NANOPARTICULATE PLATFORMS FOR MOLECULAR IMAGING OF ATHEROSCLEROSIS AND BREAST CANCER

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

By

Bryan Ronain Smith, M.S.

*****

The Ohio State University 2006

Dissertation Committee: Approved by Professor Stephen C. Lee

Professor Mauro Ferrari ______Professor Klaus Honscheid Advisors Biomedical Engineering Graduate Program

ABSTRACT

Modular nanoparticulate platforms facilitate structures amenable to multiple

usages and disease states. The present focus involves molecular imaging with attention

to potential developments in delivery of therapeutics using the devices described. The

diseases chosen for proof-of-principle of targeted, imageable nanoparticle platforms

include atherosclerosis and breast cancer; molecular imaging is performed via magnetic

resonance imaging and non-destructive evaluation ultrasound. The ultimate purpose for this work was to demonstrate proof-of-principle of 1. nanoparticle-based targeted contrast for MRI as a screening technique for heart disease for the general population and

2. early ex vivo detection of breast cancer and prediction of therapeutic response using

targeted nanoparticles and ultrasound.

Herein, the feasibility of delivery of ultrasmall superparamagnetic particulate iron

oxide (USPIO)-mediated negative contrast to atherosclerotic plaques was probed by

biochemical affinity in vivo. An array of targeting reagents was surveyed to yield the

most appropriate reagent for binding to vulnerable plaques. This work focused upon

Annexin V conjugated USPIOs via interaction with phosphatidylserine. Biofunctional

Annexin V USPIOs partitioned rapidly and deeply into plaques, primarily into plaque- localized apoptotic macrophages, which indicated lesion vulnerability. The doses of targeted USPIOs needed to generate specific contrast using MRI were minute in

ii comparison to non-targeting studies (2000-fold lower than reported to drive non-specific

uptake of untargeted USPIOs in plaques of heritable hyperlipidemic rabbits). Distinct

plaque morphologies were differentiable using targeted USPIOs (highly occlusive, stable

plaque versus vulnerable, flatter plaque in two different animal models).

Using characterization mode ultrasound (CMUS) and C-Scan systems, targeted

USPIOs along with continuum and nano-mechanics were used to probe cells and tissue

that overexpressed the Her-2/neu growth receptor. Studies showed that USPIOs could be successfully bound to their targets (via microscopy and flow cytometry) and that CMUS successfully differentiated between cells targeted with USPIOs and controls.

Furthermore, tissue biopsy sections were sensitively and specifically identified using the other ultrasound mode, C-Scan, and targeted USPIOs. Thus, malignancy information was reconstructed by quantification of the molecular expression of Her-2/neu using anti-

Her-2/neu conjugated USPIOs. The mechanical properties of tissues were thereby correlated with their malignancy and Her-2/neu status.

The work provides proof-of-principle for the platform nature of USPIOs. The modularity of the USPIO platform enabled precise and quantitative molecular information to be collected for two disparate disease states using two different imaging modalities.

iii

Dedicated to my wonderful and loving parents

iv

ACKNOWLEDGEMENTS

While the completion of any dissertation requires the cooperation,

support, and assistance of many individuals, the reality of a multi-disciplinary

dissertation in the burgeoning discipline of biomedical nanotechnology demands

assistance and integration between individuals from many fields of science and medicine.

In my case, I am indebted to excellent scientists in radiology, cell biology, electrical engineering, oncology, nuclear medicine, statistics, pathology, chemistry, physics, and of course my home department of biomedical engineering.

I must first thank my advisors, Professors Stephen Lee and Mauro Ferrari. From them I learned a great deal not only about how to do research, but how to communicate it professionally and how to think of it in the “big picture” context. Dr. Lee guided me and helped me to realize what is necessary and expected of a Ph.D. student when I needed it most – toward the beginning of my graduate education – for which I am exceedingly grateful. Our often differing perspectives, his ingrained from biology and mine from the physical sciences, caused many a long discussion in his office, yet ultimately allowed me to develop as an interdisciplinary scientist. Further, his vast knowledge of, and willingness to talk about, biology was instrumental in preparing me for a career in the biomedical facets of biomedical nanotechnology. I am deeply appreciative of the time

v and effort he spent on feedback for my dissertation drafts and on my career options and direction.

I am indebted to Dr. Ferrari, as he brought me to Ohio State. He was eternally a great source of ideas and someone whom I could count on for feedback about my own technological ideas. I learned much from his scientific vision and leadership. He was always inspiring about research goals and the drive toward achieving clinical reality by the realization of our ideas.

I am also appreciative of the direction and discussions I was privileged to have with Professor Michael Knopp of Radiology. Without his expertise and support, much of the imaging portion of my work could not have been accomplished.

I am certainly indebted to the research group members I have been fortunate to work with over the years: John Shapiro, Drew Kebbel, Michael Sprintz, Xuewu Liu,

Piyush Sinha, Mark Cheng, Jasper Nijdam, Amy Pope-Harman, Jason Sakamoto, and

Sadhana Sharma. I am a more complete scientist for having had the opportunity to interact with, learn from, and discuss scientific principles with them on a daily basis, and could not have done it without them.

I am truly grateful to Dr. Johannes Heverhagen for his assistance in MR imaging and interpretation. His suggestions, knowledge, and work were critical in propeling the atherosclerosis portion of my work.

I appreciate biological assistance and insights from Drs. Greg Lesinski and Kim

Varker, who were crucial in helping to train me to work independently on cell biology- based experiments. John Shapiro was likewise valuable, replete with a sarcastic

vi comment or five, for discussing appropriate experiments and controls especially for

molecular biology experiments.

Many thanks to Drs. Xuewu Liu, Mark Cheng, and Jasper Nijdam, as they were

always available for a helpful hint in the clean room, or to discuss new ideas in silicon

microfabrication. Bin Xie and Jason Sakamoto were vital for advancing the ultrasound

imaging portion of the breast cancer work.

I would also like to thank Ed Herderick and Vlad Maruhlenko, who were always

on hand to help with computer software or hardware issues. Furthermore, Ed provided helpful comments and statistical analyses that helped verify and evaluate findings.

I must also thank Melanie Senitko, Cheryl Kulmala, Jill Connelly, Kirsten

Gibbons, Anita Bratcher, and David Morelli for always assisting in whatever way they could, whether keeping me on the administrative path to graduation, scheduling a meeting, or simply helping out. They were always there when I needed them.

Of course I would like to thank my friends and my family, without whom I could not have done this work. Rosi Matlhabaphiri, Michael Sprintz, Zalman and Sarah

Deitsch, Steve Sullivan, Pamela Habib, and many others, I am thankful for your belief in me, your conversation, and your friendship.

I am eternally grateful for the love and support of my family, who have always been there for me no matter the situation. My sisters have seen me complete my graduate studies firsthand, being in Columbus with me and always available to talk, play tennis, or advise. My grandparents instilled the values of pursuing knowledge and education and my grandmothers were always eager for news about how my work was going. As for my

vii parents, I do not have the words to express my appreciation for their love, sacrifices, and willingness to put up with me during difficult times. I hope that I have made them proud.

viii

VITA

April 13, 1978……………………………Born – New York, New York 2000……………………………………....B.S. Physics, Mathematics, and Biomedical Engineering, Tufts University 2002………………………………………M.S. Biomedical Engineering, The Ohio State University

PUBLICATIONS

1. B.R. Smith, M. Ruegsegger, P. Barnes, M. Ferrari, S.C. Lee. Nanodevices in Biomedical Applications in BioMEMS and Biomedical Nanotechnology, 1st edition, Volume 1. M. Ferrari, ed. Springer.

2. M. Sprintz, B.R. Smith, J. Sakamoto, M. Ferrari. Oncological Nanotechnology in BioMEMS and Biomedical Nanotechnology, 1st edition. M. Ferrari, ed. Springer.

3. J. Sakamoto, B.R. Smith, B. Xie, S. Rokhlin, S. Lee, M. Ferrari (2005). The Molecular Analysis of Breast Cancer Utilizing Targeted Nanoparticulates (USPIOs): Proof-of-Principle with Ultrasound. Technology in Cancer and Research Treatment. 4:6, 627-636.

4. B.R. Smith, J. Heverhagen, M. Knopp, S. Lee (2005). Biologically-Targeted Superparamagnetic Nanoparticulates to Atheromatous Plaque. Molecular Imaging. 4:3, 312.

5. M. Sprintz, J. Sakamoto, B.R. Smith, M. Ferrari (2004). Opportunities in Oncological Nanotechnology. Proceedings of the 1st International Symposium on Micro and Nano Technology. ISMNT, Honolulu, Hawaii. March, 2004.

6. S.C. Lee, M.A. Ruegsegger, P.A. Barnes, B.R. Smith, M. Ferrari (2004). Therapeutic Nanodevices in Springer Handbook of Nanotechnology, Bharat Bhushan, ed. First Edition. Springer Verlag, 279-322.

7. B.R. Smith, A.J. Nijdam, M.C. Cheng, X. Liu, S.C. Lee, M. Ferrari (2004). A Biological Perspective of Particulate Nanoporous Silicon. Materials Technology. 19:1, 16-20.

ix 8. V.T. Granik, B.R. Smith, S.C. Lee, M. Ferrari (2002). Osmotic Pressures for Binary of Non-electrolytes. Biomedical Microdevices. 4:4, 309-321.

9. BR Smith (2001). Chapters on Biomedical Equipment Technology, Dynamic Spatial Reconstructor, Electrosurgery Machines, Endoscopy, Impedance Plethysmography, Medical Electrodes in The Gale Encyclopedia of Nursing & Allied Health, ed. Kristine Krapp, Thomson Gale.

FIELDS OF STUDY

Major Field: Biomedical Engineering Minor Field: Physics

x

TABLE OF CONTENTS Page ABSTRACT...... ii ACKNOWLEDGEMENTS...... v VITA...... ix LIST OF TABLES...... xiii LIST OF FIGURES ...... xv

Chapters:

1. Introduction and Literature Review...... 1

1.1 Nanoparticulate Platforms ...... 1 1.1.1 Bioconjugation...... 10 1.1.2 Silicon Particles ...... 12 1.2 Atherosclerosis...... 14 1.2.1 Clinical Background ...... 20 1.2.2 Modes of Detection and Intervention ...... 22 1.2.3 MRI and SPECT ...... 24 1.2.4 The Debate: Stable Versus Unstable Plaque...... 29 1.2.5 Animal Models of Atherosclerosis ...... 36 1.3 Breast Cancer...... 41 1.3.1 Clinical Background ...... 42 1.3.2 Modes of Detection, Evaluation, and Intervention ...... 43 1.3.3 Her-2/Neu Biomarker and Molecular Analysis ...... 45 1.3.4 Ultrasound Modalities...... 49

2. Atherosclerosis...... 57

2.1 Introduction...... 57 2.2 Experimental...... 59 2.2.1 Imaging ...... 64 2.2.2 Histology...... 67 2.3 Results...... 68 2.3.1 Annexin V USPIOs...... 68 2.3.2 Other Targets ...... 86 2.4 Discussion...... 88 2.4.1 Annexin V...... 89

xi 2.4.2 Other Targets ...... 102 2.5 Experimental Limitations...... 103

3. Breast Cancer...... 107

3.1 Introduction...... 107 3.2 Experimental...... 109 3.2.1 Tissue Phantoms ...... 110 3.2.2 Cell Line Studies...... 111 3.2.3 Tissue ...... 115 3.2.4 Ultrasound Systems and Theory ...... 117 3.3 Results...... 123 3.3.1 Tissue Phantom Studies ...... 124 3.3.2 Cell Line Studies...... 129 3.3.3 Human Breast Tissue Studies ...... 154 3.4 Discussion...... 158 3.4.1 Tissue Phantoms and Cell Line Studies...... 160 3.4.2 Human Breast Tissue Studies ...... 177 3.5 Experimental Limitations and Future Work ...... 180

4. Nanoparticulate Platforms ...... 186

4.1 Experimental...... 189 4.1.1 Preparation and Bioconjugation...... 190 4.1.2 Characterization ...... 192 4.2 Silicon Platforms...... 196 4.2.1 Fabrication ...... 196 4.2.2 Characterization ...... 198 4.3 Experimental Limitations...... 200 APPENDIX A...... 207 Statistics: Analysis of Variance and Tukey Tests For Agarose Tissue Phantoms...... 207 APPENDIX B ...... 217 SKBR-3 Cell Specimen IV ...... 217

BIBLIOGRAPHY...... 222

xii

LIST OF TABLES

Table 1.1: Definitions for terminology generally used in atherosclerotic diseases. Adapted from Fuster et. al. [(66)]...... 30

Table 2.1: Displays the set of targeting reagents chosen for conjugation to USPIOs. .... 60

Table 2.2: Annexin V on USPIOs is bioactive...... 70

Table 2.3: Region-of-interest (ROI) analysis...... 78

Table 3.1: Experimental particle conditions for PDMS tissue phantom study...... 126

Table 3.2: This table exhibits the relationships between the means of the four particle conditions on a single specimen for each mechanical parameter evaluated by C-Scan ...... 127

Table 3.3: Displays reconstructions of doublet mechanical parameters and their continuum mechanical counterparts for agarose tissue phantoms...... 128

Table 3.4: Tissue Reconstruction continuum mechanical properties for cell specimen I ...... 148

Table 3.5: Percent difference relative to the agarose alone sample for continuum mechanics, cell specimen I ...... 148

Table 3.6: Tissue reconstruction continuum mechanical properties for cell specimen II ...... 149

Table 3.7: Percent difference relative to the agarose alone sample for continuum mechanics, cell specimen II ...... 149

Table 3.8: Tissue reconstruction continuum mechanical properties for cell specimen III ...... 150

Table 3.9: Percent difference relative to the cell sample for continuum mechanics, cell specimen III ...... 150

Table 3.10: Her-NP exhibits greater density and stiffness than other conditions...... 151

xiii Table 3.11: Her-NP causes greater density and stiffness change by percent than other conditions...... 152

Table 3.12: A summary of reconstructed continuum mechanics attenuation factor #1 (AL) values ...... 152

Table 3.13: A summary of reconstructed continuum mechanics attenuation factor #2 (AT) values ...... 153

Table 4.1: Displays the assayed iron and protein contents ...... 195

Table 4.2: Gives the hydrodynamic diameter of many of the nanoparticles used in this dissertation...... 195

Table B.1: Her-NP exhibits greater density and stiffness parameters than other conditions...... 221

Table B.2: Her-NP displays greater percent difference in density and stiffness than other conditions...... 221

xiv

LIST OF FIGURES

Figure 1.1: The image represents a hypothetical, modular nanotherapeutic ...... 5

Figure 1.2: Cartoon of the pathogenesis of atherosclerosis from the perspective of the field of molecular imaging...... 16

Figure 1.3: Illustrates the 5 phases and types of lesions in atherosclerotic development.18

Figure 1.4: Designates the histology of atherosclerotic lesions as defined by the AHA in Stary et. al. [(67)]...... 19

Figure 1.5: Nanomechanical reflection spectra ...... 52

Figure 2.1: Jurkat cells are apoptotic after camptothecin treatment ...... 69

Figure 2.2: WHHL rabbits exhibit plaque using MRI and gadolinium contrast ...... 72

Figure 2.3: NZW rabbit displays no MR signal cancellation due to Annexin V USPIO ...... 73

Figure 2.4: NZW rabbit displays no MR signal cancellation due to Annexin V USPIO injection...... 73

Figure 2.5: Young WHHL displays no MR signal cancellation when injected with Annexin V USPIOs...... 74

Figure 2.6: WHHL rabbit exhibits MR signal cancellation within 5 minutes of Annexin V USPIO administration...... 74

Figure 2.7: Same WHHL rabbit exhibits similar MR signal cancellation after 70 days . 75

Figure 2.8: A second WHHL displays MR signal cancellation after Annexin V USPIO administration ...... 75

Figure 2.9: WHHLMI rabbit displays MR signal cancellation within 5 minutes of Annexin V USPIO administration ...... 76

Figure 2.10: WHHLMI rabbit exhibits MR signal cancellation within 5 minutes of injection of 6 times the previous doses of Annexin V USPIOs...... 76

xv

Figure 2.11: WHHLMI exhibits MR signal cancellation within 5 minutes of injection of 6 times the previous doses of Annexin V USPIOs at multiple aortic sites...... 77

Figure 2.12: SPECT and CT fusion image of WHHL rabbit reveals the biodistribution of 99mTc labeled Annexin V USPIOs in bladder, kidney, spleen, liver, and stomach 78

Figure 2.13: Sections obtained from the region of WHHL aorta which caused MRI signal reduction exhibit Prussian blue staining ...... 81

Figure 2.14: Representative sections obtained from the region of WHHLMI aorta ...... 82

Figure 2.15: Representative control sections taken from remote areas of the aorta...... 83

Figure 2.16: Correlation between representative Prussian blue positive section and TUNEL (apoptosis) positive section...... 84

Figure 2.17: By Hoechst assay, cells are apoptotic in regions of aortic iron localization84

Figure 2.18: EM reveals Annexin V USPIOs in cells within WHHLMI plaque in region of MR signal reduction ...... 85

Figure 2.19: Rabbit spleen stains with Prussian blue ...... 86

Figure 2.20: WHHLMI rabbit exhibits MR signal cancellation due to anti-VCAM-1 USPIO administration...... 87

Figure 2.21: WHHLMI rabbit exhibits MR signal cancellation within 5 minutes of anti- collagen type I USPIO administration ...... 87

Figure 2.22: WHHLMI rabbit exhibits signal cancellation due to administration of anti- cd-11b conjugated USPIO administration ...... 88

Figure 3.1: A flow chart illustrating the reconstruction of the DM inversion method .. 122

Figure 3.2: USPIOs on PDMS are detectable by C-Scan...... 124

Figure 3.3: Gold is detectable on PDMS using C-Scan...... 125

Figure 3.4: Silica microparticles are detectable on PDMS using C-Scan...... 126

Figure 3.5: Her-2 positive SKBR-3 cells are not appreciably bound by iso ab...... 129

Figure 3.6: Her-2 positive SKBR-3 cells are bound by Her ab ...... 130

Figure 3.7: Her-2 positive SKBR-3 cells are not appreciably bound by iso-NPs...... 131

xvi

Figure 3.8: Her-2 positive SKBR-3 cells are bound by Her-NP about half as well as Her alone...... 132

Figure 3.9: Protein G binds IgG antibodies between the 2nd and 3rd constant regions of the heavy chains...... 132

Figure 3.10: Schematic of a Her-NP nanoparticle bound to a Her-2/neu receptor...... 133

Figure 3.11: Her-2 positive SKBR-3 cells are not appreciably bound by iso ab...... 134

Figure 3.12: Her-2 positive SKBR-3 cells are not appreciably bound by iso-NPs...... 135

Figure 3.13: Her-2 positive SKBR-3 cells are bound Her ab ...... 136

Figure 3.14: Her-2 positive SKBR-3 cells are bound by Her-NP about as well as Her alone...... 137

Figure 3.15: Her-2/neu negative MDA...... 138

Figure 3.16: Her-2/neu positive SKBR-3 ...... 138

Figure 3.17: Isotype-NPs do not appreciably bind SKBR-3 cells ...... 139

Figure 3.18: Her-NP binds SKBR-3 cells...... 140

Figure 3.19: Her binds SKBR-3 cells ...... 141

Figure 3.20: Iso-NP does not appreciably bind SKBR-3 cells ...... 142

Figure 3.21: Her-NP binds SKBR-3 cells...... 143

Figure 3.22: Secondary does not appreciably bind SKBR-3 cells...... 144

Figure 3.23: Her binds SKBR-3 cells ...... 145

Figure 3.24: Iso-NP does not appreciably bind SKBR-3 cells ...... 146

Figure 3.25: Her-NP binds SKBR-3 cells...... 147

Figure 3.26: C-Scan is capable of differentiating between iron oxide NP concentrations and NP presence attenuates signal ...... 154

Figure 3.27: C-Scan is capable of differentiating between NPs bound to Her-2/neu positive tissue compared to tissue alone and NP presence attenuates signal...... 155

xvii Figure 3.28: C-Scan is capable of differentiating between NPs bound to Her-2/neu positive tissue compared to tissue alone and NP presence attenuates signal...... 156

Figure 3.29: Tissue sections and regions used for particle/ab treatments and ultrasound analysis...... 157

Figure 3.30: Her-Gold NPs bind to SKBR-3 cells better than other, control conditions by flow cytometry...... 158

Figure 4.1: Iron oxide cores of annexin V USPIOs averaged approximately 10 nm in diameter...... 196

Figure 4.2: Microfabricated silica microparticles are 2 µm diameter discs...... 198

Figure 4.3: SEMs show stain-etched porous silicon is etch-time dependent...... 199

Figure 4.4: TEM reveals thickness and approximate pore size of stain-etched porous silicon...... 199

Figure A.1: Significant differences between conditions...... 208

Figure A.2: Significant differences in density between conditions for all specimens... 209

Figure A.3: Significant percent differences in density between conditions for all specimens...... 210

Figure A.4: Significant differences in stiffness between conditions for all specimens. 211

Figure A.5: Significant percent differences in stiffness between conditions of all specimens...... 212

Figure A.6: Significant differences in density between conditions for cell specimen I 213

Figure A.7: Significant differences in density between conditions for cell specimen II214

Figure A.8: Significant differences in density between conditions for cell specimen III ...... 215

Figure A.9: Significant differences in density between conditions for cell specimen IV ...... 216

Figure B.1: Her binds SKBR-3 cells...... 218

Figure B.2: Iso-NP does not appreciably bind SKBR-3 cells...... 219

Figure B.3: Her-NP binds SKBR-3 cells ...... 220

xviii

CHAPTER 1

INTRODUCTION AND LITERATURE REVIEW

1. Introduction and Literature Review

1.1 Nanoparticulate Platforms The broad utility and potential of nanoparticles in biomedicine has emerged from interactions between complementary scientific disciplines such as materials science, chemistry, physics, biology, and the medicine. The diversity of disciplines contributing to nanoparticulate platforms is transparent upon examination of the immensely varied array of nanoparticles observed in the literature and scientific community. Nanoparticle underpinnings range in approach from organometallic chemistry to biochemistry and from polymer chemistry to semiconductor science. Nanoparticles are not appealing solely because of their minute dimensions, but their size regime does allow them to interact uniquely with biomolecules on their own scale – this feature endears them to biologists and clinicians. Meanwhile, many of the size-based emergent physicochemical properties of nanoparticles are particularly intriguing to physicists, chemists, and engineers. Therefore, when these two groups of scientists interact collaboratively, great clinical advances might be achieved. The arrangement of biomolecules in or on nanoparticles is a critical element in the creation of functional biological nanodevices, which may most efficiently comprise modular groups of biomolecules and synthetic nanomaterials interacting in concert to achieve designated objectives in clinical diagnostics and therapeutics [(1, 2)]. Because biological molecules such as DNA, proteins, and other biopolymers are dimensionally on the order of nanomaterials, they may comprise an ideal interface between apparatus and organism. Indeed, nanobiological devices (i.e., nanodevices which incorporate biomolecules for specific functionalities) may be prototypical of therapeutic platforms,

1 which have been championed by multiple federal health and science agencies [(3, 4)]. Therapeutic platforms consist of devices that display greater than two clinically significant qualities which include sensing early, macromolecular recognition, defined molecular signatures of disease, producing a detectable signal for the transmission of sensed information, allowing appropriate data analysis to provide a framework for decision-making to ascertain subsequent actions, controlled delivery of a suitable therapeutic intervention(s), and monitoring therapy and disease progression in real time [(3)]. These features may engender specific prognosis-improving emergent properties such as a reduction in systemic toxicity exposure, site-specific disease targeting, reduced disease burden due to earlier recognition, increased therapeutic activity at disease sites, identification of residual disease, and the ability to control both therapeutic dose and extravasation parameters. Optimizing the size and density of nanoplatforms improves the efficiency of delivery to a selected site and small molecules, peptides, proteins, and nucleic acids can be loaded into them in a “Trojan horse” approach such that they are not identified by the immune system [(5)]. Further definitions of therapeutic nanomaterials can be perused in [(6, 7)]. Nanoparticles for in vivo biomedicine are designed with substantial consideration of materials toxicity, size, geometry, and dispersity. All greatly contribute to the capacity of the nanodevices to perform the role for which they were devised. Toxic substances, for instance, will prohibit a device from achieving regulatory approval and size/geometry/dispersity/surface functionalization/charge are major factors in the system’s interactions with the nanodevices (e.g., how and when circulating nanoparticles are removed from the vasculature). An incredibly diverse array of fabrication approaches, spanning the fields from which they derive, has been employed to manage these clinical concerns with various levels of success. Device construction approaches can generally be separated into one of two main nanofabrication schemes: top-down and bottom-up. While top-down methods generally employ microfabrication or similar techniques to fashion structures via removal/modification of adjacent materials, bottom- up schemes commonly adapt assembly and chemical methods to construct devices from constituent elements; some strategies seek to hybridize the two approaches.

2 Microfabricated nanostructures comprise a significant segment of biologically- useful nanoparticles because of the high degree of characterization around microfabrication processes from other fields such as semiconductor processing. However, these techniques often suffer from low throughput [(7)]. Thus electron-beam, dip pen nanolithography, force microscopy, and even standard photolithographic techniques encompass a set of methods that facilitate very uniform, controlled nanoparticle creation, but may in some cases prove impractical due to the low volume and the costs necessary to furnish sufficient clinically valuable devices [(6, 7)]. However, advances in highly parallel microfabrication such as those in dip pen nanolithography may yet serve to render microfabricated structures highly effective in the clinic [(8)]. Dip pen nanolithography is fundamentally a direct-write technology in which a scanning probe microscope tip is dipped into an “ink” (comprising molecules such as peptides to be “written”) and dragged across a substrate [(7)]. Silicon microfabricated particulate structures, particularly those comprising porous silicon, are expected to make an impact on clinical care due to the variety of mechanical, chemical, luminescent, and biocompatible/biodegradable advantages they present [(9, 10)]; see section 1.1.2 for more detailed discussion of silicon-based structures in biomedicine. Microfabricated structures other than silicon that may be biomedically useful for detection modalities include, for instance, nanowires comprising wide bandgap III nitride materials because of superior properties such as its direct bandgap status, durability, refractory characteristics, and ability to alloy to emit light in the ultraviolet to visible range, among others. Conversely, bottom-up chemical fabrication methods offer high throughput synthesis reminiscent of industrial processes that are optimized to provide large quantities of materials [(7)]. However, these techniques yield structures that are generally not as uniform (in size or geometry) or controllable as microfabricated devices. Furthermore, devices constructed through bottom-up methods such as self-assembly do not yet display the level of functionality of some microfabricated nanodevices [(5)]. Nevertheless, a plethora of chemically-produced organic and inorganic nanoparticles have been developed for in vivo and analytical biomedical applications. Polymeric (organic) nanoparticles present many advantages, as their chemistries are extremely flexible; their

3 use has thus become somewhat prevalent, so a comprehensive review has been published covering polymer nanotherapeutics [(11)]. The flexibility of chemistry allows them to be molded into such diverse architectures as linear, graft, branched, cross-linked, star, and block polymer and co-polymers, as well as the versatile dendrimer-structure [(12, 13)]. Dendrimeric polymer chemistries are valuable because they enable controllable dimensions and geometries. Furthermore, dendrimers can be engineered to exhibit a representative property of one of the central tenets of controlled chemistries: self- assembly. Dendrimers can be coaxed into self-assembling into higher order aggregate structures that can be used in fields such as oncological therapeutics as in Figure 1.1 [(7)]; indeed, the dendrimer possesses many features characteristically critical to the development of a nanoparticulate platform. Monodispersity is high, solubility, size, and geometry are controllable, biocompatibility/toxicity are generally considered good from a materials perspective, and core materials and functional groups for derivatization can easily be interchanged [(12-14)]. Medical contributions of dendrimers include MRI- detectable constructs as well as multifunctional platforms that direct interventions to cancer cells [(11, 15-17)]. Tomalia et al. described the development of larger, multifunctional platform based dendrimeric structures termed “tecto-dendrimers” [(18)]. For instance, in the cancer work, Dr. Baker’s group exploited directed assembly techniques to produce such supramolecular dendrimeric structures capable of delivering site-specific contrast and therapeutic by employing targeted interaction against cancer sites [(7, 16, 17)]. Complementary strands of DNA were used to self-assemble fluorescein (FITC) and folic acid-conjugated dendrimers to target cancer cells which overexpress folate receptor [(16, 17)]. Other nanoparticle platforms, including both top- down and bottom-up structures, could also reasonably be appended to the central dendrimer via bioconjugate chemistries.

4

Figure 1.1: The image represents a hypothetical, modular nanotherapeutic modeled from the oncological, self-assembled tecto-dendrimer-based cluster agent from the Baker and Tomalia groups [(16-18)]. The dendrimer subunits are grown from initiator cores (C), and the tunable dendrimeric surface groups are represented by Z. All dendrimer subunits have specific, dedicated functions in the device: here, the inner dendrimer encapsulates small molecule therapeutics (E), whereas other functional components are isolated to other dendrimer constituents. These include biochemical functionalities such as targeting/tethering (Ta), therapeutic triggering to enable activation of pro-drug components of the device by an external operator (Tr), fluorescent, metal, or other constituents for imaging (I), and sensing functions (S) to mediate innately controlled activation/release of therapeutic. This design, which is one of a seemingly infinite quantity of potential similarly interchangeable, patient-tunable functional therapeutic configurations, is representative of a therapeutic platform [(3)] precisely because of its modular configuration [Figure from (7)].

5 Chemists designing inherently thermodynamically driven self-assembling architectures often seek to mimic or exploit biological self-assembly processes that generate highly complex, multifunctional systems [(19)]. Diatoms, for example, comprise a group of highly ordered unicellular algae which are unique due to their silica (silicon dioxide) microstructure, uniform nanopore/microchannel structures, mechanical resiliency, and chemical inertness [(20)]. The complexity of these purely biological structures, which would be exceedingly difficult to replicate using microfabrication or other synthetic methods, may be sufficient to be able to employ only slight modifications in order to apply them in membrane, , or other applications. While diatoms represent use of biological self-assembly for engineering utility, wholly synthetic self- assembled functional architectures include polymeric micelles, , metals and metal oxides, specifically designed biomolecules such as peptides, and even amphiphilic prodrugs [(7, 21-26)]. While self-assembled polymeric configurations often afford some measure of precision by controlled radical reaction, most such nanoparticle-sized structures offer reduced precision but greater options for device construction such as micelle- and membrane-templated syntheses and various deposition techniques [(19)]. Specific decoration with complementary base pairs of nucleic acids to the materials to be assembled has become an increasingly common technique to promote directed self- assembly [(27-29)]. Inorganic nanoparticles such as semiconductor quantum dots and metallics can also be prepared using a variety of chemical techniques. Quantum dots in particular are well-known for their precisely-defined dimensions, as the size of the core determines their interesting optical properties. Quantum dots are inherent high-quantum yield optical reporter structures, robustly and photo-stably emitting light when excited by the appropriate wavelength. Broad absorption and narrow emission are other desirable features integral to this technology. Furthermore, they can be conjugated through surface functional groups, allowing them to be targeting or delivery vehicles without quenching of optical attributes; it is thus not surprising that quantum dots are advocated as potential single-molecule detection modalities that have promise for highly sensitive, parallel biomolecule analyses [(30)]. Other microfabricated and chemically-prepared inorganics form an extensive class of nanostructures and include nanovectors, nanowires,

6 nanocantilevers, and metal/dielectric nanoshells used for cancer detection and therapy [(31, 32)]. Nanoshells in particular represent a category of “smart nanostructures” that can be spatiotemporally delimited in the context of targeting and delivery by an external stimulus (such as light, radiofrequency, ultrasonic, or other energy). Nanoshells can be targeted to the appropriate disease site, e.g., a tumor, using proteins covalently bound to the surface. Due to tunable optical characteristics, they are best imaged using the near infrared spectrum. Furthermore, nanoshells are harmless once they reach their destination, until they are irradiated with near infrared light triggering local photothermal ablation to destroy the diseased cells to which the nanoshells have bound [(32)]. By three-dimensionally and temporally delimiting the heating action which kills abnormal cells, nearby healthy cells are spared; this is an excellent example of an external triggering strategy – for a detailed discussion on triggering strategies by means of physiological, external, and secondary signaling stimuli see [(7)]. Another class of smart nanostructures avoids remote triggering methods. These structures involve devices that interact with their environment at an appropriate site via intrinsic physiological, chemical, or thermal sensor/reporter compositions [(5)]. In addition to dendrimers and nanoshells, another recent innovation in oncological therapeutics epitomizes the multifunctional approach characteristic of nanoparticulate platforms [(33)]. This structure is “smart” due to its intrinsically engineered temporal mode of therapeutic delivery. The designers carefully considered the biological obstacles which the nanostructure, termed a “nanocell,” would need to overcome. Major barriers are faced by standard tumor delivery systems – a delivered anti-angiogenic drug, used to destroy the vasculature feeding the tumor, may in fact hinder the delivery of a therapeutically efficacious concentration of chemotherapeutic to the tumor. The reduction in blood supply causes accretion of a hypoxic factor in the tumor which amplifies both the tumor’s invasiveness and its resistance to . To evade these difficulties, the nanocell is preferentially taken up by the tumor and employs a two- pronged temporal release strategy that is founded upon the engineered “nuclear cell” design of the particle. An extranuclear PEGylated lipomsomal structure surrounding a nuclear core is designed to enable an initial release of anti-angiogenic therapeutic

7 concealed within the lipid layer to cut off tumoral blood supply. The nuclear core, containing chemotherapeutic, is thus confined to the tumor and releases drug to destroy the tumor, an approach which reduces inadvertent toxicity and strengthens therapeutic efficiency [(33)]. Iron oxide nanoparticles comprise another class of inorganic structures which have been broadly useful for clinical means (e.g., imaging with magnetic resonance, or MRI), tissue repair, detoxification of biological fluids, hyperthermia, drug delivery, separation of cells both for research and clinical purposes, and immunoassays [(34)]. Iron oxide is chiefly valuable as a reporter for MRI because of its magnetic susceptibility, i.e., primarily its ability to modulate MRI contrast by shortening T1 and significantly

shortening T2 relaxation times with standard pulse sequences. This effect has long been utilized and is evident in MR images through signal reduction in regions of particle localization [(35, 36)]. Iron oxides are most commonly synthesized by wet chemical co-precipitation methods into nanoscale crystals and have been packaged within a variety of sugar, polymer, metallic, surfactant, and other biocompatible coatings (e.g., polyethylene glycol (PEG) and dextrans) for stability and ease of derivatization [(34)]. The utility of iron oxide nanoparticles derives from a combination of their demonstrated biocompatibility and in vivo signal cancellation properties and the flexibility of the surface functionalizations on their coatings to append a wide variety of peptides, oligonucleotides, proteins, and small molecule targeting entities and therapeutics. Such extensive bioconjugation capacity confers the ability to link a variety of targeting and drug delivery systems for various disease states to these nanoparticulate platforms. See Chapter 4 for further discussion of iron oxide nanoparticle platforms. Biodistribution, clearance, and immunogenicity are major factors determining nanoparticulate platform efficacy. Biodistribution measures the pattern of particulate localization to sites other than the intended site, while clearance refers to the mechanism, kinetics, and length of time it takes nanoparticles to clear from the blood pool. Knowledge of biodistribution is important because it serves as an alert to potential toxicity concerns and it is also imperative for identification of where and how long deposits of nanoparticles in certain organs or tissues may persist. Only slight

8 modification of the surface chemistry, charge, size, or material type can significantly impact the biodistribution of injectable nanoparticles [(7, 37)]. Likewise, these variations in physical properties can yield dramatically altered clearance effects by the reticuloendothelial system (RES) [(7)]. The RES, comprising the thymus, liver, and spleen, eliminates materials from the bloodstream by both passive diffusion and active routes such as receptor-mediated endocytosis [(7)]. Clearance data is also paramount, as it represents the length of time that nanoparticles are exposed to disease sites. Interpretation of clearance data depends upon whether it is desirable for the nanoparticles to extravasate or bind to vascular sites. While in some cases it is even desirable for the particles to clear, as when it is appropriate to image a clearance organ(s) itself, such as the kidney for glomerular disease [(38)], clearance is often an obstacle to be overcome by modulating the appropriate physical properties and using targeting molecules to bind nanoparticles to desired sites. In addition to legitimate clearance-based targeting endeavors such as via the RES, tumor-specific effects can be used to direct materials to cancer. Triggered by low tumoral vascular integrity and lymphatic drainage, Enhanced Permeability and Retention (EPR) denotes the effect which occurs when high molecular weight molecules (e.g., potentially nanoparticles) extravasate within tumors and are deposited and retained there due to the limited lymphatic system [(7)]. Last, parenterally- administered nanostructures may contain an epitope or epitopes which induce an immune response in the organism, such as the antibody response to intravascularly-administered dendrimers [(39)]. If an immune response is stimulated, subsequent administrations of the reagent will be rapidly recognized and cleared from the blood pool. Nanoparticles may also be opsonized (i.e., adsorption of proteins that interact with clearance molecules such as receptors of cells which clear foreign objects) and rapidly cleared by monocytes and tissue macrophages [(40)], which is a major reason why nanoparticles are often designed to prevent protein adsorption via surface functionalizations and attachment to molecules such as PEG (which inhibits protein adsorption). Particles must also generally be small in dimension to evade clearance by filtration in the first capillary bed crossed [(40)]. In terms of the design and construction of a nanoparticulate platform, all the above circumstances must be taken into consideration. Thus, after the fundamental

9 nanoparticulate material and configuration has been formulated, its final size, charge, surface characteristics, etc. must be carefully chosen both to avert biodistribution/clearance/immunogenicity issues and to perform the tasks (whether imaging, drug delivery, targeting, or other purpose) for which it was proposed. Furthermore, targeted nanostructures must be safe when dispensed over multiple treatments/doses, and must be of sufficient size to transport an appropriate imaging and/or therapeutic payload, yet retain the capacity to be taken up or bound by targeted cells. In order to design such a platform, a modular approach may be taken to facilitate use of the nanoparticle in multiple systems for a variety of disease states. This method can produce a convenient platform for which minor modifications and substitutions to the device can generate unique, sometimes unexpected (e.g., emergent) valuable properties. Indeed, platform modularity might be most efficiently exploited using a systems approach that describes the functional constituents of a device through logical objects- oriented approaches and modeling language [(1, 2)]. In such approaches, functional abstraction of biological components enables greater understanding of the capabilities of a nanodevice [(1, 2)], thereby stimulating the design of an improved, more efficient, and more adaptable biological nanodevice. Once the device has been designed with appropriate surface functionalization and the targeting moieties have been chosen, protocols must be established to link the targeting molecule with nanoparticle surface without hindering the functionality of the targeting agent and without causing nanoparticulate aggregation. These procedures may employ either non-covalent (e.g., hydrophobic or electrostatic interactions, van der Waals forces, etc.) interaction or, more commonly, covalent bioconjugate chemistries.

1.1.1 Bioconjugation A very broad array of surfaces and nanomaterials exists to which it may be desirable to link various targeting moieties, therapeutics, and other molecules. Because so many of these surfaces provide viable functional groups, extensive studies have been performed to generate biomolecularly compatible chemistries (i.e., chemistries and solvents compatible with individual biomolecules) amenable to biochemical derivatization of these surfaces. While bioconjugation of nanomaterials can be used for

10 diverse purposes ranging from a means to achieving self-assembly to the mediation of electron transfer in nanobioelectronics to imparting biocompatibility, the focus in this dissertation is on bioconjugation for affinity-mediated site-specific targeting of nanoparticulate platforms. Indeed, the tissue-specific or disease-specific delivery of nanostructures is commonly performed through biological affinity interactions, usually via proteins or glycoproteins specific to the tissue (known as tissue-specific antigens) or disease (e.g., overexpressed cell surface growth receptors, as in some cancers) interrogated. While covalently bound monoclonal antibodies provide a convenient means to mediate tethering of nanostructure to target cell (e.g., an antibody raised to an overexpressed disease marker), an abundance of molecule types exist for such purposes, including oligonucleotides, synthetic polymers, and other proteins and peptides. Whatever the molecule chosen for targeting, the selection of appropriate chemistry is critical. Stoichiometry, functionality, and efficacy of targeting reagents are all affected by the chemistry chosen. While most linkages involve covalent bonds between nanostructure and targeting reagent, other, less robust associations can be employed such as hydrophobic interactions, as in the case of some protein interactions with carbon nanotubes [(41)]. Covalent chemistries, a very common means of biomolecule immobilization, can be promiscuous, orthogonal, or chemoselective (chemoselectivity entails the preferential reaction of one molecule/chemical with one of two or more different functional groups). A frequently used promiscuous chemistry employs the heterobifunctional compound EDC (1-ethyl-3- [dimethylaminopropyl]carbodiimide hydrochloride). This cross-linking molecule provides a zero length bond between amines and carboxyl functional groups. It is considered “promiscuous” because it will link any amine functionality with any carboxyl group. For proteins such as some enzymes and intact antibodies, this promiscuity generally does not impact biological functionality to a great degree because only a small percentage of the molecular surface area is implicated in the required biofunctionality (i.e., it still retains the same affinity for its cognate subsequent to conjugation) [(7)]. A wide variety of hetero- and homobifunctional cross-linkers are available which link two different or identical functional groups (e.g., maleimide, thiol, carboxyl, carbonyl, amino, hydroxyl, and phosphate), respectively, at a variety of linker arm separation lengths in an

11 assortment of solvents [(42)]. Another disadvantage of promiscuous chemistries, other than the potential for lack of biofunctionality, is that they can yield unwanted products and aggregates unless carefully prearranged. However, they tend to offer relatively straightforward, succinct protocols. Orthogonal chemistries, on the other hand, are site-directed. They rely upon the selective reactivity of electrophile-nucleophile pairs to enable site-specific linkages between nanomaterials and biomolecules [(7)]. The preferential binding properties allow detailed design of the bioconjugation procedure, thereby yielding a specified three- dimensional organization of the device, so that bound molecules remain in a conformationally bioactive state. Indeed, derivatization to a functionally significant amino acid, base, or functional group (in proteins, nucleic acids, and small molecules, respectively) can neutralize biofunctionality [(7)], emphasizing the occasional necessity for site-specific methods. Orthogonal chemistries can also engender greater yields of active nanodevices, potentially making them less toxic and more cost-effective as less of the nanomaterial must be used. Therefore, orthogonal chemistries have been commonly incorporated into nanodevices [(7, 43, 44)].

1.1.2 Silicon Particles The use of microfabrication in medicine is equipped with an extensively well- characterized, established infrastructure courtesy of the computer industry. Not only have standard microfabrication equipment and processes been well-characterized (e.g., oxide and nitride growth, metal deposition, chemically or crystallographically selective etching protocols, photolithography, etc.), but so have the surfaces produced from these procedures. Thus, the biomolecular derivatization of silicon-based surfaces rapidly stimulated great advances [(10, 45-47)]. Because material surfaces are so critical in biomedicine, silicon has emerged as a key contributor to the field. Silicon, depending on its state (e.g., oxidized, porosified, etc.), can be functionalized via numerous methods including by means of hydrosilylation; this is a method commonly used to prepare monolayers of Si-C bonds on freshly etched silicon surfaces [(10)]. Not all microfabricated biomedical nanostructures require surface functionalization for targeting of a specific site or cell, but it is often imperative as a

12 means to biocompatibility and prevention of bio-fouling (a process of protein and other biomolecule surface adsorption, often hindering device function) for drug delivery devices such as those described in the following paragraph [(48)]. Silicon is a very attractive material for drug delivery due to its compatibility with, and the depth of understanding in, electrical conduction – this yields the feasibility of drug delivery vehicle, hardware, and software onboard a single device configuration. Such a structure would theoretically facilitate pre-programmed release schedules, remote activation via radio frequency, and algorithmic intelligent feedback attributes and sensors. While this dissertation is focused on parenterally-administrable targeted nanodevices, it is important to note substitute modes of drug delivery such as implantable release devices. For example, recent advances in the biomedical applications of silicon microfabrication include the Langer group’s controlled release microchip. Briefly, a solid-state microchip was configured to effect controllable pulsatile release of one or more chemical substances through microreservoirs etched via standard techniques [(49)]. Release of encapsulated therapeutic was realized through the controlled electrochemical dissolution of thin metallic anode layers atop the drug formulations. This technology must be implanted at the appropriate site and is being tested in animals using, for example, the brain cancer drug BCNU [(50)]. For similar purposes of drug delivery, another silicon technology involves membrane construction by microfabrication of uniform pore sizes on the order of 10 nm. This nanochannel gating membrane controls release of small molecules like therapeutics, which are on the dimension of the pore size itself, such that linear diffusion, a desirable feature in drug delivery, transpires [(51)]. Another use for the membrane technology includes immunoisolation of cellular transplants, wherein the pores may be sufficiently small to impede immune molecules (hence isolating them from immune response) but large enough to allow nutrient flow for cellular subsistence and to regulate their intended organ or system (e.g., in the case of a diabetic patient, glucose-insulin regulation might be performed by implanted, immunoisolated pancreatic islet cells) [(52)]. Using standard lithographic and sacrificial layer techniques, silicon-based particulates of various materials (silicon dioxide, porous silicon, etc.) have been fashioned and then chemically functionalized for biomedical applications [(5, 31, 53)].

13 They have been evaluated for size-based acute toxicity in vivo [(54)]. Photolithography is a prevalent tool which employs light and masks for the creation of highly uniform, high throughput silicon-based structures with features 1 µm or greater. While other technologies offer much smaller structures (e.g., direct-write methods such as electron beam and dip pen nanolithography), they suffer from low throughput, high cost, or both [(7)]. Many methods also employ sacrificial layer techniques, which enable construction of a device upon a layer which is subsequently etched to release all the devices on a single silicon wafer into a . Porous silicon (poSi) comprises an increasingly investigated silicon-based material due to its unique luminescent (e.g., photo-, electro-, and chemiluminescence), mechanical, biodegradable/biocompatible, and molecule uptake/release properties [(10)]. Fabrication of poSi by electrochemical anodization or chemical “stain” etching has been reviewed elsewhere [(10)], and will be described in further detail in Chapter 4. A previous review has also captured poSi functionalization uses and methods, which exploit either the native oxide layer which rapidly grows on freshly etched silicon or hydrosilylation chemistries on the freshly etched poSi surface [(10)]. One application was recently tested in a mouse tumor model; this study drew upon the biocompatibility and durability of poSi, in addition to its ability to imbibe 32Phosphorus for ensuing release and suppression of carcinoma xenografts [(9)]. Although the particles in this study were untargeted and injected directly into the tumor, this work provided proof-of principle for the utility of nanoporous poSi particulates in vivo. Particulates of silicon dioxide and porous silicon were fabricated and investigated for their properties in creating mechanical density in the ultrasound/breast cancer study of Chapter 3.

1.2 Atherosclerosis Atherosclerosis, an arterial disease associated with severe morbidity, is the most common cause of death in Western societies; indeed, more than 25 million in the US suffer from a clinical manifestation of atherosclerosis, a figure which does not consider those with presently concealed risky cardiovascular features [(55)]. While many of the details of atherosclerotic pathogenesis are as yet unidentified, it has become increasingly apparent that the disease results from a complex inflammatory cascade, rather than

14 abnormal lipid metabolism as previously believed [(56-58)]. The process of atherosclerosis involves a great diversity of cytokines, cells, and signaling molecules which produce the unstable progression of lesions that lead to disruptive, thrombotic events, e.g., heart attacks and strokes [(57, 59)]. Histologically, plaques predominantly comprise three constituents: cellular (mainly macrophages, smooth muscle cells, and endothelial cells), extracellular lipid, connective components, and lipid localized intracellularly to macrophages (i.e., foamy macrophages) [(57)]. Inflammatory insult in combination with disturbances in these components causes progression of the atherosclerotic lesions. In particular, post- inflammatory events such as smooth muscle cell and connective tissue proliferation, cytokine release, recruitment of circulating cells such as various types of leukocytes, and the rapid accretion of lipid and macrophages lead to the advancement of plaque as specified in Figure 1.2 [(57, 59, 60) and Figure 1.2]. The inflammatory sequence is an elaborate mechanism by which plaques may become vulnerable to rupture (see section 1.2.4).

15

Figure 1.2: Cartoon of the pathogenesis of atherosclerosis from the perspective of the field of molecular imaging. Monocytes and other cells roll along and adhere to endothelium via receptors such as VCAM-1, migrating into the intima and forming cells such as macrophages. Monocytes and macrophages express specific receptors that sustain the processes of cellular and molecular proliferation and atherogenesis, eventually forming lipid-laden foam cells, which release thrombogenic, degradative, and oxidative molecules. Macrophages then undergo apoptosis, releasing additional molecules and bodies relevant to the progression of atherosclerosis. VCAM-1 = vascular cellular adhesion molecule 1, CCR2 = Chemokine receptor 2, MCP-1 = monocyte chemoattractant protein 1, M-CSF = macrophage colony-stimulating factor, MMP = matrix metalloproteinase, ROS = reactive oxygen species [Figure from Jaffer et. al. [(61)].

Early stage atherosclerosis is predicated upon the recruitment of inflammatory cells and molecules to the endothelial surface due to abnormalities in the endothelium, which are not well understood. Endothelial insult (e.g., oxidized low density lipoprotein, or LDL, accumulation in the arterial wall, infection, mechanical injury) begets early inflammatory processes such as the overexpression of vascular cell adhesion molecule-1 (VCAM-1, or cd-106) and P- and E-Selectins on the endothelial surface [(60, 62)]. Thus, whereas normal endothelium is resistant to leukocyte (T-cells and monocytes) adhesion, endothelial insult effectively leads to recruitment of these cells via ligand-receptor interactions [(60)]. Leukocytes then release monocyte chemo-attractant protein-1 (MCP-

16 1), which amplifies the inflammatory process by activating and recruiting leukocytes to the endothelial surface, facilitating their transmigration through the arterial wall, and triggering the proliferation of smooth muscle cells [(57, 63)]. Signaling cascades subsequently induce monocytes continually to adhere to the vessel wall and migrate into basement membranes, which are being degraded by matrix metalloproteinases (MMPs), futher facilitating passage of inflammatory cells and molecules. Macrophages, formed from recruited monocytes, and T-cells migrate into the medial layer while releasing cytokines and MMPs, resulting in further matrix degradation. Addition of macrophage colony stimulating factor (M-CSF) stimulates the progression of atherosclerosis into the fatty streak stage [(60)]. Advancement of the plaque is dependent upon a broad array of inflammatory mediators that include cd-40L, interleukin (IL)-1ß, tumor necrosis factor (TNF) and ß, IL-6, M-CSF, MCP-1, and IL-18; for instance, antibodies directed to cd- 40L in mice have been shown to inhibit plaque progression, and further, to modify the lesional composition more closely to resemble that of a stable plaque (see section 1.2.4) [(57, 64)]. Morphologically, plaques are characterized by three major stages; first is the fatty streak (generally manifest by age 20 in humans), which at first may only be visible via lipid staining, but becomes grossly observable, raised, and yellowish-gray in appearance. These are mainly due to lipid-engorged macrophages and other cells. Next, fibroatheromas replete with a fibrous cap may be formed, caused by intimal thickening and proliferation of smooth muscle cells, extracellular matrix, and extracellular lipid (including cholesterol clefts, free cholesterol, and cholesterol esters [(60)] originating from the apoptotic and necrotic death of foam cells). Even if these lesions reduce the lumenal diameter, they are generally not considered dangerous unless they rupture or undergo other insult. A thinning fibrous cap, denudation of the overlying endothelium, and a large lipid core constitutes the final stage [(63)], the complicated lesion. This advanced lesion may erode to expose thrombogenic core materials such as procoagulants, tissue factor, and von Willebrand factor to the bloodstream, potentially leading to rupture of the thin, weakened fibrous cap (see section 1.2.4.1 on plaque vulnerability biomarkers [(55)]). The rupture causes thrombosis, which partially or completely occludes blood flow. In some cases, the plaque itself may then grow around the thrombus, yielding

17 plaques that contain sometimes multiple thrombi beneath the caps. Other studies group the phases of atherosclerosis into five phases and multiple types as displayed in Figures 1.3 and 1.4.

Figure 1.3: Illustrates the 5 phases and types of lesions in atherosclerotic development. Intitial lesions lead to early, type III fatty streaks, advancing into type IV or Va “vulnerable” lesions, which may be prone to rupture due to fibrous cap weaking. Types IV and Va, in addition to the “complicated” type VI lesion in Phase 4, are most relevant to heart attacks. Disruption of type IV or Va lesions, which contain extracellular lipid cores underneath fibrous caps, leads to thrombosis, which may be acutely occlusive [(63, 65)]. While phase 1 lesions are considered early atherosclerosis, phases 2 (which are not necessarily stenotic) through 5 (calcified or fibrotic lesions causing angina) comprise the levels of “advanced” atherosclerosis [(66)]. Figure from Fuster et. al. [(65, 66)].

While atherosclerosis may develop in arterial beds such as the cerebrovascular, renal, peripheral (e.g., lower extremities), coronary, and aortic arteries, this dissertation is focused upon detection of lesions in the abdominal aorta. The statistics for those with aortic atheromas are difficult to obtain, as the population-at-large remains generally unscreened when lacking clinical symptoms, but approximately 8% of patients

18 undergoing surveillance for routine clinical indications displayed atheromas in the aorta [(55)]. For those with significant carotid disease and those with acknowledged obstructive coronary disease, aortic atherosclerosis rates rise to 38% and 90%, respectively [(55)].

Figure 1.4: Designates the histology of atherosclerotic lesions as defined by the AHA in Stary et. al. [(67)]. The center column designates the direction and sequence in which pathways in human plaque develop and the numerals indicate the lesion type at left. Lesion type also corresponds with types illustrated in Figure 1.3. Lipid accumulation, both within macrophage foam cells and extracelullarly, accounts for type I to type IV advancement, and the loop from type V to VI represents thickness increase due to thrombosis on plaque surfaces. At right are age of onset of specific plaque morphologies and potential clinical manifestations of disease [(67)].

19 1.2.1 Clinical Background The assessment of risk factors is an essential element in the prevention and risk reduction of atherosclerotic diseases, particularly in asymptomatic individuals. Atherosclerosis is currently considered to have a number of clinical manifestations and four groups of risk factors: conventional, predisposing, conditional, and emerging [(68)]. Heredity may drastically impact any one or more of these groups – but, while the genetic risks of atherosclerosis are significant, no major gene has yet been found to be directly correlated with the disease (except possibly in rare cases of mendelian inheritance – i.e., subject to simple mendelian analysis rather than typical multifactorial disease); rather, only intermediary risk factors which each exert an often small, but additively significant role in the development of clinically significant atherosclerosis are linked to genetics [(55)]. For instance, the risk factor calcification is up to 50% established by heredity; other inherited risks such as obesity, diabetes, and hypertension also contribute to atherosclerotic predisposition [(55)]. However, it is clear that environment and lifestyle comprise a major component of the risk, particularly in the intensification of genetic factors in the development of advanced atherosclerosis. Conventional risk factors alone can describe only less than 50% of the variability in quantitatively measured atherosclerosis [(69, 70)]. These risks, which have a causal role in atherosclerotic progression and may be either modifiable or nonmodifiable, include cigarette , hypertension, high serum cholesterol, low high density lipoprotein (HDL), and diabetes [(69, 71)]. These factors are also known to interact with other types of risk factors, often in uncertain ways, to contribute to atherogenesis. However, it is notable that a thorough study showed that a single risk factor – cigarette smoking – decreased the age of a coronary event by a decade no matter the risk factor combination [(69)]. Nevertheless, though 80%-90% of people who undergo a cardiovascular event possess one or more conventional risk factor(s), many individuals lacking these risks also suffer cardiovascular events and people with multiple risk factors sometimes do not [(69, 71)]. Therefore, it would be optimal for physicians to be independently able to predict a patient’s risk of a cardiovascular event based on a genetic analysis, a circulating biomarker, or group of biomarkers [(72)], e.g., a molecular signature. The American

20 Heart Association (AHA) in 1999 established a group of biomarkers including homocysteine, fibrinogen, lipoprotein A, apoB-100, LDL particle size, and C-reactive protein to be conditional risk factors (i.e., factors that have an “association with increased risk for coronary artery disease although their causative, independent, and quantitative contributions to coronary artery disease are not well-documented”) [(55, 68)]. Immunoassays, electrophoresis, and other biological analyses are available to assess these risks with varying degrees of accuracy/sensitivity [(68)], but they are not currently clinically utilized, as present indications demonstrate only that conditional risks elevate cardiovascular risk with respect to conventional risks [(68)]. Thus, further work in the basic science of identifying relevant risk factors must be completed prior to achieving predictive capacity. See section 1.2.5 for a description of genetic animal models that facilitate elucidation of some conditional risk factors. Another group of risk factors includes the predisposing factors, which comprises environmental and genetic factors: obesity, sedentary lifestyle, heredity (including male gender and insulin resistance to some degree), physical inactivity, and behavioral and socioeconomic elements. Predisposition may exert direct, causal effects on atherosclerosis, but is generally known to modulate disease risk in interactions with conventional risks [(68)]. While emerging risk factors have not yet been well studied, they may represent the best prospects for identifying a single, or even multivariate, risk factor/biomarker highly correlated with cardiovascular events. These include lipoprotein-associated

phospholipase A2, pregnancy-associated plasma phosphatase, asymmetric dimethylarginine, myeloperoxidase, nitrotyrosine, candidate gene polymorphisms, and oxidative stress markers via reactive oxygen species [(68)]. Once the four classes of risk factors converge to develop into advanced atherosclerosis, the resulting clinically significant plaque is due to foam cell accumulation and characterized by intimal narrowing (which impedes flow) and neovascularization [(57)]. Clinical manifestations of atherosclerosis transpire very late in the inflammatory process described in the previous section, which hinders the capacity of interventions to treat underlying disease processes (unless they are identified very early, which is the objective of molecular imaging, described later); rather, treatments tend

21 only to deal with specific, symptomatic sites [(57)]. Manifestations include coronary heart disease, stroke, and peripheral vascular disease [(63)]. Stable angina pectoris is a chronic manifestation (usually as chest pain) of atherosclerosis, commonly due to stable, thick plaques in the arterial system occluding blood flow and resulting in decreased circulation of oxygen to cells. However, unstable angina is an acute condition caused by thin-capped plaque rupture and is characterized by longer, unexpected periods of chest pain compared to stable angina. This is the most common clinical manifestation of atherosclerosis and it may be lethal as when rupture causes complete blockage due to resulting thrombosis, or a myocardial infarction/heart attack, which restricts cellular oxygenation to cause tissue ischemias that sometimes precede death. Thus, the general objective of risk factor classification and stratification is to enable clinicians to predict cardiovascular event manifestations and administer appropriate interventions before the foregoing occurs.

1.2.2 Modes of Detection and Intervention Presently, atherosclerosis goes effectively unnoticed until a lesion has advanced to the point of physical symptoms, at which time it is too late in the pathogenesis of the disease to intervene except to treat the lesion itself (rather than the underlying disease). Therefore, detection of atherosclerosis prior to suggestive ailments could enable physicians to treat the disease considerably earlier, which would significantly lessen, or at least delay, the disease’s severe morbidity and mortality. Moreover, such detection schemes could decrease costs to society in terms of patient healthcare costs and lost time in employment. A detection modality, however, must be efficacious while simultaneously avoiding undue deleterious effects on the population it screens. Invasive detection methods (e.g., catheter-based angiography) are generally the “gold standard” for the diagnosis and treatment strategy of cardiovascular disease [(73)], but it is unacceptable to perform such procedures on populations in which a large number of individuals will be healthy. A critical point is that the assessment should conform to the individual’s disease risk: for very low risk populations, no testing need be performed. However, perhaps surprisingly, recent data shows that evaluation of early atherosclerosis for low risk individuals is highly practical and cost effective, while non-invasive tests for

22 intermediate risk patients are exceedingly valuable [(74)]. For general populations, prospective patient analyses must be relatively painless, low cost, and repeatable. Thus, optimal screening modalities are minimally invasive imaging procedures that allow physicians to visualize the arterial system’s morphology and/or underlying pathophysiological/biochemical processes exclusively via an external imaging system that may require at most parenteral administration of a contrast agent. Most invasive atherosclerosis procedures are catheter-based (e.g., angiography, intravascular ultrasound, thermography, and some optical coherence tomography and near infrared spectroscopy techniques [(75)]), allowing physicians directly to visualize the arterial lumen. Coronary angiography consists of an x-ray examination of the blood vessels or heart via an inserted catheter tube. While this technique has provided physicians with excellent data from which to make clinical decisions and interventions, more recent angiographic methods are non-invasive, including MRA (magnetic resonance angiography), transvascular ultrasound, CTA (computed tomographic angiography – both electron beam CT and multislice CT), and DSA (digital subtraction tomography) [(55, 75)]. CTA-multislice uses a multi-detector x-ray scanner to collect a multiple angle/multiple plane single acquisition image of soft tissues and adjacent structures while DSA affords digital reconstruction of the difference between x-ray generated pre- and post-contrast vascular images [(55)]. However, DSA offers only low resolution and both CTA and DSA employ ionizing x-rays. Conversely, MRA has the advantage that it is both non-invasive and uses non-ionizing radiation, which is a particularly valuable aspect for repeat screening procedures; for these reasons, and due to technical advances (though still relatively low resolution), it is considered nearly as good as classic angiography [(55)]. Another non-invasive method, duplex ultrasound, exploits the Doppler effect of fluid flow velocity through vessels – a B-mode real-time ultrasound technique, this enables quantification of degree of arterial occlusion and is considered an excellent technique for post-surgical patient follow-up to identify re-stenosis [(76)]. However, presently the sole indication for asymptomatic screening with MRA or CTA includes patients with a history of Marfan syndrome or certain inflammatory vascular diseases [(55)]. Other non-invasive imaging modalities, such as MRI (magnetic

23 resonance imaging) and a nuclear technique, SPECT (single photon emission computed tomography), are described in the following section. When a potentially deleterious lesion(s) is detected, interventions must be planned in consideration of the fact that atherosclerosis is a systemic disease which may be present in multiple areas of the vasculature – i.e., disseminated vascular disease (because certain interventions have been shown to be most efficacious in specific vascular sites [(55)]). The course of action must thoroughly consider the site of the lesion (e.g., carotid, aorta, coronary, peripheral, renal, etc.) because not only are many therapies site-directed (e.g., stents), but the site may help establish the type of therapeutic administered. In general, non-invasive therapies that modulate atherosclerotic risk factors, such as reducing cholesterol and hypertension (e.g., via drugs, exercise, diet, etc.) are often employed [(55)]. Other therapies often entail plaque stabilization strategies rather than promoting plaque regression [(77)]. Thrombosis (e.g., antiplatelet treatments, anticoagulants, aspirin, heparin) and matrix metalloproteinase inhibiting agents (MMP inhibitors), anti-inflammatories, vasodilators, and other compounds are being studied and may be employed [(76)]. Revascularization procedures such as endografts, endarterectomies, angioplasties, and/or stent placements may be implemented if urgent care is required or if the lesion is not responding to other therapies [(55, 76)].

1.2.3 MRI and SPECT MRI is a volume imaging modality that tomographically produces a slice-by-slice image of an object via analysis of magnetically-mediated photon emission from (generally) hydrogen nuclei within the imaged object. Hydrogen, which comprises some 63% of the , possesses specific spin states which are imageable by MRI with regard to their relaxation times. In particular, hydrogen atoms are aligned by the MRI apparatus’s magnetic field, after which a radiofrequency (RF) pulse is applied to “resonate,” or make the atoms spin in a particular frequency (the Larmor frequency, which is specific to the atom’s tissue environment) and direction due to the absorbed energy. After the RF pulse, hydrogen atoms return to their ground states by emitting the stored RF energy in the form of a detectable photon that can be correlated for image production. This allows differentiation between tissues via fourier transform and other

24 analyses. Processes affecting MRI signal include T1, T2, and T2* relaxation times, proton density, motion, flow, molecular diffusion, and magnetization transfer [(75)]. Certain paramagnetic atoms can influence the local magnetic environment of hydrogen atoms, which can affect and enhance images via relaxation shifts – this is the basis for many contrast agents. For instance, iron reduces hydrogen relaxation times such that radiologists can use iron-based colloids as a contrast agent to cancel signal in areas of iron localization. High-resolution MRI can be used without contrast to generate assessments of plaque composition that are comparable to the standard reference AHA grade-based histopathological evaluations of the same plaque [(78)]. While the preceding MRI study was performed ex vivo, it demonstrates the potential for MRI to assess plaque constituents with the possibility for automation [(78, 79)]. Other studies demonstrated the ability to evaluate certain plaque constituents such as proportions of calcified and fibrous tissues, as well as lipid cores and intraplaque hemorrhage in vivo [(80)]. Novel pulse sequences have also recently been developed to increase signal-to-noise by creating positive contrast methods [(81)]. However, imaging living organisms presents a set of problems such as motion artifacts which entail heart contraction, breathing, random movements (voluntary and involuntary), and flowing blood. Because these difficulties decrease the high resolution necessary to image atherosclerosis, MRI technologists have developed a suite of techniques to manage them such as black blood, motion reduction, and cardiac gating [(82)]. Using such technology, MRI is capable of revealing plaque morphology [(76)], a significant aspect in the determination of a lesion’s risk. Thus, due to its high resolution and non-deleterious mode of image acquisition, MRI may be useful not only for prospective screening of populations, but also for monitoring therapeutic regimens or surgeries [(83)]. However, while MRI is suitable for morphological characterization and some delineation of plaque components, MRI alone can not detect the biochemical variations within a plaque that provide deleterious indicators (e.g., risk factors and biomarkers mentioned in section 1.2.1). Morphological and biochemical information combined would afford a more comprehensive, dependable diagnosis [(84)]; for reasons such as this, the field of molecular imaging has blossomed, as described later.

25 Selective uptake of certain contrast agents (e.g., ultrasmall superparamagnetic iron oxides, USPIOs) by biological entities in plaque such as thrombi and macrophages significantly enhance the potential for MR visualization modalities [(85-87)]. Because of preferential uptake, these agents contribute to MRI studies through facilitation of both the specific imaging of those structures which take up iron as well as exploiting the effect that such agents have in either amplifying or decreasing MR signal. Thus, certain tissues become more discernible to the trained (e.g., radiologist) eye, allowing differentiation between normal and diseased tissues. Many paramagnetic agents have been tested for use as contrast agents in MRI in addition to iron compounds. Paramagnetic materials that have been used include gadolinium chelates, manganese, copper, chromium complexes, and native proteins (heme-containing) [(88)]. However, toxicity and stability issues preclude in vivo use for many of these compounds. Thus gadolinium chelates and iron-based colloids have become the most commonly used approved in vivo MR contrast agents. Iron oxide USPIOs may behave superparamagnetically (i.e., the energy required to modify the magnetic moment direction is on the order of the ambient thermal energy), affecting relaxation times more than similar paramagnetic materials and causing reduced particle aggregation. Iron oxides such as ferumoxides, ferumoxtran, ferumoxytol, and SHU555C (which chiefly reflect varied surface coatings and are marketed under various trade names) and others have been used clinically in humans [(89)]. While the generation of contrast in standard imaging relies upon heterogeneity in physiological, anatomical, and metabolic features [(76)], a great deal of current work in the field entails molecular imaging approaches for iron and other targeted and activatable agents to generate contrast based on molecular pathways, targets, and cellular or subcellular processes [(90)]. A typical molecular imaging reagent includes both a reporter (e.g., radioisotope, fluorochrome, magnetic material, or acoustic modulator) and a molecule-, fragment-, or cell-specific affinity or other relevant agent such as proteins, DNA, carbohydrates, or other biological or synthetic small molecule [(61)]. These agents are generally used in imaging modalities such as nuclear (e.g., with SPECT and positron emission tomography, or PET), MRI, optical (e.g., fluorescence molecular tomography), and acoustic approaches.

26 Molecular imaging unites recent advances in biology and imaging technologies to advance methods to visualize fundamental molecular signatures of atherosclerosis (and other diseases) in vivo [(61)]. Jaffer et. al. [(61)] anticipate the clinical advantages of cardiovascular molecular imaging primarily to include 1. the ability to identify patients highly vulnerable to plaque disruption who can not be identified via routine clinical exams such as physical exam, history, electrocardiogram, lipid profile, c-reactive protein, etc. 2. characterization of plaque vulnerability in high-risk vascular regions. 3. evaluation of innovative cardiovascular therapies that direct biologically-based, rather than lipid-based, interventions. 4. enabling preferred selection of individualized treatment strategies on the basis of the vulnerable plaque’s molecular signature. While lofty, these objectives are critical to the future of clinical cardiovascular patient care, and are being pursued by various groups via research and clinical trials. While early molecular imaging (animal) trials must likely concentrate upon validation studies of the reagent through histological corroboration of agent presence with the reporter signal, subsequent trials must focus upon assigning a “probability of rupture or erosion” to a particular plaque using specific molecular imaging reagents (with associated imaging modalities) in order to predict clinical cardiovascular risk [(61)]. Parenteral administration is the most common approach for molecular imaging agents, enabling agents either to bind to vascular sites accessible via the vascular lumen, as described herein for cardiovascular endeavors, or to extravasate to tether to extravascular lesions/disease sites. As described in section 1.1, targeting may be via peptides, proteins, DNA/RNA, polymers, or other biomolecular interactions with specific ligands, for instance to an expressed or overexpressed marker characteristic of the disease site. Previous work has demonstrated the ability to deliver and image VCAM-1 with iron oxide USPIOs/MRI, and Annexin V and antisense oligonucleotides via nuclear agents, specifically to atherosclerotic plaque sites in animal models [(88, 91, 92)]. As will be described in the following section, imaging of markers indicative of a plaque’s vulnerability to rupture has become a major thrust in nuclear, optical, acoustic, and MR molecular imaging [(76, 93)]. The general strategies for molecular imaging rely upon fundamental assumptions and discoveries in molecular biology and pathophysiology. The disease must have

27 defined molecular underpinnings, preferably as part of the cause of the ailment but molecular-level effects are also potentially targetable. Molecular imaging seeks to exploit these molecular features for key objectives in early detection, monitoring therapeutic interventions, offering disease-specific information via non-invasive diagnostic modalities, and prediction/prognosis and subset stratification of disease (e.g., Her-2/neu biomarker as in section 1.3) [(76)]. In particular, the use of MRI in molecular imaging is blossoming because of the capacity to examine proton density, diffusion, perfusion, and biochemical heterogeneity for functional and structural coregistration [(76)]. Current techniques for molecular imaging of atherosclerosis exploit such capabilities by targeting protease activity, lipoprotein and macrophage presence, thrombus/fibrin, angiogenesis, myeloperoxidase, and apoptosis [(61, 76)]. Anti-fibrin, for instance, was conjugated to a paramagnetic perfluorocarbon gadolinium chelate in

order to detect thrombus presence in a canine model using standard T1 MRI sequences [(94)]. Nuclear imaging modalities such as SPECT and PET also afford opportunities for molecular imaging, utilizing potentially harmful ionizing radiation. In SPECT, a gamma- sensitive camera detects photons emitted from a radioactive agent that is injected, ingested, or inhaled to reconstruct an image via computed tomography. SPECT and PET confer advantages such as very high sensitivity and low background, utilizing agents such as Technetium-99m (99mTc) and 18F fluorodeoxyglucose (18FDG) to identify regional uptake of very low concentrations of these radioactive materials [(95)]. However, these techniques can not provide the resolution (nor structural data) of MRI, hindering the capacity for very specific localization of disease sites (an aspect that may be significant when planning appropriate interventions). It is furthermore undesirable to repeatedly inject radioactive substances into healthy populations. Nevertheless, these modalities can provide considerably more (anatomical) information via combination with CT, which has been pursued via commercialization of specialized SPECT/CT and PET/CT scanners; such combined techniques may prove important for future cardiovascular diagnostics [(74)].

28 1.2.4 The Debate: Stable Versus Unstable Plaque The description of the vulnerable, or unstable, plaque remains hypothetical throughout the literature, gradually evolving toward a general consensus. As terminology fluctuates between “vulnerable plaque” and “vulnerable patient” and clinical and pathological definitions of the lesions coalesce, the complexity of the descriptions may reflect the lack of a single, large-scale study for definitive classification. Thus a need for standardization in nomenclature is recognized and has been suggested [(96)], though not uniformly adopted. A summary of lesion terminology appears in Table 1.1 [(66)]. In general, the vulnerability of a plaque denotes its potential to elicit thrombosis (via either plaque rupture or erosion) and the ensuing unstable angina or myocardial infarction, which may be fatal (though most plaque disruptions are clinically silent [(97)]). Patient vulnerability, on the other hand, has been defined in one large study [(98)]. According to Naghavi et. al., all atherosclerotic plaques likely to progress swiftly and thrombose are considered vulnerable, not only rupture-prone plaques. The vulnerable patient is considered to include these plaques in addition to individuals presenting with vulnerable blood (prone to thrombosis) and vulnerable myocardium (prone to lethal arrhythmias) [(98)]. This is a new paradigm from the clinical perspective, but the vulnerable plaque remains the focus for this dissertation.

29 A lesion in a coronary artery considered, on the basis of angiographic, autopsy, or other findings, to be responsible for the Culprit lesion clinical event. In unstable angina, myocardial infarction, and sudden coronary death, the culprit lesion is often a plaque complicated by thrombosis extending into the lumen.

A plaque with loss and/or dysfunction of the lumenal endothelial cells leading to thrombosis. There is usually no additional defect Eroded plaque or gap in the plaque, which is often rich in smooth muscle cells and proteoglycans. Considered a vulnerable lesion.

High-risk, vulnerable, These terms can be used as synonyms to describe a plaque that is and thrombosis-prone at increased risk of thrombosis and rapid stenosis progression. plaque Includes erosion-prone plaques.

An inflamed plaque with a thin cap covering a lipid-rich, necrotic Inflamed thin-cap core. An inflamed thin-cap fibroatheroma is suspected to be a fibroatheroma high-risk/vulnerable plaque.

A heavily calcified plaque with the loss and/or dysfunction of endothelial cells over a calcified nodule, resulting in loss of Plaque with a calcified fibrous cap, that makes the plaque at high-risk/vulnerable. This is nodule the least common of the three types of suspected high- risk/vulnerable plaques.

A plaque with deep injury with a real defect or gap in the fibrous cap that had separated its lipid-rich atheromatous core from the Ruptured plaque flowing blood, thereby exposing the thrombogenic core of the plaque. This is the most common cause of thrombosis.

A plaque with an overlying thrombus extending into the lumen of Thrombosed plaque the vessel. The thrombus may be occlusive or non-occlusive.

A patient at high risk (vulnerable, prone) for experiencing a Vulnerable patient cardiovascular ischemic event due to a high atherosclerotic burden, high-risk vulnerable plaques, and/or thrombogenic blood.

Table 1.1: Definitions for terminology generally used in atherosclerotic diseases. Adapted from Fuster et. al. [(66)].

It is important to distinguish between plaque rupture (generally occurring at less stenotic sites) and erosion (more prevalent at obstructive sites), because though rupture

30 remains the most frequent route to thrombosis, as much as 30%-40% of coronary thrombosis occurs in areas where rupture can not be identified [(66, 97)]. One study, for instance, demonstrated that 22 of 50 cases of sudden cardiac death were attributable to superficially eroded plaques in which no rupture of the fibrous cap was found [(99)]. It is thus unfortunate that “plaque erosion” has not yet entered the AHA-grade based vernacular. However, for present purposes, rupture- and erosion-prone lesions are considered vulnerable. Notably, the properties of a vulnerable plaque differ on the basis of patient age – for example, younger patients have lesions akin to that as described, while patients in their 70s and 80s have calcified lesions with diminished lipid content, few inflammatory cells, and generally trigger clinical events via lesion erosion [(93)]. Some vulnerable plaque (occasionally termed “thin cap fibroatheroma” [(100)]) characteristics comprise lipid rich cores (which contain at least 50 % of the plaque volume and are difficult to identify even for angiography [(93, 101)]), mild to moderate stenosis (indeed, the common definition designates a vulnerable plaque as retaining a fibrous cap less than 65 µm), many infiltrated macrophages (greater than 25 in a 0.3 mm diameter field of view) which are often foam cells, few smooth muscle cells, inflammation, scattered calcifications, angiogenic endothelial dysfunction, apoptosis, and by definition are thrombogenic [(66, 77, 101-103)]. Furthermore, recent evidence implicates leaky intraplaque vasa vasorum (the network of smaller blood vessels which distribute blood in the wall of larger vessels such as the aorta) in intraplaque hemorrhage, initiating erythrocyte accumulation that triggers plaque rupture [(104)]. That vulnerable plaques tend to exhibit low levels of stenosis is accentuated by the fact that culprit lesions (i.e., those resulting in acute myocardial infarction) less than 60% angiographically obstructive account for 85% of infarctions [(93)]. Stable plaques, conversely, tend to be highly obstructive, with thick fibrous caps and small lipid cores, reduced apoptosis levels, and are considered athrombogenic. Stated differently, a critical point in cardiovascular medicine is that the clinical outcome of a lesion is predicted better by plaque composition (i.e., vulnerable constituents) than by lumenal occlusion [(105)]; thus, major questions in cardiovascular medicine involve why plaques become “complicated,” or vulnerable. Macrophage apoptosis has been implicated in both plaque erosion and rupture [(77)] and thus with

31 vulnerability, though mechanisms and consequences remain hypothetical [(106, 107)]. While at least minimal apoptosis has been detailed throughout atherogenesis, it is fairly common in advanced stages in which a lipid core is present [(107)], and is more prevalent in macrophages than smooth muscle cells (SMCs) [(108)]. This process is depicted in Figure 1.1. Basically, early in atherogenesis monocyte chemoattractant protein-1 (MCP- 1) and other factors induce monocytes to migrate into atherosclerotic lesions and differentiate into macrophages with the aid of molecules such as macrophage colony- stimulating factor (M-CSF), which is also critical to macrophage proliferation [(109)]. Macrophages subsequently endocytose oxidized LDL (oxLDL, which also serves as a monocyte chemoattracant) via scavenger receptor, leading to cytoplasmic accumulation of cholesterol esters to form foam cells. While some foam cells migrate into peripheral blood, most die of apoptosis within the lesion due to the toxic levels of collected intracellular cholesterol by a variety of apoptotic mechanisms (e.g., Fas/Fas ligand, p53, cytokines, or nitric oxide) [(109)]. In human lesions apoptotic macrophages (which are the most common type of apoptotic cells in unstable atheromas) are located within the core and periphery of the lesion (as opposed to apoptotic SMC-derived foam cells, which are localized to fibrotic regions) [(110)]. Plaques rupture in the shoulder regions of macrophage and macrophage foam cell-rich areas (particularly apoptotic cells) due to the release of extracellular matrix (ECM)-degrading proteolytic enzymes such as certain matrix metalloproteinases (MMPs, including MMP-1, MMP-2, MMP-3, MMP-9, MMP- 12, and MMP-13) [(109)]. Because these enzymes significantly weaken the ECM of the fibrous cap, destroying the mechanical integrity, their degradative activity causes eventual plaque rupture, exposing the thrombogenic core [(109)]. Tissue factor, thrombin, and circulating platelets subsequently cause a thrombus to form [(63)]. These cascades are summed up in Figures 1.3 and 1.4. In addition to a multitude of molecular changes, apoptotic foamy macrophages externalize phosphatidylserine to their outer cytosolic leaflet, a typical cellular response to apoptosis. Interestingly, because of the robust association between macrophages and the molecular and cellular development of atherosclerosis, there is a thrust toward administering therapeutic interventions via macrophages and their interactions [(111)]. This feature might lend itself to future nanoscale therapeutic platform devices capable of both imaging and directing

32 interventions to macrophages (e.g., assessment of macrophage activity might drive personalized anti-macrophage molecular therapies [(61)] on the same device). As described in the previous sections, detection of atherosclerotic lesions has become a major clinical objective. However, plaques in humans are prevalent and patients survive for long periods with obstructive, stable lesions such as those causing chronic stable angina – thus the necessity for early and prompt localization of a vulnerable lesion, which might rupture at any time, therefore requiring urgent attention and treatment [(93)]. Given the molecular differentiation between vulnerable and stable plaques (described further in the following section), molecular imaging appears to be a natural fit toward objectives in early identification of lesions in imminent danger of rupture [(93)]. Because atherogenesis is an inflammatory process, and vulnerable lesions contain increased levels of inflammatory cells, agents that bind or localize to areas of inflammation can be valuable. 18FDG is a compound which preferentially localizes to areas of inflammation by targeting regions of abnormal glucose metabolism and can be visualized using PET. However, because 18FDG may not be sufficiently specific to unstable plaques, instead concentrating over a broad range of lesions [(93)], more precise molecular targeting reagents are often used. These include radiolabeled Annexin V [(112, 113)] (as large quantities of apoptotic cells are reported to be within vulnerable lesions [(93)]), antisense-labeled 99mTc, as well as conjugates (reporter plus targeting reagent) that are directed to specific molecular targets such as cell surface receptor proteins, metabolic molecules, proteases, peroxidases, lipids/lipoproteins/phospholipids, unique angiogenic molecules and endothelium, and cell trafficking entities such as monocytes, lymphocytes, and stem cells [(61)]. Thus, clearly specific molecular markers afford a foundation upon which to assemble vulnerable-lesion targeted imaging agents [(61)]. However, such agents are only as good as the biomarker upon which they are based, placing special importance on identification of specific and sensitive disease markers. Biomarkers are discussed in further detail in section 1.2.4.1.

1.2.4.1 Vulnerable Plaque Biomarkers As mentioned in the previous sections, MCP-1 and VCAM-1 are expressed during atherogenesis, and may be overexpressed in vulnerable plaques [(114)]. However, the

33 disease pathogenesis is much more intricate than discussed here, so though it remains poorly understood, a broad array of biomarkers specific to various phases and cells of atherogenesis have been identified. The search for vulnerable biomarkers is partially motivated by the evolution of molecular imaging and therapeutic targeting for the early detection and treatment of risky plaques, as described above. Therefore, it is crucial to understand a particular biomarker’s kinetics (when and how it occurs within the disease state), concentration (e.g., within the plaque as compared to elsewhere and is it expressed/overexpressed sufficiently to make it a viable target), distribution (i.e., where in the organism and is it accessible to targeting agents), time course, and localization, as these parameters may establish the utility of a marker of interest. Not all atherosclerosis biomarkers are practical targets. Secreted proteins, stationary receptors with low abundance [(61)], circulating molecules, completely intracellular molecules, and proteins common in multiple regions or in the blood all represent potentially exigent molecular targeting options. While biomarkers such as circulating inflammatory molecules may not be conducive to molecular imaging due to constant movement, for example, they may help to evaluate cardiovascular risk by other prognostic methods [(115)]. In contrast, cell surface and internalizing receptors, very specific, plentiful extracellular molecules, and low background enzyme-sensing quenched substrates may embody ideal molecules for targeting approaches [(61)]. While cardiovascular risk factors and humoral vulnerability markers have been established in relative consensus, the search for highly specific and sensitive vulnerable plaque protein profiles has been arduous. Recent efforts have advocated systematic proteomic analyses to establish potential cardiovascular biomarkers [(116)]. This study has identified proteins such as HSP-27 and Cathepsin D as potential novel biomarkers for cardiovascular risk [(116)]. However, potential biomarkers may also be uncovered by scouring the atherosclerosis basic research literature for proteins and other biomolecules associated with plaque and plaque rupture/erosion, as described below. While the molecules chosen for the atherosclerosis targeting studies described in this dissertation will be detailed in Chapter 2 (e.g., Annexin V, VCAM-1, etc.), the range of potentially targetable molecules extends far beyond this subset. As MMPs, for instance, degrade fibrous caps, collagen, and fibrin, fragments may commonly be

34 associated with the lesion. MMP activity as previously described is thus associated with weakened fibrous caps and plaque vulnerability [(117, 118)]. Therefore, targeting MMPs (e.g., using optical reporter structures that are activated via MMP enzymatic activity) or the proteolytic fragments they create may be viable opportunities (see [(94, 119)] and section 2.4). Cathepsins comprise a family of cysteine proteases that, like MMPs, are overexpressed in atherosclerotic plaque and degrade extracellular matrix. They may be targeted in the same manner as MMPs [(120)]. Myeloperoxidases (MPOs), leukocyte- derived heme peroxidases involved in inflammation, have also been biologically and clinically correlated with atherosclerosis [(121)]. While still early in its investigation, groups have begun to target promising agents to MPOs at sites of disease [(122)]. Other

targets include endothelial markers VCAM-1 and αvβ3, the extra-domain B of fibronection, and molecules indicating cellular apoptosis, which are indicative of vulnerability [(60, 61)]. The neovascularization of plaques represents a novel target that may practically be realized via angiogenic markers such as the specific expression of

integrin αvβ3 on neovascular endothelial cells localized to atherosclerotic lesions. Many targeting structures have been constructed on the basis of this molecule with imaging modalities ranging from MRI and ultrasound to optical and nuclear (SPECT, PET) [(61, 123)]. Cellular surface receptors embody a large array of proteins that relate to many as yet undefined atherosclerotic processes. SRA, cd-36, dextran, and many other receptors on macrophages, smooth muscle, endothelial, and other cells may be markers of atherosclerotic instability [(61)]. CCR-2, IG9, ICAMs (intercellular adhesion molecules), cd-25, M-CSF receptor, LOX-1, cd-22, and selectins are other surface proteins whose overexpression may be associated with vulnerable lesions [(124-132)]. Many of these molecules are overexpressed in activated endothelium for purposes of adhesion of cells from the circulation – during the recruitment of inflammatory cells, rolling and adhesion of circulating cells such as leukocytes (e.g., monocytes, T-cells) along the plaque endothelial surface have been demonstrated to be a critical aspect of atherogenesis [(131)], as these cells migrate into plaques to form macrophages/foam cells and other structures and release molecules that sustain disease pathogenesis (as seen in Figure 1.2).

35 Depending on the degree, concentration, and type of their molecular overexpression, therefore, these molecules may represent excellent targets. Metabolic (Hexokinase, GLUT-1), lipoproteins (OxLDL), and cell trafficking molecules (e.g., those expressed on monocytes, lymphocytes, and stem cells) are also considered atherosclerotic biomarkers [(61)]. Many other potential biomarkers relating to inflammation and other facets of atherogenesis exist in the literature, but are beyond the scope of this dissertation to describe. The range of potential markers notwithstanding, it is perhaps most likely that multiple markers, i.e., a convergence of biomarker evidence, will be required to make a rational vulnerable plaque diagnosis.

1.2.5 Animal Models of Atherosclerosis Molecular imaging diagnostics and the therapeutic interventions they engender are critically dependent upon the biological marker they recognize. While it is feasible to formulate hypotheses based on human observations, in order to discover and study potential biomarkers in pertinent systems, animal models are absolutely vital. They facilitate not only biomarker identification, but also controlled mechanistic/kinetic testing as well as prospective targeting and therapeutic reagents. Neither cell culture nor theoretical mathematical models are currently sufficiently sophisticated to model the complex organism-level interactions to achieve these functions. Some animal models correspond fairly well to the previously held perspective of atherosclerosis, i.e., stable plaques – high atherosclerotic burden/stenosis, lesion size, cell number, etc. However, it is unfortunate that presently there exists no single, ideal animal model of vulnerable plaque [(133)]. Nevertheless, a plethora of genetic, cholesterol-fed, and injury models of atherosclerosis and vulnerability in a variety of species are currently employed to study the molecular mechanisms, diagnostics, and therapeutic response related to atherosclerosis. An ideal animal model would closely mimic human disease – i.e., the long, intricate pathophysiologic process culminating in thrombotic occlusion – as defined by Cullen et. al. [(134)]: 1. The atherosclerotic process in the animal model should histologically match that in humans. 2. The atherosclerotic plaque in the animal model should show the same vulnerability to rupture, erosion, and thrombotic events as its human counterpart.

36 3. The events leading to vulnerability and rupture such as cap/core ratio, cellular composition, and collagen production and breakdown should be analogous to those in humans. 4. Plaque rupture in the animal model should occur genetically and/or via diet without the need for mechanical intercession such as cuffing of the artery or insertion of balloon catheters. 5. At least in some cases, plaque rupture should be accompanied by the formation of platelet-rich fibrin thrombi. 6. Therapeutic treatment response in the animal model should have the potential for replication in humans. 7. The animal species should be available for research, easy and economical to maintain, and plaque rupture should reproducibly occur within a reasonable time frame. 8. The model should be capable of accommodating genetic manipulations.

Not only are current animal models incapable of meeting these conditions, but the vast majority of them does not come close. Animals such as monkeys, baboons, pigs, guinea pigs, rabbits, swine, and pigeons develop atherosclerosis spontaneously or after feeding with high-cholesterol diets [(109)]. However, the most frequently used models include cholesterol fed rabbits, Watanabe rabbits (Watanabe Heritable Hyperlipidemic rabbit, or WHHL), apoliprotein E (apoE)-deficient (a cholesterol efflux mediator) mice, and LDL-deficient mice as they most closely imitate human disease [(109)]. While the most completely characterized animal model is the WHHL, originally developed in the late 1970s from a genetically hypercholesterolemic mutant [(135, 136)], the most flexible and commonly used animal models are genetic mouse models of atherosclerosis which were initially generated in 1992 [(137)]. One approach to animal models matches a single risk factor to a particular genetic knockout model. For instance, the presence of homocysteine, fibrinogen, lipoprotein (a), and C-reactive protein are modeled by cystathionine β-synthase, transgenic fibrinogen and lipoprotein (a) mice (only the latter being prone to atherosclerosis), and transgenic C- reactive protein mice, respectively [(68)]. These can be excellent approaches for studying the contribution of a single factor to atherogenesis, but fail to generate a complete atherosclerosis model as described above. While it is feasible to induce plaque rupture in animal models, most techniques disregard the complexity of relevant clinical triggers, instead employing direct mechanical disturbance of lesions [(133)]. One basic approach is to squeeze the aorta of ApoE-/- mice with forceps to elicit a platelet and thrombin-rich thrombus [(138)]; an

37 alternative is to ligate and cuff the carotid arteries [(139)]. Other techniques involve producing balloon injuries in hypercholesterolemic rabbits and mechanical disruption via inflation of the angioplasty balloon [(140)], and employing a photochemical reaction or to inject Russell’s viper venom followed by histamine to incite thrombus development in ApoE-/- mice and cholesterol fed rabbits, respectively [(141, 142)]. Another method employs double knockout ApoE-/- and LDL-/- mice combined with mental stress or hypoxia-incited myocardial infarction, which to some extent approaches clinical relevance [(143)] – however, in this study as in others, there is generally no correlation between atherosclerosis and myocardial infarctions (events appear most closely to resemble thrombi formed in response to plaque erosion), thereby decreasing their pertinence to human disease [(133)]. Other models have had varying levels of success in reproducing features of human disease, including inherited lesions in the Carneau pigeon, cholesterol fed models of chimpanzees, squirrel monkeys, howler monkeys, cynomolgus monkeys, and rhesus monkeys, significantly modified diets in dogs and rats, as well as the white Belgian pig, which displays sudden coronary death when under stress, a striking parallel to humans [(134)]. However, these models are difficult to accommodate, expensive, some species are on protected lists, and perhaps most significantly, none represents an adequate overall complement to humans. In fact, very few animal models display spontaneous lesion rupture and/or thrombus, and until somewhat recently, such events were considered exceedingly rare to nonexistent [(133)]. In the early 1990s, a porcine LDL hypercholesterolemia model was first shown to develop spontaneous hemorrhage and rupture in the coronary arteries in older 39-54 month old animals [(144)]. Mouse models of atherosclerosis (ApoE-/-, and LDL knockouts) were considered resistant to vulnerable plaque development until Rosenfeld et. al. and Johnson et.al. described apoE -/- models which displayed evidence of innominate artery intraplaque hemorrhage and thrombus-induced death in the brachiocephalic artery, respectively [(145, 146)]. A transgenic, salt-sensitive hypertensive and hyperlipidemic rat model produced a number of features comparable to human disease, including differentiable “culprit” and “stable” plaques, but neither plaque erosion nor rupture was correlated with the thrombi produced [(147, 148)].

38 Because of ease and expense in handling, availability, the broad physiological characterization available, a great deal of accessible genetic data and established procedures, and currently the most potential for comparable vulnerability of lesions, mice are the research model of choice [(134)]. However, there are also many limitations inherent in the murine model: their lipid physiology is fundamentally different from humans, they do not undergo atherogenesis without genetic modification, and mice are exceedingly small, which causes a plethora of obstacles [(134)]. At about 25 grams (approximately 3000 fold smaller than humans, but with cells approximating those of human cell dimensions), mice develop histologically disparate lesions from humans (often with medial layers only a few cell layers thick) and disruption of the plaques may not emulate that of human disease [(149)]. Indeed, accounts vary in the literature as to whether or not the ApoE-/- mouse definitively displays plaque rupture, and even if it does occur, the evidence is indirect and reports diverge in their descriptions of event frequency [(149-151)]; further, verification of thrombus formation and death due to rupture are lacking [(134)]. These discrepancies may also be partially due to sized-based obstacles, such as the difficulty in identifying/distinguishing plaque rupture and erosion in diminutive vessels. Furthermore, in imaging and diagnostics studies, technological improvements are sometimes necessary to enable viable imaging of the minute murine vasculature. For instance, in the realm of MRI, humans are generally imaged in a 1.5 T magnet. For mice, research magnets in the 2-11 T range become essential (with and without additional external coils), and because the sequences used for these systems are often different, direct comparisons with clinically relevant magnets may not be straightforward. In combination, these points serve to indicate that no animal model has yet been identified that can reliably link plaque rupture to thrombus formation and death as is the case in humans. A description of the WHHL and WHHLMI (Watanabe Heritable Hyperlipidemic Myocardial Infarction) rabbits follows, which are rabbits genetically predisposed toward the formation of atherosclerosis. These models fall into the same category as the aforementioned in that they are not ideal, as they can not link death with thrombus/plaque rupture. They will be described below to the extent of their relevance in research usage in this dissertation.

39 The rabbit is a more convenient model of atherosclerosis than mice in some ways because of their larger size (e.g., for histological and imaging methodologies, as well as for better direct comparison to human vessel cell layers in terms of size), but offer less capacity for genetic modifications and their general physiology is less well-characterized. The original WHHL strain of rabbit, again initiated by breeding a heritably hyperlipidemic mutant rabbit, displayed extensive plaque burden in the aorta, and was selectively bred to develop coronary atherosclerosis [(152, 153)]. A multitude of publications describe the WHHL morphologically, biochemically, and the response to diagnostics and therapeutics. Aortic atherosclerosis, the present focus, is grossly observed in WHHLs from 2 months of age with a normal diet. Aortic lesions advance from 40% at 6 months to greater than 70% aortic surface coverage at 12 months to nearly the entire length of aorta in rabbits aged over 18 months. Compositionally, early lesions contain many macrophages and few SMCs in the intima and media. While year old WHHLs exhibit macrophage-derived foam cells and a fibromuscular cap, interestingly (similar to humans) older WHHLs (above 18 months) display decreased cellular components, but increased collagen, extracullular lipids, cholesterol clefts, and sometimes calcium [(154, 155)]. Differentiable levels of cellular accumulation between coronary and aortic atherosclerosis in WHHLs, as well as differing severity of plaque burden, suggest varying mechanisms between the two regions of vasculature [(155)]. Incidentally, cerebral, pulmonary, carotid, mesenteric, celiac, renal, and other arterial (though no small arteries) atherosclerotic plaques have also been demonstrated in WHHLs. The WHHLMI rabbit, introduced around 2000, begins to die of apparent spontaneous myocardial infarction from the age of 11 months on a normal diet. As might be expected, aortic (and coronary) atherosclerosis progressed rapidly (with no gender differences) in comparison with the WHHL rabbit [(156)]. In parallel with human disease, WHHLMIs display comparable morphology (and histology shows WHHLMI plaques have many foamy macrophages and relatively few SMCs in addition to fragile thin fibrous caps) and many exhibit old myocardial infarction accompanied by fresh myocardial lesions [(156)]. Furthermore, the progression of atherosclerosis was “probably associated with the spontaneous development of myocardial infarction” in

40 WHHLMIs and ECGs (electrocardiograms) just prior to death displayed patterns characteristic of acute myocardial infarction in humans [(156, 157)]. However, though histological analysis reveals that 97% of WHHLMIs which died up to 35 months exhibited signs of myocardial infarction, it is important to understand that histology could not demonstrate actual culprit plaque rupture or thrombus. Thus, though many signs of vulnerable plaque and myocardial infarction exist in WHHLMIs (even at the cellular level [(158)]), it is clear that other factors contributed to the deaths of the WHHLMIs. This indicates that WHHLMI plaques too are not analogous to human disease. In comparison with mouse models, WHHLMI rabbits have advantages and disadvantages. Advantages include size and similarities in plaque morphology, lipid metabolism, and relevant enzyme activities (e.g., cholesterol ester transfer protein, hepatic triglyceride lipase, etc.) to humans. Disadvantages include that myocardial infarction and thrombus can be located in mice, though not necessarily correlated with the culprit plaque, and the comparative genetic flexibility and lower expense of mice. Rabbit models were chosen because of their gross morphological and biochemical similarity to human disease and ease of (molecular) imaging with MRI systems.

1.3 Breast Cancer With the exception of skin cancers, breast cancer afflicts more women in the United States than any cancer. The American Cancer Society estimates that in the US 212,920 women will develop invasive carcinoma, 61,980 women will develop in situ carcinoma, and 40,970 women will die of breast cancer in 2006 alone [(159)]. Even more troubling, since screening modalities such as mammography have been implemented, breast cancer incidence rates have soared [(160)]. This has caused concerns about overdiagnosis, i.e., that the screening detects cancers which are clinically insignificant. The causes of breast cancer are reported to be multifocal and numerous, and risk factors increasing the likelihood of breast cancer diagnosis include increasing age, genetics, early first menstruation, late age at first birth, a history of breast biopsies, menopausal hormone use, obesity, and more [(161)]. While risks also vary by ethnicity, culture, geography, and socioeconomic status, the cumulative probability of a woman

41 contracting breast cancer during a 100 year life is approximately 1 in 7. Thus, since breast cancer is a disease against which all women must be vigilant, it is important that improved screening technologies be developed to support this responsibility. In general, screening exams such as mammography lead to the acquisition of tumor tissue if the test indicates the possible presence of malignancy. Samples are generally acquired through surgical biopsy, fine-needle aspirate, or core needle biopsy, subsequently fixed, and inspected by a qualified pathologist. The pathologist confirms whether the tissue is cancerous, and if so, the specimen is further evaluated. These tumors may vary by cellular classification, by stage, and by molecular status. In terms of their cellular constitution, the most common breast tumors include carcinoma NOS (not otherwise specified), ductal (in situ, invasive, inflammatory, and further subclassifications), lobular (in situ or invasive), and nipple (generally Paget’s disease, which may be intraductal, invasive, or NOS) [(162)]. Tumors are staged based upon a combination of the size and cellular classifications of the primary tumor and the status of metastasis. Stages range from 0 through IV, inclusive of several sub-categorizations that may be reviewed in greater detail [(163)]. The molecular classification of tumors often helps establish their aggressiveness and the molecular status generally has a major effect on patient morbidity and mortality. Some biomarkers which are commonly tested include estrogen receptor, progesterone receptor, and Her-2/neu, which is discussed in detail in section 1.3.3. Breast cancer therapies include surgical removal, radiation therapy, chemotherapy, and molecular targeting (e.g., monoclonal antibodies that affect tumor growth kinetics). Often the most efficacious treatments involve combinations of these therapies (e.g., radiation therapy following surgical removal to ensure the elimination of all cancer cells).

1.3.1 Clinical Background It remains necessary to perform surgical biopsies to obtain tissue for morphologic and histologic examination and diagnosis through molecular assays. Not only do biopsy sections lead to clinical diagnoses, but more importantly, they can indicate appropriate routes of treatment and expectation of morbidity/mortality. Therefore great significance

42 is attributed to biopsies and those who interpret them (i.e., pathologists). Although the scoring systems are often sophisticated, it is nevertheless surprising that interpretations of tissue parameters can vary greatly between pathologists. Indeed, pathologists come to complete agreement as low as 9% of the time [(164-168)]. A qualitative skill dependent on the experience of the practitioner, tissue grading is sufficiently clinically significant that normalization among all specimens should be essential because of the risk that a slightly incorrect qualitative evaluation could lead to lethal consequences as therapeutic regimens are selected on the basis of such assessments. Therefore, impartial quantitative histological evaluation procedures may be desirable prior to samples reaching a pathologist. This is the basis of the work presented in Chapter 3.

1.3.2 Modes of Detection, Evaluation, and Intervention Other than breast self-examination, the most common breast cancer screening modality is mammography, a method that employs high energy x-ray photons which enable detection of cancers that are impalpable due to small size and non-invasive malignancies such as ductal carcinoma in situ (DCIS, which is only detectable by mammography). Mammograms compare the density of the breast to its environment. If abnormal white regions are observed on x-ray film, further examinations are ordered. Therefore, the higher the natural density of breast, the more difficult it can be for a radiologist to differentiate between breast and tumor. Nevertheless, while mammography can lead to early detection and treatment of breast tumors, it is not the only screening method available. While ultrasound has been used for breast cancer detection, it has been shown that “very few lesions are detected on ultrasonagraphy that are not detected on mammography [(169)].” However, though ultrasound has increased false positive and false negative rates [(169)], it has advantages in that it does not utilize ionizing radiation, and there are no known harmful effects of its energy. Ultrasound is also often able to detect the difference between benign and malignant masses [(170)]. Therefore, ultrasound is useful only as a companion screening method to mammography, but is not practical as the sole modality to screen for breast cancer. Medical and other ultrasound techniques will be discussed in further detail in section 1.3.4.

43 While the use of MRI for breast cancer screening is relatively un-established, studies have shown that, at minimum, under certain circumstances MRI possesses better sensitivity than mammography, ultrasound, or clinical breast examination (i.e., an exam performed by a qualified health professional) [(171, 172)]. However, despite exhibiting a high incidence of false positives [(173)], MRI’s sensitivity and attributes (such as non- invasive, non-ionizing, and high resolution images as discussed in section 1.2.3) generate hope that it may also produce improved results in the screening of populations. Other methods are currently being developed – for instance, radionuclide studies have been performed for prospective detection of breast cancer. These tests, termed scintimammography, are promising but suffer from having only undergone small trials; thus, due to a lack of clinical data scintimammography is currently useful as a complement to other established screening techniques [(174)]. Positron emission tomography, optical tomography, and thermography have also been adapted for use in breast cancer screening [(170)]. Once one or more of the aforementioned screening modalities has indicated the probability of cancer presence, biopsies are performed to obtain tissue samples for more sensitive and specific morphologic, histologic, and molecular evaluations by pathologists. To aid in these assessments, fluorescence in situ hybridization (FISH), immunohistochemistry (IHC), and other techniques (as detailed in section 1.3.3) can be performed to evaluate the amplification of the Her-2/neu gene or overexpression of Her- 2/neu protein, respectively. Therapeutic regimens are prompted by pathologists’ judgments on the basis of many elements such as the tumor’s specific stage/type, size, metastatic status, and molecular grade. While surgeries such as lumpectomies and mastectomies can often eliminate the macroscopic tumor, malignant cells left behind can form new tumors or lead to metastasis. Hormone replacement therapy, radiation, and chemotherapy are common both alone and in conjunction with surgery. However, side effects associated with these regimens have led to targeted molecular therapies that direct interventions specifically to the cancer cells, which are differentiated from normal cells on the molecular level; these therapies have been viewed as increasingly useful as detailed in the following section.

44

1.3.3 Her-2/Neu Biomarker and Molecular Analysis The Her-2/neu protein is the sole variant of the four human epidermal growth factor receptors (EGFR, or Her) that does not possess a known ligand. This receptor is a 185 kDa tyrosine kinase transmembrane molecule that is translated from a gene by the same name located in chromosome 17q. Signal trandsduction proteins that facilitate intercellular and cell-stroma interaction, the Her molecule family requires activation for functionality through phosphorylation of tyrosine residues – for activation to occur, a ligand must bind to a receptor and subsequently dimerize (homo- or hetero-) with another receptor in the family [(175)]. The Her-2/neu protein forms the most stable dimer of the EGFR family and is therefore the preferred dimer partner; further, dimers containing Her-2/neu yield the most effective signaling pairs [(176, 177)]. Because of these properties, and the fact that Her-2/neu overexpression (whether gene or protein) is observed in up to 34% of breast cancers, it has recently been studied heavily both from molecular biology and clinical standpoints. Her-2/neu is a predictive factor for breast cancer (i.e., predicts patient response to specific therapies such as Herceptin and , though it is expressed in low levels even in normal cells) and strongly correlates with poor prognosis (in fact, Her-2 amplification is more predictive for clinical outcome than any other prognostic factor except for the number of positive lymph nodes) [(178)]. Because of its established association with increased tumor aggression and patient morbidity and mortality, numerous methods have evolved to gauge Her-2/neu presence from the perspectives of both gene amplification and surface receptor overexpression. A recent analysis showed that 90% of studies (73 studies (representing 25,166 breast cancer patients) out of 81 total studies (27,161 total patients)) demonstrated that Her-2 gene amplification or protein overexpression predicted breast cancer outcome [(179)]. Her-2/neu status also predicts patient response to hormone therapy, chemotherapy, and drugs targeted to the Her-2 receptor itself [(178, 180)]. Consequently, Her-2/neu assessment techniques are useful clinically to help establish therapeutic regimens. Therefore, great significance is placed on the evaluations’ sensitivity and specificity profiles, as well as their general effectiveness (including cost, ease of processing, ability for quantitative measure, etc.). Two major subsets of tests are

45 utilized for Her-2/neu assessment, each with a distinct set of advantages and disadvantages: protein-based (while low cost, it suffers from many procedural difficulties) and gene-based (increased cost, but more objective than IHC) [(181)]. A protein-based method, immunohistochemistry (IHC) is the most commonly utilized method for the semi-quantitative evaluation of Her-2/neu. Though other protein methods such as ELISA have been employed [(182)], they are uncommon. IHC is often the primary system in the clinical arsenal for Her-2/neu evaluation such that only ambiguously scored IHC samples are subsequently tested using alternative means [(183)]. Indeed, IHC is remarkably accurate (as measured by correlation between gene copy and protein expression levels) when performed correctly [(181, 183)]. Unfortunately, IHC is plagued by sensitivity to processing (e.g., fixation, staining, choice of antibody, operator handling) and storage conditions [(184, 185)], which contribute to its inadequacy when practiced by organizations such as some community hospitals [(186)]. Another major concern with IHC is its lack of adequate standardization, in particular with respect to the subjectivity with which pathologists may apply the prevailing scoring system to score specimens [(185)]. IHC has been standardized using cell lines [(179, 187)]. Specifically, according to Ross et. al. [(179)], an IHC score of “0” is indicative of cells expressing less than 20,000 Her-2/neu receptors and show no staining; “1+” corresponds to cells expressing about 100,000 proteins showing partial membrane staining and less than 10% of cells exhibit complete staining of their membranes; “2+” signifies cells having about 500,000 receptors and displaying light to moderate membrane staining in greater than 10% of cells; “3+” is representative of cells that present about 2,300,000 receptors and show complete, strong staining of greater than 10% of cells. Recent advancements in image analysis have the potential to improve quantitative standardization [(187)], but interpathologist interpretations of the standardization protocol may well remain problematic. Perhaps due to this and processing issues, false negatives are commonly associated with IHC [(183)] (as are false positives [(178)]), which may have detrimental effects on the treatment of patients who would be better served incorporating a Her-2/neu therapeutic into their regimen. According to studies, IHC testing is often reliable when

46 3+ samples are obtained, but when 2+ and 1+ specimens are detected, secondary tests such as FISH are essential for corroboration [(178)]. The remainder of Her-2/neu tests comprise the gene-based subset. While FISH is by far the most commonly used of the DNA-based analyses, many others have been used. These include the first genetic Her-2/neu tests used, southern and slot blotting, as well as western blot, polymerase chain reaction (PCR), and a modified FISH assay termed “CISH” (chromogenic in situ hybridization). While FISH is often considered the best as it is the most well-characterized of the gene-based assessments, CISH is perhaps the most promising, potentially offering the benefits of both FISH and IHC. It could therefore establish both protein expression and gene copy number within the same tissue section, but is not yet sufficiently studied to have obtained FDA approval; only IHC and FISH have received such approval [(178)]. FISH is a morphology-based assay that utilizes an objective scoring system and includes inherent internal controls, but costs more than IHC and does not enable the ability to store slides for review as IHC does [(179)]. Many studies have directly compared FISH and IHC and generally conclude that FISH is the best option in terms of its ability to link Her-2/neu status to a patient’s response to Her-2/neu-targeted treatments [(178, 179, 183, 188, 189)], likely due to issues such as standardization and sensitivity to processing and storage conditions. However, earlier studies included both 2+ and 3+ specimens in the IHC analyses which performed worse than FISH [(188)], whereas predictive capacity for treatment response was equivalent when 3+ IHC scored specimens were compared with FISH positive samples. Indeed, the response rate for 2+ IHC specimens was 0% (35% for 3+) [(190)]. Therefore, the debate as to which is best and most practical persists [(178, 183)]. Much of the controversy surrounding the choice of Her-2/neu detection modality centers on the ability to quantify Her-2/neu, to predict probability of patient survival, and to predict the ability of Herceptin to treat patients’ cancers. Herceptin is the trade name of trastuzumab, a recombinant humanized mouse monoclonal antibody that selectively binds to Her-2/neu and specifically inhibits proliferation of malignant human cells that overexpress Her-2/neu [(191)]. Its clinical efficacy is poorly understood [(192, 193)], but is due to its down-regulation of Her-2/neu protein, which inhibits essential signaling

47 pathways. Herceptin also blocks the S-phase in the cell cycle, activates immune effector cells to direct antibody dependent cell-mediated cytotoxicity, inhibits angiogenesis by reducing a key growth factor, and inhibits the receptor’s cleavage [(193, 194)]. Herceptin is the front line therapy for Her-2 positive and hormone (estrogen and progesterone) receptor negative tumors, and is continuously used throughout the disease stages in combination with other treatments when feasible [(193)]. Because Her-2/neu is required to be in active form for adverse Her-2-associated effects to occur, an improved Her-2/neu testing modality may involve examining not only the presence but also the activity of the Her-2/neu surface proteins. Only protein-based methods such as IHC, which may use antibodies specific to the phosphorylated version of Her-2, are capable of identifying active (i.e., phosphorylated) proteins. Indeed, any method amenable to the use of targeting antibodies (e.g., nanostructures linked to an antibody targeting only activated biomolecules) could exploit this feature. This opportunity is supported by studies which show that only patients displaying phosphorylated Her-2 exhibited Her-2-associated adverse prognoses, indicating that overexpression of Her-2 alone is not predictive [(195)]. Thus IHC-based methods may ultimately prove most optimal over gene-only techniques such as FISH. Therefore, while FISH and IHC are currently considered the best methods for use in identifying patients most likely to respond to Herceptin treatment, there is an opportunity for other methods to exploit disease-specific characteristics. For instance, separate modalities mammography and molecular testing can identify the discrete risk factors tissue density and Her-2 status, respectively. As discussed in the following section, ultrasound modalities can be used to ascertain tissue densities, and targeted ultrasound contrast agents could simultaneously indicate the molecular status of tissue. A combination of such methods may lead to a more powerful indicator of breast cancer and its prognosis. Proof-of-principle of a method that detects amplified Her-2 expression through targeted changes in tissue density and stiffness is presented in Chapter 3. It has become clear that, no matter the modality used, the accurate identification of the Her-2 status of a breast cancer patient as early as possible is a critical determinant of patient therapeutic response to Herceptin and thus their general prognosis [(178)].

48 1.3.4 Ultrasound Modalities Ultrasound is defined as acoustic energy from 20 kHz (i.e., above the range of human hearing) to the GHz range. While many types of ultrasound exist (A-mode – amplitude scanning, B-mode – brightness scanning, C-mode – contrast-range, etc.), they are all fundamentally alike in that a transducer generates energy which propagates through materials as mechanical waves that modulate the materials’ elastic strain. Therefore, flaws, cracks, interfaces, and other material properties such as density and modulus are detectable as changes in ultrasound reflection/refraction and time course. Thus ultrasound has become a useful industrial tool for the inspection of bulk physical properties of materials, as well as for assessment of biological abnormalities with medical ultrasound. Indeed, ultrasound has become an important diagnostic imaging tool in medicine, and recent advances have shown its utility as an external trigger in the application of stimulus-responsive material for drug release [(196)]. Medical ultrasound is generally operated in the 1-10 MHz range and is commonly used because ultrasonic waves have no known deleterious effects and physical ultrasonic parameters such as reflection/refraction and the ability to focus the beams of sound waves can be utilized to visualize healthy and anomalous in vivo structures. In fact, sophisticated mathematical algorithms exist to exploit the physical properties of ultrasound waves in order to image patients for diagnostic detection of abnormalities such as tumors (albeit at relatively low resolution and decreased signal to noise compared to modalities such as MRI). In medical ultrasound, the high-frequency waves reflect at boundaries of tissue media with distinct acoustic properties, thereby mapping tissue boundaries in vivo. Ultrasound-based technologies have previously been advanced for the detection of breast cancer. New methods are being developed in which three-dimensional medical ultrasound is co-registered with mammography [(197)]. This has benefits in the ability to exploit the advantages and decrease the significance of the disadvantages of each modality; for instance, though mammography is poor at detecting tumors in women with dense breasts, ultrasound might still identify it. A molecular-level ultrasound-based analysis was recently devised to detect the Her-2/neu biomarker in vitro. Copland et. al. performed a study in which a gelatin phantom was combined with gold nanoparticles

49 targeted to Her-2 protein expressed on the SKBR-3 cell line; with this approach the feasibility of optoacoustic imaging for Her-2/neu detection was demonstrated [(198)]. Optoacoustic tomography (OAT), a blend of light and ultrasound energy, was used to generate high resolution, high contrast images. Because gold modulates both optical and ultrasonic signals, it behaved as a dual-purpose contrast agent, amplifying both the high contrast and high resolution benefits of OAT. The study presented in Chapter 3 does not utilize medical ultrasound; rather, it describes a method for adapting industrial material characterization ultrasound for use in the quantitative evaluation of biological tissue samples for malignancy and Her-2/neu protein overexpression status. Two industrially applied non-destructive evaluation (NDE) ultrasound systems were adapted for these analyses: characterization mode ultrasound (CMUS) and C-Scan, as described in section 3.2.4. These systems were equipped with specialized transducers that were used to transmit and receive ultrasonic wave reflections, which were collected, gated, and transformed from the time to the frequency domain using Fast Fourier Transforms.

1.3.4.1 Continuum Mechanics and Nanomechanics: Reconstructions A vital component of NDE ultrasound systems comprises the mathematical algorithms facilitating the reconstruction of material property values. The choice of the physical/mathematical framework critically affects the reconstructed values. For bulk solids, continuum mechanics is generally an adequate framework. However, for materials with more heterogeneous distributions of objects that can be modeled on the micro- to nanoscale, a different framework may offer additional crucial information. Therefore, because tissue consists of a heterogeneous material comprising cells and other micro- and nanoscale substances, algorithms applying both continuum mechanics and nanomechanics [(199)] were assessed in the present studies to detect cancerous tissue. Indeed, a recent effort was dedicated to the use of CMUS technology in combination with continuum and nanomechanics for the detection of cancer [(200-202)]. The models developed in these studies were constructed to correlate the reflection spectra (i.e., the tissue response) with the physical properties of the tissue [(201)]. Thus, the

50 mechanical properties of biopsied malignant tissue (such as shear, elasticity, density, and attenuation) were hypothesized to be differentiable by ultrasound from the properties of normal tissue [(201-203)]. A typical nanomechanical reflection spectrum is seen in Figure 1.5, in which the malignant (red) tissue spectra plainly differ from the normal (black) tissue spectra, validating the hypothesis. The algorithm used to create these spectra was developed based on the work of Lavrentyev and Rokhlin and Wang et. al. [(204, 205)] to reconstruct these tissue properties by curve fitting theoretical and experimental CMUS reflection spectra. This algorithm, as promulgated by Liu et. al. [(200-203)] did not account for the normal angle of incidence described in [(204, 205)], so it was modified again for use in the CMUS and C-Scan molecular imaging experiments discussed in Chapter 3. However, the mathematical basis of this algorithm as defined in references [(204, 205)] was generally unaffected by the modifications made in the present study and is delineated below. The algorithm is applicable to reconstruct the mechanical properties of layered substrates of arbitrary thickness from two angles of incidence – oblique and normal – using an inversion model; that is, six layer properties can be computed by this two-step inversion algorithm without any previous knowledge of the layer.

51

Figure 1.5: Nanomechanical reflection spectra derived from three normal (black) and three malignant (red) biopsied tissue sections as a function of frequency as reported by Liu et. al. [(200)].

Wang et. al. modeled the dependence of the acoustic response on an embedded layer with continuum mechanics using six parameters (layer properties): the elastic

moduli λ and µ, thickness h, density ρ, and longitudinal and shear attenuations α1 and αt [(205)]. Two groups of three nondimensional parameters were proposed that are determined directly from ultrasonic reflection spectra to simplify the calculation of these properties [(204)], one group per angle of incidence (normal and oblique). Three parameters were obtained from the normal angle reflection as the first step in the inversion algorithm:

52 Z l Z n = Z 0

hω0 H n = (1.1) Vl

α1

where the normalization constant ω0 = 1 MHz is a normalization factor and Vl is the

longitudinal velocity. Zl and Z0 are the acoustic impedance of the layer and substrate,

respectively, and are defined by the relations Zl =ρVl and Z0 = ρ0Vl0 where ρ0 and Vl0 signify the substrate density and velocity [(205)]. Using the three parameters obtained from the normal angle in addition to the ultrasonic spectra acquired from the oblique angle, 3 additional parameters are obtained which are simplified versions of expressions that would otherwise become complex at certain angles of incidence:

Vl Cl = Vl0

Vt Ct = (1.2) Vt0

αt

where Vl0 and Vt0 are the longitudinal and shear velocities of the substrate and

1 ⎛ u ⎞ 2 Vt = ⎜ ⎟ is the layer shear velocity [(205)]. ⎝ ρ ⎠

In addition to the attenuation parameters α1 and αt, the dimensional mechanical properties of the layer are derived using algebraic manipulation and expressed as the following functions of nondimensional parameters:

53 Z ρ ρ = n 0 Cl H C V h = n l l0 (1.3) ω0

2 2 λ + 2µ = ρCl Vl0

2 2 µ = ρCt Vt0

The computational approach used to obtain the ultrasound time domain signal is governed by the experimental configuration of one normal and two oblique transducers ultrasonically coupled to the substrate through an aqueous coupling fluid and is described in [(205)]. In practice, the determination of nondimensional parameters (Eqs. (1.1) and (1.2)) is an iterative process: first, when the bonding layer spectra can not be distinguished from those of the substrate layer due to signal interference, both the experimental and theoretical reflection normal angle spectra are gated. The experimental and theoretical spectra are compared and an error function is computed by least squares minimization of the sum of squared deviations between the spectra as defined in [(205)]. Once the normal parameters are sufficiently optimized as determined by calculation of the error function, the second inversion process is initiated to compute the oblique angle nondimensional parameters through use of the parameters determined in the first inversion. After the oblique theoretical and experimental spectra are matched and the error is calculated and optimized via least squares, all six nondimensional parameters are finally obtained. This iterative algorithm is shown in the schematic in Figure 3.1 of Chapter 3 (slightly modified in this case to complement the nanomechanical approach) and is detailed in [(204, 205)].

Nanomechanics The nanomechanical approach (also called doublet mechanics, DM, as pairs of nodes termed “doublets” form the basis of this discrete matter theory) taken by Liu et. al. provided excellent scalability to model the diverse nano- to microstructures implicit in biological samples [(201)]. Thus biological structures such as cells or subcellular organelles may be taken as nodes in the doublet mechanical method and modeled

54 accordingly. As in the continuum mechanical method, this approach was employed in order to help detect and categorize breast tumor masses taken via biopsy; it was designed as a novel, objective method, in that it is dependent only upon the intrinsic material properties of the biological sample rather than the currently used method of visual inspection by pathologists in order to achieve tissue detection/classification. Liu et. al.’s study employed the oblique angle of incidence alone with both continuum and nanomechanical models following an inversion strategy similar to that employed by Wang et. al. as described above [(202)]. In brief, Liu et. al. structured their algorithms by first obtaining a set of linear equations through modeling the displacement of ultrasonic waves as a function of amplitude, wave number, angle, and speed using a relation that satisfied both continuum and nanomechanical wave equations (in addition to the appropriate boundary and continuity conditions) [(201)]. This yielded magnitudes of wave displacements, from which reflection coefficients were computed as a ratio between the magnitudes of the reflection wave and the incident wave. The reflection spectrum was produced by interrogating the sample over a range of frequencies in order to calculate the associated reflection coefficients; a sample reflection spectrum is illustrated in Figure 1.5, in which the reflection coefficients are plotted as a function of frequency. Thus, the reflection spectrum is directly coupled to the physical properties of the tissue from which the wave displacements and reflection coefficients are resolved. Mathematically, the multi-scale nanomechanical equation which relates the axial tissue stress with strain takes the form of

pα = ∑ Aαβ ε β (1.4) β where pα is the overall nodal stress in the α-doublet, ε β is the axial nodal strain in the β- doublet, and the Aαβ, which are the doublet mechanical correlates to the continuum mechanical Lame constants, are the micromoduli between nodes α and β. The relation describing the strain in the α-doublet is

3 3 2 ∂ui 1 ∂ ui ε α = ∑τ αiτ αj + 2 ηα ∑ τ αiτ αjτ αk (1.5) i, j=1 ∂x j i, j,k=1 ∂x j ∂xk

55

where εα is the α nodal strain, τ’s indicate the direction cosines of unit vectors that link two nodes, u1/u2 are the displacements in the x1/x2 directions, respectively, and ηα is the internodal distance associated with the α node (which is equivalent to the depth of penetration of mechanical interactions between contiguous tissue constituents) [(201)]. The scalability of nanomechanics enables macrostresses (i.e., continuum level- stresses) to be written in terms of microstresses as a second-rank stress tensor. It is significant that continuum mechanical equations are retrieved when, under the appropriate conditions, nanomechanical relations are reduced. The derivation of these nanomechanical equations which govern the transmission of elastic plane waves can be found in [(199, 202)]. Thus nanomechanics comprises a completely scalable theory. Using the governing nanomechanical relations, Liu et. al.’s work demonstrated the ability to reconstruct seven nanomechanical parameters from oblique reflection spectra (as compared to the standard six continuum mechanical parameters reconstructed). It further showed the feasibility of using nanomechanics to collect additional, microstructural information (e.g., microelastic parameters and internodal distance) from a sample that is unavailable using only continuum mechanics. Indeed, a striking display of nanomechanics showed that while continuum mechanics did not generate a statistically significant differentiable signal between malignant and normal tissue, the same set of data employing the nanomechanical algorithm did (see Fig. 1.5, as the reflection spectra between the two tissue types are visibly distinct; for quantitative comparisons, see [(201)]). In particular, the nodal micromoduli A11 and A44, as well as the internodal distance η, displayed statistically significant differences between cancerous and non-cancerous tissues [(201)]. However, while all six continuum mechanical parameters can be reconstructed with only oblique angle coefficients, they can not be determined with “reasonable precision [(204)].” Therefore, while Liu et. al. proved the capability of establishing differential spectral ultrasonic signatures, the present study (Chapter 3) assessed the continuum and nanomechanical properties of contrast-enhanced nanoparticulate-targeted biological specimens with both the oblique and normal angles of incidence. For a description of the modifications made to the DM algorithm for the present study, see section 3.2.4.

56

CHAPTER 2

ATHEROSCLEROSIS

2. Atherosclerosis

2.1 Introduction Atherosclerosis is the leading cause of death in the developed world [(107, 113, 123, 206)]. The present study was designed to demonstrate proof-of-principle and validation of potential molecular imaging reagents that could allow early detection, interventions, and prediction of therapeutic response to help preclude the immense morbidity and mortality associated with the disease. The risk factors leading to cardiovascular events, the inflammatory pathogenesis, animal models presently in use, and current methods of detection and intervention were discussed in Chapter 1. Also reviewed were the differences between vulnerable and stable plaques from morphological and biochemical perspectives, as well as molecular imaging tools such as USPIOs emerging in the literature to detect them. Preliminary molecular imaging investigations must focus on validation of specific targeting reagents in atheroma models by correlation of imaging signal with histological presence of the reagent [(61)]. This is the paradigm employed in the present studies, which entail the selection of targeting reagent, development of a minimally invasive nanoscale molecular imaging agent, physical and in vitro characterization, in vivo testing via MRI, and histological corroboration of agent presence with biomarker of interest. This set of activities is sufficient for initial proof-of-principle of a diagnostic screening methodology, with the caveat that subsequent investigations must focus on the ability of the imaging reagent(s) to predict a cardiovascular event in a particular lesion [(61)]. For this purpose, high resolution MRI combined with innovative, molecularly targeted contrast agents is a promising, potentially powerful combination [(83)].

57 Thus, the major goal of the investigations presented here is to validate one or more molecular targeting reagents as selective in vivo binders of potentially vulnerable atherosclerotic lesions by corroboration between MRI and histology. The iron oxide nanoparticulate platform used consisted of either biomolecules covalently bound to the particle or antibodies linked to the particle via protein G interaction. In this way, a wide array of molecules were linked to particles and tested for their affinity capabilities to accessible areas of plaques in vivo. The targeting reagents employed were chosen on the basis of literature-driven analyses of targetable molecules on and associated with vulnerable plaques. The set of imaging markers employed and the specific motivations for use of their cognates are detailed in section 2.2. Atherosclerotic plaques present both biochemically and morphologically differentiating features between stable and vulnerable lesions. The animal models employed, WHHL and WHHLMI rabbits, demonstrate features of advanced plaques; moreover, the lesions of WHHLs are biochemically and morphologically more associated with stable lesions (thicker, more stenotic, with fewer macrophages) while WHHLMIs display some attributes of vulnerability (thinner, with increased apoptotic foamy macrophage content). Using the biochemical affinity inherent in molecular targeting reagents in combination with MRI, these studies sought to unite molecular affinity reagents with the imaging modality to provide biochemical and morphological plaque information within the same study. Recent literature confirms that corroboration of these two facets of atherosclerosis as a multiparameter approach may be critical in making an appropriate diagnosis [(207)]. It is also feasible that detection of a particular constellation of vulnerable plaque biomarkers may be necessary for clinical diagnosis. Both because of this and because of the lack of certainty about the implications of detecting any particular biomarker, a survey of potential biomarker targeting agents was constructed and experimentally tested. Because 70% of plaque ruptures are due to non-occlusive plaques [(207)], the collective aim of the present method, once it is correlated with cardiovascular “event probability,” is viewed as eventually being a prospective minimally invasive screening procedure for healthy populations who may be at some risk of plaque rupture/erosion (e.g., age, genetic, or other risk factors discussed in section 1.2.1). Molecular imaging

58 exams might also be given for periodic minimally invasive monitoring of patients who are undergoing therapeutic treatment once vulnerable plaques have been identified to identify plaque regression and monitor therapeutic response. This chapter details the use of iron oxide nanoparticulate platforms (the preparation and characterization of which are described in Chapter 4) parenterally administered to rabbits. The focus will be on the use of Annexin V conjugated USPIOs. Following presentation of the WHHL MRI studies, WHHLMI MRI data will be displayed with relevant controls. Subsequently, histological corroboration will be detailed via staining methods and electron microscopy (EM). Next, other molecular targeting reagents will be described with respect to their ability to cancel signal with MRI.

2.2 Experimental These investigations were driven by the hypothesis that accessible biomarkers in atherosclerotic plaque could be bound by appropriate parenterally administered targeting moieties. While the targeting moieties were chosen based on the pathophysiology of atherosclerosis, binding capacity in the present animal models was unknown until the system was challenged with each reagent. Therefore, when MR signal was apparently reduced due to particle administration with Annexin V USPIOs with confirmation by experiment replication, histology was commenced for corroboration. Studies were simultaneously initiated using other targeting reagents for in vivo imaging with WHHLMIs. The set of plaque biomarkers (deduced, but not necessarily previously shown, to be specific to inflammatory processes and/or vulnerable lesions) targeted is displayed in Table 2.1 and the motivation for their selection is detailed as follows.

59

Known Found in vulnerable serum or plaque circulating Targeting Markers constituent cells proteins Other vascular locations Phosphatidyl- serine Yes No Annexin V Sites of injury, inflammation Dendritic cells, cd-11b No Yes anti-cd-11b polymorphonuclear neutrophils Dendritic cells, cd-18 No Yes anti-cd-18 polymorphonuclear neutrophils HLA-DR Yes Yes anti-HLA-DR Dendritic cells, B lymphocytes LOX-1 Yes Yes anti-LOX-1 Activated platelets cd-43 No Yes anti-cd-43 Sites of injury, inflammation Collagen anti-collagen type I Yes Yes type I Distributed Collagen anti-collagen type III Yes Yes type III Distributed cd-22 No Yes anti-cd-22 B lymphocytes MCP-1 Yes Yes anti-MCP-1 Sites of injury, inflammation cd-1a No Yes anti-cd-1a No VCAM-1 Yes Yes anti-VCAM-1 Sites of injury, inflammation cd-4 No Yes anti-cd-4 T-lymphocytes Granulocytes, monocytes, cd-62L Yes Yes anti-cd-62L circulating lymphocytes Site of activation, e.g., activated cd-63 Yes Yes anti-cd-63 platelets

Table 2.1: Displays the set of targeting reagents chosen for conjugation to USPIOs. All targeted USPIOs were injected into WHHLMI rabbits. The plaque markers at left are analyzed with respect to current knowledge of their association with lesion vulnerability, presence in circulation, reagents which bind them, and other vascular sites at which they are known to exist. In combination, these imaging reagents may offer the ability to target most of the important cell and ECM constituents of potentially vulnerable plaques (endothelial, monocyte, macrophage, SMC, collagen, etc).

Apoptosis is thought to play a role in atherosclerotic plaque vulnerability [(60, 98, 110, 208, 209)] and large numbers of apoptotic cells reside in unstable lesions [(93)]. Indeed, as discussed in section 1.2.4, macrophage apoptosis is frequently cited as a major

60 contributor in the evolution of plaque vulnerability, rupture, and erosion leading to thrombus [(77, 210-214)]. Phosphatidylserine (PS) is a phospholipid that is externalized to the outer cytosolic leaflet early in the process of apoptosis. Annexin V is well known to bind PS avidly and was previously shown to bind hypercholesterolemic rabbit plaque [(215)]. Annexin V is an approximately 36 kDa protein of 320 amino acid residues folded into four separate 5 α-helical repeats [(216)]. The protein binds PS (even when PS is at very low levels) specifically and strongly under appropriate salt and calcium concentrations. The Kd of the Annexin V-PS interaction ranges from 9 to 15 nM when PS is present on cells (down to 40 pM when binding phospholipid vesicles) [(217)] and binding studies show that between 4 and 8 Annexin V molecules can bind one PS molecule. Annexin V is a potent anti-coagulant, anti-inflammatory (due to inhibition of phospholipase A2 activity), and antithrombotic agent in vivo [(218)], which could potentially be advantageous in cardiovascular applications (i.e., the targeting reagent itself may display therapeutic properties). Annexin V was thus chosen as a targeting reagent for attachment to USPIOs. Advantages of the PS biomarker include that it is not present on the surface of circulating cells and tends to be only available to vascular entities when and where apoptosis occurs. Disadvantageously, when cells undergo necrosis, PS may be exposed and bound by Annexin V (i.e., Annexin V may be specific to apoptosis and necrosis). cd-11b and cd-18 comprise a heterodimer which has been demonstrated to be upregulated under inflammatory conditions, in particular during atherogenesis, on the surfaces of monocytes, macrophages, and polymorphonuclear leukocytes [(219-221)]. While these adhesion molecules are expressed at high levels in cells resident in vulnerable plaque, they may also be present on circulating cells, potentially diminishing specificity when targeting plaque-resident cd-11b/cd-18 with USPIOs. LOX-1 (lectin-like oxidized LDL receptor) is a cell surface receptor chiefly expressed on endothelial cells. LOX-1 ligand activity has previously been associated with Watanabe rabbit plasma and lesions [(222)]; LOX-1 binds multiple classes of ligands associated with atherogenesis [(223)], implicating LOX-1 in early atherosclerosis. Further, advanced lesion intimal macrophages display dominant expression of LOX-1 [(224)], which co-localize in apoptotic cells in human lesions [(225)]. Because it is also

61 expressed on circulating cells, LOX-1 alone is unlikely to be capable of facilitating plaque diagnoses; however, co-localization of apoptosis-targeting Annexin V and LOX- 1 reagents could prove effective. The migration of monocytes through endothelium is mediated by endothelial protein cd-43, making it a very early marker of atherosclerosis prior to macrophage foam cell development [(226)]. cd-43 is upregulated in macrophage-rich areas of lesions, potentially in correspondence with the degree of atherosclerosis [(227)]. This, in addition to the therapeutic effect produced by the inhibition of monocyte adhesion via anti-cd-43 [(226)], makes cd-43 an intriguing target (though it, too, can be found in circulation). Cellular apoptosis and MMP activity are associated with degradation of the fibrous cap, as described in Chapter 1, which may engender free collagen fragments (the majority of collagen present in human lesions is types I and III [(228)]) for greater binding availability (or may allow greater penetration of targeting reagents into those areas of plaque, which may promote increased accessibility and potential for binding). Furthermore, in advanced plaques, genetic upregulation of collagen type I is observed in specific areas such as the fibrous cap and near monocytes/macrophages [(229)]. Cross- linked collagen type III is also implicated in human atherosclerotic lesions [(230)]. While collagen is definitively in all atherosclerotic fibrous caps and other specific regions (perhaps including vulnerable lesions to a greater degree with respect to accessibility for binding, though amplified collagen accumulation is also associated with more stenotic, perhaps stable, lesions [(228)]), collagen is a very common extracellular matrix (ECM) molecule. Therefore reagents for detecting collagens might obligately be combined with other agents for corroboration of plaque-type specificity. cd-1a has been demonstrated to be expressed on the surfaces of lipid-laden foamy macrophages in arteriosclerotic lesions [(231)]. However, there appear to be incongruent accounts of cd-1a presence in normal vasculature; some report no expression, while others describe its presence [(231, 232)]. If cd-1a is resolved to be expressed only in diseased tissues, it may facilitate a more specific in vivo assay of atherosclerosis, though further investigations must be performed to determine if it is specific only to vulnerable lesions.

62 VCAM-1 (cd-106) is one of a number of cellular adhesion molecules overexpressed in both early and later stage atherogenesis in activated endothelium. Along with molecules such as MCP-1, ICAMs, and Selectins (such as E- and L- Selectins), it facilitates the recruitment and internalization of inflammatory lymphocytes, monocytes, and eosinophils from the blood [(114, 233)]. VCAM-1 is considered a promising diagnostic target for vulnerable plaque in the literature due to its robust overexpression in many atherosclerosis-associated cells – e.g., endothelial cells [(91)], SMCs, and macrophages. For similar reasons, MCP-1 is a critical atherogenic protein which is also deemed a potential vulnerable plaque biomarker [(114, 234)]. Mediating a large segment of monocyte adhesion (around 80%), L-Selectin (cd-62L) is another molecule which may be of interest as a vulnerable plaque biomarker (though significant soluble L-Selectin associated with unstable angina may preclude targeting to lesions [(235)]) [(130)]. Other potential vulnerable plaque biomarkers include cd-22, cd-4, cd-63, and HLA-DR. cd-22 positive B-cells localize to areas of overexpressed VCAM-1 in the ApoE-/- mouse model [(124)], while cd-63 mRNA expression correlates directly with the progression of atherosclerosis in WHHLs (interestingly, it is a lysosomal membrane antigen which is translocated to the cell surface upon activation, but its function has not been fully elucidated) [(236, 237)]. HLA-DR is an MHC class II transmembrane antigen expressed on monocytes and macrophages, among other cells. It has been found to be strongly associated with dendritic cells in unstable lesions [(238, 239)]. cd-4 is a marker of T-lymphocytes which are found in atherosclerotic lesions, among other areas [(240)]. Once Annexin V USPIOS were prepared (see Chapter 4 for preparation and physical characterization), MR imaging experiments commenced with WHHL and then WHHLMI rabbits. Subsequently, histology and EM were performed for corroboration of iron presence and colocalization of UPSIOs with apoptotic cells, and the targeting reagents of Table 2.1 were prepared and MR imaged with WHHLMI rabbits.

63 2.2.1 Imaging 2.2.1.1 Annexin V In vitro Biochemical Characterization The ability of Annexin V nanoparticles to bind to PS, i.e., verify biofunctionality, was examined in vitro by comparing the binding of USPIOs to apoptotic Jurkat cells (ATCC, Manassas, VA) versus their ability to bind to normal Jurkat cells. Apoptosis was induced on Jurkat cells by standard methods, via treatment with camptothecin (20 µg/ml, [(241)]). Cells were verified to be apoptotic via TUNEL assay (originally developed by Gavrieli et. al., [(242)], Calbiochem, San Diego, CA) performed according to manufacturer’s instructions. In addition to comparison between the ability to bind apoptotic and non-apoptotic Jurkat cells, the inhibition of USPIO binding to apoptotic cells by known inhibitors of Annexin V biofunctionality was determined. A 100 mM solution of the calcium chelator disodium EDTA was used in the standard binding buffer (with physiologic calcium concentration). Further, Annexin V USPIOs were preincubated with 5 µg/ml of a short chain C-8 PS variant (Avanti Polar Lipids, Alabaster, AL) for 40 mins. Controls for each inhibitor were also prepared by performing the same procedures on them as in the appropriate experimental conditions, but using only the solvents of EDTA and PS. Each lot of Jurkat cells was added to 20 µl Annexin V conjugated USPIOs in addition to a physiologically appropriate concentration of calcium necessary for Annexin V binding and incubated. Annexin V USPIO bound cells were washed several times and separated from unbound cells using magnetic columns obtained from Miltenyi Biotec (Gladbach, Germany). Bound cells were eluted and counted by a Z2 Coulter Counter (Beckman Coulter, Fullerton, CA).

MRI MRI was performed with a 4.7 T small animal MRI system (Biospin, Bruker, Etlingen, Germany) to image rabbits injected with USPIOs. To minimize breathing artifacts, rabbits were situated in supine position in the scanner. Anesthesia was induced using an of 20 mg/kg Ketamine and 1 mg/kg Diazepam and maintained of 2% Isofluorane. All animals were treated according to approved ILACUC protocols.

64 The imaging protocol consisted of a 3 plane localizer, a time of flight angiography of the abdominal aorta, and an axial, fat-saturated, gradient echo sequence (repetition time: 1230ms; echo time: 11.1 ms; flip angle: 30°; Slice thickness: 3mm; field of view: 90 mm x 90 mm; Matrix 256 x 256; number of excitations: 1; acquisition time: 5min 16s). The gradient echo sequence was carried out before injection and repeated 5 min, 10 min, 15 min after injection of the particles, and again one and two days post-injection as required. The particles were intravenously administered into an ear vein. Animals were imaged repeatedly to evaluate the effects and reproducibility of replicate nanoparticulate injections. The images were analyzed using Medical Image Processing, Analysis and Visualization (MIPAV) [(243)]. Two physicians with MRI experience and a veterinarian specializing in radiology visually analyzed MRI images in consensus. Regions of interest (ROIs) were placed in the aorta, liver, kidney, and paraspinal muscle. Average signal intensities in these ROIs and their percent change over time were calculated.

WHHL and WHHLMI Imaging Approximately 0.05 mg iron USPIOs (3.6 X 1013 particles, see section 4.1.2) were injected for all Annexin V studies unless otherwise specified. About 0.24 mg iron (1.6 X 1014 particles) control protein G USPIOs were administered for all protein G control studies. As discussed previously, WHHL and WHHLMI animals were used, with New Zealand white (NZW) rabbits as the control animal. WHHLs were obtained from Covance (Princeton, New Jersey) and developed atherosclerosis representative of occlusive, stable plaque in the aorta in regions of interest, i.e., abdominal aorta. WHHLMIs were obtained directly from their developer (Masashi Shiomi, Kobe University, Japan). These animals spontaneously develop atherosclerotic lesions morphologically similar to that of vulnerable human plaques (with certain biochemical parallels) and show evidence of spontaneous myocardial infarctions leading to death. NZWs have been a standard control to WHHLs in the literature, though the original WHHL strain is derived from the Japanese white rabbit; NZWs are normal white rabbits

65 with healthy vasculatures. Stable and vulnerable plaques and normal vasculature were thus modeled by the WHHL, WHHLMI, and NZW rabbits, respectively. Rabbits were injected with appropriate amounts of USPIOs and imaged as described above and either subsequently returned to their cages for later repeat USPIO administration or submitted to sacrifice and histology as described below. All the animal studies were approved by The Ohio State University Institutional Laboratory Animal Care and Use Committee (ILACUC) and were performed observing accepted standards of laboratory animal care.

Magnetoradioisotopic Nanoparticles In order to evaluate nanoparticle localization and general biodistribution, dual magnetic/radioactive nanoscale probes were created. They were produced by labeling Annexin V USPIOs with radioisotope Technetium-99m (99mTc) for injection and imaging of rabbits with single photon emission computed tomography (SPECT), computed tomography (CT), and MRI. Modifying the protocol of Yang et. al. [(244)], Annexin V USPIOs (0.05 mg iron, approximately 30 µg Annexin V) were purified by high-gradient magnetic separation as described in section 4.1.1. A stirred solution of 0.0019 mol ethylenedicysteine (EC, ABX chemicals, Germany) was added to 1 ml 1N sodium bicarbonate (Sigma). This solution was added to a solution consisting of 0.0019 mol N- Hydroxysulfosuccinimide and 0.0019 mol 1-Ethyl-3-(3-dimethylaminopropyl)- carbodiimide (EDC). Purified Annexin V USPIOs in PBS without calcium and magnesium were added to the above solution and stirred at room temperature for 24 hours. USPIOs were magnetically separated by the high-gradient separation protocol delineated in section 4.1.1. The resulting EC-Annexin V USPIOs were stored at 4ºC until ready for use. 1-2 hours prior to injection, the required amount of 99mTc-pertechnetate

(15 mCi) was added to a solution of EC-Annexin V nanoparticles and 60 µg SnCl2. 99mTc-Annexin V USPIOs were separated from free 99mTc and eluted using a sephadex column (GE Healthcare – formerly Amersham Biosciences / Pharmacia, Uppsala, Sweden) and tested for specific activity on a gamma counter prior to injection into rabbits. All radiation safety and animal protocols were approved and followed.

66 SPECT imaging was performed in a Forte gamma camera (Phillips Medical Systems, Milpitas, CA) with a 64 X 64 X 16 matrix. 128 frames were obtained over a 360º orbit at 35 sec/frame and a 1.46 zoom. An order 6, .43 cut-off Butterworth filter was used for image processing. Anesthesia was performed as in the MRI protocol and the urethra was catheterized. Magnetoradioisotopic 99mTc-Annexin V USPIOs were injected intravenously into the ear vein, and two consecutive 30 minute scans were performed. A 16 slice Siemens computed tomography (CT) scanner (Siemens AG, Munich, Germany) with 200 mAs and 120 kVp, using 3 mm X 3 mm reconstructions was employed for anatomical co-registration of the rabbit with SPECT. A Kernel B31f was applied for medium smoothing. SPECT and CT images were fused and evaluated by OSU radiologists and technicians.

2.2.1.2 Other Targets WHHLMI Imaging Targeted USPIOs, employing the suite of targeting proteins listed in Table 2.1, were prepared either by high-gradient magnetic separation/purification or by conjugation to protein G and subsequent high-gradient magnetic separation/purification, both as detailed in section 4.1.1. Each targeting reagent was prepared using sterile PBS (without calcium and magnesium) solvent and injected (unless otherwise specified, the mass of iron prior to magnetic separation of the antibody-conjugated protein G USPIOs was approximately 0.09 mg). MRI was performed using the same imaging protocols specified in section 2.2.1.1. Histology and corroboration have yet to be performed.

2.2.2 Histology To analyze the Annexin V USPIO distribution ex vivo, after the final round of USPIOs was injected and MRI was completed, animals (whose serum and whole blood were drawn and stored) were sacrificed approximately one hour post-particle injection. Animals were heparinized and flushed with about 200 mls Ringer’s solution. Control WHHLMI, WHHL, and NZW animals who had never received iron were also included for comparative analysis. Whole aortae (resected from atrium down and co-registered for

67 areas of signal cancellation with in vivo MRI and ex vivo measurements), liver, kidney, lung, and spleen were removed from animals, placed in a 10% buffered formalin solution, and examined grossly by pathologists. Following overnight fixation, small slices of aorta were placed in 3% glutaraldehyde solution and subsequently submitted to interrogation by transmission electron microscopy. Segments of all excised tissues, including portions of aorta where MR signal cancellation was observed and where it was not, were embedded in paraffin and 4 µm sections were cut and stained with hematoxylin and eosin. Slides were also stained for iron using a standard Prussian blue iron staining protocol and counter-stained with eosin. In situ 3' nick end-labeling of DNA was performed to visualize DNA fragmentation in apoptosis using a TUNEL apoptosis detection kit. Hoechst staining (using Hoechst 33342, Sigma) was also performed to verify areas of apoptosis. TUNEL and Hoechst stains were performed on tissue sections contiguous with Prussian Blue-stained sections to examine whether USPIOs localized specifically in apoptotic regions of plaques in the aorta.

2.3 Results In this section, results of targeted USPIO biochemical, imaging, and histological studies are reported. The biochemical characterization results of Annexin V USPIOs are first displayed. Next, MR imaging Annexin V USPIO data from WHHL, WHHLMI, and control NZW rabbits are presented. The following section contains histological and electron microscopy image results elucidating USPIO localization and cellular apoptosis in rabbit aortae. Section 2.3.2 then displays imaging data of other targeting reagents that showed MR signal reduction in WHHLMI rabbits. All results are discussed in detail in section 2.4.

2.3.1 Annexin V USPIOs In vitro Biochemical Characterization The biofunctionality of Annexin V on USPIOs was validated as camptothecin treated (apoptotic, as observed in Figure 2.1) Jurkat cells were magnetically separated with greater than 15-fold more efficiency than untreated Jurkat cells (Table 2.2). This suggested that Annexin V bound to apoptotic (i.e., PS exposed) cells with much higher

68 effectiveness than to non-apoptotic cells, which indicated that binding ability was predicated on the presence of PS on the cell surface. Confirming this result, Annexin V binding to PS was clearly dependent on Ca+2 (which is well-established), as chelation of free calcium by EDTA strongly inhibited Annexin V binding to apoptotic cells. Further corroboration was obtained in the finding that magnetic separation did not transpire when Annexin V was incubated with exogenous PS. Therefore, Annexin V USPIO binding to apoptotic cells is clearly dependent upon biofunctional Annexin V and the Annexin V on USPIOs used prior to administration in rabbits was bioactive.

Jurkat Cell Apoptosis

Figure 2.1: Jurkat cells are apoptotic after camptothecin treatment. Untreated Jurkat cells are not apoptotic by TUNEL assay (left). A large percentage of Jurkat cells are apoptotic six hours after camptothecin treatment by TUNEL assay (right).

69 Annexin V Biofunctionality

Table 2.2: Annexin V on USPIOs is bioactive. Treatment conditions, designated by “+” and “-,” indicate if the condition contained a particular treatment (“+”) or was the solvent-only control (“-“). The right column displays the number (in thousands) of Jurkat cells isolated magnetically for each condition.

2.3.1.1 Imaging: WHHL, WHHLMI, NZW For initial assessment of plaque development, a medium-age WHHL rabbit was imaged using a standard contrast agent, gadolinium chelate, in a 1.5 Tesla magnet (Figure 2.2). This clearly demonstrated the utility of MRI to detect plaque. The abdominal aorta was chosen for the focus of imaging efforts because the animals consistently developed plaques in this area and the effects of respiratory, cardiac, and involuntary motion were relatively small to enable repeatable MR imaging. Imaging was performed by Dr. Johannes Heverhagen (MRI) and Dr. Nathan Hall (SPECT, below). Atherosclerosis-free NZW rabbits (negative control) imaged using Annexin V USPIOs displayed no MR signal reduction (Figures 2.3, 2.4, and Table 2.3), as did a young WHHL rabbit (prior to plaque development, see Figure 2.5). Similarly, WHHL or WHHLMI animals injected with protein G USPIOs did not exhibit signal cancellation,

70 which is represented quantitatively in Table 2.3. Parenteral administration of Annexin V USPIOs caused WHHL rabbits to develop signal cancellation rapidly (within five minutes of injection) in a region just distal to the renal artery branch. Signal cancellation increased over 15 minutes post-injection, and remained for at least an hour (Figures 2.6 and 2.7 and Table 2.3). Quantitatively, the signal cancellation due to injection of targeted contrast was roughly similar between WHHL imaging sessions (Table 2.3). The aortic region of interest was highly stenotic in WHHL rabbits in comparison to NZWs, causing the MR signal reduction to appear to fill, or mostly fill, the lumenal volume (Figures 2.6, 2.7, and 2.8). Figure 2.8 depicts a second WHHL rabbit imaged with Annexin V USPIOs in which contrast appears as both (mostly) volume filling and along the aortic wall. Repeated imaging in WHHLs was performed via administration and imaging with Annexin V USPIOs (Figure 2.6) and re-administration and imaging with Annexin V USPIOs 70 days later (Figure 2.7). As signal decayed within days, clearly after 70 days the initial USPIO-produced signal reduction was no longer apparent on MRI. Repeat administration generated qualitative and quantitative MRI results of similar magnitude and rapidity of contrast development (Figures 2.6 and 2.7 and Table 2.3). ROIs placed in regions other than aorta did not exhibit quantitative signal cancellation. Annexin V USPIOs were observed to direct contrast to the same anatomical site (i.e., in the abdominal aorta distal to the renal branch) in WHHLMIs as in WHHLs. The set of WHHLMI images was qualitatively internally consistent in displaying signal reduction (Figures 2.9), but somewhat distinct from the set of WHHL images. The distribution of signal reduction was more periannular (along the vessel wall) in nature in the WHHLMI than in the WHHL (Figures 2.9 and 2.10). While the initial gadolinium study (Figure 2.2) exemplifies the capacity of MRI to acquire morphological information from atherosclerotic plaques, the above data further reveals the ability for simultaneous extraction of biochemical data. Table 2.3 further shows that equivalent amounts of injected Annexin V USPIOs into WHHL and WHHLMI rabbits cause quantitatively greater signal cancellation in WHHLMIs than in WHHLs. Figures 2.10 and 2.11 represent axial and coronal MR images produced when 6 times the previous amount of Annexin V USPIO iron was injected. Table 2.3 represents these ROI values, as well as those of a second WHHLMI rabbit dosed with Annexin USPIOs, quantitatively.

71 SPECT was performed to monitor the whole body biodistribution of Annexin V USPIOs. 15 mCi 99mTc and EC-Annexin V USPIOs were mixed then separated by sephadex column. Unbound 99mTc was rinsed, and a gamma counter demonstrated that approximately 6 mCi 99mTc were bound to Annexin V USPIOs (about 0.05 mg iron, or about 4 X 1013 particles); it was estimated that no more than approximately 5-10% of the 99mTc-labeled Annexin V USPIO solution consisted of free 99mTc. The 99mTc Annexin V USPIOs were then injected into the animal, with CT performed for anatomical registration. SPECT and CT images were fused and interpreted by Dr. Nathan Hall and Dr. Jun Zhang in OSU Radiology. Images showed gamma activity in the liver, spleen, kidneys, stomach, and bladder, indicating the USPIO biodistribution (see Figure 2.12).

WHHL Imaging: MRI

Plaque

Figure 2.2: WHHL rabbits exhibit plaque using MRI and gadolinium contrast. Axial view of WHHL rabbit abdominal aorta imaged prior to (left) and after (right) injection of untargeted gadolinium chelate contrast agent. Arrows designate visible plaque in the aorta.

72 NZW Imaging: MRI

Figure 2.3: NZW rabbit displays no MR signal cancellation due to Annexin V USPIO injection. Axial view of negative control NZW rabbit abdominal aorta prior to (left) and 15 minutes after (right) injection with Annexin V USPIOs. Arrows indicate aorta.

NZW Imaging: MRI

Figure 2.4: NZW rabbit displays no MR signal cancellation due to Annexin V USPIO injection. Coronal view of negative control NZW rabbit abdominal aorta prior to (left) and 15 minutes after (right) injection with Annexin V USPIOs. Arrows designate the aorta.

73

WHHL Imaging: MRI

Figure 2.5: Young WHHL displays no MR signal cancellation when injected with Annexin V USPIOs. Axial section of a young WHHL rabbit (negative control) prior to (left), 5 minutes (center), and 15 minutes (right) post-injection of Annexin V USPIOs. Arrows indicate aorta.

WHHL Imaging: MRI

Figure 2.6: WHHL rabbit exhibits MR signal cancellation within 5 minutes of Annexin V USPIO administration. Axial section of plaque-laden WHHL rabbit abdominal aorta prior to (left), 5 minutes (center), and 15 minutes (right) post-injection with Annexin V USPIOs. Arrows indicate the aorta.

74 WHHL Imaging: MRI

Figure 2.7: Same WHHL rabbit exhibits similar MR signal cancellation after 70 days. Axial section of the same plaque-laden WHHL rabbit abdominal aorta as in Figure 2G, two months later, prior to (left), 5 minutes (center), and 15 minutes (right) post-injection with Annexin V USPIOs. Arrows indicate the aorta.

WHHL Imaging: MRI

Figure 2.8: A second WHHL displays MR signal cancellation after Annexin V USPIO administration. Axial section of a different plaque-laden WHHL rabbit abdominal aorta prior to (left) and 15 minutes (right) post-injection with Annexin V USPIOs. Arrows indicate the aorta in areas of signal reduction.

75 WHHLMI Imaging: MRI

Figure 2.9: WHHLMI rabbit displays MR signal cancellation within 5 minutes of Annexin V USPIO administration. Axial section of WHHLMI rabbit abdominal aorta prior to (left) and 15 minutes (right) post-injection of Annexin V USPIOs. Arrows indicate regions of vessel wall revealing signal reduction.

WHHLMI Imaging: MRI

Figure 2.10: WHHLMI rabbit exhibits MR signal cancellation within 5 minutes of injection of 6 times the previous doses of Annexin V USPIOs. Axial section of the same WHHLMI rabbit abdominal aorta as in Figure 2.9 prior to (left) and 15 minutes (right) post-injection of 6 times the amount of Annexin V USPIOs (approximately 0.3 mg iron) used in other studies. Arrows designate regions of signal reduction in the aorta.

76 WHHLMI Imaging: MRI

Figure 2.11: WHHLMI exhibits MR signal cancellation within 5 minutes of injection of 6 times the previous doses of Annexin V USPIOs at multiple aortic sites. Coronal section of the same WHHLMI rabbit imaging session as in Figure 2K prior to (left) and 15 minutes (right) post-injection of 6 times the amount of Annexin V USPIOs (approximately 0.3 mg iron) used in other USPIO administrations. Arrows designate regions of signal reduction in the aorta.

77 WHHLMI Imaging: SPECT

Liver Stomach Kidneys Spleen

Bladder

Figure 2.12: SPECT and CT fusion image of WHHL rabbit reveals the biodistribution of 99mTc labeled Annexin V USPIOs in bladder, kidney, spleen, liver, and stomach. The bright spots represent SPECT gamma imaging, while the outline of the animal is from CT imaging on a different date. The images were fused by software.

Signal change after injection of nanoparticles [%] Animal/Particle Type Imaging Day 5 min 15 min WHHL 1/Annexin V 0 -9 -11 WHHL 1/Annexin V 70 -7 -11 WHHL 1/Protein G 0 +4 +1 WHHL 2/ Annexin V 0 -18 -16 WHHLMI 1/Annexin V 0 -24 -23 WHHLMI 2/Annexin V 0 -21 -24 WHHLMI 1/Annexin V 70 -26 -36 (6X Fe)

WHHLMI 1/Protein G 0 -3 +1 NZW/Annexin V 0 -1 +2

Table 2.3: Region-of-interest (ROI) analysis of the change in signal following Annexin V and control protein G USPIO administrations in rabbit abdominal aorta. This table serves as ROI quantifications of the MRI Figures of this chapter.

78

2.3.2.2 Histology: WHHL, WHHLMI, NZW Resected aorta from WHHL and WHHLMI rabbits displayed grossly visible atherosclerotic plaque in the region of MR signal cancellation (i.e., distal to the renal areteries) as inspected by two independent pathologists. Other areas of the resected vasculature also displayed visible atheroma. Conversely, NZW rabbits exhibited normal aortic morphology. WHHL lesions were observed to exhibit marked subintimal expansion of fairly uniform thickness containing numerous acicular (cholesterol) clefts, foamy macrophages and modified SMCs, scattered pyknotic cells, and sporadic mineral deposits. Sections from regions near MR signal reduction lacked much of the endothelial layer, while sections from other areas of the same aorta displayed only intermittent endothelial gaps. These lesions were graded AHA type III, with some type IV features [(76, 77, 113)]. WHHLMI rabbits displayed everything shown by WHHLs, but revealed more features characteristic of type IV lesions (e.g., more foamy macrophages, mineral deposits). Further, these lesions would be classified as AHA type IV if they contained necrotic or lipid cores and fibrous caps, and though unambiguous necrotic cores and fibrous caps were not identified, they both have previously been reported in the WHHLMI model [(157)]. In animals which received Annexin V USPIOs, Prussian blue staining (indicating the presence of iron) was rare in tissue sections obtained from regions outside of the area of MR signal cancellation. Likewise, no iron was identified in the aortae of animals which had not received iron and NZWs. Conversely, Prussian blue staining was unmistakable and chiefly localized to macrophages in the regions displaying MR signal reduction of WHHL and WHHLMI rabbits injected with Annexin V USPIOs (see Figures 2.13 and 2.14). At least in the case of WHHLMIs, Prussian blue staining was typically observed in morphologically foamy macrophages. USPIOs were apparent in relatively shallow regions of the WHHL plaques (Figure 2.13), while they were detected deep (i.e., not near the arterial lumen) in WHHLMI lesions (see Figure 2.14). Control sections from remote sites of the same aortae did not stain with Prussian blue (see Figure 2.15). Sections proximate to WHHLMI sections containing iron were stained by TUNEL and Hoechst assays to identify apoptosis unambiguously as in Figures 2.16 and

79 2.17. The same area staining with Prussian blue and TUNEL stained positive for apoptosis via Hoechst (see Figures 2.16 and 2.17). Sections remote from the site of iron localization were also assayed via TUNEL and Hoechst (see Figures 2.15 and 2.17), and both TUNEL and Hoechst verified these sections to be free of apoptosis. While NZW sections exhibited no evidence of apoptosis, apoptotic cells were detected in WHHLMI lesions. Thus, apoptotic cells generally co-localized with iron accumulation, and lack of iron accumulation correlated with a lack of apoptosis. Transmission electron microscopy (TEM) revealed Annexin V USPIOs localized to a cell with macrophage morphology within a plaque in the region of MR signal cancellation of a WHHLMI (see Figure 2.18). In particular, USPIOs appeared tethered to a membranous structure, which would theoretically be expected of Annexin V USPIOs (i.e., binding to externalized PS on a membrane). UPSIOs were not located by TEM in other areas of the aorta nor on the aorta of a WHHLMI which had not received iron.

80 WHHL Histology

A B

C D

Figure 2.13: Sections obtained from the region of WHHL aorta which caused MRI signal reduction exhibit Prussian blue staining. Top row: H&E stained sections. A: Low-magnification view. Circles represent approximate areas where Prussian blue staining was found (as shown in bottom row). B: Higher-magnification. Bottom row: Prussian blue stained sections. Staining appears to localize to certain cells with macrophage morphology.

81 WHHLMI Histology

A B

C D

Figure 2.14: Representative sections obtained from the region of WHHLMI aorta which caused MRI signal reduction display clear Prussian blue staining localized to foamy macrophages. Top row: H&E stained sections. Low-magnification (4X) view (A). Indicated areas represent where Prussian blue staining was found (as shown in bottom row with higher magnification). 10X magnification of the designated region (B). Bottom row: Prussian blue stained sections. Staining appears to localize to foamy lipid- laden macrophages. 10X magnification of the same region in a section contiguous to that of the top row (C). 40X magnification of designated section (D).

82 WHHL Histology

A B

C

Figure 2.15: Representative control sections taken from remote areas of the aorta (i.e., no signal reduction) from the WHHL of Figure 2.13 A and B and the WHHLMI of Figure 2.14 C display no staining with Prussian blue. No staining was found in sections from the WHHL aorta just distal to the region of MRI signal reduction (A) nor in the aortic arch of WHHL (B) or WHHLMI (C).

83 WHHLMI Histology

Figure 2.16: Correlation between representative Prussian blue positive section and TUNEL (apoptosis) positive section. The iron in a Prussian blue stained section (left) corresponds with apoptosis in a contiguous TUNEL stained section (right).

WHHLMI Hoechst

Figure 2.17: By Hoechst assay, cells are apoptotic in regions of aortic iron localization, but they are not apoptotic in regions iron was not found. From the region of Figures 2.14 and 2.16 in which Prussian blue staining is observable, apoptotic nuclei are visible (left), but from other regions of the same aorta, nuclei do not appear apoptotic (right).

84 WHHLMI EM

Figure 2.18: EM reveals Annexin V USPIOs in cells within WHHLMI plaque in region of MR signal reduction. Transmission electron micrographs (TEM) of rabbit aorta plaque macrophage with Annexin V USPIOs apparently bound to a cell membrane (left, see arrows) in comparison to Annexin V USPIOs on a grid. Magnification: 20,000X in both images.

Organ Distribution MRI signal cancellation was displayed only in the areas of abdominal aorta shown (e.g., Figures 2.6 and 2.11) when employing Annexin V USPIO doses of both 0.05 and 0.3 mg of iron. No signal reduction was detected in possible clearance organs such as liver, lung, kidney, and spleen, nor in any other organ observed. While Prussian blue stained areas of the liver and spleen, this was not necessarily attributable to USPIOs – indeed, pathologists interpreted the amount of splenic Prussian blue staining to be within normal limits. Therefore, only representative micrographs of spleen are displayed in Figure 2.19. Based solely on Prussian blue staining of the spleen, it was not possible to distinguish rabbits which had been injected with USPIOs from those which had never been injected (due to high variability in staining appears between animals, as observed in Figure 2.19). Indeed, splenic Prussian blue staining is common and ascribable to hemosiderin. No Prussian blue staining was observed in other organs, except for faint traces of blue in Kupffer cells of the liver which were not necessarily due to USPIOs. Therefore, to monitor the USPIO biodistribution, magnetoradioisotopic USPIOs were

85 prepared, injected, and imaged with SPECT as detailed in section 2.2.1.1 (see Figure 2.12).

WHHL Organ Histology

A B

C

Figure 2.19: Rabbit spleen stains with Prussian blue, which is not necessarily indicative of USPIO presence. WHHL spleen, injected with Annexin V USPIOs (A). WHHLMI spleen, never injected with USPIOs (B). WHHLMI spleen, injected with USPIOs (C).

2.3.2 Other Targets Of the other targets tested (see Table 2.1), administration of anti-VCAM-1 and anti-collagen type I conjugated USPIOs caused signal cancellation (see Figures 2.20 and 2.21) in WHHLMI rabbits in regions distal to the renal branch of the aorta. Similar to the Annexin V USPIO administrations to WHHLMIs, these targeting reagents caused periannular signal cancellation which appeared to some extent to be in the aortic wall, especially in the case of anti-collagen type I USPIOs. Not all of the other targeting reagents have yet been completely analyzed for signal cancellation, so omission from the present list does not necessarily imply that a targeting reagent did not cause signal reduction.

86 WHHLMI MRI: anti-VCAM-1

Figure 2.20: WHHLMI rabbit exhibits MR signal cancellation due to anti-VCAM-1 USPIO administration. Axial section of WHHLMI rabbit abdominal aorta prior to (left) and 15 minutes (right) post-injection of anti-VCAM-1 conjugated USPIOs. Arrows indicate regions of vessel wall revealing signal reduction.

WHHLMI MRI: anti-Collagen Type I

Figure 2.21: WHHLMI rabbit exhibits MR signal cancellation within 5 minutes of anti- collagen type I USPIO administration. Axial section of WHHLMI rabbit abdominal aorta prior to (left), 5 minutes (center), and 15 minutes (right) post-injection of anti- collagen type I conjugated USPIOs. The vena cava is demarcated by the star, arrows indicate the aorta, and red circles reveal signal reduction in the plaque on the aortic wall over time.

87 WHHLMI MRI: anti-cd-11b

Figure 2.22: WHHLMI rabbit exhibits signal cancellation due to administration of anti- cd-11b conjugated USPIO administration. Axial section of WHHLMI rabbit abdominal aorta prior to (left) and 15 minutes (right) post-injection of anti-cd11b conjugated USPIOs. Arrows indicate regions of vessel wall revealing signal reduction.

2.4 Discussion The suite of studies pursued in this chapter represents a nanotechnological molecular imaging approach toward the early detection of vulnerable atherosclerotic lesions through biochemical and morphological cues. Targeted USPIOs were developed, characterized, and subsequently tested in heritable animal models of atherosclerosis to illustrate the potential to image biochemical indications of plaque vulnerability, while imaging modality MRI offered morphological verification of lesion stability. The rationale for these studies was derived from the potential for these nanodevices to become part of a screening tool for populations at risk for atherosclerosis so that treatments could be established prior to catastrophic events such as heart attacks. This discussion first details the application of Annexin V conjugated USPIOs for lesion detection from their bioactivity to in vivo MR contrast to histology and electron microscopy for corroboration of USPIO presence and the presence of apoptosis (i.e., localization of Annexin V ligand, phosphatidylserine). In this section, the biodistribution of Annexin V USPIOs will be discussed via histology and SPECT data compiled from the development of magnetoradioisotopic Annexin V USPIOs. Subsequently, the

88 development and in vivo imaging of other USPIO-based targeting reagents with MRI will be described.

2.4.1 Annexin V The biofunctionality assay for Annexin V was designed to ensure that Annexin V was bioactive on USPIOs prior to injection. This was critical because all conclusions made from imaging and histological results with these particles assume bioactive protein on the USPIO surface. A TUNEL assay verified that camptothecin-treated Jurkat cells were apoptotic for the Annexin V bioactivity assay. As shown in Table 2.2, apoptotic Jurkat cells were magnetically harvested more than 15 times better than untreated (and therefore non-apoptotic) Jurkat cells incubated with Annexin V USPIOs. This indicated that the USPIOs preferentially bound to apoptotic Jurkat cells and that the interaction between Annexin V and PS (which is externalized when cells undergo apoptosis) may have been responsible for this binding. The additional conditions in the bioactivity assay devised to assess this interaction clearly demonstrated that the Annexin V-PS association was implicated; first, the strong dependence of Annexin V binding to PS on Ca+2 ion concentration was borne out as the EDTA chelator caused Annexin V binding to diminish to control (i.e., that of Annexin V USPIOs with non-apoptotic cells) levels of cell binding. Furthermore, competitive inhibition via pre-incubation of Annexin V USPIOs with a short-chain PS variant (at a concentration less than the critical micelle concentration, or CMC) again demonstrated only control levels of Annexin V binding to apoptotic cells as determined by magnetic separation. Thus, because PS was already bound to Annexin V, the Annexin V could not bind to PS on cell surfaces. These experiments strongly implied that Annexin V UPSIOs bind to apoptotic cells due to externalized PS; therefore, biofunctional Annexin V was present on USPIOs prior to administration to rabbits. In brief, this discussion will focus on data indicating that these investigations revealed the ability to direct USPIOs to vascular plaques using biochemical affinity to vulnerable plaque constituents (the externalized PS of apoptotic cells). The data verified that both atherosclerotic plaque and Annexin V were essential to elicit signal cancellation as neither NZWs receiving Annexin V USPIOs nor plaque-laden Watanabe rabbits

89 receiving protein G USPIOs revealed differences in signal. It is further clear that signal cancellation was due to the presence of Annexin V USPIOs, as USPIOs were identified at sites of signal reduction with Prussian blue staining and electron microscopy while minimal evidence of USPIOs was detected at remote sites. Localization of USPIOs to specific sites was due to the association between Annexin V and its ligand PS, as the USPIO-bound Annexin V was bioactive and apoptotic cells co-localized with iron accumulation (DNA fragmentation, as observed via TUNEL, occurs later in the course of apoptosis than PS externalization thus implying the presence of external PS in TUNEL- stained cells; Annexin V USPIOs in these studies bound cells that stained with TUNEL), and there was no evidence of USPIO accumulation at remote sites where no apoptosis was observed. Therefore, MRI (by signal cancellation) revealed that Annexin V affinity for PS drove Annexin V USPIOs to plaques containing apoptotic cells. As previously described, atherosclerosis is an inflammatory disease state that can lead to acute cardiovascular events particularly when plaque vulnerability indicators such as apoptotic cells are present. The literature shows that both apoptotic macrophages and SMCs stain with Annexin V, though at least one study revealed Annexin V is taken up by macrophages in macrophage-laden advanced atherosclerotic plaques, but not SMCs [(113)]. Thus, in the context that apoptotic cells are present in unstable lesions, and the protein on USPIOs is bioactive, the MRI results demonstrating the ability to detect signal cancellation due to Annexin V USPIO injection of atherosclerotic rabbits are logical. Annexin V USPIOs in these studies predominantly targeted apoptotic macrophages in plaques. While this may have been partially a function of the route of entry of USPIOs (e.g., via the lumen as opposed to vasa vasorum), previous reports show that Annexin V binds to, and uptake correlates specifically with, apoptotic macrophages [(113)]. In support of the “route of entry” argument, SMCs are often further from the lumen than macrophages (SMCs are in the media, though they do proliferate and migrate into the intima during atherosclerosis, while macrophages are commonly part of the thickening intima), so if USPIOs penetrate plaques via the lumen, it may be logical that they bind apoptotic cells (often macrophages) nearer the lumen. Nevertheless, several SMCs in one plaque did exhibit Prussian blue staining. Thus, it may be possible that the Annexin V USPIOs target either apoptotic macrophages or SMCs, but that macrophages

90 more commonly experienced apoptosis in plaques which exhibited MR signal cancellation (this was the case based upon general observations; however, because a study specific to this question was not performed nor explored in detail with appropriate controls, a cogent answer to why USPIOs were not generally found in SMCs is not feasible). The broad objectives of this work entail use of biochemical affinity particulates (i.e., protein-conjugated USPIOs) to become a screening agent for the prediction of atherosclerotic events in at risk populations using MRI. As will be discussed later in this section, the capacity of a screening modality to differentiate between unstable plaques and stable plaques is likely to require more than one parameter; a multimodal analysis, such as offering biochemical and morphological information, would be superior to single parameter assessments that are in current use as described in Chapter 1. Though the present study did not seek to correlate USPIO targeting capabilities with predictive capacity for a cardiovascular event (a future objective), the goal was to demonstrate the feasibility of a multimodal biochemical (Annexin V) and morphological (via MRI) analysis of atherosclerosis. Many methods have been used to visualize atherosclerotic plaques (e.g., radiolabeled antibodies, lipoproteins, 18F-fluorodeoxyglucose, fibrinogen, antisense oligonucleotides (ASONs), radiolabeled Annexin V, targeting neovascular antigens with liposomal gadolinium structures, and bare USPIOs [(86, 113, 206, 245)]). Studies employing bare (untargeted) USPIOs have demonstrated contrast enhancement to the aortae of hyperlipidemic rabbits [(85, 86, 246)]. Targeting, however, provided potentially major advantages over those studies including dose reduction and markedly rapid contrast development. Bare USPIOs were used in quantities 100 times greater than the acceptable human dose for contrast agent iron. Indeed, the present targeting studies consistently employed 2000 times less iron on a mg iron/kg bodyweight basis than the untargeted USPIO studies; thus, 20 times less iron was used than permissible in humans. Clearly dose reduction was anticipated, as it is a known advantage of using targeting agents; however, the extent of the advantage was larger than expected. This could be partially explained by variations in the USPIO reagent used in these studies compared to bare USPIO investigations, including differences in particle size, surface coating, surface

91 charge, iron oxide composition, etc (however, note that even with effectively identical USPIO attributes, 5 times the amount of iron was used for protein G USPIO administrations, yet no signal reduction was observed). These property differences could not only result in differences in uptake into the plaque, but also in variations in particle blood half life (i.e., clearance rates). Both of these deviations would be expected to modulate dose requirements (for instance, longer clearance times would cause longer exposure times of USPIOs to plaque), although it is not known whether they would increase or decrease the quantity of USPIOs needed for contrast development. The best way to examine this would be to conjugate USPIOs used in the above untargeted studies (e.g., Feridex, which is commercially available and clinically approved) with Annexin V and determine how their properties change (or alternatively to establish how unconjugated versions of the Annexin V UPSIOs used in the present studies behave in vivo). The reduction in dose enabled by targeting could make this reagent more viable as a clinically useful screening reagent, or even as a reagent capable of monitoring therapy, as it would need to be administered repeatedly; repeat administrations would be feasible due to the miniscule dose required. Previous studies with bare UPSIOs reported contrast development within a day of injection (and increased in signal in ensuing days). Conversely, the present studies clearly and consistently demonstrated the emergence of contrast within 5 minutes of USPIO administration. The extremely rapid speed of onset may not only be of practical use for the ultimate objective (i.e., a screening procedure involving being dosed and imaged within minutes is preferable to being dosed and waiting a day or more and then returning to the imaging center for MRI), but it also implies that cells that have been resident within the plaque (pre-injection) are taking up the USPIOs. The alternative is that circulating cells take up USPIOs and penetrate the plaque, but this is unlikely first because of the depth of penetration observed for USPIO localization (which seems to be a great distance for a cell to travel within an hour) and second because apoptotic cells are not expected to circulate to a great extent. Because iron co-localized with apoptotic cells within plaque, the cells were apparently previously in the plaque. The deep, rapid localization of USPIOs into plaques is notable and may suggest increased prevalence of any of a number plaque attributes such as endothelial denudation and permeability [(87)],

92 cellular motility, enhanced ECM permeability, or other as yet unidentified events within the iron-containing lesions. For instance, evidence in the form of histological observation that endothelium was denuded at sites of signal cancellation compared to other regions of vasculature indicates that endothelial abnormalities may have played a role in the uptake of USPIOs into the plaque. Another consideration is that while imaging was performed only 5-15 minutes post-USPIO administration, animal sacrifice and tissue fixation occurred about an hour post-injection. It is therefore possible that USPIOs caused MR signal cancellation while still near lumenal surfaces, but migrated into the plaque and bound apoptotic cells deep in the lesion within the intervening 45 minutes prior to sacrifice and were thus identified there in histology; experiments to test this might include imaging and rapid sacrifice of animals around 5, 10, and/or 15 minutes post UPSIO injection and inspecting tissues for depth of particle localization. It is also feasible that the rapid uptake of USPIOs deep into lesions is because the particles are not delivered via the vascular lumen, but rather by the vasa vasorum (as mentioned above), or via a combination of the two. This might be investigated either by studying the penetration of USPIOs on a freshly resected aorta to determine if similar depth of penetration is observed by lumen, or again by sacrificing at various time points in order to determine the route of entry (e.g., Prussian blue may exhibit staining near the lumen and/or near vessels of the vasa vasorum at earlier time points, suggesting the favored path to apoptotic cells). It may also be feasible to use higher resolution MRI to determine route of entry, as larger magnets, coils, and more sophisticated sequences [(247)] may yield nearly cellular-level resolutions capable of tracking USPIO progression and cellular association. While it is unknown what the standard route of entry of USPIOs into plaques is, a survey of the literature suggests that it is generally assumed that USPIOs penetrate plaque via the lumen. However, as with reduction in dose concerns, this may vary somewhat based on USPIO properties (size, charge, etc.) and should be examined further in order to determine the precise mechanism of USPIO infiltration into plaque. Better understanding of this mechanism may enable improved USPIO design for optimal uptake and dosing characteristics. Repeat administrations of Annexin V UPSIOs produced similar sets of images at the same aortic site over the time course used throughout these studies. This not only

93 demonstrated repeatability in terms of verifying the initial imaging session, but it also proved that multiple imaging sessions with Annexin V UPSIOs in the same organism was feasible. Successful repeat UPSIO administrations showed that the lesion maintained an Annexin V-binding phenotype over the intervening 2 months, that no physiological response (such as an antibody response) occurred that would preclude repetitive imaging with Annexin V conjugated USPIOs, and indicated that repeat injections may be feasible for eventual clinical use in patients (because atherosclerosis is chronic, repeat administrations/screenings would be likely). Imaging of rabbits within days of Annexin V USPIO administration demonstrated no further signal cancellation. The mechanism of MR signal decay in this instance is presently unknown. While the USPIOs used in all the animal studies are known to biodegrade in vivo within a week [(248)], the experiments performed here indicate either more rapid degradation than previously reported or another mechanism for the return of signal to normal. The kinetics of signal decay must be determined with more precision, such as by imaging the animal every 15 minutes for the first 90 minutes post-injection, and subsequently by imaging every 3 or 4 hours. Animals might also be sacrificed at various times points (after imaging) to resolve USPIO presence at those respective times to yield an indication of the mechanism of signal decay (e.g., due to cell motility, or cells migrating out of plaques, as has been reported [(249)]). For eventual clinical screening, the ability to inject and image repeatably with this agent will be a requirement and further knowledge of the mechanism and kinetics of contrast decay will also be instrumental. The ability of Annexin V USPIOs consistently to drive signal contrast to apoptotic areas of lesions and their size (around 100 nm) may elicit new potential applications, such as delivery (replete with simultaneous monitoring of efficacy) of therapeutic. Given the reduction in dose enabled by targeting, therapeutic dose might likewise be minimized, thereby diminishing systemic exposure to the therapeutic (which may decrease toxicity and side effects associated with the drug). On the other hand, it is not currently known whether the present quantity of USPIOs delivered to plaque is sufficient to transport an efficacious dose of therapeutic, as only 4 X 1013 UPSIOs are injected. Indeed, as SPECT imaging of Technetium labeled Annexin V USPIOs (a novel, targeted dual imaging USPIO capable of simultaneously driving contrast effects by

94 imaging with MRI and SPECT) revealed, much of the USPIO biodistribution was dispersed to organs such as liver, spleen, kidneys, bladder, and stomach. Thus the therapeutic would either have to be non-toxic to these organs, or be configured as activatable drugs or linkages (e.g., by using “smart” delivery methods such as pro- enzymes or by linking drugs via enzymatically cleavable linkages to USPIOs in which the enzyme is known to be specifically present in unstable plaques – such as MMPs or urokinase plasminogen activator [(250)]). Alternatively, intelligent modification of USPIO properties such as size and surface materials may decrease distribution to the reticuloendothelial (RES) system and increase localization to relevant lesions. In terms of the choice of therapeutic to deliver, because contrast appears to localize to apoptotic macrophages, it would be natural to employ a therapeutic amenable to treating these regions of the plaques. On the other hand, once the mechanism/route of USPIO accumulation has been established, it may be feasible to deliver therapeutic based on sites through which USPIOs traverse. Thus, therapeutics affecting cholesterol levels and apoptosis (e.g., statins) or affecting ECM stability and permeability (e.g., MMPIs, or MMP inhibitors) may be viable options. Another study showed that antioxidants decrease macrophage infiltration and apoptosis, which led to diminished plaque vulnerability in a rabbit model [(251)]; similar work has been demonstrated with statins, which caused decreased macrophage apoptosis and increased plaque stabilization in rabbits [(252)]. These results suggest that it will be possible to monitor lesion severity by biochemical analysis and clinically useful to do so. In particular, it may be clinically valuable to repetitively administer an apoptosis- detection reagent such as Annexin V USPIOs to monitor the therapeutic response of a treatment whose clinical impact is (at least partially) achieved via modulation of cellular apoptosis (e.g., statins).

Gadolinium chelates, which are paramagnetic structures that induce T1 MRI enhancements (positive), have lower intrinsic relaxivities than iron oxides and thus are generally incorporated into nearly microscale (~300 nm) particulate formulations [(123,

253)]. Magnetic iron oxides were chosen due to the T2 signal enhancement and higher relaxivity produced by superparamagnetic iron structures, which yielded flexibility in choice of core size and surface materials/chemistry and some prior characterization

95 methods (as well as commercially available materials). Though iron oxides generally produce negative MR contrast, it has become feasible to create MR sequences which enable generation of positive MR contrast using iron [(254)]. These sequences add even further utility to iron, as they allow improved sensitivity – for instance, positive contrast can resolve difficulties in differentiating between signal loss caused by iron and by low tissue signal. The animal models employed in these studies were chosen for their distinct representations of human plaques: WHHLMI for vulnerable plaque, WHHL for stable plaque, and NZW for healthy vasculature. These were considered appropriate animal models because the WHHL, originally developed to produce atherosclerosis, and its descendant the WHHLMI, which was designed to die from myocardial infarction, exhibit morphologically and functionally distinct plaques that correspond to previously described morphologies of human plaque: thick, stable stenotic plaques and thinner, fibrous capped, less obstructive unstable plaques, respectively [(77)]. Other, cellular-level features of these animal models closely parallel human disease – for instance, apoptosis results suggest that comparatively more apoptosis and (particularly foamy) macrophages occur in the WHHLMI rabbit and considerably less in the stenotic WHHL rabbits [(107, 157, 255, 256)], corresponding to respective human degrees of apoptosis in vulnerable and stable lesions. Apoptosis has been observed at all levels of human atherosclerosis; amounts of apoptotic cells steadily increase as the plaques advance [(108, 257)], which corresponds to the relatively low levels in the WHHL compared to the WHHLMI animal model. Further, the majority of apoptotic cells in human disease are reported to be macrophages [(107, 108, 255)], which has significant implications on lipid core formation and the eventual fate of the plaque [(87, 107, 113, 252, 255)]. Therefore, monitoring macrophage apoptosis in vivo is potentially clinically valuable. This is supported by a number of studies. First, examination of arterial vessels of human patients who suffered sudden coronary death due to plaque rupture demonstrated prevalent macrophage apoptosis at the rupture site, while apoptosis was minimal at remote sites [(113)] (similar to WHHLMI rabbits at the imaged lesion site). It has been reported that not only may intimal apoptosis play a role in plaque vulnerability, but it may in fact trigger acute coronary events [(113)] (thus, the anti-thrombotic effects of Annexin V in a

96 clinical setting may also be useful), which is compatible with iron accumulation occurring in intimal sites of the imaged plaque of the vulnerable WHHLMI model, but not as much in the stable WHHL model nor in remote sites of WHHLMI plaques. It is very important to note that while minor macrophage apoptosis is known to occur in early atherosclerotic lesions, its significance is completely distinct from that in advanced, potentially vulnerable lesions. In particular, apoptotic macrophages are indicators of decreased plaque progression for early lesions, while they engender necrotic core formation and thrombosis at later stages [(258)]. Other biochemical elements and plaque morphology, as well as degree of apoptosis (and thus degree of signal cancellation), may therefore facilitate differentiation between a potentially dangerous plaque from an early lesion. Corresponding to their respective lesion type, multiple WHHLMIs spontaneously died of apparent cardiac arrest over the course of the studies, while no WHHL rabbit died unexpectedly. Though no thrombus was located, WHHLMI deaths were consistent with some signs of myocardial infarction, which is compatible with reports from the literature [(156-158, 259)]. WHHLMIs present electrocardiograms (ECGs) with features characteristic of myocardial infarction prior to lethal myocardial infarction-like events and display cellular features that differentiate apparent myocardial infarction-induced death from standard postmortem mortality [(157, 158)]. Hence WHHLMIs, while not the ideal animal model (see section 2.5), represent vulnerable plaque from multiple standpoints. Targeting to biochemical elements of such plaques was therefore anticipated to be potentially distinct from targeting to WHHL lesions, in addition to variations in morphology that are visualizable using MRI. Annexin V USPIOs partitioned into plaques and cancelled MR signal in both WHHL and WHHLMI animal models. In the WHHL, signal appeared to cancel throughout most, if not all, of the vascular lumen. This may have been due to a combination of the highly stenotic vessel (leaving a small space through which blood was flowing) and the resolution of the MRI system – i.e., the vessel lumen was sufficiently small so that, given the resolution, MRI imaged the lumenal region as cancelled signal. Conversely, WHHLMI rabbits displayed imaging only within the aortic walls. These observations have relevance to morphological differentiation between the two models

97 (discussed below). Controls demonstrated no signal cancellation – e.g., NZW rabbits injected with Annexin V USPIOs displayed no signal cancellation, implying that signal reduction was not due to flow/motion or other MR artifacts, which was corroborated by histology; protein G USPIO controls also caused no MR signal reduction, indicating that non-specific USPIO binding did not cause cancellation of signal. Thus, because Annexin V USPIOs were proven bioactive (i.e., bound to PS in vitro) and localized especially to areas of apoptotic foamy macrophages but not other areas as demonstrated by histology and electron micrsocopy, the implication is that the USPIOs were directed to these sites by the biochemical affinity of Annexin V to PS (and Annexin V avidity to PS likely modulates contrast effects of the USPIOs on the MR image). The ROI analysis (see Table 2.3) illustrates that WHHLMIs displayed more signal cancellation than WHHLs when given Annexin V USPIOs, and both showed clear signal reduction (pre-injection in comparison to post-injection) compared to controls. This suggested that USPIOs bound to the greater quantity of apoptotic macrophages that are resident within WHHLMI plaques than WHHLs as anticipated, because WHHLs display correspondingly fewer macrophages/apoptotic macrophages [(100, 107)]. Images of WHHLs and WHHLMIs (see Figures 2.13 and 2.14) suggest that Annexin V USPIOs penetrate more deeply into WHHLMI plaques (in Figure 2.13, Prussian blue stained cells are observed very near the WHHL lumen; in Figure 2.14, the Prussian blue stained cells are far from the WHHLMI lumen, indeed, they are near the outer edge of the vessel by the adventia). Assuming that the route of USPIO entry into plaque is equivalent in both animal models and was via the arterial lumen, it is hypothesized that, due to macrophage apoptosis and other dynamic processes in the imaged lesions, enzymes such as MMPs and cathepsins may have induced sufficiently greater permeability in WHHLMI plaques compared to WHHLs. If true, this would indicate that the USPIO size prevents entry into normal vasculature and more stable plaque phenotypes. The addition of the targeting reagent linked to the USPIOs then would yield a dual mechanism by which USPIOs target vulnerability – size-based (due to ECM permeability) and molecular biochemical tethering. If the route of entry were instead via vasa vasorum (partially or completely), the USPIO depth disparity may be a function of the specific vessel networks in these particular plaques.

98 In both animal models, observed signal cancellation was delimited to regions of abdominal aorta distal to the renal branching. While this region was focused upon due to the desire to minimize motion artifacts, another potential reason MR signal cancellation was observed there as compared to, for instance, the coronary arteries is the cellular level difference between the two vascular sites in the animal models; while the Watanabe aorta exhibits relatively numerous macrophages and few smooth muscle cells, the coronary arteries display the opposite [(155)]. Therefore, the demonstrated ability to detect apoptotic macrophages may be of diminished utility in the coronaries, at least for this animal model, while it is optimal in the aorta. It had previously been presumed that morphological characterization (e.g., lumenal stenosis) of an atherosclerotic lesion was sufficient to characterize it as vulnerable or not [(260)]. However, recent information suggests that cardiovascular events might be better predicted by inflammatory indicators including macrophage recruitment, formation into foam cells, metabolism, and apoptosis, as well as MMP activity [(260, 261)]. Nevertheless, it is most likely that a combination of biochemical and morphological information will prove best able to predict the likelihood of plaque rupture [(262, 263)], which is the ultimate goal. Proof-of-principle of the simultaneous delineation of the two (biochemical and morphological) was a goal of the present studies. Because of the diversity of signaling, recruitment, and other biochemical cues inherent in atherosclerosis, multimodal analysis may additionally require correlation of multiple biochemical markers of vulnerability in a lesion, which may be made possible by the multiple distinct targeting reagents that caused MR signal cancellation beside Annexin V (e.g., anti-VCAM-1, anti-collagen Type I, see below). The use of MR imaging for multimodal evaluation of degree/type of stenosis and plaque composition is potentially powerful for predicting future vascular events [(84)]. Current procedures, including lumenography methods, identify morphological features but not biochemical data, and radiological methods, while they can extract biochemical data can seldom give morphological data. The method described herein appears to provide advantages by offering both. While other methods currently used to identify the composition of plaques include invasive, catheter-based procedures such as recent developments in near infrared spectroscopy, angiography, and intravascular ultrasound

99 [(264)], they are not conducive to screening healthy populations. However, though non- invasive modalities such as CT, MRI, ultrasound, and optical coherence tomography can identify atherosclerosis on the basis of stenosis, quantitative lipid core content, and other macroscopic plaque features, they can not detect biochemical inflammatory markers. On the other hand, while nuclear imaging systems such as positron emission tomography (PET) and SPECT studies can demonstrate inflammatory and other vulnerable plaque markers, they are poor modalities for characterizing plaque morphologies [(93)] as they do not have the necessary high resolution MRI possesses [(207)] for differentiating between plaque morphologies (as has been shown herein). As described above, the thin, fibrous-capped WHHLMI plaque (compatible with an animal that spontaneously undergoes myocardial infarction) is associated in the WHHLMI images with a periannular distribution of signal cancellation maintaining a vulnerable morphology. Correspondingly, the obstructive, stable WHHL plaque is perceived by reference to the volume filling signal cancellation and/or the observed plaque thickness in WHHL images of the vessel. Given the evident need for a multimodal approach, MRI appears ideal to integrate high resolution morphological characterization with targeted contrast (i.e., safe and biodegradable USPIOs such as those described) to identify morphologically vulnerable regions that present biochemical indicators of vulnerability. Biochemical information within a lesion extracted via targeting is valuable. However, corresponding morphological analysis is also significant in demonstrating a plaque’s vulnerability. For instance, in 70% of human thrombotic incidents the clots responsible are known to be localized to sites of < 50% occlusion [(207)]. Thus, the ability to characterize the compositional and morphological features of WHHLMI aortae is critical in view of their similarities to human lesions [(157)]. As atherosclerosis is a chronic disease, the ability to monitor atherosclerotic lesions repetitively with a minimum of invasiveness is a critical aspect of the present study with respect to its clinical relevance in screening and monitoring therapeutic efficacy; this capability is displayed with repeated, nearly identical rounds of Annexin V USPIO injection and imaging. In particular, the localization of Annexin V USPIO binding in WHHLMIs correlated with plaque evaluated to be morphologically vulnerable (as defined in [(113)]). While WHHLMI plaques did not possess all morphological parameters of vulnerability,

100 lack of lumenal stenosis and other Type IV plaque characteristics observed led to the assessment of a morphologically vulnerable lesion. While plaque classification schemes are generally based on morphological parameters, morphology is critically affected by cellular processes, which in turn are shaped by molecular developments [59]; plaque composition thus influences the morphology. Because a major cellular marker of unstable disease is apoptosis (associated with molecular PS externalization), the spatial localization of PS-binding USPIOs can consequently be linked to that region’s morphology – this is visualizable as the signal cancellation observed on the MR images (in association with Prussian blue histological staining of those regions). Therefore, while signal cancellation will draw radiologist attention to the site due to biochemical processes of vulnerability, morphological analysis of that site can proceed independently. Biochemical and morphological signs of vulnerability can thus be detected autonomously (i.e., correlation of both areas of signal cancellation and morphological indicators) and synergistically (morphological analysis will only take place at sites of biochemically- driven signal cancellation). The feasibility of simultaneously extracting biochemical and morphological data from atherosclerotic lesions in a single minimally invasive imaging procedure is thus established in this work. According to Kolodogie et. al., the “development of future treatments targeted against plaque instability is contingent upon our ability to confidently recognise precursor lesions likely to thrombose [i.e., vulnerable plaque] [(100)].” Thus, the goal of the present work, a screening procedure, may have high impact on patient morbidity and mortality by influencing the timing, form, and continuation of therapy. Depending on future developments, it is anticipated that a broadly applicable screening procedure to identify vulnerable plaques and monitor them during therapeutic regimens could ultimately become available from the technology. The results presented here (in addition to those of Lanza et. al. [(245, 265)]) suggest that, contrary to the published assertion that “conventional imaging techniques, such as … magnetic resonance imaging cannot image the molecular processes within the atherosclerotic plaques … [(206)],” incorporation of molecular targeting agents such as Annexin V in USPIO platforms allows visualization of biochemical plaque constituents (e.g., phosphatidylserine) by MRI. Indeed, this study demonstrated a repeatable,

101 minimally invasive, potentially multimodal approach (biochemical and morphological) toward a screening method capable of differentiating between vulnerable and stable atherosclerotic plaque using Annexin V USPIOs and MRI. The following section discusses implications of other targeted USPIOs.

2.4.2 Other Targets In order to diagnose a particular vulnerable plaque, concurrence of multiple plaque biomarkers may be necessary in addition to correlation with morphology. In this study three other targeted USPIOs clearly revealed signal cancellation: anti-VCAM-1, anti-collagen type I, and anti-cd-11b (Figures 2.20-2.22). Anti-cd-11b was injected into two different WHHLMI rabbits, causing signal cancellation in both – Figure 2.22 displays the signal cancellation observed in the first rabbit tested. Some other targeting reagents tested (i.e., those in Table 2.1 which have not been presented in above figures) may have been responsible for signal cancellation, but further analysis is necessary. Though more complete investigations must be performed for anti-VCAM-1, anti-collagen type I, and anti-cd-11b (parallel to the Annexin V USPIO study, including examination of biofunctionality and histology), the initial images indicate that these targeted USPIOs direct contrast to plaque sites. The significance must be borne out through future work, but these investigations and their initial parallels to the Annexin V USPIO studies provide the impetus to continue studies with these reagents in order to furnish a catalogue of targeting reagents appropriate for defining the vulnerability state of a particular lesion. In order to help determine the implications of signal cancellation at specific plaque sites, an experiment might be configured for sequential administration and imaging with all targeted USPIOs which have been found to generate signal reduction in WHHLMIs. This experiment would have to be optimized to allow for the biodegradation/decay of contrast after each administration and imaging of targeting reagents. This would provide insight into whether similar regions of plaque are of interest according to the location and intensity of MR signal reduction caused by various targeted USPIOs.

102 2.5 Experimental Limitations While it is clear that the mass of iron normally employed (0.05 mg iron of Annexin V USPIOs) consistently produced signal cancellation with MRI, it is unclear whether all available PS was bound in the vasculature. The 6X dose acceleration experiment (6 times the standard amount Annexin V USPIOs injected) in WHHLMIs provided additional data, though it was not optimal – it can not conclusively demonstrate that more iron was delivered to plaque based on the increased signal cancellation associated with the increase in injected iron. Even though six times more iron appeared to cause an increase in quantifiable (ROI) signal reduction, the two intervening months between the 1X and 6X injections make it feasible that plaque dynamics changed – e.g., perhaps available PS binding sites increased. Thus, the experimental set-up must be modified if it is desired to test whether increasing the dose yields greater signal reduction (which, if it is demonstrated, would suggest that not all available PS is bound using the 1X dose). An experiment to examine this question might include injecting a single dose (0.05 mg Annexin V USPIOs) and MR imaging, then immediately injecting a 5X dose and imaging. If an increase in signal cancellation between the 1X and 5X doses were observed, this would indicate more Annexin V USPIOs bound the plaque. Of further interest, in the coronal view, is if this 5X injection also causes other areas of plaque to produce MR signal cancellation, and then to examine histologically whether the cancellation is caused by Annexin V-PS interaction (i.e., apoptotic cells). Moreover, it may be of interest to explore the saturation dose limits of Annexin V USPIOs in the WHHLMI and WHHL vasculatures. By determining the saturating doses, data may be obtained that explores the temporal framework of Annexin V USPIO binding to PS as a function of plaque location (i.e., where the USPIOs direct contrast first, until saturation is achieved). Further, such a study may present additional opportunities to study the differences between stable and unstable-type lesions (WHHL versus WHHLMI). The ability of Annexin V to bind PS is not limited to apoptotic cells, as observed in the Annexin V biofunctionality assay in the pre-incubation in which Annexin V bound free PS. Thus, Annexin V USPIOs might detect any cell exposing PS – in addition to apoptotic cells, necrotic cells may also present available PS. Necrosis is due to acute cellular injury, rather than the programmed fate of apoptotic cells. Apoptotic cells can be

103 differentiated from necrotic cells by staining with TUNEL and Hoechst. Thus, while it is presently unknown how much the presence of necrotic cells may decrease the specificity of the Annexin V USPIOs in the present application, visual analysis of histological stains indicated only that Annexin V USPIOs co-localized with apoptotic cells in animal plaques. It is unknown how Annexin V USPIOs would perform in a system that contained many necrotic cells, but this potential effect should be examined. The USPIOs used as controls in this study could be significantly improved. For Annexin V USPIOs, a mutated version of the Annexin V protein which does not bind phosphatidylserine would be a better control than the similarly sized, but otherwise disparate protein G molecule. The charge, geometry, and other parameters could vary between Annexin V and protein G, making protein G a less desirable control; a mutated Annexin V variant would be nearly identical to Annexin V in amino acid sequence and would thus model non-specific Annexin V USPIO binding and distribution better than protein G. Moreover, the difference between the number of Annexin V proteins compared to the number of protein G molecules per USPIO could impact in vivo distributions – proteins per USPIO should, on average, be identical between experimental samples and controls. Otherwise, differences in apparent size and charge could account for earlier uptake by the RES, for instance. For antibody conjugated targeting reagents, it would be desirable to use an isotype-matched antibody as the control. Thus, future work might include both Annexin V variants and isotype controls and use of approximately the same number of proteins per particle. For proof-of-principle, it was reasonable to focus upon the abdominal aorta in this study. However, to extend the work to general utility to detect plaques for the entire vasculature, cardiac and respiratory gating protocols must be developed and instituted to minimize motion artifacts. By synchronizing physiological movements with MR data acquisition, important vascular regions such as the coronary arteries might also be examined using the present techniques. While hypotheses have been formed, the kinetics and mechanism of the loss of signal cancellation from plaques remain unknown. Experiments indicate that contrast has decayed within one day, but further studies are needed to confirm and refine this time frame (e.g., by imaging every two or three hours after injection of targeted USPIOs).

104 Previous work indicates that USPIOs degrade within a week, so if contrast decays in less than a day in our studies, the mechanism involved must be ascertained. Once more data has been gained via imaging and histology, it will be possible to design the appropriate experiments to determine the mechanism (e.g., if histology begins to indicate that cellular motility causes USPIO-bound cells to migrate out of the plaque by viewing various time points, this might be studied in further detail by using higher resolution MRI and visualizing cell movement or by moving into mouse models and using genetic knockouts). SPECT imaging of 99mTc tagged Annexin V USPIOs showed that the majority of particles are dispersed in organs other than vasculature (e.g., liver, spleen, bladder, kidneys). It will be valuable to optimize USPIO parameters (such as size, iron content in each USPIO, charge, and surface material and chemistry) in order to maximize USPIO localization to plaque. Such a process may enable even lesser quantities of targeted USPIOs to be injected to detect potentially unstable lesions. Furthermore, if future iterations include a therapeutic, this will ensure that a maximum amount of therapeutic is delivered to the desired site rather than taken up by RES organs, where the therapeutic may be toxic. However, care must be taken with USPIO property modulation, as some of the plaque binding effects observed in this chapter may be due to those properties – for instance, it is feasible that the size of the USPIOs prevents penetration into normal vasculature and stable plaque, but when vulnerable lesions develop increased permeability, the USPIOs may more easily penetrate. Perhaps the most significant broad experimental limitation is that while it has been shown that Annexin V USPIOs can unequivocally and repeatedly be delivered to plaque sites in animal models, no connection has yet been made to the likelihood of that lesion to rupture or erode. Ideally, this technique will enable assignment of an imaged plaque’s probability of rupture or erosion – i.e., an event which begets thrombosis. As discussed, this will most likely be defined through concordance between multiple targeted USPIOs (using diverse targeting reagents) and MR-based morphological characterization, but an animal model in which myocardial infarctions (and the associated culprit lesions) can be identified is necessary to afford the required experimental framework.

105 In this mode, even though WHHL and WHHLMI rabbits are relevant models for stable and unstable plaques, respectively, they are not ideal. It is likely that an animal model such as the ApoE -/- mouse must instead be employed. Rupture and thrombus are consistently identifiable in these mice [(146)], so a probability of rupture could experimentally be determined. The size of these animals brings new challenges, such as MR imaging of the minute vessels and other issues mentioned in section 1.2.5, but suitable MRI equipment and techniques have been developed to manage some such obstacles.

106

CHAPTER 3

BREAST CANCER

3. Breast Cancer

3.1 Introduction

While CAD and breast cancer are distinct disease states, nanoparticulate platform modularity [(266)] facilitated use of the CAD platform for application in the amplification of malignant tissue mechanical properties for detection by ultrasound. The identical nanoparticulate platform was used for both CAD and breast cancer: however, by abstraction of the CAD-associated antibody and replacement with a breast cancer- targeted substitute, the platform was transformed into a tool for breast cancer detection.

The platform iron oxide nanoparticles were constructed for targeting to the prognostic

(i.e., a factor predictive of untreated patients having divergent outcomes) and predictive

(i.e., of therapeutic treatment response) biomarker protein Her-2/neu using a monoclonal antibody (mAb).

The objective of the study was to develop and integrate the technological elements necessary to evaluate breast tissue ex vivo. Non-destructive evaluation (NDE) ultrasound was employed as the imaging modality for this project. The detection modality branch of this chapter is founded, and builds upon, Liu et. al.’s initial results at

The Ohio State University that indicate that the malignancy status of tissue can be

107 established using reconstructed tissue mechanical properties via a mathematical analysis of ultrasonic reflection spectra [(200-202)]. These previous studies suggested that mechanical parameters of tissue are detectably modified by the presence of cancer.

The present studies aimed to demonstrate that amplification of the mechanical differences between malignant and normal tissue is feasible through the application of molecularly targeted nanoparticles and the modification of previously employed mechanical model algorithms. Given the novelty of the approach with respect to biological systems, the major experimental obstacle comprised devising a robust method capable of uniting the diverse biological/technological properties implicit in multidisciplinary endeavors: features involving the targeting of particles to biological materials were modified for ultrasonic investigations, the form of the experimental sample itself was iteratively developed to take into account the requirements inherent in facilitating mutual biological and ultrasonic evaluation viability, and bulk versus surface particulate physical properties were considered. In addition to preliminary ultrasound studies designed to test experimental materials and formats, experiments were performed on cells using bulk distribution, and human breast tissue using the surface distribution model. A subset of the NDE ultrasound methods, characterization mode ultrasound

(CMUS) was utilized for bulk (cells), while C-scan ultrasound was employed for the surface tissue distribution as it was shown that CMUS could not detect the presence of superficial particles on tissue.

This chapter will describe the course of these investigations beginning with the development of initial protocols that served to optimize tissue models. Subsequent sections will convey particulate targeting for cells and tissue via the creation of tailored

108 immunohistochemical methods through characterization and optimization with ultrasound. Permission for this study was granted by The Ohio State University Internal

Review Board to acquire and study human tissue under research protocol # 2003C0034, and all institutional and federal protocols and safety requirements were observed.

3.2 Experimental

The experimental design was driven by the necessity to integrate nanoparticles onto a biological specimen that would be submersed in a fluid medium. The need for submersion was driven by the necessity for a coupling agent to aid in the transmission, reflection, and propagation of ultrasonic waves. Because water was used as the coupling agent (a conventional choice in research laboratories, rather than the coupling gel often used in medical settings), the specimen comprised a water-proof sandwich structure of two optically flat glass plates on either side of the specimen with an epoxy seal between the pieces of glass. This design enabled a pocket within which the nanoparticulate and biological features of the experiment could be performed, and then sealed off for ultrasound interrogation. An additional benefit was affording an enhanced impedance mismatch between glass and specimen for robust, detectable reflections as compared with the more weakly differentiable reflections produced between tissue/tissue-like materials and water. Experiments commenced by selecting agarose and PDMS (poly dimethyl siloxane) for the creation of tissue phantoms. These materials were selected because they do not damage biological materials and their mechanical properties approximate those of real tissue. The models were used to glean information about the effects of nanoparticle material, concentration, and distribution (surface versus bulk) on ultrasound reflections

109 while nanoparticle bioconjugation, separation, and targeting methods were being developed and optimized.

3.2.1 Tissue Phantoms

PDMS Phantom: Surface Particle Distribution

Gold and iron oxide nanoparticles and silica microparticles were tested to determine which particulate material to employ in subsequent ultrasound studies involving cells and biological tissue. A tissue phantom model was therefore created to imitate the accumulation and binding of particles on the tissue surfaces. The phantom material, a 50:1 mixture of poly (dimethyl siloxane) (PDMS (Sylgard 184, Dow Corning

Corporation, Midland, MI)), and cross-linker, was sandwiched between glass plates in a

1” by 2” mold made of 150 µm thick tape (1506CW vinyl seal, Venture Tape

Corporation., Rockland, MA). The PDMS was pipetted into the mold on the bottom plate and allowed to cure for about 36 hours after placing a transparency sheet and second glass plate on top. Post-cure, the top glass plate and transparency sheet were removed, and various concentrations of the three particle types were evaporated onto the PDMS.

Particles were concentrated using magnetic separation or centrifugation as discussed in detail in Chapter 4. Using a top glass plate identical to the bottom replete with PDMS mold (such that the particles were sealed between the two PDMS molds), a sandwich structure was subsequently reformed and sealed with epoxy (Loctite, Henkel

Technologies, Rocky Hill, CT) so that the samples could be tested in the aqueous testing environment. After curing, the sample was ready for ultrasound interrogation.

110

Agarose Phantom: Bulk Particle Distribution

Agarose was used to probe the effects of nanoparticle concentration (in a bulk distribution) on ultrasound response. A tissue phantom was produced with a 1% agarose

(USB Corporation, Cleveland, OH) solution for realistic tissue imitation as the density was known to be similar to that of theoretical human breast tissue (approximately 1.02 g/ml to 0.96 g/ml, respectively). Each glass plate sample contained four testing conditions (molds of 150 µm thick tape) of 20 nm diameter colloidal gold nanoparticles

(Ted Pella, Redding, CA) distributed throughout the agarose solutions. Particles were concentrated by centrifugation for 7 minutes at 1300 g in order to generate four concentrations to be tested: 0, 7.0 X 1012, 7.0 X 1013, and 7.0 X 1014 particles/ml.

The PBS solutions containing nanoparticulates were used as the solvents for agarose solutions, which were then pipetted into tape molds on the glass. A bare glass top was firmly placed (to rid agarose of bubbles) over the bottom glass substrate containing agarose molds. The glass sandwich structure was sealed with epoxy and the agarose was allowed to solidify prior to submitting the sample to ultrasound.

3.2.2 Cell Line Studies

Targeted iron oxide nanoparticles were created by linking a monoclonal antibody

(mAb) to Her-2/neu and a control isotype-matched antibody to protein G molecules that were covalently bound to the particles’ surface. In vitro studies on cell lines were performed primarily to verify successful linkage of antibodies to nanoparticles

111 (conjugation and purification techniques are described in section 4.1.1) in viable, biofunctional form with no adverse effects on the nanoparticles. Samples created from nanoparticle-bound cells mixed with agarose were used to analyze the ability of the

CMUS system to 1. detect the presence of nanoparticles in a biological system and 2. to differentiate between various degrees of particle-cell binding. SKBR-3 and MDA-231 human breast cancer cell lines (American Type Culture Collection (ATCC), Manassas,

VA) were chosen to represent Her-2/neu positive and negative tissues, respectively.

SKBR-3 cells simulate breast cancer tissue IHC scores of 3+ (which express about 2.4 million Her-2/neu receptors/cell), while MDA-231 cells represent scores of 0,

(approximately 22,000 Her-2/neu receptors/cell) [(187, 267)]. SKBR-3 cells were employed for both preliminary and CMUS studies. However, due principally to the space constraints of ultrasound samples, MDA-231 cells were used only as a control in preliminary targeting assay optimization work.

In conjunction with studies that demonstrated the ability of ultrasound to detect signal change due to gold nanoparticles on PDMS tissue phantoms (see section 3.3), the viability of targeting gold nanoparticles was also studied using an identical cell culture and targeting protocol and characterization methods (e.g., flow cytometry) given below for SKBR-3 cells. The labeling and purification of the anti-Her-2/neu mAb and isotype- matched antibody to gold via protein A is delineated in Chapter 4. Flow cytometry was performed to verify binding of particles to cells.

The mAb Herceptin was purchased from Genentech (South San Francisco, CA) as it is a rigorously tested, clinically used antibody to the Her-2/neu antigen for cancer

112 therapy. An isotype matched control antibody was obtained from the Laboratory of

William Carson (The Ohio State University, Columbus, Ohio).

Cell Culture

SKBR-3 and MDA cells were maintained in RPMI-1640 (Cambrex, East

Rutherford, New Jersey) supplemented with 10% fetal bovine serum and 1% antimycotic in a 5% CO2 humidified atmosphere at 37 °C. After cracking frozen cells, they were handled according to manufacturers’ directions, feeding every other day in a T-175 flask.

Cell Targeting

When the cells were greater than 70% confluent (generally 5-6 days after culture commenced) media was aspirated and they were trypsinized for removal from the bottom of the flask. Trypsin was neutralized using serum containing media and the cells were pelleted at 1200 rpm for 8 minutes. Media was aspirated and cells were resuspended with about 2 ml media and then counted on a hemocytometer to determine cell concentration.

Approximately 5 X 106 to 1 X 107 cells were added to each of three flow tube conditions

(a fourth condition was maintained cell-free). If unused cells remained, they were treated using a standard freezing protocol using DMSO (dimethyl sulfoxide, Sigma) and frozen down at -130 °C. After pelleting the cells in flow tubes at 1200 rpm for 8 minutes, media was aspirated, and the cells were resuspended in 100 µl cold flow buffer (4 °C; PBS + 5%

Filtered FBS). Subsequently, primary antibody (Herceptin), nanoparticulate conjugates of primary antibody, and nanoparticulate conjugates of isotype-matched antibody (see section 4.1.1 for bioconjugation protocols; 10 µg antibody was used for each sample,

113 whether isolated antibody or antibody conjugate) were added to individual flow tubes.

The four testing conditions consisted of Herceptin alone (Her), Herceptin-conjugated nanoparticles (Her-NP), isotype-conjugated nanoparticles (Iso-NP), and cold flow buffer

(CFB). Tubes were incubated on ice for 30 minutes, after which cells were washed with

2 ml cold flow buffer and spun to pellet under the same conditions. Supernate was aspirated, and then the unbound rings of rust-colored nanoparticles (i.e., nanoparticles which had not bound to cells nor remained in the supernate) around the tubes at the interface between cells and residual supernatant was rinsed and aspirated. Care was taken not to disturb the pellets, and the pellets of all conditions were rinsed equivalently regardless whether or not a ring was visible. Cells were then resuspended in 50 µl CFB.

An optional step of adding 7 µl of 2 mg/ml normal goat IgG serum and incubating tubes on ice for 10 minutes to block was sometimes added (but found to be unnecessary according to flow data). Next, in the dark (note: all subsequent steps were performed in the dark), secondary antibody (mouse anti-human FITC κ light chain, Biosource

International, Camarillo, California) was added to the flow tubes at 50 µl per flow tube of a 1:50 dilution in CFB. Cells were incubated on ice for 30 minutes, washed with 2 ml cold flow buffer, and spun to pellet. After the supernate was aspirated, untargeted nanoparticle rings were again rinsed and aspirated. Next, if cells were to be viewed by immunohistochemistry, 50 µl cold flow buffer were added, and cells were sealed beneath a cover slip on slides. They were then viewed using fluorescence microscopy. On the other hand, if cells were to be used for CMUS interrogation, 500 µl cold flow buffer was added to each tube. The tubes were then subjected to analysis by flow cytometry using a

114 BD FACSCalibur (Becton Dickinson, Franklin Lakes, NJ) to verify particle-cell binding by fluorescent methods.

Cell-agarose Matrix Preparation

The particle- and antibody-bound cells were next prepared for ultrasound interrogation. Equal 250 µl volumes of pre-warmed (4-5 minutes at 45 °C in a water bath) cell solution (in cold flow buffer) and agarose solution (20 mM Hepes, 150 mM

NaCl, 3.5% agarose by weight) were pipetted into warmed flow tubes. Glass plates, one of which (the “bottom”) had been configured to accommodate four samples within square tape molds, were warmed in an oven to 55 °C. After a 2-3 minute equilibration for the cell-agarose solutions in the water bath, the four samples were gently mixed, then rapidly pipetted into the four tape molds on the bottom glass sample plate to accommodate the four test conditions – Her-NP, Iso-NP, Her, and CFB. Glass plates and samples were maintained above 37 °C to prevent premature solidification of agarose.

The sandwich structure was then created in a manner comparable to the agarose phantom protocol using a clean glass top plate. Care was taken affixing the top plate so that no bubbles were left within the molds; the tape was simultaneously used as an accurate 150

µm spacer (to regulate sample thickness) and mold. Furthermore, this technique preserved the cells, allowing them to be evaluated by ultrasound.

3.2.3 Tissue

Tissue Procurement centers from The Ohio State University and the Lombardi

Comprehensive Cancer Center of Georgetown University provided human breast cancer

115 tissue biopsies from a number of female individuals of varying age and race. Tissue Her-

2/neu expression IHC scores spanned the spectrum from 0 to 3+ values as evaluated by

Georgetown and an experienced OSU pathologist. Only the final, optimized protocol for tissue ultrasound interrogation is presented here. The cubic centimeter (on average) volume tissue sections were snap-frozen and cut by cryostat microtome into 7 µm thick sections. Sections were placed onto standard glass microscope slides. The protocol used above for staining cells was adapted for tissue: freshly cut slides were hydrated and washed for 10 minutes in buffer (PBS, 1.5% BSA, and 1:200 PMSF). No fixation step was employed due to incompatibilities with the preparation for ultrasound. The tissue was then washed three times in separate buffer containers for five minutes per wash.

Next, the slides were incubated for 30 minutes in 5% goat serum diluted in buffer. The slides were incubated with primary antibody (Herceptin, Herceptin-conjugate, isotype- matched antibody, or isotype-matched-conjugate) for 1.5-2 hours in a moist enclosure, after which they were washed once in buffer. Next, if images of the tissue were desired, secondary antibodies with a fluorescein tag were incubated with specimens for 45 minutes. After washing in buffer, slides were arrayed onto a glass plate by firmly fitting four slides together side by side. A glass plate with a 450 µm layer of agarose was subsequently placed over the glass slide array to form a sandwich structure, with care taken to ensure a lack of air bubbles between tissue and agarose. This assembly was sealed with epoxy to facilitate ensuing ultrasound interrogation.

116 Statistical Analysis

Analysis of variance (ANOVA) and the Tukey Honestly Significantly Different

(HSD) Post Hoc Test were performed for density and stiffness on agarose tissue phantom and cell specimen data and for doublet mechanical reconstructions. These analyses are summarized in Appendix A (see [(268)] for complete analyses).

3.2.4 Ultrasound Systems and Theory

CMUS and C-scan modes were both used in this investigation. As the viability of

CMUS for interrogation of biological tissue was previously unreported, preliminary experiments sought to ascertain the system’s capability in this field. However, since

CMUS was unable to discriminate between interfacial surface particle targeting conditions for tissue studies, the C-Scan mode was also assessed. Thus CMUS was used for tissue phantoms and agarose, while C-scan was utilized for tissue investigations.

Characterization Mode Ultrasound

The CMUS system as employed herein was developed by Dr. Stanislav Rokhlin’s laboratory at The Ohio State University to evaluate industrial materials through the unconventional non-destructive evaluation (NDE) technique. Dr. Jun Liu of The Ohio

State University recently pioneered use of this system for the assessment of biological tissues. The present studies sought to improve upon Dr. Liu’s work by modification and enhancement of several parameters. First, targeted nanoparticulates were developed to enhance the mechanical differentiation of tissue by leveraging the molecular-level variations between tissues to amplify the mechanical properties. Furthermore,

117 algorithmic adjustments were implemented for the reconstruction of continuum and nanomechanical parameters. While Dr. Liu reconstructed six non-dimensional continuum mechanical parameters and seven nanomechanical parameters from a single oblique angle of incidence, the current approach employed continuum (developed by Dr.

Rokhlin’s lab) and nanomechanical models founded upon two angles of incidence. An advantage of the nano-, or doublet, mechanical model was its scalability: that is, reconstructed microstructural tissue information was collected that would have been lost using continuum mechanical approaches.

While CMUS is unconventional, the testing procedure was similar to standard medical ultrasound procedures. As previously mentioned, the sample must be submerged in the water bath for acoustic coupling to facilitate the transmission of ultrasound waves.

These waves are produced by a pulser (Panametrics, 5052PR, Waltham, MA) and transmitted by unfocused broad band transducers (Panametrics, central frequency 10

MHz, 5” diameter). The mechanical waves interact with various constituents of the experimental sample (i.e., the glass sandwich structure), such as the top piece of glass, the cells or tissue phantom, and the bottom piece of glass, and their reflections are collected by the receiver through the transducers (Panametrics, 5052R). The mechanical reflections are converted into electrical pulses, which are then coded digitally by a waveform digitizer (SignaTec, PDA500M, Corona, California). Using a computer- controlled X,Y coordinate translation mechanism, 441 discrete time domain plots were produced per testing condition for continuum mechanical analysis (40 points were manually acquired for doublet mechanics).

118 Previous studies established that use of a single angle of incidence does not lead to accurate decoupling of the six unitless parameters of the ultrasonic signals [(205)]. Dr.

Rokhlin’s lab demonstrated the feasibility of simultaneously resolving all the properties of the layer from the normal and oblique angles: density, longitudinal and shear moduli, thickness, and attenuation using ultrasound [(204)]. The mathematics and mechanics of this continuum method are described by the Rokhlin lab [(204, 205)].

The scalability of continuum mechanics is problematic when approaching discrete, minute levels. Indeed, a pragmatic approximation of the behavior of macroscopic structures is provided by continuum mechanics, while quantum mechanics is eminently descriptive of matter on the atomic scale and below. The modern study of nanoscale structures has dictated a mechanical description bridging continuum and quantum mechanics. A mesomechanical theory, i.e., nanomechanics or Doublet

Mechanics (DM), was created to study structures from the mid-nanoscale through the microscale by revealing a crucial scaling factor. DM, for instance, resolved microelastic parameters that were unachievable through continuum mechanics and would require unattainable amounts of computing power through quantum mechanics [(200)]. DM’s scalability enables accurate recovery of continuum parameters from micro-

/nanomechanical parameters [(199, 203)]. The mathematical description of DM for biology is beyond the scope of this dissertation, but is reviewed elsewhere [(202)] as is its application in the enhancement of nanoparticle-augmented CMUS methods [(268)].

119 C-Scan

C-Scan Ultrasound (Panametrics, Multiscan Inspection System) is another NDE ultrasound modality which is commonly used to examine defects within composite materials. Due to the inability of CMUS to facilitate nanoparticle detection in tissue, C-

Scan was employed for tissue analyses. The shift to C-Scan necessitated reconfiguration and revalidation of the tissue sample and protocol in the form described in section 3.2.3

(previous CMUS protocols remain undescribed in this text). In the reconfiguration process, the most critical issues that arose concerned the thicknesses of the tissue sections and the agarose layer, the procurement and consistent IHC scoring of tissue samples, and most significantly the development of a protocol that combined the targeting of nanoparticles to tissue with the constraints implicit in the preparation of tissue samples for ultrasound interrogation. It was anticipated that the application of the increased frequency, focused 50 MHz transducer C-Scan system with sub-millimeter spot size compared to the CMUS specifications would produce a signal that enabled detection of nanoparticles on the tissue surfaces. The C-Scan generated digital images of samples including reflection intensities and frequency response.

Modifications to Doublet Mechanical Model

Comprehensive continuum mechanical models have been developed to compute acoustic and geometric properties (e.g., density, longitudinal and shear elastic moduli, layer thickness, and loss factors) of isotropic or anisotropic multilayered solids sandwiched around an embedded isotropic layer [(204, 205)]. For these continuum models both normal and oblique angles of incidence were taken into account, unlike the

120 previous doublet mechanical model that accounted for only the oblique parameter as surveyed in previous work [(200-202)] and described in Chapter 1. This subsection describes improvements made to the DM model. For background and previous work please refer to Chapter 1 or, if additional background is desired, please refer to the following references [(199-205)].

In order to enhance the DM model’s utility and accuracy, normal-derived parameters were integrated into the previous algorithm by incorporating the relevant parameters from the second inversion step of the parallel continuum model [(205)] into the DM algorithm. Insertion of these parameters into the previous DM code precluded the requirement for a second inversion in the DM algorithm. Addition of the second incident angle has been demonstrated to promote greater accuracy for obtaining the six non-dimensional parameters [(205)]. The updated single inversion DM model for two angles of incidence was adapted from Wang et. al. and a summary is given in Figure 3.1

[(205, 268)].

121 DM Inversion Method

Figure 3.1: A flow chart illustrating the reconstruction of the DM inversion method, based off the work of Wang et. al. [(205)], to generate micro-elastic parameters from two angles of incidence

From the continuum model, the acoustic response of the embedded layer of the sample is dependent on 6 unitless parameters. Thus, for each of the 441 points taken by the ultrasound system per sample, six parameters – elastic moduli (λ and µ), thickness

(h), density (ρ), and longitudinal and shear attenuations (αl,αt) – were calculated as described in Chapter 1 using mathematical relations between the acoustic response parameters and the model of the parameters. For DM reconstruction, λ, µ, h, and ρ were used to compute the DM parameters: microelastic moduli (A11 and A44, the doublet mechanical parallel to the continuum elastic moduli), attenuation factors (iA11 and iA44),

122 and the internodal distance, η. The original DM model was amended to relate the Lamé moduli λ and µ with A11 and A44, the microelastic parameters, as shown.

λ = A11 – A44 (3.1)

µ = ¼ A44 (3.2)

The density ρ was computed by reference to the Rokhlin continuum model [(205)] and the height h was held constant, given by the 150 µm spacers used in the experimental set-up.

3.3 Results

The results of the nanoparticle-targeted breast cancer/ultrasound branch of this dissertation are presented in this section. Figures and tables are numbered by the order in which they are referred to in the text. The data from the two tissue phantoms, PDMS and agarose, are displayed in the first subsection. Gold, iron oxide, and silica particles were studied on the surface of PDMS using C-Scan, while gold particles were examined in agarose in bulk using CMUS. For agarose phantoms, the averages of the mechanical properties are displayed. Next, the results of the SKBR-3/MDA cell line studies, inclusive of IHC micrographs, flow cytometry, and ultrasound-derived data, are illustrated in subsection 3.3.2. Images and ultrasound-derived mechanical data from the human breast tissue experiments are presented in subsection 3.3.3. For detailed explanation of the nanoparticulate facets of the experiment, please refer to Chapter 4; for discussion of results, see section 3.4.

123

3.3.1 Tissue Phantom Studies

3.3.1.1 PDMS

PDMS study: USPIOs

Sample 1: Iron Oxide (High) Ultrasound C-scan. Sample 2: Iron Oxide (Low) Ultrasound C-scan.

Sample 1: Iron Oxide (High) sample photo. Sample 2: Iron Oxide (Low) sample photo.

Figure 3.2: USPIOs on PDMS are detectable by C-Scan.

124 PDMS study: Gold

Sample 5: Gold (High) Ultrasound C-scan Sample 7: Gold (Low) Ultrasound C-scan

Sample 5: Gold (High) sample photo Sample 7: Gold (Low) sample photo

Figure 3.3: Gold is detectable on PDMS using C-Scan.

125 PDMS study: Silica microparticles

Sample 9: Silica (High) Ultrasound C-scan. Sample 10: Silica (Low) Ultrasound C-scan.

Sample 9: Silica (High) sample photo. Sample 10: Silica (Low) sample photo.

Figure 3.4: Silica microparticles are detectable on PDMS using C-Scan.

PDMS Tissue Phantom

8x10^14 8x10^13

Table 3.1: Experimental particle conditions for PDMS tissue phantom study.

126 3.3.1.2 Agarose

Table 3.2: This table exhibits the relationships between the means of the four particle conditions on a single specimen for each mechanical parameter evaluated by C-Scan. The relationship was evaluated with a score: High = all four conditions for the mechanical parameter of that specimen were related in increasing or decreasing order, Good = three conditions were related in increasing or decreasing order, and Inconclusive = two conditions were related in increasing or decreasing order. Each score was represented with a background and text color combination. High-Low or Low-High indicates the relative order of a mechanical parameter in relation to the High particle condition.

127

Table 3.3: Displays reconstructions of doublet mechanical parameters and their continuum mechanical counterparts for agarose tissue phantoms. Lambda and Mu (DM) were calculated from A11 and A44 doublet mechanical parameters from eqs. 3.1 and 3.2. Lambda and Mu continuum parameters were reconstructed from the two-step inversion algorithm. The percent differences between the calculations are given. The two samples highlighted in purple are outliers and were not included in the averaged values of percent differences.

128

3.3.2 Cell Line Studies

SKBR-3 IHC

Bright field image Fluorescent micrograph

Figure 3.5: Her-2 positive SKBR-3 cells are not appreciably bound by iso ab. Iso/SKBR-3– a bright field image (left) matching fluorescence image (right).

129 SKBR-3 IHC

Bright field image Fluorescent micrograph

Figure 3.6: Her-2 positive SKBR-3 cells are bound by Her ab. Her/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

130 SKBR-3 IHC

Bright field image Fluorescent micrograph

Figure 3.7: Her-2 positive SKBR-3 cells are not appreciably bound by iso-NPs. iso- NP/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

131 SKBR-3 IHC

Bright field image Fluorescent micrograph

Figure 3.8: Her-2 positive SKBR-3 cells are bound by Her-NP about half as well as Her alone due to the polyclonal secondary antibody used. Her-NP/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

Protein G-IgG Schematic

Hypervariable CDR region CDR

Light chain (e.g., κ )

Constant heavy chain 2 Protein G Constant heavy chain 3

Antibody

Figure 3.9: Protein G binds IgG antibodies between the 2nd and 3rd constant regions of the heavy chains.

132 Herceptin-ligand complex

Herceptin Polyclonal secondary antibody: steric hindrance to CH2, CH3 Her-2/neu Iron oxide positive cell nanoparticle Kappa light chain secondary antibody

Figure 3.10: Schematic of a Her-NP nanoparticle bound to a Her-2/neu receptor. To the left, a FITC-tagged secondary can not bind Herceptin toward the constant regions due to steric hindrance; toward the right of the cell, a kappa light chain only secondary binds the kappa light chain of a Herceptin antibody.

133 SKBR-3 IHC

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value HER2-FITC 0 0,1023 6436 54.28 100.00 3.60 3.78 36.15 74,3.16228 1 314,1023 67 0.57 1.04 20.35 23.01 11.80 4,18.7688

Bright field image Fluorescent micrograph

Figure 3.11: Her-2 positive SKBR-3 cells are not appreciably bound by iso ab. Iso/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

134 SKBR-3 IHC Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value HER2-FITC 0 0,1023 9635 55.45 100.00 1.31 1.49 136.38 1930,1 1 314,1023 40 0.23 0.42 110.40 179.07 35.19 2,1596.34

Bright field image Fluorescent micrograph

Figure 3.12: Her-2 positive SKBR-3 cells are not appreciably bound by iso-NPs. Iso- NP/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

135 SKBR-3 IHC

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value HER2-FITC 0 0,1023 6524 53.80 100.00 143.30 116.86 19.17 57,140.746 1 314,1023 6032 49.74 92.46 151.25 144.75 10.53 57,140.746

Bright field image Fluorescent micrograph

Figure 3.13: Her-2 positive SKBR-3 cells are bound Her ab. Her/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

136 SKBR-3 IHC

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value HER2-FITC 0 0,1023 9972 56.10 100.00 67.93 63.90 22.67 83,62.0824 1 314,1023 9126 51.34 91.52 71.69 78.65 15.14 83,62.0824

Bright field image Fluorescent micrograph

Figure 3.14: Her-2 positive SKBR-3 cells are bound by Her-NP about as well as Her alone. Her-NP/SKBR-3 – bright field image (left) matching fluorescence image (right), with associated flow cytometry data (top).

137 MDA cell specimen SKBR-3 cell specimen

aram name M Low,High Events %Total %Gated Median GMean CV Peak,Value Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value FIT 0 0,1023 9297 92.97 100.00 2.59 2.77 53.97 85,2.24679 FIT 0 0,1023 9225 92.25 100.00 2.21 2.34 56.05 147,2.07208 1 325,1023 121 1.21 1.30 25.71 27.00 10.50 5,19.2822 1 325,1023 94 0.94 1.02 29.16 34.50 19.90 5,26.896

iso MDA iso SKBR-3

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value FIT 0 0,1023 9302 93.02 100.00 2.62 2.81 55.84 97,1 FIT 0 0,1023 9267 92.67 100.00 286.44 262.50 11.37 106,327.812 1 325,1023 146 1.46 1.57 35.55 36.98 16.41 4,51.397 1 325,1023 9175 91.75 99.01 289.03 273.60 8.39 106,327.812

Herceptin MDA Herceptin SKBR-3

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value FIT 0 0,1023 9364 93.64 100.00 2.49 2.59 51.01 98,1 FIT 0 0,1023 1627 88.18 100.00 6.44 6.62 37.07 21,6.97831 1 325,1023 21 0.21 0.22 24.58 50.48 42.23 3,20.7208 1 325,1023 64 3.47 3.93 63.78 77.81 20.94 2,196.322

Her-NP SKBR-3 Her-NP MDA

Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value FIT 0 0,1023 9396 93.96 100.00 2.59 2.68 48.86 85,1.92822 1 325,1023 18 0.18 0.19 22.67 30.55 28.99 3,19.2822 Param name M Low,High Events %Total %Gated Median GMean CV Peak,Value FIT 0 0,1023 9181 91.81 100.00 2.10 2.23 61.05 113,1.98096 iso-NP MDA 1 325,1023 37 0.37 0.40 26.90 95.18 42.74 3,18.6008 iso-NP SKBR-3

Figure 3.15: Her-2/neu negative MDA Figure 3.16: Her-2/neu positive SKBR-3 cells treated with the labeled antibody cells treated as labeled, with the same lots of and particle conjugate conditions. antibody and particle conjugates as the MDA cells in Figure 3.15.

138 SKBR-3 cell specimen I

Figure 3.17: Isotype-NPs do not appreciably bind SKBR-3 cells. Flow cytometry data for cell specimen I, isotype-matched antibody conjugated nanoparticle (iso-NP) condition.

139 SKBR-3 cell specimen I

Figure 3.18: Her-NP binds SKBR-3 cells. Flow cytometry data for cell specimen I, Herceptin conjugated nanoparticle (Her-NP) condition.

140 SKBR-3 cell specimen II

Figure 3.19: Her binds SKBR-3 cells. Flow cytometry data for cell specimen II, Herceptin (Her) condition.

141 SKBR-3 cell specimen II

Figure 3.20: Iso-NP does not appreciably bind SKBR-3 cells. Flow cytometry data for cell specimen II, isotype-matched antibody conjugated nanoparticle (iso-NP) condition.

142 SKBR-3 cell specimen II

Figure 3.21: Her-NP binds SKBR-3 cells. Flow cytometry data for cell specimen II, Herceptin conjugated nanoparticle (Her-NP) condition

143 SKBR-3 cell specimen III

Figure 3.22: Secondary does not appreciably bind SKBR-3 cells. Flow cytometry data cell specimen III, cells tagged with fluorescent secondary antibody condition.

144 SKBR-3 cell specimen III

Figure 3.23: Her binds SKBR-3 cells. Flow cytometry data for cell specimen III, Herceptin (Her) condition.

145 SKBR-3 cell specimen III

Figure 3.24: Iso-NP does not appreciably bind SKBR-3 cells. Flow cytometry information for cell specimen III, isotype-matched antibody conjugated nanoparticle (iso- NP) condition.

146 SKBR-3 cell specimen III

Figure 3.25: Her-NP binds SKBR-3 cells. Flow cytometry data for cell specimen III, Herceptin conjugated nanoparticle (Her-NP) condition.

147

SKBR-3 cell specimen I

Her-NP Iso-NP Cells

Table 3.4: Tissue Reconstruction continuum mechanical properties for cell specimen I. Date: 12/07/2004.

Cells Iso-NP Her-NP

Table 3.5: Percent difference relative to the agarose alone sample for continuum mechanics, cell specimen I. Date: 12/07/2004

148 SKBR-3 cell specimen II

Her-NP Herceptin Iso-NP Cells

Table 3.6: Tissue reconstruction continuum mechanical properties for cell specimen II. Date: 01/20/2005

Cells Her Iso-NP Her-NP

Table 3.7: Percent difference relative to the agarose alone sample for continuum mechanics, cell specimen II. Date: 01/20/2005

149 SKBR-3 cell specimen III

Her-NP Her Iso-NP Cells

Table 3.8: Tissue reconstruction continuum mechanical properties for cell specimen III. Date: 02/09/2005

Cells Her Iso-NP Her-NP

Table 3.9: Percent difference relative to the cell sample for continuum mechanics, cell specimen III. Date: 02/09/2005

150 SKBR-3 cell specimens

Cell specimen IV:

-NP -NP

Cell specimen V:

-NP -NP

Cell specimen VI:

-NP NP

Table 3.10: Her-NP exhibits greater density and stiffness than other conditions. Display of the means and standard deviations for the mechanical parameters of each test condition for cell specimens IV - VI. Density and the stiffness component (Lambda + 2mu) correlate with the targeted mechanical contrast agent concentration.

151 SKBR-3 cell specimens

Cell specimen IV:

-NP NP

Cell specimen V:

-NP -NP

Cell specimen VI:

-NP NP

Table 3.11: Her-NP causes greater density and stiffness change by percent than other conditions. The percent difference of each test condition for each of cell specimens IV- VI was normalized to the cold flow buffer (CFB) condition. The CFB condition contained no particles, cells, or antibodies [only agarose and CFB]. The conditions were normalized to reduce the effect of sample preparation variation.

Agarose Tissue Phantom Study

Table 3.12: A summary of reconstructed continuum mechanics attenuation factor #1 (AL) values from the inversion algorithm for agarose tissue phantom studies.

152

Table 3.13: A summary of reconstructed continuum mechanics attenuation factor #2 (AT) values from the inversion algorithm for agarose tissue phantom studies.

153 3.3.3 Human Breast Tissue Studies

NPs on Tissue

Figure 3.26: C-Scan is capable of differentiating between iron oxide NP concentrations and NP presence attenuates signal. C-scan of invasive ductal carcinoma tissue with no, low, and high concentrations of bare (untargeted) iron oxide nanoparticles.

154 NPs on Tissue

Figure 3.27: C-Scan is capable of differentiating between NPs bound to Her-2/neu positive tissue compared to tissue alone and NP presence attenuates signal. C-scan of invasive ductal carcinoma tissue with a 3+ Her-2/neu score. Tissue was tagged with antibody or antibody-conjugated nanoparticles. The two columns of images were normalized at two different intensity scales.

155 NPs on Tissue

Figure 3.28: C-Scan is capable of differentiating between NPs bound to Her-2/neu positive tissue compared to tissue alone and NP presence attenuates signal. C-scan of invasive ductal carcinoma tissue with a Her-2/neu expression score of 3+. Tissue was tagged with antibody and antibody-conjugated nanoparticles. The column on the left displays the attenuation levels of the tissue, while the column on the right exhbits the frequency of tissue reflections.

156

Her-NPs on Tissue

Figure 3.29: Tissue sections and regions used for particle/ab treatments and ultrasound analysis. The top column depicts the regions of interest on ultrasound images of four tissue sections from a biopsied tumor. The corresponding histological sections are exhibited along the bottom row, which were stained for Her-2/neu expression using a standard, clinically used Her-2/neu staining system. All four histological sections were pathologist-scored to be 2+ for HER-2/neu expression.

157 Gold NP Conjugates with SKBR-3 Cells

Iso alone Her alone Her-NP gold-1

Iso-NP gold-1 Iso-NP gold-2 Her-NP gold-2

Figure 3.30: Her-Gold NPs bind to SKBR-3 cells better than other, control conditions by flow cytometry. Flow cytometry data of Herceptin and iso-conjugated gold particles (through interaction with protein A): “1” and “2” for iso-NPgold and Her-NPgold denote two distinct particle solvent testing conditions.

3.4 Discussion

This study aimed to demonstrate proof-of-feasibility to couple cancer-associated molecular changes with differentiable physical property changes, which were transformed to a detectable, quantitative “molecular signature” output by using a novel combination of nanoparticulate construction, ultrasound technologies, and mathematical

158 reconstruction algorithms. Because a central difficulty for breast carcinoma is the identification and localization of primary tumors prior to metastasis, the present study is intended to enable earlier, more efficient detection of the lesions. In particular, because overexpression of the growth receptor Her-2/neu in breast cancer is prognostic and predictive of increased morbidity and mortality in breast cancer, antibody-based targeting reagents to Her-2/neu were assembled and tested on appropriate cell lines in vitro. In parallel, the ultrasound system and reconstruction algorithms were optimized for biological utility.

As described in Chapter 1, Her-2/neu has previously been targeted with iron oxide nanoparticles, chiefly for detection with MRI [(269-272)], but also for optical detection schemes [(273)], in vitro and in vivo. The novelty of this study lies in the integration between targeted nanoparticles with the previously non-biological detection modality

NDE ultrasound and the elucidation of mechanical parameters.

This discussion interprets the results of the experiments described in sections 3.2.

First described in this section is the convergence between initial ultrasound testing/optimization results using PDMS/agarose tissue phantom matrices and nanoparticulate targeting data toward the creation of a functional ultrasound protocol.

Subsequently, biological data and associated ultrasonic reconstructions of cell/agarose studies are discussed in detail. Finally, breast tissue studies are discussed from the perspective of achieving the initial hypotheses.

159 3.4.1 Tissue Phantoms and Cell Line Studies

The PDMS tissue phantom study was devised to help choose the appropriate particulate material type and concentration to use on a tissue-like surface and to determine what the ultrasound response (i.e., signal) was to the material. Secondarily, relevant ultrasonic parameters were to be investigated and optimized. PDMS tissue phantoms were used to test particulates comprising three different core materials.

Dextran-coated iron oxide and colloidal gold nanoparticles and silicon microfabricated silica microparticles (see Chapter 4 for their fabrication and properties) were distributed onto PDMS surface interfaces to mimic the condition of particles that do not penetrate tissue in bulk. C-Scan mode ultrasound was used to demonstrate the capacity of each particulate type to provide surface contrast (see Figures 3.2-3.4).

Because particle concentrations and volumes were unknown for some particle materials it was not feasible to compare on the basis of concentration or volume.

Therefore, the experiment consisted of the three particulate materials at three concentrations – high, low (tenfold less than high), and no (substrate only). While subsequent innovations and data enabled approximations of particle concentrations and volumes, which allowed generation of Table 3.1, the choice of particulate was made without this information. Because iron oxide nanoparticles showed substantial ultrasound contrast and their conjugation and purification protocols were already in the process of optimization for CAD investigations associated with Chapter 2, they were chosen for use with breast cancer. Thus, the choice of iron oxide nanoparticles was supported by the fact that they consist of a highly modular material that afforded ease of conjugation and purification (as demonstrated in Chapter 4) as well as the ability to

160 generate ultrasound contrast. Only a fraction of the original nanoparticle batch was expected to bind to a surface presenting an appropriate antigen to the targeting reagent

(nanoparticles) because binding efficiency never approaches 100% – i.e., not all particles will bind, and those that do have an off-rate associated with them. Therefore, it was initially projected that a relatively concentrated batch of nanoparticles should be used for binding studies with ultrasound corresponding to the “high” iron oxide concentration.

Furthermore, the PDMS experiment established that iron oxide nanoparticles attenuate, rather than amplify, ultrasound reflection signal (analogous to the signal cancellation they cause with MRI as in Chapter 2).

Agarose tissue phantoms were evaluated to analyze the bulk distribution properties of nanoparticles in a tissue-like medium and determine if they could be differentiated by concentration. Gold was selected as the nanoparticulate material because of its increased density and greater core size than iron oxide. The sample conditions of various gold nanoparticle quantities in agarose were analyzed using continuum mechanics by reconstructing the Lamé moduli (elastic constants) λ and µ; thickness h, density ρ, and attenuation factors α1, αt via the two-step inversion algorithm described in Chapter 1. A modified one-step doublet mechanical algorithm was also used to reconstruct parameters in order to help resolve whether DM offered additional pertinent information to this problem. The data of both mechanical approaches have been summarized as average values and percent differences of mechanical properties for each experiment of the four sample conditions.

It was anticipated that nanoparticle concentration might modulate stiffness and density parameters. However, the reconstructed parameters do not include stiffness.

161 Therefore, two of the reconstructed parameters were combined to define a measure of stiffness (λ + 2µ) for all samples. Note that because the height varied no more than 3% between all samples, it was effectively a constant. Furthermore, because continuum attenuation factors had previously proven insignificant in similar studies, they were reserved for later study.

The utility of the DM model was compared to the continuum model by reconstructing five DM parameters – moduli A11 and A44, attenuation factors iA11 and iA44, and the internodal distance (η). The four gold nanoparticle concentrations were analyzed by DM, but because it was performed later primarily as an assessment of DM, only a subset of the total amount of data collected was evaluated.

The means of particle conditions, percent differences between the condition means, and statistical analyses were computed to form a conclusion about each set of data such that an inference could be made to resolve the identity of an unknown specimen.

The initial continuum and doublet mechanical results are summarized graphically in

Table 3.2 by displaying how well the mean of a given parameter corresponded with the relative concentration of nanoparticles. The correlation between nanoparticle concentration and their respective means is represented by a system of scoring in which the designation “High” indicates that the means of the specified parameter of the four concentration conditions were in precise increasing or decreasing order (represented by

“Low-High” and “High-Low” in the table, respectively, with regard to high concentration) with respect to particle concentration; “Good” denotes that three of the conditions were in increasing or decreasing order with respect to concentration;

“Inconclusive” signifies that two conditions correlated to concentration in increasing or

162 decreasing order. Moreover, the table is color coded to represent decreasing (High-Low, yellow), increasing (Low-High, green), and inconclusive (gray) results. The table demonstrates an apparent association between all seven specimens and most of the concentration/mechanical parameter pairs. In particular, while continuum parameters density, stiffness (λ + 2µ), and λ all appeared to have a correlation with nanoparticle concentration (as well as the doublet mechanical parameter A11), density clearly had the most robust association as indicated by this parameter yielding nearly all “High” correlation designations. Therefore, the discussion and statistical analyses will chiefly be centered on the density parameter. Stiffness analyses can be found in a recent dissertation [(268)].

To help draw an inference about an unknown specimen, analysis of variance

(ANOVA) was performed. This allowed an assessment of the statistical significance between the various concentrations as Table 3.2 only affords relative information about the trends of concentration values. The ANOVAs were designed to compare the particle conditions within each sample, as well as the sample densities throughout the set of specimens (this data helped characterize the appropriate conditions that lead to classification of an unknown specimen). In particular, they were used to compare the equality of greater than two means versus not all being equal with a 95% confidence level. The density ANOVAs, which are displayed as part of Appendix A, were calculated using Minitab software.

Additionally, the Tukey Honestly Significant Difference (HSD) Post Hoc Test was used to establish which differences in particle condition means (between all available pairs within a sample) were statistically different at the 0.05 level of significance without

163 inflating the alpha error rate. Representative Tukey density values in the first figure of

Appendix A (Figure A.1, an example of ANOVA, Tukey, and boxplot on an individual specimen) are highlighted to denote, for example, the differences between the upper and lower 95% limits for the high particle condition versus the medium particle condition.

This analysis is valuable because, for every two particle conditions, an assessment on whether the conditions are significantly different (with 95% confidence level) is determined by whether the interval includes “0” (if 0 is not included, the values are significantly different and vice versa). Therefore, for any set of two values, if they have the same sign (i.e., positive or negative), they are significantly different by this statistical test.

The boxplots, provided as a visual synopsis of the relationships between particle conditions, encompass interquartile ranges of 25%, 50%, and 75% (indicated by horizontal bars of each box). The red mark within each box indicates the mean and the vertical line through the center of the boxes designates the range of values obtained for that particle condition. Statistical outliers are demarcated by *s.

Statistics showed that the density means of the majority of specimens

(specifically, Specimens I-III and VII) demonstrated complete correlation with nanoparticle concentration, i.e., the ANOVA and Tukey tests revealed that each set of two conditions (in a given specimen) were statistically distinct. The other three specimens showed some correspondence. Specimen IV, for instance, displays very good correlation similar to Specimens I-III and VII with the exception of the “Medium” concentration condition. While the medium condition was revealed to be statistically distinct from “Low”, the two conditions were in reverse order by mean (i.e., the density

164 means were greatest in the “High” condition, then “Low,” after which the “Medium” and

“No” conditions were statistically indistinguishable, as the Tukey interval included 0).

While the source of error for Specimen IV was initially unidentified, improvements were formulated for the CMUS protocol and repeatability tests (i.e., the ability repetitively to gather equivalent data from the same specimen) were performed on Specimens V and VI.

The variability between repetitions on a single testing condition by consecutive analyses was within 2%. The variability between repetitions for which the sample was removed and replaced from its testing configuration was on average 3%, indicating that results were highly reproducible.

The source of the error in Specimen IV, as well as in Specimens V and VI (which displayed a lack of statistical significance between two sets and one set of conditions, respectively), was identified as a technical issue with the transducer’s focus during the ultrasound scanning process. This difficulty was resolved for future specimens by fixing in place the wires interfering with the transducer’s focus and always retesting focus after each scan. While the high correlation of Specimen VII results suggests that the adjustments were successful, it is necessary to test additional Specimens to confirm.

A summary evaluation across the conditions in the mean density data set is displayed in Appendix A (see Figures A.2-A.3). While the trend of correlation between the condition density means and particle concentrations was preserved, statistical tests proved that all proximal conditions (e.g., “High”-“Medium”) were not significantly different. However, all other comparisons between conditions (e.g., “High”-“Low”) were corroborated as significantly distinct. This indicated that the overall method was able to distinguish differences in nanoparticle concentration, but not quite at the highest level of

165 sensitivity tested (i.e., any set of two concentrations closest to each other). However because this evaluation included a set of specimens that was demonstrated to include a known error, it is feasible that use of the most recent protocol might provide the higher sensitivity sought.

A similar summary is exhibited for the percent differences (see Appenidix A).

The broad range of values present in the discrete specimens prevented this analysis from showing significant correlations, the distinct correlative appearance of the boxplot notwithstanding; thus, only the “High” and “No” nanoparticle conditions were proven statistically different. Rationale for the broad range of values that caused the statistical fuzziness between conditions include the reason specified above that this analysis included specimens with an error that was later fixed, disparities implicit in agarose formulations (as it was weighed out in form), and the potential for inhomogeneous blending of nanoparticles into the agarose bulk matrix. Agarose formulation inconsistencies could cause basal levels of reflected ultrasound signal to fluctuate – a potential solution would be to incorporate the “No” condition (i.e., agarose alone) as a normalization factor, so that each set of specimen data could be more accurately compared to other specimen data sets. The potential inhomogeneous mixing of nanoparticles would be difficult to control for and avoid, as nanoparticles might have been, for instance, unpredictably aggregating. This would likely affect the data by broadening the range of density values (e.g., density would increase where particles aggregate, and correspondingly decrease values everywhere else, as the particle concentration would be correspondingly diminished in all but regions of aggregation).

166 An analogous analysis was performed using DM. The mean values for DM parameters are displayed in Table 3.3 for all conditions of the seven specimens. The continuum mechanical equivalents, λ and µ, were computed via eqs 3.1 and 3.2. Table

3.3 provides a comparison between DM and continuum mechanics by calculating the percent difference between relevant parameters. Given that DM was constructed as a scalable theory that can provide viable mathematics for both micro- and macroscopic phenomena, it was hypothesized that the sets of both theories’ parameters should converge. Indeed, on average the difference between DM and continuum mechanical parameters (reconstructed from the modified one-step inversion DM and two-step inversion continuum algorithms, respectively) was 1-2%. This analysis is of interest not only because it suggests the convergent nature of doublet mechanics, but even though only 40 points were employed per particle condition for DM study (as opposed to the 441 points used for the continuum mechanical model), the values provided by DM still converged. However, possibly due to the decrease in data points, DM did not offer as robust a correlation between nanoparticle concentration and DM parameters as the continuum mechanical model (see [(268)] for the complete statistical analysis).

Cell studies

The goal of the breast cancer arm of this dissertation was to achieve specific and sensitive quantitative detection of Her-2/neu positive tissue biopsies using a targeted nanoparticle amplification strategy and CMUS. Therefore, given the tissue phantom results indicating ultrasonic detectability of iron oxide nanoparticles, a study was configured to investigate 1. whether a biological model of human tissue – human breast

167 cancer cell lines embedded in agarose – would produce differentiable ultrasonic signatures when targeted nanoparticles were cell-bound compared to controls and 2. the effectiveness of the nanoparticle bionconjugation and processing methods. In addition to the following discussion, details about this cell study can be found in the literature

[(274)].

Herceptin and an isotype-matched antibody to Herceptin were conjugated to iron oxide nanoparticles via conjugation protocols and assessed with characterization methods as presented and discussed in detail in Chapter 4. Initial testing of nanoparticulates on cells was performed sans ultrasound in order to optimize the nanoparticle conjugation and cell targeting techniques. Thus, it should also be noted here that modifications to the nanoparticle conjugation and separation procedure were made as needed during the cell line investigation (e.g., the shift from Miltenyi columns to permanent magnets to high gradient magnet and the variety of nanoparticle solvent conditions tested; see Chapter 4).

Such adjustments may preclude comparisons between pre- and post-ultrasound cell line experimental data (such as flow cytometry) across all experiments to some extent, as different yields of nanoparticles may be expected. However, comparisons between conditions within each experiment are valid, as are comparisons between differences in intraexperimental conditions across all cell line studies.

Detection modalities for cell studies included fluorescent microscopy and flow cytometry for both photographic imaging and quantitative evaluation of the binding between primary antibodies/conjugates and cells. Microscopy was used as visual verification of the flow cytometry, which facilitates quantitative characterization by employing a laser to produce a value indicative of the number of fluorescent counts per

168 cell-sized object. Fluorescent activation of binding events was mediated by a fluorescein isothiocyanate (FITC) tagged polyclonal secondary antibody to Herceptin and its isotypes. For these initial assessments, four conditions were employed: Herceptin- conjugated nanoparticles (Her-NP), isotype-matched conjugated nanoparticles (Iso-NP, negative control), Herceptin alone (Her, positive control), and iso alone (Iso, another negative control). Preliminary studies with Her-2/neu positive SKBR-3 cells revealed that Her-NP displayed approximately half the detected binding to cells as the Her alone condition as established by both fluorescence microscopy and flow cytometry (see

Figures 3.5-3.8). A hypothesis was generated that this was due to biochemical interactions and mechanical/steric interference, i.e., the primary antibody (Herceptin or isotype) was conjugated to nanoparticles via protein G molecules covalently bound to the nanoparticle surface. Protein G provides a high affinity “pocket” into which the Fc portion of the antibody may bind between constant regions 2 and 3 of the heavy chain

([(275)], see Figure 3.9), though it may bind weakly to Fab antibody fragments at constant region 1 of the heavy chain [(276)]. Because the secondary is polyclonal, the epitopes to which it binds may be located anywhere on the primary antibody. Therefore, when the

Her antibody binds to cells, the entire molecule (except the complementarity determining regions, or CDRs) is available to the polyclonal secondary. However, when Her-NP binds to cells, much of the Fc portion may be unavailable for binding by the secondary due to the steric hindrance engendered by protein G (see Figure 3.10). This would account for a diminished, yet still active, fluorescent signal. This hypothesis implies that in fact Her-NP is present in approximately the same quantities as Her on cells, but that the detection scheme produced an artifact in the data. Thus, an experiment was

169 configured to test the hypothesis: a FITC-tagged κ light chain only secondary antibody was used for all conditions. In the structure of an antibody, the κ light chain is present only in the upper, “forked” region (see Figure 3.9), constant region 1. Therefore, the secondary antibody would be capable of binding both Her and Her-NP with equivalent facility, as protein G would not sterically hinder binding to Her-NP. SKBR-3 cell results with the κ secondary corroborate this hypothesis, as fluorescent micrographs (matched with white light fields) and flow charts between Her and Her-NP were comparable (e.g., see Figures 3.11-3.14), as observed in all subsequent cell experiments (using the κ light chain only secondary).

While the fluorescence is comparable between Her and Her-NP, Her almost always maintains the slightly higher quantitative fluorescent count; this can be rationalized by two conjectures. First, the same amount of Her is incubated with the cells as is allowed to bind to nanoparticles during conjugation. Therefore, since conjugation yield is not 100%, losses of antibody are expected and may be predicted to influence the fluorescence of cells tagged with the conjugated nanoparticles. However, this feature may be partially offset due to the “multivalent nanoparticle” phenomenon, which denotes that nanoparticles have on average greater than one antibody conjugated to them (see

Chapter 4). Therefore, it is only necessary for one antibody on a nanoparticle to bind in order for the secondary fluorescent antibody to detect all antibodies on that particular nanoparticle, thereby correspondingly increasing the fluorescent count. This is because the intensity of fluorescence is directly proportional to the number of primary antibodies, whether bound to the cell or otherwise attached to the cell such as through a nanoparticle, as the secondary antibodies do not distinguish between the two. Thus, the fluorescence

170 of a single binding event is, in effect, amplified by the nanoparticle, which implies that even though there are less total antibodies in solution, fewer total binding events of conjugated nanoparticles are required to produce the same signal. Incidentally, these explanations may clarify why Iso-NP and Iso conditions alternated in their quantity of fluorescence, as the same number of non-specific binding interactions will produce a higher signal for Iso-NP, yet there is less overall antibody in solution. Furthermore, this phenomenon may explain the higher level of “punctateness” in the Her-NP micrographs

(see Figure 3.14); that is, Her tends to display a smooth distribution of fluorescence over cells, while Her-NP has a propensity to produce highly punctate images caused by more highly concentrated areas of nanoparticle binding. Due to multiple antibodies present on a single particle, only a small number of nanoparticles localized to a particular region of the cell membrane could generate a highly fluorescent appearance at that point (i.e., punctate).

The second conjecture as to why Her often has trivial levels of higher fluorescence than Her-NP is that previously bound nanoparticles may impart steric hindrance to new binding events by other nanoparticles. In this case, the number of antibodies bound to the particle (creating fluorescent counts) must compensate for the loss in the binding capacity due to steric concerns to create an equivalent number of binding events. The determining factor as to whether or not this matter will actually cause loss in fluorescent signal involves a complex formula involving nanoparticle binding efficiency, transient steric hindrances and probabilities of binding, as well as on- off rates, and other nanoscale interactions. While this conjecture was not tested, it could be examined by performing competitive binding experiments. These experiments would

171 entail devising ELISA configurations that would determine the binding kinetics (i.e., KD,

Kon, and Koff) of antibody conjugated nanoparticles and antibodies alone at concentrations of interest.

Binding by Iso and Iso-NP conditions were demonstrated to be negligible in comparison to Her and Her-NP both qualitatively (in images) and quantitatively (by flow) in this and all subsequent cell studies (see Figures 3.11-3.14).

The initial SKBR-3 experiments demonstrated the viability of the antibodies and indicated that the bioconjugation procedure was successful such that Herceptin antibodies had been effectively bound to the nanoparticles in bioactive conformation (i.e., the hypervariable region was in a spatial configuration conducive to binding its cognate, unhindered by steric or electrostatic inhibitors). Moreover, these studies proved that Her-

NP bound to cells in a manner comparable to Her, but significantly better than Iso or Iso-

NP. This suggested that the binding of nanoparticles to cell-associated Her-2/neu receptors was mediated by the specific Her-2/neu binding hypervariable region of the

Herceptin antibodies rather than non-specific interactions between the antibody and the cells or between the nanoparticle and the cells. A second set of control experiments was devised to generate more conclusive evidence that binding was specifically due to the molecular interaction between Herceptin and Her-2/neu through the use of a cell line with low Her-2/neu expression. This was to ensure the specificity and sensitivity of the antibodies and nanoparticle conjugates. The Her-2/neu positive cell line SKBR-3

(expressing 2.4 million receptors per cell) was compared with the Her-2/neu negative cell line MDA-MB-231 (expressing 22,000 receptors per cell) using the same nanoparticle/antibody conditions as employed previously. Using the same lot of

172 antibody/nanoparticle complexes, results demonstrated very low binding for both antibodies and nanoparticle conjugates with MDA cells and validation of the binding of antibodies and conjugates to SKBR-3 cells (only flow cytometry results are displayed, as fluorescent microscopy revealed no detectable fluorescence with MDA cells; see Figures

3.15 and 3.16). Note that the MDA study was performed prior to use of the κ light chain only secondary antibody, resulting in an approximately halved signal for Her-NP versus

Her. Furthermore, the results of Iso and Iso-NP conditions were similar between the

SKBR-3 and MDA cells, indicating that the interactions between Iso and Iso-NP with

SKBR-3 cells were in fact non-specific events between proteins and cells. This is because if the Iso and Iso-NP interactions were specific to the SKBR-3 cells’ Her-2/neu receptors, one would expect a marked decrease in fluorescence in the MDA cells for these conditions, which was not observed. Given the results of these experiments, the simplest and likeliest explanation is that Her-NP and Her were binding to SKBR-3 cells through antibody-antigen recognition, while iso and Iso-NP produced only background levels of binding.

Ultrasound analysis commenced after characterization of the nanoparticulate conjugates and their binding to cells had been completed (as described in section 3.2.2).

Although the conjugation and cell targeting protocols appeared functional, initial CMUS studies (Cell Specimens II and III, see Figures 3.19-3.25 and Tables 3.6-3.9) revealed little difference between the means of the mechanical parameters of Her-NP and Iso-NP condition (the four conditions for CMUS cell studies were Her-NP, Her (positive control), Iso-NP (negative control), and the cold flow buffer (CFB, which when mixed with agarose served as a normalization parameter to control for variations in agarose

173 specimen composition)). In addition to using the Iso-NP and Her alone as biological controls for flow cytometry, they and CFB were valuable controls for ultrasound testing as well, as they provided conditions with only antibody (which had bound to cells; Her), with relatively few particles (Iso-NP), and a baseline control (CFB).

It was soon realized that the cells were not washed sufficiently during the targeting protocol and due to drawbacks of microscopy and flow cytometry, this discrepancy had been overlooked. Realization occurred because of visual inspection of the tubes during the targeting process; even though flow cytometry revealed low amounts of fluorescence in Iso-NP, the color of both Her-NP and Iso-NP was rusty brown (indicative of the iron oxide nanoparticles) even after the washes. Further examination revealed that the cell pellet of the Her-NP condition was itself brown, while the cell pellet of the Iso-NP condition was nearly the color of a normal, antibody-only targeted cell pellet. A ring of brown (i.e., nanoparticles) was also noticed slightly above the cell pellet of Iso-NP. Thus, the difficulty was that, after incubating cells with particle/antibody solutions and centrifuging to wash, excess unbound particles had remained in the Iso-NP cell solutions. Centrifugation caused some of the particles to form the ring around the flow tube, thereby avoiding the subsequent buffer exchange/wash. This was not detected by flow cytometry because free floating, unattached nanoparticles are not in the cell size range, which is the only entity chosen for detection, nor was it observed in fluorescent micrographs. Using a pipette, an extra rinse around the upper edge of the cell pellet of all the conditions was added to the protocol after the post-primary antibody centrifugation and the post-secondary antibody centrifugation. This rinse caused the final solution of Iso-NP nanoparticles to become

174 significantly less brown than before, but did not noticeably alter the color of Her-NP.

Thus, although this issue technically affected both the Her-NP and Iso-NP particle conditions, because the majority of Herceptin conjugated particles in the Her-NP condition were bound to cells (as demonstrated by flow) they were left predominantly intact. Therefore, because the Iso-NP had not been washed well, the particles were still in the solution (albeit free-floating) and used for CMUS interrogation so that the Her-NP and Iso-NP conditions could not be differentiated when density and stiffness means were analyzed.

Subsequent CMUS evaluations utilized samples that had undergone the new protocol with additional rinses. This entailed CMUS interrogation of four additional autonomous samples (Specimens IV-VI) comprising four conditions each (see Appendix

B for flow cytometry data). The CMUS protocol used for the tissue phantoms was also applied to the cell-agarose specimens and the same set of statistical analyses were performed in the ANOVA and Tukey tests to establish the significance of the density and stiffness parameters to differentiate between specimens (see Appendix B). While density was selected for additional analysis due to its strong correlation with nanoparticle concentration in the agarose tissue phantom work, the elastic Lamé modulus λ (and its linear combination with elastic modulus µ, stiffness – λ + 2µ) also exhibited a relationship with nanoparticle presence. These correlations are apparent in Table 3.10, which presents the mean values of density, stiffness, and the components of stiffness, in addition to their standard deviations. Thus, these tests demonstrated the ability to differentiate between Her-NP and Iso-NP, Her, and CFB (in which only agarose and CFB were mixed). However, even though the trends were preserved within each cell specimen

175 for specimens IV-VI, the differences in density and stiffness between the three specimens require justification. The most probable explanation is, because new agarose was created for every test, variations in agarose composition (and thus density) produced shifts in those parameters. The effects of this can be observed by reference to the mechanical parameter values associated with baseline matrix CFB conditions, which fluctuate considerably. The mean values of density and stiffness for other conditions were therefore normalized to CFB in each experiment, as displayed in the chart of percent differences in Table 3.11. Table 3.11 further substantiates the contention that a clear trend exists when particles are bound to cells (Her-NP) in contrast to when they are not

(Her, Iso-NP), as Her-NP approximately doubles the other conditions in percent difference.

In addition to the qualitative evidence presented above, ANOVA and Tukey tests proved (which can be found, at 0.05 significance level, in the Ph.D. dissertation of Jason

Sakamoto) that the Her-NP density means are significantly different than Iso-NP, Her, and CFB [(268)]. Thus, it is clear that cell-bound nanoparticles embedded within a tissue model detectably impact the density and stiffness of samples. This conclusion led to progression of this work into human breast tissue specimens, as it furnished proof-of- principle for a quantitative molecular diagnostic ultrasound system for in situ and potential in vivo use.

Furthermore, based on the statistical data, conclusions were made on Iso-NP and

Her, which were statistically different than CFB. However, Iso-NP and Her were not statistically distinguishable from one another. Because the Her condition contained no particles, this suggested that the Iso-NP condition had only trivial quantities of non-

176 specific nanoparticle accumulation on the SKBR-3 cells (as indicated by flow cytometry).

Because Her may also cause cross linking between cells that could create mechanical effects detectable by ultrasound, this additionally provides provocative opportunities for future work (see Section 3.5).

3.4.2 Human Breast Tissue Studies

The objective of this study was to demonstrate the ability of ultrasound to differentiate biopsied breast tissue samples quantitatively using the molecular targeting capabilities of nanoparticles to enhance differences in the physical properties of tissue.

The preliminary studies, as conveyed above, served as proofs-of-principle that nanoparticles could be directed to the appropriate cell surface receptors/detected with ultrasound, and facilitated the establishment of appropriate protocols, which were adapted for use in the breast tissue studies. Among the modifications was the decision to shift from CMUS to C-scan ultrasound due to the inability of CMUS to sense particles on tissue surfaces (i.e., CMUS is optimally configured to evaluate bulk distributions). With

C-scan, focused transducers (and the requisite modifications to the reconstruction algorithms) were employed to yield a significantly narrower (sub-millimeter) spot size that improved resolution, reconstruction, and analysis capabilities. Initial experiments demonstrated that nanoparticles on tissue attenuated ultrasound signal, just as observed on PDMS (see Figures 3.2-3.4 and 3.26).

Next, preliminary targeted nanoparticle tissue studies were commenced with contiguous 10 µm sections of Her-2/neu positive tissue (IHC score 3+) using four variations of targeted nanoparticle or antibody alone (Her-NP, Her, Iso-NP, and Iso).

177 The Her-NP tissue condition exhibited very low signal compared with the other conditions, suggesting that the bound nanoparticles attenuated the ultrasound signal (this is observed in Figure 3.27 as a very weakly delineated structure, while Figure 3.28 on the right illustrates the same result in the frequency domain) as predicted from initial cell and

PDMS studies. It is important to observe, however, that Tables 3.12 and 3.13 do not support a correlation between attenuation coefficients and particle concentration in the agarose tissue phantom studies using CMUS.

Additional modifications were made to the protocol. The subsequent set of tissue studies commenced using 7 µm thick sections with no fixation step, as this was believed to lead to partial dissolution of the fat in the tissure, rendering it unpredictably weak and thin in some areas. Furthermore, the four tissue conditions were amended to include Her-

NP, Her, Iso-NP, and no treatement (i.e., no particle/antibody treatment). This study consisted of using pieces of tissue that had been independently IHC scored with values of

0, 2+, and 3+ with the intent of demonstrating the ability to differentiate between them.

Because it is not clinically significant to differentiate between 0 and 1+ IHC scores, 1+ tissue was excluded from study. Due to tissue heterogeneities and holes, it was necessary to select regions of interest on the tissue (see Figure 3.29, top, in which colored circles delineate regions) for evaluation by C-scan. These regions were matched against consecutive IHC scored tissue sections in order to include in the ultrasound analysis only areas of tissue considered to be the same as the general IHC tissue score. In order to compare the magnitudes of ultrasound reflections, the gray scale intensities of regions of interest were compared between conditions (see Figure 3.29, bottom, where increased attenuation corresponds to darker areas).

178 A quantitative test was performed for each region of each sample tested. A measure of the effect of targeted nanoparticles was obtained by subtracting the ultrasound intensity of Her-NP from the no treatment tissue condition. This measure was used to ascertain the specificity (True Negative/(False Negative + True Negative)) and sensitivity

(True Positive/(True Positive + False Positive)) of the samples and regions of interest from the perspective of being able to identify a 0, 2+, or 3+ sample consistent with its

IHC score. The sensitivity of this test conveys the ability of the ultrasound method to accurately predict an IHC score of 2+ or 3+ out of all 2+ and 3+ samples. The specificity, on the other hand, expresses the ablity of the ultrasound technique to accurately predict an IHC score of 0 out of all 0 samples. A diagnostic test is considered valid when both sensitivity and specificity are high (in addition to consideration of the

“incidence” and “positive predictive value” of the condition, which do not concern the present discussion); thus, when the test identifies most positive conditions (2+, 3+, i.e., sensitivity) and excludes most negative conditions (0, specificity) it is considered a valid measure of the ability to predict the IHC score.

The results exhibited 96.9% specificity (31 True Negatives and 1 False Negative),

100% sensitivity for 3+ tissue, 100% sensitivity for 2+ tissue, and 87.5% sensitivity for

0-scored tissues. This outcome approximately met or exceeded the stated initial objectives of the study to 1. create a targeted nanoparticle-enabled system capable of amplifying a molecular marker of breast cancer and 2. identify breast cancer samples with greater than 95% specificity, greater than 90% sensitivity for 3+ and 0/1+ tissue, and greater than 70% sensitivity for 2+ tissue. Thus, this may be a useful method for

179 detecting Her-/2neu positive breast cancer early, and in particular, may aid in predicting patient therapeutic response to Her-2 directed therapies such as Herceptin.

3.5 Experimental Limitations and Future Work

Many of the limitations of the experiments presented in this chapter were due to presently inescapable operator/instrumentation error. For instance, in final protocol form, nanoparticles were separated in a high gradient magnet. This required several washing steps and “pouring off” of supernate. Error is introduced as pouring lesser or greater amounts of supernate off may entail simultaneously pouring off decreased or increased quantities of particles. This could have led to potentially different signal response between experimental samples, which is why it was important to normalize the individual specimens within a sample to the CFB standard before comparing between samples.

Pipettes also introduced error in this work particularly because antibody was often added in increments of 0.2 µl. At this low volume level, the amount of antibody that may adhere to the outside of the pipette tip may not be trivial in comparison to the amount transferred (e.g., if 0.05 µl adheres to the pipette tip and is transferred to a solution in addition to the 0.2 µl within the pipette tip as intended, a total of 0.25 µl is transferred, which is a 25% increase, a significant amount). This may cause some variation between the amount of antibody added to any particular batch of nanoparticles (i.e., for any particular experiment). It is difficult to normalize for this error. Creation of larger stocks of nanoparticles would assuage these limitations, but introduce new limitations, such as leaving the particles more exposed to degradation by bacteria. Furthermore, because the targeted nanoparticles have an on/off rate of antibody (which is strongly dependent on the

180 isotype of antibody used; the association constant related to the interaction between protein G and IgG1 and IgG2b antibodies is on the order of 107 [(276)]) due to the conjugation chemistry employed using protein G, waiting to use the nanoparticles could entail many antibodies becoming detached from the nanoparticles. If a covalent conjugation chemistry were utilized instead (as described for the EDC/NHS chemistry in

Chapter 4), this would not be a concern, but the current protocol required use of the nanoparticles within a week (preferably within three days).

The PDMS tissue phantom study did not generate data in a form that allowed comparison between conditions. Because the various nanoparticulate materials employed in the PDMS study were not standardized on the basis of size and overall volume percent, it was difficult to assess their properties. Such normalization (i.e., standardizing all particle lots to the same volume fraction) would have yielded a more complete and accurate description of the interaction between nanoparticle physical properties and ultrasound response, but nanoparticle concentrations were only known for two of the three sets of particles.

Another limitation of this study was that only the Her-2/neu receptor was targeted. Even though up to 30% of breast cancer patients present with Her-2/neu positive lesions, it would be desirable to improve the early diagnostic capability and prognosis for a higher overall percentage of breast cancer patients. A more broadly applicable study might target additional overexpressed cell surface antigens. This would provide evidence that the technique is generally applicable with different proteins at various levels of expression. Other molecules that may be suitable for targeting in future work (for which there exist monoclonal antibodies) include p53, the abnormal clinical

181 expression of which has been correlated with breast cancer metastasis [(277)]. Phage display methods, in which phage are iteratively presented to a sample of interest in order to discover short peptide sequences that preferentially bind the sample, could also be used to select peptide sequences that selectively bind tumor tissue over normal tissue. These sequences could then be bound to nanoparticles (of any nanoparticulate material that presents available functional groups, including iron oxide, gold, polymeric, dendrimeric, nanotube/buckyball, semiconductor quantum dots, silicon, or other material) by simple bioconjugation chemistries that would enable nanoparticle targeting with a much higher level of control than the present work affords.

Experimental limitations were also implicit in the ultrasound component of the project. The doublet mechanical analysis, for instance, employed only 10% of the number of points used for continuum mechanics. An automated process/algorithm must thus be created and optimized for DM to become a realistic alternative to CM. Also, the cell and human breast tissue studies were extremely sensitive to bubbles, as ultrasound waves can not easily traverse pockets of air. Therefore, improved procedures should be established to eliminate the incidence of bubble formation.

Several opportunities for future work emerged as the investigations described in this chapter were performed. In one case, an experimental limitation led to the proposal of questions that require future work. This limitation was that only one nanoparticulate material was investigated in depth (i.e., with biological materials) using ultrasound; a variety of material densities and sizes may have been preferable. Initial studies provided the groundwork to employ other materials by performing gold nanoparticle conjugation and cell targeting protocol optimization (using protein A for antibody binding) in parallel

182 with preliminary investigations of iron oxide nanoparticle conjugation and cell targeting procedure optimizations. This work was performed in order to have another material readily available if CMUS were incapable of differentiating between cell samples employing iron oxide nanoparticles. Successful cell targeting of gold analogous to that observed with iron oxide nanoparticles was verified by flow cytometry (see Figure 3.30).

While the targeted gold nanoparticle/cell protocols were never tested with CMUS, they are discussed here for their potential in future work. In particular, gold nanoparticles were approximately 20 nm in diameter, while iron oxide cores were approximately 8-10 nm ([(278)] and see Chapter 2). Thus, although the overall iron oxide nanoparticle diameter far exceeded that of gold at ~100 nm, its dense core was considerably smaller by volume. Furthermore, the density of gold is 19,300 kg/m3, while the density of iron oxide is approximately 5,200 kg/m3. Therefore, given that gold nanoparticle cores are nearly 10 times larger by volume and greater than threefold more dense than iron oxide cores, it is hypothesized that gold nanoparticles would be capable of providing increased density and possibly stiffness amplification to the Her-NP condition in comparison to Her and Iso-NP. This hypothesis assumes that the same nanoparticle concentration is employed, which was not controlled for in the PDMS study of various nanoparticles, with an equivalent quantity of targeting antibody (i.e., on a per particle basis). Because it is now feasible to estimate the iron oxide nanoparticle concentration with the iron assay, it is suggested to test the ultrasound response to concentration-matched gold nanoparticle targeted cells and/or tissue. Direct comparison and analysis could then be executed between iron oxide and gold nanoparticles and evaluated on the basis of their inherent

183 characteristics. This could lead to better study design from the perspective of nanoparticulate material and size.

For future work, an intriguing component of Table 3.11 could be explored in further detail. The density means of Iso-NP are invariably slightly below that of Her, while the Iso-NP stiffness means are barely greater than those of Her. It has been postulated that the Her condition may create higher density within the cell-agarose matrix by effectively cross-linking adjacent cells (it is notable that this also applies to nanoparticles, which may facilitate cross-linking between cells through the multiple antibodies that could bind to different cells per particle, such that a component of the increased density and/or stiffness may be due to cross-linking effects). The mechanical effects of cross-linking agents such as glutaraldehyde on tissue are easy to feel by touch, so it was considered feasible that an antibody such as Herceptin may contribute to cross- linking effects in a manner that would be detectable by an instrument as sensitive as ultrasound. While the Her condition may create density through cross-linking effects, the

Iso-NP condition may have a small density contribution from the relatively small number of nanoparticles remaining with the cells after washing. These nanoparticles may also provide a stiffness component that the Her condition can not replicate with antibody alone. However, the differences in density and stiffness are sufficiently small to require extensive further investigations. Indeed, this hypothesis has not been evaluated in any detail, but an initial experiment to assess it must incorporate additional conditions for

CMUS, namely cells alone, Iso alone, and a small amount of unconjugated nanoparticles.

Her and Iso-NP could then be compared to these conditions, and the individual impacts

184 of antibodies, nanoparticles, and cells on the mechanical parameters could be isolated for more detailed study.

Other future studies based on this work may include a dose response experimental approach to standardize and correlate the number of Her-2/neu receptors with CMUS signal response. This study could comprise a correlation between IHC score (i.e., 0, 1+,

2+, 3+) and the ultrasonic mechanical response reconstructed from breast tissue or cell specimens that had been targeted by identical concentrations of Her-NP. This would be prerequisite to attaining the final objective of creating a substitute technology for IHC evaluation based off the present ultrasound work. It might be realized by devising a cell line-based dose response experiment that employs three standardardized Her-2/neu expressing breast cancer cell lines (SKBR-3, MDA-MB-175, and MDA-MB-231, which express 2.4 million, 92,000, and 22,000 Her-2/neu receptors per cell, respectively).

185

CHAPTER 4

NANPARTICULATE PLATFORMS

4. Nanoparticulate Platforms

Modular, platform-based nanoparticles capable of such multifunctional roles as targeting, sensing, reporting, and feedback/control of therapeutic release are among the most promising approaches for the early detection and treatment of many diseases. Broadly defined in section 1.1 in terms of their construction, functionality, in vivo behavior, and prospects for disease detection/modulation, the nanoparticulate platform focus in this Chapter will be on iron oxide and silicon-based particulate platforms. Iron oxide nanoparticles (USPIOs and SPIOs, or superparamagnetic iron oxides) have primarily been valuable due to reporter properties, i.e., their effects on MRI and ultrasound signal to facilitate disease visualization and their biocompatibility (see section 1.1 and Chapter 3). Iron oxides have been developed as an MR contrast agent with targeting properties for over 15 years [(279-281)]. The base iron oxide platform (generally an iron oxide nanocrystal wrapped in a polymer such as dextran) has been well-characterized physically and biologically [(36, 281-286)]. Iron oxides have interesting magnetic and biological properties even in unconjugated form, have various surface coatings (see section 1.1), and have been used in a variety of disease states [(61, 86, 287, 288)]. Multiple iron oxide agents with diverse sizes and surface coatings are either presently in FDA clinical trials or have been approved. Though they were initially designed for MRI contrast, in recent years USPIOs have been developed for multifunctional targeting, optical (e.g., near infrared fluorescence) visualization, drug delivery, site-specific heating/destruction of undesirable cells, and combinations thereof [(91, 289, 290)], while retaining attractive MRI susceptibility effects.

186 USPIOs have been directed to the Her-2/neu receptor, folate receptor, and VCAM-1, among abundant other targets, for use in imaging cancer and activated endothelium [(91, 269, 270, 291)]. However, combination targeting and drug delivery remains a common objective in the field. In one example, USPIOs were contrived in such a way as to target cancer cells via the particle-bound drug methotrexate, which gets subsequently cleaved in low pH conditions such as those which exist in the lysosome [(291)]. This targeted USPIO thus represents a platform capable of targeting, image enhancement, and “smart” intrinsic drug release. Furthermore, dual multimodal MRI/optical USPIOs have been produced which can be targeted and imaged with both MRI and NIRF (near infrared fluorescence). For example, VCAM-1 and phosphatidylserine have been targeted in vivo by covalent linkage between USPIOs and both VCAM-1 antibody/Annexin V and near infrared fluorochrome; this enables imaging of endothelial inflammation or apoptosis, respectively, with MRI and optical techniques [(91, 241)]. Another type of multimodal particle is based on the same USPIO platform – exploiting the cleaving ability of proteases implicated in specific disease states in order to identify the disease, an enzyme cleavable linkage is used to attach a fluorochrome to a cross-linked iron oxide (CLIO) [(292)]. Fluorescence is quenched by close proximity between fluorochromes, e.g., when they are linked to CLIOs, but proteolysis increases distance between fluorochromes, activating fluorescent signal [(293)]. This system thus simultaneously facilitates the optical technique to afford the ability to image protease activity and MRI to monitor CLIO localization [(292, 293)]. Methods of physical and biological characterization vary broadly with targeted and untargeted USPIOs. Iron mass can be quantified using methods such as atomic absorption spectroscopy, inductively-coupled plasma atomic emission spectroscopy (ICP- AES), and spectrophotometric comparison to iron standards (see section 4.1.2). For targeted USPIOs, surface protein content can be measured by such techniques as a modified P2T (pyridine-2-thione) method to monitor disulfide bonds, fluorescamine- based configurations, and BCA (Bichinchoninic acid)-based protein assays ([(293-295)], see section 4.1.2). The magnetic relaxation and susceptibilities of iron oxides are also measured, as these are key determinants of USPIO efficacy for MRI. Methods of verifying conjugated protein biofunctionality include configuring an ELISA with the

187 appropriate ligand (and using a colorimetric comparison) and targeting to a cell line known to express the ligand on its surface, then magnetically separating the cells and comparing to controls. USPIO sizing is commonly performed by a combination of electron microscopy (for iron oxide core crystals) and hydrodynamic sizing (e.g., dynamic light scattering for determining particle size in a particular solvent).

Porous Silicon Porous silicon (poSi) is an intriguing material that has begun to experience significant attention for its physical properties and biomedical usage. poSi nano- and microparticulates have excellent potential to be clinically and diagnostically significant both in and ex vivo. Investigated for use in fields as diverse as chemical and optical sensors, electronics, solar cells, and photonics [(10, 296-298)], reports reveal that poSi can simultaneously exploit several of these properties in combination with its biocompatibility and biodegradability for in vivo use [6]. The major techniques currently available to create porous silicon are electrochemical anodization (in which pore size and porosity are controllable by oxidizing the substrate by means of making the substrate the anode in an electrochemical cell and varying the magnitude and phase of the current) and stain (chemical) etching (using only chemicals in which nitric acid performs the role of oxidator). To form particulates, mechanical milling, ultrasonic fracture of poSi films, and photolithography are among the techniques which have been employed. poSi offers intriguing photoluminescent, electroluminescent, and chemiluminescent capabilities [(299)], which lead to numerous applications in sensing (e.g., based on color changes that vary predictably and detectably in response to analyte binding, local chemical changes, or local electrical variations). Further, numerous poSi chemical functionalization and derivatization techniques have been advanced to stabilize porous silicon and to facilitate linking to biomolecules [(10, 300)]. poSi exhibits a high degree of biocompatibility and stability (post-functionalization), the capacity to be surface functionalized, a high specific surface area, and it has a rapid and reversible response to stimuli. All these properties lend themselves to a variety of biomedical applications.

188 Currently pursued applications of poSi include uptake and/or binding of biomolecules and release from pores for diagnostic and particularly therapeutic clinical purposes. Pore size can be tailored to the appropriate dimension and surface characteristics (e.g., hydrophobicity), to match that of the molecules selected for release: for instance, enzymes, oxides, metals, receptors, proteins and cell permeation enhancers, and drug complexes have been imbibed and released from pores [(10)]. poSi has also been proposed as an ideal material for the harvest of low molecular weight proteomic molecules of interest from circulation [(10, 301)]. In a recent poSi advance, porous silicon particles imbibed with a radioactive 32P solution were injected near murine tumors. Complete or partial reduction in tumor volume was observed (depending on amount of 32P imbibed) compared to controls [(9)]. The ability of poSi to serve as a controlled delivery device in addition to its potential for targeting, luminescent reporter properties, potential to incorporate feedback systems, and the prospects for the capability to integrate electronics due to its silicon microfabrication origins make it perhaps an ideal candidate as a nanoparticulate platform.

4.1 Experimental The principal experimental design involved the preparation of biofunctional USPIO conjugates and a suite of characterization techniques to describe them. Conjugate preparation was driven both by the particular disease states studied (i.e., breast cancer and atherosclerosis) and the experimental designs specific to the projects (as described in Chapters 2 and 3) for which they were to be employed. Therefore, it was necessary to develop an efficient method of USPIO bioconjugation and magnetic separation. The separation protocols (designed to separate particles of interest from unbound molecules and to exchange solvents) described in this chapter reflect an iterative set of magnetic separation protocols that were aimed at achieving a relatively rapid, high-yield method – the protocols began with magnetic columns, moved to permanent magnets, and then to high-gradient magnets. Because a multitude of targeting reagents were to be employed between the two major projects (atherosclerosis and breast cancer), a modular approach to conjugation was adopted. Protocols will be described both for commercially available and collaborator-gifted USPIOs (as well as commercial gold nanoparticles), though the

189 commercially available USPIOs were predominantly used because they proved effective and rapidly processable. Subsequently, a set characterization methods to identify USPIO size, iron content, and protein content are described (a method to verify biofunctionality is described in section 2.2.1.1). Last, a description is given of silica (silicon dioxide) and porous silicon fabrication protocols developed and employed.

4.1.1 Preparation and Bioconjugation Bioconjugation Nanoparticles were either purchased (for dextran-coated USPIOs, Miltenyi Biotec, Gladbach, Germany; for 20 nm diameter protein A conjugated gold nanoparticles, Ted Pella, Redding, CA) or obtained from a collaborator (USPIOs from Miqin Zhang, University of Washington). Commercial USPIOs were purchased either with the protein pre-linked to the USPIO (Annexin V, antibodies to cd-1a, cd-11b, HLA- DR, cd-22, cd-4, cd-62L, from Miltenyi Biotec) or as protein G USPIOs which were conjugated with targeting reagents (antibodies to Her-2/neu (Herceptin, Genentech, South San Francisco, CA), VCAM-1 (Calbiochem, San Diego, CA), MCP-1 (BD Pharmingen, San Jose, CA), collagen type I, collagen type III, cd-18 (Research Diagnostics Inc., Concord, MA), cd-63 (Invitrogen, Carlsbad, CA), LOX-1 (Santa Cruz Biotechnology, Santa Cruz, CA), cd-43 (Research Diagnostics Inc.)) as described below. USPIOs from Dr. Zhang were covalently linked using a carbodiimide chemistry described below. Last, protein A gold nanoparticles were conjugated to targeting reagents in similar fashion as protein G USPIOs, as detailed; however, the separation process differed from that of the magnetic USPIOs. Magnetic separation (described below) was used to prepare pre-linked protein- conjugated USPIOs for use. Conjugation of protein G-linked USPIOs entailed mixing an aliquot of protein G USPIOs with the antibody to be conjugated (for Her-2/neu and its isotype in the breast cancer project, 100 µl protein G USPIOs and 10 µg antibody were combined; for the atherosclerosis project, in which USPIOs were injected into animals, 200 µl protein G USPIOs and approximately 40 µg antibody were combined).

Approximately 200-300 µl PBS without CaCl2 and MgCl2 (Invitrogen) was added to the USPIO/ab (antibody) solutions in 5 ml culture test tubes (Fisher Scientific, Hampton,

190 New Hampshire) which were then gently agitated for 45 minutes at room temperature or 4º C. They were subsequently submitted to the magnetic separation protocol described below to remove unbound abs. Collaborator USPIOs, consisting of an iron oxide core surrounded by a PEG (polyethylene glycol) coating, contained functional TFEE (trifluoroethylester) end groups in a sodium citrate buffer for stability and came with a nominal iron concentration of 0.1 mg/ml. Therefore, prior to conjugation, the solvent was exchanged to PBS (which is, unless otherwise noted, always without CaCl2 and MgCl2) using Miltenyi magnetic columns and TFEE groups were hydrolyzed in aqueous conditions to form carboxyl groups. 1 ml solvent-exchanged USPIOs was combined with 1 µg EDC (1-ethyl-3-(3- dimethylaminopropyl)carbodiimide) and 1.09 mg Sulfo-NHS (N- Hydroxysulfosuccinimide). 66 µg ab (22 µg for smaller proteins such as Annexin V) was added to the solution and mixed for 3 hours at room temperature with a magnetic stirrer (not directly into the solution, instead the stirrer was used agitate the tube containing the USPIO solution). Particles were then separated magnetically and resuspended in PBS.

Nanoparticle Separation Magnetic separation procedures endured many iterations. However, because this is not the focus of this dissertation, only the final protocol is described even though at various times magnetic columns, permanent magnets (1 Tesla at the surface), and centrifugation were used. Centrifugation is described as it was applied to ab-conjugated protein A gold nanoparticles. A high yield protocol for recovery of USPIOs is described for, e.g., washing of excess antibodies or solvent exchange. However, it requires 2 overnight incubations – if more rapid separation is desired, incubation times may be shortened at the expense of USPIO yield or another separation modality may be employed. For USPIO separation, a 5 ml culture test tube (tube A) containing the USPIO solution was inserted into an EasySep high gradient magnet (StemCell Technologies, Tukwila, Washington) and incubated at 4º C overnight. Next, supernate was poured off into a new 5 ml test tube (tube B) with tube A remaining in the high gradient magnet. 1.5 mls PBS were added to tube A and incubated about 2 hours at 4º C. Supernate was poured into tube B, then 1.5

191 mls PBS were again added to tube A and incubated at 4º C for about 2 hours. Supernate was poured from tube A into tube B, and tube B was placed into an EasySep magnet and incubated at 4º C overnight. Tube A was removed from the magnet, a small volume of PBS was added and mixed with USPIOs, and stored at 4º C. The next day, the supernate from tube B was poured off. After washing with 2.5 mls PBS and incubating for about 1- 2 hours at 4º C, the supernate was poured off again. Subsequently, tube B was removed from the magnet, about 200 µl PBS was added, and the side of the tube was rinsed to bring USPIOs back into solution. Tube B contents were added to tube A, and then the volume was brought to the desired level with PBS. Tube A was then stored at 4º C until use (e.g., injection or use for cell or tissue IHC). Protein A gold nanoparticles were conjugated with abs in the same manner as protein G USPIOs, and unbound abs were removed by repeated washings with centrifugation. Using a 1.5 ml eppendorff tube, gold nanoparticle-ab conjugates were placed into a microcentrifuge and spun at 4000 X G for 8 minutes. Supernate was carefully removed without disturbing the pellet (which should be red and puffy – if not, it may be impossible to resuspend nanoparticles). 500 µl PBS was added to the tube and nanoparticles were resuspended. This process of spinning and supernate removal was repeated twice, then the gold nanoparticles were suspended in the appropriate volume of PBS (for breast cancer studies, 100-200 µl) and stored at 4º C. The protocol must be optimized for each different antibody used (i.e., microfuge speed and length of time).

4.1.2 Characterization Experimental Annexin V and protein G USPIOs were characterized for protein and iron concentration, valency, and biofunctionality. Amounts of protein (Annexin V, protein G) on USPIOs were quantified using a variation of the bicinchoninic acid (BCA) assay [(295)], the reagents for which were purchased from Pierce Chemical Company (Rockford, Illinois) and the manufacturer’s directions were followed. The mass of iron in samples was quantified by modifications to the protocol of Riemer et. al. [(302)]. An iron chloride standard was purchased (Sigma) to obtain a calibration curve. Ferrous and ferric iron were detected by the assay in a linear range of

192 0.2 – 30 nmol [(302)] via subsequent spectrophotometric evaluation. The iron detection reagent was created by combining 6.5 mM ferrozine, 6.5 mM neocuproine, 2.5 M ammonium acetate, and 1 M ascorbic acid (all from Sigma) in water (note: neocuproine must first be dissolved in methanol and HCl may need to be added dropwise once mixed into aqueous solution). Iron chloride was dissolved in 10 mM HCl, and calibration standards were produced by dilution to yield 100 µl aliquots with amounts ranging from .1 nmol to 1 µmol of iron at regular intervals (the technique yields a linear distribution of absorbance from at least 0.2 nmol to 30 nmol iron [(302)]). To the 100 µl iron chloride aliquots were added 100 µl 50 mM NaOH, 100 µl 1.4 M HCl, and 30 µl iron detection reagent; these were incubated for 30 minutes. The samples (e.g., Annexin V, protein G, bare (unconjugated), Dr. Zhang’s, and cd-11b USPIOs) were prepared in a similar way so as to be able to compare the sample iron content to the range of iron chloride standard concentrations by correlation to their respective absorbance values. USPIOs were dissolved in 37% HCl heated to 70°C for 30 mins. Dilutions were then made as appropriate (1:10, 1:20, 1:100, 1:1000, etc.) to test the iron concentrations in each USPIO solution’s linear range. Known volumes of dissolved USPIOs were combined with 10 mM HCl to yield 100 µl total volumes. Added to this were 200 µl 10 mM HCl and 30 µl iron detection reagent. Solutions were then incubated for 30 minutes concurrent with the iron chloride standard incubation. Subsequently, 280 µl of each solution was transferred to a 96-well ELISA plate and excited at 490 nm on an ELx-808 absorption reader (Bio-Tek instruments, Winooski, Vermont). USPIO solutions were compared to the iron chloride standard curve for determination of iron content, and based on the initial volume of USPIOs taken, USPIO concentrations were back-calculated. Particles were sized by hydrodynamic diameter via Dynamic Light Scattering (DLS) using a Zetasizer Nano S Series instrument from Malvern Instruments (Southborough, Massachusetts). Particle size analyses were carried out on a variety of USPIOs and gold to obtain the particles’ average hydrodynamic diameters. Readings were performed by dilution of particle batches in PBS 10:1 and sonicating prior to analysis, which was performed according to manufacturer’s directions. Included in these studies were Micromod Partikeltechnologie USPIOs (Germany), which were chosen for

193 use in future conjugation studies due to the ability to purchase them with carboxyl and amine functional groups. These particles came nominally at 100 nm and 50 nm for both carboxyal and amine USPIOs – DLS was used to confirm this report. Annexin V biofunctionality was examined as described in section 2.2.1.1.

Results Analysis of the protein and iron content and hydrodynamic diameter of some of the USPIOs used is given in Tables 4.1 and 4.2. In the size evaluation it is notable that when proteins are bound to the USPIOs, the effective hydrodynamic diameter, as assessed by laser scatter evaluation of the Brownian motion of the colloids, appeared to decrease. This was uniformly observed. The protein assay showed that protein content was similar between Annexin V and protein G USPIO solutions. The iron content of particle batches showed there was about 10 times less iron in Annexin V than in protein G USPIO solutions, and cd-11b USPIO solutions exhibited about 20% less iron than those of Annexin V (Table 4.1). Basic (unconjugated) UPSIOs were demonstrated to have 0.96 mg iron/ml with this method (Table 4.1), in strong agreement with atomic abosorption spectrometry (which measured 1 mg iron/ml of basic USPIOs). Transmission electron microscopy (TEM) analysis of the iron oxide core size of Annexin V USPIOs agreed with previous estimates of about 10 nm ([(278, 303)], see Figure 4.1). Using this estimate, in addition to the assayed mass of iron, averaging the densities of

Fe3O4 and Fe2O3 (assuming an approximately equimolar mixture of the two), size of the proteins, and assuming the iron oxide cores are spherical, the average number of proteins per USPIO for Annexin V and protein G reagents were computed (see Table 4.1).

194

USPIO Iron content Protein content USPIO Valency type (mg/ml) USPIOs/ml (µg/ml) (proteins/USPIO) Annexin V 0.05 4 X 1013 29 18 Protein G 0.48 4 X 1014 28 2 Basic 0.96 8 X 1014 0 0 cd-11b 0.04 3 X 1013 - - Dr. Zhang 0.1 - - -

Table 4.1: Displays the assayed iron and protein contents, as well as calculated values for number of particles per ml and proteins per particle, for some USPIOs.

Particle Type Hydrodynamic Diameter (nm) Annexin V USPIO 96 Anti-cd-18 USPIO 72 Protein G USPIO 94 Unconjugated USPIO 108 Herceptin USPIO 66 Isotype control USPIO 68 Miqin USPIO 34 Protein A Gold 34 Herceptin Gold 35 Micromod COOH 100 105 Micromod COOH 50 82 Micromod NH2 100 124 Micromod NH2 50 96

Table 4.2: Gives the hydrodynamic diameter of many of the nanoparticles used in this dissertation.

195 USPIO TEM

Figure 4.1: Iron oxide cores of annexin V USPIOs averaged approximately 10 nm in diameter. TEM of annexin V USPIOs reveals the core diameter is about 10 nm, though some USPIOs in this figure may be aggregated. Magnification: 85,000X

4.2 Silicon Platforms Silicon microfabrication methods were developed to fashion particles for potential use as contrast agents for characterization mode ultrasound and C-Scan (silica microparticles, see Chapter 3) and for possible future use as drug delivery and imaging particulates (porous silicon). Chiefly, standard microfabrication techniques and reagents were employed to fashion these particles, with the exception of the stain etching protocol.

4.2.1 Fabrication Silica Microparticles Silica microparticles were developed by modification to the protocol of Nashat et. al. [(53)] using equipment at the Ohio MicroMD (Columbus, Ohio). Briefly, a 150 nm etch stop (silicon oxide) was thermally grown on single crystal, 100 mm p-type <100> silicon wafers at about 1000º C in wet conditions. An 1800 nm polysilicon sacrificial layer was deposited over the etch stop using low pressure chemical vapor deposition (LP- CVD). The particle layer was then formed by growing a 1000 nm low temperature oxide

196 layer over the sacrificial layer. Annealing for one hour at 1000º C in nitrogen gas was sometimes performed, but was found to be likely unnecessary. Photolithography was then performed. After spinning on photoresist and patterning with ultraviolet light using a mask (borrowed from iMEdd, Columbus, Ohio) to define 2 µm diameter circles, the photoresist was baked. Exposed areas of photoresist were etched away and particles were defined on the wafer via reactive ion etching (RIE). Subsequently, the 2 µm discs on the surface of the polysilicon were released by placing the wafer in a tub of 25 mls 6M KOH heated to approximately 80º C to produce an isotropic wet etch. Off-gassing occurred rapidly, and when bubbles were evenly distributed on the wafer, the wafer was removed (generally less than 5 minutes). The wafer was washed with PBS to rinse microparticles off into the KOH/particle solution, which was then rapidly pipetted into eppendorff tubes. Tubes were placed in a microcentrifuge and centrifuged at 8000 X G for 10 minutes. Supernates were poured off and PBS was replaced (about 1.5 mls). Tubes were spun again, and this procedure was repeated until microparticle solutions were below pH 8. Solutions were stored at room temperature and some were treated with sodium azide (0.05%) for extended storage.

Porous Silicon Porous silicon films were produced by stain (chemical) etching and electrochemical anodization. Anodization was achieved by immersion in an aqueous HF (hydrofluoric acid) solution in an electrochemical cell (built at OSU) for 3 minutes at a current of 0.99 A. Stain etching was performed on n- and p-type silicon wafers with previously grown 500 nm wet oxide and 1200 nm polysilicon over that (performed at the University of California, Berkeley, Berkeley, CA). Using a 1:3:5 solution of HF (49%):HNO3 (65%):de-ionized (DI) water, wafers were placed in the stain etching solution for 15 sec, 30 sec, 45 sec, 1.5 mins, and 2.5 mins. While in solution, wafers were gently agitated.

Alternatively, a volume ratio of 200:1 HF:HNO3 was used, with the wafers etched for 20 seconds. Immediately following exposure to etching solutions, wafers were plunged into a DI water bath.

197 4.2.2 Characterization Silica microparticles were sized by light microscopy while on the wafer. Silica microparticles are shown on the wafer prior to release and after KOH etch in PBS solution (Figure 4.2). The microparticles were generally uniform discs of about 2 µm diameter (and about 1 µm thick), except for defects due to flaws in the mask used (which caused some particles to be less than 2 µm diameter). Approximately 5.5 X 108 (550 million) microparticles were retrieved from each silicon wafer as calculated by hemocytometer analysis. This was a reasonable yield of microparticles (up to 109 could have been retrieved) given the available area for microparticle construction on the wafer and the individual particle area. Porous silicon surfaces were characterized by scanning electron microscopy (SEM) and TEM to inspect surface roughness, thickness, and the nominal pore size. As

Figures 4.3 and 4.4 demonstrate, for the 1:3:5 HF:HNO3:water solution at 15 and 30 sec, increased staining times engender rougher porous silicon surfaces, the thickness of porous silicon development is approximately up to 50-60 nm, and the pore size is evidently 5-10 nm.

Silicon Microparticles

Figure 4.2: Microfabricated silica microparticles are 2 µm diameter discs. Silica microparticles on wafer (left) and after release by KOH etch (right).

198 Porous Silicon: SEM

Figure 4.3: SEMs show stain-etched porous silicon is etch-time dependent. Representative electron micrographs show that 30 seconds of stain etching appear to create a considerably rougher surface (left) than 15 seconds of staining (right).

Porous Silicon: TEM

Figure 4.4: TEM reveals thickness and approximate pore size of stain-etched porous silicon. Representative electron micrograph shows that porosity is observed up to 50-60 nm in depth with a pore size of about 5-10 nm.

199

4.3 Experimental Limitations Limitations existed not only in the contexts of the individual atherosclerosis and breast cancer projects, which were discussed in previous chapters, but also in the methods of fabrication, conjugation, and characterization of the materials used in that work. Limitations involved in USPIO and silicon microfabrication experiments are described below.

Iron Oxide Nanoparticles One of the essential aspects of work with USPIOs was the creation of an efficient separation mechanism. While various methods have been used in the literature, from centrifugation to size exclusion, the most common technique for magnetic materials involves exploitation of magnetic properties to separate the magnetic materials from other compounds in a solution (e.g., unconjugated entities in a conjugation scheme) or to effect a solvent exchange. Many commonly-used modalities were tested as part of the present work (including centrifugation, size-based exclusion, and magnetism) as increased USPIO yield and efficiency were optimized. While the best method that was assessed employed high-gradient magnetic fields to draw USPIOs to the sides of culture tubes, this method too had limitations. Primarily, the high-gradient magnet technique is heavily operator dependent. For instance, because a series of steps is required in pouring off supernatants, it may be easy for an inexperienced operator to incur increased particle losses. Further, this limitation may cause even the experienced operator to incur a different percent loss each time the procedure is performed (though losses remain relatively low). Therefore, for experiments sensitive to the precise amount of iron used, it will be essential to measure the iron concentration both before and after magnetic separation. Another limitation of magnetic separation with the USPIOs used in this dissertation is the amount of time it takes. Two nights of incubation are required; thus, it is not feasible to produce separated, usable USPIOs rapidly even if they are urgently required. A much stronger magnetic field or size-based exclusion protocols (e.g., sephadex , which were demonstrated to function successfully as a USPIO separation method in the magnetoradioisotopic USPIO protocol

200 delineated in section 2.2.1.1, though they do not allow as high a yield as magnetic separation and they necessitate much higher dilution of USPIOs) would be expected to solve this issue by rapidly yielding separated USPIOs, if necessary. Protein G conjugation of USPIOs was employed in part to standardize the conjugation process in order to use the USPIOs as a platform across projects with a variety of antibodies. Thus, protein G USPIOs facilitated a survey of many injectable targeting reagents (see section 2.2) and were concurrently useful for conjugation of Herceptin (and controls). The benefits of this system included that a new protocol for conjugation did not have to be developed for each individual antibody, it simplified and shortened the conjugation process, it normalized the number of antibodies per USPIO (antibodies could not exceed protein G binding sites per particle), and antibodies were structurally required to be in bioactive position, as the antibody constant regions bind to protein G, liberating the hypervariable regions for binding to their ligand. If bioconjugation chemistries had instead been used, it may have been necessary either to employ a site-specific chemistry (to promote linkage in biofunctional configuration), or to link sufficient amounts of protein to overcome any bioactivity losses caused by attachments of some proteins to ligand binding sites (which would hinder or eliminate binding capacity). A constraint of using protein G USPIOs was that only antibody subtypes whose constant regions were known to bind to protein G could be used (all antibodies used were known to bind protein G avidly, hence this was not a concern in the present studies). Using bioconjugation chemistries, the flexibility in choice of biomolecule binding would be significantly increased to include all biomolecules, rather than the subset comprising antibodies that bind to protein G. This might become important, for instance, in contriving USPIOs that bind to peptides or aptamers that recognize disease-specific molecules. The mode of UPSIO sizing employed, dynamic light scattering, enables estimation of nanoscale particle diameters via the laser scatter of light by particles experiencing Brownian motion. While this method has some advantages, such as being relatively rapid and straightforward in producing calculated values for particle hydrodynamic size distributions (computed by an algorithm that theoretically correlates particle diffusion with size), it is also prone to inaccuracies. The size distribution is

201 dependent upon many parameters, including the solvent viscosity and temperature, as well as ionic strength, salt concentration, pH, etc. In particular, for particles which are not bulk solids, such as USPIOs that have a polymeric dextran coating, the properties of the solvent may critically influence the analysis because dextran strands are influenced by solvent chemical parameters. This can produce slightly erroneous size estimates, particularly between runs if the solvents differ (though these parameters are controlled as much as possible with careful sample preparation). Further, aggregates of particles are computed as single, larger particulates. While this can be advantageous, using DLS to examine aggregation distributions, it can also affect a true evaluation of the particle population in terms of assessing the size of the average single particulate. These limitations could also be used beneficially by using DLS to monitor changes in a particle’s aggregation and size characteristics by systematically varying solvent conditions. A peculiar issue with DLS, as previously mentioned, was paradoxically that conjugated USPIOs appear smaller than their unconjugated counterparts. That is, linkage of other molecules – such as proteins – to USPIOS appeared to affect their DLS-based size in unexpected and possibly unreported ways, i.e., when proteins were linked to UPSIOs, they were shown invariably to contract in diameter. However, when proteins were linked with gold, they were slightly enlarged (see Table 4.2). This is likely because in contrast to the non-solid dextrans adorning USPIOs, gold particles are solid throughout. It is theorized that electrostatic interactions, van der Waal forces, hydrophobic forces, and/or other intermolecular interactions may be responsible for USPIO size decrease, literally causing the dextran strands around the iron oxide cores to have a propensity to shift toward the core when proteins are linked. Alternatively, it is hypothesized that the linked proteins may cause an unanticipated phantom effect on USPIOs, causing them to “appear” smaller, as measured by Brownian motion, due to diffusion based effects of the proteins on the particles even though diameter reduction does not physically occur. A further consequence of the fact that Brownian motion is used to approximate particulate diameters may be the 8 nm size differential between Annexin V UPSIOs and protein G USPIOs – even though Annexin V and protein G proteins themselves are of nearly identical size and the unconjugated particles are

202 identical, they appear to be different diameters. A hypothesis explaining this disparity is that the effective hydrodynamic diameter of Annexin V nanoparticles is larger because of its higher protein content per particle. More proteins per nanoparticle might make the particle appear larger because of the measurement technique, which is dependent upon particle diffusion, which would change based in part on how many proteins are attached to it (not only the size of the proteins). If this hypothesis could be corroborated, these results might also serve as confirmation that more proteins are present on Annexin V USPIOs than on protein G USPIOs – however, this hypothesis may be contradictory to the hypothesis stated above that USPIO conjugated proteins cause strands to contract and make the USPIOs appear smaller. An alternative method to DLS used for particle sizing, electron microscopy, is actually a complementary technology as it is limited to elucidation of only the electron dense USPIO core. Thus, the dextran coating is not considered in this analysis. In contrast to DLS, which provides a broad size distribution for many particles, the accuracy of EM is highly sensitive to sample size. Obtaining an appropriate sample size can be time-consuming and expensive, and still perhaps not entirely accurate in obtaining a realistic estimation of average particle volume (due to the statistical method used and the fact that in practice, even though small particles outnumber larger ones, they are given equal weight). Even though the sample size used in the present studies was not sufficient for statistical significance, it is reassuring that previous studies resolved approximately the same iron oxide crystal size as obtained here [(278, 303)]. Thus, DLS and EM provide very different information on USPIO parameters; while DLS allows investigation of overall particle population parameters (i.e., the complete particle, including surface coatings and conjugated molecules), EM elucidates particle core geometry, surface structure, and size (with appropriate sample size). A third modality, atomic force microscopy (AFM), can also be used to produce data on a particle’s surface features, topography, etc. in terms of its molecular interactions with the AFM tip. It is useful because it can provide an overall depiction of the entire particle, including carbohydrates and biomolecules such as dextrans and the proteins conjugated to them. However, like EM it too is time-consuming and expensive.

203 Iron interference is a known difficulty in spectroscopic measurements to quantify biomolecules linked to iron-containing materials. Therefore, measures were taken to minimize interference effects in the µBCA protein assay to measure protein content on USPIOs. These included minimizing the amount of iron by using sets of particle dilutions to establish the minimum iron concentration at which it was possible to measure protein concentration and using basic (i.e., bare, unconjugated) USPIOs to normalize for the zero protein level. While these actions facilitate reasonable estimation of protein concentrations, an optimal protein assay would circumvent iron interference effects. This might be achievable, for example, by using an agent to cleave the covalent bonds connecting protein to USPIO (and leaving the protein intact for protein assay), then removing USPIOs by magnetic separation, and measuring protein content by the µBCA or Coomassie Blue assay. In theory, this assay would be capable of highly sensitive and accurate quantification of proteins on USPIOs. While ICP-AES and atomic absorption spectroscopy are very sensitive methods capable of very accurately identifying a wide range of iron concentrations, these approaches are both time-consuming and expensive. For these reasons, the spectrophotometric method of establishing iron concentration of USPIOs via chelation was developed. This technique is sensitive both to ferrous and ferric iron and is not affected by other divalent metal cations. While more rapid and less expensive, the limitations of this spectrophotometric method are first that it is only known to be valid (i.e., in the linear range) from 0.2 nmol to 30 nmol iron and second, that it is critically sensitive to inaccuracies in iron chloride standard quantities (thus great care must be taken to weigh and handle the standard accurately and properly). Limitations also existed in the method used to compute the average number of proteins per USPIO. The technique is as accurate as its assumptions: the first assumptions are that the assayed values of protein and iron were correct (the implicit limitations of which were discussed above). Given these values, the USPIOs were assumed to be spherical, which is invariably untrue in practice. However, it provides a rough estimation of the size of the nanocrystal. Furthermore, the USPIOs were assumed to consist of equal amounts of ferrous (Fe+2) and ferric (Fe+3) iron, which may or may not be the case. An assay might be configured to determine the ratio of ferrous to ferric iron

204 by, for example, using a modified version of the phenanthroline method [(304)] to assay for ferrous iron and subtracting from the total iron as assayed by the method described above to yield ferric iron. This analysis would increase the accuracy of the iron estimate, but only moderately, as the difference in percent iron between ferrous and ferric states is relatively small. In addition, iron oxide crystals were assumed to be 10 nm in diameter on average. This is also a critical assumption (though verified in part by EM and previous reports), because, due to the spherical assumption, linear increases in diameter cause cubic increases in particle volume – thus, particle diameter assumptions have a cubic relationship with the calculated amount of protein per particle, as the amount of iron in the particle is cubically correlated with particle diameter. Last, one and only one iron oxide nanocrystal was assumed to be enmeshed within each dextran particle. This is not known, though EM micrographs (see Figure 4.1) generally appear to demonstrate single crystal particles. Neverthelesss, if multiple crystal USPIOs were present, it would effectively increase the average amount of protein per dextran-coated particle.

Silicon While the microfabrication protocol for producing silicon dioxide microparticles was relatively optimized, one drawback included imperfections in the mask that was employed (which, due to costs, was borrowed). This caused some small disparities in particle diameter; however, the applications for which the silica particles were employed did not required precise uniformity. Presumably, if a better mask were used, particles would be very highly uniform as this is a key advantage of microfabrication procedures. One feature that could be further optimized is the KOH etch for particle release. Because KOH etches silica at a much slower rate than silicon (600 times slower), it is an appropriate etchant; however, because it does etch silica, minimizing the exposure of particles is critical. It would therefore be suitable first to optimize the time and temperature of the KOH etch (i.e., to determine the amount of time required to release the vast majority of particles, so as to minimize silica exposure to KOH) and to purify the particles immediately after their release, such as by membrane filtration. This was attempted, but was not successful. Therefore, the particle KOH solution was cooled and diluted with PBS and centrifugation was employed. Particles were recovered, but it is

205 unknown whether they were appreciably etched – examination of released particles was not able to reconcile whether or not the KOH had an effect because of expected variation in particle diameter due to the flaws in the mask. Future improvements to the porous silicon stain etching protocol were also planned. While hydrofluoric acid (HF) is standard in stain etching, there was a perceptible lack of uniformity across the stain etched surface. Lewis acids such as HBF4 and HSbF6 have been reported to achieve considerably higher pore uniformity and are known not to affect photoresists (which is imperative for protocols that have been developed to generate unique porous silicon particulates). The literature contains many reports of stain etching of single crystal silicon with HF. However, the protocol developed herein involves etching polycrystalline silicon. Therefore, considerable optimization remains necessary. Because neither HF nor the other lewis acids above have been properly characterized for polysilicon-based poSi formation, the classically- employed HF (for single crystal application) loses its chief advantage of having been well-characterized. Thus, it is suggested to use and optimize HBF4 and HSbF6 (in combination with nitric acid and DI water) in future poSi polycrystalline silicon stain etching protocols.

206

APPENDIX A

Statistics: Analysis of Variance and Tukey Tests For Agarose Tissue Phantoms

207 Agarose Specimen I ANOVA and Tukey

Figure A.1: Significant differences between conditions. Statistical ANOVA, Tukey post test, and boxplot analysis of density for agarose specimen I.

208 Analysis between the density means of agarose specimens I-VII

Figure A.2: Significant differences in density between conditions for all specimens. Statistical ANOVA, Tukey post test, and boxplot analyses of density for agarose specimens I-VII.

209

Analysis between the percent differences in the density means of agarose specimens I-VII

Figure A.3: Significant percent differences in density between conditions for all specimens. Statistical ANOVA, Tukey post test, and boxplot analyses of the percent difference of density means normalized to the “No” particle condition for agarose specimens I-VII.

210 Analysis between the stiffness means of agarose specimens I-VII

Figure A.4: Significant differences in stiffness between conditions for all specimens. Statistical ANOVA, Tukey post test, and boxplot analyses of stiffness for agarose specimens I-VII.

211 Analysis between the percent differences in the stiffness means of agarose specimens I-VII

Figure A.5: Significant percent differences in stiffness between conditions of all specimens. Statistical ANOVA, Tukey post test, and boxplot analyses of the percent difference of stiffness means normalized to the “No” particle condition for agarose specimens I-VII.

212 Cell specimen I: Statistics

NP

NP

NP

Figure A.6: Significant differences in density between conditions for cell specimen I. Statistical ANOVA, Tukey post test, and boxplot analyses of the density of cell specimen I.

213 Cell specimen II: Statistics

NP

NP

NP

NP

Figure A.7: Significant differences in density between conditions for cell specimen II. Statistical ANOVA, Tukey post test, and boxplot analyses of the density of cell specimen II.

214 Cell specimen III: Statistics

NP

NP

NP

NP

Figure A.8: Significant differences in density between conditions for cell specimen III. Statistical ANOVA, Tukey post test, and boxplot analyses of the density of cell specimen III.

215 Cell specimen IV: Statistics

NP

NP

NP

NP

N N

Figure A.9: Significant differences in density between conditions for cell specimen IV. Statistical ANOVA, Tukey post test, and boxplot analyses of the density of cell specimen IV.

216

APPENDIX B

SKBR-3 Cell Specimen IV

217

SKBR-3 cell specimen IV

Figure B.1: Her binds SKBR-3 cells. Flow cytometry data for cell specimen IV, Herceptin (Her) condition.

218

SKBR-3 cell specimen IV

Figure B.2: Iso-NP does not appreciably bind SKBR-3 cells. Flow cytometry data for cell specimen IV, isotype-matched antibody conjugated nanoparticle (iso-NP) condition.

219

SKBR-3 cell specimen IV

Figure B.3: Her-NP binds SKBR-3 cells. Flow cytometry data for cell specimen IV, Herceptin conjugated nanoparticle (Her-NP) condition.

220

SKBR-3 cell specimen IV

Cells Herceptin Iso-NP Her-NP

Table B.1: Her-NP exhibits greater density and stiffness parameters than other conditions. Tissue reconstruction continuum mechanical properties for cell specimen IV. Date: 04/08/2005

Cells Herceptin Iso-NP Her-NP

Table B.2: Her-NP displays greater percent difference in density and stiffness than other conditions. Percent difference relative to the cell sample for continuum mechanics, cell specimen IV. Date: 04/08/2005

221

BIBLIOGRAPHY

1. Lee, S. C., Bhalerao, K. & Ferrari, M. (2004) Ann N Y Acad Sci 1013, 110-23.

2. Bhalerao, K. D., Eteshola, E., Keener, M. & Lee, S. C. (2005) Applied Physics Letters.

3. Goldin, D. S., Dahl, C. A., Olsen, K. L., Ostrach, L. H. & Klausner, R. D. (2001) Science 292, 443-4.

4. Buxton, D. B., Lee, S. C., Wickline, S. A. & Ferrari, M. (2003) Circulation 108, 2737-42.

5. LaVan, D. A., McGuire, T. & Langer, R. (2003) Nat Biotechnol 21, 1184-91.

6. Lee, S., Ruegsegger, M., Barnes, P., Smith, B. & Ferrari, M. (2004) in Springer Handbook of Nanotechnology, ed. Bhushan, B. (Springer Verlag, pp. 279-322.

7. Smith, B., Ruegsegger, M., Barnes, P., Ferrari, M. & Lee, S. (2006) in BioMEMS and Biomedical Nanotechnology, ed. Ferrari, M. (Springer, Vol. 1.

8. Ginger, D. S., Zhang, H. & Mirkin, C. A. (2004) Angew Chem Int Ed Engl 43, 30- 45.

9. Zhang, K., Loong, S. L., Connor, S., Yu, S. W., Tan, S. Y., Ng, R. T., Lee, K. M., Canham, L. & Chow, P. K. (2005) Clin Cancer Res 11, 7532-7.

10. Smith, B., Nijdam, A., Cheng, M., Liu, X., Lee, S. & Ferrari, M. (2004) Material Technology 19, 16-20.

11. Duncan, R. (2003) Nat Rev Drug Discov 2, 347-60.

12. Shi, X., Majoros, I. J. & Baker, J. R., Jr. (2005) Mol Pharm 2, 278-94.

13. Qiu, L. Y. & Bae, Y. H. (2006) Pharm Res 23, 1-30.

14. Duncan, R. & Izzo, L. (2005) Adv Drug Deliv Rev 57, 2215-37.

15. Kobayashi, H. & Brechbiel, M. W. (2005) Adv Drug Deliv Rev 57, 2271-86.

16. Choi, Y., Thomas, T., Kotlyar, A., Islam, M. T. & Baker, J. R., Jr. (2005) Chem Biol 12, 35-43.

222 17. Choi, Y. & Baker, J. R., Jr. (2005) Cell Cycle 4, 669-71.

18. Tomalia, D. A., Brothers, H. M., 2nd, Piehler, L. T., Durst, H. D. & Swanson, D. R. (2002) Proc Natl Acad Sci U S A 99, 5081-7.

19. Tirrell, M. (2005) AIChE Journal 51, 2386-2390.

20. Wee, K. M., Rogers, T. N., Altan, B. S., Hackney, S. A. & Hamm, C. (2005) J Nanosci Nanotechnol 5, 88-91.

21. Holowka, E. P., Pochan, D. J. & Deming, T. J. (2005) J Am Chem Soc 127, 12423-8.

22. Rodriguez-Hernandez, J., Babin, J., Zappone, B. & Lecommandoux, S. (2005) Biomacromolecules 6, 2213-20.

23. Ng, C. C., Cheng, Y. L. & Pennefather, P. S. (2004) Biophys J 87, 323-31.

24. Li, X., Li, Y. & Yang, C. (2004) Langmuir 20, 3734-9.

25. Saghatelian, A., Yokobayashi, Y., Soltani, K. & Ghadiri, M. R. (2001) Nature 409, 797-801.

26. Jin, Y., Tong, L., Ai, P., Li, M. & Hou, X. (2006) Int J Pharm 309, 199-207.

27. Pinto, Y. Y., Le, J. D., Seeman, N. C., Musier-Forsyth, K., Taton, T. A. & Kiehl, R. A. (2005) Nano Lett 5, 2399-402.

28. Seeman, N. C. (2005) Methods Mol Biol 303, 143-66.

29. Demers, L. M., Ginger, D. S., Park, S. J., Li, Z., Chung, S. W. & Mirkin, C. A. (2002) Science 296, 1836-8.

30. Alivisatos, P. (2004) Nat Biotechnol 22, 47-52.

31. Ferrari, M. (2005) Nat Rev Cancer 5, 161-71.

32. Hirsch, L. R., Gobin, A. M., Lowery, A. R., Tam, F., Drezek, R. A., Halas, N. J. & West, J. L. (2006) Ann Biomed Eng 34, 15-22.

33. Sengupta, S., Eavarone, D., Capila, I., Zhao, G., Watson, N., Kiziltepe, T. & Sasisekharan, R. (2005) Nature 436, 568-72.

34. Gupta, A. K. & Gupta, M. (2005) Biomaterials 26, 3995-4021.

35. Weissleder, R., Elizondo, G., Josephson, L., Compton, C. C., Fretz, C. J., Stark, D. D. & Ferrucci, J. T. (1989) Radiology 171, 835-9.

223 36. Josephson, L., Lewis, J., Jacobs, P., Hahn, P. F. & Stark, D. D. (1988) Magn Reson Imaging 6, 647-53.

37. Wunderlich, G., Gruning, T., Paulke, B. R., Lieske, A. & Kotzerke, J. (2004) Nucl Med Biol 31, 87-92.

38. Nahman, N. S., Drost, T., Bhatt, U., Sferra, T., Johnson, A., Gamboa, P., Hinkle, G., Haynam, A., Bergdall, V., Hickey, C., Bonagura, J.D., Brannon-Pappas, L., Ellison, J., Mansfield, A., Shiwe, S., and Shen, N. (2002) Biomedical Microdevices 4, 189-196.

39. Lee, S. C., Parthasarathy, R., Duffin, T. D., Botwin, K., Zobel, J., Beck, T., Lange, G., Kunneman, D., Janssen, R., Rowold, E. & Voliva, C. F. (2001) Biomedical Microdevices 3, 53-59.

40. Moghimi, S. M., Hunter, A. C. & Murray, J. C. (2001) Pharmacol Rev 53, 283- 318.

41. Chen, R. J., Bangsaruntip, S., Drouvalakis, K. A., Kam, N. W., Shim, M., Li, Y., Kim, W., Utz, P. J. & Dai, H. (2003) Proc Natl Acad Sci U S A 100, 4984-9.

42. Hermanson, G. (1996) Bioconjugate Chemistry (Academic Press, San Diego).

43. Malkoch, M., Thibault, R. J., Drockenmuller, E., Messerschmidt, M., Voit, B., Russell, T. P. & Hawker, C. J. (2005) J Am Chem Soc 127, 14942-9.

44. Olafsen, T., Cheung, C. W., Yazaki, P. J., Li, L., Sundaresan, G., Gambhir, S. S., Sherman, M. A., Williams, L. E., Shively, J. E., Raubitschek, A. A. & Wu, A. M. (2004) Protein Eng Des Sel 17, 21-7.

45. Onclin, S., Ravoo, B. J. & Reinhoudt, D. N. (2005) Angew Chem Int Ed Engl 44, 6282-304.

46. Shirahata, N., Hozumi, A. & Yonezawa, T. (2005) Chem Rec 5, 145-59.

47. Li, Y. H. & Buriak, J. M. (2006) Inorg Chem 45, 1096-102.

48. Sharma, S., Johnson, R. W. & Desai, T. A. (2004) Langmuir 20, 348-56.

49. Santini, J. T., Jr., Cima, M. J. & Langer, R. (1999) Nature 397, 335-8.

50. Li, Y., Ho Duc, H. L., Tyler, B., Williams, T., Tupper, M., Langer, R., Brem, H. & Cima, M. J. (2005) J Control Release 106, 138-45.

51. Lesinski, G. B., Sharma, S., Varker, K. A., Sinha, P., Ferrari, M. & Carson, W. E., 3rd (2005) Biomed Microdevices 7, 71-9.

52. Desai, T. A., Chu, W. H., Tu, J. K., Beattie, G. M., Hayek, A. & Ferrari, M. (1998) Biotechnol Bioeng 57, 118-20.

224 53. Nashat, A. H., Moronne, M. & Ferrari, M. (1998) Biotechnol Bioeng 60, 137-46.

54. Martin, F. J., Melnik, K., West, T., Shapiro, J., Cohen, M., Boiarski, A. A. & Ferrari, M. (2005) Drugs R D 6, 71-81.

55. Faxon, D. P., Creager, M. A., Smith, S. C., Jr., Pasternak, R. C., Olin, J. W., Bettmann, M. A., Criqui, M. H., Milani, R. V., Loscalzo, J., Kaufman, J. A., Jones, D. W. & Pearce, W. H. (2004) Circulation 109, 2595-604.

56. Libby, P. (2006) Am J Clin Nutr 83, 456S-460S.

57. Crowther, M. A. (2005) Hematology 2005, 436-441.

58. Libby, P. & Theroux, P. (2005) Circulation 111, 3481-8.

59. Young, J. L., Libby, P. & Schonbeck, U. (2002) Thromb Haemost 88, 554-67.

60. Libby, P. (2002) Nature 420, 868-74.

61. Jaffer, F. A., Libby, P. & Weissleder, R. (2006) J Am Coll Cardiol 47, 1328-38.

62. Hansson, G. K. (2001) Arterioscler Thromb Vasc Biol 21, 1876-90.

63. Faxon, D. P., Fuster, V., Libby, P., Beckman, J. A., Hiatt, W. R., Thompson, R. W., Topper, J. N., Annex, B. H., Rundback, J. H., Fabunmi, R. P., Robertson, R. M. & Loscalzo, J. (2004) Circulation 109, 2617-25.

64. Schonbeck, U., Sukhova, G. K., Shimizu, K., Mach, F. & Libby, P. (2000) Proc Natl Acad Sci U S A 97, 7458-63.

65. Fuster, V., Fayad, Z. A. & Badimon, J. J. (1999) Lancet 353 Suppl 2, SII5-9.

66. Fuster, V., Moreno, P. R., Fayad, Z. A., Corti, R. & Badimon, J. J. (2005) J Am Coll Cardiol 46, 937-54.

67. Stary, H. C., Chandler, A. B., Dinsmore, R. E., Fuster, V., Glagov, S., Insull, W., Jr., Rosenfeld, M. E., Schwartz, C. J., Wagner, W. D. & Wissler, R. W. (1995) Circulation 92, 1355-74.

68. Kullo, I. J. & Ballantyne, C. M. (2005) Mayo Clin Proc 80, 219-30.

69. Khot, U. N., Khot, M. B., Bajzer, C. T., Sapp, S. K., Ohman, E. M., Brener, S. J., Ellis, S. G., Lincoff, A. M. & Topol, E. J. (2003) Jama 290, 898-904.

70. Peyser, P. A., Bielak, L. F., Chu, J. S., Turner, S. T., Ellsworth, D. L., Boerwinkle, E. & Sheedy, P. F., 2nd (2002) Circulation 106, 304-8.

71. Greenland, P., Knoll, M. D., Stamler, J., Neaton, J. D., Dyer, A. R., Garside, D. B. & Wilson, P. W. (2003) Jama 290, 891-7.

225 72. Fu, Q. & Van Eyk, J. E. (2006) Expert Rev Proteomics 3, 237-49.

73. White, C. J. (2005) Minerva Cardioangiol 53, 473-84.

74. Berman, D. S., Hachamovitch, R., Shaw, L. J., Friedman, J. D., Hayes, S. W., Thomson, L. E., Fieno, D. S., Germano, G., Slomka, P., Wong, N. D., Kang, X. & Rozanski, A. (2006) J Nucl Med 47, 74-82.

75. Rudd, J. H., Davies, J. R. & Weissberg, P. L. (2005) Trends Cardiovasc Med 15, 17-24.

76. Nighoghossian, N., Derex, L. & Douek, P. (2005) Stroke 36, 2764-72.

77. Stoneman, V. E. & Bennett, M. R. (2004) Clin Sci (Lond) 107, 343-54.

78. Clarke, S. E., Beletsky, V., Hammond, R. R., Hegele, R. A. & Rutt, B. K. (2006) Stroke 37, 93-7.

79. Auer, M., Stollberger, R., Regitnig, P., Ebner, F. & Holzapfel, G. A. (2006) IEEE Trans Med Imaging 25, 345-57.

80. Hofman, J. M., Branderhorst, W. J., ten Eikelder, H. M., Cappendijk, V. C., Heeneman, S., Kooi, M. E., Hilbers, P. A. & ter Haar Romeny, B. M. (2006) Magn Reson Med 55, 790-9.

81. Cunningham, C. H., Arai, T., Yang, P. C., McConnell, M. V., Pauly, J. M. & Conolly, S. M. (2005) Magn Reson Med 53, 999-1005.

82. Yuan, C. & Kerwin, W. S. (2004) J Magn Reson Imaging 19, 710-9.

83. Desai, M. Y. & Lima, J. A. (2006) Curr Atheroscler Rep 8, 131-9.

84. Fayad, Z. A. & Fuster, V. (2001) Circ Res 89, 305-16.

85. Ruehm, S. G., Corot, C., Vogt, P., Cristina, H. & Debatin, J. F. (2002) Acad Radiol 9 Suppl 1, S143-4.

86. Ruehm, S. G., Corot, C., Vogt, P., Kolb, S. & Debatin, J. F. (2001) Circulation 103, 415-22.

87. Schmitz, S. A. (2003) Rofo 175, 469-76.

88. Vymazal, J., Bulte, J. W., Frank, J. A., Di Chiro, G. & Brooks, R. A. (1993) J Magn Reson Imaging 3, 637-40.

89. Murillo, T. P., Sandquist, C., Jacobs, P. M., Nesbit, G., Manninger, S. & Neuwelt, E. A. (2005) Therapy 2, 871-882.

226 90. Choudhury, R. P., Fuster, V. & Fayad, Z. A. (2004) Nat Rev Drug Discov 3, 913- 25.

91. Tsourkas, A., Shinde-Patil, V. R., Kelly, K. A., Patel, P., Wolley, A., Allport, J. R. & Weissleder, R. (2005) Bioconjug Chem 16, 576-81.

92. Kelly, K. A., Allport, J. R., Tsourkas, A., Shinde-Patil, V. R., Josephson, L. & Weissleder, R. (2005) Circ Res 96, 327-36.

93. Strauss, H. W., Dunphy, M. & Tokita, N. (2004) J Nucl Med 45, 1106-7.

94. Flacke, S., Fischer, S., Scott, M. J., Fuhrhop, R. J., Allen, J. S., McLean, M., Winter, P., Sicard, G. A., Gaffney, P. J., Wickline, S. A. & Lanza, G. M. (2001) Circulation 104, 1280-5.

95. Kuhl, H. P., Beek, A. M., van der Weerdt, A. P., Hofman, M. B., Visser, C. A., Lammertsma, A. A., Heussen, N., Visser, F. C. & van Rossum, A. C. (2003) J Am Coll Cardiol 41, 1341-8.

96. Schaar, J. A., Muller, J. E., Falk, E., Virmani, R., Fuster, V., Serruys, P. W., Colombo, A., Stefanadis, C., Ward Casscells, S., Moreno, P. R., Maseri, A. & van der Steen, A. F. (2004) Eur Heart J 25, 1077-82.

97. Davies, M. J. (1996) Circulation 94, 2013-20.

98. Naghavi, M., Libby, P., Falk, E., Casscells, S. W., Litovsky, S., Rumberger, J., Badimon, J. J., Stefanadis, C., Moreno, P., Pasterkamp, G., Fayad, Z., Stone, P. H., Waxman, S., Raggi, P., Madjid, M., Zarrabi, A., Burke, A., Yuan, C., Fitzgerald, P. J., Siscovick, D. S., de Korte, C. L., Aikawa, M., Juhani Airaksinen, K. E., Assmann, G., Becker, C. R., Chesebro, J. H., Farb, A., Galis, Z. S., Jackson, C., Jang, I. K., Koenig, W., Lodder, R. A., March, K., Demirovic, J., Navab, M., Priori, S. G., Rekhter, M. D., Bahr, R., Grundy, S. M., Mehran, R., Colombo, A., Boerwinkle, E., Ballantyne, C., Insull, W., Jr., Schwartz, R. S., Vogel, R., Serruys, P. W., Hansson, G. K., Faxon, D. P., Kaul, S., Drexler, H., Greenland, P., Muller, J. E., Virmani, R., Ridker, P. M., Zipes, D. P., Shah, P. K. & Willerson, J. T. (2003) Circulation 108, 1664-72.

99. Farb, A., Burke, A. P., Tang, A. L., Liang, T. Y., Mannan, P., Smialek, J. & Virmani, R. (1996) Circulation 93, 1354-63.

100. Kolodgie, F. D., Virmani, R., Burke, A. P., Farb, A., Weber, D. K., Kutys, R., Finn, A. V. & Gold, H. K. (2004) Heart 90, 1385-91.

101. Fayad, Z. A. (2001) Int J Cardiovasc Imaging 17, 165-77.

102. Falk, E. (2006) J Am Coll Cardiol 47, 7-12.

227 103. Virmani, R., Burke, A. P., Kolodgie, F. D. & Farb, A. (2003) J Interv Cardiol 16, 267-72.

104. Virmani, R., Kolodgie, F. D., Burke, A. P., Finn, A. V., Gold, H. K., Tulenko, T. N., Wrenn, S. P. & Narula, J. (2005) Arterioscler Thromb Vasc Biol 25, 2054-61.

105. Robbie, L. & Libby, P. (2001) Ann N Y Acad Sci 947, 167-79; discussion 179-80.

106. Kolodgie, F. D., Narula, J., Guillo, P. & Virmani, R. (1999) Apoptosis 4, 5-10.

107. Hegyi, L., Hardwick, S. J., Siow, R. C. & Skepper, J. N. (2001) J Hematother Stem Cell Res 10, 27-42.

108. Akishima, Y., Akasaka, Y., Ishikawa, Y., Lijun, Z., Kiguchi, H., Ito, K., Itabe, H. & Ishii, T. (2005) Mod Pathol 18, 365-73.

109. Takahashi, K., Takeya, M. & Sakashita, N. (2002) Med Electron Microsc 35, 179- 203.

110. Geng, Y. J. & Libby, P. (1995) Am J Pathol 147, 251-66.

111. Choudhury, R. P., Lee, J. M. & Greaves, D. R. (2005) Nat Clin Pract Cardiovasc Med 2, 309-15.

112. Kietselaer, B. L., Reutelingsperger, C. P., Heidendal, G. A., Daemen, M. J., Mess, W. H., Hofstra, L. & Narula, J. (2004) N Engl J Med 350, 1472-3.

113. Kolodgie, F. D., Petrov, A., Virmani, R., Narula, N., Verjans, J. W., Weber, D. K., Hartung, D., Steinmetz, N., Vanderheyden, J. L., Vannan, M. A., Gold, H. K., Reutelingsperger, C. P., Hofstra, L. & Narula, J. (2003) Circulation 108, 3134-9.

114. Corti, R., Hutter, R., Badimon, J. J. & Fuster, V. (2004) J Thromb Thrombolysis 17, 35-44.

115. Avanzas, P., Arroyo-Espliguero, R., Garcia-Moll, X. & Kaski, J. C. (2005) Panminerva Med 47, 81-91.

116. Vivanco, F., Martin-Ventura, J. L., Duran, M. C., Barderas, M. G., Blanco-Colio, L., Darde, V. M., Mas, S., Meilhac, O., Michel, J. B., Tunon, J. & Egido, J. (2005) J Proteome Res 4, 1181-91.

117. Rouis, M. (2005) Curr Drug Targets Cardiovasc Haematol Disord 5, 541-8.

118. Kopka, K., Breyholz, H. J., Wagner, S., Law, M. P., Riemann, B., Schroer, S., Trub, M., Guilbert, B., Levkau, B., Schober, O. & Schafers, M. (2004) Nucl Med Biol 31, 257-67.

119. Chen, J., Tung, C. H., Allport, J. R., Chen, S., Weissleder, R. & Huang, P. L. (2005) Circulation 111, 1800-5.

228 120. Chen, J., Tung, C. H., Mahmood, U., Ntziachristos, V., Gyurko, R., Fishman, M. C., Huang, P. L. & Weissleder, R. (2002) Circulation 105, 2766-71.

121. Nicholls, S. J. & Hazen, S. L. (2005) Arterioscler Thromb Vasc Biol 25, 1102-11.

122. Querol Sans, M., Chen, J. W., Weissleder, R. & Bogdanov, A. A., Jr. (2005) Mol Imaging Biol 7, 403-10.

123. Winter, P. M., Morawski, A. M., Caruthers, S. D., Fuhrhop, R. W., Zhang, H., Williams, T. A., Allen, J. S., Lacy, E. K., Robertson, J. D., Lanza, G. M. & Wickline, S. A. (2003) Circulation 108, 2270-4.

124. Zhou, X. & Hansson, G. K. (1999) Scand J Immunol 50, 25-30.

125. Hosono, M., de Boer, O. J., van der Wal, A. C., van der Loos, C. M., Teeling, P., Piek, J. J., Ueda, M. & Becker, A. E. (2003) Atherosclerosis 168, 73-80.

126. Morawietz, H., Rueckschloss, U., Niemann, B., Duerrschmidt, N., Galle, J., Hakim, K., Zerkowski, H. R., Sawamura, T. & Holtz, J. (1999) Circulation 100, 899-902.

127. Charo, I. F. & Peters, W. (2003) Microcirculation 10, 259-64.

128. Inaba, T., Yamada, N., Gotoda, T., Shimano, H., Shimada, M., Momomura, K., Kadowaki, T., Motoyoshi, K., Tsukada, T., Morisaki, N. & et al. (1992) J Biol Chem 267, 5693-9.

129. Wenzel, K., Felix, S., Kleber, F. X., Brachold, R., Menke, T., Schattke, S., Schulte, K. L., Glaser, C., Rohde, K., Baumann, G. & et al. (1994) Hum Mol Genet 3, 1935-7.

130. Giuffre, L., Cordey, A. S., Monai, N., Tardy, Y., Schapira, M. & Spertini, O. (1997) J Cell Biol 136, 945-56.

131. Blankenberg, S., Barbaux, S. & Tiret, L. (2003) Atherosclerosis 170, 191-203.

132. Calderon, T. M., Factor, S. M., Hatcher, V. B., Berliner, J. A. & Berman, J. W. (1994) Lab Invest 70, 836-49.

133. Rekhter, M. D. (2002) Cardiovasc Res 54, 36-41.

134. Cullen, P., Baetta, R., Bellosta, S., Bernini, F., Chinetti, G., Cignarella, A., von Eckardstein, A., Exley, A., Goddard, M., Hofker, M., Hurt-Camejo, E., Kanters, E., Kovanen, P., Lorkowski, S., McPheat, W., Pentikainen, M., Rauterberg, J., Ritchie, A., Staels, B., Weitkamp, B. & de Winther, M. (2003) Arterioscler Thromb Vasc Biol 23, 535-42.

135. Kondo, T. & Watanabe, Y. (1975) Jikken Dobutsu 24, 89-94.

229 136. Aliev, G., Castellani, R. J., Petersen, R. B., Burnstock, G., Perry, G. & Smith, M. A. (2004) J Submicrosc Cytol Pathol 36, 225-40.

137. Lawn, R. M., Wade, D. P., Hammer, R. E., Chiesa, G., Verstuyft, J. G. & Rubin, E. M. (1992) Nature 360, 670-2.

138. Reddick, R. L., Zhang, S. H. & Maeda, N. (1998) Atherosclerosis 140, 297-305.

139. Sasaki, T., Kuzuya, M., Nakamura, K., Cheng, X. W., Shibata, T., Sato, K. & Iguchi, A. (2006) Arterioscler Thromb Vasc Biol.

140. Gertz, S. D., Fallon, J. T., Gallo, R., Taubman, M. B., Banai, S., Barry, W. L., Gimple, L. W., Nemerson, Y., Thiruvikraman, S., Naidu, S. S., Chesebro, J. H., Fuster, V., Sarembock, I. J. & Badimon, J. J. (1998) Circulation 98, 580-7.

141. Eitzman, D. T., Westrick, R. J., Xu, Z., Tyson, J. & Ginsburg, D. (2000) Arterioscler Thromb Vasc Biol 20, 1831-4.

142. Constantinides, P. & Chakravarti, R. N. (1961) Arch Pathol 72, 197-208.

143. Caligiuri, G., Levy, B., Pernow, J., Thoren, P. & Hansson, G. K. (1999) Proc Natl Acad Sci U S A 96, 6920-4.

144. Prescott, M. F., McBride, C. H., Hasler-Rapacz, J., Von Linden, J. & Rapacz, J. (1991) Am J Pathol 139, 139-47.

145. Rosenfeld, M. E., Polinsky, P., Virmani, R., Kauser, K., Rubanyi, G. & Schwartz, S. M. (2000) Arterioscler Thromb Vasc Biol 20, 2587-92.

146. Johnson, J. L. & Jackson, C. L. (2001) Atherosclerosis 154, 399-406.

147. Herrera, V. L., Makrides, S. C., Xie, H. X., Adari, H., Krauss, R. M., Ryan, U. S. & Ruiz-Opazo, N. (1999) Nat Med 5, 1383-9.

148. Herrera, V. M., Didishvili, T., Lopez, L. V., Zander, K., Traverse, S., Gantz, D., Herscovitz, H. & Ruiz-Opazo, N. (2001) Mol Med 7, 831-44.

149. Calara, F., Silvestre, M., Casanada, F., Yuan, N., Napoli, C. & Palinski, W. (2001) J Pathol 195, 257-63.

150. Rosenfeld, M. E., Carson, K. G., Johnson, J. L., Williams, H., Jackson, C. L. & Schwartz, S. M. (2002) Curr Atheroscler Rep 4, 238-42.

151. Jawien, J., Nastalek, P. & Korbut, R. (2004) J Physiol Pharmacol 55, 503-17.

152. Watanabe, Y., Ito, T. & Shiomi, M. (1985) Atherosclerosis 56, 71-9.

153. Watanabe, Y. (1980) Atherosclerosis 36, 261-8.

230 154. Ito, T., Tsukada, T., Ueda, M., Wanibuchi, H. & Shiomi, M. (1994) J Atheroscler Thromb 1, 45-52.

155. Shiomi, M., Ito, T., Tsukada, T., Yata, T. & Ueda, M. (1994) Arterioscler Thromb 14, 931-7.

156. Ito, T., Yamada, S. & Shiomi, M. (2004) Exp Anim 53, 339-46.

157. Shiomi, M., Ito, T., Yamada, S., Kawashima, S. & Fan, J. (2003) Arterioscler Thromb Vasc Biol 23, 1239-44.

158. Ito, T., Yamada, S., Tamura, T. & Shiomi, M. (2005) Exp Anim 54, 413-9.

159. (2006) (American Cancer Society, Atlanta).

160. Ries LA, E. M., Kosary CL, et al. (2002) (National Cancer Institute, Bethesda, Maryland).

161. Mintzer, D., Glassburn, J., Mason, B. A. & Sataloff, D. (2002) Oncologist 7, 547- 54.

162. (2002) in AJCC Cancer Staging Manual (Springer, New York, NY), pp. p. 171- 180.

163. Singletary, S. E., Allred, C., Ashley, P., Bassett, L. W., Berry, D., Bland, K. I., Borgen, P. I., Clark, G., Edge, S. B., Hayes, D. F., Hughes, L. L., Hutter, R. V., Morrow, M., Page, D. L., Recht, A., Theriault, R. L., Thor, A., Weaver, D. L., Wieand, H. S. & Greene, F. L. (2002) J Clin Oncol 20, 3628-36.

164. Verkooijen, H., Peterse, J., Schipper, M., Buskens, E., Hendriks, J., Pijnappel, R., Peeters, P., Rinkes, B., Mali, W. & Holland, R. (2003) European Journal of Cancer 39, 2187-2191.

165. Dunne, B. & Going, J. J. (2001) Histopathology 39, 259-265.

166. Piver, M., Tsukada, Y., Werness, B., DiCioccio, R., Whittemore, A. & Ponder, B. (2000) Gynecologic Oncology 78, 166-170.

167. Schlemper, R., Dawsey, S., Itabashi, M., Iwashita, A., Kato, Y., Koike, M., Lewin, K., Riddell, R., Shimoda, T., Sipponen, P., Stolte, M. & Watanabe, H. (2000) Cancer 88, 996-1006.

168. Tsuda, H., Sasano, H., Akiyama, F., Kurosumi, M., Hasegawa, T., Osamura, R. & Sakamoto, G. (2002) Pathology International 52, 126-134.

169. Teh, W. & Wilson, A. R. (1998) Eur J Cancer 34, 449-50.

170. Agnese, D. M. (2005) Surg Technol Int 14, 51-6.

231 171. Kriege, M., Brekelmans, C. T., Boetes, C., Besnard, P. E., Zonderland, H. M., Obdeijn, I. M., Manoliu, R. A., Kok, T., Peterse, H., Tilanus-Linthorst, M. M., Muller, S. H., Meijer, S., Oosterwijk, J. C., Beex, L. V., Tollenaar, R. A., de Koning, H. J., Rutgers, E. J. & Klijn, J. G. (2004) N Engl J Med 351, 427-37.

172. Warner, E., Plewes, D. B., Hill, K. A., Causer, P. A., Zubovits, J. T., Jong, R. A., Cutrara, M. R., DeBoer, G., Yaffe, M. J., Messner, S. J., Meschino, W. S., Piron, C. A. & Narod, S. A. (2004) Jama 292, 1317-25.

173. Lawrence, W. F., Liang, W., Mandelblatt, J. S., Gold, K. F., Freedman, M., Ascher, S. M., Trock, B. J. & Chang, P. (1998) J Natl Cancer Inst 90, 1792-800.

174. Mathieu, I., Mazy, S., Willemart, B., Destine, M., Mazy, G. & Lonneux, M. (2005) J Nucl Med 46, 1574-81.

175. Yarden, Y. & Sliwkowski, M. X. (2001) Nat Rev Mol Cell Biol 2, 127-37.

176. Karunagaran, D., Tzahar, E., Beerli, R. R., Chen, X., Graus-Porta, D., Ratzkin, B. J., Seger, R., Hynes, N. E. & Yarden, Y. (1996) Embo J 15, 254-64.

177. Tzahar, E., Waterman, H., Chen, X., Levkowitz, G., Karunagaran, D., Lavi, S., Ratzkin, B. J. & Yarden, Y. (1996) Mol Cell Biol 16, 5276-87.

178. Chorn, N. (2006) Oncol Nurs Forum 33, 265-72.

179. Ross, J. S., Fletcher, J. A., Linette, G. P., Stec, J., Clark, E., Ayers, M., Symmans, W. F., Pusztai, L. & Bloom, K. J. (2003) Oncologist 8, 307-25.

180. Dowsett, M. (2001) Endocr Relat Cancer 8, 191-5.

181. Schnitt, S. J. (2001) Mod Pathol 14, 213-8.

182. el-Ahmady, O., el-Salahy, E., Mahmoud, M., Wahab, M. A., Eissa, S. & Khalifa, A. (2002) Anticancer Res 22, 2493-9.

183. Masood, S. & Bui, M. M. (2002) Microsc Res Tech 59, 102-8.

184. Press, M. F., Hung, G., Godolphin, W. & Slamon, D. J. (1994) Cancer Res 54, 2771-7.

185. Rhodes, A., Jasani, B., Anderson, E., Dodson, A. & Balaton, A. (2002) American Journal of Clinical Pathology 118, 408-17.

186. Paik, S., Bryant, J., Tan-Chiu, E., Romond, E., Hiller, W., Park, K., Brown, A., Yothers, G., Anderson, S., Smith, R., Wickerham, D. L. & Wolmark, N. (2002) J Natl Cancer Inst 94, 852-4.

187. Rhodes, A., Borthwick, D., Sykes, R., Al-Sam, S. & Paradiso, A. (2004) Am J Clin Pathol 122, 51-60.

232 188. Mass, R., Press, M. F., Anderson, S. & Slamon, D. J. (2001) Breast Cancer Research and Treatment 69, 213-213.

189. Mass, R. D., Press, M. F., Anderson, S., Cobleigh, M. A., Vogel, C. L., Dybdal, N., Leiberman, G. & Slamon, D. J. (2005) Clin Breast Cancer 6, 240-6.

190. Vogel, C., Cobleigh, M., Tripathy, D., Gutheil, J., Harris, L., Fehrenbacher, L., Slamon, D. J., Murphy M, Novotny, W., Burchmore, M., Shak, S., Stewart, S. & Press, M. F. (2002) Journal of Clinical Oncology 20, 719-726.

191. Carter, P., Presta, L., Gorman, C. M., Ridgway, J. B., Henner, D., Wong, W. L., Rowland, A. M., Kotts, C., Carver, M. E. & Shepard, H. M. (1992) Proc Natl Acad Sci U S A 89, 4285-9.

192. Flaherty, K. T. & Brose, M. S. (2006) Curr Oncol Rep 8, 90-5.

193. Toi, M., Horiguchi, K., Bando, H., Saji, S. & Chow, L. W. (2005) Cancer Chemother Pharmacol 56 Suppl 1, 94-9.

194. Albanell, J., Codony, J., Rovira, A., Mellado, B. & Gascon, P. (2003) Adv Exp Med Biol 532, 253-68.

195. Thor, A. D., Liu, S., Edgerton, S., Moore, D., 2nd, Kasowitz, K. M., Benz, C. C., Stern, D. F. & DiGiovanna, M. P. (2000) J Clin Oncol 18, 3230-9.

196. Norris, P., Noble, M., Francolini, I., Vinogradov, A. M., Stewart, P. S., Ratner, B. D., Costerton, J. W. & Stoodley, P. (2005) Antimicrob Agents Chemother 49, 4272-9.

197. Smith, A. P., Hall, P. A. & Marcello, D. M. (2004) Radiol Manage 26, 16-24; quiz 25-7.

198. Copland, J. A., Eghtedari, M., Popov, V. L., Kotov, N., Mamedova, N., Motamedi, M. & Oraevsky, A. A. (2004) Mol Imaging Biol 6, 341-9.

199. Ferrari, M., Granik, V. T., Imam, A. & Nadeau, J. C. (1997) Advances in Doublet Mechanics (Springer-Verlag New York, Inc., New York).

200. Liu, J. (2002) in Biomedical Engineering (The Ohio State University, Columbus).

201. Liu, J. & Ferrari, M. (2002) Disease Markers 18, 175-183.

202. Liu, J. & Ferrari, M. (2003) CMES 4, 421-430.

203. Ferrari, M. (2000) Biomedical Microdevices 2, 273-281.

204. Lavrentyev, A. & Rokhlin, S. I. (1997) J. Acoust. Soc. Am. 102, 3467-3477.

205. Wang, L., Xie, B. & Rokhlin, S. I. (2002) J. Acoust. Soc. Am. 111, 2644-2653.

233 206. Qin, G., Zhang, Y., Cao, W., An, R., Gao, Z., Li, G., Xu, W., Zhang, K. & Li, S. (2005) Eur J Nucl Med Mol Imaging 32, 6-14.

207. Bhatia, V., Bhatia, R., Dhindsa, S. & Virk, A. (2003) J Postgrad Med 49, 361-8.

208. Hutter, R., Valdiviezo, C., Sauter, B. V., Savontaus, M., Chereshnev, I., Carrick, F. E., Bauriedel, G., Luderitz, B., Fallon, J. T., Fuster, V. & Badimon, J. J. (2004) Circulation 109, 2001-8.

209. Libby, P. (1995) Circulation 91, 2844-50.

210. Ball, R. Y., Stowers, E. C., Burton, J. H., Cary, N. R., Skepper, J. N. & Mitchinson, M. J. (1995) in Atherosclerosis, Vol. 114, pp. 45-54.

211. Tabas, I. (2004) in Cell Death Differ, Vol. 11 Suppl 1, pp. S12-6.

212. Libby, P., Geng, Y. J., Aikawa, M., Schoenbeck, U., Mach, F., Clinton, S. K., Sukhova, G. K. & Lee, R. T. (1996) in Curr Opin Lipidol, Vol. 7, pp. 330-5.

213. Hutter, R., Valdiviezo, C., Sauter, B. V., Savontaus, M., Chereshnev, I., Carrick, F. E., Bauriedel, G., Luderitz, B., Fallon, J. T., Fuster, V. & Badimon, J. J. (2004) in Circulation, Vol. 109, pp. 2001-8.

214. Kolodgie, F. D., Burke, A. P., Farb, A., Gold, H. K., Yuan, J., Narula, J., Finn, A. V. & Virmani, R. (2001) in Curr Opin Cardiol, Vol. 16, pp. 285-92.

215. Moldovan, N. I., Moldovan, L. & Simionescu, N. (1994) Blood Coagul Fibrinolysis 5, 921-8.

216. Huber, R., Berendes, R., Burger, A., Schneider, M., Karshikov, A., Luecke, H., Romisch, J. & Paques, E. (1992) J Mol Biol 223, 683-704.

217. Coil, D. A. & Miller, A. D. (2004) J Virol 78, 10920-6.

218. Thiagarajan, P. & Benedict, C. R. (1997) Circulation 96, 2339-47.

219. Mazzone, A. & Ricevuti, G. (1995) Haematologica 80, 161-75.

220. Berliner, S., Rogowski, O., Rotstein, R., Fusman, R., Shapira, I., Bornstein, N. M., Prochorov, V., Roth, A., Keren, G., Eldor, A. & Zeltser, D. (2000) in Cardiology, Vol. 94, pp. 19-25.

221. Kassirer, M., Zeltser, D., Prochorov, V., Schoenman, G., Frimerman, A., Keren, G., Shapira, I., Miller, H., Roth, A., Arber, N., Eldor, A. & Berliner, S. (1999) in Am Heart J, Vol. 138, pp. 555-9.

222. Kakutani, M., Ueda, M., Naruko, T., Masaki, T. & Sawamura, T. (2001) Biochem Biophys Res Commun 282, 180-5.

234 223. Chen, M., Masaki, T. & Sawamura, T. (2002) Pharmacol Ther 95, 89-100.

224. Kume, N. (2002) Nippon Ronen Igakkai Zasshi 39, 264-7.

225. Li, D. Y., Chen, H. J., Staples, E. D., Ozaki, K., Annex, B., Singh, B. K., Vermani, R. & Mehta, J. L. (2002) J Cardiovasc Pharmacol Ther 7, 147-53.

226. McEvoy, L. M., Jutila, M. A., Tsao, P. S., Cooke, J. P. & Butcher, E. C. (1997) Blood 90, 3587-94.

227. Tuomisto, T. T., Korkeela, A., Rutanen, J., Viita, H., Brasen, J. H., Riekkinen, M. S., Rissanen, T. T., Karkola, K., Kiraly, Z., Kolble, K. & Yla-Herttuala, S. (2003) Arterioscler Thromb Vasc Biol 23, 2235-40.

228. Rekhter, M. D. (1999) Cardiovasc Res 41, 376-84.

229. Rekhter, M. D., Zhang, K., Narayanan, A. S., Phan, S., Schork, M. A. & Gordon, D. (1993) Am J Pathol 143, 1634-48.

230. Bode, M. K., Mosorin, M., Satta, J., Risteli, L., Juvonen, T. & Risteli, J. (1999) in Arterioscler Thromb Vasc Biol, Vol. 19, pp. 1506-11.

231. Melian, A., Geng, Y. J., Sukhova, G. K., Libby, P. & Porcelli, S. A. (1999) Am J Pathol 155, 775-86.

232. Geng, Y. J. & Libby, P. (2002) Arterioscler Thromb Vasc Biol 22, 1370-80.

233. Lind, L. (2003) Atherosclerosis 169, 203-14.

234. Harrington, J. R. (2000) Stem Cells 18, 65-6.

235. Atalar, E., Aytemir, K., Haznedaroglu, I., Ozer, N., Ovunc, K., Aksoyek, S., Kes, S., Kirazli, S. & Ozmen, F. (2001) Int J Cardiol 78, 69-73.

236. Saving, K. L. & Mankin, P. E. (2003) J Pediatr Hematol Oncol 25, 266-9.

237. Sohma, Y., Suzuki, T., Sasano, H., Nagura, H., Nose, M. & Yamamoto, T. (1994) Cell Struct Funct 19, 219-25.

238. Yilmaz, A., Lochno, M., Traeg, F., Cicha, I., Reiss, C., Stumpf, C., Raaz, D., Anger, T., Amann, K., Probst, T., Ludwig, J., Daniel, W. G. & Garlichs, C. D. (2004) Atherosclerosis 176, 101-10.

239. Narula, J. & Strauss, H. W. (2005) Eur J Nucl Med Mol Imaging 32, 1-5.

240. Zhdanov, V. S., Chumachenko, P. V. & Drobkova, I. P. (2004) Kardiologiia 44, 40-4.

235 241. Schellenberger, E. A., Bogdanov, A., Jr., Hogemann, D., Tait, J., Weissleder, R. & Josephson, L. (2002) Mol Imaging 1, 102-7.

242. Gavrieli, Y., Sherman, Y. & Ben-Sasson, S. A. (1992) J Cell Biol 119, 493-501.

243. McAuliffe, M., Lalonde, F., McGarry, D., Gandler, W., Csaky, K. & Trus, B. (2001) in IEEE Computer-Based Medical Systems (CBMS), pp. 381-386.

244. Yang, D. J., Azhdarinia, A., Wu, P., Yu, D. F., Tansey, W., Kalimi, S. K., Kim, E. E. & Podoloff, D. A. (2001) Cancer Biother Radiopharm 16, 73-83.

245. Lanza, G. M., Winter, P., Caruthers, S., Schmeider, A., Crowder, K., Morawski, A., Zhang, H., Scott, M. J. & Wickline, S. A. (2004) Curr Pharm Biotechnol 5, 495-507.

246. Hyafil, F., Laissy, J. P., Mazighi, M., Tchetche, D., Louedec, L., Adle-Biassette, H., Chillon, S., Henin, D., Jacob, M. P., Letourneur, D. & Feldman, L. J. (2006) Arterioscler Thromb Vasc Biol 26, 176-81.

247. Mirrashed, F., Sharp, J. C., Cheung, I. & Tomanek, B. (2004) Magma 16, 167-73.

248. Pirko, I., Johnson, A., Ciric, B., Gamez, J., Macura, S. I., Pease, L. R. & Rodriguez, M. (2004) Faseb J 18, 179-82.

249. Ludewig, B. & Laman, J. D. (2004) Proc Natl Acad Sci U S A 101, 11529-30.

250. Lijnen, H. R. (2002) Biochem Soc Trans 30, 163-7.

251. Li, W., Hellsten, A., Jacobsson, L. S., Blomqvist, H. M., Olsson, A. G. & Yuan, X. M. (2004) J Mol Cell Cardiol 37, 969-78.

252. Hartung, D., Sarai, M., Petrov, A., Kolodgie, F., Narula, N., Verjans, J., Virmani, R., Reutelingsperger, C., Hofstra, L. & Narula, J. (2005) J Nucl Med 46, 2051-6.

253. Jaffer, F. A. & Weissleder, R. (2004) Circ Res 94, 433-45.

254. Mani, V., Briley-Saebo, K. C., Itskovich, V. V., Samber, D. D. & Fayad, Z. A. (2006) Magn Reson Med 55, 126-35.

255. Hegyi, L., Skepper, J. N., Cary, N. R. & Mitchinson, M. J. (1996) J Pathol 180, 423-9.

256. Han, D. K., Haudenschild, C. C., Hong, M. K., Tinkle, B. T., Leon, M. B. & Liau, G. (1995) Am J Pathol 147, 267-77.

257. Kockx, M. M. & Herman, A. G. (1998) Eur Heart J 19 Suppl G, G23-8.

258. Tabas, I. (2005) Arterioscler Thromb Vasc Biol 25, 2255-64.

236 259. Shiomi, M., Ito, T., Yamada, S., Kawashima, S. & Fan, J. (2004) J Atheroscler Thromb 11, 184-9.

260. Hartung, D. & Narula, J. (2004) Z Kardiol 93, 97-102.

261. Davies, J. R., Rudd, J. H., Weissberg, P. L. & Narula, J. (2006) J Am Coll Cardiol 47, C57-68.

262. Wilensky, R. L., Song, H. K. & Ferrari, V. A. (2006) J Am Coll Cardiol 47, C48- 56.

263. Sirol, M., Fuster, V., Toussaint, J. F. & Fayad, Z. A. (2003) Arch Mal Coeur Vaiss 96, 1219-24.

264. Moreno, P. R., Lodder, R. A., Purushothaman, K. R., Charash, W. E., O'Connor, W. N. & Muller, J. E. (2002) Circulation 105, 923-7.

265. Lanza, G. M., Winter, P. M., Caruthers, S. D., Morawski, A. M., Schmieder, A. H., Crowder, K. C. & Wickline, S. A. (2004) J Nucl Cardiol 11, 733-43.

266. Lee, S. C., Bhalerao, K. & Ferrari, M. (2004) Annals of the New York Academy of Sciences 1013, 110-123.

267. Koeppen, H. K. W., Wright, B. D., Burt, A. D., Quirke, P., McNicol, A. M., Dybdal, N. O., Sliwkowski, M. X. & Hillan, K. J. (2001) Histopathology 38, 96- 104.

268. Sakamoto, J. (2005) in Biomedical Engineering (The Ohio State University, Columbus), pp. 233.

269. Artemov, D., Mori, N., Okollie, B. & Bhujwalla, Z. M. (2003) Magn Reson Med 49, 403-8.

270. Artemov, D., Mori, N., Ravi, R. & Bhujwalla, Z. M. (2003) Cancer Res 63, 2723- 7.

271. Ito, A., Kuga, Y., Honda, H., Kikkawa, H., Horiuchi, A., Watanabe, Y. & Kobayashi, T. (2004) Cancer Lett 212, 167-75.

272. Funovics, M. A., Kapeller, B., Hoeller, C., Su, H. S., Kunstfeld, R., Puig, S. & Macfelda, K. (2004) Magn Reson Imaging 22, 843-50.

273. Montet, X., Ntziachristos, V., Grimm, J. & Weissleder, R. (2005) Cancer Res 65, 6330-6.

274. Sakamoto, J. H., Smith, B. R., Xie, B., Rokhlin, S. I., Lee, S. C. & Ferrari, M. (2005) Technol Cancer Res Treat 4, 627-36.

237 275. Sauer-Eriksson, A. E., Kleywegt, G. J., Uhlen, M. & Jones, T. A. (1995) Structure 3, 265-78.

276. Oda, M., Kozono, H., Morii, H. & Azuma, T. (2003) Int Immunol 15, 417-26.

277. Chen, Y. H., Cai, J. J. & Lu, Y. D. (1994) Zhonghua Zhong Liu Za Zhi 16, 266-8.

278. Ahrens, E. T., Feili-Hariri, M., Xu, H., Genove, G. & Morel, P. A. (2003) Magn Reson Med 49, 1006-13.

279. Josephson, L., Groman, E. V., Menz, E., Lewis, J. M. & Bengele, H. (1990) Magn Reson Imaging 8, 637-46.

280. Weissleder, R., Elizondo, G., Wittenberg, J., Lee, A. S., Josephson, L. & Brady, T. J. (1990) Radiology 175, 494-8.

281. Weissleder, R., Elizondo, G., Wittenberg, J., Rabito, C. A., Bengele, H. H. & Josephson, L. (1990) Radiology 175, 489-93.

282. Wang, Y. X., Hussain, S. M. & Krestin, G. P. (2001) Eur Radiol 11, 2319-31.

283. Bengele, H. H., Palmacci, S., Rogers, J., Jung, C. W., Crenshaw, J. & Josephson, L. (1994) Magn Reson Imaging 12, 433-42.

284. Rogers, J., Lewis, J. & Josephson, L. (1994) Invest Radiol 29 Suppl 2, S81-2.

285. Rogers, J., Lewis, J. & Josephson, L. (1994) Magn Reson Imaging 12, 631-9.

286. Reynolds, F., O'Loughlin, T., Weissleder, R. & Josephson, L. (2005) Anal Chem 77, 814-7.

287. Moore, A., Marecos, E., Bogdanov, A., Jr. & Weissleder, R. (2000) Radiology 214, 568-74.

288. Josephson, L., Bigler, J. & White, D. (1991) Magn Reson Med 22, 204-8; discussion 213-5.

289. Sonvico, F., Mornet, S., Vasseur, S., Dubernet, C., Jaillard, D., Degrouard, J., Hoebeke, J., Duguet, E., Colombo, P. & Couvreur, P. (2005) Bioconjug Chem 16, 1181-8.

290. Jain, T. K., Morales, M. A., Sahoo, S. K., Leslie-Pelecky, D. L. & Labhasetwar, V. (2005) Mol Pharm 2, 194-205.

291. Kohler, N., Sun, C., Wang, J. & Zhang, M. (2005) Langmuir 21, 8858-64.

292. Kircher, M. F., Weissleder, R. & Josephson, L. (2004) Bioconjug Chem 15, 242- 8.

238 293. Josephson, L., Kircher, M. F., Mahmood, U., Tang, Y. & Weissleder, R. (2002) Bioconjug Chem 13, 554-60.

294. Zhao, M., Kircher, M. F., Josephson, L. & Weissleder, R. (2002) Bioconjug Chem 13, 840-4.

295. Smith, P. K., Krohn, R. I., Hermanson, G. T., Mallia, A. K., Gartner, F. H., Provenzano, M. D., Fujimoto, E. K., Goeke, N. M., Olson, B. J. & Klenk, D. C. (1985) Anal Biochem 150, 76-85.

296. Kaminska, K., Brown, T., Beydaghyan, G. & Robbie, K. (2003) Appl Opt 42, 4212-9.

297. De Stefano, L., Rendina, I., Moretti, L., Tundo, S. & Rossi, A. M. (2004) Appl Opt 43, 167-72.

298. Cunin, F., Schmedake, T. A., Link, J. R., Li, Y. Y., Koh, J., Bhatia, S. N. & Sailor, M. J. (2002) Nat Mater 1, 39-41.

299. Cullis, A. G., Canham, L. T. & Calcott, P. D. J. (1997) J. Appl. Physics 82, 909- 915.

300. Stewart, M. P., Robins, E. G., Geders, T. W., Allen, M. J., Choi, H. C. & Buriak, J. M. (2000) phys. stat. sol. (a) 182, 109-115.

301. Liotta, L. A., Ferrari, M. & Petricoin, E. (2003) Nature 425, 905.

302. Riemer, J., Hoepken, H. H., Czerwinska, H., Robinson, S. R. & Dringen, R. (2004) Anal Biochem 331, 370-5.

303. Kenning, G. G., Rodriguez, R., Zotev, V. S., Moslemi, A., Wilson, S., Hawel, L., Byus, C. & Kovach, J. S. (2005) Review of Scientific Instruments 76, 014303.

304. Pronk, J. T., Liem, K., Bos, P. & Kuenen, J. G. (1991) Appl Environ Microbiol 57, 2063-2068.

239