The Role of Innate Immunity in the Response to Intracortical

Home , Nature Materials

THE ROLE OF INNATE IMMUNITY IN THE RESPONSE

TO INTRACORTICAL MICROELECTRODES

JOHN KARL HERMANN

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Dissertation Advisor: Dr. Jeffrey R. Capadona

Department of Biomedical Engineering CASE WESTERN RESERVE UNIVERSITY

August 2018

CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of John Hermann, candidate for the degree of Doctor of Philosophy*.

Jeffrey R. Capadona Research Advisor

Abidemi B. Ajiboye Committee Chair

Dawn M. Taylor Committee Member

Dominique M. Durand Committee Member

Nicholas P. Ziats Committee Member

4/25/2018 Date of Defense

*We also certify that written approval has been obtained for any proprietary material contained therein.

Table of Contents List of Tables ...... vi

List of Figures ...... viii

Acknowledgements ...... xi

List of Abbreviations ...... xiv

Abstract ...... 1

Chapter 1 Specific Aims ...... 3

Aim 1: Investigate the effect of genetic and therapeutic inhibition of CD14 on

intracortical microelectrode performance and tissue integration...... 5

Aim 2: Characterize the role of TLR2 and TLR4 in the neuroinflammatory response to

intracortical microelectrodes...... 6

Chapter 2 Introduction ...... 8

2.1 Applications of Neural Interfacing ...... 8

2.2 Neural Interfacing Methods ...... 10

2.3 Intracortical microelectrodes ...... 13

2.4 Failure of intracortical microelectrodes ...... 18

2.5 Foreign body response to intracortical microelectrodes ...... 23

2.6 Interventions to mitigate biological failures ...... 27

2.7 Studies linking neuroinflammation to electrode failure ...... 44

2.8 Innate Immunity ...... 50

2.9 Toll-like Receptors and CD14 ...... 53

2.10 Toll-like Receptors and CD14 in Neurodegenerative disorders ...... 64

2.11 TLRs/CD14 in the foreign body response to intracortical microelectrodes ...... 67

2.12 How to inhibit TLR2, TLR4, and CD14 ...... 72

Chapter 3 Inhibition of the cluster of differentiation 14 innate immunity pathway with

IAXO-101 improves chronic microelectrode performance*# ...... 79

3.1 Abstract ...... 79

3.2 Introduction ...... 80

3.3 Methods ...... 85

3.4 Results ...... 95

3.5 Discussion ...... 112

3.6 Conclusions ...... 123

3.7 Acknowledgements ...... 124

Chapter 4 The role of toll-like receptor 2 and 4 innate immunity pathways in intracortical microelectrode induced neuroinflammation...... 125

4.1 Abstract ...... 125

4.2 Introduction ...... 126

4.3 Materials and methods ...... 129

4.4 Results ...... 136

4.5 Discussion ...... 150

iii

4.6 Conclusions ...... 158

4.7 Acknowledgements ...... 159

4.8 Funding...... 159

Chapter 5 Conclusions and Future Directions ...... 160

Appendix ...... 168

Supporting Author Papers ...... 169

Supporting Author Paper 1 Targeting CD14 on blood derived cells improves intracortical microelectrode performance* ...... 170

6.1 Abstract ...... 170

6.2 Introduction ...... 171

6.3 Results ...... 174

6.4 Discussion ...... 182

6.5 Conclusion ...... 189

6.6 Methods ...... 189

6.7 Acknowledgements ...... 199

Supporting Author Paper 2 Implantation of Neural Probes in the Brain Elicits Oxidative

Stress* ...... 200

7.1 Abstract ...... 200

7.2 Introduction ...... 201

7.3 Materials and Methods ...... 205

7.4 Results ...... 212

7.5 Discussion ...... 219

7.6 Conclusion ...... 223

7.7 Conflict of Interest Statement ...... 224

7.8 Authors and Contributors ...... 224

7.9 Funding...... 225

7.10 Acknowledgements ...... 225

Supplemental Information ...... 226

8.1 Supplemental information from Chapter 3 ...... 226

8.2 Supplemental information for Supporting Author Paper 1 ...... 230

Bibliography ...... 236

List of Tables

Table 1. Major types of interfaces with the brain with varying degrees of invasiveness. 12

Table 2. Summary of major pathogen associated molecular patterns (PAMPs) recognized

by toll-like receptors (TLRs)...... 56

Table 3. Summary of major damage associated molecular patterns (DAMPs) recognized

by toll-like receptors (TLRs)...... 58

Table 4. Strategies for the inhibition of TLR2, TLR4, and CD14...... 77

Table 5. Summary of immunohistochemistry targets, antigens, antibodies, and

concentrations. Table updated from erratum [444]...... 93

Table 6. Statistical summary of the recording performance metrics compared between

Cd14−/− and WT mice...... 96

Table 7. Statistical summary of the recording performance metrics compared between

IAXO-101 and WT mice...... 102

Table 8. Summary of immunohistochemical reagents used in histology...... 133

Table 9. Statistical summary for the recording performance of laminar, silicon IME

comparing all experimental groups (top) and WT versus BdCd14-/- (bottom)...... 177

Table 10. Histological markers for oxidative stress...... 211

Table 11. Oxidative stress relative gene expression...... 214

Table 12. Daily sample size details for Cd14-/- mice, wildtype mice, and mice administered

IAXO-101 used to quantify the metrics Units per Channel, % Channels Detecting Single

Units, and Noise...... 229

Table 13. Daily sample size details for Cd14-/- mice, wildtype mice, and mice administered

IAXO-101 used to quantify the metrics Amplitude and SNR...... 230

Table 14. Complete blood count (CBC) analysis on whole blood samples from WT, Cd14-

/-, BdCd14-/- chimera, MgCd14-/- at two weeks post implantation...... 234

Table 15. Primary antibodies used in immunohistochemistry to assess inflammation. . 234

vii

List of Figures

Figure 1. “Schematic representation of the generation design of leading microelectrode array technologies, including (A) microwires, (B) Michigan-style microelectrodes, [and]

Figure 2. “Major failure modes of MEAs.” ...... 19

Figure 3. “Electrode implantation results in localized pro-inflammatory cellular and biochemical events.” “ ...... 22

Figure 4. “Serum protein recognition propagates inflammatory response to intracortical microelectrodes.” ...... 25

Figure 5. “Oxidative stress following neural probe implantation.” ...... 26

Figure 6. Summary of findings from Kozai et al...... 29

Figure 7. Summary of findings from Patel et al...... 30

Figure 8. Summary of findings from Luan et al. “ ...... 32

Figure 9. Major findings from Nguyen et al...... 34

Figure 10. Major findings of Simon et al. “ ...... 36

Figure 11. “Sinusoidal probe microfabrication and surgery preparation.” ...... 37

Figure 12. Major findings from Du et al...... 39

Figure 13. Major findings from Eles et al...... 40

Figure 14. Major findings from Cody et al...... 40

Figure 15. Major findings of Oakes et al...... 41

Figure 16. Major findings from Potter et al...... 44

Figure 17. Major findings from Rennaker et al...... 46

Figure 18. Major findings from Harris et al...... 47

viii

Figure 19. “Schematic illustration and the working model of chronic electrode failure.”

...... 48

Figure 20. Major findings of Kozai et al...... 49

Figure 21. Cellular localization of innate immunity of innate immunity receptors...... 52

Figure 22. Potential role of innate immunity pattern recognition receptors in the foreign body response to intracortical microelectrodes...... 53

Figure 23. TLR signal transduction pathways...... 60

Figure 24. Potential role of TLR2, TLR4, and CD14 in the foreign body response to intracortical microelectrodes...... 71

Figure 25. Recording performance of intracortical microelectrodes in Cd14−/− mice versus wildtype mice...... 97

Figure 26. Recording performance of intracortical microelectrodes in wildtype mice treated with IAXO-101 versus untreated wildtype mice...... 103

Figure 27. Immunohistochemical evaluation of Cd14−/− mice implanted with intracortical microelectrodes...... 109

Figure 28. Immunohistochemical evaluation of mice administered IAXO-101 and implanted with intracortical microelectrodes...... 111

Figure 29. Immunohistochemical staining in sham animals...... 134

-/- -/- Figure 30. Acute immunohistochemical evaluation of Tlr2 , Tlr4 , and wildtype mice two weeks after probe implantation...... 139

-/- -/- Figure 31. Chronic immunohistochemical evaluation of Tlr2 , Tlr4 , and wildtype mice sixteen weeks after probe implantation...... 142

Figure 32. Changes in immunohistochemical markers in wildtype mice over time...... 145

-/- Figure 33. Changes in immunohistochemical markers in Tlr2 mice over time...... 147

-/- Figure 34. Changes in immunohistochemical markers in Tlr4 mice over time...... 149

Figure 35. Recording performance for all four animal models...... 176

Figure 36. Recording performance for removing BdCD14 versus WT...... 177

Figure 37. Immunohistochemical evaluation of inflammatory activated microglia and

macrophages...... 178

Figure 38. Immunohistochemical evaluation of blood brain barrier permeability...... 179

Figure 39. Immunohistochemical evaluation of astrocyte encapsulation...... 180

Figure 40. Immunohistochemical evaluation of neuronal density...... 181

Figure 41. Representative SEM images of post-explant and non-implanted laminar, silicon

IMEs...... 188

Figure 42. Sex as a Biological Variable...... 190

Figure 43. Oxidative stress following neural probe implantation...... 202

Figure 44. Oxidative stress relative gene expression...... 217

Figure 45. Oxidative stress histological markers...... 218

Figure 46. Representative electrophysiological recording...... 231

Figure 47. Recording performance for all four conditions (continued)...... 231

Figure 48. Schematic of creation of bone marrow chimeras...... 232

Figure 49. Representative H&E stain of motor cortex about ~640 µm deep from surface of

brain...... 234

Acknowledgements

I would like to thank all those that have helped to make the completion of my

doctorate degree possible. First and foremost, I would like to thank my research and academic advisor, Dr. Jeff Capadona. Your guidance, support, and mentorship have helped me grow as a researcher. Your motivation has picked me up through the various challenges we faced throughout the course of my degree. You believed in me through all the times where my belief in myself faltered.

Next, I would like to thank my committee, consisting of Dr. Ajiboye, Dr. Taylor,

Dr. Durand, and Dr. Ziats, for challenging me to be a better scientist and engineer. I

appreciate the time you put forth to guide my dissertation. I would like to thank Dr. Taylor

for the numerous hours you spent teaching me electrophysiology, including many evenings and weekends. I would also like to thank Dr. Landreth for his participation in my committee, earlier along in my PhD program.

Additionally, I would like to thank the graduate students in the Capadona lab that have trained me, learned with me, or accepted my mentorship during my time in the lab.

Your intellectual and moral support have been a large factor in my success. These students and alumni include Dr. Kelsey Potter-Baker, Dr. Madhu Ravikumar, Dr. Jessica Nguyen,

Jen Keene, Hillary Bedell, Griffin Rial, Sydney Song, and Youjong Kim.

Further, I would like to thank the post-docs and investigators in the lab, including

Dr. John Skousen, Dr. Evon Ereifej, and Dr. Andrew Shoffstall, who shared their experience and knowledge that was beneficial to my progress and development as a researcher.

Also, thank you to all the lab technicians and lab managers in the Capadona lab, including Kyle Kovach and Monika Goss. Your work keeping the lab in order and assisting on projects is greatly appreciated. I would also like to thank Bill Marcus for your help caring for the animals and overseeing the NEC surgical facilities.

I am also appreciative of all the work that the undergraduate and high school student

volunteers in the Capadona lab have helped me out with. The students that I have directly

mentored or who have been a great help on my projects include Gigi Protasiewicz, Priya

Srivastava, Jeremy Chang, Arielle Soffer, Frankie Wong, Smrithi Sunil, Shruti Sudhakar,

Shushen Lin, Patrick Smith, and Bill Tomaszewski. I would also like to thank Seth Meade,

Emily Molinich, Jacob Rayyan, Andres Robert, Owen Zhuang, Cara Smith, and Keying

Chen for their help counting neurons that went unrecognized.

I would like to thank Dr. Kirsch and Dr. Durand for upholding the quality of the

BME department and Neural Engineering Center. Additionally, I would like to thank all

the faculty in the NEC, Biomaterials department, BME department, and Case Western

Reserve University for providing mentorship both inside and outside the classroom. I am

especially grateful for the words of wisdom and lively conversations with Dr. Mortimer. I would also like to thank the BME administrative staff, including Carol Adrine, Sheryl

Dugard, Debra Rudolph, Anita Banks, Brian Wollenzier, and Bill Marx.

Further, I would like to thank Dr. Peachey of Louis Stokes Cleveland VA Medical

Center Research and Dr. Triolo of the APT Center for providing an excellent research environment. I would also like to thank Holly Henry, John Schaffer, Lynette Dowdy, and the rest of Research, as well Vi Huynh, Rebecca Polito, Kevin Tloczynski, and the rest of

the APT center for their help throughout the years.

xii

Last but not least, I would like to thank my parents for all the support they gave given me over the course of my PhD. Everything ranging from warm meals and laundry

service to words of encouragement and life advice. This would not have been possible

without you.

xiii

List of Abbreviations

8-OHdG 8-hydroxy-2’-deoxyguanosine

Aβ Amyloid beta

Ag/AgCl Silver/silver chloride

ALS Amyotrophic lateral sclerosis

ANOVA Analysis of variance

AP Action potential

AP1 Activator protein 1

Au Gold

B2M Beta-2 microglobulin

BBB Blood-brain barrier

BCI Brain computer interface

BdCd14-/- Chimera mouse with CD14 knocked out on blood-derived cells

BM Bone marrow

BMI Brain machine interface

CBC Complete blood count

CCL2 Chemokine (C-C motif) ligand 2 (also known as MCP-1)

CCL5 Chemokine (C-C motif) ligand 5 (also known as RANTES)

CMS Chronic modified state

COX-2 Cyclooxygenase-2

CXCL1 Chemokine (C-X-C motif) ligand 1 (also known as KC)

CD14 Cluster of differentiation 14

CD68 Cluster of differentiation 68

xiv

Cd14-/- Cluster of differentiation 14 knockout

CE Conformité Européene (European Conformity)

CHG Chlorhexidine gluconate

CLR C-type lectin receptor

CNS Central nervous system

CpG Cysteine triphosphate deoxynucleotide with a phosphodiester link

to a guanine triphosphate deoxynucleotide

CREB Cyclic AMP-responsive element-binding protein

CVD Chemical vapor deposition

DAB 3,3’-Diaminobenzidine

DAMP Damage (or danger) associated molecular pattern

DAPI 4′,6-diamidino-2-phenylindole

DBS Deep brain stimulation

DLB Dementia with Lewy bodies

DNA Deoxyribonucleic acid dsRNA Double-stranded ribonucleic acid

ECM Extracellular matrix

Ehd2 EH-domain containing 2

EEG Electroencephalogram

ECoGs Electrocorticogram

FACS Fluorescence activated cell sorting

Fc Fragment crystallizable

FDA Food and Drug Administration

FES Functional Electrode Stimulation

GCaMP No abbrev.; a calcium indicator

GFAP Glial fibrillary acidic protein

Gp96 Glycoprotein 96 (also known as HSP90B1, HSP 90kDa beta

member 1)

HCMV Human cytomegalovirus

HMGB1 High mobility group box 1

HNE Hydroxynonenal

HSP(22, 60, 70) Heat shock protein(22, 60, 70)

HSV1 Herpes simplex virus type 1

IAXO(-101, - No abbrev.; Name for small-molecule CD14 inhibitor

102)

Iba1 Ionized calcium-binding adapter molecule 1

IC14 No abbrev.; Name for an anti-CD14 antibody

IFN(-α,-β) Interferon(-α)

IgG Immunoglobulin G

IKK Inhibitor of nuclear factor-κB-kinase

IL(-1α, -1β, -6, - Interleukin(-1α, -1β, -6, -10, -12)

10, -12)

IME Intracortical microelectrode iNOS Inducible nitric oxide synthase

IP Intraperitoneal

IRAK Interleukin 1-receptor-associated kinase family

xvi

IRF3 Interferon-regulatory factor 3

JNK JUN N-terminal kinase

KC No abbrev.; a chemokine (also known as CXCL1)

LI No abbrev.; Cell adhesion molecule

LFP Local field potential

LPS Lipopolysaccharide

LRR Leucine-rich repeat

LTA Lipoteichoic acid

MAL MYD88-adaptor-like protein

MAP(K) Mitogen-activated protein (kinase)

MCP-1 Monocyte chemoattractant protein 1 (also known as CCL2)

MD-2 No abbrev.; Co-receptor for TLR4 that binds LPS

MgCd14-/- Chimera mouse with CD14 knocked out on resident brain cells,

including microglia

MHC Major histocompatibility complex

MIP(-1α, -1β, -2) Macrophage inflammatory protein(-1α, -1β, -2)

MKK MAP kinase

MnTBAP Manganese (III) tetrakis (4-benzoic acid) porphyrin

chloride

MMP(3,9) Matrix-metalloprotease (3,9)

MMTV Mouse mammary tumor virus

MSA Multiple system atrophy

MTE Microthread electrodes

xvii

MyD88 Myeloid differentiation primary-response protein 88

ND Not determined

NGS Normal goat serum

NF-κB Nuclear factor kappa-light-chain-enhancer of activated B cells

NET Nanoelectronic thread

NeuN Neuronal nuclei

NLR Nod-like receptor

NO Nitric oxide

NOX1 NADPH oxidase

Noxa1 NADPH oxidase activator 1

NT Nitrotyrosine

OCT Optimal cutting temperature p38 No abbrev.; a subset of MAPKs

PACAP Pituitary adenylate cyclase-activating polypeptide

PAMP Pathogen associated molecular pattern

PAP Peroxidase anti-peroxidase

PBS Phosphate-buffered saline

PCR Polymerase chain reaction

PD Parkinson’s disease

PDCD4 Programmed cell death protein 4

PEDOT-PEG Poly(3,4-ethylenedioxythiophene)-polyethylene glycol

PEDOT:PSS Poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate)

PEG Polyethylene glycol

xviii

PFA Paraformaldehyde

PG Peptidoglycan

PRR Pattern recognition receptor

Prnp Prion protein gene

R16PA Pre-amp model number

RAA Reactive accelerated aging

RANTES Regulated on activation, normal T cell expressed and secreted;

CCL5

RIG-I Retinoic acid-inducible gene-I

RIP1 Receptor-interacting protein 1

RLR Retinoic acid-inducible gene-I-like receptor

RNA Ribonucleic acid

RNS Reactive nitrogen species

ROS Reactive oxygen species rRNA Ribosomal ribonucleic acid

RSV Respiratory syncytial virus

RT-PCR Real time PCR

RX5 Processor model number

S100A8/9 S100 calcium-binding proteins A8 and A9

Scd1 Stearoyl-Coenzyme A desaturase 1

SCI Spinal cord injury

SD Standard deviation

SEM Scanning electron microscope

xix

SMP Shape memory polymer

SNR Signal to noise ratio

SOD1 Superoxide dismutase 1

SQ Subcutaneous

SSL3 Staphylococcal superantigen-like protein 3 ssRNA Single-stranded RNA

STN Subthalamic nucleus

SU-8 No abbrev.; Photoresist material derived from epoxy

TAB(1,2) TAK1-binding protein (1,2)

TAK1 Transforming growth factor-β-activated kinase 1

TAK-242 No abbrev.; TLR4 inhibitor also known as resatorvid

TAP2 TLR4 antagonistic peptide 2

TBK1 TANK-binding kinase 1

TBS Tris buffered saline

TCMDA Tricyclo[5.2.1.02,6]decanedimethanol diacrylate

TDT Tucker Davis Technologies tGPI Toxoplasma gondii serin protease inhibitor

TIR Toll/IL-1 receptor

TLR Toll-like receptor (general)

TLR(1-13) Toll-like receptor family member

Tlr2-/- Toll-like receptor 2 knockout

Tlr4-/- Toll-like receptor 4 knockout

TMAH Tetramethylammoniahydroxide

TNF(α) Tumor necrosis factor (α)

TRAF6 Tumor-necrosis-factor-receptor-associated factor6

TRAM TRIF-related adaptor molecule

TRIF TIR-domain-containing adaptor protein inducing IFN-β

UEA Utah Electrode Array

WT Wildtype

WTi Tungsten-titanium alloy

ZC16 Headstage model number

xxi

The Role of Innate Immunity in the Response to Intracortical

Microelectrodes

by JOHN KARL HERMANN

Abstract

Intracortical microelectrodes exhibit enormous potential for researching the nervous system, steering assistive devices and functional electrode stimulation (FES) systems for severely paralyzed individuals, and augmenting the brain with computing power. Unfortunately, intracortical microelectrodes often fail to consistently record signals over clinically useful time frames. Biological mechanisms, such as the foreign body response to intracortical microelectrodes and self-perpetuating neuroinflammatory cascades contribute to the inconsistencies and decline in recording performance. The overall goal of this work was to investigate the role of innate immunity signaling in the foreign body response to intracortical microelectrodes. In this dissertation I examined the effect of cluster of differentiation 14 (CD14) inhibition via a systemically administered, small-molecule antagonist, as well as knockout mouse models on intracortical microelectrode recording performance and tissue integration. Mice receiving the small molecule antagonist to CD14 (IAXO-101) exhibited a significant improvement in recording performance over the 16-week experiment. Additionally, CD14 knockout mice exhibited significant improvements in recording performance in the first two weeks after implantation, but not the remainder of the study. These findings suggest that full removal of CD14 is helpful in the first two weeks after implantation, but a limited amount of CD14 signaling may be useful at later time points. Further, we investigated the role of two dominant co-receptors to CD14, Toll-like receptor 2 (TLR2) and Toll-like receptor 4

(TLR4), in the foreign body response to intracortical microelectrodes. The TLR4 knockout

mice exhibited significant decreases in blood-brain barrier permeability at 2 and 16 weeks after implantation, despite exhibiting significantly reduced neuronal survival at 16 weeks after implantation. These results suggest that TLR4 plays a role in the mediation of blood- brain barrier integrity as well as neuroprotective mechanisms, so full removal of TLR4 is also detrimental to chronic integration of intracortical microelectrodes. The TLR2 knockout mice did not exhibit any histological differences from wildtype mice at 2 or 16 weeks, suggesting that TLR2 does not play a major role in the foreign body response to intracortical microelectrodes. The findings of this work suggest the involvement of CD14 and TLR4 in the foreign body response to intracortical microelectrodes. However, implementation of innate immunity inhibition to improve the long-term performance of intracortical microelectrodes requires temporal refinement of inhibition strategies.

Chapter 1

Specific Aims

The goals of the work described in this dissertation are to (1) investigate the effect

of systemic innate immunity inhibition on intracortical microelectrode performance and

tissue integration and (2) characterize the role of innate immunity signaling pathways in

the foreign body response to intracortical microelectrodes.

Intracortical microelectrodes are devices that enable communication with neurons in the brain for applications in basic research, rehabilitation, and emerging commercial applications. Unfortunately, declining and inconsistent recording performance over time hinder the long-term application of intracortical microelectrodes. In addition to mechanical

and material failure mechanisms, biological failure mechanisms are hypothesized to

contribute to inconsistent and declining recording performance [1, 2]. The foreign body

response to implanted intracortical microelectrodes results in several outcomes potentially

detrimental to recording performance, including neuronal dieback, astrocytic

encapsulation, blood-brain barrier disruption, and material degradation. Self-perpetuating

chronic inflammatory cascades may facilitate these detrimental effects over time.

Many researchers have attempted to mitigate biological failure utilizing various

strategies [1] such as smaller probe designs [3-5], more compliant materials [6-8] , biomimetic coatings [9, 10], and antioxidant treatments [11, 12]. Anti-inflammatory drugs such as minocycline and dexamethasone have demonstrated promising improvements in recording performance [13] and/or tissue integration [13, 14]. However, long-term administration of these broadly-acting anti-inflammatory drugs may lead to some detrimental side effects [15-18]. Narrowing the scope of anti-inflammatory treatment by

identifying therapeutic targets more specific to the chronic inflammatory response to

intracortical microelectrodes may improve long-term treatment outcomes.

Innate immunity is a fast-acting component of the immune system that utilizes

pattern recognition receptors to identify and respond to threats. Toll-like receptors (TLRs) are a family of innate immunity pattern recognition receptors that recognize molecular patterns associated with pathogenic threats and tissue damage to promote inflammatory activation [19]. Toll-like receptor 2 (TLR2) and toll-like receptor 4 (TLR4) are two TLRs that are associated with neurodegenerative and neuroinflammatory disorders and recognize pathogen associate molecular patterns (PAMPs) and damage associated molecular patterns

(DAMPs) that may be present at the electrode tissue interface [20-23]. Cluster of differentiation 14 (CD14) is a co-receptor to both TLR2 and TLR4 that coordinates ligand binding to TLR2 and TLR4 [24-26]. Detection of tissue damage by TLR2, TLR4, and

CD14 on microglia and macrophages and subsequent inflammatory activation may perpetuate the chronic inflammatory cascades associated with intracortical microelectrodes. The primary ligand for CD14 and TLR4, lipopolysaccharide (LPS) or bacterial endotoxin, has demonstrated detrimental effects on recording performance [27], neuronal survival [27, 28], microglial activation [28], astrocytic encapsulation [28], and blood-brain barrier permeability [28]. Thus, we hypothesize that TLR2, TLR4, and/or

CD14 play roles in the foreign body response to intracortical microelectrodes. Further, we

hypothesize that attenuation of TLR2, TLR4, and/or CD14 will improve intracortical

microelectrode performance and tissue integration.

Aim 1: Investigate the effect of genetic and therapeutic inhibition of CD14 on intracortical microelectrode performance and tissue integration.

Working hypothesis: Systemic inhibition and full genetic removal of CD14 will improve

tissue integration and recording performance of intracortical microelectrodes.

Rationale: Self-perpetuating chronic inflammatory mechanisms are hypothesized to

contribute to intracortical microelectrode failure [11]. The co-receptor CD14 aids both

TLR2 and TLR4 in the recognition of pathogens and tissue damage [24]. Inhibiting or removing CD14 may reduce the capability of TLR2 and TLR4 expressed on microglia and macrophages to recognize ligands and promote inflammatory mechanisms. Attenuation of inflammatory mechanisms may improve the recording performance and tissue integration intracortical microelectrodes [13, 14].

Design: Silicon planar NeuroNexus intracortical microelectrodes were implanted into three

groups of mice: 1) wildtype mice systemically administered a small-molecule CD14

inhibitor (IAXO-101, Innaxon) every other day, 2) CD14 knockout mice completely

lacking(Cd14-/-), and 3) no treatment wildtype control mice. Awake recordings were

obtained daily through post-implantation day 5, and twice weekly until 16 weeks after

implantation. Single unit activity was extracted using the un-supervised sorting algorithm wave_clus. Endpoint histology was evaluated at 16 weeks after implantation using

Immunohistochemical markers for neurons, microglial activation, astrocytic encapsulation, and blood-brain barrier permeability (Chapter 3) [29].

Aim 2: Characterize the role of TLR2 and TLR4 in the neuroinflammatory response to intracortical microelectrodes.

Working hypothesis: Knockout mice lacking TLR2 (Tlr2-/-) or TLR4 (Tlr4-/-) will exhibit enhanced neuronal survival and decreased microglial activation, astrocytic encapsulation, and blood-brain barrier permeability in response to intracortical microelectrode implantation.

Rationale: The findings of Aim1 indicated that partial inhibition of CD14 via a small molecule inhibitor improved recording performance over a longer time frame than full genetic removal of CD14 [29]. Further, Bedell et al. indicated inhabiting CD14 only on blood-derived cells improved recording performance. These findings suggest that partial inhibition of CD14 promotes better outcomes than full removal. Another way to partially inhibit CD14 signaling is to remove one of its major downstream effectors. Thus, we wanted to demonstrate the effects of removing TLR2 or TLR4 on the foreign body response to intracortical microelectrodes.

Self-perpetuating inflammatory cascades are hypothesized to cause poor tissue integration of intracortical microelectrodes, resulting in neuronal dieback, microglial activation, astrocytic encapsulation, and blood-brain barrier permeability [11]. The innate immunity receptors TLR2 and TLR4 recognize PAMPs and DAMPs and subsequently promote pro- inflammatory responses, such as the release of cytokines, chemokines, and reactive oxygen species [20, 21, 26]. These factors have detrimental effects on neurons and blood vessels

[30-33]. The absence of TLR2 or TLR4 in knockout mice may reduce the recognition of damage caused by insertion trauma, micromotion, and chronic inflammation. Reduction of damage recognition in microglia and macrophages may reduce further pro-inflammatory

6 activation and subsequent neuronal loss, astrocytic encapsulation, and blood-brain barrier permeability.

Design: Silicon shanks with the same geometry as NeuroNexus planar arrays were implanted into Tlr2-/-, Tlr4-/-, and wildtype control mice. Endpoint histology was evaluated at 2 and 16 weeks after electrode implantation using immunohistochemical markers or 3,3'-

Diaminobenzidine (DAB) for neurons, microglial activation, astrocytic encapsulation, and blood-brain barrier permeability (Chapter 4) [34].

Chapter 2

Introduction

2.1 Applications of Neural Interfacing

Neural interfaces facilitate the transfer of information between the nervous system and technology and have demonstrated increasing usefulness in basic research [35] and

rehabilitation [36]. Several companies are developing neural interfaces for both

rehabilitation and augmentation purposes [37, 38]. Thus, improving the performance and

longevity of usefulness of neural interfacing technology is a growing area of biomedical

research.

Neural interfaces may interact with the central nervous system via the brain [39],

spinal cord [40], and dorsal root ganglia [41], or with the peripheral nervous system via

motor [42], sensory [43], and autonomic nerves [44]. Information may be extracted from

the nervous system by recording electrical activity [39, 42, 44], and information may be

sent into the nervous system by stimulation with electrical pulses [41, 43], magnetic fields

[45], light [46-48], or chemicals [47, 49]. Advanced neural interfacing systems may extract information from one part of the body apply control signals to other parts of the body [39].

One major application of neural interfacing is the restoration of motor function.

Injuries to the spinal cord, amputations, and neurodegenerative disorders may result in loss

of motor control. Signals recorded from motor nerves, the spinal cord, or the motor cortex

may be decoded and utilized to steer assistive devices to partially compensate for the loss

of motor control [50]. Whereas motor control may be improved using control signals from

motor nerves and muscles after certain amputations [50], fewer control signals are available to individuals with high-cervical spinal injuries, and signals from the brain may be useful

for achieving control with higher degrees of freedom [36]. Signals recorded from the brain

have been decoded to control computer cursors [51], robotic arms [52], prosthetic limbs

[53], and a patient’s own arm via functional electrical stimulation [39] in clinical trials.

Experiments in non-human primates have also demonstrated control of lower-limb exoskeletons [54] and wheelchairs [55, 56] with brain machine interfaces. Neural interfacing technology has demonstrated promising results in the restoration of motor function.

Another major application of neural interfaces is the attenuation of brain pathologies. Deep-brain stimulation (DBS) has been successfully implemented in the treatment of Parkinson’s disease [57]. The utilization of DBS to treat epilepsy [58],

Tourette syndrome [59], depression [60], and obsessive compulsive disorder is under clinical investigation [61]. Effective stimulation targets have been identified for a variety of pathologies, however the exact mechanisms of action are still under investigation.

Also, neural interfaces have been widely implemented in hearing restoration via

cochlear implants [62]. These devices record audio signal externally, decode the signal

into a series of current pulses that vary in amplitude over time, and deliver the pulses to the

auditory nerve from electrode channels spanning different depths of the cochlea, sorted by

frequency [62]. Cochlear implants enable some restoration of speech perception [62].

Additionally, neural interfaces have utilized in the mitigation of chronic pain.

Electrical stimulation pulses may be applied to the spinal cord [63] or peripheral nerves

[64] to disrupt problematic neural firing associated with pain. Stimulation may be applied via implanted electrodes [63, 64] or transcutaneous systems. Further, neural interfacing

technology has been applied to the control of bladder function. Electrical stimulation may

initiate micturition when needed [65] or prevent micturition when unwanted [66].

In addition, neural interfaces are under investigation for the restoration of sight.

Some degree of vision restoration has been achieved through stimulation of the retina [67], optic nerve [68], lateral geniculate nucleus [68], or visual cortex [68]. One retinal implant, the Argus II, has received FDA approval as well as the European CE marking [67].

Another system, the Alpha-IMS, also has the CE marking, and several other devices are being tested in animals [67]. Despite the success of retinal implants, they require functioning retinal ganglion cells for use [68]. Thankfully, the other modalities of sight restoration are under investigation [68].

Finally, the alteration bodily functions innervated by the vagus nerve has garnered interest recently. The vagus nerve connects the brain with the autonomic, cardiovascular, respiratory, gastrointestinal, immune, and endocrine systems via afferent and efferent fibers and regulates various bodily functions [69]. Vagus nerve stimulation is an FDA approved intervention for epilepsy and depression and is under investigation for the treatment of inflammation, and asthma [70], as well as psychiatric disorders such as dementia and schizophrenia [71].

Overall, communication with the nervous system via neural interfaces provides many promising opportunities for therapeutic intervention in a variety of injury and disease states.

2.2 Neural Interfacing Methods

A variety of neural interfacing methodologies enable communication between the nervous system and technology, varying in component of the nervous system, level of

10 invasiveness, and type of information conferred, and the direction of information flow.

Neural interfaces interact multiple components of the central and peripheral nervous systems, including the brain [1, 45, 57, 72], spinal cord [73, 74], and nerves [75-79]. Types of information transferred across neural interfaces include electricity, magnetic fields, light, and chemicals. Information may flow from the body into a device instances of recording and measurement or flow from a device into the body in instances of stimulation.

The wide variety of neural interfaces give rise to versatile solutions to the applications covered in Section 2.1.

Neural interfaces be achieved with varying degrees of invasiveness. More invasive neural interfaces operate at a closer proximity to neurons and typically confer higher spatial resolution [80], allowing more degrees of freedom in control signals [81] or more selectivity in neural stimulation [82]. Less invasive neural interfaces may be sufficient for certain applications, such as environmental control, yes or no communication, and simple neuroprostheses [83], and bypass the harmful side-effects of device implantation [81]. The trade-offs between invasiveness and spatial resolution must be considered in the context of the neural interface application.

In the brain, the least invasive neural interfaces are located outside the skull.

Electroencephalograms (EEGs) can be recorded on the surface of the scalp using surface electrodes (Table 1) [1, 72, 84]. Stimulation of neurons in the brain is also achievable outside of the skull using trans-cranial magnetic stimulation (Table 1) [45]. Increasing the level of invasiveness and spatial resolution, electrocorticograms (ECoGs) record neural activity below the skull, either above or below the dura (Table 1) [1]. The most invasive and highest resolution neural interfaces in the brain are intracortical microelectrodes, which

protrude into the cortical parenchyma and allow single unit recordings (Table 1) [1].

Intracortical microelectrodes, the main type of neural interface studied in this thesis, are covered in greater detail in Section 2.3. Additionally, Deep brain stimulation (DBS) electrodes stimulate structures deep within the brain with a high degree of selectivity

(Table 1) [57].

Table 1. Major types of interfaces with the brain with varying degrees of invasiveness. Information from [1, 45, 57].

Brain Interface Location Types of Brain Interfaces External EEG Trans-cranial magnetic stimulation

Epidural Epidural ECoG

Subdural Subdural ECoG

Cortex Intracortical Microelectrode DBS electrode

In the spinal cord, neural interfaces with varying levels of invasiveness may also

be implemented. Spinal cord stimulation typically occurs outside the dura with epidural

electrodes in humans [73] but may occur below the dura with electronic dura mater

electrodes [74] or inside the spinal cord with intraspinal electrodes such as microwires or

microelectrode arrays [73] in animal experiments.

Interfaces with nerves also exhibit varying degrees of invasiveness. Recording or

stimulation of nerves may occur outside the nerve via cuffs [75] or flat interface nerve

electrodes [76]. Increasing the degree of invasiveness, recording or stimulation inside of

nerves but outside of fascicles may occur using extrafascicular electrodes such as Slowly

Penetrating Interfascicular Nerve Electrodes [77]. Further increasing the degree of

invasiveness, recording or stimulation of nerves can occur inside of fascicles using

intrafascicular electrodes such as transverse intrafascicular multichannel electrodes [78] or

longitudinal intrafascicular electrodes [79].

Traditionally, neural interfaces communicate with electricity, but alternative signals are being implemented. Electrodes may record changes in extracellular voltage resulting from neuronal action potentials [1] or emit electrical waveforms that alter transmembrane potential and open voltage-gated ion channels to induce action potentials

[85]. Magnetic fields are being utilized to stimulate regions of the brain non-invasively in humans [45]. Magnetic fields sent to the brain induce electrical currents in the brain, as described by Faraday’s law of electromagnetic induction, which stimulate the axons of neurons [45]. Additionally, light may be employed in direct stimulation of neurons [86]

or as an activator of optogenetic channels [87]. Finally, chemicals may be utilized as a neural interface signals, including neurochemical sensors used to monitor the release of neurotransmitters [88] or microfluidic probes that can deliver pharmacological agents into neural tissue [49]. Neural interfacing technology is moving towards incorporating multiple types of signaling within a single device [89]. A plethora of options for transferring information between the nervous system and technology are available.

A variety of strategies have been implicated to communicate with many neural

tissues. Strategies that enable higher resolution communication typically come at a cost of higher invasiveness. Neural interfaces may employ one or more types of signals facilitate communicate between technology and tissue. A wealth of options for neural interfacing is promising for next generation nervous system technologies.

2.3 Intracortical microelectrodes

As introduced in Section 2.2, intracortical microelectrodes are neural interfacing devices that record electrical signals from neurons in the cortex. Intracortical

microelectrodes penetrate the cortex and are thus highly invasive [1]. Penetration of the

cortex allows the recording contacts to be in close proximity with populations of neurons

without low-pass filtering of signals by the meninges, skull, and scalp [90]. Thus,

intracortical microelectrodes are capable of recording electrophysiological signals with

high spatial resolution, including single unit activity [91] and threshold crossings [92].

Additionally, intracortical microelectrodes may also record local field potentials (LFPs)

[93]. Common intracortical electrode types, device function, and implementation will be

covered in this section.

Several types of intracortical microelectrodes are commercially available and widely implemented in vivo (Figure 1), including microwires, Michigan-style arrays, and

Utah electrode arrays. These types of intracortical microelectrodes will be discussed in

detail.

Microwires are composed of strands of conductive metal, such as iridium, coated

with an insulating layer, such as parylene-c, except for an exposed tip for recording [94]

(Figure 1A). To compensate for the single recording site present on a microwire, they are

commonly arranged into arrays of multiple microwires to acquire more data over a 2

dimensional area or different depths [95]. Microwire arrays are commercially available

from a variety of vendors, such as Tucker Davis Technologies. Microwires and their

precursors have been used to record signals from small populations of neurons for nearly

80 years, and they remain in use for both acute and chronic electrophysiological

experiments [1]. Progressing from blunt Ag/AgCl electrodes in the 1940s [96], to

electrolytically pointed stainless steel electrodes in the 1950s, and to platinum-iridium

alloys coated in Parylene-C in the 1970s [94], microwire arrays have demonstrated longer

and longer functional lifespans [1]. Schmidt et al. were even able to achieve single unit

recordings out to 223 days after implantation back in 1976 [97]. Although these devices

showed early promise, they are still hindered by inconsistent and declining performance

described in Section 2.4 to this day [1, 98, 99].

Figure 1. “Schematic representation of the generation design of leading microelectrode array technologies, including (A) microwires, (B) Michigan-style microelectrodes, [and] (C) Utah electrode arrays (EUA).” Figure and caption reproduced from [1], with permission from IOP Publishing. The next major type of electrode is the Michigan-style electrode, composed of a planar sheet of silicon with multiple recording contacts along its length (Figure 1B) [100,

101]. This design allows for recordings at multiple depths along a single shank, although arrays of multiple shanks may allow for sampling two or three dimensions (Figure 1B)

[101]. Michigan-style electrodes were developed by Wise and colleagues at the University of Michigan in the 1980s, building off of previous work on silicon-based electrodes at Bell

Telephone Laboratories in the 1960s [102]. Michigan style electrodes achieve micron scale designs, such as 123 µm x 15µm cross-section shanks with 15μm or 30 µm diameter

contacts, by various microfabrication techniques [101]. Michigan-style arrays are

commercially available from NeuroNexus. Michigan style arrays have been chronically

implanted in a variety of animal models [29, 91, 103-105], but have not been approved for

human use to this date. Similar to microwire electrodes, Michigan-style electrodes exhibit

inconsistent and declining performance over time [1, 106, 107].

The third major type of intracortical electrode is the Utah Electrode Array (UEA),

composed of a silicon-based grid of insulated tines with conductive recording tips (Figure

1C) [108]. Current iterations feature sputtered iridium oxide contacts [109] and Parylene-

C insulation [110]. The standard design of the UEA allows for sampling within a 2

dimensional area at a single depth, but slanted designs allow for some variability in depth

(Figure 1C) [111]. The UEA was developed at the University of Utah by Normann and

colleagues and published on in the early 1990s [108].Since then, UEA have been

chronically implanted in a variety of animal models, including non-human primates [112,

113]. Utah arrays are the only type of intracortical microelectrode with a CE mark and

FDA approval for use in humans, and are currently being employed in the BrainGate

clinical trials to acquire control signals from the brains of tetraplegic individuals to control

assistive devices [52, 53] or functional electrical stimulation systems [39]. Utah-style

arrays are commercially available from Blackrock Microsystems, LLC. Much like the microwire and Michigan-style electrodes, UEAs are hindered by inconsistent and declining

recording performance as well [2, 114].

This list of electrode types discussed above was limited to widely distributed styles.

Emerging designs of intracortical microelectrodes will be covered in Section 2.6.

Intracortical microelectrodes are a useful tool for recording neuronal activity.

Neuronal action potentials feature the flow of ions across neuronal membranes, and the

resulting changes in extracellular ionic concentration are detected as changes in voltage by

intracortical microelectrodes [1]. The voltage signals recorded by intracortical

microelectrodes are comprised of the action potentials and other electrical activity of many

neurons plus various sources of noise [115, 116]. Extracellular signal amplitude decays

with distance from the electrode, so the action potentials of neurons in close proximity to

the electrode are detected with higher amplitude than action potentials from distant neurons

[117]. After filtering neural signals, spiking events that exceed a threshold value may be

used to calculate threshold crossing rates [52, 92] or isolated and sorted into single units

manually [51] or with various algorithms [118]. Single unit recordings distinguish spiking

events based on neuron of origin using characteristics of waveforms, enabling the

monitoring of individual neurons. Threshold-crossing spiking activity is hypothesized to be detectable from neurons within a radius of 140 μm from an electrode, and single unit activity is hypothesized to be detectible from neurons within a radius of 50 μm from an electrode [117]. Alternatively, changes in LFPs may also be monitored with intracortical

microelectrodes [93]. Local field potentials are extracellular electrical potential fields that

encompass electrical activity from neuronal and non-neuronal sources in a given volume

[115]. Threshold crossings [52, 92], single unit activity [51, 119], or LFPs [93] may be

decoded for use as control signals in brain-computer and brain-machine interfaces.

As described in Section 2.1, signals recorded from intracortical microelectrodes implanted in humans have demonstrated usefulness as control signals for cursors on a computer screen [51], robotic and prosthetic arms [52, 53], as a patient’s own paralyzed

arm via functional electrode (FES) stimulation [39]. In addition to rehabilitation

applications, companies such as Neuralink and Kernel are developing technology to

augment human capabilities beyond normal function [38]. Specific applications remain to

be revealed.

Despite the exciting potential of intracortical microelectrodes, they do not record

signals consistently over extended periods of time. Intracortical microelectrode failure is

covered in detail in in Section 2.4.

2.4 Failure of intracortical microelectrodes

Intracortical microelectrodes should ideally operate consistently over time ranges

of years to decades to prevent the repetition of highly invasive brain surgeries.

Unfortunately, intracortical microelectrodes fail to consistently record neural signals over

extended periods of time. Many labs studying various chronically implanted intracortical

microelectrodes in several animal models and humans have demonstrated trends of

decreasing number of units detected [9, 114, 120-123], number of channels detecting units

[2, 98, 99, 107, 114, 120, 124], signal amplitudes [2, 122, 125], and decoder performance

[122], as well as generally inconsistent recording performance [98, 120, 126] and

recording longevity [2, 124, 126]. Even at more acute time points, intracortical microelectrodes have been characterized as unstable [124, 126]. These characteristics are not ideal for the long-term performance of brain-computer and brain-machine interfaces.

Declining and inconsistent recording performance have been ascribed to many failure mechanisms, typically grouped into mechanical, material, and biological mechanisms [2]. As defined in a retrospective analysis of microelectrode array failures in monkeys by Barrese et al. (Figure 2), mechanical failures included physical relocation or

18 damage of the electrode array and hardware, material failures included breakdown of electrode array materials, and biological failures included consequences of the foreign body response and implantation trauma following device insertion, as well as clinical issues with the implant [2].

Figure 2. “Major failure modes of MEAs.” “Ideal placement in cortical tissue, about 1 (or 1.5) mm into the cortex. A thin layer of arachnoid overgrowth encases the platform that sits on the pia-arachnoid surface and helps to stabilize the array. (b) Biological failures: bleeding, cell death, hardware infection, meningitis, gliosis, or meningeal encapsulation and extrusion. Macrophages originating in the subarachnoid space may mediate the encapsulation response. (c) Material failures: broken electrode tips, insulation leakage, or parylene cracks and delamination. Note that the latter three would lead to lower impedances and spike amplitudes due to shunting. (d) Mechanical failures: wire bundle damage, connector damage, and mechanical removal. A dural stitch is shown as one possible source of tethering that results in electrodes being pulled out of the brain.” Figure and caption reproduced from [2], with permission from IOP Publishing.

Mechanical failures observed by Barrese et al. included the breakdown of array- anchoring materials, loosening of microelectrode array, damage to microelectrode array connector pins, disconnection and breakage of wiring, and removal of electrode arrays from the brain [2]. Many of these observed mechanical failure resulted from collisions with monkey subjects with their cages, altercations with other monkeys, and handling by

researchers. Although compliance is likely less of an issue in human subjects and patients,

the possibility of accidents involving users and caretakers must be factored into future

intracortical microelectrode array anchoring schemes.

In addition to macroscale mechanical disruptions of intracortical microelectrode

arrays, Kozai et al. examined the effects of mechanical mismatch between the components

of planar silicon intracortical microelectrode arrays [127]. High strain was observed at the interfaces between iridium and silicon, especially at electrical lead traces, which coincided with cracking and delamination [127]. Cracking and delamination of traces may lead to open circuits or increased electrode surface area, respectively [127]. Mechanical mismatch of array components must be considered in future microelectrode array designs.

Material failure mechanisms have been studied by a variety of research groups for multiple electrode styles. Prasad et al. examined tungsten microwire arrays previously implanted in rats using SEM and observed peeling and cracking of insulation layers, as well as deterioration of electrode recording sites [98]. In a follow-up study with platinum- iridium microwire arrays previously implanted in rats, Prasad et al. observed corrosion of recording sites and deterioration of insulation, and insulation delamination coincided with a decrease in electrical impedance and poor recording performance [99]. Gilgunn et al.

observed fissures in the insulation material on recording Utah arrays implanted into

monkeys [128]. As stated above, Kozai et al. observed cracking of conductive traces on

planar silicon arrays implanted into mice [127]. Barrese et al. observed initially defective

electrode arrays, shorting of connectors due to fluid infiltration, and damage of wiring due

to flawed connector design [2]. Arrays can be tested prior to implantation to screen for

defects, fluid infiltration was mitigated with better cleaning protocols and protective caps,

and flawed connectors were replaced with a more stable solution [2]. However, the degradation of electrode sites and insulating material may be more problematic for long- term device functions, and require attention in future microelectrode array designs.

Biological failures are typically associated with the interruption of electrical signals between cortical neurons and recording contacts. Neuronal death that accompanies intracortical microelectrode implantation and chronic presence may reduce the number of neurons within detectable distance (Figure 3) [129], estimated to be within a 50 μm radius for single units or 140 μm for threshold-crossing spikes [117]. Signals detected from neurons beyond that distance may have amplitudes that are indistinguishable from background neuronal hash. Further, studies investigating multi-shank microelectrode arrays have observed tissue cavitation around implanted microelectrodes, contributing to the loss of neurons at the electrode-tissue interface [123]. In general, neurodegeneration around implanted microelectrode arrays correlates with chronic inflammatory mechanisms associated with the implant [130], which will be covered in detail in Section 2.5. A decrease in neural activity may also contribute to intracortical microelectrode failures

[131]. In addition to the loss of firing neurons, astrocytes of the brain parenchyma become hypertrophic and form a dense encapsulation layer around the implanted electrode [132,

133], which may increase the electrical impedance between the electrode contact and remaining neurons [1, 97, 134, 135]. Further, permeability of the blood-brain barrier following intracortical microelectrode implantation has been tied to poor recording performance [136]. Saxena et al. hypothesized that blood-brain barrier permeability promotes chronic inflammation, leading to neurodegeneration and electrode failure [136].

Blood-brain barrier permeability, especially within the context of chronic

neuroinflammation, will be examined in more detail in Section 2.5. Additionally, Barrese et al. observed cerebral edema related to surgical complications, infection, meningeal encapsulation and subsequent extrusion from the brain [2]. Meningeal encapsulation may lead to similar impedance issues as hypothesized with astrocytic encapsulation, and extrusion from the brain would move the electrode contacts away from the neurons of interest. Biological failure mechanisms pose serious threats to long-term recording performance. Detailed description of the causes of biological failure will be discussed in

Section 2.5, including self-perpetuating mechanisms and contributions to other failure mechanisms.

Figure 3. “Electrode implantation results in localized pro-inflammatory cellular and biochemical events.” “Early after implantation, activated microglia begin to attach to the surface of the electrode and locally release pro-inflammatory factors. Glia cell adhesion is followed by astrocytic encapsulation along the entire shaft of the electrode (formation of the glial scar). These events, as well as localized hemorrhaging, have been shown to be correlated with neurodegeneration at the interface. Representative IHC images of the dominant cell types are shown left. Scale = 100 mm.” Figure and caption reproduced from [1], with permission from IOP publishing.

Finally, electrode failures may not cleanly fall into one category. Barrese et al. observed failures that could not be distinguished, and thus were labeled as unknown failure mechanisms [2]. These unknown failures may result from a combination of failure types.

For example, mechanical mismatch contributes to material breakdowns [127]. Further, biological mechanisms may also promote material failures. Oxidative factors released during inflammation have been shown to promote material breakdowns in advanced aging experiments [137]. Multiple factors must be considered when preventing intracortical microelectrode failure.

Overall, a variety of factors contribute to intracortical microelectrode failure, including mechanical failures, material failures, biological failures, and combinations of failures. The foreign body response to intracortical microelectrodes will be covered in more detail in Section 2.5, and a variety of previously attempted strategies to address biological failure mechanisms will be covered in Section 2.6.

2.5 Foreign body response to intracortical microelectrodes

As described in Section 2.4, intracortical microelectrodes may fail, in part, due to biological mechanisms, many of which result from the foreign body response to the implant and subsequent chronic inflammatory cascades (Figure 4, for review see Jorfi et al. [1]).

As a microelectrode array is inserted in the brain, blood vessels are ruptured [138], releasing blood proteins into the brain parenchyma. Several components of blood, including proteins are neurotoxic [139]. Further, tissue is displaced and neurons and glia along the insertion path are damaged or killed [132]. Blood proteins will adsorb to the surface of the electrode array and denature, upon which they are recognized by the brain’s major inflammatory cells, the microglia [11, 140]. Subsequently, microglia transition

from a dormant ramified state to a pro-inflammatory amoeboid phenotype by retracting processes and upregulating lytic enzymes [132]. In the activated state, microglia are able to release a variety of soluble pro-inflammatory and cytotoxic factors, such as pro- inflammatory cytokines and chemokines, nitric oxide, reactive nitrogen species, and reactive oxygen species [33, 129, 141, 142]. Evidence of oxidative gene expression and

cellular oxidative damage have been detected at the electrode tissue interface (Appendix

7.1) [30]. In addition to the mechanical damage caused by electrode insertion, factors

released by pro-inflammatory microglia may contribute to neuronal death and

neurodegeneration (Figure 5) [33, 129, 141, 142]. Additionally, soluble factors released

by microglia may damage the blood-brain barrier, by disrupting tight and adherens junctions connecting endothelial cells of the vasculature [143].Further, oxidative factors may contribute to material degradation (Figure 5) [137]. Thus, consequences of electrode implantation and early inflammatory activation may contribute to the decline or unreliability of electrode performance.

Figure 4. “Serum protein recognition propagates inflammatory response to intracortical microelectrodes.” “After intracortical microelectrode implantation, the damage of localized vasculature can result in two mechanistic paradigms at the interface of the implanted device. Left: Extravasated serum proteins from damaged vasculature become adsorbed onto the surface of the implanted device and dispersed throughout the cortical tissue. Denatured serum proteins then activate resident microglia cells and result in the release of pro-inflammatory molecules. Release of pro-inflammatory molecules, capable of activating more inflammatory processes, can also facilitate a self-perpetuation of blood–brain barrier (BBB) breakdown around the implant. Right: Extravasated serum proteins can also be recognized by neurons, resulting in neuronal apoptosis. Cellular debris can also mediate further microglia activation and initiate further BBB instability.” Image and caption reproduced from [11], with permission from Elsevier.

Following the detrimental effects of electrode implantation and early inflammatory

activation, chronic inflammatory mechanisms may also contribute to recording instability

and failure (Figure 4). Attracted by the release of soluble factors by nearby microglia,

more microglia may migrate from the parenchyma and monocytes and other myeloid cells

may be recruited from circulation across the leaky blood-brain barrier [136, 144]. Blood

proteins and necrotic cells from insertion damage, as well as the early inflammatory

response and leaky blood-brain barrier may promote pro-inflammatory activation in the

newly recruited inflammatory cells [11, 136, 145]. One pathway by which microglia and

macrophages recognize such damage and enact pro-inflammatory responses is through

Toll-like receptors [20]. Section 2.9 and onward will discuss these receptors in depth.

Soluble factors from the total collection of inflammatory cells around the implant may

cause further damage to neurons, recording materials (Figure 5), and the blood-brain

barrier [33, 129, 141, 142]. Repeated damage to cells and blood-vessels resulting in the

recruitment and activation of more inflammatory cells is hypothesized to propagate self-

perpetuated chronic inflammatory cascades indefinitely [11, 136]. The neuronal and

material damage resulting from self-perpetuating inflammatory cascades may contribute to

long-term biological failure mechanisms [11, 136].

Figure 5. “Oxidative stress following neural probe implantation.” “The implantation of neural probes leads to the overproduction of reactive oxygen species (ROS) which can consequently (1) perpetuate the foreign body response, (2) facilitate neuronal death, and (3) facilitate corrosion and delamination of the microelectrode surface.” Image and caption reproduced from [30] with permission from the publisher.

In addition to the chronic-inflammatory mechanisms initiated by the initial implantation damage, mismatch of the mechanical properties between the electrode and surrounding brain tissue may also contribute to neuronal damage and inflammatory propagation. Mismatch in the elastic modulus of microelectrode array substrates and brain tissue may induce strain on the surrounding cells and tissue during micromotion of the brain against a tethered electrode [146-149], related to respiration and vascular pulsation

[150]. Tissue strain may damage neurons, other cells, and vasculature, contributing to self- perpetuating inflammatory mechanisms [151-153]. Many labs have attempted to mitigate

26 this issue by using compliant microelectrode array materials [6, 154-159], which will be covered in Section 2.6.

As mentioned in Section 2.4, encapsulation of electrodes by astrocytes and meningeal fibroblasts may be detrimental to intracortical microelectrode performance.

Following microelectrode array implantation, astrocytes of the brain will proliferate, become hypertrophic, and migrate to the implantation site, resulting in a dense encapsulation layer around the electrode [132]. Astrocytic responses typically start wider and more diffuse, and then become denser and more compact several weeks later [132].

Astrocytic encapsulation tends to isolate the inflammatory region around the electrode from the surrounding parenchyma. In addition to astrocytes, meningeal fibroblasts may migrate to the electrode-tissue interface and contribute to electrode encapsulation [2, 9].

As stated in Section 2.4, encapsulation of the electrode may increase the electrical impedance between the electrode and neurons [160], increase the distance between neurons and electrodes [126], or in extreme cases promote extrusion of electrodes arrays [2].

Overall, chronic inflammation, neuronal loss, blood-brain barrier permeability, and encapsulation of microelectrode arrays may stem from the foreign body response to implanted intracortical microelectrodes. These factors may contribute to intracortical microelectrode instability and failure, and many researchers have tried to mitigate these factors utilizing strategies covered in Section 2.6.

2.6 Interventions to mitigate biological failures

In recent years, a variety of design strategies have been implemented to overcome biological failure mechanisms of intracortical microelectrodes. A few categories of strategies to overcome biological failures include decreasing implant size, increasing

implant compliance, implementation of bioactive coatings, and antioxidant treatments.

Each strategy category and the various attempts within each category achieved varying

degrees of success. Here, attempts to mitigate biological failures are reviewed with

particular focus on studies investigating long-term recording performance or histology.

One strategy that has garnered a lot of attention recently is the reduction of

intracortical microelectrode probe size. Reducing probe size or surface area has been

hypothesized to lessen the severity of foreign body responses for several reasons [1],

including reduction of iatrogenic injury [161, 162], more favorable mechanics [163],

reduction in surface for inflammatory cell binding to confer unfavorable

mechanotransduction [163] or secrete soluble pro-inflammatory factors with high concentration [164], as well as additional unknown factors [1]. Regardless of mechanism, reducing probe size has demonstrated improved tissue integration and recording performance [3-5, 165]. Many of the ultrasmall electrodes are, at least in part, composed

of carbon fibers [4, 5, 165]. Kozai et al. developed ultrasmall composite electrodes,

referred to as microthread electrodes (MTEs) (Figure 6A), with a diameter of 7 µm, which

featured a carbon fiber core, poly (p-xylylene)-based dielectric layer, poly (thiophene)-

based recording pad, and PEDOT:PSS (Poly(3,4-

ethylenedioxythiophene):poly(styrenesulfonate)) recording sites [5]. The MTEs exhibited

stable recordings over five weeks in the brain, with high single unit yields and significant

improvements in SNR over larger silicon microelectrode probes (Figure 6B/C). MTEs

demonstrated reduced microglial accumulation, blood-brain barrier disruption, and

neuronal dieback compared to larger silicon planar arrays; however, no statistical

difference were reported. Unfortunately, this study was terminated after 5 weeks.

Figure 6. Summary of findings from Kozai et al. “(A) SEM images of a fully assembled, functional MTE. (B) Percentage of active chronically implanted MTEs (black, N = 7) and silicon probes (red, N = 80 sites, 5 probes) able to detect at least 1 single unit (solid line) or at least 2 single units (dashed line) as a function of weeks post-implantation. (C) Mean SNR of the largest single unit detected on each electrode for MTE (black, N = 7) and silicon probes (red, N = 80). Error bars indicate s.e.m.; * and ** indicate two-tailed, unequal variance statistical significance, p < 0.05 and p < 0.01, respectively.” Figure and caption reproduced and adapted from [5], with permission from Springer Nature.

Later, Patel et al. evaluated the long-term performance of multi-shank arrays of the

composite carbon fiber probes utilized by Kozai et al. [4]. Carbon fiber arrays

demonstrated higher percentages of channels detecting single units and higher mean unit

amplitudes compared to larger NeuroNexus arrays (15 µm x 123 µm cross-section), without any reported statistical differences. (Figure 7A/B). Statistical comparisons may

be confounded by the extremely poor performance of the NeuroNexus arrays, which may

be attributed to brain tissue swelling in the craniotomy [4]. Regardless, carbon fiber arrays

demonstrated single unit recording capability out to 120 days after implantation [4]. The observed improvements in recording performance over NeuroNexus arrays are promising, but the longevity is far from ideal for long-term neural interfacing applications in humans.

Device lifetimes on the order of years to decades, or the remainder of a patient’s life would

eliminate the need for additional highly invasive implantation surgeries. Further, carbon

fiber arrays demonstrated minimal microglia/macrophage presence and astrocytic

encapsulation, with trends of decreased expression relative to NeuroNexus controls and

significant reductions at various distance from the implant [4]. Carbon fiber electrodes also demonstrated significantly higher neuronal survival compared to silicon probes at moderate (~60-110) or far (~250 µm) distances from the implanted probes (Figure

7C/D/E). Strangely, neuronal survival was well above (15x) background density with high variability close to the carbon fiber electrodes [4], indicating potential issues in histology and quantification methods.

Figure 7. Summary of findings from Patel et al. “(A) On average 20%–40% of viable carbon fiber electrodes detected unit activity across time, while silicon electrodes did so with a peak of 9.5% at day 10 and most other days detecting no unit activity. After day 91 only two rats remained in the study and the loss of their unit activity is likely explained by brain tissue swelling into the craniotomy which was discovered post mortem. (B) Carbon fiber electrodes detected an average unit amplitude of 200 μV across three months. Units detected on silicon electrodes had a mean amplitude of 50–100 μV. All values are mean ± standard error of the mean. The exact number of units detected and used for amplitude analysis at each time point can

be seen in figure S3. (C) and (D) NeuN (neuron) staining around the implanted carbon fiber array and silicon electrode in ZCR19. Neural density appears much more diminished around the silicon electrode as compared to the carbon fiber array. (E) Normalized neural density around each electrode type (n = 2 images/electrode type), illustrating the healthy neuronal population surrounding the carbon fiber arrays and a lack of neurons around the silicon electrodes. *indicates significance at p < 0.05.” Figure and caption reproduced and adapted from [4], with permission from the IOP Publishing.

Additionally, Guitchounts et al. reported a microelectrode array with 5 µm carbon

fibers insulated with parylene-C for use in zebra finches [165]. These microelectrode arrays demonstrated stable multiunit recordings for several months, however no comparisons were made in recording performance or endpoint histology with commonly used control probes [165].

Breaking from the trend of ultra-small microelectrode arrays made of carbon fiber,

Luan et al. examined the in vivo integration and recording performance of ultra-small microelectrode arrays composed of SU-8 insulating layers, platinum or gold electrodes and interconnects, and no substrate (Figure 8). Two sizes of the so-called nanoelectronic thread (NET) electrodes were available, with cross-sections of 50 µm x 1 µm for the NET-

50 and 10 µm x 1.5 µm for the NET-10. The NET-50 has 8 electrodes in a linear row

(Figure 8A/C) and the NET-10 has a total of 4 electrodes with two on the front and two on the back (Figure 8B/D). NET electrodes consistently detected spikes on 75% of electrodes and sortable single units on 25% of electrodes for 4 months after implantation

(Figure 8E). Controls were not included for comparison in this experiment. Percentages of channels detecting single units was lower than reported by Patel et al., but it was more consistent and did not exhibit failure over the time course of the experiment [3, 4]. In vivo two photon imaging indicated that minor vascular damage resolved after a month and astrocytes maintained normal morphology and density near the probe. The authors reported normal neuronal density and microglia with resting morphology near the probe without any quantitative histology. Both carbon fiber and SU-8 microelectrodes exhibit 31 minimal or nonexistent foreign body reactions, but the same trends may not persist when devices are scaled up to accommodate BCIs and BMIs with higher degrees of freedom, perhaps via more shanks or multiple arrays. Ultra-small arrays have not been evaluated in functional tasks in non-human primates or humans.

Figure 8. Summary of findings from Luan et al. “Structures of NET neural probes. (A and B) As- fabricated NET-50 and NET-10 probes on substrates. (C and D) Zoom-in views of two electrodes as marked by the dashed boxes in (A) and (B), respectively. Arrows denote “vias.” Arrows denote the probes. Scale bars, 100 mm (A), 50 mm (B, G, and H), and 10 mm (C and D). (E) The number (left) and percentage (right) of electrodes that recorded unit activities (red) and sortable single-unit APs (orange) as a function of time. (F) Average peak-valley amplitude (red) and SNR (blue) of single-unit APs recorded by n = 19 electrodes as a function of time. Error bars indicate the SD.” Figure and caption reproduced and adapted from [3], under the terms of the Creative Commons Attribution-NonCommercial license.

Another major strategy for overcoming biological failure mechanisms is by increasing intracortical microelectrode compliance. As described in Section 2.5, mechanical mismatch between intracortical microelectrode arrays and brain tissue is hypothesized to contribute to the biological failures of intracortical microelectrodes [1,

147, 157, 166-169]. Silicon, a common substrate material for intracortical microelectrode arrays, has a modulus between 130 and 185 GPa, which is much stiffer than brain tissue modulus at 6 kPa. Jeffrey Capadona and colleagues previously developed dynamic materials inspired by sea cucumber dermis that remain stiff during insertion to prevent buckling and soften over time to approach tissue stiffness [155, 170, 171]. A combination

of poly(vinyl acetate) with cellulose nanocrystals with a dry modulus of 5.1 GPa softens

under physiological conditions to 12 MPa due to fluid absorption and disruption of

cellulose nanocrystal interactions, inducing phase transition [1, 171]. Harris et al.

evaluated endpoint histology of nanocomposite probes implanted into rats and observed

significantly improved neuronal survival around the probes compared to stiffer microwire

arrays at four weeks after implantation [155]. In a follow-up study, Nguyen et al. observed enhanced neuronal survival around nanocomposite probes at moderate and far distances from the probe at 4 and 8 weeks after implantation and within 50 µm of the probe at 16 weeks after implantation [6]. Neuronal survival is most critical within the first 50 microns from the probe for single unit recordings [117]. Additionally, Nguyen et al. observed decreased astrocytic encapsulation, microglial activation, and blood-brain barrier permeability at 4, 8, and 16 weeks after implantation (Figure 9) [6]. Further, release of the anti-oxidant resveratrol from nanocomposite probes resulted in increased neuronal survival close to the probe and decreased microglial activation compared to nanocomposites without drug release, improving short term tissue integration. Lastly, in- vivo force measurements revealed that the nanocomposite probes exhibited reduced strain, strain rates, and micromotion induced stresses on brain tissue compared to stiff silicon probes [172]. Despite the promising improvements in tissue integration, comparison of recording performance between nanocomposite probes and commonly used stiffer arrays has not been published.

Figure 9. Major findings from Nguyen et al. “Immunohistochemical analysis of neuronal nuclei (NeuN) around the implant site. (A, B) Representative fluorescence microscopy images of stained tissue show that neuronal dieback around the stiff PVAc-coated silicon implant (A) was significantly higher than in case of the compliant nanocomposite implant (B) at 16 weeks. (C) The bar graphs show quantification of neuron densities. Statistical analysis identified several regions with significantly different neuron populations, which varied between time points. * Denotes significance between stiff and compliant samples; # Denotes significance between noted implant and age-matched sham control (p < 0.05). Scale bar = 100 μm. The horizontal dashed line represents the 100% neuron level as determined by quantification of age-matched sham animals. Error bars represent standard error.” Figure and caption reproduced from [6], with permission from IOP Publishing.

Aside from the NET electrodes described by Luan et al., which exhibit reduced

bending stiffness and tissue displacement [3], long term recordings from compliant

intracortical microelectrodes have been under characterized in literature. However, Simon

et al. recently demonstrated that shape memory polymer (SMP) intracortical microelectrode arrays with tunable moduli can detect single units for at least 77 days after implantation in rats (Figure 10) [7]. The SMP electrodes (Figure 10A-C), developed by the Voit lab, are variations of thiol-ene/acrylate substrates that have tunable ranges of

Young’s modulus. Similar to the nanocomposite array, SMPs change modulus after insertion into tissue due to heating of the SMP substrate by the tissue (Figure 10D), to above the glass transition temperature. The SMPs typically have a Young’s modulus around 1.8 GPa or shear modulus of ~600MPa prior to insertion and a shear modulus of 30

MPa after dwelling in tissue. So far the long-term recording performance of only two rats implanted with SMPs have been published (Figure 10E). Studies comparing the recording

performance and histology of SMPs and stiffer electrodes involving larger cohorts of

animals remain to be published, but are underway in collaborations between the Voit,

Pancrazio and Capadona laboratories. Additionally, Sohal et al. combined flexible probe materials with sinusoidal architecture to accommodate motion of the brain and decouple the recording contacts from the tethered end of the probe (Figure 11) [8]. The sinusoidal probe substrate was parylene-C, so insertion was aided with hypodermic needle guides adhered with dissolvable PEG (Figure 11). Stable high voltage spindle recordings were obtained in rabbits out to 678 days after implantation [8]. However, that duration of recording was only observed on one probe out of 20-25 electrodes (4-5 electrodes per animal in 5 rabbits, 3 recording sites per electrode). A sample of single unit activity was displayed however no long-term characteristics of single unit recording quality were described or compared to controls. Comparisons of LFP and high voltage spindle responses between sinusoidal electrodes and microwires were made on one rabbit each.

Histological evaluation revealed reduced astrocytic encapsulation relative to microwires at

6 months and 24 months after implantation, reduced microglial activation and 12 and 24

months after implantation, and enhanced neurofilament expression at 12 months after

implantation [8]. However, promising histological results do not always translate to

consistent recordings. These promising results indicate the detrimental effects of

micromotion and tethering forces play a role over extended periods of time.

Figure 10. Major findings of Simon et al. “Fabrication and packaging of the novel SMP‐based intracortical probe using 47 mol % TCMDA [Tricyclo[5.2.1.02,6]decanedimethanol diacrylate]. Scanning electron microscopy image of the thin film layers of Au and parylene‐C created using photolithography to produce traces and recording sites. To reduce the impedance of each microelectrode, a layer of PEDOT:PSS was deposited on each site. (b) Photomicrograph of the penetrating portion of the approximately 2.5 mm long intracortical probe showing several of the 16 evenly spaced microelectrode sites along the lower 1.5 mm region of the device. (c) Image showing the custom circuit board designed and fabricated to interface the SMP probe with the head stage of recording system (d) Thermomechanical characterization of the thiol‐ ene/acrylate polymer compositions. Comparison for 31 and 47% TCMDA of the storage modulus (e) Novel intracortical probe performance in vivo over 11 weeks of implantation. With the exception of a brief period within the first 2 weeks of implantation, both of the animals demonstrated active single units over the duration of the study.” Figures and captions reproduced from [7] with permission from John Wiley and Sons.

Figure 11. “Sinusoidal probe microfabrication and surgery preparation.” “(A–F) Sequential microfabrication steps for the sinusoidal probe concentrating on the recording end (illustrative and not to scale): (A) 1 μm aluminum beam deposition as a sacrificial layer on a silicon substrate. (B) Parylene-C deposition through a CVD [chemical vapor deposition] process. (C) WTi sputter deposition and patterning to define electrode tracks, bondpad and recording site regions. (D) Second layer parylene-C deposition through CVD. (E) Patterning of both parylene-C layers using oxygen plasma. (F) Device release using a TMAH [Tetramethylammoniahydroxide] based photoresist developer and addition of 3D polyimide ball anchor. (G) Mask layout for the 3 mm version of the probe. Three metal tracks ran within an electrode shaft that was 20 μm deep, 35 μm wide and had a sinusoidal profile with cycles of 100 μm amplitude and 500 μm period. (H) Successful device release in TMAH. (I) Optical microscopy image of the recording end showing three separate and isolated electrode recording sites, pre-polyimide anchor addition. The protruding electrode recording sites may appear thicker due to the underlying first parylene-C layer, which is not removed during WTi etching. (J) Attached sinusoidal probe to improvised insertion carrier to allow for successful brain penetration and electrode insertion. (K) Electrodes under sterile conditions, placed in an autoclaved container and then into surgical bag primed for UV light sterilization.” Figure and caption reproduced from [8] with under the terms of the Creative Commons Attribution License.

Recently, other labs have assessed the tissue integration of other probes made from other compliant materials. Du et al. investigated the tissue integration of soft polymer wires assembled from PEDOT-PEG (Poly(3,4-ethylenedioxythiophene)-polyethylene

glycol) conducting polymer and polydimethylsiloxane with gold sputter-coating and a dip- coated fluorosilicone insulating layer [173]. The Young’s modulus of the soft conduction wires was 974 kPa and the cross-sectional area was 12,270 µm2 [173]. Since the soft polymer wires did not exhibit any dynamic softening properties, they relied on stiff insertion shuttles constructed from 27G needles adhered to the wires penetrate the cortex without buckling (Figure 12) [173]. Soft polymer wires were adhered to the shuttles using

PEG, which dissolved after approximately 30 seconds, upon which the shuttles were

removed. Soft polymer wires exhibited decreases in microglia/macrophage expression and

astrocytic encapsulation relative to stiffer tungsten wires 8 weeks after implantation, but

not at 1 week after implantation [173]. Axonal density was actually lower around soft polymer wires at 1 week after implantation with no differences at 8 weeks after implantation [173]. Neuronal density was significantly higher around soft polymer wires at 60-90 and 120-210 µm from the probe interface at 8 weeks after implantation (Figure

12), with no differences at 1 week after implantation [173]. This is a similar trend in

neuronal survival as observed in Nguyen et al., where neuronal densities were similar close

to hole but significantly higher around soft implants further from the hole (100-300 µm) at

8 weeks after implantation, with no difference at the earliest time point (2 weeks).

Additionally, blood-brain barrier permeability was decreased around soft polymeric wires

at 8 weeks after implantation (Figure 12). Electrode function was verified with impedance

spectroscopy and stimulation of the STN. Although these soft polymeric wires were not

used for recording, lessons learned about the tissue response may be beneficial to

improving intracortical microelectrodes.

Figure 12. Major findings from Du et al. “Soft polymer wires implanted deep into the brain via a PEG- anchored insertion shuttle exhibited improved tissue integration, including significantly improved neuronal survival and significantly reduced blood-brain barrier permeability.” Figure reproduced from [173] with permission from Elsevier.

Bio-inspired coatings have provided interesting solutions to biological failure

mechanisms. Eles et al. examined the effects of an L1 cell adhesion molecule on

intracortical microelectrodes using two photon imaging [174]. The L1 cell adhesion

molecule is a transmembrane glycoprotein that selectively promotes neuronal cell

attachment [174-176]. Studies with L1 coatings on intracortical microelectrodes have

demonstrated enhanced neuronal attachment and attenuated glial scarring over an 8 week

time span [177, 178]. Eles et al. observed that L1-coated NeuroNexus planar arrays

demonstrated significantly reduced probe coverage with microglia and reduced radius

microglial activation via in vivo two photon imaging at six hours after implantation (Figure

13) [174]. Included in another study by the Cui lab investigating impedance signatures of

encapsulation types, the in vivo recording performance of Utah arrays coated with the L1

cellular adhesion molecule was evaluated [9]. Unfortunately, L1 coating only conferred

benefits in recording performance on the first week after implantation (Figure 14),

specifically exhibiting higher single unit SNR amplitude ratio (Figure 14C) [9]. Although a group-wise significant effect of coating treatment was observed for multi-unit yield and

39 single-unit SNR, but this was not interpreted as an improvement or detriment (Figure

14A/C). The benefits from acute microglial coverage and activate time points does not appear extend beyond very acute time points. Degradation of the coating within the first day or so would explain a return to baseline performance. Despite promising histological and in vivo imaging findings, biological failure mitigations strategies may not translate to recording performance.

Figure 13. Major findings from Eles et al. “L1 coating continues to prevent microglial surface coverage through 6 h post-implant. A) 2D projections of probes at 2 and 6 h post-implant B) The percentage of the probe's surface that was covered by microglia was significantly less for L1-coated probes compared to control probes at 2 h and 6 h post-implant (Welch's T-test; **p < 0.001, *p < 0.05). Bar graph data presented as mean ± SEM.” Figure and caption reproduced from [174] with permission from Elsevier.

Figure 14. Major findings from Cody et al. “Electrophysiological performance of L1 coated arrays (blue) compared to uncoated control arrays (black dashed). Average evoked multi-unit yield (a), single-unit yield (b), and single-unit signal-to-noise amplitude ratio (c) between treatment groups from weekly recordings over a 12-week period. Brackets indicate significant overall (group-wise) treatment effects with p = 0.0114 and p < 0.0001 for a and c respectively. * indicates significant difference with p < 0.001 from Bonferroni corrected pair-wise test.” Figure and caption reproduced from [9] with permission from Elsevier.

Another bio-inspired coating investigated by Oakes et al. was extracellular matrix

(ECM) derived from astrocytes (Figure 15A-D) [10]. The ECM coating was chosen for its capability to promote hemostasis and pro-regenerative activation of macrophages [10,

179-187]. The astrocyte ECM coating and a commercially available decellularized ECM coating called Avitene significantly decreased clotting time in serum, but only the astrocyte

ECM coating significantly increased the percentage of ramified microglia [10]. Eight weeks after implantation, GFAP expression was significantly reduced in the astrocyte

ECM coated electrodes (Figure 15E-G) but not Avitene coated electrodes. Also at the 8- week time point, blood-brain barrier permeability was significantly increased around

Avitene-coated electrodes and trending higher in astrocyte ECM coated electrodes.

Neither coating affected microglial activation or neuronal dieback in vivo. Promising in

vitro results may not always even translate to better histological results.

Figure 15. Major findings of Oakes et al. “Representative images of: (A) a clean, sterile silicon planar microelectrode array; (B) a stereoscopic light micrograph of an astrocyte-derived ECM coated microelectrode; (C) an Avitene-MCH coated microelectrode array immunolabeled for collagen type I. Scale bar 100 μm; and, (D) a higher magnification image of C. Scale bar 10 μm. (E) GFAP immunofluorescence as a function of distance from the electrode/biotic interface in 50 μm bins compared to uncoated controls 8 weeks after implantation (n = 5). (*) denotes significant difference form controls at a p < 0.05. Data shown as mean and standard error. (F, G) Representative images of horizontal sections 8 weeks after implantation showing immunoreactivity for the intermediate filament GFAP found in astrocytes in the cortex layers IV- VI surrounding (F) uncoated microelectrodes compared to (G) astrocyte-derived ECM coated microelectrodes. The spatial distribution and intensity of GFAP was similar when comparing uncoated electrodes and those coated with Avitene (data not shown). Scale bar 100 μm.” Figure and captions reproduced and adapted from [10] with permission of Elsevier.

Further, antioxidant therapies have demonstrated varying degrees of success in the integration of intracortical microelectrodes into the brain. As mentioned in Section 2.5, oxidative factors released by microglia and macrophages are hypothesized to contribute to chronic inflammation, blood-brain barrier permeability, neuronal death, and electrode material damage following intracortical microelectrode implantation [11, 12]. More so, evidence of enhanced oxidative stress gene expression and oxidative cellular damage has been observed at the electrode-tissue interface [30]. Systemic administration of the antioxidant resveratrol before and after implantation resulted in enhanced neuronal survival around implanted intracortical microelectrodes at two weeks after implantation (Figure

16) [11]. Additionally, qualitative observation of fluorojade-c expression indicated reduced neurodegeneration in rats administered resveratrol. Further, microglia activation, microglial expression, and cellular expression were all upregulated in rats receiving resveratrol, but astrocytic encapsulation was decreased at four weeks after implantation.

Also, decreased blood-brain barrier permeability was observed around rats receiving resveratrol at two weeks after implantation, suggesting a link between neurodegeneration and blood-brain barrier permeability. Since the benefits of resveratrol administration likely wore off after the two-week time point, daily systemic administration of resveratrol was also investigation. Potter-Baker et al. demonstrated that daily systemic administration of resveratrol did not improve neuronal survival over no injection and diluent controls [12].

However, qualitative reductions in neurodegeneration were observed. Additionally, an increase in blood-brain barrier permeability was exhibited two weeks after implantation.

The most troubling aspect of chronic daily resveratrol administration was the observation of liver damage [12]. Localized delivery of antioxidants may mitigate the detrimental

effects of systemic administration. Release of resveratrol from nanocomposite probes

resulted in enhanced neuronal survival and decreased microglia/macrophage activation

around the probes at two weeks after implantation, relative to control nanocomposites

[188]. Switching to another antioxidant, Potter-Baker et al. demonstrated that localized

release of curcumin from poly (vinyl alcohol) probes resulted in enhanced neuronal

survival near the probe at two and four weeks after implantation and significantly lower

neuronal survival further from the implant at 16 weeks after implantation [189]. Release of curcumin from probes decreased blood-brain barrier permeability at two and four weeks after implantation [189]. Astrocytic encapsulation was increased at two weeks and decreased at 4 weeks after implantation in rats with curcumin-releasing probes [189].

Finally, microglial activation was decreased at 4 weeks after implantation in rats with

curcumin-releasing implants [189]. The localized release of curcumin resulted in early

improvements in microelectrode integration that wore off over time, presumably when the

drug was consumed. To extend the effectiveness of antioxidant effects without the harmful

effects of chronic systemic administration, anti-oxidative enzymes were attached to probes

[190]. Surfaces coated with the superoxide dismutase mimetic MnTBAP attenuated

reactive oxygen species release from microglia as well as cytotoxicity in vitro [190]. In

vivo implementation of MnTBAP coated intracortical microelectrode probes has yet to be

published. There is much evidence that oxidative stress affects intracortical microelectrode

integration, but delivery of antioxidant treatments still remains a challenge.

Figure 16. Major findings from Potter et al. “Effect of resveratrol administration on neuronal nuclei densities after microelectrode implantation. Neuronal nuclei (NeuN) density was investigated two and four weeks following resveratrol administration/microelectrode implantation. Here, resveratrol dosed animals had significant increases in neuronal density up to 100 μm away from the device interface in comparison to control animals (♯p < 0.006), two weeks after implantation (A–B, E). For all binning intervals at two weeks post-implantation, control animals had significant (*p < 0.04) decreases in neuronal density in comparison to the background neuronal density of non-surgery age-matched controls. In addition, at two weeks, neuronal densities in resveratrol dosed animals were significant (**p < 0.04) from non-surgical controls until 100 μm away from the interface of the device.” Figure and caption reproduced from [11] with permission from Elsevier.

Overall, strategies mitigating the biological failure mechanism have demonstrated varying degrees of success in improving intracortical microelectrode performance and tissue integration. Promising in vitro and histological results do not always translate to improved recording performance, let alone success in functional tasks. Many of the strategies exhibited time variant effectiveness possibly stemming from delivery challenges and the dynamic nature of the foreign body response. Strategies to overcome biological failure mechanisms will likely benefit from combinations of strategies with time-variant application.

2.7 Studies linking neuroinflammation to electrode failure

Despite all the efforts to overcome biological intracortical failure mechanisms, only a handful of studies have directly compared neuroinflammation to electrode performance.

Interestingly, these studies may be linked to the innate immunity, the body’s fast-acting

response to pathogenic threats. Here we briefly discuss these studies and their links to

innate immunity.

Rennaker et al. investigated the effects of minocycline, chosen for its

neuroprotective and neurorestorative effects, on intracortical microelectrode recording

performance [13]. Rats administered minocycline via water two days before and 5 days

after surgery exhibited improved recording performance over controls after the first week

of implantation, upon which the SNR of controls steadily dropped (Figure 17) [13].

Endpoint histological analyses revealed that astrocytic encapsulation was decreased at 1-

and 4-week time points after implantation (Figure 17) [13]. Improved recording performance was speculated to be caused by decreased inflammation, neuronal dieback, and microglia activation in addition to the observed decrease in astrocyte encapsulation.

Minocycline was later shown to inhibit the pro-inflammatory phenotype of macrophages

[191]. Alternative explanations could factor in the antibiotic activity of minocycline.

Perhaps a lower bacterial load activated less inflammation via innate immunity pathways.

Regardless of the mechanism of action, minocycline is not suitable for long-term administration due to detrimental side effects [15, 16].

Figure 17. Major findings from Rennaker et al. “(A) SNR data. There were no significant differences between the groups during the first 6 days. On day 7, the control group SNR decreased significantly, while the minocycline group did not change. Error bars are the 95% confidence interval. ( p < 0.05, p < 0.001).” (B) “Quantitative histology. Using automated image analysis software (ImageJ), the area occupied by each cell, the total number of activated cells and the total area occupied by activated∗ astrocytes∗∗∗ were measured. The 1 week data are presented in the left column and the 4 week data are in the right column. At both 1 and 4 weeks, the control group exhibited an increase in the total number of activated astrocytes and the total area occupied by these cells. However, the activated astrocyte size was larger in the minocycline group at week 1 only. Error bars are 95% confidence interval. ( p < 0.05; p < 0.001).” Figure and caption reproduced from [13] with permission from the IOP Publishing. ∗ ∗∗∗ In contrast to the previous study administering an anti-inflammatory compound,

Harris et al. examined the effects of a pro-inflammatory compound on intracortical microelectrode performance in an un-published thesis study [27]. At the end of a four- week study, rats administered a one-time surgical dose of the pro-inflammatory agent lipopolysaccharide (LPS) exhibited significantly lower neuronal density within the first 50 microns away from the implant (Figure 18) [27]. Additionally, administration of LPS significantly reduced firing rates in evoked neural recordings (Figure 18) [27]. Thus, inflammation is associated with poor neuronal survival and recording performance.

Interestingly, the pro-inflammatory agent LPS is derived from the bacterial cell walls of gram-negative bacteria, and is predominantly recognized by the innate immunity receptors cluster of differentiation 14 (CD14) and Toll-like receptor 4 (TLR4) to induce robust

46 inflammatory responses [192]. This indicates that activation of innate immunity can result in detrimental effects in recording performance and tissue integration.

Figure 18. Major findings from Harris et al. “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.” (B) “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.” Figures and captions reproduced from [27], an open access ETD published by Case Western Reserve University and OhioLINK.

Further, Saxena et al. examined the effects of blood-brain barrier permeability on recording performance [136]. In this study, extravasation of blood proteins and myeloid cells into the brain parenchyma around intracortical microelectrodes implanted in rats coincided with poor recording performance [136]. Saxena et al. hypothesized that chronic blood-brain barrier permeability following intracortical microelectrode implantation was part of a positive feedback look with chronic inflammation, resulting in neurodegeneration

(Figure 19) [136]. One mechanism by which microglia and macrophages may recognize blood proteins and promote inflammation is through the innate immunity receptor Toll- like receptors [20, 193]. Thus, innate immunity may play a role in the positive feedback

47 loop of blood-brain barrier permeability and chronic inflammation that leads to intracortical microelectrode failure.

Figure 19. “Schematic illustration and the working model of chronic electrode failure.” “Intracortical electrodes induce a chronic BBB breach, which leads to the extravasation of neurotoxic serum proteins, and an infiltration of myeloid cells. Reactive gliosis and myeloid cell activation occurs in a bimodal manner, leading to the production of neurotoxic and proinflammatory cytokines. These cytokines increase BBB permeability, and cause chronic inflammation resulting in a positive feedback loop, that leads to neurodegeneration and loss of chronic electrode function.” Figure and caption reproduced from [136] with permission from Elsevier.

Finally, Kozai et al. investigated the role of caspase-1 in the recording performance of intracortical microelectrodes [107]. Caspase-1 was studied for its role in the activation of the pro-inflammatory cytokine IL-1β, as well as its role in neuronal death related to ischemia and chronic neurodegeneration [107]. Knockout mice lacking caspase-1 implanted with intracortical microelectrodes exhibited significantly improved single unit

yields over wildtype controls out to ~150 days after implantation with similar yields

extending to the end of the experiment around 180 days after implantation (Figure 20)

[105]. Additionally single unit single to noise ratios were mostly significantly higher in

caspase knockout mice out to around 140 days after implantation (Figure 20) [105].

Coincidentally, caspase-1 is involved in several innate immunity mechanisms, including

inflammasome, RIG-like receptor, and TLR signaling [105]. In this instance, attenuation

of innate immunity improved chronic intracortical microelectrode performance. Thus,

innate immunity appears to play a role in the failure of intracortical microelectrodes.

Figure 20. Major findings of Kozai et al. “Chronic electrode performance comparison between caspase-1 KO (blue dashed) and WT (black solid) mice. * indicates p < 0.05. (A) Single-unit yield over days. (B) Single-unit SNR (mean amplitude over 2*STD noise).” Figures and captions reproduced from [107] with permission from Elsevier.

Overall, studies linking inflammation to intracortical microelectrode performance reveal several connections by which innate immunity may be involved in intracortical microelectrode failures.

2.8 Innate Immunity

As indicated in the previous section, innate immunity has ties to the foreign body response to intracortical microelectrodes. Thus, further understanding of the innate immunity may inform strategies to mitigate intracortical microelectrode failures.

Innate immunity is often described as the body’s first line of defense against pathogenic threats [194]. Innate immunity should be distinguished from the body’s other defense system, adaptive immunity, by several characteristics. Innate immunity activates much more quickly, responding within minutes or hours, as compared to several days [194,

195]. The quickness in responses is, in part, due to the widely distributed germline encoded effectors [195]. In contrast the adaptive immunity requires genetic recombination and clonal expansion of specific effectors [194, 195]. The disadvantage of germline encoded effectors is that innate immunity has less diversity and specificity of responses [195].

However, the recognition of general patterns conserved among classes of pathogens allows the innate immunity to be versatile and have capable effector cells distributed throughout the body [194, 195]. In further contrast, the adaptive immunity features memory, meaning that adaptive responses to a previously encountered threat may be activated more quickly and with greater intensity [195]. Innate immunity is not traditionally thought to feature memory, so innate immune responses enact with a relatively consistent activation time and intensity. However, there is growing evidence that innate immunity may feature some degree of memory [196]. The innate and adaptive immunity are not completely

independent, as the innate immunity often aids in the activation and regulation of adaptive immunity effectors [195].

Innate immunity is comprised of physical barriers, chemical barriers, and cellular

responses [195]. Physical barriers include epithelial layers, mucosal tissues, and glandular

tissues [195]. Chemical barriers include acidic fluids, anti-microbial proteins, and anti-

microbial peptides that reside near the physical barriers [195]. Cellular innate immune

responses are typically mediated by phagocytic cells, such as macrophages, neutrophils,

dendritic cells, monocytes, and microglia, but also may involve natural killer cells,

leukocytes, epithelial cells, and endothelial cells [195]. Additionally, complement

glycoproteins found in serum are also included in innate immunity [195].

In the event that pathogens bypass the physical and chemical barriers of the body,

cellular effectors may address the threat. Cellular effectors of the innate immunity are

generally activated via pattern recognition receptors (PRRs), which recognize molecular

patterns common to categories of pathogens referred to as pathogen associated molecular

patterns (PAMPs). Some pattern recognition receptors may also recognize molecular

patterns on endogenous molecules released by the body called damage (or danger)

associated molecular patterns (DAMPs) [195]. The family of PRRs employed by the innate

immunity are Toll-like receptors (TLRs) [197], C-type lectin receptors [198], Retinoic

acid-inducible gene-I-like receptors (RLRs) [199, 200], and Nod-like receptors (NLRs)

[201, 202]. The TLRs and CLRs are transmembrane proteins expressed across plasma membranes (Figure 21), however TLRs may also be expressed on endosomes and

lysosomes (Figure 21) [195]. In contrast, RLRs and NLRs are expressed in the cytosol

(Figure 21) [195]. Both TLRs and NLRs have demonstrated recognition of DAMPs in

addition PAMPs [195]. Activation of the various PRRs may result in pro-inflammatory activation, apoptosis, phagocytosis, coagulation cascades, opsonization, or complement activation [194].

Figure 21. Cellular localization of innate immunity of innate immunity receptors. Toll-like receptors (TLRs) and C-type lectin receptors (CLRs) are expressed across plasma membranes with domains located in the extracellular space and cytosol. Some TLRs are expressed across the membranes of endsomes and lysosomes, with domains located inside the endosome/lysosome and in the cytosol. Nod-like receptors (NLRs) and Retinoic acid-inducible gene-I-like receptors are expressed in the cytosol [195, 203]. Modified from [203], with permission to reproduce figure from The American Society for Microbiology.

In the context of foreign body responses to implanted intracortical microelectrodes,

TLRs may be of the most relevance. Some TLRs are expressed on plasma membranes

[204, 205], so they may be able to respond to external threats directly. The TLRs also have

the capability to recognize DAMPs and PAMPs [20, 206], enabling the potential detection of tissue, cellular, and vascular damage associated with intracortical microelectrode

implantation, in addition to pathogens introduced with the microelectrode (Figure 22).

Thus, Section 2.9 focuses on TLRs and their adaptor molecule CD14.

Figure 22. Potential role of innate immunity pattern recognition receptors in the foreign body response to intracortical microelectrodes. Due to the capability to recognize both pathogen associated molecular patterns (PAMPs) and damage associated molecular patterns (DAMPs) and initiate inflammatory responses, as well as expression on cell membranes, toll-like receptors (TLRs) are the innate immunity pattern recognition receptors most likely to be involved in the foreign body response to intracortical microelectrodes. Zoomed in portion modified from [203], with permission to reproduce figure from The American Society for Microbiology.

2.9 Toll-like Receptors and CD14

One major group of effectors in innate immunity is a class of pattern-recognition receptors (PRRs) called Toll-like receptors (TLRs) [19]. The TLRs are transmembrane proteins that recognize molecular patterns associated with either pathogens or tissue damage and enact downstream inflammatory activation [19, 207]. The TLRs are expressed on peripheral immune cells, such as macrophages and dendritic cells, as well as several cells of the central nervous system, including microglia, astrocytes, and neurons [19, 208,

209].

Toll-like receptors are named for their similarity to Toll [210, 211], a protein responsible for directing dorsal-ventral patterning in Drosophila embryos [212] and involved in innate immunity in adult Drosophila [213]. The TLRs were later identified in humans based on their shared molecular structure and downstream signaling pathways

[214].

Toll-like receptors are type 1 transmembrane proteins that feature leucine-rich repeat (LRR) domains on the extracellular side and a Toll/IL-1 receptor (TIR) domain [19,

210, 214, 215]. The LRRs are sequence motifs with frequent interspersed instances of the amino acid leucine, and are found on many proteins involved in protein-protein interactions such as signal transduction [19, 216]. The other major component of Toll-like receptors, the TIR domain, is a protein-protein interaction module found involved in the host responses of both plants and animals [19, 217].

Building off the common structural elements of LRR and TIR domains, the TLR family gives rise to at least 12 different functional members in mice and 10 different functional members in humans [19, 197, 199]. The TLRs 1-9 are functional in both mice and humans, TLR10 is functional in humans but not mice [218], and TLRs 11-13 have been identified in mice but not humans [197]. The members of the TLR family vary in membrane location and ligand specificity [19]. Some TLRs, such as TLR1, 2, 4-6, 8, and

11, are located on the cell membrane, with the LRR domain facing the extracellular environment and the TIR domain facing the cytosol [197, 219, 220]. With an outward facing ligand binding domain, TLRs on the cell membrane monitor for external threats like bacteria. Other TLRs, such as TLR3, 7, 8, and 9, are located on intracellular vesicles, with the LRR domains facing the interior of the vesicle and the TIR domain facing the cytosol

[197, 219, 220]. With a vesicle facing ligand binding domain, TLRs on vesicle membranes

monitor for internal threats, such as viruses. Each TLR recognizes its own set of ligands,

which may be pathogen associated molecular patterns (PAMPs) or damage/danger

associated molecular patterns (DAMPs).

The PAMPs recognized by TLRs and other PRRs are molecular patterns shared amongst broad categories of pathogens, allowing rapid and versatile innate immune responses with a handful of widely-available receptor types [19]. The broad, fast-acting nature of TLRs contrasts against the highly specialized receptors and antibodies of the adaptive immunity, which require time-consuming clonal expansion to distribute effector cells to combat pathogens [19]. The PAMPs are effective targets for immune recognition because they occur in pathogens but not host cells, allowing inflammatory effector cells to discriminate between self and non-self [19]. Further, PAMPs are typically molecular patterns involved in critical pathogenic survival mechanisms [19]. Mutations in pathogens removing PAMPs are typically deadly, allowing TLR signaling to remain effective across many generations of evolution [19].

The PAMPs can be utilized in the recognition of bacterial, fungal, parasitic, and viral pathogens (Table 2) [221]. Bacterial PAMPs include the gram negative bacterial cell wall component lipopolysaccharide (LPS) [222], the gram positive bacterial cell wall components lipoteichoic acid and peptidoglycan [223-227], diacyl lipopeptides [225, 226], triacyl lipopeptides [225, 226], flagellum protein flagellin [228], and unmethylated CpG motifs (cysteine triphosphate deoxynucleotide with a phosphodiester link to a guanine triphosphate deoxynucleotide) of bacterial DNA [221, 229]. The status of peptidoglycan as a PAMP has been disputed due to concerns over bacterial contamination in

peptidoglycan samples [230], however it remains to be used as a TLR2 agonist [24]. As

stated earlier, bacterial PAMPs are contained in features not found in mammalian cells,

such as cell walls, flagella, and unmethylated CpG motifs to enable discrimination between

self and non-self. Fungal PAMPs include the yeast cell wall component zymosan [231],

the yeast cell wall surface glycolipid phospholipomannan [232], fungal cell wall polysaccharide mannan [233], and yeast capsular polysaccharide glucuronoxylomannan

[221, 234], which are critical fungal components not found in mammalian cells. However,

it is interesting to note that glucuronoxylomannan demonstrated TLR2 and TLR4 binding

without cytokine release [234]. Parasitic PAMPs include the protozoan glycoprotein tGPI-

mucin [235], protozoan gycoinositolphospholipids [236], the malaria parasite heme

polymer hemozoin [237], and the protozoan actin-binding protein homolog profilin-like

molecule [221, 238]. Viral PAMPs include DNA [239-243], double-stranded RNA [244],

single-stranded RNA [245, 246], envelope proteins [247], and hemagglutinin protein [221,

248, 249]. Viral PAMPs located on nucleic acids are typically recognized in endosomes by TLR3, 7, 8, and 9, whereas viral PAMPs located on proteins are recognized by cell- surface TLRs [221].

Table 2. Summary of major pathogen associated molecular patterns (PAMPs) recognized by toll-like receptors (TLRs). Modified from [221], with permission to reproduce table from Elsevier. PAMP Pathogen TLR Bacteria LPS Gram-negative bacteria TLR4 Diacyl lipopeptides Mycoplasma TLR6/TLR2 Triacyl lipopeptides Bacteria and mycobacteria TLR1/TLR2 LTA Group B Streptococcus TLR6/TLR2 PG Gram-positive bacteria TLR2 Porins Neisseria TLR2 Lipoarabinomannan Mycobacteria TLR2 Flagellin Flagellated bacteria TLR5 CpG-DNA Bacteria and mycobacteria TLR9 ND Uropathogenic bacteria TLR11

Fungus Zymosan Saccharomyces cerevisiae TLR6/TLR2

Phospholipomannan Candida albicans TLR2 Mannan Candida albicans TLR4 Glucuronoxylomannan Cryptococcus neoformans TLR2 and TLR4

Parasites TLR2 tGPI-mucin Trypanosoma Glycoinositolphospholipids Trypanosoma TLR4 Hemozoin Plasmodium TLR9 Profilin-like molecule Toxoplasma gondii TLR11

Viruses DNA Viruses TLR9 dsRNA Viruses TLR3 ssRNA RNA viruses TLR7 and TLR8 Envelope proteins RSV, MMTV TLR4 Hemagglutinin protein Measles virus TLR2 ND HCMV, HSV1 TLR2

ND = not determined.

As opposed to PAMPs, DAMPs are patterns found in endogenous molecules

released in the event of non-infectious tissue damage (Table 3) from [20, 206, 207]. The

DAMPs recognized by TLRs include heat shock proteins (HSP), extracellular matrix

(ECM) components, components of necrotic cells, surfactant protein A, RNA, DNA, and fibrinogen [20, 207]. Heat shock proteins are molecules released when cells are exposed to various stresses [20, 206, 250, 251], and TLRs can recognize HSP22 [252], HSP60 [253-

255], HSP70 [256, 257], and Gp96 [258]. Molecules may be cleaved from the ECM by

proteolytic enzymes during the inflammatory response to traumatic injury [20], and TLRs can recognize heparan sulfate [259-261], hyaluronan-derived oligosaccharide [262, 263], biglycan [264], and fibronectin extra domain A [265]. Necrotic cells promote inflammation via TLR recognition of the released chromatin binding protein high mobility group box 1 (HMGB1) [145, 266-269] or unidentified ligands [270, 271]. Surfactant protein A is a component of pulmonary surfactant and may activate TLRs [272]. Host

RNA [273-275] and DNA [273, 276] can activate TLRs, but the method of endosome

57 localization is unclear [273]. Fibrinogen escapes vasculature during inflammation and can bind TLRs to further propagate inflammatory mechanisms [193].

Table 3. Summary of major damage associated molecular patterns (DAMPs) recognized by toll-like receptors (TLRs). Modified from [20] and updated from [206, 207] and individual studies (see References), with permission to reproduce table from John Wiley and Sons. DAMP TLR Factors produced References

Heat shock proteins (HSPs) HSP60 TLR2, TLR4 TNF, NO [253-255] HSP70 TLR2, TLR4 TNF, IL-1β, IL-6, IL-12 [256, 257] Gp96 TLR2, TLR4 TNF, IL-12 [258] HSP22 TLR4 TNF, IL-6, IL-12 [252]

Extracellular matrix components Heparan sulfate TLR4 TNF [259-261] Hyaluronan-derived TLR2, TLR4 TNF, MIP-1α, MIP-2, KC [262, 263] oligosaccharide Biglycan TLR2, TLR4 TNF, MIP-2 [264] Fibronectin extra domain A TLR4 MMP-9 [265] Decorin TLR2, TLR4 PDCD4, IL-10 [277] Versican TLR2, TLR6 TNF-α, IL-6 [278]

Necrotic cells TLR2, TLR3 TNF, IL-8, MIP-2, KC, MMP-3, [270, 271] iNOS

RNA TLR3, TLR7 IL-12, IFN-α [273, 274]

DNA (in the form of immune TLR9 IFN-α [273, 276] complexes)

High mobility group box 1 TLR2, TLR4 TNF, IL-1α, IL-1β, IL-6, IL-8, MIP- [145, 266- protein (HMGB1) 1α, MIP-1β, COX-2, iNOS 269]

Lung surfactant protein A TLR4 TNF-α, IL-10 [272]

Fibrinogen TLR4 MIP-1α, MIP-1β, MIP-2, MCP-1 [193]

S100A8 TLR4 TNF-α [279]

S100A9 TLR4 IL-1β, TNF-α, IL-6, IL-8 and IL-10 [280]

Fibrillar β-amyloid TLR2, TLR4 Superoxide radical [26]

α-synuclein TLR4 TNF-α, IL-6, CXCL1, ROS [281]

Upon binding of a PAMP or DAMP to the sequence of LRR domains on a TLR, cytoplasmic signaling events occur (Figure 23, for a more detailed review see [282]).

Communication to the interior of the cell is facilitated through the TIR domain. TLRs

dimerize after ligand binding, conformational changes occur, and myeloid differentiation

primary-response protein 88 (MyD88) is recruited to TIR domain of the TLR complex.

The intracellular protein MyD88 interacts with members of the interleukin1-receptor- associated kinase family (IRAK), which in turn activate tumor-necrosis-factor-receptor- associated factor 6 (TRAF6). IRAK1 and TRAF6 join a complex with transforming growth factor-β-activated kinase 1 (TAK1), TAK1-binding protein 1 (TAB1), and TAB2, which activates both mitogen-activated protein (MAP) kinases and the inhibitor of nuclear factor-κB-kinase (IKK) complex. The IKK complex goes on to activate nuclear factor-κB

(NF-κB), which translocates to the nucleus and promotes the expression of pro- inflammatory genes [282].

Activation of TLRs may also result in signal transduction by mechanisms independent of MyD88, utilizing adaptor molecules with a TIR domain, such as the aptly named TIR-domain-containing adaptor protein inducing IFN-β (TRIF) and TRIF-related adaptor molecule (TRAM) [282]. The MyD88-independent signaling pathways typically do not activate NF-κB, and instead promote the generation of interferons via the transcription factor interferon-regulatory factor 3 (IRF3). One exception to the trend is recognition of LPS by TLR4, which can activate NF-κB without MyD88, but much more slowly [282].

Figure 23. TLR signal transduction pathways. “TLR5, TLR11, TLR4, and the heterodimers of TLR2– TLR1 or TLR2–TLR6 bind to their respective ligands at the cell surface, whereas TLR3, TLR7–TLR8, TLR9 and TLR13 localize to the endosomes, where they sense microbial and host-derived nucleic acids. TLR4 localizes at both the plasma membrane and the endosomes. TLR signaling is initiated by ligand-induced dimerization of receptors. Following this, the Toll–IL-1-resistence (TIR) domains of TLRs engage TIR domain-containing adaptor proteins (either myeloid differentiation primary-response protein 88 (MYD88) and MYD88-adaptor-like protein (MAL), or TIR domain-containing adaptor protein inducing IFNβ (TRIF) and TRIF-related adaptor molecule (TRAM)). TLR4 moves from the plasma membrane to the endosomes in order to switch signaling from MYD88 to TRIF. Engagement of the signaling adaptor molecules stimulates downstream signaling pathways that involve interactions between IL-1R-associated kinases (IRAKs) and the adaptor molecules TNF receptor-associated factors (TRAFs), and that lead to the activation of the mitogen-activated protein kinases (MAPKs) JUN N-terminal kinase (JNK) and p38, and to the activation of transcription factors. Two important families of transcription factors that are activated downstream of TLR signaling are nuclear factor-κB (NF-κB) and the interferon-regulatory factors (IRFs), but other transcription factors, such as cyclic AMP-responsive element-binding protein (CREB) and activator protein 1 (AP1), are also important. A major consequence of TLR signaling is the induction of pro- inflammatory cytokines, and in the case of the endosomal TLRs, the induction of type I interferon (IFN). dsRNA, double-stranded RNA; IKK, inhibitor of NF-κB kinase; LPS, lipopolysaccharide; MKK, MAP kinase kinase; RIP1, receptor-interacting protein 1; rRNA, ribosomal RNA; ssRNA, single-stranded RNA; TAB, TAK1-binding protein; TAK, TGFβ-activated kinase; TBK1, TANK-binding kinase 1.” Figure and caption reproduced from [283] with permission from Nature Publishing Group. 60

Signaling pathways downstream of TLR activation typically result in the induction

of pro-inflammatory genes via MyD88/NF-κB activation or induction of type I interferons

via MyD88 independent pathways/IRF3. Genes induced by TLR activation have been

implicated in the production of pro-inflammatory cytokines, chemokines, major

histocompatibility complex (MHC), co-stimulatory molecules, and antimicrobial peptides

[19]. Whereas cytokines and chemokines promote innate immunity and inflammatory

responses, MHC and co-stimulatory molecules facilitate adaptive immune responses.

Cytokines released in response to TLR activation include the interleukins (IL) IL-1α [266],

IL-1β [266], IL-6 [252, 284], and IL-12 [257, 273], tumor necrosis factor (TNF, also listed

as TNF-α) [254, 262-264, 266, 268, 270, 273, 284], interferon-α (IFN- α) [273], and the

chemokines macrophage inflammatory protein-1α (MIP-1α) [262, 266, 271], MIP-1β

[266], MIP-2 [262, 264, 271], monocyte chemoattractant protein 1 (MCP-1) [285], and KC

[262, 271]. Cytokines can promote further inflammatory activation and edema, and chemokines can promote cellular extravasation and trafficking [286]. Other factors released after TLR activation include the free radical signaling molecule nitric oxide (NO)

[253, 254] and an enzyme that produces it, inducible nitric oxide synthase (iNOS) [268,

270]. Additionally, TLR activation may result in release of the matrix metalloprotease 3

(MMP-3) [271] or MMP-9, which are involved in tissue repair [265]. Finally, TLR activation may release the pro-inflammatory enzyme cyclooxygenase 2 (COX-2) [268], which activates arachidonic acid/prostaglandin inflammatory mechanisms [287]. Overall,

TLR signaling pathways utilize a broad range of downstream effectors to respond to pathogenic and endogenous threats.

Due to established roles of particular PRR in neurodegenerative disorders covered

in Section 2.10, we have taken particular interest in the potential role of TLR2, TLR4, and the co-receptor cluster of differentiation 14 (CD14) in the foreign body response to

intracortical microelectrodes.

The cell-surface receptor TLR2 is expressed on endothelial cells and antigen- presenting cells, including macrophages and microglia [19, 231, 288]. Rather than forming homodimers to activate downstream signaling pathways, TLR2 typically forms homodimers with TLR1 or TLR6 [19]. The main function of TLR2 is the recognition of gram-positive bacteria via peptidoglycan (disputed by Travassos et al. [230]) and lipoteichoic acid [223-227], but it also binds the PAMPs lipoproteins/lipopeptides [289], lipoarabinomannan [290], phenol-soluble modulin [291], glycoinositolphospholipids

[292], porins [293], atypical LPS [294, 295], and zymosan [231, 282]. In addition to the aforementioned PAMPs, TLR2 recognizes the DAMPs HSP60 [255], HSP70 [256, 257],

Gp96 [258], hyaluronan-derived oligosaccharide [262], biglycan [264], necrotic cells [270,

271], and HMGB1 [267, 269]. The broad range of recognized ligands makes TLR2 an effective innate immunity activator throughout the body, but the important role of TLR2 in CNS inflammation and neurodegeneration will be discussed in Section 2.10.

In contrast to TLR2, TLR4 communicates via homodimerization and recognizes gram-negative bacteria [19, 222, 296, 297]. The cell-surface receptor TLR4 is expressed on macrophages, microglia, and several immune cells [19, 214, 298]. In addition to bacterial recognition via LPS, TLR4 also recognizes several other PAMPs [221, 282], including mannan [233], glucuronoxylomannan [221, 234], viral fusion protein [299], and viral envelope proteins [247]. Also, TLR4 is also involved in the recognition of several

DAMPs [20, 207], such as HSP60 [253, 254], HSP70 [256, 257], Gp96 [258], HSP22

[252], heparan sulfate [259-261], hyaluronan-derived oligosaccharide [262, 263], biglycan

[264], fibronectin extra domain A [265], HMGB1 [267, 269], and fibrinogen [193].

Involvement of TLR4 signaling in CNS inflammation and neurodegeneration will also be discussed in Section 2.10.

The co-receptor CD14, a 55 kD glycoprotein closely associated with TLR2 and

TLR4, is predominantly known for its role in the recognition of LPS [300]. Briefly, CD14 binds LPS monomers extracted from LPS aggregates by LPS binding protein (LBP) [301,

302], and the TLR4-MD-2 complex subsequently binds LPS [300, 303]. Although the exact mechanism of LPS transfer from CD14 to TLR4-MD-2 has not been elucidated, the presence of CD14 greatly improves the sensitivity of LPS recognition by effector cells

[304]. Macrophages and to a lesser extent microglia express CD14 [305-307], either linked to a membrane by a glycophosphoinositol or released in a soluble form [19, 300]. In addition to LPS recognition, CD14 acts as a co-receptor to TLR2 in the recognition of the various PAMPs, such as virions [249] and the cell wall components peptidoglycan

(disputed by Travassos et al. [230]) and LTA [24, 227]. The co-receptor CD14 is thought to be an adaptor molecule to the TLRs for the binding of both PAMPs and DAMPs [24].

For example, CD14 acts as a co-receptor to TLR2 and/or TLR4 in the recognition of

HMGB1 [308], β amyloid plaques [26], and HSP70 [25, 256]. Further, CD14 signals independently of TLR2 and TLR4, such as in the endosomal recognition of viral nucleic acids by TLR7 and TLR9 [309], and the non-TLR mediated recognition of necrotic cells

[102] and apoptotic cells [310-312]. Finally, CD14 facilitates the internalization of several

molecules, such as phosphatidylinositol [313]. The role of CD14 in CNS inflammation and neurodegeneration will be discussed further in Section 2.10.

2.10 Toll-like Receptors and CD14 in Neurodegenerative disorders

The innate immunity receptors TLR2, TLR4, and CD14 play a prominent role in

several neurodegenerative disorders due to their expression in CNS tissue and ability to promote potent inflammatory mechanisms [23]. The receptors TLR2 and TLR4 are expressed on microglia and astrocytes in both human and mice and on neurons in mice only [208, 314]. Additionally, TLR2 is expressed on oligodendrocytes [314]. Although constitutive expression of CD14 in the brain is limited to perivascular, leptomeningeal, and choroid plexus macrophages, it can be upregulated in parenchymal microglia following injury [315, 316]. Subsequently, activation of TLR2, TLR4, and/or CD14 on microglia can induce neurodegeneration [317-319]. In addition to parenchymal cells of the brain,

TLR2 and TLR4 are expressed on cerebral endothelial cells [320], and soluble CD14 may participate in the ligand recognition of endothelial cells lacking membrane-bound CD14

[321]. Injuries of the CNS may also recruit CD14 positive macrophages to enact and propagate inflammatory responses [315]. The receptors TLR2 and TLR4 are commonly expressed in macrophages as well. Pro-inflammatory activation of TLR2, TLR4, and

CD14, including ligands and downstream effectors are covered in detail in Section 2.9.

Neuroinflammatory mechanisms involving TLR2, TLR4, and CD14 activation have been demonstrated to contribute to neurodegeneration [317-319]. Resident microglia or infiltrating macrophages may respond to TLR2, TLR4, and CD14 ligands linked to CNS disorders and injuries, such as α-synuclein [281, 322], fibrillar β-amyloid [26], heparan sulfate proteoglycans [323], heat shock proteins [253, 324, 325], necrotic neurons [270,

326], and whole and partial bacteria [317, 327]. Activation of TLRs on microglia can lead to microglial activation and the subsequent release of cytokines, chemokines, reactive oxygen species [281], and nitric oxide [317], as well as induce phagocytosis of both damaged and viable neurons [328]. The release of reactive oxygen species by activated microglia can cause damage to neuronal membranes and dendrites [329]. Additionally, activation of TLRs has been shown to interfere with remyelination [330]. In summary, over-reaction of microglia to TLR2, TLR4, and CD14 can cause neuronal damage in many ways.

Contrary to the major role of TLR2, TLR4, and CD14 signaling in neurodegeneration, these signaling pathways have also been linked to neuroprotection

[331]. Activation of TLR2, TLR4, and CD14 may protect neurons by reverting microglia to a pro-healing activation state [332], inducing cytokine release in a beneficial context

[333], or clear dangerous molecules from the CNS [334]. Thus, well-regulated innate immune signaling is useful for maintaining neuronal health.

Activation of TLR2, TLR4, and CD14 has been implicated in the neurodegeneration caused by a variety of CNS diseases and injuries, including Parkinson’s disease (PD) [335], dementia with Lewy bodies (DLB) [335], multiple system atrophy

(MSA) [335], multiple sclerosis (MS) [336], amyotrophic lateral sclerosis (ALS) [337],

Alzheimer’s disease [26, 338], neuropathic pain [339], subarachnoid hemorrhage [340,

341], traumatic brain injury [342], focal cerebral ischemia [343], spinal cord injury [57], and brain injury [60]. Synucleinopathies (PD, DLB, and MSA), Alzheimer’s disease, and

ALS will be discussed in further detail due to the involvement of multiple receptors of interest and heavy innate immunity involvement.

Alzheimer’s disease is a memory disorder involving plaques made of β amyloid in the parenchyma of the brain [21, 344]. Bystander damage caused by the chronic inflammatory mechanisms directed against β amyloid plaques is hypothesized to be a major source of neurodegeneration [345]. The receptors TLR2, TLR4 and CD14 on microglia recognize fibrillar β amyloid and induce a pro-inflammatory state involving phagocytosis and ROS release [26]. Inhibiting TLR2 and TLR4 by various means has resulted in decreased generation of pro-inflammatory factors and subsequent neurotoxicity [21, 319,

346-348]. On the other hand, activation of TLR2, TLR4, and CD14 signaling has been shown to facilitate the clearance of β amyloid [349], and the removal of neurons damaged by Aβ [350]. Thus, some degree of TLR activation while limiting neuronal damage is likely the optimal scenario for mitigating Alzheimer’s disease.

Recent studies have identified participation of TLR2 and TLR4 in synucleinopathies, such as Parkinson’s disease, dementia with Lewy bodies, and multiple system atrophy [281, 322, 334, 335]. These diseases involve misfolding and buildup of the protein α-synuclein, resulting in TLR-mediated neuroinflammation and neurodegeneration [281, 351]. Activation of TLR2 disrupts the clearance of α-synuclein via autophagy [322]. Contrarily, activation of TLR4 promotes the clearance of α-synuclein via phagocytosis, while simultaneously promoting neurodegenerative inflammatory mechanisms [281, 334]. Thus, a proper balance of clearing harmful materials and limiting self-inflicted damage must be maintained to combat synucleinopathies. Alternatively, a

TLR4 ligand similar to LPS, mono-phosphoryl lipid A, has been identified to selectively promote clearance and not destructive inflammatory cascades [335]. Therefore,

66 approaches more nuanced than simply shutting down the signaling of entire TLRs must be considered in the mitigation of TLR-mediated neuroinflammation and neurodegeneration.

Amyotrophic lateral sclerosis (ALS) is a disorder involving death of motoneurons in the motor cortex and spinal cord resulting in muscle atrophy and paralysis [21, 352, 353]

. Although the initial cause of motoneuron damage is unclear, activated microglia with mutations in superoxide dismutase 1 (SOD1) have been shown to exacerbate neurodegeneration via the release of superoxide, nitrate, and nitrite [21, 354]. TLR2, TR4, and CD14 have been shown to activate microglia by binding extracellular mutated SOD1

[355]. Further, knocking out TLR4 resulted in increased survival and improved functional outcomes in a mouse model of ALS [356]. Recently, natural and synthetic TLR4 antagonists have been shown to mitigate neurodegeneration and nitric oxide release

(natural TLR4 antagonist only) by microglia in vitro [337]. Overall, persistent activation of microglia via TLR2, TLR4, and CD14 in ALS leads to further neurodegeneration, and inhibiting these pathways mitigates harmful outcomes.

In summary, TLR2, TLR4, and CD14 are expressed on tissues of the CNS, recognize PAMPS and DAMPS in the CNS, and promote inflammatory mechanisms involved in neurodegenerative disorders. Inhibition of TLR2, TLR4, and CD14 holds the potential to mitigate neurodegenerative disorders, but clearance of the ligands that generate inflammation must be considered.

2.11 TLRs/CD14 in the foreign body response to intracortical microelectrodes

Although there is currently little direct evidence for TLR2, TLR4, and CD14 signaling in the foreign body response to intracortical microelectrodes, there are several connections that suggest a role in the recognition of damage to promote and perpetuate

67 inflammation. Particularly, expression in cells at the electrode tissue interface [208, 314-

316], presence of ligands at the electrode tissue interface [28, 265], and outcomes of activation [209, 214, 318, 357-362] may suggest participation of TLR2, TLR4, and CD14.

As discussed in Section 2.10, resident parenchymal microglia express TLR2 and

TLR4 constitutively [208, 314], and may express CD14 following injury [315, 316].

Additionally, infiltrating macrophages constitutively express TLR2, TLR4, and CD14.

Following intracortical microelectrode implantation, both resident microglia and infiltrating macrophages accumulate at the electrode-tissue interface and revert to a pro- inflammatory activated state [144]. The receptors TLR2, TLR4, and CD14 may play a role in the pro-inflammatory activation of microglia and macrophages through the recognition of ligands.

Several of the ligands that activate TLR2, TLR4, and CD14 may be present at the electrode tissue interface. Starting with PAMPs, residual endotoxin (LPS) has been detected at levels above FDA guidelines for CNS devices on intracortical microelectrode probes even following standard sterilization protocols with ethylene oxide [28].

Ravikumar et al. compared sterilization methods, and found that probes with higher endotoxin levels exhibited higher microglial activation, astrocytic encapsulation, blood- brain barrier permeability, and neuronal dieback at acute time points [28]. Since TLR4 and CD14 are involved in the main pathway of endotoxin recognition [19], their findings indicate that activation of TLR4 and CD14 may contribute to unfavorable inflammatory mechanisms at the electrode tissue interface. Immunohistochemical differences between sterilization methods were not observed 16 weeks after implantation, suggesting that endotoxin or LPS may not be a significant activating ligand for chronic inflammatory

68 responses once the LPS introduced by the electrode has been cleaned up. Further, Harris et al. administered LPS systemically to rats at the time of intracortical microelectrode implantation, and observed decreased neuronal survival paired with lower firing rates in evoked recordings compared to control rats [27]. This indicates that activation of TLR4 and CD14 may contribute to declining recording performance as well as the foreign body response to the electrode. Gram-positive bacterial cell wall components, recognized by

TLR2, were not quantified in the sterilization study, but may be present following imperfect sterilization methods.

In addition to PAMPs, several DAMPs may be present at the electrode tissue interface due to the damage caused by insertion trauma, micromotion, and chronic inflammatory mechanisms [28]. Enhanced expression of fibronectin, a ligand to TLR4

[265], has been detected in the reactive tissue around implanted intracortical microelectrodes [149], as well as cortical impact injures [363]. HMGB1, recognized by

TLR2 [267, 269], TLR4 [267, 269], and CD14 [308], was observed to be upregulated around implanted intracortical microelectrodes [189] as well as in the brain following ischemia [268, 364, 365]. Further, necrotic cells, recognized by TLR2 [270, 271] and

CD14 [102], have been observed around implanted intracortical microelectrodes [366].

Other DAMPs have been observed in various brain injuries and illnesses. A ligand to

TLR4, HSP70, has been observed in the brain following focal cerebral ischemia [367, 368].

In vitro studies of CNS cells also indicate evidence of DAMP release following nervous system injuries. The ligands HSP60 and HSP70, recognized by TLR2 [255-257], TLR4

[253, 254, 256, 257], and CD14 (HSP60 only) [25, 256], were observed on the axons of cultured DRG neurons that were explanted from rats following sciatic nerve injury [324,

325]. Further, HSP60 was observed being released from apoptotic and necrotic cells in mixed CNS culture [253]. Additionally, fibrinogen, a blood protein that activates TLR4

[193], has been speculated to be present around implanted intracortical microelectrodes [1,

129, 164]. The disruption of vasculature upon intracortical microelectrode implantation

[138] and chronic permeability of the blood brain barrier likely results in the presence of

fibrinogen in the brain at both acute and chronic time points [136, 369]. Moreover,

fibrinogen has been observed around peripheral implants [370, 371]. Overall, a variety of

DAMPs may be released in the brain following intracortical microelectrode implantation.

Finally, the downstream consequences of TLR2, TLR4, and CD14 activation may

indicate a role in the foreign body response to implanted intracortical microelectrode.

Activation of these pathways on microglia and macrophages leads to the activation of NF-

κB (Figure 23), which mediates pro-inflammatory genes responsible for the release of

cytokines, chemokines, cyclooxygenase-2, and matrix metalloproteinase-9 (MMP-9) [214,

372, 373]. These factors may perpetuate chronic inflammatory mechanisms around the

implanted device. Activated microglia and macrophages are present at the electrode tissue

interface long after device implantation [144, 369]. Activation of TLRs has the capability

to induce neuronal damage [209, 318], and neuronal dieback is commonly observed along

with the foreign body response to intracortical microelectrodes [129, 130, 369]. Activation

of TLRs by DAMP recognition may contribute to the loss of neurons around the implant.

Additionally, the neurons killed around implanted intracortical microelectrodes may

provide further stimulus for TLR activation via necrosis or the release of HSP60, HSP70,

or HMGB1, thus self-perpetuating chronic inflammation. Further, activation of TLRs may

contribute to chronic blood-brain barrier permeability via the release of nitric oxide pro-

70 inflammatory cytokines, NO, and reactive oxygen species (ROS), and MMP-9 [143, 214,

357-362]. Blood-brain barrier permeability is a persistent problem of the foreign body response to intracortical microelectrodes and is associated with poor recording performance [136, 369].

Overall, the evidence presented here suggests that TLR2, TLR4, and CD14 play a role in the foreign body response to intracortical microelectrodes via the recognition of

PAMPs and DAMPs by microglia and macrophages and subsequent pro-inflammatory, neurotoxic, and blood-brain barrier damaging actions (Figure 24).

Figure 24. Potential role of TLR2, TLR4, and CD14 in the foreign body response to intracortical microelectrodes. Microglia and macrophages at the electrode tissue interface may recognize PAMPs and DAMPs at the electrode tissue interface via TLR2, TLR4, and CD14, resulting in the release of soluble factors such as cytokines, chemokines, reactive oxygen species, and reactive nitrogen species. These soluble factors may damage neurons, blood vessels, and electrode materials, leading to poor intracortical microelectrode performance. Cells damaged and killed by the soluble inflammatory factors may release DAMPs that activate microglia and macrophages. Blood proteins released following damage to the blood brain barrier may also act as DAMPs. Inflammatory cells infiltrating across the blood-brain barrier permeability may become activated by DAMPs and contribute to the foreign body response. Thus, TLR2, TLR4, and CD14 signaling may contribute to the activation and perpetuation of chronic inflammatory mechanisms in the foreign body response to intracortical microelectrodes.

2.12 How to inhibit TLR2, TLR4, and CD14

Several strategies exist for inhibiting TLR2, TLR4, and CD14 mediated signaling,

varying in target and molecular composition. The most direct strategies target the receptors

themselves, however, inhibition strategies may target other components of the

TLR2/TLR4/CD14 signal transduction pathways. A variety of peptides, antibodies, and

RNA sequences may disrupt signaling at the various targets. In this this section, several

TLR2/TLR4/CD14 inhibition strategies under different stages of investigation will be reviewed (Table 4).

The first major option for inhibiting TLR2/TLR4/CD14 signaling is antagonism of the binding domain with a small molecule. The glycolipid IAXO-101 has demonstrated success as a CD14-TLR4 antagonist by competitively occupying CD14 [374], resulting in mitigation of sepsis, neuropathic pain, and experimental malaria in mice [339, 375, 376].

A similar compound IAXO-102 attenuated the development of an experimental form of aneurysm [377]. The naturally derived antagonist to TLR2 staphylococcal superantigen- like protein 3 (SSL3), may inhibit TLR2 signaling by both blocking binding sites and disrupting heterodimerization [378]. Due to the diversity of molecular structures recognized by the same TLR, the effectiveness of small molecule antagonists to interfere with ligand-receptor interactions may vary between ligand [379]. Small molecule antagonists to the TLR2/TLR4/CD14 pathways have demonstrated varying degrees of success. The capability of small molecule antagonists to mitigate TLR2/TLR4/CD14 activity and attenuate detrimental inflammatory outcomes is promising for their utilization in intracortical microelectrode integration.

Another mechanism of inhibiting TLR2/TLR4/CD14 signaling by blocking the

ligand-receptor interactions utilizing blocking antibodies. Antibodies that bind a cellular

receptor and prevent ligand binding have demonstrated success in various inflammatory

disorders [380]. Anti-CD14 antibodies may disrupt CD14 signaling by blocking the

binding of LPS to CD14 [381, 382] or through mechanisms independent of LPS binding

[381]. The commercially available monoclonal antibody IC14 recognizes human CD14

and saturates CD14 on monocytes and granulocytes [383]. Intravenous administration of

IC14 in humans reduced responsiveness to LPS, specifically attenuating the release of pro-

inflammatory cytokines (TNF, IL-6, and IL-10), the activation of leukocytes and endothelial cells, the acute phase protein response, and various symptoms [383]. The

administration of IC14 has produced less promising results in severe cases of sepsis [384,

385]. One potential setback for the use of blocking antibodies is activation of the

complement pathway by the Fc region of the antibody [386]. Recent studies employing hybrid antibodies with inert Fc regions have been successful in attenuating CD14-mediated inflammatory responses without activating complement-related side effects in both in vivo porcine models and ex vivo human whole blood models of sepsis [386]. Further, pairing complement inhibition with hybrid anti-CD14 antibodies has demonstrated enhanced survival in a porcine model of polymicrobial sepsis [387]. Finally, hybrid anti-CD14 antibodies with and without concurrent complement inhibition has outperformed the synthetic small molecule antagonist eritoran in mitigating monocyte activation in response to sepsis [388]. Therapeutic antibodies have demonstrated success in mitigating

TLR2/TLR4/CD14 mediated disorders despite potential complications. Thus, therapeutic antibodies may be applicable for integrating intracortical microelectrodes.

Alternative inhibition strategies to inhibit TLR2 and TLR4 focus on the non-CD14

co-receptors. The protein MD-2 (myeloid differentiation 2) is the other major cell-surface

co-receptor involved in the recognition of LPS. Where CD14 is thought to facilitate

recognition of both PAMPs and DAMPs, MD-2 is thought to facilitate response to PAMPs

only [24]. Limiting the types of ligands inhibited will likely result in different outcomes.

Small molecule antagonists to TLR4-MD-2 may be derived from synthetic and natural

sources, such as olive oil and curcumin [379]. Synthetic antagonists to TLR4-MD-2 are

typically similar in structure to the LPS component lipid A that is recognized by the TLR4

receptor complex [379]. Eritoran is a synthetic small molecule antagonist derived from

lipid A that binds MD-2 without promoting the dimerization of TLR4 [379, 389]. Despite

promising in vitro [390] and animal model results [391], eritoran was unable to reduce

mortality in clinical trials of severe sepsis [392]. More recently, the TLR4/MD-2

antagonist peptide TAP2 (TLR4 antagonistic peptide 2), hypothesized to bind the LPS- binding pocket of MD-2, was shown to attenuate LPS-induced inflammatory activation in vitro and in mice [393]. Further, clinical trials with TAP2 have not been reported. Due to the predominant interaction of MD-2 with PAMPs [24], I would not expect TLR4-MD-2 antagonists to be effective in mitigating chronic inflammation associated with the damage caused by intracortical microelectrodes. However, Potter et al previously demonstrated that the natural TLR4-MD-2 antagonist curcumin improves neuronal survival and blood- brain barrier stability around intracortical microelectrodes implanted in rats [189]. This may indicate a larger role of pathogens at the electrode tissue interface, involvement of

MD-2 in DAMP recognition, or promiscuous activity of curcumin. Regardless of the

mechanism, this finding indicates the potential for TLR4-MD-2 inhibitors in the integration

of intracortical microelectrodes.

Additional inhibition strategies to inhibit TLR2/TLR/CD14 signaling target downstream intracellular signal transduction pathways. Antagonists have been developed

or identified to target the adaptor molecules MyD88 adaptor-like (Mal) and TRIF-related

adaptor molecule (TRAM) [394], the TLR2 TIR domain [395], JNK [327], ubiquitin chains

[396], TLR4-MyD88 interactions [397], TLR1-TLR2 heterodimerization [398], and the intracellular domain of TLR4 [399, 400]. Of note, administration of the flavonoid wogonin attenuated LPS-induced inflammation in dorsal root ganglion neurons [397] and improved outcomes in a model of traumatic brain injury [372] by interfering with TLR4-MyD88 complex formation [397]. Further, cyclohexene-derived TAK-242/resatorvid inhibits

TLR4 signaling by binding the intracellular domain of TLR4 and preventing interactions with downstream adapter molecules [399, 401]. In vivo administration of TAK-242 for the treatment of sepsis [402], endotoxic shock [401], traumatic brain injury [403], cerebral ischemia [399], intracerebral hemorrhage [399], and optic nerve crush [404] resulted in positive outcomes, such as attenuated inflammation and neuroprotection. A phase III clinical trial (NCT00633477) investigated the utilization of TAK-242 to treat sepsis, but the trial was discontinued due to business concerns unrelated to safety [405]. Strategies to inhibit TLR2/TLR4/CD14 signaling have demonstrated preliminary efficacy but have not demonstrated clinical success to this date. Nonetheless, many of the in vivo experiments have demonstrated efficacy in conditions related to the foreign body response to intracortical microelectrodes. Traumatic brain injury involves similar phenomena exhibited in the foreign body response to intracortical microelectrodes, such as blood-brain

barrier permeability, inflammation, and neuronal death [406]. Additionally, ischemic conditions have been observed around intracortical microelectrodes [407]. The efficacy of inhibiting intracellular signal transduction pathways downstream of TLR2/TLR4/CD14 in

CNS applications is promising for the utilization in the integration of intracortical microelectrodes.

Other strategies to inhibit TLR2/TLR4/CD14 signaling may alter the cell-surface expression of these receptors, either by downregulation or internalization. Administration of rosmarinic acid [408], siRNA [409], oxymatrine [342], glycyrrhizin [410] , argon [411], pituitary adenylate cyclase-activating polypeptide (PACAP) [412], melatonin [413], apigenin [414], and ursolic acid [415], resulted in reduced TLR2, TLR4, and/or CD14 expression, whereas walnut extract was shown to promote receptor internalization [416].

Of note, intrathecal siRNA that reduces TLR4 expression attenuated neuropathic pain in a rat model [409]. Oxymatrine and PACAP conferred neuroprotection in a rat models of traumatic brain injury by inhibiting the expression and upregulation of TLR4 [342, 412].

Glycyrrhizin inhibited TLR2-HMGB1 signaling, resulting in reduced lymphocyte trafficking in experimental autoimmune thyroiditis in mice [410]. Melatonin, apigenin, and ursolic acid attenuated the effects of subarachnoid hemorrhage, including neurobehavioral deficits, blood-brain barrier permeability, and neuronal apoptosis [413-

415]. Reducing the expression of TLR2, TLR4, and/or CD14 effectively attenuates the negative effects of overactive TLR signaling. Again, the success of cell-surface expression in CNS injuries and disorder demonstrate a potential application for integrating intracortical microelectrodes.

Overall, TLR2, TLR4, and CD14 may be attenuated at different points along their respective signaling pathways (Table 4). Many promising findings have been observed in vitro and in vivo, but strong clinical performance in humans has not been demonstrated.

Human clinical trials mainly failed in severe cases of sepsis, which is likely not the best predictor of efficacy mitigating the foreign body response to intracortical microelectrodes.

However, promising in vivo results in CNS injuries and diseases may indicate the potential success of these strategies for integrating intracortical microelectrodes.

Table 4. Strategies for the inhibition of TLR2, TLR4, and CD14. Inhibition Target Type of therapeutic Receptor targeted Therapeutic Application(s) Mechanism Ligand binding Small Molecule CD14/TLR4 IAXO-101 Sepsis [375], neuropathic pain [339], malaria [376] Ligand binding Small Molecule CD14/TLR4 IAXO-102 Aneurysm [377] Ligand binding/ Small Molecule TLR2 SSL3 [378] N/A receptor dimerization

Ligand binding Antibody CD14 IC14 LPS responsiveness [383], sepsis [384, 385] Co-receptors (non- Small molecule TLR4 (via MD-2) Eritoran Sepsis [392] CD14) Co-receptors (non- Small molecule TLR4 (via MD-2) TAP2 Inflammation CD14) [393] Co-receptors (non- Small molecule TLR4 (via MD-2) Curcumin Integration of CD14) (natural) intracortical microelectrodes [189] TLR4-MyD88 Small molecule TLR4 Wogonin Inflammation in Complex formation DRG neurons [397], traumatic brain injury [372] Intracellular TLR Small molecule TLR4 TAK- Sepsis [402], domain interactions 242/resatorvid endotoxic shock [401], traumatic brain injury [403], cerebral ischemia [399], intracerebral hemorrhage

[399], optic nerve crush [404] Receptor siRNA TLR4 N/A Neuropathic expression at cell pain [409] surface Receptor Small molecule TLR4 Oxymatrine Traumatic brain expression at cell injury surface [342] Receptor Small molecule TLR4 PACAP Traumatic brain expression at cell injury surface [412] Receptor Small molecule TLR2 Glycyrrhizin Autoimmune expression at cell thyroiditis [410] surface Receptor Small molecule TLR4 Melatonin Subarachnoid expression at cell hemorrhage surface [413] Receptor Small molecule TLR4 Apigenin Subarachnoid expression at cell hemorrhage surface [414] Receptor Small molecule TLR4 Ursolic acid Subarachnoid expression at cell hemorrhage surface [415]

Chapter 3

Inhibition of the cluster of differentiation 14 innate immunity pathway with IAXO-101 improves chronic microelectrode performance*#

*The following chapter is reproduced, with permission by IOP Publishing, from: John K Hermann, Madhumitha Ravikumar, Andrew J Shoffstall, Evon S Ereifej, Kyle M Kovach, Jeremy Chang, Arielle Soffer, Chun Wong, Vishnupriya Srivastava, Patrick Smith, Grace Protasiewicz, Jingle Jiang, Stephen M Selkirk, Robert H Miller, Steven Sidik, Nicholas P Ziats, Dawn M Taylor, and Jeffrey R Capadona. Journal of Neural Engineering, (2018) 15 025002. https://doi.org/10.1088/1741- 2552/aaa03e. © IOP Publishing. Reproduced with permission. All rights reserved.

#Please note an erratum for this article has been published in 2018 J. Neural Eng. 15 039601 (http://iopscience.iop.org/article/10.1088/1741-2552/aab5bf). The changes from the erratum are reflected in this chapter.

3.1 Abstract

Objective. Neuroinflammatory mechanisms are hypothesized to contribute to

intracortical microelectrode failures. The CD14 molecule is an innate immunity receptor

involved in the recognition of pathogens and tissue damage to promote inflammation. The

goal of the study was to investigate the effect of CD14 inhibition on intracortical

microelectrode recording performance and tissue integration. Approach. Mice implanted

with intracortical microelectrodes in the motor cortex underwent electrophysiological

characterization for 16 weeks, followed by endpoint histology. Three conditions were examined: 1) wildtype control mice, 2) knockout mice lacking CD14, and 3) wildtype control mice administered a small molecule inhibitor to CD14 called IAXO-101. Main

Results. The CD14 knockout mice exhibited acute but not chronic improvements in intracortical microelectrode performance without significant differences in endpoint histology. Mice receiving IAXO-101 exhibited significant improvements in recording performance over the entire 16-week duration without significant differences in endpoint

histology. Significance. Full removal of CD14 is beneficial at acute time ranges, but limited CD14 signaling is beneficial at chronic time ranges. Innate immunity receptor

inhibition strategies have the potential to improve long-term intracortical microelectrode

performance.

3.2 Introduction

Signals recorded by intracortical microelectrodes can be used to control computer

cursors, robotic arms, as well as functional electrical stimulation of a patient’s own arm

[39, 51, 52]. Unfortunately, intracortical microelectrodes fail to consistently record

neurological signals over longer time frames [124]. A number of failure modes likely influence chronic recording stability and quality including: 1) direct mechanical damage of the microelectrode; 2) corrosion of electrical contacts and degradation of passivation layers and insulating coatings; and 3) the neuroinflammatory response that the brain mounts against chronically implanted devices [13, 98, 136, 417]. In a retrospective analysis of microelectrode failures over 28 years in non-human primates, the Donoghue group identified biological driven failure modes (largely inflammatory) as the largest class of chronic microelectrode failures [2].

Biological failure mechanisms span from the tissue and vascular damage associated with device implantation into the cortical tissue, through the progression of the neurodegenerative inflammatory response [1, 138]. Biological failure also includes glial scar formation isolating the microelectrode from the viable tissue [168, 418], as well as the breakdown of the blood-brain barrier (BBB) [11, 136]. Specifically, blood proteins released from the damaged vasculature adsorb to the surface of the microelectrode and promote inflammatory activation of microglia and infiltrating macrophages [11, 140].

Cytokines, chemokines, and other soluble factors released by the microglia or macrophages can damage nearby neurons, as well as recruit more inflammatory cells and promote vascular permeability [33, 129, 141, 142]. Recognition of necrotic cells and blood proteins can promote further inflammatory activation in microglia and macrophages [140, 145], leading to cycles of self-perpetuating inflammation [11]. Chronic inflammation can cause neuronal damage and dysfunction throughout the duration of the microelectrode’s residence in the brain [129, 130]. Loss of neurons within 50 µm of the microelectrode may reduce the capability of the microelectrode to detect single units [117]. As a result, groups like the Bellamkonda, Cui, Rennaker, and Tyler labs have begun to report a direct correlation between the neuroinflammatory response and recording performance [13, 27,

107, 419].

Mitigating the neuroinflammatory response to microelectrodes is a common strategy to improve intracortical microelectrode integration and performance [1].

Unfortunately, a varying degree of success was observed through these approaches [13,

133, 420-424]. Some of the most successful approaches to mitigate the neuroinflammatory response indicate a dominant role of reactive microglia cells and infiltrating macrophages, as well as the stability of the local blood-brain barrier [136, 144, 425-428]. For example, the anti-inflammatory/antibiotic Minocycline has been shown to increase the longevity and quality of functional neural recordings [13]. Additionally, to minimize the effects of surgical trauma and localized hemorrhaging, others have implemented systemic and local application of the anti-inflammatory glucocorticoid dexamethasone and shown it can reduce the inflammatory response to inserted microelectrodes [429-432]. Despite promising results when targeting neuroinflammation, chronic application of general anti-

inflammatory agents is dangerous due to harmful side-effects such as bone-softening [16]

or unanticipated damage to other organs [12]. Thus, a better understanding of the

neuroinflammatory response to implanted intracortical microelectrodes is required to

develop specific therapeutic targets for safe long-term anti-inflammatory strategies.

Innate immunity pathways receptors have been increasingly associated with

neuroinflammatory and neurodegenerative disorders [23, 208, 319, 352]. Innate immunity

is the body’s fast-acting defense against invading pathogens that involves recognition of

general molecular patterns [194]. A major component of innate immunity involves

signaling through Toll-like receptors (TLRs). Toll-like receptors are a family of

transmembrane proteins that recognize general molecular patterns characteristic to

pathogens (pathogen associated molecular patterns – PAMPs), such as bacteria and viruses,

and enact downstream pro-inflammatory changes [19]. The TLRs are expressed on

peripheral immune cells, such as macrophages and dendritic cells, as well as several cells

of the central nervous system, including microglia, astrocytes, and neurons [19, 208].

Downstream effects of TLR signaling include upregulation of the pro-inflammatory

transcription factor NF-κB and release of pro-inflammatory cytokines, such as TNF, NO,

IL1-α, IL-1β, IL-6, or MCP-1 [20, 433]. Growing evidence suggests that TLRs also

recognize endogenous molecules associated with non-infectious tissue damage (damage

associated molecular patterns – DAMPs), including blood proteins, heat shock proteins,

and proteins released by necrotic cells (HMGB1) [207].

Toll-like receptor 2 (TLR2) and Toll-like receptor 4 (TLR4), have been linked to

neuroinflammatory disorders, such as Alzheimer’s and multiple sclerosis [208, 352, 434].

The receptor TLR2 recognizes gram-positive bacteria, as well as necrotic and dying cells

[227, 271, 435]. The receptor TLR4 recognizes gram-negative bacteria via

lipopolysaccharide (LPS), as well as fibrinogen, fibronectin, and other endogenous

molecules [21, 193, 222, 265]. Bacterial endotoxin, necrotic cells, blood proteins, and other factors released by damaged tissue have been observed or hypothesized to be prevalent at the microelectrode–tissue interface, other cortical injuries, or peripheral

implantation sites [1, 11, 28, 129, 132, 149, 189, 363, 366, 370, 371, 436]. Activation of

TLR2 and TLR4 results in the downstream upregulation of the pro-inflammatory

transcription factor NF-κB and the subsequent release of TNF-α and MCP-1 [20, 433].

Activation of TLR4 can lead to neurodegeneration [318]. The receptors TLR2 and TLR4 likely play a role in the chronic inflammatory responses to intracortical microelectrodes, due to their expression on cells resident to, or infiltrating into, the CNS. More importantly,

TLR2 and TLR4 are prevalent and facilitate the release of inflammatory factors upon activation.

Cluster of differentiation 14 (CD14) is a co-receptor to both TLR2 and TLR4, and coordinates ligand binding [24, 26, 227, 303]. Macrophages, and to a lesser extent microglia, express CD14 [305-307]. The co-receptor CD14 is primarily involved in the recognition of bacterial endotoxin with TLR4, but it has also been implicated in the recognition of HMGB1, heat shock proteins, apoptotic cells, necrotic cells, and β amyloid plaques [24-26, 300, 308, 310-312, 437]. The co-receptor CD14 has also been identified to play a role in Alzheimer’s disease [338]. The close association of CD14 with TLR2 and TLR4 in the recognition of PAMPs and DAMPs suggests that CD14 likely plays a role in the neuroinflammatory response to implanted intracortical microelectrodes.

In a study of intracortical microelectrode sterilization methods, we found that

elevated bacterial endotoxin (LPS) presence corresponded with enhanced

neuroinflammation and reduced neuronal survival [28]. Further, activation of CD14 via

administration of LPS correlated with poor recording quality and reduced neuronal survival

[27]. Together, these studies suggested that increased activation of CD14 had a negative

effect on intracortical microelectrode recording performance and tissue integration.

Therefore, we hypothesized that inhibition of the innate immunity pathway for CD14

would attenuate the foreign body response to intracortical microelectrodes and subsequent

recording failure. To test this hypothesis, microelectrodes were implanted into the cortex

of knockout mice lacking CD14 expression to evaluate recording performance and

endpoint histology, compared to wild type controls over 16 weeks. In order to develop a

translational approach to CD14 inhibition, we also investigated the efficacy of systemic

CD14 inhibition in improving intracortical microelectrode performance. IAXO-101

(Innaxon) is a commercially available antagonist to the CD14/TLR4 complex [374].

Studies by Piazza and Bettoni investigating sepsis and neuropathic pain utilized IAXO-

101, with promising results [339, 375]. We therefore hypothesized that systemic inhibition of CD14 via a small molecule antagonist would improve intracortical microelectrode recording performance and tissue integration. To test this hypothesis, we implanted microelectrodes into the cortex of wild type mice systemically administered IAXO-101, and evaluated recording performance and endpoint histology after 16 weeks, compared to non-treated control animals.

3.3 Methods

3.3.1 Intracortical Probe Implantation Procedure

Single-shank 16-channel planar microelectrode arrays (NeuroNexusA1x16-3mm-

50-177) were implanted into the motor cortex of nine Cd14-/-mice (Jackson Laboratory

strain # 000664) lacking the CD14 co-receptor, five wild type mice (Jackson Laboratory strain # 003726) and five wild type mice administered IAXO-101 (Innaxon) aged 8-14

weeks and weighing at least 20g. Prior to implantation, Cd14-/- mice were verified as

knockouts via tail snips and standard PCR analysis methods.

Mice were handled inside a microisolator hood and prepared for surgery using

standard aseptic techniques to minimize pathogen exposure to immune compromised

Cd14-/- mice. Mice were anesthetized with 3% isoflurane, restrained in a stereotaxic frame

with ear bars, and maintained under anesthesia with 0.5-1% isoflurane. Mice were

administered .02mL of Marcaine under the scalp and Meloxicam subcutaneously at a

dosage of 2mg/kg. the skull was exposed with a midline incision, cleaned and dried with

3% hydrogen peroxide and covered with surgical adhesive (3M Vetbond® Tissue

Adhesive) to promote cement adhesion.

Craniotomies for ground and reference wires were drilled with a 0.45 mm diameter

drill bit 1-2 mm left of midline. The ground wire craniotomy was placed 1-2 mm posterior

to bregma and the reference wire was placed 1-2 mm anterior to bregma. A craniotomy

for the microelectrode probe was drilled with a 0.45 mm diameter drill bit 1.5 mm to the

right of midline and 0.5 mm posterior to bregma. These coordinates correspond to motor

activity of the mouse forelimb [438]. All craniotomies were kept hydrated with sterile saline as needed.

Ground and reference wires of the ethylene-oxide-sterilized electrode were inserted first, sealed with silicone elastomer (Kwik-sil), and secured into place with dental cement

(Stoelting). A hand-driven micromanipulator was then used to gradually insert the silicone microelectrode shank into the center of the motor cortex craniotomy. The electrode was inserted in 50 micron increments to an approximate depth of 800-850 microns. At this depth, the 16 electrode contacts should span cortical layers I-V. Each 50-micron insertion step took 1-2 seconds, and the cortex was allowed to rest for at least 5-10 seconds between insertion steps to provide time for any strain on the tissue to relax. The noise level on all channels was monitored throughout the process to verify the electrode remained intact and to determine when all channels were in the cortex. All electrode contacts were determined to be in the cortex when the noise level of the top channel transitioned from a high noise level (indicative of being in open air) to a lower physiological level.

3.3.2 Administration of a small molecule CD14 Inhibitor.

IAXO-101 (Innaxon), a small-molecule inhibitor to CD14, was diluted 1:5 with a sterile, nonpyrogenic 5% dextrose injection using aseptic technique according to manufacturer protocol. IAXO-101 was administered at a dosage of 3mg/kg subcutaneously every other day starting 16-24 hours before microelectrode implantation.

Doses were spaced 48±2 hours apart.

3.3.3 Electrophysiological Recordings.

Electrophysiological recordings were obtained the day of surgery, 5 days post- surgery, and twice weekly until 16 weeks post-implantation. During each recording session, the headstage (TDT ZC16) was connected after briefly exposing the animal to 3% isoflurane to minimize head movement and strain on the implant. After recovery, neural

86 recordings were collected while the mice walked freely in a large bowl for 3 minutes. The bowl was rotated by hand to counter-act tangling of the headstage cable.

Electrophysiological activity was recorded with a TDT R16PA Medusa Pre-amp and TDT

RX5 Pentusa Processor. Afterwards, the headstage was disconnected after brief exposure to 3% isoflurane.

3.3.4 Signal Processing

Electrophysiological data were sampled at 24.4 kHz, bandpass filtered between

300Hz and 3 kHz and then common average referenced to remove global noise. Brief time segments containing artifacts were removed post hoc by a reviewer blinded to time point and test group. Occasionally, channels or entire recording sessions were removed by the blinded reviewer if it appeared that the cable was improperly connected (i.e. extreme high noise) or water drips from the water bottle had shorted the contacts to ground (i.e. extreme low noise). Spikes were detected when the signal crossed a lower threshold set at 3.5 standard deviations from the mean. Spike waveforms consisted of 12 samples before and

24 samples after each threshold crossing and were sorted into different single units using the unsupervised sorting algorithm, Wave_clus [439].

To focus analysis on cortical layers with large firing neurons (layer V) and account for inherent variance in implantation depth, only the eight consecutive channels with the highest sum of average units over time were included in the calculation of the following five performance metrics for each recording session: 1) average number of units per working channel, 2) percentage of working channels detecting single units, 3) background noise level averaged across all working channels, 4 & 5) the average signal amplitude and signal-to-noise ratio (SNR) of the subset of the eight channels detecting single units.

Signal amplitude was defined as the peak-to-peak amplitude of that unit’s mean

waveform. Noise amplitude for each three-minute recording session was calculated as two

times the standard deviation of the background electrophysiological activity after time

windows containing spikes and artifacts were removed. Standard deviation of the

background activity without spikes was estimated using the median of the absolute

deviation of the voltage divided by 0.6745—an efficient metric equivalent to one standard

deviation for Gaussian distributions [118, 439]. Signal-to-noise ratio for each sorted unit

was that unit’s signal amplitude divided by the noise amplitude calculated for that channel.

Units with an SNR less than 3 were excluded from analysis.

Prior to spike detection and sorting, artifacts were removed from the recorded data

by a reviewer blinded to the animal, test condition, and post-op day. For consistency, the same person reviewed all data files in random order. Voltage traces for all electrode channels (bandpass filtered 300-3,000Hz) were common average referenced and viewed to identify and exclude specific brief sections of time containing obvious large artifacts (e.g. from the mouse shaking its head or the cable hitting the edge of the testing chamber). The reviewer identified a time segment as an artifact if it contained a sudden large spike in voltage (typically > 20 times the background noise) that appeared across most or all channels simultaneously. For each detected artifact, an additional one second window on either side of the visually-identified artifact was also excluded to ensure any remaining

vibrations in the cable/connector were also excluded from the analysis. In total, 2.6% of

the recording data was removed in this step.

Additionally, individual days of recording were removed from analysis by the

blinded reviewer in cases of obvious recording problems. Specifically, drips from the

88 animal’s water bottle sometimes shorted channels to ground creating extremely low voltages or the connector would start coming loose during recording creating a very large increase in noise across the channels. Bedding and debris in the connector was a likely cause of the headstage connector sometimes not fitting properly and becoming loose during recording. In total 9.7% of the recording days were entirely removed from analysis and these were randomly distributed across time and animal group.

On occasion, smaller bits of debris would get in the connector preventing good contact with the headstage on just one or a few channels as indicated by extremely large amplitude noise on individual channels. Overall, 3.8% of individual channels falling in the

‘best eight’ range (see below) were excluded from analysis due to poor connections, and these were randomly distributed across channels, days, and animal groups.

3.3.5 Identification of channels in recordable layers.

Since the 16 recording channels spanned cortical layers I-V, some of the channels were in cortical layers that did not have significant neural cell bodies from which to record action potentials. Including channels from non-spiking cortical layers would unnecessarily dilute the neural recording metrics. Therefore, only a subset of eight sequential channels was used in the analysis. Specifically, for each animal, the eight sequential channels that had the highest number of sorted units where used (calculated as unit counts for each channel averaged across all time points and then summed across eight sequential channels).

The occasional days and/or channels excluded due to artifacts did not contribute to the average across time when calculating the best eight. Once determined, the best eight channels were fixed across all time points and didn’t shift if a channel was dropped on a given day. Instead, daily performance metrics were defined in a way where the occasional

channel drop out would not systematically skew the results (e.g. percentage of working

channels with detectable units; average number of sorted units per working channel).

As expected, the ‘best eight’ sequential channels for each mouse included primarily

the deeper channels. These channels excluded the upper acellular layers and encompassed

layer V, which has the largest cells with the most prevalent action potentials. Note, there

was some inevitable variability in insertion depth due to difficulty visualizing when the

hair thin electrode tip first contacted the cortical surface and due to variability of the

thickness of the fluid layer within the craniotomy. Using the numbering scheme with the

deepest channel labelled as one and the shallowest channel as 16, the deepest channel in

the ‘best eight’ range had a median channel number of 1, mean of 2.2, mode of 1, and

maximum of 7 across all mice indicating the best eight were primarily in the deeper layers.

3.3.6 Approximation of Best 8 Channel Slice Depth

The approximate depth of the tissue sections was estimated by counting the number

of slices between the first slice where the probe hole appears to the current slice and

multiplying by slice thickness (16 µm). The depth of the electrode was estimated by counting the number of slices between the first slice where the probe hole appears and the last slice where the probe hole appears and multiplying by slice thickness. The depth boundaries of the best 8 channels were estimated using the approximate depth of the electrode and subtracting the distance from the tip to the deepest and shallowest of the best

8 channels. Distances between channels and the tip were estimated using the tip length (50

µm), the contact to contact distance (50 µm), and the number of contacts between the tip and the best 8 range boundaries. A range spanning from approximately 50 µm above the

top channel to 50 µm below the bottom of the best 8 channels was included in

immunohistochemical analysis.

Unfortunately, due to the variability in curvature of the brain due to exact electrode

placement, difference in the % swelling of the tissue during fixation, and difference in the extent of tissue lost at the top of the brain during sectioning, it is difficult to provide the mean and standard deviation with quantitative certainty. Therefore, specific electrode

depth should be considered a confounding factor when interpreting the data.

3.3.7 Recording Statistics

Recording data was statistically evaluated using a general linear model in Minitab

with time and treatment group (CD14 or WT; IAXO-101 or WT) as fixed factors. Time

was grouped into two ranges, acute (weeks 1-2, days 0-11) and chronic (weeks 3-16, days

16-109) to represent different phases of the neuroinflammatory response to implanted

intracortical microelectrodes [369, 440]. Individual mouse was nested within treatment as

a random factor to account for repeated measures. Treatment group, time range, mouse,

and the interaction between time range and treatment group (i.e. treatment group * time

range) were used as terms in the model. Combinations of treatment group and time range

were compared using mean and 95% confidence interval calculated by the statistical

software program R [441, 442]. Two groups with 95% confidence intervals calculated

from a general linear model that did not overlap were considered statistically different.

3.3.8 Tissue Processing

Mice were sacrificed via transcardial perfusion following similar methods to

Ravikumar et al. 16 weeks after microelectrode implantation [58]. Mice were anesthetized

using 0.5 ml of a ketamine-xylazine solution (10 mg/ml ketamine, 1mg/ml xylazine) and

transcardially perfused with 1X PBS to clear blood from the mouse. Next 4%

formaldehyde from paraformaldehyde in 1X PBS was perfused through the mouse for 10

minutes to fix the brain.

After perfusion, the mouse head was removed and stored in 4% formaldehyde in

1X PBS for 24-48 hours at 4°C to post-fix the brain. Then, the microelectrode array was carefully pulled straight up out of the brain and the brain was gently removed from the skull. (Several labs have indicated that tissue is removed with the electrode, and remains adhered to the implant. Since the tissue adhered to the implant cannot be quantified, or pieced back into images of the hole to determine the extent of missing tissue, it is common practice to define the hole edge as the implant tissue interface.) The brain was cryoprotected using a gradient of 10%, 20%, and 30% sucrose solutions in 1X PBS until the brain equilibrated at 4°C. Brains were frozen in optimal cutting temperature (OCT) gel on dry ice and transferred to a -80°C freezer. Horizontal sections of brain tissue were sliced

16 µm thick at roughly -20 to -25°C and immediately mounted on microscope slides.

Mounted microscope slides were set out at room temperature overnight and transferred to a -80°C freezer until staining.

3.3.9 Immunohistochemistry

Slices of mouse cortical tissue were stained by immunohistochemistry methods adapted from Ravikumar et al. [28]. Full details of immunohistochemistry methods are included in supplemental Section 8.1.1.1. Tissue sections were blocked for one hour at room temperature using blocking buffers containing 4% serum (chicken or goat) and 0.3%

Triton-X 100 in 1XPBS. Staining targets, antigens, primary antibodies, secondary antibodies, and antibody concentrations are summarized in Table 5. Tissue sections were

incubated in primary antibody solutions overnight at 4°C. Secondary antibody solutions

were incubated for two hours at room temperature. Tissue autofluorescence was dampened

through treatment with a copper sulfate solution [443]. Tissue sections were mounted with

Fluoromount-G (SouthernBiotech) and cover slipped. Tissue sections were allowed to dry

and subsequently stored at 4°C.

Table 5. Summary of immunohistochemistry targets, antigens, antibodies, and concentrations. Table updated from erratum [444].

3.3.10 Fluorescent Microscopy

Fluorescent images of mouse cortical tissue stained with immunohistochemical

markers were captured using an inverted fluorescence microscope (Zeiss AxioObserver

Z1) using similar methods as Potter et al. [443]. The images were centered on the

microelectrode hole and 4 by 4 mosaic tiles of 10x images were captured. The mosaic tiles

were stitched together and the final images were exported as 16-bit tiff files. Exposure

times were kept constant for each stain quantified.

3.3.11 Quantification of Immunohistochemical Markers

Histological images were analyzed using a combination of custom-built Matlab

GUIs that leverage the Image Processing Toolbox. We have previously published analyses with the program MINUTE to analyze histological images of neuroinflammation [6, 28,

443]. Here, we used an updated version of the Matlab GUI named SECOND. Differences

between MINUTE and SECOND are elucidated in the supplemental Section 8.1.2. The

fluorescence intensity of a given IHC stain was measured as a function of distance from the edge of the explanted microelectrode track [443]. Mean pixel intensities were

calculated within concentric rings spaced at 5 µm distance intervals from the hole. For

continuous staining the response as a function of distance is normalized to the background

expression level defined at 600-650 µm away from the hole. For stains with constitutive

expression (e.g. GFAP), normalization is set to one, whereas for non-natively expressed

stains in the brain (e.g. IgG) normalization is set to zero. The fluorescence intensity of

three to six tissue sections from the cortex at the approximate depth of the best 8 channels

were averaged together for each animal.

3.3.12 Quantification of Neuronal Density

Neuronal density was assessed with an in-house Matlab code called AfterNeuN.

After defining the implant hole, tissue edges, and tissue artifacts with SECOND, a blinded

user clicked the position of every NeuN-positive cell out to 550 µm from the

microelectrode hole. The program quantified the minimum distance of each click from the

edge of the microelectrode hole and number of clicks within 50 µm concentric bins out to

550 µm from the microelectrode hole. Counts were divided by bin area to assess neuronal

density. Density values were divided by neuronal density in the 500-550 µm to assess percentage of background neuronal density. The percentage of background neuronal density of three to six tissue sections from the cortex at the approximate depth of the best

8 channels were averaged together for each animal.

3.3.13 Immunohistochemistry Statistics

The normalized fluorescence intensity of 50-100 µm bins was compared between groups using General Linear Model ANOVA in the software program Minitab. The average normalized fluorescence intensity of the 3-6 slices from each animal was treated as an independent measurement.

3.3.14 Neuronal Density Statistics

The percentage of background density values of 50 µm bins were compared

between groups using General Linear Model ANOVA in the software program Minitab.

The average percentage of background density of the 3-6 slices from each animal was

treated as an independent measurement. Additionally, neuronal densities were compared

against background densities. The neuronal densities of 50 µm bins were compared against

the background bin (500-550 µm) groups using General Linear Model ANOVA in the

software program Minitab. The average neuronal density of the 3-6 slices from each

animal was treated as an independent measurement.

3.4 Results

3.4.1 Recording performance of intracortical microelectrodes in Cd14-/- mice

The number of single units detected per working channel, percentage of working

channels detecting single units, single unit signal to noise ratio, single unit amplitude, and

noise were metrics plotted versus time to compare recording performance between Cd14-/-

and wildtype mice implanted with identical NeuroNexus microelectrodes. Statistical

comparisons were made between treatment groups (Cd14-/- vs. wildtype for entire study length as a whole; ξ indicates significance), time range (acute vs. chronic for both conditions together, as a metric of change over time; @ indicates significance), and the interaction between animal group and time range (i.e. animal group crossed with time range). For animal group crossed with time range, we will only discuss relevant comparisons, namely: 1) Cd14-/- acute versus Cd14-/- chronic; $ indicates significance, 2)

wildtype acute versus wildtype chronic; % indicates significance, 3) Cd14-/- acute versus

wildtype acute; * indicates significance, and 4) Cd14-/- chronic versus wildtype chronic; δ

indicates significance. The acute time range includes the first two weeks of recording (days

0-11) [369], and the chronic time range includes the third through sixteenth weeks of

recording (days 16-109) [11]. P values are summarized in Table 6. P values are

unavailable for the comparisons of combinations of treatment group and time range, so

>0.05 and <0.05 are listed to represent overlapping and non-overlapping 95% confidence

intervals, respectively.

Table 6. Statistical summary of the recording performance metrics compared between Cd14−/− and WT mice. P-values are shown for differences between the various subcategories assessed in the general linear model. Significant values <0.05 are highlighted in grey.

3.4.1.1 Number of single units per working channel for intracortical microelectrodes in

Cd14-/- mice and wildtype controls

The number of units per working channel is displayed as mean ± standard error for

Cd14-/- (N = 6-9) and wildtype mice (N =3-5) over a 16-week time range (Figure 25A). A

full breakdown of daily sample size is located in Table 12. When comparing the entire

time range of the study, the number of units per working channel was not significantly

different between Cd14-/- mice and wildtype mice (Table 6). However, there was a

significant difference (p<0.001) when comparing the acute versus chronic time ranges,

irrespective of the animal condition (Figure 25A, @ indicates significance; Table 6). As a whole, more units were detected at acute than chronic time points, indicating a decay in the quantity of obtained signal with time. The same trend of more units detected per 96

channel at acute time points was seen for Cd14-/- mice ($, p<0.05), but not for wildtype

mice (Figure 25A; Table 6), which started lower than Cd14-/- mice. Additionally, Cd14-/-

mice exhibited a significantly higher (* p<0.05) number of units per working channel than

wildtype mice over the acute time range (Error! Reference source not found.). However,

the number of units per channel for Cd14-/- mice and wildtype mice were similar over the

chronic time range (weeks 3-16, Figure 25A; Table 6). Finally, of note, the interaction

between animal condition and time range factors was significant (p<0.001, Table 6).

Figure 25. Recording performance of intracortical microelectrodes in Cd14−/− mice versus wildtype mice. The number of single units detected per working channel (A), percentage of working channels detecting

single units (B), single unit SNR (C), single unit amplitude (D), and noise (E) were plotted versus time, to compare recording performance between Cd14−/− mice and wildtype mice implanted with identical NeuroNexus microelectrodes. Statistical comparisons were made as a function of time and treatment condition, both within and across groups. N for each plot varies and can be found in the text for the corresponding section. Statistical comparisons were made between treatment groups (Cd14−/− versus wildtype for entire study length as a whole; ξ indicates significance), time range (acute versus chronic for both conditions together, as a metric of change over time; @ indicates significance), and treatment group crossed with time range. For treatment group crossed with time range, we will only discuss relevant comparisons, namely: (1) Cd14−/− acute versus Cd14−/− chronic; $ indicates significance, (2) wildtype acute versus wildtype chronic; % indicates significance, (3) Cd14−/− acute versus wildtype acute; * indicates significance, and (4) Cd14−/− chronic versus wildtype chronic; δ indicates significance.

3.4.1.2 Percentage of channels detecting single units for intracortical microelectrodes in

Cd14-/- mice and wildtype controls

The percentage of channels detecting single units is displayed as mean ± standard

error for Cd14-/- (N = 6-9) and wildtype mice (N =3-5) over a 16-week time range (Figure

25B). A full breakdown of daily sample size is located in Table 12. When comparing the

entire time range of the study, the percentage of channels detecting single units was not significantly different between Cd14-/- mice and wildtype mice (Table 6). However, there

was a significant difference (p<0.05) when comparing the acute versus chronic time ranges,

irrespective of the animal condition (Figure 25B, @ indicates significance; Table 6). As

a whole, a higher percentage of channels were detecting single units at acute than chronic

time points, again indicating a decay in the quantity of obtained signal with time (Table

6). The same trend of a higher percentage of channels detecting single units at acute time

points than chronic time points was seen for Cd14-/- mice ($, p<0.05), but not for wildtype

mice (Figure 25B; Table 6). In addition, Cd14-/- mice exhibited a significantly higher (*

p<0.05) percentage of channels detecting single units than wildtype mice over the acute

time range (Table 6). However, the percentage of channels detecting single units for Cd14-

/- mice and wildtype mice were similar over the chronic time range (weeks 3-16, Figure

25B; Table 6). Finally, the interaction between animal condition and time range factors

was significant (p<0.001, Table 6).

3.4.1.3 Signal to noise ratio for intracortical microelectrodes in Cd14-/- mice and wildtype

controls

Single unit signal to noise ratio (SNR) is displayed as mean ± standard error for

Cd14-/- (N = 4-9) and wildtype mice (N =1-5) over a 16-week time range (Figure 25C). A

full breakdown of daily sample size is located in Table 13. SNR was significantly higher

(p<0.05) in Cd14-/- mice than in wildtype mice across the entire time range of the study

(Figure 25C; ξ indicates significance; Table 6). There was also a significant difference

(p<0.001) when comparing the acute versus chronic time ranges, irrespective of the animal

condition (Figure 25C, @ indicates significance; Table 6). SNR was overall significantly

higher over the acute time range than over the chronic time range (Table 6). Specifically,

SNR decreased significantly ($ p<0.05) for Cd14-/- mice from the acute time range to the

chronic time range (Figure 25C; Table 6) and SNR remained consistent for wildtype mice

between the acute and chronic time ranges (Figure 25C; Table 6). The SNR was

significantly higher (p<0.05) for Cd14-/- mice than for wild type mice over the acute time range (Figure 25C; * indicates significance; Table 6). However, SNR was similar between Cd14-/- mice and wildtype mice over the chronic time range. Finally, the

interaction between animal condition and time range factors was significant (p<0.05, Table

6).

3.4.1.4 Single unit amplitude for intracortical microelectrodes in Cd14-/- mice and wildtype

controls

Single unit amplitude is displayed as mean ± standard error for Cd14-/- (N = 4-9) and wildtype mice (N =1-5) over a 16-week time range (Figure 25D). A full breakdown of daily sample size is located in Table 13. When comparing the entire time range of the study, single unit amplitude was not significantly different between Cd14-/- mice and

wildtype mice (Table 6). When comparing the acute versus chronic time ranges for single

unit amplitude, irrespective of the animal condition, overall acute and chronic time ranges

were not significantly different (Figure 25D; Table 6). Further, Cd14-/- and wildtype mice

did not exhibit any significant changes from the acute range to the chronic time range

(Table 6). Finally, there were no significant differences in the single unit amplitude for

Cd14-/- versus wildtype mice at either the acute or chronic time intervals (Table 6).

3.4.1.5 Noise for intracortical microelectrodes in Cd14-/- mice and wildtype controls

Noise is displayed as mean ± standard error for Cd14-/- (N = 6-9) and wildtype mice

(N =3-5) over a 16-week time range (Figure 25E). A full breakdown of daily sample size

is located in Table 12.When comparing the entire time range of the study, noise was not

significantly different between Cd14-/- and wildtype mice (Figure 25E; Table 6).

However, there was also a significant difference (p<0.01) between the acute and chronic time ranges, irrespective of the animal condition (Figure 25E, @ indicates significance;

Table 6). Specifically, noise was overall significantly higher (@ p<0.01) over the chronic time group than over the acute time group (Table 6). Despite this, neither the Cd14-/- mice

nor the wildtype mice exhibited any significant changes from the acute range to the chronic

time range (Figure 25E; Table 6). Finally, there were also no significant differences

100

between Cd14-/- mice and wildtype mice at either the acute or chronic time point (Figure

25E; Table 6).

3.4.2 Recording performance of intracortical microelectrodes in wildtype mice

administered IAXO-101

Recording performance for wildtype mice administered IAXO-101 was presented

in a similar format to the mice in Section 3.4.1 and Figure 25. Once again, the number of

single units detected per working channel, percentage of working channels detecting single

units, single unit signal to noise ratio, single unit amplitude, and noise were plotted versus

time to compare recording performance between wildtype mice receiving IAXO-101,

compared to control wildtype mice not receiving IAXO-101. Here, mice were implanted

with identical NeuroNexus microelectrodes to the above section. Untreated wildtype

animals displayed here are the same animals used as controls above. Statistical

comparisons were made between treatment groups (wildtype mice administered IAXO-101

vs. wildtype for entire study length as a whole; ξ indicates significance), time range (acute vs. chronic for both conditions together, as a metric of change over time; @ indicates significance), and animal group crossed with time range. For animal group crossed with

time range, we will discuss: 1) wildtype mice administered IAXO-101 acute versus wildtype mice administered IAXO-101 chronic; $ indicates significance, 2) wildtype acute versus wildtype chronic; % indicates significance, 3) wildtype mice administered IAXO-

101 acute untreated wildtype acute; * indicates significance, and 4) wildtype mice administered IAXO-101 chronic versus untreated wildtype chronic; δ indicates

significance. As used above, the acute time range includes the first two weeks of recording

(days 0-11) [369], and the chronic time range includes the third through sixteenth weeks

101

of recording (days 16-109) [11]. P values are summarized in Table 7. P values are

unavailable for the comparisons of combinations of treatment group and time range, so

>0.05 and <0.05 are listed to represent overlapping and non-overlapping 95% confidence

intervals, respectively.

Table 7. Statistical summary of the recording performance metrics compared between IAXO-101 and WT mice. P-values are shown for differences between the various subcategories assessed in the general linear model. Significant values <0.05 are highlighted in bold.

3.4.2.1 Number of single units per working channel for intracortical microelectrodes in

mice administered IAXO-101 and wildtype controls

The number of units per working channel is displayed as mean ± standard error for

mice administered IAXO-101 (N = 3-5) and wildtype mice (N =3-5) over a 16-week time

range (Figure 26A). A full breakdown of daily sample size is located in Table 12.When

comparing the entire time range of the study, the number of units per working channel was

not significantly different between mice receiving IAXO-101 and wildtype mice (Table

7). There were also no significant differences when comparing the acute versus chronic

time ranges, irrespective of the animal condition (Table 7). The same lack of significance

was noted for both wildtype mice administered IAXO-101 and wildtype mice not receiving

IAXO-101, when comparing acute versus chronic units per working channel, within a treatment group (Table 7). Finally, there were no significant difference noted at a given time interval, across treatment group (Table 7).

102

Figure 26. Recording performance of intracortical microelectrodes in wildtype mice treated with IAXO-101 versus untreated wildtype mice. The number of single units detected per working channel (A), percentage of working channels detecting single units (B), single unit SNR (C), single unit amplitude (D), and noise (E) were plotted versus time, to compare recording performance between wildtype mice administered IAXO-101 (green) mice and wildtype mice (blue) implanted with identical NeuroNexus microelectrodes. Statistical comparisons were made as a function of time and treatment condition, both within and across groups. N for each plot varies and can be found in the text for the corresponding section. Statistical comparisons were made between treatment groups (IAXO-101 versus wildtype for entire study length as a whole; ξ indicates significance), time range (acute versus chronic for both conditions together, as a metric of change over time; @ indicates significance), and treatment group crossed with time range. For treatment group crossed with time range, we will only discuss relevant comparisons, namely: (1) IAXO-101

103

acute versus IAXO-101 chronic; $ indicates significance, (2) wildtype acute versus wildtype chronic; % indicates significance, (3) IAXO-101 acute versus wildtype acute; * indicates significance, and (4) IAXO- 101 chronic versus wildtype chronic; δ indicates significance. 3.4.2.2 Percentage of channels detecting single units for intracortical microelectrodes in

mice administered IAXO-101 and wildtype controls

The percentage of channels detecting single units is displayed as mean ± standard

error for mice administered IAXO-101 (N = 3-5) and wildtype mice (N =3-5) over a 16-

week time range (Figure 26B). A full breakdown of daily sample size can be found in

Table 12. When comparing the entire time range of the study, the percentage of channels

detecting single units was significantly higher for mice administered IAXO-101 than untreated wildtype mice (p<0.05; Figure 26B, ξ indicates significance; Table 7).

However, there was no significant difference comparing the acute versus chronic time ranges, irrespective of the animal condition (Table 7). There was also a lack of significance for both wildtype mice administered IAXO-101 and wildtype mice not receiving IAXO-101, when comparing acute versus chronic percent channels detecting single units, within a treatment group (Table 7). Finally, there were also no significant difference noted at a given time interval, across treatment group (Table 7). In summary, statistical breakdowns of subsets of comparisons failed to identify statistical significance, but the data set as a whole between treatment conditions demonstrated a significant improvement in recording performance metric for IAXO-101 treatment.

3.4.2.3 Signal to noise ratio for intracortical microelectrodes in mice administered IAXO-

101 and wildtype controls

Single unit signal to noise ratio (SNR) is displayed as mean ± standard error for mice administered IAXO-101 (N = 2-5) and wildtype mice (N =1-5) over a 16-week time range (Figure 26C). A full breakdown of daily sample size can be found in

104

Table 13. SNR was not significantly different between wildtype mice receiving

IAXO-101 and wildtype mice across the entire time range of the study (Table 7).

However, SNR was overall significantly higher ($ p<0.01) over the acute time group than

over the chronic time group, irrespective of the animal condition (Figure 26C, @ indicates

significance; Table 7). Within a treatment group, neither mice receiving IAXO-101 or

untreated wildtype mice demonstrated a significant difference for acute SNR versus

chronic SNR, within the treatment condition (Table 7). Finally, SNR was statistically insignificant between treatment groups at both the acute and chronic time intervals (Table

7).

3.4.2.4 Single unit amplitude for intracortical microelectrodes in mice administered IAXO-

101 and wildtype controls

Single unit amplitude is displayed as mean ± standard error for mice administered

IAXO-101 (N = 2-5) and wildtype mice (N =1-5) over a 16-week time range (Figure 26D).

A full breakdown of daily sample size can be found in Table 13. Single unit amplitude

was significantly higher (ξ p<0.05) in wildtype mice receiving IAXO-101 compared to untreated wildtype mice across the entire time range of the study (Table 7). The single unit amplitude was overall not significantly different between the acute time group and the chronic time group, irrespective of the animal condition (Table 7). This was not unexpected, seeing as though there were no significant changes from acute to chronic time groupings for either of the two treatment groups (Table 7). However, animals treated with

IAXO-101 displayed a significantly higher single unit amplitude than untreated wildtype animals at both acute (p<0.05, *indicates significance) and chronic time intervals (p<0.05,

δ indicates significance) (Table 7).

105

3.4.2.5 Noise for intracortical microelectrodes in mice administered IAXO-101 and

wildtype controls

Noise is displayed as mean ± standard error for mice administered IAXO-101 (N =

3-5) and wildtype mice (N =3-5) over a 16-week time range (Figure 26E). A full

breakdown of daily sample size can be found in Table 12. Noise is significantly higher (ξ

p<0.005) in mice administered IAXO-101 versus untreated wildtype mice, over entire time

range of the study (Table 7). Noise was overall significantly higher (@ p<0.001) over the

chronic time group than over the acute time group, irrespective of the animal treatment

condition (Figure 26E; Table 7). However, there were no significant differences in noise

levels between acute or chronic time intervals for either treatment groups (Table 7).

Statistical comparison also revealed that mice administered IAXO-101 displayed

significantly higher noise levels than untreated wildtype mice (p<0.05) at both the acute (*

indicates significance) and chronic (δ indicates significance) time intervals (Table 7).

3.4.3 Immunohistochemical evaluation of Cd14-/- mice implanted with intracortical

microelectrodes

3.4.3.1 Neuronal Nuclei

It is believed that neuronal cell bodies must be within the first 50 to 140 µm of the

intracortical microelectrode in order to maintain recordings of single unit action potentials

from individual neurons [117]. Therefore, we quantified the number of neurons around the

implants utilizing the NeuN antibody, which selectively stains for neuronal nuclei [445].

Here, neuronal survival at the microelectrode-tissue interface was evaluated as percentage of background density with respect to distance from the hole left from explanted the microelectrode hole (µm) (Figure 3A) [369]. Both Cd14-/- mice (N=9) and wildtype mice

106

(N=5) exhibit trends of decreased neuronal survival close to the microelectrode hole with

a gradual increase in neuronal survival approaching background density. The percentage

of background density is only significantly different between Cd14-/- and wildtype mice

between 450 and 500 µm from the microelectrode hole, *p<0.05. Despite trends of higher

neuronal survival in Cd14-/- mice from 0 to 450 µm, there are no additional significant

differences between Cd14-/- and wildtype. Neuronal density is significantly different from background for Cd14-/- mice between 0 and 50 µm from the microelectrode hole, # p<0.05.

Despite trends of neuronal dieback out to 350 µm from the microelectrode hole, neuronal

survival for wildtype mice is not significantly different from background.

3.4.3.2 Astrocytes

Glial fibrillary acidic protein (GFAP) was used to assess both immature and mature

resting or activated astrocytes [446]. Here, astrocytic encapsulation was evaluated as

GFAP activation with respect to distance from the microelectrode hole (Figure 27B). Both

Cd14-/- (N=9) and wildtype mice (N=5) exhibit elevated GFAP expression close to the microelectrode hole with decaying expression further away from the microelectrode hole.

No significant differences were observed between Cd14-/- and wildtype mice, regardless of the distance from the microelectrode surface.

3.4.3.3 Activated Microglia and Macrophages

Microglia/macrophages are a major component of the innate immune response in the CNS. Microglia/macrophage-released inflammatory factors sustain the innate immune/inflammatory response and recruit additional cell types. The cytosolic antigen

CD68 is found in the lysosomal compartment of activated microglia and macrophages

[447]. Microglial activation at the microelectrode-tissue interface was evaluated as CD68

107

expression with respect to distance from the microelectrode hole (Figure 27C). Both

Cd14-/- (N=9) and wildtype mice (N=5) exhibit elevated CD68 expression close to the

microelectrode hole with decaying expression further away from the microelectrode hole.

Cd14-/- mice express significantly less CD68 between 100 and 500 µm from the

microelectrode hole. No significant differences were seen within the first 100 µm from the microelectrode hole.

108

Figure 27. Immunohistochemical evaluation of Cd14−/− mice implanted with intracortical microelectrodes. (A) Neuronal survival at the microelectrode-tissue interface evaluated as percentage of background density with respect to distance from the microelectrode hole (μm). The percentage of background density is significantly different between Cd14−/− and wildtype mice between 450 and 500 μm from the microelectrode hole, * p < 0.05. Neuronal density is significantly different from background for Cd14−/− mice between 0 and 50 μm from the microelectrode hole, # p < 0.05. (B) Astrocytic encapsulation evaluated as GFAP activation with respect to distance from the microelectrode hole (μm). No significant differences were observed between Cd14−/− and wildtype mice. (C) Microglial activation evaluated as CD68 expression with respect to distance from the microelectrode hole (μm). Cd14−/− mice express significantly less CD68 between 100 and 500 μm from the microelectrode hole. (D) BBB permeability evaluated as IgG expression with respect to distance from the microelectrode hole. No significant differences were observed between Cd14−/− and wildtype mice. Cd14−/−: N = 9; wildtype: N = 5.

109

3.4.3.4 Blood-brain barrier permeability

Several labs have utilized the amount of IgG present within the tissue surrounding

explanted microelectrodes to assess the integrity of the blood-brain barrier [6, 130].

Therefore, here, the blood-brain barrier permeability was evaluated as IgG expression with respect to distance from the microelectrode hole (Figure 27D). Both Cd14-/- (N=9) and

wildtype mice (N=5) exhibit elevated IgG expression close to the microelectrode hole with

decaying expression further away from the microelectrode hole. No significant differences

were observed between Cd14-/- and wildtype mice.

3.4.4 Immunohistochemical evaluation of mice administered IAXO-101 and implanted

with intracortical microelectrodes

3.4.4.1 Neuronal Nuclei

Neuronal survival at the microelectrode-tissue interface was evaluated as

percentage of background density with respect to distance from the microelectrode hole

(Figure 28A). Both mice administered IAXO-101 (N=5) and wildtype mice (N=5) exhibit

trends of decreased neuronal survival close to the microelectrode hole with a gradual

increase in neuronal survival approaching background density. No significant differences

were observed between mice administered IAXO-101 and wildtype mice or between either

of the conditions and background.

3.4.4.2 Astrocytes

Astrocytic encapsulation was evaluated using positive GFAP activation/expression

with respect to distance from the microelectrode hole. Both mice administered IAXO-101

(N=5) and wildtype mice (N=5) exhibit elevated GFAP expression close to the

microelectrode hole with decaying expression further away from the microelectrode hole

110

(Figure 28B). No significant differences were observed between mice administered

IAXO-101 and wildtype mice.

Figure 28. Immunohistochemical evaluation of mice administered IAXO-101 and implanted with intracortical microelectrodes. (A) Neuronal survival at the microelectrode-tissue interface evaluated as percentage of background density with respect to distance from the microelectrode hole (μm). No significant differences were observed between mice administered IAXO-101 and wildtype mice or between either condition with background. (B) Astrocytic encapsulation evaluated as GFAP activation with respect to distance from the microelectrode hole (μm). No significant differences were observed between mice administered IAXO-101 and wildtype mice. (C) Microglial activation evaluated as CD68 expression with respect to distance from the microelectrode hole (μm). No significant differences were observed between

111

mice administered IAXO-101 and wildtype mice. (D) BBB permeability evaluated as IgG expression with respect to distance from the microelectrode hole. No significant differences were observed between mice administered IAXO-101 and wildtype mice. IAXO-101, WT: N = 5.

3.4.4.3 Activated Microglia and Macrophages

Microglial activation was evaluated using positive CD68 expression with respect

to distance from the microelectrode hole (Figure 28C). Both mice administered IAXO-

101 (N=5) and wildtype mice (N=5) exhibit elevated CD68 expression close to the

microelectrode hole with decaying expression further away from the microelectrode hole.

No significant differences were observed between mice administered IAXO-101 and wildtype mice.

3.4.4.4 Blood-brain barrier permeability

Blood-brain barrier permeability was evaluated using positive IgG expression with respect to distance from the microelectrode hole. Both mice administered IAXO-101

(N=5) and wildtype mice (N=5) exhibit elevated IgG expression close to the microelectrode hole with decaying expression further away from the microelectrode hole

(Figure 28D). No significant differences were observed between mice administered

IAXO-101 and wildtype mice.

3.5 Discussion

While strategies to improve intracortical microelectrode integration and performance by broadly inhibiting the neuroinflammatory response have demonstrated promising results, they have also been prone to harmful side effects [12, 13, 16]. Thus, in order to more safely inhibit microelectrode-induced neuroinflammation, we are investigating therapeutic approaches with cellular or subcellular specificity and have found

CD14 to be a viable target. The CD14 molecule is an innate immunity receptor involved

112

in the recognition of several PAMPs and DAMPs that may be present at the microelectrode

tissue interface to promote inflammation [24-26, 300, 308, 310-312, 437]. In the present

study, we examined the effects of inhibiting CD14 on intracortical microelectrode

performance and tissue integration using both a knockout mouse model and systemic

administration of a small molecule CD14 inhibitor.

Upon evaluation of electrophysiology, Cd14-/- mice exhibit higher numbers of units per channel and higher percentages of channels detecting single units over the acute range, but not the chronic time range (Figure 25A/B). Cd14-/- mice exhibit higher single unit

SNR over the entire time rage of the study and specifically during the acute time range

(Figure 25C). Further electrophysiological evaluation revealed that mice administered

IAXO-101 exhibited significant improvements in the percentage of channels detecting single units and single unit amplitude over the entire time range of the study (Figure

26B/D). Additionally, mice administered IAXO-101 exhibited significantly higher noise over the entire time range of the study (Figure 26E).

Significant improvements in the number of single units per working channel and percentages of channels detecting single units in CD14-/- mice suggest that the complete

absence of CD14 improves recording quality over the first two weeks after implantation.

Thus, CD14 may play a role in acute inflammatory mechanisms that are detrimental to

recording performance. Enhanced inflammatory activation can potentially reduce neuronal

survival, damage neuronal health, decrease neuronal firing, increase glial encapsulation,

and facilitate microelectrode material breakdown [1, 11, 129]. Pro-inflammatory

activation of CD14 has demonstrated detrimental effects on intracortical microelectrode

recording quality and tissue integration in two studies. First, activation of CD14 via

113

administration of its primary ligand LPS in a rat model led to reduced recording quality

and neuronal survival [27]. Exacerbation of inflammation via LPS at the acute time point

may account for the poor recording performance in wildtype mice at the acute time point.

Cd14-/- mice would have a diminished inflammatory response to LPS [304], which could

result in better recording performance. More recently, when comparing neural probe

sterilization methods, we demonstrated that residual endotoxin levels coincided with

greater neuronal dieback, microglial activation, astrocytic encapsulation, and blood-brain

barrier permeability at 2 weeks after implantation, but not 16 weeks after implantation [28].

However, in that study, ethylene oxide sterilized microelectrodes (same treatment used here) exhibited the lowest level of neuroinflammation even at acute time points.

Lipopolysaccharide is not the only ligand of CD14 that may be contributing to detrimental recording performance as CD14 also recognizes DAMPs, or endogenous molecules released in the event of tissue injury [24]. CD14 plays in the recognition of

HMGB1, heat shock proteins, apoptotic cells, and necrotic cells [24, 25, 300, 308, 310-

312, 437]. Damage caused by the implantation and chronic presence of the microelectrode may produce several of these factors at the microelectrode-tissue interface [11, 24, 25, 107,

308, 310]. Additionally, CD14 is a co-receptor to several TLRs, which recognize other endogenous DAMPs, including fibrinogen, fibronectin and other endogenous molecules

[44-46]. These proteins can be released into cortical tissue from vascular damage caused by the implantation and chronic presence of the implanted microelectrodes [11, 28, 136].

Activation of TLRs often involves the assistance of CD14 [53, 68-70]. Overall, a wide variety of ligands at the microelectrode tissue interface may activate CD14-mediated

114

pathways to promote neuroinflammation, which can be detrimental to recording

performance.

The lack of differences in the number of units per channel and percentages of

channels detecting single units between Cd14-/- mice and wildtype mice suggests that

CD14-independent inflammatory responses dominate at later time points, or that complete

inhibition triggers the activation of a redundant pathway that is not activated with partial

inhibition with IAXO-101.

Additionally, the Cd14-/- group exhibited significantly higher SNR over the entire

time range of the study, in contrast to the number of units per working channel and

percentage of channels detecting single units. Single unit SNR is only evaluated on

channels detecting single units. Isolated units below SNR of 3 are excluded from analysis.

If lower amplitude neurons die, stop firing, or become separated from the microelectrode

by a glial scar, remaining large neurons close to a recording contact may dominate the

metric, despite lower numbers of units or channels detecting units. SNR can be broken up

into the influence of the signal and noise. Despite significant improvements in SNR, neither single unit amplitude nor noise exhibited significant changes in Cd14-/- mice from

wildtype mice.

When comparing Cd14-/- and wildtype mice, the number of single units per working

channel, percentage of channels detecting single units, and single unit SNR all exhibit

significant decreases from acute to chronic time ranges irrespective of the animal condition.

Upon examining combinations of time group and treatment group, Cd14-/- mice exhibit a

significant decrease from the acute range to the chronic range. The decline in recording

performance can be attributed to several phenomena, including but not limited to a decrease

115

in the number of healthy, firing neurons within 50 µm of the microelectrode, increase of

glial encapsulation around the microelectrode, increase of noise from various biological

and electrical sources, or material breakdown of microelectrode and insulating components

[1, 2, 117, 127, 448]. Neuronal dieback and glial encapsulation are evident in both

treatment groups at 16 weeks after implantation (Figure 27A/B). Histology was not

evaluated at or before 2 weeks after implantation to determine the cause of the change in

units per channel from acute to chronic. We and others have observed that rats implanted

with neural probes exhibit significant neuronal dieback and elevated astrocytic

encapsulation at two weeks after implantation [1]. Noise significantly increases from the

acute to the chronic time point (Figure 25E).

An overall increase in noise was shown from the acute time range to the chronic

time range, but neither treatment group exhibited changes from the acute to the chronic

time range. There were no differences between groups across the entire time range or within acute or chronic time ranges. Noise can arise from external electrical sources, high impedance tissue in the form of thermal noise, and biological sources such as muscle activity and nearby neuronal firing [116, 448-451]. Electrical noise sources should be similar among all mice since all recordings took place in the same environment with the same equipment. Biological noise consists primarily of the overlapping spiking activity from many distant neurons beyond 100-150 µm from the recording site [448, 452-454].

Large-scale changes in neuronal populations would likely be necessary to affect biological noise. Significant dieback of neurons in the environment around the implanted microelectrode could potentially affect biological noise [129, 369]. Neuronal loss would theoretically result in less biological noise, so other factors may describe the significant

116

increase in overall noise from the acute to the chronic time point. Additionally, biological

noise can vary with the location of the microelectrode in the brain and the degree of correlation between neuronal firing [448]. Variation in implantation site may affect the variation in noise level between mice. Thermal noise will vary based on the degree of encapsulation and thus can vary from mouse to mouse. With an increase in glial encapsulation around an implanted intracortical microelectrode over time [369], an

increase in thermal noise over time would make sense.

It is noteworthy that mice administered IAXO-101 exhibited significant

improvements in percentage of channels detecting single units over wildtype controls

across the entire time range of the study. Attenuation of CD14 signaling promotes the ability to detect and resolve single unit activity over a 16-week time range. Fully intact

CD14 signaling has a detrimental effect on the ability to detect single unit activity over the full 16-week time range. Attenuating CD14 with a small molecule antagonist improves recording over the entire time range (Figure 26A) whereas the complete absence of CD14 in a knockout mouse only improves recording in the acute time range (Figure 25A). Some degree of CD14 signaling in the chronic time range may be beneficial to detecting single unit activity. Perhaps CD14 is involved in wound healing mechanisms that promote a stable environment for single unit recordings.

Single unit SNR did not exhibit any differences between mice administered IAXO-

101 and wildtype mice. However, there was an overall trend of decreasing SNR from the

acute time range to the chronic time range as single unit SNR is affected by single unit

amplitude and noise. Single unit amplitude did not exhibit any changes from the acute

time range to the chronic time range, but noise exhibited a significant increase from the

117

acute time range to the chronic time range. SNR decreased with increasing noise and

constant amplitude. Additionally, single unit amplitude and noise were both significantly higher for mice administered IAXO-101. A similar increase in numerator and denominator of SNR resulted in no difference in SNR between mice administered IAXO-101 and wildtype mice.

Single unit amplitude was significantly higher in mice administered IAXO-101

than wildtype controls across the entire 16-week time range. High single unit amplitude

suggests the presence of large, healthy, firing neurons in close proximity to the

microelectrode contacts [117]. Perhaps small molecule inhibition of CD14 provides

neuroprotective benefits. Pro-inflammatory activation of CD14 with LPS results in reduced neuronal survival at the microelectrode tissue interface [27, 28]. Mice

administered IAXO-101 did not exhibit higher neuronal survival at the microelectrode

tissue interface; however, histology is only evaluated at 16 weeks after implantation

(Figure 28A). Amplitude appears to drop in the 15th and 16th weeks, so neuronal survival

may have been higher prior to this time point (Figure 26D). High single unit amplitude

may also suggest low impedance between the microelectrode and surrounding neurons as

glial encapsulation increases impedance at the microelectrode tissue interface [10, 75].

Attentuation of CD14 signalling can reduce glial encapsulation. At 16 weeks after implantation, there were no significant differnces in astrocytic encapsulation between mice administered IAXO-101 and wildtype controls (Figure 28B). Mice administered IAXO-

101 exhibited a trend of increased astrocytic encapsulation and therefore we suggest that reduced encapsulation and impedance are not likely the cause(s) of improved single unit amplitude.

118

Mice administered IAXO-101 exhibited significantly higher noise across the entire

16-week time range. As discussed above, noise can be generated from several different

sources, including external electrical sources, high impedance tissue in the form of thermal noise, and biological sources such as muscle activity and nearby neuronal firing [116, 448-

451]. Although there is no significant difference in astrocytic encapsulation, mice

administered IAXO-101 exhibited a trend of increased astrocytic encapsulation at 16 weeks

after implantation. Elevated encapsulation may increase the level of thermal noise.

Despite no significant differences in neuronal survival in mice administered IAXO-101

and wildtype mice, there is a trend of higher neuronal survival 50-100 µm away from the

microelectrode hole. Biological noise is affected by neurons out to 100-150 µm from the

microelectrode interface [70, 80-82].

Aditionally, there is an overall trend of increasing noise from the acute to the

chronic time group. An increase in glial encapsulation around implanted miceloeleectrodes

over time may lead to increased thermal noise to account for this trend [106, 369, 449,

450]. Potter et al. demonstrated an increase in neuronal density around microelectrodes

implanted in rats between 2 and 4 weeks after implantation [369]. Stabilization of neuronal

populations beyond the limit of single unit detection (50 µm) after the acute inflammatory

phase can lead to an increase in biological noise [70, 80-82].

The Cd14-/- mice did not exhibit any changes in neuronal survival, astrocytic

encapsulation, or blood-brain barrier permeability. Compared to recording performance,

complete removal of CD14 does not appear to have an effect at chronic time points.

Endpoint histology was only evaluated at 16 weeks after implantation. Future studies

119

evaluating histology around 2 weeks after microelectrode implantation should be more

instructive about the mechanisms causing enhanced recording performance.

The Cd14-/- mice exhibited significantly lower microglial activation over a range of

100-500 µm away from the microelectrode hole. This may indicate that without CD14,

microglia become less likely to become pro-inflammatory activated due to hindered

recognition of DAMPs and PAMPs at distances further away from the microelectrode

interface [50-58]. These differences likely have minimal effect on the neurons close to the

microelectrode interface. Again, histology evaluated within the first two weeks of

implantation would be more instructive about mechanisms causing improved recording

performance.

Mice administered IAXO-101 did not exhibit any significant differences versus

controls for any of the investigated stains, despite improvements in recording performance

over the full 16-week time range. Mice administered IAXO-101 exhibit better neuronal

health, firing, or network connectivity, despite neuronal survival similar to wildtype

controls. Additionally, the efficacy of the antibody NeuN in assessing neuronal cell loss

has been disputed [455].

Overall Cd14-/- mice and mice administered IAXO-101 exhibited different trends compared to wildtype control mice. Cd14-/- mice exhibited higher numbers of units per

channel and higher percentages of channels detecting single units at acute but not chronic

time ranges. Mice administered IAXO-101 exhibited higher percentages of channels

detecting single units across the entire time range but no significant differences in the

number of units per channel. Cd14-/- mice exhibited higher SNR across the entire time

range, whereas mice administered IAXO-101 exhibited no change in SNR. Cd14-/- mice

120

exhibited no differences compared to wildtype mice for single unit amplitude and noise.

Mice administered IAXO-101 exhibited significant increases over wildtype mice in both single unit amplitude and noise.

The main difference between Cd14-/- mice and mice administered IAXO-101, is that CD14 is completely absent from Cd14-/- mice. Mice administered IAXO-101 have

intact CD14 and a portion of the administered drug will reach the environment around the

implanted microelectrode. Some IAXO-101 will transiently bind with CD14 until it is cleared from the body. IAXO-101 is hypothesized to selectively target CD14 and competitively occupy the binding site for endotoxins and other ligands [374]. Binding of

IAXO-101 with CD14 reduces the transfer of endotoxin to TLR4 and its associated co- receptor MD-2 [374]. The interaction between IAXO-101 and DAMPs is not well understood. Many CD14 receptors will remain unbound to IAXO-101 and binding of

IAXO-101 with CD14 will be temporary. Thus, CD14 signaling should be attenuated, but not completely removed.

A limited amount of CD14 signaling may be beneficial for recording performance and tissue integration.CD14 may play a role in wound healing mechanisms that provide a stable interface for detecting single unit activity. Also, a limited amount of CD14 may be needed to protect the body against bacterial pathogens. Undetected bacteria may cause damage to neurons and supporting tissue before being cleared by other components of the immune system.

In order to obtain more consistent improvements in recording quality and tissue integration, other methods of inhibiting CD14 should be investigated. IAXO-101 (used here) is no longer the only CD14-TLR4 antagonist commercially available. Several other

121

IAXO compounds have since been made available. IAXO-102, a similar glycolipid molecule with a different functional group, has demonstrated neuroprotective effects in a model for ALS [377]. Additionally, IAXO-102 has been demonstrated to inhibit the development of aortic aneurysms [377]. IAXO-102 has also demonstrated success in inhibiting cerebral vasospasm after subarachnoid hemorrhaging. In addition to small molecules, monoclonal antibodies to CD14 may me an effective method of inhibiting

CD14 to improve microelectrode recording quality and tissue integration. CD14 monoclonal antibodies have been administered to healthy and diseased individuals in sepsis studies [385, 456]. The healthy individuals exhibited inhibition of LPS-induced gene expression, and the individuals with severe sepsis exhibited variable results that warranted further clinical investigation.

Full removal of CD14 exhibited time-dependent effects on recording performance.

Limiting the expression of CD14 on specific cellular sub-populations may improve recording performance more effectively. Our lab has previously established methods of distinguishing blood-derived and resident inflammatory cells using a bone marrow chimera model [144], and are currently examining the effects of knocking out CD14 on blood derived inflammatory cells and/or resident brain inflammatory cells on intracortical microelectrode recording performance and tissue integration.

Alternatively, other innate immunity receptors that work with CD14 may play roles that are more significant in intracortical microelectrode failure. For example, CD14 coordinates ligand binding for TLR2 and TLR4 [30, 36-38]. TLR2 and TLR4 have been linked to a wider variety of DAMPs, and thus may have more opportunities to promote inflammation in response to the damage caused by intracortical microelectrodes [207].

122

Therefore, we are also currently examining the effects of knocking out TLR2 or TLR4 in

the foreign body response to intracortical microelectrodes.

Finally, IAXO-101 is also commonly regarded as a TLR4 antagonist because it

inhibits the CD14 / TLR4 complex. Inhibiting other pathways associated with TLR4 may

be more beneficial to recording quality and tissue integration. MD-2 is another co-receptor

to TLR4 involved in the recognition of LPS. MD-2 focuses on PAMPs rather than DAMPs,

and is the main pharmacological target for sepsis [24, 457]. If bacterial endotoxins prove to be more relevant to the microelectrode-tissue interface than DAMPs, targeting MD-2

prove to be more beneficial. Other ligands to TLR4 that bind independently of CD14 and

MD-2 may also be involved in the foreign body response to intracortical microelectrodes.

There are a growing number of synthetic and natural TLR4 antagonists, including

derivatives of lipid A (a component of LPS), olive oil extracts, and curcumin [379]. In

fact, we have previously shown that curcumin improves neuronal survival and blood-brain

barrier stability around implanted intracortical microelectrodes [189]. Unfortunately, no

studies investigating the effect of curcumin on microelectrode recording performance have

been carried out to this date.

3.6 Conclusions

Complete removal of CD14 results in improvements in intracortical microelectrode

recording at acute, but not chronic time points post implantation. Complete removal of

CD14 reduces microglial activation distant from the implanted microelectrode, but does

not affect neuronal survival, astrocytic encapsulation, or blood-brain barrier permeability

at 16 weeks after implantation. Partial inhibition of CD14 signaling with a small molecule

antagonist results in improved recording performance over a 16-week time range. Partial

123

inhibition of CD14 with a small molecule antagonist did not affect neuronal survival,

microglial activation, astrocytic encapsulation, or blood-brain barrier permeability. Full

removal of CD14 signaling is beneficial over the acute time range to attenuate

inflammatory mechanisms, but some degree of CD14 signaling may be necessary over

chronic time ranges to facilitate wound healing mechanisms. A better understanding of the

mechanism and efficacy that facilitates CD14-mediated improvements in microelectrode

recording performance should be completed to further improves therapeutic outcomes, and

achieve more consistent improvements in microelectrode recording performance.

3.7 Acknowledgements

This work was supported in part by the Department of Biomedical Engineering and

Case School of Engineering at Case Western Reserve University through laboratory start-

up funds, the National Institute of Health, National Institute of Neurological Disorders and

Stroke, (Grant # 1R01NS082404-01A1), the NIH Neuroengineering Training Grant 5T-

32EB004314-14. Additional support was provided by the Presidential Early Career Award

for Scientists and Engineers (PECASE, JR. Capadona) and by Merit Review Award

B1495-R from the United States (US) Department of Veterans Affairs Rehabilitation

Research and Development Service. None of the funding sources aided in collection,

analysis and interpretation of the data, in writing of the manuscript, or in the decision to

submit the manuscript for publication. The authors have no conflict of interest related to this work to disclose. The contents do not represent the views of the U.S. Department of

Veterans Affairs or the United States Government.

124

Chapter 4

The role of toll-like receptor 2 and 4 innate immunity pathways in intracortical microelectrode induced neuroinflammation.

*The following chapter is from a manuscript under review, featuring contributions from: John K Hermann, Shushen Lin, Arielle Soffer, Chun Wong, Vishnupriya Srivastava, Jeremy Chang, Smrithi Sunil, Shruti Sudhakar, William Tomaswzeski, Grace Protasiewicz, Stephen Selkirk, MD, PhD, Robert Miller, PhD, and Jeffrey R Capadona, PhD

4.1 Abstract

We have recently demonstrated that partial inhibition of the cluster of

differentiation 14 (CD14) innate immunity co-receptor pathway improves the long-term performance of intracortical microelectrodes better than complete inhibition. We hypothesized that partial activation of the CD14 pathway was critical to a neuroprotective response to the injury associated with initial and sustained device implantation. Therefore, we investigated the role of two innate immunity receptors that closely interact with CD14 in inflammatory activation. We implanted silicon planar non-recording neural probes into knockout mice lacking TLR2 (Tlr2-/-), knockout mice lacking TLR4 (Tlr4-/-), and wildtype

(WT) control mice, and evaluated endpoint histology at 2 and 16 weeks after implantation.

The Tlr4-/- mice exhibited significantly lower BBB permeability at acute and chronic time

points, but also demonstrated significantly lower neuronal survival at the chronic time

point. Inhibition of the TLR2 pathway had no significant effect compared to control

animals. Additionally, when investigating the maturation of the neuroinflammatory

response from 2 to 16 weeks, transgenic knockout mice exhibited similar histological

trends to WT controls, except that knockout mice did not exhibit changes in microglia and

macrophage activation over time. Together, our results indicate that complete genetic

125

removal of TLR4 was detrimental to the integration of intracortical neural probes, while

inhibition of TLR2 had no impact within the tests performed in this study. Therefore,

approaches focusing on incomplete or acute inhibition of TLR4 may still improve

intracortical microelectrode integration and long term recording performance.

4.2 Introduction

Brain machine interfaces are a growing area of interest for basic research,

rehabilitation, and commercial applications [37, 38, 458]. Intracortical microelectrodes remain a high-resolution tool for extracting information from the brain [39], critical for current and future applications. Unfortunately, inconsistent recording performance remains a barrier to long-term utilization in any animal model, including humans [4, 459,

460].

The correlation between the neuroinflammatory response to intracortical microelectrodes and recording performance remains a commonly debated topic for over a decade [1, 129, 133]. The consensus of the field is that in order to maintain viable recordings, the integrity of both the implanted electrodes and the neural tissue must remain intact. Several labs have shown that delamination of the insulation layers or corrosion of the electrode contacts are common in both accelerated aging and upon explanting of a variety of microelectrode types [1, 2, 98, 99, 127, 137]. Additionally, the loss of neuronal cell bodies and dendrites within the distance required for single unit detection [117] is well documented [1]. While subtle difference exist across all microelectrode types, the typical response to intracortical microelectrodes can be generalized [1]; upon implantation of the microelectrodes, tissue and cells are damaged resulting in both wound healing and scar formation. Most importantly, the robust response from microglia and macrophages leads

126

to neuronal dieback, astrocytic encapsulation, and blood-brain barrier permeability, each

of which have been implicated in biological failure mechanisms of single unit recordings

from intracortical microelectrodes [1, 132, 136, 461].

As the failure modes of intracortical microelectrodes are further elucidated, one

mechanism that has been suggested to play a key role in several failure modes is oxidative

stress and/or Fenton chemical reactions (iron catalyzed peroxide formation) at the

microelectrode-tissue interface [2, 11, 12, 30, 98, 99, 189, 190, 462]. Pro-inflammatory cells (activated microglia, macrophages and astrocytes) remain reactive on and around the intracortical microelectrodes for the duration of implantation [130, 144, 188].

Furthermore, it is understood that these pro-inflammatory cells release cytokines [132], free radicals, reactive oxygen species (ROS) and reactive nitrogen species (RNS) when activated [463-465].

Many attempts have been made to alter the design or materials properties of the intracortical microelectrodes to minimize the neuroinflammatory response (for review see

Jorfi et al. [1]). We have utilized many antioxidative strategies to specifically attenuate oxidative damage, resulting in higher densities of neuronal nuclei and more viable neurons at the intracortical microelectrode / tissue interface [1, 11, 188-190]. In parallel, we have also attempted to understand the subcellular mechanisms at play in the initiation of reactive oxygen species generation, in response to the implantation and chronic indwelling of intracortical microelectrodes [30].

In that respect, we have identified the innate immunity receptor CD14 as a molecule of interest in the chronic neuroinflammatory response to implanted intracortical microelectrodes [29, 466]. CD14 is molecule associated with the recognition of pathogen

127

associated molecular patterns (PAMPs) and damage associated molecular patterns

(DAMPs) to promote inflammation, including the release of numerous cytokines,

chemokines, and reactive oxygen species [26, 467]. Hermann et al., first observed acute

but not chronic improvements in intracortical microelectrode recording performance in

knockout mice lacking CD14, and chronic improvements in recording performance in mice

receiving a small-molecule inhibitor to the CD14 pathway [29]. More recently, using a

bone marrow chimera model, Bedell et al. demonstrated that inhibiting CD14 from only

the blood-derived macrophages, and not resident brain derived glial cells improves

recording quality over the 16-week long study [466]. Together, these two studies indicated

that partial inhibition of CD14 pathways resulted in a greater improvement to

microelectrode performance than complete inhibition. Therefore, we are interested in

developing a better understanding of the mechanism, to optimize the natural wound healing

response, yet still inhibit the over-excitation of the pathway that can lead to decreased microelectrode performance.

In the current study, we will focus on the complementary receptors associated with

CD14 activation, Toll-like receptors 2 and 4 (TLR2 and TLR4). The innate immunity

receptor TLR4, which is closely associated with CD14, is involved in the recognition of

PAMPs and DAMPs to promote inflammation [21, 26, 256]. Another innate immunity receptor closely associated with CD14, TLR2, is involved in the recognition of PAMPs and DAMPs to promote inflammation [26, 208, 256, 468]. Both TLR2 and TLR4 have been associated with neurodegenerative disorders [21, 23, 208, 344, 352]. Of note, the small molecule inhibitor to CD14 (IAXO-101, Innaxon) is also listed as a TLR4 inhibitor.

Thus, we hypothesize that TLR2 and TLR4 differentially play a role in the

128

neuroinflammatory response to implanted intracortical microelectrodes. To test this

hypothesis, we implanted silicon planar non-recording neural probes in the shape of

Michigan-style intracortical microelectrode arrays into knockout mice lacking TLR2 (Tlr2-

/-), knockout mice lacking TLR4 (Tlr4-/-), and WT control mice, and evaluated endpoint

histology at 2 and 16 weeks after implantation. Gaining a more detailed understanding of

innate immunity receptors associated with the CD14 pathway should further our

understanding of CD14 mediated neuroinflammation to intracortical microelectrodes.

4.3 Materials and methods

4.3.1 Animal Model

Tlr2-/- mice (B6.129-Tlr2tm1Kir/J, stock no. 004650), Tlr4-/- mice (B6.B10ScN-

Tlr4lps-del/JthJ, stock no. 007227), and WT mice (C57BL/6J, stock no. 000664) were

acquired from the Jackson laboratory and bred in-house. Both male and female mice were used as to not bias the results based on sex. We have previously shown that sex does not impact the quality of electrophysiological recordings with functional microelectrodes of the type used in this study [466]. Mice were handled according to the approved Case

Western Reserve University IACUC protocol and the NIH Guide for Care and Use of

Laboratory Animals.

4.3.2 Genotyping

Strains of mice were verified prior to surgery by extracting DNA from tail snips, running PCR, and running gel electrophoresis. Genotyping protocols were performed as suggested by the mouse vendor (Jackson Laboratories), following similar protocols described in previous studies within the lab [466].

129

4.3.3 Probe Implantation Surgery

Mice were implanted with neural probes (described in detail below) using methods

adapted from Ravikumar et al. [28, 144, 469]. Mice were aged to between five and nine

weeks; and weighed between fourteen and twenty-nine grams at the time of surgery. Each

mouse was induced with 3% isoflurane in an induction chamber. While under anesthesia,

mice were mounted to the ear bars of the stereotaxic frame, and anesthesia was lowered to

1% isoflurane for maintenance. Mice were kept on a heating pad while under anesthesia

to maintain body temperature. Ophthalmic ointment was applied to the eyes of the mouse

to prevent drying, followed by shaving of the scalp with a hand-held beard trimmer. The scalp of the mouse was sterilized with three applications of betadine alternated with 70% isopropanol. Marcaine (0.02 ml, 0.25%) was injected below the scalp, at the surgical site, as a local anesthetic. Either buprenorphine (0.1 mg/kg) or meloxicam (0.07 ml, 1.5 mg/mL) were administered subcutaneously as an analgesic. Choice of analgesic changed during the study due to availability from the vendor at the time. Substitutions were chosen following consultation with staff veterinarian. Additionally, cefazolin (0.2 mL, 2mg/mL) was injected subcutaneously as an antibiotic, to prevent post-operative infection. The proper surgical plane of anesthesia was verified using a toe pinch throughout the surgery.

Mouse skulls were exposed by a midline incision on the scalp and retraction of the skin using tissue spreaders. Craniotomies were carried out using a 3mm biopsy punch

(PSS select) lateral to midline, and between lambda and bregma, to minimize heat associated with drilling [428]. One ethylene oxide-sterilized silicon probe in the shape of a Michigan-style microelectrode array (2 mm long × 123 μm wide (tapered) × 15 μm thick,

1 mm x 1 mm bond tab) was inserted 2 mm into the craniotomy via forceps to avoid

130 vasculature. Prior to sterilization, electrodes were cleaned in 70% ethanol and de-ionized water. Craniotomies were sealed with a biocompatible silicone elastomer (Kwik-sil) and closed with a UV-cured liquid dentin (Fusio/Flow-it ALC, Pentron dental). Protruding bond tabs were encased in the liquid dentin to anchor the implant. Incisions were sutured closed with 5-0 monofilament polypropylene suture (Butler Schein). Antibiotic ointment was applied to the incision to prevent infection.

Mice were administered cefazolin (0.2 mL, 2mg/mL) subcutaneously twice on the first day after surgery. Mice were administered meloxicam (0.07 ml, 1.5 mg/mL) once or buprenorphine (0.1 mg/kg) twice on the first day after surgery and as needed thereafter.

Mice were monitored daily for 5 days after the operation for signs of pain and distress and then weekly thereafter.

4.3.4 Tissue Processing

Mice were transcardially perfused 2 or 16 weeks after probe implantation using protocols adapted from Ravikumar et al. [28]. Following perfusions, the mouse heads were removed and post-fixed in 4% paraformaldehyde dissolved in 1xPBS overnight. Liquid dentin skull caps were carefully removed up from the skulls to remove implanted electrodes and minimize damage to the tissue. Brains were gently extracted from skulls and transferred to a series of sucrose solutions with concentrations of 10%, 20%, and two rounds of 30% in 1xPBS for cryoprotection. Upon equilibration (typically overnight), brains were advanced to the next solution in the series and stored at 4°C. Next, brains were frozen in blocks of Optical Cutting Temperature gel over dry ice and moved to a -80°C freezer. Finally, brains were cryostat sectioned into 16 µm horizontal slices and directly

131

mounted to slides at approximately -20 to -25 °C, and stored at -80°C until removed for

immunohistochemical staining.

4.3.5 Immunohistochemical Staining

Staining protocols were adapted from Ravikumar et al. [28]. Microscope slides were removed from the freezer and equilibrated to room temperature in a humidity chamber. Brain slices were blocked in buffers composed of 4% goat or chicken serum,

0.3% Triton-X 100, and 0.1% sodium azide dissolved in 1x PBS for one hour at room temperature. Primary and secondary stain selections and dilutions are summarized in

Table 8. Brain slices were incubated in primary antibody solutions dissolved in blocking buffers containing serum matching the species of secondary antibody. Brain slices were incubated in secondary antibody solutions for 2 hours at room temperature. Tissue autofluorescence was reduced via the application of a copper sulfate solution, as described by

Potter et al. [443]. Microscope slides of brain slices were coverslipped using fluromount-

G mounting medium.

4.3.6 DAB Staining for Neuronal Nuclei

A majority of the tissue slices stained for neuronal nuclei using NeuN were followed with a DAB chromogen to make the cells visible under brightfield light.

Protocols to stain neuronal nuclei were adapted from Ravikumar et al. [28]. Brain slices were blocked in goat blocking buffer for 1 hour at room temperature. Slices were then incubated in blocking buffer solutions containing NeuN antibodies (Millipore MAB377) diluted 1:250 for one hour at room temperature. Horseradish peroxidase polymer and DAB chromogen were added to the brain slices according to the manufacturer protocols (Life

Technologies Super PicTure Polymer Detection Kit, Ref 879163). Hematoxylin was

132

applied to the brain slices as a counter stain for all cell nuclei. Slides were coverslipped

with Histomount mounting medium.

Additional NeuN sections were added using a superior protocol involving a

fluorescent secondary antibody. These sections were stained following the protocols

established in 4.3.5. Since NeuN was quantified by hand, regardless of the protocol, no

difference in results were achieved with the two protocols, just ease of quantification for

the counter.

Table 8. Summary of immunohistochemical reagents used in histology.

4.3.7 Imaging

Fluorescent images were captured using a Zeiss AxioObserver Z1 inverted

fluorescent microscope with a 10x objective and an AxioCam MRm monochrome camera.

As described in Hermann et al., 4 x 4 mosaic images centered on the probe hole were

assembled using AxioVision and Zen software [29].

Color images were captured for NeuN stains utilizing a DAB chromogen, as described by Ravikumar et al. [28]. The same AxioObserver Z1 inverted microscope was used with a 10x objective and an Erc5 color camera. As described by Ravikumar et al.,

4x4 mosaic images were assembled using Axiovision software [28].

Representative images of sham animals not implanted with neural probes are displayed in Figure 29. This figure demonstrates constitutive expression of NeuN, CD68,

GFAP, and IgG in Tlr2-/-, Tlr4-/-, and WT mice. No differences were seen in control, non-

implanted shams.

133

Figure 29. Immunohistochemical staining in sham animals. WT (A, D, G, J), Tlr2-/- (B, E, H, K), and Tlr4-/- (C, F, I, L) mice without implanted probes were stained for neuronal nuclei (NeuN, A-C), activated microglia and macrophages (CD68, D-F), astrocytes (GFAP, G-I), and blood-brain barrier permeability (IgG, J-L). Scale bars are provided for each row of images, scale = 200 µm.

4.3.8 Neuronal Nuclei Counting

Neuronal nuclei counts were obtained using our freely available custom Matlab

scripts Second and AfterNeuN as described in Hermann et al. [29]. A user traced the

outline of the probe hole and defined artifacts (tissue edge, etc.) using Second and subsequently defined NeuN positive cell positions using AfterNeuN. Users counted out to a background defined as 400 µm from the probe hole plus a 50 µm buffer zone. Neuronal

134

density was quantified as cell count per area, in concentric 50 µm bins extending from the

probe hole. Percent of background density was quantified as bin neuronal density divided

by the neuronal density of the 300-400 µm region times 100. Mean and standard error were defined for each group based on animal averages. Statistics are calculated using animal averages.

4.3.9 Immunohistochemical Marker Quantification

Immunohistochemical markers were quantified using the Matlab script Second, as described in Hermann et al. [29]. A user traced the outline of the probe hole and defined artifacts using Second. Second calculated fluorescence intensity in 5 µm concentric bins out to a distance of 650 µm from the probe. Background intensity was defined as the fluorescence intensity in the 600-650 µm region. Fluorescence intensity was normalized to a value of 0 at background for CD68 and IgG as no CD68 or IgG expression is seen in the non-implanted sham (Figure 29), and normalized to a value of 1 at background for

GFAP, as GFAP is expressed in native tissue (Figure 29). Area under the curve was calculated for 50 µm bins extending away from the probe hole; 5 µm normalized intensity bins and 50 µm area under the curve bins were averaged by animal across 3-6 slices. Mean and standard error for each group were based on animal averages. Statistics were calculated using animal averages of 50 µm area under the curve bins.

4.3.10 Statistics

All statistical comparisons were carried out using Minitab software. For each time point neuronal nuclei percent background density and normalized fluorescence intensity area under the curve animal averages are compared between knockout mice and WT controls. Animal average values for a given stain were entered into a general linear model

135 with animal average values from other strains at the same time point. Comparisons are made between Tlr2-/- and WT mice and between Tlr4-/- and WT mice using a Bonferroni test. Significance was defined as a p value less than 0.05.

Additionally, statistical comparisons were made between mice of the same strain at different time points. Neuronal nuclei percent background density and normalized fluorescence intensity area under the curve animal averages are compared between 2-week and 16-week time points. Animal average values from a given strain and time point are entered into a general linear model with animal average values from the same strain at the opposite time point. Comparisons are made between 2-week and 16-week mice for a given strain using a Tukey test. Significance was defined as a p value less than 0.05.

4.4 Results

Two cohorts of animals were implanted with neural probes to assess both the acute and chronic neuroinflammatory response to implanted intracortical microelectrodes. For the purposes of this study, four histological markers were investigated. First, since neurons are the source of electrical signals recorded by intracortical microelectrodes, sections of cortical tissue were stained with an antibody directed against NeuN, a nuclear protein specific to neurons [445]. Neuronal dieback around implanted intracortical microelectrode arrays is commonly attributed to the release of soluble factors by inflammatory activated microglia and macrophages [132]. Therefore, to understand how the absences of TLR2 and TLR4 affect inflammatory activation of microglia and macrophages in response to implanted neural probes, sections of cortical tissue were stained with an antibody directed against CD68 (macrosialin), a sialoglycoprotein found in activated microglia and macrophages [470]. Another byproduct of chronic inflammatory mechanisms potentially

136

detrimental to intracortical microelectrode activation is astrocytic encapsulation.

Astrocytes may become hypertrophic in response to implanted microelectrode arrays and encase the array in a sheath that impedes electrical signals [134]. Therefore, to understand how the absences of TLR2 and TLR4 affect astrocytic encapsulation in response to implanted neural probes, sections of cortical tissue were stained with an antibody directed against glial fibrillary associated protein (GFAP), an astrocytic intermediate filament that is upregulated during inflammatory activation [446]. Finally, another component of chronic inflammatory mechanisms correlated to poor recording performance is blood-brain barrier permeability [136]. To understand how the absences of TLR2 and TLR4 affect blood-brain barrier permeability in response to implanted neural probes, sections of cortical tissue were stained with an antibody directed against IgG, a blood protein not normally found in healthy brain tissue.

4.4.1 Acute (2-week) neuroinflammatory response to implanted intracortical probes in

Tlr2-/-, Tlr4-/-, and WT mice

Plots of percentage of background neuronal density (Neuronal survival) or

normalized fluorescence intensity (microglial activation, astrocytic encapsulation, blood-

brain barrier permeability) with respect to distance from probe hole for Tlr2-/-, Tlr4-/-, and

WT are shown in Figure 30. All three strains of mice exhibited trends of increasing

neuronal density, decreasing microglial activation, decreasing astrocytic encapsulation,

and decreasing blood-brain barrier permeability with increasing distance from the probe

hole at the acute 2-week time point. Examination of neuronal survival via the NeuN stain

in Tlr2-/-, Tlr4-/-, and WT mice (Figure 30A) revealed no statistical differences between groups at the counted distance intervals (Tlr2-/-: N=7; Tlr4-/-: N=5; WT: N=5). Similarly,

137

examination of the accumulation of inflammatory activated microglia and macrophages

via CD68 revealed no significant differences between either of the knockout conditions

-/- -/- and WT mice (Tlr2 : N=8; Tlr4 : N=4; WT: N=6). Additionally, examination of the

chronic glial scar as a function of GFAP expression also indicated no significant

differences were observed between either of the knockout conditions and WT mice (Figure

30C). Unlike glial cell density, blood-brain barrier permeability (as a function of IgG expression) indicated significant differences between experimental and control group at the acute 2-week time point (Figure 30D). Specifically, Tlr4-/- mice exhibit significantly

less IgG expression compared to WT mice at the distance intervals 0 to 50 and 550-600

µm from the probe hole, indicating reduced blood-brain barrier permeability, $ p<0.05

-/- -/- (Tlr2 : N=5; Tlr4 : N=4; WT: N=6). Tlr2-/- mice did not exhibit any differences in IgG

expression from WT controls.

138

-/- -/- Figure 30. Acute immunohistochemical evaluation of Tlr2 , Tlr4 , and wildtype mice two weeks after probe implantation. (A) Neuronal survival in cortical tissue around implanted probes presented as percent of background density with respect to distance from probe hole (µm). No significant differences were observed between either of the knockout conditions and wildtype mice. Tlr2-/-: N=7; Tlr4-/-: N=5; WT: N=5. (B) Microglia and macrophage activation in cortical tissue around implanted probes presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence intensity indicates expression of CD68. No significant differences were observed between either of the knockout conditions and wildtype mice. Tlr2-/-: N=8; Tlr4-/-: N=4; WT: N=6. (C) Astrocyte encapsulation in cortical 139

tissue around implanted probes presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence intensity indicates expression of GFAP. No significant differences were observed between either of the knockout conditions and wildtype mice. Tlr2-/-: N=8; Tlr4- /-: N=4; WT: N=6. (D) Blood-brain barrier permeability presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence intensity indicates expression of IgG. Tlr4-/- mice exhibited significantly less IgG expression within distances of 0-50 and 550-600 µm from the probe hole, indicating reduced blood-brain barrier permeability, $ p<0.05. Tlr2-/-: N=5; Tlr4-/-: N=4; WT: N=6. Orange = Tlr2-/-, Green = Tlr4-/-, and Blue = WT. Scale bars are provided for each set of images, scale = 200 µm.

4.4.2 Chronic (16-week) neuroinflammatory response to implanted intracortical probes in

Tlr2-/-, Tlr4-/-, and WT mice

An additional cohort of animals was implanted with identical non-functional probes

for 16 weeks, to assess the chronic neuroinflammatory response [369]. Examination of

neuronal density via the NeuN stain in Tlr2-/-, Tlr4-/-, and WT mice revealed a trend of

increasing neuronal survival with distance from the probe at the 16-week time point

(Figure 31A). Neuronal survival in Tlr4-/- mice was significantly lower than neuronal

survival in WT mice in the distance interval 0-50 µm from the probe hole. Neuronal

survival in Tlr2-/- mice was significantly higher than neuronal survival in WT mice in the

distance intervals 200-250 and 250-300 µm from the probe hole (Tlr2-/-: N=5; Tlr4-/-: N=5;

WT: N=7). Examination of the accumulation of inflammatory activated microglia and

macrophages via CD68 expression indicated that Tlr2-/-, Tlr4-/-, and WT mice all exhibit a trend of decreasing CD68 expression with distance from the probe hole at the chronic 16- week time point (Figure 31B). However, no significant differences were observed between either of the knockout conditions and WT mice (Tlr2-/-: N=5; Tlr4-/-: N=5; WT:

N=7). Similarly, examination of the chronic glial scar as a function of GFAP expression

also indicated no significant differences between either of the knockout conditions and WT

-/- -/- mice (Tlr2 : N=8; Tlr4 : N=4; WT: N=6) (Figure 31C). Unlike glial cell density, blood- brain barrier permeability as a function of IgG expression indicated significant differences

140

between experimental and control group at 16 weeks post-implantation (Figure 31D).

Specifically, Tlr4-/- mice exhibit significantly less IgG expression compared to WT mice

at each of the interval examined from 0 to 350 µm from the probe hole, indicating reduced

blood-brain barrier permeability, $ p<0.05. Additionally, Tlr2-/- mice exhibit significantly

less IgG expression within a distance of 450-500 µm from the probe hole, indicating

-/- -/- reduced blood-brain barrier permeability, * p<0.05 (Tlr2 : N=5; Tlr4 : N=4; WT: N=6).

141

-/- -/- Figure 31. Chronic immunohistochemical evaluation of Tlr2 , Tlr4 , and wildtype mice sixteen weeks after probe implantation. (A) Neuronal survival in cortical tissue around implanted probes presented as percent of background density with respect to distance from probe hole (µm). Neuronal survival in Tlr4-/- mice was significantly lower than neuronal survival in wildtype mice in the distance interval 0-50 µm from the probe hole. Neuronal survival in Tlr2-/- mice was significantly higher than neuronal survival in wildtype mice in the distance intervals 200-250 and 250-300 µm from the probe hole. Tlr2-/-: N=5; Tlr4-/-: N=5; WT: N=7. (B) Microglia and macrophage activation in cortical tissue around implanted probes presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence

142

intensity indicates expression of CD68. No significant differences were observed between either of the knockout conditions and wildtype mice. Tlr2-/-: N=5; Tlr4-/-: N=5; WT: N=7. (C) Astrocyte encapsulation in cortical tissue around implanted probes presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence intensity indicates expression of GFAP. No significant differences were observed between either of the knockout conditions and wildtype mice. Tlr2-/-: N=5; Tlr4-/-: N=5; WT: N=7. (D) Blood-brain barrier permeability presented as normalized fluorescence intensity with respect to distance from probe hole (µm). Normalized fluorescence intensity indicates expression of IgG. Tlr4-/- mice exhibited significantly less IgG expression within each distance interval examined from of 0-350 µm from the probe hole, indicating reduced blood-brain barrier permeability, $ p<0.05. Additionally, Tlr2-/- mice exhibited significantly less IgG expression within a distance of 450-500 µm from the probe hole, indicating reduced blood-brain barrier permeability, * p<0.05. Tlr2-/-: N=5; Tlr4-/-: N=5; WT: N=7. Tlr2-/-: N=5; Tlr4-/-: N=4; WT: N=6. Orange = Tlr2-/-, Green = Tlr4-/-, and Blue = WT. Scale bars are provided for each set of images, scale = 200 µm.

4.4.3 The progression of neuroinflammation and neurodegeneration following

intracortical microelectrode implantation

It is also important to understand how the progression of the neuroinflammatory and neurodegenerative response to intracortical microelectrodes is effected by the removal

of either TLR2 or TLR4 from the innate immunity process. Therefore, expression of

immunohistochemical markers for neuronal survival, microglia and macrophage activation, astrocytic encapsulation, and blood brain barrier permeability were compared

between the 2-week and 16-week time points for WT (Figure 32), Tlr2-/- (Figure 33), and

Tlr4-/- mice (Figure 34), independent of each other.

4.4.3.1 The progression of neuroinflammation and neurodegeneration in WT mice

Changes in immunohistochemical markers between the acute 2-week and chronic

16-week time points in WT mice will indicate the standard progression of chronic

neuroinflammatory mechanisms in response to implanted neural probes in mice.

Examination of neuronal density via the NeuN stain in WT mice exhibited significantly

higher neuronal density at 2 weeks after probe implantation in distance intervals between

150 and 300 µm from the probe hole, *p<0.05 (2wk WT: N=5; 16wk WT: N=7) (Figure

32A). Additionally, examination of the accumulation of inflammatory activated microglia

and macrophages via CD68 expression indicated WT mice exhibit significantly higher

143

CD68 expression at 2 weeks after probe implantation in distance intervals between 0-100

µm from the probe hole, *p<0.05 (2wk WT: N=6; 16wk WT: N=7) (Figure 32B). In contrast, examination of the chronic glial scar as a function of GFAP expression revealed

WT mice exhibit significantly higher GFAP expression at 16 weeks after probe implantation in distance intervals between 0 and 200 µm from the probe hole, *p<0.05

(2wk WT: N=6; 16wk WT: N=7) (Figure 32C). Similar to microglial activation, blood- brain barrier permeability as a function of IgG expression revealed WT mice exhibit significantly higher IgG expression at 2 weeks after probe implantation in distance intervals 0-50 and between 400 and 600 µm from the probe hole, *p<0.05 (2wk WT: N=6;

16wk WT: N=7) (Figure 32D).

144

Figure 32. Changes in immunohistochemical markers in wildtype mice over time. -/- Figures A-D show immunohistochemical marker expression in Tlr2 mice at 2 and 16 weeks after probe implantation. (A) Neuronal survival displayed as percent of background neuronal density with respect to distance from the probe hole (µm). Wildtype mice exhibit significantly higher neuronal survival at two weeks after implantation in distance intervals 150-200, 200-250, and 250-300 µm from the probe hole, *p<0.05. 2wk WT: N=5; 16wk WT: N=7. (B) Microglia and macrophage activation (CD68) displayed as normalized fluorescence intensity with respect to distance from the probe hole (µm). Wildtype mice exhibited significantly higher microglia and macrophage activation at 2 weeks after probe implantation at distance intervals 0-50 and 50-100 µm from the probe hole, *p<0.05. 2wk WT: N=6; 16wk WT: N=7. (C) Astrocytic encapsulation (GFAP) displayed as normalized fluorescence intensity with respect to distance from the probe hole (µm). Wildtype mice exhibit significantly higher astrocytic encapsulation at 16 weeks after probe implantation at distance intervals 0-50, 50-100, 100-150, and 150-200 µm from the probe hole, *p<0.05. 2wk WT: N=6; 16wk WT: N=7. (D) Blood-brain barrier permeability (IgG) as normalized fluorescence intensity with respect to distance from the probe hole (µm). Wildtype mice exhibit significantly blood-brain barrier permeability at two weeks after probe implantation at distance intervals 0-50, 400-450, 450-500, 500- 550, and 550-600 µm from the probe hole, *p<0.05. 2wk WT: N=6; 16wk WT: N=7.

145

4.4.3.2 The progression of neuroinflammation and neurodegeneration in Tlr2-/- mice

Examining the time course of immunohistochemical markers in Tlr2-/- mice will

identify potential effects of TLR2 removal on the standard progression of chronic

neuroinflammatory mechanisms in response to implanted neural probes. Contrary to the

trend in WT mice, examination of neuronal density in Tlr2-/- mice exhibited significantly

higher neuronal density at 16 weeks than 2 weeks after probe implantation in the distance

interval 250-300 µm from the probe hole, *p<0.05 (2wk Tlr2-/-: N=7; 16wk Tlr2-/-: N=5)

(Figure 33A). While this is a significant finding, it is likely not appreciable for electrode

performance. Similar the trend observed in WT mice, the accumulation of inflammatory

activated microglia and macrophages in Tlr2-/- mice exhibited significantly higher CD68

expression at 2 weeks than 16 weeks after probe implantation. However, the distance

intervals with higher CD68 expression were between 500 and 600 µm from the probe hole,

*p<0.05 (2wk Tlr2-/-: N=8; 16wk Tlr2-/-: N=5) (Figure 33B), which likely does not impact device performance. Similar to the trend observed in WT mice, examination of the chronic glial scar as a function of GFAP expression exhibited significantly higher GFAP expression at 16 weeks after probe implantation at distance intervals between 0 and 200

µm from the probe hole, *p<0.05 (2wk Tlr2-/-: N=8; 16wk Tlr2-/-: N=5) (Figure 33C).

Astrocytic encapsulation increases over time in Tlr2-/- mice in similar distance ranges as

WT mice. As seen in WT mice, blood-brain barrier permeability as a function of IgG in

Tlr2-/- mice revealed significantly higher IgG expression at 2 weeks after probe

implantation compared to 16 weeks after probe implantation at the distance intervals 0-50

µm from the probe hole, *p<0.05 (Figure 33D) (2wk Tlr2-/-: N=5; 16wk Tlr2-/-: N=5).

146

-/- Figure 33. Changes in immunohistochemical markers in Tlr2 mice over time. Figures A-D show -/- immunohistochemical marker expression in Tlr2 mice at 2 and 16 weeks after probe implantation. (A) Neuronal survival displayed as percent of background neuronal density with respect to distance from the -/- probe hole (µm). Tlr2 mice exhibit significantly higher neuronal survival at 16 weeks after probe -/- -/- implantation in the distance interval 250-300 µm from the probe hole. 2wk Tlr2 : N=7; 16wk Tlr2 : N=5. (B) Microglia and macrophage activation (CD68) displayed as normalized fluorescence intensity with respect -/- to distance from the probe hole (µm). Tlr2 mice exhibit significantly higher microglia and macrophage -/- -/- activation at 2 weeks after probe implantation*p<0.05. 2wk Tlr2 : N=8; 16wk Tlr2 : N=5. (C) Astrocytic encapsulation (GFAP) displayed as normalized fluorescence intensity with respect to distance from the probe -/- hole (µm). Tlr2 mice exhibit significantly higher astrocytic encapsulation at 16 weeks after probe implantation at distance intervals 0-50, 50-100, 100-150, and 150-200 µm from the probe hole, *p<0.05. -/- -/- 2wk Tlr2 : N=8; 16wk Tlr2 : N=5 (D) Blood-brain barrier permeability (IgG) as normalized fluorescence -/- intensity with respect to distance from the probe hole (µm). Tlr2 mice exhibit significantly higher blood- brain barrier permeability at two weeks after probe implantation at the distance intervals 0-50 µm from the -/- -/- probe hole, *p<0.05. 2wk Tlr2 : N=5; 16wk Tlr2 : N=5.

147

4.4.3.3 The progression of neuroinflammation and neurodegeneration in Tlr4-/- mice

Examining the time course of immunohistochemical markers in Tlr4-/- mice will

identify potential effects of TLR4 removal on the standard progression of chronic

neuroinflammatory mechanisms in response to implanted neural probes. Unlike WT mice

and Tlr2-/- mice, examination of neuronal density and accumulation of inflammatory

activated microglia and macrophages in Tlr4-/- mice exhibited no significant differences between time points (2wk Tlr4-/-: N=5; 16wk Tlr4-/-: N=5) (Figure 34A,B). Similar to

trends observed in WT and Tlr2-/- mice, examination of the chronic glial scar in Tlr4-/- mice

exhibited significantly higher GFAP expression at 16 weeks than 2 weeks after probe

implantation. Unlike WT and Tlr2-/- mice, significantly higher GFAP expression only occurred in the distance interval 0-50 µm from the probe hole, *p<0.05 (2wk Tlr4-/-: N=4;

16wk Tlr4-/-: N=5) (Figure 34C). Similar to WT and Tlr2-/- mice, blood-brain barrier

permeability as a function of IgG expression in Tlr4-/- mice exhibited significantly higher

IgG expression at 2 weeks than at 16 weeks after probe implantation. Unlike WT and

Tlr2-/- mice, significant differences occurred over the distance intervals between 0 and 150

µm from the probe hole, *p<0.05 (2wk Tlr4-/-: N=4; 16wk Tlr4-/-: N=5) (Figure 34D).

148

-/- Figure 34. Changes in immunohistochemical markers in Tlr4 mice over time. Figures A-D show -/- immunohistochemical marker expression in Tlr4 mice at 2 and 16 weeks after probe implantation. (A) Neuronal survival displayed as percent of background neuronal density with respect to distance from the -/- probe hole (µm). Tlr4 mice exhibit no significant difference in neuronal survival between two and 16 weeks -/- -/- after probe implantation. 2wk Tlr4 : N=5; 16wk Tlr4 : N=5. (B) Microglia and macrophage activation (CD68) displayed as normalized fluorescence intensity with respect to distance from the probe hole (µm). -/- Tlr4 mice exhibit no significant differences in microglia and macrophage activation between time points. -/- -/- 2wk Tlr4 : N=4; 16wk Tlr4 : N=5. (C) Astrocytic encapsulation (GFAP) displayed as normalized -/- fluorescence intensity with respect to distance from the probe hole (µm). Tlr4 mice exhibit significantly higher astrocytic encapsulation at 16 weeks after probe implantation the distance intervals 0-50 µm from the -/- -/- probe hole, *p<0.05. 2wk Tlr4 : N=4; 16wk Tlr4 : N=5. (D) Blood-brain barrier permeability (IgG) as -/- normalized fluorescence intensity with respect to distance from the probe hole (µm). Tlr4 mice exhibit significantly higher blood-brain barrier permeability at two weeks after probe implantation at the distance -/- -/- intervals 0-50, 50-100, and100-150 µm from the probe hole, *p<0.05. 2wk Tlr4 : N=4; 16wk Tlr4 : N=5.

149

4.5 Discussion

This study sought to interpret the roles of TLR2 and TLR4 in the

neuroinflammatory response to implanted intracortical microelectrodes. Overall, this

study reveals that full removal of TLR4 results in reduced blood-brain barrier permeability

at an acute (2 weeks) and chronic time point (16 weeks), and reduced neuronal survival at

chronic time points, compared to WT animals. Further, microglial activation and blood-

brain barrier permeability significantly decreased from acute to chronic time points, whereas astrocytic encapsulation significantly increased from acute to chronic time points in WT mice. Mice lacking TLR2 or TLR4 exhibited similar trends in decreasing blood- brain barrier permeability and increasing astrocytic encapsulation, but no significant changes in microglial activation from 2 to 16 weeks post-implantation. The findings presented here introduce more questions regarding the role of innate immunity receptors in the neuroinflammatory response to intracortical microelectrodes.

The first major finding of this study indicated that knocking out TLR4 resulted in decreased blood-brain barrier permeability around implanted intracortical microelectrode at both acute and chronic time points. The blood-brain barrier is a network of endothelial cells with tight junctions that protects parenchymal brain tissue from neurotoxic molecules and infiltrating inflammatory cells. Damage to the blood-brain barrier following intracortical microelectrode implantation has been linked to poor recording performance

[136], potentially through neuronal damage, altered extracellular ionic concentrations, or propagation of inflammatory mechanisms [1]. Blood-brain permeability in response to implanted neural probes may be related to TLR4 signaling on several levels: TLR4 signaling in endothelial cells, release of cytokines in response to TLR4 activation, and

150

oxidative damage caused by factors released in response to TLR4 activation. An

upregulation of TLR4 has been observed in vascular endothelial cells in response to renal

ischemia reperfusion injuries, and an increase in co-localization of TLR4 and vascular endothelial cells was observed in response to subarachnoid hemorrhages [414, 471].

Ischemic injury has been tied to neurodegeneration around implanted intracortical microelectrodes [105]. It is possible that signaling of TLR4 on vascular endothelial cells in response to the implanted intracortical microelectrodes contributes to permeability of the blood-brain barrier.

Activation of TLR4 on secondary cells may also be responsible for increased permeability of the blood-brain barrier. In addition to vascular endothelial cells, neurons and microglia also exhibited elevated co-localization with TLR4 in response to subarachnoid hemorrhages [414]. Leow-Dyke et al. demonstrated that factors released by neurons conditioned with the TLR4 ligand lipopolysaccharide (LPS), including RANTES

(CCL5), KC (CXCL1), tumor necrosis factor-α (TNFα), and IL-6, promoted the migration of neutrophils across an endothelial monolayer in vitro, and prior application of a TLR4 antagonist to the neurons significantly reduced this effect [327]. Additionally, activation of TLR4 on microglia may lead to activation of the inflammatory NF-κB pathway, which can induce the release of pro-inflammatory cytokines, such as TNFα, IL-6, IL1-β [20, 472].

TNF-α and IL1-β have been shown to increase permeability of the blood-brain barrier

[473]. Bennett et al. recently detected upregulation of genes encoding pro-inflammatory cytokines paired with downregulation of genes encoding junction proteins of the blood- brain barrier at acute time points following intracortical microelectrode implantation [143].

Coincidentally, Bennett et al identified enhanced expression of TNFα, IL-6, and KC

151

(CXCL1) following intracortical microelectrode implantation, potentially indicating a role

of neuronal cytokine and chemokine release [143, 327]. Activation of TLR4 on microglia may also lead to the release of reactive oxygen species [26]. Reactive oxygen species can promote leakiness of the blood-brain barrier [361, 474]. In the absence of TLR4, less pro- inflammatory cytokines and reactive oxygen species are likely released, and thus less damage to the blood brain barrier occurs. Conversely, Tlr4-/- mice did not exhibit any

significant differences in microglia/macrophage activation. Differences in CD68

expression may not be sensitive enough to confer differences in cytokine and ROS release

by activated microglia and macrophages, other infiltrating myeloid cells or neurons may

be driving the release of factors damaging the blood-brain barrier, or TLR4 signaling on

endothelial cells may facilitate blood-brain barrier permeability.

The next major finding of this study, indicating a reduction in neuronal survival

around implanted intracortical microelectrodes in mice lacking TLR4, is more difficult to

explain. Neurons are the source of signals recorded by intracortical microelectrodes and

hypothesized to be needed within 50 µm of the microelectrode to record single units [117].

Although neuronal dieback has frequently been observed around implanted intracortical

microelectrodes [129, 369, 461] and neuronal dieback has been hypothesized to cause

intracortical microelectrode failure [129], the relationship between neuronal dieback and

recording performance has not been fully elucidated [1]. Typically, knocking out TLR4

results in neuroprotective effects [343, 475]. Here the opposite trend is observed. Perhaps the reduced capability to detect and fight pathogens makes mice more susceptible to localized infections; however, no indications of localized infection were seen in this study.

Robust sterilization methods such as ethylene oxide sterilization do not always reduce

152

endotoxin levels below the FDA requirement for devices implanted in the brain [28].

Hermann et al. proposed that the reduced capability to detect and respond to tissue damage may hinder wound healing mechanisms beneficial to integrating devices into the brain [29].

Studies investigating the role of TLRs in neurodegenerative disorders such as Alzheimer’s and synucleinopathies suggested that some amount of TLR signaling was necessary to clean up the accumulation of abnormal protein deposits [281, 334, 349] and neurons damaged by the proteins [350]. Damaged matrix proteins and necrotic cells resulting from the implantation and chronic presence of an intracortical microelectrode that are normally recognized and disposed of by TLR4 mediated pathways may be detrimental to neurons directly or through the activation of redundant inflammatory mechanisms.

The third major finding of this study identified differences in the time course of the foreign body response to implanted intracortical microelectrodes in the absence of TLR2 and TLR4. The Tlr2-/- and Tlr4-/- mice exhibited most of the same trends as the WT mice

between the 2- and 16-week time points, except for microglial activation. The WT mice

exhibit a significant reduction in microglia and macrophage activation whereas the Tlr2-/-

and Tlr4-/- mice do not exhibit significant changes over time. The trend in WT mice

indicate that TLR2 and/or TLR4 may play an important role in the activation of microglia

and macrophages at the acute time point, despite the lack of significant differences between

either the Tlr2-/- or Tlr4-/- mice and WT mice. The Tlr2-/- and Tlr4-/- mice both demonstrated lower peak CD68 intensities, but the intensities decayed to similar values over a short span of distance (~10 µm). Differences in CD68 expression may be limited to the first layer of cells around the neural probe hole. The decrease in CD68 expression in WT mice over time may indicate a diminishing importance of TLR2 and TLR4 in the chronic

153

neuroinflammatory response to intracortical microelectrodes over time, or that lacking

either TLR2 or TLR4 retards the rate of wound healing / scar progression. For example,

TLR4 deficient mice exhibited delayed skin wound closure paired with decreased Il-1β and

IL-6 production [476], indicating the importance of TLR4 activation and subsequent cytokine release in wound healing. Similarly, Tlr2-/-, Tlr4-/-, and Tlr2/4-/- mice exhibited

larger skin wound areas paired with a reduction in infiltrating macrophages and decreased

expression of TGF-β and CCL5 (RANTES) [477]. On the contrary, Tlr2-/- and Tlr4-/- mice

exhibited improved wound healing in response to diabetic skin injuries [478]. Opposing outcomes of knocking out TLRs have been attributed to differences in acute and chronic injures, where chronic inflammation, as found in diabetic injuries, may hinder wound healing [479, 480]. The unresolved presence of an implanted intracortical microelectrode would likely behave like other chronic injuries. The role of TLR2 and TLR4 on wound

healing in the brain would be difficult to predict, since activation of TLRs promotes injury

or wound healing in a variety of injuries throughout the body [481], and the effects are

hypothesized to be dose dependent [482], timing dependent [483], and location dependent

[481, 483]. In the context of spinal cord injury, Tlr4-/- mice exhibited deficits in locomotor

recovery and elevated demyelination, astrogliosis, and macrophage activation, and Tlr2-/-

mice exhibited deficits in locomotor recovery paired with abnormal myelin patterning

[484]. These findings suggest that TLR2 and TLR4 may play a beneficial role in the

recovery of CNS injuries. However, the exact role of TLR2 and TLR4 in wound healing

of the CNS remains to be elucidated.

Comparing the findings of this study to previous studies in our lab inhibiting the

TLR co-receptor will provide a greater understanding of the role of innate immunity

154

receptors in the chronic inflammatory response to intracortical microelectrodes. Hermann

et al. previously observed that knockout mice lacking CD14 exhibited enhanced recording

performance over the acute (0 – 2 weeks) but not chronic (2 – 16 weeks) time range, with

no differences in neuronal survival, microglial activation, astrocytic encapsulation, or

blood-brain barrier permeability at 16 weeks after implantation [29]. Here we observe that knockout mice completely lacking TLR4 exhibit significantly reduced blood-brain barrier permeability at both acute and chronic time points. Although TLR4 and CD14 are closely associated in the recognition of ligands, knocking out TLR4 but not CD14 resulted in reduced blood-brain barrier permeability. Activation of TLR4 to promote blood-brain barrier permeability does not require CD14. TLR4 is able to bind and respond to its ligand

LPS in the absence of CD14, although with drastically less sensitivity [467]. There is evidence that TLR4 may bind and respond to DAMPs without CD14 [485], but the exact role of CD14 in the recognition of structurally diverse DAMPs by TLR4 remains to be elucidated. Further, knocking out TLR4 but not CD14 resulted in significantly lower neuronal survival [29]. It appears that fully removing TLR4 is beneficial at acute time points, as with CD14, but fully removing TLR4 at chronic time points is also detrimental.

The presence of functioning TLR4 signaling may be more critical for long-term wound healing than CD14, since Tlr4-/- mice exhibited significantly decreased neuronal survival

and Cd14-/- mice exhibited no differences from wildtype mice at the chronic time point.

Alternatively, chronic decreases in neuronal survival in Tlr4-/- mice may be a carry-over effect from improper wound healing at earlier time points. Experiments stopping and starting or delaying the administration of TLR4 antagonists to mice with implanted intracortical microelectrodes could potentially delineate the time-dependent role of TLR4

155

in the neuroinflammatory response. TLR4 and CD14 play related but independent roles

that vary over time in the foreign body response to intracortical microelectrodes.

In addition to investigating the complete removal of CD14 via a knockout mouse,

Hermann et al. observed that administering a small molecule inhibitor to the CD14-TLR4

complex improved recording performance at acute time points and out to chronic time

points (16 weeks), without any differences in 16-week endpoint histology. Inhibition via

a small molecule antagonist is less complete as full removal of a receptor via knockout.

Residual TLR4 signaling from incomplete inhibition may protect neurons from the

processes detrimental to neuronal survival in Tlr4-/- mice at chronic time points after

intracortical electrode implantation. The role of innate immunity signaling in the foreign body response to intracortical microelectrodes is dependent on the degree of receptor inhibition.

Further studies by Bedell et al. investigated the effects of knocking out CD14 in specific cell populations on intracortical microelectrode recording performance [466].

Mice with infiltrating blood-derived cells lacking CD14 and resident cells featuring intact

CD14 exhibited significantly improved recording performance over wildtype mice over the 16-week study without any differences in 16-week endpoint histology. TLR4 signaling may have different roles on resident cells than on infiltrating myeloid cells. Since TLR4 is constitutively expressed in parenchymal microglia as opposed to CD14, knocking out

TLR4 on resident or infiltrating cells only may produce vastly different results from CD14.

The roles of innate immunity signaling are cell-specific.

Initially, we hypothesized that TLR2 and TLR4 play a role in the neuroinflammatory response to implanted intracortical microelectrodes. Although Tlr2-/-

156

mice did not exhibit any significant differences in endpoint histology, changes in blood-

brain barrier permeability and neuronal survival in Tlr4-/- mice would suggest a role of

TLR4 in the neuroinflammatory response to intracortical microelectrodes. Differences in

the time course of microglial activation observed in Tlr2-/- mice suggest a subtler role in the neuroinflammatory response to intracortical microelectrodes. Further, we proposed

TLR2 and TLR4 signaling as a mechanism for the activation of microglia and macrophages and subsequent release of pro-inflammatory cytokines, ROS, and RNS in response to an implanted intracortical microelectrode. Here, we did not observe significant changes in microglial activation via expression of CD68 in ether Tlr2-/- or Tlr4-/- mice at acute or

chronic time points. However, the paradoxical decrease in blood-brain barrier permeability

and increase in neuronal survival observed in Tlr4-/- mice may be affected by decreased or

increased release of pro-inflammatory factors. As stated earlier, differences in expression

of CD68 may not be sensitive enough to detect functional differences in cytokine, ROS,

and NOS release. Alternatively, activation of TLR4 on other infiltrating myeloid cells not

expressing CD68, endothelial cells, or neurons may be responsible for the observed

histological changes [144]. Future studies investigating the effects of innate immunity

inhibition on gene and mRNA expression following the implantation of intracortical

microelectrodes may elucidate changes in the production of cytokines, ROS, and NOS [30,

486]. Since this study employed intracortical microelectrode arrays without functional

recording sites, conductive traces, and insulating layers, the effects of knocking out TLR2

and TLR4 on oxidative damage to these structures could not be determined via SEM. Much

remains to be learned about the specific roles of TLR2 and TLR4 in the neuroinflammatory

response to intracortical microelectrodes.

157

Building off of the previous studies of Hermann et al. and Bedell et al., innate

immunity signaling pathways appear to play a role in the neuroinflammatory response

against intracortical microelectrodes [29, 466]. The findings of this study suggest that fully removing TLR4 is beneficial at acute time points and fully removing TLR4 at chronic time points is detrimental. Thus, incomplete inhibition of TLR4 and its co-receptors via small

molecule antagonists or antibodies may be of further interest for improving the long-term performance of intracortical microelectrodes. Delaying or stopping and starting the

administration of TLR4 inhibitors may be more appropriate to address the time variant role

in the foreign body response to implanted intracortical microelectrodes. Targeting TLR4 on specific cell populations may be more beneficial than inhibiting TLR4 in the whole body. The Toll-like receptors and their adapter molecules affect the foreign body response in a time, method/extent of inhibition, and cellular subset–dependent manner. Future

strategies to integrate intracortical microelectrodes using modulation of innate immunity

signaling pathways should consider these parameters to optimize the preservation of

electrode and tissue integrity.

4.6 Conclusions

Complete removal of TLR4 via genetic knockout results in reduced blood brain

barrier permeability in response to implanted neural probes at acute and chronic time

points, as well as reduced neuronal survival around implanted neural probes. Benefits of

fully removing TLR4 are overshadowed by detrimental effects on neuronal survival.

Inhibition of TLR4 without complete removal of the pathway or for intermittent time

courses may still be promising as an intervention for improving intracortical

microelectrode integration.

158

4.7 Acknowledgements

We would like to thank Jessica Nguyen, Kelly Buchanan, Monika Goss, Seth

Meade, Emily Molinich, Jacob Rayyan, Andres Robert, Zishen Zhuang, Cara Smith, and

Keying Chen for their help in earlier iterations of the project.

4.8 Funding