Download Jonathan A. N. Fisher

LINEAR AND NON-LINEAR FLUORESCENCE
IMAGING OF NEURONAL ACTIVITY
Jonathan A. N. Fisher
A Dissertation
in
Physics and Astronomy
Presented to the Faculties of the University of Pennsylvania in Partial
Fulfillment of the Requirements for the Degree of Doctor of Philosophy
2007
Arjun G. Yodh
Supervisor of Dissertation
Ravi K. Sheth
Graduate Group Chairperson
c Copyright 2007
°
by
Jonathan A. N. Fisher
Dedication
To Shirley, Mom, Dad, and Hana
iv
Acknowledgements
My graduate work at Penn has spanned an extremely dynamic period in my life, both
scientifically and personally. There are several individuals who I am particularly indebted
to for guidance and mentorship. First and foremost, I must acknowledge my mentor, Arjun
Yodh. I have learned an amazing amount about science, professionalism, and emotional
maturity from Arjun. Since the beginning of our work together, when he lent me his
copy of Francis Crick’s The Astonishing Hypothesis and said “Let’s learn about the brain
together,” it has truly been an exciting and fruitful Odyssey.
One of the best parts about working in Arjun’s lab has been my wonderful fellow labmates. Alper Corlu and I have shared countless laughs and Turkish internet videos (most
notably the Cola Turka ad campaign, amazingly featuring Chevy Chase). I thank Alper for
his friendship (and the fez he brought back for me from Istanbul), and wish him and his
wife Canan the best in Pittsburgh. It should be noted that these sentiments apply outside
of Pittsburgh, as well. Regine Choe is one of the kindest, most giving, and thoughtful
people I’ve met. Regine has always been there to listen and talk through things. She also
happens to be a leading scientist in her field, a licensed pilot, a figure skater, a terrific chef,
and a purveyor of fine dining in Philadelphia. I actually first met Turgut Durduran when I
was still an undergrad. It was at a party thrown by our mutual friend Naoki Mihara. I was
publishing the first issue of the Journal of Natural Philosopy, and Naoki said that Turgut
should write an article because he was an amazing scientist. Years later, Turgut is that
much more amazing as a scientist. It was a nice surprise to find that Turgut worked in the
v
lab that I eventually joined at Penn. We have shared numerous crazy conference experiences, and at least two intoxicated nights stumbling down Berlin’s Oranienburger Straße.
Kijoon Lee has been an excellent teacher of nonlinear optics techniques and theory, among
other things. He is additionally a great friend, and I wish him the best in his new position
in Singapore. Soren Konecky and I have shared several amusing experiences attempting
bootstrap science. I will never forget the feeling of complete and utter frustration we experienced when trying to resuscitate the plumbing of an old DCR pulsed Nd:YAG laser.
All in pursuit of some preliminary data when writing a grant proposal for a novel optical
“fez” device (data not shown). In addition to scientific help and advice, I thank my labmates Guoqiang Yu and Chao Zhou for driving me to and from San Francisco during a
conference at San Jose. I will never forget those leftover salt-baked shrimp you saved for
me for the ride back to San Jose. I enjoyed conversations both scientific and music-related
with Hsing-Wen Wang, and wish her the best in her scientific career in Taiwan, or wherever she may finally settle. I also thank David Busch, Erin Buckley, Xiaoman Xing, Meeri
Kim, and Monika Grosicka-Koptyra for their friendship and support.
I am also thankful to members of the “other side” of Arjun’s group, the soft condensed
matter group. Daniel Chen has been a great scientific colleague and artistic accomplice.
Our collaborations have included a theatrical adaptation of Bowie’s Lady Stardust, and a
composition for the Philadelphia Fringe Festival. I am additionally grateful to my “soft
matter” colleagues Mateusz Bryning, Ahmed Alsayed, Zvonimir Dogic, and Mohammad
Islam.
vi
My research experience at Penn actually began in the experimental cosmology laboratory of Mark Devlin, when I was an undergrad. Mark provided me with an incomparable
background in experimental technique and showed me how to make something and have
it actually work. I will never forget the late nights we all spent in the lab (together with
Bernard Jones, Hans Eberhart, Jonathan Bruzzi, Harvey Chapman, Simon Dicker, and
others) trying to make those things work.
It was in Mark’s lab that I met Jeff Klein, who has since become a great friend and
scientific mentor. Jeff is one of those rare individuals who actually knows everything, and
can explain it all from first principles. Additionally, he is a gourmet chef and athlete probably as close to the Greek ideal as one will find among the mortals. The fact that he
wears no coat at any point in the cold seasons suggests, however, that he may in fact be
immortal. We have shared many stimulating conversations at the intersection of physics
and philosophy (very near the intersection of 33rd and Walnut), typically over coffee after experimenting with yet another new Philadelphia BYO (or, if with BJ, likely vegan
Chinese or other intestinal flora-friendly establishment).
I have enjoyed the guidance and friendship with other faculty members of the department of Physics and Astronomy, most notably Randy Kamien, Bhuvnesh Jain, and Mark
Goulian. In addition to Randy Kamien and Mark Goulian, I consider Paul Soven, Phil
Nelson, Nigel Lockyer and Fay Selove to be my most significant teachers of physical
theory.
I am indebted to members of the physics department staff for helping me parse through
some of the University’s bureaucratic mechanics along the way. Most notably, Audrey,
vii
Angela, Tom, and Jackie. I must thank, as well, the members of the physics machine shop
Tim and Buddy. I feel like I have known them forever (nearly 10 years at this point). I am
thankful for their friendship and guidance in learning the art of machining.
Thus far I have not mentioned my neuroscience collaborators. Ultimately, my experiments required that much of my time was spent in Penn’s neuroscience department. Brian
Salzberg has been a terrific mentor throughout the course of our experiments. It has truly
been fun to work with him through several projects, especially during our exciting recent
work on two-photon imaging of action potentials in the pituitary preparation. I look forward to more occasions to discuss science (and non-science) over beers. Especially if they
occur at Brian’s place. I must thank, additionally, Ana Lia Obaid and the members of
Brian’s lab including Gi-Ho Kim and Paul Kosterin.
And then there is Diego Contreras. Research and friendship with Diego cannot be
summed up with any one neat adjective. It is more like living a Gabriel Garcia Marquezinspired story. Diego’s mind contains universes within universes of thought. I will always
remember his assertion that scientists and musicians pursue their work with the same
driving passion. While I think that this is not necessarily true in general, for Diego it
most definitely is. I thank the members of Diego’s lab for welcoming me into their lab
and making late nights actually sort of fun. Eugene Civillico, Ashlan Reid, Cristin Welle,
Noah Roy, Jason Wester, Jessica Cardin and Mike Higgley, Larry Palmer, Esther Garcia
de Yebenes and Bryan Willent have all made me feel like family.
I must thank Leif Finkel for being a great PI on the Packard-funded projects, and for
also helping me get a bit of press for both science and music. I have enjoyed working with
viii
the next generation Jonathan, Jonathan Barchi, which resolves some of the bugs found in
the previous Jonathan. I am confident he will lead the Starship Packard to new heights and
frontiers.
I thank Mike Therien and his lab for the great collaborative experience measuring
two-photon absorption properties of an infinite number of novel compounds. In particular,
it has been a pleasure to work with Kimihiro Susumu. I wish him the best in lonely
Washington D.C. I additionally am grateful to Peter Ghoroghchian for being the one to
help get me started on this project.
My time in Philadelphia would not have been so pleasant if not for the community representing my piano life. I thank Bob Durso, Brenna, Bjorn, Victoria, Jeff, and Aaron B. in
particular for their friendship. I also thank my New York friends (Quinn, Adam, Thomas,
Brian, to mention a few) for not giving up on me while I pursued voltage-sensitive dye
signals in Philadelphia. I look forward to my return to The City.
I would have achieved none of this work without the loving support of my family.
I thank Mom, Dad, and Hana for really being the best. I think it was my mother who
first stimulated my interest in science when she showed me the “chemical reaction” that
happens when you mix baking soda and vinegar. To this day, I am amazed at the variety
of engineering applications I found that usefully employed that same reaction. I thank my
parents for then funding my purchase of Edmund Scientific’s “labware bonanza” kit a year
or two later. As I begin my postdoctoral research position, I am pleased to say that I think
the period of family-funded research may be nearing an end. I also must thank Grandma,
Grandpa, Savta, Saba, the extended Fisher and Nastasi familes, and the Zilberstein family
ix
for all their support.
Lastly, my wife Shirley has seen me through all of this work. Our relationship began
just before I started graduate school, and we were married just under a year before I
finished. We have watched each other grow up during these years. Whether in New York,
Philadelphia, Washington D.C., San Francisco, Rome, Florence, Paris, Berlin, Belgium,
Amsterdam, Madrid, Barcelona, the experiences we have shared over the past seven years
have been the best of my life. Because each year with Shirley is better than the year before,
I am excited to spend the rest of our lives together.
J.A.N. Fisher, Philadelphia, PA, 2007
x
Abstract
Linear and non-linear fluorescence imaging of neuronal
activity
Jonathan A. N. Fisher
Arjun G. Yodh
Version : July 15, 2007
Optical imaging of neuronal activity offers new possibilities for understanding brain
physiology. The predominant methods in neuroscience for measuring electrical activity
require electrodes inserted into the tissue. Such methods, however, provide limited spatial
information and are invasive. Optical methods are less physically invasive and offer the
possibility for simultaneously imaging the activity of many neurons. In this thesis oneand two-photon fluorescence microscopy techniques were applied to several in vivo and
in vitro mammalian preparations.
Using one-photon absorption fluorescence microscopy and gradient index (GRIN) lens
optics, cortical electrical activity in response to electric stimulation was resolved in threedimensions at high-speed in the primary somatosensory cortex of the mouse in vivo using
voltage-sensitive dyes. Imaging at depths up to 150 µm below the cortex surface, it was
possible to resolve depth-dependent patterns of neuronal activity in response to cortical
and thalamic electric stimulation. The patterns of activity were consistent with known
xi
cortical cellular architecture.
In a qualitatively different set of experiments, one-photon fluorescence microscopy via
voltage-sensitive dyes was successfully employed to image an in vitro preparation of the
perfused rat brainstem during the process of respiratory rhythmogenesis. Imaging results
yielded insights into the spatial organization of the central respiratory rhythm generation
region in the ventrolateral medulla.
A multifocal two-photon scanning microscope was constructed, and design and operation principles are described. Utilizing the novel device, anatomical and functional
two-photon imaging via potentiometric dyes and calcium dyes is described, and the results of in vivo versus in vitro imaging are compared. Anatomical imaging results used
either functional probe background fluorescence or green fluorescent protein (GFP) expression. Spectroscopic experiments measuring the two-photon absorption (TPA) cross
sections, Σ2 of various fluorophores are described as well.
Utilizing single-beam two-photon microscopy, action potentials were recorded optically from individual (∼1 µm) nerve terminals of the intact mouse neurohypophysis, in
a single sweep. Single-trial recordings of action potentials exhibited signal-to-noise ratios ∼5 and fractional fluorescence changes of up to ∼10%. These results represent the
first single-trial optical recording of action potentials from individual nerve terminals in
an intact mammalian preparation using 180 ◦ detection, and may serve as an alternative to
invasive electrode arrays for studying neuronal systems in vivo.
xii
Contents
Dedication
iv
Acknowledgements
v
Abstract
xi
List of Tables
xx
List of Figures
xxv
1 Introduction
1
1.1
Optical Imaging of Brain Function . . . . . . . . . . . . . . . . . . . . .
1
1.2
Background theory: One- and two-photon absorption . . . . . . . . . . .
3
1.2.1
One-photon absorption . . . . . . . . . . . . . . . . . . . . . . .
3
1.2.2
Two-photon absorption . . . . . . . . . . . . . . . . . . . . . . .
5
Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
1.3
2 Imaging Systems
2.1
12
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xiii
12
2.2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2.2.1
Illumination sources for one-photon fluorescence microscopy . .
13
2.2.2
Illumination sources for two-photon fluorescence microscopy . .
18
2.2.3
Detectors: CCDs and EMCCDs . . . . . . . . . . . . . . . . . .
22
2.3
Fiber-Coupled Epi-Illumination Imaging System . . . . . . . . . . . . .
23
2.4
Fiber-Coupled Gradient Index (GRIN) Lens Fluorescence Endoscope . .
25
2.4.1
Gradient Index (GRIN) lenses . . . . . . . . . . . . . . . . . . .
26
2.4.2
Optical configuration of imaging device . . . . . . . . . . . . . .
29
Multifocal Multiphoton Microscope . . . . . . . . . . . . . . . . . . . .
31
2.5.1
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
2.5.2
Overview of system optics . . . . . . . . . . . . . . . . . . . . .
33
2.5.3
Galvanometer scanning mirrors . . . . . . . . . . . . . . . . . .
37
2.5.4
Multifocal scanning strategies . . . . . . . . . . . . . . . . . . .
38
2.5.5
Device control and data acquisition . . . . . . . . . . . . . . . .
43
2.5.6
Advantages and challenges in multifocal microscopy techniques .
49
Appendix: Computational optical sectioning and deconvolution techniques
51
2.6.1
Background theory . . . . . . . . . . . . . . . . . . . . . . . . .
52
2.6.2
Nearest-neighbor approximation . . . . . . . . . . . . . . . . . .
56
2.6.3
Example: Nearest-neighbor deblurring of VSD signals in mouse
2.5
2.6
2.7
Components
barrel cortex in vivo . . . . . . . . . . . . . . . . . . . . . . . .
58
Appendix B: Alignment procedure for multifocal two-photon microscope
60
2.7.1
62
General alignment of excitation laser beam throughout microscope
xiv
2.7.2
General imaging protocol . . . . . . . . . . . . . . . . . . . . .
3 Molecular Indicators
3.1
3.2
4
66
68
Voltage-Sensitive Dyes . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
3.1.1
Brief History . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
3.1.2
Physical mechanisms of potentiometric dyes . . . . . . . . . . .
70
Ion indicator probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
In vivo fluorescence microscopy of neuronal activity in three dimensions using
voltage-sensitive dyes
76
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
4.2
Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . .
78
4.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
80
4.4
Appendix: Confocal microscopy study of voltage-sensitive dye staining
penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
5 Spatiotemporal activity patterns during respiratory rhythmogenesis in the rat
ventrolateral medulla
87
5.1
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88
5.2
Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
5.2.1
General preparation . . . . . . . . . . . . . . . . . . . . . . . . .
92
5.2.2
Imaging system . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
5.2.3
Dye staining . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
xv
5.3
5.4
5.5
5.2.4
Measurement protocol . . . . . . . . . . . . . . . . . . . . . . .
96
5.2.5
Image analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.3.1
General trends in the optical signal . . . . . . . . . . . . . . . . . 101
5.3.2
Timing of respiratory phase-locked signals . . . . . . . . . . . . 101
5.3.3
Influence of baseline (F ) on sign of ∆F/F . . . . . . . . . . . . 103
5.3.4
Spatiotemporal trends in event-triggered imaging data . . . . . . 103
5.3.5
Correlation coefficient analysis . . . . . . . . . . . . . . . . . . . 104
5.3.6
Extracellular recordings within imaged field of view . . . . . . . 108
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.4.1
Virtues of a perfused preparation . . . . . . . . . . . . . . . . . . 108
5.4.2
pBC region contains a variety of phase-locked fluorescence patterns109
5.4.3
Role of inhibition . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.4.4
Tissue optics considerations and interpretation of the optical signals 111
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6 Two-Photon Spectroscopy of Functional Probes for Bioimaging
6.1
115
Measurement of two-photon absorption (TPA) cross-sections . . . . . . . 115
6.1.1
Background Theory . . . . . . . . . . . . . . . . . . . . . . . . 115
6.1.2
Two-Photon Spectroscopy Setup . . . . . . . . . . . . . . . . . . 117
6.1.3
Calculation of Collection Efficiency . . . . . . . . . . . . . . . . 120
6.1.4
σ2 Measurement Procedure . . . . . . . . . . . . . . . . . . . . . 122
xvi
6.2
6.3
Near Infrared Two-Photon Excitation Cross-sections of Voltage-Sensitive
Dyes
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.2.1
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.2.2
Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.2.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2.4
Linearity Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2.5
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2.6
Prospects for Two-Photon Imaging using VSDs . . . . . . . . . . 130
6.2.7
Conclusion (VSD) . . . . . . . . . . . . . . . . . . . . . . . . . 132
Near-Infrared Two-Photon Cross-Sections of Novel Conjugated Porphyrin
Compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.3.2
Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.3.3
6.3.2.1
Materials . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.3.2.2
Measurement Procedures . . . . . . . . . . . . . . . . 137
6.3.2.3
Extraction of TPE Cross-Sections from Power-law tests 138
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.3.3.1
TPA Cross-Section Spectra . . . . . . . . . . . . . . . 141
6.3.3.2
Power-law Tests . . . . . . . . . . . . . . . . . . . . . 143
6.3.3.3
Extracted values of ² and σ2−photon . . . . . . . . . . . 145
6.3.4
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.3.5
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
xvii
7 Multifocal Two-Photon Microscopy
149
7.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.2
Anatomical two-photon imaging . . . . . . . . . . . . . . . . . . . . . . 150
7.2.1
Anatomical two-photon imaging in vitro . . . . . . . . . . . . . . 151
7.2.1.1
Surgical preparation of thalamocortical slices . . . . . . 152
7.2.1.2
In vitro Anatomical imaging via voltage-sensitive dye
staining . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.2.2
Anatomical two-photon imaging in vivo . . . . . . . . . . . . . . 156
7.2.2.1
Surgical preparation and in vivo imaging protocol . . . 156
7.2.2.2
In vivo Anatomical imaging via voltage-sensitive dye
staining . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.2.2.3
In vivo enhanced Green Fluorescent Protein (EGFP) imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8 Two-photon fluorescence recording of action potentials from individual mammalian nerve terminals in situ
162
8.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.2
Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.2.1
Two-Photon Imaging Optics . . . . . . . . . . . . . . . . . . . . 165
8.2.2
Two-Photon Functional Recordings . . . . . . . . . . . . . . . . 168
8.2.3
Action Spectra Measurements . . . . . . . . . . . . . . . . . . . 171
8.2.4
One-Photon Spectroscopy . . . . . . . . . . . . . . . . . . . . . 172
xviii
8.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.3.1
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
8.3.2
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9 Summary
180
Bibliography
183
xix
List of Tables
2.1
Specifications of Coherent Mira 900 Basic and Chameleon ultrafast Ti:Al2 O3
lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
20
Mechanical and electrical specifications and input parameters for Cambridge Technology Model 6220 galvanometer mirrors and MicroMax 671
Control board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
6.1
Summary of voltage-sensitive dye sample sources and concentrations . . 125
6.2
Values of σ2 for voltage-sensitive dyes at selected wavelengths . . . . . . 127
6.3
Linearity tests for power-square fluorescence dependence of voltage-sensitive
dyes.
6.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Extracted values of TPA cross-sections for selected compounds. . . . . . 146
xx
List of Figures
1.1
Linear and non-linear absorption of plane-wave illumination in absorptive
solution.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
Jablonsky diagram for one- and two-photon absorption.
2.1
Illumination and detection configurations for one-photon fluorescence microscopy
. . . . . . . . .
5
7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
2.2
Power spectra comparison of mercury and xenon arc lamp sources . . . .
16
2.3
Laser speckle phenomenon . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.4
Optical layout of Mira 900 Basic Ti:Al2 O3 laser
21
2.5
Comparison of CCD and EMCCD detector architecture
. . . . . . . . .
23
2.6
Fiber-Coupled Epi-Illumination Imaging System . . . . . . . . . . . . .
25
2.7
Qualitative optical properties of a converging gradient index (GRIN) rod
. . . . . . . . . . . . .
lens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
2.8
Geometrical optics of a cylindrical GRIN lens . . . . . . . . . . . . . . .
28
2.9
GRIN lens fluorescence endoscopy setup
30
. . . . . . . . . . . . . . . . .
2.10 Multifocal two-photon microscope optics schematic
xxi
. . . . . . . . . . .
33
2.11 Code V modeling of microlens array function . . . . . . . . . . . . . . .
34
2.12 Inside view of multifocal two-photon microscope optics device
35
. . . . .
2.13 Ray-tracing diagram for single off-axis beam in multifocal two-photon
microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.14 Code V modeling of demagnification optics for multifocal detection . . .
37
2.15 Scanning schemes for multifocal two-photon microscope . . . . . . . . .
41
2.16 Flowchart for device control and data acquisition software . . . . . . . .
45
2.17 Multifocal device synchronization timing sequence . . . . . . . . . . . .
47
2.18 Image reconstruction of multifocal data . . . . . . . . . . . . . . . . . .
48
2.19 Geometry of simplified two-dimensional imaging scenario for deconvolution techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
2.20 Experimentally measured point-spread-function (PSF) and optical transfer
function (OTF) of GRIN lens fluorescence imaging device. . . . . . . . .
56
2.21 Graphical illustration of nearest-neighbor de-blurring procedure . . . . .
59
2.22 Effects of nearest neighbor de-blurring on surface VSD ∆F/F measurements in mouse barrel cortex in vivo . . . . . . . . . . . . . . . . . . . .
61
2.23 Multifocal 2-photon microscope optical alignment diagram . . . . . . . .
63
3.1
Shift of absorption and emission spectra in voltage-sensitive dyes as a
function of membrane potential
. . . . . . . . . . . . . . . . . . . . . .
69
3.2
Voltage-sensitive dye response to action potential in squid giant axon
. .
69
3.3
Dimerization mechanism for voltage-sensitive dye spectral changes . . .
71
3.4
Indicators of Ca2+ concentration . . . . . . . . . . . . . . . . . . . . . .
74
xxii
3.5
Fluorescence response of Calcium Green-1 and Fura-2 to Ca2+ concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
4.1
GRIN lens fluorescence endoscopy experimental configuration . . . . . .
79
4.2
RH 795 excitation and emission spectra . . . . . . . . . . . . . . . . . .
80
4.3
Depth-dependent neural activity in response to cortical and thalamic stimulation
4.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
82
Activity response patterns compared with known cortical cellular architecture
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
4.5
Confocal survey imaging of RH 795-stained coronal brain slices . . . . .
85
4.6
Confocal imaging of RH 795 staining depth penetration. . . . . . . . . .
86
5.1
Anatomy of brainstem regions involved in respiratory rhythm generation
89
5.2
Firing patterns of respiratory neurons in the brainstem . . . . . . . . . . .
91
5.3
Respiratory activity imaging system and physiological validation
95
5.4
Fluorescent respiratory-related signals from the rat VLM . . . . . . . . . 100
5.5
Spatiotemporal correlation analysis of activity in the rat VLM . . . . . . 107
6.1
Cross-section measurement setup
6.2
Chemical structures of voltage-sensitive dyes . . . . . . . . . . . . . . . 126
6.3
Log-scale plots of two-photon excitation (TPE) cross-sections of voltage-
. . . .
. . . . . . . . . . . . . . . . . . . . . 119
sensitive dyes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.4
Comparison of linear (one-photon) absorbance and two-photon excitation
(TPE) cross-sections (σ2 ) for Nile Blue A. . . . . . . . . . . . . . . . . . 129
6.5
Comparison of σ2 for di-8-ANEPPDHQ, RH795, and di-8-ANEPPS.
xxiii
. . 130
6.6
Chemical structures of novel conjugated (porphinato)zinc(II) compounds
6.7
Log-scale plots of two-photon excitation (TPE) cross-sections of (porphi-
136
nato)zinc(II) compounds . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.8
Power-law tests for fluorescence dependence on incident excitation laser
beam intensity performed at varying laser excitation wavelength.
6.9
. . . . 144
Power-law tests for fluorescence dependence on incident excitation laser
beam intensity performed at varying sample concentration. . . . . . . . . 145
7.1
Photograph of two-photon slice imaging chamber . . . . . . . . . . . . . 153
7.2
Multifocal two-photon background image of di-4-ANEPPDHQ-stained
thalamocortical slice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.3
Imaging in depth through di-4-ANEPPDHQ-stained thalamocortical slices
154
7.4
3D volumetric rendering of two-photon depth-scanned data in a di-4-ANEPPDHQstained thalamocortical slice . . . . . . . . . . . . . . . . . . . . . . . . 155
7.5
Photograph of in vivo two-photon imaging configuration detail . . . . . . 158
7.6
Two-photon background fluorescence image of di-4-ANEPPDHQ-stained
somatosensory cortex in vivo . . . . . . . . . . . . . . . . . . . . . . . . 159
7.7
Imaging EGFP-labeled inhibitory interneurons in mouse cortex in vivo
and in vitro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.1
Optical setup for two-photon imaging of the neurohypophysis . . . . . . 167
8.2
Anatomy of the neurohypophysis . . . . . . . . . . . . . . . . . . . . . . 169
xxiv
8.3
Two-photon recording of action potentials from single mammalian nerve
terminals in situ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.4
Comparison of one- and two-photon action spectra for potentiometric dye
di-3-ANEPPDHQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.5
Signal-to-noise and ∆F/F action spectrum for potentiometric dye di-3ANEPPDHQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.6
One-photon excitation and emission spectra for potentiometric dye di-3ANEPPDHQ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
xxv
Chapter 1
Introduction
1.1 Optical Imaging of Brain Function
Neurons communicate information via millisecond-scale changes in membrane potential.
The computational power of the brain arises, in large part, from the diversity of electrical
information provided to each neuron by its unique connections. Additionally, the work of
Llinas and others has shown that large-scale aspects of brain function can originate from
electrical properties of individual neurons [115,116]. The ideal experimental technique for
studying brain function, therefore, enables simultaneous monitoring of electrical activity
in many neurons from networks distributed throughout the volume of the brain.
The traditional method for measuring electrical activity with high temporal resolution
is by the use of intracellular and extracellular electrodes inserted into the tissue. Measurement of macroscopic brain function, however, is impractical even with large electrode
arrays. At the same time, measurement of membrane potential in small neurons and their
1
processes is extremely difficult using these traditional electrode techniques. Functional
imaging of neuronal activity in three dimensions thus offers new possibilities for understanding the physiology of the brain. Since the functional architecture of the brain is threedimensional, one of the goals of neuroimaging is to accurately resolve neuronal activity in
three dimensions with high spatial and temporal resolution.
Direct imaging of millisecond neuronal electrical activity at high spatial resolution,
however, is difficult. Measuring electrical activity in individual cells requires a technique
sensitive to highly localized changes in membrane potential. Spatial resolution, at minimum, must be on the order of the size of a neuron, ∼ 10 µm soma diameter. Imaging
techniques such as electroencephalography (EEG), or magnetoencephalography (MEG)
are sensitive to electrical activity, but they average the activities of thousands of cells
across millimeters of tissue. Multi-electrode recordings sample cells located at significant
distances from each other, but cannot reveal intracellular voltages, crucial for understanding the underlying biophysical mechanisms.
Imaging modalities sensitive to a host of “intrinsic” optical signals that accompany
or follow electrical activity have been developed. Fast (∼ 2 ms) changes in tissue lightscattering properties correlated with electrical signals can be detected directly in vitro
and in vivo [168]. In response to neuronal electrical activity, cellular metabolic rates
increase. These changes, occurring on a slower timescale (seconds), can be detected in
the form of absorption changes due to blood hemodynamic response. Functional magnetic
resonance imaging (fMRI) [176], positron emission tomography (PET) [152], and diffuse
optical tomography (DOT) using near-infrared light [199] are all effective modalities for
2
monitoring hemodynamic or metabolic brain function at the tissue level.
Fluorescence microscopy using functional probes combines the advantage of high spatial resolution afforded by single-scattering optical imaging techniques with the ability to
quantitatively measure membrane potential directly. In this thesis I describe applications
of fluorescence microscopy to the study of neural systems in vivo and in vitro. The novel
imaging devices employed in these experiments utilize either one- or two-photon excitation of fluorescent indicators. One-photon excitation (linear) fluorescence microscopy was
used to study the spatiotemporal patterns of activity in the mouse primary somatosensory
cortex in vivo and in the ventrolateral medulla of the juvenile rat in vitro. Two-photon excitation (nonlinear) fluorescence microscopy, utilizing a novel multi-beam scanning microscope, was employed for anatomical and functional imaging in vivo and in vitro. In-depth
spectroscopic studies of fluorophore two-photon cross-sections, crucial for functional twophoton imaging, are presented as well.
1.2 Background theory: One- and two-photon absorption
1.2.1 One-photon absorption
The intensity I(z) of light passing through a solution containing absorptive molecules
with number density N (cm−3 ) and effective absorption cross section σ (cm2 ) (see Figure
1.1) experiences an incremental change in illumination intensity, dI. This behavior in
3
one-dimension is governed by the standard Beer’s law equation
dI
= −IσN.
dz
(1.1)
Integrating over the sample thickness d (cm) and adding the boundary condition I(0) =
I0 , Equation 1.1 can be rearranged and integrated to yield
ln
I0
= σN d.
I
(1.2)
More typically, this is expressed in terms of log10 and the molar extinction coefficient
² (M −1 ·cm−1 ), i.e.
log10
I0
= ²Cd,
I
(1.3)
where C is the molar concentration (M ).1 Thus the relationship between the extinction
coefficient and the absorption cross-section is
σ = 2.303
1
²C
.
N
(1.4)
The units for the molar extinction coefficient, ², depend on the units used for C. When C is in units of
moles (M ), ² is in units of M −1 · cm−1 .
4
d
z
Figure 1.1: One- and two-photon absorption of light in a solution containing absorptive
molecules at a number density of N (cm−3 ) and with one-photon cross section σ and
two-photon absorption coefficient β.
1.2.2 Two-photon absorption
In two-photon absorption, originally predicted by Göppert-Mayer [77], an atom or molecule
simultaneously absorbs two photons and makes a transition from its ground state to an excited state (see Jablonsky diagram representation in Fig. 1.2). A quantum mechanical
description of nonlinear processes can be generally described with or without quantizing
the electric field. For the present discussion, two-photon absorption can be characterized
in terms of the macroscopically observed nonlinear absorption of light passing through homogenous sample (for a review of the quantum field derivation of two-photon absorption,
see [150]). Assuming no linear (i.e. one-photon) absorption, the two-photon absorption of
light propagating through an optically thin sample of C molecules per unit volume (cm−3 )
is characterized by the following differential equation for the input light intensity I:
5
dI
= −βI 2
dz
(1.5)
Here I is the source intensity (erg cm−2 sec−1 ), and β is the two-photon absorption
coefficient. The two-photon absorption coefficient, β, can also be expressed in terms of
the two-photon cross-section, Σ, i.e.
β = 2C
Σ
~ω
(1.6)
where ω is the angular frequency of the incident light field; the factor of 2 arises
because two photons are absorbed, and Σ has cgs units of cm4 ·sec. In Equation 1.6, ~ω is
the energy per photon (erg·photon−1 ) at the excitation wavelength. The informal unit for
Σ is the Göppert-Mayer (GM), where 1 GM = 10−50 ·cm4 ·sec·photon−1 . Σ is described
as a “cross-section” in order to establish a two-photon analog to the linear (one-photon)
absorption cross-section, which has true units of area. The extra factor of cm2 is due to
the extra factor of I in Equation 1.5.
Following the theoretical description by Burris and McIlrath [26], absorption of the
incident electric field can also be described in terms of the complex index of refraction,
kc
= η + iκ,
ω
(1.7)
where k is the magnitude of the wave vector at angular frequency ω, and η and κ are,
respectively, the real and complex components of the index of refraction. The complex
6
2-photon
1-photon
S2
internal conversion
S1
S0
Figure 1.2: Jablonsky diagram showing one- and two-photon absorption processes followed by fluorescence. S0 is the ground state, S1 and S2 are excited states. The multiple
lines at each state represent different vibrational energy levels. ω1 is the angular frequency of a photon at the peak of a one-photon absorption resonance, ω2 = 12 ω1 , and
h
~ = 2π
≈ 1.054 × 10−34 J·s.
part of the index of refraction, κ, is responsible for absorption of the incident electric
field. The index of refraction can, in turn, be described by the macroscopic susceptibility
as
k 2 c2
= 1 + 4πχ.
ω2
(1.8)
Describing the nonlinear susceptibility as a power series expansion in order of the
incident electric field strength, E,
χ = χ(1) + χ(2) E + χ(3) E 2 + . . . ,
(1.9)
the complex component of the index of refraction can be related to the imaginary
7
component of χ by
2ηκ = 4πIm[χ(1) + χ(3) E2 + . . .].
(1.10)
Note that χ(2) is excluded because it makes no contribution to the susceptibility in
an isotropic medium because of inversion symmetry. Thus the first nonlinear absorptive
contribution arises from χ(3) . In the following discussion of two-photon absorption, it
is assumed that experimental measurements are concerned with regions with effectively
no one-photon absorption, thus χ(1) = 0. It is also assumed that Reχ(1) À Reχ(3) E2 ,
indicating that there is no intensity-dependent change in the real refractive index. With
these assumptions, Eq. 1.10 can be rearranged to yield κ directly in terms of χ(3) and I:
κ=
4π 2 IImχ(3)
,
η2c
(1.11)
where I = cηE 2 /2π. Now relating the absorption coefficient β and κ,
β=
2ω
κ,
cI
(1.12)
if the left hand side of Eq. 1.12 is replaced with the result of Eq. 1.6, and κ with Eq.
1.11, Σ can be defined in terms of the material third-order susceptibility, χ(3) , i.e.
Σ=
4π 2 ~ω 2 Imχ(3)
.
Cη 2 c2
(1.13)
Here η is the real part of the linear index of refraction at ω, and c is the speed of light.
8
In order to describe the terms in χ(3) leading to two-photon absorption, it is sufficient
to consider a density matrix calculation of the quantum mechanical perturbation theory expansion for a non-interacting assembly of distinguishable molecules [119]. For simplicity,
it is also assumed that there is only one intermediate state and that only the ground state
(0)
(0)
g is populated, ρgg = 1, ρll = 0 for l 6= g. χ(3) is a fourth-rank tensor. Thus the most
(3)
general description of the third-order nonlinear susceptibility χijkl (ω4 = ω1 +ω2 +ω3 ) contains 81 terms. Significant simplifications, however, arise when the medium is an isotropic
collection of random molecules, and when nonlinear terms such as χijkl (ω = ω + ω − ω)
are large compared to the other terms, e.g. as a result of small resonant denominators. In
this case, χ(3) in Eq. 1.13 is well described as a frequency-dependent scalar [85].
1.3 Thesis Outline
In this thesis I describe background principles and present novel instrumentation employing functional one- and two-photon fluorescence imaging of neuronal activity. Functional
imaging experiments investigated spatiotemporal patterns of both stimulated and spontaneous neuronal activity in a variety of mammalian in vivo and in vitro preparations.
Physiological results and implications are also presented.
Chapter 2 details the instrumentation for both one- and two-photon fluorescence microscopy. A fiber-coupled white-light source-based imaging device used for conventional
(2-D) fluorescence microscopy of an in vitro preparation of the juvenile rat brainstem is described in Section 2.3. A gradient index (GRIN) lens one-photon fluorescence endoscopy
9
device for high speed 3-D imaging of the mouse cortex in vivo is described in Section 2.4.
Lastly, a multifocal two photon microscope for 3-D anatomical and functional imaging of
neuronal activity in vivo and in vitro is described in Section 2.5. Chapter 3 presents an
overview of the mechanisms and history of the fluorescent indicators used throughout the
experiments described in this thesis.
Chapters 4 and 5 each describe a major experiment employing one-photon fluorescence microscopy and voltage-sensitive dyes. Chapter 4 describes the first in vivo highspeed 3-D imaging of electrical neuronal activity in the first few layers of the cortex (up
to ∼ 150 µm deep). Our approach uses a gradient index (GRIN) lens epi-illumination
endoscope. Chapter 5 describes an application of 2-D (conventional) one-photon fluorescence imaging to visualize spatiotemporal patterns of electrical activity during the process
of breathing rhythm generation (rhythmogenesis) in an in vitro preparation of the juvenile
rat brainstem.
Measurement of two-photon absorption and excitation (TPA and TPE, respectively)
cross-sections (Σ2 and σ2 ) of fluorescent molecules via an emission-based ratiometric
spectroscopy technique is detailed in Chapter 6. Background theory, experimental setup,
and measurement procedure are described in Section 6.1. The results of two-photon absorption cross-section measurement in a series of voltage-sensitive fluorescent compounds
are described in Section 6.2. Cross-section measurement results and an analysis of relative one- and two-photon absorption contributions in a series of novel conjugated (porphinato)zinc(II) compounds are described in Section 6.3.
Chapter 7 presents the results of two-photon imaging in vitro and in vivo using the
10
novel multifocal microscope device described in Section 2.5. Results of 3D anatomical
imaging using either potentiometric dye background fluorescence or green fluorescent
protein (GFP) expression are described in Section 7.2. Single-beam scanning two-photon
excited fluorescence measurements of action potentials from single nerve terminals in the
mammalian neurohypophysis in situ are described in Chapter 8.
11
Chapter 2
Imaging Systems
2.1 Introduction
This chapter describes the devices used for all neuroimaging experiments in this thesis.
First, common optical components including illumination sources and detectors are discussed and compared. The following sections describe one of three main imaging devices
utilized in this thesis. Two of the devices, a fiber-coupled epi-illumination imaging system
for conventional 2D fluorescence imaging and a gradient-index (GRIN) lens fluorescence
endoscope imaging device for 3D imaging, employed one-photon excitation fluorescence.
These two devices are described first. The remaining device, a multifocal two-photon
microscope, employed two-photon excitation and is discussed last. This device is by far
the most complex of the three. Experiments utilizing the one-photon excitation devices
are described in detail in Chapters 4 & 5, and those utilizing the two-photon microscope
are described in Chapter 7. An additional section devoted to specific device operational
12
aspects of the two-photon device are included in the Appendix.
2.2 Components
2.2.1 Illumination sources for one-photon fluorescence microscopy
Fluorescence microscopy relies on the absorption and re-emission of light by fluorescent
molecules. The collected fluorescence emission signal (photons s−1 ) is always a small
fraction of the excitation power because of limited absorption, emission, and detection
probabilities. Low fluorescence signal imposes an upper limit on imaging speed in functional fluorescence measurements. At short integration times, the functional fluorescence
signal may be comparable to the noise level of the statistical uncertainty in arrival time of
photons (shot noise1 ).
Illumination sources for functional fluorescence imaging must therefore first and foremost provide sufficient power at the wavelengths corresponding to the fluorophore absorption peak. This ensures that the largest possible fraction of the excitation illumination will
be re-emitted as detectable fluorescence signal. Secondly, temporal intensity fluctuations
in the illumination source must be smaller than the signal size, or at least have very different time-scales. In that case, high- or low-pass filtering could effectively remove such
noise factors from the data.
1
For a photon detection system, the shot noise detected (in photoelectrons) is equal
√ to the square root of
the number of total detected photoelectrons for a given integration time (Nshot = Ndetected ).
13
Three main species of illumination source are currently in widespread use for functional one-photon fluorescence microscopy: (1) white light sources (e.g. arc lamps, halogen lamps, etc.), (2) lasers, and (3) light-emitting diodes (LEDs). White light sources continue to be the most common illumination source in commercial fluorescence microscopy
systems. Typical optical configurations for these illumination sources are shown in Figure
2.1.
White light sources offer the most flexibility in excitation wavelengths, but they must
be carefully selected on the basis of power spectrum and noise level to ensure that sufficient photons will be delivered at the desired wavelength (taking into account all excitation
filters, collimation/steering optics, etc.). Figure 2.2 compares the spectra of two common
arc lamp gases, mercury and xenon, used in commercially available white light sources.
Laser sources afford high power at many excitation wavelengths spanning the UV–IR
and beyond. Because laser sources are typically monochromatic (or polychromatic over
a limited set of spectral lines), there is no need for excitation filters in the optical path;
one simply chooses a laser operating at a wavelength close to the peak absorption of the
fluorophore of interest. This approach can require multiple lasers, especially if multiple
dyes with considerably different absorption spectra are employed. As solid-state lasers
become more affordable, the laser is increasingly attractive as a high-power, low-noise
illumination sources for fluorescence imaging. Lasers are particularly useful in situations
where the fluorophore employed has either a low quantum yield (q) or if the functional
fluorescence changes are exceptionally small, as in the case of voltage-sensitive dyes.
Laser speckle observed on illuminated surfaces can make laser illumination sources
14
A
B
White light
source
White light
source
excitation
filter
emission
filter
excitation
filter
dichroic
mirror
CCD
thick sample
thin sample
C
Laser
emission
filter
CCD
beam expansion
optics
emission
filter
dichroic
mirror
CCD
thick sample
Figure 2.1: Standard illumination and detection configurations for one-photon fluorescence microscopy. A: Epi-illumination configuration using white light source. For
narrow-band LED illumination sources, the excitation filter is removed. B: Bright-field
illumination configuration using white light source. As in A, excitation filter is removed
for LED illumination. Bright-field illumination is most suitable for thin-sample imaging. C: Epi-illumination configuration using a laser illumination source. Beam expansion
optics are required for full-field illumination.
15
Mercury
% of total
tal outpu
output power
% of total output power
Xenon
wavelength (nm)
wavelength (nm)
Figure 2.2: Power spectra comparison of mercury (left) and xenon (right) arc lamp white
light sources. Note comparatively more level power spectrum along visible wavelength
range (∼400–700 nm). Adapted from Cairn Research Ltd. online material [29].
non-ideal for wide-field illumination. Laser “speckle” is caused by constructive and destructive interference of coherent light scattering off of a non-uniform surface terrain (Figure 2.3A). This typically causes an unwanted “speckled” spatial heterogeneity in illumination. This effect can be effectively countered with simple additions to the optical path,
such as the use of a vibrating mirror to temporally average out (or “smear”) the observed
speckle. Alternatively, an optical diffuser element, consisting of a volume of highly scattering (but otherwise non-absorptive) media, can be employed (Fig. 2.3C). It should be
noted that the spatiotemporal intensity fluctuations of laser speckle observed on tissue surface can in fact be utilized to extract physiological information such as blood flow [24,57].
Light-emitting diodes (LEDs) are semiconductor devices that emit non-coherent light
over a narrow wavelength band via electroluminescence. LEDs emitting wavelengths from
UV to near-infrared are commonly available. LEDs at visible wavelengths have been utilized for both fluorescence and intrinsic signal imaging of neural activity, but are limited
16
A
C
B
turbid
media
vibrating
mirror
Figure 2.3: Laser speckle phenomenon. A: Illustration of laser speckle generation. (Left)
Coherent light (e.g. from a laser) illuminates a screen containing some surface texture.
(Right) Close-up view of the illuminated screen showing the screen surface texture as a
distribution of scattering “points”. Each point illuminated within illumination spot (shown
in faint red solid circle) generates a spherically radiating point wavefront, and the overlapping wavefronts interfere constructively and destructively, creating a random speckle
pattern. B: Photograph of laser speckle pattern on a concrete wall. C: Strategies for eliminating laser speckle. (Top) Use of a diffusing medium to reduce spatial and temporal
coherence; (Bottom) Reflecting laser off of a vibrating mirror to average-out (“smear”)
the speckle pattern.
17
by long-term power stability and limited output power. While standard LEDs tend to output relatively modest power levels (∼ 1 mW), newer superluminescent diodes (SLDs)
boast superior power stability and can reach output power levels of 50–100 mW. These
illumination sources have been shown to be effective light sources for imaging small intrinsic signals in in vitro preparations, and they are superior to tungsten halogen lamps for
monitoring low-frequency optical signals [170].
Two basic illumination and detection configurations for one-photon fluorescence microscopy (shown in Fig. 2.1) are epi-illumination and bright-field illumination. In epiillumination (Fig. 2.1 A and C), fluorescence signal is detected back through the same
objective used for focusing excitation light. The emission signal is then separated from
the excitation light by a dichroic mirror which reflects the fluorescence signal toward the
detector. In bright-field illumination (Fig. 2.1 B), the fluorescence signal is detected on
the reverse side of the specimen, so the fluorescence must traverse the thickness of the
sample. Because the fluorescence signal experiences absorption and scattering in thick
tissue samples, bright-field illumination is appropriate only for thin specimens such as
vibratome sliced tissue, slide-mounted cultured cells, etc. Epi-illumination, on the other
hand, is quite suitable for thick-specimen imaging, and is mandatory for in vivo imaging.
2.2.2 Illumination sources for two-photon fluorescence microscopy
The rate of two-photon absorption is proportional to the square of excitation intensity
(see Chapter 1 for background theory). Thus pulsed modelocked laser sources offering
18
high peak power, but moderate average power are ideal for these applications because
while there is sufficient photon flux to cause detectable levels of two-photon absorption
during the pulse peaks, the low average power prevents excessive photodamage to the
sample. While CW lasers can be used to generate two-photon excitation of fluorescent
molecules [196], the rates of two-photon absorption attainable by modelocked lasers are
easily several orders of magnitude higher than CW lasers. Additionally, pulsed modelocked lasers enable superior noise-rejection through the use of lock-in detection (e.g. via
lock-in amplifier) at the laser repetition frequency.
The most useful modelocked laser sources for two-photon excitation microscopy are
the ultrafast pulsed lasers, offering femtosecond (fs) pulse durations. By far, the most
popular laser for this application is the modelocked Ti:Al2 O3 (Titanium:Sapphire) laser.
This laser employs visible CW laser light to excite the Ti:Al2 O3 crystal gain medium to
a series of excited states. Required pump laser wavelengths are in the range of 514–532
nm, and the most frequently used sources include Argon-ion gas lasers and frequencydoubled solid-state Nd:YAG, Nd:YLF, and Nd:YVO4 lasers. A birefringent filter within
the Ti:Al2 O3 laser cavity is used to select (or “tune”) the emission wavelength. The
Ti:Al2 O3 laser is widely tunable over near-infrared wavelengths (see specifications in Table 2.1), although one has to change cavity optics in order to access the full wavelength
range.
In CW mode, the Ti:Al2 O3 gain medium amplifies all wavelengths between 680–1100
nm. Cavity geometry, cavity optics (Fig. 2.4), and added wavelength selectivity of components such as tunable birefringent filters narrow this wavelength range. Additionally,
19
Coherent Mira 900B (passive modelocking)
Property
Value
Pump Source
8W Argon ion laser
Tuning Range
700–960 nm
Max Average Output Power (Pav )
1W
Pulse Width
150 fs
Peak Pulse Power (at Pav )
88 kW
Repetition Rate
76 MHz
Coherent Chameleon (active modelocking)
Pump Source
solid-state Nd:YVO4
Tuning Range
720–955 nm
Max Average Output Power (Pav )
1.4 W
Pulse Width
140 fs
Peak Pulse Power (at Pav )
111 kW
Repetition Rate
90 MHz
Table 2.1: Specifications of Coherent Mira 900 Basic and Chameleon ultrafast Ti:Al2 O3
lasers
Note that peak pulse power Ppeak can be calculated from the average output power, Pav ,
the pulse width, τ , and the repetition rate f : Ppeak = Pτav
.
f
20
for each lasing wavelength, an integral number of half wavelengths must fit within the
main lasing cavity. Each wavelength fitting this condition constitutes a longitudinal mode.
In CW operation, these longitudinal modes are not temporally synchronized, and the laser
output is approximately constant in time containing all of the longitudinal modes. Modelocking refers to the synchronization, or phase-locking of these modes such that the output
fields of the longitudinal modes add up constructively to generate a single pulse of much
higher power. The output power of a modelocked laser time-averaged over a full duty cycle (the reciprocal of the pulse separation) is therefore essentially the same as the average
power during CW operation, but the peak power at the points of pulse maxima is orders
of magnitude higher (see specifications in Table 2.1).
auxiliary alignment cavity
M8
"back" high reflector of
main cavity
M9
M7
BP1
M4
M5
L1
M6
BP2
"front" partial reflector
of main cavity
M2
M1
BRF
M3
Figure 2.4: Layout of optical components comprising the main and auxiliary alignment
cavities. Nomenclature of optical components (mirrors M1, M2, etc.; lens L1; Brewster Prism BP1, BP2) as described in Coherent product documentation. Ti:Al2 O3 lasing
medium indicated in red. Alignment cavity is only used for initial alignment of mirrors
M1–M5.
21
2.2.3 Detectors: CCDs and EMCCDs
In Charge-Coupled Device (CCD) detectors, incident photons of sufficient energy are absorbed by a silicon diode photosensor (pixel), and consequently generate electrons in the
silicon crystal via the photoelectric effect. The electrons are then shifted to a charge storage region that is, in turn, connected to an amplifier that reads out the accumulated charge.
CCD detectors consist of arrays of thousands to millions of such photosensitive pixels,
and are relatively inexpensive to manufacture. They constitute extremely practical detectors for video-rate (∼ 30 frames / second) 2D imaging at illumination levels where
measurements are not shot noise limited.
The main limitation of conventional CCDs is the relatively large amount of read-out
noise introduced by the charge amplifier stage. This is particularly disadvantageous for
high-speed imaging (> 30 frames / second) because noise scales with the bandwidth of the
amplifier, and high-speed measurements require high-bandwidth amplifiers. To avoid this
constraint, electron-multiplying CCDs (EMCCDs) introduce an additional step of lownoise charge multiplication before the charges are read off of the chip via the amplifier
(see Figure 2.5 for a comparison of CCD and EMCCD architecture). EMCCDs achieve
this electron multiplication by maximizing the rate of impact ionization, which occurs
when a charge has sufficient energy to create another electron-hole pair and free electron
charge in the conduction band creates another charge. Typically, impact ionization is seen
as source of noise because it is relatively infrequent and not a homogeneous effect applied
to all “signal” charges. In EMCCD detectors, however, the probability of a charge creating
22
a secondary electron is increased by giving the electron charge more energy by clocking
the charge with a higher voltage. Additionally, there are hundreds of cells in the electron
multiplication register in which impact ionization can occur. By controlling the process in
this way, on-chip gain levels of up to thousands are possible. This enables faster readout
rates (at low noise) and makes it possible to detect small signals (down to single photon
counting) with a high signal-to-noise ratio [7].
Electron Multiplying CCD
(EMCCD)
Conventional CCD
Figure 2.5: A comparison of Charge-Coupled Device (CCD) and electron-multiplying
CCD (EMCCD) detector architectures. In the EMCCD, before the charges stored in the
charge storage region are read out by the output amplified, they go through a step of onchip low-noise electron multiplication before being read-off. Reproduced from Andor
Technology Manual [7].
2.3 Fiber-Coupled Epi-Illumination Imaging System
A fiber-coupled epi-illumination imaging device was employed for imaging neural activity
in the ventrolateral medulla (in the brainstem) during the process of spontaneous breathing
in a perfused preparation of the juvenile rat [68] (Chapter 5).
The imaging optics consisted of a custom 1:1 focal length ratio lens couplet (Fig.5.3A).
23
Two identical achromatic lenses (Edmund Optics, Barrington, NJ) served as objective and
focusing lenses. Light from a 200-W halogen white light source (Thermo Oriel, Newport Corporation, Irvine, CA) was filtered with a 480 ± 20 nm excitation filter (Chroma
Technology Corporation, Rockingham, VT). The filtered light was then coupled into the
imaging device through a multimode fiber bundle which was recollimated with a collimating lens triplet (Melles Griot, Barloworld Scientific, Carlsbad, CA) and reflected off
of a dichroic mirror (<500 nm R, Chroma Technology Corporation, Rockingham, VT) to
provide epi-illumination through the achromatic objective lens (Edmund Optics, Barrington, NJ). The fluorescence signal was filtered with a >590 nm long-pass filter (Chroma
Technology Corporation, Rockingham, VT). This relatively inexpensive optics configuration provides a large imaging area with sufficiently high numerical aperture (N.A. = 0.38)
for voltage-sensitive dye fluorescence imaging. Additionally, the imaging device has a
long working distance because of the large objective aperture. This allowed for considerable electrophysiological, pharmacological, and surgical manipulation of the imaging area
during recording as needed.
Digital CCD image data and analog electrophysiological data were simultaneously acquired by a PC computer using custom-written software (for Linux operating system) and
hardware [155, 156]. The device was capable of continuous image and data recording at
high speeds (up to 1 kHz frame rate for video data, 20 kHz/channel for electrophysiology
data) for long periods of time (> 1 hour), limited only by the maximum file size allowed
by the operating system (2GB, for the Linux OS used in these experiments). The detector
was a CCD camera (TC255, Texas Instruments, Dallas, TX, USA; 10 × 10 µm pixels, 336
24
Figure 2.6: Optical setup for one-photon fluorescence imaging of membrane potential
dynamics in the ventral medullary surface of a perfused preparation of the juvenile rat
(see Chapter 5).
× 243 pixel array). To achieve appreciable signal levels, the frame rate was set to 100 Hz.
2.4 Fiber-Coupled Gradient Index (GRIN) Lens Fluorescence Endoscope
A gradient index (GRIN) lens fluorescence endoscope device was used to image electrical
neuronal activity in three dimensions in the superficial layers of the primary somatosensory cortex of the mouse in vivo using a fluorescent voltage-sensitive dye as a contrast
agent [67] (experiment discussed in Chapter 4). The device utilized a GRIN lens as the
main objective and 532 nm illumination was coupled from a diode-pumped solid state
CW Nd:YVO4 (neodymium doped yttrium orthovanadate) laser (Verdi, Coherent, Inc.,
25
Santa Clara, CA, USA) delivered via a multimode fiber bundle. This section discusses
the optical properties of the particular type of gradient index (GRIN) lenses utilized in the
device.
2.4.1 Gradient Index (GRIN) lenses
Conventional refractive optical components utilize abrupt changes in the refractive index
and curved surfaces to manipulate the spatial profile of electromagnetic fields. Gradient index (GRIN) optics refer to any sort of refractive optical component that features a
graded change in refractive index. One result of using graded (or “gradient”) refractive
index changes is that curved surfaces are no longer necessarily needed to achieve geometrical optical effects such as focusing, defocusing, collimation, etc. Instead, these effects
can be achieved through careful selection of gradient index profiles. While production
techniques vary for GRIN lenses, one of the most popular methods of manufacture is via
ion exchange. In this process, ions embedded in a host glass (sodium, for example) are
exchanged for silver or lithium ions by diffusion from a surrounding bathing medium. Because glasses containing different ions have different refractive indices, the result of the
ion exchange process is a graded index change between the outside glass (in contact with
the surrounding medium) and the glass on the inside.
Because the refractive index profiles of GRIN lenses are almost always nonlinear, the
geometrical optical properties tend to be fairly complex. The cylindrical “converging”
GRIN lens will be discussed here. In this case the refractive index varies parabolically as
26
A
B
C
Figure 2.7: A: Qualitative ray-propagation through “converging” gradient index rod lens.
This lens consists of a cylinder of glass whose refractive index varies with the radial distance from the center axis. B: Converging GRIN rod lenses. Reproduced from [79]. C:
Rays of light follow a sinusoidal path through such cylindrical GRIN lenses and form
images periodically within the lens. Reproduced from [129].
a function of radial distance from the cylinder’s axis of symmetry, i.e.
n(r) = n0 (1 −
A 2
r ),
2
(2.1)
where n0 is the design refractive index on the optical axis, A is a positive “steepness”
constant, and r is the radial distance from the cylindrical axis of symmetry. The trajectory
of a light ray traveling through such a refractive index gradient can be solved analytically
using the calculus of variations and Fermat’s principle of “least time” [98]. The resulting
solution is a sinusoidal trajectory, illustrated in Figure 2.7C. Thus in a rod lens with an
refractive index profile described by Equation 2.1, an image will be formed periodically
within the lens, given that the lens is sufficiently long. The number of times an internal
√
image will form is defined as the “period”, P , of the GRIN lens, where P = 2π/ A.
27
Cylindrical GRIN rod lenses can therefore be effectively employed as a form of relay
lens, and in fact GRIN lenses are often used as objectives for medical endoscopes.
Figure 2.8 depicts the geometrical optics of a cylindrical GRIN lens with a refractive
index profile of the form given by Eq. 2.1. For such a configuration, the image distance,
d2 is defined in terms of the object distance, d1 , as
n0
1
√
√
√
√
n0 Ad1 cos(L A) + sin(L A)
√
√ ,
× √
A n0 Ad1 sin(L A) − cos(L A)
(2.2)
where L is the length of the rod lens.
Figure 2.8: Geometrical optics of a cylindrical GRIN lens with a refractive index profile
of the form n(r) = n0 (1 − A2 r2 ). φ is the lens diameter (typically on the order of mm),
L is the length of the lens, d1 is the object distance, d2 is the image distance, dW is the
working distance (here defined as the image distance when the object distance is zero),
and f is the effective focal length. [129].
When employed as objectives for fluorescence imaging in brain tissue, there are three
main advantages that cylindrical converging GRIN lenses afford: (1) They offer the ability to focus at “zero” object distance. This capability permits the objective to physically
contact the tissue which minimizes sample-to-detector vibration artifacts. Additionally,
contact between the objective and tissue maintains the tissue at a set distance from the
28
detector such that back-focusing (moving the detector plane to focus at different object
planes) is feasible. (2) GRIN lenses offer the unique combination of low magnification
(∼ ×3) and reasonably high numerical aperture (∼ 0.4). This is particularly useful for
the application of imaging neuronal systems, where it is desirable to measure the activity
of a large number of cells distributed over millimeters of tissue. (3) The mechanical dimensions of GRIN lenses make them ideal for minimally invasive imaging, as illustrated
by the compact endoscopic imaging system described in this section. GRIN lenses with
small diameters (< 1 mm) can be injected into tissue, thus allowing ultra deep imaging of
neuronal tissue, barring significant tissue damage.
GRIN lenses were first effectively employed as devices for imaging fast intrinsic optical changes in neural tissue in vivo in a study by Rector et al. [154] in which a GRIN
lens endoscopic imaging device was used to measure optical signals in response to electric stimulation in the cardiorespiratory areas of the rat dorsal medulla. GRIN lenses
have since been utilized for 3D one-photon fluorescence imaging of neural activity via
voltage sensitive dyes [67] and have recently been successfully combined with scanning
multiphoton fluorescence microscopy by Schnitzer and colleagues, yielding promising results [70, 71, 99].
2.4.2 Optical configuration of imaging device
The optical setup of our focusing GRIN lens fluorescence microscope is shown in Figure
2.9; the setup allowed us to alternate between dark-field (Fig. 2.9B) and epi-illumination
29
(Fig. 2.9A) and to compare the two illumination schemes. The total lens length was 160
mm. The numerical aperture of 0.42 yielded a depth of field (in fluorescence mode) of 6.8
µm. A diode-pumped solid state CW Nd:YVO4 Verdi laser (Coherent, Inc., Santa Clara,
CA, USA) operating at 532 nm was used for excitation, illuminating the entire field of
view in epi-illumination and dark-field configurations. A secondary objective relayed the
inverted image through the emission filters (RG695 glass absorption filters) and onto the
CCD. The dark-field illumination utilized excitation light coupled through parallel fibers
(250 µm-core multimode) surrounding the perimeter of the lens (inset, Fig. 2.9B). For epiillumination, fiber-coupled excitation light was collimated with a two-element condenser
lens (Melles Griot Inc., Carlsbad, CA, USA) and reflected into the GRIN lens by a dichroic
mirror (long pass >545 nm, Chroma Technology Corp., Rockingham, VT, USA). For epiillumination, laser light was reflected from a vibrating mirror prior to fiber-coupling in
order to reduce speckle.
b
EF
Laser
CCD
RO
RO
EF
BS
GRIN
CCD
Laser
a
VM
CL
IF
GRIN
Mouse
Figure 2.9: Optics setup for, a, epi-fluorescence and, b, dark-field illumination schemes.
Inset in b, detail of perimeter fiber illumination. BS, beam splitter; CL, collimating objective; RO, relay objective; EF, emission filter; VM, vibrating mirror; IF, illuminating
fibers.
30
Images were recorded with a MiCam 64 × 96 pixel 14-bit CCD camera (Brain Vision,
Inc., Tsukuba, Japan) with a pixel area of 12 µm2 and frame rates above 1 kHz. The 16µm spatial resolution was limited by the CCD pixel size. The maximum integration time
per frame was 3 ms (333-Hz frame rate).
2.5 Multifocal Multiphoton Microscope
A multifocal two-photon microscopy was used for anatomical and functional imaging of
the mouse cortex in vivo and rat thalamocortical slice preparations Chapter 7). This section
describes the device optical configuration and principles of operation.
2.5.1 Background
Two-photon laser scanning microscopy [49] has become a powerful tool in the field of
neuroscience because it enables true three-dimensional imaging and because the longer
wavelength excitation wavelengths (NIR) can penetrate deeper into highly scattering tissues. The conventional one-beam technique, however, requires “step-and-stare” scanning
a focused laser beam across a sample to generate a full image and is not optimal for generating fast functional images of many pixels in the field of view simultaneously. The
technique is further limited when used in conjunction with fluorophores possessing small
two-photon absorption cross sections, and still further limited when looking for small
functional signals using such fluorophores.
The concept of scanning multiple beams with a single multiphoton microscope was
31
first introduced by Hell and colleagues [8, 20, 60] as a method of increasing the speed
of conventional two-photon microscopy. By scanning multiple excitation points simultaneously, the speed of image acquisition is increased. There have been three main techniques for multiplexing a single excitation laser beam: (1) passing an expanded laser
beam through a rotating microlens-containing disk (similar to a Yokogawa microlens disk)
which scans a sample in two-dimensions over the course of rotating [20], (2) passing an
expanded laser beam through a stationary microlens array and scanning the excitation
pattern over a sample via galvonometric mirrors [103], and (3) multiplexing a single unexpanded excitation laser beam by progressively dividing the beam through a system of
beamsplitters [136]. All of these multifocal techniques require that the detected fluorescence must be re-focused onto a multiple-pixel detector in order to spatially separate the
fluorescence signals attributed to the different excitation foci. The magnification of the
re-focused fluorescence image of the foci must be sufficiently large in order to avoid spatial overlap of the fluorescence signal, mostly due to scattering. Such overlap leads to
cross-talk between the foci. A detailed analysis of the relationship between foci distance
and cross-talk in multifocal multiphoton microscopy is presented in [60].
The technique has since been employed for fast full-frame imaging of highly-scattering
tissue samples [103], and for monitoring functional calcium signals in the visual motionprocessing area in the brain of the blowfly (Calliphora vicina) [111] and in slices of the
developing mouse hippocampus [45].
32
2.5.2 Overview of system optics
steering
mirror
Modelocked Ti:Al2O3
laser
iris
Beam expansion
optics
Microlens array
Demagnification
optics
Emission
Filter
EMCCD
Dichroic mirror
folding
mirror
Objective
(20X, NA = 0.98,
water immersion)
X-Y scanning galvo
mirrors
Objective
translator
Figure 2.10: Schematic of optical components in the multifocal two-photon microscope.
The system features a microlens array as the beam multiplexing element, and employs a
cooled EMCCD camera as the multianode two-photon fluorescence detector. Two-photon
excited fluorescence is detected in epi-fluorescence mode.
Figure 2.10 displays a schematic of the basic optical components in the multifocal twophoton microscope used in the experiments described in Chapter 7. Section 2.7 includes
a complete list of lens manufacturer’s part numbers and focal properties.
Excitation laser light from a mode-locked Ti:Sapphire laser (Chameleon, Coherent
Inc., Santa Clara, CA, USA) operating at 90 MHz was guided into the entrance aperture
of the microscope through a system of steering mirrors. Upon entering the microscope
housing (shown in a photograph in Figure 2.12), the beam was expanded (×5) by the
combination of a diverging and converging lens (all glass lenses except for the objective
33
were purchased from Melles Griot, Inc., Carlsbad, CA, USA).
A microlens array was employed to generate multiple excitation beamlets. Following
beam expansion, the expanded beam was sent through a stationary array of 100 × 100
microlenses (focal length = 18 mm, Adaptive Optics Associates, Inc., Cambridge, MA,
USA) of which typically 4 × 4 microlenses were fully evenly illuminated by the expanded
excitation beam.
Figure 2.11: Code V ray-tracing rendering of an expanded, collimated excitation laser
beam being split into several beamlets and focused at the focal point of a microlens array
(f = 17 mm). While the focal length, microlens spacing, and aperture sizes are all faithfully incorporated into the computer model, specific glass types and thicknesses are likely
not the same as those of the Adaptive Optics sample array used in this device. Code V
optical modeling software (Optical Research Associates, Pasadena, CA, USA) was used
to model and optimize most of the main optical systems in the multifocal two-photon
microscope described in this thesis.
The focused beamlets then passed through a collimation lens (all converging lenses
except for the objective and demagnification optics were achromatic doublets), through
a dichroic mirror (R < 700 nm, Chroma Technology Corp., Rockingham, VT, USA),
and into the two-axis galvanometer scanning mirror assembly (Cambridge Tecnhology,
34
Inc., Lexington, MA, USA) at the point where the collimating, but converging, beams
converged at the optical axis. The redirected beamlets were then relayed again through
an achromatic doublet and then collimated a final time by a doublet acting as a tube lens
for an infinity-corrected water immersion objective (Olympus XLum PlanFl, 20×, NA =
0.98) mounted on a computer-controlled objective translator (P-725, Physik Instrumente,
L.P., Aubern, MA, USA). Figure 2.12 displays the inside optics and scanning mirrors of
the multifocal device (the top dust covers are removed).
Figure 2.12: In this photograph, the top dust cover panels are removed to reveal the multiplexing, relaying, and scanning optics.
The emitted fluorescence signal was detected in epi-fluorescence1 mode back through
1
The epi-fluorescence detection mode here is technically “inverted” compared to the conventional excitation / emission wavelength relationships. In “conventional” epi-fluorescence detection, the dichroic
mirror transmits higher-energy (shorter wavelength) excitation light, and reflects (toward a detector) lowerenergy (longer wavelength) fluoresced light. Here, the excitation wavelength is longer than the emission
wavelength.
35
the same objective, after which the fluorescence light essentially re-traced (in reverse) the
same steps that the excitation beam took through the optical components, back through
the scanning mirrors up until the dichroic mirror. Approaching the dichroic mirror from
the opposite direction of the excitation light, the fluorescence signal was reflected at a
right angle to the incident excitation light. A detailed ray-tracing diagram for one off-axis
beamlet is shown in Fig. 2.13.
Emission Filter
Galvo Mirrors
Objective
Dichroic Mirror
Microlens
Array
Figure 2.13: Ray-tracing diagram for a single off-axis beam in the multifocal two-photon
microscope. Only optics post microlens array are displayed (i.e. beam expansion optics
and entrance aperture not shown). The red rays indicate (NIR) excitation light from a
modelocked Ti:Sapphire laser. Green rays indicate fluorescence (colors are not necessarily indicative of typical fluorescence spectra). Following excitation on the sample plane,
fluorescence emission re-traces its steps back through the post-galvo optics, is then descanned by the galvo mirror, and is finally re-directed by the dichroic mirror toward the
detector. Prior to detection, the beam passes through demagnification optics to increase
the number of foci that fit on the CCD chip.
Finally, prior to being focused onto the detector chip of a cooled electron-multiplying
CCD (EMCCD) camera (Andor iXon DU860, Belfast, Ireland), the fluorescence signal
was passed through a series of demagnification optics (to allow the re-focused fluorescence
foci pattern to fit on the EMCCD detector chip) and filtered through an emission filter (T <
750 nm). Figure 2.14 details the specific lens configuration employed for demagnification.
36
The lens system was modeled and optimized using Code V optical modeling software.
LAO 083
detector plane
LAO 019
LMP 001 (X2)
Figure 2.14: Code V modeling of a demagnification optical lens system for fitting more
of the re-focused emission foci on the EMCCD detector. The part numbers in the figures
are the Melles Griot (Carlsbad, CA, USA) lens part numbers. Code V software (ORA,
Pasadena, CA, USA) has a built-in lens library that includes all of the Melles Griot catalog
lenses.
2.5.3 Galvanometer scanning mirrors
The lateral (X-Y) beam scanning device in the multifocal two-photon microscope consisted of two galvanometer moving-magnet actuator mirrors (Model 6220, Cambridge
Technology, Lexington, MA, USA) controlled via two MicroMax Series 671 Single Axis
boards (Cambridge Technology). A basic galvanometer consists of a central rotating coil
element mounted in a permanent magnetic field. Current flowing through the moving coil
element causes the central element to rotate about its fixed axis of rotation such that the
magnetic dipole fields align. A spring device mechanically damps and limits the moving
coil motion. In this way, the moving coil element deflection angle can be characterized
as a function of input current. In the galvanometer-based scanning mirror, a mirror is
mounted on the moving coil element. Some of the device specifications for the Model
37
6220 scanning mirror and Series 671 Control board are tabulated in Table 2.2.
Model 6220 Galvanometer Mirror
Mechanical Specification
Value
Tolerance
Rated Excursion, Rotor
±20
minimum
Optical Aperture
5
–
Coil Resistance
3.0
±10%
Coil Inductance
160
±10%
Maximum RMS Current
2.6
maximum
Maximum RMS Power
30
maximum
Repeatability
8
maximum
Model 671 Single Axis Controller Board
Mechanical Specification
Value
Tolerance
Command Input Scale Factor
0.50
–
Analog Command Input Range
±10
maximum
Position Output Scale Factor
0.50
–
Input Voltage Requirements
±15 to ±28
–
RMS Drive Current Limit
10
peak
Units
degrees
mm
Ohms
µHenries
Amperes
Watts
µradians
Units
Volts/◦
Volts
Volts/◦
Volts DC
Amperes
Table 2.2: Mechanical and electrical specifications and input parameters for Cambridge
Technology Model 6220 galvanometer mirrors and MicroMax 671 Control board
2.5.4 Multifocal scanning strategies
Generating a continuous background image in single-point scanning microscopy simply
requires scanning the excitation beam in an evenly-spaced 2D grid of points. If knowledge of the image scale is not crucial to the application (e.g. if the physical scale of
all observed features are well-known a priori), the scanning point location as a function
of galvanometer (or other beam steering device) voltage need not be well characterized.
To take advantage of the multiplexing capabilities of a multifocal scanning microscope,
38
though, the sub-regions scanned by each excitation point in the focal plane must be reassembled (or “patched together”) by a computer following data acquisition. Scanning a
sample by moving a pattern of multiple excitation points thus requires special consideration even for generating continuous background images. If the scanning device deflection
angle is not calibrated, the reconstructed images may exhibit repeated discontinuities of
either overlapping or “missing” regions between the sub-images.
The ideal utilization of a multifocal scanning microscope is, in fact, not to scan the
foci at all but to have a sufficiently large number of foci spaced very close together in
the focal plane such that a high-resolution image can be acquired with a single exposure.
In practice, this is typically impractical in fluorescence microscopy because positioning
excitation foci very close to each other leads to excessive 2D cross-talk between foci.
Such cross-talk is primarily caused by blurring of the fluorescence origin due to scattering
in the sample tissue. A secondary technical cause for the cross-talk is the difficulty of
spatially separating the fluorescence foci images on a detector of limited area. Egner
and Hell found that lateral and axial resolution in a multifocal multiphoton microscope
system are optimal when the excitation foci are spaced at least 7λ apart from each other
[60], where λ is the excitation wavelength. Considering λ ≈ 1 µm for two-photon and
second-harmonic microscopies, foci spacing of 7 µm is not sufficiently high-resolution for
imaging neurons (e.g. excitatory cells in the neocortex are ∼ 10 µm diameter). Scanning
of the foci pattern across the inter-focus distance is therefore required to resolve even
cellular-scale details. Furthermore, nonlinear microscopy requires high excitation powers
and short pulse-widths at each excitation point. Detectors and collection optics in the
39
average two-photon scanning microscope require on the order of ∼ 10 mW of average
power to generate appreciable signals (assuming fluorophores possess average two-photon
absorption cross sections, ∼ 50 GM). Given the fact that ultrafast pulsed near-infrared
lasers offer 1–3 W of average power, it is not currently practical to simultaneously scan
many more than 100 foci.
We therefore consider the basic raster-scanning schemes for imaging neurons using a
square pattern of 4 excitation beams separated by at least 7λ in the focal plane. The three
scanning routines used in experiments described in Chapter 7 are presented in Figure 2.15.
The maximum field of view will be considered to be equal to the square with sides 4× the
inter-focus distance (indicated by the surrounding dotted square line in Fig 2.15A), unless
the excitation foci pattern is to be scanned in steps which are multiples of the full row
width of all the foci (which was not the case for the experiments described in this thesis).
To raster-scan a single continuous background image at a spatial resolution of ∆L, the
foci points are stepped N = L/∆L points in the “fast” axis (see Fig. 2.15A) for every
point on the “slow” axis, where L is the distance between the foci on the object plane. To
generate a square image, the fast and slow axes are scanned a total of N points.
To generate square higher-resolution images in the same amount of total scan time
(i.e. assuming the integration or “stare” time per point, ∆t, is held constant and that the
total acquisition time, N 2 × ∆t, where N 2 is the total number of scanned points – ignoring the small time it takes to move the scanning mirror between points–is held constant
as well), N is held constant while ∆L is reduced. When the sub-regions are combined
post-acquisition, this has the effect of generating a higher-resolution, but discontinuous
40
A
B
"fast" axis
L
"slow" axis
L = 0.5 V, ∆L = 0.005 V
L = 1 V, ∆L = 0.01 V
Volt / µm relationship: 0.01V = 1 µm
C
Example:
L1 = 0.1 V
∆L = 0.01 V
L1
Figure 2.15: Scanning schemes for multifocal two-photon microscope. A: Full-field scanning. In this scanning routine, a continuous background image is formed. B: Sub-image
scanning. This routine, used for higher-resolution scanning without sacrificing time resolution, yields a series of discontinuous sub-images. C: Line-scanning. The excitation foci
are stepped a specified number of points along a line of an given length (both are user
inputs).
41
background image (see Fig. 2.15B).
Full-frame functional imaging of fast events (e.g. membrane potential or intercellular
ion concentration) requires that N 2 ×∆t be small compared to the timescale of the activity
being measured. This imposes an upper limit on the N which is a function of signal size,
fluorophore two-photon absorption cross-section, excitation power, and system collection
efficiency (see Chapter 6 Section 6.2.6 for a sample calculation). In practice, to achieve
the highest signal-to-noise ratio in optically-measured functional measurements, it is most
feasible to produce a high-resolution background image and then to perform functional
imaging on a small number of points within the larger field of view.
The very simplest example of this technique would be to monitor just one point on the
object plane, also called “beam-parking.” Because beam-parking often causes excessive
photodamage and photobleaching of the sample at the excitation points, this scanning routine is not always feasible. Rather than this “zero-dimensional” scanning, one-dimensional
line-scanning is by far the most common scanning routine of this species (Fig. 2.15C). In
this scanning routine, N points are scanned along a line segment of length L1 at step resolution ∆L such that N × ∆t is small compared to the functional activity. This allows a
“relaxation time” of N × ∆t for each point on the line, assuming the beam returns to scan
the first point immediately after scanning the last point in the line.
42
2.5.5 Device control and data acquisition
The multifocal two-photon microscope system consists of active and passive components.
The main active components are the galvo scanning mirrors, EMCCD camera, objective
translator, and the laser shutter. The Coherent Chameleon Ti:Sapphire laser is an active
self-modelocking and climate-maintaining device. Laser maintenance functions are fully
performed by embedded systems and the only necessary input is laser shutter control. The
passive components include most of the microscope optics. Following an initial manual
alignment procedure (see Section 2.7), control and synchronization of all active devices
are performed by custom software written in LabVIEW 8 (National Instruments Corp.,
TX, USA) visual programming environment1 running on a desktop PC machine with Microsoft Windows XP operating system.
The data acquisition and system control software consists of a main LabVIEW program (file ending with suffix “.VI” for “virtual instrument”) and a library of other programs (sub-.VIs) that are called by the main program. A simplified flowchart displaying
the basic functions of the main program and one particular scan-routine sub-VI program
is shown in Figure 2.16. The main program (represented by the middle column of the
flowchart in Fig. 2.16) initializes all active components and executes the default “wait”
mode while-loop. If no command to begin scanning (in any mode) is given, the program
continuously executes this main while-loop. During this wait time, the EMCCD camera
temperature is slowly stabilized (the EMCCD temperature during scanning is usually set
1
N.B. While LabVIEW 8 utilizes “Data MX” functions for data acquisition and hardware control, earlier
versions of LabVIEW utilize a completely different library of functions. Programs (.VI files) are generally
neither reverse- nor forward-compatible among LabVIEW versions.
43
to −75◦ C). The software has four scan modes: (1) single-frame raster scan, in which the
user inputs the desired scanning dimensions and resolution, (2) line-scan mode, which
also triggers external electrophysiology equipment for functional imaging, (3) volumescan mode, which takes a series of raster-scanned background images over the course
of a user-defined depth and depth resolution, (4) time-series scan mode, which repeatedly takes a background image a user-defined amount of times. One particular scanning
operation (sub-VI), a single-frame scan, is detailed in Fig. 2.16 (right column). After
a series of initialization steps, the sub-VI executes its main scanning while-loop, which
ends after all points have been scanned. Acquired data is then parsed, re-shaped, and
displayed using custom-written Matlab (MathWorks, Natick, MA, USA) M-files executed
within LabVIEW mathscript-nodes. The other three scanning functions follow the basic
sequence structure of this sub-VI, similarly featuring initialization steps, a main scanning
while-loop, and image-reconstruction steps.
Digital and analog synchronization communication signals are read/written by a multifunction data-acquisition card (NI PCI-6052E, 333 kS/s, 16-Bit, 16-Analog-Input Multifunction DAQ, National Instruments Corp., TX, USA) installed in a PCI slot of the device
PC computer. Communication and control of the EMCCD camera and objective translator are performed by individual PCI cards from each manufacturer. All devices, regardless
of hardware, can be controlled via LabVIEW. The main scanning while-loop involves
synchronization of scanning galvo mirrors, EMCCD camera, and laser shutter. Figure
2.17 shows the relevant communication signals occurring at each device during the raster
44
Generate array of
scan point
voltages
Start Program
Set EMCCD
acquisition
params
Device
Initialization
(Camera ON,
Objective
translator
ON,Laser shutter
CLOSED)
Initialize Analog
Channels
(for galvos)
Get EMCCD Temp
Laser shutter
OPEN
Set EMCCD Temp
Start EMCCD
acquisition
Get objective pos
scan single point
Set objective pos
frame acquired?
YES
YES
OFF command?
All points
scanned?
NO
NO
YES
Do scan routine?
NO
YES
Laser shutter
CLOSED
Extract foci data
from EMCCD
images
Close devices
Reshape image
data to form
combined image
End Program
Save and Display
Image
Figure 2.16: Flowchart for device control and data acquisition.
45
scanning of four neighboring lines. This scanning routine is strictly for the purpose of illustrating the synchronization principles, as scanning a 4 × 20 pixel image is rarely useful.
The DU860 EMCCD camera is operated in “kinetic series” acquisition mode, in which a
specified number of frames are acquired at a set frame rate before all the data is written
to disk. In this mode, the DU860 outputs a regular series of “high” pulses on the “Fire”
digital output line corresponding to integration time. This train of output pulses acts as the
synchronization “clock,” where each digital high signal serves as a “point done” flag for
the scanning while-loop.
Scanning with multiple excitation points necessitates being able to individually keep
track of the fluorescence returning from each of these foci. When the excitation pattern is
scanned, the fluorescence “image” of the excitation pattern shifts as well (since the fluorescence only originates from the focal volumes of the scanning points). This movement of
the fluorescence signal must somehow be accounted for during image reconstruction (i.e.
by knowing which pixels correspond to which foci as a function of time). In practice this
is very difficult and imposes additional limitations on the detector’s physical dimensions
and speed (mostly because binning is no longer an option).
An easier alternative, employed in this device, is to “de-scan” the epi-fluorescence signal, whereby the fluorescence signal re-traces its steps back through the scanning mirrors
prior to entering the detector. De-scanning yields a stationary foci pattern on the detector
plane, simplifying reconstruction.
Figure 2.18 illustrates the basic principles of multifocal image reconstruction. The
product of any scanning routine is a series of EMCCD frames (such as the ones at the
46
EMCCD line out
TTL
pulses
-10
0
10
20
30
40
50
60
70
80
90
50
60
70
80
90
60
70
80
90
60
70
80
90
time (ms)
laser shutter
open
closed
-10
0
10
20
30
40
time (ms)
"fast" galvo axis
voltage
-10
0
10
20
30
40
50
time (ms)
"slow" galvo axis
voltage
-10
0
10
20
30
40
50
time (ms)
Figure 2.17: Device synchronization timing sequence for raster-scanning using an EMCCD camera detector. The “Fire” digital output line of the Andor iXon DU860 EMCCD
camera goes high during frame acquisition, and low during the readout / wait time. Each
scan point (most easily visualized as the voltage “steps” on the fast galvo mirror voltage)
is followed by a “high” signal flag from the EMCCD.
47
Frame 1
(t = 1 ms)
Frame 2
(t = 2 ms)
Frame 3
(t = 3 ms)
Figure 2.18: Basic principles of image reconstruction of multifocal data. De-scanned epifluorescence signal is re-focused onto the EMCCD detector to produce images such as
the top three frames. For each time point (each scan step of the “fast” scan axis), the
peak intensity at each focus is used as the pixel value for the corresponding pixel in the
reconstructed image (where each sub-region corresponds to one of the excitation foci).
48
top of Fig. 2.18), each corresponding to a particular scanned point on the sample. By
comparing the time-stamp on each EMCCD frame with the indexed list of scan-points
generated by the scanning software (see right column of Fig. 2.16), the intensity value for
each fluorescence focus is attributed to a particular pixel in its corresponding sub-image.
2.5.6 Advantages and challenges in multifocal microscopy techniques
Multifocal scanning microscopy techniques offer three major advantages for imaging fast
neuronal activity using functional fluorescent dyes. First, as discussed in previous sections, scanning multiple points enables faster scanning by a factor of the number of foci
employed. Secondly, full-field background images consisting of N 2 points scanned with a
standard 1-beam scanning device can be scanned with a multifocal device, at the same fullframe scan rate, with longer integration (“stare”) time per point. The factor of increase in
integration time will be equal to the number of excitation foci. This is particularly useful in
applications which employ fluorophores with moderate to low two-photon cross-sections
(as is the case for most voltage-sensitive dyes). Lastly, for line-scanning applications,
scanning multiple points generates a grid of line-scans distributed over the entire field of
view, permitting 2D spatiotemporal tracking of fast activity.
Nonetheless, several challenges still remain for multifocal techniques. Most of these
challenges are technical and can be remedied via choice of alternative devices (as they become available or more affordable). One of the biggest limitations for multifocal scanning
49
techniques is the lack of suitable commercially available detectors of sufficient sensitivity
and speed. Photomultiplier tube (PMT) arrays boast extremely high gain levels (∼ 107 )
and nanosecond time resolution, but have relatively small quantum efficiencies (often
<5% at emission wavelengths >550 nm). This makes small signals in shot noise limited measurements exceedingly difficult to measure at long emission wavelengths. Backilluminated electron-multiplying CCD (EMCCD) detectors feature high gain (∼1000) and
very high quantum efficiencies (> 90% at wavelengths >550 nm) and are currently one of
the best detector options for this application. Full-frame acquisition rates for these EMCCDs (typically 100s of Hz), however, fall somewhat short of the ideal temporal resolution
desired for use as scanning system detectors. Binning and region-of-interest acquisition
modes for these EMCCD detector chips increase the acquisition rates (sub-frame acquisition rates can reach 1000s of Hz), but complicate image reconstruction (see Section
2.5.5). Additionally, high-speed EMCCD chips have small physical dimensions (< 4 mm
side length) making it difficult to spatially separate re-focused multifocal fluorescence and
avoid lateral cross-talk.
Scattering poses a limitation to multifocal imaging in scattering tissues. As optical sections are acquired from deeper depths in scattering tissue, the direction of epi-fluorescence
photons originating from the “point sources” (the excitation foci volumes) becomes increasingly randomized. This causes the de-scanned fluorescence foci image to become increasingly diffuse (i.e. the width of the focused fluorescence points increases), decreasing
the peak count rate at the foci maxima and increasing fluorescence overlap (cross-talk).
Additionally, de-scanning the epi-fluorescence signal (see Section 2.5.5) attenuates the
50
fluorescence signal by imposing additional reflective surfaces and apertures.
An optical engineering difficulty introduced by scanning multiple beams simultaneously is the added complexity of a system with significant off-axis beam components.
Rays most distant from the optical axis on lens surfaces suffer from the most significant
aberration effects (coma, principally). As evident from Fig. 2.10, collimated or focused
off-axis beamlets pass through nearly all of the optics in the microscope. While employing
non-focusing beam multiplexing techniques (e.g. via a system of beamsplitters [136]) can
reduce much of the coma, curvature of field due to the objective (and emission demagnification optics, if present) must still be accounted for.
2.6 Appendix: Computational optical sectioning and deconvolution techniques
One-photon fluorescence microscopy provides the benefit of relatively large signal levels
(assuming the fluorophore possesses a reasonably high quantum efficiency, i.e. q > 10−3 )
and relatively simple optical design compared with two-photon microscopy. Depth resolution suffers in the former case, however, as a result of excessive fluorescence signal
originating from regions above and below the focal plane position. This is the “signalsummation” problem, which leads to uncertainty in the depth of the signal origin.
The degree to which out-of-focus fluorescence in the object plane contributes to an
image can be quantified by measuring the point spread function (PSF) of the optical system. The PSF measured at a particular degree of defocus is the spot image of a point
51
source. One solution to the “signal summation” problem is to employ computational optical sectioning, or “deblurring”, methods that utilize experimentally measured system PSFs
to remove components of fluorescence in the image originating from neighboring object
planes. Such techniques have been effectively employed to achieve quasi “confocal” performance using conventional 3-D epifluorescence microscopy [3, 92] and to remove outof-focus fluorescence signal in functional voltage-sensitive dye imaging studies [171,197].
2.6.1 Background theory
A simplified two-dimensional imaging geometry is shown in Figure 2.19 (this background
section follows the description in [31]).
z
x'
image plane
y'
Image Plane
gfoc'(x', y', zfoc')
Lens system
Object Plane
f (x, y, zobj)
object
ct plane
zobj
Focal Plane
focal plane
zfoc
Actual optical Reverse projection
of image plane =
density = 0
x
Origin
y
gfoc (x, y, zfoc)
Figure 2.19: Simplified geometry of two-dimensional specimen imaging where the optical
density function of the sample, f (x, y, z) is zero at all depths except for the object plane
at depth z1 .
Here, the object plane is at depth z = zobj and the focal plane is at depth z = zf oc . In
52
this scenario, there is no object optical density distribution (e.g. fluorescence, absorption,
reflectivity, etc.) except at the depth zobj , i.e.
f (x, y, z) = fobj (x, y)δ(z − zobj ),
(2.3)
where f (x, y, z) is the optical density distribution throughout all space on the object
side of the lens system. The resulting image is the function gf0 oc (x0 , y 0 , zf0 oc ), where the
primed variables refer to image plane coordinates following transformation by the lens
system. While there is no actual optical density at the focal plane (unless zf oc = zobj ), we
define gf oc as the “reverse projection” of the image onto the focal plane, which amounts
to essentially “what the image plane sees when focused at z = zf oc .” This is nonzero
because of the point spread function (PSF) of the optical system. This definition is useful
because it avoids variable transformations (i.e. x −→ x0 , y −→ y 0 , z −→ z 0 are no longer
necessary).
In terms of the PSF, h(x, y, zobj − zf oc ), which is the PSF at the defocus amount zobj −
zf oc , the image “reverse projection” is defined as the two-dimensional convolution (over x
and y) of the optical density distribution with this PSF, i.e.
Z
∞
Z
∞
gf oc (x, y, zf oc ) =
fobj (u, v, zobj )h(x − u, y − v, zobj − zf oc )dudv.
−∞
(2.4)
−∞
The convolution operation, written in two-dimensions in the RHS of Eq. 2.4, is often
53
abbreviated using asterisk notation (“∗”). For example, in one-dimension, p(x) ∗ q(x) is
the operation
Z
∞
k(x) = p ∗ q =
p(u)q(x − u)du.
(2.5)
−∞
One way of modeling a thick specimen imaging scenario is to envision a sample of
thickness T (in depth, i.e. the z axis) comprised of a series of two-dimensional optical
density distribution “slices” occurring at small intervals, ∆z, over the depth of the sample
thickness. In this case, there are N =
T
∆z
planes of optical density functions.
Focusing through the sample thickness in steps of ∆z , the jth focal plane image,
i.e. the image “reverse projection” when the focal plane is set at depth zf oc = j∆z, is a
summation of Eq. 2.4 over all the sample thickness, i.e.
gf oc (x, y, j∆z) =
N
X
f (x, y, i∆z) ∗ h(x, y, i∆z − j∆z)∆z,
(2.6)
i=1
where 1 ≤ j ≤ N .
For the purposes of computational image processing (particularly when using Matlab),
convolution equations such as Eq. 2.6 can be simplified by replacing the functions with
their Fourier transforms. This is because of the Convolution Theorem, which states that
convolution in one domain corresponds to multiplication in the Fourier domain, and vice
versa. Thus making the substitutions
54
Z
∞
Z
∞
Gj (u, v) = F{gf oc (x, y, z = j∆z)} =
−∞
gf oc (x, y, z = j∆z)e−i2π(ux+uy) dxdy.
−∞
(2.7)
Z
Z
∞
∞
Fi (u, v) = F{fobj (x, y, z = i∆z)} =
−∞
fobj (x, y, z = i∆z)e−i2π(ux+uy) dxdy.
−∞
(2.8)
Z
∞
Z
∞
Hi (u, v) = F{h(x, y, z = i∆z)} =
−∞
h(x, y, z = i∆z)e−i2π(ux+uy) dxdy, (2.9)
−∞
where the notation F{} indicates the Fourier transform operation, enables us to rewrite
Eq. 2.6 as
N −j
Gj =
X
Fi+j Hi .
(2.10)
i=1−j
In these equations, H(x, y, z) is the 2D Fourier transform of the point-spread-function
(PSF), h(x, y, z), and is known as the Optical Transfer Function (OTF). Figure 2.20 compares an experimentally-measured in-focus PSF of a GRIN lens epi-fluorescence imaging
system (see Section 2.4) with the corresponding OTF image generated using Matlab software (MathWorks, Natick, MA). The PSF was measured by imaging a single fluorescent
polystyrene microscophere (Molecular Probes / Invitrogen Corporation, Carlsbad, CA)
55
with a diameter of 0.1 µm. When measuring the PSF of an optical system using a CCD
camera as a detector, it is important that the size of the point source is significantly smaller
than the detection pixel size so that the measured PSF is not dominated by the shape of the
point source itself. In practice, the source size should be chosen to be as physically small
as allowed by the sensitivity of the detector given that small point sources yield small
fluorescence signals [92].
0.5 mm
Figure 2.20: Experimentally measured point-spread-function (PSF) (left) and optical
transfer function (OTF) (right) of GRIN lens fluorescence imaging device (see Section
2.4). The left image is the in-focus PSF which was measured by imaging a single fluorescent polystyrene microscophere (Molecular Probes / Invitrogen Corporation, Carlsbad,
CA) with a diameter of 0.1 µm. When measuring the PSF of an optical system using
a CCD camera as a detector, it is important that the size of the point source is significantly smaller than the detection pixel size so that the measured PSF is not dominated by
the shape of the point source itself. In practice, the source size should be chosen to be as
physically small as allowed by the sensitivity of the detector given that small point sources
yield small fluorescence signals [92].
2.6.2 Nearest-neighbor approximation
These results seem to indicate that the only way to effectively reconstruct the desired
f (x, y, z) utilizing an experimentally determined h(x, y, zobj −zf oc ) would involve making
measurements of gf0 oc (x0 , y 0 , zf0 oc ) at N planes of defocus, and then solving the system of N
56
simultaneous linear equations in N unknowns. To the contrary, approximate methods can
often achieve solutions that are nearly equivalent, especially given that the experimentally
measured PSF is rarely error-free and that there is always an extra noise factor which
cannot be accounted for.
Dropping excess variable notation, Equation 2.6 can be rearranged as
fj ∗ h0 = gj −
−1
X
N −j
fi+j ∗ hi −
i=1−j
X
fi+j ∗ hi
(2.11)
i=1
where the i = 0 term has been taken out of the sum and negative and positive i have been
separated into 2 terms. h0 is the in-focus PSF of the microscope system. The goal is to
somehow approximate fj . Three approximations will help simplify this goal significantly.
First, we assume contributions from neighboring object planes (the last two terms on the
right hand side of Equation 2.11) are primarily low spatial-frequency contributions. Second, we assume the specimen place fj can be approximated as a high-pass filtered version
of image gj such that
fj ≈ gj ∗ k0 .
(2.12)
Third, and finally, we assume that in-focus PSF, h0 , has negligible effect on the image.
With these three approximations, Equation 2.11 can be approximated as
fj ≈ gj −
−1
X
gi+j ∗ k0 ∗ hi −
i=1−j
n−j
X
i=1
57
gi+j ∗ k0 ∗ hi ,
(2.13)
and if we consider only a subset M of these N object planes, Equation 2.13 can be
written
fj ≈ gj −
M
X
(gj−i ∗ h−i + gj+i ∗ hi ) ∗ k0 .
(2.14)
i=1
Finally, if only the planes above and below the focal plane are considered, Equation
2.14 can be reduced to
fj ≈ c1 gj − c2 (gj−1 + gj+1 ) ∗ h1 ,
(2.15)
where coefficients c1 and c2 are empirically determined and reflect the relative contributions of out-of-focus signal from neighboring planes [2].
2.6.3 Example: Nearest-neighbor deblurring of VSD signals in mouse
barrel cortex in vivo
As an illustrative example of nearest-neighbor de-blurring, we describe the procedure and
results of removing out-of-focus fluorescence contributions from voltage-sensitive dye
signals recorded in the mouse barrel cortex in response to cortical electrical stimulation
(see Section 2.4 for device details and Chapter 4 for experimental protocol details).
Eq. 2.15 involves convolution operations, which can require significant processing
power when applied to a series of many images. Taking the Fourier transform of Eq. 2.15
yields
58
Fj
Gj
Hj
Fj = c1 [Gj - c2 (Hj Gj+1 + Hj Gj-1)]
0.5 mm
Fj
fj
Figure 2.21: Graphical illustration of nearest-neighbor de-blurring procedure. Here, Eq.
2.16 is shown with sample images for the main functions. Gj is the 2-dimensional Fourier
transform of a measured ∆F/F image at the surface of the mouse barrel cortex in vivo.
The depth spacing, ∆z, was ∼12 µm here. Hj is the optical transfer function (OTF) which
is the Fourier transform of the experimentally measured point spread function (PSF) at the
jth plane. In this figure, Hj is the Fourier transform of the PSF shown in Fig. 2.20. The
constants c1 and c2 are high-pass filtering constants and values chosen here were in the
range of c1 ≈ 0.9 and c2 = 0.25 − −0.45.
59
Fj ≈ c1 [Gj − c2 (Hj Gj+1 + Hj Gj−1 )],
(2.16)
which is an easily computed convolution-free equation. This equation is illustrated
graphically in Fig. 2.21. Here, Hj is the optical transfer function (OTF) at the jth plane
of focus. To generate Fj , one needs to experimentally measure the PSF at the jth plane of
focus, the image at j (gj ), the plane above (gj+1 ), and the plane below (gj−1 ). Generation
of the de-blurred image fj then simply requires taking the inverse 2-dimensional Fourier
transform of Fj (see Fig. 2.21, bottom).
Figure 2.22 shows the effects of nearest-neighbor computational de-blurring on a timeseries of ∆F/F images acquired during cortical electrical stimulation of the mouse barrel
cortex in vivo. The de-blurring led to an apparent increase in signal-to-noise of ∆F/F
responses to individual electrical stimuli (4.5 de-blurred, vs. 3.5 un-corrected).
2.7 Appendix B: Alignment procedure for multifocal twophoton microscope
Here we describe the alignment procedure for the multifocal two-photon microscope described in previous sections. Refer to Figure 2.23 for component names and configuration
throughout this description.
WARNING!! WAVELENGTH-SPECIFIC PROTECTIVE EYEWEAR MUST BE
60
A
2
B
X10-3
5
0
-5
-2
X10
-4
red = deconvolved
blue = uncorrected (raw)
0
-10
0.5 mm
-3
0
120
360
240
time (ms)
480
C
D
uncorrected (raw)
with nearest-neighbor
subtraction
Figure 2.22: Effects of nearest neighbor de-blurring on surface VSD ∆F/F measurements in mouse barrel cortex in vivo. A: Sample de-blurred image from a time-series of
∆F/F images taken during the course of 4 electric stimuli directed into the cortex near
the imaging location. This sample fluorescence image was taken while focusing at the
surface of the cortex. B: Traces of ∆F/F through the entire time series (data represents
the average of ∼30 trials) for both uncorrected (blue trace) and de-blurred (red trace, via
nearest-neighbor deconvolution) images. The nearest-neighbor de-blurring led to an increase in signal-to-noise from 3.5 (blue) to 4.5 (red). Traces represent the pixel-averaged
activity from all pixels within the area delimited by the dotted white circle in (A). C: Montage of uncorrected (raw) ∆F/F images (arranged left-to-right, top-to-bottom) during the
course of electric stimulation in the mouse cortex. D: De-blurred montage of ∆F/F images. Note that neuronal responses to individual stimuli are easier to identify visually.
61
WORN AT ALL TIMES! The modelocked Ti:Sapphire laser is a Class IV laser (the highest classification in the laser safety scale) and is capable of causing immediate damage
to the eye or skin! Even scattered light (off of lens surfaces, beam picks, prisms, mirrors, filters, etc) can cause immediate and permanent injury. Particular precaution must be
taken at tuning wavelengths beyond the visible (particularly > 825 nm for the modelocked
Ti:Sapphire laser).
2.7.1 General alignment of excitation laser beam throughout microscope
1. The key on the front panel of the Chameleon Ti:Al2 O3 pump laser should be in
STANDBY mode and the pump laser should be operating. If the laser is not on
(i.e. if the pump laser is off and there is no message on the front LCD display), the
laser must be turned on according to the Chameleon manual. Do not attempt to turn
the laser on without reading and understanding the full system power-up procedure.
Running the laser without the necessary circulation systems (water and air) can (and
almost definitely will) damage the laser.
2. Tune the Chameleon to 750 nm (so that the beam is visible) and turn the key on
the front panel of the Chameleon pump laser to “ON”. The laser should modelock
within 10–15 minutes, and the power should stabilize within 30 minutes.
3. Turn on the PI objective translator control box. Turn on the power switches for the
two galvo axis boards. NOTE: The galvo box generates a wealth of electric noise.
62
Modelocked Ti:Al2O3
laser
M4
M3
periscope
assembly
M1
M2
BB
iris
L1
L2
Wall Set 1
MLA
L3
L5
L4
EMCCD
BSA
M7
M5 (top mirror)
M6 (bottom mirror)
L6
Figure 2.23: Optical alignment layout for the multifocal two-photon microscope. This
optical layout represents the current microscope setup on an L-shaped air table in an in
vivo electrophysiology experiment room. Mirrors are denoted by the “M” in labels; lenses
by “L”. BB = beam block; MLA = microlens array; BSA = beamsplitter assembly; the
periscope assembly consists of mirrors at two height levels (optical table level and microscope level). Lens elements (except for microarray) are all from Melles Griot (Carlsbad,
CA, USA). Lens catalog part numbers are: L1 = LDK 001, L2 = LAO 134, L3 = LAO
083/76, L4 = LAO 083, L5 = LAO 119/77. Mirrors M1–M4 are coated IR mirrors from
Edmund Industrial Optics (Barrington, NJ, USA). The two short arrows pointing to corners indicate screw locations that will secure/release Wall Set 1.
63
It may be necessary to connect the galvo box to a different power circuit to avoid
heavy line noise in sensitive electrophysiological measurements.
4. Open the main scanning labview .vi file. Press the RUN button and set the camera
temperature to −75◦ C.
5. Using an IR viewer scope or IR viewing card (the use of an IR viewer is highly
recommended and may be necessary for some steps), verify that the beam reflecting
off of M1 is fully blocked by the beam block (BB). If there is significant stray light,
or if the beam is obviously directed at one of the walls, M1 is grossly misaligned
and must be fixed.
6. Using the vertical and horizontal positioning controls on the M1 mount, direct the
beam onto the center of M2, and from M2 onto the center of M3, and so on up the
periscope in numbered order. Depending on how accurately the mirrors are aligned,
a centered beam may not be possible for all mirrors.
7. Make sure the galvos are pointed to the X,Y coordinates (0, 0). This can either
be done by directly communicating with the galvos via the National Instruments
Measurement and Control program, or, using the multifocal scanning LabView VI
program, by putting the camera into continuous acquisition mode.
8. This step is only necessary if the system is being completely re-aligned from scratch
(however we assume that the galvos have not been removed). Remove Wall Set 1
by removing the 4-40 screws from the corners indicated by the arrows in Fig. 2.23.
64
Remove all optics between the first iris and the galvo scanning mirrors (L1, L2,
MLA, L3, and BSA). Open the iris fully. Using an IR viewer, adjust the vertical
and horizontal controls on M4 to direct the beam onto the middle of the top galvo
axis, such that the beam does not “leak” out of the mirror aperture and is completely
reflected at a 90-degree angle in the plane of the table. Return the removed optics
to their holders on the dove-tail rail mount. Make sure that the beam more or less
passes through the center of each optic up to the galvos. Close the entrance iris and
adjust the iris hight to center around the beam.
9. If L4 is not in its correct place, it should be equidistant with L3 from the galvo
mirrors, since they are essentially the same lens. The L5 position should be finetuned to ensure that the resulting beamlets are collimated and do not focus at some
far point. This is because the objective is “infinity-corrected” and thus designed to
accept parallel rays as the input. NOTE: If lens distances have been corrupted, they
can be corrected by looking up the focal length of each Melles Griot lens and using
Fig. 2.23 as a guide for when beams need to focus / collimate. Lens distances can
be measured with metric dial calipers (focal distances are typically quoted in mm).
10. Regardless of beam position on M7, the crucial thing is that the beams must exit the
objective symmetrically. This can be verified by placing a white sheet of paper flat
approximately 5 inches below the objective and observing the focal pattern with an
IR viewer. If the pattern is asymmetric, use the horizontal and vertical adjustment
controls on M5. For daily operation, it is typically sufficient to simply use the M4
65
steering controls to “tweak” the symmetry of the unfocused beamlet pattern below
the objective.
2.7.2 General imaging protocol
1. Tune the Chameleon laser to the peak excitation wavelength of the fluorophore of
interest in the sample to be imaged. Note that the fluorescence signal is directly
proportional to the two-photon cross-section, but proportional to the square of the
average excitation power. Refer to the tuning curve in the Coherent Chameleon
product manual; if the peak excitation of the fluorophore is near one of the extremes
of the tuning range, there may not be sufficient excitation power to merit using that
wavelength.
2. Unless the sample is highly fluorescent, it is advisable to first “find” the focal plane
(in order to create a “foci image”) using a fluorescent test slide or other highlyfluorescent uniform sample. This can be positioned below the objective using a
combination of a manual micrometer-driven stage and the PI objective translator
(the objective translator, however, only has a maximum travel range of 400 µm).
3. Apply a drop or two of DI water to the test sample. Then turn the room lights off.
Advance the vertical stage micrometer until the two-photon excited fluorescence
spot is detected. If the sample is extremely weakly fluorescent (e.g. GFP-labeled
cells, autofluorescence, bulk calcium indicator-loaded cells), it will be necessary to
monitor the fluorescence signal detected by the EMCCD by watching the monitor
66
during Continuous Acquisition mode, making sure that integration time values are
set to ∼3 ms for multifocal mode, and ∼0.5 ms for single-point scanning mode (the
microlens array must be removed in this case). If the sample is highly fluorescent
(e.g. TPA cross-section ≥ 10 GM; quantum yield ≥ 0.5), the two-photon excited
fluorescence will be visible by eye in a dark room.
4. Select “Take Foci Image” in the multifocal scanning program. This will present an
image of the foci pattern (typically 4 × 4). With the computer mouse curser, select
the center points of foci in the following order: top-to-bottom, then left-to-right.
Then press enter. If the foci points have been chosen in a recent experiment and
the foci positions have probably not changed, you can opt to “Revert to Saved Foci
Data.”
5. Image the desired region using the scanning program options.
67
Chapter 3
Molecular Indicators
3.1 Voltage-Sensitive Dyes
Voltage-sensitive dyes (VSDs)1 are organic dyes that typically embed themselves in the
neuronal plasma membrane. They exhibit shifts in absorption and emission spectra in
response to changes in the trans-membrane electric field (see Figure 3.1). Use of voltagesensitive dyes as contrast agents (see [38] and [167] for reviews) offers the possibility
of simultaneously measuring membrane potential at many sites (or cells) non-invasively.
Such a measurement using electrical techniques requires impractically large electrode arrays to be inserted into the tissue.
The use of voltage-sensitive dyes as molecular voltmeters is currently the only optical
technique enabling direct measurements of neuronal membrane potential. The sensitivity
1
also frequently referred to as “potentiometric” dyes
68
Figure 3.1: Shift of absorption and emission spectra in voltage-sensitive dyes as a function
of membrane potential. Reproduced from [25].
of this methodology ranges from ∼ 10 mV, relevant for subthreshold membrane potential dynamics, to ∼ 100 mV, which is associated with action potentials (see Figure 3.2).
Voltage-sensitive dyes have proven to be effective for measuring electrical activity in neurons in vitro [43,80,169,201] and in vivo [82,145,149]. To date, several studies have visualized voltage-sensitive dye responses in three-dimensions using one-photon fluorescence
in vivo [105, 149], and recently gradient-index (GRIN) lens optics and computational optical sectioning techniques have been used to achieve high-speed three-dimensional microscopy with voltage-sensitive dyes in near surface tissues [67].
Figure 3.2: Voltage-sensitive dye absorption response during an action potential in the
squid giant axon. Optical absorption signal (black dots) superimposed over electrical
recording trace (red line). Reproduced from [164].
69
3.1.1 Brief History
Changes in intrinsic optical tissue properties (scattering and birefringence) that accompanying fast, millisecond-scale electrical activity were first reported by Cohen et al. [39].
Shortly thereafter, the method of optically measuring membrane potential by monitoring
fluorescence intensity was introduced by Tasaki et al. [188]. However the fluorescence
signal generated by the dye 8-anilinonaphthalene-1-sulfonic acid (ANS) were small and
averaging was required. Extensive efforts were then put forth to find fluorescent dyes that
yielded larger signals. The fluorescent compound “merocyanine 540” was the first dye
to permit measurement of changes in fluorescence during single action potentials with a
large signal-to-noise ratio (> 101 ); this experiment used the squid giant axon [47].
3.1.2 Physical mechanisms of potentiometric dyes
The physical mechanisms underlying optical changes exhibited by potentiometric dyes are
not yet completely understood. It has been established that these optical changes, which
occur as a function of membrane potential, Vm , are the result of several different possible
physical mechanisms. Spectroscopic measurements of VSDs at varying membrane potential suggest that, at least for some dyes, a combination of multiple physical mechanisms
seems likely [36, 72].
A review by Waggoner [192] described four principal physical mechanisms of fast response dyes2 : dimerization, relocation of aggregation (the so-called “on-off” mechanism),
2
“fast” response dyes respond to membrane potential on the order of microseconds, as opposed to “slow”
response dyes which respond on the order of tens to hundreds of milliseconds.
70
chromophore reorientation, and the Stark effect.
The dimerization, or “rotation-dimer” mechanism, refers to a membrane-potential dependent shift in equilibrium populations of monomer and dimerized dye molecules. Illustrated in figure 3.3, depolarization leads to a reorientation of the membrane-bound dye
molecule from perpendicular to parallel relative to the membrane surface tangent. The
parallel orientation then can dimerize with dissociation constant KD . The resultant optical change is due to the fact that the monomer and dimer forms of the dye have different
absorption spectra and quantum yields [163]. This mechanism, proposed and confirmed
by Dragsten and Webb [52], explains the potential-dependent changes in absorption in the
dye Merocyanine 540.
KD
K(Vm)
Figure 3.3: Dimerization (“rotation-dimer” mechanism underlying spectral changes in response to membrane potential in Merocyanine 540. Depolarization causes a reorientation
of the dipolar dye molecule from perpendicular to parallel (relative to the membrane) with
a potential-dependent equilibrium constant K(Vm ). The parallel orientation is susceptible
to a dimerization reaction with dissociation constant KD , and the monomer and dimer
species of the dye have different absorption and emission spectra.
The “on-off” mechanism was proposed by Waggoner et al. [193] based on studies of
the cyanine dye diS-C2 -(5) in an artificial lipid bilayer membrane. The mechanism essentially refers to solvent-dependent changes in absorption spectra. The proposed scenario
is that at “resting” membrane potential, the dye aggregates on the hydrocarbon interior
71
of the bilayer membrane; then, in response to changes in membrane potential, the dye
moves outward into the aqueous solution surrounding the bilayer membrane. As the dye
has different absorption spectra in hydrocarbon environments versus water environments,
a change in absorption, measured at a constant excitation wavelength, is observed. This
may explain the rapid absorption response in some cyanine and oxonol dyes. It should
be noted that dye aggregation can also occur purely via Nernstian changes in intracellular
dye concentration, particularly for dyes that are readily internalized by the cell [177]. This
effect, however, is extremely slow (minutes) in comparison.
Some form of potential-dependent chromophore reorientation may explain the absorption and fluorescence changes of certain dyes, especially at low staining concentrations where aggregation effects are unlikely. However, there are few observed situations
in which this mechanism appears to be the largest contributing factor to the optical signal [192].
Finally, the Stark effect mechanism (or “electrochromism”) refers to quantum mechanical changes in molecular absorption spectrum due to the interaction of the molecular
dipole moment with the membrane DC electric field. The Stark shift voltage-sensitivity
mechanism has gained particular interest because it is physically well-understood and
provides guidelines for a priori designed probes via quantum-chemical techniques [117].
Purely electrochromic probes have proven to be quite elusive, however, and even highly
electrochromic probes typically display optical changes that deviate from those predicted
by the Stark effect alone [118]. The search has centered around a class of hemicyanine dyes. While these probes display Stark shift-like absorption changes as a function
72
of membrane potential, remarkable variability in optical trends has been found among
structurally similar amphiphilic hemicyanine dyes [72]. Detailed study of the potentialdependent spectral changes in at least one particular hemicyanine dye, RH421, bound
in unilamellar phospholipid vesicles concluded that the observed potential-dependent absorbance changes were not consistent with purely electrochromic phenomena, and proposed that dye reorientation mechanisms likely played an important role too [36]. The
two common potential-dependent traits repeatedly identified in hemicyanine dyes are that
membrane-bound excitation spectra are blueshifted and that fluorescence quantum yield
decreases in response to depolarization. Recent investigation of a novel class of anellated
hemicyanine dyes [96] has identified a few voltage-sensitive probes that display optical
changes consistent with pure Stark shift effect [109]. In contrast with previous probes,
these dyes additionally display large fluorescence changes as a function of membrane potential when excited via one- or two-photon excitation near the red spectral edge of the
absorption curve [110].
3.2 Ion indicator probes
While voltage-sensitive dyes unambiguously monitor membrane potential optically, fluorescent indicators that are sensitive to electrophysiologically-relevant ionic concentrations
can also be used to track electrical activity. Local concentrations of K+ , Na+ , Cl− , and
Ca2+ ions all play a role in determining membrane potential and in synaptic transmission; fluorescent ion indicator probes are commercially available for each of these ions.
73
Of these, Ca2+ indicators have been utilized the most in neuroscience research, primarily
because of the large optical signal associated with large changes in Ca2+ concentration
during synaptic transmission (see [190] for a review). Na+ indicators have also been effectively employed to investigate Na+ dynamics in cultured neurons and in in vitro preparations [161].
BAPTA
Fura-2
Calcium Green
Figure 3.4: Structures of the chelator BAPTA (1,2-bis(o-aminophenoxy)ethaneN,N,N’,N’-tetraacetic acid) and two common probes which consist of fluorescent
molecules bound to BAPTA: Calcium Green-1 and Fura-2 (Molecular Probes / Invitrogen Corporation, Carlsbad, CA, USA).
Ca2+ indicators typically consist of a non-fluorescent chemical chelator such as EGTA
(glycol-bis(2-aminoethylether)-N,N,N’,N’-tetraacetic acid) or BAPTA (1,2-bis(o-aminophenoxy)ethane-N,N,N’,N’-tetraacetic acid) bound to a fluorescent chromophore. The chelation
(from the Greek word χηλ´η, meaning “claw”) process in BAPTA consists of a Ca2+ ion
binding reversibly to the two nitrogens, four carboxylates, and two oxygens (see structures
in Fig. 3.4). Binding of Ca2+ to an indicating probe can affect the optical properties by
74
any one of three physical mechanisms: (1) a shift of excitation or emission peaks (Fig.
3.5A), (2) a change in fluorescence intensity (Fig. 3.5B), and (3) a change in fluorescence
resonance energy transfer (FRET) [1].
A
B
Figure 3.5: A: Changes in fluorescence emission spectrum of Fura-2 with varying [Ca2+ ];
B: Changes in fluorescence intensity in Calcium Green-1as a function of [Ca2+ ] (Reproduced from [133]).
75
Chapter 4
In vivo fluorescence microscopy of
neuronal activity in three dimensions
using voltage-sensitive dyes
4.1 Introduction
Neurons integrate and transmit information by means of changes in the electrical potential across their membranes. The predominant methods for measuring such activity in
neuroscience use electrodes inserted in the tissue. Such methods, however, provide limited spatial information and are invasive. Optical methods capitalizing on changes in the
transmembrane electric field visualized through intrinsic signals [168, 182] and voltagesensitive dyes [38, 43, 200] (VSDs) offer the possibility for simultaneously imaging the
activity of many neurons [200]. This is a key experimental step for investigation of more
76
complex neural functions such as learning, memory, and the fast dynamics of the cortical
networks on which such functions depend [43].
Unfortunately, brain tissue scattering [19] and the small signal-to-noise ratio (< 1%)
of optical measurements limit the depth penetration and require long integration times
[9, 93, 149]. Experiments with intrinsic signals [9, 182] employ dark-field illumination
[93, 182] to detect variations in cellular morphology, either measured as a change in cell
volume [9] or as a change in extra-cellular space [93]. These studies typically integrate
the responses of the first few hundred micrometers of tissue below the surface. In contrast, most fluorescence microscopy has been limited to nonspecific two-dimensional surface measurements (25 µm, 2.5-ms resolution) of neuronal activity in vivo [149] and in
vitro [43, 201]. More recently, multiphoton microscopy schemes have demonstrated improved depth of penetration [104, 187], but signal-to-noise considerations have limited
their application to slower processes in the brain, with a limited field of view [142]. Finally, diffuse optical methods [199] show promise for detecting changes in deep structures
of brain tissue but have poorer spatial resolution than the microscopies.
Here we describe three-dimensional in vivo epifluorescence functional microscopy of
neuronal activity in layers one through three of the mouse barrel cortex, a region of the
primary somatosensory cortex where the whiskers are represented. The measurements had
∼ 16 µm spatial resolution in the specimen plane, 3-millisecond temporal resolution, and
probed to depths of ∼ 150 µm below the cortex surface with a transverse field of view of
approximately 1.3 mm × 3.2 mm. The depth-dependent differential fluorescence images
were consistent with known cortical architecture. To our knowledge these are the first
77
in vivo optically sectioned neuronal activity measurements based on VSD fluorescence.
The microscopy employed a custom gradient-index (GRIN) lens probe. A critical improvement over previous work [154] was the use of epi-illumination rather than dark-field
illumination.
4.2 Methods and Materials
The optical layout for the novel endoscopic imaging device employed for these measurements is shown in Figure 4.1. A more detailed description of the device is presented in
Chapter 2, Section 2.4. Frames of optical data were acquired as differences in fluorescence
from a 14-bit reference image, F , (the first frame in the movie) on-line. This yielded a
movie consisting of frames that show ∆F . To correct for irregularities in staining, these
frames were divided (off-line) by the reference image to produce fractional fluorescence
data (∆F/F ). RH795 (Molecular Probes, 530 nm in methanol) exhibits a spectral emission shift in response to depolarization, and ∆F/F was typically negative. Because our
CCD was gray scale, this spectral shift was measured as variation in intensity (∆F/F ).
The mice were anesthetized with a combination of ketamine and xylazine (100 and 20
mg/kg, respectively, intraperitoneally), a 5-mm-diameter craniotomy was performed, the
dura was removed, and the surface of the brain was exposed. Tungsten microelectrodes
were placed in the cortex and the thalamus for recording and electrical stimulation (Fig.
4.1E). The exposed surface of the barrel cortex was stained with RH 795 (spectra shown
in Figure 4.2) for ∼45 min. The cortex was then washed with saline solution to remove
78
B
A
epi-illumination
C
darkfield
illumination
Laser
EF
Laser
CCD
CCD
RO
RO
EF
BS
VM
CL
GRIN
IF
1mm
GRIN
0 ms
E
200 ms 300 ms 400 ms
Sh
ut
te
rc
lo
se
st
im
ul
us
st
im
ul
us
st
im
ul
us
D
Sh
ut
te
ro
pe
n
Mouse
~ 800 ms
time
cortical electrode
thalamic electrode
Figure 4.1: GRIN lens fluorescence endoscopy optics setup and experimental timing scenario. Optical setup for (A) epi-fluorescence and (B) dark-field illumination schemes.
Inset in B, detail of perimeter fiber illumination. BS, beam splitter; CL, collimating objective; RO, relay objective; EF, emission filter; VM, vibrating mirror; IF, illuminating fibers.
C: Cross-section of RH795-stained brain slice (top, bright field; bottom, fluorescence) revealing uniform staining. D: Time course of events for single experimental recording trial.
E: Illustration of relative locations of thalamic and cortical stimulating electrodes.
79
unbound dye. Staining was fairly homogeneous as determined from sliced cross sections
of the brain examined after the experiments (Fig. 4.1C).
RH 795
excitation
emission
Figure 4.2: Phospholipid membrane-bound excitation and emission spectra for RH 795
voltage-sensitive dye.
The experimental timing sequence is shown in Fig. 4.1D. Trials were separated by
a rest period of 6 s. To increase the signal-to-noise ratio, optical recordings of multiple
identical trials (10–30 depending on signal size) were averaged frame by frame. Only one
focal plane was imaged per set of single-layer averages, but the order in which the layers
were imaged was random.
4.3 Results
In Fig. 4.3 we display images of ∆F/F from recordings at 0, 75, and 150 µm below the
cortical surface. All images were processed spatially with a 3 × 3 Gaussian smoothing
80
filter. The time traces in Fig. 4.3 were smoothed temporally with a three-frame sliding window filter. Image processing was performed only to clarify large scale qualitative
changes in neuronal activity. The time traces represent changes in fluorescence derived
from all pixels thresholded below −0.3%. The frames and traces display neuronal activity in response to single electrical stimuli of the cortex (Figs. 4.3A and 4.3C) and the
thalamus (Figs. 4.3B and 4.3D). Each led to qualitatively different three-dimensional spatiotemporal activation of the cortex. In Figs. 4.3A and 4.3B a series of nonconsecutive
frames shows the response to a single stimulus. Pixel-averaged ∆F/F time traces in Figs.
4.3C and 4.3D are derived from all frames.
The cortex is laminar, and the cellular architecture varies with depth. Since fluorescence is emitted from essentially all stained tissue, software tools were used to reduce the
contributions from out-of-plane fluorescence. To this end, we employed a computational
deconvolution procedure developed by Yae et al [197] (see Section 2.6). This algorithm
required an experimentally measured point-spread function at varying depths of defocus
as an input parameter [31]. Experimental point-spread functions were obtained by the
imaging of fluorescent beads (3-µm diameter) in situ at varying degrees of defocus. Our
deconvolution techniques resulted in a reduced fractional fluorescence level as well as an
increase in the signal-to-noise ratio in plots of pixel value versus time.
The discriminating power of the optical device is illustrated in Fig. 4.4. We exhibit
three of the image planes during responses to single electrical stimuli applied to the cortex
in the vicinity of the recording area and to the ventrobasal nucleus of the thalamus. Images
of ∆F/F following cortical stimulation (Fig. 4.4B) showed a band of activation along the
81
C0
A
168ms
-3
-0.2
172ms
∆F/F
164ms
x 10
-0.4
-0.6
0um
75um
150um
-0.8
-1
-1.2
0 30 60 90 120 150 180 210 240
B
time (ms)
1.5 mm
2 mm
D
x 10
-4
0
∆F/F
-2
-4
0um
75um
150um
-6
-8
0 30 60 90 120 150 180 210 240
Epi-illumination (surface)
12ms
21ms
F
18ms
15ms
27ms
24ms
Darkfield-illumination (surface)
time (ms)
G
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
-3
-3.5
-4
-3
x 10
Epi-illumination
Darkfield
∆F/F
E
0
15
30
45
60
75
90
time (ms)
x 10-3
-1
0
-0.5
Figure 4.3: Measured ∆F/F activity at three different depths in response to, (A), local
cortical and, (B), thalamic stimulation. Single electrode stimulus was delivered at 150
ms in A & B, and at 0 ms in E & F. Time stamps for B are as in A, F as in E. Frames
from left to right: ∆F/F one frame after stimulation, halfway up the rising edge of the
response, and maximum ∆F/F . C & D: Traces of thresholded pixel-averaged ∆F/F activity show trends as a function of depth (C, response to cortical stimulus, D, response to
thalamic stimulus). E & F: Comparison of signal-to-noise with epi-illumination and darkfield illumination: Response to thalamic stimulation imaged using (E) Epi-illumination,
and (F)dark-field illumination. G: Signal-to-noise comparison for epi- vs. darkfield illumination.
82
field of view stretching away from the stimulation electrode. The highest response amplitude was at the surface, and it decreased smoothly with depth. In contrast, the response to
thalamic stimulation (Fig. 4.4C) showed a much more widespread area of activation with
the highest values at 150-µm depth, limited immediately above (at 75 µm) by the lowest
response values, and finally by intermediate fluorescence values at the surface.
C
B
150 µm
75 µm
0 µm
A
E
D
I
II / III
IV
V
VI
THALAMUS
THALAMUS
Figure 4.4: Peak response ∆F/F images at three focal planes 0, 75, and 150 µm below
the cortex surface (background images at these depths shown in A) during the spread
of electrical stimulation (B, cortical; C, thalamic). D, Known thalamocortical and, E,
intracortical projections. Shading shows approximate range of GRIN lens measurements
with respect to cortical layers and indicates the strength relative to the amplitude of the
responses to, D, cortical and, E, thalamic stimulation.
Such distribution of activity is compatible with the known anatomy and physiology of
the superficial layers of the cortex [32], in particular with the geometry of cortical cells and
the distribution of inputs. Clearly the novel optical probe is capable of spatiotemporally
distinguishing layer-dependent physiological responses.
83
Finally, we compared epi-illumination and dark-field illumination schemes (Fig. 4.1A&B).
A priori, the detrimental effects of speckle and reflected background light were not known.
Fig. 4.3E shows epi-illumination and dark-field illumination ∆F/F recordings of the cortical surface in response to a single stimulus of the thalamus averaged over ten trials.
The plots in Fig. 4.3F exhibit corresponding pixel-averaged ∆F/F traces for each case;
the selected pixels are at the center of the activity. The signal-to-noise ratio of the epiillumination trace was nearly an order of magnitude larger than the corresponding darkfield recording. This is because in the dark-field illumination scheme, only light backscattered from the tissue at and below the image plane returned through the GRIN lens and
was detected. To reach an appreciable signal, significantly higher levels of illumination
are required in dark-field-illumination (∼500 mW/cm2 ) compared with epi-illumination
(∼100 mW/cm2 ). Our results suggest that in VSD-based fluorescence measurements too
much illumination light is lost in dark-field.
To conclude, we have demonstrated epi-illumination, GRIN-lens-based, three-dimensional
in vivo fluorescence microscopy of neural activity by use of VSDs. To our knowledge these
are the first in vivo optically sectioned neural activity measurements.
4.4 Appendix: Confocal microscopy study of voltage-sensitive
dye staining penetration
Topical application of voltage-sensitive dyes is typically the most practical method of
staining for in vivo preparations where surface structures of the brain (e.g. cortex) are to
84
be studied. When imaging at depths below the stained surface, it is crucial to determine
whether the imaged depths are sufficiently stained such that they provide meaningful contributions to the fluorescence signal.
Following the GRIN lens imaging experiments described in the preceding sections,
mice were sacrificed and their brains were removed and sliced coronally (see Fig. 4.5 for
slicing geometry). Wide-field confocal imaging revealed the relative depth of penetration
of RH 795 voltage-sensitive dye (Fig. 4.5).
Figure 4.5: Confocal survey imaging of RH 795-stained coronal brain slices. Images are
from slices prepared from fixed tissue following an imaging experiment. Thin “sliver”
of staining indicates region where cortex was stained in vivo (via diffusion of topicallyapplied dye). Note that image color is not indicative of emission spectrum (RH 795 has
an emission peak near 700 nm).
85
Analysis of reconstructed depth-profile images is shown in Figure 4.6. Averaged intensity pixel-plots through the first few hundred µm of cortex reveal that the staining intensity
drops-off significantly after 100 µm and only a quarter or the peak staining intensity remains at 200 µm below the cortex surface.
Intensity (A.U.)
0
100
200
~ 60 µm
100
200
300
Figure 4.6: Confocal images of RH 795-stained (fixed) mouse somatosensory cortex coronal slice. The full depth profile was obtained by digitally patching together single ∼ 60
µm-side images. To generate wide-field confocal images, a 63X water-immersion objective was used in conjunction with a 15 µm pinhole. Image analysis of the staining profile
is shown in the traces (traces). The thick black trace is the average of several 1-pixel-wide
lines drawn down the length of the depth profile (middle image; individual line profiles
are plotted as the thin colored lines).
86
Chapter 5
Spatiotemporal activity patterns during
respiratory rhythmogenesis in the rat
ventrolateral medulla
One of the most important brain rhythms is that which generates involuntary breathing
movements. The lower brainstem contains neural circuitry for respiratory rhythm generation in mammals. To date, microsectioning and selective lesioning studies have revealed
anatomical regions necessary for respiratory rhythmogenesis. Although respiratory neurons distributed within these regions can be identified by their firing patterns in different
phases of the respiratory cycle, conventional electrophysiology techniques have limited
the investigation of spatial organization within this network. Optical imaging techniques
offer the potential for monitoring the spatiotemporal activity of large groups of neurons
simultaneously. Using high-speed voltage-sensitive dye imaging and spatial correlation
87
analysis in an arterially-perfused in situ preparation of the juvenile rat, we have determined
the spatial distribution of respiratory neuronal activity in a region of the ventrolateral respiratory group containing the pre-Bötzinger complex (pBC) during spontaneous eupneic
breathing.
While distinctly pre- and post-inspiratory related responses were spatially localizable
on length scales less than 100 µm, we found that the investigated area on the whole exhibited a spatial mixture of phase-spanning and post-inspiratory related activity. Additionally,
optical recordings revealed significant widespread hyperpolarization, suggesting inhibition in the same region during expiration. This finding is consistent with the hypothesis
that inhibitory neurons play a crucial role in the inspiration-expiration phase transition in
the pBC. To our knowledge this is the first optical imaging of a near fully-intact in situ
preparation that exhibits both eupneic respiratory activity and functional reflexes.
5.1 Background
Breathing is an essential function of all higher vertebrates. It is under both voluntary
and involuntary control, with the involuntary component consisting of a basic rhythm
generator within the central nervous system, with central and peripheral reflexive control
elements. The neuronal center for generating rhythmic breathing in mammals is located
in the medulla of the brainstem ( [69, 121]; see [21, 41, 61] for reviews). The neurons
required for rhythmogenesis are localized primarily in a limited region of the ventrolateral medulla [144, 179], but their exact location and the mechanism of primary rhythm
88
generation remains a subject of active inquiry. Figure 5.1 shows the basic anatomy of the
medullary regions of interest in this study and details the relative locations of the subregions involved in respiration rhythm generation.
Figure 5.1: Anatomy of brainstem regions involved in control of breathing rhythm generation. Adapted from [122].
Among these ventromedullary regions, the site known as the pre-Bötzinger Complex
(pBC) (see relative location in Fig. 5.1) [179] is an important region for rhythmogenesis. This region contains many different respiratory neuron types [172, 184], and has
been established in vivo as essential for rhythmic breathing [78, 151, 153]. While the
exact nature of the mammalian breathing rhythm generator is not known, it is suspected
89
that both intrinsic cellular and network properties contribute to rhythmogenesis with statedependent [165], and age-varying degrees of importance [91, 126, 127, 148]. Pharmacological studies in arterially perfused rats have revealed that synaptic inhibition is vital for
respiratory rhythm generation in adults but not for neonates, suggesting an overall developmental shift in rhythmogenesis mechanism to a network mechanism [87, 148]. Many of
these studies, though, are difficult to reconcile because each in vitro preparation defines
a different operational condition compared to in vivo preparations [160]. Clearly, there
are differences in the generation of respiratory rhythm between neonates and adult rats, as
well as between in vivo and in vitro models.
The pBC has been described in vitro as a major respiratory rhythm-generating region
containing (early) inspiratory activity [97, 108, 179]. On the other hand, in vivo electrophysiological investigation of the adult cat [172] and rat [184] has revealed that the pBC is
neuronally mixed area containing phase-spanning propriobulbar neurons that fire through
the respiratory phase transition (see Fig. 5.2 for an overview of the firing patterns of the
various respiratory neurons). These studies hypothesize that such phase-spanning neurons in the pBC promote the expiratory-inspiratory phase transition and are thus critical to
rhythmogenesis in vivo. Thus, results obtained in vivo vs. in vitro suggest very different
functional roles for the pBC.
To achieve a more global picture of neuronal distribution, some groups have employed
optical imaging techniques to visualize large-scale activity. Such studies, however, have
concentrated on in vitro preparations of the neonatal rat, either of brainstem slices [108,
189], or en bloc preparations of the brainstem and spinal cord [144, 189]. Unfortunately,
90
Inspiration
Expiration
Expiration "phase I"
Expiration "phase II"
Early Inspiratory
Inspiratory
Augmenting
Late Inspiratory
Post Inspiratory
Expiratory
Augmenting
Pre-Inspiratory
Integrated
Phrenic Nerve
Recording
Figure 5.2: Idealized firing patterns of the various respiratory neuronal subtypes. The
respiratory cycle is divided into inspiration (red) and expiration (blue), and the various
neuronal subtypes found in the respiratory regions within the medulla are classified by
comparing their firing dynamics to the two respiratory phases. Adapted from [21].
91
interpretation of results from these neonatal or reduced preparations, with respect to adult
in vivo mechanisms, incurs the same disadvantages described above.
We hypothesize that in a more mature and intact rat preparation, the area of the pBC
would be functionally mixed and exhibit significant activity throughout all phases of respiration, as in in vivo preparations. In addition, we expect that inhibitory mechanisms will
be evident. To test this hypothesis, we used voltage-sensitive dye imaging to investigate
the large-scale electrical spatiotemporal activity in the pBC area of a largely-intact juvenile rat preparation. We performed high-speed wide-field fluorescence video microscopy
of the surface of the medulla immediately ventral to the VRG, in an arterially-perfused
preparation of the juvenile rat [147, 181] using the voltage-sensitive dye, di-8-ANEPPS.
Using correlation coefficient analysis [14, 57], we identified non-overlapping spatial regions within the pBC exhibiting fractional fluorescence signals at different phases of respiration.
5.2 Methods and Materials
5.2.1 General preparation
All procedures were performed with the approval of the University of Delaware Institutional Animal Care and Use Committee. Following the “working heart-brainstem preparation” [147], juvenile rats (72–100g; n = 5) were bisected just below the diaphragm, decerebrated at the precollicular level, and skinned under deep isoflurane anesthesia in 4 ◦ C
92
artificial cerebrospinal solution. Deep anesthesia was assured by the absence of limb withdrawal and reflexive changes in breathing following limb pinch. They were then bilaterally
vagotomized, with the diaphragm, heart, lungs, and portions of the rib cage removed, and
appropriate vasculature ligated to prevent excessive leakage of the perfusate.
The descending aorta was cannulated and rats were perfused with a 95% O2 / 5%
CO2 -saturated solution consisting of 125 mM NaCl, 1.25 mM NaHCO3 , 4 mM KCl, 2.5
mM CaCl2 , 1.25 mM MgSO4 , 1.25 mM KH2 PO4 , 20 mM dextrose, 0.05 mg/ml vecuronium bromide, and 2.5% dextran (230 kDa) at 95–105 mmHg and gradually warmed to
31–33 ◦ C. This experimental preparation has been shown to produce eupneic breathing
patterns in mice [147] as well as rats [181].
Phrenic nerve (300–3000 Hz) and single unit (500–3000 Hz) recordings (10,000 samples/s) were made with silver wire hook electrodes and glass electrodes (2M NaCl), respectively. Signals were amplified and filtered using conventional electronics (Neurolog,
Digitimer, Ltd.). Raw phrenic nerve recordings were used to monitor the respiratory state
of the preparation (e.g. eupnea, gasping, etc.) and to adjust perfusate flow accordingly.
5.2.2 Imaging system
Simultaneous image and electrophysiological data acquisition was performed with custombuilt hardware (including camera) and custom-written software [155,156] capable of continuous image and data recording at high speeds (up to 1 kHz frame rate for video data, 20
kHz/channel for electrophysiology data) for long periods of time (> 1 hour). The optical
93
imaging setup is in Figure 5.3A and described in detail in Section 2.3.
5.2.3 Dye staining
Stock solutions of voltage-sensitive dye di-8-ANEPPS dissolved in F-127 Pluronic (20%
in DMSO) (both from Molecular Probes / Invitrogen Corporation, Carlsbad, CA, USA)
were diluted to final concentrations of 500 µg/mL in Ringer’s solution. The diluted dye
solution was loaded into the tissue systemically, via a side port of the aortic perfusion
line. Following perfusion with di-8-ANEPPS, reduced rat preparations continued to exhibit eupneic breathing. In addition, it was possible to elicit the Hering-Breuer reflex whereby inspiration is terminated and expiration extended - through electrical stimulation
of the vagus nerve (Fig.5.3B). We used the time latency and efficacy of the Hering-Breuer
reflex and the overall phrenic activity (amplitude, duration, pattern, and rhythmicity) as
indicators of possible toxic effects of systemic dye introduction. In all preparations, these
parameters were not changed more than ±5% after 15–90 min of dye injection. In addition, the presence of stereotypical respiratory-related single-neuron activity (Fig.5.3C,D)
provided further confidence that the effects of the dye-staining were inconsequential. Because the perfusate is optically transparent at the excitation and emission wavelengths
(480 ± 20 nm and >590 nm, respectively) used in the experiment, erythrocyte-dependent,
oxygenation-related optical artifacts, i.e. large sources of noise in in vivo conditions, were
eliminated.
94
Figure 5.3: Experimental system and physiological validation. A: optical setup for fluorescence measurements from the ventral medullary surface. B: Hering-Breuer reflex
demonstration 15 min after injection of di-8-ANEPPS into perfusate. Region between
arrows (VN stim) indicates electrical stimulation (0.2 ms, 50 Hz, 100 ms) of central end
of transected cervical vagus nerve, leading to termination of inspiration (raw neurogram).
C and D: single unit recordings of inspiratory (C) and expiratory (D) neurons in the preBötzinger complex (pBC) region after injection of potentiometric dye. Traces from top
in C and D: extracellular neurogram; integrated phrenic activity; raw phrenic neurogram;
inspiratory cycletriggered histogram of activity. Bin widths: 0.25 ms. Sweeps: C = 32, D
= 67. Note time scales differ slightly in C and D.
95
5.2.4 Measurement protocol
The imaged region included areas just rostral and lateral to the rostral-most XII rootlets,
and the region indicated in Figure 5.4A (rectangle) was used in all experiments. The
focal plane of the camera was set to ∼200 µm below the ventral surface. After the dye
was loaded, raw images and electrophysiological data were collected for long durations
(∼40 min), during which several hundred eupneic breaths typically occurred. Triggered
averaging of video and electrophysiology was performed off-line using custom written
functions in Matlab software (MathWorks, Natick. MA). Integrated phrenic recordings (τ
= 50 ms) were used to create trigger markings for averaging, and the final processed data
was averaged over 20–50 breaths. Off-line averaging enabled arbitrarily long pre-trigger
times.
Because of the variability in phrenic burst duration, off-line triggered averaging was
performed using both early- and late-phase triggering points. The onset of the integrated
phrenic burst (defined by a threshold set just above the noise level) served as the “early”
trigger point; the peak (i.e., turning point) of the integrated phrenic burst defined the “late”
trigger point. For characterizing onset times and peak response values, early-phase triggered averaging was used for optical responses occurring just before the onset of the
phrenic burst (pre-inspiratory period), and late-phase triggered averaging was used for
responses during and after inspiration.
Fractional fluorescence (∆F/F) images were produced by subtracting a reference image (see below for details), F, from averaged video images to generate ∆F. This quantity
96
was then divided by F to yield normalized changes and correct for irregularity in staining.
Using our emission filter (>590 nm), di-8-ANEPPS exhibits a decrease in amplitude of
fluorescence emission during membrane depolarization. Thus, depolarizations are accompanied by negative changes in fractional fluorescence (∆F/(F)), and vice versa for membrane hyperpolarization. Fractional fluorescence CCD recordings were smoothed spatially
with a 5 × 5 Gaussian filter and temporally with a 5-frame sliding box car average. Other
than this simple spatiotemporal smoothing, all data traces are presented unfiltered.
Following the imaging experiment, dye was preferentially applied to the two rostralmost XII rootlets by placing a small di-8-ANEPPS-saturated piece of gauze directly onto
the rootlets with a pair of forceps for the purpose of generating a high-contrast background
image including identifiable anatomical landmarks in the imaging field of view. After ∼5
min, the gauze was removed and unbound dye was washed away by the constant outflow of
perfusate in the brainstem area. Several seconds’ worth of frames of this background were
averaged in order to resolve anatomical details in the low-contrast tissue. Thresholded
montage images (Fig.5.4C) are displayed superimposed over this background image.
5.2.5 Image analysis
Voltage-sensitive dyes yield fluorescence changes linearly proportional to membrane potential changes and can therefore be used as a molecular voltmeters [25,38,167]. However,
because the polarity reported by ∆F/F depends on the choice of reference image, F , we
97
compared ∆F/F using two different reference images. One reference image was generated by averaging the first 10 frames of the triggered average CCD recording. To achieve a
reference fluorescence image more likely close to steady-state potentials, the preparation
was cooled to a temperature of 16◦ C, during which no rhythmic or tonic phrenic activity
was observed and fluorescence images were recorded and averaged to generate F .
Correlation coefficient images [14, 57] were computed by calculating the correlation
coefficient for each pixel with a given correlation function. The correlation coefficient,
cci , is defined as
PN
− µf )(ri − µr )
qP
N
2
2
n=1 (fi − µf )
n=1 (ri − µr )
cci = qP
N
n=1 (fi
(5.1)
where fi (i = 1 . . . N ) is the time course of ∆F/F for a given pixel, ri (i = 1 . . . N )
is the function representing the desired correlation function during respiration, µf and µr
are the mean values of f and r, and N is the number of frames.
Two types of functions were used in our correlation analysis: functions that approximate firing frequency rates for 5 basic types of respiratory neurons (pre-inspiratory, lateinspiratory, post-inspiratory, ramp-inspiratory, and expiratory-augmenting) and functions
that approximate the membrane potential trajectories for those same neuron types. Approximations were based on descriptions by [21]. Following the description of the respiratory phases in [159], three points on the integrated phrenic burst recording were used
to define the inspiratory and expiratory phases (Fig.5.4B) and to define correlation functions (Fig. 5.5A): the onset of the inspiratory phrenic burst, the peak / turning point of
98
the inspiratory phrenic trace, and the return to baseline of the phrenic trace. The period
of “inspiration” spanned the duration between the phrenic burst onset point and the peak /
turning point. “Post-inspiration” (i.e., E1) was defined as the period between the integrated
phrenic peak / turning point and the return to baseline of the phrenic trace. “Expiration”
(i.e., E2), spanned the duration between the return to baseline of the integrated phrenic
trace and the onset of the next breath’s phrenic burst. Because depolarization resulted in
a negative change in fluorescence while the value of correlation functions followed the
sign convention of intracellular neuron recordings (depolarization represented by a positive change), maximal cc values are negative when the sign of the correlation function is
opposite that of the optical response. A cc value of 1 would imply that the time course of
∆F/F is identical to the correlation function multiplied by an arbitrary constant; a value
of 0 indicates no similarity between the ∆F/F time course and the correlation function.
Maximum values of cc for our data ranged from 0.4–0.7, due to mismatch between model
functions and real activity and due to the presence of physiological and CCD noise in
optical recordings.
99
Figure 5.4: Fluorescent respiratory-related signals from the rat VLM. A: Anatomical location of imaged region. Rectangle outline indicates CCD imaging area. Dashed circle:
approximate outline of pre-Bötzinger complex. Dashed lines: rostral and caudal limits of
XII roots. Blue and red dots show locations used to generate corresponding traces in B.
Scale bar, bottom, in mm. Inset: cross-sectional view taken from slice 1.05 mm rostral to
calamus scriptorius. Red region: nucleus tractus solitarius. Green region: lateral tegmental field. Blue circle: pBC. Small Square: approximate location of extracellular recordings
in Fig. 5.3. B, top and middle: Fractional fluorescence signal from corresponding points
indicated in A. Note that axis scales have been inverted to mimic corresponding depolarization such that negative change in fractional fluorescence corresponds to an increase on
these axes (i.e., depolarization). B, bottom: Corresponding averaged, integrated phrenic
nerve activity plotted on same time-scale. Shading: pre-inspiratory and post-inspiratory
portions of respiratory cycle. Dotted lines: relevant peaks in ∆F/F . C: Montage of selected frames from a recording showing clear spatiotemporal separation of pre- and postinspiratory signals. Scale bar in first frame = 1 mm. Blue and red dots in first frame (same
as in A) show regions of ∆F/F traces in B. Images are superimposed upon background
fluorescence image displaying two rostral-most XII rootlets. Below each frame is integrated phrenic trace with time-marker (vertical bar) representing the temporal relationship
of current frame with respect to respiratory cycle. Note emerging widespread excitation
during inspiration (second row) and widespread inhibition during expiration
100
5.3 Results
5.3.1 General trends in the optical signal
For all rats (n = 5) in a region that included the pBC, we observed consistent fractional
fluorescence responses during the pre-inspiratory period, during inspiration, during postinspiration, and during expiration (see METHODS section for detailed definition of periods). The fractional fluorescence responses during pre-inspiration, inspiration, and postinspiration were all (statistically significant) decreases in fluorescence, indicating depolarization (Fig.4.3B). The fractional fluorescence response during expiration was a statistically significant increase in fluorescence, indicating hyperpolarization (Fig.5.4B). While
peak response regions were spatially localizable on length scales less than 100 µm, these
responses were also readily detectible in time-traces of ∆F/F averaged over the entire
field of view.
5.3.2 Timing of respiratory phase-locked signals
Averaging over all animals, the integrated phrenic burst duration, defined here as the
time between the onset and the return to baseline of the phrenic burst, was 935 ± 68
ms (mean ± SD). The range of breath durations was between 870 and 1000 ms and the
standard deviation of the phrenic burst duration was less than 9% within individual experiments. Because of the variability in phrenic burst duration, onset times (in ms) of optical
responses will be described both in terms of mean ± SD (averaging over all animals) and
in terms of percentages of the inspiratory burst duration (i.e. onset values are divided by
101
individual experiment phrenic burst duration prior to group averaging) to give more qualitatively general descriptions. As described in the METHODS section, the variability in
phrenic burst durations necessitated the use of early- and late-phase triggered averaging.
Averaging using either of the trigger points did not erase any of the responses, but did lead
to smaller standard deviations when the trigger closest to the optical response was used.
The “pre-inspiratory” response onset preceded the phrenic burst onset by 178 ± 85 ms, or
19 ± 9% of the phrenic burst duration. This initial gross fluorescence response had an average peak value of −0.029 ± 0.014% and always ended before the middle of inspiration.
The ∆F/F response (triggered off of the onset of inspiration) that peaked during inspiration began 296 ± 107 ms (27 ± 9%) after the phrenic burst onset, and reached peak values
of −0.037 ± 0.015%. The ∆F/F response (triggered off of the phrenic peak) that reached
maximum during the post-inspiratory period (e.g., Fig.5.4B, last peak in red trace) began
329 ± 174 ms prior to the phrenic burst peak (29 ± 12% of the burst width), and reached
∆F/F values of −0.038 ± 0.024%. We found a small but significant overall increase in
fluorescence of +0.023 ± 0.020% during the expiratory period indicating hyperpolarization (Fig.5.4C; note the graph scales and colorbars have been inverted so a decrease in
fluorescence indicates membrane depolarization).
102
5.3.3 Influence of baseline (F ) on sign of ∆F/F
Using either the average of the first 10 trigger-averaged frames or the average of 10 random
frames in a continuous fluorescence recording of the cooled brainstem (of the same preparation), the pre- and late-inspiratory ∆F/F changes were negative, indicating widespread
depolarization (e.g., Fig.5.4C, middle row last 2 frames). In addition, using the cooledbrainstem fluorescence signal (no phrenic activity) as a baseline, F, revealed gross depolarization during the presence of tonic activity1 (data not shown).
5.3.4 Spatiotemporal trends in event-triggered imaging data
On the scale of our field of view (∼1 mm×1.2 mm), we were able to distinguish both
global and spatially-localized phasic ∆F/F responses without the aid of image processing. Figure 5.4B shows ∆F/F responses at two different locations within the field of
view. The top trace (blue) shows the ∆F/F time trace from the blue marker on the field
of view in Figure 5.4A. While this trace is representative of the gross post-inspiratory
∆F/F changes seen in the frame montage in Figure 5.4C, the red trace in Figure 5.4B,
corresponding to the area near the rostral-most XII rootlet (Fig.5.4A, red dot), shows
an additional pre- and post-inspiratory signal not present in other regions. Figure 5.4C
demonstrates the typical degree of spatiotemporally-restricted fluorescence changes from
the region over the course of a breath.
1
i.e. spontaneous spiking activity as recorded by extracellular electrodes inserted into the pBC region
103
5.3.5 Correlation coefficient analysis
While some spatiotemporal variations were apparent without any image-processing, other
features could be identified with correlation coefficient imaging. Fig. 5.5 shows the results
of correlation coefficient analysis using functions approximating both firing frequency and
membrane potential patterns for five basic respiratory neuron types. The histograms in
Fig. 5.5A and B represent correlation coefficients of all pixels in all experiments, using
both early- and late-phase triggered-averaging. Some of the histograms have one main
peak, while others are clearly multi-modal, indicating high or low consistency among experiments, respectively. Early- and late-phase triggered averaging revealed differences
in this spread of correlation values; triggered averaging led to a decrease in the cc value
spread (see Eq. 5.1 for definition of cc) when the selected trigger-event was near the main
feature (if any) of the correlation function (e.g., pre-I pattern using phrenic onset triggering). Because of this triggering-dependent effect, we additionally analyzed correlation
statistics for the initial portion (beginning immediately after the return of the integrated
phrenic trace to baseline) of the expiratory-augmenting firing frequency function, which
has main features coincident with both early and late trigger events. For both firing frequency and membrane potential functions, the least experimental variability was observed
in the correlation analysis using pre-inspiratory (“pre-I”), late-inspiratory (“late-I”), and
initial expiratory-augmenting (“exp-aug (initial)”) functions (functions shown in left-hand
columns of Fig. 5.5A & B). Of those, pre-I and exp-aug (initial) firing-frequency functions revealed nonzero mean values, indicating ubiquity of some degree of similarity to
104
those functions among pixels in all experiments. A negative mean cc value was observed
among experiments to the pre-I firing-frequency correlation function and a positive mean
cc value was observed using the exp-aug firing frequency function, interpreted as depolarization and hyperpolarization in those temporal patterns, respectively. Compared with the
truncated exp-aug (initial) function, the full exp-aug correlation function revealed considerable experimental variability and no significant non-zero mean cc value.
Examination of an individual experiment (Fig. 5.5C-E) reveals a spatial mixture of
regions correlated with significantly different functions. The top two images in Fig. 5.5C
show a thresholded map of cc values for pre-I and exp-aug (initial) correlation functions,
and reveal that both regions of high correlation cover a similar portion of the field of view.
The bottom image in Fig. 5.5C displays the combined regions of the top two images
(minimum threshold set to 1 standard deviation of the mean value), illustrating that the
highly-correlated (>2 S.D.) sub-regions are largely non-overlapping. For all experiments,
there was less than 5% overlap for regions with cc values greater than 2 S.D. for pre-I and
exp-aug firing-frequency functions, indicating spatial separation between the regions with
the strongest temporal correlations.
Region-averaged plots of ∆F/F (Fig. 5.5D) revealed that values larger than 2 S.D.
showed significantly different time-traces for the two correlation functions. While the
region-averaged plot of pre-I-correlated activity did not exhibit the exp-aug (initial) characteristic increase in fluorescence (note again the reversed y-axis), some degree of pre-Ilike ∆F/F response (peak at ∼300 ms; Fig. 5.5D) was seen in pixels highly-correlated
with the exp-aug function. Additionally, similar to the traces in Fig.5.4B (in a different
105
106
Figure 5.5: Spatiotemporal correlation analysis of activity. A: Correlation analysis (over
all experiments) using firingR frequency-based correlation functions. A, Left column: Averaged integrated phrenic ( Phr) trace (red) shown along with each different correlation
function based on the firing-frequency traces of respiratory neurons (blue). A, Middle column: Histograms of calculated pixel correlation coefficients (for all experiments) based on
the correlation function (blue) on the left using phrenic onset- triggered averaging (“early
phase triggering”). A, Right column: Histograms of calculated pixel correlation coefficients (based on the same correlation functions as the middle column) using phrenic
peak-triggered averaging (“late phase triggering”). B: Correlation analysis (over all experiments) using subthreshold membrane
R potential-based correlation functions. B, Left
column: Averaged integrated phrenic ( Phr) trace (red) shown along with each different
correlation function based on subthreshold membrane potential traces of respiratory neurons (blue). B, Middle column: Histograms of calculated pixel correlation coefficients
based on the correlation function (blue) on the left using early-phase triggered averaging
(see METHODS for details). B, Right column: Histograms of calculated pixel correlation
coefficients (based on the same correlation functions as the middle column) using latephase triggered averaging. C, Top 2 images: Sample images of highly-correlated regions
for expiratory-augmenting (“exp-aug (initial),” blue-cyan colorbar) and pre-inspiratory
(“pre-I,” red-yellow colorbar) firing frequency-based correlation functions (shown at the
bottom of D) in a single experiment. Colorbar below defines three ranges of correlationcoefficient (cc) values: (Mean cc value) < |cc| <1SD, |cc| > 1SD, |cc| >2SD. Correlation
regions are thresholded as indicated by colorbar. Mean cc values and standard deviations
are derived from gaussian fits of cc value histograms (shown in E). Images are superimposed on fluorescence background image of di-8-ANEPPS-labeled XII rootlets, scale bar
= 0.5 mm. Dotted lines: outline of two rostral-most XII rootlets. C, Bottom image: Combined correlation regions (as in images above) showing only values larger than 1 SD. D,
Top: Color-coded region-averaged time traces corresponding to regions in C. Traces represent averaged activity in all pixels with values of greater than 2 SD of the mean cc value.
D, Bottom: Corresponding averaged, integrated phrenic recording used in C and D. Colorcoded functions below phrenic show firing frequency-based functions used in correlation
analysis for C. E: Histograms of pixel cc values for individual experiment analysis.
107
experiment), the pre-I-correlated region contained smaller optical responses in the postinspiratory and expiratory periods.
5.3.6 Extracellular recordings within imaged field of view
During simultaneous optical recordings, we recorded extracellularly from 13 neurons that
exhibited respiratory phase-related activity, all in the CCD field of view and within 340
µm from the ventral surface. Among these, were neurons with “pre-I” activity (onset of
activity 65.2±8.4 ms prior to phrenic onset with maximum firing frequency at the phrenic
onset), early-I and late-I activity (frequency peaks; see Fig. 5.3C for examples). In regions
exhibiting “post-I” activity, several expiratory-decrementing units were also found (onset
of activity ∼mid-inspiration; max firing at phrenic post-I onset). Figure 5.3D shows an
example of one of these units. Thus, neurons with or without “prototypical” firing patterns
for a given region may be observed within that region, indicating a fair amount of spatial
mixture, particularly on scales > 200 µm.
5.4 Discussion
5.4.1 Virtues of a perfused preparation
In efforts to dissect the underlying mechanisms of respiratory rhythmogenesis, different in
vitro preparations have been developed and each has its own particular rhythmic properties which differ from an in vivo preparation [160]. Ramping, eupneic integrated phrenic
108
recordings combined with the ability to elicit the Hering-Breuer reflex in this artificially
perfused preparation (even after the injection of voltage-sensitive dye di-8-ANEPPS into
the circulatory system) suggests superior degrees of functional relevance compared with
brainstem/spinal cord en bloc or in vitro medullary slice preparations. The results clearly
show the feasibility and utility of optical imaging of the working heart-brainstem preparation via voltage-sensitive dyes. We also show that the components of an important portion
of the rhythm generator may be spatiotemporally characterized using these methods. This
represents the first step in dissecting this system as it functions in its intact form, with the
ultimate goal of creating a realistic quantitative model.
5.4.2 pBC region contains a variety of phase-locked fluorescence patterns
This is the first study, to our knowledge, to demonstrate correlation in subregions of the
optical image to specific known firing patterns of respiratory neurons in a relatively intact
preparation. The results shown in Figs. 5.4 and 5.5 suggest that the ∆F/F response in
the area of the pBC can be divided into at least four periods of response: pre-inspiratory,
inspiratory, post-inspiratory, and expiratory. The presence of ∆F/F signals during all four
of these periods is fully consistent with recent voltage-sensitive dye imaging results in the
brainstem-spinal cord preparation [144] and with known membrane-potential dynamics
for respiratory neuron types [18, 63, 86, 158, 166].
Several reports have found that among neighboring regions in the VRG, the pBC is
109
a region containing a relatively high variety of respiratory neuron types including pre-I,
post-I, phase-spanning, and exp-2 neuron types [65, 95, 184]. While it was apparent that
distinctly pre-inspiratory and post-inspiratory ∆F/F responses could be spatially distinguished within our imaging field of view (Fig. 5.4), the largest homogeneous subregions
observed in one experiment were ∼ 400 µm in diameter. Additionally, correlation coefficient images showed that regions exhibiting strong pre-inspiratory ∆F/F signals also exhibited significant post-inspiratory activity (Fig 5.5D, blue trace with small peak at post-I
onset), akin to recent optical imaging studies of the ventral surface of the para-facial respiratory group (pFRG) [144]. Our findings are therefore in agreement with the model of
the rat pBC as a region consisting of predominantly mixed neuron types, and involved in
phase-transition [184].
The results of correlation coefficient image analysis revealed widespread correlation
of the optical signal with pre-I and exp-aug (initial) firing frequency functions. The least
speculative interpretation of this finding is that the optical signal tended to take on the
same general form as those particular functions. The extent to which this finding depends
on integration time remains unclear, and is beyond the scope of this study.
110
5.4.3 Role of inhibition
Our findings of significant increase in fluorescence, corresponding to hyperpolarization,
directly following the termination of inspiration in a region containing the pBC represent the first optical imaging evidence in support of the prevalence of inhibitory activity during rhythmogenesis. Glycinergic inhibition of distinct subsets of respiratory neurons in the VRG has been increasingly identified as crucial for respiratory rhythmogenesis [27, 28, 58, 64, 124]. The region of the pBC, specifically, has been shown pharmacologically to be rich in glycine receptors [34]. That similar increases in fluorescence were
not seen in previous imaging studies of the brainstem-spinal cord preparation [144, 189]
suggests further functional differences between neonatal in vitro and our juvenile in situ
preparations.
5.4.4 Tissue optics considerations and interpretation of the optical
signals
The measurements in this report utilized linear (one-photon) fluorescence. Thus the contribution of out-of-focus light and scattering of the returning fluorescence must be considered. Voltage-sensitive dyes exhibit linear changes in fluorescence and absorption spectra
as a function of membrane potential, but the definite spatial and temporal localization of
the signal depend on the collection optics and integration time, respectively. Our system
specifications (integration time = 8 ms; N.A. = 0.38) enabled us to discern fast optical
signals with a wide field-of-view and long working distance. These specifications lead to
111
uncertainty in both temporal and spatial localization of the signal origins. The observed
signal at each pixel represents a spatial and temporal summation of the fluorescence activity in all of the stained membrane surface area contained within the volume defined by the
lateral width (10 µm×10 µm) and by the fluorescence depth-of-field (∼10 µm). Calibration tests with fluorescent polystyrene beads placed at varying degrees of defocus show
that the optical signal at each pixel reflects fluorescence from up to ∼50 µm surrounding
the focal plane [67].
Because the optical signal at each pixel represents a summation over a potentially
unknown distribution of membrane area, the electrophysiological interpretation of the
voltage-sensitive dye signal warrants further discussion. High-speed microscopy using
voltage-sensitive dyes can provide optical recordings of fast membrane potential dynamics such as action potentials. However, studies in which single detector pixels correspond
to a mixture of neuronal architecture have found that voltage-sensitive dye measurements
largely reflect a superposition of subthreshold membrane potential dynamics rather than
spiking [83, 105, 106, 189]. Other voltage-sensitive dye studies employing spike-triggered
imaging found that the optical signal was most similar to local field potential (LFP) measurements [13, 82], further concluding that the optical signal likely reflects primarily dendritic synaptic potentials rather than spiking. Unfortunately, the basic respiratory neurons are not nearly as well-characterized in terms of their dendritic electrical properties as
they are in terms of spiking frequency. Likewise, there is less certainty about the consistency of subthreshold membrane potential dynamics of respiratory neurons than of firing
frequency. The best interpretation of our imaging results, therefore, is as indicative of
112
summed membrane potential dynamics.
Since the imaged plane was ∼ 200 µm below the ventral surface of the brainstem, scattering of the returning fluorescence signal can be expected to account for some further loss
of spatial resolution. In the rat brainstem, the scattering properties increase significantly
after 3 weeks of life due to increased myelination [102]. To avoid this potential limitation,
our study employed rats just under 3 weeks old. Taking into account the absence of blood
in the artificially perfused preparation, as well as the fact that in vitro scattering coefficient
values are almost always higher than in vivo values in brain tissue [19], it is reasonable to
assume that there are relatively few scattering events in the first few hundred µm of tissue.
In light of these interpretive issues associated with the optical response, the thresholded regions in Fig. 5.4 and the significantly correlated regions in Fig. 5.5 should not
necessarily be interpreted as the locations of cell bodies. Respiratory neurons in the VRG
send their axonal projections and branching collaterals over large distances, frequently
exceeding a diameter of several millimeters [62]. A large fraction of the resulting ∆F/F
signal likely contains contributions from portions of neurons other than the somata, particularly from the dendrites, which contain a large fraction of the surface area of these
neurons [62].
Overall, our results show that the area of the pBC contains spatially mixed subregions that exhibit optical responses during all phases of respiration, and that those subregions’ optical responses exhibit further similarities to the electrical properties of specific
respiratory neuron types. These results suggest that even at the juvenile developmental
stage, the pBC region already demonstrates stereotypical network-driven phenomena (i.e.,
113
widespread inhibition), and may participate in the generation of respiratory phase transitions.
5.5 Conclusion
In conclusion, our findings support the model of the pBC in the juvenile rat as a network of
neurons that is spatially mixed in its organization, but with identifiable sub-structural organization that receives near-global excitatory and inhibitory influences during inspiration
and expiration, respectively. In addition, the sub-regions that show the strongest pre-I and
E-aug correlations are not spatially overlapping. We expect the methods utilized herein
will lead to higher-resolution qualitative and quantitative models of the respiratory rhythm
generator.
114
Chapter 6
Two-Photon Spectroscopy of Functional
Probes for Bioimaging
6.1 Measurement of two-photon absorption (TPA) crosssections
6.1.1 Background Theory
The two-photon absorption cross-section, Σ, can be measured by absorption and excitation spectroscopy techniques [22]. Absorption spectroscopy measures the attenuation of
the incident light as it passes through a sample of known concentration and thickness. Excitation spectroscopy measures the light emitted by the molecules following absorption.
115
Generally, absorption measurements are difficult because the high incident intensities required to generate measurable absorption signals also lead to saturation, excited-state absorption, and photobleaching [4]. On the other hand, for molecules with reasonably large
quantum efficiencies (q > 10−3 ), the fluorescence signal is readily detectable and can
yield accurate measurements of Σ if q is known. Here q is the number of emitted photons
by the molecule per absorbed photon. The two-photon excitation cross-section, σ2 is the
product of Σ and the two-photon quantum efficiency, q2 , defined as the number of emitted
photons by the molecule per pair of absorbed photons:
σ2 = q2 Σ
(6.1)
In 2-photon microscopy the measured signal is typically fluorescence emission, and
therefore σ2 is generally a more useful figure of merit than Σ.
Xu and Webb [196] have shown that for mode-locked laser sources, the time-averaged
fluorescence signal hF (t)i (photons·sec−1 ) detected by a spectroscopic system with collection efficiency ϕ and time-averaged illumination power hP (t)i (photons·sec−1 ) is
1
gp 8ηhP (t)i2
hF (t)i ≈ q2 ΣC
2
fτ
πλ
(6.2)
Here C is the concentration of fluorophores in the sample (cm−3 ), f is the repetition
rate of the pulsed laser, τ is the laser pulse width, gp is a numerical constant of order
unity, λ is the center wavelength of the laser pulse, and η is the real part of the linear
index of refraction at ω. The quantity
gp 8ηhP (t)i2
fτ
πλ
116
accounts for the temporal shape of the
pulses and the spatial profile of the beam in the excitation volume [94, 196]; to compute
this quantity the sample thickness is assumed much larger than the depth of focus of the
imaging system, and diffraction limited high numerical aperture focusing [157, 174] is
also assumed.
Since the rate of two-photon absorption (TPA) depends on the spatial and temporal
excitation beam profiles, Albota et al. [4] introduced a ratiometric technique whereby σ2 of
an unknown sample is deduced from identical measurements made of a reference sample
with known values of σ2 . Within this approach the two-photon excitation cross-section,
σ2S , of an unknown sample is
σ2S (λ) = q2S ΣR (λ)
ϕR CR hP (t)i2R hF (t)iS ηR
ϕS CS hP (t)i2S hF (t)iR ηS
(6.3)
where subscripts S and R indicate “sample” and “reference,” respectively, and the
cross sections are written explicitly as a function of wavelength to remind us about their
wavelength-dependence. In our measurements we monitored the second harmonic signal
SHG(t) from a small fraction of the incident beam that was diverted through a BetaBarium Borate (β-BaB2 O4 , or “BBO”) SHG crystal (Inrad, Northvale, New Jersey, USA);
the time-averaged second harmonic, hSHG(t)i, is directly proportional to hP (t)i2 .
6.1.2 Two-Photon Spectroscopy Setup
The optical setup for cross-section measurements is shown in Fig. 6.1. In the experiment,
CR and CS are known, ΣR , ηR , and ηS are taken from the literature, and hF (t)i , ϕR , ϕS ,
117
and hP (t)i2 are measured for each sample. The laser excitation source was a mode-locked
Ti:Sapphire laser (Mira Basic, Coherent Inc., Santa Clara, California, USA) with a 76
MHz repetition rate. The laser was pumped by an 8 W multi-line Argon-ion laser (Innova
310, Coherent Inc., Santa Clara, California, USA). While two different optics sets were
used to span the spectral region from 790 nm to 960 nm, optics sets that span the range
of 720 nm to 980 nm are now available. The wavelength was confirmed with an external
spectrum analyzer (Ocean Optics Inc., Dunedin, Florida, USA). As noted above, a small
fraction of the excitation beam was focused onto a BBO SHG crystal (Inrad, Northvale,
New Jersey, USA) to record a signal proportional to hP (t)i2 , which was time-averaged
over the measurement period. The ratio of reference and sample values of SHG(t) was
within the range of 1 ± 0.01. To minimize distortion of the beam profile, a λ/2 plate and
Glan-Thompson polarizer were used to adjust the average excitation power at the sample
to be ∼10 mW.
Pulse widths were measured with a home-built background-free autocorrelator [11,
12]. The pulse-width measured just before the sample was ∼200 fs. The beam was
expanded with a 5× Gallilean beam-expander in order to back-fill an N.A. 0.4 objective (Carl Zeiss AG, Göttingen, Germany); the approximate spot size in the sample was
∼3 µm. TPA-induced fluorescence was collected at 90◦ re-focused onto the slit of a
monochromator (Thermo Jarrell-Ash Corporation, Franklin, Massachusetts, USA). The
fluorescence signal was detected by a photomultiplier tube (Hamamatsu R943-02, Hamamatsu Photonics, Hamamatsu City, Japan) which then fed a gated photon counter (SR400,
Stanford Research Systems, Sunnyvale, California, USA). The excitation beam passed
118
Modelocked Ti:Sapphire
laser
Argon ion pump laser
(multiline 488 - 502 nm)
Spectrum Analyzer
BG 39
Photon Counter
PMT
SHG crystal
λ/2
Monochromator
P
PMT
slit
Autocorrelator
Calibrated power
meter
chopper
BE
sample
N.A. = 0.4
Figure 6.1: Cross-section measurement setup. BG 39 = Schott glass absorption filter, PMT
= photomultiplier tube, SHG = second harmonic generating BBO crystal (CASTECH),
λ/2 = half wave plate, P = Glan-Thompson polarizer, BE = beam expander (5X).
119
through a chopper which provided a reference signal enabling the photon counter to perform background subtracted integration. Average power levels of the incident light were
also monitored by refocusing the excitation beam onto a calibrated power meter (Newport
Corporation, Irvine, California, USA).
One centimeter thick sample cells were used to ensure that the entire focal volume of
the excitation light was contained within the sample. We accounted for small absorptive
losses by measuring the location of the focal volume within the cuvette and using linear
absorption theory with the known molar extinction coefficients ² of the fluorophores. For
all measurements the linear absorption was negligible (i.e. its effects were much smaller
than the systematic uncertainty in the measurements).
6.1.3 Calculation of Collection Efficiency
Because the monochromator grating efficiency is wavelength-dependent, it was necessary
to measure ϕ for each wavelength setting used. ϕ depends on the product of several factors: photomultiplier tube quantum efficiency, solid angle of collection (expressed as a
percentage of 4π radians), monochromator transmission, quartz cuvette transmission, and
the fluorophore emission spectrum. Monochromator transmission was measured using
a white light source emitting into the 90-degree fluorescence collection path to mimic
fluorescence emission. To this end, white light was focused at the point of two-photon
excitation in the sample cuvette, and care was taken to ensure that all subsequently diverging light was contained within the fluorescence collection solid angle (a function of
120
fluorescence collection optics and the acceptance numerical aperture of the monochromator). First, the reference intensity was measured without the monochromator in the optical
path; then the monochromator was placed in the position used throughout the entire experiment. A spectrophotometer (Ocean Optics, Dunedin, Florida, USA) replacing the
fluorescence-collection PMT was used to generate transmission curves as the monochromator scanned over all relevant wavelength settings. The uncertainty in calculation of
ϕ was by far the largest source of systematic error. This error reflects uncertainty in
monochromator transmission as a function of wavelength, and errors due to numerical integration of the transmission spectrum, since the fluorescein reference and experimental
sample were typically measured at different monochromator settings. By repeating the
monochromator transmission measurement procedure using a focused white light source,
we obtained a standard deviation for this aspect of the measurement. Error introduced by
the numerical integration of the dye emission spectrum was estimated by comparing the
results of numerical integration while varying (1) numerical method, (2) interpolation resolution, and (3) integration limits. We estimated the systematic error in measurements of
two-photon cross sections to be ±30%, comparable to the uncertainty reported by [4]. This
number represents the standard deviation obtained as a result of propagating estimated errors in both sample and reference fluorescence collection efficiencies, ϕS and ϕR . Other
significant sources of systematic error were constant for both sample and reference, and
thus cancelled out.
121
6.1.4
σ2 Measurement Procedure
For each wavelength, the fluorescein reference sample was first measured, and then the
rest of the samples were measured. Each measurement was followed by a blank sample
in order to measure background stray light. Then the fluorescein reference was measured
again. The monochromator was tuned to the peak of the emission spectrum for each fluorophore. Before placing the sample in the cuvette holder, the average excitation power was
measured with the calibrated power meter. For each sample at each wavelength, integrated
fluorescence photon counts from the SR400 and the averaged SHG signal were recorded.
Background counts were subtracted from the total integrated fluorescence photon counts.
Integration times were 20 - 40 sec. Intensity tests confirming the power-squared dependence of the fluorescence were performed by measuring the integrated fluorescence at
various excitation intensities. Excitation intensities were measured with the calibrated
power meter. The results were plotted on a log-log scale and fit to a line, the slope of
which yielded the power-dependence exponent.
6.2 Near Infrared Two-Photon Excitation Cross-sections
of Voltage-Sensitive Dyes
Microscopy based on voltage-sensitive dyes has proven effective for revealing spatiotemporal patterns of neuronal activity in vivo and in vitro. Two-photon microscopy using
122
voltage-sensitive dyes offers the possibility of wide-field visualization of membrane potential on sub-cellular length scales, hundreds of microns below the tissue surface. Very
little information is available, however, about the utility of voltage-sensitive dyes for twophoton imaging purposes. Here we report on measurements of two-photon fluorescenceexcitation cross-sections for nine voltage-sensitive dyes in a solvent, octanol, intended to
simulate the membrane environment. Ultrashort light pulses from a Ti:sapphire laser were
used for excitation from 790 nm to 960 nm, and fluorescein dye was used as a calibration
standard. Overall, dyes RH795, RH421, RH 414, di-8-ANEPPS, and di-8-ANEPPDHQ
had the largest two-photon excitation cross-sections in this wavelength region and are
therefore potentially useful for two-photon microscopy. Interestingly, di-8-ANEPPDHQ,
a chimera constructed from the potentiometric dyes RH795 and di-8-ANEPPS, exhibited
larger cross-sections than either of its constituents.
6.2.1 Background
Voltage-sensitive dyes have proven to be effective for measuring electrical activity in neurons in vitro [43, 80, 169, 201] and in vivo [82, 145, 149]. However, in all these studies optical sectioning is inherently limited by one-photon fluorescence microscopy techniques;
i.e. the images are sensitive to fluorescence from the entire depth of focus and do not explicitly reject out-of-focus light. Two-photon laser scanning microscopy [49] enables true
three-dimensional imaging because of its intrinsic optical sectioning properties. Additionally, high contrast images can be obtained from deeper within biological tissues because
123
there is less scattering in tissue at the longer excitation wavelengths used in two-photon
excitation.
Albota, Xu and Webb have measured 2-photon excitation cross-sections for a variety of biologically relevant molecular fluorophores [4, 195, 196]. Very little information, however, exists on the suitability of voltage-sensitive dyes for two-photon imaging purposes [90]. To this end, we have analyzed the two-photon spectral properties
of some of the most common voltage-sensitive dyes, including recent novel dyes di-8ANEPPDHQ [140] and RH1692 [175]. We also included in our study Nile Blue A, which
is a lipophilic membrane-permeant potentiometric dye [16, 40, 191]. Using a ratiometric
method (Albota et al., 1998) with fluorescein as a reference, two-photon excitation crosssections were obtained for nine voltage-sensitive dyes at incident wavelengths ranging
from 790 nm to 960 nm. We used octanol as a solvent to approximate the environment
of these dyes when bound to membrane [177] and we identified dyes with comparatively
high cross-sections in this spectral region. We also investigated a novel naphthylstyrylpyridinium “chimera” dye di-8-ANEPPDHQ, which is a combination of an RH795 [82]
quaternary ammonium head group and a di-8-ANEPPS [15] chromophore; we found its
two-photon excitation cross-section in this spectral region was larger than that of either of
its constituent components.
124
6.2.2 Sample Preparation
Sample preparation and the sample sources are listed in Table 6.1. Selected chemical structures are shown in Fig. 6.2. Large volumes of high-concentration stock solutions were prepared and then diluted to ∼100 µM. The reference sample was fluorescein in H2 O at pH 13. Two-photon cross-sections are known for this reference sample [4, 195]. For all other samples, octanol was used as a solvent. For solubility reasons,
some samples were first prepared at high concentration in other solvents before adding
octanol. For example, RH1692 was first dissolved in a small quantity of ethanol; di-8ANEPPS and di-8-ANEPPDHQ were first prepared in a highly concentrated stock solution of DMSO/Pluronic. In these cases, solvents other than octanol make up less than 1%
of the total solution volume.
Fluorophore
di-8-ANEPPS
di-8-ANEPPDHQ
RH795
RH414
RH237
RH421
RH1692
Merocyanine 540
Nile Blue A
Source
L. Loew
L. Loew
Molecular Probes
Molecular Probes
Molecular Probes
Molecular Probes
R. Hildesheim
Eastman Kodak†
Chroma Gesellschaft‡
λabs/em
433 / 625
500 / 620
530 / 640
525 / 636
550 / 676
532 / 648
600 / 680
533 / 615
649 / 660
Concentration (µM)
90
16
95
150
142
97
117
128
118
Reference
[15]
[140]
[82]
[81]
[76]
[80]
[175]
[47]
[191]
Table 6.1: Summary of voltage-sensitive dye sample sources and concentrations
Sample summary. Absorption and emission wavelengths are experimentally measured
values in octanol. †Eastman Kodak, Rochester, New York ‡Chroma Gesellschaft-Schmid
& Co., Stuttgart, Germany.
125
Figure 6.2: Chemical structures of voltage-sensitive dyes used in this investigation.
6.2.3 Results
Two Photon Excitation Cross-sections TPE cross-section measurements, σ2 , for all samples are shown in Fig. 6.3. Cross-section values are in units of the Gppert-Mayer (GM),
where 1 GM = 10−50 ·cm4 ·sec·photon−1 . Values at selected wavelengths are listed in Table
6.2. All of the blue-green-absorbing dyes (RH795, RH414, RH421, RH237, di-8 dyes,
and Merocyanine 540) exhibit increases in cross-section between 900 nm and 960 nm.
This is almost certainly because the excitation photon wavelength approaches twice the
peak one-photon absorption wavelength of the blue-green-absorbing dyes. In fact, all of
the blue-green-absorbing dyes except for RH237 exhibited increases of over an order of
magnitude as λex varied from 790 nm to 960 nm. Both Nile Blue A and RH1692, the two
red-absorbing dyes in the study, exhibit an overall decrease in cross-section between 790
126
nm and 960 nm. The high values near 790 nm are most likely due to one-photon absorption in the red absorption tail, which extends to 750 nm in both dyes. Very little spectral
discontinuity was observed when changing optics sets.
Dye
di-8-ANEPPS
di-8-ANEPPDHQ
RH795
RH414
RH421
RH237
RH1692
Nile Blue A
Merocyanine 540
800 nm
1.4 ± 0.4
3.0 ± 0.9
0.6 ± 0.2
0.10 ± 0.03
1.9 ± 0.5
8.9 ± 2.6
1.5 ± 0.4
0.6 ± 0.2
0.27 ± 0.08
840 nm
1.7 ± 0.5
4.0 ± 1.2
1.0 ± 0.3
0.69 ± 0.21
10 ± 3.0
2.3 ± 0.7
–
0.13 ± 0.04
0.83 ± 0.25
880 nm
2.5 ± 0.8
8.8 ± 2.6
2.5 ± 0.8
1.6 ± 0.5
4.0 ± 1.2
0.42 ± 0.1
0.27 ± 0.08
0.24 ± 0.07
0.41 ± 0.1
920 nm
5.6 ± 1.7
16 ± 4.8
6.3 ± 1.9
4.6 ± 1.4
13.6 ± 4.1
0.84 ± 0.2
0.68 ± 0.2
–
0.74 ± 0.2
960 nm
10 ± 3.0
19 ± 5.9
10 ± 3.0
12 ± 3.6
16 ± 4.8
3.9 ± 1.2
–
–
4.4 ± 1.3
Table 6.2: Values of σ2 for voltage-sensitive dyes at selected wavelengths
Cross-sections are given in GM (1 GM = 10−50 ·cm4 ·sec·photon−1 ), and dashes represent
measurements with insufficient signal-to-noise.
6.2.4 Linearity Tests
Table 6.3 reports the results of linearity tests for several fluorophores. The deviation from
a pure intensity-squared law is within ±6% for all samples. These results confirm that the
measured fluorescence emission was indeed true two-photon induced fluorescence.
6.2.5 Discussion
Comparison with One-Photon Absorption Spectra When plotted as a function of λex /2, all
of the TPE cross-section spectra mimic the trends of their one-photon absorption spectra.
Di-8-ANEPPS was the only fluorophore for which the two-photon absorption peak was
127
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Wavelength (nm)
Figure 6.3: Values of σ2 are plotted in units of 10−50 ·cm4 ·sec·photon−1 as a function of
wavelength (nm). (•) mid-wavelength optics set (790 nm - 900 nm), (H) long-wavelength
optics set (900 nm - 960 nm). Systematic error for σ2 is ±30% for each point (not shown
on graphs)
Dye
RH795
RH421
di-8-ANEPPS
di-8-ANEPPDHQ
Slope
2.05 ± 0.06
1.96 ± 0.06
2.06 ± 0.08
1.95 ± 0.08
Table 6.3: Linearity tests for power-square fluorescence dependence of voltage-sensitive
dyes.
Values are slope of logarithmic plot of fluorescence (A.U.) as a function of peak excitation
intensity at 920 nm.
128
within our tuning range. A small peak in TPE cross-section at 850 nm is observed in
this case, which is likely a blueshifted correlate of twice the one-photon absorption peak
expected at 866 nm. Other non-resonance features in Fig. 6.3 are also likely to be physical
trends corresponding to one-photon absorption spectra. For instance, the small peak at 425
nm in the one-photon absorption spectrum of Nile Blue A is found in the TPE spectrum
as a blueshifted peak at 830 nm (Fig. 6.4).
2-Photon Excitation Wavelength (nm)
760
0.35
800
840
880
920
960
1000
1-photon a bsorption
2-photon a bsorption
2
0.3
1
0.25
0.2
380
400
420
440
460
480
TPE Cross-Section ( GM)
One-Photon Absorbance
Nile Blue
0
500
1-Photon Excitation Wavelength (nm)
Figure 6.4: Absorbance (defined as − log10 (I/I0 ), where I/I0 is the sample transmittance) and σ2 (in units of 10−50 ·cm4 ·sec·photon−1 ) are plotted as a function of excitation wavelength (nm) (bottom and top axes correspond to one- and two-photon excitation
wavelengths, respectively).
A striking result in this study is the greatly increased cross-section of the chimeric
molecule di-8-ANEPPDHQ compared with that of its constituent components (Fig. 6.5).
No peak in the TPE spectrum in this range is observed, and none is expected since the
one-photon absorption maximum is at 500 nm. However, the chimera TPE cross-section
129
values were double those of RH795 and di-8-ANEPPS over the entire tuning range. Di8-ANEPPDHQ has recently been shown to offer advantages for imaging of mammalian
neuronal networks (Obaid et al., 2004). Combined with the fact that 790 nm to 960 nm
is a typical tuning range for most of the commercially available laser sources for twophoton microscopy, di-8-ANEPPDHQ shows significant promise for two-photon voltagesensitive dye imaging.
25
di-8-ANEPPDHQ
di-8-ANEPP
S
RH 795
10-50cm4s / photon
20
15
10
5
0
780
800
820
840
860
880
900
Wavelength (nm)
920
940
960
Figure 6.5: Comparison of σ2 for di-8-ANEPPDHQ, RH795, and di-8-ANEPPS. Solid
line = di-8-ANEPPDHQ, dashed line = di-8-ANEPPS, dotted line = RH795. (•) midwavelength optics set (790 - 900nm), (H) long-wavelength optics set (900 - 960 nm)
6.2.6 Prospects for Two-Photon Imaging using VSDs
Fluorescence-based scanning microscopy measures a relatively small number of photons
per image pixel compared with reflectance or transmission bright-field video microscopy.
130
Therefore to assess the practical utility of two-photon microscopy using voltage-sensitive
dyes for measurement of membrane potential, we consider the case of shot-noise limited
measurements. Fractional fluorescence (∆F/F ) response following two-photon excitation of voltage-sensitive dyes varies considerably between dyes [90, 110]. While these
studies have reported ∆F/F changes from 5%–40% per 100 mV change in the membrane potential of cultured neurons in vitro, for the purpose of a conservative estimate
of expected signal-to-noise ratio, we assume that with some expected non-specific dyebinding, the in vivo ∆F/F is ∼ 1% (as in one-photon fluorescence microscopy). To detect fractional fluorescence changes of 1% in a single measurement with a signal-to-noise
ratio of greater than 1, it is necessary to detect > 104 photons during the physiologically relevant sampling time. Using Equation 6.3, we calculated the number of collected
photons for a realistic epi-fluorescence microscope system (integration time was 100 µs,
average power at objective focus was 10 mW, fluorescence collection efficiency was 10%,
excitation wavelength was 930 nm, membrane-bound dye concentration was 10 µM, TPE
cross-section was 16 GM, gp was 0.588, laser repetition rate was 76 MHz, and laser pulsewidth was 200 fsec). Such a system would be capable of detecting 1% changes fluorescence with a signal-to-noise ratio of ∼2. This calculation suggests that the most suitable
voltage-sensitive dyes for two-photon microscopy of neural network responses are di-8ANEPPDHQ, di-8-ANEPPS, RH795, RH414 and RH421 at excitation wavelengths above
930 nm.
131
6.2.7 Conclusion (VSD)
We have measured the TPE cross-sections in the range of 790 nm to 960 nm of seven commercially available voltage sensitive dyes and two novel dyes obtained from other laboratories. Several of the fluorophores exhibited blueshifted spectral features. The novel naphthylstyryl pyridinium chimeric probe di-8-ANEPPDHQ was found to have large crosssections compared with its constitutive structural elements, di-8-ANEPPS and RH795.
Work is underway to study the potentiometric properties of this set of voltage sensitive
dyes when excited via two-photon absorption.
6.3 Near-Infrared Two-Photon Cross-Sections of Novel Conjugated Porphyrin Compounds
We report the measurement of near-infrared two-photon absorption (TPA) cross sections
for a novel series of conjugated (porphinato)zinc(II) compounds. The cross-sections were
measured ratiometrically using fluorescein as a reference. While apparently very large
TPA cross-section values (up to ∼ 104 × 10−50 ·cm4 ·sec·photon−1 ) were measured in the
two-photon Soret (or B-band) resonance region, measurements of fluorescence as a function of incident photon flux indicate significant one-photon absorption in the same region
for all of the compounds in the series. Extraction of TPA cross-sections using a model with
both one- and two-photon absorption for selected samples yielded TPA cross-section values of ∼ 10 GM. This study presents a detailed investigation of the one- and two-photon
132
absorption contributions to these signals in the Zn-based porphyrins. The findings suggest
that large TPA cross-sections reported in the Soret resonance region of similar compounds
likely contain significant contributions from one-photon absorption processes.
6.3.1 Introduction
New materials with large two-photon absorption cross-sections are critical for a variety
of optical and electro-optical technologies. Recent two-photon absorption (TPA) applications include optical data storage [46], fluorescence microscopy [49], and lithographic microfabrication [125]. Porphyrins are among the most highly polarizable organic molecules
reported to date, and because of their enhanced nonlinear susceptibilities [5, 114], porphyrins are promising molecules for use as optical limiters [128], for photodynamic therapy [23], for targeted delivery of drugs, and as contrast agents for in vivo imaging [74].
Conjugated porphyrin molecules possess large third-order nonlinear susceptibilities
[6, 141], χ(3) . The imaginary part of χ(3) is responsible for the process of two-photon
absorption, and a number of recent studies have reported enormous (up to ∼ 104 ×
10−50 ·cm4 ·sec·photon−1 ) two-photon absorption cross-sections in conjugated porphyrin
molecules [55, 56]. The linear absorption spectrum of porphyrins consists of a strong
transition to the second excited state (S0 −→S2 ) at UV - visible wavelengths, called the
Soret or B-band, and a weaker transition to the first excited state (S0 −→S1 ) at visible
- near-IR wavelengths, called the Q-band. Large molecular cross-sections have been reported in the two-photon resonance of the Soret band, and resonance enhancement due
133
to Q-band transitions has been suggested to be the root cause of these large two-photon
cross-sections [53, 55, 56, 101].
Typically, in the two-photon Soret resonance region of conjugated porphyrin monomers
and dimers, one-photon absorptivity is negligible. For two-photon absorption-based applications, however, the extremely high peak illumination intensities provided by modelocked lasers can lead to significant one-photon absorption as well as other photophysical processes such as excited-state absorption and photobleaching. In fact, several investigations of the linear and nonlinear optical properties of porphyrins and conjugated
porphyrins have reported sizeable excited-state absorption effects at relatively low excitation intensities [17, 84, 194]. While large TPA cross-sections are crucial for TPA-based
applications, the simultaneous presence of one-photon absorption or excited-state absorption processes can undo these gains. Two-photon fluorescence microscopy, in particular,
will suffer undesirable increases in excitation volume and photobleaching [59, 180] in the
presence of such processes. The two-photon resonance of the Soret band in porphyrins
(∼800–900nm) is of particular interest because of the large reported TPA cross-sections
and because this wavelength range represents a region readily available to the modelocked
Ti:Sapphire laser.
We present results of wideband TPA cross-section measurements in the wavelength
region from 750 nm to 960 nm. We explore a series of novel (porphinato)zinc(II) compounds. Additionally, we present fluorescence linearity-tests which indicate the relative contributions of one- and two-photon absorption. Extremely large absorption crosssections were measured, consistent with previous reports; however, our results indicate
134
that for the vast majority of these porphyrins, significant linear absorption in the twophoton resonance region corresponding to the Soret band is present. While the nonlinear
absorption component increases in some porphyrin monomers and dimers at wavelengths
progressively longer than the Q-band absorption tail, the effect is qualitatively a small
perturbation on an otherwise linear dependence on excitation power. Our results therefore suggest that in the absence of clearly demonstrated intensity-squared absorption dependence, the enormous TPA cross-section values reported in similar compounds likely
reflect significant one-photon absorption contributions in the Soret resonance region.
6.3.2 Experimental
6.3.2.1 Materials
Synthesis and characterization data for the compounds in this study are described in [185],
and structural diagrams for the 15 porphyrin compounds investigated are shown in Fig.
6.6. Porphyrin samples were prepared at concentrations of ≤ 100 µM in tetrahydrofuran
solvent, then experimentally verified based on absorptivity measurements performed after two-photon cross-section measurements. The reference sample for TPA cross-section
measurements was fluorescein in H2 O at pH 13. Two-photon cross-sections are known for
this reference sample [4, 195].
135
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
N N
TMS
TMS
Zn
N N
O
O
O
O
O
O
O
O
O
O
O
O
O
O
N N
N N
N N
N N
N N
N N
N N
N N
N N
N N
Zn
O
Zn
Zn
O
O
O
O
Zn
O
DD
Zn
N N
O
O
C3F7
N N
Zn
O
O
O
O
O
O
O
O
O
O
DDD
PZnE2
TMS
O
O
TMS
N N
N N
N N
N N
Zn
C3F7
O
Zn
H
C3F7
O
O
O
C3F7
O
O
C3F7
N N
N N
N N
N N
N N
N N
Zn
O
Zn
Zn
C3F7
O
O
O
RfPZnE2
DAD
DA
S
N N
R
N N
R
S
N N
S
N N
N N
Zn
Zn
R
R
R
N N
N N
N N
Zn
Zn
N N
N N
N N
N N
N N
R
R
R
R
R
(BTDt)2PZn
R
R
Zn
N N
N N
Zn
N
N
R=
R
R
O
D-P-A
R
R
R
R
R
N N
N N
N N
N N
N N
N N
N N
N N
N N
N N
R
R
R
R
R
Zn
Zn
Zn
N
Zn
N
R
R
R
N N
N N
N N
N N
N N
N N
R
R
R
Zn
Zn
NO2
D-PP-A
D-PPP-D
Zn
NO2
N N
O
D-P-D
N
O
PZnEEPZn
PZnEBTDEPZn
N N
N
O
R=
Zn
N
Zn
R
R
R
N N
N N
N N
N N
N N
N N
R
R
R
Zn
Zn
Zn
NO2
D-PPP-A
TBT
Figure 6.6: Chemical structures of novel conjugated (porphinato)zinc(II) compounds measured in this study.
136
6.3.2.2 Measurement Procedures
The background theory and procedure used for measuring two-photon excitation crosssection spectra in this study has been previously described [66] (see Section 6.1 for details). Briefly, the measurement technique directly measures the two-photon excitation
(TPE) cross-section, σ2−photon , which is equal to the product of the two-photon absorption (TPA) cross-section, Σ2 , multiplied by the two-photon fluorescence quantum yield,
q2 : σ2−photon = q2 Σ2 . The two-photon fluorescence quantum yield, q2 , is defined as
the number of emitted photons by the molecule per pair of absorbed photons. Values of
TPA and TPE cross-sections therefore both have the units of cm4 sphoton−1 , but the TPA
cross-sections can be substantially higher, depending on q2 . The informal unit for Σ2 and
σ2−photon is the Göppert-Mayer (GM), where 1 GM = 10−50 ·cm4 ·sec·photon−1 . In this
study, one-photon fluorescence quantum yield data was available for 9 out of 15 samples,
and corresponding TPA cross-sections were calculated assuming that the one-photon and
two-photon fluorescence quantum yields were equivalent, i.e. q1 = q2 . Results are specified as either TPA or TPE cross-section on the y-axis labels of Fig. 6.7. Tests of the input
power dependence of the fluorescence were performed by measuring the integrated fluorescence as a function of excitation intensity. Excitation intensities were measured with
the calibrated power meter. The results were plotted on a log-log scale and fit to a line; the
slope of this line yielded the power-dependence exponent.
137
6.3.2.3 Extraction of TPE Cross-Sections from Power-law tests
The data yielded by the power-law tests can be used to derive approximate values for the
two-photon excitation (TPE) cross-section [120]. Assuming only one- and two-photon
absorption, the time-averaged fluorescence signal detected by the system, hF (t)i (photons
s−1 ) is the sum of the respective fluorescence contributions, i.e.
hF (t)i = hF (t)i1−photon + hF (t)i2−photon .
(6.4)
The detected time-averaged fluorescence signal, hF (t)i, is related to the incident timeaveraged power hP (t)i (photons s−1 ) through one- and two-photon absorption coefficients
c1 and c2 , respectively:
hF (t)i = c1 hP (t)i + c2 hP (t)i2 .
(6.5)
The coefficients c1 and c2 can be solved for by fitting the power-law test data (Figs. 6.8
& 6.9) to the function in Eq. 6.5. The coefficients c1 and c2 can, in turn, be used to solve for
the one- and two-photon absorption cross sections utilizing known relationships between
input power hP (t)i and fluorescence contributions hF (t)i1−photon and hF (t)i2−photon . For
one-photon absorption, the fluorescence signal detected is linearly dependant on the excitation intensity over the focal volume, i.e.
Z
F (t)1−photon = q1 φσ1−photon C
I(~r, t)dV,
V
138
(6.6)
where q1 is the one-photon quantum yield, φ is the system fluorescence collection efficiency, σ1−photon is the one-photon absorption cross section, C is the sample concentration
(cm−3 ), and excitation intensity, I(~r, t), is integrated over the focal volume. We assume
that the sample concentration is spatially homogeneous. For one-photon absorption, the
integral of I(~r, t) can be approximated as the time-varying intensity at the focus, I(t),
multiplied by the focal volume, Vf ocal , thus
F (t)1−photon ≈ q1 φσ1−photon CVf ocal I(t).
(6.7)
By approximating Vf ocal as a cylinder with diameter equal to the diffraction-limited
focal spot of a Gaussian beam and length equal to the focal depth, we can relate the timeaveraged detected fluorescence to the input power as follows:
F (t)1−photon = c1 hP (t)i ≈ q1 φσ1−photon C
0.4186π 2 λn
hP (t)i,
N.A.2
(6.8)
where N.A. is the numerical aperture of the objective, and n is the linear refractive
index at excitation wavelength λ. Here, we have defined the intensity at the focal point as
hI(t)i =
πN.A.2
hP (t)i.
λ2
Equating the coefficients of hP (t)i, we can solve for c1 in terms
of the spectroscopy system properties, i.e.
c1 = q1 φσ1−photon C
0.4186π 2 λn
,
N.A.2
(6.9)
and we can rearrange Eq. 6.9 to solve for the one-photon absorption cross-section,
139
σ1−photon , yielding
σ1−photon =
c1
N.A.2
×
.
q1 φC 0.4186π 2 λn
(6.10)
The one-photon absorption cross-section can be used to calculate the molar absorptivity, ² (mol−1 cm−1 ), by using the relationship ² =
σ1−photon C
,
2.303 Cm
where Cm is the molar
concentration. ² was measured as a function of excitation wavelength for all of the compounds in this study.
Additionally, for mode-locked laser sources, the time-averaged two-photon fluorescence signal hF (t)i2−photon detected by a spectroscopic system with collection efficiency
φ and time-averaged illumination power hP (t)i is given by Eq. 6.3, which we now write
side-by-side with the nonlinear component of Eq. 6.5:
1
gp 8nhP (t)i2
hF (t)i2−photon = c2 hP (t)i2 ≈ φq2 Σ2 C
.
2
fτ
πλ
(6.11)
Equating the coefficients of hP (t)i2 in Eq. 6.11, we can show that
1
gp 8n
c2 ≈ φq2 Σ2 C
,
2
f τ πλ
(6.12)
and can then solve for the TPE cross-section, q2 Σ2 ,
q2 Σ2 = σ2−photon = 2c2
f τ 1 πλ 1
,
gp C 8n φ
(6.13)
where all parameters on the RHS are known except for the system collection efficiency,
140
φ, which must be estimated. The combination of error in the data fit and the uncertainty
in φ makes this method of calculating σ2−photon somewhat error-prone compared with the
ratiometric measurement approach used in this report. However, the reference sample can
still be employed to approximate φ using Eq. 6.3, since all values except for φ are known.
6.3.3 Results
6.3.3.1 TPA Cross-Section Spectra
Two-photon absorption and emission cross-section spectra for all 15 compounds are plotted as a function of excitation wavelength in Fig. 6.7. Most of the compounds exhibit large
cross-section values at the shortest wavelengths. Nonlinear absorption features clearly
corresponding to the two-photon Soret resonance are found in three of the monomers
(RfPZnE2 and (BTDt)2 PZn, and D-P-D). These nonlinear absorption features are clearly
blueshifted in RfPZnE2 . Nonlinear absorption features exhibiting a slight plateau before
rising to very high values at shorter wavelengths are found in one monomer (D-P-A),
three dimers (DA, PZnEEPZn, and D-PP-A), and one trimer (D-PPP-A). These plateaus
occur at wavelengths slightly shorter than two times the one-photon Soret resonance wavelength. Increases in nonlinear absorption toward the longest wavelengths were observed
in the monomer D-P-A, the dimer PZnEEPZn, and the trimer DDD.
141
6
20
35
RfPZnE2
10
5
(BTDt)2PZn
30
-1
15
Σ2 (10-50cm4 s photon )
-1
q2Σ2 (10-50cm4 s photon )
-1
Σ2 (10-50cm4 s photon )
PZnE2
4
2
25
20
15
10
5
0
780
820
860
900
Wavelength (nm)
940
0
980
5
x 10
780
820
860
900
Wavelength (nm)
-1
500
0
860
900
Wavelength (nm)
940
780
980
820
860
900
Wavelength (nm)
940
980
940
980
2
x 10
820
860
900
Wavelength (nm)
5
DAD
2.5
Σ2 (10-50cm4 s photon )
-1
Σ2 (10-50cm4 s photon )
-1
Σ2 (10-50cm4 s photon )
-1
2000
1500
1000
500
0
780
820
860
900
Wavelength (nm)
940
5000
780
820
860
900
Wavelength (nm)
940
980
780
820
860
900
Wavelength (nm)
940
25
0
980
6000
780
820
860
900
Wavelength (nm)
940
2000
780
820
860
900
Wavelength (nm)
940
980
780
820
860
900
Wavelength (nm)
940
980
940
980
2500
D-PPP-D
-1
5000
4000
3000
2000
1000
0
D-PP-A
0
980
q2Σ2 (10-50cm4 s photon )
q2Σ2 (10-50cm4 s photon-1)
4000
860
900
Wavelength (nm)
500
6000
D-PPP-A
820
q2Σ2 (10-50cm4 s photon-1)
D-P-A
0
780
1000
-1
q2Σ2 (10-50cm4 s photon )
Σ2 (10-50cm4 s photon-1)
D-P-D
1
0
0
980
20
2
1.5
0.5
50
q2Σ2 (10-50cm4 s photon-1)
940
4
780
DDD
10000
980
6
3
PZn-EE-PZn
940
0
980
3000
2500
860
900
Wavelength (nm)
PZn-BTD-PZn
Σ2 (10-50cm4 s photon )
1
820
820
4
x 10
8
-1
2
780
780
DD
Σ2 (10-50cm4 s photon )
-1
Σ2 (10-50cm4 s photon )
0
980
1000
DA
0
940
0
780
820
860
900
Wavelength (nm)
940
980
TBT
2000
1500
1000
500
0
780
820
860
900
Wavelength (nm)
Figure 6.7: Two-photon absorption (TPA), Σ2 , and excitation (TPE), q2 Σ2 , cross-sections
measured for all 15 porphyrin compounds. Labels on y-axes indicate either Σ2 (TPA)
or q2 Σ2 (TPE). Solid lines: one-photon absorptivity (A.U., y-axis scale not shown) as a
function of 2× the incident laser excitation wavelength (nm). Dashed lines connecting
data points: two-photon absorption or emission cross-sections (10−50 cm4 s photon−1 ) as
calculated using a ratiometric emission-based measurement technique. Error bars (±30%)
represent the estimated systematic error.
142
6.3.3.2 Power-law Tests
The input power dependence studies for fluorescence were performed at various laser excitation wavelengths (for selected samples). The results are shown in Fig. 6.8. A slope
of one on a log-log scale indicates purely linear absorption, and a slope of two indicates
two-photon absorption. A slope between one and two is an indication that both one- and
two-photon processes are occurring. Significant changes in the fitted slope as a function of wavelength were observed in the four compounds shown in Fig. 6.8: monomers
(BTDt)2 PZn and D-P-A, and dimers PZnEEPZn and DD. The largest slope values measured among all samples at any wavelength was m = 1.4, measured in PZnEEPZn at 800
nm and DD at 825 nm. For all other samples, the measured slopes were equal to 1 within
the estimated fit error (±0.1). For three out of these four compounds exhibiting a mixture
of linear and nonlinear absorption (all but PZnEEPZn), incrementally higher slope fit
values were found at longer wavelengths (∼800 nm–825 nm) with respect to the shortest
wavelength studied (750 nm). The slopes did not continue to increase beyond 825 nm,
and at the longest wavelength studied (960 nm), the absorption rates for these compounds
were too small to provide sufficient signal-to-noise ratio for cross-section measurements.
While the other larger compounds did have appreciable levels of absorption up through
960 nm, the fitted slopes were equal to 1 within the estimated fit error.
Increased TPA as a result of self-aggregation has been measured experimentally in
similar porphyrin compounds [100], thus we performed power-law tests at various sample
concentrations. Fig. 6.9 shows the results of fluorescence emission versus input power at
143
Figure 6.8: Power-law tests for fluorescence dependence on incident excitation laser beam
intensity performed at varying laser excitation wavelength. Straight dotted lines show fits
to Eq. 6.5 of the measured data points on a log-log scale plot. Values of excitation laser
wavelength (nm) are indicated in the chart legend. Slope fit values (m) are indicated above
each dotted line. Inset: (data points on dashed line) Measured two-photon excitation
(TPE) cross-section spectrum (GM) as a function of wavelength (nm); (solid line) onephoton absorptivity (A.U., y-axis scale not shown) as a function of 2× the incident laser
excitation wavelength (nm).
144
different sample concentrations for two compounds, with the excitation wavelength held
constant. While the monomer D-P-A shows little significant change in power-law from
slope fit, the dimer DD does show a significant increase in nonlinear absorption at a higher
concentration (m = 1.4 at a concentration of 1.63 × 10−4 M).
Figure 6.9: Power-law tests for fluorescence dependence on incident excitation laser beam
intensity performed at varying sample concentration. Straight dotted lines show fits to Eq.
6.5 of the measured data points on a log-log scale plot. Values of sample concentration
are indicated in the chart legend. Slope fit values (m) are indicated above each dotted line.
Inset: compound structural diagram.
6.3.3.3 Extracted values of ² and σ2−photon
For the samples with the largest power-law slopes (DD, PZn-EE-PZn, and (BTDt)2 PZn),
extracted ² and σ2−photon (= q2 Σ2 ) are tabulated in Table 1. The three samples exhibited
extremely small one-photon absorption at the excitation wavelengths; consequently values
of ² measured via conventional spectrophotometer are presented in order of magnitude
range due to comparatively large shot noise in the absorptivity measurements. Extracted
145
² for compounds DD and (BTDt)2PZn was equal to these measured values, while the
extracted ² for PZn-EE-PZn was an order of magnitude less than the measured value.
All three samples had values of extracted q2 Σ of order 10 GM. All extracted values had
relatively large estimated uncertainty (50%) due to approximation errors.
Compound λex (nm)λem (nm)Slope
DD
825
PZn-EE-PZn800
(BTDt)2PZn 790
713
732
689
Extracted ²
(mol−1 cm−1 )
1.4 ± 0.1 24 ± 12
1.4 ± 0.1 29 ± 15
1.3 ± 0.1 36 ± 18
Measured ²
(mol−1 cm−1 )
∼101 –102
∼102
∼101
Extracted
q2 Σ2 (GM)
10 ± 5
25 ± 13
17 ± 9
Table 6.4: Extracted values of TPA cross-sections for selected compounds.
Data from Figs.6.8 and 6.9 was fit to the function hF (t)i = c1 hP (t)i + c2 hP (t)i2 and
values of ² and σ2−photon in extracted values were calculated from c1 and c2 , respectively.
Systematic error of 50% in extracted values represents uncertainty in collection efficiency
and data fitting. Measured values of ² are presented in order of magnitude range due to
comparably large shot noise in absorptivity measurements at these excitation wavelengths.
6.3.4 Discussion
The largest TPA cross-sections measured in our study are on the order of 104 –105 GM,
comparable to the largest values measured to date via ratiometric methods in similar porphyrin oligomers. Additionally, the general spectral trends of our measured cross-section
plots (Fig. 6.7) closely match those of studies on similar compounds [55, 56]. The results of our fluorescence power-law tests suggest, however, that a significant portion of the
detected fluorescence is the result of other absorptive processes.
By fitting the power-law test data to a model that assumed one- and two-photon absorption processes, we extracted TPE cross-section values that were on the order of 10 GM.
146
While such fitting techniques introduce fairly large errors [120], these values are likely
much closer to the true TPA contributions of these compounds since they take into account
the presence of simultaneous one-photon absorption. Extracted values of (one-photon)
molar absorptivity, ² (mol−1 cm−1 ), were equal to or within one order of magnitude difference from measured values (one-photon absorption spectra for these compounds typically
span 5 orders of magnitude between visible and near-infrared excitation wavelengths).
The order of magnitude difference between extracted and measured values for PZn-EEPZn may be due to sample decomposition, as the power-law tests were performed several
days later than the one-photon absorptivity measurements.
All methods of calculating two-photon absorption cross-sections based on measuring
intensity-dependent absorption or emission require assumptions about the relevance of different physical processes. Previous studies of this compound class have typically assumed
no one-photon absorption nor excited-state absorption, and the TPA and TPE cross-section
ratiometric measurement techniques used in this thesis and other studies (e.g. [54] [56])
require that two-photon absorption is the only significant source of absorption. Our powerlaw tests, however, clearly indicate that these are not sound assumptions.
The results of power-law tests as a function of wavelength (at constant concentration)
in Fig. 6.8 show significant departures from pure TPA even at wavelengths seemingly beyond the one-photon absorption tail, based on one-photon absorption spectra. The results
suggest that while there may indeed be large two-photon absorption at wavelengths approaching the two-photon Soret resonance, other processes such as one-photon absorption
and excited-state absorption must be taken into consideration. The finding of significantly
147
increased nonlinear absorption at higher concentrations in the dimer DD, but none in the
monomer D-P-A (Fig. 6.9) is consistent with reports of increased optical nonlinearities
due to self-aggregation effects in similar compounds [42, 48, 100].
6.3.5 Conclusion
The increasing interest in nonlinear optical applications of porphyrins has led to a demand
for accurate measurements of nonlinear optical properties. While certain optical applications (e.g. optical limiting devices) may not be hindered by the occurrence of excitedstate absorption, other applications (e.g. two-photon microscopy) necessitate the purely
power-squared dependence of TPA. From the viewpoint of synthesis, one of the goals of
investigating the nonlinear optical properties of porphyrins has been to enable predictions
about two-photon absorption patterns based on one-photon absorption spectra. The results
presented in this thesis suggest that while some general two-photon absorption properties
may be predicted, experimental care must be taken to consider other photophysical processes that may occur in these electronically complex molecules.
148
Chapter 7
Multifocal Two-Photon Microscopy
7.1 Introduction
Two-photon laser scanning microscopy [49] enables true three-dimensional imaging because of its intrinsic optical sectioning properties. Additionally, high contrast images
can be obtained from deeper within biological tissues compared to conventional and even
confocal microscopy [33]. Since its inception, two-photon laser scanning microscopy
has found wide-spread applicability throughout the field of neuroscience, both in vitro
[123,201] and in vivo [89,183,187,198]. Calcium Green, green fluorescent protein (GFP),
and other fluorophores with absorption peaks in the blue are among the most widely
used dyes for in vivo studies, in part because their peak absorption wavelength can be
reached by two photons within the tuning range of most commercially available pulsed
laser sources. Calcium-indicators yield large signals following two-photon excitation and
149
can reveal rapid intracellular Ca2+ concentration changes dependent upon action potentials [186].
Second harmonic generation (SHG) imaging [88, 173] also provides inherent optical sectioning, and has recently been used in conjunction with voltage-sensitive dyes
[30, 50, 131] to produce exceptional images of activity in cultured neurons. However,
because SHG is a parametric nonlinear process, the resulting second harmonic wave travels predominantly in the same direction as the incident light [130] and yields signal in the
reflectance direction only after backscattering. By contrast, even at the scale of a single
fluorophore, fluorescence emission following two-photon absorption of polarized light by
membrane-bound voltage-sensitive dyes leads to a symmetric dipole radiation distribution [113]. In sub-surface tissue imaging conditions, where the excitation volume is small
compared to the imaging depth, photons are generally assumed to be emitted isotropically [142]. These factors make two-photon microscopy preferable to SHG for backward
detection, a critical criterion for in vivo imaging.
7.2 Anatomical two-photon imaging
Two-photon microscopy was used for 3D anatomical investigation of near-surface tissues
in vivo and in vitro up to depths of ∼ 250 µm below the surface. All measurements
were performed with a multifocal two-photon microscope described in Chapter 2. Two
qualitatively different forms of 3D anatomical imaging have been performed: (1) imaging
150
of voltage-sensitive dye baseline fluorescence, and (2) imaging of enhanced Green Fluorescent Protein (EGFP) expression in specific sub-populations of neurons in genetically
modified animals.
7.2.1 Anatomical two-photon imaging in vitro
Imaging in vitro preparations (such as the ones described herein) is a lot easier than in
vivo imaging. First and foremost, in vitro preparations permit unsurpassed experimental
degrees of freedom. Pharmacological agents can be used without concern for overall
impact on an animal’s health, and, for the benefit of imaging experiments, tissue can be
cut or prepared to reveal ordinarily inaccessible tissue regions. In addition, following what
may be exceedingly intricate surgery and preparation, maintenance of health in many in
vitro preparations is less critical, if not simpler, than in vivo preparations (e.g. anesthesia
is not required). Furthermore, some in vitro preparations (e.g. slice preparations) afford
many individual experiments per single animal, reducing overall costs and reducing the
number of animals sacrificed.
All optical microscopy techniques for investigating the brain in vivo are limited to the
first few mm of the cerebral cortex due to scattering effects. This is a severe limitation on
the ability to understand the functional architecture of the brain, as all sensory information (except for that of olfaction) must pass through sub-cortical nuclei in the depths of the
brain and brainstem. Anatomical imaging of living thalamocortical brain slices provides
high-resolution functional imaging access to all of the cortical layers as well as connected
151
thalamic regions. When maintained in good physiological condition, slice preparations
remain electrically viable for electrophysiological measurements via electrode or optical
measurements. Slice preparations have been employed for numerous one-photon fluorescence measurements via voltage-sensitive dyes (e.g. [43, 107, 112, 201]) and for twophoton investigation via calcium sensitive dyes (e.g. [123, 183]).
7.2.1.1 Surgical preparation of thalamocortical slices
Sprague-Dawley rat pups aged P14-P24 were deeply anesthetized with isofluorane [sodium
pentobarbital (30–40 mg/kg)] and killed by decapitation. The forebrain was rapidly removed and 400-µm-thick thalamocortical slices were cut on a vibratome according to
Gibson et al. [75]. The ice-cold bathing medium contained 126 mM NaCl, 3 mM KCl, 2
mM MgSO4 , 1.25 mM NaH2 PO4 , 2 mM CaCl2 , 26 mM NaHCO3 , and 10 mM dextrose
and was aerated with 95% O2 / 5% CO2 .
Aeration of slices with 95% O2 / 5% CO2 was maintained at all times except briefly
during bath-application of voltage-sensitive dyes (see photograph of slice imaging setup
in Figure 7.1). In this case, live slices were incubated in 1–2 ml of dilute dye solution
(∼ 2–4 µM) of either RH795, di-3-ANEPPDHQ, or di-4-ANEPPDHQ (Molecular Probes
/ Invitrogen Corporation, Carlsbad, CA, USA) for 5–20 min, then washed with ACSF.
Calcium dye staining was performed following a pressure ejection-based dye delivery
protocol described by Stosiek et al. [183]. Calcium Green-1 AM ester (Molecular Probes
/ Invitrogen Corporation, Carlsbad, CA, USA) was dissolved at a concentration of 10 mM
in DMSO with 20% Pluronic F-127. A micropipette was filled with this solution and
152
inserted into the desired cortical region of a slice. A pressure pulse [1 min, 0.7 bar (1 bar
= 100 kPa); Picospritzer II, General Valve, Fairfield, NJ] was applied to the pipette to eject
∼ 400 fl femtoliters of the solution.
Figure 7.1: Photograph of chamber used for two-photon imaging of thalamocortical slices.
Stimulating electrode is seen here not yet inserted into the tissue.
7.2.1.2
In vitro Anatomical imaging via voltage-sensitive dye staining
Figure 7.2 displays a typical background fluorescence image produced from imaging di4-ANEPPDHQ-stained slices. Because the dye stains only the membrane and is not internalized, cells appear as dark holes in a fluorescent background. Larger cells, ∼ 10 µm in
diameter, are likely pyramidal cells. Processes rising through layers 1 and 2/3 are faintly
visible, as well.
Scanning in the Z-direction into the tissue in thalmocortical slices has the effect of
153
100 µm
slice edge
layer 2/3
(arrow indicates putative
pyramidal cell body)
layer 1
Figure 7.2: Multifocal background image displaying cortical layers. Expanded sub-region
(corresponding to an image generated by one particular excitation focus) shows cells visualized with high spatial resolution. The right-side edge in the main background image
(right) is the edge of the slice, thus the surface of the cortex.
sectioning rostro-caudally, keeping all layers in the same view. Figure 7.3 shows optically
sectioned images from three depths through slice tissue, revealing changes in cellular features. Blurring and cross-talk in images due to scattering became highly limiting at depths
greater than ∼ 200 µm into the tissue.
depth = -100 µm
100 µm
depth = -125 µm
100 µm
depth = -150 µm
100 µm
Figure 7.3: Multifocal two-photon images scanned from three focal depths through di-4ANEPPDHQ voltage-sensitive dye-stained thalamocortical slice.
Depth-scanning image stacks acquired at fine depth resolution were rendered into 3D
154
volume images using Voxx volume rendering software [37]. Volumetric reconstruction
allows off-line investigation by computationally “passing through” the reconstructed volume. Figure 7.4 shows a volume rendered from three neighboring sub-regions of a larger
sectioned background image. The side profile clearly displays cellular profiles which
change in depth, illustrating the optical sectioning capability of the multifocal device.
edge of slice
dark holes =
cell bodies
100 µm
cell bodies beneath
tissue surface
visualized via optical
sectioning
Figure 7.4: 3D rendered volume of a depth-scanned two-photon image stack (∼ 6 µm
steps) in di-4-ANEPPDHQ voltage-sensitive dye-stained thalamocortical slice region.
Volume region corresponds to the red shaded region in the sample background image
(top).
155
7.2.2 Anatomical two-photon imaging in vivo
Two-photon microscopy in vivo using fluorescent dyes enables anatomical investigations
of tissues in their natural context. For brain tissue, this ensures that the cellular mechanical integrity is maintained, whereas significant warping of tissue geometry often occurs
in slice preparations. Additionally, in vitro preparations lack the electrophysiological connectivity found in vivo. Thus experimental results yielded by in vitro preparations are of
limited conclusiveness regarding natural in vivo physiological function. In vivo imaging,
however, poses unique optical challenges for two-photon microscopy and in particular
for scanning techniques that cannot utilize scattered emission signals (see discussion in
Section 2.5.6). Imaging deep into the layers of the cortex, for instance, is particularly
challenging because of the ∼ 100 µm of cell-sparse scattering layer 1 that must be penetrated by both excitation and emission light before cell bodies, mostly in layer 2/3, are
visible. This situation is readily visualized in the thalamocortical slice image in Fig. 7.2,
which shows that vast majority of cell bodies are separated from the cortex surface by a
thick layer of highly fluorescent featureless tissue.
7.2.2.1 Surgical preparation and in vivo imaging protocol
Mice were anesthetized with a combination of ketamine and xylazine (100 and 20 mg/kg,
respectively, intraperitoneally), a 5-mm-diameter craniotomy was performed, the dura
was removed, and the surface of the brain was exposed. The exposed surface of the
156
barrel cortex was stained with voltage-sensitive dye (RH795, di-3-ANEPPDHQ, or di-4ANEPPDHQ) for ∼ 45 min. The cortex was then washed with saline solution to remove
unbound dye. To maintain a meniscus for the water immersion objective, a reservoir well
of inside diameter ∼ 8 mm was then constructed with dental acrylic surrounding the craniotomy. The top of the cortex was immersed in a thin layer of 2% agarose (∼ 100 µL
in liquid form), and a ∼ 4 mm-diameter fragment of a No. 0 glass coverslip was placed
directly on top of the craniotomy region. These steps significantly suppressed physiological motion of the brain (e.g. heartbeat, respiration). For functional imaging, tungsten
microelectrodes were then placed in the cortex and the thalamus for recording and electrical stimulation, and the multifocal two-photon microscope objective was then centered
over the craniotomy and lowered with coarse vertical adjustments until the objective drew
a meniscus with a small quantity of saline in the reservoir. The mouse stereotaxic assembly was then raised with the vertical axis of a 3-axis micrometer while watching the
real-time full-frame EMCCD image acquisition until two-photon fluorescence foci were
observed. A photograph of the experimental imaging configuration is shown in Figure
7.5. In a darkened room, the voltage-sensitive dye two-photon fluorescence signal on the
craniotomy was usually bright enough to see with the un-aided eye.
7.2.2.2
In vivo Anatomical imaging via voltage-sensitive dye staining
The most obvious feature of two-photon voltage-sensitive dye fluorescence background
images of the first ∼ 100 µm of cortex tissue in vivo is the cortical vascular microcircuitry
(Figure 7.6. Locations where blood vessels move vertically through the tissue appear as
157
Figure 7.5: Photograph of in vivo two-photon imaging configuration. Details of the meniscus reservoir and electrode configuration are visible.
“dark spots” in the field of view.
7.2.2.3
In vivo enhanced Green Fluorescent Protein (EGFP) imaging
In another set of two-photon anatomical imaging experiments, specific sub-populations
of interneurons were imaged in vivo in the cortex of transgenic mice expressing enhanced
green fluorescent protein (EGFP). EGFP [44] is a mutant form of green fluorescent protein
(GFP) [134] in which a set of amino acid substitutions lead to enhanced variants that
exhibit higher fluorescence yields. The novel line of transgenic mice [143] mostly express
EGFP in hippocampal interneuronal sub-populations, but also express EGFP in cortical
interneurons, albeit comparatively sparsely.
Volumetric reconstructions of two-photon depth-scans (Figure 7.7) reveal a number
of small, oblong bright cellular features in the first 150 µm of cortex tissue. Figure 7.7B
158
VSD-stained
cortex surface
blood
vessels
100 µm
rising / plunging
blood vessel
Figure 7.6: Two-photon background fluorescence image of di-4-ANEPPDHQ-stained somatosensory cortex in vivo at focal depth ∼ 100 µm below the cortex surface. Cortical
vasculature is clearly visible. Dark “spots” are locations where blood vessels move vertically through the cortex.
159
depicts several stages of “passing through” a reconstructed imaged volume, whereby one
of the lateral axes of the rendered volume data is penetrated. Vertical white arrows in
Fig. 7.7 highlight locations of EGFP-labeled interneurons which wind vertically through
layer 1 of the cortex. Similar process morphology is visible in 1-photon confocal images
of fixed thalamocortical slice slides (Fig. 7.7C). Two-photon images with higher contrast
(Fig. 7.7D) were achieved via 1-focus excitation (by removing the microlens array and
collimation optics).
160
A
B
surface
C
D
Figure 7.7: A: Surface background two-photon image of cortex in transgenic mouse (in
vivo) expressing enhanced GFP (EGFP) in sub-populations of interneurons. Image shown
is a set of sub-regions taken from a larger (400 × 400 µm) background image. B: Series
of side-on views sectioning through a volumetric reconstruction of two-photon depthscan image stack. Vertical white arrows highlight locations of putative EGFP-expressing
interneurons. C: Confocal microscopy (one-photon) image of fixed slice from the same
series of transgenic EGFP-expressing mice. Note similarities in “meandering” process
morphology between cells visualized in cross-sections of (B) and confocal image from
(C) . Image courtesy E. G. Hughes and C.G. Welle. D: Higher magnification two-photon
images of fixed EGFP-expressing cortical mouse slice. Depths represent depth of imaging
below slice surface (rather than dura surface).
161
Chapter 8
Two-photon fluorescence recording of
action potentials from individual
mammalian nerve terminals in situ
We report the optical recording of action potentials from individual (∼1 µm) nerve terminals of the intact mouse neurohypophysis, in a single sweep. The measurements utilized two-photon excitation along the “blue” edge of the two-photon absorption spectrum
of a fluorescent voltage-sensitive naphthylstyryl pyridinium dye. Single-trial recordings
of action potentials exhibited signal-to-noise ratios (S:N) ∼5 and fractional fluorescence
changes of up to ∼10%. These results represent the first single-trial optical recording
of action potentials from individual nerve terminals in an intact mammalian preparation
using 180 ◦ detection. As such, this technique offers clear advantages over other optical
approaches (e.g., voltage sensitive second harmonic generation (SHG)) currently used to
162
monitor voltage changes in localized neuronal regions, and may also serve as an alternative
to invasive electrode arrays for studying neuronal systems in vivo.
8.1 Introduction
The basic unit of information flow in the nervous system is the fast, all-or-nothing millisecondscale propagating wave of membrane depolarization termed the action potential (AP). At
this fundamental digital level of neuronal communication, signal averaging can rarely be
used to study the dynamics of neuronal communication and AP timing synchronization,
especially in measurements of spontaneous activity. In these cases, single-trial detection
of fast, discrete electrical events at single-cell resolution is crucial. Such measurements
have long been dominated by electrode technology, which is invasive, and limited in spatial resolution. The most advanced electrode arrays (e.g. [146]) can resolve membrane
potential dynamics at multiple sites with micron-scale spatial resolution, but they require
careful interfacing with neuronal preparations and are not capable of recording at different
depths within thick samples.
Imaging techniques employing potentiometric dyes as molecular indicators of membrane voltage [38], [167] permit non-invasive measurements of electrical activity from
a wide range of excitable structures ranging from individual dendritic spines [138] to
intact, complex neuronal circuits, in vitro as well as in vivo, with extremely high temporal resolution. Previous voltage-sensitive dye imaging of neuronal networks in vitro
163
(e.g. [43], [10], [139], [140]) and in vivo (e.g. [149], [35], [67]) have employed conventional one-photon (linear) excitation. While one-photon fluorescence microscopy can be
used to measure neuronal activity with high S:N, these measurements represent an integration over the depth of the focal plane as well as the activity encoded by out-of-focus
fluorescence signals. This spatial summation of signals complicates the interpretation of
fluorescence patterns from thick tissue samples, and precludes the detection of potentially
localized events such as activity in individual nerve terminals and dendrites.
Nonlinear optical microscopy techniques such as two-photon microscopy [49] and
second harmonic generation microscopy [88], [173], [30], [138] offer intrinsic optical
sectioning at micron-scale resolution, thereby virtually eliminating spatial signal summation. However, single-trial measurements of fast electrical events in intact preparations (as
opposed to neurons in culture) with appreciable S:N have not yet been achieved using nonlinear optical techniques [178]. The small size of the fluorescence changes (∆F/F ∼ 1% /
100 mV) in response to membrane depolarization, the limited two-photon absorption and
scattering cross-sections of potentiometric probes, and the constraints on incident light
intensity imposed by the need to avoid phototoxicity, all conspire against this goal. In addition, with the exception of a few studies (e.g. [66], [132]), the wavelength-dependence
of potentiometric, nonlinear optical signals has not been explored.
The neurohypophysis (or posterior pituitary) is a classic in vitro preparation for studying evoked release of peptide hormones [51]. Magnocellular neurons, with their cell bodies in the hypothalamus, project their axons as a bundle of fibers through the infundibular stalk, to terminate in the neurohypophysis where they ramify extensively. In the rat,
164
∼20,000 neurons give rise to ∼40 million terminals and secretory swellings [137]. When
this in vitro preparation is stained with voltage sensitive dyes, and stimulated electrically,
it generates large optical signals. These compound responses, changes in one-photon absorption [73] or fluorescence [135], represent the synchronous firing of action potentials
from a large population of terminals.
Here we present single-trial two-photon imaging of action potentials from individual
(∼1 µm) nerve terminals in an intact mouse neurohypophysis stained with the naphthylstyryl pyridinium potentiometric probe di-3-ANEPPDHQ [140]. To optimize the recording conditions, we investigated the wavelength dependence of the signal upon excitation
between 800 nm and 900 nm, and upon emission between 490 and 700 nm. The best twophoton single-trial measurements exhibited high S:N (up to 5:1) and large fractional fluorescence changes (up to ∼10%). Beyond representing the first single-trial optical recordings of electrical activity at sub-micron resolution from an intact mammalian preparation
using 180 ◦ detection, these results lay the methodological foundation for in depth optical
recording of electrical activity in neuronal circuits in vivo.
8.2 Methods and Materials
8.2.1 Two-Photon Imaging Optics
Two-photon imaging was performed with a custom-built microscope (Fig. 8.1). The laser
excitation source was a mode-locked Titanium:Sapphire laser (Chameleon, Coherent Inc.,
165
Santa Clara, CA, USA). To minimize distortion of the beam profile, a λ/2 plate and GlanThompson polarizer were used to adjust the average excitation power at the sample to be
∼10-20 mW. A small fraction of the beam was picked off with a microscope coverslide and
focused onto a BBO (β-BaB2 O4 ) second harmonic generation (SHG) crystal in order to
monitor fluctuations in the square of the laser power, hP (t)i2 . This SHG signal was filtered
with a BG39 Schott glass absorption filter and detected by a photomultiplier tube (R955,
Hamamatsu Photonics, Hamamatsu City, Japan) operating in analog mode. Galvonometer
scanning mirrors (Cambridge Technology Inc., Lexington, MA, USA) were used to scan
the excitation beam in the sample plane, and a piezoelectric-driven objective translator
(P721, Physik Instrumente GmbH & Co. KG, Karlsruhe/Palmbach, Germany) scanned
the beam along the vertical (illumination) axis. The combination of scan lens and tube lens
served to expand the excitation laser beam and back-fill a microscope objective (20×/0.95
NA XLUMPlanFl, Olympus, Melville, NY). Two-photon absorption-induced fluorescence
was collected in the backward direction (180 ◦ ) by the objective, and redirected to a second
photomultiplier tube (R943, Hamamatsu Photonics, Hamamatsu City, Japan) by a dichroic
mirror (R < 700, Chroma Technology Corp., Rockingham, VT, USA) followed by an
emission filter (typically interference-coated glass absorption filter). Analog signals from
both PMTs were amplified prior to digitization by a multifunction data acquisition board
(PCI-6052E, National Instruments, Austin, TX, USA). The PMT-detected fluorescence
voltage signal was divided by the SHG signal to correct for laser power fluctuation. Device
control and data acquisition were performed by custom-written software using LabVIEW
visual programming environment (National Instruments, Austin, TX, USA).
166
BG 39 filter
SHG crystal (BBO)
steering
mirror
Modelocked Ti:Al2O3
laser
iris
emission filter
tube lens
folding
mirror
dichroic mirror
Objective
(20X, NA = 0.98,
water immersion)
scan lens
X-Y scanning galvo
mirrors
Objective
translator
+
- stimulation leads
sample chamber
Figure 8.1: Optical setup for two-photon imaging and functional recording.
167
8.2.2 Two-Photon Functional Recordings
The neurointermediate lobe of the pituitary, that comprises pars nervosa (neurohypophysis) and pars intermedia, was obtained [N = 6], as previously described [168], from
female CD-1 mice that had been anesthetized by CO2 inhalation and decapitated in accordance with institutional guidelines. Briefly, the mouse head was pinned to the bottom
of a Sylgard-lined dissection dish, and, following removal of the skin, the skull was opened
along the dorsal midline and the bone removed bilaterally. Cutting the optic and olfactory
nerves allowed caudal reflexion of the brain and rupture of the infundibular stalk, leaving the infundibular stump together with the entire pituitary gland attached to the base
of the skull. The gland, gently removed using fine forceps and iridectomy scissors, was
transferred to a dish containing oxygenated mouse Ringer’s solution (in mM: 154 NaCl,
5.6 KCl, 1.0 MgCl2 , 2.2 CaCl2 , 10 glucose, 20 HEPES, adjusted to pH 7.4 with NaOH)
where the anterior pituitary (pars anterior) was separated from the neurointermediate lobe
and discarded. The pars intermedia, which consists of a delicate lacework of cells supporting the neurohypophysis, provided a convenient border through which the specimen
could be pinned to the Sylgard bottom of a small Petri dish (see Fig. 8.1) while preserving
the integrity of the neurohypophysis. Stimulation was achieved using a pair of Tefloncoated Pt-Ir (90%-10%) electrodes clasping the infundibular stump, and consisted of brief
(100-500 ms) shocks delivered through a stimulus isolator.
Figure 8.2A shows a dissection microscope image of an isolated neurohypophysis attached to stimulation leads in a recording chamber. Mouse Ringer’s solution was changed
168
C
A
stimulating wire
200 µm
pars nervosa
pars intermedia
B
RBC
3.5 µm
Figure 8.2: Anatomy of the neurohypophysis. A: dissection microscope image of isolated
neurohypophysis with surrounding pars intermedia, and infundibular stalk connecting to
stimulation leads (scale bar = 200 µm). B: Two-photon background image of neurohypophysis stained with di-3-ANEPPDHQ voltage-sensitive dye. Excitation wavelength
= 850 nm. C: Wide-field scanning electron microscope image of mouse neurohypophysis. Individual axon terminals contain peptide hormone-carrying vesicles. Red blood
cell (RBC) provides scale (length ∼7.5 µm).
169
every 20-30 minutes to maintain tissue viability. Prior to functional imaging, two-photon
anatomical survey images were scanned. Functional imaging consisted of short line-scans
(typically ∼3 points over a line ∼0.5 µm long, as in Fig. 8.3A) integrating for ∼100 µs at
each point. Each point in the final data traces (e.g. Fig. 8.3C) represents the average of the
∼3 points long the short line. To minimize photodamage, line scanning was performed
rather than static single-point “beam-parking” functional recording. Electric stimuli (∼8
mA) were delivered beginning ∼50 ms after the beginning of the line-scan in order to minimize the effect of photobleaching. When multiple stimuli were delivered (maximum 16
stimuli), the stimulation frequency was 16 Hz. Otherwise, all trials were single-stimulus
experiments. Single trial recording durations were between 100 and 500 ms, and wait time
between trials was 5 seconds for single trials, and 1 minute for multiple stimulus trains.
Averaging, when needed, was performed offline and only utilized responses within a single multi-stimulus trial. The number of trials averaged (if any) is indicated in individual
figure legends. For each point scanned in functional recordings, fractional fluorescence
(∆F/F ) was calculated by subtracting the baseline, F , from each point in the recording,
and then dividing by this same background. F , in our experiments, was the average of the
first 10 data points in the recording. Functional recording data was not processed in any
way; all traces represent raw data.
170
8.2.3 Action Spectra Measurements
One-photon action spectrum measurements were generated from the results of 10 fluorescence ∆F/F responses at 19 wavelengths. Excitation wavelength dependence was measured by varying excitation filters (bandpass) and comparing peak fluorescence ∆F/F
response. Because excitation and emission wavelengths are more narrowly shifted in onephoton measurements compared with two-photon measurements, several dichroic mirror
and filter combinations were used to span the excitation wavelength range of 360 nm 580 nm. The filter combinations were as follows: for 360 nm to 440 nm excitation, a R <
450 nm dichroic and Schott OG475 glass emission filter were used; for 450 nm to 578.5
nm excitation, a R < 585 nm dichroic and Schott OG570 glass emission filter were used;
for 590 nm to 630 nm excitation, a 695 nm dichroic and Schott RG715 glass emission
filter were used. Fractional fluorescence (∆F/F ) was measured as a function of excitation wavelength rather than signal-to-noise because the system excitation and collection
efficiency varied with different filter sets. The two-photon action spectrum was generated
from the results of all two-photon functional recordings. The excitation laser wavelength
was tuned between 800 nm and 900 nm to investigate the positive peak of the action spectrum for di-3-ANEPPDHQ; the corresponding two-photon excitation wavelengths for the
negative ∆F/F peak (see Fig. 8.4A) were beyond the tuning range of our mode-locked
laser.
171
8.2.4 One-Photon Spectroscopy
Absorption and emission spectra for the VSD di-3-ANEPPDHQ were measured using
octanol as a solvent to approximate the environment of the dye when bound to membrane
[177]. Final sample concentration was less than 100 µM. Absorption measurements were
performed on a spectrophotometer (Ocean Optics, Dunedin, FL, USA) and excitation and
emission spectrum measurements were performed on a spectrofluorometer (DeltaRAM;
Photon Technology International).
8.3 Results
Figure 8.3 shows representative functional imaging results of both single-trial and trialaveraged data. Fig. 8.3C shows single-trial data from a typical experiment in which
multiple stimuli were delivered, and Fig. 8.3D shows a representative recording from a
single-stimulus trial. Characteristic action potential shape, including afterhyperpolarization, is clearly visible in the averaged response shown in Fig. 8.3B, which shows the averaged response of 16 stimuli delivered in a single recording trial. Anatomical two-photon
imaging of the neurohypophysis is compared with wide-field electron microscopy images
in Fig. 8.2. Features resembling the individual axon terminals visible in Fig. 8.2C are
evident in the two-photon image (Fig. 2B). Fig. 8.3A shows the two-photon background
image and recording location used for Fig. 8.3B & C.
Action spectra for one- and two-photon excitation are shown in Figs. 8.4 and 8.5,
respectively. Fig. 8.5A shows statistics of single-trial S:N as a function of two-photon
172
A
B
0.03
0.02
0.01
0
-0.01
0 10 20 30 40 50 60 70
C
D
time (ms)
0.1
0.04
0.05
0.02
0
0
-0.02
250 300 350 400 450
-0.05
20
40
60
80 100
time (ms)
time (ms)
Figure 8.3: Two-photon recording of action potentials in individual nerve terminals. A:
Background two-photon image of region of neurohypophysis used for functional optical
probing (scale = 5 µm). Inset: Magnified view (scale bar = 1 µm) of imaged region; short
red line indicates actual location of line-scan excitation. B: Average of 16 putative action
potentials from a single trial (16 stims delivered at 16 Hz). C: Segment of single-trial data
from the same trial used for average in B. D: Single trial response to single stimulation
with both high S:N and large ∆F/F signal (nearly 10%). For all data in figure, excitation
wavelength = 850 nm and emission wavelength = 530/90 nm.
173
excitation wavelength, averaging over all experiments. Here, S:N is defined as the signal amplitude (peak-to-peak) divided by the RMS noise, which is calculated from data
recorded before the stimuli were delivered (i.e. before 50 ms). Both standard deviation
(red) and standard error of mean (blue) error bars are shown. While there was significant
statistical variability in single-trial two-photon measurements, there is a clear local maximum at 850 nm. Fig. 8.5B shows the corresponding statistics (same database as Fig.
8.5A) for fractional fluorescence (∆F/F ) as a function of two-photon excitation wavelength. Although there is a similar peak in the ∆F/F response at 850 nm to that in the
S:N figure, it is not as statistically significant and the main statistical trend is a downward
slope toward 900 nm. Assuming constant laser intensity over the excitation spectrum, Fig.
8.5A is consistent with Figs. 8.5A and C, as the signal-to-noise is proportional to ∆F/F
and the TPE cross-section, σ2 , i.e. S/N ∝
∆F √
σ2 .
F
A
B
2-Photon Excitation Wavelength (nm)
780 800 820 840 860 880 900 920 940
0.05
0.04
0.07
0.03
0.06
0.02
0.05
0.01
0.04
0
-0.01
0.03
-0.02
0.02
-0.03
0.01
-0.04
0
-0.05
350
2-Photon
1-Photon
400
450
500
550
600
Excitation Wavelength (nm)
390 400 410 420 430 440 450 460 470
1-Photon Excitation Wavelength (nm)
Figure 8.4: Comparison of one- and two-photon action spectra for potentiometric dye di3-ANEPPDHQ. A: One-photon action spectrum of di-3-ANEPPDHQ. B: Comparison of
one- and two-photon action spectrum. Note that two-photon action spectral data is shown
on a different scale (top axis) for wavelength comparison purposes.
The one-photon fluorescence emission spectrum for di-3-ANEPPDHQ measured in
174
octanol had a peak at ∼613 nm for excitation at 425 nm (Fig. 8.5D). The optimal twophoton fluorescence emission detection wavelength for the excitation wavelength range of
800 nm - 900 nm was experimentally determined to be ∼530 nm. While the fluorescence
emission spectrum extends much further into the red (and is excitation wavelength dependent - see Fig. 8.6), band-pass filters centered at longer wavelengths as well as long-pass
filters yielded measurements with significantly smaller S:N.
The corresponding one-photon action spectrum (Fig. 8.4A) displays a bipolar shape
with both positive and negative peaks. The sign change occurs because the absorption
spectrum of di-3-ANEPPDHQ is blueshifted in response to depolarization, thus fluorescence responses recorded during excitation near the absorption peak around 500 nm (see
Fig. 8.5D) are either positive or negative depending on the side of the peak chosen as the
excitation wavelength. One- and two-photon action spectra are compared in Fig. 8.4B.
The one-photon action spectrum in the corresponding wavelength range of two-photon
excitation (i.e. two-photon excitation wavelength divided by 2) indicates that the sign of
the two-photon signal is appropriately positive. The data additionally indicate that twophoton signals with larger ∆F/F can be expected at excitation wavelengths beyond 1000
nm.
8.3.1 Discussion
The neurohypophysis consists nearly entirely of the nerve terminals and axonal varicosities of magnocellular neurons (see Fig. 8.2). The remainder of the tissue consists of
175
A
B
S/N
6
0.08
0.07
s.d.
s.e.m.
5
4
0.05
3
0.04
2
0.03
1
0.02
emission wavelength = 530±45 nm
0
C
800
emission wavelength = 530±45 nm
820
840
860
880
Excitation Wavelength (nm)
0.01
900
TPE Cross-Section (σ 2)
800
820
840
860
880
Excitation Wavelength (nm)
900
D
absorption
fluorescence
20
10-50cm4s / photon
s.d.
s.e.m.
0.06
16
12
8
4
0
800
840
880
Excitation Wavelength (nm)
300
920
400
500
600
700
800
Wavelength (nm)
Figure 8.5: Signal-to-noise and ∆F/F action spectrum for di-3-ANEPPDHQ. A. Signalto-noise in two-photon excited fluorescence signal as a function of excitation wavelength
in single-trial optical recordings of action potentials. Red error bars represent standard
deviation of measurements, blue bars represent standard error of mean. B: ∆F/F of twophoton excited fluorescence signal as a function of excitation wavelength in single-trial
optical recordings of action potentials (same trials as S:N plot used for analysis). Red error bars represent standard deviation of measurements, blue bars represent standard error
of mean. C: Measured two-photon excitation cross-sections as a function of excitation
wavelength for voltage-sensitive dye di-8-ANEPPDHQ (data from [66]; see Chapter 6).
D: One-photon absorption and fluorescence emission spectra for di-3-ANEPPDHQ in octanol. Emission spectra was measured with excitation at 425 nm.
176
electrically inactive pituicytes (glial cells), and capillaries, into the vicinity of which the
nerve terminals secrete the peptide hormones oxytocin and vasopressin. Because the axonal varicosities and nerve terminals represent the only electrically excitable tissue in the
organ, there is no post-synaptic electrical activity to confound the interpretation of the
optical responses, which must represent action potentials. Single-trial responses were 2-3
ms wide at half maximum, and averaged responses were often 4-5 ms wide. This broadening is likely due to physiological jitter in the timing of the response, since the recordings
represent the activity of just a single nerve terminal. This statistical variability in response
is reflected in the error bars in the two-photon action spectra (Fig. 8.5). Error bars on
the one-photon action spectra are significantly smaller, reflecting the fact that those optical responses represent the summed synchronous firing of hundreds of thousands, if not
millions of terminals.
B
A
absorption
fluorescence
excitation
300
400
500
600
700
Excitation Wavelength (nm)
530 nm Excitation
425 nm Excitation
400
800
450
500
550
600
Wavelength (nm)
650
700
Figure 8.6: One-photon excitation and emission spectra data for potentiometric dye di-3ANEPPDHQ. A: Comparison of absorption (black) and excitation spectrum (gray), which
measured fluorescence emission at 610 nm as a function of excitation wavelength. B:
Fluorescence emission in response to 425 nm excitation (gray) and 530 nm excitation
(black).
177
The results summarized in Fig. 8.3 clearly demonstrate the ability to record action potentials in single trials from sub-micron patches of membrane. Interestingly, two-photon
excitation yielded fractional fluorescence changes approximately five times larger than
one-photon excited ∆F/F at corresponding (
λ2photon
)
2
excitation wavelengths (see Fig.
8.4A). Such increases in ∆F/F following two-photon excitation are consistent with reported measurements in cultured neurons stained with a hemicyanine VSD [110]. The
magnitude of VSD signals varies widely between phyla and even species [162]; the finding of large ∆F/F signals with high signal-to-noise in an intact mammalian preparation
described in this report are therefore particularly relevant to the fields of systems neuroscience and neurology, where studying the mammalian brain is of primary interest. Additionally, the VSD employed, di-3-ANEPPDHQ, has been shown to exhibit comparatively
moderate levels of phototoxicity, and of dye internalization, which reduces S:N [140].
There has been substantial interest in functional two-photon imaging of VSDs at wavelengths beyond 960 nm, primarily because long excitation wavelengths permit greater
penetration into scattering tissue samples and generate less phototoxicity. Further, VSD
peak two-photon absorption cross-sections apparently lie beyond 960 nm [66]. While the
one-photon action spectrum clearly suggests that larger signals can be expected for twophoton excitation between 1000 nm and 1150 nm, increased water absorption beyond 900
nm may mitigate some of the advantages of much longer wavelengths. Measurement of
two-photon absorption cross-sections at these longer wavelengths would resolve this issue and permit quantitative prediction of the potential advantages of infrared two-photon
excitation of VSDs.
178
8.3.2 Conclusion
The ability to record action potentials from nerve terminals and, by implication, other submicron features of neurons has been realized in an intact mammalian preparation. Action
spectra in the physiological “window” of low water absorption have been characterized
for one- and two-photon excitation. We expect the technique to yield comparable imaging
results in other in vitro and in vivo mammalian preparations.
179
Chapter 9
Summary
In summary, this thesis has described neuroimaging experiments utilizing novel one- and
two-photon imaging devices. As part of the detailed description of the novel devices, I
have reviewed the background physical principles and the relevant scientific motivation
leading to the utility of these imaging modalities.
The one-photon fluorescence imaging experiments have addressed questions regarding
the processing of both stimulated and spontaneous electrical neuronal activity, in the layers
of the neocortex and in the ventrolateral medulla, respectively. One-photon investigation
using a GRIN lens endoscopic imaging device produced the first 3D high-speed imaging
of membrane potential dynamics in vivo, and yielded depth patterns of neuronal activity
consistent with cellular architecture of the first few cortical layers. Future experimental
directions for GRIN lens imaging in vivo include (1) imaging neuronal response to more
sophisticated sensory stimuli (e.g. visual or whisker stimuli), (2) further miniaturization
of device optics to enable head-mounting of the device in awake, behaving animals (this
180
would likely necessitate fiber-coupling the fluorescence emission to a separate stationary
detector), and (3) implementing nonlinear (i.e. two-photon) scanning capabilities in the
endoscopic device to achieve true optical sectioning of fluorescence signals.
One-photon imaging of spontaneous respiratory-related activity in the pre-Bötzinger
region of the ventrolateral medulla in vitro indicated that the putative respiratory rhythm
generation region consisted of a distributed network, rather than a single pacemaking
point. This study represents the first imaging of respiratory-related membrane-potential
dynamics patterns in a near fully-intact in situ preparation, and is the first step in dissecting
this system as it functions in its intact form. New research directions following this work
should involve (1) using faster and more sensitive detectors for similar triggered-averaging
measurements of spontaneous respiratory-related activity for better temporal resolution
and higher signal-to-noise, (2) higher spatial resolution imaging, (3) functional imaging
using calcium indicator dyes to improve localization of activity to individual cells, and (4)
using functional (i.e. voltage or calcium sensitive) dyes in reduced preparations of transgenic rats which express GFP variants in individual respiratory neuron sub-populations.
An in-depth spectroscopic study of the two-photon absorption properties of fluorescent probes for bioimaging was performed. This study identified the best current candidate voltage-sensitive dyes for two-photon imaging (di-8-ANEPPDHQ, di-8-ANEPPS,
and RH795). Further investigation of this series of compounds (as well as new VSDs, as
they become available) should identify the peak TPA cross-sections using pulsed lasers
with extended tuning ranges. A similar spectroscopic study was attempted for a novel
181
series of conjugated (porphinato)zinc(II) compounds with electrical properties suggesting promise for use as voltage-sensitive dyes. Results indicated that these compounds
generally exhibited a combination of linear and non-linear absorption, thus making them
non-ideal for two-photon imaging. Based on this study, porphyrin monomer and dimer
compounds seem to be the most likely candidates for fluorescence two-photon imaging
probes, as the larger conjugated compounds exhibit too much linear absorption throughout the tuning range of most ultrafast pulsed lasers. The potentiometric properties of these
porphyrin-based compounds, however, have yet to be investigated. Such investigation will
require (1) modification of the compounds such that they embed themselves in neuronal
membranes, and (2) construction of an experimental apparatus for testing voltage sensitivity (e.g. artificial planar bilayers or vesicles, cultured neurons, etc.).
The application of multifocal two-photon microscopy to several imaging experiments
in vivo and in vitro was described. Anatomical imaging using voltage-sensitive dye background fluorescence as well as EGFP expression revealed the optical sectioning and resolution properties of the novel microscope system. Functional imaging with the current
multifocal device configuration has been difficult, however, due to fluorescence collection
limitations. Multifocal detection requires rejection of scattered light at each emission focus, making imaging through scattering tissue difficult. Additionally, de-scanning the fluorescence emission signal, required in order to maintain a stationary multifocal detection
scheme, imposes excessive collection efficiency losses on the system. This necessitates
extremely high excitation power levels, which leads to rapid dye bleaching and photodamage. Functional neuroimaging via multifocal two-photon microscopy using fluorescent
182
dyes would likely yield improved results by (1) implementing more efficient techniques
for achieving multifocal excitation (e.g. beamsplitter array based beam multiplexer, as
in [136], [45]) in order to avoid off-axis aberrations and distortion of laser pulse-width,
(2) modification of fluorescence collection to optimize use of scattered fluorescence emission and to avoid cross-talk between foci, and (3) the use of lasers with higher average
power at longer excitation wavelengths.
Finally, using single-beam scanning techniques, we have described single-trial twophoton excited fluorescence recordings of action potentials from individual (∼1 µm) nerve
terminals of the intact mouse neurohypophysis. Utilizing two-photon excitation along
the “blue” edge of the two-photon absorption spectrum of a fluorescent voltage-sensitive
naphthylstyryl pyridinium dye, single-trial recordings of action potentials exhibited signalto-noise ratios (S:N) ∼5 and fractional fluorescence changes of up to ∼10%. These results
represent the first single-trial optical recording of action potentials from individual nerve
terminals in an intact mammalian preparation using 180 ◦ detection. As such, this technique offers clear advantages over other optical approaches (e.g., voltage sensitive second
harmonic generation (SHG)) currently used to monitor voltage changes in localized neuronal regions, and may also serve as an alternative to invasive electrode arrays for studying
neuronal systems in vivo.
183
Bibliography
[1] S. Adams. Imaging neurons. A laboratory manual chapter How calcium indicators
work, pages 30.1–30.7. Cold Spring Harbor Press (2000).
[2] D. Agard. Optical Sectioning Microscopy: Cellular Architecture in Three Dimensions. Annual Review of Biophysics and Bioengineering 13(1), 191–219 (1984).
[3] D. Agard and J. Sedat. Three-dimensional architecture of a polytene nucleus. Nature 302(5910), 676–681 (1983).
[4] M. Albota, C. Xu, and W. Webb. Two-photon fluorescence excitation cross sections
of biomolecular probes from 690 to 960 nm. Applied Optics 37(31), 7352–7356
(1998).
[5] H. Anderson. Building molecular wires from the colours of life: conjugated porphyrin oligomers. Chemical Communications 2323, 2330 (1999).
[6] H. Anderson, S. Martin, and D. Bradley. Synthesis and Third-Order Nonlinear
Optical Properties of a Conjugated Porphyrin Polymer. Angewandte Chemie International Edition in English 33(6), 655–657 (1994).
184
[7] Andor Technology PLC Belfast, Ireland. Andor Technology Digital Camera Fundamentals.
[8] V. Andresen, A. Egner, and S. Hell. Time-multiplexed multifocal multiphoton microscope. Optics Letters 26, 75–77 (2001).
[9] R. D. Andrew and B. A. Mac Vicar. Imaging cell volume changes and neuronal
excitation in the hippocampal slice. Neuroscience 62(2), 371–383 (1994).
[10] C. Ang, G. Carlson, and D. Coulter. Hippocampal CA1 Circuitry Dynamically
Gates Direct Cortical Inputs Preferentially at Theta Frequencies. Journal of Neuroscience 25(42), 9567–9580 (2005).
[11] W. Angerer. SHG spectroscopy of gallium nitride thin films on sapphire with ultrashort pulses. PhD thesis University of Pennsylvania (1998).
[12] W. Angerer, N. Yang, A. Yodh, M. Khan, and C. Sun. Ultrafast second-harmonic
generation spectroscopy of GaN thin films on sapphire.
Physical Review B
59(4), 2932–2946 (1999).
[13] A. Arieli, D. Shoham, R. Hildesheim, and A. Grinvald.
Coherent spatiotem-
poral patterns of ongoing activity revealed by real-time optical imaging coupled
with single-unit recording in the cat visual cortex. Journal of Neurophysiology
73(5), 2072–2093 (1995).
185
[14] P. Bandettini, A. Jesmanowicz, E. Wong, and J. Hyde. Processing strategies for
time-course data sets in functional MRI of the human brain. Magnetic Resonance
in Medicine 30(2), 161–73 (1993).
[15] R. Bedlack, M. Wei, and L. Loew. Localized membrane depolarizations and localized calcium influx during electric field-guided neurite growth. Neuron 9(3), 393–
403 (1992).
[16] T. Beeler, R. Farmen, and A. Martonosi. The mechanism of voltage-sensitive dye
responses on sarcoplasmic reticulum. Journal of Membrane Biology 62(1), 113–
137 (1981).
[17] D. Beljonne, G. E. O’Keefe, P. J. Hamer, R. Friend, H. L. Anderson, and J. L.
Brédas. Investigation of the linear and nonlinear optical response of edge-linked
conjugated zinc porphyrin oligomers by optical spectroscopy and configuration interaction techniques. Journal of Chemical Physics 106(23), 9439–9460 (1997).
[18] M. Bellingham. Driving respiration: the respiratory central pattern generator. Clinical and Experimental Pharmacology and Physiology 25(10), 847–56 (1998).
[19] F. Bevilacqua, D. Piguet, J. D. Gross, B. J. Tromberg, and C. Depeursinge. In
vivo local determination of tissue optical properties: applications to human brain.
Applied Optics 38 (1999).
[20] J. Bewersdorf, R. Pick, and S. Hell. Multifocal multiphoton microscopy. Optics
Letters 23(9), 655–657 (1998).
186
[21] A. Bianchi, M. Denavit-Saubie, and J. Champagnat. Central control of breathing in
mammals: neuronal circuitry, membrane properties, and neurotransmitters. Physiological Reviews 75(1), 1–45 (1995).
[22] R. Birge and D. Kliger.
[23] R. Bonnett. Photosensitizers of the porphyrin and phthalocyanine series for photodynamic therapy. Chemical Society Reviews 24, 19–33 (1995).
[24] J. Briers and A. Fercher. Retinal blood-flow visualization by means of laser speckle
photography. Investigative Ophthalmology & Visual Science 22(2), 255–259
(1982).
[25] A. Bullen and P. Saggau. High-Speed, Random-Access Fluorescence Microscopy:
II. Fast Quantitative Measurements With Voltage-Sensitive Dyes. Biophysical Journal 76(4), 2272–2287 (1999).
[26] J. Burris and T. McIlrath. Theoretical study relating the two-photon absorption
cross section to the susceptibility controlling four-wave mixing. Journal of the
Optical Society of America B 2, 1313–1317 (1985).
[27] D. Büsselberg, A. Bischoff, K. Becker, C. Becker, and D. Richter. The respiratory
rhythm in mutant oscillator mice. Neuroscience letters 316(2), 99–102 (2001).
[28] D. Büsselberg, A. Bischoff, J. Paton, and D. Richter. Reorganisation of respiratory network activity after loss of glycinergic inhibition. Pflügers Archiv European
Journal of Physiology 441(4), 444–449 (2001).
187
[29] Cairn Research Ltd. Online Documentation. About arc lamps. http://www.
cairnweb.com/tech/tech_lamp2.html (2003).
[30] P. Campagnola, M. Wei, A. Lewis, and L. Loew. High resolution non-linear optical
microscopy of living cells by second harmonic generation. Biophysical Journal
77, 3341–3349 (1999).
[31] K. R. Castleman. Digital Image Processing. Prentice Hall (1979).
[32] L. Cauller and B. Connors. Synaptic physiology of horizontal afferents to layer I in
slices of rat SI neocortex. Journal of Neuroscience 14(3), 751–762 (1994).
[33] V. Centonze and J. White. Multiphoton Excitation Provides Optical Sections from
Deeper within Scattering Specimens than Confocal Imaging. Biophysical Journal
75(4), 2015–2024 (1998).
[34] V. Chitravanshi and H. Sapru. Microinjections of glycine into the pre-Botzinger
complex inhibit phrenic nerve activity in the rat. Brain Research 947(1), 25–33
(2002).
[35] E. Civillico and D. Contreras. Integration of Evoked Responses in Supragranular Cortex Studied With Optical Recordings In Vivo. Journal of Neurophysiology
96(1), 336 (2006).
[36] R. Clarke, A. Zouni, and J. Holzwarth. Voltage sensitivity of the fluorescent probe
RH421 in a model membrane system. Biophysical Journal 68(4), 1406–1415
(1995).
188
[37] J. Clendenon, C. Phillips, R. Sandoval, S. Fang, and K. Dunn. Voxx: a PC-based,
near real-time volume rendering system for biological microscopy. American Journal of Physiology – Cell Physiology 282(1), C213–8 (2002).
[38] L. B. Cohen and B. M. Salzberg. Optical measurement of membrane potential.
Reviews of Physiology, Biochemistry and Pharmacology 83 (1978).
[39] L. Cohen, R. Keynes, and B. Hille. Light scattering and birefringence changes
during nerve activity. Nature 218(140), 438–41 (1968).
[40] L. Cohen, B. Salzberg, H. Davila, W. Ross, D. Landowne, A. Waggoner, and
C. Wang. Changes in axon fluorescence during activity: Molecular probes of membrane potential. Journal of Membrane Biology 19(1), 1–36 (1974).
[41] M. Cohen. Central determinants of respiratory rhythm. Annual Review of Physiology 43, 91–104 (1981).
[42] E. Collini, C. Ferrante, and R. Bozio. Strong enhancement of the two-photon absorption of tetrakis (4-sulfonatophenyl) porphyrin diacid in water upon aggregation.
Journal of physical chemistry. B, Condensed matter, materials, surfaces, interfaces,
& biophysical chemistry 109(1), 2–5 (2005).
[43] D. Contreras and R. Llinas. Voltage-sensitive dye imaging of neocortical spatiotemporal dynamics to afferent activation frequency. Journal of Neuroscience
21(23), 9403–9413 (2001).
189
[44] B. Cormack, R. Valdivia, and S. Falkow. FACS-optimized mutants of the green
fluorescent protein. Gene 173(1), 33–38 (1996).
[45] V. Crépel, D. Aronov, A. Represa, Y. Ben-Ari, and R. Cossart. A parturitionassociated nonsynaptic coherent activity pattern in the developing hippocampus.
Neuron 54.
[46] B. Cumpston, S. Ananthavel, S. Barlow, D. Dyer, J. Ehrlich, L. Erskine, A. Heikal,
S. Kuebler, I. Lee, D. McCord-Maughon, et al. Two-photon polymerization initiators for 3D optical data storage and microfabrication. Nature 398(6722), 51–54
(1999).
[47] H. Davila, B. Salzberg, L. Cohen, and A. Waggoner. A large change in axon fluorescence that provides a promising method for measuring membrane potential.
Nature: New Biology 241(109), 159–60 (1973).
[48] D. Delmarre, N. Hioka, R. Boch, E. Sternberg, and D. Dolphin. Aggregation studies
of benzoporphyrin derivative. Canadian Journal of Chemistry 79(5-6), 1068–1074
(2001).
[49] W. Denk, J. Strickler, and W. Webb. Two-photon laser scanning fluorescence microscopy. Science 248(4951), 73 (1990).
[50] D. Dombeck, M. Blanchard-Desce, and W. Webb. Optical Recording of Action
Potentials with Second-Harmonic Generation Microscopy. Journal of Neuroscience
24(4), 999 (2004).
190
[51] W. Douglas. A possible mechanism of neurosecretion-release of vasopressin by
depolarization and its dependence on calcium. Nature 197(4862), 81–82 (1963).
[52] P. Dragsten and W. Webb. Mechanism of the membrane potential sensitivity of the
fluorescent membrane probe merocyanine 540. Biochemistry 17(24), 5228–5240
(1978).
[53] M. Drobizhev, A. Karotki, M. Kruk, and A. Rebane. Drastic enhancement of twophoton absorption in porphyrins associated with symmetrical electron-accepting
substitution. Chemical Physics Letters 361(5–6), 504–512 (2002).
[54] M. Drobizhev, A. Karotki, and A. Rebane. Persistent spectral hole burning by
simultaneous two-photon absorption. Chemical Physics Letters 334(1), 76–82
(2001).
[55] M. Drobizhev, Y. Stepanenko, Y. Dzenis, A. Karotki, A. Rebane, P. Taylor, and
H. Anderson. Understanding strong two-photon absorption in-conjugated porphyrin dimers via double-resonance enhancement in a three-level model. Journal
of the American Chemical Society 126(47), 15352–15353 (2004).
[56] M. Drobizhev, Y. Stepanenko, Y. Dzenis, A. Karotki, A. Rebane, P. Taylor, and
H. Anderson. Extremely strong near-IR two-photon absorption in conjugated porphyrin dimers: Quantitative description with three-essential-states model. Journal
of Physical Chemistry B 109(15), 7223–7236 (2005).
191
[57] T. Durduran, M. Burnett, G. Yu, C. Zhou, D. Furuya, A. Yodh, J. Detre, and
J. Greenberg. Spatiotemporal Quantification of Cerebral Blood Flow During Functional Activation in Rat Somatosensory Cortex Using Laser-Speckle Flowmetry.
Journal of Cerebral Blood Flow & Metabolism 24, 518–525 (2004).
[58] M. Dutschmann and J. Paton. Glycinergic inhibition is essential for co-ordinating
cranial and spinal respiratory motor outputs in the neonatal rat. The Journal of
Physiology 543(2), 643–653 (2002).
[59] C. Eggeling, A. Volkmer, and C. Seidel. Molecular Photobleaching Kinetics of
Rhodamine 6 G by One- and Two-Photon Induced Confocal Fluorescence Microscopy. ChemPhysChem 6(5), 791–804 (2005).
[60] A. Egner and S. Hell. Time multiplexing and parallelization in multifocal multiphoton microscopy. Journal of the Optical Society of America A 17(7), 1192–1201
(2000).
[61] C. Euler. Brain-stem mechanisms for generation and control of the breathing.
Handbook of Physiology, The Respiratory System 2, 1–67 (1986).
[62] K. Ezure.
Synaptic connections between medullary respiratory neurons and
considerations on the genesis of respiratory rhythm. Progress in Neurobiology
35(6), 429–50 (1990).
[63] K. Ezure, M. Manabe, and H. Yamada. Distribution of medullary respiratory neurons in the rat. Brain Research 455(2), 262–70 (1988).
192
[64] K. Ezure, I. Tanaka, and M. Kondo. Glycine Is Used as a Transmitter by Decrementing Expiratory Neurons of the Ventrolateral Medulla in the Rat. Journal of
Neuroscience 23(26), 8941 (2003).
[65] K. Ezure, I. Tanaka, and Y. Saito. Activity of brainstem respiratory neurones just
before the expiration-inspiration transition in the rat. The Journal of Physiology
547(2), 629–640 (2003).
[66] J. Fisher, B. Salzberg, and A. Yodh. Near infrared two-photon excitation crosssections of voltage-sensitive dyes. Journal of Neuroscience Methods 148(1), 94–
102 (2005).
[67] J. Fisher, E. Civillico, D. Contreras, and A. Yodh. In vivo fluorescence microscopy
of neuronal activity in three dimensions by use of voltage-sensitive dyes. Optics
Letters 29(1), 71–73 (2004).
[68] J. Fisher, V. Marchenko, A. Yodh, and R. Rogers. Spatiotemporal activity patterns
during respiratory rhythmogenesis in the rat ventrolateral medulla. Journal of Neurophysiology 95(3), 1982–1991 (2006).
[69] M. Flourens. Note sur le point vital de la moelle allongée. Comptes rendus des
séances de la Société de biologie et de ses filiales 33, 437–439 (1851).
[70] B. Flusberg, E. Cocker, W. Piyawattanametha, J. Jung, E. Cheung, and
M. Schnitzer. Fiber-optic fluorescence imaging. Nature Methods 2, 941–950
(2005).
193
[71] B. Flusberg, J. Jung, E. Cocker, E. Anderson, and M. Schnitzer. In vivo brain imaging using a portable 3.9 gram two-photon fluorescence microendoscope. Optics
Letters 30(17), 2272–2274 (2005).
[72] P. Fromherz and C. Muller. Voltage-sensitive fluorescence of amphiphilic hemicyanine dyes in neuron membrane. Biochimica et biophysica acta 1150(2), 111–22
(1993).
[73] H. Gainer, S. Wolfe Jr, A. Obaid, and B. Salzberg.
Action potentials and
frequency-dependent secretion in the mouse neurohypophysis. Neuroendocrinology 43(5), 557–63 (1986).
[74] P. Ghoroghchian, P. Frail, K. Susumu, D. Blessington, A. Brannan, F. Bates,
B. Chance, D. Hammer, and M. Therien. Near-infrared-emissive polymersomes:
Self-assembled soft matter for in vivo optical imaging. Proceedings of the National
Academy of Sciences 102(8), 2922–2927 (2005).
[75] J. Gibson, M. Beierlein, and B. Connors. Two networks of electrically coupled
inhibitory neurons in neocortex. Nature 402, 75–79 (1999).
[76] P. Gogan, I. Schmiedel-Jakob, Y. Chitti, and S. Tyc-Dumont. Fluorescence imaging
of local membrane electric fields during the excitation of single neurons in culture.
Biophysical Journal 69(2), 299–310 (1995).
[77] M. Göppert-Mayer. Über Elementarakte mit zwei Quantensprüngen. Annalen der
Physik 401(3), 273–294 (1931).
194
[78] P. Gray, W. Janczewski, N. Mellen, D. McCrimmon, and J. Feldman. Normal
breathing requires preBötzinger complex neurokinin-1 receptor-expressing neurons. Nature Neuroscience 4(9), 927–30 (2001).
[79] GRINTECH GmbH. Grintech gmbh “gradient index optics technology”. http:
//www.grintech.de (2007).
[80] A. Grinvald, A. Fine, I. Farber, and R. Hildesheim. Fluorescence monitoring of
electrical responses from small neurons and their processes. Biophysical Journal
42(2), 195–198 (1983).
[81] A. Grinvald, R. Frostig, E. Lieke, and R. Hildesheim. Optical imaging of neuronal
activity. Physiological reviews 68(4), 1285–1366 (1988).
[82] A. Grinvald, E. Lieke, R. Frostig, and R. Hildesheim. Cortical point-spread function and long-range lateral interactions revealed by real-time optical imaging of
macaque monkey primary visual cortex. Journal of Neuroscience 14(5), 2545–
2568 (1994).
[83] A. Grinvald, A. Manker, and M. Segal. Visualization of the spread of electrical
activity in rat hippocampal slices by voltage-sensitive optical probes. J Physiol
333, 269–91 (1982).
[84] S. Guha, K. Kang, P. Porter, J. Roach, D. Remy, F. Aranda, and D. Rao. Thirdorder optical nonlinearities of metallotetrabenzoporphyrins and a platinum polyyne. Optics Letters 17, 264–266 (1992).
195
[85] D. Hanna, M. Yuratich, and D. Cotter.
Nonlinear optics of free atoms and
molecules. Springer-Verlag New York (1979).
[86] F. Hayashi, S. Coles, and D. McCrimmon. Respiratory Neurons Mediating the
Breuer-Hering Reflex Prolongation of Expiration in Rat. Journal of Neuroscience
16(20), 6526 (1996).
[87] F. Hayashi and J. Lipski. The role of inhibitory amino acids in control of respiratory
motor output in an arterially perfused rat. Respiration Physiology 89(1), 47–63
(1992).
[88] R. Hellwarth and P. Christensen. Nonlinear optical microscopic examination of
structure in polycrystalline ZnSe. Optics Communications 12(3), 318–322 (1974).
[89] F. Helmchen, M. Fee, D. Tank, and W. Denk. A Miniature Head-Mounted TwoPhoton Microscope High-Resolution Brain Imaging in Freely Moving Animals.
Neuron 31(6), 903–912 (2001).
[90] S. Hess and W. Webb. Measurement of fluorescence signal of a voltage-sensitive
dye using two-photon excitation. Biophysical Journal 74, A201 (1998).
[91] G. Hilaire and B. Duron. Maturation of the Mammalian Respiratory System. Physiological Reviews 79(2), 325–360 (1999).
[92] Y. Hiraoka, J. Sedat, and D. Agard. Determination of three-dimensional imaging
properties of a light microscope system. Partial confocal behavior in epifluorescence microscopy. Biophysical Journal 57(2), 325–333 (1990).
196
[93] J. Holthoff and O. Witte. Intrinsic optical signals in rat neocortical slices measured
with near-infrared dark-field microscopy reveal changes in extracellular space.
Journal of Neuroscience 16(8), 2740–2749 (1996).
[94] A. Hopt and E. Neher. Highly Nonlinear Photodamage in Two-Photon Fluorescence
Microscopy. Biophysical Journal 80(4), 2029–2036 (2001).
[95] B. Howard and M. Tabatabai. Localization of the medullary respiratory neurons in
rats by microelectrode recording. Journal of Applied Physiology 39(5), 812–817
(1975).
[96] G. Hubener, A. Lambacher, and P. Fromherz. Anellated hemicyanine dyes with
large symmetrical solvatochromism of absorption and fluorescence. Journal of
Physical Chemistry B 107(31), 7896–7902 (2003).
[97] S. Johnson, N. Koshiya, and J. Smith. Isolation of the kernel for respiratory rhythm
generation in a novel preparation: the pre-Botzinger complex” island”. Journal of
Neurophysiology 85(4), 1772–6 (2001).
[98] K. Jones, S. Lundgren, and A. Chakravorty. A calculus of variations demonstration:
The gradient index lens. Am. J. Phys. 56(12), 1099 (1988).
[99] J. Jung and M. Schnitzer. Multiphoton endoscopy. Optics Letters 28(11), 902–904
(2003).
197
[100] A. Karotki, M. Drobizhev, Y. Dzenis, P. Taylor, H. Anderson, A. Rebane, and
S. Matter. Dramatic enhancement of intrinsic two-photon absorption in a conjugated porphyrin dimer. Phys. Chem. Chem. Phys 6, 7–10 (2004).
[101] A. Karotki, M. Drobizhev, M. Kruk, C. Spangler, E. Nickel, N. Mamardashvili, and
A. Rebane. Enhancement of two-photon absorption in tetrapyrrolic compounds.
Journal of the Optical Society of America B 20(2), 321–332 (2003).
[102] S. Kasparov, A. Teschemacher, and J. Paton. Dynamic confocal imaging in acute
brain slices and organotypic slice cultures using a spectral confocal microscope
with single photon excitation. Experimental Physiology 87(06), 715–724 (2002).
[103] D. Kim, K. Kim, S. Yazdanfar, and P. So. Optical biopsy in high-speed handheld
miniaturized multifocal multiphoton microscopy. Proceedings of SPIE 5700, 14
(2005).
[104] K. H. Kim, C. Buehler, and P. T. C. So. High-speed, two-photon scanning microscope. Applied Optics 38 (1999).
[105] D. Kleinfeld and K. Delaney. Distributed representation of vibrissa movement in
the upper layers of somatosensory cortex revealed with voltage-sensitive dyes. The
Journal of Comparative Neurology 375(1), 89–108 (1996).
[106] D. Kleinfeld, K. Delaney, M. Fee, J. Flores, D. Tank, and A. Gelperin. Dynamics
of propagating waves in the olfactory network of a terrestrial mollusk: an electrical
and optical study. Journal of Neurophysiology 72(3), 1402–1419 (1994).
198
[107] A. Konnerth, A. Obaid, and B. Salzberg. Optical recording of electrical activity
from parallel fibres and other cell types in skate cerebellar slices in vitro. The
Journal of Physiology 393(1), 681–702 (1987).
[108] N. Koshiya and J. Smith. Neuronal pacemaker for breathing visualized in vitro.
Behavioral Ecology and Sociobiology 20, 307–311 (1987).
[109] B. Kuhn and P. Fromherz. Anellated hemicyanine dyes in a neuron membrane:
molecular stark effect and optical voltage recording. Journal of Physical Chemistry
B 107(31), 7903–7913 (2003).
[110] B. Kuhn, P. Fromherz, and W. Denk. High Sensitivity of Stark-Shift VoltageSensing Dyes by One-or Two-Photon Excitation Near the Red Spectral Edge. Biophysical Journal 87(1), 631–639 (2004).
[111] R. Kurtz, M. Fricke, J. Kalb, P. Tinnefeld, and M. Sauer. Application of multiline
two-photon microscopy to functional in vivo imaging. Journal of Neuroscience
Methods 151(2), 276–286 (2006).
[112] N. Laaris and A. Keller. Functional Independence of Layer IV Barrels. Journal of
Neurophysiology 87(2), 1028–1034 (2002).
[113] J. Lakowicz. Principles of Fluorescence Spectroscopy Second Edition. Kluwer
Academic (1999).
199
[114] V. Lin, S. DiMagno, and M. Therien.
porphyrins:
Highly conjugated, acetylenyl bridged
New models for light-harvesting antenna systems.
Science
264(5162), 1105 (1994).
[115] R. Llinas, U. Ribary, D. Contreras, and C. Pedroarena. The Neuronal Basis for Consciousness. Philosophical Transactions: Biological Sciences 353(1377), 1841–
1849 (1998).
[116] R. Llinas. The intrinsic electrophysiological properties of mammalian neurons:
insights into central nervous system function. Science 242(4886), 1654 (1988).
[117] L. Loew. Design and characterization of electrochromic membrane probes. Journal
of Biochemical and Biophysical Methods 6(3), 243–60 (1982).
[118] L. Loew, L. Cohen, B. Salzberg, A. Obaid, and F. Bezanilla. Charge-shift probes of
membrane potential. Characterization of aminostyrylpyridinium dyes on the squid
giant axon. Biophysical Journal 47(1), 71–77 (1985).
[119] H. Lotem, R. Lynch Jr, and N. Bloembergen. Interference between Raman resonances in four-wave difference mixing. Physical Review A 14(5), 1748–1755
(1976).
[120] A. Ludlow, H. Nelson, and S. Bergeson. Two-photon absorption in potassium niobate. Journal of the Optical Society of America B 18(12), 1813–1820 (2001).
[121] T. Lumsden. Observations on the respiratory centres in the cat. J Physiol 57(34), 153–160 (1923).
200
[122] Lynch, R. Human physiology course notes (iphy 3430), university of colorado,
boulder. http://www.colorado.edu/kines/Class/IPHY3430-200/
12respir.html (2007).
[123] Z. Mainen, M. Maletic-Savatic, S. Shi, Y. Hayashi, R. Malinow, and K. Svoboda.
Two-Photon Imaging in Living Brain Slices. Methods 18(2), 231–239 (1999).
[124] U. Markstahler, E. Kremer, S. Kimmina, K. Becker, and D. Richter. Effects of functional knock-out of alpha 1 glycine-receptors on breathing movements in oscillator
mice. Respiratory Physiology and Neurobiology 130(1), 33–42 (2002).
[125] S. Maruo, O. Nakamura, and S. Kawata. Three-dimensional microfabrication with
two-photon-absorbed photopolymerization. Optics Letters 22(2), 132–134 (1997).
[126] D. McCrimmon, A. Monnier, F. Hayashi, and E. Zuperku. Pattern Formation And
Rhythm Generation In The Ventral Respiratory Group. Clinical and Experimental
Pharmacology and Physiology 27(1-2), 126–131 (2000).
[127] D. McCrimmon, J. Ramirez, S. Alford, and E. Zuperku. Unraveling the mechanism
for respiratory rhythm generation. BioEssays 22(1), 6–9 (2000).
[128] K. McEwan, G. Bourhill, J. Robertson, and H. Anderson. The Nonlinear Optical
Characterization of Meso-Substituted Porphyrin Dyes. Journal of Nonlinear Optical Physics and Materials 9(4), 451–468 (2000).
[129] Melles Griot, Inc. Components Catalog. Gradient-index lenses. http://www.
mellesgriot.com/pdf/0015.16-15.20.pdf (2007).
201
[130] J. Mertz and L. Moreaux. Second-harmonic generation by focused excitation of
inhomogeneously distributed scatterers. Optics Communications 196(1), 325–330
(2001).
[131] A. Millard, L. Jin, A. Lewis, and L. Loew. Direct measurement of the voltage
sensitivity of second-harmonic generation from a membrane dye in patch-clamped
cells. Optics Letters 28(14), 1221–1223 (2003).
[132] A. Millard, L. Jin, J. Wuskell, D. Boudreau, A. Lewis, and L. Loew. Wavelengthand Time-Dependence of Potentiometric Non-linear Optical Signals from Styryl
Dyes. Journal of Membrane Biology 208(2), 103–111 (2005).
[133] Molecular Probes / Invitrogen Corporation. Long-Wavelength Calcium Indicators
September (2005).
[134] H. Morise, O. Shimomura, F. Johnson, and J. Winant. Intermolecular energy transfer in the bioluminescent system of Aequorea. Biochemistry 13(12), 2656–2662
(1974).
[135] M. Muschol, P. Kosterin, M. Ichikawa, and B. Salzberg. Activity-Dependent Depression of Excitability and Calcium Transients in the Neurohypophysis Suggests a
Model of HStuttering Conduction. Journal of Neuroscience 23(36), 11352–11362
(2003).
202
[136] T. Nielsen, M. Fricke, D. Hellweg, and P. Andresen. High efficiency beam splitter
for multifocal multiphoton microscopy. Journal of Microscopy 201(3), 368–376
(2001).
[137] J. Nordmann. Ultrastructural morphometry of the rat neurohypophysis. Journal of
Anatomy 123(1), 213–8 (1977).
[138] M. Nuriya, J. Jiang, B. Nemet, K. Eisenthal, and R. Yuste. Imaging membrane
potential in dendritic spines. Proceedings of the National Academy of Sciences
103(3), 786–790 (2006).
[139] A. Obaid, T. Koyano, J. Lindstrom, T. Sakai, and B. Salzberg. Spatiotemporal
Patterns of Activity in an Intact Mammalian Network with Single-Cell Resolution:
Optical Studies of Nicotinic Activity in an Enteric Plexus. Journal of Neuroscience
19(8), 3073–3093 (1999).
[140] A. Obaid, L. Loew, J. Wuskell, and B. Salzberg. Novel naphthylstyryl-pyridium
potentiometric dyes offer advantages for neural network analysis. Journal of Neuroscience Methods 134(2), 179–90 (2004).
[141] K. Ogawa, T. Zhang, K. Yoshihara, and Y. Kobuke. Large third-order optical nonlinearity of self-assembled porphyrin oligomers. Journal of the American Chemical
Society 124(1), 22–3 (2002).
203
[142] M. Oheim, E. Beaurepaire, E. Chaigneau, J. Mertz, and S. Charpak. Two-photon
microscopy in brain tissue: parameters influencing the imaging depth. Journal of
Neuroscience Methods 111 (2001).
[143] A. Oliva Jr, M. Jiang, T. Lam, K. Smith, and J. Swann. Novel Hippocampal Interneuronal Subtypes Identified Using Transgenic Mice That Express Green Fluorescent Protein in GABAergic Interneurons. Journal of Neuroscience 20(9), 3354
(2000).
[144] H. Onimaru and I. Homma. A Novel Functional Neuron Group for Respiratory
Rhythm Generation in the Ventral Medulla. Journal of Neuroscience 23(4), 1478
(2003).
[145] H. Orbach and L. Cohen. Optical monitoring of activity from many areas of the in
vitro and in vivo salamander olfactory bulb: a new method for studying functional
organization in the vertebrate central nervous system. Journal of Neuroscience
3(11), 2251–2262 (1983).
[146] F. Patolsky, B. Timko, G. Yu, Y. Fang, A. Greytak, G. Zheng, and C. Lieber. Detection, Stimulation, and Inhibition of Neuronal Signals with High-Density Nanowire
Transistor Arrays. Science 313(5790), 1100 (2006).
[147] J. Paton. A working heart-brainstem preparation of the mouse. Journal of Neuroscience Methods 65(1), 63–8 (1996).
204
[148] J. Paton, J. Ramirez, and D. Richter. Mechanisms of respiratory rhythm generation change profoundly during early life in mice and rats. Neuroscience Letters
170(1), 167–70 (1994).
[149] C. C. H. Petersen, A. Grinvald, and B. Sakmann. Spatiotemporal dynamics of sensory responses in layer 2/3 of rat barrel cortex measured in vivo by voltage-sensitive
dye imaging combined with whole-cell voltage recordings and neuron reconstructions. Journal of Neuroscience 23(4), 1298–1309 (2003).
[150] W. Peticolas. Multiphoton Spectroscopy. Annual Review of Physical Chemistry
18(1), 233–260 (1967).
[151] O. Pierrefiche, S. Schwarzacher, A. Bischoff, and D. Richter. Blockade of synaptic inhibition within the pre-Botzinger complex in the cat suppresses respiratory
rhythm generation in vivo. The Journal of Physiology 509(1), 245–254 (1998).
[152] M. Posner and M. Raichle. The Neuroimaging of Human Brain Function. Proceedings of the National Academy of Sciences of the United States of America
95(3), 763–764 (1998).
[153] J. Ramirez, S. Schwarzacher, O. Pierrefiche, B. Olivera, and D. Richter. Selective
lesioning of the cat pre-Botzinger complex in vivo eliminates breathing but not
gasping. The Journal of Physiology 507(3), 895–907 (1998).
[154] D. M. Rector, R. F. Rogers, and J. S. George. A focusing image probe for assessing
neural activity in vivo. Journal of Neuroscience Methods 91(1-2), 135–145 (1999).
205
[155] D. Rector and J. George. Continuous image and electrophysiological recording
with real-time processing and control. Methods 25(2), 151–163 (2001).
[156] D. Rector, R. Rogers, J. Schwaber, R. Harper, and J. George. Scattered-light imaging in vivo tracks fast and slow processes of neurophysiological activation. Neuroimage 14(5), 977–94 (2001).
[157] B. Richards and E. Wolf. Electromagnetic Diffraction in Optical Systems. II. Structure of the Image Field in an Aplanatic System. Proceedings of the Royal Society
of London. Series A, Mathematical and Physical Sciences 253(1274), 358–379
(1959).
[158] D. Richter. Neural regulation of respiration: rhythmogenesis and afferent control.
Comprehensive Human Physiology 2, 2079–2095 (1996).
[159] D. Richter, D. Ballantyne, and J. Remmers. How Is the Respiratory Rhythm Generated? A Model. News in Physiological Sciences 1, 109–112 (1986).
[160] D. Richter and K. Spyer. Studying rhythmogenesis of breathing: comparison of in
vivo and in vitro models. Trends in Neuroscience 24(8), 464–72 (2001).
[161] C. J. Ross, W.N. and N. Lasser-Ross. Imaging neurons. A laboratory manual chapter Imaging sodium dynamics in neurons, pages 52.1–52.8. Cold Spring Harbor
Press (2000).
[162] W. Ross and L. Reichardt. Species-specific effects on the optical signals of voltagesensitive dyes. Journal of Membrane Biology 48(4), 343–356 (1979).
206
[163] W. Ross, B. Salzberg, L. Cohen, and H. Davila. A large change in dye absorption
during the action potential. Biophysical Journal 14(983), 86 (1974).
[164] W. Ross, B. Salzberg, L. Cohen, A. Grinvald, H. Davila, A. Waggoner, and
C. Wang. Changes in absorption, fluorescence, dichroism, and birefringence in
stained giant axons: Optical measurement of membrane potential. Journal of Membrane Biology 33(1), 141–183 (1977).
[165] I. Rybak, J. Paton, R. Rogers, and W. John. Generation of the respiratoryrhythm:
state-dependencyand switching. Neurocomputing 44(46), 605–614 (2002).
[166] I. Rybak, J. Paton, and J. Schwaber. Modeling Neural Mechanisms for Genesis
of Respiratory Rhythm and Pattern. I. Models of Respiratory Neurons. Journal of
Neurophysiology 77(4), 1994–2006 (1997).
[167] B. Salzberg. Optical recording of electrical activity in neurons using molecular
probes. Current Methods in Cellular Neurobiology 3, 139–187 (1983).
[168] B. Salzberg. Large and rapid changes in light scattering accompany secretion by
nerve terminals in the mammalian neurohypophysis. The Journal of General Physiology 86(3), 395–411 (1985).
[169] B. Salzberg, H. Davila, and L. Cohen. Optical recording of impulses in individual
neurones of an invertebrate central nervous system. Nature 246(5434), 508–509
(1973).
207
[170] B. Salzberg, P. Kosterin, M. Muschol, A. Obaid, S. Rumyantsev, Y. Bilenko, and
M. Shur. An ultra-stable non-coherent light source for optical measurements in
neuroscience and cell physiology. Journal of Neuroscience Methods 141(1), 165–
9 (2005).
[171] K. Sato, Y. Momose-Sato, T. Sakai, A. Hirota, and K. Kamino. Responses to glossopharyngeal stimulus in the early embryonic chick brainstem: spatiotemporal patterns in three dimensions from repeated multiple-site optical recording of electrical
activity. Journal of Neuroscience 15(3), 2123–2140 (1995).
[172] S. Schwarzacher, Z. Wilhelm, K. Anders, and D. Richter. The medullary respiratory
network in the rat. The Journal of Physiology 435(1), 631–644 (1991).
[173] C. Sheppard, R. Kompfner, J. Gannaway, and D. Walsh. The scanning harmonic
optical microscope, vol. QE13. IEEE/OSA Conf Laser Eng Appl. Washington, DC.
p 100D (1977).
[174] H. Sheppard, CJR Matthews. Imaging in high-aperture optical systems. Journal of
the Optical Society of America A 4, 1354–1360 (1987).
[175] D. Shoham, D. Glaser, A. Arieli, T. Kenet, C. Wijnbergen, Y. Toledo,
R. Hildesheim, and A. Grinvald. Imaging Cortical Dynamics at High Neurotechnique Spatial and Temporal Resolution with Novel Blue Voltage-Sensitive Dyes.
Neuron 24, 791–802 (1999).
208
[176] A. Silva and A. Koretsky. Laminar specificity of functional MRI onset times during
somatosensory stimulation in rat. Proceedings of the National Academy of Sciences
99(23), 15182–15187 (2002).
[177] P. Sims, A. Waggoner, and C. Wang. Hoffman, 3. F. 1974. Studies on the mechanism by which cyanine dyes measure membrane potential in red blood cells and
phosphatidylcholine vesicles. Biochemistry 13, 3315–3323.
[178] L. Sjulson and G. Miesenbock. Optical Recording of Action Potentials and Other
Discrete Physiological Events: A Perspective from Signal Detection Theory. Physiology 22(1), 47 (2007).
[179] J. Smith, H. Ellenberger, K. Ballanyi, D. Richter, and J. Feldman. Pre-Botzinger
complex: a brainstem region that may generate respiratory rhythm in mammals.
Science 254(5032), 726 (1991).
[180] L. Song, C. Varma, J. Verhoeven, and H. Tanke. Influence of the triplet excited state
on the photobleaching kinetics of fluorescein in microscopy. Biophysical Journal
70(6), 2959–2968 (1996).
[181] W. St-John and J. Paton. Characterizations of eupnea, apneusis and gasping in a
perfused rat preparation. Respiratory Physiology 123(3), 201–13 (2000).
[182] R. A. Stepnoski, A. LaPorta, F. Raccuia-Behling, G. E. Blonder, R. E. Slusher, and
D. Kleinfeld. Noninvasive detection of changes in membrane potential in cultured
209
neurons by light scattering. Proceedings of the National Academy of Sciences USA
88(21), 9382–9386 (1991).
[183] C. Stosiek, O. Garaschuk, K. Holthoff, and A. Konnerth. In vivo two-photon calcium imaging of neuronal networks. Proceedings of the National Academy of Sciences 100(12), 7319–7324 (2003).
[184] Q. Sun, A. Goodchild, J. Chalmers, and P. Pilowsky. The pre-Botzinger complex
and phase-spanning neurons in the adult rat. Brain Research 809(2), 204–13
(1998).
[185] K. Susumu, T. Duncan, and M. Therien. Potentiometric, electronic structural,
and ground- and excited-state optical properties of conjugated Bis [(Porphinato)
zinc (II)] compounds featuring proquinoidal spacer units. Journal of the American
Chemical Society 127(14), 5186–5195 (2005).
[186] K. Svoboda, W. Denk, D. Kleinfeld, and D. Tank. In vivo dendritic calcium dynamics in neocortical pyramidal neurons. Nature 385(6612), 161–165 (1997).
[187] K. Svoboda, F. Helmchen, W. Denk, and D. Tank. Spread of dendritic excitation
in layer 2/3 pyramidal neurons in rat barrel cortex in vivo. Nature Neuroscience 2
(1999).
[188] I. Tasaki, A. Watanabe, R. Sandlin, and L. Carnay. Changes in Fluorescence, Turbidity, and Birefringence Associated with Nerve Excitation. Proceedings of the
National Academy of Sciences 61(3), 883–888 (1968).
210
[189] M. Tokumasu, Y. Nakazono, H. Ide, K. Akagawa, and H. Onimaru. Optical recording of spontaneous respiratory neuron activity in the rat brain stem. Japanese Journal of Physiology 51(5), 613–9 (2001).
[190] R. Tsien. Calcium as a Cellular Regulator chapter Monitoring cell calcium, pages
28–54. Oxford University Press (1999).
[191] J. Vergara. Nile blue fluorescence signals from cut single muscle fibers under voltage or current clamp conditions. The Journal of General Physiology 72(6), 775–
800 (1978).
[192] A. Waggoner. Dye indicators of membrane potential. Annual Review of Biophysics
and Bioengineering 8, 47–68 (1979).
[193] A. Waggoner, C. Wang, and R. Tolles. Mechanism of potential-dependent light absorption changes of lipid bilayer membranes in the presence of cyanine and oxonol
dyes. Journal of Membrane Biology 33(1), 109–140 (1977).
[194] G. Wood, M. Miller, and A. Mott. Investigation of tetrabenzporphyrin by the Z-scan
technique. Optics Letters 20(9), 973–975 (1995).
[195] C. Xu. Imaging neurons. A laboratory manual chapter Two-photon cross sections
of indicators, pages 19.1–19.9. Cold Spring Harbor Press (2000).
[196] C. Xu and W. Webb. Measurement of two-photon excitation cross sections of
molecular fluorophores with data from 690 to 1050 nm. Journal of the Optical
Society of America B 13(3), 481–491 (1996).
211
[197] H. Yae, S. A. Elias, and T. J. Ebner. Deblurring of 3-dimensional patterns of evoked
rat cerebellar cortical activity: a study using voltage-sensitive dyes and optical sectioning. Journal of Neuroscience Methods 42(3), 195–209 (1992).
[198] E. Yoder and D. Kleinfeld. Cortical imaging through the intact mouse skull using two-photon excitation laser scanning microscopy. Microscopy Research and
Technique 56(4), 304–305 (2002).
[199] A. G. Yodh and D. A. Boas. Biomedical Photonics. CRC Press (2003).
[200] R. Yuste, F. Lanni, and A. Konnerth. Imaging Neurons. Cold Spring Harbor Laboratory (2000).
[201] R. Yuste, D. W. Tank, and D. Kleinfeld. Functional study of the rat cortical microcircuitry with voltage- sensitive dye imaging of neocortical slices. Cerebral Cortex
7 (1997).
212