THE GLIA-NEURONAL RESPONSE TO CORTICAL ELECTRODES:
INTERACTIONS WITH SUBSTRATE STIFFNESS AND ELECTROPHYSIOLOGY
by
JAMES PATRICK HARRIS
Submitted in partial fulfillment of the requirements
For the degree of Doctor of Philosophy
Dissertation Adviser: Dr. Dustin J. Tyler
Department of Biomedical Engineering
CASE WESTERN RESERVE UNIVERSITY
January, 2012
CASE WESTERN RESERVE UNIVERSITY
SCHOOL OF GRADUATE STUDIES
We hereby approve the thesis/dissertation of
James Patrick Harris
candidate for the Doctor of Philosophy degree *.
(signed) Dustin J. Tyler (chair of the committee)
Jeffrey R. Capadona
Robert H. Miller
Dawn Taylor
Christoph Weder
(date) August 26, 2011
*We also certify that written approval has been obtained for any
proprietary material contained therein.
To Mom and Dad, the people that make me the proudest.
Table of Contents
List of Tables ...... vii List of Figures...... viii Acknowledgements ...... xii Abstract...... xvi CHAPTER 1. SPECIFIC AIMS...... 1 Aim 1 ...... 2 Aim 2 ...... 3 Aim 3 ...... 3 Outline...... 4 CHAPTER 2. INTRODUCTION...... 5 2.1. Statement of Problem ...... 5 2.2. Importance of Problem ...... 6 2.3. Review...... 7 2.3.1. Brain Recording Technologies...... 7 2.3.1.1. Electrical Versus Other Recording Paradigms ...... 7 2.3.1.2. Electrical Electrode Recording Types ...... 9 2.3.1.3. Information Transfer Rates of Electrodes and Other Assistive Technologies ...... 12 2.3.1.4. Intracortical Electrode Technologies...... 14 2.3.1.5. Importance of Intracortical Electrode Recording Site Size...... 15 2.3.1.6. Types and Characteristics of Intracortical Electrode Recordings ...... 16 2.3.1.7. Performance and Limitations of Intracortical Electrodes...... 17 2.3.2. Trauma from Microelectrode Insertion ...... 19 2.3.3. Tissue Response ...... 21 2.3.3.1. Acute Phase of Injury ...... 22 2.3.3.2. Chronic Phase of Injury...... 22 2.3.3.2.1. Chronically Activated Inflammatory Cells (Macrophages/Microglia)...... 22 2.3.3.2.2. Blood Brain Barrier (BBB)...... 25 2.3.3.2.3. Astrocytes ...... 27 2.3.3.2.4. Chemokines and Cytokines ...... 29 2.3.3.2.5. Other Cell Types and Factors Associated in Response ...... 29 2.3.3.2.6. Summary of Chronic Response ...... 30 2.3.3.3. Mechanical effects...... 31 2.3.3.3.1. Micro Mechanical Effects on Cells ...... 31 2.3.3.3.2. Macro Mechanical Effects on Tissue ...... 32 2.3.3.3.3. Pathways for Mechanical Effects on Cells ...... 34 2.3.3.3.4. Mechanical Effects on Device and Packaging...... 35 2.3.3.4. Intracortical Electrode Recording Longevity ...... 36 2.4. Summary...... 41 CHAPTER 3. IN VIVO DEPLOYMENT OF MECHANICALLY ADAPTIVE NANOCOMPOSITES FOR INTRACORTICAL MICROELECTRODES*...... 42 3.1 Abstract...... 42 3.2. Introduction...... 43 3.3. Materials and Methods...... 47
iv 3.3.1. Device Fabrication...... 47 3.3.2. Surgical Procedures...... 49 3.3.3. Mechanical Test Procedures...... 50 3.3.3.1. Microtensile Testing Setup...... 50 3.3.3.2. Moisture and Temperature Maintenance...... 51 3.3.3.3. Dimensional Changes Due to Swelling...... 52 3.3.4. Insertion Force Measurements ...... 52 3.3.5. Buckling Load Calculations ...... 53 3.3.6. Buckle Testing...... 53 3.3.7. Histology ...... 54 3.4. Results...... 55 3.4.1. Microprobe Insertion into Rat Cortex ...... 55 3.4.2. Dry NC Modulus Measurements...... 57 3.4.3. In vivo Mechanical Switching...... 58 3.4.4. Examination of Ex Vivo Modulus Testing Moisture ...... 59 3.4.5. Buckling and Insertion of Neat and NC Polymers ...... 61 3.4.6. Implanted NC Microprobe Histology...... 65 3.5. Discussion ...... 67 3.6. Conclusions...... 74 3.7. Acknowledgements ...... 75 CHAPTER 4. Mechanically adaptive intracortical implants improve the proximity of neuronal cell bodies* ...... 76 4.1 Abstract...... 76 4.2. Introduction...... 77 4.3. Materials and Methods...... 80 4.3.1. Implant Fabrication and Imaging ...... 80 4.3.2. Water Contact Angle ...... 82 4.3.3. Surgical Procedures...... 82 4.3.4. Fixation, Tissue Preparation, and Immunohistochemistry...... 85 4.3.5. Image Acquisition and Analysis...... 87 4.3.6. Statistics...... 90 4.4. Results...... 90 4.4.1. Water Contract Angle...... 90 4.4.2. Neuronal Nuclei...... 91 4.4.3. Astrocytes...... 92 4.4.4. Proteoglycan...... 93 4.4.5. Vimentin...... 94 4.4.6. Macrophages/Microglia...... 95 4.4.7. IHC intensity based on NC side ...... 97 4.5. Discussion ...... 98 4.6. Conclusions...... 106 4.7. Acknowledgements ...... 106 CHAPTER 5. LPS-induced inflammation degrades neural recordings...... 108
v 5.1. Abstract...... 108 5.2. Introduction...... 109 5.3. Materials and Methods...... 114 5.3.1. Surgical Procedures...... 114 5.3.2. Electrical Measurements ...... 118 5.3.2.1. Anesthetized Neural Recordings ...... 118 5.3.2.2. Impedance Spectroscopy ...... 120 5.3.2.3. Free Moving Neural Recordings ...... 121 5.3.3. Histology ...... 121 5.3.4. Image Acquisition and Analysis...... 123 5.3.5. Electrophysiology Analysis...... 125 5.3.5.1. Non-Evoked Analysis...... 125 5.3.5.2. Evoked Analysis...... 128 5.4. Results...... 129 5.4.1. Inflammatory Cells: Macrophages and Microglia...... 129 5.4.2. Astrocytes...... 132 5.4.3. All Cell Nuclei: DAPI ...... 133 5.4.4. Neuronal Nuclei (NeuN) ...... 134 5.4.5. Dendrites and Neurodegeneration (MAP2 and Fluoro Jade C)...... 135 5.4.6. Electrode Impedance ...... 136 5.4.7. Electrophysiology...... 137 5.4.7.1. Overall Characteristics of Recordings...... 137 5.4.7.2. Unit Recordings with SNR>=1 ...... 138 5.4.7.3. Evoked Recordings...... 138 5.5. Discussion ...... 140 5.6. Conclusions...... 150 5.7. Acknowledgements ...... 150 CHAPTER 6. CONCLUSION...... 152 Appendix...... 160 A.1. Mechanical Inserter and Load Cell (Force) Setup Procedures ...... 160 A.2. Surgical Procedures for Rats ...... 165 A.3. Perfusion + Slicing + Labeling Brain Tissue ...... 171 A.4. Image Analysis (in Matlab)...... 174 References...... 178
vi
List of Tables
Table 1. Review of current assistive technologies and information transfer rate (ITR) of
those technologies. ______13
Table 2. Summary of intracortical electrode types. ______15
Table 3. Summary of Primary Antibodies. ______86
Table 4. Summary of IHC analysis when subdivided based on side. ______97
Table 5. Summary of Primary Antibodies. ______123
Table 6. Contingency table used for Fisher’s Exact Test for count data. ______147
Table 7. Table for Calibration ______164
Table 8. Representative data tables of peak intensities measured across several animals
to be given to statistical program to utilize a general linear model to analyze
statistically significant effects______177
!
vii
List of Figures
Figure 1. Diagram illustrating several types of electrodes used in BMIs ______10
Figure 2. Diagram of trauma and acute damage to brain parenchyma caused by
electrode insertion into brain. ______20
Figure 3. Representative diagram illustrating a lack of response to an implanted
electrode (“No Response”) and the typical chronic response to an indwelling
intracortical electrode. ______23
Figure 4. Number of units recorded in 2 subjects from BrainGate project. ______37
Figure 5. Diagram of Microtensile Testing (MT) instrumentation. ______49
Figure 6. Insertion force recorded as a function of microprobe position during all
microprobe insertion trials. ______56
Figure 7. Stress-strain plots for dynamic microprobes.______58
Figure 8. The Young’s modulus of mechanically dynamic materials as a function of time
in medium.______59
Figure 9. The maximum force was recorded by a computer controlled load cell during
several insertion attempts. ______62
Figure 10. Snapshots of Supplemental Video S1 in (Harris et al., 2011) of insertion
attempts of the neat polymer and the NC microprobes. ______64
Figure 11. Snapshots of Supplemental Video S2 in (Harris et al., 2011) of insertion,
retraction, and buckling of NC probe. ______65
Figure 12. Histological evaluation of the microprobe-tissue interface obtained H+E and
DAPI. ______67
Figure 13. Parameter space for microprobe insertions.______71
viii Figure 14. Materials bilaterally implanted in rodent cortex. ______81
Figure 15. Quantification of fluorescent immunohistochemistry (fIHC) staining via two
methods. ______89
Figure 16. Representative Images and Analysis of NeuN Immunohistochemistry. ____ 91
Figure 17. Representative Images and Analysis of GFAP-Reactive Astrocytes as a
function of distance from the tissue-implant border. ______93
Figure 18. Representative Images and Analysis of CS56-CSPGs as a function of distance.
______94
Figure 19. Representative Images and Analysis of vimentin as a function of distance. 95
Figure 20. Representative Images and Analysis of IBA1-Microglia
immunohistochemistry as a function of distance from the tissue-implant border.
______96
Figure 21. Representative Images and Analysis of ED1-Activated Macrophage and
Microglia immunohistochemistry as a function of distance from the tissue-implant
border. ______97
Figure 22. Implantation site on brain for electrode in two different animals ______117
Figure 23. Experimental Setup of whisker stimulus. ______120
Figure 24. Representative diagram of electrophysiological analysis for a single sample
electrode or channel. ______128
Figure 25. Analysis of IBA1-Microglia immunohistochemistry and representative images
of microglial and activation. ______131
ix Figure 26. Analysis of ED1-Activated macrophages and microglia
immunohistochemistry as a function of distance from the tissue-implant border.
______132
Figure 27. Analysis of GFAP-Reactive astrocyte immunohistochemistry as a function of
distance from the tissue-implant border. ______133
Figure 28. Analysis of DAPI-all cell nuclei histochemistry as a function of distance from
the tissue-implant border. ______134
Figure 29. Analysis of NeuN Immunohistochemistry. ______135
Figure 30. Quantification of degradation of dendrite density and neural degeneration via
MAP2 and Fluoro Jade C. ______136
Figure 31. Bar graphs for signal and noise measures across all electrodes. ______138
Figure 32. Bar graphs for evoked recordings showing single spike and LFP recordings.
______140
Figure 33. Impedance measurements over four week time course of implanted electrodes
measured at 1kHz with animal anesthetized. ______144
Figure 34. Measures of non-evoked electrophysiology quality in unanesthetized animal
over four week time course of implanted electrodes showing units per site and
average noise voltage for all electrodes. ______148
Figure 35. Peristimulus time histograms (PSTH) in anesthetized animal over four week
time course of implanted electrodes showing changes in spiking rates based on
stimulus for all electrodes. ______148
Figure 36. Snapshot of Inserter program before start of the program. ______161
x Figure 37. Popup window that is shown when the motor has not been previously
initialized. ______162
Figure 38. Example of Indicator lights when and movement error has occurred. ___ 157
xi
Acknowledgements
I would not have made it through graduate school without the gracious help of
many people. I feel lucky to have had the help from them all. Thank you to everyone for
helping me out and giving me the chance to figure it out.
Even though, I’ve thanked them in the dedication page, I need to thank my parents again. They have provided me with everything great. Everything that I have accomplished or enjoyed in life is because of them. Countless times they have put me before themselves, and I am truly grateful. It is a pleasure to make them proud.
Coming to Case was not a clear decision, but the graduate students here created a culture of inclusion, excitement, enjoyment, and exploration. These people were the reason that I chose Case. For Christa Wheeler Moss, Michael Ackermann, Scott
Lempka, Adam Boger, and Stephen Foldes, thank you for making me feel so excited about the potential of a CWRU Ph.D. Likewise, thank you for being great friends and colleagues. From discussing neurons to baseball, it was pleasure learning from you in every aspect. Likewise, thank you to all the people that started with me in the program that helped me through. From Lee Fisher, I now know how annoying one can be without actually being annoying while being incredibly smart. From Tom Bulea, I have learned countless things. He has not only been an incredible gentleman, but also exemplifies how to look like you aren’t working hard at all, but in actuality, are working harder than anyone else.
Thank you to all the other graduate students that have helped me along the way.
Molly Fuller will always stand out as the first person that started to make sense of immunohistochemistry for me. Sarah Busch was always available to help make sense of
xii glial scars. Bernadette Erokwu for teaching me how to perfuse an animal. Andrew
Caprariello for passing on the tricks of the cryostat. Kadhiravan Shanmuganathan and
Lorraine Hsu for making the basic materials I needed to conduct my research. Hearty thanks go to Allison Hess and Jeremy Dunning as well. Their hard work has always been timely and incredibly thorough, my graduation thanks you. Kelsey Potter has provided excess help at every request. Andrew Shoffstall helped me experiment with the H+E staining. Additionally, to all the other graduate students that I have interacted with, thank you for helping here or there.
Additionally, I want to thank many of the graduate students that have taught me lessons that are useful outside of the lab. Most notably, Kevin Speer and Quentin
Jamieson. Presidents of the Graduate Student Senate (GSS), they both have taught me about being passionate about a cause as well as controlling that passion towards a productive end. For many of the graduate students in the GSS, thank you for your time and dedication that pushed me to do more.
I would be remiss in not thanking Emily Hornack who not only did a great job in her service in the GSS, but also in being a great friend and partner in the Brite Winter
Festival. Through her help and support, I have learned that anything is possible, everyone works differently, and sometimes that it is more important to complete something small rather than just dreaming big. For all the other people that I have met in
Cleveland, thank you for making life outside of the lab interesting and informative.
I have had the opportunity to interact with many graduate students and postdocs outside of Case, and I would like to thank a couple in particular. First is Nick Langhals.
He was always able to answer each question with more information than I could imagine.
xiii Likewise, it was through his help and gracious sharing of his code that I was able to analyze my data. Additionally, I want to thank Yan Wong for his countless arguments regarding the zen-like discussion of what is really noise and what is really signal, a conversation that stimulates the scientist and philosopher in me.
Thank you for the help of Tina Emancipator and Jenifer Mikulan whose animal expertise helped me make the jump from circuit boards to BME. They helped in many ways, big and small. Similarly, Paul Marasco was the first person, albeit five years into my Ph.D., that sat me down and told me how to “correctly” perform surgeries. His input was incredibly helpful and elucidating. Anne DeChant rescued me from the vibratome abyss. Through her help and knowledge, she made sure that I could perform histology and navigate Bob’s schedule with the help of Sara VanDommelen. Thank you both.
Which brings me to Bob Miller, who I need to thank for all his very efficient meetings that provided clear advice, clearer insights, and succinct lessons. Thank you to
Dawn Taylor for her insights into neural recordings and her gracious attitude regarding my monopolization of her equipment. Thank you for Chris Weder for his straightforward comments about the course of the research. Additionally, while Erin Lavik has not been at CWRU as long as I would have liked, I thank her for taking the time to help me at any point. She has constantly opened her door and asked me straightforward questions that I should be asking myself. She has been a great mentor. Thank you to Dustin Tyler for providing the initial path of this project, comments on grantsmanship, and taking time to write superlative recommendations for future employment opportunities.
There are many faculty members and staff of the University that I have had the benefit of working with in my career at CWRU. Thank you to all of them, but special
xiv thanks go to Doreen Thibodeau and Colleen Barker-Williamson. Both are tireless advocates for students and are always able to help. Additionally, I want to thank Carol
Adrine who has always helped, regardless of her workload or duties.
Penultimately, I want to thank Jeff Capadona for being a fantastic mentor through the whole process of graduate school. He has been there from the beginning of my program. I have enjoyed our conversations and the opportunities to work on projects together. Every time I needed anything, he always made the time to help, even though it did not help his career. He has constantly pushed me while helping me. If I hadn’t had his support, I would have never stayed. For that I believe that I am very lucky, and I thank him for helping me to see this through.
Lastly, I want to thank Paige Cramer, an inspirational person and special someone.
She is the one of the hardest working people that I know, and I strive to be half the scientist and person that she is. Insightful conversations with her have been crucial to interpreting my data. Likewise, she has challenged my thinking and actions in all respects. I am very thankful to have had her support in the lab as well as life.
xv The Glia-Neuronal Response to Cortical Electrodes: Interactions with Substrate Stiffness and Electrophysiology
Abstract
by
JAMES PATRICK HARRIS
The overall goal of this work is to improve intracortical electrodes for chronic recordings from the brain. The studies in the dissertation examine material stiffness, the tissue response, and electrode recording quality to enable improvements. Intracortical electrodes can improve the quality of life of patients with severe paralysis via brain machine interfaces (BMIs), but factors limit electrode widespread clinical usage. One factor of implanted intracortical electrodes is that the number of recorded signals decreases over time, limiting the longevity of electrodes. The decrease is hypothesized to be from the tissue response to the electrodes. In this dissertation, we investigate the tissue response and its effect on electrode recordings. Specifically, we examine the role of material stiffness on the tissue response. Recently developed, a mechanically adaptive nanocomposite decreases its stiffness from 5 GPa to 12 MPa in vitro. Confirming in vitro work, we show that the nanocomposite is stiff enough for insertion, but softer than traditional BMI electrodes after brain implantation. The nanocomposite enables our examination of the role of material stiffness on tissue response. Implanting the soft nanocomposite and surface-matched stiff microwire, we examine the effect of stiffness on tissue response. At one month, greater neural density around the nanocomposite versus the microwire accompanies changes in mechanically associated factors, intermediate filaments and extracellular matrix components. At two months, the neural
xvi densities are similar between implant types, and neural density around the nanocomposite
is maintained despite macrophage activation. The soft nanocomposite modifies the tissue
response, and therefore, we examine whether the tissue response affects neural
recordings. By using lipopolysaccharide (LPS) to promote the inflammatory response,
LPS-treated animals show an increase in non-neural cells and a decrease in the neural
cells near the implant. Additionally, the quality of recordings is reduced in LPS-treated
animals. Though further research is necessary, the results of this work support the
hypothesis that a softer material can improve long-term neural recording. The work guides electrode development to improve electrodes to offer patients with severe paralysis and other neurological deficits a better quality of life.
xvii CHAPTER 1. SPECIFIC AIMS
The basis of the work in this dissertation is to further understanding of the brain’s
glia-neuronal response to intracortical electrodes. Implanted electrodes offer the ability
to restore and augment function via a brain machine interface (BMI) (Donoghue, 2008).
Intracortical electrode technologies also allow examination of brain functions and
dysfunction. Therefore, research to improve understanding on the glia-neuronal response
to intracortical electrodes may have a far-reaching impact on several neurological conditions.
The response to electrodes and their recordings are highly complex, sensitive, and variable, nevertheless, electrodes have been employed in a clinical setting to help patients with spinal cord injury regain function (Hochberg et al., 2006). While significant work has been performed over the years, the electrode technology still cannot be deployed in a widescale clinical setting due to issues with quality and longevity of neuronal signals
(Schwartz, 2004; Tresco and Winslow, 2011).
Several different studies have been performed to improve the response to electrodes or the performance of electrodes, but a clear understanding of the response to implanted electrodes and the impact of the glial scar on recordings is still lacking (Purcell et al., 2009; Zhong and Bellamkonda, 2007; Leach et al., 2010). The work in this dissertation will examine fundamental aspects of the glia-neuronal response. Specifically, the main areas of the work will examine how the substrate stiffness modifies the tissue response and how tissue response modifies neural recordings in vivo.
1 The role of material stiffness on cells and tissue has been examined in vitro and in silico, and a few studies have indirectly examined mechanical effects in vivo (Subbaroyan et al., 2005; Cullen et al., 2007; Biran et al., 2007). Using a novel nanocomposite that changed from a stiff state to enable insertion but switched to a compliant state at body temperature while wet, we were able examine the impact of stiffness on the glia-neuronal response.
Though several factors play a role in the tissue response to intracortical electrodes, reviewed in Chapter 2, the inflammatory stimulant lipopolysaccharide has a well-published history and was utilized to modify the tissue response. Additionally, the role of the in vivo tissue response in electrode recordings has not been well examined as the majority of studies have been focused on immunohistochemical analysis or electrophysiological analysis, not both (Winslow et al., 2010; Williams et al., 1999).
Therefore, the goal of this work is to better understand the interactions between implant stiffness, tissue response, and electrode recordings in vivo. To accomplish this goal, we pursued three specific aims.
Aim 1
Quantify the mechanical properties of a novel mechanically adaptive nanocomposite in vivo
The focus of this aim was to compare the material performance between two conditions: in vitro and in vivo. The hypothesis was that the nanocomposite material will decrease its stiffness in vivo, similar to in vitro results. Rodents were implanted with the nanocomposite material, and a previously verified microtensile tester (Hess et al., 2009) provided measurements of the nanocomposite stiffness after various implant times. In
2 vivo experiments examined similar sized neat polymer and nanocomposite probes, and the in vivo nanocomposite polymer results were compared to previous in vitro results
(Shanmuganathan, 2010).
Aim 2
Quantify the effect of substrate stiffness on cortical tissue
The hypothesis was that the mechanical mismatch between brain tissue and microelectrodes modifies the inflammatory response. A novel nanocomposite dynamically alters its stiffness from 5 GPa to 12 MPa upon exposure to in vitro conditions (Capadona et al., 2008). The material provided a dynamic range not present in other materials. The nanocomposite facilitated insertion into the brain when in a stiff state. When the nanocomposite was in its soft state, the material stiffness was closer to the brain’s stiffness than traditionally used materials (Grill et al., 2009). The electrode in the compliant state should alter the tissue response by altering the forces on the tissue.
The compliant nanocomposite of poly(vinyl acetate) (PVAc) and tunicate whiskers and stiff PVAc-coated wire implants were implanted into the cortex of rodents for 4 or 8 weeks. After the implant duration, the brain tissue was analyzed via image analysis of tissue immunohistochemistry (IHC). Examination of the IHC labeling indicated the effect of material stiffness on the glia-neuronal response.
Aim 3
Determine the effect of the inflammatory response on cortical neural recordings
It was hypothesized that an increased inflammatory response will decrease quality of cortical neural recordings. The molecule lipopolysaccharide was used to augment the
3 inflammatory response. The tissue response including the inflammatory response separates neurons and axons from the recording site (Biran et al., 2005). Quantification of recording characteristics, impedance, and the tissue response was performed to investigate the effects of inflammatory response on electrode recordings.
Outline
The dissertation work addresses the specific aims and hypotheses discussed above. The work is directed at the goal of investigating interactions with the glial scar further.
Specifically, the two aspects to be examined in more detail are the effect of mechanical stiffness on tissue response and the effect of tissue response on electrode recordings.
Chapter 2 provides significance and background of the work. Chapter 3 addresses Aim 1 and enables a full examination of the effect of substrate stiffness on tissue response, covered in Aim 2. Experiments examining Aim 2 are covered in Chapter 4.
Additionally, examination of Aim 3 is covered in Chapter 5 to investigate the effect of inflammatory tissue response on electrode recordings. Chapter 6 covers a discussion of the implications of the work covered in the dissertation and discusses several broad categories and specific areas for future work for this project and for the field. The appendix provides further details about procedures conducted in the course of this dissertation.
4 CHAPTER 2. INTRODUCTION
2.1. Statement of Problem
Applications for intracortical electrodes scale from the laboratory to the clinical setting. In a laboratory setting, intracortical electrodes have been used to study the function and dysfunction of the brain. The understanding has lead to valuable insights regarding many neurological conditions. Likewise, clinical applications of brain-machine interfaces (BMI) offer the ability to improve the quality of life of many patients afflicted with high level paralysis from spinal cord injury, brain stem stroke, amyotrophic lateral sclerosis, or other conditions. Intracortical electrodes used in BMIs have shown the best promise to control complex movements to improve the quality of life for patients with high-level paralysis.
Intracortical electrodes provide the most detailed control, but limitations with longevity and stability have hindered laboratory and clinical applications. Previous research has hypothesized that the tissue response to the electrodes causes intracortical electrode longevity limitations. These issues hinder wide scale clinical application, and in this work, we examined the tissue response to intracortical electrodes in order to improve future intracortical electrode technologies. While many aspects of the tissue response are important to improving electrode technologies, this work will specifically examine the effect of implant stiffness on the tissue response and the impact of tissue response on electrode recording performance. The research will provide a foundation to improve electrodes and further development of clinical BMI technologies.
5 2.2. Importance of Problem
There is an opportunity to improve quality of life in complete paralysis patients
by creating technologies to translate thoughts into action. Brain-Machine Interfaces or
Brain-Computer Interfaces, BMIs or BCIs, translate a patient’s thoughts to control an external device, be it a computer cursor, a robotic arm, or their own muscles. BMIs can improve the independence and quality of life of patients with spinal cord injury (SCI),
Amyotrophic Lateral Sclerosis (ALS), or other patients with complete paralysis.
The number of spinal cord injuries and ALS patients is small in comparison to other neurological conditions. There are 12,000 new SCI injuries each year in the U.S. and 5,600 new ALS diagnoses each year in the U.S. In comparison, there are 650,000 new stroke survivors each year in the U.S. and 200,000 new cases of seizure or epilepsy disorders each year in the U.S. (Roger et al., 2011; Epilepsy Foundation, 2011; ALS
Association, 2011; National Spinal Cord Injury Statistical Center, 2011). In comparison to the relative size of the population, the fiscal and emotional burden is high. The first year after a high tetraplegia injury the average yearly expenses are about one million dollars, and the lifetime expected costs are between 2.4 and 4.4 million dollars (National
Spinal Cord Injury Statistical Center, 2011). Though patients with high level paralysis are a driver for much research in this area, military research into BMIs is another driver.
Mind control for combat missions is one focus, and BMIs for assistance of injured veterans is another focus. Regardless of the arena, improved BMIs technologies have the opportunity to improve the quality of life of individuals.
Though much of the clinical work regarding intracortical electrodes has been focused on BMIs, work regarding intracortical electrodes is applicable to other
6 neurological conditions. In addition to the 650,000 yearly stroke survivors (Roger et al.,
2011) and 200,00 new cases of seizure or epilepsy (Epilepsy Foundation, 2011), patients
with multiple sclerosis, Alzheimer’s disease, Parkinson’s disease, Huntington’s disease,
traumatic brain injury, and many other neurological conditions form a large patient
population. Intracortical electrodes can be used to understand the development of
conditions and their impact on neural networks (Alves et al., 2005; Buzsáki, 2004;
Ortinski et al., 2010). More robust intracortical electrodes may also allow for treatment of many of these neurological conditions, but electrodes can also be used to understand
the native function of the brain and memory formation (Churchland et al., 2010;
Scherberger et al., 2005).
2.3. Review
2.3.1. Brain Recording Technologies
Several invasive and non-invasive technologies have been employed to record the activity of the brain and improve understanding of the function of the brain. A general recording setup consists of a detector (electrical, chemical, magnetic, or optical sensor), a conduit or channel in which the detector’s signal needs to be conducted through (radio waves, wire, optical fiber), and a decoder (a computer or signal processor). Additionally, when control of a device is desired with a brain machine interface (BMI), there will be an external device, an effector, which can take the form of a computer cursor, robotic arm, or other device to produce action. The external effector may require additional equipment or software to decode/encode signals to control the external device.
2.3.1.1. Electrical Versus Other Recording Paradigms
The majority of recording research has occurred in the electrical regime, and the classical setup is a metal electrode as the detector with an insulated wired cable as the
7 conduit to the computer/signal processor though wireless transmission is becoming more common (Wise et al., 2004; Mohseni et al., 2005; Martel et al., 2001; Kipke, 2004). The computer/signal processor may create a signal to manipulate a computer cursor, a mechanical arm, or other device (Hochberg et al., 2006; Taylor et al., 2002; Velliste et al., 2008).
There are several reasons that the electrical regime has been employed in the clinic instead of other technologies. Chemical recording can have issues with chemical specificity and depending on the technology used, has low temporal resolution, shows only relative changes instead of absolute, exhibits breakdown of the material to analyze a chemical, or fails because of clogging of microdialysis ports (Bakker, 2004; Kennedy,
2010). Several of these deficiencies result in a short usage life of these technologies or poor performance tracking. Non-invasive technologies such as positron emission tomography (PET) can also provide chemical information. The low temporal resolution and the size and expense for supporting hardware make it difficult to employ in an everyday setting. Similar problems exist for magnetic technologies that use functional
Magnetic Resonance Imaging (fMRI) to extract thoughts from the brain. Another type of magnetic based technology is magnetoencephalography (MEG) using Superconducting
Quantum Interference Devices (SQUIDs). This technology requires a great deal of external hardware but does not suffer from temporal issues. MEG technology suffers from spatial resolution issues at depths further from the skull surface (Hillebrand and
Barnes, 2002). Another technology using optics to record brain signals is limited by the need for the repeated introduction of exogenous agents to transmute chemical or electrical activity into an optical signal.
8 Due to the limitations of these other technologies, the discussion of brain recording will be concerned with the classical electrical paradigm (metal electrode, wired or wireless transmission, and computer/signal processor). The paradigm has been the most researched, and it offers the current best path towards clinical implementation of a
BMI (Simeral et al., 2011a; Kim et al., 2008; Hochberg et al., 2006; Kennedy et al.,
2004; Kennedy et al., 2000; Simeral et al., 2011b).
2.3.1.2. Electrode Location and Recording Types
There are tradeoffs between different electrode technologies. Typical electrode types include intracortical electrodes, electrocorticogram (ECoG) grids, bone screws, and electroencephalography (EEG) electrodes (Figure 1).
9
Figure 1. Diagram illustrating several types of electrodes used in BMIs (wires or transmission packaging not pictured). The brain is pictured as the curved gray region, and the skull is pictured as the curved surface above it (bone is pictured as dotted surface). Electroencephalogram (EEG) electrodes are placed on the scalp of the patient. The bone screw is placed in the bone of the skull. The electrocorticogram (ECoG) is placed on the surface of the brain underneath the dura (dura and skull opening not pictured). Intracortical electrodes are placed into the brain tissue. Two larger scale versions of intracortical electrodes are pictured. The Neuronexus/Michigan planar silicon electrode is depicted on the left while a photograph of a microwire array is pictured on the right (33 µm diameter wire without insulation, 3mm long wire).
There is an inherent tradeoff between the invasiveness of the electrode into the brain and the spatial selectivity of the electrode technology. Less invasive electrodes are less spatially selective. Although clinical and research experience is varied for each electrode type, the body’s response to the electrode affects electrode performance greater with increasing invasiveness. The typical issue with EEGs is from donning and doffing and abrasions from repeatedly placing electrodes on the head (Ferree et al., 2001).
10 Encapsulation occurs with any device implanted into the body. The degradation is more significant for intracortical electrodes since they record from the smallest region of tissue.
Therefore, they are most sensitive to increases in distance and impedance due to encapsulation. Similarly, ECoG grids can elicit a response, but research has shown that the tissue impedance stabilizes within the first week of the implant and remains stable until 18 weeks (Henle et al., 2011). The same researchers also show that the response is limited to the leptomeninges, and neurons in the cortex were not significantly affected by the implant (Henle et al., 2011). In certain cases, the usage of ECoG grids results in issues of infection, bleeding, or hemorrhaging (Simeral et al., 2011b). It is believed that the negative response is due, in part, to the large area of skull removed to enable implantation. Regardless, the potential of infections and complications is increased when the skin and/or skull are opened to allow placement of devices.
The functional impact of reduced spatial selectivity is still being researched, and many paradigms exist to extract the highest level of control from the different quality and quantity of signals from each electrode type. Going from least invasive and least selective to most: EEG, bone screws, ECoG, and intracortical electrodes. Additionally, intracortical electrodes can record local field potential (LFP) recordings, multiunit recordings, and single unit recordings in increasing spatial selectivity. It should be noted that in addition to spatial selectivity, each electrode and recording type signifies a slightly different component of neural activity. For example, local field potentials represent activity from a collection of dendrites while single units are recorded from the neural soma (Mitzdorf, 1985).
11 2.3.1.3. Information Transfer Rates of Electrodes and Other Assistive Technologies
Though a comprehensive study comparing all modalities available to SCI patients
has not been performed, previous research has often quantified the quality of an interface
in terms of information transfer rate (bits/sec). In addition to the lack of a comprehensive
review, slightly different definitions for information transfer rate have been used
including Fitt’s Law and Wolpaw’s definition (Fitts, 1954; Wolpaw et al., 1998; Kronegg
et al., 2005). Depending on the application, the bit rate may also not be the best measure
as it does not measure accuracy which would be an important component of complex
motor neural prosthetics (Kronegg et al., 2005). In an effort to compare assistive
technologies for patients with partial or complete paralysis, Table 1 was compiled.
For patients that lack all voluntary movement, BMIs offer the only method to
interact with the outside world as they lack speech and the ability to control eye or other
movements. In the case of patients that still possess some movement control, non-
invasive technologies such as sip-puff, head movement, tongue systems, speech
recognition, muscle activity (EMG), tooth click, or eye movement/electrooculography
(EOG) are available. Certain movements such as eye movement may be reduced in
certain patient populations, e.g. patients with ALS (Birbaumer, 2006).
Technologies using the tongue to control a computer cursor have been limited due
to lack of acceptance of a mouthpiece, but a magnetically-based and unobtrusive tongue device has addressed some of these issues (Huo and Ghovanloo, 2010). With the recent improvement in price and quality of eye tracking devices, several studies have examined eye gaze to control devices. The typical problem with eye gaze devices is that there is a difficulty determining a click. Different methods have been developed to function as a
12 click including using face muscle activation, eye dwell time, or the spacebar. The
spacebar option is not possible in many patients with high-level paralysis. These eye movement and gaze technologies have achieved performance comparable, if not better, to the mouse. It is notable that the similar gaze techniques can yield very different numbers, and these differences are somewhat attributable to the farther targets in the Surakka et al. study (Surakka et al., 2004) that translated to an increase in information transfer rate because of the quickness of eye movement over far distances. Since these tests were performed in uninjured patients, it is unknown how these technologies would perform in patients with paralysis. Studies have shown that other assistive devices show a reduced performance in patients with SCI (Simpson et al., 2010).
Table 1. Review of current assistive technologies and information transfer rate (ITR) of those technologies. Assistive Technology Bits/sec Reference EEG 0.42 (Wolpaw et al., 2002). Tooth click device 0.46 (Simpson et al., 2008) Sip puff 0.59 (Simpson et al., 2008) EEG and ECoG 0.5-1.3 (Ryu and Shenoy, 2009)
EMG of head and neck muscles 1.0 (Williams and Kirsch, 2008) Head tracking with tooth click 1.52 (Simpson et al., 2010). Magnetically-based tongue device 1.58 (Huo and Ghovanloo, 2010) Electrooculography (uninjured) 3.3 (Kherlopian et al., 2006) Eye gaze with EMG clicking 4.9 (Mateo et al., 2008) (uninjured) Intracortical electrodes 4.8-5.5 (Taylor et al., 2003)
Intracortical electrodes 6.5 (Ryu and Shenoy, 2009) Mouse (uninjured) 6.9 (Simpson et al., 2010) Keyboard (uninjured) 10 (Ryu and Shenoy, 2009) Intracortical electrodes 10 (Ryu and Shenoy, 2009) (theoretical max) Eye gaze with EMG clicking (uninjured) 12.6 (Surakka et al., 2004)
13 In comparison to electrodes translating thoughts to action, some of these assistive
technologies can perform better. It is unknown whether certain technologies would perform better as a motor prosthesis or communication prosthesis. A motor prosthesis uses speed and path, but a communication prosthesis uses a “keypress” or endpoint.
While intracortical electrodes may outperform EEG, certain assistive technologies outperform intracortical electrodes and indicate that current intracortical electrode technologies are only advantageous to patients with severe paralysis. Significant improvements will be needed to intracortical technologies to enable their usage in other patient populations or applications.
2.3.1.4. Intracortical Electrode Technologies
Intracortical electrodes will be the focus of this review and research. Though the most invasive technology, they have been shown to provide the most selectivity and highest information transfer rate of brain electrodes. The increased information transfer rate would be needed to control complex systems. The clearest demonstration of these advantages is in non human primate trials where a primate can feed himself by controlling an arm with four degrees of freedom (Velliste et al., 2008).
At present there are two types of electrode array architectures employed, sites at the tip and sites along the shank (Table 2). Additionally, the materials used in these arrays have usually consisted of a four different types a) insulated metal microwire
(Williams et al., 1999), b) microfabricated silicon structures with of metalized sites
(Kipke et al., 2003), c) microfabricated structures consisting of a ceramic substrate with metalized sites (Moxon et al., 2004), d) microfabricated structures consisting of a polymer based substrate with metalized sites (Rousche et al., 2001). Novel technologies
14 are often a variation on these types. One standard variation is an additional chemical coating on the substrate or recording site (Ludwig et al., 2006). Though some have discussed polymers as conductors, their use has been limited to coatings on metal sites leaving metal and polysilicon conducting along the length of the electrode.
Table 2. Summary of intracortical electrode types. Electrode Topology Example Electrodes Utah/Cyberkinetics/Blackrock (Jones et al., 1992) Site at the tip of the electrode Microwire (Schmidt et al., 1988) Michigan/Neuronexus (Drake et al., 1988) Sites along shank, planar Moxon/Ceramic (Moxon et al., 2004)
2.3.1.5. Importance of Intracortical Electrode Recording Site Size
Another tradeoff, important for intracortical arrays, is electrode size versus electrode impedance. Previous modeling has shown that a smaller electrode with the same material will increase the impedance of the electrode, and hence, and thermal noise
(Lempka et al., 2006). The same modeling study showed that smaller electrodes are able to record a larger peak voltage of the signal when placed the same distance away from the neuronal soma (Lempka et al., 2006). For a given current density, the smaller electrode with higher impedance would record a larger voltage because voltage equals current times impedance. However, it is also believed that increased impedance from smaller electrodes can lead to a decrease in recording amplitude due to capacitive shunting along the length of the electrode, but the capacitive effect has not been confirmed in modeling or in vivo (Moffitt and McIntyre, 2005; Lempka et al., 2006). Smaller electrodes also record from a much smaller volume because the recorded voltage of a neural dipole decreases with the square of the distance between electrode and dipole.
15 These factors indicate that smaller electrodes have the benefit of recording less
neural “noise” and functionally increasing the signal to noise ratio (Marzullo, 2008). At a
certain size, decreasing the size of the electrode is no longer feasible since the thermal
noise would increase, decreasing the signal to noise ratio. Previous research has
indicated that the lower limit for site size is 380 µm2, but this value does not include the capacitive shunting effect and has not been validated in vivo (Lempka et al., 2006).
Therefore, single unit recording electrodes do not necessarily need exhibit a high impedance, though most do. The total electrode including size, material, and surface site roughness determines recording ability. Previous research and anecdotal experience suggest electrode site sizes from 380-1000 µm2 (Lempka et al., 2006; Marzullo, 2008),
though some have successfully used microwires with site sizes twice as large (Nicolelis
et al., 2003).
2.3.1.6. Types and Characteristics of Intracortical Electrode Recordings
The placement and construction of the electrode determines the spatial range and
type of signals recorded. Additionally, frequency filtering plays a role in the spatial
range and type of signal recorded. Filtering allows for the recording of a signal that can
be composed from one to thousands of neurons (Katzner et al., 2009). The majority of
previous work using intracortical electrodes recorded extracellular potentials from single
neurons, and frequency filtering is usually centered around the fundamental frequency of
single unit action potentials, 1kHz. The same electrode and placement can be used to
record a broad spatial range of dendritic activity in the form of local field potentials,
LFPs, over the frequency range less than 100-200 Hz. The LFPs can be from sources up to 250 µm from the electrode (Katzner et al., 2009). In comparison, recordings from
16 single units with an intracortical wire tetrode can theoretically record ~100s neurons within 50 µm of the electrode or ~1000 neurons within 140 µm of the electrode (Henze et al., 2000).
In practice, the theoretical performance is not achieved since current technologies lack the ability to distinguish between many signals. Neurons further from the electrode provide fainter recordings. As the distance increases from the electrode, the number of neurons increases as well. As the number of neurons increases, the uncorrelated signals deconstructively interfere with each other. These factors imply that the indistinguishable signals are further away from the electrode. These indistinguishable neural signals are classified as noise or “neural hash”, a significant contributor to the overall noise of the system.
2.3.1.7. Performance and Limitations of Intracortical Electrodes
Recent research in patients with complete paralysis has shown that BMIs are feasible long after the onset of paralysis (Hochberg et al., 2006). The cursor control and click performance demonstrated in humans does not achieve the same four degrees of control as in animals, but intracortical electrodes have served as an effective way to control external effectors (Velliste et al., 2008; Taylor et al., 2002; Nicolelis et al., 2003).
2D control of a computer cursor needed 18-29 neurons, and approximately 39 directionally-tuned neurons allowed for 3D control by a non human primate (Simeral et al., 2011b; Taylor et al., 2003). Additionally, 3D control with the addition of a gripper (4 degrees of freedom) used 60-120 neurons for control (Velliste et al., 2008). More complex movements will require more neurons. The possibility of translating these
17 technologies to humans offers the hope to improve quality of life, but issues with array technologies have hindered clinical translation.
The most well known human study to date uses the Utah array, originally developed in the early 1990’s (Jones et al., 1992; Hochberg et al., 2006). Other technologies have been developed since, but most suffer from the same problems described by Dr. Schwartz in his 2004 review:
“The largest remaining obstacle to the successful implementation of cortical neural prostheses is the chronic recording electrode… an informal survey (A. Schwartz, personal communication) of the laboratories using CNP [cortical neural prostheses] suggests that, on average, a chronic electrode implanted in monkey cortex has only a 40% to 60% chance of recording unit activity. Although each lab has an example of an all-star animal with good recordings for multiple years, electrode recordings usually deteriorate after several months.”(Schwartz, 2004)
Though the technology can be a significant help to patients, the long-term performance of these electrodes has prevented broad application in the clinical setting.
To enable widespread clinical implementation of BCI intracortical electrodes, the longevity of electrodes will need to be improved to last at least a decade. What causes the deterioration of performance as noted by Dr. Schwartz? One hypothesis is related to the insertion of intracortical electrodes and the subsequent cascade of events. The resulting tissue response encapsulates and creates a glial scar around the electrode decreasing neuronal density, proximity, and function that detriment recording quality.
The following sections will go through the insertion, acute, and chronic stages of the response to implanted electrodes to provide a foundation for understanding the response to implanted electrodes and the effect on electrode recordings.
18 2.3.2. Trauma from Microelectrode Insertion
During microelectrode insertion into the brain, the electrode will damage
(Bjornsson et al., 2006) the neurovascular unit consisting of the microvasculature,
astrocytes, other glial cells, and neurons (Abbott et al., 2006; Abbott, 2002). Damage to
the blood vessels results in microhemorrhaging (Turner et al., 1999; Edell et al., 1992;
Bjornsson et al., 2006) and release of several plasma components, including blood-borne macrophages. The macrophages are normally isolated from cortical tissue by the blood- brain-barrier (Figure 2). The trauma initiates an acute response that leads to the chronic response to the electrode.
Current methods are unable to avoid insertion damage due to the density of the vasculature and the size of current intracortical electrode technology. In examining the near surface vasculature with 2-photon microscopy, researchers have shown that it is possible to minimize insertion trauma by avoiding surface, and consequently, underlying vasculature (Kozai et al., 2010b). To avoid insertion damage, smaller devices have been developed. Since neurons are usually within 15 µm of a blood vessel, it is inherently difficult to get an electrode close to record a quality single neuron or unit without causing trauma (Tsai et al., 2009).
19
Figure 2. Diagram of trauma and acute damage to brain parenchyma caused by electrode insertion into brain. Minutes after insertion, blood vessels are damaged and blood enters the brain parenchyma. Neurons are damaged during insertion event. Activation of microglia and infiltration of blood-borne macrophages occurs within hours with continued neurodegeneration. After a day(s), the activation of macrophages and microglial continues with the leaky blood brain barrier in conjunction with the presence of reactive astrocytes. Further neurodegeneration continues.
The issue of insertion damage will only increase since improved control will
require more signals from more implanted electrodes in electrode arrays. Current array
technologies place the electrodes in a fixed-geometrical pattern that allows one to introduce more electrodes, but also causes more insertion trauma. Since each electrode
20 in the array cannot be independently moved, insertion damage cannot be totally avoided
by microscope-based positioning.
Future arrays might deviate from the classic fixed-geometry format and reduce tissue damage. Recent work at the University of Michigan using electroactive polymers to steer an electrode around blood vessels is motivated by minimizing vasculature damage (D. Kipke, personal communication). Two other tactics that have gained some attention but have not grown past initial experiments are: 1) to grow conductive polymers in situ (Richardson-Burns et al., 2007) and 2) to thread electrodes through blood vessels to avoid trauma (Llinas et al., 2005). Both methods present other issues that need to be addressed, including polymer toxicity and thrombosis, respectively. Additionally, further research needs to investigate the ability to record quality signals, the bioresponse to the electrode, and the performance over the long-term.
2.3.3. Tissue Response
After the acute insertion event, a tissue response is initiated to contain the injury and repair the damage incurred. The tissue response impacts neurons in two separate but intertwined processes: neurodegeneration and neuroinflammation (Carson et al., 2006).
These two processes are an intricate balance of several chemokine and cytokine pathways that are still not completely understood, but result in neuronal dieback from the electrode and glial scar encapsulation of the device (Figure 3, right).
The quality of intracortical recordings has been implicated to be directly related to the proximity of neurons, the functioning of the neuron network, and the impedance between the electrode and neuron (Biran et al., 2005; Johnson et al., 2004; McConnell,
21 2008). Therefore, research of the tissue response is a significant part of BMI and
intracortical electrode development. Research has classified the tissue response into the acute phase and the chronic phase where the acute phase lasts for the first 1-2 weeks
(Renault-Mihara et al., 2008; McConnell et al., 2007a).
2.3.3.1 Acute Phase of Injury
An electrode insertion causes damage to blood vessels that allows foreign plasma constituents into the brain. The disturbance in the chemical milieu initiates many events, but the initial markers of the response are localized edema and activation of “resting” microglial cells and infiltration of macrophages released from the damaged vasculature
(Figure 2) (Kettenmann, 2007). Denatured proteins adhere to the electrode surface and cause the further release of chemokines and cytokines, including glutamate and adenosine triphosphate (ATP). The chemicals put many neurons in a distressed state (Ortinski et al., 2010; Anderson et al., 2008; Raghavendra Rao et al., 2000). The positive feedback loop activates astrocytes and many other cell types to limit the extent of the insult
(Sofroniew, 2005; Faulkner et al., 2004).
2.3.3.2. Chronic Phase of Injury
2.3.3.2.1. Chronically Activated Inflammatory Cells (Macrophages/Microglia)
As the brain tries to remove the foreign implant, the activated microglia and macrophages (activated inflammatory cells) try to clean up and remove the implant while astrocytes and other extracellular matrix cells try to rebuild the blood brain barrier around the implant. The consequences of the neuroinflammation response are dependent on several factors and have been implicated to be both neuroprotective and neurodegenerative (Lenzlinger et al., 2001; Carson et al., 2006).
22 Over time periods from one to 7 days, microglia have been shown to be
“alternatively” activated (M2 state). The M2 state has shown neuroprotective effects.
The state where microglia are neurodegenerative has been named “classically” activated
(M1 state). The M1 state is associated with the detrimental effects of activated macrophages and microglia (Kigerl et al., 2009). When the microglia and macrophages are unable to remove the damage or device, these “frustrated” cells release cytokines and chemokines that lead to neurodegeneration (Anderson et al., 2008).
Figure 3. Representative diagram illustrating a lack of response to an implanted electrode (“No Response”) and the typical chronic response to an indwelling intracortical electrode. Fibroblasts and a new extracellular matrix (ECM) are created around the electrode that is surrounded by phagocytic macrophages and microglia. Outside of the phagocytic macrophages/microglia, reactive astrocytes form a glial scar. A decreasing gradient of chondroitin sulfate proteoglycan (CSPG) is around the electrode. Blood vessels have been damaged and have a leaky blood brain barrier (BBB). The density of neurons is decreased around the implanted electrode.
23 Several studies have indicated a large number of activated macrophage and microglia cells on the implant surface long after the initial wound healing response is complete (Szarowski et al., 2003; McConnell et al., 2009). The studies also showed that continued activation is correlated with decreased neural density (Szarowski et al., 2003;
McConnell et al., 2009). Szarowski et al. suggest the microglia response would last as long as the material remained in the brain tissue, indicated by continued presence of activated microglia in the tight cellular sheet surrounding explanted electrodes implanted for up to twelve weeks (Szarowski et al., 2003). Additionally, Biran et al. (Biran et al.,
2005) have demonstrated that a “stab” (insertion and immediate removal) of a neural implant shows greater neuronal density around implant sites than chronic implants, suggesting repair can happen in the absence of a chronic agitator such as a implanted electrode.
Some researchers have tried to control the microglial response by developing coatings or systemically administering dexamethasone, minocycline, and flavopiridol among others (Rennaker et al., 2007; Zhong and Bellamkonda, 2007; Zhong et al., 2005;
Purcell et al., 2009). Dexamethasone coatings were able to improve neurofilament around electrodes up to 10%. The coatings decreased macrophage activation (ED1) at the 1 week time point, but the effect did not last at 4 weeks. The dexamethasone coated electrodes still incurred some damage, but the effect on recordings was not assessed. Other researchers examined the systemic effect of dexamethasone on recordings. It was shown that dexamethasone could improve recordings, but the recordings still decreased over time (Anderson, 2008). Usage of minocycline yielded similar results, and flavopiridol did not change the quality of recordings or quantity of neural or non-neural cells. While
24 coatings allow for direct, non-systemic application, the duration of drug elution can be limited.
Other researchers have down regulated macrophages/microglia with certain materials such as electrospun polycaprolactone scaffolds. The activation levels are lesser than a stab by 3 weeks and return to similar levels as no insertion by 9 weeks (Nisbet et al., 2009). Though the polycaprolactone study did not quantify the neuronal density or perform electrode recordings, the results indicate that certain materials in a specific topology can modify the tissue response. Regardless of the implant, an insertion causes a complex inflammatory cascade to clean up debris from damaged cells, repair the vasculature, and reestablish the blood brain barrier.
2.3.3.2.2. Blood Brain Barrier (BBB)
It is hypothesized that a portion of the inability of the tissue to repair itself or deconstruct the glial scar is due to the chronic disruption and dysfunction of the blood brain barrier. A leaky BBB compromises the immune privilege of the brain.
Macrophages and other chemicals not normally found in the central nervous system tissue are allowed to enter the brain tissue. The pathway that leads to a leaky BBB has not been determined. Many of the chemicals produced in response to implanted electrode have been shown to contribute to the dysfunction of the BBB including chemicals such as
ATP, TGF!, IL-6, and TNF" (Abbott et al., 2006). Therefore, an implanted electrode causes an initial and sustained response through the interruption of the BBB. Previous research has indicated a leaky BBB around an implanted electrode up to 12 weeks
(Winslow et al., 2010).
25 A chronic influx of foreign materials and macrophages would cause a continued activation of chemical signaling pathways that would result in further glial scar formation and neuronal degeneration while limiting the regrowth of neuronal processes.
Additionally, the vasculature and BBB alters the efflux of chemicals that have detrimental effects in high concentrations. Proper function of the BBB to help control chemical concentrations is key to neuronal survival and function (Abbott et al., 2006).
Activated macrophages have been implicated in neuronal degradation, and the infiltration of these cells through the BBB is thought to be an issue with an impaired
BBB. Conclusive studies have not been able to determine whether chronically activated inflammatory cells are activated microglia or infiltrating macrophages. Current immunohistochemistry (IHC) techniques are unable to distinguish between macrophages and microglia though experiments using chimera mice have been used distinguish between the two in mice (Tanaka et al., 2003). Notably, in the Tanaka et al. study, the researchers saw the extrinsic macrophages travel outside of the infarct area further than expected.
One portion of the BBB is the boundary formed by meningeal cells forming the exterior layers of the central nervous system encapsulating the brain, spinal cord, and cerebral spinal fluid (CSF). An implanted electrode perforates the meninges. The extent of meningeal repair after electrode insertion is still unknown since there are no specific immunohistochemical markers for meningeal cells.
One preliminary study indirectly examined the meningeal response by examining the immunohistological response among a flexibly-tethered probe, a rigidly-tethered
26 probe, and untethered probe. Since there were similarities between both tethered probes
and differences between the untethered and tethered probes. Specifically, the neuronal
loss and non-neuronal cell counts were similar between tethered probe types but improved with the untethered probe type. The researchers suggested that meningeal growth down the tether of the electrode was the source of the differences. Since there are no meningeal histological labels, the results cannot be confirmed (Subbaroyan, 2007).
The research suggested that the glial scar and chronic tissue response to an electrode is
the attempt to reestablish a robust blood brain barrier.
2.3.3.2.3. Astrocytes
Astrocytes directly interact with cells composing the BBB, and astrocytes are another significant cell type involved in the glial scar response. Astrocytes extend endfeet to contact pericytes, the BBB, and neurons, and the astrocytic response separates the electrode from the brain parenchyma. Astrocytes also buffer neurotransmitters and chemicals to regulate levels (Abbott et al., 2006).
Traditionally, reactive astrocyte proliferation has been correlated with a decrease in neural density, and consequently, astrocytes have been assumed to be neurodegenerative. Additionally, reactive astrocytes have been viewed as detrimental for recordings as they have been shown to be a diffusion limiter and barrier to remylination
(Roitbak and Sykova, 1999; Fawcett and Asher, 1999). Reactive astrocytes increase their expression of Glial Fibrillary Acidic Protein (GFAP), and change into a more stellate morphology as they encircle and shield the implant from the brain parenchyma.
27 Though reactive astrocytes have classically been seen as detrimental, recent
research has demonstrated positive aspects of the astrocytic response in limiting the
expanse of neural injury. Transgenic animals without GFAP were shown to have worse
functional deficits after injury (Sofroniew, 2005). Another function of astrocytes is to
buffer excessive concentrations. Excessive concentrations of glutamate have been seen
to be neurotoxic (Salinska et al., 2005), and it has been shown that astrocytes can regulate
levels of glutamate in a neuroprotective manner (Liang et al., 2008). Previous research
has also indicated that the ability of astrocytes to regulate glutamate levels is impaired
when astrocytes become activated (Ortinski et al., 2010). Therefore, astrocytes play an
important part in response to an implanted electrode.
While astrocytes are important on their own, there are additional important
feedback mechanisms between microglia and astrocytes. The response mediated by
cytokines and chemokines can increase production of chemicals that have a vast array of
effects on glial and neural cells. Positive feedback loops can lead to an exponential and
quick response, but the magnitude of the final response may limit repair of tissue after the
initial quick response. For example, a positive feedback loop between microglia and
astrocytes has been shown to be important in the production of proteoglycans that have
shown to limit axonal growth (Yin et al., 2009). While feedback cycles implicate both
microglia and astrocytes in production of chemicals, it has been shown that extracellular
matrix components, such as fibronectin and chondroitin sulfate proteoglycans (CSPGs)
are primarily made by astrocytes (Yin et al., 2009; McKeon et al., 1999; Shearer et al.,
2003).
28 2.3.3.2.4. Chemokines and Cytokines
Involved in much of the response to the electrode, chemokines and cytokines are
released by astrocytes, macrophages, microglia, and other cells. While some cytokines
such as neural growth factor (NGF) have been shown to promote neural cell migration
and proliferation (Elkabes et al., 1996; Nakajima et al., 2001), chemicals released by
“classically” activated microglia (M1) have been shown to be neurodegenerative. Biran
et al. have shown that microglia cells isolated on the surface of explanted cortical
electrodes release monocyte chemotactic protein-1 (MCP-1) and tumor necrosis factor-!
(TNF!) when incubated in vitro (Biran et al., 2005). TNF! has been shown to lead to neurodegeneration in vitro and in vivo (Munoz-Fernandez and Fresno, 1998). A varied list of chemicals plays a role in controlling the response, such as TNF" and IL-1! to simple molecules such as Ca2+ and ATP (Koizumi et al., 2007; Salinska et al., 2005;
McConnell, 2008). An electrode insertion will cause a vast array of chemical signaling to
direct cell differentiation, cell proliferation, and many other cellular responses to control
the extent of trauma and mediate the repair process. Microglia and other cells create a
positive feedback cycle of events that will direct the nervous system’s response. After
initial trauma, chemical signaling will continue to play a role as debris cleanup and tissue
remodeling occur.
2.3.3.2.5. Other Cell Types and Factors Associated in Response
Oligodendrocytes/myelin and extracellular matrix cells are important cells in the
brain that are associated with the response to an implanted electrode. Other factors like
proteoglycans and intermediate filaments can be expressed by these cells or other cell
types. Chemicals such as NoGo, myelin associated glycoprotein (MAG), and chondroitin
29 sulfate proteoglycans (CSPGs) have been shown to be a chemical blockade to regrowth.
Others have advanced a slightly different view, indicating NG2+ cells, a proteoglycan, exhibiting molecules such as Vimentin, an intermediate filament, provide a chemical oasis that stabilize neuronal axons preventing further axonal dieback (Silver and Miller,
2004; Busch et al., 2010).
2.3.3.2.6. Summary of Chronic Response
Refocusing on the interplay between neurodegeneration and neuroinflammation, the inflammatory encapsulation tissue composed of astrocytes, microglial, extracellular matrix proteins/cells, and other cell types have multiple beneficial and negative effects.
For example, astrocytes have been shown to limit the extent of tissue degeneration after an injury (Sofroniew, 2005), but chronic astrocyte injury has been shown to be neurodegenerative (Abbott et al., 2006). Microglia and astrocytes coat the surface of electrodes and create a barrier between the neurons and the electrode (Szarowski et al.,
2003; Grill and Mortimer, 1994; Turner et al., 1999). The glial scar creates an overall inhospitable climate for neurons and regrowth (Szarowski et al., 2003; Turner et al.,
1999; Zhong and Bellamkonda, 2007; Lu et al., 2007; Hampton et al., 2004; Fawcett and
Asher, 1999). Neuronal density and proximity is decreased and tissue resistivity is increased while modifying neuronal network and BBB function (Winslow et al., 2010;
McConnell, 2008). The acute inflammatory response is beneficial in promoting vascular healing, removing cellular debris, and containing injury. The chronic response is accompanied by a loss of neurons near the probe and excessive encapsulation that may limit electrode efficacy. Whether electrical impedance, neuronal distance, or neuronal
30 function is ultimately the root cause for degradation of long term performance of electrodes is uncertain.
2.3.3.3. Mechanical effects
Mechanical signaling is another mechanism that occurs in response to an implanted electrode. Traditional materials used for neural probes include tungsten, platinum, steel, silicon, glass, and ceramic that are all near 9 orders of magnitude stiffer than the brain tissue (Young’s modulus in the range of 100s of GPa versus kPa). While stiff neuronal probes facilitate insertion, an insertion of any probe causes damage initiating a signaling cascade that initiates a cascade of events leading to glial scar formation and neuronal dieback. In vitro and in vivo studies have shown cell differentiation and proliferation are modified by mechanical factors such as shape, porosity, surface roughness, and stiffness (Szarowski et al., 2003; Engler et al., 2006;
Georges et al., 2006; Leach et al., 2007; Zaman et al., 2006; Nisbet et al., 2009; Li et al.,
2007). Mechanical effects can be categorized into micro or macro effects. Micro effects are where the cells sense the modulus of the material. Macro effects are where the cell responds to an applied force. In many cases, the material modulus, configuration, and geometry determine the force exerted on the tissue and cause a response by the cells.
2.3.3.3.1. Micro Mechanical Effects on Cells
Many in vitro experiments have been performed examining the micro signaling, defined as mechanotransduction. Previous experiments have shown that material stiffness affects stem cell differentiation, cell proliferation, tumor migration, neuron growth cone development, oligodendrocyte spreading, and neuron branching among
31 other effects (Engler et al., 2006; Georges et al., 2006; Zaman et al., 2006; Hampton et al., 2004; Flanagan et al., 2002).
2.3.3.3.2. Macro Mechanical Effects on Tissue
The other effects in comparison to micro mechanical are macro mechanical effects. The macro pathway is concerned with the force exerted on the tissue that may damage tissue or cells. A clear demonstration of the response of cells to a force is an in vitro experiment showing a calcium wave spreading through astrocytes in reaction to an applied mechanical force (Ostrow and Sachs, 2005). Other in vitro research has shown that astrocytes become reactive based on the magnitude of shear force exerted on cells
(Cullen et al., 2007). Since material modulus, cross sectional area, and shape determine the force of a bending probe, research has been directed at two methods to reduce the force electrodes exert on tissue: 1) reduction of cross sectional area and shape and 2) reduction of material modulus.
Both methods try to reduce forces that the probe exerts on tissue. The standard practice of tethering an electrode to the skull with dental cement and bone screws transfers forces to brain tissue. Previous in vivo experiments have compared tethered and untethered implants showing that the response to the tethered implants has less of a glial scar and a greater neural density around untethered implants (Biran et al., 2007). The tethered versus untethered experiment is the most direct method in published literature of examining the effect of force on tissue response in an in vivo setting. Another manner where forces are transferred to brain tissue is when the animal quickly moves its head.
Since the brain can move within the skull, forces can occur. Softer electrodes reduces the forces of an electrode, as shown in computer modeling (Subbaroyan et al., 2005).
32 Additionally, forces come from micromotion. Micromotion from cardiac factors cause as
much as 1-3 µm of movement in rats, and pulmonary factors cause 2-25 µm in movement
(Gilletti and Muthuswamy, 2006; Subbaroyan et al., 2005).
Some researchers have been focused at creating very small electrodes to minimize
forces. Research from the Kipke lab has examined feature size. Their work has shown a
greater neural density and a lower non-neural density with reduced device size. Whether
this is because of a reduction in force or a reduction in contact area between tissue and
implant was not examined (Seymour and Kipke, 2007). Additionally, the Kipke lab has
created 4 µm wide carbon fibers to record single unit extracellular action potentials
(Kozai et al., 2010a). Preliminary findings have shown that the ultra-small fibers have
been shown to reduce the tissue response. Creating ultra small electrodes has its own
problems since decreasing the size increases impedance to the point where recordings
degrade because of a significant thermal noise contribution. To solve the large
impedance problem, additional processing steps can be utilized to coat the electrode to
decrease the impedance and reduce thermal noise. Unfortunately, the additional
processing steps likely lead to further instability in the electrode as current electrode
treatments and insulation deteriorate over time (Sanchez et al., 2006; Yamato et al.,
1995). Similar in vivo research investigated an open-lattice structure for implants
(Skousen et al., 2010). The research has implied beneficial effects of the open lattice
structure, but it is uncertain whether the positive effects of the open-lattice structure are from mechanical, diffusion, or surface chemistry effects.
The other method to reduce force on the brain tissue is to utilize substrates with a lower Young’s Modulus. In silico modeling has shown that a material with a lower
33 Young’s Modulus (less stiff) provides significantly less strain on the surrounding brain
tissue (Subbaroyan et al., 2005). Other research has examined novel electrode substrates
and substrate coatings from materials that are more compliant than materials traditionally
used to create electrodes. The materials include polyimide, polydimethylsiloxane
(PDMS), and parylene. These materials have not been able to reduce glial scar formation,
but many of these materials still have moduli 6 orders of magnitude larger than that of the
brain. Some introduce different surface chemistries that may detriment the tissue-
material interaction (Lee et al., 2004; Nikles et al., 2003; Wester et al., 2009; Takeuchi et
al., 2005).
2.3.3.3.3. Pathways for Mechanical Effects on Cells
Most of the micro effects are attributed to internal cytoskeletal reorganization and
a linkage of the cytoskeleton to the cell nucleus and protein synthesis (Ingber, 2006;
Peyton et al., 2007). Mechanotransduction in cells that are better known to respond to
mechanical stimuli have been well studied, and a combination of stretch-activated ion channels, integrins, growth factor receptors, myosin motors, cytoskeletal filaments, and several other structures have shown to be involved with mechanotransduction responses.
The forces associated with mechanotransduction can come in different varieties.
Some forces are applied to cells such as a flow passing through a blood vessel. Other
forces are from the network of cell attachments to balance cell forces on the cytoskeleton.
Every mechanotransduction pathway has a slightly different mechanism, but the
translation of mechanical stimulus into cell action usually transpires through the change
in intracellular ion concentration. In the case of hair cells in the ear, mechanical forces
34 translate to an influx of calcium ions. Mechanical activation activates other cells with a similar influx of calcium ions. Calcium dependent proteins then phosphorylate their substrates including myosin II, a protein involved in cytoskeletal tension (Clark et al.,
2007). Calcium affects the actomyosin complex that governs cell adhesion and migration that affect cell function in many ways.
2.3.3.3.4. Mechanical Effects on Device and Packaging
Another mechanical factor is the packaging of the neural device and its possible degradation over time. The publishing rate regarding packaging and intracortical electrode development has been quite low. The total package consists of the electrode, packaging, as well as any interconnect to allow for the electrode to be easily interfaced to an external system (wired or wirelessly). Issues with the total package can occur along the whole span, but the forces on a wired connector may also transmit forces to the electrode in the brain. It is possible that the meninges utilize the connector and/or electrode shank as a pathway into the brain, further degrading electrode performance
(Subbaroyan, 2007). Regardless, many current technologies suffer from connector and mounting issues including breakages at the connector portion of the electrode or interface between the electrode and external connection. While direct research has not been performed on this issue, it has been specifically mentioned in previous studies
(Williams et al., 1999; Hochberg et al., 2006; Vetter et al., 2004). Similar issues of electrode failure due to electrode breakages and degradation of insulation are also highlighted by other researchers (Sanchez et al., 2006; Simeral et al., 2011b). Early results from Michigan’s carbon fiber electrode also indicate packaging issues based on
35 rapid changes in performance in spite of a limited immunohistochemical response (D.
Kipke, personal communication).
2.3.3.4. Intracortical Electrode Recording Longevity
One issue with intracortical electrodes is the limited longevity of the devices.
Previous research has indicated that a typical implant does not exceed 6 weeks in a rat or
guinea pig (Vetter et al., 2004; Williams et al., 1999; Rennaker et al., 2005b; Rennaker et
al., 2005a). Work by Williams et al. (Williams et al., 1999) has indicated that only 30%
of electrodes were still active at 5 weeks. In another study, automated insertion of
electrodes in the rat auditory cortex lasted through six weeks while hand inserted
electrodes did not (Rennaker et al., 2005b).
A recent literature review examined several decades of intracortical electrode
work across all types. The review indicates that although microwire experiments in the
1980s have had some of the best success, the longevity of devices in animals are usually
in the best case 2-3 years though one non-human primate has lasted 7 years (Tresco and
Winslow, 2011). Though the 2-3 years represents the typical maximum values, the
average lifetime is much less. Only 40-60% of electrode sites record from the initial
implant, and usually, the signals deteriorate after several months in non-human primates
(Schwartz, 2004; Ryu and Shenoy, 2009).
Though experience in humans with intracortical electrodes is less than with animals, work from the Donoghue lab and the BrainGate project has shown that the number of units (distinct neurons) degrades over time, as well (Figure 4). Although early data predicted that subject S3 would have no recordable units by day 500, recent work by the same group has shown subject S3 to still be recording from about 24 neurons (units)
36 on a given day around 1000 days (2.7 years)(Simeral et al., 2011b). From research of their lab, that number of neural signals allows for clicking and 2D control of a computer cursor since 18-29 units are needed for good 2D control. Additionally, other research has shown 39 directionally-tuned units allowed for 3D control by a non-human primate
(Taylor et al., 2003). Further research with a non human primate for 3D control with the addition of a gripper (4 degrees of freedom) used 60-120 units for control (Velliste et al.,
2008). Therefore, previous research indicates that more neuronal signals will be needed to control complex external systems via a BMI.
Figure 4. Number of units recorded in 2 subjects from BrainGate project (Simeral et al., 2011b; Simeral et al., 2011a; Kim et al., 2008). Trendlines are fit to a power equation Units = A*DayB.
A great deal of research has been focused on improving the number of units that electrode arrays can record. Many researchers have indicated that the tissue response is the main reason for the degradation of electrode performance instead of mechanical or
37 electrical failure (Leach et al., 2010; Schwartz, 2004; Grill et al., 2009). As discussed
previously, there are many mechanisms through which an electrode can effect the tissue
response, but there are three main pathways that the tissue response affects neural
recordings: 1) proximity of neurons to the electrode 2) tissue impedance between the
electrode and the neuron 3) the function of the neurons and their network around the
electrode (neural health) (McConnell, 2008).
Previous research has largely been focused on two parts of the response to
intracortical electrodes: 1) the electrode-tissue response or 2) the electrode and the quality
of neural recordings. There is a comparatively small amount of research that has
examined the whole span of the electrode-tissue response-neural recordings interaction.
One reason stems from the inherent difficulty of examining the electrode-tissue-neural recording interaction altogether since the tissue response is primarily characterized by immunohistochemistry labeling that can only be performed at one point in time with a given animal. Additionally, the highly vascularized nature of the brain has been assumed to contribute to the highly variable response to intracortical electrodes, thereby necessitating large numbers of animals for studies. Regardless, some research has been performed to examine the electrode-tissue-neural recording interface.
One notable effort has been to use complex impedance spectroscopy (also called
EIS for electrode impedance spectroscopy) to quantify the impedance of the tissue- electrode interface over time (Williams et al., 2007; Mercanzini et al., 2009). This technique has the advantage of analysis over time instead of the one point in time with histology. EIS provides detailed analysis of the resistive and capacitive components of the tissue response over time (Williams et al., 2007; Otto et al., 2006). Though
38 researchers have shown an inverse relationship between electrode recording quality and tissue impedance, the effect of impedance on recordings has shown to be weak (Purcell et al., 2009) . Studies have suggested a relationship between EIS data and glial scar response, but not neuronal density (Mercanzini et al., 2009; Williams et al., 2007). The method therefore offers the opportunity to indirectly examine a portion of the tissue response over time while examining the quality of neural recordings. Previous studies have shown a weak inverse relationship, if any, between impedance measurements and recordings (Purcell et al., 2009; Suner et al., 2005).
Other examinations into the electrode-tissue-neural recording interaction have been performed by treating animals with a chemical known to improve the tissue response. Dexamethasone has been shown in many studies to reduce the glial scar response and increase neural density around implants (Zhong and Bellamkonda, 2007;
Wadhwa et al., 2006; Kim and Martin, 2006; Zhong et al., 2005; Spataro et al., 2005).
One study used dexamethasone investigated the impact on neural recordings (Anderson,
2008). The findings of the research indicate that dexamethasone maintains neural recordings in guinea pigs better than untreated animals over a four week time period, but the tissue response was not examined in this study (Anderson, 2008).
A similar study examined the performance of electrode recordings using the drug minocycline, a neuroprotective antibiotic (Rennaker et al., 2007). The study indicated that minocycline treated animals had improved neural recordings in comparison to untreated animals where the SNR and number of units were improved with the minocycline treatment (Rennaker et al., 2007). Though minocycline is claimed to be neuroprotective, no immunohistochemistry for neuronal nuclei was performed. The
39 recording data was coupled to histology of only GFAP, reactive astrocytes. The results
indicate that minocycline reduces astrocyte activation at 1 week, but the activation of
astrocytes is only slightly, but significantly, lesser with minocycline at 4 weeks.
Therefore, the reduced activation of GFAP accompanies improved recordings. It appears
that the differences in GFAP histology are not of the same magnitude as differences in
electrophysiology (Rennaker et al., 2007).
Another electrophysiology study used flavopiridol, a drug that inhibits the cell
cycle and proliferation. Electrode recordings and histology were performed (Purcell et
al., 2009). In this work, animals treated with the drug saw lower impedance electrodes,
but no significant changes in recordings were seen. A lack of differences in recordings was accompanied by similarity between neuronal and non-neuronal densities around electrodes between treated and non-treated animals.
Another recent example examining the relationship between tissue response and electrical recordings is work from a Utah group (Parker-Ure et al., 2008). The work presented at an annual conference examined the Utah array implanted into one cat and the correlation between recordings at 8 months and immunohistochemical labeling. Though the data was limited to one cat, almost 100 electrodes were examined. Remarkably, the research showed no correlation between neural density around the implant and performance of recordings. Similarly, labeling for GFAP and microglia did not correlate
with recording performance. The researchers did find Isolectin B4 inversely correlated
with neural recording performance. Though the presentation noted that Isolectin B4 was
a marker for leaky blood brain barrier, the abstract for the presentation notes that it is an
inflammatory biomarker. Other researchers have used it to label microglia and
40 endothelial cells (Alonso, 2005). Regardless, their findings suggest that neuronal proximity and density may not be the principal determinants for neural recording performance.
2.4. Summary
The research has indicated an implanted intracortical electrode impacts the tissue response and affects neural recordings through many pathways. An implanted electrode causes damage, initiates an acute response, and agitates a chronic response. Several researchers have examined different factors to control the chronic response to an implant including size, shape, surface roughness, porosity, and chemical modification. Since research in vitro and in silico has indicated that material stiffness affects cells and tissue, the first focus of research will be concerned with softer materials and the effect of softer materials on tissue response.
Ultimately, the goal of the research is to improve electrode recordings. To that goal, the second focus of the research is concerned with the effect of the inflammatory response on neural recordings. Since any novel electrode design, using softer materials or otherwise, is aimed at modifying the inflammatory response, it is important to understand the effect of the inflammatory response on neural recording quality.
Examinations of both foci will enable further understanding of key design components that would allow the future development of better electrode technologies.
The improved technologies offer the opportunity to greatly improve understanding of many neurological conditions, capabilities of BMI systems, and the quality of life of patients with severe paralysis.
41
CHAPTER 3. IN VIVO DEPLOYMENT OF MECHANICALLY ADAPTIVE NANOCOMPOSITES FOR INTRACORTICAL MICROELECTRODES*
*The following chapter is reproduced, with permission, from Harris J, Hess AE, Rowan S, Weder C, Zorman C, Tyler D and Capadona J 2011 Journal of Neural Engineering 8 046010 DOI: 10.1088/1741-2560/8/4/046010
3.1 Abstract
We recently introduced a series of stimuli-responsive, mechanically-adaptive
polymer nanocomposites. Here, we report the first application of these bio-inspired
materials as substrates for intracortical microelectrodes. Our hypothesis is that the ideal
electrode should be initially stiff to facilitate minimal trauma during insertion into the
cortex, yet becomes mechanically compliant to match the stiffness of the brain tissue and
minimize forces exerted on the tissue, attenuating inflammation. Microprobes created
from mechanically reinforced nanocomposites demonstrated a significant advantage
compared to model microprobes composed of neat polymer only. The nanocomposite
microprobes exhibit a higher storage modulus (E’ = ~5 GPa) than the neat polymer
microprobes (E’ = ~2 GPa) and could sustain higher loads (~12 mN), facilitating penetration through the pia mater and insertion into the cerebral cortex of a rat. In contrast, the neat polymer microprobes mechanically failed under lower loads (~7 mN) before they were capable of inserting into cortical tissue. Further, we demonstrated the material’s ability to morph while in the rat cortex to more closely match the mechanical properties of the cortical tissue. Nanocomposite microprobes that were implanted into the rat cortex for up to 8 weeks demonstrated increased cell density at the microelectrode-
42 tissue interface and a lack of tissue necrosis or excessive gliosis. This body of work
introduces our nanocomposite-based microprobes as adaptive substrates for intracortical
microelectrodes and potentially other biomedical applications.
3.2. Introduction
Intracortical microelectrodes for neural unit recording have led to many
significant insights into the behaviour of the brain in animal models (Nicolelis, 2003;
Schwartz, 2004). There have been several notable efforts to develop probes that can be
implanted chronically in humans to directly interface with the brain to achieve command,
control, and feedback in many clinical applications (Kennedy et al., 2004; Hochberg et
al., 2006). Intracortical interfaces that could be implanted and remain stable for many years offer many significant opportunities to improve human health, as well as our understanding of the brain. Unfortunately, at this time, this technology is not feasible for wide-scale clinical applications due to the variability in the reliability of the recordings over both short and long time periods, regardless of the type of electrode used (Ward et al., 2009).
While it has been shown that intracortical microelectrodes can record the activity of individual or small populations of neurons early after implantation (Buzsáki, 2004;
Schwartz, 2004), challenges still remain in maintaining the ability to record neural signals from individual or small groups of neurons for extended periods (McConnell et al., 2009;
Ward et al., 2009; Williams et al., 1999). This is largely attributed to the inflammatory
response initiated by the trauma of implantation, and the foreign body reaction spanning
the life of the chronically implanted microelectrode (Polikov et al., 2005; He and
Bellamkonda, 2007; Leach et al., 2010). Support for the importance of reducing the
43 chronic inflammatory response to penetrating neural microelectrodes has been reported previously in both in vitro and in vivo models (Polikov et al., 2006; Polikov et al., 2005).
Both of these systems demonstrate electrode encapsulation by neural inflammatory cells, and significant reduction in the number of neurons at the electrode tissue interface
(“neuronal dieback”). Although the dominant mechanism is still being debated, it is important to point out that the ability of cortical microelectrodes to record from individual neurons is directly related to the proximity of viable neurons and the characteristics of non-neural, encapsulation tissue between the electrode and neuron
(Biran et al., 2005).
Traditionally, microelectrodes have been composed of metals, silicon and/or ceramics. These materials have a high mechanical stiffness (high modulus) relative to brain tissue. While the high modulus has enabled facile electrode implantations
(Szarowski et al., 2003), the stiffness difference between brain tissue (~6 kPa) and typical electrode material (~200 GPa) results in chronic shear and differential motion between the microelectrode and the neurons (Subbaroyan et al., 2005). Previous research with in silico modeling and in vivo studies has indicated that indwelling microelectrodes become mechanically coupled to the brain tissue and exert forces on local populations of cells
(Lee et al., 2005; Subbaroyan et al., 2005; McConnell et al., 2007b). The effects of the mechanical mismatch are thought to play a significant role in the cell-mediated inflammatory response impacting the glial scar, neuronal health, and the microelectrode– cortical tissue interface (Williams et al., 2007; Polikov et al., 2005; He and Bellamkonda,
2007; Leach et al., 2010).
44 Several research groups have attempted to investigate the effects of mechanical
mismatch between the microelectrodes and the cortical tissue, and many have developed
microelectrode substrates and substrate coatings from materials such as polyimide, SU-8, polydimethylsiloxane (PDMS), and parylene that are more compliant than materials traditionally used to create electrodes (Rousche et al., 2001; Subbaroyan and Kipke,
2006; Takeuchi et al., 2005; Wester et al., 2009; Fernandez et al., 2009; Mercanzini et al., 2009; Lu et al., 2009). However, in general, these materials have had limited success in attenuating glial scar formation or improving neural recordings. A drawback to most of these studies is that such materials still have Young’s moduli 6 orders of magnitude larger than that of the brain (Takeuchi et al., 2005; Lee et al., 2004; Nikles et al., 2003;
Takeuchi et al., 2004; Wester et al., 2009). Additionally, the slight decrease in stiffness
of these polymeric based microelectrodes results in difficulties in implanting the
microelectrodes (Takeuchi et al., 2005; Kozai and Kipke, 2009; Lee et al., 2003; Kee-
Keun and et al., 2004), requiring insertion shuttles that can cause increased volumes of
tissue damage during insertion and retraction (Kozai and Kipke, 2009).
Therefore, we hypothesize that the ideal microelectrode should be composed of a
material that is initially stiff to facilitate insertion into the cortex, yet becomes
mechanically compliant to match the stiffness of the brain tissue, to minimize forces
exerted on the tissue, and to attenuate inflammation around the implant. To this end, we
have recently developed a new class of bio-inspired, chemo-responsive, mechanically-
adaptive polymer nanocomposite that can controllably and selectively be switched
between stiff and compliant states (Capadona et al., 2009; Capadona et al., 2008;
Capadona et al., 2007; Shanmuganathan et al., 2010b, a, 2009; van den Berg et al.,
45 2007). The bio-inspired adaptive nanocomposite material design was based on the dermis of the echinoderm holothuroidea (sea cucumber). These invertebrates feature soft connective tissues with mutable mechanical properties; the animal can switch between low and high stiffness states on a sub-second time scale. Our materials exhibit this behaviour by mimicking the architecture and proposed switching mechanism at play in the sea cucumber dermis by utilizing a polymer nanocomposite consisting of a controllable structural scaffold of rigid cellulose nanofibres embedded within a soft polymeric matrix. When the nanofibres percolate they interact with each other through hydrogen bonding and form a nanofibre network that becomes the load-bearing element, leading to a high overall stiffness of the nanocomposite. When combined with a polymer system that additionally undergoes a phase transition at physiologically relevant temperatures, a contrast of over 2 orders of magnitude for the tensile elastic modulus is exhibited. Initial material characterization has demonstrated that the polymer nanocomposite reduces its tensile storage modulus from ~5 GPa to 12 MPa within 15 minutes upon exposure to simulated physiological conditions in artificial cerebral spinal fluid (ACSF) at 37oC (Shanmuganathan, 2010; Capadona et al., 2009; Capadona et al.,
2008). Additionally, advanced bioMEMS fabrication processes have been developed to create cortical microprobes from this novel material (Hess et al., 2009; Hess et al., 2011).
Here, we report on the first demonstration of the application of our mechanically- dynamic material towards intracortical microelectrodes. This study shows that the stiffness of the nanocomposite facilitates insertion into the cortex of a living rodent, that the nanocomposite mechanically morphs within the animal to achieve a compliant state
46 more similar to the mechanical properties of the brain, and that the nanocomposite can
integrate within the cortical tissue over many weeks.
3.3. Materials and Methods
3.3.1. Device Fabrication
Nanocomposite (NC) implants consisted of poly(vinyl acetate) and cellulose
nanocrystals derived from tunicates (tunicate whiskers or whiskers) that were created by
casting films from a dimethylformamide (DMF) solution of the polymer and whiskers
(Capadona et al., 2008). Poly(vinyl acetate) (PVAc, weight-average molecular weight,
MW = 113,000 g/mol, density, d = 1.19 g/cm3) was purchased from Sigma-Aldrich. The
isolation of cellulose whiskers largely followed our published protocols with minor
modifications (van den Berg et al., 2007). NC films consisted of 15% w/w tunicate
whiskers, and the films were used to create sheets via compression-molding in a hot press
(Carver, Wabash, IN) to yield 53-110 µm thin films, depending on the experiment.
Subsequently, two types of devices were created, one for mechanical testing and buckling
tests, and another for insertion force measurements and histology.
Starting with compression-molded sheets, for both types of devices, the NC sheets
were cut into the appropriate shapes. For insertion force and histology, 100 µm thick
sheets were cut into 3 mm long probes that were 200 µm wide with an approximately 45°
angle tip. Probes for mechanical testing and buckling tests were created from a solution-
cast poly(vinyl acetate)-tunicate whisker film compressed to a specified thickness (53 µm
for buckling tests, 110 µm for mechanical measurements) and weakly adhered to a bare
Si wafer by applying gentle pressure and heating on a hotplate set to 70°C for 3 minutes.
Using a direct-write CO2 laser at a laser power of 0.5 W, speed of 56 mm/s, and a
47 resolution of 1000 pulses per inch (Hess et al., 2009; Hess et al., 2011), microprobes for
mechanical testing were fabricated as beams 6 mm in length and 180-210 µm in width.
For buckling testing, beams with a tip opening of 30° were cut using the same settings for
the laser (3mm in shank length, 125 µm in width, 53 µm thick). Neat polymer probes for
buckle testing were fabricated to a similar size (3mm in shank length, 125 µm wide, and
55 µm thick) from spin-cast films. Neat PVAc films were fabricated by first dissolving
30 g PVAc in 50 mL dimethylformamide (DMF). Approximately 5 mL solution was
statically dispensed onto a bare Si wafer, and then spin-cast onto the wafer at a spin speed
of 2000 rpm for 30 seconds. The film was then dried in a vacuum oven for 4 hours at
65°C. Films were then immersed in DI water to aid in peeling them from the wafer. After
drying overnight in ambient conditions, neat PVAc microprobes were laser-patterned using the same parameters as were developed for patterning the NC microprobes.
Separately, acrylic mounts designed to serve as the sample clamp during tensile testing were also fabricated and marked with an identification number and two lines indicating where the sample would be attached to the mount. A 1.5 mm-long portion at one end of the NC samples was adhered to the acrylic mount using a cyanoacrylate-based adhesive, leaving a 4.5 mm-long cantilever structure hanging from the mount (Figure 5, digital image). After curing the adhesive overnight, the width and thickness of each sample were measured under a microscope and recorded along with the identifying number.
48
Figure 5. Diagram of Microtensile Testing (MT) instrumentation. Sample is held between two clamps where a load is applied to the sample, and both force and displacement are measured. The light and humidifier maintain the conditions similar to that within the rodent. The digital image depicts the model microprobes attached to the clamp (scale = 1mm).
3.3.2. Surgical Procedures
The same surgeon performed all procedures to minimize variability. The Case
Western Reserve University Institutional Animal Care and Use Committee (CWRU
IACUC) approved all procedures and all efforts to minimize the pain and discomfort of
the animals were employed.
Sprague-Dawley rats weighing between 200-340g were used for all experiments. ketamine (80 mg/kg) and xylazine (10 mg/kg) provided anesthesia induction via intraperitoneal injection (IP). When the animal was unresponsive to a toe pinch, the animal was prepared for surgery via head shaving and eye protection with ocular lubricant. Upon transfer of the animal to the stereotaxic frame fitted with a gas mask, the animal was ventilated with oxygen. Vitals were measured by a pulse oximeter attached to
49 the hind limb paw. The animal was kept warm via a circulating water mat. When the
animal began to whisk, react to toe pinch, or exhibit a raised heartbeat, introduction of
Isoflurane (1-3%) into the oxygen flow maintained a surgical level of anesthesia.
Access to the brain was gained by first creating a midline incision via a scalpel.
Cotton swabs were used to move periosteum from the skull, and either a 3 mm biopsy
punch or dental drill with stainless steel burr bit was used to remove the bone of the skull.
Both types of openings were approximately centered at 3 mm lateral from the midline
and 4.5 mm caudal to the bregma. After opening the skull, the dura was removed using a
fine 45o angle microprobe and fine forceps using a surgical microscope. Implants were attached to ceramic forceps tips, and implants were manually lowered to near the surface of the brain. A computer-controlled mechanical inserter controlled by a custom LabView
7.1 program inserted implants 2-3 mm into the brain at 2 mm/s. For microtensile testing, implants were inserted into the brain with a micromanipulator instead of the computer.
In the case of animals used for histology (chronic survival animals), probes were left implanted in the animal. The skull was sealed with Kwik-Sil (WPI, Inc. Sarasota,
FL) followed by dental cement adhered to three stainless steel screws implanted into the brain. The skin of the scalp was sutured closed over the dental cement and closely monitored for recovery.
3.3.3. Mechanical Test Procedures
3.3.3.1. Microtensile Testing Setup
Samples were implanted in the rat cortex for 1 to 30 minutes, then removed and immediately inserted into the microtensile tester (MT) that had previously been validated
(Hess et al., 2009; Hess et al., 2011). A computer-controlled, custom-built microtensile
50 tester (Figure 5) was utilized to measure the stress-strain relationship of NC samples that were used for in vivo experiments. The samples were gripped over a length of 1.5 mm length on each end between acrylic blocks clamped by metal bolts, leaving a gauge length of 3 mm. One grip was anchored to an immobile structural component of the microtensile tester, and the other grip was attached to the drive rod of a linear piezomotor that applied a strain to the sample. The displacement distance was measured with an indicator with a resolution of 0.5 mm, and the force required to pull the sample was measured with a load cell with a resolution of 49 mN. The setup was computer-controlled via data acquisition software.
Each sample was tensile tested by straining at a rate of 8.9 mm/s until failure. The time from sample removal from the brain until completion of tensile test was recorded, typically between 3 to 4.5 minutes. After testing was completed, the stress and strain were calculated and the Young’s modulus was determined from the slope of the curve in the initial linear portion of the stress-strain curve.
3.3.3.2. Moisture and Temperature Maintenance
In order to keep the environmental conditions of the test setup similar to those of the biological environment of the implant, considerations were made for temperature and humidity. Temperature was controlled with the radiant heat from a focused light source using a thermocouple placed near the sample to set the local temperature to 37°C. A fluid mist, via a commercial airbrush, maintained the humidity of the environment to prevent sample drying (Humidifier, Figure 5). This is important, as moisture is the primary stimulus of changing mechanical properties of the NC. Effects of using or not using the
51 humidifier were examined using dry and wet samples to examine the effect of the
humidifier on mechanical testing results.
3.3.3.3. Dimensional Changes Due to Swelling
To determine the cross-sectional dimensions of implanted samples for stress
calculations, sample thickness and width were measured before and after soaking samples
similar to the implanted samples in artificial cerebrospinal fluid (ACSF) for 30 minutes at
37°C. The width or thickness of dehydrated samples was measured under a light
microscope before the samples were immersed in ACSF at a temperature of 37°C. After
30 minutes, the width and thickness of the samples were measured again. The percentage
change for each dimension was calculated and used to calculate the cross-sectional dimensions of implanted samples.
3.3.4. Insertion Force Measurements
A computer-controlled mechanical inserter was used to insert implants 2-3 mm into the brain at 2 mm/s (as described above). The load cell was calibrated before each use via loading the load cell with determined weights, creating a voltage-force calibration curve. The computerized setup controlled a stepper motor outfitted with an optical encoder to track the position of the stepper motor while a load cell was mounted on the front of the stepper motor to record the forces during the insertion. The data acquisition board, synchronized to a motion control board, allowed for synchronous recording of position and load cell voltage. Data from a single insertion was post-processed to use the calibration curve to convert recorded voltage into force. Graphs and measurements of force versus position were used to determine the needed force for insertion as determined
52 by the first peak in the insertion graph after the artefact from the initiation of movement
(Jensen et al., 2006; Paralikar et al., 2006; Najafi and Hetke, 1990).
3.3.5. Buckling Load Calculations
Following analysis of insertion force data, statistical analysis was completed in the statistical package R (R Foundation for Statistical Computing, Vienna, Austria) (R
Development Core Team, 2009). An analysis was performed to analyze the distribution of forces and the force needed to insert into the brain. The 95% confidence level and required insertion force was computed using a t-distribution and analysis of the mean and standard deviation of observed data.
From the computed force to implant, the microprobe modulus was predicted by
Euler’s buckling formula (see below and Najafi et al. (Najafi and Hetke, 1990)). The parameter space was graphically investigated using Matlab R2009b (MathWorks, Natick,
MA) by varying device length, width, and thickness, and insertion force predicted.
3.3.6. Buckle Testing
During buckle testing, a video camera connected to a surgical scope recorded the insertion and retraction movements. The implants used had a 3mm long shaft and were
125µm wide with a thickness of 53µm (NC) or 55µm (neat polymer), both with a 30° opening angle. After three minutes, the implanted NC microprobes were removed from the brain tissue via the computer-controlled motor and manually moved to a new insertion site. Immediately, the computer-controlled inserter was initiated to lower the probe into the brain at 2 mm/s. Between repeated insertions, the brain was not continually hydrated in order to minimize sample wetting by exogenous applied saline.
53 3.3.7. Histology
As a preliminary examination into the chronic response to the NC, a pair of
animals was implanted with the NC microprobe and was euthanized after four or eight
weeks to examine histology of the brain-implant interface. Following anesthesia
induction, transcardial perfusion was performed via a mechanical pump with
approximately 500 mL of Dulbecco’s Phosphate-Buffered Saline followed by 250 mL of
10% buffered formalin fixing the tissue. The extracted brain was placed in formalin for
24 hours, after which the brain was then transferred to fresh formalin where it remained
for up to one week, until it was cryoprotected in 30% sucrose solution. To perform
slicing of the brains, brains were frozen in embedding molds (Electron Microscopy
Products, Hatfield, PA) containing Optimal Cutting Temperature (OCT) compound
(Sakura, Tokyo, Japan) that were then mounted and sectioned horizontally using a
cryostat to create 30 µm thick slices. Following a standard haematoxylin and eosin
(H+E) protocol, slices were stained, mounted, and cover-slipped. Briefly, sections were rinsed in distilled water, and nuclei were stained with the haematoxylin. Following a rinse in tap water, tissue was differentiated with acid alcohol and rinsed in tap water.
Finally, samples were stained with eosin for 2 minutes, then dehydrated and mounted(Lillie, 1965). Separate slices were counterstained with 4',6-diamidino-2- phenylindole (DAPI) dilactate (Invitrogen, Carlsbad, CA) at room temperature. All slices were imaged via a microscope outfitted with a grayscale CCD camera.
54 3.4. Results
3.4.1. Microprobe Insertion into Rat Cortex
Implants fabricated from the PVAc/cellulose nanowhisker nanocomposite (NC)
or, for the purpose of comparison, the neat PVAc polymer, were attached to ceramic
forceps tips and manually lowered to near the surface of the brain. They were then
advanced by computer control 3 mm towards the exposed brain at 2 mm/s. The
implantation path typically exhibited a slight motion start artefact; an example can be seen in the first peak of Figure 6a,b. Generally, there was no increase in force before the microprobe came into contact with the surface of the brain. As the pia mater was indented, the force exerted on the tissue grew prior to puncture of either the pia mater tissue (microprobe insertion through the tissue, Figure 6a), or microprobe failure
(material buckling, Figure 6b). Video examples of an insertion and a buckle can be seen in the supplemental material for (Harris et al., 2011b). A full explanation of the videos and still frames is below.
In the case of the NC microprobe, the pia was punctured, allowing the microprobe to penetrate into the brain, facilitating implantation when the force exerted was lower than 14 mN (Figure 6a). After microprobe implantation was achieved, the force was quickly relaxed. After insertion, the microprobe continued to advance through the cortex, generating increased resistance against the microprobe, as depicted graphically by the second gradual rise (~1500µm – 3000µm) in force after the initial insertion.
55
Figure 6. The insertion force was recorded as a function of microprobe position during all microprobe insertion trials. Representative trials of both the insertion of a 3mm long NC microprobe into the dura-removed cortex at a rate of 2mm/s (a), and the buckling of a 3mm long neat PVAc microprobe upon failed insertion into the dura-removed cortex at a rate of 2mm/s (b). Under identical conditions, the neat PVAc polymer microprobes created for the purpose of comparison failed to penetrate the pia mater, but first bent and eventually buckled when the force exceeded 5.6 mN (Figure 6b). The bowing out and eventual buckling of the microprobe resulted in a reduction in measured force where the buckling force was less than the required penetration force. In this case, the pia continued to indent as the microprobe was unable to penetrate the tissue, causing a second rise in force after the initial insertion or buckling event. We defined the maximum force as the force before the microprobe starts bowing out.
56 3.4.2. Dry NC Modulus Measurements
Displacement and force data are converted to strain and stress, respectively, by
dividing displacement by the initial length, and dividing force by the initial cross-
sectional area of the sample, respectively (Equation 1a):
and (Equation 1b):
,
where ! is strain, d is the displacement measured by the indicator, l0 is the initial gauge length (3 mm), # is the sample stress, F is the force measured by the load cell, w is the sample width, and t is the sample thickness.
First, a dry NC microprobe sample was tested at room temperature to confirm the as-fabricated sample properties with our previous studies (Hess et al., 2009; Capadona et al., 2008; Shanmuganathan et al., 2010b, a, 2009). The Young’s modulus (E) of the sample is defined as the slope of the linear portion of the stress-strain curve. Using our test set-up, the Young’s modulus was measured to be 3411 ± 98 MPa (number of samples n=5 for this composition, Figure 7), consistent with that previously reported(Hess et al.,
2011).
57
Figure 7. Stress-strain plots for both pre-inserted dynamic microprobes (dry, room temperature; E = 3411 MPa) and the identical composition of microprobe that had been implanted in living rat cortex (30 min implantation, explanted and tested at 37°C; E = 33 MPa). The inset shows the stress-strain plot of the explanted sample on a different scale. 3.4.3. In vivo Mechanical Switching
To confirm that the NC microprobes remained mechanically dynamic upon implantation into the rat cortex, as previously shown in vitro for this bio-mimetic design
(Capadona et al., 2008), we explanted microprobes and re-examined their Young’s modulus using our microtensile device (Figure 7). Implantation into the rat cortex reduced the Young’s modulus of the NC by several orders of magnitude.
The ex vivo Young’s modulus of the mechanically dynamic microelectrodes was plotted as a function of implantation time (Figure 8). There was no significant difference between samples that had been implanted for 5, 15, or 30 minutes, which collectively had a Young’s modulus of 32 ± 13 MPa (Figure 8). There was no statistical difference between data collected from microtensile testing (MT) and dynamic mechanic analysis
(DMA) after fifteen minutes of immersion (Shanmuganathan, 2010). The time required
58 for set up of the tensile test plus the time required to strain the sample to break generally required 3-4.5 minutes after removal from the brain.
Figure 8. The Young’s modulus of the mechanically dynamic materials were measured with a DMA (open squares; bulk materials(Shanmuganathan, 2010)) and with MT (open circles; explanted microprobes) to confirm mechanical switching from stiff to compliant. DMA samples were placed into a submersion clamp with ACSF preheated to 37oC (Shanmuganathan, 2010) while microprobes were implanted into the rat cortex and explanted for MT testing. The x-axis indicates time of exposure to either ACSF or implanted in the rat cortex, respectively.
3.4.4. Examination of Ex Vivo Modulus Testing Moisture
If allowed to dry upon explanting from physiological conditions, the reinforcing network within the mechanically dynamic NCs will return to their original stiff state
(Capadona et al., 2008). Therefore, to ensure an accurate measurement of the explanted modulus, materials were humidified via a fluid mister using water to prevent drying. To examine the effect of the fluid misting on measured samples, a sample was repeatedly tested before, during, and after the application of a misting spray. In the case of a dry NC
59 sample, the modulus was not statistically different between before and during humidifier application over 6 minutes (data not shown). After the 6 minutes of continued misting, the modulus did become significantly different than the modulus before or during humidifier application (p=0.03, 2 Sample t-test); though the modulus only decreased by at most 10% after exposure to the humidifier moisture. This modulus decrease could be from the applied moisture, or the decrease could also be due to the sample being taken very slightly out of the elastic region causing a reduction in the measured modulus.
Alternatively, the difference could be from an increase in temperature after the humidifier was turned off, as temperature was not controlled in these tests. Further tests were performed to examine the effects of the humidifier (data not show). After determining a setting at which the humidifier would contribute a minimal reduction in the Young’s modulus of the NC, the effect of the mister on wet samples was also investigated.
Samples like those used in the ex vivo mechanical testing were soaked in DI water for 10 minutes, followed by loading into the microtensile tester. For both the “humidifier off” and “humidifier on” conditions, the Young’s modulus of the NC as a function of time removed from water was determined using the microtensile tester in dynamic mode.
When the humidifier was not used to prevent the sample from drying, the Young’s modulus quickly increased to ~400 MPa by the time the microtensile testing started, and after 5 minutes removed from the DI water, the Young’s modulus was ~1100 MPa. When the humidifier was used, the Young’s modulus remained <100 MPa for nearly 4 minutes, and leveled out at ~600 MPa after 7 minutes. It should be noted that the linear portion of the stress-strain curve was complete within ~30 seconds after the start of the ex vivo mechanical testing. At this point, the sample had been removed from the brain for no
60 more than 3 minutes, a time at which there would be a considerable increase in Young’s
modulus from drying if the humidifier were not used.
3.4.5. Buckling and Insertion of Neat and NC Polymers
A summary of the recorded forces during the computer-controlled movement of
both types of microprobes was created (Figure 9). Measurements were made in one
animal by alternating between microprobe types. To ensure that the microprobes were
responding to just the cortical tissue, the surface of the brain was not supplemented with
moisture and slowly dried during these experiments. As a result, the critical insertion
force increased during the experiment. However, the graph (Figure 9) demonstrates that the unreinforced (neat polymer) microprobes failed to insert into the cortical tissue at a lower critical force than the mechanically reinforced dynamic microprobes.
61
Figure 9. The maximum force was recorded by a computer controlled load cell during each insertion attempt. The critical insertion force increased with subsequent trials. Circles represent neat PVAc microprobes, while squares represent dynamic microprobes. Filled shapes were inserted successfully while open shapes buckled. The grey bars represents the critical buckling force (Fcrit) for the microprobe, the theoretical maximum
force the microprobe could endure before buckling. The Fcrit is based on the material’s modulus and measured dimensions with standard deviation. The height of the grey bars is determined by the standard deviation of the microprobe width and thickness. Using the device dimensions and Young’s modulus of the neat or NC polymer, a
critical buckling force was predicted using Euler’s buckling formula (Equation 2a):
and (Equation 2b):
where Fcrit= the critical force required to buckle the implant, I= area moment of inertia, w=width, t=thickness, E=Young’s Modulus of the material, Leff= effective length. The
62 implant was modeled as a beam with one fixed end, and one hinged end, thus making Leff
= Length/"2. For these calculations, the thickness is defined as being less than or equal to the width.
The critical buckling force is plotted as a grey horizontal bar where the standard deviation of sample dimensions determines the bar’s height in Figure 9. Although the
NC microprobes are thinner than the neat polymer, the critical force of the NC is much higher because of the NC’s greater Young’s Modulus. For both polymers, insertions were successful when the maximum force was less than the critical buckling force as predicted by Euler’s buckling formula. When the maximum force exceeded the critical buckling force, microprobes would buckle. The NC was able to withstand a buckling force greater than the neat polymer (Figure 10, Supplemental Video S1 in (Harris et al.,
2011b)). Importantly, the neat polymer probes would need to be significantly bigger than
NC microprobes, as a neat microprobe would need to be almost 40% thicker or three times wider than NC microprobes to allow insertion.
To better examine the functional consequences of the change in modulus of the dynamic nanocomposite, NC microprobes were implanted for several minutes. A NC
microprobe was removed from the brain and attempted to be reinserted through intact pia.
The computer-controlled reinsertion resulted in a typical buckling behaviour as the probe
started to bow out, then buckle, and collapse upon itself (Figure 11, Supplemental Video
S2 in (Harris et al., 2011b)).
63
Figure 10. Snapshots of Supplemental Video S1 in (Harris et al., 2011b) of insertion attempts of the neat polymer and the NC microprobes. (a) Before initial insertion, the NC is about a millimeter above the brain’s surface. (b) During the movement, the NC indents and then penetrates the pia and cortex. (c)After the completion of movement, the NC is implanted in the brain and the indentation is relaxed. (d) Before initial insertion, the neat polymer is about a millimeter above the brain’s surface. (e) During the movement, the neat polymer indents the pia but quickly buckles. (f)After the completion of movement, the neat polymer is completely buckled.
64
Figure 11. Snapshots of Supplemental Video S2 in (Harris et al., 2011b) of insertion, retraction, and buckling of NC probe. (a)Before initial insertion, the NC is about a millimeter above the brain’s surface. (b) During the movement, the NC indents and then penetrates the pia and cortex. (c)After the completion of movement, the NC rests in the brain for 3 minutes. (d)The NC is removed, and moved to a new location above the cortex. (e)During the movement of the reinsertion, the brain-softened NC, indents the pia and cortex. (f) The NC is not stiff enough to puncture the pia and cortex, and the NC buckles. Inset depicts close up of curved buckled microprobe when removed from brain’s surface.
3.4.6. Implanted NC Microprobe Histology
To examine the chronic inflammatory-mediated tissue response to the NC microprobes, we performed standard histological methods on animals that had been implanted with microelectrodes and allowed to survive for four or eight weeks. Classic haematoxylin and eosin (H+E) staining was performed on horizontal tissue sections from animals that had been implanted for eight weeks (Figure 12a). Cell nuclei were stained by haematoxylin and are pictured as black dots and light black dots. The surrounding
65 cortical tissue stained by eosin is pictured as grey, and blood vessels are throughout the
section and appear as white holes. The image with NC still embedded in the section
shows intimate contact of the tissue with the microprobe. The NC appears striated as a
result of the ethanol dehydration during tissue processing. The dark ring around the NC
shows increased density of cells at the tissue-microprobe interface. The response suggests
the lack of significant necrosis or gliosis around the implant. Histological staining for total cell nuclei with DAPI (Figure 12b) supports our finding from H+E staining. In this
case, the NC was removed before DAPI imaging.
66
Figure 12. Histological evaluation of the microprobe-tissue interface was obtained with (a) H+E and (b) fluorescent staining with DAPI. All images were obtained from horizontal sections of brain tissue approximately 1mm deep into the cortex of animals implanted for four (b, DAPI) or eight (a, H+E) weeks. (a) Nuclei are stained various shades of black, other tissue is gray, and blood vessels are white holes. (b) Fluorescent view of DAPI. Cell nuclei are pictured white. (Scale = 100µm).
3.5. Discussion
In this study, we have for the first time been able to validate the feasibility of our mechanically adaptive NC as a microprobe that could form the basis of an intracortical microelectrode. We have shown the applicability of standard mechanics formulas in the insertion of microprobes. Euler’s formula suggests that with our unique NC, we can
67 decrease the size of the implant to minimize surgical trauma. Though this enables us to implant the material and understand its mechanical performance, we also demonstrated in preliminary histology studies that the implant does not create a significant glial scar.
Collectively, this work lays the foundation for continued cortical electrode development to improve cortical neural recordings.
While our initial interests were in exploiting our dynamic mechanical materials in biomedical applications, specifically as adaptive substrates for intracortical microelectrodes, exposure to brain tissue had previously only been simulated by immersing the samples into artificial cerebral spinal fluid (ACSF) and heating to a physiological temperature of 37 °C (Capadona et al., 2008; Shanmuganathan et al., 2009,
2010a, b). Therefore, to examine the performance in the rodent brain, we used a microtensile tester that was developed specifically to test the small microprobes that have been developed towards intracortical microelectrodes. Previous results have validated that the microtensile tester provides similar results to DMA testing, a standard in the mechanical testing field (Hess et al., 2009; Hess et al., 2011). Since the microprobes quickly change in response to temperature and water uptake (Capadona et al., 2008), we were concerned that small samples would dry out once explanted from the rat brain.
Therefore, the microtensile tester was paired with an airbrush and heat lamp to minimize drying and cooling when implants were removed from the brain (Figure 5).
Our ex vivo microtensile testing confirmed that the NC rapidly decreases its modulus when implanted into the rodent brain (Figure 8). Though the fabrication method provides some variability between samples, results consistently showed that the modulus switch occurred within the span of five minutes. Both DMA and MT testing showed
68 similar modulus decreases, though DMA testing showed a more gradual change. The rate at which the dynamic material decreased its modulus was measured on the MT after extraction from the cortical tissue (Figure 8), and compared to the bulk samples measured in solution on a DMA. Differences are presumably due to the much larger surface area- to-volume ratio of the micromachined samples used with the MT (Hess et al., 2009; Hess et al., 2011; Capadona et al., 2008). Additionally, for DMA measurements, dry samples were placed into a submersion clamp system to measure the dynamic change in modulus as the material both heated and swelled to decouple the reinforcing network and lower the modulus. In the DMA studies, the change in the dimensions of the material was not taken into account during swelling. This may account for slight difference in modulus between the two methods of testing.
Though we have confirmed that the material reduces its modulus once implanted, it is important to optimize the design of the electrode to minimize surgical trauma. Some research has suggested that microelectrode size does not affect the inflammatory tissue response (Szarowski et al., 2003), yet other research has shown that very small electrodes do reduce the glial scar response (Seymour and Kipke, 2007). Collectively, recent literature suggests that size differences on the same order of magnitude are negligible while orders of magnitude size differences are significant factors in the glial scar response. Therefore, if the microelectrode size can be significantly minimized, through the use of proper design or new materials, one would be able to reduce the surgical trauma, presumably contributing to a decreased chronic inflammatory reaction.
Using design principles and methods, we sought to minimize the size of the implant needed for insertion by first measuring the required insertion force. We
69 implanted microprobes into the cortex of many animals. Through the use of our custom
computer-controlled inserter and force measurements, our experiments have shown that
our hand-cut NC microprobe (100µm thick, 200µm wide) provides a range of insertion
forces from 1.26 mN to 3.35 mN (95% confidence interval). Alternatively, the mean
force for insertion is 2.30 mN ± 0.38 mN (standard error). The range of animal sizes varied from 206g to 335g, but there was no correlation between animal size and required insertion force. A lack of correlation between size and force also occurred in prior pilot
tests of the NC microprobe where the animal range was 227g to 353g (data not shown).
Given the expected maximum insertion force, we can use the classical mechanical
formula, Euler’s buckling formula (Equation 2a), to compute the required modulus to
ensure a buckle-free implantation into the cortex. Figure 13 shows the calculated
modulus based on different device sizes (length, width, and thickness). It is apparent that
thickness is the main factor in buckling (Figure 13). Theoretical microprobes above a
given plane would insert while microprobes below would buckle. It is also notable that
the needed material modulus rapidly increases at smaller dimensions; this important
factor has dictated that current standard cortical microelectrodes be made out of stiff
metal or silicon (Schwartz, 2004).
70
Figure 13. Parameter space for microprobe insertions. Based on a given force, the length, width, and thickness are parameters for Euler’s buckling formula to create a manifold to examine the required Young’s modulus to prevent buckling and enable insertion into the rat cortex. The highest plane has a length = 4mm, the next highest 3mm, and the lowest plane has a length = 2mm. The scale bar represents modulus in MPa. At the dimensions investigated in this study, both implants easily inserted into the cortex at low insertion force less than ~5.6 mN (Figure 9). This is consistent with the findings of Najafi et al. Their study reported a penetration stress through the rat pia of
1.2 x 107 dynes/cm2 for a Si probe that was 40 µm in thickness and 80 µm in width, corresponding to a penetration force of 3.84 mN (Najafi and Hetke, 1990). Figure 9 also clearly denotes an increased insertion force for subsequent microprobe implantations.
Unreinforced polymer microprobes were unable to penetrate the pia tissue when the insertion force reached ~7 mN, while the mechanically reinforced dynamic NC
71 microprobes were capable of penetrating the pia tissue with above 12 mN of force
applied; both are consistent with predictions from Euler’s buckling formula. The
difference compared to the Najafi study was likely a result of our experimental conditions. Specifically, due to the rate at which the novel materials uptake moisture and decouples the reinforcing network resulting in significant decreases in modulus(see above and (Capadona et al., 2008; Shanmuganathan et al., 2010b, a, 2009)), it was important to ensure that the microprobes were responding to the cortical tissue, and not the increased moisture on the surface of the brain. Thus, the brain was not supplemented with moisture between trials. Unfortunately, over the time course of repeated implantations, the surface of the brain slowly dried and collected dried blood, and most likely increased in the elastic modulus of the pia. As a result, the critical insertion force increased during the experiment. While the absolute values for the force recorded during the critical insertion are not reflective of a single insertion model, Figure 9 clearly demonstrates that the unreinforced (neat polymer) microprobes failed to insert into the cortical tissue before the mechanically reinforced dynamic microprobes.
The polymer NC consists of a controllable structural scaffold of rigid nanofibres embedded within a soft polymeric matrix. The network becomes the load-bearing element and leads to a high overall stiffness of the NC, as compared to the neat polymer materials. Therefore, the microprobes created from the dynamic materials are capable of sustaining an increased force during insertion. The improved mechanical properties of the NC over the neat polymer would therefore allow for a smaller device. As discussed above, to enable the insertion of a neat microprobe, the device would need to be almost
40% thicker or three times wider than NC microprobes to allow insertion.
72 In conjunction with the Boolean-type inserted/buckled results, we captured video of movements to show the complete functional impact of the properties of our NC. As shown above, the NC withstands greater insertion forces; demonstrated by videos showing the neat polymer buckle, directly followed by the insertion of the NC in the same area of brain (Figure 10 and Supplementary Video S1 in (Harris et al., 2011b)).
Additionally, the decrease in the modulus of the dynamic microprobe can be readily viewed by the naked eye within only a few minutes after the NC was implanted into the rat cortex. When the NC is then extracted and moved to another spot on the brain, due to the rapid decrease in modulus, the microprobe buckles and fails to penetrate into the pia tissue and reinsert into the cortical tissue (Figure 11, Supplemental Video S2 in (Harris et al., 2011b)). The time between the two insertions attempts were on the order of tens of seconds. The drying of pia was minimal since the dry NC implanted with a force of
2.8mN followed by the NC (after remaining in the brain for a few minutes) buckled upon reinsertion at about 3mN. These numbers agree well with what is predicted from the formulas.
Though we have shown how the properties of the brain influence the microprobe and its design, it is also important to examine the response of the brain tissue to the microprobe. In preliminary studies, we have used standard histological stains to investigate the material-tissue interface, with special interest in examining potential tissue necrosis. Histological staining of rats that had been implanted with dynamic NC microprobes and allowed to survive for four or eight weeks post-implantation suggest no tissue necrosis as a result of the implanted materials (Figure 12). Further, both H+E and
DAPI staining clearly show cell bodies at the implant interface, without excessive
73 overgrowth, suggesting the lack of excessive glial scar tissue formation. Obviously, the
extent of the histological examination presented on this tissue is not robust enough to
fully understand the cortical tissue response to our novel bio-inspired mechanically
dynamic microprobes. It is also important to point out that while this current generation
of our dynamic nanocomposites is capable of reducing its storage modulus to 12 MPa,
the cortical brain tissue has been reported on the kPa range. However, in silico research
has shown that a soft substrate on the order of 6 MPa significantly decreases the strain on
surrounding tissue by up to two orders of magnitude in comparison to silicon (E=200
GPa) or polyimide electrodes (E~3GPa)(Subbaroyan et al., 2005). Therefore, further
studies are underway to examine the more detailed response to these materials, including
a complete time course to examine the various stages of inflammation, as well as specific
cellular (neuronal cell nuclei, astrocytes, and microglia) and extracellular markers.
3.6. Conclusions
The goal in this work was to investigate the first biomedical application of our
novel bio-inspired mechanically-dynamic polymer nanocomposites. The intrinsic
properties of our materials were designed with the intended application of penetrating
intracortical microprobes and to afford us the distinctive opportunity to investigate the
role of material stiffness in the tissue response to the implant. Our previous work has
clearly characterized the materials properties of our novel class of material, in a
controlled laboratory setting. Here, we have for the first time, validated these bench-top studies, and have shown the initial feasibility of our material towards an intracortical microelectrode. Specifically, we have shown that the dynamic polymer nanocomposite is sufficiently stiff enough to penetrate the cortical tissue without the need for assistive
74 devices. Further, the initially stiff microprobe utilizes the in vivo environment to rapidly
become compliant to more closely match the surrounding cortical tissue while initial
histological examination suggest cellular integration at the microprobe-tissue interface.
Taken collectively, this work has established a solid foundation to complete further
studies to investigate the effect of substrate stiffness on glial scar formation and brain
tissue response.
3.7. Acknowledgements
This work was supported by Grant Number R21-NS053798 and F31-NS063640 from the
National Institute of Neurological Disorders and Stroke and T32-EB004314-06 for the
National Institute of Biomedical Imaging and Bioengineering. Additional support was from the Department of Veterans Affairs grant numbers C3819C, F4827H and B6344W, as well as the National Science Foundation under Grant Number ECS-0621984. The authors acknowledge the support of Anne DeChant and Kadhiravan Shanmuganathan.
None of the funding sources aided in the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the paper for publication. The authors have no conflicts of interest related to this work to disclose.
75
CHAPTER 4. Mechanically adaptive intracortical implants improve the proximity of neuronal cell bodies*
*The following chapter is reproduced, with permission, from Harris J, Capadona J, Miller R, Healy B, Shanmuganathan K, Rowan S, Weder C and Tyler D 2011 Journal of Neural Engineering 8 066011 DOI: 10.1088/1741-2560/8/6/066011
4.1 Abstract
The hypothesis is that mechanical mismatch between brain tissue and microelectrodes influences the inflammatory response. Our unique, mechanically- adaptive polymer nanocomposite enabled this study within the cerebral cortex of rats.
The initial tensile storage modulus of 5 GPa decreases to 12 MPa within 15 minutes under physiological conditions. The response to the nanocomposite was compared to surface-matched, stiffer implants of traditional wires (411 GPa) coated with the identical polymer substrate and implanted on the contralateral side. Both implants were tethered.
Fluorescent immunohistochemistry labeling examined neurons, intermediate filaments, macrophages, microglia, and proteoglycans. We demonstrate, for the first time, a system that decouples the mechanical and surface chemistry components of the neural response.
The neuronal nuclei density within 100 µm of the device at four weeks post implantation was greater for the compliant nanocomposite compared to the stiff wire. At eight weeks post implantation, the neuronal nuclei density around the nanocomposite was maintained, but the density around the wire recovered to match the nanocomposite. The glial scar response to the compliant nanocomposite was less vigorous than to the stiffer wire. The results suggest that mechanically associated factors such as proteoglycans and
76 intermediate filaments are important modulators of the response of the compliant nanocomposite.
4.2. Introduction
Despite the increasing evidence of the importance of cellular mechanotransduction on tissue repair and homeostasis (Clark et al., 2007; Ingber, 2006), the role of mechanical mismatch has not been fully elucidated in an in vivo study to explain the interplay of chemical and mechanical factors that contribute to glial scarring surrounding intracortical implants. Previous research in vitro and in silico has supported the importance of mechanical signaling in several cell types in the brain. Cultured astrocytes have been shown to respond to mechanical stimuli via calcium signaling
(Ostrow and Sachs, 2005). Investigations have shown that higher strain rates for cultured astrocytes lead to an increased reactivity (Cullen et al., 2007). Cell types have responded differently to substrate stiffness as well. Notably, the rate of astrocyte and neuron proliferation as well as oligodendrocyte spreading and neuronal branching were influenced by substrate stiffness (Kippert et al., 2009; Flanagan et al., 2002; Georges et al., 2006). Substrate stiffness is also known to shift cell differentiation in mesenchymal stem cells to be neurogenic, myogenic, or osteogenic (Engler et al., 2006).
In addition to the effect of substrate stiffness, in silico modeling studies indicate that indwelling electrodes exert forces on local populations of cells (Lee et al., 2005;
Subbaroyan et al., 2005; McConnell et al., 2007b). Over time, the implanted electrode is anchored to the tissue via the extracellular matrix and neural inflammatory cells
(including microglia and astrocytes), resulting in cellular attachments to the electrode modifying the forces exerted on the brain tissue (McConnell et al., 2007b). Micromotion
77 associated with electrode movement within the tissue and the mechanical properties of the electrode dynamically change the level of exerted forces on the cortical tissue during scar maturation. This phenomenon can also be exacerbated as a result of mechanically tethering the electrodes to the skull. In vivo studies which focus on the effects of electrode tethering have shown that untethered implants reduce the extent of the glial scar
(Biran et al., 2007; Kim et al., 2004; Subbaroyan, 2007). It has been suggested that a reduced glial scar results from the reduction in forces applied to the tissue. Alternately, it has been proposed that less scarring is the result of limited meningeal ingrowth in response to untethered implants (Subbaroyan, 2007).
To investigate the effects of mechanical mismatch between the electrodes and the cortical tissue, several groups have developed electrode substrates and substrate coatings from materials such as polyimide, polydimethylsiloxane (PDMS), and parylene, which are more compliant than materials traditionally used to create electrodes. However, these materials have had limited success in attenuating glial scar formation. A limitation to many of these studies is that such materials still have moduli 6 orders of magnitude larger than that of the brain, while also introducing different surface chemistries that could confound the tissue-material interaction (Lee et al., 2004; Nikles et al., 2003; Takeuchi et al., 2004; Wester et al., 2009; Takeuchi et al., 2005) and complicate the interpretation of these results. Of interest, surface treatments of electrode materials with a poly(vinyl alcohol)/poly(acrylic acid) hydrogel have reduced astrocyte reactivity, measured through the expression of glial fibrillary acidic protein (GFAP) (Lu et al., 2009).
Given the evidence for the contribution of mechanical force to the glial scar from in vitro, in silico, and in vivo experiments, we hypothesize that the in vivo glial scar
78 response is influenced by a chronic stiffness mismatch between the soft brain tissue and stiff intracortical implants. To create an in vivo device to characterize the importance of mechanical compliance of intracortical implants on both the chronic inflammatory reaction and the long-term electrode performance, we used our recently developed mechanically adaptive polymer nanocomposite, previously described in detail (Capadona et al., 2007; Capadona et al., 2008; Capadona et al., 2009; Shanmuganathan et al., 2010a, b, 2009). Our previous work has demonstrated that the polymer nanocomposite can reduce its tensile storage modulus from 5 GPa to 12 MPa within 15 minutes upon exposure to simulated physiological conditions in artificial cerebral spinal fluid (ACSF) at 37oC (Capadona et al., 2007; Capadona et al., 2008; Capadona et al., 2009;
Shanmuganathan et al., 2010a, b, 2009). The polymer nanocomposite has modest aqueous swelling of 60-75%, in comparison to the several hundred percent exhibited by hydrogel approaches. Here, we have created model electrodes from the adaptive polymer nanocomposite (NC) material. We have demonstrated that the nanocomposite material performs similar in vitro and in vivo (Harris et al., 2011b). While the previous work performed modest histological analysis, this is a more complete histological report on their use as cortical implants demonstrating that there are both foreign body elements, as well as mechanical effects, contributing to the overall tissue response at the biotic-abiotic interface. These components of the inflammatory response appear to be separate, but intricately coupled to the neuronal response.
79 4.3. Materials and Methods
4.3.1. Implant Fabrication and Imaging
Implants consisted of two types, referred to as nanocomposite (NC) and wire. NC implants were created by casting films from a dimethylformamide (DMF) solution of poly(vinyl acetate) and tunicate whiskers as reported previously (Capadona et al., 2008).
The NC had a cellulose whisker content of 15% w/w. The resulting films were used to create sheets via a custom mold in a hot press (Carver, Wabash, IN). Sheets were used to create 3 mm long probes that measured 100 ± 5 µm in thickness and 203 ± 19 µm in width with an approximately 45 ± 3o angle tip (Figure 14a) via a cutting process. The cutting process created a nanocomposite with two types of surfaces: two pressed sides
(Figure 14c) and two cut sides (Figure 14d). Both implant types were single shank probes with nearly the same cross-sectional area. Refer to (Harris et al., 2011b) for a characterization of implant mechanical properties of the NC material.
80
Figure 14. Materials bilaterally implanted in rodent cortex. a) Microscopic picture of PVAc-NC polymer nanocomposite implant, 3mm in length, 200µm wide, 100µm thick. b) Microscopic picture of PVAc-coated tungsten wire. The tungsten wire is 50µm in diameter before coating. The diameter after coating is ~160µm. Light white lines outline the wire within the PVAc coating. c-e) SEM images of implants. c) SEM image of pressed side of PVAc-NC implant. d) SEM image showing cut side of PVAc-NC implant that is rough. The image also shows the pressed side of the PVAc-NC implant at the bottom of the image. Inset, a higher magnification SEM image shows the cellulose whiskers on the cut side of the PVAc-NC implant. e) SEM image of the PVAc-coated tungsten wire. Wire implants consisted of 50 µm diameter tungsten wires (AM-Systems,
Sequim, WA) from which the Teflon insulation was mechanically stripped using cutting tweezers. Wire implants (Figure 14b) were dip-coated in a solution of poly (vinyl acetate) (PVAc, MW 113,000g/mol, density 1.19g/cm3, Sigma Aldrich) in toluene at a concentration of 10% w/w at 70oC. A droplet of PVAc was placed on the edge of a glass
dish and the tungsten wire was steadily dragged through the drop repeatedly about twenty
times or until the coated wire had a diameter of ~ 160 µm. Samples were then cut into
3mm long sections, and with the aid of a microscope, sections of uniform diameter of 160
µm were selected for implantation.
81 A scanning electron microscope (SEM) imaged the NC and wire implants at high
magnification. Both the NC and wire implants were covered with a thin layer of
Palladium, approximately 7-10 nm. The samples were then attached to a chuck with
Teflon tape and placed inside the S-4500 Hitachi SEM. Samples were imaged at various
magnifications and angles (Figure 14c-e).
4.3.2. Water Contact Angle
To analyze the surface hydrophilicity of the two implants, water contact angle measurements were performed on sheets of the nanocomposite and neat PVAc. Sheets of nanocomposite were made as described above, and the fabrication of neat polymer sheets was performed as described in previous literature (Harris et al., 2011b). The FTA 200
(First Ten Angstroms, Portsmouth, VA) instrument dispensed and imaged droplets on the sheets. Three to five samples from sheets were imaged, and the FTA32 software program analyzed the water contact angle. On the image, the user set the base of the droplet and one fourth of the envelope of the drop. The program calculated the angle using the base and envelope of water droplet.
4.3.3. Surgical Procedures
All surgeries were performed under sterile conditions with autoclaved instruments. Both wire and NC implants were sterilized via an autoclave before implantation. The modulus of NC implants was measured and verified to be unaltered by the autoclave process (data not shown). The same surgeon performed all surgeries while an assistant monitored anesthesia levels and manipulated equipment. All procedures were approved by the Case Western Reserve University Institutional Animal Care and Use
Committee (CWRU IACUC) and minimized pain and discomfort of the animal.
82 Male Sprague-Dawley rats from 206-335g were implanted bilaterally with a wire and a NC probe. The implants were affixed to sterile ceramic forceps tips (Fine Science
Tools, Foster City, CA) with glucose. Anesthesia was induced via an intraperitoneal injection of ketamine (80 mg/kg) and xylazine (10 mg/kg). After the animal reached the surgical plane, as determined by lack of response to toe pinch, the animal’s head was shaved and eyes protected via sterile ocular lubricant. A circulating water mat was placed underneath the animal while the animal was fitted with a thermometer and pulse oximeter
(Surgivet, Waukesha, WI) to monitor vitals. The animal was secured on a stereotaxic frame (Stoelting Co., Wood Dale, IL) fitted with a gas mask. The animal was ventilated
(Engler, Hialeah, FL) with oxygen; once the animal started to come out of the initial ketamine/xylazine induction (determined by whisking, reacting to toe pinch, raised heart beat), Isoflurane (1-3%) was introduced to oxygen flow to maintain a surgical level of anesthesia.
After securing the animal, marcaine (~0.1 mL, diluted 1:5 with saline) was injected before incision as a local anesthetic, and the head area was sterilized with betadine and alcohol. An incision along the midline exposed the skull. Skin was retracted using an eye speculum. Bilaterally, a 3 mm biopsy punch (Miltex, York, PA) was used to drill 2 implant sites centered 4.5 mm caudal to the bregma, 3 mm lateral to the midline. The biopsy punch prevented heat damage caused by motorized drills and provided a well- defined 3 mm opening. Three screws were implanted for later fixation of dental cement.
For both openings, one screw was placed 1 mm caudal to the biopsy punch-produced openings. The remaining screw was placed 4mm rostral to the biopsy hole on the animal’s left side. After three 0-80 X 3/32 screws were attached to the skull, a fine 45o
83 angle microprobe (Fine Science Tools, Foster City, CA) was used to create an opening in
the dura (approximately 1 mm diameter) for both implant sites. After manually lowering
the implant to the surface of the brain, an automated mechanical inserter controlled by
LabView 7.1 (National Instruments, Austin, TX) inserted the implants 2 mm into the
tissue at a rate of 2 mm/s. A 1 mm tab was left exposed above the brain tissue to enable
tethering via Kwik-Sil (WPI Inc, Sarasota, FL). Kwik-Sil sealed the implant sites, and
dental cement (Stoelting, Wood Dale, IL) was applied to cover the implant sites and
screws. The skin was sutured with 5-0 polypro sutures (Henry Schein, 1016409,
Melville, NY) enclosing the dental cement headcap. Each animal received one wire and one NC implant in random order and side placement.
After surgery, an animal was isolated in a clean cage on a warm water blanket and
monitored every thirty minutes until sternal. One antibiotic shot of cefazolin (20 mg/kg)
and one analgesic shot of buprenorphine (0.01 mg/kg) were given on the day of surgery.
Two cefazolin shots, same dose, were given the day following surgery, while
buprenorphine was given as needed. No steroids were applied at any time throughout the
implant period. Each rat was monitored regularly after surgery for signs of distress,
weight loss, and dehydration. In rare cases, complications with infections and sutures
caused early euthanasia and these animals were removed from the study. In the
remaining animals, some animals exhibited a large void. The response was analyzed and
determined to be resulting from excessive bleeding after implantation. These
hemispheres were excluded from analysis since the response was due to insertion trauma
rather than the response to the implant. In the remaining animals, some animals exhibited a large void around the implant. The response was analyzed and determined to be a result
84 of excessive bleeding after implantation. These hemispheres were excluded from analysis
since the response was due to insertion trauma rather than the response to the implant.
4.3.4. Fixation, Tissue Preparation, and Immunohistochemistry
After the designated timeframe, four or eight weeks, animals were euthanized by
transcardial perfusion. Animals were first anesthetized to a surgical plane as described
above. Perfusion rate was controlled via a mechanical pump (VWR, West Chester, PA)
with 500 mL of D-PBS (Invitrogen, Carlsbad, CA) or when no trace of blood was seen
exiting the right atrium. Approximately 250 mL of 10% buffered formalin (Fisher,
Pittsburgh, PA) fixed the tissue of the animal. The brain and probes were placed in
formalin for 24 hours, after which the brain was then transferred to fresh formalin. The
brain remained in fresh formalin for up to one week, until it was cryoprotected in 30%
sucrose solution.
After removing the probes, brains were frozen in embedding molds (Electron
Microscopy Products, Hatfield, PA) containing OCT compound (Sakura, Tokyo, Japan)
and mounted and sectioned horizontally by a cryostat (Thermo Fisher Scientific,
Waltham, MA) creating 30 µm thick slices. Slices chosen were spread over the whole
length of the implant while remaining slices were preserved in cryoprotectant solution
(Watson et al., 1986). Following procedures outlined previously (Biran et al., 2005),
slices were labeled via a free-floating protocol.
Briefly, slices were placed in blocking solution overnight at 4oC. Blocking solution was composed of 4% (v/v) normal goat serum (Sigma-Aldrich, St. Louis, MO),
0.3% (v/v) Triton-X-100 (Sigma-Aldrich), and 0.1% (w/v) sodium azide (Sigma-Aldrich)
85 in 1x D-PBS (Invitrogen, Carlsbad, CA). Slices were then placed in fresh blocking
solution with primary antibodies (Table 3) overnight at 4oC. On the third day slices were placed in blocking solution with secondary antibodies, goat anti-rabbit IgG or anti-mouse
IgG1 or IgM, Alexa Fluor 594, 488, and DAPI dilactate (Invitrogen, Carlsbad, CA) for one hour at room temperature. After primary and secondary application periods, three, 15 minute washes in PBS were performed on a shaker. Slices were mounted and cover slipped using Fluoromount-G (Southern Biotech, Birmingham, AL).
Table 3. Summary of Primary Antibodies.
Antibody Vendor (Part #) Dilution Isotype Cell Type/Labeling
GFAP Chemicon 1:1000 Rb IgG Astrocytes (AB5804)
NeuN Chemicon 1:500 Ms IgG1 Neurons (MAB377)
CD68 Chemicon 1:250 Ms IgG1 Activated Macrophages, (ED1) (MAB1435) Microglia
IBA1 Wako 1:615 Rb IgG Microglia (019-19741)
CS56 Sigma 1:500 Ms IgM Chondroitin Sulfate (C8035) Proteoglycans
Vimentin Sigma 1:200 Ms IgG1 Varied- Immature/reactive (V6389) astrocytes, microglia, endothelial cells, and fibroblasts
86 4.3.5. Image Acquisition and Analysis
Images of slices were taken by an Axio Imager.Z1M Microscope and AxioCam
MRm CCD camera (Zeiss, Thornwood, NY). Implant sites were centered in the field,
and exposure settings were set to prevent overexposure. The settings were then
maintained the same for the whole set of stained images. Four automated exposures were
performed on every site: one bright field and 3 fluorescent colours. Seventeen animals
with bilateral implants were analyzed at four weeks. After removing hemispheres with
excessive insertion trauma, there were 11 trials for the wire and 16 for the NC. Seven
animals were analyzed after eight weeks. Statistical analysis based on prior work
predicted that about 10 animals were needed for analysis. On average, three slices were
used per animal for a given label.
A custom Matlab R2009b (MathWorks, Natick, MA) script was developed to
measure the intensity around the implant. The script automatically detected the border
and area of the implant based on the bright field image. Bright field images were used
for border detection since it was difficult to see the actual border in fluorescent images.
Therefore, some fluorescent images appeared to have a larger implant area than actual.
This effect and the variable shrinkage of tissue during histology contributed to the
apparent variation in size of implants pictured in fluorescent images. Calculating the
size of the implants in images, the average sizes of the implants were similar. 23,329 ±
3279 µm2 for NC and 24,513 ± 5264 µm2 for wire.
One hundred, equally spaced, radial lines emanated from the centroid of the
implant area. For each one of the 100 rays, the intensity (pixel values) over the first 200
microns immediately external to the border were measured around the implant site
87 (Figure 15a). The average intensity of the 100 rays was divided by the background
measurement of the image to account for differences between images due to
photobleaching and other effects inherent in fluorescent immunohistochemistry.
Background intensity values were measured per image at the edge of the image, at least
300 µm from the implant and outside the region measured for the response. Several
measurements were made for each image including peak intensity, near integral intensity
(0-50 µm), and far integral intensity (50-100 µm) as well as certain intervals dictated by the data and highlighted in the text.
For certain IHC labels, an additional evaluation of IHC intensity was performed to determine the effect of the nanocomposite side on the response. A subset of the 100 rays was averaged to measure intensity of the pressed sides versus the cut sides of the nanocomposite. Based on the cutting technique used to create the probes, the pressed side was the longer of the two sides. The majority of images analyzed had clearly discernable long and short sides. Images with indistinguishable sides were left out of analysis. Approximately 10 of the 100 rays were used to determine image intensity based on a single side. For each slice, the two long sides were averaged together and the two short sides were averaged together. Therefore, approximately 20 of the 100 rays determined the reaction to the cut sides, and a separate 20 of the rays determined the reaction for the pressed sides. The first 20 µm and 50 µm of the intensity were examined for DAPI and ED1 to examine the effect of the surface roughness on the tissue response.
A mean intensity was computed over the distance as well as the 95% confidence interval of the mean intensity. DAPI analysis was performed to analyze the total cell density
88 around the implants. Analysis of ED1 was performed since previous literature has
indicated that ED1 positive cells are often closest to the implant (Biran et al., 2005).
The automatic detection removed human error across samples, but could result in
the inclusion of up to 3-5 µm of the image without tissue, which could slightly distort data in this range. The distance between the implant and the peak tissue response error is within 5 µm, but is consistent between the experimental conditions and does not affect the results of this work.
Figure 15. Quantification of fluorescent immunohistochemistry (fIHC) staining via two methods. Horizontal slices of representative brains shown. a) Representative image indicating the 100 radial lines drawn to compute the pixel intensity radiating outward from electrode – tissue interface. b) Representative image showing computer and user selected cell bodies to compute the distance of cell bodies from the electrode – tissue interface. Inset, Close-up of marked cell bodies. Another custom Matlab script used the same border detection algorithm, but computationally isolated and highlighted nuclei (Figure 15b). After the automated marking, an investigator manually reviewed the identification and added or subtracted nuclei missed by the algorithm. After marking, Matlab computed the distance from each nucleus to the border along the ray emanating from the centroid of the implant site.
These distances were binned into 10 µm increments, and a histogram was created for the
89 number of nuclei. The counts were scaled by dividing by the area of the corresponding
10 µm-wide concentric polygonal donut to provide the neuron density.
4.3.6. Statistics
In order to estimate the effect of the implant type on the outcome, the differences between measured values (e.g. the peak intensity, integrals, cell count, etc.) were modeled using a multilevel model. The model (Equation 3) included a baseline level for
the outcome (!0), a fixed effect for the difference in the outcome based on the implant
type (!1), a random effect for each animal (b0,i), a random effect for each slice within each
animal (c0,ij), and residual error (eijk):
Outcomeijk = "0 + "1 * IMPLANT _ TYPEijk + b0,i + c0,ij + eijk (3)
where i indexes the rat, j indexes the slice, and k indexes the side. If !1 was significantly
different than zero, the implant had a significant effect on the outcome of interest. All
statistical analysis was completed in the statistical package R (R Foundation for
Statistical Computing, Vienna, Austria) (R Development Core Team, 2009; Pinheiro et
al., 2009).
4.4. Results
4.4.1. Water Contract Angle
Water contact angle measures show that there is no significant difference between
the hydrophobicity of the PVAc wire coating and the PVAc-NC material being studied.
The nanocomposite had a contact angle of 70.3 ± 2.2o and the neat material had a contact
angle of 64.0 ± 3.1o. A two sample t-test did not show any significant differences in the contact angle (p=0.22).
90 4.4.2. Neuronal Nuclei
NeuN (neuronal nuclei) labeled cells showed neuronal nuclei around the probe
(Figure 16). There is a reduction of neuronal density towards the tissue-implant border for both implants. At four weeks, however, the neuronal density within 200 µm of the nanocomposite (NC) implant is significantly greater than it is around the wire (Figure
16a,b). At eight weeks, the neuronal density around both implants types is no longer significantly different at any distance from the implant (Figure 16a,b). For the wire, the neuronal density within a radius of 20 µm from the interface is significantly greater at eight weeks than four weeks (p=0.04). The data at other intervals is not statistically significant, but trends indicate that neuronal density around rigid wire implants increases between four and eight weeks to a density similar to the NC at both four and eight weeks.
Further, there are no trends in decreasing neuronal nuclei density between four and eight weeks with the NC.
Figure 16. Analysis of NeuN Immunohistochemistry. a) Quantification of NeuN based on distance from the tissue-implant border in response to implants: four week NC (square), four week wire (asterisk), eight week NC (cross), eight week wire (circle). Points represent histogram counts in 10 µm intervals. Counts have been scaled based on
91 area as well as background count per image. b) Average count of NeuN as a function of distance from the tissue-implant border ± standard error. Four week NC (white), four week wire (black), eight week NC (gray), eight week wire (striped) *= p<0.05 for intraweek comparisons, **<0.05 for interweek comparisons. Representative images of NeuN for c) four week NC, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm.
4.4.3. Astrocytes
Analysis of GFAP labeling indicates the reactive astrocytic response. The
intensity of GFAP labeling peaks approximately 10-20 µm from the tissue-implant border and gradually decreases further away (Figure 17a,b). For both four and eight week data, the 0-50 µm region from the implant exhibits hypertrophied astrocytes with a reactive morphology (Figure 17c-f, c inset). Astrocytes outside of 200 µm do not exhibit GFAP labeling (Figure 17c-f). At four weeks, the maximum intensity and 0-50 µm integral of intensity are not significantly different between tissue around wire and NC implants, though the 50-100 µm intensity integral is significantly greater (p=0.05) for wire samples compared to NC at four weeks. At eight weeks, the intensity maximum is significantly greater around the wire implants than the NC implants (p=0.04), and the intensity integral from 25-100 µm around wire implants is significantly less than the NC (p=0.03). There are no statistical differences for GFAP intensity with wire samples between four and eight weeks, but the labeling for GFAP around the nanocomposite at eight weeks is significantly greater over the 20-80 µm interval than at the four week timepoint (p=0.03).
Additionally, GFAP labeling tends to be more diffuse or less compact around the NC than the wire at eight weeks (Figure 17b, e, f).
92
Figure 17. Analysis of GFAP-Reactive Astrocytes as a function of distance from the tissue-implant border. a) Relative intensity ± standard error as a function of distance from border of four week NC (gray) and wire (black). Brackets indicate an integral of intensity. b) Relative intensity ± standard error as a function of distance from border of eight week NC (gray) and wire (black). Representative images of GFAP for c) four week NC, inset 50µm wide close-up of reactive astrocyte, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm. *=p<0.05.
4.4.4. Proteoglycan
Chondroitin Sulfate Proteoglycan (CSPG) is a structural component of extracellular matrix and its upregulation within the glial scar has been linked to inhibition of axonal outgrowth (Silver and Miller, 2004). Intensity for CSPGs, via labeling for
CS56 antibodies (Table 3), peaks within 15 µm from the border (Figure 18a-f). The peak around the NC implant returns to baseline by 50 µm for the four and eight week samples implanted with NC; levels remain elevated around the four week wire implant up to
150 µm. The wire implants at eight weeks feature a response that returns to baseline by
100 µm from the tissue-implant border (Figure 18b,e,f). CS56 labeling at four weeks was significantly greater around the wire than NC starting at 10 µm and extending out to 150
µm (p<0.01) (Figure 18a,c,d). At eight weeks, the peaks and integrals were similar for tissue around NC and wire implants.
93
Figure 18. Analysis of CS56-CSPGs as a function of distance. a) Relative intensity ± standard error as a function of distance from border of four week NC (gray) and wire (black). Brackets indicate an interval of integral of intensity calculation. b) Relative intensity ± standard error as a function of distance from border of eight week NC (gray) and wire (black). Representative images of CS56 for c) four week NC, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm.*=p<0.05.
4.4.5. Vimentin
Vimentin labeling is associated with immature astrocytes, extracellular matrix
components, and developing axons (Alonso, 2005; Pekny, 2003; Levin et al., 2009). At
both timepoints, vimentin expression is upregulated with a peak at 10 µm from the
implant (Figure 19a-f) and decreases to baseline levels by 100-150 µm from the implant
border. At four weeks (Figure 19a,c,d), the vimentin intensity maximum and 0-32 µm intensity integral for NC implants is significantly greater than for the wire implants
(p<0.01). Note that at 32 µm there is an intersection of the two implant profiles. At eight weeks (Figure 19b,e,f), measures of vimentin labeling are not significantly different between the NC and wire. In the interval of 10-150 µm from the tissue-implant border, the intensity integral around the wire significantly decreased from four to eight weeks
(p=0.03).
94
Figure 19. Analysis of vimentin as a function of distance. a) Relative intensity ± standard error as a function of distance from border of four week NC (gray) and wire (black). Brackets indicate an integral of intensity. b) Relative intensity ± standard error as a function of distance from border of eight week NC (gray) and wire (black). Representative images of vimentin for c) four week NC, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm.*= p<0.05.
4.4.6. Macrophages/Microglia
Analysis of IBA1 and ED1 labeling quantified the density of activated microglia and macrophages. IBA1 labels all microglia (Figure 20a-f) and ED1 labels (Figure 21a-f) activated microglia and activated macrophages. Microglia as shown by IBA1 labeling appeared punctate or amoeboid near the border and a ramified morphology away from the implant site. The microglia response was characterized by intense IBA1 labeling 10
µm from the border that rapidly decreased to near background levels at more than 100
µm from the border. ED1 labeling (Figure 21a-f) exhibited similar features but was limited mainly to amoeboid cells close to the implant interface.
The four and eight week IBA1 maximum intensity (Figure 20a-f) from tissue around the NC is significantly greater than the intensity from tissue with wire implants at the same time point (p=0.02, p<0.01). The 0-150 µm intensity integral is significantly greater from samples around the eight week NC than from the eight week wire (Figure
95 20b,e,f). Response changes for a particular implant type are not significant between four
and eight weeks.
The peak intensities of ED1 at four and eight weeks are not significantly different
between the NC and the wire (Figure 21a-f), but the 17-150 µm intensity integral is significantly greater for wire implants compared to NC at four weeks (p=0.03). The 0-
150 µm ED1 intensity integral significantly decreases for the wire from four to eight weeks (p=0.04). The 0-50 µm ED1 intensity integral is significantly greater for the nanocomposite compared to wire implants at eight weeks (p=0.03) (Figure 21b,e,f).
Figure 20. Analysis of IBA1-Microglia immunohistochemistry as a function of distance from the tissue-implant border. a) Relative intensity ± standard error as a function of distance from border of four week NC (gray) and wire (black). Brackets indicate an integral of intensity. b) Relative intensity ± standard error as a function of distance from border of eight week NC (gray) and wire (black). Representative images of IBA1 for c) four week NC, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm. *= p<0.05.
96
Figure 21. Analysis of ED1-Activated Macrophage and Microglia immunohistochemistry as a function of distance from the tissue-implant border. a) Relative intensity ± standard error as a function of distance from border of four week NC (gray) and wire (black). Brackets indicate an integral of intensity. b) Relative intensity ± standard error as a function of distance from border of eight week NC (gray) and wire (black). Representative images of vimentin for c) four week NC, d) four week wire, e) eight week NC, and f) eight week wire. Scale bars 100 µm. *= p<0.05.
4.4.7. IHC intensity based on NC side
To examine the effect of surface roughness on the IHC response, we examined
DAPI and ED1 labeling across many animals. Several different examinations of intensity were examined, but no examination yielded significant effects based on side. The complete numbers are shown in Table 4. Similar intensity between sides indicates that the difference of mean intensities between sides is less than the 95% confidence interval.
A sign test for the paired data was performed to determine if the there was a statistically significant difference based on side. The results indicate that side does not have a significant effect on the tissue response.
Table 4. Summary of IHC analysis when subdivided based on side. The number in the table represents the number of sections analyzed that showed a measurable difference in normalized stain intensity for one side or the other. The results are further divided into measures within the first 50 µm and the first 20 µm. These correspond to the distance related to unit recordings and approximately the first 3-4 cell layers, respectively. There
97 does not appear to be any particular bias in the results to one side or the other and that surface roughness is not significantly contributing to the response.
Label DAPI ED1 Greater Long Short No Long Short No side side side Difference Side side Difference Within 13 12 13 6 13 12 50 µm Within 10 9 19 7 12 12 20 µm
4.5. Discussion
The results presented here support the hypothesis that the mechanical mismatch between brain tissue and microelectrodes influences the inflammatory response.
Specifically, we found that the implant stiffness affects the neuronal response and composition of the glial scar, as the NC implanted animals had a significantly greater neuronal density than the wire implants at four weeks. These results aligned with previous studies that reported neuronal nuclei density was reduced around the implant in conjunction with similar up-regulation of identical glial scar factors (Winslow and
Tresco, 2009; Biran et al., 2005; Zhong and Bellamkonda, 2007).
In this work, we implanted nanocomposite probes and neat PVAc coated tungsten wires. To confirm that surface roughness and morphology were not factors in the response, we analyzed the nanocomposite and neat PVAc. Examination via SEM showed that the neat polymer had a smooth surface (Figure 14e), and the nanocomposite had two sides of different roughness. The pressed side of the nanocomposite had a smooth surface with only minor indentations (Figure 14c). The cut side of the nanocomposite had a rough surface (Figure 14d). Under 50,000x magnification (Figure 14d, inset), the
98 tunicate whiskers of the nanocomposite were clearly evident and protruded on the cut edge.
Next, we examined whether differences in surface roughness were contributing to the tissue response. There are two distinct sides with different surface roughness, and we investigated whether the side affected the tissue response. The examination of the response based on side indicated that the response was similar between the two sides of the nanocomposite. These results support that surface roughness did not affect the tissue response. From SEM images, the surface roughness differences between wire and nanocomposite were much less than the differences between sides of the nanocomposite.
Collectively, the data suggests that our observed differences in tissue response are not due to differences in surface roughness. Our findings agree with previous research that has shown that the surface roughness of implant does not effect the tissue response past 4 weeks (Szarowski et al., 2003).
The tissue response in our results was similar to previous reports (Winslow et al.,
2010; Winslow and Tresco, 2009). However, in our case, the classical markers of the glial scar (GFAP and ED1) do not fully explain the differences seen in neuronal density around each of the implant types at four weeks. Though the softer nanocomposite implant had a significantly greater neuronal density around the probe than the rigid wire after four weeks, there was not a significant difference in GFAP or ED1 labeling (peak or 0-50 µm interval). In the 0-50 µm interval nearest to the device, considered critical for unit recording, the brain tissue around the NC implants had nearly twice the neuronal density than that around the wire implants at four weeks that may improve recording of single units from neurons located close to the implant at that time period (Buzsáki, 2004).
99 Examining the correspondence of GFAP and neural density in our results, GFAP
did not account for all the differences seen between neuronal densities around the implant
types. Specifically at four weeks, the GFAP peak intensity and the 0-22 µm intensity integrals were similar for both implant types while there was a significant difference over the same range for neuronal density. This may be due to the wide assortment of roles, both positive and negative associated with astrocytes. Astrocytes have been implicated in several different functions associated with cytoskeletal structure including neurotransmitter maintenance and blood brain barrier function (Simard et al., 2003;
Ortinski et al., 2010). GFAP positive cell functions can be neuroprotective, as transgenic animals without GFAP show a larger functional deficit after injury (Faulkner et al.,
2004). Detrimental to neural protection, astrocytes have been implicated as a diffusion barrier (Roitbak and Sykova, 1999), remyelination limiter (Fawcett and Asher, 1999), and producer of cytokines and extracellular matrix components such as fibronectin
(Shearer et al., 2003) and CSPGs (McKeon et al., 1999). Since astrocytes have been tied to many structural components including cytoskeletal arrangement and secretion of ECM components, it is believed that GFAP would be a key component in mechanical signaling. Therefore, it is notable that the rigid wire created a wider radius of GFAP activation at four weeks, but a more compact scar at eight weeks. A wider radius is suggested by a nearly significant difference over 22-150 µm intensity interval (p=0.06) and a significant difference over the 50-100 µm intensity integral (p=0.05) at four weeks
(Figure 17a,c,d). Further, the peak intensity of GFAP expression was larger for the wire
(p=0.04), yet the distribution of GFAP expression was broader for the NC at eight weeks
(25-100 µm, p = 0.03, Figure 17b,e,f). From four to eight weeks, the neuronal response
100 (NeuN) and GFAP labeling around the NC implant was unchanged. This suggests the
NC has formed a static scar. Yet, with wire implants, the GFAP expression at eight weeks became more compact, and NeuN positive cells in the same region of tissue appeared to increase around wire implants. The results from wire implants suggest that the scar is continuing to compact and remodel in response to the greater mechanical stiffness.
Cytoskeletal components play a major role in the ability of a cell to interface with and respond to the extracellular environment. The type and quantity of the filament are important in determining the ability of the cell to respond to the mechanical environment and move through the extracellular space (Alberts, 1994). In addition to GFAP, vimentin is an intermediate filament expressed in astrocytes and was investigated here (see (Pekny,
2003) for a review). In contrast to GFAP, the peak intensity and 0-32 µm intensity integral for vimentin labeling were significantly greater at four weeks for the NC, compared to the wire. Historically, vimentin has been tied to immature astrocytes, but recent research has linked vimentin expression to rapid neurite extension in response to damage (Levin et al., 2009). Likewise, NG2+ cells that also express vimentin have been proposed to support repair of central nervous system (CNS) damage, and stabilize axons in response to dieback from ED1+ cells (Alonso, 2005; Nishiyama, 2007; Busch et al.,
2010). Here, the increase in vimentin labeling around four week implants was associated with more neurons around NC implants than wire implants (Figure 19a,c,d, Figure
16a,c,d). This is consistent with potential neuroprotective implications for vimentin positive cells. Thus, we believe the correlation is due to positive effects associated with the intermediate filament vimentin. At eight weeks, the contrast in vimentin levels was no longer evident, which also correlates with similarity in neuronal densities around each
101 implant type, further supporting the role of the intermediate filament, vimentin in supporting the neuronal density around the cortical implant. We believe that these results suggest that the expression of intermediate filaments in the tissue surrounding our implants is related to the mechanical properties of the electrode material.
To further examine the role of mechanics in the wound healing response, we examined another component of the glial scar, proteoglycans. Proteoglycans are negatively charged polysaccharide chains, and are an important structural constituent of the extracellular matrix (ECM) (Alberts, 1994). When found in elevated concentrations within tissue, the high density of negative charges found across their backbone have been shown to sequester cations, causing large amounts of water to diffuse out of adjacent tissue, into the proteoglycan rich areas. This phenomenon creates a swelling pressure that enables the tissue to withstand compressive forces. A subclass of proteoglycans secreted by astrocytes, Chondroitin Sulfate Proteoglycan (CSPG), has been shown to be deposited by astrocytes in gradients of increasing concentration towards the injury site
(Silver and Miller, 2004). Interestingly, our results indicated that the area with elevated
CS56 labeling (label for CSPGs,) is much larger around the wire implants than around the NC at four weeks (Figure 18a,c,d). This is in agreement with previous research demonstrating that CSPGs inhibit neuronal outgrowth (Silver and Miller, 2004; Busch and Silver, 2007; Fawcett and Asher, 1999), further supported by our four week NeuN staining (Figure 16a,b,c,d). Additionally, in vitro experiments have shown increased
CSPG labeling in response to increased strain (Cullen et al., 2007), further supporting our hypothesis that the mechanical mismatch between brain tissue and microelectrodes influences the inflammatory response.
102 While our results provide interesting insights into the NC, reduced mechanical
forces, and improved neuronal densities, we cannot ignore the cofounding effects of the
inflammatory mediating microglia cells. Microglia and macrophage cells mediate the
inflammatory and immune response to minimize bacterial/viral invaders (Kreutzberg,
1996). Macrophages have also been implicated in axonal dieback (Horn et al., 2008).
Many studies of the tissue response to intracortical electrodes have focused on
macrophage activation in response to indwelling implants (Biran et al., 2005). Activated
macrophages and microglia, as labeled by ED1, were similar for peak intensity and the 0-
17 µm (border to divergence point) intensity integral at four weeks (Figure 21a,c,d).
Similar to the GFAP and CSPG responses, there is a greater radius of ED1 activation in
tissue around the wire implants at four weeks; the 17-150 µm intensity integral was greater for wire implants at four weeks. At four weeks, increased ED1 activation can be loosely associated with decreased NeuN labeling over similar distance from the electrode tissue interface. At eight weeks, ED1 activity was less pronounced around wire implants
(Figure 21b,e,f), which corresponded to an apparent recovery in NeuN positive tissue.
While ED1 labels only activated macrophages and microglia, IBA1 marks all microglia and macrophages. The four and eight week IBA1 maximum intensity (Figure
20a-f) from tissue around the NC was significantly greater than the intensity from tissue with wire implants at the same time point (p=0.02, p<0.01). Differences between IBA1 and ED1 labeling can therefore be attributed to the presence of either resting microglia or a subset of beneficial, “alternatively activated” (M2), macrophages (Kigerl et al., 2009).
The ED1 peak intensity was similar between the two implant types at four weeks, but the
IBA1 peak for NC was 18% greater than the wire, indicating the presence of microglia
103 cells independent of levels of activated cells (ED1) (Figure 20a,c,d, Figure 21a,c,d). For
the NC implants, the presence of additional non-ED1 reactive microglia and higher
neuronal density at four weeks may indicate a beneficial role of microglia. Regardless of
the state of the macrophages/microglia, the results suggest an ongoing reaction for both
implant types since activated macrophages should return to baseline 21 days after a stab
injury to cortex (Fujita et al., 1998; Nisbet et al., 2009). Our results are consistent with
research showing a sustained response to indwelling implants (Winslow et al., 2010;
McConnell et al., 2009).
Indicators of the glial scar components suggest that mechanics continues to
modify the glial scar at eight weeks. Although, the mechanical-associated labels, GFAP
and CS56, no longer show a similar increased radius of activation around wire implants
at eight weeks, GFAP labeling (Figure 17b,e,f) around wire implants exhibit a more
compact, higher peak than the compliant NC implants. Prior work (Frampton et al., 2010)
hypothesizes that a more compact GFAP response increases the impedance of an
electrode which may decrease the quality of electrode recordings. Aligning with previous
research (Fujita et al., 1998; Nisbet et al., 2009), both implant types appear to lower ED1
activation from four to eight weeks (Figure 21a-f) suggesting an ongoing reaction that lessens over time. The 0-50 µm intensity integral for ED1 and IBA1 was significantly greater around NC implants than the wire implants at eight weeks (Figure 20b,e,f and
Figure 21b,e,f), but further work is needed to determine whether the difference is neuroprotective, neurotoxic, or neither at longer time points. Regardless of the similar neural densities at eight weeks, the nanocomposite continued to modify the overall tissue response up to eight weeks.
104 Another possible interpretation of the data is that softer implants are affecting the time-course of the response rather that final results, and hence, similar neuronal densities at eight weeks. It is known that the insertion event causes trauma, but it is unknown how stiffness might affect the repair of this trauma. Previous research has shown the response past four weeks is independent of size, shape, and surface roughness (Szarowski et al.,
2003), but the softer nanocomposite may allow for quicker repair of insertion damage.
The increased levels of vimentin and IBA1 at four weeks could indicate a more robust healing process. Additionally, the similar neuronal densities between NC and wire at eight weeks coupled with the recovery of the neuronal density around wire implants to nanocomposite levels could suggest a quicker repair around NC implants. In this paper, our goal was to demonstrate that there is an effect based on probe stiffness. Further experiments are necessary to clearly describe the time course of the response and the specific effects of probe mechanics on that time course.
In summary, the four and eight week results show that the nanocomposite material modifies the glial scar and neuronal density around implants in comparison to stiffer wire implants. Four week nanocomposite implants feature greater neuronal densities than wire implants, and several glial scar labels such as GFAP, CSPG, and ED1 show a larger radius of activation from the implant-tissue border around wire implants than NC implants. Notably GFAP and CSPGs, structural components, suggest the involvement of mechanics in the tissue response. Eight week results provide further evidence that the response to implants is ongoing, dynamic, and modified by the nanocomposite. Results show that the nanocomposite can maintain its neuronal density for at least eight weeks despite macrophage/microglia activation.
105 4.6. Conclusions
By utilizing our mechanically adaptive polymer NC materials, it is possible to separate the foreign body response to the implanted electrodes from the effect of material stiffness. While mechanical properties of the materials correlate to differences between neuronal densities at four weeks, we also demonstrate that the NC parallels the response to the field-leading wire at eight weeks (Winslow and Tresco, 2009; Winslow et al.,
2010). Our data also suggest the importance of controlling vimentin and CSPG response to improve neuronal density. Through the use of the mechanically adaptive nanocomposite, we demonstrate a system that can decouple the mechanical and surface chemistry components of the neural response. We expect this to be a valuable tool in future research regarding cortical tissue. Our findings indicate two parallel tracks to capitalize on our work: 1) Further development of advanced materials to modulate the response to mechanical and molecular factors and 2) Research into mechanisms and neuroprotective pathways, specifically M2 macrophage/microglia presence, CSPG reduction, and vimentin upregulation.
4.7. Acknowledgements
This work was supported by Grant Number R21-NS053798 and F31-NS063640 from the National Institute of Neurological Disorders and Stroke and T32-EB004314-06 for the National Institute of Biomedical Imaging and Bioengineering. Additional support was from Veteran's Affairs grants C3819C, F4827H and B6344W. The authors acknowledge the support of Anne DeChant. None of the funding sources aided in the collection, analysis, and interpretation of data, in the writing of the report, or in the
106 decision to submit the paper for publication. The authors have no conflicts of interest related to this work to disclose.
107
CHAPTER 5. LPS-induced inflammation degrades neural recordings
5.1. Abstract
Intracortical electrodes and the brain-machine interfaces (BMI) utilizing them offer the ability to restore brain-controlled movement for patients with reduced motor function. Researchers have noted that implanted electrodes are unable record neuronal signals for long periods of time, and research has associated the glial scar with the degradation in performance of arrays. The functional impact of the tissue response on electrode recordings is not well studied in literature. The intent of this work is to examine the effect of the tissue response on recordings by stimulating an inflammatory response with lipopolysaccharide (LPS). Our hypothesis is that an increased response will decrease the quality of intracortical recordings. Male Sprague-Dawley rats were implanted with four shank silicon electrodes. Recordings were made with the electrodes periodically over four weeks for evoked and non-evoked neural signals. After four weeks, animals were perfused, and brain slices were labeled via florescent immunohistochemistry (fIHC) for neuronal cell bodies, dendrites, reactive astrocytes, microglia, and activated macrophages. Results indicate that LPS-treated animals had an elevated cellular response at four weeks as evidenced by DAPI. Labeling indicated that the LPS activation of macrophages and microglia was not evident at the four week point. Increased impedance measurements at four weeks aligned with increased non-neural cell density. The neural and dendritic density around LPS-treated animals was reduced. Correlated with neural labeling, neural recording quality at the four weeks was decreased in LPS-treated animals as shown by evoked recording quality. The noise levels in LPS-treated animals were
108 increased, consistent with the increased impedance. The results support the hypothesis that the increased tissue response decreased recording performance.
5.2. Introduction
Much attention has been garnered by the ability of brain computer interfaces to increasingly perform complex tasks, as illustrated by the ability of a monkey BMI to control a robot arm to feed himself (Velliste et al., 2008). The technology of brain computer interface (BCI) would allow a user to transform thoughts into action, whether it is a computer cursor, robotic arm, and even their own muscles (Hochberg et al., 2006).
The BCI modality offers much hope for many patients that do not have full use of their extremities, either from spinal cord injury, brain stem stroke, ALS, amputation, or other neurological conditions. While BCI devices have performed well in the laboratory setting, improvement is needed for full-scale clinical deployment for these patient populations.
Electrodes to record the brain signals are the foundation of the BCI system.
Electrodes used in BCIs come in different shapes and sizes (Kipke et al., 2008; Ward et al., 2009), including EEG, ECoG, and intracortical electrodes going from least to most spatially selective (Grill et al., 2009). While the invasiveness of intracortical electrodes presents issues, intracortical electrodes have garnered the most attention in BCI work to allow for control of complex actions because of their ability to capture more spatially precise signals. A recent paper from the Hochberg and Donoghue (Simeral et al., 2011b) features a lengthy discussion highlighting the performance advantages of intracortical electrodes versus other less spatially selective electrodes. The recent paper also
109 highlights that the implanted intracortical electrode array is functional after 1000 days, a
milestone in human implants.
The publishing record of intracortical electrodes, however, indicates longevity
rarely lasts out to 1000 days, and often only 40-60% of the implanted electrodes record after implantation (Schwartz, 2004). A recent literature review examined the published record of many types of intracortical electrodes. The research indicated that metal microwires from the early 1980’s showed the best performance, but the longevity of any electrode technology rarely lasted more than a couple of years, even in exceptional cases
(Tresco and Winslow, 2011). While certain electrodes may last to 1000 days, there is a decrease in performance of those electrode arrays over time. Even in the case of the human electrode implant that lasted 1000 days (2.7 years), the number of neuronal signals, i.e. units, decreased over time. The subject had ~160 units at day 50 but only
~24 units at day 1000 (Simeral et al., 2011b; Kim et al., 2008). Though the 24 units were still able to perform 2D control of a computer cursor, other research has indicated that more units will be required for more complex functions (Velliste et al., 2008; Taylor et
al., 2003).
Most research with intracortical electrodes used them to monitor action potentials
from single neurons or small sets of neurons, but intracortical electrodes were also used
to monitor local field potentials (LFPs). LFPs captured the dendritic activity of neurons
while single unit action potentials were recorded from the neural cell soma. Additionally,
LFPs covered a spatial range out to 250 µm while single unit action potentials were
recorded from up to 50 - 140 µm (Henze et al., 2000; Katzner et al., 2009). In addition to
110 recording a different signal, data supports the hypothesis that LFPs may be more resistant
to the tissue response than single unit action potentials.
Several researchers have postulated why electrode performance decreased over
time, but the majority of researchers interpreted previous data to indicate that the tissue
response to the implanted electrode caused recording degradation. The degradation of recordings was usually associated with one of three items associated with the tissue response 1) decreased neuronal proximity, 2) increased tissue impedance, and 3) reduced neural connectivity, function, or health (McConnell, 2008). Electrode implantation caused damage to the neovascular unit (Bjornsson et al., 2006; Kozai et al., 2010b) that initiated an acute response to the injury (Szarowski et al., 2003; Shain et al., 2003;
Turner et al., 1999). If the electrode was immediately removed, as in a stab wound, the injury was usually repaired in approximately 4 weeks (Biran et al., 2005). In contrast, if the electrode remained implanted in the brain, the acute response was not resolved and the injury site became a site of chronic injury that was surrounded by activated macrophages and microglia, reactive astrocytes, as well as other factors and cells that were disturbed from their natural gradients and concentrations (Silver and Miller, 2004;
Leach et al., 2010; Szarowski et al., 2003). Since activated macrophages and reactive astrocytes were usually correlated with a decrease in neuronal density around the electrode, many researchers assumed a causative and detrimental relationship between activated macrophages, reactive astrocytes, and the reduction of neuronal cell density around the electrode. Some research, however, indicated positive effects associated with astrocytes and macrophages/microglia. Researchers showed that microglia can produce neurotrophic factors and “alternatively” activated microglial can be beneficial (Kigerl et
111 al., 2009; Liang et al., 2010). Animals without glial fibrillary acidic protein (GFAP), a primary intermediate filament in reactive astrocytes, had a worse functional outcome after trauma (Sofroniew, 2005). The research suggested that astrocytes are important to contain and minimize damage from an injury. While other in vitro work indicated the detrimental roles of astrocytes and macrophages on neural cells, the full in vivo picture has yet to be elucidated (Polikov et al., 2006; Wanner et al., 2008).
The majority of intracortical electrode studies examined either the tissue response or intracortical recordings, and only a few studies were directed at the understanding how tissue response affects electrode recordings. The most fundamental study used electrode impedance spectroscopy (EIS) to correlate the impedance of the electrode to tissue response (Williams et al., 2007; Mercanzini et al., 2009). While EIS provided longitudinal data of the response to electrode, EIS did not provide information on the proximity or function of neuronal cell bodies around the electrode.
Recent research using the drug flavopiridol decreased impedance even though electrical recordings were unchanged between control and drug-treated animals at four weeks (Purcell et al., 2009). Notably, the neuronal proximity was unchanged by the drug treatment, but no assessment of neuronal function or degradation was assessed around electrodes at 4 weeks. Another study examined the usage of the drug minocycline, a neuroprotective antibiotic, over four weeks and the performance of electrode recordings
(Rennaker et al., 2007). Minocycline-treated animals had improved neural recordings versus untreated animals. Only labeling for reactive astrocytes with GFAP was presented.
The data showed that minocycline decreased astrocyte activation at both 1 and 4 weeks though the differences were slight at four weeks. While minocycline altered the reactive
112 astrocyte response at four weeks, no impedance measurements or labeling for neuronal nuclei were examined.
Another example examining electrical recordings and tissue response featured only one animal. The relationship between tissue response and electrical recordings was examined and presented at an annual conference (Parker-Ure et al., 2008). Remarkably, the research showed no correlation between neural density around the implant and performance of recordings at 8 months. Similarly, labeling for reactive astrocytes and microglia did not correlate with decreased performance. The researchers showed labeling for Isolectin B4 inversely correlated with neural recording performance. It was uncertain whether Isolectin B4 labeled the leaky blood brain barrier or microglia and endothelial cells (Alonso, 2005).
Different researchers indicated the importance of different parts of the glial neuronal response to intracortical electrodes. Considering the conflicting data as well as the importance of the longevity and efficacy of intracortical recordings, the value of assessing the impact of tissue response on intracortical recordings was highlighted.
Previous studies used drugs with wide ranging cellular effects (Rennaker et al., 2007;
Purcell et al., 2009). Similarly, drugs were aimed at reducing and slowing down the tissue response. The research that will be discussed here was aimed at understanding the effect of hastening the tissue response on electrical recordings. We hypothesize that increased inflammation will negatively affect the quality of intracortical recordings.
113 5.3. Materials and Methods
5.3.1. Surgical Procedures
All procedures were designed to minimize the pain and discomfort of the animal
and were approved by the Case Western Reserve University Institutional Animal Care
and Use Committee. The same surgeon performed all procedures to minimize variability.
14 Sprague-Dawley rats weighing between 231-309g pre-implant were used for the experiments. All instruments were sterilized, primarily through steam sterilization while some instruments not suitable for steam were sterilized by gas (ethylene oxide). After induction with ketamine (80 mg/kg) and xylazine (10 mg/kg) via intraperitoneal injection
(IP), the animal's head was shaved. After the animal was unresponsive to toe pinch, ocular lubricant was applied to protect the eyes. Six of the thirteen animals were chosen randomly and were injected IP with lipopolysaccharide (LPS) (EMD Bioscience,
437627-5MG) at 2.5 mg/kg. The dosage used was double the dosage used by others where a response was detected in the days after injection (Rivest, 2009). The animal was placed in a stereotaxic frame fitted with a gas mask and ventilated with oxygen. A pulse oximeter attached to the hind limb paw aided in monitoring of animal health and anesthesia level. The animal was kept warm via a circulating water mat. When the animal began to whisk, react to toe pinch, or exhibit a raised heartbeat, Isoflurane (1-3%) was introduced into the oxygen flow. After a brief cessation, the surgery was continued when the animal did not react to toe pinch.
A 9-12mm incision along the scalp midline was made via a scalpel, and no skin was removed. Cotton tipped applicators pushed the periosteum from the skull. Using a pair of forceps, the scalp was held a centimeter lateral from the head. Curved scissors,
114 held inverted and perpendicular to the ground, were inserted underneath the scalp to
separate the muscle from its fascia until an iridescent color was apparent on the
remaining muscle. A pair of 90O curved hemostats was placed on the side of the head to retract the skin and muscle. With the aid of a periosteum elevator, the muscle was separated from the side of the skull by pushing the tool up and over the ridge of the skull and down the side of the skull. The side of the skull was cleared of any remaining muscle, and an ample area was exposed to allow for entry of the drill tool and drill site to be exposed. Using a sterile ruler and pen, an outline of the area to drill was made. A dental drill with a stainless steel burr bit was circled repeatedly around the outlined shape to remove the bone of the skull. Liberal application of saline and a slow drill speed avoided excessive heating of the skull. The opening, approximately 4 mm in diameter, was centered at 4 mm lateral from the midline and 3 mm caudal to the bregma to target implantation in the rat barrel cortex. After opening the skull, three stainless steel screws were screwed into the skull to provide mechanical anchors for the dental cement as well as a distant ground for the electrode. One screw was placed contralateral to the implant site and the other two screws were placed rostral and caudal of the bregma and lambda on the implant side, approximately 4-5mm from the edge of the skull opening. The dura was removed using a fine 45o angle microprobe and fine forceps using a surgical microscope.
A micromanipulator with a 100 µm (gross) and 10 µm (fine) resolution was
mounted to a metal pole positioned in line with the implant site. The angle of the
micromanipulator was set so that the implant was inserted perpendicularly to the brain’s
surface. A third of a meter long metal rod with an alligator clip rigidly attached at the end
was placed in the micromanipulator. For all implants a Neuronexus (part #A4x4-4mm-
115 200-200-1250-Z16, Ann Arbor, MI) 4 shank, 16 site 1250 µm2 site size, electrode array was used. The electrode was placed in the alligator clip so that the electrode sites faced the midline of the animal and the shanks were arrayed along the length of the animal. The exact target of implantation was judged based on the rat’s vasculature and was located caudal of the bifurcation of the two large blood vessels exposed after opening the skull
(Figure 22).
The device was lowered to near the skull surface, and the ground and reference wires were wrapped around the contralateral bone screw. The micromanipulator was lowered until the electrode was touching the surface of the brain, and the array was inserted with the gross dial until the topmost electrode disappeared into the brain tissue
(about 700 µm). The gross dial was used to minimize tissue dimpling during insertion.
Then, the electrode was advanced slowly by 10 µm increments to a total implant depth of
1000 µm. This depth was chosen so that the two middle electrodes along the length of the shank were located around layer V of the cortex.
116
Figure 22. Implantation site on brain for electrode in two different animals a)Pre implantation site illustrating bifurcation of blood vessels on the left of the image that identified the area for implantation for the rat barrel cortex. b)Post implantation site showing the 4-shank Michigan electrode where the shanks were arrayed along the length of the animal. In both images, the midline is towards the top of the image and the rostral is located to the left of the image. Bifurcation marked by arrow. The skull was sealed with Kwik-Sil (WPI, Inc. Sarasota, FL) where care was
taken to minimize sticking of the Kwik-Sil to tools or skin. After allowing Kwik-Sil to dry for a few minutes, dental cement was applied in multiple batches to slowly build up the head cap, insulate the ground and reference wires, and minimize edges that could be used to pry off the head cap. Care was also taken to ensure that the dental cement was not built up too high preventing attachment of the electrode headstage during recordings.
Before allowing the dental cement to dry for about 5 minutes, the skin was manipulated to ensure it was on top of the dement cement. No sutures were used to hold the skin tight next to the headcap. Some researchers noted that sutures can get snagged and cause complications. No animals were euthanized because of the headcap coming loose or from complications from a lack of sutures.
117 5.3.2. Electrical Measurements
5.3.2.1. Anesthetized Neural Recordings
To perform recordings, animals were placed in an induction chamber, and
Isoflurane (5%) was administered with oxygen until the animal was anesthetized, unresponsive to toe pinch. The animal was transferred to a nearby stereotaxic frame fitted with a gas mask where the animal was ventilated on oxygen and Isoflurane (1%).
The gas mask was loosely placed around the rat’s snout to not disturb the whiskers. The headstage was attached to the animal, and the metal table was grounded to minimize noise. Anesthetized recordings were performed on a weekly basis.
A custom program on a TDT RX5 Pentusa base station controlled all recordings.
The headstage was connected from the animal to a high impedance preamp located a third of a meter from the animal. An anti-aliasing filter in the preamplifier filtered signals before digitization. The preamp converted the signal to an optical signal that was recorded and processed in the RX5. The custom program performed several functions including: 1) recording the signals to a hard drive, 2) additional filtering of the signal for real time visualization, and 3) an interface to control the stimulus to the whiskers and recording of the timing of the stimulus. The signal was recorded at 24.1 kHz without additional software filters in order to minimize signal loss from filters as well as enable all spectrums of the signal to be analyzed offline (Purcell et al., 2009).
The TDT program controlled the stimulation of the whiskers. At an arbitrary interval, the whisker stimulus would be activated for 1 second, and the timing of the event would be recorded synchronously with the neural recordings. The whisker stimulus was a custom built device controlled by an output from the TDT RX5 (Figure 23). The
118 stimulus device consisted of a stereo speaker 10 cm in diameter. Applying a voltage to the speaker caused the cone of the speaker to oscillate up and down without an audible sound (Figure 23c). Attached to the speaker cone was a metal 17 gauge catheter about
30 cm long that was secured to the speaker with surgical suture. The metal catheter was cantilevered on the side of the speaker, and therefore, a small movement in the speaker translated to a large movement at the end of the metal catheter that disturbed several whiskers. Since the 4 shank Michigan probe recorded from a broad volume of cortical tissue, the ability to stimulate many whiskers at once was desirable. A single animal recording trial lasted 3 minutes. A session included four trials total: two trials with the whisker stimulus positioned to move the whiskers (Figure 23a) and two trials with the stimulus positioned to the side of the whiskers, not touching (Figure 23b). After neural recordings and impedance measurements, the animal was placed back in his cage and monitored until sternal.
119 *: !"#$ ;: %&'!"#$(!)*$
$
& . ,
01 01 2&34 2&34
5'!,6 5'!,6 5'!,6 5'!,6 7/, 28/#49 7/, 28/#49
-.,/),*+ !#+, 6:
Figure 23. Experimental Setup of whisker stimulus. The experiment overhead view is shown in a) and b) and the side view of the whisker stimulation apparatus is shown in c). a) Stimulation of whiskers with moving metal catheter attached to speaker where stimulation lasts for 1 second. b) Sham stimulation of the animal where the metal catheter still moved but the whiskers were not touched by the metal catheter. c) Side view of the whisker stimulation apparatus. 5.3.2.2. Impedance Spectroscopy
After the neural recordings were completed, the headstage was disconnected from the animal, and another headstage was connected to the anesthetized animal. The notable difference of the second headstage was there were no active electronics in the headstage, enabling impedance measurements. The headstage was connected to a custom-built breakout box connected to a multiplexor unit (SCNR16) that was connected and controlled by an Autolab Potentiostat (PGSTAT12) and computer software (Frequency
Response Analyzer, FRA v4.8). Since the electrodes were implanted in the head of the animal, the impedance measurements used a 2 cell configuration. The working electrode was the electrode on the array to be tested. The counter and reference electrodes were the
120 ground wire of the array. The ground was affixed to the upright portion of the mouth bar on the stereotaxic frame . Using the Autolab supplied software, a custom script was used to sequentially measure the impedance of the 16 electrodes with a 10mV sinusoid at 31 different frequencies from 10 Hz to 10kHz, logarithmically spaced.
5.3.2.3. Free Moving Neural Recordings
Between the weekly, anesthetized recordings, additional neural recordings were performed. The animal was connected to the TDT RX5 recording system and was able to freely move about the cage. A trial would last for 3 minutes, and two trials per day were recorded. No overt attempt to stimulate whiskers was made during recordings.
5.3.3. Histology
As an examination into the chronic response to the implanted electrodes, immunohistochemistry was performed on the brains of animals at four weeks. Animals were perfused directly after the final recording session. At four weeks, transcardial perfusion was performed via a mechanical pump with approximately 500 mL of
Dulbecco’s Phosphate-Buffered Saline followed by 250 mL of 10% buffered formalin to fix the tissue. The extracted brain was placed in formalin for 24 hours when replaced by fresh formalin. Brains were cryoprotected in 30% sucrose solution and were frozen in embedding molds (Electron Microscopy Products, Hatfield, PA). Molds contained
Optimal Cutting Temperature (OCT) compound (Sakura, Tokyo, Japan). Brains were sectioned horizontally with a cryostat, cutting 30 µm thick slices. Some slices were labeled immediately while others were preserved in cryoprotectant solution (Watson et al., 1986). For a given label, all animals were processed in the same batch to minimize
121 differences across animals. Following these procedures (Biran et al., 2005), slices were
labeled using free-floating sections.
Slices were placed in goat blocking solution overnight at 4oC composed of 4%
(v/v) normal goat serum (Sigma-Aldrich, St. Louis, MO), 0.3% (v/v) Triton-X-100
(Sigma-Aldrich), and 0.1% (w/v) sodium azide (Sigma-Aldrich) in 1x D-PBS
(Invitrogen, Carlsbad, CA). Next, slices were placed in blocking solution with primary
antibodies (see Table 5) overnight at 4oC. Following three, 15 minute washes in PBS,
slices were placed in blocking solution with secondary antibodies, goat anti-rabbit IgG or
anti-mouse IgG1 or Alexa Fluor 594, 488, and DAPI dilactate (Invitrogen, Carlsbad, CA)
for one hour at room temperature. Labeling for GFAP required 2 hours of secondary
incubation for all animals. After three, 15 minute washes in PBS, slices were mounted
and cover slipped using Fluoromount-G (Southern Biotech, Birmingham, AL).
Some samples were labeled with Fluoro Jade C (FJC) to mark neurodegeneration.
FJC has been shown to stain other cell types such as astrocytes, blood vessels, and
microglia, as well (Chidlow et al., 2009; Damjanac et al., 2007). The method used was a
slight variation on the standard FJC protocol. The frozen sections were cut at a thickness
of 30 µm. Then sections were mounted from PBS onto Superfrost Plus Slides (Fisher
Scientific) to improve adherence of slices to slides. Slides were dried in a vacuum oven at
50oC. After drying, the slides were immersed into a basic alcohol solution consisting of
1% sodium hydroxide in 80% ethanol for 5 minutes. Slides were rinsed for 2 minutes in
70% ethanol followed by 2 minutes in distilled water. Slides were then incubated in
0.06% potassium permanganate solution for 10 minutes. Slides were transferred for 10 minutes to a 0.0002% solution of Fluoro-Jade C (Millipore, Billerica, MA) dissolved in
122 0.1% acetic acid vehicle (1:50 of 0.01% FJC stock solution). Rinsing was performed
through three changes of distilled water for 1 minute each. Excess water was drained
onto a paper towel, and further drying was performed in a vacuum oven at 50oC for at least 5 minutes. The slides were then cleared in Xylene for a minute and then coverslipped with Permount mounting media (Fisher Scientific).
Table 5. Summary of Primary Antibodies. Antibody Vendor (Part #) Dilution Isotype Cell Type/Labeling
GFAP Chemicon 1:1000 Rb IgG Astrocytes (AB5804)
NeuN Chemicon 1:500 Ms IgG1 Neurons (MAB377)
CD68 Chemicon 1:250 Ms IgG1 Activated (ED1) (MAB1435) Macrophages, Microglia
IBA1 Wako 1:615 Rb IgG Microglia (019-19741)
Fluoro Chemicon 1:50 N/A Neurodegeneration/ Jade C (AG325) (0.0002%) Astrocytes/ Blood Vessels
MAP2 Chemicon 1:1000 Ms IgG1 Neurons (Dendrites) (MAB3418)
5.3.4. Image Acquisition and Analysis
Images of slices were taken with an Axio Imager.Z1M Microscope and AxioCam
MRm CCD camera in grayscale with 12 bits of digitization (Zeiss, Thornwood, NY).
Sections chosen for acquisition and analysis were along the length of the implant shaft
before it tapered to a point. To minimize depth-associated effects, the sections were
123 located on average at 1900 µm ± 200 µm below the topmost point of the cortex. Though
some variation is expected based on slight differences in implant placement and the
curved surface of the cortex, the slice depth was not significantly different between the
control and LPS-treated groups. Several sites were imaged including: the 4 sites of the 4 shanks, the contralateral side matching the area of the 4 shanks, and an image located at
the caudal end of the brain on both the implant and contralateral side. The caudal end
was approximately located 2.5 mm anterior and 2.5 mm from the midline of the
separation of the two hemispheres. Exposure times for a specific label were set to
prevent overexposure. For a given label, exposure times were the same for all animals.
Four automated exposures were performed on every image: one bright field and 3
fluorescent colours for 3 different labels. Eleven animals were analyzed after four weeks,
but only six animals were taken into account for IHC and electrophysiology
quantification.
Two animals, one of each group, were sacrificed early at two weeks because they
were not recording single unit action potentials. Two animals, one of each group, were
removed from IHC analysis because their electrode arrays were deemed to be broken
because their impedance was similar to the shunt capacitance of the probes.
Additionally, three animals, one control and two LPS-treated, were excluded because
upon perfusion the electrodes were seen to not be protruding the skull and therefore not
implanted into the cortex.
As explained in previous work, a custom Matlab R2009b (MathWorks, Natick,
MA) script was developed to measure the intensity around the implant (Harris et al.,
2011a). Due to the small size of the implant, the researcher defined the border of the
124 implant in the bright field image. One hundred radial lines emanated from the centroid of the implant area to measure the intensity (grayscale values) over the first 100 microns immediately external to the border. The average intensity of the 100 rays was divided by the background measurement of the image contralateral to the implant to account for differences between images due to photobleaching and other effects. The peak intensity and intensity integrals were examined between control animals and those injected with
LPS. A two sample t-test was used to examine statistical significance.
Another custom Matlab script was used to count neuronal cell nuclei around each shank. As discussed in previous work, the script computationally isolated and highlighted nuclei (Harris et al., 2011a). After automated marking, nuclei were manually added or subtracted. After marking, Matlab computed the distance from each nucleus to the border along the ray emanating from the centroid of the implant site. These distances were binned into 10 µm increments; the counts were normalized by area of the corresponding 10 µm-wide concentric polygonal donut to provide the neuron density.
Further, the density was scaled to a percentage of the neuron density from the contralateral non-implanted side at a similar position as the implanted shanks.
5.3.5. Electrophysiology Analysis
5.3.5.1. Non-Evoked Analysis
To process the data, a trial was run through the analysis program (TDTAnalysis v.21) that was modified. The program performed analysis published by the Kipke lab previously (Purcell et al., 2009; Ludwig et al., 2009). Signal processing was performed on each trial before averaging of signals. The electrodes were spaced 200 µm, and single unit action potentials are typically recorded from neurons within 50 µm (Henze et al.,
125 2000). Therefore, the findings presented here treated the 16 electrodes on one array as 16
independent measures (Williams et al., 2007).
To go into more detail of the analysis, the program loaded the trial for all 16
electrodes. The offline filtering and processing in Matlab was performed to match
previous methods of Michigan researchers (Ludwig et al., 2009; Purcell et al., 2009).
Filters separated the signals into two different components: one for local field potentials
(LFPs) and one for neural spikes. The spike filters were 1st order Butterworth in order to mimic analog real time filters. LFP bandpass filters were implemented as 8th order
Butterworth filters to minimize noise from frequencies not in the bandpass. LFPs were
filtered from 0.1 Hz to 140 Hz, and spikes were bandpass filtered over 300-5000 Hz.
Previous research has shown the benefits of computing a common average reference
(CAR) from all of the 16 electrodes (Ludwig et al., 2009), so the LFP CAR and spike
CAR signals were subtracted from the LFP and spike signals.
For spike analysis, the program automatically created a negative threshold. The threshold was the 3.5 * standard deviation of the signal. A sample that passed the threshold joined a list of candidate samples. The absolute minimum value of those candidate samples over a 2.4 ms window was chosen to be the minimum and center of a spike snippet window. The minimum was aligned to the 1.2 ms mark of the 2.4 ms snippet window that had data 1.2ms before and after the minimum. These spike snippet windows were removed from the signal leaving a remaining signal. The remaining signal was used to calculate the noise voltage peak to peak of each electrode channel. Based on previous research from Michigan researchers (Purcell et al., 2009; Ludwig et al., 2009), the noise voltage peak to peak of each electrode was defined as 6 * standard deviation of
126 the remaining noise signal that was formed from removing snippet windows. The
computation of noise was similar to techniques employed by other researchers (Quiroga
et al., 2004).
For each electrode, the collection of spike snippet windows underwent a principal
component analysis (PCA) to remap the features of a spike snippet to the PCA space.
The remapping to PCA space facilitates clustering of the signals. Clustering was
performed on the first three principal components of the spike snippet. The clustering
was done automatically by the program via Fuzzy C-Means clustering, similar to K-
Means clustering. Clustering techniques iterated through the collection of snippet windows and assigned a given snippet window into a certain cluster based on its description in PCA space and its distance to the cluster centers. Based on the closest cluster center, the spike was assigned to that cluster. K-Means clustering iteratively determined the cluster centers based only on the membership of all member points. In
Fuzzy C-Means, cluster centers were iteratively determined by the position of the points in the cluster as well as to the degree to which that point is part of that cluster, e.g. a point closer to the cluster center has more weight than those further away. The iterative clustering process binned snippet windows for a given electrode into different clusters.
After clustering, a spike snippet window was considered part of a cluster only if it had a membership index greater than 80%. Of the remaining spike snippet windows in the cluster, a cluster was considered a neural signal if there were more than 30 instances in the cluster. For the neural clusters with more than 30 instances, a mean spike waveform was created from the spikes in the cluster. The voltage peak to peak of the
127 mean waveform created the signal voltage peak to peak. The overall process is depicted
in Figure 24.
Figure 24. Representative diagram of electrophysiological analysis for a single sample electrode or channel. Neural data was recorded, and offline analysis was completed via Matlab. After filtering, only waveforms that exceeded the threshold were analyzed further. The waveforms were then remapped via principal component analysis (PCA) into the three dimensional PCA space. The clustering algorithm assigned membership into clusters for the candidate waveforms. Each cluster had its associated average signal. Each channel had a noise voltage associated with it. Therefore, the signal to noise ratio (SNR) was be computed for each cluster. The top of the gray bar shows a sample neural recording and its filtered version. The filtered signal shows a magnification of a candidate spike. The length of the full recordings were about 3 minutes and span from -1 mV to 1 mV. The magnified single spikes are -200 µV to 100 µV with a length of about 3 seconds. 5.3.5.2. Evoked Analysis
Evoked recording analysis was similar to the non-evoked recording analysis.
Instead of examining the entire trial, evoked recording analysis analyzed the periods of time around a whisker stimulus. Within each evoked trial, there were several instances
(about 20 per trial) where the rat’s whiskers were stimulated. In the analysis, we
128 examined difference between the signals before the stimulus and signals during the
stimulation.
The first parameter examined was the change in spiking rate (spikes/second) in
response to the stimulus. Here, we examined the difference in mean spiking rate for the
second during the stimulus minus the mean spiking rate one second before the stimulus
onset.
The other examined parameter was the change in LFP power in response to the
whisker stimulation. The analysis consisted of the power during the one second stimulus
minus the power one second before the whisker stimulation. Several frequency bands
within the LFP were examined, including: low frequency (0.1-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), beta (6-24 Hz), low gamma (35-55 Hz), and the high gamma band (65-
140 Hz) (Khawaja et al., 2009). The analysis computed the power and/or power spectral
density for each segment before computing the mean of all the segments before/during
the stimulus.
5.4. Results
5.4.1. Inflammatory Cells: Macrophages and Microglia
Analysis of IBA1 and ED1 labeling quantified activation of microglia and
macrophages. IBA1 labeled microglia in all states (Figure 25), and ED1 labeled activated
microglia and activated macrophages (Figure 26). Examining contralateral levels for
IBA1 labeling, control and LPS-treated animals had similar contralateral intensities
(p=0.65). For controls, IBA1 images had an average contralateral intensity of 4.83 ± 0.23
relative units. LPS-treated animals had an average contralateral intensity of 4.92 ± 0.26
129 relative units for IBA1 images. ED1 labeling had similar contralateral intensities for
control and LPS-treated animals, as well (p=0.46). For controls, ED1 images had an average contralateral intensity of 7.61 ± 0.44 relative units. LPS-treated animals had an average contralateral intensity of 7.11 ± 0.48 relative units for ED1 images.
Microglia labeled by IBA1 normally appeared to have a central spheroid shape with some branching emanating from the central body. In response to the implant, the
IBA1 labeled microglial had more branching around the electrode (Figure 25c).
Similarly, IBA1 labeling often appeared punctate or amoeboid near the electrode border
(Figure 25d) with a more ramified morphology away from the implant site (Figure
25b,c). The punctate staining often coincided with ED1 labeling and exhibited a different morphology than the branched morphology (Figure 25d). ED1 labeling (Figure 26)
exhibited similar features that were limited mainly to amoeboid cells within 50 µm to the
implant interface.
The microglia response was characterized by intense IBA1 labeling 10 µm from
the border. The response decreased at 25 µm from the border for LPS treated animals
and 50 µm for control animals (Figure 25). IBA1 levels stayed elevated from
background levels past 100 µm from the border. On average, the labeling of ED1 in
control animals was nearly significantly greater than in LPS animals over the integral
region examined (0-100 µm, p=0.07). The peak intensities for IBA1 and ED1 labeling
were similar for control animals and LPS-treated animals (p=0.31, p=0.35).
130
Figure 25. Analysis of IBA1-Microglia immunohistochemistry and representative images of microglial and activation. a) Analysis of microglia intensity as a function of distance from the tissue-implant border. Relative intensity ± standard error (outline) as a function of distance from border of four week control (blue-grey) and LPS-treated (black) animals. b)Resting microglia was labeled by IBA1 indicating some branching in a ramified state. One specific microglial cell indicated by arrow. c) Reactive microglia were labeled by IBA1 showing increased branching. One specific branched cell indicated by arrow. d) Activated or amoeboid morphology with little branching was shown by labeling for IBA1. One specific amoeboid cell indicated by arrow. Scale bar is 50 µm.
131
Figure 26. Analysis of ED1-Activated macrophages and microglia immunohistochemistry as a function of distance from the tissue-implant border. Relative intensity ± standard error (outline) as a function of distance from border of four week control (blue-grey) and LPS-treated (black) animals. 5.4.2. Astrocytes
Labeling for GFAP indicated the reactive astrocyte response. LPS-treated animals and control animals showed different patterns of labeling. Control animals showed a level of GFAP labeling that had a very slight peak near the border (around 10 #m). LPS- treated animals showed a larger difference from its baseline, but generally exhibited a lower level of activation further from the implant (Figure 27). Examining contralateral levels for GFAP labeling, control and LPS-treated animals had different contralateral intensities (p<0.01). For controls, GFAP images had an average contralateral intensity of
13.01 ± 0.77 relative units. LPS-treated animals had an average contralateral intensity of
17.81 ± 1.41 relative units for GFAP images. Since previous research indicated that
GFAP levels are similar to controls after LPS treatments (Weberpals et al., 2009), it was assumed that differences were based on variability of animals that was more pronounced by the small number of animals in the trial. The peak for GFAP labeling in LPS-treated animals was located around 25 #m that reached baseline levels by 50 #m. There was no statistically significant difference between the peak intensities for control or LPS animals.
Further examination of the morphology of the reactive astrocytes labeling in control and
132 LPS-treated animals indicated a difference in the astrocyte cell area. Astrocyte area
around the implant shanks in LPS-treated animals was 26% greater than the astrocytes
around implants in the control animals (data not shown). The difference was statistically
significant (p=0.05).
Figure 27. Analysis of GFAP-Reactive astrocyte immunohistochemistry as a function of distance from the tissue-implant border. Relative intensity ± standard error (outline) as a function of distance from border of four week control (blue -grey) and LPS-treated (black) animals. 5.4.3. All Cell Nuclei: DAPI
During the labeling for specific cell types, all cell nuclei were counterstained with
DAPI dilactate. DAPI was analyzed via image intensity. Bunching of cells occurred around the shanks of LPS-treated animals more significantly than control animals. The
DAPI intensity in LPS-treated animals was increased above standard levels out to 60 #m from the electrode-tissue interface. Control animals also had elevated levels of cellular density to around 60 #m. The peak intensity level of labeling for DAPI around the shanks was significantly less than LPS-treated animals (p<0.001). The control animals had 35% of the DAPI peak intensity in LPS-treated animals.
133
Figure 28. Analysis of DAPI-all cell nuclei histochemistry as a function of distance from the tissue-implant border. Relative intensity ± standard error (outline) as a function of distance from border of four week control (blue-grey) and LPS-treated (black) animals. 5.4.4. Neuronal Nuclei (NeuN)
Labeling for NeuN (neuronal nuclei) was used to investigate neuronal cell density within 100 µm of an electrode shank (Figure 29). Similar to previous literature, neuronal density was reduced towards the tissue-electrode border in both control and LPS-treated animals (Biran et al., 2005). However, the average neuronal density within 50 µm of the implants in control animals was significantly greater than around implants in LPS-treated animals (p=0.02, Figure 29). Though it appears that there was a greater neural cell density in the 10 µm region, the difference was not statistically significant (p=0.13).
134 (!!"# '!"# &!"# *+,-.+/# %!"# /01# $!"#
"#+2#*+,-.3/3-4.3/#54,16-7 !"# !# )!# (!!# 86*.+94-4.1#2.+9#:,-4.23*4 Figure 29. Analysis of NeuN Immunohistochemistry. Quantification of NeuN based on distance from the tissue-implant border in response to implants: four week control animals (blue-grey diamond), four week LPS-treated (black square). Points represent histogram counts in 10 µm intervals. Counts have been scaled based on area as well as background count for the contralateral, non-implanted side. Average count of NeuN as a function of distance from the tissue-implant border ± standard error. 5.4.5. Dendrites and Neurodegeneration (MAP2 and Fluoro Jade C)
To examine the state of neurons in the zone near the implanted electrodes, the neural network and neuron health were assessed through dendrite labeling (MAP2, Figure
30a-c) and neurodegeneration labeling, Fluoro Jade C (FJC, Figure 30d). Similar to neural density, the density of the dendrites decreased approaching the electrode. Notably, the quantification of dendrite intensity (Figure 30a) showed that the LPS-treated animals had a more rapid decrease in dendrites than control animals. The rapid decrease in LPS- treated animals formed a sharp elbow at the 35 µm mark (Figure 30b) that was not apparent in control animals (Figure 30c). In Figure 30b, the black area where dendrites are lacking is filled with DAPI positive cells (Figure 28).
In consecutive slices to MAP2 labeled slices, labeling for FJC was performed to analyze the current health of neurons around the electrode shanks. Labeling showed a few, if any, cell bodies labeled via FJC indicating few actively degenerating neurons
(Figure 30d). Additionally, FJC often labeled a different morphology other than cell
135 bodies. In some slices, the FJC labeled astrocytes as determined by the labeling shape
and pattern (data not shown). In other cases, the labeling consisted of an indistinct ring
that did not correspond to astrocyte, microglia, or other labels performed in this trial
(Figure 30d).
Figure 30. Quantification of degradation of dendrite density and neural degeneration. a) Analysis of MAP2-dendrite histochemistry as a function of distance from the tissue- implant border. Relative intensity ± standard error (outline) as a function of distance from border of four week control (blue-grey) and LPS-treated (black) animals. b) Representative image from LPS-treated animal with labeling for dendrites with MAP2. The four shanks were pictured, and the open/black space at center was filled with non- neural cells as marked by DAPI. c)Representative image from control animal with labeling for dendrites with MAP2. The four shanks were pictured. d) Magnification of FJC labeling in slice following slice pictured in b). The top most site was magnified. A degenerating neuron indicated by arrow, and the arrowhead indicated a ring of FJC labeling of unknown cell origin. Scale bars are 50 µm.
5.4.6. Electrode Impedance
Electrode impedance measurements were performed on each electrode site to examine the capacitance and resistance between the electrode and the tissue. At four weeks, the average impedance for the electrodes sites in LPS-treated animals was
2.10 ± 0.2 MOhms. The average impedance for electrode sites in control animals was
0.681 ± 0.03 MOhms. The difference between groups was statistically significant
(p<0.001) and included 4 control animals and 2 LPS-treated animals. Since there are 16 electrodes per animal, there are 64 electrodes for control animals and 32 electrodes for
LPS animals.
136 5.4.7. Electrophysiology
5.4.7.1. Overall Characteristics of Recordings
After recording and processing electrophysiology recordings, all clustered units
from recordings from all channels were examined to understand the general
characteristics of the recordings (Figure 31). The average SNR across all electrode sites
was similar between LPS-treated and control animals (p=0.23) (Figure 31a). When the
distribution of individual points was examined, the SNR had a greater maximum in
control animals. The average of the values above the 3rd quartile for all values was statistically greater in control animals than LPS-treated animals (p=0.04, Figure 31b).
The average signal voltage, peak to peak, across all sites was similar between control and
LPS-treated animals. The noise voltage, peak to peak, was trending towards significantly greater in LPS-treated animals than control animals (p=0.06, Figure 31d). Similarly, the
RMS noise in LPS-treated animals was significantly greater than control animals
(p=0.05, data not shown).
137
Figure 31. Bar graphs for signal and noise measures across all electrodes. a) Signal to Noise Ratio (SNR) for all electrodes in control (blue-grey) and LPS-treated (white) animals. Measurements of individual electrode were shown on left triangles for control animals and on right squares for LPS-treated animals and showed the distribution of measurements. b) Histogram of SNR values for control and LPS-treated animals as depicted by points in a). LPS and control animals have same number of occurrences at 1. c) Average voltage peak to peak of signals from recordings across all electrodes for control (blue-grey) and LPS-treated animals (white) ± standard error. c) Average noise voltage peak to peak computed after removing spike signals from original signal for control (blue-grey) and LPS-treated animals (white) ± standard error.
5.4.7.2. Unit Recordings with SNR>=1
In order to examine the highest quality signals for a BMI system, clustered units that had a signal to noise ratio (SNR) greater than or equal to 1 were examined for control animals and LPS-treated animals. No units with a SNR>=1 were recorded from LPS- treated animals. All signals recorded from the LPS animals were of low signal amplitude
(on average 40 µV peak to peak across all electrodes, Figure 31c). 0.15 ± 0.05 units per site were recorded for control animals with an average SNR of 1.87 ± 0.194. A Fisher’s
Exact Test for count data showed the proportion of electrode sites with an SNR>=1 in control animals was nearly significantly different than LPS-treated animals (p=0.09).
5.4.7.3. Evoked Recordings
To further quantify the quality of neural recordings in both types of animals, evoked responses were analyzed. The evoked recordings allow investigation of detecting
138 a definitive event. Examining all channels and all spikes that exceed the threshold, there is a greater change in firing rate for control animal electrodes than LPS-treated animal electrodes (p<0.01). The difference in firing rate (during minus pre stimulus) was 154.95
±23.93 spikes/sec for electrodes in control animals. The change in firing rate was 52.81
± 15.39 spikes/sec for electrodes in LPS-treated animals, nearly one third of the control rate (Figure 32a, left columns). For units with an SNR>=1, control animals exhibited
49.88 ± 19.9 spikes/sec more during the stimulus than the second before the stimulus
(Figure 32a, right column). Since LPS-treated animals had no units with an SNR>=1, there was no firing rate computed for LPS-treated animals with SNR>=1.
Another measure of neural recordings, local field potentials (LFPs) record a larger collection of neurons than spike recordings. LFPs provide an alternative measure of the quality of recordings. Several frequency bands were examined including low frequency
(0.1-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), beta (6-24 Hz), low gamma (35-55 Hz), and the high gamma band (65-140 Hz), (Figure 32b). Several of the bands had statistically significant differences (p<0.05) between groups, including 0.1-4, 4-8 Hz, 16-24, 35-55, and 65-140 Hz. While the LPS treated animals generally had a constant level of power, the control animals saw an increasing power with increasing frequency. Therefore, the low frequencies (0.1-4 and 4-8 Hz) had a significantly lesser power in controls compared to LPS-treated animals while the higher frequencies (16-24, 35-55, and 65-140 Hz) had a significantly greater power than LPS-treated animals.
139 *+ &$$#$$% ,+ )#$$% (#$$% ")$#$$% '#$$% @%04(%A "$$#$$% &#$$% !#3 "#$$%
36/7'893':%0*$ )$#$$% !"#$#%&'($)*+,$ -.(/012#('$34/5$$ $#$$% $#$$% !"#$$% -.. 456 $#"!( (!9 9!"& ":!&( ')!)) :)!"($ /01230# 78" "(';.'0:<$+=0*$)>?,
Figure 32. Bar graphs for evoked recordings showing single spike and LFP recordings. a) The average difference in firing rate (spikes/sec) across all electrodes during the stimulus minus before the stimulus in implanted control and LPS-treated animals. The left set of columns show the firing rate for all spikes that exceeded a threshold (control=blue-grey, LPS=white) while the right column shows the firing rate for control animals (blue-grey). The LPS-treated animals had no units with an SNR>=1 and consequently no firing rate. b) Average LFP power difference (during minus pre stimulus) plotted in decibels (dB) ± standard error over several frequency bands for control (blue-grey) and LPS-treated animals (white). 5.5. Discussion
The preliminary findings detailed in this study examine the role of tissue response in intracortical electrode recordings. Through the use of LPS, we show a higher cell density around implants that correlate with higher impedance measurements at 4 weeks.
Additionally, the LPS-treated animals exhibit decreased neuronal proximity and dendrite density. Though the overall characteristics of the recordings are similar between groups, we show that LPS-treated animals have a decrease in the clinically relevant evoked responses. This preliminary work indicates that an LPS-stimulated response may have multiple pathways to affect the quality of neural recordings. The possible modes include decreased dendrites, decreased neural proximity, increased non-neural density, and increased tissue impedance. The work shows that a stimulated tissue response detriments neural recordings.
140 Neural recordings from animals and humans show a maximum lifetime typically
2-3 years (Tresco and Winslow, 2011). To enable widespread clinical efficacy,
technological advances are needed to improve electrode longevity (Schwartz, 2004;
Tresco and Winslow, 2011; Vetter et al., 2004; Williams et al., 1999). There are
extensive studies on either a) the tissue response or b) electrode recordings, but few
studies examine the direct effect of tissue response on neural recordings. Some of the
studies that feature both histology and electrophysiology do not perform a full
immunohistochemical study with glial and neural markers (Anderson, 2008; Rennaker et
al., 2007). The studies are therefore limited in ability to determine the effect of the tissue
response on electrode recordings.
The most complete study to date by Purcell et al. (Purcell et al., 2009) examines
several labels for cell types in control and drug-treated animals. They hypothesize that the drug-treated animals would improve recordings, but no significant differences are shown between groups for non-neural density, neural density, or recording quality.
Regardless of the treatment paradigm used, the classic response to an intracortical electrode is similar. Chronically, the intracortical electrode causes a tissue response characterized by a glial scar (Silver and Miller, 2004; Leach et al., 2010). The glial scar is composed of activated macrophages and microglia encircling the electrode. The microglia and macrophages try to repair damage from the electrode insertion and respond to plasma and blood proteins adhered to the electrode. The general belief is that the response creates an overall inhospitable environment for neuronal cells with dendritic pruning (McConnell et al., 2009). The distance and the impedance between the electrode and neuron increase over time that decreases recording performance.
141 Therefore, we hypothesize that an increased response will negatively affect the
quality of intracortical recordings. We examine this hypothesis via the usage of the
inflammatory stimulant LPS. LPS is an endotoxin on the cell wall of gram-negative bacteria, and many researchers use LPS to stimulate microglia and macrophages in the brain (Fujioka and Akema, 2010; Herber et al., 2007; Hinkerohe et al., 2010; Kipp et al.,
2008; Qin et al., 2007).
Though many have characterized the LPS response in the days after an LPS injection, we examine several histological labels to characterize the response 4 weeks after LPS administration. Since LPS activates microglia (Qin et al., 2007), we examine microglia by labeling with IBA1 (Figure 25) and activated macrophages and microglia by labeling for ED1 (Figure 26). Remarkably, at the four week time point, control and LPS- treated animals have similar peak levels of activation of macrophages and microglia
(ED1) and microglia intensity (IBA1). The levels of ED1 are nearly significantly greater in control animals than LPS-treated animals (p=0.07). Since the intensity of staining on the contralateral side from implants is similar between control and LPS-treated animals, the data suggests that the LPS treatment ran its course by the four week time point.
GFAP labels reactive astrocytes and show peak levels that are similar between control animals and LPS-treated animals (Figure 27). Reactive astrocytes usually encircle activated macrophages and microglia and are a principal constituent in the glial scar.
Levels 50-100 µm from the electrode are significantly less in LPS-treated animals, p=0.02. While GFAP activation usually accompanies microglial activation, other researchers show that GFAP is similar between untreated and LPS treated animals after 2 months despite microglia activation (Weberpals et al., 2009). Since intensity
142 measurements do not assess the morphology of the astrocytes, other researchers examine
the astrocyte spreading to measure activation (Georges et al., 2006; Wanner et al., 2008).
We assess astrocyte cell body area as a measure of spreading. The astrocyte cell body
area is 26% greater in LPS-treated animals (p=0.05). The data suggests that LPS in
combination with the electrode modified the astrocyte response.
Since IHC can only show a snapshot in time, the response to the LPS injection at
earlier timepoints is not directly known. The LPS dosage in this study is twice the
concentration as other studies that show microglia activation after 24 hours (Rivest,
2009). The LPS dosage in this study uses half the dosage that shows microglial
activation at 2 months (Weberpals et al., 2009). Therefore, it is possible that the
activation of cells, such as microglia, subsides by four weeks. Similar activation levels
contralateral to the implant in both groups support the hypothesis that the inflammatory
response subsided by four weeks.
Regardless of the specific pathway, the labeling of all cell nuclei via DAPI shows
that LPS causes a greater cellular response around the electrode at the four week
timepoint (Figure 28). There is a statistically greater intensity around shanks in LPS- treated animals (a greater than threefold increase in peak intensity). With NeuN labeling
(Figure 29), the DAPI labeling indicates that the cells are predominately non-neural.
Future work needs to be completed to examine the type and source of the DAPI labeled cells.
Increased impedance measurements correlate with increased DAPI labeling. The electrodes in LPS-treated animals had significantly higher impedance at 1kHz, as well as
143 10 and 100 Hz. The focus of this study is to examine the correlations between tissue
response and electrode recordings at 4 weeks, but impedance measurements were
gathered weekly (Figure 33). These results show a rapid change in impedance between
the two groups, suggesting that the origin of the DAPI cells are from an acute influx of
cells due to LPS activation. LPS has been shown to increase the permeability of the
BBB, so it is possible that the LPS treatment allows more blood borne cells to enter the
parenchyma (Mayhan, 1998). The current data does not allow us to conclusively say
whether changes within the first week are through a more permeable BBB or from an
increase in activation of microglia and macrophages in the brain parenchyma. Further
labeling for T-cells, neutrophils, fibroblasts, or endothelial cells may allow for the
determination of the source and type of the DAPI cells around electrodes.
Figure 33. Impedance measurements over four week time course of implanted electrodes measured at 1kHz with animal anesthetized. Electrodes in LPS-treated animals (white) showed an increased impedance versus electrodes implanted in control (blue-grey) animals. Error bars are shown as standard error. Significant differences between groups for a given day are marked by * (p<0.05). Intragroup differences from the preimplant levels are marked by a % (p<0.05). Intragroup differences in the control group are marked by a ! (p<0.01). Intragroup differences in the LPS group are marked by a # (0.05 144 In addition to changes in total cell density, LPS-treated animals exhibit differences in neuronal proximity around shanks (Figure 29). The average density of neural cells (NeuN) less than 50 µm from the electrode in LPS-treated animals is less than half than levels in control animals. Based on previous research the 0-50 µm range is the most important region to record spikes (Henze et al., 2000). Aligning with the neural density trends, dendrite intensity (MAP2, Figure 30a-c) around implants in LPS-treated animals is significantly less than control animals. Dendritic pruning occurs early in the process of neurodegeneration (McConnell et al., 2009). Similar to previous literature, few neural cell bodies are in the process of neurodegenerating at the 4 week time point (FJC- Figure 30d) (McConnell et al., 2009). This data suggests that progressively degenerating neurons are not the principal factor decreasing recording quality at longer time points. It is possible that neurons are slowly degenerating over a long period of time, and therefore few neurons are labeled by FJC. It is also possible that the neurons migrate away from the implant site. Either would steadily decrease electrode performance over time. The possibility that LPS has a method of action on the neural network is also an option. After a LPS injection, researchers do not see a change in NeuN staining levels between treated and untreated animals at 2 months despite activated microglia (Weberpals et al., 2009). Despite similar neuron densities, the researchers see long-term cognitive deficits in LPS treated animals. The researchers show support that the cognitive deficits are from LPS-induced changes in synaptic proteins. Similar research investigating LPS treatments show other subtle changes in neurons. After a low dosage of LPS, data suggested that some neuron function was altered from inhibitory to 145 excitatory (Zhihong Chen and Bruce Trapp, unpublished results). These more subtle changes in neural function are not examined via IHC in this study. We examine the non-evoked recording characteristics of the channels to investigate the overall characteristics of the electrode performance in the two animal groups. Though SNR across all electrodes is similar (Figure 31a, bar graphs), the average of the upper quartile is lesser in the LPS-treated animals. Illustrations of the differences are shown by the point plots in Figure 31a and by the histogram in Figure 31b. The data indicates that control animals can record higher quality units. The data also indicates that control animals can record low quality units that the LPS-treated animals cannot record. As suggested by data, both changes in recording high and low quality units are due to decreased neural density and increased impedance in LPS treated animals, compared to controls. The signal voltages are similar at four weeks (Figure 31c), but the noise voltages increase for electrodes in LPS-treated animals (Figure 31d), agreeing with previous research (Purcell et al., 2009). The immunohistological data indicates that there is a higher non-neural cell density around the electrode, increased separation between the electrode and neural cell bodies, and decreased dendrite density near the electrode in LPS-treated animals. In addition, the impedance of electrodes in LPS-treated animals is greater. All indicate that neural recording should be degraded in LPS-treated animals. Several measures indicate that recordings are worse in LPS-treated animals. The electrodes in LPS-treated animals record 0 quality (SNR>=1) units while control animals record 0.15 ± 0.06 units per site with an average SNR of 1.87 ± 0.194 for quality units. Since there are no quality units 146 recorded in LPS-treated animals, statistical tests rely on the relative populations of each group using Fisher’s Exact Test for Count Data (Table 6). Table 6. Contingency table used for Fisher’s Exact Test for count data. Electrodes w/ SNR>=1 Electrodes w SNR<1 Totals Control Electrodes 7 54 61 LPS Electrodes 0 31 31 Total 7 85 92 In examining the proportion of electrodes in each group that could record a quality unit (SNR>=1), results are nearly significantly different between groups (p=0.09). For evoked spikes with SNR>=1, Fisher’s Exact Test is also applied for the spike rate (Figure 32a, right graph). The test provides the same nearly significant result (p=0.09). Examining spikes regardless of the SNR, the firing rate of electrodes in LPS- treated animals is one third of controls (Figure 32a, left graph, p<0.01). The periodic recordings (Figure 34 and Figure 35) over four weeks indicate that differences between the electrodes in the different groups occur within the first two weeks. The data supports that the action of LPS occurs before the four week time point of the histology. 147 *+ ,+ !"'!# (!!# !"&!# '!# &!# :1"$;12 !"%!# %!# <39 !"#$%&%#$'( !"$!# $!# !"!!# !# 71#%'8(,22(9#$'%( !# (!# $!# )!# !# (!# $!# )!# 06( ,-./(012$*.'(343(5 )*+( )*+( Figure 34. Measures of non-evoked electrophysiology quality in unanesthetized animal over four week time course of implanted electrodes showing units per site and average noise voltage for all electrodes. a)Average units (SNR>=1) per site across all electrodes implanted in animals for the LPS-treated animals (black) and control animals (blue-grey). b)Average noise voltage peak to peak in millivolts across all electrodes implanted animals for the LPS-treated (black) and control animals (blue-grey). Stars (*) indicate statistical significant between groups on the marked measurement day (p<0.05), pluses (+) indicate trending towards statistical significance (p<0.1) )* +* (!!" #'!" '!!" &!!" #!!" %!!" -./01.2 $!!" '!" 34! !"#$%&'&%() #!!" !"#$%&'&%() !" !" !" #!" $!" %!" !" #!" $!" %!" *+,) *+,) Figure 35. Peristimulus time histograms (PSTH) in anesthetized animal over four week time course of implanted electrodes showing changes in spiking rates based on stimulus for all electrodes. a)Average change in spiking rate for the 1 second during stimulus minus the 1 second before the stimulus for the LPS-treated animals (black) and control animals (blue-grey). All spikes that exceeded the automatically generated threshold are plotted. b)Average change in spiking rate for the 1 second during stimulus minus the 1 second before the stimulus for the LPS-treated animals (black) and control animals (blue- grey). Only spike clusters that had an SNR >=1 are plotted. Stars (*) indicate statistical significant between groups on the marked measurement day (p<0.05), pluses (+) indicate trending towards statistical significance (p<0.1) 148 In addition to evoked spiking rates, we examine evoked LFPs. In contrast to spikes that are usually recorded from cell bodies within a sphere of 50 µm from an electrode (Henze et al., 2000), LFPs record signals within 250 µm of the electrode and associated with activity in the dendritic trees of neurons (Mitzdorf, 1985; Xing et al., 2009). We analyze six different frequencies within LFPs. In response to a whisker stimulus, previous research shows that the power increases with frequency (Jones and Barth, 1999). That trend is exhibited here (Figure 32b) for the control animals, but the same trend is not seen in the relatively flat power across frequencies in LPS treated animals. For power comparisons between treatment groups, power is significantly different between groups for all frequency bands except 8-12 Hz (p<0.01). The LFP power in LPS-treated animals is generally flat across frequency ranges, and the control data aligns with previous research(Jones and Barth, 1999). The data suggests that electrodes in control animals are better able to detect changes in LFP power. It is unknown why the LFP power difference in LPS animals is non-zero. It is possible that the whisker stimulation system causes a non-zero offset due to noise when the electrode cannot record a sufficient signal. Immunohistological data indicates that LPS creates a more pronounced response that resulted in a greater non-neural cell density. The larger cell density correlates with an increase in electrode impedance. In addition, the LPS treated animals have less neurons and dendrites around the electrode. The increased impedance and decreased neural proximity ultimately leads to a decrease in several measures of recording quality for single units and LFPs. It is unknown whether the determining factor for recording quality is impedance, neural density, or network function. Since LFPs measure dendritic 149 activity, the differences between control and LPS groups for LFP activity indicate dendritic function has been altered. The lack of LFP power modulation in LPS-treated animals correlates to differences in dendrite labeling, MAP2. Neural recording ability degrades with LPS administration, but labeling for neuronal health indicates that neurodegeneration is not widespread in either group at four weeks. 5.6. Conclusions The goal of this work was to examine the effect of the tissue response on electrode recordings in order to improve intracortical electrode design and performance. We have shown that LPS modified the tissue response as evidenced by DAPI and other immunohistochemical measures. The increase in cell density around electrodes correlates with increased impedance. The decrease in neuronal density <50 µm from the electrode in LPS-treated animals correlates with decreases in several measures of neural recording, most notably the clinically-relevant evoked response. Collectively, we have supported our hypothesis that an increased tissue response does degrade neural recordings through usage of the inflammatory stimulant LPS. The work suggests that LPS treatment is a method to artificially hasten the degradation of recordings to complete accelerated lifetime testing of novel electrode technologies. LPS-treatment provides an important development tool. More complete studies correlating immunohistochemistry and electrophysiology for individual electrodes should be completed to assess the impact of tissue impedance, neural density, or neural function on neural recordings. 5.7. Acknowledgements This work was supported by F31-NS063640 from the National Institute of Neurological Disorders and Stroke. Additional support was from the Department of 150 Veterans Affairs grant numbers C3819C, F4827H and B6344W. The authors acknowledge the support of Nick Langhals and the Center for Neural Communication Technology (NIBIB P41 EB002030) as the source of the bulk of the electrophysiology analysis code used here. The authors acknowledge the support of Paul Marasco, Dawn Taylor, and Cameron McIntyre as well in their support and equipment. None of the funding sources aided in the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the paper for publication. The authors have no conflicts of interest related to this work to disclose. 151 CHAPTER 6. CONCLUSION The work in the dissertation examined the glial scar response to intracortical electrodes. Specifically, the work investigated how substrate stiffness modified the glial scar and how modification of the glial scar affected electrode recordings. The work improved understanding of the glial scar response to intracortical electrodes and highlighted the important questions to be answered to enable continued development of electrodes for clinical and research applications. Previous in vitro research indicated that material stiffness affects cell proliferation, cell differentiation, and other cell functions (Engler et al., 2006; Georges et al., 2006; Ostrow and Sachs, 2005). The effect of material stiffness on tissue response was not directly examined in vivo in previous work. Therefore, we examined the material stiffness and the tissue response towards the development of intracortical electrodes. A novel nanocomposite enabled the study of the effect of material stiffness on the tissue response to an intracortical implant. The nanocomposite exhibited a higher storage modulus (E’ = ~5 GPa) than the neat polymer microprobes (E’ = ~2 GPa) and dynamically changed its modulus from 5 GPa to 12 MPa. The mechanically dynamic properties were previously well-established in bulk samples in vitro (Capadona et al., 2008; Shanmuganathan, 2010). Here, we showed that the behavior and properties are the same for microprobes tested in vivo. The nanocomposite enabled insertion without additional support structures, but chronically, the nanocomposite was soft. The chronic 152 softness enabled a study of the effect of material stiffness on the tissue response to a non- functional electrode. The effect of substrate stiffness on the glial scar response was examined through chronic trials in the rat cortex. The nanocomposite and the surface-matched, stiff wire coated with the polymer substrate were implanted. Results indicated each implant type induced different responses. Softer implants were able to preserve more neuronal nuclei density around the device at four weeks post implantation. Correlating with the increased neural density several mechanically associated factors, chondroitin sulfate proteoglycan (CSPG), glial fibrillary acidic protein (GFAP), and vimentin, were modified around the nanocomposite. At eight weeks, the neural density was maintained around the nanocomposite, but the neural density around the implant and wire were similar. The nanocomposite neural density matched the best in class wire (Tresco and Winslow, 2011). Notably, the neural density around the nanocomposite was maintained despite elevated levels of activated macrophages. While some of the mechanical factors, CSPG and vimentin, were similar at eight weeks, a compaction of GFAP around wire implants suggests a difference in morphology around the two implant types. Previous research indicated that astrocytes are a diffusion limiter, so the compaction may lead to increased impedance and worse recordings (Roitbak and Sykova, 1999). By coating a stiff wire with the same polymer used in the nanocomposite, we were able to isolate the mechanics of the system while maintaining the same material interface with the brain tissue. The results suggest that we were able to separate the chemically and mechanically mediated response to implanted materials. The soft nanocomposite resulted in a different inflammatory response than from the stiff wire. 153 While it is generally believed that an elevated tissue response degrades the electrode recordings, no direct in vivo evidence was found in an exhaustive search of the literature. Therefore, the effect of stimulating the tissue response on electrode recordings was examined. Results show that the quantity of neuronal cells was decreased and the quantity of non-neuronal cells was increased around electrodes in LPS-stimulated animals. The increase in non-neuronal cell density correlated with increased impedance measurements in LPS-treated animals. Additionally, several measures of neural recording quality were worse in LPS-treated animals that correlated with decreases in neuronal density. The results showed that a greater tissue response decreased the quality of cortical neural recordings. The electrophysiology results indicated that the glial scar response is an important factor in determining electrode performance. The next logical question to continue this work is to ask whether even softer materials could further improve the tissue response. While the nanocomposite is several orders of magnitude softer than traditionally used materials, the nanocomposite is still two or three orders of magnitude stiffer than the brain. The similarity between neural density and other labels at eight weeks indicates that even softer materials are needed. To fully understand the response, more factors and more time points will need to be assessed. Accounting for differences between the wire and the nanocomposite, another alternative is that the nanocomposite sped up the response. Both implant types exhibited the same neural density at eight weeks. The neural density improved from four to eight weeks around wire implants. If a similar neural density improvement did occur around the nanocomposite, it would have happened before the four week time point. Therefore, the results may indicate that the nanocomposite hastened the response to an 154 intracortical insertion. Investigations at shorter time points would determine if the nanocomposite quickens the response. Investigations at longer time points would determine if the nanocomposite continues to preserve neuron density in spite of activated macrophages. Though the nanocomposite and wire exhibited similar neural densities at eight weeks, the mechanism for the increase in neural density around the wire from four to eight weeks is unknown. One possible reason for the increase in density around wire implants from four to eight weeks was the compaction of the glial scar. Examining DAPI labeling for all cells, the results did not show marked compaction from four to eight weeks. Though neurodegeneration was not performed with Fluoro Jade C (FJC) labeling, previous researchers have rarely seen marked FJC labeling after 2 weeks (McConnell et al., 2009). Therefore, the data suggests that the neurons were able to migrate back to the wire. The nanocomposite showed the ability to modify the tissue response, but current methods are unable to extract electrode performance from the tissue response. We stimulated the tissue response with LPS, and increased levels of DAPI indicated a greater non-neural density. Though LPS has been shown to activate microglia (Weberpals et al., 2009), microglia and macrophage levels were similar between controls and LPS-treated animals at four weeks. The increased DAPI levels indicate that LPS created a response, and one possibility is that the LPS-induced response ended by the four week time point. Similar levels of ED1 and IBA1 on the contralateral side support the hypothesis that LPS action ceased by four weeks. Another alternative is that the LPS caused a more 155 permeable BBB after the implant, and the permeable BBB allowed the entry of blood borne cells that culminated in the increase in DAPI levels. LPS has been shown to have many effects, including action on synaptic proteins, as well. Therefore, it is uncertain whether LPS affected the neural recordings through decreased neural density, increased impedance of the glial scar, or degraded neural health. To examine the cause of decreased neural density at four weeks, Fluoro Jade C (FJC), was used to investigate neurodegeneration. The label showed a paucity of degenerating neurons. The results indicate that migration rather than neural degeneration is the important factor contributing to recording ability. The studies completed in this dissertation support several interesting new hypotheses and point to continuing research directions. At this time, tissue histochemistry alone cannot predict electrode recording quality. A strong focus needs remain on electrode recordings with increasing emphasis on understanding the relationship between electrode function and immunohistochemistry. Others have started this work, but continued progress is important (Purcell et al., 2009; Williams et al., 2007). The most important question to address is which part of the glial scar response is causing the degradation of neural recordings (neural proximity versus glial scar impedance versus neural health). Results from our work and others show little neurodegeneration after 2-4 weeks. Our work and others also indicate that neurons increase around the electrode after the initial 2 week implant period. The data suggests a migration of neurons based on the surrounding chemical milieu. Assessment of the mechanisms controlling the density of neurons around an implant is important. 156 One way to address which part of the glial scar is causing the degradation of recordings is to leverage the variability of intracortical electrode responses. Performing detailed IHC and neural recordings, on a large set of animals, one could correlate electrode performance to distinct immunohistochemical measures. The correlation or lack of correlation would indicate the relative importance of different factors given a large enough data set. The analysis could occur within treatment groups contained in a larger study. The study may require large numbers, but the study would provide a great deal of information to improve the design of electrodes. Modeling tools and techniques have not been employed extensively in the area of cortical electrodes. Models, both electrical and mechanical, could help direct experiments and provide deeper insight into the interactions between the tissue and electrode. Improving and extending published models from 2005, electrode design could be advanced through iterative testing of models followed by validation of model predictions in vivo (Subbaroyan et al., 2005; Lee et al., 2005; Lempka et al., 2006; Moffitt and McIntyre, 2005). Some of the more intriguing mechanical-oriented questions include: how much does micromotion affect recording variability, how does the glial scar buffer the mechanical effects of an implant, and does the glial scar limit diffusion of cytokines, blood, or drug agents. The value of electrical/neural analysis on the impact of glial scar in stimulation has been previously demonstrated. Similar computational studies related to recordings would be beneficial. Specifically, an analysis of the effect on recordings from a change in neuronal proximity, neuronal density, neuronal firing, and scar impedance would be useful questions to analyze in simulations. A recent publication has completed an initial analysis of simulations regarding these electrical/neural issues. Further work 157 needs to be completed to improve models and understand the in vivo and clinical implications of the simulations (Lempka et al., 2011). The field could further benefit by adopting techniques to improve understanding of the complex glial scar reaction. Improved techniques are needed to quantify the response, gather longitudinal data, and examine hundreds or thousands of factors at a time. Pooling of data from many diverse labs would benefit progress, as well. The field should also look to modern neuroscience for tools to control certain aspects of the response in order to provide greater understanding. Knock out mice and other tools from biomedical science that allow for a more controlled model can improve understanding of the actors in the glial scar response. Knock out mice and accompanying technologies can control the type and timing of the response. Two particular interesting areas that have not been well studied in the cortical electrode field are the blood brain barrier and the relative importance of macrophages vs. microglia. Through the use of dextrans with controlled sizes or through the use of chimera animals, the tools exist to analyze the integrity of the blood barrier as well as to determine whether activated inflammatory cells in the brain are of brain (microglial) or blood (macrophages) in origin. Techniques could also be used to better assess the relative neurodegenerative and neuroprotective contributions of these cells. Proceeding from the basic science to the engineering, significant work in the field needs to be conducted on connector technology as well. Even though almost every research lab has encountered connector problems and many have noted problems in published studies, there is limited published research and controlled studies to better 158 examine the failures of electrode connectors. There are easily applied methodologies from other fields to understand mean time to failure in connector technologies, and it is important to understand and improve these failures, as they are just as important as any failure on the way to clinical deployment. In conclusion, the field of cortical electrode development continues to grow in new directions with an ever-growing number of researchers. To this growing base of literature, we add support to two long held beliefs that material stiffness affects the tissue response and tissue response affects the quality of electrode recordings. Novel techniques and experiments will continue to develop. This information will enable novel techniques to interface with the neural tissue and improve understanding of the relationship between the brain and electrodes. 159 Appendix A.1. Mechanical Inserter and Load Cell (Force) Setup Procedures Components -Small Break out box - The motion (UMI) break out box (white) -Large break out box - The Data Acquisition (DAQ) break out box (white long rectangle) -Power Supply -Motor Driver (labeled PDO box) -Stepper Motor -Encoder attached to Motor -Load cell (gold square) -Amplifiers and Filters (Preamp and CEDs) -Computer + Software -Grippers (to hold the motor) Connecting the Wires (see external documentation for more information) Position Stepper Motor into the Grippers (on free stand or stereotaxic arm, adjust the stereotaxic arm so that the motor rests firmly on the upright metal rod). Screw on Load cell onto end of Stepper Motor. Connect 4 wire, green connector from Motor to Motor Driver Connect 6 wire (13 connection slots) connector from UMI box to Motor Driver Connect Gray (unbraided) cable from power supply and preamp to CEDs (into transducer Channel 1) Connect preamp to +12V and GND Connect silver (braided) cable from pre-amp to Load cell Connect Wide Blue connector from Computer (from Motion board) to back of UMI box Connect UMI to power supply (+5V and GND) Connect Channel 1 CED amp out to Channel 0 (top left of DAQ breakout box) via a coax cable. Make sure that nothing else is connected to channel 1 of CED (only trigger and amp out connected) Plug in Power Supply and Motor Driver Connect CEDs to computer (make sure serial cable is attached in the back of the CEDs and to the Serial Port hub connected via USB to the computer, both CEDs must be connected to serial cable and turned on for proper function) (If large white DAQ breakout box is not connected to computer, make sure that it connects to large blue connector from the DATA I/O port on the computer output) Turn on Computer Turn on CEDs 160 Software Setup Login Run Try1902 (if you can’t find it in Start menu, go to Program Files/Try1902/ Set to Com 3, channel 1 (Com # changes depending on where CED serial port is plugged in, so if it isn’t on Com 3, switch until you find it) Set to Gain to 100x, -600mV offset, Low Freq: 30Hz. Hit ok Run Labview Load Single Axis Test (either under open > recent files or My Documents/dashboard) Operation of LabView Program Figure 36 shows how the LabView program will look when you load up Lab View. Hit white arrow at top left to run program Figure 36. Snapshot of Inserter program before start of the program. 161 Then the arrow should turn from white to black. If you see a box (Figure 37) when you start it, hit Do It. Figure 37. Popup window that is shown when the motor has not been previously initialized. Select Test Off (Figure 36) to end program The main buttons to pay attention to are: Green Start Button Calibrate Load Slow Up Params Load Fast Down Params Load Up Params Other areas to be familiar with: Trajectory Parameters (will need to convert millimeters to steps ) Graphs of Load Cell Voltages To move down: Hit Load Fast Down Params button that will set the Trajectory Parameters to move the inserter 2mm downwards (see end for conversions from mm to steps). Click Start (the velocity is 2mm/s and the distance is 2mm) To move up: Hit Load Up Params button that will set the Trajectory Parameters to move the inserter 2mm upwards (see end for conversions from mm to steps). Select Start. (the velocity is 2mm/s and the distance is 2mm) NOTE: negative # in Target position= downward motion (extending arm of motor) positive # in Target position = upward motion (retracting arm of motor) To Record to File Unhighlight Record to Dummy File and enter the filename you want to record to. Each time you hit Start, the program will overwrite the file. Select the movement type and hit Start. 162 To Calibrate Hit Calibrate, then select Start. The box below Average will indicate the average voltage measured without a movement. (Note if the signal maxed out the system with no weight, you may be able to fix the issue by switching the DAQ motion box from SE to DIFF by flipping the switch on the front of the box near channel 0) Error If there is an error, the lights will usually look like Figure 38, hit HALT to reset things. It is possible that the motor is all the way at it max extended or max retracted and cannot move in that direction anymore. Full Calibration Procedure Weigh out small weights such as paper clips, record weight (see Calibration table below). The inserter should have the ceramic forceps tip held on by the small cut end of a paper clip. Hit Calibrate, make sure the things are steady and free from vibrations. Hit Start. Record value shown in the Average box as Voltage in table below. Exchange other weights onto forceps tip, and hit Start when paper clips are settled. Record the weight. Continue to change the weight (paper clip) and record the voltage until done. Create Calibration Curve Plot mass (x) vs. Voltage (y) in Excel Add a linear trendline and show the equation: y=mx+b The value of m will be the Volts/gram for the load Cell that you will need to enter into the post processing macro (zero_smooth.xls) Figure 38. Example of Indicator lights when and movement error has occurred. Conversion Of Steps <--> mm Position: 25600 steps/rev *(12 rev/inch) * (1 inch/ 25.4 mm) = 12094.48819 steps/mm Speed: 1 rev/min = 3.527 x 10-2 mm/s 163 Post Process Run filter.m in Matlab to filter the recorded signal further (will produce same filename with _2 on end, also it fills in values that falsely get recorded as 0 Volts in the raw file) Run macro in zero_smooth.xls in Excel and enter in calibration data (volts/gram from calibration curve. The macro plots the results of force vs. time ! produces .xls file from what you gave it (choose _2 version you just made with Matlab as input) Run macro in graph in position.xls in Excel to graph force based on position instead of time. Confirm overwrite of file, when asked. Warning, the newer version of Excel has problems with large amounts of data. Table 7. Table for Calibration Item Weight (g) Voltage (V) 164 A.2. Surgical Procedures for Rats Implantation of Cortical Probes Supplies Implant Surgical Pack-Autoclave (amount for 1 animal) Scalpel Dura Hooks (microprobe) Forceps (7s) Gauze (2 or so inches) Cotton-tipped applicators (25+) Small Screws in vial (x4) Screwdriver Handheld drill and drill bit Tin Pans (x2) Needle Holders Forceps to Suturing (teeth on end to grip skin) Suture Scissors Drape Scissors Drape on stereotaxic frame (x2) Drape on animal Disposable Biopsy Punches (3 mm diameter, one per hole ) Syringe for Saline (10 cc) Needle for syringe (19 gauge) Sterile Saline Blade for Scalpel (15) Probe Assembly Microprobes in vials Ceramic Forceps tips Paper clip connectors Blunt forceps Drape for assembly area Gauze Implant Rig/Mechanical Inserter: Setup C clamp to mount motor to stereotaxic frame See Mechanical Inserter Setup 165 Closure Kwik-Sil Cotton-tipped applicators (Q-tips) Sutures (Henry Schein, polypropylene blue monofilament; 5-0; 3/8 circle. REF# 101-6409, simple interrupted suture) Dental Cement Microprobe Preparation Microprobes were autoclaved in glass vials before surgery along with other tools listed under Probe Assembly supplies. All tools were spread out on a sterile cloth on a table. A supersaturated solution of glucose in DI water was created. A small glass vial containing DI water was heated to 50oC. Twice the weight of the water was added in grams of glucose into the DI water. After mixing well, the mixture was allowed to cool. The solution was filtered through a 0.22µm sterile filter into a sterile glass vial. The process therefore created a sterile supersaturated glucose solution. A ceramic forceps tip was placed on the sterile cloth. Using the blunt forceps, a drop of glucose was placed on the end of the ceramic forceps tip. After wiping excess glucose off the forceps, a microprobe was placed on the end of the ceramic forceps tip. The preparation was allowed to dry overnight. Surgical Overview (more details are listed in Chapters 3,4) Pre Surgery Sterilize microprobes and prepare (see above) Setup Mechanical Inserter + Load Cell + Amplifiers, mount stepper motor on stereotaxic arm Calibrate Load Cell Lay out Pack Get dental cement and Kwik-Sil out Setup surgical scope, attach camera if desired (camera driver needs to be plugged in. The only connotations are coax cables running to camera and recorder) Setup dental drill and area (if being used instead of biopsy punches) 166 Position stereotaxic frame to leave room for drill at front of table, position in well lit and accessed by surgical scope Surgery Plan Knock Down Animal Open sterile pack and open disposable tools (scalpel, needle, syringe, etc.) Give cefazolin/marcaine Shave Animal head Insert Animal in Ear Bars Betadine and Alcohol to sterilize head area Surgeon Sterilize (put on face mask and cap, wash hands with scrub, dry, put on robe, put on sterile gloves) Draw saline into syringe If using dental drill instead of biopsy punch With help of assistant lay out drape for dental drill handpiece+ insert drill bit + Put stockinette on Scope and dental drill Rest dental drill on head of table on clean drape Hydrate animal head with saline as needed Midline incision with scalpel, move away periosteum with cotton tipped applicators Use eye speculum to keep skin/muscle out of way Use biopsy punch as mock ruler, and make shallow mark for implant site Drill holes for anchor screws and for ground screws, one rostral to each hole - leaving 3mm or so between the implant hole and insert screws Insert screws, being careful to not go too deep as to indent brain Open skull with biopsy punch (or dental drill) Use rongeurs to make slightly bigger hole in skull, if needed Use dura pick to make hole in dura, use jeweler's forceps to clip back dura (5's with tip slightly bent to be a clipping claw) Use surgical scope or jeweler's glasses to aid visualization Assistant/Inserter Portion Swing inserter mounted on Stereotaxic arm into surgical field Attach electrode (ceramic forceps tip) onto load cell with paper clip Manually lower electrode to area near implant site Surgeon uses sterile gauze to lower it very near surface of brain Assistant activates inserter program on command of surgeon If you are implanting more than one item, drill next hole, and implant before applying Kwik-Sil Apply Kwik-Sil, try to make sure it doesn't adhere to skin and/or instruments. If it does, it is better to wait until it dries - it has problems sticking to wet things Mix dental cement in tins with cotton tipped applicators Apply dental cement to head covering skull defects, making sure to get dental cement underneath anchor screws Wait until dry Depending on surgery type, suture the wound closed with suture 167 Implantation of Intracortical Electrodes Supplies Surgery Surgical Pack-Autoclave (amount for 1 animal) Dura Hooks (microprobe) Forceps (7s) Dental Drill + bit Handheld drillholder + bit Screwdriver Small Screws in vial (x4) Teflon coated forceps Jeweler's forceps Flat forceps 2 5's forceps Fine scissors (to cut skin if necessary) curved scissors to separate fascia Drape Scissors Scalpel handle Small Rongeurs Periosteum elevator 2 hemostats (90 degree curved top, small and big to hold skin away) Eye speculum (retractor) 1 inch diameter Stockinette - 1 long for drill , 2 short for scope Sterile Drapes (x4) Gauze (2 or so inches) Cotton-tipped applicators (25+) Items to Gas Sterilize (Ethylene Oxide-EtO) Ruler Tape Pen Disposable 15 scalpel blades 10cc syringe 19.5 Gauge needle tip Surgical Spears Syringe (10cc) (x2) Other tools Surgical scope Jeweler's glasses Closure Kwik-Sil + tips Dental Acrylic Cotton tipped applicators (Q-tips) to mix dental acrylic Tin Pans (x4) to mix dental acrylic 168 Triple antibiotic ointment Forceps Electrical Recordings Electrodes Cables/Connectors to electrodes Magnetic Base on weight and metal bars Micromanipulator to mount on Magnetic Base and bars Metal rod with alligator clip mounted on end Grounding wires Surgical Plan Pre Surgery Gas Sterilize Electrodes Lay out Pack Get table for electrode staging area Put out manipulator and weight, adjust to proper angle and insert electrode holder Get dental cement and Kwik-Sil out Setup surgical scope, attach camera if desired (camera driver needs to be plugged in, and only other things are coax cables running to camera and recorder) Setup dental drill and area Position stereotaxic frame to leave room for drill at front of table, position in well lit and accessed by surgical scope Surgery Plan Knock Down Animal Open sterile pack and open disposable tools (scalpel, needle, syringe, etc.) Inject Animal with LPS or saline, and give cefazolin/marcaine Shave Animal Insert Animal in Ear Bars Betadine and Alcohol to sterilize head area Surgeon Sterilize (put on face mask and cap, wash hands with scrub, dry, put on robe, put on sterile gloves) Draw saline into syringe With help of assistant lay out drape for dental drill handpiece+ insert drill bit + Put stockinette on Scope and dental drill Rest dental drill on head of table on clean drape Hydrate with saline as needed Midline incision with scalpel, move away periosteum with cotton tipped applicators Use curved scissors and forceps to separate skin from muscle + fascia (need to see iridescent color) Use hemostats to keep skin/muscle out of way (just lean very curved hemostats) Use Periosteum elevator + skin forceps to separate muscle from skull (elevator) 169 Mark midline with pen, area to drill (4mm lateral, -3 Bregma - basically a 3mm circle around skull ridge) Use Dental Drill to open skull Use rongeurs to make slightly bigger hole in skull if needed Once skull open, drill holes for anchor screws and for ground screws, one on contralateral hemisphere for wire, isolateral: one in rostral of bregma, another caudal lambda Use dura microprobe to make hole in dura, use jeweler's forceps to clip back dura (5's with tip slightly bent to be a clipping claw) Use surgical scope or jeweler's glasses Non Sterile Portion (touch anything you want except sterile stuff) Have a second syringe and needle for saline, draw it up with saline Move scope aside Slide stereotaxic frame over to make room for inserter Move weight and micromanipulator into position Attach electrode to electrode holder Insert electrode and holder into micromanipulator (make sure that the manipulator has room to descend, and the fine manipulator is set to 0) Note electrode # on sheet Take photos (before insertion and after insertion – ideally before Kwik-Sil application to document spot of insertion) Lower electrode to a few millimeters above skull Tie wires to ground screw (use certain forceps that are not as sterile as the rest) Lower electrode 500-700um deep into brain (1mm total for Michigan, insert with big scale until you barely see the top electrode and then advance with small for 300um more) Seal with Kwik-Sil - wait ~5mins, slowly move skin and hemostat away from Kwik-Sil Seal with dental acrylic/ make sure to cover all screws well, and minimize edges where animal could pry up dental acrylic. (will take multiple batches) Let dental acrylic dry. While holding electrode holder open, loosen micromanipulator holder, remove holder Collect recordings (headstage + TDT + impedance, etc.) Move animal to recovery Move weight and manipulator back to resting position Remove drapes Reglove (Sterile again) Put down new drapes, being careful that the hemostats that hold the side of the skull always stay out and are folded under an old drape. ! ! ! ! ! ! 170 A.3. Perfusion + Slicing + Labeling Brain Tissue Perfusion (full details and illustrations are available in Powerpoint presentation) Equipment: Anesthetic: ketamine + xylazine 6 ml syringe 21-gauge hypodermic needle Surgery: 1 pair large blunt/blunt curved scissors (~14.5 cm) 1 pair straight iris scissors (~9 cm) 1 pair standard tweezers 2 large hemostat forceps - straight (~19 cm) 1 small hemostat forceps - straight or curved serrated (14 cm) 1 15-gauge blunt or olive-tipped needle Perfusion: Pump + Tygon Tubing ~500 ml phosphate buffered saline, pH 7.4 250 ml fixative: 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 Shallow tray, approximately 10” x 10” Dissection: 1 pair standard sharp/blunt heavy duty straight scissors 1 pair medium curved or straight rongeurs (14-16 cm) or skull bone removal pliers, 1 pair small rongeurs 1 pair straight iris scissors (~9 cm) 1 micro-spatula (double 2” flat ends, one rounded, one tapered to 1/8”) Scalpel and blade (10) Postfixation & storage: 1 50 ml glass vial Brief Overview of Perfusion Procedure (use attached datasheet to make notes): Prepare pumping system Anaesthetize animal Access midsection Open ribcage Clamp vessels Open left ventricle and insert needle Cut right atrium Start pump, continue with PBS until no blood seen left exiting atrium, then switch to formalin (PFA) Extract brain from skull case and place in vial with PFA Embedding and Slicing of Rat Brain (for free floating sections) Rat perfused and fixed with 4%PFA Brain extracted 171 Put into 4%PFA (transfer to 30% sucrose 48-72 hours before slicing) Section using cryostat Remove probe/electrode, place in dish with PBS or cryoprotectant Trim brain - cut off cerebellum + olfactory bulb, remove any extra dura Use india ink to move a mark in front of implant on left hemisphere, behind on right hemisphere Place brain level in small square mold Add OCT compound Place OCT + mold into crushed dry ice within cooler ~10 minutes Ready cryostat (-20 deg C), roll plate + tray + chucks clean, brushes loaded Remove mold from dry ice, cut slits into brain mold Place OCT onto chuck, affix brain, let rest in cryostat for a few minutes Mount chuck in cryostat, make sure level Set thickness: 30um Use buttons to advance cryostat until close to mold Based on how it is cutting, re-level chuck Reset counter when you get the first slice Use flat tweezers to move slices into PBS wells or cryoprotectant Labeling with Immunohistochemical Markers Day1 block step aspirate PBS add blocking solution (see below for formula) leave in fridge overnight Day2 aspirate add primaries (diluted in blocking solution), fridge overnight Day3 aspirate, add PBS rinse on shaker (3x 15mins) ~ 150 RPM add secondaries (diluted in blocking solution) 1hr @ room temp aspirate, add PBS rinse on shaker (3x 15mins) ~ 150 RPM Mount on slide with 1.5 coverglass and Fluoromount-G onto glass slide- let dry Block Solution (250 mL total) 4% (v/v) normal goat serum - 10mL 0.3% (v/v) Triton-X-100 - 0.75mL 0.1% (w/v) sodium azide - 25mL of 1% sodium azide in 1xPBS Shake well until all dissolved 172 Cryoprotectant (1L) - from Watson RE, Jr., Wiegand SJ, Clough RW, Hoffman GE. Use of cryoprotectant to maintain long-term peptide immunoreactivity and tissue morphology. Peptides 1986;7(1):155-9. 500mL PBS (50% v/v) 300g sucrose (30% w/v) 10g polvinylpryrrolidone (1% w/v PVP-40 sigma) 300mL ethylene glycol (30% v/v/ Fisher) ! 173 A.4. Image Analysis (in Matlab) Microscope – Collection of Images Photograph portions of slices as desired. Site options include: • Implant site (suggested that it be center of the image) • Contralateral side, similar position to implant • Location away from implant site, either contralateral or ipsilateral It is suggested that photos are taken at a 10x magnification, and the scope automation is utilized to take one bright field image for each fluorescent image. A 1x or 4x image can be taken to visualize the location of the implant or overall condition of the slice. 20x or 40x images (with or without aid of apotome or confocal microscope) can be taken as well to better examine specific morphology of certain cell types. Files can be saved as .ZVI or .JPG, but make sure that .JPGs are annotated with the proper scaling and camera settings. Conversion of Images (for ZVI images) Images saved as ZVI need to be converted to an appropriate format. Open AxioVision and run File>Export -Select or unselect Create Project Folder as you see fit, select all for mode, make sure only “use channel names” under Channel Selection is selected, add the files, hit run batch -Select JPG or TIF as file style. TIF is suggested since it is lossless. This creates all your JPG Black and White images Image Analysis – Intensity Detection of Implant Area – Bright field analysis – bf_background.m Given a bright field image that corresponds to a fluorescent image Detect the edges in the image Smooth the edges in the image by applying an erosion operation to the inverse of the image through applying three pixel disk shape for the erosion. If the program chooses the incorrect implant site, the user can manually trace the implant area Save the coordinates of the implant area to a file (.mat) Image Intensity Analysis – Fluorescent Analysis – thresh.m Given a fluorescent image that has a corresponding bright field image where the implant border has been defined Depending on the analysis desired and the choice of the scaling for background intensity, input the background intensity from another image. Load the file storing the border area of the implant from bright field Compute the centroid of the implant area Compute and draw 100 radial lines emanating from centroid 174 Compute the intersection of each radial line with the border of implant area Measure intensity on each radial line for the portion external to implant border, take mean Compute in-image average based on user input (left, right, or both sides of image far from implant site) Store results to file (.mat or .csv) Image Analysis – Cell Count (Neuron cell bodies) Detection of Implant Area – Bright field analysis (see above in Intensity analysis) Image Cell Count Background Count Analysis - Fluorescent Analysis – histo_backg.m Given a florescent image Define a representative area in the image in order to compute the cell density in the image Manually mark cells Cell density is computed based on marked cells and area of defined area Save the result (.mat) Image Concentric Ring Area Analysis- Fluorescent Analysis – true_area.m Given a bright field image where the implant area outline has been defined Load the file storing the border area of the implant from bright field Compute the area of the 10 µm ring around the implant area Extend the border by 10 µm (the radius increases by 10 µm each step). Compute the area of the 10 µm ring or donut Continue to extend and compute the area until the whole image is covered Save the area of each concentric ring (.mat) Image Cell Count Analysis – Fluorescent Analysis – histo_neun.m Given a fluorescent image that has a corresponding bright field image where the implant border has been defined Depending on the analysis desired and the choice of the scaling for background count, input the background count from another image or from the current image previously computed in histo_backg.m. Automatically detect cell bodies Convert grayscale image to black and white image by putting image through threshold where only intensities over a certain limit are assigned as white where all other are assigned black. Morphologically close the image (fill in black holes) with a 1 pixel wide disk shape. Label white-connected regions (cell bodies) Compute centroid of each cell body Manually add or subtract cell bodies that were selected erroneously Load the file storing the area of the concentric rings (true_area.m) Load the file storing the border area of the implant (bf_background.m) Compute the centroid of the implant area 175 Compute each line that connects the centroid of a cell body and the centroid of the implant area Compute the intersection of each line with the border of implant area Record the distance from each cell body centroid to the border Bin each distance into 10um bins (histogram) Scale bin counts Scale based on area (from true_area.m) Scale based on background counts (from histo_backg.m) Store unscaled and scaled results to file (.mat or .csv) Statistical Analysis After the intensity or cell count analysis, there are two options to statistically examine the data 1)Simple Two Sample T-Test In this method, there is either one sample per item (animal or shank) or you average several measures together to create one sample per item Create two groups, each with a set of measures Compute Two Sample T-Test Example in R: t.test(control, lps, var.equal = TRUE) 2)General Linear Model (see Chapter 4 for other information) In this method, you may have several samples per item (animal or shank), but you do not want to average the several samples together. In this case, you can create a table of values and feed the table to a statistical program. Though an additional sample for an animal does not carry as much statistical weight as an additional animal, one does not lose information by averaging the several samples together Example Model from Chapter 4: Outcomeijk = "0 + "1 * IMPLANT _ TYPEijk + b0,i + c0,ij + eijk The intensity or cell count is the outcome, and the model included a baseline level for the outcome (b0), a fixed effect for the difference in the outcome based on the implant type (b1), a random effect for each animal (b0,i), a random effect for each slice within each animal (c0,ij), and residual error (eijk). The letters: i indexes the rat, j indexes the slice, and k indexes the side. If b1 is significantly different than zero, the implant type created a significantly different outcome (intensity or cell count). A representative data table (Table 8) that would be fed to a statistical program is included. Example in R: summary(lme(Intensity~Implant,random=list(~1|Animal,~1|Slice), data=trial_data)) 176 Table 8. Representative data tables of peak intensities measured across several animals to be given to statistical program to utilize a general linear model to analyze statistically significant effects Animal (i) Slice (j) Side (k) Implant Type Peak Intensity 3F1 12 2 NC 2.6933 3F1 12 1 WIRE 2.2431 3F1 20 2 NC 2.1184 3F2 27 2 NC 2.8445 3F3 13 1 NC 3.0624 3F3 13 2 WIRE 3.1715 3F4 27 1 NC 2.8193 3F4 31 1 NC 1.7035 177 References Abbott N J 2002 Astrocyte-endothelial interactions and blood-brain barrier permeability J Anat 200 629-38 Abbott N J, Ronnback L and Hansson E 2006 Astrocyte-endothelial interactions at the blood-brain barrier Nature Reviews. Neuroscience 7 41-53 Alberts B 1994 Molecular Biology of the Cell, ed B Alberts, et al. (New York, New York: Garland Publishing Inc.) pp 948-1009 Alonso G 2005 NG2 proteoglycan-expressing cells of the adult rat brain: possible involvement in the formation of glial scar astrocytes following stab wound Glia 49 318-38 ALS Association 2011 Who Gets ALS. Washington, DC. Alves O L, Bullock R, Clausen T, Reinert M and Reeves T M 2005 Concurrent monitoring of cerebral electrophysiology and metabolism after traumatic brain injury: an experimental and clinical study Journal of neurotrauma 22 733-49 Anderson D J 2008 Penetrating multichannel stimulation and recording electrodes in auditory prosthesis research Hear Res 242 31-41 Anderson J M, Rodriguez A and Chang D T 2008 Foreign body reaction to biomaterials Seminars in immunology 20 86-100 Bakker E 2004 Electrochemical Sensors Analytical Chemistry 76 3285-98 Biran R, Martin D C and Tresco P A 2005 Neuronal cell loss accompanies the brain tissue response to chronically implanted silicon microelectrode arrays Exp Neurol 195 115-26 Biran R, Martin D C and Tresco P A 2007 The brain tissue response to implanted silicon microelectrode arrays is increased when the device is tethered to the skull J Biomed Mater Res A 82 169-78 Birbaumer N 2006 Brain-computer-interface research: coming of age Clin Neurophysiol 117 479-83 Bjornsson C S, Oh S J, Al-Kofahi Y A, Lim Y J, Smith K L, Turner J N, De S, Roysam B, Shain W and Kim S J 2006 Effects of insertion conditions on tissue strain and vascular damage during neuroprosthetic device insertion J Neural Eng 3 196-207 Busch S A, Horn K P, Cuascut F X, Hawthorne A L, Bai L, Miller R H and Silver J 2010 Adult NG2+ cells are permissive to neurite outgrowth and stabilize sensory axons during macrophage-induced axonal dieback after spinal cord injury J Neurosci 30 255-65 Busch S A and Silver J 2007 The role of extracellular matrix in CNS regeneration Current opinion in neurobiology 17 120-7 Buzsáki G 2004 Large-scale recording of neuronal ensembles Nat Neurosci 7 446-51 Capadona J R, Shanmuganathan K, Trittschuh S, Seidel S, Rowan S J and Weder C 2009 Polymer nanocomposites with nanowhiskers isolated from microcrystalline cellulose Biomacromolecules 10 712-6 Capadona J R, Shanmuganathan K, Tyler D J, Rowan S J and Weder C 2008 Stimuli- Responsive Polymer Nanocomposites Inspired by the Sea Cucumber Dermis Science 319 1370 178 Capadona J R, Van Den Berg O, Capadona L A, Schroeter M, Rowan S J, Tyler D J and Weder C 2007 A versatile approach for the processing of polymer nanocomposites with self-assembled nanofibre templates Nat Nano 2 765-9 Carson M J, Cameron Thrash J and Walter B 2006 The cellular response in neuroinflammation: The role of leukocytes, microglia and astrocytes in neuronal death and survival. Division of Biomedical Sciences, University of California- Riverside, Riverside, CA 92521, USA Chidlow G, Wood J P, Sarvestani G, Manavis J and Casson R J 2009 Evaluation of Fluoro-Jade C as a marker of degenerating neurons in the rat retina and optic nerve Exp Eye Res 88 426-37 Churchland M M, Yu B M, Cunningham J P, Sugrue L P, Cohen M R, Corrado G S, Newsome W T, Clark A M, Hosseini P, Scott B B, Bradley D C, Smith M A, Kohn A, Movshon J A, Armstrong K M, Moore T, Chang S W, Snyder L H, Lisberger S G, Priebe N J, Finn I M, Ferster D, Ryu S I, Santhanam G, Sahani M and Shenoy K V 2010 Stimulus onset quenches neural variability: a widespread cortical phenomenon Nat Neurosci 13 369-78 Clark K, Langeslag M, Figdor C G and van Leeuwen F N 2007 Myosin II and mechanotransduction: a balancing act Trends in cell biology 17 178-86 Cullen D K, Simon C M and LaPlaca M C 2007 Strain rate-dependent induction of reactive astrogliosis and cell death in three-dimensional neuronal-astrocytic co- cultures Brain Res 1158 103-15 Damjanac M, Rioux Bilan A, Barrier L, Pontcharraud R, Anne C, Hugon J and Page G 2007 Fluoro-Jade B staining as useful tool to identify activated microglia and astrocytes in a mouse transgenic model of Alzheimer's disease Brain Res 1128 40-9 Donoghue J P 2008 Bridging the brain to the world: a perspective on neural interface systems Neuron 60 511-21 Drake K L, Wise K D, Farraye J, Anderson D J and BeMent S L 1988 Performance of planar multisite microprobes in recording extracellular single-unit intracortical activity IEEE Trans Biomed Eng 35 719-32 Edell D J, Toi V V, McNeil V M and Clark L D 1992 Factors influencing the biocompatibility of insertable silicon microshafts in cerebral cortex IEEE Trans Biomed Eng 39 635-43 Elkabes S, DiCicco-Bloom E M and Black I B 1996 Brain microglia macrophages express neurotrophins that selectively reglulate microglial proliferaction and function. Journal of Neuroscience 16 2508-21 Engler A J, Sen S, Sweeney H L and Discher D E 2006 Matrix elasticity directs stem cell lineage specification Cell 126 677-89 Epilepsy Foundation 2011 Epilepsy and Seizure Statistics. Landover, MD Faulkner J R, Herrmann J E, Woo M J, Tansey K E, Doan N B and Sofroniew M V 2004 Reactive astrocytes protect tissue and preserve function after spinal cord injury J Neurosci 24 2143-55 Fawcett J W and Asher R A 1999 The glial scar and central nervous system repair Brain research bulletin 49 377-91 Fernandez L J, Altuna A, Tijero M, Gabriel G, Villa R, Rodriguez M J, Batlle M, Vilares R, Berganzo J and Blanco F J 2009 Study of functional viability of SU-8-based 179 microneedles for neural applications Journal of Micromechanics and Microengineering 19 025007 Ferree T C, Luu P, Russell G S and Tucker D M 2001 Scalp electrode impedance, infection risk, and EEG data quality Clin Neurophysiol 112 536-44 Fitts P M 1954 The information capacity of the human motor system in controlling the amplitude of movement J Exp Psychol 47 381-91 Flanagan L A, Ju Y E, Marg B, Osterfield M and Janmey P A 2002 Neurite branching on deformable substrates Neuroreport 13 2411-5 Frampton J P, Hynd M R, Shuler M L and Shain W 2010 Effects of Glial Cells on Electrode Impedance Recorded from Neural Prosthetic Devices In Vitro Annals of biomedical engineering Fujioka H and Akema T 2010 Lipopolysaccharide acutely inhibits proliferation of neural precursor cells in the dentate gyrus in adult rats Brain Res 1352C 35-42 Fujita T, Yoshimine T, Maruno M and Hayakawa T 1998 Cellular dynamics of macrophages and microglial cells in reaction to stab wounds in rat cerebral cortex Acta neurochirurgica 140 275-9 Georges P C, Miller W J, Meaney D F, Sawyer E S and Janmey P A 2006 Matrices with compliance comparable to that of brain tissue select neuronal over glial growth in mixed cortical cultures Biophysical journal 90 3012-8 Gilletti A and Muthuswamy J 2006 Brain micromotion around implants in the rodent somatosensory cortex J Neural Eng 3 189-95 Grill W M and Mortimer J T 1994 Electrical properties of implant encapsulation tissue Annals of biomedical engineering 22 23-33 Grill W M, Norman S E and Bellamkonda R V 2009 Implanted Neural Interfaces: Biochallenges and Engineered Solutions Annual review of biomedical engineering Hampton D W, Rhodes K E, Zhao C, Franklin R J and Fawcett J W 2004 The responses of oligodendrocyte precursor cells, astrocytes and microglia to a cortical stab injury, in the brain Neuroscience 127 813-20 Harris J, Capadona J, Miller R, Healy B, Shanmuganathan K, Rowan S, Weder C and Tyler D 2011a Mechanically Adaptive Intracortical Implants Improve the Proximity of Neuronal Cell Bodies Journal of Neural Engineering under review Harris J, Hess AE, Rowan S, Weder C, Zorman C, Tyler D and Capadona J 2011b In vivo deployment of mechanically adaptive nanocomposites for intracortical microelectrodes Journal of Neural Engineering 8 046010 He W and Bellamkonda R V 2007 Indwelling Neural Implants: Strategies for Contending with the In Vivo Environment ed W M Reichert Henle C, Raab M, Cordeiro J G, Doostkam S, Schulze-Bonhage A, Stieglitz T and Rickert J 2011 First long term in vivo study on subdurally implanted micro-ECoG electrodes, manufactured with a novel laser technology Biomed Microdevices 13 59-68 Henze D A, Borhegyi Z, Csicsvari J, Mamiya A, Harris K D and Buzsaki G 2000 Intracellular features predicted by extracellular recordings in the hippocampus in vivo J Neurophysiol 84 390-400 Herber D L, Mercer M, Roth L M, Symmonds K, Maloney J, Wilson N, Freeman M J, Morgan D and Gordon M N 2007 Microglial activation is required for A beta 180 clearance after intracranial injection of lipopolysaccharide in APP transgenic mice J Neuroimmune Pharm 2 222-31 Hess A, Capadona J, Shanmuganathan K, Hsu L, Rowan S, Weder C, Tyler D and Zorman C 2011 Development of a stimuli-responsive polymer nanocomposite toward biologically-optimized, MEMS-based neural probes. Journal of Micromechanics and Microengineering 21 054009 Hess A, Dunning J, Harris J, Capadona J R, Shanmuganathan K, Rowan S J, Weder C, Tyler D J and Zorman C A 2009 A bio-inspired, chemo-responsive polymer nanocomposite for mechanically dynamic microsystems. In: Solid-State Sensors, Actuators and Microsystems Conference, 2009. TRANSDUCERS 2009. International, Hillebrand A and Barnes G R 2002 A quantitative assessment of the sensitivity of whole- head MEG to activity in the adult human cortex Neuroimage 16 638-50 Hinkerohe D, Smikalla D, Schoebel A, Haghikia A, Zoidl G, Haase C G, Schlegel U and Faustmann P M 2010 Dexamethasone prevents LPS-induced microglial activation and astroglial impairment in an experimental bacterial meningitis co-culture model Brain Res 1329 45-54 Hochberg L R, Serruya M D, Friehs G M, Mukand J A, Saleh M, Caplan A H, Branner A, Chen D, Penn R D and Donoghue J P 2006 Neuronal ensemble control of prosthetic devices by a human with tetraplegia Nature 442 164-71 Horn K P, Busch S A, Hawthorne A L, van Rooijen N and Silver J 2008 Another Barrier to Regeneration in the CNS: Activated Macrophages Induce Extensive Retraction of Dystrophic Axons through Direct Physical Interactions. Huo X and Ghovanloo M 2010 Evaluation of a wireless wearable tongue-computer interface by individuals with high-level spinal cord injuries J Neural Eng 7 26008 Ingber D E 2006 Cellular mechanotransduction: putting all the pieces together again Faseb J 20 811-27 Jensen W, Yoshida K and Hofmann U G 2006 In-Vivo Implant Mechanics of Flexible, Silicon-Based ACREO Microelectrode Arrays in Rat Cerebral Cortex Biomedical Engineering, IEEE Transactions on 53 934-40 Johnson M, Otto K, Williams J and Kipke D 2004 Bias voltages at microelectrodes change neural interface properties in vivo Conf Proc IEEE Eng Med Biol Soc 6 4103-6 Jones K E, Campbell P K and Normann R A 1992 A glass/silicon composite intracortical electrode array Annals of biomedical engineering 20 423-37 Jones M S and Barth D S 1999 Spatiotemporal organization of fast (>200 Hz) electrical oscillations in rat Vibrissa/Barrel cortex J Neurophysiol 82 1599-609 Katzner S, Nauhaus I, Benucci A, Bonin V, Ringach D L and Carandini M 2009 Local origin of field potentials in visual cortex Neuron 61 35-41 Kee-Keun L and et al. 2004 Polyimide-based intracortical neural implant with improved structural stiffness Journal of Micromechanics and Microengineering 14 32 Kennedy P R, Bakay R A, Moore M M, Adams K and Goldwaithe J 2000 Direct control of a computer from the human central nervous system IEEE Trans Rehabil Eng 8 198-202 Kennedy P R, Kirby M T, Moore M M, King B and Mallory A 2004 Computer control using human intracortical local field potentials Ieee T Neur Sys Reh 12 339-44 181 Kennedy R 2010 Measuring Neurotransmitters in Vivo. In: Center for Neural Communication Technology 2010, Ann Arbor, MI: University of Michigan Kettenmann H 2007 Neuroscience: The brain's garbage men Nature 446 987-9 Khawaja F A, Tsui J M and Pack C C 2009 Pattern motion selectivity of spiking outputs and local field potentials in macaque visual cortex J Neurosci 29 13702-9 Kherlopian A R, Gerrein J P, Yue M, Kim K E, Kim J W, Sukumaran M and Sajda P 2006 Electrooculogram based system for computer control using a multiple feature classification model. In: Engineering in Medicine and Biology Society, 2006. EMBS '06. 28th Annual International Conference of the IEEE, Kigerl K A, Gensel J C, Ankeny D P, Alexander J K, Donnelly D J and Popovich P G 2009 Identification of two distinct macrophage subsets with divergent effects causing either neurotoxicity or regeneration in the injured mouse spinal cord J Neurosci 29 13435-44 Kim D H and Martin D C 2006 Sustained release of dexamethasone from hydrophilic matrices using PLGA nanoparticles for neural drug delivery Biomaterials 27 3031-7 Kim S P, Simeral J D, Hochberg L R, Donoghue J P and Black M J 2008 Neural control of computer cursor velocity by decoding motor cortical spiking activity in humans with tetraplegia J Neural Eng 5 455-76 Kim Y T, Hitchcock R W, Bridge M J and Tresco P A 2004 Chronic response of adult rat brain tissue to implants anchored to the skull Biomaterials 25 2229-37 Kipke D R 2004 Implantable neural probe systems for cortical neuroprostheses. In: Engineering in Medicine and Biology Society, 2004. EMBC 2004. Conference Proceedings. 26th Annual International Conference of the, Kipke D R, Shain W, Buzsaki G, Fetz E, Henderson J M, Hetke J F and Schalk G 2008 Advanced neurotechnologies for chronic neural interfaces: new horizons and clinical opportunities J Neurosci 28 11830-8 Kipke D R, Vetter R J, Williams J C and Hetke J F 2003 Silicon-substrate intracortical microelectrode arrays for long-term recording of neuronal spike activity in cerebral cortex IEEE Trans Neural Syst Rehabil Eng 11 151-5 Kipp M, Norkute A, Johann S, Lorenz L, Braun A, Hieble A, Gingele S, Pott F, Richter J and Beyer C 2008 Brain-region-specific astroglial responses in vitro after LPS exposure J Mol Neurosci 35 235-43 Kippert A, Fitzner D, Helenius J and Simons M 2009 Actomyosin contractility controls cell surface area of oligodendrocytes BMC cell biology 10 71 Koizumi S, Shigemoto-Mogami Y, Nasu-Tada K, Shinozaki Y, Ohsawa K, Tsuda M, Joshi B V, Jacobson K A, Kohsaka S and Inoue K 2007 UDP acting at P2Y6 receptors is a mediator of microglial phagocytosis Nature 446 1091-5 Kozai T D and Kipke D R 2009 Insertion shuttle with carboxyl terminated self-assembled monolayer coatings for implanting flexible polymer neural probes in the brain J Neurosci Methods 184 199-205 Kozai T D, Langhals N B, Patel P R, Deng X, Zhang H, Lahann J, Kotov N and Kipke D R 2010a In vivo chronic cortical recordings using novel ultra-small carbon fiber based implantable microthread ultramicroelectrodes In: Soceity for Neuroscience Annual Meeting, San Diego, CA 182 Kozai T D, Marzullo T C, Hooi F, Langhals N B, Majewska A K, Brown E B and Kipke D R 2010b Reduction of neurovascular damage resulting from microelectrode insertion into the cerebral cortex using in vivo two-photon mapping J Neural Eng 7 046011 Kreutzberg G W 1996 Microglia: a sensor for pathological events in the CNS Trends in neurosciences 19 312-8 Kronegg J, Voloshynovskiy S and Pun T 2005 Analysis of bit-rate definitions for brain computer interfaces. In: 2005 International Conference on Human-computer interaction, Leach J, Achyuta A K H and Murthy S K 2010 Bridging the divide between neuroprosthetic design, tissue engineering and neurobiology Frontiers in Neuroengineering 2 1-19 Leach J B, Brown X Q, Jacot J G, Dimilla P A and Wong J Y 2007 Neurite outgrowth and branching of PC12 cells on very soft substrates sharply decreases below a threshold of substrate rigidity J Neural Eng 4 26-34 Lee H, Bellamkonda R V, Sun W and Levenston M E 2005 Biomechanical analysis of silicon microelectrode-induced strain in the brain J Neural Eng 2 81-9 Lee K, He J, Clement R, Massia S and Kim B 2004 Biocompatible benzocyclobutene (BCB)-based neural implants with micro-fluidic channel Biosens Bioelectron 20 404-7 Lee K-K, He J, Singh A and Kim B 2003 Benzocyclobutene (BCB) based intracortical neural implant MEMS, NANO and Smart Systems, 2003. Proceedings. International Conference on 418- 22 Lempka S F, Johnson M D, Barnett D W, Moffitt M A, Otto K J, Kipke D R and McIntyre C C 2006 Optimization of Microelectrode Design for Cortical Recording Based on Thermal Noise Considerations. In: Engineering in Medicine and Biology Society, 2006. EMBS '06. 28th Annual International Conference of the IEEE, Lempka S F, Johnson M D, Moffitt M A, Otto K J, Kipke D R and McIntyre C C 2011 Theoretical analysis of intracortical microelectrode recordings J Neural Eng 8 045006 Lenzlinger P, Morganti-Kossmann M-C, Laurer H and McIntosh T 2001 The duality of the inflammatory response to traumatic brain injury Molecular Neurobiology 24 169-81 Levin E C, Acharya N K, Sedeyn J C, Venkataraman V, D'Andrea M R, Wang H Y and Nagele R G 2009 Neuronal expression of vimentin in the Alzheimer's disease brain may be part of a generalized dendritic damage-response mechanism Brain Res 1298 194-207 Li G N, Livi L L, Gourd C M, Deweerd E S and Hoffman-Kim D 2007 Genomic and morphological changes of neuroblastoma cells in response to three-dimensional matrices Tissue engineering 13 1035-47 Liang J, Takeuchi H, Doi Y, Kawanokuchi J, Sonobe Y, Jin S, Yawata I, Li H, Yasuoka S, Mizuno T and Suzumura A 2008 Excitatory amino acid transporter expression by astrocytes is neuroprotective against microglial excitotoxicity Brain Res 1210 11-9 183 Liang J, Takeuchi H, Jin S, Noda M, Li H, Doi Y, Kawanokuchi J, Sonobe Y, Mizuno T and Suzumura A 2010 Glutamate induces neurotrophic factor production from microglia via protein kinase C pathway Brain Res 1322 8-23 Lillie R D 1965 Histopathologic Technic and Practical Histochemistry (New York, NY: McGraw-Hill Book Co.) Llinas R R, Walton K D, Nakao M, Hunter I and Anquetil P A 2005 Neuro-vascular central nervous recording/stimulating system: Using nanotechnology probes Journal of Nanoparticle Research 7 111-27 Lu P, Jones L L and Tuszynski M H 2007 Axon regeneration through scars and into sites of chronic spinal cord injury Exp Neurol 203 8-21 Lu Y, Wang D, Li T, Zhao X, Cao Y, Yang H and Duan Y Y 2009 Poly(vinyl alcohol)/poly(acrylic acid) hydrogel coatings for improving electrode-neural tissue interface Biomaterials 30 4143-51 Ludwig K A, Miriani R M, Langhals N B, Joseph M D, Anderson D J and Kipke D R 2009 Using a Common Average Reference to Improve Cortical Neuron Recordings From Microelectrode Arrays J Neurophysiol 101 1679-89 Ludwig K A, Uram J D, Yang J, Martin D C and Kipke D R 2006 Chronic neural recordings using silicon microelectrode arrays electrochemically deposited with a poly(3,4-ethylenedioxythiophene) (PEDOT) film J Neural Eng 3 59-70 Martel S, Hatsapoulos N, Hunter I, Donoghue J, Burgert J, Malasek J, Wiseman C and Dyer R 2001 Development of a wireless brain implant: the Telemetric Electrode Array System (TEAS) project. In: Engineering in Medicine and Biology Society, 2001. Proceedings of the 23rd Annual International Conference of the IEEE, Marzullo T 2008 On the Relationship Between Impedance, Site Size and Unit Activity Report Briefs Neuronexus Technologies Mateo J C, Agustin J S and Hansen J P 2008 Gaze beats mouse: hands-free selection by combining gaze and emg. In: CHI '08 extended abstracts on Human factors in computing systems, Florence, Italy: ACM Mayhan W G 1998 Effect of lipopolysaccharide on the permeability and reactivity of the cerebral microcirculation: role of inducible nitric oxide synthase Brain Res 792 353-7 McConnell G C 2008 Chronic inflammation surrounding intra-cortical electrodes is correlated with a local neurodegenerative state. In: Biomedical Engineering, Atlanta, GA: Georgia Institute of Technology McConnell G C, Rees H D, Levey A I, Gutekunst C A, Gross R E and Bellamkonda R V 2009 Implanted neural electrodes cause chronic, local inflammation that is correlated with local neurodegeneration J Neural Eng 6 56003 McConnell G C, Schneider T M and Bellamkonda R V 2007a Acute spatiotemporal changes in neuronal density surrounding microelectrode arrays implanted in rat motor cortex. In: Neural Engineering, 2007. CNE '07. 3rd International IEEE/EMBS Conference on, McConnell G C, Schneider T M, Owens D J and Bellamkonda R V 2007b Extraction force and cortical tissue reaction of silicon microelectrode arrays implanted in the rat brain IEEE Trans Biomed Eng 54 1097-107 184 McKeon R J, Jurynec M J and Buck C R 1999 The chondroitin sulfate proteoglycans neurocan and phosphacan are expressed by reactive astrocytes in the chronic CNS glial scar J Neurosci 19 10778-88 Mercanzini A, Colin P, Bensadoun J C, Bertsch A and Renaud P 2009 In vivo electrical impedance spectroscopy of tissue reaction to microelectrode arrays IEEE Trans Biomed Eng 56 1909-18 Mitzdorf U 1985 Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena Physiol Rev 65 37-100 Moffitt M A and McIntyre C C 2005 Model-based analysis of cortical recording with silicon microelectrodes Clin Neurophysiol 116 2240-50 Mohseni P, Najafi K, Eliades S J and Xiaoqin W 2005 Wireless multichannel biopotential recording using an integrated FM telemetry circuit Neural Systems and Rehabilitation Engineering, IEEE Transactions on [see also IEEE Trans. on Rehabilitation Engineering] 13 263-71 Moxon K A, Leiser S C, Gerhardt G A, Barbee K A and Chapin J K 2004 Ceramic-based multisite electrode arrays for chronic single-neuron recording IEEE Trans Biomed Eng 51 647-56 Munoz-Fernandez M A and Fresno M 1998 The role of tumor necrosis factor, interlukin- 6, interfuron-gamma, and inducible nitric oxide synthase in the development and pathology of the nervous system Prog. Neurobiol 56 307-40 Najafi K and Hetke J F 1990 Strength characterization of silicon microprobes in neurophysiological tissues IEEE Trans Biomed Eng 37 474-81 Nakajima K, Honda S, Tohyama Y, Imai Y, Kohsaka S and Kurihara T 2001 Neurotrophin Secretion From Cultured Microglia Journal of Neuroscience Research 65 322-31 National Spinal Cord Injury Statistical Center 2011 Spinal Cord Injury Facts and Figures at a Glance. Birmingham, Alabama Nicolelis M A 2003 Brain-machine interfaces to restore motor function and probe neural circuits Nature reviews 4 417-22 Nicolelis M A, Dimitrov D, Carmena J M, Crist R, Lehew G, Kralik J D and Wise S P 2003 Chronic, multisite, multielectrode recordings in macaque monkeys Proc Natl Acad Sci U S A 100 11041-6 Nikles S A, Pellinen D S, Kitagawa J, Bradley R M, Kipke D R and Najafi K 2003 Long term in vitro monitoring of polyimide microprobe electrical properties. In: Engineering in Medicine and Biology Society, 2003. Proceedings of the 25th Annual International Conference of the IEEE, Nisbet D R, Rodda A E, Horne M K, Forsythe J S and Finkelstein D I 2009 Neurite infiltration and cellular response to electrospun polycaprolactone scaffolds implanted into the brain Biomaterials Nishiyama A 2007 Polydendrocytes: NG2 cells with many roles in development and repair of the CNS Neuroscientist 13 62-76 Ortinski P I, Dong J, Mungenast A, Yue C, Takano H, Watson D J, Haydon P G and Coulter D A 2010 Selective induction of astrocytic gliosis generates deficits in neuronal inhibition Nat Neurosci 13 584-91 Ostrow L W and Sachs F 2005 Mechanosensation and endothelin in astrocytes-- hypothetical roles in CNS pathophysiology Brain Res Brain Res Rev 48 488-508 185 Otto K J, Johnson M D and Kipke D R 2006 Voltage pulses change neural interface properties and improve unit recordings with chronically implanted microelectrodes IEEE Trans Biomed Eng 53 333-40 Paralikar K J, Lawrence J K and Clement R S 2006 Collagenase-aided Insertion of Intracortical Microelectrode Arrays: Evaluation of Insertion Force and Chronic Recording Performance. In: Engineering in Medicine and Biology Society, 2006. EMBS '06. 28th Annual International Conference of the IEEE, Parker-Ure R A, Christensen M B, House P, Tresco P A and Greger B 2008 Micro- electrode functionality inversely correlated with markers of inflammation and not with markers of neuronal cell viability In: Soceity for Neuroscience Annual Meeting, Washington, DC Pekny M 2003 Glial Cell Function, ed B C Lopez and M Nieto-Sampedro (Amsterdam: Elsevier) Peyton S R, Ghajar C M, Khatiwala C B and Putnam A J 2007 The emergence of ECM mechanics and cytoskeletal tension as important regulators of cell function Cell biochemistry and biophysics 47 300-20 Pinheiro J, Bates D, DebRoy S, Sarkar D and the R Core team 2009 nlme: Linear and Nonlinear Mixed Effects Models. Polikov V S, Block M L, Fellous J M, Hong J S and Reichert W M 2006 In vitro model of glial scarring around neuroelectrodes chronically implanted in the CNS Biomaterials 27 5368-76 Polikov V S, Tresco P A and Reichert W M 2005 Response of brain tissue to chronically implanted neural electrodes J Neurosci Methods 148 1-18 Purcell E K, Thompson D E, Ludwig K A and Kipke D R 2009 Flavopiridol reduces the impedance of neural prostheses in vivo without affecting recording quality J Neurosci Methods Qin L, Wu X, Block M L, Liu Y, Breese G R, Hong J S, Knapp D J and Crews F T 2007 Systemic LPS causes chronic neuroinflammation and progressive neurodegeneration Glia 55 453-62 Quiroga R Q, Nadasdy Z and Ben-Shaul Y 2004 Unsupervised spike detection and sorting with wavelets and superparamagnetic clustering Neural Comput 16 1661- 87 R Development Core Team 2009 R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing Raghavendra Rao V L, Rao A M, Dogan A, Bowen K K, Hatcher J, Rothstein J D and Dempsey R J 2000 Glial glutamate transporter GLT-1 down-regulation precedes delayed neuronal death in gerbil hippocampus following transient global cerebral ischemia Neurochemistry international 36 531-7 Renault-Mihara F, Okada S, Shibata S, Nakamura M, Toyama Y and Okano H 2008 Spinal cord injury: emerging beneficial role of reactive astrocytes' migration The international journal of biochemistry & cell biology 40 1649-53 Rennaker R L, Miller J, Tang H and Wilson D A 2007 Minocycline increases quality and longevity of chronic neural recordings J Neural Eng 4 L1-5 Rennaker R L, Ruyle A M, Street S E and Sloan A M 2005a An economical multi- channel cortical electrode array for extended periods of recording during behavior J Neurosci Methods 142 97-105 186 Rennaker R L, Street S, Ruyle A M and Sloan A M 2005b A comparison of chronic multi-channel cortical implantation techniques: manual versus mechanical insertion J Neurosci Methods 142 169-76 Richardson-Burns S M, Hendricks J L and Martin D C 2007 Electrochemical polymerization of conducting polymers in living neural tissue J Neural Eng 4 L6- L13 Rivest S 2009 Regulation of innate immune responses in the brain Nat Rev Immunol 9 429-39 Roger V L, Go A S, Lloyd-Jones D M, Adams R J, Berry J D, Brown T M, Carnethon M R, Dai S, de Simone G, Ford E S, Fox C S, Fullerton H J, Gillespie C, Greenlund K J, Hailpern S M, Heit J A, Ho P M, Howard V J, Kissela B M, Kittner S J, Lackland D T, Lichtman J H, Lisabeth L D, Makuc D M, Marcus G M, Marelli A, Matchar D B, McDermott M M, Meigs J B, Moy C S, Mozaffarian D, Mussolino M E, Nichol G, Paynter N P, Rosamond W D, Sorlie P D, Stafford R S, Turan T N, Turner M B, Wong N D and Wylie-Rosett J 2011 Heart Disease and Stroke Statistics--2011 Update: A Report From the American Heart Association Circulation 123 e18-209 Roitbak T and Sykova E 1999 Diffusion barriers evoked in the rat cortex by reactive astrogliosis Glia 28 40-8 Rousche P J, Pellinen D S, Pivin D P, Williams J C, Vetter R J and Kipke D R 2001 Flexible polyimide-based intracortical electrode arrays with bioactive capability IEEE Trans Biomed Eng 48 361-71 Ryu S I and Shenoy K V 2009 Human cortical prostheses: lost in translation? Neurosurgical focus 27 E5 Salinska E, Danysz W and Lazarewicz J W 2005 The role of excitotoxicity in neurodegeneration Folia neuropathologica / Association of Polish Neuropathologists and Medical Research Centre, Polish Academy of Sciences 43 322-39 Sanchez J C, Alba N, Nishida T, Batich C and Carney P R 2006 Structural modifications in chronic microwire electrodes for cortical neuroprosthetics: a case study IEEE Trans Neural Syst Rehabil Eng 14 217-21 Scherberger H, Jarvis M R and Andersen R A 2005 Cortical local field potential encodes movement intentions in the posterior parietal cortex Neuron 46 347-54 Schmidt E M, McIntosh J S and Bak M J 1988 Long-term implants of Parylene-C coated microelectrodes Medical & biological engineering & computing 26 96-101 Schwartz A B 2004 Cortical neural prosthetics Annu Rev Neurosci 27 487-507 Seymour J P and Kipke D R 2007 Neural probe design for reduced tissue encapsulation in CNS Biomaterials 28 3594-607 Shain W, Spataro L, Dilgen J, Haverstick K, Retterer S, Isaacson M, Saltzman M and Turner J N 2003 Controlling cellular reactive responses around neural prosthetic devices using peripheral and local intervention strategies IEEE Trans Neural Syst Rehabil Eng 11 186-8 Shanmuganathan K 2010 Bio-Inspired Stimuli Responsive Mechanically Dynamic Nanocomposites. In: Macromolecular Science and Engineering, Cleveland, OH: Case Western Reserve University 187 Shanmuganathan K, Capadona J R, Rowan S J and Weder C 2009 Stimuli-Responsive Mechanically Adaptive Polymer Nanocomposites ACS Applied Materials & Interfaces 2 165-74 Shanmuganathan K, Capadona J R, Rowan S J and Weder C 2010a Bio-inspired mechanically-adaptive nanocomposites derived from cotton cellulose whiskers Journal of Materials Chemistry 20 180 Shanmuganathan K, Capadona J R, Rowan S J and Weder C 2010b Biomimetic mechanically adaptive nanocomposites Progress in Polymer Science 35 212-22 Shearer M C, Niclou S P, Brown D, Asher R A, Holtmaat A J, Levine J M, Verhaagen J and Fawcett J W 2003 The astrocyte/meningeal cell interface is a barrier to neurite outgrowth which can be overcome by manipulation of inhibitory molecules or axonal signalling pathways Molecular and cellular neurosciences 24 913-25 Silver J and Miller J H 2004 Regeneration beyond the glial scar Nature reviews 5 146-56 Simard M, Arcuino G, Takano T, Liu Q S and Nedergaard M 2003 Signaling at the gliovascular interface J Neurosci 23 9254-62 Simeral J, Hochberg L, Donoghue J, Friehs G and Black M 2011a Point-and-Click Cursor Control With an Intracortical Neural Interface System in Humans With Tetraplegia IEEE Trans Neural Syst Rehabil Eng Simeral J D, Kim S P, Black M J, Donoghue J P and Hochberg L R 2011b Neural control of cursor trajectory and click by a human with tetraplegia 1000 days after implant of an intracortical microelectrode array J Neural Eng 8 025027 Simpson T, Broughton C, Gauthier M J and Prochazka A 2008 Tooth-click control of a hands-free computer interface IEEE Trans Biomed Eng 55 2050-6 Simpson T, Gauthier M and Prochazka A 2010 Evaluation of Tooth-Click Triggering and Speech Recognition in Assistive Technology for Computer Access Neurorehabilitation and Neural Repair 24 188-94 Skousen J L, Winslow B D, Merriam M E, Srivannavit O, Perline G, Wise K D and Tresco P A 2010 Electrode constitutive properties modulate the brain tissue foreign body response In: Soceity for Neuroscience Annual Meeting, San Diego, CA Sofroniew M V 2005 Reactive astrocytes in neural repair and protection Neuroscientist 11 400-7 Spataro L, Dilgen J, Retterer S, Spence A J, Isaacson M, Turner J N and Shain W 2005 Dexamethasone treatment reduces astroglia responses to inserted neuroprosthetic devices in rat neocortex Exp Neurol 194 289-300 Subbaroyan J 2007 Investigations of tethering induced injury response in brain tissue by intracortical implants through modeling and in vivo experiments. In: Biomedical Engineering, Ann Arbor, MI: University of Michigan Subbaroyan J and Kipke D R 2006 The role of flexible polymer interconnects in chronic tissue response induced by intracortical microelectrodes--a modeling and an in vivo study Conf Proc IEEE Eng Med Biol Soc 1 3588-91 Subbaroyan J, Martin D C and Kipke D R 2005 A finite-element model of the mechanical effects of implantable microelectrodes in the cerebral cortex J Neural Eng 2 103- 13 188 Suner S, Fellows M R, Vargas-Irwin C, Nakata G K A N G K and Donoghue J P A D J P 2005 Reliability of signals from a chronically implanted, silicon-based electrode array in non-human primate primary motor cortex Neural Systems and Rehabilitation Engineering, IEEE Transactions on [see also IEEE Trans. on Rehabilitation Engineering] 13 524-41 Surakka V, Illi M and Isokoski P 2004 Gazing and frowning as a new human--computer interaction technique ACM Trans. Appl. Percept. 1 40-56 Szarowski D H, Andersen M D, Retterer S, Spence A J, Isaacson M, Craighead H G, Turner J N and Shain W 2003 Brain responses to micro-machined silicon devices Brain Res 983 23-35 Takeuchi S, Suzuki T, Mabuchi K and Fujita H 2004 3D flexible multichannel neural probe array Journal of Micromechanics and Microengineering 14 104-7 Takeuchi S, Ziegler D, Yoshida Y, Mabuchi K and Suzuki T 2005 Parylene flexible neural probes integrated with microfluidic channels Lab Chip 5 519-23 Tanaka R, Komine-Kobayashi M, Mochizuki H, Yamada M, Furuya T, Migita M, Shimada T, Mizuno Y and Urabe T 2003 Migration of enhanced green fluorescent protein expressing bone marrow-derived microglia/macrophage into the mouse brain following permanent focal ischemia Neuroscience 117 531-9 Taylor D M, Tillery S I H and Schwartz A B 2002 Direct Cortical Control of 3D Neuroprosthetic Devices. Taylor D M, Tillery S I H and Schwartz A B 2003 Information conveyed through brain- control: cursor versus robot Neural Systems and Rehabilitation Engineering, IEEE Transactions on 11 195-9 Tresco P A and Winslow B D 2011 The challenge of integrating devices into the central nervous system Crit Rev Biomed Eng 39 29-44 Tsai P S, Kaufhold J P, Blinder P, Friedman B, Drew P J, Karten H J, Lyden P D and Kleinfeld D 2009 Correlations of neuronal and microvascular densities in murine cortex revealed by direct counting and colocalization of nuclei and vessels J Neurosci 29 14553-70 Turner J N, Shain W, Szarowski D H, Andersen M, Martins S, Isaacson M and Craighead H 1999 Cerebral astrocyte response to micromachined silicon implants Exp Neurol 156 33-49 van den Berg O, Capadona J R and Weder C 2007 Preparation of homogeneous dispersions of tunicate cellulose whiskers in organic solvents Biomacromolecules 8 1353-7 Velliste M, Perel S, Spalding M C, Whitford A S and Schwartz A B 2008 Cortical control of a prosthetic arm for self-feeding Nature 453 1098-101 Vetter R J, Williams J C, Hetke J F, Nunamaker E A and Kipke D R 2004 Chronic neural recording using silicon-substrate microelectrode arrays implanted in cerebral cortex IEEE Trans Biomed Eng 51 896-904 Wadhwa R, Lagenaur C F and Cui X T 2006 Electrochemically controlled release of dexamethasone from conducting polymer polypyrrole coated electrode J Control Release 110 531-41 Wanner I B, Deik A, Torres M, Rosendahl A, Neary J T, Lemmon V P and Bixby J L 2008 A new in vitro model of the glial scar inhibits axon growth Glia 56 1691- 709 189 Ward M P, Rajdev P, Ellison C and Irazoqui P P 2009 Toward a comparison of microelectrodes for acute and chronic recordings Brain Res 1282 183-200 Watson R E, Jr., Wiegand S J, Clough R W and Hoffman G E 1986 Use of cryoprotectant to maintain long-term peptide immunoreactivity and tissue morphology Peptides 7 155-9 Weberpals M, Hermes M, Hermann S, Kummer M P, Terwel D, Semmler A, Berger M, Schafers M and Heneka M T 2009 NOS2 gene deficiency protects from sepsis- induced long-term cognitive deficits J Neurosci 29 14177-84 Wester B A, Lee R H and LaPlaca M C 2009 Development and characterization of in vivo flexible electrodes compatible with large tissue displacements J Neural Eng 6 024002 Williams J C, Hippensteel J A, Dilgen J, Shain W and Kipke D R 2007 Complex impedance spectroscopy for monitoring tissue responses to inserted neural implants J Neural Eng 4 410-23 Williams J C, Rennaker R L and Kipke D R 1999 Long-term neural recording characteristics of wire microelectrode arrays implanted in cerebral cortex Brain Res Protoc 4 303-13 Williams M R and Kirsch R F 2008 Evaluation of head orientation and neck muscle EMG signals as command inputs to a human-computer interface for individuals with high tetraplegia IEEE Trans Neural Syst Rehabil Eng 16 485-96 Winslow B D, Christensen M B, Yang W K, Solzbacher F and Tresco P A 2010 A comparison of the tissue response to chronically implanted Parylene-C-coated and uncoated planar silicon microelectrode arrays in rat cortex Biomaterials Winslow B D and Tresco P A 2009 Quantitative analysis of the tissue response to chronically implanted microwire electrodes in rat cortex Biomaterials Wise K D, Anderson D J, Hetke J F, Kipke D R and Najafi K 2004 Wireless implantable microsystems: high-density electronic interfaces to the nervous system Proceedings of the IEEE 92 76-97 Wolpaw J R, Birbaumer N, McFarland D J, Pfurtscheller G and Vaughan T M 2002 Brain-computer interfaces for communication and control Clin Neurophysiol 113 767-91 Wolpaw J R, Ramoser H, McFarland D J and Pfurtscheller G 1998 EEG-based communication: improved accuracy by response verification IEEE Trans Rehabil Eng 6 326-33 Xing D, Yeh C I and Shapley R M 2009 Spatial spread of the local field potential and its laminar variation in visual cortex J Neurosci 29 11540-9 Yamato H, Ohwa M and Wernet W 1995 Stability of polypyrrole and poly(3,4- ethylenedioxythiophene) for biosensor application Journal of Electroanalytical Chemistry 397 163-70 Yin J, Sakamoto K, Zhang H, Ito Z, Imagama S, Kishida S, Natori T, Sawada M, Matsuyama Y and Kadomatsu K 2009 Transforming growth factor-beta1 upregulates keratan sulfate and chondroitin sulfate biosynthesis in microglias after brain injury Brain Res 1263 10-22 Zaman M H, Trapani L M, Sieminski A L, Mackellar D, Gong H, Kamm R D, Wells A, Lauffenburger D A and Matsudaira P 2006 Migration of tumor cells in 3D 190 matrices is governed by matrix stiffness along with cell-matrix adhesion and proteolysis Proc Natl Acad Sci U S A 103 10889-94 Zhong Y and Bellamkonda R V 2007 Dexamethasone-coated neural probes elicit attenuated inflammatory response and neuronal loss compared to uncoated neural probes Brain Res 1148 15-27 Zhong Y, McConnell G C, Ross J D, DeWeerth S P and Bellamkonda R V 2005 A Novel Dexamethasone-releasing, Anti-inflammatory Coating for Neural Implants. In: Neural Engineering, 2005. Conference Proceedings. 2nd International IEEE EMBS Conference on, 191