COMBINED OPTICAL STIMULATION AND ELECTRICAL RECORDING IN IN VIVO NEUROMODULATION
BY
JING WANG
B. Sc., NANJING UNIVERSITY, 2004
M. Eng., NANJING UNIVERSITY, 2006
Sc. M., BROWN UNIVERSITY, 2008
SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN THE DEPARTMENT OF PHYSICS AT BROWN UNIVERSITY
PROVIDENCE, RHODE ISLAND
MAY 2012
i
@ Copyright 2012 by Jing Wang
ii
This dissertation by Jing Wang is accepted in its present form by the
Department of Physics as satisfying the dissertation requirement for the
degree of Doctor of Philosophy
Date ______
Arto V. Nurmikko, Advisor
Recommended to the Graduate Council
Date ______
Rebecca D. Burwell, Reader
Date ______
James M. Valles, Jr, Reader
Approved by the Graduate Council
Date ______
Peter M. Weber, Dean of the Graduate School
iii
VITA
Jing Wang was born in Jiangxi, China on October 2nd, 1983. She received her B. Sc. and M. Eng. in Material Science and Engineering from Nanjing University in 2004 and
2006. She subsequently started her graduated study at Brown University and received her
Sc. M. in Physics in May 2008. Her scientific publications include:
PEER REVIEWED PUBLICATIONS
Jing Wang, Fabien Wagner, David A. Borton, Jiayi Zhang, Ilker Ozden, Rebecca D. Burwell, Arto V. Nurmikko, Rick van Wagenen, Ilka Diester, and Karl Deisseroth, “Integrated Device for Combined Optical Neuromodulation and Electrical Recording for Chronic In Vivo Applications”. Journal of Neural Engineering, 9: 016001, (2012). Jing Wang, Ilker Ozden, Mohamed Diagne, Fabien Wagner, David Borton, Benjamin Brush, Naubahar Agha, Rebecca Burwell, David Sheinberg, Ilka Diester, Karl Deisseroth, Arto Nurmikko “Approaches to Optical Neuromodulation from Rodents to Non-Human Primates by Integrated Optoelectronic Devices”, Invited paper, Conf Proc IEEE Eng Med Biol Soc., (2011). Jing Wang, David A. Borton, Jiayi Zhang, Rebecca D. Burwell, and Arto V. Nurmikko, “A Neurophotonic Device for Stimulation and Recording of Neural Microcircuits”, Conf Proc IEEE Eng Med Biol Soc. 2935-8 (2010). Qiang Zhang, Cuong Dang, Hayato Urabe, Jing Wang, Shouheng Sun, and Arto Nurmikko “Large ordered arrays of single photon sources based on II-VI semiconductor colloidal quantum dot”, Optics Express 16: 19592-19599 (2008). Jing Wang, Zheng-Bin Gu, Ming-Hui Lu, Di Wu, Chang-Sheng Yuan, Shan-Tao Zhang, Yan-Feng Chen, Shi-Ning Zhu, and Yong-Yuan Zhu, “Giant magnetoresistance in transition-metal-doped ZnO films”, Applied Physics Letter 88: 252110 (2006). Zheng-Bin Gu, Minghui Lu, Jing Wang, Di Wu, Shan-Tao Zhang, Xiang-Kang Meng, Yong-Yuan Zhu, Shi-Ning Zhu, and Yan-Feng Chen, Xiao-Qing Pan. “Structure,
iv optical, and magnetic properties of sputtered Mn and N-codoped ZnO films”, Applied Physics Letters 88: 082111 (2006). Zheng-Bin Gu, Chang-Sheng Yuan, Ming-Hui Lu, Jing Wang, Di, Wu, Shan-Tao Zhang, Shi-Ning Zhu, Yong-Yuan Zhu, and Yan-Feng Chen, “Magnetic and transport properties of (Mn,Co)-codoped ZnO films prepared by radio-frequency magnetron cosputtering”, Journal of Applied Physics 98: 053908 (2005).
v
ACKNOWLEDGMENTS
I would like to truly thank my advisor, Professor Arto Nurmikko, whose support, guidance and encouragement have been invaluable. His enthusiasm and dedication in research have always been inspiring to me. My sincere gratitude must go to Dr. İlker Özden, who has become my best friend and whose accompanyment has been the highlight of my graduate life. He has been an excellent scientist, great mentor and humorous friend. I would like to thank Professor Rebecca Burwell for her guidance and encouragement in research. Her vibrant energy, generosity and patience have always been unforgettable to me. I would like to thank the members of the Nurmikko lab: Dr. Cuong Dang, Dr. Joonhee Lee, Dr. Dave Borton, Fabien Wagner, Dr. Ming Yin, Dr. Juan Aceros, Sunmee Park, Farah Laiwalla, Andy Blaeser, Ben Brush, Yao Lu, Naubahar Agha, Travis May, Jacob Komar, Rizwan Huq, Kwangdong Roh, Chris Heelan, Christopher Bull, Dr. Qiang Zhang, Dr. Yoon-Kyu Song. Dr. Tolga Atay, Dr. Jiayi Zhang, Dr. Hayato Urabe, Dr. Heng Xu, Dr. Yanqiu Li, Dr. Hyunjin Kim, and all the friends at Brown who have been great professional and companions. I would like to thank all the members of Burwell lab and Connors lab: Dr. Scott Cruikshank, Dr. Tim Zolnik, Dr. Sharon Furtak, Kristin Kerr, Dr. Jonathan Ho, Devon Poeta and Fang-Chi Yang for their generous help and discussion. I would like to thank all the members of Sheinberg lab and Donoghue lab: Valerie Yorgan, Dr. Ji Dai, Dr. Dan Brooks and Dr. Carlos Vargas-Irwin for being inspiring and supportive all the time. I would like to thank Dr. Harper and the Brown IACUC, Tony McCormick, Micheal Jibitsky, Sandra Van Wagoner, Barbara Dailey and John Lee for their excellent technical and administrative support. Thank you everyone here at Brown!
vi
To my family who are the source of joy and strength within me.
献给爸爸妈妈和王佳！
vii
Table of Contents CHAPTER 1 INTRODUCTION ...... 1
CHAPTER 2 DEVICE CONCEPT OF THE SINGLE COAXIAL OPTRODE ...... 6
2. 1 The fabrication of the single coaxial optrode ...... 7
2. 2 Electrical and optical characterization of the optrode ...... 11
2. 3 Stimulation and recording functionality of the optrode ...... 18
2. 4 In-situ fluorescence detection using the single coaxial optrode ...... 23
2. 5 Summary ...... 26
CHAPTER 3 LIGHT DELIVERY AND LIGHT INDUCED EFFECTS IN BRAIN
TISSUE 27
3. 1 Optical Properties of Neural Tissue ...... 27
3. 2 Monte Carlo Simulation of Light in Brain Tissue ...... 33
3. 3 Summary ...... 47
CHAPTER 4 RECORDING FUNCTIONALITY OF THE COAXIAL OPTRODE ... 48
4. 1 Electromagnetic field generated by the neuronal activity...... 49
4. 2 Sensitivity field of electrode...... 54
4. 3 Modeling the recording sensitivity field of optrode using finite element (FE)
method 58
4. 4 Discussion ...... 69
CHAPTER 5 NEURONAL CORRELATES OF SPATIAL INFORMATION
RECORDED USING THE MEA DEVICES ...... 70
viii
5. 1 Introduction ...... 71
5. 2 Behavioral Training ...... 76
5. 3 Surgery ...... 77
5. 4 Data Acquisition ...... 78
5. 5 Result and Discussion ...... 79
CHAPTER 6 INTEGRATED DEVICE FOR COMBINED OPTICAL NEURO-
MODULATION AND ELECTRICAL RECORDING FOR CHRONIC IN VIVO
APPLICATIONS ...... 83
6. 1 Introduction ...... 83
6. 2 Materials and Methods ...... 86
6. 3 Results ...... 95
6. 4 Summary ...... 113
CHAPTER 7 CONCLUSION AND FUTURE WORKS ...... 115
SUPPLEMENTARY Chapter: Large Scale Ordered Structures of Single Photon Sources
Based on II-VI Semiconductor Colloidal Quantum Dots ...... 119
8. 1 Introduction ...... 120
8. 2 Methods and Results: QD array fabrication and characterization ...... 123
8. 3 Summary ...... 134
APPENDIX ...... 136
BIBLIOGRAPHY ...... 143
ix
List of Figures
Figure 2.1 Concept of the single coaxial optrode device...... 8
Figure 2.2 Photograph of the illumination pattern from a coaxial single optrode device, ...... 13
Figure 2.3 (a) An electric double layer forms at the tip, which is a capacitor effectively ...... 15
Figure 2.4 An example of in vivo optrode recording, measured from an anesthetized rat...... 19
Figure 2.5 Peri-stimulus time histograms (bottom in each panel) and corresponding raster plots (top in
each panel) of spiking activity recorded by the single coaxial optrode...... 20
Figure 2.6 Histology results. Left shows the representative images of anti-FP staining in LIP...... 22
Figure 2.7 Photograph of the M1 areas on both hemispheres of the transduced rat brain. I...... 24
Figure 2.8 The schematics of the fluorescence detection module attached to an optrode...... 25
Figure 2.9 Detecting the depth dependent fluorescence intensity ...... 25
Figure 3.1 Optical absorption spectra of various tissue components in the UV,...... 30
Figure 3.2 The absorption spectrum (in arbitrary unit) of ChR2 and NpHR, adapted from Ref [8]...... 31
Figure 3.3 A flowchart of the Monte Carlo method used in tracking photons in the scattering media...... 35
Figure 3.4 Schematics of weight dropping and redirecting of a single photon during the scattering...... 37
Figure 3.5 Examples of light intensity distribution in brain tissue obtained from the simulation ...... 39
Figure 3.6 Comparison of the stimluation volumes...... 40
Figure 3.7 Spatial and temporal resolved distribution of temperature change...... 45
Figure 3.8 The peak temperature (as indicated by arrows in Figure 3.7(b) ) evolves in response to a 200
ms continuous stimulation (green)...... 46
Figure 4.1 The schematic of potential difference measurement...... 53
Figure 4.2 The schematics for deriving the reciprocity theorem (figure reproduced from [58])...... 57
Figure 4.3 A function (red curve) is approximately expressed as the linear combination of the basis
functions. In this example, the basis are piecewise functions (blue curves)...... 60
Figure 4.4 The geometry of the optrode modeled by COMSOL...... 62
Figure 4.5 The materials assigned for each domain / boundary of the model...... 63
x
Figure 4.6 Simulating the reciprocal current density distribution (the sensitivity field)...... 64
Figure 4.7 The relative signal amptitude of four types of electrode geometry (see text), obtained from the
FE modeling...... 66
Figure 4.8 Estimated signal amplitude as a function of the recording area. Four types of geometry are
simulated: ring shaped (240.8 µm2), half-ring shaped (120.4 µm2), quarter-ring shaped (60.2 µm2),
and conical electrodes (33.5 µm2)...... 67
Figure 5.1 Schematic diagram of connections between PPC and other cortical/subcortical areas...... 74
Figure 5.2 An animal (M6) is performing the DNMTP task in a floor projection maze...... 75
Figure 5.3 Left panel shows the layout of the Y-shaped maze...... 77
Figure 5.4 Schematic of the implanted MEA in PPC...... 79
Figure 5.5 Delay dependent performance...... 80
Figure 5.6 (a) Rastergrams and peri-evnet histograms of two “encoding” cells...... 82
Figure 6.1 A single optrode as an in vivo electrophysiological recording tool...... 87
Figure 6.2 Overview of the optrode-MEA...... 91
Figure 6.3 The ChR2 expressing volume and the optical excitation volume...... 99
Figure 6.4 Representative examples of light activation of single units and LFPs...... 104
Figure 6.5 Spatially and temporally resolved neuronal activities from a large cortical area...... 107
Figure 6.6 Examples of single unit recordings under light modulation...... 111
Figure 6.7 Results from histology studies...... 112
Figure 7.1 (a) A 10x10 MEA integrated with a single POF...... 117
Figure 8. 1 The “bottom-up” and “top-down” fabrication approaches...... 122
Figure 8. 2 High resolution transmission electron microscope (HR-TEM) images demonstrate epitaxial
core-shell growth of CdSe-core CdS/Zn0.5Cd0.5S/ZnS shell nanoparticles...... 126
Figure 8. 3 Silica-encapsulated II-VI semiconductor colloidal QDs...... 128
Figure 8. 4 Electrostatic force self-assembly of the silica-encapsulated QDs at single-particle resolution.
...... 130
Figure 8. 5 Photon statistics measurement of the emission from single silica encapsulated II-VI
semiconductor colloidal QDs in self-assembled array...... 133
xi
CHAPTER 1 INTRODUCTION
Brain, as a vital part of the nervous system, contains a vast number of interconnected neurons (1011 neurons with 1014 synaptic connections for human brain).
Each neuron acts as a signal processing basic unit that transmits and modulates information from other parts of the brain or body. The connections and their activity- dependent plasticity induce the complex and dynamical properties of the neural system, which are believed to be the microscopic basis of many macroscopic phenomena, ranging from animal behavior to human cognition. One of the straightforward ways to study this information processing-storage machinery is to perturb the system and readout its activity. The specific goal of this thesis is to create such multi-functional tools – the
‘optrode’ and hence the optrode-microelectrode array (optrode-MEA) devices, to facilitate the neuroscience research and, ultimately, innovation in the treatments of neurological diseases.
Neural stimulation by means of electrical current injection through the conductive brain tissue has been well established in electrophysiology and clinical treatments [1-3].
However, there are uncertainties associated with the complex current pathways and non- 1 specific activation of cell bodies and processes. A new approach to neural modulation occurred with the discovery of a light-sensitive ion channel, Channelrhodopsin-2 (ChR2)
[4], followed by the discovery of an optically activated chloride pump, Halorhodopsin
(NpHR) [5] and a proton pump, Archaerhodopsin (Arch) [6]. Through the combination of genetic and optical methods these discoveries were rapidly advanced in mammalian animal models as fundamentally new methods for targeting neurons, now termed
‘optogenetics’ [7-10].
Such light induced modulation enables both neuronal excitation and inhibition, each for inducing well-defined neuronal events with millisecond time resolution [9].
Among the practical advantages of optical stimulation over electrical stimulation, the optical method has minimal instrumental interference with simultaneous electrophysiological recording techniques. Furthermore, the method is technically scalable for accessing potentially multiple target areas for controllable spatiotemporal modulation across the cortex and deeper brain structures [11-14]. To advance such prospects, development of “dual-function” chronically implanted devices for simultaneous light delivery and electrophysiological recording are needed. For localized light delivery, glass optical fibers have seen wide use in optogenetics to date, given their abundant commercial availability as flexible (though somewhat fragile) low-loss optical waveguides. In addition, an implanted optical fiber allows in vivo fluorescence detection in the intact brain for minimally invasive assessment of local opsin expression [15]. So far, recording of optically evoked neural activity in vivo in rodent and non-human primate models has been mainly limited to the use of individual extracellular electrodes. In these experiments, the electrodes are adhesively attached in parallel to the optical fiber [14-18].
2
There are, of course, a variety of techniques in electrophysiology for neural population recording, ranging from multi-electrode arrays (intracortical and electrocochleography) to optical imaging [19-21], but these methods now need integration with fiber optics for chronic implants in the freely behaving animal experiments, which is the major motivation behind this work.
In this thesis, a tapered coaxial optical waveguide construct termed “optrode” is introduced. This dual functional device provides simultaneous light delivery and electrophysiological recording capabilities. As described in Chapter 2, we first introduce the concept and design of the latest version of the single coaxial optrode. Then, we discuss the electrical and optical properties of the optrode, and their dependence on the fabrication process. We then apply the single coaxial optrode as a stand-alone unit in vivo in optical stimulation and recording in various experimental animal models - including transgenic mice, optogenetically transduced rodents and non-human primates.
Chapter 3 discusses the interaction between light and brain tissue. We will begin with the investigation of optical properties of the brain, and then employ a computational model to simulate the process of light propagation and the associated heating of brain tissue. Finally, we will review and discuss the strategies for meeting the requirements for particular experimental settings via the spatial-temporal control of light delivery in optogenetic studies.
In Chapter 4, we construct a model that describes the electric behavior of optrode, which enables mathematical analysis of its recording properties. We start with characterizing the nature of bioelectric sources and conductors. Based on a reciprocity theorem, we then discuss the source-field-probe relationship, where our unconventional
3 coaxial optrode is modeled as a conductive ring with insulated shaft in the conducting medium. We implement a finite element (FE) method to simulate the sensitivity field of recording, which varies among different types of optrodes. This model provides an insightful way of estimating and interpreting measured signals from different optrode configurations. The results of the study provide foundation for refining optrode design for the simultaneously optical stimulation and electrical recording applications.
In Chapter 5, we introduce the multi-channel recording device - Utah Micro- electrode Array (MEA). We demonstrate preliminary data of neuronal correlates of spatial representation recorded by using the MEA devices. These findings motivate us further to innovate an integrated optrode-MEA device that could help us understand the underlying neurocircuits that dictate behavioral outcome.
In Chapter 6, we demonstrate an integrated optrode-MEA device that enables neuromodulation and simultaneous mapping of electrophysiological responses of neuronal populations in ChR2-expressing rats in vivo. We will show the use of this device as an optogenetic tool for optical stimulation and electrical mapping of neural activity in a configuration which can be applied to other animal models.
Chapter 7 summarizes the main work of this dissertation and some ongoing work in our laboratory.
The Supplementary Chapter presents a separate project on II-VI semiconductor colloidal quantum dots (QDs) which was conducted earlier in our laboratory. This project introduced me to microscopy and optical systems that was a great preparation for my later work in optogenetics. We have developed a novel and efficient method of deterministically organizing colloidal particles on structured surfaces over macroscopic
4 areas. Later, as a specific demonstration, we have created highly ordered 2D arrays of single II-VI semiconductor colloidal QDs by this method. Individually, the QDs display triggered single photon emission at room temperature with characteristic photon antibunching statistics, suggesting a pathway to scalable quantum optical radiative systems.
5
CHAPTER 2 DEVICE CONCEPT OF THE
SINGLE COAXIAL OPTRODE
Optrode – “optical electrode” – comes from the idea of creating a dual modality device which is able to deliver optical stimulus and record extracellular neural signal simultaneously. An interesting note on the term “optrode” is that it is not the original optrode that has appeared in scientific literature. Back in the 1980s, the optrode was first developed as an optical-chemical sensor to detect changes in pH [22] or chemicals [23,
24] when the amount of sample is small. It consists of an optical fiber and phosphors placed to the tip of the fiber. The detected signals are the dynamic change/quenching of luminescence from the phosphors. Our optrode, created for in vivo applications of optogenetics techniques, has a completely different functionality.
In this chapter, I will introduce our single coaxial optrode, a structure that consists of a gold-coated tapered optical fiber fixed in a stainless steel reinforcing tube. The single coaxial optrode is suitable for both in vitro and in vivo electrophysiological recording. In
6 addition to its versatility in terms of optical stimulation and recording, we also emphasized its mechanical robustness, e.g. multiple penetrations of the dura mater in nonhuman primates can be accomplished with the same device. Another prominent feature of the single coaxial optrode is its capability of detecting the fluorescence (XFP that co-expressed with opsins). We will describe a fluorescence detection module, incorporated with the optogenetics excitation system, and present preliminary data showing the expression variation in transgenic and injected animals.
2. 1 The fabrication of the single coaxial optrode
We have developed several types of single coaxial optrodes, catering to various scientific needs. The basic structure of the optrode is illustrated in Figure 2.1 (a). In order to improve electrical and mechanical property of the optrode, the design and fabrication steps are tailored for various experiment settings and they are constantly under revision.
However, the general fabrication procedures, illustrated in Figure 2.1 (b), can be summarized as follows.
2. 1. 1 Shaping the fiber tip
The optrode is fabricated from a small-core step-index multimode optical fiber
(10/125 μm core/cladding, HPSC10, Thorlabs Inc.). The fiber is cut into approximately
15 cm in length. The polymer buffer on both ends is carefully removed by mechanical stripping. Fiber tips are cleaved and cleaned with acetone and methanol before processing. One end of the fiber is sharpened to minimize the neuronal damage during
7 insertion. We developed three different ways for tip shaping dependent on the application: wet chemical etching, mechanical polishing and laser-assisted heat pulling.
(a)
(b)
Figure 2.1 Concept of the single coaxial optrode device. (a) A schematics of the design and (b) the process steps of the single coaxial optrode.
8
1). In the wet etching procedure, the tapered end is created in the aqueous hydrofluoric acid (HF, 49%) at room temperature (22-24 oC). Tip angle (6-30 degrees) is controlled by the pulling rate (0 – 20 μm/min). It is critical that this step should be done in a ventilated chemical hood. We rinse the fiber with de-ionized water right after taking out from HF solution. 2). The fiber is slided into a 18 G guide tube and left 1mm of fiber exposed. We then polish it with an angle of 25 - 30 degree on a rotating polishing plate, which is controlled by a pipette beveler (BV-10, Sutter instruments). Meanwhile, the fiber is rotated constantly to form a conical shape. The tip is then ultrasonically cleaned in acetone and then in water. The mechanical polishing approach is more time consuming than other two methods, however it gives better tip strength and adhesion to the metal coating. 3). The laser-assisted pulling method gives better control of tip shape. We remove the polymer jacket at center of the fiber and mount both ends on the pullers.
Make sure that fiber is aligned to the groove and the center part sits in the scanning zone.
Fiber is heated by a scanning laser beam and a trenching force is applied to the ends.
Being thinner in diameter along the heated portion, the fiber maintains a constant ratio of the core/cladding diameter. This implies that fiber will lose its waveguiding ability along the tapered shank and a final polishing step becomes necessary. Optical connection is finished on the other end of the fiber. We then connectorize the distal end in a LC miniature connector (Ziconia ferrule, with stainless steel flange, OptoEquip) using a heat- curable epoxy (Thorlabs).
9
2. 1. 2 Coating with conductive metal layers
The metal coating step determines the recording quality of the optrodes. We remove the rest of polymer jacket after leaving the fiber in acetone overnight, and clean the fibers again with acetone and methanol. For the full ring coating we vertically mount them on the sample holder in the Angstrom thermal evaporator. For the half side coating ones, we mount them on a clean silicon wafer in the electron beam evaporator. The half side coating is accomplished by using the shade effect of electron beam evaporation. A layer of chromium or titanium (10 - 20 nm) is initially deposited to improve the adhesion between glass and gold layers. Chromium is used in most cases. A gold layer is deposited sequentially at an optimal deposition rate of 1.5 Å/s. The thickness of gold layer ranges from 100 to 300 nm, depending on the specific optrode configuration and impedance requirement. Finally, the coated fibers are carefully transferred and the coating is inspected under optical microscope.
2. 1. 3 Electrical insulation and mechanical
reinforcement
To insulate the coating except for the tip part, we use a transparent polyimide tube with inner diameter slightly larger than the fiber size. Then, the whole piece is reinforced by inserting into a 29G stainless steel tube. The insulation at the tip is finished with the
UV curable epoxy (NOA81, Thorlabs).
10
2. 1. 4 Electrical connection
A magnetic copper wire is attached to the coating on the other end. We use two part mixing silver epoxy for the electrical connection. The whole connection pathway is insulated again with a polymer tube. Finanlly, we bake the optrodes at 60 °C overnight for best mechanical performance of UV epoxy and silver epoxy.
2. 1. 5 Final polishing to minimize the optical artifact
Since the metal coating partially covers the tapered tip where the fiber core is potentially exposed, large and transient photo-electrical artifact induced by light on/off can be expected. Although the downstream processing (e.g. filtering the recorded data) could help alleviate the issue, we always perform a final polishing step which was proven to be most effective in reduceing the artifact. The polishing is conducted again on the beveler. We carefully ‘open’ the optical pathway of the tip. Instead of water, we use 0.9% saline as polishing media in order to monitor artifact amplitude as well as its impedance change. The photo-induced artifact is constantly checked using oscilloscope until it is less than 100 μV with ~1 mW total light throughput. Finally, the optrodes were encapsulated around the fiber connector by a heat shrink tube in order to prevent future fracture and separation during bending.
2. 2 Electrical and optical characterization of the
optrode
Since most fabrication steps are manually applied, the electrical properties of the
11 optrodes may vary among the fabrication batches made by different personnel. Electrical and optical characterizations are usually conducted in the following order: 1. Inspect the shape of tip and light emitting profile in dye solution under optical microscope. 2. Make electrical impedance measurements 3. Perform in vivo stimulation and recording test for selective optrode in transgenic mouse model. 4. Perform in vivo applications in other animal models, such as transduced primates. Without compromising its dual functionality and while minimizing the overall size, we are constantly modifying the design and fabrication to improve its properties.
2. 2. 1 Optical throughput
We collect the optical power output from both the aperture and possible sidewall leakage, using an integrating sphere (ILX Lightwave). The total optical throughput has been experimentally determined to be ~50 -70%. The major loss comes from the mis- coupling at the connector end. To quantify the emission profile at the tapered end, we place the optrode in a non-scattering medium Rhodamine-6G (R6G, Sigma Aldrich) solution. Green light (532 nm; World Star Tech) was coupled into the optrode for fluorescence excitation (see Figure 2. 2).
The small core fiber (10/125 μm) has numerical aperture of 0.1, therefore, the divergent angle is estimated to be 9.5o by the following relation:
( 2. 1 )
, where n is the index of refraction of solution and is the half angle of divergence.
Comparing with regular fiber (62.5/125 μm, N.A. = 0.22) whose divergent angle is 20o, the optrode has less sidewall leakage along the tapering, assuming the same tip profile.
However, once entering brain tissue, a strongly scattering medium in the visible
12 spectrum, the light would lose its coherence quickly. A Monte Carlo computational model was introduced to investigate this problem, which will be described in a later chapter.
Figure 2.2 Photograph of the illumination pattern from a coaxial single optrode device, imaged from the induced fluorescence in the Rhodamine 6G dye solution.
2. 2. 2 Impedance measurement
The measured impedance values of the optrode are within a range of 200 k to
1.5 M, enabling reliable detection of single or multi-unit activities. Impedance measurement is an important estimation of the recording quality because the impedance is an indirect indicator of exposed recording area, assuming the materials used and fabrication proceduce are identical. In this section, I discuss the simplified electric circuit of recording and its relationship with the measured impedance.
When a metal layer is immersed in the electrolyte, chemical reactions occur at the interface. If the metal is partially soluble, its ions may be involved in the reactions. In our case, we use gold, an inert metal, so that it only transfers the charge carriers- electrons through the interface. As a result, an electric potential gradient forms at the interface, 13 which opposes further movement of charges driven by chemical potential gradient. The trapped charges at the interface form an electric double layer (Figure 2.3(a)). In brief, the exposed metal tip-CSF interface acts like a leaking electrolytic capacitor [25]. Therefore, the tip of the optrode in the CSF solution can be represented as a parallel resistance Re and capacitance Ce, as shown in Figure 2.3(b). The capacitance component is related to the coating materials and preparation methods, for example, the plated platinum in saline has been reported to be on the order of 0.2 – 1 pF/μm2 at 1 kHz [25]. It has been experimentally shown that both Re and Ce are not constant, but vary as , where is the signal frequency. Although an accurate representation would be a distributed circuit, we can simplify it as a lumped circuit diagram shown in Figure 2.3 (b). The equivalent tip impedance Z can be estimated as [26]
( 2. 2 )
It has been shown that the capacitance and resistance components have similar frequency dependence, and the input-output phase shift for various metal electrodes is near 45 degree over the frequency of interest (1 Hz -10 kHz) [25, 26]. This implies that the magnitude of the impedance also varies ~ .
( 2. 3 )
In practice, the impedance measured at commonly accepted frequency 1 kHz is adequate to specify the tip properties. The conventional microelectrodes usually have 0.5-
3 MΩ impedance for good signal/noise ratio in single unit recording. In fact, contrary to popular myth, which views that higher impedance as a sign of better recording, the ideal electrode would be one with zero impedance and infinitely small tip dimensions [26].
14
Unfortunately, it is impossible to achieve this ideal case as the small tip size is, to certain extent, necessarily correlated with high impedance.
(a)
(b)
Figure 2.3 (a) An electric double layer forms at the tip, which is a capacitor effectively in parallel with a resistor. (b) The equivalent circuit of recording from a metal electrode. Za is the input impedance of the amplifier system. It is usually larger than 10 MΩ, seldomly a problem in recording. Cs is the shunt capacitance from the tip to the input of amplifier. Cs includes the capacitance from the insulation and the accumulated capacitance of all wires and connectors. It can be an issue if the insertion is deep or the insulating material has high dielectric constant ε . Rm is the resistance of the metallic part of the electrode, it’s usually very small ~1Ω. Re and Ce is the leakage resistance and capacitance of the electric double layer (see text). Rs is the speading resistance between the electrode tip and the saline bath. (figure adapated from [25] )
In Figure 2.3 (b), the circuit representation of the electrical properties of the recording system shows that the electric bilayer is essentially a simple low pass filter [26].
15
The input signal ( ) is the potential field change (due to spiking or other membrane event) at the electrode tip, and the output signal ( ) connects to the input of amplifier system. The absolute magnitude of the output can be evaluated by a simple calculation:
( 2. 4 )
Cs is the shunt capacitance from the tip to the input of amplifier. The materials used for insulation is polyimide in our design. Polyimide has low dielectric constant (ε =
3.5 at 1kHz) [27]. If the whole length covered by polyimide/stainless steel needle is 10cm, the resulting shunt capacitance is about 68.2 pF at 1kHz [25], which corresponds to shunt impedance ( ) of 15 MΩ. This indicates that signal attenuation due to Cs is usually not a problem. It can be an issue if the electrode tip impedance is large which becomes comparable with the shunting impedance (as implied by equation (2.4)). For regular metal electrode, a common practice to reduce the tip impedance without sacrificing the tip size is by platining [25]. The deposition of platinum may reduce the impedance to only 1/10 of that of unplated ones [25, 28]. In our case, the gold coating is similar to the platinum coating that it provides excellent interface properties [28, 29]. In addition, the amplitude of Johnson noise (thermal noise) generated across the electrode is estimated by [26]:
( 2. 5 ) by assuming the shunt impedance and amplifier input impedance is infinitely large. In equation (2.5), k = 1.38 ×10-23 JK-1 is the Boltzmann’s constant, T is the temperature in degree Kelvin, Z is the electrode impedance and is the signal bandwidth in Hz. Thus, with an impedance of 1 MΩ and a recording band of 10 kHz, one will expect noise level
16 on the order of 15μV. This is another reason why the optrode impedance should keep low without compromising the exposed area.
The histological data shows that the majority of cells within the mammalian CNS are with soma size of 10-20 μm. The extracellular potentials by the small cells may exceed an amplitude of 200-300 μV [26], and they are detectable only in immediate vicinity of the cell bodies (at the risk of destroying the cell by the recording probes). A failure to record the spiking activity often occurs when the exposed part has larger dimensions than that of the steep part of voltage gradient. At best, the recorded potential would be some average of that along electrode exposed region. Therefore, probing in the spreading potential field of a firing neuron by optrode, one should consider not only the impedance characterization, but also the geometry of the exposed area of the optrode.
Metal microelectrodes are easily made into a conical tip shape by convention and sufficient reproducibility. However, the optrode tips are subject to modification according to different requirements. In the following chapter, we will model and discuss the issue about how geometry of exposed coating will affect its recording properties.
A miscellaneous point should be made is the choice of reference electrode. If both the spiking and local field potential (LFP) bands are of interest, it is desirable to use for the reference a non-polarizable material, such as Ag/AgCl wire. We prepare the chloride silver reference electrode by placing the silver wire in saline and passing a DC current
[26]. Usually for experiment done in an open craniotomy or recording chamber, a large diameter Ag/AgCl wire is immersed into the CSF as a reference, For recording through burhole, we can use the stainless steel reinforce tube as reference, which was found out to have good noise level, possibly due to its local noise (e.g. respiration-related)
17 cancellation from the reinforce tube. In all cases, the lead from the optrode to the input tage of the amplifier should be kept as short as possible to reduced signal shunting and ambient noise.
2. 3 Stimulation and recording functionality of the optrode
As a first step to validate the recording functionality, we employed the single optrode in the anesthetized transgenic mouse and transduced rat model. The transgenic mouse line was Thy1-ChR2-YFP (line 18, Jackson Labs), and the transduced rats were expressing AAV5-CAMKII-C1V1-YFP. We observed contralateral twitching responses in the transgenic mouse model when the corresponding M1 forelimb area was stimulated by blue light (473 nm, ~0.2 mW, 10-20Hz 10 pulse/train). Under low anesthesia (~0.5% isoflurane or low dose of ketamine/xylazine), we found that twitching response was as reliable as the electrical stimulation (300Hz, 200μs, 20 pulses at 150μA) [30], in which we used the regular metal microelectrodes. The simultaneously electrical recording from the optrode shows the spiking activity was also modulated by light pulses (data not shown here).
Similar optical neuromodulation experiments were conducted in the rat model.
The objects were two Long-Evan rats, whose M1 forelimb region was transduced with an
AAV virus construct (AAV5-CAMKII-C1V1-YFP). We did optical stimulation and electrophysiology in rats 31 days after the viral injection, under anesthesia. We could reliably obtain modulated single/multi unit activity recording from the M1 region (Figure
2.4). However, we failed to observe the twitching behavior in the transduced rat model
18
(under low dose ketamine/xylazine anesthesia) that we so clearly observed in the transgenic mice. The optical power used for stimulation in transduced rat model is usually in the range of 1 mW to 2 mW. In a control (non-transduced area) experiment, the spontaneous spiking activity recorded from the optrode was not affected by the optical stimulation, even when the power output was ramped up to 6.7 mW (not shown here).
This data suggests that the response due to photo-toxicity is not an issue in our case.
Figure 2.4 An example of in vivo optrode recording, measured from an anesthetized rat. The virus construct (AAV5-CAMKII-C1V1-YFP) was targeting at M1 forelimb area. The recording and optrical stimulation was conducted 31 days after the viral injection. The light power output from the tip was 2 mW at 561 nm. Light modulated multi-unit acitivty was recording by the same optrode (500 kΩ) and an extracellular recording amplifier system (A-M Systems, Model 1800). The signal was band pass filtered at 300 Hz-10 kHz and the Vpp noise amplitude was about 40 μV.
19
Figure 2.5 Peri-stimulus time histograms (bottom in each panel) and corresponding raster plots (top in each panel) of spiking activity recorded by the single coaxial optrode. It shows the spiking modulation in response to 4 different light stimulus paradigms in the LIP of a ChR2 expressing non-human primate under anesthesia (isoflurane, 1.4% MAC). The light stimulus is delivered by a blue laser (473 nm, ~2mW) coupled to the optrode. The stimulus times and durations are indicated by blue bars in each panel. The recording and stimulation was performed 4 weeks after viral injection(AAV5-Thy1-ChR2-YFP).
We employed the anesthetized rhesus macaque primates to test modulation of neural activity optically [31]. The depth of injected location was 700 µm from the cortical surface. Several AAV5-virally packaged optogenetic constructs (ChR2 and NpHR) were injected to multiple sites of the cortex. We employed specially designed hollowed-out Ti-
20 screws in the skull of the primate which acted as cannula for the viral injection and subsequent repositioning of the optrode in repeated experiments over weeks. The injection protocol consisted of hydraulic minipumps in a setup where the injected viral volumes per site (along cortical columns) were 1 L with typical viral titers of ~1010
IU/mL. The single coaxial optrode device was employed in a number of cortical areas to locate and characterize sites with maximal photoactivated response. The light induced neural responses in lateral intraparietal (LIP) area were particularly prominent, which is probably related to the variability of viral injections or anesthesia levels. As an example,
Figure 2.5 shows rasters and peri-stimulus time histograms of a representative firing activity of an isolated unit in the LIP, in response to the laser pulse train stimulation at various frequencies. So far, there is no conlusive report on behavioral modification on non-human primates with a single fiber illuminating limited volume of brain tissue, althoughat the same site elicited clear behaviors[15, 16]. This may be due to 1). the limited targeting volume of transduced tissue, and/or 2). the distinct mechanism of optogenetical stimulation. For example, ChR2 fails to drive high-frequency firing of projection axons that is usually needed for evoking a movement behavior. With continued effort and enthusiasm in the field, we should be able to unravel these questions in the near future. As a first step here, we didn’t intend to look for behavior response because the optrode device and viral construct were tested in non-human primates under anesthesia.
To validate the opsin expression, we euthanized the animal eight weeks after viral injection. The brain was fixed and sectioned through the services of Neurosience
Associates. Brain slices were immunostained with antibodies against the co-expressing
21
XFP (Figure 2.6). Histology showed that 1µL of viral injections at five different depths could cause strong expression (~5 mm lateral spreading) in frontal cortex (data not shown here). The AAV5-Thy1-ChR2 expression in LIP was more densely packed in layer V
(Figure 2.6) as opposed to a column shape dictated by the diffusion of virus. After two months of optical modulated the neural recordings, we saw widespread expression of
ChR2 and NpHR, with no histological abnormalities in neurons or glia. Supported by other groups’ findings [15, 16], this study provides some evidence of the non-toxicity of the optogenetics in the non-human primate, which may provide the next generation neuromodulation therapy for humans.
Figure 2.6 Histology results. Left shows the representative images of anti-FP staining in LIP. Right shows the corresponding fluorescence images, indicating the expression was well localized on the plasma membrane at cell bodies and processes.
22
Regarding the cortical damage by multiple penetration tracks, we found a cylindrical trace of diameter 150-200 µm possibly due to the viral injection and optrode insertion (Figure 2.6 left panel). More systematic histology study should be done in other animal models and is necessary to understand the extent of possible cortical damage by the single coaxial optrode device. In the mean time, for the future experiments, we will work on the optimal procedure/design that would minimize the damage due to optrode insertion.
2. 4 In-situ fluorescence detection using the single
coaxial optrode
In optogenetics experiments, it is important to identify the regions of opsin expression for proper light targeting. In rodents, it is possible to see the injected spots on the brain surface by inspecting the fluorescence emission under a fluorescence microscope (Figure 2.7 right panel). However, it is not possible to identify the depth of expression. On the other hand, in primates, the non-transparent dura mater prevents the identification of the transduced areas even on the brain surface. This becomes an issue in optical stimulation experiments where the sites of opsin expression might be easily missed.
23
Figure 2.7 Photograph of the M1 areas on both hemispheres of the transduced rat brain. Images were taken under the stereomicroscope (left panel) and the fluorescence microscope (right panel) .
To overcome this problem, we designed and implemented a fluorescence detection module (Figure 2.8(a)) for the single coaxial optrode which will allow us to track the local YFP fluorescence intensity level (indicates opsin expression levels) around the optrode in the brain. The fluorescence module does not require any change in the design of the optrodes. It uses the optrode to deliver light into the brain and excite fluorescent proteins and collect the resulting fluorescence. With suitable spectrum separating filters, the fluorescence signal is guided to a photon counting module (SPCM-
AQR-1X, PerkinElmer) which is a highly sensitive avalanche photodiode (APD) based photodetector converting each photon to a short (35 ns) pulse of 2 Volts. Therefore, it is possible to monitor the count rate of the photons arriving at the detector which is proportional to the fluorescence intensity.
24
Figure 2.8 The schematics of the fluorescence detection module attached to an optrode. (b) The fluorescence detection module. (Courtesy of Ilker Ozden et al [32])
Figure 2.9 Detecting the depth dependent fluorescence intensity (obtained from the photon count rates) variation under different experimental conditions (see text). The red arrow indicates the location of injection in the transduced mice.( Courtesy of Ilker Ozden et al [32])
25
The implementation of the fluorescence module is shown in Figure 2.8 (b). We tested the module in the wild-type transduced (Lenti-Syn-ChR2-YFP) and transgenic
(Thy1-ChR2-YFP) mice with the optrode (Figure 2.9). In all mice we recorded the fluorescence intensity as a function of depth. Positive depths indicate recordings in the brain. The wild-type and the non-injected areas of the transduced mouse were used as controls (black and cyan circles). We observed 50x fluorescence increase in the transgenic mice and 20x in the transduced mouse compared with the control cases. The intensity change is comparable to that observed in the transduced primate with similar design of fluorescence detection system [15]. In transgenic mouse, the fluorescence intensity varied with depth, but it was significantly higher than the control levels at all depths. In transduced animal, the fluorescence intensity dropped to control level at a depth of 1.4 mm indicating the border of transduced area. The injection location in this animal was at a depth of 700 μm, suggesting a spread of expression of about 700 μm. As a summary, with the fluorescence detection module, it was possible to monitor the YFP fluorescence intensity as an indicator of opsin expression deep in the brain using an optrode.
2. 5 Summary
In summary, we have demonstrated the concept and fabrication of the single coaxial optrode device, and shown its wide applications for in vivo optical stimulation and recording in various experimental animal models. We will now focus on the physics related to the device and its application in the following chapters.
26
CHAPTER 3 LIGHT DELIVERY AND
LIGHT INDUCED EFFECTS IN BRAIN
TISSUE
This chapter will address the issue of interaction between light and brain tissue.
We will begin with the investigation of optical properties of the brain, and then employ a computational model to simulate the process of light propagation and the associated issue of heating in brain tissue. Finally, we will review and discuss the strategies for meeting the requirements for different experimental settings by controlling the spatial-temporal patterns of light delivery in optogenetics studies.
3. 1 Optical Properties of Neural Tissue
Understanding the optical properties of the brain is important not only for improving non-invasive optical methods in neuroscience research such as optogenetics,
27 functional 1-photon and 2-photon imaging but also for developing medical applications such as optical diffusion tomography for diagnostics and therapy. The photo-activation of a transduced neuron depends on many factors, including the properties of the opsin being expressed, the wavelength, intensity and duration of the stimulation light, and even the illumination history (e.g. the dark-adapted states [9, 33]). In all cases, however, the rate of absorption at given wavelength is proportional to the local photon flux, that is the number of photons per unit area per unit time. Most mammalian brains are optically non- transparent. When designing a light delivery system, one wishes to spatially control the illumination in the brain in vivo. Examining the exact illumination patterns in the transduced tissue are unfortunately beyond the capability of techniques currently available. Instead, we are trying to tackle the problem first from a theoretical perspective.
The optical scattering and absorption properties of the brain and the rationale of choosing particular parameters for later computational method will be discussed in the following section.
3. 1. 1 Optical Absorption
Optical absorption by molecules (chromophores) typically occurs via two transitions: electronic and vibrational. Electronic transition molecules are able to absorb photons whose energy match with their π-orbital resonance frequency. An example of the electronic transition molecule is the retinal whose peak absorption wavelength is around
380 nm. Vibrational transitions typically occur in the infrared where a variety of the chemical bonds can resonantly vibrate or twist upon absorption of a photon. The conventional way of quantifying the strength of absorption is using an absorption cross section σa , which is defined as
28
( 3. 1 )
, in which Qa is the absorption efficiency and A is the geometrical area of the chromophore. Therefore, the absorption coefficient μa is defined as
( 3. 2 )
ρV is the volume density of the chromophore particles in the medium.
There are several major contributions in the absorption spectrum of biological tissue. In the ultraviolet, the absorption occurs mainly due to proteins and DNAs, which increases with shorter wavelength. In the red to near-infrared (~700 nm to 1000 nm), the absorption is minimum and often referred to as “diagnostic and therapeutic window”
[34]. In the infrared, the absorption occurs mainly due to tissue water content, which increases with longer wavelength. Figure 3.1 gives a summary of absorption spectra in biological tissue. The optical absorption spectrum of brain tissue is similar to that of other biological tissue. The most common chromophores in the brain research include hemoglobin and mitochondrial pigments [35], which play important roles in optical imaging [36].
The absorption coefficient per mole of hemoglobin, which is dominated by HbO2, at 470 nm is 32,000 cm-1M-1 [37]. But the volume fraction of blood vessels in gray mater is so small that the averaged absorption coefficient of brain tissue, which is 0.7 cm-1 at
470 nm and 1.4 cm-1 at 630 nm for human grey matter [38], is much smaller than that of hemoglobin in blood vessel. Note again that experimental data shows the absorption coefficient of brain tissue is typically much smaller than the scattering coefficient. Since light absorption properties of the light sensitive proteins play important role in the optogenetic neuromodulation, the absorption for both the light sensitive chromophores,
29 i.e. all-trans retinal and the co-labeled fluorescent proteins (e.g. GFP and EYFP) will be estimated as follows.
Figure 3.1 Optical absorption spectra of various tissue components in the UV, visible and infrared frequency range. The peak absorption wavelength of several commonly used opsins were indicated by arrows.
The number of ChR2 proteins per neuron is estimated based on the single channel conductance (50-250 fS) and the saturation photocurrent at voltage clamp in cortical and hippocampal neurons (1-3 nA at resting potential (-65mV) reported in reference [39, 40]).
Therefore, the total conductance (C) of a single neuron is
( 3. 3 )
, assuming the average distance between two adjacent neurons is 50 um. Therefore, the total number (N) of channels per neuron and the molar concentration (M) of the channels are estimated to be:
30
( 3. 4 )
( 3. 5 )
The maximal extinction coefficient, i.e. the probability that the all-trans retinal molecule will absorb a photon at blue, is around 50,000 cm-1M-1 [41]. The absorption coefficient of the retinal molecules in transduced tissue is
( 3. 6 )
The absorption cross section of the retinal molecule is thus
2 1 Å ( 3. 7 )
, which is small comparing with the physical area of the retinal molecule ( 30 Å2 [42]).
The absorption spectrum shifts when the all-trans retinal molecules were embedded in different rhodopsin constructs. Figure 3.2 shows the activation spectrum of two typical opsins, the ChR2 and NpHR.
Figure 3.2 The absorption spectrum (in arbitrary unit) of ChR2 and NpHR, adapted from Ref [8].
31
The volume density of the co-expressed fluorescence proteins (EYFP etc.) is the same as that of retinal molecules. The absorption coefficient of EYFP has been experimentally reported to be 84,000 cm-1M-1 at its maximum absorption wavelength
(514nm) [43]. By using the similar calculations as all-trans retinal, we can derive that:
( 3. 8 )
The absorption coefficient of the all-trans retinal and EYFP, as estimated in equation (3.6) and (3.8), are 4 orders of magnitude smaller than that of intact brain tissue
(0.7 cm-1 at 470 nm [38] ). Hence, we can legitimately claim that in the opsin-expressing brain tissue the fraction of photons absorbed by light sensitive channels is negligible in comparison to other native chromophores. We will use this assumption in Monte Carlo simulation of photon absorption and scatting in later section.
3. 1. 2 Optical Scattering
Brain tissue contains various structures including cell membrane, subcellular organelles and molecules. These structures have different geometrical sizes from 10 μm for cells to 10 nm for membranes. Light scattering occurs in medium which has fluctuations in the refractive index. Such fluctuation could be due to discrete particles in the media or continuous variations in n, the index of refraction. In order to quantitatively define the light scattering, it is important to describe the scattering parameters first. In
1908, Gustav Mie was the first one to publish the mathematical solutions to the optical absorption and scattering properties of particles by solving Maxwell’s equation, using multipole expansions of the electromagnetic (EM) waves [44]. The solutions were formalized mathematically later into a more common form [45]. In the simplified scattering model, the medium contains spherical particles. The incident light is redirected
32 by the particle and casts a shadow at the incident direction whose area is described as the scattering cross section (σs). Similarly, the scattering coefficient, µS, can be defined. The inverse of µS describes the mean free path of the photon in the medium before it gets scattered. The scattering coefficient of the intact brain tissue is measured to be 100 cm-1 at 470 nm [38], much larger than the absorption coefficient. Photons usually undergo multiple scattering before it is finally absorbed by the brain tissue. The distribution of scattered light is not necessarily isotropic. The preference of forward directed light after the scattering event can be described by the anisotropy factor, g. A reduced scattering coefficient, µS’, describes the mean free path in the incident direction.
3. 2 Monte Carlo Simulation of Light in Brain
Tissue
3. 2. 1 General Idea and Implementation of Monte
Carlo Method
Brain tissue is a complicated system from the point of view of optical interaction, being scattering, absorption, or birefringence (the change of light polarization). Without sufficient information about various cellular or sub-cellular structures, we cannot arrange the scattering medium in such a way that each scatterer has its definitive location and optical properties. Instead, our model generalizes the photon-tissue interaction using several parameters (absorption coefficient, scattering coefficient, anisotropic factor and etc.), which have been obtained from measurements (for example [38]).
33
The first major exploration of the Monte Carlo (MC) method was in the research work to develop the first atomic bomb in 1944 [46]. The scientists working on the
Manhattan Project had intractably difficult equations to solve in order to calculate the probability with which a neutron from one fissioning Uranium atom would cause another to fission. They solved the problem when realizing that they could follow the trajectories of individual neutrons, one at a time, using teams of humans implementing the calculation with mechanical calculators [47]. At each step, they could compute the probabilities that a neutron was absorbed or escaped from the bomb, so that one could recreate a virtual experiment to answer some of the important questions.
Although computationally intensive, Monte Carlo methods are widely applied in simulating photon transport in biomedical imaging and other applications [48]. The idea of MC is simple and straightforward, and the use of Monte Carlo in modern physics allows us to examine many complex systems. As a first step to introduce our approach to the problem, we present a flowchart (Figure 3.3) of how to implement the MC method in tracking the trajectory of a single photon in the scattering tissue. As discussed before, light is strongly scattered, placing limits on the activation volume of opsin-expressing brain tissue emanating from fiber optics. We first implemented this computational model to estimate the intensity at any location from various light sources.
We performed a Monte Carlo simulation that treats the brain tissue as heterogeneous scattering and absorptive medium. Accordingly, we employed an anisotropic scattering model based on the most commonly used Henyey-Greenstein (HG) phase function (Equation 3.9) [49],
34
( 3. 9 )
This function gives a model distribution of the direction change of scattered photons. It is an empirical approximation for Mie scattering from particles with a distribution of sizes.
It was first proposed to model the scattering of light from distant galaxies by dust.
Nevertheless, HG phase function is a widely accepted approximation and agrees well with experimental observation (e.g. The goniometric measurement of scattered light by
Figure 3.3 A flowchart of the Monte Carlo method used in tracking photons in the scattering media.
thin tissue samples. is the polar angle in the spherical coordinates. Conventionally,
expresses the probability of direction change falling into the range of
. The anisotropy factor indicates the spatial uniformity of the
35 scattering, with g = 0 being no preference to any direction and g = 1 being mostly forward scattering. We employed an anisotropy factor of g = 0.88 in our simulation, the scattering and absorption coefficients used were 100 cm-1 and 0.7 cm-1, respectively, for blue light as obtained from the diffuse reflectance and transmittance measurements of human brain tissue [38]. Therefore, the brain tissue is considered a continuum of scatterers and absorbers distributed uniformly.
In most fiber optics based light delivery system, light traveling down the fiber is confined until it is emitted through the tip. To simplify the model in our case, we started with treating the very tip of the optrode as a point light source with a given initial beam divergence, while placing it in the center of a cube with dimension of 2×2×2 mm3. The cubic region of interest (ROI) was divided into 200×200×200 volume differential elements (voxels), so that voxel spatial resolution (10 μm, approximately one tenth of the mean free path of a photon traveling in brain tissue) was sufficient to map the distribution of the light intensity. A packet of 106 photons was launched from the source, uniformly distributed according to a chosen angle of divergence. The maximum divergence angle chosen was 30 degrees to the z-axis of the fiber, acquired from a separate experimental observation we made for the emission profile of the optrode immersed and imaged in a fluorescent dye solution [50, 51].
The distance traveled before each photon interacts with the tissue is based on the random number and the local attenuation coefficients (absorption and scattering). A practical approach to implement the model is to drop the weight at end of each step as shown in Figure 3.4. The remaining non-absorbed photons are redirected again based on
36 random numbers and scattering parameters. The photon changes direction according to the HG scattering phase function.
Figure 3.4 Schematics of weight dropping and redirecting of a single photon during the scattering. W(n) is the photon weight at the nth step, s is the traveling distance at each step, and µa is the absorption coefficient.
By repeating the random process, the trajectories of all photons, their absorbed weights and the location of individual events are registered. The simulated result will approach the realistic scenario if one generates enough samples of the possible photon- tissue interactions. Light intensity distribution, which is immediate related to photon density distribution, can be obtained by knowing the number of absorbed photons at each spatial location.
One should notice that the random number generation is never truly random. Any random number sequence, generated by computaional algorithms, is in fact determintistic.
It is usually determined by the initial value, which is known as the ‘seed’. In our implementation of the ‘psueduo-random’ number generator, we use the system time of
Windows OS as the seed, which is believed to be ‘random’ enough to initialize the random number generator.
37
3. 2. 2 The simulation of light intensity and affected
volume
The simulation was implemented by homemade Matlab code (refer to the appendix). The intensity value at any spatial point was stored in a 3d matrix. The results from two types of commonly used fibers were compared in Figure 3.5 (a). We plot the color-coded contour lines in order to guide the eyes. The intensity of both cases, plot along the z-axis, were presented in Figure 3.5(b). Simulation shows that for the same light power output, light intensity of small core fiber is almost two orders higher near the optrode tip comparing with that of the 200 µm core fiber case. However, light beam loses its high intensity and directionality due to the strongly scattering and weakly absorption of the brain. At 500 µm away from the fiber ends, light intensity becomes comparable for both cases (Figure 3.5(a) and (b)).
What is the exact amount of neurons being affected is another important question asked frequently while researchers apply various stimulation paradigms. On one hand, expression specificity and transduction volume are well controlled during the viral delivery, and on the other hand, the stimulation volume is related to the geometry and power of light emitting from optics. We also compare the effective volumes based on previous simulation result (as shown in figure 3.5), assuming the threshold intensity for activating the light-sensitive channel is 1 mW/mm2[8]. Figure 3.6 shows that small fiber has larger effected volume than that of the 200 um fiber when output power is lower than
6 mW (the typical safe power range). For instance, for the same power of 2 mW, two types of fiber will effect a nearly spherical range with radius of 0.55 mm and 0.39 mm respectively. For power output of 1 mW, the radii will be 0.49 mm and 0.37 mm. It is
38 valid that, when one starts ramping the power, intensity surround the small fiber should pass the threshold before that of the larger one.
Figure 3.5 Examples of light intensity distribution in brain tissue obtained from the simulation (a) light intensity in the x–z plane. Left, the nearly point source case (e.g. tapered 10 µm core fiber or optrode); Right, the case of large core (200µm) multimode fiber. (b) Intensity plot along the z-axis.
39
Figure 3.6 Comparison of the stimluation volumes. It shows that the volume, obtained from MC simulation, is related to the light power as well as the fiber size.
3. 2. 3 The simulation of light induced heating effect
In conjunction with light absorption and local energy storage, there will be energy dissipation in the tissue. Two primary mechanisms of energy storage are encountered most frequently during light stimulation: sensible and latent. Sensible storage results in change of temperature and latent storage results in change in structure. The two processes may occur simultaneously, depending on the initial state of tissue and intensity of stimulation. Phase changes, such as protein denaturation, are often the direct source of damage to tissue. A local elevation of temperature will cause the diffusion of heat to surrounding areas. The diffusive distribution of heat depends on the spatial pattern of light absorption as well as the temporary stimulation scheme. Therefore, the analysis of
40 heat transport under various experiment conditions is an important aspect of understanding some of the side effect of light stimulation.
We begin with evaluating heat transport from the perspective of conservation of energy applied to the tissue-photon system. One can refer to any textbook (e.g. [35]) for theoretical treatment of light-tissue interaction and detailed formulation. For our purpose, we employed a simplified model starting with the energy conservation:
( 3. 10 )
, in which indicates the rate of heat change. Assuming the venous blood equilibrates with the tissue temperature, so that it can be combined into the heat conduction terms. Another assumption is that the metabolic heat generated by tissue itself is negligible.
If the phase transition effects are neglected, then the energy storage term is expressed by ( 3. 11 )
The heat diffusion occurs along the temperature gradients.
( 3. 12 )
Where T is the spatially dependent temperature function, and dA is the infinitesimal area normal to the temperature gradient. This expression (3.12) is known as the Fourier’s law of heat conduction. The relevant constitutive property is the thermal conductivity, k, which is the measure of how tissue facilitates the heat transport. It is related to chemical composition and molecular structures as may be manifested microscopically and macroscopically. For this reason, changes in temperature and pressure can cause significant alterations of heat transport properties.
41
The thermal conductivity is in the cerebral cortex of human
brain. It can be obtained via experiments in which heat transport is caused in specific materials under tightly controlled conditions [52].
The heat conduction within a microscopic dimension dx, dy and dz can also be expressed via Taylor expansion, with terms higher than the first order neglected.
( 3. 13 )
The individual conductions across the x, y and z dimensions are described by
Fourier’s law.
( 3. 14 )
, which are substituted into equation (3.13)
( 3. 15 )
Combining equations (3.10), (3.11) and (3.15)
( 3. 16 )
, in which light power being absorbed is , tissue mass density
. [52].
However, a complete solution to the partial differential equation (3.16) requires the specification of the spatial-temporal configuration of the source, as well as boundary conditions. The treatment here is that the tissue can be considered as a 2x2x2mm cube of
42 uniform scattering/absorbing medium regardless of any details of the cellular or cortical structure.
The first step is to simulate a simplest case – what is the spatial-temporal behavior of thermal response to a light impulse from a point source, i.e. Pabs is a delta function in space and time. Since the system is in the linear response regime, solutions to any source configuration or stimulation protocol could, in principle, be derived from the impulse response function (see equations (3.17) and (3.18) below).
( 3. 17 )
G(r, t) is the Green’s function of equation (3.16), which is defined as
( 3. 18 )
An example of implementing such ‘multi-dimensional’ convolution is given in the appendix as well. Figure 3.7 illustrates the temperature increases generated by a 1 mW blue laser pulse of 1 ms duration (in log scale for better visualization) emanating from fiber of 10 μm and 200 μm core diameter. The temperature changes in axial direction
(along the dashed line) for both cases are also shown at the bottom. The peak temperature increase generated in this condition is about 0.04 °C for 10 μm fiber, while it increases only 0.0007°C for 200 μm fiber. The plots with decreasing peak values correspond to the temperature profiles at each 1 ms time step in the following 50 ms after the initial pulse.
This data indicates a fast decay time by heat conduction for the 10 μm fiber (Figure 3.7
(b) left ), where 20 ms after optical pulse delivery the peak temperature increase in the tissue drops to less than 0.005 °C. Therefore, we expect relatively mild temperature
43 increases in response to the typical optical stimulation paradigms in current use. For example, for 1 mW and 1 ms pulses delivered at 100 Hz, the peak temperature increase is
0.05 °C.
However, the continuous stimulation is an extreme situation that one should be cautious with the amount of light power. We then plot in Figure 3.8 the peak temperature
(indicated by arrows in Figure 3.7) change as the function of time during the continuous stimulation. For the 10 μm fiber, we find that the maximum temperature increase reaches
0.2 °C after 200 ms of 1 mW light. The magnitude of temperature increase might not have damaging effects on the tissue in sub-acute experiments. For the 200 μm fiber case, it is shown that 200 ms continuous pulse would generate a peak temperature increase of
0.05 °C. We always performed the simulations at a light power level of 1 mW.
Theoretically, the temperature change linearly relates to the power intensity. Therefore, the temperature changes at other power levels can be calculated by using this simple relationship. Accordingly, if we assume a 1 °C rise in temperature is acceptable for brain tissue as regulated by FDA on RF energy deposition in MRI application[53], we are able to deliver up to 20 mW light power from a 200 μm core fiber and up to 5 mW from a 10
μm core fiber under this continuous condition.
44
Figure 3.7 Spatial and temporal resolved distribution of temperature change. (a) The spatial distribution of temperature increase in brain tissue in response to a 1 ms light pulse delivered through a 10 μm (left) and 200 μm fiber (right). The fiber is at the center of the heat map. (b) The temperature profile along the z-axial direction (dashed lines in (a)) at different elapsed time after the 1 ms pulse stimulation.
45
Figure 3.8 The peak temperature (as indicated by arrows in Figure 3.7(b) ) evolves in response to a 200 ms continuous stimulation (green). It shows that 10 μm fiber generates faster rising temperature (blue curve) than that from the 200 μm fiber (red curve). The output power is kept constant at 1mW for both cases.
We have used different temporal patterns to better exam the pulse sequencing that will be employed in real experiments. The temperature in rodent brain varies naturally over a range of several degrees, due to circadian rhythm, exercise, and environmental variation [54, 55]. In general, the expected heating from light absorption under typical experimental conditions is much less than this range according to the simulation results.
However, when investigators resort to other light sources (e.g. the cortical surface mounted LEDs [56]) with additional heat contribute from the device itself, heating effect will become relevant, and hence we should re-emphasize the need for opsin-negative controls and the importance of heating control in future [57].
46
3. 3 Summary
In this chapter, we have focused on the light delivery and the effect of optical power in the optogenetically transduced brain, defining a broad guideline for safe levels of laser-based optical excitation. This issue is increasingly important to the community and beyond given the transition of multiple experiments from the acute to chronic regime in both rodents and non-human primates. We have demonstrated a Monte Carlo-based computer simulation of the light intensity and heat distribution generated by light within the brain tissue. We compared the result from two types of fiber optical tools: tapered optrodes, elaborated in previous chapters, and large area blunt fibers used by other laboratories. This work, which is complementary to the optical stimulation- electrophysiology experiments in rodent and monkey to test the effects of light on neural responses, is our first step in developing effective and safe stimulation protocols.
47
CHAPTER 4 RECORDING
FUNCTIONALITY OF THE COAXIAL
OPTRODE
To investigate the recording capability of various designs of optrodes, we construct a model that describes the electric behavior of optrode in brain tissue which allows its recording properties to be rigorously analyzed.
The following chapter will start with characterizing the nature of bioelectric sources and conductors. This discussion points out that, in contrast to electronic circuits, in which the elements are isolated and concentrated, the brain consists of distributed volume sources and volume conductors. The volume sources are neurons that provide time-dependent transmembrane currents generated during action potentials. On the recording side, the recorded voltage signal depends not only on the electrode location relative to the firing neuron, but also on the geometry and electrical properties of the
48 active recording area. It is shown that a reciprocity theorem applies to this source-probe problem. Based on this theorem, we will discuss the optrode-tissue model, where our optrode is modeled as a conductive ring (or partial ring) with an insulated shaft in a conducting medium. An important characterization of the recording capability is the sensitivity field, which is a measure of how easily an optrode can record neural activity and how well it can spatially discreminate the origin of activity.We then implement a finite element (FE) method to simulate the sensitivity field of recording, which varies among different types of optrodes. This model analysis shall provide an insightful way of estimating and interpreting measured signals by different optrode configurations.
4. 1 Electromagnetic field generated by the
neuronal activity.
In electrophysiology, the resistive components (cytoplasm and extracellular medium), capacitive components (neuron membrane), and batteries (active membrane events) are not discrete but distributed. That is, the conducting medium extends continuously in three-dimensional space and is referred to as a volume conductor [58].
When a neuron fires, it undergoes an increase in conductivity over the excitable regions of its membrane, usually at the axon hillock and/or soma. The current that flows into the membrane will flow along the cell, and exit at various regions of the inactive membrane to return to the initial current by way of numerous diverse pathways though the extracellular medium. The flow of current across the resistive medium generates a complex time-dependent potential field around the neuron. The properties of this field depend on the geometry of the cell, and the location and time-course of the membrane
49 conductance change. The extracellular field refers to the complex field around a discharging neuron, and the EM field in the volume conductor is governed by Maxwell equations:
( 4. 1 )
Where E is the electric field, is the displacement field, and is the permittivity of the medium. is the volume charge density, B is the magnetic field, and J is the current density.
Assuming the volume conductor is homogeneous and infinite, the total current density derived by the volume source is given by [59]:
( 4. 2 )
The quantity is referred to be the return current, introduced by the primary source – the impressed current [60]. The return current is necessary to avoid building up of charges due to the source. is the membrane current due to the channel opening
(resistive) or membrane potential change (capacitative). The individual segment of neuron behaves as an electric current dipoles during action potential. Hence the overall impressed current is equivalent to a current dipole. The primary source establishes an electric field . The potential field is linked to electric field by:
( 4. 3 )
50
The scalar and vector potentials ( ) can be defined unambiguously by choosing the Lorentz gauge:
( 4. 4 )
By taking the divergence of equation (4.2), and combing equation (4.1), (4.3) and
(4.4), the scalar potential can be calculated by solving the following
( 4. 5 )
Since the capacitive component of tissue impedance is negligible in the frequency band of neuronal events (1Hz-10kHz), according to experimental evidence[61], this implies that the time-varying electric activities in the brain tissue can be examined in the conventional quasi-static limit. Equation (4.5) is therefore reduced to
Poisson’s equation:
( 4. 6 )
The solution for potential has an integral form:
( 4. 7 )
The element behaves like a point charge, in that it sets up a field, that
varies as . Because we seek the solution for the field outside the volume source, equation
(4.7) can be transformed to:
( 4. 8 )
51
, using the identity: , assuming vanishes at
boundaries infinitely far from source. The source element in equation (4.8) behaves like a
dipole element, with a field that varies as the dot product with .
In general, for arbitrary shaped active cell, the field generated by the distributed membrane current can be rigorous formulated by equation (4.8). However, there are two simple types of current source, namely monopole or dipole. Here briefly describes the fields generated by monopole or dipole current source[58].
Monopole source: If we consider a point current source of magnitude lying in a uniform conducting medium of conductivity , the impressed current flow has the form:
( 4. 9 )
And accordingly, the potential field must have a spherical symmetry.
( 4. 10 )
Dipole source: Bioelectric events can never have a monopole current source because of charge conservation. The dipole consists of two monopole of opposite polarity
(often termed source and sink) separated by distance, d. In fact, the strict definition requires being finite[58]. Consequently, the field arising from dipole current source can be obtained from the monopole field. That is approximated by first order term of Taylor expansion with respect to spatial coordinates.
( 4. 11 )
is the dipole moment and is the unit vector in radial direction. If the coordinate is oriented so that the dipole is along the z axis and placed in the origin of a
52 cylindrical coordinate, assuming the recording electrode is placed at a distance of r and an inclination angle of :
( 4. 12 )
Definition of the lead vector[58]: The measurement of extracellular potential signals resulting from the neuronal activity is usually done by metal probes, such as single unit recording electrodes or EEG electrodes. Suppose that at electrode location i the potential due to a current dipole can be simplified as (equation 4.11)
( 4. 13 )
Then the potential difference between any two points, e.g. i and j (representing the recording electrode and reference).
( 4. 14 )
is defined as the lead vector, which implicately includes structure about conductive properties of tissue, current source distribution, electrode/reference geometry and their spatial configuration. Figure 4.1 illustructes the lead vector.
Figure 4.1 The schematic of potential difference measurement. A dipole current element p contributes to the potential difference by the product of the dipole moment and lead vector at the exact location. 53
An important assumption made for above discussion is that the potential field should not be significantly disturbed by the presence of recording electrodes, which is generally true for single-unit recording electrodes. This is also one of the reasons why we appreciate those electrodes plated with noble metals. For instance, the measured exchange-current density of gold is about 10-9 A/cm2 , which for a recording area of 100
µm2 would represent a current of only 10-15 Ampres [26, 29]. It is the same order of magnitude compared to that of tungsten metal. The exchange current density is a measure of current flow across the metal-electrolyte interface at certain electrode potential. It reflects the intrinsic rate of carrier transfer between electrolyte and the metal electrode.
4. 2 Sensitivity field of electrode.
As noted before that one can assign to each possible location of point source a lead vector. In this way we establish a lead vector field which is distributed throughout the entire volume conductor. It’s necessary to introduce the concept of sensitivity field
[58], an extension of the concept of lead vector. Because the lead vector indicates the sensitivity of the recording potential by , the distribution of (as a function of the location and orientation of ) defines the sensitivity field to the source . An isosensitivity surface is a surface where the magnitude of is constant. An important evaluation of the optrode recording capability is the sensitivity field, which tells how easily an optrode can record and how well it can spatially discreminate the origin of activity. Other evaluations of the performance of optrode, such as impedance and optical throughput, will be discussed in other sections of this thesis.
54
4. 2. 1 Reciprocity theorem of Helmholtz
It is not practical to directly solve the sensitivity field, without information about source density distribution and electrode/reference geometry. However, an alternative reciprocity-based approach shows that the distribution of the actual sensitivity field of electrode may be obtained indirectly without using the lead vector concept. In other words, to determine the recorded potential of certain electrode configuration, we could generate a reciprocal potential field, as opposed to the sensitivity field, by feeding a virtual current through the recording electrode. As shown later in this section, the contribution from each volume source equals the dot product of the reciprocal current density and the source element. Such explanation was based on Helmholtz’s principle of duality (1853)[62]. Following section gives the mathematical proof of the reciprocity theorem as described by Plonsey (1963)[63].
4. 2. 2 Proof of the Reciprocity theorem
If Φ1 and Φ2 are any two scalar fields in arbitrary volume v, the following vector identities must be satisfied:
( 4. 15 )
Subtract the second equation from the first one, integrate over the volume v, and use the divergence theorem. We obtain
( 4. 16 )
55
We assume that Φ1 is the scalar potential due to a current source , i.e. the impressed membrane current generated by neurons during action potentials. We assume
further that Φ2 is the potential produced by a current flow , which is the reciprocal current generated by feeding a virtual current through the recording site, i.e.
( 4. 17 )
Usually an electrode has much higher conductivity compared with that of neural tissue, so that the direction of is always normal to the surface of recording site. The scalar potential is then identified as the reciprocal potential Any enclosure in the volume has
( 4. 18 )
The reciprocal current must be solenoidal in the volume conductor, hence
.
Therefore, ( 4. 19 )
Assume there is negligable current crossing the boundary, i.e.,
( 4. 20 )
We can rewrite Equation (4.16) by equations (4.17), (4.18), (4.19), and (4.20).
( 4. 21 )
A simple form of reciprocity theorem can be derived from equation (4.21) in the following way. Consider that the reciprocal potential arises from an inflow of unit current at point a and b (e.g. a point recording electrode and reference electrode) on the boundary shown in figure 4.2. Mathematically:
( 4. 22 )
56
Similarly, consider the unit current (a current dipole) consists of a point source and point sink at and , hence:
( 4. 23 )
Substituting equation (4.22) and (4.23) into equation (4.21), we obtain:
( 4. 24 )
Figure 4.2 The schematics for deriving the reciprocity theorem (figure reproduced from [58]).
Which implies that the potential difference between two arbitrary boundary point a and b due to a unit dipole current source, supplied internally between points 1 and 2, equals the potential difference between point 1 and 2 due to a unit current applied externally (reciprocally) between points a and b. This is the simplest form of the reciprocity theorem.
The recorded potential signal is essentially:
( 4. 25 )
57
It could be generalized to the case of electrode with arbitrary geometry. By imposing an uniformly distributed and unit reciprocal current on the electrode-electrolyte interface, we can spatially average the field and rewrite equation (4.21) into:
( 4. 26 )
Therefore,
( 4. 27 )
Note that above essentially describes the sensitivity field as expressed in
, an integral form of equation (4.14). Since no assumption has been made concerning the geometry of volume conductor, we can use the method for quantitatively evaluating the sensitivity field of arbitrary electrode configuration. If the volume conductor is homogeneous, the conductivity is spatially independent. i.e.
( 4. 28 )
Equation (4.28) simply implies that the recorded signal is proportional to the integral of dot product of reciprocal field and source current density.
4. 3 Modeling the recording sensitivity field of
optrode using finite element (FE) method
As described in previous section, the recording sensitivity field of the recording tools can be essentially mapped out by solving its reciprocal electric field. Here we establish a computational model to simulate this property by using the FE method.
58
The FE method is a numerical technique for finding approximate solutions to
PDEs [64], for instance the Poisson’s equation in our case, of which analytical solutions only exist in few lucky situations. This section describes how the FE approximates a PDE with a problem that has finite number of unknown parameters, that is the discretization of the original problem. In preparation, a simple illustration of how FE works is described.
Consider a potential function in 1-d, the Poisson equation is
( 4. 29 )
Aussming a set of functions ( the basis functions vi) defined on the same domain, i.e. region of interest of the model, that an arbiarty function can be represented by this basis
( 4. 30 )
For integrable real-valued functions, the exact form of inner product of on the interval [a,b] can be
( 4. 31 )
Other forms of inner product also can be defined (please refer to a textbook on
Functional Analysis). Rewrite PDE (4.29) as
( 4. 32 )
The second part in above equation could be proven using integration by part.
Define the matrices
( 4. 33 )
59
Combing equation (4.30) and (4.32), and using the defination (4.33), we can convert the PDE (4.29) into
( 4. 34 )
, in which and are vector represntation on the basis space. Therefore, this PDE becomes an eigenvalue problem, which we could solve numerically. Since the domain is represented by a discrete mesh, piecewise functions related to the meshing are commonly choosen as the basis (Figure 4.3). Although we consider 1-d function space here, this analysis extends to 2-d, 3-d and higher dimensional funciton space (Hilbert spaces) that allows the distance and inner product of two functions to be measured.
Figure 4.3 A function (red curve) is approximately expressed as the linear combination of the basis functions. In this example, the basis are piecewise functions (blue curves).
We build a realistic model of the optrode using a commercial FE software
(COMSOL Multiphysics). The physics problem we are interested is to solve, as discussed before, the reciprocal potential field generated by a virtual current fed into the conducting medium through the optrode. We use the Electric Currents interface, under the AC/DC module in the COMSOL. Current conservation is the main feature. It defines a Poisson’s
60 equation, boundary conditions, and current sources for modeling steady electric currents, and it provides options for defining the associated material properties. It also provides many physics quantities by solving the electric potential.
The geometry of our model is shown in Figure 4.4. All objects are defined in a cylindrical conducting medium with a 1 mm diameter. The model optrode, consisting of cone and cylinder objects, has fiber core size of 10 µm and cladding size of 125 µm
(Figure 4.4(b)). A conductive ring (exposed gold coating) is defined on the tapered tip
Figure 4.4(c)). The ring is acturally approximated by the outside wall of a slab of the cone. Its small circular base has a radius of 5.2 µm (which matches the size of 10 µm core fiber) and height of 5 µm. Materials was defined for each component of the geometry. Individual material properties (refractive index, electric conductivity and permittivity) and boundary conditions can be assigned accordingly (Figure 4.5).
61
Figure 4.4 The geometry of the optrode modeled by COMSOL. (a) The whole model cosists of a cylinder (diameter of 1mm and height of 1mm) filled with conductive media and an optrode on the center axis. (b) A close view of the tapered optrode tip. (c) a sideview (top) and bottomview (bottom) of the defined coating area.
62
Figure 4.5 The materials assigned for each domain / boundary of the model.
The reciprocal current density field is essentially the same mapping as sensitivity field. As demonstrated in previous section, since the recording signal is related to this
current field by equation (4.28), i.e. , one can estimate the recording
performance of optrode before any measurement. Figure 4.6 illustrates the simulated reciprocal current density distribution for the optrode with a conical ring-shaped coating.
Red lines (in 3d) indicate the flow direction of the reciprocal current field. The color image shows the magnitude of the current density in one slice (yz plane, along the optrode axis). The value is in logarithm scale to guide the eyes.
63
Figure 4.6 Simulating the reciprocal current density distribution (the sensitivity field). The color map (in log10 scale) is a 2d section of the density at yz-plane. Red lines are the current flow lines viewed in the 3d perspective.
Experimentally, we observed that the signal was improved by the optrodes with half side coating. In order to understand the relation between geometry of coating and its recording quality, we compare the simulation result of four types of optrode configurations: the ring shaped (The conductive coating area is acturally approximated by the outside wall of a slab of the cone. Its small circular base has a radius of 5.2 µm, which matches the size of 10 µm core fiber, and slab height of 5 µm as shown in figure
4.4(c). It corresponds to an exposed area of 240.8 µm2), half ring shaped (a half of the previously defined ring coating), quarter ring (a quarter area of the previously defined
64 ring), shaped and a cone shaped coating (conventional conical tip electrode with a base of dia = 4.6 µm and height of 4 µm, which has the metal exposed area of 33.5 µm2 [26] ).
The total reciprocal current is identical in all cases. The coating materials and, therefore, surface current distribution are different (equation 4.21 and 4.26) depending on the specific optrode design.
Regardless the exact form of generated by specific type or shape of neuron, we will discuss, as a first step, the simplified case as follows. A neuron initates an action potential at its axon hollock, which can be viewed as a point current source
. According to equation (4.26), i.e. , when such activated neuron is placed at location , the recorded potential signal should be proportional to the reciprocal potential at the activated region . Based on this observation, we obtain the relative amptitude of recorded signal for four types of probes, showing in figure 4.7.
As a straightforward way of illustrating the recorded signal, we plot the onto a xy-plane, the same plane where the ring seats. The result is summarized in figure 4.7.
Each dot represents a node (in fact, the node closest to the section plane) in the tetrahedral mesh. The color coded value indicates the amptitude of signal, if an activated neuron was place at this node. It shows that, for neuron firing at around 50 µm distance away from the recording ring, different optrodes would detect signal with similar amplitude. However, when the neuron approaches exposed coating, the recording signal will be significantly different for optrode with different coatings.
65
Figure 4.7 The relative signal amptitude of four types of electrode geometry (see text), obtained from the FE modeling.
66
Figure 4.8 Estimated signal amplitude as a function of the recording area. Four types of geometry are simulated: ring shaped (240.8 µm2), half-ring shaped (120.4 µm2), quarter-ring shaped (60.2 µm2), and conical electrodes (33.5 µm2).
Based on the numerical simulation result, we can also obtain the relationship between signal amplitude and the size of recording area (shown in figure 4.8). For instance, the maximum potential recorded from the metal electrode is expected to be almost four times of that from the ring coated optrode, comparing their recording area which is around 7 times difference. However, the point source approximation becomes questionable when it approaches the recording area. One should take into account the morphology of particular type of neurons, and it also requires certain knowledge of their membrane current distribution during action potentials [26]. In a coupled neuron-FEM model, the authors described the solution as a matrix equation [65]. A non-linear cable model of a layer V pyramidal cell was represented as a set of point currents (531 transmembrane segments comupted in NEURON) at the appropriate spatial location in 67 the FE electric field model. The recorded potential at different time instances is computed by changing the scale factors corresponding to each point source. Mathematically, we can formulate as
( 4. 35 )
, in which is the potential varying in time, is essential the sensitivity field of electrode, and is not exact membrane current, instead it contains the scaling factors of currents in space (at various compartments of a neuron) and time (during the course of action potential). J is simulated directly in NEURON and its polarity may switch from outward to inward (or vise versa) during an action potential. K is derived from the FE model. K is obtained by reciprocity approach because direct application of numerious point currents to FE model is not practical.
By using the neuron-FEM model, the authors found that small (<1000 µm2) and large
(10k µm2) silicon electrode contacts had similar volumes of recording sensitivity, but small contacts exhibited higher signal amplitudes (~50%) when neurons were in close proximity (50 µm) to the electrode[65]. We might expect similar effect that the contact area will not change the recording amplitude too much, if the size of model neuron is much larger than the contact size. However, I am expecting that the sensitivity of recording varies among different optrodes regardless of the neuron structure. Therefore, we could easily isolate single unit activity from background activity. The optimal design of the optrode is application-dependent. How the optrode contact size affects its recording properties will be exploited in the design process in the future.
68
4. 4 Discussion
Computational models are useful tools for understanding processes and refining methodologies, but they cannot exactly represent experimental conditions. In this FE model study, limitations and assumptions must be considered during interpretation of the results. Firstly, we used a static, resistive and linear in the context of electrical recording.
We did not include the double-layer capacitive interface at the electrode-electrolyte interface. Secondly, we assumed a uniform volume conductor while, in realty, the local electrical inhomogeneity around the electrode can substantially influence the recordings.
Finally, the effects of noise are not included in this initial study. Noise that is dependent on exposing contact will be particularly important. A thorough study should incorporate models of physiological and non-physiological noise into existing model.
Despite those limitations, this model analysis shall provide foundation for refining optrode design for the simultaneously optical stimulation and electrical recording applications.
69
CHAPTER 5 NEURONAL CORRELATES
OF SPATIAL INFORMATION RECORDED
USING THE MEA DEVICES
In this chapter, we introduce the multi-channel recording device - Utah Micro- electrode Array (MEA). We demonstrate preliminary data of neuronal correlates of spatial representation recorded by using the MEA devices. We will discribe the training of rats to perform a sustained attention task, and electrophysiological recording from the posterior parietal cortex (PPC) that demonstrate correlates between neuronal activity and animal behavior. This finding provides the neuroscientific ground for our next step of developing an integrated optrode-MEA device for optical neuromodulation and potentially behavioral modulation.
70
5. 1 Introduction
The ability of an animal to accurately navigate from one place to another requires integration of multiple sources of information, including the awareness of one’s position and direction, relations to landmarks, movement velocity, and the perceived location of the goal. Animals appear to use two basic and complementary processes for localization and orientation. Path integration is the process by which current position is estimated by performing an integration of movement velocity over time since the last known position
[66-68]. During locomotion, the movement velocity can be estimated using cues such as vestibular signals, visual cues or motor efferent copy. While such cues for path integration are readily available, any errors during the integration process will tend to accumulate, leading to inaccuracies. The second method of estimating one’s position is the landmark navigation, which relies upon the presence of stable landmark cues in the environment [67]. Perhaps the best example in the laboratory is the Morris water maze
[69], where animal is placed in a water pool at random position and must learn to locate the position of a hidden platform based on the visual cues in the room. While the landmark navigation appears to be more accurate than the path integration, the availability of familiar landmarks may be limited. So a compromise strategy is to use the landmarks to accurately locate one’s position when they are available, and use the path integration to fill the gap in the absence of landmarks.
Researchers have examined the role that the posterior parietal cortex (PPC) plays in relating the sense of spatial information with a navigational goal and in formulating a plan to attain that goal [68]. In early electrophysiological experiments, evidence from non-human primates suggested that the PPC executes a ‘matching function’ between
71 sensory input and the goal state, and also performs the function of shifting attention from one stimulus to another [70] [71]. It was later considered the visual associated cortex
[72], and included as part of the dorsal stream important for processing the ‘where’ aspect of visual perception. Human patients with PPC brain injury show deficits e.g. unilateral neglect and errors in reaching visual targets, which seems a combination of both sensory and motor deficits [73].
Nakamura recorded from rat PPC using a task similar to those used for monkey recordings [74]. Specifically, the head of the animals were immobilized and the animals were required to remember the spatial location of stimuli in a delayed-non-matching-to- sample task. In this paradigm, nearly half of PPC cells were spatially tuned, i.e., they responded preferentially to a particular location. Furthermore, most of the spatially tuned cells exhibited a ‘memory’ response, where they maintained an elevated rate of activity during the delay between the sample and matching phases. When the animal was rotated, the cells maintained the same directional preference relative to the arena, suggesting that the spatial stimuli encoded was using allocentric (landmark centered) frame of reference, while this finding of allocentricity in rat PPC is in contrast to the typically egocentric responses seen in primate PPC [75, 76]. Researchers did electrophysiological studies in freely behaving rats have also found cells that respond to navigationally important task features [77, 78]. McNaughton and colleagues have recorded cells in the PPC showing preferential activity for specific responses on a radial arm maze [77]. Some cells became active during movement towards or away from the center, others during turns or motionless. Using more elaborate task, Nitz found that activity of most PPC cells could not be explained by single variable such as direction or location [79]. Rather, activity was
72 coupled to the progression along the route. As an example, a cell that fired at the outset of outward movement would also fire at the outset of return trip. Nitz suggested that these
PPC cells encoded the order of both individual and multiple navigational epochs in a route and served as one of the neural substrates for navigation.
Early attempts to characterize the rat PPC were mainly based on cytocharchitecture and there was much disagreement as to the extent and location of the parietal areas (see
[80] for a review). This lack of clarity has led to the methodological difference among researchers using a lesion-based approach. In localizing their lesion sites, most have used similar coordinates in the approximately 1.5-5mm lateral to the midline and 2 to 6mm posterior to the Bregma. Despite the differences in defining the PPC area, it is generally consistent finding that rats with lesions of this area showed impaired at allocentric navigation.
The connections of PPC with other cortical sensory and motor areas along with limbic areas provide further evidence that PPC may play an important role in navigational behavior. Figure 5.1 presents a diagram of these connections [68]. PPC shares connections with primary and secondary visual cortex, somatosensory cortex, frontal association and retrosplenial cortex. While the connections with sensory and motor areas provide the navigational information reaching PPC, the connection between
PPC and retrosplendial cortex could act as a conduit for information transfer between cortical and subcortical structures. Retrosplenial cortex projects to postsuniculum, an area where head direction (HD) cells were first found.
73
Figure 5.1 Schematic diagram of connections between PPC and other cortical/subcortical areas. Arrows represent direction of information flow. Areas where grid, place and HD cells have been found are indicated by the legend. (Figure adapted from [68] )
Retrosplenial cortex also projects to entorhinal cortex, which contains grid cells.
Grid cells fire in grid-like pattern based on the position of the animal and have been conceptualized as a cellular constituent of the path integrator. The entorhinal cortex is a primary input to the hippocampal complex where place cells were initially discovered by
O’keefe and Dostrovsky in 1971 [81]. To summarize, the extensive connections of PPC
74 with other cortical areas provides the basis of various spatial correlates, it also explains the integral role of PPC in spatial processing.
In an effort to better understand PPC functioning in spatial attention and navigation in the neural substrate. we have adapted the Delayed-NonMatching-To-Position
(DNMTP) task. The behavior chamber was a Floor Projection Maze and a Y-shaped
Plexiglas walls [82]. Figure 5.2 shows an animal is performing the task in the maze. The maze was interfaced with three computers, one for tracking, one for data acquisition, and one for behavioral control. Tracking was accomplished with a Cineplex Digital Video
Recording and Tracking system (Plexon Inc., Dallas TX). Custom Matlab programs
(MathWorks Inc., Natick MA) translated the coordinates into behaviorally relevant position and relayed this information to a computer dedicated to behavioral control using
Med-PC IV and Med Associate hardware (Med Associate Inc.)
Figure 5.2 An animal (M6) is performing the DNMTP task in a floor projection maze.
75
5. 2 Behavioral Training
All procedures were conducted in accordance with national Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee at Brown
University. Subjects were three male Long-Evans rats (M6, M7 and M8, 6-10 months), mildly food-deprived to 85% of free-feeding body weight and trained to perform the
DNMTP task in the Y-shaped maze. Figure 5.3 is the diagram of behavioral events in the
DNMTP. Specifically, before the task training, the animals were pre-trained to hold at the fixation location and wait for the visual stimulus. Each trial of the DNMTP task consisted of two phases. In the sample phase, animals were directed to one arm with stimulus. After a delay period, the animals have to choose between two identical stimuli (choice phase).
To obtain the reward, they were required to enter the goal arm not visited (the non- matching one) during the sample phase. The chocolate milk reward was located at the south pole of the maze. The reward was also signaled by a tone (frequency = 1.5 kHz) when animal made the correct choice. When the animals failed to reach the correct choice, the result would be a brief period of blank floor and no reward was delivered.
This task requires animal to hold spatial attention (left or right arms) over variable delay intervals [83]. After two days of habituation to the maze and seven days of shaping, animals were given several training sessions of 30 trials daily until they reached criterion, defined as three consecutive days during which at least 70% are correct for trails with delay <10 sec.
76
Figure 5.3 Left panel shows the layout of the Y-shaped maze. Right panel shows timeline of a correct trial and an error trial. The fixation signal triggers the present of sample and choice stimuli. The delay time is defined as the interval between end of sample present and onset of choice stimli.
5. 3 Surgery
The implantation procedures were similar to that of the optrode-MEA device, which will be decribed in next chapter in details. Briefly, the rats were anaesthetized with 2% isofluorane-oxygen mixture, placed in a stereotaxic apparatus. The craniotomy was centered at the following coordinates: 4.5mm P to bregma and 4.5mm L to the midline. In our experience, the quality of the surgical procedure plays a decisive role the outcome of recordings. By carefully drilling a 3x3 mm2 craniotomy window and peeling off the duramater, the PPC was exposed. The devices targeted at right posterior parietal cortex
(as shown in Figure 5.4) Prior to implantation, the devices were sterilized using ultraviolet light to minimize the risk of infection. A series of holes were drilled for placement of 6 skull screws which provide mechanical support and common ground
77 reference for electro-physiological recordings. The MEA was slowly driven (100
μm/min) into the cortex by a micropositioner, without causing major tissue depression.
Once the device reached the desired layer of cortex, the craniotomy was sealed with silicone elastomer (World Precision Instruments) and the entire device was fixed in place using dental cement. After they had recovered from surgery for a week, we put the animals back on a restricted diet and training schedule.
5. 4 Data Acquisition
The behavioral data presented here were recorded from eleven sessions from one animal
(M6), and subsequent analysis was using Matlab. Real-time neural activity was recorded and processed by using Multineuron Acquisition Processor (Plexon Inc, Dallas TX), a parallel processor capable of digitizing up to 32 channels simultaneously at 40 kHz rate.
Voltage-time threshold and waveform template algorithm were used to identify single units [84]. Using these criteria, we routinely record from the same units throughout each
2 hr session. In off-line analysis, manual clustering of waveforms were further discriminated in principle components space (Offline Sorter, Plexon Inc.). Figure 5.4 left shows spike waveforms of 19 isolated units geographically mapping on the MEA channels in a typical recording session. Time-stamp records of spikes, stimulus onset and behavioral events were saved digitally, as well as all sampled spike waveforms. Local field potential (LFP) signals of selected channels were band pass filtered between 0.7 and
300 Hz and digitized at 1 kHz sampling rate. Subsequent analyses employed custom
Matlab programs and NeuroExplorer (Nex software).
78
Figure 5.4 Schematic of the implanted MEA in PPC. Left panel shows a 6x6 element MEA and its physical size related to the cortex (image was adapted from [85] and [86]. PTLp indicates the Posterior Parietal area). Right panel shows the spike waveforms from individual electrode in one recording session. 5. 5 Result and Discussion
Figure 5.5(a)shows the delay dependent performance within one day. Top panel of
Figure 5.5(a) is the number of trials and bottom of Figure 5.5(a) is the correct rate distributed at each delay periods, which shows that the performance dropped to the 50% chance level when the delay was larger than 25 seconds. Figure 5.5(b) shows the average performance curve over eleven recording sessions. Delay-dependent performance was exhibited for delays of 1-30 sec (decrease from 0.76 to 0.55, P<0.01).
79
(a)
(b)
Figure 5.5 Delay dependent performance. (a) top graph shows the number of trials distributed among various the delay periods. Bottom graph shows the averaged rate of correct trials within one day recording. (b) The average performance curve over eleven recording sessions and the standard deviations (SDs) indicated by vertical bars. Trails were grouped and plotted by 5 sec intervals across the 1-35 sec delay periods.
80
(a)
(b)
81
Figure 5.6 (a) Rastergrams and peri-evnet histograms of two “encoding” cells. Left column shows the spiking ativity of cell 1 and 2 responding to the left-correct choice. Right colum shows the response to the right-correct choice. Trials were aligned to the event timestamps of choice stimulus present (green ticks). (b) Spiking activity of the cells ( the same cells as shown in (a) ) mapped on the floor maze during the delay period.
Figure 5.6(a) shows the rastergrams and perievent histograms of two neurons during the task. Each neuron was identified with respect to its functional type that encodes different task features [87]. For example, cell #1 increased its firing just before the appropriate stimulus (left non-match choice) presented, while cell #2 shows response to the right non-match choice event only. We also plotted their firing activity according to the location of the animal during the delay periods. Figure 5.6(b) demonstrated the spatial firing activity of each cell at different choice events. The dark blue region indicated the location that animal had visited during the delay and the averaged spiking rate was color coded on the map. Since the recording area PPC has strong connection with retrosplenial cortex and other subcortical structures (see Figure 5.1), we speculate that the event encoded feature of individual cell is correlated with the head direction of the animal while preparing for the preferred choice. We propose that one could modulate the behavioral outcome by modulating the activity of such event encoded neurons optogenetically. Hence, the behavioral and physiological correlated data from bare MEAs motivated us furthermore for innovating the optrode-MEA version of the devices in the genetically modified animals, which will be elaborated in the next chapter.
82
CHAPTER 6 INTEGRATED DEVICE FOR
COMBINED OPTICAL NEURO-
MODULATION AND ELECTRICAL
RECORDING FOR CHRONIC IN VIVO
APPLICATIONS
6. 1 Introduction
Studying brain function and its local circuit dynamics requires neural interfaces that can record and stimulate the brain with high spatiotemporal resolution. Optogenetics, a technique that genetically targets specific neurons to express light sensitive channel proteins, provides the capability to control central nervous system (CNS) neuronal activity in mammals with millisecond time precision. This technique enables precise
83 optical stimulation of neurons and simultaneous monitoring of neural response by electrophysiological means, both in the vicinity of and distant to the stimulation site. We previously demonstrated, in-vitro, the dual capability (optical delivery and electrical recording) while testing a novel hybrid device (optrode-MEA), which incorporates a tapered coaxial optical electrode (‘optrode’) and a 100 element microelectrode array
(MEA). In this chapter, we will decribe a fully chronic implant of a new version of this device in ChR2-expressing rats, and demonstrate its use in freely moving animals over periods up to eight months. In its present configuration, we show the device delivering optical excitation to a single cortical site while mapping neural response from surrounding 30 channels of the 6×6 element MEA, thereby enabling recording of optically modulated single-unit and local field potential (LFP) activity across several millimeters of the neocortical landscape[88].
Previously, we introduced an initial design and fabrication methods of an optoelectronic device where an optical fiber was integrated to a 100 element (10×10) intracortical microelectrode array (MEA). This device was tested in-vitro, to monitor neural activity in response to single site optical stimulation as a proof-of-concept application for triggering and mapping 2D epileptic wave propagation in optogenetically transduced mouse cortical slices [50]. Here we report on an advanced version of this device, with specially designed optical/electrical interface that was chronically implanted and tested in two rats. This device was used for stimulation and recording in freely moving animals over eight and two month experimental period respectively. The transition from an in vitro experimental condition to an in vivo one is a major achievement for this type of device. It took us more than a year to optimize the design of
84 the device, as well as to develop proper surgical procedures, for chronic implantation, viral injection and light targeting. As a result, the presented device was in many aspects different than the one described in our previous work [50]. A comparison of these two devices is the following:
1). In the presented version of the device, the MEA was integrated with a cannula, which allowed viral injections after the implantation of the MEA. The same cannula was used to mount the optrode following the viral injections during the implantation surgery.
The sequential implantations of the optrode and MEA allowed us to target the optrode to the exact same location as the virally injected spot. In the previous optrode-MEA device
[50], optrode and MEA were integrated as a single unit and attached to a micromanipulator for targeting nonspecific regions in an in vitro condition (brain slices).
This configuration was not suitable for chronic implantations with precise targeting.
2). In the presented version of the device, the optrode and its optical and electrical connections were miniaturized and protected for application in awake behaving rats. The laser light pathway uses a relay fiber, while in the previously reported version, the optrode was directly connected to the laser output. Although there was a 30% light loss at the second stage coupling, we gained flexibility by using the miniature optrode and fiber connector, which could be secured in the cone shaped protective head-ware while the animal is freely behaving.
3). The fiber optics based devices are very fragile and it is not uncommon that they break during the implantation surgery or behaving animal experiments. One of the prominent features of the current device is that the optrode and multi-electrode array are independent agents. This allowed us to replace the optrodes whenever necessary. The
85 other equivalent designs (e.g. optoelectrodes from NeuroNexus) integrating the fiber onto the same substrate of electrode array might have limited access to the fiber optics once the device is implanted.
4). The established method for Utah MEA insertion is using a pneumatically actuated impactor [89]. For the current modified MEA construct, we instead advanced the
MEA slowly within the small craniotomy. After having performed the lowering method many times, we observed less dimpling of the brain and less bleeding than after pneumatic insertion. In addition, we had precise control of electrode position within the brain. These all contribute to a higher success rate for chronic recordings.
In this chapter, we demonstrate that the device enabled neuromodulation and simultaneous mapping of electrophysiological responses of neuronal populations in
ChR2-expressing rats in vivo. We show below the use of this device as an optogenetic tool for optical stimulation and electrical mapping of neural activity in a configuration which can be applied to other animal models.
6. 2 Materials and Methods
6. 2. 1 Integrated Optrode-MEA Device Design and
Fabrication for Chronic Use
The design and fabrication of the integrated optrode-MEA device consisted of three steps: (a) the creation of a tapered optical waveguide used as a means of light delivery as a cortically embedded point source to optogenetically transducted targets, (b) the physical integration of such an optrode to the MEA, and (c) means of affixing the
86 integrated device to the head of the animal subject with suitable interconnects for tethered optical and electrical wiring to external instrumentation during experiments.
For rodent use, the optrode has no stainless steel tube reinforecement. The fabrication process was decribed in details in Chapter 2. The scanning electron microscope image (figure 6. 1(b)) shows the conductive coating covering most of optrode leaving only the very tip exposed for light delivery. The current version of the optrode is suitable for in vivo electrophysiological recording. The impedance values were within a range of 200 k to 1.5 M , enabling reliable detection of single or multi-unit activities
Figure 6.1 A single optrode as an in vivo electrophysiological recording tool. (a) Optical images of an optrode being slowly driven into cerebellar cortex of an anesthetized mouse. The inset shows the detailed structure of the optrode. Light could be locally delivered through the aperture of the tapered optical fiber, while simultaneously the neural activities of nearby cells are recorded. (b) SEM image showing that the diameter of the optical aperture is about 10 μm in this optrode. Metal coating around the tapering tip, except the aperture, is indicated. (c) Examples of in vivo recording from an optrode. The left trace is the band pass filtered (300 Hz-10 kHz) recording of multiple unit activity from cerebellar cortex of an anesthetized mouse as shown in (a). The Vpp
87 noise amplitude was about 30μV. Right panel shows the waveform samples from two isolated single units recorded by the optrode
(figure 6. 1(c)). An example of in vivo recording, measured with an extracellular recording amplifier (A-M Systems, Model 1800), from anesthetized mouse cerebellum is shown in figure 6. 1(a) and 1(c). The optrode was mechanically strong enough to penetrate through the dura mater – an important design criteria for projected use in non- human primates.
In the next key fabrication step toward a chronic implant, the optrode (with or without recording capability) was integrated within a modified conventional multi- electrode intracortical array (figure 6. 2(a)). We note that whereas the physical integration described next was important to fabricate and test the optrode-MEAs on a benchtop (with immersion to fluorescent dye-doped saline or agar), in the actual animal experiments, the sequence of integration was operationally different due to the intermediate steps of surgical implantation of the MEA, as well as the use of the laser drilled aperture for injection of optogenetic viral constructs (see next section 2.2). The MEAs were based on commercial silicon-based electrode arrays, fabricated for our chronic rodent experiments as 6×6 arrays of 1mm long tapered Pt-plated microelectrodes (Blackrock Microsystems) with an inter-electrode distance of 400 μm[90]. The MEAs were further processed by ablative laser drilling to remove one silicon electrode at a chosen site (usually near the center), leaving a clear 200μm diameter round hole in the structure. A Teflon cannula
(C316GA, Plastic 1) was centered with and bonded to the open hole on the backside of the MEA, with an approximately 10 μm tolerance. The function of the cannula was to provide a spatially stationary alignment guide for the viral injections and subsequent
88 incorporation of the optrode as a light delivery tool for the chronic in vivo experiments.
The integrated optrode-MEA construct in its final form is shown in figure 6. 2(a) and (b).
The design and the form factor of the single optrodes were chosen such that their shape and dimensions matched the individual intracortical electrodes of the MEAs, though other fiber shapes are also possible, for minimizing tissue damage upon insertion into the cortex.
6. 2. 2 ChR2 transduction and Optrode-MEA Implantation
All procedures were conducted in accordance with the National Institutes of
Health guidelines and approved by the Institutional Animal Care and Use Committee
(IACUC) at Brown University. Subjects were two male Long-Evans rats (6-10 months,
300-320g body weight), mildly food-deprived to 85% of free-feeding body weight and habituated to an operant chamber two weeks before the surgery. Surgeries were performed under aseptic conditions. The animals were anaesthetized with 2% isofluorane-oxygen mixture, prepped, and placed in a stereotaxic apparatus (David Kopf
Instruments). For one subject (M8), the craniotomy was centered on the posterior parietal cortex: anteroposterior (AP) = -4.5 mm from bregma and lateral (ML) = 4.5 mm from the midline. The craniotomy was centered at AP= -3.0 mm and ML= 2.5 mm for the other animal (M9). Since the planar electrode array has dimensions of 2.4×2.4 mm2, it could cover sizable cortical areas, including the posterior parietal cortex (PPC), somatosensory cortex, and partially visual and association areas, on the right hemisphere. Prior to implantation, the MEA, optrodes and injection hypodermic needles were sterilized using ultraviolet light (Hand held UV EF-160C, Spectroline) for 15 seconds to minimize the risk of infection. A series of holes were drilled for placement of 6 skull screws to provide
89 mechanical support and a common ground reference was connected to two of them. A craniotomy window of 3×3 mm2 was opened using a micro-drill (OmniDrill35 World
Precision Instruments Inc.), mounted on the stereotaxic frame. The dura mater was carefully peeled away at the implant location to exposed cortex. The MEA (with its laser drilled hole and guide cannula) was inserted slowly in to the cortex (100 μm/min) using a micromanipulator. This method of insertion caused less dimpling and bleeding of the brain compared to the impact insertion method [91] commonly used for this type of arrays. In addition, slow insertion allowed more precise control of the depth of the bare
MEA in the brain. These all contributed to higher success rate in obtaining chronic neural recordings. After insertion about 500 μm -1 mm into the cortex (including the slight brain swelling), the craniotomy was sealed with silicon elastomer (Kwik-Cast, World Precision
Instruments Inc.) and the entire device was fixed in place using dental acrylic.
Viral injection for ChR2 transduction was conducted shortly after the acrylic was cured. A lentiviral vector was used to express ChR2 and a reporter fluorescent protein (EYFP) under the control of the human synapsin promoter. Plasmid
DNA encoding this transgene (pLenti-Synapsin-hChR2(H134R)-EYFP-WPRE,
Optogenetics.org) was obtained from Stanford University and amplified using standard methods in molecular biology (MegaPrep, QIAGEN). VSV-G pseudo-typed lentivirus was produced at the University of Pennsylvania Vector Core. Typical viral titers were
~1010 IU/ml. A convection driven injector with a hypodermic needle (G32 Hamilton) was slowly driven into the pre-defined cortical location through the guide cannula. Prior to actual injection, we advanced the needle beyond the target depth by 100 μm and retracted it by the same distance to create a vacuous cavity. Column injection targeting at two
90 depths (600 μm and 900 μm below cortical surface) was conducted by injecting viral solution at a speed of 0.1 μL/min and volume of 1 μL at each site.
Then, to complete the optrode-MEA integration during one and the same surgical procedure, the optrode element was gently held by a micromanipulator and inserted through the cannula guide. We ensured that the tip of the optrode was flush with
Figure 6.2 Overview of the optrode-MEA. (a) Image of the 6×6 multi-electrode array device with one element being replaced by an optrode (arrow). The spacing between neighbor electrodes is 400 μm and electrode shank length is 1 mm. (b) A close- up view of the optrode shows the laser light emitted from the tip of optrode. (c) Schematic of the optrode-MEA implant, which shows the cannulated tube used to guide
91 the optrode as well as the injection needle. (d) One of the subjects with the optrode-MEA implanted. The fiber optics and headstage connector are protected by a cone-shaped plastic cylinder. The optical fiber can be coiled and secured inside the cone after each recording session.
the tips of the MEA electrodes plane as pre-determined by calibrating the depth of insertion. Finally, the craniotomy was closed with silicone and acrylic. The connector for a compact headstage was mounted vertically atop the implant (figure 6. 2(c)), protected by a cone-shaped plastic cylinder (figure 6. 2(d)).
6. 2. 3 Instrumentation for optical stimulation and electrophysiological recording by the Optrode-MEA devices.
The optrode fiber was coupled to a relay optical fiber using a standard fiber coupler. At the distal end, the relay fiber was coupled to a 473 nm blue solid-state laser
(OptoEngine LLC.) using a beam collimator (Thorlabs, FiberPort f=4.6 mm). Although there is 30% light loss at the interface between the optrode and relay fiber, we gained flexibility by using the miniature optrode and fiber connector. The laser was triggered by
TTL pulses generated by a Matlab program via a digital I/O interface (NI USB-6501
National Instruments). Maximum optical power exiting the optrode optical aperture to the brain was measured to be 1.8 mW in the actual implant, sufficient to excite the ChR2 expressing neurons at the tip of optrode, where the estimated average power density is about 916 mW/mm2 (See section 6.2.5 and 6.3.1). Time-stamp records of spikes, optical stimulus onset and behavioral events were saved digitally, as well as all sampled spike waveforms.
92
For electrophysiology, a miniature headstage amplifier (20× gain) was used between the preamplifier module and the head-mounted connector. Signals pass through the preamplifier, which provides programmable gain and filtering (set at 150 to 8000 Hz bandpass-), were then recorded and processed by a Plexon Multichannel Acquisition
Processor (Plexon Inc, Dallas TX), a parallel processor capable of digitizing up to 32 channels simultaneously at 40 kHz rate [92]. LFP signals of selected channels were also recorded with band pass filtered between 0.7 Hz and 170 Hz and digitized at 1 kHz sampling rate.
6. 2. 4 Data analysis of the MEA recorded neural signals.
Online waveform selection was performed by time-amplitude window discrimination [92]. Individual units were discriminated by employing principal component analysis or template methods (Offline sorter, Plexon Inc.). Further analysis employed commercial software (Nex software) and custom Matlab (MathWorks Inc.) programs. Peri-stimulus time histograms (PSTHs) were calculated with a bin size of 5 ms
(except for determining the spike latency, where we used 1ms bin), and firing rate was defined as the trial-average spike counts divided by the bin size. Power spectrum analysis on the LFP was performed with frequency range of 1-500 Hz and spectrograms were plotted with 1 second time shift window. The mean LFP power was defined as the integral of the squared LFP values over one period of light stimulation.
93
6. 2. 5 Light Scattering in the Brain: Monte Carlo simulation of photon propagation in tissue
Blue light (473nm) is strongly scattered in neural tissue, placing limits on the activation volume of ChR2-expressing brain tissue emanating from an ideal point source as opposed to the actual absorption by ChR2-transduced neurons at a given cortical target site. We implemented a computational model to estimate the intensity at various locations from the light source, i.e. the optrode. We performed a Monte Carlo simulation as described in Chapter 3. A packet of 106 photons was launched from the source, uniformly distributed according to a chosen angle of divergence. The maximum divergence angle chosen was 30 degrees to the z-axis of the fiber, acquired from a separate experimental observation we made for the emission profile of the optrode immersed and imaged in a control fluorescent dye solution.
Figure 6. 3(b) below shows the results of the simulation, a cross-sectional image of photon density (proportional to light intensity) in the x-z plane, in relation to the optrode source and adjacent microelectrodes of the MEA. The simulation of light distribution in brain tissue was in good agreement with photoimaging experiments, which we performed in vitro in the slice preparation (data not shown), except in the immediate vicinity of the optrode (within one mean free path for photons to the point source). We note that, while alternative models for numerical solving of the photon diffusion equation exist, they are generally considered less accurate near the light source due to the invalid approximations made at this region [48]. We return to the utility of these numerical simulation results below in Section 6.3.
94
6. 2. 6 Histology and fluorescence imaging
At the conclusion of the experiments, subjects were euthanized with pentobarbital sodium (Beuthanasia-D) and transcardially perfused. Brains were fixed in 4% paraformaldehyde for 4 days and then equilibrated in 28% sucrose in PBS. 40μm–thick sections were prepared with a freezing microtome and stored in PBS at 4°C. DAPI staining (1:15000) was conducted following standard immunohistochemical procedures.
Sections were mounted on glass slides, and sealed by the mounting medium
(VECTASHIELD, Vector Laboratories, Inc.) to prevent rapid loss of fluorescence during microscopic examination. Confocal fluorescence images were acquired with a microscope (LSM510, Carl Zeiss) using the 5× and 63× (oil-immersion) objectives. Nissl body staining (in 0.1% cresyl violet solution) was performed on the two subjects (M9 and
M8) implanted for two and eight months, respectively. The Nissl stain images were acquired from an optical microscope. Images were further analyzed and processed using two software packages (ImageJ and Adobe Illustrator). The images are shown in figure 6.
3 and figure 6. 7, and commented on further below.
6. 3 Results
6. 3. 1 Observation and Characterization of in vivo
Neuromodulation Via Optogenetic Activation in Freely Moving
Rats
In this section we describe results from chronic implant experiments, with the aim of demonstrating now the optrode-MEA device can be used for in vivo, freely moving
95 rodents to map spatio-temporal patterns of the optically induced neuromodulation in the cortical area defined by the microelectrode reach of the MEA. We place less emphasis in this thesis on the recording of the single optrode itself, i.e. at the very point of light delivery, since this is a complex topic, including the possible light induced damage to neurons in the immediate zone at the end of the optrode tip. Likewise, there are ongoing investigations which aim to study possible photoinduced detrimental effects to the cells in the immediate vicinity of the small aperture tip during long term light exposure in chronic applications.
For each animal, we began the optical stimulation and neural recording session from about two weeks after initial viral injections, so as to allow reasonable time for
ChR2 expression to develop. As an empirical measure of the level of ChR2 expression, we employed the amplitude of light induced LFP. We found that light-induced responses ramped up to their maximal amplitude in about 4 weeks after injection and persisted at a relatively stable level in the following 2-4 weeks. This observation agrees with those reported recently via in vivo tracking of fluorescence protein expression by a fiber optics opsin detector [15].
At the end of experiments, animals were perfused and brains sectioned for immunohistochemistry to (a) visualize the location and transduction efficiency of ChR2 expression, and (b) examine the anatomical imprints of the chronic intracortical arrays.
Most importantly, the center of injection site showed strong expression as indicated by fluorescence from co-expressed EYFP in the fusion plasmid. Confocal microscope images, co-labeling with YFP (opsin marker) and DAPI (nuclear marker), were shown in figure 6. 3(a). The estimated spread of ChR2 was 700 μm in the horizontal direction
96 and 1 mm in the vertical plane as indicated in this particular brain section. The human synapsin promoter, which exclusively targets neurons, resulted in strong expression of the opsin in cell bodies, processes as well as fibers that project to subcortical structures
(see figure 6. 3(c)) [15].
Conventionally, the methodology for in vivo optogenetic stimulation and neural recording is using an optical fiber mechanically attached to an electrode [93]. Therefore the distance between recording and stimulation sites is predefined and it can only sample one recording site at a time. Another commercial available optrode recording unit
(NeuroNexus) combines the multi-shank silicon probes with a bare optical fiber [11, 94].
The loose confinement of the etched waveguide might cause the problem of low spatial resolution, as discussed in reference [11]. The principal advantage of the optrode-MEA hybrid device lies in its capability to optically modulate at one site (or with more than one fiber, at multi-sites) while recording extracellularly from an ensemble of those neurons that define the microcircuitry environment in the vicinity of the stimulus site.
One important aspect of this approach is the ability to control the size of excited brain volume by controlling the emitted optical power and its distribution within the brain. In free space, the point source optical output from the optrode tip spreads with a divergence angle of 30 degrees. This conical emission, if preserved, would target only a confined region of brain tissue within a conical volume below the optrode tip. The power density of the emission cone would drop approximately with the inverse of the square distance between targeting neurons and optrode tip. However, the laser wavelength (473 nm) optimized for ChR2 excitation is strongly scattered, and weakly absorbed in neural
97 tissue. These properties of the tissue are expected to have significant effects on the spatial distribution of light power.
As described above in Section 6. 2. 5, to enable the estimation of the volume of light-induced ChR2 activation in the cortex with the tapered fiber optrode, we implemented a Monte Carlo model, to calculate the spatial distribution of light intensity at various locations around a point source with an emission divergence angle of 30 degrees, as an approximation to optrode tip (see Materials and Methods). Returning to
Figure 6. 3(b), the numerical simulations show the 10% and 1% contours of iso-density of photons (proportional to light intensity) as a cross-sectional image in the x-z plane, positioned relatively close to the light source and adjacent to microelectrodes of MEA.
The simulations for the taper in question showed how the iso-intensity plane from the hypothetical optrode becomes oval shape at 300 μm (about three times of photon mean free path length for scattering events) away from the tip. According to previous in vitro and in vivo work [9, 18], the minimal light intensity required for spiking of ChR2- transduced neurons is on the order of 1 mW/mm2. Under our stimulus parameters (1.8 mW power at the optrode tip), the light intensity is estimated to be 5 mW/mm2 at the site
400 μm lateral to the stimulus, enough to activate ChR2 expressing cells around the neighbor electrodes. Therefore, the optrode was able to deliver a more than adequate level of light for optogenetic excitation to a volume whose size is primarily determined by tissue scattering and the power output at the tip of optrode. (We note that high optical intensities do, however, exist near the tip aperture, though considerably less e.g. than typically employed in two-photon microscopy).
98
These results indicate that the total volume of excitation through the optrode was determined mainly by the optical power distribution in the brain rather than the expression volume of ChR2. In addition, our optrode-MEA device covered a larger area than both the ChR2 expressing region and the optical distribution, thereby making it possible to monitor the spread of activity in a network of neurons in response to local stimulation.
Figure 6.3 The ChR2 expressing volume and the optical excitation volume. (a) shows a fluoresence image of a coronal slice that resolves individual EYFP-opsin expressing neurons. The cell nuclei were indicated by DAPI staining. (b) The Monte Carlo simulation of photon counts (proportional to light intensity) distribution in brain tissue. The contours of 10% and 1% of the iso-intensity are indicated by yellow dots. The light intensity reaching the neighbor electrodes is 5 mW/mm2 estimated from the simulation (see details in the text). The output light from optrode is approximated as a point source and has divergent angle of 30 degrees. (c) The enlarged images to display cell bodies (left) and fibers (right) that project to subcortical structures.
99
6. 3. 2 Observation and Characterization of in vivo
Neuromodulation via Optogenetic activation in Freely Moving
Rats
In this section we demonstrate the performance of the optrode-MEA hybrid device through selected examples of its utility of operation in freely moving rats, with a focus on recording and mapping the optically induced modulation of neural activity across the 2.4x2.4 mm2 cortical area spanned by the planar MEA.
Optical Neuromodulation of single unit and LFP activity
In optogenetics, targeted cortical illumination can modify spiking activity in many neurons, as well as LFPs, which were here monitored across the MEA recording sites. Figure 6. 4(a) displays an example of one light-responsive neuron in animal M8, including raster plots of the spikes, their firing rates from PSTHs, as well as trial- averaged LFPs. Each light train lasted 5 seconds with peak intensity of 5 mW/mm2 at the specific MEA electrode as estimated by our Monte Carlo simulation. The blue marks indicate the individual pulses of 20 ms duration. For example, a neuron could reliably respond to the stimulation with an approximately threefold induced increase in the probability of spiking (figure 6. 4(a) top), while spike waveform shape remained unaltered. The lowest trace of figure 6. 4(a) represents the trial-averaged LFP, recorded at the same site, in response to pulse train stimulation. The evoked LFP showed a negative deflection followed by a small positive rebound, in accordance with the expected current flow. Since the proximal MEA electrodes were close to the excitatory
ChR2-expressing volume, this type of LFP behavior thus reflected the inward ion flow
100 due to excitation in this area (i.e. near the current sinks) at the initial phase of LFP response.
We further used the simultaneously measured LFPs in animal M8 while asking the question whether optical excitation might be able to induce collective response of neuronal ensembles in response to the pulsed stimulation at specific repetition rates
(frequencies). The extracted power spectrogram (figure 6. 4(b) left) demonstrates the enhanced activity at the stimulation frequency and its harmonics. Investigating the strength of neural activity at varying stimulation frequencies could in principle give another dimension of information about how the neural network responds to optogenetic stimulation. Figure 6. 4(c) illustrates the LFP power as a function of light intensity at the recording electrode. The variation of LFP power was calculated based on recording from
10 sequential pulse trains with same stimulation parameters. The actual output power variation at the optrode tip was measured in an in vitro setting by intentionally bending or twisting the fiber optics and the range of power densities at the recorded site
(horizontal error bar in figure 6. 4(c)) was then obtained from the Monte Carlo simulation. This was meant to give a measure of variability of output intensity due to animal movement. The approximately linear relation between LFP power and light power indicates that we were not obviously “over-driving” the neural network by saturating the optical stimulation response. The high degree of repeatability of the responses allowed us to drive stimulation rates up to 30 Hz for synchronous response
(data not shown here), which suggests how the optrode-MEA could be a useful tool for activating and probing the collective neural network behavior such as rhytmicity across cortical areas of specific interest.
101
Comments on Photo-induced artifacts and Use of Controls for
Validating Stimulation
One caution raised by the combination of optical stimulation and simultaneous electrical recording is that the light stimulus can induce electrical artifacts [95]. Even if much less of a hindrance than the interference which is always present in the case of recording during direct electrical stimulation [96, 97], optical artifacts are caused by direct photoelectronic or photo-induced temperature effects depending on the optoelectronic properties of materials at the recording interface. We have found it convenient both to quantitatively calibrate as well as to discriminate against such finite artifacts e.g. from field potentials, by conducting separate control measurements in the cortex of a freshly sacrificed animal. An example of optical artifacts (top trace labeled
“Control” in figure 6. 4(d)) was recorded from an electrode in the MEA ≈ 0.57 mm away from the optrode. We see a fast recovery (< 5 ms) to the small, if finite, effect, whereas the actual LFP response in the rat’s brain (lower black trace) has much slower effect (~50 ms) elicited by the light stimulus.
We also used a simple “brain phantom” made from 2.5% by weight of agarose in saline solution to emulate and quantify the optical artifact as well as light scattering properties of brain tissue. We obtained similar results with the “brain phantom” (data not shown here) as with the control brain tissue. As another control, we also tested the neuromodulation of single units and population in vivo by stimulating with yellow light
(561 nm, and same parameters as the blue excitation light). Due to the finite absorption tail of the ChR2 chromophore (retinal) at this wavelength, we did observe a finite
102 response, but with LFP amplitude that was decreased to 19% of that using blue light stimulus (data not shown here).
103
Figure 6.4 Representative examples of light activation of single units and LFPs. (a) Raster plots and PSTHs from a light-responsive cell after 5 months of implantation. Blue ticks indicate the pulse train stimulation (473 nm, 8 Hz pulse frequency and 20 ms pulse duration). A sample of overlaid spike waveforms is shown on the right. Lower panel shows that trial averaged LFP has negative deflections with a positive rebound in response to each pulse stimulation. (b) Power spectrogram and power density plot of optically modulated LFP. Note that the power is significantly enhanced at the light stimulation frequency and its harmonics. (c) LFP power around the pulse stimulation frequency as a function of estimated light intensity at the recording electrode site (see text for details). The arrow indicates the light intensity used for neural stimulation in the rest of the article. (d) Optically induced LFP response and comparison with control recordings. “De-sensitization” is shown as the amplitude of negative peak decreases in response to repeatedly optical stimulation.
Mapping of Ensemble Neuronal Activity across Cortical Areas in Rats by the Optrode-MEA Device.
As another example of illustrating how we used the optrode-MEA to acquire real-time spatiotemporal maps of the neural response across a finite population of neurons, figure 6. 5 and 6 shows sample results from animal M9, where the implant was now centered at the somatosensory cortex, with the stimulus and recording taking place at a depth of approximately 1 mm. To further demonstrate the ability of our dual function device to monitor multiple units and collective neural activity in response to local optical neuromodulation, figure 6. 5(a) shows an illustration of the “topographical” mapping of waveforms of 17 isolated units acquired within a recording session from animal M9. In order to discriminate the firing behavior of neurons in figure 6. 5(a), we demonstrated the modulated firing rates of single units as a function of their distance from the site of optrode stimulation, as shown in the scatter plot of figure 6. 5(b). Each dot in figure 6.
5(b) indicates the fraction of the firing rate change for an individual unit, which is defined as :
104
A paired t-test was performed on the firing rates of the baseline epochs (500 ms period before stimulus) and that of stimulation epochs. Most recorded neurons changed their firing rate significantly in response to stimulus as indicated by the filled dots in figure 6. 5(b) (p<0.05). In addition, neurons within 600 µm from the optrode, i.e. those recorded from the nearest and next nearest electrodes, had larger fractional change in their firing rate than those further away from the stimulation site (p<0.01). Data shown in figure 6. 5(b) was from one set of units from a sample recording session. As pointed out previously, it’s generally difficult to hold stable recordings from same individual neurons during recording sessions even in the same day. Whereas, modulations of LFP (figure 6.
5(c) and (d) ) in each individual channel were highly consistent, in terms of relative amplitudes, time-deflection profile and delay time.
The optrode-MEA device was able to consistently monitor on average 13±5 isolated units at each recording session. However, the recording sites, and probably the recorded units, varied from session to session. Claims that the same neurons are recorded necessarily assume some metrics based on, for instance, waveform shapes and inter-spike interval histograms. However, each of these measures can be subject to controversy. In one of a comprehensive study on this issue, the authors estimated that about 39% of single-units were stable across 15 days in their primate recordings [98]. Comparing such epi-cortical “floating” MEA with our skull-anchored optrode-MEA, we expect the chance of stable recording from same units would be less in the latter case. However, a similar systematic study in rodent model is beyond the scope of this presented work which aimed at a proof-of-concept device demonstration, but we note that in future planned
105 deployment of the optrode-MEAs in primates, skull mounting of the devices is unlikely to be used due to larger anatomical workspace.
We emphasize here that the optrode-MEA can be used for chronic recordings in rodents, in the same way that standard MEAs are used routinely in primates. It is important to note that this variability can also be due to neurobiological factors (such as plasticity), as well as mechanical motion. This preliminary work is a first step towards applications to primates where the use of MEAs has been more documented.
The capability of recording whole bandwidth signals, spikes and LFPs, with high spatiotemporal precision gives us the opportunity to study the dynamics of neuromodulation by light by concentrating on specific spectral power bands. Focusing next on LFPs, we recall that spatial mapping of LFPs enables tracking the propagation of activity through the network at a more global level. Therefore, the optrode-MEA hybrid device offers the opportunity to study the dynamics of network affected by neuromodulation via LFP recordings. To explore this aim for animal M9, we plot the averaged LFPs, aligned to the onset of 1 ms stimulus light pulse (vertical line in figure 6.
5(c)), recorded from various locations at the top left corner of the device (refer to figure
6. 5(d) for ease of visualization). The “void” channels, indicated by black in figure 6.
5(d), were due to the laser drilled hole and wiring configuration of our current recording system. All the PSTH and LFP signals discussed above represent averages over 30 trials.
Interestingly, the data shows quite different activation profiles for proximal and distal fields in response to the light stimulus. The “nearest neighbor” electrode, 400 µm away from the excitation spot, has a rapid negative deflection followed by a slow positive
106 rebound, whereas distal electrodes demonstrate opposite polarity at beginning and shifted peak (red dots) according to their distance from the optrode.
Figure 6.5 Spatially and temporally resolved neuronal activities from a large cortical area. (a) The mapping of averaged (N=100) spike waveforms on each input channel obtained from a sample recording session. (b) The fractional change of firing rate (see details in the text) of single units as a function of the distance between the recording electrode and stimulating optrode. The firing rate data were collected from same set of isolated units as in (a). (c) Pulse-triggered LFPs at various locations show both proximal and distal field potential in response to the 1ms pulse stimulation. Vertical line indicates the onset of light pulse and red dots indicate the negative peaks. Output power of optrode is estimated to be 1.2mW. The distances between recording electrode and the optrode are labeled on the individual traces. (d) Mapping the full-band power of pulse-triggered LFP over 2.4×2.4 mm2 cortical area. Black colors indicate the sites without recordings. All the PSTH and LFP signals in figure 6. 4 are averaged over 30 trials.
107
A more detailed current source density analysis of these types of LFP recordings by multi-electrode arrays should allow for more accurate identification of current sinks and sources over space and time [99]. However, in this work we were limited by the 400
µm fixed-grid spacing between MEA electrodes as well as the limited number of available LFP recording sites to conduct such an in-depth analysis. We report that the strongest LFP modulations occurred on the same channels that were in the vicinity of the optrode. It is also difficult to assess the variability within a session due to the coupling efficiency between the optical fiber and the optrode can vary in a freely behaving animal.
These variations did not prevent to elicit consistent responses during the sessions. Using a rotary joint (optical commutator) to couple light into the optrode could possibly minimize this effect.
Cell-type Diversity Suggested by Variability Observed in
Neuromodulatory Circuit Response to Optical Stimulation
We now comment on our observations where pronounced variation has been seen in the nature and magnitude of the modulated electrophysiological responses across a population of single units. To first order, these MEA accessible (excitatory and inhibitory) neurons can be assumed to be a mixture of those cells which can be directly modulated by light and those which lie outside the optrode excitation ‘field of view’
(given the estimated light emission pattern of e.g. figure 6. 3(b)). The latter population is then likely to be only indirectly modulated by the blue laser excitation, reporting via a local network effect triggered from within the photoexcitation volume (~ 1 mm3).
Consider first a given cell and its direct modulated responses to the 1 ms light pulses in figure 6. 6(a), resulting in an increase in instantaneous firing rate as shown in the PSTH
108 data in the figure. The pulse-triggered average of spike delay, plotted as latency in the right hand inset of figure 6. 6(a), shows well-defined, time-locked spiking with 3 ms delay following the blue light pulses, which is an indicator of direct channel activation.
Here, we defined a ‘direct zone’ amongst the recording electrodes that are within 0.57 mm distance away from the stimulation center. We have seldom encountered such directly activated units outside the optrode ‘field of view’, which is also consistent with the distances from Monte Carlo light scattering/absorption simulation results as well as the somewhat larger ChR2-expressing volume acquired deduced from later histological studies. As for variability, we show by contrast another neuron, depicted in figure 6. 6(b) which displayed light induced responses, but now with a ‘silent’ period for about 30 ms after light pulse, probably caused by an indirect-pathway inhibitory input from ChR2- expressing neurons.
Finally, we also employed continuous stimulation, whereupon the blue laser was constantly on for the entire period of 500 ms. Using this condition as a case example, we show in figure 6. 6(c) the time-resolved mapping of neural response from multiple sites across the MEA from a sample recording. The images demonstrate various patterns of modulated single unit activities in terms of the light triggered raster plots of spiking and their PSTHs in the individual panels, with the relative locations of the MEA electrodes to the stimulus site indicated by the accompanying insets.
109
110
Figure 6.6 Examples of single unit recordings under light modulation. (a) Raster plots and PSTHs from a neuron activated by the pulse train (blue ticks). The spike count histogram (right panel) shows the time-locked spikes evoked by the stimulation. Samples of selected spike waveforms are plotted in the right. (b) A unit showing inhibitory response to the same stimulation. (c) Examples of activation patterns of single units at selected sites across the microelectrode array (insets indicate locations), in response to 500 ms continuous light stimulation. The last plot is an example of LFP response.
The diversity of responses exemplified above illustrates the intricate dynamics of the photomodulated cortical network. Some cells were tonically excited or inhibited throughout optical stimulation, whereas other neurons showed a phasic response followed by a more sustained activity. For example, the cell on the second row, first column
(figure 6. 6(c)) increased its firing rate for a few milliseconds after the onset of light stimulation, before going back to its baseline firing rate. Another cell (first row, second column in figure 6. 6(c)) was first inhibited and then increased its firing rate above baseline for the rest of the stimulation.
The different types of neurons are distinguishable by their width of waveforms, firing rates and interspike interva l [100]. We have tentatively classified two types of units, the fast-spiking (FS) and regular-spiking (‘RS’), based on a scatter plot of their baseline firing rate and width between trough to peak (data is not shown here). A factor that limited the analysis was that the waveforms often did not reach their peak, because the recording system stored each spike waveform in only 1ms time window. In general, we didn’t see strong correlation between modulated behavior and neuron types (e.g. in the figure below, both RS and FS firing can be suppressed by light, and their appearance were not clearly organized according to the distance to the optrode). We believe that the observed variety of responses is likely due to a sequence of direct and indirect (synaptic) optical modulation of activity. In these first chronic experiments, the animals were not
111 subject to any trained task so that no explicit correlation with behavior was investigated rather the emphasis here is to underscore the potential utility of the optrode-MEA device for mapping optically modulated activity. Experiments that connect with specific training paradigms are under way.
Figure 6.7 Results from histology studies. Nissl staining shows the anatomical impact of the chronic optrode-MEA devices after explants. Cortical depression of approximately 100 μm was observed in (a) one subject two-month after surgery, and about 350 μm depression was observed in (b) another subject eight-month post-implant. Arrows point to the clearly identifiable traces of individual microelectrodes and the optrode.
The Chronic Optrode-MEA Implant and Possible Tissue Damage
As an indicator of the impact of the chronic implant of the optrode-MEA devices on cortical tissue, figure 6. 7 shows examples of two cases of Nissl stained slices from two subjects after two and eight months of duration. No apparent behavioral changes were observed in either subject during these time periods during their recording sessions or in home cages, respectively. We did not find any obvious signatures of tissue/array interaction, which was consistent with stable recording extracellularly from single units over months by standard MEAs as shown in many experiments in non-human primates
112
[101], however, clear traces of microelectrode/optrode tracks can be seen in figure 6.
7(a), not unlike in chronic monkey experiments with the very same type of MEAs [102].
In the present case, effects of possible additional light induced damage in contributing to the “mechanical” damage remained unclear. Rather, a significant depression of the cortex is present in the images, suggesting that improvements must be made in future experiments with the approach of flexibly anchoring the optrode/MEA devices in rodent use. We believe that the cortical depression is not caused by the arrays themselves per se
(even if their footprint for a rodent is smaller than the approximately 4x4 mm2 are of the
100 element MEAs for primates), but rather the rigid way that the arrays were anchored to the animal skull in these first experiments in freely moving rodents.
6. 4 Summary
We have developed an intracortical device for simultaneous in vivo optical stimulation and multisite extracellular recording which integrates a single localized custom-designed and fabricated optrode into a commercially available microelectrode array. We described above the device design, fabrication, and its surgical implantation into rats. The experiments demonstrate the capability to measure optogenetically mediated neural stimulation in freely moving animals over periods of months. While we have emphasized the device engineering in this thesis, we believe that the device construct described above can lead to extraction of new scientific information about neural microcircuit dynamics under in vivo optical stimulation such that optical neuromodulation can combined with both population recording and neural signal decoding, in the context of specific behaviors and tasks. Short term applications of this device can include investigating basic network properties and characterizing the
113 propagation of activity throughout this network at a spatial scale intermediary between intracellular recordings and ECoG or EEG recordings such as those of importance for example in the study of epilepsy [103]. This optrode-MEA device has been designed to be scalable for accommodating additional optrode sites to enable spatiotemporal patterned light activation or inhibition for targeting specific cortical regions or subcortical structures. Such a device could be desirable for potential clinical trials in human patients.
114
CHAPTER 7 CONCLUSION AND FUTURE
WORKS
In this thesis work, we have presented the concept and fabrication of the single coaxial optrode. Its dual functionality of simultaneous light delivery and electrophysiological recording has been demonstrated. We applied the coaxial optrode as a stand-alone unit in in vivo optical stimulation and recording in various experimental animal models - including transgenic mice, optogenetically transduced rodents and non- human primates. To better understand light distribution through the optrode and compare optical stimulation among different fiber optical structures, we investigated the interaction between light and brain tissue, and then employed the Monte Carlo method to simulate the process of light propagation and associated issue of heating. As suggested by the simulation result, we expect relatively mild temperature increases in response to stimulation paradigms currently used in our lab. We also construct a model that describes the electric behavior of optrode, which allows the spatial sensitivity of optrodes with
115 various coating configurations to be mathematically analyzed. We speculate that, in the vicinity of the membrane potential change, optrode with less exposed coating would have higher recording selectivity. Optogenetics have been applied to many studies and opened up new landscapes in neuroscience. Advance in tool functionality, such as our stimulation/readout enabling optrode, will allow the goal of optogenetics to be achieved: millisecond-scale optical control of defined devents in specified cellular populations while these populations remain embedded and functioning within freely moving animals or other intact and complex biological systems.
In the second part of this thesis, we demonstrate an integrated optrode- microelectrode array (optrode-MEA) device that enables neuromodulation and simultaneous recording of electrophysiological responses of neuronal populations in
ChR2-expressing rodents in vivo. We have showed the use of this device as an optogenetic tool for single site optical stimulation and electrical mapping of neural activity in a configuration which can be applied to other animal models. Recently, to translate these optical neuromodulation studies into non-human primates, we prepared a version of this intracortical device which consists of a 10x10 microelectrode array (MEA) integrated with a polymer optic fiber (POF) for light delivery (Figure 7.1(a)). The device includes an extra cereport (Blackrock Microsystems Inc) pedestal for optical connections in addition to the standard cereport pedestal for electrical connections
(Figure 7.1 (c) and (d)).
Polymer fiber has numerous advantages in short-haul applications over glass- based fibers and foremost it is more pliable. POF has less bend radius and is more resilient to mechanical damage than glass due to its intrinsic material characteristics. The
116 perticular type of POF we use is a step-index multimode POF (MMPOF250, Paradigm
Optics). with 240 µm core diameter ( poly(methyl methacrylate) (PMMA) doped with
3% polystyrene, n=1.4923) and cladding of 5 µm (acrylic coating n=1.4905). The numerical aperture is large (N. A. = 0.5 ) giving better stability of light coupling.
Figure 7.1 (a) A 10x10 MEA integrated with a single POF. The MEA is attached to two cereport pedestals: one for optical and the other for electrical interface. (b) A close view of the POF inserting in a holey array. The backside of the array is relatively flat suitable for the usage of pneumatic inserter. (c) and (d) The schematics of a MEA integrated with two fibers. Note that a single cereport pedestal is enough for two or more optical connections.
We were able to connect the polymer fiber to the device without sacrificing the flatness of the backside of the array (Figure7.1 (a) and (b)). This is important for the usage of this type of devices in primates since the insertion into the brain will be
117 accomplished by a pneumatic inserter which requires a flat device surface. Since planar array device will be floating between cortex and the dura mater, we should keep the backside thin that minimizes the pressure to the cortex. On the other side, the optical interface should be steady so that light coupling is stable. Before the initial test of the
POF-MEA device in primates, the performance of this device has been validated using rodents. First, we used pneumatic inserter to test the insertion of this device in to a rat brain without damaging the brain or the device. After the demonstration of satisfactory device performance, we will proceed to use a MEA integrated with two or more fibers for multi-site stimulation (Figure 7.1 (c)). Multi-site temporally ordered stimulation are important that it can provide insightful information of how certain brain areas coordinate in timing while animal is conducting behavioral task. The POF-MEA device could help unravel the mystery of brain functions. Such device is also a prototype of neuroprosthetics desirable for potential clinical trials in human patients.
118
SUPPLEMENTARY Chapter: Large Scale
Ordered Structures of Single Photon Sources
Based on II-VI Semiconductor Colloidal Quantum
Dots
In this chapter, a separate project on II-VI semiconductor colloidal quantum dots
(QDs) which were conducted earlier in our laboratory [104] is presented. First of all, we have developed a novel and efficient method of deterministically organizing colloidal particles on structured surfaces over macroscopic areas. Our approach utilizes integrated solution-based processes of dielectric encapsulation and electrostatic-force-mediated self- assembly, which allow precisely controlled placement of sub-10nm sized particles at single particle resolution. Later, as a specific demonstration, motivated by application to single photon sources, we have created highly ordered 2D arrays of single II-VI semiconductor colloidal QDs by this method. Individually, the QDs display triggered 119 single photon emission at room temperature with characteristic photon antibunching statistics, suggesting a pathway to scalable quantum optical radiative systems.
8. 1 Introduction
Ordered assemblies of mesoscopic scale colloid particles are an important class of nanomaterials that provide means to achieve one, two and three dimensional structures for a wide variety of applications, ranging from photonics devices [105, 106] to microelectronics [107], memory [108], storage [109] and sensor [110] type of devices, etc. In order to fulfill their potential, it is a prerequisite to understand and gain control over the relevant growth and ordering parameters of the underlying nanometric building units. This becomes increasingly challenging when the particle size migrates into the quantum regime and the desired functional properties are yet only apparent at the individual particle level, requiring the need for precise placement of nanoparticles at single particle resolution.
In specific application areas of quantum key distribution and quantum computing, the development of practical non-classical light sources [111, 112] is a vigorous subject of ongoing research. A number of reports have appeared recently, demonstrating how single photon emission can be possible in solid state systems [113-116]. Among them, II-
VI semiconductor colloidal QDs are highly fluorescent nanocrystals which are prepared through organometallic synthesis [117] in solution phase.
Recent discoveries show that the suppression of bi-exciton and multi-exciton emissions in CdSe-based QDs can be quite effective, as a result of the enhanced non- radiative Auger recombination processes [118, 119] with respect to their bulk
120 counterpart. It follows that the lowest quasi-zero dimensional Wannier exciton state radiates much like a two-level atomic system, making the colloidal QDs one appealing material candidate as room-temperature-operating single photon source [120, 121], owing to its large exciton binding energy commonly seen in II-VI compounds. To exploit such property in practice, integration of the QDs at well-defined position within functional photonic device structures is a prerequisite. However, their translation to device implementation is currently hampered by the lack of precise spatial control of individual emitters on solid surfaces at macroscopic scale.
Thermodynamically driven self-assembly is an efficient method capable of arranging large number of colloidal particles in parallel to create complex nanostructures. It has been explored extensively on the basis of a variety of inter-particle and external forces
[122]. However, precise and reproducible organization of small nanoparticles into versatile geometries with long range order poses severe challenges in controlling the particle-particle and particle-substrate interactions. Perhaps the most promising approach so far is based on capillary forces [123], which has demonstrated large-area high resolution patterning of metallic particles down to 50 nm in size [124, 125]. But its susceptibility to the influence of thermal (Brownian motion) effect in solution has rendered this method ineffective for particles of even smaller size and lighter mass, such as QDs. The sensitivity on detailed solvent hydrodynamic properties also makes it hard to apply to particles suspended in nonaqueous solutions.
Here, we report on a way to overcome these limitations by resorting to indirect self- assembly approach in conjunction with efficacious surface engineering of the nanoparticles in process. Our approach starts with the well-established solution-based
121 synthesis of II-VI colloidal QDs, but now follows the key idea of adapting their surface chemistry so as to substantially enlarge these optically active “seed” nanoparticles by the subsequent growth of a thick (about 100 nm) optically transparent (passive) dielectric shell. The enlarged particles, each containing a single QD, are then individually self- assembled onto pre-defined templates by employing an electrostatic “adhesion” approach. In the end, the process results in geometrically precise planar 2D arrays of silica-clad QDs with controlled spatial specificity of location. In broad terms, we symbiotically combine the so-called “bottom-up” and “top-down” fabrication approaches, as illustrated in Figure 8.1.
Figure 8. 1 The “bottom-up” and “top-down” fabrication approaches. (a) The synthesis procedure of the CdSe-core CdS/Zn0.5Cd0.5S/ZnS-multishell QDs. (b) Indirect self-assembly of surface-engineered nanoparticles. First, uniformly thick shells of optically transparent dielectric material (silica) are grown onto the nanoparticle “seeds” in microemulsion, which is illustrated here as CdSe-based core-shell colloidal QDs; then,
122 the resulted composite particles, each containing a single QD, are individually anchored onto lithographically defined templates by electrostatic force self-assembly in solution.
8. 2 Methods and Results: QD array
fabrication and characterization
8. 2. 1 Synthesis of Colloidal Semiconductor QDs
In terms of improving the optical properties of colloidal QDs, the chemical synthesis of colloidal QDs is the most crucial step one should extensively investigate. The carriers generated within QD during excitation process, being optical or electrical, suffer from the surface traps, and recombination of the trapped carriers creates a deep trap emission band on the longer wavelength side of band edge luminescence. The quantum yield (i.e. emission efficiency) is a common indication of how well the surface is passivated. By controlling organic ligand (usually amine group) and the complex mixed-solvent in fabrication, remarkably high QY can be achieved in CdSe colloidal QDs [126]. However, the stability of these colloidal QDs could not be guaranteed, because of the unstable nature of organic bonding. Epitaxial growth of inorganic shell on highly monodisperse
CdSe nanocrystals is a common technique, which not only overcomes the instability problem, but also enhances the luminescence quantum efficiency (typically almost an order of magnitude compared to the organic capped starting nanocrystal core). In a particular case of CdSe colloidal QD, the shell materials are usually higher band gap semiconductor ZnS [127] or CdS [128]. Therefore, the band gap of the core materials is
123 energetically within the band gap of shell materials and both the excited carriers (electron and hole) are confined mainly inside the CdSe core.
In our case, the CdSe-core CdS/Zn0.5Cd0.5S/ZnS-multishell QDs used in this study were synthesized following a procedure described in Ref. [129], as illustrated in Figure
8.1(a), in which the core was formed by precipitation reaction first and then the desired multishell growth was achieved by successive ion layer adhesion and reaction (SILAR) technique [130]. The synthesis of such core-shell structured colloidal QDs usually consists of two steps:
CdSe core nanocrystal synthesis: CdSe core nanocrystals were synthesized using a standard method. A mixture of 4.0 g TOPO (Trioctylphosphine Oxide 90%) and 58 mg
CdO powder (99.99%) in a 50ml three neck flask under nitrogen flow was gradually heated to 330 OC to form a clear solution. Se precursor prepared by dissolving 80 mg Se powder (99.5%) in 2.4 ml TOP (99%) was injected into reaction flask, the temperature
O O should immediately drop to 260 C, and then it was raised to 280 C. After 190 seconds,
O the heating mantle was removed and when temperature dropped to 100 C, 20ml of methanol was injected to start precipitation of nanocrystals. After centrifugation and decantation, the purified CdSe nanocrystals were dispersed in hexane as a stock solution for shell growth process.
Multishell CdS/Zn0.5Cd0.5S/ZnS growth by SILAR technique: All precursor solutions of 0.1M were freshly prepared under nitrogen atmosphere. The amount of materials and prepared temperature for four precursors are given in Table 8.1. After making the precursor solutions at high temperatures, Cd, Zn and Cd/Zn precursor solutions were kept at 80 OC, while S precursor solution was cooled to room temperature.
124
The typical procedure to grow a multi-shell nanostructure was performed. 3 mL of ODE
O 19 and 1g of ODA were loaded into a 50mL three neck flask, heated to 100 C under nitrogen atmosphere for an hour, then cooled to room temperature.
Table 8. 1 Conditions to prepare precursor solutions for graded multishell
growth by SILAR technique.
Precursor Materials/ ODE (g) Oleic acid Temperature Amount (mg) (mL) (OC)
Cd precursor CdO/ 320.4 6.18 18 300 S precursor S / 80 0 25 180 Cd/Zn precursor ZnO/ 1.7 6.18 18 300 CdO/ 60.2 Zn precursor 203.4 6.18 18 310
The CdSe core nanocrystals in hexane (2.01x10-7mol of particles) were loaded into the flask and temperature was set at 100oC for about 30 minutes to remove hexane and unexpected materials. Then the solution was heated to 235 OC for shell growth. The process started with the Cd precursor injection, and then followed by alternating injections of S, Cd, Cd/Zn, Zn precursor solutions. The amounts of injections were calculated according to growing size of core/shell structure [129]. Because the numbers of Cd and Se atoms in the surface of CdSe core nanocrystal are equal, the first injection
(Cd precursor) used just half calculated amount to complete the first shell monolayer. A period of 10 minutes between each injection is sufficient to complete the reaction. After
O final injection, the solution was kept at 260 C for another 30 minutes and subsequently cooled to room temperature. A 10ml of hexane was added, and then the unreacted compounds and byproducts were removed by successive methanol extraction. The complete shell, which consists of 2 monolayers of CdS, 3.5 monolayers of Zn0.5Cd0.5S
125 and 2 monolayers of ZnS, was epitaxial grown on CdSe core nanocrystals. Examples of high resolution TEM images (Figure 8.2) show the epitaxial core-shell structure of the nanoparticles. The final product dispersed in hexane has a relatively high quantum yield of 50%.
Figure 8. 2 High resolution transmission electron microscope (HR-TEM) images demonstrate epitaxial core-shell growth of CdSe-core CdS/Zn0.5Cd0.5S/ZnS shell nanoparticles. The images are from reference [129].
8. 2. 2 QD encapsulation
The as-synthesized QDs were capped with octadecylamine (ODA). They had an averaged diameter of 8nm and measured fluorescent quantum yield of about 50%.
Individual QDs exhibited the typical blinking widely reported; however in terms of main goal of this work we did not attempt to control or reduce this behavior. We do note that recent results report significant reduction or elimination of blinking [131]. A thick shell of silica was subsequently grown onto the QD utilizing water-in-oil (W/O) microemulsion growth technique, which has been widely adopted to synthesize silica colloidal particles, but so far mainly in the sub-100nm size range [132]. The microemulsion process began in a non-polar solvent (cyclohexane) in which the ODA-
126 capped QDs and an excessive amount of non-ionic surfactant polyoxyethylene(12) nonylphenyl ether (NP-12) were dissolved, which typically contains 1.5-24 nM QDs, 167 mM NP-12, and 797 mM hexanol. Adding a slight amount (105 mM) of NH4OH aqueous solution (30 wt.%) triggered the spontaneous formation of reverse micelles. Through interaction with the reverse micelles, the loosely bonded ODA molecules tended to leave the QD’s surface, driving the QDs into the reverse micelles in a one-to-one correspondence. The micelles then behaved as a microreactor where polycondensation of the hydrolyzed silica precursor tetraethyl orthosilicate (TEOS) took place and nucleated onto the QDs. The reaction started when 41mM of TEOS was applied dropwise and it was allowed to proceed for 8 hours before the particles were isolated by centrifugal precipitation. The thickness of the silica shell can be effectively tuned by controlling the concentrations of the QDs and/or of the TEOS.
Similar silica-cladding approaches have recently been initiated on QDs [133, 134], as well as on magnetic nanoparticles[135], but issues related to particle size control and encapsulation uniformity are reported to be considerable challenges. However, by our technique, we have achieved significant improvement in the final particle size distribution, as seen from the transmission electron microscope (TEM) images in Figure
8.3(a)-(d) where up to 95% of the particles have single QD core precisely positioned at the center. One key to this advance is the choice of the surfactant NP-12 with relatively large unit length polyoxyethylene hydrophilic group. As NP-12 helps to reinforce the stability of the micelle against ethanol, a by-product of the reaction, we could tune the final particle size up to 220 nm in a one-pot synthesis without generating secondary silica nuclei. Nevertheless, shorter polyoxyethylene surfactant, such as polyoxyethylene(5)
127 nonylphenyl ether (NP-5), is still useful to obtain thin silica shells with thickness below
15 nm (Figure 8.3 (a)).
Figure 8. 3 Silica-encapsulated II-VI semiconductor colloidal QDs. a-d, Transmission electron microscope images of the synthesized silica-clad QDs with various total particle diameters of 28 nm (a), 75nm (b), 95nm (c), and 180 nm (d), obtained via microemulsion synthesis with NP-5 (a) and NP-12 (b-d) as the surfactants, respectively. Single QDs of about 8nm in diameter are visible at the core of the composite particles, appearing as small dark dots. The scale bars in the images correspond to 20 nm (a) and 100 nm (b-d). (e) and (f), the synthesized silica-clad QDs (right, 180nm diameter) in cyclohexance under ambient (e) and UV (f) exposures. Bare QD controls (left) with similar concentration was used as a reference. g, photoluminescence spectra of the silica- clad QDs in cyclohexane (blue) and in ethanol (green), in comparison to the bare QD control (red), under the excitation at 380nm.
Important for applications, the optical quality of the QDs is largely preserved after thick silica encapsulation, attributed to the improved QD crystalline and photochemical stabilities [129] offered by the graded multishell synthesis approach in comparison to the traditional single ZnS shell structure. As seen in Figure 8.3 (g), the photoluminescence
128
(PL) spectrum of the QDs after silica encapsulation shows clean QD ground state exciton emission signature centered around 613nm (slightly red-shifted by about 3nm compared to the “bare” QD control), without introducing any noticeable impurity emission background. The PL efficiency of the silica-clad QDs is nearly unchanged if they are dispersed in non-polar solvent; however, it drops by about 60% upon transferring into ethanol in which the final 2D assembly process is carried out (described next). This effect is primarily because of the incomplete passivation of the QD surface as a result of the porosity of the silica material grown by the current method, an issue that we anticipate to address in future work.
8. 2. 3 Self-assembly of the encapsulated QDs
In order to achieve the goal of controlled spatial organization of individual silica- clad QDs, we developed an electrostatic force self-assembly method (abbreviated as
EFSA) whose process flow is sketched in Figure 8.4(a)-(e). The EFSA approach relies on utilizing the electrostatic interactions between the silica-clad QDs and an oppositely charged polyelectrolyte template whose principle has been demonstrated previously with
µm -sized particles [136]. It is implemented here as follows. First, a template consisting of an array of SiO2 circular pads, defining the capture sites for the particles, was created on a silicon substrate, using electron-beam lithography followed by SiO2 deposition and lift-off. Prior to lift-off, a monolayer of cationic polyelectrolyte poly(diallyldimethylammonium chloride) (PDDA) was self-assembled [137] onto the top surface of the pads by dipping into a PDDA aqueous solution for 20mins, which contains a mixture of PDDA (molecular weight 400K- 500K) 0.58 mM and NaCl 0.5 M with pH value of 7.0. Then, immersion of the templates into the ethanol solution with the
129
Figure 8. 4 Electrostatic force self-assembly of the silica-encapsulated QDs at single-particle resolution. (a-e), schematics of the process flow of the electrostatic force self-assembly, which starts off with patterning of the PMMA resist on silicon substrate by electron-beam lithography (a), followed by deposition of SiO2 film (20 nm thick) by ion beam sputtering (b) and assembly of a monolayer of polyelectrolyte (PDDA) molecules on the SiO2 surface by dip-coating (c). Lift-off of the PMMA film in acetone leads to
PDDA covered SiO2 pad (d). Finally, immersion of the template into silica-clad QD ethanol solution results in spatially selective settlement of single silica-clad QD particle on each pad (e) via electrostatic interactions. (f), scanning electron microscope image of a highly ordered array of individual silica-clad QDs formed by electrostatic force self- assembly. A small fraction of the particles are accidentally displaced from the target site, exposing the PDDA-covered SiO2 pad underneath. (g-j), SEM image examples of EFSA results when the pad size is varied from 200 nm (g), 250 nm (h), 300 nm (i) to 400 nm (j). The diameter of the silica-clad QD is 220 nm. The scale bar in (f)-(j) is 2 µm.
negatively surface-charged and electrostatically stabilized QD/silica particles leads to their attraction onto the pads by electrostatic interactions. An immersion time setting of
15mins is typical for a particle concentration on the order of 1012/cm3. For best result, solvents with moderate ionic strength are preferred to avoid over-screening of the electric field. A Debye screening length commensurate with composite particle size helps to improve binding specificity. Once a pad is occupied by a single silica encapsulated QD particle, Coulomb repulsion between the particles overpowers the attraction and in optimal conditions prevents a second particle attachment. Similarly, the silicon substrate carrying a net negative charge due to the hydrolysis of its native oxide surface also
130 remains free of particles. Thus, by fine tuning the pad geometry for a given particle size, we have been able to achieve controlled placement of single QDs into individual sites with very high accuracy, as shown in the scanning electron microscope (SEM) images in
Figure 8.4 (f).
To illustrate the influence of the pad sizes on the kinetics of the particle assembly and the quality of the final result, we varied the pad diameter D from 200 nm to 400 nm for QD/silica particles with diameter d =220 nm, as shown in Figure 8.4 (g)-(j). We found that the particle capture probability decreased when D became smaller than d, whereas multiple particles could be accommodated on a single pad when D was increased to nearly twice the size of the particles. For a wide window of parameter values in between, single particle capture probability up to 95% over macroscopic area (>100x100 µm2) was found, demonstrating a good tolerance of our method to the finite pad and particle size fluctuations from fabrication uncertainties.
It is worth noting that the EFSA method takes substantially less time (typically a few minutes) to implement in comparison to the assembly technique based on capillary forces
[125] (typically a few hours, scaling with substrate size). It is also insensitive to fluctuations in parameters such as temperature and colloidal concentration. Furthermore, the fact that EFSA method relies on surface charge contrast rather than the trapping force exerted by the surface profile suggests certain flexibility in designing the template geometry for practical purposes. As a complementary method, we also developed well- shaped traps (as opposed to pads) within poly(methyl methacrylate) (PMMA) templates which could also capture single QD/silica particles as efficiently (not shown here).
131
8. 2. 4 Optical characterization of the assembled QDs
The optical properties of the geometrically arrayed single QDs were studied and characterized, firstly by fluorescence microscopy, and secondly, through photon statistical measurements, all at room temperature. Figure 8.5(a) shows the confocal fluorescence microscope image collected from a square-lattice array of silica-clad QDs, which spans over an area of 80x80 µm2 with 2 µm pitch. It needs to be stressed that this particular planar geometry is chosen only for demonstration purpose. Virtually any pattern arrangement defined by electron-beam lithography can be used to create the arrays. In acquiring the image, a 405 nm laser diode excitation source and a single photon counting avalanche photodiode (APD) detector were used. Subsequent photon statistical analysis shows that the majority of the emitters sitting precisely at the lattice sites in the image correspond to single QDs. Their fluorescence intensity variations are mostly inherited from the QD themselves, as similar level of fluctuations has been observed in our bare QDs that might links to the variation of QD sizes as well as the randomness of the QD crystal orientation [138]. There are also a small fraction of the spots (less than 7% of the total number) in the image which are significantly brighter than the others. These
“defect” sites tend to be occupied by multiple QDs, either in a single composite particle or as a group of particles.
The ability to create ordered QD assemblies where inter-particle separation larger than the wavelength of light can be specified has given us an easy optical access to targeted individual quantum dots to evaluate their performance as single photon emitters.
For this measurement, we illuminated the sample through our high resolution (N.A. =1.3) confocal microscope with a diode laser operating at 405 nm in pulsed mode (individual
132
Figure 8. 5 Photon statistics measurement of the emission from single silica encapsulated II-VI semiconductor colloidal QDs in self-assembled array. (a), fluorescence image of an array of single silica-clad QDs captured by a confocal microscope. The particles reside on a square lattice with 2 µm pitch. The scale bar is 10 µm. (b), a close-up fluorescence image of a 3x3 silica-clad QD array. The confocal excitation/detection geometry allows easy optical access to any individual silica-clad QD in the array. The scale bar is 2 µm. (c), the Hanbury-Brown and Twiss intensity correlation apparatus for photon statistics measurement of the QD emissions. The QDs are excited by a 2.5 MHz pulsed diode source operating at 405nm. (d) and (e), a typical photoluminescence intensity correlation histogram collected from a single silica-clad QD in the array at room temperature (d), in comparison to the similar data collected from a bare QD control (e). The number in the graph indicates the ratio of the area under the peak at zero time delay to the averaged area of the neighboring peaks.
pulse width τp =70 ps) at 2.5 MHz repetition rate. The collected photoluminescence was fed into a Hanbury-Brown and Twiss intensity correlation instrument as depicted in
Figure 8.5(c). In Figure 8.5(d), a typical photoluminescence intensity correlation histogram acquired from a randomly chosen single silica-clad QD particle in the array is displayed. The suppression of the amplitude of the peak near zero time delay is a clear indication of the photon antibunching characteristics [111]. We define the figure of merit
133 here as the ratio of the area under this peak to the area of the neighboring peak, which yields A(0)= 0.101. In other words, the probability of this silica-clad QD emitting a multi-photon state upon each triggered excitation is suppressed by an order of magnitude relative to a coherent laser source. The result is compared to the data collected from a bare control QD as shown in Figure 8.5(e) with A(0)=0.096. The closeness of the two values demonstrates the compatibility of the silica encapsulation and the assembly processes with the goal of achieving single photon emission characteristics of the individual colloidal QDs. Incomplete suppression of A(0) is attributed to the decreasing
(1/R3) of Auger recombination rate as the QD radius R increases [118]. Smaller CdSe
QDs (R =1.8 nm) with emission peak at 570 nm have shown significantly lower values
A(0) =0.04 was reported by X. Brokmann et al [120].
8. 3 Summary
In summary, we have successfully developed an experimental approach which holds the prospect of alleviating some of the major limitations currently associated with the patterning of sub-10 nm sized nanoparticles at single particle resolution in precise geometries with long range order. This approach is made possible by employing novel methods to first encapsulate individual nanoparticles in large and uniformly sized silica particles via microemulsion growth, and subsequently self-assemble them into predefined patterns utilizing the electrostatic interactions in solution in a massively parallel fashion.
While this method is potentially applicable to a wide variety of materials, a specific example based on II-VI colloidal QDs is given in this thesis demonstrating how 2D single
QD arrays can be created. These highly ordered QDs are readily accessible by confocal fluorescence microscopy for site-selective photoexcitation and collection of spontaneous
134 emission. Most importantly, the effectiveness of these patterned QDs as single photon emitters has been demonstrated in practice at room temperature. We believe that the ability of precise spatial control of semiconductor colloidal QDs on a solid substrate will facilitate the study of these optically active nanoparticles, for example in promoting their integration with advanced microcavity structures and related optoelectronic devices as a possible avenue towards practical single-photon-on-command sources for quantum optical applications.
135
APPENDIX
1. Matlab code for Monte Carlo simulation of light intensity and instant temperature change in the brain tissue.
Main function
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%% % Monte Carlo Simulation of light intensity % by Jing Wang, (v.1.6-Matlab, 64bit system ) Jan. 2012 updated. % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%% function [t]=MC_intensity() %% Parameters photons=1e6; % number of launching photons mu_a = 0.00007;%Absorption Coefficient in 1/um !!non-zero!! mu_s=0.010; % Scattering Coefficient in 1/um g=0.88; %anisotropic factor mfp=1.0/(mu_a + mu_s); %mean free path in um mesh= mfp/10; % mesh size in um Maxsize= 10*mfp; % ROI dimensions MAX = Maxsize/mesh; % max grid dimension, interger
power= 1; % power= 1mW spot_size=200; % in um NA=0.2;% numerical aperture n_re=1.36; %index of refraction theta0=asin(NA/n_re); % half divergent angle,generate launching direction, half of divergent angle in rad
ro=spot_size/2; %ro_ = mesh*ceil(ro/mesh) ; %io=power*1e6/(4*ro_*ro_); io=power*1e6/(pi*ro*ro); % peak intensity, in mW/mm^2, assum emission distribution is gaussian
136 rho= 1.07*1e-3; % density of brain tissue in g/mm3 c=3.6; % specific heat, J/(gK) power_= 1e-3; % power= 1mW in w t_=1e-3; % stimulation time= 1msec to=t_*power_/(rho*c*photons*1e-9*mesh*mesh*mesh); % coefficient, in K
Maxstep=100; % number of max scattering steps
%% trajectory and intensity recording rand('twister', sum(100*clock)); % initialize the random number generator, system time as seed, "rand" uniformly distribute between(0,1) tic
X(Maxstep,photons,3)=0; %position vector(x,y,z) of each photon at each scattering position %initial position vector(x,y,z) in um phi=2*pi*rand(size(X(1,:,1))); r=(ro)*sqrt(rand(size(X(1,:,1)))); % % uniform light spot % r=(ro)*sqrt(-log(rand(size(X(1,:,1))))); % % gaussian spot X(1,:,1) = r.*cos(phi); X(1,:,2) = r.*sin(phi); X(1,:,3)=0;
L(Maxstep,photons) = 0; % travel length
abs_=zeros(2*MAX,2*MAX,2*MAX); % probability of absorption
psi= 2*pi*rand(photons,1); % direction vector (u,v,w) w = cos(theta0*rand(photons,1));% uniformly divergent beam u = sqrt(1-w.*w).*cos(psi); v = sqrt(1-w.*w).*sin(psi);
for it=2:Maxstep psi = 2*pi*rand(photons,1); % generate new direction cosp = cos(psi); sinp = sin(psi); %cost = CTheta(g,photons); temp_ = (1-g*g)./(1-g+2.*g.*rand(photons,1)); cost = (1+g*g - temp_.*temp_)/(2*g); sint = sqrt(1.0 - cost.*cost);
temp = sqrt(1.0 - w.*w); temp_u =u; temp_v =v; temp_w =w; u = sint.*(temp_u.*temp_w.*cosp - temp_v.*sinp)./temp + temp_u.*cost;
137
v = sint.*(temp_v.*temp_w.*cosp + temp_u.*sinp)./temp + temp_v.*cost; w = -sint.*cosp.*temp + temp_w.*cost;
s = (-mfp.*log(rand(photons,1)));% generate a step size from random number L(it,:)=L(it-1,:)+ s'; % total travel path length abs_weight=exp(-L(it-1,:)'*mu_a).*(mu_a/(mu_a+mu_s)); % abs_weight=exp(-L(it-1,:)'*mu_a).*(1-exp(-s*mu_a)); X(it,:,:)=squeeze(X(it-1,:,:))+ [s.*u s.*v s.*w]; % move a step, record position
inds=squeeze(ceil(X(it,:,:)./mesh)+MAX); %indices of spatial location in_ROI=find(not(sum((inds<1 | inds>2*MAX),2)));% indices of photon within ROI
for n=1:length(in_ROI) in=in_ROI(n); % index of photon abs_(inds(in,1),inds(in,2),inds(in,3))= abs_(inds(in,1),inds(in,2),inds(in,3))+ abs_weight(in); end end To=to*abs_; Io=io*abs_./max(max(max(abs_)));
% 2D plot grid=(-MAX+0.5: 1: MAX-0.5)*mesh; Io2=squeeze(mean(Io(MAX:MAX+1,:,:),1)); %x-z plan profile int_plot(Io2,grid,[1e-5 1]); % intensity plot t=toc; % counting the caculation time assignin('base','mesh',mesh); assignin('base','MAX',MAX); assignin('base','Maxstep',Maxstep); assignin('base','Io',Io); assignin('base','To',To); %filename=strcat('MC_',num2str(power),'_',num2str(spot_size)); %save(filename, 'mesh','MAX','Maxstep','Io','To'); end % %% costheta function % function costheta = CTheta(g,n) % end % if(g ==0) % costheta = 2*rand(n,1)-1; % else % temp = (1-g*g)./(1-g+2.*g.*rand(n,1)); % costheta = (1+g*g - temp.*temp)/(2*g); % % end % end
138
Function for taking a section of intensity matrix and plotting in 2d
% plot a 2D intensity profile and equ-intensity contour, in log10 scale function []=int_plot(int,grid, colorbar_limit) figure; imagesc(grid,grid, int); colorbar xlabel('\bf z (\mum)') ylabel('\bf x (\mum)') title ('\bf Intensity Distribution (mW/mm^2)') int=int/max(max(int)); % normalization figure;imagesc(grid,grid, log10(int),log10(colorbar_limit)); colorbar xlabel('\bf z (\mum)') ylabel('\bf x (\mum)') title ('\bf Intensity Distribution (in log10 scale)') figure; contourf(grid,grid,log10(int),log10([0.0001 0.001, 0.01 0.1 1])); colormap hot colorbar
2. Matlab code for simulation of light induced heating effect
Temperature function in response to the impulse stimulation
%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % light induced thermal response % Temperature distribution, relaxation after 1ms 1mW impulse stimulus % by J. Wang, Oct.2011 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function [time]=T_impulse()
MAX = evalin('base', 'MAX'); mesh= evalin('base', 'mesh'); Ti= evalin('base', 'To'); k=0.56*1e-3;% heat conductivity, in W/mm/K rho= 1.07*1e-3; % density of brain tissue in g/mm3 c=3.6; % specific heat, J/(gK) % power= 1e-3; % power= 1mW in w % t=1e-3; % stimulation time= 1msec % mu_a= 0.1; % abs coeff in mm^-1 alpha=k/rho/c; % in mm/sec dt=0.5e-3; % time interval =1msec nt=100;% max response time elapsed coeff=alpha*dt/(mesh*mesh*1e-6);% increment
139 grid=mesh*(-MAX+0.5: 1: MAX-0.5);
%% initial intensity distribution
T=cell(nt,1); T{1,1}= Ti; % initial temperature elavation : rho*c*dT/dt=mu_a*intensity tic for i=2:nt T{i,1}= coeff.*laplace(T{i-1,1}) +(1-coeff).*T{i-1,1}; end assignin('base','Tend',T{end,1});
T2(nt,2*MAX,2*MAX)=0; for i=1:nt temp=T{i,1}; T2(i,:,:)=(mean(temp(MAX:MAX+1,:,:),1)); end assignin('base','dt',dt); assignin('base','T2',T2);
% figure; % for i=1:nt % subplot(10,10,i) % %surf(squeeze(T2(i,:,:))); % im=squeeze(T2(i,:,:)); % %imagesc(im,[0 max(max(im))]); % image(25600*im,'CDataMapping','direct') % axis off ; % colormap hot % title(['t= ',num2str(dt*(i-1)),'s']); % end figure; n_t=round(1e-3/dt); for i=1:n_t:nt plot(grid,squeeze(T2(i,MAX,:)),'-'); hold on xlim([-100 1000]) %ylim([0 0.05]) end xlabel('\bfz (\mum)') ylabel('\bfdT (K)') hold off legend(['dt=',num2str(n_t*dt*1e3),'msec']) time=toc; end function [P]=laplace(p)
140
%P(size(p))=0; % if length(size(p))==3; %padx=zeros(size(p(1,:,:))); px1=cat(1, p(1,:,:), p(1:end-1,:,:)); px2=cat(1, p(2:end,:,:), p(end,:,:)); % pady=zeros(size(p(:,1,:))); py1=cat(2, p(:,1,:), p(:,1:end-1,:)); py2=cat(2, p(:,2:end,:), p(:,end,:)); %padz=zeros(size(p(:,:,1))); pz1=cat(3, p(:,:,1), p(:,:,1:end-1)); pz2=cat(3, p(:,:,2:end), p(:,:,end)); P=(px1+px2+py1+py2+pz1+pz2)./6; % else return % end end
e.g. Simulation of temperature response to a pulse train stimulation
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % apply pulse train stimulation %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% power=7; % Power in mW nt=1000;% max time elapse 1sec nstim=nt; %stim=zeros(nstim,1); f=50; % frequency Hz duration=5; % in ms stim(1:nstim)= (mod((1:1:nstim),round(nt/f)) <= duration);% stimulation duration 100ms, 100Hz 10 pulses
%R=cell(nt,1); % response function %% time convolution %for i=1:nt Rt=zeros(size(squeeze(T2(1,:,:)))); for j=1:nt Rt= Rt + stim(j)*squeeze(T2((nt-j+1),:,:)); end Rt=Rt*power; % R{i,1}=Rt*power; %end grid=mesh*(-MAX+0.5: 1: MAX-0.5); R2=squeeze(mean(Rt(MAX:MAX+1,:),1));
figure; bar(stim);xlabel('\bfTime(ms)') % int_plot(R2,grid,[1e-5 1]); figure; for i=1:nt plot(grid,squeeze(T2(i,MAX,:)),'-'); hold on xlim([-100 1000]) ylim([0 0.01])
141 end xlabel('\bfz (\mum)') ylabel('\bfdT (K)') hold off legend(['dt=',num2str(dt*1e3),'msec'])
figure; plot(grid,R2,'-'); xlim([-500 1000]) ylim([0 0.5]) xlabel('\bfz (\mum)') ylabel('\bfdT (K)') legend(['t= ', num2str(nt*dt*1e3),'ms']); figure; imagesc(grid,grid, Rt); colorbar xlabel('\bf z (\mum)') ylabel('\bf x (\mum)') title ({'\bf Temperature Distribution after 1sec stimulation (in Kevin)';' (50um core dia, 50Hz, 7mW )'})
142
BIBLIOGRAPHY
[1] Perlmutter J S and Mink J W 2006 Deep brain stimulation Annu Rev Neurosci 29 229-57 [2] Cruikshank S J, Lewis T J and Connors B W 2007 Synaptic basis for intense thalamocortical activation of feedforward inhibitory cells in neocortex Nat Neurosci 10 462-8 [3] Mayberg H S, Lozano A M, Voon V, McNeely H E, Seminowicz D, Hamani C, Schwalb J M and Kennedy S H 2005 Deep brain stimulation for treatment- resistant depression Neuron 45 651-60 [4] Nagel G, Szellas T, Huhn W, Kateriya S, Adeishvili N, Berthold P, Ollig D, Hegemann P and Bamberg E 2003 Channelrhodopsin-2, a directly light-gated cation-selective membrane channel Proceedings of the National Academy of Sciences of the United States of America 100 13940-5 [5] Schobert B and Lanyi J K 1982 Halorhodopsin is a light-driven chloride pump Journal of Biological Chemistry 257 10306-13 [6] Chow B Y, Han X, Dobry A S, Qian X, Chuong A S, Li M, Henninger M A, Belfort G M, Lin Y, Monahan P E and Boyden E S 2010 High-performance genetically targetable optical neural silencing by light-driven proton pumps Nature 463 98-102 [7] Deisseroth K, Feng G, Majewska A K, Miesenböck G, Ting A and Schnitzer M J 2006 Next-Generation Optical Technologies for Illuminating Genetically Targeted Brain Circuits The Journal of Neuroscience 26 10380-6 [8] Zhang F, Wang L-P, Brauner M, Liewald J F, Kay K, Watzke N, Wood P G, Bamberg E, Nagel G, Gottschalk A and Deisseroth K 2007 Multimodal fast optical interrogation of neural circuitry Nature 446 633-9 [9] Boyden E S, Zhang F, Bamberg E, Nagel G and Deisseroth K 2005 Millisecond- timescale, genetically targeted optical control of neural activity Nat Neurosci 8 1263-8 [10] Zhang F, Aravanis A M, Adamantidis A, de Lecea L and Deisseroth K 2007 Circuit-breakers: optical technologies for probing neural signals and systems Nat Rev Neurosci 8 577-81 [11] Royer S, Zemelman B V, Barbic M, Losonczy A, Buzsáki G and Magee J C 2010 Multi-array silicon probes with integrated optical fibers: light-assisted perturbation and recording of local neural circuits in the behaving animal European Journal of Neuroscience 31 2279-91 [12] Zorzos A N, Boyden E S and Fonstad C G 2010 Multiwaveguide implantable probe for light delivery to sets of distributed brain targets Opt. Lett. 35 4133-5 [13] Maesoon I, Il-Joo C, Fan W, Wise K D and Euisik Y 2011 Neural probes integrated with optical mixer/splitter waveguides and multiple stimulation sites.
143
In: Micro Electro Mechanical Systems (MEMS), 2011 IEEE 24th International Conference on, pp 1051-4 [14] Gradinaru V, Thompson K R, Zhang F, Mogri M, Kay K, Schneider M B and Deisseroth K 2007 Targeting and Readout Strategies for Fast Optical Neural Control In Vitro and In Vivo The Journal of Neuroscience 27 14231-8 [15] Diester I, Kaufman M T, Mogri M, Pashaie R, Goo W, Yizhar O, Ramakrishnan C, Deisseroth K and Shenoy K V 2011 An optogenetic toolbox designed for primates Nat Neurosci 14 387-97 [16] Han X, Qian X, Bernstein J G, Zhou H-h, Franzesi G T, Stern P, Bronson R T, Graybiel A M, Desimone R and Boyden E S 2009 Millisecond-Timescale Optical Control of Neural Dynamics in the Nonhuman Primate Brain Neuron 62 191-8 [17] Gradinaru V, Mogri M, Thompson K R, Henderson J M and Deisseroth K 2009 Optical Deconstruction of Parkinsonian Neural Circuitry Science 324 354-9 [18] Aravanis A M and et al. 2007 An optical neural interface: in vivo control of rodent motor cortex with integrated fiberoptic and optogenetic technology Journal of Neural Engineering 4 S143 [19] Crone N E, Miglioretti D L, Gordon B and Lesser R P 1998 Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. II. Event-related synchronization in the gamma band Brain 121 2301-15 [20] Schwartz A B 2004 CORTICAL NEURAL PROSTHETICS Annual Review of Neuroscience 27 487-507 [21] Logothetis N K, Pauls J, Augath M, Trinath T and Oeltermann A 2001 Neurophysiological investigation of the basis of the fMRI signal Nature 412 150- 7 [22] McCulloch S and Uttamchandani D 1997 Development of a fibre optic micro- optrode for intracellular pH measurements Optoelectronics, IEEE Proceedings - 144 162-7 [23] Moreno-Bondi M C, Wolfbeis O S, Leiner M J P and Schaffar B P H 1990 Oxygen optrode for use in a fiber-optic glucose biosensor Analytical Chemistry 62 2377-80 [24] Wolfbeis O S and Schaffar B P H 1987 Optical sensors: An ion-selective optrode for potassium Analytica Chimica Acta 198 1-12 [25] Robinson D A 1968 The electrical properties of metal microelectrodes Proceedings of the IEEE 56 1065-71 [26] Humphrey D R and Schmidt E M 1991 ed A A Boulton, et al.: Humana Press) pp 1-64 [27] Li L, Bowler N, Hondred P R and Kessler M R 2011 Influence of thermal degradation and saline exposure on dielectric permittivity of polyimide Journal of Physics and Chemistry of Solids 72 875-81 [28] Merrill E and Ainsworth A 1972 Glass-coated platinum-plated tungsten microelectrodes Medical and Biological Engineering and Computing 10 662-72 [29] Wise K D, Angell J B and Starr A 1970 An Integrated-Circuit Approach to Extracellular Microelectrodes Biomedical Engineering, IEEE Transactions on BME-17 238-47 [30] Sanes J N, Suner S and Donoghue J P 1990 Dynamic organization of primary motor cortex output to target muscles in adult rats I. Long-term patterns of
144
reorganization following motor or mixed peripheral nerve lesions Experimental Brain Research 79 479-91 [31] Wang J, Ozden I, Diagne M, Wagner F, Borton D, Brush B, Agha N, Burwell R, Sheinberg D, Diester I, Deisseroth K and Nurmikko A 2011 Approaches to optical neuromodulation from rodents to non-human primates by integrated optoelectronic devices. In: Engineering in Medicine and Biology Society,EMBC, 2011 Annual International Conference of the IEEE, pp 7525-8 [32] Unpublished work by Ilker Ozden et al. [33] Rickgauer J P and Tank D W 2009 Two-photon excitation of channelrhodopsin-2 at saturation Proceedings of the National Academy of Sciences 106 15025-30 [34] Anderson R and Parrish J 1983 Selective photothermolysis: precise microsurgery by selective absorption of pulsed radiation Science 220 524-7 [35] Ashley J. Welch A J C v G 2011 Optical-Thermal Response of Laser-Irradiated Tissue: Springer Science + Business Media) [36] Grinvald A, Lieke E, Frostig R D, Gilbert C D and Wiesel T N 1986 Functional architecture of cortex revealed by optical imaging of intrinsic signals Nature 324 361-4 [37] Prahl S http://omlc.ogi.edu/spectra/hemoglobin/ [38] Yaroslavsky A N, Schulze P C, Yaroslavsky I V, Schober R, Ulrich F and Schwarzmaier H J 2002 Optical properties of selected native and coagulated human brain tissues in vitro in the visible and near infrared spectral range Physics in Medicine and Biology 47 2059 [39] Bamann C, Kirsch T, Nagel G and Bamberg E 2008 Spectral Characteristics of the Photocycle of Channelrhodopsin-2 and Its Implication for Channel Function Journal of Molecular Biology 375 686-94 [40] Lin J Y, Lin M Z, Steinbach P and Tsien R Y 2009 Characterization of Engineered Channelrhodopsin Variants with Improved Properties and Kinetics Biophysical journal 96 1803-14 [41] Hegemann P and Moglich A 2011 Channelrhodopsin engineering and exploration of new optogenetic tools Nat Meth 8 39-42 [42] Palczewski K, Kumasaka T, Hori T, Behnke C A, Motoshima H, Fox B A, Trong I L, Teller D C, Okada T, Stenkamp R E, Yamamoto M and Miyano M 2000 Crystal Structure of Rhodopsin: A G Protein-Coupled Receptor Science 289 739- 45 [43] McRae S R, Brown C L and Bushell G R 2005 Rapid purification of EGFP, EYFP, and ECFP with high yield and purity Protein Expression and Purification 41 121-7 [44] Mie G 1908 Contributions to the optics of turbid media, particularly of colloidal metal solutions. Annalen der Physik 3 [45] Bohren C F and Huffman D R 2007 Absorption and Scattering of Light by Small Particles: Wiley-VCH Verlag GmbH) pp 499-519 [46] Pllana S 2000 History of Monte Carlo Method [47] Feynman R 1985 Surely You're Joking, Mr. Feynman! : W.W. Norton) [48] Wang L and Wu H-i 2007 Biomedical Optics: Principles and Imaging: Wiley) [49] Henyey L G, Greenstein, J. L. 1941 Diffuse radiation in the Galaxy Astrophys. J, 93 70-83
145
[50] Zhang J and et al. 2009 Integrated device for optical stimulation and spatiotemporal electrical recording of neural activity in light-sensitized brain tissue Journal of Neural Engineering 6 055007 [51] Wang J, Ozden I, Diagne M, Wagner F, Borton D, Brush B, Agha N, Burwell R, Sheinberg D, Diester I, Deisseroth K and Nurmikko A 2011 Approaches to optical neuromodulation from rodents to non-human primates by integrated optoelectronic devices Conf Proc IEEE Eng Med Biol Soc 2011 7525-8 [52] Holmes K R Thermal conductivity data for specific tissues and organs for humans and other mammalian species. [53] Price R R 1999 The AAPM/RSNA physics tutorial for residents. MR imaging safety considerations. Radiological Society of North America Radiographics 19 1641-51 [54] Moser E, Mathiesen I and Andersen P 1993 Association between brain temperature and dentate field potentials in exploring and swimming rats Science 259 1324-6 [55] DeBow S and Colbourne F 2003 Brain temperature measurement and regulation in awake and freely moving rodents Methods 30 167-71 [56] Grossman N, Poher V, Grubb M S, Kennedy G T, Nikolic K, McGovern B, Palmini R B, Gong Z, Drakakis E M, Neil M A A, Dawson M D, Burrone J and Degenaar P 2010 Multi-site optical excitation using ChR2 and micro-LED array Journal of Neural Engineering 7 [57] Yizhar O, Fenno Lief E, Davidson Thomas J, Mogri M and Deisseroth K 2011 Optogenetics in Neural Systems Neuron 71 9-34 [58] Malmivuo J and Plonsey R 1995 Bioelectromagnetism: Principles and applications of bioelectric and biomagnetic fields: Oxford University Press) [59] Geselowitz D B 1967 On bioelectric potentials in an inhomogeneous volume conductor Biophys J 7 1-11 [60] Plonsey R 1969 On volume conductor fields of compound nerve action potentials Med Biol Eng 7 253-4 [61] Schwan H P and Kay C F 1957 Capacitive properties of body tissues Circ Res 5 439-43 [62] Helmholtz H 1853 Ann. Physik und Chemie 89 211-33 [63] Plonsey R 1963 Reciprocity Applied to Volume Conductors and the Ecg IEEE Trans Biomed Eng 10 9-12 [64] Pelosi G 2007 The finite-element method, part 1: R. L. Courant Ieee Antenn Propag M 49 180-2 [65] Moffitt M A and McIntyre C C 2005 Model-based analysis of cortical recording with silicon microelectrodes Clin Neurophysiol 116 2240-50 [66] Mizumori S J and Williams J D 1993 Directionally selective mnemonic properties of neurons in the lateral dorsal nucleus of the thalamus of rats J Neurosci 13 4015-28 [67] Gallistel C R 1990 The organization of learning. (Cambridge, MA: MIT Press.) [68] Calton J L and Taube J S 2009 Where am I and how will I get there from here? A role for posterior parietal cortex in the integration of spatial information and route planning Neurobiol Learn Mem 91 186-96
146
[69] Morris R G M 1981 Spatial Localization Does Not Require the Presence of Local Cues Learn Motiv 12 239-60 [70] Lynch J C, Mountcastle V B, Talbot W H and Yin T C 1977 Parietal lobe mechanisms for directed visual attention J Neurophysiol 40 362-89 [71] Robinson D L and Kertzman C 1995 Covert orienting of attention in macaques. III. Contributions of the superior colliculus J Neurophysiol 74 713-21 [72] Mishkin M, Lewis M E and Ungerleider L G 1982 Equivalence of parieto- preoccipital subareas for visuospatial ability in monkeys Behav Brain Res 6 41-55 [73] Critchley M 1953 Tactile thought, with special reference to the blind Brain 76 19- 35 [74] Nakamura K 1999 Auditory spatial discriminatory and mnemonic neurons in rat posterior parietal cortex J Neurophysiol 82 2503-17 [75] Buneo C A and Andersen R A 2006 The posterior parietal cortex: sensorimotor interface for the planning and online control of visually guided movements Neuropsychologia 44 2594-606 [76] Colby C L and Goldberg M E 1999 Space and attention in parietal cortex Annu Rev Neurosci 22 319-49 [77] McNaughton B L, Mizumori S J, Barnes C A, Leonard B J, Marquis M and Green E J 1994 Cortical representation of motion during unrestrained spatial navigation in the rat Cereb Cortex 4 27-39 [78] Medendorp W P, Beurze S M, Van Pelt S and Van Der Werf J 2008 Behavioral and cortical mechanisms for spatial coding and action planning Cortex 44 587-97 [79] Nitz D A 2006 Tracking route progression in the posterior parietal cortex Neuron 49 747-56 [80] Corwin J T 1998 Identifying the genes of hearing, deafness, and dysequilibrium Proceedings of the National Academy of Sciences of the United States of America 95 12080-2 [81] O'Keefe J and Dostrovsky J 1971 The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat Brain Res 34 171-5 [82] Furtak S C, Cho C E, Kerr K M, Barredo J L, Alleyne J E, Patterson Y R and Burwell R D 2009 The Floor Projection Maze: A novel behavioral apparatus for presenting visual stimuli to rats Journal of Neuroscience Methods 181 82-8 [83] Deadwyler S, Bunn T and Hampson R 1996 Hippocampal ensemble activity during spatial delayed-nonmatch-to-sample performance in rats J. Neurosci. 16 354-72 [84] Nicolelis M A L, Ghazanfar A A, Faggin B M, Votaw S and Oliveira L M O 1997 Reconstructing the engram: Simultaneous, multisite, many single neuron recordings Neuron 18 529-37 [85] Burwell R D and Amaral D G 1998 Cortical afferents of the perirhinal, postrhinal, and entorhinal cortices of the rat The Journal of Comparative Neurology 398 179- 205 [86] Whitlock J R, Sutherland R J, Witter M P, Moser M-B and Moser E I 2008 Navigating from hippocampus to parietal cortex Proceedings of the National Academy of Sciences 105 14755-62
147
[87] Hampson R E, Simeral J D and Deadwyler S A 2002 "Keeping on track": firing of hippocampal neurons during delayed-nonmatch-to-sample performance J Neurosci 22 RC198 [88] Wang J, Wagner F, Borton D A, Zhang J, Ozden I, Burwell R D, Nurmikko A V, van Wagenen R, Diester I and Deisseroth K 2012 Integrated device for combined optical neuromodulation and electrical recording for chronic in vivo applications J Neural Eng 9 016001 [89] Rousche P and Normann R 1992 A method for pneumatically inserting an array of penetrating electrodes into cortical tissue Annals of Biomedical Engineering 20 413-22 [90] Bhandari R, Negi S and Solzbacher F 2010 Wafer-scale fabrication of penetrating neural microelectrode arrays Biomed Microdevices 12 797-807 [91] Rousche P and Normann R 1992 A method for pneumatically inserting an array of penetrating electrodes into cortical tissue Annals of Biomedical Engineering 20 413-22-22 [92] Wiebe S P and Stäubli U V 1999 Dynamic Filtering of Recognition Memory Codes in the Hippocampus The Journal of Neuroscience 19 10562-74 [93] Zhang F, Gradinaru V, Adamantidis A R, Durand R, Airan R D, de Lecea L and Deisseroth K 2010 Optogenetic interrogation of neural circuits: technology for probing mammalian brain structures Nat. Protocols 5 439-56 [94] Kravitz A V, Freeze B S, Parker P R L, Kay K, Thwin M T, Deisseroth K and Kreitzer A C 2010 Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry Nature 466 622-6 [95] Cardin J A, Carlen M, Meletis K, Knoblich U, Zhang F, Deisseroth K, Tsai L-H and Moore C I 2010 Targeted optogenetic stimulation and recording of neurons in vivo using cell-type-specific expression of Channelrhodopsin-2 Nat. Protocols 5 247-54 [96] Jimbo Y, Tateno T and Robinson H P C 1999 Simultaneous Induction of Pathway-Specific Potentiation and Depression in Networks of Cortical Neurons Biophysical Journal 76 670-8 [97] Olsson R H, III, Buhl D L, Sirota A M, Buzsaki G and Wise K D 2005 Band- tunable and multiplexed integrated circuits for simultaneous recording and stimulation with microelectrode arrays Biomedical Engineering, IEEE Transactions on 52 1303-11 [98] Dickey A S, Suminski A, Amit Y and Hatsopoulos N G 2009 Single-Unit Stability Using Chronically Implanted Multielectrode Arrays Journal of Neurophysiology 102 1331-9 [99] Mitzdorf U 1985 Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena Physiological Reviews 65 37-100 [100] Wilson F A, O'Scalaidhe S P and Goldman-Rakic P S 1994 Functional synergism between putative gamma-aminobutyrate-containing neurons and pyramidal neurons in prefrontal cortex Proc Natl Acad Sci U S A 91 4009-13 [101] Serruya M D, Hatsopoulos N G, Paninski L, Fellows M R and Donoghue J P 2002 Brain-machine interface: Instant neural control of a movement signal Nature 416 141-2
148
[102] Serruya M, & Donoghue, J. 2003 Design principles of a neuromotor prosthetic device. Design two 1158-96. [103] Truccolo W, Donoghue J A, Hochberg L R, Eskandar E N, Madsen J R, Anderson W S, Brown E N, Halgren E and Cash S S 2011 Single-neuron dynamics in human focal epilepsy Nat Neurosci 14 635-41 [104] Zhang Q, Dang C, Urabe H, Wang J, Sun S and Nurmikko A 2008 Large ordered arrays of single photon sources based on II-VI semiconductor colloidal quantum dot Opt. Express 16 19592-9 [105] Coe S, Woo W-K, Bawendi M and Bulovic V 2002 Electroluminescence from single monolayers of nanocrystals in molecular organic devices Nature 420 800-3 [106] Younan X, Gates B and Sang Hyun P 1999 Fabrication of three-dimensional photonic crystals for use in the spectral region from ultraviolet to near-infrared Lightwave Technology, Journal of 17 1956-62 [107] Chung S-W, Ginger D S, Morales M W, Zhang Z, Chandrasekhar V, Ratner M A and Mirkin C A 2005 Top-Down Meets Bottom-Up: Dip-Pen Nanolithography and DNA-Directed Assembly of Nanoscale Electrical Circuits Small 1 64-9 [108] Tseng R J, Tsai C, Ma L, Ouyang J, Ozkan C S and Yang Y 2006 Digital memory device based on tobacco mosaic virus conjugated with nanoparticles Nat Nano 1 72-7 [109] Sun S, Murray C B, Weller D, Folks L and Moser A 2000 Monodisperse FePt Nanoparticles and Ferromagnetic FePt Nanocrystal Superlattices Science 287 1989-92 [110] Ali M F, Kirby R, Goodey A P, Rodriguez M D, Ellington A D, Neikirk D P and McDevitt J T 2003 DNA Hybridization and Discrimination of Single-Nucleotide Mismatches Using Chip-Based Microbead Arrays Analytical Chemistry 75 4732- 9 [111] Brahim L and Michel O 2005 Single-photon sources Reports on Progress in Physics 68 1129 [112] Shields A J 2007 Semiconductor quantum light sources Nat Photon 1 215-23 [113] Kurtsiefer C, Mayer S, Zarda P and Weinfurter H 2000 Stable Solid-State Source of Single Photons Physical Review Letters 85 290-3 [114] Michler P, Imamoglu A, Mason M D, Carson P J, Strouse G F and Buratto S K 2000 Quantum correlation among photons from a single quantum dot at room temperature Nature 406 968-70 [115] Michler P, Kiraz A, Becher C, Schoenfeld W V, Petroff P M, Zhang L, Hu E and Imamoglu A 2000 A Quantum Dot Single-Photon Turnstile Device Science 290 2282-5 [116] Strauf S, Michler P, Klude M, Hommel D, Bacher G and Forchel A 2002 Quantum Optical Studies on Individual Acceptor Bound Excitons in a Semiconductor Physical Review Letters 89 177403 [117] Murray C B, Norris D J and Bawendi M G 1993 Synthesis and characterization of nearly monodisperse CdE (E = sulfur, selenium, tellurium) semiconductor nanocrystallites Journal of the American Chemical Society 115 8706-15 [118] Klimov V I, Mikhailovsky A A, McBranch D W, Leatherdale C A and Bawendi M G 2000 Quantization of Multiparticle Auger Rates in Semiconductor Quantum Dots Science 287 1011-3
149
[119] Wang L-W, Califano M, Zunger A and Franceschetti A 2003 Pseudopotential Theory of Auger Processes in CdSe Quantum Dots Physical Review Letters 91 056404 [120] Brokmann X, Giacobino E, Dahan M and Hermier J P 2004 Highly efficient triggered emission of single photons by colloidal CdSe/ZnS nanocrystals Applied Physics Letters 85 712-4 [121] Lounis B, Bechtel H A, Gerion D, Alivisatos P and Moerner W E 2000 Photon antibunching in single CdSe/ZnS quantum dot fluorescence Chemical Physics Letters 329 399-404 [122] Min Y, Akbulut M, Kristiansen K, Golan Y and Israelachvili J 2008 The role of interparticle and external forces in nanoparticle assembly Nat Mater 7 527-38 [123] Yin Y, Lu Y, Gates B and Xia Y 2001 Template-Assisted Self-Assembly: A Practical Route to Complex Aggregates of Monodispersed Colloids with Well- Defined Sizes, Shapes, and Structures Journal of the American Chemical Society 123 8718-29 [124] Cui Y, Björk M T, Liddle J A, Sönnichsen C, Boussert B and Alivisatos A P 2004 Integration of Colloidal Nanocrystals into Lithographically Patterned Devices Nano Letters 4 1093-8 [125] Kraus T, Malaquin L, Schmid H, Riess W, Spencer N D and Wolf H 2007 Nanoparticle printing with single-particle resolution Nat Nano 2 570-6 [126] Qu L and Peng X 2002 Control of Photoluminescence Properties of CdSe Nanocrystals in Growth Journal of the American Chemical Society 124 2049-55 [127] Hines M A and Guyot-Sionnest P 1996 Synthesis and Characterization of Strongly Luminescing ZnS-Capped CdSe Nanocrystals The Journal of Physical Chemistry 100 468-71 [128] Peng X, Schlamp M C, Kadavanich A V and Alivisatos A P 1997 Epitaxial Growth of Highly Luminescent CdSe/CdS Core/Shell Nanocrystals with Photostability and Electronic Accessibility Journal of the American Chemical Society 119 7019-29 [129] Xie R, Kolb U, Li J, Basché T and Mews A 2005 Synthesis and Characterization of Highly Luminescent CdSe−Core CdS/Zn0.5Cd0.5S/ZnS Multishell Nanocrystals Journal of the American Chemical Society 127 7480-8 [130] Li J J, Wang Y A, Guo W, Keay J C, Mishima T D, Johnson M B and Peng X 2003 Large-Scale Synthesis of Nearly Monodisperse CdSe/CdS Core/Shell Nanocrystals Using Air-Stable Reagents via Successive Ion Layer Adsorption and Reaction Journal of the American Chemical Society 125 12567-75 [131] Mahler B, Spinicelli P, Buil S, Quelin X, Hermier J-P and Dubertret B 2008 Towards non-blinking colloidal quantum dots Nat Mater 7 659-64 [132] Arriagada F J and Osseo-Asare K 1999 Synthesis of Nanosize Silica in a Nonionic Water-in-Oil Microemulsion: Effects of the Water/Surfactant Molar Ratio and Ammonia Concentration Journal of Colloid and Interface Science 211 210-20 [133] Darbandi M, Thomann R and Nann T 2005 Single Quantum Dots in Silica Spheres by Microemulsion Synthesis Chemistry of Materials 17 5720-5
150
[134] Selvan S T, Tan T T and Ying J Y 2005 Robust, Non-Cytotoxic, Silica-Coated CdSe Quantum Dots with Efficient Photoluminescence Advanced Materials 17 1620-5 [135] Yi D K, Lee S S, Papaefthymiou G C and Ying J Y 2006 Nanoparticle Architectures Templated by SiO2/Fe2O3 Nanocomposites Chemistry of Materials 18 614-9 [136] Aizenberg J, Braun P V and Wiltzius P 2000 Patterned Colloidal Deposition Controlled by Electrostatic and Capillary Forces Physical Review Letters 84 2997-3000 [137] Mashl R J, Gronbech-Jensen N, Fitzsimmons M R, Lutt M and Li D 1999 Theoretical and experimental adsorption studies of polyelectrolytes on an oppositely charged surface The Journal of Chemical Physics 110 2219-25 [138] Empedocles S A, Neuhauser R, Shimizu K and Bawendi M G 1999 Photoluminescence from Single Semiconductor Nanostructures Advanced Materials 11 1243-56
151