University of Southern California Dissertations and Theses

Improving stimulation strategies for epiretinal prostheses

(USC Thesis Other)

Improving stimulation strategies for epiretinal prostheses

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content IMPROVING STIMULATION STRATEGIES FOR
EPIRETINAL PROSTHESES
by
Andrew C. Weitz
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
May 2013
Copyright 2013 Andrew C. Weitz
Epigraph
If only we could pull out our brain and use only our eyes.
- Pablo Picasso
ii
Acknowledgments
Completing my Ph.D. would not have been possible without the generous help and sup-
port of others. I would like to thank my adviser, Dr. James Weiland, for giving me the
freedom to be independent while making sure that I stayed focused and on track. I was
very fortunate to have an adviser who always made time for discussions and who allowed
me to maintain a balanced personal and professional life. I owe many thanks to Dr. Bob
Chow, whose technical advice was instrumental throughout my work. Though I am an
engineer by training, Bob helped me develop a passion for science. I would also like to
thank the rest of my committee, Drs. Mark Humayun and A.P. Sampath, for the valuable
advice they provided over the years.
I am very grateful to Matthew Behrend, who developed the initial framework for my
research and who spent countless hours training me to perform experiments. Matthew
was an excellent teacher and helped me grasp several concepts that were fundamental
to my work. I would also like to thank my labmates, past and present, for their helpful
discussions and assistance with my research. Specifically, I owe thanks to Alice Cho,
Nancy Lee, Ashish Ahuja, Artin Petrossians, Aditi Ray, Leanne Chan, Devyani Nanduri,
Alan Horsager, Vivek Pradeep, Neha Parikh, Tim Nayar, Navya Davuluri, Samantha
iii
Cunningham, Steven Walston, Boshuo Wang, Kiran Nimmagadda, Karthik Murali, and
Jung Hwa Cho. I would also like to thank my collaborators at other universities, as well
as employees of Second Sight Medical Products, Inc. These people include Ron Klein,
Bill Hauswirth, Vince Chiodo, Loren Looger, Jasper Akerboom, Rob Greenberg, Punita
Christopher, Vara Wuyyuru, Jianing Wei, Jessy Dorn, Uday Patel, and Jordan Neysmith.
I am very thankful for my parents, who instilled good values and a strong work ethic
in me from a young age. They have been a source of encouragement throughout my
life. Last, and most importantly, I would like to thank my wife Lindsay for her constant
patience and understanding during my Ph.D. career. I am extremely fortunate to have
such a wonderful wife who takes care of me and provides me with her unconditional love
and support.
iv
Table of Contents
Epigraph ii
Acknowledgments iii
List of Tables viii
List of Figures ix
List of Abbreviations xiii
Abstract xvi
Chapter 1: Introduction 1
1.1 History of Visual Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Overview of the Visual System . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Diseases of the Eye . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Components of a Visual Prosthesis . . . . . . . . . . . . . . . . . . . . . . 11
1.4.1 Image Capture Device . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.2 Image Processor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.3 Stimulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.4 Telemetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.5 Electrode Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Safety Concerns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Cortical Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7 Optic Nerve Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.8 Retinal Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.8.1 Subretinal Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.8.2 Epiretinal Prostheses . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.9 Retinal Prosthesis Implications from Animal Studies . . . . . . . . . . . . 35
1.9.1 Direct vs. Indirect Activation of Ganglion Cells . . . . . . . . . . . 36
1.9.2 Visual Acuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.9.3 Effects of Retinal Degeneration . . . . . . . . . . . . . . . . . . . . 39
1.10 Gene Therapy Treatments for Outer Retinal Degenerations . . . . . . . . 40
1.11 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.12 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
v
1.12.1 Basic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.12.2 Summary of Results . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.12.3 Structure of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Chapter 2: Experimental Design and Methods 51
2.1 Probing Activity in Electrically Excitable Cells . . . . . . . . . . . . . . . 51
2.1.1 Electrical Recording . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.1.2 Optical Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.2 Calcium Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2.1 History and Basic Principles . . . . . . . . . . . . . . . . . . . . . 57
2.2.2 Delivering Calcium Indicators to Cells . . . . . . . . . . . . . . . . 61
2.2.2.1 Synthetic Dye Loading . . . . . . . . . . . . . . . . . . . 61
2.2.2.2 Expression of Genetically Encoded Calcium Indicators . . 63
2.3 Retrograde Dye Loading in Adult Rodent Retina . . . . . . . . . . . . . . 66
2.4 Transducing RGCs with a Custom Adeno-Associated Viral Vector . . . . 69
2.4.1 Plasmid Construction . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4.1.2 Detailed Procedure . . . . . . . . . . . . . . . . . . . . . 72
2.4.2 Vector Packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4.3 Transducing Adult Rat Retina . . . . . . . . . . . . . . . . . . . . 77
2.4.3.1 Intravitreal Injections . . . . . . . . . . . . . . . . . . . . 78
2.4.3.2 Histology . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.4.3.3 Transduction Profiles . . . . . . . . . . . . . . . . . . . . 79
2.5 Imaging Calcium Transients in Transduced RGCs . . . . . . . . . . . . . . 84
2.5.1 Axonal Activation . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.5.2 Comparing Fluorescence Responses of Different Calcium Indicators 87
2.5.3 Correlating Electrical Activity with Calcium Transients . . . . . . 88
2.5.4 Statistics about the Properties of Electrically Responsive Cells . . 88
2.5.5 Effects of Rat Strain, Temperature, and Cytomorbidity . . . . . . 90
2.6 Apparatus Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.6.1 Electrophysiology Rig . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.6.2 Electrode Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.6.2.1 Electrode Modeling . . . . . . . . . . . . . . . . . . . . . 97
2.6.2.2 Microfabrication Process Flow . . . . . . . . . . . . . . . 103
2.6.2.3 MEA Layouts . . . . . . . . . . . . . . . . . . . . . . . . 106
2.6.2.4 Electrode Characterization . . . . . . . . . . . . . . . . . 107
2.7 Data Collection and Processing . . . . . . . . . . . . . . . . . . . . . . . . 109
2.7.1 Retina Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
2.7.2 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
2.7.3 Image Data Reduction . . . . . . . . . . . . . . . . . . . . . . . . . 113
2.7.4 Calculating RGC Thresholds . . . . . . . . . . . . . . . . . . . . . 114
2.7.4.1 Repeatability of the Threshold Measurement . . . . . . . 117
2.7.5 Threshold Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
vi
Chapter 3: Using Interphase Gaps to Lower Thresholds 122
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.2.1 Animal Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.2.2 Computational Modeling . . . . . . . . . . . . . . . . . . . . . . . 124
3.2.3 Human Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.2.4 Curve Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Chapter 4: Using Pulse Width to Control Percept Shape 137
4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.2 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.3 Threshold Maps for 200-m-diameter Electrodes . . . . . . . . . . . . . . 145
4.3.1 Effects of Synaptic Blockers . . . . . . . . . . . . . . . . . . . . . . 147
4.3.2 Selectivity Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.3.3 Strength-Duration Curves . . . . . . . . . . . . . . . . . . . . . . . 153
4.3.4 Comparison of 20-Hz Sinusoidal and Pulsatile Stimulation . . . . . 156
4.4 Threshold Maps for 75- and 30-m-diameter Electrodes . . . . . . . . . . 159
4.5 Reverberating Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.6 Effects of Retinal Degeneration . . . . . . . . . . . . . . . . . . . . . . . . 162
4.7 Multielectrode Pattern Stimulation . . . . . . . . . . . . . . . . . . . . . . 166
4.7.1 Interleaved Stimulation . . . . . . . . . . . . . . . . . . . . . . . . 167
4.7.2 Line Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
4.7.2.1 Effects of Pulse Width . . . . . . . . . . . . . . . . . . . 170
4.7.3 Letters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
4.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Chapter 5: Conclusions and Future Work 179
5.1 Recommendations for Epiretinal Prostheses . . . . . . . . . . . . . . . . . 179
5.2 Future Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.2.1 Retinal Transduction . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.2.2 Retinal Stimulation . . . . . . . . . . . . . . . . . . . . . . . . . . 184
References 186
vii
List of Tables
2.1 Non-viral DNA transfection methods . . . . . . . . . . . . . . . . . . . . . 64
2.2 Percentage of RGCs and other retinal cells labeled with GCaMP3 following
intravitreal injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.1 Stimulation parameters used in the interphase gap experiments . . . . . . 128
4.1 Stimulation parameters used in the pulse width experiments . . . . . . . . 144
4.2 Number of retinas and RGCs in each threshold map (200-m-diameter
electrodes, wild-type retina) . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.3 Synaptic blocker cocktail used to pharmacologically isolate RGCs . . . . . 149
4.4 Number of retinas and RGCs in each threshold map (75- and 30-m-
diameter electrodes, wild-type retina) . . . . . . . . . . . . . . . . . . . . 160
4.5 Number of retinas and RGCs in each threshold map (200-m-diameter
electrodes, RD retina) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.6 Comparison of thresholds in WT and RD retinas for three pulse widths . 165
viii
List of Figures
1.1 Brindley’s chronic cortical prosthesis . . . . . . . . . . . . . . . . . . . . . 3
1.2 Cross section of the human eye . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Cross section of the human retina . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Components of an epiretinal prosthesis . . . . . . . . . . . . . . . . . . . . 12
1.5 Three types of electrode arrays . . . . . . . . . . . . . . . . . . . . . . . . 16
1.6 Simulations of prosthetic vision . . . . . . . . . . . . . . . . . . . . . . . . 17
1.7 Biphasic current pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.8 Argus II epiretinal prosthesis . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.1 Electrode arrays for measuring RGC spiking activity . . . . . . . . . . . . 53
2.2 Arch(D95N) genetically encoded voltage indicator . . . . . . . . . . . . . 56
2.3 Crystal structure of GCaMP . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.4 AAV pseudotype vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.5 Retrograde dye loading in adult rodent retina . . . . . . . . . . . . . . . . 69
2.6 Map of pAAV2-CAG-GCaMP . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.7 pGFP plasmid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.8 pGFP backbone gel electrophoresis . . . . . . . . . . . . . . . . . . . . . . 73
2.9 PCR primers used for GCaMP cDNA isolation . . . . . . . . . . . . . . . 73
ix
2.10 GCaMP insert gel electrophoresis . . . . . . . . . . . . . . . . . . . . . . . 74
2.11 pAAV2-CAG-GCaMP gel electrophoresis . . . . . . . . . . . . . . . . . . 75
2.12 Fluorescent imaging of AtT-20 cells transfected with pAAV-CAG-GCaMP3 76
2.13 Retinal wholemount of adult rat infected with AAV2-CAG-GCaMP3 . . . 80
2.14 Double-labeling RGCs with AAV2-CAG-GCaMP3 and Alexa 594 . . . . . 81
2.15 AAV2-CAG-GCaMP3 and AAV2/1-SYN1-GCaMP3 retinal histology . . . 83
2.16 AAV2/9-SYN1-GCaMP3 and AAV9-CAG-GFP retinal histology . . . . . 84
2.17 Axonal stimulation evoked by a 20-m-diameter electrode . . . . . . . . . 85
2.18 Axonal stimulation evoked by a 200-m-diameter electrode . . . . . . . . 86
2.19 Comparison of OGB-1, GCaMP3, and GCaMP5G fluorescence responses . 87
2.20 Correlation between GCaMP5G calcium transients and spiking activity . 89
2.21 Effects of temperature on GCaMP5G-labeled Copenhagen RGCs . . . . . 91
2.22 Electrophysiology rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.23 Fluorescence spectra for GCaMP and the GFP-4050A filter set . . . . . . 94
2.24 Fluorescence spectra for Alexa Fluor 594 and the TXRED-4040B filter set 94
2.25 Voltage divider and current source circuit diagram . . . . . . . . . . . . . 95
2.26 Custom PC boards used to interface with the retina . . . . . . . . . . . . 96
2.27 Randles cell model of the electrode-electrolyte interface . . . . . . . . . . 98
2.28 Impedance spectra of a 60-m-diameter Pt/Ir electrode . . . . . . . . . . 100
2.29 Modified Randles cell incorporating a parallel shunt capacitor . . . . . . . 101
2.30 Effect of shunt capacitance on electrode impedance . . . . . . . . . . . . . 102
2.31 Microfabrication process flow used to pattern ITO . . . . . . . . . . . . . 104
2.32 Microfabrication process flow used to pattern Si

N

and SU-8 . . . . . . . 105
x
2.33 MEA photographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
2.34 MEA photomicrographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
2.35 Effects of electrode size and material on impedance . . . . . . . . . . . . . 108
2.36 Stimulation protocol used for GCaMP5G in rat . . . . . . . . . . . . . . . 112
2.37 Marking ganglion cell bodies in ImageJ . . . . . . . . . . . . . . . . . . . 114
2.38 Raw and filtered fluorescence traces used for detecting calcium responses . 115
2.39 Post-stimulus time histograms used for rejecting false positives . . . . . . 116
2.40 Repeatability of the threshold measurement . . . . . . . . . . . . . . . . . 118
2.41 Example threshold map from a single experiment . . . . . . . . . . . . . . 120
2.42 Example threshold map from a multiple experiments . . . . . . . . . . . . 121
3.1 Circuit diagram of a tiger salamander ganglion cell membrane . . . . . . . 126
3.2 Modeling a salamander RGC’s response to stimulation . . . . . . . . . . . 127
3.3 Effects of threshold adaptation in a human prosthesis patient . . . . . . . 130
3.4 Effect of IPG duration on ganglion cell electrical thresholds . . . . . . . . 131
3.5 Effect of stimulus frequency and IPG length on anodic block . . . . . . . 133
3.6 Effect of IPG duration on perceptual thresholds . . . . . . . . . . . . . . . 134
4.1 Axonal stimulation in Argus I and II subjects . . . . . . . . . . . . . . . . 140
4.2 Pseudo-sinusoidal stimulation in an Argus I subject . . . . . . . . . . . . . 142
4.3 Threshold maps for 200-m-diameter electrodes in wild-type retina . . . . 146
4.4 Threshold vs. displacement from electrode center for 100-ms pulses . . . . 147
4.5 Effect of synaptic blockers on light flash responses . . . . . . . . . . . . . 149
4.6 Effect of synaptic blockers on RGC thresholds over a range of pulse widths 150
4.7 Selectivity ratios measured over a range of pulse widths . . . . . . . . . . 152
xi
4.8 Strength-duration curve plotted in terms of current density . . . . . . . . 155
4.9 Strength-duration curve plotted in terms of charge density . . . . . . . . . 157
4.10 Comparison of threshold maps for 20-Hz sine and square waves . . . . . . 158
4.11 Comparison of individual RGC thresholds for 20-Hz sine and square waves 158
4.12 Threshold maps for 75- and 30-m-diameter electrodes in wild-type retina 160
4.13 Comparison of selectivity ratios for three electrode sizes . . . . . . . . . . 161
4.14 Reverberating responses to high-amplitude stimulation . . . . . . . . . . . 162
4.15 GCaMP5G fluorescence responses in RD retina . . . . . . . . . . . . . . . 164
4.16 Threshold maps for 200-m-diameter electrodes in RD retina . . . . . . . 164
4.17 Multielectrode stimulation of RGCs with temporally interleaved pulses . . 168
4.18 Line pattern stimulation of RGCs with 25-ms pulses . . . . . . . . . . . . 170
4.19 Electrodes used for line pattern stimulation in Fig. 4.20 . . . . . . . . . . 171
4.20 Line pattern stimulation of RGCs with two pulse widths . . . . . . . . . . 172
4.21 Multielectrode stimulation in the shape of a V . . . . . . . . . . . . . . . 174
4.22 Multielectrode stimulation to spell the word LIFT . . . . . . . . . . . . . 174
xii
List of Abbreviations
AAV adeno-associated virus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40
AM acetoxymethyl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62
AMD age-related macular degeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
AMPA -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid . . . . . . . . . . . . . . . . . . . . 149
ANOVA analysis of variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
ASR Artificial Silicon Retina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
BAPTA 1,2-bis(2-aminophenoxy)ethane-N,N,N’,N’-tetraacetic acid . . . . . . . . . . . . . . . . . 57
bGH bovine growth hormone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
BSS balanced salt solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
CaM calmodulin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
CBA chicken -actin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
CCD charge-coupled device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
ChR2 channelrhodopsin-2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
CMOS complementary metal-oxide semiconductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
CMV cytomegalovirus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
CNQX 6-cyano-7-nitroquinoxaline-2,3-dione. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .149
d-APV d-(-)-2-amino-5-phosphonopentanoic acid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .149
DAPI 4’,6-diamidino-2-phenylindole. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79
DIP dual in-line package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
DMSO dimethyl sulfoxide. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .148
DNA deoxyribonucleic acid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
EGTA ethylene glycol-bis(2-aminoethylether)-N,N,N’,N’-tetraacetic acid . . . . . . . . . . 57
xiii
EIS electrochemical impedance spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
EMCCD electron-multiplied charge-coupled device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
FPLC fast protein liquid chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
FRET fluorescence resonance energy transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
GABA
A
-aminobutyric acid type A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
GCL ganglion cell layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
GECI genetically encoded calcium indicator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59
GFP green fluorescent protein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
HMDS hexamethyldisilazane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .103
ICMS intracortical microstimulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
ie immediate-early . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
IMI Intelligent Medical Implants, GmbH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
INL inner nuclear layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
IPG interphase gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
ITO indium tin oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
l-APB l-(+)-2-amino-4-phosphonobutyric acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
LB lysogeny broth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
LCA Leber congenital amaurosis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
LED light-emitting diode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
LGN lateral geniculate nucleus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
MEA multielectrode array. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45
mGluR6 metabotropic glutamate receptor 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
MPDA microphotodiode array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
NA numerical aperture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
NMDA N -methyl-d-aspartate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
OAT organic anion transporter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
OGB-1 Oregon Green BAPTA-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
PBS phosphate buffered saline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .78
PCB printed circuit board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
PCR polymerase chain reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
xiv
PECVD plasma-enhanced chemical vapor deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
poly(A) polyadenylation signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71
RD retinal degeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
RGC retinal ganglion cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
RIE reactive ion etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
RMS root mean square . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
RNA ribonucleic acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
RP retinitis pigmentosa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
sccm standard cubic centimeters per minute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
SD standard deviation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
SEM standard error of the mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Si

N

silicon nitride . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
SNR signal-to-noise ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
SSMP Second Sight Medical Products, Inc.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
TR terminal repeat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
TTL transistor-transistor logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
UV ultraviolet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .57
V1 primary visual cortex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
WPRE woodchuck hepatitis virus post-transcriptional regulatory element . . . . . . . . . 70
WT wild-type. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .49
YFP yellow fluorescent protein. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59
xv
Abstract
Epiretinal implants for the blind are designed to stimulate surviving retinal neurons,
thus bypassing the diseased photoreceptor layer. Single-unit or multielectrode recordings
from isolated animal retina are commonly used to inform the design of these implants.
However, such electrical recordings provide limited information about the spatial patterns
of retinal activation. Calcium imaging overcomes this limitation, as imaging enables high
spatial resolution mapping of retinal ganglion cell (RGC) activity as well as simultaneous
recording from hundreds of RGCs.
I developed a method for labeling the majority of ganglion cells in adult rat retina
with genetically encoded calcium indicators (GECIs). Intravitreal injection of an adeno-
associated viral vector targeted roughly 85% of ganglion cells with high specificity. Due
to the large fluorescence signals provided by GECIs, I was able to visualize the retina’s
response to electrical stimulation in real time. Imaging transduced retinas mounted on
multielectrode arrays revealed how varying stimulus parameters dramatically affects the
spatial extent of RGC activation.
Recent human subject testing indicated that patients often see large, elongated phos-
phenes due to stimulation of RGC axon bundles. My experiments revealed two potential
xvi
strategies for avoiding stimulation of axons: Short pulses (≤ 0.1 ms) selectively stimulate
ganglion cell somata at thresholds 40–60% lower than their axons, while long pulses
(≥ 25 ms) preferentially target bipolar cells. No axonal activation was observed during
stimulation with long pulses, regardless of amplitude. Responses were focal and gradually
became larger as stimulus amplitude was increased.
Low-frequency sinusoids can be used as an alternative to long pulses, as they require
significantly less charge to evoke responses. When stimulating with short pulses, inter-
phase gaps can be used to reduce overall thresholds by 10–25%. Retinal degeneration
appears to have no effect on response shape; however, bipolar cell thresholds in degenerate
animals were elevated by threefold.
Multielectrode stimulation with long pulses revealed that controlled patterns of gan-
glion cells can be activated with minimal electrode interactions. Lines and letters were
patterned on the retina with high spatial resolution. An array design with 75-m-diameter
electrodes on a 150-m pitch should permit a resolution of 0.26
∘ of visual field and a
Snellen acuity of 20/312.
xvii
Chapter 1
Introduction
Since the mid-1700s, humans have been trying to restore vision to the blind. The first
attempts involved gross application of electric shock to the skull (LeRoy, 1755; Marg,
1991). While unsuccessful, these efforts utilized the same principles employed by modern-
day visual prostheses. Electrical current, substituted for light, is delivered to cells along
the visual pathway in order to create the sensation of vision. Visual prostheses have
indeed come a long way since their inception in the 18th century. They have enabled blind
individuals to perceive light and perform simple visual tasks. Given the current state of
the technology, it is not unreasonable to suggest that within the next 10 to 20 years,
people with implants who were once blind will be reading small print and recognizing
faces. In this chapter (Sections 1.1–1.11), we review several important aspects of visual
prostheses and discuss how the field of artificial vision has progressed over the years
(Weitz and Weiland, in press).
1
1.1 History of Visual Prostheses
Successful artificial vision requires placing stimulating electrodes in close proximity to
their target cells. Such placement was first attempted by the German neurosurgeon
Otfrid Foerster in 1929. Foerster applied direct electrical stimulation to the visual cortex
(posterior pole of the brain’s occipital cortex) of a normally sighted patient under local
anesthesia. The patient reported seeing a spot of light, known as a phosphene, whose
location depended on the position of the electrode over the cortex (Foerster, 1929). Two
years later, this same result was demonstrated in a patient who had been blind for eight
years (Krause and Schum, 1931).
Although crude, these experiments demonstrated three principles that are crucial to
the success of visual prostheses: 1) Electrical stimulation can be substituted for light in
order to create visual perception; 2) The visual cortex is spatially organized such that
stimulating a given area of the cortex produces a phosphene in the corresponding part of
the visual field. This so-called retinotopic map originates in the retina and is conserved
throughout the brain’s visual centers; and 3) The ability of the visual cortex to perceive
light is retained in individuals who have become blind. We now know that this ability
persists for decades (Dobelle et al., 1974; Schmidt et al., 1996).
Having understood these principles, the British physiologist Giles Brindley designed
the first chronically implanted visual prosthesis in the late 1960s (Fig. 1.1). A blind
woman, 52 years of age, was the first to receive the device. An array of 80 platinum
electrodes was surgically positioned on her visual cortex. An array of 80 receivers, im-
planted between the skull and scalp, was wired to each electrode. In order to activate
2
a given receiver and its corresponding electrode, radiofrequency signals were delivered
through the scalp from an oscillator coil held above the receiver (Brindley and Lewin,
1968; Karny, 1975).
Figure 1.1: Left: Brindley’s chronic cortical prosthesis shown prior to implantation. Right: An
X-ray image of the same device post-implantation. In each image, the arrow points to the receiver
array, and the arrowhead points to the electrode array (Brindley and Lewin, 1968).
The woman implanted with the device reported seeing phosphenes in 39 separate
locations, indicating that nearly half of the electrodes were functional. For the most
part, the position of each phosphene in the visual field corresponded with the position
of each electrode on the cortex. Amazingly, Brindley reported six years later that many
parts of the device were still functioning (Brindley and Rushton, 1974; Karny, 1975).
Brindley’s work was pioneering in the field of prosthetic vision; he was the first to
successfully demonstrate a chronic multielectrode visual prosthesis. In the 1970s, William
Dobelle, an American biomedical engineer, made another major contribution to the field.
3
Dobelle designed a new prosthesis that enabled patients to recognize simple patterns,
including letters (Dobelle et al., 1974). He even integrated a television camera into the
setup; images from the camera were converted into electrical signals and delivered by the
electrodes. One patient was able to use the camera to detect horizontal and vertical lines
(Dobelle et al., 1979).
Despite the impressive results obtained by Brindley and Dobelle, their work never led
to a commercial device. The use of relatively large surface electrodes (∼ 1 mm
2
) required
delivering relatively high currents (milliamps). Consequently, when adjacent electrodes
were activated simultaneously, patients reported seeing a single phosphene. The inability
of multielectrode stimulation to elicit discrete phosphenes limited the practical usefulness
of the devices (Brindley, 1982; Maynard, 2001).
The next major advancement for visual prostheses came in 1990. Researchers found
that intracortical stimulation (i.e., using electrodes that penetrate the cortex) produced
phosphenes with currents 10 to 100 times lower than when surface electrodes were used.
They also found that individual phosphenes could be resolved when two electrodes, spaced
only 700m apart, were activated simultaneously (Bak et al., 1990). Together, these two
findings demonstrated that high-resolution stimulation of the visual cortex was indeed
possible. Furthermore, the recent advent of silicon micromachining enabled fabrication
of densely packed arrays with small electrodes (Campbell et al., 1991; Maynard, 2001;
Wise et al., 1970). The availability of these technologies, coupled with the prospect of
high-resolution vision, prompted many researchers to begin investigating what are now
considered modern-day visual prostheses.
4
Although stimulating the visual cortex had shown much promise, it became evident
that the cortex may not be the best site for a prosthesis. Different parts of the visual
cortex are specialized for different functions, such as recognizing color, orientation, and
shape (Margalit et al., 2002). Furthermore, organization of the visual cortex is complex
and does not follow a simple retinotopic map (i.e., two adjacent cortical areas might not
correspond to two adjacent areas of the visual field). This led researchers to investigate
prostheses placed at different locations along the visual pathway, specifically the retina
and optic nerve. Each type of prosthesis is named after the location it stimulates: retinal
(consisting of epiretinal, subretinal, and suprachoroidal), optic nerve, and cortical. To
date, retinal prostheses have received the most attention and are the farthest along in
development. The primary focus of this chapter is retinal prostheses; however, all types
of visual prostheses are covered.
1.2 Overview of the Visual System
Vision is the most complex human sense and is arguably the most important. Sight is
made possible by a complex network of interconnected structures, mostly located inside
the brain. In fact, nearly half of the brain’s processing power is devoted to vision (Walker,
2009).
The eye is the first component of the visual system. As shown in Fig. 1.2, the eye is
a spherical structure mostly composed of the vitreous gel. The outer layer of the eye is
comprised of two collagenic structures called the sclera and cornea. When light strikes
the eye, it passes through the transparent cornea and lens to form an image on the retina,
5
a 200-m-thick sheet that lines the back of the eye (W¨ assle, 2004). Extraocular muscles
rotate the eye to align the image on the fovea, the central part of the retina responsible
for high-acuity vision. Through a process called phototransduction, the retina converts
incoming light into electrical signals that are carried by the optic nerve to the brain’s
higher visual centers.
Figure 1.2: Cross section of the adult human eye. Light enters through the cornea and is focused
on the retina by the lens. Visual information is transmitted by the optic nerve to the higher visual
centers of the brain. (Image courtesy of Webvision, http://webvision.med.utah.edu/.)
As shown in Fig. 1.3, the retina is a complex structure formed of many layers of cells.
Light passes through each layer before reaching the photoreceptors. The photoreceptor
layer contains two types of cells: rods and cones. Rods detect dim light and are mostly
responsible for night vision. The human retina contains roughly 100 million rods, most
of which are located outside the fovea. Cones, which have the highest concentration
inside the fovea, are responsible for color vision in ambient daylight levels. There are
6
three types of cones—red, green, and blue—each tuned to detect photons of different
wavelengths. The human retina contains roughly 4.6 million cones (Curcio et al., 1990).
Retinal pigment epithelial cells, which line the back of the retina, provide metabolic and
functional support to the photoreceptors.
Figure 1.3: Cross section of the human retina. Light entering the retina travels through all the
layers before reaching the photoreceptors. Photoreceptors convert the incoming light to electrical
signals, which are passed through each layer of the retina to the ganglion cells. After significant
modulation of the incoming signals, ganglion cells carry the information to the brain via their
axons, which form the nerve fiber layer and optic nerve. (Image courtesy of Webvision,http:
//webvision.med.utah.edu/.)
When light strikes a photoreceptor, photons are absorbed by G protein-coupled re-
ceptors called opsins. This initiates a second messenger cascade, known as phototrans-
duction, which amplifies the incoming light stimulus and converts it to chemical and
electrical signals. The final step in the phototransduction cascade causes ion channels in
7
the photoreceptor cell membrane to close, which leads to hyperpolarization of the mem-
brane. Photoreceptors are among the few cell types in the body that hyperpolarize in
response to external stimuli (most other cell types depolarize).
When photoreceptors hyperpolarize, there is a reduction in the neurotransmitter glu-
tamate that is released from their synaptic terminals. Bipolar cells, which synapse with
photoreceptors, detect this reduction and relay the signal to ganglion cells (either directly
or through communication with amacrine cells). Ganglion cells convert the graded elec-
trical signal into a series of action potentials. Ganglion cell axons, which form the optic
nerve, transmit the action potentials to the cortex for further processing (W¨ assle, 2004).
The optic nerve terminates in a region of the thalamus called the lateral geniculate
nucleus (LGN). Axons leaving the LGN transmit the visual signals to the primary visual
cortex (V1). Visual information is processed in V1 and is also passed on to other higher
cortical visual centers. The collective activity of these centers is what creates our sense
of vision.
Although the visual cortex plays the largest role in visual processing, a considerable
amount of processing happens within the retina (which is considered part of the brain).
There are five classes of retinal cells, and some have more than 10 subtypes, each special-
ized for a unique function (Masland, 2001). Furthermore, the retina contains separate
pathways that carry different types of visual information. For example, the ON and OFF
pathways carry information about whether a spot of light is brighter (ON pathway) or
dimmer (OFF pathway) than the background light intensity. Finally, there is a signifi-
cant amount of adaptation and neural convergence (i.e., data compression) in the retina.
While there are more than 100 million photoreceptor cells for detecting light, there are
8
only one million ganglion cells for transmitting that visual information to the cortex
(W¨ assle, 2004).
1.3 Diseases of the Eye
All types of blindness can be divided into two broad classes: 1) those in which the retina
loses its ability to transduce light into electrical signals or 2) those in which the electrical
signals fail to reach the cortex (Maynard, 2001). Each class can be treated by a different
type of visual prosthesis. The former can be treated by retinal, optic nerve, and cortical
prostheses, while the latter can be treated only by cortical prostheses.
Most blinding conditions are related to diseases of the retina, most commonly aris-
ing from inherited retinal degenerations (RDs). The two most prevalent types of retinal
degenerative disorders are age-related macular degeneration (AMD) and retinitis pig-
mentosa (RP). Together, they account for millions of cases of blindness around the world
(Chader et al., 2009).
AMD, which affects persons mostly over the age of 50, relates to degeneration of
photoreceptor cells in the central visual field (the macula). As a result, patients with
AMD lose their high-acuity central vision but retain peripheral vision. Roughly 30–
50 million people around the world have AMD, 14 million of whom are blind or severely
visually impaired (Gehrs et al., 2006). Though the exact causes of AMD are still unknown,
risk factors include age, cigarette smoking, obesity, and genetic predisposition.
AMD occurs in two forms: neovascular (wet) and non-neovascular (dry). The dry
form, which constitutes roughly 90% of AMD cases, can be treated only with antioxidants,
9
which provide limited effectiveness (Tan et al., 2008). Treatment options for wet AMD
are also limited. Antineovascular drugs can be administered, but they must be injected
into the eye every two to six weeks (Menon and Walters, 2009).
Retinitis pigmentosa is the second most prevalent inherited retinal degeneration, af-
fecting roughly one in four thousand individuals worldwide (Hartong et al., 2006). RP
refers to more than 200 identified mutations that can cause a similar phenotype. The
disease primarily affects rod photoreceptors, causing night blindness and tunnel vision.
As is the case with AMD, treatment options are limited. Gene therapy techniques may
be promising for the future, but such techniques are still in their infancy. To this point,
gene therapy has only been successful in treating a rare form of RP that accounts for less
than 1% of all cases (Cideciyan, 2010; Stein et al., 2011).
In order for retinal prostheses to be effective for AMD and RP, a significant number
of retinal neurons must survive; otherwise there would be no cells for the implants to
electrically stimulate. Fortunately, it has been found that while AMD and RP cause loss
of photoreceptors in the outer retina, inner retinal cells remain largely intact. In the case
of RP, it has been reported that 78–88% of bipolar cells and 30–48% of ganglion cells
survive degeneration (Santos et al., 1997). Similarly, studies of AMD patients have shown
that inner retinal cells are relatively well preserved (Kim et al., 2002a,b). Nevertheless,
degeneration does take its toll on the inner retina. Surviving cells exhibit abnormalities
such as neurite sprouting, migration, and rewiring (Fariss et al., 2000; Marc et al., 2003).
It is still unknown to what extent these abnormalities affect prosthetic vision. However,
given the encouraging results obtained in retinal prosthesis patients, it does appear that
the effects of the abnormalities can be surmounted.
10
While retinal prostheses are effective for treating outer retinal degenerative disorders,
they are not suitable for treating diseases that affect the inner and/or whole retina. This
is also the case for vision loss caused by damage to the optic nerve, which can result from
diabetic retinopathy, glaucoma, tumors, ischemia, inflammation, or other complications.
For these patients, cortical prostheses remain the only viable option (Margalit et al.,
2002).
1.4 Components of a Visual Prosthesis
All visual prostheses contain the same basic elements: a device for capturing images,
hardware for processing those images, a stimulator for generating electrical impulses, a
telemetry system for delivering power and data to the stimulator, and an electrode array.
Fig. 1.4 depicts some of those components in the case of an epiretinal prosthesis. Each
component is described below.
1.4.1 Image Capture Device
In most cases, a digital camera is used to capture images. As shown in Fig. 1.4, the
camera can be concealed by mounting it to a pair of sunglasses. However, this requires
patients to move their heads in order to change gaze. A more elegant solution would
be to implant a miniature camera inside the eye so that gaze can be controlled with eye
movements. This method is currently under investigation and is a likely prospect for the
future (Chai et al., 2008; Hauer et al., 2007; Nasiatka et al., 2005; Stiles et al., 2007; Zhou
et al., 2010).
11
Figure 1.4: Components of an epiretinal prosthesis. Images are captured by a camera that is
mounted to a pair of sunglasses worn by the patient. Image data is processed (not shown) and
transmitted wirelessly to the retinal implant, which sits on the eye wall. The implant drives stim-
ulation of the electrode array, which is tacked to the retina. (Image courtesy of the Department
of Energy.)
One type of subretinal prosthesis was designed to operate without an external camera.
This prosthesis employs an array of microphotodiodes (devices that convert light into
electrical current) that are integrated with stimulating chip. The main advantage of
this type of prosthesis is its simplified design. However, the low quantum efficiency of
microphotodiodes poses a major problem: unnaturally bright light is needed in order to
generate enough stimulation current. These issues are discussed in more detail later in
the chapter.
12
1.4.2 Image Processor
The images captured by the camera must be converted to a stimulus pattern that can be
delivered by the electrode array. At present, this is accomplished by custom hardware
worn by the patient. In the future, it is possible that the hardware will be miniaturized
enough to be implanted inside the eye (Humayun et al., 2001).
In its simplest form, an image processing unit would merely pixelate the images and
assign each pixel to a different electrode. However, given the limited number of available
electrodes, more intelligent processing schemes have been proposed. One strategy is
to process images in ways that mimic retinal processing. For example, we know the
retina performs functions such as gain control, contrast enhancement, edge enhancement,
and motion detection. Since many visual prostheses bypass the retinal circuitry (by
stimulating bipolar or ganglion cells directly), it is logical to include these functions in
the image processor (Weiland et al., 2005).
Given the limited resolution of current-generation prostheses, some investigators are
working to incorporate object recognition algorithms into the devices (Fink et al., 2004;
Parikh et al., 2010). The algorithms identify important features or objects in an image
and notify patients via auditory or tactile cues. Combined with the visual information
delivered by the prosthesis, these cues can help patients navigate their environments and
avoid obstacles. At present, image recognition is a difficult computer vision problem and
is somewhat unreliable. As the algorithms improve over time, it is possible that they will
become a standard component of visual prostheses.
13
The most advanced image processing strategy makes use of something called a retina
encoder (Eckmiller et al., 2005; Nirenberg and Pandarinath, 2012). When ganglion cells
transmit an image to the brain, they do so by altering their rate of action potential firing.
Researchers have created models that predict ganglion cell firing patterns in response to
any input image. (This is done by projecting images onto isolated animal retina while
measuring the electrical activity of ganglion cells (Keat et al., 2001; Pillow et al., 2005).)
Applying these models to the images captured by prosthesis cameras would enable retinal
prostheses to deliver “natural” stimuli to ganglion cells, arguably resulting in the highest
quality vision.
1.4.3 Stimulator
A major component of any implantable visual prosthesis is a microelectronic chip that
provides stimulus current to the electrode array. This chip receives input from the image
processor and delivers appropriate signals to the electrodes. Since the electrode-tissue
interface is high impedance, there must be ample power for driving enough output current
to generate a phosphene. This requires a chip that can supply high voltages, which means
that large transistors must be used. Consequently, chip designers must make tradeoffs
between stimulator size, power consumption, and output capability (Weiland et al., 2005).
1.4.4 Telemetry
The implantable stimulator must have a means of receiving data from the external image
processor. In the case of cortical prostheses, a percutaneous connector can be secured to
the skull, and physical wires can be used to connect the stimulator and image processor
14
(Hochberg et al., 2006). However, percutaneous connections pose a risk of infection and
are not practical for retinal prostheses. A better solution is to transmit data wirelessly.
Similarly, the stimulator must have a means of receiving power. While an implantable
battery would be optimal, such batteries cannot supply enough power over a long enough
lifetime (Weiland et al., 2005). For this reason, power must also be transmitted wirelessly.
At present, the best solution for delivering power and data to an implantable prosthesis
is through inductive coupling. This involves the use of two conductive coils (inductors),
one placed outside and the other inside the body. Current passed through the external
(primary) coil generates a magnetic field, which induces a voltage across the internal
(secondary) coil. Because the magnetic field strength decays exponentially with distance,
the coils must be placed in close proximity (Weiland et al., 2005).
One group recently reported the design of a 256-channel epiretinal prosthesis stimula-
tor and telemetry system. Two separate radio frequencies are used to deliver power and
data. Data is transmitted at a rate of 2 Mb/s with a frequency of 22 MHz. Up to 100
mW of power is delivered at a frequency of 2 MHz. All 256 channels can be activated
simultaneously with a range of stimulus pulse widths and amplitudes (Chen et al., 2010).
In the future, systems that support thousands of electrodes will need to be developed.
1.4.5 Electrode Array
The electrode array is the only prosthesis component that makes a functional interface
with the body. Its design largely depends on the cells being targeted for stimulation.
Epiretinal prostheses, which target ganglion and bipolar cells, employ an array of elec-
trodes placed near the inner surface of the retina. Optic nerve prostheses use spiral cuff
15
electrodes that surround the optic nerve to stimulate ganglion cell axons. Cortical pros-
theses use arrays of penetrating microelectrodes to activate neurons that lie beneath the
surface of the visual cortex. Fig. 1.5 shows three types of electrode arrays that differ
vastly in their designs.
Figure 1.5: Left: Fundus photograph of an implanted Argus II epiretinal prosthesis (Second
Sight Medical Products, Inc.). The array contains 60 platinum disk electrodes, each 200 m
in diameter. (Image courtesy of Second Sight Medical Products, Inc.) Center: Fundus pho-
tograph of an implanted Active Subretinal Device (Retina Implant AG). The device contains
1550 microphotodiodes, each coupled to a titanium-nitride electrode (Besch et al., 2008). Right:
Scanning electron microscope image of the Utah Electrode Array cortical prosthesis. The array
consists of 100 1.5-mm-long penetrating electrodes. The shaft of each electrode is insulated with
polymer so that only the tip is exposed (Normann et al., 2009).
The resolution of a visual prosthesis is correlated with the number of electrodes in
its array. Each electrode can be considered as a pixel in an image. With a 4× 4 array
of 16 electrodes (pixels), as was the case with the first epiretinal implant to undergo
clinical trials, one would expect low-resolution vision. This raises the question of how
many electrodes are needed to restore vision to a useful level. It has been suggested that
a visual acuity of 20/80 should be the target resolution for a visual prosthesis (Palanker
et al., 2005). However, a more practical assessment of visual improvement might be based
on functional criteria. For example, to what extent can patients fitted with prostheses
navigate their environments, identify objects, read text, and recognize faces? Clinical
16
testing may indeed be the only way to determine the level of visual improvement that
patients find acceptable.
Since clinical testing with high-resolution electrode arrays is not yet possible, investi-
gators have used virtual reality simulations to determine how many electrodes are needed
for useful vision. Studies have indicated that 625 electrodes implanted in a 1-cm
2
area are
sufficient for enabling tasks such as guided mobility and reading (Cha et al., 1992a,b,c).
In reality, these tasks might be possible with even fewer electrodes. Subjects overcome the
limitation of having too few electrodes by scanning their head-mounted cameras across
the visual environment, thus enabling spatial and temporal integration of visual infor-
mation (Hayes et al., 2003). Fig. 1.6 demonstrates how three images might appear to a
retinal prosthesis user when represented by different numbers of electrodes.
Figure 1.6: Illustration of how three images might appear to retinal prosthesis patients fitted
with different numbers of electrodes: 100, 700, and 2500 (Freeman et al., 2011).
17
1.5 Safety Concerns
Assuming that a visual prosthesis can be implanted safely, it is important to ensure that
the device operates within safe levels of electrical stimulation. Applying too much cur-
rent to a given area could potentially damage electrodes or nearby cells. When charge
is delivered by an electrode, electrochemical redox reactions take place at the electrode
surface. If the reaction products (usually H
2
or O
2
gas) diffuse away from the surface,
irreversible corrosion is caused to the electrode. To mitigate this, most visual prostheses
stimulate with charge-balanced biphasic current pulses (Fig. 1.7). The leading phase,
typically the cathodic phase, injects charge in order to evoke responses in target cells.
This is followed by the anodic phase, which removes charge and reverses electrochemical
processes. Placing an interphase gap (IPG) between the two phases can lower the thresh-
old of excitation (Weitz et al., 2011). Using current to stimulate, rather than voltage,
ensures that charge is balanced across both phases of the pulse (charge = amplitude× pulse width), thereby preventing corrosion to the electrode. If voltage pulses are used to
stimulate, a blocking capacitor is typically used to ensure charge balance.
Figure 1.7: The most common electrical stimulation pulse used in visual prostheses is a biphasic
current pulse. The biphasic nature enables charge-balanced pulses to be delivered, which prevents
buildup of toxic byproducts that arise from electrochemical reactions at the electrode surface.
Addition of an interphase gap between the two phases can lower the threshold of excitation.
18
Even when charge-balanced biphasic current pulses are used, it is still possible to
damage an electrode by delivering too much charge per phase. The amount of charge that
can be delivered safely is known as the charge injection limit. This limit depends on the
electrode material and its surface area. Two commonly used electrode materials, platinum
and iridium oxide, have charge injection limits of 0.1–0.35 mC/cm
2
and 1 mC/cm
2
,
respectively (Beebe and Rose, 1988; Brummer and Turner, 1977; Rose and Robblee,
1990).
While it is important to prevent damage to stimulating electrodes, it is also impor-
tant to ensure that nearby tissue is not damaged by electrical stimulation. Fortunately,
the threshold for neural damage appears to be well above the charge injection limits
of platinum and iridium oxide. In one study, investigators applied excessive electrical
stimulation to rat retina using charge densities as high as 2.2 mC/cm
2
. They found that
electrical stimulation alone did not cause damage. Rather, the only damage they observed
was induced by mechanical pressure of the electrode array on the retina (Colodetti et al.,
2007; Ray et al., 2009).
Besides electrode damage, there are other safety concerns that designers of visual
prostheses must consider. Implantable materials must be chosen carefully in order to
prevent infection and inflammation that could arise from toxic substances. The materials
must also be resistant to corrosion caused by biological fluids. As is the case with other
implants, visual prostheses can be hermetically encased in titanium or ceramic or sealed
with thin films to protect the components from corrosion. Another safety consideration
for prostheses relates to heat damage. Heat dissipated by the electronic components
could potentially injure neural tissue. In the case of retinal prostheses, researchers have
19
found that avoiding direct contact between the electronics and the retina enables heat
to dissipate safely and prevents neural damage (Margalit et al., 2002; Piyathaisere et al.,
2001).
1.6 Cortical Prostheses
Cortical prostheses have the potential to restore vision to the greatest number of pa-
tients. When the retina or optic nerve undergoes permanent injury, as is the case with
the majority of irreversible blinding conditions, cortical prostheses are the only option.
Glaucoma is one example of a prevalent disease that could be treated by cortical pros-
theses. Glaucoma accounts for vision loss in nearly one third of Caucasians over the age
of 65 (Pardhan and Mahomed, 2002; Winter et al., 2007).
Stimulation of the visual cortex has been under investigation for more than 80 years
(Foerster, 1929; Krause and Schum, 1931). Early cortical stimulation experiments used
surface electrodes to evoke phosphenes. Because the electrodes were located atop the
pia and arachnoid membranes that overlay the visual cortex, relatively large currents
(milliamps) were required (Normann et al., 2009). Furthermore, phosphenes sometimes
flickered and had strange chromatic effects (Dobelle and Mladejovsky, 1974). In the worst
cases, the high currents needed to evoke phosphenes also induced seizures (Normann et al.,
2009). These undesirable effects led researchers to investigate the use of penetrating
electrode arrays (Fig. 1.5, right), a strategy known as intracortical microstimulation
(ICMS) (Bak et al., 1990; Schmidt et al., 1996). Proof of concept came when an acute
20
study in humans undergoing occipital craniotomies demonstrated that ICMS could elicit
phosphenes with 100 times less current than surface stimulation (Bak et al., 1990).
Despite the promising effects of ICMS, only one human subject has been implanted
with an ICMS array to date. The study took place in the mid-1990s, when a woman who
had been blind from glaucoma for 22 years was implanted with an array of 38 penetrating
microelectrodes for a period of four months. Results were encouraging: Phosphenes
could be elicited with currents as low as 1.9 A. Phosphene size could be controlled by
modulating stimulus amplitude, duration, and frequency. Electrodes spaced 500m apart
were able to evoke separate phosphenes when activated simultaneously, a 5x improvement
over surface stimulation (Schmidt et al., 1996).
Before a chronic cortical prosthesis can be realized, more animal and acute human
studies must be conducted. Biocompatibility and long-term stability must be demon-
strated in animal models before the FDA will even consider a human implant. Behavioral
experiments in non-human primates are already underway (Normann et al., 2009; Troyk
et al., 2003), but more data are needed to better understand the nature of cortically
evoked phosphenes. Finally, there is still much work left to be done on the hardware
and software components of a cortical prosthesis. Nevertheless, researchers are hopeful
that ICMS experiments in human volunteers will commence within the next few years
(Dagnelie, 2011; Normann et al., 2009).
21
1.7 Optic Nerve Prostheses
The optic nerve is ∼ 2 mm thick and contains roughly one million ganglion cell axons,
each less than 1 m in diameter (Mikelberg et al., 1989; Repka and Quigley, 1989). In
the visual pathway, the optic nerve is the one place where the visual field is represented
over a relatively small area (Maynard, 2001). Stimulation of the optic nerve offers the
advantage that the entire visual field can be covered with a small array of electrodes.
The first approach to an optic nerve prosthesis used cuff electrodes wrapped tightly
around the nerve. It had been shown previously that cuff electrode stimulation of the
sciatic nerve could be used to selectively recruit individual muscles (Grill and Mortimer,
1996). This prompted researchers in Brussels to investigate the feasibility of an optic nerve
prosthesis. They implanted a 59-year-old woman blinded by RP with a cuff electrode
containing four contacts. Stimulation elicited phosphenes with currents as low as 30 A.
Each stimulus produced a group of phosphenes, arranged in rows or clumps, whose size
depended on the stimulus amplitude and duration. Phosphene location differed with
each contact, indicating a coarse retinotopic organization of the optic nerve that could
be exploited by the prosthesis (Veraart et al., 1998). Despite having only four electrode
contacts (pixels), integration of a head-mounted camera enabled the patient to perform
tasks such as pattern recognition, object localization, and object discrimination. Each
task was reported to take tens of seconds (Brel´ en et al., 2005; Duret et al., 2006; Veraart
et al., 2003).
Since this study, researchers have investigated other approaches to optic nerve stimu-
lation. One group is using penetrating electrodes, rather than cuff electrodes (Chai et al.,
22
2008). Another group has proposed moving the stimulation site to the optic nerve head,
the location where the optic nerve exits the eye (Fang et al., 2006). To this point, all
work conducted by these groups has been in animal models. However, it is possible that
their efforts will eventually be translated into human prostheses.
The main drawback to optic nerve stimulation is that producing form vision could
prove difficult, if not impossible. Form vision would require selective and focal stimu-
lation of individual axons or small groups of axons, which is a tremendous challenge in
the optic nerve. It would also require the axons to be organized into a fine retinotopic
structure. While the Belgian study cited above described a coarse retinotopic mapping of
phosphenes, fine organization of mammalian optic nerves has not been reported (Horton
et al., 1979; Simon and O’Leary, 1991). This could be problematic for optic nerve pros-
theses in that stimulating two adjacent axons could produce phosphenes in two separate
parts of the visual field.
1.8 Retinal Prostheses
Because of the difficulties associated with cortical and optic nerve prostheses, several
groups are investigating implants that restore vision through electrical stimulation of the
retina (Rizzo et al., 2007). Retinal prostheses offer many advantages over their cortical
and optic nerve counterparts. Because the retina is the earliest part of the visual pathway,
stimulating retinal neurons enables implants to leverage the considerable amount of visual
processing that takes place in the retina. Furthermore, the retinotopic organization of the
retina means that activating a two-dimensional pattern of cells should elicit phosphenes
23
that resemble the same pattern. Finally, surgical procedures for implanting retinal pros-
theses are less complicated compared to other types of visual prostheses. Morbidity and
mortality rates are expected to be lower, and the transparency of the eye permits retinal
implants to be monitored over time.
There are three types of retinal prostheses, each named after the location in which
they are implanted. Subretinal prostheses are embedded between the retina and choroid
(the vascular layer of the eye; see Fig. 1.2), adjacent to the dying photoreceptor layer.
Epiretinal prostheses are attached to the inner surface of the retina, adjacent to the
ganglion cells. Suprachoroidal prostheses are implanted between the sclera and choroid.
Because suprachoroidal prostheses are still in early stages of development relative to the
other types of retinal prostheses, they are not covered in this chapter. The reader is
referred to a recent publication by Fujikado et al. (2011).
1.8.1 Subretinal Prostheses
The subretinal approach offers several advantages over epiretinal stimulation. While
epiretinal prostheses aim to stimulate ganglion cells, most subretinal prostheses target
bipolar cells. Stimulating bipolar cells allows more retinal processing to take place, since
bipolar cells are situated earlier in the visual pathway. Furthermore, direct stimulation of
bipolar cells avoids activation of ganglion cell axon bundles, which can create streak-like
phosphenes (Behrend et al., 2011; Nanduri et al., 2011; Weiland et al., 1999). Another
advantage of the subretinal approach is that the subretinal space holds the implant in
place without it having to be secured. In contrast, epiretinal electrode arrays must be
tacked to the retina, which is known to cause mechanical damage (Basinger et al., 2009).
24
There are two basic approaches to subretinal stimulation. The first uses an electrode
array to activate bipolar and/or ganglion cells, and the second uses a microphotodi-
ode array (MPDA). MPDAs are analogous to miniature solar cells—they detect incident
light and convert that light into electrical signals. MPDAs offer an elegant solution to
prosthetic vision because they eliminate the need for an external camera and image pro-
cessor. Furthermore, since MPDAs are implanted inside the eye, patients can change
gaze by moving their eyes rather than their heads.
Several groups are currently investigating subretinal approaches to electrical stimu-
lation (Rizzo et al., 2007). Three of those efforts are covered here—two using MPDAs,
and the third using an electrode array. Optobionics was the first company to attempt
an FDA-approved clinical trial for a subretinal prosthesis (Chader et al., 2009). Their
device, called the Artificial Silicon Retina (ASR), consists of approximately five thousand
microphotodiodes, each coupled to an iridium oxide electrode. The entire chip mea-
sures 2 mm in diameter and is only 25 m thick. The ASR is completely passive and
self-contained, requiring no external wires or power (Chow et al., 2004).
The Optobionics ASR was implanted in more than 40 patients with RP as part of
phase I and phase II clinical trials. While safety of the device was demonstrated, efficacy
was a more complicated matter. It was reported that patients implanted with the ASR
exhibited improvements in visual perception. However, those improvements were in areas
of the retina far from the implant site. The investigators concluded that the implant was
not electrically stimulating retinal neurons. Rather, the presence of the device in the
subretinal space was likely rescuing damaged retina through neurotrophic (i.e., growth
factor) effects (Chow et al., 2004; DeMarco et al., 2007). Theoretical calculations of the
25
ASR’s output capacity suggest that each microphotodiode can only output roughly one
nanoamp of current, at least one thousand times less than the amount needed to activate
retinal neurons (Palanker et al., 2005). Because Optobionics was unable to demonstrate
efficacy of the ASR, the company has ceased operations.
A more successful approach to a subretinal MPDA implant has been undertaken by a
German company, Retina Implant AG. Shown in Fig. 1.5 (center), their device consists
of 1550 microphotodiodes, each coupled to an amplifier and titanium-nitride electrode.
The MPDA is roughly 3 mm in diameter and 70m thick, covering approximately 15
∘ of
visual field (Besch et al., 2008; Stingl et al., 2013; Zrenner et al., 2011). The photodiodes
are externally powered, which solves Optobionics’ problem of not being able to supply
enough current to activate retinal neurons. Wireless power is supplied via inductive
coupling, with the secondary coil implanted subdermally behind the ear (Stingl et al.,
2013).
To date, Retina Implant’s device has been tested in patients for periods up to nine
months (Stingl et al., 2013). While hardware problems led to device failure in the first
few patients, testing in the remaining subjects has demonstrated the implant’s ability
to restore functional vision. Subjects have been able to successfully identify, locate, and
discriminate between sets of objects. One subject, who had been blind from RP for
several years, was able to distinguish between 16 different white letters (5–8 cm high)
on a black background with roughly 60% accuracy. He was able to recognize groups of
letters as words and could discern shades of gray with contrast differences of only 15%
(Zrenner et al., 2011).
26
Acute testing with Retina Implant’s MPDA demonstrated its feasibility as an effective
subretinal prosthesis. Before their device can be used chronically, some technical issues
must be addressed. Among these issues are hardware reliability and hermetic packag-
ing. In order to maximize the intensity of incident light, MPDAs must be encased in a
thin, transparent, hermetic film. At present, such films do not exist. Unless a suitable
material can be found, Retina Implant’s approach is unlikely to be viable in the long-
term. Nonetheless, the company is currently pursuing CE mark approval. If granted,
they would be allowed to market and sell the implant as an approved medical device in
Europe.
A third group to develop a subretinal implant has taken a markedly different approach.
Rather than use photodiodes for stimulation, this group uses a multielectrode array. Their
device, known as the Boston Retinal Implant, began as an epiretinal prosthesis. The
group switched to a subretinal approach because the surgical technique for implantation
is less invasive (although more complicated) and results in better biocompatibility (Rizzo
et al., 2011a,b). While the epiretinal version of the device was acutely tested in six human
patients (Rizzo et al., 2003a,b), the subretinal version has been tested only in animals.
The Boston Retinal Implant group is currently working to develop a human-grade
version of their device with more than 200 electrodes (from Rizzo presentation, IEEE
Engineering in Medicine and Biology Conference, 2011). They are investigating the use
of penetrating microelectrodes (see Fig. 1.5, right) coated with iridium oxide. This novel
approach will ensure that the tip of each electrode is in close proximity to ganglion and/or
bipolar cells. Furthermore, since the electrodes are fully contained within the retina,
electric fields will decay much slower than the fields of stimuli applied from epiretinal
27
electrodes. (Epiretinal electrodes lie within the vitreous, which is much more conductive
than the retina.) The penetrating electrode approach should enable more focal activation
of ganglion and bipolar cells with lower currents. However, there are increased safety
concerns with this approach that must be addressed before a human version of the implant
can be realized.
1.8.2 Epiretinal Prostheses
While subretinal prostheses have shown great potential, the epiretinal approach offers
a number of advantages. Epiretinal implantation surgeries are easier and are at risk of
fewer complications. The subretinal space is small and contains limited room for placing
electronics. Because of its proximity to the retina, any electronics placed in the subretinal
space are at risk of damaging the retina through thermal injury (Weiland et al., 2005). In
contrast, electronics for epiretinal prosthesis are implanted in the orbit of the eye, which
is relatively large and more efficient at dissipating heat. Finally, placing the electronics
and their hard packaging materials in the vitreous reduces the risk of mechanical damage
to the retina (Lakhanpal et al., 2003).
Several groups are currently investigating epiretinal prostheses. The three who have
made the most progress are Second Sight Medical Products, Inc. (SSMP), Intelligent
Medical Implants GmbH, and EpiRet GmbH. The latter two groups are based in Germany
and have completed clinical trials. The first group, SSMP, was awarded CE mark for its
Argus II epiretinal implant in 2011 and received FDA approval in early 2013, making it
the first commercially available retinal prosthesis in the world.
28
Basic operation of SSMP’s device is depicted in Fig. 1.4. A camera mounted to a pair
of sunglasses captures images of the surrounding environment. The images are processed
by custom hardware (not shown in the figure) and are transmitted wirelessly to the stim-
ulator implanted on the eye wall. The stimulator provides current to the electrode array,
which is attached to the inner surface of the retina. SSMP’s first-generation implant,
the Argus I, used an electrode array with 16 platinum microelectrodes, 260 or 520 m in
diameter, arranged in a 4× 4 grid. The Argus II array contains 60 electrodes, 200 m in
diameter, arranged in a 6× 10 grid (see Fig. 1.5, left).
The Argus I was implanted in six blind subjects with RP between 2002 and 2004. All
were able to perceive light with the device activated, even after being blind for several
years. Safety of the prosthesis was demonstrated in every patient with no unexpected
adverse events. (One subject had the device removed because of unrelated health prob-
lems (Chader et al., 2009).) Each patient has demonstrated the ability to perform simple
visual tasks, such as detecting objects, counting them, and discriminating between ob-
ject forms (Yanai et al., 2007). Some have reported using the device for more practical
purposes, such as following the crosswalk lines when crossing the street.
Several important findings came from the Argus I clinical trial. First, it was noted that
phosphenes produced from activation of single electrodes were generally small and oval-
or circular-shaped (de Balthasar et al., 2008; Horsager et al., 2009), although elongated
phosphenes have also been reported (Nanduri et al., 2008, 2011). Most were yellow or
white in color (de Balthasar et al., 2008; Horsager et al., 2009; Humayun et al., 2003;
Mahadevappa et al., 2005). Their size and brightness tended to increase with stimulus
amplitude (de Balthasar et al., 2008; Greenwald et al., 2009; Humayun et al., 2003;
29
Nanduri et al., 2008, 2012). Increasing stimulus frequency also increased brightness but
had a smaller effect on size (Nanduri et al., 2012). Visual acuity in one subject was
measured to be 20/3240, which is the highest possible resolution that could be restored
given the spacing of the electrodes on the array (Caspi et al., 2009).
Thresholds required to elicit phosphenes were almost always within the safe limit of
platinum (0.35 mC/cm
2
) and were usually well below that limit (Brummer and Turner,
1977; de Balthasar et al., 2008; Mahadevappa et al., 2005). Increasing the pulse duration
and/or stimulus frequency generally lowered stimulus thresholds (Horsager et al., 2009).
By measuring the distance between the electrodes and the retina using a non-invasive
imaging technology called optical coherence tomography, investigators found that thresh-
olds correlated with electrode-retina distance; the farther an electrode was from the retina,
the more current that was required to elicit a phosphene. Surprisingly, threshold did not
correlate with electrode size, electrode impedance, or retinal thickness (de Balthasar et al.,
2008). These findings illustrate the importance of implanting the electrode array in close
proximity to the retina, ideally within tens or hundreds of microns. This is currently a
challenge, since the array is held in place with a single tack. In the future, it may be
possible to use bioadhesives (Tunc et al., 2008) or tiny magnets to attach the array to
the retina.
One goal of the Argus I studies was to investigate how interactions between multi-
ple electrodes limit the resolution of the prosthesis. For example, can subjects see two
distinct phosphenes when two electrodes are activated simultaneously? The answer to
this question is still unclear. In some cases, patients reported seeing one phosphene; in
other cases, they reported two (Nanduri et al., 2008). This behavior likely results from
30
stimulating retinal ganglion cell (RGC) axon bundles, which is known to create streak-
like phosphenes that follow the paths of the axons (Behrend et al., 2011; Nanduri et al.,
2011; Weiland et al., 1999). Activating two electrodes lying along the same axon pathway
might produce one phosphene, while activating electrodes along different axon pathways
might produce two (Nanduri, 2011).
Having achieved success with the Argus I implant, SSMP set out to build a higher
resolution device called the Argus II. In addition to increasing the number of electrodes
from 16 to 60, the array was also made larger to cover more of the visual field (roughly
20
∘ ). Thirty subjects were implanted with devices between 2007 and 2009 as part of
a phase II clinical trial. Twenty-eight of those devices are still functioning (Humayun
et al., 2012). Fig. 1.8 shows the Argus II Retinal Prosthesis System. The sunglasses
(left) contain the camera and primary inductive coil. They interface with the Visual
Processing Unit (right), which filters the camera’s images, downsamples them to a 6× 10
grid, and creates a series of stimuli based on pixel brightness values and look-up tables
customized for each user (Ahuja et al., 2011).
Results from the Argus II clinical trial have been encouraging. All 30 patients have
been able to perceive light during electrical stimulation. Safety was demonstrated with
no unexpected adverse events. Visual acuity was reliably restored in multiple subjects,
with the best patient scoring 20/1260 (Humayun et al., 2012). To assess patients’ abilities
to perform spatial-motor tasks, they were asked to locate and touch a white square on a
black computer screen. With the system on, more than 90% of the subjects performed
this task with greater accuracy and repeatability than with the system off (Ahuja et al.,
2011).
31
Figure 1.8: The Argus II epiretinal prosthesis (Second Sight Medical Products, Inc.). Left:
The sunglasses contain a camera for capturing images and a coil for transmitting signals to the
implant. Right: The Visual Processing Unit filters the camera’s images and converts them to a
series of stimuli. (Image courtesy of Second Sight Medical Products, Inc.)
Tests have also demonstrated patients’ abilities to read high-contrast letters. An
experiment with 22 subjects showed they could correctly identify a set of eight letters
with 72.6% success, compared to 16.8% success with the system off. However, letter
recognition often took tens of seconds (da Cruz et al., 2010). Some subjects were also
able to read three-letter words and short sentences (Sahel et al., 2011; Stanga et al.,
2010). Again, reading speed was slow but improved with practice (Sahel et al., 2011).
Interestingly, it was reported that several Argus II subjects were able to perceive
phosphenes of different colors. A test with nine subjects demonstrated their abilities to
see eight different colors (red, orange, yellow, green, blue, pink, gray, and white). Results
were repeatable. The perceived color depended on which electrode was stimulated and
the stimulus parameters that were used (Stanga et al., 2011).
In order to create meaningful form vision in human prosthesis patients, repetitive
stimuli will need to be delivered. One Argus II study investigated the effect of repetitive
32
stimulation on phosphene brightness. Subjects reported phosphenes that were initially
bright but faded with time. There were two components to the fading: a fast com-
ponent lasting hundreds of milliseconds, and a slow component lasting several seconds
(Perez Fornos et al., 2010). This behavior is undesirable and could limit temporal reso-
lution.
The Argus clinical trials represent the largest study of retinal prostheses ever con-
ducted. Although there is significant room for improvement, results from the trials
demonstrate the true feasibility of high-resolution prosthetic vision. The main prob-
lems that need to be addressed relate to phosphene persistence/fading, positioning the
array on the retina, and avoiding stimulation of retinal ganglion cell axons. Overcoming
these problems will likely improve the resolution of devices for future patients. Further-
more, use of an intraocular camera would allow patients to view their environments more
naturally without having to scan their heads across the visual field.
Intelligent Medical Implants (IMI) is the second of three companies developing an
epiretinal prosthesis. Their device operates similarly to SSMP’s Argus implant. A camera
mounted to a pair of eyeglasses captures images that are processed by a wearable unit
called the Pocket Processor. The processed images and power are transmitted wirelessly
to the implant. While the standard method of inductive coupling is used to carry power,
infrared light is used to transmit visual information. (The advantage of using infrared
light is that eyelid closing causes an interruption of data transfer, the same way it does in
normally sighted individuals.) Stimulation is carried out by a 49-electrode array tacked
to the inner surface of the retina (Hornig et al., 2008).
33
A major component of IMI’s prosthesis is an image processing algorithm called the
Retina Encoder. It is responsible for predicting the firing pattern of ganglion cells in
response to a given image. The encoder takes the visual information captured by the
camera and converts it to stimulation commands that emulate those firing patterns.
Since each patient requires different parameters for optimal vision, the Retina Encoder
can be tuned individually for each patient (Hornig et al., 2008).
IMI began implanting their device in humans in 2005. Early reports indicated that the
devices were well tolerated by the body. Thresholds fell within the safe limit of electrical
stimulation and remained relatively stable over time (Richard et al., 2009). In regards to
single-electrode stimulation, patients were able to distinguish between phosphenes elicited
by different electrodes. Multielectrode stimulation enabled patients to recognize simple
patterns, such as a horizontal bar (Richard et al., 2007). While it is likely that IMI
has obtained many more results in human patients, those results have not been shared
publicly.
EpiRet is the third company developing an epiretinal prosthesis. Their latest-gener-
ation implant is called the EPIRET3. Unlike the other epiretinal implants, which place
the electronics on the eye wall, EpiRet’s device is designed to fit entirely inside the eye
(excluding the camera and image processor). This eliminates the need for a transscleral
(across the eye) cable that connects the stimulator to the electrode array (Klauke et al.,
2011). The EPIRET3 electrode array contains 25 penetrating iridium oxide electrodes,
each 100 m in diameter and 25 m high (Roessler et al., 2009).
EpiRet’s device was implanted in six human subjects in 2006. By study design,
the implants were left in place for only four weeks. During this time, studies showed
34
that stimulation thresholds required to evoke phosphenes were within the safe limit of
electrical stimulation. The ability of a stimulus to evoke perception depended more on
pulse duration than current amplitude or total charge. By activating different pairs
of electrodes, investigators found that subjects could discriminate between unique pairs
(Klauke et al., 2011).
EpiRet’s major challenge is to develop a device that can be chronically implanted
inside the eye. Power requirements make this difficult, as does hermetic sealing of the
electronics. Nevertheless, as electronics become smaller and more power-efficient, it is
possible that EpiRet will one day achieve their goal of a fully implantable prosthesis.
1.9 Retinal Prosthesis Implications from Animal Studies
Despite the successes reported in retinal prosthesis patients, subject performance is highly
variable. This is especially true of more complex tasks, such as letter reading. There are
several possible explanations for this variability: 1) Location of the electrode array and
its distance from the retina is inconsistent across subjects; 2) The type and severity
of RP differ among subjects, causing variable levels of retinal degeneration; 3) There
may be varying levels of cortical reorganization in the higher visual centers; and 4) It is
possible that adjacent electrodes activate overlapping regions of cells, causing distortion
in the perceived signal. All these factors are difficult, if not impossible, to control in the
clinic. Consequently, several groups are studying electrical stimulation of animal retina
in the laboratory setting, enabling them to control each factor independently. Findings
from animal studies can be directly applied and tested in human subjects. This section
35
reviews some of those findings and their implications for retinal prostheses. For a more
comprehensive review, we refer readers to a publication by Freeman et al. (2011).
1.9.1 Direct vs. Indirect Activation of Ganglion Cells
Visual perception in prosthesis patients depends largely on which cell type(s) are stim-
ulated. Animal experiments have shown that when ganglion cells are activated directly,
they generally fire one action potential per stimulus pulse (Fried et al., 2006; Sekirn-
jak et al., 2008). Alternatively, ganglion cells can be activated indirectly by stimulating
presynaptic retinal neurons, such as bipolar cells. Exciting bipolar cells causes post-
synaptic ganglion cells to fire bursts of spikes (Jensen et al., 2005a). These bursts most
likely resemble the neural code that the ganglion cells would generate in response to a
spot of light. For this reason, it could be argued that retinal prostheses should target
bipolar cells; doing so would produce more natural spiking patterns in the ganglion cells.
While this may be true, one difficulty with stimulating bipolar cells is that responses can
become desensitized (Ahuja et al., 2008a; Freeman and Fried, 2011; Jensen and Rizzo,
2007). When a bipolar cell is targeted with a train of stimuli, the first pulse in the train
might evoke several action potentials in a postsynaptic ganglion cell, whereas a later pulse
might evoke only one (or even none). This effect has been attributed to amacrine cell
inhibition, as well as other unidentified mechanisms. It may limit the temporal resolution
of indirect activation and might also explain the phenomenon of fading phosphenes in
human patients (Freeman and Fried, 2011).
36
In contrast to indirect activation, direct activation experiments have shown that gan-
glion cells can follow high-frequency stimulation (i.e., fire one spike per pulse) at frequen-
cies up to 500 Hz (Ahuja et al., 2008a) (although desensitization in response to direct
activation has also been reported (Sekirnjak et al., 2006; Tsai et al., 2011)). When cou-
pled with a retina encoder (Eckmiller et al., 2005; Nirenberg and Pandarinath, 2012),
direct stimulation might be able to elicit “natural” firing patterns in ganglion cells. The
main problem, however, is that direct stimulation of ganglion cells also leads to activa-
tion of their axons, which are arranged in bundles. Stimulating one axon bundle can
activate hundreds of ganglion cell bodies, creating a streak-like response. This limits the
spatial resolution that can be achieved with direct activation. Whether novel stimulation
paradigms can be used to avoid activation of passing axons is still unknown.
Selective targeting of individual cell types can be achieved with different stimulus
waveforms. Animal experiments have shown that when using rectangular pulses, such
as the one in Fig. 1.7, short pulse widths (i.e.,≤ 0.2 ms) preferentially target ganglion
cells (Behrend et al., 2011; Fried et al., 2006; Jensen et al., 2005b; Sekirnjak et al., 2006).
Longer pulse widths target both ganglion and bipolar cells (Behrend et al., 2011; Jensen
et al., 2005b; Margalit and Thoreson, 2006). To selectively activate bipolar cells, 10–25-Hz
sinusoidal waveforms can be delivered (Freeman et al., 2010). Most retinal prostheses are
currently using rectangular pulses of relatively long durations. Therefore, it is likely
that both ganglion and bipolar cells are being activated. Future generations of retinal
prostheses will conceivably leverage the stimulation strategies learned from animal studies
in order to selectively activate one cell type or the other. It remains to be seen whether
that cell type will be ganglion or bipolar cells.
37
1.9.2 Visual Acuity
Another important issue for retinal prostheses relates to the quality of vision that can be
restored to patients. One way to measure this is through visual acuity (i.e., spatial reso-
lution). Acuity is highly dependent on the number of ganglion cells that are stimulated
by a single electrode. Stimulating fewer ganglion cells should cause an increase in spatial
resolution.
Animal studies are well suited for investigating the spatial resolution that can be
achieved through electrical stimulation of the retina. One study in salamander found that
the smallest area that could be activated by a single electrode was 150 m in diameter
(corresponding to a visual acuity of 20/660), even for electrodes as small as 10 m in
diameter (Behrend et al., 2011). However, another study in monkey found that single
ganglion cells could be activated with 10-m-diameter electrodes (Sekirnjak et al., 2008).
Although these results conflict, they do suggest that spatial resolution can be improved
in future prosthesis patients. (The best reported visual acuity to date is 20/546 in one
Retina Implant subject (Stingl et al., 2013).) Doing so will likely require the use of smaller
electrodes that are situated closer to the target cells. Stimulation of axon bundles must
also be avoided.
Even if these conditions are met, there are factors that may place an upper limit
on spatial resolution. The retinal and cortical reorganization that accompany retinal
degenerative disorders may limit the ability of humans to regain their vision. Furthermore,
electric field interactions between neighboring electrodes can impede the resolution of
38
multielectrode stimulation (Horsager et al., 2011). Because of these issues, it is possible
that vision in prosthesis patients can be restored only to a certain extent.
1.9.3 Effects of Retinal Degeneration
One explanation for the variable thresholds observed in human prosthesis patients might
relate to the conditions of their retinas. Retinal reorganization can differ among patients,
depending on the type and severity of RP (Humayun et al., 1999b; Santos et al., 1997).
It is possible that patients with more advanced stages of RP have higher stimulation
thresholds.
Animal models of RP have enabled scientists to investigate the potential effects of
RP on retinal ganglion cell electrical thresholds. These models, most commonly found
in rodents, carry transgenic mutations similar to those found in individuals with RP.
The most widely studied mouse models, rd1 (Keeler, 1924; Pittler et al., 1993) and rd10
(Chang et al., 2002), carry mutations that resemble recessive forms of RP. In rats, the
most widely studied models are P23H (Machida et al., 2000) and S334ter (Liu et al.,
1999), which both resemble autosomal dominant RP. Each of these four models differs in
its rate of disease progression and effect on each retinal cell type. For example, rd10 mice
exhibit complete loss of photoreceptors over time, yet ganglion cells remain intact (at
least up to nine months of age) (Mazzoni et al., 2008). P23H and S334ter rats, however,
have been shown to lose ganglion cells (and photoreceptors) with age, much like humans
with RP (Chan et al., 2011; Garc´ ıa-Ayuso et al., 2010; Kolomiets et al., 2010). For this
reason, rat models may be better suited for studying how retinal degeneration affects
electrical thresholds.
39
There have been conflicting reports as to whether the electrical thresholds of ganglion
cells change as a result of retinal degeneration. Most studies have reported that thresholds
in degenerate retina are significantly higher than in normal retina, in mice (Jensen and
Rizzo, 2008; Suzuki et al., 2004), rats (Chan et al., 2011; Jensen, 2012), rabbits (Humayun
et al., 1994), and even humans (Humayun et al., 1999a; Rizzo et al., 2003a). However,
a recent study by Sekirnjak et al. (2009) reported that ganglion cell electrical thresholds
did not significantly differ between normal and degenerate rat retinas. In this study, the
authors used very small electrodes (7–16m) that were in close contact with the ganglion
cell layer. Electrodes used by the other studies were larger and farther away from the
ganglion cells, as is the case with clinical retinal prostheses. Despite these conflicting
results, the Sekirnjak study provides strong motivation for investigating new methods for
achieving close coupling between electrodes and the retina.
1.10 Gene Therapy Treatments for Outer Retinal
Degenerations
To this point, we have focused on techniques that use electrical stimulation to restore
vision to the blind. However, fundamentally different approaches are under investigation
and may hold promise for the future. Several of these approaches involve gene therapy,
in which a viral vector is used to deliver a gene to diseased cells. In most cases, recom-
binant adeno-associated virus (AAV) is used because it is safe (i.e., non-pathogenic and
non-toxic), effective at transducing retinal cells, and leads to long-term gene expression
(Grimm and Kay, 2003). More than 600 patients have already participated in at least
40
48 clinical trials involving AAV without a single serious adverse event (Retracing events,
2007).
Leber congenital amaurosis (LCA) is a rare form of RP in which mutations to any of
at least 14 genes cause loss of vision and eventual blindness (den Hollander et al., 2008).
One of those genes, RPE65, is the subject of gene replacement therapy in LCA patients.
A mutation in RPE65 affects the viability of retinal pigment epithelial cells (see Fig. 1.3)
and can lead to blindness in children and young adults. In order to treat individuals
with this condition, scientists are delivering non-mutated RPE65 to their retinal pigment
epithelia. RPE65 is packaged into an AAV vector and injected into the subretinal space.
Using this approach, scientists have been able to improve visual function in many LCA
patients. Clinical trials are still ongoing (Cideciyan, 2010; Stein et al., 2011).
While the LCA trials have shown promise, the disease constitutes a rare form of RP
that affects only one in 81,000 people worldwide (Stone, 2007). Furthermore, the RPE65
mutation is only present in roughly 6% of individuals with LCA (Stein et al., 2011). It
therefore accounts for less than 0.3% of all RP cases, limiting its usefulness to the general
RP population.
One gene therapy approach, which could benefit all individuals with RP, involves the
use of optogenetics to restore light sensitivity to surviving retinal cells. Optogenetics is
an emerging field that integrates optics and genetics in order to control specific events
in targeted cells, such as the firing of action potentials (Deisseroth, 2010). The most
widely used optogenetics tool, channelrhodopsin-2 (ChR2), is a bacterial cation channel
that is gated with light (Boyden et al., 2005; Nagel et al., 2003). When ChR2 absorbs
blue light, it undergoes a conformational change that permits cations (e.g., Na
+
and K
+
)
41
to flow across the electrochemical gradient. Cells expressing ChR2 in their membranes
therefore depolarize in response to blue light and fire action potentials if the magnitude
of the depolarizations is large enough.
Investigators are currently exploring the use of ChR2 for imparting light sensitivity to
retinal neurons in animal models of retinal degeneration. Studies have shown that ChR2
delivered to ganglion and/or bipolar cells via AAV restores light sensitivity in rd1 mouse
models (Bi et al., 2006; Doroudchi et al., 2011). Stable expression was observed for at
least 6–10 months. In one study, treated mice demonstrated the ability to navigate a
water maze (Doroudchi et al., 2011).
Although promising, ChR2 is not yet ready for human gene therapy. Unnaturally
bright light is needed in order to activate the channel. Light sensitivity can be increased
by mutating the gene; however, this often has negative impacts on the kinetics of the
channel (Berndt et al., 2008; Lin et al., 2009). Additionally, more animal studies are
needed to confirm the safety of virally delivered optogenes before human trials commence.
AAV has been shown to be safe for human gene transfer, but data on immune reactions
to ChR2 in non-human primates are still lacking (Busskamp and Roska, 2011). Assuming
safety and reasonable light sensitivity can be proven, it is conceivable that optogenetic
gene therapy approaches will be used in the future for treating outer retinal degenerations.
1.11 Conclusion
Despite the progress made in recent years, visual prostheses still face significant chal-
lenges. Fortunately, most of these challenges are engineering problems that can be solved
42
with technical advances. The main goal at this time is to improve visual acuity in pros-
thesis patients, which is still poor compared to normally sighted individuals. Acuity is
expected to increase in future generations of prostheses that employ greater numbers
of electrodes and smarter stimulation strategies. However, factors such as retinal and
cortical remodeling may inherently limit the quality of vision that can be restored.
There are multiple approaches to artificial vision, and it is unknown whether electrical
prostheses will be the best solution. Optogenetic techniques have shown promise in
animal models and may one day be used in human therapies. Other approaches are
under development, including neurotransmitter-based prostheses (Finlayson and Iezzi,
2010), retinal sheet transplantation (Radtke et al., 2004), and stem cell therapies (Lund
et al., 2006). There are advantages and disadvantages to each approach; given the many
forms of retinal blindness, it is likely that different forms will require different treatments.
Visual prostheses have advanced tremendously since the days of Foerster, Brindley,
and Dobelle. In particular, developments in the past decade have demonstrated the
practicality of retinal prostheses as a viable treatment for outer retinal degenerations.
While there is still much work to be done, early results have been encouraging. Patients
who were otherwise blind have been able to perform complex tasks such as letter reading.
This is quite an impressive achievement and is truly pioneering in the field of artificial
vision. While only a select few have benefited from the technology so far, hundreds, if
not thousands, are expected to benefit in years to come. With continued efforts, visual
prostheses have the potential to restore independence to blind individuals around the
world.
43
1.12 Thesis Overview
Results from the Argus I and II clinical trials have been encouraging, though there remains
significant room for improvement (see Section 1.8.2). Patient performance will likely
improve as prostheses begin to incorporate larger numbers of electrodes that are placed
in closer contact with the retina. However, it may be years before we see a clinical device
that contains hundreds or thousands of electrodes. Furthermore, if the percept elicited
by a single electrode cannot be controlled (and is not focal), patient performance will not
scale with the number of electrodes.
This thesis aims to attain better control of the percepts evoked by epiretinal electrical
stimulation. It addresses four specific problems that arose during clinical testing with
Argus I and II subjects:
1. High Thresholds. On average, nearly half of the electrodes in each Argus II patient
are unable to evoke phosphenes without exceeding the electrochemical safety limit
(Humayun et al., 2012).
2. Elongated Phosphenes. Rather than creating a punctate percept, many stimulat-
ing electrodes evoke elongated phosphenes. In some cases, these phosphenes subtend
tens of degrees of visual angle (Nanduri, 2011).
3. Electrode Interactions. During multielectrode stimulation, interactions between
neighboring electrodes have a significant influence on the quality (shape and bright-
ness) of the evoked percept(s) (Horsager et al., 2010, 2011; Nanduri et al., 2008).
44
4. Patient Variability. Subject performance is highly variable, especially with complex
vision tasks such as letter reading. It is possible that this arises from varying degrees
of retinal degeneration among patients.
Taken together, these problems make it difficult, if not impossible, to produce mean-
ingful form vision in prosthesis patients. Form vision requires stimulating with multiple
electrodes to activate ganglion cells arrayed in a precise pattern. To investigate the extent
to which this is possible, I used an in vitro model of the retinal prosthesis to measure the
spatial properties of RGC activity during epiretinal electrical stimulation. My specific
hypothesis was that by tuning stimulus parameters, multielectrode pattern stimulation of
normal and degenerate retina could activate focal regions of ganglion cells that conform
to the stimulus pattern. Indeed, I found that simply modifying the shape of the stimulus
waveform was enough to lower stimulus thresholds, produce focal responses, and limit
electrode interactions. Furthermore, I showed for the first time that focal stimulation
was possible in degenerate retina. Results from my work are readily translatable to the
Argus II prosthesis, as no physical changes to the device’s design would be required.
1.12.1 Basic Approach
Most electrophysiology studies for retinal prostheses use in vitro preparations of animal
retina. Isolated retina can be maintained in culture and studied for several hours. Mount-
ing the retina on a multielectrode array (MEA) enables RGC population activity to be
recorded during electrical stimulation. This provides precise temporal measurement of
RGC responses properties; however, spatial resolution is quite limited.
45
Because we are interested in measuring the spatial properties of RGC activation, our
group developed an optical method to record from populations of RGCs in isolated retina.
Calcium imaging reports RGC activity during stimulation with one or more MEA elec-
trodes (Behrend et al., 2009). We recently demonstrated the power of this technique for
identifying the spatial properties of RGC activation during electrical stimulation (Behrend
et al., 2011).
Our method relies on being able to selectively label large populations of RGC somata
with fluorescent calcium indicator. In the past, we accomplished this by retrogradely
loading synthetic calcium dye via the cut optic nerve (Behrend et al., 2009). This ap-
proach labeled nearly all ganglion cells in amphibian retina, yet the same approach did
not succeed in mature mammalian retina due to dye extrusion. Thus, a major goal of
my work was to develop a method for labeling mammalian ganglion cells with calcium
indicator. This would enable us to use genetic animal models of retinal degeneration for
studying the influence of degeneration on electrically stimulated RGCs. I was able to
accomplish this by designing a custom AAV vector to selectively transduce ganglion cells
in adult rat retina with genetically encoded calcium indicators. The large fluorescence
signals provided by these indicators enabled us for the first time visualize the retina’s
response to electrical stimulation in real time. This greatly expedited the process of test-
ing different stimulation strategies and ultimately enabled me to study multielectrode
pattern stimulation of the retina.
46
1.12.2 Summary of Results
This thesis focuses on addressing problems that arose during clinical testing with retinal
prosthesis patients (see above). My first goal was to investigate a method for lowering
stimulus thresholds. Prior studies in cochlear implant had shown that adding an inter-
phase gap between the two phases of biphasic current pulses reduced perceptual thresholds
(Carlyon et al., 2005; McKay and Henshall, 2003). I tested the effect of interphase gap
duration on RGC electrical thresholds using the same electrode size (200 m) and pulse
width (0.46 ms) as the Argus II (Weitz et al., 2011). Experiments were performed in
salamander retina. Results showed that with an IPG duration equivalent to the pulse
width, RGC thresholds became ∼ 20% lower than when no gap was used. Gap lengths
longer than 0.46 ms further decreased thresholds, but only marginally. Gaps shorter than
0.46 ms were less effective at stimulating RGCs. To validate these results, I generated a
Hodgkin-Huxley-type model and found its predictions to align well with the experimental
data. Finally, I worked with Second Sight Medical Products, Inc. to test the effect of IPG
in retinal prosthesis patients. Results from five subjects demonstrated that interphase
gap has a similar effect on perceptual thresholds as it does on RGC electrical thresholds.
In a second study, I systematically investigated the effect of stimulus pulse width
on RGC response patterns. Recent human and animal work from our group showed
that elongated percepts arose from activation of ganglion cell axon bundles (Behrend
et al., 2011; Nanduri et al., 2011). Two potential solutions to this problem would be to
preferentially stimulate ganglion cell bodies over their axons or to selectively stimulate
inner retinal cells (i.e., bipolar cells). I hypothesized that stimulus pulse width might
47
provide a means for achieving such selectivity—prior studies suggested that pulse width
affects the target of epiretinal electric stimulation: short pulses excite RGCs directly,
while longer pulses target bipolar cells (Freeman et al., 2010; Fried et al., 2006; Greenberg,
1998; Jensen et al., 2005b; Margalit and Thoreson, 2006; Shah et al., 2006). Despite
these findings, each study relied on traditional electrical recording methods that could
not precisely measure the pattern of activated cells. Furthermore, they were conducted
in normal (non-diseased) retina, and results were inconclusive.
The nature of our calcium imaging approach lends itself well to studying the spatial
extent of RGC activation. We can visualize responses of cells both local and distant to
the stimulating electrode. By sweeping pulse width across more than three orders of
magnitude (from 0.06 to 100 ms), I was able to determine the amount of somatic (local)
and axonal (distant) activation at each pulse width. With 200-m-diameter electrodes, I
found that very short pulses (0.06–0.1 ms) selectively stimulated RGC somata at thresh-
olds roughly 40% lower than their axons. As pulse width was increased from 0.1 to 4 ms,
this selectivity gradually diminished. Interestingly, selectivity was regained with a pulse
width of 8 ms. Doubling pulse width to 16 ms yielded very little axonal activation. Pulses
25 ms and longer produced no measurable axon responses, resulting in focal activation.
To elucidate the mechanism of this behavior, I used synaptic blockers to measure the
contribution from inner retinal stimulation at each pulse width. I found that 0.06-ms
pulses activated only ganglion cells, while pulse widths between 0.1 and 4 ms activated
both RGCs and inner retinal cells. At 8 ms, RGC thresholds in the presence of block-
ers nearly doubled, suggesting a large contribution from inner retinal stimulation and
explaining why selectivity was regained at this pulse width. Pulses longer than 16 ms
48
did not stimulate ganglion cells when blockers were present, indicating that long pulses
activate inner retinal cells, which synaptically drive RGCs.
To test whether these effects persisted with smaller electrodes, I delivered three pulse
widths (0.06, 1, and 25 ms) from 75- and 30-m-diameter electrodes. Results were consis-
tent with those obtained from 200-m electrodes: 0.06- and 1-ms pulses activated axons,
but only 0.06-ms pulses provided selectivity for ganglion cell somata. 25-ms pulses pro-
duced focal responses. As expected, the size of the response area decreased with electrode
size, though only to a certain extent.
I concluded my study of single-electrode stimulation by mapping thresholds in the
S334ter-line-3 rodent model of retinal degeneration. I found that the size and shape of
RGC activation patterns were similar to those from wild-type (WT) retina. Thresholds
for 0.06-ms pulses were also similar to WT but increased by 10–20% for 1-ms pulses and
200% for 25-ms pulses. This suggests that bipolar cell thresholds become higher in RD
retina, while direct ganglion cell thresholds remain unchanged. Results from this study
may explain the inconsistent findings in the literature as to whether thresholds become
elevated during retinal degeneration.
Having shown that 25-ms pulses achieve focal activation of RGCs, I set out to de-
termine whether multielectrode stimulation could produce controlled patterns of RGC
activation. Indeed, I was able to form lines parallel and perpendicular to axon bundles
without activating the axons. Electrode interactions were negligible when using 30-m
electrodes on a 75-m pitch and 75-m electrodes on a 150-m pitch, regardless of stim-
ulus amplitude. I also demonstrated the ability to activate patterns of RGCs forming
letters with a critical detail size of 0.26
∘ of visual field. This size corresponds to 9-mm-tall
49
letters viewed from typical reading distance and a Snellen acuity of 20/312. Collectively,
my results suggest that multielectrode stimulation in epiretinal prosthesis patients has
the potential to restore meaningful form vision.
1.12.3 Structure of Thesis
Chapter 2 describes the experimental setup and methods common to all experiments.
Extra detail is devoted to multielectrode array and AAV vector design. Methods for data
collection and processing are also discussed.
Chapter 3 describes how interphase gap can be used to lower RGC electrical thresholds
in animal retina and perceptual thresholds in human subjects. A computational model
that validates these results is provided.
Chapter 4 presents a detailed study of how pulse width can be used to control the
pattern of excited cells. Effects of electrode size and retinal degeneration are explored.
Successful multielectrode pattern stimulation is also demonstrated.
Chapter 5 discusses the significance of my findings and includes suggestions for future
experiments.
50
Chapter 2
Experimental Design and Methods
To inform the design of retinal prostheses, several groups are using in vitro retina prepara-
tions to measure population activity of retinal ganglion cells during electrical stimulation.
Though these preparations can differ vastly in their design, all involve an extracellular
stimulation electrode for delivering current to the tissue. What makes each preparation
unique is how RGC responses are measured. The following section discusses common
modalities for recording neural activity from electrically excitable cells.
2.1 Probing Activity in Electrically Excitable Cells
There are two distinct approaches for monitoring neural activity. The first uses electrodes
to measure changes in membrane potential that occur during spiking. This technique
enables precise temporal measurements—single spikes and even subthreshold activity can
be recorded. The second approach utilizes optical probes to obtain an indirect measure
of electrical activity. Optical imaging permits recordings to be made with high spatial
precision. Both approaches are discussed below.
51
2.1.1 Electrical Recording
Cell membranes contain specialized channels that allow specific ions (e.g., Na
+
and K
+
)
to pass through. Some channels remain open at all times, while others are gated by
membrane potential, neurotransmitters, temperature, or mechanical force. When neurons
are at rest, the majority of gated channels is closed. The concentration of Na
+
ions outside
the cell is greater than the concentration inside the cell; the opposite is true for K
+
ions.
The K
+
concentration gradient creates an electric potential across the cell membrane of
approximately− 70 mV.
An action potential occurs when a neuron is excited by an external stimulus. Na
+
channels open and allow Na
+
ions to diffuse down the electrochemical gradient (from the
outside to the inside of the cell). This depolarizes the cell and causes more voltage-gated
Na
+
channels to open, resulting in a positive feedback loop that shifts the membrane
potential to a positive value. Eventually, the cell membrane is repolarized by the opening
of voltage-gated K
+
channels and inactivation of Na
+
channels. This entire process can
happen within 1 ms.
Electrical recording techniques measure changes in voltage and/or current that occur
when neurons spike. Recordings can be made intracellularly or extracellularly. Intracellu-
lar recordings involve a fine-tipped microelectrode or micropipette that is inserted inside
a cell or patched onto its membrane. This enables precise measurement of membrane
potential or ionic currents during spiking. Furthermore, current through the cell can be
driven externally, or the cell membrane can be clamped to a specific voltage. Chemicals
and dyes can also be loaded into the cell via iontophoresis or microinjection. Despite these
52
advantages, intracellular techniques are generally limited to recording from a single cell
at a time. For this reason, other approaches are often preferred for measuring population
activity of neurons.
Multielectrode arrays (Fig. 2.1, left) offer a means to record simultaneous activity
from a large number of cells. These arrays contain tens or hundreds of microelectrodes
that measure extracellular potentials relative to a common reference electrode. Because
the potentials decay rapidly with distance, an electrode can detect signals from cells only
up to∼ 80 m away (Segev et al., 2004). The small signals (on the order of microvolts)
must then be amplified by specialized hardware. Many retinal prosthesis studies employ
multielectrode arrays, since activity can be recorded from dozens of RGCs simultaneously
(Ahuja et al., 2008a; Grumet et al., 2000; Sekirnjak et al., 2008; Stett et al., 2000). Placing
the retina ganglion cell-side-down on an MEA enables a tight interface between the RGCs
and electrodes (Fig. 2.1, right).
Figure 2.1: Left: A multielectrode array used to record spikes from a large fraction of RGCs
in a retinal patch (Segev et al., 2004). Right: Cross section of a retina on top of an MEA (Litke
et al., 2004). The ganglion cells are within a few microns of the recording electrodes.
53
Although MEAs are well suited for measuring temporal response properties of RGCs,
spatial resolution is limited by the area covered by the electrodes, the spacing between
each electrode, and the proximity of RGCs to electrodes. State-of-the-art high-density
MEAs (30-m electrode spacing) can record from nearly all ganglion cells in a small
patch of retina, yet the number of electrodes is limited to∼ 500 (Field et al., 2010; Litke
et al., 2004; Segev et al., 2004). This number should increase as high-resolution CMOS-
based MEAs become available (Eversmann et al., 2003; Ferrea et al., 2012; Hutzler et al.,
2006); however, MEA recordings have other limitations. Complex spike sorting algorithms
(Segev et al., 2004) and stimulus artifact subtraction techniques (Sekirnjak et al., 2006)
are needed to assign spikes to individual RGCs. These techniques are biased toward iden-
tifying primarily larger neurons that are closer to the recording electrodes (Briggman and
Euler, 2011; Segev et al., 2004). Cell position can only be inferred through triangulation,
rather than directly observed.
2.1.2 Optical Recording
Optical recording techniques overcome many of the limitations inherent to traditional
electrical recording methods. These techniques usually require cell loading with fluores-
cent or luminescent probes. The probes report cellular activity via changes in brightness
and/or wavelength. The changes are captured by photodetectors, such as a CCD array,
which can image hundreds or thousands of cells at a time.
There are two types of neural activity probes: small-molecule dyes and genetically
encoded indicators. Small-molecule probes are synthetic chemicals containing an ion
54
chelator bound to a chromophore. Genetically encoded indicators are proteins that con-
tain an ion-binding domain fused to green fluorescent protein (GFP) or one of its variants.
Both types of probes operate in a similar fashion: upon binding to their target molecule,
they undergo a conformational change that modulates their excitation and/or emission
properties.
Optical probes can measure a wide range of cellular activity, including calcium con-
centration, membrane voltage, and pH. Calcium indicators are the most commonly used
neural activity probes and are discussed in the next section. Voltage sensors have been
around for nearly three decades (Homma et al., 2009) though have only recently be-
gun to achieve the performance necessary for common use (Looger and Griesbeck, 2011).
Early generations of voltage sensors were synthetic dyes that suffered from poor signal-to-
noise ratios (SNRs), especially in mammalian preparations (Peterka et al., 2011). They
were difficult to load into cells and were often toxic (Looger and Griesbeck, 2011). For
these reasons, voltage-sensitive dyes have achieved little practical use. More recently,
genetically encoded voltage indicators have begun to overcome the problems inherent to
voltage-sensitive dyes. The latest-generation voltage indicators, Arch and Arch(D95N),
have shown the ability to image single action potentials and subthreshold depolarizations
in cultured neurons with acceptable SNR (Fig. 2.2) (Kralj et al., 2011).
Although voltage indicators provide a direct measurement of membrane voltage, the
rapid signals that occur during action potential firing (on the order of 1 ms) make it
difficult to image simultaneous activity from neuronal populations. Doing so requires
both high temporal and spatial resolution, which limits the amount of photons that can
55
be captured. For this reason, voltage imaging is usually performed on cellular aggregates
rather than on individual cells in a population (Looger and Griesbeck, 2011).
Figure 2.2: Left: Fluorescence micrograph of a rat hippocampal neuron expressing Arch(D95N),
showing Arch(D95N) fluorescence (gray) and regions of voltage-dependent fluorescence (red).
Scale bar is 10 m. Right: Recording of whole-cell membrane potential (blue) and weighted
Arch(D95N) fluorescence (red) during a train of action potentials (Kralj et al., 2011).
2.2 Calcium Imaging
Calcium imaging is the most widely used optical technique for measuring neural activity.
Action potential firing leads to an influx of Ca
2+
ions that enter the cell through voltage-
and/or neurotransmitter-gated ion channels. The correlation between calcium bursts and
spiking has been well-established (Lohmann et al., 2002; Smetters et al., 1999; Wong,
1998). Measurements of [Ca
2+
]
i
can thus serve as a proxy for neural activity (Looger and
Griesbeck, 2011). Because calcium dynamics inside the soma are slow relative to the rate
of spiking, calcium imaging has limited temporal resolution. However, the slow nature of
calcium transients also helps to increase the overall SNR (Looger and Griesbeck, 2011).
56
2.2.1 History and Basic Principles
The first calcium indicator, aequorin, is a blue luminescent protein extracted from the
jellyfishAequorea aequorea (Shimomura et al., 1962). In 1967, researchers showed that
the protein could measure calcium transients in single muscle fibers (Ridgway and Ash-
ley). This marked the first use of calcium imaging to measure neural activity. Aequorin
was cloned in 1985 (Inouye et al.) and still receives widespread use, along with its many
mutants. Unlike fluorescent probes, aequorin requires no optical excitation. However,
it suffers from low light output and is irreversibly destroyed upon binding to calcium
(Miyawaki et al., 1999). For these reasons, other calcium indicators are generally pre-
ferred.
The majority of advancements in calcium indicator biotechnology were pioneered by
Roger Tsien, who was later awarded the Nobel Prize in chemistry for his work with GFP.
Tsien’s group developed the first small-molecule calcium dyes in the 1980s. These dyes
were based off the calcium chelator BAPTA, a derivative of EGTA. Both EGTA and
BAPTA are UV-excitable and have a high (>10
5
) affinity for Ca
2+
over Mg
2+
. BAPTA
is the preferred chelator due to its faster kinetics, better calcium selectivity, and smaller
pH dependence (Cobbold and Rink, 1987; Tsien, 1980).
Quin-2 was the first calcium indicator synthesized by Tsien (1980). When excited by
339-nm light, it exhibits a five- to sixfold increase in emission fluorescence upon binding
to calcium. It also exhibits a decrease in fluorescence when excited at 365 nm, allowing
for dual-excitation ratiometric [Ca
2+
]
i
measurements (Cobbold and Rink, 1987). Quin-2
57
was the first calcium sensor to provide estimates of [Ca
2+
]
i
in many important cell types,
however its quantum yield is low compared to other indicators (Cobbold and Rink, 1987).
In 1985, Tsien’s group synthesized two more calcium sensors, indo-1 and fura-2, that
are still widely used today. These dyes provide a 30-fold increase in brightness over
quin-2 (Grynkiewicz et al., 1985). Upon binding to calcium, indo-1 undergoes a shift in
emission wavelength, while fura-2 exhibits a shift in its absorption spectra. Both dyes are
ratiometric and UV-excitable. They have lower Ca
2+
affinities than quin-2, which results
in less calcium buffering and makes it easier to observe Ca
2+
kinetics (Grynkiewicz et al.,
1985; Takahashi et al., 1999).
Despite the utility of UV-excitable dyes, ultraviolet light is toxic to cells. For this
reason, many prefer to use visible-wavelength indicators. The first of these dyes, rhod-2
(red emission) and fluo-3 (green emission), were also synthesized by Tsien’s group. When
bound to calcium, rhod-2 and fluo-3 exhibit a 15- and 40-fold increase in brightness,
respectively (Minta et al., 1989). Because the dyes do not undergo a shift in wavelength,
they cannot be used for ratiometric imaging unless cells are colabeled with a second dye
of a different color.
Several calcium dye variants have been developed throughout the last 30 years. The
newer versions are significantly brighter and are available in a wide range of colors, cal-
cium affinities, and molecular weights. Fluo-4 and Oregon Green BAPTA-1 (OGB-1)
are among the brightest visible-wavelength indicators and have achieved widespread use.
Still, these dyes suffer from issues such as leakage, photobleaching, compartmentalization,
and lack of cell-type specificity, limiting their use to acute imaging experiments.
58
Genetically encoded calcium indicators (GECIs) overcome many of the problems as-
sociated with synthetic dyes. These sensors are chimeric proteins consisting of a calcium-
binding domain fused to one or two fluorescent molecules. Cameleon was the first GECI,
also developed by Tsien and colleagues (Miyawaki et al., 1997). The sensor consists of
two GFP mutants, each with a different emission wavelength (e.g., cyan and yellow).
The fluorophores are bridged by the Ca
2+
-binding protein calmodulin (CaM) and the
CaM-binding M13 fragment of myosin light chain kinase. Upon binding to calcium, CaM
wraps around the M13 domain, causing the two fluorophores to become closer. This in-
creases the amount of fluorescence resonance energy transfer (FRET) between the donor
(cyan) and acceptor (yellow) fluorophores. By monitoring changes in the fluorescence
intensity ratio between each fluorophore, one can obtain an accurate measure of calcium
concentration.
The first generation of chameleons exhibited relatively small ratio changes (< 2), mak-
ing them far less sensitive than most calcium dyes (Takahashi et al., 1999). Later gener-
ations of FRET-based GECIs—namely YC3.60, D3cpv, and TN-XXL—allowed for much
better Ca
2+
sensitivity. These sensors are essentially variants of the original cameleons
that incorporate different fluorescent and/or calcium-binding proteins (e.g., troponin C).
Still, their performance lags behind that of the best synthetic dyes (Mank and Griesbeck,
2008).
Single-wavelength GECIs offer an alternative to FRET-based proteins. In 1999,
Tsien’s group discovered that the -barrel around the GFP chromophore tolerated in-
sertion of entire proteins without quenching fluorescence (Baird et al., 1999; Mank and
Griesbeck, 2008). Insertion of CaM into yellow fluorescent protein (YFP) led to the first
59
single-fluorophore GECI, camgaroo-1 (Baird et al., 1999). Due to its low Ca
2+
affinity
and high sensitivity to changes in temperature and pH (Griesbeck et al., 2001), this in-
dicator was never widely adopted. However, its basic design served as the backbone for
future generations of single-wavelength sensors.
GCaMP was among the first of these improved sensors, incorporating a GFP fluo-
rophore and the CaM-M13 Ca
2+
-binding complex (Fig. 2.3) (Nakai et al., 2001). This
GECI was also plagued by problems with pH and temperature sensitivity, as well as
many other issues (Mank and Griesbeck, 2008). However, iterative improvements in its
design helped to overcome these limitations. Random and site-directed mutagenesis led
to improved GCaMP variants (GCaMP1.6, GCaMP2, GCaMP3, GCaMP5, and finally
GCaMP6), as well as mutants that emit different colors (i.e., red and blue) (Zhao et al.,
2011).
Figure 2.3: Crystal structure of GCaMP in its Ca
2+
-bound state. The protein consists of a
circularly permuted GFP linked to CaM and the CaM-binding M13 fragment of myosin light
chain kinase. Upon binding to calcium, CaM wraps around the M13 peptide and blocks solvent
access to the cpGFP chromophore, causing it to become brighter (Akerboom et al., 2009).
GCaMP3 was the first GECI to truly rival synthetic calcium dyes in terms of SNR
and photostability (Borghuis et al., 2011; Tian et al., 2009). When expressed in vivo, it
can reliably detect short bursts of three or more action potentials. A newer version of this
60
sensor, GCaMP5, improved upon GCaMP3’s SNR by two- to threefold but still lagged
synthetic dyes in terms of detecting sparse neural activity (Akerboom et al., 2012). The
latest GCaMP variant, GCaMP6, can reliably detect single spikes in vivo with near 100%
accuracy. It is the first GECI to have sensitivity, dynamic range, and speed exceeding
those of the best synthetic indicators (Chen et al., 2012a,b).
2.2.2 Delivering Calcium Indicators to Cells
Numerous methods have been developed for introducing calcium indicators into cell pop-
ulations. Choice of the appropriate method depends on several factors including cell type,
indicator type, and the desired length of imaging. This section reviews calcium indicator
loading techniques with special attention given to the retina.
2.2.2.1 Synthetic Dye Loading
Calcium dyes are hydrophilic and therefore cannot readily cross cell membranes. Accord-
ingly, several procedures have been developed for disrupting cell membranes and allowing
the dyes to enter. Microinjection involves the use of intracellular or patch pipettes for
injecting dye into cells (Wu et al., 2004). This was the standard labeling method during
the early days of calcium imaging. However, the technique is labor-intensive and can be
used only on single cells.
Biolistics and electroporation are two techniques that disrupt the membranes of cell
populations. Biolistic delivery uses a “gene gun” to bombard cultured cells or living
tissue with dye-coated particles (Kettunen et al., 2002). The high velocity of the particles
enables them to penetrate cell membranes. Biolistics has been used in the retina to label
61
photoreceptors and RGCs in adult mice (Roizenblatt et al., 2006). However, staining is
sparse and is not cell type-specific.
Electroporation involves application of large electric fields (kV/m) to transiently cre-
ate pores in cell membranes. A recovery period following the procedure is necessary
to allow for membrane resealing and recovery from induced shock and swelling. Early
attempts to electroporate RGCs in ex vivo retina with calcium dye resulted in sparse
labeling (Yu et al., 2009). However, a more recent study demonstrated staining of the en-
tire ganglion cell layer in adult mice (Briggman and Euler, 2011). Nonetheless, displaced
amacrine cells, which constitute roughly 50–60% of cells in the rodent ganglion cell layer
(GCL) (Jeon et al., 1998; Perry, 1981), were labeled indiscriminately.
In addition to methods that involve disrupting cell membranes, noninvasive dye load-
ing approaches have been developed. Many calcium dyes are available in acetoxymethyl
(AM) ester forms, which cap the hydrophilic carboxylic acids with lipophilic AM ester
groups. This causes the dyes to become uncharged and enables them to permeate cell
membranes. Once inside the cell, endogenous esterases cleave the AM bonds and activate
the dye (Paredes et al., 2008).
Loading AM indicators is a simple process that entails bath incubation of the dye
with cells or tissue. The problem with this approach is that it lacks cell-type specificity.
Furthermore, the dyes are not well-retained in adult mammalian tissue, including retina
(Wong, 1998; Wong and Oakley, 1996; Yuste and Katz, 1991). They carry a high negative
charge and are thought to be pumped out of cells by organic anion transporters (OATs)
(Behrend et al., 2009).
62
Increasing the size of calcium dyes via dextran conjugation can make them more
resistant to extrusion. Dextrans are hydrophilic polysaccharides that have low toxicity
and are relatively inert. Dextran-conjugated calcium dyes are available in a variety of
molecular weights. However, they cannot cross cell membranes and are therefore not
trivial to load.
Staining of retinal ganglion cells with dextran conjugates has been achieved via ret-
rograde loading. In the first implementation of this method, investigators punctured
isolated rabbit retina with a dye-laden needle; any axons cut by the needle would take
up the dye. Over a two-hour period, the dye diffused down the axons and into gan-
glion cell bodies. Though useful, this method can label only a small percentage of RGCs
(Baldridge, 1996; Hartwick et al., 2004, 2005).
Our group later developed an improved retrograde loading method, which works by
applying calcium dye to the cut optic nerve. This technique stains the majority of ganglion
cells in isolated retina and is selective for RGCs (Behrend et al., 2009). However, the
approach is not effective in mature mammalian retina, presumably owing to dye extrusion
by OATs. I have since tried variations of this method in adult rat and mouse without
success (see Section 2.3).
2.2.2.2 Expression of Genetically Encoded Calcium Indicators
To express foreign genes in cells, the DNA (or RNA) encoding those genes must be
delivered. Cellular uptake of naked DNA is highly inefficient, primarily owing to its high
negative charge (which is repelled by anionic cell membranes). Inside the cell, DNA must
be trafficked to the nucleus before being converted to protein. However, naked DNA can
63
be degraded by endogenous nucleases before it reaches the nucleus (Patil et al., 2005).
For these reasons, a number of DNA delivery systems have been developed.
DNA delivery systems are typically classified as viral or non-viral. Viral approaches
involve the use of recombinant viruses (i.e., viral vectors) and are the most effective means
for gene delivery (Luo and Saltzman, 2000). However, they can lead to problems with
toxicity and immunogenicity. Non-viral approaches are generally safer but are also less
effective. They achieve transfection through mechanical, electrical, or chemical means
(Table 2.1).
Table 2.1: Non-viral DNA transfection methods (adapted from Luo and Saltzman, 2000).
Approach Method
Mechanical Microinjection
Biolistics
Sonoporation
Electrical Electroporation
Chemical Artificial lipids
Cationic polymers
Choosing the appropriate delivery system is an important design consideration for
gene-targeting studies. In the retina, gene delivery is typically accomplished via electro-
poration, viral vectors, or generation of transgenic animals. In vivo (Dezawa et al., 2002;
Mo et al., 2002) and in utero (Garcia-Frigola et al., 2007) electroporation can transfect
RGC populations with plasmid DNA, but the extent of labeling is limited and is not
RGC-specific. Transgenic mouse lines offer an alternative means to express GECIs in
neuronal populations. The Pvalb-2A-Cre:Ai38 line, for example, expresses the GECI
GCaMP3 in RGCs, horizontal cells, and M¨ uller glia (Zariwala et al., 2012). However, the
64
lack of a pan-ganglion-cell-specific promoter makes it difficult to generate a mouse line
that expresses genes exclusively in RGCs (Feng et al., 2000). Furthermore, establishing
stable transgenic lines is costly and takes many months, making it impractical to incor-
porate new GECIs as they become available. Viral vectors overcome these limitations.
They can be designed to target specific neuronal classes and produced within a matter
of weeks. Of all available gene delivery systems, viral vectors have shown the greatest
cellular specificity in the retina (Borghuis et al., 2011; Hellstr¨ om et al., 2008).
Recombinant adeno-associated viral vectors are most commonly used for retinal gene
transfer (Grimm and Kay, 2003). AAV is a small (25-nm diameter), single-stranded DNA
virus that is non-enveloped and is replication-incompetent (Daya and Berns, 2008). Its at-
tractive properties include lack of pathogenicity and toxicity, as well the ability to achieve
long-term gene expression. Twelve naturally occurring AAV serotypes (AAV1–AAV12)
and over 100 variants have been isolated to date (Li et al., 2009), each with different
tissue tropisms. Tropism can further be altered through a process called pseudotyping,
which cross-packages one serotype’s vector genome into anther serotype’s capsid (Fig.
2.4) (Heilbronn and Weger, 2010). Mutant vectors with enhanced transduction proper-
ties have also been developed (Kwon and Schaffer, 2008; Petrs-Silva et al., 2010). The
packaging limit for recombinant AAV is roughly 4.7 kb (Le Bec and Douar, 2006), much
larger than the size of genetically encoded calcium indicators.
Compared to other viruses, AAV has a superior ability to target specific retinal cell
types (Grimm and Kay, 2003). Transduction of the outer and inner retina can be achieved
by subretinal and intravitreal injections, respectively. AAV tropism in the retina is largely
dictated by viral capsid, which differs among serotypes (Hellstr¨ om et al., 2008). By using
65
Figure 2.4: Pseudotyped AAV vectors cross-package one serotype’s vector genome (AAV2 in
this example) into anther serotype’s capsid. These vectors display altered tropism and improved
transduction efficiency in certain tissues (Heilbronn and Weger, 2010).
various combinations of serotypes and promoters, Borghuis et al. delivered GCaMP3 to
the five major neuron classes in mouse retina (2011). Intravitreal injection of AAV2/1-
SYN1-GCaMP3 targeted RGCs with high specificity, although some horizontal cells were
also labeled. In general, labeling was described as patchy, with >70% of RGCs in a patch
showing GCaMP3 expression (Borghuis et al., 2011).
2.3 Retrograde Dye Loading in Adult Rodent Retina
A major goal of this thesis was to determine the effects of retinal degeneration on the
retina’s response to electrical stimulation. This required transgenic animal models of RP,
which exist only in mammals. Loading calcium dyes into mature mammalian RGCs has
been notoriously difficult (Behrend et al., 2009; Wong, 1998; Wong and Oakley, 1996;
Yuste and Katz, 1991). In a prior study, our group demonstrated retrograde loading
66
of OGB-1 488 dextran 10 kDa into neonatal rat RGCs; however, the dye was not re-
tained after eye opening (P14–P15) (Behrend et al., 2009). Application of the organic
anion transporter inhibitor probenecid (Cunningham et al., 1981; Khamdang et al., 2004)
slightly improved staining in adult rats, suggesting that OATs were externalizing the dye.
OAT activity was not a problem in amphibian retina—OGB-1 488 dextran 10 kDa was
well-retained by tiger salamander RGCs. Thus, this animal model was used in our early
calcium imaging experiments; however, it does have several drawbacks. Compared to
rodents, amphibians are physiologically less similar to humans. For example, both human
and rodent retinas share at least 10–15 types of retinal ganglion cells (Masland, 2001). In
contrast, only six distinct types have been identified in salamander retina (Segev et al.,
2006). Furthermore, salamanders are harvested from the wild, and seasonal changes affect
the health of their retinas. Salamanders also undergo natural retinal degeneration whose
degree appears to correlate with age (Townes-Anderson et al., 1998). In our experiments,
RGC electrical thresholds between salamanders varied by nearly twofold (Behrend, 2009).
Rodent models overcome these problems. Rodents are bred in a controlled setting,
allowing their age to be controlled and ensuring that the animals (and their retinas) are
genotypically and phenotypically similar. Furthermore, a number of transgenic rodent
models are available that mimic human forms of RP (see Section 1.9.3). These factors
make rodents an attractive model to use for studying retinal prostheses in vitro.
Before designing an AAV vector to transduce RGCs with GECIs, I attempted varia-
tions of our retrograde dye loading technique in adult rodents. In addition to OGB-1 488
dextran 10 kDa, I tested three other calcium dyes in both rat and mouse (only rats had
been used previously). I also applied the OAT blocker sulfinpyrazone, which was shown
67
to inhibit fura-2 extrusion in cell lines at a concentration 10 times lower than probenecid
(Di Virgilio et al., 1988).
All dyes were dextran-conjugated potassium salts: fluo-4 dextran 10 kDa (20 mM),
fura dextran 10 kDa (20 mM), and Calcium Green-1 dextran 70 kDa (5 mM) (Life
Technologies, Grand Island, NY). Required loading times were predicted by a finite ele-
ment model based off the diffusion equation (Behrend et al., 2009). Ten-kilodalton dyes
were loaded for 4–14 hours. Calcium Green-1 dextran 70 kDa was loaded for 15 hours.
Probenecid and sulfinpyrazone were added to the superfusate at concentrations of 2.5
mM and 250 M, respectively. Experiments were performed in adult Long Evans rats
(Harlan Laboratories, Indianapolis, IN) and C57B16/J mice (The Jackson Laboratory,
Bar Harbor, ME).
As a control experiment, I applied an equimolar mixture (10 mM) of Alexa Fluor 594
dextran 10 kDa and OGB-1 dextran 10 kDa to the cut optic nerve. The weakly charged
Alexa tracer was retained by RGC somata (Fig. 2.5A), but the highly charged Oregon
Green was not (Fig. 2.5B), further supporting the hypothesis that OATs are responsible
for calcium dye extrusion. Attempts to load fluo-4 dextran 10 kDa (Fig. 2.5 C ) and fura
dextran 10 kDa (Fig. 2.5D) were also unsuccessful. Reasoning that a heavier dye might
be more resistant to externalization, I tried loading Calcium Green-1 dextran 70 kDa;
however, this dye was not retained (Fig. 2.5E).
68
Figure 2.5: Attempts to load dextran-conjugated calcium dyes via the cut optic nerve were
unsuccessful in adult rat (Weitz et al., in press). Retrograde loading with an equimolar mixture
of Alexa Fluor 594 dextran 10 kDa and OGB-1 dextran 10 kDa led to retention of the Alexa dye
(A) but not Oregon Green (B). Very weak OGB-1 staining can be seen in some RGC somata, but
I could not evoke calcium transients in these cells. Fluo-4 dextran 10 kDa (C ), fura dextran 10
kDa (D), and Calcium Green-1 dextran 70 kDa (E) were also not retained by adult rat RGCs,
even in the presence of probenecid and sulfinpyrazone. Results in adult mice were similar. Scale
bars are 100 m.
2.4 Transducing RGCs with a Custom Adeno-Associated
Viral Vector
The inability to load synthetic calcium dyes into mature mammalian retina led me to
investigate genetically encoded calcium indicators. I chose the latest generation of single-
wavelength GECIs, GCaMP3 (Tian et al., 2009) and GCaMP5G (Akerboom et al., 2012),
69
due to their high sensitivity and photostability. AAV was selected as the delivery vehicle
because it is easily administered (via intravitreal injection), is not pathogenic, and can
be designed to target specific classes of retinal cells (Borghuis et al., 2011).
I designed a custom AAV vector with the aim to achieve labeling that was dense and
widespread, yet selective to RGCs. This required selecting the appropriate viral serotype
and promoter, as these elements dictate AAV transduction profiles. Of the many AAV
serotypes that have been identified (Schmidt et al., 2008), AAV2 has been reported to
be best for labeling RGCs (Auricchio et al., 2001; Hellstr¨ om et al., 2008). AAV2 also
transduces the greatest number of cells following intravitreal injection (Hellstr¨ om et al.,
2008). When coupled with the CAG promoter, AAV2 vectors can target the overwhelming
majority of RGCs in mammalian retina (Martin et al., 2002).
I modeled my vector off the construct, AAV2-CAG-GFP, which transduces ∼ 85% of
adult rat RGCs with GFP following intravitreal injection (Martin et al., 2002). By re-
placing the GFP transgene with GCaMP3 and GCaMP5, I created AAV2-CAG-GCaMP3
and AAV2-CAG-GCaMP5G, respectively (Fig. 2.6). Both vectors contain the strong
and ubiquitous CAG promoter (Niwa et al., 1991) followed by GCaMP3 or GCaMP5G
cDNA. Woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) is
placed downstream of the transgene to enhance protein translation (Loeb et al., 1999).
Each cassette is flanked by AAV2 inverted terminal repeats.
70
Figure 2.6: Map of pAAV2-CAG-GCaMP. The CMV enhancer, CBA promoter, exon, and intron
collectively form the CAG promoter. WPRE is placed downstream of the GCaMP3/GCaMP5G
transgene to increase protein translation. The cassette is flanked by AAV2 inverted terminal
repeats. TR: terminal repeat; CMV ie enhancer: cytomegalovirus immediate-early enhancer;
CBA promoter: chicken -actin promoter; WPRE: woodchuck hepatitis virus post-transcriptional
regulatory element; bGH poly(A): bovine growth hormone polyadenylation signal.
2.4.1 Plasmid Construction
2.4.1.1 Overview
The pGFP plasmid (Fig. 2.7) (Klein et al., 2002; Wu et al., 2003) was used as a backbone
for constructing pAAV-CAG-GCaMP3 and pAAV-CAG-GCaMP5G. The GFP transgene
was replaced with GCaMP3 and GCaMP5G according to the following steps:
1. Restriction enzyme digestion, followed by gel extraction, to remove GFP from the
pGFP backbone.
2. PCR to isolate the GCaMP insert from its expression vector and add the appropriate
restriction sites (for subsequent mating with the pGFP backbone).
3. Ligation of the pGFP backbone and GCaMP insert.
4. Screening to select ligation product containing GCaMP in the forward (5
′
–3
′
) ori-
entation.
71
Figure 2.7: The pGFP plasmid was used as a backbone for constructing pAAV-CAG-GCaMP.
The GFP transgene is flanked by HindIII restriction sites. (Adapted from Martin et al., 2003.)
2.4.1.2 Detailed Procedure
The first step in creating pAAV-CAG-GCaMP was to isolate the pGFP backbone and
GCaMP insert for subsequent ligation. The GFP transgene was cut from pGFP via
restriction enzyme digestion with HindIII (which flanked both ends). Digestion was
performed at 37
∘ C for two hours. The restriction product was run on a 1% gel to
separate the pGFP backbone from the GFP cDNA (Fig. 2.8). The 5.8-kb backbone was
cut from the gel with a razor blade and purified with a GenCatch Gel Extraction Kit
(Epoch Life Science, Sugar Land, TX).
Before ligating GCaMP to the pGFP backbone, HindIII sticky ends needed to be
added to both ends of the GCaMP transgene. Extraction of GCaMP from its expression
vector and addition of the HindIII sites was accomplished with PCR. Forward and re-
verse primers (Fig. 2.9) were designed and ordered from Integrated DNA Technologies
(Coralville, IA). Thirty cycles of PCR were performed with Taq DNA Polymerase (Life
72
Figure 2.8: pGFP was digested with HindIII and run on a gel. The cut fragments (middle and
right lanes) are shown alongside a 1-kb DNA ladder (left lane). The top bands represent the
pGFP backbone (5.8 kb), and the bottom bands represent the GFP transgene (0.75 kb).
Technologies, Grand Island, NY). The amplified product was extracted with a QIAquick
PCR Purification Kit (Qiagen, Hilden, Germany) and digested with HindIII enzyme to
obtain sticky ends. The DNA was run on a gel to confirm proper length (Fig. 2.10) and
then isolated with a GenCatch Gel Extraction Kit.
Figure 2.9: Forward and reverse primers were designed to extract the GCaMP transgene from
its expression vector and add HindIII restriction sites.
73
Figure 2.10: The GCaMP insert, flanked by HindIII, was run on a gel to confirm proper length
(1.3 kb). The insert (middle and right lanes) is shown alongside a 100-bp DNA ladder (left lane).
Prior to ligation, the pGFP backbone was treated with Antarctic Phosphatase (New
England Biolabs, Ipswich, MA) to remove the 5
′
phosphoryl termini and prevent self-
ligation. Ligation was performed by incubating a 1:3 ratio of pGFP backbone to GCaMP
insert in the presence of T4 DNA Ligase (New England Biolabs, Ipswich, MA). The
reaction was allowed to proceed overnight at a temperature of 16
∘ C.
Since the GCaMP insert was flanked by HindIII on both ends, it could ligate in
the forward and reverse orientations. It was therefore necessary to screen for plasmid
DNA containing forward-orientated GCaMP. SURE 2 Supercompetent Cells (Agilent
Technologies, Santa Clara, CA) were transformed with 1 L of ligation product on LB-
carbenicillin agar plates and left to grow overnight at 37
∘ C. Transformation colonies
were purified by the streak dilution method (Miller, 1992) and again incubated overnight.
Purified single colonies were picked and inoculated in microcentrifuge tubes containing
2 mL LB medium with 0.1% carbenecillin. Tubes were shaken overnight at 225 rpm
74
and 37
∘ C to allow bacterial growth. DNA from each tube was isolated with a Plasmid
Miniprep Kit (Bioland Scientific, Paramount, CA).
To determine which samples had forward-oriented GCaMP, a portion of DNA from
each was digested with BglII and XhoI enzymes. This produced fragments of different
lengths, depending on transgene orientation (1.9, 2.0, and 3.2 kb for forward orientation;
1.0, 2.9, and 3.2 kb for reverse orientation). Fragments were run on a gel to determine
their lengths (Fig. 2.11).
Figure 2.11: Purified ligation product was digested with BglII and XhoI and run on a gel
alongside a 1-kb DNA ladder. The middle lane shows fragments from plasmid containing forward-
oriented GCaMP. The fragments are 1.9, 2.0, and 3.2 kb, though only two bands are visible due
to overlap between the 1.9- and 2.0-kb fragments. The right lane shows fragments from plasmid
DNA containing reverse-oriented GCaMP. The fragments are 1.0, 2.9, and 3.2 kb. Again, only
two bands are visible due to overlap between the 2.9- and 3.2-kb fragments.
Plasmid DNA containing forward-oriented GCaMP was amplified and sequenced
(Eton Bioscience, San Diego, CA) to confirm lack of PCR-induced mutations. Upon
confirmation, the plasmid was tested in AtT-20 cells, a pituitary cell line that exhibits
spontaneous calcium oscillations. Cells were transfected with Lipofectamine 2000 (Life
75
Technologies, Grand Island, NY) according to the manufacturer’s protocol and imaged on
an epifluorescence microscope. As shown in Fig. 2.12, GCaMP reported AtT-20 calcium
transients through large changes in fluorescence intensity.
Figure 2.12: AtT-20 cells transfected with pAAV-CAG-GCaMP3 were imaged to monitor cal-
cium transients. Left: Fluorescence image of transfected cells in culture. Scale bar is 50 m.
Right: GCaMP3 fluorescence of three AtT-20 cells arising from spontaneous calcium oscillations.
Each trace represents a different cell.
2.4.2 Vector Packaging
Recombinant AAV vectors were produced at the University of Florida Vector Core by
the two-plasmid cotransfection method (Zolotukhin et al., 1999). Briefly, one CellSTACK
(Corning, Inc., Corning, NY) with approximately 1× 10
9
HEK 293 cells was cultured in
Dulbecco’s Modified Eagle’s Medium (Hyclone Laboratories, Logan, UT), supplemented
with 5% fetal bovine serum and antibiotics (cDMEM). A calcium phosphate precipita-
tion transfection was set up by mixing a 1:1 molar ratio of pAAV-CAG-GCaMP and
a serotype-2-specific rep-cap helper plasmid, pDG. This precipitate was added to 1100
76
mL of cDMEM, and the mixture was applied to the cell monolayer. The transfection
was allowed to incubate at 37
∘ C/7% CO
2
for 60 hours. The cells were then harvested
and lysed by three freeze/thaw cycles. The crude lysate was clarified by centrifugation,
and the resulting vector-containing supernatant was divided among four discontinuous
iodixanol step gradients. The gradients were centrifuged at 350,000 g for 1 hour. 5 mL
of the AAV containing 40–60% interface was removed from each gradient and combined.
This combined iodixanol fraction was further purified and concentrated by column chro-
matography on a 5-mL HiTrap Q Sepharose (anion exchange) column using an AKTA
FPLC system (Pharmacia, Piscataway, NJ). The vector was eluted from the column using
215-mM NaCl, pH 8.0, and the AAV peak was collected. The AAV-containing fraction
was then concentrated and buffer exchanged in Alcon Balanced Salt Solution (BSS) with
0.014% Polysorbate 20, using a Biomax 100K concentrator (Millipore, Billerica, MA).
The AAV was titered for DNase-resistant vector genomes by real-time PCR relative to a
standard. Purity of the final vectors was assayed by polyacrylamide gel electrophoresis to
determine the fraction of total protein that was AAV viral capsids VP1, VP2, and VP3
(>95%). Final concentrations of AAV2-CAG-GCaMP3 and AAV2-CAG-GCaMP5G were
≥ 3.2× 10
12
vector genomes/mL.
2.4.3 Transducing Adult Rat Retina
Adult rats received an intravitreal injection of AAV2-CAG-GCaMP3 or AAV2-CAG-
GCaMP5G. Ganglion cells began to show green fluorescence after one week. Expression
77
levels became strong after two weeks, at which point retinas were harvested for exper-
iments. Transduced retinas were dissected out and examined histologically or mounted
on an MEA for imaging and electrophysiology.
2.4.3.1 Intravitreal Injections
Rats were anesthetized via intraperitoneal injection of ketamine (60 mg/kg) and xylazine
(8 mg/kg). Phenylephrine (2.5%) and tropicamide (1%) were dropped into each eye
to induce pupil dilation. Tetracaine (0.5%) was applied as a local anesthetic. A 30G
needle was used to make a pilot hole through the sclera, choroid, and retina, 1–2 mm
posterior to the corneal limbus. A microliter syringe (Hamilton, Reno, NV) attached
to a blunt 32G needle was used to inject 4–5 L of virus into the vitreous. Care was
taken to avoid injury of the lens. Injections were given slowly over a period of 30 seconds
to allow diffusion of the virus. The needle was left in place for 30 seconds after the
injection and withdrawn slowly to prevent leakage. The injection site was visualized with
an ophthalmic microscope to confirm absence of cataract, intraocular bleeding, and air
bubbles. Antibiotic eye ointment (neomycin and polymyxin B sulfates and bacitracin
zinc) was applied to prevent infection.
2.4.3.2 Histology
Eyes treated with AAV were enucleated, and their anterior segments were carefully re-
moved. Posterior eyecups were fixed in 4% paraformaldehyde in 1x phosphate buffered
saline (PBS) for 1 hour. Following fixation, eyecups were washed three times in 1x PBS for
10 minutes each. Samples were transferred to 30% sucrose in 1x PBS and kept overnight
78
at 4
∘ C. Eyecups were then embedded in O.C.T. compound (Tissue-Tek; Sakura Finetek,
Torrance, CA) and sectioned to a thickness of 8m. Sections were collected on slides and
stored at− 80
∘ C until imaging. Prior to imaging, slides were thawed at room temper-
ature and immersed in 1x PBS for 10 minutes. Nuclei were stained with ProLong Gold
Antifade Reagent with DAPI (Life Technologies, Grand Island, NY). Imaging was per-
formed at the Doheny Eye Institute Specialized Microscopy Core on a Zeiss (Thornwood,
NY) LSM 510 confocal laser scanning microscope equipped with a Plan-Neofluar 1.3-NA
40x objective. GCaMP fluorescence was excited with a 488-nm argon laser and collected
through a 505–530 nm band-pass filter. DAPI was excited at 800 nm by a Ti:sapphire
laser and collected through a 390–465 nm band-pass filter.
2.4.3.3 Transduction Profiles
Retinal wholemounts showed widespread GCaMP expression, with anywhere from one
quarter to the entire wholemount exhibiting green fluorescence (Fig. 2.13). Ganglion
cell bodies were generally brighter than their axons, which permitted imaging of somata
through the superficial nerve fiber layer (see Fig. 2.14 A). When experiments were per-
formed less than four weeks post-injection, fluorescence remained predominantly localized
to RGC cytoplasms (see Fig. 2.14C, inset). As time progressed, baseline fluorescence
increased and became apparent in ganglion cell nuclei, indicating GCaMP overexpression
and cytomorbidity (Akerboom et al., 2012; Borghuis et al., 2011; Tian et al., 2009). Elec-
trical recordings confirmed that very bright cells and/or ones with filled nuclei did not fire
spikes or exhibit GCaMP fluorescence transients during electrical stimulation. To limit
79
overexpression, viral stock was diluted in BSS prior to administration. Optimal dose was
roughly 1–5× 10
9
vector genomes per injection.
Figure 2.13: Retinal wholemount of an adult rat infected with AAV2-CAG-GCaMP3 (14 days
post-injection). The retina was mounted on a multielectrode array and imaged with an inverted
epifluorescence microscope. GCaMP3 expression is visible throughout the ganglion cell layer. The
mosaic was created by stitching together 88 10x images. The two black circles are 200-m-diameter
Pt/Ir electrodes.
80
Double-labeling RGCs with the red tracer dye Alexa 594 enabled me to determine
whether transduced cells were RGCs or other cell types (Fig. 2.14). I compared ex-
pression profiles induced by AAV2-CAG-GCaMP3 to those of AAV2/1-SYN1-GCaMP3,
which labels mouse RGCs with high efficiency (Borghuis et al., 2011). Cell counting
revealed that both vectors had a high specificity for RGCs. Although differences were
not statistically significant (unpaired t-tests), AAV2-CAG-GCaMP3 labeled more RGCs
and less non-RGCs than AAV2/1-SYN1-GCaMP3 (Table 2.2). Roughly 83% of ganglion
cells were transduced by AAV2-CAG-GCaMP3, which achieved a labeling density of 1813
± 343 cells/mm
2
. Similarly, Martin et al. found that AAV2-CAG-GFP targeted 84.5%
of adult rat RGCs and labeled 1828± 299 cells/mm
2
(2002). Given that 85.3% of all
AAV2-CAG-GCaMP3-labeled cells (i.e., ∼ 1550 cells/mm
2
) were RGCs (see Table 2.2),
my counts are consistent with counts of∼ 1500–1600 RGCs/mm
2
obtained by Fluorogold
backfilling through rat optic nerve and superior colliculus (Salinas-Navarro et al., 2009).
Figure 2.14: To investigate the vector’s specificity for transducing RGCs, ganglion cells were
double-labeled with the red tracer dye Alexa 594. A: Green fluorescence image of rat retina
infected with AAV2-CAG-GCaMP3. Axons are weakly labeled. B: Red fluorescence image of
ganglion cells retrogradely loaded with Alexa 594 (same field of view asA). C : Merging A and B
reveals that the majority of RGCs are transduced with GCaMP3. The inset shows how GCaMP3
fluorescence is localized to the cytoplasm, a characteristic indicative of cells that have not been
damaged by GCaMP overexpression (Akerboom et al., 2012; Borghuis et al., 2011; Tian et al.,
2009). Scale bar is 200 m.
81
Table 2.2: Percentage of RGCs and other retinal cells labeled with GCaMP3 following intravitreal
injection.
AAV2-CAG AAV2/1-SYN1 P-value
% RGCs Labeled 82.9± 9.2 75.4± 7.5 0.14
% Labeled Cells that Were
Not RGCs
14.7± 7.6 17.6± 3.9 0.39
Labeled Cell Density
(cells/mm
2
)
1813± 343 1553± 367 0.24
# Counted Cells 2714 1492 —
Cells were counted in regions with dense GCaMP3 labeling. I counted eight 400× 400 m regions
from two retinas transduced with AAV2-CAG-GCaMP3 and five 400× 400 m regions from two
retinas transduced with AAV2/1-SYN1-GCaMP3. RGCs were identified after retrograde load-
ing with Alexa 594. Percentages indicate mean± SD. P-values imply no significant differences
between each vector’s labeling profile.
To determine which types of non-RGCs were targeted by AAV2-CAG-GCaMP3 and
AAV2/1-SYN1-GCaMP3, I examined cross sections of retina transduced by each vector.
With rare exception, I found AAV2-CAG-GCaMP3 labeling to be confined to the ganglion
cell layer, implying transduction of RGCs and displaced amacrine cells (Fig. 2.15A). In
contrast, AAV2/1-SYN1-GCaMP3 expression often extended into the inner nuclear layer
(INL), where it labeled amacrine cells (Fig. 2.15B). Coupled with my double-labeling
and cell counting results, these data indicate that the 14.7% of non-RGCs transduced by
AAV2-CAG-GCaMP3 (see Table 2.2) were displaced amacrine cells. Similarly, Harvey et
al. found that 12–13% of all cells transduced by intravitreal injection of AAV2-CMV-GFP
in adult rat were amacrine cells (2002).
In addition to AAV2-CAG-GCaMP3/5G and AAV2/1-SYN1-GCaMP3, two other
AAV vectors were tested in adult rat. The first was AAV2/9-SYN1-GCaMP3, which
82
Figure 2.15: Representative sections of rat retina infected with AAV2-CAG-GCaMP3 (A) and
AAV2/1-SYN1-GCaMP3 (B). AAV2-CAG-GCaMP3 labeling is limited to the GCL (RGCs and
displaced amacrine cells), while AAV2/1-SYN1-GCaMP3 labeling extends into the INL (amacrine
cells). Top images show GCaMP3 fluorescence;bottom images show GCaMP3+DAPI fluorescence.
Scale bar is 50 m.
labels RGCs, photoreceptors, and horizontal cells in mouse retina (Borghuis et al., 2011).
Following intravitreal injection in rat, this vector transduced RGCs and displaced ama-
crine cells, similar to AAV2/1-SYN1-GCaMP3 (Fig. 2.16A). No fluorescent photorecep-
tors or horizontal cells were observed. The last vector tested was AAV9-CAG-GFP, which
shows enhanced expression in the brain compared with AAV2 (Dayton et al., 2012). In
the retina, AAV9-CAG-GFP was found to produce sparse labeling of photoreceptors and
amacrine cells (Fig. 2.16B).
AAV2-CAG-GCaMP3 and AAV2-CAG-GCaMP5G were ultimately selected for my
mammalian electrophysiology studies. These vectors were designed in-house and achieved
the most widespread and selective labeling of RGCs. Prior studies have also reported
83
Figure 2.16: Representative sections of rat retina infected with AAV2/9-SYN1-GCaMP3 (A) and
AAV9-CAG-GFP (B). AAV2/9-SYN1-GCaMP3 labeled RGCs and amacrine cells, while AAV9-
CAG-GFP sparsely labeled photoreceptors and amacrine cells. Top images show GCaMP3/GFP
fluorescence;bottom images show GCaMP3/GFP+DAPI fluorescence. Scale bar is 50m.
superior RGC labeling with AAV2 versus other serotypes (Auricchio et al., 2001; Hell-
str¨ om et al., 2008). Though the vectors I tested contained different promoters, CAG and
synapsin-1 are both strong promoters that produce similar amounts of transgene expres-
sion (Shevtsova et al. 2005). Furthermore, the extent of retinal transduction is limited
by the ability of the AAV particles to penetrate the retina, which is determined by the
capsid serotype, not the promoter (Dalkara et al., 2009; Petrs-Silva et al., 2010).
2.5 Imaging Calcium Transients in Transduced RGCs
Retinas transduced with GCaMP3 or GCaMP5G were dissected out, mounted ganglion
cell-side-down on a transparent MEA, and imaged with an inverted epifluorescence mi-
croscope. Subsequent electrical stimulation with MEA electrodes evoked large increases
84
in fluorescence intensity—much larger than we observed in amphibian retina with syn-
thetic calcium dyes (Behrend et al., 2009). This enabled us, for the first time, to visualize
spatial patterns of RGC activation in real time.
2.5.1 Axonal Activation
The calcium transients evoked by electrical stimulation were clearly visible when viewed
through the microscope. As shown in Fig. 2.17, burst stimulation with a 20-m-diameter
electrode (arrow) elicited strong fluorescence responses in GCaMP3-expressing RGCs. A
rapid train of 120 pulses (1.2 A, 0.4-ms pulse width, 333 Hz) stimulated a bundle of
passing axons, antidromically activating a streak-like pattern of RGC somata. Normal-
ized change in fluorescence (Δ/ ) for responding cells was 75.9± 21.5% ( = 34).
Application of 1-mM CdCl
2
abolished all fluorescence signals within minutes.
Figure 2.17: Electrical stimulation activates a streak pattern of RGCs, as revealed through
large changes in GCaMP3 fluorescence intensity. A: Before stimulation, cells are at baseline
fluorescence. B: Delivering a burst of 120 suprathreshold pulses from a 20-m-diameter electrode
(arrows) causes electrically activated cells to become visibly brighter. C : Image subtraction of A
from B highlights the pattern of responding cells. Stimulation with 0.4-ms pulses activates an
axon bundle, causing a streak-like antidromic response that extends away from the optic disc.
Average Δ/ for responding RGCs was 75.9± 21.5% ( = 34). Scale bar is 100 m.
Avoiding stimulation of RGC axon bundles is one of the biggest challenges faced by
epiretinal prostheses (Behrend et al., 2011; Nanduri et al., 2011; Weiland et al., 1999). In
85
the human eye, ganglion cell fibers travel from the periphery of the retina or the horizontal
raphe (a line dividing the inferior and superior retina) to the optic disc. Prior analysis
of percept shape in Argus I and II subjects revealed that axonal stimulation produces
elongated phosphenes, which extend from the electrode location (in the macula) to the
horizontal raphe (where the axons terminate; see Fig. 4.1) (Nanduri et al., 2011). In an
attempt to reproduce this behavior in rat, ganglion cells labeled with GCaMP5G were
stimulated using an electrode size (200-m diameter) and pulse width (0.4 ms/phase)
similar to those of the Argus II. Indeed, axonal activation produced an elongated pattern
of RGC responses subtending from the electrode position to the edge of the retina (rats
do not have a horizontal raphe) (Fig. 2.18).
Figure 2.18: A burst of 120 pulses (14 A, 0.4-ms pulse width, 167 Hz) was delivered from
a 200-m-diameter transparent electrode (red circle). Stimulation activated nearby bundles of
RGC axons, creating an elongated excitation pattern that extended to the edge of the retina.
Left: Fluorescent image mosaic of a rat retina transduced with AAV2-CAG-GCaMP5G. Right:
Background-subtracted fluorescence responses to electrical stimulation. Brightness and contrast
were adjusted to accentuate the responses.
86
2.5.2 Comparing Fluorescence Responses of Different Calcium
Indicators
I compared calcium transients from GCaMP3-, GCaMP5G-, and Oregon Green-labeled
RGCs by delivering bursts of 40 suprathreshold pulses (0.06-ms pulse width, 333 Hz)
through a 200-m-diameter electrode (Fig. 2.19). GCaMP3 fluorescence responses in
rat RGCs were similar in amplitude to Oregon Green responses in salamander RGCs
(Behrend et al., 2009) but were generally less noisy. GCaMP5G signals in rat were roughly
3–4 times larger than those of GCaMP3. After rising to full magnitude, GCaMP3 and
GCaMP5G fluorescence decayed with time constants of 0.91 s± 0.10 SEM ( = 24) and
0.50 s± 0.03 SEM ( = 66), respectively.
Figure 2.19: Normalized change in fluorescence (Δ/ ) plotted for RGCs in response to re-
peated stimuli (tick marks). Each stimulus was a rapid burst of 40 suprathreshold pulses. The
bottom traces are from salamander RGCs ( = 34) labeled with OGB-1 dextran 10 kDa (Behrend
et al., 2009). The middle traces ( = 24) and top traces ( = 66) are from rat RGCs labeled
with GCaMP3 and GCaMP5G, respectively. GCaMP3 and Oregon Green signals are similar in
amplitude, while GCaMP5G provides a 3–4x improvement in Δ/ .
87
2.5.3 Correlating Electrical Activity with Calcium Transients
To confirm that calcium transients arose from RGC spiking activity, I recorded extracellu-
lar potentials from a GCaMP5G-labeled RGC during stimulation with a nearby electrode
(Fig. 2.20). Recorded signals were amplified with a MEA1060-Inv-BC amplifier (Multi
Channel Systems, Reutlingen, Germany). Delivering a burst of 20 pulses (0.06-ms pulse
width, 250 Hz) at an amplitude just below threshold caused only 5 of the pulses to evoke
action potentials. Spike shape was revealed by subtracting the stimulus artifact (Sekirnjak
et al., 2006) and averaging the 5 traces (Fig. 2.20, blue trace). GCaMP5G fluorescence
rose by roughly 4% Δ/ (Fig. 2.20, green trace) in response to the 5 elicited spikes.
Single action potentials never elevated GCaMP5G fluorescence above the baseline noise
level; a rapid burst of 5–10 spikes was needed. For RGCs expressing GCaMP3, 20–30
action potentials were needed to detect a calcium transient.
2.5.4 Statistics about the Properties of Electrically Responsive Cells
I gathered statistics about the properties of electrically responsive cells across seven
GCaMP5G-labeled retinas. Retinas were stimulated with a transparent 200-m-diameter
electrode, and only RGCs lying directly above the electrode were used for analysis
( = 418). In response to high-amplitude stimulation, 82.3± 10.5% of cells responded
with detectable calcium transients (stimulus amplitude was 2–3 times greater than the
mean threshold of responding RGCs). For responding cells, there was little to no cor-
relation between baseline fluorescence,
0
, and response magnitude, Δ =

−
0
(correlation coefficient = 0.19± 0.24). Within each retina, baseline fluorescence of RGC
somata was highly uniform (mean coefficient of variation = 12.4± 4.9%).
88
Figure 2.20: GCaMP5G calcium transients are correlated with a rapid series of evoked spikes.
Near-threshold stimuli were delivered to a ganglion cell (arrowhead) by a transparent 200-m-
diameter electrode (dotted outline). Extracellular spikes were recorded by a 10-m-diameter
electrode (black circle). Artifact subtraction was used to isolate spike waveforms from the electri-
cally evoked responses. The blue trace shows a subtracted spike that was averaged from 5 traces.
Its shape matches that of a spontaneous spike recorded from the same cell (red trace). The 5
elicited action potentials caused a ∼ 4% increase in GCaMP5G fluorescence intensity (green trace).
To investigate whether GCaMP overexpression (i.e., very bright baseline fluorescence)
was the cause of non-responsiveness (Borghuis et al., 2011; Tian et al., 2009), I compared

0
of responding cells to that of non-responding ones. Baseline fluorescence between
these two populations varied by only 1.0± 5.9% ( = 0.67, unpaired t-test), indicating
that cytomorbidity was not the primary cause. Indeed, cells that were bright (
0
> 1
standard deviation from the mean) and not responsive made up only 3.1% of the total
population.
89
2.5.5 Effects of Rat Strain, Temperature, and Cytomorbidity
All data reported in this thesis were collected from outbred rat strains (Long Evans;
Harlan Laboratories, Indianapolis, IN) and inbred/outbred crosses (S334ter-line-3 het-
erozygous; bred in-house). Injecting AAV2-CAG-GCaMP into the vitreous of inbred
Copenhagen rats (Charles River Laboratories International, Wilmington, MA) also led
to widespread RGC transduction. However, GCaMP-induced cytomorbidity (Borghuis
et al., 2011; Tian et al., 2009) was much more pronounced in Copenhagen retina, espe-
cially at higher temperatures. At 23
∘ C, Copenhagen RGCs showed very dim baseline
fluorescence (Fig. 2.21, left). Electrical stimulation evoked calcium transients in these
cells, but fluorescence changes were much smaller than those of Long Evans. When
bath temperature was increased to 28–30
∘ C, Copenhagen RGCs became very bright and
stopped responding to stimulation (Fig. 2.21, right). This was accompanied by cessation
of spontaneous spiking and a gradual increase in nuclear fluorescence. Reducing GCaMP
expression levels with lower viral titers did not eliminate cytomorbidity. These findings
resemble those of Newman, in which temperatures above 24
∘ C altered calcium indica-
tor fluorescence in rat M¨ uller glia and caused deterioration of retinal function (2003).
For experiments in Long Evans and S334ter-line-3 rats, temperature had little to no ef-
fect on calcium transients (though higher temperatures sometimes led to increased basal
fluorescence, especially with higher viral doses).
90
Figure 2.21: Effects of temperature on GCaMP5G-labeled Copenhagen RGCs. Left: At 23
∘ C,
ganglion cells showed dim baseline fluorescence. Many were not visible. Right: When temperature
was increased to 28–30
∘ C, RGCs became very bright and stopped responding to stimulation. Most
RGCs exhibited nuclear fluorescence, indicating GCaMP-induced cytomorbidity. Both images
show the same field of view and were captured 8 minutes apart.
Prior studies have also reported problems with inbred lines that do not occur with
outbreds, possibly due to deleterious homozygous recessive alleles that result from in-
breeding. For example, Carlson et al. found that overexpressing Alzheimer amyloid pre-
cursor protein (Kang et al., 1987) was lethal to inbred mice but not to outbreds (1997).
Similarly, Kinney and Sidman found that a mutation causing spongiform encephalopathy
killed inbred mice within three months but was well-tolerated outbred strains (1986).
Though the focus of these studies is unrelated, their findings demonstrate how phenotype
can be affected by genetic background.
Despite the relative ease of AAV vector production and delivery to the eye, GCaMP-
induced cytomorbidity limits the types of experiments that can be performed. In agree-
ment with another study, I found that retinas harvested after∼ 4 weeks exhibited abnor-
mal cellular physiology (Borghuis et al., 2011). This has been suggested to arise from
91
GCaMP overexpression and interaction of the sensor’s calmodulin and/or M13 motifs
with endogenous proteins (Hasan et al., 2004). It may be possible to limit overexpres-
sion by systemically administering hyperosmotic mannitol prior to AAV injection (Burger
et al., 2005; Kuhn et al., 2012), though this might also cause GCaMP expression to extend
beyond the GCL. Chronic experiments would require the use of transgenic GECI knock-
ins, as expression levels in these animals remain stable for at least 10 months (Zariwala
et al., 2012).
2.6 Apparatus Design
Performing the experiments in this thesis required specialized hardware for interfacing
with the retina. Some of this hardware was designed previously (Behrend, 2009), while the
rest had to be custom built. New MEA layouts were needed for performing multielectrode
pattern stimulation of the retina. Modifications to the MEA process flow were necessary
for preventing parasitic capacitance associated with the electrode traces. Additional
circuitry was required for delivering small-amplitude signals to MEA electrodes.
2.6.1 Electrophysiology Rig
The electrophysiology rig (Fig. 2.22) was designed to measure calcium fluorescence from
ex vivo retina during stimulation with MEA electrodes. The retina is held in a record-
ing chamber atop a thin, transparent MEA and placed on an inverted microscope stage.
Electrical stimulation is provided by a computer-controlled stimulus generator, which
interfaces with the electrode array through a custom printed circuit board (PCB). Fluo-
rescence emitted by the calcium indicator passes through the MEA and is captured by an
92
electron-multiplied CCD (EMCCD) camera. A gravity-fed perfusion system enables the
retina to be preserved in in heated, oxygenated saline over the course of each experiment.
Figure 2.22: Electrophysiology rig. Left: Image of the inverted microscope stage with the MEA,
recording chamber, and interface board mounted on top. The EMCCD camera is attached to the
right-side port of the microscope. Right: Cartoon depiction of the stimulation and recording
setup.
The microscope is an Eclipse TE2000-U (Nikon, Tokyo, Japan) equipped with two
objective lenses: Plan Apo 0.75 NA 20x and 0.45 NA 10x. Fluorescence excitation
is provided by a super bright white LED. Excitation and emission light are filtered
through Semrock (Rochester, NY) filter sets: #GFP-4050A for GCaMP (Fig. 2.23)
and #TXRED-4040B for Alexa Fluor 594 (Fig. 2.24). Fluorescence images are captured
by an Andor (Belfast, Northern Ireland) Ixon DV885KCS-VP EMCCD.
Electrical stimulation is controlled by an 8-channel stimulus generator (STG2008;
Multi Channel Systems, Reutlingen, Germany). Voltage stimuli are fed through a custom
93
Figure 2.23: Fluorescence spectra for GCaMP3/GCaMP5G and the GFP-4050A filter set.
Blue = excitation, red = emission, and green = dichroic. (Image courtesy of Semrock, http:
//www.semrock.com/.)
Figure 2.24: Fluorescence spectra for Alexa Fluor 594 and the TXRED-4040B filter set. Blue
= excitation, red = emission, and green = dichroic. (Image courtesy of Semrock, http://www.
semrock.com/.)
94
16-channel current source board. Each current source has± 25-A range with 50-nA
resolution and uses a conversion factor of 0.1 V = 1 A (Behrend, 2009). Channels
can be connected in parallel to perform stimulation with amplitudes >25 A. Electrode
voltage and current are monitored on an oscilloscope during stimulation to ensure that
the water window is not exceeded.
The stimulus generator can accurately deliver a range of voltages from 0.2 to 8 V,
corresponding to a minimum current (after V-to-I conversion) of 2A. This was a problem
for experiments involving long pulses (up to 100 ms), which required amplitudes as low as
0.2 A. A 16-channel voltage divider PC board was built to enable stimulation at levels
below 2 A. When switched on, the board divides the source voltage by a factor of 10
before relaying it to the current source (Fig. 2.25). An image of the assembled voltage
divider and current source boards is shown in Fig. 2.26 (left).
Figure 2.25: Circuit diagram of the switchable voltage divider and programmable current source.
The voltage output from the stimulus generator can be divided by 10 before being converted to
current.
Each MEA contains 60 electrodes, which can be addressed individually or in groups.
A custom PC board that rests on the microscope stage is used to interface with the
95
electrodes (Fig. 2.26, right) (Behrend, 2009). The board contains 60 DIP switches and
input sockets, each connected to a pogo pin that contacts a different MEA pad. The
board also houses the perfusion inlet and outlet glass capillary tubes, which keep saline
flowing through the recording chamber at a constant rate. The saline is warmed by an
in-line heater and temperature controller (Warner Instruments, Hamden, CT) prior to
entering the chamber. To maintain a tight interface between the retina and MEA, the
retina is held flat by a porous Teflon membrane (#JVWP01300; Millipore, Billerica, MA)
glued to a titanium ring. A platinum wire encircling the top of the recording chamber
serves as the return electrode.
Figure 2.26: Custom PC boards used to interface with the retina. Left: A 16-channel switchable
voltage divider (bottom) and programmable current source (top) tune the signals coming from the
stimulus generator. Right: A custom PC board interfaces with the multielectrode array. The
MEA and recording chamber are placed in the center of the board, which uses pogo pins to
connect with each of the 60 MEA pads. The retina is held flat on the MEA surface by a porous
membrane and is superfused with heated, oxygenated saline.
A single computer controls the electrical stimulation and image acquisition param-
eters. Stimulus protocols for the STG2008 are programmed in software provided by
the manufacturer (MC Stimulus II; Multi Channel Systems, Reutlingen, Germany). A
TTL sync channel drives the camera exposure trigger at 200-ms intervals. Another sync
96
channel triggers the oscilloscope to display the electrode voltage and current waveforms.
Images from the camera are acquired in Andor (Belfast, Northern Ireland) Solis and are
streamed directly to disk.
2.6.2 Electrode Arrays
To image calcium indicator fluorescence through the MEAs, they had to be constructed
from thin, transparent materials. MEAs used in prior studies were patterned from indium
tin oxide (ITO) on #1 cover glass substrates (Behrend et al., 2009, 2011). A 600-nm-thick
layer of silicon nitride (Si

N

) insulated the electrode traces from the overlying saline.
This insulation layer was sufficient in many cases; however, it was too thin for experiments
involving small, high-impedance electrodes. Charge buildup along the electrode traces
shunted a significant portion of current away from the electrodes.
2.6.2.1 Electrode Modeling
The electrode-electrolyte interface can be modeled by an RC circuit known as a Randles
cell (Fig. 2.27) (Randles, 1947). The classical Randles cell contains three circuit elements:
1. Double-layer capacitor. C
dl
models the charge redistribution that occurs as ions
in the electrolyte combine with the electrode. A plane of charge becomes trapped
at the electrode surface and is opposed by a plane of opposite charge in the elec-
trolyte. The two planes are separated by polar molecules (e.g., water) that act as
an insulator. This creates a capacitance at the interface, enabling charge to flow
without electron transfer (Merrill et al., 2005).
97
2. Charge transfer resistor. R
ct
models the Faradaic redox reactions that occur at
the electrode-electrolyte interface during charge transfer. These reactions can be
reversible or irreversible, depending on the relative rates of kinetics (electron trans-
fer at the interface) and mass transport (delivery of reactants to the interface).
When the rate of kinetics is high relative to the rate of mass transport, reaction
products remain close to the electrode surface, and reactions are reversible. In
the opposite case, reaction products irreversibly diffuse away from the electrode
interface, causing corrosion to the electrode (Merrill et al., 2005).
3. Solution resistor. R
sln
models the resistance of the bulk electrolyte. This resistance
depends only on the size of the electrode and the conductivity of the electrolyte
(Franks et al., 2005).
Figure 2.27: Randles cell model of the electrode-electrolyte interface (Randles, 1947). C
dl
=
double-layer capacitance, R
ct
= charge transfer resistance, and R
sln
= solution resistance.
Electrochemical impedance spectroscopy (EIS) is a technique used for characterizing
electrodes. A small-signal AC voltage is applied over a wide range of frequencies while
measuring the electrode current (amplitude and phase shift). Ohm’s law is then used
98
to calculate impedance at each frequency, which can be represented on a Bode plot
(Barsoukov and Macdonald, 2005):
=
()
()
=

0
cos()

0
cos(− )
=
0
cos()
cos(− )
(2.1)
Before building my own MEAs, I characterized a 60-m-diameter Pt/Ir electrode used
in our prior calcium imaging studies (Fig. 2.28). EIS measurements were performed with
a Gamry FAS1 potentiostat (Gamry Instruments, Warminster, PA) in 1x PBS. Impedance
spectra were determined from 10 mHz to 100 kHz in response to a± 10 mV
rms
(versus
Ag/AgCl reference) AC signal. To better simulate experimental conditions, a piece of
salamander retina was placed over the electrode. This likely caused an increase in the
impedance, especially at higher frequencies (Grill and Mortimer, 1994; Shah et al., 2007;
Stein et al., 1978).
Once EIS data is obtained, it is possible to determine the values of R
ct
, R
sln
, and C
dl
.
At low frequencies, C
dl
acts like an open circuit and R
ct
dominates (since R
ct
≫ R
sln
).
Thus, the impedance magnitude at 10 mHz (10 MΩ in Fig. 2.28) corresponds to R
ct
. At
high frequencies, C
dl
is shorted and R
sln
dominates; the high-frequency plateau (55 kΩ
in Fig. 2.28) corresponds to R
sln
. The value of C
dl
can be determined from the cutoff
frequency, f
c
, where the phase crosses− 45
∘ :

=
1
2 ×

×

(2.2)

=
1
2 ×

×

=
1
2 × 55 kΩ× 31.8 Hz
= 91 nF (2.3)
99
Figure 2.28: Impedance spectra of a 60-m-diameter Pt/Ir electrode. Model predictions are
plotted alongside the experimental data. Predictions for phase angle are inaccurate at low fre-
quencies due to the constant phase element behavior of C
dl
(MacDonald, 1984; McAdams et al.,
1995). As discussed below, deviations from the model at high frequencies arise from parasitic
capacitance associated with the electrode trace.
The electrode impedance can then be modeled by solving for the total impedance of
the Randles cell:

=
(︂

//
1

)︂ +

(2.4)
As shown in Fig. 2.28, the model predictions roughly align with the experimental EIS
measurements, though there are some exceptions. First, the prediction of phase angle
is inaccurate at low frequencies (< 1 Hz). This is due to C
dl
behaving like a constant
phase element rather than an ideal capacitor (MacDonald, 1984; McAdams et al., 1995).
Second, predictions of both magnitude and phase are inaccurate at higher frequencies.
100
This can be explained as follows: When current flows through the circuit, charge builds
up along the electrode trace and attracts opposite charge in the electrolyte. The opposing
charges are separated by a thin Si

N

insulation layer, creating a capacitance along the
trace. This so-called parasitic capacitance shunts current away from the electrode. Its
effect can be modeled by adding a parallel capacitor, C
shunt
, to the Randles cell (Fig.
2.29) (Shah et al., 2007). This changes the total impedance of the Randles cell to the
following:

=
[︂(︂

//
1

)︂ +

]︂ //
ℎ
(2.5)
Figure 2.29: Modified Randles cell incorporating a parallel shunt capacitor. C
dl
= double-layer
capacitance, R
ct
= charge transfer resistance, R
sln
= solution resistance, and C
shunt
= shunt
capacitance.
To solve Eq. 2.5, the value of C
shunt
needed to be determined. This was accomplished
by covering the electrode with a dab of silicone and measuring the trace’s capacitance
with an LCR meter (C
shunt
= 50 pF at 1 kHz). After solving for Z
tot
and plotting the
theoretical impedance, it was found to align better with the measured EIS data at high
frequencies (Fig. 2.30; compare with Fig. 2.28).
The shunt capacitance is a major problem for small electrodes (e.g., 15- or 30-m
diameter) that have high impedances. As electrode impedance increases, so does the
101
Figure 2.30: EIS spectra of a 60-m-diameter Pt/Ir electrode plotted alongside a Randles cell
model incorporating a shunt capacitor (Fig. 2.29). Compared to the original model (Fig. 2.28),
this new model aligns well with the EIS data at high frequencies.
amount of shunted current. When stimulating with short pulses (e.g., 0.06 ms), a rela-
tively long amount of time is spent charging the electrode trace, which prevents current
from being delivered to the tissue.
Shunt capacitance can be reduced by using thicker MEA insulation layers (capacitance
∝ 1/thickness). Silicon nitride is an excellent insulator for many applications, but it is
difficult to deposit thick films that are free of pinholes. Another good insulating material
is SU-8, an epoxy-based photoresist. SU-8 is biocompatible and can be deposited in a
wide range of thicknesses. Although it does not adhere well to MEA substrates, which
are made of glass, it does form good adhesion with silicon nitride. For this reason, SU-8 is
102
often deposited atop Si

N

to create a dual insulation layer (Ahuja et al., 2008a; Gholmieh
et al., 2006). These layers are mechanically stable and provide the thickness necessary
to prevent current shunting. By constructing my MEAs with a dual Si

N

/SU-8 layer, I
was able to reduce the shunt capacitance to a negligible amount.
2.6.2.2 Microfabrication Process Flow
MEAs were fabricated at the W.M. Keck Photonics Laboratory, a class 100 clean room.
Fabrication was a two-mask photolithographic process: The first mask was used to pattern
the ITO electrodes and their traces. The second was used to create vias in the insulation
layer over the electrodes and perimeter contact pads. All MEAs were constructed from
thin, transparent materials for imaging purposes.
Processing began with sheets of ITO-coated #1 cover glass (12 Ω/square, 300-nm-
thick ITO layer; Vaculayer, Mississauga, ON, Canada). Individual substrates were formed
by cutting sheets into 28× 28 mm slides with a dicing saw. A sacrificial layer of Si

N

(180
nm thick) was deposited atop the ITO by plasma-enhanced chemical vapor deposition
(PECVD) at 275
∘ C for 15 min (30 W, 425 mTorr, flow rates of 40 sccm for silane, 20 sccm
for NH
3
, and 60 sccm for N
2
). AZ1518 positive photoresist (Clariant, Somerville, NJ)
was spin coated on the Si

N

using an HMDS adhesion promoter. Slides were soft baked
at 100
∘ C for 40 sec to remove excess solvent, aligned with the first photolithographic
mask, and exposed to UV light (365 nm, 70 mJ/cm
2
) (Fig. 2.31, left). Photoresist was
developed with AZ 400K (Clariant, Somerville, NJ) to define the electrodes, traces, and
perimeter contact pads. The uncovered Si

N

was removed with CF
4
reactive ion etching
(RIE) for 8 min (100 W, 200 mTorr) in order to expose the underlying ITO (Fig. 2.31,
103
middle). The ITO was then patterned by wet etching with aqua regia (60% HCl, 5%
HNO
3
, 35% H
2
O). The remaining photoresist and sacrificial Si

N

were removed with
acetone and RIE, respectively. At this point, all that remained was the patterned ITO
and the cover glass substrate (Fig. 2.31, right).
Figure 2.31: Microfabrication process flow used to pattern the ITO electrodes and traces on the
cover glass substrates.
To create the dual Si

N

/SU-8 insulation layer, a 300-nm-thick film of Si

N

was
deposited by PECVD at 275
∘ C for 30 min (30 W, 450 mTorr, same flow rates as above).
The film was ashed with O
2
RIE for 1 min to clean the surface (100 W, 100 mTorr).
SU-8 2001 negative photoresist (MicroChem, Newton, MA) was spun onto the Si

N

to
a thickness of 1.5–1.75 m. Slides were soft baked at 75
∘ C for 30 min to evaporate
the SU-8 solvent. They were then aligned with the second photolithographic mask and
exposed to UV light (365 nm, 120 mJ/cm
2
) (Fig. 2.32, left). A post-exposure bake was
performed for 2 min at 95
∘ C to cross-link the SU-8 polymer. Slides were then immersed
in SU-8 developer (MicroChem, Newton, MA) to dissolve the SU-8 above the electrodes
and perimeter contact pads. Undeveloped SU-8 was cured with a hard bake at 200
∘ C
for 25 min. To expose the ITO, Si

N

was dry etched with CF
4
RIE for 5 min (100 W,
200 mTorr) (Fig. 2.32, right). A final O
2
RIE ashing step was performed to clean the
ITO surface (20 sec, 100 W, 100 mTorr).
104
Figure 2.32: Microfabrication process flow used to pattern the dual Si

N

/SU-8 insulation layer.
Following fabrication, MEA slides were bonded with silver epoxy (CW2400; ITW
Chemtronics, Kennesaw, GA) to a custom PCB that mates with the interface board on
the microscope stage (Fig. 2.33, left). All electrodes≤ 30 m in diameter, as well as
some larger ones, were electroplated with Pt/Ir to increase their charge injection limits
(also rendering them opaque). Plating was performed with cyclic voltammetry from
− 0.55 to +0.9 V versus an Ag/AgCl reference electrode (100 mV/sec sweep rate). MEA
construction was completed by gluing a polyoxymethylene recording chamber to the array
surface (Fig. 2.33, right).
Figure 2.33: MEA slides were attached to a custom PCB that mates with the interface board on
the microscope stage. Left: ITO contact pads were bonded to traces on the bottom of the PCB
with conductive epoxy. Right: A recording chamber was glued to the MEA surface. When placed
in the interface board, pogo pins make contact with the 60 pads around the PCB perimeter. Each
pad is wired to a different electrode.
105
2.6.2.3 MEA Layouts
MEA layouts were designed to enable specific types of experiments. Electrode diameter
was varied from 10 to 200 m in order to investigate the effect of electrode size on RGC
responses. Studies involving 200-m-diameter electrodes were particularly important, as
that size is used by the Argus II. To enable multielectrode pattern stimulation experi-
ments, smaller electrodes (15-, 30-, and 75-m diameter) were arranged in hexagonal and
rectangular grids. Groups of electrodes could then be activated to form different shapes,
such as lines and letters. Fig. 2.34 shows photomicrographs of three MEAs that were
used in my experiments.
Figure 2.34: Photomicrographs of three MEAs. Each array contains 60 ITO electrodes, some of
which are plated with Pt/Ir. Scale bars are 200 m. A: 75-m-diameter electrodes on a 150-m
pitch (upper left). 30-m-diameter electrodes on a 75-m pitch (lower right). B: 15-m-diameter
electrodes on a 50-m pitch. All 60 electrodes can be imaged in a single 20x field of view. C :
Various electrode sizes ranging from 10 to 200 m in diameter.
106
2.6.2.4 Electrode Characterization
Following MEA construction, electrodes of different sizes were characterized by EIS (Fig.
2.35). Impedance spectra were measured for 20-, 60-, 100-, and 200-m-diameter elec-
trodes (10 mHz to 100 kHz in 1x PBS). Data were obtained for both plated (Pt/Ir) and
unplated (ITO) electrodes.
As shown in Fig. 2.35, impedance was clearly affected by electrode size. Consistent
with prior reports (Ahuja et al., 2008b; Shah et al., 2007), smaller electrodes had higher
impedances over the entire frequency range. For a given stimulus current, this increased
impedance will a cause larger voltage drop across the electrode-electrolyte interface. If
large enough, irreversible Faradaic reactions will release toxic byproducts into the elec-
trolyte, causing damage to both the electrode and the tissue (Brummer et al., 1977).
Fig 2.35 also shows how impedance is affected by electrode material. Pt/Ir electrodes
have lower impedances than ITO electrodes of the same size, but only at lower frequencies.
For a given electrode size, the impedance magnitude spectra for Pt/Ir and ITO converge
at∼ 100 kHz, where R
sln
dominates. This happens because electroplating increases surface
roughness, affecting both R
ct
and C
dl
. Since R
sln
depends on geometric surface area rather
than real surface area, it is unaffected by surface roughness (Franks et al., 2005).
It is important to note that the effects of C
shunt
are no longer present in the EIS
spectra, even for the smallest electrodes. This is evidenced by the lack of high-frequency
poles, such as the one in Fig. 2.29 at 40 kHz. It therefore apparent that the dual
Si

N

/SU-8 insulation layer was effective in reducing the shunt capacitance to a negligible
amount.
107
Figure 2.35: EIS spectra of 20-, 60-, 100-, and 200-m-diameter electrodes (plated and un-
plated). Impedance magnitude decreases with increasing electrode size. For frequencies domi-
nated by R
ct
and C
dl
, Pt/Ir electrodes have lower impedances than ITO electrodes of the same
size.
108
2.7 Data Collection and Processing
2.7.1 Retina Preparation
Calcium imaging experiments were conducted using two different animal models. Larval
tiger salamanders (Ambystoma tigrinum) were used to study the effect of interphase gap
on RGC thresholds (Chapter 3). These experiments were performed with synthetic Ca
2+
dye before I found a method for transducing mammalian RGCs with GECIs. Detailed
protocols for dissection and dye loading in salamander are given elsewhere (Behrend et al.,
2009).
The remainder of experiments were performed in adult rats. Retinas were harvested
2–4 weeks after injecting AAV2-CAG-GCaMP into their vitreous cavities (see Section
2.4.3). Since the extent of GCaMP transduction depended on the quality of injection,
retinas from both eyes were kept alive until their expression profiles could be examined.
All procedures were approved by the Institutional Animal Care and Use Committee and
the Institutional Biosafety Committee at the University of Southern California.
Rats were deeply anesthetized by intraperitoneal injection of ketamine (60 mg/kg)
and xylazine (8 mg/kg). The first eye was enucleated, and connective tissue around the
optic nerve was trimmed with dissection scissors. A razor blade was used to make make
a small incision through the sclera and retina, 1 mm posterior to the corneal limbus. The
optic nerve was cut to a length of 0.5 mm, and a segment of thin-wall polyimide tubing
was glued around it. Next, a small circular sheet of polystyrene (∼ 1 cm diameter) with
a hole in the center was glued around the tubing to the posterior sclera. DI water was
pipetted into the tubing to keep the optic nerve nerve wet. The eye was then submerged
109
in calcium-free Ames’ medium (Baldridge, 1996) and hemisected with dissection scissors.
Vitreous was removed with a custom extractor (Sekirnjak et al., 2006) to allow for a tight
interface between the retina and MEA. Next, the posterior eyecup was transferred to a
Petri dish, and the saline in the tubing was replaced with Alexa 594 hydrazide sodium
salt (30 mM, 1.5 L; Life Technologies, Grand Island, NY). The assembly was then
moved to a perfusion chamber and left for 1 hour to enable retrograde loading of the dye.
Labeling the ganglion cells with a red dye was important for determining whether cells
transduced with GCaMP were RGCs or other cell types. During dye loading, the eyecup
was superfused at 29–30
∘ C and a flow rate of 4–5 mL/min. Shortly after transferring the
first eye to the perfusion chamber, the rat was rapidly decapitated, and all procedures
were repeated on the second eye.
Following dye loading, the retina from the first eyecup was dissected out. Four relief
slits were made so that the tissue could lie flat. The retina was then mounted on a porous
membrane attached to a titanium ring. The ring was inverted and placed on an MEA such
that the ganglion cell side faced down. The MEA was transferred to the microscope, and
GCaMP expression profiles were examined by fluorescence imaging. If poor transduction
was observed, the retina was discarded, and the second retina was used. Once a region
with dense GCaMP expression was identified, a smaller piece of retina containing that
region was isolated and used for recording. Pieces larger than∼ 5× 5 mm were difficult to
work with, as they did not always lie flat on the MEA. When isolating a piece of retina,
the optic disc was always left intact so that its position could later be identified during
data processing.
110
Retinas were superfused with bicarbonate-buffered Ames’ medium (Sigma-Aldrich,
Saint Louis, MO) during the course of each experiment (30–33
∘ C, flow rate of 4–5
mL/min). Media was supplemented with penicillin-streptomycin to prevent bacterial
growth, equilibrated with 5% CO
2
/95% O
2
gas, and adjusted to pH 7.4 and 280 mOsm.
Ames’ medium used during dissection contained 2.5-mM MgCl
2
and 5-mM EGTA in
order to chelate calcium (Baldridge, 1996) and prevent Ca
2+
-dependent resealing of the
optic nerve (Yawo and Kuno, 1985).
2.7.2 Data Acquisition
The goal of each experiment was to determine electrical thresholds of RGCs for a given
set of stimulus parameters, such as pulse width, waveform shape, and electrode size.
Calcium indicators were used to report RGC activity during stimulation. Threshold was
defined as the minimum amplitude needed to evoke a calcium response. Simultaneous
recordings were made from approximately 300 RGCs at a time, depending on the density
of indicator labeling.
Stimulation protocols were different for rat and salamander, since calcium indicators
with different sensitivities were used in each animal. Experiments in salamander were
performed with OGB-1 488 dextran 10 kDa, which is far less sensitive than the GECI
GCaMP5G used in rat (see Fig. 2.19). Stimulus protocols for salamander were identical
to those described by Behrend et al. (2009; 2011), except for the changes discussed in
Section 3.2.1. Protocols for rat were optimized for GCaMP5G and are given below.
Each stimulation protocol (Fig. 2.36) lasted 10 minutes. During this time, current of
progressively increasing amplitude was delivered to the retina. The minimum amplitude
111
was chosen to be subthreshold, and the maximum amplitude was dictated by the stimula-
tion hardware and electrochemical safety limits. A total of 10 amplitudes were used, each
lasting 1 minute. For each amplitude, 14 identical stimuli were delivered on 3.6-second
intervals. The purpose of repeating the stimuli was to obtain redundancy and reduce
detection noise. No stimuli were delivered during the first 5 and last 4.6 seconds of each
minute; data collected during these times represented baseline noise levels, which were
later used to set response detection thresholds relative to the noise (see Section 2.7.4).
Figure 2.36: Stimulation protocol used for GCaMP5G in rat (adapted from Behrend, 2009).
Stimuli were delivered 14 times on 3.6-second intervals and were repeated over 10 monotonically
increasing amplitudes. Each stimulus was a burst of pulses, a square wave, or a sine wave designed
to evoke a burst of spikes and generate a detectable calcium transient.
All stimuli were repetitive in order to evoke a burst of spikes and generate a detectable
calcium transient (Akerboom et al., 2012; Behrend et al., 2009; Borghuis et al., 2011; Tian
et al., 2009). Rectangular pulse trains (20 pulses, 120-ms duration), square waves (5–
125 Hz,≤ 200-ms duration), and sine waves (20 Hz, 200-ms duration) were applied in
different experiments. Rectangular pulses consisted of a cathodic phase followed by an
112
anodic phase of twice the duration and half the amplitude. Square waves and sine waves
were also cathodic-first. All stated amplitude and pulse width values are for the cathodic
phase.
Epifluorescence images were continuously acquired at 5 Hz with an 85% exposure
duty cycle. Images were either synchronized with the onset of each stimulus or were
phase shifted by 20 ms relative the onset. Stimulus durations were always≤ 200 ms,
appearing nearly instantaneous relative to the rate of image acquisition.
2.7.3 Image Data Reduction
Images collected from each experiment typically totaled tens of gigabytes. To reduce the
file size, ganglion cell fluorescence intensities and coordinates were extracted from each
image series. RGCs were marked in ImageJ (NIH, Bethesda, MD) with a custom tool
written in-house (Fig. 2.37).
1
For mammalian experiments, only cells double-labeled
with GCaMP5G and Alexa 594 were marked to ensure that they were RGCs (see Fig.
2.14). Once all regions of interest were defined, ImageJ’s Multi Measure plugin was used to
extract the center coordinates and average fluorescence intensity from every marked region
in each image frame. These data were imported into MATLAB (MathWorks, Natick,
MA), which was used for the remainder of data processing. Except for the identification
of ganglion cell bodies, all processing steps were automated and were identical for each
dataset.
1
The Soma ROI Tool was uploaded to the NIH website and is available from http://rsbweb.nih.
gov/ij/macros/tools/SomaRoiTool.txt.
113
Figure 2.37: Ganglion cell bodies were marked in ImageJ. For mammalian experiments, only
cells double-labeled with GCaMP5G (green) and Alexa 594 (red) were marked to ensure that
they were RGCs. The center coordinates and average fluorescence intensity of each region were
extracted from every image frame.
2.7.4 Calculating RGC Thresholds
Electrically evoked responses were detected as sudden increases in calcium fluorescence
temporally correlated with the stimuli. Raw fluorescence signals from each cell were high-
pass filtered to reveal sharp rises in fluorescence. Filtering was performed by convolving
the signals with a difference filter, [2 1− 1− 2]. The difference-filtered waveform for each
cell was then thresholded at the RMS noise level for that cell. Any point that crossed
this threshold was considered a possible calcium response (Fig. 2.38).
Thresholding at the RMS noise level gave a fairly large number of false positives
(threshold crossings that were not due to stimulation-evoked calcium transients). Since
GCaMP5G signals were so large, one possibility was to threshold at a multiple of the
RMS noise. This was indeed attempted, however it resulted in stimulation thresholds
114
Figure 2.38: Top: Normalized fluorescence traces from a single RGC in response to stimulation
with three amplitudes (0.5, 1.1, and 1.6 A). At each amplitude, 14 stimuli were delivered on
3.6-second intervals. Responses became progressively larger as amplitude increased. Bottom:
Fluorescence traces were high-pass filtered and thresholded at the RMS noise level (horizontal
lines) to detect possible calcium responses.
becoming highly variable. It was found that this variability arose from differences in
GCaMP5G expression levels among RGCs, such that higher expression levels caused
larger calcium transients. By thresholding at 1× the RMS noise, stimulation thresholds
among cells/retinas remained independent of GCaMP5G expression.
False positives were rejected by binning calcium responses (threshold crossings) ac-
cording to latency into post-stimulus time histograms (Fig. 2.39). For each amplitude,
the response percentage (i.e., the fraction of stimuli that evoked calcium transients) was
calculated by averaging the number of responses occurring within two image frames of
stimulus onset and subtracting the average number of responses occurring after these
frames. In other words, response percentage represents the number of calcium responses
115
temporally correlated with the stimuli, less the number of uncorrelated responses (false
positives), normalized to the number of stimuli.
Figure 2.39: Post-stimulus time histograms for the RGC in Fig. 2.38. Histogram bin size is 200
ms, equivalent to the camera exposure period. As stimulation amplitude increased (from left to
right), more responses accumulated in the first two bins. Responses occurring in later bins were
assumed to be false positives.
After calculating response percentages, a dose response curve was generated for every
RGC by plotting response percentage against stimulation amplitude. A sigmoidal func-
tion (Eq. 2.6) was fit to each curve, and threshold was defined as the amplitude that
yielded a 50% response. Threshold for the RGC in Figs. 2.38 and 2.39 was found to be
0.92 A.
=
1
1 +
− (−
0
)
(2.6)
For experiments involving GCaMP5G, there were often diffuse calcium responses from
fluorescent RGC dendrites that were out of focus to the camera. These responses inter-
fered with the fluorescence signals of other cells that were overlapped by the dendrites. In
some cases, this could cause a cell which did not respond to stimulation to be classified as
responding (if the dendritic response was large enough). To prevent these false positives,
a minimum threshold was set as a multiple of the RMS noise level. The specific multiple
was chosen based on the overall magnitude of calcium responses but was always at least
3× the RMS noise (dendritic responses rarely reached this level). The response percent-
age was then recalculated for every cell (see above) using the new threshold. Any cells
116
that did not have a 50% response percentage for at least one amplitude were classified
as non-responding. For responding cells, thresholds were always determined from the
value calculated using 1× the RMS noise (multiples of the RMS noise were used only for
rejecting false positives).
2.7.4.1 Repeatability of the Threshold Measurement
Repeatability of the threshold measurement was assessed by performing two identical
stimulation runs 45 minutes apart. Thresholds were calculated for 74 rat RGCs labeled
with GCaMP5G (Fig. 2.40). Thresholds between runs were found to change by only
− 3.9± 10.0% (correlation coefficient = 0.92). This is not significantly different from
threshold changes measured previously in salamander retina with OGB-1 488 dextran 10
kDa ( = 0.16, unpaired t-test) (Behrend, 2009). Overall, the threshold measurement
appears to be highly repeatable, regardless of animal model or calcium indicator type.
The small differences in thresholds might arise from variance in individual cells, system
noise, or adaptation of the retina to stimulation (Behrend, 2009).
2.7.5 Threshold Maps
Once stimulation threshold is calculated for every RGC, a spatial threshold map is gener-
ated (Fig. 2.41A). These maps are the primary output of every experiment. They provide
a visual representation of RGC thresholds relative to the position of the stimulating elec-
trode and optic disc.
The size of each threshold map is limited by the imaging objective’s field of view,
which is roughly 400× 400 m. It was often necessary to measure thresholds from larger
117
Figure 2.40: Changes in RGC thresholds were measured between two identical stimulation runs
spaced 45 minutes apart. Data were collected from 74 rat RGCs labeled with GCaMP5G. Stimulus
pulse width was 0.4 ms. Thresholds between runs were found to change by only− 3.9± 10.0%.
areas, especially when imaging the effects of axonal stimulation. To accomplish this, two
or more identical stimulation runs were imaged from adjacent fields of view. Threshold
maps from individual fields were then stitched together, as shown in Fig. 2.41 A.
The large fluorescence signals provided by GCaMP enabled patterns of electrical ac-
tivity to be viewed during stimulation. In addition to threshold maps, background-
subtracted images were generated from every stimulation run for the highest amplitude
(Fig. 2.41C ). The 14 images captured immediately after each stimulus event were aver-
aged, and the baseline fluorescence intensity of every pixel was subtracted. Brightness
and contrast were globally adjusted to accentuate the responses.
To identify trends in the threshold responses, it was helpful to combine threshold maps
from multiple experiments involving the same stimulus parameters. Since the orientation
118
of axons relative to the stimulating electrode and optic disc varied among experiments,
maps had to be aligned to the same reference frame before thresholds could be pooled.
Each map was therefore shifted and rotated such that the electrode lay at the center
and the optic disk was positioned to the left. Cells were binned in a grid according to
spatial location relative to the electrode, and thresholds in each bin were averaged (Fig.
2.42A). Background-subtracted images were also combined by transforming images from
each experiment to the same reference frame and averaging (Fig. 2.42B).
As discussed in Section 2.3, thresholds measured in salamander retinas varied from one
experiment to another by nearly twofold. Salamanders are harvested from the wild, and
their retinas are affected by both seasonal changes and age (which cannot be controlled).
In order to combine threshold maps from different salamander experiments, each map
had to be normalized before thresholds were averaged (Behrend, 2009).
Thresholds in rat were found to be much more uniform. Under the same experimental
conditions, thresholds among animals generally varied by≤ 25–30%, though often much
less. This was true of both wild-type rats and retinal degenerative models (of similar
age). It was therefore not necessary to normalize thresholds when combining maps from
multiple rat retinas. It is possible that thresholds would have been even more uniform
if an inbred strain were used, though issues with GCaMP-induced cytomorbidity would
need to be resolved (see Section 2.5.5).
119
Figure 2.41: A: Threshold map of rat RGCs in response to stimulation with 0.4-ms pulses.
Stimuli were delivered by a transparent 200-m-diameter electrode (teal circle). RGCs are drawn
as colored circles according to their stimulus thresholds. Empty circles indicate cells that did not
respond to stimulation. Data were acquired from two adjacent image frames and stitched together.
Of the 461 cells that were imaged, 239 responded to stimulation. B: Baseline fluorescence images of
GCaMP5G expression in the same two fields of view asA. C : Background-subtracted fluorescence
responses to the highest stimulus amplitude (6.0 A). The electrode perimeter in B and C is
outlined in gray.
120
Figure 2.42: A: Threshold maps from three rat retinas (1253 somata) were combined into a
single composite map. Maps from each retina were rotated and shifted into the same reference
frame such that the optic disc lies to the left of each image. Cells were binned into a grid,
and thresholds of cells in each grid bin were averaged. The map shows RGC responses to 0.4-
ms pulses delivered by 200-m-diameter electrodes (black circle). RGCs are drawn as colored
circles according to their stimulus thresholds. Empty circles indicate cells that did not respond
to stimulation. Small, medium, and large circles indicate 1–2, 3–4, and 5+ cells, respectively.
B: Background-subtracted fluorescence responses to the highest stimulus amplitude. Images from
each retina were transformed into the same reference frame and averaged. The electrode perimeter
is outlined in gray.
121
Chapter 3
Using Interphase Gaps to Lower Thresholds
3.1 Introduction
The goal of my experiments was to gather data that would help improve the design and
operation of epiretinal prostheses for human patients. Thus, many of my experiments
were motivated by specific problems in the clinic. One such problem relates to the safety
limits of electrical stimulation. In many patients, a significant number of electrodes are
unable to evoke percepts without exceeding their charge injection limits (de Balthasar
et al., 2008; Humayun et al., 2012). A recent study with 30 Argus II recipients found
that an average of 44.5% electrodes required more current to elicit phosphenes than
was electrochemically safe (Humayun et al., 2012), effectively rendering those electrodes
unusable. This prompted me to investigate methods for lowering ganglion thresholds,
which would potentially enable some electrodes to elicit phosphenes that were previously
unable to do so.
Reducing thresholds would also have other important implications for the prosthesis.
Clinical testing has shown that phosphenes tend to get larger and brighter as stimulation
122
amplitude increases (de Balthasar et al., 2008; Greenwald et al., 2009; Humayun et al.,
2003; Nanduri et al., 2012). Dynamic range of phosphene size and brightness is therefore
defined as the difference between an electrode’s safety limit and its threshold current.
Lowering threshold current would cause dynamic range to increase accordingly. It would
also reduce the overall power consumption of the prosthesis, since less current would be
required to elicit responses.
Most retinal implants stimulate with symmetric biphasic current pulses (see Section
1.5) in order to maintain a balance of charge. A drawback to using biphasic pulses is
that the trailing phase partially opposes the depolarization induced by the leading phase,
thereby increasing the stimulus amplitude needed to evoke an action potential. Prior
studies have found that adding an interphase gap between the two phases (see Fig. 1.7)
reduces this effect. Interphase gaps have been used to decrease electrical thresholds in
auditory nerve (Prado-Guitierrez et al., 2006; Shepherd and Javel, 1999) and motor nerve
(Gorman and Mortimer, 1983; van den Honert and Mortimer, 1979), as well as perceptual
thresholds in cochlear implant patients (Carlyon et al., 2005; McKay and Henshall, 2003).
Interphase gaps were also used by the Argus I (de Balthasar et al., 2008; Greenwald
et al., 2009; Mahadevappa et al., 2005), though a formal study of their effect was never
conducted. These gaps were later eliminated from Argus II trials for reasons that are
unclear (Horsager et al., 2009; Nanduri et al., 2012).
I performed a thorough study to investigate the effect of interphase gap on ganglion
cell electrical thresholds (Weitz et al., 2011). Calcium imaging was used to report RGC
activity during stimulation with different IPG lengths. Results were corroborated through
123
computational modeling. Testing was also performed in five Argus II subjects to measure
the effect of interphase gap on phosphene perceptual thresholds.
3.2 Methods
3.2.1 Animal Experiments
All experiments were conducted in tiger salamander retina ( = 5) with OGB-1 488
dextran 10 kDa. Electrode size and pulse width were chosen to match those of the Argus
II (200-m-diameter electrodes, 0.46-ms pulses). Interphase gap duration was varied
between 0.12 ms (26% of pulse width) and 1.84 ms (400% of pulse width). Each cell’s
threshold at a particular gap length, , was compared to its threshold when no IPG was
used:
change in threshold =

=
−
=0

=0
× 100% (3.1)
Thresholds were calculated as described in Section 2.7.4. All stimulus and data acquisition
parameters were identical to those described by Behrend et al. (2009; 2011), except for
stimulus burst length (reduced from 40 to 30) and frequency (reduced from 333 Hz to
167 Hz). Spacing pulses farther apart in time was necessary in order to limit interference
between successive pulses (see Fig. 3.5).
3.2.2 Computational Modeling
I generated a Hodgkin-Huxley-type model to further study the effect of IPG on ganglion
cell electrical thresholds. Equations and parameters were taken from prior physiology and
modeling studies of tiger salamander (Fohlmeister and Miller, 1997; Fohlmeister et al.,
124
1990) and cat (Benison et al., 2001) RGCs. As shown in Fig. 3.1, the salamander
model incorporates four types of voltage-gated ion channels (Na
+
, delayed rectifier K
+
,
A-type K
+
, and L-type Ca
2+
), as well as Ca
2+
-activated K
+
channels. The cat model
includes these same channels in addition to N-type Ca
2+
channels. Differential equations
for membrane voltage (Eq. 3.2) and intracellular calcium concentration (Eq. 3.3 for
salamander, Eq. 3.4 for cat) were solved in MATLAB (MathWorks, Natick, MA) using
the forward Euler method with a time step of 0.01 ms. Current clamp was used to
simulate extracellular stimulation.

=
1

(−

−

−

−

−

−

−

−

+

) (3.2)
[
2+
]

=
− 3

2
− [
2+
]

− [
2+
]

,
where = the Faraday constant,
3

= surface area-to-volume ratio of the cell,
2 = valency of calcium,
[
2+
]

= residual calcium level,
and = time constant of calcium removal/sequestration
(3.3)
[
2+
]

=

(︂ −

+

2

− [
2+
]

− [
2+
]

−

)︂ ,
where

= fraction of free calcium in the cytosol,

= volume of the cell,
and

= loss of calcium due to Ca
2+
pumps
(3.4)
Solving the differential equations required values for the channel currents,

, at each
time step. These currents can be expressed in terms of Ohm’s law:

=

(−

), (3.5)
125
Figure 3.1: Components of the Hodgkin-Huxley-type model include four types of voltage-gated
ion channels (Na
+
, delayed rectifier K
+
, A-type K
+
, and L-type Ca
2+
), Ca
2+
-activated K
+
channels, leak channels, a membrane capacitance, and a current stimulation source. N-type Ca
2+
channels (not pictured) were included only in the cat model.
where

is the channel conductance and

is the reversal potential. The reversal po-
tential is constant for every channel type except Ca
2+
(

must be updated to reflect
changes in [Ca
2+
]
i
, which is done using the Nernst equation). Voltage-gated conductances
are governed by the activation and inactivation kinetics of each channel:

= ¯

ℎ

, (3.6)
where

is the maximum conductance, is the activation gating variable, ℎ is the
inactivation gating variable, and and are integers. The Ca
2+
-activated K
+
channel
is unique in that it is gated by calcium rather than voltage. Its conductance depends on
both the intracellular calcium concentration and the binding affinity of Ca
2+
ions to the
channel (equations given in Benison et al., 2001; Fohlmeister et al., 1990).
The model’s stimulation parameters were set to match the experimental conditions
as closely as possible. A burst of 30 pulses was delivered at 167 Hz. Pulse width was
0.46 ms/phase, and IPGs ranged from 0.12 to 1.84 ms. The cell’s membrane potential
was calculated in response to each stimulus burst. Action potentials were identified
126
as peaks in the membrane potential that exceeded 20 mV (salamander) or 30 mV (cat).
Threshold was defined as the minimum stimulus amplitude that caused at least 50% of the
pulses in the burst to elicit action potentials. Fig. 3.2 shows the membrane voltage and
intracellular calcium concentration of a modeled salamander RGC, computed in response
to a stimulus burst at threshold. As was the case with the animal experiments, threshold
at a particular IPG length was compared to threshold when no IPG was used. Adding
10% uniformly distributed pseudorandom noise to the model parameters had minimal
effect on predicted threshold changes.
Figure 3.2: Membrane voltage (top) and intracellular calcium concentration (bottom) of a model
salamander RGC in response to a stimulus burst (middle). Fifteen of the 30 stimuli in the burst
evoked action potentials, a condition that was defined as threshold.
127
3.2.3 Human Experiments
The effect of interphase gap on perceptual threshold (the current required to evoke a
phosphene) was measured in five Argus II subjects. Two electrodes were tested in each
patient. Electrodes were chosen based on prior knowledge of their thresholds. We se-
lected electrodes with relatively low thresholds to limit the possibility of exceeding elec-
trochemical safety limits as patients adapted to stimulation (see below). Thresholds were
measured with a method of adjustment procedure, as described previously (de Balthasar
et al., 2008). All testing was performed by Second Sight Medical Products, Inc. (Sylmar,
CA). Informed consent was obtained from every patient, and all procedures complied
with the Declaration of Helsinki.
Stimulation parameters used for human testing were similar to those used in the
salamander experiments (Table 3.1). Each stimulus was a burst of 30 pulses delivered at
120 Hz (the maximum allowable frequency of the prosthesis). Pulse width was fixed at
0.46 ms/phase, and IPG was varied between 0.15 and 2.0 ms in a random order. Pulses
containing no interphase gap were also presented in order to obtain baseline thresholds.
Table 3.1: Comparison of stimulation parameters used in the salamander and human interphase
gap experiments.
Salamander Human
IPG Lengths (ms)
0, 0.12, 0.24, 0.34,
0.46, 0.92, 1.38, 1.84
0, 0.15, 0.46, 1.0, 2.0
Pulse Width (ms) 0.46 0.46
# of Pulses per Burst 30 30
Pulse Frequency (Hz) 167 120
Electrode Diameter (m) 200 200
Threshold Electrical Perceptual
128
Initial testing with one patient revealed that thresholds gradually rose during the
course of the test session, implying that responses were becoming desensitized. Such
desensitization has been reported previously in prosthesis patients; the brightness of per-
cepts gradually fades during sustained stimulation (Perez Fornos et al., 2010; Zrenner
et al., 2011). The mechanism of this fading is still unknown, however electrophysiology
studies have suggested that it arises from amacrine cell inhibition (Freeman and Fried,
2011), depletion of synaptic vesicles (Jensen and Rizzo, 2007), and/or RGC Na
+
channel
inactivation (Tsai et al., 2011). Although these mechanisms may contribute to desensi-
tization, phosphene fading has also been reported during intracortical microstimulation
(Schmidt et al., 1996). Therefore, adaptation might result from a combination of both
retinal and cortical mechanisms.
Response desensitization was a problem for our interphase gap experiments, since it
masked any possible threshold-lowering effects of IPG (Fig. 3.3). To overcome this, we
used a bracketing technique in which each IPG threshold was measured between two
stimulus trials with gapless pulses. The effect of IPG was then determined by comparing
the threshold measured from the IPG stimulus to the average of the two thresholds
measured when no gap was used. (For example, the order of IPGs tested in a patient
might be: 0 ms, 1.0 ms, 0 ms, 0.15 ms, 0 ms, 2.0 ms, 0 ms, 0.46 ms, 0 ms.) After
implementing this method, the effects of interphase gap on perceptual thresholds became
much more pronounced.
129
Figure 3.3: Effects of threshold adaptation in a human prosthesis patient. Left: Perceptual
thresholds were measured in response to pulses containing different interphase gap lengths. The
order in which pulses were presented is shown on the horizontal axis (from left to right). Adap-
tation clearly masks any effects IPG has on threshold. Right: Five threshold measurements were
performed with identical pulses containing no IPG. Thresholds steadily rose over time, indicating
that responses were becoming desensitized.
3.2.4 Curve Fitting
All curves were fit with decaying exponentials (Eq. 3.7). Fits were optimized to minimize
the sum of squared error. Weighted fits were used for the salamander and human data,
with the weights defined as the reciprocal of the variance.
=
− /
+ (3.7)
3.3 Results
Results from the salamander experiments, computational modeling, and human testing
consistently showed that including an interphase gap decreases thresholds. Fig. 3.4
summarizes the effect of IPG duration on RGC electrical thresholds for the animal exper-
iments and computational models. In the animal data, IPG durations 0.24 ms and longer
caused a significant reduction in threshold versus no IPG ( < 0.05, unpaired t-tests).
130
With an IPG duration equivalent to the pulse width (0.46 ms), RGC threshold was 21.1
± 13.9% lower than when no gap was used ( < 0.01). Gap lengths longer than 0.46 ms
further decreased threshold, but only marginally: With 0.92-, 1.38-, and 1.84-ms gaps,
mean change in threshold fell by only an additional 3.8%. Pulses with gaps shorter than
0.46 ms were less effective at stimulating RGCs. For example, an IPG duration of 0.24 ms
(52% of pulse width) lowered threshold by only 12.4± 16.4% versus no gap ( < 0.05).
Figure 3.4: Effect of interphase gap duration on ganglion cell electrical thresholds. Data from
five salamander retinas (green circles) indicated that IPGs reduce thresholds by up to∼ 25%.
Computational modeling also predicted the threshold-lowering effects of IPG. Two models were
generated: one for a salamander RGC (purple triangles) and another for a cat RGC (red di-
amonds). The salamander model aligned much more closely with the animal data. Decaying
exponentials were fit to each data set. Error bars indicate SEM. * means < 0.05, and ** means
< 0.01.
Consistent with the animal data, the computational models also predicted lower
thresholds with larger IPGs. The effects of interphase gap were more pronounced in
131
the cat model than in the salamander model, especially with longer gap lengths. Though
the reason for this is unclear, discrepancies likely arise from the different distributions of
ion channels in each modeled cell. As anticipated, predictions from the salamander model
were more in line with the experimental data (which was also from salamander retina);
predicted change in threshold fell within the standard deviation of the animal data for
every IPG duration that was studied.
The salamander model revealed an interesting behavior that may have implications
for high-frequency stimulation with long IPGs. As shown in Fig. 3.4 (purple triangles),
1.38-ms gaps were more effective at stimulating the model RGC than 1.84-ms gaps. I
hypothesized that this was due to interactions between successive pulses; as IPG becomes
longer, pulses become closer in time. This causes a blocking effect, in which the anodic
phase of one pulse raises the threshold of action potential initiation for the next pulse.
It has been suggested that this block results from the anodic phase depolarizing or hy-
perpolarizing the cell membrane, which affects sodium channel sensitivity (Bhadra and
Kilgore, 2004; Ranck, 1975).
To test this hypothesis, I solved the model using lower stimulation frequencies (154
and 143 Hz), spacing pulses farther apart in increments of 0.5 ms (Fig. 3.5). As expected,
the blocking effect was present only at longer IPG lengths and was reduced as stimulus
frequency decreased. By solving for the membrane voltage of the modeled cell in response
to a single biphasic pulse, it became clear that the anodic phase was hyperpolarizing the
cell membrane (Fig. 3.5, inset).
Having shown that interphase gaps can reduce ganglion cell electrical thresholds, the
next step was to test their effect on perceptual thresholds in human prosthesis users.
132
Figure 3.5: Modeling the effect of stimulus frequency and IPG length on anodic block. The
blocking effect is most pronounced when pulses are spaced close together (i.e., at high frequencies
and long IPG durations). The inset shows how the cell membrane is hyperpolarized by the anodic
phase of the stimulus pulse. Data were generated from the salamander model.
Data were acquired from five Argus II subjects, using two electrodes per subject. After
controlling for effects of adaptation to repeated stimulation (see Section 3.2.3), it became
evident that IPG had a similar effect on perceptual thresholds as it did on RGC electrical
thresholds. Fig. 3.6 (blue curve) shows how perceptual thresholds changed as interphase
gap length was varied between 0 and 2 ms. Regardless of IPG length, there was a∼ 10–15%
reduction in thresholds compared to gapless pulses. These reductions were statistically
significant ( < 0.05, unpaired t-tests) for all IPG durations except 2 ms (possibly due to
anodic block). All four interphase gaps that were tested had a statistically similar effect
on threshold ( = 0.81, one-way ANOVA).
133
Figure 3.6: Effect of interphase gap duration on perceptual thresholds in five Argus II subjects
(blue squares). Data were fit with a decaying exponential. Results are plotted alongside a curve
fit to the salamander experimental data (green trace; see Fig. 3.4). Error bars indicate SEM. **
means < 0.01, and * means < 0.05.
3.4 Conclusions
Taken together, results from the animal experiments, computational modeling, and hu-
man testing demonstrate that interphase gaps can be used to reduce stimulation thresh-
olds by 10–25%. In general, longer IPGs lead to lower thresholds; however, once IPG
duration exceeds a certain length, change in threshold further decreases only marginally,
if at all. This is likely due to the timing of the trailing phase of the stimulus pulse and
where it occurs in time with respect to the rising phase of the action potential. At a
certain point during the rising phase, positive feedback from voltage-gated Na
+
channels
causes the action potential to become self-sustaining. If the trailing phase occurs before
134
this happens, it can cause the action potential to cease firing. If the trailing phase occurs
after the action potential is self-sustaining, it will finish firing.
My modeling results suggest that the effects of interphase gap depend on both stimulus
frequency and IPG duration. Though I investigated only a single pulse width (0.46 ms),
it is likely that pulse width also plays a role in determining how thresholds are affected
by IPG. Indeed, other studies have found the effects of IPG to be more pronounced with
shorter pulses (McKay and Henshall, 2003; Shepherd and Javel, 1999). I also observed
that the combination of high-frequency stimulation and long interphase gaps produced a
blocking effect that caused thresholds to increase. A similar finding has been reported
in cochlear implant users (Carlyon et al., 2005). There is clearly a tradeoff that must be
made between stimulus frequency, pulse width, and IPG duration.
The effects of interphase gap were not as pronounced in the human studies as they
were in the salamander experiments and computational models. There are a few possible
explanations for this. Human testing was performed on a small sample size, and there
was a large amount of noise arising from response desensitization. It is also unknown the
extent to which cortical processing affects perceptual thresholds. Furthermore, there are
more variables to consider when measuring thresholds in humans versus animals: Each
subject may have a different degree of retinal degeneration, and placement of the electrode
array on the retina may vary. Prior studies have found that thresholds and phosphene
brightness both vary with retinal eccentricity (Humayun et al., 1999a, 2003).
In the future, more human testing is needed to determine the optimal interphase gap
length for the Argus II. Safety studies might also be required to confirm that IPGs are
135
electrochemically safe. Nevertheless, it appears that interphase gaps offer a simple solu-
tion to the problem of high thresholds in human patients. They also provide a means to
reduce device power consumption and increase the available dynamic range of phosphene
size and brightness. Finally, they can be implemented without making any changes to
existing hardware.
136
Chapter 4
Using Pulse Width to Control Percept Shape
4.1 Background
Over the last two decades, there have been significant efforts to understand and control
the shapes of percepts elicited by epiretinal stimulation. The first of these efforts were
acute studies in which transscleral probes or MEAs were used to stimulate the retinas of
humans with RP (Humayun et al., 1996, 1999a; Weiland et al., 1999). Patients remained
under local anesthesia so they could describe the nature of evoked phosphenes. Following
stimulation with a single electrode, nearly all patients reported seeing a small spot of
light the size of a pea, pin, or match head at arm’s length.
The fact that patients saw focal percepts, rather than elongated ones, was puzzling.
Given that the stimulating electrodes were placed near the epiretinal surface, one might
expect ganglion cell axon fibers to become activated since they are the closest neural
elements to the electrodes. However, patients did not report streak-shaped phosphenes,
so it was assumed that ganglion cell bodies were the target of stimulation. Either the cell
137
bodies had lower thresholds than their axons, or stimulation was activating bipolar cells
and causing synaptic activation of RGCs.
Greenberg was the first to explore these two possibilities. Through computational
modeling, he found that ganglion cell body thresholds were lower than the thresholds
of their axons, but only by a factor of less than two—not enough to explain the focal
percepts (Greenberg et al., 1999). This led him to investigate whether bipolar cells were
the target of epiretinal stimulation. By performing experiments in isolated frog retina,
he found that pulse duration was the key factor in determining which cell type was
stimulated; pulses longer than 0.5 ms selectively stimulated bipolar cells, while pulses
shorter than 0.5 ms directly stimulated ganglion cells (including their axons) (Greenberg,
1998). Greenberg hypothesized that the focal percepts observed in humans arose from
bipolar cell stimulation. Indeed, the acute human studies all involved pulses that were 1
ms or longer (Humayun et al., 1996, 1999a; Weiland et al., 1999).
Greenberg’s hypothesis was first tested by Rizzo et al. in three human subjects (2003a;
2003b). Rizzo varied pulse duration between 0.25 and 16 ms and asked patients to classify
evoked phosphenes as round or elongated. Surprisingly, patients reported both phosphene
shapes over the entire range of stimulus durations (though round percepts were 2.5x more
common). This was the first published evidence of axonal activation in humans. There
was, however, no apparent dependence of percept shape on pulse width.
The first chronic epiretinal prosthesis, the Argus I, was surgically implanted in 2002
(Humayun et al., 2003). Over the next few years, several patients with RP received
the device (de Balthasar et al., 2008; Mahadevappa et al., 2005; Yanai et al., 2007).
Despite Rizzo’s findings, no subjects reported elongated phosphenes (pulse width was
138
1 ms). It was therefore concluded that deeper retinal neurons must be the target of
epiretinal stimulation (Humayun et al., 2003). However, not everyone was convinced of
this. Investigators continued to perform experiments in isolated animal retina to test
whether different cell classes were excited by different pulse widths.
Several important findings came out of the pulse width animal studies. First, it was
shown that short-duration pulses (≤ 0.2 ms) selectively stimulated ganglion cells (Behrend
et al., 2011; Fried et al., 2006; Jensen et al., 2005b; Sekirnjak et al., 2006). These pulses
could also target bipolar cells, but much larger currents were needed (2.5–20x, depending
on the study). It was also found that short pulse widths provided selectivity for ganglion
cell bodies over axon bundles (Behrend et al., 2011; Jensen et al., 2003, 2005b). This
selectivity ranged from ∼ 2–4x and improved as pulses became shorter. Taken together,
these results suggest that short pulses might provide a means to preferentially stimulate
ganglion cell bodies in prosthesis users, which should produce focal percepts. However,
such stimulation would be possible only over a small range of stimulus amplitudes, as
larger amplitudes would activate RGC axons.
Another possibility for avoiding axons is to selectively stimulate bipolar cells and
cause indirect RGC activation. Several studies have reported bipolar cell stimulation
with pulse widths≥ 1 ms (Behrend et al., 2011; Fried et al., 2006; Jensen et al., 2005b;
Margalit and Thoreson, 2006; Shah et al., 2006). However, ganglion cells and their axons
can also be activated directly by long pulses. One study found that even with a duration
of 50 ms, ganglion cell thresholds were just 1.25–2x higher than bipolar cell thresholds,
and those differences were not statistically significant (Jensen et al., 2005b). Nonetheless,
direct activation causes only a single spike per pulse, whereas indirect activation causes
139
RGCs to fire bursts of spikes. Thus, even if passing axons are activated during bipolar
cell stimulation, the highest concentration of spiking activity would be proximal to the
electrode.
The inconsistencies between the human and animal data prompted further investiga-
tion into phosphene shape. More detailed questioning of Argus I and II subjects revealed
that percepts were in fact elongated in some cases. To measure shape, a tracing method
was developed in which test subjects “drew” the perceived phosphene on a board or
touchscreen. It was found that percepts were not randomly oriented but rather followed
the paths of RGC axon bundles (Fig. 4.1) (Nanduri, 2011; Nanduri et al., 2011).
Figure 4.1: Axonal stimulation in Argus I and II subjects produces elongated phosphenes. A:
Two electrodes were activated in an Argus I subject (left image, red circles). Percept shapes
were predicted to align with the pathways of RGC axon bundles (center image). Percepts drawn
by the subject (right image) matched the predictions, suggesting that axons were indeed being
stimulated. Axis labels indicate degrees of visual angle. B: Several electrodes were activated in
an Argus II subject at different times (red circles). Black lines show the percepts drawn by the
subject, which are clearly elongated. (Adapted from Nanduri, 2011.)
140
Axonal stimulation is a major problem for epiretinal prostheses, as it greatly di-
minishes spatial resolution. Thus, there is a clear need for novel methods or stimulus
paradigms that prevent RGC axons from being activated. The most promising of these
paradigms was recently proposed by Freeman et al. (2010). Rather than using traditional
rectangular pulses, they stimulated isolated rabbit retina with sinusoidal waveforms of
different frequencies. They discovered that 10–25-Hz sinusoids selectively activated bipo-
lar cells without causing direct activation of ganglion cell bodies/axons. Even at the
highest stimulation amplitudes they tested (7–10x greater than bipolar cell thresholds),
RGC axons were still avoided. To explain this behavior, they generated computational
models of ganglion and bipolar cells. The models predicted that L-type Ca
2+
channels,
which underlie glutamate release from bipolar cells (Thoreson, 2007), respond strongly to
10–25-Hz sine waves. Na
+
channels, which underline RGC spiking, become inactivated
at those frequencies.
Our group recently tested Freeman’s findings in a single Argus I subject (Nanduri,
2011). Percepts evoked by 0.45-ms biphasic pulses were compared to those evoked by 20-
Hz sinusoids. (Due to hardware limitations, pseudo-sinusoids were delivered rather than
true sinusoids; see Fig. 4.2, red trace.) Phosphenes elicited by sinusoidal stimulation
were generally much rounder than those elicited by biphasic stimulation, though not on
every electrode. Furthermore, activating two electrodes along the same axon pathway
produced two distinct phosphenes when sinusoids were used but only a single phosphene
with biphasic pulses. Results are summarized in Fig. 4.2.
141
Figure 4.2: Pseudo-sinusoidal stimulation in an Argus I subject produces punctate phosphenes.
Left: Phosphene drawings following single-electrode stimulation. The top and bottom rows repre-
sent two different electrodes. Pseudo-sinusoidal percepts are shown in red, and biphasic percepts
are shown in blue. Pseudo-sinusoidal percepts are clearly more focal. Right: Phosphene draw-
ings following dual-electrode stimulation. Pseudo-sinusoids elicited two distinct phosphenes, while
biphasic pulses elicited only one. (Adapted from Nanduri, 2011.)
4.2 Experimental Design
The animal studies described above provide strong evidence that stimulus design affects
the spatiotemporal patterns of RGC activity. Nevertheless, each study relied on single-
or multielectrode array recordings that were unable to measure the spatial extent of RGC
activation. In fact, all studies involving pulse widths greater than 1 ms utilized single-unit
recordings, so the authors could not determine how many RGCs were activated by each
stimulus. The same is true of Freeman’s sinusoidal stimulation experiments. Further-
more, each study was conducted in wild-type retina. The extent to which degeneration
affects the retina’s response to different pulse widths and stimulus shapes has yet to be
determined.
Our calcium imaging approach is unique in that it provides precise detail about the
location of every activated RGC. It is ideally suited for measuring the spatial patterns of
RGC activation, which may be good predictors of human perception. Given the recent
142
analyses of phosphene shape in humans (Nanduri, 2011; Nanduri et al., 2011), there is
presently a need to relate percept shape to stimulus design (Behrend et al., 2011). We can
now do this in both wild-type and degenerate retina through the use of virally delivered
GECIs.
I conducted a comprehensive study to investigate how stimulus duration affects the
spatial responses of RGCs. My initial intent was to measure RGC activation patterns for
sinusoidal stimuli over a wide range of frequencies. However, preliminary results revealed
that 20-Hz square waves were as effective at producing focal responses as 20-Hz sine
waves. I therefore hypothesized that stimulus duration, rather than waveform shape, was
the factor underlying these responses.
To test this idea, I stimulated isolated rat retina with pulse widths ranging from 0.06
to 100 ms in duration (Table 4.1). By creating spatial threshold maps (see Fig. 2.42) for
each pulse width, I was able to determine the effect of pulse width on response shape. I
was also able to compare threshold differences between ganglion cell bodies and axons.
To determine whether responses were mediated by RGC and/or inner retinal stimulation,
I measured how thresholds changed in the presence of synaptic blockers at every pulse
width.
Initial experiments were performed in Long Evans rats (aged P81–P260) with the
GECI GCaMP5G. Electrode diameter was chosen to match the size of Argus II electrodes
(200-m diameter). Threshold maps were generated for all 12 pulse widths in Table 4.1.
To determine how responses were affected by electrode size, maps were also generated for
75- and 30-m-diameter electrodes using a subset of pulse widths (0.06, 1, and 25 ms).
The effects of retinal degeneration on response shape were studied by mapping thresholds
143
Table 4.1: Stimulation parameters used in the pulse width experiments.
Pulse
Width (ms)
Pulse
Shape
# of
Pulses
Frequency (Hz) Duration (ms)
0.06 20 167 120
0.1 20 167 120
0.2 20 167 120
0.4 20 167 120
1 20 167 120
2 20 167 120
4 25 125 200
8 12 62.5 192
16 6 31.25 192
25 4 20 200
50 2 10 200
100 1 5 200
in three S334ter-line-3 rats (aged∼ P600) with 200-m-diameter electrodes and the same
subset of pulse widths. Data from at least three retinas were used to create each map.
After characterizing the retina’s response to single-electrode stimulation, I demonstrated
successful multielectrode stimulation by activating controlled patterns of ganglion cells
with groups of 75- and 30-m-diameter electrodes.
As described in Section 2.5.3, repetitive stimuli were needed to evoke a burst of spikes
and generate a detectable calcium transient. Each stimulus was therefore a rectangular
pulse train or square wave. Stimulus duration was limited by the image acquisition rate to
200 ms, which placed constraints on the waveform type used for each pulse width. Pulse
durations≤ 2 ms were delivered as 167-Hz rectangular pulse trains. To stay consistent
with prior studies (Behrend et al., 2011; Sekirnjak et al., 2006, 2008, 2009), each pulse
contained a cathodic phase followed by an anodic phase of twice the duration and half
144
the amplitude. Pulses longer than 2 ms required symmetric square waves, as the 200-ms
time window was not long enough to permit asymmetric pulses or interpulse gaps.
A control experiment was performed to verify that response shape depended only on
pulse width—not stimulus waveform design. Indeed, RGC responses were found to be
spatially consistent for a given pulse width, regardless pulse symmetry or interpulse delay.
For example, 500-Hz trains of symmetric 1-ms pulses activated the same cells as 167-Hz
trains of asymmetric 1-ms pulses. Similarly, 10-Hz trains of asymmetric 25-ms pulses
activated the same cells as 20-Hz square waves (i.e., trains of symmetric 25-ms pulses).
4.3 Threshold Maps for 200-m-diameter Electrodes
Fig. 4.3 shows threshold maps for pulse widths ranging from 0.06 to 100 ms. All pulses
≤ 8 ms in duration activated axon bundles, causing elongated responses extending away
from the optic disc. 16-ms pulses also stimulated axons, but to a much lesser extent.
Pulse widths≥ 25 ms produced no detectable axon activity, resulting in focal activation.
Charge density was used as a measure of threshold in all experiments, since its value
accounts for electrode size, pulse width, and waveform shape. Fig. 4.3 shows how thresh-
old charge density became higher as pulse width was increased, indicating that longer
pulses were less efficient at evoking responses. For some pulse widths, there were also
clear threshold differences between ganglion cell bodies (cells above the electrode) and
axons (cells to the right of the electrode). For example, 0.06-, 0.1-, and 8-ms pulses pro-
vided the best selectivity for RGC somata, while 1-, 2-, and 4-ms pulses provided little
to no selectivity.
145
Figure 4.3: Threshold maps for 200-m-diameter electrodes in wild-type retina. The color bar
shows thresholds in terms of charge density on a log
2
scale. The numbers of responding and
non-responding cells at each pulse width are listed in Table 4.2.
Table 4.2: Number of retinas and RGCs in each threshold map (200-m-diameter electrodes,
wild-type retina).
Pulse Width (ms)
0.06 0.1 0.2 0.4 1 2 4 8 16 25 50 100
# of Retinas 4 3 3 3 3 3 3 3 3 4 3 3
# of Respond-
ing RGCs
578 539 550 552 542 471 659 472 233 229 320 344
# of Non-Re-
sponding RGCs
1342 932 837 701 929 1040 852 1035 1663 1971 1419 1395
146
For pulses≥ 25 ms, thresholds directly above the electrode were highly uniform.
Thresholds beyond the electrode perimeter were more variable but tended to decay with
distance (Fig. 4.4). These results suggest that long pulses might permit an amplitude
coding strategy in which percept size is controlled by stimulus strength. Amplitude coding
has already been demonstrated in an Argus I subject with 0.45-ms pulses; however, higher
amplitudes also led to stimulation of passing axons (Nanduri et al., 2012). I did not
observe axonal activation for pulses≥ 25 ms, regardless of amplitude.
Figure 4.4: Threshold vs. displacement from electrode center for 100-ms pulses. Ganglion cell
thresholds ( = 344) were taken from the 100-ms response map in Fig. 4.3. Electrode diameter
is 200 m.
4.3.1 Effects of Synaptic Blockers
As shown in Fig. 4.3, response shapes transitioned from being elongated to focal at a
pulse width of∼ 16 ms. I speculated that this transition represented a shift from direct
to indirect activation of RGCs, such that shorter pulses stimulated RGCs directly, while
147
longer pulses targeted deeper retinal neurons. To test this idea, I applied synaptic blockers
and measured their effect on thresholds at every pulse width.
Prior blocking experiments in salamander retina utilized kynurenic acid, a broad-
spectrum glutamate receptor antagonist (Behrend et al., 2011; Massey and Miller, 1988).
Though partially effective, kynurenic acid does not block synaptic transmission com-
pletely and is therefore not an ideal agent. Attempts were made to use stronger blocking
agents in salamander, however their autofluorescence masked the Oregon Green signals.
Because GCaMP5G signals were much larger than those of Oregon Green (see Fig. 2.19),
I was able to use stronger blockers without noticeable interference from autofluorescence.
I used a blocker cocktail consisting of five ingredients (Table 4.3). CNQX,d-APV,
and l-APB were used to block excitatory input to RGCs, while strychnine and picrotoxin
were applied to block inhibitory input (Margolis et al., 2008; Sekirnjak et al., 2006, 2011).
All blockers were dissolved in the superfusate and were washed in for at least 10 minutes
prior to recording. CNQX has poor water solubility and was therefore dissolved in DMSO
before being added to the superfusate (final DMSO concentration was 0.1%). I verified
that the cocktail abolished synaptic transmission by confirming absence of light responses
(Fig. 4.5).
To determine how thresholds changed in the presence of blockers, I performed iden-
tical stimulation runs before and after adding blockers to the superfusate. Stimuli were
delivered by a 200-m-diameter electrode over the entire range of pulse widths (0.06–
100 ms). The contribution from inner retinal stimulation was determined by comparing
thresholds in the presence and absence of blockers at every pulse width (Fig. 4.6). As
shown in the figure, thresholds for 0.06-ms pulses were unchanged after blocking synaptic
148
Table 4.3: Synaptic blocker cocktail used to pharmacologically isolate RGCs.
Blocking Agent Concentration (M) Effect
CNQX 70 AMPA/kainate receptor antagonist
d-APV 50 NMDA receptor antagonist
l-APB 50 mGluR6 receptor agonist
strychnine 10 glycine receptor antagonist
picrotoxin 50 GABA
A
receptor antagonist
Figure 4.5: Light responses were recorded extracellularly from RGCs in dark-adapted retina.
Strong light responses were observed (left image) prior to adding synaptic blockers (Table 4.3).
These responses were abolished six minutes after washing in the blockers (right image) and were
restored roughly 10 minutes after wash-out (data not shown). Arrows show the times at which
brief light flashes were presented to the retina.
transmission ( = 0.48, unpaired t-test), indicating that this pulse width stimulated only
RGCs. Thresholds for pulse durations between 0.1 and 4 ms rose by 20–30% once blockers
were applied ( < 0.001), implying contributions from inner retinal stimulation. These
contributions were highest for 8-ms pulses, in which thresholds nearly doubled in the
presence of blockers ( < 0.001). Pulses longer than 8 ms evoked no calcium transients
once synaptic transmission was blocked, indicating that responses to these long-duration
149
pulses were predominantly synaptic in origin. Results are consistent with the notion that
long pulses produce focal responses due to stimulation of the inner retina.
It is likely that responses to long pulses were mediated by bipolar cells, rather than
photoreceptors. Photoreceptors were bleached by the fluorescence excitation light, caus-
ing them to remain hyperpolarized and become harder to stimulate (Behrend, 2009;
Greenberg, 1998). Bright light prevents the cells from releasing glutamate, effectively
blocking their output.
Figure 4.6: Effect of synaptic blockers on ganglion cell thresholds over a range of pulse widths.
Pulses longer than 8 ms did not evoke responses in the presence of blockers. Thresholds for all
pulse widths except 0.06 ms rose significantly after blockers were applied (*** means < 0.001).
Error bars indicate SEM. Data were fit with a weighted inverse sigmoid, =− ln(

−
− 1) +.
4.3.2 Selectivity Ratios
Some response maps in Fig. 4.3 show clear differences between soma and axon thresholds.
These differences were quantified through a measure called selectivity ratio (Eq. 4.1).
150
This ratio provides an estimate for the amplitude range over which an electrode can
operate without stimulating axons.
selectivity ratio =
mean axon threshold
mean soma threshold
(4.1)
Because calcium measurements were not recorded directly from axon bundles, thresh-
olds of antidromically stimulated somata were taken to represent axon thresholds
(Behrend et al., 2011). Any responding cells≥ 100 m to the right of the electrode
perimeter were assumed to be antidromically stimulated. Soma thresholds were defined
as thresholds of cells directly above the electrode. These thresholds most likely represent
those of the axon initial segment—a region of the proximal axon that contains the high-
est density of voltage-gated Na
+
channels. Because axon initial segments are the most
electrically excitable parts of RGCs (Fried et al., 2009), one would expect their thresholds
to be lower than those of passing axons.
Fig. 4.7 summarizes the selectivity ratios for every pulse duration that was studied.
Compared to a repeat stimulation control (see Section 2.7.4.1), axon thresholds were
significantly higher than soma thresholds, regardless of pulse width ( < 0.001). The
ratios never exceeded 1.8 and were often smaller than 1.2. These values are slightly
lower than those reported by another study in rabbit retina (1.2–4; Jensen et al., 2005b).
Any inconsistencies are likely attributed to the way in which thresholds were measured
(electrically vs. optically) and use of a different animal model (rabbit vs. rat).
Consistent with prior reports (Behrend et al., 2011; Jensen et al., 2005b), selectivity
was found to be best for very short pulses (0.06 and 0.1 ms) and gradually diminished as
151
Figure 4.7: Selectivity ratios were plotted for every pulse duration. Axon thresholds were
significantly higher than soma thresholds for all pulse widths ( < 0.001). The inset shows
minimum selectivity ratios for pulse widths that caused little to no axonal activation. Arrows
indicate that values are lower bounds. Error bars indicate SEM.
pulse width was increased to 4 ms. As predicted by a modeling study, pulses shorter than
∼ 0.4 ms are selective because they induce a greater depolarization in the bend of the axon
relative to axons of passage (Schiefer and Grill, 2006). Fig. 4.7 also indicates that 8-ms
pulses provided relatively good selectivity for RGC somata. Results from the synaptic
blocker experiments (see above) showed that contribution from inner retinal stimulation
was at least 3x greater for 8-ms pulses than it was for shorter pulses. This suggests that
8-ms pulses were selective because they activated inner retinal cells at thresholds lower
than those of RGC axons.
Selectivity ratio could not be measured for pulse durations between 16 and 100 ms,
since these pulses produced little to no measurable axonal activation. Instead, minimum
selectivity ratios were calculated by taking the highest stimulus amplitude delivered at
152
each pulse width and dividing by the mean threshold of cells directly above the electrode
(Fig. 4.7). Minimum ratios represent a lower bound over which long pulses might evoke
focal responses without activating RGC axons.
1
Their values ranged from 18–45, depend-
ing on pulse width and were roughly 2–6 times higher than the minimum ratios reported
for 10- and 25-Hz sinusoidal stimulation (Freeman et al., 2010).
Although pulse widths≥ 25 ms appeared to avoid axons, it is possible that these
pulses caused antidromic spiking that was too sparse to detect with GCaMP5G. This
can be explained by the following example: The highest stimulus amplitude delivered for
100-ms pulses was 8 A. Thus, the total charge delivered within the first millisecond of
each pulse was 8 nC. This is nearly 7 times as high as the threshold charge of RGC axons
for 1-ms pulses (1.2 nC). We would therefore expect 100-ms pulses to excite RGC axons
within one millisecond. However, if only a single spike were elicited in the axons, any
changes in GCaMP5G fluorescence would be too small to detect.
4.3.3 Strength-Duration Curves
A strength-duration curve is a plot of threshold amplitude versus pulse width. These
plots are widely used in electrophysiological and psychophysical studies to characterize
electrically excitable cells or tissue. Two important parameters can be extracted from
a strength-duration curve: rheobase and chronaxie. Rheobase represents the threshold
amplitude for an infinitely long pulse width and corresponds to the asymptote of the
curve. Chronaxie is the pulse duration along the curve corresponding to an amplitude of
twice the rheobase.
1
16-ms pulses did activate axons, but only rarely. Of the 901 somata located more than 100 m to
the right of the electrode perimeter, only 42 responded at this pulse width.
153
Determining a cell’s chronaxie is useful for two reasons: First, stimulus energy is
minimized at a pulse duration equal to the chronaxie (Geddes, 2004). Second, there is
a strong correlation between a cell’s chronaxie and its membrane time constant (Ranck,
1975). For example, retinal ganglion cells have fast integration time constants and short
chronaxies and are therefore most easily stimulated by short pulses. Bipolar cells have
relatively slow time constants and long chronaxies and therefore respond best to long
pulses (Freeman et al., 2011).
Strength-duration curves were plotted for all pulse durations in Table 4.1 (0.06–100
ms). Two types of curves were generated: one showing threshold current (Fig. 4.8) and
another showing threshold charge (Fig. 4.9). The amplitude plotted for each pulse width
represents the average threshold of cells directly above the electrode. Antidromically
stimulated RGCs were not considered, as chronaxie values for axons and somata have
been shown to differ (Jensen et al., 2005b; Tehovnik et al., 2006).
Strength-duration curves are often fit with a single decaying exponential function
(Chan et al., 2011; Lapicque, 1907). However, the data in Fig. 4.8 showed a clear drop-
off in threshold current at 16 ms (inset), representing a shift from ganglion to bipolar cell
stimulation at this pulse width. For this reason, data were fit with two exponentials (Eq.
4.2)—the first to pulse widths between 0.06 and 2 ms and the second to pulse widths
between 16 and 100 ms. Fits were weighted with the reciprocal of the variance and were
optimized to minimize sum of squared error.
154
=

1−
− /
,
where is the rheobase
and ln 2 is the chronaxie
(4.2)
Figure 4.8: Strength-duration curve plotted in terms of current density. The inset shows a
higher-magnification view of pulse widths between 1 and 100 ms in order to portray the drop-off
in threshold current at 16 ms. The blue and red curves are decaying exponentials that were fit to
the data. Error bars indicate SD.
Rheobase and chronaxie values were extracted from the two exponential fits in Fig.
4.8. The blue curve represents direct RGC stimulation and indicates a chronaxie of 0.6
ms. This is slightly higher than ganglion cell chronaxies reported in the literature (0.1–0.5
ms; Ahuja et al., 2008a; Chan et al., 2011; Jensen et al., 2005b; Sekirnjak et al., 2006). The
red curve represents inner retinal stimulation and indicates a chronaxie of 21 ms. This is
155
also slightly higher than previously reported values of bipolar cell chronaxies (14–18 ms;
Jensen et al., 2005b). These inconsistencies might arise from differences in the method
used to measure thresholds: Calcium imaging measures RGC activity indirectly, whereas
the other studies obtained direct threshold measurements via electrical recordings.
Fig. 4.9 shows the strength-duration curve plotted in terms of threshold charge den-
sity. Again, there is a clear transition in the curve at 16 ms, representing the shift to
bipolar cell stimulation. The plot also shows how longer pulses were less efficient at evok-
ing responses than shorter ones. For example, 25-ms pulses (the shortest pulse width
to achieve focal stimulation) required more than 10 times the amount of charge as 0.06-
and 0.1-ms pulses. Despite this fact, threshold charge densities for all pulse durations
were well below the electrochemical safety limit of platinum (0.35 mC/cm
2
; Brummer
and Turner, 1977; Rose and Robblee, 1990).
4.3.4 Comparison of 20-Hz Sinusoidal and Pulsatile Stimulation
A recent study by Freeman et al. showed that 10–25-Hz sine-wave stimulation is selec-
tive for bipolar cells and avoids passing axons (2010). The authors, however, could not
measure the spatial pattern of activation because they relied on single-unit recordings.
To investigate this, I imaged RGC responses to 20-Hz sinusoidal stimulation. As shown
in Fig. 4.10, the response shape was largely focal and looked qualitatively similar to
that of 20-Hz square waves (i.e., 25-ms pulses). Of the 1418 cells that were imaged, only
five showed antidromic responses. Based on the maximum stimulus amplitude delivered,
minimum selectivity ratio (see p. 152) for 20-Hz sine waves was 16.7.
156
Figure 4.9: Strength-duration curve plotted in terms of charge density. Error bars indicate SD.
The color bars in Fig. 4.10 indicate that thresholds for 20-Hz sine waves were lower
than those of 20-Hz square waves. These differences were quantified by imaging the same
set of RGCs in two retinas during stimulation with both waveform shapes. Overall, square
wave thresholds were found to be 22.8± 35.5% higher than sine wave thresholds ( = 121,
< 0.001; Fig. 4.11). Though the standard deviation is relatively large, these results
suggest that 20-Hz sinusoids are generally more effective at stimulating bipolar cells than
square waves of the same frequency. Others have also reported lower thresholds for sine
versus square waves, though at much higher stimulation frequencies (Suzuki et al., 2004).
157
Figure 4.10: Comparison of threshold maps for 20-Hz sine and square waves. Electrode diameter
is 200 m. The sine wave threshold map (top) contains 1418 cells from three retinas (a total of
239 cells responded). The square wave map (bottom) is the same one shown in Fig. 4.3 for 25-ms
pulses, except that the color scale has been changed.
Figure 4.11: Comparison of individual RGC thresholds for 20-Hz sine and square waves. Each
data point represents a different cell ( = 121). Data were combined from experiments in two
retinas. Square wave thresholds were 22.8± 35.5% higher than sine wave thresholds ( < 0.001).
158
4.4 Threshold Maps for 75- and 30-m-diameter Electrodes
Threshold maps were generated for 75- and 30-m-diameter electrodes to examine the
effect of electrode size on response shape. Three pulse durations were tested: 0.06, 1, and
25 ms. These durations were chosen because they exhibited markedly different behaviors
in the 200-m electrode experiments: 0.06-ms pulses had some selectivity for RGC somata
over axons, 1-ms pulses had poor selectivity, and 25-ms pulses produced focal responses.
Fig. 4.12 shows response maps for the three electrode sizes (200 m, 75 m, 30 m)
and pulse widths. Threshold charge density gradually increased as electrodes became
smaller. The general shape of responses was the same regardless of electrode size; however,
responses to 25-ms pulses were more focal for 75- and 30-m-diameter electrodes than for
200-m electrodes. Response maps also indicate that axon bundle thresholds were more
variable for smaller electrodes.
Consistent with a prior calcium imaging study in salamander retina (Behrend et al.,
2011), response size did not scale with electrode diameter for electrodes smaller than 75
m. 30-m electrodes activated fewer cells than 75-m electrodes (Table 4.4), but the
activated areas were similar in size for both electrode diameters. This suggests that there
may be a lower limit to the size of percepts that can be elicited by epiretinal stimulation.
Selectivity ratios were determined for 75- and 30-m-diameter electrodes according to
Eq. 4.1. Axon thresholds were measured from cells≥ 100m to the right of the electrode
perimeter, and soma thresholds were measured from cells within 50 m of the electrode
center. Fig. 4.13 compares the selectivity ratios for the three electrode sizes. Selectivity
was best for 75-m electrodes; with pulse width of 0.06 ms, axon thresholds were roughly
159
Figure 4.12: Threshold maps for three electrode sizes and pulse widths in wild-type retina.
The 200-m-diameter maps were taken from from Fig. 4.3. The numbers of responding and
non-responding cells for 75- and 30-m-diameter electrodes are listed in Table 4.4. The bottom
row shows averaged background-subtracted images for 30-m electrodes.
Table 4.4: Number of retinas and RGCs in each threshold map (75- and 30-m-diameter elec-
trodes, wild-type retina).
Pulse Width (ms)
0.06 1 25 0.06 1 25
75-m-diameter 30-m-diameter
# of Retinas 3 3 3 3 3 3
# of Responding RGCs 490 477 111 239 295 81
# of Non-Responding RGCs 1249 1035 1401 1445 1074 1603
160
2.3x as high as thresholds of local somata. Taken together, these results suggest that 75
m might be an optimal electrode diameter for the epiretinal prosthesis. This diameter
provides the best selectivity for RGC somata, and response size does not decrease with
smaller electrodes.
Figure 4.13: Comparison of selectivity ratios for three electrode sizes. Error bars indicate SEM.
*** means < 0.001, and * means < 0.05.
4.5 Reverberating Responses
Stimulating with amplitudes well above threshold caused oscillatory retinal activity that
lasted several seconds. As shown in Fig. 4.14, responses to 25-ms pulsatile stimulation
were initially focal and gradually spread radially over time. These reverberations were
observed over a wide range of pulse widths, provided that the stimulation amplitude was
sufficiently high.
161
Reverberating activity has been reported previously in the retina, but only over a
period of several hundred milliseconds (Ahuja et al., 2008a; Crapper and Noell, 1963;
Fried et al., 2006; Shimazu et al., 1999). These studies measured spiking activity from
a single recording electrode, which likely explains why they did not observe distal RGC
responses over a longer time scale. It has been suggested that reverberations arise from
feed-forward and -backward communication between bipolar, amacrine, and ganglion cells
(Fried et al., 2006). Interestingly, several epiretinal prosthesis patients have described
percepts that begin as a punctate spot and are followed by shooting lights or fireworks
that move outward (personal communication, Ashish Ahuja). These descriptions are
consistent with the reverberating patterns of calcium responses illustrated in Fig. 4.14.
Figure 4.14: Reverberating responses to high-amplitude stimulation. A 200-m-diameter elec-
trode (black circle) delivered a burst of 25-ms pulses to GCaMP5G-labeled retina. Stimulation
amplitude (0.32 mC/cm
2
) was well above threshold. Background-subtracted images show how
RGC responses were initially centered around the electrode and gradually spread over time. Im-
ages were acquired from four fields of view and stitched together.
4.6 Effects of Retinal Degeneration
Stimulation thresholds were mapped in RD retinas to determine how degeneration affects
response shapes and thresholds. Maps were generated for 0.06-, 1-, and 25-ms pulses and
an electrode diameter of 200 m. Three S334ter-line-3 rats (aged P591, P611, and P619)
162
were used in these experiments. S334ter is a mutation that causes truncation of the
rhodopsin protein. Retinas expressing this mutation degenerate over time and are a good
model of autosomal dominant RP (Steinberg et al., 1996).
Heterozygous S334ter-line-3 rats were bred in-house by crossing homozygous breeders
with Copenhagen rats. Heterozygous animals undergo a slower degeneration than ho-
mozygous strains and are therefore a better model of human RP (Sekirnjak et al., 2009).
RD rats injected with AAV2-CAG-GCaMP5G showed widespread ganglion cell trans-
duction. Though their retinas were not examined histologically, others have reported
that AAV2 transduction profiles in S334ter-line-3 retina are similar to those of wild-type
rats (Kolstad et al., 2010). Ganglion cells in RD retinas showed no signs of temperature-
induced cytomorbidity that were typical of inbred Copenhagen strains (see Section 2.5.5).
Absence of light flash responses was confirmed prior to imaging.
Response shapes in RD retina were similar to those of wild-type rats. 0.06- and 1-ms
pulses activated passing axons (Fig. 4.15A), while 25-ms pulses produced focal responses
(Fig. 4.15B). The fact that RD retinas responded to 25-ms pulses is further evidence
that this pulse width stimulates bipolar cells rather than photoreceptors (see p. 150).
Consistent with results in wild-type retina, RD threshold maps (Fig. 4.16) indicated
that 0.06-ms stimulation preferentially targeted RGC somata. Selectivity ratio (Eq. 4.1)
for this pulse width was 1.63± 0.04 SEM (compared to 1.67± 0.04 in WT). For 1-ms
pulses, the ratio fell to 1.06± 0.03 SEM, slightly lower than that of WT (1.14± 0.03).
Table 4.6 compares soma and axon thresholds between WT and RD retinas for every
pulse width. For 0.06-ms pulses, thresholds between WT and RD were similar, indicating
that direct ganglion cell thresholds do not change during degeneration. For 1-ms pulses,
163
Figure 4.15: GCaMP5G fluorescence responses in RD retina. A: Averaged background-
subtracted responses to 1-ms pulsatile stimulation in three retinas. B: Merged GCaMP5G/Alexa
594 image from a single retina showing RGC responses to 25-ms pulses. A high stimulation am-
plitude (1.3 mC/cm
2
) was delivered in B to evoke large calcium transients. Electrode diameter
is 200 m.
Figure 4.16: Comparison of RD and WT threshold maps for 200-m-diameter electrodes. The
WT maps were taken from Fig. 4.3. The numbers of responding and non-responding cells for RD
maps are listed in Table 4.5.
164
Table 4.5: Number of retinas and RGCs in each threshold map (200-m-diameter electrodes,
RD retina).
Pulse Width (ms)
0.06 1 25
# of Retinas 3 3 3
# of Responding RGCs 792 761 311
# of Non-Responding RGCs 1690 1721 2542
RD thresholds for somata and axons were elevated by∼ 19% and 11%, respectively. The
most striking difference in thresholds occurred with a pulse width of 25 ms; thresholds in
RD retina were roughly three times as high as in WT (but were still below the 1 mC/cm
2
safety limit of iridium oxide; Beebe and Rose, 1988). Taken together, these results suggest
that bipolar cell thresholds increase during degeneration, while RGC thresholds remain
unchanged. This may be a consequence of the extensive remodeling that affects bipolar
cell morphology but leaves ganglion cells relatively well intact (Gargini et al., 2006; Jones
et al., 2003; Mazzoni et al., 2008; Sekirnjak et al., 2009). Another possibility is that lack
of photoreceptor input to bipolar cells raises their excitation thresholds (Jensen, 2012).
Table 4.6: Comparison of thresholds in WT and RD retinas for three pulse widths.
0.06 ms 1 ms 25 ms
Somata Axons Somata Axons Somata Axons
% Change 5.57 3.18 19.47 11.00 204.43 —
-value 0.10 0.007 < 0.001 < 0.001 < 0.001 —
Percentages indicate how much higher RD thresholds were than WT thresholds. -values < 0.05
imply significant differences between WT and RD thresholds (unpaired t-tests).
165
Most electrophysiology studies have reported higher stimulation thresholds in RD
retina versus wild-type controls. Thresholds of rd1 mouse retinas were 1.2–7.4 times
higher than those of WT retina (Chen et al., 2006; Jensen and Rizzo, 2008, 2009; O’Hearn
et al., 2006; Suzuki et al., 2004). In P500–P700 S334ter-line-3 rats (the same age used
in this study), thresholds were elevated by up to 4x versus normal controls (Chan et al.,
2011). Each of these studies used relatively long pulse widths and/or large electrodes that
were far from the epiretinal surface. Long-latency spikes were often recorded, indicative
of inner retinal stimulation. It is therefore likely that the reported thresholds represent
a combination of ganglion and bipolar cell responses.
To date, only one other group has measured direct RGC responses to epiretinal stim-
ulation in RD retina (P23H rats; Sekirnjak et al., 2009). By using short pulse widths
(0.05–0.1 ms) and small MEA electrodes (7–16 m), the authors avoided stimulating
bipolar cells. They found that direct RGC thresholds were unaffected by degeneration.
Collectively, these studies are in agreement with my finding that bipolar cell thresholds
become elevated during degeneration, while direct ganglion cell thresholds remain the
same.
4.7 Multielectrode Pattern Stimulation
During multielectrode stimulation in human prosthesis patients, interactions between
neighboring electrodes have a significant influence on phosphene shape and brightness
(Horsager et al., 2010, 2011; Nanduri et al., 2008). These interactions have been attributed
to interfering electric fields as well as neuronal interactions (Horsager et al., 2010). Very
166
few animal studies have examined the responses of retinal ganglion cells to stimulation
with more than one electrode simultaneously (Jensen et al., 2009; Sekirnjak et al., 2006;
Stett et al., 2000). In each of these studies, the authors measured responses of only
one or a few RGCs at a time. Furthermore, there were conflicting conclusions about
whether current from adjacent electrodes interfered with the temporal spiking patterns
of activated RGCs.
Calcium imaging provides the unique ability to visualize spatial patterns of RGC
activation in real time. Doing so requires a highly sensitive calcium probe, such as
GCaMP5G. By imaging RGCs transduced with this sensor, I was able to determine
whether precise patterns of cells could be activated by multielectrode stimulation. All
multielectrode experiments were conducted in normal retina.
4.7.1 Interleaved Stimulation
When stimulating with multiple electrodes, pulses can be interleaved in time to prevent
electric field interactions. This strategy is common in cochlear implants (Wilson et al.,
1993) and has also been tested in Argus I subjects (Horsager et al., 2010, 2011). Results
from my single-electrode experiments indicated that very short (≤ 0.1 ms) and very long
(≥ 25 ms) pulses were the most selective for RGC somata. Compared to short pulses,
long pulses are more difficult to interleave in time but are also better at avoiding axons.
To determine the degree of interactions for short-duration pulses, I activated groups of
two and four electrodes while imaging RGC calcium transients. Pulse duration was fixed
at 0.06 ms, and electrodes were 75 m in diameter on a 150 m pitch. Burst stimulation
(120 pulses, 200 Hz) was used in order to elevate GCaMP5G fluorescence to a detectable
167
level. Stimuli from each electrode were delivered pseudosynchronously (phase-shifted by
1 ms) in order to prevent electric field interactions. Amplitude (6.8C/cm
2
) was chosen
to be slightly above threshold.
Figure 4.17A shows the baseline fluorescence of GCaMP5G-labeled RGCs. Each of the
four transparent stimulating electrodes is outlined in blue. First, the top two electrodes
were activated pseudosynchronously, and background-subtracted GCaMP5G fluorescence
responses were calculated (Fig. 4.17B, red). This was then repeated for the bottom two
electrodes (Fig. 4.17B, green). As can be seen in the merged image (Fig. 4.17B), the top
two electrodes activated only nearby cells, while responses to the bottom two electrodes
were less focal (presumably owing to the axon bundle that was stimulated by the lower-
right electrode; Fig. 4.17A, arrowhead). Three somata near the upper-left electrode
(Fig. 4.17B, yellow) responded to stimulation with the top electrodes as well as with the
bottom electrodes.
Figure 4.17: Multielectrode stimulation of RGCs with temporally interleaved stimulation. A:
Baseline fluorescence of GCaMP5G-labeled cells that were stimulated by four 75-m-diameter
electrodes (outlined in blue). The arrowhead points to an axon bundle that was stimulated by
the bottom-right electrode. B: Merged, background-subtracted image of RGCs that responded to
stimulation with the top two electrodes (red) and bottom two electrodes (green). C : Background-
subtracted image of RGCs that responded to stimulation with all four electrodes. The arrowheads
in B and C point to cells that responded differently in both cases.
168
To explore the effects of electrode interactions, all four electrodes were activated pseu-
dosynchronously. Background-subtracted GCaMP5G fluorescence responses are shown in
Fig. 4.17C. If there were no interactions, the pattern of activated RGCs would match that
of Fig. 4.17B. Though this was largely the case, the two patterns are not identical: The
arrowhead in Fig. 4.17B shows a cell that responded to stimulation with the bottom two
electrodes but not to stimulation with all four. Similarly, the arrowheads in Fig. 4.17C
show several cells that responded only to stimulation with all four electrodes. Because
stimuli were interleaved in time, these differences suggest that electrode interactions are
present at the neural level. Neural interactions have been reported in epiretinal prosthesis
patients and are suggested to arise from lateral connections mediated by amacrine cells
(Horsager et al., 2010).
4.7.2 Line Patterns
Lines are among the simplest patterns that can be produced by multielectrode stimula-
tion. By activating groups of four electrodes along the same axis, I was able to stimulate
linear arrangements of ganglion cells. These experiments utilized 30-m-diameter elec-
trodes on a 75-m pitch and 25-ms pulses. Stimulus amplitudes were the same on each
electrode.
Fig. 4.18 shows background-subtracted fluorescence images recorded during line stim-
ulation. All images were rotated such that the optic disc lies to the left and axon bundles
traverse horizontally. The top row of the figure illustrates responses to lines oriented
almost perpendicular to the axons. Four stimulus amplitudes were tested. The mini-
mum amplitude, I
min
(707 C/cm
2
on each electrode), was chosen to be slightly above
169
threshold. Responses generally became stronger as amplitude was increased. No axonal
activation was observed, even at an amplitude of 7.5× I
min
.
Line stimulation was also effective when electrodes were oriented parallel to axons.
The bottom row of Fig. 4.18 shows four electrodes that were situated directly along the
path of an axon bundle. Stimulating at 7.5× I
min
activated a linear pattern of RGCs
without causing an elongated streak response.
Figure 4.18: Line pattern stimulation of RGCs with 30-m-diameter electrodes and 25-ms
pulses (a burst of 12 pulses). The top row shows background-subtracted responses to a line
oriented nearly perpendicular to the axons. Different stimulation amplitudes were tested (I
min
=
707 C/cm
2
on each electrode). The bottom row illustrates responses to a line oriented parallel
to the axons. The Alexa 594 image on the left shows that electrodes were situated directly below
an axon bundle.
4.7.2.1 Effects of Pulse Width
The ability to activate linear patterns of RGCs was compared for short- and long-duration
pulses. Burst stimuli were delivered by groups of four 30-m-diameter electrodes oriented
170
perpendicular to the axon bundles (Fig. 4.19). Each electrode was activated individually,
and then all four were activated together. Amplitudes were the same on all electrodes,
and pulses were not interleaved in time.
Figure 4.19: GCaMP5G (left) and Alexa 594 (right) images showing the four 30-m-diameter
electrodes used in the line pattern stimulation experiments in Fig. 4.20. Images were rotated such
that the optic disc lies to the left.
Fig. 4.20 shows RGC activation patterns for 0.1- and 25-ms pulses over a range of
amplitudes. I
min
(10C/cm
2
for 0.1-ms pulses; 707 C/cm
2
for 25-ms pulses) represents
an amplitude slightly above threshold. As shown in the figure ( top two rows), responses
to single-electrode stimulation with 25-ms pulses were largely focal, regardless of ampli-
tude. Activating all four electrodes simultaneously produced a linear pattern of activated
RGCs; however, the pattern was not a simple summation of the single-channel responses,
indicating that electrode interactions were present. Nonetheless, these interactions were
not significant enough to distract from the overall shape of the line.
When stimulating with 0.1-ms pulses near threshold (I
min
), three of the four electrodes
activated single ganglion cells (Fig. 4.20, middle row). Interestingly, these cells were
located up to 40 m away from the electrodes, which is the distance between the soma
and axon initial segment (Fried et al., 2009). Single-cell stimulation has been reported
171
Figure 4.20: Line pattern stimulation with short and long pulses. Burst stimuli were delivered
through groups of four 30-m-diameter electrodes along the same axis (see Fig. 4.19). Responses
to 25-ms pulses (20 Hz, 600-ms duration) were compared to those of 0.1-ms pulses (167 Hz, 720-ms
duration). Each of the four electrodes was activated individually, and then all four were activated
together. Different stimulation amplitudes were tested (I
min
= 10 C/cm
2
for 0.1-ms pulses
and 707 C/cm
2
for 25-ms pulses). Images show background-subtracted GCaMP5G fluorescence
responses.
172
previously in primate retina with 9–15-m-diameter electrodes and similar pulse durations
(Sekirnjak et al., 2008).
Delivering 0.1-ms pulses from all four electrodes at I
min
evoked responses in approx-
imately 10 ganglion cells, forming a quasi-line pattern (Fig. 4.20, middle row). Again,
this pattern was not a simple summation of the single-channel responses, implicating the
effects of electrode interactions. Increasing the amplitude above I
min
by only 40% caused
significant axonal stimulation, especially when the four electrodes were activated simul-
taneously (Fig. 4.20, bottom two rows). These antidromic responses extended far outside
the camera’s field of view. Results from this experiment clearly demonstrate the disad-
vantages of using short-duration pulses for both single- and multi-electrode stimulation.
4.7.3 Letters
The ability to form more complex patterns on the retina was assessed by activating groups
of electrodes in the shape of letters. Fig. 4.21 shows ganglion cell responses to the letter
V. Stimulation was performed with seven 30-m-diameter electrodes and 25-ms pulses.
Electrodes were arranged hexagonally and spaced 75-m apart (center-to-center). The
height of the V on the retina was 220 m, corresponding to 0.78
∘ of human visual angle
(Drasdo and Fowler, 1974). This is equivalent to viewing a 5-mm-tall letter from typical
reading distance (40 cm).
The letters L, F, and T were generated by activating sets of transparent 75-m-
diameter electrodes on a 150-m pitch (25-ms pulse width). Each letter spanned 375
m on the retina, representing 1.3
∘ of human visual field (Drasdo and Fowler, 1974) and
corresponding to 9-mm-tall letters viewed from typical reading distance. Their shapes
173
Figure 4.21: Multielectrode stimulation in the shape of a V. Left: Alexa 594 image showing the
seven electrodes that were activated to form the letter. Right: Background-subtracted GCaMP5G
fluorescence responses to stimulation. The V in the lower-right corner shows the actual size of
the letter if viewed from typical reading distance.
conform to the definition of Snellen letters, which require a critical detail size (stroke and
gap width) that subtends 1/5 of the overall height. In this case, the critical detail size
is 75 m, corresponding to 0.26
∘ visual field and a Snellen acuity of 20/312. This acuity
is approximately four times as high as the best visual acuity reported among Argus II
subjects (Humayun et al., 2012). To spell the word LIFT (Fig. 4.22), the letters L, F,
and T were combined with a line pattern from Fig. 4.18.
Figure 4.22: Multielectrode stimulation to spell the word LIFT. Four background-subtracted
images (one per letter) were stitched together. The L, F, and T were formed by transparent
75-m-diameter electrodes, and the I was formed by Pt/Ir 30-m-diameter electrodes. The scale
bar in the left image is 100 m. The right image shows the actual size of the word at typical
reading distance.
174
4.8 Conclusions
Neural prostheses have classically employed short-duration pulses, and years of in vitro
retina studies have suggested that short pulses are the best option for avoiding axons. By
mapping RGC responses to different pulse widths and electrode sizes, I have shown that
the opposite may in fact be true. I demonstrated for the first time that long-duration
stimuli can produce focal activation patterns in degenerate retina over a wide range of
amplitudes.
Consistent with prior reports, I found that short pulse widths (≤ 0.1 ms) target
ganglion cells directly (Behrend et al., 2011; Fried et al., 2006; Jensen et al., 2005b;
Sekirnjak et al., 2006). When the amplitude is tuned properly, short pulses can activate
a single ganglion cell soma. The problem is that once stimulation current is increased
by a factor of two, hundreds or thousands of RGCs are activated antidromically. This
means that amplitude coding can operate only over a small range of amplitudes without
stimulating axon bundles. Furthermore, thresholds in human subjects gradually rise
during repetitive stimulation (see Fig. 3.3), making it difficult to adaptively select an
amplitude that will excite RGC somata and not their axons.
In contrast to short pulses, long pulse widths (≥ 25 ms) target bipolar cells and
cause focal RGC activation. Thresholds are highly uniform above the electrode and
gradually decay with distance beyond its perimeter, thus permitting an amplitude coding
strategy for controlling percept size. Importantly, multielectrode stimulation with long
pulses can activate precise patterns of ganglion cells at high spatial resolution. Electrode
interactions appear to have a minimal effect on the overall response shape. It is therefore
175
not necessary to interleave pulses on adjacent electrodes, which would be difficult with
long pulse durations.
At first glance, there appear to be several disadvantages associated with long pulses.
For one, they require at least 10 times more charge as short pulses, especially in de-
generate retina. Nevertheless, thresholds to 25-ms pulses in my experiments were below
the electrochemical safety limit of platinum for electrode diameters of 75 and 200 m.
This may not be the case in human prostheses, where the array is often several hundred
microns from the retina; however, new methods for achieving close coupling between the
MEA and retina are under investigation (Tunc et al., 2008). Novel electrode materials
and surface coatings with higher charge injection limits are also being developed (Green
et al., 2013; Petrossians et al., 2011; Venkatraman et al., 2011). Furthermore, it is possible
that long pulses will allow a higher electrochemically safe charge density, since reversible
reactions have more time to proceed.
Another apparent disadvantage of long pulses is that they do not permit frequency
coding or use of a retina encoder. Though this may be true to a certain extent, long
pulses target bipolar cells and cause RGCs to fire bursts of spikes (Freeman and Fried,
2011; Jensen and Rizzo, 2007; Jensen et al., 2005a). These bursts may represent the
natural firing patterns of ganglion cells in response to a spot of light. It could therefore
be argued that prostheses should target bipolar cells since they are situated earlier in the
visual pathway.
Others have argued that bipolar cell stimulation can be ineffective because responses
quickly become desensitized (Ahuja et al., 2008a; Freeman and Fried, 2011; Jensen and
Rizzo, 2007; Jensen et al., 2009). However, these studies used maximum pulse durations of
176
1–2 ms. It is unknown whether desensitization occurs with much longer pulses. The only
study to report spike rates during repetitive stimulation with long-duration waveforms
(5–25-Hz sinusoids) did not indicate any effects of desensitization (see Freeman et al.,
2010, Fig. 6).
Only two other studies to date have tested pulse durations >10 ms (Jensen et al.,
2005b; Margalit and Thoreson, 2006). Both concluded that bipolar cells were targeted by
long pulses; however, direct RGC stimulation was also observed. A typical ganglion cell
response would consist of a single spike elicited by direct activation, followed by a burst
of spikes due to indirect activation. Though I did not observe axonal responses to long
pulses, it is possible that distant cells were in fact stimulated antidromically and that their
spiking was too sparse to detect with GCaMP5G. Nevertheless, threshold maps indicated
that the largest concentration of spiking activity was localized to an area around the
electrode. Coupled with our human patient data showing that longer-duration stimuli
cause rounder percepts (see Fig. 4.2), it is not unreasonable to suggest that the large
number of bursting RGCs near the electrode dominates the perceptual response.
Compared to 25-ms pulses (i.e., 20-Hz square waves), I found that 20-Hz sine waves
produced similar activation patterns at slightly lower thresholds. This suggests that
low-frequency sinusoids may be a better alternative to long pulse widths. Furthermore,
electrical recordings have confirmed that 20-Hz square waves completely avoid RGC axons
at thresholds at least 7x higher than those of bipolar cells (Freeman et al., 2010).
By mapping responses to different electrode sizes, I found that response shapes did
not become smaller once electrode diameter fell below 75 m. With short pulses, this
electrode size also provided the best selectivity for RGC somata. 75 m might therefore
177
be an optimal electrode diameter for an epiretinal prosthesis. Spacing electrodes 150 m
apart (center-to-center) should enable a visual acuity of 20/312.
Finally, I have reported for the first time that retinal degeneration has little to no effect
on response shape. Bipolar cell thresholds are approximately 200% higher in degenerate
retina, while direct ganglion cell thresholds remain unchanged. Coupled with results from
my pattern stimulation experiments, my data suggest that multielectrode stimulation in
blind humans has the potential to restore high-acuity form vision.
178
Chapter 5
Conclusions and Future Work
Through the use of novel imaging tools, I have demonstrated several improvements to
stimulation strategies for epiretinal prostheses. My studies in mammalian and degenerate
retina were made possible by the development of specialized viral vectors for transducing
RGCs with genetically encoded calcium indicators. The large signals provided by these
indicators enabled me to visualize ganglion cell activity in real time. I built custom
MEAs arrays with small and transparent electrodes to investigate high-resolution pattern
stimulation of the retina. By leveraging my findings regarding stimulus pulse width, I was
able to evoke focal responses during stimulation with one or more electrodes. My work
culminated in a proof-of-concept demonstration of multielectrode pattern stimulation in
which I showed that controlled patterns of ganglion cells could be activated to form letters
and words.
5.1 Recommendations for Epiretinal Prostheses
My studies were motivated by the recent finding that epiretinal stimulation produces
elongated percepts in human patients, resulting from activation of RGC axon bundles
179
(Nanduri, 2011; Nanduri et al., 2008). I found that 0.45-ms pulses, which are currently
used by the prosthesis, provide essentially no means for avoiding these bundles. My
experiments did, however, reveal two competing strategies for producing focal responses:
At one extreme, short pulses can be delivered to preferentially activate RGC somata. At
the other, long-duration stimuli can be used to selectively target bipolar cells. Each of
these strategies has advantages and disadvantages and may be appropriate under different
circumstances.
Short pulses can stimulate ganglion cells at high frequencies. This makes it possible
to encode visual information by modulating ganglion cell spike rates. Short pulses also
require significantly less charge. Because epiretinal prostheses currently face the challenge
of remaining within electrochemical safety limits (Humayun et al., 2012), these pulses may
be more appropriate until closer coupling between the retina and MEA can be achieved.
When stimulating with short pulse widths, interphase gaps should be used in order to
reduce ganglion cell thresholds by up to 25%.
The problem with short-duration pulses is that they provide limited selectivity for
RGC somata. Increasing the stimulus amplitude only 60–70% above threshold causes
significant axonal activation. Given the effects of threshold adaptation in human patients,
it may be difficult to keep the stimulus amplitude within a range that excites ganglion
cell bodies and not their axons. For this reason, long-duration stimuli may be preferred.
By mapping spatial responses to long pulses and low-frequency sine waves, I found
that these stimuli produced focal activation patterns. Blocking synaptic transmission in-
dicated that responses were mediated by inner retinal stimulation. The fact that responses
persisted in degenerate retina suggests that long pulses target bipolar cells. Thresholds to
180
20-Hz sine waves were lower than those of 20-Hz square waves, indicating that sinusoids
may be a more effective waveform. Although long-duration stimuli may evoke sparse
neural activity in passing axons (Jensen et al., 2005b; Margalit and Thoreson, 2006),
response maps indicated that the highest concentration of spiking occurred in an area
around the electrode.
Without testing each stimulus in humans, it is difficult to recommend an optimal
stimulus waveform. In vitro response maps are good predictors of phosphene shape and
size, but they are an oversimplification of a highly complex problem. Response maps do
not account for threshold adaption, response desensitization, or phosphene fading. It is
also unknown to what extent cortical reorganization affects the spatiotemporal aspects
of phosphene perception. Furthermore, simultaneous activation of the ON and OFF
pathways may limit the quality of prosthetic vision. It is possible that stimulus design
affects each of these factors differently. For example, threshold adaptation may be more
pronounced with short pulses than with long ones. The only way to resolve this is to test
each stimulus in human subjects. I therefore propose that epiretinal prosthesis designs
should permit stimulation with short pulses (≤ 0.1 ms), long pulses (≥ 25 ms), and low-
frequency sinusoids (≤ 20 Hz). Human testing can then be performed to determine which
stimulus produces the highest-quality vision.
With regard to electrode size, I found that response shapes did not become smaller
once electrode diameter fell below 75m. This diameter also provided the best selectivity
when short pulses were used for stimulation. Furthermore, threshold charge densities for
75-m electrodes were electrochemically safe, regardless of pulse width. Based on these
results, I propose that next-generation epiretinal prostheses should incorporate an MEA
181
layout with 75-m-diameter electrodes on a 150-m grid. I have already demonstrated
successful pattern stimulation with this layout in vitro and have shown that it can achieve
a Snellen acuity of 20/312 with minimal electrode interactions.
5.2 Future Experiments
5.2.1 Retinal Transduction
By designing an AAV vector that transduces the majority of ganglion cells in rodent
retina, I have created a framework for delivering any gene of interest to RGCs. For
example, neurotrophic factors could be overexpressed in ganglion cells, which may slow
their loss during degeneration (Harvey et al., 2006; Martin et al., 2003). By taking a
viral delivery approach, rather than a transgenic one, it should be relatively quick and
easy to incorporate new genetically encoded calcium indicators as they become available.
GCaMP6 has recently been made available to the public (Chen et al., 2012a,b) and may
be sensitive enough to detect single spikes or spike doublets in our retina preparation.
As direct and indirect epiretinal activation strategies are further explored, it would
be useful to monitor bipolar cell activity during electrical stimulation. AAV-mediated
transduction of bipolar cells has been notoriously inefficient; however, new advances in
tyrosine-mutant AAV vectors have led to 20-fold increases in retinal transduction (Petrs-
Silva et al., 2010). Self-complementary AAV2/8-Y733F mutant vectors can selectively
target a significant portion of bipolar cells following subretinal injection (Doroudchi et al.,
2011). Coupled with the recent advent of red, yellow, and blue GECIs (Looger and
Griesbeck, 2011; Zhao et al., 2011), it should now be possible to transduce RGCs and
182
bipolar cells with indicators of different colors. The fluorescence of each cell class could
then be monitored simultaneously through a dual band-pass dichroic during electrical
stimulation. Such a strategy would enable direct determination of whether ganglion
and/or bipolar cells are excited by different types of stimuli. Indirect responses could
also be traced from bipolar cells to the ganglion cells with which they synapse.
Although AAV is a powerful gene delivery tool, there are some drawbacks to its use.
Experiments must be planned ahead of time, since ganglion cell transduction takes ap-
proximately two weeks. Furthermore, the extent of transduction varies among animals
and may depend on the quality of injection. In some cases, retinas exhibit little to no
GECI expression, presumably owing to viral leakage through the injection site. The suc-
cess rate of experiments could be improved by screening for sufficient retinal transduction
prior to animal sacrifice. A custom-built fundoscope would make this possible (Schejter
et al., 2012).
Another drawback to virally delivered GECIs is that overexpression leads to gan-
glion cell cytomorbidity as early as four weeks post-injection. Transgenic animals do not
have this problem, as expression levels remain stable over time (Zariwala et al., 2012).
Transgenic mice expressing GCaMP3 in RGCs have already been generated (Pvalb-2A-
Cre:Ai38; Zariwala et al., 2012). It may be beneficial to develop transgenic rats or mice
expressing newer GECIs such as GCaMP6. These animals could then be crossed with
rodent models of retinal degeneration. The downside to using transgenic lines is that
they are expensive and time-consuming to generate (Tong et al., 2011). There is also no
way to limit GECI expression to the ganglion cell layer.
183
5.2.2 Retinal Stimulation
My experiments have revealed improved methods for focal epiretinal stimulation; however,
some questions remain unanswered. First, it is unknown whether long-duration stimuli
cause ganglion cell responses to become desensitized. Desensitization has been reported
previously with indirect stimulation and may be responsible for the fading percepts seen
by humans (Ahuja et al., 2008a; Freeman and Fried, 2011; Jensen and Rizzo, 2007).
However, response desensitization has never been investigated for pulses longer than 1
ms or low-frequency sinusoids. A useful experiment would be to determine if and how
the degree of desensitization varies with pulse width. Doing so would require traditional
electrical recording techniques, as calcium imaging cannot be used for counting spikes.
Another open question regarding long-duration stimuli relates to their ability to avoid
ganglion cell axons. The only other study to report thresholds to long pulses (> 10 ms)
found that bipolar cell thresholds were only slightly lower than those of RGCs. However,
this study was conducted in rabbit retina and involved single-electrode recordings from
only a few cells. Throughout my experiments, I did not observe any axonal activation with
pulses≥ 25 ms, regardless of amplitude. It is possible that axons were in fact stimulated
and that GCaMP5G was not sensitive enough to detect their sparse responses. Further
investigation is needed to determine whether this was the case. One option would be to
express GCaMP6 in RGCs, as this sensor can reliably detect sparse neural activity (Chen
et al., 2012a). Otherwise, patch clamp or MEA recordings could be used for this purpose.
GECIs have proven to be a valuable tool for measuring the retina’s response to elec-
trical stimulation. Another possibility is to use them for reporting RGC activity during
184
light stimulation (Borghuis et al., 2011). This could be accomplished by projecting vis-
ible light onto the photoreceptors and using infrared two-photon excitation to measure
GCaMP responses. It would be interesting to image the pattern of cells excited by a visual
stimulus, such as a letter, and to see if that pattern could be replicated by multielectrode
stimulation.
185
References
Ahuja, A.K., Behrend, M.R., Kuroda, M., Humayun, M.S., and Weiland, J.D. An in vitro model
of a retinal prosthesis. IEEE Transactions on Biomedical Engineering, 55(6):1744–1753, 2008a.
Ahuja, A.K., Behrend, M.R., Whalen, J.J., Humayun, M.S., and Weiland, J.D. The depen-
dence of spectral impedance on disc microelectrode radius. IEEE Transactions on Biomedical
Engineering, 55(4):1457–1460, 2008b.
Ahuja, A.K., Dorn, J.D., Caspi, A., McMahon, M.J., Dagnelie, G., Dacruz, L., Stanga, P., Hu-
mayun, M.S., and Greenberg, R.J. Blind subjects implanted with the Argus II Retinal Prosthe-
sis are able to improve performance in a spatial-motor task. British Journal of Ophthalmology,
95(4):539–543, 2011.
Akerboom, J., Rivera, J.D.V., Guilbe, M.M.R., Malav´ e, E.C.A., Hernandez, H.H., Tian, L., Hires,
S.A., Marvin, J.S., Looger, L.L., and Schreiter, E.R. Crystal structures of the GCaMP calcium
sensor reveal the mechanism of fluorescence signal change and aid rational design. Journal of
Biological Chemistry, 284(10):6455–6464, 2009.
Akerboom, J., Chen, T.W., Wardill, T.J., Tian, L., Marvin, J.S., Mutlu, S., Calder´ on, N.C.,
Esposti, F., Borghuis, B.G., Sun, X.R., Gordus, A., Orger, M.B., Portugues, R., Engert, F.,
Macklin, J.J., Filosa, A., Aggarwal, A., Kerr, R., Takagi, R., Kracun, S., Shigetomi, E., Khakh,
B.S., Baier, H., Lagnado, L., Wang, S.S.H., Bargmann, C.I., Kimmel, B.E., Jayaraman, V.,
Svoboda, K., Kim, D.S., Schreiter, E.R., and Looger, L.L. Optimization of a GCaMP calcium
indicator for neural activity imaging. Journal of Neuroscience, 32(40):13819–13840, 2012.
Auricchio, A., Kobinger, G., Anand, V., Hildinger, M., O’Connor, E., Maguire, A.M., Wilson,
J.M., and Bennett, J. Exchange of surface proteins impacts on viral vector cellular specificity
and transduction characteristics: the retina as a model. Human Molecular Genetics, 10(26):
3075–3081, 2001.
Baird, G.S., Zacharias, D.A., and Tsien, R.Y. Circular permutation and receptor insertion within
green fluorescent proteins. Proceedings of the National Academy of Sciences, 96(20):11241–
11246, 1999.
Bak, M., Girvin, J.P., Hambrecht, F.T., Kufta, C.V., Loeb, G.E., and Schmidt, E.M. Visual
sensations produced by intracortical microstimulation of the human occipital cortex. Medical
and Biological Engineering and Computing, 28(3):257–259, 1990.
Baldridge, W.H. Optical recordings of the effects of cholinergic ligands on neurons in the ganglion
cell layer of mammalian retina. Journal of Neuroscience, 16(16):5060–5072, 1996.
Barsoukov, E. and Macdonald, J.R. Impedance spectroscopy: theory, experiment, and applications.
Wiley-Interscience, New York, NY, 2 edition, 2005.
186
Basinger, B.C., Rowley, A.P., Chen, K., Humayun, M.S., and Weiland, J.D. Finite element
modeling of retinal prosthesis mechanics. Journal of Neural Engineering, 6:055006, 2009.
Beebe, X. and Rose, T.L. Charge injection limits of activated iridium oxide electrodes with 0.2
ms pulses in bicarbonate buffered saline. IEEE Transactions onBiomedical Engineering, 35(6):
494–495, 1988.
Behrend, M.R. Viewing the picture we paint. PhD thesis, The University of Southern California,
2009.
Behrend, M.R., Ahuja, A.K., Humayun, M.S., Weiland, J.D., and Chow, R.H. Selective labeling
of retinal ganglion cells with calcium indicators by retrograde loading in vitro. Journal of
Neuroscience Methods, 179(2):166–172, 2009.
Behrend, M.R., Ahuja, A.K., Humayun, M.S., Chow, R.H., and Weiland, J.D. Resolution of the
epiretinal prosthesis is not limited by electrode size. IEEE Transactions on Neural Systems
and Rehabilitation Engineering, 19(4):436–441, 2011.
Benison, G., Keizer, J., Chalupa, L.M., and Robinson, D.W. Modeling temporal behavior of
postnatal cat retinal ganglion cells. Journal of Theoretical Biology, 210(2):187–199, 2001.
Berndt, A., Yizhar, O., Gunaydin, L.A., Hegemann, P., and Deisseroth, K. Bi-stable neural state
switches. Nature Neuroscience, 12(2):229–234, 2008.
Besch, D., Sachs, H., Szurman, P., G¨ ulicher, D., Wilke, R., Reinert, S., Zrenner, E., Bartz-
Schmidt, K.U., and Gekeler, F. Extraocular surgery for implantation of an active subretinal
visual prosthesis with external connections: feasibility and outcome in seven patients. British
Journal of Ophthalmology, 92(10):1361–1368, 2008.
Bhadra, N. and Kilgore, K.L. Direct current electrical conduction block of peripheral nerve. IEEE
Transactions on Neural Systems and Rehabilitation Engineering, 12(3):313–324, 2004.
Bi, A., Cui, J., Ma, Y.P., Olshevskaya, E., Pu, M., Dizhoor, A.M., and Pan, Z.H. Ectopic
expression of a microbial-type rhodopsin restores visual responses in mice with photoreceptor
degeneration. Neuron, 50(1):23–33, 2006.
Borghuis, B.G., Tian, L., Xu, Y., Nikonov, S.S., Vardi, N., Zemelman, B.V., and Looger, L.L.
Imaging light responses of targeted neuron populations in the rodent retina. Journal of Neu-
roscience, 31(8):2855–2867, 2011.
Boyden, E.S., Zhang, F., Bamberg, E., Nagel, G., and Deisseroth, K. Millisecond-timescale,
genetically targeted optical control of neural activity. Nature Neuroscience, 8(9):1263–1268,
2005.
Brel´ en, M.E., Duret, F., G´ erard, B., Delbeke, J., and Veraart, C. Creating a meaningful visual
perception in blind volunteers by optic nerve stimulation. Journal of Neural Engineering, 2:
S22–S28, 2005.
Briggman, K.L. and Euler, T. Bulk electroporation and population calcium imaging in the adult
mammalian retina. Journal of Neurophysiology, 105(5):2601–2609, 2011.
Brindley, G.S. Effects of electrical stimulation of the visual cortex. Human Neurobiology, 1(4):
281–283, 1982.
Brindley, G.S. and Lewin, W.S. The sensations produced by electrical stimulation of the visual
cortex. Journal of Physiology, 196(2):479–493, 1968.
187
Brindley, G.S. and Rushton, D. Implanted stimulators of the visual cortex as visual prosthetic
devices. Transactions of the American Academy of Ophthalmology and Otolaryngology, 78:
741–745, 1974.
Brummer, S.B. and Turner, M.J. Electrical stimulation with Pt electrodes: II—estimation of
maximum surface redox (theoretical non-gassing) limits. IEEE Transactions on Biomedical
Engineering, 24(5):440–443, 1977.
Brummer, S.B., McHardy, J., and Turner, M.J. Electrical stimulation with Pt electrodes: trace
analysis for dissolved platinum and other dissolved electrochemical products. Brain, Behavior
and Evolution, 14:10–22, 1977.
Burger, C., Nguyen, F.N., Deng, J., and Mandel, R.J. Systemic mannitol-induced hyperosmolal-
ity amplifies rAAV2-mediated striatal transduction to a greater extent than local co-infusion.
Molecular Therapy, 11(2):327–331, 2005.
Busskamp, V. and Roska, B. Optogenetic approaches to restoring visual function in retinitis
pigmentosa. Current Opinion in Neurobiology, 21:1–5, 2011.
Campbell, P.K., Jones, K.E., Huber, R.J., Horch, K.W., and Normann, R.A. A silicon-based,
three-dimensional neural interface: manufacturing processes for an intracortical electrode array.
IEEE Transactions on Biomedical Engineering, 38(8):758–768, 1991.
Carlson, G.A., Borchelt, D.R., Dake, A., Turner, S., Danielson, V., Coffin, J.D., Eckman, C.,
Meiners, J., Nilsen, S.P., and Younkin, S.G. Genetic modification of the phenotypes produced
by amyloid precursor protein overexpression in transgenic mice. Human Molecular Genetics, 6
(11):1951–1959, 1997.
Carlyon, R.P., van Wieringen, A., Deeks, J.M., Long, C.J., Lyzenga, J., and Wouters, J. Effect of
inter-phase gap on the sensitivity of cochlear implant users to electrical stimulation. Hearing
Research, 205(1):210–224, 2005.
Caspi, A., Dorn, J.D., McClure, K.H., Humayun, M.S., Greenberg, R.J., and McMahon, M.J.
Feasibility study of a retinal prosthesis: Spatial vision with a 16-electrode implant. Archives of
Ophthalmology, 127(4):398–401, 2009.
Cha, K., Horch, K., and Normann, R.A. Simulation of a phosphene-based visual field: visual
acuity in a pixelized vision system. Annals of Biomedical Engineering, 20(4):439–449, 1992a.
Cha, K., Horch, K.W., and Normann, R.A. Mobility performance with a pixelized vision system.
Vision Research, 32(7):1367–1372, 1992b.
Cha, K., Horch, K.W., Normann, R.A., and Boman, D.K. Reading speed with a pixelized vision
system. Journal of the Optical Society of America A, 9(5):673–677, 1992c.
Chader, G.J., Weiland, J., and Humayun, M.S. Artificial vision: needs, functioning, and testing
of a retinal electronic prosthesis. Progress in Brain Research, 175:317–332, 2009.
Chai, X., Li, L., Wu, K., Zhou, C., Cao, P., and Ren, Q. C-sight visual prostheses for the blind.
IEEE Engineering in Medicine and Biology Magazine, 27(5):20–28, 2008.
Chan, L.L.H., Lee, E.J., Humayun, M.S., and Weiland, J.D. Both electrical stimulation thresholds
and SMI-32-immunoreactive retinal ganglion cell density correlate with age in S334ter line 3
rat retina. Journal of Neurophysiology, 105(6):2687–2697, 2011.
188
Chang, B., Hawes, N.L., Hurd, R.E., Davisson, M.T., Nusinowitz, S., and Heckenlively, J.R.
Retinal degeneration mutants in the mouse. Vision Research, 42(4):517–525, 2002.
Chen, K., Yang, Z., Hoang, L., Weiland, J., Humayun, M., and Liu, W. An integrated 256-channel
epiretinal prosthesis. IEEE Journal of Solid-State Circuits, 45(9):1946–1956, 2010.
Chen, S.J., Mahadevappa, M., Roizenblatt, R., Weiland, J., and Humayun, M. Neural responses
elicited by electrical stimulation of the retina. Transactions of the American Ophthalmological
Society, 104:252–259, 2006.
Chen, T.W., Wardill, T.J., Hasseman, J.P., Tsegaye, G., Fosque, B.F., Schreiter, E.R., Kimmel,
B.E., Kerr, R.A., Jayaraman, V., Svoboda, K., Looger, L.L., and Kim, D.S. Engineering next
generation GCaMP calcium indicators using neuron-based screening. In Neuroscience 2012,
2012a.
Chen, T.W., Yu, J., Kerr, R.A., Jayaraman, V., Looger, L.L., Svoboda, K., and Kim, D.S. Reliable
detection of single action potentials and synaptic calcium signals using improved genetically-
encoded calcium indicators. In Neuroscience 2012, 2012b.
Chow, A.Y., Chow, V.Y., Packo, K.H., Pollack, J.S., Peyman, G.A., and Schuchard, R. The
artificial silicon retina microchip for the treatment of vision loss from retinitis pigmentosa.
Archives of Ophthalmology, 122(4):460–469, 2004.
Cideciyan, A.V. Leber congenital amaurosis due to RPE65 mutations and its treatment with gene
therapy. Progress in Retinal and Eye Research, 29:398–427, 2010.
Cobbold, P.H. and Rink, T.J. Fluorescence and bioluminescence measurement of cytoplasmic free
calcium. Biochemical Journal, 248(2):313–328, 1987.
Colodetti, L., Weiland, J.D., Colodetti, S., Ray, A., Seiler, M.J., Hinton, D.R., and Humayun,
M.S. Pathology of damaging electrical stimulation in the retina. Experimental Eye Research,
85(1):23–33, 2007.
Crapper, D.R. and Noell, W.K. Retinal excitation and inhibition from direct electrical stimulation.
Journal of Neurophysiology, 26(6):924–947, 1963.
Cunningham, R.F., Israili, Z.H., and Dayton, P.G. Clinical pharmacokinetics of probenecid.
Clinical Pharmacokinetics, 6(2):135–151, 1981.
Curcio, C.A., Sloan, K.R., Kalina, R.E., and Hendrickson, A.E. Human photoreceptor topography.
Journal of Comparative Neurology, 292(4):497–523, 1990.
da Cruz, L., Coley, B., Christopher, P., Merlini, F., Wuyyuru, V., Sahel, J.A., Stanga, P., Filley,
E., and Dagnelie, G. Patients blinded by outer retinal dystrophies are able to identify letters
using the Argus II Retinal Prosthesis System. In ARVO 2010, 2010.
Dagnelie, G. Visual Prosthetics: Physiology, Bioengineering, Rehabilitation. Springer, 1 edition,
2011.
Dalkara, D., Kolstad, K.D., Caporale, N., Visel, M., Klimczak, R.R., Schaffer, D.V., and Flannery,
J.G. Inner limiting membrane barriers to AAV-mediated retinal transduction from the vitreous.
Molecular Therapy, 17(12):2096–2102, 2009.
Daya, S. and Berns, K.I. Gene therapy using adeno-associated virus vectors. Clinical Microbiology
Reviews, 21(4):583–593, 2008.
189
Dayton, R.D., Wang, D.B., and Klein, R.L. The advent of AAV9 expands applications for brain
and spinal cord gene delivery. Expert Opinion on Biological Therapy, 12(6):757–766, 2012.
de Balthasar, C., Patel, S., Roy, A., Freda, R., Greenwald, S., Horsager, A., Mahadevappa, M.,
Yanai, D., McMahon, M.J., and Humayun, M.S. Factors affecting perceptual thresholds in
epiretinal prostheses. Investigative Ophthalmology and Visual Science, 49(6):2303–2314, 2008.
Deisseroth, K. Optogenetics. Nature Methods, 8(1):26–29, 2010.
DeMarco, P.J., Yarbrough, G.L., Yee, C.W., McLean, G.Y., Sagdullaev, B.T., Ball, S.L., and
McCall, M.A. Stimulation via a subretinally placed prosthetic elicits central activity and induces
a trophic effect on visual responses. Investigative Ophthalmology and Visual Science, 48(2):916–
926, 2007.
den Hollander, A.I., Roepman, R., Koenekoop, R.K., and Cremers, F.P.M. Leber congenital
amaurosis: genes, proteins and disease mechanisms. Progress in Retinal and Eye Research, 27
(4):391–419, 2008.
Dezawa, M., Takano, M., Negishi, H., Mo, X., Oshitari, T., and Sawada, H. Gene transfer into
retinal ganglion cells by in vivo electroporation: a new approach. Micron, 33(1):1–6, 2002.
Di Virgilio, F., Fasolato, C., and Steinberg, T.H. Inhibitors of membrane transport system for
organic anions block fura-2 excretion from PC12 and N2A cells. Biochemical Journal, 256(3):
959–963, 1988.
Dobelle, W.H. and Mladejovsky, M.G. Phosphenes produced by electrical stimulation of human
occipital cortex, and their application to the development of a prosthesis for the blind. Journal
of Physiology, 243(2):553–576, 1974.
Dobelle, W.H., Mladejovsky, M.G., and Girvin, J.P. Artificial vision for the blind: electrical
stimulation of visual cortex offers hope for a functional prosthesis. Science, 183(4123):440–444,
1974.
Dobelle, W.H., Quest, D.O., Antunes, J.L., Roberts, T.S., and Girvin, J.P. Artificial vision for
the blind by electrical stimulation of the visual cortex. Neurosurgery, 5(4):521–527, 1979.
Doroudchi, M.M., Greenberg, K.P., Liu, J., Silka, K.A., Boyden, E.S., Lockridge, J.A., Arman,
A.C., Janani, R., Boye, S.E., and Boye, S.L. Virally delivered channelrhodopsin-2 safely and
effectively restores visual function in multiple mouse models of blindness. Molecular Therapy,
19(7):1220–1229, 2011.
Drasdo, N. and Fowler, C.W. Non-linear projection of the retinal image in a wide-angle schematic
eye. British Journal Of Ophthalmology, 58(8):709–714, 1974.
Duret, F., Brel´ en, M.E., Lambert, V., G´ erard, B., Delbeke, J., and Veraart, C. Object localization,
discrimination, and grasping with the optic nerve visual prosthesis. Restorative Neurology and
Neuroscience, 24(1):31–40, 2006.
Eckmiller, R., Neumann, D., and Baruth, O. Tunable retina encoders for retina implants: why
and how. Journal of Neural Engineering, 2:S91–S104, 2005.
Eversmann, B., Jenkner, M., Hofmann, F., Paulus, C., Brederlow, R., Holzapfl, B., Fromherz, P.,
Merz, M., Brenner, M., and Schreiter, M. A 128 x 128 CMOS biosensor array for extracellular
recording of neural activity. IEEE Journal of Solid-State Circuits, 38(12):2306–2317, 2003.
190
Fang, X., Sakaguchi, H., Fujikado, T., Osanai, M., Ikuno, Y., Kamei, M., Ohji, M., Yagi, T., and
Tano, Y. Electrophysiological and histological studies of chronically implanted intrapapillary
microelectrodes in rabbit eyes. Graefe’s Archive for Clinical and Experimental Ophthalmology,
244(3):364–375, 2006.
Fariss, R.N., Li, Z.Y., and Milam, A.H. Abnormalities in rod photoreceptors, amacrine cells, and
horizontal cells in human retinas with retinitis pigmentosa. American Journal of Ophthalmology,
129(2):215–223, 2000.
Feng, G., Mellor, R.H., Bernstein, M., Keller-Peck, C., Nguyen, Q.T., Wallace, M., Nerbonne,
J.M., Lichtman, J.W., and Sanes, J.R. Imaging neuronal subsets in transgenic mice expressing
multiple spectral variants of GFP. Neuron, 28(1):41–51, 2000.
Ferrea, E., Maccione, A., Medrihan, L., Nieus, T., Ghezzi, D., Baldelli, P., Benfenati, F., and
Berdondini, L. Large-scale, high-resolution electrophysiological imaging of field potentials in
brain slices with microelectronic multielectrode arrays. Frontiers in Neural Circuits, 6:1–14,
2012.
Field, G.D., Gauthier, J.L., Sher, A., Greschner, M., Machado, T.A., Jepson, L.H., Shlens, J.,
Gunning, D.E., Mathieson, K., and Dabrowski, W. Functional connectivity in the retina at the
resolution of photoreceptors. Nature, 467(7316):673–677, 2010.
Fink, W., Tarbell, M., Weiland, J.D., and Humayun, M.S. DORA: digital object recognition
audio-assistant for the visually impaired. Invest. Ophthalmol. Vis. Sci., 45(4201), 2004.
Finlayson, P.G. and Iezzi, R. Glutamate stimulation of retinal ganglion cells in normal and
S334ter-4 rat retinas: a candidate for a neurotransmitter-based retinal prosthesis. Investigative
Ophthalmology and Visual Science, 51(7):3619–3628, 2010.
Foerster, O. Beitrage zur pathophysiologie der sehbahn und der spehsphare. J Psychol Neurol,
39:435–63, 1929.
Fohlmeister, J.F. and Miller, R.F. Impulse encoding mechanisms of ganglion cells in the tiger
salamander retina. Journal of Neurophysiology, 78(4):1935–1947, 1997.
Fohlmeister, J.F., Coleman, P.A., and Miller, R.F. Modeling the repetitive firing of retinal gan-
glion cells. Brain Research, 510(2):343–345, 1990.
Franks, W., Schenker, I., Schmutz, P., and Hierlemann, A. Impedance characterization and mod-
eling of electrodes for biomedical applications. IEEE Transactions on Biomedical Engineering,
52(7):1295–1302, 2005.
Freeman, D.K. and Fried, S.I. Multiple components of ganglion cell desensitization in response
to prosthetic stimulation. Journal of Neural Engineering, 8:016008, 2011.
Freeman, D.K., Eddington, D.K., Rizzo, J.F., and Fried, S.I. Selective activation of neuronal
targets with sinusoidal electric stimulation. Journal of Neurophysiology, 104(5):2778–2791,
2010.
Freeman, D.K., Rizzo, J.F., and Fried, S.I. Encoding visual information in retinal ganglion cells
with prosthetic stimulation. Journal of Neural Engineering, 8:035005, 2011.
Fried, S.I., Hsueh, H.A., and Werblin, F.S. A method for generating precise temporal patterns of
retinal spiking using prosthetic stimulation. Journal of Neurophysiology, 95(2):970–978, 2006.
191
Fried, S.I., Lasker, A.C.W., Desai, N.J., Eddington, D.K., and Rizzo, J.F. Axonal sodium-
channel bands shape the response to electric stimulation in retinal ganglion cells. Journal of
Neurophysiology, 101(4):1972–1987, 2009.
Fujikado, T., Kamei, M., Sakaguchi, H., Kanda, H., Morimoto, T., Ikuno, Y., Nishida, K.,
Kishima, H., Maruo, T., and Konoma, K. Testing of semichronically implanted retinal prosthe-
sis by suprachoroidal-transretinal stimulation in patients with retinitis pigmentosa. Investigative
Ophthalmology and Visual Science, 52(7):4726–4733, 2011.
Garc´ ıa-Ayuso, D., Salinas-Navarro, M., Agudo, M., Cuenca, N., Pinilla, I., Vidal-Sanz, M., and
Villegas-P´ erez, M.P. Retinal ganglion cell numbers and delayed retinal ganglion cell death in
the P23H rat retina. Experimental Eye Research, 91:800–810, 2010.
Garcia-Frigola, C., Carreres, M., Vegar, C., and Herrera, E. Gene delivery into mouse retinal
ganglion cells by in utero electroporation. BMC Developmental Biology, 7(1):103, 2007.
Gargini, C., Terzibasi, E., Mazzoni, F., and Strettoi, E. Retinal organization in the retinal
degeneration 10 (rd10) mutant mouse: a morphological and ERG study. Journal of Comparative
Neurology, 500(2):222–238, 2006.
Geddes, L.A. Accuracy limitations of chronaxie values. IEEE Transactions on Biomedical Engi-
neering, 51(1):176–181, 2004.
Gehrs, K.M., Anderson, D.H., Johnson, L.V., and Hageman, G.S. Age-related macular
degeneration-emerging pathogenetic and therapeutic concepts. Annals of Medicine, 38(7):450–
471, 2006.
Gholmieh, G., Soussou, W., Han, M., Ahuja, A., Hsiao, M.C., Song, D., Tanguay Jr., A.R., and
Berger, T.W. Custom-designed high-density conformal planar multielectrode arrays for brain
slice electrophysiology. Journal of Neuroscience Methods, 152(1-2):116–129, 2006.
Gorman, P.H. and Mortimer, J.T. The effect of stimulus parameters on the recruitment char-
acteristics of direct nerve stimulation. IEEE Transactions on Biomedical Engineering, 30(7):
407–414, 1983.
Green, R.A., Matteucci, P.B., Hassarati, R.T., Giraud, B., Dodds, C.W.D., Chen, S., Byrnes-
Preston, P.J., Suaning, G.J., Poole-Warren, L.A., and Lovell, N.H. Performance of conducting
polymer electrodes for stimulating neuroprosthetics. Journal of Neural Engineering, 10(1):
016009, 2013.
Greenberg, R.J. Analysis of electrical stimulation of the vertebrate retina: work towards a retinal
prosthesis. PhD thesis, The Johns Hopkins University, 1998.
Greenberg, R.J., Velte, T.J., Humayun, M.S., Scarlatis, G.N., and de Juan Jr., E. A computational
model of electrical stimulation of the retinal ganglion cell. IEEE Transactions on Biomedical
Engineering, 46(5):505–514, 1999.
Greenwald, S.H., Horsager, A., Humayun, M.S., Greenberg, R.J., McMahon, M.J., and Fine, I.
Brightness as a function of current amplitude in human retinal electrical stimulation. Inves-
tigative Ophthalmology and Visual Science, 50(11):5017–5025, 2009.
Griesbeck, O., Baird, G.S., Campbell, R.E., Zacharias, D.A., and Tsien, R.Y. Reducing the
environmental sensitivity of yellow fluorescent protein. Journal of Biological Chemistry, 276
(31):29188–29194, 2001.
192
Grill, W.M. and Mortimer, J.T. Electrical properties of implant encapsulation tissue. Annals of
Biomedical Engineering, 22(1):23–33, 1994.
Grill, W.M. and Mortimer, J.T. Quantification of recruitment properties of multiple contact cuff
electrodes. IEEE Transactions on Rehabilitation Engineering, 4(2):49–62, 1996.
Grimm, D. and Kay, M.A. From virus evolution to vector revolution: use of naturally occurring
serotypes of adeno-associated virus (AAV) as novel vectors for human gene therapy. Current
Gene Therapy, 3(4):281–304, 2003.
Grumet, A.E., Wyatt, J.L., and Rizzo, J.F. Multi-electrode stimulation and recording in the
isolated retina. Journal of Neuroscience Methods, 101(1):31–42, 2000.
Grynkiewicz, G., Poenie, M., and Tsien, R.Y. A new generation of Ca
2+
indicators with greatly
improved fluorescence properties. Journal of Biological Chemistry, 260(6):3440–3450, 1985.
Hartong, D.T., Berson, E.L., and Dryja, T.P. Retinitis pigmentosa. The Lancet, 368(9549):
1795–1809, 2006.
Hartwick, A.T.E., Lalonde, M.R., Barnes, S., and Baldridge, W.H. Adenosine A1-receptor mod-
ulation of glutamate-induced calcium influx in rat retinal ganglion cells. Investigative Ophthal-
mology and Visual Science, 45(10):3740–3748, 2004.
Hartwick, A.T.E., Zhang, X., Chauhan, B.C., and Baldridge, W.H. Functional assessment of
glutamate clearance mechanisms in a chronic rat glaucoma model using retinal ganglion cell
calcium imaging. Journal of Neurochemistry, 94(3):794–807, 2005.
Harvey, A.R., Kamphuis, W., Eggers, R., Symons, N.A., Blits, B., Niclou, S., Boer, G.J., and
Verhaagen, J. Intravitreal injection of adeno-associated viral vectors results in the transduction
of different types of retinal neurons in neonatal and adult rats: a comparison with lentiviral
vectors. Molecular and Cellular Neuroscience, 21(1):141–157, 2002.
Harvey, A.R., Hu, Y., Leaver, S.G., Mellough, C.B., Park, K., Verhaagen, J., Plant, G.W., and
Cui, Q. Gene therapy and transplantation in CNS repair: the visual system. Progress in Retinal
and Eye Research, 25(5):449–489, 2006.
Hasan, M.T., Friedrich, R.W., Euler, T., Larkum, M.E., Giese, G., Both, M., Duebel, J., Waters,
J., Bujard, H., and Griesbeck, O. Functional fluorescent Ca
2+
indicator proteins in transgenic
mice under TET control. PLoS Biology, 2(6):0763–0775, 2004.
Hauer, M.C., Nasiatka, P.J., Stiles, N.R., Lue, J.C.L., Agrawal, R., Weiland, J.D., Humayun,
M.S., and Tanguay Jr., A.R. Intraocular camera for retinal prostheses: optical design. Frontiers
In Optics, Optical Society of America, 2007.
Hayes, J.S., Yin, V.T., Piyathaisere, D., Weiland, J.D., Humayun, M.S., and Dagnelie, G. Visually
guided performance of simple tasks using simulated prosthetic vision. Artificial Organs , 27(11):
1016–1028, 2003.
Heilbronn, R. and Weger, S. Viral vectors for gene transfer: current status of gene therapeutics.
Handbook of Experimental Pharmacology, 197:143–170, 2010.
Hellstr¨ om, M., Ruitenberg, M.J., Pollett, M.A., Ehlert, E.M.E., Twisk, J., Verhaagen, J., and
Harvey, A.R. Cellular tropism and transduction properties of seven adeno-associated viral
vector serotypes in adult retina after intravitreal injection. Gene Therapy, 16(4):521–532, 2008.
193
Hochberg, L.R., Serruya, M.D., Friehs, G.M., Mukand, J.A., Saleh, M., Caplan, A.H., Branner,
A., Chen, D., Penn, R.D., and Donoghue, J.P. Neuronal ensemble control of prosthetic devices
by a human with tetraplegia. Nature, 442(7099):164–171, 2006.
Homma, R., Baker, B.J., Jin, L., Garaschuk, O., Konnerth, A., Cohen, L.B., and Zecevic, D.
Wide-field and two-photon imaging of brain activity with voltage- and calcium-sensitive dyes.
Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1529):2453–2467,
2009.
Hornig, R., Zehnder, T., Velikay-Parel, M., Laube, T., Feucht, M., and Richard, G. The IMI
Retinal Implant System. Artificial Sight: Basic Research, Biomedical Engineering, and Clinical
Advances. Springer, 2008.
Horsager, A., Greenwald, S.H., Weiland, J.D., Humayun, M.S., Greenberg, R.J., McMahon, M.J.,
Boynton, G.M., and Fine, I. Predicting visual sensitivity in retinal prosthesis patients. Inves-
tigative Ophthalmology and Visual Science, 50(4):1483–1491, 2009.
Horsager, A., Greenberg, R.J., and Fine, I. Spatiotemporal interactions in retinal prosthesis
subjects. Investigative Ophthalmology and Visual Science, 51(2):1223–1233, 2010.
Horsager, A., Boynton, G.M., Greenberg, R.J., and Fine, I. Temporal interactions during paired-
electrode stimulation in two retinal prosthesis subjects. Investigative Ophthalmology and Visual
Science, 52(1):549–557, 2011.
Horton, J.C., Greenwood, M.M., and Hubel, D.H. Non-retinotopic arrangement of fibres in cat
optic nerve. Nature, 282(5740):720–722, 1979.
Humayun, M., Propst, R., de Juan Jr., E., McCormick, K., and Hickingbotham, D. Bipolar
surface electrical stimulation of the vertebrate retina. Archives of Ophthalmology, 112(1):110–
116, 1994.
Humayun, M.S., de Juan Jr., E., Dagnelie, G., Greenberg, R.J., Propst, R.H., and Phillips,
D.H. Visual perception elicited by electrical stimulation of retina in blind humans. Archives of
Ophthalmology, 114(1):40–46, 1996.
Humayun, M.S., De Juan Jr., E., Weiland, J.D., Dagnelie, G., Katona, S., Greenberg, R., and
Suzuki, S. Pattern electrical stimulation of the human retina. Vision Research, 39(15):2569–
2576, 1999a.
Humayun, M.S., Prince, M., de Juan Jr., E., Barron, Y., Moskowitz, M., Klock, I.B., and Milam,
A.H. Morphometric analysis of the extramacular retina from postmortem eyes with retinitis
pigmentosa. Investigative Ophthalmology and Visual Science, 40(1):143–148, 1999b.
Humayun, M.S., Weiland, J.D., Justus, B., Merrit, C., Whalen, J., Piyathaisere, D., Chen, S.J.,
Margalit, E., Fujii, G., and Greenberg, R.J. Towards a completely implantable, light-sensitive
intraocular retinal prosthesis. In 23rd Annual International IEEE EMBS Conference, pages
3422–3425. IEEE, 2001.
Humayun, M.S., Weiland, J.D., Fujii, G.Y., Greenberg, R., Williamson, R., Little, J., Mech, B.,
Cimmarusti, V., Van Boemel, G., and Dagnelie, G. Visual perception in a blind subject with
a chronic microelectronic retinal prosthesis. Vision Research, 43(24):2573–2581, 2003.
Humayun, M.S., Dorn, J.D., da Cruz, L., Dagnelie, G., Sahel, J.A., Stanga, P.E., Cideciyan, A.V.,
Duncan, J.L., Eliott, D., and Filley, E. Interim results from the international trial of Second
Sight’s visual prosthesis. Ophthalmology, 2012.
194
Hutzler, M., Lambacher, A., Eversmann, B., Jenkner, M., Thewes, R., and Fromherz, P. High-
resolution multitransistor array recording of electrical field potentials in cultured brain slices.
Journal of Neurophysiology, 96(3):1638–1645, 2006.
Inouye, S., Noguchi, M., Sakaki, Y., Takagi, Y., Miyata, T., Iwanaga, S., and Tsuji, FI. Cloning
and sequence analysis of cDNA for the luminescent protein aequorin. Proceedings of the National
Academy of Sciences, 82(10):3154–3158, 1985.
Jensen, R.J. Activation of ganglion cells in wild-type and P23H rat retinas with a small subretinal
electrode. Experimental Eye Research, 99:71–77, 2012.
Jensen, R.J. and Rizzo, J.F. Responses of ganglion cells to repetitive electrical stimulation of the
retina. Journal of Neural Engineering, 4:S1–S6, 2007.
Jensen, R.J. and Rizzo, J.F. Activation of retinal ganglion cells in wild-type and rd1 mice through
electrical stimulation of the retinal neural network. Vision Research, 48(14):1562–1568, 2008.
Jensen, R.J. and Rizzo, J.F. Activation of ganglion cells in wild-type and rd1 mouse retinas with
monophasic and biphasic current pulses. Journal of Neural Engineering, 6(3):035004, 2009.
Jensen, R.J., Rizzo, J.F., Ziv, O.R., Grumet, A., and Wyatt, J. Thresholds for activation of
rabbit retinal ganglion cells with an ultrafine, extracellular microelectrode. Investigative Oph-
thalmology and Visual Science, 44(8):3533–3543, 2003.
Jensen, R.J., Ziv, O.R., and Rizzo, J.F. Responses of rabbit retinal ganglion cells to electrical
stimulation with an epiretinal electrode. Journal of Neural Engineering, 2:S16–S21, 2005a.
Jensen, R.J., Ziv, O.R., and Rizzo, J.F. Thresholds for activation of rabbit retinal ganglion cells
with relatively large, extracellular microelectrodes. Investigative Ophthalmology and Visual
Science, 46(4):1486–1496, 2005b.
Jensen, R.J., Ziv, O.R., Rizzo, J.F., Scribner, D., and Johnson, L. Spatiotemporal aspects of
pulsed electrical stimuli on the responses of rabbit retinal ganglion cells. Experimental Eye
Research, 89(6):972–979, 2009.
Jeon, C.J., Strettoi, E., and Masland, R.H. The major cell populations of the mouse retina.
Journal of Neuroscience, 18(21):8936–8946, 1998.
Jones, B.W., Watt, C.B., Frederick, J.M., Baehr, W., Chen, C.K., Levine, E.M., Milam, A.H.,
Lavail, M.M., and Marc, R.E. Retinal remodeling triggered by photoreceptor degenerations.
Journal of Comparative Neurology, 464(1):1–16, 2003.
Kang, J., Lemaire, H.G., Unterbeck, A., Salbaum, J.M., Masters, C.L., Grzeschik, K.H.,
Multhaup, G., Beyreuther, K., and M¨ uller-Hill, B. The precursor of Alzheimer’s disease amyloid
A4 protein resembles a cell-surface receptor. Nature, 325:733–736, 1987.
Karny, H. Clinical and physiological aspects of the cortical visual prosthesis. Survey of Ophthal-
mology, 20(1):47–58, 1975.
Keat, J., Reinagel, P., Reid, R.C., and Meister, M. Predicting every spike: a model for the
responses of visual neurons. Neuron, 30(3):803–817, 2001.
Keeler, C.E. The inheritance of a retinal abnormality in white mice. Proceedings of the National
Academy of Sciences, 10(7):329–333, 1924.
195
Kettunen, P., Demas, J., Lohmann, C., Kasthuri, N., Gong, Y., Wong, R.O.L., and Gan, W.B.
Imaging calcium dynamics in the nervous system by means of ballistic delivery of indicators.
Journal of Neuroscience Methods, 119(1):37–43, 2002.
Khamdang, S., Takeda, M., Shimoda, M., Noshiro, R., Narikawa, S., Huang, X.L., Enomoto, A.,
Piyachaturawat, P., and Endou, H. Interactions of human- and rat-organic anion transporters
with pravastatin and cimetidine. Journal of Pharmacological Sciences, 94(2):197–202, 2004.
Kim, S.Y., Sadda, S., Humayun, M.S., de Juan Jr., E., Melia, B.M., and Green, W.R. Mor-
phometric analysis of the macula in eyes with geographic atrophy due to age-related macular
degeneration. Retina, 22(4):464–470, 2002a.
Kim, S.Y., Sadda, S., Pearlman, J., Humayun, M.S., de Juan Jr., E., Melia, B.M., and Green,
W.R. Morphometric analysis of the macula in eyes with disciform age-related macular degen-
eration. Retina, 22(4):471–477, 2002b.
Kinney, H.C. and Sidman, R.L. Pathology of the spongiform encephalopathy in the gray tremor
mutant mouse. Journal of Neuropathology and Experimental Neurology, 45(2):108–126, 1986.
Klauke, S., Goertz, M., Rein, S., Hoehl, D., Thomas, U., Eckhorn, R., Bremmer, F., and Wachtler,
T. Stimulation with a wireless intraocular epiretinal implant elicits visual percepts in blind
humans. Investigative Ophthalmology and Visual Science, 52(1):449–455, 2011.
Klein, R.L., Hamby, M.E., Gong, Y., Hirko, A.C., Wang, S., Hughes, J.A., King, M.A., and
Meyer, E.M. Dose and promoter effects of adeno-associated viral vector for green fluorescent
protein expression in the rat brain. Experimental Neurology, 176(1):66–74, 2002.
Kolomiets, B., Dubus, E., Simonutti, M., Rosolen, S., Sahel, J.A., and Picaud, S. Late histological
and functional changes in the P23H rat retina after photoreceptor loss. Neurobiology of Disease,
38(1):47–58, 2010.
Kolstad, K.D., Dalkara, D., Guerin, K., Visel, M., Hoffmann, N., Schaffer, D.V., and Flannery,
J.G. Changes in adeno-associated virus-mediated gene delivery in retinal degeneration. Human
Gene Therapy, 21(5):571–578, 2010.
Kralj, J.M., Douglass, A.D., Hochbaum, D.R., Maclaurin, D., and Cohen, A.E. Optical recording
of action potentials in mammalian neurons using a microbial rhodopsin. Nature Methods, 9(1):
90–95, 2011.
Krause, F. and Schum, H. Die epileptischen Erkrankungen. Neue Deutsche Chirurgie, 49a:482–
486, 1931.
Kuhn, B., Ozden, I., Lampi, Y., Hasan, M.T., and Wang, S.S.H. An amplified promoter system for
targeted expression of calcium indicator proteins in the cerebellar cortex. Frontiers in Neural
Circuits, 6(49):1–12, 2012.
Kwon, I. and Schaffer, D.V. Designer gene delivery vectors: molecular engineering and evolution
of adeno-associated viral vectors for enhanced gene transfer. Pharmaceutical Research, 25(3):
489–499, 2008.
Lakhanpal, R.R., Yanai, D., Weiland, J.D., Fujii, G.Y., Caffey, S., Greenberg, R.J., de Juan Jr.,
E., and Humayun, M.S. Advances in the development of visual prostheses. Current Opinion
in Ophthalmology, 14(3):122–127, 2003.
196
Lapicque, L. Recherches quantitatives sur l’excitation ´ electrique des nerfs trait´ ee comme une
polarisation. Journal de Physiologie et de Pathologie G´ en´ erale, 9:620–635, 1907.
Le Bec, C. and Douar, A.M. Gene therapy progress and prospects — vectorology: design and
production of expression cassettes in AAV vectors. Gene Therapy, 13(10):805–813, 2006.
LeRoy, C. O` u l’on rend compte de quelques tentatives que l’on a faites pour gu´ erir plusieurs
maladies par l’´ electricit´ e. Hist. Acad. Roy. Sciences (Paris), M´ emoires Math. Phys., 60:87–95,
1755.
Li, C., Hirsch, M., DiPrimio, N., Asokan, A., Goudy, K., Tisch, R., and Samulski, R.J. Cytotoxic-
T-lymphocyte-mediated elimination of target cells transduced with engineered adeno-associated
virus type 2 vector in vivo. Journal of Virology, 83(13):6817–6824, 2009.
Lin, J.Y., Lin, M.Z., Steinbach, P., and Tsien, R.Y. Characterization of engineered channel-
rhodopsin variants with improved properties and kinetics. Biophysical Journal, 96(5):1803–
1814, 2009.
Litke, A.M., Bezayiff, N., Chichilnisky, E.J., Cunningham, W., Dabrowski, W., Grillo, A.A.,
Grivich, M., Grybos, P., Hottowy, P., and Kachiguine, S. What does the eye tell the brain?:
development of a system for the large-scale recording of retinal output activity. IEEE Trans-
actions on Nuclear Science, 51(4):1434–1440, 2004.
Liu, C., Li, Y., Peng, M., Laties, A.M., and Wen, R. Activation of caspase-3 in the retina
of transgenic rats with the rhodopsin mutation S334ter during photoreceptor degeneration.
Journal of Neuroscience, 19(12):4778–4785, 1999.
Loeb, J.E., Cordier, W.S., Harris, M.E., Weitzman, M.D., and Hope, T.J. Enhanced expression
of transgenes from adeno-associated virus vectors with the woodchuck hepatitis virus posttran-
scriptional regulatory element: implications for gene therapy. Human Gene Therapy, 10(14):
2295–2305, 1999.
Lohmann, C., Myhr, K.L., and Wong, R.O.L. Transmitter-evoked local calcium release stabilizes
developing dendrites. Nature, 418(6894):177–181, 2002.
Looger, L.L. and Griesbeck, O. Genetically encoded neural activity indicators. Current Opinion
in Neurobiology, 22:18–23, 2011.
Lund, R.D., Wang, S., Klimanskaya, I., Holmes, T., Ramos-Kelsey, R., Lu, B., Girman, S.,
Bischoff, N., Sauv´ e, Y., and Lanza, R. Human embryonic stem cell-derived cells rescue visual
function in dystrophic RCS rats. Cloning and Stem Cells, 8(3):189–199, 2006.
Luo, D. and Saltzman, W.M. Synthetic DNA delivery systems. Nature Biotechnology, 18(1):
33–37, 2000.
MacDonald, J.R. Note on the parameterization of the constant-phase admittance element. Solid
State Ionics, 13(2):147–149, 1984.
Machida, S., Kondo, M., Jamison, J.A., Khan, N.W., Kononen, L.T., Sugawara, T., Bush,
R.A., and Sieving, P.A. P23H rhodopsin transgenic rat: correlation of retinal function with
histopathology. Investigative Ophthalmology and Visual Science, 41(10):3200–3209, 2000.
Mahadevappa, M., Weiland, J.D., Yanai, D., Fine, I., Greenberg, R.J., and Humayun, M.S. Per-
ceptual thresholds and electrode impedance in three retinal prosthesis subjects. IEEE Trans-
actions on Neural Systems and Rehabilitation Engineering, 13(2):201–206, 2005.
197
Mank, M. and Griesbeck, O. Genetically encoded calcium indicators. Chemical Reviews, 108(5):
1550–1564, 2008.
Marc, R.E., Jones, B.W., Watt, C.B., and Strettoi, E. Neural remodeling in retinal degeneration.
Progress in Retinal and Eye Research, 22(5):607–655, 2003.
Marg, E. Magnetostimulation of vision: direct noninvasive stimulation of the retina and the visual
brain. Optometry and Vision Science, 68(6):427–440, 1991.
Margalit, E. and Thoreson, W.B. Inner retinal mechanisms engaged by retinal electrical stimula-
tion. Investigative Ophthalmology and Visual Science, 47(6):2606–2612, 2006.
Margalit, E., Maia, M., Weiland, J.D., Greenberg, R.J., Fujii, G.Y., Torres, G., Piyathaisere,
D.V., O’Hearn, T.M., Liu, W., and Lazzi, G. Retinal prosthesis for the blind. Survey of
Ophthalmology, 47(4):335–356, 2002.
Margolis, D.J., Newkirk, G., Euler, T., and Detwiler, P.B. Functional stability of retinal ganglion
cells after degeneration-induced changes in synaptic input. Journal of Neuroscience, 28(25):
6526–6536, 2008.
Martin, K.R.G., Klein, R.L., and Quigley, H.A. Gene delivery to the eye using adeno-associated
viral vectors. Methods, 28(2):267–275, 2002.
Martin, K.R.G., Quigley, H.A., Zack, D.J., Levkovitch-Verbin, H., Kielczewski, J., Valenta, D.,
Baumrind, L., Pease, M.E., Klein, R.L., and Hauswirth, W.W. Gene therapy with brain-derived
neurotrophic factor as a protection: retinal ganglion cells in a rat glaucoma model. Investigative
Ophthalmology and Visual Science, 44(10):4357–4365, 2003.
Masland, R.H. Neuronal diversity in the retina. Current Opinion in Neurobiology, 11(4):431–436,
2001.
Massey, S.C. and Miller, R.F. Glutamate receptors of ganglion cells in the rabbit retina: evidence
for glutamate as a bipolar cell transmitter. Journal of Physiology, 405(1):635–655, 1988.
Maynard, E.M. Visual prostheses. Annual Review of Biomedical Engineering, 3(1):145–168, 2001.
Mazzoni, F., Novelli, E., and Strettoi, E. Retinal ganglion cells survive and maintain normal
dendritic morphology in a mouse model of inherited photoreceptor degeneration. Journal of
Neuroscience, 28(52):14282–14292, 2008.
McAdams, E.T., Lackermeier, A., McLaughlin, J.A., Macken, D., and Jossinet, J. The linear and
non-linear electrical properties of the electrode-electrolyte interface. Biosensors and Bioelec-
tronics, 10(1):67–74, 1995.
McKay, C.M. and Henshall, K.R. The perceptual effects of interphase gap duration in cochlear
implant stimulation. Hearing Research, 181:94–99, 2003.
Menon, G. and Walters, G. New paradigms in the treatment of wet AMD: the impact of anti-
VEGF therapy. Eye, 23:S1–S7, 2009.
Merrill, D.R., Bikson, M., and Jefferys, J.G.R. Electrical stimulation of excitable tissue: design
of efficacious and safe protocols. Journal of Neuroscience Methods, 141(2):171–198, 2005.
Mikelberg, F.S., Drance, S.M., Schulzer, M., Yidegiligne, H.M., and Weis, M.M. The normal
human optic nerve. Axon count and axon diameter distribution. Ophthalmology, 96(9):1325–
1328, 1989.
198
Miller, J.H. A short course in bacterial genetics: a laboratory manual and handbook for Escherichia
coli and related bacteria. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1992.
Minta, A., Kao, JP, and Tsien, R.Y. Fluorescent indicators for cytosolic calcium based on rho-
damine and fluorescein chromophores. Journal of Biological Chemistry, 264(14):8171–8178,
1989.
Miyawaki, A., Llopis, J., Heim, R., McCaffery, J.M., Adams, J.A., Ikurak, M., and Tsien, R.Y.
Fluorescent indicators for Ca
2+
based on green fluorescent proteins and calmodulin. Nature,
388:883–887, 1997.
Miyawaki, A., Griesbeck, O., Heim, R., and Tsien, R.Y. Dynamic and quantitative Ca
2+
mea-
surements using improved cameleons. Proceedings of the National Academy of Sciences, 96(5):
2135–2140, 1999.
Mo, X., Yokoyama, A., Oshitari, T., Negishi, H., Dezawa, M., Mizota, A., and Adachi-Usami,
E. Rescue of axotomized retinal ganglion cells by BDNF gene electroporation in adult rats.
Investigative Ophthalmology and Visual Science, 43(7):2401–2405, 2002.
Nagel, G., Szellas, T., Huhn, W., Kateriya, S., Adeishvili, N., Berthold, P., Ollig, D., Hegemann,
P., and Bamberg, E. Channelrhodopsin-2, a directly light-gated cation-selective membrane
channel. Proceedings of the National Academy of Sciences, 100(24):13940–13945, 2003.
Nakai, J., Ohkura, M., and Imoto, K. A high signal-to-noise Ca
2+
probe composed of a single
green fluorescent protein. Nature Biotechnology, 19(2):137–141, 2001.
Nanduri, D. Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stim-
ulation. PhD thesis, The University of Southern California, 2011.
Nanduri, D., Humayun, M.S., Greenberg, R.J., McMahon, M.J., and Weiland, J.D. Retinal
prosthesis phosphene shape analysis. In 30th Annual International IEEE EMBS Conference,
pages 1785–1788. IEEE, 2008.
Nanduri, D., Fine, I., Greenberg, R., Horsager, A., Boynton, G.M., and Weiland, J. Predicting
the percepts of electrical stimulation in retinal prosthesis subjects. In Cosyne 2011, 2011.
Nanduri, D., Fine, I., Horsager, A., Boynton, G.M., Humayun, M.S., Greenberg, R.J., and Wei-
land, J.D. Frequency and amplitude modulation have different effects on the percepts elicited
by retinal stimulation. Investigative Ophthalmology and Visual Science, 53(1):205–214, 2012.
Nasiatka, P., Ahuja, A.., Stiles, N.R.B., Hauer, M.C., Agrawal, R.N., Freda, R., Guven, D.,
Humayun, M.S., Weiland, J.D., and Tanguay Jr., A.R. Intraocular camera design for retinal
prostheses. Frontiers in Optics, Optical Society of America, 2005.
Newman, E.A. Glial cell inhibition of neurons by release of ATP. Journal of Neuroscience, 23(5):
1659–1666, 2003.
Nirenberg, S. and Pandarinath, C. Retinal prosthetic strategy with the capacity to restore normal
vision. Proceedings of the National Academy of Sciences, 109(37):15012–15017, 2012.
Niwa, H., Ken-ichi, Y., and Jun-ichi, M. Efficient selection for high-expression transfectants with
a novel eukaryotic vector. Gene, 108(2):193–199, 1991.
Normann, R.A., Greger, B.A., House, P., Romero, S.F., Pelayo, F., and Fernandez, E. Toward
the development of a cortically based visual neuroprosthesis. Journal of Neural Engineering,
6:035001, 2009.
199
O’Hearn, T.M., Sadda, S.R., Weiland, J.D., Maia, M., Margalit, E., and Humayun, M.S. Electrical
stimulation in normal and retinal degeneration (rd1) isolated mouse retina. Vision Research,
46(19):3198–3204, 2006.
Palanker, D., Vankov, A., Huie, P., and Baccus, S. Design of a high-resolution optoelectronic
retinal prosthesis. Journal of Neural Engineering, 2:S105–S120, 2005.
Pardhan, S. and Mahomed, I. The clinical characteristics of Asian and Caucasian patients on
Bradford’s Low Vision Register. Eye, 16(5):572–576, 2002.
Paredes, R.M., Etzler, J.C., Watts, L.T., Zheng, W., and Lechleiter, J.D. Chemical calcium
indicators. Methods, 46(3):143–151, 2008.
Parikh, N., Itti, L., and Weiland, J. Saliency-based image processing for retinal prostheses. Journal
of Neural Engineering, 7:016006, 2010.
Patil, S.D., Rhodes, D.G., and Burgess, D.J. DNA-based therapeutics and DNA delivery systems:
a comprehensive review. The AAPS Journal, 7(1):61–77, 2005.
Perez Fornos, A., Sommerhalder, J., and Pelizzone, M. Dynamics of visual perception upon
electrical stimulation of the retina. In ARVO 2010, 2010.
Perry, V.H. Evidence for an amacrine cell system in the ganglion cell layer of the rat retina.
Neuroscience, 6(5):931–944, 1981.
Peterka, D.S., Takahashi, H., and Yuste, R. Imaging voltage in neurons. Neuron, 69(1):9–21,
2011.
Petrossians, A., Whalen, J.J., Weiland, J.D., and Mansfeld, F. Surface modification of neural
stimulating/recording electrodes with high surface area platinum-iridium alloy coatings. In
33th Annual International IEEE EMBS Conference, pages 3001–3004. IEEE, 2011.
Petrs-Silva, H., Dinculescu, A., Li, Q., Deng, W.T., Pang, J., Min, S.H., Chiodo, V., Neeley,
A.W., Govindasamy, L., and Bennett, A. Novel properties of tyrosine-mutant AAV2 vectors in
the mouse retina. Molecular Therapy, 19(2):293–301, 2010.
Pillow, J.W., Paninski, L., Uzzell, V.J., Simoncelli, E.P., and Chichilnisky, E.J. Prediction and
decoding of retinal ganglion cell responses with a probabilistic spiking model. Journal of Neu-
roscience, 25(47):11003–11013, 2005.
Pittler, S.J., Keeler, C.E., Sidman, R.L., and Baehr, W. PCR analysis of DNA from 70-year-old
sections of rodless retina demonstrates identity with the mouse rd defect. Proceedings of the
National Academy of Sciences, 90(20):9616–9619, 1993.
Piyathaisere, D.V., Margalit, E., and Chen, S.J. Effects of short-term exposure to heat on the
retina. Investigative Ophthalmology and Visual Science, 42:S814, 2001.
Prado-Guitierrez, P., Fewster, L.M., Heasman, J.M., McKay, C.M., and Shepherd, R.K. Effect of
interphase gap and pulse duration on electrically evoked potentials is correlated with auditory
nerve survival. Hearing Research, 215(1-2):47–55, 2006.
Radtke, N.D., Aramant, R.B., Seiler, M.J., Petry, H.M., and Pidwell, D. Vision change after sheet
transplant of fetal retina with retinal pigment epithelium to a patient with retinitis pigmentosa.
Archives of Ophthalmology, 122(8):1159–1165, 2004.
200
Ranck, J.B. Which elements are excited in electrical stimulation of mammalian central nervous
system: a review. Brain Research, 98(3):417–440, 1975.
Randles, J.E.B. Kinetics of rapid electrode reactions. Discussions of the Faraday Society, 1:11–19,
1947.
Ray, A., Colodetti, L., Weiland, J.D., Hinton, D.R., Humayun, M.S., and Lee, E.J. Immunocyto-
chemical analysis of retinal neurons under electrical stimulation. Brain Research, 1255:89–97,
2009.
Repka, M.X. and Quigley, H.A. The effect of age on normal human optic nerve fiber number and
diameter. Ophthalmology, 96(1):26–32, 1989.
Retracing events. Nature Biotechnology, 25(9):949, 2007.
Richard, G., Hornig, R., Keser¨ u, M., and Feucht, M. Chronic epiretinal chip implant in blind
patients with retinitis pigmentosa: long-term clinical results. In ARVO 2007, 2007.
Richard, G., Keserue, M., Hornig, R., Post, N., and Zeitz, O. Long-term stability of stimulation
thresholds obtained from a human patient with a prototype of an epiretinal retina prosthesis.
In ARVO 2009, 2009.
Ridgway, E.B. and Ashley, C.C. Calcium transients in single muscle fibers. Biochemical and
Biophysical Research Communications, 29(2):229–234, 1967.
Rizzo, J.F., Wyatt, J., Loewenstein, J., Kelly, S., and Shire, D. Methods and perceptual thresholds
for short-term electrical stimulation of human retina with microelectrode arrays. Investigative
Ophthalmology and Visual Science, 44(12):5355–5361, 2003a.
Rizzo, J.F., Wyatt, J., Loewenstein, J., Kelly, S., and Shire, D. Perceptual efficacy of electrical
stimulation of human retina with a microelectrode array during short-term surgical trials.
Investigative Ophthalmology and Visual Science, 44(12):5362–5369, 2003b.
Rizzo, J.F., Snebold, L., and Kenney, M. Development of a Visual Prosthesis. Visual Prosthesis
and Ophthalmic Devices: New Hope in Sight. Springer, 2007.
Rizzo, J.F., Shire, D.B., Kelly, S.K., Troyk, P., Gingerich, M., McKee, B., Priplata, A., Chen, J.,
Drohan, W., Doyle, P., Mendoza, O., Theogarajan, L., Cogan, S., and Wyatt, J.L. Development
of the Boston Retinal Prosthesis. In 33th Annual International IEEE EMBS Conference, pages
3135–3138. IEEE, 2011a.
Rizzo, J.F., Shire, D.B., Kelly, S.K., Troyk, P., Gingerich, M., McKee, B., Priplata, A., Chen, J.,
Drohan, W., Doyle, P., Mendoza, O., Theogarajan, L., Cogan, S., and Wyatt, J.L. Overview
of the Boston Retinal Prosthesis: challenges and opportunities to restore useful vision to the
blind. In 33th Annual International IEEE EMBS Conference, pages 7492–7495. IEEE, 2011b.
Roessler, G., Laube, T., Brockmann, C., Kirschkamp, T., Mazinani, B., Goertz, M., Koch, C.,
Krisch, I., Sellhaus, B., and Trieu, H.K. Implantation and explantation of a wireless epiretinal
retina implant device: observations during the EPIRET3 prospective clinical trial. Investigative
Ophthalmology and Visual Science, 50(6):3003, 2009.
Roizenblatt, R., Weiland, J.D., Carcieri, S., Qiu, G., Behrend, M., Humayun, M.S., and Chow,
R.H. Nanobiolistic delivery of indicators to the living mouse retina. Journal of Neuroscience
Methods, 153(1):154–161, 2006.
201
Rose, T.L. and Robblee, L.S. Electrical stimulation with Pt electrodes. VIII. electrochemically
safe charge injection limits with 0.2 ms pulses. IEEE Transactions on Biomedical Engineering,
37(11):1118–1120, 1990.
Sahel, J.A., da Cruz, L., Hafezi, F., Stanga, P., Merlini, F., Coley, B., and Greenberg, R. Subjects
blind from outer retinal dystrophies are able to consistently read short sentences using the Argus
II Retinal Prosthesis System. In ARVO 2011, 2011.
Salinas-Navarro, M., Mayor-Torroglosa, S., Jimenez-Lopez, M., Avil´ es-Trigueros, M., Holmes,
T.M., Lund, R.D., Villegas-P´ erez, M.P., and Vidal-Sanz, M. A computerized analysis of the
entire retinal ganglion cell population and its spatial distribution in adult rats. Vision Research,
49(1):115–126, 2009.
Santos, A., Humayun, M.S., de Juan Jr., E., Greenburg, R.J., Marsh, M.J., Klock, I.B., and
Milam, A.H. Preservation of the inner retina in retinitis pigmentosa: a morphometric analysis.
Archives of Ophthalmology, 115(4):511–515, 1997.
Schejter, A., Tsur, L., Farah, N., Reutsky-Gefen, I., Falick, Y., and Shoham, S. Cellular resolution
panretinal imaging of optogenetic probes using a simple funduscope. Translational Vision
Science and Technology, 1(2):1–8, 2012.
Schiefer, M.A. and Grill, W.M. Sites of neuronal excitation by epiretinal electrical stimulation.
IEEE Transactions on Neural Systems and Rehabilitation Engineering, 14(1):5–13, 2006.
Schmidt, E.M., Bak, M.J., Hambrecht, F.T., Kufta, C.V., O’rourke, D.K., and Vallabhanath, P.
Feasibility of a visual prosthesis for the blind based on intracortical micro stimulation of the
visual cortex. Brain, 119(2):507–522, 1996.
Schmidt, M., Voutetakis, A., Afione, S., Zheng, C., Mandikian, D., and Chiorini, J.A. Adeno-
associated virus type 12 (AAV12): a novel AAV serotype with sialic acid-and heparan sulfate
proteoglycan-independent transduction activity. Journal of Virology, 82(3):1399–1406, 2008.
Segev, R., Goodhouse, J., Puchalla, J., and Berry, M.J. Recording spikes from a large fraction of
the ganglion cells in a retinal patch. Nature Neuroscience, 7(10):1155–1162, 2004.
Segev, R., Puchalla, J., and Berry, M.J. Functional organization of ganglion cells in the salamander
retina. Journal of Neurophysiology, 95(4):2277–2292, 2006.
Sekirnjak, C., Hottowy, P., Sher, A., Dabrowski, W., Litke, AM, and Chichilnisky, EJ. Elec-
trical stimulation of mammalian retinal ganglion cells with multielectrode arrays. Journal of
Neurophysiology, 95(6):3311–3327, 2006.
Sekirnjak, C., Hottowy, P., Sher, A., Dabrowski, W., Litke, A.M., and Chichilnisky, E.J. High-
resolution electrical stimulation of primate retina for epiretinal implant design. Journal of
Neuroscience, 28(17):4446–4456, 2008.
Sekirnjak, C., Hulse, C., Jepson, L.H., Hottowy, P., Sher, A., Dabrowski, W., Litke, A.M., and
Chichilnisky, EJ. Loss of responses to visual but not electrical stimulation in ganglion cells of
rats with severe photoreceptor degeneration. Journal of Neurophysiology, 102(6):3260, 2009.
Sekirnjak, C., Jepson, L.H., Hottowy, P., Sher, A., Dabrowski, W., Litke, A.M., and Chichilnisky,
E.J. Changes in physiological properties of rat ganglion cells during retinal degeneration.
Journal of Neurophysiology, 105(5):2560–2571, 2011.
202
Shah, H.A., Montezuma, S.R., and Rizzo, J.F. In vivo electrical stimulation of rabbit retina:
effect of stimulus duration and electrical field orientation. Experimental Eye Research, 83(2):
247–254, 2006.
Shah, S., Hines, A., Zhou, D., Greenberg, R.J., Humayun, M.S., and Weiland, J.D. Electrical
properties of retinal-electrode interface. Journal of Neural Engineering, 4(1):S24–S29, 2007.
Shepherd, R.K. and Javel, E. Electrical stimulation of the auditory nerve: II. Effect of stimulus
waveshape on single fibre response properties. Hearing research, 130(1):171–188, 1999.
Shimazu, K., Miyake, Y., and Watanabe, S. Retinal ganglion cell response properties in the
transcorneal electrically evoked response of the visual system. Vision Research, 39(13):2251–
2260, 1999.
Shimomura, O., Johnson, F.H., and Saiga, Y. Extraction, purification and properties of aequorin,
a bioluminescent protein from the luminous hydromedusan, Aequorea. Journal of cellular and
comparative physiology, 59(3):223–239, 1962.
Simon, D.K. and O’Leary, D.D.M. Relationship of retinotopic ordering of axons in the optic
pathway to the formation of visual maps in central targets. Journal of Comparative Neurology,
307(3):393–404, 1991.
Smetters, D., Majewska, A., and Yuste, R. Detecting action potentials in neuronal populations
with calcium imaging. Methods, 18(2):215–221, 1999.
Stanga, P.E., Coley, B., Merlini, F., Biswas, S., Turner, G., and Greenberg, R. Initial word
recognition results with the Argus II Retinal Prosthesis. In ARVO 2010, 2010.
Stanga, P.E., Hafezi, F., Sahel, J.A., da Cruz, L., Merlini, F., Coley, B., and Greenberg, R.
Patients blinded by outer retinal dystrophies are able to perceive color using the Argus II
Retinal Prosthesis System. In ARVO 2011, 2011.
Stein, L., Roy, K., Lei, L., and Kaushal, S. Clinical gene therapy for the treatment of RPE65-
associated Leber congenital amaurosis. Expert Opinion on Biological Therapy, 11(3):429–439,
2011.
Stein, R.B., Charles, D., Gordon, T., Hoffer, J.A., and Jhamandas, J. Impedance properties of
metal electrodes for chronic recording from mammalian nerves. IEEE Transactions on Biomed-
ical Engineering, 25(6):532–537, 1978.
Steinberg, R.H., Flannery, J.G., Naash, M., Oh, P., Matthes, M.T., Yasumura, D., Lau-Villacorta,
C., Chen, J., and LaVail, M.M. Transgenic rat models of inherited retinal degeneration caused
by mutant opsin genes. Investigative Ophthalmology and Visual Science, 37(3):S698, 1996.
Stett, A., Barth, W., Weiss, S., Haemmerle, H., and Zrenner, E. Electrical multisite stimulation
of the isolated chicken retina. Vision Research, 40(13):1785–1795, 2000.
Stiles, N.R., Hauer, M.C., Lee, P., Nasiatka, P.J., Lue, J.C., Weiland, J.D., Humayun, M.S., and
Tanguay Jr., A.R. Intraocular camera for retinal prostheses: design constraints based on visual
psychophysics. Frontiers In Optics, Optical Society of America, 2007.
Stingl, K., Bartz-Schmidt, K.U., Besch, D., Braun, A., Bruckmann, A., Gekeler, F., Greppmaier,
U., Hipp, S., H¨ ortd¨ orfer, G., Kernstock, C., Koitschev, A., Kusnyerik, A., Sachs, H., Schatz, A.,
Stingl, K.T., Peters, T., Wilhelm, B., and Zrenner, E. Artificial vision with wirelessly powered
subretinal electronic implant alpha-IMS. Proceedings of the Royal Society B, 280:20130077,
2013.
203
Stone, E.M. Leber congenital amaurosis-a model for efficient genetic testing of heterogeneous
disorders: LXIV Edward Jackson Memorial Lecture. American Journal of Ophthalmology, 144
(6):791–811, 2007.
Suzuki, S., Humayun, M.S., Weiland, J.D., Chen, S.J., Margalit, E., Piyathaisere, D.V., and
de Juan Jr., E. Comparison of electrical stimulation thresholds in normal and retinal degener-
ated mouse retina. Japanese Journal of Ophthalmology, 48(4):345–349, 2004.
Takahashi, A., Camacho, P., Lechleiter, J.D., and Herman, B. Measurement of intracellular
calcium. Physiological Reviews, 79(4):1089–1125, 1999.
Tan, J.S.L., Wang, J.J., Flood, V., Rochtchina, E., Smith, W., and Mitchell, P. Dietary antiox-
idants and the long-term incidence of age-related macular degeneration: the Blue Mountains
Eye Study. Ophthalmology, 115(2):334–341, 2008.
Tehovnik, E.J., Tolias, A.S., Sultan, F., Slocum, W.M., and Logothetis, N.K. Direct and indirect
activation of cortical neurons by electrical microstimulation. Journal of Neurophysiology, 96
(2):512–521, 2006.
Thoreson, W.B. Kinetics of synaptic transmission at ribbon synapses of rods and cones. Molecular
Neurobiology, 36(3):205–223, 2007.
Tian, L., Hires, S.A., Mao, T., Huber, D., Chiappe, M.E., Chalasani, S.H., Petreanu, L., Aker-
boom, J., McKinney, S.A., and Schreiter, E.R. Imaging neural activity in worms, flies and mice
with improved GCaMP calcium indicators. Nature Methods, 6(12):875–881, 2009.
Tong, C., Huang, G., Ashton, C., Li, P., and Ying, Q.L. Generating gene knockout rats by
homologous recombination in embryonic stem cells. Nature Protocols, 6(6):827–844, 2011.
Townes-Anderson, E., Colantonio, A., and St. Jules, R. S. Age-related changes in the tiger
salamander retina. Experimental Eye Research, 66(5):653–667, 1998.
Troyk, P., Bak, M., Berg, J., Bradley, D., Cogan, S., Erickson, R., Kufta, C., McCreery, D.,
Schmidt, E., and Towle, V. A model for intracortical visual prosthesis research. Artificial
Organs, 27(11):1005–1015, 2003.
Tsai, D., Morley, J.W., Suaning, G.J., and Lovell, N.H. Frequency-dependent reduction of voltage-
gated sodium current modulates retinal ganglion cell response rate to electrical stimulation.
Journal of Neural Engineering, 8:066007, 2011.
Tsien, R.Y. New calcium indicators and buffers with high selectivity against magnesium and
protons: design, synthesis, and properties of prototype structures. Biochemistry, 19(11):2396–
2404, 1980.
Tunc, M., Humayun, M., Cheng, X., and Ratner, B.D. A reversible thermosensitive adhesive
for retinal implants: in vivo experience with plasma-deposited poly(N-isopropyl acrylamide).
Retina, 28(9):1338–1343, 2008.
van den Honert, C. and Mortimer, J.T. The response of the myelinated nerve fiber to short
duration biphasic stimulating currents. Annals of Biomedical Engineering, 7(2):117–125, 1979.
Venkatraman, S., Hendricks, J., King, Z.A., Sereno, A.J., Richardson-Burns, S., Martin, D., and
Carmena, J.M. In vitro and in vivo evaluation of PEDOT microelectrodes for neural stimulation
and recording. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 19(3):
307–316, 2011.
204
Veraart, C., Raftopoulos, C., Mortimer, J.T., Delbeke, J., Pins, D., Michaux, G., Vanlierde, A.,
Parrini, S., and Wanet-Defalque, M.C. Visual sensations produced by optic nerve stimulation
using an implanted self-sizing spiral cuff electrode. Brain Research, 813(1):181–186, 1998.
Veraart, C., Wanet Defalque, M.C., G´ erard, B., Vanlierde, A., and Delbeke, J. Pattern recognition
with the optic nerve visual prosthesis. Artificial Organs , 27(11):996–1004, 2003.
Walker, J.S. Neuropsychological assessment. Wiley Encyclopedia of Forensic Science, 2009.
Weiland, J.D., Humayun, M.S., Dagnelie, G., de Juan Jr., E., Greenberg, R.J., and Iliff, N.T.
Understanding the origin of visual percepts elicited by electrical stimulation of the human retina.
Graefe’s Archive for Clinical and Experimental Ophthalmology, 237(12):1007–1013, 1999.
Weiland, J.D., Liu, W., and Humayun, M.S. Retinal prosthesis. Annu. Rev. Biomed. Eng., 7:
361–401, 2005.
Weitz, A.C. and Weiland, J.D. Visual Prostheses. In Neural Computation, Neural Devices, and
Neural Prostheses. Springer, New York, in press.
Weitz, A.C., Behrend, M.R., Humayun, M.S., Chow, R.H., and Weiland, J.D. Interphase gap
decreases electrical stimulation threshold of retinal ganglion cells. In 33th Annual International
IEEE EMBS Conference, pages 6725–6728. IEEE, 2011.
Weitz, A.C., Behrend, M.R., Lee, N.S., Klein, R.L., Chiodo, V.A., Hauswirth, W.W., Humayun,
M.S., Weiland, J.D., and Chow, R.H. Imaging the response of the retina to electrical stimulation
with genetically encoded calcium indicators. Journal of Neurophysiology, in press.
Wilson, B.S., Finley, C.C., Lawson, D.T., Wolford, R.D., and Zerbi, M. Design and evaluation of
a continuous interleaved sampling (CIS) processing strategy for multichannel cochlear implants.
Journal of Rehabilitation Research and Development, 30:110–116, 1993.
Winter, J.O., Cogan, S.F., and Rizzo, J.F. Retinal prostheses: current challenges and future
outlook. Journal of Biomaterials Science, Polymer Edition, 18(8):1031–1055, 2007.
Wise, K.D., Angell, J.B., and Starr, A. An integrated-circuit approach to extracellular microelec-
trodes. IEEE Transactions on Biomedical Engineering, 17(3):238–247, 1970.
Wong, R.O.L. Calcium imaging and multielectrode recordings of global patterns of activity in the
developing nervous system. The Histochemical Journal, 30(3):217–229, 1998.
Wong, R.O.L. and Oakley, D.M. Changing patterns of spontaneous bursting activity of on and
off retinal ganglion cells during development. Neuron, 16(6):1087–1095, 1996.
W¨ assle, H. Parallel processing in the mammalian retina. Nature Reviews Neuroscience, 5(10):
747–757, 2004.
Wu, C.C., Reilly, J.F., Young, W.G., Morrison, J.H., and Bloom, F.E. High-throughput morpho-
metric analysis of individual neurons. Cerebral Cortex, 14(5):543–554, 2004.
Wu, K., Klein, R.L., Meyers, C.A., King, M.A., Hughes, J.A., Millard, W.J., and Meyer, E.M.
Long-term neuronal effects and disposition of ectopic preproNGF gene transfer into the rat
septum. Human Gene Therapy, 14(15):1463–1472, 2003.
Yanai, D., Weiland, J.D., Mahadevappa, M., Greenberg, R.J., Fine, I., and Humayun, M.S. Visual
performance using a retinal prosthesis in three subjects with retinitis pigmentosa. American
Journal of Ophthalmology, 143(5):820–827, 2007.
205
Yawo, H. and Kuno, M. Calcium dependence of membrane sealing at the cut end of the cockroach
giant axon. Journal of Neuroscience, 5(6):1626–1632, 1985.
Yu, J., Daniels, B.A., and Baldridge, W.H. Slow excitation of cultured rat retinal ganglion cells
by activating group I metabotropic glutamate receptors. Journal of Neurophysiology, 102(6):
3728–3739, 2009.
Yuste, R. and Katz, L.C. Control of postsynaptic Ca
2+
influx in developing neocortex by excita-
tory and inhibitory neurotransmitters. Neuron, 6(3):333–344, 1991.
Zariwala, H.A., Borghuis, B.G., Hoogland, T.M., Madisen, L., Tian, L., De Zeeuw, C.I., Zeng,
H., Looger, L.L., Svoboda, K., and Chen, T.W. A Cre-dependent GCaMP3 reporter mouse for
neuronal imaging in vivo. Journal of Neuroscience, 32(9):3131–3141, 2012.
Zhao, Y., Araki, S., Wu, J., Teramoto, T., Chang, Y.F., Nakano, M., Abdelfattah, A.S., Fujiwara,
M., Ishihara, T., and Nagai, T. An expanded palette of genetically encoded Ca
2+
indicators.
Science, 333(6051):1888–1891, 2011.
Zhou, C., Tao, C., Chai, X., Sun, Y., and Ren, Q. Implantable imaging system for visual pros-
thesis. Artificial Organs , 34(6):518–522, 2010.
Zolotukhin, S., Byrne, B.J., Mason, E., Zolotukhin, I., Potter, M., Chesnut, K., Summerford, C.,
Samulski, R.J., and Muzyczka, N. Recombinant adeno-associated virus purification using novel
methods improves infectious titer and yield. Gene Therapy, 6(6):973–985, 1999.
Zrenner, E., Bartz-Schmidt, K.U., Benav, H., Besch, D., Bruckmann, A., Gabel, V.P., Gekeler, F.,
Greppmaier, U., Harscher, A., and Kibbel, S. Subretinal electronic chips allow blind patients
to read letters and combine them to words. Proceedings of the Royal Society B, 278(1711):
1489–1497, 2011.
206

Abstract (if available)

Abstract Epiretinal implants for the blind are designed to stimulate surviving retinal neurons, thus bypassing the diseased photoreceptor layer. Single-unit or multielectrode recordings from isolated animal retina are commonly used to inform the design of these implants. However, such electrical recordings provide limited information about the spatial patterns of retinal activation. Calcium imaging overcomes this limitation, as imaging enables high spatial resolution mapping of retinal ganglion cell (RGC) activity as well as simultaneous recording from hundreds of RGCs. ❧ I developed a method for labeling the majority of ganglion cells in adult rat retina with genetically encoded calcium indicators (GECIs). Intravitreal injection of an adeno-associated viral vector targeted roughly 85% of ganglion cells with high specificity. Due to the large fluorescence signals provided by GECIs, I was able to visualize the retina's response to electrical stimulation in real time. Imaging transduced retinas mounted on multielectrode arrays revealed how varying stimulus parameters dramatically affects the spatial extent of RGC activation. ❧ Recent human subject testing indicated that patients often see large, elongated phosphenes due to stimulation of RGC axon bundles. My experiments revealed two potential strategies for avoiding stimulation of axons: Short pulses (≤ 0.1 ms) selectively stimulate ganglion cell somata at thresholds 40–60% lower than their axons, while long pulses (≥ 25 ms) preferentially target bipolar cells. No axonal activation was observed during stimulation with long pulses, regardless of amplitude. Responses were focal and gradually became larger as stimulus amplitude was increased. ❧ Low-frequency sinusoids can be used as an alternative to long pulses, as they require significantly less charge to evoke responses. When stimulating with short pulses, interphase gaps can be used to reduce overall thresholds by 10–25%. Retinal degeneration appears to have no effect on response shape

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Manipulation of RGCs response using different stimulation strategies for retinal prosthesis

PDF

Stimulation strategies to improve efficiency and temporal resolution of epiretinal prostheses

PDF

Understanding the degenerate retina's response to electrical stimulation: an in vitro approach

PDF

Viewing the picture we paint

PDF

Cortical and subcortical responses to electrical stimulation of rat retina

PDF

Towards a high resolution retinal implant

PDF

Synaptic second wind: phosphene fading of the Argus II and glutamate secretion in the inner retina

PDF

Electrical stimulation of degenerate retina

PDF

Dependence of rabbit retinal synchrony on visual stimulation parameters

PDF

Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stimulation

PDF

Investigation of the electrode-tissue interface of retinal prostheses

PDF

PET study of retinal prosthesis functionality

PDF

Encoding of natural images by retinal ganglion cells

PDF

Oxygen therapy for the treatment of retinal ischemia

PDF

Neuromorphic motion sensing circuits in a silicon retina

PDF

Ultrasound neuromodulation and its applications for noninvasive vision restoration

PDF

On the electrophysiology of multielectrode recordings of the basal ganglia and thalamus to improve DBS therapy for children with secondary dystonia

PDF

Adaptive event-driven simulation strategies for accurate and high performance retinal simulation

PDF

Therapeutic electrical stimulation strategies for neuroregeneration and neuroprotection of retinal neurons

PDF

Electronics design and in vivo evaluation of a wirelessly rechargeable fetal micropacemaker

Asset Metadata

Creator Weitz, Andrew C. (author)

Core Title Improving stimulation strategies for epiretinal prostheses

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 12/06/2012

Defense Date 03/05/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag calcium imaging,electrophysiology,multielectrode arrays,OAI-PMH Harvest,retina,retinal ganglion cells,retinal prostheses

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Weiland, James D. (committee chair), Chow, Robert H. (committee member), Humayun, Mark S. (committee member), Sampath, Alapakkam P. (committee member)

Creator Email andrew.weitz@gmail.com,aweitz@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-246023

Unique identifier UC11288410

Identifier etd-WeitzAndre-1600.pdf (filename),usctheses-c3-246023 (legacy record id)

Legacy Identifier etd-WeitzAndre-1600.pdf

Dmrecord 246023

Document Type Dissertation

Rights Weitz, Andrew C.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA