Low-Cost, Rapid, Sensitive Detection of Pathogenic Bacteria Using
Phage-Based Magnetoelastic Biosensors
by
Shin Horikawa
A dissertation submitted to the Graduate Faculty of
Auburn University
in partial ful llment of the
requirements for the Degree of
Doctor of Philosophy
Auburn, Alabama
May 5, 2013
Keywords: magnetoelastic biosensor, phage, pathogenic bacteria, food safety,
biosecurity
Copyright 2013 by Shin Horikawa
Approved by
Bryan A. Chin, Chair, Professor of Materials Engineering
Zhongyang Cheng, Professor of Materials Engineering
Dong-Joo Kim, Associate Professor of Materials Engineering
Valery A. Petrenko, Professor of Pathobiology
Sang-Jin Suh, Associate Professor of Biological Sciences
George T. Flowers, Dean of the Graduate School
Abstract
As part of the ongoing e orts to secure food safety as well as to guard against
possible bioterrorism, the role of pathogen detection technologies has become vi-
tal. However, conventional and standard detection methods, including culture-,
immunology-, and polymerase chain reaction-based methods, are generally expen-
sive, time-consuming, and labor-intensive. Hence, there is a need for new detection
technologies that outperform the conventional methods and enable the rapid, on-site
detection of pathogenic substances. Although label-free biosensors have proven to be
among the most promising methods, meeting various performance criteria (e.g., sen-
sitivity, selectivity, assay time, thermal stability, and longevity) simultaneously still
remains a challenge. Hence, further research and development are essential before
biosensors become a reliable, alternative solution.
Phage-based magnetoelastic (ME) biosensors, a novel class of wireless, mass-
sensitive biosensors, are among potential candidates that could overcome the above
performance challenge. These biosensors are not only thermally robust, but their
wireless nature of detection o ers great  exibility in design and use, which facili-
tates on-site pathogen detection. In addition, the sensitivity of ME biosensors can
be improved by reducing their dimensions, and the fabrication cost per sensor can
be reduced via batch fabrication. Hence, this dissertation presents investigations
into the performance improvement of phage-based ME biosensors, in terms of cost-
e ectiveness, rapidness, and sensitivity, and into the enhanced detection of pathogenic
bacteria, Salmonella Typhimurium and Bacillus anthracis spores, for food safety and
biosecurity.
ii
To enhance both cost-e ectiveness and sensitivity, micron- to millimeter-scale
ME biosensors were batch-fabricated and used. In this way, the fabrication cost per
sensor was reduced to a fraction of a cent. In addition, the following two method-
ologies were employed to dramatically shorten assay time: (1) direct detection of S.
Typhimurium on fresh spinach leaves and (2) detection of B. anthracis spores with
the aid of a designed micro uidic  ow cell, which ensures e cient physical contact
between a biosensor and  owing spores. By using these methodologies with low-cost,
miniature ME biosensors, (1) S. Typhimurium cells on the order of 104 cells/cm2
were detected with 150- m long sensors in 45 min, and (2) down to 106 B. anthracis
spores were detected with 200- m long sensors in 10 min.
Additionally, to further enhance the detection capabilities of phage-based ME
biosensors, the following e ects were studied: (1) the e ects of mass position on
the sensitivity of ME biosensors and (2) the e ects of surface functionalization on
surface phage coverage. The mass sensitivity of ME biosensors was found to be
largely dependent on the dimensions of the sensors as well as on the position of
attached masses. From numerical simulation results, a formula that predicts the
mass-position-dependent sensor response for a single localized mass was also derived.
In addition, surface phage coverage on bare and surface-functionalized ME biosen-
sors was quanti ed by atomic force microscopy. The results showed that activated
carboxyl-based covalent attachment produced a surface phage coverage of  50%,
which is comparable to that obtained through physical adsorption, the traditional
method of phage immobilization. By contrast, much lower surface phage coverages
( 5%) were obtained for aldehyde- and methyl-terminated sensor surfaces. These
di erences in surface phage coverage was also found to a ect the quantity of a sub-
sequently captured analyte. Hence, by properly functionalizing the sensor surface,
both surface phage coverage and the quantity of the captured analyte can be con-
trolled. Finally, with the results of the mass-position-dependence of sensor response,
iii
a concept of phage layer patterning was introduced. Phage may be patterned onto
desired parts of the sensor surface to further enhance the detection capabilities of ME
biosensors.
iv
Acknowledgments
This dissertation would not have been possible without the guidance and help of
a great number of individuals. First and foremost, I am truly indebted to my advisor,
Dr. Bryan A. Chin, who has provided invaluable assistance in the preparation and
completion of this research. I will never forget his generosity and encouragement. I
am also grateful to all my committee members, Dr. Zhongyang Cheng, Dr. Dong-Joo
Kim, Dr. Valery A. Petrenko, and Dr. Sang-Jin Suh for their sincere support. Special
thanks to Dr. Valery A. Petrenko, who has helped me improve my attitude towards
scienti c research.
I would like to thank Dr. Maria L. Auad, Dr. Michael J. Bozack, and Dr. Michael
E. Miller for their technical assistance in atomic force microscopy, x-ray photoelectron
spectroscopy, and confocal scanning laser microscopy, respectively. With their expert
guidance and assistance, I have been able to acquire sound engineering practices.
I am also thankful for all the support from Dr. James M. Barbaree, Kiril A.
Vaglenov, and I-Hsuan Chen, who have provided all biological samples and shared a
great deal of knowledge regarding microbiology over the past several years.
Yating Chai, Dr. John Shu, Michael L. Johnson, Leslie C. Mathison, Dr. Shichu
Huang, Dr. Jiehui Wan, Dr. Suiqiong Li, and Steve Best have shared with me
countless discussions on sensing principles, microelectronic fabrication, and testing
methodologies. We have worked together in our experiments and constantly tried to
gain greater understanding of our work. I would like to thank them for always being
there for me.
Last but not least, I owe earnest thankfulness to my parents and sister for their
love and support.
v
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background and need . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Target pathogenic bacteria to be detected . . . . . . . . . . . . . . . 6
1.3.1 Salmonella Typhimurium . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Bacillus anthracis spores . . . . . . . . . . . . . . . . . . . . . 8
1.4 Organization of this dissertation . . . . . . . . . . . . . . . . . . . . . 10
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Review of the Literature on Bacterial Detection Methods . . . . . . . . . 14
2.1 Conventional detection methods . . . . . . . . . . . . . . . . . . . . . 14
2.1.1 Culture-based methods . . . . . . . . . . . . . . . . . . . . . . 14
2.1.2 Immunology-based methods . . . . . . . . . . . . . . . . . . . 15
2.1.3 Polymerase chain reaction-based methods . . . . . . . . . . . 17
2.2 Biosensors as promising bacterial detection methods . . . . . . . . . . 20
2.2.1 De nition of a biosensor . . . . . . . . . . . . . . . . . . . . . 20
2.2.2 Biomolecular-recognition elements . . . . . . . . . . . . . . . . 22
2.2.3 Signal transducers . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Conventional detection methods vs. biosensors . . . . . . . . . . . . . 23
vi
2.3.1 Probability of detection: PCR vs. biosensors . . . . . . . . . . 25
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3 Phage-Based Magnetoelastic (ME) Biosensors . . . . . . . . . . . . . . . 37
3.1 Landscape phages as biomolecular-recognition elements . . . . . . . . 37
3.2 Magnetoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 Joule magnetostriction . . . . . . . . . . . . . . . . . . . . . . 42
3.2.2 Magnetization and Joule magnetostriction in ferromagnets . . 44
3.2.3 ME signal transducers . . . . . . . . . . . . . . . . . . . . . . 47
3.3 Fabrication of ME sensor platforms . . . . . . . . . . . . . . . . . . . 48
3.3.1 Dicing method . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3.2 Co-sputtering-based method . . . . . . . . . . . . . . . . . . . 49
3.3.3 Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.4 Fabrication cost per sensor platform . . . . . . . . . . . . . . . 53
3.4 Fabrication of phage-based ME biosensors . . . . . . . . . . . . . . . 55
3.4.1 Immobilization of a phage on the ME sensor platforms . . . . 55
3.4.2 Surface blocking of the ME biosensors with bovine serum albumin 55
3.5 Principle of detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5.1 Minimum detectable number of bacterial cells . . . . . . . . . 57
3.5.2 Measurement of the resonant frequency of the ME biosensors . 59
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4 Direct Detection of S. Typhimurium on Fresh Spinach Leaves . . . . . . 64
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Material and methods . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 E2 phage and S. Typhimurium . . . . . . . . . . . . . . . . . 66
4.2.2 Confocal re ectance imaging of spinach leaf surfaces . . . . . . 66
4.2.3 Fabrication of ME sensor platforms with three di erent sizes . 67
4.2.4 Fabrication of phage-based ME biosensors . . . . . . . . . . . 68
vii
4.2.5 Determination of the concentration of BSA for surface blocking 68
4.2.6 Direct detection of S. Typhimurium on fresh spinach leaves . . 69
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.1 Observation of Salmonella-inoculated leaf surfaces . . . . . . . 71
4.3.2 Resonant frequency measurement . . . . . . . . . . . . . . . . 72
4.3.3 Dose-response of the ME biosensors . . . . . . . . . . . . . . . 74
4.3.4 Determination of the LOD . . . . . . . . . . . . . . . . . . . . 76
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.1 Topography of leaf surfaces and its e ects on the LOD . . . . 79
4.4.2 E ects of the number of biosensors on the LOD . . . . . . . . 83
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5 Detection of B. anthracis Spores with the Aid of A Micro uidic Flow Cell 95
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Material and methods . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.2.1 JRB7 phage and B. anthracis Sterne spores . . . . . . . . . . 97
5.2.2 Batch-fabrication of micron-scale ME resonators . . . . . . . . 97
5.2.3 Micro uidic  ow cells . . . . . . . . . . . . . . . . . . . . . . . 98
5.2.3.1 Design and fabrication of the Type I micro uidic  ow
cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.2.3.2 Design and fabrication of the Type II micro uidic  ow
cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.3 Valve actuation and sample injection . . . . . . . . . . . . . . . . . . 104
5.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.4.1 Con rmation of valve actuation . . . . . . . . . . . . . . . . . 105
5.4.2 Manipulation of  uorescent-labeled micro-spheres . . . . . . . 105
5.4.3 Manipulation of a sensor . . . . . . . . . . . . . . . . . . . . . 106
viii
5.4.4 Detection of B. anthracis spores with the Type II micro uidic
 ow cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6 Enhancing the Detection Capabilities of Phage-Based ME Biosensors . . 115
6.1 E ects of mass position on the mass sensitivity of ME biosensors . . . 115
6.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.1.2 Material and methods . . . . . . . . . . . . . . . . . . . . . . 116
6.1.2.1 Microcontact printing . . . . . . . . . . . . . . . . . 116
6.1.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 117
6.1.3.1 Finite element modal simulation . . . . . . . . . . . 118
6.2 E ects of surface functionalization on surface phage coverage . . . . . 123
6.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.2.2 Material and methods . . . . . . . . . . . . . . . . . . . . . . 125
6.2.2.1 Preparation of biological samples . . . . . . . . . . . 126
6.2.2.2 Manufacture of gold-coated ME resonators . . . . . . 126
6.2.2.3 Surface functionalization of gold-coated ME resonators 126
6.2.2.4 X-ray photoelectron spectroscopy . . . . . . . . . . . 127
6.2.2.5 Loading of the phage on the ME resonators . . . . . 128
6.2.2.6 Resonant frequency measurement . . . . . . . . . . . 128
6.2.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 129
6.2.3.1 Veri cation of the surface functionalization of gold . 129
6.2.3.2 Surface phage coverage . . . . . . . . . . . . . . . . . 131
6.2.3.3 SEM observation and dose-response results . . . . . . 135
6.2.3.4 Enhancing the detection capabilities of ME biosensors
with a patterned phage layer . . . . . . . . . . . . . 138
6.2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
ix
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
x
List of Figures
1.1 Economic cost - bene t assessment. . . . . . . . . . . . . . . . . . . . . . 2
1.2 Civilian biodefense funding by  scal year, FY2001 - FY2013 (in a36millions). 3
1.3 S. Typhimurium cells on a spinach leaf surface. . . . . . . . . . . . . . . 8
1.4 B. anthracis spores on a gold surface. . . . . . . . . . . . . . . . . . . . . 9
2.1 Conventional methods for bacterial detection. . . . . . . . . . . . . . . . 14
2.2 Typical procedure for sandwich ELISA. . . . . . . . . . . . . . . . . . . 16
2.3 Schematic illustration of the PCR cycle. . . . . . . . . . . . . . . . . . . 18
2.4 Number of research articles published between 1985 and 2005 on di erent
bacterial detection methods. . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Schematic diagram of a biosensor. . . . . . . . . . . . . . . . . . . . . . . 21
2.6 Classi cation of biosensors. . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7 Probability of detection with respect to the inhalation dose of the target
pathogen for PCR-based detectors. The following values were used in the
calculations: Ws = 1,000 l/min, Ke = 0.8, I = 15, and m = 1. . . . . . . 28
2.8 Probability of detection with respect to the inhalation dose of the target
pathogen for biosensor-based detectors. The following values were used in
the calculations: Ws = 1,000 l/min, Ke = 0.8, Vs = 1 ml, and n = 1. . . 29
xi
3.1 Schematic illustration of the wild-type fd phage and its genetically engi-
neered form, displaying a foreign peptide on the major coat protein pVIII. 38
3.2 Sequences of amino acid residues of the fd coat proteins. The N-terminus
is to the left. The hydrophobic domains are underlined, whereas charged
residues are indicated by + or -. . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Sequences of amino acid residues of the wild-type and fusion pVIII proteins. 41
3.4 Joule magnetostriction of a spherical ME material. . . . . . . . . . . . . 43
3.5 Hypothetical  eld dependencies of  k and  ?. . . . . . . . . . . . . . . . 44
3.6 Magnetic domains and magnetization processes in a ferromagnet. . . . . 45
3.7 (a) E ects of magnetizing  eld and mechanical stress on the distribution
of magnetic moments in a ferromagnet and (b)  eld dependence of  k
under a compressive stress. . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.8 Temperature dependence of normalized  s in Fe80B20. . . . . . . . . . . . 46
3.9 Diced sensor platforms stored in dry methanol. . . . . . . . . . . . . . . 49
3.10 Procedure for the co-sputtering-based method. . . . . . . . . . . . . . . . 50
3.11 Denton sputter coater. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.12 Scanning electron micrograph of batch-fabricated sensor platforms with a
size of 100  m  25  m  4  m on a gold-coated wafer. . . . . . . . . . 52
3.13 Uniform attachment of bacterial cells or spores on a phage-immobilized
ME biosensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.14 Measurement setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
xii
3.15 Response of a typical 500  m  100  m  4  m ME biosensor in air. . . 60
4.1 Scanning electron micrographs of various produce surfaces: (a) tomato,
(b) eggshell, and (c) spinach leaf. A close-up view of a spinach leaf spiked
with S. Typhimurium is shown in (d). . . . . . . . . . . . . . . . . . . . 65
4.2 Di erently sized sensor platforms used (top view). . . . . . . . . . . . . . 67
4.3 E ects of BSA concentration on (a) resonant frequency changes for mea-
surement and control sensors (2-mm long) and on (b) the con dence level
of di erence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4 Schematic illustration of the test procedure: (a) spot-inoculation of S.
Typhimurium on the leaf surface and measurement of the initial resonant
frequency of biosensors, (b) placement of both measurement and control
sensors on the Salmonella-inoculated sites (after drying the Salmonella
drops and misting the leaf surface), (c) measurement of the  nal resonant
frequency of the biosensors, and (d) typical responses of the biosensors. . 70
4.5 Scanning electron micrographs of a spinach leaf surface inoculated with a
40- l drop of S. Typhimurium with various concentrations: (a) 5  108
cells/ml, (b) 5  107 cells/ml, (c) 5  106 cells/ml (with a 150  m-long
ME biosensor), and (d) 0 cells/ml (reference). . . . . . . . . . . . . . . . 71
4.6 Response of a typical 150  m 30  m 4  m sensor in air: (a) raw data
set (10-time averaged) and its smoothed curve and (b) Lorentizan  tting
of the smoothed curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Resonant peaks for a 150- m long sensor before and after placing on a
leaf surface inoculated with S. Typhimurium at a concentration of 5  
108 cells/ml. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
xiii
4.8 Dose-response plots for the di erently sized biosensors (2 mm-, 0.5 mm-
and, 150  m-long sensors). The plots on the left and right are the results
for the adaxial and abaxial surfaces, respectively. . . . . . . . . . . . . . 75
4.9 (a) Sigmoidal curve and (b) the determination of the LOD. . . . . . . . . 76
4.10 Background-subtracted data for the dose-response plots in Fig. 4.8. These
data were  tted with sigmoidal functions (red solid curves). The R2 values
were all close to one. The values of  fAVE + 3 are shown in blue text. 78
4.11 Typical height maps (a & b) and associated averaged pro les (c & d)
of a leaf surface obtained along di erent sampling lengths. A three-
dimensional representation of a leaf surface expressed by Eq. 4.2 is shown
in (e). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.12 Finite well model of a Salmonella-inoculated leaf surface. The total sur-
face area is x2LW: There are three types of wells: sensor-containing wells,
cell-containing wells, and empty wells. . . . . . . . . . . . . . . . . . . . 84
4.13 Finite well model with n = 1 (i.e., one cell-containing well). . . . . . . . 85
4.14 Probability of detection with respect to the number of biosensors, m, and
the number of cell-containing wells, n. Biosensors with lateral dimensions
of 50  m  10  m were placed on a leaf surface of 0.26 cm2. . . . . . . . 87
4.15 Dependence of the LOD on the number of biosensors (50  m  10  m  
2  m), m, for various values of P(D). . . . . . . . . . . . . . . . . . . . . 90
5.1 Scanning electron micrographs of 200  m 40  m 4  m ME resonators
fabricated on a  at wafer: (a) batch-fabricated resonators and (b) a close-
up view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
xiv
5.2 Procedure for the fabrication of a micro udic chip by multilayer soft
lithography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.3 Design of the Type I  ow cell: (a) a three-dimensional view of the whole
chip and (b) a close-up view of the key elements of the chip. . . . . . . . 100
5.4 Push-up and push-down valves. . . . . . . . . . . . . . . . . . . . . . . . 102
5.5 Design of the Type II  ow cell: (a) a three-dimensional view of the whole
chip on a slotted glass slide and (b) a close-up view of the key elements
of the chip. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.6 A close-up top view of a fabricated chip (Type I), connected to external
pressure sources for valve actuation and sample injection. . . . . . . . . . 104
5.7 Close-up top views of a Type I  ow cell: (a) All the valves (#1 through
#4) are closed, and (b) only the #2 and #4 valves are closed such that
the green dye injected through the vertical channels can only  ll out the
center chamber. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.8 Fluorescent micrographs showing that the  ow cell can separate a few
micro-spheres into the reaction chamber: (a) separation of four spheres,
(b) two spheres, and (c) one sphere. . . . . . . . . . . . . . . . . . . . . 106
5.9 Injection of a 200  m  40  m  4  m sensor into the chamber through
the horizontal channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.10 Optical micrograph of the Type II micro uidic  ow cell. All the push-
down valves are closed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.11 Fluorescent micrograph showing B. anthracis spores are bound on a 200
 m  40  m  4  m ME biosensor. The chamber was outlined with a
yellow solid line for better visualization. . . . . . . . . . . . . . . . . . . 109
xv
5.12 Streamlines in the micro uidic  ow cell. The color bar below indicates
the magnitude of  ow velocity. . . . . . . . . . . . . . . . . . . . . . . . 110
5.13 Responses of ME biosensors (200  m 40  m 4  m) to various numbers
of spores. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.1 PDMS stamps and microcontact printing. SA beads (not to scale) were
placed in the middle (stamping area: 1 mm  0.8 mm) or (b) at both
ends (stamping area: 0.5 mm  0.8 mm each) of a phage-immobilized
ME biosensor using the Type A or Type B stamp, respectively. BSA is
not shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.2 Phage-based ME biosensors loaded with SA beads. Beads are attached
(a) in the middle, (b) at both ends, or (c) uniformly on both sides of the
ME biosensors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.3 Three-dimensional FE model: (a) geometry, (b) meshed geometry, and
(c) resultant mode shape. . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.4 Dependence of the mass sensitivity,  f= m, on the longitudinal position
of the attached mass and on the dimensions of the ME biosensors. . . . . 121
6.5 Surface functionalization of gold (1 1 1) with three SAM chemicals, based
on the sulfur - gold chemistry: (a) AC (activated carboxyl-terminated),
(b) ALD (aldehyde-terminated), and (c) MT (methyl-terminated). . . . . 125
xvi
6.6 Schematic illustration of the frequency measurement setup for ME biosen-
sors. (a) The setup consists of a copper solenoid coil, a bar magnet, and a
network analyzer (not shown). (b) A phage-immobilized ME biosensor is
placed in the glass capillary  ow cell and positioned in the coil center. A
suspension of SA beads was passed at 25  l/min over the sensor, and the
resonant frequency change of the biosensor was monitored in a wireless,
magnetic manner. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.7 (a) Water contact angles for bare and surface-functionalized gold surfaces.
(b) XPS S2p peaks at around 162 eV, indicating the formation of sulfur
{ gold bonded systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.8 AFM images (2  m  2  m) of the SAE10 phage on bare and surface-
functionalized ME resonators: (a) reference, (b) bare gold (physical phage
adsorption), (c) AC- (covalent phage attachment), (d) ALD-, and (e) MT-
functionalized ME resonators. The white lines on the photographs were
the paths from which the height pro les were measured. (f) The surface
phage coverage values were 46.8%, 49.4%, 4.2%, and 5.2% for bare, AC-,
ALD-, and MT-functionalized ME resonators, respectively. . . . . . . . . 132
6.9 Water contact angles for bare and AC-functionalized gold surfaces loaded
with/without phages. The surfaces were washed with a 100-mM HEPES
bu er containing various concentrations of Tween 20 (0 to 5 % v/v) and
 nally with DI water. The e ect of washing was found to be large on the
immobilization stability of the physically adsorbed phages. . . . . . . . . 134
xvii
6.10 SEM images showing SA beads captured by the SAE10 phage on bare
and surface-functionalized ME biosensors (all BSA blocked): (a) bare, (b)
AC-, (c) ALD-, and (d) MT-functionalized ME biosensors. Non-speci c
adsorption of SA beads was greatly reduced by BSA blocking: (e) bare
and (f) AC-functionalized ME biosensors without phage. . . . . . . . . . 136
6.11 Dose-response plots showing the comparable performance of (a) bare and
(b) AC-functionalized ME biosensors (1 mm  0.2 mm  15  m). . . . . 137
xviii
List of Tables
1.1 Major performance criteria for biosensors. . . . . . . . . . . . . . . . . . 4
1.2 Infectious doses and incubation periods for the target pathogenic bacteria. 6
1.3 Recent Salmonella outbreaks in various food products in the United States. 7
1.4 Types of anthrax infection and associated fatality rates. . . . . . . . . . 9
2.1 Examples of culture-based bacterial detection. . . . . . . . . . . . . . . . 15
2.2 Examples of ELISA-based bacterial detection. . . . . . . . . . . . . . . . 17
2.3 Examples of PCR-based bacterial detection. . . . . . . . . . . . . . . . . 19
2.4 Comparison of the performance of major bacterial detection methods. . . 24
2.5 Variables and their values used for the calculations of the probability of
detection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Lethal doses of B. anthracis spores in humans. . . . . . . . . . . . . . . 28
2.7 Minimum detectable number of pathogens for PCR- and biosensor-based
methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1 Longevity of a landscape phage and monoclonal antibody at various tem-
peratures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Comparison between landscape phages and antibodies in terms of selec-
tivity and production cost. . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Phage clones used in this research. . . . . . . . . . . . . . . . . . . . . . 42
3.4 Materials properties for Metglas 2826MB and Fe80B20. . . . . . . . . . . 48
3.5 Sputtering conditions used for the fabrication of micron-scale sensor plat-
forms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.6 Material costs for the sputtering targets. . . . . . . . . . . . . . . . . . . 54
xix
3.7 Di erently sized ME biosensors and their theoretical detection limits. . . 58
4.1 Mean resonant frequencies for the di erently sized biosensors. . . . . . . 74
4.2 LODs of the di erently sized biosensors for the adaxial and abaxial sur-
faces of spinach leaves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3 Surface geometric parameters for the adaxial and abaxial surfaces of spinach
leaves. The values are the averages of 15 samples. . . . . . . . . . . . . . 81
4.4 Required number of biosensors, m, to obtain desired LODs for various
values of P(D) (0.1, 0.5, and 0.8). . . . . . . . . . . . . . . . . . . . . . . 89
5.1 Number of bound spores and corresponding resonant frequency changes
for measurement and control sensors. . . . . . . . . . . . . . . . . . . . . 111
6.1 Resonant frequency changes in Hz. . . . . . . . . . . . . . . . . . . . . . 118
6.2 Materials constants used in FE simulations. . . . . . . . . . . . . . . . . 120
xx
List of Abbreviations
AFM Atomic force microscopy
AUDFS the Auburn University Detection and Food Safety Center
BSA Bovine serum albumin
CDC the Centers for Disease Control and Prevention
DI Deionized
DMF dimethylformamide
DNA Deoxyribonucleic acid
dNTP Deoxynucleotide triphosphate
ELISA Enzyme-linked immunosorbent assay
FE Finite element
LOD Limit of detection
ME Magnetoelastic
PCR Polymerase chain reaction
PDMS Poly(dimethylsiloxane)
RNA Ribonucleic acid
SAM Self-assembled monolayer
SEM Scanning electron microscopy
xxi
TBS Tris-bu ered saline
VBNC Viable but non-culturable
XPS X-ray photoelectron spectroscopy
xxii
Chapter 1
Introduction
1.1 Background and need
Food is essential for each individual to grow and stay healthy in daily life. How-
ever, the past decades have been marked by a global increase in the outbreaks of
food poisoning and associated illnesses. These public health problems are caused
by the accidental supply and consumption of contaminated food, largely due to im-
proper safety knowledge, perspectives, and practices of food producers [1] as well
as insu cient consumer awareness [2]. Although substantial progress on food safety
regulations has been made worldwide [3], up to 30% of the population even in indus-
trialized countries su er from foodborne illnesses each year [4]. In the United States,
for example, approximately 48 million cases of foodborne illnesses are estimated to
occur annually, resulting in 128,000 hospitalizations, 3,000 deaths, and a3651.0 to a3677.7
billion economic losses [5{7].
At present, 31 foodborne pathogens, including bacteria, parasites, and viruses,
are identi ed in the United States [7]. According to the Centers for Disease Control
and Prevention (CDC) [5], most foodborne illnesses in the United States are caused by
norovirus (58%), followed by nontyphoidal Salmonella spp. (11%), Clostridium per-
fringens (10%), and Campylobacter spp (9%). In addition, the leading cause of both
hospitalization and death is nontyphoidal Salmonella spp. (35% and 28%, respec-
tively), which cause Salmonellosis, a major foodborne disease in most countries [4].
Food can be contaminated by these identi ed as well as unidenti ed pathogens at any
stages of the supply chain (e.g., production, packaging, transportation, and retail).
As a result, the food industry, the main party that is concerned with the presence
1
Figure 1.1: Economic cost - bene t assessment.
of foodborn pathogens, is responsible for controlling the quality of food products.
However, zero risk for all food products is unlikely to be achievable. Hence, one of
the biggest challenges is to put e ective controls in place without unnecessarily in-
creasing costs. In other words, the optimal level of food safety must be determined
through economic cost - bene t assessments [8]. Figure 1.1 shows a typical represen-
tation of the relationship between implicit price per unit of safety and level of safety.
As can be seen from the upward sloping line of marginal social cost, it is inexpen-
sive to improve safety at low levels, but further improvements are more costly. By
contrast, marginal social bene t (i.e., society?s additional willingness to pay to avoid
ill-health and the costs of treating ill-health) decreases as the level of safety increases,
represented by the downward sloping line. At point A, the amount the society is
willing to pay exceeds the amount it would cost to improve safety, indicating that it
is worth allocating resources to produce more safety. By contrast, costs exceed ben-
e ts at point B, meaning that too many resources are being devoted to safety. Only
at point Qm are costs equal to bene ts per unit of safety, representing that e cient
resource allocation occurs [8]. Hence, this point Qm is the sought safety level (Pm is
2
the corresponding price per unit of safety). From the above example of assessment,
it is understandable that cost is an important factor that cannot be disregarded for
the management of food safety risks.
Foodborne illness is not the only problem that poses a severe risk to public
safety. Since the 2001 anthrax mail attacks in the United States, bioterrorism, which
makes use of bacteria, viruses, fungi, and/or toxins as a bioweapon, has been publicly
recognized as an emerging danger. The CDC has, thus far, identi ed 35 potential
bioterrorism agents and classi ed them into three categories [9]. For example, Bacillus
anthracis, the etiologic agent of anthrax, is listed among the high-priority Category
A agents, which have the potential for major public health impact. Comprehensive
attempts to control these deadly biological agents have been made internationally
by prohibiting their use and proliferation since the war era in the last century [10].
However, as is evident from the recent anthrax attacks, the attempts have not been
Figure 1.2: Civilian biodefense funding by  scal year, FY2001 - FY2013 (in a36millions).
3
entirely successful. As a result, the government of the United States has been en-
hancing national biosecurity. In fact, the funding for civilian biodefense dramatically
increased after the 2001 anthrax attacks, and over a364 billion of funding has been
maintained since the  scal year of 2002 as shown in Fig. 1.2 [11]. The budget transi-
tion clearly indicates that there is a need for comprehensive biodefense systems that
enable the nationwide surveillance and prevention of bioterrorism. In addition, a por-
tion of the funding is dedicated to food defense, including the prevention of deliberate
food contamination with pathogenic agents.
As part of the ongoing e orts to secure food safety as well as to guard against
possible bioterrorism, the role of pathogen detection technologies has become vi-
tal. However, conventional and standard detection methods, including culture-,
immunology-, and polymerase chain reaction-based methods, are generally expen-
sive, time-consuming, and labor-intensive [3]. Hence, much research has been recently
focused on developing label-free biosensors, which are meant to be low-cost, rapid,
Table 1.1: Major performance criteria for biosensors.
Criterion Description
Sensitivity Ability to detect a small amount of pathogens in a reasonably small
sample volume
Selectivity Ability to distinguish among pathogens
Assay time Short for a single test
Thermal stability Ability to function at a wide range of temperatures
Longevity Ability to retain detection capabilities for a fair period of time
Assay protocol No reagent addition needed
Measurement Direct and without pre-enrichment
Format Highly automated format
Operator No expertise needed
Cost Inexpensive
Size Compact and portable for on-site detection
4
and user-friendly, adequate for on-site pathogen detection for both food safety and
biosecurity. Table 1.1 lists the major performance criteria for biosensors (partially
adapted from [12]). Although some existing biosensors possess excellent performance,
meeting various performance criteria simultaneously still remains a challenge. Hence,
further research and development are essential before biosensors become a reliable,
alternative solution.
1.2 Research objectives
Magnetoelastic (ME) biosensors, a novel class of wireless, mass-sensitive biosen-
sors, are among potential candidates that could overcome the above-mentioned perfor-
mance challenge. In recent years, the Auburn University Detection and Food Safety
Center (AUDFS) has begun research into the detection of pathogenic bacteria us-
ing freestanding, strip-shaped ME biosensors combined with a landscape phage (i.e.,
genetically engineered phage) [13] as the biomolecular-recognition element [14{17].
These phage-based ME biosensors are not only rapid and thermally robust [16], but
their wireless nature of detection o ers great  exibility in design and use, which fa-
cilitates on-site bacterial detection. In addition, the sensitivity of ME biosensors can
be improved by reducing their dimensions [18], and the fabrication cost per sensor
can be reduced via batch fabrication. Hence, the primary objectives of this research
are (1) to further improve the cost-e ectiveness, rapidness, and sensitivity of the
phage-based ME biosensors and (2) to demonstrate enhanced detection of pathogenic
bacteria (i.e., Salmonella Typhimurium and Bacillus anthracis spores). In order to
improve both cost-e ectiveness and sensitivity, micron- to millimeter-scale ME biosen-
sors were batch-fabricated and used. In addition, the following two methodologies
were employed to shorten assay time:
5
1. Direct detection of S. Typhimurium on fresh spinach leaves without any pre-test
sample preparation (i.e., collection and puri cation of Salmonella-containing
samples, followed by enrichment)
2. Detection of B. anthracis spores with the aid of a designed micro uidic  ow
cell, which ensures e cient physical contact between a biosensor and  owing
spores.
Additionally, as potential ways to further enhance the detection capabilities of phage-
based ME biosensors, the following e ects were studied:
1. E ects of mass position on the sensitivity of ME biosensors
2. E ects of surface functionalization of ME biosensors on surface phage coverage.
1.3 Target pathogenic bacteria to be detected
S. Typhimurium and B. anthracis spores are target pathogenic bacteria to be
detected in this research. Their median infectious doses and incubation periods are
summarized in Table 1.2.
Table 1.2: Infectious doses and incubation periods for the target pathogenic bacteria.
Target pathogen Infectious dose Incubation period Ref.
S. Typhimurium 100 to 1,000 cells (ingestion) 6 to 72 hr [19]
B. anthracis 8,000 to 50,000 spores (inhalation) < 7 days [20]
1.3.1 Salmonella Typhimurium
Nontyphoidal Salmonella spp. are important foodborne pathogens that cause
gastroenteritis, bacteremia, and subsequent focal infection [21]. They are responsible
for 11% of all foodborn illnesses in the United States, resulting in roughly 20,000
hospitalizations and 400 deaths each year [5]. Salmonellosis, caused by the ingestion
6
Table 1.3: Recent Salmonella outbreaks in various food products in the United States.
Source Cause(s) Cases Hospitalizations Year Ref.
Tomatoes S. Typhimurium 183 22 2006 [24]
Peanut butter S. Tennessee 425 71 2007 [25]
Cantaloupes S. Litch eld 51 > 16 2008 [26]
Jalape~no peppers S. Saintpaul 1,442 > 286 2008 [27]
Peanut butter S. Typhimurium 714  170 2009 [28]
Alfalfa sprouts S. Saintpaul 235  7 2009 [29]
Shell eggs S. Enteritidis 1,939 N/A 2010 [30]
Cantaloupes S. Panama 20 3 2011 [31]
Sprouts S. Enteritidis 25 3 2011 [32]
Ground turkey S. Heidelberg 136 37 2011 [33]
Chicken livers S. Heidelberg 190 30 2011 [34]
Ground beef S. Typhimurium 20 8 2011 [35]
Ground tuna S. Bareilly &
S. Nchanga
425 55 2012 [36]
Ground beef S. Enteritidis 46 12 2012 [37]
Cantaloupes S. Typhimurium &
S. Newport
270 101 2012 [38]
Live poultry S. Montevideo 76 17 2012 [39]
Mangoes S. Braenderup 121 25 2012 [40]
Peanut butter S. Bredeney 30 4 2012 [41]
of nontyphoidal Salmonella spp., is a major foodborne disease in most countries today
[4] and usually contracted from various sources [22], including eggs, meat, poultry,
and fresh produce as shown in Table 1.3. Among over 2,500 serovars capable of
infecting humans and animals, S. Typhimurium is becoming one of the most prevalent
serovars [23]. Hence, this pathogenic bacterium has been selected as one of the target
pathogens to be detected in this research. Figure 1.3 shows a scanning electron
micrograph of S. Typhimurium cells on a spinach leaf surface.
7
Figure 1.3: S. Typhimurium cells on a spinach leaf surface.
1.3.2 Bacillus anthracis spores
B. anthracis, the etiologic agent of anthrax, is a rod-shaped, spore-forming bac-
terium [42, 43]. Due to it?s ability to form a resistant spore (i.e., dehydrated, thick-
walled cell), this bacterium can populate a wide range of environments, including soil,
bodies of water and animal hosts [42]. Although B. anthracis spores are metabolically
dormant, they can germinate and grow to a vast number of vegetative cells once en-
tering a nutrient-rich host, which in turn causes the disease anthrax with high fatality
rates (Table 1.4). Hence, the potential use of B. anthracis spores as a bioweapon is a
signi cant public safety concern. When used in an aerosolized form, they could enter
the bodies of individuals through inhalation. For example, the recent anthrax attacks
that occurred in the United States in 2001 resulted in 11 cases of inhalational anthrax,
8
Table 1.4: Types of anthrax infection and associated fatality rates.
Type Fatality rate Ref.
Inhalational As high as 90% (< 50% with appropriate treatment) [44{46]
Cutaneous 20% (< 1% with appropriate treatment) [45,46]
Gastrointestinal 25 to 60% [45,46]
5 of whom died. Although early, proper antibiotic treatments have proven e ective
in reducing the high fatality rates, such treatments are often di cult to provide due
to initial, non-speci c symptoms in infected patients [44]. Hence, anthrax infection
must be prevented through early detection of B. anthracis spores. Figure 1.4 shows
a scanning electron micrograph of B. anthracis spores on a gold surface.
Figure 1.4: B. anthracis spores on a gold surface.
9
1.4 Organization of this dissertation
In this chapter, the need for high-performance biosensors for on-site pathogen
detection was described, and the objectives of the present research were stated. The
rest of this dissertation is organized as follows:
Chapter 2 brie y reviews major bacterial detection methods and discusses rea-
sons for the current shift towards the development of label-free biosensors.
Chapter 3 describes the fundamentals, detection principle, and fabrication meth-
ods of phage-based ME biosensors in depth.
Chapter 4 presents an investigation into rapid, direct detection of S. Typhimurium
on fresh spinach leaves. Various e ects, including the topography of spinach leaf sur-
faces, the distribution of S. Typhimurium cells, and the size and number of ME
biosensors, on the limit of detection will also be discussed.
Chapter 5 presents an investigation into rapid, sensitive detection of B. anthracis
spores using micron-scale ME biosensors in combination with a designed micro uidic
 ow cell.
Chapter 6 investigates the e ects of mass position on the sensitivity of ME biosen-
sors. Experimental and numerical results will be  rst compared, and then, a formula
predicting the sensor response for a single point-mass will be derived. In addition, the
e ects of surface functionalization on surface phage coverage will be studied. Based
on the results of these investigations, a concept of the patterning of the phage layer
onto desired parts of the sensor surface will be introduced.
Finally, Chapter 7 presents an overall summary and conclusions of this disserta-
tion.
10
Bibliography
[1] M. L. L. Ivey, J. T. LeJeune, and S. A. Miller, \Vegetable producers perceptions
of food safety hazards in the midwestern USA," Food Control, vol. 26, no. 2,
pp. 453 { 465, 2012.
[2] C. Losasso, V. Cibin, V. Cappa, A. Roccato, A. Vanzo, I. Andrighetto, and
A. Ricci, \Food safety and nutrition: Improving consumer behaviour," Food
Control, vol. 26, no. 2, pp. 252 { 258, 2012.
[3] V. Velusamy, K. Arshak, O. Korostynska, K. Oliwa, and C. Adley, \An overview
of foodborne pathogen detection: In the perspective of biosensors," Biotechnology
Advances, vol. 28, no. 2, pp. 232 { 254, 2010.
[4] WHO, \Food safety and foodborne illness," 2007. fact sheet no. 237.
[5] E. Scallan, R. M. Hoekstra, F. J. Angulo, R. V. Tauxe, M.-A. Widdowson, S. L.
Roy, J. L. Jones, and P. M. Gri n, \Foodborne illness acquired in the United
States - major pathogens," Emerging Infectious Diseases, vol. 17, no. 1, pp. 7 {
15, 2011.
[6] E. Scallan, P. M. Gri n, F. J. Angulo, R. V. Tauxe, and R. M. Hoekstra,
\Foodborne illness acquired in the United States - unspeci ed agents," Emerging
Infectious Diseases, vol. 17, no. 1, pp. 16 { 22, 2011.
[7] R. L. Schar , \Economic burden from health losses due to foodborne illness in
the United States," Journal of food protection, vol. 75, no. 1, pp. 123 { 131, 2012.
[8] W. B. Traill and A. Koenig, \Economic assessment of food safety standards:
Costs and bene ts of alternative approaches," Food Control, vol. 21, no. 12,
pp. 1611 { 1619, 2010.
[9] M. Shannon, \Management of infectious agents of bioterrorism," Clinical Pedi-
atric Emergency Medicine, vol. 5, no. 1, pp. 63 { 71, 2004.
[10] D. P. Fidler, \Facing the global challenges posed by biological weapons," Mi-
crobes and Infection, vol. 1, no. 12, pp. 1059 { 1066, 1999.
[11] C. Franco and T. K. Sell, \Federal agency biodefense funding, FY2012 -
FY2013," Biosecurity and Bioterrorism: Biodefense Strategy, Practice, and Sci-
ence, vol. 10, no. 2, pp. 162 { 181, 2012.
11
[12] D. Ivnitski, I. Abdel-Hamid, P. Atanasov, and E. Wilkins, \Biosensors for de-
tection of pathogenic bacteria," Biosensors and Bioelectronics, vol. 14, no. 7,
pp. 599 { 624, 1999.
[13] V. A. Petrenko, \Landscape phage as a molecular recognition interface for de-
tection devices," Microelectronics Journal, vol. 39, no. 2, pp. 202 { 207, 2008.
[14] S. Huang, H. Yang, R. Lakshmanan, M. Johnson, J. Wan, I.-H. Chen, H. W.
III, V. Petrenko, J. Barbaree, and B. Chin, \Sequential detection of Salmonella
typhimurium and Bacillus anthracis spores using magnetoelastic biosensors,"
Biosensors and Bioelectronics, vol. 24, no. 6, pp. 1730 { 1736, 2009.
[15] R. S. Lakshmanan, R. Guntupalli, J. Hu, D.-J. Kim, V. A. Petrenko, J. M.
Barbaree, and B. A. Chin, \Phage immobilized magnetoelastic sensor for the de-
tection of Salmonella typhimurium," Journal of Microbiological Methods, vol. 71,
no. 1, pp. 55 { 60, 2007.
[16] J. Wan, H. Shu, S. Huang, B. Fiebor, I.-H. Chen, V. A. Petrenko, and B. A. Chin,
\Phage-based magnetoelastic wireless biosensors for detecting Bacillus anthracis
spores," IEEE Sensors Journal, vol. 7, pp. 470 { 477, 2007.
[17] J. Wan, M. L. Johnson, R. Guntupalli, V. A. Petrenko, and B. A. Chin, \De-
tection of Bacillus anthracis spores in liquid using phage-based magnetoelastic
micro-resonators," Sensors and Actuators B: Chemical, vol. 127, no. 2, pp. 559
{ 566, 2007.
[18] M. L. Johnson, J. Wan, S. Huang, Z. Cheng, V. A. Petrenko, D.-J. Kim, I.-
H. Chen, J. M. Barbaree, J. W. Hong, and B. A. Chin, \A wireless biosensor
using microfabricated phage-interfaced magnetoelastic particles," Sensors and
Actuators A: Physical, vol. 144, no. 1, pp. 38 { 47, 2008.
[19] http://www.msdsonline.com/resources/msds-resources/
free-safety-data-sheet-index/salmonella-spp.aspx.
[20] http://www.msdsonline.com/resources/msds-resources/
free-safety-data-sheet-index/bacillus-anthracis.aspx.
[21] D. Acheson and E. L. Hohmann, \Nontyphoidal Salmonellosis," Clinical Infec-
tious Diseases, vol. 32, no. 2, pp. 263 { 269, 2001.
[22] http://www.harrisonspractice.com/practice/ub/view/Harrisons%20Practice/
141052/all/Nontyphoidal Salmonellosis.
[23] http://www.safe-poultry.com/ParatyphoidSalmonella.asp.
[24] http://www.cdc.gov/ncidod/dbmd/diseaseinfo/salmonellosis 2006/
110306 outbreak notice.htm.
12
[25] http://www.cdc.gov/ncidod/dbmd/diseaseinfo/salmonellosis 2007/
030707 outbreak notice.htm.
[26] http://www.cdc.gov/salmonella/litch eld/.
[27] http://www.cdc.gov/salmonella/saintpaul/jalapeno/index.html.
[28] http://www.cdc.gov/salmonella/typhimurium/update.html.
[29] http://www.cdc.gov/salmonella/saintpaul/alfalfa/.
[30] http://www.cdc.gov/salmonella/enteritidis/index.html.
[31] http://www.cdc.gov/salmonella/panama0311/062311/index.html.
[32] http://www.cdc.gov/salmonella/sprouts-enteritidis0611/070611/index.html.
[33] http://www.cdc.gov/salmonella/heidelberg/111011/index.html.
[34] http://www.cdc.gov/salmonella/heidelberg-chickenlivers/011112/index.html.
[35] http://www.cdc.gov/salmonella/typhimurium-groundbeef/020112/index.html.
[36] http://www.cdc.gov/salmonella/bareilly-04-12/index.html.
[37] http://www.cdc.gov/salmonella/enteritidis-07-12/index.html.
[38] http://www.cdc.gov/salmonella/typhimurium-cantaloupe-08-12/index.html.
[39] http://www.cdc.gov/salmonella/montevideo-06-12/index.html.
[40] http://www.cdc.gov/salmonella/braenderup-08-12/index.html.
[41] http://www.cdc.gov/salmonella/bredeney-09-12/index.html.
[42] A. Driks, \The Bacillus anthracis spore," Molecular Aspects of Medicine, vol. 30,
no. 6, pp. 368 { 373, 2009.
[43] T. M. Koehler, \Bacillus anthracis physiology and genetics," Molecular Aspects
of Medicine, vol. 30, no. 6, pp. 386 { 396, 2009.
[44] T. V. Inglesby, T. O?Toole, D. A. Henderson, J. G. Bartlett, M. S. Ascher,
E. Eitzen, A. M. Friedlander, J. Gerberding, J. Hauer, J. Hughes, J. McDade,
M. T. Osterholm, G. Parker, T. M. Perl, P. K. Russell, and K. Tonat, \Anthrax
as a biological weapon, 2002: Updated recommendations for management," The
Journal of the American Medical Association, vol. 287, no. 17, pp. 2236 { 2252,
2002.
[45] http://www.bt.cdc.gov/agent/anthrax/faq/.
[46] http://www.upmc-biosecurity.org/website/our work/
biological-threats-and-epidemics/fact sheets/anthrax.html.
13
Chapter 2
Review of the Literature on Bacterial Detection Methods
This chapter reviews major bacterial detection methods and discusses reasons
for the current shift towards the development of label-free biosensors.
2.1 Conventional detection methods
Conventional methods for the detection of pathogenic bacteria rely on speci c
microbiological or biochemical identi cation. Three major conventional methods are
culture-, immunology-, and polymerase chain reaction-based methods as shown in
Fig. 2.1. Although these methods can be highly sensitive, selective, and reliable, their
application to on-site bacterial detection is greatly restricted by several drawbacks,
including long assay times, high cost, and cumbersome procedures, requiring trained
personnel.
Figure 2.1: Conventional methods for bacterial detection.
2.1.1 Culture-based methods
Culture-based methods remain the most reliable and commonly used techniques
for bacterial detection. These methods are capable of identifying a small number of
14
pathogenic bacteria (down to single bacteria). However, cumbersome assay steps,
including pre-enrichment, selective enrichment, colony counting, biochemical screen-
ing, and serological con rmation, are generally required [1]. As a result, depending
on bacterial species and/or strains, these culture-based methods may take days to
weeks to yield results, which hinders their use in on-site bacterial detection. In addi-
tion, some viable bacteria in the environment may enter a dormant state and become
non-culturable (i.e., viable but non-culturable (VBNC) state), which leads to an un-
derestimation of the quantity of the bacteria or a failure to identify the bacteria
in a contaminated sample [2]. Table 2.1 shows examples of culture-based bacterial
detection and their assay times.
Table 2.1: Examples of culture-based bacterial detection.
Detected pathogen Assay time Ref.
Escherichia coli O157:H7 2 days [3]
Salmonella Enteritidis 4 to 8 days [4]
Listeria monocytogenes up to 7 days [5,6]
Campylobacter fetus 14 to 16 days [7]
2.1.2 Immunology-based methods
Immunology-based methods, the majority of which rely on an antibody - antigen
binding, have been widely used for the detection of pathogenic bacteria, including
Escherichia coli, Salmonella spp., Listeria monocytogenes, Campylobacter spp., and
Staphylococcal enterotoxins [2]. Among existing methods, enzyme-linked immunosor-
bent assay (ELISA), which is relatively rapid and versatile, is the most commonly
used technique. Figure 2.2 illustrates a typical procedure for sandwich ELISA, a
commonly used ELISA variant. The assay steps are as follows:
15
Figure 2.2: Typical procedure for sandwich ELISA.
1. Immobilize a capture antibody on the surface of each well of a microtiter plate
(often called an ELISA plate). Then, wash the plate so that any unbound
antibodies are removed.
2. Block any non-speci c adsorption sites on the well surface with a surface block-
ing agent (usually, bovine serum albumin or casein).
3. Apply a sample that contains a target antigen to the plate and allow the capture
antibody to bind with the antigen. Then, wash the plate to remove any unbound
antigens.
4. Add a primary antibody and allow it to bind with the antigen. Then, wash the
plate.
5. Add an enzyme-linked secondary antibody that binds with the primary anti-
body. Then, wash the plate.
16
6. Finally, add a substrate that can be converted by the enzyme into a color or
electrochemical signal for measurement.
Although ELISA-based methods are much more rapid than culture-based methods,
hours to days are still required to yield results [8{10]. In addition, their limits of detec-
tion (LODs) are not competitive with those of culture-based methods. Furthermore,
cumbersome assay procedures (i.e., a series of washing and addition of reagents) make
ELISA-based methods unsuitable for on-site bacterial detection. Table 2.2 shows ex-
amples of ELISA-based bacterial detection and their LODs.
Table 2.2: Examples of ELISA-based bacterial detection.
Detected pathogen LOD Ref.
Escherichia coli O157 103 to 104 cfu/ml [11,12]
Salmonella serovars 106 cells/ml [13]
Listeria monocytogenes 103 cells/ml [14]
Campylobacter fetus 105 cells/ml [7]
2.1.3 Polymerase chain reaction-based methods
The polymerase chain reaction (PCR) is a biochemical technique to produce mil-
lions of copies of a fragment of a nucleic acid (usually, deoxyribonucleic acid (DNA),
which is more stable than ribonucleic acid (RNA)). PCR-based methods have been
widely used to identify or detect pathogenic bacteria, including Salmonella aureus,
Listeria monocytogenes, Bacillus cereus, Escherichia coli O157:H7, and Campylobac-
ter jejuni [2]. As shown in Fig. 2.3, the PCR typically requires a series of 20 to 40
thermal cycles. In each cycle, there are three discrete temperature steps as described
below:
1. Denaturation: separating double-stranded DNA into a pair of single-stranded
DNA templates at a temperature of around 95  C.
17
Figure 2.3: Schematic illustration of the PCR cycle.
2. Annealing: allowing annealing of forward and reverse primers to the single-
stranded DNA templates at a temperature of 50 to 65  C.
3. Elongation: extending the primers with the aid of DNA polymerase to synthe-
size complementary strands at a temperature of around 70  C.
In this way, the number of amplicons doubles after each cycle, resulting in exponential
ampli cation in the amount of the target DNA sequence. Since each cycle requires
only several minutes, millions of amplicons can be produced within a few hours. After
thermal cycling, the  nal PCR products are typically analyzed by gel electrophoresis.
PCR-based methods possess the capability of detecting a small amount of target
DNA (down to a few DNA molecules [15, 16]) as well as o er high speci city and
18
accuracy. In addition, they are relatively rapid (i.e., a few hours of assay time) when
compared with culture- and immunology-based methods. However, their use in on-site
bacterial detection is restricted by a number of shortcomings. They require pure DNA
samples and speci c primers for avoiding false ampli cation, expensive reagents (e.g.,
DNA polymerase, deoxynucleotide triphosphate (dNTP), and other additives), and
hours of thermal cycles, followed by a gel electrophoresis-based analysis, which usually
takes one additional hour. In other words, these PCR-based methods are complex,
expensive, and still time-consuming. Although newer PCR variants, including real-
time PCR [17], digital PCR [18], and micro uidic PCR [19], can o er a much shorter
assay time with less volumes of reagents, the use of a  uorescent-labeled DNA probe
as well as an optical detector for the acquisition of  uorescence signals is additionally
needed, leading to an increase in cost and assay complexity. Furthermore, PCR-based
methods cannot generally discriminate between viable and non-viable cells [20], as
well as the extraction of DNA from a resistant bacterial spore, such as a B. anthracis
spore, remains a challenge [21]. Examples of PCR-based bacterial detection and their
LODs are shown in Table 2.3.
Table 2.3: Examples of PCR-based bacterial detection.
Detected pathogen LOD Ref.
Escherichia coli 102 cells/ml [22]
Salmonella Enteritidis 1 cfu/25 g or ml of food samples [23]
Listeria monocytogenes 3 cfu/g in ground beef [24]
Campylobacter spp. 100 to 150 cfu/ml [25]
Legionella pneumophila < 10 cfu/ml [26]
19
2.2 Biosensors as promising bacterial detection methods
Figure 2.4 shows the number of research articles published between 1985 and
2005 on di erent bacterial detection methods [20]. As can be seen, the most popular
detection methods were PCR-, culture-, and ELISA-based methods, which is due to
their low LODs and high reliability as mentioned in the previous sections. However,
in addition to these conventional methods, emerging biosensor technologies have been
drawing much attention in recent years. In fact, the global market for biosensors in
2012 is estimated to be a368.5 billion and projected to reach a3616.8 billion by 2018 [27].
Biosensor technologies come with promises of equally reliable results in much shorter
times [20].
Figure 2.4: Number of research articles published between 1985 and 2005 on di erent
bacterial detection methods.
2.2.1 De nition of a biosensor
A biosensor is an analytical device that converts a biological response into an
electrical signal. Two principal components of a biosensor are: (1) a biomolecular-
recognition element, which recognizes and speci cally binds with a target analyte,
20
and (2) a signal transducer, which converts the recognition event into a measurable
electrical signal.
Figure 2.5 shows a schematic diagram of a biosensor. When a biomolecular-
recognition event occurs, a signal can be instantaneously generated by the transducer.
This initial, small input signal from the transducer is, then, ampli ed, processed,
and sent to an output system for display or further analyses. Biosensors can be
rapid, sensitive, target-speci c, and portable, which makes them suitable for use in a
variety of  elds, including medical care, environmental monitoring, food safety, and
biosecurity. Biosensors can be classi ed by their recognition elements and/or signal
transduction methods as shown in Fig. 2.6.
Figure 2.5: Schematic diagram of a biosensor.
21
Figure 2.6: Classi cation of biosensors.
2.2.2 Biomolecular-recognition elements
Biomolecular-recognition elements are responsible for speci cally binding a tar-
get analyte to a biosensor. They are generally immobilized on the surface of a sig-
nal transducer. Antibodies, nucleic acids, and enzymes are three major types of
biomolecular-recognition elements as can be seen in Fig. 2.6. In recent years, how-
ever, a number of attempts have been made in employing landscape phages (i.e.,
genetically engineered phages) as the biomolecular-recognition element for the detec-
tion of pathogenic bacteria [28{31]. Landscape phages are highly tailorable, speci c,
relatively inexpensive, and thermally robust (much better than commonly used anti-
bodies) [32]. Hence, in this research, landscape phages were a nity-selected for the
target pathogenic bacteria (i.e., S. Typhimurium and B. anthracis spores) and used.
22
2.2.3 Signal transducers
Signal transducers are responsible for converting a biomolecular-recognition event
into a measurable electrical signal. In the past decades, a wide variety of transduction
methods has been developed. Among them, optical, electrochemical, and mass-based
methods are the most commonly used methods (Fig. 2.6). Each of these major types
of transducers contains many di erent subtypes. In addition, they can be further
classi ed into labeled and label-free methods. While the labeled methods depend on
the detection of a speci c label (e.g.,  uorescent, chemiluminescent, and radioactive
labels), the label-free methods are based on the direct measurement of a phenomenon
occurring on a transducer surface [2]. In this research, freestanding, strip-shaped mag-
netoelastic (ME) transducers, a novel class of mass-based transducers, were combined
with the above-mentioned landscape phages and used. Compared with conventional
mass-based transducers (e.g., piezoelectric transducers), ME transducers possess ad-
vantageous features, such as low-cost production, wireless signal transduction, and
thus,  exibility in biosensor design. These unique characteristics of ME transducers
facilitate on-site bacterial detection.
2.3 Conventional detection methods vs. biosensors
Table 2.4 compares the LODs and assay times of major bacterial detection meth-
ods. As mentioned earlier, the conventional detection methods possess low LODs, and
even single pathogenic bacteria can be detected with culture-based methods. How-
ever, the conventional methods are all time-consuming (i.e., up to weeks of assay
time required), and thus, the  nal results can not be obtained rapidly. By contrast,
biosensors are generally much more rapid (i.e., on the order of minutes and up to
hours of assay time), which is a distinct advantage for on-site bacterial detection
for food safety and biosecurity. However, their LODs need to be improved so that
they can be competitive with the conventional detection methods. In addition, other
23
Table 2.4: Comparison of the performance of major bacterial detection methods.
Detection method LOD Assay time Ref.
Conventional methods
Culture Down to a single cell Days to weeks [3{7]
ELISA 103 to 106 cfu/ml Hours to days [7,11{14]
PCR Down to a single cell Hours [16,22{26]
Optical biosensors
Surface plasmon resonance 50 to 105 cfu/ml 15 min to hours [33{36]
Resonant mirror 103 spores 10 min [37]
Interferometer 5  106 cfu/ml < 40 min [38]
Ring resonator 105 cfu/ml < 60 min [39]
Bioluminescence 10 cfu/ml 20 min [40]
Electrochemical biosensors
Amperometric 101 to 103 cfu/ml Minutes to hours [41{45]
Potentiometric 101 to 103 cfu/ml 30 min to 1.5 hr [46{48]
Impedimetric 2 cfu/ml 45 min [49]
Conductometric 61 cfu/ml 8 min [50]
Mass-based biosensors
Quartz-crystal microbalance 103 spores/ml < 30 min [51]
Love-wave < 200 spores/ml 5 min [52]
Silicon cantilevers 1 spore in air hours [53]
50 spores in water hours [53]
Piezoelectric cantilevers 300 spores/ml < 20 min [54]
Magnetoelastic cantilevers 105 cfu/ml < 120 min [55]
Magnetoelastic strips 103 cfu/ml < 30 min [56]
performance criteria previously shown in Table 1.1 need to be met before biosensors
become reliable alternatives.
24
2.3.1 Probability of detection: PCR vs. biosensors
Although there is still room for improvement, biosensors are promising tools for
pathogen detection, which may be explained with a statistical model reported by
Sabelnikov et al [15]. The model can be used to estimate the probability of detection
for an aerosolized pathogen (e.g., aerosolized B. anthracis spores) using a model
detector or its network. A model detector is de ned as a single device that consists of
an aerosol sampler and a detection device based on any of known bacterial detection
methods, such as PCR-, immunology-, and biosensor-based methods. In addition,
a network of model detectors consists of m single model detectors, each of which
operates in the same way and deals with the same amount of a sample that contains
a target pathogen. The assumptions for this statistical model are as follows [15]:
1. There is a space that contains aerosolized particles of a pathogen. These
pathogen particles are distributed in the space according to a Poisson distri-
bution with a parameter,  , which is equal to the mean concentration of the
pathogen.
2. The aerosol sampler intakes the air with a  ow rate, Ws, and concentrates the
aerosolized pathogen with an e ciency, Ke, into a liquid collective sample with
a volume, Vc.
3. The time of sampling is set equal to the time of inhalation of the pathogen by
an individual. Since the total number of the pathogens inhaled by the exposed
individual, Di, is equal to the concentration of the pathogen in the air multiplied
by the time of exposure and the inhalation rate, Wh, this assumption allows to
exclude time and concentration factors from all calculations.
4. n individual samples of identical volume, Vs, are simultaneously tested for de-
tection.
25
5. The pathogen can be detected in a single sample (i.e, n = 1) with a probability
of 100% only if its amount in the sample is greater than or equal to a certain
threshold value, I. In other words, the value I represents the LOD of the
detection device (i.e., the minimum detectable number of pathogens per sample
volume, Vs). For simpli cation, neither false positives nor false negatives are
allowed.
Based on the above assumptions, the probability of detection of the pathogen in
a single sample (i.e., n = 1), Pds, may be expressed as
Pds = 1 
I 1X
k=0
F(k); (2.1)
F(k) = ( Vs)
ke  Vs
k! ; (2.2)
 = KeDiWsW
hVc
; (2.3)
where F(k) is the probability of  nding exactly k pathogens in the sample volume,
Vs. In addition, the probability of detection with one model detector (i.e., m = 1
with n samples), Pdn, is equal to the probability that the pathogen can be detected
in at least one of n individual samples, which can be calculated by
Pdn = 1 (1 Pds)n: (2.4)
By analogy, the probability of detection with a network of m model detectors, Pdm,
can be computed by
Pdm = 1 (1 Pdn)m: (2.5)
The use of the above equations allows one to compare di erent bacterial detection
methods in terms of their probability of detection with respect to the inhalation dose
of a target pathogen. Here, as an example, comparisons will be made for PCR-
26
and biosensor-based methods. In all calculations, Ke and Vc were kept constant and
equal to 0.8 and 10 ml, respectively, which are currently used in the most advanced
commercial samplers [58, 59]. In addition, an inhalation rate (Wh) of 11 l/min for
adult humans [60] was used. For convenience, all the variables and their values used
are summarized in Table 2.5.
Table 2.5: Variables and their values used for the calculations of the probability of
detection.
Variable Values Ref.
Di: total number of inhaled pathogens 1 to 100
Ws: intake  ow rate of the sampler 1,000 l/min [57,58]
Ke: e cacy coe cient of the sampler (ratio of
the number of concentrated pathogens to the
number of sampled pathogens)
0.8 [58]
Vc: volume of a liquid collective sample 10 ml [58,59]
Vs: volume of an individual sample for detec-
tion
(1) 50 to 500  l for PCR
(2) 1 ml for biosensors
[15]
n: number of individual samples (1) 96 or 384 for PCR
(2) 1 for biosensors
[15]
I: detectable number of pathogens per sample
volume, Vs
(1) 15 for PCR
(2) 50 for biosensors
[15]
m: number of model detectors (1) 1 for PCR
(2) 1 to 4 for biosensors
Wh: inhalation rate for adult humans 11 l/min [60]
Figure 2.7 shows the results for PCR-based detectors. These detectors can si-
multaneously test 96 or 384 samples (i.e., n = 96 or 384, based on the standard 96- or
384-well format of PCR technologies) with I = 15, which is of the best commercial,
 eld-operated PCR device [15]. In addition, various values of Vs (50, 100, and 500  l)
were used in the calculations. For a standard PCR-based detector (n = 96 and Vs =
50  l), it was found that 36 and more pathogens can be detected with a probability
of 100% (solid curve). In addition, a slightly better result was obtained with n =
384, which reduces the minimum detectable number of pathogens with a probability
27
Figure 2.7: Probability of detection with respect to the inhalation dose of the target
pathogen for PCR-based detectors. The following values were used in the calculations:
Ws = 1,000 l/min, Ke = 0.8, I = 15, and m = 1.
of 100% to be 28 (dashed curve). However, for both detectors, detection of less than
10 pathogens was found to be hardly possible, which may lead to certain risks of
lethality in humans for some existing pathogens. For instance, Table 2.6 shows lethal
doses of B. anthracis spores in humans for di erent levels of lethality. The LD10, the
lethal dose at which 10% of the population is expected to die, may be as low as 50
to 98 spores, and even lower lethal doses for the LD5 and LD1 have been reported.
Comparing the above calculated and these reported lethal doses clearly indicates that
both of the standard PCR-based detectors would fail to detect B. anthracis spores in
doses that could still cause a lethality of 5%. Although an increase in the volume of
Table 2.6: Lethal doses of B. anthracis spores in humans.
Level of lethality Lethal dose [spores] Ref.
LD10 50 to 98 [61]
LD5 14 to 28 [61]
LD1 1 to 3 [61]
28
individual samples, Vs, further reduces the minimum detectable number of pathogens
(solid curves with squares and circles in Fig. 2.7), the use of such larger volumes per
sample (i.e., Vs = 100 or 500  l) is not practical because it obviously increases costs
due to a corresponding increase in the volumes of reagents required for the PCR (i.e.,
DNA polymerase, dNTPs, and other additives).
Figure 2.8: Probability of detection with respect to the inhalation dose of the tar-
get pathogen for biosensor-based detectors. The following values were used in the
calculations: Ws = 1,000 l/min, Ke = 0.8, Vs = 1 ml, and n = 1.
Figure 2.8 shows the results for biosensor-based detectors with various values of
I (10, 50, and 100 pathogens per Vs). Biosensors generally deal with a much larger
individual sample volume (1 ml) than standard PCR-based methods (50  l). Hence,
if the value of I is su ciently small, they could outperform standard PCR-based
detectors as shown in Table 2.7. It can be seen that the minimum detectable number
of pathogens for the biosensor-based detectors dramatically decreases as the value of
I is decreased. In addition, when a network of four detectors (i.e., m = 4) with I =
10 is used, down to four pathogens can be detected. Hence, these calculation results
indicate that the use of biosensors can be a better choice than PCR-based detection
29
methods. Particularly, the use of a network of biosensors o ers a promise in rapid,
sensitive detection of target pathogens.
Table 2.7: Minimum detectable number of pathogens for PCR- and biosensor-based
methods.
Detection method I Vs n m Minimum detectable number of
pathogens
Standard PCR 15 50  l 96 1 36
Biosensor 100 1 ml 1 1 29
Biosensor 50 1 ml 1 1 19
Biosensor 10 1 ml 1 1 9
Biosensor 10 1 ml 1 4 4
30
Bibliography
[1] J. J. Gooding, \Biosensor technology for detecting biological warfare agents:
Recent progress and future trends," Analytica Chimica Acta, vol. 559, no. 2,
pp. 137 { 151, 2006.
[2] V. Velusamy, K. Arshak, O. Korostynska, K. Oliwa, and C. Adley, \An overview
of foodborne pathogen detection: In the perspective of biosensors," Biotechnology
Advances, vol. 28, no. 2, pp. 232 { 254, 2010.
[3] G. Ngwa, R. Schop, S. Weir, C. Le on-Velarde, and J. Odumeru, \Detection
and enumeration of E. coli O157:H7 in water samples by culture and molecular
methods," Journal of Microbiological Methods, vol. 92, no. 2, pp. 164 { 172, 2013.
[4] B. Lungu, W. D. Waltman, R. D. Berghaus, and C. L. Hofacre, \Comparison of
a real-time PCR method with a culture method for the detection of Salmonella
enterica serotype Enteritidis in naturally contaminated environmental samples
from integrated poultry houses," Journal of food protection, vol. 75, no. 4, pp. 743
{ 747, 2012.
[5] E. de Boer and R. R. Beumer, \Methodology for detection and typing of food-
borne microorganisms," International Journal of Food Microbiology, vol. 50,
no. 12, pp. 119 { 130, 1999.
[6] S. Artault, J. L. Blind, J. Delaval, Y. Dureuil, and N. Gaillard, \Detecting
Listeria monocytogenes in food," International Food Hygiene, vol. 12, p. 23,
2001.
[7] B. Brooks, J. Devenish, C. Lutze-Wallace, D. Milnes, R. Robertson, and
G. Berlie-Surujballi, \Evaluation of a monoclonal antibody-based enzyme-linked
immunosorbent assay for detection of Campylobacter fetus in bovine preputial
washing and vaginal mucus samples," Veterinary Microbiology, vol. 103, no. 12,
pp. 77 { 84, 2004.
[8] K. D. King, J. M. Vanniere, J. L. LeBlanc, K. E. Bullock, and G. P. Anderson,
\Automated  ber optic biosensor for multiplexed immunoassays," Environmental
Science & Technology, vol. 34, no. 13, pp. 2845 { 2850, 2000.
[9] F. S. Ligler, G. P. Anderson, P. T. Davidson, R. J. Foch, J. T. Ives, K. D. King,
G. Page, D. A. Stenger, and J. P. Whelan, \Remote sensing using an airborne
biosensor," Environmental Science & Technology, vol. 32, no. 16, pp. 2461 {
2466, 1998.
31
[10] K. Tsuchiya and C. C. dUrsel, \Development of a sensitive ELISA using
three monoclonal antibodies against lipopolysaccharides to detect Xanthomonas
campestris pv. vesicatoria, the causal agent of bacterial spot of tomato and pep-
per," Journal of General Plant Pathology, vol. 70, pp. 21 { 26, 2004.
[11] B. W. Blais, J. Leggate, J. Bosley, and A. Martinez-Perez, \Comparison of  uo-
rogenic and chromogenic assay systems in the detection of Escherichia coli O157
by a novel polymyxin-based ELISA," Letters in Applied Microbiology, vol. 39,
no. 6, pp. 516 { 522, 2004.
[12] M. Feng, Q. Yong, W. Wang, H. Kuang, L. Wang, and C. Xu, \Development of
a monoclonal antibody-based ELISA to detect Escherichia coli O157:H7," Food
and Agricultural Immunology, pp. 1 { 7, 2012.
[13] L. P. Mans eld and S. Forsythe, \The detection of Salmonella serovars from
animal feed and raw chicken using a combined immunomagnetic separation and
ELISA method," Food Microbiology, vol. 18, no. 4, pp. 361 { 366, 2001.
[14] E. L. Crowley, C. K. O?Sullivan, and G. G. Guilbault, \Increasing the sensitivity
of Listeria monocytogenes assays: Evaluation using ELISA and amperometric
detection," Analyst, vol. 124, pp. 295 { 299, 1999.
[15] A. Sabelnikov, V. Zhukov, and R. Kempf, \Probability of real-time detection
versus probability of infection for aerosolized biowarfare agents: A model study,"
Biosensors and Bioelectronics, vol. 21, no. 11, pp. 2070 { 2077, 2006.
[16] K. M. Hong, H. Najjar, M. Hawley, and R. D. Press, \Quantitative real-time
PCR with automated sample preparation for diagnosis and monitoring of cy-
tomegalovirus infection in bone marrow transplant patients," Clinical Chemistry,
vol. 50, no. 5, pp. 846 { 856, 2004.
[17] M. Kubista, J. M. Andrade, M. Bengtsson, A. Forootan, J. Jon ak, K. Lind,
R. Sindelka, R. Sj oback, B. Sj ogreen, L. Str ombom, A. St ahlberg, and N. Zoric,
\The real-time polymerase chain reaction," Molecular Aspects of Medicine,
vol. 27, no. 23, pp. 95 { 125, 2006.
[18] B. Vogelstein and K. W. Kinzler, \Digital PCR," Proceedings of the National
Academy of Sciences, vol. 96, no. 16, pp. 9236 { 9241, 1999.
[19] J. Liu, M. Enzelberger, and S. Quake, \A nanoliter rotary device for polymerase
chain reaction," Electrophoresis, vol. 23, no. 10, pp. 1531 { 1536, 2002.
[20] O. Lazcka, F. J. D. Campo, and F. X. M. noz, \Pathogen detection: A per-
spective of traditional methods and biosensors," Biosensors and Bioelectronics,
vol. 22, no. 7, pp. 1205 { 1217, 2007.
[21] P. Belgrader, M. Okuzumi, F. Pourahmadi, D. A. Borkholder, and M. Northrup,
\A micro uidic cartridge to prepare spores for PCR analysis," Biosensors and
Bioelectronics, vol. 14, no. 1011, pp. 849 { 852, 2000.
32
[22] J. A. Higgins, S. Nasarabadi, J. S. Karns, D. R. Shelton, M. Cooper, A. Gbakima,
and R. P. Koopman, \A handheld real time thermal cycler for bacterial pathogen
detection," Biosensors and Bioelectronics, vol. 18, no. 9, pp. 1115 { 1123, 2003.
[23] C. Almeida, L. Cerqueira, N. Azevedo, and M. Vieira, \Detection of Salmonella
enterica serovar Enteritidis using real time PCR, immunocapture assay, PNA
FISH and standard culture methods in di erent types of food samples," Inter-
national Journal of Food Microbiology, vol. 161, no. 1, pp. 16 { 22, 2013.
[24] P. G. Klein and V. K. Juneja, \Sensitive detection of viable Listeria monocyto-
genes by reverse transcription-PCR.," Applied and Environmental Microbiology,
vol. 63, no. 11, pp. 4441 { 4448, 1997.
[25] M. Lund, S. Nordentoft, K. Pedersen, and M. Madsen, \Detection of Campy-
lobacter spp. in chicken fecal samples by real-time PCR," Journal of Clinical
Microbiology, vol. 42, no. 11, pp. 5125 { 5132, 2004.
[26] P. Villari, E. Motti, C. Farullo, and I. Torre, \Comparison of conventional culture
and PCR methods for the detection of Legionella pneumophila in water," Letters
in Applied Microbiology, vol. 27, no. 2, pp. 106 { 110, 1998.
[27] http://www.pharmaceutical-market-research.com/publications/ nance
investment/biosensors global market overview.html.
[28] S. Huang, H. Yang, R. Lakshmanan, M. Johnson, J. Wan, I.-H. Chen, H. W.
III, V. Petrenko, J. Barbaree, and B. Chin, \Sequential detection of Salmonella
typhimurium and Bacillus anthracis spores using magnetoelastic biosensors,"
Biosensors and Bioelectronics, vol. 24, no. 6, pp. 1730 { 1736, 2009.
[29] M. L. Johnson, J. Wan, S. Huang, Z. Cheng, V. A. Petrenko, D.-J. Kim, I.-
H. Chen, J. M. Barbaree, J. W. Hong, and B. A. Chin, \A wireless biosensor
using microfabricated phage-interfaced magnetoelastic particles," Sensors and
Actuators A: Physical, vol. 144, no. 1, pp. 38 { 47, 2008.
[30] R. S. Lakshmanan, R. Guntupalli, J. Hu, D.-J. Kim, V. A. Petrenko, J. M.
Barbaree, and B. A. Chin, \Phage immobilized magnetoelastic sensor for the de-
tection of Salmonella typhimurium," Journal of Microbiological Methods, vol. 71,
no. 1, pp. 55 { 60, 2007.
[31] J. Wan, H. Shu, S. Huang, B. Fiebor, I.-H. Chen, V. A. Petrenko, and B. A. Chin,
\Phage-based magnetoelastic wireless biosensors for detecting Bacillus anthracis
spores," IEEE Sensors Journal, vol. 7, pp. 470 { 477, 2007.
[32] V. A. Petrenko, \Landscape phage as a molecular recognition interface for de-
tection devices," Microelectronics Journal, vol. 39, no. 2, pp. 202 { 207, 2008.
[33] G. C. Bokken, R. J. Corbee, F. van Knapen, and A. A. Bergwer , \Immunochem-
ical detection of Salmonella group B, D and E using an optical surface plasmon
33
resonance biosensor," FEMS Microbiology Letters, vol. 222, no. 1, pp. 75 { 82,
2003.
[34] B.-K. Oh, Y.-K. Kim, K. W. Park, W. H. Lee, and J.-W. Choi, \Surface plasmon
resonance immunosensor for the detection of Salmonella typhimurium," Biosen-
sors and Bioelectronics, vol. 19, no. 11, pp. 1497 { 1504, 2004.
[35] D. Zhang, Y. Yan, Q. Li, T. Yu, W. Cheng, L. Wang, H. Ju, and S. Ding, \Label-
free and high-sensitive detection of Salmonella using a surface plasmon resonance
DNA-based biosensor," Journal of Biotechnology, vol. 160, no. 34, pp. 123 { 128,
2012.
[36] Y. Wang, W. Knoll, and J. Dostalek, \Bacterial pathogen surface plasmon reso-
nance biosensor advanced by long range surface plasmons and magnetic nanopar-
ticle assays," Analytical Chemistry, vol. 84, no. 19, pp. 8345 { 8350, 2012.
[37] T. E. Love, C. Redmond, and C. N. Mayers, \Real time detection of anthrax
spores using highly speci c anti-EA1 recombinant antibodies produced by com-
petitive panning," Journal of Immunological Methods, vol. 334, no. 12, pp. 1 {
10, 2008.
[38] B. H. Schneider, J. G. Edwards, and N. F. Hartman, \Hartman interferometer:
versatile integrated optic sensor for label-free, real-time quanti cation of nucleic
acids, proteins, and pathogens," Clinical Chemistry, vol. 43, no. 9, pp. 1757 {
1763, 1997.
[39] A. Ramachandran, S. Wang, J. Clarke, S. Ja, D. Goad, L. Wald, E. Flood,
E. Knobbe, J. Hryniewicz, S. Chu, D. Gill, W. Chen, O. King, and B. Little, \A
universal biosensing platform based on optical micro-ring resonators," Biosensors
and Bioelectronics, vol. 23, no. 7, pp. 939 { 944, 2008.
[40] J. Qiu, Y. Zhou, H. Chen, and J.-M. Lin, \Immunomagnetic separation and
rapid detection of bacteria using bioluminescence and micro uidics," Talanta,
vol. 79, no. 3, pp. 787 { 795, 2009.
[41] Z. Muhammad-Tahir and E. Alocilja, \Fabrication of a disposable biosensor for
Escherichia coli O157:H7 detection," Sensors Journal, IEEE, vol. 3, pp. 345 {
351, Aug. 2003.
[42] Z. Muhammad-Tahir and E. Alocilja, \A disposable biosensor for pathogen de-
tection in fresh produce samples," Biosystems Engineering, vol. 88, no. 2, pp. 145
{ 151, 2004.
[43] C. Ruan, H. Wang, and Y. Li, \A bienzyme electrochemical biosensor coupled
with immunomagnetic separation for rapid detection of Escherichia coli O157:H7
in food samples," Transactions of the ASABE, vol. 45, no. 1, pp. 249 { 255, 2002.
34
[44] L. Yang, C. Ruan, and Y. LI, \Rapid detection of Salmonella typhimurium in
food samples using a bienzyme electrochemical biosensor with  ow injection,"
Journal of Rapid Methods & Automation in Microbiology, vol. 9, no. 4, pp. 229
{ 240, 2001.
[45] O. Laczka, J.-M. Maesa, N. Godino, J. del Campo, M. Fougt-Hansen, J. P.
Kutter, D. Snakenborg, F.-X. Munoz-Pascual, and E. Baldrich, \Improved bac-
teria detection by coupling magneto-immunocapture and amperometry at  ow-
channel microband electrodes," Biosensors and Bioelectronics, vol. 26, no. 8,
pp. 3633 { 3640, 2011.
[46] C. Ercole, M. D. Gallo, L. Mosiello, S. Baccella, and A. Lepidi, \Escherichia
coli detection in vegetable food by a potentiometric biosensor," Sensors and
Actuators B: Chemical, vol. 91, pp. 163 { 168, 2003.
[47] A. G. Gehring, D. L. Patterson, and S.-I. Tu, \Use of a light-addressable po-
tentiometric sensor for the detection of Escherichia coli O157:H7," Analytical
Biochemistry, vol. 258, no. 2, pp. 293 { 298, 1998.
[48] G. A. Zelada-Guill en, J. L. Sebasti an-Avila, P. Blondeau, J. Riu, and F. X. Rius,
\Label-free detection of Staphylococcus aureus in skin using real-time potentio-
metric biosensors based on carbon nanotubes and aptamers," Biosensors and
Bioelectronics, vol. 31, no. 1, pp. 226 { 232, 2012.
[49] M. dos Santos, J. Agusil, B. Prieto-Simn, C. Sporer, V. Teixeira, and J. Sami-
tier, \Highly sensitive detection of pathogen E. coli O157:H7 by electrochemi-
cal impedance spectroscopy," Biosensors and Bioelectronics, In Press, Accepted
Manuscript, 2013.
[50] Y. Luo, S. Nartker, H. Miller, D. Hochhalter, M. Wiederoder, S. Wiederoder,
E. Setterington, L. T. Drzal, and E. C. Alocilja, \Surface functionalization of
electrospun nano bers for detecting E. coli O157:H7 and BVDV cells in a direct-
charge transfer biosensor," Biosensors and Bioelectronics, vol. 26, no. 4, pp. 1612
{ 1617, 2010.
[51] R. Hao, D. Wang, X. Zhang, G. Zuo, H. Wei, R. Yang, Z. Zhang, Z. Cheng,
Y. Guo, Z. Cui, and Y. Zhou, \Rapid detection of Bacillus anthracis using mon-
oclonal antibody functionalized QCM sensor," Biosensors and Bioelectronics,
vol. 24, no. 5, pp. 1330 { 1335, 2009.
[52] D. W. Branch and S. M. Brozik, \Low-level detection of a Bacillus anthracis
simulant using Love-wave biosensors on 36 YX LiTaO3," Biosensors and Bio-
electronics, vol. 19, no. 8, pp. 849 { 859, 2004.
[53] A. P. Davila, J. Jang, A. K. Gupta, T. Walter, A. Aronson, and R. Bashir,
\Microresonator mass sensors for detection of Bacillus anthracis Sterne spores
in air and water," Biosensors and Bioelectronics, vol. 22, no. 12, pp. 3028 { 3035,
2007.
35
[54] G. A. Campbell and R. Mutharasan, \Detection of Bacillus anthracis spores and
a model protein using PEMC sensors in a  ow cell at 1 mL/min," Biosensors
and Bioelectronics, vol. 22, no. 1, pp. 78 { 85, 2006.
[55] L. Fu, K. Zhang, S. Li, Y. Wang, T.-S. Huang, A. Zhang, and Z.-Y. Cheng,
\In situ real-time detection of E. coli in water using antibody-coated magne-
tostrictive microcantilever," Sensors and Actuators B: Chemical, vol. 150, no. 1,
pp. 220 { 225, 2010.
[56] J. Wan, M. L. Johnson, R. Guntupalli, V. A. Petrenko, and B. A. Chin, \De-
tection of Bacillus anthracis spores in liquid using phage-based magnetoelastic
micro-resonators," Sensors and Actuators B: Chemical, vol. 127, no. 2, pp. 559
{ 566, 2007.
[57] M. Jantunen, O. H anninen, K. Koistinen, and J. Hashim, \Fine PM measure-
ments: personal and indoor air monitoring," Chemosphere, vol. 49, no. 9, pp. 993
{ 1007, 2002.
[58] P. M. Irving, \Air sampling technology: A requirement for biodetection," in
Materials of the Second International Symposium on Detection Technologies,
Knowledge Foundation, (Arlington, VA, USA), December 2002.
[59] http://www.sceptorindustries.com.
[60] M. Allan and G. M. Richardson, \Probability density functions describing 24 h
inhalation rates for use in human health risk assessments," Human and Ecological
Risk Assessment, vol. 4, no. 2, pp. 379 { 408, 1998.
[61] C. J. Peters and D. M. Hartley, \Anthrax inhalation and lethal human infection,"
The Lancet, vol. 359, pp. 710 { 711, 2002.
36
Chapter 3
Phage-Based Magnetoelastic (ME) Biosensors
Biosensors can be potential alternatives to the conventional bacterial detection
methods. However, as described in the previous chapters, the performance of existing
biosensors still needs to be improved. In this chapter, the fundamentals of phage-
based ME biosensors, a novel class of wireless, mass-sensitive biosensors, will be
described in depth.
3.1 Landscape phages as biomolecular-recognition elements
For the past decades, antibodies have been the most commonly used biomolecular-
recognition elements [2]. However, their use in on-site bacterial detection might be
restricted by such factors as thermal stability, selectivity, and production cost [1,3,4].
Hence, as emerging alternatives, landscape phages have attracted growing attention,
and their application to various biosensing systems has been recently reported [3,5{8].
Table 3.1 compares the longevity of a landscape phage and monoclonal antibody spe-
ci c for  -galactosidase tested at various temperatures [1]. As can be seen, both
Table 3.1: Longevity of a landscape phage and monoclonal antibody at various tem-
peratures.
Temperature Landscape phage Antibody Ref.
Room temp. > 6 months > 6 months [1]
37  C 950 days (half-life) 107 days (half-life) [1]
50  C 5 weeks (half-life) 5 weeks [1]
63  C 6 weeks 24 hr [1]
76  C 2.4 days No binding activity [1]
37
phage and antibody retain their binding activities at room temperature for greater
than 6 months. However, the antibody degrades much faster than the phage at higher
temperatures. For on-site bacterial detection, where a wide range of temperatures are
anticipated, a high thermal stability with a fair life time is essential for a biomolecular-
recognition element. In addition, as shown in Table 3.2, a high selectivity and low
production cost are distinct advantages of landscape phages over antibodies.
Table 3.2: Comparison between landscape phages and antibodies in terms of selec-
tivity and production cost.
Recognition element Selectivity Production cost Ref.
Landscape phage High Low [1,3,4]
Monoclonal antibody High Very high [1,4]
Polyclonal antibody Low High [1,4]
Landscape phages are genetically engineered phages that can be synthesized
through the phage display technology [9, 10]. This technology, primarily developed
Figure 3.1: Schematic illustration of the wild-type fd phage and its genetically engi-
neered form, displaying a foreign peptide on the major coat protein pVIII.
38
for the Ff class of  lamentous phage strains (i.e., fd, f1, and M13), enables one to
construct billions of phage clones that display engineered sequences of peptides on
their outer surfaces. For example, Fig. 3.1 shows a schematic illustration of the
wild-type fd phage (top) and its genetically engineered form (bottom), displaying a
foreign peptide on the major coat protein pVIII.
The wild-type Ff phage strains, which possess virtually identical DNA sequences,
are  exible, thread-like particles about 800 to 900 nm long and 6.5 nm in diameter [1].
They consist of a circular single-stranded DNA ( 6,400 nucleotides) enclosed in a
tube of helically arranged molecules of coat proteins (the N-termini exposed on the
outer surface and C-termini in the lumen) [10, 11]. There are approximately 2,700
copies of the major coat protein pVIII along the tube?s length, accounting for 98%
by mass [1]. In addition,  ve copies each of the minor coat proteins cap both ends
Figure 3.2: Sequences of amino acid residues of the fd coat proteins. The N-terminus
is to the left. The hydrophobic domains are underlined, whereas charged residues are
indicated by + or -.
39
of the tube (The minor coat protein pIII and pVI are at one end, whereas the pVII
and pIX proteins are at the other end.). Figure 3.2 shows the sequences of amino
acid residues of the fd coat proteins (The N-terminus is to the left.) [12]. The sizes
of these proteins are: pVIII, 50 residues; pIII, 406 residues; pVI, 112 residues; PVII,
33 residues; and pIV, 32 residues. The N-terminal portion of the protein pIII can
be considered as three distinct domains, separated by striking glycine-rich tandem
repeat linkers of GGGS and EGGGS between the domains [12]. The numbers within
the circles represent amino acid residues assigned to each domain. In the above  gure,
apolar, hydrophobic domains are underlined, whereas charged residues are indicated
by + or -. The major coat protein pVIII, for example, has a hydrophobic domain
of continuous 19 amino acid residues (YIGYAWAMVVVIVGATIGI) in the interior
of its sequence. Adjacent copies of this protein in the phage virion are held together
by hydrophobic interactions between these domains [12]. In addition, the positively
charged residues near the C-terminus neutralize the negative charge of the DNA
core. Furthermore, all the four minor coat proteins also possess hydrophobic domains
similar in length to the hydrophobic domain of pVIII, suggesting that these minor
coat proteins may associate with pVIII by hydrophobic interactions [12]. According to
Endemann and Model [13], the minor coat proteins pIII, pVI, and pVII all interact
with the major coat protein pVIII in phage. Also, the pIII and pIX proteins are
exposed to the environment, whereas the pVI and PVII proteins are shielded from
the environment.
These wild-type Ff phages are viruses that infect the bacterium Escherichia coli
bearing F pili. Infection is initiated by the attachment of the N-terminal domain
of the pIII protein to the tip of the pilus [10]. As the process continues, the coat
proteins dissolve into the surface envelope of the cell, and the viral DNA alone enters
the cytoplasm, where a vast number of progeny viral DNA molecules are synthesized
by host machinery. These progeny viral DNA molecules are, then, extruded through
40
the cell envelope, acquiring the coat proteins from the cell membrane and emerging
as completed virions [10]. Up to 1,000 progeny virions per cell per division can be
secreted continuously without killing the host cell [1], leading to a low production
cost. The yield of virions can exceed 0.3 mg/mL [10].
In phage display constructions, a foreign coding sequence is spliced in-frame into
one of the  ve coat protein genes. The resultant foreign peptide encoded by this
sequence can be fused to the coat protein and, thereby, displayed on the surface of
the virion. In addition, the subsequent length of the phage capsid (i.e., the protein
shell of the phage) is altered to match the size of the enclosed recombinant DNA by
adding proportionally more pVIII subunits during phage assembly [14].
In this research, three phage clones (E2, JRB7, and SAE10), displaying foreign
octamers or nanomers in approximately 4,000 copies of the major coat protein pVIII,
were derived from the landscape phage libraries f8/8 and f8/9 [14{16] and used as
biomolecular-recognition elements. As shown in Fig. 3.3, three amino acid residues
(EGD) of the wild-type pVIII are replaced by a random octamer in the f8/8 library,
whereas four residues (EGDD) are replaced by a random nanomer in the f8/9 library,
bringing the total size of both fusion pVIII proteins to 55 amino acids. Here, the
symbol x represents any amino acid residue. About a half of the pVIII peptide
Figure 3.3: Sequences of amino acid residues of the wild-type and fusion pVIII pro-
teins.
41
sequence is exposed to the environment, whereas the other half is buried in the capsid
[1]. The foreign peptide sequence for each phage clone is shown in Table 3.3.
Table 3.3: Phage clones used in this research.
Phage Foreign peptide sequence Target pathogen or analyte Ref.
E2 VTPPTQHQ S. Typhimurium [15]
JRB7 EPRLSPHS B. anthracis [16]
SAE10 VPVGAYSDT Streptavidin [14]
3.2 Magnetoelasticity
Any magnetic materials exhibit magnetoelastic (ME) behaviour. In other words,
the dimensions and elastic properties of these materials are dependent upon their
magnetic states, and their magnetic properties are, by contrast, in uenced by internal
as well as applied mechanical stresses [17]. Magnetoelasticity has been observed not
only in ferromagnets but also in ferrimagnets, antiferrimagnets, paramagnets, and
even daimagnets with low susceptibilities [17]. However, from a technical point of
view, ferromagnets have been extensively studied for the past centuries. In this work,
two amorphous ferromagnets, Metglas Alloy 2826MB and Fe79B21, were used for the
construction of ME signal transducers.
3.2.1 Joule magnetostriction
An ME material undergoes a change in its dimensions during the process of
magnetization. This phenomenon is known as Joule magnetostriction, discovered by
and named after James P. Joule [17]. A spherical ME material, for example, may be
transformed into an ellipsoid when subjected to an externally applied magnetic  eld,
H, as illustrated in Fig. 3.4. The induced strains measured in the directions parallel
and perpendicular to the  eld,  k and  ?, are largely dependent on the  eld strength
42
Figure 3.4: Joule magnetostriction of a spherical ME material.
and can be positive or negative, depending on whether the material?s deformation is
expansive or compressive. These strains reach their limiting values when the material
becomes magnetically saturated. From the hypothetical  { H relationships shown
in Fig. 3.5, the saturation Joule magnetostriction,  s, can be de ned as [18]
 s = 23( k  ?); (3.1)
where  k and  ? are determined by the extrapolation of the tangents of the linear
 eld dependencies of the strains at the saturated state down to H = 0.  s is equal
to  k when the demagnetized state of the material is isotropic (i.e.,  k/2 ? = 1).
However, for actual materials, the value of  k/2 ? is often not equal to 1, depending
on both intrinsic parameters (e.g., magnetocrystalline anisotropy) and sample pa-
rameters (e.g., demagnetizing  eld and internal stresses) [17]. Hence, Eq. 3.1, which
43
remains independent of the demagnetized state of the material, is preferably used to
determine  s.
Figure 3.5: Hypothetical  eld dependencies of  k and  ?.
3.2.2 Magnetization and Joule magnetostriction in ferromagnets
When demagnetized at temperatures lower than its Curie temperature, a fer-
romagnetic material is divided into a number of magnetic domains to minimize the
material?s internal energy as schematically illustrated in Fig. 3.6. Each magnetic do-
main is magnetized in a di erent direction such that the net magnetization is zero or
small. Between any two adjacent domains, the elementary magnetic moments rotate
gradually from one easy magnetization direction (Di) to another (Dj) [17]. When
a magnetic  eld, H, is applied in any given direction, the following processes occur
within the material:
1. The domain D1, whose magnetization direction is the closest to the  eld direc-
tion, expands at the expense of the domain D3 through the displacement of the
180 domain wall.
44
2. The magnetization in the domain D4 is reversed so that the net magnetization
is minimized.
3. When a stronger  eld is applied, the domain D1 further expands at the expense
of the domains D2 and D4 (90 domain wall displacement), resulting in the
formation of only one single domain D1.
4. Finally, the magnetization in this single domain rotates out of its easy magne-
tization direction (D1) and aligns along the  eld direction.
Figure 3.6: Magnetic domains and magnetization processes in a ferromagnet.
During the above magnetization processes, the material is strained due to the
ME coupling since the distribution of the elementary magnetic moments becomes
anisotropic (upper left illustration in Fig. 3.7a). This anisotropic distribution of mag-
netic moments may also be induced when an ME material is mechanically stressed
(upper right illustration in Fig. 3.7a). As a result, spontaneous Joule magnetostric-
tion is also induced, again due to the ME coupling. When a magnetic  eld is, then,
applied to this material,  k becomes a function of the applied pre-stress [17]. In
the case of positive Joule magnetostriction, a compressive stress orients the mag-
netic moments into the direction perpendicular to the stress direction. Such  eld
45
and stress dependence of  k is shown in Fig. 3.7b. As can be seen,  k is increased
with larger pre-stresses, which rotate more magnetic moments (Note that  s is, by
contrast, usually nearly stress-independent [17].).
Figure 3.7: (a) E ects of magnetizing  eld and mechanical stress on the distribu-
tion of magnetic moments in a ferromagnet and (b)  eld dependence of  k under a
compressive stress.
Figure 3.8: Temperature dependence of normalized  s in Fe80B20.
46
Joule magnetostriction is also temperature dependent. The thermal dependence
of  s is usually monotonous as in the case for Fe80B20, shown in Fig. 3.8, where Tc is
the Curie temperature [17].
3.2.3 ME signal transducers
An ideal material for the construction of ME signal transducers is a magnetically
soft material that possesses a high saturation Joule magnetostriction,  s, and a high
magnetomechanical coupling factor, k, which can be de ned as [17]
k = EmepE
eEm
; (3.2)
whereEme, Ee, andEm are the mutual elastic and magnetic energy density, elastic self-
energy density, and magnetic self-energy density, respectively. This coupling factor
can be used to characterize the material?s ability to convert a magnetic energy into an
elastic energy and vice versa. Traditionally, Metglas Alloy 2826MB (from Honeywell
International) with a composition of Fe40Ni38Mo4B18 has been the material for ME
signal transducers. This amorphous, ferromagnetic alloy is mechanically robust (e.g.,
tensile strength of 1 to 2 GPa), and it possesses reasonably high  s and k values
[19{22]. In addition, the alloy has a low material cost, allowing ME signal transducers
made of this alloy to be used on a disposable basis [23].
In recent years, another amorphous, ferromagnetic alloy with a composition close
to Fe80B20 has also been used for the construction of ME signal transducers [7, 24].
This iron-rich alloy can be easily produced through physical or electrochemical de-
position processes. Hence, by combining with standard microelectronic fabrication
techniques, batch fabrication of miniature ME signal transducers is also possible,
which further reduces the fabrication cost. Some important materials properties for
the above alloys are summarized in Table 3.4.
47
Table 3.4: Materials properties for Metglas 2826MB and Fe80B20.
Property Metglas 2826MB Ref. Fe80B20 Reference
Elastic modulus [GPa] 100 to 110 [22] 166 [25]
Density [ 103 kg/m3] 7.9 [22] 7.4 [25]
Poisson?s ratio 0.33 [26] 0.3 [27]
Saturation magnetostriction,  s
[ppm]
12 [22] 32 [28]
Magnetomechanical coupling fac-
tor, k
 0.98 [23]  0.64 [29]
3.3 Fabrication of ME sensor platforms
Depending on the size of sensor platforms (i.e., ME signal transducers), two dif-
ferent fabrication methods were employed. A dicing method was used for millimeter-
scale sensor platforms (0.5 to 4-mm long), while a co-sputtering-based method was
used for micron-scale sensor platforms (100 to 500- m long).
3.3.1 Dicing method
A ribbon of Metglas Alloy 2826MB was purchased from Honeywell International.
Small pieces with a size of 50 mm 12.7 mm 30  m were cut from the ribbon and
double-side polished down to a thickness of 15  m. The polished pieces were, then,
diced into millimeter-scale strip-shaped sensor platforms (Fig. 3.9) using a automated
dicing saw. After cleaning with acetone and ethanol in an ultrasonic bath, these diced
sensor platforms were successively coated with thin layers of Cr (90 nm) and Au (150
nm) by electron-beam induced deposition. The Cr layer acts as an adhesive interlayer
between the Au layer and sensor platform. The Au layer provides corrosion resistance
as well as a ready surface for the immobilization of the phages.
48
Figure 3.9: Diced sensor platforms stored in dry methanol.
3.3.2 Co-sputtering-based method
Micron-scale ME sensor platforms were batch-fabricated using a co-sputtering-
based method reported previously [7]. The fabrication procedure is diagrammed in
Fig. 3.10. First, a four-inch gold-coated wafer was photolithographically patterned
with 10- m thick rectangular islands of the STR-1045 photoresist (from Rohm and
Haas Electronic Materials, LLC). The lateral dimensions of these photoresist islands
were kept the same as those of sensor platforms of target size. Next, the patterned
wafer was successively deposited with 50-nm thick Au, 4- m thick Fe79B21 (i.e., co-
sputtering of Fe and B), and another 50-nm thick Au using a Denton sputter coater
(Fig. 3.11). Then, the Au-enclosed Fe79B21 alloy on the photoresist islands was
lifted o the wafer to become freestanding ME sensor platforms by dissolving the
underlying photoresist with acetone. This method not only ensures the dimensional
49
Figure 3.10: Procedure for the co-sputtering-based method.
consistency of the fabricated sensor platforms but also greatly reduces the fabrication
cost per sensor platform due to the batch fabrication process (See Fig. 3.12, showing
batch-fabricated sensor platforms with a size of 100  m  25  m  4  m.).
50
Figure 3.11: Denton sputter coater.
51
Figure 3.12: Scanning electron micrograph of batch-fabricated sensor platforms with
a size of 100  m  25  m  4  m on a gold-coated wafer.
The sputtering conditions used are summarized in Table 3.5. Prior to sputtering,
the chamber was pumped down to 7 10 7 Torr in order to minimize residual oxygen
in the fabricated sensor platforms. In addition, during the deposition process, the
sample stage was rotated so that the uniformity of the deposits can be guaranteed.
Table 3.5: Sputtering conditions used for the fabrication of micron-scale sensor plat-
forms.
Target Cathode type Power [W] Time [sec] Ar  ow rate [sccm]
Cr DC 50 100 25
Au DC 100 200  3 times 12
Fe DC 41 64,000 30
B RF 100 64,000 30
52
3.3.3 Annealing
The fabricated sensor platforms were  nally annealed in vacuum at 220  C for 2
h to relieve residual internal stresses and minimize the e ects of any surface defects
from the fabrication processes [30]. This temperature was chosen because it is high
enough to e ectively anneal the sensor platforms within a reasonable time, yet it is
still far below the Curie temperatures as well as the recrystallization temperatures of
the ME materials [7].
3.3.4 Fabrication cost per sensor platform
As discussed in Chapter 1, cost is an important factor for the on-site detection
of pathogens because proper allocation of limited resources is essential to improving
overall safety. Through the batch fabrication of micron-scale sensor platforms, great
reduction in fabrication cost can be realized. For a rough estimation of the fabrication
cost per sensor platform, the following assumptions were made:
1. The size of sensor platforms to be fabricated is 200  m  40  m  4  m, and
their composition is Fe79B21.
2. The size of a silicon wafer on which sensor platforms will be fabricated is four
inches in diameter. 80% of the wafer surface area will be used.
3. Only the material costs for metals to be sputtered (i.e., Cr, Au, Fe, and B as
shown in Table 3.6) and the silicon wafer ( a3620) are considered. In other
words, the costs for the STR-1045 photoresist and other required chemicals are
not considered.
4. The metals will be sputter-deposited only on the wafer.
5. The electricity costs for photolithography, annealing, and other fabrication pro-
cesses are not considered. Only the energy charges for sputtering are considered.
53
An electricity rate of 7.49a162/kWh paid by Auburn University in FY 2009 [31] is
used.
Table 3.6: Material costs for the sputtering targets.
Material Cost [a36/g] Density [g/cm3] Deposit?s thickness Ref.
Cr  1 7.2 10 nm [32]
Au  75 19.3 150 nm (50 nm  3) [33]
Fe  2 7.9 3.2  m (79% of 4  m) [34]
B  2 2.3 0.8  m (21% of 4  m) [35]
Hence, the fabrication cost per sensor platform, Csensor, can be calculated by
Csensor = material costs + energy chargesnumber of sensors
=
 X
(Mi  i Vi) + wafer cost
+
X
(Pi Ti $0:0749=kWh)
 
=Nsensor; (3.3)
i = Cr; Au; Fe; or B;
where M,  , and V represent the material cost (per gram), density, and volume for
each sputtered metal; P and T denote the power (in kilo-watts) and time (in hours)
for sputtering; and Nsensor is the number of sensors to be fabricated on the wafer. By
using the values shown in Tables 3.5 and 3.6,
Csensor = ($5:8 10 4 + $1:8 + $4:1 10 1 + $3:0 10 2 + $20
+ $1:0 10 4 + $1:3 10 3 + $5:5 10 2 + $1:3 10 1)=8:1 105 sensors
 $2:8 10 5: (3.4)
54
3.4 Fabrication of phage-based ME biosensors
3.4.1 Immobilization of a phage on the ME sensor platforms
The annealed sensor platforms were individually immersed in 330  l of a phage
suspension (usually, phage in a tris-bu ered saline (TBS) bu er solution, 5  1011
vir/ml) in a polypropylene PCR tube. The tubes were, then, rotated with a Barn-
stead LabQuake tube rotator (from Fisher Scienti c, Inc.) at 8 rpm for 1 h. In
this way, the phage was allowed to uniformly attach to platform surfaces via phys-
ical adsorption. Finally, these phage-immobilized ME biosensors (i.e., measurement
sensors) were thoroughly rinsed with sterile distilled water to remove any TBS bu er
components as well as loosely attached phages from the platform surfaces. Covalent
immobilization of a phage on the surface of sensor platforms will be described in
Chapter 6.
3.4.2 Surface blocking of the ME biosensors with bovine serum albumin
In order to reduce non-speci c adsorption of S. Typhimurium cells or B. anthracis
spores on biosensor surfaces, surface blocking with bovine serum albumin (BSA) was
performed. The prepared measurement sensors were individually immersed in a 330-
 l solution of BSA (0.01 to 1 % w/v in sterile distilled water) in a PCR tube. After
40 min of tube rotation at 8 rpm, the biosensors were collected from the solution and
thoroughly rinsed with sterile distilled water to be ready for use. Control sensors,
which are not immobilized with phage but only surface-blocked with BSA, were also
prepared and used for background subtraction.
55
3.5 Principle of detection
The fabricated ME biosensors are made from one of the magnetostrictive alloys
(i.e., Metglas 2826MB or Fe79B21). Hence, the biosensors can be placed into magneto-
mechanical resonance when subjected to an externally applied magnetic  eld that
alternates at the right frequency. For a freestanding, strip-shaped biosensor, the
fundamental resonant frequency of longitudinal vibration, f, can be expressed by [26]
f = 12L
s
E
 (1  ); (3.5)
where L, E,  , and  denote the length, modulus of elasticity, density, and Poisson?s
ratio of the biosensor, respectively. When the biosensor and bacterial cells (or spores)
come into contact with each other, the phage that is immobilized on the biosensor
binds with the cells (or spores) (Fig. 3.13), thereby increasing the total mass of the
biosensor. This change in mass causes a corresponding decrease in the biosensor?s
resonant frequency, which is the principle of detection. In addition, for uniform
mass attachment, the mass sensitivity of the biosensor, Sm, de ned as the ratio of
the resonant frequency change,  f, to the mass change  m, can be approximated
Figure 3.13: Uniform attachment of bacterial cells or spores on a phage-immobilized
ME biosensor.
56
by [23,36]
Sm =  f m  14L2WT
s
E
 3(1  ); (3.6)
where W and T represent the width and thickness of the biosensor, respectively.
Equation 3.6 describes that the mass sensitivity is largely dependent on the size of
the biosensor and inversely proportional to L2WT.
3.5.1 Minimum detectable number of bacterial cells
Table 3.7 shows ME biosensors of di erent size and their theoretical detection
limits in terms of the minimum detectable number of bacterial cells, Nmin. The  rst
four biosensors from the top are made of Metglas 2826MB, whereas the rest is made
of Fe79B21. By rearranging Eq. 3.6, the minimum detectable mass change,  mmin,
can be calculated by
 mmin =  fminS
m
; (3.7)
where  fmin represents the corresponding minimum detectable frequency change. If
the mass of a single bacterial cell is assumed to be 1 pg [37], and a frequency shift of
1,000 Hz (i.e.,  fmin = 1,000 Hz) can be resolved, the minimum detectable number
of bacterial cells, Nmin, can then be computed by
Nmin =  mmin1 pg =
 fmin
Sm
1 pg =
1;000 Hz
1 pg  
1
Sm; (3.8)
where Nmin is an integer. As can be seen in the table, detection of a single bacterial
cell may be possible with a 50  m  10  m  2  m sensor.
57
Table 3.7: Di erently sized ME biosensors and their theoretical detection limits.
L [ m] W [ m] T [ m] Sm [Hz/pg]  mmin [pg] Nmin Material
4,000 800 15 7.34  10 4 1.36  106 1,362,212 Metglas
2,000 400 15 5.87  10 3 1.70  105 170,277 Metglas
1,000 200 15 4.70  10 2 2.13  104 21,285 Metglas
500 100 15 3.76  10 1 2.66  103 2,661 Metglas
500 250 4 7.65  10 1 1.31  103 1,308 Fe79B21
500 167 4 1.15 8.73  102 874 Fe79B21
500 125 4 1.53 6.54  102 654 Fe79B21
500 100 4 1.91 5.23  102 523 Fe79B21
200 100 4 1.20  101 8.37  101 84 Fe79B21
200 67 4 1.78  101 5.61  101 57 Fe79B21
200 50 4 2.39  101 4.18  101 42 Fe79B21
200 40 4 2.99  101 3.35  101 34 Fe79B21
150 75 4 2.83  101 3.53  101 36 Fe79B21
150 50 4 4.25  101 2.35  101 24 Fe79B21
150 38 4 5.59  101 1.79  101 18 Fe79B21
150 30 4 7.08  101 1.41  101 15 Fe79B21
100 50 4 9.56  101 1.05  101 11 Fe79B21
100 33 4 1.45  102 6.90 7 Fe79B21
100 25 4 1.91  102 5.23 6 Fe79B21
100 20 4 2.39  102 4.18 5 Fe79B21
50 10 2 3.82  103 0.26 1 Fe79B21
58
3.5.2 Measurement of the resonant frequency of the ME biosensors
In order to determine the resonant frequency of the biosensors, they were in-
dividually placed into the center of a copper solenoid coil that is connected to a
network analyzer (HP/Agilent 8751A from Agilent Technologies, Inc.), operated in
the S11 re ection mode (Fig. 3.14). An incident AC signal was, then, applied across
the coil to magnetically excite the longitudinal vibration of the biosensor, and the
resultant re ected signal was compared with the incident signal over a proper range
of frequencies. Finally, the resonant frequency of the biosensor was determined to
be the frequency at which the largest change in normalized jS11j occurs, due to the
magneto-mechanical resonance of the biosensor. To enhance the magnitude of the
resonance peak, a proper bias magnetic  eld was also applied to the biosensor with
a bar magnet. Figure 3.15 shows a response of a typical 500  m  100  m  4  m
ME biosensor in air.
Figure 3.14: Measurement setup.
59
Figure 3.15: Response of a typical 500  m  100  m  4  m ME biosensor in air.
60
Bibliography
[1] V. A. Petrenko, \Landscape phage as a molecular recognition interface for de-
tection devices," Microelectronics Journal, vol. 39, no. 2, pp. 202 { 207, 2008.
[2] S. Goodchild, T. Love, N. Hopkins, and C. Mayers, \Engineering antibodies for
biosensor technologies," vol. 58 of Advances in Applied Microbiology, pp. 185 {
226, Academic Press, 2005.
[3] J. Wan, M. L. Johnson, R. Guntupalli, V. A. Petrenko, and B. A. Chin, \De-
tection of Bacillus anthracis spores in liquid using phage-based magnetoelastic
micro-resonators," Sensors and Actuators B: Chemical, vol. 127, no. 2, pp. 559
{ 566, 2007.
[4] V. A. Petrenko and I. B. Sorokulova, \Detection of biological threats. a challenge
for directed molecular evolution," Journal of Microbiological Methods, vol. 58,
pp. 147 { 168, 2004.
[5] C. Mao, A. Liu, and B. Cao, \Virus-based chemical and biological sensing,"
Angewandte Chemie International Edition, vol. 48, pp. 6790 { 6810, 2009.
[6] J. Wan, H. Shu, S. Huang, B. Fiebor, I.-H. Chen, V. A. Petrenko, and B. A. Chin,
\Phage-based magnetoelastic wireless biosensors for detecting Bacillus anthracis
spores," IEEE Sensors Journal, vol. 7, pp. 470 { 477, 2007.
[7] M. L. Johnson, J. Wan, S. Huang, Z. Cheng, V. A. Petrenko, D.-J. Kim, I.-
H. Chen, J. M. Barbaree, J. W. Hong, and B. A. Chin, \A wireless biosensor
using microfabricated phage-interfaced magnetoelastic particles," Sensors and
Actuators A: Physical, vol. 144, no. 1, pp. 38 { 47, 2008.
[8] R. S. Lakshmanan, R. Guntupalli, J. Hu, D.-J. Kim, V. A. Petrenko, J. M.
Barbaree, and B. A. Chin, \Phage immobilized magnetoelastic sensor for the de-
tection of Salmonella typhimurium," Journal of Microbiological Methods, vol. 71,
no. 1, pp. 55 { 60, 2007.
[9] V. A. Petrenko, G. P. Smith, X. Gong, and T. Quinn, \A library of organic
landscapes on  lamentous phage," Protein Engneering, vol. 9, pp. 797 { 801,
1996.
[10] G. P. Smith and V. A. Petrenko, \Phage display," Chemical Reviews, vol. 97,
pp. 391 { 410, 1997.
61
[11] V. Petrenko, G. Smith, M. Mazooji, and T. Quinn, \ -helically constrained
phage display library," Protein Engineering, vol. 15, no. 11, pp. 943 { 950, 2002.
[12] D. Marvin, \Filamentous phage structure, infection and assembly," Current
Opinion in Structural Biology, vol. 8, no. 2, pp. 150 { 158, 1998.
[13] H. Endemann and P. Model, \Lcoation of  lamentous phage minor coat proteins
in phage and in infected cells," Journal of Molecular Biology, vol. 250, no. 4,
pp. 496 { 506, 1995.
[14] G. Kuzmicheva, P. Jayanna, I. Sorokulova, and V. Petrenko, \Diversity and cen-
soring of landscape phage libraries," Protein Engineering, Design and Selection,
vol. 22, pp. 9 { 18, 2009.
[15] I. B. Sorokulova, E. V. Olsen, I.-H. Chen, B. Fiebor, J. M. Barbaree, V. J.
Vodyanoy, B. A. Chin, and V. A. Petrenko, \Landscape phage probes for
Salmonella typhimurium," Journal of Microbiological Methods, vol. 63, no. 1,
pp. 55 { 72, 2005.
[16] J. Brigati, D. D. Williams, I. B. Sorokulova, V. Nanduri, I.-H. Chen, C. L.
Turnbough, and V. A. Petrenko, \Diagnostic probes for Bacillus anthracis spores
selected from a landscape phage library," Clinical Chemistry, vol. 50, no. 10,
pp. 1899 { 1906, 2004.
[17] E. du Tr emolet de Lacheisserie, Magnetostriction: Theory and Applications of
Magnetoelasticity. CRC Press, 1993.
[18] G. Vlas ak, \Direct measurement of magnetostriction of rapidly quenched thin
ribbons," Journal of Magnetism and Magnetic Materials, vol. 215-216, pp. 479
{ 481, 2000.
[19] A. Hernando, M. Vazquez, and J. M. Barandiaran, \Metallic glasses and sensing
applications," Journal of Physics E: Scienti c Instruments, vol. 21, pp. 1129 {
1139, 1988.
[20] C. Modzelewski, H. T. Savage, L. T. Kabaco , and A. E. Clark, \Magnetome-
chanical coupling and permeability in transversely annealed Metglas 2605 alloys,"
IEEE Transactions on Magnetics, vol. 17, pp. 2837 { 2839, 1981.
[21] R. C. O?Handley, Modern Magnetic Materials: Principles and Applications. New
York: John Wiley & Sons Inc., 2000.
[22] http://www.metglas.com/products/page5 1 2 7.htm.
[23] C. A. Grimes, C. S. Mungle, K. Zeng, M. K. Jain, W. R. Dreschel, M. Paulose,
and K. G. Ong, \Wireless magnetoelastic resonance sensors: A critical review,"
Sensors, vol. 2, pp. 294 { 313, 2002.
62
[24] S. Li, Development of novel acoustic wave biosensor platforms based on magne-
tostriction and fabrication of magnetostrictive nanowires. PhD thesis, Auburn
University, 2007.
[25] L. Davis, R. Ray, C.-P. Chou, and R. O?Handley, \Mechanical and thermal
properties of Fe80B20 glass," Scripta Metallurgica, vol. 10, pp. 541 { 546, 1976.
[26] C. Liang, S. Morshed, and B. C. Prorok, \Correction for longitudinal mode
vibration in thin slender beams," Applied Physics Letters, vol. 90, p. 221912,
2007.
[27] H.-U. K unzi, \Mechanical properties of metallic glasses," in Glassy Metal II,
vol. 53 of Topics in Applied Physics, pp. 169{216, Springer Berlin Heidelberg,
1983.
[28] C. China, \Magnetic properties of sputter deposited Fe based amorphous thin
 lms for resonator application," Master?s thesis, the University of Central
Florida, 2006.
[29] M. A. Mitchell, A. E. Clark, H. T. Savage, and R. J. Abbundi, \ E e ect and
magnetomechanical coupling factor in Fe80B20 and Fe78Si10B12 glassy ribbons,"
IEEE Transactions on Magnetics, vol. 14, pp. 1169 { 1171, 1978.
[30] S. Huang, J. Hu, J. Wan, M. Johnson, H. Shu, and B. Chin, \The e ect of
annealing and gold deposition on the performance of magnetoelastic biosensors,"
Materials Science and Engineering: C, vol. 28, no. 3, pp. 380 { 386, 2008.
[31] http://rs.acupcc.org/site media/uploads/cap/699-cap 1.pdf.
[32] http://www.lesker.com/newweb/deposition materials/depositionmaterials
sputtertargets 1.cfm?pgid=cr1.
[33] http://www.lesker.com/newweb/deposition materials/depositionmaterials
sputtertargets 1.cfm?pgid=au1.
[34] http://www.lesker.com/newweb/deposition materials/depositionmaterials
sputtertargets 1.cfm?pgid=fe1.
[35] http://www.lesker.com/newweb/deposition materials/depositionmaterials
sputtertargets 1.cfm?pgid=bor1.
[36] C. A. Grimes, S. C. Roy, S. Rani, and Q. Cai, \Theory, instrumentation and
applications of magnetoelastic resonance sensors: A review," Sensors, vol. 11,
no. 3, pp. 2809 { 2844, 2011.
[37] http://hypertextbook.com/facts/2003/LouisSiu.shtml.
63
Chapter 4
Direct Detection of S. Typhimurium on Fresh Spinach Leaves
4.1 Introduction
Frequent outbreaks of bacterial food poisoning as well as associated illnesses are
signi cant public health concerns. Hence, to secure the safety of our food supply,
proper, routine testing of food products must be implemented throughout the sup-
ply chain. Conventionally, such testing has been conducted with phenotypic and/or
genotypic assays, including culture- and PCR-based assays [1{10]. Although capable
of providing con rmatory results with low detection limits, these methods are often
complex, expensive, labour-intensive, and thus, not suitable for the on-site detection
of pathogenic bacteria. In addition, rapidness of the testing is among the most impor-
tant requirements particularly for dealing with a large volume of fresh produce and
other essential food items that are consumed daily. To overcome the drawbacks of
the conventional detection methods and facilitate on-site bacterial detection, much re-
search has been recently focused on developing label-free biosensors [11{19]. However,
even for these biosensors, sample preparation, including the collection, puri cation,
and enrichment of a pathogen-containing sample, is generally required prior to the
testing. Hence, there is a motivation for eliminating any pre-test sample preparation
steps to simplify the procedure and further reduce the total time and cost of testing.
Attempts using phage-based ME biosensors have been recently made for the
rapid, direct detection of S. Typhimurium on fresh produce (i.e., tomatoes and shell
eggs [20{22]). These biosensors are composed of a freestanding, strip-shaped ME
signal transducer coated with the E2 phage [23], which speci cally binds with S.
Typhimurium. These biosensors can be directly placed on produce surfaces due to
64
their wireless, freestanding nature and used to monitor the presence of the bacterium
without pre-test sample preparation. In this investigation, the methodology was
employed to test Salmonella-spiked spinach leaves.
Figures 4.1a to 4.1c show scanning electron micrographs of the surfaces of various
produce: (a) a tomato, (b) a shell egg, and (c) a spinach leaf. As can be seen, surface
topography varies from one produce to another. In addition, since spinach leaves are
likely to possess complex surface topography as shown in Fig. 4.1d (a close-up view
of a leaf surface), these surface features may a ect the physical contact between the
Figure 4.1: Scanning electron micrographs of various produce surfaces: (a) tomato,
(b) eggshell, and (c) spinach leaf. A close-up view of a spinach leaf spiked with S.
Typhimurium is shown in (d).
65
biosensors and S. Typhimurium cells. Hence, three di erent sizes of biosensors (2-
mm, 500- m, and 150- m long) were fabricated and tested to investigate the e ects
of sensor size on the limit of detection (LOD). Furthermore, a formula describing the
probability of detection as a function of the size and number of biosensors and the
surface density of S. Typhimurium was derived. By using the formula, the required
number of biosensors to obtain a desired LOD can be determined.
4.2 Material and methods
4.2.1 E2 phage and S. Typhimurium
Suspensions of the E2 phage (5  1011 virions/ml in a TBS bu er) and S. Ty-
phimurium cells (ATCC 13311 at a concentration of 5  108 cells/ml in sterile dis-
tilled water) were provided by Dr. James Barbaree?s group at Auburn University.
The concentrated Salmonella suspension was diluted with sterile distilled water as
desired prior to use.
4.2.2 Confocal re ectance imaging of spinach leaf surfaces
Both adaxial and abaxial surfaces of fresh spinach leaves were imaged with a
confocal scanning laser microscope (Nikon A1 from Nikon Corp.). The leaves were
cut into 10 mm 10 mm pieces without including any major leaf veins and mounted
on a glass slide with double-sided tape. These samples were prepared right before the
imaging and individually surface-scanned with a 488-nm line laser at 21  C and 37%
relative humidity. To minimize the degradation of the samples, a low laser power of
1.5% was used exclusively, and the imaging per sample was completed within 30 min.
Re ectance from the sample surfaces was collected at magni cations of 40 and 400
with a bandpass  lter of 482/35 nm. The collected digital data were saved as images
with a pixel resolution of 1,024 1,024. The unit pixel length was 3.11  m and 0.311
 m for images taken at 40 and 400 magni cations, respectively.
66
To reconstruct and characterize the topography of the sample surfaces, a series
of plane images through the thickness of spinach leaf surfaces was captured. The
separation between adjacent slices was 0.1  m. Fifteen samples were prepared for
both adaxial and abaxial surfaces and imaged at the above-mentioned magni cations.
The ImageJ software with the SurfCharJ plugin [24] was, then, used to reconstruct
the surface topography and extract surface height data associated with their location
of pixels. Finally, surface pro les were regenerated and tilt-corrected by subtracting
an overall increasing or decreasing linear trend along the sampling length.
4.2.3 Fabrication of ME sensor platforms with three di erent sizes
Freestanding, strip-shaped ME sensor platforms with three di erent sizes were
fabricated of either Metglas 2826MB or Fe79B21, both of which are amorphous, fer-
romagnetic alloys with ME properties [25,26]. Millimeter-long sensor platforms with
two di erent sizes (2 mm  0.4 mm  15  m and 0.5 mm  0.1 mm  15  m)
were manufactured by polishing and dicing a sheet of Metglas 2826MB. By contrast,
Figure 4.2: Di erently sized sensor platforms used (top view).
67
micrometer-long sensor platforms with a size of 150  m 30  m 4  m were fabri-
cated of Fe79B21 using the co-sputtering-based method described in Chapter 3. Figure
4.2 illustrates the three di erently sized sensor platforms used in this investigation.
4.2.4 Fabrication of phage-based ME biosensors
The fabrication of phage-based ME biosensors was completed by following the
procedures described in Chapter 3. In addition to measurement sensors, control
sensors, which are not immobilized with the E2 phage but only surface-blocked with
BSA, were also prepared and used for background subtraction.
4.2.5 Determination of the concentration of BSA for surface blocking
In order to determine a reasonable concentration of BSA for surface blocking,
measurement and control sensors (2-mm long) exposed to three di erent concentra-
tions of BSA (0.01, 0.1, and 1 % w/v) were prepared and tested. These sensors
were placed on a wet spinach leaf surface inoculated with S. Typhimurium (5  108
cells/ml, 40  L) and, then, allowed for binding. Figure 4.3a shows responses for both
measurement and control sensors. As can be seen, the resonant frequency changes
for both types of sensors decreased with increased concentrations of BSA. Hence, in
terms of surface blocking, a high BSA concentration is preferable. However, the re-
sponse of control sensors must be minimized without unnecessarily reducing that of
measurement sensors. Hence, it is important to  nd a BSA concentration at which
the di erence between the responses of measurement and control sensors is statisti-
cally maximized. For the three pairs of data in Fig. 4.3a, the con dence level of
di erence was calculated with a standard t-test [27,28]. The best result (i.e., highest
con dence level of di erence) was obtained at a BSA concentration of 0.1 % w/v as
shown in Fig. 4.3b. This concentration was, hence, used for surface blocking.
68
Figure 4.3: E ects of BSA concentration on (a) resonant frequency changes for mea-
surement and control sensors (2-mm long) and on (b) the con dence level of di erence.
4.2.6 Direct detection of S. Typhimurium on fresh spinach leaves
Pre-washed, bagged, fresh baby spinach leaves (Kroger brand) were purchased
from a local grocery store and used as-received. They were,  rst, individually ad-
hered to a clean,  at surface with double-sided tape. Forty-microliter drops of S.
Typhimurium with various concentrations (i.e., ten-fold serial dilutions of 5  108
cells/ml) were, then, spot-inoculated on the leaf surface as illustrated in Fig. 4.4a.
Any major leaf veins, which cause a sudden, large change in surface topography, were
avoided. Yet, the total surface area of the major leaf veins was found to be small (7.3
69
Figure 4.4: Schematic illustration of the test procedure: (a) spot-inoculation of S.
Typhimurium on the leaf surface and measurement of the initial resonant frequency of
biosensors, (b) placement of both measurement and control sensors on the Salmonella-
inoculated sites (after drying the Salmonella drops and misting the leaf surface),
(c) measurement of the  nal resonant frequency of the biosensors, and (d) typical
responses of the biosensors.
 2.9 % of a leaf). The inoculated Salmonella drops were, then, allowed to dry in
air for 90 to 120 min. Next, the leaf surface was uniformly misted by spraying ster-
ile distilled water ( 20  l/cm2), and both measurement and control sensors with a
pre-determined resonant frequency were directly placed on the Salmonella-inoculated
sites (Fig. 4.4b). After 25 min to allow for binding, the biosensors were collected with
a magnet, and measurement of their  nal resonant frequency was completed within
20 min (Figs. 4.4c and 4.4d) (The method for resonant frequency measurement was
described in Chapter 3). The total test time was, hence, roughly 45 min. In this
investigation, the test was performed at 23  C and 35% relative humidity. The three
70
di erently sized biosensors (i.e., 2 mm-, 0.5 mm-, and 150  m-long sensors) were
used to test both adaxial and abaxial surfaces of the leaves. Ten measurement and
control sensors each were used for each concentration of S. Typhimurium. In order
to convert cells/ml into cells/cm2 (i.e., surface density of S. Typhimurium), the area
of the inoculation sites was measured and found to be 0.22  0.02 cm2 and 0.30  
0.03 cm2 for the adaxial and abaxial surfaces, respectively.
4.3 Results
4.3.1 Observation of Salmonella-inoculated leaf surfaces
Figure 4.5: Scanning electron micrographs of a spinach leaf surface inoculated with a
40- l drop of S. Typhimurium with various concentrations: (a) 5 108 cells/ml, (b)
5 107 cells/ml, (c) 5 106 cells/ml (with a 150  m-long ME biosensor), and (d) 0
cells/ml (reference).
71
Figure 4.5 shows representative scanning electron micrographs of a spinach leaf
surface inoculated with a 40- l drop of S. Typhimurium with various concentrations.
At a concentration of 5 108 cells/ml, the leaf surface was nearly completely covered
by S. Typhimurium cells (Fig. 4.5a). The number of observable cells, then, decreased
with decreased concentrations of inoculated cells as anticipated, and the distribution
of cells became non-uniform for lower concentrations (Figs. 4.5b and 4.5c). This
localization of cells may be attributed to localized availability of nutrients as well as
leaf surface conditions (e.g., hydrophobicity and surface charges), which a ect the
motility and attachment of the cells [29,30]. Furthermore, the leaf surface was found
to possess complex topography (Fig. 4.5d), which plays a crucial role in physical
contact between a biosensor and S. Typhimurium cells.
4.3.2 Resonant frequency measurement
Figure 4.6 shows a response of a typical 150  m 30  m 4  m sensor, which
was measured in air with the setup described in Chapter 3. It can be seen that the
raw data set (dim gray curve in Fig. 4.6a) contains a high degree of noise even after a
10-time averaging operation. This noisy raw curve is common particularly for small
sensors with small peak amplitudes, which requires additional smoothing of the data
set. Hence, to further reduce the noise level, the Savitzky-Golay smoothing (25 points,
second-order) was performed (black curve in Fig. 4.6a). In addition, Lorentzian  tting
of the smoothed curve was  nally performed to determine the resonant frequency of
the sensor as shown in Fig. 4.6b. This method of data arrangement was also used for
larger sensors. The mean resonant frequencies for the three di erently sized biosensors
are summarized in Table 4.1.
72
Figure 4.6: Response of a typical 150  m  30  m  4  m sensor in air: (a) raw
data set (10-time averaged) and its smoothed curve and (b) Lorentizan  tting of the
smoothed curve.
73
Table 4.1: Mean resonant frequencies for the di erently sized biosensors.
Sensor size Mean resonant frequency
2 mm  0.4 mm  15  m 1.12  0.04 MHz
0.5 mm  0.1 mm  15  m 4.41  0.04 MHz
150  m  30  m  4  m 13.06  0.12 MHz
To determine a change in the resonant frequency of a sensor, both the initial
and  nal resonant frequencies (finitial and f nal) need to be measured. An example
is shown in Fig. 4.7, where the resonant peaks for a 150- m long sensor before and
after placing on a leaf surface inoculated with S. Typhimurium (5 108 cells/ml) are
shown. It can be seen that the  nal peak appears at a lower frequency due to the
attachment of S. Typhimurium cells.
Figure 4.7: Resonant peaks for a 150- m long sensor before and after placing on a
leaf surface inoculated with S. Typhimurium at a concentration of 5  108 cells/ml.
4.3.3 Dose-response of the ME biosensors
Dose-response relationships for the di erently sized biosensors are shown in Fig.
4.8. The plots on the left and right are the results for the adaxial and abaxial surfaces
74
Figure 4.8: Dose-response plots for the di erently sized biosensors (2 mm-, 0.5 mm-
and, 150  m-long sensors). The plots on the left and right are the results for the
adaxial and abaxial surfaces, respectively.
of spinach leaves, respectively. Resonant frequency changes of measurement sensors
(circles) were found to be largely dependent on the surface density of S. Typhimurium.
By contrast, control sensors (squares) showed much smaller responses, indicating
that selective binding of S. Typhimurium on the measurement sensors occurred. In
75
addition, the standard error was found to be large at high surface densities of S.
Typhimurium, which may be attributed to (1) the complex topography of leaf surfaces
and (2) the random locations of the sensors, resulting in non-uniform, inconsistent
physical contact between the sensors and S. Typhimurium cells.
4.3.4 Determination of the LOD
The LODs for the above dose-response plots were determined as follows:
1. The responses of the control sensors were subtracted from those of the measure-
ment sensors (i.e., background subtraction).
Figure 4.9: (a) Sigmoidal curve and (b) the determination of the LOD.
76
2. The resultant data were curve- tted with sigmoidal functions, which are com-
monly used for the description of the response patterns of bioassays [31]. On a
sigmoidal curve, there are usually two concentration-independent regions (i.e.,
lower and upper plateaus) and one linear concentration-dependent region de-
 ned with two bend points as shown in Fig. 4.9a.
3. Finally, the LOD was determined as the concentration at which the  tted re-
sponse deviates from the average response in the lower plateau region,  fAVE,
by a multiple of the standard error,  , in the region [32]. The value of the
multiple is dependent on a required statistical signi cance level. In this work, a
multiple of three was used (i.e., 3 ) as shown in Fig. 4.9b. Note that the lower
plateau region was determined through a linear regression analysis with an R2
value of greater than 0.95.
Figure 4.10 shows background-subtracted data for the dose-response plots shown
in Fig. 4.8. These data were  tted with sigmoidal functions (red solid curves), which
can be de ned as [31]
Y = a b1 + (X=c)d +b; (4.1)
where Y, X, a, b, c, and d represent the response, concentration (i.e., surface density
of S. Typhimurium), lower asymptote, upper asymptote, in ection point, and slope
factor, respectively. The R2 values were found to be all close to the unity, indicating
that the curve  tting was well performed. To determine the LOD, the  fAVE + 3 
values were extrapolated to the  tted curves as shown in Fig. 4.10 (blue dashed lines).
Table 4.2 summarizes the LODs of the di erently sized biosensors for both adaxial
and abaxial surfaces of spinach leaves.
77
Figure 4.10: Background-subtracted data for the dose-response plots in Fig. 4.8.
These data were  tted with sigmoidal functions (red solid curves). The R2 values
were all close to one. The values of  fAVE + 3 are shown in blue text.
As can be seen in Table 4.2, the LOD was found to be on the order of 104
to 105 cells/cm2. The best results were obtained with the 150- m long sensors al-
though, among the di erently sized biosensors, these micron-scale biosensors possess
the smallest surface area (i.e., the product of the length and width of the biosensor,
LW) to cover a Salmonella-inoculated leaf surface. In addition, the LODs for the
larger biosensors were found to be higher by half an order to one order of magnitude,
78
and there were not large di erences among them. One possible explanation for these
results is that there is a trade-o between the surface area and mass sensitivity of
a biosensor as can be seen in Eq. 3.6. In other words, while the surface area of a
biosensor increases proportionally to LW, the mass sensitivity decreases proportion-
ally to L2WT. Hence, the use of a large biosensor, which covers a large area of a
leaf surface, may not always lead to sensitive detection of S. Typhimurium. Rather,
a properly small biosensor may be able to show a measurable response even though
a small number of S. Typhimurium cells may be bound on this biosensor. In addi-
tion, from the preceding microscopic observation, the topography of leaf surfaces and
distribution of S. Typhimurium cells are both anticipated to a ect the LOD.
Table 4.2: LODs of the di erently sized biosensors for the adaxial and abaxial surfaces
of spinach leaves.
LOD [cells/cm2]
Sensor size Adaxial surface Abaxial surface
2 mm  0.4 mm  15  m 1.28  105 4.12  105
0.5 mm  0.1 mm  15  m 1.42  105 1.37  105
150  m  30  m  4  m 4.77  104 5.22  104
4.4 Discussion
4.4.1 Topography of leaf surfaces and its e ects on the LOD
In order to characterize the topography of the leaf surfaces, mean roughness (Ra),
mean  atness (Fa), and associated major periodicities (TR and TF) were quanti ed.
Mean roughness, Ra, is a commonly used roughness parameter and can be de ned
as the arithmetic average of the absolute values of height deviations measured from
the mean plane of a surface [33, 34]. In addition, the same de nition can be given
to mean  atness, Fa, which was, however, measured with a longer sampling length
(i.e., The sampling lengths for Ra and Fa measurements were 318.5  m and 3185
79
Figure 4.11: Typical height maps (a & b) and associated averaged pro les (c &
d) of a leaf surface obtained along di erent sampling lengths. A three-dimensional
representation of a leaf surface expressed by Eq. 4.2 is shown in (e).
 m, respectively.). Figures 4.11a to 4.11d show typical height maps and associated
averaged pro les of a leaf surface obtained along di erent sampling lengths. From
the pro le with a short sampling length (Fig. 4.11c), it can be seen that the leaf
surface possesses a certain degree of roughness, Ra. By contrast, the pro le with a
long sampling length (Fig. 4.11d) provides insight into the  atness of the surface,
Fa. Furthermore, for both surface pro les, the major periodicities, TR and TF, can
be determined by performing a Fourier transform of the pro le ordinates and taking
the reciprocal of the main frequency mode. The experimentally determined values for
these surface geometric parameters are summarized in Table 4.3. The values are all
non-zeros, indicating that leaf surfaces are rough, non- at surfaces with topographic
periodicities. Hence, with all the geometric parameters, the height at an arbitrary
location of a leaf surface, H (x, y), can be expressed by the superposition of sine
80
Table 4.3: Surface geometric parameters for the adaxial and abaxial surfaces of
spinach leaves. The values are the averages of 15 samples.
Parameter Adaxial surface Abaxial surface
Ra ( m) 1.2  0.2 1.6  0.3
Fa ( m) 8.7  3.7 7.0  1.9
TR ( m) 56.3  18.4 67.4  27.0
TF ( m) 1376.7  309.0 874.7  196.1
waves (i.e., when Fa << TF) as
H(x;y) = 12[Ra sin( 2 T
R
x) +Fa sin( 2 T
F
x)
+Ra sin( 2 T
R
y) +Fa sin( 2 T
F
y)]; (4.2)
where x and y are the lateral coordinates of the leaf surface, respectively. The  rst and
third terms on the right-hand side of Eq. 4.2 describe roughness pro les, whereas the
second and forth terms represent  atness pro les along the lateral dimensions of the
leaf surface, respectively. A three-dimensional representation of this mathematically
expressed surface is shown in Fig. 4.11e, where a periodic arrangement of peaks and
valleys can be seen. The periodicities, TR and TF, represent the lateral distances
between two adjacent peaks of di erent scale, respectively (i.e., roughness-related
small peaks and  atness-related large peaks).
In order to investigate e ects of a leaf?s surface topography on the physical con-
tact between a biosensor and S. Typhimurium cells, the surface area of the biosensor,
LW, was compared with two characteristic areas of the leaf surface, AR and AF,
which are de ned by
AR = TR TR; (4.3)
AF = TF TF; (4.4)
AR <<AF (*TR <<TF): (4.5)
81
In other words, AR and AF are the areas surrounded by four adjacent small and large
peaks, respectively. When LW >>AF, the biosensor is likely to stay on large peaks,
which greatly reduces the degree of physical contact with the leaf surface (and thus,
with S. Typhimurium cells). By contrast, when AF >> LW >> AR, the biosen-
sor can  t among large peaks (but stays on small peaks), which may improve the
physical contact. Furthermore, when LW << AR, the e ects of the leaf?s surface
topography can be neglected, and thus, the degree of physical contact can be maxi-
mized. In this case, all that matters is the distribution of S. Typhimurium cells on
the leaf surface. Hence, the characteristic area, AR, and its associated length, TR,
are important parameters that allow one to choose a properly small biosensor free
from the surface topographic e ects. As shown in Table 4.3, the values of TR for the
adaxial and abaxial surfaces of spinach leaves were 56.3  18.4  m and 67.4  27.0
 m, respectively. Among the di erently sized biosensors used in this investigation,
the 150- m long sensors possess the closest lateral dimensions to these characteristic
lengths, indicating that the topographic e ects for these micron-scale biosensors were
the smallest. By contrast, the millimeter-scale biosensors (i.e., 2-mm and 0.5-mm
long sensors) possess lateral dimensions much larger than TR and, rather, close to
TF. Hence, it is understandable that much larger topographic e ects were posed to
these larger biosensors. Furthermore, for real leaf surfaces, there exists microscopic
irregularity in topography as can be seen in Figs. 4.11a to 4.11d, which is likely to
cause reduced degrees of physical contact particularly for large biosensors. Hence,
with the aforementioned merit of high mass sensitivity, the use of su ciently small
biosensors (i.e., LW <<AR) in proper quantity (number of sensors) may be the key
to improving the LOD.
82
4.4.2 E ects of the number of biosensors on the LOD
Now that a properly small size of biosensor can be determined from the previous
section, a remaining key question is how many biosensors of such a size are actually
needed to detect S. Typhimurium cells on a leaf surface? To answer this question,
a simple model that describes the probability of detection will be presented. The
probability of detection is dependent on various factors, including the size, num-
ber, and mass sensitivity of the biosensor, the surface density and distribution of S.
Typhimurium cells, and the topography of the leaf surface. Although spinach leaf
surfaces were found to be rough, non- at surfaces, they can be treated as smooth,
 at surfaces if the size of the biosensor is su ciently small (i.e., LW <<AR). In this
case, the probability of detection, P(D), can be simpli ed to
P(D) = P(ms)P(pc); (4.6)
where P(ms) is a probability function related to the mass sensitivity of the biosen-
sor, and P(pc) is the probability of physical contact between the biosensor and S.
Typhimurium cells. P(ms) is given as
P(ms) =
8
>>>
><
>>>
>:
1 (N Nmin)
0 (N <Nmin)
; (4.7)
where N is the number of S. Typhimurium cells bound on a single biosensor, and
Nmin is the minimum detectable number of S. Typhimurium cells for this biosensor.
Nmin can be computed using Eq. 3.6 with the minimum detectable response,  fAVE
+ 3 , de ned in Fig. 4.9 and the mass of a single S. Typhimurium cell, mST. By
83
rearranging Eq. 3.6 and replacing  f with  fAVE + 3 , Nmin can be de ned as
Nmin =  mm
ST
 4(j fAVE + 3 j)L
2WT
mST
r
 3(1  )
E : (4.8)
Figure 4.12: Finite well model of a Salmonella-inoculated leaf surface. The total
surface area is x2LW: There are three types of wells: sensor-containing wells, cell-
containing wells, and empty wells.
Now, let?s consider a simple  nite well model of a Salmonella-inoculated leaf
surface with a total surface area of x2LW as shown in Fig. 4.12. Let m, n, and x2 be
the number of sensor-containing wells, the number of cell-containing wells, and the
total number of wells, respectively. For this model, the following assumptions were
made:
1. The area of each well is equal to that of a sensor (i.e., LW). Hence, one well can
contain one sensor at most. In other words, the number of sensor-containing
wells, m, is equal to the number of sensors placed on the surface.
2. The orientation of a sensor and a well matches to each other.
3. There is no overlapping among sensors.
84
4. Cell-containing wells are randomly located.
5. In each cell-containing well, there are N cells, which are uniformly distributed.
When a sensor is placed, all the N cells are attached to the sensor uniformly.
In order to calculate P(pc),  rst, the multiplication law of probability was used
to derive the complement probability, P(pc). When n = 1, there is only one cell-
containing well among x2 wells as can be seen in Fig. 4.13. Hence, the probability
that the  rst sensor is not placed in the cell-containing well, P(pc1), is
P(pc1) = x2 1C1
x2C1
= x
2 1
x2 : (4.9)
Likewise, the probability that the second sensor is also not placed in the cell-containing
well, P(pc2), is
P(pc2) = x2 2C1
x2 1C1
= x
2 2
x2 1; (4.10)
and the probability that the ith sensor is also not placed in the cell-containing well,
P(pci), is
P(pci) = x2 iC1
x2 i+1C1
= x
2 i
x2 i+ 1: (4.11)
Figure 4.13: Finite well model with n = 1 (i.e., one cell-containing well).
85
Finally, the probability that none of m sensors is placed in the cell-containing well,
P(pc)n=1, can be given as
P(pc)n=1 = P(pc1) P(pc2)     P(pcm)
=
mY
i
( x
2 i
x2 i+ 1): (4.12)
Similarly, when n = 2 and n = 3, the compliment probabilities can be calculated
respectively as
P(pc)n=2 =
mY
i
(x
2 i 1
x2 i+ 1) (4.13)
and
P(pc)n=3 =
mY
i
(x
2 i 2
x2 i+ 1): (4.14)
By analogy, when there are n cell-containing wells among x2 wells, the probability
that none of m sensors is placed in any of the n cell-containing wells, P(pc), can be
expressed as
P(pc) =
mY
i
(x
2 i n+ 1
x2 i+ 1 ): (4.15)
Finally, P(pc) can be obtained by subtracting P(pc) from the unity as
P(pc) = 1 P(pc)
= 1 
mY
i
(x
2 i n+ 1
x2 i+ 1 ): (4.16)
By substituting Eq. 4.16 into Eq. 4.6, the probability of detection, P(D), is, then,
given as
P(D) = P(ms)[1 
mY
i
(x
2 i n+ 1
x2 i+ 1 )]; (4.17)
which is a function of P(ms), m, n, and x2. Figure 4.14, for example, shows the
probability of detection, P(D), with respect to the number of biosensors, m, and the
86
Figure 4.14: Probability of detection with respect to the number of biosensors, m,
and the number of cell-containing wells, n. Biosensors with lateral dimensions of 50
 m  10  m were placed on a leaf surface of 0.26 cm2.
number of cell-containing wells, n, for 50  m  10  m biosensors (i.e., LW <<AR)
placed on a leaf surface of 0.26 cm2, provided that N  Nmin (i.e., P(ms) = 1).
Note that the value of x2 was calculated by 0.26 cm2/(LW). As can be seen, P(D)
increases with increased numbers of m and n.
Now, theoretical LODs with desired threshold values of P(D) can be determined
for a su ciently small biosensor (i.e., LW <<AR). From the assumption #5, the to-
tal number of S. Typhimurium cells, Ntotal, on the Salmonella-inoculated leaf surface
can be calculated by
Ntotal = nN: (4.18)
Hence, when N = Nmin, the LOD can be calculated in terms of the surface density
of S. Typhimurium by
LOD = Ntotaltotal surface area = Ntotalx2LW = nNx2LW = nNminx2LW: (4.19)
87
When Nmin = 1, Eq. 4.19 can be further simpli ed to
LOD = nx2LW; (4.20)
which is the case for biosensors as small as 50  m  10  m  2  m, provided that
 fAVE+3 of 1,000 Hz can be resolved (See Table 3.7.). In addition, from Eq. 4.17,
the number of cell-containing wells, n, becomes merely a function of the number of
biosensors, m, when the values of P(D) and x2 are given. Hence, for a certain size
of biosensor (i.e., for known values of L and W with LW <<AR), the LOD can be
 nally expressed as a function of m by
LOD = nx2LW = nconstant = f(m): (4.21)
Table 4.4, for example, shows the values of m for 50  m  10  m  2  m
biosensors to obtain desired LODs for various values of P(D) (0.1, 0.5, and 0.8). It
can be seen that a higher number of biosensors is needed to obtain a lower LOD. In
addition, the values of m were found to be largely dependent on the values of P(D),
which can also be seen in Fig. 4.15. Hence, by using the derived Eqs. 4.17 and 4.20,
one can estimate the LOD with a threshold probability of detection for a certain size
and number of biosensors.
88
Table 4.4: Required number of biosensors, m, to obtain desired LODs for various
values of P(D) (0.1, 0.5, and 0.8).
Dimensions [ m] Number of wells
L W T x2 n P(D) m LOD (cells/cm2)
50 10 2 52,000 1 0.1 5,200 3.85  100
50 10 2 52,000 10 0.1 545 3.85  101
50 10 2 52,000 100 0.1 55 3.85  102
50 10 2 52,000 1,000 0.1 6 3.85  103
50 10 2 52,000 1,0000 0.1 1 3.85  104
50 10 2 52,000 1 0.5 26,000 3.85  100
50 10 2 52,000 10 0.5 3,482 3.85  101
50 10 2 52,000 100 0.5 359 3.85  102
50 10 2 52,000 1,000 0.5 36 3.85  103
50 10 2 52,000 1,0000 0.5 4 3.85  104
50 10 2 52,000 1 0.8 41,600 3.85  100
50 10 2 52,000 10 0.8 7,730 3.85  101
50 10 2 52,000 100 0.8 830 3.85  102
50 10 2 52,000 1,000 0.8 83 3.85  103
50 10 2 52,000 1,0000 0.8 8 3.85  104
89
Figure 4.15: Dependence of the LOD on the number of biosensors (50  m  10  m
 2  m), m, for various values of P(D).
4.5 Conclusions
By placing freestanding, strip-shaped ME biosensors on leaf surfaces, rapid de-
tection of S. Typhimurium (< 45 min) without any pre-test sample preparation was
realized. The LOD was found to be on the order of 104 to 105 cells/cm2 for the three
di erently sized biosensors (2-mm, 0.5-mm, and 150- m long sensors). Although the
best results were obtained with the 150- m long sensors (4.77  104 cells/cm2 and
5.22  104 cells/cm2 for the adaxial and abaxial surfaces of leaves), the LOD should
be dependent not only on the size and mass sensitivity of biosensors, but also on
the number of biosensors, the surface distribution of S. Typhimurium cells, and the
topography of leaf surfaces.
In order to investigate the e ects of the topography of leaf surfaces on the LOD,
surface geometric parameters of mean roughness (Ra), mean  atness (Fa), and asso-
ciated major periodicities (TR and TF) were quanti ed. Then, the characteristic sizes
of biosensors free from the surface topography e ects were determined.
90
Finally, a formula describing the probability of detection as a function of the size
and number of biosensors and the surface density of S. Typhimurium was derived.
By using the formula, the required number of biosensors to obtain a desired LOD can
be determined.
91
Bibliography
[1] B. Lungu, W. D. Waltman, R. D. Berghaus, and C. L. Hofacre, \Comparison of
a real-time PCR method with a culture method for the detection of Salmonella
enterica serotype Enteritidis in naturally contaminated environmental samples
from integrated poultry houses," Journal of food protection, vol. 75, no. 4, pp. 743
{ 747, 2012.
[2] S. Artault, J. L. Blind, J. Delaval, Y. Dureuil, and N. Gaillard, \Detecting
Listeria monocytogenes in food," International Food Hygiene, vol. 12, p. 23,
2001.
[3] B. Brooks, J. Devenish, C. Lutze-Wallace, D. Milnes, R. Robertson, and
G. Berlie-Surujballi, \Evaluation of a monoclonal antibody-based enzyme-linked
immunosorbent assay for detection of Campylobacter fetus in bovine preputial
washing and vaginal mucus samples," Veterinary Microbiology, vol. 103, no. 12,
pp. 77 { 84, 2004.
[4] E. L. Crowley, C. K. O?Sullivan, and G. G. Guilbault, \Increasing the sensitivity
of Listeria monocytogenes assays: Evaluation using ELISA and amperometric
detection," Analyst, vol. 124, pp. 295 { 299, 1999.
[5] L. P. Mans eld and S. Forsythe, \The detection of Salmonella serovars from
animal feed and raw chicken using a combined immunomagnetic separation and
ELISA method," Food Microbiology, vol. 18, no. 4, pp. 361 { 366, 2001.
[6] B. W. Blais, J. Leggate, J. Bosley, and A. Martinez-Perez, \Comparison of  uo-
rogenic and chromogenic assay systems in the detection of Escherichia coli O157
by a novel polymyxin-based ELISA," Letters in Applied Microbiology, vol. 39,
no. 6, pp. 516 { 522, 2004.
[7] P. G. Klein and V. K. Juneja, \Sensitive detection of viable Listeria monocyto-
genes by reverse transcription-PCR.," Applied and Environmental Microbiology,
vol. 63, no. 11, pp. 4441 { 4448, 1997.
[8] P. Villari, E. Motti, C. Farullo, and I. Torre, \Comparison of conventional culture
and PCR methods for the detection of Legionella pneumophila in water," Letters
in Applied Microbiology, vol. 27, no. 2, pp. 106 { 110, 1998.
[9] J. A. Higgins, S. Nasarabadi, J. S. Karns, D. R. Shelton, M. Cooper, A. Gbakima,
and R. P. Koopman, \A handheld real time thermal cycler for bacterial pathogen
detection," Biosensors and Bioelectronics, vol. 18, no. 9, pp. 1115 { 1123, 2003.
92
[10] M. Lund, S. Nordentoft, K. Pedersen, and M. Madsen, \Detection of Campy-
lobacter spp. in chicken fecal samples by real-time PCR," Journal of Clinical
Microbiology, vol. 42, no. 11, pp. 5125 { 5132, 2004.
[11] G. C. Bokken, R. J. Corbee, F. van Knapen, and A. A. Bergwer , \Immunochem-
ical detection of Salmonella group B, D and E using an optical surface plasmon
resonance biosensor," FEMS Microbiology Letters, vol. 222, no. 1, pp. 75 { 82,
2003.
[12] B.-K. Oh, Y.-K. Kim, K. W. Park, W. H. Lee, and J.-W. Choi, \Surface plasmon
resonance immunosensor for the detection of Salmonella typhimurium," Biosen-
sors and Bioelectronics, vol. 19, no. 11, pp. 1497 { 1504, 2004.
[13] H. J. Watts, C. R. Lowe, and D. V. Pollard-Knight, \Optical biosensor for mon-
itoring microbial cells," Analytical Chemistry, vol. 66, no. 15, pp. 2465 { 2470,
1994.
[14] B. H. Schneider, J. G. Edwards, and N. F. Hartman, \Hartman interferometer:
versatile integrated optic sensor for label-free, real-time quanti cation of nucleic
acids, proteins, and pathogens," Clinical Chemistry, vol. 43, no. 9, pp. 1757 {
1763, 1997.
[15] J. Lehtinen, S. J arvinen, M. Virta, and E.-M. Lilius, \Real-time monitoring
of antimicrobial activity with the multiparameter microplate assay," Journal of
Microbiological Methods, vol. 66, no. 3, pp. 381 { 389, 2006.
[16] C. Ruan, H. Wang, and Y. Li, \A bienzyme electrochemical biosensor coupled
with immunomagnetic separation for rapid detection of Escherichia coli O157:H7
in food samples," Transactions of the ASABE, vol. 45, no. 1, pp. 249 { 255, 2002.
[17] L. Yang, C. Ruan, and Y. LI, \Rapid detection of Salmonella typhimurium in
food samples using a bienzyme electrochemical biosensor with  ow injection,"
Journal of Rapid Methods & Automation in Microbiology, vol. 9, no. 4, pp. 229
{ 240, 2001.
[18] L. Yang, Y. Li, C. L. Gri s, and M. G. Johnson, \Interdigitated microelectrode
(IME) impedance sensor for the detection of viable Salmonella typhimurium,"
Biosensors and Bioelectronics, vol. 19, no. 10, pp. 1139 { 1147, 2004.
[19] D. W. Branch and S. M. Brozik, \Low-level detection of a Bacillus anthracis
simulant using Love-wave biosensors on 36 YX LiTaO3," Biosensors and Bio-
electronics, vol. 19, no. 8, pp. 849 { 859, 2004.
[20] M.-K. Park, H. C. W. III, Y. Chai, S. Horikawa, W. Shen, and B. A. Chin, \The
e ect of incubation time for Salmonella Typhimurium binding to phage-based
magnetoelastic biosensors," Food Control, vol. 26, no. 2, pp. 539 { 545, 2012.
93
[21] S. Li and Z.-Y. Cheng, \Nonuniform mass detection using magnetostrictive
biosensors operating under multiple harmonic resonance modes," Journal of Ap-
plied Physics, vol. 107, p. 114514, 2010.
[22] Y. Chai, S. Li, S. Horikawa, M.-K. Park, V. Vodyanoy, and B. A. Chin, \Rapid
and sensitive detection of Salmonella Typhimurium on eggshells by using wireless
biosensors," Journal of Food Protection, vol. 75, no. 4, pp. 631 { 636, 2012.
[23] I. B. Sorokulova, E. V. Olsen, I.-H. Chen, B. Fiebor, J. M. Barbaree, V. J.
Vodyanoy, B. A. Chin, and V. A. Petrenko, \Landscape phage probes for
Salmonella typhimurium," Journal of Microbiological Methods, vol. 63, no. 1,
pp. 55 { 72, 2005.
[24] G. Chinga, P. O. Johnssen, R. Dougherty, E. L. Berli, and J. Walter, \Quanti ca-
tion of the 3-D micro-structure of SC surfaces," Journal of Microscopy, vol. 227,
pp. 254 { 265, 2007.
[25] M. L. Johnson, J. Wan, S. Huang, Z. Cheng, V. A. Petrenko, D.-J. Kim, I.-
H. Chen, J. M. Barbaree, J. W. Hong, and B. A. Chin, \A wireless biosensor
using microfabricated phage-interfaced magnetoelastic particles," Sensors and
Actuators A: Physical, vol. 144, no. 1, pp. 38 { 47, 2008.
[26] S. Li, S. Horikawa, M. kyung Park, Y. Chai, V. J. Vodyanoy, and B. A. Chin,
\Amorphous metallic glass biosensors," Intermetallics, vol. 30, no. 0, pp. 80 {
85, 2012.
[27] http://o ce.microsoft.com/en-us/excel-help/ttest-HP005209325.aspx.
[28] projectile.sv.cmu.edu/research/public/talks/t-test.htmn#types.
[29] R. M. Harshey, \Bacterial motility on a surface: Many ways to a common goal,"
Annual Review of Microbiology, vol. 57, pp. 249 { 273, 2003.
[30] B. Li and B. E. Logan, \Bacterial adhesion to glass and metal-oxide surfaces,"
Colloids and Surfaces B: Biointerfaces, vol. 36, pp. 81 { 90, 2004.
[31] J. L. Sebaugh and P. D. McCray, \De ning the linear portion of a sigmoid-shaped
curve: bend points," Pharmaceutical Statistics, vol. 2, no. 3, pp. 167 { 174, 2003.
[32] R. P. Buck and E. Lindner, \Recomendations for nomenclature of ion-selective
electrodes," Pure and Applied Chemistry, vol. 66, pp. 2527 { 2536, 1994.
[33] S. Sheen, G. Bao, and P. Cooke, \Food surface texture measurement using re-
 ective confocal laser scanning microscopy," Journal of Food S, vol. 73, pp. E227
{ E234, 2008.
[34] H. Wang, H. Feng, W. Liang, Y. Luo, and V. Malyarchuk, \E ect of surface
roughness on retention and removal of Escherichia coli O157:H7 on surfaces of
selected fruits," Journal of Food Science, vol. 74, pp. E8 { E15, 2009.
94
Chapter 5
Detection of B. anthracis Spores with the Aid of A Micro uidic Flow Cell
5.1 Introduction
The bacterium B. anthracis, among the Category A bioterrorism agents, has the
ability to form a spore under starvation conditions. This spore is a dehydrated, thick-
walled cell that can remain metabolically dormant for years and survive in a wide
range of harsh environments [1]. Yet, once entering a nutrient-rich host, the spore
can germinate and grow to a vast number of vegetative cells, which in turn causes the
disease anthrax with high mortality rates. Although early, proper antibiotic treat-
ments have proven e ective in reducing the high mortality rates, such treatments are
often di cult to provide due to initial, non-speci c symptoms in infected patients [2].
Hence, anthrax infection is a signi cant public health-concern and must be prevented
through early detection of B. anthracis spores preferably before their entry into the
human body.
Since the 2001 anthrax mail attacks in the United States, substantial progress
has been made in the development of label-free biosensors for the low-cost, rapid,
on-site detection of B. anthracis spores. Among the most successful examples are
mass-sensitive biosensors, such as those based on a cantilever and a quartz-crystal
microbalance (QCM) as the signal transducer. These biosensors combined with an
antibody as the biomolecular-recognition element have shown the potential to rapidly
detect the spores at low concentrations (e.g., as few as 50 spores in water with 20
 m-long silicon cantilevers [3]; 300 spores/ml with piezoelectric-excited millimeter-
sized cantilever sensors [4]; and 103 spores/ml with a QCM [5]). However, due to
low thermal stability of antibodies in general, the outdoor use as well as longevity of
95
these antibody-based biosensors may be problematic [6]. As a result, there is a need
for thermally robust biosensors that can function outdoors for a fair period of time
while possessing cost-e ectiveness, rapidness, and high sensitivity as the minimum
performance requisites.
Phage-based ME biosensors, a class of wireless, mass-sensitive biosensors, are
among potential candidates that could meet the above-mentioned need. These biosen-
sors are composed of a freestanding, strip-shaped ME resonator as the wireless signal
transducer and a thermally stable landscape phage as the biomolecular-recognition el-
ement [7{10]. Previously, a landscape phage with high binding a nity for B. anthracis
spores has been reported (i.e., the JRB7 phage with a binding peptide sequence of
EPRLSPHS [11]). ME biosensors based on this phage have shown the retention of
binding activity at various temperatures (25, 45, and 65  C) for at least a hundred
days. By contrast, the best antibody-based ME biosensors tested alongside have re-
tained binding activity for a maximum of 5 days at 65  C [12]. In addition, the phage-
based ME biosensors are not only inexpensive and rapid in response time, but they
can be easily replaced after use due to their wireless nature, which may facilitate the
on-site monitoring of the presence of the spores. However, a remaining key issue has
been sensitivity. Theoretically, sensitivity improves as the size of the ME resonator
is reduced [13]. Hence, e orts have been made to manufacture and use small-sized
resonators for the sensitive detection of B. anthracis spores [12, 14{16]. However,
due to the challenge in acquiring a measurable transduced signal from small-sized
resonators, their length has been limited to millimeters to half a millimeter (5 to 0.5
mm), which are not signi cantly small to enable the detection of a few spores (e.g., on
the order of 10 or less). In other words, further reduction in the size of the resonator is
needed for the ME biosensors to be competitive in sensitivity. In addition, when such
miniature ME biosensors are used with a traditional large  ow cell, the chances of
physical contact between the biosensor and spores in the  ow cell is greatly reduced.
96
This speculation points to a need for a properly designed micro uidic  ow cell that
e ciently guides a stream of the spores towards the biosensor. Hence, this chapter
presents an investigation into the potential use of a micron-scale phage-based ME
biosensor combined with a micro uidic  ow cell for the rapid, sensitive detection of
B. anthracis spores. In this work, 200  m-long ME resonators were batch-fabricated
and used (i.e.,  106 resonators on a four-inch wafer). In this way, the fabrication
cost per resonator was reduced to a fraction of a cent, which allows the biosensors to
be disposable, and thus, free of cross-contamination.
5.2 Material and methods
5.2.1 JRB7 phage and B. anthracis Sterne spores
Suspensions of the JRB7 phage (5  1011 virions/ml in a PBS bu er, pH 7.2)
and B. anthracis Sterne spores (5  108 spores/ml in sterile distilled water) were
kindly provided by Dr. James Barbaree?s group at Auburn University. The Sterne
strain of B. anthracis is an attenuated strain that is incapable of causing anthrax
infection in humans and, thus, convenient for safe laboratory experiments. Yet, all
antigenic markers on the external surface of a Sterne spore are common with those of
a pathogenic spore. Hence, the binding characteristics of the a nity-selected JRB7
phage to the Sterne and pathogenic spores are expected to be identical. The con-
centrated spore suspension was diluted with sterile distilled water as desired prior to
use.
5.2.2 Batch-fabrication of micron-scale ME resonators
ME resonators with a size of 200  m  40  m  4  m were batch-fabricated
using the co-sputtering-based method described in Chapter 3. Figure 5.1 shows scan-
ning electron micrographs of the fabricated ME resonators. After lift-o , these ME
resonators were annealed and immobilized with the JRB7 phage.
97
Figure 5.1: Scanning electron micrographs of 200  m 40  m 4  m ME resonators
fabricated on a  at wafer: (a) batch-fabricated resonators and (b) a close-up view.
5.2.3 Micro uidic  ow cells
In order to ensure e cient physical contact between the biosensor and B. an-
thracis spores, micro uidic  ow cells were designed and fabricated of a silicone elas-
tomer, poly(dimethylsiloxane) (PDMS), by multilayer soft lithography (Fig. 5.2). In
this investigation, the following two types of  ow cells were fabricated and used:
Figure 5.2: Procedure for the fabrication of a micro udic chip by multilayer soft
lithography.
98
1. Type I micro uidic  ow cell with pneumatic push-up valves for a proof-of-
concept experiment (i.e., manipulation of sensors and micro-spheres).
2. Type II micro uidic  ow cell with pneumatic push-down valves integrated with
an ME biosensor for the rapid, sensitive detection of B. anthracis spores.
5.2.3.1 Design and fabrication of the Type I micro uidic  ow cell
The design of the Type I micro uidic  ow cell is shown in Fig. 5.3. The  ow
cell is composed of one reaction chamber (300  m  300  m  50  m) and four
channels that were used for the injection and removal of a sensor and micro-spheres.
Pneumatic push-up valves were used to open and close the connections between the
chamber and channels. Two PDMS layers, the  uidic and control layers, are needed
to realize this actuation mechanism. The control layer mold was fabricated using the
SPR 220-7.0 photoresist (Rohm and Haas Electronic Materials, LLC.) with a typical
height of 15  m. By contrast, the mold for the  uidic layer was manufactured using
two di erent photoresists, AZ P4620 (AZ Electronic Materials USA Corp.) and SU-8
2025 (Microchem Corp.) for micro-features of di erent height. The fabricated micro-
features shown in blue and green lines in Fig. 5.3 have typical heights of 20  m and
50  m, respectively.
99
Figure 5.3: Design of the Type I  ow cell: (a) a three-dimensional view of the whole
chip and (b) a close-up view of the key elements of the chip.
100
Commercially available GE RTV 615 was used to make the PDMS micro uidic
 ow cell. Typically, 5 parts of GE RTV 615 part A (base) was mixed thoroughly with
1 part of part B (curing agent), followed by degasi cation under vacuum for 1 to 2
hours. This mixture was, then, gently poured onto the  uidic layer mold and cured
at 80  C for 45 min. When cooled down, the PDMS replica was gently peeled o the
mold. After access holes were punched into this replica, this PDMS  uidic layer was
stored in a clean-room chamber until use. For the fabrication of the PDMS control
layer, 20 parts of GE RTV 615 A was mixed thoroughly with 1 part of part B, followed
by degasi cation under vacuum for 15 minutes. The mixture was, then, spin-coated
onto the control layer mold at a rotation speed of 3,200 rpm for 120 sec and cured at
80  C for 30 min. The PDMS  uidic layer, which was pre-cleaned with methanol and
dried, was, then, aligned to the control layer using an optical microscope. Afterwards,
the aligned chip was cured at 80  C for 4 hours to complete the bonding between the
two layers. The  nal chip was sealed against a clean glass slide, which was pre-coated
with a thin layer of PDMS. The  nal bonding between the chip and glass slide was
performed at 80  C for 4 hours.
5.2.3.2 Design and fabrication of the Type II micro uidic  ow cell
The design of the Type II micro uidic  ow cell is similar to that of Type I. The
di erences between them are as follows:
1. The Type II  ow cell has push-down valves (Fig. 5.4). In other words, the
control layer is located on the  uidic layer.
2. The Type II  ow cell was sealed against a slotted glass slide, which facilitates
the collection of an ME biosensor after exposure to B. anthracis spores for
subsequent resonant frequency measurement. Figure 5.5 shows the design of
the Type II  ow cell.
101
Figure 5.4: Push-up and push-down valves.
102
Figure 5.5: Design of the Type II  ow cell: (a) a three-dimensional view of the whole
chip on a slotted glass slide and (b) a close-up view of the key elements of the chip.
103
5.3 Valve actuation and sample injection
The actuation of pneumatic valves was tested with a controlled pressure from a
nitrogen gas source. Fluidic pressure controllers and manifolds (Fluidigm corporation)
were used to control the pressure supply to the control channels  lled with water. For
the  uidic channels, a much smaller pressure was needed to control the travel speed of
analytes (i.e., micro-spheres and spores). Hence, a digital pressure gauge (DPG1203-
005 from Omega Engineering Inc.) connected with a low pressure regulator (Go
Regulator Company) was used to adjust the pressure required for the manipulation
of the analytes. Flexible polymeric tubing (TYGON from Saint-Gobain PPL Corp.)
and metal pins were connected to inject liquid samples into chips as shown in Fig.
5.6.
Figure 5.6: A close-up top view of a fabricated chip (Type I), connected to external
pressure sources for valve actuation and sample injection.
104
5.4 Results and discussion
5.4.1 Con rmation of valve actuation
Valve actuation was con rmed by optical microscopic observation. As shown
in Fig. 5.7a, all the four valves of a Type I  ow cell were successfully closed at a
pressure of 15 psi. In this way, the center chamber can be separated from all the
connected  uidic lines, which enables the mechanical separation of any tiny analyte
when injected into the  ow cell. To further test the separation capabilities of the
 ow cell, a green food dye was injected into the chamber through the vertical  uidic
channels while the #2 and #4 valves were kept closed. As can be seen in Fig. 5.7b,
the green dye only  lled the chamber, and no leakage towards the horizontal  uidic
channels occurred.
Figure 5.7: Close-up top views of a Type I  ow cell: (a) All the valves (#1 through
#4) are closed, and (b) only the #2 and #4 valves are closed such that the green dye
injected through the vertical channels can only  ll out the center chamber.
5.4.2 Manipulation of  uorescent-labeled micro-spheres
Fluorescent-labeled micro-spheres that have equivalent dimensions to B. an-
thracis spores (about 1.5  m) were used to test the manipulation capabilities of the
Type I micro uidic  ow cell. The micro-spheres were injected through the vertical
105
 uidic channels and traced under an inverted  uorescence microscope (Axiovert 40
CFL from Carl Zeiss, Inc.). Once a desired number of spheres reached the chamber,
they were isolated by closing valves. The results in Fig. 5.8 show the feasibility of ma-
nipulating a small number of micro-spheres. Figures 5.8a to 5.8c show the separation
of four spheres, two spheres, and one sphere, respectively.
Figure 5.8: Fluorescent micrographs showing that the  ow cell can separate a few
micro-spheres into the reaction chamber: (a) separation of four spheres, (b) two
spheres, and (c) one sphere.
5.4.3 Manipulation of a sensor
Similarly, a sensor can be injected through the horizontal channels and positioned
in the reaction chamber as shown in Fig. 5.9.
106
Figure 5.9: Injection of a 200  m 40  m 4  m sensor into the chamber through
the horizontal channels.
5.4.4 Detection of B. anthracis spores with the Type II micro uidic  ow
cell
With the Type I mico uidic  ow cell, the potential usefulness of a polymeric
micro udic system with pneumatic valves was veri ed. For the enhanced detection
of B. anthracis spores, the Type II  ow cell in combination with a 200  m-long ME
biosensor was employed. The Type II  ow cell was sealed on a slotted glass slide,
which facilitates the collection of the ME biosensor after exposure to B. anthracis
spores for subsequent resonant frequency measurement. Figure 5.10 shows an optical
micrograph of the key elements of the Type II micro uidic  ow cell. In this  ow
cell, B. anthracis spores were injected and removed though the vertical channels at
a  ow velocity of 100 to 150  m/s in the circular reaction chamber, whereas the
107
Figure 5.10: Optical micrograph of the Type II micro uidic  ow cell. All the push-
down valves are closed.
ME biosensor was moved through the horizontal channels using a bar magnet. After
exposure to B. anthracis spores in the chamber, the sensor was collected through the
slot in the glass slide, which is located under the  ow cell. A representative image
of the testing is shown in Fig. 5.11, where the binding of  uorescent-labeled spores
on a 200  m  40  m  4  m ME biosensor can be seen. Interestingly, the binding
process occurred very quickly (on the order of minutes), which can be monitored in
a real-time manner using the microscope. After this binding step, the biosensor was
collected, and measurement of it?s  nal resonant frequency was completed within 5
min. Hence, the total assay time was only about 10 min.
108
Figure 5.11: Fluorescent micrograph showing B. anthracis spores are bound on a 200
 m  40  m  4  m ME biosensor. The chamber was outlined with a yellow solid
line for better visualization.
Figure 5.12 shows a result of a streamline analysis using the Caedium software
[17]. The streamlines represent the instantaneous paths of massless particles. Due to
the limited paths that spores can take in the micron-scale  ow cell, the chances of
physical contact between the sensor and spores were greatly enhanced. In addition,
the  ow velocity in the reaction chamber was found to be low when compared with
those in the connected  uidic channels. This decreased velocity is also likely to
facilitate the binding of spores on the sensor surfaces.
109
Figure 5.12: Streamlines in the micro uidic  ow cell. The color bar below indicates
the magnitude of  ow velocity.
Figure 5.13: Responses of ME biosensors (200  m  40  m  4  m) to various
numbers of spores.
110
Finally, Fig. 5.13 shows responses of 200  m 40  m 4  m ME biosensors to
various numbers of B. anthracis spores. As can be seen, resonant frequency changes
for measurement sensors (black circles) were largely dependent on the number of
bound spores. At 106 spores, a minimum frequency change of 1,993 Hz was obtained.
In addition, the same experimental procedure was used to test control sensors, which
were not immobilized with the JRB7 phage but only surface-blocked with BSA. The
number of non-speci cally bound spores on the control sensors was found to be very
small (6 to 28 spores), and the average of the corresponding resonant frequency
changes was 683 240 Hz. This average response is much smaller than the minimum
response of the measurement sensors (i.e., 1,993 Hz for 106 spores). Hence, it can be
concluded that 200  m-long ME biosensors were capable of detecting 106 spores. All
the dose-response data are summarized in Table 5.1.
Table 5.1: Number of bound spores and corresponding resonant frequency changes
for measurement and control sensors.
Sample # Number of bound spores Resonant frequency change [Hz]
Measurement 1 238 6,890
Measurement 2 282 3,299
Measurement 3 124 2,865
Measurement 4 534 8,270
Measurement 5 562 12,390
Measurement 6 106 1,993
Measurement 7 1,142 15,299
Measurement 8 296 3,968
Measurement 9 502 7,562
Measurement 10 236 5,739
Measurement 11 378 8,768
Control 1 6 638
Control 2 28 943
Control 3 14 469
111
5.5 Conclusions
PDMS micro uidic  ow cells were designed, fabricated, and used with 200  m 
40  m 4  m ME biosensors for the enhanced detection of B. anthracis spores. Due
to the enhanced chances of physical contact between the biosensors and spores, the
time required for the testing was only about 10 min. In addition, with the micron-
scale ME biosensors, down to 106 spores were experimentally detected.
112
Bibliography
[1] A. Driks, \The Bacillus anthracis spore," Molecular Aspects of Medicine, vol. 30,
no. 6, pp. 368 { 373, 2009.
[2] T. V. Inglesby, T. O?Toole, D. A. Henderson, J. G. Bartlett, M. S. Ascher,
E. Eitzen, A. M. Friedlander, J. Gerberding, J. Hauer, J. Hughes, J. McDade,
M. T. Osterholm, G. Parker, T. M. Perl, P. K. Russell, and K. Tonat, \Anthrax
as a biological weapon, 2002: Updated recommendations for management," The
Journal of the American Medical Association, vol. 287, no. 17, pp. 2236 { 2252,
2002.
[3] A. P. Davila, J. Jang, A. K. Gupta, T. Walter, A. Aronson, and R. Bashir,
\Microresonator mass sensors for detection of Bacillus anthracis Sterne spores
in air and water," Biosensors and Bioelectronics, vol. 22, no. 12, pp. 3028 { 3035,
2007.
[4] G. A. Campbell and R. Mutharasan, \Detection of Bacillus anthracis spores and
a model protein using PEMC sensors in a  ow cell at 1 mL/min," Biosensors
and Bioelectronics, vol. 22, no. 1, pp. 78 { 85, 2006.
[5] R. Hao, D. Wang, X. Zhang, G. Zuo, H. Wei, R. Yang, Z. Zhang, Z. Cheng,
Y. Guo, Z. Cui, and Y. Zhou, \Rapid detection of Bacillus anthracis using mon-
oclonal antibody functionalized QCM sensor," Biosensors and Bioelectronics,
vol. 24, no. 5, pp. 1330 { 1335, 2009.
[6] V. A. Petrenko, \Landscape phage as a molecular recognition interface for de-
tection devices," Microelectronics Journal, vol. 39, no. 2, pp. 202 { 207, 2008.
[7] Y. Chai, S. Li, S. Horikawa, M.-K. Park, V. Vodyanoy, and B. A. Chin, \Rapid
and sensitive detection of Salmonella Typhimurium on eggshells by using wireless
biosensors," Journal of Food Protection, vol. 75, no. 4, pp. 631 { 636, 2012.
[8] S. Horikawa, D. Bedi, S. Li, W. Shen, S. Huang, I.-H. Chen, Y. Chai, M. L.
Auad, M. J. Bozack, J. M. Barbaree, V. A. Petrenko, and B. A. Chin, \E ects
of surface functionalization on the surface phage coverage and the subsequent
performance of phage-immobilized magnetoelastic biosensors," Biosensors and
Bioelectronics, vol. 26, no. 5, pp. 2361 { 2367, 2011.
[9] S. Li and Z.-Y. Cheng, \Nonuniform mass detection using magnetostrictive
biosensors operating under multiple harmonic resonance modes," Journal of Ap-
plied Physics, vol. 107, p. 114514, 2010.
113
[10] M.-K. Park, H. C. W. III, Y. Chai, S. Horikawa, W. Shen, and B. A. Chin, \The
e ect of incubation time for Salmonella Typhimurium binding to phage-based
magnetoelastic biosensors," Food Control, vol. 26, no. 2, pp. 539 { 545, 2012.
[11] J. Brigati, D. D. Williams, I. B. Sorokulova, V. Nanduri, I.-H. Chen, C. L.
Turnbough, and V. A. Petrenko, \Diagnostic probes for Bacillus anthracis spores
selected from a landscape phage library," Clinical Chemistry, vol. 50, no. 10,
pp. 1899 { 1906, 2004.
[12] J. Wan, H. Shu, S. Huang, B. Fiebor, I.-H. Chen, V. A. Petrenko, and B. A. Chin,
\Phage-based magnetoelastic wireless biosensors for detecting Bacillus anthracis
spores," IEEE Sensors Journal, vol. 7, pp. 470 { 477, 2007.
[13] C. A. Grimes, C. S. Mungle, K. Zeng, M. K. Jain, W. R. Dreschel, M. Paulose,
and K. G. Ong, \Wireless magnetoelastic resonance sensors: A critical review,"
Sensors, vol. 2, pp. 294 { 313, 2002.
[14] S. Huang, H. Yang, R. Lakshmanan, M. Johnson, J. Wan, I.-H. Chen, H. W.
III, V. Petrenko, J. Barbaree, and B. Chin, \Sequential detection of Salmonella
typhimurium and Bacillus anthracis spores using magnetoelastic biosensors,"
Biosensors and Bioelectronics, vol. 24, no. 6, pp. 1730 { 1736, 2009.
[15] M. L. Johnson, J. Wan, S. Huang, Z. Cheng, V. A. Petrenko, D.-J. Kim, I.-
H. Chen, J. M. Barbaree, J. W. Hong, and B. A. Chin, \A wireless biosensor
using microfabricated phage-interfaced magnetoelastic particles," Sensors and
Actuators A: Physical, vol. 144, no. 1, pp. 38 { 47, 2008.
[16] J. Wan, M. L. Johnson, R. Guntupalli, V. A. Petrenko, and B. A. Chin, \De-
tection of Bacillus anthracis spores in liquid using phage-based magnetoelastic
micro-resonators," Sensors and Actuators B: Chemical, vol. 127, no. 2, pp. 559
{ 566, 2007.
[17] http://www.symscape.com/.
114
Chapter 6
Enhancing the Detection Capabilities of Phage-Based ME Biosensors
In this chapter, the following two e ects will be investigated to further enhance
the detection capabilities of phage-based ME biosensors:
1. E ects of mass position on the sensitivity of ME biosensors
2. E ects of surface functionalization on surface phage coverage.
6.1 E ects of mass position on the mass sensitivity of ME biosensors
6.1.1 Introduction
One of the ultimate goals to be reached for phage-based ME biosensors is the
real-time detection of single pathogenic bacteria. Although the principle of detection
for freestanding, strip-shaped ME biosensors has been well documented, conventional
theories describe only the case where attached masses are uniformly distributed over
the ME sensors [1]. For non-uniformly distributed masses, the mass sensitivity of
ME sensors is largely dependent on the position of the masses attached to the sensor
surfaces [2]. Considering this dependence is crucial to detection of dilute analytes
(e.g., low-concentration bacterial samples) because their local attachment may cause
varying sensor responses. To address the issue, three-dimensional  nite element (FE)
models were constructed for di erently sized phage-based biosensors with a single
localized mass. The FE simulations allow one to predict the mass-position-dependent
resonant frequency changes of biosensors.
115
6.1.2 Material and methods
6.1.2.1 Microcontact printing
In order to address the issue, resonant frequency measurements of phage-based
ME biosensors with localized masses were  rst performed. Gold-coated ME sensor
platforms (4 mm  0.8 mm  30  m), made of Metglas 2826MB (from Honeywell
International, Inc.), were manufactured with the procedure described in Chapter 3.
The sensor platforms were, then, loaded with the SAE10 phage, which speci cally
binds with streptavidin (See Table 3.3), and  nally with BSA (0.1 % w/v in 330- l,
100-mM HEPES) for surface blocking. Microcontact printing was used to locally
place streptavidin-coated polystyrene beads (hereafter, SA beads, 0.68  0.11  m in
diameter, from Spherotech, Inc.) on the phage-based ME biosensors as shown in Fig.
Figure 6.1: PDMS stamps and microcontact printing. SA beads (not to scale) were
placed in the middle (stamping area: 1 mm 0.8 mm) or (b) at both ends (stamping
area: 0.5 mm  0.8 mm each) of a phage-immobilized ME biosensor using the Type
A or Type B stamp, respectively. BSA is not shown.
116
6.1. Two types of printing stamps, made of PDMS (from Momentive Performance
Materials, Inc.), were fabricated by soft lithography and used. The Type A stamp
places SA beads in the middle of a sensor (1 mm  0.8 mm), whereas the type B
stamp locates SA beads at both ends of a sensor (0.5 mm  0.8 mm). The total
stamping areas were kept the same for both types such that equivalent total masses
could be placed at di erent locations on the sensor surfaces. A suspension of SA beads
(3.6  1010 beads/ml) was pipetted onto the islands of the PDMS stamps. In this
way, the beads were transferred to the surfaces of the ME biosensors. This procedure
was followed by washing with  ltered DI water three times to remove unbound and
loosely bound SA beads.
Resonant frequency measurements of the ME biosensors were performed in air
with the setup described in Chapter 3. The fundamental resonant frequency of lon-
gitudinal vibration was measured in this investigation.
6.1.3 Results and discussion
Figure 6.2 shows scanning electron micrographs of the phage-immobilized ME
biosensors (4 mm  0.8 mm, top view) loaded with SA beads. As a reference, an
ME biosensor with uniform coverage of beads is also shown (Fig. 6.2c). The brightly
colored regions on the sensors are where the SA beads are loaded. The average
surface bead coverage in these regions was found to be 51.9%. As shown in Table
6.1, nonequivalent resonant frequency changes were experimentally obtained for the
ME biosensors with localized bead masses at di erent positions (middle vs. ends).
The mass sensitivity was low when the mass was loaded in the middle of the ME
biosensors. By contrast, a higher mass sensitivity was obtained for the placement of
the mass at both ends of the sensors. These results are reasonable because the nodal
point (i.e., the point with zero displacement) is located in the middle of the biosensors
for the fundamental mode of longitudinal vibration [2].
117
Figure 6.2: Phage-based ME biosensors loaded with SA beads. Beads are attached (a)
in the middle, (b) at both ends, or (c) uniformly on both sides of the ME biosensors.
Table 6.1: Resonant frequency changes in Hz.
Mass position Experimental result Numerical result
Middle -33  17 -10
Ends -212  48 -179
Uniform1 -797  52 -755
1 Beads are attached to the back side as well.
6.1.3.1 Finite element modal simulation
Three-dimensional  nite element (FE) analysis was performed to simulate the
above phenomena (CalculiX, ver. 2.5, Convergent Mechanical Solutions, Seattle,
WA). Convergence tests were conducted for all FE models, prior to eigenfrequency
analysis, to ensure the element sizes that were chosen were  ne enough to simulate
the ME biosensor (i.e., quadratic brick elements with a size of 30 to 50  m). In this
work, the SA beads attached to the sensor surfaces were modeled as an orthotropic
 lm mass (0.68- m thick) as shown in Fig. 6.3a. Because of the discontinuities
among the SA beads in reality, the materials constants for the  lm mass were given
accordingly as summarized in Table 6.2. For simpli cation,  at rectangular  lms (1
mm 0.8 mm in the middle and 0.5 mm 0.8 mm at both ends) were modeled. The
118
Figure 6.3: Three-dimensional FE model: (a) geometry, (b) meshed geometry, and
(c) resultant mode shape.
metal deposits [3] (90-nm thick Cr and 150-nm thick Au) and phage layer ( 10 nm)
on the ME biosensor platforms were neglected because of their small e ects on the
sensors? resonant frequencies. As shown in Table 6.1, the simulation results showed
close agreement with the experimental results.
After this validation of the FE method, similar FE models were constructed and
used to predict the mass sensitivity of phage-immobilized ME biosensors loaded with
119
Table 6.2: Materials constants used in FE simulationsa.
[GPa] [GPa] [GPa] [GPa] [g/cm3]
Part E EL = EW ET GLW=WT=TL    LW=WT=TL
Beadsb | 10 5 1.56 10 5 0.54 | 0
Bacteriac 0.025 | | | 1.05 0.499 |
Sensord
(Metglas)
105 | | | 7.9 0.33 |
Sensore
(Fe80B20)
166 | | | 7.4 0.3 |
a E, G,  , and  denote the elastic modulus, shear modulus, density, and Poisson?s ratio,
respectively. The subscripts L, W, and T represent the axes along the length, width,
and thickness of the ME biosensor.
b ET was estimated by the rule of mixture from the surface bead coverage (51.9%) and a
rough elastic modulus of a polystyrene bead (3 GPa [5]).  was calculated from both the
surface bead coverage and density of a polystyrene bead (1.05 g/cm3 from Spherotech,
Inc.) to keep the mass of the  lm equivalent to that of the SA beads attached to the
ME biosensors.
c E from [6] and  from [7]. Due to the high water content in a bacterial cell,   0.5 was
used.
d Values obtained from Ref. [4].
e Values previously shown in Table 3.4.
a single bacterial mass. To investigate the e ects of the sensor?s dimensions on the
mass sensitivity, FE models with varying lengths (100 { 500  m), widths (lateral
aspect ratios of 5 { 100), and thicknesses (1 { 15  m) were built. The bacterial mass,
treated as an isotropic mass (Table 6.2), was placed in the middle of the width and at
di erent longitudinal positions of the ME biosensors (the e ect of the mass position
along the width was small). The dimensions of the mass were chosen to be 2  m  
0.4  m  0.4  m, which is of typical of the bacteria Listeria [6].
Figure 6.4 shows representative results of the dependence of mass sensitivity,
 f/ m, on the longitudinal position of the attached mass and dimensions of ME
biosensors. Similar to the results presented above, the mass sensitivity was largely
dependent on the longitudinal mass position. Symmetric double-sigmoidal curves
were found to show the best  t to the simulation data points (Fig. 6.4a). It was also
120
Figure 6.4: Dependence of the mass sensitivity,  f= m, on the longitudinal position
of the attached mass and on the dimensions of the ME biosensors.
121
found that the mass sensitivity is inversely proportional to L2, W, and T as shown
in Figs. 6.4b { 6.4d, where L, W, and T denote the length, width, and thickness of
the ME biosensors, respectively.
From the FE simulation results, the mass sensitivity of a phage-immobilized ME
biosensor loaded with a single bacterium was derived to be
 f
 m[Hz/pg]  
1
L2WT
2
4 2:805 108
1 + exp
 
L0 L=4
 0:082L 0:239
  1:310 107
3
5; (6.1)
where  f,  m, and L0 denote the resonant frequency change, mass of the attached
bacterium, and longitudinal distance of the attached bacterium from the center of
the ME biosensor, respectively. Equation (6.1) describes a Boltzmann function whose
amplitude is in uenced by the dimensional variables, L, W, and T (the unit is  m).
Hence, this equation allows one to estimate the mass sensitivity and, thus, resonant
frequency change of di erently sized ME biosensors upon the attachment of a single
bacterium ( m here is 2  m  0.4  m  0.4  m  1.05 g/cm3 = 0.336 pg).
In summary, the mass sensitivity of ME biosensors is largely dependent on the
longitudinal mass position and dimensions of the sensors. The equation derived in
this investigation can be used to predict the mass sensitivity and resonant frequency
change of di erently sized ME biosensors for the detection of single pathogenic bac-
teria.
122
6.2 E ects of surface functionalization on surface phage coverage
Traditionally, phage-based ME biosensors have been constructed by immobilizing
a landscape phage on gold-coated ME resonators via physical adsorption. Although
the physical adsorption method is simple, the immobilization stability and surface
coverage of a phage on di erently functionalized sensor surfaces need to be evaluated
as a potential way to enhance the detection capabilities of the biosensors.
6.2.1 Introduction
The phage display technology, primarily developed for the Ff class of  lamentous
phages [8,9], enables one to construct billions of phage clones that display engineered
sequences of peptides on their outer surfaces. Today, such phage clones are called
landscape phages, and they are being used in a variety of applications due to their
diversity, low cost, ease of production, and environmental robustness [10, 11]. In
bacterial detection, for instance, a number of attempts have been recently made to
employ landscape phages as biomolecular recognition elements, which substitute for
commonly used antibodies.
A biomolecular recognition element is a key component of a biosensor used to
speci cally target and bind a pathogen of interest to the signal transducer [12]. Hence,
in phage-based biosensing systems, robust and e cient immobilization of a phage
has been one of the spotlighted issues. Widely used immobilization methods include
physical adsorption, biotin { avidin coupling, and covalent attachment through a
self-assembled monolayer (SAM). Among these methods, physical adsorption is the
simplest method, and it has been traditionally used to construct our phage-based ME
biosensors. However, the immobilization stability and quantity of physically adsorbed
phages may be lower than those of biotin { avidin coupled and covalently bound
phages because of the weak, non-speci c bonding of physical adsorption. Recently,
a covalent immobilization of a landscape phage has been reported [13]. A SAM of
123
N-hydroxysuccinimide thioctic ester was used to covalently bind M13 phage onto
a gold-coated quartz-crystal microbalance. The immobilization chemistry yielded a
surface phage density of 1.1 1011 virions/cm2, which is 70% higher than the highest
reported surface density of a physically adsorbed fd-tet phage [14]. In addition, a
study of chemical immobilization of T4 phage on gold surfaces has shown that phage
immobilization through glutaraldehyde-activated cysteine/cysteamine produced a 37-
fold increase in surface phage density and a 9-fold increase in the surface density
of subsequently captured E. coli host cells, compared to those with physical phage
adsorption [15]. Hence, these pieces of recent research have indicated that surface
functionalization may a ect the stability and quantity of immobilized phages.
The primary objectives of this work are: (1) to investigate the e ects of surface
functionalization on the surface coverage of a  lamentous fd-tet phage and (2) to
 nd the correlation between the observed surface phage coverage and the quantity
of a subsequently captured target analyte. As a model study, an fd-tet phage that
speci cally binds with streptavidin (the SAE10 phage in Table 3.3) was adsorbed
on gold-coated ME resonators with a size of 1 mm  0.2 mm  15  m. The sur-
faces of the resonators were either bare or functionalized with three di erent SAMs as
shown in Fig. 6.5: (a) AC (activated carboxyl-terminated SAM: 1-[5-(1,2-dithiolan-3-
yl)pentanoyl]oxypyrrolidine-2,5-dione); (b) ALD (aldehyde-terminated SAM: N-[19-
(1,2-dithiolan-3-yl)-15-oxo-4,7,10-trioxa-14-azanonadec-1-yl]-4-formylbenzamide); and
(c) MT (methyl-terminated SAM: hexanethiol). Surface functionalization with these
SAMs was based on the sulfur - gold chemistry, which was veri ed by contact angle
goniometry and X-ray photoelectron spectroscopy (XPS). Atomic force microscopy
(AFM) was used to quantify surface phage coverage and to visualize the distribution
of the  lamentous phage on the ME resonator surfaces. In addition, dose-responses
of the phage-immobilized ME biosensors to streptavidin-coated microbeads were ob-
tained through resonant frequency measurements.
124
Figure 6.5: Surface functionalization of gold (1 1 1) with three SAM chemicals, based
on the sulfur - gold chemistry: (a) AC (activated carboxyl-terminated), (b) ALD
(aldehyde-terminated), and (c) MT (methyl-terminated).
6.2.2 Material and methods
All chemical and biological reagents were purchased from Fisher Scienti c (Pitts-
burgh, PA, USA), Sigma-Aldrich (St.Louis, MO, USA), or SoluLinK (San Diego, CA,
USA) and used as-received unless otherwise noted. Filtered deionized (DI) water (pH
7.4) was prepared with SimPak 1 Puri cation Pack Kit (Millipore, Billerica, MA,
USA) and used for solution making and sample washing. A HEPES bu er (100-mM,
pH 7.3) was prepared and  ltered with Stericup  lter units (Millipore, Billerica, MA,
USA) prior to use.
125
6.2.2.1 Preparation of biological samples
The SAE10 phage, which speci cally binds with streptavidin [16], was prepared
in a 100 mM HEPES bu er (6.5  1011 virions/ml, pH 7.3). Streptavidin-coated
polystyrene microbeads (hereafter, SA beads, 0.68 0.11  m in diameter (measured),
3.6  1010 beads/ml) were purchased from Spherotech, Inc. (Lake Forest, IL, USA),
transferred to a 100-mM HEPES bu er (pH 7.3) by centrifugation, and used for dose-
response experiments of ME biosensors. Serial dilutions of the bead suspension were
also prepared with a 100-mM HEPES bu er (pH 7.3).
6.2.2.2 Manufacture of gold-coated ME resonators
ME resonators with a size of 1 mm 0.2 mm 15  m were manufactured with
the dicing-based procedure described in Chapter 3. After annealing, chromium (90-
nm thick) and gold (150-nm thick) were successively deposited on all the resonator
surfaces by sputtering at 2.5 10 6 Torr (Denton Vacuum, Moorestown, NJ, USA).
The predominant crystallographic orientation of the gold deposit was (111), veri ed
by x-ray di raction analysis (D/MAX-B, Rigaku, Inc., Danvers, MA, USA). The
outer-coat gold deposit protects ME resonators from corrosion as well as provides
suitable surfaces for surface functionalization (i.e., sulfur { gold chemistry-based)
and immobilization of the phage.
6.2.2.3 Surface functionalization of gold-coated ME resonators
Surface functionalization, based on the sulfur { gold chemistry [17], was per-
formed. The three SAM chemicals, AC, ALD, and MT, possess either 1,2-dithiolane
(C3H6S2) or a thiol group ({SH) in their structures as the head group (Figure 6.5).
AC and ALD were individually dissolved (16.5 mM each) in dry dimethylformamide
(DMF, 99.8% pure), whereas MT was dissolved (1 mM) in dry ethanol ( 99.5%
pure). Freshly prepared gold-coated ME resonators were immersed in each of the
126
SAM chemical solutions for 36 hours at 23 C in a desiccator, protected from UV ex-
posure. The surface-functionalized ME resonators were, then, washed with dry DMF
once and dry ethanol twice (The MT-functionalized ME resonators were washed with
dry ethanol three times.). After drying with high quality argon gas ( 99.999% pure,
Airgas South, Kennesaw, GA, USA), the surface-functionalized ME resonators were
stored in a desiccator, protected from UV exposure, until used.
6.2.2.4 X-ray photoelectron spectroscopy
Photoemission measurements were performed for bare and SAM-modi ed gold
surfaces in a load-locked Kratos XSAM 800 surface analysis system (Kratos Analytical
Inc., New York, USA). The base pressure of the system was 1  10 10 torr. XPS
spectra were recorded in the  xed analyzer transmission (FAT) mode with a pass
energy of 20 eV, appropriate for acquisition of high resolution, high signal-to-noise
spectra. The magni cation of the analyzer in the FAT mode was selected to collect
electrons from the smallest allowable area (5 mm2) on the specimen. The resolution
of the instrument at the operating parameters was measured from the full-width-at-
half-maximum of the Ag3d5=2 peak to be 1.0 eV. The XPS energy scale was calibrated
by setting the Ag3d5=2 line on clean silver to exactly 368.3 eV referenced to the Fermi
level. Because of specimen charging during x-ray irradiation of the organic specimens,
the energy axis of each XPS spectra was shifted to make the C1s binding energy line
equal to 284.7 eV, a standard hydrocarbon energy (C-H and C-C bonds) used to
reference charge-a ected materials. The uncertainty of the binding energies recorded
was 0.1 eV. The photoelectrons were excited by a water-cooled Kratos Model WG-170
dual anode x-ray gun equipped with an aluminum window. The angle of the incidence
of the x-ray beam with the specimen normal was 51.5 . Unmonochromatized MgK 
(1253.6 eV) radiation was used exclusively. The x-ray power was kept relatively low
127
(260W, 13 kV at 20 mA) so as to minimize sample heating. Raw XPS spectra were
curve- tted by the Savitzky-Golay method (10-point, second-order  tting).
6.2.2.5 Loading of the phage on the ME resonators
The prepared ME resonators were individually immersed in 330  L of the phage
suspension in a polypropylene PCR tube. The tubes were, then, rotated with a
Barnstead LabQuake tube rotator (from Fisher Scienti c, Inc.) at 8 rpm for 1 h,
protected from UV exposure. In this way, the phage was allowed to uniformly attach
to resonator surfaces via either physical adsorption or covalent attachment. After
loading the phage, these resonators were washed thoroughly to remove any HEPES
bu er component as well as loosely attached phages from the resonator surfaces.
The surface-functionalized ME resonators loaded with phage were washed with a
wash bu er (Tween 20 in a 100-mM HEPES bu er (0.5% v/v, pH 7.3)) three times,
whereas the bare ME resonators loaded with phage were washed with DI water three
times, following the traditional way of sensor preparation described in Chapter 3.
Finally, all the phage-loaded resonators were surface-blocked with BSA (0.1% w/v in
a 100-mM HEPES bu er), followed by washing with  ltered DI water three times.
Samples for AFM imaging were not BSA blocked.
6.2.2.6 Resonant frequency measurement
Resonant frequency measurements of ME biosensors were performed with the
setup described in Chapter 3. For dose-response experiments, a phage-immobilized
ME biosensor placed in the center of the solenoid coil was exposed to a suspension of
SA beads with various concentrations (3.6  105 to 3.6  1010 beads/ml in 100-mM
HEPES) at a  ow rate of 25  l/min (Ismatec CP 78017-10, Cole-Parmer Instrument
Company, Vernon Hills, IL, USA) as shown in Figure 6.6. The resonant frequency
of the ME biosensors was recorded every 15 seconds with a resolution of 25 Hz. In
128
addition, BSA-blocked ME biosensors (without phage) were tested as controls. A 100-
mM HEPES bu er was  owed before and after the bead exposure to eliminate any
viscosity e ects on the resonant frequency change and to remove unbound and loosely
bound SA beads from the biosensor surfaces. UV exposure during measurements
was minimized by covering the solenoid coil with an opaque plastic box. Selective
capture of SA beads on the phage-immobilized ME biosensors was veri ed by scanning
electron microscopy (SEM) (JSM-7000F, JOEL, Ltd., Tokyo, Japan).
Figure 6.6: Schematic illustration of the frequency measurement setup for ME biosen-
sors. (a) The setup consists of a copper solenoid coil, a bar magnet, and a network
analyzer (not shown). (b) A phage-immobilized ME biosensor is placed in the glass
capillary  ow cell and positioned in the coil center. A suspension of SA beads was
passed at 25  l/min over the sensor, and the resonant frequency change of the biosen-
sor was monitored in a wireless, magnetic manner.
6.2.3 Results and discussion
6.2.3.1 Veri cation of the surface functionalization of gold
The surface functionalization of gold was veri ed by contact angle goniometry
(Model 200, Ram^e-Har, Inc., Mountain Lakes, NJ, USA) and XPS (Kratos XSAM
800, Kratos Analytical, Inc., New York, NY, USA). Since the gold-coated ME res-
onators (1 mm  0.2 mm) were too small for these analyses, gold surface specimens
129
Figure 6.7: (a) Water contact angles for bare and surface-functionalized gold surfaces.
(b) XPS S2p peaks at around 162 eV, indicating the formation of sulfur { gold bonded
systems.
were, instead, prepared on polished silicon wafer pieces (1 cm  1 cm), surface-
functionalized in the same ways described earlier, and tested. As shown in Figure
6.7a, the water contact angle varied as a result of the surface functionalization (mea-
sured at 23 C and 31% relative humidity with water drops of pH 7.8). The polar
activated carboxyl- and aldehyde-terminated surfaces (AC- and ALD-functionalized
surfaces) showed lower contact angles than the reference gold surface, whereas the
130
non-polar methyl-terminated surface (MT-functionalized surface) showed a dramatic
increase in contact angle. XPS analysis was also performed to con rm the contact
angle results. As shown in Figure 6.7b, the S2p peaks for the surface-functionalized
gold surfaces were close to the expected binding energy ( 162 eV) in a sulfur { gold
bonded system [18]. In other words, the gold surfaces were properly functionalized
with the SAM chemicals, based on the sulfur { gold chemistry.
6.2.3.2 Surface phage coverage
After the loading of the phage on both bare and surface-functionalized ME res-
onators, the surface phage coverage was measured by AFM. The AFM imaging was
performed at 23 C and 30% relative humidity using a scanning probe microscope
(Dimension 3100, Veeco Instrument, Inc., Santa Barbara, CA, USA), operated in the
tapping mode. Silicon probes with a tip radius of curvature of < 8 nm (FM probes,
NanoWorld AG, Neuch^atel, Switzerland) were used to image 2  m  2  m areas at
a resolution of 256  256 pixels.
As shown in Figs. 6.8b to 6.8e, the phage was present on the ME resonator
surfaces in the form of phage bundles (49.3 4.8 nm in lateral thickness) and can be
distinguished from the background gold deposit, based on the height pro les (Figure
6.8a is of the reference gold surface). The surface phage coverage was calculated from
the surface areas occupied by the phage to be 46.8%, 49.4%, 4.2%, and 5.2% for
(b) bare, (c) AC-, (d) ALD-, and (e) MT-functionalized ME resonators, respectively
(Fig. 6.8f). As anticipated, relatively high surface phage coverage was obtained for
the bare gold-coated ME resonator, which is attributed to the physical adsorption of
the phage. In addition, comparable surface phage coverage was obtained for the AC-
functionalized ME resonator. At a pH of above 7, the N-hydroxysuccinimide of the
chemical AC leaves the group. The resultant activated carboxyl group, thus, forms a
covalent amide bond with the N-terminus of a phage coat protein [13,19]. Even after
131
Figure 6.8: AFM images (2  m  2  m) of the SAE10 phage on bare and surface-
functionalized ME resonators: (a) reference, (b) bare gold (physical phage adsorp-
tion), (c) AC- (covalent phage attachment), (d) ALD-, and (e) MT-functionalized
ME resonators. The white lines on the photographs were the paths from which the
height pro les were measured. (f) The surface phage coverage values were 46.8%,
49.4%, 4.2%, and 5.2% for bare, AC-, ALD-, and MT-functionalized ME resonators,
respectively.
washing with the wash bu er containing Tween 20, a commonly used detergent for
blocking and cleaning of non-speci cally adsorbed proteins [20], the covalently bound
phages were found to be immobilized robustly on the resonator. In this work, a sig-
ni cant improvement in surface phage coverage was not observed for the AC-based
132
covalent phage immobilization, compared to the physical phage adsorption. Possible
reasons are: (1) similar space- lling mechanisms of both physically and covalently
adsorbed phages (i.e., the  lamentous phage probe lay on the ME resonator surfaces
upon adsorption) and (2) the e ects of washing with/without the Tween 20 deter-
gent on the quantity of the immobilized phages. In both cases, the major coat protein
pVIII ( 4,000 copies) seemed to provide the largest contribution in adsorption and
immobilization stability. To investigate the e ects of washing with the Tween 20 de-
tergent, both physcially and covalently adsorbed phages on bare or AC-functionalized
gold surfaces (1 cm  1 cm) were washed with a 100-mM HEPES bu er containing
various concentrations of Tween 20 (0 to 5% v/v) and  nally with DI water. Contact
angle changes due to the washing were, then, measured for these surfaces. The e ect
of washing was found to be large on the immobilization stability of the physically
adsorbed phages. At a Tween 20 concentration of over 0.5% v/v, dramatic changes
in water contact angle were observed for the gold surfaces with physically adsorbed
phages, whereas much small contact angle changes occurred for those with covalently
bound phages (Fig. 6.9). By contrast, a much smaller phage coverage was obtained
for ALD- and MT-functionalized ME resonators (Figs. 6.8d and 6.8e). The surface
phage coverage values were one order of magnitude smaller than those for the bare
and AC-functionalized ME resonators (Figs. 6.8b and 6.8c). The coupling of the
aldehyde-terminated SAM (ALD) and the N-terminus of a coat protein of the SAE10
phage seemed not to occur extensively under the bu er condition used in this work
(100-mM HEPES, pH 7.3). Although Peelen and Smith has reported that an imine
and a carbonyl can co-exist for an aldehyde { amine coupling reaction in aqueous
bu er, the reactivity depends on the pH and bu er compositions [21]. In fact, the best
condition for their model system was 10-mM HEPES mixed with 60-mM NaBH3CN
at pH 10, which produced a stable secondary amine product. Washing of the ME
133
Figure 6.9: Water contact angles for bare and AC-functionalized gold surfaces loaded
with/without phages. The surfaces were washed with a 100-mM HEPES bu er con-
taining various concentrations of Tween 20 (0 to 5 % v/v) and  nally with DI water.
The e ect of washing was found to be large on the immobilization stability of the
physically adsorbed phages.
resonators with DI water after the phage loading was found to cause the imine prod-
uct to be excessively hydrolyzed. Furthermore, the aldehyde- and methyl-terminated
SAMs also have been found to block cystein-based covalent phage attachment [14] to
gold-coated ME resonator surfaces. Hence, the ALD and MT-functionalized surfaces
can be described as \anti-phage surfaces" under proper bu er and sample preparation
conditions.
134
6.2.3.3 SEM observation and dose-response results
Figures 6.10a to 6.10d show SEM images for phage-immobilized ME biosen-
sors ((a) bare, (b) AC-, (c) ALD-, and (d) MT-functionalized ME biosensors, all
BSA-blocked) exposed to an SA bead suspension (3.6 1010 beads/ml), followed by
washing with  ltered DI water three times. The surface bead coverage for the bare
and AC-functionalized ME biosensors was much higher than that for the ALD- and
MT-functionalized ME biosensors, which can be explained by the large di erences
in surface phage coverage (See Figure 6.8.). Although the adsorption of the SAE10
phage onto the sensor surface is random in nature, approximately 4,000 copies of
the binding sites along the length of the phage massively bound with SA beads (as
high as 1.42  0.26 beads/ m2). This large number of binding sites is one of the
advantages of  lamentous landscape phages. In addition, non-speci c adsorption of
SA beads was greatly prevented by BSA as can be seen in Figs. 6.10e and 6.10f.
135
Figure 6.10: SEM images showing SA beads captured by the SAE10 phage on bare
and surface-functionalized ME biosensors (all BSA blocked): (a) bare, (b) AC-, (c)
ALD-, and (d) MT-functionalized ME biosensors. Non-speci c adsorption of SA
beads was greatly reduced by BSA blocking: (e) bare and (f) AC-functionalized ME
biosensors without phage.
136
Based on the SEM observation results, dose-response plots were constructed for
only the bare and AC-functionalized ME biosensors with high surface phage coverage
( 50%) in order to compare their performance. The ME biosensors were exposed to
various concentrations of SA bead suspensions (3.6  105 to 3.6  1010 beads/ml),
and the resultant resonant frequency changes were plotted in Figs. 6.11a and 6.11b.
Figure 6.11: Dose-response plots showing the comparable performance of (a) bare
and (b) AC-functionalized ME biosensors (1 mm  0.2 mm  15  m).
137
The comparable performance of the bare and AC-functionalized ME biosensors (mea-
surement sensors) indicates again that the surface phage coverage plays a crucial role
in analyte capture. In addition, the measured negligible changes in control sensor
resonant frequencies veri ed the prevention of non-speci c adsorption of SA beads on
the BSA-blocked ME biosensors.
6.2.3.4 Enhancing the detection capabilities of ME biosensors with a pat-
terned phage layer
One of the ultimate goals to be reached for phage-immobilized ME biosensors is
the detection of single pathogenic bacteria. However, the mass sensitivity of freestand-
ing, strip-shaped ME biosensors operating longitudinal-vibration modes is largely
dependent on the position of masses attached to the sensor surfaces [2]. In fact,
for non-uniform mass attachment, which may be caused by the absence of su cient
quantity of target analytes, the mass sensitivity cannot be described by Eq. 3.6
(i.e., This equation is valid only for uniform mass attachment). It has been recently
reported that for fundamental resonant frequency measurements of ME sensors, 1)
the mass sensitivity is close to zero when the mass is attached in the middle of the
sensor?s longest dimension and 2) a high mass sensitivity is, by contrast, obtained
for the mass attached at both ends of the sensor [2]. Considering this dependence is
crucial to detection of low-concentration bacterial targets, including single pathogenic
bacteria, because their local attachment may cause varying sensor responses. In a
worst-case scenario, the resultant changes in resonant frequency may be too small and
undetectable despite the use of micron-scale ME biosensors (i.e., The mass sensitivity
increases as the size of sensor is decreased.). Although measurements of higher-mode
resonant frequencies may solve the above-mentioned problem [2], the amplitude of the
higher-mode frequency peaks is generally too small to be detected for micron-scale
ME sensors, particularly in a liquid environment. Hence, when fundamental resonant
138
frequency measurements are considered, the phage layer that speci cally binds with
target bacteria may need to be patterned onto desired parts of sensor surface to en-
hance detection capabilities. One potential way to achieve this goal is to properly
functionalize the sensor surface to maximize or minimize the surface phage coverage.
Therefore, the results presented in this chapter may be useful. Since various methods
of SAM patterning (e.g., photobleaching and scanning probe-based lithography) are
readily available [17], micron-scale ME biosensors interfaced with a patterned phage
layer may enable the detection of single pathogenic bacteria in the future.
6.2.4 Conclusions
Surface phage coverage on bare and surface-functionalized ME biosensors was
quanti ed by AFM. The activated carboxyl-based covalent attachment (AC-based
covalent attachment) produced a phage coverage of  50%, comparable to that ob-
tained through physical phage adsorption. The results can be attributed to the space-
 lling nature of the  lamentous SAE10 phage upon adsorption and e ects of washing
with/without the Tween 20 detergent. By contrast, aldehyde- and methyl-terminated
surfaces (ALD- and MT-functionalized surfaces) did not yield high phage coverage.
These surface functionalization may be used to construct \anti-phage surfaces" under
proper bu er and sample preparation conditions. In addition, the large di erences in
the quantity of the captured SA beads on the di erently functionalized sensor surfaces
as well as the comparable dose-response results of the bare and AC-functionalized ME
biosensors indicated that the surface phage coverage is a key factor in analyte capture.
The results obtained here should be applicable to other fd-tet phage-based biosensing
systems. Finally, a phage probe layer can be patterned onto desired parts of the
sensor surface to enhance detection capabilities by properly functionalizing sensor
surfaces.
139
Bibliography
[1] C. A. Grimes, C. S. Mungle, K. Zeng, M. K. Jain, W. R. Dreschel, M. Paulose,
and K. G. Ong, \Wireless magnetoelastic resonance sensors: A critical review,"
Sensors, vol. 2, pp. 294 { 313, 2002.
[2] S. Li and Z.-Y. Cheng, \Nonuniform mass detection using magnetostrictive
biosensors operating under multiple harmonic resonance modes," Journal of Ap-
plied Physics, vol. 107, p. 114514, 2010.
[3] S. Huang, J. Hu, J. Wan, M. Johnson, H. Shu, and B. Chin, \The e ect of
annealing and gold deposition on the performance of magnetoelastic biosensors,"
Materials Science and Engineering: C, vol. 28, no. 3, pp. 380 { 386, 2008.
[4] C. Liang, S. Morshed, and B. C. Prorok, \Correction for longitudinal mode
vibration in thin slender beams," Applied Physics Letters, vol. 90, p. 221912,
2007.
[5] U. Rai and R. Singh, \Synthesis and mechanical characterization of polymer-
matrix composites containing calcium carbonate/white cement  ller," Materials
Letters, vol. 58, pp. 235 { 240, 2003.
[6] A. Gupta, D. Akin, and R. Bashir, \Detection of bacterial cells and antibod-
ies using surface micromachined thin silicon cantilever resonators," Journal of
Vacuum Science & Technology B, vol. 22, pp. 2785 { 2791, 2004.
[7] X. Yao, M. Jericho, D. Pink, and T. Beveridge, \Thickness and elasticity of
gram-negative murein sacculi measured by atomic force microscopy," Journal of
Bacteriology, vol. 181, pp. 6865 { 6875, 1999.
[8] V. A. Petrenko, G. P. Smith, X. Gong, and T. Quinn, \A library of organic
landscapes on  lamentous phage," Protein Engneering, vol. 9, pp. 797 { 801,
1996.
[9] G. P. Smith and V. A. Petrenko, \Phage display," Chemical Reviews, vol. 97,
pp. 391 { 410, 1997.
[10] C. Mao, A. Liu, and B. Cao, \Virus-based chemical and biological sensing,"
Angewandte Chemie International Edition, vol. 48, pp. 6790 { 6810, 2009.
[11] V. A. Petrenko, \Landscape phage as a molecular recognition interface for de-
tection devices," Microelectronics Journal, vol. 39, no. 2, pp. 202 { 207, 2008.
140
[12] O. Lazcka, F. J. D. Campo, and F. X. M. noz, \Pathogen detection: A per-
spective of traditional methods and biosensors," Biosensors and Bioelectronics,
vol. 22, no. 7, pp. 1205 { 1217, 2007.
[13] L.-M. C. Yang, J. E. Diaz, T. M. McIntire, G. A. Weiss, and R. M. Penner,
\Covalent virus layer for mass-based biosensing," Analytical Chemistry, vol. 80,
pp. 933 { 943, 2008.
[14] E. V. Olsen, I. B. Sorokulova, V. A. Petrenko, I.-H. Chen, J. M. Barbaree,
and V. J. Vodyanoy, \A nity-selected  lamentous bacteriophage as a probe for
acoustic wave biodetectors of Salmonella typhimurium," Biosensors and Bioelec-
tronics, vol. 21, pp. 1434 { 1442, 2006.
[15] A. Singh, N. Glass, M. Tolba, L. Brovko, M. Gri ths, and S. Evoy, \Immobi-
lization of bacteriophages on gold surfaces for the speci c capture of pathogens,"
Biosensors and Bioelectronics, vol. 24, pp. 3645 { 3651, 2009.
[16] G. Kuzmicheva, P. Jayanna, I. Sorokulova, and V. Petrenko, \Diversity and cen-
soring of landscape phage libraries," Protein Engineering, Design and Selection,
vol. 22, pp. 9 { 18, 2009.
[17] R. K. Smith, P. A. Lewis, and P. S. Weiss, \Patterning self-assembled monolay-
ers," Progress in Surface Science, vol. 75, pp. 1 { 68, 2004.
[18] W. Fabianowski, L. C. Coyle, B. A. Weber, R. D. Granata, D. G. Castner,
A. Sadownik, and S. L. Regen, \Spontaneous assembly of phosphatidylcholine
monolayers via chemisorption onto gold," Langmuir, vol. 5, pp. 35 { 41, 1989.
[19] P. Wagner, M. Hegner, P. Kernen, F. Zaugg, and G. Semenza, \Covalent immo-
bilization of native biomolecules onto Au(111) via N-hydroxysuccinimide ester
functionalized self-assembled monolayers for scanning probe microscopy," Bio-
physical Journal, vol. 70, pp. 2052 { 2066, 1996.
[20] Z.-H. Wang, A. S. Viana, G. Jin, and L. M. Abrantes, \Immunosensor interface
based on physical and chemical immunoglobulin G adsorption onto mixed self-
assembled monolayers," Bioelectrochemistry, vol. 69, pp. 180 { 186, 2006.
[21] D. Peelen and L. M. Smith, \Immobilization of amine-modi ed oligonucleotides
on aldehyde-terminated alkanethiol monolayers on gold," Langmuir, vol. 21,
pp. 266 { 271, 2005.
141
Chapter 7
Conclusions
In order to improve the cost-e ectiveness and sensitivity of phage-based ME
biosensors, batch fabrication of micron- to millimeter-scale sensors was performed.
Two di erent sensor fabrication methods used were:
1. the dicing-based method for millimeter-scale sensors
2. the co-sputtering-based method for micron-scale sensors.
Particularly, the co-sputtering method was found to be suitable for fabricating micron-
scale sensors with dimensional consistency. In addition, by performing batch fabrica-
tion, the cost per sensor can be reduced to a fraction of a cent.
Rapidness of testing was also improved by using two di erent methods of bacte-
rial detection as follows:
1. Direct detection of S. Typhimurium on fresh spinach leaves
2. Detection of B. anthracis spores with the aid of a designed micro uidic  ow
cell.
For the  rst method, by placing freestanding biosensors on spinach leaf surfaces,
rapid, direct detection of S. Typhimurium was realized. This method does not re-
quire any pre-test sample preparation. Hence, total assay time was only 45 min. In
addition, by characterizing the topography of leaf surfaces, the characteristic sizes of
biosensors free from the surface topography e ects were determined. Furthermore, a
formula describing the probability of detection as a function of the size and number
of biosensors and the surface density of S. Typhimurium was derived. By using the
142
formula, the required number of biosensors to obtain a desired LOD can be deter-
mined.
For the second method, PDMS micro uidic  ow cells were designed, fabricated,
and tested with 200  m-long phage-based ME biosensors for the enhanced detection
of B. anthracis spores. Due to the enhanced chances of physical contact between a
biosensor and spores, the time required for the testing was only about 10 min. In
addition, with the micron-scale ME biosensors, a small number of spores, down to
106 spores, were detected.
Additionally, to further enhance the detection capabilities of phage-based ME
biosensors, the following e ects were studied:
1. E ects of mass position on the sensitivity of ME biosensors
2. E ects of surface functionalization of ME biosensors on surface phage coverage.
The mass sensitivity of ME biosensors was found to be largely dependent on the
longitudinal mass position and dimensions of the sensors. The formula derived in this
work can be used to predict the mass sensitivity and resonant frequency change of
di erently sized ME biosensors for the detection of single pathogenic bacteria.
Surface phage coverage on di erently surface-functionalized ME biosensors was
quanti ed by atomic force microscopy. The activated carboxyl-based covalent at-
tachment produced a phage coverage of 50%, comparable to that obtained through
physical phage adsorption, a traditional way of phage immobilization. By contrast,
aldehyde- and methyl-terminated surfaces did not yield high phage coverage. These
surface functionalization may be used to construct \anti-phage surfaces" under proper
bu er and sample preparation conditions. In addition, it was found that surface phage
coverage is a key factor in analyte capture.
143
Finally, based on the above results, a concept of phage layer patterning was in-
troduced. Phage may be patterned onto desired parts of biosensor surface to enhance
detection capabilities.
144