Dynamic Contrast-Enhanced Magnetic Resonance Imaging & Fluorescence Microscopy of Tumor Microvascular Permeability

Item Type text; Electronic Dissertation

Authors Jennings, Dominique Louise

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 03/10/2021 18:57:45

Link to Item http://hdl.handle.net/10150/193555

DYNAMIC CONTRASTENHANCED MAGNETIC RESONANCE IMAGING &

FLUORESCENCE MICROSCOPY OF TUMOR MICROVASCULAR

PERMEABILITY

Dominique Louise Jennings

______

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF BIOMEDICAL ENGINEERING

In Partial Fulfillment of the Requirements For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2008

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by: Dominique L. Jennings entitled: Dynamic ContrastEnhanced Magnetic Resonance Imaging & Fluorescence Microscopy of Tumor Microvascular Permeability and recommend that it be accepted as fulfilling the dissertation requirement for the

Degree of Doctor of Philosophy

______Date: 01/16/08 Robert J. Gillies

______Date: 01/16/08 Theodore P. Trouard

______Date: 01/16/08 Natarajan Raghunand

______Date: 01/16/08 Robert A. Gatenby

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: 01/16/08 Dissertation Director: Robert J. Gillies

______Date: 01/16/08 CoDissertation Director: Theodore P. Trouard

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: Dominique L. Jennings

ACKNOWLEDGMENTS

I would like to thank my mentor and sole collegiate advisor for his patience, mentorship, and for training me to be independent, the most valuable lesson I learned in academic research. Through his training, I have learned to persevere in spite of disappointing and unexpected experimental results and appreciate those that were positive, but most importantly to “treat those two imposters just the same” ( §).

I would like to thank another close mentor, Natarajan Raghunand, for endlessly encouraging me to understand more and to be a better student. Despite my fears and frustration with concepts such as transcytolemmal water exchange, the challenge of doing so has elevated my standard of expectations for myself and the scientific method.

I would like to attribute my success with the window chamber experiments to Bethany Skovan for her meticuolous and tireless efforts to transform the window chamber preparation into the success that it is today. The experiments that comprise this dissertation would not have been possible without her surgical talent.

I would to thank my professor and mentor, Ted Trouard, for his enthusiasm patience for teaching NMR and especially MRI, making it an exciting concept to learn. I would also like to thank him for showing us that it was alright to be confused, to ask more questions when you are because an attempt to understand the lesson is a contribution to that lesson. As one of the smartest scientists I know, I have taken this lesson in earnest.

There are many people that contributed to my technical learning of NMR and MRI. I would like to thank Constantine Job for long tutorials on NMR/MRI, scanner hardware and coil electronics, but mostly for his friendship. I would also like to thank Jingyu Guo for his patience and willingness to teach an undergraduate Biochemistry student the basic, functional aspects of running an animal research spectrometer.

Nonacademic thanks go to my family for their support, especially my brother for convincing me to go to college and become more than I thought I deserved to be, and my mother for always helping and supporting me with her unconditional love. Finally, I have to thank my husband for his patience, understanding, love, but most of all for his contribution to the scientist I am today and the accomplishments I could not have made without this support. I would never have come this far without him.

§ Rudyard Kipling (1895). If—. In Rewards and Fairies .

DEDICATION

Dedicated to my husband, Nathaniel D. Kirkpatrick, Ph.D.

TABLE OF CONTENTS

1. LIST OF ILLUSTRATIONS ...... 9 2. LIST OF TABLES ...... 11 3. ABSTRACT...... 12 4. CHAPTER 1...... 13 Section 1 1.1. Defining Vascular Permeability and Perfusion...... 13 1.2. Macrocirculation – Overview...... 13 1.2.1. Morphology...... 13 1.2.2. Vessel Regulation of Flow...... 14 1.3. Capillary Microcirculation ...... 16 1.3.1. Passive Capillary Transport ...... 18 1.4. Angiogenesis ...... 19 1.4.1. Molecular Mechanisms of Angiogenesis...... 21 1.4.2. Microcirculatory Assays ...... 23 1.4.3. Matrix Implant Assays...... 25 1.4.4. Ex Vivo Assays...... 26 1.5. Tumor angiogenesis ...... 26 1.6. Using Imaging to Estimate Microvascular Permeability ...... 28 Section 2 2.1. Imaging the Tumor Microvasculature: Survey of Methods...... 30 2.2. Measurement of Hemodynamics with PET/SPECT ...... 30 2.2.1. Response to Antivascular Therapy ...... 31 2.2.2. Response to Cytotoxic Therapy...... 32 2.2.3. Kinetic Modeling of PET/SPECT Data ...... 35 2.3. Measurement of Hemodynamics with CT ...... 36 2.3.1. Kinetic Modeling of Perfusion CT Data...... 39 2.4. Measurement of Hemodynamics with Ultrasound...... 40

2.4.1. Modeling of Perfusion Ultrasound Data...... 44 2.5. Measurement of Hemodynamics with Optical Imaging ...... 45 2.5.1. Kinetic Modeling of Dynamic Optical Imaging Data...... 48 2.6. Nuclear Magnetic Resonance...... 49 2.7. Magnetic Resonance Imaging...... 56 2.8. Measurement of Hemodynamics with DCEMRI...... 58 2.8.1. MRI Contrast Agents ...... 59 2.8.2. Kinetic Modeling of DynamicContrast Enhanced MRI Data ...... 60 2.8.3. Transcytolemmal Water Exchange...... 63 Section 3 3.1. Antiangiogenic & Antivascular Therapies...... 72 3.2. Imaging Response to Antivascular vs. Antiangiogenic Therapies...... 74 Section 4 4.1. Endothelial Transport of Contrast Agents: Macromolecular Transport ...... 75 4.1.1. Paracellular Transport...... 76 4.1.2. Transcellular Transport...... 77 4.1.2.1. Caveolae ...... 77 4.1.2.2. Transendothelial Channels ...... 78 4.1.2.3. Vacuolar Organelles ...... 78 Conclusions...... 80 5. CHAPTER 2...... 93 6. CHAPTER 3...... 120 7. CHAPTER 4...... 143 8. CONCLUDING REMARKS ...... 163 APPENDICES AD 9. APPENDIX A ...... 164 9.1. Derivation of the full 2compartment, 3parameter and SI Methods...... 164 10. APPENDIX B ...... 169 10.1. MRI Data Acquisition Flowchart...... 169

11. APPENDIX C ...... 170 11.1. Graphical Data Analysis Flowchart ...... 170 12. APPENDIX D ...... 171 12.1. Fluorescence Microscopy Data Acquisition Flowchart ...... 171 13. APPENDIX E...... 172 13.1. Graphical Data Analysis Flowchart ...... 172 14. REFERENCES...... 173

LIST OF ILLUSTRATIONS

FIGURE 1.1, Anatomical crosssection of major vessels...... 82 FIGURE 1.2, Endothelial cell transport mechanisms ...... 83 FIGURE 2.1a, T1weighted series of biotinBSAGdDTPA phantom...... 110 FIGURE 2.1b, Recovery curves for biotinBSAGdDTPA phantom...... 110

FIGURE 2.1c, Fit for r1 relaxivity ...... 110 FIGURE 2.2a, T2weighted series of biotinBSAGdDTPA phantom...... 110 FIGURE 2.2b, Decay curves for biotinBSAGdDTPA phantom...... 111

FIGURE 2.2c, Fit for r2 relaxivity ...... 111 FIGURE 2.3, Polyacetal resin & titanium window chambers ...... 111 FIGURE 2.4, Polyacetal resin chamber implanted on animal ...... 112 FIGURE 2.5, Transillumination images overlaid with GFP fluorescence...... 112 FIGURE 2.6, Representative DCEMRI scans ...... 112 FIGURE 2.7, Vessel input function from DCEMRI data...... 113 FIGURE 2.8, Representative dynamic FM scans ...... 113 FIGURE 2.9, Vessel input function from dynamic FM data ...... 114

FIGURE 2.10ab, Representative DCEMRI and FMderived v p maps...... 114

FIGURE 2.10c, Coregistration of DCEMRI and FMderived v p maps...... 114 FIGURE 2.11ab, Representative DCEMRI and FMderived K trans maps ...... 114 FIGURE 2.11c, Coregistration of DCEMRI and FMderived K trans maps...... 114 FIGURE 2.12, Representative K trans histograms for DCEMRI and FM data ...... 115 FIGURE 2.13, Percent threshold histograms of K trans for DCEMRI and FM data...... 116

FIGURE 2.14, Representative v p histograms for DCEMRI and FM data...... 116

FIGURE 2.15, Percent threshold histograms of v p for DCEMRI and FM data...... 117 FIGURE 2.16, Graphical representation of Table 2.1...... 119 FIGURE 3.1a, T1weighted series of biotinOvAGdDTPA phantom...... 133 FIGURE 3.1b, Recovery curves for biotinOvAGdDTPA phantom...... 133

FIGURE 3.1c, Fit for r1 relaxivity ...... 133

FIGURE 3.2a, T2weighted series of biotinOvAGdDTPA phantom...... 134 FIGURE 3.2b, Decay curves for biotinOvAGdDTPA phantom...... 134

FIGURE 3.2c, Fit for r2 relaxivity ...... 134 FIGURE 3.3, Transillumination image overlaid with GFP fluorescence ...... 135 FIGURE 3.4, DCEMRI average of biotinOvAGdDTPA & corresponding AIF ...... 135 FIGURE 3.5, Comparison SNR/CNR of biotinBSA/OvAGdDTPA for DCEMRI.... 136 FIGURE 3.6, Dyanmic FM average of biotinOvAGdDTPA & corresponding AIF.... 136 FIGURE 3.7, Comparison SNR/CNR of biotinBSA/OvAGdDTPA for FM...... 137

FIGURE 3.8ab, Representative DCEMRI and FMderived v p maps...... 137

FIGURE 3.8c, Coregistration of DCEMRI and FMderived v p maps ...... 137 FIGURE 3.9ab, Representative DCEMRI and FMderived K trans maps ...... 138 FIGURE 3.9c, Coregistration of DCEMRI and FMderived K trans maps...... 138 FIGURE 3.10, Representative K trans histograms for DCEMRI and FM data ...... 138 FIGURE 3.11, Percent threshold histograms of K trans for DCEMRI and FM data...... 139

FIGURE 3.12, Representative v p histograms for DCEMRI and FM data...... 139

FIGURE 3.13, Percent threshold histograms of v p for DCEMRI and FM data...... 140 FIGURE 3.14, Graphical representation of Table 3.2 biotinOvAGdDTPA ...... 141 FIGURE 3.15, Graphical representation of Table 3.2 biotinBSAGdDTPA ...... 142 FIGURE 4.1, Images pre & posttherapy using biotinBSAGdDTPA...... 153 FIGURE 4.2, Histograms pre & posttherapy using biotinBSAGdDTPA...... 155 FIGURE 4.3, Graphical representation of Table 4.1...... 157 FIGURE 4.4, Images pre & posttherapy using biotinOvAGdDTPA...... 158 FIGURE 4.5, Histograms pre & posttherapy using biotinOvAGdDTPA...... 160 FIGURE 4.6, Graphical representation of Table 4.2...... 162

FIGURE 4.7, Representative v p map pre & postTx using biotinOvaGdDTPA...... 162

LIST OF TABLES

TABLE 1.1, Comparison of Imaging Systems ...... 84 TABLE 1.2, Measurement of Hemodynamics with PET/SPECT ...... 84 TABLE 1.3, Measurement of Hemodynamics with CT ...... 86 TABLE 1.4, Measurement of Hemodynamics with Ultrasound...... 87 TABLE 1.5, Measurement of Hemodynamics with Optical Imaging ...... 89 TABLE 1.6, Measurement of Hemodynamics with DCEMRI...... 89 TABLE 1.7, Antivascular Therapies: Targets & Mechanisms of Action...... 90 TABLE 1.8, Antiangiogenic Therapies: Targets & Mechanisms of Action...... 90 TABLE 1.9, Imaging Response to AntiVascular & Antiangiogenic Therapies...... 92 trans TABLE 2.1, K and vp values in the tumor and nontumor ROIs ...... 118 TABLE 3.1, Characterization of biotinBSAGdDTPA vs. biotinOvAGdDTPA ...... 135 trans TABLE 3.2, Comparison of v p and K from biotinBSAGdDTPA vs. biotinOvA GdDTPA...... 141 trans TABLE 4.1, MR and FMderived K and v p biotinBSAGdDTPA ...... 156 trans TABLE 4.2, MR and FMderived K and v p biotinOvAGdDTPA ...... 161

ABSTRACT

Microvascular permeability is a pharmacologic indicator of tumor response to therapy, and it is expected that this biomarker will evolve into a clinical surrogate endpoint and be integrated into protocols for determining patient response to antiangiogenic or antivascular therapies. The goal of this research is to develop a method by which trans microvascular permeability ( K ) and vascular volume (vp) as measured by DCEMRI were directly compared to the same parameters measured by intravital fluorescence microscopy in an MRIcompatible window chamber model. Dynamic contrast enhanced MRI (DCEMRI) is a noninvasive, clinically useful imaging approach that has been used extensively to measure active changes in tumor microvascular hemodynamics. However, uncertainties exist in DCEMRI as it does not interrogate the contrast reagent (CR) itself, but the effect of the CR on tissue water relaxivity. Thus, direct comparison of DCEMRI with a more quantitative measure would help better define the derived parameters. The combined imaging system was able to obtain both dynamic contrastenhanced MRI data high spatiotermporal resolution fluorescence data following injection of fluorescent and gadolinium colabeled albumin. This approach allowed for the crossvalidation of vascular permeability data, in relation tumor growth, angiogenesis and response to therapy in both imaging systems.

CHAPTER 1

Introduction to Imaging of Microvascular Permeability

1 Section 1

1.1 Defining Vascular Permeability and Perfusion

Perfusion is the process by which the circulatory system delivers blood to tissues. It is

1 1 functionally equivalent to flow, expressed in ml 100g tissue min . By Poiseuille’s Law, flow is a function of the pressure gradient over a given length of vessel and the resistance of that vessel to the flow of viscous fluid (blood). By contrast, vascular permeability is expressed as a transfer coefficient of a solute per unit time, per volume of tissue between the vascular compartment (plasma) and the extracellular extravascular space (EES).

Permeability, expressed in units cm sec 1, is functionally defined as the flux (mmol s 1) per unit area of intervening membrane (cm 2) and per unit concentration difference across that membrane (mmol cm 3).

1.2 Macrocirculation – Overview

1.2.1. Morphology

A physiologically normal blood vessel is an unctuous conduit for blood flow. Both arteries and veins have three identifiable or distinct layers: tunica intima, media and adventitia (1). The tunica intima is the innermost layer or the layer most proximal to the lumen of the vessel and itself is composed of three layers: the endothelium,

14 subendothelial layer and the internal elastic membrane. The endothelium is a confluent monolayer of endothelial cells that are aligned in the direction of flow. The abluminal surface of the endothelial layer is embedded by a basement membrane (subendothelial layer) composed of collagen, fibronectin and laminin. In some vessels, the basement membrane may rest directly on the internal elastic membrane. This membrane interfaces with the central layer or the tunica media.

The tunica media constitutes the greatest volume for the vessel, comprised mainly of smooth muscle cells. In arteries, one may also find reticular collagenous fibers and an external elastic membrane, all of which regulate the vasoconstrictive and vasodilatory properties of the vessel important for maintaining blood pressure. In veins, besides smooth muscle cells, one may only find fibroelastic connective tissue.

The outermost layer of the vessel, the tunica adventitia is comprised mostly of collagenous tissue, nerves and fibroblasts (1). Arterial adventitia sometimes presents with a vascular network known as the vasa vasorum. In the veins, however, there is evidence of connective tissue, fibroblasts, elastic fibers and smooth muscle cells. Another distinct difference between veins and arteries is that veins may possess valves which prevent reflux of blood. The anatomical differences between veins and arteries are diagrammed in

Figure 1.

1.2.2. Vessel Regulation of Flow

A pressure gradient (expressed in mm Hg) down the vessel imparts flow (2). In its primary physical sense, flow ( F, ml sec 1), is the product of the velocity, ν, or displacement of a substance per unit time (cm sec 1) and the crosssectional area of the carrier ( A, cm 2): F = ν A . When discussing blood flow in the context of vascular physiology, units can be ambiguous because most generally, it is regional flow or flow standardized to a unit quantity that is being discussed. Thus, in the literature, a physical expression of flow is expressed as quantity over time (ml min 1), whereas in general, the vascular physiology expression of flow is expressed as quantity over time over quantity

1 1 (ml 100g tissue min ) (3).

The characteristic hierarchal branching of the cardiovascular system is designed such that the more extensive the branching, the greater the crosssectional area. As the total flow rate must remain constant throughout the system, the blood flow velocity, ν, must decrease as it moves through arteries and arterioles down to the capillary level (2).

Resistance is another important factor that affects blood flow. Hemodynamic resistance is a measure of the opposition to blood flow through a vessel (2). Throughout the cardiovascular system, pressure increases with increased resistance to maintain a constant flow rate. Resistance is dependent on: 1) blood viscosity, η; 2) vessel length, and; 3) vessel radius (1). Although the viscosity of blood is shearrate dependent, it remains relatively constant throughout the vascular system and therefore, does not contribute a great deal to the control of resistance. Similarly, the length of vessels in a system is also

16 constant. Thus, the main variable determinant of resistance is vessel radius. A small change in the radius leads to a significant change in flow as resistance is inversely proportional to the fourth power of the radius:

1 R ∝ Equation 1 r 4

The greatest control over vessel radii, and hence resistance, occurs at the level of the arteriole. This is because extensive downstream capillary branching leads to a steep post arteriole pressure gradient.

Poiseuilles’s Law describes the flow rate Φ (mL sec 1) of an incompressible fluid as:

π Pr 4 Φ = , Equation 2 8ηL where P/ L represents the pressure gradient (P 1 P 2) over a given length, L. Poiseuilles’s

Law is limited by a few assumptions: 1) the radius of the tube must be constant, and much smaller than the length of the tube, 2) the fluid viscosity must be uniform and constant, 3) flow is laminar, and, 4) the velocity at the tube edge is zero or negligible, assured by a nonzero viscocity. The second and third assumptions may not be met in the arterial circulation and hence, Poiseuilles’s Law is only a useful approximation of total blood flow (2).

1.3 Capillary Microcirculation

Imaging studies of vascular permeability primarily reflect microcirculatory (capillary) phenomena. It is estimated that the human body is equipped with 1040 billion

17 capillaries, which provide over 600 m 2 of surface area for exchange of materials.

Capillary walls are thin (~1 m) and the average diameter of a capillary is small (~3 m) compared to that of a red blood cell (~8 m). In normal tissues, the abundance of capillaries is such that no cell is farther than 100 m away from a capillary (2).

Extensive data have shown that this is clearly not the case in tumors (described below).

There are fewer cell types that constitute a capillary compared to other vessels, but the underlying biology of the cells is more complex. A single, flat layer of endothelial cells joined together by tight junctions constitutes the inner lining of capillaries, and these cells are as heterogeneous as the tissues they perfuse. Oh et al. developed a discovery and validation approach to understand the heterogeneity in lung endothelial cells (4, 5). This productive research led to discovery of about 50 differentially expressed proteins, two of which may permit immunotargeting and imaging in vivo . An outer basement membrane surrounds these cells, and in some capillary beds, perivascular cells or pericytes surround that basement membrane.

Capillary endothelia are categorized according to porosity. Continuous capillaries are the most common, and maintain a tight barrier between the blood and the interstitial fluid.

These are observed in tissues that require a significant amount of protection from blood constituents such as the central nervous system, muscle and lung. Fenestrated capillaries are more porous, with clefts between endothelial cells and fenestrations at the surface of the endothelial cells. These allow passage of watersoluble substances such as glucose and amino acids, but they are not permeable to plasma proteins, which are larger.

Fenestrae can be completely open, like a pore, or may be covered by a thin, non membranous diaphragm across the opening. Sinusoidal capillaries are specialized vessels found in organs (e.g. bone marrow, spleen or liver) that require the involution or excretion of whole cells, large molecules or assorted particles from the blood. The diameters of these capillaries are large (~50 m). The lumen is irregularly shaped and lined by specialized phagocytic cells. These vessels have large intercellular clefts and fenestrations that lack diaphragms. In some vessels the basal lamina may be partially or completely absent. Of all microcirculatory vessel types, these sinusoidal vessels most resemble capillaries found in tumors.

1.3.1. Passive Capillary Transport

Poiseuille’s Law is useful for describing the flow of whole blood through blood vessels.

Both plasma and solute molecules contained in blood also undergo passive transport across the capillary wall, by either diffusive or convective mechanisms (1). Active transport mechanisms are discussed later in this review. In solution or in tissues without a permeability barrier, Fick’s law of diffusion simply states:

dQ ∂C = −DS , Equation 3 dt ∂x where dQ is the diffusive flux (mmol) in time dt across a plane of area, S (cm 2), under an instantaneous concentration gradient ( ∂C , mmol cm 3, ∂x , cm 1), and D is the diffusion coefficient (cm 2 sec 1) at a given temperature in a given medium (6).

The expression changes in the context of two compartments ( C1 and C2) separated by a semipermeable membrane. Across the membrane, Q depends on permeability of the membrane to the materials being exchanged, and the ratio of the area ( S) to the thickness

(H) of the membrane becomes important:

dQ S = −D ()C − C . Equation 4 dt H 1 2

This model assumes that compartments C 1 and C 2 are homogenous and perfectly mixed, and diffusion occurs only in the membrane. In the treatment of a capillary and its surrounding stroma, or the interstitial space, these conditions are met, and modeling permeability of the capillary membrane requires that the influence of diffusion on the tracer becomes negligible and that the two compartments (vascular and interstitial compartments) are well mixed at all times. Through simple mass balance, uptake of tracer in tissue can be expressed as the difference between the quantity of tracer delivered to the tissue from the arterial phase ( Ca) and the quantity removed by the venous phase

(Cv):

Q = F(Ca − Cv ) . Equation 5

These concepts are developed more extensively in the treatment of Dynamic Contrast

Enhanced MRI estimation of tracer concentration later in this review.

1.4 Angiogenesis

Angiogenesis is an important component of tumor formation. The angioblast is commonly believed to be the progenitor of endothelial cells, yet a multipotential precursor, the hemangioblast, has been identified that can differentiate into either

20 hematopoietic or endothelial cells (7). Recent studies have confirmed an alternative pathway not involving erythropoiesis (810), though it remains unclear when the divergence of these two lineages occurs.

Vasculogenesis is the differentiation of the angioblast in the embryonic and extra embryonic mesoderm and subsequent formation of primordial capillary plexus from angioblasts at their site of origin (11). Vascular endothelial growth factor (VEGF) and its associated receptors, VEGFReceptor 1 and 2 (VEGFR1 and R2) and the angiopoietin/Tie receptor system are the known principal endotheliumspecific factors and associated receptors that regulate vasculogenesis during embryogenesis (12).

Following successful elaboration of the cardiovascular system, the expansion of the microvasculature in embryogenesis (and in adult tissue remodeling) occurs by angiogenesis. During normal, postnatal development, there is little need for angiogenesis except during reproductive cycling and wound healing (1316). Angiogenesis on an established vascular scaffold includes pruning of preexisting vessels and the formation of new capillaries, by both sprouting and nonsprouting angiogenesis (intussusception)

(1720).

Sprouting angiogenesis is a relatively slow process characterized by the extension of capillary sprouts invading the mesenchyme. Sprouting is initiated by the proteolytic degradation of the basement membrane on an established capillary, followed by the migration and proliferation of endothelial cells into the extracellular matrix and along the

21 chemotactic gradient to a proximal capillary. The sprouts internally reorganize to create a vascular lumen and form a stable connection with an adjacent capillary. Though the entire extension process requires little more than 24 hours, perfusion of the new capillary does not occur for 35 days following its inauguration into the vascular system (20).

1.4.1. Molecular Mechanisms of Angiogenesis

Mediators of angiogenesis have excitatory or inhibitory effects on vessel initiation and growth. These mediators vary spatially and temporally throughout the angiogenic process, and this differential regulation generally manifests as an inhibition of angiogenesis at any given point in time (21, 22). If new vessel growth or remodeling is required, this inhibition is attenuated and activators of angiogenesis are expressed and mobilized. The entire process of angiogenesis can be organized into two general categories: initiation and resolution (23, 24). The initiation process constitutes a period of growth and expansion following a stimulus to generate, expand or remodel the local vasculature. The resolution phase involves the stabilization of nascent blood vessels by reinstatement of balanced levels of angiogenic growth factors that favor angiogenic quiescence. Both processes are tightly regulated and inextricably linked to one another.

The initiation phase of angiogenesis is characterized by endothelial cell activation of substances that promote vasodilation, such as nitric oxide (NO), and permeability, such as VEGF (23, 24). The ensuing processes involve the disassembly of the vessel wall and subsequent degradation of the basement membrane and the extracellular matrix (ECM).

These processes are predominantly mediated by angiopoietin (Ang)2 and matrix metalloproteinases (MMPs), which are found in the endothelial cells of existing blood vessels. Ang2 induces the detachment of smooth muscle cells and its associated basement membrane, thereby relieving interendothelial cell contacts. Interestingly, Ang

1 has profoundly the opposite effect, contributing to the stabilization and maturation of vessels (25). MMPs, which are secreted by endothelial cells, stromal cells or in the case of tumorigenic angiogenesis, tumor cells themselves, degrade the capillary basement membrane and if necessary, induce the detachment of the pericytes that surround the capillary (22). They are thought to not only provide space for migration, but also to release growth factors like VEGF and basic fibroblast growth factor (bFGF), which are normally sequestered in the matrix. Other proteases involved in the ECM degradation process include the serine proteases, tissue plasminogen activator (tPA) and urokinase type plasminogen activator (uPA), which convert plasminogen to plasmin (26). Plasmin can degrade several ECM components such as fibronectin, laminin and proteoglycans, and it may possibly activate MMPs in vivo .

In order for endothelial cells to emigrate, the interendothelial cell contacts that stabilize the vessel must be modulated. Cellular adhesion molecules are important for numerous processes including cell growth, differentiation, immunecell mediated inflammatory responses, metastases and angiogenesis. The four classes of endothelial cell adhesion molecules are integrins, cadherins, selectins and members of the immunoglobin

23 superfamily. All have been implicated in the angiogenic process, either through observations of coincident events or through direct association (27).

Once a provisional scaffold has been established on which endothelial cells can migrate, small groups of pioneering endothelial cells negotiate the new pathway to the interstitium

(21). They are closely followed by proliferating endothelial cells that are stimulated to divide by number angiogenic inducers, some of which are released from the degraded

ECM. These proliferating endothelial cells begin to organize into hollow tubes that eventually mature into a functional blood vessel (21, 23, 24, 28).

1.4.2. Microcirculatory Assays

Chronic Transparent Chamber Assay

The rabbit ear chamber was initially developed to study wound healing (29). The transparent chamber assay has since been used in the hamster cheek pouch for diffusion studies (30), and the dorsal skinfold window chamber was later introduced into mice to study tumor angiogenesis (31, 32). The advantage of the rabbit ear over the hamster cheek pouch and the dorsal skinfold is that the rabbit ear is thinner and more conducive to optical transillumination imaging. However, the rabbit ear and hamster cheek sites are not immunodeficient, and therefore, interrogation of tumor angiogenesis is possible only with the dorsal skinfold preparation. Cranial windows are a form of chronic transparent chamber, but they require epiillumination and administration of contrast agents (e.g. chromophores) for inspection of the circulation. Despite this disadvantage (and the fact

24 that it is not an immunodeficient preparation), it lasts longer in animals (33) and it permits investigation of the brain microvasculature (34). There are limitations to the chronic transparent chamber assay, however. In comparison to other preparations, it is technically challenging and timeconsuming, and any observed angiogenic effects may not initially be deconvolved from inflammation and the process of wound healing (35).

Tissue Flaps & CAM Assays

For the hamster cheek pouch, rat cremaster muscle (36) and rabbit mesentery (37) preparations, the skin above the tissue of interest is separated from the underlying fascia followed by a transplantation of the cells or pellet containing the angiogenic agent into the area. Following the necessary amount of time for that particular experiment, the skin flap is “everted” again, the degree of angiogenesis is evaluated, and the skin flap is replaced and sutured for protection until the next observation. The disadvantage of this approach is that the eversion is traumatic, potentially confounding observations of angiogenesis (38).

The chick chorioallantoic membrane (CAM) assay is perhaps the most commonly used angiogenic assay because it is relatively inexpensive (39, 40). In the explant preparation, the entire egg contents are placed into a culture dish after a 72 hour incubation period

(other preparations keep the shell and shell membrane intact except for an opening through which agents are applied). Stimulation of the system with angiogenic agents

(contained in grafts, sponges, filters, or sustainedrelease membranes) over a span of

25 several days can be visualized throughout the development of the embryo. Of course, a major limitation is that the cells must originate from the chicken (Gallus gallus).

In Situ Assays

The advantage of the in situ assay is that it can be used in rabbits, rats and mice. This system consists of a pocket in the cornea distal to the limbus. When angiogenesis activators are placed in the pocket, peripheral limbal vascularization is stimulated (41).

The limitations of this assay are that the cornea is otherwise avascular, and physiological angiogenesis does not typically occur in avascular regions. Furthermore, in longterm observation studies, sustained release formulations can induce inflammatory reactions

(42).

1.4.3. Matrix Implant Assays

Matrigel Assay

Matrigel® (BD Biosciences) is a solubule basement membrane preparation extracted from EHS mouse sarcoma cells. A Matrigel plug is created when Matrigel is injected into subcutaneous tissue where it solidifies (43). Specific growth factors can be suspended in the liquid Matrigel which, when implanted, induce angiogenesis (44). Disadvantages to this system include its low reproducibility and the undefined composition of Matrigel.

Injection of identical amounts of the liquid Matrigel rarely forms the 3dimensional scaffold in exactly the same way.

1.4.4. Ex Vivo Assays

Ex vivo assays involve light or electron microscopy or histological examination of excised tissue. Light or electron microscopy involves casting, and histological examination can be done by staining for endothelial cells or perfusing the network intravascular contrast agents (e.g. colloidal carbon, radioactively labeled red blood cells

(RBCs) and biotinylated highmolecular weight contrast agents) (4549). The main disadvantages of this technique are that it is invasive, and it cannot be done in the same tissue serially over time, for example, before and after stimulation with angiogenesis inducers or inhibitors.

However, an important and compelling argument for use of ex vivo assays is that it is currently the only method by which to analyze angiogenesis in humans. A clinical method for noninvasively observing, evaluating or assessing angiogenesis in humans will most likely come in the form of imaging techniques such as Positron Emission

Tomography (PET), SinglePhoton Emission Computed Tomography (SPECT),

Computed Tomography (CT), Doppler ultrasound, Dynamic ContrastEnhanced

Magnetic Resonance Imaging (DCEMRI) and optical methods, discussed below.

1.5 Tumor angiogenesis

Pathological conditions that involve uncontrolled angiogenesis include rheumatoid arthritis (50), diabetic retinopathy (51), psoriasis (52), chronic inflammation (53), and tumor growth and metastases (28, 54, 55). Tumorigenic angiogenesis is a requirement

27 for tumor growth beyond the clinically detectable size of 12 mm 3 (56), making it an acquired capability that is shared by most types of the invasive human tumors (16, 57). It is critical to the growth and metastases of malignant tumors, and with increasing clinical observations and subsequent investigation of tumor blood vessels, more and more evidence points to correlations between highly vascularized solid tumors (i.e. a high intratumoral microvessel density) and poor prognosis (i.e. increased incidence of metastases and reduced patient survival) (58, 59). Thus, robust measures of tumor angiogenesis will serve as potential endpoints for targeted therapies.

Tumor blood vessel permeability is an area of intense research because physiological angiogenesis and tumorigenic angiogenesis are significantly different, particularly in the context of mechanical features and the regulation of chemical mediators. Physiological angiogenesis is meticulously regulated, and new vessel growth is carefully initiated or inhibited depending on the metabolic needs of the tissue. Tumor angiogenesis is an aberrantly regulated process and is stimulated by competition for limited resources amongst actively proliferating tumor cells (28, 60). Tumor vessels themselves exhibit vast irregularities when compared to normal vessel characteristics (57). Early reaction diffusion models characterized oxygen diffusion distances to be less than 150 m from a vessel (i.e. the Krogh cylinder) (61). As the tumor grows beyond these lifesustaining limits, the tumor is denuded of important nutrients, such as oxygen and glucose.

Carcinomas in situ , which are encapsulated by a basement membrane, are physically separated from vessels (60). These avascular tumors may stay dormant and undetected

28 for years. Once they escape basement membrane encapsulation, they may induce the formation of a customized vasculature, providing the necessary capability for tissue invasion and metastases. This evolution from avascularity to incipient neovascularization depends on the counterbalance between the inhibitory and excitatory mediators of angiogenesis in the tumor’s armamentarium. Imbalances of angiogenic mediators like

VEGF and the angiopoietins may contribute to the excessive branching and shunting, which paradoxically results in even more hypoxia and acidosis. The excessive leakiness or hyperpermeability of tumor vessels is in part attributed to cytokines and angiogenic factors, which presumably alter the dynamic microvessel wall (57).

1.6 Using Imaging to Estimate Microvascular Permeability

This review is focused on the measurement of microvessel hemodynamics. Of particular interest is microvascular permeability, particularly in the context of the abnormal tumor microvascular network. Flux of a tracer across the endothelium is a complex system of exchange between the vascular and extravascular space dependent on oncotic and hydraulic pressure, active and passive transport and vessel surface area. The microvessel membrane is a fluid, dynamic machine, whereas the watery membrane described in the original Kety diffusion models assumed a plane of homogenous porosity, a significant limitation to obtaining a true estimate of permeability. There is no consensus among the imaging community regarding what measured permeability veritably represents. In truth, imaging offers only an estimation of permeability, or an ‘effective’ permeability based on tracer enhancement kinetics (discussed in more detail in later sections), and thus, on this

29 macroscopic level, the rate of enhancement is related to microvascular permeability through complex kinetic models. However, it is prudent to consider that there are many other factors affecting this process that are involuted into the assumptions of these tracer kinetic models.

Although a measure of effective permeability can be theoretically derived from quantitative imaging data, there has been limited progress in the development of in vivo imaging systems that can truly quantify this parameter. The nuclear medicine modalities such as PET and SPECT, as well as CT have dynamic tracer imaging capabilities for measuring tumor microvessel permeability, but exposure to continuous radiation limits their use in longitudinal studies and acquisition of images with high spatiotemporal resolution. Optical measurements are also robust in terms of quantitation but a major limitation is a lack of penetration depth, which is a significant impediment to clinical translation. Ultrasound is noninvasive and nonionizing and is clinically ubiquitous, but it only measures perfusion (not permeability) moderately well and suffers from low signal to noise ratio, poor dynamic range and low spatial resolution at clinically used frequencies. DCEMRI is also nonionizing, but poor reproducibility in both image acquisition and analysis across image sites and institutions reduces confidence levels.

This review will present the basic methodology, strengths and weaknesses for each of these systems with respect to their ability to produce reliable and robust measures of effective tumor vascular permeability.

2 Section 2

2.1 Imaging the Tumor Microvasculature: Survey of Methods

Useful imaging systems have been developed to monitor angiogenesis and the microvasculature in vivo , including PET and SPECT, CT, Doppler ultrasound, optical imaging methods and DCEMRI. Table 1.1 contains a listing of the overall coverage, resolution and sensitivity associated with each imaging system. Each system has strengths and limitations in terms of their ability to measure microvascular hemodynamics, all of which will be reviewed in their respective sections.

2.2 Measurement of Hemodynamics with Nuclear Medicine (PET/SPECT)

Positron emission tomography is based on the approximate collinearity of two photons simultaneously emitted by the annihilation of a positron and electron. The resolution of

PET is limited by uncertainties arising from the approximation of collinearity and annihilation position due to individual positron ranges (62).

Generally, PET measures of tumor perfusion have used ( 15 O)labeled radiotracers. An important study compared steadystate and dynamic methods for determining blood flow in 22 patients with liver metastases (63). The steadystate method required inhalation of

15 15 [ C]O2 and the dynamic method required an intravenous bolus injection of [ O]H2O.

Although both tracers allowed for quantification of a blood flow parameter, the steady

15 state method using [ C]O2 subject to errors arising from a builtin assumption of the partition coefficient for water, and is thus inferior to its ability to quantify blood flow

15 compared to the dynamic method using [ O]H2O. The dynamic method also delivered close correlations to angiographic data, demonstrating that the degree of vascularization, as determined by PET, can reliably differentiate between tumor grades.

A requirement for quantification of perfusion using dynamic enhancement imaging methods is an accurate determination of an arterial input function, which can be obtained noninvasively in a purely arterial region of interest, such as the aorta (64, 65). This also allows for determination of the volume of distribution parameter ( Vd), which is the proportion of exchanging water space between tissue compartment and vascular compartments. With these corrections, the coefficients of variation can be minimized.

15 15 Using Olabeled carbon dioxide ([ O]CO 2), intrapatient coefficients of variation for measurement of flow and Vd were 11% and 6%, respectively (66).

2.2.1. Response to Antivascular Therapy

One of the first randomized clinical trials to use 15 Olabeled tracers measured vascular hemodynamics in a Phase I doseescalation study of combrestatin A4 phosphate (CA4P), an antivascular therapy (67). In 13 patients with a variety of solid tumors, significant

15 reductions in perfusion were observed as measured by [ O]H2O 30 minutes and 24 hours posttreatment. Significant reductions in blood volume could be measured by 15 O labeled carbon monoxide ([ 15 O]CO), which binds to hemoglobin and is thus intravascular. These also show similar reductions within 30 minutes of CA4P, yet these were followed by a vascular recovery 24 hours later. In a Phase II clinical trial, 12

15 patients with solid primary or metastatic renal tumors were evaluated with [ O]H2O, and [ 15 O]CO in response to the antiangiogenic therapy, razoxane (68). Perfusion and blood volume data were obtained from only six of the total enrolled patients. Although the mean posttreatment perfusion was lower than pretreatment, these were not statistically significant. For this drug, there was virtually no change in blood volume pre versus posttreatment. There was also no correlation found between pretreatment perfusion and tumor size (as estimated by CT), nor was there a correlation found between perfusion and patient response or survival. Another study investigated the effects of

SU5416, a small molecule inhibitor of the tyrosine kinase domain of the VEGF receptor,

Flk1, in renal cell carcinoma patients (69). Following treatment, no patients had a clinical response, but 15 of 23 patients demonstrated stable disease. In patients with

15 available data, perfusion (as determined by [ O]H2O) and metabolism (as determined by

18 FDG) decreased or minimally changed in patients with stable disease and increased in a patient with progressive disease.

2.2.2. Response to Cytotoxic Therapy

PETmeasured perfusion and blood volume changes following treatment with cytotoxic therapy were quantified in 19 patients bearing stage II or III breast tumors (70). This

15 study found no correlation between response to therapy and perfusion using [ O]H2O, but found a positive correlation in total blood volume as measured by 11 Clabeled carbon monoxide ([ 11 C]CO). Another study involved investigation of improved delivery of a cytotoxic therapy, 5fluorouracil (5FU), using a vasopressor (nicotinamide) concomitant

33 with carbogen inhalation. The hypothesis was that such a treatment would enhance the delivery of 5fluorouracil by increasing tumor blood flow (71). Patients with colorectal

15 metastases were intravenously administered [ O]H2O, imaged for 10 minutes, then given [ 18 F]5FU intravenously and imaged for another 90 minutes. Nicotinamide and carbogen were administered only on the second imaging session, concurrent with the second cycle of chemotherapy. Both tumor perfusion and delivery of 5FU to the tumor were observed to increase significantly in response to nicotinamide/carbogen.

Finally, an interesting study was designed to examine the relationship between hypoxia and perfusion in 11 patients with various brain tumors (72). 18 Ffluoromisonidazole ( 18 F

FMISO) is a nitroimidazole derivative that is trapped by hypoxic cells and this was

15 injected following measurement of [ O]H2O distributions to evaluate perfusion. There was a positive correlation between tumor perfusion and 18 FFMISO tumor uptake within the first 5 minutes of the 18 FFMISO injection, probably reflecting delivery. At later time points, 18 FFMISO uptake occurred in areas that were not perfused (as measured by

15 [ O]H2O), suggesting that hypoxia developed in areas with poor perfusion.

Single photon emission computed tomography (SPECT) is routinely used to measure perfusion in the clinic. Although it has lower sensitivity compared to PET, the spatial resolution is theoretically higher, as it is not dependent on positron annihilation (62).

Dynamic SPECT is an improvement to conventional SPECT, offering higher resolution, highsensitivity dynamic and static collimators and filtered backprojection reconstruction.

Dynamic SPECT can be used to measure cerebral blood vessel dynamics (73). SPECT of

Nisopropylp[123 I]iodoamphetamine (IMP) has been used to positively differentiate between blood flow in brain tumors and normal gray matter, indicating a disruption in tumorassociated vessel permeability (74). The intravascular agents, Technetium99m labeled albumin ( 99m TcHSA), has been used to measure permeability in patients with hepatic tumors, and perfusion was shown to increase compared to normal tissue following epinephrine infusion in 67% of the patients (75). Another study demonstrated that hepatic arterial perfusion scintigraphy (HAPS) using 99m TcHSA was useful in assessment of catheter placement and perfusion, changes in perfusion during chemotherapy and arteriovenous shunting to the lung (76). Thallium201 ( 201 TlCl) chloride and 99m Tc diethylenetriaminepentaacetic acid labeled HSA ( 99m TcHSAD) were used to assess tumor viability and tumor vascular permeability, respectively, in 17 patients. Within one month postradiotherapy, both systems delivered values that were significantly reduced compared to pretherapy values (77). In another study, patients with various soft tissue and bone tumors were imaged using 99m Tc labeled hexakis2methoxyisobutylisonitrile ( 99m Tc

MIBI) (78). The perfusion index was derived by dividing the peak count of the arterial phase by that of the contralateral tissue, and the MIBIuptake ratio was derived by dividing the count density of the tumor by that of the contralateral tissue. Both the perfusion index and the MIBIuptake ratio of benign lesions were significantly lower compared to malignant lesions. Setting a threshold cutoff for the perfusion index of 10.0, sensitivity, specificity, accuracy, PPV and NPV values of 40%, 100%, 71%, 100% and

64% respectively, were possible. This cutoff was not dependent on tumor size. Though

35 many studies have used 99m TcMIBI to evaluate tumor perfusion, they have primarily used planar scintigraphy to evaluate the perfusion characteristics. For example, patients with various cancers have been evaluated with 99m Tchexamethylpropylenamineoxime

(HMPAO), and a hypoxiarelated tracer, iodoazomycin arabinoside (IAZA) labeled with

123 I (79). In general, an inverse correlation between perfusion and uptake of IAZA was observed in all tumor types except for glioblastoma multiforme. Table 1.2 summarizes the list of studies using PET and SPECT to measure hemodynamic parameters according to the system and parameters measured along side an abbreviated version of the critical findings from each study.

2.2.3. Kinetic Modeling of PET/SPECT Data

Radioisotope measurements of tracer distribution into tissues are wellcharacterized. The relationship between blood flow and tracer clearance was originally described by Fick in

1870. In 1978, Ohno et al. modified the Fick equations of tracer diffusion to describe the distribution of 14 Clabeledlow molecular weight compounds (80). In this formalism, the plasma concentration can be represented as a sum of decaying exponentials:

n −mit C p (t) = ∑ A1e . Equation 6 i=1

When blood flow ( F) is much more than the permeability surface area product ( PS ), transport across the capillary is independent of flow, and the tissue tracer uptake per unit gram of tissue can be expressed as:

dC  C  t  t  Equation 7 = PSC p − , dt  ve 

36 where Ct is the tissue concentration, ve is the extracellular extravascular space per unit volume of tissue and Ce=C t/v e. Ohno et al. did not include the tissue density ( ρ) in their equations, stating that brain specific gravity (a dimensionless quantity) was approximately

1.0. Accounting for tissue density and solving this differential equation leads to:

n A1 −(PSρ /ve )t −mit Ct (t) = PSρ∑ e − e , Equation 8 i=1 mi − (PSρ / ve ) and early in the time course, when flow of tracer is unidirectionally moving out of blood into the tissue,

C t Equation 9 PS = t . ρ C t )'( dt' ∫ p 0

Essentially, tracer cleared by the tissue over time is a function of the perfusion of the tracer and the extraction of the tracer.

2.3 Measurement of Hemodynamics with Computed Tomography

In Xray computed tomography (CT), tissue contrast is based on variable attenuation coefficients (Hounsfield units) of the object absorbing the Xrays (62). Hemodynamic parameters may be extracted from dynamic changes in Xray attenuation caused by the intravenous injection of a contrast agent. Perfusion CT data can deliver quantitative hemodynamic information, such as blood volume, blood flow, permeability surfacearea product and mean transit time (MTT). An example includes diagnostic differentiation of rectal cancers from normal tissue based on elevated blood flow and attenuated MTT in the tumor and measurable posttreatment changes in blood flow and MTT (81). Notably, it

37 was even possible with this small data set to make general conclusions about the predictive value of this imaging system for patient response, as it appeared that a higher pretherapy perfusion index and shorter MTT was associated with a poorer therapeutic outcome.

In a larger study, 77 patients with primary oropharyngeal and oral carcinoma, questionable recurrent disease and suspected metastatic lymph nodes were evaluated with perfusion CT

(82). In this study, a deconvolution model with a 2 compartment analysis was applied for capillary permeabilitysurface area product (PS) measurements. Other parametric maps were obtained for blood flow, blood volume and MTT. All parameters except for MTT were significantly higher in tumors compared to normal tissue. MTT was significantly lower than that of normal tissue. Furthermore, the perfusion measurements sufficiently differentiated between primary and recurrent tumors. In head and neck squamous cell carcinomas, perfusion CT was able to reliably and significantly differentiate between tumor and normal tissue (83). Parametric maps of blood flow, blood volume and permeability demonstrated significantly larger values compared to normal tissue and the mean transit time was significantly attenuated compared to normal tissue. In hepatic metastases, median values of arterial perfusion were greater in the metastatic tissue compared to adjacent normal tissue (84). An advantage of CT is the excellent spatial resolution, allowing observation of microenvironmental heterogeneities. In just under half of the patients evaluated, peripheral perfusion of the tissue surrounding the tumor was higher than the ROI containing only tumor, suggesting that neovascularization is occurring well beyond the margin of the tumor. In some patients, low perfusion was observed in the

38 center of the tumor. Notably, there were statistically significant correlations established between patient survival and perfusion of the metastatic lesions and adjacent liver tissue; i.e. a higher perfusion index was correlated to longer survival times (p < 0.05, r = 0.69). In a separate study conducted on 21 patients with lymphoma, median perfusion values were higher in patients with active, high or intermediate lymphoma, and perfusion values below

0.2 ml/min/ml indicated inactive disease (85). In this study, values of permeability did not produce information allowing for the differentiation between grades or lymphoma activity status.

A small study was conducted to assess the reproducibility of CT measured hemodynamics within a Phase IB clinical trial of combrestatin A4 phosphate in 10 patients with inoperable nonsmall lung cell cancer (NSCLC) (86). Parametric maps of permeability and blood volume were derived based on a Patlak analysis (87), and both parameters demonstrated good agreement between studies, with a CV of 9.49% and 26.31% for permeability and blood volume, respectively. In this study, an increase or decrease in tumor permeability over 8.3% would have been significant. Goh et. al. assessed intra and interobserver reproducibility for blood volume, blood flow, mean transit time and permeability from dynamic CT in patients with colorectal carcinoma (88). The correlation coefficients for the interobserver variability ranged from 0.8 to 0.89 for all perfusion parameters, and the intraobserver correlation coefficients ranged from 0.86 to 0.98, indicating excellent agreement in both cases. The same group also found that when imaging abdominal cancers, acquisition times under 45 minutes would result in significant

39 degradation of permeability measurements but not blood flow, blood volume or mean transit time (89). Dynamic CT has also been used to assess patient response to bevacizumab (90) and for a Phase I dose finding study of MEDI522, a humanized anti body to the αvβ3 integrin (91). In this latter study, of all the measured parameters (blood flow, blood volume, permeability surface area product and MTT) only the MTT demonstrated a significant change with increasing dose of MEDI522. Table 1.3 summarizes the list of studies using CT to measure hemodynamic parameters according to the system and parameters measured along side the critical findings for each study.

2.3.1. Kinetic Modeling of Perfusion CT Data

A widely used model of perfusion CT kinetic parameters which also accounts for diffusive resistances and concentration gradients (i.e. a distributed parameter model) was described by Johnson et al. (92) and later modified for a solution with an adiabatic approximation by

St. Lawrence et al. (93). According to the model, a contrast agent enters the intravascular space and is described as having a concentration, Cb(x,t), which decreases from the arterial inflow of the capillary, C a(t), to the venous outflow, C v(t). C e(t) represents the change in the contrast agent concentration in the extravascular, extracellular space over time, V e represents the interstitial volume, V b represents the blood volume, F represents the flow rate and PS represents the permeabilitysurface area product. The convective and diffusive transport of the contrast agent is represented by (81):

∂Cb (x,t) FL ∂Cb (x,t) PS + ⋅ + [Cb (x,t) − Ce (t)] = 0 , Equation 10 ∂t Vb ∂x Vb

40 and the evolution of the contrast agent concentration in the interstitial compartment is described by:

dC (t) PS L V e = [C (x,t) − C (t)]dx . Equation 11 e ∫ b e dt L 0

Using these equations to model enhancement kinetics, four vascular parameters, vascular permeability, vascular blood volume, MTT and blood flow, can be estimated simultaneously (81).

2.4 Measurement of Hemodynamics with Ultrasound

Medical ultrasonography is based on the pulseecho and backscattered echo waveforms.

A transducer which converts electric power to acoustic power is used to transmit short and relatively broadband pulses through the tissue, which are attenuated at tissue interfaces due to absorption and scattering. The pressure from the backscattered signal is collected by the same or a second phasesensitive transducer, and the output voltage is a radiofrequency trace which is recorded as a function of the acoustical travel time in the tissue. Because the trace is an amplitudemodulated display of the backscattered waveform, it is referred to as an Amode display or Ascan. The ultrasound image is created by converting the Amode display into a brightness mode display along a vertical axis. The horizontal profile of waveform spikes is now converted into a vertical series of bright dots. Moving the transducer along the specimen will create a 2 dimensional image of bright dots in the same manner, hence the term Bmode display. The fundamental limit to resolution in ultrasonography is the frequencydependent attenuation. Higher

41 transducer frequencies offer better resolution but greater tissue attenuation. Thus, any improvement in lateral resolution implies a loss in depth of penetration.

There are several different ultrasonic approaches designed specifically to measure blood flow including transit time, continuouswave Doppler, pulsed and color Doppler and power Doppler flowmeters. Transittime flowmeters operate under the phaseshift principle and utilize two transducers to measure blood flow velocity. However, the transittime flowmeter is infrequently used for routine estimation of flow because invasive surgery is required to expose the vessel. Unlike transittime flowmeter, the continuouswave (CW) Doppler flowmeter requires particulate matter, such as blood cells, for scattering. In CW Doppler ultrasound, red blood cells are considered reflecting targets, and when those targets recede from a stationary sound source (transducer), the frequency of the received sound is shifted proportionally to their velocity relative to the transmitter and receiver (94). The Doppler frequency shift is expressed as:

2 f u cosθ f = 0 , Equation 12 d c where f0 is the source frequency, c is the velocity of sound, u is the target velocity and θ is the angle between acoustic transmission and the long axis of the blood vessel. Pulsed

Doppler ultrasound is the most broadly applied ultrasound technique. In this, the delay between a brief transmitted wave and its reception is directly proportional to distance, so a complete plot of reflections across the blood vessel can plotted, generating a velocity profile based on Doppler shifts. The tradeoff for high resolution (short pulses) is a better signaltonoise ratio (long pulses). Furthermore, since the pulse system is broadband, it is

42 inherently more sensitive to electronic noise, resulting in low dynamic range and rendering it less capable of detecting small volumes of blood. Color Doppler is termed as such because pulse Dopplerderived velocity information is color mapped onto grayscale ultrasound images. The most common estimation of the velocities uses consecutive transmit pulses at fixed depths to find the phase difference between received echoes.

Color Doppler has been successfully used to evaluate abnormal perfusion profiles in liver

(95), ovarian (96), breast (48, 49) and prostate carcinomas (47). In a newer modification, called Power Doppler, color is encoded based on the integrated power of the Doppler signal rather than the mean Doppler frequency shift (97). The advantages of power

Doppler over color Doppler include a lower noise variance and ability to spectrally encode noise in order to differentiate it from information signal. Other advantages include angle independence and immunity to aliasing because the power remains the same as a function of phase shift, and depthcompensation (94).

Contrastenhanced ultrasound has become increasingly popular for measuring perfusion.

The first ultrasound contrast agent was reported in 1968 (98). Since then, the explosion of microbubble research has made ultrasonography a useful contrastenhanced clinical imaging utility. Microbubbles are comprised of an outer shell, generally composed of lipids, proteins, or polymers, and a gas inner core filled with air, perfluorocarbon, sulfur hexafluoride or nitrogen (99). Microbubbles expand and contract due to pressure from the acoustical transmit pulse, and the primary mode of echogenicity is the impedance mismatch between the microbubbleblood interface, making them significantly more

43 echogenic than normal tissue. However, there are many other manifestations of microbubble contrast including increased scattering crosssection, nonlinearities from increased acoustic pressure that give rise to unique harmonics, and transient, nonlinear signal emission from bubble disruption (a.k.a. cavitation).

Although it is not routine, contrastenhanced ultrasound is used to measure hemodynamics in human tumors. In recurrent breast cancer patients, contrast agent uptake after RF ablation therapy has been used to successfully predict responders, indicating a strong ability to estimate tumor vascularity (100). In solid breast tumors, contrast enhanced ultrasound has also been used to provide a measure of solid breast tumor vascularity as compared to histological analysis of vascularity (101). Compared to conventional Bmode scanning and CT, contrastenhanced ultrasound sensitively differentiated patients with metastatic liver lesions (~95%, compared to 50% and 59% for Bmode scanning and CT, respectively) (102). Contrastenhanced ultrasound is capable of differentiating between nonfunctional islet tumors (often misdiagnosed as ductal adenocarcinoma) from pancreatic carcinoma. The differential diagnosis criteria included age, size, margin, level of vascularity and enhancement timing (103). A notable quantitative study conducted on untreatable hepatocellular carcinomas showed that posttreatment with thalidomide, the percent blood volume, average red blood cell velocity and the product of the two, perfusion, all decreased significantly following therapy, even in patients followed out to 2 years (104). Table 1.4 summarizes the list of studies using ultrasound to measure

44 hemodynamic parameters according to the system and parameters measured along side the critical findings for each study.

2.4.1. Modeling of Perfusion Ultrasound Data

Hertz was the first to provide the idea of estimating perfusion using continuous wave

Doppler ultrasound (105), and this was closely followed by Dymling (106) with a mathematical estimation of perfusion based on the measured Doppler power spectrum.

Perfusion was expressed as follows:

Qc P = ≈ nc E{vc }, Equation 13 Vs where Qc is total capillary flow, Vs is the sample volume of tissue, nc is the number of

capillaries per unit volume tissue and E{vc } is the average blood velocity in the capillaries. To establish the proportionality between perfusion and a Dopplerderived measurement, the assumption was made that red blood cells were the primary, isotropic scattering source:

∞ fS(){}f df ≈N 0 E vc . Equation 14 ∫0

Such an expression relates the mean number of red blood cells in some volume of interest and their mean speed along the zaxis to the first moment of the Doppler power spectrum.

This expression can be used to estimate velocity using any of the above ultrasound techniques, but the modern flow mapping techniques estimate perfusion by quantifying the density of colored pixels or number of visible vessels in a given region of interest relative to normal tissue.

An interesting method to estimate perfusion was demonstrated in myocardium using contrastenhanced ultrasound and selective highpower pulses to destroy microbubbles and observe the inflow of unaffected microbubbles during a continuous infusion. The so called ‘refilling’ signal reflected the mean myocardial microbubble velocity. The asymptotic maximum was related to fractional vascular volume, and the product of the two parameters was related to perfusion (104, 107). This approach has yet to be used in cancers, although focused ultrasonic cavitation of microbubbles has been identified as a promising technique for imaging and delivery of therapies (108111).

2.5 Measurement of Hemodynamics with Optical Imaging

“Optical imaging” describes a broad range of techniques. Methods specifically designed for in vivo imaging include photoluminescence ( fluorescence) and chemiluminescence

(commonly referred to as bioluminescence ). An additional approach that may have clinical utility is diffuse optical (DOT) tomography. This method utilizes the diffusion equation to model propagation of light through tissue using the absorption and scattering coefficients.

A major limitation of optical imaging is the depth of penetration because scattering and absorption increase as a function of tissue depth. Thus, capturing images from outside the body is difficult. However, sophisticated optical images can be obtained by detectors attached to endoscopes. The major advantages of optical imaging are its temporal, spatial and spectral resolution. Using reflected light and fluorescence light at a micrometer

46 resolution (for example, confocal microscopy), cancerous cell morphology can be interrogated, and multispectral imaging with millimeter resolution can be used to interrogate general tissue morphology (112). Optical imaging is capable of capturing real time dynamics particularly with recent improvements in the quality of fast and highly sensitive cameras.

The advancement of fluorescence microscopy in biomedical research is partly attributed to the development of methods to conjugate proteins to fluorochromes (113, 114), and partly to improvements in thin film technology and optoelectronics, which increased the sensitivity and specificity of detection of emitted fluorescent light (115). Most filters sets are not totally efficient in eliminating damaging and unwanted UV wavelengths, so fluorescence imaging can be damaging to tissues. The primary mechanisms of damage are phototoxicity (i.e. absorption of photons by exogenous and endogenous fluorochromes) and creation of reactive oxygen species (ROS). This can be mitigated somewhat by twophoton or nearIR imaging which uses longer wavelengths for excitation. A further advantage of NIR imaging is that the range of wavelengths (700

900 nm) maximizes depth of penetration while minimizing scatter and background autofluorescence (116). Chemiluminescence imaging is based on the catalysis of light emission from a substrate. The most common and fully developed reporter is the luciferase enzyme which oxidizes luciferin. Light emitted from this reaction is between

530640 nm (broadband) (117).

In general, optical imaging of vascular permeability has been limited to preclinical studies. However, there are currently more than 10 ongoing clinical trials that are using optical imaging to study human tumors (118). Fluorescence/reflectance imaging and confocal microscopy are used to detect surgical resections of nonmelanoma skin cancers; noninvasive, scanning twophoton fluorescence microscopy is being used to replace the skin biopsy; optical spectroscopy is being investigated as a potential use for intraoperative margin assessment during breast cancer lumpectomy; and transillumination breast spectroscopy is being developed as breast cancer risk predictor to replace and or augment mammographic parenchymal density.

Diffuse optical tomography (DOT) is potentially useful for measuring hemodynamics.

DOT propagates light through the tissue at multiple projections to yield a 3dimensional tomographic image of deep tissue organs and, depending on the mathematical approach taken, maps of absorption, scattering, vascularization, oxygenation, and contrast agent uptake can be generated (119). In a recent study that used DOT concurrently with dynamic contrastenhanced MRI, the distribution of fluorescent indocyanine green (ICG) was compared to that of an MRcontrast agent, GdDTPA, and found that they were correlated

(120).

The transillumination limitation of optical imaging can be overcome with a preclinical, in vivo chronic transparent chamber angiogenesis assay, the dorsal skinfold window chamber technique. Coupled with intravital fluorescence microscopy, this technique has provided

48 tremendous insight into basic vessel physiology and is capable of providing quantitative measurements of blood flow, vascular volume, permeability, vessel density, gene expression and drug delivery in animals (121). Tozer et al. has used the window chamber model and intravital fluorescence microscopy to show combretastatin A4 3Ophosphate and nitric oxide synthase effects on vascular permeability (122). Table 1.5 summarizes the list of studies using optical imaging to measure hemodynamic parameters according to the system and parameters measured along side the critical findings for each study.

2.5.1. Kinetic Modeling of Dynamic Optical Imaging Data

A very comprehensive study using fluorescence video microscopy to measure vascular permeability was performed in the window chamber model (123). The derivation of vascular permeability as measured optically begins with the assumption that the relationship between number of fluorochromes ( N(t) ) in the vascular and interstitial space and the average fluorescent intensity ( I(t) ) in the vascular and interstitial space is linear:

k ⋅N (t) I (t) = v v , Equation 15 v A

k ⋅ N (t) I (t) = i i , Equation 16 i A where kv (vascular) and ki (interstitial) relate the number of fluorochromes in area A to the average light intensity in A.

From Fick’s Law, the average permeability coefficient ( P) is the ratio of the flux of tracer

(Js) to the product of the vessel surface area and difference between the plasma concentration ( Cp) and the average interstitial concentration ( Ci):

J P = s . Equation 17 S ⋅(C p − Ci )

Assuming that the flux is approximately equal to the rate of accumulation of tracer in the interstitial space, and substituting Cp and Ci for parameters that account for vascular volume ( Vv), tube hematocrit ( HT) and volume to surface area ratio ( λ), the following equation is derived:

t −β (t−τ ) Ii (t) = α ⋅ e ⋅I v (τ )dτ , Equation 18 ∫0 which relates the constants α and β to permeability as follows:

  k   v  . Equation 19 P = α ⋅ λ ⋅()1− HT ⋅    ki 

A complete description of the derivation is provided elsewhere (123).

2.6 Nuclear Magnetic Resonance

Nuclear magnetic resonance is a spectroscopic technique designed to assess the chemical and physical nature of molecules using electromagnetic radiation. In the physical and chemical context, nuclei may possess spin angular momentum and orbital angular momentum. Spin angular momentum describes the rotation of the nucleus about its internal axis and orbital angular momentum describes the rotation of the nucleus about an external axis. The magnitude and direction vector ρ , describes the cumulative effects of

50 the two angular momenta. Such charged particles with angular momenta generate their own local magnetic fields or magnetic dipole moments. The dipole moment, , and its angular momentum, ρ , have discrete values and can be characterized by the nuclear spin quantum number, I. If the nucleus has an odd number of protons or neutrons, the spin is nonzero, and depending on the isotope, an integer or halfinteger. The hydrogen isotope,

1H, is the most often exploited NMRactive isotope due to its abundance in humans. 1H consists of only one proton, and is a spin½ nucleus.

Assuming an external magnetic field, B 0, is aligned along the zaxis of a Cartesian coordinate frame, zcomponents of the two vectors and ρ do not align exactly along the direction of the external field but rather at some offset angle (e.g. for 1H, that angle is

54.7 °). The magnitude of the zcomponents of these vectors (vectors coincident with B 0) is a function of Planck’s constant divided by 2π, h and the gyromagnetic ratio, γ of the nucleus:

h γh ρ = ± and = ± Equation 20 ± z 2 ± z 2

Furthermore, there are two possible orientations of and ρ for a spin ½ nucleus: spin up (parallel to B 0) or spindown (antiparallel to B 0). There are two corresponding energy states for these orientations, and the potential energy of each is dependent on field strength:

γhB E = ± ⋅ B = ± 0 Equation 21 0 2

Spins oriented along the main magnetic field occupy a lower energy state, while spins oriented against the main magnetic field occupy a higher energy state. The equilibrium distribution of the two energy states is illustrated by the Boltzmann distribution:

N = e(−E / kT ) , Equation 22 where N is the number of nuclei, T is the temperature in degrees Kelvin, and k is the

Boltzmann constant (1.38 x 10 23 Joules/K). Boltzmann statistics indicates that the ratio of the two orientation populations is equivalent to the energy difference between the spin states (E). That population difference between the energy states can be expressed in terms of the total number of nuclei in the specimen, N T:

N γhB N = T 0 . Equation 23 2kT

The NMR signal or more accurately, the net magnetization is related to that population difference:

γhN M = . Equation 24 0 2

When an electromagnetic (EM) radiation frequency (later conceptualized as a radiofrequency (RF) pulse) resonates with the energy level difference between spin states, the nuclei will absorb the energy and transition between energy states (from low to high). The difference in the energy of each state is proportional to the frequency of the applied EM radiation, υ :

E = hυ . Equation 25

The EM frequency that satisfies the resonance condition can then be specified as:

γ υ = B . Equation 26 2π 0

From a classical perspective, when a 1H is placed in a static magnetic field, the combination between the torque exerted on the nucleus by B 0 and the offset alignment of the nucleus in B 0 force the nucleus to precess about B 0 according to the Larmor equation:

ω = −γB0 . Equation 27

This equation is functionally equivalent to Equation 26, though dimensionally, Equation

26 is expressed in rad/sec and Equation 27 is expressed in Hertz. Thus, RF energy applied at the Larmor frequency will also satisfy the resonance condition.

Briefly suggested earlier, the net magnetization, M0, is cumulative sum of all the individual components of , and can be characterized as the population difference between spin states. It is more practical when describing spin behavior to express this difference as a single vector. The net magnetization, M 0, can then be considered the component of the M vector that is coincident with B 0. The longitudinal net magnetization, M z, is equivalent to M 0. The M vector precesses about B 0 in the same manner as the individual magnetic moments.

To create signal from a sample of NMRactive nuclei, a current induced magnetic field is applied perpendicular to B 0. When the frequency of this applied magnetic field, B 1, matches the Larmor frequency, the net magnetization, M 0, precesses into the transverse

(xy) plane. The angle ( θ ) at which M 0 rotates away from the zaxis into the transverse plane is calculated by:

θ = γB1τ , Equation 28 where τ is the time for which B 1 is applied. A traditional spin echo pulse sequence begins with a 90 ° “excitation” or initiation pulse. An excitation pulse strength and duration equivalent to 180 ° would tip the spins past the transverse plane onto the negative zaxis.

Following the removal of the applied RF energy, M z will relax exponentially back to

Boltzmann equilibrium according to the intrinsic relaxation properties of those spins. The

T2* relaxation is a combination of the intrinsic T 2 relaxation of the nuclei and T 2†, which is a combination of field inhomogeneity, magnetic susceptibility and molecular diffusion.

A sinusoidally oscillating signal called the free induction decay (FID) is recorded by a receiver after the removal of the excitation RF energy. The FID is a convolution of a sinusoid and a decaying exponential. The Fourier transform of that FID would be a

Lorentzian function. The FID decays due to dephasing or loss of coherence of the spins in the transverse plane. The spins lose coherence because the nuclei precess at different frequencies, some more slowly than the Larmor frequency and some more quickly. In a traditional spinecho sequence, at a given time, τ , another RF energy pulse is applied at a tip angle of 180 ° in order to change the position of those spins about the yaxis by 180 °.

At this point, spins that were moving more slowly than the Larmor frequency are closer to the initial position instantly following the 90 ° pulse, and the spins that were moving

54 faster than the Larmor frequency are farther away from that instantaneous position.

Ultimately, both classes of spins converge on the yaxis (axis of convergence depends on the phase of the 90º and 180 º RF pulses), resulting in the characteristic FID echo.

In 1946, Felix Bloch, described the various components of the M vector (M x, M y & M z) in the presence of B 0 and B 1 using a profile of differential equations:

dM x M x = γB0 M y + γB1M z sin(ωt) − , Equation 29 dt T2

dM y M y = −γB0 M x + γB1M z sin(ωt) − , Equation 30 dt T2

dM z M z − M 0 = −γB1[M z sin(ωt) + M y cos(ωt)]− . Equation 31 dt T1

For simplicity, when referencing these equations, the frame of reference is often described as rotating at the Larmor frequency (as opposed to remaining stationary). In this case, the equations are simplified (i.e., the B 0 components become zero), and classically, the description of the net magnetization vector following an RF pulse is more easily depicted. Furthermore, in the absence of a B1 field (e.g. following an excitation pulse), the B 1 components also become zero, and the Bloch equations can be rewritten to express the magnetization vector in terms of the relaxation rate constants, T 1 and T 2:

dM xy M xy = − , Equation 32 dt T2

dM M − M z = − z 0 . Equation 33 dt T1

The two main relaxation mechanisms are T 1 (logitundinal, spinlattice) and T 2

(transverse, spinspin) relaxation. Since nuclei in a biological sample can have a variety

55 of different relaxation time constants, these differences can contribute to a wide assortment of contrasting pixels in an image.

Longitudinal or T 1 relaxation occurs because the nuclei release their potential energy into the surrounding molecular and atomic environment. Free water has an exceptionally long relaxation rate constant because as a species, it is small and therefore has a fast rotational rate (short correlation time). When transiently bound to large molecules, that rotational rate slows and the T1 relaxation is considerably shortened.

Transverse or T 2 relaxation processes are a function of the static or slowly fluctuating local magnetic fields in the specimen. When nuclei experience those local magnetic fields, they adapt to the frequency and temporarily resonate at that frequency. This results in a phase shift and ultimately, a loss of phase coherence of the net magnetization in the transverse plane. Free water also has an exceptionally long T 2 relaxation rate constant.

Larger molecules are associated with slow molecular mobility and slower rotational frequencies and short T 2 relaxation times. When sufficient time has passed following an

RF excitation source, the composite vector of magnetization will dephase in the xy plane according to the T2* decay mechanisms particular to that environment. In a standard spin echo sequence, a 180 ° pulse would be used to refocus that T2* decay, and continuous

180 ° pulses could be used to refocus the T2* decay as long as the pure T2 decay has not completely attenuated the magnetization. Measured decay in the signal recorded through the FID should thus result exclusively from intrinsic T 2 relaxation.

2.7 Magnetic Resonance Imaging

Magnetic resonance imaging is a 2 and 3dimensional imaging technique based on the principles of NMR. The history of MRI is rich and interestingly, spawned from the interest of delineating diseased tissues from normal tissues. In 1969, Raymond Damadian conceived of the concept of imaging using NMR principles, but it wasn’t until 1971 that

Damadian proposed that “Spin Echo NMR” was capable of detecting large relaxation differences between malignant tumors and normal tissue in the rat, and interestingly, between malignant tumor tissue and benign fibroadenoma tissues (124). He concluded that these differences were apparent due to the motion of water in these tissues. One year later, Paul Lauterbur proposed and published the use of a gradient for spatial localization, though the concept for that use had been previously published by Carr & Purcell in 1954

(125). In 1974, Peter Mansfield published a 3dimensional scanning method and in 1975

(126), Ernst introduced the concept of phase and frequency encoding, and the use of the

Fourier transform (127). In 1977, the first image of a human was produced by Damadian et al. (128) and again one year later in a patient with a lung lesion from metastatic breast adenocarcinoma (129).

MR Imaging is based on the ability to spatially localize the spins in a sample through the use of magnetic field gradients. That information is encoded into the phase and frequency of the NMR signal. The first important gradient for localization is the slice select gradient. This gradient varies linearly with distance from its center. From one end of the

57 specimen to the other, spins are resonating according to the strength of the gradient at that position. Slice selective pulses define a slice of spins in the specimen by applying an RF pulse at the resonant frequency of the spins in that slice. Since spins on either side of the pulse resonate at different frequencies (due to their position along the linearly varying gradient), they will not absorb the energy from the slice excitation pulse. The bandwidth of that RF pulse describes the band of frequencies contained in that excitation pulse profile. The thickness of the slice can be manipulated by changing the bandwidth (narrow

BW produce narrow slices) or increasing the gradient strength (increases in gradient strength produce thinner slices).

The frequency endcode gradient (or readout gradient) is another linear magnetic field gradient. This gradient is applied along one axis perpendicular to a selected slice through the specimen, encoding frequency information in that slice during acquisition of the

NMR signal. As with slice selection, the precessional frequencies of the nuclei that experience this gradient will vary linearly with the gradient strength, and those spins are frequency encoded according to their position along that gradient. Assuming no other modifications are necessary, the composite signal from all of the spins in that slice could be Fourier transformed into frequency and amplitude information. However, in order to establish spatial information in more than one dimension, a third spatial gradient is applied before the frequency encode gradient and after the slice encode gradient.

The phase encode gradient applies a finite phase shift to all the spins in the slice a certain number of times during repetition time (TR). Like the other gradients mentioned, the phase encode gradient varies linearly from its center, so spins at the center of the phase encode gradient do not experience any phase change. The first in a series (e.g. 256) of phase encode gradient applications (in a single TR) imparts a spatially dependent angular frequency on the spins in the selected slice, and when the application is terminated, those have a spatially dependent variation in phase shift when they return to the Larmor frequency. The following views or phase encode gradient applications incrementally change in gradient strength thereby producing an array of spatially encoded phase variations that are ultimately be decoded by the Fourier transform.

The raw data from a single experiment for every slice are stored in a raw data matrix referred to as “kspace”. Image data are obtained through Fourier transform of the raw data matrix. This topic has been described extensively elsewhere (130).

2.8 Measurement of Hemodynamics with Magnetic Resonance Imaging

Dynamic contrast enhanced (DCE)MRI has been used to measure microvascular permeability in many human tumors, such as bone sarcomas (131), headneck and prostate tumors (132), breast tumors (133), cerebral gliomas (134), and breast tumors

(135). DCEMRI is a noninvasive, functional method to measure active changes in hemodynamic parameters occurring following treatment before macroscopic changes, such as decreases in tumor volume, become evident. Currently, there are at least 11

59 clinical trials in all four phases dedicated to the interrogation of tumors and tumor response to therapy using information derived from DCEMRI (118). DCEMRI is used to experimentally characterize the tumor microvasculature by modeling the pharmacokinetics of injected contrast agents. These unique MRI contrast agents are chelated with Gadolinium and generate very large magnetic fields in the immediate neighborhood of the complex, greatly shortening the T1 or longitudinal relaxation of water protons that approach the paramagnetic center. This lends to the unique contrast between tissues observed using this imaging approach. The efficiency with which a contrast medium can shorten T1 relaxation is determined by its inherent relaxivity. Table

6 summarizes the list of studies using DCEMRI to measure hemodynamic parameters according to the system and parameters measured along side the critical findings for each study.

2.8.1. MRI Contrast Agents

The challenges remaining in the development and application of MR contrast agents are

1) to find agents that are most sensitive to microvascular permeability and; 2) to quantify the values obtained from images. Generally, DCEMRI measurements using high molecular weight contrast agents (HMWCA, i.e. > 30 kDa) are more sensitive to the transendothelial capillary permeability, while DCEMRI measurements using small molecular weight contrast agents (SMWCA, i.e. < 1 kDa) are more sensitive to flow.

Although SMWCA are commonly used in the clinic, they rapidly extravasate in tumors and normal tissues with fenestrated or sinusoidal capillaries. Quantification of this rapid

60 extravasation requires rapid image acquisition, at the expense of reduced spatial resolution. Furthermore, extravasation rates are flowdominated, so SMWCA are insensitive to druginduced changes in microvascular permeability. HMWCA do not extravasate rapidly, and measurements made using them are more sensitive to changes in microvascular permeability. Imaging performed using HMWCA can be at higher spatial resolution at the expense of temporal resolution, given the slow extravasation rates.

Thus, there is an interest in developing intravascular compounds that demonstrate a large dynamic range and sensitivity to changes following interventional therapies in cancer

(136). To be clinically useful, these agents must be nontoxic and have high relaxivity.

Compounds under development are of a larger molecular weight than the clinically used agents (~1 kDa), ranging from 30100 kDa, offer a longer blood pool residence time and permit acquisition of steady state images for more than an hour. These high molecular weight contrast agents remain intravascular in all tissues except for leaky tumor vasculature (137139).

2.8.2. Kinetic Modeling of DynamicContrast Enhanced MRI Data

A major difference between DCEMRI data and data from other techniques is the fact that DCEMRI measures the effect of the contrast agent on water relaxation and not the contrast agent itself, thus deriving concentration of the contrast agent from signal intensities requires the characterization of a specific relaxivity for that contrast agent.

Once the relationship between concentration and signal intensity has been established via the relaxivity, pharmacokinetic models can be to quantitatively evaluate the concentration

61 versus time curves to extract values for the volume transfer constant ( Ktrans ) between the plasma compartment and the interstitial space and the fractional volume of the interstitial

1 trans space ( ve) (140). The rate constant, kep (min ), is the ratio of the transfer constant (K ,

1 min ) to the extravascular extracellular fractional volume, ve:

trans kep = K / ve . Equation 34

This rate constant can be directly derived from the shape of the contrast agent

trans concentration versus time curve, but deriving K and ve require knowledge of absolute values of contrast agent concentration.

If the intrinsic permeability of the capillary is high, then transport of a solute across the capillary is dependent on blood flow rather than diffusion, and as such is flowlimited .

Tofts et. al. have extended this interpretation to include flow per unit mass of tissue as opposed to total blood flow, and mass tissue concentration ( Ct/ρ) instead of volume tissue concentration ( Ct) (140):

dC t = Fρ(C − C ) . Equation 35 dt a v

In this case, ρ (g ml 1) is needed because F is now expressed in ml g 1 min 1. The arterial and venous whole blood tracer concentrations are represented by Ca and Cv (mM), respectively, and the tissue tracer concentration is represented by Ct (mM). For more details on parameters and the following derivations, please see Appendix A. The partition coefficient, λ, relates the concentration of the tracer in the tissue to the concentration of the tracer in the blood:

Ct = λ Cv . Equation 36

It then follows that:

dC F t = − (C − λρC ) . Equation 37 dt λ t a

In a flowlimited regime, the venous whole blood concentration, Cv, or the venous plasma concentration (= Cv/(1Hct)) can be assumed equal to the concentration in the extravascular extracellular space ( Ce, i.e. C v = (1Hct)/C e). In this case, λ is derived in terms of ve, and the final model is expressed as:

dC t = Fρ 1( − Hct)(C − C / v ) . Equation 38 dt p t e

This model is used in situations where capillary permeability to the contrast agent is very high (i.e., when small molecular weight contrast agents are used). The rate of uptake of the contrast agent into the tissue is related to the ve and the contrast agent concentrations in the tissue ( Ct) and the arterial blood plasma ( Cp). Importantly, the transfer constant,

Ktrans , is a function of the perfusion of whole blood per unit mass of tissue ( F, ml g 1 min

1), the density of tissue ( ρ, g ml 1) and the hematocrit ( Hct ) (i.e.,K trans = Fρ 1( − Hct ) )

(140).

In situations where capillary permeability to the contrast agent is very low (i.e. when high molecular weight contrast agents are used), Ktrans is a function of the permeability surface area product ( PS , ml min1 g1) of the capillary wall and ρ, (i.e. K trans = PSρ ), and the rate of uptake is related to the ve, Ct and Cp:

dC t = PSρ(C − C / v ) . Equation 39 dt p t e

Finally, in situations where capillary permeability is intermediate, Ktrans is equal to the fractional reduction in capillary blood concentration during the passage of the contrast agent through the tissue (or the extraction ratio, E) and the hematocrit (i.e. K trans = EFρ ), and the rate of uptake is related to the ve, Ct, and Cp:

dC t = EFρ 1( − Hct)(C − C / v ) . Equation 40 dt p t e

In summary, the contrast agent concentration kinetics are a function of the plasma

trans trans ep concentration, K and ve, or K and k . This is expressed by a generalized form for each of the three previous models:

dC t = K trans (C − C / v ) = K transC − k C , Equation 41 dt p t e p ep t which describes the change in the concentration of the contrast agent as a function of

trans time in terms of K and ve. It is important to note here that each model (high, low or intermediate capillary permeability to contrast agent) ignore the contribution of intravascular contrast agent to tissue concentration, and therefore, limit the fitted model

trans parameters to only K and ve. Contrast agents of different molecular weights will require the use of different models for permeability. Of the listed parameters, Cp and Ct

trans are actually measured and K , kep and ve are calculated from the fitted model equations.

2.8.3. Transcytolemmal Water Exchange

Most DCEMRI analyses are conducted based on assumption that the exchange of water across the cell membrane occurs of infinitely fast. In 1999, Landis et al. pointed out that such an exchange in most cells was not occurring at a rate beyond the temporally detectable limits of NMR (141). Consequently, these researchers developed a model to account for transmembrane water exchange by modifying the Bloch equations to represent two separate extravascular compartments, the extra and intracellular compartments. Following subsequent empirical demonstrations of these modifications, they contended that in some cases, the exchange of protons from the extracellular extravascular space to the intracellular environment did not occur infinitely fast and that such an assumption led to considerable underestimations of vascular hemodynamics. This discovery has contributed to the beginning of a logical reevaluation of experimental and mathematical approaches to imaging vascular hemodynamics by the MRI community

(142145). In 2003, Yankeelov et al. modified the signal equation to estimate vascular permeability under these conditions (146). Because of the NMRmeasurable exchange of water across the cell membrane, the relationship between the relaxation rate constant R 1

1 (T 1 ) and concentration of a gadoliniumchelated contrast agent was not linear. The relaxation rate constant R 1 (1/T 1) is used to calibrate the relationship between contrast agent and signal intensity:

R1 = r1 ⋅[CR] + R10 , Equation 42 where r1 is the relaxivity, and R 10 is the precontrast relaxation rate constant. In order to describe the relationship between contrast agent concentration and signal intensity in the

65 plasma (the only compartment in which the contrast agent resides in the whole blood) the equation is modified to give:

R1b = r1p ⋅ 1( − Hct) ⋅[CR] + R1b0 , Equation 43

The term b denotes whole blood, p denotes whole plasma and Hct is the hematocrit.

However, half of the blood water population is intracellular thus, this equation is not sufficient. The next two equations represent a 2site exchange for a water molecule in an erythrocyte:

H 2Oi ↔ H 2Oo , Equation 44

H 2Oo + CRo ↔ H 2O ⋅CRo . Equation 45

The first equation is true in homogenous solutions (where i and o denote inside and outside the cell), and the second equation is true because the exchange of water across the erythrocyte membrane is in the fastexchange limit (FXL) of the NMR time scale at practical contrast agent concentrations. Both equations are FXL expressions.

In the interstitium, an equation can be derived as well, since the contrast agent remains only in the interstitial space:

R1 = r1 ⋅ po ⋅[CR] + R10 , Equation 46 where p o is the fraction of tissue water that is extracellular and r1 is the interstitial relaxivity of the contrast agent.

In the case of biological tissues, the major assumptions are that the interstitial space behaves as a wellmixed, homogenous solution, and that the exchange of water protons

66 across the cell membrane is occurring infinitely fast within the FXL. Since we know that the interstitial space is far from a well mixed, homogenous solution, that it is in fact, highly compartmentalized and that in most cells which are larger and less permeable than erythrocytes, the exchange of water across the cell membrane is not occurring within the

FXL at any contrast agent concentration. For these situations, the return of M z magnetization can be characterized by a biexponential recovery:

−t R I 1L −tI R1S M z = M 0 1[ − (2 aLe + aS e )] , Equation 47 where M z is the instantaneous magnetization, M 0 is the Boltzmann equilibrium value, a L and R 1L are the apparent fraction and rate constant for the nuclear system with the longer

T1 and a S and R 1S are the apparent fraction and rate constant for the nuclear system with the shorter T 1.

In 1958, McConnell generalized the Bloch equations to include the effects of chemical exchange (147). He started by using the conventions that Bloch used, that is, letting u, v, and Mz denote the components of the nucleus’ magnetizations, which are in phase with the effective rotating component of the radiofrequency field, out of phase with the rotating radiofrequency field and in the direction of the large stationary field, respectively. Bloch showed that:

M x = u cosωt − υ sin ωt , Equation 48

M y = ±(u sin ωt + υ cosωt) . Equation 49

McConnell stated that these magnetizations can be written as the sum of the contributions of the A and B systems:

u = u A + u B

υ = υ A + υ B Equation 50 A B M z = M z + M z

McConnell then described the modified Bloch equations as follows:

• u A u B u A + ω Aυ A = − + τ 2 A τ B • u B u A u B + ω Bυ B = − + τ 2B τ A • υ A υ B A υ A + ω Au A = − + − ω1M z τ 2 A τ B • Equation 51 υ B υ A B υ B + ω B u B = − + − ω1M z τ 2B τ A

• A A B A M 0 M z M z M z − ω1υ A = + + T1A τ 1A τ B

• B B A B M 0 M z M z M z − ω1υ B = + + T1B τ 1B τ A

A B where M 0 and M 0 are the equilibrium z magnetizations of the nuclei in A and B (dots

correspond to vector notation), ω1 = γH1 , and

1 1 1 = + τ1A T1A τ A Equation 52 1 1 1 = + , τ 2 A T2 A τ A

where T1A and T2 A are the longitudinal and transverse relaxation times of the nucleus in

A. Similar definitions apply for τ 1B and τ 2B .

Then, in 1961, Woessnner made the same modficiations with different notations (148). In this case, the nucleus makes an instantaneous transfer between two different state

68 environments a and b. The average lifetimes of a given nucleus in states a and b are given as reciprocals of the probabilities per second or rates Ca and Cb for leaving these states.

The fractions of nuclei in states a and b are Pa and Pb.

The modified Woessner equations are:

• U a U a = −(ω a − ω)Va − − C aU a + CbU b T2a • U b U b = −(ωb − ω)Vb − − CbU b + C aU a T2b • Va V a = (ω a − ω)U a − − C aVa + CbVb − ω1 M za T2a Equation 53 • Vb V b = (ωb − ω)U b − − CbVb + C aVa − ω1 M zb T2b • M 0a M za M za = − − C a M za + Cb M zb + ω1Va T1a T1a • M 0b M zb M zb = − − Cb M zb + C a M za + ω1Vb T1b T1b

1 where Ca = (from Woessner), ω1 = γH1 , ωa is the precessional frequency, T1a and τ A

T2a are the longitudinal and transverse relaxation times for the nucleus in state a, and

M 0a is the equilibrium magnetization along the direction of the stationary field for nuclei in a. Same is true for the b states.

In the absence of the radiofrequency field, ω1 = 0, and the last two equations above can be rewritten as:

• • M 0a − M za = −α1a (M 0a − M za ) + Cb (M 0b + M zb ) Equation 54 • • M 0b − M zb = −α1b (M 0b − M zb ) + Ca (M 0a + M za )

The sum of the solutions to the equations above evaluated at time τ (no, I did not solve

these equations) is expressed below. This function, F(T1 ) , describes the time dependence of M z under the influence of spinlattice relaxation and the environmental state transitions:

F(T1 ) = [g'exp(−c'τ ) + 1( − g )' exp(c'τ )]]× exp(−φ1τ )  1 1  φ = 1  + C + C  1 2 a b   T1a T1b 

  1 1   1 − 1 P − P  −  + C + C 2 4 ()b a   a b    T1a T1b   g' = Equation 55 c'

1 2 2  1 1   c'= 1  − + C + C  + 4C C  2  T T a b  a b  1a 1b   Recalling the magnetization equation from Yankeelov et al.:

M z = M 0 [1− 2(aL exp(−tI R1L ) + aS e(−tI R1S ))] Equation 56 and the Woessner derivation:

F(T1) = [g'exp([−c'−φ1]τ ) + 1( − g )' exp([c'−φ1]τ )]] Equation 57 it can be said that:

R1L = φ1 + c' and Equation 58

R1S = φ1 − c'

To consider only the case of R1L , we can see that:

1  2 2    1 1    1 1    R = 1  + + C + C  + 1  − + C + C  + 4C C  Eq. 59 1L  2  a b   2  a b  a b    T1a T1b    T1a T1b       

Considering these conventions, the last equation can be expressed as follows:

1  2 2    1 1    1 1  1 1   R = 1  R + R + +  + 1  R − R + +  + 4  Eq. 60 1L  2  1i 1o   2  1i 1o     τ i τ o    τ i τ o  τ i τ o      

Further substitution of R1o results in:

  [R + 1( − p )R ] 1 1 − p  R = 1  R + r [CR ] + 10 o 1i + + o  1L  2  1i 1o o    po τ i τ i po  1 Eq. 61  2 2    [R + 1( − p )R ] 1 1 − p  1 1 − p   + 1  R − r [CR ] + 10 o 1i + + o  + 4 o   2  1i 1o o     po τ i τ i po  τ i τ i po      

Simplifying results in:

  [R + R + ]1  R = 1 2R + r [CR ](t) + 10 1i  − 1L  2 1i 1o o    τ i po  1 Equation 62  2  2    2 [R10 + R1i + ]1  1( − po )   − r1o [CRo ](t) +  + 4   τ τ p  2  i i o  τ i po     

aS (t) The same substitutions can be applied to R1S and R1S . aS + aL

  [R + R + ]1  R = 1 2R + r [CR ](t) + 10 1i  − 1S  2 1i 1o o    τ i po  1  2  2    2 [R10 + R1i + ]1  1( − po )   − r1o [CRo ](t) +  + 4   τ τ p  2  i i o  τ i po     

These modifications to the Bloch equations for NMR relaxation in the presence of 2site exchange can be applied to standard pharmacokinetic models to provide a more comprehensive and accurate estimation of hemodynamic parameters using DCEMRI.

3 Section 3

3.1 Antivascular &. Antiangiogenic Therapies

Antivascular therapies target an established vessel network. In general, they induce collapse of a mature vasculature, reduce perfusion, blood volume and vascular tortuosity

(149). Available antivascular therapies have been tabulated (Table 7).

Antiangiogenic therapies, on the other hand, target actively proliferating and newly forming vessel networks. Most antiangiogenic therapies are small molecule inhibitors of vascular growth factors that support the neovasculature. In general, antiangiogenic therapies reduce permeability, perfusion and blood volume (Table 8).

The table of antiangiogenic inhibitors above is a conglomeration of direct and indirect inhibitors. Indirect inhibitors of angiogenesis target tumor cell production of angiogenic growth factors and their receptors. This type of inhibitor is prone to development of resistance as there is a substantial opportunity for the cell to pursue alternative pathways of angiogenesis. Direct inhibitors of angiogenesis target tumor endothelial cells and inhibit endothelial cell proliferation, migration, tube formation and/or induce endothelial cell apoptosis (150).

Successful clinical trial design for conventional cytotoxic agents is based on the following concepts: 1) the agent is associated with a dosedependent toxicity; 2) an upper limit for doseescalation will be set by a doselimiting toxicity (DLT) (e.g. anemia); 3)

73 the maximum tolerated dose (MTD) has the highest probability of producing objective remission (shrinking the tumor) and improving palliation of symptoms and; 4) the agent, alone or in combination, is capable of prolonging survival. The standard paradigm for selecting the dose of chemotherapeutic agents, the MTD, is not appropriate for antiangiogenic drugs because a) as these are targeted therapies, the optimal dose may be much less than the tolerated dose, b) they are associated with low incidence of side effects and; c) the DLT may not be biochemically related to the agent’s mechanism of action. Since antiangiogenic and antivascular therapies are cytostatic, measuring tumor volumes is impractical because tumor shrinkage or disease progression may not occur or could take weeks or months to be visible. Furthermore, heterogeneity in individual responses to cytostatic therapies would require large populations to obtain significance, leading to long and expensive clinical trials. Therefore, many potential antiangiogenic therapies have not lived up to the expectations of preclinical trials, in part because of the difficulties in designing clinical trials for antiangiogenic drugs.

Thus, there is a need to establish sensitive and reliable biomarkers of early responses to antiangiogenic or antivascular drugs, for in vivo pharmacodynamics (i.e. action of a drug on its target). Drug action can then be compared to subsequent clinical responses to ascertain if it is effective. Because antiangiogenic or antivascular therapies are designed to affect the abnormal blood vessels recruited by tumors, changes in blood volume, blood flow, or other hemodynamic parameters may be promising biomarkers that allow for an in vivo assessment of the biological activity of these angiogenic modulators.

3.2 Imaging Response to Antivascular vs. Antiangiogenic Therapies

Methodological details about the imaging methods that have exploited these physiological parameters are described elsewhere in this review. However, Table 1.9 presents a list of preclinical studies and clinical trials that have used these modalities to evaluate the effects of several important antiangiogenic or antivascular therapies.

4 Section 4

4.1 Endothelial Transport of Contrast Agents: Macromolecular Transport

Thus far, we have discussed the mechanisms regulating ion and small molecule transport through passive mechanisms, though it was indicated that active transport mechanisms may be regulating transport as well. Indeed, the concept of the endothelium as a passive, simple barrier to blood constituents has long since been replaced with the idea of an endothelium that can sense, modulate and control macromolecular transport. This section will focus on the transport of macromolecules and the characteristics that determine capillary permeability to these macromolecules.

The consensus among morphologists is that large molecules require a transcytotic event or active, transcellular shuttle mechanism to get from the luminal end to the abluminal end and vice versa. The first system proposed described the shuttling of plasmalemmal vesicles or caveolae across the endothelial cells between the luminal and abluminal ends

(151). An alternative form of transport is mediated by socalled “vesicularvacuolar organelles” (VVO), which create transendothelial channels that span the luminal to the abluminal end (152). Independent observation of a membranebound tubular system evidently used for transport sparked a debate as to whether or not all three systems were describing the same phenomena (153).

The regulation of transport naturally begins at the level of the luminal plasmalemma. As the forefront of the transport process, this luminal membrane interacts with all blood

76 constituents and functions in sorting these molecules to the designated location. The luminal plasmalemmal surface is comprised of glycoconjugates (e.g. glycoproteins, glycolipids and proteoglycans), enzymes (e.g. lipoprotein lipase and angiotensin convertin enzymes) and receptors for specific plasma molecules (e.g. vasoactive amines, hormones, procoagulants, anticoagulants, carrier proteins and lipoproteins) (154). This expression profile differs significantly from the abluminal surface of the plasmalemma, suggesting that the cell surface protein distribution of both interfaces is important for moving blood constituents either from the luminal compartment of the capillary to the interstitial compartment in the tissue or from the interstitial compartment to the lumen of the capillary (155).

Independent transport mechanisms mediate the trafficking of materials across the plasma membrane. The functional units of these mechanisms, including interendothelial junctions (paracellular transport), caveolae (transcytosis) and their transendothelial channels, and vesiculovacuolar organelles (VVO), are diagrammed in Figure 2 and discussed in greater detail below.

4.1.1. Paracellular Transport

Paracellular transport can be described as transport through the interendothelial junctions between adjacent endothelial cells. This system limits transport to ions with radii less than 3 nm, such as urea, glucose and small ions. However, under certain conditions, these tightly regulated interendothelial junctions will transform to allow intercellular gaps to

77 form, through which select macromolecules like plasma proteins can pass into the extravascular extracellular space (155, 156). This process is mediated by binding of

VEGF and certain inflammatory mediators to their receptors.

4.1.2. Transcellular Transport

Transcellular transport or t ranscytosis describes the process whereby molecules are translocated across the cell to the interstitial fluid. Transcytosis can occur via non specific mechanisms (fluid phase and/or adsorptive) or specific mechanisms (receptor mediated) (155). The process of fluid phase transcytosis is described as endocytosis of the plasma by caveolae and the subsequent relocation of the contents to the abluminal surface where the vesicle is disencumbered. Adsorptive nonspecific transcytosis involves an electrostatic interaction between the vesicle and the molecule before the translocation event can occur. Specific transcytosis is entirely receptormediated.

4.1.2.1. Caveolae

Caveolae are small vesicles (mean diameter ~70 nm) that are open to the endothelial cell membrane on either end—luminal or abluminal—via a thin neck of 1040 nm in diameter, endowing them with the “flaskshape” distinction often referred to in the literature. The aperture between the endothelial cell membrane and the caveolae is sometimes enclosed by a thin diaphragm displaying a central knob and spokelike features, which act as a size and ion filter (157, 158). These vesicles have been implicated in endocytotic and transcytotic transport mechanisms. The latter of the two will be

78 described in more detail. A review on the endocytotic transport pathway can be found elsewhere (155).

Caveolae have a heavy lipid composition, particularly, sphingomyelin and cholesterol.

The expression of caveolin1 is also an integral part of the caveolar composition. This protein mediates proteinprotein interaction with Gprotein subunits, HaRas, Src tyrosine kinases and eNOS (159).

4.1.2.2. Transendothelial Channels

The original postulation for transendothelial channels was not married to substantial evidence for their existence (158) until years later when it was evidenced more clearly

(160). These channels are created when a sequence of caveolae line up from one end of the plasma membrane to the other to form an open channel. It is presumed that these structures regulate function by providing a dynamic hydrophobic channel or pore.

4.1.2.3. Vacuolar Organelles

Vesiculovacuolar organelles are clusters of vesicles that can arrange into continuous channels that span across the endothelial cell membrane, and in some cases, have been observed spanning to the lateral endothelial cell interface. It is not entirely clear whether they are morphologically distinct from transendothelial channels (AL Baldwin, personal communication). Nevertheless, they were originally described by Dvorak’s group (152,

161) and have been reported elsewhere (162164). The size of these vesicles can range

79 from 80140 nm in diameter. They are clusters of smaller vesicles that are 12 m in the longest dimension. They are hypothesized to have some involvement in the transportation of large molecules because treatment of normal animals with a permeabilizing agent resulted in an increased extravasation of ferritin in normal venules by way of VVO structures (165). VVO structures also contain diaphragms that separate the luminal and abluminal interfaces from the interior of the vesicle. The only aspect that seems to significantly deviate from a caveolinlike morphology is that the behavior of these vesicles is completely normal in caveolindeficient mice, indicating that these structures do not require caveolin1 for function (166, 167).

CONCLUSIONS

This review has described both the physical basis of microvascular permeability as well as the mathematical methods required for a clinically practical imaging approach to estimating this parameter. It has presented methods that the imaging community has developed to deliver an estimation of microvascular permeability in the attempt to establish reliable and sensitive measures of response to therapy in human tumors. It is anticipated that microvascular permeability will evolve into a clinically useful biomarker that can 1) aid in the “Go/No Go” decision making process in drug development; 2) become a useful tool in establishing a drug’s mechanism of action and; 3) ultimately transform into a surrogate endpoint for drug approval purposes. A logical initiation point for this review was to define vascular permeability and perfusion, diagram major vessel architecture and mammalian regulation of blood flow, and survey the various morphologies of capillaries and modes of transport, both passive and active. In order to associate microvascular permeability with angiogenesis, mechanisms of normal physiological and tumor angiogenesis and the litany of in vitro and in vivo assays used to observe angiogenesis were described. These issues provided a reasonable segue into the current status of clinical imaging of perfusion, blood flow, permeability, distribution and vascular volumes using PET, SPECT, DCEMRI, ultrasound and optical imaging, particularly in the context of their ability to estimate tumor response to antiangiogenic and antivascular therapies.

There is no ‘silver bullet’ for consummate quantification of microvascular hemodynamics. No standard protocols for measuring microvascular hemodynamics exist and reproducibility is simply not adequate across image sites and different imaging centers. Perhaps a profile of hemodynamic data obtained by an array of imaging assays would lead to more reliable and robust indicators of therapy response than any one imaging modality could alone provide. Nevertheless, the contribution of each modality as an adjuvant to traditional tumor response indices is powerful and incontrovertible. At the very least, researchers interested in advancing a particular technique alone or in combination with other imaging modalities will continue to find fellowship in this large community of hypothesisgenerating research.

FIGURES

Figure 1.1 Crosssectional diagram of the distinct layers of a general artery (left) and vein (right): tunica intima, media and adventitia.

Figure 1.2 Two adjacent endothelial cells separated by interendothelial cell junctions (IEJ). Endothelial cells are characterized by extensive plasmalemmal vesicles or caveolae (C). Individual caveolae can be open to the luminal front (CL) or abluminal front (CA) or free in the cytoplasm. Caveolae can aggregate endtoend to create a transendothelial channel (TEC). The transcellular pathway mediates the transport of plasma proteins such as albumin via fluid phase or receptor mediated mechanisms, while small molecules such as glucose and urea are transported via the paracellular pathway.

TABLES Table 1.1: Comparison of Imaging Systems

Modality Contrast Penetration Resolution Sensitivity Agent/Tracer 15 15 11 11 12 PET H2 O, C O2, C Whole Body 13 mm 10 10 mol/L SPECT IMP, 99m TcAlb Whole Body 13 mm 10 10 10 11 sesamibii mol/L CT Iodinated Tracers Whole Body 0.050.8 mm Not well characterized Ultrasound Microbubbles 9 cm @ 7 MHz 0.1 mm Single bubbles MRI GdDTPA Whole Body 0.11.5 mm 10 310 5 mol/L Optical Fluorochromes <1 cm 0.21.0 m 10 910 12 Imaging mol/L

Table 1.2: Measurement of Hemodynamics with PET/SPECT

Tracer System Parameter Results Reference Measured PET ([ 15 O] Liver metastases Perfusion Comparing dynamic versus (63) 15 H2O, [ C] steady state methods,

CO 2) differentiation between tumor grades PET ([ 15 O] Breast tumors Perfusion and Evaluated noninvasive AIF (64)

H2O) distribution measurement and obtained volume better estimations of distribution volume PET ([ 15 O] NSCLC Perfusion and Large interpatient variations, (65) 18 H2O, FDG) tumor localization inaccurate estimation of perfusion PET ([ 15 O] Solid tumors Perfusion and Phase I doseescalation (67)

15 H2O, [ C]CO) blood volume study, perfusion and blood volume decreases in response to CA4P PET ([ 15 O] Solid/metastatic Perfusion and Phase II study, insignificant (68) 15 H2O, [ C]CO) renal tumors blood volume decreases in perfusion and BV in response to razoxane PET ([ 15 O] Renal tumors Perfusion and Perfusion decreases in (69)

H2O) metabolism patients with stable disease in response to SU5416 PET ([ 15 O] Breast tumors Perfusion and No significant association (70) 11 H2O, [ C]CO) blood volume between perfusion and response to treatment; insignificant increase in blood volume in non responding tumors PET ([ 15 O] Colorectal Perfusion Enhanced delivery of 5FU (71)

H2O) metastases via carbogen and nicotinamide, significant increase in perfusion post treatment PET ([ 15 O] Brain tumors Perfusion and Early uptake correlations (72) 18 H2O, F hypoxia between perfusion and FMISO) hypoxia, no late uptake correlations SPECT (IMP) Brain tumors Perfusion Differentiation between (74) perfusion of tumor and perfusion of normal tissue SPECT ( 99m Tc Liver tumors Perfusion Increased perfusion in tumor (75) HSA ) following infusion of epinephrine SPECT Brain metastases Permeability and Significant posttreatment (77) (201 TICI, 99m Tc viability reductions in permeability

HSA D) and viability SPECT ( 99m Tc Soft tissue and Perfusion and Perfusion and MIBI uptake (78) MIBI ) bone tumors MIBI uptake ratio ratio lower in benign versus malignant tumors, achievable predictive value of these parameters SPECT ( 99m Tc SCLC, GM, brain Perfusion and Inverse correlation between (79) HMPAO, metastases, hypoxia perfusion and uptake of IAZA ) prostate and head IAZA in all tumors but GM and neck tumors tumors

Table 1.3: Measurement of Hemodynamics with CT

Tracer System Parameters Results Reference Measured Iopamidol Rectal tumors Perfusion and mean Decreased perfusion and (81) (Isovue) transit time increased mean transit time posttreatment; high perfusion index and short MTT index poor response to therapy. Iomeron Oropharyngeal Permeabilit y, blood All parameters elevated in (82) (Iomeprol) and oral tumors flow, volume and tumors compared to normal mean transit time tissue except for low mean transit times; differentiation between primary and recurrent tumors based on blood flow Iodinated Head and neck Permeability, blood All parameters were (83) contrast agent tumors flow, blood volume significantly different in and mean transit tumors compared to normal time tissue

Iopamidol Hepatic Perfusion Perfusion elevated in (84) (Isovue 300) metastases metastases compared to contralateral tissue; significant correlation between survival and perfusion of metastasis (higher perfusion, increased survival) Iopamidol Lymphoma Perfusion and Perfusion correlated well (85) (Isovue 300) permeability with grade and activity but permeability did not reliably differentiate between grade or activity Iobitridol Nonsmall cell Perfusion, Reproducibility of whole (86) (Xenetix 30) lung cancer permeability and tumor perfusion both blood volume parameters was acceptable Iopamidol Colorectal Blood volume, Correlation coefficients for (88) (Niopam 340) tumors blood flow, mean interobserver variability and transit time and intraobserver variability permeability indicated excellent agreement in both cases Iopamidol Colorectal Perfusion Early changes in perfusion of (90) (Niopam 340) tumors tumors responding to bevacizumab demonstrated Iodinated Advanced solid Blood flow, blood Phase I doseescalation study (91) contrast agent tumors volume, of MEDI522; only mean refractory to permeability and transit time demonstrated therapy mean transit time significant changes post treatment

Table 1.4: Measurement of Hemodynamics with Ultrasound

Tracer System Parameter Results Reference Measured Sonovue® Power Doppler Percent intratumor Following RF (100) (microbubbles) US contrast agent ablation therapy, uptake DUPC accurately estimated degree of vascularization in 80% of cases Levovist® CEUS Tumor vascularity Postcontrast (101) (microbubbles) vascularity correlated well with histological assessment of vascularity Sonovue® CEUS, Number of CEUS outperformed (102) (microbubbles) conventional B metastases conventional US and mode US, spiral CT in its ability to CT accurately detect more metastatic lesions Sonovue® CEUS, color Enhancement Accurate (103) (microbubbles) Doppler US timing & pattern differentiation (related to between vascularity) nonfunctional islet tumors and pancreatic carcinoma Sonovue® CEUS Percent blood Posttreatment (104) (microbubbles) volume fraction, changes in these RBC velocity & parameters perfusion following treatment

with thalidomide

Table 1.5: Measurement of Hemodynamics with Optical Imaging

Tracer System Parameter Results Reference Measured ICG (indocyanine DOT Tracer Comparable (120) green) distribution distribution kinetics with GdDTPA (as measured by MRI)

Table 1.6: Measurement of Hemodynamics with DCEMRI

Tracer System Parameter Results Reference Measured GdDTPA Bone sarcoma Permeability Associations found between (131) (Magnevist) expression of VEGF and increases in permeability GdDTPA Head & neck, Rate of Demonstrated (132)

(Magnevist) prostate & uptake, k ep reproducibility of k ep brain neoplasms GdDTPA Breast tumors Permeability Significant change in (133) (Magnevist) permeability pre and post treatment Gadodiamide Cerebral Permeability Demonstrated (134) (GdDTPA glioma reproducibility of BMA) permeability GdDTPA Breast tumors Exchange Significant differentiation (135) (Magnevist) rates between malignant and

benign tissue

Table 1.7: Antivascular Therapies: Targets & Mechanisms of Action

Antivascular Agents Drug Target General Mechanism of Action Disrupts tubulin cytoskeleton, induces rapid Combretastatin A4 phosphate Tubulin vascular shutdown NAcetylcolchinolO Disrupts tubulin cytoskeleton, changes in Phosphate Tubulin cell morphology of proliferating endothelial (ZD6126) cells 5,6dimethylxanthenone4 Host immune Stimulation of host immune system, acetic acid (DMXAA) system induction of vascular collapse

Table 1.8: Antiangiogenic Therapies: Targets & Mechanisms of Action

Antiangiogenic Agents Drug Target General Mechanism of Action VEGFR1 & 2, AEE788 Dual inhibitor of EGFR and VEGFR EGFR Vandetanib VEGFR2, EGFR, Dual inhibitor of EGFR and VEGFR (ZD6474) FGFR1, RET VEGFR1, 2 & Inhibits VEGFR autophosphorylation, Axitinib 3, PDGFRβ, c inhibits endothelial cell proliferation and (AG013736) Kit survival VEGFR1, 2 & Sunitinib Inhibits endothelial cell proliferation and 3, PDGFRβ, c (SU11248) survival Kit VEGFR1, 2 & Recentin 3, PDGFRβ, c Inhibits VEGFinduced angiogenesis in vivo (AZD2171) Kit

VEGFR1, 2 & Vatalanib Inhibits endothelial cell proliferation and 3, PDGFRβ, c (PTK787/ZK222584) survival Kit VEGFR2 & 3, Sorafenib Raf kinase, Inhibits tumor and endothelial cell (BAY 439006) PDGFRβ, cKit, proliferation and survival RET O(chloroacetylcarbamoyl) Relatively specifically cytostatically inhibits EC growth (TNP470/AGM1470) endothelial cell growth (UI 8297716) Inhibits PI3kinase pathway, cellular PX866 PI3Kinase proliferation and survival Inhibits PI3kinase pathway, cellular LY294002 PI3Kinase proliferation and survival Inhibits PI3kinase pathway, cellular Wortmannin PI3Kinase proliferation and survival Bevacizumab VEGF Monoclonal antibody to VEGF (rhuMAb VEGF) Volociximab AAB1 targets AAB1, a component protein of α5β1 recombinant humanized IgG4k monoclonal HuMV833 VEGF antibody targeting VEGF Fusion of human IgG1 Fc to extracellular “VEGFtrap” VEGF domains of VEGFR1 and VEGFR2; binds (AVE 0005) VEGF IMC1121B VEGFR2 IgG antibody targeting VEGFR2 (antiVEGFR2MAb) α5β1, αvβ3 & Interferes with migration and proliferation of Endostatin αvβ5 endothelial cell Tumstatin αvβ3 Inhibits EC protein synthesis Competitively binds to integrins, inhibiting Angiostatin Integrin αvβ3 interaction with matrix ligands, interfering w/ cell attachment and adhesion Neovastat MMPs Inhibition of MMPs

Table 1.9: Imaging Response to Antivascular vs. Antiangiogenic Therapies

Drug Imaging Study Reference

CA4P PET Perfusion (67)

Razoxane PET Perfusion and blood volume (68)

SU5416 PET Perfusion and metabolism (69)

Bevacizumab CT Perfusion (90)

MEDI522 CT Blood flow, blood volume, PS, MTT (91)

AZD2171 DCEMRI – Decreases in K trans (168, 169)

SU11248 DCEMRI – Decreases in K trans & fPV 24h posttreatment (170)

AG013736 DCEMRI – Decreases in K trans 48h posttreatment (171)

PTK787/ZK222584 DCEMRI – Dose dependent reduction in K trans & AUC (172, 173)

AEE788 DCEMRI – Decrease in area under enhancement curve (174)

ZD6474 DCEMRI – Dose dependent reduction in K trans (175)

HuMV833 DCEMRI – Decreases in K trans 48h posttreatment (176)

trans Endostatin DCEMRI – No consistent change in K or ve (177)

CHAPTER 2

Combined Dynamic Contrast Enhanced Magnetic Resonance Imaging and Fluorescence Microscopy of Tumor Microvascular Permeability

INTRODUCTION

Tumor microcirculation differs significantly from a physiological vasculature.

Physiological angiogenesis in the adult is transient, limited to wound healing, inflammatory processes and reproductive cycling (1315). Physiological angiogenesis is equipoised on the fulcrum between inhibitors of angiogenesis like angiostatin, endostatin, thrombospondin1, angiopoietin2, and activators of angiogenesis, like VEGF, bFGF,

PDGF, TGFβ and angiopoietin1 (28). In contrast, angiogenesis in tumors is driven by disrupted morphogenetic gradients of these angiogenic factors, which are produced by actively proliferating tumor cells, and this imbalance may contribute to excessive branching and shunting and the characteristic hyperpermeability of tumor vessels (28, 57,

60).

DCEMRI is capable of measuring active changes in hemodynamic parameters occurring following treatment before macroscopic changes, such as decreases in tumor volume, become evident (131, 133, 134, 138, 178). DCEMRI measures the effect of the contrast agent on water relaxation rather than the effect on contrast agent itself. Hence, concentrations of contrast agent must be inferred from in vitro calibration to determine the relationship between the intrinsic T 1 relaxation rate and contrast agent concentration.

Further complex mathematical models are required to derive parameters relevant to microvascular hemodynamics. The evolution from timeactivity curves to vascular permeability or vascular volume fraction is modeldependent and is subject to assumptions that may not be appropriate for every application (179, 180). On a more macroscopic level, reproducibility/repeatability across image sites and centers further reduces confidence levels (3). Consequently, MRI measures of hemodynamics has not been validated by, e.g. comparison to an independent method.

We propose that validation of algorithms and assumptions can be achieved with an alternative metric of microvascular hemodynamics at high spatiotemporal resolution in an identical system. The purpose of the current work was the development of an MRI compatible dorsal midline skinfold window chamber which was used to deliver spatial and temporal contrast agent activity data from both DCEMRI and dynamic intravital fluorescence microscopy in the same preparation. Intravital fluorescence microscopy has been used extensively in the dorsal midline window chamber to estimate microvascular permeability (123, 181186). Thus, intravital microscopy is suitable for validating measured hemodynamic data against DCEMRI hemodynamic data. Ultimately, multimodal imaging will provide complimentary information to help elucidate the full spectrum of events occurring during tumor angiogenesis and, potentially, response to therapy. This research design will allow development of a more complete understanding of the mechanisms underlying changes in the tumor vasculature during tumor growth and in response to antiangiogenic therapies.

MATERIALS AND METHODS

Dorsal Midline Skinfold Window Chamber. Novel polyacetal resin (Delrin®) window chambers were built by the University Research Instrumentation Center at the University of Arizona and implanted into the dorsal midline skinfold of 89 week old male SCID mice. The chamber design was based on the commercial metal prototype used in a considerable number of studies (32, 187, 188). Mice were housed in disposable static microisolator cages (Innovive Inc., El Cajon, CA) and maintained under specific pathogenfree conditions. The mice were fed NIH31 irradiated pellets (Tekland

Premier, Madison, Wisconsin) and autoclaved water was freely available. All procedures were completed in accordance with the University of Arizona Institutional Animal Care and Use Committee (IACUC) protocol (Protocol # 04110, Approved August 19, 2004).

Subcutaneous injection 2.5mg/kg of analgesic buprenorphrine (Infusion Solutions,

Tucson, AZ) was given at least 1 hour prior to surgery and every 612 hours p.r.n. for at least 2 days. Animals were shaved and a depilatory cream was used to remove all the fur on the back. Animals were anesthetized with isoflourane (Phoenix Pharmaceuticals, Inc.,

St. Joseph, MO) and a povidone iodine topical solution (Cardinal Health, McGraw Park,

IL) was used to clean the surgical area. Mice were placed in a prone position and a plastic chamber was sutured in place and the skin from the left side of the dorsal midline skinfold was excised. The open tissue was bathed with 1X PBS (Mediatech, Inc.,

Herndon, VA) then covered with a sterile, 12 mm round cover slip (Fisher Scientific,

Houston, TX). Animals were given 1 cc of sterile saline subcutaneously 3 days prior to

96 and 7 days post surgery in order to keep both the animal and the chamber hydrated. Two days following the implantation of the window chamber, the GFPtransfected PC3 cells were transplanted into the sterile plastic window chambers by temporarily removing the glass coverslip and injecting the cells centrally into the chamber using a 28G monoject insulin syringe.

Cells.

Green fluorescent protein (GFP)transfected PC3 (human prostate cancer) cells were grown in IMDM (VWR, Phoenix, AZ) with 10% FBS (Omega Scientific, Tarzana, CA), and 0.5 mg/mL G418 Sulfate (Geneticin®, Molecular Probes, Eugene, OR) and maintained in 5% CO295% air humidified atmosphere at 37˚C. Subconfluent cells which were harvested by using 0.25% trypsinEDTA (HyClone, Logan, Utah) and were counted using the trypan blue assay technique. Cells (98% viability) were resuspended at the concentration of 3 x 10 6 cells/20 l of sterile saline.

Synthesis of biotinyl-BSA-GdDTPA 23 . Bovine serum albumin (~67 kDa) was labeled with either GadoliniumDiethylenetriaminepentaacetate (GdDTPA) for DCEMRI experimentation and additionally with AlexaFluor  633 (Molecular Probes, Eugene, OR) for our intravital fluorescence microscopy experimentation. This protein was also labeled with biotin for subsequent immunohistochemistry. The protocol for labeling these proteins is described in detail elsewhere (177, 189). Briefly, the protein dissolved in sodium bicarbonate was reacted in a molar ratio 1:9 with Nhydroxysuccinimidobiotin

(Sigma Aldrich, St. Louis, MO) dissolved in demethylformamide. DTPAA (Sigma

Aldrich, St. Louis, MO) suspended in DMSO was added to the solution followed by mixing with gadolinium (III) chloride dissolved (Sigma Aldrich, St. Louis, MO) in sodium acetate buffer. The lyophilized biotinylBSAGdDTPA product was stored at 4˚C and reconstituted in PBS when needed. For fluorescence labeling, AlexaFluor  633 was conjugated to the biotinylBSAGdDTPA at pH 8.3, followed by size exclusion chromatography according to manufacturer instructions (A20005, Invitrogen, Carlsbad,

CA ).

In Vitro Relaxivity Measurements. A range (0 – 2 x 10 4 M) of biotinBSAGdDTPA concentrations were suspended in PBS, loaded into microcentrifuge tubes and maintained at 37˚C using an airflow heater. Sample temperatures were monitored using a fluoroptic thermometer. Longitudinal relaxation ( T1) times were measured from contiguous 2DSE images acquired on a 40 cm horizontal bore Bruker Biospec Avance® 4.7 Tesla imaging spectrometer (Bruker, Karlsuhe, Germany) at 20 repetition times (TR) ranging from 40

8000 ms (TE = 9.6 ms, NEX = 2, FOV = 2.56 cm x 2.56 cm, slice thickness = 2 mm, matrix 128 x 128; Figure 2.1a). The concentration of biotinBSAGdDTPA is expected to be linearly proportional to the relaxation rate ( 1/T 1) at low concentrations, and thus the longitudinal water proton relaxivity r1 of biotinBSAGdDTPA can be calculated by fitting those parameters to the following equation:

1 1 = + r1 ⋅[biotin − BSA − GdDTPA], Equation 63 T1 T10

98 where 1/T 10 is the relaxation rate in the absence of contrast agent. The r1 relaxivity of biotinBSAGdDTPA was calculated to be 8.5 mM 1 s1 (expressed in terms of

Gadolinium), which was very similar to that reported in the literature (expressed in terms of BSA) (189). It should be noted here that this relaxivity is used later in vivo to estimate concentration, and there are uncertainties regarding the in vivo relaxivity of GdDTPA due to 1) the influence of the in vivo microenvironment on the intrinsic Gadolinium relaxivities and; 2) the nonlinear relationship between the relaxation rate (1/T 1) and concentration at high Gadolinium concentrations.

To find the transverse water proton relaxivity r2 of biotinBSAGdDTPA, images were acquired at 10 different echo times (TE) ranging from 9150 ms (TR = 5000 ms, FOV =

2.56 cm x 2.56 cm, slice thickness = 2 mm, matrix 128 x 128; Figure 2.2a) and transverse relaxation ( T2) times were plotted against concentration. The r2 relaxivity was derived from the transverse relaxation rates in the presence and absence of contrast agent and the concentration:

1 1 = + r2 ⋅[biotin − BSA − GdDTPA], Equation 64 T2 T20 1 1 The r2 relaxivity of biotinBSAGdDTPA was calculated to be 25.1 mM s (expressed in terms of Gadolinium).

In Vivo MRI Acquisition. All MRI experiments were performed on a 4.7 Tesla imaging spectrometer (Bruker, Karlsuhe, Germany) equipped with an actively shielded gradient

99 coil capable of 150 mT/m. All animals were anesthetized with 1.5% Isoflurane delivered in O 2, at 1.0 LPM, and temperatures were continuously monitored using a rectal

Luxtron® fluoroptic thermometer. An external heater was used to maintain core body temperature between 3637 °C during the course of the imaging experiments. A series of

2Dspin echo (SE) images at eight different repetition time (TR) values were acquired prior to injection of the contrast agent (TR = 100, 200, 500, 800, 1000, 2000, 3000, 5000 ms, TE = 9.6 ms, NEX = 1, FOV = 2.56 mm x 2.56 mm, slice thickness = 5 mm, matrix

= 256 x 256, inplane resolution = 100 m) in order to find the preinjection T 1 for all pixels (T 1pre ). The postcontrast, dynamic series of T 1weighted images were collected before and up to 30 minutes following bolus administration (200 L of 0.01 mmol/kg) of the contrast agent (TR = 100 ms, TE = 9.6 ms, NEX = 2, FOV = 2.56 cm x 2.56 cm, slice thickness = 5 mm, matrix 256 x 256, inplane resolution = 100 m).

MRI Data Processing. DCEMRI analyses were conducted based on assumption of infinitelyfast transcytolemmal exchange kinetics (146). For tissue pixels, the precontrast

2DSE images acquired at 8 different TR values were used to calculate pixelbypixel precontrast 1/T 1 maps using:

−TR /T1 S = S0 (1− e ), Equation 65 where S0 reflects the spin density of the excited slice and S is the measured signal intensity as a function of TR.

100

In the blood, however, T2 relaxation has a significant effect on signal intensity, and this effect is dependent on blood oxygenation (190). Furthermore, T2 effects from the contrast agent may be significant at concentrations above 0.05 mM, well below the concentration of the issued dose in this work, 0.01 mmol/kg (≈ 0.15 mM). The tracer kinetic models used in this work require access the arterial input function, which is acquired from a region of interest (ROI) drawn over an available large vessel. In this region, T2 effects were modeled as follows:

−TR / T1 −TE /T 2 S = S0 (1− e )⋅(e ), Equation 66 where an endogenous T2 in the blood of 0.10 s was estimated based on temperature and oxygenation saturation status (190).

Since tissue pixels require only the incorporation of T1 relaxation effects, the post contrast 1/T 1 values and r1 relaxivity are used to calculate concentration simply using

Equation 1. However, in the reference vessel ROI, T2 effects must still be modeled, and concentration is iteratively solved using the following equation:

 1   1  −TR +r1[biotin−BSA−GdDTPA] −TE +r2 [biotin−BSA−GdDTPA]  T1pre   T 2 pre  S post = S pre 1( − e ) ⋅ (e ) Eq. 67 where T 1pre is the preinjection T 1 (see In Vivo MRI Acquisition ) and T 2pre is the pre injection T 2 relaxation time.

Quantification of Hemodynamic Parameters from DCE-MRI Data. Dynamic contrast enhanced imaging enables the acquisition of a signal enhancement versus time curve, and

101 quantitative evaluation of that curve provides estimation of the extravascular extracellular

trans fractional volume ( ve) and the volume transfer constant ( K ) between the blood plasma and the ve (140). Changes in these parameters are of great importance in terms of establishing reliable, sensitive and quantitative biomarkers of response to therapy.

There are two principle regimes that reflect capillary permeability. In a situation where the intrinsic permeability of the capillary is high, then transport of a solute across the capillary is dependent on blood flow rather than diffusion, and as such is flowlimited

(140):

dC dC p t = v + Fρ 1( − Hct)(C − C ) . Equation 68 dt p dt p e

The rate of uptake of the contrast agent into the tissue is related to the ve, the contrast agent concentrations in the tissue ( Ct) and the arterial blood plasma ( Cp). Importantly, the transfer constant, Ktrans , is a function of the perfusion of whole blood per unit mass of tissue ( F, ml g 1 min 1), the density of tissue ( ρ, g ml 1) and the hematocrit ( Hct )

(i.e.K trans = Fρ 1( − Hct ) ).

In situations where capillary permeability to the contrast agent is very low, the Ktrans is a function of the permeability surface area product (PS , ml min 1 g1) of the capillary wall

trans and ρ (i.e. K = PSρ ), and the rate of uptake is related to the ve, Ct and Cp:

dC dC p t = v + PSρ(C − C ) . Equation 69 dt p dt p e In a theoretically perfect flowlimited regime, the arterial input function matches the rate of tracer accumulation in the tissue. In a theoretically perfect permeabilitysurface area

102 productlimited region, the arterial input function is a step function, and the rate of tracer accumulation in the tissue is linear in the early stages of uptake when the extracellular concentration, Ce, is approximately equal to zero. Equation 70 then simplifies to:

dC dC p t = v + K transC Equation 70 dt p dt p

Solving for dC t/dt with initial conditions (C p = C t = 0 at t = 0) allows for simplification down to a first order polynomial equation:

trans Ct (t) = v p C p + K C p t , Equation 71 where the slope of the tissue tracer accumulation rate is related to permeability ( Ktrans ), and the intercept is related to the fractional vascular volume ( vp). For more details on this derivation, please see Appendix A.

Due to these assumptions, the time period immediately following the bolus delivery and distribution of the contrast agent in the circulation is ignored (i.e. the first 2.5 minutes) and only the steadystate equilibrium of the contrast agent in the circulation is modeled

(i.e. 2.528.5 minutes).

Fluorescence Microscopy. Fluorescence microscopy was performed on an upright research microscope (Eclipse E600, Nikon, Melville, NY) using Nikon’s 1X objective and a high pressure Hg lamp for GFP excitation. For excitation at 633 nm of the fluorophore (Alexa Fluor 633), an onboard Red HeNe laser and 640 bandpass filter were used, and for excitation of the GFPtransfected cells at 480 nm, an Argon laser and

103

515/30 bandpass filter were used. Data were captured with a 12bit CCD camera from

Optronics (MicroFire Megapixel) using the native application software. Four averages at

1.92 s pixel dwell were obtained for background fluorescence, then a 200 L bolus injection of fluorescently labeled contrast agent was administered over 80 frames as fast as possible up to 5 minutes for washin kinetics followed by a postwash acquisition at a frame rate matching the MRacquisition times (i.e. 38 s/frame).

Quantification of Hemodynamic Parameters from Fluorescence Microscopy Data.

The relative concentration of the injected contrast agent in the vascular space is directly proportional to the signal intensity obtained from fluorescence microscopy data and can be measured directly (123). Eight standard concentrations of the fluorophore in 1X PBS were prepared and captured in capillary tubes of a known inner diameter. The average signal intensity for each concentration was recorded at three gains (typically between a gain of 75105), with one gain identical to the gain used in vivo . The signal intensity versus concentration at the gain used in vivo was used to find the proportionality constant

(the slope of the linear part of this nonlinear (at high concentration) relationship) needed to convert raw signal intensity to concentration in every pixel for the raw signal activity

trans curves. Once concentration activity was determined, vp and K maps were generated using the same models described above.

Image Processing. All image processing was performed using developed routines in

Interactive Data Language (IDL, Boulder, CO). Vascular volume and permeability maps

104 derived from DCEMRI and fluorescence microscopy were coregistered in order to provide more accurate estimates of hemodynamic parameters in the tumor. More specifically, the fluorescence microscopyderived vascular volume and permeability maps were warped based on the irregular grid of the DCEMRIderived maps as the basis for triangulation. Covariance analyses on the coregistered maps generally resulted in correlation coefficients between 0.75 and 0.90.

RESULTS

The longitudinal relaxivity was determined by imaging serial dilutions of the contrast agent at multiple repetition times (Figure 2.1a), using the magnetization recovery information to calculate the T 1 for each concentration (Figure 2.1b) and plotting those concentrations against the corresponding calculated T 1 value (Figure 2.1c). The r1 relaxivity of the contrast agent was calculated to be 7.5 mM 1 s1. The transverse relaxivity was determined by imaging serial dilutions of the contrast agent at multiple echo times (Figure 2.2a), using the magnetization decay information to calculate the T 2 for each concentration (Figure 2.2b) and plotting those concentrations against the corresponding calculated T 2 value (Figure 2.2c). The r2 relaxivity of the contrast agent

1 1 was calculated to be 22.8 mM s .

Polyacetal resin window chambers were recapitulated from the original titanium design

(Figure 2.3). Mice were less encumbered when bearing the polyacetal resin chambers compared to metal window chambers (0.95 g versus 2.5 g, respectively) (Figure 2.4).

105

In all cases, transillumination images of the window chamber were acquired to obtain a highresolution, brightfield image of the vascular network encasing the tumor. These transillumination images were overlaid with GFP fluorescence of the tumor at 10X

(Figure 2.5a) and 20X (Figure 2.5b).

Representative examples of the dynamic MR scans demonstrate the enhancement of the tumor region in the upper left quadrant of the chamber over time (Figures 2.6ac). This enhancement is more clearly demonstrated by a subtraction image displayed in Figure

2.6d. Dynamic scans were acquired up to 30 minutes following bolus administration of the triplylabeled albumin contrast agent (200 L of a 0.01 mmol/kg dose). An example of a timeactivity curve from an ROI placed over a major blood vessel is shown in Figure

2.7. Representative examples of the dynamic fluorescence microscopy scans demonstrate enhancement in an identical field of view to the DCEMRI data (Figures 2.8ac). An example of a timeactivity curve from an ROI placed over the same major vessel found in the MR image of the same FOV is shown in Figure 2.9. The vessel input function here is significantly less confounded in the noise than the MR vessel input function (Figure 2.7).

In planeresolution is 24.8 m compared to 100 m inplane resolution on the dynamic

MR scans.

Time activity curves for DCEMRI and fluorescence microscopy data were analyzed by

2compartment models to generate pixelbypixel maps of vp, as described in the Methods

106

& Materials section (Figures 2.10 a and b, respectively), and those maps were registered using a canned warping routine in IDL. Pixelbypixel covariance analysis of the co registration delivered correlation coefficients between 0.750.90, and qualitatively, the coregistration was acceptable (Figure 2.10c). Similarly, analysis of the timeactivity curves for DCEMRI and fluorescence microscopy data were used to generate pixelby pixel maps of Ktrans (Figures 2.11a and b, respectively), and coregistration of the two

Ktrans maps delivered correlation coefficients between 0.750.90. The low permeability associated with the vessels within the chamber is consistent with a normal vessel selectivity to macromolecules that have long intravascular retention times (e.g. macromolecules with a MW>~20kDa), and high permeability to the macromolecular contrast agent was observed in the tumorassociated vessels.

Figure 2.12 displays the K trans distribution for the DCEMRI data (blue) and dynamic fluorescence microscopy data (black) for the tumor ROI (a) and the nontumor ROI (i.e. an ROI that covered the entire chamber excluding the tumor) (b). Another approach was taken to represent of the histogram data in order to more clearly demonstrate difference between two populations (e.g., a comparison between the K trans distribution from DCE

MRI data and the K trans distribution from FMdata). In a given bin from a histogram, the value for that bin was normalized to the total number of pixels in the ROI and systematically subtracted from the pixel count in the bin before it. Since the first bin has no prior bin pixel count, the value would assume 100%. Effectively, this results in a percentage of pixels for each bin that is greater than the next. As the number of pixels in

107

trans each bin (beginning at the first bin of zero) increases with increasing K or v p, the percent above that value decreases. Figures 2.13a and b represent this form of analysis for the K trans pixel distribution in the tumor and nontumor ROIs, respectively. Figure 2.14 displays the v p distribution for the DCEMRI data (blue) and dynamic fluorescence microscopy data (black) for the tumor ROI (a) and the nontumor ROI (b). Figures 2.15a and b represent the percent above threshold distributions for v p for the tumor and non tumor ROIs, respectively.

Table 2.1 provides the individual values for each subject for both imaging systems in the tumor ROI and the nontumor ROI, and Figure 2.16 is a graphical representation of the table.

DISCUSSION

The purpose of this work was to establish a dualimaging system whereby an alternative metric of microvascular permeability, dynamic fluorescence microscopy, could be used to validate data derived from DCEMR imaging. Quantitative results from both imaging systems were achievable, and with standard coregistration techniques, those results could be compared and used to evaluate overestimations or underestimations of microvascular permeability as quantified by DCEMRI.

Since DCEMRI contrast agent concentrations are derived through indirect methods, and permeability is calculated based on these concentrations, it is expected that such distance

108 between signal intensity and calculated concentration may introduce undisclosed variability in the data. This variability was expected to be elucidated with the addition of an alternative metric of permeability, dynamic fluorescence microscopy, an imaging system that provides a directly proportional relationship between signal intensities and concentration.

The data presented in this chapter offer a sufficient spatiotemporal resolution to cross correlate the rate and extent of extravasation of macromolecular contrast agents using both MRI and optical imaging modalities. As shown in Table 2.1, these data are correlated, suggesting that the assumptions made for analysis of DCEMRI data, as enumerated in the Methods & Materials, are appropriate for this preparation. A future goal outside of this dissertation is to collect data from both systems simultaneously. Such dualmodality imaging will improve quantitation of MR signal because injection volumes, concentrations and timing would be identical, eliminating additional complexity in the data collection and analysis.

Currently available clinical contrast agents are small molecular weight compounds, and have high pretherapy variance compared to larger molecular weight contrast agents.

Once optimized, this multimodal imaging system may be useful in evaluating contrast agents of different molecular weights to identify sizes that produce the best reproducibility and highest sensitivity/specificity to changes in the tumor microvasculature following antiangiogenic therapy. Optical imaging is particularly

109 appropriate for such preclinical screening. Optical imaging provides a solution in applications where comparisons (at higher sensitivity, specificity and resolution) against clinically available diagnostics are required, and there are a plethora of optical probes already available that can aid in the development and validation of MRactive probes.

Furthermore, the intrinsic optical properties of tissue offer additional noninvasive imaging opportunities (e.g., imaging collagen dynamics during angiogenesis using secondharmonic generation).

Future applications of this system include pH imaging of the tumor microenvironment

(191), interrogation of ECM/MMP interactions (i.e. enzymeactivatable probes directly report enzyme activity, and optical imaging can be used to localize and identify changes to the environment simultaneously with MRI, opening the door for development of MR sensitive techniques for probing MMP activity) (192), or imaging apoptosis with optical probes for apoptosis (e.g. FLIVO) to validate the ability of diffusionweight MRI to differentiate between apoptosis and necrosis/mitotic catastrophe (193).

110

FIGURES

Figure 2.1 a Crosssectional view of 6 tubes containing a serial dilution of the biotin BSAGdDTPA. The series of images were acquired at a constant echo time (TE = 9.6 ms) and multiple repetition times (TR = 10 s, 5 s, 3 s, 1 s, 0.8 s, 0.5 s, 0.3 ms, 0.1 s). b. c. 2e+7 3.0

r2 = 0.9998 2.5 r = 8.5 mM 1 s1 2e+7 1 2.0 ) 1 1.5

1e+7 [ 0.29 ] (s

[ 0.14 ] 1 [ 0.07 ]

[ 0.04 ] 1/T 1.0 5e+6 [ 0.02 ] Signal Signal Intensity [ 0.01 ] 0.5

0 0.0 0 2 4 6 8 10 0.00 0.05 0.10 0.15 0.20 0.25 0.30 TR (msec) Concentration (mM)

Figure 2.1 b-c Corresponding recovery curves for each concentration (displayed in brackets in terms of mM) at multiple repetition times (b). Calculated T 1 plotted against corresponding concentration of biotinBSAGdDTPA (c).

111

Figure 2.2 a Crosssectional view of 6 tubes containing a serial dilution of the biotin BSAGdDTPA. The series of images were acquired at a constant repetition time (TR = 5000 ms) and multiple echo times (TE = 9.6 ms, 20 ms, 30 ms, 40 ms, 50 ms, 60 ms, 80 ms, 120 ms, 200 ms, 300 ms, 500 ms).

b. c. 16 2e+7 r2 = 0.9969 [ 0.29 ] 14 1 1 r1 = 25.1 mM s [ 0.14 ] 2e+7 [ 0.07 ] ) [ 0.04 ] 12 1 [ 0.02 ] 1e+7 (s [ 0.01 ] 2 10 1/T 5e+6 Signal Intensity Signal 8

0 6 0.0 0.1 0.2 0.3 0.4 0.5 0.00 0.05 0.10 0.15 0.20 0.25 0.30 TE (msec) Concentration (mM)

Figure 2.2 b-c Corresponding decay curves for each concentration (displayed in brackets in terms of mM) at multiple echo times (b). Calculated T 2 plotted against corresponding concentration of biotinBSAGdDTPA (c).

Figure 2.3 Polyacetal resin (a.k.a. Delrin) replicas of the original titanium design were created to improve tumor growth in mice. Animals were less encumbered by the plastic chambers than the metal cohorts (0.95g—polyactal resin; 2.5g—titanium) and in general, the plastic chambers reduced morbidity of the preparation.

112

Figure 2.4 Plastic window chamber implanted on animal.

a b

Figure 2.5 a-b Transillumination images overlaid with GFP fluorescence at 10X (a) and 20X. Tumor (upper left quadrant) is highly vascularized and exhibits areas of hyperpermeability.

a b c d

113

Figure 2.6 a-d Representative DCEMRI scans of window chamber over time (tumor region located in the upper left quadrant of the window) at 0 (a), 5 (b), and 30 minutes (c). A subtraction image shows pixels that enhanced from the baseline or the 0 minute image (d).

0.012

0.010

0.008

0.006

0.004

0.002 [biotinBSAGdDTPA] (mM) [biotinBSAGdDTPA]

0.000 0 5 10 15 20 25 30 Time (min)

Figure 2.7 Example of the timeactivity curve obtained from an ROI (redfilled object) drawn over a major blood vessel throughout the DCEMR image series. This vessel input function was a characteristic stepfunction for the biotinBSAGdDTPA contrast agent.

a b c

Figure 2.8 a-c Representative dynamic fluorescence microscopy scans of window chamber over time (tumor region located in the upper left quadrant of the window) at 0 (a), 2 (b), and 30 minutes (c).

114

0.0025

0.0020

0.0015

0.0010

0.0005 [biotinBSAGdDTPA] (mM)

0.0000 0 5 10 15 20 25 30 35 Time (min)

Figure 2.9 Example of the timeactivity curve obtained from an ROI drawn over a major blood vessel throughout the dynamic fluorescence microscopy image series. a 100 b 100 c

80 80

60 60

(%) (%) p p

v 40 v 40

20 20

0 0

Figure 2.10 a-c Dynamic fluorescence microscopyderived (a) and DCEMRIderived

(b) v p maps obtained from concentrationtime activity curves on a pixelbypixel basis.

Warped fluorescence microscope v p maps were blended with (or overlaid on) the reference DCEMRI v p map (c). Coregistration of the two maps ensured that ROIs drawn over the tumor and the nontumor space would result in inclusion of near completely registered pixels in those designated areas (correlation coefficients for registration ranged between 0.75 and 0.90).

115

a 0.038 b 0.038 c

0.031 0.031

) ) 1 1 - - 0.023 0.023

(min (min 0.015 0.015 trans trans K K 0.008 0.008

0.000 0.000

Figure 2.11 a-c Dynamic fluorescence microscopyderived (a) and DCEMRIderived (b) Ktrans maps obtained from concentrationtime activity curves on a pixelbypixel basis. Warped fluorescence microscope K trans maps were blended with (or overlaid upon) the reference DCEMRI derived K trans map (c). Coregistration of the two maps ensured that ROIs drawn over the tumor and the nontumor space would result in inclusion of near completely registered pixels in those designated areas (correlation coefficients for registration ranged between 0.75 and 0.90).

a trans 1 b trans 1 70 K (min ) 600 K (min ) Tumor Non Tumor 60 500 50 400 40 FM Mean: 0.00223 FM Mean: 0.00158 DCEMRI Mean: 0.00394 300 DCEMRI Mean: 0.00342 30 No. Pixels No. Pixels 200 20

10 100

0 0 0.000 0.005 0.010 0.015 0.020 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 )

Figure 2.12 a-b Histograms of K trans derived from DCEMRI data (blue) and dynamic FM data (black) for both the tumor ROI (a) and the nontumor ROI (b) (different scales were used for each ROI to maintain the integrity of an adequate comparison).

116

a b trans 1 trans 1 120 K (min ) 120 K (min ) Tumor Non Tumor 100 100

80 80 FM Mean: 0.00223 FM Mean: 0.00158 60 DCEMRI Mean: 0.00394 60 DCEMRI Mean: 0.00342 % Pixels % 40 Pixels % 40

20 20

0 0 0.000 0.005 0.010 0.015 0.020 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 )

Figure 2.13 a-b Distributions for DCEMRI data (blue circles) and dynamic fluorescence microscopy data (black circles) were created by normalizing histograms to percentage of pixels above the respective K trans value. Distributions are shown for the tumor (a) and nontumor (b) ROIs.

a Vascular Volume Fraction (%) b Vascular Volume Fraction (%) Tumor Non Tumor 160 1400

140 1200 120 1000 100 FM Mean: 19.96 FM Mean: 18.89 DCEMRI Mean: 17.11 800 DCEMRI Mean: 22.00 80 600 60 No. No. Pixels No. Pixels No. 40 400 20 200

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

117

Figure 2.14 a-b Histograms of v p derived from DCEMRI data (blue) and dynamic FM data (black) for both the tumor ROI (a) and the nontumor ROI (b) (different scales were used for each ROI to maintain the integrity of an adequate comparison).

Vascular Volume Fraction (%) a b Vascular Volume Fraction (%) Tumor Non Tumor 120 120

100 100

80 FM Mean: 19.96 80 FM Mean: 18.89 DCEMRI Mean: 17.11 DCEMRI Mean: 22.00 60 60 No. No. Pixels 40 No. Pixels 40

20 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

Figure 2.15 a-b Distributions for DCEMRI data (blue circles) and dynamic fluorescence microscopy data (black circles) were created by normalizing histograms to percentage of pixels above the respective vp value. Distributions are shown for the tumor (a) and non tumor (b) ROIs.

118

Imaging Tumor NonTumor System v v (%) ktrans (min 1 ) v v (%) ktrans (min 1 ) MR 17.11 0.003944 22.00 0.003423 subject 1 FM 19.96 0.002231 18.89 0.001577 MR 7.04 0.003140 9.77 0.003969 subject 2 FM 14.27 0.003703 20.23 0.006228 MR 7.64 0.001175 12.35 0.001747 subject 3 FM 11.65 0.002470 13.44 0.003067 MR 22.90 0.001753 16.35 0.003350 subject 4 FM 32.40 0.001504 26.85 0.003484 MR 8.31 0.003855 10.87 0.003023 subject 5 FM 18.29 0.002217 15.06 0.001463 MR 12.24 0.002179 6.33 0.001492 subject 6 FM n/a n/a n/a n/a MR 12.14 0.003559 15.77 0.006662 subject 7 FM 13.34 0.002233 15.08 0.002368 MR 19.68 0.004635 20.93 0.004905 subject 8 FM 11.67 0.002250 12.50 0.002406 MR 15.52 0.001202 15.66 0.001019 subject 9 FM 12.03 0.001575 13.53 0.001497 MR 11.76 0.001659 14.92 0.001867 subject 10 FM 10.22 0.001521 16.79 0.002185

trans Table 2.1 Individual vp and K values from both DCEMRI and dynamic FM dynamic data in the tumor and nontumor ROIs. For one data set, the fluorescence microscopy data were not available, but DCEMRI data were still included and standard deviation and standard error calculations (shown in Figure 2.16) reflected the omission of the fluorescence microscopy data.

119 a b trans 1 Vascular Volume Fraction (%) K (min )

Tumor Tumor 20 NonTumor 0.004 NonTumor )

15 1 0.003 (%)

10 (min 0.002 p v trans 0.001 5 K

0 0.000 MRI FM MRI FM

Figure 2.16 a-b Graphical representation of table 2.1. Means of individual values for v p and K trans for both regions (tumor and nontumor) are expressed coincident with standard errors for 10 animals measured using DCEMRI and 9 animals using dynamic fluorescence microscopy.

120

CHAPTER 3

Combined Dynamic Contrast Enhanced Magnetic Resonance Imaging and Fluorescence

Microscopy of Paramagnetically Labeled Proteins of Different Molecular Weights

INTRODUCTION

The aim of this research project was to create a biotinylated, paramagnetic globular protein at a lower molecular weight than the biotinBSAGdDTPA. Since large molecular weight contrast agents are not completely characterized and in some cases immunogenic, they are relegated to the preclinical evaluation of experimental tumors. Small molecular weight agents such as GdDTPA have been used extensively in the clinic and provide meaningful adjuvant information to existing diagnostic protocols. However, studies have shown the considerable improvement in resolution between pre and posttherapy tumor hemodynamics with larger molecular weight contrast agents (e.g. > 17 kDa) (194199).

This study investigated the rate and extent of extravasation of two MR contrast media of different molecular weights using the previously described multimodal imaging system.

Ovalbumin was biotinylated and subsequently conjugated to GadoliniumDTPA using an identical synthesis protocol to the biotinBSAGdDTPA. The resulting molecular weight of the biotinOvalbuminGdDTPA (~47 kDa) was slightly more than half that of the biotinBSAGdDTPA (~80 kDa). It was anticipated that the signal to noise ratio (SNR) and contrast to noise ratio (CNR) between images acquired using the two proteins would

121 be comparable and that results would provide some insight as to the relationship between

trans molecular weight and K and v p.

METHODS & MATERIALS

Synthesis of biotinyl-Ovalbumin-GdDTPA 7. Ovalbumin (from chicken egg, Sigma

Aldrich, ~44 kDa) was also labeled with either GdDTPA for our DCEMRI experimentation and additionally with AlexaFluor  633 for our intravital fluorescence microscopy experimentation and biotin for immunohistochemistry. The protocol for labeling Ovalbumin was identical to that for labeling BSA. Briefly, a cold solution of N hydroxysuccinimidobiotin (Sigma Aldrich, St. Louis, MO) in dry DMF was added to

Ovalbumin in aqueous NaHCO 3 solution at 0 °C. The reaction mixture was stirred at 0 °C for 1h, at room temperature for 2h, then loaded into dialysis bags and dialyzed against

DDW at 4°C for 5 days. A suspension of DTPA anhydride (Sigma Aldrich, St. Louis,

MO) in dry DMSO was added to the biotinylOvalbumin solution adjusted to 0.1 M in

Hepes buffer at a constant pH of 8.5. The solution was then stirred at 4 °C for 2h and dialyzed against citrate buffer at 4°C for 5 days. GdCl 36H 2O (Sigma Aldrich, St. Louis,

MO) dissolved in sodium acetate buffer was slowly added to the biotinylOvalbumin

DTPA in citrate buffer. The solution was then stirred at 4 °C for 24h and dialyzed at 4°C against citrate buffer for 5 days. The final lyophilized biotinylOvalbuminGdDTPA product was stored at 4˚C and reconstituted in PBS when needed. A difference between the BSA and Ovalbumin of paramount importance was the number of available surface lysine groups on Ovalbumin (89 available lysine groups on BSA compared to 44.4 on

122

Ovalbumin) (200). Thus, in order to avoid saturation of available lysine groups on

Ovalbumin, Nhydroxysuccinimidobiotin, DTPAA and gadolinium (III) chloride were halved from ratio used for labeling BSA. For fluorescence labeling, AlexaFluor  633 was conjugated to the biotinylOvalbuminGdDTPA at pH 8.3, followed by size exclusion chromatography. These steps were calibrated to a protocol that provided a known amount of Gadolinium or fluorophore per molecule.

In Vitro Relaxivity Measurements. A range (0 – 1.3 x 10 3 M) of biotinOvalbumin

GdDTPA concentrations were suspended in PBS, loaded into microcentrifuge tubes and maintained at 37˚C using an airflow heater. Sample temperatures were monitored using a fluoroptic thermometer. Longitudinal relaxation ( T1) times were measured from contiguous 2DSE images acquired on a 40 cm horizontal bore Bruker Biospec Avance®

4.7 Tesla imaging spectrometer (Bruker, Karlsuhe, Germany) at 20 repetition times (TR) ranging from 0.1015 s (TE = 9.6 ms, NEX = 2, FOV = 2.56 cm x 2.56 cm, slice thickness = 5 mm, matrix 256 x 256, inplane resolution = 100 m). The concentration of biotinOvalbuminGdDTPA is expected to be linearly proportional to the relaxation rate

(1/T 1) at low concentrations of Gadolinium, and thus the longitudinal water proton relaxivity r1 of biotinOvalbuminGdDTPA and can be calculated by fitting those parameters to the following equation:

1 1 = + r1 ⋅[biotin − Ovalbumin − GdDTPA], Equation 72 T1 T10

123 where 1/T 10 is the relaxation rate in the absence of contrast agent. The r1 relaxivity of biotinOvalbuminGdDTPA (in terms of Gadolinium) was calculated to be 8.06 mM 1 s 1.

To find the transverse water proton relaxivity r2 of biotinOvalbuminGdDTPA, 10 echo times (TE) ranging from 10 1000 ms (TR = 15000 ms, FOV = 2.56 cm x 2.56 cm, slice thickness = 5 mm, matrix 256 x 256, inplane resolution = 100 m) were acquired and transverse relaxation ( T2) times were plotted against concentration. The r 2 relaxivity was derived from the transverse relaxation rates in the presence and absence of contrast agent and the concentration:

1 1 = + r2 ⋅[biotin − Ovalbumin − GdDTPA], Equation 73 T2 T20

The r2 relaxivity of biotinOvalbuminGdDTPA (in terms of Gadolinium) was calculated to be 39.8 mM 1 s1.

In Vivo MRI Acquisition. Details on the in vivo experiments were detailed elsewhere

(please see Chapter 1, Methods & Materials). Briefly, a series of 2Dspin echo (SE) images at eight different repetition time (TR) values were acquired prior to injection of the contrast agent (TR = 100, 200, 500, 800, 1000, 2000, 3000, 5000 ms, TE = 9.6 ms,

NEX = 1, FOV = 2.56 mm x 2.56 mm, slice thickness = 5 mm, matrix = 256 x 256, in plane resolution = 100 m) in order to find the preinjection T 1 for all pixels (T 1pre ). The postcontrast, dynamic series of T 1weighted images were collected before and up to 30 minutes following bolus administration (200 L of 0.03 mmol/kg) of the biotin

124

OvalbuminGdDTPA (TR = 100 ms, TE = 9.6 ms, NEX = 2, FOV = 2.56 cm x 2.56 cm, slice thickness = 5 mm, matrix 256 x 256, inplane resolution = 100 m).

MRI Data Processing. Details on the MRI data processing were detailed elsewhere

(please see Chapter 2, Methods & Materials), however, it is important to reiterate that for tissue pixels, the precontrast T 1 map was determined using:

−TR / T1 S = S0 (1− e ), Equation 74 whereas the T1 in blood pixels (i.e., in an vessel input function ROI) was modeled to account for T2 relaxation effects, which may have a significant effect on the estimated T1 in each pixel, which was ultimately used to determine concentration:

−TR / T1 −TE /T 2 S = S0 (1− e )⋅(e ), Equation 75 where a literaturebased T2 in the blood of 0.10 s was estimated based on temperature and oxygenation saturation status at 37ºC and 4.7T (190).

Following injection of the contrast agent, the T1 relaxation rate in tissue pixels post contrast and was modeled using Equation 72, and concentration could then subsequently be calculated using the relationship between the T1 relaxation rate pre and postcontrast and the in vitro r1 relaxivity:

1 1 = + r1⋅[biotin − BSA / OvA − GdDTPA] . Equation 76 T1 post T1pre

125

However, to estimate concentration in the blood pixels (i.e., the vessel input function

ROI), it was necessary to include effects from T 2 relaxation in the blood and the r2 relaxivity of biotinOvalbuminGdDTPA:

 1   1  −TR +r1[biotin−BSA−GdDTPA] −TE +r2 [biotin−BSA−GdDTPA]  T1pre   T 2 pre  S post = S pre 1( − e ) ⋅ (e ), Eq. 77 where T 1pre is the preinjection T 1 (please see In Vivo MRI Acquisition ) and T 2pre is the preinjection T 2 relaxation time. Since this type of analysis is done only in a blood vessel

ROI, a T 2 value for whole blood was obtained from the literature (190).

Quantification of Hemodynamic Parameters from DCE-MRI Data. The details of this section were discussed elsewhere (see Chapter 2, Methods & Materials). However, it is important to note that the kinetics of the biotinBSAGdDTPA generally resulted in flat or step vessel input function, allowing for simplification of the standard 2compartment,

3parameter tracer kinetic model to a 2parameter modification occasionally referred to as the SlopeIntercept (SI) Method. The pharmacokinetics of the biotinOvalbumin

GdDTPA contrast agent, however, precluded the use of the simplified model by virtue of the gammavariate vessel input function. Nevertheless, for either contrast agent, tracer kinetic models provided for estimation of parameters related to permeability (K trans ) and the vascular volume fraction (v p).

Quantification of Hemodynamic Parameters from Fluorescence Microscopy Data.

The details of this section were discussed extensively elsewhere (please see Chapter 2,

Methods & Materials). It is important to note here that concentration of biotin

126

OvalbuminGdDTPA was derived from capillarytube phantoms containing the agent diluted in PBS which were imaged after every session to account for variations in laser output energy, and the same tracer kinetic models used for the analysis of DCEMRI were used for the analysis of the corresponding fluorescence microscopy study.

RESULTS

The longitudinal relaxivity was determined by imaging serial dilutions of the biotin

OvalbuminGdDTPA at multiple repetition times (Figure 3.1a), using the magnetization recovery information to calculate the T 1 for each concentration (Figure 3.1b) and plotting those concentrations against the corresponding calculated T 1 value (Figure 3.1c). The r1 relaxivity of the biotinOvalbuminGdDTPA (in terms of Gadolinium) was calculated to be 8.1 mM 1 s1. The transverse relaxivity was determined by the same phantom at multiple echo times (Figure 3.2a), using the magnetization decay information to calculate the T 2 for each concentration (Figure 3.2b) and plotting those concentrations against the corresponding calculated T 2 value (Figure 3.2c). The r2 relaxivity of biotinOvalbumin

1 1 GdDTPA (in terms of Gadolinium) was calculated to be 39.8 mM s .

A comparison of characteristics between the biotinBSAGdDTPA and biotin

OvalbuminGdDTPA has been tabulated (Table 3.1). With only 44 surface lysines available to conjugate Ovalbumin with DTPA, and thus Gadolinium, compared to 89 available lysines on BSA, a 2% labeling of the Gadolinium atom to Ovalbumin

127

(compared to 4% for BSA) was expected. This translated to almost 3fold fewer

Gadolinium atoms per molecule.

Figure 3.3 is a representative transillumination image of the polyacetal window chamber overlaid with a fluorescence image of the GFPtransfected PC3 prostate carcinoma cells at 10X (a) and 20X (b). Figure 3.4a is an average of the 10 final frames from a DCEMRI series of images following bolus injection of the biotinOvalbuminGdDTPA. Figure 3.4b is the corresponding vessel input function taken from a large vessel near the tumor. In order to demonstrate the difference in SNR and CNR between the biotinBSAGdDTPA

(Figure 3.5a) and biotinOvalbuminGdDTPA (Figure 3.5b), the SNR was calculated for last scan of the biotinBSAGdDTPA injection series (SNR = 10.5) and the last scan of the biotinOvalbuminGdDTPA injection series (SNR = 4.3). This resulted in a CNR

(CNR = SNR a – SNR b) of 6.3.

Figure 3.6a is an average of the 10 final frames from a dynamic fluorescence microscopy series of images following bolus injection of the AlexaFluor 633®labeled biotin

OvalbuminGdDTPA. Figure 3.6b is the corresponding vessel input function taken from a large vessel near the tumor. Very little noise has contaminated the vessel input function compared to noise inclusion from DCEMRI data with the same protein (Figure 3.4b). In order to demonstrate the difference in SNR and CNR between the AlexaFluor 633® labeled biotinBSAGdDTPA (Figure 3.7a) and AlexaFluor 633®labeled biotin

OvalbuminGdDTPA (Figure 3.7b), the SNR was calculated for last scan of the biotin

128

BSAGdDTPA injection series (SNR = 25.7) and the last scan of the biotinOvalbumin

GdDTPA injection series (SNR = 10.2). This resulted in a CNR of 15.5.

Time activity curves for DCEMRI and fluorescence microscopy data were analyzed by a full, 2compartement, 3parameter pharmacokinetic model to generate pixelbypixel maps of vp, as described in the Methods & Materials section of Chapter 2 (Figures 3.8 a and b, respectively), and those maps were registered using a canned warping routine in

IDL. Pixelbypixel covariance analysis of the coregistration delivered correlation coefficients between 0.750.90, and qualitatively, the coregistration was acceptable

(Figure 3.8c). Similarly, analysis of the timeactivity curves for DCEMRI and fluorescence microscopy data were used to generate pixelbypixel maps of Ktrans

(Figures 3.9a and b, respectively), and coregistration of the two Ktrans maps delivered correlation coefficients between 0.750.90.

Figure 3.10 displays the K trans distribution for the DCEMRI data (blue) and dynamic fluorescence microscopy data (black) for the tumor ROI (a) and the nontumor ROI (i.e. an ROI that covered the entire chamber excluding the tumor) (b). Another approach was taken to represent the histogram data in order to more clearly demonstrate difference between two populations (e.g., a comparison between the K trans distribution from DCE

MRI data and the K trans distribution from FMdata). In a give bin from a histogram, the value for that bin was normalized to the total number of pixels in the ROI and systematically subtracted from the pixel count in the bin before it. Since the first bin has

129 no prior bin pixel count, the value would assume 100%. Effectively, this results in a percentage of pixels for each bin that is greater than the next. As the number of pixels in

trans each bin (beginning at the first bin of zero) increases with increasing K or v p, the percent above that value decreases. Figures 3.11a and b represent this form of analysis for the K trans pixel distribution in the tumor and nontumor ROIs, respectively.

Figure 3.12 displays the v p distribution for the DCEMRI data (blue) and dynamic fluorescence microscopy data (black) for the tumor ROI (a) and the nontumor ROI (b).

Figures 3.13a and b represent the percent above threshold distributions for v p for the tumor and nontumor ROIs, respectively.

trans Table 3.2 is a comparison between mean v p and K values in the tumor and nontumor

ROIs for both biotinBSAGdDTPA and biotinOvalbuminGdDTPA with associated

trans standard errors. When considering both contrast agents, DCEMRImeasured v p and K values (in the tumor) are quite similar to the fluorescence microscopyderived v p and

trans K values. In the case of both contrast agents, the DCEMRIderived vp estimates are lower than the FMderived vp estimates. However, for the biotinOvalbuminGdDTPA, the DCEMRI derived Ktrans estimates are higher than those estimated using fluorescence microscopy, whereas they are lower than the fluorescence microscopy derived estimates using biotinBSAGdDTPA.

130

Figures 3.14a and b are graphical representations of Table 3.2 for biotinOvalbumin

trans GdDTPA. Means of individual values for v p and K for both regions (tumor and non tumor) are expressed coincident with standard errors for DCEMRI and dynamic fluorescence microscopy. Figures 3.15a and b were adapted from Chapter 2 (Figure 2.16 ab) in order to provide a comparison with the graphical mean data in Figures 3.14 ab.

They represent the same data in Table 3.2 for biotinBSAGdDTPA. The data are

trans expressed as the means of individual values for v p and K for both regions (tumor and nontumor) coincident with standard errors for DCEMRI and dynamic fluorescence microscopy.

DISCUSSION

A method to more precisely assess microvascular hemodynamics (specifically microvascular permeability) may require variablysized contrast agents. Because small molecules extravasate freely through the tumor endothelium and conversely, molecules which are too large may not extravasate enough to record a true estimate of K trans , changes pre and posttherapy may not be distinctly evident with either small or large (i.e.

> 80 kDa) molecular weight contrast agents. This study sought to combine the previously described multimodality imaging approach with a large molecular weight protein (~80 kDa) and an intermediate molecular weight protein (~47 kDa), both triply labeled with the same MR and FMactive conjugates. This approach allowed for assessment of these contrast agents in the same experimental model of tumor vascularization with DCEMRI and fluorescence microscopy.

131

The first topic to be addressed here is the comparison between the contrasttonoise ratio

(CNR) of the DCEMR images of biotinBSAGdDTPA (Figure 3.5a) and the DCEMR images of biotinOvalbuminGdDTPA (Figure 3.5b). The SNR of images of biotin

OvalbuminGdDTPA were much lower, yielding a CNR of 6.3 at comparable time points following the bolus injection. Importantly, concentrations required to get this CNR were

3fold higher than those used for the biotinBSAGdDTPA (compare 0.01 mmol/kg for biotinBSAGdDTPA and 0.03 mmol/kg for biotinOvalbuminGdDTPA). The SNR of the fluorescence microscopy images of AlexaFluor 633 ®labeled biotinOvalbumin

GdDTPA (Figure 3.7a) was lower than the AlexaFluor 633 ®labeled biotinBSA

GdDTPA (Figure 3.7b), producing a CNR of 15.5. Overall, the CNR was better between these two contrast agents using fluorescence microscopy than using DCEMRI.

The second important observation from these studies was the characteristic biotin

OvalbuminGdDTPA vessel input function (as illustrated from a representative animal in

Figure 3.4b). In general, these vessel input functions peaked within several minutes and quickly lost amplitude following the peak, then remained quite stable after ~1520 minutes. This behavior is consistent with what is expected for a lower molecular weight contrast agent and consistent with behavior observed in dynamic fluorescence microscopy data (Figure 3.6b). However, the noise embedded in the DCEMRI input function exceeds the noise in the dynamic fluorescence microscopy input functions.

132

trans In spite of these limitations, the K and v p values were comparable between DCEMRI and fluorescence microscopy. In the case of both contrast agents, the v p parameter derived from the DCEMRI time activity data was lower than that derived from the FM time activity data. Also in the case of both contrast agents, the K trans parameter derived from the DCEMRI time activity data was higher than that derived from the FM time activity data. Thus, although the overall difference in the magnitude of the two

trans parameters (v p and K ) derived from the two contrast agents is different, the parameters both trend toward the same direction.

Based on these observations, this current multimodal imaging methodology may then be applied in a treatment/response regimen to investigate differences in permeability changes measured with an array of contrast agent in order to further develop our understanding of changes associated with response to antiangiogenic drugs.

133

FIGURES

Figure 3.1 a Crosssectional view of 8 tubes containing a serial dilution of the biotin OvalbuminGdDTPA. The series of images were acquired at a constant echo time (TE = 9.6 ms) and multiple repetition times (TR = 15 s, 10 s, 5 s, 2 s, 1 s, 0.8 s, 0.5 s, 0.3 ms, 0.2 ms, 0.1 s). b. c. Signal Intensity vs. Repetition Time (TR) Signal Intensity vs. Repetition Time (TR) 2.5e+9 10 r = 8.06 mM 1 s 1 2.0e+9 [1.26] 8 1 r2 = 0.9877

[0.92] )

1.5e+9 [0.58] 1 6

[0.35] (s 1 1.0e+9 [0.19] 4 [0.10] 1/T

SI (arbitrary SI units) 5.0e+8 [0.05] 2 [0.02] 0.0 0 0 3 5 8 10 13 15 0.00 0.25 0.50 0.75 1.00 TR (s) [biotinOvalbuminGdDTPA] (mM)

134

Figure 3.1 b-c Corresponding recovery curves for each concentration (displayed in brackets in terms of mM) at multiple repetition times (b). Calculated T 1 plotted against corresponding concentration of biotinOvalbuminGdDTPA (c).

Figure 3.2 a Crosssectional view of 8 tubes containing a serial dilution of the biotin OvalbuminGdDTPA. The series of images were acquired at a constant repetition time (TR = 15000 ms) and multiple echo times (TE = 9.6 ms, 20 ms, 30 ms, 40 ms, 60 ms, 80 ms, 100 ms, 150 ms, 200 ms, 250 ms, 300 ms, 400 ms, 500 ms, 750 ms, 1000 ms). b. c. Signal Intensity vs. Repetition Time (TE) Signal Intensity vs. Repetition Time (TE) 2.2e+9 50 [1.26] ~1 1 r = 39.8 mM s [0.92] 40 2 1.7e+9 r2 = 0.9747

[0.58] )

[0.35] 1 30 1.1e+9 (s [0.19] 2 20

[0.10] 1/T 5.5e+8 [0.05] SI (arbitrary units) (arbitrary SI [0.02] 10

0.0 0 0.0 0.2 0.4 0.6 0.8 1.0 0.00 0.25 0.50 0.75 1.00 TE (s) [biotinOvalbuminGdDTPA] (mM)

Figure 3.2 b-c Corresponding decay curves for each concentration (displayed in brackets in terms of mM) at multiple echo times (b). Calculated T 2 plotted against corresponding concentration of biotinOvalbuminGdDTPA (c).

135

biotinBSAGdDTPA 23 biotinOvaGdDTPA 7

Molecular Weight s 78,615 gmol 1 47,454 gmol 1 Formula Weight Gd s 3,479 gmol 1 6,749 gmol 1 % Gadolinium s 4.52 2.33 (Elemental Analysis) s # Gd per molecule s 23 7

Table 3.1 Table of composition characteristics between biotinBSAGdDTPA and biotin OvalbuminGdDTPA.

Figure 3.3 a-b Representative transillumination image overlaid with GFP fluorescence at 10X (a) and 20X.

a b

0.07

0.06

0.05

0.04

0.03

0.02

0.01

[biotinOvalbuminGdDTPA] (mM) [biotinOvalbuminGdDTPA] 0.00 0 5 10 15 20 25 30 Time (min)

136

Figure 3.4 a-b Average of several frames from a DCEMRI series of the biotin OvalbuminGdDTPA injection (a). Example of the timeactivity curve obtained from an ROI (redfilled object) drawn over a major blood vessel throughout the DCEMR image series. This vessel input function was a characteristic for the biotinOvalbuminGdDTPA contrast agent.

a b

SNR = 10.5 SNR = 4.3

CNR = 6.2

Figure 3.5 a-b Comparison between SNR and CNR of a DCEMRI series following injection with biotinBSAGdDTPA (a) (taken from Chapter 2, Figure 2.6c) and biotin OvalbuminGdDTPA (b). SNR is 2.5 fold higher for the biotinBSAGdDTPA, resulting in a CNR of 6.2 between the two contrast agents.

a b 0.007

0.006

0.005

0.004

0.003

0.002

0.001

[biotinOvalbuminGdDTPA] (mM) [biotinOvalbuminGdDTPA] 0.000 0 5 10 15 20 25 30 Time (min)

Figure 3.6 a-b Average of several frames from a dynamic fluorescence microscopy series of the AlexaFluor 633 ®labeled biotinOvalbuminGdDTPA injection (a). Example

137 of the timeactivity curve obtained from an ROI (redfilled object) drawn over a major blood vessel throughout the series.

a b

SNR = 25.7 SNR = 10.2

CNR = 15.5

Figure 3.7 a-b Comparison between SNR and CNR of a dynamic fluorescence microscopy series following injection with AlexaFluor 633 ®labeled biotinBSA GdDTPA (a) (taken from Chapter 2, Figure 2.8c) and biotinOvalbuminGdDTPA (b). SNR is 2.5 fold higher for the biotinBSAGdDTPA, resulting in a CNR of 15.5 between the two contrast agents.

a 100 b 100 c

80 80

60 60

(%) (%) p p

v 40 v 40

20 20

0 0

Figure 3.8 a-c Dynamic fluorescence microscopyderived (a) and DCEMRIderived (b) vp maps obtained from concentrationtime activity curves on a pixelbypixel basis.

138 completely registered pixels in those designated areas (correlation coefficients for registration ranged between 0.75 and 0.90).

a 0.0072 b 0.0072 c

0.0057 0.0057 ) ) 1 1 - - 0.0043 0.0043

(min (min 0.0033 0.0033 trans trans K K 0.0018 0.0018

0 0

Figure 3.9 a-c Dynamic fluorescence microscopyderived (a) and DCEMRIderived (b) Ktrans maps obtained from concentrationtime activity curves on a pixelbypixel basis. Warped fluorescence microscope K trans maps were blended with (or overlaid on) the reference DCEMRI K trans map (c). Coregistration of the two maps ensured that ROIs drawn over the tumor and the nontumor space would result in inclusion of near completely registered pixels in those designated areas (correlation coefficients for registration ranged between 0.75 and 0.90).

a trans 1 b trans 1 300 K (min ) 300 K (min ) Tumor Non Tumor 250 250

200 200 FM Mean: 0.000436 FM Mean: 0.00040 150 DCEMRI Mean: 0.000432 150 DCEMRI Mean: 0.00077 No. No. Pixels No. No. Pixels 100 100

50 50

0 0 0.000 0.001 0.002 0.003 0.004 0.005 0.000 0.001 0.002 0.003 0.004 0.005 Ktrans (min 1 ) Ktrans (min 1 )

139

Figure 3.10 a-b Histograms of K trans derived from DCEMRI data (blue) and dynamic FM data (black) for both the tumor ROI (a) and the nontumor ROI (b) (different scales were used for each ROI to maintain the integrity of an adequate comparison).

a trans 1 b trans 1 120 K (min ) 120 K (min ) Tumor Non Tumor 100 100

80 80 FM Mean: 0.000436 FM Mean: 0.00040 60 DCEMRI Mean: 0.000432 60 DCEMRI Mean: 0.00077 No. No. Pixels 40 No. Pixels 40

20 20

0 0 0.000 0.005 0.010 0.015 0.020 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 )

Figure 3.11 a-b Distributions for DCEMRI data (blue circles) and dynamic fluorescence microscopy data (black circles) were created by normalizing histograms to percentage of pixels above the respective K trans value. Distributions are shown for the tumor (a) and nontumor (b) ROIs.

Vascular Volume Fraction (%) Vascular Volume Fraction (%) a Tumor b Non Tumor 3000 1600 1400 2500 1200 2000 FM Mean: 13.76 1000 FM Mean: 17.87 DCEMRI Mean: 13.75 DCEMRI Mean: 09.83 1500 800 600 No. Pixels No. No. Pixels No. 1000 400 500 200

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

140

Figure 3.12 a-b Histograms of v p derived from DCEMRI data (blue) and dynamic FM data (black) for both the tumor ROI (a) and the nontumor ROI (b) (different scales were used for each ROI to maintain the integrity of an adequate comparison).

Vascular Volume Fraction (%) Vascular Volume Fraction (%) a Tumor b Non Tumor 120 120

100 100

80 FM Mean: 13.76 80 FM Mean: 17.87 DCEMRI Mean: 13.75 DCEMRI Mean: 09.83 60 60 No.Pixels No.Pixels 40 40

20 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

Figure 3.13 a-b Distributions for DCEMRI data (blue circles) and dynamic fluorescence microscopy data (black circles) were created by normalizing histograms to percentage of pixels above the respective v p value. Distributions are shown for the tumor (a) and non tumor (b) ROIs.

biotinBSAGdDTPA 23 biotinOvaGdDTPA 7 DCEMRI FM DCEMRI FM 2.79E03 ± 2.16E03 ± 2.72E03 ± 1.72E03 ± Ktrans (tumor) 1.4E04 8.8E05 5.7E04 3.0E04 3.36E03 ± 2.63E03 ± 1.51E03 ± 5.19E04 ± Ktrans (nontumor) 1.9E04 1.9E04 1.8E04 3.9E05 12.07 ± 16.46 ± 08.24 ± 21.10 ± v (tumor) p 0.68 0.82 0.31 0.76 14.83 ± 17.56 ± 12.21 ± 21.87 ± v (nontumor) p 0.53 0.54 0.37 0.69

141

trans Table 3.2 Comparison between mean vp and K values in the tumor and nontumor ROIs for both biotinBSAGdDTPA and biotinOvalbuminGdDTPA with associated trans standard errors. For both contrast agents, DCEMRImeasured v p and K values in the trans trans tumor are quite similar to FMderived v p and K values, with DCEMRIderived K estimates slightly elevated compared to FMderived K trans estimates, and DCEMRI derived vp estimates slightly lower compared to FMderived vp estimates. a b Vascular Volume Fraction (%) Ktrans (min 1 ) Tumor 25 Tumor NonTumor 0.004 NonTumor 20 )

1 0.003 15 (%) (min p 0.002

v 10

trans 0.001 5 K

0 0.000 MRI FM MRI FM

Figure 3.14 a-b Graphical representation of Table 3.2 for biotinOvalbuminGdDTPA. trans Means of individual values for v p and K for both regions (tumor and nontumor) are expressed coincident with standard errors for DCEMRI and dynamic fluorescence microscopy.

142

trans 1 a Vascular Volume Fraction (%) b K (min ) Tumor 20 Tumor NonTumor 0.004 NonTumor

15 )

1 0.003

(%) 10 (min

p 0.002 v

5 trans 0.001 K

0 0.000 MRI FM MRI FM Figure 3.15 a-b Graphical representation of Table 3.2 for biotinBSAGdDTPA. Means trans of individual values for v p and K for both regions (tumor and nontumor) are expressed coincident with standard errors for DCEMRI and dynamic fluorescence microscopy. The identical graph was presented originally in Chapter 2, Figure 2.16 ab, but has been reproduced here for comparison with Figure 3.14 ab.

143

CHAPTER 4

Tumor Response to Sutent® Using Contrast Media of Variable Molecular Weight and

Assessment of Two Tracer Kinetic Models

INTRODUCTION

Suntinib (SU11248) is a small molecule inhibitor of multiple receptor tyrosine kinases, and Sutent® is the malate salt derived from sunitib. In general, Sutent® is a paninhibitor of endothelial cell proliferation and survival by targeting cellular receptors such as

Vascular Endothelial Growth Factor Receptor (VEGFR)1, 2 & 3, Platelet Derived

Growth Factor Receptor (PDGFR)β, stem cell factor receptor (cKit), colonystimulating factor type 1 (CSF1R). In animals, Sutent® has been shown to effectively lower permeability in experimental colon tumors using either low or high molecular weight contrast agents (201), inhibition of tumor growth in a bone metastases model (202), and inhibition of experimental breast tumor growth and prolonged survival (203). In humans,

Sutent® is effective against gastrointestinal stromal tumors (204), renal cell carcinomas

(205) and a variety of hematological malignancies (206). However, as with other anti angiogenic therapies, the clinical translation of tumor response measurements as imaging biomarkers remains limited primarily due to lack of experimental cross validation techniques.

Based on the methods described in the previous two chapters, this study extended the use of the MRIcompatible dorsal midline skinfold window chamber and data from both

144

DCEMRI and dynamic intravital FM to determine if there was a relationship between druginduced changes in tumor hemodynamics and molecular weight of the contrast agent. Since the imaging system is designed to assess changes in microvascular

trans hemodynamics, changes in K or v p following treatment with Sutent® were expected to be readily evident. Using data analyzed from the dualmodality imaging, it was possible to investigate differences in K trans between fluorescence microscopy and MRI as well as between contrast agents and applied mathematical models.

METHODS & MATERIALS

See Chapters 2 and 3 for details on the majority of these Methods and Materials.

Administration of Sutent®. A daily dose of 40 mg/kg Sutent® (Pfizer, New York, NY) was administered to animals via oral gavage (p.o.) for 710 days.

Tumor Response. Tumor response to therapy was indicated by changes in tumor volumes pre and posttherapy. Volumes were calculated by drawing ROIs and calculating tumor areas on GFPtumor fluorescence images.

RESULTS

Figure 4.1a and b display the transillumination images for a representative animal overlaid with GFP fluorescence at 10X pre and posttherapy, respectively. Following

145 injection of the biotinBSAGdDTPA, timeactively curves obtained using DCEMRI were quantitatively analyzed on a pixelbypixel basis for vp, and results are shown pre therapy (Figure 4.1c) and posttherapy (Figure 4.1d). Corresponding v p maps derived from dynamic fluorescence microscopy following injection of the fluorophorelabeled biotinBSAGdDTPA are shown in the same animal pretherapy (Figure 4.1e) and post therapy (Figure 4.1f). Pretherapy (Figure 4.1g) and posttherapy (Figure 4.1h) DCE

MRIderived K trans maps are shown for the same animal. Finally, the FMderived K trans maps pretherapy (Figure 4.1i) and posttherapy (Figure 4.1j) are shown for the same animal.

trans For all pre and posttherapy v p and K maps, associated histograms were generated.

Histograms from the representative data sets in Figure 4.1 are displayed as pretherapy distributions in black and posttherapy distributions in blue. Figure 4.2a shows the representative histogram in the tumor ROI for the DCEMRIderived v p map. A secondary approach (described in Chapter 3) was taken to represent the histogram data in order to more clearly demonstrate difference between two populations. To reiterate, in a given bin from a histogram, the value for that bin was normalized to the total number of pixels in the ROI and systematically subtracted from the pixel count in the bin before it.

Since the first bin has no prior bin pixel count, the value would assume 100%.

Effectively, this results in a percentage of pixels for each bin that is greater than the next.

As the number of pixels in each bin (beginning at the first bin of zero) increases with

trans increasing K or v p, the percent above that value decreases. Figures 4.2b is the percent

146 above threshold histogram for the pre and posttherapy histogram data from the DCE

trans MRIderived v p maps. Histograms were generated for the DCEMRIderived K maps pre and posttherapy (Figure 4.2e) and corresponding percent above threshold distributions are shown (Figure 4.2f). For the fluorescence microscopy data, histograms

trans for the pre and posttherapy v p (Figure 4.2c) and K (Figure 4.2g) maps were generated and displayed alongside their corresponding percent above threshold distributions (Figure 4.2d and h).

For the experiments using the biotinBSAGdDTPA (all representative data illustrated

trans above), tumor response to treatment as indicated by v p and K parameters (in the tumor

ROI) is tabulated (Table 4.1). The K trans parameter appears to be the most sensitive indicator of response to therapy with an overwhelming and consistently lower K trans estimate posttherapy for all animals. Figures 4.3a and b are graphical representations of the information in Table 4.1. It was expected that this expression would aid in clarifying trends in the data.

For animals receiving biotinOvalbuminGdDTPA (n = 4), MR images were acquired pre and posttherapy with Sutent, immediately followed by fluorescence microscopy using the AlexaFluor 633 ®labeled biotinOvalbuminGdDTPA. Figure 4.4a and b display the transillumination images for a representative animal overlaid with GFP fluorescence at 10X pre and posttherapy, respectively. Following injection of the biotin

OvalbuminGdDTPA, timeactively curves obtained using DCEMRIwere quantitatively

147 analyzed on a pixelbypixel basis for v p, and results are shown pretherapy (Figure 4.4c) and posttherapy (Figure 4.4d). Corresponding v p maps derived from dynamic fluorescence microscopy following injection of the fluorophorelabeled biotinBSA

GdDTPA are shown in the same animal pretherapy (Figure 4.4e) and posttherapy

(Figure 4.4f). Pretherapy (Figure 4.4g) and posttherapy (Figure 4.4h) DCEMRI derived K trans maps are shown for the same animal. Finally, the FMderived K trans maps pretherapy (Figure 4.4i) and posttherapy (Figure 4.4j) are shown for the same animal.

trans For all pre and posttherapy v p and K maps generated using biotinOvalbumin

GdDTPA, histograms were created. Representative pretherapy distributions are shown in black, and posttherapy distributions are shown in blue. Figure 4.5a shows the representative histogram in the tumor ROI for the DCEMRIderived v p map and its associated percent above threshold distribution is displayed as Figure 4.5b. Histograms were generated for the DCEMRIderived K trans maps pre and posttherapy (Figure 4.5e) as well and the corresponding percent above threshold distributions are shown (Figure

4.5f). For the fluorescence microscopy data, histograms for the pre and posttherapy v p

(Figure 4.5c) and K trans (Figure 4.5g) maps were generated and displayed alongside their corresponding percent above threshold distributions (Figure 4.5d and h).

For the experiments using the biotinOvalbuminGdDTPA (all representative data

trans illustrated above), tumor response to treatment as indicated by v p and K parameters (in the tumor ROI) is tabulated (Table 4.2). Neither parameter appears to be the most

148 sensitive indicator of response to therapy. Indeed, according to tumor response data (% change tumor volumes), there appears to be no perceivable effect of Sutent on the tumor microvasculature with DCEMRI or dynamic FM. Figures 4.6a and b are graphical representations of the information in Table 4.2 quantifying the mean change in Ktrans and vp parameters in the tumor for both the DCEMR data and the fluorescence microscopy data pre and posttreatment. It was expected that this expression would aid in clarifying trends in the data.

Figures 4.7 a and b display the pre (a) and posttherapy (b) v p maps from a representative animal that demonstrated a clear decrease in the v p parameter despite large increases in tumor response data (the tumor volume more than doubled).

As demonstrated in Chapter 3, due to a difference in the molecular weight of biotin

OvalbuminGdDTPA, the vessel input function differed from biotinBSAGdDTPA, precluding the use of the simplified slopeintercept method. The more rigorous, 2 compartment, 3parameter model was more appropriate in this case (for details on the derivation of the 2compartment, 3parameter kinetic model, please see Appendix A).

The biotinOvalbuminGdDTPA vessel input function data were lownoise using fluorescence microscopy (Figure 3.5b), and in the case for both DCEMRI and dynamic

FM, both vessel input functions demonstrate a distinct clearance of the tracer after an initial peak (Figure 3.3b and 3.5b). Due to this behavior, data were modeled according to the full 2compartment, 3parameter model.

149

DISCUSSION

The most notable contribution from this study is the biotinBSAGdDTPA estimated posttherapy K trans changes using both DCEMRI and dynamic FM. Following therapy, significant decreases in K trans were observed using both DCEMRI and dynamic FM. The vp parameter did not show appreciable change posttherapy, but for dynamic FM, there appeared to be a trend towards increases in v p posttherapy.

Estimates of hemodynamic parameters using biotinOvalbuminGdDTPA did obtain significance, however, the indication from these parameters was a lack of response following therapy with the antiangiogenic. For example, for both imaging systems, large increases in K trans were observed posttherapy. Interestingly, a trend toward decreases in vp posttherapy were observed using both imaging systems, though those decreases were not significant. This is interesting because an increase in K trans suggests certain physiological activity geared towards survival or maintenance of the tumor, for example, remodeling, increases in vessel diameter or tortuosity, which would all presume a corresponding increase in the vascular volume fraction parameter, v p.

In order to understand this behavior, it is more important to view each subject independently to better understand the source of this potential discrepancy (Table 4.2). A decrease in the v p parameter was observed in only two of the subjects following therapy.

The first largest decrease in v p was recorded for a tumor that demonstrated a 70%

150 decrease in its volume. Such a change in tumor volume might accurately reflect change in vascular physiology as measured by the biotinOvalbuminGdDTPA, however, K trans is still larger posttherapy as measured by both imaging systems. It may be explained that the dynamic FMderived Ktrans , albeit larger posttherapy, is not considerably so, and the

DCEMRIderived K trans , albeit considerably larger posttherapy, may be slightly inaccurate (compared to dynamic FM as well as imaging performed using the other contrast agent, biotinBSAGdDTPA) due to SNR and CNR as discussed in Chapter 3.

The second decrease in v p posttherapy was recorded for a tumor that more than doubled in size. However, when looking at the tumor, it can be seen that the pretherapy v p map displays large, healthy vessels with superb intravascular signal (Figure 4.7a). Post therapy, one of the largest vessels (that presumably contributed significantly to high estimates of v p pretherapy) has now been recruited to almost completely circumscribe the tumor (Figure 4.7b). However, that vessel appears to have either 1) a reduced capacity to allow for adequate flow of the tracer through its lumen or; 2) experienced a fast washout or sluggish washin. Either case would reflect a lower estimate of v p when the data are fit to current pharmacokinetic models. It is hypothesized that this is the cause of a substantially reduced v p posttherapy in spite of the survival and growth of the tumor in this case.

151

trans For the other two subjects, both the v p and K parameter appear to decrease following therapy, consistent with a nonresponse to therapy and consistent with tumor response data (which are a strong indication a lack of response to therapy).

For the single subject demonstrating a lack of response to therapy (as indicated by v p and

Ktrans ) in spite of significant decreases in tumor volume, such behavior may be explained by one of two hypotheses: 1) therapy was not actually administered in the manner of which was expected (i.e. the oral gavage method of 40 mg/kg every day for 710 days) or; 2) the new biotinOvalbuminGdDTPA extravasated so rapidly, that it could not accurately reflect vessel physiology (i.e. vascular normalization, decreases in vessel density, diameter and permeability) in response to therapy. The preparation of the drug is a challenging task, requiring vigorous shaking and sonication immediately prior to administration. Technically, sonicating and vortexing the drug in its solution is possible up until the point of loading it into a syringe. Once the solution is loaded, the animal is immobilized and the drug is administered. During this time, the drug generally settles inside the syringe, making accurate dosages almost impossible. A logical solution for actually determining whether the biotinOvalbuminGdDTPA is an adequate contrast

trans agent for providing robust measures of v p and K in response to therapy is to perform the experiment with an antiangiogenic that is more easily and accurately administered and use a larger number of subjects.

152

In conclusion, the use of biotinBSAGdDTPA for both DCEMR and dynamic FM

trans imaging provides estimates for v p and K that change according to response to anti

trans angiogenic therapy. Imaging with biotinOvalbuminGdDTPA produced v p and K estimates that were closely related between the two imaging modalities, and in most cases, changes in those parameters were consistent with tumor response data (i.e. most animals did not respond to therapy). These data substantiate the need to improve specifically the biotinOvalbuminGdDTPA contrast agent SNR and CNR issues as well as the drug administration techniques for this type of study as the information provided by this study would not have been easily discovered without the combined imaging system.

153

FIGURES

a b

c d 100.0

80.0

60.0

(%) p 40.0 v 20.0

e f 0 100.0

80.0

60.0

(%) p 40.0 v 20.0

0 0.010 g h

) 0.008 1 - 0.006 (min

trans 0.004

K 0.002

0 0.010 i j

) 0.008 1 - 0.006 (min

trans 0.004

K 0.002

154

Figure 4.1 a-j For experiments using the biotinBSAGdDTPA, a representative set of images pre and posttreatment with Sutent is displayed. Transillumination images overlaid with GFP fluorescence at 10X pretherapy (a) and posttherapy (b); DCEMRI derived v p maps pretherapy (c) and posttherapy (d); fluorescence microscopy (FM) trans derived v p maps pretherapy (e) and posttherapy (f); DCEMRIderived K maps pre therapy (g) and posttherapy (h); FMderived K trans maps pretherapy (i) and posttherapy (j).

155

Vascular Volume Fraction (%) Vascular Volume Fraction (%) a DCEMRI b DCEMRI 10 120

8 100

PreTx Mean: 12.01 80 6 PreTx Mean: 12.14 PostTx Mean: 15.54 PostTx Mean: 15.52 60 4 No. Pixels No. No. Pixels 40 2 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

Vascular Volume Fraction (%) Vascular Volume Fraction (%) c FM d FM 140 120

120 100 100 PreTx Mean: 13.34 80 PreTx Mean: 13.34 PostTx Mean: 12.01 80 PostTx Mean: 12.01 60 60 No. Pixels No. Pixels 40 40

20 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%) e f trans 1 trans 1 12 K (min ) 120 K (min ) DCEMRI DCEMRI 10 100

8 80 PreTx Mean: 0.00356 PreTx Mean: 12.14 6 PostTx Mean: 0.00121 60 PostTx Mean: 15.52 No. Pixels No. 4 Pixels No. 40

2 20

0 0 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 )

g trans 1 h trans 1 60 K (min ) 120 K (min ) FM FM 50 100

40 80 PreTx Mean: 0.00223 PreTx Mean: 0.00223 30 PostTx Mean: 0.00157 60 PostTx Mean: 0.00157 No. Pixels 20 Pixels No. 40

10 20

0 0 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 ) Figure 4.2 a-h For experiments using the biotinBSAGdDTPA, a representative set of histograms and percent above threshold distributions pre and posttreatment with Sutent is displayed. For DCEMRIderived v p information, histograms display the pretherapy

156

(black) and posttherapy (blue) distributions of v p in the tumor ROI (a), and an alternative representation of those distributions, the percent above a specified threshold distribution, is displayed in (b), also differentiating between pretherapy (black) and posttherapy

(blue); FMderived v p histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (c) and associated percent above threshold distributions (d) pretherapy (black) and posttherapy (blue); DCEMRIderived K trans histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (e) and associated percent above threshold distributions (f) pretherapy (black) and posttherapy (blue); FMderived K trans histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (g) and associated percent above threshold distributions (h) pretherapy (black) and posttherapy (blue).

1 tumor imaging vp (%) ktrans (min ) volume system pretx posttx % v p ∆ pretx posttx % ktrans ∆ % tv ∆ mr 8.31 12.24 0.47 0.00386 0.00218 0.43 subject #1 0.24 fm 18.29 n/a 0.00222 n/a mr 12.14 15.52 0.28 0.00356 0.00120 0.66 subject #2 0.80 fm 13.34 12.03 0.10 0.00223 0.00157 0.29 mr 19.68 15.52 0.21 0.00463 0.00120 0.74 subject #3 1.04 fm 11.67 12.03 0.03 0.00225 0.00157 0.30 mr 11.76 2.09 0.82 0.00166 0.00140 0.16 subject #4 0.01 fm 10.22 16.37 0.60 0.00152 0.00135 0.11

Table 4.1 For experiments using the biotinBSAGdDTPA, tumor response to treatment trans trans as indicated by v p and K parameters (in the tumor ROI) has been tabulated. The K parameter appears to be the most sensitive indicator of response to therapy with an overwhelming and consistently lower K trans estimate posttherapy for all animals.

157

a b Vascular Volume Fraction (%) Ktrans (min 1 )

PreTherapy PreTherapy 20 PostTherapy 0.008 PostTherapy

15 ) 1 (%)

10 (min 0.004 p v 5 trans K

0 0.000 MRI FM MRI FM

trans Figure 4.3 a-b Graphical representation of Table 4.1. Data represent v p and K maps generated from timeactivity data obtained using biotinBSAGdDTPA and AlexaFlour 633labeled biotinBSAGdDTPA.

158

a b

100.0 c d 80.0

60.0

(%) p 40.0 v 20.0

0 e f 100.0

80.0

60.0

(%) p 40.0 v 20.0

0 g h 0.010

) 0.008 1 - 0.006 (min

trans 0.004

K 0.002

0 i j 0.010

) 0.008 1 - 0.006 (min

trans 0.004

K 0.002

0 Figure 4.4 a-j For experiments using the biotinOvalbuminGdDTPA, a representative set of images pre and posttreatment with Sutent is displayed. Transillumination images overlaid with GFP fluorescence at 10X pretherapy (a) and posttherapy (b); DCEMRI

159 derived v p maps pretherapy (c) and posttherapy (d); fluorescence microscopyderived v p maps pretherapy (e) and posttherapy (f); DCEMRIderived K trans maps pretherapy (g) and posttherapy (h); fluorescence microscopyderived K trans maps pretherapy (i) and posttherapy (j).

160

Vascular Volume Fraction (%) Vascular Volume Fraction (%) a DCEMRI b DCEMRI 50 120

40 100

PreTx Mean: 3.75 80 30 PreTx Mean: 3.75 PostTx Mean: 9.43 PostTx Mean: 9.43 60 20 No. Pixels No. Pixels 40

10 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%) c Vascular Volume Fraction (%) d Vascular Volume Fraction (%) FM FM 3000 120

2500 100

2000 PreTx Mean: 13.76 80 PreTx Mean: 13.76 PostTx Mean: 33.38 PostTx Mean: 33.38 1500 60 No. No. Pixels 1000 No. Pixels 40

500 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 Vascular Volume Fraction (%) Vascular Volume Fraction (%)

e trans 1 f trans 1 25 K (min ) 120 K (min ) DCEMRI DCEMRI 100 20

80 15 PreTx Mean: 0.00043 PreTx Mean: 0.00043 PostTx Mean: 0.00083 60 PostTx Mean: 0.00083 10 No. Pixels No. Pixels 40

5 20

0 0 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 ) g trans 1 h trans 1 350 K (min ) 120 K (min ) FM FM 300 100 250 80 200 PreTx Mean: 0.00044 PreTx Mean: 0.00044 PostTx Mean: 0.00136 60 PostTx Mean: 0.00136 150

No. No. Pixels No. Pixels 40 100

50 20

0 0 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.000 0.005 0.010 0.015 0.020 Ktrans (min 1 ) Ktrans (min 1 ) Figure 4.5 a-h For experiments using the biotinOvalbuminGdDTPA, a representative set of histograms and percent above threshold distributions pre and posttreatment with

Sutent is displayed. For DCEMRIderived v p information, histograms display the pre

161 therapy (black) and posttherapy (blue) distributions of v p in the tumor ROI (a), and an alternative representation of those distributions, the percent above a specified threshold distribution, is displayed in (b), also differentiating between pretherapy (black) and post therapy (blue); FMderived v p histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (c) and associated percent above threshold distributions (d) pretherapy (black) and posttherapy (blue); DCEMRIderived K trans histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (e) and associated percent above threshold distributions (f) pretherapy (black) and posttherapy (blue); FMderived K trans histograms pretherapy (black) and posttherapy (blue) in the tumor ROI (g) and associated percent above threshold distributions (h) pretherapy (black) and posttherapy (blue).

trans 1 tumor imaging vp (%) k (min ) volume system pretx posttx % v p ∆ pretx posttx % ktrans ∆ % tv ∆ mr 11.56 4.12 0.64 0.00052 0.00088 0.69 subject #1 0.70 fm 31.45 0.66 0.98 0.00028 0.00033 0.17 mr 2.99 9.43 2.16 0.00043 0.00083 0.94 subject #2 0.34 fm 16.86 33.38 0.98 0.00039 0.00136 2.47 mr 9.82 10.62 0.08 0.00188 0.02449 12.03 subject #3 1.39 fm 19.04 21.58 0.13 0.00107 0.01292 11.03 mr 15.96 6.53 0.59 0.00040 0.00083 1.06 subject #4 1.04 fm 23.13 18.19 0.21 0.00046 0.00069 0.52

Table 4.2 For experiments using the biotinOvalbuminGdDTPA, tumor response to trans treatment as indicated by v p and K parameters (in the tumor ROI) has been tabulated. Neither parameter appears to be the most sensitive indicator of response to therapy. Indeed, according to tumor response data (% change tumor volumes), there appears to be no perceivable effect of Sutent on the tumor microvasculature with DCEMRI or dynamic FM.

162

a b Vascular Volume Fraction (%) Ktrans (min 1 )

PreTherapy PreTherapy 30 PostTherapy 0.012 PostTherapy 25 ) 20 1 0.008

(%) 15 (min p v

10 trans 0.004

5 K 0 0.000 MRI FM MRI FM

trans Figure 4.6 a-b Graphical representation of Table 4.1. Data represent v p and K maps generated from timeactivity data obtained using biotinOvalbuminGdDTPA and AlexaFlour633labeled biotinOvalbuminGdDTPA.

100 100

80 80

60 60 (%) (%) (%) (%) p p v v 40 40

20 20

0 0

Figure 4.7 a-b Representative animal demonstrating a clear decrease in the v p parameter despite large increases in tumor response data (the tumor volume more than doubled).

Pretherapy (a) image of v p map and posttherapy (b) image of v p map.

163

CONCLUDING REMARKS

These studies have demonstrated that the fullyfunctional dualimaging system developed in this dissertation is capable of producing high quality imaging data from both DCE

MRI and dynamic fluorescence microscopy in an identical field of view to cross correlated contrast agent extravasation from the vasculature and subsequent quantification of tumor microvascular hemodynamics. A crosscomparison of the distribution of a duallabeled contrast agent will ultimately allow for better quantitation of MR signal through improvement and optimization of tracer kinetic models and as well as provide a better understanding of microscopic partial volumes underlying MR signal behavior. These studies have also shown the utility of a new, biotinylated MRactive contrast agent of a smaller molecular weight, particularly in the context of its ability to measure posttherapy induced changes in microvascular hemodynamics using fluorescence microscopy. Such an imaging system might help to optimize other synthesized contrast agents, which may eventually lead to a contrast agent or array of contrast agent sizes that will most accurately define tumor pathophysiology in the context of antivascular and antiangiogenic therapy.

164

APPENDIX A

The following is an introduction to the threecompartment mass balance equation, and the evolution from this equation to a solution for the movement of a contrast agent from one compartment to the next for both macromolecules and small molecules.

1. The mass balance equation:

Ct = v pC p + veCe + viCi 0 assume contrast agent does not cross the cell membrance Ct = [tracer] in the tissue Cp = [tracer] in the blood plasma Ce = [tracer] in the EES Ci = [tracer] in the intracellular space vp = blood plasma volume (per unit volume tissue) ve = EES volume (per unit volume tissue) vi = intracellular volume (per unit volume tissue)

An implicit assumption here is that there is no distinction between arterial and venous

compartments, there is only a plasma compartment.

2. Take the derivative of the mass balance equation to find the rate of change of tissue

concentration over time:

dC dC dC t = v p + v e dt p dt e dt

Fρ 1( − Hct)(C p − Ce ) PSρ(C p − Ce )

(mM) (g/mL) (mM) perfusion of whole blood total permeability of capillary wall per unit mass tissue (cm/min) (mL/gmin)

tissue density surface area per unit mass tissue (g/mL) (cm 2/g)

165

3. The transfer constant, K trans , can be used to describe different physiologic scenarios,

all of which are a function of the capillary permeability and blood flow in the tissue.

In the FlowLimited – High Permeability context (for example, small molecular

weight contrast agent), venous blood leaves with a [tracer] that is at all times in

equilibrium with tissue [tracer] (since permeability is high), therefore, after injection,

arterial [tracer] is high and venous [tracer] is low because most of the tracer is being

removed from the blood as it moves through the tissue. In this situation,

K trans = Fρ 1( − Hct)

In the Permeability Surface Area Product (PS)Limited – Low Permeability context

(for example, a large molecular weight contrast agent), the rate of uptake is

determined by the PS of the capillary. If flow is high, the arterial and venous tracer

concentrations are considered equal and the transport of tracer out of the vasculature

is slow enough not to deplete the intravascular concentration, and

K trans = PSρ

4. A generalized form of the kinetic model can be written as follows:

dC dC t = v p + K trans ()C − C dt p dt p e

5. Now, C e can be substituted with C e from the mass balance equation (C t=v pCp+v eCe):

Ct − v pC p Ce = ve

and

166

dC dC   v  C  t p trans  p  t = v p + K C p 1+  −  dt dt   ve  ve 

and

dC dC  v  C t p trans  p  trans t = v p + K C p 1+  − K dt dt  ve  ve

6. Now solve for dC t/dt:

trans Ct (τ ) C p (τ )  v  τ K τ dC (t) = v dC (t) + K trans 1+ p  C (t)dt − C (t)dt ∫0 t p ∫0 p  ∫0 p ∫0 t  ve  ve

7. The solution to this equation, with initial conditions (C p = C t = 0 @ t=0), is as

follows:

trans  v  τ K τ C (τ ) − C )0( = v ()C (τ ) − C )0( + K trans 1+ p  C (t)dt − C (t)dt t t p p p  v ∫0 p v ∫0 t 0 0  e  e

8. Approximating the integrals using summation of trapezoids and MLR yields:

 v  τ K trans τ trans  p  Ct (τ ) = v pC p (τ ) + K 1+ ∑C pt − ∑Ct t  ve  0 ve 0

In a theoretically perfect flowlimited regime (high permeability, small molecules),

the arterial input function matches the rate if tracer accumulation in the tissue.

However, in a theoretically perfect permeabilitysurface area productlimited regime

(low permeability, macromolecules), the arterial input function is a step function, and

167

the rate of tracer accumulation in the tissue is linear in the early stages of uptake

when C e is approximately equal to zero:

1. Going back to the generalized form of the kinetic model:

dC dC t = v p + K trans ()C − C dt p dt p e

2. Since C e is approximately equal to zero, the model takes the form:

dC dC t = v p + K transC dt p dt p

3. Again solve for dC t/dt:

C (τ ) C (τ ) τ t p trans dCt (t) = v p dC p (t) + K C p (t)dt ∫0 ∫0 ∫0

4. The solution to this equation, with initial conditions (C p = C t = 0 @ t=0), is as

follows:

τ trans Ct (τ ) − Ct )0( = v p (C p (τ ) − C p )0( )+ K C p (t)dt ∫0 0 0 5. Finally, the equation is simplified down to a firstorder polynomial (linear equation):

trans Ct (t) = v p C p + K C p t

trans And K and v p (after being normalized to the tracer concentration in the plasma,

Cp) are the slope and intercept, respectively. The effective microvascular permeability

is related to K trans by the following equation:

K trans = Q(1− e−Peff S /Q ),

168 where Q is the plasma flow rate per unit tissue weight and S is the surface area of the microvessel per unit tissue weight. For large macromolecules, P eff S << Q, and in

trans such cases the K is related to P eff simply by:

trans K = Peff S .

169

APPENDIX B

MR Data Acquisition

T1map_VTR 5s, 3s, 2s,1s, 0.8s, 0.5s, 0.2, 0.1

Absorbance Spectroscopy to determine concentration of injectate

Series Scan TE = 9.648 ms, TR = 100 ms, NR = 45, Scantime = 30 min Acquire 2 repetitions preinjection, & for a duration of 3 min, inject 200 L of ~0.1 mmol/kg biotinBSAGdDTPA

Remove animal, collect blood in capillary tube, spin down using Hct microcentrifuge for 5 min

MR Data Acquisition

T1map_VTR 5s, 3s, 2s,1s, 0.8s, 0.5s, 0.2, 0.1

msme, TE = 10 ms msme, TE = 20 ms . . . msme, TE = 150 ms

170

APPENDIX C

171

APPENDIX D

FM Data Acquisition

Transillumination imaging White light, GFP, white light + GFP 1X, 2X and 4X

Series Scan Argon laser excitation of GFP, Red HeNe laser excitation of Alexa Acquire 26 frames preinjection, inject 100 l of ~0.0001 mmol/kg biotinBSAGdDTPAAlexa Washin: 300 frames as fast as possible, 1.92 s dwell time, scantime ≈ 5 min PostWashin: 30 frames with 38 sec fixed delay, scantime ≈ 19 min

Remove animal, collect blood in capillary tube, spin down using Hct microcentrifuge for 5 min

FM Data Acquisition

Red HeNe laser excitation of Alexa For blood : Acquire 2 frames as fast as possible at the appropriate gain to avoid saturation and at the gain used in the series scan

For phantoms (contrast agent suspended in PBS and raised in capillary tube same size as blood sample tube) : Image the concentration that has the same gain as the blood phantom at 2 gains: gain appropriate for that concentration and the gain used in the series scan

172

APPENDIX E

173

REFERENCES

1. Sherwood, L. Human Physiology, 4th edition, p. 360369. Brooks/Cole Publishing, 2001. 2. Patton, H. D., Fuchs, A. F., Hille, B., Scher, A. M., and Steiner, R. Textbook of Physiology: Circulation, Respiration, Body Fluids, Metabolism and Endocrinology, 21st edition. Philadelphia: W.B. Saunders Company, Harcourt Brace Jovanocivh, Inc., 1989. 3. Laking, G. R., West, C., Buckley, D. L., Matthews, J., and Price, P. M. Imaging vascular physiology to monitor cancer treatment. Crit Rev Oncol Hematol, 58: 95113, 2006. 4. Oh, P., Li, Y., Yu, J., Durr, E., Krasinska, K. M., Carver, L. A., Testa, J. E., and Schnitzer, J. E. Subtractive proteomic mapping of the endothelial surface in lung and solid tumours for tissuespecific therapy. [see comment]. Nature, 429: 629 635, 2004. 5. Durr, E., Yu, J., Krasinska, K. M., Carver, L. A., Yates, J. R., Testa, J. E., Oh, P., and Schnitzer, J. E. Direct proteomic mapping of the lung microvascular endothelial cell surface in vivo and in cell culture. Nature Biotechnology, 22: 985992, 2004. 6. Kety, S. S. The theory and applications of the exchange of inert gas at the lungs and tissues. Pharmacological Reviews, 3: 141, 1951. 7. Murray, P. D. F. The development in vitro of the early chick embryo. Strangeways Research Laboratory Cambridge : 497521, 1932. 8. Pardanaud, L., Luton, D., Prigent, M., Bourcheix, L. M., Catala, M., and DieterlenLievre, F. Two distinct endothelial lineages in ontogeny, one of them related to hemopoiesis. Development, 122: 13631371, 1996. 9. Palis, J., McGrath, K. E., and Kingsley, P. D. Initiation of hematopoiesis and vasculogenesis in murine yolk sac explants. Blood, 86: 156163, 1995. 10. Mills, K. R., Kruep, D., and Saha, M. S. Elucidating the origins of the vascular system: a fate map of the vascular endothelial and red blood cell lineages in Xenopus laevis. Developmental Biology, 209: 352368, 1999. 11. Risau, W. and Flamme, I. Vasculogenesis. [Review] [122 refs]. Annual Review of Cell & Developmental Biology, 11: 7391, 1995. 12. Ferguson, J. E., III, Kelley, R. W., and Patterson, C. Mechanisms of endothelial differentiation in embryonic vasculogenesis. [Review] [107 refs]. Arteriosclerosis, Thrombosis and Vascular Biology, 25: 22462254, 2005. 13. Tonnesen, M. G., Feng, X., and Clark, R. A. Angiogenesis in wound healing. [Review] [65 refs]. Journal of Investigative Dermatology, Symposium Proceedings. 5: 4046, 2000. 14. Reynolds, L. P., Killilea, S. D., and Redmer, D. A. Angiogenesis in the female reproductive system. [Review] [66 refs]. FASEB Journal, 6: 886892, 1992.

174

15. Li, W. W. Angiogenesis in wound healing. Contemporary Surgery, Supplement : 136, 2003. 16. Hanahan, D. and Weinberg, R. A. The hallmarks of cancer. [Review] [94 refs]. Cell, 100: 5770, 2000. 17. Schlatter, P., Konig, M. F., Karlsson, L. M., and Burri, P. H. Quantitative study of intussusceptive capillary growth in the chorioallantoic membrane (CAM) of the chicken embryo. Microvascular Research, 54: 6573, 1997. 18. Risau, W. Mechanisms of angiogenesis. [Review] [53 refs]. Nature, 386: 671 674, 1997. 19. Djonov, V., Schmid, M., Tschanz, S. A., and Burri, P. H. Intussusceptive angiogenesis: its role in embryonic vascular network formation. Circulation Research, 86: 286292, 2000. 20. Djonov, V., Baum, O., and Burri, P. H. Vascular remodeling by intussusceptive angiogenesis. [Review] [59 refs]. Cell and Tissue Research, 314: 107117, 2003. 21. Liekens, S., De Clercq, E., and Neyts, J. Angiogenesis: regulators and clinical applications. [Review] [225 refs]. Biochemical Pharmacology, 61: 253270, 2001. 22. Kalluri, R. Basement membranes: structure, assembly and role in tumour angiogenesis. [Review] [221 refs]. Nature Reviews, Cancer. 3: 422433, 2003. 23. Conway, E. M., Collen, D., and Carmeliet, P. Molecular mechanisms of blood vessel growth. [Review] [140 refs]. Cardiovascular Research, 49: 507521, 2001. 24. Carmeliet, P. Mechanisms of angiogenesis and arteriogenesis. [Review] [71 refs]. Nature Medicine, 6: 389395, 2000. 25. Thurston, G., Rudge, J. S., Ioffe, E., Zhou, H., Ross, L., Croll, S. D., Glazer, N., Holash, J., McDonald, D. M., and Yancopoulos, G. D. Angiopoietin1 protects the adult vasculature against plasma leakage. Nat Med, 6: 460463, 2000. 26. Lijnen, H. R. Extracellular proteolysis in the development and progression of atherosclerosis. [Review] [49 refs]. Biochemical Society Transactions, 30: 163 167, 2002. 27. Bischoff, J. Cell adhesion and angiogenesis. [Review] [27 refs]. Journal of Clinical Investigation, 100: S37S39, 1997. 28. Papetti, M. and Herman, I. M. Mechanisms of normal and tumorderived angiogenesis. [Review] [266 refs]. American Journal of Physiology Cell Physiology, 282: C947C970, 2002. 29. Sandison, J. C. A new method for the microscopic study of living growing tissues by introduction of a transparent chamber in the rabbit's ear. Anatomical Record, 28 281287, 1924. 30. Greenblatt, M. and Shubi, P. Tumor angiogenesis: transfilter diffusion studies in the hamster by the transparent chamber technique. Journal of the National Cancer Institute, 41: 111124, 1968. 31. Algire, G. H. and Chalkley, H. W. Vascular reactions of normal and malignant tissues in vivo. Journal of the National Cancer Institute, 3 7385, 1945. 32. Algire, G. H. An adaptation of the transparent chamber technique to the mouse. Journal of the National Cancer Institute, 4 111, 1943.

175

33. Yuan, F., Salehi, H. A., Boucher, Y., Vasthare, U. S., Tuma, R. F., and Jain, R. K. Vascular permeability and microcirculation of gliomas and mammary carcinomas transplanted in rat and mouse cranial windows. Cancer Research, 54: 45644568, 1994. 34. Dellian, M., Witwer, B. P., Salehi, H. A., Yuan, F., and Jain, R. K. Quantitation and physiological characterization of angiogenic vessels in mice: effect of basic fibroblast growth factor, vascular endothelial growth factor/vascular permeability factor, and host microenvironment.[see comment]. American Journal of Pathology, 149: 5971, 1996. 35. Auerbach, R., Akhtar, N., Lewis, R. L., and Shinners, B. L. Angiogenesis assays: problems and pitfalls. [Review] [30 refs]. Cancer and Metastasis Reviews, 19: 167172, 2000. 36. Heuser, L. S. and Miller, F. N. Differential macromolecular leakage from the vasculature of tumors. Cancer, 57: 461464, 1986. 37. Zeidman, I. The fate of circulating tumors cells. I. Passage of cells through capillaries. Cancer Research, 21: 3839, 1961. 38. Jain, R. K., Schlenger, K., Hockel, M., and Yuan, F. Quantitative angiogenesis assays: progress and problems. [Review] [82 refs]. Nature Medicine, 3: 1203 1208, 1997. 39. Nguyen, M., Shing, Y., and Folkman, J. Quantitation of angiogenesis and antiangiogenesis in the chick embryo chorioallantoic membrane. Microvascular Research, 47: 3140, 1994. 40. Auerbach, R., Arensman, R., Kubai, L., and Folkman, J. Tumorinduced angiogenesis: lack of inhibition by irradiation. International Journal of Cancer, 15: 241245, 1975. 41. Gimbrone, M. A., Jr., Leapman, S. B., Cotran, R. S., and Folkman, J. Tumor dormancy in vivo by prevention of neovascularization. Journal of Experimental Medicine, 136: 261276, 1972. 42. Langer, R., Brem, H., Falterman, K., Klein, M., and Folkman, J. Isolations of a cartilage factor that inhibits tumor neovascularization. Science, 193: 7072, 1976. 43. Akhtar, N., Dickerson, E. B., and Auerbach, R. The sponge/Matrigel angiogenesis assay. Angiogenesis, 5: 7580, 2002. 44. Passaniti, A., Taylor, R. M., Pili, R., Guo, Y., Long, P. V., Haney, J. A., Pauly, R. R., Grant, D. S., and Martin, G. R. A simple, quantitative method for assessing angiogenesis and antiangiogenic agents using reconstituted basement membrane, heparin, and fibroblast growth factor. Laboratory Investigation, 67: 519528, 1992. 45. Weidner, N., Semple, J. P., Welch, W. R., and Folkman, J. Tumor angiogenesis and metastasiscorrelation in invasive breast carcinoma. New England Journal of Medicine, 324: 18, 1991. 46. Srivastava, A., Laidler, P., Hughes, L. E., Woodcock, J., and Shedden, E. J. Neovascularization in human cutaneous melanoma: a quantitative morphological and Doppler ultrasound study. European Journal of Cancer & Clinical Oncology, 22: 12051209, 1986.

176

47. Louvar, E., Littrup, P. J., Goldstein, A., Yu, L., Sakr, W., and Grignon, D. Correlation of color Doppler flow in the prostate with tissue microvascularity. Cancer, 83: 135140, 1998. 48. Cosgrove, D. O., Bamber, J. C., Davey, J. B., McKinna, J. A., and Sinnett, H. D. Color Doppler signals from breast tumors. Work in progress. Radiology, 176: 175180, 1990. 49. Adler, D. D., Carson, P. L., Rubin, J. M., and QuinnReid, D. Doppler ultrasound color flow imaging in the study of breast cancer: preliminary findings. Ultrasound in Medicine and Biology, 16: 553559, 1990. 50. Koch, A. E. Review: angiogenesis: implications for rheumatoid arthritis. [Review] [104 refs]. Arthritis & Rheumatism, 41: 951962, 1998. 51. Sebag, J. and McMeel, J. W. Diabetic retinopathy. Pathogenesis and the role of retinaderived growth factor in angiogenesis. [Review] [86 refs]. Survey of Ophthalmology, 30: 377384, 1986. 52. Creamer, D., Sullivan, D., Bicknell, R., and Barker, J. Angiogenesis in psoriasis. [Review] [37 refs]. Angiogenesis, 5: 231236, 2002. 53. Majno, G. Chronic inflammation: links with angiogenesis and wound healing. [comment] [Review] [27 refs]. American Journal of Pathology, 153: 10351039, 1998. 54. Ryschich, E., Schmidt, J., Hammerling, G. J., Klar, E., and Ganss, R. Transformation of the microvascular system during multistage tumorigenesis. International Journal of Cancer, 97: 719725, 2002. 55. Holash, J., Wiegand, S. J., and Yancopoulos, G. D. New model of tumor angiogenesis: dynamic balance between vessel regression and growth mediated by angiopoietins and VEGF. [Review] [36 refs]. Oncogene, 18: 53565362, 1999. 56. Folkman, J., Watson, K., Ingber, D., and Hanahan, D. Induction of angiogenesis during the transition from hyperplasia to neoplasia. Nature, 339: 5861, 1989. 57. Carmeliet, P. and Jain, R. K. Angiogenesis in cancer and other diseases. [Review] [75 refs]. Nature, 407: 249257, 2000. 58. Weidner, N. Tumour vascularity and proliferation: clear evidence of a close relationship. [comment]. Journal of Pathology, 189: 297299, 1999. 59. Fox, S. B. Tumour angiogenesis and prognosis. [see comment] [Review] [112 refs]. Histopathology, 30: 294301, 1997. 60. Gatenby, R. A. and Gillies, R. J. Why do cancers have high aerobic glycolysis? [Review] [96 refs]. Nature Reviews, Cancer. 4: 891899, 2004. 61. Krogh, A. The number and distribution of capillaries in muscles with calculations of the oxygen pressure head necessary for supplying the tissue. Journal of Physiology, 52: 409415, 1919. 62. Cho, Z. H., Jones, J. P., and Singh, M. Foundations of Medical Imaging, 1st edition. New York: John Wiley & Sons, Inc., 1993. 63. Yamaguchi, A., Taniguchi, H., Kunishima, S., Koh, T., and Yamagishi, H. Correlation between angiographically assessed vascularity and blood flow in hepatic metastases in patients with colorectal carcinoma. Cancer, 89: 12361244, 2000.

177

64. Wilson, C. B., Lammertsma, A. A., McKenzie, C. G., Sikora, K., and Jones, T. Measurements of blood flow and exchanging water space in breast tumors using positron emission tomography: a rapid and noninvasive dynamic method. Cancer Research, 52: 15921597, 1992. 65. Hoekstra, C. J., Stroobants, S. G., Hoekstra, O. S., Smit, E. F., Vansteenkiste, J. F., and Lammertsma, A. A. Measurement of perfusion in stage IIIAN2 nonsmall cell lung cancer using H(2)(15)O and positron emission tomography. Clinical Cancer Research, 8: 21092115, 2002. 66. Wells, P., Jones, T., and Price, P. Assessment of inter and intrapatient variability in C15O2 positron emission tomography measurements of blood flow in patients with intraabdominal cancers. Clinical Cancer Research, 9: 63506356, 2003. 67. Anderson, H. L., Yap, J. T., Miller, M. P., Robbins, A., Jones, T., and Price, P. M. Assessment of pharmacodynamic vascular response in a phase I trial of combretastatin A4 phosphate.[see comment]. Journal of Clinical Oncology, 21: 28232830, 2003. 68. Anderson, H., Yap, J. T., Wells, P., Miller, M. P., Propper, D., Price, P., and Harris, A. L. Measurement of renal tumour and normal tissue perfusion using positron emission tomography in a phase II clinical trial of razoxane. British Journal of Cancer, 89: 262267, 2003. 69. Lara, P. N., Jr., Quinn, D. I., Margolin, K., Meyers, F. J., Longmate, J., Frankel, P., Mack, P. C., Turrell, C., Valk, P., Rao, J., Buckley, P., Wun, T., Gosselin, R., Galvin, I., Gumerlock, P. H., Lenz, H. J., Doroshow, J. H., Gandara, D. R., and California Cancer, C. SU5416 plus interferon alpha in advanced renal cell carcinoma: a phase II California Cancer Consortium Study with biological and imaging correlates of angiogenesis inhibition. Clinical Cancer Research, 9: 4772 4781, 2003. 70. Miller, K. D., Soule, S. E., Calley, C., Emerson, R. E., Hutchins, G. D., Kopecky, K., Badve, S., Storniolo, A., Goulet, R., and Sledge, G. W., Jr. Randomized phase II trial of the antiangiogenic potential of doxorubicin and docetaxel; primary chemotherapy as Biomarker Discovery Laboratory. Breast Cancer Research and Treatment, 89: 187197, 2005. 71. Gupta, N., Saleem, A., Kotz, B., Osman, S., Aboagye, E. O., Phillips, R., Vernon, C., Wasan, H., Jones, T., Hoskin, P. J., and Price, P. M. Carbogen and nicotinamide increase blood flow and 5fluorouracil delivery but not 5 fluorouracil retention in colorectal cancer metastases in patients. Clinical Cancer Research, 12: 31153123, 2006. 72. Bruehlmeier, M., Roelcke, U., Schubiger, P. A., and Ametamey, S. M. Assessment of hypoxia and perfusion in human brain tumors using PET with 18F fluoromisonidazole and 15OH2O. Journal of Nuclear Medicine, 45: 18511859, 2004. 73. Stokely, E. M., Sveinsdottir, E., Lassen, N. A., and Rommer, P. A single photon dynamic computer assisted tomograph (DCAT) for imaging brain function in multiple cross sections. Journal of Computer Assisted Tomography, 4: 230240, 1980.

178

74. Namba, H., Yanagisawa, M., Yui, N., Togawa, T., Kinoshita, F., Iwadate, Y., and Sueyoshi, K. Quantifying brain tumor blood flow by the microsphere model with Nisopropylp[123I]iodoamphetamine superearly SPECT. Annals of Nuclear Medicine, 10: 161164, 1996. 75. Andrews, J. C., WalkerAndrews, S. C., Juni, J. E., Warber, S., and Ensminger, W. D. Modulation of liver tumor blood flow with hepatic arterial epinephrine: a SPECT study. Radiology, 173: 645647, 1989. 76. Iannotti, F. Functional imaging of blood brain barrier permeability by single photon emission computerised tomography and positron emission tomography. [Review] [51 refs]. Advances & Technical Standards in Neurosurgery, 19: 103 119, 1992. 77. Seo, Y., Fukuoka, S., Nakagawara, J., Takanashi, M., Suematsu, K., and Nakamura, J. Early effects of gamma knife radiosurgery on brain metastases: assessment by 201TlCl SPECT and 99mTcDTPAhuman serum albumin SPECT. Neurologia MedicoChirurgica, 37: 2530, 1997. 78. Ishiyama, K., Tomura, N., Okada, K., Nagasawa, H., Sashi, R., Sasaki, K., Sato, K., and Watarai, J. Evaluating benign and malignant musculoskeletal lesions with radionuclide angiography and SPECT using Tc99m MIBI. Clinical Nuclear Medicine, 30: 598603, 2005. 79. Groshar, D., McEwan, A. J., Parliament, M. B., Urtasun, R. C., Golberg, L. E., Hoskinson, M., Mercer, J. R., Mannan, R. H., Wiebe, L. I., and Chapman, J. D. Imaging tumor hypoxia and tumor perfusion. Journal of Nuclear Medicine, 34: 885888, 1993. 80. Ohno, K., Pettigrew, K. D., and Rapoport, S. I. Lower limits of cerebrovascular permeability to nonelectrolytes in the conscious rat. American Journal of Physiology, 235: H299H307, 1978. 81. Sahani, D. V., Kalva, S. P., Hamberg, L. M., Hahn, P. F., Willett, C. G., Saini, S., Mueller, P. R., and Lee, T. Y. Assessing tumor perfusion and treatment response in rectal cancer with multisection CT: initial observations. Radiology, 234: 785 792, 2005. 82. Bisdas, S., Baghi, M., Smolarz, A., Pihno, N. C., Lehnert, T., Knecht, R., Mack, M. G., Vogl, T. J., Tuerkay, S., and Koh, T. S. Quantitative measurements of perfusion and permeability of oropharyngeal and oral cavity cancer, recurrent disease, and associated lymph nodes using firstpass contrastenhanced computed tomography studies. Investigative Radiology, 42: 172179, 2007. 83. Gandhi, D., Hoeffner, E. G., Carlos, R. C., Case, I., and Mukherji, S. K. Computed tomography perfusion of squamous cell carcinoma of the upper aerodigestive tract. Initial results. Journal of Computer Assisted Tomography, 27: 687693, 2003. 84. Miles, K. A., Leggett, D. A., Kelley, B. B., Hayball, M. P., Sinnatamby, R., and Bunce, I. In vivo assessment of neovascularization of liver metastases using perfusion CT. [see comment]. British Journal of Radiology, 71: 276281, 1998. 85. Dugdale, P. E., Miles, K. A., Bunce, I., Kelley, B. B., and Leggett, D. A. CT measurement of perfusion and permeability within lymphoma masses and its

179

ability to assess grade, activity, and chemotherapeutic response. Journal of Computer Assisted Tomography, 23: 540547, 1999. 86. Ng, Q. S., Goh, V., Fichte, H., Klotz, E., Fernie, P., Saunders, M. I., Hoskin, P. J., and Padhani, A. R. Lung cancer perfusion at multidetector row CT: reproducibility of whole tumor quantitative measurements. Radiology, 239: 547 553, 2006. 87. Patlak, C. S., Blasberg, R. G., and Fenstermacher, J. D. Graphical evaluation of bloodtobrain transfer constants from multipletime uptake data. J Cereb Blood Flow Metab, 3: 17, 1983. 88. Goh, V., Halligan, S., Hugill, J. A., Bassett, P., and Bartram, C. I. Quantitative assessment of colorectal cancer perfusion using MDCT: inter and intraobserver agreement. AJR, American Journal of Roentgenology. 185: 225231, 2005. 89. Goh, V., Halligan, S., Hugill, J. A., Gartner, L., Bartram, C. I., Goh, V., Halligan, S., Hugill, J. A., Gartner, L., and Bartram, C. I. Quantitative colorectal cancer perfusion measurement using dynamic contrastenhanced multidetectorrow computed tomography: effect of acquisition time and implications for protocols. Journal of Computer Assisted Tomography, 29: 5963, 2005. 90. Koukourakis, M. I., Mavanis, I., Kouklakis, G., Pitiakoudis, M., Minopoulos, G., Manolas, C., and Simopoulos, C. Early antivascular effects of bevacizumab anti VEGF monoclonal antibody on colorectal carcinomas assessed with functional CT imaging. American Journal of Clinical Oncology, 30: 315318, 2007. 91. McNeel, D. G., Eickhoff, J., Lee, F. T., King, D. M., Alberti, D., Thomas, J. P., Friedl, A., Kolesar, J., Marnocha, R., Volkman, J., Zhang, J., Hammershaimb, L., Zwiebel, J. A., and Wilding, G. Phase I trial of a monoclonal antibody specific for alphavbeta3 integrin (MEDI522) in patients with advanced malignancies, including an assessment of effect on tumor perfusion. Clinical Cancer Research, 11: 78517860, 2005. 92. Johnson, J. A. and Wilson, T. A. A model for capillary exchange. American Journal of Physiology, 210: 12991303, 1966. 93. St. Lawrence, K. S. and Lee, T. Y. An adiabatic approximation to the tissue homogeneity model for water exchange in the brain: I. Theoretical derivation. Journal of Cerebral Blood Flow and Metabolism, 18: 13651377, 1998. 94. Clark, J. W., Neuman, M. R., Olson, W. H., Peura, R. A., Primiano, F. P., Siedband, M. P., Webster, J. W., Wheeler, L. A., and John, G. W. Medical Instrumentation, 3rd edition.: John Wiley & Sons, 1998. 95. Tanaka, S., Kitamura, T., Fujita, M., Nakanishi, K., and Okuda, S. Color Doppler flow imaging of liver tumors. AJR, American Journal of Roentgenology. 154: 509514, 1990. 96. Bourne, T., Campbell, S., Steer, C., Whitehead, M. I., and Collins, W. P. Transvaginal colour flow imaging: a possible new screening technique for ovarian cancer.[see comment]. BMJ, 299: 13671370, 1989. 97. Rubin, J. M., Bude, R. O., Carson, P. L., Bree, R. L., and Adler, R. S. Power Doppler US: a potentially useful alternative to mean frequencybased color Doppler US. Radiology, 190: 853856, 1994.

180

98. Gramiak, R. and Shah, P. M. Echocardiography of the aortic root. Investigative Radiology, 3: 356366, 1968. 99. Blomley, M. J., Cooke, J. C., Unger, E. C., Monaghan, M. J., and Cosgrove, D. O. Microbubble contrast agents: a new era in ultrasound. [Review] [30 refs]. BMJ, 322: 12221225, 2001. 100. Lamuraglia, M., Lassau, N., Garbay, J. R., Mathieu, M. C., Rouzier, R., Jaziri, S., Roche, A., and Leclere, J. Doppler US with perfusion software and contrast medium injection in the early evaluation of radiofrequency in breast cancer recurrences: a prospective phase II study. European Journal of Radiology, 56: 376381, 2005. 101. Chaudhari, M. H., Forsberg, F., Voodarla, A., Saikali, F. N., Goonewardene, S., Needleman, L., Finkel, G. C., and Goldberg, B. B. Breast tumor vascularity identified by contrast enhanced ultrasound and pathology: initial results. Ultrasonics, 38: 105109, 2000. 102. Piscaglia, F., Corradi, F., Mancini, M., Giangregorio, F., Tamberi, S., Ugolini, G., Cola, B., Bazzocchi, A., Righini, R., Pini, P., Fornari, F., and Bolondi, L. Real time contrast enhanced ultrasonography in detection of liver metastases from gastrointestinal cancer. BMC Cancer, 7: 171, 2007. 103. Yang, W., Chen, M. H., Yan, K., Wu, W., Dai, Y., and Zhang, H. Differential diagnosis of nonfunctional islet cell tumor and pancreatic carcinoma with sonography. European Journal of Radiology, 62: 342351, 2007. 104. Bertolotto, M., Pozzato, G., Croce, L. S., Nascimben, F., Gasparini, C., Cova, M. A., and Tiribelli, C. Blood flow changes in hepatocellular carcinoma after the administration of thalidomide assessed by reperfusion kinetics during microbubble infusion: preliminary results. Investigative Radiology, 41: 1521, 2006. 105. Hertz, C. H. Fourth European Congress on Ultrasound in Medicine, Dubrovnik, 1981. 106. Dymling, S., Persson, H. W., and Hertz, C. H. The measurement of blood perfusion in tissue. Fifth World Congress of Ultrasound in Medicine and Biology, Brighton, UK, 1982. 107. Wei, K., Jayaweera, A. R., Firoozan, S., Linka, A., Skyba, D. M., and Kaul, S. Quantification of myocardial blood flow with ultrasoundinduced destruction of microbubbles administered as a constant venous infusion. Circulation, 97: 473 483, 1998. 108. Unger, E. C., Matsunaga, T. O., McCreery, T., Schumann, P., Sweitzer, R., and Quigley, R. Therapeutic applications of microbubbles. [Review] [17 refs]. European Journal of Radiology, 42: 160168, 2002. 109. Unger, E. C., Hersh, E., Vannan, M., Matsunaga, T. O., and McCreery, T. Local drug and gene delivery through microbubbles. [Review] [50 refs]. Progress in Cardiovascular Diseases, 44: 4554, 2001. 110. Borden, M. A., Martinez, G. V., Ricker, J., Tsvetkova, N., Longo, M., Gillies, R. J., Dayton, P. A., and Ferrara, K. W. Lateral phase separation in lipidcoated microbubbles. Langmuir, 22: 42914297, 2006.

181

111. Borden, M. A., Caskey, C. F., Little, E., Gillies, R. J., and Ferrara, K. W. DNA and polylysine adsorption and multilayer construction onto cationic lipidcoated microbubbles. Langmuir, 23: 94019408, 2007. 112. Sokolov, K., Aaron, J., Hsu, B., Nida, D., Gillenwater, A., Follen, M., MacAulay, C., AdlerStorthz, K., Korgel, B., Descour, M., Pasqualini, R., Arap, W., Lam, W., and RichardsKortum, R. Optical systems for in vivo molecular imaging of cancer. [Review] [112 refs]. Technology in Cancer Research & Treatment, 2: 491504, 2003. 113. McDevitt, H. O. and Coons, A. H. Methods for preparation of fluorescent proteins. Methods in Medical Research, 10: 142148, 1964. 114. Coons, A. H., Creech, H. J., and Jones, R. N. Immunological properties of an antibody containing a fluorescent group Proceedings of The Society for Experimental Biology and Medicine, 42: 200202, 1941. 115. Murphy, D. B. Fundamentals of Light Microscopy and Electronic Imaging, 1st edition. New York, NY: John Wiley & Sons, Inc., 2001. 116. Weissleder, R. and Mahmood, U. Molecular imaging. [Review] [141 refs]. Radiology, 219: 316333, 2001. 117. Sadikot, R. T. and Blackwell, T. S. Bioluminescence imaging. [Review] [43 refs]. Proceedings of the American Thoracic Society, 2: 537540, 2005. 118. http://clinicaltrials.gov/ . 119. Ntziachristos, V. and Chance, B. Probing physiology and molecular function using optical imaging: applications to breast cancer. [Review] [29 refs]. Breast Cancer Research, 3: 4146, 2001. 120. Ntziachristos, V., Yodh, A. G., Schnall, M., and Chance, B. Concurrent MRI and diffuse optical tomography of breast after indocyanine green enhancement. Proceedings of the National Academy of Sciences of the United States of America, 97: 27672772, 2000. 121. Brown, E. B., Campbell, R. B., Tsuzuki, Y., Xu, L., Carmeliet, P., Fukumura, D., and Jain, R. K. In vivo measurement of gene expression, angiogenesis and physiological function in tumors using multiphoton laser scanning microscopy.[erratum appears in Nat Med 2001 Sep;7(9):1069]. Nature Medicine, 7: 864868, 2001. 122. Tozer, G. M., Prise, V. E., Wilson, J., Cemazar, M., Shan, S., Dewhirst, M. W., Barber, P. R., Vojnovic, B., and Chaplin, D. J. Mechanisms associated with tumor vascular shutdown induced by combretastatin A4 phosphate: intravital microscopy and measurement of vascular permeability. Cancer Research, 61: 64136422, 2001. 123. Wu, N. Z., Klitzman, B., Rosner, G., Needham, D., and Dewhirst, M. W. Measurement of material extravasation in microvascular networks using fluorescence videomicroscopy. Microvascular Research, 46: 231253, 1993. 124. Damadian, R. Tumor detection by nuclear magnetic resonance. Science, 171: 11511153, 1971. 125. Lauterbur, P. C. Image formation by induced local interactions. Examples employing nuclear magnetic resonance. 1973. Clin Orthop Relat Res : 36, 1989.

182

126. Garroway, A. N., Grannell, P. K., and Mansfield, P. Image formation in NMR by a selective irradiative process Journal of Physics C: Solid State Physics 7: L457L462, 1974. 127. Kumar, A., Welti, D., and Ernst, R. R. NMR Fourier zeugmatography. Journal of Magnetic Resonance, 18: 6983, 1975. 128. Damadian, R., Goldsmith, M., and Minkoff, L. NMR in cancer: XVI. FONAR image of the live human body. Physiol Chem Phys, 9: 97100, 108, 1977. 129. Damadian, R., Goldsmith, M., and Minkoff, L. NMR in cancer: XX. FONAR scans of patients with cancer. Physiol Chem Phys, 10: 285287, 1978. 130. Haacke, E., Brown, R., Thompson, M., and Venkatesan, R. Magnetic Resonance Imaging: Physcial principles and sequence design. New York: WileyLiss; John Wiley & Sons, Inc. , 1999. 131. Hoang, B. H., Dyke, J. P., Koutcher, J. A., Huvos, A. G., Mizobuchi, H., Mazza, B. A., Gorlick, R., and Healey, J. H. VEGF expression in osteosarcoma correlates with vascular permeability by dynamic MRI. Clinical Orthopaedics and Related Research : 3238, 2004. 132. Rijpkema, M., Kaanders, J. H., Joosten, F. B., van der Kogel, A. J., and Heerschap, A. Method for quantitative mapping of dynamic MRI contrast agent uptake in human tumors. Journal of Magnetic Resonance Imaging, 14: 457463, 2001. 133. Hayes, C., Padhani, A. R., and Leach, M. O. Assessing changes in tumour vascular function using dynamic contrastenhanced magnetic resonance imaging. NMR in Biomedicine, 15: 154163, 2002. 134. Li, K. L., Zhu, X. P., Checkley, D. R., Tessier, J. J., Hillier, V. F., Waterton, J. C., and Jackson, A. Simultaneous mapping of blood volume and endothelial permeability surface area product in gliomas using iterative analysis of firstpass dynamic contrast enhanced MRI data. British Journal of Radiology, 76: 3950, 2003. 135. Knopp, M. V., Weiss, E., Sinn, H. P., Mattern, J., Junkermann, H., Radeleff, J., Magener, A., Brix, G., Delorme, S., Zuna, I., and van Kaick, G. Pathophysiologic basis of contrast enhancement in breast tumors. Journal of Magnetic Resonance Imaging, 10: 260266, 1999. 136. Turetschek, K., Preda, A., Novikov, V., Brasch, R. C., Weinmann, H. J., Wunderbaldinger, P., and Roberts, T. P. Tumor microvascular changes in antiangiogenic treatment: assessment by magnetic resonance contrast media of different molecular weights. Journal of Magnetic Resonance Imaging, 20: 138 144, 2004. 137. Raatschen, H. J., Fu, Y., Shames, D. M., Wendland, M. F., and Brasch, R. C. Magnetic resonance imaging enhancement of normal tissues and tumors using macromolecular Gdbased cascade polymer contrast agents: preclinical evaluations. Investigative Radiology, 41: 860867, 2006. 138. Jordan, B. F., Runquist, M., Raghunand, N., Gillies, R. J., Tate, W. R., Powis, G., and Baker, A. F. The thioredoxin1 inhibitor 1methylpropyl 2imidazolyl disulfide (PX12) decreases vascular permeability in tumor xenografts monitored

183

by dynamic contrast enhanced magnetic resonance imaging. Clinical Cancer Research, 11: 529536, 2005. 139. Gillies, R. J., Raghunand, N., Karczmar, G. S., and Bhujwalla, Z. M. MRI of the tumor microenvironment. [erratum appears in J Magn Reson Imaging 2002 Dec;16(6):751] [Review] [225 refs]. Journal of Magnetic Resonance Imaging, 16: 430450, 2002. 140. Tofts, P. S., Brix, G., Buckley, D. L., Evelhoch, J. L., Henderson, E., Knopp, M. V., Larsson, H. B., Lee, T. Y., Mayr, N. A., Parker, G. J., Port, R. E., Taylor, J., and Weisskoff, R. M. Estimating kinetic parameters from dynamic contrast enhanced T(1)weighted MRI of a diffusable tracer: standardized quantities and symbols. [Review] [50 refs]. Journal of Magnetic Resonance Imaging, 10: 223 232, 1999. 141. Landis, C. S., Li, X., Telang, F. W., Molina, P. E., Palyka, I., Vetek, G., and Springer, C. S., Jr. Equilibrium transcytolemmal waterexchange kinetics in skeletal muscle in vivo.[see comment]. Magnetic Resonance in Medicine, 42: 467478, 1999. 142. Yang, C., Karczmar, G. S., Medved, M., and Stadler, W. M. Multiple reference tissue method for contrast agent arterial input function estimation. Magn Reson Med, 58: 12661275, 2007. 143. Kim, S., Quon, H., Loevner, L. A., Rosen, M. A., Dougherty, L., Kilger, A. M., Glickson, J. D., and Poptani, H. Transcytolemmal water exchange in pharmacokinetic analysis of dynamic contrastenhanced MRI data in squamous cell carcinoma of the head and neck. J Magn Reson Imaging, 26: 16071617, 2007. 144. Faranesh, A. Z., Kraitchman, D. L., and McVeigh, E. R. Measurement of kinetic parameters in skeletal muscle by magnetic resonance imaging with an intravascular agent. Magn Reson Med, 55: 11141123, 2006. 145. Eyal, E. and Degani, H. Modelbased and modelfree parametric analysis of breast dynamiccontrastenhanced MRI. NMR Biomed, 2007. 146. Yankeelov, T. E., Rooney, W. D., Li, X., and Springer, C. S., Jr. Variation of the relaxographic "shutterspeed" for transcytolemmal water exchange affects the CR bolustracking curve shape. Magnetic Resonance in Medicine, 50: 11511169, 2003. 147. McConnell, H. M. Reaction Rates by Nuclear Magnetic Resonance. Journal of Chemical Physics, 28: 430431, 1958. 148. Woessner, D. E. Nuclear Transfer Effects in Nuclear Magnetic Resonance Pulse Experiments. Journal of Chemical Physics, 35 4148, 1961. 149. Leach, M. O., Brindle, K. M., Evelhoch, J. L., Griffiths, J. R., Horsman, M. R., Jackson, A., Jayson, G., Judson, I. R., Knopp, M. V., Maxwell, R. J., McIntyre, D., Padhani, A. R., Price, P., Rathbone, R., Rustin, G., Tofts, P. S., Tozer, G. M., Vennart, W., Waterton, J. C., Williams, S. R., Workman, P., and Workman, P. Assessment of antiangiogenic and antivascular therapeutics using MRI: recommendations for appropriate methodology for clinical trials. [Review] [10 refs]. British Journal of Radiology, 76 Spec No 1: S87S91, 2003.

184

150. Tandle, A., Blazer Iii, D. G., and Libutti, S. K. Antiangiogenic gene therapy of cancer: recent developments. Journal of Translational Medicine. 2006. 151. Palade, G. E. Transport in quanta across the endothelium of blood capillaries. Anatomical Record, 136: 254, 1960. 152. Kohn, S., Nagy, J. A., Dvorak, H. F., and Dvorak, A. M. Pathways of macromolecular tracer transport across venules and small veins. Structural basis for the hyperpermeability of tumor blood vessels. Laboratory Investigation, 67: 596607, 1992. 153. Bendayan, M. and Rasio, E. A. Evidence of a tubular system for transendothelial transport in arterial capillaries of the rete mirabile. Journal of Histochemistry and Cytochemistry, 45: 13651378, 1997. 154. Simionescu, M. and Simionescu, N. Endothelial transport of macromolecules: transcytosis and endocytosis. A look from cell biology. [Review] [407 refs]. Cell Biology Reviews, 25: 578, 1991. 155. Simionescu, M., Gafencu, A., and Antohe, F. Transcytosis of plasma macromolecules in endothelial cells: a cell biological survey. [Review] [60 refs]. Microscopy Research and Technique, 57: 269288, 2002. 156. Mehta, D. and Malik, A. B. Signaling mechanisms regulating endothelial permeability. [Review] [1112 refs]. Physiological Reviews, 86: 279367, 2006. 157. Bruns, R. R. and Palade, G. E. Studies on blood capillaries. II. Transport of ferritin molecules across the wall of muscle capillaries. Journal of Cell Biology, 37: 277299, 1968. 158. Bruns, R. R. and Palade, G. E. Studies on blood capillaries. I. General organization of blood capillaries in muscle. Journal of Cell Biology, 37: 244276, 1968. 159. Okamoto, T., Schlegel, A., Scherer, P. E., and Lisanti, M. P. Caveolins, a family of scaffolding proteins for organizing "preassembled signaling complexes" at the plasma membrane. [Review] [72 refs]. Journal of Biological Chemistry, 273: 54195422, 1998. 160. Simionescu, M., Simionescu, N., and Palade, G. E. Segmental differentiations of cell junctions in the vascular endothelium. The microvasculature. Journal of Cell Biology, 67: 863885, 1975. 161. Dvorak, A. M., Kohn, S., Morgan, E. S., Fox, P., Nagy, J. A., and Dvorak, H. F. The vesiculovacuolar organelle (VVO): a distinct endothelial cell structure that provides a transcellular pathway for macromolecular extravasation. Journal of Leukocyte Biology, 59: 100115, 1996. 162. Kumar, P. and Timoney, J. F. Histology, immunohistochemistry and ultrastructure of the equine palatine tonsil. Anatomia, Histologia, Embryologia: Veterinary Medicine Series C, 34: 192198, 2005. 163. Kumar, P. and Timoney, J. F. Histology and ultrastructure of the equine lingual tonsil. II. Lymphoid tissue and associated high endothelial venules. Anatomia, Histologia, Embryologia: Veterinary Medicine Series C, 34: 98104, 2005.

185

164. Caruso, R. A., Speciale, G., Inferrera, A., Rigoli, L., and Inferrera, C. Ultrastructural observations on the microvasculature in advanced gastric carcinomas. Histology and Histopathology, 16: 785792, 2001. 165. Feng, D., Nagy, J. A., Hipp, J., Dvorak, H. F., and Dvorak, A. M. Vesiculo vacuolar organelles and the regulation of venule permeability to macromolecules by vascular permeability factor, histamine, and serotonin. Journal of Experimental Medicine, 183: 19811986, 1996. 166. Minshall, R. D., Sessa, W. C., Stan, R. V., Anderson, R. G., and Malik, A. B. Caveolin regulation of endothelial function. [Review] [57 refs]. American Journal of Physiology Lung Cellular and Molecular Physiology, 285: L1179L1183, 2003. 167. Feng, D., Nagy, J. A., Dvorak, H. F., and Dvorak, A. M. Ultrastructural studies define soluble macromolecular, particulate, and cellular transendothelial cell pathways in venules, lymphatic vessels, and tumorassociated microvessels in man and animals. [Review] [150 refs]. Microscopy Research and Technique, 57: 289326, 2002. 168. Hormigo, A., Gutin, P. H., and Rafii, S. Tracking normalization of brain tumor vasculature by magnetic imaging and proangiogenic biomarkers. [comment] [Review] [13 refs]. Cancer Cell, 11: 68, 2007. 169. Batchelor, T. T., Sorensen, A. G., di Tomaso, E., Zhang, W. T., Duda, D. G., Cohen, K. S., Kozak, K. R., Cahill, D. P., Chen, P. J., Zhu, M., Ancukiewicz, M., Mrugala, M. M., Plotkin, S., Drappatz, J., Louis, D. N., Ivy, P., Scadden, D. T., Benner, T., Loeffler, J. S., Wen, P. Y., and Jain, R. K. AZD2171, a panVEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients.[see comment]. Cancer Cell, 11: 8395, 2007. 170. Marzola, P., Degrassi, A., Calderan, L., Farace, P., Nicolato, E., Crescimanno, C., Sandri, M., Giusti, A., Pesenti, E., Terron, A., Sbarbati, A., and Osculati, F. Early antiangiogenic activity of SU11248 evaluated in vivo by dynamic contrast enhanced magnetic resonance imaging in an experimental model of colon carcinoma. Clinical Cancer Research, 11: 58275832, 2005. 171. Liu, G., Rugo, H. S., Wilding, G., McShane, T. M., Evelhoch, J. L., Ng, C., Jackson, E., Kelcz, F., Yeh, B. M., Lee, F. T., Jr., Charnsangavej, C., Park, J. W., Ashton, E. A., Steinfeldt, H. M., Pithavala, Y. K., Reich, S. D., Herbst, R. S., Liu, G., Rugo, H. S., Wilding, G., McShane, T. M., Evelhoch, J. L., Ng, C., Jackson, E., Kelcz, F., Yeh, B. M., Lee, F. T. J., Charnsangavej, C., Park, J. W., Ashton, E. A., Steinfeldt, H. M., Pithavala, Y. K., Reich, S. D., and Herbst, R. S. Dynamic contrastenhanced magnetic resonance imaging as a pharmacodynamic measure of response after acute dosing of AG013736, an oral angiogenesis inhibitor, in patients with advanced solid tumors: results from a phase I study. [see comment]. Journal of Clinical Oncology, 23: 54645473, 2005. 172. Thomas, A. L., Morgan, B., Horsfield, M. A., Higginson, A., Kay, A., Lee, L., Masson, E., PuccioPick, M., Laurent, D., and Steward, W. P. Phase I study of the safety, tolerability, pharmacokinetics, and pharmacodynamics of PTK787/ZK

186

222584 administered twice daily in patients with advanced cancer. Journal of Clinical Oncology, 23: 41624171, 2005. 173. Morgan, B., Thomas, A. L., Drevs, J., Hennig, J., Buchert, M., Jivan, A., Horsfield, M. A., Mross, K., Ball, H. A., Lee, L., Mietlowski, W., Fuxuis, S., Unger, C., O'Byrne, K., Henry, A., Cherryman, G. R., Laurent, D., Dugan, M., Marme, D., Steward, W. P., and Steward, W. P. Dynamic contrastenhanced magnetic resonance imaging as a biomarker for the pharmacological response of PTK787/ZK 222584, an inhibitor of the vascular endothelial growth factor receptor tyrosine kinases, in patients with advanced colorectal cancer and liver metastases: results from two phase I studies. [see comment]. Journal of Clinical Oncology, 21: 39553964, 2003. 174. Traxler, P., Allegrini, P. R., Brandt, R., Brueggen, J., Cozens, R., Fabbro, D., Grosios, K., Lane, H. A., McSheehy, P., Mestan, J., Meyer, T., Tang, C., Wartmann, M., Wood, J., and Caravatti, G. AEE788: a dual family epidermal growth factor receptor/ErbB2 and vascular endothelial growth factor receptor tyrosine kinase inhibitor with antitumor and antiangiogenic activity. Cancer Research, 64: 49314941, 2004. 175. Checkley, D., Tessier, J. J., Kendrew, J., Waterton, J. C., and Wedge, S. R. Use of dynamic contrastenhanced MRI to evaluate acute treatment with ZD6474, a VEGF signalling inhibitor, in PC3 prostate tumours. British Journal of Cancer, 89: 18891895, 2003. 176. Jayson, G. C., Zweit, J., Jackson, A., Mulatero, C., Julyan, P., Ranson, M., Broughton, L., Wagstaff, J., Hakannson, L., Groenewegen, G., Bailey, J., Smith, N., Hastings, D., Lawrance, J., Haroon, H., Ward, T., McGown, A. T., Tang, M., Levitt, D., Marreaud, S., Lehmann, F. F., Herold, M., Zwierzina, H., and European Organisation for Research and Treatment of Cancer Biological Therapeutic Development, G. Molecular imaging and biological evaluation of HuMV833 antiVEGF antibody: implications for trial design of antiangiogenic antibodies. Journal of the National Cancer Institute, 94: 14841493, 2002. 177. Eder, J. P., Jr., Supko, J. G., Clark, J. W., Puchalski, T. A., GarciaCarbonero, R., Ryan, D. P., Shulman, L. N., Proper, J., Kirvan, M., Rattner, B., Connors, S., Keogan, M. T., Janicek, M. J., Fogler, W. E., Schnipper, L., Kinchla, N., Sidor, C., Phillips, E., Folkman, J., and Kufe, D. W. Phase I clinical trial of recombinant human endostatin administered as a short intravenous infusion repeated daily.[see comment]. Journal of Clinical Oncology, 20: 37723784, 2002. 178. Jordan, B. F., Runquist, M., Raghunand, N., Baker, A., Williams, R., Kirkpatrick, L., Powis, G., and Gillies, R. J. Dynamic contrastenhanced and diffusion MRI show rapid and dramatic changes in tumor microenvironment in response to inhibition of HIF1alpha using PX478. Neoplasia, 7: 475485, 2005. 179. Tofts, P. S. Modeling tracer kinetics in dynamic GdDTPA MR imaging. [Review] [56 refs]. Journal of Magnetic Resonance Imaging, 7: 91101, 1997. 180. Taylor, J. S., Tofts, P. S., Port, R., Evelhoch, J. L., Knopp, M., Reddick, W. E., Runge, V. M., and Mayr, N. MR imaging of tumor microcirculation: promise for the new millennium. J Magn Reson Imaging, 10: 903907, 1999.

187

181. Yuan, F., Leunig, M., Huang, S. K., Berk, D. A., Papahadjopoulos, D., and Jain, R. K. Microvascular permeability and interstitial penetration of sterically stabilized (stealth) liposomes in a human tumor xenograft. Cancer Res, 54: 3352 3356, 1994. 182. Yuan, F., Leunig, M., Berk, D. A., and Jain, R. K. Microvascular permeability of albumin, vascular surface area, and vascular volume measured in human adenocarcinoma LS174T using dorsal chamber in SCID mice. Microvasc Res, 45: 269289, 1993. 183. Yuan, F., Dellian, M., Fukumura, D., Leunig, M., Berk, D. A., Torchilin, V. P., and Jain, R. K. Vascular permeability in a human tumor xenograft: molecular size dependence and cutoff size. Cancer Res, 55: 37523756, 1995. 184. Seshadri, M., Spernyak, J. A., Maiery, P. G., Cheney, R. T., Mazurchuk, R., Bellnier, D. A., Seshadri, M., Spernyak, J. A., Maiery, P. G., Cheney, R. T., Mazurchuk, R., and Bellnier, D. A. Visualizing the acute effects of vascular targeted therapy in vivo using intravital microscopy and magnetic resonance imaging: correlation with endothelial apoptosis, cytokine induction, and treatment outcome. Neoplasia (New York), 9: 128135, 2007. 185. Menger, M. D., Wolf, B., Hobel, R., Schorlemmer, H. U., and Messmer, K. Microvascular phenomena during pancreatic islet graft rejection. Langenbecks Arch Chir, 376: 214221, 1991. 186. Leunig, M., Yuan, F., Menger, M. D., Boucher, Y., Goetz, A. E., Messmer, K., and Jain, R. K. Angiogenesis, microvascular architecture, microhemodynamics, and interstitial fluid pressure during early growth of human adenocarcinoma LS174T in SCID mice. Cancer Res, 52: 65536560, 1992. 187. Dewhirst, M. W., Ong, E. T., Braun, R. D., Smith, B., Klitzman, B., Evans, S. M., and Wilson, D. Quantification of longitudinal tissue pO2 gradients in window chamber tumours: impact on tumour hypoxia. Br J Cancer, 79: 17171722, 1999. 188. Lehr, H. A., Leunig, M., Menger, M. D., Nolte, D., and Messmer, K. Dorsal skinfold chamber technique for intravital microscopy in nude mice. Am J Pathol, 143: 10551062, 1993. 189. Dafni, H., Gilead, A., Nevo, N., Eilam, R., Harmelin, A., and Neeman, M. Modulation of the pharmacokinetics of macromolecular contrast material by avidin chase: MRI, optical, and inductively coupled plasma mass spectrometry tracking of triply labeled albumin. Magn Reson Med, 50: 904914, 2003. 190. Atalay, M. K., Reeder, S. B., Zerhouni, E. A., and Forder, J. R. Blood oxygenation dependence of T1 and T2 in the isolated, perfused rabbit heart at 4.7T. Magn Reson Med, 34: 623627, 1995. 191. Gatenby, R. A., Gawlinski, E. T., Gmitro, A. F., Kaylor, B., and Gillies, R. J. Acidmediated tumor invasion: a multidisciplinary study. Cancer Res, 66: 5216 5223, 2006. 192. Bremer, C., Tung, C. H., and Weissleder, R. In vivo molecular target assessment of matrix metalloproteinase inhibition. Nat Med, 7: 743748, 2001. 193. Griffin, R. J., Williams, B. W., Bischof, J. C., Olin, M., Johnson, G. L., and Lee, B. W. Use of a fluorescently labeled polycaspase inhibitor for in vivo detection

188

of apoptosis related to vasculartargeting agent arsenic trioxide for cancer therapy. Technol Cancer Res Treat, 6: 651654, 2007. 194. Corot, C., Violas, X., Robert, P., and Port, M. Pharmacokinetics of three gadolinium chelates with different molecular sizes shortly after intravenous injection in rabbits: relevance to MR angiography. Invest Radiol, 35: 213218, 2000. 195. de Lussanet, Q. G., Langereis, S., BeetsTan, R. G., van Genderen, M. H., Griffioen, A. W., van Engelshoven, J. M., and Backes, W. H. Dynamic contrast enhanced MR imaging kinetic parameters and molecular weight of dendritic contrast agents in tumor angiogenesis in mice. Radiology, 235: 6572, 2005. 196. Henderson, E., Sykes, J., Drost, D., Weinmann, H. J., Rutt, B. K., and Lee, T. Y. Simultaneous MRI measurement of blood flow, blood volume, and capillary permeability in mammary tumors using two different contrast agents. J Magn Reson Imaging, 12: 9911003, 2000. 197. Su, M. Y., Muhler, A., Lao, X., and Nalcioglu, O. Tumor characterization with dynamic contrastenhanced MRI using MR contrast agents of various molecular weights. Magn Reson Med, 39: 259269, 1998. 198. Su, M. Y., Wang, Z., Carpenter, P. M., Lao, X., Muhler, A., and Nalcioglu, O. Characterization of NethylNnitrosoureainduced malignant and benign breast tumors in rats by using three MR contrast agents. J Magn Reson Imaging, 9: 177 186, 1999. 199. Turetschek, K., Preda, A., Novikov, V., Brasch, R. C., Weinmann, H. J., Wunderbaldinger, P., and Roberts, T. P. Tumor microvascular changes in antiangiogenic treatment: assessment by magnetic resonance contrast media of different molecular weights. J Magn Reson Imaging, 20: 138144, 2004. 200. Goodno, C. C., Swaisgood, H. E., and Catignani, G. L. A fluorimetric assay for available lysine in proteins. Anal Biochem, 115: 203211, 1981. 201. Marzola, P., Degrassi, A., Calderan, L., Farace, P., Nicolato, E., Crescimanno, C., Sandri, M., Giusti, A., Pesenti, E., Terron, A., Sbarbati, A., and Osculati, F. Early antiangiogenic activity of SU11248 evaluated in vivo by dynamic contrast enhanced magnetic resonance imaging in an experimental model of colon carcinoma. Clin Cancer Res, 11: 58275832, 2005. 202. Murray, L. J., Abrams, T. J., Long, K. R., Ngai, T. J., Olson, L. M., Hong, W., Keast, P. K., Brassard, J. A., O'Farrell, A. M., Cherrington, J. M., and Pryer, N. K. SU11248 inhibits tumor growth and CSF1Rdependent osteolysis in an experimental breast cancer bone metastasis model. Clin Exp Metastasis, 20: 757 766, 2003. 203. Abrams, T. J., Murray, L. J., Pesenti, E., Holway, V. W., Colombo, T., Lee, L. B., Cherrington, J. M., Pryer, N. K., Abrams, T. J., Murray, L. J., Pesenti, E., Holway, V. W., Colombo, T., Lee, L. B., Cherrington, J. M., and Pryer, N. K. Preclinical evaluation of the tyrosine kinase inhibitor SU11248 as a single agent and in combination with "standard of care" therapeutic agents for the treatment of breast cancer. Molecular Cancer Therapeutics, 2: 10111021, 2003.

189

204. George, S. Sunitinib, a multitargeted tyrosine kinase inhibitor, in the management of gastrointestinal stromal tumor. Curr Oncol Rep, 9: 323327, 2007. 205. Motzer, R. J., Michaelson, M. D., Redman, B. G., Hudes, G. R., Wilding, G., Figlin, R. A., Ginsberg, M. S., Kim, S. T., Baum, C. M., DePrimo, S. E., Li, J. Z., Bello, C. L., Theuer, C. P., George, D. J., and Rini, B. I. Activity of SU11248, a multitargeted inhibitor of vascular endothelial growth factor receptor and platelet derived growth factor receptor, in patients with metastatic renal cell carcinoma. J Clin Oncol, 24: 1624, 2006. 206. Ikezoe, T., Nishioka, C., Tasaka, T., Yang, Y., Komatsu, N., Togitani, K., Koeffler, H. P., and Taguchi, H. The antitumor effects of sunitinib (formerly SU11248) against a variety of human hematologic malignancies: enhancement of growth inhibition via inhibition of mammalian target of rapamycin signaling. Mol Cancer Ther, 5: 25222530, 2006.