty and repeatability using the piezoelectric-excited millimeter-sized cantilever (PEMC) sensors. We also showed that PEMC sensors have the sensitivity to monitor the viability and measure specific growth rate of EC in Chapter 8. In Chapter 9,
PEMC sensors of higher sensitivity were developed and used for continuous detection of
Bacillus anthracis spores under both batch and flow configuration at 300/mL. Detection of a single cell has been the focus of several researchers, specifically in the case of drinking water.282 Some of the sensor platforms that have been employed are the atomic force microbalance (AFM),114,118 light-addressable potentiometric sensor (LAPS),345 and porous polymers.346 Although these techniques showed results that are reliable, they do not lend themselves to detecting very low pathogen concentration. The infectious dosage of EC O157:H7 has been reported as 10 cells and the United States Environmental
Protection Agency standard for drinking water is 40 cells per liter.165,167
In this Chapter, we describe three approaches for the detection of EC at 1 cell/mL using piezoelectric-excited millimeter-sized cantilevers: first a batch configuration, 323 second at modest flow rates, and finally a flow and stop measurement at high flow rate.
Flow was used to increase the contact of the pathogen with the sensor since the pathogen count was very low. The batch experiments were carried out with 1 EC in 1 mL of buffer, while the flow experiments were done using a 1 liter sample buffer containing 1000 EC.
Due to the size of the sensor (a few mm2), dipping it into a large volume of liquid is not an effective means of contacting the target analyte with the sensor surface, particularly for particulate antigen such as EC.
18.2 Materials and Methods
18.2.1 Sensors
Two different PEMC sensor designs were used in this investigation: PEMC-a of design Figure 4.1B and PEMC-b of design Figure 4.1C. Details on their fabrication can be found in Chapter 4. The dimensions of both the PZT and borosilicate glass in PEMC-a were 1 ± 0.05 x 1 ± 0.05 x 0.127 ± 0.005 mm3 (L x W x t) and 3 ± 0.05 x 1 ± 0.05 x
0.160 ± 0.005 mm3, respectively. PEMC-b design has only the PZT anchored at one end and at the free-end a 1.65 x 1.0 x 0.60 mm3 (L x W x t) fused quartz layer was attached.
The PZT dimensions were 2.54 x 1.0 x 0.127 mm3. The length of the PZT monolayer between the bi-layer and the anchored section was 0.89 mm. The exposed PZT surfaces were coated with a thin layer of polyurethane (~ 30 μm thick) for electrical insulation.
Also, in each cantilever design a 1 mm length of the PZT layer was epoxied behind the anchored point to which the top and bottom electrodes were soldered.
18.2.2 Experimental Setup
The experiments were carried out in a flow configuration as shown in Figure 18.1, the flow apparatus consisted of a one-liter sample, antibody reservoir, hydroxylamine (Sigma 324
Aldrich) reservoir, phosphate buffered saline (PBS, Sigma-Aldrich) reservoir, peristaltic pumps, and a sensor flow cell (SFC). The EC sample container (1 L) was stirred using a
2˝ magnetic stirrer (at 150 rpm) to ensure concentration homogeneity. PEMC-b was installed vertically into the SFC containing PBS and was secured by a Swagelok® fitting.
For batch studies PEMC-a was used. PEMC-a was mounted on a XYZ position stage
(Optosigma Corporation, Santa Ana, CA) between two vertical parallel plates such that the sensor was positioned perpendicular to the table on which the experiment was setup
(see a schematic in Figure 12.1). The cantilever was positioned vertically into the sample liquid at 2 mm immersion depth with ± 10 µm accuracy. The sensors were connected to an impedance analyzer (Agilent, HP 4192A). Both the impedance and phase angle measurements were acquired at an excitation voltage of 100 mV at various frequencies with a LabVIEW® data acquisition application.
18.2.3 Experimental procedure
The cantilevers glass surfaces were cleaned and functionalized with 3-aminopropyl- triethoxysilane (APTES, Sigma-Aldrich) and subsequently antibody to EC (affinity purified anti-EC, KPL, Gaithersburg, MD) was immobilized as discussed in previous
Chapters. Working solution of formaldehyde-killed EC (strain B1409, USDA, PA) was prepared from a stock solution of 1 x 109 EC cells/mL by serial dilution in phosphate buffered saline (PBS; 10 mM, pH 7.4) to a final concentration of 70, 10, and 1 EC/mL.
Two types of experiment were carried out in this investigation: batch detection at 70,
10, and 1 EC in 1 mL of PBS buffer using PEMC-a and flow detection of 1 cell/mL in 1 liter of PBS buffer using PEMC-b. The flow experiments were conducted in two modes.
In the first, various flow rates were used and continuous measurements of resonant 325 frequency was performed. In the second, a high flow rate was used for a specific time period and then stopped for sufficient time to measure resonant frequency. The latter method was used as the measurements were too noisy at flow rates greater than 5 mL/min. The flow circuitry was first primed with PBS to remove any air bubbles. Then, the APTES functionalized cantilever was installed into the SFC containing PBS (running buffer). The running buffer was re-circulated through the SFC (by opening V2) at 1.5 mL/min and the sensor’s resonant frequency was monitored until it reached a steady value, which was usually achieved in 10 to 20 minutes. Upon stabilization, activated anti-
EC solution at 10 μg/mL was re-circulated through the SFC at 1.5 mL/min by opening valve V3 and closing V2 to allow antibody immobilization for 2 h at 25 ± 0.2 °C.
Hydroxylamine was flowed through the SFC (by opening V4), and then PBS (by opening valve V2 and closing V4) to flush the system. Upon completion of the rinsing step, the
EC sample was flowed once through the SFC by opening V1 and closing V2. The EC detection experiments were carried out at flow rates of 1.5, 2.5, 3 and 17 mL/min. The experiments carried out at 1.5, 2.5 and 3 mL/min were done under continuous flow conditions. The bound EC was released from the sensor by a low pH solution. For the highest flow rate (17 mL/min) the EC sample was flowed for five minutes, stopped, and measurements were made of three resonant peaks. This procedure was repeated until the entire 1 L sample solution was flown through the SFC. A control experiment was carried out for each detection study using an anti-EC functionalized sensor exposed to PBS solution flowing at a rate identical to the detection experiment.
18.3 Results and Discussion
326
18.3.1 Resonance Spectra
A typical resonant spectrum of PEMC-a sensor in both air and liquid can be found in
Chapter 10. The spectrum, a plot of phase angle versus excitation frequency, showed only two dominant bending modes in the frequency range of 1 to 100 kHz. Basically, the first or fundamental resonance occurred between 14 and 25 kHz, and the second bending mode between 50 to 100 kHz. For the sensor used in this investigation the first and second bending mode frequencies in air were 21.5 kHz and 85.5 kHz, respectively. Upon liquid immersion the fundamental mode decreased by ~ 7.5 kHz, while the second mode decreased by ~ 20 kHz. Dominant higher order resonant modes beyond 100 kHz were rarely seen in PEMC-a design. In Chapter 6, we showed that higher modes are more sensitive and therefore, in this study the second resonant bending mode was used to monitor the detection of EC. The resonant spectrum of PEMC-b in air showed resonance peaks upon to 2 MHz, which is a distinguishing feature compared to PEMC-a. In Figure
18.2 the resonant spectra of PEMC-b in both air and under liquid are presented. The fundamental resonant frequency in air was 51.2 kHz which is two to three times higher than PEMC-a sensor. The resonant frequencies of the dominant modes in air (blue curve) were located at 51.2, 186.5, 883.5, and 1778.2 kHz, respectively. Upon fully submerging the sensor in PBS, the resonant frequencies shifted to the left (pink curve). The resonant frequencies of the dominant modes decreased to 37.5, 162.5, 800.0, and 1725.1 kHz, respectively. The decrease in the resonant frequencies is due to the added mass effect, since the liquid adjacent to sensor also oscillates with it. The difference in frequency between air and liquid is a measure of resonant peak sensitivity. That is, a larger frequency change generally implies a greater level of sensitivity. Therefore, a change in 327 frequency of 83.5 kHz for the third dominant resonant peak suggests that it is the most sensitive peak among the identified four.
The sensor behaves as a capacitor at non-resonant frequencies, and thus its baseline phase angle is close to -90°. At resonance, there is a shape increase in phase angle as the
PZT becomes more resistive due to increase charge accumulation as a result of higher than normal mechanical stress. Under liquid immersion the height of each resonant peak decreased due to viscous damping, illustrated in Figure 18.2 (pink curve). The base of a resonant peak is a measure of the dissipative energy of the sensor in the surrounding medium and is defined by the peak’s quality factor (Q value). Q is determined from the ratio of resonant frequency to the peak width at half the peak height. In other words, the
Q value gives a measure of peak sharpness. The Q-values of the dominant resonant peaks
(51.2, 186.5, 883.5, and 1778.2 kHz) of PEMC-b in air were 63.8, 15.3, 51.9, and 18.9, respectively, and in liquid the Q-values decreased to 46.9, 13.6, 32.0, and 11.0. For
PEMC-a the Q-value in air of the first and second modes were 29 and 38, respectively, and 25 and 32 under liquid. One notes that the quality factor of the fundamental bending mode was larger in PEMC-b and furthermore, the fundamental mode is located at a higher frequency, which is an indication that a large stress is generated at resonance in
PEMC-b. However, a Q value change of up to 40% was observed in PEMC-b and only a maximum of ~ 20% was seen in PEMC-a design. This is an important advantage of
PEMC sensors over microcantilever under liquid conditions. For microcantilever the Q- value reduced to about 1 under liquid immersion.347 In addition, a stable resonant peak with a high Q-value will enable a more accurate determination of resonant frequency. In this study, the 65.5 kHz peak of PEMC-a, and the 162.5, 800.0, and 1725.1 kHz resonant 328 peaks of PEMC-b under liquid were used to detect EC. We carried out each of the detection experiments at least twice and the data shown is typical of the results obtained.
18.3.2 Batch Detection of EC Using PEMC-a
The functionalization of the sensor with antibody for the detection of pathogen EC was done through covalent bonds. APTES reacts with the hydroxyl groups on the glass surface leaving an exposed amine group, which binds covalently with the activated carboxyl groups of the anti-EC. One milliliter of PBS buffer containing 70, 10, and 1 EC were prepared and antibody functionalized PEMC-a was immersed 2 mm for ~ 55 minutes. The frequency response of EC binding to the cantilever is presented in Figure
18.3. These experiments were repeated 8 times. The results showed an increase in the total frequency change with the increase in EC concentration and correspondingly, the binding rate also increased. The total frequency change observed for 70 and 10 EC/mL was 58 ± 5 (n=8) Hz and 19 ± 2 (n=8) Hz, respectively. For the 1 cell/mL EC sample no significant change (0 ± 1 Hz) in the resonant frequency was observed. That is the response was similar to the control (0 ± 2 Hz), which may suggest one of three things: first, the cantilever was not sensitive enough to detect 1 cell and secondly, the cell did not attach to the sensor. Thirdly, there was no cell in the sample. To eliminate the latter, the 1
EC/mL experiment was repeated 7 times and no response was observed in any of the experiments. Both the 70 and 10 EC/mL experiments were carried out 8 times. The 70
EC/mL sample gave positive detector response 6 out of the 8 times, and the 10 EC/mL results gave positive detection in only 2 of the 8 experiments. These results indicate that the detection of low pathogen concentration in a batch configuration is not an effective modality because the cells may not be in close proximity with the sensor for attachment. 329
By introducing flow, one can overcome the transport of the pathogen to the sensor surface, which we investigate next.
18.3.3 Detection of EC Using PEMC-b At 1 EC/mL Under Various Flow Rates
The response of PEMC-b, using the 800.0 kHz peak under full liquid immersion, to the binding of EC at various flow rates is presented in Figure 18.4A. The freshly prepared antibody-functionalized sensor was exposed to a 1 L sample of EC at 1 cell/mL and sample flow rates of 1.5, 2.5, and 3 mL/min. In all the experiments, the sensor responded with an initial rapid decrease in its resonant frequency followed by a slower decrease until the frequency reached a steady value. The rate of EC binding and the total resonant frequency change increased with increasing flow rate. The steady state frequency changes for the binding of EC (1 cell/mL) at 1.5, 2.5, and 3 mL/min were 2,229.5 ± 11.2, 3,068.9
± 45.1, and 4,685.6 ± 96.6 Hz, respectively. We note that the time to reach steady state is a weak function of the flow rate used. The steady state frequencies at 1.5, 2.5, and 3 mL/min were achieved in 192, 230, and 290 minutes, respectively. Since the number of
EC in the samples were equal and about 20 minutes before the EC sample flow was initiated the resonant frequency was constant, it is reasonable to conclude that higher flow rate improved contact of EC with the sensor. At each flow rate a control was also carried out. The control was an antibody functionalized cantilever exposed to PBS buffer at the same flow rate as the detection experiment. The control shown in Figure 18.4A was carried out at 3 mL/min. The frequency change decreased slightly 77 ± 81 Hz, which is slightly higher than the noise level we observed with cantilevers of this design. However, we note that the change is insignificant when compared to the frequency response of EC 330 binding. The data in Figure 4A suggests that PEMC sensors have the sensitivity to detect
EC at 1 EC/mL in a flow configuration.
To confirm the binding of EC to the cantilever, in each experiment, after completion of the detection segment the flow circuit was rinsed with PBS followed by a pH 2.02 solution, and finally PBS was re-introduced to return the environment prior to EC attachment. Figure 18.4B shows the release sensor response for the bound EC from 1.5 mL/min sample. The PBS rinsed after the EC attachment did not show any noticeable change in frequency. In fact, the frequency increased by only 30 Hz, which is below the control response. Conversely, exposure of the sensor to the release solution resulted in an immediate increase in frequency and a constant frequency change of 1,904.8 ± 16.8 Hz was reached. One notes that the total frequency change obtained during the release is less than that of the attachment response; this may be due to the high affinity of the antibody to the antigen that resulted in an incomplete release of the bound EC. However, flushing the SFC with PBS afterwards caused a further increase in the resonant frequency by 392
± 20 Hz, indicating that the remaining EC, after the release episode, was detached. In addition, the final frequency change reached was ~ 110 Hz higher than the value before initiation of the EC sample. It is reasonable to conclude that this difference is small considering the magnitude of EC response and the variation observed in the control studies.
18.3.4 EC Detection In A Flow and Stop Modality
In this modality we show the use of multiple resonant peaks in the detection of EC.
Because resonance characteristics of PEMC sensors are very unstable under high liquid flow rates (17 mL/min), we carried out the experiment in a flow-stop-measure modality. 331
We note that the sample volume was 1 liter and the EC sample (1 cell/mL) was flowed at
17 mL/min. We also measured the behavior of the sensor’s three higher-order resonant modes by a constant mass. The primary motivation is that the frequency response will give a measure of the sensitivity of the various peaks. In addition, a non-flow condition ensures that the measured resonant frequency changes were not due to flow effects, and only mass changes caused by EC attachment.
In Figure 18.5, the 162.5, 800.0, and 1,725.1 kHz resonant peaks, under liquid immersion, were used in the detection of EC (at 1 EC/mL). The plot shows the frequency response of the sensor versus the sample volume that flowed through the SFC prior to resonant frequency measurement. The sample was flowed for 5 minutes, stopped, and then the resonant frequencies of the three modes were monitored individually until they stabilized, which typically took place in 5 to 10 minutes. This step was repeated until the sample volume was completely flown through the SFC. As shown in the figure, the resonant frequencies of the different modes decreased rapidly after the initial 5 minutes of sample flow (sample volume of 85 mL) and ultimately reached a steady state total frequency change of 4,340 ± 50, 7,188 ± 50, and 25,850 ± 60 Hz corresponding to the
162.5, 800.0, and 1725.1 kHz resonant peaks under liquid immersion. The significance of these results is that for the same mass change different resonant modes responded differently. For each detection experiment a control was conducted in the same fashion.
The control experiments, including an antibody functionalized PEMC-b sensor exposed to PBS buffer solution, in each case the resonant frequency fluctuated around zero; 56 ±
154, 97 ± 225, and 21 ± 328 Hz for the 162.5, 800.0, and 1,725.1 kHz resonant peaks under liquid immersion, respectively. If we compare the frequency response presented in 332
Figure 18.4, for the 800.0 kHz peak under liquid immersion, a larger change in frequency is observed at 17 mL/min, which further confirms that higher flow rate increases the EC attachment. These results suggest that the cantilever was mechanically robust and at the same time very sensitive.
18.3.5 Kinetics of EC O157:H7 binding
The binding kinetics of EC has been characterized in earlier chapters using Langmuir kinetic model. The model is expressed as Eq. (9-4). A fit of the experimental data presented in Figure 18.4A to the kinetic model yielded straight lines of correlation coefficient ranging from 0.986 to 0.996 and is shown in Figure 18.6. We limit the
analysis to the first 10 minutes to avoid diffusion effects. The kobs values were determined to be 0.021, 0.015, and 0.009 min-1 corresponding to sample flow rates of 3,
2.5, and 1.5 mL/min. These results suggest that flow increases contact with sensor surface, and as a result increase the initial rate of EC attachment.
18.4 Conclusion
This Chapter demonstrates the detection of 1 EC/mL at various sample flow rates using antibody functionalized piezoelectric-excited millimeter-sized cantilever sensor.
Detection of 1 EC/mL in a batch configuration did not show a positive response. On the other hand, flow increased the sensor response and provided a reliable measurement modality. In addition, we observed that binding rate increased with flow rate. We also found that the time to reach a steady state frequency is a weak function of flow rate. We also demonstrated that higher-order resonant modes are more sensitive than lower-order modes for the same mass change.
333
Circulation tube
Cantilever 5.2 Polylysine
80,000 -50
70,000 -55
60,000 -60
50,000 -65
40,000 -70
30,000 -75 Amp litud e (ohm s) 20,000 -80 (deg) Angle Phase
10,000 -85
0 -90 0 50 100 150 200 250 Frequency
Impedance (ohms) Phase (deg)
On/Off valves Peristaltic pump #2
V1 V2 V3 V4 Sample outlet Sample
Sample container 4-Port manifold
Heating fluid inlet
Stirrer Heating fluid outlet
Sample inlet
Peristaltic pump #1
Constant temperature bath
Figure 18.1: Schematic of experimental setup. 334
50
30 In Air
10
-10 In PBS
-30
Phase angle [degrees] angle Phase -50
-70
-90 0.0E+00 4.0E+05 8.0E+05 1.2E+06 1.6E+06 2.0E+06 Excitation frequency [Hz]
Figure 18.2: Resonant spectra of, phase angle versus excitation frequencies, PEMC-b sensor in air (blue) and fully immersed in PBS (pink). At resonance, the phase angle of the oscillating cantilever (100 mV excitation) moves away from -90°. Under liquid the resonant peaks shifted to the left. 335
10
0
-10
-20
-30 70 cells/mL 10 cells/mL -40 1 cell/mL Control -50 Resonant frequency change[Hz] -60
-70 0 102030405060
Time [min]
Figure 18.3: Frequency response of PEMC-a to the detection of EC O157:H7 in a batch configuration at 70, 10 and 1 EC in 1 mL of PBS buffer. No response was observed for the 1 EC sample in over six repeated experiments, suggesting that it was almost impossible to detect 1 EC in 1 mL of sample buffer. 336
1,000 Initiation of EC sample
0 Control
-1,000
-2,000 1. 5 mL/min
2.5 mL/min -3,000 Resonant frequency change [Hz] -4,000 A 3 mL/min
-5,000 0 50 100 150 200 250 300 350 Time [min]
500 S Sample initiated at 1.5 mL/min 0 PB
-500
-1,000
-1,500 S Release solution Release PB -2,000 B Resonant frequency change [Hz] change frequency Resonant
-2,500 0 100 200 300 400 500 600 Time [min]
Figure 18.4: Panel A: Resonant frequency change, of the 800.0 kHz peak under liquid, for the binding of E.coli (concentration 1 EC/mL in 1 L buffer solution) to PEMC-b. The sample liquid was flowed at 1.5, 2.5, and 3 mL/min. The control was an antibody functionalized PEMC-b sensor exposed to 1 L of PBS buffer flowing at 3 mL/min. Panel B: Sensor response to the attachment of EC at sample flow rate of 1.5 mL/min and the subsequent release of the bound EC. 337
2500 0 -2500 -5000 -7500
-10000 Control-187kHz -12500 Control-884kHz -15000 Control-1.78MHz EC-187kHz -17500 EC-884kHz -20000 EC-1.78MHz
Resonant frequency change [Hz] change frequency Resonant -22500 -25000 -27500 0 200 400 600 800 1000 1200
Sample volume flown [mL]
Figure 18.5: Resonant frequency response of PEMC-b sensor to the binding of EC O157:H7 at concentration of 1 EC/mL (total sample volume was 1 L) for the dominant resonant modes investigated. The control experiments were an anti-EC PEMC-b sensor exposed to only PBS solution in the same fashion as the EC sample.
338
0.00
-0.05 ) ∞ f -0.10 Δ f)
Δ - ∞ f Δ -0.15 ln(( 3 mL/min -0.20 2.5 mL/min 1.5 mL/min
-0.25 024681012 τ [min]
Figure 18.6: Initial kinetic analysis of EC attachment at various flow rates. The analysis is of the experiment data presented in Figure 4A. The correlation coefficient ranged from 0.983 to 0.996. The slope of each line gives the observed characteristic binding rate constant kobs. 339
Chapter 19: Future Work
This thesis work presented a novel approach for the continous detection of biologics at very low concentrations in the liquid phase of both batch and flow configuration. To evaluate the feasibility of PEMC sensors in detecting a single molecule or a virus is of great interest, and therefore, things that could enhance the findings of this research are summarized below.
1) Sensitivity
Typically, the experimental determination of PEMC sensors mass change sensitivity
in both air and liquid is on the picograms levels. However, we are able to detect tens
of attograms molecules. Some of the experiments that can be done to enhance the
experimental measure of sensitivity are:
• Attach known mass of protein (example BSA) to the cantilever as (1) a point mass
and (2) a distributed mass.
• Investigate the sensitivity in both air and under liquid.
• Explore various cantilever geometries such as a beam because the sensor would
be anchored on both sides this may enhance its resonant frequencies. One can also
investigate a triangular cantilever with and without a hole in the tip.
• Use known amount of congujated polystyrene spheres to increase the weight on
the sensor.
• After the addition of a known mass make measurements of sensor deflection and
compute its mass change sensitivity.
340
2) Optimize the immobilization process
In this thesis work, the estimated maximum surface coverage of antigen detected was
approximately 40%. Therefore, there is need to optimize the immobilization protocol.
• The detection of an antigen involves functionalization of the cantilever with
recognition molecules. For a glass surface, the surface is first aminated (amine-
terminated silane) followed by the covalent linking of an antibody. To enhance
antigen coverage the amination protocol as to be optimized so that the antibody
can be oriented properly; Fab region point away from sensor surface.
• Surface coverage can also be enhanced by increasing the contact of the analyte
with the sensor by stirring or at optimium flow rate for a particular biologic.
3) Minimize false alarm; whether positive or neative.
• The best way of minimizing false alarms is by using a reference sensor.
• The reference sensor could be independently operated or can be in parallel to the
sample detector. The advantage of the parallel connection is that only one
instrument will be needed or one does not have to switch between sensors when
taking a measurement (example with a control box); as this will introduce some
noise in the signal.
• Also in parallel the acquisition program can be configured so that the reference
signal is substracted for the detection response automatically.
• The only drawback to this technique is that both sensors (reference and sample)
have to have resonance a few kHz apart and therefore, nearly identical
sensitivities so the upon immersion the resonance peaks does not merge and will
be separated so that frequency shift between the two peaks can be obtained. 341
List of References
1. Binning, G.; Quate, C. F.; Gerver, Ch., 1986. Phys. Rev. Lett. 930-935.
2. Gupta, A.; Akin, D.; Bashir, R., 2004. Appl. Phys. Lett. 84, 1976-1978.
3. Ilic, B.; Yang, Y.; Graighead, H. G. 2004. Appl. Phys. Lett. 85, 2604-2606.
4. Nie S., Zare R. N., 1997. Annu. Rev. Biophys. Biomol. Struct. 26, 567-96.
5. http://optics.org/articles/news/8/3/27/1
6. Hartman P. A., Swaminathan B., Curiale M. S., Firstenberg-Eden R. A., Sharpe N., Cox N. A., Fung D. Y. C., Goldschimidt M. C., 1992. Rapid Methods and Automation. In: C Vanderzant and DF Splittstoesser (Eds.). Compendium of Methods for the Microbiological Examination of Foods 3rd ed. American Public Health Association, Washington, DC, 665-746.
7. Park S-J., Taton T. A., Mirkin C. A., 2002. Science 295, 1503-1506.
8. J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H.-J. Guntherodt, Ch. Gerber, J. K Gimzewski, Science, 288, 316-8.
9. Schemmel H. , Gaud E., 1999. Rev. Sci. Instrum. 70, 1313-1317.
10. Baselt D. R., Lee G. U., Colton R. J., 1996. J. Vac. Sci. Technol. B 14, 798-803.
11. Lee G. U., Kidwell D. A., Colton R. J., 1996. U. S. Pat. Appl. 38 pp. Avail. NTIS Order No. PAT-APPL-8-505547.
12. Han W., Lindsay S. M., Jing T., 1996. Appl. Phys. Lett. 69, 4111-4113.
13. Sakti S. P., Rösler S., Lucklum R., Hauptmann P., Bühling F., Ansorge S., 1999. Sensors and Actuators, 76, 98.
14. Ward M. D., Buttry D. A., 1990. Science, 249, 1000-1007.
15. Lin Z., Yip C. M., Josheph I. S., Ward M. D., 1993. Anal. Chem., 65, 1546-1551.
342
16. Berger, R.; Delamarche, E.; Lang, H. P.; Gerber, C.; Gimzewski, J. K.; Meyer, E.; Guntherodt, H. J., 1998. Appl. Phys. A 66, S55-S59.
17. Chen, G. Y.; Thundat, T.; Wachter, E. A.; Warmack, R. J., 1995. J. Appl. Phys. 77, 3618-3622.
18. Lang, H. P.; Baller, M. K.; Berger, R.; Gerber, C.; Gimzewski, J. K.; Battiston, F. M.; Fornaro, P., Ramseyer, J. P., Meyer, E.; Guntherodt, H. J., 1995. Anal. Chim. Acta. 393, 59-65.
19. Thundat, T.; Warmack, R. J.; Chen, G. Y.; Allison, D. P., 1994. Appl. Phys. Lett. 64, 2894-2896.
20. Thundat, T.; Chen, G. Y.; Warmack, R. J. Allison, D. P.; Wachter, E. A., 1995. Anal. Chem. 67, 519-521.
21. Raiteri, R.; Grattarola, M.; Butt, H.-J.; Skladal, P., 2001. Sensors and Actuators B 79, 115-126.
22. Preissing, F. J., 1989. J. Appl. Phys. 66, 4262.
23. Rugar, D.; Mamin, H. J.; Erlandsson, R.;Stern, J. E.; Terris, B. D., 1988. Rev. Sci. Instrum. 59, 2337-2340.
24. Blanc, N.; Brugger, J.; de Rooij, N. F.; Duerig, U., 1996. J. Vac. Sci. Technol. B 14, 901.
25. Miller, S.; Turner, K.; MacDonald, N., 1997. Rev. Sci. Instrum. 68, 4155.
26. Martinez, R. E.; Augustyniak, W. M.; Golovchenko, J. A., 1990. Phys. Rev. Lett. 64, 1035-1038.
27. Lai, J.; Perazzo, T.; Shi, Z.; Majumdar, A. 1997. Sensors and Actuators A 58, 113-119.
343
28. Gimzewski, J. K.; Gerber, C.; Meyer, E.; Schlittler, R. R., 1994. Chem. Phys. Lett. 217, 589-594.
29. Reid, S.; Cagnoli, G.; Crooks, D. R. M.; Hough, J.; Murray, P.; Rowan, S.; Fejer, M. M.; Route, R.; Zappe, S. 2006. Phys. Lett. A, AID 15103.
30. Marie, R.; Thaysen, J.; Christensen, C. B. V.; Boisen, A. 2003. Microelect. Eng. 67-68, 893-898.
31. Yang, Z.; Li, X.; Wang, Y.; Bao, H.; Liu, M. 2004. Microelect. J. 35, 479-483.
32. Chua, D. H. C.; Tay, B. K.; Zhang, P.; Teo, E. H. T.; Lim, L. T. W.; O’Shea, S.; Miao, J.; Milne, W. I., 2004. Diamond & Related Materials 13, 1980-1983.
33. Mertens, J.; Finot, E.; Thundat, T.; Fabre, A.; Nadal, M.-H.; Eyraud, V.; Bourillot, E., 2003. Ultramicroscopy 97, 119-126.
34. Ansorge, E.; Schimpf, S.; Hirsch, S.; Sauerwald, J.; Fritze, H.; Schmidt, B. 2006. Sensors and Actuators A, SNA-5130.
35. Hiroya, T.; Inagaki, K.; Tanda, S.; Tsuneta, T.; Yamaya, K., 2003. Pysica B 329- 33, 1635-1637.
36. Lai, J.; Perazzo, T.; Shi, Z.; Majumdar, A., 1996. Proceedings of ASME Conference on Micro-electro-mechanical Systems (MEMS), 55-60.
37. Datskos, P. G.; Oden, P. I.; Thundat, T.; Wachter, E. A.; Hunter, S. R., 1996. Appl. Phys. Lett. 69, 2986-2988.
38. Oden, I.; Datskos, P. G.; Thundat, T.; Wachter, E. A., 1996. Appl. Phys. Lett. 69, 3277-3279.
39. Wachter, E. A.; Thundat, T.; Oden, P. I.; Warmack, R. J.; Datskos, P. G.; Sharp, S. L., 1996. Rev. Sc. Instrum. 67, 3434-3439.
344
40. Markidou, A.; Shih, W. Y.; Shih, W.-H., 2005. Rev. Sci. Instrum. 76, 064302-6.
41. Rasmussen, P. A.; Thaysen, J.; Hansen, O.; Eriksen, S. C.; Boisen, A., 2003. Ultramicroscopy 97, 371-376.
42. Lemoine, P.; Zhao, J. F.; Bell, A.; Maguire, P.; McLaughlin, J., 2001. Diamond and Related Materials 10, 94-98.
43. Tabard-Cossa, V.; Godin, M.; Beaulieu, L. Y.; Grütter, P., 2005. Sensors and Actuators B 107, 233-241.
44. Abu-Lail, N. I.; Kaholek, M.; LaMattina, B.; Clark, R. L.; Zauscher, S., 2005. Sensors and Actuators B, doi: 10.1016/l.snb.2005.06.003.
45. Cardinale, G. F.; Howitt, D. G.; Clift, W. M.; McCarty, K. F.; Medlin, D. L.; Mirkarimi, P. B.; Moody, N. R., 1996. Thin Solid Films 287, 214-219.
46. Khaled, A.-R. A.; Vafai, K.; Yang, M.; Zhang, X.; Ozkan, C. S., 2003. Sensors and Actuators B 94, 103-115.
47. Naik, T.; Longmire, E. K.; Mantell, S. C., 2003. Sensors and Actuators A 102, 240-254.
48. Chen, G. Y.; Warlock, R. J.; Huang, A.; Thundat, T., 1995. J. Appl. Phys. 78, 1465-1469.
49. Ahmed, N.; Nino, D. F.; Moy, V. T., 2001. Rev. Sci. Instrum. 72, 2731-2734.
50. Shih, W. Y.; Li, X.; Gu, H.; Shih, W.-H.; Aksay, I.A., 2001. J. Appl. Phys. 89, 1497-1505.
51. Sader, J. E., 1998. J. Appl. Phys. 84, 64-76.
52. Chu, W.-H., 1963. Tech. Rep. No. 2, DTMB, Contact NObs-86396(X), Southwest Reseacrh Institute, San Antonio, Texas.
345
53. Bergaud, C.; Nicu, L., 2000. Rev. Sci. Instrum. 71, 2487-2491.
54. Goedeke, S. M.; Allison, S. W.; Datskos, P. G., 2004. Sensors and Actuators A, doi:10.1016/j.sna.2003.12.002.
55. Shibata, T.; Unno, K.; Makino, E.; Shimada, S., 2004. Sensors and Actuators A, doi:10.1016/j.sna.2003.11.025.
56. Jeon, Y. B.; Sood, R.; Jeong, J.-H.; Kim, S.-G., 2005. Sensors and Actuators A 122, 16-22.
57. Kanno, I.; Kotera, H.; Wasa, K., 2003. Sensors and Actuators A 107, 68-74.
58. Zhang, Q. Q.; Gross, S. J.; Tadigadapa, S.; Jackson, T. N.; Djuth, F. T.; Trolier- McKinstry, S., 2003. Sensors and Actuators A 105, 91-97.
59. Cattan, E.; Haccart, T.; Vélu, G.; Rémiens, D.; Bergaud, C.; Nicu, L., 1999. Sensors and Actuators 74, 60-64.
60. Itoh, T.; Lee, C.; Chu, J.; Suga, T., 1997. IEEE, 78-83.
61. Brissaud, M.; Ledren, S.; Gonnard, P., 2003. J. Micromech. Microeng. 13, 832- 844.
62. Davis, Z. J.; Abadal, G.; Helbo, B.; Hansen, O.; Campabadal, F.; Pérez-Murano, F.; Esteve, J.; Figueras, E.; Ruiz, R.; Barniol, N.; Boisen, A., 2001. High mass and spatial resolution mass sensor based on resonating nano-cantilevers integrated with CMOS. The 11th International Conference on Solid-State Sensors and Actuators, Munich, Germany, June 10-14.
63. Zurn, S.; Hsieh, M.; Smith, G.; Markus, D.; Zang, M.; Hughes, G.; Nam, Y.; Arik, M.; Polla, D., 2001. Smart Mater. Struct. 10, 252-263.
64. Elmer F.-J.; Dreier, M., 1997. J. Appl. Phys. 81, 7709-7711.
65. Yaman, M., 2004. Materials and Design doi:10.1016/j.matdes.2004.10.009. 346
66. Teva, J.; Abadal, G.; Davis, Z. J.; Verd, J.; Borrise, X.; Boisen, A.; Pérez- Murano, F.; Barniol, N., 2004. Ultramicroscopy 100, 225-232.
67. Passian, A.; Muraligharan, G.; Mehta, A.; Simpson, H.; Ferrel, T. L.; Thaundat, T., 2003. Ultramicroscopy 97, 391-399.
68. Li, X.; Shih, W. Y.; Vartuli, J.; Millus, D. L.; Prud’homme, R.; Aksay, I. A., 2002. J. Appl. Phys. 92, 1-6.
69. Fabian, J.-H.; Scandella, L.; Fuhrmann, H.; Berger, R.; Mezzacasa, T.; Musil, C. H.; Gobrecht, J.; Meyer, E., 2000. Ultramicroscopy 82, 69-77.
70. Abadal, G.; Davis, Z. J.; Borrise, X.; Hansen, O.; Boisen, A.; Barniol, N.; Perez- Murano, F.; Serra, F., 2003. Ultramicroscopy 97, 127-133.
71. Davis, Z. J.; Abadal, G.; Hansen, O.; Borise, X.; Barniol, N.; Perez-Murano, F.; Boisen, A., 2003. Ultramicroscopy 97, 467-472.
72. Adams, J. D.; Rogers, B.; Manning, L.; Hu, Z.; Thundat, T.; Cavazos, H.; Minne, S. C., 2005. Sensors and Actuators A 121, 457-461.
73. Britton Jr., C. L.; Jones, R. L.; Oden, P. I.; Hu, Z.; Warmack, R. J.; Smith, S. F.; Bryan, W. L.; Rochelle, J. M., 2000. Ultramicroscopy 82, 17-21.
74. Maute, M.; Raible, S.; Prins, F. E.; Kern, D. P.; Ulmer, H.; Weimar, U.; Gopel, W., 1999. Sensors and Actuators B 58, 505-511.
75. Fadel, L.; Lochon, F.; Dufour, I.; Francais, O., 2004. J. Micromech. Microeng. 14, S23-S30.
76. Battiston, F. M.; Ramseyer, J.-P.; Lang, H. P.; Baller, M. K.; Gerber, Ch.; Gimzewski, J. K.; Meyer, E.; Guntherodt, H.-J., 2001. Sensors and Actuators B 77, 122-131.
347
77. Kim. B. H.; Prins, F. E.; Kern, D. P.; Raible, S.; Weimar, U., 2001. Sensors and Actuators B 78, 12-18.
78. Betts, T. A.; Tipple, C. A.; Sepaniak, M. J.; Datskos, P. G., 2000. Analytica Chimica Acta 422, 89-99.
79. Porter, T. L.; Eastman, M. P.; Macomber, C.; Delinger, W. G.; Zhine, R., 2003. Ultramicroscopy 97, 365-369.
80. Lang, H. P.; Baller, M. K.; Berger, R.; Gerber, Ch.; Gimzewski, Battison, F. M.; Fornaro, P.; Ramseyer, J. P.; Meyer, E.; Güntherodt, H.-J., 1999. Analytica Chimica Acta 393, 59-65.
81. Headrick, J. J.; Sepaniak, M. J.; Lavrik, N. V.; Datskos, P. G., 2003. Ultrmicroscopy 97, 417-424.
82. Kooser, A.; Gunter, R. L.; Delinger, W. D.; Porter, T. L.; Eastman, M. P., 2004. Sensors and Actuators B, doi: 10.1016/j.snb.2003.12.057.
83. Fabre, A.; Finot, E.; Demoment, J.; Contreras, S., 2003. Ultramicroscopy 97, 425- 432.
84. Chen, G. Y.; Warmack, R. J.; Thundat, T.; Allison, D. P., 1994. Rev. Sci. Instrum. 65, 2532-2537.
85. Zhou, J.; Li, P.; Zhang, S.; Long, Y.; Zhou, F.; Huang, Y.; Yang, P.; Bao, M., 2003. Sensors and Actuators B 94, 337-342.
86. Zhou, J.; Li, P.; Zhang, S.; Huang, Y.; Yang, P.; Bao, M.; Ruan, G., 2003. Microelectronic Engineering 69, 37-46.
87. Zhao, Q.; Zhu, Q.; Shih, W. Y.; Shih, W.-H., 2005. Sensors and Actuators B, doi: 10.10116/jsna.2005.10.048.
88. Pinnaduwage, L. A.; Thundat, T.; Gehl, A.; Wilson, S. D.; Hedden, D. L.; Lareau, R. T., 2004. Ultramicroscopy 100, 211-216. 348
89. Pinnaduwage, L. A.; Thundat, T.; Hawk, J. E.; Hedden, D. L.; Britt, P. F.; Houser, E. J.; Stepnowski, S.; McGill, R. A.; Budd, D., 2004. Sensors and Actuators B 99, 223-229.
90. Alvarez, M.; Calle, A.; Tamayo, J.; Lechuga, L. M.; Abad, A.; Montoya, A., 2003. Biosensors and Bioelectronics 18, 649-653.
91. Vidic, A.; Then. D.; Ziegler, C. H., 2003. Ultramicroscopy 97, 407-416.
92. Adams, J. D.; Parrott, G.; Bauer, C.; Sant, T.; Manning, L.; Jones, M.; Rogers, B., 2003. Appl. Phys. Lett. 83, 3428-3430.
93. Tamayo, J.; Humphris, A. D. L.; Malloy, A. M.; Miles, M. J., 2001. Ultramicroscopy 86, 167-173.
94. Fagan, B. C.; Tipple, C. A.; Xue, Z.; Sepaniak, M. J.; Datskos, P. G., 2000. Talanta 53, 599-608.
95. Cherian, S.; Gupta, R. K.; Mullin, B. C.; Thundat, T., 2003. Biosensors and Bioelectronics 19, 411-416.
96. Xu, X.; Thundat, T. G.; Brown, G. M.; Ji, H.-F., 2002. Anal. Chem. 74, 3611- 3615.
97. Boiabjiev, V. I.; Brown, G. M.; Pinnaduwage, L. A.; Goretzki, G.; Bonnesen, P. V.; Thundat, T., 2005.. Langmuir 21, 1139-1142.
98. Ji, H.-F.; Thundat, T., 2002. Biosensors and Bioelectronics 17, 337-343.
99. Ji, H.-F.; Hansen, K. M.; Hu, Z.; Thundat, T., 2001. Sensors and Actuators B 72, 233-238.
100. Hu, K.; Bard, A. J., 1998. Langmuir 14, 4790-4794.
349
101. Liu, F.; Zhang, Y.; Ou-Yang, Z., 2003. Biosensors and Bioelectronics 18, 655- 660.
102. Kwak, K. J.; Kudo, H.; Fujihira, M., 2003. Ultramicroscopy 97, 249-255.
103. Marie, R.; Jensenius, H.; Thaysen, J.; Christensen, C. A.; Boisen, A., 2002. Ultramicroscopy 91, 29-36.
104. Hansen, K. M.; Thundat, T., 2005. Methods 37, 57-64.
105. Calleja, M.; Tamayo, J.; Johansson, A.; Rasmussen, P.; Lechuga, L. M.; Boisen, A., 2003. Sensor Letters 1, 20-24.
106. Moulin, A. M.; O’Sea, S. J.; Welland, M. E., 2000. Ultramicroscopy 82, 23-31.
107. Kooser, A.; Manygoats, K.; Eastmann, M. P.; Porter, T, L., 2003. Biosensors and Bioelectronics 19, 503-508.
108. Takahara, A.; Hara, Y.; Kojio, K.; Kajiyama, T., 2002. Colliods and Surface B: Biointerfaces 23, 141-152.
109. Wu, G.; Datar, R. H.; Hansen, K. M.; Thundat, T.; Cote, R. J.; Majumdar, A., 2001. Nature 19, 856-860.
110. Lee, H. J.; Hwang, K. S.; Park, J.; Yoon, K. H.; Yoon, D. S.; Kim, T. S., 2005. Biosensors and Bioelectronics 20, 2157-2162.
111. Wee, K. W.; Kang, G. Y.; Park, J.; Kang, J. Y.; Yoon, D. S.; Park, J. H.; Kim, T. S., 2004. Biosensors and Bioelectronics, doi: 10.1016/j.bios.2004.09.023.
112. Lee, J. H.; Yoon, K. H.; Hwang, K. S.; Park, J.; Ahn, S.; Kim, T. S., 2004. Biosensors and Bioelectronics 20, 269-275.
113. Gad, M.; Itoh, A.; Ikai, A., 1997. Cell Biology International 21, 697-706.
350
114. Bowen, W.R.; Lovitt, R.W.; Wright, C.J., 2001. Journal of Colloid and Interface Science 237, 54-61.
115. Ikai, A.; Idiris, A.; Sekiguchi, H.; Arakawa, H.; Nishida, S., 2002. Applied Surface Science 188, 506-512.
116. Stuart, J. K.; Hlady, V., 1999. Colliods and Surfaces B: Biointerfaces 15, 37-55.
117. Kim, H.; Arakawa, H.; Osada, T.; Ikai, A., 2003. Ultramicroscopy 97, 359-363.
118. Kim, H.; Arakawa, H.; Osada, T,; Ikai, A., 2002. Applied Surface Science 188, 493-498.
119. Ghatnekar-Nilsson, S.; Lindahl, J.; Dahlin, A.; Stjernholm, T.; Jeppesen, S.; Hook, F.; Montelius, L., 2005. Nanotechnology 16, 1512-1516.
120. Gfeller, K.Y.; Nugaeva, N.; Hegner, M., 2004. Biosensors and Bioelectronics, doi: 10.1016/j.bios.004.11.018.
121. Zhang, J.; H.-F., Ji, 2004. Analytical Sciences 20, 585-587.
122. Yi, J. W.; Shih, W. Y.; Mutharasan, R.; Shih, W.-H., 2003. Journal of Applied Physics 93, 1-7.
123. Grant, A.; McDonnell, L., 2003. Ultramicroscopy 97, 177-184.
124. Chakraborty, A. K.; Golumbsfskie, A. J., 2001. Annu. Rev. Phys. Chem. 52, 537- 573.
125. Wu, G.-H.; Ji, H.-F.; Hansen, K.; Thundat, T.; Datar, R.; Cote, R.; Hagan, M. F.; Chakrabory, A. K.; Majumdar, A., 2001. Proc. Natl. Acad. Sci. USA 98, 1560- 1564.
126. Allemand, J.-F.; Bensimon, D.; Jullien, L.; Bensimon, A.; Croquette, V., 1997. Biopys. J. 73, 2064.
351
127. Gaus. K.; Hall, E. A. H., 1999. Anal. Chem. 71, 2459-2467.
128. Gaus. K.; Hall, E. A. H., 1997. J. Colloid Interface Sci. 194, 364-372.
129. Gere, J. M.; Timoshanko, S. P., 1984. Mechanics of materials. Wadsworth Inc. USA, 3rd Ed.
130. Verissimo, M. I. S.; Mantas, P. Q.; Senos, A. M. S. R; Oliveira, J. A. B. P., Gomes M.T.S.R, 2003. Sensors and Actuators B 95, 25-31.
131. Fukuda, K.; Irihama, H.; Tsuji, T.; Nakamoto, K.; Yamanaka, K., 2003. Surface Science 535, 1145-1151.
132. IEEE Standard on Piezoelectricity, IEEE Standard 176-1987. Institute of Electrical and Electronic Engineers, 345 East 47th St., NY.
133. IRE Standard 1961. Measurements of piezoelectric ceramics. Proc. IRE Standards on Piezoelectric Crystals, 1161-1169.
134. IEEE Standards on piezoelectricity 1984. Special issue of Trans. Sonics and Ultrason. SU no 33-2 pt II ANSI-IEEE Std. 176, NY.
135. Kirstein, S.; Mertesdorf, M.; Schonhoff, M., 1998. J. Appl. Phys. 84, 1782-1790.
136. Cass, T.; Ligler F. S., 1998. Immobilized Biomolecules in Analysis. Oxford University Press Inc., New York.
137. Hermanson, G. T., 1996. Bioconjugate Techniques. Elsevier Science, San Diego, California, pg. 438-439.
138. Sakharov, V. E.; Kuznetsov, S. A., Zaitsev, B. D.; Kuznetsova, I. E.; Joshi, S. G., 2003. Ultrasonics 41, 319-322.
139. Chern, E. J.; Djordjevic, B. B., 1990. Mater. Eval. 48, 481-485.
352
140. Osborne, R.; Ward, M.; Dawkins, K., 2004. Sensors and Actuators A 115, 385- 391.
141. Bates, K. S.; Chang, D. B., 1993. US patient No. 5, 251-482.
142. Donlagić, D.; Zavrśnik, M., 1996. Sensors and Actuators A 55, 99-106.
143. Musayev, E.; Karlik, S. E., 2003. Sensors and Actuators A 109, 21-24.
144. Yang, C.; Chen, S., 2001. Sensors and Actuators A 94, 69-75.
145. Golnabi, H., 2004. Optics and Lasers in Engineering 41, 801-812.
146. Raatikainen, P.; Kassamakov, I.; Kakanakov, R.; Luukkala, M., 1997. Sensors and Actuators A 58, 93-97.
147. Vázquez, C.; Gonzalo, A. B.; Vargas, S.; Montalvo, J., 2004. Sensors and Actuators A 116, 22-32.
148. Iwamoto, K.; Kamata, I., 1992. Appl. Opt. 31, 51-54.
149. Weiss, D. J., 2000. Opt. Eng. 39, 2198-2213.
150. Yakymyshyn, C. P.; Pollock, C. R., 1987. J. Lightwave Technol. LT-5, 7, 941- 946.
151. Morris, J. A.; Yakymyshyn, C. P.; Pollock, C. R., 1987. ISA Trans. 26, 25-32.
152. Morris, J. F. A.; Pollock, C. R., 1987. J. Lightwave Technol. LT-5, 7, 920-925.
153. Svirid, S. V.; Leon,V.; Khotiaintsev, S. N., 2000. IEICE Trans. Electron E83-C, 303-308.
154. Haddock, H. S.; Shanker, P. M.; Mutharasan, R., 2003. Sensors and Actuators B 88, 67-74.
353
155. Wiejata, P. J.; Shanker, P. M.; Mutharasan, R., 2003. Sensors and Actuators B 96, 315-320.
156. Zhou, C.; Pivarnik, P.; Auger, S.; Rand, A.; Letcher, S., 1997. Sensors and Actuators B42, 169-175.
157. Kishen, A.; John, M. S.; Lim, C. S.; Asundi, A., 2003. Biosensors and Bioelectronics 18, 1371-1378.
158. Keirsse, J.; Plédel, C. B.; Loreal, O.; Sire, O.; Bureau, B.; Turlin, B.; Leroyer, P.; Lucas, J., 2003. J. of Non-Crystalline Solids 326-327, 430-433.
159. Keirsse, J.; Plédel, C. B.; Loreal , O.; Sire, O.; Bureau, B.; Turlin, B.; Leroyer, P.; Lucas, J., 2003. Vibrational Spetroscopy 32, 23-32.
160. Oden. I.; Chen, G. Y.; Steele, R. A.; Warmack, R. J.; Thundat, T., 1986. Appl. Phys. Lett. 68, 3814-3816.
161. Fogler, H. S., 1999. Elements of Chemical Reaction Engineering, New Jersey, 3rd ed., 700-705.
162. Yi, J. W.; Shih, W. Y.; Shih, W-H., 2002. J. Appl. Phys. 91, 1680-1686.
163. Campbell, G. A.; Mutharasan, R., 2005. Biosensors and Bioelectronics 21, 462- 473.
164. Centers for Disease Control and Prevention, 1997. Escherichia coli O157:H7 infections associated with eating a nationally distributed commercial brand of frozen ground beef patties and burgers. Colorado, Weekly Rep. 43, 777.
165. Johnson, J. L.; Rose, B. E.; Sharar, A. K.; Ransom, G. M.; Lattuada, C. P.; McNamara, A. M., 1994. J. Food Prot. 58, 597-603.
166. American Association for Food Hygiene Veterinarians, 1996. FSIS publishes final rule on pathogen reduction and HACCP systems. New-O-Gram 20, 16.
354
167. Greenberg, A. E.; Trussel, R. R.; Clesceri, L. S.; Franson, M. A. H., 1992. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC.
168. Federal Register, 1991. Drinking water: National primary drinking water regulations: total coliform proposed rule. Federal Register 54, 27544-27567.
169. Gunter, R. L.; Delinger, W. G.; Manygoats, K.; Kooser, A.; Porter, T.L., 2003. Sensors and Actuators A 107, 219-224.
170. Djuric, Z.; Jaksic, O.; Randjelovic, D., 2002. Sensors and Actuators A 96, 244- 251.
171. Florin, E.L.; Moy, V.T.; Gaub, H.E., 1994. Science 264, 415-417.
172. Lee, G.U.; Kidwell, D.A.; Colton, R.J., 1994. Langmuir 10, 354-357.
173. Hinterdorfer, P.; Baumgarter, W.; Gruber, H.J.; Schilcher, K.; Schindler, H., 1996. Pro. Natl. Acad. Sci. U.S.A. 93, 3477-3481.
174. Dammer, U.; Hegner, M.; Anselmetti, D.; Wagner, P.; Dreier, M.; Huber, W.; Guntherodt, H.J., 1996. Biophys. J. 70, 2437-2441.
175. Browning-Kelley, M.E.; Wadu-Mesthrige, K.; Hari, V.; Chen, G.Y., 1997. Langmuir 13, 343-350.
176. Lee G.U., Chrisey L.A., Colton R.J., 1994. Science 266, 771-773.
177. Pember, A.; Smith, R.; Kemhadjian, H., 1995. Sensors and Actuators A 46, 51-57.
178. Li, X.; Ono, T.; Lin, R.; Esashi, M., 2003. Microelectronic Engineering 65, 1-12.
179. Carbonell, X.; Corchero, J. L.; Cubarsí, R.; Vila, P.; Villaverde, A., 2002. Microbiol. Res. 157, 257-265.
355
180. Ross, T.; Ratkowsky, D. A.; Mellefont, L. A.; McMeekin, T. A., 2003. International Journal of Food Microbiology 82, 33-43.
181. Teng, L-J.; Hsueh, P-R.; Liaw, S-J.; Ho, S-W.; Tsai, J-C., 2004. J. Microbiol Immumol. Infect. 37, 327-334.
182. Buchanan, R. L.; Bagi, L. K.; Goins, R. V.; Phillips, J. G., 1993. Food Microbiol. 10, 303-315.
183. Presser, K. A.; Ratkowsky, D. A.; Ross, T., 1997. Appl. Environ. Microbiol. 63, 2355-2360.
184. Heitzer, A.; Kohler, H. E.; Reichert, P.; Hamer, G., 1991. Appl. Environ. Microbiol. 57, 2656-2665.
185. Rosso, L.; Lobry, J. R.; Bajard, S.; Flandrois, J. P., 1995. Appl. Environ. Microbiol. 61, 610-616.
186. Salter, M.; Ross, T.; McMeekin, T. A., 1998. J. Appl. Microbiol. 85, 357-384.
187. Campbell, G. A.; Mutharasan, R., 2005. Sensors and Actuators A 122, 326-334.
188. Kazan, D.; Çamurdan, A.; Hortaçsu, A., 1995. Process Biochemistry. 30, 269- 273.
189. Maraldo, D.; Shankar, P. M.; Mutharasan, R., 2006. Biosensors and Bioelectronics 21, 1339-1344.
190. Fujikawa, H.; Kai, A.; Morozumi, S., 2004. Food Microbiology 21, 501-509.
191. Bailey, J. E.; Ollis, D. F., 1986. Biochemical Engineering Fundamentals. McGraw Hill, New York, NY.
192. Atkinson, B.; Mavituna, F., 1983. Biochemical Engineering and Biotechnology Handbook. Macmillan Publishers, New York, NY.
356
193. Chen, K.; Le, D.; Zhang, H.; Nie, L.; Yao, S., 1996. Analytica Chimica Acta, 329:83-89.
194. Hunter, L.; Corbett, W.; Grindem, C., 1989. J. Am. Vet. Med. Asso. 8, 1028-1031.
195. Stragier, P.; Losick, R., 1996. Annu. Rev. Genet. 30, 297-341.
196. Turnbull, P. C. B., 1990. Amthrax. In: Principles of Bacteriology, Virology and Immunity, 8th edn., 365-379. Arnold, London.
197. James, C. P.; John, D. M.; Edward, M. E.; Arthur M. F., 1998. Arch. Intern. Med. 158, 429-434.
198. Tims, T. B.; Lim D. V., 2004. Journal of Microbiological Methods 59, 127-130.
199. Branch, D. W.; Brozik, S. M., 2004. Biosensors and Bioelectronics 19, 849-859.
200. Fasanella, A.; Losito, S.; Trotta, T.; Adone, R.; Massa, S.; Ciuchini, F.; Chiocco, D., 2001. Vaccine 19, 4214-4218.
201. Kim, K.; Seo, J.; Wheeler, K.; Park, C.; Kim, D.; Park, S.; Kim, W.; Chung, S-I.; Leighton, T., 2005. FEMS Immunology and Microbiology 43, 301-310.
202. Makino, S-i.; Cheun, H-i., 2003. Journal of Microbiological Methods 53, 141- 147.
203. Ramisse, V.; Patra, G.; Garrigue, H.; Guesdon, J-L.; Mock, M., 1996. FEMS Microbiology Letters 145, 9-16.
204. Sjöstedt, A.; Eriksson, U.; Ramisse, V.; Garrigue, H., 1997. FEMS Microbiology Ecology 23, 159-168.
205. Wang, S-H.; Wen, J-K.; Zhou, Y-F.; Zhang, Z-P.; Yang, R-F.; Zhang, J-B.; Chen, J.; Zhang, X-E., 2004. Biosensors and Bioelectronics 20, 807-813.
206. Turnbough Jr., C. L., 2003. Journal of Microbiological Methods 53, 263-271. 357
207. Floriano, P. N.; Christodoulides, N.; Romanovicz, D.; Bernard, B.; Simmons, G. W.; Cavell, M.; McDevitt, J. T., 2005. Biosensors and Bioelectronics 20, 2079- 2088.
208. Mainelis G., Adhikari A., Willeke K., Lee S-A., Reponen T., Grinshpun S. A., 2002. Aerosol Science 33, 1417-1432.
209. Bruno, G. J.; Kiel, J. L., 1999. Biosensors and Bioelectronics 14, 457-464.
210. Gatto-Menking, D. L.; Yu, H.; Bruno, J. G.; Goode, M. T.; Miller, M.; Zulich, A. W., 1995. Biosensors and Bioelectronics 10, 501-507.
211. Lee, S-H.; Stubbs, D. D.; Cairney, J.; Hunt, W. D., 2003. IEEE, 1194-1198.
212. Davey H. M., Kell D. B., 1997. Cytometry 28, 311-315.
213. Francis C. A., Lockley A. C., Sartory D. P., Watkins J., 2001. Water Res. 35, 3758-3761.
214. Fox, A.; Anderson, M.; Dunn, J.; Guenther, B.; Parks, L.; Pinnick, R.; Reed, C.; Rowe, J.; Lufting, R.; Long, G. W.; Lontz, R.; Marchin, G. L.; McManus, A. T.; Setlow, P.; Siedow, J.; Sievers, A. J.; Tchikindas, M. L.; Tove, S., 2002. J. Microbiol. Methods 51, 247-254.
215. Morris, B. A.; Sadana, A., 2005. Biophysical Chemistry 113, 67-81.
216. Karpovich, D. S.; Blanchard, G. J., 1994. Langmuir 10, 3315-3322.
217. Subramanian, R.; Lakshminarayanan, V., 2000. Electrochimica Acta. 45, 4501.
218. Yan, D.; Saunders, J. A.; Jennings, K., 2002. Langmuir, 18, 10202-10212.
219. Chuang, H.Y. K., 1987. Interaction of Plasma Proteins with artificial surfaces. Blood Compatibility 1, CRC Press, Boca Raton, FL, 87-102.
358
220. Horbett, D. J.; Brash, J. L., 1987. Proteins at interfaces. Physicochemical and Biochemical studies. ACS symposium series 343, Washington, DC.
221. Marvin, J.; Corcoran, E.; Hattangadi, N.; Zhang, J.; Gere, S.; Hellinga, H., 1997. Biosensors and Bioelectronics 8, 12.
222. Prime, K.L.; Whitesides, G. M., 1991. Science 252, 1164-1167.
223. Buiijs, J.; Britt, D.W.; Hadley, V., 1998. Langmuir 14, 335-341.
224. Dahint, R.; Seigel, R.R.; Harper, P.; Grunze, M.; Josse, F., 1996. Sensors and Actuators B 35-36, 497-505.
225. Fong, C-C.; Wong, M-S.; Fong, W-F.; Yang, M., 2002. Analytica Chimica Acta 456, 201-208.
226. Brown, R.E.; Jarvis, K.L.; Hyland, K. J., 1989. Anal. Biochem. 180, 136-139.
227. Bradford, M.M., 1976. Anal. Biochem. 72, 248-254.
228. Lowry, O.H.; Rosebrough, N.J.; Farr, A.R., 1951. J. Biol. Chem. 193, 265-275.
229. Knoll, W.; Park, H.; Sinner, E-k.;Yao, D.; Yu, F., 2004. Surface Science 570, 142-154.
230. Morag, E.; Bayer, E.A; Wilchek, M., 1996. Biochem. J. 316, 193-199.
231. Morag, E.; Bayer, E.A.; Wilchek, M., 1996. Analytical Biochemistry 243, 257- 263.
232. Schreiber, F., 2000. Surf. Sci. 65, 151.
233. Mrksich, M.; Whitesides, G.M., 1996. Annu. Rev. Biophys. Biomol. Struct. 25, 55- 78.
359
234. Ulman, A., 1991. From Langmuir-Blodgett to self-assembled. San Diego, CA, USA: Academic Press.
235. Bain, C.D.; Troughton, E.B.; Tao, Y.T.; Evall, J.; Whitesides, G. M.; Nuzzo, R. G. J., 1989. Am. Chem. Soc. 111, 321.
236. Kluth, G. J.; Carraro, C.; Maboudian, R., 1999. Phys. Rev. B 59, 10449.
237. Jin, Q.; Rodriguez, J. A., 1999. Surface Science 425, 101.
238. Dong, S.; Li, J., 1997. Bioelectrochem. Bioenergetics 42, 7.
239. Gooding, J. J.; Hibbert, D. B., 1999. Trends Anal. Chem. 18, 525.
240. Spangler, B.D.; Tyler, B. J., 1999. Anal. Chem. Acta. 399, 51.
241. Yourdshahyan, Y.; Zhang, H. K.; Rappe, A.M., 2001. Phys. Rev. 63, 405.
242. Tengvall, P.; Lundstrom, I.; Liedberg, B., 1998. Biomaterials 19, 407.
243. Sigal, G.B.; -57, M.; Whitesides, G. M., 1998. J. Am. Chem. Soc. 120, 3464.
244. Silin, V.; Weetall, H.; Vanderah, D. J., 1997. Colloid Interface Sci. 185, 94.
245. Pan, W.; Durning, C. J.; Turro, N. J., 1996. Langmuir 12, 4469.
246. Shen, D.; Huang, M.; Chow, L.; Yang, M., 2001. Sensor and Actuators B 77, 664- 670.
247. Shimazu, K.; Yagi, I.; Sato, Y.; Uosaki, K., 1992. Langmuir 8, 1385.
248. Taborelli, M.; Eng, L.; Descouts, P.; Ranieri, J. P.; Bellamkonda, R.; Aebisher, P. J. Biomed. Mater. Res. 29, 707.
249. Wadu-Mesthrige, K.; Amro, N. A.; Liu, G., 2000. Scanning 22, 380-388.
360
250. Margel, S.; Vogler, E.A.; Firment, L.; Watt, T.; Haynie, S.; Sogah, D.Y.J., 1993. Biomed. Mater. Res. 27, 1463.
251. Tidwell, C. D.; Ertel, S. I.; Ratner, B. D., 1997. Langmuir 13, 3404.
252. Jung, L. S.; Campbell, C. T., 2000. Phys. Rev. Lett. 84, 5164.
253. Pan, S.; Belu, A.M.; Ratner, B. D., 1999. Materials Science and Engineering C 7, 5158.
254. De-Bono, R. F.; Loucks, G. D.; Manna, D.D.; Krull, U.J., 2000. Can. J. Chem. 47, 1197.
255. Nagaoka, H.; Imae, T. J., 2003. Colloid and Interfaces Science 264, 335.
256. Rahn, J. R.; Hallock, R. B., 1995. Langmuir 11, 650.
257. Damos, F. S.; Luz, R. C.S.; Kubota, L.T., 2005. Langmuir 21, 602.
258. Takada, K.; Storrier, G.D.; Moŕan, M.; Abruńa, H.D., 1999. Langmuir 15, 7333.
259. Peterlinz, K.A.; Georgiadis, R., 1996. Langmuir 12, 4731.
260. Yan, J.; Dong, S. J., 1997. Electroanalytical Chem. 440, 229-238.
261. Ostuni, E.; Yan, L.; Whitesides, G.M., 1999. Ostuni and Surfaces B 15, 3-30.
262. Prime, K.L.; Whitesides, G.M., 1993. J. Am. Chem. Soc. 115, 10714-10721.
263. Andrade, J. D., 1985. Surface and Interfacial Aspects of Biomedical Polymers: Protein Adsorption. Plenum, New York, NY.
264. Tang, L.; Jiang, W.; Welty, S. E., 2002. J. Biomed Mater. Res. 62, 471-7.
265. Mrksich, M.; Whitesides, G.M., 1995. Trend Biotechnol.13, 228-235.
361
266. Kalltorp, M.; Carlén, A.; Thomen, P.; Olsson, J.; Tengvall, P.J., 2000. Mater Sci. Mater. Med. 11, 191-9.
267. Martins, M.C.L.; Fonseca, C.; Barbosa, M.A.; Ratner, B.D., 2003. Biomaterials 24, 3697-3706.
268. Shen, D.; Wu, X.; Liu, X.; Kang, Q.; Chen, S., 1999. Microchemical Journal 63, 322-332.
269. Kapur, R.; Rudolph, A.S., 1998. Exp Cells Res 244, 275-285.
270. Faucheux, N.; Schweiss, R.; Lützow, K.; Werner, C.; Groth, T., 2004. Biomaterials, 25, 2721-2730.
271. Blawas, A.S.; Reichert, W.M., 1998. Protein patterning. Biomaterials 19, 595- 609.
272. Hederos, M.; Konradsson, P.; Liedberg, B., 2005. Langmuir 21, 2971-2980.
273. Holmlin, R.E.; Chen, X.; Chapman, R.G.; Takayama, S.; Whitesides, G.M., 2001. Langmuir 17, 2841-2850.
274. Ostuni, E.; Chapman, R.G.; Liang, M.N.; Meluleni, G.; Pier, G.; Ingber, D.E.; Whitesides, G.M., 2001. Langmuir 17, 6336-6343.
275. Musser, M.J.; Kapur, V.; Szeto, J.; Pan, X.; Swanson, D.S.; Martin, D.R., 1995. Infect. Immun. 63, 994-1003.
276. Reid, F.H.; Bassett, D.C.; Gaworzewska, E.; Colman, G.; Poon-King, T., 1990. J. Med. Microbiol. 32, 111-114.
277. Gaunt, N. P.; Seal D.V., 2003. J. Infect. 15, 5-20.
278. Miwa, K.; Fukuyama M.; Sakai R.; Shimizu S.; Ida N.; Endo M.; Igarashi H., 2000. Microbiol. Immunol. 44, 519-523.
362
279. Liang, H.; Cordova, S.E.; Kieft, T.L.; Rogelj, S., 2003. J. Immuno. Methods 279, 101-110.
280. Bourbeau, P.P.; Heiter, B.J., 2004. Journal of Clinical Microbiology 42, 3207- 3211.
281. Bourbeau, P.P.; Heiter B.J., 2003. Journal of Clinical Microbiology 41, 2686- 2689.
282. Szewzyk, U.; Szewzyk, R.; Manz, W.; Schleifer, K.-H., 2003. Annu. Rev. Microbiol. 54, 81-127.
283. Kim, N.; Park, I-S., 2003. Biosensors and Bioelectronics 18, 1101-1107.
284. Michalzik, M.; Wilke, R.; Büttgenbach, S., 2005. Sensors and Actuators B 111- 112, 410-415.
285. Spangler, B.D.; Wilkinson, E.A.; Murphy, J.T.; Tyler, B.J., 2001. Analytica Chimica Acta. 444, 149-161.
286. Sota, H.; Yoshimine, H.; Whittier, R.F.; Gotoh, M.; Shinohara, Y.; Hasegawa, Y., 2002. Anal. Chem. 74, 3592-3598.
287. Pei, J.; Tian, F.; Thundat, T., 2004. Anal. Chem. 76, 292 – 297.
288. Kawazumi, H.; Gobi, K.V.; Ogino, K.; Maeda, H.; Miura, N., 2005. Sensors and Actuators B 108, 791-796.
289. Medina, M.B.; Houten, L.V.; Cooke, P.H.; Tu, S.I., 1997. Biotechnology Techniques 11, 173-176.
290. Yu, J.C.C.; Lai, E.P.C., 2005. Reactive & Functional Polymers 63,171-176.
291. Padhye, N.V.; Doyle, M.P., 1991. Appl. Environ. Microbiol. 57, 2693-2698.
292. Tortorello, M.L.; Gendel, S.M., 1993. J. Food Protect. 56, 672-677. 363
293. Tortorello, M.L.; Stewart, D.S., 1994. Appl. Environ. Microbiol. 60, 3553-3559.
294. Tu, Shu-I.; Golden, W.; Andreotti, P.; Irwin, P., 2002. Journal of Rapid Methods and Automation in Microbiology 10, 37-48.
295. Tu, Shu-I.; Uknalis, J.; Gore, M.; Irwin, P.; 2002. Journal of Rapid Methods and Automation in Microbiology 10, 185-195.
296. Neill, M.A.; Tarr, P.I.; Clausen, C.R.; Christie, D.L.; Hickman, R.O., 1987. Pediatrics 1, 37-40.
297. Riley, L.W.; Remis, R.S.; Helgerson, S.D.; McGee, H.B.; Wells, J.G.; Davis, B.R.; Herbert, R.J.; Olcott, E.S.; Johnson, L.M.; Hargrett, N.T.; Blake, P.A.; Cohen M. L., 1983. N Engl J Med. 12, 681-5.
298. Tarr, P.I., 1994. Review of 1993 Escherichia coli O157:H7 outbreak: Westen US, Dairy Food Envon. Sanit. 14, 372-373.
299. Karmali, M.A.; Petric, M.; Lim, C.; Fleming, P.C.; Steele, B.T., 1983. Lancet i, 619-620.
300. Mohle-Boetani, J.C.; Farrar, J.A.; Werner, S.B.; Minassian, D.; Bryant, R.; Abbott, S.; Slutsker, L.; Vugia, D.J., 2001. Escherichia coli O157:H7 and Salmonella infections associated with sprouts in California, 1996-1998. Ann. Intern. Med. 135, 239-247.
301. Wells, J.G.; Shipman, L.D.; Sowers, E.G.; Green, J.H.; Cameron, D.N.; Downes, F.P.; Martin, M.L.; Griffin, P.M.; Ostroff, S.M.; Potter, M.E.; Tauxe, R.V.; Wachsmuth I.K., 1991. J. Clin. Microbiology 29, 985-989.
302. Lee, Moon-S.; Kaspar, C.W.; Brosch, R.; Shere, J.; Luchansky, J.B., 1996.. Veterinary Microbiology 48, 223-230.
303. Tu, Shu-I.; Uknalis, J.; Irwin, P.; Yu, L.S.L., 2000. Journal of Rapid Method and Automation in Microbiology 8, 95-109. 364
304. US Department of Agriculture-Food Safety and inspection service, 1996. National federal plant raw ground beef microbiological survey, August 1993-March 1994.
305. Rowe, P.M., 1995. Lancet 345, 117.
306. Irwin P.; Damert, W.; Brewster, J.; Gehring, A.; Tu Shu-I, 2002. Journal of Rapid Methods and Automation in Microbiology 10, 129-147.
307. Tu, Shu-I.; Uknalis, J.; Gore, M.; Irwin, P.; Feder, I., 2003. Journal of Rapid Methods and Automation in Microbiology 11, 35-46.
308. Wang, H.; Sharpe, N., 1998. Food Microbiology 15, 559-565.
309. Alefantis, T.; Grewal, P.; Ashton, J.; Khan, A.S.; Valdes, J.J.; Del Vecchio, V.G., 2004. Molecular and Cellular Probes 18, 379-382.
310. McCormick, J.K.; Yarwood, J.M.; Schlievert, P.M., 2001. Annu. Rev. Microbiol. 55, 77-104.
311. Savransky, V.; Pinelis, D.; Korolev, S.; Ionin, B.; Fegeding, K., 2004. Toxicon 43, 433-438.
312. Aloisi, R.M.; Dowd, J.C., 1981. Am. J. Med. Technl. 47, 913-918.
313. Park, C.E.; Aktar, M.; Rayman, M.K., 1993. Appl. Environ. Microbiol. 59, 2210- 2213.
314. Piletsky, S.A.; Pilestska, E.V.; Bossi, A.; Karim, K.; Lowe, P.; Turner, A.P., 2001. Biosensors and Bioelectronics 16, 701-707.
315. Homola, J.; Yee, S.S.; Gauglitz, G., 1999. Sensors and Actuators B 54, 3-15.
316. Medina, M.B., 2003. J. of Rapid Methods and Automation in Microbiology 11, 225-243.
365
317. Medina, M.B., 2005. J. of Rapid Methods and Automation in Microbiology 13, 37-55.
318. Rasooly, A., 2001. J. Food Prot. 64, 37-43.
319. Slavík, R.; Homola, J.; Brynda, E., 2002. Biosensors and Bioelectronics 17, 591- 595.
320. King, K.D.; Anderson, G.P.; Bullock, K.E.; Regina, M.J.; Saaski, E.W.; Ligler, F.S., 1999. Biosensors and Bioelectronics 14, 163-170.
321. Ogert, R.A; Brown, J.E.; Singh, B.R.; Shriver-Lake, L.C.; Ligler, F.S., 1992. Anal. Biochem. 205, 306-312.
322. Tempelman, L.A.; King, K.D.; Anderson, G.P.; Ligler, F.S., 1996. Anal. Biochem. 233, 50-57.
323. Trouillet, A.; Ronot-Trioli, C.; Veillas, C.; Gagnaire, H., 1996. Pure Applied Optics 5, 227-237.
324. Khan, A.S.; Cao, C.J.; Thompson, R.G.; Valdes, J.J., 2003. Molecular and Cellular Probes 17, 125-126.
325. Peruski, A.H.; Johnson III, L.H.; Peruski Jr., L.F., 2002, J. Immunological Methods 263, 35-41.
326. Soini, E.; Kojola, H., 1983. Clin. Chem. 29, 65.
327. Mikola, H.; Takalo, T.; Hemmilia, I., 1995. Bioconjugate Chem. 6, 235.
328. Ruan, C.; Zeng, K.; Varghese, O.K.; Grimes, C.A., 2004. Biosensors and Bioelectronics 20, 585-591.
329. Lin, H-C.; Tsai, W-C., 2003. Biosensors and Bioelectronics 18, 1479-1483.
366
330. La Duc, M.T.; Satomi, M.; Agata, N.; Venkateswaran, K., 2004. Journal of Microbiological Methods 56, 383.
331. Carretto, E.; Barbarini, D.; Poletti, F.; Marzani, F.C.; Emmi, V.; Marone, P. J., 2000. Infect. Dis. 32, 98.
332. Agata, N.; Ohta, M.; Arakawa, Y.; Mori, M., 1995. Microbiology 141, 983.
333. Carlson, C. R.; Caugant, D.; Kolstø, A., 1994. Appl. Environ. Microbiol. 60, 1719.
334. Abdel-Hameed, A.; Landen, R.J., 1994. Microbiol. Biotechnol. 10, 406.
335. Watson, J.; Koya, V.; Leppla, S.H.; Daniel, H., 2004. Vaccine 22, 4374.
336. Jernigan, J.A.; Stephens, D.S.; Ashford, D.A.; Omenaca, C.; Topiel, M.S.; Galbraith, M.; Tapper, M.; Fisk, T.L.; Zaki, S.; Popovic, T.; Meyer, R.F.; Quinn, C. P.; Harper, S.A.; Fridkin, S.K.; Sejvar, J.J.; Shepard, C.W.; McConnel, M.; Guanner, J.; Shieh, W.J.; Malecki, J.M.; Gerberding, J.L.; Hughes, J.M.; Perkins, B. A., 2001. Emerg. Infect. Dis. 7, 933.
337. Morais, S.; GoncalezMartinez, M.A.; Abad, A.; Montoya, A.; Maquieira, A.; Puchades R.J., 1997. Immunol. Methods 208, 75.
338. Sperveslage, J.; Stackerbrandt, E.; Lembke, F.W.; Koch, C., 1996. Journal of Microbiological Methods 26, 219.
339. Lee, S-H.; Stubbs, D.D.; Cairney, J.; Hunt, W.D., 2003. IEEE 1194-1198.
340. Chen, C.; Nace, G. W.; Irwin, P. L., 2003. J. Microbiological Methods 55, 475-
479.
341. Wei, J.; Okerberg, E.; Dunlap, J.; Ly, C.; Shear, J. B., 2000. Anal. Chem. 72,
1360. 367
342. Pal, A.; Acharya, D.; Saha, D.; Dhar, T. K., 2004. Anal. Chem. 76, 4237.
343. Ahn-Yoon, S.; DeCory, T. R.; Baeumner, A. J.; Durst, R. A., 2003. Anal. Chem.
75, 2256.
344. Kanda, V.; Kitov, P.; Bundle, D. R.; McDermott, M. T., 2005. Anal. Chem. 77,
7497.
345. Xu, G.; Ye, X.; Qin, L.; Xu, Y.; Li, Y.; Li, R.; Wang, P. Biosensors and
Bioelectronics 2005, 20, 1757.
346. Manjooran, N. J.; Pickrell, G.R. J. Material Processing Technology 2005, 168,
225.
347. Lee, G. U.; Kidwell, D. A.; Colton, R. J. U. S. Pat. Appl. 38 pp. Avail. NTIS
Order No. PAT-APPL-8-505547, 1996.
348. Rosen, D. L.; Sharpless, C.; McGown, L. B. Anal. Chem. 1997, 69, 1082. 368
APPENDIX 369
Appendix A: Cantilever sensors for physical and chemical measurements
A.1 Physical measurement of cantilever sensors
The physical properties are summarized in Table A.1, along with the sensor sensitivity and/or detection limit, the cantilever type, measured parameter, and references.
A.1.1 Temperature
Temperature measurements by bi-metallic cantilevers having resolutions down to 3 pm were reported by Lai et al.27 and Gimzewski et al.28. The change in temperature causes the cantilever to undergo bending due to the difference in expansion coefficients of the sandwich metals. Here, the change in cantilever deflection is a direct measure of temperature change. In the study conducted by Lai et al., the thickness ratio of the bi- layer composite structure was optimized to enhanced the temperature resolution by an order of magnitude; 10-6 K compared to 10-5 K achieved by Gimzewski et al. The cantilever used by Lai et al. was constructed from silicon nitride (the cantilever base material) and aluminum (AL-SiNx). At the optimized thickness ratio the authors determined the thermal time constant as 0.33 ms, which suggests that the heat conducted from the base layer is only to the aluminum layer (top layer) and not to the surrounding air. Therefore, precise determination of temperature change sensitivity can be obtained.
Reid et al.29 demonstrated the improved sensitivity of silicon based cantilevers for the long-baseline gravitational wave detectors application by reducing the thermal noise.
Their results indicated that the mechanical dissipation is a direct function of temperature.
They have showed that mechanical losses involved thermoelastic damping, material, surface layer, loss associated with clamping and damping due to the surrounding medium. The authors observed mechanical losses over the temperature range 85 to 300 K 370 with dissipation resolution of 4.4 x 10-7. In addition, their results also indicate that at 215
K thermal dissipation is negligible, which may be due to the fact that at this temperature the linear thermal expansion coefficient of silicon approaches zero. Similarly Marie et al.30 have used a piezoresistive based cantilever sensor in monitoring the temperature of polymerase chain reaction during temperature cycles of 20 and 70 °C. In their experiments they used the output voltage of the sensor to correlate the temperature changes. Although the authors demonstrated the use of the cantilever at 20 and 70 °C the operating temperature range was 0 to 100 °C. Furthermore, the experimental results showed that the output voltage is inversely proportional to temperature and the voltage stabilized when the temperature plateaus.
In the field of information technology piezoresistive cantilevers have been used extensively for high density data storage. Yang et al.31 have reported the use of an integrated piezoresistor, heater, and an electro-thermal nano-tip. As a result both data writing and reading can be accomplished. In data writing, a force is exerted on the cantilever as the heater is heated with a current. Then the electro-thermal tip gains contact with a spinning polymer to form memory indent. In data reading, a force is applied to the cantilever. As the tip moves to a memory indent, the piezoresistor senses the displacement of the cantilever. The authors illustrated the technique at 463.15 K and found the sensitivity of piezoresistivity to be 5.4 x 104 /m with an applied force of 2 x 10-
7 N. Unlike Yang et al., Chua et al.32 have demonstrated the effect of increasing temperature (up to 240 °C) on resonant frequency response of a diamond-like amorphous carbon (DL-aC) cantilever. With the high values of the material properties (Young’s modulus and density) of DL-aC films, it is expected that the stress generated by these 371 materials should be high and therefore, larger resonant frequency values should be achieved. The authors demonstrated that by increasing the temperature from 0 to 240 °C the cantilever’s resonant frequency increased from 109.5 kHz to 111.5 kHz. Similarly,
Mertens et al.33 demonstrated the effect of temperature (290-390 K) on the resonant frequency response of a mono-material and a bi-material microcantilever in an atmosphere of helium and dry nitrogen. The authors found that the resonant frequency decreased with an increase in temperature. However, for the single material cantilever, the resonant frequency change was a linear function of temperature change due to the decrease in Young’s modulus with increase in temperature, and for the double layer cantilever the frequency change was nonlinear due to the different thermal expansion coefficients of the materials. The thermal sensitivity of the single material cantilever was determined as 1.23 Hz/K for a temperature change of 30 K.
Piezoresistive, silicon, diamond-like amorphous carbon, and bi-metallic cantilevers have been shown to detect temperature changes and operate over a wide range of temperatures. Ansorge et al.34 showed that the piezoelectric lanthanum gallium silicate
(langasite, La3Ga5SiO14) cantilever is functional up to temperatures of 700 °C. The performance of langasite at such high temperatures has proven to outrun the well known lead zirconate titanate piezoelectric material, which has a curry temperature of about 350
°C. In addition, langasite’s has shown to have piezoelectric properties up to at least 1050
°C. Conversely, cantilevers have been shown to operate in a cryogenic environment by
Hiroya et al.35 In their study the cantilever’s tip was coated with thin films of titanium oxide in one case and indium tin oxide in another. They demonstrated successfully that the two oxide layers were conductive at temperatures as low as 4.2 K and that the 372 cantilever was functional at 1.8 K. This technique is useful in the applications of nanoscale electrical transport, for example resistance measurement where earlier work reported measurements at room temperature.
A.1.2 Power
The bi-metallic cantilever discussed earlier as a temperature sensor has been used to measure power as low as 76 pW27. However, several researchers have shown piezoresistive microcantilevers to have a lower power sensitivity ranging from 3.5-70 nW/Hz1/2. These cantilevers are coated with a heat adsorbing layer that creates a differential stress between the top layer (coating) and the substrate due to temperature fluctuation36-39. The bending causes a change in the piezoresistance and is proportional to the amount of heat absorbed. The microcantilever IR sensor exhibits two distinct thermal responses: a fast one (< ms) and a slower one (~ 10 ms). A noise equivalent power (at a modulation frequency of 30 Hz) was estimated to be ~ 70 nW/Hz1/2.37
A.1.3 Pressure
Mertens et al.33 who reported temperature effects on resonant frequency discussed in the previous section had investigated the effects of varying pressure on resonance of the same silicon nitride (substrate) gold coated microcantilever. The study was conducted in helium or nitrogen rich environment. They studied the pressure effect by introducing helium or nitrogen in a vacuum chamber containing the cantilever. The authors found the resonant frequency to decrease with increasing pressure. The frequency response was nonlinear and they showed three regimes of frequency change: (1) the free molecule regime where no change in resonant frequency or cantilever deflection was observed
(pressures below 20 Pa), (2) the transition regime, at 1 x 103 Pa the resonant frequency 373 decreased rapidly suggesting that the cantilever was detecting the viscous property of the gas and finally (3) at higher pressures the resonant frequency continues to decrease at a slower rate due to the density of the gas, which induces an added mass effect. Although the authors were able to show distinctly the effects of temperature and pressure on the resonant frequency, the added mass induced by nitrogen adsorption was not observed.
The authors suggested that this may be due to the high thermal noise.
A.1.4 Elastic properties
Markidou et al.40 reported the measurements of both Young’s modulus and the shear modulus of a tissue-like soft material (gelatin) using a piezoelectric/stainless steel cantilever. The authors did compressive and shear tests on the tissue-like material and they determined the Young’s modulus of 0.05, 0.07, and 0.14 g/mL gelatin sample to be
40 ± 5, 63 ± 6, and 110 ± 10 kPa, and shear modulus as 14 ± 5, 23 ± 5, and 43 ± 8 kPa, respectively. They also, determined the Young’s and shear moduli of CL2000X rubber as
98 ± 8, and 32 ± 5 kPa, respectively. The operation of the cantilever was based on the converse piezoelectric effect. That is, the application of an alternating potential across the
PZT material generates a stress that causes the cantilever structure to bend. Then optically the cantilever tip displacement was measured and used to determine the moduli.
In addition, the authors verified the moduli values determined by the cantilever for the rubber sample with a mechanical test alternative and found them to be in good agreement.
A.1.5 Stress
Rasmussen et al.41 presented an optimized piezoresistive cantilever sensor for measuring surface stress. The cantilever was fabricated from silicon nitride integrated 374 with four polysilicon resistors and was wired using titanium silicide. The sensor’s
performance was measured by the minimum detectable surface stress(σ s,min )given by:
V σ = noise (A-1) s,min 1 ()Kεσ −1 V 4 s in
where Vnoise is the voltage noise which is a combination of the mechanical and thermal
noise, σ s the surface stress, ε the strain of piezoresistor, Vin the input voltage, and K is the gauge factor. Therefore, to minimize the surface stress one needs to reduce the noise, increase the gauge factor and increase the input voltage. The authors found the minimum detectable surface stress to be 6.5 x 10-3 N/m, which was due to the improved wiring
(titanium silicide, more stable at higher temperature) over conventional ones (doped silicon or metal). As a result, annealing could be carried out at a higher temperature without wire damage. The authors suggested that the sensitivity of the cantilever could be improved by increasing the annealing temperature and time by using gold wiring. In such case the gold wire was deposited after the annealing process and a sensitivity of 1 x 10-3
N/m was obtained. The authors demonstrated the enhanced sensitivity of the sensor in liquid (wires were coated to prevent short circuitry) by detecting the adsorption of single stranded DNA on the gold coated cantilever. They found that changes in surface stress of
0.1 N/m were detectable with high levels of accuracy.
Other stress measurements were demonstrated by Lemoine et al.42, Tabard-Cossa et al.43, Abu-lail et al.44, Cardinale et al.45, and Klaled et al.46 using silicon microcantilevers.
Each of these research groups uses the same principle of measurement. That is, the cantilever was coated on one side with a recognition layer and the adsorption of the target species induces differential surface stress causing the cantilever to bend. The deflection 375 of the cantilever is measured, using either contact profiler or optical techniques, from which the differential stress is determined. The surface stress change ()Δσ is in turn directly proportional to the cantilever’s deflection(ΔZ ) from Eq. (2-2). Lemoine et al. deposited a thin layer of an amorphous carbon film on the cantilever tip and found the compressive stress to be 2.60 and 2.54 GPa for a 20 and 110 nm thick layer. Interestingly,
Tabard-Cossa et al. used two cantilever sensors simultaneously in their study. One of the cantilevers served as the sample electrode and the other as the reference cantilever to measure unwanted noise due to mechanical and thermal vibration, and uncontrolled chemical reaction. Both cantilevers were coated with titanium and gold on opposite sides to be equally sensitive to temperature due to the difference in thermal expansion coefficients. It is important to note that their experiments were carried out in an aqueous environment and therefore, all electrical contact was protected from possible electrolyte contact. The experiment was setup with the two cantilevers side by side in an electrolytic solution of perchloric acid, the voltage of the sample or working cantilever was cycled between 0 and 700 mV to initiate the electrochemical reaction. No electrical contact was made to the reference cantilever and therefore, the noise could be detected. The difference between the two signals was used to determine the stress. The authors found the cantilevers to have a surface stress sensitivity of 1 x 10-4 N/m. Abu-Lail et al. have reported the use of the cantilever in sensing the conformational change of poly-N- isopropylacrylamide (pNIPAAM), due to the changes in pH and solvent type, which causes significant surface stress changes. The authors showed that the deflection of the cantilever was greater in solvent that caused the polymer to swell than solvent that causes it to collapse. The differential surface stress values were found to be between 6.4 and 376
52.2 N/m. Also, they found that the copolymer was sensitive to pH (stress increases with the increase in pH) and that the cantilever deflection was linear in the pH range of 4 to 6 with a sensitivity of -121 nm/pH unit. However, the authors acknowledge the fact that the cantilever deflection detection scheme is a slow process and it strongly depends on the solvent temperature. Furthermore, they note that resonant frequency detection is faster but viscous damping will reduce the cantilever sensitivity.
Khaled et al. investigated the main cause of deflection in cantilever sensors during chemical reaction and oscillation flow conditions. The authors showed that the reaction of an analyte to receptor molecules on the cantilever produces a stress differential that causes the cantilever to bend, which is the principle of measurements used by the above mentioned research groups. They also found that flow conditions created substantial deflection in the cantilever at high frequencies. In addition, the authors found that biomaterial influenced the cantilever deflection at low frequency of turbulence and when this effect is absent the turbulence increases the deflection. Like Abu-Lail et al.44 the authors concluded that resonant frequency shifts are better for sensing the interaction between analyte and surface receptor, because the binding of analyte to the cantilever surface will increase its effective mass and decrease the resonant frequency, see Eq. (3).
A.1.6 Viscosity and Density
The dynamic response of a bimorph millimeter-sized cantilever sensor in various viscous fluids was investigated by Naik et al.47. The study was done in air and in
Fluorinerttm liquids of nearly identical density and different viscosities. Also, they vary the gap height between the cantilever and a solid surface from millimeter to micrometer distance. The dynamics of a cantilever in a liquid medium is affected both by an added 377 mass effect and a viscous effect. The added mass is an inertial force that is proportional to the cantilever acceleration and the viscosity causes a dissipative force proportional to the cantilever’s velocity. The added mass decreases the resonant frequency and increases with the increase in density. The dissipative force increases with the increase in viscosity of the fluid, which decrease the quality of the resonance response. The authors experimentally measured the vibration amplitude as a function of resonant frequency.
They found that in air, the cantilever’s response was better (higher resonant frequencies and sharper resonant peaks) compared to liquid. This was due to the larger density and viscosity of the liquid. The authors observed a 15% decrease in resonant frequencies in going from air to liquid. Their results also showed that the resonant frequency further decreased with increase in liquid density; the frequency shift between liquid was 1%. In addition, the increase in viscosity decreases the resonant peak amplitude and broadens the peak due to the increasing viscous damping. They also investigated the gap height between the cantilever and a solid surface in both the air and liquid. In air no change in the cantilever’s resonance response was observed, however, in the liquid the resonant frequency, peak amplitude, and peak quality decrease with the decrease in gap height (in the same liquid). In an earlier publication by Chen et al.48 they studied the effects of gap height (nanometer distance) on the vibration amplitude and found that liquid damping increases (and amplitude decreases) as the gap height decreased. However, their experiment was not design to investigate the effect of mass loading as a function of gap height as in the study conducted by Naik et al.
Ahmed et al.49 reported the use of an atomic force microscope cantilever in the measurement of viscosity of various aqueous solutions. In this study, the authors used the 378 cantilever’s resonant frequency to demonstrate the effect of increasing viscosity of glycerol, sucrose, ethanol, sodium chloride, polyethylene glycol, and bovine plasma albumin solution of know viscosity and density. Their results showed that the resonant frequency decreased with an increase in viscosity for all the solutions investigated. The authors also presented the effect of cantilever spring constant to viscosity. They found that at viscosity values of glycerol between 1 and 10 mPa s, the cantilever with the lowest spring constant was more sensitive to viscous variations and at higher viscosity the stiffer cantilever is more sensitive. Furthermore, Chen et al.84 formulates an expression for the resonant frequency of a cantilever in a viscous liquid as:
1 ω = ⎜⎛ 9()Kηρ 4 + 64ω 2 − 3()Kηρ 2 ⎟⎞ (A-2) 8 ⎝ 0 ⎠ where ω is the angular frequency, K the spring constant, and η the viscosity. From the above one can see that the resonant frequency is a strong function of viscosity.
Shih et al.50 used a lead zirconate titanate (PZT)/stainless steel cantilever to simultaneously measure the viscosity and density of various water-glycerol mixtures. In their study, they used the phase angle to track resonance within the cantilever via an HP impedance analyzer. The resonant spectrum, a plot of phase angle versus resonant frequency, gave a number of peaks and each peak represents a mode of vibration. Also, the authors modeled the cantilever as an oscillating sphere immersed in a liquid and used the model to predict the cantilever resonant frequencies and the peak widths. They found the resonant frequency to decrease and the resonant peak width to increase with increasing glycerol content. In addition, they used the oscillating-sphere model to predict the density and viscosity from the experimentally determined resonant frequency and peak width and found them to be in good agreement with the known values. The main 379 advantage of the PZT cantilever is that both the sensing and actuating can be done by the piezoelectric material. However, in this study the authors used two electrodes where one served as the driving electrode and the other, a PZT ceramic, served as the sensing electrode, which may account for the error levels that they observed in the experimental values of viscosity and density. Also, their theoretical values may have been off because the charactistic dimension of the cantilever was in millimeters and thus, the Reynolds number was greater than one (Re>>1) resulting negligible viscous effects on the sensor performance. Furthermore, the cantilever was only partially immersed in the liquid and there was no correction in the model for this dynamic; as the cantilever section above and below the liquid may have vibrated in a slightly different manner.
Most analytical and numerical model of a cantilever dynamic resonant frequency in liquid assumes an inviscid liquid64, which is not a justifiable approach for practical applications. Sader51 has developed a detailed theoretical analysis for the resonant frequency response of a cantilever vibrating in a viscous liquid. He made the following assumptions: small vibration amplitude, incompressible fluid, uniform cross section, isotropic material, and a large length to width ratio. Also, the author includes a hydrodynamic term which accounts for the cantilever’s geometric cross section. The author developed an expression of the hydrodynamic function as:
4iK1 (− i i Re ) Γ()ω = 1+ (A-3) i ReK 0 ()− i i Re
ωρb 2 where K and K are the modified Bessel functions, Re = is the Reynolds 0 1 4μ number, μ the viscosity, and b is the width or diameter of a rectangular or circular cantilever structure, respectively. The author suggested that as the Re → ∞ , Γ()ω = 1 and 380
− 4i Γ()ω = : Re → 0 . Also, the author showed that the real part of the Reln()− i i Re hydrodynamic function is related to the angular resonant frequency of the cantilever at small dissipative effects as:
−1 2 ω ⎛ πρb 2 ⎞ f ,n ⎜ ⎟ = ⎜1+ Γr ()ω f ,n ⎟ (A-4) ωv,n ⎝ 4μ ⎠
Sader’s also demonstrated that as Re → ∞ the angular frequency can be represented as:
−1 2 ω f ,n ⎛ πρb ⎞ = ⎜1+ ⎟ (A-5) ωv,n ⎝ 4ρc h ⎠
where ρc is the density of the cantilever and h is the thickness. The above equation was first developed by Chu et al.52 for the dynamics of a rectangular cantilever sensor in a inviscid fluid. Using the above mentioned equations, Sader was able to determine the resonant frequencies of a cantilever in a viscous and an inviscid fluid from the known density and viscosity of the fluid prior to the cantilever’s construction. In another study reported by Bergaud and Nicu53, the validity of Sader’s51 analytical model (Eq.A-4) for the determination of liquid viscosity was investigated. In their work the authors predicted the resonant frequencies of the first two resonant modes in air, ethanol, and water and found them to be in good agreement with the experimentally determined values. The authors then used cantilevers of varying lengths to measure the viscosity of ultrapure ethanol. They first obtained the experimental resonant frequency of the cantilevers in ethanol and then used Eq. (A-4) to calculate the viscosity. They determined an average viscosity of 1.05 ± 0.31 x 10-3 kg/m s which was in good agreement with the known value of 1.35 x 10-3 kg/m s. In addition, the authors investigated the dynamics of the cantilever in a highly viscous liquid: silicone oil. They found that the lower resonant 381 peaks (<110 kHz), which were present upon immersion in ethanol and water disappeared in silicon oil. The authors suggested that this phenomenon was due to the high viscosity of the medium, which significantly increases the dissipative effects in the cantilever.
Also, the authors indicated that the effects of high viscosity can be reduced or overcome by tailoring the cantilever dimensions and geometry.
A.1.7 Current
Goedeke et al.54 demonstrated the use of a cobalt coated microcantilever for the detection of current induced by a magnetic field. As discussed, in the above physical measurements, small changes to the cantilever change its mechanical properties and therefore, the cantilever has to be sensitive to these small changes. In this study, the authors coated both sides of the cantilever with a thin layer of ferromagnetic cobalt. The cantilever’s response was measured by a dual optical fibers system; one of the fiber served as the light source carrier and the other picks up the reflected light from the cantilever surface. The two fibers were placed parallel to each other and normal to the cantilever’s surface. The cantilever was place close to a current carrying wire in which the current was varied from 0.2 to 1.6 A in increments of 0.1A. As the magnetic field was induced into the sensor, the cantilever deflects which changes the light intensity picked up by the retrieving fiber (response was displayed on an oscilloscope). They found that the cantilever was able to detect 0.1 A change with a sensitivity of 0.625 to 327 A/V.
Also, the authors found that the amplitude of the response decreased with increasing separation distance from the current carrying wire up to 5 mm.
A.1.8 Piezoelectric charge coefficient
The piezoelectric charge constant (d) is defined as a ratio of the mechanical strain 382 along a particular axis to the applied electric field parallel to a specific axis, it can also be expressed as the ratio of the charge generated per unit area to the applied mechanical stress in a particular direction of the piezoelectric material. The piezoelectric ceramic responds to an electrical and mechanical excitation in different directions depending on its polarization. The directional conventions are represented as 1, 2, 3, 4, 5, and 6, see
Figure A.1. Typically, the material is poled in the 3 direction and strain in the 1 and 2 directions. Directions 4, 5, and 6 refer to the torsional properties associated with 1, 2, and
3. The piezoelectric charge constant is written with two numeric subscripts. The first designates the direction of polarization or the applied electric field strength and second
indicates the strain direction or applied stress. For instance, d31 indicates that polarization was induced in the 3 direction per unit stress applied in the 1 direction or it could be interpreted as a mechanical strain in the 1 direction was induced per unit applied electric field in the 3 direction.
Several research groups have investigated the piezoelectric charge constant (d) of
55-59 55 PZT thin film . Shibata et al. measured the piezoelectric charge constant (d31) of a 1
μm thick PZT layer coated on a AFM probe. They found d31 to be -90 pC/N on a flat silicon substrate in both nitrogen and oxygen ambient. However, on a diamond substrate the piezoelectric charge constant was -20 pC/N. Furthermore, after poling the PZT material they found the charge constant to increase to -65 pC/N. By increasing the annealing temperature of the PZT layer on the AFM tip they found the charge constant to increase. In addition, the authors demonstrated the high piezoelectric charge constant on an AFM probe and showed an increase in the sensing resolution of the AFM cantilever with minimum imaging at 0.4 nm at charge of 1 x 10-15 C and applied force of 17 μN. 383
56 Conversely, Jeon et al. reported that the d33 mode of PZT generates 20 times more higher voltage than d31 mode in the same cantilever design. In the d31 mode the applied voltage is limited to the thickness of the PZT layer while in the d33 mode the voltage is limited to the length of the PZT (which can be several times greater than the thickness).
Their cantilever was designed with three different layers: a bottom silicon nitride layer and protective layer of zirconium oxide followed by a PZT thin film (0.48 μm thick). The cantilever gave three resonant peaks (13.9, 21.9, and 48.5 kHz) in the frequency range of
0-200 kHz. The authors concluded that the first and third resonant peaks were bending modes and the second peak was due to torsion. Their result showed that the charge generated in the cantilever for the fundamental mode increased linearly with an increase in tip displacement at 4.14 pC/μm. They also found the power density to be 0.74 mWh/cm2. Similar studies were done by Kanno et al.57, Zhang et al.58, and Cattan et al.59 and they found the piezoelectric coefficient as -4.7 to -4.9 C/m2, 26 μC/cm2, and -4.7
C/m2, respectively.
Itoh et al.60 developed an array of 200 μm long sol-gel piezoelectric cantilevers for a multiprobe scanning force microscope. The PZT layer was attached to a thick thermal
SiO2 layer to which the cantilever tip was bonded. Top surface of the PZT was coated with Au/Cr and the bottom with Pt/Ti. The cantilever array consists of four identical cantilevers and each cantilever has a reference sensor. The cantilevers operate individually by applying a scan voltage across the PZT and measurement of the tip displacement was determined from the generated piezoelectric current. The authors found their cantilevers to have a fundamental resonant mode at 64 kHz and an actuation sensitivity of 150 nm/V. Also, they have estimated the piezoelectric charge constant as - 384
60 pC/N.
A.1.9 Resonant frequency
Several research groups62-71 have developed and characterized cantilever sensors for resonant frequency measurements. Davis et al.62 were the first to report the fabrication and resonant frequency response of a nano-meter sized cantilever integrated in a CMOS circuit. The cantilever was operated under high vacuum in a SEM chamber that provides two advantages: (1) the amplitude of vibration can be enhanced and (2) the vacuum environment significantly reduces damping. To reduce parasitic capacitance due to external wiring and other electronics, the cantilever was integrated into a CMOS circuit
(see Ref 62 for more detail). The authors showed that the resonant spectrum (a plot of amplitude vs. driving frequency) shifted to the right in going from air to vacuum with and without the CMOS circuit integration. However, with the CMOS circuitry the resonant peak height increased in vacuum (peak became sharper). As a result the quality factor
(the term used to define peak sharpness) increased from 50-100 in air to 30,000 in vacuum. They also suggested that the cantilever was able to detect mass at 1 attogram per
Hertz in vacuum. On the other hand Zurn et al.63 have fabricated and characterized a PZT cantilever using a different technique from Davis et al.62 The cantilever consisted of a low-stressed silicon nitride layer coated with a thin layer of PZT. The authors modeled the cantilever harmonics in air and found the first ten resonant modes to be 10% higher than their corresponding experimental values. The first seven eigenfrequencies of a rectangular AFM microcantilever in both air and in water was predicted by Elmer et al.64
The authors combined the elastomechanical and hydrodynamic equation to model the dynamic resonant frequencies of the cantilever in both media. The authors assumed that 385 the cantilever is infinitely thin and that the number of nodes of an eigenmode is very large. The equation they formulated is identical to the Sader51 equation. They found the model to predict the experimental resonant frequencies within 4% error in cases where the cantilever length to width ratio was less than 0.05. The authors showed results that were similar to the above mentioned frequency response in a more viscous medium. That is, the resonant frequency of a particular mode decreases in the following order vacuum>air>liquid.
Li et al.68 have investigated both theoretically and experimentally the resonant frequency of a PZT/brass unimorph cantilever with a layer of ice and water on the brass layer, respectively. They worked from the fundamental equation⎜⎛ fα D ⎟⎞ , where D is ⎝ m ⎠ the bending modulus per unit length and m is the cantilever’s effective mass. D depends on the Young’s modulus and the moment of inertia. They showed experimentally that the added mass of liquid on the brass layer decreases the resonant frequency and that further increase in the amount of liquid decreases the resonant frequency more. However, decreasing the temperature of the liquid on the brass layer causes the resonant frequency to increase until the water froze. As the ice melted, the resonant frequency cycled back to the value prior to the initial temperature change. They indicated that, as the ice forms, it bonds to the brass surface causing an increase in the bending modulus. In addition, as the ice layer thickness increases the resonant frequency increased. One notes that an increase in the ice thickness, increases the cantilever mass. Since the resonant frequency increased, the bending modulus (D) must have been increasing at a faster rate.
Theoretically the authors used the material properties and dimensions to predict the frequency response and found them to be in line with the experimental values. 386
A.2 Chemical measurements cantilever sensors
Many different cantilever sensors have been developed and used in the measurements of several chemicals and chemical properties such as gases72-90 (which includes vapor concentration72,73, gas concentration84-88, nerve agent89, and explosive90), pH91, pesticide concentration93, ethanol/water concentration94, and ion concentration96-99. These chemicals and chemical properties are summarized in Table A.2, along with cantilever sensitivity and/or detection limit, the cantilever type, detection principle, and references.
It should be noted that in Table A.2, most of the sensors used an optical readout for deflection measurements, as bending occurs, to sense the adsorbed mass of chemical on the sensor surface.
A.2.1 Gas and vapor detection
Two studies described the detection of mercury vapor on a single microcantilever sensor72,73. Adams et al.72 used a zinc oxide piezoelectric microcantilever; the piezoelectric layer bonded to an overhanging silicon layer to which a tip was attached.
The piezoelectric thin film acts as both an actuator and a sensing element. The silicon layer was gold coated. The cantilever tip was brought in contact with a surface, as the tip bends away from the surface its vibration amplitude increased and as it bends toward the surface the amplitude decreased. The deflection of the cantilever tip generates a potential on the piezoelectric material, which is directly proportional to the amplitude of vibration.
The cantilever operated at its fundamental frequency of 51.3 kHz. Also, experiment was conducted in nitrogen. The authors demonstrated the high sensitivity of the ZnO piezoelectric microcantilever in detecting mercury vapor at 50 ppb. They also found that exposing the cantilever to water vapor resulted in a three fold increase in response in less 387 than a minute compared to mercury vapor. However, after removing the water vapor from the chamber the cantilever went to its response prior to water vapor exposure.
Conversely, mercury vapor gave a constant response after the vapor was removed. The authors did not suggest a possible explanation, but the phenomenon may be explained by the larger density of condensed water vapor on the cantilever surface that caused the tip deflection to increase significantly. However, removing the vapor reduced the vapor pressure which causes the water to be evaporated off the cantilever surface (less mass oscillation amplitude decrease), while in the case of mercury the vapor binds to the surface and therefore the cantilever’s mass remains constant. Britton Jr. et al.73 reported an array microcantilever system for the detection of mercury vapor, in a ambient hydrogen, on a gold coated surface. Like Adams, Britton Jr. used the stress induced bending principle as the sensing mechanism except a capacitor is used to sense the cantilever’s deflection. The authors found the cantilever to detection mercury vapor in ambient hydrogen in the concentration range of 10 to 70 ppb.
Several research groups have developed sensors for measuring the concentration of volatile organic compounds (VOC)74-81. The VOCs of interest were n-octane, toluene, n- butanol, alcohols, alkanes, aniline, and methylene chloride. The principle of detection in each of the studies was identical. That is, a chemical sensitive polymer was coated on one side of the cantilever and upon exposure to the analyte vapor the vapor diffuses into the polymer causing it to smell. The change in the mass of the coating creates an induced- differential surface stress that causes the cantilever to bend. The cantilever’s deflection was measured optically and was used to determine the frequency of oscillation and the adsorbed mass of analyte. For example, in a study conducted by Maute et al.74 the 388 polymeric chemical sensitive coating used was polydimethylsiloxane (PDMS). The authors monitored the resonant frequency of the cantilever using a laser beam deflection setup as the targeted vapor was adsorbed. The resonant frequency decreased due to mass loading. They found the cantilever to have a detection sensitivity of –0.0988 Hz/ppm. In the experiments the authors used the fundamental resonant mode at 46.21 kHz without the polymer coating. After coating, the frequency shifted to 24.73 kHz and after exposure to n-octane the frequency further decreased to 23.67 kHz. This is an indication that the decrease in frequency was caused by the absorption of the analyte as compared to results of control studies without polymer coating. In a similar investigation by Fadel et al.75 both a piezoelectric and an electromagnetic excited sensors were reported. The piezoelectric actuation was done by the converse piezoelectric effect (the application of an ac voltage across the piezoelectric element causes deformation and bending), while the electromagnetic actuation was carried out by an induced electromagnetic current into a ferroelectric layer on the cantilever, which causes it to bend. Unlike Maute’s work, Fadel et al. used a piezoresistive gauge to track resonance within the cantilever. Fadel et al. found the cantilever to have a VOC gas resolution of 14 ppm.
Battiston et al.76 reported a comprehensive study on the effect of various polymer coating on a cantilever resonant frequency to various alcohols and alkanes absorption.
The polymer coatings investigated were polyvinylalcohol (PVA), carboxymethylcelloluse
(CMC), polyvinylpyridine (PVP), polyvinylchloride (PVC), polyurethane (PU), polystyrene (PS), and polymethylmethacrylate (PMMA). The authors showed that a desirable polymer for a particular application should have a high partition coefficient to the target vapor and a low partition coefficient to other gases. Therefore, the chemical 389 nature of a polymer surface determines it selectivity; hydrophobicity, hydrophilicity, and polarity which dictates the swelling capacity of the polymer. The authors illustrated that
PVA, CMC, PVC, and PVP strongly adsorb water while the others are insensitive to water. They used PVC polymer coating to demonstrate the effect of cantilever response to various alcohols and alkanes. Battiston et al. found that for the lower carbon chain alcohols and alkanes resonant frequency response was greater. They have suggested that the smaller molecules were absorbed at a faster rate by the polymer and thus, resulted in a larger response. Also, confirmation of the absorption was done by desorbing the absorbed analyte and showing that the cantilever response went back to the original value prior to the absorption process. A similar gas sensor, described by Betts et al.78, used a polar
(poly(bis-cyanopropylsilane, SP-2340) and a non-polar (poly(phenyl- methyldimethylsiloxane, OV-25) polymeric coating for the sensing of alcohols, alkanes, methylene chloride, water vapor, and aniline. In each of their experiments a reference uncoated cantilever was used. The authors found that the presence of the polymeric coating on the cantilever significantly enhances its sensitivity and that different polymers influence selectivity depending on the molecular recognition. They also reported that the
SP-2340 coating was most sensitive to alcohols (example ethanol) and methylene chloride and slightly more sensitive to the absorption of the other analyte gases compared to OV-25. In addition, they showed that the analyte gases adsorption onto the uncoated side of the cantilever which reduces the overall signal (smaller differential surface stress) however, over a finite time period the surface was saturated (constant surface stress on the uncoated side). Furthermore, the authors conducted studies of varying polymer thickness and found the signal to noise ratio to decrease up to 200 nm thick polymer 390 layer. They concluded that the cantilever response to gases depends on the polarity of the polymer, the polymer thickness, and the vapor pressure of the gas. The authors verified the absorption of the various gases by exposing the cantilever after sensing the analyte gases to nitrogen and the cantilevers went back to their original responses prior to absorption.
Porter et al. developed a gas sensor to measure the concentration of VOC vapors79.
Unlike the above mentioned gas sensors that used a coated polymer on the cantilever,
Porter et al. embedded a piezoresistive microcantilever into a chemically sensitive polymer. The expansion of the polymer due to the absorption of analyte gas causes the cantilever to bend, which changes the resistance of the piezoresistor element and is therefore a direct measure of the amount of absorbent gas. The polymeric coating used in the authors experiments was lithium perchlorate doped poly(ethylene oxid) (PEO). The authors investigated the response time of the sensor to the detection of 50% water vapor, toluene, acetone, ethanol and hexane, and found that the sensor has a quick response time of about 15 seconds and a recovery time of 185 seconds. In addition, the sensor response decreased in the following order: water vapor, acetone, ethanol, toluene, and hexane.
They explained the phenomenon based on the solubility of the polymer (22.7 MPa1/2) and the analyte gases. Water vapor gave the largest response although its partition coefficient has a greater difference from the polymer, however, the other gases fall in the correct order. The idea here is that the closer the two partition coefficients more gas will penetrate the polymer, resulting in a larger cantilever deflection and thus, resistance change. The authors suggested that the high response due to water vapor was caused by the lithium perchlorate that was added to the polymer. 391
Headrick et al.81 have reported microcantilever sensor for improved volatile organic compound sensing. The surface of the cantilever was modified with microchannels along its width by a focused ion beam milling technique. The channels were filled with a chemical sensitive polymer (2,3-dihydroxynaphthalene). Upon exposure of the modified cantilever to the various VOCs (trichloroethylene, chloroform, and methylene) concentrations the smallest response observed was 980 ppb. This response was without optimization. The authors suggested that after optimization the detection limit of the sensor is going to be significantly lower.
The effects of gas viscosity on V-shaped microcantilevers were investigated by Chen et al.84. They reported on the damping of various gases including air, nitrogen, argon, and heliem. Damping of a cantilever is measured by the quality factor of the resonance characteristic (frequency or amplitude). The quality factor (Q-value) can be defined as the dissipative losses of the cantilever during oscillation in a viscous medium. The authors showed analytically that the Q-value in a viscous gas is proportional to the cantilever’s resonant frequency as:
ω m* Q = f (A-6) αμ where m* is the cantilevers effective mass and α the drag proportional coefficient. α is a constant for a given cantilever because it only depends on the cantilever’s geometry.
However, in a low damped environment the authors deduced that the quality factor can
ω be determined asQ = f , where Δυ is the peak width at half the maximum peak height Δυ of the amplitude curve. They found helium to have the highest Q factor, which was in good agreement with theory because helium has the lowest viscosity of all the gases 392 investigated.
Zhou et al.85,86 developed and reported the use of a self-excited piezoelectric microcantilever for the detection of Freon-12. As discussed earlier the piezoelectric material is used as both an actuator and a sensing element. The cantilevers were coated with a layer of zeolite (a shape selective matrix for the absorption of gases). The authors illustrated that upon exposure of the cantilever to Freon the sensor responded with a decrease in its natural frequency of oscillation due to the added mass of Freon. The author found the cantilever to have a sensitivity of -0.0024%/ppm (24/ppt) and a minimum mass detection limit of 3.5 ng. Similarly, Zhao et al.87 reported on a self excited piezoelectric cantilever sensor for dimethyl-methylphosphonate (DMMP) detection. The piezoelectric layer was bonded to a stainless steel layer that was of a longer length. DMMP is a nerve agent stimulant. The cantilevers were coated with microporous SiO2 or mesoporous Al2O3. The authors showed that the exposure of
DMMP to the SiO2 coated cantilever exhibited a rapid decrease in resonant frequency over time and the Al2O3 cantilever responded very slowly. In addition, the authors found that the uncoated cantilever gave no response when exposed to DMMP. Furthermore, they demonstrated the selectivity of the cantilever by exposing the uncoated sensor to ammonia, which showed a large resonant frequency response compared to the SiO2 coated cantilever exposed to DMMP. From the cantilever’s response the authors estimated a mass sensitivity of 2.5 x 10-8 g/Hz, which is three orders on magnitude lower than most of the cantilevers discussed above.
Pinnaduwage et al.88 discussed the desorption characteristics of explosive and non- explosive vapors from silicon microcantilevers in ambient air. In these studies the authors 393 did not apply a polymer coat to the cantilever. The principle of mass measurement was based on the deflagration of the deposited material on the cantilever, which causes thermal vibration. The motion of the cantilever was monitored by a laser beam using a two quadrant positive-sensitive detector that was integrated with a spectrum analyzer for resonant frequency measurement. In all the experiments, the authors monitored the absorption and desorption of the analyte vapor. The explosives investigated by the authors were trinitrotoluene (TNT), pentaerythritol tetranitrate (PETN), and hexahydro-
1,3,5-triazine (RDX) and the non-explosives were ethanol, acetone, and water vapor. The authors found TNT to desorb within 50 minutes in air, while PETN and RDX took several hours to observe significant cantilever resonant frequency response. Conversely, the authors could not measure desorption of the non-explosives due to their rapid rate of desorption. The authors indicated that there is a direct relationship between the analyte’s vapor pressure and the desorbed rate. The vapor pressures of PETN and RDX are two orders of magnitude smaller than that of TNT. The authors estimated a desorption rate of
2.8 ± 0.2 pg/s or 7.4 ± 0.5 x 109 s-1. Pinnaduwage et al.89 also reported a gas sensor for the measurement of 2,4-dinitrotoluene (DNT) concentration. The cantilever and its principle of detection was the same as discussed above. The sensor was coated with
SXFA-[poly(1-4-hydroxy-4-trifluoromethyl-5,5,5-trifluoro)pent-1-enyl)methylsilane] and used to sense the presence of DNT. The authors determined the detection sensitivity of the sensor as 300 ppt. They also showed that the adsorption process was reversible and that the detection took place within a few seconds. In addition, they suggested that the
SXFA coat can be repeatedly exposed to varying levels of DNT concentration for over a year and respond effectively. 394
Alvarez et al.90, first reported a pesticide detector for the measurements of DDT concentration. The cantilever sensor was of micro-meter dimensions and operated on the principle of bending due to differential surface stress caused by the binding of the analyte on one side of the cantilever. One side of the cantilever was gold coated, while to other was functionalized with an inert to the target. The gold coated side of the cantilever was functionalized with a thiol self-assembled monolayer (cystamine) which forms density monolayer coverage. To the monolayer the authors covalently linked glutaraldehyde
(contains two aldehydes group, one on both ends of the molecule) that acts as the recognition layer for DDT. DDT was modified to a hapten derivative, which was further conjugated with bovine serum albumin (BSA). Upon exposure of the hapten-DDT-BSA to the glutaraldehyde functionalized cantilever a downward bending of the cantilever was observed and measured. This indicates that the binding of the pesticide to the surface caused an increase in the differential surface stress. Exposing the cantilever to an antibody to DDT caused a further decrease in the sensor deflection suggesting binding of the antibody to the DDT derivative. In addition, exposing the DDT derivatized cantilever to a non-specific antibody produces no change in the sensor response. They determined the cantilever’s sensitivity to DDT and anti-DDT as 25 and 33 nM, respectively.
Several research groups have developed chemical sensors for both gas and liquid detection91,92,93. Adams et al.91 designed a piezoelectric microcantilever array, of three cantilevers, for the detection of ethanol/water mixture. One of the cantilevers was coated with uncured NovolacTM, one with 1% fluoropolymer solution in fluorosolvent, and the third was uncoated. The ethanol adsorption was measured by the change in resonant frequency due to the increase in the cantilever’s effective mass cause by the absorbed 395 analyte. Exposure of the cantilevers to ethanol vapor resulted in a decrease in the measured resonant frequencies of the sensors, which is attributed to the added mass of analyte on the cantilever surface. Their result also indicated that the quality factor for both the uncoated and fluoropel coated cantilevers decreased, while the Q-value of the
Novolac coated cantilever increased. The authors suggested that this phenomenon may be due to the loosening of the uncured coating or stiffing of the polymer upon swelling.
Similarly, Vidic et al.92 and Tamayo et al.93 have reported the use of a microcantilever in the detection of varying ethanol concentration in water. Both research groups used the resonant frequency to measure the sensor response to the analyte. It is important to note that these experiments were done in the liquid phase. As discussed earlier, both the Q- value and frequency of a resonant mode decreased in going from air to liquid. In the work of Vidic et al. they designed an oscillation circuit that enhanced the Q-value by two orders of magnitude in air and one under liquid conditions. Vidic et al. showed detection of as low as 2% ethanol/water solution. Tamayo et al. built a positive feedback Q-control that enhanced the quality factor by three orders of magnitude under liquid. To the best of my knowledge Tamayo et al. are the first to have reported such an increase in quality factor in liquid. They demonstrated the enhancement in the detection of ethanol/water mixture at 0.5% ethanol. The authors also found the cantilever to have a detection limit of
1 pg.
A.2.2 Ion concentrations
The measurements of ionic concentration have been reported in the literature by several research groups using microcantilever sensors95-98. Cherian et al.95 demonstrated the detection of heavy metal ions such as Zn2+, Ni2+, Co2+, Co2+ Cu2+, Cd2+, and Hg2+ 396 using silicon nitride cantilevers that were functionalized with the metal binding protein
AgNt84-6. One side of the cantilever was gold coated to which a stable self-assembled
4ATP monolayer was formed by sulfur-gold interaction. EDC/NHS chemistry was used to activate the carboxyl groups on the protein, which was later coupled covalently to the
4ATP monolayer. The cantilever was rinsed between each step to remove any weakly bound residue and after the protein (AgHt84-6) was attached the available reactive sites on the gold surface was exposed to chicken eggwhite lysozyme to reduce non-specific binding of the metal ions. The cantilever’s deflection was used to track the binding of the protein-metal ion interaction. Upon exposure of the protein derivatized cantilever to Hg2+ and Zn2+ ions the authors observed increasing bending of the cantilever. However, Hg2+ ion gave the largest deflection. In addition, they found that after exposing the cantilever to Zn2+ and then Hg2+ ion the bending of the cantilever increased significantly. The authors suggest that Hg2+ binds stronger to the protein and may have replaced some of the bound Zn2+ ions. They also speculated that there may be specific binding sites (15 histidine residues) on the protein for Hg2+ ions. Furthermore, the authors tried to regenerate the protein surface with EDTA after exposure to the ions and they found that
EDTA does not effectively release the bound ions from the protein surface. Confirmation of the protein-ion interactions was illustrated using electrophoresis technique in which the protein-ion complexes were shown to have a higher mobility through the gel, while the control migrated slowly. The authors indicated that this may be due to the conformational change of the protein upon binding with the metal ions. Their results showed that they were able to detect ion concentration of 100 μM. A similar study was done by Xu et al.96, here they reported on the detection of Hg2+ ions using 1- 397 dodecanethiol self-assembled monolayer (SAM) on a gold coated silicon nitride microcantilever. As in the previous study, one side of the cantilever was gold coated and modified with the SAM and the other side was functionalized with triethoxymethylsilane to reduce non-specific binding. The deflection was used to track the SAM-ion interaction.
The authors showed that the cantilever was able to detect Hg2+ at 10 pM. They also illustrated that the exposure of the cantilever to other metal ions such as Zn2+, Ni2+, Na+,
K+ Cu2+, Cd2+, Ca2+ and Pb2+ showed no response. These results indicate that the 1- dodecanethiol SAM functionalized sensor was selective to Hg2+ ions.
Boiadjiev et al.97 developed a photochemical hydrosilylation coated microcantilever for the detection of Cr(VI). The gold coated side of the cantilever was functionalized with
11-undecenyltriethylammonium bromide via photochemical hydrosilylation to form free ammonium groups for chromate interaction. The authors showed that their sensor has a
2- 98 detection limit of 1 nM CrO4 ions. Ji and Thundat showed that bis(11- mercaptooundecyl) phosphate SAM on a gold coated cantilever can be used to detect
Ca2+ ions at 1 nM concentration. They also found that the modified cantilever was not selective to other cations such as Na+ and K+ ions. It is important to note that all of these experiments of ion concentration measurement were carried out in a flow cell at flow rates ranging from 2 to 10 mL/h.
A.2.3 pH
Ji et al.99 described a microcantilever pH sensor. The cantilever was constructed from silicon nitride. One side of the cantilever surface was both chemically and metal modified in the various studies. Chemical modification was done by functionalizing the cantilever with 4-aminobutyltriethoxysilane and 11-mercaptoundecanoic acid, and metal coating 398 used was Au/Al. The detection mechanism was based on the ionization of the coated species of the cantilever surface. As the pH sensitive surface accumulates charges the cantilever bends due to the differential surface stress that is generated by proton-surface interaction. The deflection of the cantilever is proportional to the pH of the surrounding medium and therefore, measurement of the deflection gave quantitative measures of the surrounding liquid pH. The authors demonstrated that the aminosilane modified cantilever operated well in the pH range of 2-8, with a detection sensitivity of 49 nm/pH unit, while the metal-modifed sensor performed best in the pH range of 2-6 and 8-12 with a detection sensitivity of 30 nm/pH unit. Furthermore, they found that at pH values below
2 both type of modified surfaces gave negligible change in the sensor deflection. In addition, the Al2O3 surface was not stable at high or low pH values. For instance, at low pH (HCl) the metal oxide dissolves forming AlCl3 and at high pH (NaOH) it also dissolves producing NaAlO3. As a result the authors found the Aluminum coated surface to be effective for pH measurements in the range of 4 to 12, with a deflection rate of 34 nm/pH unit.
A.2.4 Rate of reaction
Hu and Bard100 reported a sensor to determine the adsorption rate constant of
- mercaptoundecanoic acid (HSC10COO ) on a gold substrate using an atomic force microscope cantilever. The authors were the first to report quantitatively the kinetics of thiol adsorption on a gold surface using an AFM cantilever. They measured the
- interfacial force between the deprotonated HSC10COO surface and a modified negatively charged silica sphere tip of the cantilever as a function of adsorption time. The deflection of the cantilever was measured as a function of separation distance from the gold surface 399 as the monolayer formed, from which they obtained the surface coverage. By applying the first order Langmuir kinetics to fit the data of surface coverage versus time, the
- authors found the average observed adsorption rate constant of HSC10COO to be 0.045
±0.005 (0.5 mM) and 0.02 ±0.003 (0.05 mM) min-1. 400
(6)
(3)
(2)
(1) (5)
(4)
Figure A.1: Direction of excitation for a piezoelectric material
401
Table A.1: Cantilever sensors for physical measurements.
Measurement Resolution Cantilever type Measured References Parameters parameter Temperature 2 x10-6 K [27] Bimetallic Deflection [27, [27-33] 1 x10-5 K [28] [27, 28] 28, 30] 85-300 K [29] Piezoelectric Frequency [29, 0-100 °C [30] [29,32,33] 32,33] 5.4 x 10-4 @ 463.15 K [31] Piezoresistive Resistance [30. 240 °C [32] [30,31] 31] 290-390 K [33] 700 °C [34] Power 76 pW [27] Bimetallic Deflection [27] [36-39] 3.5-70 nW/Hz1/2 [36-39] [27] Resistance [36- Piezoresistive 39] [36-39] Pressure 20 Pa Piezoelectric Frequency [33] Moduli Young’s modulus 98 ± 8 kPa Piezoelectric Frequency [40] Shear modulus 31 ± 5 kPa Stress 1 x 10-3 N/m [41] Piezoresistive [41-46] 0.97-0.78 GPa [42] [41-46] Resistance 1 x 10-3 N/m [43] 6.4-52.2 N/m [44] 0.1-1.3 GPa [45] 402
Table A.1: Cantilever sensors for physical measurements (continues).
Measurement Resolution Cantilever type Measured References Parameters parameter Viscosity and Air [47] Frequency [47-53], density 0.89 mPa s [48] Piezoelectric [50-53] 1.8405 x 103 Pa [49] [50,53] 3 H2O 0.01 P & 1 g/cm [50] Ethanol 1.05±0.31 x 10-3 kg m-1 s-1 & 0.79 g.cm3 [53] Current ΔI = 0.1A electromagnetic Current [54] Piezoelectric -65 to -90 pC/N [55], Piezoelectric Force [55] [55-61] charge 4.14 pC/μm [56] Frequency constant -4.7 to -4.9 C/m2 [57] [56,59] 26 μC/cm2 [58] Deflection -4.7 C/m2 [59] [57,58,60,61] -60 pC/N & [60] 32.7 pC/N [61] Frequency 1 attogram/Hz [62] Piezoelectric Frequency [61-71]
403
Table A.2: Microcantilever sensors for chemical measurements.
Measurement Sensitivity/Range Cantilever Measured References Parameters type parameter
Mercury vapor 50 ppb in N2 [72] Piezoelectric Frequency [72] [72-73]
10-70 ppb in H2 [73] Amplitude [73] VOCs (alcohols, 0.0998 Hz/ppm [74] Bimetallic Frequency 74- [74-81] toluene, n-octane,..) 14 ppm or 0.06 Hz/ng [74,76] 75,76,80] [75] 200 ppm [76] Piezoresistive Deflection 0.0055 Hz/ppm [77] [75,79] [77-79, 81] -0.54 to -26 V/atm [78] AFM [78] 1 x 10-6 nm-1 [79] 6.5 gp/Hz [80] Piezoelectric 1-1000 ppm [81] [80] Carbon monoxide 0.7 x 10-6 nN-1 Piezoresistive Resistance [82]
Hydrogen @ 66% H2 stress was Deflection and [83] 1.1 x 10-4 Nm-1 Frequency
Air, H2, N2, Ar, He AFM Frequency [84] 404
Table A.2: Microcantilever sensors for chemical measurements (continues).
Measurement Sensitivity/Range Cantilever Measured References Parameters type parameter Freon 3.5 x 10-9g or - Piezoelectric Frequency [85,86] 0.0024%/ppm [85,86] [85,86] [85,86] Dimethyl 2.8 ± 0.2 pg/s Piezoelectric Frequency [87] methylphosphonate 7.4 x 109 s-1 bending (DMMP) TNT, PETN, & 2.8 ± 0.2 pg/s or 7.4 ± Piezoresistive Frequency [88] RDX explosives 0.5 x 109 s-1 bending 2,4-dinitrotoluene 300 ppt Optical Bending [89] (DNT) DDT 10nM Optical Bending [90] Ethanol/water 2% EtOH/water [91] Magnetic [91] Bending [91,93] [91-93] 370 ppm ethanol [92] Piezoelectric Frequency [92] 100 ppm EtOH [93] [93] Silane Creates selective Optical Bending [94] surface Zn2+, Ni2+, Co2+, 1.7 to 30.8 μM [94] Optical Bending [95] Cd2+, Hg2+, Cu2+ 10 pM Hg2+ [95] [96] 2- CrO4 1 nM Optical Bending [97] Ca2+ 1nM Optical Bending [98] pH 2-12 Optical Bending [99] Rate of reaction 0.045 ±0.005 min-1 Optical Bending [100] - (HSC10COO , 0.5mM)
405
Vita
Gossett A. Campbell received his Bachelor of Science with honors in Chemical
Engineering from Drexel University in 2002. After graduating, he started graduate school at Drexel University in the Department of Chemical and Biological Engineering, where he works under the supervision of Dr. Raj Mutharasan on biosensors. In his field of study he specialized in the design and fabrication of MEMS, detection, and quantification of low concentrations of molecules, proteins, and pathogens. His progression was very quick and his work had led him to several awards over the years such as the best presentation award at the Mid-Atlantic Biochemical Engineering
Consortium (2005) , Drexel College of Engineering graduate student research award
(2005), and the technical excellence award from the National Society of Black Engineers
(2006). Gossett has presented his work at the AIChE and the MABEC national conferences every year, since 2003.
Gossett has excellent experimental skills in which he demonstrated his expertises in the novel research he does. His accomplishments include the publication of fourteen papers to some of the major scientific journals (Langmuir, Analytical Sciences,
Biosensors and Bioelectronics, and Journal of Analytical Chemistry) and the disclosure of five patents of which three of them are filed.