<%BANNER%>

Random Cortical Networks Grown on Microelectrode Arrays as Cell-Based Biosensors for Neurotoxins as Demonstrated by Botu...

Permanent Link: http://ufdc.ufl.edu/UFE0021727/00001

Material Information

Title: Random Cortical Networks Grown on Microelectrode Arrays as Cell-Based Biosensors for Neurotoxins as Demonstrated by Botulinum Toxin Type A
Physical Description: 1 online resource (113 p.)
Language: english
Creator: Scarlatos, Amber C
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: biosensor, botulism, cell, clostridium, microelectrode, mouse
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This work presents a potential biosensor for neurotoxin detection that employs living neural tissue cultures grown in-vitro on microelectrode arrays (MEAs). An MEA is a tissue culture dish with a grid of electrodes embedded in the surface of the dish, enabling extracellular recording of action potentials produced by neural cultures grown over the array. Botulism toxin (BoNT) is an extremely potent neurotoxin produced by toxigenic strains of Clostridium botulinum. Botulism toxin continues to pose a major threat since it may be employed in a deliberate attack on the US food supply. Furthermore, BoNT may be liberated in any insufficiently processed food containing a reduced oxygen atmosphere. Hence, rapid and reliable detection of BoNTs in foods is necessary to reduce the risk posed through food contamination. In this study, a bolus of 200 to 1000 pg of pharmaceutical grade BoNT A was applied to the media bath of mature cortical networks cultured on microelectrode arrays. It was hypothesized that a detectable change in burst activity would be observed of treatment of the cortical cultures with BoNT A. The observed change in burst activity would enable reliable detection of the toxin as the basis for cell-based BoNT biosensor. After treatment of the tissue cultures with BoNT A, both spontaneous and evoked activity were then monitored over a one week period to quantify changes in the neural population produced by BoNT A. Introduction of BoNT A resulted in an increase in the duration and number of spikes in the spontaneous and evoked bursts relative to control cultures. Application of BoNT A also induced unique oscillatory behavior within each burst that is reminiscent of early developmental activity patterns rather than the mature cultures used here. Additionally, changes in connectivity during evoked bursts quantified using Granger Causality (GC), a tool borrowed from economics. Neural cultures dosed with botulism toxin were found to have an overall decrease in GC values indicating a decrease in effective connectivity. Based upon the observed changes, an algorithm for BoNT detection was developed which enabled detection of BoNT A in 13 of 15 neural cultures by 48h and all 15 neural cultures by 72h. No control dishes were positive for BoNT A up to 72h after dosage with BoNT A carrier solution. Thus the activity of these cortical networks measured with MEAs could provide a valuable substrate for BoNT detection.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Amber C Scarlatos.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Welt, Bruce A.
Local: Co-adviser: Cooper, Brian Y.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021727:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021727/00001

Material Information

Title: Random Cortical Networks Grown on Microelectrode Arrays as Cell-Based Biosensors for Neurotoxins as Demonstrated by Botulinum Toxin Type A
Physical Description: 1 online resource (113 p.)
Language: english
Creator: Scarlatos, Amber C
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: biosensor, botulism, cell, clostridium, microelectrode, mouse
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This work presents a potential biosensor for neurotoxin detection that employs living neural tissue cultures grown in-vitro on microelectrode arrays (MEAs). An MEA is a tissue culture dish with a grid of electrodes embedded in the surface of the dish, enabling extracellular recording of action potentials produced by neural cultures grown over the array. Botulism toxin (BoNT) is an extremely potent neurotoxin produced by toxigenic strains of Clostridium botulinum. Botulism toxin continues to pose a major threat since it may be employed in a deliberate attack on the US food supply. Furthermore, BoNT may be liberated in any insufficiently processed food containing a reduced oxygen atmosphere. Hence, rapid and reliable detection of BoNTs in foods is necessary to reduce the risk posed through food contamination. In this study, a bolus of 200 to 1000 pg of pharmaceutical grade BoNT A was applied to the media bath of mature cortical networks cultured on microelectrode arrays. It was hypothesized that a detectable change in burst activity would be observed of treatment of the cortical cultures with BoNT A. The observed change in burst activity would enable reliable detection of the toxin as the basis for cell-based BoNT biosensor. After treatment of the tissue cultures with BoNT A, both spontaneous and evoked activity were then monitored over a one week period to quantify changes in the neural population produced by BoNT A. Introduction of BoNT A resulted in an increase in the duration and number of spikes in the spontaneous and evoked bursts relative to control cultures. Application of BoNT A also induced unique oscillatory behavior within each burst that is reminiscent of early developmental activity patterns rather than the mature cultures used here. Additionally, changes in connectivity during evoked bursts quantified using Granger Causality (GC), a tool borrowed from economics. Neural cultures dosed with botulism toxin were found to have an overall decrease in GC values indicating a decrease in effective connectivity. Based upon the observed changes, an algorithm for BoNT detection was developed which enabled detection of BoNT A in 13 of 15 neural cultures by 48h and all 15 neural cultures by 72h. No control dishes were positive for BoNT A up to 72h after dosage with BoNT A carrier solution. Thus the activity of these cortical networks measured with MEAs could provide a valuable substrate for BoNT detection.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Amber C Scarlatos.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Welt, Bruce A.
Local: Co-adviser: Cooper, Brian Y.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021727:00001


This item has the following downloads:


Full Text





RANDOM CORTICAL NETWORKS GROWN ON MICROELECTRODE
ARRAYS AS CELL-BASED BIOSENSORS FOR NEUROTOXINS
AS DEMONSTRATED BY BOTULINUM TOXIN TYPE A





















By

AMBER CHRISTINE SCARLATOS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2007

































2007 Amber Christine Scarlatos

































This work is dedicated to my husband, Chris Scarlatos, my son, Nathan Scarlatos, and my
parents Allen and Arleen Austin









ACKNOWLEDGMENTS

I thank Dr Bruce Welt for his enthusiasm, patience and support. I thank Dr. Thomas

DeMarse who kindly guided me during my many adventures in the Neural Computation and

Robotics Laboratory. I thank Dr. Brian Cooper for introducing me to the expansive field of

electrophysiology and for keeping me (mostly) sane during the early years of my PhD work. I

thank Dr. Archer and Dr. Chau for serving on my committee and for helping to guide me through

my Ph.D. work. I thank Alex Cadotte who shared my enthusiasm and frustrations with cell

culture and guided me in application of Granger's Causality analysis to data generated on

multielectrode arrays.

Additionally, I thank my friends and family for their ongoing support. I would especially

like to thank my husband for patiently waiting four more years, and my son for giving me a

unique perspective.









TABLE OF CONTENTS

A CK N O W LED G M EN T S ...................................................... ..............................................

T A B L E O F C O N T E N T S ........................................................ .........................................5

L IST O F T A B L E S ...................................... ..................................... ................ .. 8

LIST OF FIGURES .................................. .. .... ..... ................. 10

L IST O F A B B R E V IA T IO N S ........................................................... ....... .............................. 11

A B S T R A C T ......... ....................... ............................................................ 13

CHAPTER

1 INTRODUCTION AND BACKGROUND ........................................ ....................... 15

In tro d u ctio n ................... ...................1...................5..........
O b j e ctiv e ................... ...................1...................6..........
B ack g rou n d ............................ .................................................................................................16
Electrophysiology of N euronal Tissue .................................. .............................. ...... 16
B otulism T oxin ....................................................................... ................ 20
BoNT Action on the Central Nervous System ............................................ ............... 25

2 REVIEW OF BOTULINUM TOXIN DETECTION METHODS .............................27

In tro d u ctio n ................... ...................2...................7..........
M ou se B ioassay .............. .. .....................................................................................2 7
Enzym e Linked Im munosorbent Assay........................................................ ...............29
Polym erase Chain R action A ssay ............................... ......... ......... ..............................31
Protease A activity A says .................................. .. .......... .. ............36
M ass Spectrom etry ..................................... ... .. .... ........ ........ ........ 38
D isc u ssio n ................... .............................................................. ................4 0

3 NEURAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELL-
B A SE D B IO SE N SO R S ............................ .................................................. ......................43

Introduction .................................. .......... .............................................. 43
Hardware and Software .................................... ...... ....... ........... ..... 45
M icro-electrode arrays (M EA) ......................................................... ............... 45
D ata acquisition ..................................... ................................ ........... 46
Physiological conditions........... ............................................................ .. .... .... ..... 46
D associated N eural C cultures ......................................................................... ....................47
Tissue Culture M methods ............................................. ...... ............. 47
D evelopm ent and activity .......... ................................................................. 48
N etw ork B u resting ........................... ................................................................................. 50
M EAs as Biosensors ...................................................................... ........ 51









4 DEFINING THE BONT A FINGERPRINT.............................................. ............... 52

In tro d u ctio n ................... ...................5...................2..........
M materials and M methods ........................ .. ........................ .. .... ........ ........ 52
P ro c e d u re ................................................................................................................... 5 2
D rug application .................................................................... 53
A analysis and statistics............................................. 53
R e su lts ............... ......... ......... ............................................................................................... 5 4
Spontaneous activity................................................... 54
Intra-burst neural dynam ics ................................................................. .. .. ........ 56
Evoked response following intoxication with BoNT A ..................................... 58
Discussion ....................................... ............... 62

5 REFINING THE BONT A FINGERPRINT AND DEVELOPMENT OF A BONT
DETECTION ALGORITHM .................. .................. .......... .. ......................... 64

In tro du ctio n ................... ...................6...................4..........
B ack g rou n d ................... ...................6...................4..........
P la stic ity ..............................................................6 4
G ranger's C au sality ..................................................................................... .65
M methods ............................................ ......... .....................67
E x p erim mental p roto col ............................................................................................... 6 7
Plasticity protocol ..... ........... ......... .................... 67
R e su lts an d D iscu ssio n ..................................................................................................... 6 8
S p o n tan eou s activ ity .................................................................................................. 6 8
E limited activity .................................................................................................... 68
P la stic ity ................................................................7 0
P G C A n aly sis ............................................................72
B oN T D election A lgorithm ....................................................................................... 72

6 CONCLUSION AND FUTURE WORK ....................................................... 77

Future W ork .................................................. 77
Reduction of detection tim e....................................................... 77
Serotype differentiation ............................................................ .................... 78
Use of neural networks with crude samples ..........................................78
A alternative and im m ortal cell lines ....................................................... 79
Conclusion ......................................... .. .............. ............... 79

APPENDIX

A SPONTANEOUS ACTIVITY IN BONT A AND CARRIER DOSED CULTURES ........... 81

B ELLICITED ACTIVITY IN BONT A AND CARRIER DOSED CULTURES ...................87

C BURST DETECTION ALGORITH M .............. .........................................................91

D BONT A DETECTION ALGORITHM ............ ..... ........ .........95


6









R E F E R E N C E S .................................................................................................................. 10 5

BIOGRAPHICAL SKETCH ................................................................... ........... 113









LIST OF TABLES


Table page

2-1 Immunoassays for the detection of BoNTs................................. ......................... 30

2-2 PCR primers and DNA probes for detection of C. botulinum and toxigenic C.
buty ricum an d C bara ti ......................................................................... ...................... 32

5-1 Identification of BoNT in unknown cultures ........................................ ...............76

A-i Average spike height for neural cultures dosed with BoNT carrier or 200pg BoNT A ...81

A-2 Average spike width for neural cultures dosed with BoNT carrier or 200 pg BoNT A....81

A-3 Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg
B oN T A .......................................................................................... 8 1

A-5 Average inter-burst interval for neural cultures dosed with BoNT carrier or 200 pg
B oN T A ........................................................................................... 8 2

A-6 Average number of spikes per burst for neural cultures dosed with BoNT carrier or
200 pg B oN T A ............................................................................82

A-7 Average spike height for neural cultures dosed with BoNT carrier or 600pg BoNT A ...82

A-8 Average spike width for neural cultures dosed with BoNT carrier or 600pg BoNT A.....83

A-9 Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg
B oN T A ........................................................................................... 8 3

A-10 Average burst duration for neural cultures dosed with BoNT carrier or 600pg BoNT
A ............................................................ .................................. . 8 3

A-11 Average inter-burst interval for neural cultures dosed with BoNT carrier or 600 pg
B oN T A ........................................................................................... 8 4

A-12 Average number of spikes per burst for neural cultures dosed with BoNT carrier or
600 pg B oN T A ............................................................................84

A-13 Average number of channels per burst for neural cultures dosed with BoNT carrier or
600 pg B oN T A ............................................................................84

A-14 Average burst duration for neural cultures dosed with BoNT carrier or 1000 pg
B oN T A ........................................................................................... 8 5

A-15 Average inter-burst interval for neural cultures dosed with BoNT carrier or 1000 pg
B oN T A ........................................................................................... 8 5









A-16 Average number of spikes per burst for neural cultures dosed with BoNT carrier or
1000 pg B oN T A ...........................................................................85

A-17 Average number of channels per burst for neural cultures dosed with BoNT carrier or
1000 pg B oN T A ........................................................................ ... ..... .. 86

B-l Average spikes per stimulation for neural cultures dosed with BoNT carrier or
2 0 0p g B oN T A ......................................................... ................ 87

B-2 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
2 0 0p g B oN T A ......................................................... ................ 87

B-3 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 200pg BoNT A .........................87

B-4 Percent change in the number of spikes 150 to 300 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 200pg BoNT A .........................88

B-5 Average spikes per stimulation for neural cultures dosed with BoNT carrier or 600
p g B o N T A ............................................................................ 8 8

B-6 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
600 pg B oN T A ............................................................................88

B-7 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 600pg BoNT A .........................89

B-8 Percent change in the number of spikes 150 to 300 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 600pg BoNT A.........................89

B-9 Average spikes per stimulation for neural cultures dosed with BoNT carrier or 1000
p g B o N T A ............................................................................ 8 9

B-10 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
1000 pg B oN T A ........................................................................ ... ..... .. 90

B-11 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A...................90

B-12 Percent change in the number of spikes 100 to 300 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A...................90









LIST OF FIGURES


Figure p e

1-1 Botulism toxin cleaves SNARE proteins necessary for vesicle fusion docking ...............26

2-1 Botulinum B micromechanosensor................. .......... ............................ 38

3-1 A micro-electrode array (MEA) from Multichannel Systems.......................................45

3-2 D ata acquisition hardw are.............................................. .................... ............... 47

3-3 Screen shot of typical raw activity during a network-wide burst ................................49

4-1 Spontaneous burst activity in cortical networks after treatment with BoNT A or toxin
c a rrie r ................... ............................................................ ................ 5 5

4-2 Return plots generated by sorting spikes into 100 ms bins and plotting bins at time t
against bins at tim e t 50 m s.................................................. ............................... 57

4-3 Stimulation-response for cortical culture grown on MEA 9053................... ..............59

4-4 Mean change in the number of evoked spikes 100 to 300 ms post stimulation in the
B oN T and control cultures............................................ ................... ............... 60

4-5 Average stimulation response for Control (Panel A) and BoNT dosed (Panel B)
tissue cultures ............................................................... ..... ...... ........ 61

5-1 Percent change in spikes with respect to baseline activity after treatment with 600 pg
B oN T A (A ) and 1000 pg B oN T A (B ) ......................................................................... ...69

5-2 N etw ork plasticity induced by tetanus ...................................................... .............. 71

5-3 Granger's Causality difference plot after dosage with BoNT A carrier (A), and 1000
pg B oN T A (B ) ............................................................................73

5-4 Average stimulation response (blue) and smoothed stimulation response (red) for
control dish 7079 (A D) and dish 9299 after dosage with 200 pg BoNT A (E H)........75









LIST OF ABBREVIATIONS


APRS:

AR:

BoNT:

BHI:

CNS:

DIV

DMEM

ELISA:

ES-MS:

Fab:

GalNAc-T--

GABA:

GC:

HA:

LC-ES-MS:

LD50:

LTD:

LTP:

MALDI-MS:

MEA:

MLD:

mAbs

NTNH:

pAbs:


array of presynaptic receptors

autoregressive

botulinum toxin

Brain Heart Infusion

central nervous system

days in vitro

Dulbucco's Modified Eagles Medium

enzyme-linked immunosorbent assay

electron spray mass spectrometry

antigen binding antibody fragment

GalNAc transferase negative

y-Aminobutyric acid

Granger's Causality

heamagluttins

liquid chromatography and electrospray mass spectrometry

lethal dose for 50 % of the population

long term depression

long term potentiation

matrix assisted laser desorption ionization mass spectrometry

multielectrode array

minimum leathal dose

monoclonal antibodies

nontoxic non-heamagluttins

polyclonal antibodies









PCR: polymerase chain reaction

PGC: Pairwise Granger's Causality

RT-PCR: reverse transcriptase polymerase chain reaction

SNARE soluble NSF attachment receptor

SNAP-25 synaptosome-associated protein of 25,000 daltons

SV2 synaptic vesicle protein 2

SYT: synaptogamin

TYP-C: tryptone-peptone-yeast extract

VAMP: synaptobrevin









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

RANDOM CORTICAL NETWORKS GROWN ON MICROELECTRODE
ARRAYS AS CELL-BASED BIOSENSORS FOR NEUROTOXINS
AS DEMONSTRATED BY BOTULINUM TOXIN TYPE A



By

Amber C Scarlatos

December 2007

Chair: Bruce Welt
Cochair: Brian Cooper
Major: Agricultural and Biological Engineering

This work presents a potential biosensor for neurotoxin detection that employs living

neural tissue cultures grown in-vitro on microelectrode arrays (MEAs). An MEA is a tissue

culture dish with a grid of electrodes embedded in the surface of the dish, enabling extracellular

recording of action potentials produced by neural cultures grown over the array. Botulism toxin

(BoNT) is an extremely potent neurotoxin produced by toxigenic strains of Clostridium

botulinum. Botulism toxin continues to pose a major threat since it may be employed in a

deliberate attack on the US food supply. Furthermore, BoNT may be liberated in any

insufficiently processed food containing a reduced oxygen atmosphere. Hence, rapid and reliable

detection of BoNTs in foods is necessary to reduce the risk posed through food contamination. In

this study, a bolus of 200 to 1000 pg of pharmaceutical grade BoNT A was applied to the media

bath of mature cortical networks cultured on microelectrode arrays. It was hypothesized that a

detectable change in burst activity would be observed of treatment of the cortical cultures with

BoNT A. The observed change in burst activity would enable reliable detection of the toxin as

the basis for cell-based BoNT biosensor.









After treatment of the tissue cultures with BoNT A, both spontaneous and evoked activity

were then monitored over a one week period to quantify changes in the neural population

produced by BoNT A. Introduction of BoNT A resulted in an increase in the duration and

number of spikes in the spontaneous and evoked bursts relative to control cultures. Application

of BoNT A also induced unique oscillatory behavior within each burst that is reminiscent of

early developmental activity patterns rather than the mature cultures used here. Additionally,

changes in connectivity during evoked bursts quantified using Granger Causality (GC), a tool

borrowed from economics. Neural cultures dosed with botulism toxin were found to have an

overall decrease in GC values indicating a decrease in effective connectivity.

Based upon the observed changes, an algorithm for BoNT detection was developed which

enabled detection of BoNT A in 13 of 15 neural cultures by 48h and all 15 neural cultures by

72h. No control dishes were positive for BoNT A up to 72h after dosage with BoNT A carrier

solution. Thus the activity of these cortical networks measured with MEAs could provide a

valuable substrate for BoNT detection.











CHAPTER 1
INTRODUCTION AND BACKGROUND

Introduction

Efficient detection of food-borne toxins is necessary to secure the food supply from both

intentional and unintentional contamination. Botulism toxin, associated with food-borne

botulism, is a potent neurotoxin that may be liberated in any unprocessed or insufficiently

processed food containing an oxygen deficient atmosphere. Furthermore, BoNTs may be used as

a bioweapon in a deliberate attack on the US food supply. Currently, seven antigenic strains of

botulism toxin (A-G) have been identified, which are produced by the anaerobic bacterium

Clostridium botulinum. Botulinal neurotoxin is extremely toxic, with an estimated oral lethal

dose of as little as 70 pg of botulism toxin type A (BoNT A) for a 70 kg human (Amon et al.,

2001).

Containment of a botulism outbreak is dependant upon rapid identification of the toxin in

suspect foods. While there are many assays available for detecting specific toxins, toxin

fragments, and DNA of causative organisms, the FDA continues to primarily rely on the "mouse

bioassay" for the detection of botulinum toxicity in foods (Scarlatos et al., 2005). The mouse

bioassay detects the presence of toxins by observing death rates of animals injected with suspect

food samples. Identification of toxin serotypes, determined through neutralization assays using

equine antiserum, require additional time and animal use. The mouse bioassay is extremely

sensitive, detecting as little as a few picograms of toxin, and capable of identifying the presence

of virtually any type of biologically active toxin (Aron et al., 2001). It is also time consuming,

prohibitively expensive due to the large number of animals required, and available at only a

limited number of facilities nationwide. Alternative assays tend to focus on specific toxins and









may provide false positive indications when fragments of inactive toxin proteins or DNA are

present (Scarlatos et al., 2005).

Objective

It was hypothesized that addition of BoNT A to the cortical networks grown on MEAs

would produce a discernable change in the burst dynamics. This change in network activity

would be unique to previously reported chemically-induced changes in network activity,

providing a unique toxin fingerprint for BoNT A.

The objective of this work is to characterize activity changes resulting from dosage of

random cortical neuron networks grown in vivo with botulinum toxin, and to develop reliable

methods for discerning BoNT toxin from native activity. This work provides the groundwork for

development of a cell-based BoNT biosensor.

Background

Electrophysiology of Neuronal Tissue

Membrane biology. A fundamental characteristic of all cells is an outer membrane, which

defines the intracellular and extracellular environment. The cell membrane has many remarkable

properties which enable cell function. Cell membranes are flexible, self sealing and have limited

permeability. The primary component of cell membranes is a phospholipid, which is comprised

of a hydrophilic head along with two long hydrophobic fatty-acid tails. Within the cell

membrane the phospholipids form a lipid bilayer, where the hydrophilic heads of the

phospholipids are located on the outermost and innermost surface of the cell membrane, while

the hydrophobic tails are located within the bilayer core. This structure enables the cell

membrane to be passively impermeable to most polar and charged solutes. However,

permeability of membranes is not simply a passive property, since an array of important protein









structures reside within the cell membrane. These proteins play an important role in membrane

function.

Because cell membranes are impermeable to ions, intracellular solute concentrations are

different from extracellular concentrations. In particular, ions cannot passively diffuse across the

cell membrane, and thus exist at different concentrations within intracellular and extracellular

spaces. Since ions have an associated electrical charge, a potential difference across cell

membranes results from the differential concentration of ions. This potential difference is an

important aspect of cell membranes, and is actively maintained through pumps which exchange

ions from one side of the membrane to the other. When a net ion flux across a cell's membrane is

at zero, due to balanced electrical and chemical forces, the membrane potential is constant and is

known as the resting potential. The equilibrium potential for a single ion can be described by the

Nernst Equation.

RT [ion]
E= n (1-1)
zF [ion],

However, the resting potential is dependant on several ions. The most common ions

found within animal fluids are sodium (Na+), potassium (K+), Calcium (Ca2+), Magnesium

(Mg2+), Chlorine (C1-), and Bicarbonate (HCO3-). To obtain the resting potential when multiple

ion species are relevant the Goldman-Hodgkin-Katz equation can be used, where P is the

permeability of an ion.

N
Y [ionj ] P
SRT t
E= Inc r (1-2)

J-1

Sodium and Potassium ions have the most striking contrast between extracellular and

intracellular concentrations. In mammalian tissue, [K ]iz 140 mM while [K],o ] 4 mM.









Conversely, [Na ]i z 12 mM while, [Na+]o ] 145 mM (Nelson and Cox, 2005). The large

sodium and potassium gradients are primarily due to active transport of sodium and potassium

ions by an integral membrane transport protein known as Na+K+ ATPase. This protein uses

energy from ATP hydrolysis to pump three Na+ out of the cell, and two K ions into the cell.

Neuronal tissue has a resting potential around -70 mV which is close to the K+ resting potential.

The action potential. Cells can be classified as excitable or inexcitable based upon how

they react to a depolarizing current. A depolarizing current applied to the membrane of an

unexcitable cell results in a proportional change in the membrane potential. In this instance the

membrane mimics an RC-circuit. Excitable cells have a similar RC-response to sufficiently small

depolarizing currents. However, when depolarizing currents exceed a threshold, a large

disproportionate change in membrane potential occurs, known as the action potential. Animals

have a variety of excitable cells including neuronal, and muscular cells.

The action potential is an electrochemical disturbance generated within excitable cells and

is the basic unit of neural communication. The action potential is characterized by three phases:

the excitation phase, the recovery phase and a refractory period. These phases are a result of

voltage gated sodium and potassium channels in the membrane. A depolarizing current that

exceeds the excitability threshold causes voltage-gated sodium channels to open. This causes a

rapid influx of sodium ions, further depolarizing the membrane potential. Time dependant

inactivation of voltage gated sodium channels causes sodium channels to temporarily enter a

state from which they cannot reopen. At the same time, voltage gated activation of potassium

channels occurs, and potassium ions flow outward to repolarize the cell membrane back towards

the resting potential. Potassium channels are voltage dependant and can be time dependant as

well. As the voltage returns to the resting potential, it often hyperpolarizes the membrane due, in









part, to delayed closing of potassium channels. After the voltage gated sodium channels have

closed there is a period in which they cannot be activated by excitable input no matter how large.

This is known as the refractory period, which is typically around one millisecond. The Na+K+

ATPase actively pumps sodium out and potassium back into the cell to return intracellular ion

concentrations to physiological levels. Action potentials are an all or nothing event. They can be

generated through application of an external stimulus, or may occur spontaneously due to

fluctuations from the resting potential. Depolarization of the membrane by an action potential

can activate a variety of cellular processes including activation of additional voltage gated

channels, such as voltage gated calcium channels. Voltage gated calcium channels play an

essential role in cell to cell communication at the chemical synapse as described below.

Neuronal Networks. Typically, neurons are comprised of three basic components: the

soma, the dendritic tree and the axon. The soma is a cell body where the cell nucleus and the

majority of cellular organelles reside. The dendritic tree is a large branched portion of a neuron,

and is variable in shape and morphology. The dendrite is typically close to the cell body and

primarily receives input from other neurons though chemical synapses. Neuronal axons typical

consist of a long myelinated process that ends at a synapse. The axon primarily communicates

with target neurons through neurotransmission at chemical synapses. For example, at the

neuromuscular junction, the primary target of BoNTs, release of acetylcholine from chemical

synapses results in muscle contraction.

Within the body, neurons do not behave as an independent unit, but rather act as a subunit

of a larger network of thousands of neurons. Communication between neurons occurs at

synapses. When an action potential reaches the axon terminal of a chemical synapse, voltage-

gated calcium channels are activated. This initiates a rapid influx of calcium, which in turn,









enables membrane fusion of neurotransmitter containing vesicles. Neurotransmitter is released

into the synaptic cleft, where it diffuses across to postsynaptic receptors, binds to appropriate

postsynaptic proteins to depolarize (or sometimes hyperpolarize) the target neuron. Excitatory

input at a single synapse may be insufficient to generate an action potential that will propagate

from the dendritic tree to the axon terminal. Typically, excitatory inputs from multiple synapses

must be integrated in order to generate an excitatory input which exceeds the threshold necessary

to generate an action potential.

A sufficiently large excitatory input in neuronal networks comprised purely of excitatory

cells would simply cascade along the network triggering all downstream cells. This would result

in a domino effect, where the entire network would be activated and small perturbations would

be rapidly amplified. Fortunately, neuronal networks are not purely excitatory. Rather a small

percentage of neurons contain inhibitory synapses that suppress activity in target neurons. This

drastically increases the complexity of activity within neuronal networks, enabling neurons to

perform the many complex tasks necessary within the brain and nervous system.

Botulism Toxin

The seven serotypes of botulinum toxin (type A-G) are responsible for three naturally

occurring forms of botulism: food-borne botulism, wound botulism, and infant botulism. Due to

weaponization of botulinum toxins, a fourth form of botulism has been identified as inhalation

botulism. The seven serotypes are produced by the bacteria species Clostridium botulinum,

Clostridium baratti (type E), and Clostridium butyrium (type F). Serotypes A, B, and E account

for almost all cases of human botulism (Simpson 1996). Botulism is characterized by

descending flaccid muscle paralysis, beginning with cranial nerves. Current treatment of

botulism consists of supportive care and passive immunization (Arnon et al., 2001). Equine

derived polyvalent antitoxin is available for treatment of clinically diagnosed botulism cases.









Since therapeutic antitoxin neutralizes toxin molecules not already bound to neural tissue, it is

most effective when administered during early onset of disease (Caya et al., 2004). Complete

recovery can take weeks to months and often requires extensive medical support including

mechanical ventilation (Solomon and Lilly 2001).

Botulism toxin (BoNT) is a potent neurotoxin that targets peripheral cholinergenic

synapses primarily at the neuromuscular junction. The toxin internalizes and enzymatically

blocks neurotransmitter release resulting in flaccid paralysis. Synaptic transmission is blocked

through a multi-step mechanism involving binding, internalization and intracellular cleavage of

soluble NSF attachment receptor (SNARE) proteins necessary for vesicle fusion docking and

subsequent neurotransmitter release into the synaptic cleft.

The toxin is a 150 kDa metaloprotease comprised of a 100 kDa heavy chain and a 50 kDa

light chain linked by a disulfide bond (Simpson, 1996). The light chain includes the enzymatic

portion of the toxin, which cleaves SNARE proteins necessary for vesicle fusion docking. The

heavy chain has been identified as relevant in toxin binding and internalization, although the

light chain may play an important role as well. The toxin may be part of up to a 900 kDa

complex, formed with a family of haemagluttins (HAs) and notoxic non-haemagluttins (NTNHs)

that play a role in targeting the toxin to neuromuscular junctions, and have been shown to

increase toxin potency (Ahnert-Hilger and Bignalke 1995). However, these auxiliary proteins are

not essential for intoxication.

Binding. Both proteins and gangliosides have been proposed as BoNT receptors.

Gangliosides were initially proposed as receptors for BoNTs and a multitude of studies have

supported the theory that gangliosides play an important role in binding BoNTs to the

presynaptic cleft. Polysialylated ganglioside GT1b has been shown to inhibit BoNT A binding to









synaptosomes, to quench BoNT A fluorescence, and to bind to immobilized BoNT A.

Furthermore, monoclonal antibody which bound to GT1b was shown to antagonize BoNT A

binding in rat superior cervical ganglion cells (Yowler et al., 2002). In a study by Bullens et al

(2002), presynaptic ACh release was studied in wild-type and GalNAc-transferase negative

(GalNAc-T-/-) mice was studied. The GalNAc-T-/- mice lacked complex gangliosides at the

neuromuscular junction, but exhibited unregulated expression of simple gangliosides GD3 and

GM3. Wild-type mice exhibited sensitivity to BoNT A, however, GalNAc-T-- mice lacked toxin

sensitivity. The specificity of BoNTs for neuronal membranes may be due to high concentrations

of complex gangliosides found there. Crystollographic studies of BoNT A and B have revealed a

lactose binding pocket located near the c-terminus of the BoNT heavy chain (Montecucco et al.,

2004). This site contains a lactose-binding peptide motif that is highly conserved among the

seven BoNTs. Rummel et al.(2004b) used mass spectrometry to demonstrate that the lactose

binding site, bound to a single molecule of ganglioside GT1b. Mutations of this binding site

abolished the GT1b heavy-chain complex. Furthermore, GD1b and GT1b were identified as

functional receptors for BoNT C, while phosphatidylethanolamine was reported as a receptor for

BoNT D (Verderio et al., 2006). However, in many cases, binding affinity of the neurotoxin to

gangliosides under physiological conditions is low (Nishiki et al., 1994). Additionally, an

ultrastructural audioradiographic study of BoNT A an B binding sites on motor nerves indicates

that binding sites are unique to each serotype (Black and Dolly 1986; Simpson 2004). Thus, it is

unlikely that GT1b is the sole receptor for BoNT A and B.

Proteins have also been proposed as a receptor for BoNT. Nishiki et al. (1994) first

identified synaptogamin, a synaptic vesicle membrane associated protein as a high affinity

receptor for BoNT B. Labeled BoNT B bound to protein from a crude synapsomal fraction in the









presence of gangliosides in this study. No binding was observed in preparations containing only

gangliosides. Bound extract was then purfied by DEAE-Sepharose chromatography, Phenyl-

Toyopearl chromatography, and Heparin-Toyopearl chromatography. Binding affinity of the

heprin elutate closely matched that of the synapsome. However, antibodies raised against the

elutate were unable to block binding of labeled BoNT B to synapsomes. In a later study, Nishiki,

demonstrated both synapotogamin I and II bound BoNT B, but with different binding affinities

(1996). Additionally, binding of BoNT B to synaptosomes was inhibited using a monoclonal

antibody specific for the N-terminus of synaptogamin II. This indicated that BoNT B associates

with synaptogamin within 20 amino acids of the N-terminus. Similarly synaptogamins I and II

were found to be receptors for BoNT G (Rummel et al., 2004a). Recently, synaptic vesicle

protein 2 (SV2) was found to be the protein receptor for BoNT A (Dong et al., 2006). To date no

protein receptor has been identified for the remaining toxin serotypes. This was established using

SV2A/B knockout mice, in which BoNT A binding was abolished, while SV2B knock-down

mice displayed a reduced sensitivity to BoNT A.

The picture that is emerging is that binding of BoNTs at the neuromuscular junction is a

much more complicated process than that of a single high affinity receptor. Many researchers

have concluded that binding ofBoNTs involves gangliosides as an initial low affinity receptor

and a protein as a high affinity receptor. Toxin may first bind weakly with low-affinity receptors,

which induces conformational changes that then bring the toxin in contact with a high affinity

receptor. Alternatively, Montecucco et al. (2004) proposed that BoNT receptors are comprised

of an array of presynaptic receptors (APRs) including polysialogangliosides, other lipids, and

glycophosphoinostol anchored proteins. In this mechanism, the toxin is first captured by the

oligosaccharide portion of the polysialogangliosides, which extends from the membrane surface.









Because the polysialoganglioside is latterly motile it can then rapidly bring the toxin to the APR.

The association constant of the toxin would be the product of the association constant of each

interaction within the APR, thus a large association constant would result even when weak

interactions are involved.

Internalization. Toxin internalization may be due to either receptor mediated

endocytosis or the vesicle recycling pathway. While receptor mediated endocytosis was widely

assumed as the mechanism by which the toxin was internalized, recent evidence of vesicle

associated proteins as high affinity receptors for BoNT suggests it is likely that toxins utilize the

vesicle retrieval pathway to internalize. This is further supported by studies that demonstrate that

internalization of toxins is activity dependant, and that strong depolarization, such as in an

elevated potassium solution, drastically increases neurotoxin uptake (Verderio et al., 2006).

Regardless of the mechanism by which endocytosis occurs; the catalytic domain must

completely dissociate from the heavy chain, and enter the cytosol prior to cleavage of relevant

substrates. The mechanism by which BoNTs escape the lumen of the endosome is primarily

achieved through pH induced translocation. This was demonstrated through a series of

experiments with drugs known to neutralize endosomal pH. When these drugs were applied to

neuronal preparations, uptake of BoNTs was blocked (Simpson, 1983; Simpson, 2004). Thus, the

toxins were shown to contain hydrophobic domains that are exposed when the pH is low. These

hydrophobic domains aid in translocation to the cytosol. Two mechanisms for the pH-induced

translocation have been proposed. The first involves formation of a channel by the heavy chain

N-terminus. This theory is based upon an experiment in which an artificial lipid bilayer was

exposed to the heavy chain N-terminus of BoNT B, and changes in membrane conductance were

measured (Hoch et al., 1985). It was determined that addition of toxin when the pH of solutions









on both sides of the membrane was balanced, produced only a small change in membrane

conductance. However, a large change in membrane conductance was produced when a pH

gradient similar to that of an acidified endosomal compartment was introduced, indicating

channel formation. Channels were not formed by intact BoNT B. A second model has been

proposed in which the heavy chain forms a cleft that enables passage of the light chain to the

cytosol (Simpson 2004). In this case, changes in membrane conductance may be explained as an

after effect of residual changes in the membrane.

Ultimately, translocation of toxin from the endosomal compartment involves a pH-

induced change in toxin structure, translocation of the light chain to the cytosol, reduction of the

disulfide bond that links the heavy and light chain, uncoupling of all non-covalent forces

between the heavy and light chains, and restoration of the light chain to its hydrophobic

structure. Additionally, pH induced changes in structure may result in changes to the zinc-

binding motif, which must be restored once the light chain has entered the cytosol.

Cleavage. The light-chain is a zinc-dependant metaloprotease responsible for cleavage of

SNARE proteins that are essential for vesicle fusion docking. Each toxin serotype cleaves one of

three SNARE proteins at a specific site. BoNT B, D, F and G cleave synaptobrevin (VAMP), a

vesicle associated protein. BoNT A, C and E cleave SNAP-25 and BoNT Cl cleaves syntaxin.

Substrate cleavage can only occur prior to SNARE complex formation. Figure 1-1 shows the

cleavage sites for the seven BoNT serotypes.

BoNT Action on the Central Nervous System

Botulinum neurotoxins primarily target peripheral nerve terminals since they do not cross

the blood brain barrier. However, central nervous system (CNS) tissue is susceptible to BoNT

intoxication. Synaptic vesicle protein 2 (SV2), the protein receptor for BoNT A, is expressed in

both excitatory and inhibitory neurons of the hippocampus and neocortex (Verderio et al., 2006).

























Figure 1-1. Botulism toxin cleaves SNARE proteins necessary for vesicle fusion docking. A
SNARE proteins cleaved by the seven BoNT serotypes. B. The SNARE complex is
necessary for vesicle fusion docking. C. The substrates for the seven botulinum
serotypes.

Additionally, synaptogamin (SYT)-I and -II, the protein receptors for BoNT B and G, are

expressed in both excitatory and inhibitory synapses of the neocortex and hippocampus.

However, GABAergic synapses have been shown to be less susceptible to BoNT A and E then

glutameric synapses (Verderio et al., 2004). Therefore cortical neurons are a suitable tissue, for a

cell-based BoNT biosensor









CHAPTER 2
REVIEW OF BOTULINUM TOXIN DETECTION METHODS

Introduction

The standard for detection of BoNT toxins is the mouse bioassay, which is able to detect as

little as 0.02 ng of toxin. Strengths of the mouse bioassay include conceptual simplicity and

sensitivity. While the non-selectivity of the mouse bioassay enables it to detect any BoNT

serotype, additional neutralization assays are necessary to determine a specific serotype. Other

limitations of the mouse bioassay include expense, expertise related to maintaining mouse

rearing facilities, and time, since as much as four days may be required to obtain results (Hallis

et al., 1996; Witcome et al., 1999). Several attempts to replace the mouse bioassay have been

made. Methods that have been developed and hold promise for future replacement of the mouse

bioassay include immunoassays, PCR assays, and assays based upon protease activities of

BoNTs. Currently, no single assay appears to be capable of replacing the broadly applicable

mouse bioassay. Therefore, novel toxin detection methods are needed.

Mouse Bioassay

Currently, the standard for detecting BoNTs and the only method accepted by regulatory

agencies such as CDC and FDA is the mouse toxicity and neutralization assay. The mouse

bioassay was first developed by John Trevan in 1927 and was immediately adopted by the

scientific community as the standard for toxicity testing (Trevan, 1927). The mouse bioassay is

exquisitely sensitive and has been shown to detect as little as 0.02 to 0.03 ng of toxin (Ferreira,

2004). This assay is applicable to a large range of toxin concentrations, will detect only active

toxin, and is insensitive to contamination by food, sera, and fecal material. However, it can take

several days to perform the assay since it requires separate tests for detection of BoNTs,









establishment of the minimum lethal dose (MLD) as estimate of toxicity, and neutralization

assays to determine the BoNT serotype.

Both the Center for Disease Control (CDC) and the Federal Drug Administration (FDA)

have outlined procedures for the mouse bioassay (Center for Disease Control and Prevention

2001, Solomon and Lilly 2001). In general, the sample is suspended in gelatin phosphate buffer,

centrifuged and clarified supernatant is used for subsequent testing. For each assay, two mice

are injected intraperitoneally with the specimen. The CDC method simultaneously tests for heat

liability, trypsin activation, and neutralization by antitoxin, but does not require serial dilution

assays be performed. If mice display botulism symptoms only in the trypsinated samples,

neutralization tests are repeated. The FDA method requires testing be conducted in three stages:

toxin screening, toxin titer, and toxin neutralization with monovalent antitoxins. Initially, serial

dilutions in gelatin phosphate buffer of 1:5, 1:10, and 1:100 must be performed as well as tests

using undiluted sample. Trypsin activation of each dilution as well as heat liability of undiluted

sample is also tested. If all mice, except those receiving heated preparations, die, testing is

repeated at higher dilutions. Once toxicity tests have been performed, the preparation which

yielded the highest MLD (trypsinized or untreated) is used for neutralization assays. To perform

neutralization assays monovalent antisera (types A, B, C, D, E, and F), which are available from

CDC, are used. FDA procedure requires protection of mice by injecting antisera prior to toxicity

testing, while the CDC method recommends neutralizing toxin by adding antisera directly to the

sample. Mice are observed at regular intervals for a period of 4 to 6 days. Botulinum toxin will

typically kill mice within 6 to 24 hours, although death may be delayed. Under the CDC method,

if a sufficient amount of toxin is present, only mice receiving samples with the appropriate

antiserum survive. Clinical signs of botulism in mice are ruffled fur, followed by labored









breathing, weakness of limbs and total paralysis. These clinical symptoms must be present to

confirm presence of botulinum toxin (Center for Disease Control and Prevention, 1998).

Although relatively simple in concept, the mouse bioassay is expensive to perform, requires use

of a large number of animals, and can take several days to obtain results. Moreover, it can only

be performed at the CDC or about 20 other qualified laboratories in the United States. Thus, the

mouse bioassay cannot meet the current need for rapid diagnosis of botulism. Additionally, it

cannot be used for high volume screening to detect deliberate contamination of US food, dairy

and/or water supply.

Enzyme Linked Immunosorbent Assay

Due to the need for rapid and specific in vitro assays for detection of botulinum toxins, a

variety of antibody based bioassays have been developed. A summary of antibody based

detection methods is given in Table 1. These assays employ the use of polyclonal and

monoclonal antibodies to directly detect the presence of BoNTs. Polyclonal antibodies (pAb) are

produced by immunization of a host. Purified immunoglobins obtained from the sera of host

animals are then used for bio-detection. While polyclonal antibodies enable sensitive detection of

antigen, the immune response of each animal is unique. Thus, each batch must be validated and

calibrated to ensure minimal variability between assays using different antibody batches.

Additionally, a finite supply of antisera can be produced by a single animal, necessitating an

ongoing effort to produce antibodies. Monoclonal antibodies (mAb) are produced by fusion of a

single antibody-producing B cell with a meyloma cell producing a hybridoma cell line (Emanuel

et al., 2000). While development of monoclonal antibodies is more expensive then polyclonal

antibodies they enable standardized mass production of antibodies. However, they may be

sensitive to strain-specific variations since reactivity can be lost with even minor changes in

antigen. Thus biosensors employing monoclonal antibodies may have more variability in











Table 2-1. Immunoassays for the detection of BoNTs
BoNT Immunosensor Label Matrix Sensitivity Specificity Reference
serotypee)


Alkaline Aqueous 10 MLD/ml mouse Shone


phosphatase suspension,
salmon, and
corned beef,


RVV-XA
labeling
enzyme


Aqueous
suspension and
culture filtrates


BoNT A Amplified ELISA
with capture mAb
BAll and pAb
guinea pig anti-Bot
A
BoNT ELISA&Enzyme
A,B,E Linked
Coagulation Assay
(ELCA) using
affinity purified
horse pAb
BoNT Amplified ELISA
A,B,E,F using affinity
purified pAb IgG

Bot toxoid Multi-analyte
A, B fluorescent
biosensor using
Goat pAb IgG

BoNT Bi-diffractive
A,B grating biosensor
(BDG) with goat
pAb anti-BOT A/B
and rabbit pAb
anti-BOT A/B
BoNT A Hand-held lateral
flow immunoassay
utilizing mouse
Bot Fab5

BoNT Time-resolved
A,B fluorometry assay
using affinity
purified goat pAb
anti-BOT A/B
BoNT E Chemi-
luminescent slot
blot immunoassay
using affinity
purified rabbit pAb
and donkey anti-
rabbit DAb


Biotin
(capture)
Cy5
Bisfunctional
Reactive Dye
colloidal gold





colloidal gold


Lanathide
(Eu3+)


Aqueous
suspension



Aqueous
suspension




Aqueous
suspension


Aqueous
suspension, urine,
human serum,
sewage, dirt


Horseradish Enrichment
peroxidase cultures filtrates
from fish


(66 pg/ml)1 bioassay 1985


A 8 pg/ml, B-
8 pg/ml, E-
8pg/ml



10 MLD/ml
(0.20 to 0.30
ng/ml type A
toxin)2
toxoid A 40
ng/ml; toxoid
B 200 ng/ml


100 ng/ml





.25 ng/ml


200 pg/ml


mouse
bioassay


Dollegast
1993


mouse Ferreira
bioassay 2003,
Ferreira
2004
ND Rowe-
Taitt 2000



ELISA O'Brien
2000


similar
assay with
Bot-01
mAb


Emanuel
2000


ELISA Peruski
2002


4 MLD (72 mouse Cadieux
pg)3 bioassay 2005


1. Estimate of toxin amount is based on the published specific toxicity of 1.5 108 MLD mg protein'1 for purified
BoNT A (Shone 1985)
2. Estimate of toxin amount is based upon the published estimate of 0.02 to 0.03 ng MLD-1 for purified type A toxin
complex (Ferrriera 2004)
3. Estimate of toxin amount is based upon the specific toxicity of 2.8 *107 MLD mg protein1 for pure BoNT E
complex, published in van Barr (2005)


Alkaline Culture filtrates,
phosphatase chili









subspecies detection limits. Both monoclonal and polyclonal antibodies may be expensive to

produce. Recently, a third alternative has been developed. Recombinant antibodies, consisting of

only the antigen binding domain, produced through recombinant DNA technology, have been

developed (Emanuel et al., 2000). To produce recombinant antibodies, cDNA, made from

messenger RNA of an immunized host, is introduced in to Escherichia coli bacteria. Bacteria are

coinfected with a bacteriophage that displays the fragment of antigen binding antibody (Fab) on

the bacteriophage's surface with normally occurring coat protein. The use of Fab in a biosensor

as alternative to mAb or pAb may significantly reduce cost, while providing a stable genetic

source. However, like monoclonal antibodies, Fabs may be sensitive to strain-specific variations.

Antibody based assays include enzyme linked immunsorbent assays, time resolved fluorescence

assays, and lateral flow devices. These assays may employ pAbs, mAbs or Fabs as well as any

combination of them.

Polymerase Chain Reaction Assay

Assays employing the polymerase chain reaction (PCR) assays have also been proposed for

detection of BoNT production by C. botulinum in food samples (Ferreira et al., 1993; Fach et al.,

1993; Campbell et al., 1993; Fach et al., 1995; Sciacchitano and Hirshfield 1997; Aranda et al.,

1997; McGrath et al., 2000; Lindstrom et al., 2001; Braconnier et al., 2001). Polymerase chain

reactions assays use DNA polymerase, a heat stable enzyme that is only active within a specified

range of temperatures, along with synthetic primers to rapidly amplify a target DNA sequence.

Thermocyclers and programmable incubators enable DNA to be denatured, annealed with

primers, and elongated with DNA polymerase. Because DNA is double stranded, each cycle

produces two DNA strands from every one DNA strand present. Thus, the amount of DNA

present in a sample increases exponentially. The PCR technique provides a sensitive method for

detection of specific, known DNA sequences. It is rapid and reliable and does not require the











Table 2-2. PCR primers and DNA probes for detection of C. botulinum and toxigenic C.


butyricum and C. barati
Primer/Probe Method


R-BoNT A, L-BoNT A
P136, P137
P138

BoNT 1, 2
Con.BoNT
Cbot. B
Cbot. E
Cbot. F
Cbar. F
P260, P261
Pr265
Pr266
Pr270
Pr272
Pr280
E3, E4

S1-F, S1-R
T7
CBMLA1, CBMLA2
CBMLB1, CBMLB2
CBMLE1, CBMLE2
CBMLF1, CBMLF2
fB, rB
fBn, rBn
fEF, rEF
fEFn, rEFn
BE1430F, BE1709R
BE1571FP

fl6S, rl6s
16Sprobe

cntBprobe
TQ Al, TQ A2
TQ B1, TQ B2
TQ El, TQ E2
A

E
B


PCR
PCR
Digoxigenin- 1 dUTP
labeled probe
PCR
Fluoresceine
dUTP labeled
probe



PCR
Digoxigenin-
11dUTP
labeled probe



PCR/Capillary
Electrophoresis
Competitive RT-PCR
control cDNA
Multiplex PCR




Nested PCR




Quantitative PCR
FAM/TAMRA
labeled probe
Quantitative RT-PCR
reporter quencher
labeled probe

real time PCR



FAM/TAMRA
labeled probe

VIC/TAMRA labeled
Drobe


Target Gene (size/type)
BoNT A (1.34 kbp/amplicon)
BoNT A(798 bp/amplicon)
BoNT(24 bp probe)

BoNT (1.1 kbp/amplicon)
BoNT (22 bp probe)
BoNT B (24 bp probe)
BoNT E (28 bp probe)
BoNT F (28 bp probe)
BoNT F (29 bp probe)
BoNT(260 bp/amplicon)
BoNT A(19 bp probe)
BoNT B(22 bp probe)
BoNT E(23 bp probe)
BoNT F(24 bp probe)
BoNT G(21 bp probe)
BoNT E(513 bp/amplicon)

BoNT E(250 bp/amplicon)
cDNA(231 bp)
BoNT A(782 bp/amplicon)
BoNT B(205 bp/amplicon)
BoNT E(389 bp/amplicon)
BoNT F(543 bp/amplicon)
BoNT B(480 bp/amplicon)
BoNT B(220 bp/amplicon)
BoNT E,F(630 bp/amplicon)
BoNT E,F(200 bp/amplicon)
BoNT E(280 bp)


16sRNA(210 bp/amplicon)
16sRNA(25 bp probe)

cntB(26 bp probe)
BoNT A(amplicon)
BoNT B(amplicon)
BoNT E(amplicon)
BoNT A(37 bp probe)


BoNT E(31 bp probe)
BoNT B(33 bp probe)


Reference
Ferreira and others 1993
Fach and others 1993


Campbell and others 1993







Fach and others 1995







Sciacchiato and Hirshfield 1996

McGrath and others 2000


Lindstrom and others 2001




Dahlenborg and others 2001




Kimura and others 2001


Lovenklev and others 2004




Akbulut and others 2004









production of neurotoxin, in culture, for detection. Therefore, the incubation time required to

determine presence of C. botulinum may be significantly reduced. All PCR assays discussed

here target the BoNT gene, enabling the assay to detect only organisms carrying the BoNT toxin

gene. Table 2 contains a list of PCR primers and molecular probes used for the detection of C.

botulinum in food samples.

Early PCR techniques were limited to detection of a single strain of C. botulinum and thus

were not applicable to detection of unknown BoNT samples (Ferreira et al., 1993; Fach et al.,

1993; Sciacchitano and Hirshfield 1996). Additionally, primers proposed for use in these assays

are highly variable in melting temperature, requiring separate tests for each serotype, adding

significant reagent costs and testing time. Alternative assays, employed a single primer for

detection of BoNT gene and separate DNA probes to identify each serotype (Campbell et al.,

1993; Fach et al., 1995). However, the probing step for identification of serotype extends

detection time. To enable simultaneous and specific detection of more than one serotype of C.

botulinum a multiplex PCR was developed (Lindstrom, 2001). This assay was able to detect 102

cells per reaction mixture for types A, E, and F and 10 cells per reaction mixture for type B.

Using a two-step enrichment stage the detection limit was as little as 10-2 spores/g of sample for

types A, B, and F, and 10-1 spores/g of sample for type E (Lindstrom, 2001). The detection limit

varied depending upon strain, sample material, temperature, and enrichment time. Relatively low

concentrations of beef inhibited amplification of types A and F, while higher concentrations

inhibited type B and E. An enrichment time of up to 5 days was necessary depending on BoNT

strain and sample type. All of the above PCR assays require an enrichment phase to detect the C.

botulinum organism at sufficiently sensitive levels. This allows inhibitory substances in foods to

be diluted, and prevents false positive results due to non-culturable bacteria (Aranda et al., 1997).









However, addition of an enrichment phase can extend the time required to confirm presence of

C. botulinum by several days. It was also found that detection of C. botulinum in food samples

was complicated by difficulties including low number of target bacteria, contamination by other

bacteria, and presence of other proteases and chelators (Fach, 1993). Also, PCR assays are not

inherently quantitative, therefore additional post-assay processing is required for all such assays

to assist in determination of extent of contamination. A procedure for PCR detection of C.

botulinum neurotoxin A, B, E, and F gene fragments is described in the Bacteriological Analytic

Manual (Solomon and Lilly 2001). It may be necessary to test dilutions since DNA

concentrations outside of the range from 0.34ng-5160ng DNA/100 ul total PCR reaction volume

may yield false negative results. Additionally, since PCR assays do not directly detect botulinum

toxin, they are unable to detect food contamination where insufficient DNA is present. Thus, a

deliberate attack on the food supply using purified or pharmaceutical grade BoNTs would not be

detected. PCR may also detect silent neurotoxin genes resulting in a positive detection even in

the absence of toxic Clostridia spp.

Recently, real-time PCR assays have been developed (Kimura et al., 2001; Akbulut et al.,

2004). Real-time PCR has advantages of being faster and less expensive than the mouse

bioassay. Additionally, fluorescence based real time PCR allows for online monitoring of

amplified gene fragments during each cycle of PCR, thus amplification and detection can be

performed simultaneously. The real-time PCR assay developed by Akbulut et al. (2004)

correlated with the mouse bioassay 75% of the time. Some inhibition of PCR from DNA extracts

was detected by a Taq-Man Exogenous Internal Positive Control Kit. However, further

improvement of the DNA extraction process may significantly improve results.









The PCR assay is used to detect the BoNT gene rather than BoNT toxin. Therefore, it

cannot give a direct measure of toxicity. Additionally, because PCR may be capable of detecting

spores as well as vegetative cells, application of traditional PCR testing for C. botulinum is

insufficient for evaluation of conditions leading to toxin production. To evaluate whether

microorganisms released toxin, reverse transcription PCR (RT-PCR) assays have been proposed

(McGrath et al., 2000; Lovenklev et al., 2004). In this case, gene expression is necessary for

detection, thus only organisms actively producing toxin are detected. McGrath et al. (2000)

developed a competitive reverse transcriptase polymerase chain reaction assay (RT-PCR) for

evaluation of C. botulinum Type E toxin gene expression in type E broth and brain heart infusion

(BHI) broth. In this assay, exogenously synthesized control RNA is introduced with the test

RNA prior to reverse transcription by reverse transcriptase. This allows for the quantification of

target nucleic acid by allowing competitive coamplification of the target nucleic acid along with

a control c-DNA. Results of PCR are then compared to a template with known concentrations of

the competitor (McGrath et al., 2000). This assay correlated well with the mouse bioassay and

was able to detect a 100-fold increase in toxin production in type E broth in comparison to BHI

broth. Lovenklev et al. (2004) developed a quantitative RT-PCR for the evaluation of BoNT B

gene expression in TYP-C broth (tryptone-peptone-yeast extract broth supplemented with .4%

glucose, 0.1% maltose, 0.1% cellobiose, and 0.1% soluble starch). A polyclonal ELISA was used

to distinguish between biologically active and inactive toxin. The RT-PCR method is a valuable

tool that can be used for monitoring BoNT gene expression and conditions that lead to toxin

production. However, because it only detects organisms that are actively producing toxin, it is

inapplicable to broader toxin detection. Thus it will not only fail to detect contamination with

purified toxin, it will fail to detect contamination where toxin is present, but the organisms are









not actively producing toxin as well. Additionally, RT-PCR has not been applied to detection of

toxin production in food and fecal material. Obtaining high-quality RNA from food matrices

may pose a significant challenge to application of RT-PCR to BoNT detection.

Protease Activity Assays

Botulinum toxins are highly specific endopeptidases. Therefore, several assays have been

developed based upon the proteolytic activity of the BoNTs (Hallis et al., 1996; Ekong et al.,

1997; Keller et al., 1999; Witcome et al., 1999; Schmidt et al., 2001; Schmidt et al., 2003; Liu et

al., 2003; Anne et al., 2001). These assays use cleavage of a synthetic substrate to detect the

presence of BoNTs. A summary of protease activity assays is given in Table 2-3. In an assay

developed by Hallis et al. (1996), immobilized VAMP (residues 60-94) or SNAP-25 (137-206)

peptide substrates were cleaved by BoNT B and BoNT A, respectively. Cleaved substrate was

detected by antibodies specific to the cleaved substrate. Sensitivity of this assay, when used with

an amplification system, was 0.17 and 0.13 ng/ml for BoNT B and A, respectively. Similarly, a

51-residue peptide homologous to human VAMP2 (residues 44 to 94) was used to assess the

catalytic activity of BoNT B (Keller et al., 1999). A similar assay was developed for BoNT A

using SNAP25 residues 134-206 (Ekong et al., 1997). Detection of the equivalent of 0.2 1.0

mouse LD50 for BoNT A was achieved in a purified clinical preparation without the need of an

amplification system. These assays provide an additional advantage over previously described

ELISA techniques since results directly determine the amount of catalytically active BoNT

present in a sample. Also, such techniques may be adapted to a high throughput format to rapidly

screen for BoNTs. However, these methods rely on antibodies which may be inhibited by some

foods similar to traditional ELISA assays (Center for Disease Control and Prevention. 1998). To

obtain an assay that is sufficiently robust to a variety of foodstuffs, Witcome et al. (1999)

developed an immunoaffinity column with highly specific monoclonal antibodies to capture









BoNT B in food extracts. The immunoaffinity column was then washed and a solution

containing peptide substrate (VAMP residues 60 -94) was added to the column and in the

presence of BoNT B the substrate was cleaved. The peptide solution in the immunoaffinity

column was then added to a streptavidin-coated microtiter plate where antibodies specific to

cleaved substrate could be used to detect cleaved substrate. As little as 1.5 pg/ml of BoNT B

could be detected in toxin spiked foods. Because multiple monoclonal antibodies were employed

both proteolytic and non-proteolytic botulinum strains could be detected.

Several assays using fluorogenic synthetic substrates have also been developed (Anne et

al., 2001; Schmidt et al., 2001; Schmidt et al., 2003). A high-throughput assay for BoNTs A, B,

D, and F was developed using immobilized substrates consisting of (1) a cleavable synthetic

peptide core based upon the peptide substrate of the BoNT protein, (2) glycine spacer groups

added to both ends, (3) a fluorescine molecule covalently bound to the N-terminus and (4) a

cystine-amide residue as the C-terminus (Schmidt, 2001). For all toxins, as little as 2 ng/ml

produced significant fluorescence above background. This assay did not require use of antibodies

or reverse phase extraction, which enabled it to be easily adapted to a high-throughput automated

environment. While these assays are extremely sensitive, they cannot distinguish between

holotoxin toxins and cleaved light chain, thus they do not evaluate receptor-binding or

internalization of toxin. Also, use of synthetic substrates, that are not readily available, adds

significantly to assay costs.

A novel micro-electromechanical sensor (shown in Figure 2-1) was recently developed for

detection of BoNT B (Liu et al., 2003). The sensor uses nickel-agarose beads functionalized

with synaptobrevin 2-His6, as well as micro-cantilevers functionalized with syntaxin 1A-His6.

The beads are tethered to cantilevers through a complex formed by the N-terminus of










recombinant synaptobrevin 2 and the N-terminus of the recombinant syntaxin 1A. Resonant

frequencies of cantilever vibrations with attached beads are specific and measurable. Subsequent

cleavage of synaptobrevin 2 by BoNT B in the presence of zinc ions produces a change in the

resonant frequency from about 70 Hz to 1 kHz. This system could detect catalytic BoNT B

within 15 min at concentrations of 8 nM or greater.



Silicone Nitride
Cantilever BoNT B
Syntaxin 1A-Hfis, Znr


^ Snaptobrevin 2-His, k.



Nickel-Agarose Bead Tethered Cantilever Tip Detached Tip


Figure 2-1. Botulinum B micromechanosensor. The C-terminus of Syntaxin 1 A-His6 is
attached to a silicon nitride cantilever. The C-terminus of Synaptobrevin 2-His6 is
attached to a nickel agarose bead. The N-terminus of Syntaxin 1 A-His6 forms a
complex with the N-terminus of Synaptobrevin 2-His6 tethering the cantilever tip to
the agarose bead. This generates resonant frequencies of about 70 Hz. Addition of
BoNT B in the presence of zinc ions (Zn2+) cleaves Synaptobrevin 2 such that the
small segment of Synaptobrevin 2 which remains on the bead cannot interact with
Syntaxin 1A. This alters the resonant frequency to 1 kHz.

Mass Spectrometry

Another alternative assay for BoNTs that has been developed is mass spectrometry. Mass

spectrometry has been widely used for both characterization of unknown samples as well as

identification and characterization of macro-biomolecules van Baar et al. (2002, 2004) was able

to identify both BoTX A, B, D, E, and F using both matrix assisted laser desorption ionization

mass spectrometry (MALDI-MS) and coupled liquid chromatography and electrospray mass

spectrometry (LC-ES-MS). MALDI is a sensitive method where the desired analyte is combined

with a matrix compound and evaporated onto a metal plate (Mikkelsen and Cort6n, 2004). Short









pulses of laser radiation are used to release ions from the sample which are passed through a

mass analyzer, producing a mass spectrum plot. A database search of known protein sequences

enables a confidence parameter of the best matches to be calculated. Electron spray (ES-MS)

ionization enables ionization of peptides without fragmentation. Liquid samples pass through an

electric field dispersing the sample into a fine spray of charged droplets. Solvent is evaporated

by passing droplets through an inert drying gas resulting in increased surface charge. When the

surface charge equals columbic forces the droplets explode into smaller droplets. This continues

until sufficient surface charge energy is present to charge ionize sample molecules. Addition of

liquid chromatography enables separation of peptides preventing suppression effects of

ionization. Mass spectrometry data using LC-ES-MS enabled even higher probabilities for

identification of BoNTs. Identification of BoNTs can be complicated due to the presence of HA

and NTHA proteins. However, identification of auxiliary proteins enables subsequent database

searches to be limited to C. botulinum proteins. No data were published on the sensitivity of this

method. Additionally, only limited strains of neat toxin proteins were tested. Identification of

BoNTs in crude samples, including protein matrices may prove difficult.

Recently Barr et al. (2005) developed a mass spectrometry assay that detected synthetic

substrates cleaved by BoNTs A, B, E, and F. This enabled rapid serotype specific detection of

BoNTs. Additionally, a multiplexed method was developed by combining all four substrate

peptides into a single sample. The product peptides were then distinguished based upon their

mass. Detection limits for this assay were approximately 0.62 MLD50/ml for BoNTs A and B and

0.31 MLD50/ml for BoNTs E and F. Primary limitations of this assay are equipment, and high

level of expertise required to operate equipment. Further validation of this assay is necessary

with crude samples.









Discussion

Over the past decade tremendous progress has been made in development of novel assays

for detection of botulinum toxins, however, none of these assays appears capable of challenging

the conceptual simplicity, non-selectivity and sensitivity of the mouse bioassay. Many areas

would benefit from availability of rapid, robust and inexpensive BoNT assays including food

processing, packaging and distribution, food packaging related research and development,

therapeutic testing, basic research focused on evaluating conditions leading to toxin production

or inhibition, and confirmation of clinical botulism. However, because deliberate release of

BoNTs poses a significant threat to the United States population, the most critical need for BoNT

assays may be rapid and reliable detection of weaponized BoNTs. The September 11, 2001

terrorist attacks in the US demonstrated that such events are carefully planned and orchestrated.

It is believed that attacks on the US food supply will likely be carried out with similar

sophistication. Thus, assays to detect deliberate release of BoNTs must be sensitive, robust, and

capable of detecting serologically and antigenically varied BoNTs. Additionally, as demonstrated

by the 2001 anthrax events laboratories screening for BoNTs must be capable of performing tens

or hundreds of thousands of assays or more over a period of a few weeks. Since purified toxin

may be used in a biological attack, direct detection of toxin is preferred to other indicators, such

as nucleic acids. While the majority of the assays described here are capable of detecting BoNTs

well below estimated human lethality limits, very few have been shown to maintain sensitivity in

complex food matrices and against multiple BoNT strains. Those that have maintained

sensitivity are currently limited to single serotypes.

Both ELISA and PCR assays may provide a crude screening device for weaponized

BoNTs. However, careful consideration should be made for inhibition due to sample

contamination. Additionally, PCR assays may be inhibited by the presence of a wide variety of









bacteria. The presence and frequency of naturally occurring C. botulinum species should also be

considered to limit the number of positive PCR assay results. PCR screening for weaponized

BoNTs may be inadequate since purified toxin may not be sufficiently contaminated with DNA

to enable detection. Assays requiring enrichment stages may be inappropriate since weaponized

BoNTs may not contain viable organisms. Assays employing mAbs and Fabs may be sensitive to

strain-specific variations. Exploiting multiple mAbs or Fabs raised against a single BoNT

serotype may enable more sensitive and robust BoNT detection.

Routine screening assays at critical points along supply chains may help ensure safety

while discouraging terrorists from using foods as a vehicle for terrorism. For an assay to be

applicable for routine-screening, it must be rapid, robust and inexpensive while minimizing false

positive and negative indications. Routine screening assays would ideally consist of a bench-top

or portable device that requires little equipment and expertise to operate.

Rapid identification of the agent responsible for a suspected bioterrorism attack is also

necessary to reduce further illness and issue proper recalls while minimizing panic. While much

work can be accomplished through appropriate epidemiological studies, final confirmation of a

suspect food contaminated with a weaponized BoNT currently requires a mouse bioassay. This

may lead to both loss of valuable time and money. Thus, replacement of the mouse bioassay with

a more rapid, but equally sensitive assay remains necessary. It may be possible to reduce the

number of suspect foods with a crude field test, and then perform a more complex assay that

matches sensitivity of the mouse bioassay and mouse neutralization assay at select laboratories.

In this case, confirmation of the suspect food could be made in a few hours at multiple

laboratories nationwide by trained laboratory technicians. Thus a more elaborate assay, such as

the assay proposed by Witcome and others (1999) could be used for confirmation.









Clearly, there is a continuing need for new and better assays for BoNT toxin detection as

well as additional validation of existing assays. Assays developed should be inexpensive and

sufficiently robust such that they will detect any toxin serotype in a variety of substrates

including sera, food, and fecal material. Assays should provide broad-based indications, like the

mouse bioassay, together with capability to identify specific toxins involved. They should be

conceptually simple, easily performed, inexpensive, reproducible, and rapid.









CHAPTER 3
NEURAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELL-BASED
BIOSENSORS

Introduction

One of the first living biosensors, the canary in the mine shaft, was a rapid and reliable

method for detecting toxic conditions within coal mines. However, most modern "living"

biosensors have little resemblance to the canary in the mine shaft. Rather, they rely on the

biological complexity of intact cells or tissue for rapid and reliable detection of desired analytes.

Cell-based biosensors incorporate whole cells, or intact tissue from bacteria, animals, and plants.

They are often capable of specificity and sensitivity, which is difficult to mimic with more

conventional biochemical detection methods. Furthermore, they are capable of detecting more

complex analytes. For example, botulinum toxin detected using the cell-based biosensor

developed in this work, must be capable of a three-step mechanism (binding, internalization and

cleavage of SNAP-25) prior to detection, eliminating false positives from toxin fragments, or

denatured toxin. Furthermore, cell-based assays are often capable of mimicking in vivo

conditions, thus providing valuable insight into how analytes behave in vivo.

Neural cultures, grown over micro-electrode arrays (MEAs) represent a potential cell-

based assay that may be applicable for a broad assortment of tissue. An MEA is a tissue culture

dish with a grid of 60 independent extracellular electrodes embedded under the surface. Each

electrode measures electrophysiological activity (i.e., action-potentials) of the neurons (Figure 3-

1) for extended periods of hours to weeks (Potter and DeMarse, 2001). Micro electrode arrays

have been employed to measure the activity from a variety of electrically active cell lines such as

cortical, hippocampal, and spinal neurons (Segev et al., 2002; Marom and Shahaf, 2002; Selinger

et al., 2004; Eytan et al., 2003; Wagenaar et al., 2006a, Wagenaar et al., 2006b), cardiac









myocytes (Reppel et al., 2007) and retinal ganglion cells (Meister et al., 1995; Meister et al.,

1994).

Use of cortical cultures grown on MEAs for detection of environment threats is an

application currently in development at a variety of labs (Gholmieh et al., 2001; Gholmieh et al.,

2003; Gross et al., 1997; Gramowski et al., 2004; Keefer et al., 2001; Pancrazio et al., 2003;

Stenger et al., 2001; Selinger et al., 2004). MEAs have thus far been used to detect a variety of

environmental threats including mercuric chloride, sodium arsenate, phosdrin, chloridimeform,

strychnine and brevetoxin. These threats come from a broad assortment of contaminants

including heavy metals, organophosphates, insecticides and marine neurotoxins. Additionally,

MEAs have been used with a variety of pharmacological agents including bicuculline, valproate,

atropine, and methanandamide (Cadotte and DeMarse 2005; Cadotte 2004; Morefeild et al.,

2000; Gramowski et al., 2004).

These biosensors typically use changes in the spatial and temporal patterns of activity

across a network of neurons as an indicator of toxicity. One of the most prominent patterns of

activity in these cultures, known as population bursting or bursts, consists of network wide

synchronous volleys of action potentials (i.e. spikes) observed on nearly all electrodes that last

anywhere from 100 ms to several seconds. Both spiking and bursting activity are highly sensitive

to environmental changes, including the introduction of chemical agents to the culture.

Moreover, measuring network-wide activity enables detection of subtle changes in neuronal

dynamics that may be missed by other single neuron based assays. Hence, monitoring changes in

dynamics of these neural cultures following application of test agents could provide a valuable

diagnostic tool and practical biosensor for a variety of neurologically active agents.









Hardware and Software


Micro-electrode arrays (MEA)

The MEAs (Fig. 3-1) used in this study obtained from Multichannel Systems GmBH. The

MEAs are comprised of a glass wafer over which gold was deposited and lithographically etched

to form electrode pads traces. Silicon nitride is vapor deposited to form an insulating layer over

the surface of the MEA. Holes in the silicon nitride insulation are plasma etched at the electrode

pads and the gold electrodes are coated with a thin layer of Titanium nitride (TiN), to form a low

impedance interface between recording amplifiers and culture media. Electrodes have a 30 |tm

diameter and are spaced 200 |tm apart. The entire array covers about a 1.6 mm2 area located near

the center of the dish. A large reference electrode is located on the right side of the dish. Each

electrode records changes in ionic voltage within an approximate radius of 60 tm.























Figure 3-1. A micro-electrode array (MEA) from Multichannel Systems. An 8 x 8 grid of
electrodes 200 [tM apart (right panel) enables the recording of changes in
extracellular voltages produced by activity (action potentials, lower left) from cortical
networks grown in vitro.









Recordings were conducted in an environmental chamber that was maintained at 35.5C

and 5% CO2 to duplicate the environment within the incubator in which the cultures were

maintained.

Data acquisition

Connection to the dishes is made using the MEA60 amplifier system made by

Multichannel Systems (Ruetlingen, Germany). The close proximity of the amplifier to signal

source enables good signal to noise ratio. The MEA60 amplifier system consists of an amplifier

with a set gain and a bandpass filter. Data from the MEA 60 amplifier system is digitized using

MCCard at 25 kHz. A custom stimulation board, shown in Figure 3-3B developed by Dr.

Thomas DeMarse and Vihn Trihn enables the system to rapidly toggle between application of a

biphasic stimulation current and data acquisition. Real time data is processed using MEA Bench

OSX, a custom software package that interfaces with the Mac OSX platform (available at

http://neural.bme.ufl.edu/index.html). This software enables online viewing of raw data, real-

time spike detection and recording of raw and spike data. It also provides a user interface for

stimulation and stimulation artifact suppression using techniques detailed in Wagenaar and

Potter (2002).

Physiological conditions

Recordings were conducted within a custom made environmental chamber shown in

Figure 3-3 A. Temperature was maintained at 35.5 C and CO2 levels were maintained at 5%

during recordings using Multichannels Systems TCO2 temperature controller. Media osmolarity

was maintained by minimizing evaporation from the dish during recordings. To prevent

evaporation, a wet towel was placed in the micro-incubation chamber to increase humidity

within the enclosure. Additionally, to prevent a temperature gradient from forming between the

bottom and the surface of the dish, the MEA 60 recording board was elevated one inch above the









bottom of the chamber using four legs. This enabled heated air to circulate both above and below

the dish.












A. B

Figure3-2 Data acquisition hardware. A. Stimulation board. B. Environmental chamber

Dissociated Neural Cultures

Tissue Culture Methods

Detailed methods used for neuronal cultures have been previously described (Potter and

DeMarse, 2001). Briefly, cortical hemispheres from E17 E18 Wistar rat embryos were

dissociated using papain and mechanical trituration. After dissociation, approximately 50,000

cells (neurons and glia) were placed in the center of a MEA. The surface of the MEA was

treated with polyethyleneimine (Sigma-Aldrich, St Louis MO) to make the surface preferable for

cell adhesion. Polyethylenimine is a branched organic copolymer containing a large number

of positively charged amino groups. Treatment of dishes with polyethylenimine enables long

term adhesion to the dish surface, and prevents clumping of neural cultures. Dishes were also

treated with laminin (Sigma-Aldrich, St Louis MO) prior to cell placement to promote neurite

outgrowth. Laminin was isolated, in the intact native form, from Engel Broth-Holm-Swarm

tumor, a mouse tumor which produces large amounts of basement membranes. Laminin contains

active domains for binding, cell adhesion, heparin binding and neurite outgrowth.









Neurons were cultured in medium consisting of 90% Dulbecco's modified eagle's medium

(DMEM) (Invitrogen Corporation, Carlesbad, CA) and 10% Equine Serum (HyClone, Logan

UT) that was equilibrated to 5% carbon dioxide and 35.5C. The medium for each culture was

exchanged twice per week and cultures were incubated continuously in 5% CO2 at 35.5 C.

To enable long term development, sterile conditions were maintained within the tissue

culture chamber of the MEA. To maintain sterility, MEAs were fitted with a Teflon ring that

seals around the outer wall of the tissue culture chamber. The top of the teflon ring was covered

with a FEP Teflon film. This film was permeable to gases such as CO2 and 02 but was

impermeable to water and other contaminants. Long-term tissue culture methods are described in

detail in Potter and DeMarse (2001).

Development and activity

Cultured cortical neurons, when plated on MEA dishes, begin organizing into networks

within a few hours of plating. Persistent tonic firing is usually observed within 2 to 5 days in

vitro (DIV). Synchronized bursting typically appears within 7 DIV, but may appear as late as 14

DIV. Bursting may be preceded by large single channel or small multi-channel (<5) bursts of

activity that persist for up to two days. During the initial period when bursting appears, bursts are

often large, lasting 1-5 seconds. These large burst are followed by long quiescent periods of 3-15

min, in which no bursting occurs. As cultures mature, larger bursts are replaced by shorter, more

regular bursts. Often, around 10 to 11 DIV, clusters of bursts appear. These bursts are frequently

referred to in literature as super bursts. Superbursts include stable clusters of 5-7 bursts (up to

12) with an inter-burst interval (IBI) much smaller than the mean IBI. Alternatively, longer less

stable volleys of bursts may appear. Occasionally these super bursts persisted through the life of

the culture.









Activity typically stabilizes around 30 DIV, with short regular bursts. Burst rates are

variable between cultures, although much less variable within batches plated at the same

Al 21 31 51 b 1 11, 1















Figure 3-3. Screen shot of typical raw activity during a network-wide burst. Each window











represents a single electrode in the 8 by 8 grid of electrodes, and is 200 ms wide and
50 mV high.

time than between batches. Cultures may burst as little as a few times per minute, or as often as

every 5 seconds. Activity also depends on culture density, which may partially explain

variability between batches. Additionally, cultures with higher levels of tonic spiking activity

typically achieve maturity faster. Thinly plated cultures generally exhibit delays in maturation

while sparse cultures may never achieve stable short bursts. At four weeks in vitro the cultures

were considered mature (Marom and Shahaf, 2002; Habets et al., 1987; Muramoto et al., 1993).

All experiments discussed here were performed on mature cultures (> 28 DIV).
'4- t 84A-44-7r--b4




*^ I--- -L l"-d. L. p --- k--- 5-- -7 --- -- A-1-




117 ----- y-."-7- p -- ^ -- ^---71"T7 --- --j77--r--II 11 1-W l






Figure 3-3. Screen shot of typical raw activity during a network-wide burst. Each window














were considered mature (Marom and Shahaf, 2002; Habets et al., 1987; Muramoto et al., 1993).

All experiments discussed here were performed on mature cultures (> 28 DIV).









Network Bursting

Network wide bursts are the dominant pattern of activity observed in cortical-networks

grown in vitro. Spontaneous synchronized patterns of activity are found in a variety of tissues

including the developing visual system, epileptogenesis, and in the central nervous system

development. However, mechanisms leading to bursting activity have been the subject of broad

debate. Several theories have been proposed. One theory was that activity is the result of a class

of specific pacemaker cells that form a trigger network, or that bursting is the result some

chemical factor X which is present in the culture. Medea et al. (1995) disproved the factor X

theory by slicing networks into quarters and observing bursting that was independent even when

initiated close to borders. He concluded that bursting was governed by the level of presynaptic

firing, degree of connectivity, and by excitation and recovery processes. Shao et al. (2006) used a

model to demonstrate that burst-like activity could be generated independent of ion channel

properties, or synaptic mechanisms. Thus, he concluded bursting is critically dependent upon

connectivity. This is, at least in part, in agreement with what is observed in the laboratory since

sparse cultures did not produce bursting.

Many theories have also been proposed about mechanisms involved in burst termination in

random cultured networks. These include synaptic disinhibition as the mechanism by which

bursts terminate. However, disinhibition was shown to induce slow bursting rather than

suppressing it. Alternatively, Streit et al. (2005) proposed that slow spike frequency adaptation

may be the primary mechanism by which bursts terminate. Spike frequency termination may be

due to accumulation of Ca2+ dependant K+ currents or slow inactivation of Na+ currents. Since

increases in extracellular Ca2+ did not terminate bursts, it was concluded that progressive

inactivation of sodium channels was the mechanism by which bursts terminate.









MEAs as Biosensors

A variety of inherent network properties have been used to measure the changes in pattern

activity. Some of the more common methods used include changes in burst rate, spike rate, and

burst duration, mean spikes per burst, interspike interval, burst amplitude, and spike amplitude

(Gramowski et al., 2004; Keefer et al., 2001; Pancrazio et al., 2003).

In attempting to identify unknown substances using neural networks grown on MEAs

Gramowski et al. (2004) calculated the coefficient of variation (CV) across the network and

across time for native activity and five pharmacological agents. Additionally, an autocorrelogram

of a spike timestamps was fitted to a Gabor function. Amplitude, decay, exponent, and offset of

this function were then added to standard data features, such as spike rate and burst rate, to create

a database of standard activities for the six pharmacological agents. The database was then

compared to activity patterns of six unknowns. Using the Euclidian distance, the similarity

between unknown agents and the database of agents was analyzed. This enabled correct

identification of two of the agents and the native state 100% of the time, two agents 94% of the

time and one agent 84% of the time. This work demonstrated how a biological fingerprint of

several analytes can be generated.









CHAPTER 4
DEFINING THE BONT A FINGERPRINT

Introduction

In this chapter the potential of cortical networks cultured on MEAs as a biosensor for

BoNT A by bath application of the toxin is assessed. The activity of these cortical networks over

the course of one week, was measured to characterize changes in the spatiotemporal patterns of

activity relative to untreated controls indicative of BoNT A.

Materials and Methods

Procedure

Neurons began producing spontaneous action potentials within the first few days after

plating on MEAs. Cultures were considered mature at four weeks in vitro, at which point

neurons expressed the full complement of receptors (Marom and Shahaf, 2002; Habets et al.,

1987; Muramoto et al., 1993). All cultures used were older than 30 DIV to allow for cultures to

be considered fully developed. Recordings were conducted within an environmental chamber

that was maintained at 35.5 C and 5% CO2 to duplicate the environment within the incubator in

which cultures were maintained.

Ten MEAs were randomly assigned to either a BoNT A group (N=5) which received toxin

or a Control group (N=5) without toxin. Twenty minutes of spontaneous neural activity recorded

prior to treatment with and without BoNT, was used as a baseline activity measurement. Activity

was then recorded immediately following application of BoNT for the BoNT A group and during

a similar period for the Control group. This was followed by twenty minute spontaneous activity

recordings at 12, 24, 36, 48, and 72 hours, and 1 week after treatment.

Following each recording, five of the 60 channels were stimulated in a randomized order to

measure any changes in amount of evoked activity between the BoNT and Control groups. Each









stimulation consisted of a single 600 mV, 200 |iS, biphasic pulse, that typically evoked a short

burst of activity lasting 100 ms or more. The five stimulation sites were selected prior to the first

recording with the criterion that a single stimulation pulse would produce activity on at least

50% of the channels. Each site was stimulated with five pulses at 0.10 Hz. Stimulation order

was randomized during each session. At the end of the week of recordings, MEAs in the Control

group that had not been dosed with BoNT A were then dosed with BoNT and recorded for a one

week period similar to the BoNT group. One of the control cultures during this reversal had to be

removed from the experiment due to infection.

Drug application

Lyophilized BoNT A complex (BOTOX, Allergen Pharmaceuticals) was reconstituted in

0.9% sterile saline at a concentration of 200 U/ml (approximately 10 ng/ml). Each vial of

BOTOX contains 100 U of botulinum toxin type A, 0.5mg of human albumin, and 0.9 mg

sodium chloride (Allergens Pharmaceutical Ireland 2004). Twenty micro-liters of toxin solution

(4U of toxin) was applied directly to the 1 ml culture bath. The concentration chosen was well

below human lethality limits for safety reasons. In addition, it was sufficiently low to

demonstrate sensitivity of the networks for BoNT A. This is particularly important since dilution

of crude samples will likely be necessary prior to dosage of networks in future studies. Toxin

carrier solution, consisting of albumin diluted in 0.9% sterile saline to a concentration of 1

mg/ml, was applied to control cultures.

Analysis and statistics.

Action potentials (i.e. spikes) produced by neurons grown on the surface of the array were

detected as voltage deviations greater than five standard deviations above or below the mean

noise level for each electrode. Activity was also stimulated through the application of 600 mv,

200 itS, bi-phasic voltage pulses at select electrode sites (Wagenaar and others 2004). Spike









events were extracted and stored online for further offline processing to identify burst events.

Burst events were detected by smoothing the number of spike times from all channels into 0.5

ms bins. Consecutive bins that exceeded a threshold number of spikes based upon the mean bin

size, were then identified as bursts. The end of each burst was detected as a 10 to 20 ms period in

which no activity was detected. Data is presented as a mean percentage of the baseline activity

+/- S. E.M. The separate variance t-test was used to establish significance between control and

BoNT dosed groups (cultures) with P < 0.05 considered as significant. The Paired Student's t-test

was used to compare post-treatment activity to baseline (within groups). The Paired Student's t-

test was used as well as to compare BoNT dosed control dishes from the reversal to control

activity obtained the week prior. In this case, results were confounded with blocks (week 1 and

week 2).

Results

Spontaneous activity

Application of the toxin resulted in a significant changes in the spontaneous activity before

versus after treatment in the BoNT group and relative to the Control group. In 89% of BoNT

treated cultures application of BoNT resulted in an increase in duration of spontaneous bursts

and the number of spikes within those bursts with respect to activity recorded during baseline

(before the toxin had been delivered). Figure 2 shows the mean duration of each burst (Panel A)

and average number of spikes per burst (Panel B) in the BoNT and control groups from 12h to 1

week following BoNT application. Dishes dosed during the control reversal also had a

significant increase in burst duration and spikes per burst (data not shown). Increases were

observed within 12 to 24h after BoNT dosing and continued to increase until 72h. Increases were

significant 48h after dosage with BoNT A (P < 0.05). There was, however, no significant

difference between spike rate during each burst or inter-burst interval immediately following the













A. Burst duration
E Control N = 5
* BoNT N = 5


12h 24h 36h 48h 72h 1wk
Time Post Treatment


200

180

S160

w 140

f 120

W 100

80
60
w 60

0. 40

20

0


B. Spikes per burst


12h 24h 36h 48h 72h 1wk
Time Post Treatment


Figure 4-1 Spontaneous burst activity in cortical networks after treatment with BoNT A or toxin
carrier. A. Average burst duration as a percent of baseline activity. B. Average spikes
per burst as a percentage of baseline activity._A. Average burst duration as a percent
of baseline activity. B. Average spikes per burst as a percentage of baseline activity. *
Indicates a significant increase in burst duration over baseline activity (Students t-test,
p < 0.05), control N =5, BoNT N = 9

application of the toxin (data not shown in the figure, P > 0.05). Finally, there was no significant


difference between the BoNT and control group prior to the application of the toxin for spike


rate, burst rate, inter-burst interval and burst duration (P > 0.05).









Intra-burst neural dynamics

Application of the toxin not only changed the number of spikes in a burst and the burst

duration, it also affected the distribution of spikes within each burst event. To illustrate this

change return plots were generated in which the activity during a burst was plotted as a series of

points comparing the rate of activity at time t to the activity observed 50 ms in the past. In a

normal culture, a typical burst would consist of tonic spiking observed across the majority of the

array in which a rapid increase in activity peaks within 50ms followed by a slowly decay

appearing in a return plot as a loop beginning at the origin (Wagenaar and others 2006b; Segev

and others 2001; Tabak and Latham 2003; van Pelt and others 2004; van Pelt and others 2005;

Kamioka and othersl996; Jimbo and others 2000).

Figure 3 shows return plots for the control group (left column) and BoNT group (right

column) for three subjects during baseline, 24, 48, and 72 hours. The normal burst patterns are

illustrated in the left column of the figure for two subjects in the control group. A similar pattern

can also be seen for the baseline activity in the BoNT group before the application of the toxin

(right column, blue line). In contrast, the shape of the trajectories during each burst was altered

after delivery of the toxin in the BoNT group (right column). After 24 hr, slight oscillations in

the rate of activity during the burst begin to appear that became more pronounced over time. The

oscillatory behavior represents the appearance of multiple peaks of activity (three or more)

within each burst. This was true particularly after 48 and 72 hours and was apparent in 50%

BoNT cultures, but never observed in controls.

Note the scale of the left upper panel of Figure 3, labeled as control 7079, is nearly on

order if magnitude smaller than the other panels. This is because the network covered a much

smaller portion of the array, approximately 50%. However despite the much smaller magnitude











observed in the burst trajectory, the shape of the burst trajectory remained stable over the seventy


two hours shown.


x lo3 Control 7097 lo1 BoNT A 9299
Base
424h
-72h












x 103 104




x 104 Control 9059 x 104 BoNT A 9059




2 I
0








0 05 1 1.5 2 25 3 0 05 1 15 2 25 3
Spikes per 100 ms at time t x 10 Spikes per 100 ms at time t 104

Figure 4-2. Return plots generated by sorting spikes into 100 ms bins and plotting bins at time t
against bins at time t 50 ms. Responses are from baseline recordings (dark blue), as
well as 24h (light blue), 48h (pink) and 72h (red) in two carrier dosed cultures (left
column) and two BoNT dosed cultures (right column). The bottom row is the same
tissue culture with and without BoNT treatment

(right column, blue line). In contrast, the shape of trajectories during each burst was altered after


delivery of toxin in the BoNT group (right column). After 24 hr, slight oscillations in the rate of


activity during the burst begin to appear that became more pronounced over time. The oscillatory


behavior represents the appearance of multiple peaks of activity (three or more) within each









burst. This was true particularly after 48 and 72 hours and was apparent in five of the nine BoNT

cultures, but never observed in controls.

Evoked response following intoxication with BoNT A

The toxin's effect on spontaneous activity was also compared with activity evoked through

a series of five brief biphasic pulses after each recording session. Application of toxin resulted in

significant changes to the overall rate of spiking of activity in cortical culture during bursts, and

the pattern of spiking relative to the control group. Figure 4-3 shows an example of action

potentials (plotted as points) produced for each channel following a single stimulation pulse

(panel A, B) and the average activity across channels (panel C, D). There was no significant

difference in the number of evoked spikes between the BoNT and Control group during baseline,

or 12 hours after toxin delivery. However, there was a significant increase in number of spikes

evoked in the BoNT vs. Control group 24, 48, and 72 hours following the toxin which persisted

up to 1 week. The increase in spike activity was most apparent in the late phase of the evoked

burst (100 to 300 ms post stimulation). Figure 4-5 shows the average number of late phase spikes

evoked through stimulation before, 12, 24, 48, 72 hours, and 1 week after toxin delivery in the

BoNT and Control group. There was no significant difference between the number of

stimulations that resulted in a burst (Control 20.0+/-0.91, BoNT 21.0+/-0.92) .

A comparison between patterns of evoked activity 300 ms following stimulation in the

control group (top panel) versus the BoNT group (bottom panel) is shown in Figure 4-5. There

was no significant change in activity observed in BoNT A dosed cultures during baseline or 12

hours following administration of the toxin. However, beginning at 24 hours, eight of the nine

BoNT dosed cultures displayed elevated activity during the later part of the elicited burst (>100

ms post stimulation) that persisted up to 1 week following toxin delivery. Interestingly, after 36h

multiple peaks in the rate of activity was observed in five of nine BoNT dosed cultures and eight
















50


40
-i
c
30


20


S 50 100 150 200 250 300
Time Post Stim (ms)



C


50j

40-

E2o-

A5 20-l \ ~


4 r rr50r r- frr r .. V *




10 ,r-"...... M *f, v


40 rtrnwwr.- -p....t r vr






Post Stim Time (ms)


Time (ms) Channel Number


300 7
Time Post Stim (ms) Channel Number

Figure 4-3.Stimulation-response for cortical culture grown on MEA 9053 during the baseline
recording (panels A, C) and 72h after treatment with 200 pg BoNT A (panels B, D).
Raster plot after a single stimulation (top panels), and average stimulation-response
plots for each stimulation channel (bottom panels).


. etm








*.-r.ws.:
'"-'m'
*rB.ss..
aaW.


.7..mtJ
rnrr
a **










Total Spikes 100 to 300ms Post Stimulation as a Percent of Baseline Activity

1000 -
350
an 3 T Control, N =
900 300
E BoNTA,N=5
Q o 250
o 800 8100
M 300 -
o 0 150 *
o 700
V--










*. 100
cm .2 600 50-













12h 24h 36h 48h 72h 1 week
Time Post BoNT A or Carrier Treatment

Figure 4-4 Mean change in the number of evoked spikes 100 to 300 ms post stimulation in the
BoNT (closed bar, N=5) and control cultures (open bar, N=5). Due to culture to
culture variability the data was normalized as a percent of baseline activity, where
100 percent indicates no change in activity. The baseline average of spikes 100 to 300
ms post stimulation for control and BoNT groups is shown in the panel inset. There
was no significant difference between baseline activity of BoNT A dosed and control
cultures.

of the nine cultures at 72h recording. Initial peaks in the rate of activity observed in both the

experimental and control groups were followed by a series of smaller peaks in the BoNT but not

control cultures. Each peak in the activity rate of the BoNT group was slightly smaller, and

slightly longer in duration than the previous peak, lasting 50 to 100 ms and spaced 50 to 100 ms

apart. This is similar to effects observed during spontaneous recordings. Moreover, this

oscillatory effect was still apparent even one week after toxin application in the BoNT group.





















0 300










300 Baseline
50


300 Baseline
Post Stim Time (ms)

B 30 2
--24h
S--48h
-72h
\ week
SB I Baseline


3300






25250





300 Baseline

Post Stim Time (ms)


Figure 4-5. Average stimulation response for Control (Panel A) and BoNT dosed (Panel B)
tissue cultures. Note the increase in spike rate during the latter portion of the
stimulation response (>100 ms) as well as the emergence of oscillatory behavior most
prominent at 72h.
prominent at 72h.









Discussion

Use of cultured neural networks as a biosensor depends upon identification of distinctive

modifications to native activity, which can serve as a fingerprint for biological agents of interest.

A variety of inherent network properties have been used to measure the changes in activity.

Some of the more common methods used include changes in burst rate, spike rate, and burst

duration, mean spikes per burst, interspike interval, burst amplitude, and spike amplitude

(Gramowski et al., 2004; Keefer et al., 2001; Pancrazio et al., 2003). The effects on neural

activity of an important food-borne toxin, BoNT A, have been examined. Application of BoNT

A to the culture media resulted in an increase in number of spikes per burst within 36h after

BoNT dosage. Furthermore, an oscillatory change in bursting activity was observed, which to

our knowledge has not been previously reported with other biological agents. These changes are

present not only in spontaneous bursts, but also in bursts elicited through stimulation.

As shown in Figure 4-2, the size and shape of spontaneous burst activity produced in a

single recording session can be highly variable (Wagenaar et al., 2006b). Thus, long recording

periods are necessary to obtain representative samples of activity. However, elicited bursts tend

to be less variable in size and shape and are produced reliably over time. Because of this, the

amount of recording time necessary to obtain representative samples of activity is lower for

stimulation experiments then spontaneous recordings. Therefore, it is anticipated that monitoring

stimulated burst activity will greatly aid in the detection of BoNT A by reducing variability and

time. Additionally, return plots of spike activity are a simple method to rapidly visualize changes

in burst activity. These plots can easily be obtained during online data processing since they do

not require a prior knowledge of burst activity. Thus, they may be a valuable tool for visually

identifying changes in activity associated with toxin or drug application.









Why this oscillatory behavior appeared following toxin application during both

spontaneous and evoked bursts is not clear. Previous studies have demonstrated that the primary

target of BoNTs in the hippocampal and neocortical tissues are excitatory synapses, while

GABAergic synapses remain much less effected (Verderio et al., 2006). For example, BoNT has

been shown to block presynaptic release of not only acetylcholine, but glutamate, glycine,

noardrenaline, dopamine, serotonin and neuropeptides (Keller et al., 2004; Verderio et al., 1999;

Verderio et al., 2006). Factors such as structural connectivity, synaptic strength, and the balance

between excitatory and inhibitory contributions within these cultures have been shown to

produce a profound effect on the pattern of activity observed, especially during development

(Nakanishi and Kukita 1998; Streit et al., 2005; Wagenaar et al., 2006b). Thus, it is not

surprising that this toxin would alter the pattern of activity within each burst.

In this experiment at least 36 hours after toxin exposure was required before a significant change
in activity was observed. This is similar to the mouse bioassay which requires from 12 to 72h for
BoNT detection perhaps reflecting the slow toxin uptake (Scarlatos et al., 2005). Additionally,
ongoing changes in activity observed after the 24h recording may be a result of further SNAP-25
cleavage after BoNT uptake with the maximum level of SNAP cleavage not occurring until 48 to
72h post treatment.









CHAPTER 5
REFINING THE BONT A FINGERPRINT AND DEVELOPMENT OF A BONT DETECTION
ALGORITHM

Introduction

Work highlighted in the previous chapter demonstrated that MEAs are a suitable substrate

for detecting BoNT A action on neocortical neurons. A possible explanation for the oscillatory

changes observed during bursts was a reduction in synaptic strength. Synaptic strength between

connected neurons can be modified by a variety of factors including the amount of presynaptic

neurotransmitter release and the post-synaptic response to neurotransmitter release. An approach

to assessing changes in synaptic strength associated with BoNT intoxication is to evaluate

stimulus dependant plasticity. Additionally, a tool adapted to neuroscience from economics,

known as Granger's Causality (GC) can be used to evaluate changes in connectivity associated

with intoxication. Finally, an application of an algorithm for BoNT detection was used to

determine BoNT A in tissue cultures dosed with 200pg, 600pg and 1000 pg of BoNT A.

Background

Plasticity

Plasticity of the nervous system through activity-dependent modification of synaptic

strength is central to development (Meister et al., 1991), learning, and memory (Chirwa et al

2001; Dan and Poo, 2006). Specifically, long term potentiation (LTP) and long term depression

(LTD) are two key mechanisms by which synaptic efficacy may be either increased or decreased

respectively. Synaptic strength depends on several parameters, including initial release

probability of the synapse, size of the readily releasable vesicle pool and release probability of

each vesicle.

It has been demonstrated that spike timing plays a central role in plasticity. When

postsynaptic spiking precedes presynaptic spiking (post-pre) within a short interval, the synapse









is depressed via LTD. Alternatively, if presynaptic spiking precedes postsynaptic (pre-post)

spiking within a short interval the synapse is enhanced via LTP (Poo et al., 2001). Spike intervals

necessary for LTP and LTD are dependent upon neuronal type, and range from 10 to 100 ms.

Jimbo et al. (1999) developed a method for detecting stimulus driven LTP and LTD in

random neural networks grown on MEAs. In essence a 600 mV biphasic probe sequence is

applied to the 60 MEA channels ten times. Then a 600 mV 20 Hz tetanic stimulus is applied at a

single electrode site. Then the pre-tetanic probe is repeated. Pre-tetanic spike rates for each

channel are compared to post-tetanic spike rates. Interestingly, it was found that changes induced

by tetanic stimuli were pathway dependant.

Granger's Causality

Botulism likely reduces synaptic efficacy, leading to changes in effective connectivity.

Traditionally, evaluation of structural information relating to activity observed within MEAs has

been limited to cross-correlation and coherence in the spectral domain. These techniques can

demonstrate statistical relationships between multivariate data, but lack directional information.

Thus, they provide little evidence about functional connectivity. However, recent adaptation of

Granger's Causality (GC), traditionally an economics tool, to the field of neuroscience has

enabled causal influences between neural time series to be examined. The underlying principle

behind GC, is improved prediction of a time series by including information from a second time

series if the second time series has causal influence on the first time series (Ding et al., 2006).

Granger adapted this principle to linear regression models using the variance of the

autoregressive prediction error to determine if prediction of the first time series was improved by

the second. Similarly, the second time series can be evaluated using the first time series enabling

bidirectional causality to be identified. It is important to note that this is not considered to be true

causality, but rather Granger's Causality. This method has recently been adapted by our









laboratory at UF for use with multivariate data from microelectrode arrays. Additionally, a

biologically plausible computer simulation has been used to demonstrate that values obtained

from Pairwise Granger's Causality analysis (PGC) scales linearly to synaptic weights,

demonstrating that PGC provides a realistic estimate of connectivity (Cadotte et al., 2006). It was

hypothesized that BoNT A would reduce the effective connectivity. This would result in a

network wide decrease in Granger Causality values after BoNT addition.

To calculate Pairwise Granger's Causality (PGC), 200 ms post stimulation spike trains

were low pass filtered and fitted to a bivariate autoregressive (AR) model:

N N
X, = -AA(j)Xj_ + cA (j)Ytj +Ex, (5-1)
J=1 J=1

N N
Y = ZA, (j)Xj + A (j) +E, (5-2)
J=1 J=1

The coefficient matrix A, and error matrix E was then estimated using Morfs method

(Morf, 1978). Following methods detailed in Ding et al. (2006), the bivariate AR model is

Fourrier transformed to the spectral domain to generate the autospectra Sxx, and Syy and the cross

spectra Sxy. Gewekes's transform, applied to autospectra from the same bivariate model is used to

calculate the intrinsic power term. The causal influence for X--Y and Y--X in the spectral

domain is then defined as:


fx zr() = In ()(5-3)
sl (5-3)



x () = In S(u) (5-4)


The values obtained represent PGC in the spectral domain. To represent PGC between

electrode pairs, the maximum value between 1-30 Hz is used.









Methods


Experimental protocol

Tissue culture methods during these experiments were the same as those detailed in

Chapter 4. MEAs were dosed with 600 pg (n = 4) and 1 ng (n = 6) ofBoNT A and compared to

control cultures dosed with BoNT carrier were recorded over the same period. Twenty minutes

of spontaneous neural activity, recorded prior to treatment with either BoNT or no BoNT, was

used as a baseline activity measurement. Activity was then recorded immediately following

application of BoNT for the BoNT A group and during a similar period for the Control group.

This was followed by twenty minute spontaneous activity recordings at 12, 24, 36, 48, and 72

hours after treatment.

Following each spontaneous recording session, ten channels were stimulated for the 600 pg

and 1 ng cultures using a 600 mV 200 |.s biphasic pulse. Stimulations were replicated ten times

and the stimulation order was randomized.

Plasticity protocol

The Jimbo protocol was applied 72h after treatment to the four cultures dosed with 600 pg

BoNT A and four cultures dosed with BoNT A carrier (Jimbo et al., 1999). Briefly, a test

stimulus consisting of a bipolar 600 mV and 200 [s pulse was applied at 58 of the 60 electrode

sites. Each pulse was applied at a three-second interval, and ten replicates were performed with

three seconds between replicates. The stimulation order was randomized. After the test stimulus

was applied, a tetanic stimulation pulse consisting of 20 trains of 10 pulses (600 mV and 200 |.s)

at 20 Hz was applied to a single electrode site. The test stimulus was reapplied following tetanic

stimulation. Plasticity changes induced by the tetanic stimulation were evaluated by comparing

the average spike rate for each test stimulis site.









Results and Discussion


Spontaneous activity

Cultures dosed with 600 pg and 1 ng ofBoNT A exhibited similar behavior to results

characterized in Chapter 4. No significant change in activity was observed up to 24h post BoNT

dosage. At the 24 h period, most cortical networks dosed with BoNT A exhibited a slight

increase in burst duration compared to baseline activity. An increase was also observed in three

of the controls. Interestingly, at the 48h period two of the 600 pg BoNT A dosed cultures and

three of the 1 ng dosed cultures exhibited superburst behavior. By the 72h recording period,

seven of the ten cultures dosed with 600 pg or 1 ng BoNT A exhibited superburst behavior.

Three control cultures exhibited superburst behavior as well, though the size and duration of the

bursts within the superbursts of control cultures were much smaller and more regular in duration.

Burst durations in networks that exhibited superbursts were drastically reduced relative to

cultures with regular bursting. This resulted in a decrease in burst duration for these cultures.

Thus the burst duration of spontaneously recorded bursts did not reliably increase over the

recording session for either 600pg or 1 ng dosed cultures, and was not stable during spontaneous

recording sessions. Additionally, the burst detection algorithm did not reliably detect the burst

onset and burst termination of bursts in cultures exhibiting superburst behavior. Because of this,

it was determined that burst statistics obtained during spontaneous recordings may not be

suitable for BoNT detection.

Elicited activity

Cortical cultures were also stimulated using a 600 mV 200 [is biphasic pulse. Stimulation

channels were selected that produced a reliable response. Stimulations that were already bursting

immediately prior to a stimulation pulse were excluded from the analysis. On average, one to












three stimulations were excluded from each recording for this reason. There was no significant


change in the number of stimulations with prior bursts between control cultures (P >



A.
2500.0 -
Control
ft
S200 600 pg BoNT
CL 2000.0 -
0)
,,
a-
1500.0
aJ


S1000.0
m


1 500.0
-.


10m 6h 12h 24h 36h 48h 72h
Time Post Treatment


B

900.0

800.0


1" 700.0







400-
t
600.0
a-
*S 500.0



U-
C
= 400.0

f 300.0
0
* 200.0-
oa
100.0 -


* Control

* 1000 pg BoNT A


10m 6h 12h 24h 36h
Time Post Treatment


48h 72h


Figure 5-1. Percent change in spikes with respect to baseline activity after treatment with 600 pg
BoNT A (A) and 1000 pg BoNT A (B). (* indicates p < 0.05)


0.0 -









0.05). Occasionally, a stimulation did not elicit a burst. Approximately 80 percent of stimulations

elicited bursts during each recording session. There were no significant differences between

Control and BoNT dosed cultures. Burst sizes associated with stimulations were smaller in

cultures dosed with 600 pg and 1 ng BoNT A and their respective control cultures than those

reported in the previous chapter for cultures dosed with 200 pg BoNT A. This is attributed to a

reduction in the signal to noise ratio due to the age of the MEAs used. The number of spikes

during the late phase of the burst (100 to 300 ms post stimulation) was found to increase in both

cultures dosed with 600 pg and 1000 pg BoNT A, as observed in BoNT dosed cultures in

Chapter 4. Figure 5-1 shows the percent change in the average number of late phase spikes for

cultures dosed with 600 pg and 1000 pg. A significant increase in late phase spikes (p > 0.05),

occurred 48h after dosage in both cases.

Plasticity

To further characterize activity changes induced by BoNT intoxication, plasticity changes

associated with LTP and LTD were assessed using the protocol established by Jimbo et al.

(1999). Results for culture 9059, shown in Figure 5-2, were consistent with all four cultures. The

color map shown in Figure 5-2 indicates the difference in average number of spikes between pre

and post tetanus application. The x-ordinate indicates probe site, and the y-ordinate indicates

response site. As demonstrated in Jimbo et. al. (1999), changes in the response channel varied

according to the stimulation site, while activity in response to a single probe site was uniformly

depressed or enhanced. Jimbo's experiment demonstrated that activity changes associated with

tetanic stimulus-induced LTP and LTD were pathway dependant. Results obtained for both

control and BoNT dosed cultures were consistent with Jimbo's observations. Similar to control

cultures both stimulus induced LTP and LTD were present in BoNT dosed cultures. However,

numbers of probe sites that were depressed, and magnitude of the changes in activity (both














9059 BoNT
10
0


-10


3 10 20 30 40 50 60

--if 0m" .706. i"I "m "



1 .






60L "


10 20 30 40 50 60
probe


9059 Carrier
In


lO
10
0 10 20 30 40 50 60


10 : A r l ._

20 -
m--. .- -
30






30
50



10 20 30 40 50 60
probe


0

10 -

20

30

40

50

30
-20 0 20


10

10

20

30

40

50

60
-10 0 10


Figure 5-2. Network plasticity induced by tetanus (A) 72 h after dosage with 600 pg BoNT A,
and (B) 72h after dosage with BoNT A carrier solution, results shown are for dish
9059 as a representative of results obtained for 5 cultures. The color map indicates the
difference in the average number of spikes between pre and post tetanus application.
The x axis indicates the probe site, and the y axis indicates response site. The bar
above the color map indicates changes in response channel 40 for each probe site. As
demonstrated in Jimbo et. al. (1999), changes in the response channel varied
according to the stimulation site. The bar to the right of the color map indicates
changes in spike rate for each response channel for probe channel number 47 (red)
and 9 (blue). Consistent with results obtained by Jimbo, for a given probe site
response channels were either all depressed (LTD) or potentiated (LTP) indicating
that plasticity induced changes are site specific. Interestingly, plasticity changes in
BoNT cultures were much larger than responses in control cultures.









depression and potentiation) increased. This may indicate that the mechanism by which LTP and

LTD are produced is altered by BoNT A. Alternatively, this may simply be the result of effective

connectivity changes in the culture.

PGC Analysis

To determine changes in network connectivity PGC was applied to stimulation data from

control and BoNT dosed cultures. For the PGC analysis, after each stimulation a 200 ms window

of data was used. Data from stimulations in which the culture was bursting prior to stimulation,

and stimulations which did not elicit a burst were excluded from the analysis. In both control and

BoNT dosed cultures, both increases and decreases in PGC values were observed over the

seventy-two hour period in which activity was monitored. However, in the majority of BoNT

dosed cultures, a decrease in PGC values was observed across the majority of the network 48h to

72h after treatment. This suggests a decrease in overall network connectivity over time.

BoNT Detection Algorithm

A BoNT detection algorithm was developed based upon three criteria: the post stimulation

spike rate, the number of local activity peaks in post stimulation response, and average change in

connectivity. All cultures were evaluated with respect to baseline activity. Based on these

criteria, cultures received a score from 0 to 3 with higher scores indicating likely BoNT

intoxication. Cultures with score two or above were considered to be intoxicated with BoNT.

To assess changes in the late phase spike rate, spikes after each burst were sorted into 10

ms bins with an eight ms overlap. Replicates of the same stimulation site were then averaged

together to yield the average response for each stimulation. Bins from 100 to 300 ms post

stimulation were summed together to obtain late phase stimulation response. If 60 percent or

more stimulation sites had a 75% or greater increase over the baseline response the late phase












A 24h Post Treatment


- a I


- -: I


A 9


! *I -- i i J


40 II -- i i [- .
60
20- -_- i
10 20 30 40 50 60 10 20 30 40 50 60 10 20 30 40 50 60
Source Channel


B 24h Post Treatment
0

2 I
OF~


48h Post Treatment


72h Post Treatment


20[ iL
40 -- -
60L ~ 1 L
20
40
60
,, I,


4 02 0 I t


2O
20
40! -
2ot- ri! -. -: [

620 F


10 20 30 40 50 60


* ==. !fi~~ii1 K


ti -


z 7


777

10 20 30 40 50 60
Source Channel


2 -1.5 -1 -0 5 0 05 1 1 52


Figure 5-3. GC difference plot after dosage with carrier (A), and 1000 pg BoNT A (B). Results
shown are for dish 9299 as a representative of results obtained for fourteen carrier
dosed cultures and fifteen BoNT A dosed cultures. The color map indicates the
difference in the Granger Causality score from baseline activity. The x axis indicates
the source channel, and the y axis indicates response channel. Each box represents the
response for a single probe channel (1-10). After dosage with BoNT A connectivity
as evaluated using Granger's Causality tended to decrease. After dosage with carrier
solution connectivity in some networks also decreased although changes were less
consistent between stimulation sites. About 25% of the time a large decrease in
connectivity was observed within 72h after dosage with BoNT carrier. In contrast, a
large decrease in connectivity was observed about 75% of the time after dosage with
BoNT A.


IF


- L

-F^


-3 r


F


lI


I

F


i -77n


I- tA~4 ~ia


10 20 30 40 50 60
10 70 30 40 t^O RO


I ~ ~ ~ ~ ~ = l T -T i: -


' = i r "


r


: : 3


--I


48h Post Treatment


72h Post Treatment


. -f


I FT-


Z : -


7 il


c~-- -~-~~









stimulation response was scored as 1. Otherwise the late phase stimulation response was scored

as 0.

The number of local maxima was determined using the grand average of stinimulation

responses. Results were smoothed using a first-order Butterworth filter with a cutoff frequency

of 50 Hz. Local maxima were detected as a change in slope from positive to negative. Each

recording time was scored as one if the number of local maxima detected exceeded the number

of local maxima in the baseline recording by one. Figure 5-4 shows the unfiltered and filtered

post-stimulation response for control dish 7079 and 200 ng BoNT A dosed dish 9299 as an

example of typical results.

Changes in connectivity were evaluated using Granger Causality for each stimulation site.

Results were summed across stimulation sites to yield a 60 by 60 matrix probe verses response

matrix. As was noted previously, a large decrease in connectivity was observed in the majority of

the cultures. Thus, if 80% of probe-response sites that exhibited an absolute change in PGC

value greater than 2 were decreased (-2.0 or less), the connectivity score was 1. Otherwise

connectivity changes were scored as 0.













A. Baseline 7079


500

400



200

1o0

0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time Post Treatment



Figure 5-4. Average stimulation response (blue) and smoothed stimulation response (red) for
control dish 7079 (A D) and dish 9299 after dosage with 200 pg BoNT A (E H).
As a criterion for the BoNT toxin detection algorithm the average response to
stimulation was smoothed using a first-order Butterworth filter with a cutoff
frequency of 50 Hz and the number of local maxima was detected using the change in
the slope. One local maxima was detected for control dish 7079 during the Baseline,
24h, 48h, and 72h recordings so it did not meet the criteria for BoNT detection. In
contrast while one local maxima was detected during baseline recording of dish 9272,
after dosage with 200 pg BoNT A two local maxima were detected at 24h, and three
local maxima were detected for the 48h and 72h recordings.


B. 24h Carrier Treatment 7079
Average Response
Smoothed Response


S172


F. 24h BoNT A Treatment 9299













H. 72h BoNT A Treatment 9299


G. 48h BoNT A Treatment 9299
600,









Results of the BoNT detection algorithm are summarized in Table 5-1. No false positive

results were obtained for any of the control cultures. However, only 33% of BoNT A dosed

cultures were detected as positive 24h after dosage, while at 72h, all of the BoNT dosed cultures

were identified. This demonstrates that stimulation induced bursts, are suitable for BoNT

detection. Bursts induced through stimulation are stable over time, and require shorter recording

periods to obtain sufficient data for BoNT detection. Furthermore, variability due to poor burst

detection, and variability in activity over time are drastically reduced using elicited activity.



Table 5-1. Identification of BoNT in unknown cultures
Number of
Time Post positive Number of Number of Number of False
Dose Stim cultures negative cultures False Positives Negatives
24h 2/5 4/4 0/4 3/5
200 pg 48h 4/5 4/4 0/4 1/5
72h 5/5 4/4 0/4 0/5

24h 3/4 4/4 0/4 1/4
600 pg 48h 3/4 4/4 0/4 1/4
72h 4/4 4/4 0/4 0/4

24h 1/6 6/6 0/6 5/6
1000 pg 48h 6/6 6/6 0/6 6/6
72h 6/6 6/6 0/6 6/6

24h 5/15 14/14 0/14 9/15
Overall 48h 13/15 14/14 0/14 2/15
72h 15/15 13/14 0/14 0/15









CHAPTER 6
CONCLUSION AND FUTURE WORK

Future Work

Reduction of detection time

Currently, this method of toxin detection requires 24 to 72h. While this is longer than rapid

detection methods such as ELISA and protease activity assays, it is comparable to the mouse

bioassay. Additionally, many rapid methods rely on enrichment phases that drastically increase

time to detection. Thus, despite the time to detection, this method may still be suitable as a

replacement to the mouse bioassay. However, reducing time required for BoNT detection would

drastically improve use of cortical networks as a practical biosensor.

Slow toxin uptake is a likely explanation for the time required before changes in activity

are detected. Recent studies have indicated that botulism toxin uptake is coupled to vesicle

recycling (Dong et al., 2006; Verderio et al., 2006). Therefore it may be possible to increase the

rate of vesicle recycling, increasing uptake and hence, decrease the detection time for this toxin.

For example, Keller et al., (2004) used elevated potassium media for studies of BoNT uptake in

spinal cord cultures. In this case, optimal BoNT A uptake was achieved after 4 minutes of toxin

exposure in elevated potassium media with a significant level of SNAP-25 cleavage

demonstrated at 2.5 hours after toxin exposure. While less than twenty percent of SNAP-25 was

cleaved in media containing 3 mM KC1, more than fifty percent was cleaved in media containing

50 to 80 mM KC1. However, concentrations of BoNT A used in the current study were two to

four orders of magnitude smaller than concentrations used by Keller et al., (2004). By increasing

potassium in the media during toxin dosing and subsequently increasing native activity in our

cultures, time for toxin detection may be drastically reduced. Alternatively, repeated stimulation

may also increase rates of vesicle recycling. Furthermore, repeated measures provided by more









stimulations would provide a more stable and higher resolution measure than provided in this

study. This may have the added advantage of targeting BoNT to active synapses of a specific

pathway. However, changes induced by repeated stimulation would have to be carefully

distinguished from toxin activity.

Finally, inhibition of gamma-aminobutyric acid (GABA)ergic synapses using a treatment

such a biccuculine may be used to increase network activity. Since the effect of biccuculine is

reversible, temporary dosage of the network with biccuculine may be sufficient to elevate network

activity, enabling more rapid BoNT uptake.

Serotype differentiation

This work was limited to characterization of BoNT A on network activity. However,

identification of all serotypes is ideal for a BoNT biosensor. It may be possible to distinguish

BoNT serotypes based solely upon network activity. Alternatively, a neutralization assay may be

used to determine toxin serotypes. Similar to the mouse bioassay, samples would be positive for

BoNT A serotype when the base-line recording is altered for all samples except those containing

antiserum for the correct serotype.

Use of neural networks with crude samples

This work was limited to use of pharmaceutical preparations of botulinum type A toxins.

However, for neural networks to be used as practical biosensors they must be applicable for use

with crude samples, such as detection of toxin in food. Specific challenges to the use of MEA

cultures with crude samples include sensitivity of the cultured neurons to changes in pH and ion

concentrations, sensitivity of cultures to contamination by food materials, and sensitivity of

cultures to infection. Thus careful preparation of samples is necessary to address these issues.

Samples will likely need to be centrifuged and filtered to minimize contamination and infection

of neural cultures.









Refinement of MEAs
Further refinement of the hardware and protocols used will drastically improve the use of

MEAs as biosensors. As computer processing speed continues to improve it may be possible to

develop arrays of microelectrode arrays, in which sister cultures could be recorded from

simultaneously. This could enable simultaneous analysis of samples and controls, along with

automated toxin detection using algorithms similar to the one used in this work. Alternatively,

limiting intoxication to only a portion of the cortical network may enable a single culture to serve

as both a test subject and control. This may be achieved through micro-fluidic steering.

Additionally, changes to MEAs which reduce cost and increase durability are desirable prior to

commercial application as a biosensor.

Alternative and immortal cell lines

Detection of BoNT toxin and reduction of detection time may be achieved using an

alternative cell line. Many cell cultures are suitable for MEAs including cardiac myocytes, spinal

neurons, and retinal ganglion cells. It may be that one of these alternatives is more suitable for

BoNT detection. Additionally, an adult tissue slice may be used as an alternative to dissociated

tissue. Finally, as an alternative to animal derived tissue cultures it may be possible to use an

immortal cell-line, completely eliminating animal use. This may also reduce variability between

samples. Furthermore, neurons may be altered to express a large number of toxin receptor sites,

improving toxin uptake.

Conclusion

Use of cortical networks as a biosensor for botulinum toxin enables detection of only intact

physiologically active toxin that is capable of binding, internalization and substrate cleavage.

Thus, similar to the mouse bioassay this assay may be suitable for evaluation of biological

activity of the toxin. This is advantageous to alternative assays such as ELISA and PCR since









they are incapable of assessing toxin activity, and may even be positive in the absence of active

toxin. This work demonstrates that cortical tissue grown on MEAs provides a suitable substrate

for a cell-based BoNT A biosensor.

While this study is currently limited to exploring the effect of BoNT A on network activity

this technique could also be useful for detecting and perhaps even differentiating the effects of

other BoNT strains (types B-G) as well as detection of a plethora of other neurotoxins.

Furthermore, while the mouse bioassay requires several animals for a single assay, with this

technique a single animal could provide tissue for multiple neural network based assays on

MEAs. Since the cortical tissue used in this work is commercially available no onsite mouse

rearing facilities are necessary.

This work has established the time course and a suitable concentration range for BoNT

detection using neural networks. It remains conceivable that an assay based on this technique

could provide a powerful method for detecting a wide variety of food borne agents including

both known and currently unknown biological threats.










APPENDIX A
SPONTANEOUS ACTIVITY IN BONT A AND CARRIER DOSED CULTURES

Table A-i. Average spike height for neural cultures dosed with BoNT carrier or 200pg BoNT A


Time Post
Treatment
Base
12h
24h
36h
48h
72h
1wk

Table A-2.


Time Post
Treatment
Base
12h
24h
36h
48h
72h
1wk


Control
Mean
N=5
(mV)
-3.486
-3.388
-3.059
-3.316
-3.566
-3.737
-3.937


BoNT A
Mean
N=9
(mV)
-3.964
-3.781
-3.991
-3.864
-3.726
-3.358
-2.992


Average spike width for


Control
Mean
N=5
(ms)
0.323
0.315
0.286
0.284
0.264
0.302
0.332


BoNT A
Mean
N=9
(ms)
0.348
0.333
0.318
0.311
0.308
0.299
0.323


Control
Standard
Deviation
(mV)
0.125
0.113
0.119
0.124
0.107
0.095
0.129


BoNT A
Standard
Deviation
(mV)
0.105
0.097
0.096
0.107
0.092
0.099
0.130


neural cultures dosed with BoNT carrier or 200 pg BoNT A


Control
Standard
Deviation
(ms)
0.125
0.113
0.119
0.124
0.107
0.095
0.129


BoNT A
Standard
Deviation
(ms)
0.105
0.097
0.096
0.107
0.092
0.099
0.130


Table A-3. Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg
BoNT A
Control BoNT A Control BoNT A
Mean Mean Standard Standard
Time Post N=5 N=9 Deviation Deviation
Treatment (ms) (ms) (ms) (ms)
Base 16.926 21.762 25.243 31.459
12h 22.401 21.286 25.700 30.275
24h 27.534 29.675 37.940 41.908
36h 26.366 20.703 22.438 27.572
48h 31.200 29.191 30.831 45.076
72h 40.353 27.376 35.553 42.268
0
1wk 29.584 29.518 34.879 51.924










Table A-4. Average burst duration for neural cultures dosed with BoNT carrier or 20pg BoNT A
Control BoNT A Control BoNT A


Time Post
Treatment
Base
12h
24h
36h
48h
72h
1wk


Mean
N=5
0.204
0.204
0.216
0.225
0.253
0.249
0.172


Mean
N=9
0.231
0.256
0.309
0.311
0.368
0.436
0.394


Standard
Deviation
0.080
0.082
0.090
0.096
0.136
0.121
0.073


Standard
Deviation
0.108
0.111
0.130
0.126
0.138
0.197
0.204


Table A-5. Average inter-burst interval for neural cultures dosed with BoNT carrier or 200 pg
BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation


Base
12h
24h
36h
48h
72h
1wk


15.14
17.52
19.21
19.50
21.37
22.53
13.50


11.52
13.61
17.59
16.06
15.29
17.30
16.62


12.13
11.20
7.26
9.36
6.99
8.27
7.12


5.37
8.14
6.69
8.94
4.41
8.65
5.04


Table A-6. Average number of spikes per burst for neural cultures dosed with BoNT carrier or
200 pg BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation


Base
12h
24h
36h
48h
72h
1wk


1986
2012
2012
2102
2066
1985
1390


1713
1761
2226
2499
2793
3380
3329


2546
2455
2442
2597
2673
2734
2061


1628
1685
1907
2139
2428
2929
2959


Table A-7. Average spike height for neural cultures dosed with BoNT carrier or 600pg BoNT A

Control BoNT A


Time
Post
Treatment
Base
12h
24h
36h
48h
72h


Mean
N=4
(mV)
-2.9183
-2.7094
-2.7940
-2.3781
-2.5932
-2.9813


Mean
N=4
(mV)
-2.7380
-2.4200
-1.4691
-1.4858
-2.5963
-2.8765


Control
Standard
Deviation
1.1164
0.9534
0.7612
0.7816
0.9513
1.1365


BoNT A
Standard
Deviation
0.4644
0.3139
2.7080
2.4462
0.6585
0.4444










Table A-8. Average spike width for neural cultures dosed with BoNT carrier or 600pg BoNT A

Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (ms) (ms) Deviation Deviation
Base 0.1622 0.1725 0.0195 0.0051
12h 0.1508 0.1749 0.0144 0.0094
24h 0.1554 0.1430 0.0105 0.0074
36h 0.1711 0.1702 0.0177 0.0233
48h 0.1743 0.1528 0.0143 0.0185
72h 0.1820 0.1386 0.0095 0.0167



Table A-9. Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg
BoNT A

Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (ms) (ms) Deviation Deviation


Base
12h
24h
36h
48h
72h


0.0137
0.0150
0.0154
0.0102
0.0131
0.0087


0.0095
0.0121
0.0088
0.0084
0.0131
0.0192


0.0049
0.0039
0.0050
0.0028
0.0027
0.0027


0.0005
0.0036
0.0034
0.0033
0.0053
0.0069


Table A-10. Average burst duration for neural cultures dosed with BoNT carrier or 600pg BoNT


Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (ms) (ms) Deviation Deviation
Base 124.2260 118.3149 21.2908 21.7382
12h 129.1907 130.8544 25.6628 36.8718
24h 126.0828 137.1658 26.4672 49.5016
36h 126.2304 139.1466 23.9707 41.6674
48h 133.4856 147.6418 15.9434 26.3732
72h 122.2815 53.3322 13.3289 5.1737










Table A-11. Average inter-burst interval for neural cultures dosed with BoNT carrier or 600 pg
BoNT A

Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (ms) (ms) Deviation Deviation
Base 5.066 3.911 1.487 0.470
12h 5.697 4.821 1.715 2.166
24h 5.909 2.307 0.798 0.330
36h 4.434 2.006 1.329 0.493
48h 5.749 2.492 1.303 0.686
72h 3.464 3.080 0.586 0.734


Table A-12. Average number of spikes per burst for neural cultures dosed with BoNT carrier or
600 pg BoNT A
Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (Sk/Bst) (Sk/Bst) Deviation Deviation
Base 278.5 308.4 114.7 58.0
12h 288.9 297.1 116.2 87.6
24h 285.1 235.3 91.7 155.7
36h 349.5 245.9 131.6 207.0
48h 354.4 148.7 91.3 48.1
72h 329.9 72.0 137.0 21.3


Table A-13. Average number of channels per burst for neural cultures dosed with BoNT carrier
or 600 pg BoNT A
Control BoNT A
Time Mean Mean Control BoNT A
Post N=4 N=4 Standard Standard
Treatment (Ch/Bst) (Ch/Bst) Deviation Deviation
Base 37.1935 39.8215 7.4324 4.9276
12h 37.9359 38.3629 6.2427 6.7225
24h 38.1660 36.6311 7.1915 7.9138
36h 40.2464 36.4478 5.0727 9.0970
48h 41.2229 33.2713 4.4958 3.5410
72h 39.6653 23.5968 3.6450 2.2500










Table A-14. Average burst duration for neural cultures dosed with BoNT carrier or 1000 pg
BoNT A

Control BoNT A Control BoNT A
Time Mean Mean Standard Standard
Post N=4 N=4 Deviation Deviation
Treatment (ms) (ms) (ms) (ms)
Base 112.5372 150.2687 24.4398 46.1023
12h 124.3151 165.4070 25.7942 51.9701
24h 124.0459 178.6280 34.2769 67.6356
36h 132.7705 202.2882 43.6598 103.6945
48h 127.0323 169.7188 36.1385 66.0071
72h 124.7127 203.5174 16.2268 30.4046


Table A-15. Average inter-burst interval for neural cultures dosed with BoNT carrier or 1000 pg
BoNT A

Control BoNT A Control BoNT A
Time Mean Mean Standard Standard
Post N=4 N=4 Deviation Deviation
Treatment (ms) (ms) (ms) (ms)
Base 5.5481 6.4824 3.2419 2.9963
12h 5.2274 7.2141 3.4010 3.3540
24h 5.1237 4.7685 3.4082 2.8610
36h 5.4598 5.3077 4.4351 4.1595
48h 4.0069 3.9431 3.2410 1.4940
72h 5.3891 3.9288 3.8194 2.7445


Table A-16. Average number of spikes per burst for neural cultures dosed with BoNT carrier or
1000 pg BoNT A

Control BoNT A Control BoNT A
Time Mean Mean Standard Standard
Post N=4 N=4 Deviation Deviation
Treatment (Sk/Bst) (Sk/Bst) (Sk/Bst) (Sk/Bst)
Base 341.9895 491.2265 119.2052 308.0769
12h 342.2862 400.8340 122.9619 262.8882
24h 304.1719 428.1824 103.9649 254.7530
36h 301.6341 535.5256 141.3596 434.1746
48h 246.6684 331.2163 142.3897 178.8936
72h 284.6426 310.8945 93.1311 124.6812










Table A-17. Average number of channels per burst for neural cultures dosed with BoNT carrier
or 1000 pg BoNT A

Control BoNT A Control BoNT A
Time Mean Mean Standard Standard
Post N=4 N=4 Deviation Deviation
Treatment (Ch/Bst) (Ch/Bst) (Ch/Bst) (Ch/Bst)
Base 29.6113 32.5754 10.7625 14.8220
12h 29.3288 31.3244 10.3198 14.0974
24h 28.6892 31.5278 11.0753 13.0683
36h 28.1165 32.7809 9.7323 13.8754
48h 26.1800 32.3135 10.1581 13.8349
72h 27.8608 31.4127 9.4488 12.1135










APPENDIX B
ELLICITED ACTIVITY IN BONT A AND CARRIER DOSED CULTURES

Table B-1. Average spikes per stimulation for neural cultures dosed with BoNT carrier or 200pg
BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation


Base
12h
24h
36h
48h
72h


1842.0
1846.9
1864.5
1885.2
1971.9
1893.3


1743.7
1768.5
1951.5
2201.7
2542.8
2798.1


2522.9
2426.9
2439.3
2472.0
2638.2
2580.3


1631.1
1617.7
1747.9
1924.8
2355.4
2602.1


Table B-2. Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
200pg BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation
Base 73.3 77.6 18.8 13.7
12h 79.5 83.0 12.6 9.3
24h 82.0 84.1 7.7 22.3
36h 83.9 84.6 4.9 14.9
48h 81.0 87.7 10.1 4.9
72h 79.2 87.5 12.8 13.9



Table B-3. Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 200pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=5
1.8
5.9
5.6
6.5
-2.8


BoNT A
Mean
N=9
5.2
16.3
25.6
22.3
24.6


Control
Standard
Deviation
9.6
15.7
20.2
38.2
26.3


BoNT A
Standard
Deviation
12.9
37.5
39.9
47.7
51.9










Table B-4. Percent change in the number of spikes 150 to 300 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 200pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=5
36.0
-22.6
-18.0
-1.4
26.8


BoNT A
Mean
N=9
58.7
72.4
117.7
281.2
671.9


Control
Standard
Deviation
71.8
33.5
34.5
41.5
73.3


BoNT A
Standard
Deviation
90.0
116.0
114.5
261.3
765.8


Table B-5. Average spikes per stimulation for neural
pg BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation


Base
12h
24h
36h
48h
72h


157.7
179.5
213.5
204.8
202.8
223.3


272.0
309.1
260.9
216.3
244.5
354.8


85.0
107.7
88.9
106.6
46.6
77.7


cultures dosed with BoNT carrier or 600


79.7
37.3
53.3
63.6
39.9
220.8


Table B-6. Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
600 pg BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=5 N=9 Deviation Deviation
Base 80.8 78.3 4.1 2.1
12h 83.8 81.5 3.8 1.7
24h 81.8 85.4 2.8 4.4
36h 80.3 86.0 4.5 3.5
48h 82.3 85.4 4.5 0.8
72h 79.9 85.8 3.0 2.9










Table B-7. Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 600pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=4
13.9
46.6
36.5
48.6
55.7


BoNT A
Mean
N=4
18.8
-4.6
-25.3
-13.8
1.5


Control
Standard
Deviation
15.2
36.5
32.5
56.1
39.0


BoNT A
Standard
Deviation
30.4
33.9
42.6
23.8
29.4


Table B-8. Percent change in the number of spikes 150 to 300 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 600pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=4
118.9
90.6
59.8
96.1
85.7


BoNT A
Mean
N=4
176.7
653.1
1107.2
1059.3
1923.3


Control
Standard
Deviation
110.6
63.0
28.8
47.2
52.6


BoNT A
Standard
Deviation
116.9
674.3
1312.1
471.5
1108.1


Table B-9. Average spikes per stimulation for neural cultures dosed with BoNT carrier or 1000
pg BoNT A


Time Post
Treatment
Base
12h
24h
36h
48h
72h


Control
Mean
N=6
374.2
396.7
413.7
410.7
360.8
342.1


BoNT A
Mean
N=6
324.6
390.6
447.8
329.2
303.5
253.0


Control
Standard
Deviation
235.8
248.8
302.0
257.8
221.0
204.7


BoNT A
Standard
Deviation
188.2
225.7
368.0
182.1
187.8
177.7










Table B-10. Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or
1000 pg BoNT A
Control BoNT A Control BoNT A
Time Post Mean Mean Standard Standard
Treatment N=6 N=6 Deviation Deviation
Base 86.1 87.6 2.7 3.5
12h 86.1 86.2 3.0 3.0
24h 83.2 90.0 5.5 1.5
36h 85.3 89.5 6.0 3.0
48h 82.2 88.9 5.7 2.5
72h 83.6 91.5 5.1 3.7



Table B-l 1. Percent change in the number of spikes 0 to 150 ms post stimulation from baseline
activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=6
7.0
10.2
12.2
0.5
-1.8


BoNT A
Mean
N=6
20.2
39.1
1.8
-9.0
4.3


Control
Standard
Deviation
40.2
104.4
31.0
38.9
78.5


BoNT A
Standard
Deviation
7.6
5.5
7.6
10.4
5.0


Table B-12. Percent change in the number of spikes 100 to 300 ms post stimulation from
baseline activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A


Time Post
Treatment
12h
24h
36h
48h
72h


Control
Mean
N=6
19.4
32.3
29.1
22.1
36.6


BoNT A
Mean
N=6
17.5
231.8
196.1
284.8
547.2


Control
Standard
Deviation
20.6
423.2
192.4
255.5
455.8


BoNT A
Standard
Deviation
29.4
47.3
18.9
87.9
211.1











APPENDIX C
BURST DETECTION ALGORITHM

The following algorithm, Bst Det is a burst detection algorithm written for use with Matlab and
requires data in the form of a spike file from MEABench. The program returns burst statistics in
the variable datastorage, and burst times in a 3 x N matrix, in which the first column is the burst
start, the second column is the burst stop, and the third column is the burst duration.

function [data storage,bst] = Bst Det(datal)
% THIS PROGRAM SORTS ALL THE SPIKES INTO A HISTOGRAM WITH BIN LENGTHS
% SPECIFIED BY BIN S. A THRESHOLD IS SET BASED UPON THE MEAN SPIKES PER
% BIN. A BURST IS DETECTED EACH TIME THE SPIKES IN A BIN EXCEED THE
% THRESHOLD. AFTER A BURST IS DETECTED AT LEAST 0.15 SECONDs MUST ELAPSE
BEFORE
% A NEW BURST CAN BE DETECTED.THE BURST END IS DETECTED AS THE FIRST 0.5 MS
% BIN AFTER A BURST WITH NO SPIKES IN IT.

%SET THE INTERVAL OVER WHICH YOU ARE INTERESTED IN CALCULATING
%PARAMETERS IN SECONDS
strt = 1; %start 10 min into the recording
stp = 1201; %stop 20 min into the recording

tic
bin s = 0.0005;

P = find(datal.time > strt & datal.time < stp);
data.time = datal.time(P);
data.channel = datal.channel(P);

N = strt-bin s:bin s:stp-bin s;
y = hist(data.time, N);
R = mean(y)+ std(y)
if R < 0.4
thresh = 1
elseif R > .4 && R < 1
thresh = 2
else
thresh = 3
end

%Bin size = 0.01;
Bin size = 0.01;

C = find(y > thresh);
D = size(C,2);
Burst = 0;
p = 1;
bst(l,l) = 0;




for i = 1:D-1 %Detects bursts as consecutive
bins











if C(l,i+l) C(l,i) == 1 && Burst ==
bst(l,p) = N(1,C(l,i));
p = p +;
Burst = 1;
elseif Burst == 1 && C(l,i+l) C(l,i)
if N(1,C(l,i))-bst(l,p-1) > 0.20
as 200 ms
Burst = 0;
end


exceeding the threshold


S2
sets the min time between bursts


end


%Detect the end of the burst
i = 1;
Cntr = size(bst, 2);
while i < Cntr +1
Bin size2(i) = Bin size;
Sk post = find(data.time > bst(l,i) & data.time < bst(l,i) + 5);
ch = 0;
while ch == 0
r = hist(data.time(Sk post), 5/Bin size2(i));
D = find(r == 0);
if numel(D) == 0
Bin size2(i) = Bin size2(i) .005;
elseif Bin size2(i) <= 0
ch = 1;
else
ch = 1;
end


if Bin size2(i) <= 0
bst(2,i) = bst(l,i) + 5;
else
bst(2,i) = bst(l,i) + D(1)*Bin size2(i);
end


i = i +1;


%Check that the same burst has not been detected more than once
Error check = bst(1,2:Cntr) bst(2,1:Cntr-1);
C = find(Error check <= 0);
if numel(C) > 0
for i = 1:size(C,2)+1
if i == 1


newbst
elseif i >
newbst
elseif i ==
newbst
end


= bst(:,l:C(i));
1 && i < size(C,2) + 1
= [newbst(:,:) bst(:,C(i-1)+2:C(i))];
Ssize(C,2) + 1
= [newbst(:,:) bst(:, C(i-1)+2:Cntr)];














bst = newbst;
end

%Quick plot to check BD alogrithm
i = 1;
N = (stp-strt)*1000;
new time = linspace(strt, stp, 2*N);
BUR = zeros(size(new time));
while i < size(bst(1,:), 2)+1
Row Col = find(new time > bst(l,i) & new time < bst(2,i));
BUR(Row Col) = 60;
i = i+1;
end

%PRINT THE FIGURES IN 100 SECOND INTERVALS
I=1;%round(strt/100)+1;
while I < 13 % round(stp/100)+1
figure()
plot(data.time, data.channel,'.','markersize',1),axis([1+((I-1)*100)
100+((I-1)*100) 0 64]);
hold on
plot(new time, BUR,'r-');
I = I +1;

end

%Calc the bst duration
bst(3,:) = bst(2,:)-bst(l,:);

D = find(bst(l,:) > strt & bst(l,:) < stp);

%CALCULATE THE BURST DURATION
dur in int = bst(3,D);% results(1,1);
data storage(1,1)= mean(dur in int);
data storage(2,1) = std(dur in int);
data storage(3,1) = ((std(dur in int))^2)/sqrt(size(D,2));

%SPIKE RATE
SPKS = find(data.time>strt & data.time TOT SPKS = size(SPKS, 1);
data storage(4,1)= TOT SPKS/(stp-strt);%-results(3,1);
data storage(5,1) = TOT SPKS/((stp-strt)/60);%-results(4,1);

%MEAN SPIKES/BST & CHAN/BST
M = size(D,2);
Chan Cnt = zeros(M, 1);

for i = 1:M
Spike bsts = find(data.time>bst(1,D(i))&data.time< bst(2,D(i)));
Sk per bst(i) = size(Spike bsts,1);

for k = 1:60
B = find(data.channel(Spike bsts) == k-1);











if numel(B > 0)
Chan Cnt(i,1) =Chan Cnt(i,1)+1;
end


data storage(6,1)
data storage(7,1)
data storage(8,1)

%PERCENT SPIKES IN
data storage(9,1)
data storage(10,1)


Smean(Sk per bst);%-results(5,1);
Smean(Chan Cnt(:,1));%-results(6,1);
Sstd(Chan Cnt(:,1));%-results(6,1));

BURSTS
= sum(Sk per bst);
= 100*data storage(9,1)/TOT SPKS;% -


%MEAN BURST RATE (PER SECOND)
data storage(11,1) = M/(stp-strt);%-results(10,1);
data storage(12,1) = M/((stp-strt)/60);%-results(11,1);


%Burst Period START TO START
P = size(D,2);
IBI st st = (bst(l,D(1,2:P))-bst(1,D(1,1:P-1)));
data storage(13,1) = mean(IBI st st);
data storage(14,1) = std(IBI st st);


-results(12,1);


%INTERBURST INTERVAL STOP TO START
IBI sp st = (bst(l,D(1,2:P))-bst(2,D(1,1:P-1)));%-results(14,1);
data storage(15,1) = mean(IBI sp st);
data storage(16,1) = std(IBI sp st);


storage(17,1)
storage(18,1)
storage(19,1)
storage(20,1)


std(Sk per bst);
(std(Sk per bst))^2/sqrt(size(D,2));
thresh;
mean(Bin size2,2);


results(9,1);


data
data
data
data











APPENDIX D
BONT A DETECTION ALGORITHM

The following algorithm, BotCheckStim is an algorithm written for use with Matlab and
requires data in the form of a spike file from MEABench. This program returns the variable tally,
a 3 by 1 matrix which contains a score from 1 to 3 indicating the likely hood of BoNT
intoxication. A score greater than 2 is considered positive for BoNT intoxication. BotCheck is
the main program. Functions written by me are listed below Botcheckstim in the order they are
called

function tally = Bot check stim(fnl, fn2, fn3, fn4)

[Base stimdat, Base count, Base avg, Base results] = Score Stim(fnl);

[B24h stimdat, B24h count, B24h avg, B24h results] = Score Stim(fn2);

[B48h stimdat, B48h count, B48h avg, B48h results] = Score Stim(fn3);

[B72h stimdat, B72h count, B72h avg, B72h results] = Score Stim(fn4);

tally(l:3,1) = 0;


% Criteria #1, it was observed that in a large number of BoNT dosed cultures%
% there was a drastic decrease in connectivity associated with BoNT %
o % intoxication. This criteria check whether the majority of channels had a%
% decrease in connectivity as determined using Granger's Causality. %

fnl = strcat(fnl,' 2');
fn2 = strcat(fn2,' 2');
fn3 = strcat(fn3,' 2');
fn4 = strcat(fn4,' 2');
scorel = Eval connect(fnl,fn2,fn3,fn4);
for i = 1:3
if scorel(i,l) > 0.80
tally(i,l) = tally(i,l)+l;
end
end


% Criteria #2, Check if there is an increase in activity observed in the
% late phase of the burst. If there a 100% or greater increase in the late%
% phase for the majority of the stim sites, the tally is increased by one.%

check late spikes(1,:) = B24h results Base results;
check late spikes(2,:) = B48h results Base results;
check late spikes(3,:) = B72h results Base results;

K = 0.75* mean(Base results(1,:));
for j = 1:3
row col = find(check late spikes(j,:) > K );
score2(j,l) = size(row col,2)/size(check late spikes,2);
if score2(j,l) > 0.6












end
end




% Criteria #3, Multiple peaks of activity the grand average of all the
% is used to determine if there are multiple local maxima present in the
% response to stimulation. If there is an increase in local maxima
% relative to baseline activity, the tally is increased.


Grand avg(1,1:150) = mean(Base avg,1);
Grand avg(2,1:150) = mean(B24h avg,l);
Grand avg(3,1:150) = mean(B48h avg,l);
Grand avg(4,1:150) = mean(B72h avg,l);
for i = 1:4
[b,a] = butter(l,0.05);
Smooth avg = filter(b,a,Grand avg(i,1:100));
figure()
plot(Grand avg(i,:),'b-');
hold on
plot(Smooth avg(1,:),'r-');
Inflect detect = Smooth avg(:, 2:75) Smoo
row col = find(Inflect detect > 0);
P = size(row col,2);
find local max = row col(l,2:P) row col(l,
row col2 = find(find local max > 3);
if numel(row col2) > 0
max count(i,l)= size(row col2,2)+1;
else
max count(i,l) = 1;
end
end


for i = 1:3
score3(i,l) = max count(i+1,1)
if score3(i,1) > 0
tally(i,l) = tally(i,l) +1;
end


th avg(:, 1: 74);


1:P-1) ;


max count(1,1);


tally(j,l)


tally(j,l)+l;











The following algorithm, Score_Stim is an algorithm written for use with Matlab. This program
loads data in the form of spike files written by MEA bench. It determines the stimulation times
and channels from a data file recorded at the time the stimulation was performed (stim_dat),
finds the number of stimulations eliciting a burst and returns a count for each stimulation site
(count), determines the average stimulation response for each stimulation site (avg_stim), and
determines the total late phase response (100 to 300 ms post simulation) for each stimulation site

function [stim dat, count, avg stim, results] = Score Stim(fn)
tic

fn spike = strcat(fn,'.spike');

%LOADS DATA FROM SPIKE FILE
data = PCloadspike noc(fn spike, 3, 25);

fn data = strcat(fn,'.data');

%DETERMINES THE STIM TIMES AND SORTS SPIKES INTO 200 MS bins
[stim time, Final stim time] = Stimbstdet(data);

%SORT DATA BY STIM SITE, RETURN A 100 BY 3 MATRIX WITH COL 1 BEING THE STIM
%REPLICATE # (1-10), COL 2 BEING THE STIM SITE #(1-60), AND COL 3 BEING THE
%STIM TIME
[stim dat] = Stimdata(data, fn data);

% DETERMINES THE TIMES THE CULTURe WAS ALREADY BURSTING PRIOR TO THE BURST
ONSET
% AND THE TIMES WHEN THE CULTURE DID NOT BURST AFTER THE STIMULATION,
STIM DAT
% IS ADJUSTED SO THAT THESE TIMES ARE SET TO ZERO, COUNT SHOWS THE TOTAL
% NUMBER OF STIMS ELLICITING A BURST FOR EACH STIM SITE, AVG STIM IS THE
% AVERAGE STIM RESPONSE UP TO 300 MS AFTER THE BURST, WHERE A 10 ms SLIDING
% WINDOW IS USED

[stim dat, count, avg stim] = stim response(data, Final stim time,
stim dat);

%AVERAGE STIM RESPONSE 100 to 300 MS AFTER THE STIMULATION, (LATE PHASE
%STIMULATION RESPONSE)
N = size(avg stim,l);
for i = 1:N
results(i,l) = sum(avg stim(i, 50:150));
end











The following algorithm, Stimbstdet is an algorithm written for use with Matlab. This program
useses data in the form of spike files written by MEA bench.

function [stim time, Final stim time] = Stimbstdet(datal)
% THIS PROGRAM SORTS ALL THE SPIKES INTO A HISTOGRAM WITH BIN LENGTHS
% SPECIFIED BY BIN S. A THRESHOLD IS SET BASED UPON THE MEAN SPIKES PER
% BIN. A BURST IS DETECTED EACH TIME THE SPIKES IN A BIN EXCEED THE
% THRESHOLD. AFTER A BURST IS DETECTED AT LEAST 0.15 SECONDs MUST ELAPSE
BEFORE
% A NEW BURST CAN BE DETECTED.THE BURST END IS DETECTED AS THE FIRST 0.5 MS
% BIN AFTER A BURST WITH NO SPIKES IN IT.

row col = find(datal.channel == 59);
M = size(row col);
strt = datal.time(row col(1,1))-5;
stp = datal.time(row col(M(1,1),M(1,2)))+5;
row col = find(datal.time > strt & datal.time < stp);
data.time = datal.time(row col);
data.channel = datal.channel(row col);


tic
bin s = 0.0005;
N = strt-bin s:bin s:stp-bin s;
y = hist(data.time, N);
R = mean(y)+ std(y)
if R <= 1
thresh = 1
elseif R > 1 && R <= 1.5
thresh = 2
else
thresh = 3
end


%Find the number of stims in the recording by detecting spikes on
%channel 60
[Row Col] = find(data.channel == 60);
Num ch63 sks = size(Row Col,2);
stim time(1,1) = data.time(Row Col(1,1));
stim time(1,2) = 61;
p=2;
for i = 2:Num ch63 sks
if data.time(Row Col(l,i)) data.time(Row Col(1,i-1)) > .3 %
spikes for
stim time(p,1) = data.time(Row Col(1,i)); %up to


stim time(p,2)
p=p+l;


[Row Col] = find(data.channel =
false spikes
time chl4 = data.time(Row Col);


store all


300ms after


stim


14); %Channel 14 is the ground channel,

%Correspond to spikes on this channel











num stim = size(stim time,1);
Stim Sk cnt = zeros(1,num stim);
for i = 1:num stim %R
false spike


emove all spikes for up to 10 ms after a


[Row Col] = find(data.time > stim time(i,1)+.005 & data.time <
stim time(i,1)+.5);
stim dat = data.time(Row Col);
[Row Col] = find(time chl4 > stim time(i,1)+.005 & time chl4 <
stim time(i,1)+.5);
if numel(Row Col) > 0
disturbance = time chl4(Row Col);
stt = min(disturbance);
stp = stt +.010;
[Row Col] = find(stim dat < stt I stim dat > stp); %Find sp:


stim dat2


stim dat(Row Col);


else


stim dat2


likes


;corresponding
;the to thee stim
;excluding those
,10 ms after a
;spike on chan 14


stim dat;


end
Stim Sk cnt(i) = size(stim dat2,2);
N=ceil(max((stim dat2-stim time(i,1))/0.001));
Stim bin(i,1:N+2) = hist((stim dat2-stim time(i,1))/.001, 0:N+1);
if N+2 <300
Stim bin(i,N+2:300) = 0;


end


%Check 30ms before each stim to ensure that the culture is not already
Bursting


p=l;
prestim(p,l) = 0;
for i = 1:num stim
[Row Col] = find(data.time > stim time(i,l)
stim time(i,l)-.010);
num spikes(l,i) = size(Row Col,2);


Snum spikes;
i = 1:num stim
if num spikes(1,i)
prestim(p,l)
p=p+l;


.200 & data.time <


> 20*thresh
i;


p=l;
Thresh = mean(Stim Sk cnt)-std(Stim Sk cnt)
Stim Sk cnt

%Throw out stims that did not burst
for i = 1:size(Stim Sk cnt,2)
if Stim Sk cnt(i) > Thresh


end












Final stim bin(p, 1:200) = Stim bin(i,1:200);
Final stim time(p,l) = stim time(i,l);


p=p+l;
if prestim(l) > 0;
for j = l:size(prestim,l)
if prestim(j,l) == i
C = Final stim bin(l:p-1,1:200);
D = Final stim time(1:p-1,1);
clear Final stim bin Final stim time;
Final stim bin = C;
Final stim time = D;
clear C
end
end
end
end
end












The following algorithm, Stim Response is an algorithm written for use with Matlab. This
program useses data in the form of spike files written by MEA bench.

function [stim dat, count, stim avg] = Stim Response(data, Final stim time,
stim dat)

%Now spikes for each stim are sorted according to channel. In this case
%there are 300 ms of data, sorted into s histogram using a 10 ms moving
%window, which shifts 2 ms

R = size(stim dat,1);
N = 0;
Finaltime = 200 0.5;
G = -.5:1:Finaltime;
Win = 200;
S=0;

grangerdata = zeros(60,Win,11,10);
C = floor(Final stim time);
D = floor(stim dat(:,3));
for T = 1:R
row col = find(C == D(T,1));
if numel(row col) == 0
stim dat(T,3) = 0;
end
end

m= 1;
for T = 1:R
if stim dat(T,1) == 1
N = N+1;
if T > 1
count(m,l) = num ;
m = m+1;
end
num =0;
end

if stim dat(T,3) >0
num = num +1;

A = find(data.time >= stim dat(T,3) & data.time <= stim dat(T,3)+.300

newtime = (data.time(A)- stim dat(T,3))*1000;
newchan = data.channel(A);
B = find(newchan > 59);
if numel(B) > 0
remove false = max(newtime(B));
C = find(newtime > remove false);
P = size(C,2);
spiketime(1:P) = newtime(C);
spikechan(1:P) = newchan(C);
else
spiketime(1:P) = newtime;
spikechan(1:P) = newchan;











end
N1


num ;


G = 1;
for M =1:30
Ql=find(spiketime >=
hist data(N,G,N1) =
Q2=find(spiketime >=
hist data(N,G+1,N1)
Q3=find(spiketime >=
hist data(N,G+2,N1)
Q4=find(spiketime >=
hist data(N,G+3,N1)
Q5=find(spiketime >=
hist data(N,G+4,N1)
G = G + 5;


:(M-1)*10 & spiketime < M 10);
size(Ql,2);
:(M-1)*10+2 & spiketime < M 10+2);
= size(Q2,2);
:(M-1)*10+4 & spiketime < M 10+4);
= size(Q3,2);
:(M-1)*10+6 & spiketime < M 10+6);
= size(Q4,2);
:(M-1)*10+8 & spiketime < M 10+8);
= size(Q5,2);


clear spiketime spikechan
end
end
count(m,l) = num ;


%p = 1;
for i = 1:N
P = count(i,1);
if P > 0
dat = hist data(i,:,1:P);
st avg = mean(dat,3); collapsee across stim avg
stim avg(i,1:150) = st avg;
%p = p+l;
else
stim avg(i,1:150) = 0;











The algorithm Evalconnect loads granger causality data, which has been stored as a 60 by 60 by
N matrix in a Mat file, where N = the number of stimulation sites.


function score


Eval connect(fnl,fn2,fn3,fn4)%dose, culture)


load(fnl, 'G*')
base = abs(GrangerMatrix)-0.5;
row col = find(base <= 0);
base(row col) = 0;

load(fn2, 'G*')
B24h = abs(GrangerMatrix)-0.5;
row col = find(B24h <=0);
B24h(row col) = 0;

load(fn3, 'G*')
B48h = abs(GrangerMatrix)-0.5;
row col = find(B48h <=0);
B48h(row col) = 0;

load(fn4, 'G*')
B72h = abs(GrangerMatrix)-0.5;
row col = find(B72h <=0);
B72h(row col) = 0;


reshape(sum(base,3),60,60);
reshape(sum(B24h,3),60,60);
reshape(sum(B48h,3),60,60);
reshape(sum(B72h,3),60,60);


Pr 24h
Pr 48h
Pr 72h


% collapse across probe
collapse across probe
collapse across probe
collapse across probe


Pr Base;
Pr Base;
Pr Base;


find(Pr 24 Base(:,:) <
find(Pr 48 Base(:,:) <
find(Pr 72 Base(:,:) <


2 Pr 24 Base(:,:) > 2);
2 Pr 48 Base(:,:) > 2);
2 Pr 72 Base(:,:) > 2);


if size(row coll,1) > 2
row col2 = find(Pr 24
score(l,l) = size(row


row col2 =
score (2,1)

row col2 =
score(3,1)


Base(:,:) < -2);
col2,1)/size(row coll,l);


find(Pr 48 Base(:,:) < -2);
= size(row col2,1)/size(row col3,1);

find(Pr 72 Base(:,:) < -2);
= size(row col2,1)/size(row col4,1);


figure ()
subplot(1,3,1)
imagesc(Pr 24 Base(:,:), [-10 10]);


Pr Base =
Pr 24h =
Pr 48h =
Pr 72h =

Pr 24 Bas
Pr 48 Bas
Pr 72 Bas



row coll
row col3
row col4


e
e
e











subplot(1,3,2)
imagesc(Pr 48h Base(:,:),[-10 10]);
subplot(1,3,3)
imagesc(Pr 72h Base(:,:),[-10 10]);









REFERENCES


Ahnert-Hilger, G. and H. Bignalke. 1995. Molecular aspects of tetnus a botulism neurotoxin
poisoning. Prog Neurobiol. 46(1), 83-96.

Akbulut, D., K. Grant, and J. McLauchin. 2004. Development and application of real-time PCR
assays to detect fragments of the Clostridium botulinum types A, B, and E neurotoxin
genes for investigation of human foodborne and infant botulism. Foodborne Pathog Dis.
125(4), 247-257.

Anne, C., F. Cornille, C. Lenoir, and B. Roques. 2001. High-throughput fluorogenic assay for
determination of botulinum type B neurotoxin protease activity. Anal Biochem. 291,253-
261.

Aranda, E., M. Rodriguez, M. Asensio, and J. Cordoba. 1997. Detection of Clostridium
botulinum types A, B, E and F in foods by PCR and DNA probe. Lett Appl Microbiol.
25, 186-190.

Arnon, S., R. Schechter, T. Inglesby, D. Henderson, J. Bartlett, M. Ascher, E. Eitzen, A. Fine, J.
Hauer, M. Layton, S. Lillibridge, M. Osterholm, T. O'Toole, G. Parker, T. Perl, P.
Russell, D. Swerdlow, and K. Tonat. 2001. Botulism Toxin as a Biological Weapon.
JAMA. 285, 1059-1070.

Barr J., H. Moura, A.E. Boyer, A.R. Woolfitt, S.R. Kalb, A. Pavlopoulos, L.G. McWilliams, J.G.
Schmidt, R.A. Martinez, D.L. Ashley. 2005. Botulinum neurotoxin detection and
differentiation by mass spectrocopy. Emerg Infect Dis. 11(10): 1578-83

Black, J. D. and J. O. Dolly. 1986. Interaction of 125-Iabeled botulinum neurotoxins with nerve
terminals: Ultrastructural Autoradiographical localization and quantification of distinct
membrane acceptors for types A and B motor nerves. J Cell Biol. 103(2), 521-534.

Braconnier, A., V. Brousoule, S. Perelle, P. Fach, C. Ngyun-The, and F. Carlin. 2001. Screening
for Clostridium botulinum type A, B and E in cooked chilled foods containing vegetables
and raw material using polymerase chain reaction and molecular probes. J Food Protect.
64(2), 201-207.

Bullens, R., G. O'Hanlon, E. Wagner, P. Molenaar, K. Furukawa, K. Furukawa, J. J. Plomp, and
H. J. Wilson. 2002. Complex gangliosides at the neuromuscular junction are membrane
receptors for autoantibodies and botulinum neurotoxin. J Neurosci. 22(16), 6876-6884.

Cadieux, B., B. Blanchfield, J. Smith, and J. Austin. 2005. A rapid chemiluminescent slot blot
immunoassay for the detection and quantification of Clostridium botulinum neurotoxin
type E, in cultures. Int J Food Microbiol. 101, 9-16.

Cadotte, A. J. 2004. Polymeric Drug Delivery of Antiepileptic Drugs to Neuronal Networks
Cultured on Multielectrode Arrays. Ph. D. Dissertation, University of Florida,
Biomedical Engineering.









Cadotte, A. and T. B. DeMarse. 2005. Poly-Hema a a drug delivery device for in vitro neural
networks on micro-electrode arrays. J Neural Eng. 2, 114-122.

Cadotte, A., M Ding, T Demarse.2006. Quantification of neural interaction via granger causality.
In: EEG/MEG Studies of Functional Neural Connectivity [Poster] 2006 Neuroscience
Meeting.

Campbell, K., M. Collins, and A. East. 1993. gene probes for identification of the botulinal
neurotoxin gene and specific identification of neurotoxin types B, E, and F. J Clin
Microbiol. 31(9), 2255-2262.

Caya, J., R. Angi, and J. Miller. 2004. Clostridium botulinum and the clinical labratorian: a
detailed review of botulism, including biological warfare ramifications of botulinum
toxin. Arch Pathol Lab Med. 128, 653-662.

Center for Disease Control and Prevention. 1998. Center for Disease Control and Prevention:
botulism in the United States, 1899-1996. Handbook for epidimiologists, clinicians and
laboratory workers. CDC, Atlanta, GA.

Chirwa, S., J. Mack, R. Park, K. Dennis, and A. Aduonum. 2001. An in vivo meodel for
investigating bilateral synaptic plasticity across CA3/CA1 synapses in guinea pig dorsal
hippocampus. J Neurosci Meth. 110, 25-30.

Dahlenborg, M., E. Borch, and P. Radstrome. 2001. Development of a combined selection and
enrichment PCR procedure for Clostridium botulinum types B, E and F and it use to
determine prevalence in fecal samples from slaughtered pigs. Appl and Environ
Microbiol. 67, 4781-4788.

Dan, Y., and M.M. Poo. 2006. Spike timing-deendant plasticity: from synapse to perception.
Physiol Rev. 86: 1033-1048.

Ding, M., Y. Chen, and S. Bressler. 2006. Granger causality: basic theory and application to
neuroscience, 437-459. In B. Schelter, M. Winterhalder, and J. Timmer (eds.), Handbook
of Time Series Anlysis: Recent Theoretical Developments and Applications. Wiley,
Verlag, Berlin.

Dollegast, G., M. Triscott, G. Beard, D. Bottoms, T. Cheng, B. Roh, M. Roman, P. Hall, and J.
Brown. 1993. Sensitive enzyme-linked immunosorbent assay for the detection of
Clostridium botulinum neurotoxins A, B, and E using signal amplification via enzyme
linked coagulation assay. J Clin Microbiol. 31, 2402-2409.

Dong, M., Y. Felix, W. Tepp, C. Dean, E. Johnson, R. Janz, and E. Chapman. 2006, 12 March.
SV2 is the protein receptor for botulism neurotoxin A. Science [Online] Science Express.
Available, www.scienceexpress.org.









Ekong, T., I. Feavers, and D. Sesardic. 1997. Recombinant SNAP-25 is an effective substrate for
Clostridium botulinum type A toxin endpeptidase activity in vitro. Microbiol. 143, 3337-
3347.

Emanuel, P., J. Dang, J. Gebhardt, J. Aldrisch, E. Garber, H. Kulga, P. Stopa, J. Valdes, and A.
Dion-Schultz. 2000. Recombinant antibodies: a new reagent for biological agent
detection. Biosens Bioelectron. 14, 751-759.

Eytan, D., N. Brenner, and S. Marom. 2003. Selective Adaptation in Networks of Cortical
Neurons. JNeurosci. 23(28), 9349-9356.

Fach, P., M. Gibert, R. Griffais, J. guillou, and M. Popoff =. 1995. PCR and gene pobe
identification of botulinum neurotoxin A-, B-, E-, F- and G- producing Clostridium spp.
and the evaluation in food samples. Appl Envioron Microbiol. 61, 389-392.

Fach, P., D. Hauser, J. Guillou, and M. Popoff. 1993. Polymerase chanin reaction for the
rapididentification of Clostridium Botulinum type A strains in food samples. J Appl
Bacteriol. 75, 234-239.

Ferreira, J., B. Baumstark, M. Hamdy, and S. McCay. 1993. Polymerase chain reaction for
detection of type A Clostridium botulinum in foods. J Food Protect. 56(1), 18-20.

Ferreira, J., S. Eliasberg, P. Edmonds, and M. Harrison. 2004. Comparision of the mouse
bioassay and enzyme-linked immunosorbent assay procedures for the detection of type
A botulinal toxin in food. J Food Protect. 67((1)), 203-206.

Ferreira, J., S. Maslanka, E. Johnson, and M. Goodenough. 2003. Detection of botulinal
neurotoxins types A, B, E and F by amlified enzyme-linked immunosorbent assay:
collaborative study. JAOAC Int. 86(2), 314-331.

Gholmieh G, Soussou W, Courellis S, Marmarlis V, Berger T, and Baudry M. 2001. A
biosensor for detecting changes in cognative processing based on nonlinear system
analysis. Biosens Bioelectron. 16, 491-501.

Gholmieh G, Spiros C, Fakheri S, Cheung E, Marmarelis V, Baudry M, and Berger T. 2003.
Detection and classification of neurotoxins using novel short-term plasticity
quantification method. Biosens Bioelectron. 18, 1467-1478.

Gramowski, A., K. Jugelt, D. G. Weiss, and G. W. Gross. 2004. Substance identification by
quantitative characterization of oscillatory activity in murine spinal cord networks on
microelectrode arrays. Eur J of Neurosci. 19, 2815-2825.

Gross, G. W., B. K. Rhoades, H. M. E. Azzaay, and M.-C. Wu. 1995. The use of neuronal
networks on multielectrode arrays as biosensors. Biosens Bioelectron. 10, 553-567.









Gross, G. W., A. Harsch, B. K. Rhoades, and W. Gopel. 1997. Oder, drug and toxin analysis
with neuronal networks in vitro: extracellular recording of network responses. Biosens
Bioelctron. 12(5), 373-393.

Habets, A., A. Van Dingen, F. Van Huizen, and Corner MA. 1987. Spontaneous neurnal firing
patterns in fetal rat cortical networks during development in vitro: a quantitative analysis.
Exp Brain Res. 69, 43-52.

Hallis, B., B. James, and C. Shone. 1996. Development of novel assays for botulinum type A and
B neurotoxins based on their endopeptidase activities. J Clin Microbiol. 34, 1934-1938.

Hoch, D. H., M. Romero-Mira, B. E. Ehrlich, A. Finkelstein, B. R. DasGupta, and L. Simpson.
1985. Channels formed by botulinum, tetnus, and diptheria toxins in planar lipid bilayers:
relevence to translocation of proteins across membranes. PNAS. 82, 1692-1696.

Jimbo Y, T. Tateno, H. P.C. Robinson. 1999. Simultaneous induction of pathway-specific
potentiation and depression in networks of cortical neurons. Biophysical J 76: 670-678.

Jimbo, Y., A. Kawana, P. Parodi, and V. Torre. 2000. The dynamics of a neuronal culture of
dissociated cortical neurons of neonatal rats. Biol Cybern. 83, 1-20.

Kamioka, H., E. Maeda, Y. Jimbo, and H. Robinson, Kawana. 1996. Spontaneous periodic
synchronized bursting during formation of mature patterns of connections in cortical
cultures. Neurosci Lett. 206, 109-112.

Keefer, E., S. Norton, N. A. J. Boyle, V. Talesa, and G. W. Gross. 2001. Acute Toxicity
Screening of Novel AChe Inhibitors Using Neuronal Networks on Microelectrode
Arrays. Neurotoxicology. 22, 3-12.

Keller, J. E., F. Cai, and E. A. Neale. 2004. Uptake of Botulinum Neurotoxin into Cultured
Neurons. Biochemistry. 43, 526-532.

Keller, J., J. Nowakowski, M. Filbert, and M. Alder. 1999. Rapid microplate assay for mintoring
botulinum neurotoxin B catalytic activity. J Appll Toxicol 19,S13-S17.

Kimura, B., S. Kawaski, and H. F. Nakano, T. 2001. Rapid, quantitiative PCR monitoring
growth of Clostridium botulinum type E in modified-atmosphere-packaged fish. Appl
Environ Microbiol. 67(1), 206-216.

Lindstrom, M., R. Keto, M. Annukka, M. Nevas, S. Heilm, and H. Korkeala. 2001. Multiplex
PCR assay for the detection and identification of Clostridium botulinum types A, B, E,
and F in food and fecal material. Appl Envioron Microbiol. 67(12), 5694-5699.

Liu, W., V. Montana, E. Chapman, U. Mohideen, and V. Parpura. 2003. Botulinum toxin type B
micromechanosensor. PNAS. 100, 13621-13625.









Lovenklev, M., E. Hoist, E. Borch, and P. Radstom. 2004. Relative neurotoxin gene expression
in Clostridium botulinum type B, determined using quantitative reverse transcription
PCR. Appl Environ Microbiol. 70(5), 2919-2927.

Maeda, E., P. C. Robinson, and A. Kawana. 1995. The Mechanisms of generation and
propagation of synchronized bursting in developing networks of cortical neurons. The
Journal of Neuroscience. 15, 6834-6845.

Marom, S. and G. Shahaf. 2002. Development, learning and memory in large random networks
of cortical neurons: lessons beyond anatomy. Q Rev Biophysics. 35, 63-87.

McGrath, S., J. Dooley, R. Haylock, and. 2000. Quantification of Clostridium botulinum toxin
gene expression by cometetive reverse transcription-PCR. Appl Envioron Microbiol.
66(4), 1423-1428.

Meister, M., R. Wong, Baylor DA, and C. Shatz. 1991. Synchronous burst of action potentials in
gnglion cells of the developing mammalian retina. Science 252, 939-943.

Meister, M., L. Lagnado, and D. A. Baylor. 1995. Concerted Signaling by Retinal Ganglion
Cells. Science 270, 1207-1210.

Meister, M., J. Pine, and D. A. Baylor. 1994. Multi-neuronal signals from the retina: acquisition
and analysis. JNeurosci Meth. 51, 95-106.

Montecucco, C., O. Rossetto, and G. Schiavo. 2004. Presynaptic receptor arrays for clostridial
neurotoxins. Trends Microbiol. 12(10), 442-446.

Morefield, S. I., E. W. Keefer, K. D. Chapman, and G. W. Gross. 2000. Drug evaluations using
neuronal networks cultured on microelectrode arrays. Biosens Bioelectron. 15, 383-396.

Morf, M., A. Vieira, D.T.L. Lee, and T. Kailath. 1978. Recursive multichannel maximum
entropy spectral estimation. IEEE Trans on Geosci Electron. G-16(2): 85- 94.

Muramoto, K., M. Ichikawa, K. Kobayashi, M. Kawahara, and Y. Kuroda. 1993. Formation and
maturation of synapses in primary cultures of rat cereberal cortical cells; an electron
microscopic study. Neurosience Research. 16(2), 95-103.

Nakanishi, K. and F. Kukita. 1998. Functional synapses in synchronized bursting of neocortical
neurons in culture. Brain Res. 795, 137-146.

Nelsen, D. and M. Cox. 2005. Lehninger Principles of Biochemistry. W H Freeman, New York.

Nishiki, T., Y. Kamata, Y. Nemoto, A. Omori, T. Ito, M. Takahashi, and S. Kozaki. 1994.
Identification of a protein receptor for Clostridium botulinum Type B Neurotoxin in Rat
Brain Synapsomes. JBiol Chem. 269(14), 10498-10503.









Nishiki, T., Y. Tokuyama, Nemboto, A. Yoshida, K. Sato, M. Sekiguchi, M. Takahashi, and S.
Kozaki. 1996. The high-affinity binding of Clostridium botulinum B neurotoxin to
synaptogamin II associated with gangliosides GTlb/GDla. FEBS Lett. 378, 253-257.

O'Brien, T., L. Johnson, Aldrich, S. Allen, L. Liang, A. Plummer, S. Krak, and A. Boiarski.
2000. The development of immunoassays to four biological threat agents in a bidiffractive
grating biosensor. Biosens Bioelectron. 14, 815-828.

Pancrazio, J. J., S. A. Gray, Y. S. Shubin, N. Kulagina, D. S. Cuttino, K. M. E. Shaffer, K, A.
Curran, B. Zim, G. W. Gross, and T. J. O'Shaughnessy. 2003. A portable microelectrode
array recording system incorporating cultured neuronal networks for neurotoxin
detection. Biosens Bioelectron. 18, 1339-1347.

Peruski, A., L. Johnson, and L. Peruski. 2002. Rapid and sensitive detection of biological
warfare agents using time-resolved Fluorescence assays. J Immunol Meth 263, 35-41.

Poo, M. and G. Bi. 1998. Synptic modifications by correlated activity, Hebb's postulate revisited.
Ann Rev Neurosci. 24, 139-166.

Potter, S. and T. DeMarse. 2001. A new approach to neuronal cell culture for long-term studies. J
Neurosci Meth. 110, 17-24.

Reppel, M., P. Igelmund, U. Egert, F. Juchelka, J. Hescheler, and I. Drobinskaya. 2007. Effect of
Cardioactive Drugs on Action Potential Generation and Propagation in Embryonic Stem
Cell Derived Cardiomyocytes. Cell Physiol Biochem. 19, 213-224.

Rowe-Taitt, C., J. Golden, M. Feildstein, J. Crass, K. Hoffman, and F. Liger. 2000. Array
biosensor for detection of biohazards. Biosens Bioelectron. 14, 785-794.

Rummel A, T. Karnath, H. Bigalke, and T. Binz. 2004a. Synaptogamins I and II act as nerve cell
receptors for botulinum neurotoxin G. J Biol Chem. 279, 30865-30870.

Rummel, A., S. Mahrold, H. Bigalke, and T. Binz. 2004b. The Hcc domain of botulinum
neurotoxins A and B exhibits a singular ganglioside binding site displaying serotype
specific carbohydrate interaction. Mol Microbiol. 51(3), 631-643.

Scarlatos, A., B. W. Welt, B. Y. Cooper, D. Archer, T. DeMarse, and K. V. Chau. 2005.
Methods for Detecting Botulinum Toxin with Applicability to Screening Foods Against
Biological Terrorist Attacks. J Food Sci. 70(8), R121-R130.

Schmidt, J. and R. Staford. 2003. Fluorogenic substrates for the protease activities of botulinum
neurotoxins, serotypes A, B, and E. Appl Envioron Microbiol 50, 63-67.

Schmidt, J., R. Staford, and C. Millard. 2001. High-throughput assays for botulinum neurotoxin
proteolytic activity: serptypes A, B, D, and F. Anal Biochem. 296, 130-137.









Sciacchitano, C. and I. Hirshfield. 1996. Molecular detecton of Clostridium botulinum type E
neurotoxin gene in smoked fish by polymerase chain reaction and cappillary
electrophoresis. J AOAC Int. 79(4), 861-865.

Segev, R., M. Beveniste, E. Hulata, N. Cohen, and A. Palevski. 2002. Long term behavior of
Lithographically Prepared In Vitro Neuronal Networks. Phys Rev Lett. 88(11), 118102.

Selinger, J., J. J. Pancrazio, and G. W. Gross. 2004. Measuring synchronization in neuraonal
networks for biosensor applications. Biosens Bioelectron.19, 675-683.

Shao, J., T. Tso, and R. Butera. 2006. Bursting with Slow Kinetics: A Role for Small World.
Neural Comput. 18, 2029-2035.

Shone, C., P. Smith, N. Appelton, P. Hambelton, N. Modi, S. Gatley, and J. Melling. 1985.
Monoclonal antibody-based immunoassay for type A Clostridium botulinum toxin is
comparable to the mouse bioassay. Appl and Environ Microbiol. 50(63-67).

Simpson, L. L. 1983. The interaction between aminoquinolines and presynaptically acting
neurotoxins. J Pharmacol and Exp Ther. 222, 43-48.

Simpson, L. L. 2004. Identification of the major stepsin botulinum toxin action. Annu Rev
Pharmacol. 44, 167-193.

Simpson, L. 1996. Botulinum toxin: a deadly poison sheds its negative image. Ann Int Med. 125,
616-617.

Solomon, H. and T. Lily. 2001. Clostridium botulinum. Available:
http//www.cfsan.fda.gov/-ebam-17.html.

Stenger, D. A., G. W. Gross, E. W. Keefer, K. M. Shaffer, J. D. Andreadis, W. Ma, and J. J.
Pancrazio. 2001. Detection of physiologically active compounds using cell-based
biosensors. Trends Biotechnol. 19(8), 304-309.

Streit, J., A. Tescherter, and P. Darbon. 2005. Rythem Generation in Spinal Cultures is it the
Neuron or the Network, 26. In M. Taketani and M. Baudry (eds.), Advances in Network
Electrophysiology Using Multi-Electrode Arrays. New York.

Tabak, J. and P. Latham. 2003. Analysis of spontaneous bursting activity in random neural
networks. Neuroreport. 14(11), 1445-1449.

Trevan, J. 1927. The error of determining toxicicity. Proc R Soc. 101B, 483-514.

van Barr LM, A.G. Hulst, A.L. de Jong, E.R.J. Wils. 2002. Characterization of botulinum toxins
A and B, by matrix-assisted laser desorption ionisation and electrospray mass
spectrometry. J Chromatogr A 1035:95-115.









van Barr LM, A.G. Hulst, A.L. de Jong, E.R.J. Wils. 2004. Characterization ofbotulinum toxins
C, D, E, and F by matrix-assisted laser desorption ionisation and electrospray mass
spectrometry. J Chromatogr A 970:97-114.

van Pelt, J, P.S. Woters, M.A. Corer, W.L. Rutten, G.J.A. Ramakers. 2004. Long-term
characterization of firing dynamics of spontaneous bursts in culltured neural networks.
IEEE Transiomed Eng 51 (11):2051-62.

van Pelt, J., I. Vajda, P. Wolters, M. Corner, and G. Ramakers. 2005. Dynamics and placticity in
developing neuronal networks in vitro. Prog Brain Res. 147, 173-188.

Verderio, C., S. Coco, O. Rossetto, C. Montecucco, and M. Matteoli. 1999. Internalization and
proteolytic action in CNS neurons and astrocytes. J Neurochem. 73(1), 372-379.

Verderio, C., O. Rossotto, C. Grumelli, C. Frassoni, C. Montecucco, and M. Matteoli. 2006.
Entering neurons: botulinum toxins and synaptic vescle recycling. EMBO Reports.
7(10):995-999.

Wagenaar, D., Z. Nadasdy, and S. Potter. 2006. Persistent dynamic attractors in activity patterns
of cutured neuronal networks. Phys Rev E. 73:0151907-1-8.

Wagenaar, D., J. Pine, and S. Potter. 2006. An extremely rich repertoire of bursting patterns
during the development of cortical cultures. BMC Neurosci. 7.

Wagenaar, D. and S. Potter. 2002. Real-time multi-channel stimulus artifact suppression by local
curve fitting. J Neurosci Meth. 120:113-120.

Witcome, M., K. Newton, K. Jameson, B. Hallis, P. Dunnigan, E. Mackay, S. Clarke, R. Taylor,
J. Gaze, K. Foster, and C. Shone. 1999. Development of an In Vitro Bioassay for
Clostridium Botulinum Type B Neurotoxin in Foods That is More Sensitive than the
Mouse Bioassay. Appl and Environ Microbiol. 65(9, Sept):3787-3792.

Yowler, B. C., R. D. Kensinger, and C. L. Schengrund. 2002. Botulinum neurotoxin A activity is
dependant upon the presence of specific gangliosides in neuroblastoma cells expressing
Synaptogamin I. J Biol Chem. 277(36):32815-32819.









BIOGRAPHICAL SKETCH

Amber Scarlatos graduated cum laude with a Bachelor of Science degree in chemical

engineering from the University of Florida in May 2003. In August 2003 she received an alumni

fellowship to pursue her Doctorate of Philosophy in agricultural and biological engineering at the

University of Florida. Her research interest involves the development of living tissue cultures as

broadly applicable biosensors for biological agents. In particular, her current research involves

the use of cortical tissue grown on a microelectrode array to detect Botulism toxin.





PAGE 1

RANDOM CORTICAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELL-BASED BIOS ENSORS FOR NEUROTOXINS AS DEMONSTRATED BY BOTULINUM TOXIN TYPE A By AMBER CHRISTINE SCARLATOS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007 1

PAGE 2

2007 Amber Christine Scarlatos 2

PAGE 3

This work is dedicated to my husband, Chris S carlatos, my son, Nathan Scarlatos, and my parents Allen and Arleen Austin 3

PAGE 4

ACKNOWLEDGMENTS I thank Dr Bruce Welt for his enthusiasm, patience and support. I thank Dr. Thomas DeMarse who kindly guided me during my many adventures in the Neural Computation and Robotics Laboratory. I thank Dr. Brian Cooper fo r introducing me to the expansive field of electrophysiology and for keeping me (mostly) sane during the early years of my PhD work. I thank Dr. Archer and Dr. Chau for serving on my committee and for helping to guide me through my Ph.D. work. I thank Alex Cadotte who shared my enthusiasm and frustrations with cell culture and guided me in application of Granger's Causality analysis to data generated on multielectrode arrays. Additionally, I thank my friends and family for their ongoing support. I would especially like to thank my husband for patiently waiting fo ur more years, and my son for giving me a unique perspective. 4

PAGE 5

TABLE OF CONTENTS ACKNOWLEDGMENTS ............................................................................................................... 4TABLE OF CONTENTS ............................................................................................................. ....5LIST OF TABLES ...........................................................................................................................8LIST OF FIGURES .......................................................................................................................10LIST OF ABBREVIATIONS ........................................................................................................1 1ABSTRACT ...................................................................................................................... .............13 CHAPTER 1 INTRODUCTION AND BACKGROUND ...........................................................................15Introduction .................................................................................................................. ...........15Objective ..................................................................................................................... ............16Background .................................................................................................................... .........16Electrophysiology of Neuronal Tissue ............................................................................16Botulism Toxin ................................................................................................................ 20BoNT Action on the Central Nervous System .......................................................................252 REVIEW OF BOTULINUM TO XIN DETECTI ON METHODS ........................................27Introduction .................................................................................................................. ...........27Mouse Bioassay ................................................................................................................ ......27Enzyme Linked Immunosorbent Assay ..................................................................................29Polymerase Chain Reaction Assay .........................................................................................31Protease Activity Assays ........................................................................................................36Mass Spectrometry .................................................................................................................38Discussion .................................................................................................................... ...........403 NEURAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELLBASED BIOSENSORS ..........................................................................................................43Introduction .................................................................................................................. ...........43Hardware and Software ..........................................................................................................45Micro-electrode arrays (MEA) ........................................................................................45Data acquisition ...............................................................................................................46Physiological conditions ..................................................................................................46Dissociated Neural Cultures ................................................................................................... 47Tissue Culture Methods ...................................................................................................47Development and activity ................................................................................................48Network Bursting ....................................................................................................................50MEAs as Biosensors ...............................................................................................................51 5

PAGE 6

6 4 DEFINING THE BONT A FINGERPRINT ..........................................................................52Introduction .................................................................................................................. ...........52Materials and Methods ...........................................................................................................52Procedure ..................................................................................................................... ....52Drug application ..............................................................................................................53Analysis and statistics. .....................................................................................................53Results .....................................................................................................................................54Spontaneous activity ........................................................................................................54Intra-burst neural dynamics .............................................................................................56Evoked response following intoxication with BoNT A ..................................................58Discussion .................................................................................................................... ...........625 REFINING THE BONT A FINGERPRINT AND DEVELOPMENT OF A BONT DETECTION ALGORITHM .................................................................................................64Introduction .................................................................................................................. ...........64Background .................................................................................................................... .........64Plasticity .................................................................................................................... ......64Granger's Causality ..........................................................................................................65Methods ..................................................................................................................................67Experimental protocol .....................................................................................................67Plasticity protocol ............................................................................................................67Results and Discussion ........................................................................................................ ...68Spontaneous activity ........................................................................................................68Elicited activity ................................................................................................................68Plasticity .................................................................................................................... ......70PGC Analysis .................................................................................................................. 72BoNT Detection Algorithm .............................................................................................726 CONCLUSION AND FUTURE WORK ...............................................................................77Future Work ............................................................................................................................77Reduction of detection time .............................................................................................77Serotype differentiation ...................................................................................................78Use of neural networks with crude samples ....................................................................78Alternative and immortal cell lines .................................................................................79Conclusion .................................................................................................................... ..........79APPENDIX A SPONTANEOUS ACTIVITY IN BONT A AND CARRIER DOS ED CULTURES ...........81B ELLICITED ACTIVITY IN BONT A AND CARRIER DOSED CULTURES ...................87C BURST DETECTION ALGORITHM ...................................................................................91D BONT A DETECTION ALGORITHM .................................................................................95

PAGE 7

7 REFERENCES .................................................................................................................... ........105BIOGRAPHICAL SKETCH .......................................................................................................113

PAGE 8

LIST OF TABLES Table page 2-1 Immunoassays for the detection of BoNTs ........................................................................30 2-2 PCR primers and DNA pr obes for detection of C. botulinum and toxigenic C. butyricum and C. barati .....................................................................................................32 5-1 Identification of BoNT in unknown cultures .....................................................................76 A-1 Average spike height for neural cultures dosed with BoNT carrier or 200pg BoNT A ...81 A-2 Average spike width for neural cultures dos ed with BoNT carrier or 200 pg BoNT A ....81 A-3 Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg BoNT A ..............................................................................................................................81 A-5 Average inter-burst interv al for neural cultures dose d with BoNT carrier or 200 pg BoNT A ..............................................................................................................................82 A-6 Average number of spikes per burst for neural cultures dosed with BoNT carrier or 200 pg BoNT A ..................................................................................................................82 A-7 Average spike height for neural cultures dosed with BoNT carrier or 600pg BoNT A ...82 A-8 Average spike width for neural cultures dosed with BoNT carrier or 600pg BoNT A .....83 A-9 Average inter spike interval for neural cultures dosed with BoNT carrier or 200pg BoNT A ..............................................................................................................................83 A-10 Average burst duration for neural cultur es dosed with BoNT carrier or 600pg BoNT A ............................................................................................................................. ............83 A-11 Average inter-burst interv al for neural cultures dose d with BoNT carrier or 600 pg BoNT A ..............................................................................................................................84 A-12 Average number of spikes per burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A ..................................................................................................................84 A-13 Average number of channels per burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A ..................................................................................................................84 A-14 Average burst duration for neural cult ures dosed with BoNT carrier or 1000 pg BoNT A ..............................................................................................................................85 A-15 Average inter-burst interv al for neural cultures dose d with BoNT carrier or 1000 pg BoNT A ..............................................................................................................................85 8

PAGE 9

9 A-16 Average number of spikes per burst for neural cultures dosed with BoNT carrier or 1000 pg BoNT A ................................................................................................................85 A-17 Average number of channels per burst for neural cultures dosed with BoNT carrier or 1000 pg BoNT A ................................................................................................................86 B-1 Average spikes per stimulation for neur al cultures dosed with BoNT carrier or 200pg BoNT A ...................................................................................................................87 B-2 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or 200pg BoNT A ...................................................................................................................87 B-3 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 200pg BoNT A .........................87 B-4 Percent change in the number of spikes 150 to 300 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 200pg BoNT A .........................88 B-5 Average spikes per stimulation for neural cultures dosed with BoNT carrier or 600 pg BoNT A .........................................................................................................................88 B-6 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A ..................................................................................................................88 B-7 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 600pg BoNT A .........................89 B-8 Percent change in the number of spikes 150 to 300 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 600pg BoNT A .........................89 B-9 Average spikes per stimulation for neural cultures dosed with BoNT carrier or 1000 pg BoNT A .........................................................................................................................89 B-10 Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or 1000 pg BoNT A ................................................................................................................90 B-11 Percent change in the number of spikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A ......................90 B-12 Percent change in the number of spikes 100 to 300 ms post stimulation from baseline activity for neural cultures dosed with BoNT carrier or 1000 pg BoNT A ......................90

PAGE 10

LIST OF FIGURES Figure page 1-1 Botulism toxin cleaves SNARE proteins necessary for vesicle fusion docking ................26 2-1 Botulinum B micromechanosensor ....................................................................................38 3-1 A micro-electrode array (MEA ) from Multichannel Systems ...........................................45 3-2 Data acquisition hardware ..................................................................................................47 3-3 Screen shot of t ypical raw activity during a network-wide burst ......................................49 4-1 Spontaneous burst activity in cortical networks after treat ment with BoNT A or toxin carrier ....................................................................................................................... ..........55 4-2 Return plots generated by sorting spikes in to 100 ms bins and plotting bins at time t against bins at time t 50 ms .............................................................................................57 4-3 Stimulation-response for cort ical culture grown on MEA 9053 ........................................59 4-4 Mean change in the number of evoked spik es 100 to 300 ms post stimulation in the BoNT and control cultures ................................................................................................60 4-5 Average stimulation response for Control (Panel A) and BoNT dosed (Panel B) tissue cultures ............................................................................................................... ......61 5-1 Percent change in spikes with respect to baseline activ ity after treatment with 600 pg BoNT A (A) and 1000 pg BoNT A (B) .............................................................................69 5-2 Network plasticity induced by tetanus ...............................................................................71 5-3 Granger's Causality difference plot afte r dosage with BoNT A carrier (A), and 1000 pg BoNT A (B) ................................................................................................................. .73 5-4 Average stimulation response (blue) a nd smoothed stimulation response (red) for control dish 7079 (A D) and dish 9299 after dosage with 200 pg BoNT A (E H). .......75 10

PAGE 11

LIST OF ABBREVIATIONS APRS: array of presynaptic receptors AR: autoregressive BoNT: botulinum toxin BHI: Brain Heart Infusion CNS: central nervous system DIV days in vitro DMEM Dulbucco's Modified Eagles Medium ELISA: enzyme-linked immunosorbent assay ES-MS: electron spray mass spectrometry Fab: antigen binding antibody fragment GalNAc-T-/GalNAc transferase negative GABA: -Aminobutyric acid GC: Granger's Causality HA: heamagluttins LC-ES-MS: liquid chromatography a nd electrospray mass spectrometry LD50: lethal dose for 50 % of the population LTD: long term depression LTP: long term potentiation MALDI-MS: matrix assisted laser de sorption ionization mass spectrometry MEA: multielectrode array MLD: minimum leathal dose mAbs monoclonal antibodies NTNH: nontoxic non-heamagluttins pAbs: polyclonal antibodies 11

PAGE 12

12 PCR: polymerase chain reaction PGC: Pairwise Granger's Causality RT-PCR: reverse transcriptase polymerase chain reaction SNARE soluble NSF attachment receptor SNAP-25 synaptosome-associated protein of 25,000 daltons SV2 synaptic vesicle protein 2 SYT: synaptogamin TYP-C: tryptone-pept one-yeast extract VAMP: synaptobrevin

PAGE 13

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RANDOM CORTICAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELL-BASED BIOS ENSORS FOR NEUROTOXINS AS DEMONSTRATED BY BOTULINUM TOXIN TYPE A By Amber C Scarlatos December 2007 Chair: Bruce Welt Cochair: Brian Cooper Major: Agricultural and Biological Engineering This work presents a potential biosensor fo r neurotoxin detection that employs living neural tissue cultures grown in-v itro on microelectrode arrays (MEAs). An MEA is a tissue culture dish with a grid of el ectrodes embedded in the surface of the dish, enabling extracellular recording of action potentials produced by neural cultures grown over the array. Botulism toxin (BoNT) is an extremely potent neurot oxin produced by toxigenic strains of Clostridium botulinum Botulism toxin continues to pose a majo r threat since it may be employed in a deliberate attack on the US f ood supply. Furthermore, BoNT may be liberated in any insufficiently processed food containing a reduce d oxygen atmosphere. Hence, rapid and reliable detection of BoNTs in foods is necessary to re duce the risk posed through food contamination. In this study, a bolus of 200 to 1000 pg of pharmaceutic al grade BoNT A was applied to the media bath of mature cortical networks cultured on mi croelectrode arrays. It was hypothesized that a detectable change in burst activ ity would be observed of treatmen t of the cortical cultures with BoNT A. The observed change in burst activity w ould enable reliable dete ction of the toxin as the basis for cell-based BoNT biosensor. 13

PAGE 14

14 After treatment of the tissue cultures with BoNT A, both spontaneous and evoked activity were then monitored over a one week period to quantify changes in the neural population produced by BoNT A. Introduction of BoNT A resulted in an increase in the duration and number of spikes in the spontaneous and evoked bursts relative to contro l cultures. Application of BoNT A also induced unique os cillatory behavior within each burst that is reminiscent of early developmental activity patterns rather th an the mature cultures used here. Additionally, changes in connectivity during e voked bursts quantifie d using Granger Causality (GC), a tool borrowed from economics. Neural cultures dosed with botulism toxin were found to have an overall decrease in GC values indicating a decrease in effective connectivity. Based upon the observed changes, an algorithm for BoNT detection was developed which enabled detection of BoNT A in 13 of 15 neural cultures by 48h and a ll 15 neural cultures by 72h. No control dishes were positive for BoNT A up to 72h after dosage with BoNT A carrier solution. Thus the activity of these cortical networks measured with MEAs could provide a valuable substrate for BoNT detection.

PAGE 15

CHAPTER 1 INTRODUCTION AND BACKGROUND Introduction Efficient detection of food-borne toxins is necessary to secure the food supply from both intentional and uninten tional contamination. Botulism toxi n, associated with food-borne botulism, is a potent neurotoxin that may be liberated in a ny unprocessed or insufficiently processed food containing an oxyge n deficient atmosphere. Furtherm ore, BoNTs may be used as a bioweapon in a deliberate attack on the US food supply. Currently, seven antigenic strains of botulism toxin (A-G) have been identified, which are produced by the anaerobic bacterium Clostridium botulinum. Botulinal neurotoxin is extremely toxi c, with an estimated oral lethal dose of as little as 70 g of botulism toxin type A (BoNT A) for a 70 kg human (Arnon et al., 2001). Containment of a botulism outbreak is dependa nt upon rapid identification of the toxin in suspect foods. While there are many assays av ailable for detecting specific toxins, toxin fragments, and DNA of causative organisms, the FDA continues to primarily rely on the mouse bioassay for the detection of botulinum toxicity in foods (Scarlatos et al., 2005). The mouse bioassay detects the presence of t oxins by observing death rates of animals injected with suspect food samples. Identification of toxin serotypes, determined through neutralization assays using equine antiserum, require additional time and an imal use. The mouse bioassay is extremely sensitive, detecting as little as a few picogram s of toxin, and capable of identifying the presence of virtually any type of biologi cally active toxin (Arnon et al., 2001). It is also time consuming, prohibitively expensive due to the large number of animals re quired, and available at only a limited number of facilities nati onwide. Alternative assays tend to focus on specific toxins and 15

PAGE 16

16 may provide false positive indications when fragments of inactive toxi n proteins or DNA are present (Scarlatos et al., 2005). Objective It was hypothesized that addi tion of BoNT A to the cortical networks grown on MEAs would produce a discernable change in the burs t dynamics. This change in network activity would be unique to previously reported ch emically-induced changes in network activity, providing a unique toxin fi ngerprint for BoNT A. The objective of this work is to characterize activity changes resulting from dosage of random cortical neuron networks grown in vivo with botulinum toxin, an d to develop reliable methods for discerning BoNT toxin from native activity. This work provides the groundwork for development of a cell-based BoNT biosensor. Background Electrophysiology of Neuronal Tissue Membrane biology. A fundamental characteristic of all cells is an outer membrane, which defines the intracellular and ex tracellular environment. The cell membrane has many remarkable properties which enable cell func tion. Cell membranes are flexible, self sealing and have limited permeability. The primary component of cell membranes is a phospholipid, which is comprised of a hydrophilic head along with two long hydrophobic fatty-acid ta ils. Within the cell membrane the phospholipids form a lipid bila yer, where the hydrophilic heads of the phospholipids are located on the outermost and i nnermost surface of the cell membrane, while the hydrophobic tails are located within the bilayer core. This structure enables the cell membrane to be passively impermeable to most polar and charged solutes. However, permeability of membranes is not simply a passive property, since an array of important protein

PAGE 17

17 structures reside within the cell membrane. These proteins play an important role in membrane function. Because cell membranes are impermeable to i ons, intracellular solute concentrations are different from extracellular concen trations. In particular, ions ca nnot passively diffuse across the cell membrane, and thus exist at different concen trations within intracellular and extracellular spaces. Since ions have an associated electr ical charge, a potential difference across cell membranes results from the differential concentra tion of ions. This potential difference is an important aspect of cell membranes, and is ac tively maintained through pumps which exchange ions from one side of the membrane to the other. When a net ion flux across a cell's membrane is at zero, due to balanced electri cal and chemical forces, the membrane potential is constant and is known as the resting potential. The equilibrium pot ential for a single ion can be described by the Nernst Equation. i oion ion zF RT E ln (1-1) However, the resting potential is dependant on several ions. The most common ions found within animal fluids are sodium (Na+), potassium (K+), Calcium (Ca2+), Magnesium (Mg2+), Chlorine (Cl-), and Bicarbonate (HCO3 -). To obtain the resting potential when multiple ion species are relevant the Goldman-HodgkinKatz equation can be used, where P is the permeability of an ion. N j j o j N j j i jPion Pion F RT E1 1ln (1-2) Sodium and Potassium ions have the most st riking contrast between extracellular and intracellular concentrations. In mammalian tissue, [K+]i 140 mM while [K+]o 4 mM.

PAGE 18

18 Conversely, [Na+]i 12 mM while [Na+]o 145 mM (Nelson and Cox, 2005). The large sodium and potassium gradients are primarily du e to active transport of sodium and potassium ions by an integral membrane transport protein known as Na+K+ ATPase. This protein uses energy from ATP hydrolysis to pump three Na+ out of the cell, and two K+ ions into the cell. Neuronal tissue has a resting potential around -70 mV which is close to the K+ resting potential. The action potential. Cells can be classified as exci table or inexcitable based upon how they react to a depolarizing cu rrent. A depolarizing current appl ied to the membrane of an unexcitable cell results in a propor tional change in the membrane potential. In this instance the membrane mimics an RC-circuit. Excitable cells have a similar RC-response to sufficiently small depolarizing currents. However, when depolar izing currents exceed a threshold, a large disproportionate change in membrane potential occurs, known as the action potential. Animals have a variety of excitable cells incl uding neuronal, and muscular cells. The action potential is an electrochemical dist urbance generated within excitable cells and is the basic unit of neural communication. The ac tion potential is charac terized by three phases: the excitation phase, the recovery phase and a refractory period. These phases are a result of voltage gated sodium and potassium channels in the membrane. A depolarizing current that exceeds the excitability threshold causes voltage-g ated sodium channels to open. This causes a rapid influx of sodium ions, further depolariz ing the membrane potential. Time dependant inactivation of voltage gated sodium channels ca uses sodium channels to temporarily enter a state from which they cannot reopen. At the sa me time, voltage gated activation of potassium channels occurs, and potassium ions flow outward to repolarize the cell membrane back towards the resting potential. Potassium channels are voltage dependant and can be time dependant as well. As the voltage returns to the resting potential, it often hyperpolarizes the membrane due, in

PAGE 19

19 part, to delayed closing of pota ssium channels. After the voltage gated sodium channels have closed there is a period in whic h they cannot be activated by exci table input no matter how large. This is known as the refractory period, wh ich is typically around one millisecond. The Na+K+ ATPase actively pumps sodium out and potassium back into the ce ll to return intracellular ion concentrations to physiological le vels. Action potentials are an all or nothing event. They can be generated through application of an external stimulus, or may occur spontaneously due to fluctuations from the resting potential. Depolarization of the membrane by an action potential can activate a variety of cellular processes in cluding activation of additional voltage gated channels, such as voltage gated calcium channe ls. Voltage gated calcium channels play an essential role in cell to cell communication at the chemical synapse as described below. Neuronal Networks. Typically, neurons are comprised of three basic components: the soma, the dendritic tree and the axon. The soma is a cell body where the cell nucleus and the majority of cellular organelles reside. The dendritic tree is a large branch ed portion of a neuron, and is variable in shape and morphology. The de ndrite is typically clos e to the cell body and primarily receives input from other neurons t hough chemical synapses. Neuronal axons typical consist of a long myelinated process that ends at a synapse. The axon primarily communicates with target neurons through neurotransmission at chemical synapses. For example, at the neuromuscular junction, the primary target of BoNTs, release of acetylcholine from chemical synapses results in muscle contraction. Within the body, neurons do not behave as an independent unit, but ra ther act as a subunit of a larger network of thousands of neur ons. Communication between neurons occurs at synapses. When an action potential reaches the axon terminal of a chem ical synapse, voltagegated calcium channels are activated. This initiates a rapid influx of calcium, which in turn,

PAGE 20

20 enables membrane fusion of neurotransmitter containing vesicles. Neurotransmitter is released into the synaptic cleft, where it diffuses across to postsynaptic receptors binds to appropriate postsynaptic proteins to depolarize (or sometimes hyperpolarize) the target neuron. Excitatory input at a single synapse may be insufficient to ge nerate an action potentia l that will propagate from the dendritic tree to the axon terminal. T ypically, excitatory inputs from multiple synapses must be integrated in order to generate an excitatory input which exceeds the threshold necessary to generate an action potential. A sufficiently large excitatory input in neur onal networks comprised purely of excitatory cells would simply cascade along th e network triggering all downstr eam cells. This would result in a domino effect, where the entire network would be activated and small perturbations would be rapidly amplified. Fortunately, neuronal networks are not purely excitatory. Rather a small percentage of neurons contain inhi bitory synapses that suppress activity in target neurons. This drastically increases the complexity of activity within neuronal networks, enabling neurons to perform the many complex tasks necessary within the brain and nervous system. Botulism Toxin The seven serotypes of botulinum toxin (typ e A-G) are responsible for three naturally occurring forms of botulism: food-borne botulism, wound botulism, and infant botulism. Due to weaponization of botulinum toxins, a fourth form of botulism has been identified as inhalation botulism. The seven serotypes are produced by the bacteria species Clostridium botulinum, Clostridium baratti (type E), and Clostridium butyrium (type F). Serotypes A, B, and E account for almost all cases of human botulism (Simpson 1996). Botulism is characterized by descending flaccid muscle paralysis, beginning with cranial nerves. Current treatment of botulism consists of supportiv e care and passive immunizatio n (Arnon et al., 2001). Equine derived polyvalent antito xin is available for treatment of clinically diagnosed botulism cases.

PAGE 21

21 Since therapeutic antitoxin neutralizes toxin molecules not alr eady bound to neural tissue, it is most effective when administered during early onset of disease (Caya et al., 2004). Complete recovery can take weeks to months and of ten requires extensive medical support including mechanical ventilation (Solomon and Lilly 2001). Botulism toxin (BoNT) is a potent neurotox in that targets peripheral cholinergenic synapses primarily at the neuromuscular juncti on. The toxin internalizes and enzymatically blocks neurotransmitter release resulting in flac cid paralysis. Synaptic transmission is blocked through a multi-step mechanism involving binding, internalization and intracellular cl eavage of soluble NSF attachment receptor (SNARE) prot eins necessary for vesicle fusion docking and subsequent neurotransmitter rele ase into the synaptic cleft. The toxin is a 150 kDa metaloprotease comp rised of a 100 kDa heavy chain and a 50 kDa light chain linked by a disulfide bond (Simpson, 1996). The light chain includes the enzymatic portion of the toxin, which cleaves SNARE prot eins necessary for vesicle fusion docking. The heavy chain has been identified as relevant in toxin binding and intern alization, although the light chain may play an important role as we ll. The toxin may be pa rt of up to a 900 kDa complex, formed with a family of haemaglu ttins (HAs) and notoxic non-haemagluttins (NTNHs) that play a role in targeting the toxin to ne uromuscular junctions, and have been shown to increase toxin potency (Ahnert-Hilger and Bignalk e 1995). However, these auxiliary proteins are not essential for intoxication. Binding. Both proteins and gangliosides have been proposed as BoNT receptors. Gangliosides were initially proposed as receptors for BoNTs and a multitude of studies have supported the theory that gang liosides play an important ro le in binding BoNTs to the presynaptic cleft. Polysi alylated ganglioside GT1b has been shown to inhibit BoNT A binding to

PAGE 22

22 synaptosomes, to quench BoNT A fluorescen ce, and to bind to immobilized BoNT A. Furthermore, monoclonal antibody which bound to GT1b was shown to antagonize BoNT A binding in rat superior cervical ganglion cells (Yowler et al., 2002) In a study by Bullens et al (2002), presynaptic ACh release was studied in wild-type and GalNAc-transferase negative (GalNAc-T-/) mice was studied. The GalNAc-T-/mice lacked complex gangliosides at the neuromuscular junction, but exhi bited unregulated expression of simple gangliosides GD3 and GM3. Wild-type mice exhibited sensit ivity to BoNT A, however, GalNAc-T-/mice lacked toxin sensitivity. The specificity of BoNTs for neuron al membranes may be due to high concentrations of complex gangliosides found there. Crystollogra phic studies of BoNT A and B have revealed a lactose binding pocket located near the c-terminus of the BoNT heavy chain (Montecucco et al., 2004). This site contains a la ctose-binding peptide motif that is highly conserved among the seven BoNTs. Rummel et al.( 2004b) used mass spectrometry to demonstrate that the lactose binding site, bound to a single molecule of ganglioside GT1b. Mutations of this binding site abolished the GT1b heavy-chain complex. Furthermore, GD1b and GT1b were identified as functional receptors for BoNT C, while phosphatidylethanolamine was reported as a receptor for BoNT D (Verderio et al., 2006). However, in many cases, binding affinity of the neurotoxin to gangliosides under physiological conditions is lo w (Nishiki et al., 1994). Additionally, an ultrastructural audioradiographic study of BoNT A an B binding sites on motor nerves indicates that binding sites are unique to each serotype (Black and Do lly 1986; Simpson 2004). Thus, it is unlikely that GT1b is the sole receptor for BoNT A and B. Proteins have also been proposed as a receptor for BoNT. Nishiki et al. (1994) first identified synaptogamin, a synaptic vesicle me mbrane associated protein as a high affinity receptor for BoNT B. Labeled BoNT B bound to prot ein from a crude synapsomal fraction in the

PAGE 23

23 presence of gangliosides in this study. No bi nding was observed in preparations containing only gangliosides. Bound extract was then purfied by DEAE-Sepharose chromatography, PhenylToyopearl chromatography, and Heparin-Toyopearl chromatography. Binding affinity of the heprin elutate closely matched that of the syna psome. However, antibodies raised against the elutate were unable to bl ock binding of labeled BoNT B to synapsomes. In a later study, Nishiki, demonstrated both synapotogamin I and II bound BoNT B, but with different binding affinities (1996). Additionally, binding of BoNT B to synaptosomes was inhibited using a monoclonal antibody specific for the N-terminus of synaptogami n II. This indicated that BoNT B associates with synaptogamin within 20 amino acids of th e N-terminus. Similarly synaptogamins I and II were found to be receptors for BoNT G (Rummel et al., 2004a). Recently, synaptic vesicle protein 2 (SV2) was found to be the protein receptor for BoNT A (Dong et al., 2006). To date no protein receptor has been identified for the remain ing toxin serotypes. This was established using SV2A/B knockout mice, in which BoNT A bi nding was abolished, while SV2B knock-down mice displayed a reduced sensitivity to BoNT A. The picture that is emerging is that binding of BoNTs at the neuromuscular junction is a much more complicated process than that of a single high affinity receptor. Many researchers have concluded that binding of BoNTs involves gangliosides as an initial low affinity receptor and a protein as a high affinity receptor. Toxin may first bind weakly with low-affinity receptors, which induces conformational changes that then bring the toxin in contact with a high affinity receptor. Alternatively, Montecucco et al. ( 2004) proposed that BoNT receptors are comprised of an array of presynaptic recep tors (APRs) including polysial ogangliosides, other lipids, and glycophosphoinostol anchored proteins. In this me chanism, the toxin is first captured by the oligosaccharide portion of the polysialoganglioside s, which extends from the membrane surface.

PAGE 24

24 Because the polysialoganglioside is latterly motile it can then rapidly bring the toxin to the APR. The association constant of the toxin would be the product of th e association constant of each interaction within the APR, thus a large asso ciation constant would result even when weak interactions are involved. Internalization. Toxin internalization may be due to either receptor mediated endocytosis or the vesicle recy cling pathway. While receptor me diated endocytosis was widely assumed as the mechanism by which the toxin was internalized, recent evidence of vesicle associated proteins as high affinity receptors for BoNT suggests it is likely that toxins utilize the vesicle retrieval pathway to internalize. This is further supported by studies that demonstrate that internalization of toxins is ac tivity dependant, and that strong de polarization, such as in an elevated potassium solution, drastically increa ses neurotoxin uptake (V erderio et al., 2006). Regardless of the mechanism by which endocyt osis occurs; the cata lytic domain must completely dissociate from the heavy chain, and en ter the cytosol prior to cleavage of relevant substrates. The mechanism by which BoNTs escape the lumen of the endosome is primarily achieved through pH induced translocation. Th is was demonstrated through a series of experiments with drugs known to neutralize endosomal pH. When these drugs were applied to neuronal preparations, uptake of BoNTs was blocked (Simpson, 1983; Simpson, 2004). Thus, the toxins were shown to contain hydrophobic domains that are exposed when the pH is low. These hydrophobic domains aid in translocation to the cytosol. Two mechanisms for the pH-induced translocation have been proposed. The first invol ves formation of a channel by the heavy chain N-terminus. This theory is based upon an experi ment in which an artificial lipid bilayer was exposed to the heavy chain N-terminus of BoNT B, and changes in membrane conductance were measured (Hoch et al., 1985). It was determined that addition of toxin when the pH of solutions

PAGE 25

25 on both sides of the membrane was balanced, produced only a small change in membrane conductance. However, a large change in membrane conductance was produced when a pH gradient similar to that of an acidified endosomal compartment was introduced, indicating channel formation. Channels were not formed by intact BoNT B. A second model has been proposed in which the heavy chain forms a cleft th at enables passage of the light chain to the cytosol (Simpson 2004). In this case, changes in membrane conductance may be explained as an after effect of residual changes in the membrane. Ultimately, translocation of toxin from the endosomal compartment involves a pHinduced change in toxin structure, translocation of the light chain to the cyto sol, reduction of the disulfide bond that links the heavy and light chain, uncouplin g of all non-covalent forces between the heavy and light chains, and restoration of the light chain to its hydrophobic structure. Additionally, pH induced changes in structure may result in changes to the zincbinding motif, which must be restored once th e light chain has entered the cytosol. Cleavage. The light-chain is a zinc-dependant me taloprotease responsible for cleavage of SNARE proteins that are essential for vesicle fusion docking. E ach toxin serotype cleaves one of three SNARE proteins at a specific site. BoNT B, D, F and G cleave s ynaptobrevin (VAMP), a vesicle associated protein. BoNT A, C and E cleave SNAP-25 and BoNT C1 cleaves syntaxin. Substrate cleavage can only occur prior to S NARE complex formation. Figure 1-1 shows the cleavage sites for the seven BoNT serotypes. BoNT Action on the Central Nervous System Botulinum neurotoxins primarily target peripheral nerve terminals since they do not cross the blood brain barrier. However, central nervous system (CNS) tissue is susceptible to BoNT intoxication. Synaptic vesicle prot ein 2 (SV2), the protein receptor for BoNT A, is expressed in both excitatory and inhibitory neurons of the hi ppocampus and neocortex (V erderio et al., 2006).

PAGE 26

26 Figure 1-1. Botulism toxin clea ves SNARE proteins necessary for vesicle fusion docking. A SNARE proteins cleaved by th e seven BoNT serotypes. B. The SNARE complex is necessary for vesicle fusion docking. C. The substrates for the seven botulinum serotypes. Additionally, synaptogamin (SYT)-I and -II, the protein receptors for BoNT B and G, are expressed in both excitatory and inhibitory synapses of the neocortex and hippocampus. However, GABAergic synapses have been shown to be less susceptible to BoNT A and E then glutameric synapses (Verderio et al., 2004). Therefore cortical neurons are a suitable tissue, for a cell-based BoNT biosensor

PAGE 27

27 CHAPTER 2 REVIEW OF BOTULINUM TO XIN DETECTION METHODS Introduction The standard for detection of BoNT toxins is th e mouse bioassay, which is able to detect as little as 0.02 ng of toxin. Strengths of the mouse bioassay include conceptual simplicity and sensitivity. While the non-selectivity of the mouse bioassay enables it to detect any BoNT serotype, additional neutralization assays are nece ssary to determine a specific serotype. Other limitations of the mouse bioassay include expens e, expertise related to maintaining mouse rearing facilities, and time, since as much as fo ur days may be required to obtain results (Hallis et al., 1996; Witcome et al., 1999). Several attemp ts to replace the mouse bioassay have been made. Methods that have been developed and hold promise for future replacement of the mouse bioassay include immunoassays, PCR assays, and assays based upon protease activities of BoNTs. Currently, no single assay appears to be capable of replacing the broadly applicable mouse bioassay. Therefore, novel to xin detection methods are needed. Mouse Bioassay Currently, the standard for detecting Bo NTs and the only method accepted by regulatory agencies such as CDC and FDA is the mouse toxicity and neutralizat ion assay. The mouse bioassay was first developed by John Trevan in 1927 and was immediately adopted by the scientific community as the standard for toxicity testing (Trevan, 1927). The mouse bioassay is exquisitely sensitive and has been shown to detect as little as 0.02 to 0.03 ng of toxin (Ferreira, 2004). This assay is applicable to a large range of toxin concentr ations, will det ect only active toxin, and is insensitive to cont amination by food, sera, and fecal material. However, it can take several days to perform the assay since it requi res separate tests for detection of BoNTs,

PAGE 28

28 establishment of the minimum lethal dose (MLD) as estimate of toxicity, and neutralization assays to determine the BoNT serotype. Both the Center for Disease Control (CDC ) and the Federal Drug Administration (FDA) have outlined procedures for the mouse bioass ay (Center for Disease Control and Prevention 2001, Solomon and Lilly 2001). In general, the samp le is suspended in gelatin phosphate buffer, centrifuged and clarified supernatant is used for subsequent testing. Fo r each assay, two mice are injected intraperitoneally wi th the specimen. The CDC method simultaneously tests for heat lability, trypsin activation, and neutralization by antitoxin, but does not require serial dilution assays be performed. If mice display botulis m symptoms only in the trypsinated samples, neutralization tests are re peated. The FDA method requires tes ting be conducted in three stages: toxin screening, toxin titer, and toxin neutralization with monovalent antitoxins. Initially, serial dilutions in gelatin phosphate buffer of 1:5, 1:10, and 1:100 must be perf ormed as well as tests using undiluted sample. Trypsin acti vation of each dilution as well as heat liability of undiluted sample is also tested. If all mice, except thos e receiving heated preparations, die, testing is repeated at higher dilutions. Once toxicity test s have been performed, the preparation which yielded the highest MLD (trypsiniz ed or untreated) is used for ne utralization assays. To perform neutralization assays monovalent anti sera (types A, B, C, D, E, and F), which are available from CDC, are used. FDA procedure requires protection of mice by injecting antisera prior to toxicity testing, while the CDC method recommends neutrali zing toxin by adding antisera directly to the sample. Mice are observed at regular intervals fo r a period of 4 to 6 days. Botulinum toxin will typically kill mice within 6 to 24 hours, although death may be delayed. Under the CDC method, if a sufficient amount of toxin is present, only mice receiving samples with the appropriate antiserum survive. Clinical signs of botulism in mice are ruffled fur, followed by labored

PAGE 29

29 breathing, weakness of limbs and total paralysis. These clinical symptoms must be present to confirm presence of botulinum toxin (Center for Disease Control and Prevention, 1998). Although relatively simple in concept, the mouse bioassay is expensive to perform, requires use of a large number of animals, and can take severa l days to obtain result s. Moreover, it can only be performed at the CDC or about 20 other quali fied laboratories in the United States. Thus, the mouse bioassay cannot meet the current need fo r rapid diagnosis of botulism. Additionally, it cannot be used for high volume screening to det ect deliberate contamination of US food, dairy and/or water supply. Enzyme Linked Immunosorbent Assay Due to the need for rapid and specific in vitro assays for detection of botulinum toxins, a variety of antibody based bioassays have be en developed. A summary of antibody based detection methods is given in Table 1. Th ese assays employ the use of polyclonal and monoclonal antibodies to directly detect the presence of BoNTs. Polyclonal antibodies (pAb) are produced by immunization of a host. Purified immunoglobins obtained from the sera of host animals are then used for bio-detection. While polyclonal antibodies enable sensitive detection of antigen, the immune response of each animal is unique. Thus, each batch must be validated and calibrated to ensure minimal variability betw een assays using different antibody batches. Additionally, a finite supply of antisera can be produced by a single animal, necessitating an ongoing effort to produce antibodie s. Monoclonal antibodies (m Ab) are produced by fusion of a single antibody-producing B cell with a meyloma cell producing a hybridoma cell line (Emanuel et al., 2000). While development of monoclonal antibodies is more expe nsive then polyclonal antibodies they enable standardized mass production of antibodi es. However, they may be sensitive to strain-specific variat ions since reactivity can be lo st with even minor changes in antigen. Thus biosensors employing monoclonal antibodies may have mo re variability in

PAGE 30

30 Table 2-1. Immunoassays fo r the detection of BoNTs BoNT Immunosensor Label Matrix Sensitivity Specificity Reference (serotype) BoNT A Amplified ELISA with capture mAb BA11 and pAb guinea pig anti-Bot A Alkaline phosphatase Aqueous suspension, salmon, and corned beef, 10 MLD/ml (66 pg/ml)1 mouse bioassay Shone 1985 BoNT A,B,E ELISA&Enzyme Linked Coagulation Assay (ELCA) using affinity purified horse pAb RVV-XA labeling enzyme Aqueous suspension and culture filtrates A 8 pg/ml, B8 pg/ml, E8pg/ml mouse bioassay Dollegast 1993 BoNT A,B,E,F Amplified ELISA using affinity purified pAb IgG Alkaline phosphatase Culture filtrates, chili 10 MLD/ml (0.20 to 0.30 ng/ml type A toxin)2 mouse bioassay Ferreira 2003, Ferreira 2004 Bot toxoid A, B Multi-analyte fluorescent biosensor using Goat pAb IgG Biotin (capture) Cy5 Bisfunctional Reactive Dye Aqueous suspension toxoid A 40 ng/ml; toxoid B 200 ng/ml ND RoweTaitt 2000 BoNT A,B Bi-diffractive grating biosensor (BDG) with goat pAb anti-BOT A/B and rabbit pAb anti-BOT A/B colloidal gold Aqueous suspension 100 ng/ml ELISA O'Brien 2000 BoNT A Hand-held lateral flow immunoassay utilizing mouse Bot Fab5 colloidal gold Aqueous suspension .25 ng/ml similar assay with Bot-01 mAb Emanuel 2000 BoNT A,B Time-resolved fluorometry assay using affinity purified goat pAb anti-BOT A/B Lanathide (Eu3+) Aqueous suspension, urine, human serum, sewage, dirt 200 pg/ml ELISA Peruski 2002 BoNT E Chemiluminescent slot blot immunoassay using affinity purified rabbit pAb and donkey antirabbit pAb Horseradish peroxidase Enrichment cultures filtrates from fish 4 MLD ( 72 pg)3 mouse bioassay Cadieux 2005 1. Estimate of toxin amount is based on the published specific toxicity of 1.5 *108 MLD mg protein-1 for purified BoNT A (Shone 1985) 2. Estimate of toxin amount is based upon the published estimate of 0.02 to 0.03 ng MLD-1 for purified type A toxin complex (Ferrriera 2004) 3. Estimate of toxin amount is based upon the specific toxicity of 2.8 *107 MLD mg protein-1 for pure BoNT E complex, published in van Barr (2005)

PAGE 31

31 subspecies detection limits. Both monoclonal and polyclonal antibodies may be expensive to produce. Recently, a third alternative has been developed. Recombinant antibodies, consisting of only the antigen binding domain, produced th rough recombinant DNA technology, have been developed (Emanuel et al., 2000) To produce recombinant antibodies, cDNA, made from messenger RNA of an immunized host, is introduced in to Escheric hia coli bacteria. Bacteria are coinfected with a bacteriophage that displays the fragment of antigen binding antibody (Fab) on the bacteriophage's surface with normally occurring coat protein. The use of Fab in a biosensor as alternative to mAb or pAb may significantly reduce cost, while providing a stable genetic source. However, like monoclonal antibodies, Fabs may be sensitive to stra in-specific variations. Antibody based assays include enzyme linked imm unsorbent assays, time resolved fluorescence assays, and lateral flow devices. These assays ma y employ pAbs, mAbs or Fabs as well as any combination of them. Polymerase Chain Reaction Assay Assays employing the polymerase chain reaction (PCR) assays have also been proposed for detection of BoNT production by C. botulinum in food samples (Ferreira et al., 1993; Fach et al., 1993; Campbell et al., 1993; Fach et al., 1995; Sciacchitano and Hirshfield 1997; Aranda et al., 1997; McGrath et al., 2000; Lindstr m et al., 2001; Braconnier et al., 2001). Polymerase chain reactions assays use DNA polymerase, a heat stable enzyme that is only active within a specified range of temperatures, along with synthetic primers to rapidly amplify a target DNA sequence. Thermocyclers and programmable incubators en able DNA to be denatured, annealed with primers, and elongated with DNA polymerase. Because DNA is double stranded, each cycle produces two DNA strands from every one DNA strand present. Thus, the amount of DNA present in a sample increases exponentially. The PCR technique provides a sensitive method for detection of specific, known DNA sequences. It is rapid and reliable and does not require the

PAGE 32

32 Table 2-2. PCR primers and DNA probes for detection of C. botulinum and toxigenic C. butyricum and C. barati Primer/Probe Method Target Gene (size/type) Reference R-BoNT A, L-BoNT A PCR BoNT A (1.34 kbp/amplicon) Ferreira and others 1993 P136, P137 PCR BoNT A(798 bp/amplicon) Fach and others 1993 P138 Digoxigenin-11dUTP labeled probe BoNT(24 bp probe) BoNT 1, 2 PCR BoNT (1.1 kbp/amplicon) Campbell and others 1993 Con.BoNT Fluoresceine BoNT (22 bp probe) Cbot. B dUTP labeled BoNT B (24 bp probe) Cbot. E probe BoNT E (28 bp probe) Cbot. F BoNT F (28 bp probe) Cbar. F BoNT F (29 bp probe) P260, P261 PCR BoNT(260 bp/amplicon) Fach and others 1995 Pr265 DigoxigeninBoNT A(19 bp probe) Pr266 11dUTP BoNT B(22 bp probe) Pr270 labeled probe BoNT E(23 bp probe) Pr272 BoNT F(24 bp probe) Pr280 BoNT G(21 bp probe) E3, E4 PCR/Capillary Electrophoresis BoNT E(513 bp/amplicon) Sci acchiato and Hirshfield 1996 S1-F, S1-R Competitive RT-PCR BoNT E(250 bp/amplicon) McGrath and others 2000 T7 control cDNA cDNA(231 bp) CBMLA1, CBMLA2 Multiplex PCR BoNT A(782 bp/amplicon) Lindstrm and others 2001 CBMLB1, CBMLB2 BoNT B(205 bp/amplicon) CBMLE1, CBMLE2 BoNT E(389 bp/amplicon) CBMLF1, CBMLF2 BoNT F(543 bp/amplicon) fB, rB Nested PCR BoNT B(480 bp/amplicon) Dahlenborg and others 2001 fBn, rBn BoNT B(220 bp/amplicon) fEF, rEF BoNT E,F(630 bp/amplicon) fEFn, rEFn BoNT E,F(200 bp/amplicon) BE1430F, BE1709R Quantitative PCR BoNT E(280 bp) Kimura and others 2001 BE1571FP FAM/TAMRA labeled probe f16S, r16s Quantitative RT-PCR 16sRNA(210 bp/amplicon) Lvenklev and others 2004 16Sprobe reporter quencher labeled probe 16sRNA(25 bp probe) cntBprobe cntB (26 bp probe) TQ A1, TQ A2 real time PCR BoNT A(amplicon) Akbulut and others 2004 TQ B1, TQ B2 BoNT B(amplicon) TQ E1, TQ E2 BoNT E(amplicon) A FAM/TAMRA labeled probe BoNT A(37 bp probe) E BoNT E(31 bp probe) B VIC/TAMRA labeled probe BoNT B(33 bp probe)

PAGE 33

33 production of neurotoxin, in cult ure, for detection. Therefore, the incubation time required to determine presence of C. botulinum may be significantly reduced All PCR assays discussed here target the BoNT gene, enabling the assay to detect only organisms carrying the BoNT toxin gene. Table 2 contains a list of PCR primers a nd molecular probes used for the detection of C. botulinum in food samples. Early PCR techniques were limited to detection of a single strain of C. botulinum and thus were not applicable to detection of unknown BoNT samples (Ferreira et al., 1993; Fach et al., 1993; Sciacchitano and Hirshfield 1996). Additionally, primers propos ed for use in these assays are highly variable in melting temperature, re quiring separate tests fo r each serotype, adding significant reagent costs and testing time. Alte rnative assays, employed a single primer for detection of BoNT gene and separate DNA probe s to identify each serotype (Campbell et al., 1993; Fach et al., 1995). However, the probing st ep for identification of serotype extends detection time. To enable simultaneous and specif ic detection of more than one serotype of C. botulinum a multiplex PCR was developed (Lindstrom, 2001). This assay was able to detect 102 cells per reaction mixture for types A, E, and F and 10 cells per reacti on mixture for type B. Using a two-step enrichment stage the detection limit was as little as 10-2 spores/g of sample for types A, B, and F, and 10-1 spores/g of sample for type E (L indstrom, 2001). The detection limit varied depending upon strain, sample material, temp erature, and enrichment time. Relatively low concentrations of beef inhibite d amplification of types A and F, while higher concentrations inhibited type B and E. An enrichment time of up to 5 days was necessary depending on BoNT strain and sample type. All of the above PCR assays require an enrichment phase to detect the C. botulinum organism at sufficiently sensitive levels. This allows inhibitory substances in foods to be diluted, and prevents false positive results due to non-culturable bacteria (Aranda et al., 1997).

PAGE 34

34 However, addition of an enrichment phase can extend the time required to confirm presence of C. botulinum by several days. It was also found that detection of C. botulinum in food samples was complicated by difficulties including low numbe r of target bacteria, contamination by other bacteria, and presence of other proteases and chelators (Fach, 1993). Also PCR assays are not inherently quantitative, therefore additional postassay processing is required for all such assays to assist in determination of extent of c ontamination. A procedure for PCR detection of C. botulinum neurotoxin A, B, E, and F gene fragments is described in the Bacteriological Analytic Manual (Solomon and Lilly 2001). It may be necessary to test dilutions since DNA concentrations outside of the range from 0.34ng-5160ng DNA/100 l total PCR reaction volume may yield false negative results. Additionally, since PCR assays do not directly de tect botulinum toxin, they are unable to detect food contamin ation where insufficient DNA is present. Thus, a deliberate attack on the food supply using purified or pharmaceuti cal grade BoNTs would not be detected. PCR may also detect silent neurotoxin genes resulting in a positive detection even in the absence of toxic Clostridia spp. Recently, real-time PCR assays have been developed (Kimura et al., 2001; Akbulut et al., 2004). Real-time PCR has advantages of being faster and less expensive than the mouse bioassay. Additionally, fluorescence based real time PCR allows for online monitoring of amplified gene fragments during each cycle of PCR, thus amplification and detection can be performed simultaneously. The real-time PCR assay developed by Akbulut et al. (2004) correlated with the mouse bioassay 75% of the ti me. Some inhibition of PCR from DNA extracts was detected by a Taq-Man Exogenous Internal Positive Control Kit. However, further improvement of the DNA extraction pro cess may significantly improve results.

PAGE 35

35 The PCR assay is used to detect the BoNT ge ne rather than BoNT toxin. Therefore, it cannot give a direct measure of t oxicity. Additionally, because PCR may be capable of detecting spores as well as vegetative cells, application of traditional PCR testing for C. botulinum is insufficient for evaluation of conditions leadi ng to toxin production. To evaluate whether microorganisms released toxin, reverse transcri ption PCR (RT-PCR) assays have been proposed (McGrath et al., 2000; Lvenklev et al., 2004). In this case, gene expression is necessary for detection, thus only organisms actively producin g toxin are detected. McGrath et al. (2000) developed a competitive reverse transcriptase po lymerase chain reaction assay (RT-PCR) for evaluation of C. botulinum Type E toxin gene expression in ty pe E broth and brain heart infusion (BHI) broth. In this assay, exogenously synthe sized control RNA is in troduced with the test RNA prior to reverse transcription by reverse tran scriptase. This allows for the quantification of target nucleic acid by allowing competitive coamp lification of the target nucleic acid along with a control c-DNA. Results of PCR are then compared to a template with known concentrations of the competitor (McGrath et al., 2000). This assay correlated we ll with the mouse bioassay and was able to detect a 100-fold increase in toxin production in type E broth in comparison to BHI broth. Lvenklev et al. (2004) developed a quantitative RT-PCR for the evaluation of BoNT B gene expression in TYP-C brot h (tryptone-peptone-yeast extract broth supplemented with .4% glucose, 0.1% maltose, 0.1% cellobiose, and 0.1% soluble starch). A polyclonal ELISA was used to distinguish between biologically active and inactive toxin. The RT-PCR method is a valuable tool that can be used for monitoring BoNT gene expression and conditions that lead to toxin production. However, because it only detects orga nisms that are actively producing toxin, it is inapplicable to broader toxin detection. Thus it will not only fail to detect contamination with purified toxin, it will fail to detect contamina tion where toxin is present, but the organisms are

PAGE 36

36 not actively producing toxin as we ll. Additionally, RT-PCR has not been applied to detection of toxin production in food and fe cal material. Obtaining high-qu ality RNA from food matrices may pose a significant challe nge to application of RT -PCR to BoNT detection. Protease Activity Assays Botulinum toxins are highly specific endopeptid ases. Therefore, several assays have been developed based upon the proteolytic activity of the BoNTs (Hallis et al., 1996; Ekong et al., 1997; Keller et al., 1999; Witcome et al., 1999; Schmid t et al., 2001; Schmidt et al., 2003; Liu et al., 2003; Anne et al., 2001). Thes e assays use cleavage of a synthetic substrate to detect the presence of BoNTs. A summary of protease activity assays is gi ven in Table 2-3. In an assay developed by Hallis et al. (1996), immobilized VAMP (residues 60-94) or SNAP-25 (137-206) peptide substrates were cleaved by BoNT B and BoNT A, respect ively. Cleaved substrate was detected by antibodies specific to the cleaved substrat e. Sensitivity of this assay, when used with an amplification system, was 0.17 and 0.13 ng/ml for BoNT B and A, respectively. Similarly, a 51-residue peptide homologous to human VAMP2 (re sidues 44 to 94) was used to assess the catalytic activity of BoNT B (Keller et al., 1999). A similar assay was developed for BoNT A using SNAP25 residues 134-206 (Ekong et al., 1997). Detection of the equivalent of 0.2 1.0 mouse LD50 for BoNT A was achieved in a purified clinical preparati on without the need of an amplification system. These assays provide an additional advantage over previously described ELISA techniques since results directly determine the amount of catalytically active BoNT present in a sample. Also, such techniques may be adapted to a high thro ughput format to rapidly screen for BoNTs. However, these methods re ly on antibodies which may be inhibited by some foods similar to traditional ELISA assays (Cen ter for Disease Control and Prevention. 1998). To obtain an assay that is sufficiently robust to a variety of foodstuffs, Witcome et al. (1999) developed an immunoaffinity column with hi ghly specific monoclonal antibodies to capture

PAGE 37

37 BoNT B in food extracts. The immunoaffinity column was then washed and a solution containing peptide substrate ( VAMP residues 60 -94) was added to the column and in the presence of BoNT B the substrate was cleave d. The peptide solution in the immunoaffinity column was then added to a streptavidin-coate d microtiter plate where antibodies specific to cleaved substrate could be used to detect cleaved substrate. As little as 1.5 pg/ml of BoNT B could be detected in toxin spiked foods. Beca use multiple monoclonal antibodies were employed both proteolytic and non-proteolytic bot ulinum strains could be detected. Several assays using fluorogenic synthetic substrates have al so been developed (Anne et al., 2001; Schmidt et al., 2001; Schmidt et al., 2003). A high-throughput assay for BoNTs A, B, D, and F was developed using immobilized substr ates consisting of (1) a cleavable synthetic peptide core based upon the peptid e substrate of the BoNT protei n, (2) glycine spacer groups added to both ends, (3) a fluorescine molecule covalently bound to the N-terminus and (4) a cystine-amide residue as the C-terminus (Schmi dt, 2001). For all toxins, as little as 2 ng/ml produced significant fluorescence above background. This assay did not requi re use of antibodies or reverse phase extraction, which enabled it to be easily adapted to a high-throughput automated environment. While these assays are extremel y sensitive, they cannot distinguish between holotoxin toxins and cleaved li ght chain, thus they do not evaluate receptor-binding or internalization of toxin. Also, use of synthetic substrates, that are not readily available, adds significantly to assay costs. A novel micro-electromechanical sensor (shown in Figure 2-1) was recently developed for detection of BoNT B (Liu et al., 2003). The sensor uses nickel-agarose beads functionalized with synaptobrevin 2-His6, as well as micro-cantilevers f unctionalized with syntaxin 1A-His6. The beads are tethered to can tilevers through a complex formed by the N-terminus of

PAGE 38

38 recombinant synaptobrevin 2 and the N-terminus of the recombinant syntaxin 1A. Resonant frequencies of cantilever vibrations with attached beads are specific and measurable. Subsequent cleavage of synaptobrevin 2 by BoNT B in the presence of zinc ions produces a change in the resonant frequency from about 70 Hz to 1 kHz. This system could detect catalytic BoNT B within 15 min at concentrations of 8 nM or greater. Figure 2-1. Botulinum B micromechanosensor. The C-terminus of Syntaxin 1 A-His6 is attached to a silicon nitride cantilever. The C-terminus of Syna ptobrevin 2-His6 is attached to a nickel agarose bead. The Nterminus of Syntaxin 1 A-His6 forms a complex with the N-terminus of Synaptobrev in 2-His6 tethering the cantilever tip to the agarose bead. This generates resonant frequencies of about 70 Hz. Addition of BoNT B in the presence of zinc ions (Zn2+ ) cleaves Synaptobrevin 2 such that the small segment of Synaptobrevin 2 which remains on the bead cannot interact with Syntaxin 1A. This alters the resonant frequency to 1 kHz. Mass Spectrometry Another alternative assay for BoNTs that has been developed is mass spectrometry. Mass spectrometry has been widely used for both characterization of unknown samples as well as identification and characterization of macro-biomolecules van Baar et al. (2002, 2004) was able to identify both BoTX A, B, D, E, and F using both matrix assisted laser desorption ionization mass spectrometry (MALDI-MS) and coupled li quid chromatography and electrospray mass spectrometry (LC-ES-MS). MALDI is a sensitive method where the desired analyte is combined with a matrix compound and evaporated onto a metal plate (Mikkelsen and Cortn, 2004). Short

PAGE 39

39 pulses of laser radiation are used to release ions from the sample which are passed through a mass analyzer, producing a mass spectrum plot. A da tabase search of known protein sequences enables a confidence parameter of the best matc hes to be calculated. Electron spray (ES-MS) ionization enables ionization of peptides without fragmentation. Liquid samples pass through an electric field dispersing the sample into a fine sp ray of charged droplets. Solvent is evaporated by passing droplets through an inert drying gas re sulting in increased surface charge. When the surface charge equals columbic forces the droplets explode into smaller droplets. This continues until sufficient surface charge energy is present to charge ionize sample molecules. Addition of liquid chromatography enables separation of peptides preventing suppression effects of ionization. Mass spectrometry data using LC-ES-MS enabled ev en higher probabilities for identification of BoNTs. Identification of BoNTs can be complicated due to the presence of HA and NTHA proteins. However, iden tification of auxiliary proteins enables subsequent database searches to be limited to C. botulinum proteins. No data were publishe d on the sensitiv ity of this method. Additionally, only limited strains of neat toxin proteins were tested. Identification of BoNTs in crude samples, including pr otein matrices may prove difficult. Recently Barr et al. (2005) developed a mass sp ectrometry assay that detected synthetic substrates cleaved by BoNTs A, B, E, and F. Th is enabled rapid serotype specific detection of BoNTs. Additionally, a multiplexed method was developed by combining all four substrate peptides into a single sample. The product pept ides were then distinguished based upon their mass. Detection limits for this assay were approximately 0.62 MLD50/ml for BoNTs A and B and 0.31 MLD50/ml for BoNTs E and F. Primary limitations of this assay are equipment, and high level of expertise required to operate equipment. Further validation of th is assay is necessary with crude samples.

PAGE 40

40 Discussion Over the past decade tremendous progress has been made in development of novel assays for detection of botulinum toxins, however, none of these assays appears capable of challenging the conceptual simplicity, non-selectivity and se nsitivity of the mouse bioassay. Many areas would benefit from availability of rapid, r obust and inexpensive BoNT assays including food processing, packaging and dist ribution, food packaging relate d research and development, therapeutic testing, basic research focused on evaluating conditions leading to toxin production or inhibition, and confirmation of clinical botu lism. However, because deliberate release of BoNTs poses a significant threat to the United Stat es population, the most critical need for BoNT assays may be rapid and reliable detecti on of weaponized BoNTs. The September 11, 2001 terrorist attacks in the US demonstrated that such events are carefully planned and orchestrated. It is believed that attacks on the US food s upply will likely be carried out with similar sophistication. Thus, assays to detect deliberate release of BoNTs must be sensitive, robust, and capable of detecting serologically and antigenically varied BoNTs. Additionally, as demonstrated by the 2001 anthrax events laborato ries screening for BoNTs must be capable of performing tens or hundreds of thousands of assays or more over a period of a few weeks. Since purified toxin may be used in a biological attack, direct detectio n of toxin is preferred to other indicators, such as nucleic acids. While the majority of the assays described here are capable of detecting BoNTs well below estimated human lethality limits, very fe w have been shown to maintain sensitivity in complex food matrices and against multiple BoNT strains. Those that have maintained sensitivity are currently limited to single serotypes. Both ELISA and PCR assays may provide a crude screening device for weaponized BoNTs. However, careful consideration shoul d be made for inhibition due to sample contamination. Additionally, PCR assays may be inhibited by the presence of a wide variety of

PAGE 41

41 bacteria. The presence and fre quency of naturally occurring C. botulinum species should also be considered to limit the number of positive PCR assay results. PCR screening for weaponized BoNTs may be inadequate since purified toxin may not be sufficiently contaminated with DNA to enable detection. Assays requiring enrichme nt stages may be inappropriate since weaponized BoNTs may not contain viable or ganisms. Assays employing mAbs and Fabs may be sensitive to strain-specific variations. Expl oiting multiple mAbs or Fabs raised against a single BoNT serotype may enable more sensitive and robust BoNT detection. Routine screening assays at critical points along supply chai ns may help ensure safety while discouraging terrorists from using foods as a vehicle for terrorism. For an assay to be applicable for routine-screening, it must be ra pid, robust and inexpensiv e while minimizing false positive and negative indications. Routine screeni ng assays would ideally consist of a bench-top or portable device that requires little equipment and expertise to operate. Rapid identification of the agent responsible fo r a suspected bioterrori sm attack is also necessary to reduce further illness and issue prop er recalls while minimizing panic. While much work can be accomplished through appropriate epid emiological studies, final confirmation of a suspect food contaminated with a weaponized Bo NT currently requires a mouse bioassay. This may lead to both loss of valuable time and money. Thus, replacement of the mouse bioassay with a more rapid, but equally sensitive assay remain s necessary. It may be possible to reduce the number of suspect foods with a crude field test, and then perf orm a more complex assay that matches sensitivity of the mouse bioassay and mouse neutralization assay at select laboratories. In this case, confirmation of the suspect f ood could be made in a few hours at multiple laboratories nationwide by trained la boratory technicians. Thus a mo re elaborate assay, such as the assay proposed by Witcome and others (1 999) could be used for confirmation.

PAGE 42

42 Clearly, there is a continuing need for new and better assays for BoNT toxin detection as well as additional validation of existing assays Assays developed should be inexpensive and sufficiently robust such that they will detect any toxin serotype in a variety of substrates including sera, food, and fecal material. Assays should provide broad-based indications, like the mouse bioassay, together with cap ability to identify specific toxins involved. They should be conceptually simple, easily performed, inexpensive, reproducible, and rapid.

PAGE 43

43 CHAPTER 3 NEURAL NETWORKS GROWN ON MICROELECTRODE ARRAYS AS CELL-BASED BIOSENSORS Introduction One of the first living biosensors, the canary in the mine shaft, was a rapid and reliable method for detecting toxic conditio ns within coal mines. However, most modern "living" biosensors have little resemblance to the canary in the mine shaft. Rather, they rely on the biological complexity of intact ce lls or tissue for rapid and reliab le detection of desired analytes. Cell-based biosensors incorporate whole cells, or in tact tissue from bacteria, animals, and plants. They are often capable of specificity and sensit ivity, which is difficult to mimic with more conventional biochemical detecti on methods. Furthermore, they ar e capable of detecting more complex analytes. For example, botulinum toxi n detected using the cell-based biosensor developed in this work, must be capable of a th ree-step mechanism (binding, internalization and cleavage of SNAP-25) prior to de tection, eliminating false positives from toxin fragments, or denatured toxin. Furthermore, cell-based assays are often capable of mimicking in vivo conditions, thus providing valuable insight into how analytes behave in vivo Neural cultures, grown over micro-electrode arrays (MEAs) represent a potential cellbased assay that may be applicable for a broad a ssortment of tissue. An MEA is a tissue culture dish with a grid of 60 indepe ndent extracellular electrodes em bedded under the surface. Each electrode measures electrophysiologi cal activity (i.e., action-potent ials) of the neurons (Figure 31) for extended periods of hours to weeks (Pot ter and DeMarse, 2001). Mi cro electrode arrays have been employed to measure the activity from a variety of electrically ac tive cell lines such as cortical, hippocampal, and spinal neurons (Seg ev et al., 2002; Marom and Shahaf, 2002; Selinger et al., 2004; Eytan et al., 2003; Wagenaar et al., 2006a, Wa genaar et al., 2006b), cardiac

PAGE 44

44 myocytes (Reppel et al., 2007) and retinal ganglio n cells (Meister et al., 1995; Meister et al., 1994). Use of cortical cultures grown on MEAs fo r detection of environment threats is an application currently in developmen t at a variety of labs (Gholmieh et al., 2001; Gholmieh et al., 2003; Gross et al., 1997; Gramowsk i et al., 2004; Keefer et al., 2001; Pancrazio et al., 2003; Stenger et al., 2001; Selinger et al., 2004) MEAs have thus far been us ed to detect a variety of environmental threats including mercuric chlori de, sodium arsenate, phosdrin, chloridimeform, strychnine and brevetoxin. These threats come from a broad assortment of contaminants including heavy metals, organophosphates, insectic ides and marine neurotoxins. Additionally, MEAs have been used with a variety of pharmaco logical agents including bi cuculline, valproate, atropine, and methanandamide (Cadotte and DeMarse 2005; Cadotte 2004; Morefeild et al., 2000; Gramowski et al., 2004). These biosensors typically use changes in the spatial and temporal patterns of activity across a network of neurons as an indicator of toxici ty. One of the most prominent patterns of activity in these cultures, known as population bu rsting or bursts, consists of network wide synchronous volleys of action potentials (i.e. spik es) observed on nearly all electrodes that last anywhere from 100 ms to several seconds. Both sp iking and bursting activity are highly sensitive to environmental changes, incl uding the introduction of chemical agents to the culture. Moreover, measuring network-wide activity enables detection of subtle changes in neuronal dynamics that may be missed by other single neur on based assays. Hence, monitoring changes in dynamics of these neural cultures following application of test agents could provide a valuable diagnostic tool and practical biosensor for a variety of neurologically active agents.

PAGE 45

45 Hardware and Software Micro-electrode arrays (MEA) The MEAs (Fig. 3-1) used in this study obtained from Multichannel Systems GmBH. The MEAs are comprised of a glass wafer over whic h gold was deposited and li thographically etched to form electrode pads traces. Silicon nitride is vapor deposited to form an insulating layer over the surface of the MEA. Holes in the silicon nitride insulation are plasma etched at the electrode pads and the gold electrodes are coated with a thin layer of Titanium nitride (TiN), to form a low impedance interface between recording amplifie rs and culture media. Electrodes have a 30 m diameter and are spaced 200 m apart. The entire array covers about a 1.6 mm2 area located near the center of the dish. A large reference electrode is located on the right side of the dish. Each electrode records changes in ionic voltage within an approximate radius of 60 m. Figure 3-1. A micro-el ectrode array (MEA) from Multichann el Systems. An 8 x 8 grid of electrodes 200 M apart (right panel) enables the recording of changes in extracellular voltages produced by activity (action potentials, lower left) from cortical networks grown in vitro.

PAGE 46

46 Recordings were conducted in an environmen tal chamber that was maintained at 35.5C and 5% CO2 to duplicate the environment within th e incubator in whic h the cultures were maintained. Data acquisition Connection to the dishes is made using the MEA60 amplifier system made by Multichannel Systems (Ruetlingen, Germany). Th e close proximity of the amplifier to signal source enables good signal to noise ratio. The MEA 60 amplifier system consists of an amplifier with a set gain and a bandpass filter. Data from the MEA 60 amplifier system is digitized using MCCard at 25 kHz. A custom stimulation boa rd, shown in Figure 33B developed by Dr. Thomas DeMarse and Vihn Trihn enables the system to rapidly toggle between application of a biphasic stimulation current and data acquisition. Real time data is processed using MEA Bench OSX, a custom software package that interfa ces with the Mac OSX platform (available at http://neural.bme.ufl.edu/index.html ). This software enables onlin e viewing of raw data, realtime spike detection and recording of raw and sp ike data. It also provi des a user interface for stimulation and stimulation artifact suppression using techniques detailed in Wagenaar and Potter (2002). Physiological conditions Recordings were conducted within a custom made environmental chamber shown in Figure 3-3 A. Temperature was maintained at 35.5 oC and CO2 levels were maintained at 5% during recordings using Multichannels Systems TCO2 temperature controller. Media osmolarity was maintained by minimizing evaporation from the dish during recordings. To prevent evaporation, a wet towel was placed in the micro-incubation chamber to increase humidity within the enclosure. Additionally, to prevent a temperature gradient from forming between the bottom and the surface of the dish, the MEA 60 recording board was elevated one inch above the

PAGE 47

47 bottom of the chamber using four legs. This enab led heated air to circulate both above and below the dish. A. B. Figure3-2 Data acquisition hardware. A. St imulation board. B. Environmental chamber Dissociated Neural Cultures Tissue Culture Methods Detailed methods used for neuronal cultures ha ve been previously described (Potter and DeMarse, 2001). Briefly, cortical hemisphere s from E17 E18 Wistar rat embryos were dissociated using papain and m echanical trituration. After di ssociation, approximately 50,000 cells (neurons and glia) were placed in the ce nter of a MEA. The surface of the MEA was treated with polyethyleneimine (Sigma-Aldrich, St Louis MO) to make the surface preferable for cell adhesion. Polyethylenimine is a branched organic copolymer containing a large number of positively charged amino gr oups. Treatment of dishes with polyethylenimine enables long term adhesion to the dish surface, and prevents clumping of neural culture s. Dishes were also treated with laminin (Sigma-Ald rich, St Louis MO) prior to cell placement to promote neurite outgrowth. Laminin was isolated in the intact native form, from Engel Broth-Holm-Swarm tumor, a mouse tumor which produces large amount s of basement membranes. Laminin contains active domains for binding, cell adhesion, heparin binding and neurite outgrowth.

PAGE 48

48 Neurons were cultured in medi um consisting of 90% Dulbeccos modified eagles medium (DMEM) (Invitrogen Corporation, Carlesbad, CA) and 10% Equine Serum (HyClone, Logan UT) that was equilibrated to 5% carbon dioxide and 35.5oC. The medium for each culture was exchanged twice per week and cultures were incubated continuously in 5% CO2 at 35.5 oC. To enable long term development, sterile conditions were maintain ed within the tissue culture chamber of the MEA. To maintain steril ity, MEAs were fitted with a Teflon ring that seals around the outer wall of the tissue culture chamber. The top of the teflon ring was covered with a FEP Teflon film. This film was permeable to gases such as CO2 and O2 but was impermeable to water and other contaminants. L ong-term tissue culture methods are described in detail in Potter and DeMarse (2001). Development and activity Cultured cortical neurons, when plated on ME A dishes, begin organizing into networks within a few hours of plating. Persistent tonic firing is usually observed within 2 to 5 days in vitro (DIV). Synchronized bursting typi cally appears within 7 DIV, but may appear as late as 14 DIV. Bursting may be preceded by large single channel or small multi-channel (<5) bursts of activity that persist for up to two days. During th e initial period when burst ing appears, bursts are often large, lasting 1-5 seconds. These large burst are followed by long quiescent periods of 3-15 min, in which no bursting occurs. As cultures matu re, larger bursts are repl aced by shorter, more regular bursts. Often, around 10 to 11 DIV, clusters of bursts appear. Thes e bursts are frequently referred to in literature as super bursts. Superbursts include stable clusters of 5-7 bursts (up to 12) with an inter-burst interval (IBI) much smaller than the mean IBI. Alternatively, longer less stable volleys of bursts may appear. Occasionally these super bursts persisted through the life of the culture.

PAGE 49

49 Activity typically stabilizes around 30 DIV, with short regular burst s. Burst rates are variable between cultures, alt hough much less variable within batches plated at the same Figure 3-3. Screen shot of t ypical raw activity during a netw ork-wide burst. Each window represents a single electrode in the 8 by 8 grid of electrodes, and is 200 ms wide and 50 mV high. time than between batches. Cultures may burst as lit tle as a few times per minute, or as often as every 5 seconds. Activity also depends on cultu re density, which may partially explain variability between batches. Additionally, cultures with higher levels of tonic spiking activity typically achieve maturity faster. Thinly plated cultures generally exhib it delays in maturation while sparse cultures may never achieve stable short bursts. At four weeks in vitro the cultures were considered mature (Marom and Shahaf, 2002; Habets et al., 1987; Muramoto et al., 1993). All experiments discussed here were pe rformed on mature cultures (> 28 DIV).

PAGE 50

50 Network Bursting Network wide bursts are the dominant pattern of activity observed in cortical-networks grown in vitro Spontaneous synchronized patterns of ac tivity are found in a variety of tissues including the developing visual system, epileptogenesis, and in the central nervous system development. However, mechanisms leading to bu rsting activity have been the subject of broad debate. Several theories have been proposed. One th eory was that activity is the result of a class of specific pacemaker cells that form a trigger network, or that bursting is the result some chemical factor X which is pres ent in the culture. Medea et al (1995) disproved the factor X theory by slicing networks into quarters and observing bursting that was independent even when initiated close to borders. He concluded that bursting was governed by the level of presynaptic firing, degree of connectivity, and by excitation and recovery proce sses. Shao et al. (2006) used a model to demonstrate that burst-like activity co uld be generated independent of ion channel properties, or synaptic mechanisms. Thus, he concluded bursting is critically dependent upon connectivity. This is, at least in part, in agreemen t with what is observed in the laboratory since sparse cultures did not produce bursting. Many theories have also been proposed about mechanisms involved in burst termination in random cultured networks. These include synaptic disinhibition as the mechanism by which bursts terminate. However, di sinhibition was shown to induce slow bursting rather than suppressing it. Alternatively, St reit et al. (2005) proposed that slow spike frequency adaptation may be the primary mechanism by which bursts te rminate. Spike frequency termination may be due to accumulation of Ca2+ dependant K+ currents or slow inactivation of Na+ currents. Since increases in extracellular Ca2+ did not terminate bursts, it wa s concluded that progressive inactivation of sodium channels was th e mechanism by which bursts terminate.

PAGE 51

51 MEAs as Biosensors A variety of inherent network properties have been used to measure the changes in pattern activity. Some of the more common methods used include changes in burst rate, spike rate, and burst duration, mean spikes per burst, interspike interval, burst amplitude, and spike amplitude (Gramowski et al., 2004; Keefer et al., 2001; Pancrazio et al., 2003). In attempting to identify unknown substances using neural networks grown on MEAs Gramowski et al. (2004) calculate d the coefficient of variation (CV) across the network and across time for native activity and five pharmaco logical agents. Additionally, an autocorrelogram of a spike timestamps was fitted to a Gabor func tion. Amplitude, decay, exponent, and offset of this function were then added to standard data feat ures, such as spike rate and burst rate, to create a database of standard activities for the six pharmacological agents. The database was then compared to activity patterns of six unknowns. Using the Euclid ian distance, the similarity between unknown agents and the database of agents was analyzed. This enabled correct identification of two of the agen ts and the native state 100% of the time, two agents 94% of the time and one agent 84% of the time. This work demonstrated how a biol ogical fingerprint of several analytes can be generated.

PAGE 52

52 CHAPTER 4 DEFINING THE BONT A FINGERPRINT Introduction In this chapter the potential of cortical networks cultured on MEAs as a biosensor for BoNT A by bath application of the toxin is assess ed. The activity of these cortical networks over the course of one week, was measured to character ize changes in the spatiotemporal patterns of activity relative to untreated c ontrols indicative of BoNT A. Materials and Methods Procedure Neurons began producing spontaneous action potentials within the first few days after plating on MEAs. Cultures were considered mature at four weeks in vitro at which point neurons expressed the full complement of rece ptors (Marom and Shahaf, 2002; Habets et al., 1987; Muramoto et al., 1993). All cultures used were older than 30 DIV to allow for cultures to be considered fully developed. Recordings we re conducted within an environmental chamber that was maintained at 35.5 C and 5% CO2 to duplicate the environment within the incubator in which cultures were maintained. Ten MEAs were randomly assigned to either a BoNT A group (N=5) which received toxin or a Control group (N=5) without to xin. Twenty minutes of spontan eous neural activity recorded prior to treatment with and wit hout BoNT, was used as a baseline activity measur ement. Activity was then recorded immediately following applica tion of BoNT for the BoNT A group and during a similar period for the Control group. This wa s followed by twenty minute spontaneous activity recordings at 12, 24, 36, 48, and 72 hours, and 1 week after treatment. Following each recording, five of the 60 channels were stimulated in a randomized order to measure any changes in amount of evoked activit y between the BoNT and Control groups. Each

PAGE 53

53 stimulation consisted of a singl e 600 mV, 200 S, biphasic pulse, that typically evoked a short burst of activity lasting 100 ms or more. The five stimulation sites we re selected prior to the first recording with the criterion that a single stimulation pulse would produce activity on at least 50% of the channels. Each site was stimulated with five pulses at 0.10 Hz. Stimulation order was randomized during each session. At the end of the week of r ecordings, MEAs in the Control group that had not been dosed with BoNT A were then dosed with BoNT and recorded for a one week period similar to the BoNT group. One of the control cultures during this reversal had to be removed from the experiment due to infection. Drug application Lyophilized BoNT A complex (BOTOX, Allergen Pharmaceuticals) was reconstituted in 0.9% sterile saline at a concen tration of 200 U/ml (approximat ely 10 ng/ml). Each vial of BOTOX contains 100 U of botulinum toxin type A, 0.5mg of human albumin, and 0.9 mg sodium chloride (Allergens Pharmaceutical Irel and 2004). Twenty micro-liters of toxin solution (4U of toxin) was applied directly to the 1 ml culture bath. Th e concentration chosen was well below human lethality limits for safety reasons. In addition, it was sufficiently low to demonstrate sensitivity of the networks for BoNT A. This is particularly important since dilution of crude samples will likely be necessary prior to dosage of networks in future studies. Toxin carrier solution, consisting of albumin diluted in 0.9% sterile saline to a concentration of 1 mg/ml, was applied to control cultures. Analysis and statistics. Action potentials (i.e. spikes) produced by neur ons grown on the surface of the array were detected as voltage deviations greater than five standard deviations above or below the mean noise level for each electrode. Activity was also stimulated through the application of 600 mv, 200 S, bi-phasic voltage pulses at select el ectrode sites (Wagen aar and others 2004) Spike

PAGE 54

54 events were extracted and stored online for further offline processing to identify burst events. Burst events were detected by sm oothing the number of spike times from all channels into 0.5 ms bins. Consecutive bins that exceeded a th reshold number of spikes based upon the mean bin size, were then identified as bursts. The end of each burst was detected as a 10 to 20 ms period in which no activity was detected. Data is presented as a mean percentage of the baseline activity +/S. E.M. The separate variance t-test was us ed to establish significance between control and BoNT dosed groups (cultures) with P < 0.05 consider ed as significant. The Paired Student's t-test was used to compare post-treatment activity to baseline (within groups). The Paired Student's ttest was used as well as to compare BoNT dosed control dishes from the reversal to control activity obtained the week prior. In this case, results were c onfounded with blocks (week 1 and week 2). Results Spontaneous activity Application of the toxin resulted in a significant changes in th e spontaneous activity before versus after treatment in the BoNT group and relative to the Control group. In 89% of BoNT treated cultures applicati on of BoNT resulted in an increase in duration of spontaneous bursts and the number of spikes within those bursts with respect to activity recorded during baseline (before the toxin had been delivered). Figure 2 sh ows the mean duration of each burst (Panel A) and average number of spikes pe r burst (Panel B) in the BoNT and control groups from 12h to 1 week following BoNT application. Dishes dos ed during the control reversal also had a significant increase in burst duration and spikes per burst (data not s hown). Increases were observed within 12 to 24h after BoNT dosing and continued to increase until 72h. Increases were significant 48h after dos age with BoNT A (P < 0.05). Th ere was, however, no significant difference between spike rate during each burst or inter-burst interval immediately following the

PAGE 55

55 0 20 40 60 80 100 120 140 160 180 200 12h24h36h48h72h1wk Time Post TreatmentPercent Change in Burst Duration Control N = 5 BoNT N = 5* A Burst duration 0 20 40 60 80 100 120 140 160 180 200 12h24h36h48h72h1wk Time Post TreatmentPercent of Baseline Spike Rate Control N = 5 BoNT N = 5B. Spikes per burst* Figure 4-1 Spontaneous burst activity in cortical networks after tr eatment with BoNT A or toxin carrier. A. Average burst duration as a percen t of baseline activity. B. Average spikes per burst as a percentage of baseline activity. A. Average burst duration as a percent of baseline activity. B. Average spikes per burst as a percentage of baseline activity. Indicates a significant increase in burst dura tion over baseline activity (Students t-test, p < 0.05), control N =5, BoNT N = 9 application of the toxin (data not shown in the fi gure, P > 0.05). Finally, there was no significant difference between the BoNT and control group pr ior to the application of the toxin for spike rate, burst rate, inter-burst interv al and burst duration (P > 0.05).

PAGE 56

56 Intra-burst neural dynamics Application of the toxin not onl y changed the number of spik es in a burst and the burst duration, it also affected the dist ribution of spikes within each burst event. To illustrate this change return plots were generated in which the activity during a burst was plotted as a series of points comparing the rate of activ ity at time t to the activity observed 50 ms in the past. In a normal culture, a typical burst woul d consist of tonic sp iking observed across the majority of the array in which a rapid increase in activity peaks within 50ms follo wed by a slowly decay appearing in a return plot as a loop beginning at the origin (W agenaar and others 2006b; Segev and others 2001; Tabak and Latham 2003; van Pelt and others 2004; van Pelt and others 2005; Kamioka and others1996; Jimbo and others 2000) Figure 3 shows return plots for the contro l group (left column) and BoNT group (right column) for three subjects during baseline, 24, 48, and 72 hours. The normal burst patterns are illustrated in the left column of the figure for tw o subjects in the control group. A similar pattern can also be seen for the baseline activity in the BoNT group befo re the application of the toxin (right column, blue line). In contrast, the shape of the trajectories duri ng each burst was altered after delivery of the toxin in the BoNT group (right column). After 24 hr, slight oscillations in the rate of activity durin g the burst begin to appear that became more pronounced over time. The oscillatory behavior represents the appearance of multiple peaks of activity (three or more) within each burst. This was tr ue particularly after 48 and 72 hours and was apparent in 50% BoNT cultures, but never observed in controls. Note the scale of the left upper panel of Fi gure 3, labeled as control 7079, is nearly on order if magnitude smaller than the other panels This is because the network covered a much smaller portion of the array, approximately 50% However despite the much smaller magnitude

PAGE 57

57 observed in the burst trajectory, the shape of the burst trajectory remained stable over the seventy two hours shown. Figure 4-2. Return plots generated by sorting spikes in to 100 ms bins and pl otting bins at time t against bins at time t 50 ms. Responses ar e from baseline recordi ngs (dark blue), as well as 24h (light blue), 48h (pink) and 72h (re d) in two carrier dosed cultures (left column) and two BoNT dosed cultures (right column). The bottom row is the same tissue culture with and without BoNT treatment (right column, blue line). In contrast, the shape of trajectories during each burst was altered after delivery of toxin in the BoNT group (right column). After 24 hr, slight oscillations in the rate of activity during the burst begin to appear that became more pronoun ced over time. The oscillatory behavior represents the appearance of multipl e peaks of activity (three or more) within each

PAGE 58

58 burst. This was true particularly after 48 and 72 hours and was appare nt in five of the nine BoNT cultures, but never observed in controls. Evoked response following in toxication with BoNT A The toxin's effect on spontaneous activity wa s also compared with activity evoked through a series of five brief biphasic pulses after each recording session. Application of toxin resulted in significant changes to the overall ra te of spiking of activity in co rtical culture during bursts, and the pattern of spiking relative to the control group. Figure 4-3 shows an example of action potentials (plotted as points) produced for each channel following a single stimulation pulse (panel A, B) and the average activity across ch annels (panel C, D). There was no significant difference in the number of evoked spikes betw een the BoNT and Control group during baseline, or 12 hours after toxin delivery. However, there was a significant increase in number of spikes evoked in the BoNT vs. Control group 24, 48, and 72 hours following the toxin which persisted up to 1 week. The increase in sp ike activity was most apparent in the late phase of the evoked burst (100 to 300 ms post stimula tion). Figure 4-5 shows the average number of late phase spikes evoked through stimulation before, 12, 24, 48, 72 hours, and 1 week after toxin delivery in the BoNT and Control group. There was no signifi cant difference between the number of stimulations that resulted in a burst (Control 20.0+/-0.91, BoNT 21.0+/-0.92) A comparison between patterns of evoked act ivity 300 ms following stimulation in the control group (top panel) versus the BoNT group (bottom panel) is shown in Figure 4-5. There was no significant change in activity observed in BoNT A dosed cultures during baseline or 12 hours following administration of the toxin. Howeve r, beginning at 24 hours, eight of the nine BoNT dosed cultures displayed elevated activity dur ing the later part of the elicited burst (>100 ms post stimulation) that persis ted up to 1 week following toxin delivery. Interestingly, after 36h multiple peaks in the rate of activity was observed in five of nine BoNT dosed cultures and eight

PAGE 59

59 Figure 4-3.Stimulation-response for cortical culture grown on MEA 9053 during the baseline recording (panels A, C) and 72h after trea tment with 200 pg BoNT A (panels B, D). Raster plot after a single stimulation (top panels), and average stimulation-response plots for each stimulation ch annel (bottom panels).

PAGE 60

60 Figure 4-4 Mean change in the number of evoked spikes 100 to 300 ms post stimulation in the BoNT (closed bar, N=5) and control cultu res (open bar, N=5). Due to culture to culture variability the data was normalized as a percent of baseline activity, where 100 percent indicates no change in activity. The baseline average of spikes 100 to 300 ms post stimulation for control and BoNT gr oups is shown in the panel inset. There was no significant difference between baseline activity of BoNT A dosed and control cultures. of the nine cultures at 72h record ing. Initial peaks in the rate of activity observed in both the experimental and control groups were followed by a series of smaller peaks in the BoNT but not control cultures. Each peak in the activity rate of the BoNT group was slightly smaller, and slightly longer in duration than the previous peak, lasting 50 to 100 ms and spaced 50 to 100 ms apart. This is similar to effects observed during spontaneous recordings. Moreover, this oscillatory effect was still appa rent even one week after toxin application in the BoNT group.

PAGE 61

61 Figure 4-5. Average stimulati on response for Control (Panel A) and BoNT dosed (Panel B) tissue cultures. Note the increase in sp ike rate during the latter portion of the stimulation response (>100 ms) as well as th e emergence of oscillatory behavior most prominent at 72h.

PAGE 62

62 Discussion Use of cultured neural networks as a biosensor depends upon identification of distinctive modifications to native activity, which can serve as a fingerprint for biological agents of interest. A variety of inherent ne twork properties have been used to measure the changes in activity. Some of the more common methods used include changes in burst rate, spike rate, and burst duration, mean spikes per burst, interspike interval, burst amplitude, and spike amplitude (Gramowski et al., 2004; Keefer et al., 2001; Pa ncrazio et al., 2003). The effects on neural activity of an important food-borne toxin, BoNT A, have been examined. Application of BoNT A to the culture media resulted in an increase in number of spikes per burst within 36h after BoNT dosage. Furthermore, an oscillatory change in bursting activity was observed, which to our knowledge has not been previously reported w ith other biological agents. These changes are present not only in spontaneous bursts, but al so in bursts elicited through stimulation. As shown in Figure 4-2, the size and shape of spontaneous burst activity produced in a single recording session can be highly variable (Wagenaar et al., 2006b). Thus, long recording periods are necessary to obtain re presentative samples of activity. However, elicited bursts tend to be less variable in size and shape and are produced reliably over time. Because of this, the amount of recording time necessary to obtain repr esentative samples of activity is lower for stimulation experiments then spont aneous recordings. Therefore, it is anticipated that monitoring stimulated burst activity will great ly aid in the detection of BoNT A by reducing variability and time. Additionally, return plots of spike activity are a simple met hod to rapidly visualize changes in burst activity. These plots can easily be obta ined during online data processing since they do not require a prior knowledge of burst activity. T hus, they may be a valuable tool for visually identifying changes in activity associat ed with toxin or drug application.

PAGE 63

63 Why this oscillatory behavior appeared following toxin application during both spontaneous and evoked bursts is not clear. Previous studies have demonstrated that the primary target of BoNTs in the hippocampal and neocor tical tissues are excitatory synapses, while GABAergic synapses remain much less effected (Verderio et al., 2006). For example, BoNT has been shown to block presynaptic release of not only acetylcholine, but glutamate, glycine, noardrenaline, dopamine, serotonin and neuropeptid es (Keller et al., 2004; Verderio et al., 1999; Verderio et al., 2006). Factors such as structural conn ectivity, synaptic strength, and the balance between excitatory and inhibito ry contributions with in these cultures have been shown to produce a profound effect on the pattern of activ ity observed, especia lly during development (Nakanishi and Kukita 1998; Streit et al., 2005; Wagenaar et al., 2006b). Thus, it is not surprising that this toxin would alter the pattern of activity within each burst. In this experiment at least 36 hours after toxin exposure was requi red before a significant change in activity was observed. This is similar to th e mouse bioassay which requires from 12 to 72h for BoNT detection perhaps reflecting the slow t oxin uptake (Scarlatos et al., 2005). Additionally, ongoing changes in activity observed after the 24h recording may be a result of further SNAP-25 cleavage after BoNT uptake with the maximum le vel of SNAP cleavage not occurring until 48 to 72h post treatment.

PAGE 64

64 CHAPTER 5 REFINING THE BONT A FINGERPRINT AND DEVELOPMENT OF A BONT DETECTION ALGORITHM Introduction Work highlighted in the previous chapter dem onstrated that MEAs are a suitable substrate for detecting BoNT A action on neocortical neuron s. A possible explanation for the oscillatory changes observed during bursts was a reduction in s ynaptic strength. Synap tic strength between connected neurons can be modified by a variety of factors including the amount of presynaptic neurotransmitter release and the post-synaptic response to neurotransmitter release. An approach to assessing changes in synaptic strength associated with BoNT intoxication is to evaluate stimulus dependant plasticity. Additionally, a t ool adapted to neuroscience from economics, known as Granger's Causality (GC) can be used to evaluate changes in connectivity associated with intoxication. Finally, an application of an algorithm for BoNT detection was used to determine BoNT A in tissue cultures dosed with 200pg, 600pg and 1000 pg of BoNT A. Background Plasticity Plasticity of the nervous system through act ivity-dependent modification of synaptic strength is central to development (Meister et al., 1991), learning, and memory (Chirwa et al 2001; Dan and Poo, 2006). Specifically, long term potentiation (LTP) and long term depression (LTD) are two key mechanisms by which synaptic efficacy may be either increased or decreased respectively. Synaptic strength depends on se veral parameters, including initial release probability of the synapse, size of the readily re leasable vesicle pool and release probability of each vesicle. It has been demonstrated that spike timing plays a central role in plasticity. When postsynaptic spiking precedes pres ynaptic spiking (post-pre) within a short interval, the synapse

PAGE 65

65 is depressed via LTD. Alternativ ely, if presynaptic spiking precedes postsynaptic (pre-post) spiking within a short interval the synapse is enhanced via LTP (Poo et al., 2001). Spike intervals necessary for LTP and LTD are dependent upon neuronal type, and range from 10 to 100 ms. Jimbo et al. (1999) developed a method for de tecting stimulus driven LTP and LTD in random neural networks grown on MEAs. In essence a 600 mV biphasic probe sequence is applied to the 60 MEA channels ten times. Then a 600 mV 20 Hz tetanic stimulus is applied at a single electrode site. Then the pre-tetanic probe is repeated. Pre-tetanic spike rates for each channel are compared to post-tetanic spike rates. Interestingly, it was found that changes induced by tetanic stimuli were pathway dependant. Granger's Causality Botulism likely reduces synaptic efficacy, lead ing to changes in effective connectivity. Traditionally, evaluation of structural information relating to activity observed within MEAs has been limited to cross-correlation and coherenc e in the spectral domain. These techniques can demonstrate statistical relationships between multivariate data, but lack directional information. Thus, they provide little eviden ce about functional c onnectivity. However, recent adaptation of Granger's Causality (GC), traditionally an econo mics tool, to the field of neuroscience has enabled causal influences betwee n neural time series to be examined. The underlying principle behind GC, is improved prediction of a time series by including information from a second time series if the second time series has causal influence on the first time series (Ding et al., 2006). Granger adapted this principle to linear regression models using the variance of the autoregressive prediction error to determine if prediction of the first time series was improved by the second. Similarly, the second time series can be evaluated using the first time series enabling bidirectional causality to be identified. It is important to note that this is not considered to be true causality, but rather Granger' s Causality. This method has recently been adapted by our

PAGE 66

66 laboratory at UF for use with multivariate data from microelectrode arrays. Additionally, a biologically plausible computer si mulation has been used to demonstrate that values obtained from Pairwise Granger's Causality analysis (PGC) scales linearly to synaptic weights, demonstrating that PGC provides a realistic estimate of connectivity (Cadotte et al., 2006). It was hypothesized that BoNT A would reduce the effe ctive connectivity. This would result in a network wide decrease in Granger Ca usality values after BoNT addition. To calculate Pairwise Granger's Causality (PGC), 200 ms post stimulation spike trains were low pass filtered and fitted to a bi variate autoregressive (AR) model: xt N j jtxy N j jtxx tEYjAXjAX 1 1)( )( (5-1) yt N j jtyy N j jtyx tEYjAXjAY 1 1)( )( (5-2) The coefficient matrix A, and error matrix E was then estimated using Morf's method (Morf, 1978). Following methods de tailed in Ding et al. (2006) the bivariate AR model is Fourrier transformed to the spectral domain to generate the autospectra Sxx, and Syy and the cross spectra Sxy. Gewekes's transform, applied to autospectra from the same bivariate model is used to calculate the intrinsic power term. The causal influence for X Y and Y X in the spectral domain is then defined as: )( )( ln)(int yy yy YXS S f (5-3) )( )( ln)(int xx xx XYS S f (5-4) The values obtained represent PGC in the spectral domain. To represent PGC between electrode pairs, the maximum value between 1-30 Hz is used.

PAGE 67

67 Methods Experimental protocol Tissue culture methods during these experiment s were the same as those detailed in Chapter 4. MEAs were dosed with 600 pg (n = 4) and 1 ng (n = 6) of BoNT A and compared to control cultures dosed with BoNT carrier were recorded over the same period. Twenty minutes of spontaneous neural activity, r ecorded prior to treatment with either BoNT or no BoNT, was used as a baseline activity measurement. Activi ty was then recorded immediately following application of BoNT for the BoNT A group and during a similar period for the Control group. This was followed by twenty minute spontaneous activity recordings at 12, 24, 36, 48, and 72 hours after treatment. Following each spontaneous recording session, ten channels were stimulated for the 600 pg and 1 ng cultures using a 600 mV 200 s biphasic pulse. Stimulati ons were replicated ten times and the stimulation or der was randomized. Plasticity protocol The Jimbo protocol was applied 72h after treatm ent to the four cultur es dosed with 600 pg BoNT A and four cultures dosed with BoNT A carrier (Jimbo et al., 199 9). Briefly, a test stimulus consisting of a bipolar 600 mV and 200 s pulse was a pplied at 58 of the 60 electrode sites. Each pulse was applied at a three-second interval and ten replicates were performed with three seconds between replicates. The stimulation order was randomized. After the test stimulus was applied, a tetanic stimulation pulse consis ting of 20 trains of 10 pulses (600 mV and 200 s) at 20 Hz was applied to a single electrode site. The test stimulus was reapplied following tetanic stimulation. Plasticity change s induced by the tetanic stimula tion were evalua ted by comparing the average spike rate for each test stimulis site.

PAGE 68

68 Results and Discussion Spontaneous activity Cultures dosed with 600 pg and 1 ng of BoNT A exhibited similar behavior to results characterized in Chapter 4. No significant cha nge in activity was observed up to 24h post BoNT dosage. At the 24 h period, most cortical netw orks dosed with BoNT A exhibited a slight increase in burst duration compared to baseline activity. An increase was also observed in three of the controls. Interestingly, at the 48h period two of the 600 pg BoNT A dosed cultures and three of the 1 ng dosed cultures exhibited superb urst behavior. By the 72h recording period, seven of the ten cultures dosed with 600 pg or 1 ng BoNT A exhibited superburst behavior. Three control cultures exhibited superburst behavior as well, though the size and duration of the bursts within the superbursts of c ontrol cultures were much smalle r and more regular in duration. Burst durations in networks that exhibited su perbursts were drastically reduced relative to cultures with regular bursting. Th is resulted in a decrease in burst duration for these cultures. Thus the burst duration of spontaneously reco rded bursts did not reliably increase over the recording session for either 600pg or 1 ng dosed cu ltures, and was not stable during spontaneous recording sessions. Additionally, the burst detection al gorithm did not reliably detect the burst onset and burst termination of bursts in cultures exhibiting superburst behavior. Because of this, it was determined that burst statistics obtained during spontaneous recordings may not be suitable for BoNT detection. Elicited activity Cortical cultures were also stimulated usi ng a 600 mV 200 s biphasic pulse. Stimulation channels were selected that produced a reliable response. Stimulations th at were already bursting immediately prior to a stimulation pulse were ex cluded from the analysis. On average, one to

PAGE 69

69 three stimulations were excluded from each reco rding for this reason. There was no significant change in the number of stimul ations with prior bursts betw een control cultures (P > Figure 5-1. Percent change in spik es with respect to baseline ac tivity after treatment with 600 pg BoNT A (A) and 1000 pg BoNT A (B). (* indicates p < 0.05)

PAGE 70

70 0.05). Occasionally, a stimulation did not elicit a burst. Approximately 80 percent of stimulations elicited bursts during each recording session. There were no significant differences between Control and BoNT dosed cultures. Burst sizes as sociated with stimulations were smaller in cultures dosed with 600 pg and 1 ng BoNT A and th eir respective control cultures than those reported in the previous chapter for cultures dosed with 200 pg BoNT A. This is attributed to a reduction in the signal to noise ratio due to the age of the ME As used. The number of spikes during the late phase of the burst (100 to 300 ms post stimulation) was found to increase in both cultures dosed with 600 pg and 1000 pg BoNT A, as observed in BoNT dosed cultures in Chapter 4. Figure 5-1 shows the percent change in the average number of late phase spikes for cultures dosed with 600 pg and 1000 pg. A significant increase in late phase spikes (p > 0.05), occurred 48h after dosage in both cases. Plasticity To further characterize activity changes induced by BoNT intoxication, plasticity changes associated with LTP and LTD were assessed us ing the protocol established by Jimbo et al. (1999). Results for culture 9059, shown in Figure 5-2, were consistent with all four cultures. The color map shown in Figure 5-2 indicates the diffe rence in average number of spikes between pre and post tetanus application. The x-ordinate indicates probe site and the y-ordinate indicates response site. As demonstrated in Jimbo et. al (1999), changes in the response channel varied according to the stimulation site, while activity in response to a single probe site was uniformly depressed or enhanced. Jimbo's experiment demonstrated that activity changes associated with tetanic stimulus-induced LTP and LTD were pathway dependant. Results obtained for both control and BoNT dosed cultures were consistent with Jimbo's observations. Similar to control cultures both stimulus induced LTP and LTD were present in BoNT dosed cultures. However, numbers of probe sites that were depressed, a nd magnitude of the changes in activity (both

PAGE 71

71 A B A Figure 5-2. Network plasticity induced by tetanus (A) 72 h af ter dosage with 600 pg BoNT A, and (B) 72h after dosage with BoNT A carrier solution, results shown are for dish 9059 as a representative of results obtained for 5 cultures. The co lor map indicates the difference in the average number of spikes between pre and post tetanus application. The x axis indicates the probe site, and th e y axis indicates re sponse site. The bar above the color map indicates changes in res ponse channel 40 for each probe site. As demonstrated in Jimbo et. al. (1999), ch anges in the response channel varied according to the stimulation site. The bar to the right of the color map indicates changes in spike rate for each response channel for probe channel number 47 (red) and 9 (blue). Consistent wi th results obtained by Jimbo, for a given probe site response channels were either all depressed (LTD) or potentiated (LTP) indicating that plasticity induced changes are site spec ific. Interestingly, plasticity changes in BoNT cultures were much larger th an responses in control cultures.

PAGE 72

72 depression and potentiation) increased. This may indicate that the mechanism by which LTP and LTD are produced is altered by BoNT A. Alternativ ely, this may simply be the result of effective connectivity changes in the culture. PGC Analysis To determine changes in netw ork connectivity PGC was applie d to stimulation data from control and BoNT dosed cultures. For the PGC analysis, after ea ch stimulation a 200 ms window of data was used. Data from stimulations in wh ich the culture was bursting prior to stimulation, and stimulations which did not el icit a burst were excluded from th e analysis. In both control and BoNT dosed cultures, both increases and decreases in PGC values were observed over the seventy-two hour period in which activity was monitored. However, in the majority of BoNT dosed cultures, a decrease in PGC values was obser ved across the majority of the network 48h to 72h after treatment. This suggests a decrease in overall network connectivity over time. BoNT Detection Algorithm A BoNT detection algorithm wa s developed based upon three cr iteria: the post stimulation spike rate, the number of local acti vity peaks in post stimulation response, and average change in connectivity. All cultures were evaluated with respect to baseline activity. Based on these criteria, cultures received a score from 0 to 3 with higher scores indicating likely BoNT intoxication. Cultures with score two or above were considered to be intoxicated with BoNT. To assess changes in the late phase spike rate, spikes after each burst were sorted into 10 ms bins with an eight ms overlap. Replicates of the same stimulation site were then averaged together to yield the average response for each stimulation. Bins from 100 to 300 ms post stimulation were summed together to obtain late phase stimulation res ponse. If 60 percent or more stimulation sites had a 75% or greater increase over the base line response the late phase

PAGE 73

73 Figure 5-3. GC difference plot after dosage with carrier (A), and 1000 pg BoNT A (B). Results shown are for dish 9299 as a representative of results obtained for fourteen carrier dosed cultures and fifteen BoNT A dose d cultures. The color map indicates the difference in the Granger Causality score fr om baseline activity. The x axis indicates the source channel, and the y axis indicates response channel. E ach box represents the response for a single probe channel (1-10) After dosage with BoNT A connectivity as evaluated using Granger's Causality tende d to decrease. After dosage with carrier solution connectivity in some networks also decreased although changes were less consistent between stimulation sites. A bout 25% of the time a large decrease in connectivity was observed with in 72h after dosage with BoNT carrier. In contrast, a large decrease in connectivity was observed about 75% of the time after dosage with BoNT A.

PAGE 74

74 stimulation response was scored as 1. Otherwise the late phase stimulation response was scored as 0. The number of local maxima was determined using the grand average of stinimulation responses. Results were smoothed using a first-order Butterworth filter with a cutoff frequency of 50 Hz. Local maxima were detected as a chan ge in slope from positive to negative. Each recording time was scored as one if the number of local maxima detected exceeded the number of local maxima in the baseline recording by one. Figure 5-4 shows the unfiltered and filtered post-stimulation response for control dish 7079 and 200 ng BoNT A dosed dish 9299 as an example of typical results. Changes in connectivity were evaluated using Granger Causality for each stimulation site. Results were summed across stimulation sites to yield a 60 by 60 matrix probe verses response matrix. As was noted previously, a large decrease in connectivity was observed in the majority of the cultures. Thus, if 80% of probe-response site s that exhibited an absolute change in PGC value greater than 2 were decreased (-2.0 or less), the connectivity score was 1. Otherwise connectivity changes were scored as 0.

PAGE 75

75 Figure 5-4. Average stimulati on response (blue) and smoothed stimulation response (red) for control dish 7079 (A D) and dish 9299 af ter dosage with 200 pg BoNT A (E H). As a criterion for the BoNT toxin det ection algorithm the average response to stimulation was smoothed using a firstorder Butterworth filter with a cutoff frequency of 50 Hz and the number of local maxima was detected using the change in the slope. One local maxima was detected for control dish 7079 during the Baseline, 24h, 48h, and 72h recordings so it did not m eet the criteria for BoNT detection. In contrast while one local maxi ma was detected during base line recording of dish 9272, after dosage with 200 pg BoNT A two local ma xima were detected at 24h, and three local maxima were detected for the 48h and 72h recordings.

PAGE 76

76 Results of the BoNT detection algorithm are summarized in Table 5-1. No false positive results were obtained for any of the control cultures. However, only 33% of BoNT A dosed cultures were detected as positive 24h after dosage, while at 72h, all of the BoNT dosed cultures were identified. This demonstrates that stim ulation induced bursts, are suitable for BoNT detection. Bursts induced through stimulation are stable over time and require shorter recording periods to obtain sufficient data for BoNT detect ion. Furthermore, variab ility due to poor burst detection, and variability in activity over time are drastically reduced using elicited activity. Table 5-1. Identification of BoNT in unknown cultures Dose Time Post Stim Number of positive cultures Number of negative cultures Number of False Positives Number of False Negatives 24h 2/5 4/4 0/4 3/5 200 pg 48h 4/5 4/4 0/4 1/5 72h 5/5 4/4 0/4 0/5 24h 3/4 4/4 0/4 1/4 600 pg 48h 3/4 4/4 0/4 1/4 72h 4/4 4/4 0/4 0/4 24h 1/6 6/6 0/6 5/6 1000 pg 48h 6/6 6/6 0/6 6/6 72h 6/6 6/6 0/6 6/6 24h 5/15 14/14 0/14 9/15 Overall 48h 13/15 14/14 0/14 2/15 72h 15/15 13/14 0/14 0/15

PAGE 77

77 CHAPTER 6 CONCLUSION AND FUTURE WORK Future Work Reduction of detection time Currently, this method of toxin detection requires 24 to 72h. Wh ile this is longer than rapid detection methods such as ELISA and protease activity assays, it is comparable to the mouse bioassay. Additionally, many rapid methods rely on enrichment phases that drastically increase time to detection. Thus, despite the time to detection, this method may still be suitable as a replacement to the mouse bioassay. However, re ducing time required for BoNT detection would drastically improve use of cortical networks as a practical biosensor. Slow toxin uptake is a likely explanation fo r the time required before changes in activity are detected. Recent studies have indicated that botulism toxi n uptake is coupled to vesicle recycling (Dong et al., 2006; Verderio et al., 200 6). Therefore it may be possible to increase the rate of vesicle recycling, increasing uptake and he nce, decrease the detection time for this toxin. For example, Keller et al., (2004) used elevated potassium media for studies of BoNT uptake in spinal cord cultures. In this case, optimal BoNT A uptake was achieved after 4 minutes of toxin exposure in elevated potassium media with a significant level of SNAP-25 cleavage demonstrated at 2.5 hours after toxin exposure. While less than twenty percent of SNAP-25 was cleaved in media containing 3 mM KCl, more than fifty percent was cleaved in media containing 50 to 80 mM KCl. However, concentrations of BoNT A used in the current study were two to four orders of magnitude smaller than concentrations used by Keller et al., (2004). By increasing potassium in the media during toxin dosing and subsequently increasing native activity in our cultures, time for toxin detection may be drastica lly reduced. Alternatively, repeated stimulation may also increase rates of vesicle recycling. Fu rthermore, repeated measures provided by more

PAGE 78

78 stimulations would provide a more stable and hi gher resolution measure than provided in this study. This may have the added advantage of targ eting BoNT to active synapses of a specific pathway. However, changes induced by repeated stimulation would ha ve to be carefully distinguished from toxin activity. Finally, inhibition of gamma-aminobutyric acid (GABA)ergic synapses using a treatment such a biccuculine may be used to increase netw ork activity. Since the ef fect of biccuculine is reversible, temorary dosage of the network with biccuculine may be suffici ent to elevate network activity, enabling more rapid BoNT uptake. Serotype differentiation This work was limited to characterization of BoNT A on network activity. However, identification of all sero types is ideal for a BoNT biosensor. It may be possible to distinguish BoNT serotypes based solely upon network activity Alternatively, a neut ralization assay may be used to determine toxin serotypes. Similar to the mouse bioassay, samples would be positive for BoNT A serotype when the base-line recording is altered for all samples except those containing antiserum for the correct serotype. Use of neural networks with crude samples This work was limited to use of pharmaceutical preparations of botulinum type A toxins. However, for neural networks to be used as practical biosensors they must be applicable for use with crude samples, such as detection of toxin in food. Specific challeng es to the use of MEA cultures with crude samples include sensitivity of the cultured neurons to changes in pH and ion concentrations, sensitivit y of cultures to contamination by fo od materials, and sensitivity of cultures to infection. Thus careful preparation of samples is necessary to address these issues. Samples will likely need to be centrifuged and filtered to minimize contamination and infection of neural cultures.

PAGE 79

79 Refinement of MEAs Further refinement of the hardware and protocols used will drastically improve the use of MEAs as biosensors. As computer processing sp eed continues to improve it may be possible to develop arrays of microelectrode arrays, in which sister cult ures could be recorded from simultaneously. This could enable simultaneous analysis of samples a nd controls, along with automated toxin detection using algorithms similar to the one used in this work. Alternatively, limiting intoxication to only a portion of the cortic al network may enable a single culture to serve as both a test subject and control. This may be achieved through mi cro-fluidic steering. Additionally, changes to MEAs which reduce cost and increase durability are desirable prior to commercial application as a biosensor. Alternative and immortal cell lines Detection of BoNT toxin and reduction of detection time may be achieved using an alternative cell line. Many cell cu ltures are suitable for MEAs including cardiac myocytes, spinal neurons, and retinal ganglion cells. It may be that one of these a lternatives is more suitable for BoNT detection. Additionally, an adult tissue slice may be used as an alternative to dissociated tissue. Finally, as an alternative to animal derived tissue cultures it may be possible to use an immortal cell-line, completely eliminating animal use. This may also reduce variability between samples. Furthermore, neurons may be altered to express a large number of toxin receptor sites, improving toxin uptake. Conclusion Use of cortical networks as a biosensor for botulinum toxin en ables detection of only intact physiologically active toxin that is capable of binding, inte rnalization and substrate cleavage. Thus, similar to the mouse bioassay this assa y may be suitable for evaluation of biological activity of the toxin. This is advantageous to alternative assays such as ELISA and PCR since

PAGE 80

80 they are incapable of assessing toxin activity, and may even be positive in the absence of active toxin. This work demonstrates that cortical tissue grown on MEAs provides a suitable substrate for a cell-based BoNT A biosensor. While this study is currently limited to explor ing the effect of BoNT A on network activity this technique could also be useful for detecti ng and perhaps even differe ntiating the effects of other BoNT strains (types B-G) as well as dete ction of a plethora of other neurotoxins. Furthermore, while the mouse bioassay requires se veral animals for a single assay, with this technique a single animal could provide tissue for multiple ne ural network based assays on MEAs. Since the cortical tissue used in this work is commercially available no onsite mouse rearing facilities are necessary. This work has established the time course and a suitable concentration range for BoNT detection using neural networks. It remains conceivable that an assay based on this technique could provide a powerful method for detecting a wide variety of food borne agents including both known and currently unknow n biological threats.

PAGE 81

81 APPENDIX A SPONTANEOUS ACTIVITY IN BONT A AND CARRIER DOSED CULTURES Table A-1. Average spike height for neural cultures dosed with BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 (mV) BoNT A Mean N=9 (mV) Control Standard Deviation (mV) BoNT A Standard Deviation (mV) Base -3.486 -3.964 0.125 0.105 12h -3.388 -3.781 0.113 0.097 24h -3.059 -3.991 0.119 0.096 36h -3.316 -3.864 0.124 0.107 48h -3.566 -3.726 0.107 0.092 72h -3.737 -3.358 0.095 0.099 1wk -3.937 -2.992 0.129 0.130 Table A-2. Average spike width for neural cultures dosed with BoNT carrier or 200 pg BoNT A Time Post Treatment Control Mean N=5 (ms) BoNT A Mean N=9 (ms) Control Standard Deviation (ms) BoNT A Standard Deviation (ms) Base 0.323 0.348 0.125 0.105 12h 0.315 0.333 0.113 0.097 24h 0.286 0.318 0.119 0.096 36h 0.284 0.311 0.124 0.107 48h 0.264 0.308 0.107 0.092 72h 0.302 0.299 0.095 0.099 1wk 0.332 0.323 0.129 0.130 Table A-3. Average inter spike interval for neural cultures dosed w ith BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 (ms) BoNT A Mean N=9 (ms) Control Standard Deviation (ms) BoNT A Standard Deviation (ms) Base 16.926 21.762 25.243 31.459 12h 22.401 21.286 25.700 30.275 24h 27.534 29.675 37.940 41.908 36h 26.366 20.703 22.438 27.572 48h 31.200 29.191 30.831 45.076 72h 40.353 27.376 35.553 42.2680 1wk 29.584 29.518 34.879 51.924

PAGE 82

Table A-4. Average burst duration for neural cultures dosed with BoNT carrier or 20pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 0.204 0.231 0.080 0.108 12h 0.204 0.256 0.082 0.111 24h 0.216 0.309 0.090 0.130 36h 0.225 0.311 0.096 0.126 48h 0.253 0.368 0.136 0.138 72h 0.249 0.436 0.121 0.197 1wk 0.172 0.394 0.073 0.204 Table A-5. Average inter-burst in terval for neural cultures dosed with BoNT carrier or 200 pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 15.14 11.52 12.13 5.37 12h 17.52 13.61 11.20 8.14 24h 19.21 17.59 7.26 6.69 36h 19.50 16.06 9.36 8.94 48h 21.37 15.29 6.99 4.41 72h 22.53 17.30 8.27 8.65 1wk 13.50 16.62 7.12 5.04 Table A-6. Average number of spikes per burst for neural cultures dosed with BoNT carrier or 200 pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 1986 1713 2546 1628 12h 2012 1761 2455 1685 24h 2012 2226 2442 1907 36h 2102 2499 2597 2139 48h 2066 2793 2673 2428 72h 1985 3380 2734 2929 1wk 1390 3329 2061 2959 Table A-7. Average spike height for neural cultures dosed with BoNT carrier or 600pg BoNT A Time Post Treatment Control Mean N=4 (mV) BoNT A Mean N=4 (mV) Control Standard Deviation BoNT A Standard Deviation Base -2.9183 -2.7380 1.1164 0.4644 12h -2.7094 -2.4200 0.9534 0.3139 24h -2.7940 -1.4691 0.7612 2.7080 36h -2.3781 -1.4858 0.7816 2.4462 48h -2.5932 -2.5963 0.9513 0.6585 72h -2.9813 -2.8765 1.1365 0.4444 82

PAGE 83

Table A-8. Average spike width for neural cultures dosed with BoNT carrier or 600pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation BoNT A Standard Deviation Base 0.1622 0.1725 0.0195 0.0051 12h 0.1508 0.1749 0.0144 0.0094 24h 0.1554 0.1430 0.0105 0.0074 36h 0.1711 0.1702 0.0177 0.0233 48h 0.1743 0.1528 0.0143 0.0185 72h 0.1820 0.1386 0.0095 0.0167 Table A-9. Average inter spike interval for neural cultures dosed w ith BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation BoNT A Standard Deviation Base 0.0137 0.0095 0.0049 0.0005 12h 0.0150 0.0121 0.0039 0.0036 24h 0.0154 0.0088 0.0050 0.0034 36h 0.0102 0.0084 0.0028 0.0033 48h 0.0131 0.0131 0.0027 0.0053 72h 0.0087 0.0192 0.0027 0.0069 Table A-10. Average burst durati on for neural cultures dosed w ith BoNT carrier or 600pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation BoNT A Standard Deviation Base 124.2260 118.3149 21.2908 21.7382 12h 129.1907 130.8544 25.6628 36.8718 24h 126.0828 137.1658 26.4672 49.5016 36h 126.2304 139.1466 23.9707 41.6674 48h 133.4856 147.6418 15.9434 26.3732 72h 122.2815 53.3322 13.3289 5.1737 83

PAGE 84

Table A-11. Average inter-burst in terval for neural cultures dosed with BoNT carrier or 600 pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation BoNT A Standard Deviation Base 5.066 3.911 1.487 0.470 12h 5.697 4.821 1.715 2.166 24h 5.909 2.307 0.798 0.330 36h 4.434 2.006 1.329 0.493 48h 5.749 2.492 1.303 0.686 72h 3.464 3.080 0.586 0.734 Table A-12. Average number of sp ikes per burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A Time Post Treatment Control Mean N=4 (Sk/Bst) BoNT A Mean N=4 (Sk/Bst) Control Standard Deviation BoNT A Standard Deviation Base 278.5 308.4 114.7 58.0 12h 288.9 297.1 116.2 87.6 24h 285.1 235.3 91.7 155.7 36h 349.5 245.9 131.6 207.0 48h 354.4 148.7 91.3 48.1 72h 329.9 72.0 137.0 21.3 Table A-13. Average number of channels per burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A Time Post Treatment Control Mean N=4 (Ch/Bst) BoNT A Mean N=4 (Ch/Bst) Control Standard Deviation BoNT A Standard Deviation Base 37.1935 39.8215 7.4324 4.9276 12h 37.9359 38.3629 6.2427 6.7225 24h 38.1660 36.6311 7.1915 7.9138 36h 40.2464 36.4478 5.0727 9.0970 48h 41.2229 33.2713 4.4958 3.5410 72h 39.6653 23.5968 3.6450 2.2500 84

PAGE 85

Table A-14. Average burst durati on for neural cultures dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation (ms) BoNT A Standard Deviation (ms) Base 112.5372 150.2687 24.4398 46.1023 12h 124.3151 165.4070 25.7942 51.9701 24h 124.0459 178.6280 34.2769 67.6356 36h 132.7705 202.2882 43.6598 103.6945 48h 127.0323 169.7188 36.1385 66.0071 72h 124.7127 203.5174 16.2268 30.4046 Table A-15. Average inter-burst in terval for neural cultures dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=4 (ms) BoNT A Mean N=4 (ms) Control Standard Deviation (ms) BoNT A Standard Deviation (ms) Base 5.5481 6.4824 3.2419 2.9963 12h 5.2274 7.2141 3.4010 3.3540 24h 5.1237 4.7685 3.4082 2.8610 36h 5.4598 5.3077 4.4351 4.1595 48h 4.0069 3.9431 3.2410 1.4940 72h 5.3891 3.9288 3.8194 2.7445 Table A-16. Average number of spikes per burst for neural cultures dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=4 (Sk/Bst) BoNT A Mean N=4 (Sk/Bst) Control Standard Deviation (Sk/Bst) BoNT A Standard Deviation (Sk/Bst) Base 341.9895 491.2265 119.2052 308.0769 12h 342.2862 400.8340 122.9619 262.8882 24h 304.1719 428.1824 103.9649 254.7530 36h 301.6341 535.5256 141.3596 434.1746 48h 246.6684 331.2163 142.3897 178.8936 72h 284.6426 310.8945 93.1311 124.6812 85

PAGE 86

Table A-17. Average number of channels per burst for neural cultures dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=4 (Ch/Bst) BoNT A Mean N=4 (Ch/Bst) Control Standard Deviation (Ch/Bst) BoNT A Standard Deviation (Ch/Bst) Base 29.6113 32.5754 10.7625 14.8220 12h 29.3288 31.3244 10.3198 14.0974 24h 28.6892 31.5278 11.0753 13.0683 36h 28.1165 32.7809 9.7323 13.8754 48h 26.1800 32.3135 10.1581 13.8349 72h 27.8608 31.4127 9.4488 12.1135 86

PAGE 87

APPENDIX B ELLICITED ACTIVITY IN BONT A AND CARRIER DOSED CULTURES Table B-1. Average spikes per stimulation for ne ural cultures dosed with BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 1842.0 1743.7 2522.9 1631.1 12h 1846.9 1768.5 2426.9 1617.7 24h 1864.5 1951.5 2439.3 1747.9 36h 1885.2 2201.7 2472.0 1924.8 48h 1971.9 2542.8 2638.2 2355.4 72h 1893.3 2798.1 2580.3 2602.1 Table B-2. Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 73.3 77.6 18.8 13.7 12h 79.5 83.0 12.6 9.3 24h 82.0 84.1 7.7 22.3 36h 83.9 84.6 4.9 14.9 48h 81.0 87.7 10.1 4.9 72h 79.2 87.5 12.8 13.9 Table B-3. Percent change in the number of sp ikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed wi th BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation 12h 1.8 5.2 9.6 12.9 24h 5.9 16.3 15.7 37.5 36h 5.6 25.6 20.2 39.9 48h 6.5 22.3 38.2 47.7 72h -2.8 24.6 26.3 51.9 87

PAGE 88

88 Table B-4. Percent change in the number of spik es 150 to 300 ms post stimulation from baseline activity for neural cultures dosed wi th BoNT carrier or 200pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation 12h 36.0 58.7 71.8 90.0 24h -22.6 72.4 33.5 116.0 36h -18.0 117.7 34.5 114.5 48h -1.4 281.2 41.5 261.3 72h 26.8 671.9 73.3 765.8 Table B-5. Average spikes per stimulation for ne ural cultures dosed with BoNT carrier or 600 pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 157.7 272.0 85.0 79.7 12h 179.5 309.1 107.7 37.3 24h 213.5 260.9 88.9 53.3 36h 204.8 216.3 106.6 63.6 48h 202.8 244.5 46.6 39.9 72h 223.3 354.8 77.7 220.8 Table B-6. Percent of stimulations with a burst for neural cultures dosed with BoNT carrier or 600 pg BoNT A Time Post Treatment Control Mean N=5 BoNT A Mean N=9 Control Standard Deviation BoNT A Standard Deviation Base 80.8 78.3 4.1 2.1 12h 83.8 81.5 3.8 1.7 24h 81.8 85.4 2.8 4.4 36h 80.3 86.0 4.5 3.5 48h 82.3 85.4 4.5 0.8 72h 79.9 85.8 3.0 2.9

PAGE 89

Table B-7. Percent change in the number of spikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed wi th BoNT carrier or 600pg BoNT A Time Post Treatment Control Mean N=4 BoNT A Mean N=4 Control Standard Deviation BoNT A Standard Deviation 12h 13.9 18.8 15.2 30.4 24h 46.6 -4.6 36.5 33.9 36h 36.5 -25.3 32.5 42.6 48h 48.6 -13.8 56.1 23.8 72h 55.7 1.5 39.0 29.4 Table B-8. Percent change in the number of spik es 150 to 300 ms post stimulation from baseline activity for neural cultures dosed wi th BoNT carrier or 600pg BoNT A Time Post Treatment Control Mean N=4 BoNT A Mean N=4 Control Standard Deviation BoNT A Standard Deviation 12h 118.9 176.7 110.6 116.9 24h 90.6 653.1 63.0 674.3 36h 59.8 1107.2 28.8 1312.1 48h 96.1 1059.3 47.2 471.5 72h 85.7 1923.3 52.6 1108.1 Table B-9. Average spikes per stimulation for ne ural cultures dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=6 BoNT A Mean N=6 Control Standard Deviation BoNT A Standard Deviation Base 374.2 324.6 235.8 188.2 12h 396.7 390.6 248.8 225.7 24h 413.7 447.8 302.0 368.0 36h 410.7 329.2 257.8 182.1 48h 360.8 303.5 221.0 187.8 72h 342.1 253.0 204.7 177.7 89

PAGE 90

Table B-10. Percent of stimulations with a burst for neural culture s dosed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=6 BoNT A Mean N=6 Control Standard Deviation BoNT A Standard Deviation Base 86.1 87.6 2.7 3.5 12h 86.1 86.2 3.0 3.0 24h 83.2 90.0 5.5 1.5 36h 85.3 89.5 6.0 3.0 48h 82.2 88.9 5.7 2.5 72h 83.6 91.5 5.1 3.7 Table B-11. Percent change in the number of spikes 0 to 150 ms post stimulation from baseline activity for neural cultures dosed wi th BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=6 BoNT A Mean N=6 Control Standard Deviation BoNT A Standard Deviation 12h 7.0 20.2 40.2 7.6 24h 10.2 39.1 104.4 5.5 36h 12.2 1.8 31.0 7.6 48h 0.5 -9.0 38.9 10.4 72h -1.8 4.3 78.5 5.0 Table B-12. Percent change in the number of spikes 100 to 300 ms post stimulation from baseline activity for neural cultures do sed with BoNT carrier or 1000 pg BoNT A Time Post Treatment Control Mean N=6 BoNT A Mean N=6 Control Standard Deviation BoNT A Standard Deviation 12h 19.4 17.5 20.6 29.4 24h 32.3 231.8 423.2 47.3 36h 29.1 196.1 192.4 18.9 48h 22.1 284.8 255.5 87.9 72h 36.6 547.2 455.8 211.1 90

PAGE 91

APPENDIX C BURST DETECTION ALGORITHM The following algorithm, Bst_Det is a burst detect ion algorithm written fo r use with Matlab and requires data in the form of a spike file from MEABench. The program returns burst statistics in the variable data_storage, and burst times in a 3 x N matrix, in whch the first column is the burst start, the second column is th e burst stop, and the third co lumn is the burst duration. function [data_storage,bst] = Bst_Det(data1) % THIS PROGRAM SORTS ALL THE SPIKES INTO A HISTOGRAM WITH BIN LENGTHS % SPECIFIED BY BIN_S. A THRESHOLD IS SET BASED UPON THE MEAN SPIKES PER % BIN. A BURST IS DETECTED EACH TIME THE SPIKES IN A BIN EXCEED THE % THRESHOLD. AFTER A BURST IS DETECTED AT LEAST 0.15 SECONDs MUST ELAPSE BEFORE % A NEW BURST CAN BE DETECTED.THE BURST END IS DETECTED AS THE FIRST 0.5 MS % BIN AFTER A BURST WITH NO SPIKES IN IT. %SET THE INTERVAL OVER WHICH YOU ARE INTERESTED IN CALCULATING %PARAMETERS IN SECONDS strt = 1; %start 10 min into the recording stp = 1201; %stop 20 min into the recording tic bin_s = 0.0005; P = find(data1.time > strt & data1.time < stp); data.time = data1.time(P); data.channel = data1.channel(P); N = strt-bin_s:bin_s:stp-bin_s; y = hist(data.time, N); R = mean(y)+ std(y) if R < 0.4 thresh = 1 elseif R > .4 && R < 1 thresh = 2 else thresh = 3 end %Bin_size = 0.01; Bin_size = 0.01; C = find(y > thresh); D = size(C,2); Burst = 0; p = 1; bst(1,1) = 0; for i = 1:D-1 %Detects bursts as consecutive bins 91

PAGE 92

92 if C(1,i+1) C(1,i) == 1 && Burst == 0 %exceeding the threshold bst(1,p) = N(1,C(1,i)); p = p + 1; Burst = 1; elseif Burst == 1 && C(1,i+1) C(1,i) > 2 if N(1,C(1,i))-bst(1,p-1) > 0.20 %sets the min time between bursts as 200 ms Burst = 0; end end end %Detect the end of the burst i = 1; Cntr = size(bst, 2); while i < Cntr +1 Bin_size2(i) = Bin_size; Sk_post = find(data.time > bst(1,i) & data.time < bst(1,i) + 5); ch_ = 0; while ch_ == 0 r = hist(data.time(Sk_post), 5/Bin_size2(i)); D = find(r == 0); if numel(D) == 0 Bin_size2(i) = Bin_size2(i) .005; elseif Bin_size2(i) <= 0 ch_ = 1; else ch_ = 1; end end if Bin_size2(i) <= 0 bst(2,i) = bst(1,i) + 5; else bst(2,i) = bst(1,i) + D(1)*Bin_size2(i); end i = i +1; end %Check that the same burst has not been detected more than once Error_check = bst(1,2:Cntr) bst(2,1:Cntr-1); C = find(Error_check <= 0); if numel(C) > 0 for i = 1:size(C,2)+1 if i == 1 newbst = bst(:,1:C(i)); elseif i > 1 && i < size(C,2) + 1 newbst = [newbst(:,:) bst(:,C(i-1)+2:C(i))]; elseif i == size(C,2) + 1 newbst = [newbst(:,:) bst(:, C(i-1)+2:Cntr)]; end

PAGE 93

end bst = newbst; end %Quick plot to check BD alogrithm i = 1; N = (stp-strt)*1000; new_time = linspace(strt, stp, 2*N); BUR = zeros(size(new_time)); while i < size(bst(1,:), 2)+1 Row_Col = find(new_time > bst(1,i) & new_time < bst(2,i)); BUR(Row_Col) = 60; i = i+1; end %PRINT THE FIGURES IN 100 SECOND INTERVALS I=1;%round(strt/100)+1; while I < 13 % round(stp/100)+1 figure() plot(data.time, data.channel,'.','markersize',1),axis([1+((I-1)*100) 100+((I-1)*100) 0 64]); hold on plot(new_time, BUR,'r-'); I = I +1; end %Calc the bst duration bst(3,:) = bst(2,:)-bst(1,:); D = find(bst(1,:) > strt & bst(1,:) < stp); %CALCULATE THE BURST DURATION dur_in_int = bst(3,D);% results(1,1); data_storage(1,1)= mean(dur_in_int); data_storage(2,1) = std(dur_in_int); data_storage(3,1) = ((std(dur_in_int))^2)/sqrt(size(D,2)); %SPIKE RATE SPKS = find(data.time>strt & data.timebst(1,D(i))&data.time< bst(2,D(i))); Sk_per_bst(i) = size(Spike_bsts,1); for k = 1:60 B = find(data.channel(Spike_bsts) == k-1); 93

PAGE 94

if numel(B > 0) Chan_Cnt(i,1) =Chan_Cnt(i,1)+1; end end end data_storage(6,1) = mean(Sk_per_bst);%-results(5,1); data_storage(7,1) = mean(Chan_Cnt(:,1));%-results(6,1); data_storage(8,1) = std(Chan_Cnt(:,1));%-results(6,1)); %PERCENT SPIKES IN BURSTS data_storage(9,1) = sum(Sk_per_bst); data_storage(10,1) = 100*data_storage(9,1)/TOT_SPKS;% results(9,1); %MEAN BURST RATE (PER SECOND) data_storage(11,1) = M/(stp-strt);%-results(10,1); data_storage(12,1) = M/((stp-strt)/60);%-results(11,1); %Burst Period START TO START P = size(D,2); IBI_st_st = (bst(1,D(1,2:P))-bst(1,D(1,1:P-1)));%-results(12,1); data_storage(13,1) = mean(IBI_st_st); data_storage(14,1) = std(IBI_st_st); %INTERBURST INTERVAL STOP TO START IBI_sp_st = (bst(1,D(1,2:P))-bst(2,D(1,1:P-1)));%-results(14,1); data_storage(15,1) = mean(IBI_sp_st); data_storage(16,1) = std(IBI_sp_st); data_storage(17,1) = std(Sk_per_bst); data_storage(18,1) = (std(Sk_per_bst))^2/sqrt(size(D,2)); data_storage(19,1) = thresh; data_storage(20,1) = mean(Bin_size2,2); 94

PAGE 95

APPENDIX D BONT A DETECTION ALGORITHM The following algorithm, Bot_Check_Stim is an algorithm written for use with Matlab and requires data in the form of a spike file from MEABench. This program retu rns the variable tally, a 3 by 1 matrix which contains a score from 1 to 3 indicating the likely hood of BoNT intoxication. A score greater than 2 is considered positive for BoNT intoxication. Bot_Check is the main program. Functions written by me are list ed below Bot_check_stim in the order they are called function tally = Bot_check_stim(fn1, fn2, fn3, fn4) [Base_stimdat, Base_count, Base_avg, Base_results] = Score_Stim(fn1); [B24h_stimdat, B24h_count, B24h_avg, B24h_results] = Score_Stim(fn2); [B48h_stimdat, B48h_count, B48h_avg, B48h_results] = Score_Stim(fn3); [B72h_stimdat, B72h_count, B72h_avg, B72h_results] = Score_Stim(fn4); tally(1:3,1) = 0; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Criteria #1, it was observed that in a large number of BoNT dosed cultures% % there was a drastic decrease in connectivity associated with BoNT % % intoxication. This criteria check whether the majority of channels had a% 95% decrease in connectivity as determined using Granger's Causality. % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% fn1 = strcat(fn1,'_2'); fn2 = strcat(fn2,'_2'); fn3 = strcat(fn3,'_2'); fn4 = strcat(fn4,'_2'); score1 = Eval_connect(fn1,fn2,fn3,fn4); for i = 1:3 if score1(i,1) > 0.80 tally(i,1) = tally(i,1)+1; end end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Criteria #2, Check if there is an increase in activity observed in the % % late phase of the burst. If there a 100% or greater increase in the late% % phase for the majority of the stim sites, the tally is increased by one.% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% check_late_spikes(1,:) = B24h_results Base_results; check_late_spikes(2,:) = B48h_results Base_results; check_late_spikes(3,:) = B72h_results Base_results; K = 0.75* mean(Base_results(1,:)); for j = 1:3 row_col = find(check_late_spikes(j,:) > K ); score2(j,1) = size(row_col,2)/size(check_late_spikes,2); if score2(j,1) > 0.6

PAGE 96

96 tally(j,1) = tally(j,1)+1; end end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Criteria #3, Multiple peaks of activity the grand average of all the % % is used to determine if there are multiple local maxima present in the % % response to stimulation. If there is an increase in local maxima % % relative to baseline activity, the tally is increased. % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Grand_avg(1,1:150) = mean(Base_avg,1); Grand_avg(2,1:150) = mean(B24h_avg,1); Grand_avg(3,1:150) = mean(B48h_avg,1); Grand_avg(4,1:150) = mean(B72h_avg,1); for i = 1:4 [b,a] = butter(1,0.05); Smooth_avg = filter(b,a,Grand_avg(i,1:100)); % figure() plot(Grand_avg(i,:),'b-'); hold on plot(Smooth_avg(1,:),'r-'); Inflect_detect = Smooth_avg(:, 2:75) Smooth_avg(:, 1: 74); row_col = find(Inflect_detect > 0); P = size(row_col,2); find_local_max = row_col(1,2:P) row_col(1,1:P-1); row_col2 = find(find_local_max > 3); if numel(row_col2) > 0 max_count(i,1)= size(row_col2,2)+1; else max_count(i,1) = 1; end end for i = 1:3 score3(i,1) = max_count(i+1,1) max_count(1,1); if score3(i,1) > 0 tally(i,1) = tally(i,1) +1; end end

PAGE 97

The following algorithm, Score_Stim is an algorit hm written for use with Matlab. This program loads data in the form of spike files written by MEA bench. It determines the stimulation times and channels from a data file recorded at th e time the stimulation was performed (stim_dat), finds the number of stimulations eliciting a burst and returns a count for each stimulation site (count), determines the average stimulation response for each stimulatio n site (avg_stim), and determines the total late phase response (100 to 300 ms post simulation) fo r each stimulation site function [stim_dat, count, avg_stim, results] = Score_Stim(fn) tic fn_spike = strcat(fn,'.spike'); %LOADS DATA FROM SPIKE FILE data = PCloadspike_noc(fn_spike, 3, 25); fn_data = strcat(fn,'.data'); %DETERMINES THE STIM TIMES AND SORTS SPIKES INTO 200 MS bins [stim_time, Final_stim_time] = Stimbstdet(data); %SORT DATA BY STIM SITE, RETURN A 100 BY 3 MATRIX WITH COL 1 BEING THE STIM %REPLICATE # (1-10), COL 2 BEING THE STIM SITE #(1-60), AND COL 3 BEING THE %STIM TIME [stim_dat] = Stimdata(data, fn_data); % DETERMINES THE TIMES THE CULTURe WAS ALREADY BURSTING PRIOR TO THE BURST ONSET % AND THE TIMES WHEN THE CULTURE DID NOT BURST AFTER THE STIMULATION, STIM_DAT % IS ADJUSTED SO THAT THESE TIMES ARE SET TO ZERO, COUNT SHOWS THE TOTAL % NUMBER OF STIMS ELLICITING A BURST FOR EACH STIM SITE, AVG_STIM IS THE % AVERAGE STIM RESPONSE UP TO 300 MS AFTER THE BURST, WHERE A 10 ms SLIDING % WINDOW IS USED [stim_dat, count, avg_stim] = stim_response(data, Final_stim_time, stim_dat); %AVERAGE STIM RESPONSE 100 to 300 MS AFTER THE STIMULATION, (LATE PHASE %STIMULATION RESPONSE) N = size(avg_stim,1); for i = 1:N results(i,1) = sum(avg_stim(i, 50:150)); end 97

PAGE 98

98 The following algorithm, Stimbstdet is an algor ithm written for use with Matlab. This program useses data in the form of sp ike files written by MEA bench. function [stim_time, Final_stim_time] = Stimbstdet(data1) % THIS PROGRAM SORTS ALL THE SPIKES INTO A HISTOGRAM WITH BIN LENGTHS % SPECIFIED BY BIN_S. A THRESHOLD IS SET BASED UPON THE MEAN SPIKES PER % BIN. A BURST IS DETECTED EACH TIME THE SPIKES IN A BIN EXCEED THE % THRESHOLD. AFTER A BURST IS DETECTED AT LEAST 0.15 SECONDs MUST ELAPSE BEFORE % A NEW BURST CAN BE DETECTED.THE BURST END IS DETECTED AS THE FIRST 0.5 MS % BIN AFTER A BURST WITH NO SPIKES IN IT. row_col = find(data1.channel == 59); M = size(row_col); strt = data1.time(row_col(1,1))-5; stp = data1.time(row_col(M(1,1),M(1,2)))+5; row_col = find(data1.time > strt & data1.time < stp); data.time = data1.time(row_col); data.channel = data1.channel(row_col); tic bin_s = 0.0005; N = strt-bin_s:bin_s:stp-bin_s; y = hist(data.time, N); R = mean(y)+ std(y) if R <= 1 thresh = 1 elseif R > 1 && R <= 1.5 thresh = 2 else thresh = 3 end %Find the number of stims in the recording by detecting spikes on %channel 60 [Row_Col] = find(data.channel == 60); Num_ch63_sks = size(Row_Col,2); stim_time(1,1) = data.time(Row_Col(1,1)); stim_time(1,2) = 61; p=2; for i = 2:Num_ch63_sks if data.time(Row_Col(1,i)) data.time(Row_Col(1,i-1)) > .3 %store all spikes for stim_time(p,1) = data.time(Row_Col(1,i)); %up to 300ms after the stim_time(p,2) = 61; p=p+1; %stim end end [Row_Col] = find(data.channel == 14); %Channel 14 is the ground channel, false spikes time_ch14 = data.time(Row_Col); %Correspond to spikes on this channel

PAGE 99

num_stim = size(stim_time,1); Stim_Sk_cnt = zeros(1,num_stim); for i = 1:num_stim %Remove all spikes for up to 10 ms after a false spike [Row_Col] = find(data.time > stim_time(i,1)+.005 & data.time < stim_time(i,1)+.5); stim_dat = data.time(Row_Col); [Row_Col] = find(time_ch14 > stim_time(i,1)+.005 & time_ch14 < stim_time(i,1)+.5); if numel(Row_Col) > 0 disturbance = time_ch14(Row_Col); stt = min(disturbance); stp = stt +.010; [Row_Col] = find(stim_dat < stt | stim_dat > stp); %Find spikes %corresponding stim_dat2 = stim_dat(Row_Col); %the to thee stim %excluding those else %10 ms after a %spike on chan 14 stim_dat2 = stim_dat; end Stim_Sk_cnt(i) = size(stim_dat2,2); N=ceil(max((stim_dat2-stim_time(i,1))/0.001)); Stim_bin(i,1:N+2) = hist((stim_dat2-stim_time(i,1))/.001, 0:N+1); if N+2 <300 Stim_bin(i,N+2:300) = 0; end end %Check 30ms before each stim to ensure that the culture is not already %bursting p=1; prestim(p,1) = 0; for i = 1:num_stim [Row_Col] = find(data.time > stim_time(i,1) .200 & data.time < stim_time(i,1)-.010); num_spikes(1,i) = size(Row_Col,2); end %C = num_spikes; for i = 1:num_stim if num_spikes(1,i) > 20*thresh prestim(p,1) = i; p=p+1; end end p=1; Thresh = mean(Stim_Sk_cnt)-std(Stim_Sk_cnt) Stim_Sk_cnt %Throw out stims that did not burst for i = 1:size(Stim_Sk_cnt,2) if Stim_Sk_cnt(i) > Thresh 99

PAGE 100

Final_stim_bin(p, 1:200) = Stim_bin(i,1:200); Final_stim_time(p,1) = stim_time(i,1); p=p+1; if prestim(1) > 0; for j = 1:size(prestim,1) if prestim(j,1) == i C = Final_stim_bin(1:p-1,1:200); D = Final_stim_time(1:p-1,1); clear Final_stim_bin Final_stim_time; Final_stim_bin = C; Final_stim_time = D; clear C end end end end end 100

PAGE 101

The following algorithm, Stim_Response is an algorithm written for use with Matlab. This program useses data in the form of spike files written by MEA bench. function [stim_dat, count, stim_avg] = Stim_Response(data, Final_stim_time, stim_dat) %Now spikes for each stim are sorted according to channel. In this case %there are 300 ms of data, sorted into s histogram using a 10 ms moving %window, which shifts 2 ms R = size(stim_dat,1); N = 0; Finaltime = 200 0.5; G = -.5:1:Finaltime; Win = 200; S=0; grangerdata = zeros(60,Win,11,10); C = floor(Final_stim_time); D = floor(stim_dat(:,3)); for T = 1:R row_col = find(C == D(T,1)); if numel(row_col) == 0 stim_dat(T,3) = 0; end end m = 1; for T = 1:R if stim_dat(T,1) == 1 N = N+1; if T > 1 count(m,1) = num_; m = m+1; end num_=0; end if stim_dat(T,3) > 0 num_ = num_+1; A = find(data.time >= stim_dat(T,3) & data.time <= stim_dat(T,3)+.300 ); newtime = (data.time(A)stim_dat(T,3))*1000; newchan = data.channel(A); B = find(newchan > 59); if numel(B) > 0 remove_false = max(newtime(B)); C = find(newtime > remove_false); P = size(C,2); spiketime(1:P) = newtime(C); spikechan(1:P) = newchan(C); else spiketime(1:P) = newtime; spikechan(1:P) = newchan; 101

PAGE 102

end N1 = num_; G = 1; for M =1:30 Q1=find(spiketime >=(M-1)*10 & spiketime < M 10); hist_data(N,G,N1) = size(Q1,2); Q2=find(spiketime >=(M-1)*10+2 & spiketime < M 10+2); hist_data(N,G+1,N1) = size(Q2,2); Q3=find(spiketime >=(M-1)*10+4 & spiketime < M 10+4); hist_data(N,G+2,N1) = size(Q3,2); Q4=find(spiketime >=(M-1)*10+6 & spiketime < M 10+6); hist_data(N,G+3,N1) = size(Q4,2); Q5=find(spiketime >=(M-1)*10+8 & spiketime < M 10+8); hist_data(N,G+4,N1) = size(Q5,2); G = G + 5; end clear spiketime spikechan end end count(m,1) = num_; %T %p = 1; for i = 1:N P = count(i,1); if P > 0 dat = hist_data(i,:,1:P); st_avg = mean(dat,3); %colapse across stim_avg stim_avg(i,1:150) = st_avg; %p = p+1; else stim_avg(i,1:150) = 0; end end 102

PAGE 103

The algorithm Eval_connect loads gr anger causality data, which has been stored as a 60 by 60 by N matrix in a Mat file, where N = the number of stimulation sites. function score = Eval_connect(fn1,fn2,fn3,fn4)%dose, culture) load(fn1, 'G*') base = abs(GrangerMatrix)-0.5; row_col = find(base <= 0); base(row_col) = 0; load(fn2, 'G*') B24h = abs(GrangerMatrix)-0.5; row_col = find(B24h <=0); B24h(row_col) = 0; load(fn3, 'G*') B48h = abs(GrangerMatrix)-0.5; row_col = find(B48h <=0); B48h(row_col) = 0; load(fn4, 'G*') B72h = abs(GrangerMatrix)-0.5; row_col = find(B72h <=0); B72h(row_col) = 0; Pr_Base = reshape(sum(base,3),60,60); % collapse across probe Pr_24h = reshape(sum(B24h,3),60,60); % collapse across probe Pr_48h = reshape(sum(B48h,3),60,60); % collapse across probe Pr_72h = reshape(sum(B72h,3),60,60); % collapse across probe Pr_24_Base = Pr_24h Pr_Base; Pr_48_Base = Pr_48h Pr_Base; Pr_72_Base = Pr_72h Pr_Base; row_col1 = find(Pr_24_Base(:,:) < -2 | Pr_24_Base(:,:) > 2); row_col3 = find(Pr_48_Base(:,:) < -2 | Pr_48_Base(:,:) > 2); row_col4 = find(Pr_72_Base(:,:) < -2 | Pr_72_Base(:,:) > 2); if size(row_col1,1) > 2 row_col2 = find(Pr_24_Base(:,:) < -2); score(1,1) = size(row_col2,1)/size(row_col1,1); row_col2 = find(Pr_48_Base(:,:) < -2); score(2,1) = size(row_col2,1)/size(row_col3,1); row_col2 = find(Pr_72_Base(:,:) < -2); score(3,1) = size(row_col2,1)/size(row_col4,1); end figure() subplot(1,3,1) imagesc(Pr_24_Base(:,:), [-10 10]); 103

PAGE 104

104 subplot(1,3,2) imagesc(Pr_48h_Base(:,:),[-10 10]); subplot(1,3,3) imagesc(Pr_72h_Base(:,:),[-10 10]);

PAGE 105

105 REFERENCES Ahnert-Hilger, G. and H. Bignalke. 1995. Molecular aspects of tetnus a botulism neurotoxin poisoning. Prog Neurobiol. 46(1), 83-96. Akbulut, D., K. Grant, and J. McLauchin. 2004. Development and application of real-time PCR assays to detect fragments of the Clostrid ium botulinum types A, B, and E neurotoxin genes for investigation of human foodborne and infant bot ulism. Foodborne Pathog Dis. 125(4), 247-257. Anne, C., F. Cornille, C. Lenoir, and B. Roques. 2001. High-throughput fluorogenic assay for determination of botulinum type B neurotox in protease activity. Anal Biochem. 291,253261. Aranda, E., M. Rodriguez, M. Asensio, and J. Cordoba. 1997. Detection of Clostridium botulinum types A, B, E and F in foods by PCR and DNA probe. Le tt Appl Microbiol. 25, 186-190. Arnon, S., R. Schechter, T. Ingles by, D. Henderson, J. Bartlett, M. Ascher, E. Eitzen, A. Fine, J. Hauer, M. Layton, S. Lillibridge, M. Oster holm, T. O'Toole, G. Parker, T. Perl, P. Russell, D. Swerdlow, and K. Tonat. 2001. Botulism Toxin as a Biological Weapon. JAMA. 285, 1059-1070. Barr J., H. Moura, A.E. Boyer, A.R. Woolfitt, S.R. Kalb, A. Pavlopoulos, L.G. McWilliams, J.G. Schmidt, R.A. Martinez, D.L. Ashley. 2005. Botulinum neurotoxin detection and differentiation by mass spectrocopy. Emerg Infect Dis. 11(10):1578. Black, J. D. and J. O. Dolly. 1986. Interaction of 125-Iabeled botulinum neurotoxins with nerve terminals: Ultrastructural Autoradiographical localization and quantification of distinct membrane acceptors for types A and B motor nerves. J Cell Biol. 103(2), 521-534. Braconnier, A., V. Brousoule, S. Perelle, P. Fach, C. Ngyun-The, and F. Carlin. 2001. Screening for Clostridium botulinum type A, B and E in cooked chilled foods containing vegtables and raw material using polymerase chain re action and molecular probes. J Food Protect. 64(2), 201-207. Bullens, R., G. O'Hanlon, E. Wagner, P. Molenaar K. Furukawa, K. Furukawa, J. J. Plomp, and H. J. Wilson. 2002. Complex gangliosides at the neuromuscular junction are membrane receptors for autoantibodies and botulinum neurotoxin. J Neurosci. 22(16), 6876-6884. Cadieux, B., B. Blanchfield, J. Smith, and J. Austin. 2005. A rapid chemiluminescent slot blot immunoassay for the detection and quantifica tion of Clostridium botulinum neurotoxin type E, in cultures. Int J Food Microbiol. 101, 9-16. Cadotte, A. J. 2004. Polymeric Drug Delivery of Antiepileptic Drugs to Neuronal Networks Cultured on Multielectrode Arrays. Ph. D. Dissertation, University of Florida, Biomedical Engineering.

PAGE 106

106 Cadotte, A. and T. B. DeMarse. 2005. Poly-Hem a a a drug delivery device for in vitro neural networks on micro-electrode ar rays. J Neural Eng. 2, 114-122. Cadotte, A., M Ding, T Demarse.2006. Quantification of neural interaction via granger causality. In: EEG/MEG Studies of Functional Neural Connectivity [Poste r] 2006 Neuroscience Meeting. Campbell, K., M. Collins, and A. East. 1993. ge ne probes for identification of the botulinal neurotoxin gene and specific identification of neurotoxin types B, E, and F. J Clin Microbiol. 31(9), 2255-2262. Caya, J., R. Angi, and J. Miller. 2004. Clostrid ium botulinum and the clinical labratorian: a detailed review of botulism, including biol ogical warfare remifications of botulinum toxin. Arch Pathol Lab Med. 128, 653-662. Center for Disease Control and Prevention. 1998. Ceneter for Disease Control and Prevention: botulism in the United States, 1899-1996. Handbook for epidimiologists, clinicians and labratory workers. CDC, Atlanta, GA. Chirwa, S., J. Mack, R. Park, K. Dennis, and A. Aduonum. 2001. An in vivo meodel for investgating bilateral synaptic plasticity ac ross CA3/CA1 synapses in guinea pig dorsal hippocampus. J Neurosci Meth. 110, 25-30. Dahlenborg, M., E. Borch, and P. Radstrome. 2001. Development of a combined selection and enrichment PCR procedure for Clostridium botulinum types B, E and F and it use to determine prevalence in fecal samples fr om slaughtered pigs. Appl and Environ Microbiol. 67, 4781-4788. Dan, Y., and M.M. Poo. 2006. Spike timing-deenda nt plasticity: from synapse to perception. Physiol Rev. 86: 1033-1048. Ding, M., Y. Chen, and S. Bressler. 2006. Granger causality: basic theo ry and appliction to neuroscience, 437-459. In B. Schelter, M. Winterhalder, and J. Timmer (eds.), Handbook of Time Series Anlysis: Recent Theoretical Developments and Applications. Wiley, Verlag, Berlin. Dollegast, G., M. Triscott, G. Beard, D. Bottoms, T. Cheng, B. Roh, M. Roman, P. Hall, and J. Brown. 1993. Sensitive enzyme-linked imm unosorbent assay for the detection of Clostridium botulinum neurotoxins A, B, and E using signal amplification via enzyme linked coagulation assay. J Clin Microbiol. 31, 2402-2409. Dong, M., Y. Felix, W. Tepp, C. Dean, E. Johnson, R. Janz, and E. Chapman. 2006, 12 March. SV2 is the protein receptor for botulism neur otoxin A. Science [Online] Science Express. Available, www.scienceexpress.org.

PAGE 107

Ekong, T., I. Feavers, and D. Sesardic. 1997. Reco mbinant SNAP-25 is an effective substrate for Clostridium botulinum type A toxin endpeptidase activity in vitro. Microbiol. 143, 33373347. Emanuel, P., J. Dang, J. Gebhardt, J. Aldrisch, E. Garber, H. Kulga, P. Stopa, J. Valdes, and A. Dion-Schultz. 2000. Recombinant antibodies: a new reagent for biological agent detection. Biosens Bioelectron. 14, 751-759. Eytan, D., N. Brenner, and S. Marom. 2003. Selective Adaptation in Ne tworks of Cortical Neurons. J Neurosci. 23(28), 9349-9356. Fach, P., M. Gibert, R. Griffais, J. guillou, and M. Popoff =. 1995. PCR and gene pobe identification of botulinum ne urotoxin A-, B-, E-, Fand Gproducing Clostridium spp. and the evaluation in food samples. A ppl Envioron Microbiol. 61, 389-392. Fach, P., D. Hauser, J. Guillou, and M. Popoff. 1993. Polymerase chanin reaction for the rapididentification of Clostridium Botulinum type A strains in food samples. J Appl Bacteriol. 75, 234-239. Ferreira, J., B. Baumstark, M. Hamdy, and S. McCay. 1993. Polymerase chain reaction for detection of type A Clostridium botulinum in foods. J Food Protect. 56(1), 18-20. Ferreira, J., S. Eliasberg, P. Edmonds, and M. Harrison. 2004. Comparision of the mouse bioassay and enzyme-linked immunosorbent assa y proceedures for the detection of type A botulinal toxin in food. J Food Protect. 67((1)), 203-206. Ferreira, J., S. Maslanka, E. Johnson, and M. Goodenough. 2003. Detection of botulinal neurotoxins types A, B, E and F by amlifi ed enzyme-linked immunosorbent assay: collborative study. JAOAC Int. 86(2), 314-331. Gholmieh G, Soussou W, Courellis S, Marmarlis V, Berger T, and Baudry M. 2001. A biosensor for detecting changes in cogna tive processing based on nonlinear system analysis. Biosens Bioelectron. 16, 491-501. Gholmieh G, Spiros C, Fakheri S, Cheung E, Marmarelis V, Baudry M, and Berger T. 2003. Detection and classification of neurot oxins using novel short-term plasticity quantification method. Biosen s Bioelectron. 18, 1467-1478. Gramowski, A., K. Jugelt, D. G. Weiss, and G. W. Gross. 2004. Substance identification by quantitative characterization of oscillatory activity in murine spinal cord networks on microelectrode arrays. Eur J of Neurosci. 19, 2815-2825. Gross, G. W., B. K. Rhoades, H. M. E. Azzaay, and M.-C. Wu. 1995. The use of neuronal networks on multiel ectrode arrays as biosensors. Biosens Bioelectron. 10, 553-567. 107

PAGE 108

Gross, G. W., A. Harsch, B. K. Rhoades, a nd W. Gopel. 1997. Oder, drug and toxin analysis with neuronal networks in vi tro: extracellular recording of network responses. Biosens Bioelctron. 12(5), 373-393. Habets, A., A. Van Dingen, F. Van Huizen, a nd Corner MA. 1987. Spontaneous neurnal firing patterns in fetal rat cortical networks during development in vitro: a quantitative analysis. Exp Brain Res. 69, 43-52. Hallis, B., B. James, and C. Shone. 1996. Development of novel assays for botulinum type A and B neurotoxins based on their endopeptidase activities. J Clin Mi crobiol. 34, 1934-1938. Hoch, D. H., M. Romero-Mira, B. E. Ehrlich, A. Finkelstein, B. R. DasGupta, and L. Simpson. 1985. Channels formed by botulinum, tetnus, and diptheria toxins in pl anar lipid bilayers: relevence to translocation of protei ns across membranes. PNAS. 82, 1692-1696. Jimbo Y, T. Tateno, H. P.C. Robinson. 1999. Simultaneous induction of pathway-specific potentiation and depression in networks of cortical neurons. Biophysical J 76: 670-678. Jimbo, Y., A. Kawana, P. Par odi, and V. Torre. 2000. The dynamics of a neuronal culture of dissociated cortical neurons of neonatal rats. Biol Cybern. 83, 1-20. Kamioka, H., E. Maeda, Y. Jimbo, and H. Robinson, Kawana. 1996. Spontaneous periodic synchronized bursting during formation of matu re patterns of connections in cortical cultures. Neurosci Lett. 206, 109-112. Keefer, E., S. Norton, N. A. J. Boyle, V. Talesa, and G. W. Gro ss. 2001. Acute Toxicity Screening of Novel AChe Inhibitors Us ing Neuronal Networks on Microelectrode Arrays. Neurotoxicology. 22, 3-12. Keller, J. E., F. Cai, and E. A. Neale. 2004. Uptake of Botulinum Neurotoxin into Cultured Neurons. Biochemistry. 43, 526-532. Keller, J., J. Nowakowski, M. F ilbert, and M. Alder. 1999. Rapid microplate assay for mintoring botulinum neurotoxin B catalytic ac tivity. J Appll Toxicol 19,S13-S17. Kimura, B., S. Kawaski, and H. F. Nakano, T. 2001. Rapid, quanti tiative PCR monitoring growth of Clostridium botulinum type E in modified-atmosphere-packaged fish. Appl Environ Microbiol. 67(1), 206-216. Lindstrom, M., R. Keto, M. Annukka, M. Nevas, S. Heilm, and H. Korkeala. 2001. Multiplex PCR assay for the detection and identification of Clostridium botulinum types A, B, E, and F in food and fecal material. Appl Envioron Microbiol. 67(12), 5694-5699. Liu, W., V. Montana, E. Chapman, U. Mohideen, and V. Parpur a. 2003. Botulinum toxin type B micromechanosensor. PNAS. 100, 13621-13625. 108

PAGE 109

Lovenklev, M., E. Holst, E. Borch, and P. Radstom. 2004. Relative neurotoxin gene expression in Clostridium botulinum type B, determin ed using quantitative reverse transcription PCR. Appl Environ Mi crobiol. 70(5), 2919-2927. Maeda, E., P. C. Robinson, and A. Kawa na. 1995. The Mechanisms of generation and propagation of synchronized bursting in deve loping networks of cortical neurons. The Journal of Neuroscience. 15, 6834-6845. Marom, S. and G. Shahaf. 2002. Development, learning and memory in large random networks of cortical neurons: lessons beyond an atomy. Q Rev Biophysics. 35, 63-87. McGrath, S., J. Dooley, R. Haylock, and 2000. Quantification of Clostridium botulinum toxin gene expression by cometetive revearse tran scription-PCR. Appl Envioron Microbiol. 66(4), 1423-1428. Meister, M., R. Wong, Baylor DA, and C. Shatz. 1991. Synchronous burst of action potentials in gnglion cells of the developing mamm alian retina. Science 252, 939-943. Meister, M., L. Lagnado, and D. A. Baylor 1995. Concerted Signali ng by Retinal Ganglion Cells. Science 270, 1207-1210. Meister, M., J. Pine, and D. A. Baylor. 1994. Mu lti-neuronal signals from the retina: acquisition and analysis. J Neurosci Meth. 51, 95-106. Montecucco, C., O. Rossetto, and G. Schiavo. 20 04. Presynaptic receptor arrays for clostridial neurotoxins. Trends Mi crobiol. 12(10), 442-446. Morefield, S. I., E. W. Keefer, K. D. Chapma n, and G. W. Gross. 2000. Drug evaluations using neuronal networks cultured on microelectrode arrays. Biosens Bioelectron. 15, 383-396. Morf, M., A. Vieira, D.T.L. Lee, and T. Kailath. 1978. Recursive multichannel maximum entropy spectral estimation. IEEE Trans on Geosci Electron. G-16(2): 8594. Muramoto, K., M. Ichikawa, K. Kobayashi, M. Kawahara, and Y. Kuroda. 1993. Formation and maturation of synapses in primary cultures of rat cereberal cortical cells; an electron microscopic study. Neurosience Research. 16(2), 95-103. Nakanishi, K. and F. Kukita. 1998. Functional synapses in sync hronized bursting of neocortical neurons in culture. Brain Res. 795, 137-146. Nelsen, D. and M. Cox. 2005. Lehninger Principl es of Biochemistry. W H Freeman, New York. Nishiki, T., Y. Kamata, Y. Nemoto, A. Omori, T. Ito, M. Takahashi, and S. Kozaki. 1994. Identification of a protein receptor for Clostr idium botulinum Type B Neurotoxin in Rat Brain Synapsomes. J Biol Chem. 269(14), 10498-10503. 109

PAGE 110

Nishiki, T., Y. Tokuyama, Nemboto, A. Yoshida, K. Sato, M. Sekiguchi, M. Takahashi, and S. Kozaki. 1996. The high-affinity binding of Clostridium botulinum B neurotoxin to synaptogamin II associated with ganglios ides GT1b/GD1a. FEBS Lett. 378, 253-257. O'Brien, T., L. Johnson, Aldrich, S. Allen, L. Liang, A. Plummer, S. Krak, and A. Boiarski. 2000. The development of immunoassays to four biologcal threat agen ts in a bidiffractive grating biosensor. Biosens Bioelectron. 14, 815-828. Pancrazio, J. J., S. A. Gray, Y. S. Shubin, N. Kulagina, D. S. Cuttino, K. M. E. Shaffer, K, A. Curran, B. Zim, G. W. Gross, and T. J. O'Shaughnessy. 2003. A portable microelectrode array recording system incorporating cultured neuronal networks for neurotoxin detection. Biosens Bioelectron. 18, 1339-1347. Peruski, A., L. Johnson, and L. Peruski. 2002. Rapid and sensitive detection of biological warefare agents using time-resolved Fluor escence assays. J Immunol Meth 263, 35-41. Poo, M. and G. Bi. 1998. Synptic modifications by correlated activity, Hebb' s postulate revisited. Ann Rev Neurosci. 24, 139-166. Potter, S. and T. DeMarse. 2001. A new approach to neuronal cell culture for long-term studies. J Neurosci Meth. 110, 17-24. Reppel, M., P. Igelmund, U. Egert, F. Juchelka, J. Hescheler, and I. Drobi nskaya. 2007. Effect of Cardioactive Drugs on Action Potential Gene ration and Propagation in Embryonic Stem Cell Derived Cardiomyocytes. Cell Physiol Biochem. 19, 213-224. Rowe-Taitt, C., J. Golden, M. Feildstein, J. Crass, K. Hoffman, and F. Liger. 2000. Array biosensor for detection of biohazards Biosens Bioelectron. 14, 785-794. Rummel A, T. Karnath, H. Bigalke, and T. Binz. 2004a. Syna ptogamins I and II act as nerve cell receptors for botulinum neurotoxin G. J Biol Chem. 279, 30865-30870. Rummel, A., S. Mahrold, H. Bigalke, and T. Binz. 2004b. The Hcc domain of botulinum neurotoxins A and B exhibits a singular ganglioside binding site displaying serotype specific carbohydrate interaction. Mol Microbiol. 51(3), 631-643. Scarlatos, A., B. W. Welt, B. Y. Cooper, D. Archer, T. DeMarse, and K. V. Chau. 2005. Methods for Detecting Botulinum Toxin with Applicability to Screening Foods Against Biological Terrorist Attacks. J Food Sci. 70(8), R121-R130. Schmidt, J. and R. Staford. 2003. Fluorogenic subs trates for the protease activities of botulinum neurotoxins, serotypes A, B, and E. Appl Envioron Microbiol 50, 63-67. Schmidt, J., R. Staford, and C. Millard. 2001. High-throughput assays for botulinum neurotoxin proteolytic activity: serptypes A, B, D, and F. Anal Biochem. 296, 130-137. 110

PAGE 111

Sciacchitano, C. and I. Hirshfield. 1996. Molecu lar detecton of Clostridium botulinum type E neurotoxin gene in smoked fish by pol ymerase chain reaction and cappillary electrophoresis. J AOAC Int. 79(4), 861-865. Segev, R., M. Beveniste, E. Hulata, N. Cohen, and A. Palevski. 2002. Long term behavior of Lithographically Prepared In Vitro Neurona l Networks. Phys Rev Lett. 88(11), 118102. Selinger, J., J. J. Pancrazio, and G. W. Gro ss. 2004. Measuring synchronization in neuraonal networks for biosensor applicati ons. Biosens Bioelectron.19, 675-683. Shao, J., T. Tso, and R. Butera. 2006. Bursting with Slow Kinetics: A Role for Small World. Neural Comput. 18, 2029-2035. Shone, C., P. Smith, N. Appelton, P. Hambelton, N. Modi, S. Gatley, and J. Melling. 1985. Monoclonal antibody-based immunoassay for type A Clostridium botulinum toxin is comparable to the mouse bioassay. Appl and Environ Microbiol. 50(63-67). Simpson, L. L. 1983. The interaction between aminoquinolines and pr esynaptically acting neurotoxins. J Pharmacol and Exp Ther. 222, 43-48. Simpson, L. L. 2004. Identification of the ma jor stepsin botulinum toxin action. Annu Rev Pharmacol. 44, 167-193. Simpson, L. 1996. Botulinum toxin: a deadly poi son sheds its negative im age. Ann Int Med. 125, 616-617. Solomon, H. and T. Lily. 2001. Cl ostridium botulinum. Available: http//www.cfsan.fda.gov/~ebam-17.html. Stenger, D. A., G. W. Gross, E. W. Keefer, K. M. Shaffer, J. D. Andreadis, W. Ma, and J. J. Pancrazio. 2001. Detection of physiologica lly active compounds using cell-based biosensors. Trends Bi otechnol. 19(8), 304-309. Streit, J., A. Tescherter, and P. Darbon. 2005. Rythem Generation in Spinal Cultures is it the Neuron or the Network, 26. In M. Taketani and M. Baudry (eds.), Advances in Network Electrophysiology Using Multi-Electrode Arrays. New York. Tabak, J. and P. Latham. 2003. Analysis of spontaneous bursting activity in random neural networks. Neurorepor t. 14(11), 1445-1449. Trevan, J. 1927. The error of determin ing toxicicity. Proc R Soc. 101B, 483-514. van Barr LM, A.G. Hulst, A.L. de Jong, E.R.J. Wils. 2002. Characterizati on of botulinum toxins A and B, by matrix-assisted laser deso rption ionisation and electrospray mass spectrometry. J Chromatogr A 1035:95. 111

PAGE 112

van Barr LM, A.G. Hulst, A.L. de Jong, E.R.J. Wils. 2004. Characterizati on of botulinum toxins C, D, E, and F by matrix-assisted laser de sorption ionisation and electrospray mass spectrometry. J Chromatogr A 970:97. van Pelt, J, P.S. Woters, M.A. Corner, W.L. Rutten, G.J.A. Ramakers. 2004. Long-term characterization of firing dynami cs of spontaneous bursts in culltured neural networks. IEEE Transiomed Eng 51 (11):2051-62. van Pelt, J., I. Vajda, P. Wolters, M. Corner, a nd G. Ramakers. 2005. Dynamics and placticity in developing neuronal networks in vi tro. Prog Brain Res. 147, 173-188. Verderio, C., S. Coco, O. Rossetto, C. Montecu cco, and M. Matteoli. 1999. Internalization and proteolytic action in CNS neurons and as trocytes. J Neurochem. 73(1), 372-379. Verderio, C., O. Rossotto, C. Grumelli, C. Fr assoni, C. Montecucco, and M. Matteoli. 2006. Entering neurons: botulinum toxins and synaptic vescle recycling. EMBO Reports. 7(10):995-999. Wagenaar, D., Z. Nadasdy, and S. Potter. 2006. Pers istent dynamic attractors in activity patterns of cutured neuronal neworks. Phys Rev E. 73:0151907-1-8. Wagenaar, D., J. Pine, and S. Potter. 2006. An extremely rich repertoire of bursting patterns during the development of cortic al cultures. BMC Neurosci. 7. Wagenaar, D. and S. Potter. 2002. Real-time multichannel stimulus artifact supression by local curve fitting. J Neurosci Meth. 120:113-120. Witcome, M., K. Newton, K. Jameson, B. Hallis, P. Dunnigan, E. Mackay, S. Clarke, R. Taylor, J. Gaze, K. Foster, and C. Shone. 1999. Development of an In Vitro Bioassay for Clostridium Botulinum Type B Neurotoxin in Foods That is More Sensitive than the Mouse Bioassay. Appl and Environ Microbiol. 65(9, Sept):3787-3792. Yowler, B. C., R. D. Kensinger, and C. L. Schengrund. 2002. Botu linum neurotoxin A activity is dependant upon the presence of specific gangliosides in ne uroblastoma cells expressing Synaptogamin I. J Biol Chem. 277(36):32815-32819. 112

PAGE 113

113 BIOGRAPHICAL SKETCH Amber Scarlatos graduated cum laude with a Bachelor of Science degree in chemical engineering from the University of Florida in May 2003. In A ugust 2003 she received an alumni fellowship to pursue her Doctorate of Philosophy in agricultural and biologi cal engineering at the University of Florida. Her research interest i nvolves the development of living tissue cultures as broadly applicable biosensors for biological agents. In particular, her cu rrent research involves the use of cortical tissue grown on a microe lectrode array to detect Botulism toxin.